Field of the Invention

[0001] The invention relates to particle sorting, sample processing, biological assays, biotechnology, and diagnostics.

Background of the Invention

[0002] Efficient separation, isolation and/or enrichment of particles including nano- and micro-sized spheres, emulsions, droplets, pathogens and cells at ultra-low concentration (1-1000 particles/mL) in (turbid) liquid samples, coupled to further processing and analyses such as sequencing, is a critical task in diverse research areas, ranging from clinical diagnostics and environmental monitoring to food safety and biotechnology. Whether applied to preventive routine controls across the food chain or to the prompt identification of blood-bome pathogens, next-generation techniques are expected to provide absolute quantification of low-abundant bio markers (for example down to 1-100 units/mL), while simultaneously satisfying a short (minute- to-hours) assay time requirement. They should guarantee single-cell sensitivity and specificity, high throughput, minimal sample pre-processing and compatibility with large-volume turbid or highly scattering media. Moreover, combining the detection of low-abundant pathogens with their isolation and sorting for further analysis would be of paramount importance in directing and optimizing immediate treatment. However, none of the currently employed techniques, including conventional colony-counting methods and flow cytometry, is capable of complying with all or the majority of these requirements at the same time [1,2]. For instance, FACS (Fluorescence Activated Cell Sorting) requires target concentrations higher than ~ 100,000 units/mL [3-5]. The same concentration limit applies to a variety of micro fluidic-based sorting devices [6], ranging from active systems exploiting acoustic, electric, optical or magnetic forces, to passive systems relying on inertial forces or immobilization procedures [3]. The typical adopted flow rate ( _' -pL/min) provides droplet-based micro fluidic sorting systems with low (-«2 kHz) throughput, and the small (<mL) total handled sample volume is not suitable for fast diagnostic assays on low-abundant pathogens, which require exploration of a 0.2-10 mL clinical sample volume just to identify sufficient targets for meaningful downstream analysis [6,7]. [0003] For example, sepsis resulting from Blood Stream Infections (BSIs) is a severe health problem, annually affecting over 18 million people worldwide and over 700,000 patients in the United States [8-10]. It is widely recognized that the prompt diagnosis and treatment of bacterial BSIs in the early stage of infection has a profound effect on survival rates [9-11]. However, significant challenges are associated with the necessity of identifying pathogens that typically occur at ultra- low concentration (1-100 units per mL of blood) in a total short (minutes to hours) assay time. The gold standard among currently employed techniques - bacterial culture coupled to susceptibility testing for drug resistance - is excessively time consuming due to the several days of incubation necessary for the visualization of colonies [1,2,12]. On the other hand, flow cytometry and amplification-based methods including PCR (Polymerase Chain Reaction) are capable of reducing the assay time, but do not always meet the sensitivity and specificity required to target low-abundant bacteria in a pool of heterogeneous non-target species (for example, red blood cells in unprocessed blood samples) [12,13]. Standard methods often require complex and lengthy sample pre-processing, are unsuitable for turbid media, suffer from low throughput or simply cannot handle a large (>mL) total sample volume. Furthermore, all these techniques fail in enabling the sorting and isolation of the detected pathogens for subsequent processing and analyses. The characterization of the bacteria physiological state or viability upon detection, as well as single-cell genetic analyses and the evaluation of basic properties such as metabolic activity, membrane integrity or reproductive ability, would be highly beneficial in complementing the detection of low-abundant pathogens and in directing and optimizing early therapy and treatment [14,15].

[0004] Particularly detrimental in the context of clinical diagnostics, the existing limitations in the detection and isolation of rare biomarkers in (turbid) liquid samples equally impact environmental monitoring and the food industry. A variety of low-abundant pathogenic agents (mainly bacteria, but also viruses and parasites) are associated with severe water-borne and food-borne diseases, which represent a threatening health problem worldwide. For instance, more than 320000 cases have been reported in 2015 in the European Union alone [16] by the annual summary report of the European Food Safety Authority (EFSA) and the European Centre for Disease Prevention and Control (ECDC), with consequences ranging from mild symptoms to life-threatening conditions. The incidence, outcome and cost of food-related illnesses again point out the need for novel methods and diagnostic platforms to enable efficient and reliable food screening in packaging and processing facilities, before pathogenic outbreaks may reach the delivery/distribution/commercialization chain and ultimately the consumer level [17-19].

[0005] Facing the existing limitations in the detection of low- abundant pathogens in blood and turbid fluids in general, we have recently developed a portable, low cost, high- throughput particle counter, capable of providing absolute quantification of fluorescent targets in liquids even at extremely low concentration (1-1000 particles/mL) and in very short (minute- long) sample scanning periods [20-23] (Gratton, et al. US Patent No. 7,973,294 (2011); US Patent No. 7,528,834 (2009)). The previously described particle counter achieves fast throughput by avoiding flow but instead moving the sample in front of a detection unit. In combination with droplet blood microencapsulation and DNAzyme sensor fluorescence technology (i.e., in our recently developed Integrated Comprehensive Droplet Digital Detection - IC3D- platform [13]), the device has already been successfully applied to the detection of rare E. Coli (1-10000 units/mL) in spiked blood samples with a single, culture-free, specific and fast (1 to 4 hours) reaction [13].

Brief Summary of the Invention

[0006] This invention provides a particle-sorting unit, component or device that can be complemented and integrated with a particle counter, to enable separation, isolation and/or enrichment of target particles. In alternative embodiments, sorted particles are subject to downstream analyses and processing including sequencing for genetic profiling and diagnosis.

In alternative embodiments, the said particles are nano- and micro-sized spheres, emulsions, droplets, pathogens and cells with various sizes and shapes. By the adaptation of the fluorescence staining, the automated, integrated, user-friendly particle detection and sorting unit described here can be applied to the isolation of a variety of rare food-, water- and blood-bome target cell populations, including, e.g., low-abundant (antibiotic resistant) bacteria, circulating tumor or fetal cells, and hematopoietic stem cells in fluid samples. We anticipate therefore critical enabling applications to infectious diseases, antibiotic resistance, cancer research and stem cell therapy, virus or prenatal diagnostics, biotechnology industry, agriculture and environmental monitoring.

[0007] More specifically, we provide a particle- sorter that can be complemented and integrated with a particle counter, to enable separation, isolation and/or enrichment of particles. The particle-sorter comprises a real-time particle identification procedure, a receiver for the detected particles, and a micro-injector or an electrically controlled pneumatic, solenoid, or piezo valve system that generates negative pressure to direct the detected particles into the receiver.

The receiver is a needle or a capillary for particles that are nano- or micro-sized. Examples of particles are micro-sized spheres, emulsions, droplets, pathogens or cells with various sizes and shapes, where the emulsions and droplets can comprise a cell, a plurality of cells. A switchable valve can be provided whereby extracted particles can be dispensed into one or multiple containers after extraction.

[0008] In an embodiment, the sorted particles are subject to downstream analyses and processing. In another embodiment, analyses and processing includes sequencing for genetic profiling and diagnosis. In another embodiment, the particle sorter is applied to the isolation of a variety of rare food-, water- and blood-bome target cell populations. In another embodiment, the target populations include low-abundant (antibiotic resistant) bacteria, viruses, circulating tumor cells, fetal cells, and/or hematopoietic stem cells, in fluid samples. In still another embodiment, the particle sorter can be applied to infectious diseases, antibiotic resistance, cancer research, stem cell therapy, virus diagnostics, prenatal diagnostics, biotechnology industry, agriculture and/or environmental applications.

Brief Description of the Drawings

[0009] For a more complete understanding of the present invention, reference is now made to the following descriptions taken in conjunction with the accompanying drawings, in which:

[0010] Figure 1A is a schematic of the particle-counter optical path;

[0011] Figure IB shows the scanned volume in the IC3D particle counter;

[0012] Figure 2A shows an exemplarary time trace;

[0013] Figure 2B exemplifies the pattern-recognition algorithm adopted for the particle- counter data analysis for analyzing the time-trace of Figre 2A;

[0014] Figure 3 is a schematic of the particle-sorter prototype;

[0015] Figure 4 shows fluorescence intensity time traces collected with the particle counter showing combatibility of the proposed sorting mechanism with droplet microfluidics; [0016] Figure 5 A shows results (number of counted particles) of the pattern-recognition- based analysis to exclude turbulence effects induced by the needle of the particle sorter;

[0017] Figure 5B shows results (optimal standard deviation for the Gaussian recognition filter) of the pattern-recognition-based analysis to exclude turbulence effects induced by the needle of the particle sorter;

[0018] Figure 6 shows frames from a time-lapse image sequence demonstrating successful operation of the particle sorter;

[0019] Figure 7 shows frames from a time-lapse image sequence acquired under the same experimental conditions of Figure 6 demonstrating control of the sorting mechanism (non target particles do not get sorted).

[0020] Figure 8A shows the prototype of the particle-sorter.

[0021] Figure 8B shows the cross-correlation of the signals between the detection and positioning volume.

[0022] Figure 8C shows an image of the extracted particle in the capillary used for extraction.

[0023] Figure 9A shows how particles can be collected during extraction.

[0024] Figure 9B shows how particles collected in Figure 9A can be dispensed into one or several containers.

Detailed Description of the Invention

[0025] We describe the necessary hardware modifications on the particle counter and the software requirements to perform the real-time pattern-recognition-based identification of target fluorescent particles and to trigger sorting events.

Particle counter

[0026] The particle counter developed at the LFD [20-22] (Laboratory for Fluorescence Dynamics, University of California-Irvine, Irvine, CA, USA) is an essential building block of the claimed particle sorter. We therefore start by summarizing its working principles.

[0027] Particle-counter optical path. The employed particle counter generally consists in a portable, low-cost, horizontal-geometry fluorescence confocal microscope (Figure 1A; Dl and D2, dichroic mirrors; Ll, objective lens; L2 and L3, lenses; F, emission filter; PMT, Photo- Multiplier Tube; and ADC, Analog-to-Digital converter). A visible-wavelength excitation laser beam (l=469 nm or 532 nm in our case, with ~5 mW typical output power) is focused by an under-filled low numerical aperture 20x objective (MV-20x, Newport Corp., CA, USA) inside a cylindrical glass cuvette (Abbott, IL, USA) containing the sample. Slow (1-15 mm/s) vertical translation and fast (10-1000 rpm) rotation of the cuvette (stepper motors by Vexta, Oriental Motors USA Corp., CA, USA and by Haydon Kerk, CT, USA or other appropriate vendors) allow transporting the target fluorescent particles across the Gaussian-shaped excitation volume (as the cuvette translates and rotates, target particles cross the excitation volume as if they were undergoing laminar flow with constant speed v _x ; Figure IB). The fluorescence signal emitted by the particles is collected by the same objective, transmitted through the dichroic filters and collimated onto the sensitive area of a Photomultiplier Tube (H9305-04 PMT; Hamamatsu,

Japan or from other appropriate vendors) for the acquisition of a fluorescence intensity time- trace and for subsequent data-analysis via a pattern-recognition algorithm.

[0028] A large confocal pinhole in the detection path (with adjustable size from ~2 mm down to ~20 pm), combined with the under-filled low NA excitation objective, results in a ~nanoliter size for the excitation volume (V _eXc , radial and axial beam waists co _x , co _y and co _z in Figure IB): together with the fast cuvette rotation, this is crucial to enable the high-throughput (~0.5 mL/min) required to rapidly detect pathogens in the typical relevant ultra-low

concentration range. For instance, the total observed helicoid volume V _0bs can be as high as 0.4 mL with a typical rotation frequency f=300 rpm and a scanning time t=l min (Figure IB). Measurements in turbid and highly scattering media are allowed by positioning the excitation volume close (<100 pm) to the cuvette wall.

[0029] Particle-counter data analysis: pattern recognition algorithm. The raw data of a particle-counting experiment consist of a temporal fluorescence intensity trace, with peaks of fluorescence intensity (produced by target particles when crossing the excitation volume) overlaid to an ideally zero, constant background (an exemplary time-trace is shown in Figure 2A). Such a time-trace is analyzed by the pattern-recognition algorithm we describe in Figure 2B. Upon background subtraction, at each time-point a fit to a predetermined profile (here assumed to be Gaussian, as assigned by the shape of the excitation volume) with variable amplitude A and fixed standard deviation s is performed: if the chi-square of the fit is below a pre-required value and the fit amplitude exceeds a minimum intensity threshold T a hit is counted . High-intensity spikes of different shape (e.g., those arising from cuvette impurities) do not comply with the fit chi-square requirements and are not counted as hits. Then the absolute target concentration is quantified by simply dividing the total number of counted particles by the scanned volume Vobs, which can be either precisely computed (Figure IB) or pre-calibrated by successive dilutions of a reference sample of known concentration. In the original

implementation, the pattern-recognition analysis is performed upon completion of the data collection. The analysis of a typical 2 minutes-long time trace (6 million data points for a 50 kHz fluorescence sampling frequency) can be performed in a few seconds (SimFCS software, Laboratory for Fluorescence Dynamics, CA, USA).

Particle sorter

[0030] Particle-sorting unit: structure and operation overview. The particle-sorting device invented here can be integrated with the particle counter described above and, in principle, with any particle detecting system. The setup simultaneously performs the acquisition and analysis of the experimental data. The pattern-recognition based analysis of the raw temporal fluorescence intensity trace collected by the particle counter is preserved but implemented for real-time performance: fluorescence intensity spikes are fit to the theoretically expected predetermined (Gaussian) profile during the acquisition of the intensity time-trace, so that a sorting event can be triggered immediately upon the identification of a positive particle. Sorting is performed by a detection-and-suction mechanism. In alternative embodiments, a system of (multiple) pneumatic valves, connected to a needle positioned inside the cylindrical cuvette containing the sample, is exploited to generate the vacuum pressure required to suck the detected particle inside a connection tube. We believe that such a configuration provides the most straightforward and simple possible realization of the sorting mechanism. First, it does not require any modification of the particle-counter optical path, which would be otherwise mandatory for an alternative optical tweezers-based approach. Second, insertion of sterile needle and tubing inside the cuvette does not induce sample contamination and allows avoiding addition of sorting outlets or separate micro-channels. This preserves the exploitation of a cylindrical rotating sample-holding cuvette that, as described previously (Figure 1), is crucial to scan a mL- sized volume in a few minutes. Either micromanipulation technology or simpler stage micrometers are adopted to allow the precise and repeatable positioning of the needle along the circular trajectory followed by the particles due to the fast cuvette rotational motion. Sorted droplets are maintained inside the tubing until completion of the sorting procedure and are finally transferred into a separate sealed cuvette for subsequent analysis.

[0031] Data acquisition and pattern-recognition-based analysis. While the particle- counter optical path (Figure 1A) is preserved and kept unchanged, the implementation of a particle-sorting unit requires the development of new specifically tailored software to enable the in-parallel acquisition and pattern-recognition-based analysis of the fluorescence intensity time- trace. In order to achieve this goal, the data- acquisition card (l6-bit, 1 MHz,

DaqBoard/3000USB, IoTech Inc., OH, USA or from other appropriate vendors) is operated in a specific data- streaming mode. Data acquisition proceeds indefinitely upon the start signal until the accumulation of the pre-specified number of requested data points; during the acquisition, data get transferred to the computer buffer and processed by the pattern-recognition algorithm (Figure 2B) in a slot-wise fashion. The highest speed is achieved with a data-slot size N- 4096 data points: it allows performing the data analysis every 102 ms (= A-N/f _sa mpi ) with the typical operating sampling frequency f _sa mpi- 40 kHz. Two important considerations arise from this approach. First, the exact coordinate of the detected particle within the 4096 data slot must be recorded and taken into account when defining the time-delay between detection and sorting (i.e., when triggering the generation of a sorting event). This task is easily accomplished by a position-sensitive array-based storage of the raw intensity values provided by the data- acquisition board. Second, when for convenience sake a constant time-delay between the detection of a particle and its effective suction is adopted for all the particles, D=102 ms assigns the minimum value for such a delay. The chosen temporal delay between detection and sorting is set as experimental parameter in the data-acquisition software, together with the fit chi-square and amplitude thresholds for the patter-recognition data analysis.

[0032] Sorting mechanism. Provided the real-time analysis of the raw temporal fluorescence intensity trace, a sorting event must be initiated after the identification of each target particle. Specifically, our particle sorter takes advantage of a needle positioned downstream of the excitation volume and connected via high-pressure-resistant tubing to a system of (multiple) programmable pneumatic valves. Negative vacuum pressure is generated inside the needle by triggering the valves opening, thereby inducing suction of each detected target into the tubing. Subsequent re-balancing of the pressure between successive sorting events keeps the sucked particles and prevents them from re-entering the solution. Only at the end of the sorting procedure, the sorted targets are ejected and transferred into a separate cuvette for subsequent analysis.

[0033] Needle position. Accurate positioning of the needle, with respect to the particle- counter excitation volume and along the particles circular trajectory inside the cuvette, is crucial for the success of the sorting procedure. Given a desired time delay t between detection and sorting, the needle must be positioned at the expected particle coordinate at the very same time t, as assigned by the user-defined rotational frequency /and the vertical translation speed of the cuvette motion. In our prototype we have opted for a time delay i=l/(4/) between detection and sorting events, so that each particle is sorted after completing a quarter of a revolution. This configuration offers an important advantage: in case some perturbation in the flow laminarity is induced behind the needle (though unlikely, see Examples), the particle should not undergo any deviation from its expected circular trajectory before reaching the needle and being sorted. To achieve precise positioning of the needle, a second fluorescence excitation and detection channel is added at 90 degrees by duplication of the particle-counter optical path. In this way, the system resembles a dual-spot cross-correlation setup [24]. At first, effective positioning of the second added excitation volume along the particles circular trajectory is achieved and verified by adjusting its position with a three-axis stage micrometer (~l0 pm accuracy), and by

simultaneously monitoring the cross-correlation of the fluorescence time-traces detected in the two channels. Then, once the cross-correlation signal has been maximized, the needle is positioned inside such a second excitation volume: to this aim, the needle tip is fluorescently labeled with a non- water-soluble dye (Rhodamine 6G in our case), and the detected signal is maximized by real-time monitoring of the fluorescence counts. At this point, the second excitation laser beam can be switched off and the entire sorting procedure can be carried out at the right position with the simple one-channel configuration described in Figure 1.

[0034] Particle-sorter prototype. Following this design, an instrument prototype has been constructed and demonstrated (a schematic of the setup is shown in Figure 3; note that the needle and the top and side view of the cuvette are shown not to scale). For simplicity and reasons of practical convenience, we have exploited a pneumatic programmable micro-injector operated in suction mode: connected to a suitable compressed-gas source (a 60-psi compressed- air wall outlet, in our case) and to a thin (30-gauge) needle inside the cuvette containing the sample, the micro-injector generates the negative pressure required to suck the target particle inside the needle and the connection tubing. The electrically controlled micro-injector we have employed (IM-300, Narishige, Japan) allows the pneumatic -valve activation time (i.e., the time span of the suction event) to be tuned in the broad range 10-300 ms. The suction pressure can be conveniently regulated from 1 to 60 psi. The same applies to the injection pressure, which is employed to transfer all the sorted particles from the connection tubing to a separate sealed cuvette at the end of the sorting procedure. The pressure used to counter-balance the capillary effect can be set down to <0.1 psi according to the diameter of the employed needle.

[0035] Two main advantages are associated with, and have motivated the use of, the micro-injector. First, a tiny (0.1 pL) volume can be isolated in a single sorting event with a 30- gauge needle by simply regulating a 60-psi suction pressure and a 20-ms valve activation time.

If we assume an initial target concentration of 100 units/mL, even with a 50% sorting efficiency we can achieve a (biologically and clinically relevant) 50-fold concentration increase. Overall, a 0.1 -pL suction volume defines a maximum achievable concentration of 10 ⁴ particles/mL for the sorted sample. Secondly, the micro-injector can be activated and driven by external TTL signals. Therefore, a single digital-output channel of the same data-acquisition board

(DaqBoard/3000USB, IOTech Inc., OH, USA) responsible for the collection of the temporal fluorescence intensity trace has been programmed to generate step-like trigger signals with the following features: (i) 5-V amplitude, (ii) a start point assigned by the exact time-point of the particle transit through the counter excitation volume (i.e., the exact particle coordinate in the previously mentioned 4096 data-points slot) and by the desired delay between a detection and suction event. A time span equal to the desired pneumatic-valve activation time is usually selected, even though such activation time is separately pre-set on the micro-injector. Again, an array-based approach has been exploited to transfer the desired signal to the FIFO buffer of the IOTech board.

[0036] The sorter prototype based on the pneumatic micro-injector has been employed to run the first proof-of-principle sorting experiments on fluorescent micro-heads at ultra- low concentration in water solutions (see section Examples). However, a drawback of the setup prototype consists in a long (~500 ms) and non-constant time delay between the software generation of a trigger signal and the actual activation of the desired pressure difference inside the needle. This can be ascribed to both a non-constant response time of the micro-injector, and to the long (~lm) length of the air-filled tube connecting the micro-injector valve to the needle. Aiming at an overall improvement of the sorter performance and operation repeatability, in alternative embodiments, the micro-injector-based system is replaced with electrically controlled pneumatic valves, to be positioned closer to the needle with a water- filled connecting channel. Specifications of the valves are again being selected based on the minimum volume that can be displaced in a single sorting event and the minimum valve activation and response time. [0037] In alternative embodiments, the target particles include, but are not limited to, nano- and micro-sized spheres or beads, emulsions, droplets, microorganisms, pathogens (such as bacteria, viruses, fungi), cells (such as mammalian cells, human cells, animal cells, cancer cells, immune cells, stem cells, progenitor cells, differentiated cells, prokaryote cells, yeast cells, fungal cells, bacteria, eukaryotic cells, plant cells, breeding cells, engineered cells, fused cells, hybrid cells), embryos, multicellular organoids, protoplasts, lipid vesicles (such as liposomes), extracellular vesicles (such as exosomes, micro vesicles, and apoptotic bodies).

[0038] In alternative embodiments, the claimed particle sorter can sort particles with various sizes ranging from ~ l0 nanometers (nm) to ~200 micrometers (pm). In alternative embodiments, the said particles can have different shapes or morphologies such as spherical and rod-like.

[0039] In alternative embodiments, the claimed particle sorter operates with a typical initial target concentration ranging from ~ l to 1000 particles/mL and as low as 1-10

particles/mL.

[0040] In alternative embodiments, the claimed particle sorter is amenable to be integrated, coupled or used in conjunction of down-stream, established in the art, techniques for target or biomarker processing and analysis including, for example, PCR, immunostaining, sequencing and single-cell sequencing.

[0041] It is understood that the performance such as efficiency and operation

repeatability of the particle sorter can be optimized and improved by modulating the time span of the suction event, the applied sorting pressure, the transit speed of target particles across the counter excitation volume, the time delay between detection and suction events and the initial target concentration.

[0042] Our particle sorter possesses the following performance properties, including (i) fast sorting rate and high ( -0.5 mL/min) throughput, (ii) suitability for native or complex biological fluids, (iii) capability to handle large sample volumes (100 pLs to several mLs) and highly diluted targets, (iv) simple operational procedure in the framework of a fully integrated and automated system, (v) low cost and portability.

[0043] In some embodiments, the droplets comprise a live cell. In some embodiments, the particle- sorting device is used for sorting of droplets that contain live cells, wherein the fluorescent particle in the said droplet is a live cell, wherein the fluorescent signal of the said cell is or is derived from a group consisting of a fluorescent protein expressed by the said cell, a fluorescent-dye-conjugated affinity agent that labels the said cell’s surface, a fluorescent cell tracking dye that stains the said cell, and a fluorogenic substrate provided to the said cell which directly or indirectly coverts the said fluorogenic substrate into a fluorescent signal.

[0044] In some embodiments, a plurality of droplets co-encapsulated with one, two or more types of live cells are provided for sorting by the particle-sorting device, wherein at least one type of the said cells is fluorescent. In some embodiments, the said one or more types of live cells may produce fluorescent signal of a same, a similar or a distinct excitation/emission spectrum. In some embodiments, a single live cell may comprise two or more types of fluorophores characterized by similar but different excitation/emission spectrums. As used herein, the said cell refers to a mammalian cell, an animal cell, a yeast cell, a fungal cell, a bacterium, or a derived or an engineered form of any of the above-said cell. As used herein, the fluorescent protein refers to a violet, blue, green, orange, yellow, red, or far-red fluorescent protein that is represented by green fluorescent protein (GFP), EGFP, ZsGreen, mCherry, yellow fluorescent protein (YFP), and red fluorescent protein (RFP). As used herein, the affinity reagent refers to an antibody, a target-binding protein, a target-binding peptide, a ligand, an aptamer, a RNA fragment, a DNA fragment, a derivative or an engineered or a combination form of any the-above listed reagents.

[0045] In some embodiments, the particle-sorting device is used for sorting of droplets that contain live cells, wherein the fluorescent particle in the said droplet is a fluorescent bead that is co-encapsulated with the said live cell, wherein the fluorescent bead serves as a qualitative or quantitative assay readout of a biological molecule that is secreted by the said cell. In some embodiments, the fluorescent bead is a nanobead or microbead or particle or polymer with a size ranging from about 30 nm to about 30 pm, from about 100 nm to about 15 pm, or from about 200 nm to about 10 pm. In some embodiments, the fluorescent bead is a particle coated with an affinity reagent to capture biological molecule produced and secreted by the said co-encapsulated cell. In some embodiments, the said biological molecule anchored on the beads further capture a fluorophore-containing antibody or probe. As used herein, the biological molecule that is secreted by the said cell is a protein, an antibody, a ligand, a receptor, a cytokine, a chemokine, a metabolite, a sugar, a lipid, a complex lipid, or a combination of any of the above-listed biological molecules. [0046] In some embodiment, the particle-sorting device is used to separate, isolate, or enrich cells with one or more characteristics. In some embodiments, the cells to be sorted are B cells, T cells, hybridoma cells, cancer cells, stem cells, engineered cells, fused cells, yeast cells or bacterium cells. As non-limiting example, a plurality of B cells are encapsulated in droplets which are then sorted by the particle-sorting device on the basis of antigen- specific binding or functional activity of the antibodies secreted by the B cells. As another non-limiting example, a plurality of T cells are co-encapsulated with a target-antigen presenting cell in droplets which are then sorted by the particle- sorting device on the basis of the cognate interaction between a T cell and the target-antigen presenting cell, wherein the positive interaction leads to the production of a fluorescent protein that serves as an assay readout for an activated T cell receptor in a said droplet.

[0047] In some embodiments, the sorted droplets from the particle-sorting device are dispensed individually into a collection device for downstream processes such as single cell cloning, cell culturing and growth, single-cell reverse transcription, and polymerase chain reaction (PCR). In some embodiments, the said collection device is a 96- well, 384- well plate, a multi-well plate, a multi-well device, or a multi-tube array. In some embodiments, the sorted droplets from the particle-sorting device are collected as a pool.

EXAMPLES

[0048] The invention will be further described with the following examples; however, the invention is not limited to such examples.

Example 1. Sorter calibration

[0049] We have performed a set of calibration experiments to verify the compatibility of the pressure-driven sorting mechanism with fluorescent droplets (water- in-oil emulsions)

(Figure 4). Configured with a suction and injection pressure of 60 psi, the micro-injector of the sorter prototype has been employed to suck 0.25 pL of a highly concentrated (2· 10 ⁷ positive droplets per mL) solution of fluorescent (10 pM FITC) 50-pL droplets and inject them into a pool of negative (non-fluorescent) PBS droplets (droplets were fabricated according to the standard procedure reported in [13]). Comparison of the time-traces acquired with the particle counter before (red line) and after (black line) the injection of the positive droplets into the pool of non-fluorescent ones reveals that FITC droplets are clearly visible on an unchanged background and are not damaged by the needle and the adopted 60-psi suction pressure. We note that only an 8-second portion of the entire acquired raw time-traces is shown in Figure 4 for the sake of clarity in the visualization of fluorescence peaks.

[0050] Having verified that pressure-driven sorting is potentially compatible with droplet emulsions, we then investigated the effects produced by the needle inside the cylindrical cuvette containing the sample. Due to the fast cuvette rotational motion, fluorescent target particles are directed across the excitation volume of the particle counter exactly as if they were undergoing laminar flow with constant speed. Therefore, any needle inserted along the particles trajectory might act as an obstacle and induce turbulence. We do not expect the laminarity condition to be violated provided our typical experimental Reynolds number (R=l050 for an aqueous solution and a relatively high 200-rpm rotational frequency) and the thin (300 pm) outer diameter of the custom-made 2”-long 30G needle (Hamilton, NV, USA or other appropriate needle sizes, types and vendors) we typically employ. However, we obtained further experimental confirmation by performing particle-counting experiments on water solutions of l-um [wzi][m2] yellow-green fluorescent micro-beads (Invitrogen, CA, USA), which have been selected as standard reference sample due to ease in preparation, low cost and immediate availability. Successive fluorescence intensity time-traces have been collected, under the same conditions (cuvette rotation frequency f=400 rpm, cuvette vertical translation speed 5 mm/s, signal sampling frequency 70 kHz) and on the very same sample (10000 beads/mL), in the presence and absence of the needle; to fully explore its effect on the particles trajectories, two needle external diameters (200 and 400 pm) as well as various positions of the needle before and after the counter excitation volume have been tested. For all the configurations, the results of the pattem-recognition-based analysis of the time-traces have been compared in terms of the total number of counted targets (Figure 5A) and in terms on the optimal standard deviation for the Gaussian recognition filter (Figure 5B; the optimal standard deviation for the pattern-recognition analysis has been determined as the one maximizing the number of counted particles for fixed amplitude threshold T and fixed maximum fit chi-square value. Error bars in Figure 5 represent the standard deviation over N=6 1 -minute- long repeated measurements). As reported with the exemplary results of Figure 5A-B, no effect of the needle has been observed even up to a needle diameter of 400 pm. We conclude that the presence of the needle neither biases the number of counted particles, nor it modifies the speed of the particles crossing the excitation volume (changes in the molecules speed would otherwise affect the width of the fluorescence Gaussian peaks). Turbulent flow can therefore be excluded, further confirming the feasibility of the proposed approach. Example 2. Sorting experiments of fluorescent microspheres

[0051] Particle counting and sorting experiments have been carried out. In an example, 15- pm[wz3][ _m 4] yellow-green fluorescent beads (Invitrogen, CA, USA) diluted in milliQ water have been employed, and an ultra-low target concentration (200 units/mL) has been adopted to make the probability of sorting a particle just by chance negligible (for a 0.1 pL suction volume, based on Poisson statistics such a probability amounts to 2%). Experiments have been performed on the previously described prototype of the particle-sorting device. At first, the optical path of the particle counter has been duplicated at 90°: two laser sources with equal specifications and two identical Photomultiplier Tubes (PMTs) have been accommodated to provide fluorescence excitation and detection in the particle-counter channel (hereafter referred to as channel #1) and in the newly added channel #2. The two excitation volumes have been positioned at the same height inside the sample cuvette, 90° apart along the particles circular trajectory. Such a configuration has been achieved by keeping the first excitation volume fixed, by mounting the entire optical path of channel #2 on a three-axis stage micrometer (lO-pm accuracy along each axis), and by continuously adjusting the position of the second excitation volume while maximizing the cross-correlation function of the intensity time-traces collected in the two channels. Maximization of the cross-correlation signal ensures that all the particles cross both excitation volumes. While channel #1 has been subsequently employed for particle detection and identification, channel #2 has been used as a spatial reference to properly position the needle. The needle tip has been fluorescently labeled with pM[wz3][ _m 4] Rhodamine 6G and has been placed inside the excitation volume of channel #2 by simply maximizing the detected fluorescence counts; again, a three-axis stage micrometer (lO-pm accuracy along each axis) has been employed for the needle displacement up to the final position. Particle sorting has then been performed by triggering the activation of the pressure difference inside the needle through the micro-injector pneumatic valve (60-psi suction pressure, 60-ms activation time). The time delay between consecutive detection and sorting events has been defined on the data-acquisition software according to (i) the time it takes for each particle to travel from the counter excitation volume to the needle (l/(4f), with a cuvette rotation frequency f=25 rpm), and (ii) the previously mentioned (~500 ms) delay required for the triggered pressure difference to actually reach the needle. [0052] The second excitation laser beam could in principle be switched off once the needle has been placed (i.e., it is not strictly required for the particle sorter to operate). However, we have decided to further take advantage of the second fluorescence detection channel to provide direct inspection of the success of sorting events. Once accomplished the alignment of the needle, we have replaced PMT #2 with an EMCCD (Electron Multiplying Charged Coupled Device) camera (Photometries Cascade 128+, Photometries, USA), and regulated the system magnification to achieve a 450-pm imaged field of view centered on the needle tip. Continuous (630-fps) time-lapse imaging has been carried out during the sorting procedure, to allow visualization of fluorescent particles getting closer to the needle tip and being sucked into the needle itself.

[0053] Correct operation of the particle-sorting device and the possibility of employing the proposed pneumatic-driven detection-and-suction mechanism are demonstrated in Figure 6. An exemplifying successful triggered sorting event is shown with selected frames (frames 87 in Figure 6A, 157 in B, 227 in C, 297 in D, 367 in E and 437 in F) from a time-lapse movie acquired at 630 fps on the Cascade 128+ EMCCD camera (465 pni[wz3][ _m 4] image side, intensity scale from 1832 to 4361 a.u.). The needle tip stained with Rhodamine 6G appears as a saturated fluorescent spot at the center of the field of view: a particle enters the imaging area and, as soon as the pressure difference is activated, it disappears when getting sucked into the needle. Direct visualization of the trigger signal sent to the micro-injector has been achieved by positioning a green LED close to the camera sensitive chip, and by turning the LED on and off via the data- acquisition software through the very same 5V signal used to activate the micro-injector (i.e., when the LED turns on, the micro-injector receives the electric signal to activate the valve). The LED on-time (frames 85-123) can be visualized in Figure 6G, which shows the average intensity per frame versus time. The time difference between the activation of the LED and the time-point where the particle is sorted by the needle matches with the separately measured time delay required for the pressure difference to actually appear inside the needle. This delay has been taken into account when defining the trigger- signal time on the sorter software. The LED on-time is pre-set on the software too but has no practical consequence on the pneumatic valve activation time (which is directly programmed on the micro-injector). A cuvette rotation frequency f=25 rpm and a PMT sampling frequency f _Sampi =40 kHz have been employed as acquisition parameters. [0054] In Figure 7, as a control, we show frames from a time-lapse image sequence (465 pm image side, 630 fps, intensity scale from 1811 to 4371 a.u.; frames 22, 47 and 249 in Figure 7A, B and C respectively) acquired under the same experimental conditions of Figure 6 showing a l5-pm fluorescent bead that, in the absence of a trigger signal to the micro-injector, is not sorted and is therefore detected on both sides of the needle. If the micro-injector is not activated (i.e., if a particle is counted but no trigger signal is sent to the pneumatic valve), no suction occurs and the particle crosses the entire camera field of view with unchanged trajectory.

[0055] While particle extraction with a pneumatic-driven micro injector works, as demonstrated in the previous sections, the long air-filled tube connecting to the device can cause a time delay in mediating the suction. Therefore, we implemented an alternative approach using optionally a solenoid valve (6724 2 way Whisper Valve, Buerkert, Germany) instead of the micro injector. This minimizes tubing length and air gaps can be avoided by filling the entire tubing with fluid. In Figure 8, we show the prototype device and the proof of concept for particle extraction with a solenoid valve. Figure 8A shows a photograph of the particle sorting and extraction device. A constant vacuum was applied via fluid-filled tubing, and a solenoid valve was used to trigger the suction event. Particle detection and triggering was performed as described in the previous sections. For particle extraction, a glass capillary was placed into the cuvette and the position of the tip was precisely adjusted with a XYZ micrometer stage. To verify that the tip position is correct, the fluorescence excitation/detection arm was replicated with the laser focus placed at the suction point located at approximately 90 degree angle with respect to the first arm (in principle, placement can be at any angle). Figure 8B shows the cross correlation of the fluorescence signals detected from a solution containing fluorescent particles (15 pm diameter yellow-green fluorescent microspheres, Invitrogen, USA) in both arms. The cross-correlation amplitude peak indicates that particles detected in the first laser focus appear in the second laser focus at 60 ms delay. The particle detection and subsequent extraction was verified with the same 15 pm diameter yellow-green fluorescent microspheres spiked into 3 ml of water at a concentration of -1,000 particles/ml, a fluorescence image of the capillary tip taken after extraction shows the extracted microsphere in Figure 8C. Here, a transparent tip material such as glass is advantageous because the presence of particles can be optically verified with ease. This verification can be done post-extraction, as shown here, or during extraction, by real time fluorescence detection in the capillary with the second detection arm. [0056] Finally, by using a valve with multiple ports, i.e., a 3-way valve, or an additional T-connector switch, particles can be easily dispensed after extraction by reversing the flow. The concept of particle dispensing is shown in Figure 9. During particle extraction, the dispensing port is closed (Figure 9A). A vacuum is applied to suck the particles into a reservoir, which in the simplest case can be the tubing itself. After particle extraction, the port connecting to the extraction needle/capillary is closed, the port to the dispenser opened, and positive pressure is applied to eject the particles into one or multiple containers such as a multi-well plate (Figure 9B). References

[1] Lee, A., Mirrett, S., Reller, L. B., Weinstein, M. P. Detection of bloodstream infections in adults: how many blood cultures are needed? J. Clin. Microbiol. 45:3546-3548 (2007).

[2] Klouche, M., Schroder, U. Rapid methods for diagnosis of bloodstream infection. Clin. Chem. Lab. Med. 46(7): 888-908 (2008).

[3] Wyatt Shields IV, C., Reyes, C. D., Lopez, G. P. Microfluidic cell sorting: a review of the advances in the separation of cells from debulking to rare cell isolation. Lab Chip.

15(5):1230-1249 (2015).

[4] Herzenberger, L. A., Parks, D., Sahaf, B., Perez, O., Roederer, M., Herzenberger, L.

A. The history and future of the Fluorescence Activated Cell Sorter and flow cytometry: a view from Stanford. 48(10): 1819-1827 (2002).

[5] Chen, P. et al. Hydrodynamic gating valve for microfluidic fluorescence-activated cell sorting. Anal. Chim. Acta 663: 1-6 (2010).

[6] Warkiani , M. E., Wu, L., Tay, A. K. P., Han, J. Large-volume microfluidic cell sorting for biomedical applications. Annu. Rev. Biomed. Eng. 17:1-34 (2015).

[7] Mazutis, L. et al. Single-cell analysis and sorting using droplet-based microfluidics. Nat. Prot. 8(5): 870-891 (2013).

[8] Centers for Disease Control and Prevention. Antibiotic Resistance Threats in the United States (2013).

[9] Angus, D. C., van der Poll, T. Severe sepsis and septic shock. New Engl. J. Med. 369: 840-851 (2013).

[10] Angus, D. C. et al. Epidemiology of severe sepsis in the United States: analysis of incidence, outcome, and associated costs of care. Crit. Care Med. 29:1303-1310 (2001). [11] Boedicker, J. Q., Li, L., Kline, T. R., Ismagilov, R. F. Detecting bacteria and determining their susceptibility to antibiotics by stochastic confinement in nanoliter droplets using plug-based microfluidics. Lab Chip 8:1265-1272 (2008).

[12] Liesenfeld, O., Lehman, L., Hunfeld, K. P., Kost, G. Molecular diagnosis of sepsis: new aspects and recent developments. Eur. J. Microbiol. Immunol. 4(1): 1-25 (2014).

[13] Kang, D. et al. Rapid detection of single bacteria in unprocessed blood using Integrated Comprehensive Droplet Digital Detection. Nat. Commun. 5:5427 (2014).

[14] Nebe-von-Caron, G. et al. Analysis of bacterial function by multi-colour fluorescence flow cytometry and single-cell sorting. J. Microbiol. Methods 42: 97-114 (2000).

[15] Davey, H. M„ Kell, D. B. Flow cytometry and cell sorting of heterogeneous microbial populations: the importance of single-cell analysis. Microbiol. Rev. 60(4): 641-696 (1996).

[16] EFSA (European Food Safety Authority) and ECDC (European Centre for Disease Prevention and Control), The European Union summary report on trends and sources of zoonoses, zoonotic agents and food-borne outbreaks in 2015. EFSA Journal, 14(12):4634, 231 pp. (2016).

[17] Gracias, K. S., McKillip, J. L. A review of conventional detection and enumeration methods for pathogenic bacteria in food. Can. J. Microbiol. 50:883-890 (2004).

[18] Swaminathan, B., Feng, P. Rapid detection of food-borne pathogenic bacteria. Annu. Rev. Microbiol. 48:401-426 (1994).

[19] Yoon, J., Kim, B. Lab-on-a-chip pathogens sensors for food safety. Sensors Eng. 12:10713-10741 (2012).

[20] Gratton, E. M. G., Tahari A. Methods and devices for characterizing particles in clear and turbid media. United States Patent number 7528384 (2009).

[21] Skinner, J. P. et al. Simplified confocal microscope for counting particles at low concentrations. Rev. Sci. Instrum. 84:074301 (2013).

[22] Altamore, L, Lanzano, L., Gratton, E. Dual channel detection of ultra low concentration of bacteria in real time by scanning FCS. Meas. Sci. Technol. 24:65702 (2013).

[23] Zhang, K. et al. Digital quantification ofmiRNA directly in plasma using integrated comprehensive droplet digital detection. Lab Chip 15(21): 4217-26 (2015). [24] Brinkmeier, M., Dorre, K., Stephan, J., Eigen, M. Two-beam cross-correlation: a method to characterize transport phenomena in micrometer-sized structures. Anal. Chem. 71:609-616 (1999).

[0057] Although the present invention has been described in connection with the preferred embodiments, it is to be understood that modifications and variations may be utilized without departing from the principles and scope of the invention, as those skilled in the art will readily understand. Accordingly, such modifications may be practiced within the scope of the following claims.