Dend ritic Cell Response Gene Expression, Compositions of Matters and Methods of

Use Thereof

RELATED APPLICATIONS

[0001] This application claims the benefit of U.S. Provisional Application No.

61/787,378, filed March 15, 2013, which is incorporated herein by reference in its entirety.

STATEMENT AS TO FEDERALLY SPONSORED RESEARCH

[0002J This invention was made with Government support under the following grants: Pioneer Grant DP 1 OD003958-01 awarded by National Institute of Health; Pioneer Grant 5DP1 OD003893-03 awarded by National Institute of Health; Centers of Excellence in Genomics Science Grant 1P50HG006193-01 awarded by National Institute of Health; Grant 1 P01HG005062-01 awarded by National Human Genome Research Institute; and Grant 1 F32HD075541 -01 award by the National Institute of Health. The Government has certain rights in the invention.

FIELD OF THE INVENTION

[0003] This invention relates generally to compositions and methods for identifying the regulatory network that modulates, controls or otherwise influences dendritic cell (DC) response(s), for example, dendritic cell maturation, dendritic cell antiviral response(s) and/or dendritic cell inflammatory response(s), as well compositions and methods for exploiting the regu latory network that modulates, controls or otherwise influences dendritic cell response(s) in a variety of therapeutic and/or diagnostic indications.

BACKGROUND OF THE INVENTION

[0004] Despite their importance, the molecular circuits that control dendritic cell responses, including antiviral responses, inflammatory responses, maturation, recruitment of T cells and B cells, remain largely unknown or unrefined. Recent studies that reconstructed regulatory networks in dendritic cells have focused on measurements across cell populations that can fail to detect signals across the entire population and/or can fail to distingu ish between signal(s) that are expressed only in certain subsets of cells. Accordingly, there exists a need for a better understanding of the network that modulates, controls, or otherwise influences dendritic cell response and means for exploiting this network in a variety of therapeutic and diagnostic methods.

SUMMARY OF THE INVENTION

[0005] The invention provides compositions and methods for modulating one or more dendritic cell responses. As used herein, the term "modulating" includes up-regulation of, or otherwise increasing, the expression of one or more genes; down-regulation of, or otherwise decreasing, the expression of one or more genes; inhibiting or otherwise decreasing the expression, activity and/or function of one or more gene products;

neutralizing or otherwise inactivating the expression, activity and/or function of one or more gene products; and/or enhancing or otherwise increasing the expression, activity and/or function of one or more gene products.

[0006] As used herein, the term "modulating a response of dendritic cells" includes the modulation of any of a variety of dendritic cell functions and/or activities, including by way of non-limiting example, controlling or otherwise influencing the networks that regulate dendritic cell maturation; controlling or otherwise influencing the networks that regulate an immune response of a dendritic cell; controlling or otherwise influencing the networks that regulate an antiviral immune response of a dendritic cell, for example, an antiviral immune response of a dendritic cell including a core antiviral response and/or a secondary antiviral response; controlling or otherwise influencing the networks that regulate an inflammatory immune response of a dendritic cell, for example, an induced

inflammatory response and/or a sharped peak inflammatory response; controlling or otherwise influencing the networks that regulate a Toll-like receptor (TLR) response of dendritic cells; controlling or otherwise influencing the networks that regulate T cell and B cell recruitment; controlling or otherwise influencing the networks that regulate DC promotion of T _H l -cell response(s); controlling or otherwise influencing the networks that regulate DC induction of TH2-CC11 response(s); controlling or otherwise influencing the networks that regulate DC induction, impact or other effect on any cell that is downstream of the D; controlling or otherwise influencing the networks that regulate DC induction of T cells including regulatory T cells (Tregs), Thl 7 cells, memory T cells and other T cells; controlling or otherwise influencing the networks that regulate a shift in a DC phenotype, for example, between a mature and immature phenotype and/or between subsets of DCs; manipulating or otherwise influencing at least one function or biological activity of a dendritic cell; manipulating or otherwise influencing dendritic cell control of pathogen- drive T cell polarization; and/or manipulating or otherwise influence the production of cytokines, chemokines and other molecules secreted by the DC.

[0007] The invention provides modulating agents that modulate one or more dendritic cell response(s). Suitable modulating agents include an antibody, a soluble polypeptide, a polypeptide agent, a peptide agent, a nucleic acid agent, a nucleic ac id ligand, or a small molecule agent.

[0008] The invention provides a series of gene signatures, including a "Core

Antiviral" gene signature, a "Secondary Antiviral" gene signature, a "Maturation" gene signature, an "Induced Inflammatory" gene signature, and a "Sharp Peaked Inflammatory" gene signature. These signatures were identified by clustering gene expression values across single cells, for example, coherent groups of single cells. In some embodiments, these signatures significantly refine and improve upon previously identified signatures. In some embodiments, these signatures produce signals that are absent or cannot be reliably detected in cell population measurements.

[0009] The "Core Antiviral" gene signature is induced in the earliest of the responding dendritic cells. The "Maturation" gene signature looks similar to the "Induced Inflammatory" gene signature at a population level, but using single cell analysis, it was established that the "Maturation" gene signature is expressed in only a subset of cells. The "Maturation" gene signature is responsible for allowing dendritic cells to recruit T cells and B cells, thereby bridging the gap between the innate and adaptive immunity system.

[00010] These genes are targets for use in a number of indications, for example, for treating and/or diagnosis of an immune response, for monitoring an immune response, e.g., inflammation, in transplant and other therapeutic indications and/or for vaccine

development.

[00011] In some embodiments, the one or more signature genes are selected from those listed in Tables 1 , 1A, 2, 2A, 3, 3A, 4, 4A, 5 and/or 5A (i.e., Tables 1 , 1A, 2, 2A, 3, 3 A, 4, 4A, 5 and/or 5A).

[00012] A desired target gene or combination of target genes is selected, and after determining whether the selected target gene(s) is overexpressed or under-expressed during a dendritic cell response, a suitable antagonist or agonist is used depending on the desired maturation and/or function outcome. Suitable antagonists and/or agonists include an antibody, a soluble polypeptide, a polypeptide agent, a peptide agent, a nucleic acid agent, a nucleic acid ligand, or a small molecule agent. [00013] The modulating agents are used to modulate the expression of one or more target genes or one or more products of one or more target genes that have been identified as genes responsive to dendritic cell-related perturbations. These target genes are identified, for example, by contacting a dendritic cell with a modulating agent and monitoring the effect, if any, on the expression of one or more signature genes or one or more products of one or more signature genes. In some embodiments, the one or more signature genes are selected from those listed in Tables 1 -5A. The modulating agent can act directly on the expression of one or more target genes or one or more products of one or more target genes and/or the modulating agent can act indirectly on the expression of one or more target genes or one or more products of one or more target genes by modu lating the expression, activity and/or function of a gene or a product of a gene that is known to be associated with the target gene(s).

[00014] In some embodiments, the target gene is tumor necrosis factor receptor (TNFR). In some embodiments, the modulating agent alters the expression, activity and/or function of TNFR. In some embodiments, the modulating agent alters the expression, activity and/or function of a gene that is associated with TNFR, such as, by way of non- limiting example, a gene from those shown in Table 6.

[00015] In some embodiments, the target gene is a Toll/interleukin- l receptor (TIR) domain-containing adapter protein (T1RAP). In some embodiments, the modulating agent alters the expression, activity and/or function of TIRAP. In some embodiments, the modulating agent alters the expression, activity and/or function of a gene that is associated with TIRAP, such as, by way of non-limiting example, a gene from those shown in Table 7.

[00016] In some embodiments, the target gene is Statl . In some embodiments, the modulating agent alters the expression, activity and/or function of Statl . In some embodiments, the modulating agent alters the expression, activity and/or function of a gene that is associated with Stat l , such as, by way of non-limiting example, a gene from those shown in Table 8.

[00017] In some embod iments, the target gene is interferon production regulator (1FNR). In some embodiments, the modulating agent alters the expression, activity and/or function of IFNR. In some embodiments, the modulating agent alters the expression, activity and/or function of a gene that is associated with IFNR, such as, by way of non- limiting example, a gene from those shown in Table 9. [00018] In some embodiments, the target gene is one or more genes from those listed below in Table 10, Table 1 1 or Table 12. In some embodiments, the modulating agent alters the expression, activity and/or function of the target gene(s).

[00019] In some embodiments, the invention provides a method of identifying genes or genetic elements associated with a dendritic cell response comprising: a) contacting a dendritic cell with an inhibitor of a dendritic cell response or an agent that enhances a dendritic cell response; and b) identifying a gene or genetic element whose expression is modulated by step (a). In some embod iments, the method also comprises c) perturbing expression of the gene or genetic element identified in step b) in a dendritic cell that has been in contact with an inhibitor of the dendritic cell response or an agent that the dendritic cell response; and d) identifying a gene whose expression is modulated by step c). In some embodiments, the antagonist and/or agonist is an antibody, a soluble polypeptide, a polypeptide antagonist, a peptide antagonist, a nucleic acid antagonist, a nucleic acid ligand, or a small molecule antagonist.

[00020] In some embodiments, the invention provides a method of modulating one or more dendritic cel l response(s) comprising contacting a dendritic cell with an agent that modulates expression, activity and/or function of one or more genes or one or more products of one or more genes selected from those listed in Table 1 or Table 1 A. In some embodiments, the invention provides a method of modulating one or more dendritic cell response(s) comprising contacting a dendritic cell with an agent that modulates expression, activity and/or function of one or more genes or one or more products of one or more genes selected from those listed in Table 2 or Table 2A. In some embodiments, the invention provides a method of modulating one or more dendritic cell response(s) comprising contacting a dendritic cell with an agent that modulates expression, activity and/or function of one or more genes or one or more products of one or more genes selected from those listed in Table 3 or Table 3A. In some embodiments, the invention provides a method of modulating one or more dendritic cell response(s) comprising contacting a dendritic cell with an agent that modulates expression, activity and/or function of one or more genes or one or more products of one or more genes selected from those listed in Table 4 or Table 4A. In some embodiments, the invention provides a method of modulating one or more dendritic cell response(s) comprising contacting a dendritic cell with an agent that modulates expression, activity and/or function of one or more genes or one or more products of one or more genes selected from those listed in Table 5 or Table 5A. (00U21J In some embodiments, the invention provides a method of diagnosing an immune response in a subject, comprising detecting a level of expression, activity and/or function of one or more signature genes or one or more products of one or more signature genes selected from those listed in Tables 1 , 1 A, 2, 2 A, 3, 3 A, 4, 4A, 5 and/or 5A and comparing the detected level to a control of level of signature gene or gene product expression, activity and/or function, wherein a difference between the detected level and the control level indicates that the presence of an immune response in the subject. In some embodiments, the immune response is an autoimmune response. In some embodiments, the immune response is an inflammatory response, including inflammatory response(s) associated with an autoimmune response and/or inflammatory response(s) associated with an infectious disease or other pathogen-based disorder.

[00022] In some embodiments, the invention provides a method of monitoring an immune response in a subject, compr ising detecting a level of expression, activity and/or function of one or more signature genes or one or more products of one or more signature genes, e.g. , one or more signature genes selected from those listed in Tables 1 , 1A, 2, 2A, 3, 3A, 4, 4A, 5 and/or 5A at a first time point, detecting a level of expression, activity and/or function of one or more signature genes or one or more products of one or more signature genes, e.g., one or more signature genes selected from those listed in Tables 1 , I A, 2, 2A, 3, 3A, 4, 4A, 5 and/or 5A at a second time point, and comparing the first detected level of expression, activity and/or function with the second detected level of expression, activity and/or function, wherein a change between the first and second detected levels indicates a change in the immune response in the subject. In some embodiments, the immune response is an autoimmune response. In some embodiments, the immune response is an inflammatory response.

[00023] In some embodiments, the invention provides a method of diagnosing an aberrant dendritic cell response in a subject, comprising detecting a level of expression, activity and/or function of one or more signature genes or one or more products of one or more signature genes selected from those listed in Tables 1 , 1A, 2, 2A, 3, 3A, 4, 4A, 5 and/or 5A and comparing the detected level to a control of level of signature gene or gene product expression, activity and/or function, wherein a difference between the detected level and the control level indicates that the presence of an aberrant dendritic cell response in the subject. In some embodiments, the aberrant dendritic cell response is an autoimmune response. In some embodiments, the aberrant dendritic cell response is an inflammatory response, includ ing inflammatory response(s) associated with an autoimmune response and/or inflammatory response(s) associated with an infectious disease or other pathogen- based disorder. In some embodiments, the aberrant dendritic cell response is an altered ability of the dendritic cell to recruit T cells and B cells. In some embodiments, the aberrant dendritic cell response is the absence of a response. In some embodiments, the aberrant dendritic cell response is a reduction in a dendritic cell response. In some embodiments, the aberrant dendritic cell response is an enhancement in a dendritic cell response.

[00024] In some embodiments, the invention provides a method of monitoring an aberrant dendritic cell response in a subject, comprising detecting a level of expression, activity and/or function of one or more signature genes or one or more products of one or more signature genes, e.g., one or more signature genes selected from those listed in ables 1 , 1A, 2, 2A, 3, 3 A, 4, 4A, 5 and/or 5 A at a first time point, detecting a level of expression, activity and/or function of one or more signature genes or one or more products of one or more signature genes, e.g., one or more signature genes selected from those listed in Tables 1 , 1A, 2, 2A, 3, 3A, 4, 4A, 5 and/or 5A at a second time point, and comparing the first detected level of expression, activity and/or function with the second detected level of expression, activity and/or function, wherein a change between the first and second detected levels indicates a change in the dendritic cell response in the subject. In some embodiments, the dendritic cell response is an autoimmune response. In some embodiments, the dendritic cell response is an inflammatory response. . In some embodiments, the dendritic cell response is the ability of the dendritic cell to recruit T cells and B cells.

[00025] Suitable modulating agent(s) for use in any of the compositions and methods provided herein include an antibody, a soluble polypeptide, a polypeptide agent, a peptide agent, a nucleic acid agent, a nucleic acid ligand, or a small molecule agent.

[00026] In some embodiments, the modulating agent is used to modulate the expression of one or more genes from the "Core Antiviral" gene signature, e.g. , one or more genes from those listed in Tables 1 and 1 A. These modulating agents are referred to herein as "core antiviral modulating agent(s)."

[00027] For example, in some embodiments the core antiviral modulating agent is a kinase, such as, by way of non-limiting example, a kinase selected from the group consisting of: MAPKl , EIF2AK2, TBK l , PLK4, IKBKE, PLK2, MAP3K7, CHUK, JAKl , CRKL, MKNK2, TYK2, RPS6KB2, IKBKB, MKNK1 , NEK7, PIK3R2, 1KB KG, RIPK2, MAP2K6, MET, RPS6KJB 1 , MARK2, DGKA, and BUB I B. [00028J For example, in some embodiments, the core antiviral modulating agent is a transmembrane receptor, a mammalian endogenous chemical drug, a chemical drug, e.g., a chemical kinase inhibitor drug or other chemical drug such as a chemical reagent, toxicant or other chemical drug, a bio logic drug or any combination thereof. Suitable core antiviral modulating agents include any of those described herein.

[00029J n some embodiments, the modulating agent is used to modulate the expression of one or more genes from the "Secondary Antiviral" gene signature, e.g., one or more genes from those listed in Tables 2 and 2A. These modulating agents are referred to herein as "second antiviral modulating agents."

[00030] For example, in some embodiments the secondary antiviral modulating agent is a kinase, a transmembrane receptor, a non-mammalian endogenous chemical drug, a chemical drug, e.g., a chemical kinase inhibitor drug or another chemical drug such as a chemical reagent, toxicant or other chemical drug, or any combination thereof. Suitable secondary antiviral modu lating agents include any of those described herein.

[00031J In some embodiments, the modulating agent is used to modulate the expression of one or more genes from the "Maturation" gene signature, e.g., one or more genes from those listed in Tables 3 and 3A. These modulating agents are referred to herein as "maturation modulating agents."

[00032] For example, in some embodiments the maturation modulating agent is a kinase, a transmembrane receptor, a mammalian endogenous chemical drug, a non- mammalian endogenous chemical drug, a chemical drug, e.g., a chemical kinase inhibitor drug or another chemical drug such as a chemical reagent, chemical toxicant or other chemical drug, a biologic drug, or any combination thereof. Suitable maturation modulating agents include any of those described herein.

[00033] In some embodiments, the modulating agent is used to modulate the expression of one or more genes from the "Peaked Inflammatory" gene signature, e.g. , one or more genes from those listed in Tables 4 and 4A. These modulating agents are referred to herein as "peaked inflammatory modulating agents."

[00034] For example, in some embodiments the peaked inflammatory modulating agent is a kinase, such as, by way of non-limiting example, a kinase, a transmembrane receptor, a mammalian endogenous chemical drug, a non-mammalian endogenous chemical drug, a chemical drug, e.g., a chemical kinase inhibitor or another chemical drug such as a chemical reagent, toxicant or other chemical drug, a biologic drug, or other modulating agent, or any combination thereof. Suitable peaked inflammatory modulating agents include any of those described herein.

[00035] In some embodiments, the modulating agent is used to modulate the expression of one or more genes from the "Induced Inflammatory" gene signature, e.g. , one or more genes from those listed in Tables 5 and 5A. These modulating agents are referred to herein as "induced inflammatory modulating agents."

[00036] For example, in some embodiments the induced inflammatory modulating agent is a kinase, a transmembrane receptor, a mammalian endogenous chemical drug, is a non-mammalian endogenous chemical drug, a chemical drug, such as a chemical kinase inhibitor or another chemical drug, such as, by way of non-limiting example, a chemical reagent, chemical toxicant or other chemical drug, a biologic drug, or any combination thereof. Suitable peaked inflammatory modulating agents include those described herein.

[00037] One skilled in the art will appreciate that the modulating agents have a variety of uses. For example, the modulating agents are used as therapeutic agents as described herein. The modulating agents can be used as reagents in screening assays, diagnostic kits or as diagnostic tools, or these modulating agents can be used in competition assays to generate therapeutic reagents.

BRIEF DESCRIPTION OF THE DRAWINGS

[00038] Figures 1 A- 1 H are a series of graphs and illustrations depicting that single cell NA-Seq of LPS-stimulated BMDCs revealed extensive transcriptome heterogeneity.

A color version of these figures can be found in Shalek et al., "Single-cell transcriptomics reveals bimodality in expression and splicing in immune cells." Nature 498(7453):236-40

(2013); doi: 10.1038/natureI 2172. Figures l a-lc depict correlations oftranscript expression levels (x & y-axes: log-scale TPM+1 ) between two 10,000 cell population replicates (Fig. l a), two single cells (Fig. l b), and the 'average' single cell and a population measurement

(Fig. l c). The Pearson correlation coefficient (r) is marked in the upper left corner. Figures

I d, l e, depict example transcripts. Shown are the RNA-Seq read densities in each single cell

("1 " on the y axis) and the three population replicates ("10,000" on the y-axis) for three non-variable genes (Fig. I d) and four variable ones (Fig. l e). Figures l f- l h depict RNA-

FISH of representative transcripts. Shown are micrographs (log filtered, (Fig. I f, Fig. l g)) and distributions of expression levels (Fig. l h) from RNA-FISH staining for the lower variation gene 116 (top panel, n=3193 cells) and the higher variation gene Cxcl l (bottom panel, n=31 93 cells). Cell boundaries are represented by light grey outlines. [00039] Figures 2A-2C are a series of graphs and illustrations depicting bimodal variation in expression levels across single cells. A color version of these figures can be found in Shalek et al., "Single-cell transcriptomics reveals b nodality in expression and splicing in immune cells." Nature 498(7453):236-40 (2013). Figure 2a depicts inter-cell variation at a broad range of expression levels. Shown is the relationship between the single cell expression average (μ, X axis) and single cell variability (standard deviation, σ, Y axis). B lue dashed (i.e., upper) line indicates the theoretical maximum standard deviation for an average expression level (Example 1); Grey dashed (i.e., lower) line denotes the constant Fano factor (σ/μ = 0.25). Immune response and housekeeping genes are marked in magenta and green, respectively; light blue shaded region represents single cell average TPM < 250. Notably, even at high average expression levels, BMDC response elements show substantial variability (left), while hESCs (Ramskold, D. et al. Full-length mRNA-Seq from single-cell levels of RNA and individual circulating tumor cells. Nature Biotechnology 30, 777-782, doi:papers2://publication/doi/10.1038/nbt.2282 (2012)) (right) do not. Figure 2b depicts inter-cell variation of the 522 most highly expressed genes. For each gene (rows, sorted by Fano factor from low (top) to high (bottom)) and each expression level bin (columns), shown is the number of cells (strong yellow: 1 8 cells; black: 0 cells) in which the gene is expressed at the bin defined level. The genes are chosen based on their average single cell expression level (TPM > 250, white area in (Fig. 2a)). Grey dashed line denotes the constant Fano factor (0.25) highlighted in (Fig. 2a). Figure 2c depicts average expression probability density distributions for the 281 low-variability genes (top) and the 241 highly variable genes (bottom).

[00040] Figures 3A-3D are a series of graphs and illustrations depicting variation in isoform usage between single cells. A color version of these figures can be found in Shalek et al., "Single-cell transcriptomics reveals bimodality in expression and splicing in immune cells." Nature 498(7453):236-40 (2013). Figure 3a depicts examples of genes with significant splicing differences between individual cells. Shown are the RNA-Seq read densities for each of the 1 8 single cells (1 , blue) and 3 population replicates ( 10K, grey) for two illustrative loci, each with two different isoforms (bottom). Figure 3b shows the distributions of exon inc lusion (Percent Spliced In (PSI) scores, X axis) for alternatively spliced exons of highly expressed genes (single cell TPM > 250) in individual cells (blue histogram, top) and in the populations (grey histogram, bottom). Single cells exhibit a strong bias towards expression of one particular isoform. Figure 3c depicts RNA-FISH validation of splicing variation in Irf7. Left: RNA-Seq read densities (only cells where the transcript is expressed are shown). Color boxes mark exons analyzed by RNA-F1SH. Right: RNA-FISH images from simu ltaneous hybridization with probes for two constitutive ('Constitutive' or 'Con') regions of the transcript (constitutive region A: cyan (C);

constitutive region B: magenta (M)) and one alternatively spliced exon ('Specific' : orange (O)). White arrows highlight two cells with similarly high expression levels for Irf7, but opposite preferences for the alternatively spliced exon. Histograms: The two constitutive regions (right top and right bottom panels) are detected at similar levels (bottom histogram, deviation from 0.5 is as expected due to probe design,), whereas the alternative exon (middle right panel) shows a bias towards inclusion or exclusion in individual cells (top histogram). Figure 3d demonstrates that similar results were obtained for alternative regu lation of mutually exclusive last exons for the gene Acpp.

[00041] Figures 4A-4F are a series of graphs and illu strations that depict how analysis of co-variation in single cell mRNA expression levels revealed distinct maturity states and an antiviral cell circuit. A color version of these figures can be found in Shalek et al., "Single-cell transcriptomics reveals bimodality in expression and splicing in immune cells." Nature 498(7453):236-40 (2013). Figure 4a depicts PCA of 632 LPS-induced genes. Shown are the contributions of each cell (points) to the first two principle components. PCI (X axis) discriminates 3 'semi-mature' cells (square) from 15 'maturing' cells (triangles). Light grey triangles denote the most mature cells. Figure 4b depicts clustered correlation matrix of induced genes. Left: Shown is the Pearson correlation coefficients (r, purple: negative correlation; yellow: positive correlation) between single-cell expression profiles of every pair of 632 LPS-induced genes (rows, columns). The three highlighted clusters are noted on the left along with a few representative loci. Right: The projection score (green: high; blue: low) for each gene (row) onto PC I (left) and PC2 (right). PCI differentiates semi-mature from maturing BMDCs; PC2 maps to a cluster of antiviral genes. Figure 4c depicts confirmation of correlations by RNA-FISH. Shown are the relationships between two pairs of genes (Irl7-Stat2, Irf7-Ifit l ) based on RNA-FISI I when simu ltaneously staining for the members of each pair. The square of the Pearson correlation coefficient {r ) and number of measured cel ls are denoted in the upper left corner. Figures 4d-4f depicting how Irf7 propagates variability in an interferon feedback circuit. Shown are expression levels for each of eight genes from the antiviral cluster ('antiviral' rows), along with eight non- variable immune response genes ('non-variable response' rows), in each single immature B DC (columns), measured using single-cell qRT-PCR in wild type (WT) (n = 36) (Fig. 4d), Irf7 -/- (n = 47) (Fig. 4e), and Ifnr -/- (n = 18) (Fig. 4f) BMDCs stimulated with LPS for 4h.

[00042] Figure 5 is a graph depicting global correlations in mRNA expression between single LPS stimulated BMDCs. Shown are the Pearson correlation coefficients between global expression profiles of each of 1 8 individual cells, the single cell average, and three populations of 10,000 cells each (rows, columns). All correlations were computed on log-scale expression profiles. Single cells (S) 12, 13, and 16 are Semi-Mature, while 9 and 16 are the most mature, correspond to light grey triangles in Fig. 4a.

[00043] Figure 6 is a series of graphs depicting agreement between single-cell RNA- Seq and RNA-FISH for 25 different transcripts. Shown are the distributions of gene expression levels for each of 25 transcripts in single-cell RNA-Seq of 18 cells (left, blue) and in single-cell RNA-FISH of, on average, 1 00 cells (right, red).

[00044] Figure 7 is a graph depicting robust LPS response across all cells. Shown are tracks of RNA-Seq reads from the Integrative Genomics Viewer for the levels of key response genes (columns, gene name at bottom) in each single cell (blue) and the population average (grey). The genes include key chemokines and chemokine receptors (Ccl3, Ccl4, Ccrl2), cytokines (Cxcl2), and other important components of the LPS response (Tank, Cflar).

[00045] Figure 8 is a series of graphs depicting variation in gene expression from single-cell RNA-Seq in other cell types. Shown is the relationship between the single cell expression average (μ, X axis) and single cell variability (standard deviation, σ, Y axis) in mouse embryonic stem cells (left) and mouse embryonic fibroblasts (right). These figures show a re-analysis of previously published single cell RNA-seq data (Hashimshony, T., Wagner, F., Sher, N. & Yanai, I . CEL-Seq: Single-Cell RNA-Seq by Multiplexed Linear Amplification. Cell Reports, doi: 10.10I 6/j.celrep.2012.08.003). Housekeeping genes are green. In both cases substantially less variability in single-cell gene expression was found compared to LPS-stimulated BMDCs (Fig. 2a).

[00046] Figures 9A-9D are a series of graphs and illustrations depicting

quantification of unique mRNA molecules in three single cells. Figure 9a depicts a modified protocol. The SMARTer II A oligo was modified, introducing a random four nucleotide barcode onto each mRNA molecule during reverse transcription. Shown is the structure of modified oligo (barcode is represented by NNNN). This barcode is retained through PCR amplification and library preparation. Figure 9b depicts an 1GV scrcenshot showing read densities at one locus for the three barcoded single-cell cDNA libraries (blue) as wel l as the three 10,000 cell replicate experiments (grey). Two single cells express exclusively one of two isoforms. Figure 9c depicts detailed examination of reads mapping to 5 ' end of transcript. The 8 1 reads represent 23 unique barcodes, affirming that the observed splicing result is not simply due to stochastic amplification of one or a few molecules. Figure 9d depicts the relationship between single-cell TPM (X axis, log scale) and uniquely identified barcodes (Y axis, log scale) for the three barcoded single-cell libraries. Only genes represented by at least one unique barcode are plotted. Light blue shaded area represents single cell TPM < 250, the threshold used throughout the study. The two alternate quantifications of single-cell gene expression are well correlated overall (0.82 < R < 0.86) and exhibit a tightly linear relationship for highly expressed genes (TPM > 250).

[00047J Figures l OA-l OC are a series of graphs and a table depicting variation in isoform expression between single cells based on the 3 barcoded single-cell libraries. Figure 10a depicts IGV screenshots showing read densities for 6 alternatively spl iced genes. For each gene, the alternatively spliced exon is boxed in orange. Figure 10b is a table showing the number of unique molecular barcodes counted for each transcript shown in Fig. 10a. Figure 10c depicts the distributions of exon inclusion (PS1 scores, X axis) for alternatively spliced exons in genes represented by at least 15 barcodes in single cells (blue histogram, top) and in the populations (grey histogram, bottom). Results are highly similar to the splicing analysis of highly expressed genes across the 18 cells (single-cell TPM>250; Fig. 3). Single cells exhibit a strong skew towards one isoform or the other.

[00048] Figure 1 1 is a graph depicting RNA-F1SH validation of splicing variation in Irf7 in single cells. Shown is the distribution across cells of the ratio of Irf7 transcripts displaying the isoform specific lr!7 probe (Orange, Fig. 3c) relative to the shorter constitutive probe (Magenta, Fig. 3c). The distribution is similarly bimodal to that obtained when calculating the ratio of the specific probe to the longer constitutive probe (Fig. 3c).

[00049] Figure 12 is a graph depicting IGV screenshots exhibiting the separation between semi-mature and maturing cells. These genes have either very high (positive) or low (negative) projection scores for PCI . A black vertical bar on the right highlights two cells that express both mature and maturing markers, suggesting that they are, in fact, the most mature of the maturing cells.

[00050] Figures 13A-1 3E are a series of graphs depicting confirmation of co- variation patterns by RNA-FISH. Shown are the relationships in expression levels (log (Count+1 )) for pairs of transcripts simultaneously measured by RNA-FISH. Figure 13a depicts that expression levels for Ccr7 (expressed more in maturing cells) and Ill b

(expressed more in semi-mature cells) did not correlate strongly (Pearson r ² = 0.12, n = 812). Figure 13b depicts that expression levels for Stat4 (expressed more in semi-mature cells) and Serpinb9 (expressed more in semi-mature cells) correlated more strongly (Pearson r ² = 0.28, n = 573). Figure 13c depicts that expression leve ls for Cxcll O and Tnf (both expressed more in maturing cells) correlated mildly (Pearson r ² = 0.18, n = 51 1 ). Figure 13d depicts that Ccl22 and Ir 8 (both expressed in semi-mature cells) showed moderate correlation (Pearson r ² = 0.26, n = 1 1 10). Figure 13e depicts that Stat l (antiviral, specific to neither) and Cxcl l (inflammatory, specific to neither) correlated very weakly (Pearson r ² = 0.07, n = 631 ).

[00051] Figure 14 is a graph depicting that individual LPS-stimulated BMDCs cluster into two distinct populations by single-cell qRT-PCR. Shown are the normalized expression levels (red: high; blue: low, scale on top) from single-cell qRT-PCR (Fluidigm) for 50 genes (rows) in each of 46 individual cells (columns). The cells were clustered by hierarchical agglomerative clustering based on their expression profiles (dendrogram, top) and form two main clusters (semi-mature and maturing, bottom).

[00052] Figure 15 is a graph depicting differences in expression levels of key markers between subpopulations that are positive and negative for different semi-mature and maturing cell surface markers. Shown are the differential expression levels (Y axis) of each of 10 marker genes (bars, color legend, right) measured by qRT-PCR between cells positively and negative sorted for each marker (X axis). The markers were chosen based on their ability to discriminate the 'maturing' (Red) and 'semi-mature' (Blue) subpopulations in the RNA-Seq data.

[00053] Figures 16A and 16B are a series of graphs depicting single-cell qPCR expression profiling for a signature of 1 3 genes along an LPS response time course. Figure 16a depicts the expression levels of each gene (row) in each cell (column) in unstimulated BMDCs and at 2h, 4h, and 6h post-LPS stimulation. The gene signature consists of nine antiviral cluster genes, two uniformly induced genes, and two housekeeping controls. Figure 16b depicts the percentage of cells that express each gene (rows) at each time point (column). A cell was scored as positive for a gene if the gene's expression was higher than a Ct of 23 on the Fluidigm BioMark. While some immune response genes, Cxcl l O and Clec4e, were uniformly induced in all cells and persisted across time points, the percentage of cells expressing the antiviral cluster genes increased in a time-dependent manner.

[00054J Figures 17A and 17B are a series of photographs and graphs depicting RNA- FISH and Immunofluorescence co-staining. Figure 17a depicts an example of a co-staining image for Stat l protein (green), Stat l mRNA (magenta), and Ifit l mRNA (white). Figure 17b depicts the distributions of the levels of Ifitl mRNA (black) and Stat l (red), pStatl (grey), and Stat2 (green) proteins (total fluorescence level, left histogram; average fluorescence level, middle; and percent nuclear localization, right) after exposure to LPS for 0 (top), 2 (middle) or 4 (bottom) hours. While overall protein levels increased in all cases throughout the time course, substantial variation in the induction of Statl , pStatl , and Stat2 was found. Stat l levels rose gradually while pStatl 's shifts were most pronounced early. Stat2, meanwhile, showed strong nuclear localization by 2h, followed by strong induction from 2 to 4h. By 4hr, protein levels were more homogeneous and nuclear translocation was less pronounced.

[00055] Figures 18A and 18B are a series of graphs depicting correlation between Stat protein and Ifit l mRNA expression. Figure 18a depicts representative scatter plots showing the correlation between Stat proteins (Y axis) and Ifit mRNA levels (X axis) after a 4h LPS stimulation. Top row: Statl , middle row: pStatl ; bottom row: Stat 2. Left column: total protein fluorescence; middle column: average protein fluorescence; right column: percent of nuclear protein. Figure 18b depicts heatmaps showing the correlation (r ² ; blue = 0; red = 1 ) between different measured parameters after exposure to LPS for 0 (top), 2 (middle), or 4 hours (bottom).

[00056] Figure 19 is an illustration depicting a simple model for the identified antiviral circuit. X's represent points of perturbation. Ifh feedback drives expression of IrI7 and Stat2. Variability in the expression of Irf7 propagates to variability in the expression of antiviral genes, such as Ifitl . Stat2 is implicated as well, though its relation to Irf7 cannot be established by the current experiments.

[00057] Figure 20 is a graph depicting splicing patterns for 'poison' cassette exons of the splicing factors Srsf3 and Srsf7. Shown are the RNA-Seq read densities in each individual cell (T, blue) and the population average (' 10,000', grey) for two genes encoding the splicing factors Srsf3 and Srsf7, each of which is known to have an alternatively spliced poison cassette exon (dashed box). The known annotated iso forms for each gene is shown at the bottom. One cell, S I 3, highlighted in orange at the top, expressed only the Srsfi and Srsi7 iso forms that contain the 'poisonous' exons. For each gene, 1 1 cells exclusively expressed the alternative isoform.

[00058] Figure 21 is a graph depicting expression variation in long non-coding (lnc) RNAs. Shown are the RNA-Seq read densities in each individual cell (Ί ', blue) and the population average (Ί 0,000', grey) for three previously annotated IncRNA genes. A IncRNA relatively highly expressed at the population level (Gas5, left), is bimodally expressed at the single-cell level. Two IncRNAs lowly expressed or undetectable at the population (Gm8773, 2810025M 15Rik) are in fact significantly expressed in some individual cells.

[00059] Figure 22 is a series of graphs depicting quality control for 3' bias. Shown are plots of normalized RNA-Seq coverage at each normalized transcript position from 5 ' (left) to 3 ' (right) for 6 single cells (top two rows) and all three 10,000 populations (bottom row). Both the single cells and the populations show little 3' bias.

[00060] Figure 23 is a series of photographs depicting RNA-FISH of the immune- response genes Cxcl l (Top) and Cxcl lO (Bottom) in the absence of LPS stimulation (left) and after 4h of an LPS stimulus (right). Cxcl l O and Cxcl l, although expressed at negligible levels prior to stimulation, are strongly induced by LPS.

[00061] Figures 24A-24E are a series of graphs and illustrations depicting microfluidic enabled single-cell RNA-Seq of bone marrow derived dendritic cells (BMDCs) stimulated with pathogen components. Figure 24a depicts a scanning electron micrograph of a BMDC (scale bar: 25 μιη). Figure 24b depicts a simplified schematic of Toll-Like Receptor (TLR) network for sensing of PAM3CSK (PAM, from gram-positive bacteria) by TLR2, Lipopolysaccharide (LPS. from gram-negative bacteria) by TLR4, and

polyinosinic:polycytidylic acid (PIC, poly(LC), a synthetic mimic of viral RNA) by TLR3. Figure 24c depicts microfluidic capture of a single BMDC (top, cell circled in purple) on the CI chip (CAD drawing, bottom). Figure 24d depicts principal component (PC) analysis, computed over samples from all three stimuli and time points together, for the LPS- stimulated cells (left) and the distributions of LPS -stimulated cellular scores for the first three PCs (right). Figure 24e depicts time course expression profiles for induced genes (rows) in BMDCs at 0, 1,2,4, and 6h post stimulation with PAM (green), LPS (black), and PIC (magenta) within BMDC populations (left columns) and ind ividual BMDCs (right columns). At the far right are gene projection scores onto the first 3 principle components (PCs) (PCI , PC2, and PC3, columns); on the bottom are contributions of each cell (columns)to the first three PCs (PC I , PC2, and PC3, rows).

[00062] Figures 25A-25D is a series of graphs depicting time dependent behaviors of single cells. Figure 25a depicts example single-cell expression distributions seen for three genes (one from each of the three clusters in Fig 24e), at each time po int (marked on top) after stimulation with PAM (top, green), LPS (middle, black), and PIC (bottom, magenta). Distributions are scaled to have the same maximum height. Individual cells are plotted as bars underneath each distribution. Figures 25b-d depict, for each of the three modules (labeled, top), wave plots of all of its constituent genes at 2h (left) and 6h (right) in BMDCs stimulated with LPS (top), PIC (for the "core" antiviral cluster Id, (Fig. 25b) or PAM (fot the "peaked" inflammatory cluster (Fig. 25c) and "sustained" inflammatory (Fig. 25d) clusters. X axis: expression level, ln(TPM+l ); Y axis: genes; Z axis: single-cell expression density. Genes are ordered from lowest to highest average expression value at the 2h ("peaked" inflammatory) or 6h ("core" antiviral, "sustained" inflammatory) LPS time po int.

[00063] Figures 26A-26L are a series of graphs depicting dynamic changes in variation during stimulation. Figure 25a presents a schematic rendering of the three parameters used to describe single-cell expression distributions, from left to right: μ, the mean RNA abundance levels for cells with detectable level of expression; σ, the dispersion in expression for cells with detectable expression; and a, the fraction of all cells with detectable expression (at ln(TPM+l) > 1 ). Figure 25b depicts examples of fit (grey) for measured TNF expression distributions (black) at different time points post LPS stimulation. Figure 25c depicts changes in the values of μ, σ ² , and a (Y axes, left to right) estimated for TNF at each time point (X axis). Units for μ and σ ² are ln(TPM+l). Figure 25d is a maximum likelihood estimate a. Shown are the expression distributions (black, left) of TNF at different time points following LPS stimulation, and the matching likelihood function (dotted blue line) used to determine OIMLE (green, right), when considering a null model where expression is distributed in a log-normal fashion and any deviations are due to technical detection limits. Figures 25e and 25f depict that the relationship between expression and H3 27ac binding depends on a, but not on μ. Plot shows average promoter read density (intensity; black high; white low; scale bar, bottom) for H3K27ac (LPS 2h, top), H3K27ac (Unstim, middle), and H3K4me3 (2h LPS, bottom) genes corresponding to each of 10 quantile bins of population expression (Y axis) and each of 10 quantile bins of a (Fig. 25e, X axis) or μ (Fig. 25f. X axis). The overall population correlation between expression and the H3K27ac (Fig. 25e, top, middle) largely disappears after controlling for the percentage of single cells with detectable expression levels (a; Fig. 25e, middle), but this dependency remains for H3 4Me3 levels (Fig. 25 e, bottom). In contrast, controlling for μ (Fig. 25f) does not eliminate the dependency between expression level and K27ac, since within a single range of ji (vertical stripe), the correlation between population expression level and K27ac is maintained, suggesting that iper se is not the underlying determinant of this relationship. Figure 25k depicts bar plots showing p-valucs of correlation between average expression levels and K27ac only for immune response genes either as is (red) or when controlling for μ (blue) or u (green). Figure 25h depicts dynamic changes in u and μ in each module. Bar plots showing for each module (top: core antiviral; middle: peaked inflammatory; bottom: sustained inflammatory) the fraction of genes (Y axis) with a significant change only in a (by a likelihood ratio test, P < 0.01 , blue), only in μ (Wilcoxon test, P < 0.01 , green), or in both (each test independently, light blue), at each transition (X axis), in different conditions (marked on top). In each module and condition, the proportion is calculated out of the total number of genes in the module that are significantly bimodal (by a likelihood ratio test) in at least one timepoint during the response timecourse, and are expressed in at least 10 cells in both conditions. This number is marked on top of each bar.

[00064] Figures 27A-27F are a series of graphs and illustration depicting that IFN-β feedback drives heterogeneity in expression of "core" antiviral targets. Figure 27a depicts single cell expression distributions for Rsad (top) and Stat2 (bottom) after stimulating with LPS (left, black) or IFN-β (right, red) for 2h. Figure 27b depicts wave plots showing the distribution of expression of each of the genes in the "core" antiviral cluster (Y axis;

ordered as in Fig. 25b) at 2h stimulation with LPS (left) or IFN-β (right). Whereas the expression of most genes was bimodal at 2h with LPS, most were unimodally expressed at 2h with IFN-β (akin to the 4h LPS time point in Fig. 25b). Figure 27c depicts the "core" antiviral score (Y axis) for each LPS-stimulated cell (0, 1 , 2, 4, and 6h) and cells simulated for 2h with IFN-β (rightmost). Two "precocious" cells (yellow stars) have unusually high antiviral scores at 1 h LPS. (d) Normal quantile plots of the expression of genes from the "core" (cluster Id, left) and secondary (cluster Ic, right) antiviral clusters at l h LPS. The two "precocious" cells (yellow stars) express unusually high levels of "core" antiviral genes (left) but not of secondary genes (right). Figure 27e depicts that RNA- fluorescence in situ hybridization (RNA-FISH) confirmed the presence of rare early responders (arrow; yellow star), positive for both Ifnbl (magenta) and Ifit l (cyan). Grey: cell outlines. Scale bar: 25 μιη. Figure 27f presents a Venn diagram showing the coincidence for detection (>5 copies) of both Ifn bl (magenta) and I -fit 1 (cyan) by RNA-FISH after a l h LPS stimu lation (P < 10- 25, test of equal proportions).

[00065] Figures 28A-28E are a series of illustrations and graphs depicting microfluid ic blocking of cell-to-cell signaling affects response heterogeneity in antiviral and inflammatory modules. Figure 28A depicts experimental blocking of cell-to-cell communication. Left: CI chip; Right: On-chip stimulation, followed by actuation of microfluidic valves (red bars), seals the cells at individual chambers, preventing intercellular signaling. Figure 27b depicts expression of the genes (rows) in the "core" antiviral (Id, top rows) and "peaked" inflammatory (111c, bottom rows) modules in single cells (columns) from the on-chip (left; no cell-to-cell signaling) and in-tube (right) stimulations. Colors represent scaled expression values (z-scores). Figure 27c depicts gene expression distributions for individual representative genes from the "core" antiviral (top) and "peaked" inflammatory (bottom) clusters in the on-chip (left, blue; no paracrine signaling) or in-tube (right; black) 4h LPS stimulation. Figure 27d depicts violin plots of "core" antiviral (top panel, Y axis), "peaked" inflammatory (middle panel, Y axis), and "sustained" inflammatory (bottom panel, Y axis) scores for individual cells from (left to right): LPS Oh, LPS lh, LPS 2h, LPS 4h, LPS 6h, "On-Chip" Unstimulated, "On-Chip" LPS 4h, LPS 4h with GolgiPlug (Brefeld in A) added at Oh, LPS 4h with GolgiPlug added at l h, LPS 4h with GolgiPlug added at 2h, LPS 4h with Ifnar-/- BMDCs, and LPS 4h with Stat l-/- BMDCs. The two "precocious" cells (Fig. 27) with unusually high antiviral scores at lh LPS are denoted with yellow stars.

[00066] Figures 29A and 29B are a series of illustrations and graphs depicting that population-level paracrine signaling enhances and coordinates the "core" antiviral response while dampening and desynchronizing the "peaked" inflammatory ones. Figure 29a is a gene network model showing how positive IFN-β signaling induced the antiviral response and reduced its heterogeneity, while simu ltaneously activating a negative paracrine feedback loop, possibly including IL- 10, which dampened the "peaked" inflammatory cluster and increases its heterogeneity. Figure 29b is a cell population model showing how positive and negative paracrine feedback altered antiviral (magenta) and inflammatory (green) gene expression variability across cells. Grey denotes no expression.

DETAILED DESCRIPTION

[00067] This invention relates generally to compositions and methods for identifying the regulatory networks that control dendritic cell response, including core antiviral response, secondary antiviral response, maturation, induced inflammatory response and sharp peaked inflammatory response, as well compositions and methods for exploiting the regulatory networks that control dendritic cell response(s) in a variety of therapeutic and/or diagnostic indications.

[00068J The studies provided herein used single cell nucleic acid analysis, specifically, S ingle-Cell RNA Seq, to profile the mRNA in individual dendritic cells (DCs) responding to various pathogenic components. Using the Single-Cell RNA Seq profiling methods provides a number of advantages, such as, by way of non-limiting examples, cleaner signatures, a separation of antiviral circuits from maturation ones, and refining signatures identified in cell populations.

[00069] Single-cell RNA-Seq offers an unbiased approach for understanding the extent, basis, and function of gene expression variation between seemingly identical cells. However, fulfilling this promise requires a high-throughput workflow for profiling and analyzing many cells across different experimental cond itions. The disclosure provides a microfluidics-based approach to prepare single-cell RNA-Seq libraries from over 1 ,700 primary mouse dendritic cells (DCs) stimulated with three pathogenic components.

Substantial variation between individual cells exposed to the same stimulus was found, in both the fraction of cells expressing a given mRNA transcript at a detectable level and the transcript' s levels within these expressing cells. Distinct gene modules are characterized by different temporal heterogeneity profiles. In particular, a "core" module of antiviral genes is expressed very early by a few "precocious" cells and then becomes active in all cells at later time points. By stimulating cells individually in sealed microfluidic chambers, analyzing DCs from knockout mice, and modulating secretion and extracellular signaling, this response is propagated and coordinated via interferon-mediated paracrine signaling.

Surprisingly, preventing cell-to-cell communication also substantially reduces variability in the expression of a peaked, early-induced inflammatory module, suggesting that paracrine signaling additional ly represses a portion of the inflammatory program. The compositions and methods provided herein highlight the importance of cell-to-cell communication in controlling cellular heterogeneity and reveals general strategies that multicellular populations use to establish complex dynamic responses.

[00070] Using this analysis for the first time ever, a series of refined gene signatures for different response elements, referred to herein as the "Core Antiviral" gene signature, the "Secondary Antiviral" gene signature, the "Maturation" gene signature, the "Inflammatory Induced" gene signature, and the "Inflammatory Sharp Peaked" gene signature, have been uncovered. These signatures are genes that are expressed in coherent groups of single cells. Each of these gene signatures is provided in ^' fables 1 -5 below.

[00071] The "Core Antiviral" gene signature is induced in the earliest of the responding dendritic cells. The "Maturation" gene signature looks similar to the "Induced Inflammatory" gene signature at a population level, but using single cell analysis, it was established that the "Maturation" gene signature is expressed in only a subset of cells. The "Maturation" gene signature is responsible for allowing dendritic cells to recruit T cells and B cells, thereby bridging the gap between the innate and adaptive immunity system.

Table 1 . Core Antiviral Signature Genes

Table 1 A. Subset of Core Antiviral Signature Genes

ADAR GVI 1 OASL2

Al 607873 H2-T10 PARP12

AK172683 HERC6 PARP9

AW112010 I830012O16RIK PHF11

BST2 IFI203 PTTG1

CAR 13 IFI204 PYHIN1

CASP11 IFI205 RNASET2A

CD 274 IFI35 RTP4

CD69 IFI44 SAMD9L

CMPK2 IFI47 SERPINA3G

D14ERTD668E IFIH1 SLC03A1

DAXX IFITM3 SLFN5

DDX58 IGTP SLFN8

DDX60 IRGM1 SLFN9

DHX58 IRGM2 SP100

DTX3L MITD1 SP140

E030037K03RIK M DAL TA 1

EIF2AK2 MOV10 TOR3A

ETNK1 MPA2L TREX1

FAM26F MS4A4C TRIM30A

GBP2 MX1 TRIM30D

GBP3 NLRC5 UBA7

GM12250 NT5C3 USP18

GM14446 OAS1A XAF1

GM4902 OAS1G ZBP1

GM4951 OAS2 ZNFX1

GM5431 OAS3 ZUFSP

GM8979 SLFN13 TRIM5

CA13 GBP6 GVINP1

IFI16 IFIT1B MNDA

MS4A4A OAS1 PRIC285

SERPINA3

Table 2. Secondary Antiviral Signature Genes

Table 2 A. Subset of Secondary Antiviral Signature Genes

1110018G07RIK GNB4 RNF114

1600014C10RI K GYPC RN F135

2810474O19RIK H2-T23 SAMHD1

3110001I22RI K H2-T24 SETDB2

4930523C07RIK HEATR5B SGCB

9230105E10RIK IFI27L2A SLAMF7

ADAP2 IL18BP SLC25A22

AFTPH IRF9 SLFN 1

AI DA KYNU SPRED1

AIM1 LAP3 STARD3

A1M2 LGALS9 STXBP3A

AK142678 LNP TBC1D13

AK163331 MI ER3 TDRD7

AKT3 MINPP1 TFG

ALDH1B1 MKIAA1823 TMCC3

AP3M2 MLKL TMEM 140

APOBEC3 MTHFR TMEM67

AZI2 NAA25 TOR1AI P1 BC147527 NODI TOR1AI P2

BFAR P2RY14 TRIM25

CASP7 PARP11 TRIM34

CCDC25 PARP8 TRIM5

CCNJ PCGF5 UBE2L6

CH25H PFKP VCAN

DCK PLA2G16 VCPIP1

FBXW17 PPA1 WARS

FIMDC3A PP 1K WHSC1L1

FRMD4A PRPF38A XKR8

G530011O06RIK PS B9 XRN1

GBP4 RASA4 ZC3HAV1

GBP6 RIN2 ZFP800

GBP9 FBXW12 KIAA0040

C19orfl2 KIAA1715 PPHLN 1

KIAA0317 SLFN 12

Table 3. Maturation S ignature Genes

Table 3 A. Subset of Maturation Signature Genes

1200009I06RIK EIF2C3 PPP1CB

9030625A04RIK ETS2 PVR

AKNA ETV3 PVRL2

APOL7C FA 129A RAB8B

APPL1 FAM177A REL

BC035044 GPR85 RHOB

BCL2L1 H2-Q7 RND3

BIRC3 HSD17B11 SAMS Ml

BLNK IRF8 SEMA6D

CCL22 ITGA4 SERPINB9

CCR7 KTELC1 SERPIN B9B

CD72 MKIAA0769 SRGN

CD80 P25 ST3GAL1

CD83 NFKBIB STAT3

CD86 NUDT17 SWAP70

CDKN 1A NUP62-IL4I1 TBC1D1

CHAC2 OSGIN2 TIMP1

CRLF3 PALM2 TM EM39A

CSF1 PDZK1IP1 TNIP3

DENND5A PGAP2 VCAM1

EBI3 EXOC3L4 FAM177A1

IL4I 1 LACC1

Table 4. Inflammatory Induced Signature Genes

Table 4A. Subset of Inflammatory Induced Signature Genes

6330409N04RIK GM14005 PPAP2B

A130040 12RIK GPR84 PPP4R2

AA467197 H2-M2 PROCR

ACSL1 HCK PTPRJ

AO AH IL1B RAB10

B3GNT2 IL1RN RAB32

BCL2A1A IL27 RHBDF2

BCL2A1B IL6 RNF19B

BCL2A1C !RGl RPS6KA2

BCL2A1D ITGA5 SAA3

CALCRL ITGAV SBDS

CAV1 JAK2 SDC4

CCL5 KPNA3 SLC15A3

CD 14 LASS 6 SLC2A6

CD200 LC 2 SLC7A11

CD3S MAPKAPK2 SLC7A2

CFLAR MARCKSL1 SLFN2

CLEC4E MFU00294 SOD2

CLIC4 MKIAA1673 SQSTM1

CXCL16 MMP14 ST3GAL5

CXCL3 MTPN TAGAP

DCBLD2 NAMPT TANK

EHD1 NFKB1 TAR Ml

ELL2 NOS2 TLR1

FAM 102B OLR1 TNFRSF1B

FPR2 PARP14 TNFSF15

GADD45B PIK3R5 TRAF1

GBP5 PLEK TXNRD1

C15orf48 CERS6 Table 5. Inflammatory Sharp Peaked Signature Genes

Table 5A. Subset of Inflammatory Sharp Peaked Signature Genes CDC42EP4 N FKBI E SLC16A10

CLCN7 N IACR l SLC25A25

CLEC4D N LRP3 SLC25A37

CPD N RP2 SLC39A14

DDH D1 N U P54 SPATA13

DUSP16 N UPR1 TGM2

F10 ORAI2 TLR2

FA 108C OSBPL3 TNFAI P2

FA 20C PDE4B TNFAI P3

FLRT3 PI LRA TNI P1

FPR1 PI P5K1A TOPI

GRAMD1B PLAGL2 TREM1

H1FO PLE H02 TRIM 13

ICOSL PLSCR1 TSHZ1

I L1F9 PSTPIP2 ZEB2

I NSIG 1 PTAFR ZSWIM4

I RAK-2 PTP E

[00072] A desired target gene or combination of target genes is selected, and after determining whether the selected target gene(s) is overexpressed or under-expressed during a dendritic cell response, a suitable antagonist or agonist is used depending on the desired maturation and/or function outcome. Suitable antagonists and/or agonists include an antibody, a soluble polypeptide, a polypeptide agent, a peptide agent, a nucleic acid agent, a nucleic acid ligand, or a small molecule agent.

[00073] The modulating agents are used to modulate the expression of one or more target genes or one or more products of one or more target genes that have been identified as genes responsive to dendritic cell-related perturbations. These target genes are identified, for example, by contacting a dendritic cell with a modulating agent and monitoring the effect, if any, on the expression of one or more signature genes or one or more products of one or more signature genes. In some embodiments, the one or more signature genes are selected from those listed in Tables 1 -5A. The modulating agent can act directly on the expression of one or more target genes or one or more products of one or more target genes and/or the modulating agent can act indirectly on the expression of one or more target genes or one or more products of one or more target genes by modulating the expression, activity and/or function of a gene or a product of a gene that is known to be associated with the target gene(s).

[00074] In some embodiments, the target gene is tumor necrosis factor receptor

(TNFR). In some embodiments, the modulating agent alters the expression, activity and/or function ofTNFR. In some embodiments, the modulating agent alters the expression, activity and/or function of a gene that is assoc iated with TNFR, such as, by way of non- limiting example, a gene from those shown in Table 6 below. The underlined genes in Table 6 are genes that are upregulated when TNFR is absent, e.g.. knocked out, and the non- underlined genes are genes that are down-regulated when TNFR is absent, e.g., knocked out.

Table 6.

CCL5 PNRC1 AKNA CAV1 MTHFR

ETV3 CHD1 TRIM34 MLKL FAM53C

BLNK GBP9 CXCL10 AK178429 SLC7A11

S GN BTG2 ARL5C EGR2 TMEM140

MCMBP TMEM39A OSGIN2 AZI2 9030425E11RIK

IRF8 ARID5B DENND5A A130040M12RIK VCL

MARCKSL1 EIF2C3 RSAD2 PLEKHF2 TLR3

PVRL2 CST7 SEPW1 TRAF1 MKIAA1994

IFIT2 RPS6KA2 IFIT1 G530011O06RIK MAF

KTELC1 DLGAP4 FBXW17 DUSP1 SAMSN1

CCND2 BCL2A1A SMI F RELA TLR6

9030625A04RIK PIK3AP1 RB S2 SLFN9 AK138792

CDKN1A IFIT3 GRAM DIB LDLR NAA25

ISG15 CSF1 EPSTI1 FSTL1 AK172683

GL1PR2 FAM129A BCL2A1C NFKB2 ZCCHC2

CD86 1110038F14RIK HERC6 SERPINB9 CD14

SDC4 RNF19B TRMT61B A430084P05RIK F830016B08RIK

TNFSF8 BC006779 NCOA7 MKIAA0696 FILIPIL

I FIH1 KLF7 CCL4 SAMD9L RALGDS

CD80 BCL2A1B TRA2A NFKB1 TNFAIP2

IFI27L2A ISG20 CLU LY6A A230046K03RIK

IIGP1 TMEM219 CCRN4L MAP3K8 TSHZ1

D14ERTD668E H GN3 TAR Ml RBM7 TLR7

GBP4 TPN 5031414D18RIK 2310004I24RIK SPATA13

STAT5A APPL1 MFSD7 H2-T10 AK050909

AK163331 MITD1 1110018G07RIK LRRK2 INSIG1

RGS1 ICOSL OPTN PDZK1I P1 PTGS2

LAP3 TMC03 BATF2 EIF2AK2 H3F3B

CCL22 DYNC1I2 RANBP2 PLK2 SLC7A2

SWA 70 CDYL2 I FNB1 MGAT4A FOSL2

EBI3 IL13RA1 SNX10 IRG1 DAB2

AA467197 CLN3 TRIM13 RTP4 CALCRL

SLC2A6 ALDH1A2 STAT3 1810029B16RIK SPIC [00075] In some embod iments, the target gene is a Toll/interleukin- 1 receptor (TIR) domain-containing adapter protein (TIRAP). In some embodiments, the modulating agent alters the expression, activity and/or function of TIRAP. In some embodiments, the modulating agent alters the expression, activity and/or function of a gene that is associated with TIRAP, such as, by way of non-limiting example, a gene from those shown in Table 7 below. The underlined genes in Table 7 are genes that are upregulated when TIRAP is absent, e.g., knocked out, and the non-underlined genes are genes that are down-regulated when TIRAP is absent, e.g., knocked out.

Table 7.

LYZ1 TLR3 GBP9 DDHD1 ARHGEF3

SG K1 DENND1B ST8S1A4 AW112010 BTG1

PRDX1 PMP22 UBC CD72 APPL1

ACSL1 FAM20C FOSL2 AIM RD17 MTPN

MET FAM102B M PP1 NOS2 MFU00294

PDPN BATF2 PRDM1 CD47 NFKB2

CLEC4D PTAFR GTPBP2 LDLR ILIA

PTPRE TIFA FAM53C DLGAP4 CD40

9030425 E11RIK PYH IN1 LRRK2 RELA FBXL3

MMP13 EPSTI1 JHDM1D MINA ITGA5

LY6C2 CD274 PLK2 EXT1 4930523C07RIK

MCOLN2 A430084P05RIK CRBN ANKRD57 SWAP70

DENND3 F10 ISG20 SDC4 G530011O06RIK

SLPI 1810029B16R1K MALT1 WARS CXCL1

RSAD2 SLC16A10 PLEKHN1 PPP4R2 SH3BP5

CD38 DDX60 LRCH1 CHAC2 NFKBIE

1190002H23RIK OAS2 MCA32 CAR13 2310004I24RIK

SLC7A8 TH BS1 PTG ES SLC25A22 PIK3R5

ZCCHC2 NAA25 PSMB10 LZTFL1 NFKBIA

FCGR1 PHC2 WHSC1L1 AK139528 DNAJ B6

LY6A VCL PA2L FBXOll BI RC3

MT2 BST2 H3 F3B AK138792 BCL2L11

I RG 1 CLEC4E SLFN1 HSPA5 ICOSL

PPAP2B MXD1 TGM2 JAK2 A230046K03RIK

IFIT3 HCK AK178429 SLFN2 INSIG 1

RALGDS CCL2 SLC20A1 GYPC DUSP16

EG 2 MX2 XRN1 CXCL3 PDZK1I P1

PTX3 IER3 TMEM67 TMC03 IRF8

LY6I VWA5A MS4A4C PTTGl LM04

SLFN5 MFSD7 DNAJ B4 SBDS MTDH

GPR141 PARP10 SGMS2 WDR37 SOCS3 HIF1A SAMHD1 MITF FLRT3 ST3GAL1

E030037K03RIK IL1F9 UPP1 PELI1 CD83

CD180 PSTPIP2 NUB1 RNF19B NFKB1

FABP3 RGS14 BPAG1 PGAP2 PVRL2

SGMS1 FCGR4 IL18BP CSF1 SEMA6D

STAT1 AK200837 IRF7 UBXN2A KYNU

P4HA1 MNDAL ZNFX1 TNIP1 BCL2A1A

C5AR1 PLEKH02 PNPT1 OSGIN2 EIF2C3

CMPK2 NOTCH 2 AK050909 JUNB DENND4A

CALCRL FAM26F IL12RB2 H1FO IL23A

CAV1 OPTN DNAJC13 1200009I06RIK CCNG2

FOS STXBP3A GTF2B RAB8B NFKBIB

FAM46C IFI203 SETDB2 MTMR14 BATF

CFB CCRL2 SLFNIO-PS FGL2 PALM2

SLC7A2 H2-T24 ISG15 KPNA3 EBI3

STK38L AK042010 LY75 CD86 MMP25

HK3 TMEM219 BC006779 BTG2 PNRC1

GM 14446 TAP2 USP12 MARCKSL1 CCND2

GPD2 IFIT1 NT5C3 TET2 FILIP1L

IFIT2 GM6644 SGCB BC035044 EHD1

KLF3 MERT CASP7 H2-Q7 SAMSN1

CST7 MEF2A TRIM5 SLC39A14 AY096003

SLC25A 7 6330409N04RIK NM! BCL2L1 BCL2A1D

CLCN7 OSBPL3 ZFP800 HSD17B11 CISH

CASP1 LAP3 feoOlSBOS I AA467197 NUP62-IL4I1

IL15RA PLAUR CCL4 CCR7 CCL3

JARID2 DYNC1H1 ZUFSP RNF214 TBC1D1

EGR1 P2RY13 DENND5A ETV3 9030625A04RIK

IRF1 TXNRD1 HEATR5B AK139487 CCL17

D1ERTD622E ARFGEF1 RABGEF1 MARCO PPP1CB

SNXIO LRP12 TNFAIP3 HIST1H4D TRAF1

CXCL10 PNP ZSWIM4 PMAIP1 BLNK

I830012O16RIK SAMD9L AK150559 TNI 3 CD80

2310016C08RIK OASL1 TIMP1 AK052414 TNFSF15

DRAM1 1110038F14RIK ST3GAL5 CXCL16 REL

MKIAA1994 MLKL MKIAA1823 GM6377 BCL2A1B

NPY FRMD4B TSHZ1 RGS1 PDE4B

FBXO30 ARG2 MAX SEC24B FAM177A

SLFN3 GNG12 PTPRJ ARF4 FAM129A

ADAP2 IGTP PPA1 2010106G01RI KTELC1

HIST3H2A P2RY14 ARL5C CD14 RND3

FPR1 RNF34 ITGA4 CD200 TMEM39A

MDM2 CD44 AMN1 A630001G21RI GADD45B

IL1RN ASB13 LYRM1 TMCC3 PTGS2

[00076] In some embodiments, the target gene is Stat l . In some embodiments, the modulating agent aJters the expression, activity and/or function of Stat l . In some embodiments, the modulating agent alters the expression, activity and/or function of a gene that is associated with Stat l , such as, by way of non-limiting example, a gene from those shown in Table 8 below. The underlined genes in Table 8 arc genes that are upregulated when Stat l is absent, e.g., knocked out, and the non-underlined genes are genes that are down-regulated when Stat l is absent, e.g., knocked out.

Table 8.

E030037K03RIK TMEM184B PPP2R5A RALGDS SPATA13

IFI35 GNB4 ARID5B JARID2 IRAK3

MS4A4C LARP1 RNF19A PLSCR1 APOL7C

TNFSF15 PGAP2 VWA5A CPD NIACRl

DTX3L 5-Mar ANXA7 MALT1 CXCL1

PHF11 IL7R RFFL NFKBIB PTGES

TOR3A 9930111J21RIK1 PPAP2B TPR LY6C2

RND3 RIN2 MARCO MKI67 ORAI2

TRIM25 OAS2 RAB10 APBB2 CI U

IRGM2 MAFK JHDM1D FLRT3 PTPRE

PARP12 PSMB9 IFT172 TGM2 C5AR1

OAS3 CH25H RABGEF1 TARM1 LCN2

OASL1 KPNA3 NRP2 MKIAA0769 ARG2

DDX60 PCGF5 INHBA SLC20A1 SI.PI

MXD1 RAP2C SNX10 AK042010 IL1F9

SAMHD1 MBNL2 PLK3 GNG12 PTX3

RNASET2A PARP11 TNFAIP2 DUSP1G CD38

NLRC5 FAM26F BIRC6 PILRA GM6644

ZNFX1 4930523C07RIK PPP4R2 BHLHE40 SAA3

BCL2A1D PELIl FPR2 FPR1 SOD2

[00077] In some embodiments, the target gene is interferon production regulator

(IFNR). In some embodiments, the modulating agent alters the expression, activity and/or function of IFNR. In some embodiments, the modulating agent alters the expression, activity and/or function of a gene that is associated with IFNR, such as, by way of non- limiting example, a gene from those shown in Table 9 below. The underlined genes in Table 9 are genes that are upregulated when IFNR absent, e.g., knocked out, and the non- underlined genes are genes that are down- regulated when IFNR is absent, e.g., knocked out.

Table 9.

MALT1 RGL1 TLR1 SLFN2 MXD1

LM04 NUP54 ZCCHC2 MINPP1 IGTP

TXNRD1 HIF1A CCNL1 ATF3 BC013712

IFRD1 SH3BP5 KLF7 LARP1 PGAP2

TNF PLAT MFU00294 SP140 SETDB2

9030425E11RIK 6330409N04RIK INPP5B UBE2L6 PML

CD38 GRAMD1B FAM102B AK142678 ZNFX1

EG 2 MMP14 3110043O21RIK BFAR LAP 3

BC031781 PPFIA1 NFKBIA TRIM5 PARP12

SLC20A1 FAM46C TNFAIP2 IRF8 GBP3

INHBA GP 84 TRIM13 TFG ADAP2

METRNL 1200009I06RIK CRBN KPNA3 GVIN1

PLK2 ST8SIA4 PSTPIP2 ZC3H7A SLFN9

1190002H23RIK PLEKH02 PLAGL2 P2RY14 EIF2AK2

ZSWIM4 RFFL GCNT2 NAMPT KYNU

LRP12 DNAJB4 THBS1 IKZF1 IRGM2

CXCL3 SLC7A2 AK200837 NUB1 ZUFSP

CCL2 FAM108C TPR MYD88 AI607873

FABP3 ARHGAP31 NFKB1 FAM26F MPA2L

RASA2 0LR1 ZEB2 AFF1 A230046K03RIK

MKIAA0769 RABGEF1 CDYL2 PPM1K OAS3

ALDH1A2 F10 MED21 9930111J21RIK1 IFI204

SPIC HIPK2 STK38L EHD4 MX2

TSHZ1 UPP1 ARID5B BBX E030037K03RIK

MT2 PVR MBNL2 SAMD9L FGL2

PPP1R15A IRAK-2 TLR6 ETV3 DTX3L

PRDX1 SQSTM1 APPL1 UBA7 MS4A6C

MET TLR2 CHAC2 PLEKHF2 PARP14

FBXO30 AK217941 DCBLD2 CCDC86 IFI35

MMP13 PLA2G4A RNASET2B TCF4 1110018G07RIK

PDPN CD200 FRMD4B IL15RA IL12B

CLEC4D IL20RB H3F3B KTELC1 TOR3A

CD44 H2-M2 OSBPL3 I830012O16RIK NLRC5

MC0LN2 DENND3 GTPBP2 LNPEP TOR1AIP1

SGMS1 RNASET2A AK163331 DYNC1I2 AK172683

TNFAIP3 JHDM1D RNF114 ISG15 ETNK1

AD0RA2B CLCN7 BIRC3 MTHFR LGALS9

2310016C08RIK ELL2 T IP3 RND3 MS4A4C

RALGDS RNF19A ISG20 GBP4 ZBP1

SPATA13 LASS6 MORC3 5730508 B09RIK GBP2

0RAI2 IFT172 KAT A1 TAGAP SLFN5

TNIP1 H2-Q7 FRMD4A SLAMF7 CCND2

RBM7 SLC39A14 PSMB10 STXBP3A PTTGl

BPAG1 UBE2Q2 ANKRD17 RNF19B TRIM25

[00078] In some embodiments, the target gene is one or more genes from those listed below in Table 10, Table 1 1 or Table 12. In some embodiments, the modu lating agent alters the expression, activity and/or function of the target gene(s). The underlined genes in Table 1 0, Tabic 1 1 and/or Table 12 are genes that are upregulated when a target gene absent, e.g., knocked out, and the non-underlined genes are genes that are down-regulated when the target gene is absent, e.g., knocked out. fable 10.

CCND2 TMCC3 STXBP3A SPIC ZFP36

PARP9 AIDA FAM177A HK3 LCP2

SLFN5 CEPT1 MORC3 BCL2L1 HIPK2

MS4A6C PPP1CB TRAFD1 MKIAA1673 ILIA

DHX58 BIRC3 RAP2C PSTPIP2 MPP5

PARP12 GBP4 9930111J21RIK1 AK200837 SERTAD2

CCL22 UBE2L6 CPNE3 AP4B1 MAPKAPK2

MS4A4C VCA TMEM184B SOCS3 SQSTM1

DDX58 I830012O16RIK KPNA3 PPAP2B ZCCHC2

PHF11 PLA2G16 OAS1G SLC25A37 MEF2A

RTP4 TNFAIP2 CASP7 MDM2 CCL4

CXCL10 4930523C07RIK TBC1D1 SLC3A2 PLAGL2

GBP6 AI607873 ETNK1 PILRA ARG2

IFI204 FPR2 BC013712 STAT3 RPS6KA2

ADAP2 PTTG1 IFI203 ZC3H12C LRP12

JAK2 USP25 AY096003 LASS6 NUP54

SDC4 NOTCH2 TMEM67 PRDM1 NFKBID

IFITM3 NUP62-IL4I1 1600014C10RIK TLR2 H5PA5

DDX60 INA THFR SLC7A8 CPEB4

SLFN9 SLFN2 MINPP1 FAM53C - 6330409N04RIK

KYNU BLNK NCOA7 ARHGAP31 CXCL1

NLRC5 TAGAP SGCB SGMS2 ATF3

MX2 TNFRSF1B KATNA1 ycL SLC20A1

9230105E10RIK OAS3 XRN1 SEC24B TOPI

CD69 PELIl AZI2 NRP2 PMP22

PARP14 TRI 34 SAMD9L NFKBIZ NFKBIB

BCL2A1D AK139487 IL7R OSM INHBA

GM8979 IL27 ISG15 IL1F9 PLEKH02

IFI205 ZNFX1 CCDC25 IFNB1 FAM20C

BCL2A1B OAS1A GPR141 TNFSF4 PY

E030037K03RIK MNDAL T C03 TNIP1 SERPINB2

AFTPH RNASET2A GBGT1 OSG1N2 ZSWIM4

ARL5C MAP2K1 4930453N24RIK PGF PLK3

FAM102B FAM26F SERTAD3 SLC39A2 FBXO30

FGL2 CD86 PPFIA1 MET PTX3

IRF7 INSIG1 FIO TNFAIP3 PIP5K1A

0ASL1 MARCKSL1 GNA13 PVR IFRD1

AK138792 IFI44 GYK A130040M12RIK MMP13

CCL5 DAXX LGALS9 PRDX1 FLRT3

NFKB1 ST3GAL5 AIM2 THBS1 IRF4

SLC03A1 RAB8B IL20RB OLR1 GM6644

TNFSF15 BIRC6 REL MAFK 2310016C08RIK

ITGA4 EBI3 3110001I22RIK IL12RB2 PTPRE

ZBP1 BBX AK042010 TRIM13 CISH

Table 11.

SLFN8 MINA ARFGEF1 HK2 CCL17

USP18 IRAK3 MKI67 NCK1 MNDAL

OSBPL3 UBXN2A BST2 ZC3HAV1 TNIP3

ST3GAL5 IRF9 NUDT17 SERPINB2 MDM2

IKZF1 AIDA FNDC3A CCNG2 BHLHE40

SLC03A1 DYNC1I2 TAPBP GNA13 UBE2Q2

SNX10 ZEB2 PYHIN1 CHD1 OPTN

GPR141 TLR1 PPM1K H2-Q7 4930453N24RIK

MYD88 TRI 30D NAMPT TTC39B H3F3B

LASS6 IFIT1 JHDM1D STAT5A 9030425E11RIK

PIK3R5 FILIP1L CCND2 GOLGA3 SAT1

NFKB1 MTMR14 TAGAP OSGIN2 OSM

CPD AK139487 GNG12 METRNL FBXL3

CD38 INPP5B NRP2 HK3 RNF2

STAT1 PCGF5 CAR13 PTPRE PENK

A230046K03RIK 3110043O21RIK CHAC2 PLA2G4A LRP12

RAB10 SLFN9 NLRC5 SGCB 3110001I22RIK

FAM129A XRN1 MS4A6B SLC3A2 RALGDS

RAB32 ADAR SP100 SAMHD1 KLF6

IL12B TNFRSF1B NUP54 FCGR1 FBXO30

BATF IFI47 PARP8 ANKRD57 GM 14047

SLC1GA10 MFSD7 P2RY14 LYRM1 RPS6KA2

DTX3L DDHD1 IFI35 IQSEC2 SLC39A2

FPR1 PVRL2 FAM46A IL18BP ARG2

STXBP3A C5AR1 GBP3 ANXA7 MAFK

TNIP1 CLN3 D14ERTD668E SLC7A11 DNAJB4

GPR84 CASP7 IL13RA1 FOS SLC20A1

1200009I06RIK MAP2K1 GLIPR2 NFIL3 MET

PARP9 DENND4A CLEC4E H1F1A MMP13

SLC2A6 PSME1 FBXW17 PLEKHN1 MT2

PSMB10 TLR2 CMPK2 SLC15A3 CCRN4L

ICOSL CCL5 CCR7 CASP3 FABP3

PALM2 ATXN7L1 CXCL3 PLEKH02 TNF

CCDC86 MS4A4C TRIM25 BPAG1 IL1RN

TMCC3 ZBP1 PPP4R2 PTX3 SLPI

CXCL16 DNAJB6 TET2 IFT172 EGR1

HERC6 CLIC4 F10 CIAPIN1 SQSTM1

FAM102B EPSTI1 SLC39A14 DCK ATF3

IRGM1 TBC1D1 MKIAA0694 RCAN1 CDKN1A

PSME2 MX2 SP140 SERPINB9B PPP1R15A

ACPP BC035044 ANKRD17 OLR1 FAM46C

AK150559 IL6 CD69 TRIM13 AK050909

CD40 H2- 2 KYNU BBX IFRD1

SAMSN1 SH3BP5 PLAT SGMS1 TGIF1 I PNP SEPW1 MKIAA1673 NLRP3

RTP4 TBK1 S0CS3 PMAIP1 FOSL2

LCN2 I RF7 SLC25A22 FAM82A2 NDRG1

TOR1AIP1 RNF114 ED21 ARID5B LAP 3

G BP5 AP0L7C RNF34 CD274 CCL7

IL18 EHD1 PTTG1 INSIG1 RGS1

PARP14 CLCN7 IL15RA TIPARP CXCL1

MARCKSL1 DDX58 LRCH1 SLFN1 A130040M12RIK

TRAF1 GPD2 RNASET2B TOPI DUSP1

AKNA IGTP AFTPH AP4B1 BTG1

FAM20C TBC1D13 TFG ISG20 A430084P05RIK

PTPRJ TMEM39A TNFSF9 CARHSP1 CAR4

NCOA7 TRIM26 RNF135 MIER3 ZFP36

GPR85 FAS TLE3 TGM2 PMP22

USP25 LY6C2 PNRC1 SERTAD2 BC031781

IFITM3 PGAP2 CCNL1 TIMP1 SRG N

SGK3 PPP1R15B AK163331 SLFN3 NFKBID

TOR3A M IAA1994 RNF19A PROCR ZSWIM4

KTELC1 FTSJD2 DENND3 VCAM1 UPP1

ZNFX1 ZUFSP SLFN2 I830012O16RIK JUNB

GADD45B PLAGL2 IL12RB2 H2-T24 CCL4

2010106G01RIK XAF1 PPAP2B UBC CAR2

OASL2 1110038F14RIK CMBP PRDM1 ,EG 2

TAPBPL IL15 BRAF BTG2 RABGEF1

GBP2 UBE2L6 NFKBIB RASGEF1B H2-T23

TAP1 PSMB9 MFU00294 MERTK ILIA

MTDH LARP1 TLR3 RNF139 DAB2

EIF2C3 AK138792 THBS1 IL1F9 PTGS2

CD47 EIF2AK2 PDE4B RGL1 CISH

RELA MCA 32 PFKFB3 SCARF1 CXCL2

PARP12 CD83 DENND1B GM6377 FLRT3

I FIH1 ARMC8 SBDS TMEM67 CCL2

IRAK-2 SEC24B RAB20 HIST3H2A PLK3

PILRA DAXX ARF4 AK052414 PDPN

TNFSF15 MKIAA0769 AA960436 RFFL CCL3

ITGA4 BCL2A1B NODI IFI27L2A IER3

CCL22 GTPBP2 GNB4 CPEB4 G530011O06RIK

TRAFD1 AK217941 TREM1 OAS3 2310016C08RIK

TNFAIP2 LZTFL1 EXPI 2810474O19RIK INHBA

MS4A6D PDZK1IP1 CH25H RHOB PRDX1

KPNA3 MDFIC ETS2 A630001G21RIK PLAUR

TIFA ETNK1 CD80 HI PK2 1190002H23RIK

KLF7 MTPN LY75 IL7R PLK2

TREX1 ZDHHC21 STAT3 WARS Table 12.

[00079] The sensitivity of the techniques provided herein allows for the detection and definition of closely related subpopulations of cells. These techniques allow for the identification of gene response modules, e.g., signatures, which are selectively induced in distinct subsets of cells. Correlative analyses between single cells are useful in

reconstructing cellular circuits and identifying regulators of these modules.

[00080] Recent molecular studies have revealed that, even when derived from a "homogenous" population, individual cells can exhibit significant differences in gene expression, protein levels, and phenotypic output (Spencer, S. L., Gaudet, S., Albeck, J. G., Burke, J. M. & Sorger, P. K. Non-genetic origins of cell-to-cell variability in TRA1L- induced apoptosis. Nature 459, 428-432, doi: 10.1038/nature08012 (2009); Cohen, A. A. et al. Dynamic Proteomics of Individual Cancer Cells in Response to a Drug. Science 322, 1 5 1 1 - 1516, doi: 10.1 126/science. l 160165 (2008); Niepel, M, Spencer, S . L. & Sorger, P. . Non-genetic cell-to-cell variability and the consequences for pharmaco logy. Curr. Opin. Chem. Biol. 13, 556-561 , doi: 10.1016/j.cbpa.2009.09.01 5 (2009); Sharma, S. V. et al. A chromatin-medialed reversible drug-tolerant state in cancer cell subpopu lations. Cell 141 , 69-80, doi: 10.1016/j.celL2010.02.027 (2010); Gascoigne, K. E. & Taylor, S. S. Cancer cells display profound intra- and interline variation following prolonged exposure to antimitotic drugs. Cancer cell 14, 1 1 1 -122, doi: 10.101 6/j.ccr.2008.07.002 (2008), with important functional consequences (Spencer, S. L., Gaudet, S., Albeck, J. G., Burke, J. M. & Sorger, P. K. Non-genetic origins of cell-to-cell variability in TRAIL-induced apoptosis. Nature 459, 428-432, doi: 10.1038/nature08012 (2009);Sharma, S. V. et al. A chromatin-mediated reversible drug-tolerant state in cancer cell subpopulations. Cell 141 , 69-80,

doi: 10.1016/j.cell.2010.02.027 (2010): Gascoigne, K. E. & Taylor, S. S. Cancer cells display profound intra- and interline variation following prolonged exposure to antimitotic drugs. Cancer cell 14, 1 1 1- 122, doi: 10.1016/j.ccr.2008.07.002 (2008); Feinerman, O. et al. Single-cell quantification of IL-2 response by effector and regulatory T cells reveals critical plasticity in immune response. Mo lecular Systems Biology 6, 1 -16,

doi:papers2://publication/doi/10. 1038/msb.2010.90 (2010)). Existing studies of cellular heterogeneity, however, have typically measured only a small number of pre-selected RNAs (Yu, M. et al. RNA sequencing of pancreatic circulating tumour cells implicates WNT signalling in metastasis. Nature 487, 510-513, doi: 1 0.1038/naturel l 217 (2012); Raj, A., Rifkin, S. A., Andersen, E. & Van Oudenaarden, A. Variability in gene expression underlies incomplete penetrance. Nature 463, 91 3-91 8, doi: 10.1038/nature08781 (2010)) or proteins simultaneously (Spencer, S. L., Gaudet, S., Albeck, J. G., Burke, J. M. & Sorger, P. K. Non- genetic origins of cell-to-cell variability in TRAIL-induced apoptosis. Nature 459, 428-432, doi: 10. 1038/nature08012 (2009); Cohen, A. A. et al. Dynamic Proteomics of Individual Cancer Cells in Response to a Drug. Science 322, 1 1 1 -1 516, doi: 10.1 126/science. l 1 601 65 (2008); Niepel, M., Spencer, S. L. & Sorger, P. K. Non-genetic cell-to-cell variability and the consequences for pharmacology. Curr. Opin. Chem. Biol. 1 3, 556-561 ,

doi: 10.1016/j.cbpa.2009.09.01 5 (2009); Sharma, S. V. et al. A chromatin-mediated reversible drug-tolerant state in cancer cell subpopulations. Cell 141 , 69-80,

doi: 10.1016/j.cell.2010.02.027 (2010); Gascoigne, K. E. & Taylor, S. S. Cancer cells display profound intra- and interline variation following prolonged exposure to antimitotic drugs. Cancer cell 14, 1 1 1 - 122, doi: 10.101 6/j.ccr.2008.07.002 (2008); Dalerba, P. et al. Single-cell dissection of transcriptional heterogeneity in human colon tumors. Nature Biotechnology 29, 1 120-1 127, doi: 10.1 038/nbt.2038 (201 1 ); Bendall, S. C. & Nolan, G. P. Single-Cell Mass Cytometry of Differential Immune and Drug Responses Across a Human Hematopoietic Continuum. Science (New York, NY) 332, 677-678,

doi: 10.1 126/science.1206351 (201 1 ), because genomic profiling method (Bendall, S. C. & Nolan, G. P. Single-Cell Mass Cytometry of Differential Immune and Drug Responses Across a Human Hematopoietic Continuum. Science (New York, NY) 332, 677-678, doi: 1 0.1 1 26/science.1206351 (201 1 ); Altschuler, S. J. & Wu, L. F. Cellular Heterogeneity: Do Differences Make a Difference? Cell 14 1 , 559-563, doi: 10.1016/j.cell.2010.04.033 (2010); Kalisky, T., Blainey, P. & Quake, S. R. Genomic Analysis at the Single-Cell Level. Annual review of genetics 45, 431 -445, doi:papers2://publication/doi 10.1 146/annurev- genet-102209- 163607 (201 1 ); Kalisky, T. & Quake, S. R. Single-cell genomics. Nature Methods 8, 3 1 1-314 (201 1 )) could not be applied to single cells until very recently (Islam, S. et al. Characterization of the single-cell transcriptional landscape by highly multiplex RNA-seq. Genome Research, doi:papers2://publication/doi l 0.1 101/gr. l 10882.1 10 (201 1); Tang, F. et al. RNA-Seq analysis to capture the transcriptome landscape of a single cell. Nature Protocols 5, 516-535, doi: 10.1038/nprot.2009.236 (2010); Tang, F. et al. mRNA-Seq whole-transcriptome analysis of a single cell. Nature Methods 6, 377-382,

doi: 10.1038/nmeth. l 315 (2009); Ramskold, D. et al. Full-length mRNA-Seq from single- cell levels of RNA and individual circulating tumor cells. Nature Biotechnology 30, 777- 782, doi:papers2://publication/doi/l 0.1 038/nbt.2282 (2012); Hashimshony, T., Wagner, F., Sher, N. & Yanai, I. CEL-Seq: Single-Cell RNA-Seq by Multiplexed Linear Amplification. Cell Reports, doi: 10.1016/j.celrep.2012.08.003). Here, single-cell RNA-Seq was used to investigate heterogeneity in the response of a model mammalian system, bone marrow derived dendritic cells (BMDCs) stimulated by lipopolysaccharide (LPS). Extensive, and previously unobserved, bimodal variation was discovered in both the abundance and splicing patterns of RNA transcripts, which were independently validated by RNA- fluorescence in situ hybrid ization of selected transcripts. In particular, hundreds of key immune genes are bimodally expressed across individual cells, surprisingly even for genes that are very highly expressed at the population average. Moreover, splicing patterns across single cells demonstrate previously unobserved levels of heterogeneity: for genes that have mu ltiple splice iso forms at the population level, individual cells exhibit a bias towards predominant expression of one particular isoform. As shown by the Examples provided herein, these cell-to-cell differences are driven by heterogeneity in both cell state and cell circuit usage. While some of the bimodality reflects the presence of BMDCs in closely related, yet distinct, known maturity states, other bimodal patterns exist even within cells in the same maturity state, reflecting differences in the usage of key regulatory circuits between otherwise identical cells. For example, a modu le of 137 highly variable, yet co- regulated, antiviral response genes was identified. Using BMDCs from knockout mice, the studies presented herein demonstrate that bimodality in this antiviral module may be propagated through an interferon circuit involving the master antiviral transcriptional regulators Stat2 and lrf7. This study demonstrates the power and promise of unbiased single-cell genomics in uncovering extensive functional diversity between cells and in deciphering cell states and circuits.

[00081] The above analysis provides a proof-of-concept demonstrating how covariation between transcripts across single cells in the same condition and overall state can help to identify and assemble regulatory circuits whose d ifferential usage promotes significant cellular heterogeneity. Specifical ly, in the variable circuit (Fig. 19) interferon signaling is required for induction of Stat2 and Irf7, which, in turn, act to induce the variable antiviral cluster genes. The experiments do not definitively determine, however, which component of the circuit causes the observed heterogeneity/^r je. One compelling possibility is that upstream noise is propagated from the interferon-signaling pathway first to Stat2 and lrf7 and then to the target genes. This hypothesis is supported by the variation that was observed in Stat protein levels and nuclear localization. It is also supported by recent studies (Zhao, M., Zhang, J., Phatnani, H., Scheu, S. & Maniatis, T. Stochastic Expression of the Interferon-? Gene. PLoS biology 10, el 001249 (2012); Apostolou, E. & Thanos, D. Virus Infection Induces NF-kappaB-dependent interchromosomal associations mediating monoallelic IFN-beta gene expression. Cell 134, 85-96 (2008); Rand, U. et al. Mu It i- layered stochasticity and paracrine signal propagation shape the type-I interferon response. Molecular Systems B iology 8, doi: 10.1038/msb.2012.17 (2012)) demonstrating that over expression of Irf7 during viral replication in mammalian cells reduces heterogeneity in Ifn-β production and that Irf7 translocation correlates with Ifn-β production under a viral stimu lus. Notably, variability in the expression of interferon-stimulated genes (e.g., Isg l 5) and interferon- induced proteins that correlated strongly with the levels of Irf7 and Stat2 was also observed. This was not observed in previous studies w ith uniform Ifn-β stimulation (Zhao, M., Zhang, J., Phatnani, H., Scheu, S. & Maniatis, T. Stochastic Expression of the Interferon-β Gene. PLoS biology 10, e l 001249 (2012)), supporting the hypothesis that variability in interferon feedback drives downstream heterogeneity.

[00082] A similar strategy could potentially be used to explore the consequences of bimodality in splicing. Even looking at just 18 cells, interesting examples of bimodal splicing patterns were observed for genes whose isoforms have distinct functional consequences. For example, the splicing regulators Srsf3 and Srsf7 are each known to contain a "poison cassette exon", that, when included, targets the RNA for degradation via nonsense-mediated decay (Anko, M.-L. et al. The RNA-binding landscapes of two SR proteins reveal unique functions and binding to diverse RNA classes. Genome Biology 13, doi: 10.1 186/gb-2012-13-3-rl 7 (2012)). While these exons are very weakly expressed at a population level, one of the single cells (cell S 13, Fig. 20) exclusively expressed the poisoned isoforms at high levels (for both Srsf3 and Srsf7, 1 1 cells exclusively expressed the other). Since Srs†3 itself is responsible for increasing inclusion of its own poison cassette exon in a negative feedback loop (Anko, M.-L. et al. The RNA-binding landscapes of two SR proteins reveal unique functions and binding to diverse RNA classes. Genome Biology 13, doi: 10.1 186/gb-2012- 1 3-3-r l 7 (2012)), S 13 may in fact represent the highest levels of Srsf3 activity. When armed with a larger number of cells, correlation analyses could be used to identify potential targets of Srsf3. Splicing differences in other regulatory genes, meanwhile, may further enhance expression diversity: for example, proteins encoded by different isoforms of lrf7 - bimodally spliced in the cells (Fig. 3c) - differentially activate interferon-responsive genes in vitro (Ning, S., Huye, L. E. & Pagano, J. S.

Regulation of the Transcriptional Activity of the IRF7 Promoter by a Pathway Independent of Interferon S ignaling. Journal of Biological Chemistry 280, 12262- 12270 (2005)). These examples suggest that heterogeneity in splicing may represent another potential layer of response encoding.

[00083] The studies provided herein discover extensive bimodality in the transcriptional response of BMDCs to LPS stimulation, reflected in gene expression, alternative splicing, and regulatory circuit activity. In gene expression, hundreds of bimodally expressed transcripts encoding key immune proteins, including those that are highly expressed in the population average, were found. While variation in some genes is due to a minority sub-population in a different maturation state, others reflect the bimodal activity of an anti-viral regulatory circuit. Co-variation across single cells can help dissect refined functional gene modules that may be ind istinguishable in population scale measurements. In particular, in a recent population-scale study (Garber, M. et al. A High- Throughput Chromatin Immunoprecipitation Approach Reveals Principles of Dynamic Gene Regulation in Mammals. Molecular Cell 47, 810-822,

doi: 10.101 6/j.molcel.2012.07.030 (2012)), a large cluster of 808 "late-induced" LPS genes that was enriched for maturation genes as well as antiviral genes controlled by STAT proteins was identified. These two subsets could not be teased apart based on population- level data alone, but the single-cell data from a single time point clearly distinguishes them as expressed in different single cells. Similarly, the unexpected and prevalent skewing that was discovered in alternative splicing between single cells revises the molecular view of this process. Both phenomena also allow for the treatment of each cell as a "perturbation system" for reconstructing cell circuits (Angelo, K. et al. A biophysical signature of network affiliation and sensory processing in mitral cells. Nature 488, 375-378, doi:papers2://publication/doi/ 10.1038/naturel l 291 (2012); Sachs, K., Perez, O., Pe'er, D. & al, e. Causal protein-signaling networks derived from multiparameter single-cell data. Science (New York, NY) (2005)). Indeed, even with data from just 1 8 single cells and focusing on induced genes, the studies herein demonstrated as a 'proof of concept' how different regulators could be causally connected to their co-varying targets within an interfei on-driven antiviral circuit that was subsequently validated in knockout models. Finally, although many of the analyses focused on highly expressed genes to remove the possible influence of amplification noise, the data also reveal significant bimodality amongst more moderately expressed transcripts, such as large non-coding RNAs (Fig. 21 ). This observation suggests an intriguing possibility that the lower expression levels of these transcripts in the population (Cabili, M. N. et al. Integrative annotation of human large intergenic noncoding RNAs reveals global properties and specific subclasses. Genes & development 25, 1915- 1927 (201 1 )) may be the result of a small number of cells expressing them at high levels rather than all of the cells expressing them at a low level, although further technical improvements will be necessary to disentangle these two hypotheses (Fig. 9). As such, single-cell measurements should help facilitate the discovery, annotation, and analysis of these transcripts.

[00084] Comparing these results to other single cell RNA-Seq data sets indicates that the source of the analyzed tissue {in vitro vs. ex vivo), the biological condition of the individual cells (steady state vs. dynamically responding), and the heterogeneity in cellular microenvironment all likely influence the extent of single-cell heterogeneity within any ind ividual system. When applied to complex tissues— such as unsorted hone marrow, different stages of developing embryos, heterogeneous tumors, and rare clinical samples (Spencer, S. L., Gaudet, S., Albeck, J. G ., Burke, J. M. & Sorger, P. . Non-genetic origins of cell-to-cell variability in TRAIL-induced apoptosis. Nature 459, 428-432,

doi: 10.1038/nalure08012 (2009);Todd, R. & Margolin, D. H. Challenges of single-cell diagnostics: analysis of gene expression. Trends Mol. Med. 8, 254-257 (2002)) - the variability seen through single-cell genomics may help determine new cell classification schemes, identify transitional states, discover previously unrecognized biological distinctions, and map markers that differentiate them. Fulfilling this potential would require novel strategies to address the high levels of noise inherent in single-cell genomics - both technical, due to minute amounts of input material, and biological, e.g. , due to short bursts of R A transcription (Taniguchi, Y. et al. Quantifying E. coli Proteome and Transcriptomc with S ingle-Molecule Sensitivity in Single Cells. Science 329, 533-538,

doi: 10. ] 126/science. l 188308 (2010); Cai, L., Dalai, C. K. & Elowitz, M. B. Frequency- modulated nuclear localization bursts coordinate gene regulation. Nature 455, 485-490, doi:nature07292 [pii] 10.1038/nature07292 (2008)). Future studies that couple technological advances in experimental preparation with novel computational approaches would enable analyses, based on hundreds or thousands of single cells, to reconstruct intracellular circuits, enumerate and redefine cell states and types, and fundamentally transform the

understanding of cellular decision-making on a genomic scale.

[00085J The studies provided herein also use a microfluidic system to generate and analyze more than 2,000 SMART-Seq (Ramskold, D. et al. Full-length mR A-Seq from single-cell levels of RNA and individual circulating tumor cells. Nature Biotechnology 30, 777-782, doi:papers2://publication/doi/10.1038/nbt.2282 (2012)) single cell Bone Marrow Dendrit ic Cell (BMDC) RNA-Seq libraries. BMDCs are an attractive system for studying single cell responses since they are primary, post-mitotic, and, in response to pathogenic components, elicit robust, physiologically relevant transcriptional programs for inflammatory and antiviral cytokines that are well-characterized at the population level (Amit, 1. et al. Unbiased reconstruction of a mammalian transcriptional network mediating pathogen responses. Science 326, 257-263, doi: 10.1 126/science. l 179050 (2009).; Chevrier, N. et al. Systematic Discovery of TLR Signaling Components Delineates Viral-Sensing Circuits. Cell 1 47, 853-867, doi: 10. 1016/j.cell.201 1 .10.022 (201 1 ); Garter, M. et al. A High-Throughput Chromatin Immunoprecipitation Approach Reveals Principles of Dynamic Gene Regulation in Mammals. Molecu lar Cell 47, 81 0-822,

doi: 10.1016/j.molcel.20 12.07.030 (2012); Takeuchi, O. & Akira, S. Pattern Recognition Receptors and Inflammation. Cell 140, 805-820, doi: 10.1016/j.cell.2010.01 .022 (2010); Shalek, A. K. et al. Nanow ire- mediated delivery enables functional interrogation of primary immune cells: application to the analysis of chronic lymphocytic leukemia. Nano Lett. 12( 1 2):6498-504. doi: 10.1021/nl3042917 (2012)). Initially, BMDCs were profiled pre- stimulation and at four time points ( l ,2,4,6h) after stimulation with LPS, PAM3CS , and poly C (resp. from gram-negative bacteria, gram-positive bacteria and a synthetic mimic of viral RNA). From these distinct snapshots, the temporal and response-specific structures of single cell noise were examined. To assess changes in single cell variation across stimuli and time points, a new nested statistical model was developed and used to parameterize the single cell expression distributions of each gene. While each pathogen component activates a distinct temporal program at the population level, individual responding cells display dramatically variable behaviors also within each response. In inflammatory circuits, two temporally distinct patterns of expression heterogeneity were found: some circuits are strongly synchronized early and de-phase over time, whereas others are noisily induced. Antiviral gene circuits, meanwhile, onset noisily and become tightly synchronized over time.

[00086] In particular, the studies presented herein discovered a rare population of precocious "early anti-viral responders", masked in population measurements, and hypothesize that their response is amplified throughout the population via paracrine signaling. To test this hypothesis, each cell was stimu lated individually in a sealed micro fluidic chamber, and it was found that most cells fail to induce key antiviral response genes. Surprisingly, however, the inflammatory response is less variable in these isolated cells, demonstrating that intracellular communication can both restrict and increase noise for different circuits. Analyzing DCs lacking the interferon receptor recapitulates many of these findings, showing that interferon feedback in essential for coordinating the antiviral response as well as for cross-inhibition and noise in the inflammatory response. Finally, DCs deleted for key intracellular regulators nominated by the model were tested to verify key circuit component controlling this process. This study demonstrates how to harness variability across single cells for reconstructing inter- and intracellular circu its, and for understanding of cellular decision-making on a genomic scale. [00087] The compositions and methods of the disclosure use a use a microfl idics- based approach to prepare over 1 ,700 SMART-Seq singlecell RNA-Seq libraries, sampling the dynamic response of BMDCs to different pathogen components and related

perturbations. (See e.g., Ramskold, D. et al. Full-length mRNA-Seq from single-cell levels of RNA and individual circulating tumor cells. Nature B iotechnology 30, 777-782, doi: 10.1038/nbt.2282 (2012). Distinct gene modules are characterized by different temporal variability profiles, arising from changes in both the fraction of cells that express a given mRNA transcript at a detectable level and the mRNA levels within these detectably expressing cells. The average temporal response of the BMDC population arises from an underlying asynchronous, yet continuous, process at the single-cell level: at each sampled time point, and for each module, some cells are more 'advanced' than others on the temporal continuum. In particular, a few "precocious" cells were discovered, masked in population measurements, that produce interferon and activate a core antiviral module early. Without intending to be bound by theory, it is believe that these precocious cells are responsible for driving the antiviral response in the population through interferon-mediated paracrine signaling.

[00088] To understand the role of paracrine signaling in coordinating the population response, the studies provided herein developed a new experimental approach to stimulate cells individually in sealed microfluidic chambers, preventing cell-to-cell communication. This blocks the spread and coordination of the antiviral response at later time points, suggesting that these "precocious" cells play a crucial role in initiating and coordinating the native population response. Furthermore, it was found that BMDCs deficient for interferon receptor, or treated with a secretion inhibitor (Brefeldin A, 'GolgiPlug') or a protein synthesis inhibitor (Cycloheximide), failed to induce "core" antiviral response genes when they were stimulated with LPS. Surprisingly, inhibiting paracrine signaling or just interferon signaling also resulted in a significant increase in the fraction of cells expressing an inflammatory response gene modu le with an early, sharply peaked induction pattern, highlighting how dynamic population-level positive and negative paracrine feedback loops can both promote and restrain variation in the immune response.

[00089] The behavior of individual cells within BMDC populations is highly dynamic during the immune response, with both digital and analogue variation changing across various time points, stimuli and modules. These patterns - masked in population- level measurements - reveal princip les for how a cell population can use both intra- and inter-ccllular control strategies to coordinate a complex dynam ic response. The single-cell profiling data sets presented here, obtained in d ifferent time points and stimuli, and the associated statistical analyses, and physical, genetic and biochemical perturbations, provide essential input and approaches for dissecting these intra- and intercellular control strategies.

[00090] First, the statistical analysis of single-cell expression distributions reveals that during a dynamic response both the fraction of cells expressing a particu lar transcript at a detectable level as well as the mRNA levels within expressing cells change. The interaction of these two functions can encode a rich diversity of temporal response profiles. For example, late-induced "core" antiviral genes exhibit very weak average expression at early time points, but are highly expressed in a few "precocious" cells. In contrast, the progressive dampening of "peaked" inflammatory genes reflects changes in the fraction of cells expressing these transcripts, rather than a uniform gradual decrease in the expression in all cells. The ubiquity of this behavior challenges conventional computational approaches for circuit reconstruction that tend to implicitly attribute the changes in population expression profiles solely to intra-cellular events. Rather, these observations suggest that cell populations can generate complex average responses not only through intricate intracellular circuits, which are common to all cells, but also with inter-cellular feedback mechanisms between heterogeneous single cells. The early changes in bimodality which characterize multiple response programs (Fig. 26f) could suggest that the most efficient way to generate rapid immune responses is to ask more cells to perform a given task rather than to ask any cell to perform it more efficiently.

[00091] One example of the importance of such inter -cellular control strategies is the find ing that paracrine signaling plays a crucial role in establishing several distinct temporal patterns of single-cell behavior. In particular, the studies herein have uncovered a small number of "precocious" cells that express Ifnb l and "core" antiviral genes as early as l h after LPS stimulation, and through the secretion of IFN-β, help activate "core" antiviral genes in other cells to coordinate the population response. It is noted that these cells are not distinguishable from the rest of the population, except for expression of the approximately one hundred genes in the "core" antiviral module.

[00092] The experimental data presented herein do suggest that the "precocious" cells that were observed are likely to be primed initiators that are crucial in enabling the efficient, and timely, popu lation response. First, the Brefeldin A (GolgiPlug) experiment inhibiting secretion at different time points after addition of LPS suggests that the key paracrine signal acting on the "core" antiviral response is secreted early, around l h. More importantly, the "on-chip" isolation experiment shows that, without paracrine signaling from these "precocious" cells, only a small portion (20%) of cells can initiate a diminished "core" antiviral response to LPS by themselves even after 4h of incubation. These data therefore suggest that the "precocious" cells may represent cells in a special, possibly stochastically defined, epigenetic state that are primed to express Ifnbl in response to LPS as early as l h. Paracrine signaling, including interferon-mediated communication, also acts to dampen a subset of induced genes ("peaked" inflammatory) at later time points. Taken together, these observations suggest a model (Fig. 29) for the cross- inhibit ion between the antiviral and inflammatory pathways that was observed in "on-chip", knockout and chemical modulatory experiments. In this model, anti-viral feedback from a small number of cells induces the expression and secretion of anti-inflammatory cytokines from a subset of cells, which, in turn, attenuate the inflammatory responses of nearby cells. Importantly, this model also suggests alternative therapeutic strategies that target the balance between distinct response subsets rather than presenting uniform excess extracellular signaling molecules (e.g., IFN-β) (see e.g., Banchereau, J. & Pascual, V. Type I Interferon in Systemic Lupus Erythematosus and Other Autoimmune Diseases. Immunity 25, 383-392, doi:http://dx.doi.org/10.1016/j. immuni.2006.08.010 (2006); Hall, J. C. & Rosen, A. Type I interferons: crucial participants in disease amplification in autoimmunity. Nature Reviews Rheumatology 6, 40-49, doi:http://dx.doi.org/10.1038/nrrheum.2009.237 (2010)).

Automated Procedu re for Selection of Signature Genes

[00093] The invention also provides methods of determining gene signatures that are useful in various therapeutic and/or diagnostic indications. The goal of these methods is to select a small signature of genes that will be informative with respect to a process of interest. The basic concept is that different types of information can entail different partitions of the "space" of the entire genome (>20k genes) into subsets of associated genes.

This strategy is designed to have the best coverage of these partitions, given the constraint on the signature size. For instance, in some embodiments of this strategy, there are two types of information: (i) temporal expression profiles; and (ii) functional annotations. The first information source partitions the genes into sets of co-expressed genes. The information source partitions the genes into sets of co-functional genes. A small set of genes is then selected such that there are a desired number of representatives from each set, for example, at least 10 representatives from each co-expression set and at least 10 representatives from each co-functional set. The problem of working with multiple sources of information (and thus aiming to "cover" multiple partitions) is known in the theory of computer science as Set-Cover. While this problem cannot be solved to optimality (due to its NP-hardness) it can be approximated to within a small factor. In some embodiments, the desired number of representatives from each set is one or more, at least 2, 5 or more, 10 or more, 15 or more, 20 or more, 25 or more, 30 or more, 35 or more, 40 or more, 50 or more, 60 or more, 70 or more, 80 or more, 90 or more, or 100 or more.

[00094] An important feature of this approach is that it can be given either the size of the signature (and then find the best coverage it can under this constraint); or the desired level of coverage (and then select the minimal signature size that can satisfy the coverage demand).

[00095] An exemplary embodiment of this procedure is the selection of the various gene signatures presented in Tables 1 , 1 A, 2, 2A, 3, 3 A, 4, 4A, 5 and/or 5 A.

Use of Signature Genes

[00096] The invention provides dendritic cell related gene signatures for use in a variety of diagnostic and/or therapeutic indications, as well as in a variety of methods of screening for or otherwise identifying therapeutic molecules. "Signatures" in the context of the present invention encompasses, without limitation nucleic acids, together with their polymorphisms, mutations, variants, modifications, subunits, fragments, and other analytes or sample-derived measures.

[00097] Exemplary signatures are shown in Tables 1 , 1A, 2, 2A, 3, 3A, 4, 4A, 5 and

5A and are collectively referred to herein as, inter alia, "dendritic cell-associated genes," "dendritic cell-associated nucleic acids," "signature genes," or "signature nucleic acids."

[00098] These signatures are useful in methods of diagnosing, prognosing and/or staging an immune response and/or aberrant dendritic cell response in a subject by detecting a first level of expression, activity and/or function of one or more signature genes or one or more products of one or more signature genes selected from those listed in Tables 1 , 1 A, 2, 2A, 3, 3 A, 4, 4A, 5 and/or 5 A and comparing the detected level to a control of level of signature gene or gene product expression, activity and/or function, wherein a difference in the detected level and the control level indicates that the presence of an immune response and/or aberrant dendritic cell response in the subject.

[00099] These signatures are useful in methods of monitoring an immune response and/or aberrant dendritic cell response in a subject by detecting a level of expression, activity and/or function of one or more signature genes or one or more products of one or more signature genes selected from those listed in Tables 1. 1 A, 2, 2A, 3, 3 A, 4, 4A, 5 and/or 5A at a first time point, detecting a level of expression, activity and/or function of one or more signature genes or one or more products of one or more signature genes selected from those listed in fables 1 , 1A, 2, 2A, 3, 3A, 4, 4A, 5 and/or 5A at a second time point, and comparing the first detected level of expression, activity and/or function with the second detected level of expression, activity and/or function, wherein a change in the first and second detected levels indicates a change in the immune response and/or aberrant dendritic cell response in the subject.

[000100] These signatures are useful in methods of identifying patient populations at risk or suffering from an immune response, e.g. , an aberrant immune response, an autoimmune response, and/or an inflammatory response, and/or aberrant dendritic cell response based on a detected level of expression, activity and/or function of one or more signature genes or one or more products of one or more signature genes selected from those listed in Tables 1 , 1Λ, 2, 2Λ, 3, 3A, 4, 4A, 5 and/or 5A. These signatures are also useful in monitoring subjects undergoing treatments and therapies for aberrant immune response(s) and/or aberrant dendritic cell response(s) to determine efficaciousness of the treatment or therapy. These signatures are also useful in monitoring subjects undergoing treatments and therapies for aberrant immune response(s) and/or aberrant dendritic cell response(s) to determine whether the patient is responsive to the treatment or therapy. These signatures are also useful for selecting or modifying therapies and treatments that would be efficacious in treating, delaying the progression of or otherwise ameliorating a symptom of an aberrant immune response and/or aberrant dendritic cell response. The signatures provided herein are useful for selecting a group of patients at a specific state of a disease with accuracy that facilitates selection of treatments.

[000101] These signature genes are also useful in methods of monitoring patient response to a therapy, vaccine, transplant or other therapeutic intervention. For example, the expression level of one or more signature genes can be detected at a variety of timepoints pre- and post-administration, and these levels can be analyzed using the single cell methods provided herein. By determining which genes are being expressed in cohorts or other coherent groups and/or which subpopulations of cells are exclusively expressing these genes, a practitioner will be able to determine which cohort(s) and/or which pathway(s) are responsible for generating an immune response and/or an aberrant dendritic cell response. [000102] The present invention also comprises a kit with a detection reagent that binds to one or more signature nucleic acids. Also provided by the invention is an array of detection reagents, e.g. , oligonucleotides that can bind to one or more signature nucleic acids. Suitable detection reagents include nucleic acids that specifically identify one or more signature nucleic acids by having homologous nucleic acid sequences, such as

oligonucleotide sequences, complementary to a portion of the signature nucleic acids packaged together in the form of a kit. The oligonucleotides can be fragments of the signature genes. For example the oligonucleotides can be 200, 150, 1 00. 50, 25, 10 or fewer nucleotides in length. The kit may contain in separate container or packaged separately with reagents for binding them to the matrix), control formulations (positive and/or negative), and/or a detectable label such as fluorescein, green fluorescent protein, rhodamine, cyanine dyes, Alexa dyes, luciferase, radiolabels, among others. Instructions (e.g. , written, tape, VCR, CD-ROM, etc.) for carrying out the assay may be included in the kit. The assay may for example be in the form of a Northern hybridization or a sandwich ELISA as known in the art. The kit may for example include reagents and instructions for carrying out any of the methods described herein, including PCR, nucleic acid sequencing, etc. Alternatively, the kit contains a nucleic acid substrate array comprising one or more nucleic acid sequences.

Dendritic Cells and Uses Thereof

[000103] Dendritic cells (DCs) are involved in a number of immune responses including and/or contributing to resistance to infection and modulating tolerance to self. DCs have the capacity to control T-cell recognition and/or responsiveness.

[000104] DCs are known to induce resistance to infection, as they mature in distinct ways in response to different pathogens, e.g., microbial components, and can therefore initiate different host immunity responses. (See e.g. , Steinman & Banchereau. "Taking dendritic cells into medicine." Nature, vol. 449: 419-426 (2007); doi: 10.1038/nature06175). The modulating agents provided herein can be used to disrupt these immune responses. For example, the modulating agents modulate the expression, activity, and/or function of one or more genes from Tables 1 -5A. In some embodiments, these modulating agents block or otherwise inhibit DC maturation. In some embodiments, these modulating agents alter or otherwise influence one or more functions of DCs, thereby modulating T-cell responses, for example, from a protective Τ _Μ 1 phenotype to a non-protective T _f i2 phenotype. [000105] DCs are also useful in the design and creation of a variety of vaccine indications to treat and prevent infection by enhancing immunogenesis. In these indications, the vaccine can include one or more modulating agents. For example, in some

embodiments, the vaccine delivers a modulating agent that controls or otherwise influences dendritic cell maturation. In some embodiments, the vaccine delivers a modulating agent that alters or otherwise influences one or more T-cell responses, for example, induction of the protective Tnl phenotype.

[000106] DCs are also useful in the design and creation of a variety of therapeutic vaccines against cancer due to their capacity to regulate T cell immunity (see e.g. ,

Banchereau & Palucka. Dendritic Cells as Therapeutic Vaccines Against Cancer. Mature, vo l. 5: 296-306 (2005); doi: 10.1038/nril592); see also, Palucka et al. "Building on dendritic cell subsets to improve cancer vaccines." Curr Op Immunol, 22: 258-63 (2010);

doi: 10.1016/j.coi.2010.02.010).

[000107] For example, DCs are used as adjuvants in the vaccines. Immature DCs are known to induce tolerance, while mature DCs induce immunity. Immature DCs function mainly as antigen-capturing cells, while mature DCs mainly function as antigen-presenting cells. Thus, the modulating agents can be used to modulate the maturity of a DC or population of DCs, for example, to shift the balance between mature and immature DCs based on the desired outcome. For example, the modulating agent can be used to shift toward an immature DC phenotype where tolerance is desired, and in indications where immunity is desired, the modulating agent can be used to shift toward a mature DC phenotype. In some embodiments, the modulating agent is used to modulate the plasticity of a DC or population of DCs. For example, the modulating agent can be used to shift a DC or population of DCs toward a particular subset of DCs, e.g.. toward or away from Langerhans cells, interstitial DCs and plasmacytoid DCs; or toward a particular pathway of DC differentiation, e.g. , toward or away from the myeloid pathway and/or toward or away from the lymphoid pathway.

[000108] The invention provides compositions and methods for modulating one or more dendritic cell responses. As used herein, the term "modulating" includes up-regulation of, or otherwise increasing, the expression of one or more genes, down-regulation of, or otherwise decreasing, the expression of one or more genes, inhibiting or otherwise decreasing the expression, activity and/or function of one or more gene products, and/or enhancing or otherwise increasing the expression, activity and/or function of one or more gene products.

[000109] As used herein, the term "modulating a response of dendritic cells" includes the modulation of any of a variety of dendritic cell functions and/or activities, includ ing by way of non-lim iting example, controlling or otherw ise influencing the networks that regulate dendritic cell maturation; controlling or otherwise influencing the networks that regulate an immune response of a dendritic cell; controlling or otherwise influencing the networks that regulate an antiviral immune response of a dendritic cell, for example, an antiviral immune response of a dendritic cell including a core antiviral response and/or a secondary antiviral response; controlling or otherwise influencing the networks that regulate an inflammatory immune response of a dendritic cell, for example, an induced

inflammatory response and/or a sharped peak inflammatory response; controlling or otherwise influencing the networks that regulate a Toll-like receptor (TLR) response of dendritic cells; controlling or otherwise influencing the networks that regulate T cell and B cell recruitment; controlling or otherwise influencing the networks that regulate DC promotion of TV -cell response(s); controlling or otherwise influencing the networks that regulate DC induction of Tn2-cell response(s); controlling or otherwise influencing the networks that regulate DC induction, impact or other effect on any cell that is downstream of the D; controlling or otherwise influencing the networks that regulate DC induction of T cells including regulatory T cells (Tregs), Thl 7 cells, memory T cells and other T cells; controlling or otherwise influencing the networks that regulate a shift in a DC phenotype, for example, between a mature and immature phenotype and/or between subsets of DCs; manipulating or otherwise influencing at least one function or biological activity of a dendritic cell; manipulating or otherwise influencing dendritic cell control of pathogen- drive T cell polarization; and/or manipulating or otherwise influence the production of cytokines, chemokines and other molecules secreted by the DC.

[000110] The invention provides modulating agents that modulate one or more dendritic cell response(s). Suitable modulating agents include an antibody, a soluble polypeptide, a polypeptide agent, a peptide agent, a nucleic acid agent, a nucleic acid ligand, or a small molecule agent.

[0001 11 ] In some embodiments, the modulating agent is used to modulate the expression of one or more genes from the ''Core Antiviral" gene signature, e.g. , one or more genes from those listed in Tables 1 and 1 A. These modulating agents are referred to herein as "core antiviral modulating agent(s)."

[000112] For example, in some embodiments the core antiviral modulating agent is a kinase, such as, by way of non-limiting example, a kinase selected from the group consisting of: MAPKl , EIF2AK2, TBKl , PLK4, IKBKE, PLK2, MAP3K7, CHUK, JAKl , CRKL, MKNK2, TYK2, RPS6KB2, 1KB KB, MKNK1 , NEK7, PIK3R2, IKBKG, RIPK2, MAP2K6, MET, RPS6KB 1 , MARK2, DGKA, and BUB I B.

[000113] For example, in some embodiments, the core antiviral modulating agent is a transmembrane receptor, such as, by way of non- limiting example, a transmembrane receptor selected from the group consisting of: IFNAR1, TLR3, TLR4, IL28RA, TLR9, IFNAR2, COLEC 12, SCARA3, MSR 1 , FCER1 G, and KIR2DS4.

[000114] For example, in some embodiments, the core antiviral modulating agent is a mammalian endogenous chemical drug, such as, by way of non-limiting example, tretinoin, or a non-mammalian endogenous chemical drug such as, by way of non- limiting example, a non-mammalian endogenous chemical drug selected from the group consisting of:

salmonella minnesota R595 lipopolysaccharides, mezerein, 3-deoxy-2-octulosonic acid(2)- lipid A, E. coli B5 lipopolysaccharide, and bafilomycin Al .

[000115] For example, in some embodiments, the core antiviral modulating agent is a chemical drug, such as, by way of non-limiting example, a chemical kinase inhibitor drug such as SB203580 or H-7, or another chemical drug such as a chemical reagent, toxicant or other chemical drug selected from the group consisting of: lipopolysaccharide, poly rI:rC- RNA, E. coli B4 lipopolysaccharide, stallimycin, bromodeoxyuridine, 2-aminopurine, ribavirin, CpG ODN 1668, pristane, imiquimod, decitabine, Salmonella enterica serotype abortus equi lipopolysaccharide, CpG ODN 1826, concanamycin A, poly dA-dT, ionomycin, fiicoidin, CpG ODN 221 6, AL 108, 4,4'-diaminodiphenylmethane, epigallocatechin-gallate, chloroquine, 3M-01 1 , carbimazole, 3M-001 , Pam3-Cys, rosiglitazone, and lipid A.

[000116] For example, in some embodiments, the core antiviral modulating agent is a biologic drug, such as, by way of non-limiting example, a biologic drug selected from the group consisting of: pegintron, fontolizumab and interferon beta-la.

[000117] In some embodiments, the modulating agent is used to modulate the expression of one or more genes from the "Secondary Antiviral" gene signature, e.g., one or more genes from those listed in Tables 2 and 2 A. These modu lating agents are referred to herein as "second antiviral modulating agents."

[000118] For example, in some embodiments the secondary antiviral modulating agent is a kinase, such as, by way of non-limit ing example, a kinase selected from the group consisting of: MAPK9, EIF2AK2, CRKL, MET, TBK1 , MAP3K7, and JA 1 .

[000119] For example, in some embodiments, the secondary antiviral modulating agent is a transmembrane receptor, such as, by way of non-limiting example, a

transmembrane receptor selected from the group consisting of: TLR4, TLR3, and 1FNAR2.

[000120] For example, in some embodiments, the secondary antiviral modulating agent is a non-mammalian endogenous chemical drug such as, by way of non-limiting example, salmonella minnesota R595 lipopolysaccharides.

[000121] For example, in some embodiments, the secondary antiviral modulating agent is a chemical drug, such as, by way of non-limiting example, a chemical kinase inhibitor drug such as LFM-A13, or another chemical drug such as a chemical reagent, toxicant or other chemical drug selected from the group consisting of: poly rI:rC-RNA, lipopolysaccharide, and R-W1N 55,212.

[000122] In some embodiments, the modulating agent is used to modulate the expression of one or more genes from the "Maturation" gene signature, e.g. , one or more genes from those listed in Tables 3 and 3A. These modulating agents are referred to herein as "maturation modulating agents."

[000123] For example, in some embodiments the maturation modulating agent is a kinase, such as, by way of non-limiting example, a kinase selected from the group consisting of: 1KB KB, MAP2K4, PRKCD, MTOR, MAPKAPK2, PRKCB, LYN, MAPK14, DDR1 , TGFBR1 , PRKCA, AKT1 , RAF1 , SHC 1 , CSF1R, IRAK4, PRKCQ, SPHK1 , MAP4K1 , RPS6KB 1 , GSK3B, FES, MAP3K7, MAP3K8, SRC, CHUK, PTK2, PIK3R1 , MAP2K7, MAPK9, RPS6KA5, MAPK8, BTK, EGFR, ΜΛΡ2Κ6, PDPK1 , PRKG 1 , FLT3, TYK2, CDK9, ACVR2B, CDK10, MAST2, MAPK1 1 , FGFR3, P1M 1 , ACVRL 1 , FGFR2, MARK2, PBK. PLK3, MAP3K14, NME 1 , H1PK2, and ERBB2.

[000124] For example, in some embodiments, the maturation modulat ing agent is a transmembrane receptor, such as, by way of non-limiting example, a transmembrane receptor selected from the group consisting of: CD40, TLR4, TLR9, FAS, TLR7, CD5, IL27RA, TLR2, TLR3, CD28, ICAMl . LTBR, TLR8, FCGR2A, TYROBP, TNFRSFI OA, TLR5, TREM2, IGHM, CD2, TNFRSF8, 1L6R, CLEC7A, CHRNA1 , JTGB3, AGER, TNFRSF6B, TLR6, TNFRSF 1 1 A, TRA@ (also known as TRA, T cell receptor alpha locus), FCGR2B, NGFR, 1GF1 R, TNFRSF1 A, 1L 1 RL2, CD300C, CD86, and MS4A2.

[000125] For example, in some embodiments, the maturation modulating agent is a mammalian endogenous chemical drug such as, by way of non-limiting example, a mammalian endogenous chemical drug selected from the group consisting of prostaglandin E2, hyaluronic acid, ATP, tretinoin, ethanol, hydrogen peroxide, butyric acid, arachidonic acid, uric acid, chondroitin sulfate A, adenosine, heparin, Ca2+, histamine, L-methionine, carbon monoxide, cyclic AMP, lauric acid, epinephrine, 1 1 , 12-epoxyeicosatrienoic acid, beta-estradio 1, lipoxin A4, L-glutamic acid, dihydrotestosterone, progesterone, kynurenic acid, mevalonic acid, 5,6-epoxyeicosatrienoic acid, L-ornithine, malonic acid, elaidic acid, N(omega)-hydroxyarginine, dimethylglycine, 17-epiestriol, D-galactosamine, hydrocortisone, folic acid, hemin, glucosamine, platelet activating factor,

glycosylphosphatidylinosito l, palmitoleic acid, and glutathione.

[000126] For example, in some embodiments, the maturation modulating agent is a non-mammalian endogenous chemical drug such as, by way of non-limiting example, a non-mammalian endogenous chemical drug selected from the group consisting of E. coli lipopolysaccharide, lipoteichoic acid, E. coli B5 lipopolysaccharide, N-acetylmuramyl-L- alanyl-D-isoglutamine, zymosan A, 15-deoxy-delta- 12, 14 -PGJ 2, peptidoglycan, ursolic acid, ganglioside GD3, zymosan, hemozoin, prostaglandin A l , mezerein, E. coli serotype 0127B8 lipopolysaccharide, salmonella minnesota R595 lipopolysaccharides, ricinoleic acid, tunicamycin, and apigenin.

[000127] For example, in some embodiments, the maturation modulating agent is a chemical drug, such as, by way of non-limiting example, a chemical kinase inhibitor drug such as SB203580, wortmannin, PD98059, SP600125, Sb202190, U0126, LY294002, AG490, KN 93, bisindolylmaleimide I, Ro31-8220, staurosporine, Bay 1 1 -7082, H89, Go 6976, tyrphostin AG 1478, PD 169316, PP 1 , 8-bromoguanosine 3',5'-cyclic

monophosphate, l -o-hexadecyl-2-o-methyl-rac-glycei l, myristoylated P Czeta pseudosubstrate peptide inhibitor, T 5926, and 8-chlorophenylthio-adenosine 3',5'-cyclic monophosphate.

[000128] For example, in some embodiments, the maturation modulating agent is a chemical drug, such as, by way of non-limiting example, another chemical drug such as a chemical reagent, chemical toxicant or other chemical drug selected from the group consisting of: lipopolysaccharide, ssRNA40, N-nitro-L-arginine methyl ester, caffeic acid phenethyl ester, S-nitrosoglutathione, W7, E. coli B4 lipopolysaccharide, phorbol myristate acetate, CpG ODN 2006, CpG ODN 1826, poly rI:rC-RNA, ATP-gamma-S, simvastatin, EGTA, nystatin, N-acetyl-L-cysteine, 3M-001 , tranilast, thapsigargin, Pam3-Cys-Ser-Lys4, DETA-NONOate, resiquimod, CpG ODN 1668, Salmonella enterica serotype abortus equi lipopolysaccharide, 3-methyladenine, murabutide, CpG oligonucleotide, R5020, lovastatin, sirolimus, bucladesine, epigallocatechin-gallate, melphalan, 3M-01 1 , imatinib, zVAD- FM , Pam3-Cys, aspirin, bleomycin, dexamethasone, sanglifehrin A, methoxsalen, bortezomib, camptothecin, monophosphoryl lipid A, 3M-002, paclitaxel, pyrrolidine dithiocarbamate, nickel, trichostatin A, docosahexaenoic acid, curcumin, dextran sulfate, resveratrol, forskolin, suramin, pristane, 7-ethyl- l O-hydroxy-camptothecin, Ni2+, trovafloxacin, phenanthridine, bryostatin 1 , UCN-01 , vinblastine, etoposide, cycloheximide, oxaliplatin, [Lys l 5,Arg l 6,Leu27]VIP( l -7)GRF(8-27), fluvastatin, ciglitazone, nicotine, eicosapentenoic acid, rosiglitazone, ionomycin, pentoxifylline, niflumic acid, [Ac-His l ,D- Phe2,Lys l 5,Arg l 6,Leu27]VIP-(3-7)-GRF-(8-27), mifepristone, gliotoxin, f avopiridol, tanespimycin, rotenone, GCS- 100, midazo lam, 1 -alpha, 25-dihydroxy vitamin D3, decitabine, 3,3'-diindolylmethane, A23 1 87. cntinostat, zidovudine, cytidylyl-3'-5'- guanosine, tetrandrine, valpro ic acid, cisplatin, toremifene, quinacrine, vitamin E, vorinostat, GW3965, isobutylmethylxanthine, fulvestrant, Sn50 peptide, clobetasol propionate, D609, benzene, epothilone B, spermine nitric oxide complex, methylselenic acid, deferoxamine, troglitazone, 1 '-acetoxychavicol acetate, paricalcitol, arsenic, imiquimod, GLP-l-(7-34)-amide, S-(2,3-bispalmitoyloxypropyl)-cysteine-GDPKHP SF, 9-cis-retinoic acid, cadmium, sulindac sulfide, rottlerin, 13-cis-retinoic acid, nitrofurantoin, N-Ac-Leu-Leu-norleucinal, dacinostat, Ro41 -5253, tosylphenylalanyl chloromethyl ketone, raloxifene, cerivastatin, panobinostat, fisetin, trinitrobenzenesulfonic acid, CpG ODN 2216, ochratoxin A, azoxymethane, epicatechin gal late, phorbol esters, MALP-2s, S-nitroso-N- acetyl-DL-penicillaniine, rolipram, lactacystin, reactive oxygen species, carbon tetrachloride, phorbol 12. 13-didecanoate, polyethylene glycol, diisopropanolnitrosamine, N(l )-guanyl- l ,7-diaminoheptane, aldesleukin, 4-hydroxytamoxifen, thalidomide, doxorubicin, sulforafan, methylnitronitrosoguanidine, SU6656, CGS 21680, daunorubicin, omega-N-mcthylarginine, linsidomine. fasudil. 5-fluorouracil, diethylstilbestroi, morphine, mitomycin C, ribavirin, S-nitroso-N-acetylpenicillamine, sodium orthovanadate, Am 580, prednisolone, chloroquine, galactosylceramide-alpha, gemcitabine, 9, 10-dimethyl- l ,2- benzanthracene, BAPTA-AM, methylprednisolone, indomethacin, CP-55940, docetaxel, memantine, arbutin, moxestrol, 2,2,2-trichloroelhanol, daniisertib. anastrozole. perifosine, bisphosphonate, mefenamic acid, glutathione ethyl ester, vinflunine, polyinosinic acid, sparfosic acid, retinoid, vincristine, phcnacetin, lipid A, dimethylnitrosamine, genistein, 2- deoxyglucose, pioglitazone, 06-benzylguanine, beryllium sulfate, benzo(a)pyrene 7,8- dihydrodiol, methylamphotericin B, riociguat, 0-(chloroacetylcarbamoyl)fumagillo l, dephostatin, atrasentan, tipifamib, bongkrekic acid, natamycin, 10-decarbamoylmitomycin C, phenoxodiol, potassium cyanide, 3,4-inethylenedioxyamphetamine, (-)-gallocatcchin gallate, lbeta,25-dihydroxyvitamin D3, 17-alpha-ethinylestradiol, salicylic acid, 3- deazaneplanocin, and doxycycline.

[000129] For example, in some embodiments, the maturation modulating agent is a biologic drug, such as, by way of non-limiting example, a biologic drug selected from the group consisting of: cyclosporin A, hemocyanin, etancrccpt, cntcrotoxin B, romidepsin, adalimumab, interferon beta- l b, atosiban, and defibrotide.

[000130] In some embodiments, the modulating agent is used to modulate the expression of one or more genes from the "Peaked Inflammatory" gene signature, e.g., one or more genes from those listed in Tables 4 and 4A. These modulating agents are referred to herein as "peaked inflammatory modulating agents."

[000131] For example, in some embodiments the peaked inflammatory modulating agent is a kinase, such as, by way of non-limiting example, a kinase selected from the group consisting of: IRAK4, CHUK, IKBKG, I BKB, MAP2K 1 , MAR 2, MAP3K 14, TBK 1 , IRAK3, TGFBR2, LYN, EIF2AK2, MAPK8, KIT, RIPK2, PRKCA, CDK9, SPHK1 , PRKCD, EGFR, MAP3K7, TXK, MAP3K8, ΜΑΡΚΛΡΚ2, MAPK10, IRAK2, IKBKE, RAF 1 , JAK2, ADRBKl , TEK, MAPK9, MET, MAPK14, ITK, BMPR2, FLT3, PRKDl , TYK2, PRKCQ, MERTK, MAPK1 , AKT2, MAPKAPK5, JAK1 , and PIK3CG.

[000132] For example, in some embodiments, the peaked inflammatory modulating agent is a transmembrane receptor, such as, by way of non-limiting example, a

transmembrane receptor selected from the group consisting of: TLR4, IL28RA, IFNARl , FAS, TLR7, CD14, TLR3, TNFRSF1A, TLR5, CD40, ICAM1 , TLR9, SIGIRR, MSR1 , ILI ORA, FCGR2B, FCGR2A, I L27RA, TLR2, CD28, PLAUR, MARCO, UNC5B, THBD, IFNGR 1 , ILI ORB, CD86, IL 1 R1 , FCGR1 A, IL1 RL1 , IL6R, TNFRSF 18, RARRES2, TNFRSF 1 B, EPOR, TRA@, IL17RA, TRB@ (also known as TRB, T cell receptor beta locus), and CD300C. (000133j For example, in some embodiments, the peaked inflammatory modulating agent is a mammalian endogenous chemical drug, such as, by way of non-limiting example, a mammalian endogenous chemical drug selected from the group consisting of hyaluronic acid, beta-estradiol, prostaglandin E2, uric acid, neuroprotectin Dl , platelet activating factor, stearic acid, tretinoin, palmitic acid, progesterone, D-sphingosine, spermine, hydrogen peroxide, leukotriene D4, hydrocortisone, lauric acid, fatty acid, 1 1 , 12- epoxyeicosatrienoic acid, chenodeoxycholic acid, linolenic acid, ATP, lithocholic acid, lipid, arachidonic acid, aldehyde, methyl palmitate, L-cystine, L-tartaric acid, arginine, butyric acid, D-glucose, L-ornithine, 1,4-glucan, taurolithocholic acid,

globotriaosylceramide, cerotic acid, D-erythro-C 16-ceramide, dimethylglycine, 22(R)- hydroxycho lesterol, L-triiodothyronine, mevalonic acid, alcoho l, beta-carotene, and D- galactosamine.

[000134] For example, in some embodiments, the peaked inflammatory modulating agent is a non-mammalian endogenous chemical drug such as, by way of non- limiting example, a non-mammalian endogenous chemical drug selected from the group consisting of: salmonella minnesota R595 lipopo lysaccharides, E. coli B5 lipopolysaccharide, zymosan, N-acetylmuramyl-L-alanyl-D-isoglutamine, E. coli serotype 0127B8

lipopolysaccharide, lipoteichoic acid, E. coli l ipopolysaccharide, peptidoglycan, mezerein, mannan, carrageenan, ubiquinone 9, brefeldin A, polyamines, mannosylated

lipoarabinomannan, isoquercitrin, cyclomaltodextrin, cyclopiazonic acid, 2- mercaptoacetate, bafilomycin Al , hemozo in, lipoarabinomannan, MALP-2R, Silybum marianum extract, polysaccharide, 15-deoxy-delta-12, 14 -PCI 2, phorbol 12, 13-dibutyrate, syringin, isobutylamine, and glucuronoxylomannan.

[000135] For example, in some embodiments, the peaked inflammatory modulating agent is a chemical drug, such as, by way of non-limiting example, a chemical kinase inhibitor selected from the group consisting of SP600125, U0126, SB203580, LY294002, PD98059, PP 1 , wortmannin, Bay 1 1 -7082, Go 6976, PS- 1 145, JAK inhibitor 1, merck C, bisindolylmaleimide I, and tyrphostin B56.

[000136] For example, in some embodiments, the peaked inflammatory modulating agent is another chemical drug such as a chemical reagent, toxicant or other chemical drug selected from selected from the group consisting of: lipopo lysaccharide, Salmonella enterica serotype abortus equi lipopolysaccharide, trovafloxacin, resiqu imod, dexamethasone, cyclo heximide, trinitrobenzenesulfonic acid, ALP-2s, E. coli B4 lipopolysaccharide, poly rl .-rC-RNA, camptothecin, Pam3-Cys, Pam3-Cys-Ser-Lys4, CpG ODN 1668, CpG oligonucleotide, simvastatin, paclitaxel, genistein, phorbol myristate acetate, N-nitro-L- arginine methyl ester, triamcinolone acetonide, thapsigargin, picryl chloride, 1 -alpha, 25- dihydroxy vitamin D3, 5-N-ethylcarboxamido adenosine, pyrrolidine dithiocarbamate, ceruletide, magnesium sulfate, GW3965, cortisone acetate, ranitidine, rofluinilast, 3- methyladenine, Ni2+, dextran sulfate, glucocorticoid, epigallocatechin-gallate, ozone, gemfibrozil, triciribine, famotidine, tranexamic acid, grepafloxacin, acetaminophen, daidzein, bepafant, 1DN-6556, ZFA-fmk, BQ 123, pentoxifylline, zinc, chloroquine, alpha- tocopherol, triamcinolone hexacetonide, edaravone, rabeprazole, okadaic acid, CP-55940, ionomycin, caffeic acid phenethyl ester, Z-DEVD-FMK, polymyxin B, palmitoyl-Cys((RS)- 2,3-di(palmitoyloxy)-propyl)-Ala-G ly-OH, cytidylyl-3'-5'-guanosine, BQ-788, melphalan, N-acetyl-L-cystcine, stallimycin, 25-hydroxycholesterol, bucladesine, A23187, sunitinib, lactacysttn, actinomycin D, methylprednisolone, docosahexaenoic acid, SR 144528, vitamin E, clarithromycin, salmeterol, mevastatin, bromodeoxyuridine, CpG ODN 1826, monophosphoryl lipid A, 2,4-dinitrofluorobenzene, vorinostat, TO-901317, erythromycin, misoprostol, PD184352, diethylmaleate, ammonium chloride, l -methyl-4-phenyl- l ,2,3,6- tetrahydropyridine, bleomycin, alendronic acid, parthenolide, tosylphenylalanyl chloromethyl ketone, nifedipine, rosiglitazone, desipramine, ilomastat, nicotine, 13-cis- retinoic acid, trichostatin A, cis-urocanic acid, rosuvastatin, mycophenolic acid, cyclophosphamide, 8-bromo-cAMP, eicosapentcnoic acid, estrogen, oleoyl-estrone, 8- cyclopentyl- l,3-dipropylxanthine, carteolol, N-formyl-Nle-Leu-Phe, NSC 270012, dalcetrapib, MK2206, GSK21 1 8436, dexamethasone/tobramycin, deoxyspergualin, RP 48740, fosfomycin, NSC 95397, bacitracin, tirofiban, dexanabinol, rolipram, curcumin, diclofenac, N-formyl-Met-Leu-Phe, reactive oxygen species, omega-N-methylarginine, tacrolimus, pirinixic acid, valproic acid, thioacetamide, cisplatin, propylthiouracil, 5- azacytidine, galactosylceramide-alpha, diphosphoryl lipid A, gcntamicin C I , CpG ODN M362, PF-251 802, PF-4308515, GTS 21 , compound 48/80, vesnarinonc, glyoxal, 2,4- dinitrothiocyanatobenzene, enalaprilat, trehalose dimycolate, bis-pom-pmea, BAPTA-AM, resveratrol, S-nitrosoglutathione, lovastatin, and chloropromazine.

[000137] For example, in some embodiments, the peaked inflammatory modulating agent is a biologic drug, such as, by way of non-limiting example, a biologic drug selected from the group consisting of: cyclosporin A, enterotoxin B, lisinopril, abciximab, and eptifibatide. [000138] In some embodiments, the modulating agent is used to modulate the expression of one or more genes from the "Induced Inflammatory" gene signature, e.g. , one or more genes from those listed in Tables 5 and 5A. These modulating agents are referred to herein as "induced inflammatory modulating agents."

[000139] For example, in some embodiments the induced inflammatory modulating agent is a kinase, such as, by way of non-limiting example, a kinase selected from the group consisting of: CHUK, IKBKB, TBK1 , MAP2K1 , MAPK1 , LYN, IKBKG, MAP3 8, I BKE, MAP3 7, NEK7, AKT 1 , GSK3B, MAPKAP 2, INSR, LRRK2, PR CB, JA 2, CARD] 1 , MET, MAPK9, IRA 4, MAPK14, EGFR, MAP3K14, RET, MAP2 4, PI 3R1 , RIPK2, PRKCE, MAPK8, MAP2K6, ERBB2, CSF1 R, PLK4, PLK2, PRKCD, SPHK1 , MAPK 1 1 , EIF2AK3, PIK3CA, MERTK, SYK, DR, MARK2, JAK1 , and RAF1.

[000140] For example, in some embodiments, the induced inflammatory modulating agent is a transmembrane receptor, such as, by way of non-limiting example, a

transmembrane receptor selected from the group consisting of: TLR4, TLR3, IFNAR1 , TLR9, CD40, IL28RA, TNFRSF1A, TLR2, CD14, MRC1 , CD244, NCR1 , KLRC4- KLR 1/KLRK 1 , FAS, FCER 1 G, IL1 R1 , LEPR, PGRMC1 , MSR1 , T FRSF 18, Klra4 (includes others), ITGB3, IL4R, FCGR2A, TNFRSF1B, TREM2, NCR3, TLR5, TLR7, ICAM1 , TLR8, IGF1 R, FCER2, IL6R, AGER, CD28, IL1 I RA, ITGB 1 , SIGLEC7, TYROBP, and GFRA1.

[000141] For example, in some embodiments, the induced inflammatory modulating agent is a mammalian endogenous chemical drug, such as, by way of non-limiting example, a mammalian endogenous chemical drug selected from the group consisting of ATP, prostaglandin E2, progesterone, hyaluronic acid, beta-estradiol, superoxide, lauric acid, uric acid, palmitic acid, hydrogen peroxide, tretinoin, histamine, benzylamine, poly(ADP- ribose), ethanol, oleic acid, glutathione, carbon monoxide, cholesterol, sphingosine- 1 - phosphate, arginine, N-acetylglucosamine, testosterone, phosphatidic acid, niacinamide, UDP, nitric oxide, ganglioside GDl a, gamma-linolenic acid, 8-oxo-7- hydrodeoxyguanosine, melatonin, alcohol, D-galactosamine, ganglioside, iron, leukotriene D4, leukotriene C4, 5'-methylthioadenosine, glycochenodeoxycholate, linoleic acid, neuroprotectin D l , hydrocortisone, sodium chloride, heparin, prostaglandin El , 4- phenylbutyric acid, cyclic AMP, fatty acid, chenodeoxycholic acid, UTP, cholecalciferol, lipoxin A4, thromboxane A2, acyl-coenzyme A, geranylgeranyl pyrophosphate, arachidonic acid, formaldehyde, taurine, prostaglandin D2, L-glutamic acid, anandamide, 2- methoxyestradiol. advanced glycation end-products. D-glucose, scpiaptcrin, vanillic acid, D-erythro-C 16-ceramide, citrulline, mevalonic acid, and beta-carotene.

[000142] For example, in some embodiments, the induced inflammatory modulating agent is a non-mammalian endogenous chemical drug such as, by way of non-limiting example, a non-mammalian endogenous chemical drug selected from the group consisting of: peptidoglycan, salmonella minnesota R595 lipopolysaccharidcs, E. coli serotype 0127B8 lipopolysaccharide, E. coli lipopolysaccharide, zymosan, phospholipid, bafilomycin Λ1 , luteolin, E. coli B5 lipopolysaccharide, carrageenan, ursolic acid, apigenin, 2-cyclohexen-l - one, lipoteichoic acid, geldanamycin, manganese, N-acelylmuramyl-L-alanyl-D- isoglutamine, isoliquiritigcnin, cyclomaltodextrin, benzyl isothiocyanate, piceatannol, naringenin, hemozoin, prostaglandin Λ1 , honokiol, pregna-4, 17-diene-3, 16-dione, lipoarabinomannan, D-cysteine, 8-prenylkaempferol, sinapinic acid, (S)-norcoclaurine, fumagillin, 15-deoxy-delta- 12, 14 -PGJ 2, bile acid, prostaglandin J2, iso leucine, and ginsenoside Rgl .

[000143] For example, in some embodiments, the induced inflammatory modulating agent is a chemical drug, such as, by way of non-limiting example, a chemical kinase inhibitor selected from the group consisting of Bay 1 1 -7082, PD98059, U0126, SB203580, LY294002, JAK inhibitor I, lL-6-hydroxymethyl-chiro-mosilol 2-(R)-2-0-methyl-3-0- octadccylcarbonate, tyrphostin AG 1296, wortmannin, Ro31 -8220, SC68376, PS- 1 145, SP600125, PP2/AG1879 tyrosine kinase inhibitor, AG490, PP 1 , bisindolylmaleimide I, tyrphostin AG 127, herbimycin, Go 6976, Sb202190, H89, calphostin C, Rp-cAMPS, Tpl2 kinase inhibitor, CGP77675, Ro 31 -7549, tyrphostin AG 1288, 8-bromo guano sine 3',5'- cyclic monophosphate, SB 220025, AR- 12, erbstatin, KT 5926, tyrphostin 47, and staurosporine.

[000144] For example, in some embodiments, the induced inflammatory modulating agent is another chemical drug, such as, by way of non-limiting example, a chemical reagent, chemical toxicant or other chemical drug selected from the group consisting of: lipopolysaccharide, poly rI:rC-RNA, resiquimod, CpG oligonucleotide, E. coli B4 lipopolysaccharide, lipid A, CpG ΟϋΝ 1 826, Pam3-Cys-Ser-Lys4, dexamethasone, CEP- 1347, phorbol myristate acetate, rosiglitazone, Salmonella enterica serotype abortus equi lipopolysaccharide, ciglitazone, MALP-2s, trinitrobenzenesulfonic acid, CpG ODN 1668, CGS 21680, methyl 2-cyano-3, 12-dioxoolean- l ,9-dien-28-oate, cycloheximide, pyrrolidine dithiocarbamate, lonafarnib, ferrous sulfate, lysophosphatidylcholine, Pam3-Cys, picolinic acid, tacrolimus, aspirin, dextran su lfate, carbon tetrachloride, resveratrol, 2-aminopurine, curcuinin, bleomycin, 3-methyladenine, GW3965, camptothecin, methotrexate, bortezomib, celecoxib, tributyrin, cigarette smoke, arachidonyltrifluoromethane, simvastatin, thioacetamide, epigallocatechin-gallate, lipooligosaccharide, amphotericin B, triamcinolone acetonide, pioglitazone, nystatin, 3M-002, peroxynitrite, S-(2,3-bispalmitoyloxypropyl)- cysteine-GDPKHP SF, fish oils, indomethacin, salicylic acid, arsenite, pirinixic acid, quercetin, parthenolide, fenretinide, paclitaxel, Λ23 187, temozolomide,

tetrachbrodibenzodioxin, atorvastatin, docosahexaenoic acid, N-acetyl-L-cysteine, lansoprazole, rutin, rimonabant, selenium, isoproterenol, aclinomycin D, ATP-gamma-S, vinblastine, bucladesine, cinnamaldehyde, tempol, thalidomide, topotecan,

diethylslilbestrol, fluvastatin, 13-cis-retinoic acid, proteasome inhibitor PSI, ferric nitrilotriacetate, N-Ac-Lcu-Leu-norleucinal, etoposide, mycophenolic acid, chloroquine, tannic acid, rabeprazole, 3M-01 1 , forskolin, okadaic acid, doxorubicin, SB 216763, 2',3'- dialdehyde ATP, NCX-4040, capsazepine, 5-aminosalicylic acid, hexamethoxyilavone, tosyllysine chloromethyl ketone, corticosteroid, 3M-001 , cytochalasin D, cisplatin, cryptotanshinone, methylene blue, L-N6-(l -iminoethyl)-lysine, nitroprusside, N- acetylsphingosine, mifepristone, 5-azacytidine, telmisartan, ebselen, prostaglandin, capsaicin, doxycycline, SR 144528, piperine, pravastatin, carbonyl cyanide m-chlorophenyl hydrazone, ethyl pyruvate, clenbuterol, auranofin, tamoxifen, minocycline, TGAL copolymer, cannabidiol, Sn50 peptide, benzo(a)pyrene, silibinin, 1 '-acetoxychavicol acetate, nimesulide, rofecoxib, isobutylmethylxanthine, diethylmaleate, tranilast, dipyridamole, 1- methyl-4-phenyl-l ,2,3,6-tetrahydropyridine, fulvestrant, imiquimod, 17-alpha- ethinylestradiol, triflusal, tosylphenylalanyl chloromethyl ketone, captopril, fluticasone, fisetin, nicotine, benzyloxycarbonyl-Leu-Leu-Leu aldehyde, cis-urocanic acid, fucoidin, N- nitro-L-arginine methyl ester, genistein, azoxymethane, epicatechin gallate, ionomycin, troglitazone, NS-398, ccrivastatin, allopurinol, 8-chloroadenosine, AZD8055,

chlorpheniramine, diethylthiocarbamate, LY3 1 1727, BN 50730, l -(l -glycero)dodeca- 1 ,3,5,7,9-pentaene, bisperoxo(picolinato)oxovanadate, ethyl vani llin, benznidazole, CE- 2072, metaproterenol sulfate, n-6 docosapentaenoic acid, AG 194204, choline fenofibrate, eicosapentenoic acid, losartan potassium, vancomycin, bryostatin 1 , urethane, estrogen, methylprednisolone, U73 122, metformin, bezafibrate, diclofenac, crocidolite asbestos, acetovanillone, N-formyl-Met-Leu-Phe, reactive oxygen species, SU6656, 2-cyano-3, 12- dioxoolean-l ,9-dien-28-oic acid, semaxinib, streptozocin, green tea polyphenol, S-nitroso- N-acetylpenicillainine, 5-N-ethylcarboxamido adenosine, lactacystin, N-(3- (aminomethyl)benzyl)acetamidine, pentoxifylline, tanespimycin, medroxyprogesterone acetate, sulforafan, propranolol, alpha-tocopherol, arbutin, trans-cinnamaldehyde, hesperidin, sitagliptin, des-Arg( 10)-kallidin, lysine clonixinate, bafilomycin A, soy isoflavones, hydroxyl radical, marimaslat, zileuton, bumelanide, oxazepam, metastat, felodipine, gamma tocopherol, pyri lamine, microcystin, epoxyeicosatrieno ic acid, remifentanil, laminaran, fluniso lidc, ibuprofen, 9, 10-dimethyl- l ,2-benzanthracene, morphine, pimagedine, zVAD-FMK, and S-nitrosoglutathione.

[000145] For example, in some embodiments, the induced inflammatory modulating agent is a biologic drug, such as, by way of non-limiting example, a biologic drug selected from the group consisting of: cyclosporin A, infliximab, interferon beta- l a, NF-kappaB decoy, entcrotoxin B, fontolizumab, anakinra, hemocyanin, grape seed extract, and etanercept.

Use of Modulating Agents

[000146] It will be appreciated that administration of therapeutic entities in accordance with the invention will be administered with suitable carriers, excipients, and other agents that are incorporated into formulations to provide improved transfer, delivery, tolerance, and the like. A multitude of appropriate formulations can be found in the formulary known to all pharmaceutical chemists: Remington's Pharmaceutical Sciences ( 15th cd, Mack Publishing Company, Easton, PA (1 75)), particularly Chapter 87 by Blaug, Seymour, therein. These formulations include, for example, powders, pastes, ointments, jellies, waxes, oils, lipids, lipid (cationic or anionic) containing vesicles (such as Lipofectin™), DNA conjugates, anhydrous absorption pastes, oil-in-water and water-in-oil emulsions, emulsions carbowax (polyethylene glycols of various molecular weights), semi-solid gels, and semisolid mixtures containing carbowax. Any of the foregoing mixtures may be appropriate in treatments and therapies in accordance with the present invention, provided that the active ingredient in the formulation is not inactivated by the formulation and the formulation is physiologically compatible and tolerable with the route of administration. See also Baldrick P. "Pharmaceutical excipient development: the need for preclinical guidance." Regul. Toxicol Pharmacol. 32(2):210-8 (2000), Wang W. "Lyophilization and development of solid protein pharmaceuticals." Int. J. Pharm. 203( 1 -2): 1 -60 (2000), Charman WN "Lipids, lipophilic drugs, and oral drug delivery-some emerging concepts." J Pharm Sci. 89(8):967-

78 (2000), Powel l et al. "Compendium of excipients for parenteral formulations" PDA J Pharm Sci Techno 1. 52:238-3 1 1 ( 1998) and the citations therein for additional information related to formulations, excipients and carriers well known to pharmaceutical chemists.

[000147] Therapeutic formulations of the invention, which include a modu lating agent, are used to treat or alleviate a symptom associated with an immune-related disorder, an aberrant immune response, and/or an neoplastic condition such as, for example, cancer. The present invention also provides methods of treating or alleviating a symptom associated with an immune-related disorder or an aberrant immune response. A therapeutic regimen is carried out by identifying a subject, e.g., a human patient suffering from (or at risk of developing) an immune-related disorder or aberrant immune response, using standard methods. For example, modulating agents are useful therapeutic tools in the treatment of autoimmune diseases and/or inflammatory disorders. In certain embodiments, the use of modulating agents is contemplated, for example, against certain pathogens and other infectious diseases. The modulating agents are also useful therapeutic tools in various transplant indications, for example, to prevent, delay or otherwise mitigate transplant rejection and/or prolong survival of a transplant. The modulating agents are also useful in patients who have genetic defects that exhibit aberrant dendritic cell response.

[000148] The modu lating agents are also useful in vaccines and/or as vaccine adjuvants, against autoimmune disorders, inflammatory diseases, proliferation disorders including cancers, etc. The combination of adjuvants for treatment of these types of disorders are suitable for use in combination with a wide variety of antigens from targeted self-antigens, i.e. , autoantigens, involved in autoimmunity, e.g. , myelin basic protein;

inflammatory self-antigens, e.g. , amyloid peptide protein, or transplant antigens, e.g. , alioantigens. The antigen may comprise peptides or polypeptides derived from proteins, as well as fragments of any of the following: saccharides, proteins, polynucleotides or oligonucleotides, autoantigens, amyloid peptide protein, transplant antigens, allergens, or other macromoleciilar components. In some instances, more than one antigen is included in the antigenic composition.

[000149] Autoimmune diseases include, for example, Acquired Immunodeficiency Syndrome (AIDS, which is a viral disease with an autoimmune component), alopecia areata, ankylosing spondylitis, antiphospholipid syndrome, autoimmune Addison' s disease, autoimmune hemo lytic anemia, autoimmune hepatitis, autoimmune inner ear disease (AIED), autoimmune lymphoproliferative syndrome (ALPS), autoimmune

thrombocytopenic purpura (ATP), Behcet's disease, cardiomyopathy, celiac spruc- dermatitis hepetiformis; chronic fatigue immune dysfunction syndrome (CF1DS), chronic inflammatory demyelinating polyneuropathy (CIPD), cicatricial pemphigoid, cold agglutinin disease, crest syndrome, Crohn's disease, Degos' disease, dei matomyositts- juvenile, discoid lupus, essential mixed cryoglobulinemia, fibromyalgia-fibromyositis, Graves' disease, Guillain-Barre ^" syndrome, Hashimoto's thyroiditis, idiopathic pulmonary fibrosis, idiopathic thrombocytopenia purpura (ITP), IgA nephropathy, insulin-dependent diabetes mellitus, juvenile chronic arthritis (Still's disease), juvenile rheumatoid arthritis, Meniere's disease, mixed connective tissue disease, multiple sclerosis, myasthenia gravis, pernacious anemia, polyarteritis nodosa, polychondritis, polyglandular syndromes, polymyalgia rheumatica, polymyositis and dermatomyositis, primary agammaglobulinemia, primary biliary cirrhosis, psoriasis, psoriatic arthritis, Raynaud's phenomena, Reiter's syndrome, rheumatic fever, rheumatoid arthritis, sarcoidosis, scleroderma (progressive systemic sclerosis (PSS), also known as systemic sclerosis (SS)), Sjogren's syndrome, stiff- man syndrome, systemic lupus erythematosus, Takayasu arteritis, temporal arteritis/giant cell arteritis, ulcerative colitis, uveitis, vitiligo and Wegener's granulomatosis.

[000150] In some embodiments, modulating agents are useful in treating, delaying the progression of, or otherwise ameliorating a symptom of an autoimmune disease having an inflammatory component such as an aberrant inflammatory response in a subject. In some embodiments, modulating agents are useflil in treating an autoimmune disease that is known to be associated with an aberrant dendritic cell response.

[000151] Inflammatory disorders include, for example, chronic and acute

inflammatory disorders. Examples of inflammatory disorders include Alzheimer's disease, asthma, atopic allergy, allergy, atherosclerosis, bronchial asthma, eczema,

glomerulonephritis, graft vs. host disease, hemolytic anemias, osteoarthritis, sepsis, stroke, transplantation of tissue and organs, vascu litis, diabetic retinopathy and ventilator induced lung injury.

[000152] Symptoms associated with these immune-related disorders include, for example, inflammation, fever, general malaise, fever, pain, often localized to the inflamed area, rapid pulse rate, joint pain or aches (arthralgia), rapid breathing or other abnormal breathing patterns, chills, confusion, disorientation, agitation, dizziness, cough, dyspnea, pu lmonary infections, cardiac failure, respiratory failure, edema, weight gain, mucopurulent relapses, cachexia, wheezing, headache, and abdominal symptoms such as, for example, abdominal pain, diarrhea or constipation. [000153] Efficaciousness of treatment is determined in association with any known method for diagnosing or treating the particular immune-related disorder. Alleviation of one or more symptoms of the immune-related disorder ind icates that the modulating agent confers a clinical benefit.

[000154] Administration of a modu lating agent to a patient suffering from an immune- related disorder or aberrant immune response is considered successful if any of a variety of laboratory or clinical objectives is achieved. For example, administration of a modulating agent to a patient is considered successful if one or more of the symptoms associated with the immune-related disorder or aberrant immune response is alleviated, reduced, inhibited or does not progress to a further, i.e. , worse, state. Administration of modulating agent to a patient is considered successful if the immune-related disorder or aberrant immune response enters remission or does not progress to a further, i.e. , worse, state.

[000155] A therapeutically effective amount of a modulating agent relates generally to the amount needed to achieve a therapeutic objective. The amount required to be administered will furthermore depend on the specificity of the modulating agent for its specific target, and will also depend on the rate at which an administered modulating agent is depleted from the free volume other subject to which it is administered.

[000156] Modulating agents can be administered for the treatment of a variety of diseases and disorders in the form of pharmaceutical compositions. Principles and considerations involved in preparing such compositions, as well as guidance in the choice of components are provided, for example, in Remington: The Science And Practice Of Pharmacy 19th ed. (Alfonso R. Gennaro, et al., editors) Mack Pub. Co., Easton, Pa. : 1995 ; Drug Absorption Enhancement: Concepts, Possibilities, Limitations, And Trends, Harwood Academic Publishers, Langhorne, Pa., 1 994; and Peptide And Protein Drug Delivery (Advances In Parenteral Sciences, Vol. 4), 1991 , M. Dekker, New York.

[000157] Where polypeptide-based modulating agents are used, the smallest fragment that specifically binds to the target and retains therapeutic function is preferred. Such fragments can be synthesized chemically and/or produced by recombinant DNA technology. {See, e.g. , Marasco et al., Proc. Natl. Acad. Sci. USA, 90: 7889-7893 ( 1993)). The formulation can also contain more than one active compound as necessary for the particular indication being treated, preferably those with complementary activities that do not adversely affect each other. Alternatively, or in addition, the composition can comprise an agent that enhances its function, such as, for example, a cytotoxic agent, cytokine, chemotherapeutic agent, or growth-inhibitory agent. Such molecules are suitably present in combination in amounts that are effective for the purpose intended.

[000158] All publications and patent documents cited herein are incorporated herein by reference as if each such publication or document was specifically and individually indicated to be incorporated herein by reference. Citation of publications and patent documents is not intended as an admission that any is pertinent prior art, nor does it constitute any admission as to the contents or date of the same. The invention having now been described by way of written description, those of skill in the art will recognize that the invention can be practiced in a variety of embodiments and that the foregoing description and examples below are for purposes of illustration and not limitation of the claims that follow.

EXAMPLES

[000159] The following examples, including the experiments conducted and results achieved are provided for illustrative purposes only and are not to be construed as limiting upon the present invention.

EXAMPLE 1: Materials and Methods

[000160] Cell culture, sorting, and lysis: Cultures of bone marrow derived dendritic cells (BMDCs) from 6-8 week old female B6 mice were prepared as previously described (Amit, I. et al. Unbiased reconstruction of a mammalian transcriptional network mediating pathogen responses. Science 326, 257-263, doi.T 0.1 126/science. l 1 79050 (2009); Chevrier, N. et al. Systematic Discovery of TLR Signaling Components Delineates Viral-Sensing Circuits. Cell 147, 853-867, doi: 10.1016/j.cell.201 1 .10.022 (201 1 )). At 9 days of in vitro culture, the cells were stimulated with lipopolysaccharide (LPS, Invivogen) as previously described (Ibid) for 4h, transferred the cells to a 15 mL conical tube on ice, added 5 μΜ Calcein AM and 5 μΜ Ethidium Homodimer (EthD- 1 , Invitrogen), and then sorted single Calce in-positive, EthD- 1 -negative cells into individual wells of a 96-well plate, each containing 5 μΐ TCL buffer supplemented with 1 % 2-mercaptoethanol (Qiagen, Valencia, CA). After centrifuging, the plates were frozen immediately at -80°C. The total time elapsed between removal from the incubator and lysis was less than 15 minutes. Right before cDNA synthesis, the cells were thawed on ice and purified them with 2.2x RNAClean SPRI beads (Beckman Coulter Genomics, Danvers, MA) without final elution. The beads with captured RNA were air-dried and processed immediately for cDNA synthesis. Wells with no cells were also prepared as negative controls and extracted total RNA from ensembles of 10,000 cells as population samples (see below).

[000161] cDNA synthesis and amplification: The SMARTer U ltra Low RNA Kit (Clontech, Mountain View, CA) was used to prepare amplified cDNA. 1 μΐ of 12 μΜ 3' SMART primer (5'-AAGCAGTGGTATCAACGCAGAGTACT ₍₃ o ₎ N- lN (N = A, C, G, or T; N-l = A, G, or C), SEQ ID NO: 273), 1 μΐ of H ₂ 0, and 2.5 μ ΐ of Reaction Buffer were added onto the RNA-capture beads. The beads were mixed well by pipetting. The mixture was heated at 72°C for 3 minutes and then placed on ice. First-strand cDNA was synthesized with this RNA primer mix by adding 2 μΙ of 5x first-strand buffer, 0.25 μΐ of 1 OOmM DTT, 1 μΐ of 10 mM dNTPs, 1 μΐ of 12 μΜ SMARTer II A Oligo (5'- AAGCAGTGGTATCAACGCAGAGTACXXXXX (X = undisclosed base in the proprietary SMARTer oligo sequence), SEQ ID NO: 274), 100 U SMARTScribe RT, and 10 U RNase Inhibitor in a total volume of 10 μΐ and incubating at 42°C for 90 minutes fo llowed by 10 minutes at 70°C. The first strand cDNA was purified by add ing 25 μ ΐ of room temperature AMPure XP SPRI beads (Beckman Coulter Genomics, Danvers, MA), mixing well by pipetting, incubating at room temperature for 8 minutes. The supernatant was removed from the beads after a good separation was established. All of the above steps were carried out in a PCR product- free clean room. The cDNA was amplified by adding 5 μ| of 1 Ox Advantage 2 PCR Buffer, 2 μΐ of 1 0 mM dNTPs, 2 μΐ of 12 μΜ IS PCR primer (5'- A AG C AGTG GTATC A ACGC AG AGT, SEQ ID NO: 275), 2 μΐ of 50x Advantage 2 Polymerase Mix, and 39 μΐ ¾0 in a total volume of 50 μΐ. The PCR was performed at 95°C for 1 minute, followed by 21 cycles of 15 seconds at 95°C, 30 seconds at 65°C and 6 minutes at 68°C, followed by another 10 minutes at 72°C for final extension. The amplified cDNA was purified by adding 90 μΐ of AMPure XP SPRI beads and washing with 80% cthanol. For molecule counting {see Kivioja, T. et al. Counting absolute numbers of molecules using unique molecular identifiers. Nature Methods 9, 72-74,

doi: 10.1038/nmeth. l 778 (201 1 ) (as in Fig. 9 & 1 0), the SMARTer II A Oligo was replaced with a custom RNA oligonucleotide containing four random bases (Barcoded SMARTer II A Oligo: 5'- AAGCAGTGGTATCAACGCAGAGTNNNNrGrGrG-3 ', SEQ ID NO: 276).

[000162] cDNA shearing and library construction: The purification buffer (Clontech) was added to the amplified cDNA to make a total volume of 76 μΐ. The cDNA was sheared in a 100 μΐ tube with 10% Duty Cycle, 5% Intensity and 200 Cycles/Burst for 5 minutes in the frequency sweeping mode (Covaris S2 machine, Woburn, MA). The sheared cDNA was purified with 2.2 volumes AMPure XP SPRI beads.

[000163] Indexed paired-end libraries for Illu mina sequencing were prepared as described (see Levin, J. Z. et al. Comprehensive comparative analysis of strand-specific RNA sequencing methods. Nature Methods 7, 709-715 (2010). with the following modifications. First, a different indexing adaptor (containing an 8-base barcode) was used for each library. Second, the ligation product was size-selected by using two rounds of 0.7 volume of AMPure XP SPRI bead cleanup with the first round starting volume at 100 μΐ. Third, PCR was performed with Phusion High-Fidelity DNA polymerase with GC buffer and 2 M betaine. Fourth, 55°C was used as the annealing temperature in PCR with the universal indexing primers (forward primer 5'-

AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGAC (SEQ ID NO: 277), reverse primer 5'-CAAGCAGAAGACGGCATACGAGAT (SEQ ID NO: 278)). Fifth, 12 cycles of PCR were performed. Sixth, PCR primers were removed using two rounds of 1 .0 volume of AMPure beads.

[000164] Population controls and negative controls: For positive (population) controls, 13.8 ng of total RNA was isolated, as measured by BioAnalyzer (Agilent, Santa Clara, CA), from 10,000 cells using PrepEase RNA Spin Kit (Affymetrix, Santa Clara, CA). 1 ng of total RNA was used in the above processes except that only 12 cycles were used in the cDNA amplification step. For negative controls, all of the above processes were carried out starting with zero sorted cells in TCL-buffer-containing wells. 18 cycles in the final PCR of I llumina library construction was used.

[000165] Read trimming and mapping: During reverse transcription, the SMART polymerase adds short (SMARTer II A Oligo) and long (SMART primer oligo) adapters to the beginning of the second read for fragments originating from the 5' and 3' ends of the transcript, respectively. Before mapping reads, these adapter sequences were removed using Btrim64 with command line arguments -1 1 -e 100 -v 1 -b 28 -a - 100. Adapter sequences were trimmed from approximately a third of the second reads. Trimmed reads were mapped to the mm9 version of the mouse genome using Tophat vl .4.1 (Trapnell, C, Pachter, L. & Salzberg, S. L. TopHat: discovering splice junctions with RNA-Seq. Bio informatics 25, 1 105- 1 1 1 1 , doi: 10.1093/bioinlbrmatics/btpl 20 (2009)) with default parameters. Genome mappings were used to visualize data in the Integrative Genome Viewer (Robinson, J. T. et al. Integrative genomics viewer. Nature Biotechnology 29, 24-26, doi: 10. 1038/nbt. 1 754 (201 1 )), and to compute a set of library quality metrics, as described below.

[0001661 Reads where the short adapter (5' end) was trimmed mapped at approximately equal rates to untrimmed reads. However, read pairs where the long adapter (3' end) was trimmed often contained polyA stretches even after trimming, and mapped at extremely low rates (<1 %). Since these reads should originate from the 3' end of the transcript, this low mapping percentage results in a depletion of reads from the 3' end of the transcript. This depletion may cancel out the 3' coverage bias that is a byproduct of the SMART protocol (see below) (Ramskold, D. et al. Full-length mRNA-Seq from single-cell levels of RNA and individual circulating tumor cells. Nature Biotechnology 30, 777-782, doi:papers2://publication/doi/10.1 038/nbt.2282 (2012)).

[000167] Quantifying unique mRNA molecules: When processing the three single-cell libraries where the SMARTer oligo was modified to include a tbur-nucleotide random barcode sequence, reads containing the SMARTer II A Oligo were isolated and trimmed as described above. Four additional bases (corresponding to the barcode) were then trimmed and maintained for later processing. Trimmed reads were mapped to the mouse mm9 genome as described above. For each gene, the subset of these reads that mapped to exonic sequence on the correct strand was then identified, and their orig inal tbur-nucleotide barcodes were retrieved. The unique number of barcodes for each gene was counted and used as an alternative quantification of single-cell gene expression. Both unique molecular barcode counts and TPM estimates were provided for all three cells.

[000168] Library quality metrics: Library quality metrics, including genomic mapping rates, coefficients of variation of coverage of each transcript, the fraction ofribosomal RNA in each library, and positional coverage biases, were calculated using PicardTools version 1 .42 (picard.sourceforge.net). Less 3' bias was observed in this data, compared to previous reports (Ramskold, D. et al. Full-length mRNA-Seq from single-cell levels of RNA and individual circu lating tumor cells. Nature Biotechnology 30, 777-782,

doi:papers2://publication/doi/10.1 038/nbt.2282 (2012)), likely due to the differences in library construction noted above (Fig. 22).

[000169] Expression level calculation: A Bowtie index (Langmead, B., Trapnell, C, Pop, M. & Salzberg, S. Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome Bio logy 10, doi: 10. 1 1 86/gb-2009- 10-3-r25 (2009)) was created based on the UCSC knownGene transcriptome (Fujita, P. A. et al. The UCSC Genome Browser database: update 201 1. Nucleic Acids Research, doi: 10.1093/nar/gkq963 (2010)), and paired-end reads were aligned directly to this index using Bowtie v 0.12.7 with command line options -q -phred33-quals -n 2 -e 99999999 -1 25 -1 1 -X 1000 -a -m 200. Next, RSEM vl . l 1 (Li, B. & Dewey, C. RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome. BMC Bio informatics 12,

doi: 10.1 186/1471 -2105- 12-323 (201 1 )) was ran with default parameters on these alignments to estimate expression levels. RSEM's gene level expression estimates (tau) were multiplied by 1 ,000,000 to obtain transcript per million (TPM) estimates for each gene. To transform expression levels to log-space, the ln(TPM+l) was taken. When calculating the "average" single-cell expression level, TPM levels from each of the 18 single cells were first averaged, and then this average estimate was transformed into log space.

[000170] Identical procedures were applied to a previously published dataset (Garber, M. et al. A High-Throughput Chromatin Immunoprecipitation Approach Reveals Principles of Dynamic Gene Regulation in Mammals. Molecular Cell 47, 810-822,

doi: 1 0.1016/j.molcel.2012.07.030 (2012)), consisting of an RNA-Seq time course after LPS stimulation of BMDCs. This dataset was used to identify a set of 632 genes that were induced at least two-fold in the population at 4h following LPS stimulation as compared to pre-stimulation. These genes were analyzed in Fig. 2a, Fig. 2d, and Fig. 4b.

[000171] RNA fluorescence in situ hybridization (FISH): The expression levels were measured for 25 different mRNA transcripts in situ using RNA-FISH probes (Panomics). Briefly, BMDCs were sorted on Cd l l c (Miltenyi Biotech) at 8 days in vitro and plated on poly-/-lysine coated glass coverslips. The following morning, some cells were stimulated with LPS as previously described (Amit, I. et al. Unbiased reconstruction of a mammalian transcriptional network mediating pathogen responses. Science 326, 257-263,

doi: 10.1 126/science. l 1 79050 (2009); Chevrier, N. et al. Systematic Discovery ofTLR S ignaling Components Delineates Viral-Sensing Circuits. Cell 147, 853-867,

doi: 10.1016/j. cell.201 1 .10.022 (201 1)). Ten minutes prior to fixation, cell culture media was replaced with a 1 :500 dilution of Alexa-350 Wheat Germ Agglutinin ( WGA,

Invitrogen) in HBSS. Subsequently, cells were fixed and stained according to the manufacturer's recommendations. After curing overnight, Slowfade (Invitrogen) mounted coverslips were raster scanned at 60x magnification (1.42 NA, oil immersion) in x, y, and z using an epifluorescence microscope (Olympus) outfitted with Metamorph software. On average, 100 individual 3-dimensional stacks were taken for each sample. For all samples, four-color imaging was performed to obtain the following information: excitation (ex) 405nm - WGA & DAPI stains; ex 488nm, ex 546nm, ex 647nm - Probes 1 , 2, and 3 , respectively.

[000172] The obtained images were processed in two phases. First, CellProfiler (Carpenter, A. et al. CellProfiler: image analysis software for identifying and quantifying cell phenotypes. Genome Bio logy 7, doi: 10.1 186/gb-2006-7- 10-rl 00 (2006)) was used to determine cell numbers and locations for each stack of images taken using the UV filter set (ex405nm). Brightly stained nuclear regions (DAPI) were used to identify individual nuclei and were then used as seeds for determining the extents of each cell from the duller membrane outlines (WGA). The locations and extents of individual cells were then extracted for each imaging position using the software. Next, for each color channel, individual mRNAs were identified and counted in Matlab using a previously described analysis package (Raj, A., Van Den Bogaard, P., Rifkin, S. A., Van Oudenaarden, A. & Tyagi, S. Imaging individual mRNA molecules using multiple singly labeled probes. Nature Methods 5, 877-879, doi: 1 0.1038/nmeth. l 253 (2008)). Identified mRNAs were then allotted to individual cells using the output of CellProfiler. Final analysis and plotting was also performed using Matlab. The displayed RNA-FISH images were false-colored and overlaid using Adobe Photoshop.

[000173] For all RNA FISH histograms, counts were binned (n = 50) and smoothed with a window of 5 bins in Matlab. As controls, BMDCs that were not stimulated with LPS were also analyzed to ensure the specificity of the induced-gene RNA-FISH probes (Fig. 23).

[000174] For splicing analyses, custom RNA fish probes (Panomics) were designed to either Ir†7 or Acpp as follows:

[000175] The difference in the number of bDNAs between the two constitutive Irf7 probes led to slightly better binding and thus higher counts for the constitutive probe B. As a result, the metric, probe A counts/(probe A counts + probe B counts) (used in the histogram in Fig. 3c), is normally distributed with a mean of -0.45 (instead of -0.5).

Plotting the isoform-specific probe over constitutive probe B gave a similar curve (compare Fig. 3c with Fig. 15). A cell was only included if the number of counted mRNAs for the constitutive probe (Irf7) was at least 5 or if the sum of alterative exon counts was at least 5. For Acpp, n = 615 cells; for Irf7, n = 490.

[000176] Immunofluorescence (IF) measurements'. IF co-staining was performed as previously described (Chevrier, N. et al. Systematic discovery of TLR signaling components delineates viral-sensing circuits. Cell 147, 853-867,

doi: 10.1016/j.cell.201 .10.022 (201 1); Shalek, A. K. et al. Nano w ire-Mediated Gene S ilencing in Primary Immune Cells: Identification of Patient-Specific Responses in Chronic Lymphocytic Leukemia. In Review (2012)) directly after RNA-FISH staining. Stat l, pStat l , and Stat2 antibodies, all used at 1 :200, were obtained from Santa Cruz Biotechnology. Average and total fluorescence levels, as well as the percentage of the fluorescence localized to the nucleus, were quantified from epilluorescence images using locations and extents of individual cells and their nuclei, as above (Fig. 17 & 1 8). For all protein histograms, counts were binned (n = 100) and smoothed with a window of 5 bins in Matlab. Single-plane and 3-dimensional scans yielded similar results (data not shown).

[000177] Single-cell qRT-PCR: Single BMDCs were prepared for qRT-PCR using the Single-Cell-to-Ct kit (Ambion) with minor modifications. Namely, individual BMDCs were sorted into one-fourth of the recommended lysis buffer volume and all subsequent steps were scaled to match. After specific target amplification, an exonuclease I digestion (NEB) was performed by adding 0.5 Exonuclease 1, 0.25 μί Exonuclease I Reaction Buffer, and 1.75 water to each sample, vortexing, centrifuging, and heating to 37°C for 30 minutes. After an 80°C heat inactivation for 15 minute, samples were diluted 1 :5 in Buffer TE. Single cells, negative controls, and population controls (prepared equivalently using extracted total RNA) were analyzed using 96x96 gene expression chips (Fluidigm Biomark) (Dalerba, P. et al. Single-cell dissection of transcriptional heterogeneity in human colon tumors. Nature Biotechnology 29, 1 120- 1 127, doi: 10.1038/nbt.2038 (201 1 )).

[000178] Fano factor calculation: The Fano factor (normalized standard deviation) was calculated as the ratio of the standard deviation of gene expression values (log space) across single cells and the average single cell expression level (log space, see above). The dashed grey lines in Fig. 2a,b represent a constant Fano factor of 0.25, and broadly separate highly expressed genes into two groups of variable and non- variable genes, as shown below in Table S3. Functional enrichment analysis of these two gene sets (see below) was highly robust to small changes in the Fano factor threshold (between 0.2 and 0.3) that was used.

Table S3. Highly Expressed Genes Based On Averages In Cell Populations

D21RIK

2900073- ACAD9 CCL5 CLEC4N IFI204 PILRB1 G15RIK

ATP5L SUM01 TYROBP CHMP2A PSMC4 PP1

SHF 1 PSME2 H3F3A RPL17 PLA2G7 PTGS1

GHITM CAPZA1 HMGB1 CD274 PLD3 MGL2

BC071253 RPL14 CD63 LGALS3 PRDX2 ATP6V0D2

AK163440 6720456- IL1B IL6 FAM96A TARM1

B07RIK

GU332589 VAMP8 TNFAIP3 ANXA3 GTF2B CXCL1

DAZAP2 C920009- CDKN1A IFITM3 MSR1 RSAD2

B18RIK

FTL1 MALAT1 CNBP MMP12 DAD1 THBS1

RPS19 GPI1 PRDX5 ECH1 IDH1 EMR1

CALM2 LITAF EI F3E A130040- VPS29 RBM7

M12RIK

BCL2A1A IL2RG CCL6 MGST1 INHBA TUBA1B

ATP6V0E RPL37A HNRNPC ARPC1B PFDN5 IFI205

RPL35 TPM3 FXYD5 IRG1 CFP MMP13

MT-ND4 PTAFR HN1 LDHA GRN SIRPB1B

RPL23 PPIB RPL30 DLD TC 1 AW112010

MSN SAT1 PSAP CCL2 PRDX6 TBXAS1

AK140265 T BIM6 LILRB4 SH3BGRL LI PA KLK1B11

ATP6AP1 PSMB1 ANXA4 PSMA6 SCPEP1 LY6C2

CD52 BTG1 SAMSN1 ALOX5AP SERPINB2 GLIPR2

RPS27A RPL34 AKAP13 SDHD IGSF6 CD86

ALDOA RPL7A 1SG15 SDHA NME2 C1QB

SUB1 RPS29 CYBB H2-DMA LGALS9 H2-M2

TALDOl CCDC72 PTP4A2 CLEC7A C0R01A ACSL1

CFL1 ITGB2 RPL27 RPL10A RPS27L IFNB1

CLIC1 HNRNPF NFE2L2 TNIP3 LCP2 FPR1

RPS18 BRP44L H2-K1 PI LRA TNFSF15 FPR2

ANXA5 EIF3K SBDS CXCL10 IFIT1 LGALS1

GM15450 CD44 CTSD APSA CLEC4D CCR7

ARHGDIB RPS7 UQCRB EIF4A1 HPRT 0ASL1

S100A11 IQGAP1 H2AFV PFKP TTC35 GLRX

RPL32 DSTN CYB5 UBE2L6 ATP5H CHI3L3

SRGN BZW1 MT1 LAMP2 LGMN FLRT3

ARPC3 AA467197 PLD4 ALAS1 ESD TMEM39A

RPS9 WDR1 TREX1 ATP6V1F IDI1 PF4

3110003- ATP6V1E1 ILIA ARL5C PSMA1 AK041746 A17RIK

LCP1 RPS8 H1F0 LYZ1 SEC13 EAR2 MYL6 RPS27 CD48 PLI N2 SUM02 IL23A

A 141672 ATP5G2 RPS26 LGALS3BP 0AZ1 SAA3

CSTB CCL9 ASS1 PGK1 RSU1 CD82

COPZ1 RPS17 LYZ2 TFEC PLAU R ZFP263

RPS14 ENOl ASAH1 ATP5J GBP2 LY86

CAPZA2 ERH COX4I 1 RNH 1 CCL22 UPP1

ATP6V1G 1 PRDX1 POM P PG D PSMD14 TMEM 176B

RPL24 RPL15 TM EM50A PSM B2 GYG IFIT2

EEF1A1 BN L1 MAP1LC3 B 2900010- MY01F CD69

J23RIK

CFLAR CTSC CTSZ PSMA2 CD38 GPR84

NPC2 VPS35 ARF1 EEF1G DPEP2 TNFSF4

SRSF5 ACTG1 RPL28 TNFSF9 GM6644 STM Nl

TANK LI LRB3 AP2M 1 TLR2 CD80 GM6377

RPS25 RPS20 COX6C 1810029- DAB2 I L1R2

B16RI K

FAU UQCRH TNF CTSA CCT5 TU BA1A

EI F4G2 GN B2L1 SDCBP GM 11428 ETFB CD40

EEF2 HSPA8 RASGEF1B CD53 MMP8 NIACRl

RPS5 AY096003 PKM2 HSPA5 EVI2A

[000179] The dashed blue line in Fig. 2a represents the maximum theoretical standard deviation for the 18 single cells given their single cell average. This theoretical maximum occurs when the cells are perfectly bimodally distributed about a value of (^i+log(2))/2 and is represented by the relationship: a _max = sqrt( l 8/17)*^+log(2))/2).

[000180] Functional enrichment of variable/non-variable gene sets: Functional enrichment (GO annotation) of non- variable highly expressed gene sets was performed using DAVID v6.7 (Huang, D. W., Sherman, B. T. & Lempicki, R. A. Bio informatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists. Nucleic Acids Research 37, 1 - 13, doi: 10.1093/nar/gkn923 (2009); Huang, D. W., Sherman, B. T. & Lempicki, R. A. Systematic and integrative analysis of large gene lists using

DAVID bio informatics resources. Nature Protocols 4, 44-57, doi: 10.1038/nprot.2008.21 1 (2009)). The full list of 522 highly expressed genes was used as the background set. For Fig. 2a and Table S3, two lists were combined to form a set of housekeeping genes. The first list is a set of ribosomal subunit proteins defined in GO annotations (Huang, et al., Nucleic Acids Research 2009; Huang, et al., Nature Protocols 2009) and the second list is taken from a table of commonly used mouse housekeeping genes that were downloaded from the Qiagen website. [000181] Correlation matrix and principal component analysis (PC A): PCA for 632 induced genes was performed in R using the prcomp function. The expression values of each gene were transformed to have zero mean and unit variance across single cells in order to appropriately compare variability patterns across genes with different overall abundance in the population.

[000182] A correlation matrix was calculated based on the log-scale (but non- transformed) gene-expression estimates, and clustered the matrix using k-means. A parameter of five clusters based on the "elbow method" (Diday, E. New approaches in classification and data analysis. (Springer-Verlag, 1994)) (data not shown) were chosen, but the identification of a strongly enriched antiviral cluster (and its high degree of overlap with PC2) was highly robust to the parameter choice or stochastictty of k-means.

[000183] In the set of 632 genes, a set of antiviral gene targets from previous work (Amit, I. et al. Unbiased reconstruction of a mammalian transcriptional network mediating pathogen responses. Science 326, 257-263, doi: 10.1 126/science. l 179050 (2009)) were annotated. Stat2 targets were annotated from a previously defined set of "promoter ChIP peaks" (Garber, M. et al. A High-Throughput Chromatin Immunoprecipitation Approach Reveals Principles of Dynamic Gene Regulation in Mammals. Molecular Cell 47, 810-822, doi: 10.1016/j.molcel.2012.07.030 (2012)) on a set of identically stimulated (at 4h) BMDCs. Cluster-specific enrichment analyses were performed using a hypergeometric test in R, using the full set of 632 induced genes as a background set.

[000184] Population fluorescence-activated cell-sorting (FACS) analysis and quantitative reverse transcription polymerase chain reaction (qRT-PCR): BMDCs were stimulated with LPS for 4h. Fifteen minutes prior to sorting, cells were stained with each of 1 1 antibodies from Bio legend that defined the semi-mature (S) or maturing (M) cells: Cd83 (S), Cd273 (S), Ccr7 (S), Cd40 (S), Cd20 l (S), Cd I 37 (S), Cd68 (M), Cdl20b (M), Cd53 (M), Cd88 (M), and Cd l 6/32 (M). Three groups of 1 ,000 cells either positive or negative for each of the tested surface markers were sorted in 1 00 of buffer TCL supplemented with 1 % 2-mercaptoethanal. Total RNA was then extracted from each of the 20 samples using an RNeasy Mini Kit (Qiagen) and cDNA was prepared using Sensiscript RT (Qiagen) as previously described (Shalek, A. K. et al. Vertical silicon nanowires as a universal platform for delivering biomolecules into living cells. Proceedings of the National Academy of Sciences 107, 1 870- 1875, doi: 10.1 073/pnas.0909350107 (2010)). Population-wide expression levels for different transcripts were then analyzed relative to GAPDH using qRT-PCR, as previously described (Shalek, A. K. et al. Vertical silicon nanowires as a universal platform for delivering biomolecules into living cells. Proceedings of the National Academy o f Sciences 107, 1870- 1 875, doi: 1 0.1073/pnas.0909350107 (2010)) (Fig. 15). Primers for qRT-PCR are presented below in Table S6.

Table S6. Gene List, PCR Primer Pairs For Fluidigm Single Cell qPCR Codeset

SEQ

Gene SEQ ID

Forward Primer ID Reverse Primer

Name NO:

NO:

18s gcaattattccccatgaacg 1 gggacttaatcaacgcaagc 137

28s tcatcagaccccagaaaagg 2 gattcggcaggtgagttgtt 138

Actb ctaaggccaaccgtgaaaag 3 accagaggcatacagggaca 139

Anxa7 gaacgtctcctcgtgtccat 4 ggccatctggtggttcac 140

Arbp/RPLPO actggtctaggacccgagaag 5 tcccaccttgtctccagtct 141

Arf4 gatgcgcattttgatggtt 6 ttcagtttatacagaattgtcgtcttg 142

Arg2 tatggtccagctgccattc 7 ccaaagtcttttaggtggcatc 143

Atf3 gctggagtcagttaccgtcaa 8 cgcctccttttcctctcat 144

Atf4 atgatggcttggccagtg 9 ccattttctccaacatccaatc 145

B2m ttctggtgcttgtctcactga 10 cagtatgttcggcttcccattc 146

Calcrl ctcctgagactattcccacagaa 11 caagatgttgctgtatcatcatagg 147 cavl ccagggaaacctcctcaga 12 ccggatgggaacagtgtaga 148

Ccl2 catccacgtgttggctca 13 gatcatcttgctggtgaatgagt 149

Ccl3 tgcccttgctgttcttctct 14 gtggaatcttccggctgtag 150

Ccl7 ttctgtgcctgctgctcata 15 ttgacatagcagcatgtggat 151

Ccnd2 ctgtgcatttacaccgacaac 16 cactaccagttcccactccag 152

Ccr7 ctccttgtcattttccaggtg 17 tggtattctcgccgatgtagt 153

Cdl4 aaagaaactgaagcctttctcg 18 agcaacaagccaagcacac 154

Cebpb tgatgcaatccggatcaa 19 cacgtgtgttgcgtcagtc 155

Cited2 atcgcaaagacggaagga 20 tgctgctggtgatgatgc 156

Clec4e gcctccatcctgtttctcag 21 tgagagctgcgatatgttacg 157

Cxcll ct ggg attca cctcaa gaa catc 22 cagggtcaaggcaagcctc 158

CxcllO gccgtcattttctgcctca 23 cgtccttgcgagagggatc 159

Cxcl2 aaaatcatccaaaagatactgaacaa 24 ctttggttcttccgttgagg 160

DDX58 gaagattctggaccccaccta 25 tgaatgtactgcacctcctca 161

Dnmt3a acacagggcccgttacttct 26 tcacagtggatgccaaagg 162 ets2 cagttttcgtgggacactca 27 aagggagcacagcaaacaga 163

Gnb4 ttgggatagctatacgacaaataaga 28 ggcgtaggcacaggtcat 164

Hmgn2 gctcccagcgctataaaaact 29 tgagcacggggatacagc 165

Hprt tcctcctcagaccgctttt 30 cctggttcatcatcgctaatc 166

Ifihl ctattaaccgtgttcaaaacatgaa 31 cacctgcaattccaaaatctta 167

Ifitl tctaaacagggccttgcag 32 gcagagccctttttgataatgt 168 SEQ

Gene SEQ ID

Forward Primer ID Reverse Primer

Name NO:

NO:

Ifit2 gcaagatgcaccaagatgag 33 cttctaatgaagtgctccagacc 169

Ifit3 tgaactgctcagcccaca 34 tcccggttgacctcactc 170

Ifnbl ctggcttccatcatgaacaa 35 agagggctgtggtggagaa 171

Ikbke gggagagtctttgcctgattc 36 atctcctgggcttggctatc 172

1112b gattcagactccaggggaca 37 tggttagcttctgaggacacatc 173

1115 cagctcagagaggtcaggaaa 38 catgaagaggcagtgctttg 174

IL15ra ccagtgccaacagtagtgaca 39 ttgggagagaaagcttctgg 175

Ilia ttggttaaatga ctgcaaca 40 gagcgctcacgaacagttg 176 lllb acctgtcctgtgtaatgaaagacg 41 tgggtattgcttgggatcca 177

116 gctaccaaactggatataatcagga 42 ccaggtagctatggtactccagaa 178 inhba atcatcacctttgccgagtc 43 tcactgccttccttggaaat 179

Irfl gagctgggccattcacac 44 tccatgtcttgggatctgg 180

Irf4 acagcaccttatggctctctg 45 atggggtggcatcatgtagt 181

Irf7 cttcagcactttcttccgaga 46 tgtagtgtggtgacccttgc 182

Irf8 gagccagatcctccctgact 47 ggcatatccggtcaccagt 183

Irf9 tgaggccaccattagagagg 48 agcagcagcgagtagtctga 184

Irgl gcttttgttaatggtgttgctg 49 ggcttccgatagagctgtga 185

Isgl5 agtcgacccagtctctgactct 50 ccccagcatcttcaccttta 186

Jak2 aagattgccaaggccaga 51 tgttgttccagcactctgtca 187

Jarid2 gcacttgtgctacctgtcca 52 tccaggcagaacacgacat 188

Lgals9 gcattggttcccctgagata 53 tccagtaaaggggatgatcg 189 mapkapk2 cagcaaaaattcgccctaaa 54 agtgcagctccacctctctg 190

Mt2 catggaccccaactgctc 55 agcaggagcagcagcttt 191

Mxl ttcaaggatcactcatacttcagc 56 gggaggtgagctcctcagt 192 x2 ca gttcctctca gtccca a ga t 57 tgcggttgtgagcctctt 193 yd88 tggccttgttagaccgtga 58 aagtatttctggcagtcctcctc 194

Nfe2l2 catgatggacttggagttgc 59 cctccaaaggatgtcaatcaa 195

Nfkbl cactgctcaggtccactgtc 60 ctgtcactatcccggagttca 196

Nfkbia acgagcaaatggtgaaggag 61 atgattgccaagtgcagga 197

Nfkbiz cagctggggaagtcattttt 62 ggcaacagcaatatggagaaa 198

Pa2g4 ggtcgtgaccaagtataagatgg 63 cagacacacctgagctggaa 199

Pelil ctgatcaagaaaatcatccttcc 64 accgtttgggagagatccat 200 pgkl tacctgctggctggatgg 65 cacagcctcggcatatttct 201

Plek agtggatcaaagccatccag 66 tcagtgattctcggtgtcctc 202

Plkl ttgtagttttggagctctgtcg 67 agtgccttcctcctcttgtg 203

Plk2 catcaccaccattcccact 68 tcgtaacactttgcaaatcca 204

Pml aggaaccctccgaagactatg 69 ttcctcctgtatggcttgct 205

Pnrc2 tgtgctgaggagactcgatg 70 tgagccagtctgctgatttc 206

Ppia acgccactgtcgcttttc 71 gcaaacagctcgaaggagac 207 SEQ

Gene SEQ ID

Forward Primer ID Reverse Primer

Name NO:

NO:

Ptgs2 gatgctcttccgagctgtg 72 ggattggaacagcaaggattt 208 ptx3 cgctgtgctggaggaact 73 gggaagaaaattgctgtttcac 209

Rel ttgcagagatggatactatgaagc 74 caccgaatacccaaattttgaa 210

Rpll3a atccctccaccctatgacaa 75 gccccaggtaagcaaactt 211

Rsad2 gcttcaacgtggacgaagac 76 cctcaattaggaggcactgg 212

Serpinb9 tgtggacctcagcaaggtg 77 cctcaacatcagtgctcttcat 213

Sfpil ggagaagctgatggcttgg 78 caggcga atctttttcttgc 214

Slc7all tgggtggaactgctcgtaat 79 aggatgtagcgtccaaatgc 215 slfnl cgtgctcagtagagcagcttag 80 catcggtgatgttcattttcc 216

Slfn2 aggcaactgagcaaagcaac 81 ttgcattttccagctgaatg 217

Socs3 atttcgcttcgggactagc 82 aacttgctgtgggtgaccat 218

Statl gcagcacaacatacggaaaa 83 tctgtacgggatcttcttgga 219

Stat2 ggaacagctggaacagtggt 84 gtagctgccgaaggtgga 220

Tank attccccaggaaaggctgt 85 ttggttaagaaaaggcttccaa 221

Tapbp cagcactctcttcagcctctc 86 tatgggtgaggacggtcag 222

Tbkl cctcggaggaacaaagaagtaa 87 tccagatattgcaccagacg 223

Tmem39a gacgggcttgaggaacag 88 cctggggtaattaaggctgtg 224

Tnf tcttctcattcctgcttgtgg 89 ggtctgggccatagaactga 225

Tnfaip2 ggagcctttgaaagacctcaa 90 gaacttcttaaacagcggcttc 226

Tnfrsflb gaggcccaagggtttcag 91 ggcttccgtgggaagaat 227 tnfsf4 aaaggaccctccaatccaaa 92 agttgcccatcctcacatct 228

Trexl cagggcagaccaagaattg 93 ggtctgtgagcccatgct 229

Trim 12a agcaccgtggtcacaaaac 94 cagcctttgcagaactacctg 230

Ywhaz aacagctttcgatgaagccat 95 tgggtatccgatgtccacaat 231

Zfp36ll ttcacgacacaccagatcct 96 tgagcatcttgttacccttgc 232

Arid5a cagagcaggagccagagc 97 gccaagttcatcatacacgttc 233

Bat5 acattgctgctgctacttgc 98 gtactgggggttggtccag 234

Bcl3 gaacaacagcctgaacatgg 99 tctgagcgttcacgttgg 235

Cbx4 gtgggaaccagaggagaaca 100 tcagctgctcctgccttt 236

Clic4 actgtaacctgctgcccaag 101 aggaatatcaaagttgcggtattt 237

Crkl cgccaggtttgattcttcag 102 cctcctggcgagtcactg 238

E2f5a aaccccagatgctgacaaag 103 ccacctttattttaggtttcttgg 239

Fos gggacagcctttcctactacc 104 gatctgcgcaaaagtcctgt 240

Fus aaggcctaggcgagaatgtt 105 cataggctgtccagttttcttgt 241

Gapdh ggcaaattcaacggcacagt 106 agatggtgatgggcttccc 242

Hhex tcagaatcgccgagctaaat 107 ctgtccaacgcatccttttt 243

Ifna2 atgaggaggctcccctttc 108 accttctccagggggaatc 244

Ifna4 tcaagccatccttgtgctaa 109 gtcttttgatgtgaagaggttcaa 245

1112a ccatcagcagatcattctagacaa 110 cgccattatgattcagagactg 246 SEQ

Gene SEQ ID

Forwa rd Primer ID Reverse Primer

Name NO:

NO:

isg20 ttggtgaagccaggctagag 111 cttcagggcattgaagtcgt 247

Jun ccagaagatggtgtggtgttt 112 ctga ccctctccccttgc 248

Junb ccacggagggagagaaaatc 113 agttggcagctgtgcgtaa 249

Lcp2 ccaacaggcaggaatcactc 114 cttctgctgggctcttcgt 250

Map3k7 ccatcccaatggcgtatc 115 ccatggattctttggagtttg 251

Mapk9 acgttaccagcaactgaaacc 116 gaactgtatcaaaagcagcacaa 252

Nfkb2 tggaacagcccaaacagc 117 cacctggcaaacctccat 253

Parpl4 tggagatcctagtgacaaaaatcc 118 ctggaaaggctcccatagatac 254

Phlppl cttgccctggaccacaaa 119 gtcaatcttgaagcagcgaat 255

Plagl2 catccggagcagagacca 120 atgcactggtggggtttc 256

Plk3 ggctggcagctcgattag 121 gttgggagtgccacagatg 257

Plk4 gaaaaccaaaaaggctgtgg 122 tccttcagacgcacactctc 258

Rbll gcggcaactacagcctagag 123 tgcggcaagcaacatataaa 259

Rela cccagaccgcagtatccat 124 gctccaggtctcgcttctt 260

Relb gtgacctctcttccctgtcact 125 tgtattcgtcgatgatttccaa 261

Runxl ctccgtgctacccactcact 126 atgacggtgaccagagtgc 262

Sap30 cggtgcagtgtcagcttc 127 ctcccgcaaacaacagagtt 263

Sbds ggtggtggagagtgaggact 128 gctcatcaatttctctgaagca 264

Sfrsl ggtccgagaacagagtggtt 129 cctttaagtcctgccagcttc 265

Sfrs3 tcgtcgtcctcgagatgatt 130 ctccttcttggggatctgc 266

SnxlO gccagggcttggaagatt 131 cagatggctctgcaggaag 267

Stat4 cggcatctgctagctcagt 132 tgccatagtttcattgttagaagc 268

Timeless gagtcctcagcgaga ccttg 133 tgtcttcttcttgccgatcc 269

Tmod3 ccaagagcgttttcccaat 134 gttggatttggtggctcatc 270

Zc3hl2a gcgaggccacacagatattac 135 cgaaggatgtgctggtctg 271

Zc3hl2c agcgtaatgcgagaaacctc 136 ttctttgtttccatggctca 272

[000185] Splicing analysis: a set of -67,000 previously annotated alternatively spliced events (skipped exons, mutually exclusive splice events) were downloaded (Wang, E. T. et al. Alternative isoform regulation in human tissue transcriptomes. Nature 456, 470-476, doi: 1 0.1 038/nature07509 (2008)). MISO (Katz, Y., Wang, E. T., Airoldi, E. M. & Burge, C. B. Analysis and design of RNA sequencing experiments for identifying isoform regulation. Nature Methods 7, 1009- 1015, doi: 10.1038/nmeth.1528 (2010)) was run with default parameters to estimate the percent spliced in (PSI) for every event in each of the single cells and population replicates. The vast majority of events were not expressed at sufficient depth in any of the samples to be analyzed by MISO. For the remaining 4,338 events it was noted that PSI estimates derived from 10,000 cell replicates were tightly correlated (mean r = 0.91 ). The PSJ values for the three population replicates were averaged and focused the remainder of the analyses on the 352 "alternatively spliced" events (20% < population PS1 average < 80%) in 322 genes (28 genes had at least two alternative splicing events).

[000186] The PSI distribution of these 352 alternative splicing events across single cells was then examined (Fig. 3b). To ensure that only reliable splicing events from highly expressed transcripts were examined, only PSI estimates for single cell/splice event pairs where the alternatively spliced gene was expressed at high levels (single-cell TPM > 250) within that single cell were considered. This resu lted in 89 unique alternative splice events from 79 genes. After applying this filter, a histogram of PSI estimates across single cells (Fig. 3b, top) was plotted. Fig. 3b (bottom) shows a histogram of PSI estimates from the first 10,000-cell replicate for the same 89 splice events from Fig. 3b (top).

[000187] Mice: For the high throughput Examp les provided herein, 6-8 week old female C57BL/6 wild-type (wt), Tnfrsfl a ^{~ _} x Tnfrsfl b^ (Tnfr, IrfT ^7"" , Tirap ^" ' ^" , Ill rn ^"7" , Ikbke ^"7" _, Cxcr2 ^"7" _, Egrl ^" ' ^" , Fas ^" ' ^" , NZBWF 1 /J and Ifhp i -eYFP reporter mice were obtained from Jackson Laboratory (Bar Harbor, ME). Stat 1 and 129/Sv control mice were purchased from Taconic (Hudson, NY). Irf7 ^" ' ^" bone marrow (BM) was provided by Kate Fitzgerald from University of Massachusetts Medical School. Ifnr ^" ' ^" BM was provided by Nir Hacohen from Massachusetts General Hospital. ZFP36 ^"7" (TTP ^"7" ) and control BM were provided by Perry Blackshear from NIH/NIEI IS. Ifnarl -/- (Ifhr KO) bone marrow Nir I lacohen (Massachusetts General Hospital); 1127-/- (I I27r KO) bone marrow as provided by Vijay Kuchroo (Brigham and Women's Hospital).

[000188] All animals were housed and maintained in a conventional pathogen- free facility at the MIT in Cambridge, MA (IUCAC protocol: 0609-058015). All experiments were performed in accordance to the gu idelines outlined by the MIT Committee on Animal Care (Cambridge, MA).

[000189] Cell culture, sorting, and lysis: For the high throughput Examples provided herein, cultures of bone marrow derived dendrit ic cells (BMDCs) from 6-8 week old female B6 mice were prepared as previously described (Am it, I. et al. Unbiased reconstruction of a mammalian transcriptional network med iating pathogen responses. Science 326, 257-263, doi: 10. 1 126/science. l 1 79050 (2009); Chevrier, N. et al. Systematic Discovery of TLR Signaling Components Delineates Viral-Sensing Circuits. Cell 147, 853-867,

doi: 10.1016/j.cell.201 1.10.022 (201 1 )), with minor modification. Namely, isolated bone marrow was frozen down at 5 million (M) cells per mL in pure fetal bovine serum supplemented with 10% DMSO. For each run, a single vial was thawed and cultured as previously described (Ibid). At 9 days of in vitro culture, the cells were labeled with anti- Cd l l c ant ibod ies ( iltenyi Biotech) and flow-sorted, retaining the top 10% of positive cells. Subsequently, the cells were spun down and resuspended in a 15 mL conical tube at a concentration of 2 x 10 ⁵ cells per mL in media supplemented with the relevant stimulus and placed in the incubator with the cap slightly ajar. Stimulants - PAM3CS (Invivogen), Poly(I:C) [PIC] (Enzo Life Sciences, 10 iglmh), LPS (Invivogen, 100 ng/mL), and Interferon-β (Ifn-β) (R&D Systems, 1000 units/mL) - were used as previously described (Ibid). 45 minutes prior to the specific time point, cells were spun down, resuspended at a concentration of 3 x 10^ M cell per mL of complete media supplemented with Hoechst 34580 dye (Life Technologies, according to the manufacturer's recommendations), mixed 7:3 with Ci suspension reagent (Fluidigm), and loaded onto Ci microfluidic chips. After loading, each of the Q microfluidic chip's capture ports were optically inspected for the presence of a cell. The number of cells present in each chamber was determined by counting the number of nuclei. The average single cell capture rate was 72 (average) ± 13 (standard deviation) per chip. Th average number of chambers with two or more cells was 8 ± 7. Although rare multiple capture events were not filtered out automatically (i.e., by computational analysis) in the presented analyses, any specific finding (e.g., 'precocious cells') was confirmed by manual inspection, to ensure that no cell doublet or other cell capture concerns were involved. Similarly, it was explicitly confirmed that the addition of Hoechst 34580 does not alter gene expression in the system provided herein.

Whole transcriptome amplification: After cell isolation, cells were lysed and SMART-Seq {See Ramskold, 201 1 ). Whole Transcriptome Amplified products (WTA) were prepared using the SMARTer Ultra Low RNA Kit for Illumina Sequencing (Clontech) in conjunction with the mRNA-Seq protocol was run on the C I with the following modifications:

Cell Lysis Mix:

Cycling Conditions I :

a) 72°C, 3 min

b) 4 ^u C, 10 min

c) 25°C, 1 min Reverse Transcription (RT) Reaction Mix:

Cycling Conditions II:

a) 42°C, 90 min

b) 70°C, 10 min

PCR Mix:

Cycling Conditions III:

a) 95°C, 1 min

b) 5 cycles of:

i) 95°C, 20s

ii) 58°C, 4 min

ii) 68°C, 6 min

c) 9 cycles of:

i) 95°C, 20s

ii) 64°C, 30s

ii) 68°C, 6 min

d) 7 cycles of:

i) 95°C, 30s

ii) 64°C, 30s

ii) 68°C, 7 min

e) 72°C, 10 min

[000190] Library Preparation and RNA-Seqi The WTA products were harvested from the C I chip and cDNA libraries were prepared using Nextera XT DNA Sample preparation kit (Illumina) as per the manufacturer's recommendations, with minor modifications. Namely, reactions were run at one-fourth the recommended volume and the tagmentation step was extended to 10 minutes. After the PCR step, all 96 samples were pooled without library normalization, cleaned twice with 0.9x AMPure XP SPRI beads (Beckman Coulter), and eluted in buffer TE. Finally, the pooled libraries were quantified using Quant-IT DNA High-Sensitivity Assay Kit (Invitrogen), examined using a high sensitivity DNA chip (Agilent), and run on a MiSeq (Illumina). Finally, samples were sequenced deeply using either a HiSeq 2000 or a HiSeq 2500.

[000191J RNA-Seq of Population Controls^ Population controls were generated by extracting total RNA using RNeasy plus Micro RNA kit (Qiagen) according to the manufacturer's recommendations. Subsequently, 1 μΐ, of RNA in water was added to 2 ί. of lysis reaction mix, thermocycled using cycling conditions I (as above). Next, 4 μί of the RT Reaction Mix were added and the mixture was thermocycled using cycling conditions II (as above). Finally, 1 of the total RT reaction was added to 9 μί of PCR mix and that mixture was thermocycled using cycling conditions III (as above). Products were quantified, diluted to 0.1 25 wg/^L and libraries were prepared, cleaned, and tested as above.

[000192] RNA Fluorescence in situ Hybridization (RNA-Fish)i ^A-V\m (Fig. 27) for Ifitl , Tnf, 116, B2m, and Ifnbl were performed as previously described using probes from Panomics (see e.g., Shalek, A. K. et al. "Single-cell transcriptomics reveals bimodality in expression and splicing in immune cells." Nature 498, 236-240, doi: 10.1038/nature l 2172 (2013)).

[000193] On-Chip Cell Isolation and Simulation: To block cell-to-cell communication, individual BMDCs were stimulated in the C I chip after capture. First, prior to loading the cells, the C I chip was pre-blocked with CI blocking reagent and then with complete culture media for 2h. Next, unstimulated BMDCs were loaded and then washed with complete media supplemented with the appropriate stimulus. After introduction of the stimulus-laced complete media, the chip was maintained at 37°C within the CI System until 30 minutes prior to the specific assay time point (i.e., for 3.5 hours for the 4h stimulation time point). The cells were then washed on chip with media containing Hoechst (Invitrogen), and the chip was removed from the C I System, imaged and run as above at 4h. The 30-minute interval at room temperature (equivalent to our timing of loading of "in tube" samples) accounts for cell wash (15 minutes), imaging (5 minutes), and reagent loading ( 10 minutes) prior to lysis. Lastly, a mock "on-chip" experiment was performed by loading cells as above and then introducing complete media without LPS as above. (000194] Cytokine, addition, GolgiPlug, and Cyclohexiniide experiments: Recombinant 1L-4 (Miltenyi Biotcc), 1L-6 (Miltcnyi Biotcc), IL- 10 (R&D Systems), IL- 12 (Miltenyi Biotec), IL- 15 (Miltenyi Biotec), IL-27 (R&D Systems), IL-35 (AdipoGen) were added as described at 200 ng/mL. GolgiPlug (BD Biosciences) was added at a 1 : 1 ,000 dilution at various lime points. Finally, Cycloheximide was added at 100 μg/mL from a 500x ethanolic stock at the time of stimu lation or during a standard 4h LPS control.

[0001 5] Processing RNA-Seq data: Raw sequencing data were processed as previously described (Shalek, Nature 2013), except that there was no need to trim

SMARTer short and long adapter sequences due to the Nextera library preparation (see e.g., Ramskold, D. et al. Full-length mRNA-Seq from single-cell levels of RNA and individual circulating tumor cells. Nature Biotechnology 30, 777-782, doi: 10.1038/nbt.2282 (2012). Short sequencing reads were aligned to the UCSC mm9 transcriptomc (see e.g., Fujita, P. A. et al. The UCSC Genome Browser database: update 201 1. Nucleic Acids Research, doi: 10.1093/nar/gkq963 (2010). These alignments were used to estimate transcriptomic alignment rates, and were also used as input in RSEM v 1 .12 (Li, B. & Dewey, C. RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome. BMC Bio informatics 12, doi: 10.1 186/ 1471 -2105- 12-323 (201 1)) to quantity gene expression levels (transcripts per million; TPM) for all UCSC mm9 genes in all samples. Genomic mappings were performed with Tophat v. 1.41 (Trapnell, C, Pachter, L. & Salzberg, S. L. TopHat: discovering splice junctions with RNA-Seq. B io in ormatics 25, 1 105- 1 1 1 1 , doi: 10.1093/bioinformatics/btpl 20 (2009)), and the resulting alignments were used to calculate genomic mapping rates, rRNA contamination, and 3' and 5 ' positional bias (PicardTools). All genes that were not expressed at appreciable levels (ln(TPM+l ) > 1 ) in at least 1 % of all single cells were discarded, leaving 10,313 genes for all further analyses.

[000196] Determining statistically significant associations between clusters and principal components (PCs): In order to determine which modules were significantly associated with the primary sources of variability in the data as defined by the PCs, a recently developed statistical resampling approach (Chung, N. C. & Storey, J. D. Statistical significance of variables driving systematic variation. arXiv, doi: uuid/22B6DA41 -E02D- 423F-87BC-21 1091 235 A5 1 (2013)) was used to determine genes which were associated with the first three PCs. Briefly, F-statistics were calculated for each immune response gene by independently zeroing out the contribution of each gene to the first three PCs, and examining the change in variance explained by the modified PCs. Then, a small number of random rows (n-5) in the matrix were permuted, and F-statistics were calcu lated for these synthetic null variables. This procedure was repeated 1 ,000 times to generate a set of nu ll statistics. To assess the statistical significance of each module, a one sided Mann- Whitney test was performed.

[000197] Fitting parametric models of gene expression variation: The nominal parameters α, σ2, and μ were estimated for each gene in each condition by fitting a series of nested statistical models to its expression distribution (Fig. 26a,b). All presentations were focused on the LPS response, where genes from the most modules are induced. First, it was tested whether the single cell expression distribution of immune response genes was compatible with a (μ, σ ² ) unimodal lognormal distribution, as has been previously used in the literature to describe single-cell distributions of gene expression (see e.g., Raj, A., Peskin, C. S., Tranchina, D, Vargas, D. Y. & Tyagi, S. in PLoS Biol Vol. 4 e309 (2006)). For each gene in each condition, the mean and variance of all log(TPIVB-l ) values was calculated, and a goodness-of-fit test was used to test a lognormal distribution with these parameters. Only a very small minority (2.5%) of distributions was wel l described by the two-parameter model, primarily due to the inflation of zero values in our single cell data.

[000198] Next, each single-cell gene expression distribution was parameterized by estimating values for α, σ ² , and μ. Each distribution corresponds to the observed expression values across single cells for a given gene in a given condition. The expression threshold was set at ln(TPM+l) > 1 , as it was observed that levels of expression in the range 0 < ln(TPM+l ) < 1 typically reflected very few reads that mapped to exonic sequences, and these could likely signify small amounts of contamination. Thus a values were estimated as the proportion of cells where transcript expression was detected at levels (ln(TPM+l) > 1 ). The mean (μ) and variance (σ ² ) was then calculated in log-space of all expression values where ln(TPM+l ) > 1 . The fit of this three-parameter model was assessed using an add itional goodness-of-fit test. It was found that the majority (92%) of distributions were well described by the three-parameter (μ, σ ² , ) model (p < 0.01 , goodness of fit test).

EXAMPLE 2. Stimulation of Bone Marrow Derived Dendritic Cells (BMCDs) with Lipopolysaccharide (LPS)

[000199] To characterize the extent of expression variability on a genomic scale and decipher its regulatory and functional implications, single-cell RNA-Seq was used to study heterogeneity in the response of BMDCs to LPS stimulation. BMDCs are an attractive model system for single-cell analyses for several reasons. First, LPS, a component of gram- negative bacteria and a ligand of Toll-like receptor 4, is a physiologically relevant, uniform stimulus that synchronizes cellular responses and mitigates temporal phasing (Tay, S . et al. S ingle-cell NF-κΒ dynamics reveal digital activation and analogue information processing. Nature 466, 267-271 , doi:papers2://publication/do i/10.1038/nature09145 (2010)). Second, activation by LPS evokes a robust transcriptional program for inflammatory and antiviral cytokines, and many of the components controlling this response are known from

'population-wide' studies (Takeuchi, O. & Akira, S. Pattern Recognition Receptors and Inflammation. Cell 140, 805-820, doi: 10.1016/j.celI.2010.01.022 (2010)). Third, LPS stimulation should increase the synchronization between mRNA and protein levels for induced genes, reducing a potentially confounding factor for single-cell analyses

(Taniguchi, Y. et al. Quantifying E. coli Proteome and Transcriptome with Single-Molecule Sensitivity in Single Cells. Science 329, 533-538, doi: 10.1 126/science.1 188308 (2010), Li, G.-W. & Xie, X. S. Central dogma at the single-molecule level in living cells. Nature 475, 308-315 (201 1 )). Lastly, differentiated DCs from bone marrow cultures are post-mitotic, largely removing the effects of the cell cycle (Kalisky, T., Blainey, P. & Quake, S. R.

Genomic Analysis at the Single-Cell Level. Annual review of genetics 45, 431 -445.

doi:papers2://publication/doi/10.1 146/annurev-genet- 102209- 163607 (201 1 ); Ramos, C. A. et al. Evidence for Diversity in Transcriptional Profiles of Single Hematopoietic Stem Cells. PLoS Genetics 2, el 59, doi:papers2://publication/doi/10.1371 /journal.pgen.0020159.st008 (2006)).

[000200] BMDCs with LPS were stimulated and single cells were harvested after four hours (Amit, I. et al. Unbiased reconstruction of a mammalian transcriptional network mediating pathogen responses. Science 326, 257-263, doi: 10.1 126/science. l 179050 (2009); Chevrier, N. et al. Systematic Discovery of TLR Signaling Components Delineates Viral- Sensing Circuits. Cell 147, 853-867, doi: 10.1016/j.cell.201 1.10.022 (201 1 )) (Example 1). Using SMART-Seq (Ramskold, D. et al. Full-length mR A-Seq from single-cell levels of RNA and individual circulating tumor cells. Nature Biotechnology 30, 777-782, doi:papers2://publication/doi/ 10.1038/nbt.2282 (2012)), cDNA libraries derived from 18 single BMDCs (S 1 -S I 8) were constructed, three replicate populations of 10,000 cells, and two negative controls (empty wells). Each of these libraries was sequenced to an average depth of 27-million read-pairs per sample. As expected, less than 0.25% of reads from the negative control libraries aligned to the mouse genome, and these samples were discarded from all further analyses. Library quality metrics (Levin, J. Z. et al. Comprehensive comparative analysis of strand-specific RNA sequencing methods. Nature Methods 7, 709- 715 (201 0)), such as alignment rates to the genome, ribosomal RNA contamination, and 3 ' or 5 ' coverage bias, were similar across all single-cell libraries and 10,000-cell replicates. For each sample, expression levels were calculated for all UCSC-annotated genes (Hsu, F. et al. The UCSC Known Genes. Bioinformatics (Oxford, England) 22, 1036- 1046, doi: 1 0.1093/bioinformatics/btl048 (2006)) using RSEM (Li, B. & Dewey, C. N. RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome. BMC Bioinformatics 12, 323-323 (201 1)) (Example 1 ), and discarded all genes that were not expressed at appreciable levels (transcripts per million (TPM) > 1 ) in at least three individual cells, retaining 6,313 genes for further analysis.

[000201] While gene expression levels of population replicates were tightly correlated with one another (Pearson r > 0.98, log-scale, Fig. l a), there was substantial variation in gene expression profiles between individual cells (0.29 < r < 0.62, mean: 0.48, Fig. l b, Fig. 5). Despite this extensive cell-to-cell variation, expression levels for an "average" single cell - derived by averaging transcript expression levels over all 1 8 single cells - correlated well (0.79 < r < 0.81 ) with the population samples (Fig. l c, Fig. 5). This observation confirms that the significant gene expression differences observed between single cells do average together to form the population profile.

{000202] RNA-fluorescence in situ hybridization (FISH), a single molecule imaging technique that does not require amplification (Yu, M. et al. RNA sequencing of pancreatic circulating tumour cells implicates WNT signalling in metastasis. Nature 487, 510-513, doi: 10.1038/naturel 1217 (2012); Raj, A., Rifkin, S. A., Andersen, E. & Van Oudenaarden, A. Variability in gene expression underlies incomplete penetrance. Nature 463, 913-918, doi: 10.1038/nature08781 (2010)), was used to verify that the heterogeneity in single-cell expression reflects true biological d ifferences, rather than technical noise associated with the amplification of a small amount of cellular RNA. For 25 genes, selected to cover a wide range of expression levels, variation in gene expression levels detected by RNA-FISH closely mirrored the heterogeneity observed in the sequencing data (Fig. l d-h, Fig. 6). For example, the expression of classical housekeeping genes (e.g. , Beta-Actin (Actb), Beta-2- microglobulin (B2m)) matched a log-normal distribution in both single-cell RNA-Seq and RNA-FISH measurements, consistent with previous studies (Bengtsson, M. Gene expression profiling in single cells from the pancreatic islets of Langerhans reveals lognormal distribution of mRNA levels. Genome Research 15, 1388- 1392, doi:papers2://publication/doi/10.1 101/gr.3820805 (2005)). In contrast, many genes involved in the BMDC response to LPS, although highly expressed on average, exhibited significantly greater levels of heterogeneity that do not fit a log-normal distribution. In extreme cases, the expression levels of these genes varied up to -1 ,000 fold between individual cells (Fig. le-h). More generally, it was found that high levels of variability in single-cell gene expression persisted across a wide range of population expression levels (Fig. 2a).

[000203] In particular, 281 of the 522 most highly expressed genes (single-cell average TPM > 250, Table S3) exhibited low variability, and their expression levels were well described by log-normal distributions across single cells (RNA-Seq: Fig. 2b,c top, RNA-FISH (Actb, B2m): Fig. 6). These 281 genes are enriched for housekeeping genes, encoding ribosomal and other structural proteins (Bonfen-oni-corrected ρ=1.5* 10 ^~δ ). This is consistent with previous observations in yeast (Newman, J. R. S. & Weissman, J. S.

Systems biology: many things from one. Nature 444, 561 -562 (2006); Bar-Even, A. et al. Noise in protein expression scales with natural protein abundance. Nature Genetics 38, 636- 643 (2006)) and human (Ram, O. et al. Combinatorial Patterning of Chromatin Regulators Uncovered by Genome-wide Location Analysis in Human Cells. Cell 147, 1628-1639 (201 1)) cells that highly expressed housekeeping and 'growth' genes are less variable between cells.

[000204] Surprisingly, however, most of the other highly expressed genes exhibited a bimodal expression pattern ( 185 of 241 highly variable genes, Fig. 2b,c bottom): mRNA levels for these genes were high in many of the cells, but were at least an order of magnitude lower than the single-cell average in at least three cells, where abundances were often very low or undetectable. This variation was independently verified by RNA-FISH (e.g. , Cxcll , Cxcll O, Ifit l , and others: Fig. 6), confirming that it is not a result of technical noise. This variable set was highly enriched for genes that were induced by at least two-fold upon LPS stimulation at the population level (Garber, M. et al. A High-Throughput Chromatin Immunoprecipitation Approach Reveals Principles of Dynamic Gene Regulation in Mammals. Molecular Cell 47, 810-822, doi: 10.1016/j.molcel.2012.07.030 (2012)) (p = 2.7x 10 ^"7 ; hypergeometric test), and included both antiviral and inflammatory response elements, suggesting that such widespread variability amongst highly expressed genes might be a feature of the immune response. While bimodal expression patterns characterize many immune response transcripts, some immune response genes were highly expressed in every cell (Fig. 7), demonstrating that all cells robustly responded to LPS. These include key chcmokincs and chemokine receptors (Ccl3, Ccl4, Ccrl2), cytokines (Cxcl2), and other important components of the LPS response (Tank).

[000205] This degree of variation in highly expressed transcripts has not been observed in previous studies (Islam, S. et al. Characterization of the single-cell transcriptional landscape by highly multiplex RNA-seq. Genome Research,

doi:papers2://publication/doi/10. 1 101/gr. l 10882.1 10 (201 1 ); Tang, F. et al. RNA-Seq analysis to capture the transcriptome landscape of a single cell. Nature Protocols 5, 5 16- 535, doi: 10.1038/nprot.2009.236 (2010); Tang, F. et al. mRNA-Seq whole-transcriptome analysis of a single cell. Nature Methods 6, 377-382, doi: 10.1038/nmcth. l 315 (2009); Ramskold, D. et al. Full-length mRNA-Seq from single-cell levels of RNA and individual circulating tumor cells. Nature B iotechnology 30, 777-782,

doi:papers2://publication/doi/10.1038/nbt.2282 (2012); Hashimshony, T., Wagner, F., Sher, N. & Yanai, I. CEL-Seq: Single-Cell RNA-Seq by Multiplexed Linear Amplification. Cell Reports, doi: 1 0. l 0 l 6/j.celrep.20l 2.08.003). For example, far less heterogeneity was found in expression for highly abundant (population average) genes in a published SMART-Seq dataset of eight human embryonic stem cells (Ramskold, D. et al. Full-length mRNA-Seq from single-cell levels of RNA and ind iv idual circulating tumor cells. Nature Biotechnology 30, 777-782, doi:papers2://publication/doi/10.1038/nbt.2282 (2012)) (Fig. 2a), or in single cell RNA-Seq datasets from terminally differentiated mouse embryonic fibroblasts and mouse embryonic stem cells (Hashimshony, T., Wagner, F., Sher, N. & Yanai, I. CEL-Seq: S ingle-Cell RNA-Seq by Multiplexed Linear Amplification. Cell Reports,

doi: 10.1016/j.celrep.2012.08.003) (Fig. 8). In addition, such bimodality in (on average) highly expressed genes was not observed in genome-scale studies of variation in protein expression in mid-log yeast cells and dividing human cell lines (Newman, J. R. S. & Weissman, J. S. Systems biology: many things from one. Nature 444, 561 -562 (2006); Bar- Even, A. et al. Noise in protein expression scales with natural protein abundance. Nature Genetics 38, 636-643 (2006); Sigal, A. et al. Variabil ity and memory of protein levels in human cells. Nature 444, 643-646, doi: 10.1038/nature05316 (2006)). It was thus hypothesized that the observed bimodal ity may reflect functionally important differences in the stimulated BMDC population.

[000206] Furthermore, splicing patterns across single cells also show previously unobserved levels of heterogeneity: for genes that have multiple splice isoforms at the population level, individual cells predominantly express one particular isoform. The f equency (percent spliced in, PSI) of previously annotated splicing events in each of the samples was calculated using MISO (Katz, Y., Wang, E. T., Airoldi, E. M. & Burge, C. B. Analysis and design of RNA sequencing experiments for identifying isoform regulation. Nature Methods 7, 1009-1015 (2010)), a Bayesian framework for calculating isoform ratios. Surprisingly, although the population-derived estimates were highly reproducible, single cells exhibited significant variability in exon-inclusion frequencies (Fig. 3a,b).

[000207] The possibility that the PCR amplification steps (intrinsic to the library preparation process) could potentially result in an overestimation of isoform regulation variability, particularly for weakly expressed transcripts, due to 'jackpot effects'

(Shiroguchi, K., Jia, T. Z., Sims, P. A. & Xie, X. S. Digital RNA sequencing minimizes sequence-dependent bias and amplification noise with optimized single-molecule barcodes. Proceedings of the National Academy of Sciences of the United States of America 109, 1347- 1352 (2012)) was carefully considered. However, it was found that, even when the analysis was limited to 89 alternatively spliced exons (0.2 < population PSI < 0.8) that were very highly expressed within a single cell (single cell TPM > 250), the same bimodality in splicing patterns amongst individual cells was still observed, with highly skewed expression of one or the other splice variant instead of simultaneous expression of both at comparable levels (Fig. 3b).

[000208] To further control for the possibility that stochastic overamplification of a few mo lecules could confound the splicing analyses, three additional single cell cDNA libraries were created using a slightly modified SMART-Seq protocol (Example 1 ) in which a four nucleotide barcode was introduced onto each RNA molecule during reverse transcription. This barcode was retained through PCR amplification and library preparation, allowing us to quantify the number of unique RNA transcripts that are represented in the sequencing library (Fig. 9 and Example 1). Even when limiting the splicing analyses to genes that were represented by at least 15 unique barcodes, a strong bias in isoform expression in single cells was observed compared to population averages (Fig. 10).

[000209] To date, single-cell variation in splicing patterns has rarely been studied even for single genes, and never analyzed at a genomic scale. One recent report (Waks, Z., Klein, A. M. & Silver, P. A. Cell-to-cell variability of alternative RNA splicing. Molecular Systems Biology 7, 1 - 12, doi:papers2://publ ication/doi/10.1038/msb.201 1.32 (201 1 )) used RNA-FI SH to study variation in alternative isoforms in two genes, and observed lower levels of isoform variability across single cells (the levels of heterogeneity differed in different cell types). Another study using fluorescent reporters to quantify single-cell exon inclusion levels observed highly variable and bimodal splicing patterns for one gene (Gurskaya, N. G. et al. Analysis of alternative splicing of cassette exons at single-cell level using two fluorescent proteins. Nucleic Acids Research 40, doi: 10.1093/nar/gkrl 314 (2012)).

[000210] To independently verify the extensive differences in isoform ratios between cells, RNA-FISH probes targeting constitutive and isoform-specific exons in two genes (Irf7 and Acpp, Fig. 3c,d) (Waks, Z., Klein, A. M. & Silver, P. A. Ccll-to-ccll variability of alternative RNA splicing. Molecular Systems Biology 7, 1- 12,

doi:papers2://publication/doi/10.1038/msb.201 1.32 (201 1)) were designed. Substantial expression variability in overall Irf7 levels was found between individual cells (as reflected by the 'constitutive' probes, Fig. 3c, bottom and top panels), mirroring the single-cell sequencing results (and further explored below). Additionally, within each Irf7-expressing cell, a bias toward either the inclusion or exclusion of the specific exon (Fig. 3c, Fig. 1 1, middle panel, e.g. , compare 'high' and 'low' marked cells) was observed. Similar results were obtained using two probes designed to detect mutually exclusive alternative final exons for Acpp (Fig. 3d). Thus, these studies demonstrate that splicing heterogeneity is a common mode of variation between single cells, a phenomenon often masked by the 'simultaneous expression' of alternative isoforms observed in population studies.

EXAMPLE 3. Sources and Implications of Observed Bimodalities

[000211] The studies described herein were designed to explore the sources and functional implications of the observed bimodalities. The enrichment in immune response genes amongst highly (on average), yet bimodally, expressed genes may reflect either distinct functional states {e.g. , cell subtypes) or stochastic differences in the activation of signaling circuits (Tay, S. et al. Single-cell NF-κΒ dynamics reveal digital activation and analogue information processing. Nature 466, 267-271 ,

doi:papers2://publication/doi/10.1038/nature09145 (2010)), in promoter events (Sanchez,

A., Garcia, H. G., Jones, D., Phillips, R. & Kondev, J. Effect of Promoter Architecture on the Cell-to-Cell Variability in Gene Expression. PLoS Comput Biol 7, e l 001 100-el OO l 100

(201 1 )), or in response timing (Nachman, I., Regev, A. & Ramanathan, S. Dissecting timing variability in yeast meiosis. Cell 131 , 544-556 (2007)). First, it was hypothesized that at least some of the variation may reflect distinct cell states in the in vitro differentiated BMDCs. In particular, it has been previously reported that BMDCs can acquire distinct maturation states through a developmental process in which BMDCs switch from antigen- capturing to antigen-presenting cells in order to prime the adaptive immune system

(Banchereau, J. et al. Immunobiology of Dendritic Cells. Annual Review of Immunology 18, 767-81 1 (2000)). Maturation can occur either in response to pathogen-derived ligands, such as LPS, or as a result of disrupting clusters of DCs in culture (Jiang, A. et al. Disruption of E-Cadherin-Mediated Adhesion Induces a Functionally Distinct Pathway of Dendritic Cell Maturation. Immunity 27, 610-624,

doi:papers2://publication doi 10. 1016/j. immuni.2007.08.015 (2007)), both leading to up- regulation of specific cell surface markers. The induction of cytokines that occurs in response to LPS represents an even more mature state of BMDCs.

[000212] To test how much, if any, of the transcriptional variation in immune response genes is due to distinct maturity states, an unbiased principal components analysis (PCA, Fig. 4a) was performed on the single-cell expression profiles, focusing on the 632 genes that were induced at least two-fold in the population-wide response to LPS (Garber, M. et al. A High-Throughput Chromatin Immunoprecipitation Approach Reveals Principles of Dynamic Gene Regulation in Mammals. Molecular Cell 47, 810-822,

doi: l 0.1016/j.molcel.2012.07.030 (2012)). At least two distinct subpopulations of cells were found within the dataset, clearly distinguishable by the first principal component (PC I , 15% of the total variation, Fig. 4a). One group of fifteen cells expressed a set of both antiviral and inflammatory cytokines (including: Tnf, Ili a, 111 , and Cxcl lO) at extremely high levels (TPM > 1 ,000), whereas a second group of three cells expressed far lower, albeit detectable, levels of each of these genes (TPM < 50). Other markers, such as Ccr7, Cd83, Serpinb9, and Ccl22, showed the opposite expression pattern (Fig. 4b, Fig. 12). Many of the genes that distinguish these two groups encode cell surface proteins (e.g. , Cd83, Cd86, and Ccr7) that have been previously identified as markers of BMDC maturation. These observations suggest that the two subpopulations of 15 and 3 cells represent distinct stages of DC maturation: cells with high expression of Cd83, Cd86 and Ccr7 and low expression of cytokines resemble 'semi-mature DCs' or cluster-disrupted DCs (Jiang, A. et al.

Disrupt ion of E-Cadherin-Med iated Adhesion Induces a Functionally Distinct Pathway of Dendritic Cell Maturation. Immunity 27, 610-624,

doi:papers2://publication/doi/10. 1 016/j.immuni.2007.08.015 (2007); Lutz, M. B.

Therapeutic potential of semi-mature dendritic cells for tolerance induction. Frontiers in immunology 3, 123, doi:papers2://publication/doi/] 0.3389/fimnni.20 ] 2.00123 (201 2)), whi le those with high expression of cytokines represent 'maturing or mature DCs'. In addition, two of the 1 5 maturing cells (Fig. 12) express higher levels of transcripts encoding both cytokines and surface markers, suggesting that these cells are the most mature DCs (Fig. 5).

[000213] The existence of semi-mature and maturing BMDCs in the single cells were validated in several ways. First, the same semi-mature/maturing groupings were verified with RNA-F1SH (Fig. 13), and also with single-cell quantitative reverse-transcription polymerase chain reaction (qRT-PCR: Fluidigm BioMark HD) using a signature of 96 genes selected to cover different expression levels and each of the first two principal components (Fig. 1 1 , Table S6) (Dalerba, P. et al. Single-cell dissection of transcriptional heterogeneity in human colon tumors. Nature Biotechnology 29, 1 120- 1 127,

doi:1 0.1038/nbt.2038 (201 1 )). Second, subsets of Cdl l c+ BMDCs were sorted based on the presence or absence of each of 1 1 cell surface markers whose mRNA levels in the single cell RNA-Seq discriminate between the maturing and semi-mature groups. qRT-PCR was then used in each pair of sorted populations to measure mRNA levels for the ten marker genes that also discriminate the two groups in the sequencing data, for example, Tnf and Cxcl l O (highly expressed in the maturing subpopulation) and Ccl22 and Serpinb9 (highly expressed in the semi-mature subpopulation). Indeed, for pairs of populations sorted by 8 of 1 1 cell surface markers, the expected differences in marker expression levels were detected, confirming the sequencing- based classification (Fig. 15). These results further validate the sensitivity of single-cell RNA-Seq, demonstrating how it can effectively distinguish between closely related, yet distinct, maturity states, even within the same cell type.

EXAMPLE 4. Role of variation in Regulatory Circuits Amongst Cells in the Same Cell State

[000214] S ince distinct maturity states explain only a small portion of the observed heterogeneity and bimodality, the role of variation in regulatory circuits amongst cells in the

'same' cell state (e.g. , the 1 5 maturing cells) was examined next. It was reasoned that if such variable circuits exist, co-variation across single cells between the mRNA levels of a transcription factor and the expression of its targets would represent a potential regulatory interaction, and furthermore, would suggest that the variation in the regulator's expression underlies the variability of its targets. Such a correlative approach has successfully identified regulatory connections from population-level transcription profiles measured in different cond itions (Amit, I. et al. Unbiased reconstruction of a mammalian transcriptional network mediating pathogen responses. Science 326, 257-263, doi: 10. 1 126/science. l 179050 (2009); Nachman, I., Regev, A. & Friedman, N. in Bioinformatics Vol. 20 (248 (2004)). Here, the studies were designed to apply it to multiple single cells in the same condition.

[000215] To this end, the correlation in expression profiles between every pair of induced genes across all single cells was calculated, and a cluster of 137 genes that varied in a correlated way across the cells was identified (Fig. 4b). The cluster's genes were highly enriched for members of the antiviral response (Am it, I. el al. Unbiased reconstruction of a mammalian transcriptional network med iating pathogen responses. Science 326, 257-263, doi: 10. 1 126/science. 1 179050 (2009)) (60 of 1 37 genes, p = 2.5 * 1 0 ^" \ hypergeometric test) and included the transcripts encoding two known master regulators of the antiviral response, Stat2 and Irf7. The cluster was also enriched for Stat2 targets, as were previously determined by ChlP-Seq in DCs stimulated with LPS (Garber, M. et al. A High-Throughput Chromatin lmmunoprecipitation Approach Reveals Principles of Dynamic Gene Regulation in Mammals. Molecular Cell 47 ^' , 810-822, doi: 10. 1016/j.molcel.201 2.07.030 (2012)) (73/137 genes, p = 4.5* 10 ^"5 , hypergeometric test). Genes in this 'antiviral cluster' were strongly discriminated by the second principal component of the PCA (PC2, 8% of the variation, Fig. 4a,b). The correlations between these antiviral genes were validated using both single-cell qRT-PCR (the same 96 gene signature as above) and RNA-FISH (Fig. 4c,d). Notably, most ( 100/137) of the cluster's genes exhibited bimodal expression across the cells (Fig. 2c, bottom) and were strongly expressed at the population level ( 13 genes TPM > 250; 53 genes TPM > 50).

[000216] To further characterize how the variation in the antiviral circuit may change during the response, single-cell qPCR expression profiling was performed for a signature of 13 genes (nine antiviral cluster genes, two uniformly induced genes, and two housekeeping controls) in unstimulated BMDCs and at 2h, 4h, and 6h post-LPS stimu lation (Fig. 16). The percentage of cells expressing the antiviral cluster genes increased in a time-dependent manner (Fig. 16), and was mirrored by changes in the fraction of cells that exhibit high mRNA levels for antiviral master regulators. In contrast, the uniformly induced genes (Cxcl l O, Clec4e) were robustly induced after two hours in all cells. Importantly, the quantitative correlations between the expression levels of the transcripts encoding master regulators and the downstream target genes existed in both the 4h and 6h time points. EXAMPLE 5. Differences in Levels of Stat2 and Irf7

[000217] Having observed that the use of this anti-viral response circuit is highly variable between BMDCs of the same maturity state, it was hypothesized that bimodal variation in the express ion of the cluster's genes may be related to differences in the levels of Stat2 and Irf7. In this case, it would be expected that perturbing these master regulators in BMDCs would result in reduced expression and variation in their targets. To test this hypothesis, expression of the signature genes was measured using single-cell qRT-PCR in LPS-stimulated cells from Irf7 knockout (Irf7 -/-) mice. As expected, this perturbation ablated the transcription of most signature genes in the variable antiviral cluster, while leaving constitutive elements of the antiviral response relatively unaffected (Fig. 4e).

However, Stat2 expression and variability levels were unaffected by the Irf7 knockout, implying that Stat2 may act either upstream or in parallel to Irf7 during the response (Ning, S., Huye, L. E. & Pagano, J. S. Regulation of the Transcriptional Activity of the IRF7 Promoter by a Pathway Independent of Interferon Signaling. Journal of B iological Chemistry 280, 12262- 12270 (2005); Ousman, S. S., Wang, J. & Campbell, I. L.

Differential regulation of interferon regulatory factor (IRF)-7 and IRF-9 gene expression in the central nervous system during viral infection. Journal of Virology 79, 7514-7527 (2005)). Because both Stat2 and Irf7 are targets of the interferon-signaling pathway, the effect of interferon feedback on the expression and variation of Stat2, Irf7 and the cluster genes were tested next. Indeed, when BMDCs from interferon receptor knockout (Ifnr -/-) mice (Darnell, J. E., Jr., Kerr, I. M. & Stark, G. R. Jak-STAT pathways and transcriptional activation in response to IFNs and other extracellular signaling proteins. Science (New York, N.Y.) 264, 1415- 1421 (1994); Gough, D. J. et al. Functional crosstalk between type 1 and II interferon through the regulated expression of STAT 1. PLoS biology 8, el 000361 - e l 000361 (2010)) were stimulated, drastically reduced expression for both Stat2 and Irf7, as well as all other cluster genes was observed (Fig. 4f).

[000218] One possibility is that earlier variation in Stat2 levels underlies the extensive variation in the anti-viral cluster at 4 hours, including in the Stat2 transcript itself (via autoregulation (Garber, M. et al. A High-Throughput Chromatin Immunoprecipitation Approach Reveals Principles of Dynamic Gene Regu lation in Mammals. Molecular Cell 47, 810-822, doi: 10.101 6/j.molcel.2012.07.030 (2012))). For example, while the majority of immune response genes (e.g., Ifitl ) were not expressed in unstimulated cells, the Stat2 transcript is variably expressed even prior to LPS stimulation (Fig. 1 ). Cells with high levels of Stat2 prior to stimu lation may be the most likely to express the antiviral cluster at the 4h time point.

[000219] To further examine this link, cells were co-stained for Ifit l , Stat l , and Stat2 mRNAs and Stat l , pStatl, and Stat2 proteins (Example 1 , Fig. 17 & 18), and quantified these mRN A/protein levels and protein localization in B DCs simu lated with LPS for 0, 2, and 4 hrs. While overall protein levels increased in all cases throughout the time course, substantial heterogeneity was found in the induction of Stat l , pStat l , and Stat2 (Fig. 17). At 2 hr, all three proteins showed heterogeneity in both their expression and nuclear translocation. By 4hr, protein levels were more homogeneous, and nuclear translocation was less pronounced. Ifit l mRNA distributions displayed highly similar patterns, exhibiting more bimodal expression at early time points that became more uniform by 4h. However, Stat protein and Ifitl mRNA levels within individual cells were not correlated early (0.00 < r ² < 0. 12), and only very weakly correlated at four hours (0.00 < r ² < 0.28). This may be due to the fact that a target's mRNA accumulation reflects the integrated spatiotemporal activity of a transcriptional regulator, which may not be well represented by a single temporal snapshot (Cai, L., Dalai, C. K. & Elowitz, M. B. Frequency-modulated nuclear localization bursts coordinate gene regulation. Nature 455, 485-490, doi:nature07292

[pii] 10.1038/nature07292 (2008)). Thus, in cells with high Ifit l mRNA levels, Stat proteins may already have left the nucleus. Validating such a hypothesis requires real-time tracing of protein and multiple transcripts simultaneously (Cohen, A. A. et al. Dynamic Proteomics of Individual Cancer Cells in Response to a Drug. Science 322, 151 1 - 15 16,

doi: 10.1 126/science. 1 160165 (2008)), a task significantly complicated by difficulties of adding endogenous fluorescent tags in primary immune cells (Shalek, A. . el al.

Nanowire-Mediated Delivery Enables Functional Interrogation of Primary Immune Cells: Application to the Analysis of Chronic Lymphocytic Leukemia. Nano Lett 12, 6498-6504, doi:papers2://publication/doi/10.1021/nl3042917 (2012)), and to the Stat proteins specifically (Meyer, T., Begitt, A. & Vinkemeier, U. Green fluorescent protein-tagging reduces the nucleocytoplasmic shuttling specifically of unphosphorylated STAT 1 . GFP- tagging ofSTATl 274, 815-826, doi:papers2://publication/doi/10.1 1 1 1 /j.1742- 4658.2006.05626.x (2007)). Conversely, even at 4h, Ifit l mRNA levels correlated better with Stat l and Stat2 mRNA than their protein levels (Fig. 18). Since Stat proteins autoregulate their own gene expression (Garber, M. et al. A High-Throughput Chromatin I mmunoprecipitat ion Approach Reveals Principles of Dynamic Gene Regulation in Mammals. Molecular Cell 47, 810-822, doi: 10.1 016/j.molcel.2012.07.030 (201 2)), this is consistent with the hypothesis of an earlier regu latory event.

EXAMPLE 6. High Throughput Microfluidic-Enabled Single Cell RNA-SEO

[000220] The trillions of cells in complex eukaryotes are canonically grouped in tissues and organs, and further subdivided into types that share molecules, structures and functions. In recent years, however, it has become increasingly apparent that even functionally ' identical' cells can be markedly different in their component molecules (Taniguchi, Y. et al. Quantifying E. coli Proteome and Transcriptome with Single-Molecule Sensitivity in Single Cells. Science 329, 533-538, doi: 10.1 126/science.1 188308 (2010); Tay, S. et al. Single-cell NF-κΒ dynamics reveal digital activation and analogue information processing. Nature 466, 267-271 , doi:papers2://publication/doi/10.1038/nature09145 (201 ); Raj, A. & Van Oudenaarden, A. Single-Molecule Approaches to Stochastic Gene Expression. Annual Review of Biophysics 38, 255-270,

doi: 10.1 146/annurev.biophys.37.032807.125928 (2009); Cohen, A. A. et al. Dynamic

Proteomics of Individual Cancer Cells in Response to a Drug. Science 322, 151 1 -1516, doi: 10.1 126/science. l 160165 (2008); Altschuler, S. J. & Wu, L. F. Cellular Heterogeneity:

Do Differences Make a Difference? Cell 141 , 559-563, doi: 10.1016/j.cell.2010.04.033

(2010); Warren, L., Bryder, D., Weissman, I. L. & Quake, S. R. Transcription factor profiling in individual hematopoietic progenitors by digital RT-PCR. Proceedings of the

National Academy of Sciences of the United States of America 103, 17807- 17812, doi: 10.1073/pnas.0608512103 (2006); Paszek, P. et al. Population robustness arising from cellular heterogeneity. Proceedings of the National Academy of Sciences of the United

States of America 107, 1 1644- 1 1649, doi: 10.1073/pnas.0913798107 (2010); Slack, M. D.,

Martinez, E. D., Wu, L. F. & Altschuler, S. J. Characterizing heterogeneous cellular responses to perturbations. Proceedings of the National Academy of Sciences 105, 19306-

1931 1, doi: 10.1073/pnas.0807038105 (2008); Niepel, M., Spencer, S. L. & Sorger, P. .

Non-genetic cell-to-cell variability and the consequences for pharmacology. Curr. Opin.

Chem. Biol. 13, 556-561 , doi: 10.1016/j.cbpa.2009.09.015 (2009); Sharma, S. V. et al. A chromatin-mediated reversible drug-tolerant state in cancer cell subpopulations. Cell 141 ,

69-80, doi: 10.1016/j.cell.2010.02.027 (2010); Gascoigne, K. E. & Taylor, S. S. Cancer cells display profound intra- and interline variation following prolonged exposure to antimitotic drugs. Cancer cell 14, 1 1 1 - 122, doi: 10.101 6/j.ccr.2008.07.002 (2008) and that this heterogeneity can result in substantially different responses to external stimuli (Cohen. A. A. et al. Dynamic Proteomics of Individual Cancer Cells in Response to a Drug. Science 322, 1 51 1 - 151 6, doi: 10.1 126/science. l 160165 (2008); Niepel, M., Spencer, S. L. & Sorger, P. K. Non-genetic cel l-to-cell variabi lity and the consequences for pharmaco logy. Curr. Opin. Chem. B io l. 13, 556-561 , doi: 10.1016/j.cbpa.2009.09.015 (2009); Sharma, S. V. et al. A chromatin-mediated reversible drug-tolerant state in cancer cell subpopulations. Cell 14 1. 69-80, doi: 10. 1016/j.cell.2010.02.027 (2010); Gascoigne, K. E. & Taylor, S. S. Cancer cells display profound intra- and interline variation following prolonged exposure to antimitotic drugs. Cancer cell 14, 1 1 1 - 122, doi: 10.1016/j.ccr.2008.07.002 (2008); Spencer, S. L., Gaudet, S., Albeck, J. G., Burke, J. M. & Sorger, P. K. Non-genetic origins of cell-to-cell variability in TRAIL-induced apoptosis. Nature 459, 428-432, doi: 10.1038/nature08012 (2009)). While such variability can prove detrimental in the case of therapeutic intervention (Cohen, A. A. et al. Dynamic Proteomics of Individual Cancer Cells in Response to a Drug. Science 322, 1 51 1 - 15 16, doi: 10.1 126/science. l 160165 (2008); Altschuler, S. J. & Wu, L. F. Cellular Heterogeneity: Do Differences Make a Difference? Cell 141 , 559-563, doi: 10.1016/j.cell.2010.04.033 (2010); Spencer, S. L., Gaudet, S., Albeck, J. G., Burke, J. M. & Sorger, P. K. Non-genetic origins of cell-to-cell variability in TRAIL-induced apoptosis. Nature 459, 428-432, doi: 10.1 038/nature08012 (2009); Spencer, S. L. & Sorger, P. K. Measuring and Modeling Apoptosis in Single Cells. Cell 144, 926-939,

doi: 10.1016/j.cell.201 1.03.002 (201 1 )) it likely plays an important functional role by increasing the diversity of potential population-level responses (Feinerman, O. et al. Single- cell quantification of IL-2 response by effector and regulatory T cells reveals critical plasticity in immune response. Molecular Systems B iology 6, 1 - 16,

doi:papers2://publication doi/10. 1038/msb.2010.90 (2010); Veening, J.-W., Smits, W. K. & uipers, O. P. Bistability, Epigenetics, and Bet-Hedging in Bacteria. Annual Review of Microbiology 62, 193-210, doi:papers2://publication/doi/10.1 146/

annurev.micro.62.081307.163002 (2008); Locke, .1. C. & Elowitz, M. B. Using movies to analyse gene circuit dynamics in single cel ls. Nature reviews. Microbiology 7, 383-392, doi: 10.1038/nrmicro2056 (2009); Thattai, M. & van Oudenaarden, A. Stochastic gene expression in fluctuating environments. Genetics 167, 523 (2004); Beaumont, H. J., Gallie, J., Kost, C, Ferguson, G. C. & Rainey, P. B. Experimental evolution of bet hedging. Nature 462, 90-93, doi: 10.1038/nature08504 (2009); Chalancon, G. et al. Interplay between gene expression noise and regulatory network architecture. Trends in genetics : TIG 28, 221 -232, doi: 10.1016/j.tig.2012.01.006 (2012)). [000221] The immune system is a well-established example of this: although immune cells are notoriously heterogeneous in their types and functions (Bendall, S. C. & Nolan, G. P. Single-Cel l Mass Cytometry of Differential I mmune and Drug Responses Across a Human Hematopoietic Continuum. Science (New York, NY) 332, 677-678,

doi: 10.1 126/science.1206351 (201 1 ); Hashimoto, D., Miller, J. & Merad, M. Dendritic Cell and Macrophage Heterogeneity In Vivo. Immunity 35, 323-335,

doi:papers2 ://pi]blication/doi/ 10.1016/j. immuni.201 1.09.007 (201 1 )), they must collectively generate appropriate responses to pathogens. Understand ing the strategies used to encode population-level behaviors, as well as when they fail and at what expense, is a fundamental biological problem with substantial clinical relevance. Recent molecular studies have demonstrated the potential for single cell approaches to unveil the informing mechanisms, normally masked by technical and biological noise, with sufficient sampling (Cohen, A. A. et al. Dynamic Proteomics of Individual Cancer Cells in Response to a Drug. Science 322, 1 51 1 -1516, dor.10.1 126/science. l 160165 (2008); Altschuler, S. J. & Wu, L. F. Cellular Heterogeneity: Do Differences Make a Difference? Cell 141 , 559-563,

doi: 10.1016/j.cell.2010.04.033 (2010); Niepel, M., Spencer, S. L. & Sorger, P. K. Non- genetic cell-to-cell variability and the consequences for pharmacology. Curr. Opin. Chem. Biol. 13, 556-561 , doi: 10.1016/j.cbpa.2009.09.015 (2009); Sharma, S. V. et al. A chromatin-mediated reversible drug-tolerant state in cancer cell subpopulations. Cell 141 , 69-80, doi: 10.1016/j.cell.2010.02.027 (2010); Spencer, S. L., Gaudet, S., Albeck, J. G., Burke, J. M. & Sorger, P. K. Non-genetic origins of cell-to-cell variability in TRAIL- induced apoptosis. Nature 459, 428-432, doi: 10.1038/nature08012 (2009); Feinerman, O. et al. Single-cell quantification of IL-2 response by effector and regulatory T cells reveals critical plasticity in immune response. Molecular Systems Biology 6, 1 - 16,

doi:papers2 ://publication/doi 10.1038/msb.2010.90 (2010); Bendall, S. C. & Nolan, G. P. S ingle-Cell Mass Cytometry of Differential Immune and Drug Responses Across a Human Hematopoietic Continuum. Science (New York, NY) 332, 677-678,

doi: 10.1 126/science. l 206351 (201 1 ))). Nevertheless, the majority of these studies have focused - by necessity - on well-characterized markers with available reagents and known roles, hindering unbiased discovery of the determinants of immune responses.

[000222] Emerging single cell genomics methods now open the possibility of using sequencing-based approaches to profile the behaviors of single cells in unprecedented detail (Islam, S. et al. Characterization of the single-cell transcriptional landscape by highly multiplex RNA-seq. Genome Research, doi:papers2://publication/doi/ 10.1 101/gr. l 10882.1 10 (201 1 ); Tang, F. et al. RNA-Seq analysis to capture the transcriptome landscape of a single cell. Nature Protocols 5, 516-535,

doi: 10.1038/nprot.2009.236 (2010); Tang, F. et al. mRNA-Seq whole-transcriplome analysis of a single cell. Nature Methods 6, 377-382, doi: 10.1 038/nmeth. l 31 5 (2009); Ramskold, D. et al. Full-length mRNA-Seq from single-cell levels of RNA and individual circulating tumor cells. Nature Biotechnology 30, 777-782,

doi:papers2://publication/doi/10.1038/nbt.2282 (2012); Hashimshony, T., Wagner, F., Sher, N. & Yanai, I. CEL-Seq: Single-Cell RNA-Seq by Multiplexed Linear Amplification. Cell Reports, doi: 10.1016/j.celrep.2012.08.003). In principle, genome-wide single cell approaches could help determine, ab initio, new cell classification schemes, transitional states, unrecognized biological distinctions, molecular circuits, and the like. Fulfilling this potential requires the development of new experimental strategies for achieving the scale needed to address the high levels of noise inherent in single-cel l measurements (Chalancon, G. et al. Interplay between gene expression noise and regulatory network architecture. Trends in genetics : TIG 28, 221-232, doi: 10.101 6/j.tig.2012.01.006 (2012); Newman, J. R. S. et al. in Nature Vol. 441 840-846 (2006); Munsky, B., Neuert, G. & van Oudenaarden, A. Using Gene Expression Noise to Understand Gene Regulation. Science (New York, NY) 336, 183- 187, doi: 10.1 126/science.1216379 (201 2); Balazsi, G., Van Oudenaarden, A. & Collins, J. J. Cellular Decision Making and Biological Noise: From Microbes to Mammals. Cell 144, 910-925, doi: 10. 101 6/j.cell.201 1.01.030 (201 1 )) - both technical, due to minute amounts of input material, and biological, due to bursts of RNA transcription (Taniguchi, Y. et al. Quantifying E. coli Proteome and Transcriptome with Single-Molecule Sensitivity in Single Cells. Science 329, 533-538, doi: l 0.1 1 26/science. l 1 88308 (2010); Cai, L., Dalai, C. . & Elowitz, M. B. Frequency-modulated nuclear localization bursts coordinate gene regulation. Nature 455, 485-490, doi:papers2://publication/doi/10.1038/nature07292 (2008)).

[000223] Integrated microfluidic circuits present an elegant solution for surmounting this obstacle. Indeed, methodological precedent exists for performing each of the steps implicated in a single cell whole transcriptome (WTA) amplification protocol within a microfluidic device (Taniguchi, Y. et al. Quantifying E. coli Proteome and Transcriptome with Single-Molecule Sensitivity in Single Cells. Science 329, 533-538,

doi: 10.1 126/science. l 1 88308 (2010); Tay, S. et al. Single-cell NF-κΒ dynamics reveal digital activation and analogue information processing. Nature 466, 267-271 ,

doi:papers2 ://publ ication/doi/1 0.1038/nature09145 (2010); I Jong, J. W., Studer, V., Hang, G., Anderson, W. F. & Quake, S. R. A na no liter- scale nucleic acid processor with parallel architecture. Nature Publishing Group 22, 435-439, doi: l 0. l038/nbt95 1 (2004); Huang, B. et al. Counting Low-Copy Number Proteins in a Single Cell. Science (New York, NY) 315, 81 -84, doi: 10.1 126/science. 1 1 33992 (2007); Marcus, J., Anderson, W. & Quake, S.

Microfluidic single-cell mRNA isolation and analysis. Analytical Chemistry 78, 3084-3089 (2006); Melin, J. & Quake, S. R. Microfluidic Large-Scale Integration: The Evolution of Design Rules for Biological Automation. Annual Review of Biophysics and Biomolecular Structure 36, 213-231 , doi: 10.1 146/annurev.biophys.36.040306.132646 (2007); Amit, I. et al. Unbiased reconstruction of a mammalian transcriptional network mediat ing pathogen responses. Science 326, 257-263, doi: 10.1 126/science. 1 179050 (2009)), including cell capture, imaging, lysis, reverse transcription, and amplification (PCR). In this study, a commercially available microfluidic system (C I Single Cell Auto Prep System, Fluidigm) was adapted to prepare single-cell SMART-seq mRNA transcriptome libraries. The system isolates up to 96 individual cells, applies multi-step molecular bio logy protoco ls to each isolated cell, and then outputs the reaction product to an SBS-format well on the chip carrier. The SMART-Seq (Ramskold, D. et al. Full-length mRNA-Seq from single-cell levels of RNA and individual circulating tumor cells. Nature Biotechnology 30, 777-782, doi:papers2://publication/doi/10.1038/nbt.2282 (2012)) double-stranded cDNA generated form each cell are then converted to Illumina sequencing libraries.

[000224] Single Cell RNA-Seq Profiling of Thousands of Bone Marrow Dendritic Cells: the Fluidigm C I Single-Cell Auto Prep System was utilized, combined with a high- throughput cDNA library construction protocol, to generate RNA-Seq ready libraries from a total 2000-3000 single Bone Marrow-Derived Dendritic cells (BMDCs) (Toriello, N. et al. Integrated microfluidic bioprocessor for single-cell gene expression analysis. Proceedings of the National Academy of Sciences 105, 201 73 (2008): Chevrier, N. et al. Systematic Discovery of TLR Signaling Components Delineates Viral-Sensing Circuits. Cell 147, 853- 867, doi: 10. 1016/j.cell.201 1 . 1 0.022 (201 1 ): Garber, M. et al. A High-Throughput

Chromatin Immunoprecipitation Approach Reveals Principles of Dynamic Gene Regulation in Mammals. Molecular Cell 47, 810-822, doi: 10.1016/j.molcel.2012.07.030 (2012)). BMDCs represent a good model system for studying single cell responses since they are primary, well-characterized at the population level, post-mitotic, and can be synchronized through the addition of a strong pathogenic stimulus (oriello, N. et al. Integrated microfiuidic bioprocessor for single-cell gene expression analysis. Proceedings of the National Academy of Sciences 105, 20173 (2008)). The previous Examples examin ing response variability between 18 'homogenous' stimulated, single BMDCs did not allow for the examination of the evolution of noise and its molecular determinants. Moreover, the focus on one stimulus prevented the profiling and contrasting of circuit activation and heterogeneity across different stimuli.

[000225] The studies described herein were designed to address these questions. First, genome-wide mRNA expression responses were profiled at five time po ints (0, 1 , 2, 4, & 6hr) after activating BMDC Toll-Like Receptor (TLR) signaling with three distinct pathogenic stimuli (Chevrier, N. et al. Systematic Discovery of TLR Signaling Components Delineates Viral-Sensing Circuits. Cell 147, 853-867, do i: 10.1016/j.ce 11.201 1.10.022 (201 1 )) - lipopolysaccharide (LPS ; a component of gram- negative bacteria and TLR4 agonist), Polyinosinic:polycytidylic acid (Poly(LC), PIC; viral-like double stranded RNA and TLR3 agonist), and PAM3CSK (PAM; a synthetic mimic of bacterial lipopeptides and TLR2 agonist). For each condition, a single CI IFC, capturing up to 96 cells (average 85 ± 1 0%,) was run, and libraries were also generated from 10,000 cells (population control,). In all, 3 1 1 , 212, and 146 cells responding to LPS, PIC, and PAM, respectively, as well as ~ 4000 additional cells (described below) were profiled.

[000226] Each of these samples were sequenced to an average depth of 10 million read pairs, and expression estimates (transcripts per million; TPM) were calculated for all UCSC-annotated genes using (Li, B. & Dewey, C. N. RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome. BMC

Bio informatics 12, 323-323 (201 1 )). The obtained libraries were of consistently high quality, comparable to published SMART data (Ramskold, D. et al. Full-length mRNA-Seq from single-cell levels of RNA and individual circulating tumor cells. Nature Biotechnology 30, 777-782, doi:papers2://publication/doi/10.1038/nbt.2282 (2012); Shalek, A. . et al. Nanowire-mediated delivery enables functional interrogation of primary immune cells: application to the analysis of chronic lymphocytic leukemia. Nano Lett. 12(12):6498-504. doi: 10. 1 021 /nl3042917 (2012)). Median transcriptomic mapping rates were -50-60%, while median genomic mapping rates were—70-80%. A significant fraction of reads (~ 10%) failed to map due to contaminating adaptor sequence which cannot be trimmed, suggesting that the cDNA libraries are of even higher quality than appears from the transcriptomic mapping rates. Meanwhile, 3 ' bias levels were higher than had been observed previously, but were very similar to those previously published from Nextera data (available on the Illumina website).

[000227] Expression-wise, the single cell measurements agreed closely when aggregated and compared with data from a cell population generated using a similar protocol. The correlations between in silico single-cell average RNA-Seq data and population measurements were high (R= -0.9, Fig. 24a). This represents an improvement over correlations observed for comparisons between two different library construction methods for replicates of the same bulk-population sample. (Levin, J. Z. et al.

Comprehensive comparative analysis of strand-specific RNA sequencing methods. Nature Methods 7, 709-715 (2010)). This degree of correlation was robust across the expression spectrum. The correlations tended to plateau once around 30 cells had been included in other in silico single-cell average.

[000228] Genes were clustered based on their d ifferential temporal responses to these three stimuli (Fig. 24b) within the population level sam les. Population based

measurements agreed closely with, and refined (described below), previously run microarray-based experiments (Chevrier, N. et al. Systematic Discovery of TLR Signaling Components Delineates Viral-Sensing Circuits. Cell 147, 853-867,

doi: 10.1016/j.cell.201 1 .10.022 (201 1)). In particular, the analysis recapitulated several previously-discovered clusters that were highly enriched for targets of NF-kB

(inflammatory program, Clusters VI, VII), as well distinct clusters highly enriched for interferon responsive genes (Cluster 1,11) (Fig. 24b). Broadly, while antiviral genes were typically "late-induced" at both the population and single cell levels, most inflammatory response genes were sharply peaked early (at l-2hrs). Still, there was a set of late- induced inflammatory genes (Cluster VI) that peaked late.

EXAMPLE 7. Variation Between Cells During Immune Response

[000229] Refinement of Cell Circuits from Single Cell data: From this broad definition of population-level pathways, higher resolution structure was investigated by sub-clustering genes based on their expression values in single cells, (black lines, Fig. 24b). In concert with the cluster analyses, an unbiased principal components analysis (PCA) was also performed on all -800 single cells in the timecourse dataset.

[000230] It was discovered that the high-resolution data allowed genes to be assigned to a refined set of circuits that could not be distinguished at the population level. For example, while all antiviral genes exhibited population-level induction at later timepoints after exposure to LPS and PIC, a cluster of 102 genes (Cluster I D) was observed that were distinguished not only based on their overall induction levels, but also from coherent expression within subsets of single cells (Supp. Figure). While genes in this module exhibit dramatic enrichment for antiviral and interferon response genes, genes in clusters 1A-1 C do not exhibit similar functional patterns. Genes in cluster I D are also strongly distinguished by their contribution to the first principal component (PC I ) in the PCA analysis. Thus, cluster I D was termed to represent the "core" antiviral response of BMDCs. Notably, the separation between core and non-core antiviral genes is not readily apparent from population level measurements and was not observed in either previous RNA-Seq (Amit, 1. et al. Unbiased reconstruction of a mammalian transcriptional network mediating pathogen responses. Science 326, 257-263, doi: 10. 1 126/scicncc. 1 179050 (2009) or microarray (Chevrier, N. et al. Systematic Discovery of TLR Signaling Components Delineates Viral- Sensing Circuits. Cell 147, 853-867, doi: 10.1016/j.cell.201 1.10.022 (201 1)) experiments.

[000231] Similarly, it was observed that the inflammatory program, broadly denoted by high projection scores of the second principal component (PC2), could be separated into multiple distinct circuits. Many canonical inflammatory markers (i.e., TNF, ILIA, CXCL2), exhibit "sharp peaked" responses to LPS (Takeuchi, O. & Akira, S. Pattern Recognition Receptors and Inflammation. Cell 140, 805-820, doi: 10.1016/j.celL2010.01 .022 (2010), cluster 3c) - these genes are sharply induced early and are downregulated at later timepoints in the response. Other clusters shared between LPS and PAM (clusters 3b,d), in contrast, exhibit increased levels of induction throughout the timecourse. While these two clusters appear highly similar from population level measurements, cluster 3b genes are marked by strong projection scores for the third principal component (PC3) and are highly enriched for markers of dendritic cell maturation, in particular cell surface markers, receptors and transporters (CD83, CD86, CCR7) and cytokines (CCL 17 and CCL22) which arc essential for proper communication with and activation of T cells. These genes are highly and induced in the response to LPS, but only in a distinct subset of cells.

[000232] The refined single cell circuits allow for the identification of novel molecular regulators which may play key roles in the immune response. For example, while the "maturation" cluster (denoted by high projection scores for PC3) contains many well known markers of BMDC maturation, the remainder of the genes in the signature compromise a rich list of transcription factors, G protein coupled receptors, lincRNAs, and transmembrane proteins whose strong single-cell correlations with known maturity markers implicates their role in activating the adaptive immune system. Many of these genes are do not have characterized roles in BMDC maturation, or even in the regu lation of immune response.

[000233] Others, such as the transcription factor IRF8 and the transmembrane protein TMEM39A, have been significantly associated with autoimmune disease via unknown mo lecular mechanisms. S imilarly, the refinement of a "core" antiviral module highlights the potential role of previously uncharacterized regulators, including nuclear-dot associated proteins (ex. Spl OO and Sp l40), chromatin regulators (ex. Phf] 1 ), putative transcriptional regulators (ex. Znfxl) and ubiquitin ligascs (ex. Dtx31).

[000234] Temporal and developmental heterogeneity are defined by a continuous spectrum: The principal components analysis indicates that, rather than separating into multiple distinct subgroupings, the dendritic cells represent individual points on a continuous landscape of cellular variation. For example, while the first principal component broadly separates single cells based on their stimulation time point, there is significant spread between PC I loadings for cells within any given dataset (Figure 24d).

[000235] This is particularly true early in the response ( 1 and 2hr), which is clearly separated from later timepoints as the cells begin to synchronize their core antiviral response four hours post-stimulation. In contrast to antiviral response, however, it is seen that the diversity in maturity state between single cells steadily increases during the duration of the LPS time course. While the identified circuit is only induced in a subset cells, the highly variable levels of induction result in a continuous range of intermediate states (Figure 24 f). These studies were unable to identify clearly defined, discrete subpopulations after performing separate PCA analyses on each of the three stimu lation timecourses, or even on each individual timepoint, highlighting the continuous nature of ingle cell noise observed in the system. This is likely a reflection of the experimental system having been chosen upfront as a homogenous, post-mitotic, and synchronized population of immune cells.

[000236] Parameterization of single cell data: In the previous analysis of 1 8 individual BMDC transcriptomes, extensive bimodality was observed in individual gene expression levels between single cells, observing that most transcripts were not detected in every cell either by RNA-Seq or RNA-FISH. The scale of the current experiment, however, provides sufficient scale to begin to model and parameterize single cell data from a single condition. Thus, the studies here attempted to fit a series of nested statistical models to each single cell distribution, initially focusing the efforts on the LPS response genes. While a small percentage (-5%) of transcripts were well described by a unimodal log-norma l distribution (parameterized by the mean, urn, and the standard deviation, sigma), the remainder benefited statistically (likelihood ratio test, P<0.01 ) from the introduction of a third parameter (alpha) which defined the percentage of cells expressing the transcript at non-negligible levels (ln(TPM)> l ). This explicit parameterization of single cell data as a bimodal distribution allows us to break single cell heterogeneity into two components: one level of variability is represented by the percentage of cells expressing a transcript (parameterized by alpha, as referred to herein, this is digital noise), a second layer reflects the spread in RNA levels amongst expressing cells (parameterized by sigma, which is referred to herein as analogue noise).

[000237] The vast majority (-80-90%, goodness of fit test, SM) of single cell distributions were well described by this three parameter, explicitly bimodal, distribution, implying that the new parameterization could be broadly applied to analyze changes in single cell noise systematically. Interestingly, the majority (70-80%) of transcripts that did not fit the three-parameter distribution at one LPS timepoint were well described by a mixture model of normal distributions and also failed the goodness-of-fit test at another timepoint, suggesting the existence of multiple regulated "bursting states" for these genes.

[000238] Quantitative chromatin levels agree with single cell noise parameters: While strong correlations between mRNA levels and chromatin states at have been well described (see e.g., Ram, O. et al. Combinatorial Patterning of Chromatin Regulators Uncovered by Genome-wide Location Analysis in Human Cells. Cell 147, 1628-1639 (201 1 ); Garber, M. et al. A High-Throughput Chromatin Immunoprecipitation Approach Reveals Principles of Dynamic Gene Regulation in Mammals. Molecular Cell 47, 810-822,

doi: 10.1016/j.molcel.2012.07.030 (2012) the single cell data here allow for the reanalysis of this relationship at a new level. Population maps of histone marks, often assayed with ChlP- seq, exhibit a wide quantitative range. Since chromatin marks are either present or absent from a DNA mo lecule, it was reasoned that quantitative chromatin measurements of active marks at a promoter should correlate with the digital noise levels of a gene, i.e. the percentage of cells expressing the transcript, rather than the overall population expression level. Indeed, a strong relationship was observed between the alpha parameter of the single cell distributions for a gene and the population level of K27 present at the gene's promoter, even after controlling for population expression level. In stark contrast, no relationship was observed between the population level mRNA expression and quantitative chromatin levels after controlling for the percentage of cells expressing the transcript. These relationships were robust for the active chromatin mark K27ac as well as RNA Poll I levels, but not for the H3K4me3, in line with previous observations that K27ac is more tightly correlated with active transcription.

[000239] Distinct heterogeneity profiles of immune response circuits: The parameterization of single cell distributions were applied to analyze changes in the single cell helerogeneity of immune response circu its across experimental conditions. This study started by examining the structure of the core antiviral program, which is typically classified as "late-induced" from population studies (Amit, 1. et al. Unbiased reconstruction of a mammalian transcriptional network mediating pathogen responses. Science 326, 257-263, doi: 10.1 126/science. l 179050 (2009)) and identifying substantial bimodality during a snapshot of the response in previous work (Chevrier, N. et al. Systematic Discovery ofTLR Signaling Components Delineates Viral-Sensing Circuits. Cell 147, 853-867,

doi: 10.1016/j.cell.201 1.10.022 (201 1 )). The most significant single cell patterns in the antiviral response occurred between the two hour and four hour timepoints, as key antiviral genes shifted their expression patterns from bimodal to unimodal across single cells. This shift in digital noise, however, is accompanied by a significant reduction in analogue noise- again with the most dramatic shifts in all parameters occurring between two and four hours (median sigma shift==0.6 to 0.9, pvalue=3.5xl 0 ^A~5 ). Thus, single cells tightly synchronize their antiviral response during the observed timecourse, exhibited by robust and tightly regulated expression of core antiviral genes at later timepoints.

[000240] Genes participating in the inflammatory program tend to display starkly opposite temporal heterogeneity profiles compared to their antiviral counterparts. In particular, genes exhibiting sharp peaked responses (cluster IIIc)- including canonical antiinflammatory cytokines such as Il i a and TNF-alpha were sharply induced at early timepoints, but are downregulated later in the response. The exact cause of this temporal dephasing is unknown, although cross-inhibitory feedback loops and RNA degradation factors may be responsible for creating a peaked response. Remarkably, it was observed that the dynamics of these bulk expression estimates are due almost entirely to changes in digital noise. While the percentage of cells expressing these transcripts exhibited significant change between all temporal transitions, parameters representing the distribution of expressing cells were statistically unchanged throughout the response- including at the unstimulated timepoint. [000241] A distinct cluster of inflammatory genes (cluster 11 Id) are continually induced over the timecourse, exhibiting patterns of digital noise that are similar to the core antiviral cluster- again with the most significant shift occurring between two and four hours. In contrast to antiviral synchronization, however, no change was observed in the analogue noise of this circuit. Thus while, late-induced antiviral and inflammatory genes show similar temporal profiles at the population level in LPS, the two responses exhibit different heterogeneity profiles at the 4h timcpoint, with the former resembling a tightly regulated circuit while the latter exhibits a noisier induction. Taken together, these analyses highlight the vastly different temporal heterogeneity patterns of functionally distinct LPS response modules, and exemplify the ability of single cell RNA-seq to distinguish both tightly regulated and noisy circuits.

[000242] Changes in single cell noise across stimuli: It has been previously noted that while PIC and PAM are specific antagonists of the antiviral and inflammatory pathways respectively, LPS is capable of activating both defense programs in BMDC populations (Takeuchi, O. & Akira, S. Pattern Recognition Receptors and Inflammation. Cell 140, 805- 820, dok l O. 101 6/j.cel 1.2010.01.022 (2010).). Given the non-specific nature ofTLR4 signaling, it was hypothesized that immune response circuits may behave differently in response to a more directed stimulus.

[000243] For example, it was hypothesized that exposure to PIC may reduce single cell heterogeneity in the antiviral cluster. It was observed, however, that antiviral temporal heterogeneity patterns were slightly delayed in the PIC timecourse in comparison to LPS. In particular, the core genes transitioned from bimodal to unimodal expression between the four and six hour timepoints, and the delay in antiviral synchronization indicated that PIC in fact acted as a weaker stimulus. These observations are in line with previous reports.

[000244] The temporal variability patterns of inflammatory circuits, however, differed greatly after exposure to PAM. As in the LPS response, sharp peaked response genes exhibited a sharp induction in the percentage of expressing cells at early timepoints. These genes, however, tend to "plateau' instead of "peak" at the two hour timepoint, and failed to desynchronize at later timepoints (no statistically significant change in either digital or analogue noise). Likewise, it was found that inflammatory circuits began to synchronize (significant reduction in analogue noise between T=2hr and 4h, p-value=.0014), their response at later timepoints, similar to the antiviral core circuit during the LPS response. The changing temporal noise patterns of these circuits after exposure to distinct stimuli strongly argues that single cell heterogeneity is not purely a consequence of unconstrained transcriptional slochasticty, but is instead a controlled phenomenon that is regulated during immune response. The next studies thus further investigated the role of both intracellular and intercellular determinants in driving single cell variability.

EXAMPLE 8. Environmental Determinants Of Temporal Noise

[000245] While variable levels of internal components can drive differences in response phenotype (Taniguchi, Y. et al. Quantifying E. coli Proteome and Transcriptome with Sin le-Molecule Sensitivity in Single Cells. Science 329, 533-538.

dok l O. l 126/science. l 188308 (2010); Tay, S. et al. Single-cel l NF-κΒ dynamics reveal digital activation and analogue information processing. Nature 466, 267-271,

do ^' r.papers2://publication/doi/10.1038/nature09145 (2010); Raj, A. & Van Oudenaarden, A. Single-Molecule Approaches to Stochastic Gene Expression. Annual Review of Biophysics 38, 255-270, doi: 1 0.1 146/annurev.biophys.37.032807.125928 (2009); Cohen, A. A. et al. Dynamic Protcomics of Individual Cancer Cells in Response to a Drug. Science 322, 151 1 - 1516, doi: 10.1 126/science. l 160165 (2008); Altschuler, S. J. & Wu, L. F. Cellular

Heterogeneity: Do Differences Make a Difference? Cell 141 , 559-563,

doi: 10.1016/j.cell.2010.04.033 (2010); Warren, L., Bryder, D., Weissman, I. L. & Quake, S. R. Transcription factor profiling in indiv idual hematopoietic progenitors by digital RT-PCR. Proceedings of the National Academy of Sciences of the United States of America 103, 17807-17812, doi: 10.1 073/pnas.0608512103 (2006); Paszek, P. et al. Population robustness arising from cellular heterogeneity. Proceedings of the National Academy of Sciences of the United States of America 107, 1 1644- 1 1649, doi: 10.1073/pnas.0913798107 (2010); Slack, M. D., Martinez, E. D., Wu, L. F. & Altschu ler, S. J. Characterizing heterogeneous cellular responses to perturbations. Proceedings of the National Academy of Sciences 105, 19306- 1931 1 , doi: 10.1 073/pnas.0807038105 (2008); Niepel, M., Spencer, S. L. & Sorger, P. K. Non-genetic cell-to-cell variability and the consequences for pharmacology. Curr. Opin. Chem. Biol. 13, 556-561 , doi: 10.1016/j.cbpa.2009.09.015 (2009); Sharma, S. V. et al. A chromatin-mediated reversible drug-tolerant state in cancer cell subpopulations. Cell 141 , 69-80, doi: 10. 1016/j.cell.2010.02.027 (2010); Gascoigne, K. E. & Taylor, S. S. Cancer cells display profound intra- and interline variation following prolonged exposure to antimitotic drugs. Cancer cell 14, 1 1 1 - 122, doi: 1 0.1 016/j.ccr.2008.07.002 (2008)), local differences in the cellular inicroenvironinent can afford an external, confounding source of heterogeneity (Fan, R. et al. Integrated barcode chips for rapid, multiplexed analysis of proteins in microliter quantities of blood. Nature Biotechnology 26, 1373-1378, doi: 10.1038/nbt.1507 (2008); Gomez-Sjoberg, R., Leyrat, A., Pirone, D., Chen, C. & Quake, S. Versatile, fully automated, microfluidic cell culture system. Analytical Chemistry 79, 8557-8563 (2007); Huang, S. Non-genetic heterogeneity of cells in development: more than just noise.

Development 136, 3853-3862, doi:papers2://publication/doi/10.1242/dev.035139 (2009): alisky, T., Blainey, P. & Quake, S. R. Genomic Analysis at the Single-Cell Level. Annual review of genetics 45, 43 1 -445, doi:papers2://publication/do i/10.1 146/annurev-genet- 102209-163607 (201 1); Lecault, V. et al. High-throughput analysis of single hematopoietic stem cell proliferation in microfluidic cell culture arrays. Nature Methods 8, 581 -586, doi:papers2://publication/doi/10.1038/nmeth. l 614 (201 1); Loewer, A. & Lahav, G. We are all individuals: causes and consequences of non-genetic heterogeneity in mammalian cells. Current opinion in genetics &amp; development 21 , 753-758,

doi: 10.1016/j.gde.201 1.09.010 (201 1 ); Millet, L. J., Stewart, M. E., Sweedler, J. V., Nuzzo, R. G. & Gillette, M. U. Microfluidic devices for culturing primary mammalian neurons at low densities. Lab on a Chip 7, 987, doi: 10.1039/b 705266a (2007); Raser, J. M. Control of Stochasticity in Eukaryotic Gene Expression. Science (New York, NY) 304, 181 1 - 1814, doi: 10.1 126/science.1098641 (2004)). The response of each BMDC is dominated by the expression of mRNAs for cytokines and chemokines, that can, in turn, activate additional intracellular signaling pathways. Thus, heterogeneous intercellular signaling, coupled with slow diffusion, could easily give rise to a rich local diversity in environmental conditions, forcing each cell to compute its response under different constraints.

[000246] Uniform interferon stimulus removes bimodality from antiviral response: It was previously hypothesized (Shalek, A. K. et al. Nanowire-mediated delivery enables functional interrogation of primary immune cells: application to the analysis of chronic lymphocytic leukemia. Nano Lett. 12( 12):6498-504. doi: 10.102 l/nl3042917 (2012)) that variability in a secondary wave of interferon (IFN) signaling was responsible for the widespread bimodality that was observed in the antiviral response at the 4h timepoint. To test this further, BMDCs were stimulated directly with IFN-β so as to provide all of the cells with equal access to this antiviral feedback. At 2hr after stimulation (equivalent to a 4h LPS stimulus since IFN-β peaks under LPS at 2hr (Amit, I. et al. Unbiased reconstruction of a mammalian transcriptional network mediating pathogen responses. Science 326, 257-263, doi: 10.1 126/science.1 179050 (2009))), a dramatic shift in the digital noise of the antiviral cluster was observed, with key genes shifting from a bimodal expression distribution in LPS to a unimodal one under lFN-β. This finding suggests that early heterogeneity in the expression of antiviral genes might not arise from intracellular timing differences (Nachman et al., Dissecting timing variability in yeast meiosis. Cell 131 , 544-556 (2007)) but rather from differences in IFN-β exposure. This, coupled with the early rise in Ifnbl mRNA expression seen from 0 to 2h under LPS stimulation in a select set of cells, suggests that a subset of cells may, in fact, be responsible for generating a primary wave of interferon signaling, which eventually synchronizes the antiviral response as the lFN-β enshrouds the entire population. In such a case, it would be expected the cells that produced IFN-β and their nearest neighbors to exhibit early antiviral induction due to autocrine and paracrine signaling.

[000247] A rare population of cells precociously expresses late-induced antiviral genes at early limepoints: In support of this hypothesis, three cells that exhibited precocious expression of antiviral response genes after only 1 hi of LPS stimulation were intriguing ly discovered. These cells could be clearly distinguished by robust expression the general antiviral signature, including Ifitl , as well as by their projection across the second principle component. To verify the existence of this population, RNA-FISH was performed, co- staining cells for expression of Ifitl and Ifnbl . Thus, this population exists, but it is a rare population.

[000248] Ablation of paracrine signaling dramatically alters cellular heterogeneity: While highly suggestive, the discovery of the "early responder" subpopulation does not definitively show that intracellular signaling is required for antiviral synchronization in the population. Validating this hypothesis requires methods for isolating cells and culturing them individually. In the absence of paracrine signaling, the former hypothesis would suggest a shift in the digital antiviral noise.

[000249] To accomplish this, unstimulated BMDCs were loaded and isolated onto the C I IFC, and proceeded to stimulate each cell with LPS individually inside the sealed microfluidic chamber. To closely mirror the standard stimulation protocols, the CI system was programmed to deliver LPS-laced media via one of the IFC's washing ports and then incubated the cells at 37°C for four hours prior to normal imaging, lysis, and cDNA synthesis and amplification. Importantly, the cell density for this on-chip stimulation ( 1 cell per 4.5 nL) tightly matched the normal, in tube, stimulations ( 1 cell per 5 nL), enabling direct comparison of this experiment with the existing LPS data. As originally

hypothesized, the absence of paracrine signaling strongly desynchronized the antiviral response. A dramatic increase in digital noise was observed as antiviral gene distributions shifted from unimodal (bulk LPS stimulation) to bimodal (on-chip stimulation). Notably, a subset of cells - likely analogous to the identified early responders, did exhibit robust activation of the core antiviral response. Similarly, the ablation of paracrine signaling severely restricted the maturation process for all BMDCs, ablating expression of maturation markers in all cells. This is likely due to the abrogation of TNF-mediated signaling, which is known to drive maturation in BMDCs. Still, not all induced genes behaved different in the absence of paracrine signaling: many late- induced inflammatory genes were unaffected, demonstrating that isolated cells were capable of undergoing a natural response to LPS in a micro fluidic chamber.

[000250] To test this, paracrine signaling was ablated by isolated and then stimulated individual BMDCs for 4hr inside of the C I IFC. To match the normal activating conditions, the CI system was programmed to deliver LPS-laced media via one of the IFC's washing ports and then incubated the cells at 37°C for the duration of the stimulation, before imaging and lysing as normal. Importantly, the cell density for the on chip stimulation ( 1 cell per 4.5 nL) tightly matched the normal, in tube, stimulations ( 1 cell per 5 nL), enabling direct comparison of the two. As originally hypothesized, the absence of paracrine signaling strongly desynchronized the antiviral response. A dramatic increase in digital noise was observed as limited coherent induction of key antiviral markers in a small subset of cells shifted the antiviral gene distributions from unimodal to bimodal. Importantly, not all induced genes behaved different in the absence of paracrine signaling: many late-induced inflammatory genes were unaffected, demonstrating that isolated cells were capable of undergoing a natural response to LPS in a microfluidic chamber.

[000251] While the presence of paracrine signaling is necessary for antiviral synchronization, intracellular communication has the opposite effect on other immune response circuits. Surprisingly, ablation of paracrine signaling dramatically reduced both digital and analogue noise after four hours of LPS stimulation. Canonical inflammatory markers such as TNF, Ili a, and INHBA all shifted from bimodal to unimodal distributions upon paracrine ablation - resembling their uniform expression at the two-hour timepo int. Thus, the results strongly point to paracrine signaling as an upstream determinant of this desynchronization, and highlight the extensive - and, at times, opposing - roles that intercellular communication performs in driving heterogeneity in both the antiviral and inflammatory pathways. [000252] Interferon feedback increases inflammatory heterogeneity: Since the on-chip iso lation experiment bluntly abrogates all paracrine signaling, it cannot discern the ind ividual, or combination of, paracrine signals which are responsible for the resu lts observed above. To more specifically address the roles of individual signaling pathways, this study turned to profiling knockout mice deficient for specific receptor molecules. To better understand the upstream source of inflammatory noise, this study began by profiling BMDCs from mice deficient for TNF receptor. Consistent with previous findings and hypotheses, TNFR -/- BMDCs exhibited no induction of maturation markers. However, many sharply peaked response genes exhibited highly similar distributions in both the wild type and TNFR -/- BMDCs at the four-hour timepoint. Similar results were seen when profiling BMDCs deficient for 1L1 receptor; BMDCs failed to mature, but coherent changes were not observed amongst sharp peaked response genes.

[000253] BMDCs from interferon receptor knock-out (Ifnarl -/-) mice were next profiled. As expected, and in accordance with previous findings (Shalek, A. K. et al.

Nanowire-mediated delivery enables functional interrogation of primary immune cells: application to the analysis of chronic lymphocytic leukemia. Nano Lett. 12( 12):6498-504. doi: 10.1021 /n!3042917 (2012)), inhibiting interferon signaling fully blocked expression of antiviral genes. The ablation of the antiviral pathway was essentially complete, with no cells exhibiting any antiviral response, implying that even the "early responders" may require autocrine signaling of IfnB in order to activate their antiviral response. However, once again inflammatory "sharp peaked" genes displayed strikingly reduced levels of both digital and analogue variability in these knockout cells. Ifnarl-/- clustered closely with cells from the on-chip stimulation, and shifts in noise compared to LPS were significantly correlated between both experiments. Given the known role of interferon signaling in inducing the antiviral pathway, these finding cohesively point to extensive antiviral cross-inhibition as a primary upstream mechanism of inflammatory-response de-synchronization.

EXAMPLE 9. Removal of "Cluster-Disrupted" Cells

[000254] In the Examples above, it was identified that BMDCs fell into two distinct subpopulations, corresponding to distinct maturity states. BMDC maturation is a developmental process in which BMDCs switch from antigen-capturing to antigen- presenting cells in order to prime the adaptive immune system (see e.g., Jiang, A. et al. Disruption of E-Cadherin-Mediated Adhesion Induces a Functionally Distinct Pathway of Dendritic Cell Maturation. Immunity 27, 610-624, doi: 10.1016/j. immuni.2007.08.015 (2007)). Maturation can occur either in response to pathogen-derived ligands, such as LPS, or as a result of disrupting clusters of BMDCs in cu lture (Ibid.), both of which lead to up- regulation of specific cell-surface markers. Pathogen-dependent maturation occurs over a prolonged time after pathogen exposure and cells fall along a developmental continuum in the dataset (Fig. 24d,e).

[000255] However, pathogen-independent maturation, also referred to as 'cluster disruption', is a known artifact of the culturing process, occurs prior to stimulation, and represents a distinct cellular state. Thus, to measure changes in gene expression variation from a 'homogenous' population appropriately, the studies provided herein sought to remove all cluster-disrupted cells from all further analyses.

[000256] In the previous Examples that performed PCA on 18 cells, it was found that the first principal component (PC I ) discriminated these two cellular populations. Many genes with high PC I loadings were known markers of BMDC maturation (Jiang Immunity 2007), such as the cell-surface receptors Ccr7, Cd83, and Cd86. These genes are up- regulated in both the pathogen-dependent and pathogen-independent maturation pathways, and thus are all induced at a population-level in the LPS time course (Am it, 1. et al.

Unbiased reconstruction of a mammalian transcriptional network mediating pathogen responses. Science 326, 257-263, doi: 10.1 126/science. l 179050 (2009); Shalek, Nature 2013). Among the PC I genes, Lyzl had the strongest loading, and was the best discriminator of cluster-disrupted cells. It was (ln(TPM+l) > 9) in the 1 maturing (non- cluster disrupted) cells, but was completely absent (TPM=0) in the three cluster-disrupted cells. Furthermore, Lyzl was not differentially regulated in two cells undergoing pathogen- dependent maturation, and this did not appreciably change in its single-cell or population- averaged levels throughout the LPS time course. Similarly, a complementary marker (Serpinb6b) was identified, and this marker was found to be highly expressed only in cluster-disrupted cells, but absent from all others, yet did not appreciably change its overall expression during the LPS time course. Thus, these markers are unlikely to be differentially regulated in cells undergoing pathogen-dependent maturation, and it was reasoned that the expression patterns of these marker transcripts provided a method for identifying cluster disrupted cells. To independently confirm the two markers, further qRTPCR analysis was performed on cells pre-sorted for CD83 (maturation marker) expression before stimulation and then stimulated the two sorted sub-populations (CD83+, CD83-) with LPS for 4h. The level of the two mRNAs in the two subpopulations was measured both before and after stimulation. These studies successful ly validated that these markers cleanly distinguish between the two subpopulations over the pathogen response.

Primers Used:

[000257] Accordingly, to stringently remove all potentially cluster-disrupted cells, all libraries where ln(TPM+l ) < 6 for Lyzl or ln(TPM+l ) > 4 for Serpinb6b were excluded from further analyses. This was done for each experiment without exception.

[000258] To make sure that cluster disruption was not linked to early activation of the "core" antiviral module, it was confirmed that there was no correlation between the expression of cluster disruption markers and the activation of the "core" antiviral module for both the 1 h LPS stimulation and the 4h LPS "on-chip" stimulation experiments.

[000259] The invention having now been described by way of written description and example, those of skill in the art will recognize that the invention can be practiced in a variety of embodiments and that the description and examples above are for purposes of illustration and not limitation of the following claims.