The present application claims the benefit of priority of European patent application EP 19216591.8 filed 16.12.2019 which is incorporated herein by reference.

The present invention relates to the use of a pharmaceutical agent capable of promoting angiogenesis to achieve vascular normalization in prevention or treatment of metastasis, particularly an agonistic ligand specifically binding to the EphB4 receptor or EphrinB2 or a polypeptide comprising the Ephrin B2 polypeptide sequence. The invention also relates to a method for predicting the likelihood that a tumor forms metastasis.

Background of the Invention

Circulating tumor cells (CTCs) are considered to be metastatic precursors in several cancer types, including breast cancer, but the mechanisms that lead to their intravasation are largely uncharacterized. CTCs are shed as single cells, multicellular clusters (CTC clusters), and clusters with immune cells (CTC-WBC clusters). While cluster formation generally leads to an increased metastatic ability, whether CTC clusters are shed from a cancerous lesion in a passive (i.e. stochastic) or active (i.e. as a consequence to defined event) manner is poorly understood. Several factors have been linked to the ability of cancer cells to metastasize, such as cell-autonomous upregulation of metastasis-promoting genes or genes involved in the formation of a pre-metastatic niche, interaction with the immune system and microenvironmental signals. Particularly concerning the microenvironment, intratumor hypoxia and deregulated angiogenesis have emerged as key factors in promoting metastatic spread, yet, antiangiogenic approaches in the clinic have failed to consistently inhibit breast cancer progression.

Based on the above-mentioned state of the art, the objective of the present invention is to provide means and methods to prevent CTC cluster formation and thereby treat or prevent metastasis. This objective is attained by the subject-matter of the independent claims of the present specification.

Summary of the Invention

A first aspect of the invention relates to an agent for prevention or treatment of metastasis, wherein said agent is selected from a pharmaceutical drug a polypeptide and wherein said agent promotes angiogenesis to achieve vascular normalization inside the tumor.

A second aspect of the invention relates to a nucleic acid molecule encoding the agent according to the first aspect for use in treatment or prevention of metastasis.

A third aspect of the invention relates to a method for assigning a likelihood of developing metastasis to a patient. A fourth aspect of the invention relates to the agent for use in treatment or prevention of metastasis in a patient according to the first aspect, or the nucleic acid molecule according to the second aspect, wherein a high likelihood of developing metastasis is assigned to said patient according to the method of the third aspect.

In another embodiment, the present invention relates a pharmaceutical composition comprising at least one of the agents of the present invention or a pharmaceutically acceptable salt thereof and at least one pharmaceutically acceptable carrier, diluent or excipient, for use in treatment or prevention of metastatic cancer.

Brief Description of the Figures

Fig. 1 Dynamic labeling of hypoxic CTCs and assessment of their metastatic potential.

Fig. 2 Hypoxic CTC clusters express a gene signature that is associated with a poor prognosis of Stage I breast cancer patients.

Fig. 3 Knockdown of VEGFA increases CTC cluster shedding and metastasis formation.

Fig. 4 Pro-angiogenic therapy reduces intra-tumor hypoxia and suppresses the formation of CTC clusters and metastasis.

Fig. 5 In vitro characterization of the HIF1a activity reporter.

Fig. 6 In vivo analysis of vascularized hypoxic regions of breast tumors.

Fig. 7 Analysis of eYFP expression in single CTCs and CTC clusters.

Fig. 8 Metastatic potential of hypoxic CTC clusters.

Fig. 9 Knockdown of HIF1a does not affect the metastatic potential of LM2 and BR16 cells.

Fig. 10 HypoxiaRed correlates with HIF1a activity reporter and reacts with live hypoxic CTC clusters.

Fig. 11 Single-cell resolution RNA-sequencing analysis of hypoxic and normoxic CTC clusters.

Fig. 12 Knockdown of VEGFA increases CTC cluster shedding and metastasis formation.

Fig. 13 Pro-angiogenic therapy reduces CTC clusters shedding and metastasis formation in fastgrowing tumor models.

Fig. 14 Pro-angiogenic therapy reduces CTC clusters shedding and metastasis formation in the slow- growing tumor model BR16.

Detailed Description of the Invention

Terms and definitions

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art (e.g., in cell culture, molecular genetics, nucleic acid chemistry, hybridization techniques and biochemistry). Standard techniques are used for molecular, genetic and biochemical methods (see generally, Sambrook et al., Molecular Cloning: A Laboratory Manual, 2d ed. (1989) Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y. and Ausubel et al., Short Protocols in Molecular Biology (1999) 4th Ed, John Wiley & Sons, Inc.) and chemical methods.

The term “prevention or treatment of metastasis" in the context of the present specification relates to the use of agents and compositions specified herein in the treatment of cancer, where the indication or intention is predominantly to inhibit, slow or decrease the likelihood of a tumor to metastasize, in other words to spread from one site (i.e. from a primary tumour, the initial site, or from metastatic sites that are already established at the time of treatment initiation) to a different or secondary site in the patient’s body.

The term polypeptide in the context of the present specification relates to a molecule consisting of 50 or more amino acids that form a linear chain wherein the amino acids are connected by peptide bonds. The amino acid sequence of a polypeptide may represent the amino acid sequence of a whole (as found physiologically) protein or fragments thereof. The term "polypeptides" and "protein" are used interchangeably herein and include proteins and fragments thereof. Polypeptides are disclosed herein as amino acid residue sequences.

The term peptide in the context of the present specification relates to a molecule consisting of up to 50 amino acids, in particular 8 to 30 amino acids, more particularly 8 to 15amino acids, that form a linear chain wherein the amino acids are connected by peptide bonds.

Amino acid residue sequences are given from amino to carboxyl terminus. Capital letters for sequence positions referto L-amino acids in the one-letter code (Stryer, Biochemistry, 3 ^rd ed. p. 21). Lower case letters for amino acid sequence positions refer to the corresponding D- or (2R)-amino acids. Sequences are written left to right in the direction from the amino to the carboxy terminus. In accordance with standard nomenclature, amino acid residue sequences are denominated by either a three letter or a single letter code as indicated as follows: Alanine (Ala, A), Arginine (Arg, R), Asparagine (Asn, N), Aspartic Acid (Asp, D), Cysteine (Cys, C), Glutamine (Gin, Q), Glutamic Acid (Glu, E), Glycine (Gly, G), Histidine (His, H), Isoleucine (lie, I), Leucine (Leu, L), Lysine (Lys, K), Methionine (Met, M), Phenylalanine (Phe, F), Proline (Pro, P), Serine (Ser, S), Threonine (Thr, T), Tryptophan (Trp, W), Tyrosine (Tyr, Y), and Valine (Val, V).

The term gene refers to a polynucleotide containing at least one open reading frame (ORF) that is capable of encoding a particular polypeptide or protein after being transcribed and translated. A polynucleotide sequence can be used to identify larger fragments or full-length coding sequences of the gene with which they are associated. Methods of isolating larger fragment sequences are known to those of skill in the art.

The terms gene expression or alternatively gene product referto the processes - and products thereof - of nucleic acids (RNA) or amino acids (e.g., peptide or polypeptide) being generated when a gene is transcribed and translated.

As used herein, expression refers to the process by which DNA is transcribed into mRNA and/or the process by which the transcribed mRNA is subsequently translated into peptides, polypeptides or proteins. If the polynucleotide is derived from genomic DNA, expression may include splicing of the mRNA in a eukaryotic cell. Expression may be assayed both on the level

The term Nucleotides in the context of the present specification relates to nucleic acid or nucleic acid analogue building blocks, oligomers of which are capable of forming selective hybrids with RNA or DNA oligomers on the basis of base pairing. The term nucleotides in this context includes the classic ribonucleotide building blocks adenosine, guanosine, uridine (and ribosylthymine), cytidine, the classic deoxyribonucleotides deoxyadenosine, deoxyguanosine, thymidine, deoxyuridine and deoxycytidine. It further includes analogues of nucleic acids such as phosphotioates, 2’O-methylphosphothioates, peptide nucleic acids (PNA; N-(2-aminoethyl)-glycine units linked by peptide linkage, with the nucleobase attached to the alpha-carbon of the glycine) or locked nucleic acids (LNA; 2Ό, 4’C methylene bridged RNA building blocks). Wherever reference is made herein to a hybridizing sequence, such hybridizing sequence may be composed of any of the above nucleotides, or mixtures thereof.

The term siRNA (small/short interfering RNA) in the context of the present specification relates to an RNA molecule capable of interfering with the expression (in other words: inhibiting or preventing the expression) of a gene comprising a nucleic acid sequence complementary or hybridizing to the sequence of the siRNA in a process termed RNA interference. The term siRNA is meant to encompass both single stranded siRNA and double stranded siRNA. siRNA is usually characterized by a length of 17-24 nucleotides. Double stranded siRNA can be derived from longer double stranded RNA molecules (dsRNA). According to prevailing theory, the longer dsRNA is cleaved by an endo- ribonuclease (called Dicer) to form double stranded siRNA. In a nucleoprotein complex (called RISC), the double stranded siRNA is unwound to form single stranded siRNA. RNA interference often works via binding of an siRNA molecule to the mRNA molecule having a complementary sequence, resulting in degradation of the mRNA. RNA interference is also possible by binding of an siRNA molecule to an intronic sequence of a pre-mRNA (an immature, non-spliced mRNA) within the nucleus of a cell, resulting in degradation of the pre-mRNA.

The term shRNA (small hairpin RNA) in the context of the present specification relates to an artificial RNA molecule with a tight hairpin turn that can be used to silence target gene expression via RNA interference (RNAi).

The term sgRNA (single guide RNA) in the context of the present specification relates to an RNA molecule capable of sequence-specific repression of gene expression via the CRISPR (clustered regularly interspaced short palindromic repeats) mechanism.

The term miRNA (microRNA) in the context of the present specification relates to a small non-coding RNA molecule (containing about 22 nucleotides) that functions in RNA silencing and post- transcriptional regulation of gene expression.

The term antisense oligonucleotide in the context of the present specification relates to an oligonucleotide having a sequence substantially complimentary to, and capable of hybridizing to, an RNA. Antisense action on such RNA will lead to modulation, particular inhibition or suppression of the RNA’s biological effect. If the RNA is an mRNA, expression of the resulting gene product is inhibited or suppressed. Antisense oligonucleotides can consist of DNA, RNA, nucleotide analogues and/or mixtures thereof. The skilled person is aware of a variety of commercial and non-commercial sources for computation of a theoretically optimal antisense sequence to a given target. Optimization can be performed both in terms of nucleobase sequence and in terms of backbone (ribo, deoxyribo, analogue) composition. Many sources exist for delivery of the actual physical oligonucleotide, which generally is synthesized by solid

Sequences similar or homologous (e.g., at least about 70% sequence identity) to the sequences disclosed herein are also part of the invention. In some embodiments, the sequence identity at the amino acid level can be about 80%, 85%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, 99% or higher. At the nucleic acid level, the sequence identity can be about 70%, 75%, 80%, 85%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, 99% or higher. Alternatively, substantial identity exists when the nucleic acid segments will hybridize under selective hybridization conditions (e.g., very high stringency hybridization conditions), to the complement of the strand. The nucleic acids may be present in whole cells, in a cell lysate, or in a partially purified or substantially pure form.

Calculations of "homology" or "sequence identity" or "similarity" between two sequences (the terms are used interchangeably herein) are performed as follows. The sequences are aligned for optimal comparison purposes (e.g., gaps can be introduced in one or both of a first and a second amino acid or nucleic acid sequence for optimal alignment and non-homologous sequences can be disregarded for comparison purposes). In a particular embodiment, the length of a reference sequence aligned for comparison purposes is at least 30%, particularly at least 40%, more particularly at least 50%, even more particularly at least 60%, and even more particularly at least 70%, 80%, 90%, 100% of the length of the reference sequence. The amino acid residues or nucleotides at corresponding amino acid positions or nucleotide positions are then compared. When a position in the first sequence is occupied by the same amino acid residue or nucleotide as the corresponding position in the second sequence, then the molecules are identical at that position (as used herein amino acid or nucleic acid "homology" is equivalent to amino acid or nucleic acid "identity"). The percent identity between the two sequences is a function of the number of identical positions shared by the sequences, taking into account the number of gaps, and the length of each gap, which need to be introduced for optimal alignment of the two sequences. In the case of circularly related proteins, the sequence of one of the partners needs to be appropriately split and aligned in two sections to achieve optimal alignment of the functionally equivalent residues necessary to calculate the percent identity.

In the context of the present specification, the terms sequence identity and percentage of sequence identity refer to a single quantitative parameter representing the result of a sequence comparison determined by comparing two aligned sequences position by position. Methods for alignment of sequences for comparison are well-known in the art. Alignment of sequences for comparison may be conducted by the local homology algorithm of Smith and Waterman, Adv. Appl. Math. 2:482 (1981), by the global alignment algorithm of Needleman and Wunsch, J. Mol. Biol. 48:443 (1970), by the search for similarity method of Pearson and Lipman, Proc. Nat. Acad. Sci. 85:2444 (1988) or by computerized implementations of these algorithms, including, but not limited to: CLUSTAL, GAP, BESTFIT, BLAST, FASTA and TFASTA. Software for performing BLAST analyses is publicly available, e.g., through the National Center for Biotechnology-Information (http://blast.ncbi.nlm.nih.gov/).

One example for comparison of amino acid sequences is the BLASTP algorithm that uses the default settings: Expect threshold: 10; Word size: 3; Max matches in a query range: 0; Matrix: BLOSUM62; Gap Costs: Existence 11 , Extension 1 ; Compositional adjustments: Conditional compositional score matrix adjustment. One such example for comparison of nucleic acid sequences is the BLASTN algorithm that uses the default settings: Expect threshold: 10; Word size: 28; Max matches in a query range: 0; Match/Mismatch Scores: 1.-2; Gap costs: Linear. Unless stated otherwise, sequence identity values provided herein refer to the value obtained using the BLAST suite of programs (Altschul et at,

J. Mol. Biol. 215:403-410 (1990)) using the above identified default parameters for protein and nucleic acid comparison, respectively.

In the context of the present specification, the term antibody refers to whole antibodies including but not limited to immunoglobulin type G (IgG), type A (IgA), type D (IgD), type E (IgE) or type M (IgM), any antigen binding fragment or single chains thereof and related or derived constructs. A whole antibody is a glycoprotein comprising at least two heavy (H) chains and two light (L) chains interconnected by disulfide bonds. Each heavy chain is comprised of a heavy chain variable region (VH) and a heavy chain constant region (CH). The heavy chain constant region is comprised of three domains, CH1 , CH2 and CH3. Each light chain is comprised of a light chain variable region (abbreviated herein as VL) and a light chain constant region (CL). The light chain constant region is comprised of one domain, CL. The variable regions of the heavy and light chains contain a binding domain that interacts with an antigen. The constant regions of the antibodies may mediate the binding of the immunoglobulin to host tissues or factors, including various cells of the immune system (e.g., effector cells) and the first component of the classical complement system. Similarly, the term encompasses a so-called nanobody or single domain antibody, an antibody fragment consisting of a single monomeric variable antibody domain.

In the context of the present specification, the term humanized antibody refers to an antibody originally produced by immune cells of a non-human species, the protein sequences of which have been modified to increase their similarity to antibody variants produced naturally in humans. The term humanized antibody as used herein includes antibodies in which CDR sequences derived from the germline of another mammalian species, such as a mouse, have been grafted onto human framework sequences. Additional framework region modifications may be made within the human framework sequences as well as within the CDR sequences derived from the germline of another mammalian species.

The term antibody-like molecule in the context of the present specification refers to a molecule capable of specific binding to another molecule or target with high affinity / a Kd < 10E-8 mol/l. An antibody-like molecule binds to its target similarly to the specific binding of an antibody. The term antibody-like molecule encompasses a repeat protein, such as a designed ankyrin repeat protein (Molecular Partners, Zurich), an engineered antibody mimetic proteins exhibiting highly specific and high-affinity target protein binding (see US2012142611 , US2016250341 , US2016075767 and US2015368302, all of which are incorporated herein by reference). The term antibody-like molecule further encompasses, but is not limited to, a polypeptide derived from armadillo repeat proteins, a polypeptide derived from leucine-rich repeat proteins and a polypeptide derived from tetratricopeptide repeat proteins.

The term antibody-like molecule further encompasses a specifically binding polypeptide derived from a protein A domain, fibronectin domain FN3, consensus fibronectin domains, a lipocalins (see Skerra, Biochim. Biophys. Acta 2000, 1482(1 -2):337-50), a polypeptide derived from a Zinc finger protein (see Kwan et al. Structure 2003, 11 (7):803-

813),

Src homology domain 2 (SH2) or Src homology domain 3 (SH3), a PDZ domain, gamma-crystallin, ubiquitin, a cysteine knot polypeptide or a knottin, cystatin,

Sac7d, a triple helix coiled coil (also known as alphabodies), a Kunitz domain or a Kunitz-type protease inhibitor and a carbohydrate binding module 32-2.

In the context of the present specification, the term fragment crystallizable (Fc) region is used in its meaning known in the art of cell biology and immunology; it refers to a fraction of an antibody comprising two identical heavy chain fragments comprised of a CH2 and a CH3 domain, covalently linked by disulfide bonds.

The term specific binding in the context of the present invention refers to a property of ligands that bind to their target with a certain affinity and target specificity. The affinity of such a ligand is indicated by the dissociation constant of the ligand. A specifically reactive ligand has a dissociation constant of < 10 ⁷ mol/L when binding to its target, but a dissociation constant at least three orders of magnitude higher in its interaction with a molecule having a globally similar chemical composition as the target, but a different three-dimensional structure.

As used herein, the term treating or treatment of any disease or disorder (e.g. cancer) refers in one embodiment, to ameliorating the disease or disorder (e.g. slowing or arresting or reducing the development of the disease or at least one of the clinical symptoms thereof). In another embodiment "treating" or "treatment" refers to alleviating or ameliorating at least one physical parameter including those which may not be discernible by the patient. In yet another embodiment, "treating" or "treatment" refers to modulating the disease or disorder, either physically, (e.g., stabilization of a discernible symptom), physiologically, (e.g., stabilization of a physical parameter), or both. Methods for assessing treatment and/or prevention of disease are generally known in the art, unless specifically described hereinbelow. A first aspect of the invention relates to an agent for prevention or treatment of metastasis, wherein said agent is selected from a pharmaceutical drug a polypeptide and wherein said agent promotes angiogenesis to achieve vascular normalization inside the tumor.

The pharmaceutical drug follows the Lipinski rules. In certain embodiments, the drug moiety has a molecular mass of more than 160 u but less than 1000 u, particularly less than 700 u, more particularly less than 500 u, comprises up to five hydrogen bond donators (e.g., oxygen and or nitrogen atoms with one H attached), up to ten hydrogen bond acceptors (e.g., oxygen or nitrogen atoms) and is characterized by an octanol-water partition coefficient logP of below 5.6 (any of these characteristics applied to the isolated drug moiety, without regard to the rest of compound). These are the so-called “Lipinski” rules of 5 (originally, referring to molecules between 160 and 500 u) for druglike compounds.

Vascular normalization is characterized by a homogeneous distribution of functional blood vessels in the tumor tissue, accompanied by a reduction or lack of intratumor hypoxia.

In certain embodiments, said agent is an agonistic ligand specifically binding to the EphB4 receptor, wherein said agonistic ligand induces EphB4 activation, particularly EphB4 phosphorylation.

In certain embodiments, the agonistic ligand targets endothelial cells, meaning the phosphorylation of the EphB4 receptor occurs in endothelial cells.

In certain embodiments, said agent is a polypeptide comprising or essentially consisting of

- EphrinB2 (UniProtKB - P52799 (EFNB2JHUMAN) SEQ ID NO 001), or a polypeptide which is at least 90% identical, particularly at least 95% identical, more particularly at least 98% identical to EphrinB2 and has at least (>) 80% of the biological activity, particularly > 90%, even more particularly > 95% of the biological activity of EphrinB2.

SEQ ID NO 001 (EphrinB2):

MAVRRDSVWK YCWGVLMVLC RTAISKSIVL EPIYWNSSNS KFLPGQGLVL YPQIGDKLDI ICPKVDSKTV GQYEYYKVYM VDKDQADRCT IKKENTPLLN CAKPDQDIKF TIKFQEFSPN LWGLEFQKNK DYYIISTSNG SLEGLDNQEG GVCQTRAMKI LMKVGQDASS AGSTRNKDPT RRPELEAGTN GRSSTTSPFV KPNPGSSTDG NSAGHSGNNI LGSEVALFAG IASGCIIFIV IIITLVVLLL KYRRRHRKHS PQHTTTLSLS TLATPKRSGN NNGSEPSDII IPLRTADSVF CPHYEKVSGD YGHPVYIVQE MPPQSPANIY YKV

The assay for the biological activity of Ephrin B2 is an EphB4 phosphorylation assay. When activated by Ephrin B2, EphB4 phosphorylation is increased. Antibody-mediated detection of EphB4 phosphorylation levels is performed as described in Hu et al. (Hu et al., Tumour Biol. 2014 Jul;35(7):7225-32.). In certain embodiments, said agent additionally comprises a fragment crystallizable (Fc) region of an antibody. In certain embodiments, said agent additionally comprises an IgG Fc region. In certain embodiments, said agent additionally comprises a human or humanized IgG Fc region.

In certain embodiments, said agent is selected from a peptide, an antibody, an antibody fragment, an aptamer or an antibody-like molecule and the agent is capable of specifically binding to EphB4 receptor and leading to the downstream signalling effects of Ephrin B2 binding to EphB4 receptor.

In certain embodiments, said agent is administered in combination with at least one non-agonist ligand, wherein each non-agonist ligand is specifically binding to one of

- VCAM1 (CD 106),

- CD49d,

- CD29,

IL6 receptor (CD126),

- IL6,

IL1 b receptor (IL1R1),

- IL1 b,

- F11R,

- ICAM1 , or

- ITGB2.

Alternatively, this combination may be formulated as a combination medicament.

In certain embodiments, said agent is administered in combination with at least one inhibitor nucleic acid sequence, wherein each inhibitor nucleic acid sequence is capable of downregulating or inhibiting expression of a target nucleic acid sequence encoding a protein selected from:

- VCAM1 (CD 106),

- CD49d,

- CD29,

IL6 receptor (CD126),

- IL6,

IL1 b receptor (IL1R1),

- IL1 b,

- F11R,

- ICAM1 , and

- ITGB2.

Alternatively, this combination may be formulated as a combination medicament.

In certain embodiments, the prevention or treatment of metastasis is applied for prevention or treatment of metastasis in breast cancer.

In certain embodiments, the agent is administered to a breast cancer patient who is characterized by having a high risk of developing metastasis, defined by a tumor being estrogen receptor-negative and/or progesterone receptor-negative, and/or a tumor being HER-2-positive, and/or a tumor grade according to the WHO classification of G2, G3, or G4, particularly G3 or G4, and/or a tumor size of more than 1 .5 cm in diameter, particularly more than 2 cm in diameter, and/or a tumor mitotic fraction of at least 15%, particularly at least 20%, and/or tumor cells found in the nearby lymph nodes.

The tumor grade according to the WHO classification is defined as in https://www.cancer.gov/about- cancer/diaqnosis-staqinq/proqnosis/tumor-qrade-fact-sheet.

The tumor mitotic fraction is defined as in https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1797032/.

In certain embodiments, the patient fulfils two of the above-mentioned criteria. In certain embodiments, the patient fulfils three of the above-mentioned criteria. In certain embodiments, the patient fulfils four of the above-mentioned criteria. In certain embodiments, the patient fulfils five of the above-mentioned criteria. In certain embodiments, the patient fulfils all of the above-mentioned criteria. The more criteria are fulfilled, the higher is the risk of developing or having metastasis. The agent described herein is especially beneficial for patients having metastasis or being at a high risk of developing metastasis.

In certain embodiments, a metastatic cancer cells’ capability of extravasation, infiltration into stroma of organs, survival or growth at distant sites is diminished by the agent.

In certain embodiments, the agent is administered to a patient who is characterized by the presence of CTC-clusters inside the bloodstream. In certain embodiments, the agent is administered to a patient who is characterized by the presence of CTC-WBC-clusters inside the bloodstream. In certain embodiments, the agent is administered to a patient who is characterized by the presence of CTC- neutrophil-clusters inside the bloodstream.

A second aspect of the invention relates to a nucleic acid molecule encoding the agent according to the first aspect for use in treatment or prevention of metastasis.

A third aspect of the invention relates to a method for assigning a likelihood of developing metastasis to a patient, wherein a high likelihood of developing metastasis is assigned if a. at least 17 genes, particularly 70%, of the following list of genes show a single-cell gene expression value above the detection threshold:

VEGFA, UBALD1 , CLCN2, ZNF771 , FAM13A, ABCC5, IKZF2, ZBTB43, IRS1 , BIVM, PIK3C2B, LRIG2, MLLT3, PLD6, P4HA1 , GDF15, LIPT1 , MACC1 , PPP1R3E, EPHX2, POFUT1 , NIPAL1 , CCNG2, HIF1A, and CTSS particularly wherein the detection threshold is defined as above 2 transcripts per million; or b. the expression of at least 17 genes, particularly 70%, of the following list of genes is increased inside the tumor compared to their expression in healthy tissue of the same patient: gene VEGFA increased 3.81 times (log fold change) gene UBALD1 increased 3.50 times (log fold change) gene CLCN2 increased 3.91 times (log fold change) gene ZNF771 increased 4 times (log fold change) gene FAM13A increased 4.60 times (log fold change) gene ABCC5 increased 2.47 times (log fold change) gene IKZF2 increased 3.78 times (log fold change) gene ZBTB43 increased 2.87 times (log fold change) gene IRS1 increased 2.44 times (log fold change) gene BIVM increased 3.73 times (log fold change) gene PIK3C2B increased 2.95 times (log fold change) gene LRIG2 increased 2.467 times (log fold change) gene MLLT3 increased 3.13 times (log fold change) gene PLD6 increased 2.88 times (log fold change) gene P4HA1 increased 2.23 times (log fold change) gene GDF15 increased 3.26 times (log fold change) gene LIPT1 increased 2.13 times (log fold change) gene MACC1 increased 2.90 times (log fold change) gene PPP1 R3E increased 2.78 times (log fold change) gene EPHX2 increased 3.18 times (log fold change) gene POFUT1 increased 2.14 times (log fold change) gene NIPAL1 increased 2.97 times (log fold change) gene CCNG2 increased 2.96 times (log fold change) gene HIF1 A increased 1 .05 times (log fold change) gene CTSS increased 2.27 times (log fold change).

Single-cell gene expression is assessed as described in the methods section.

Expression inside the tumor compared to healthy tissue of the same patient is assessed with RNAseq, RT-qPCR, serial analysis of gene expression (SAGE), or by hybridization microarray. For the purpose of determining whether a specific embodiment falls within the scope of the present invention, the method, unless indicated otherwise, shall by RNAseq.

A fourth aspect of the invention relates to the agent for use in treatment or prevention of metastasis in a patient according to the first aspect, or the nucleic acid molecule according to the second aspect, wherein a high likelihood of developing metastasis is assigned to said patient according to the method of the third aspect.

Similarly, a dosage form for the prevention or treatment of cancer is provided, comprising a nonagonist ligand or antisense molecule according to any of the above aspects or embodiments of the invention.

The skilled person is aware that any specifically mentioned drug may be present as a pharmaceutically acceptable salt of said drug. Pharmaceutically acceptable salts comprise the ionized drug and an oppositely charged counterion. Non-limiting examples of pharmaceutically acceptable anionic salt forms include acetate, benzoate, besylate, bitatrate, bromide, carbonate, chloride, citrate, edetate, edisylate, embonate, estolate, fumarate, gluceptate, gluconate, hydrobromide, hydrochloride, iodide, lactate, lactobionate, malate, maleate, mandelate, mesylate, methyl bromide, methyl sulfate, mucate, napsylate, nitrate, pamoate, phosphate, diphosphate, salicylate, disalicylate, stearate, succinate, sulfate, tartrate, tosylate, triethiodide and valerate. Non-limiting examples of pharmaceutically acceptable cationic salt forms include aluminium, benzathine, calcium, ethylene diamine, lysine, magnesium, meglumine, potassium, procaine, sodium, tromethamine and zinc.

Dosage forms may be for enteral administration, such as nasal, buccal, rectal, transdermal or oral administration, or as an inhalation form or suppository. Alternatively, parenteral administration may be used, such as subcutaneous, intravenous, intrahepatic or intramuscular injection forms. Optionally, a pharmaceutically acceptable carrier and/or excipient may be present.

Topical administration is also within the scope of the advantageous uses of the invention. The skilled artisan is aware of a broad range of possible recipes for providing topical formulations, as exemplified by the content of Benson and Watkinson (Eds.), Topical and Transdermal Drug Delivery: Principles and Practice (1st Edition, Wiley 2011 , ISBN-13: 978-0470450291); and Guy and Handcraft: Transdermal Drug Delivery Systems: Revised and Expanded (2 ^nd Ed., CRC Press 2002, ISBN-13: 978-0824708610); Osborne and Amann (Eds.): Topical Drug Delivery Formulations (1 ^st Ed. CRC Press 1989; ISBN-13: 978-0824781835).

Pharmaceutical Composition and Administration

Another aspect of the invention relates to a pharmaceutical composition comprising a compound of the present invention, or a pharmaceutically acceptable salt thereof, and a pharmaceutically acceptable carrier. In further embodiments, the composition comprises at least two pharmaceutically acceptable carriers, such as those described herein.

In certain embodiments of the invention, the compound of the present invention is typically formulated into pharmaceutical dosage forms to provide an easily controllable dosage of the drug and to give the patient an elegant and easily handleable product.

In embodiments of the invention relating to topical uses of the compounds of the invention, the pharmaceutical composition is formulated in a way that is suitable for topical administration such as aqueous solutions, suspensions, ointments, creams, gels or sprayable formulations, e.g., for delivery by aerosol or the like, comprising the active ingredient together with one or more of solubilizers, stabilizers, tonicity enhancing agents, buffers and preservatives that are known to those skilled in the art.

The pharmaceutical composition can be formulated for oral administration, parenteral administration, or rectal administration. In addition, the pharmaceutical compositions of the present invention can be made up in a solid form (including without limitation capsules, tablets, pills, granules, powders or suppositories), or in a liquid form (including without limitation solutions, suspensions or emulsions).

The dosage regimen for the compounds of the present invention will vary depending upon known factors, such as the pharmacodynamic characteristics of the particular agent and its mode and route of administration; the species, age, sex, health, medical condition, and weight of the recipient; the nature and extent of the symptoms; the kind of concurrent treatment; the frequency of treatment; the route of administration, the renal and hepatic function of the patient, and the effect desired. In certain embodiments, the compounds of the invention may be administered in a single daily dose, or the total daily dosage may be administered in divided doses of two, three, or four times daily.

In certain embodiments, the pharmaceutical composition or combination of the present invention can be in unit dosage of about 1-1000 mg of active ingredient(s) for a subject of about 50-70 kg. The therapeutically effective dosage of a compound, the pharmaceutical composition, or the combinations thereof, is dependent on the species of the subject, the body weight, age and individual condition, the disorder or disease or the severity thereof being treated. A physician, clinician or veterinarian of ordinary skill can readily determine the effective amount of each of the active ingredients necessary to prevent, treat or inhibit the progress of the disorder or disease.

The pharmaceutical compositions of the present invention can be subjected to conventional pharmaceutical operations such as sterilization and/or can contain conventional inert diluents, lubricating agents, or buffering agents, as well as adjuvants, such as preservatives, stabilizers, wetting agents, emulsifiers and buffers, etc. They may be produced by standard processes, for instance by conventional mixing, granulating, dissolving or lyophilizing processes. Many such procedures and methods for preparing pharmaceutical compositions are known in the art, see for example L. Lachman et al. The Theory and Practice of Industrial Pharmacy, 4th Ed, 2013 (ISBN 8123922892).

Method of Manufacture and Method of Treatment accordinp to the invention

The invention further encompasses, as an additional aspect, the use of a compound as identified herein (an agonistic ligand specifically binding to the EphB4 receptor), or its pharmaceutically acceptable salt, as specified in detail above, for use in a method of manufacture of a medicament for the treatment or prevention of metastasis.

Similarly, the invention encompasses methods of treatment of a patient having been diagnosed with a disease associated with metastasis. This method entails administering to the patient an effective amount of compound as identified herein (an agonistic ligand specifically binding to the EphB4 receptor), or its pharmaceutically acceptable salt, as specified in detail herein.

The invention further encompasses the use of primers and probes for the detection of the below- specified genes for use in the manufacture of a kit for the detection of metastasis.

Using this kit, a likelihood of developing metastasis is assigned to a patient, wherein a high likelihood of developing metastasis is assigned if a. at least 17 genes of the following list of genes show a single-cell gene expression value above the detection threshold:

VEGFA, UBALD1 , CLCN2, ZNF771 , FAM13A, ABCC5, IKZF2, ZBTB43, IRS1 , BIVM, PIK3C2B, LRIG2, MLLT3, PLD6, P4HA1 , GDF15, LIPT1 , MACC1 , PPP1 R3E, EPHX2, POFUT1 , NIPAL1 , CCNG2, HIF1A, and CTSS particularly wherein the detection threshold is defined as above 2 transcripts per million; or b. the expression of at least 17 genes of the following list of genes is increased inside the tumor compared to their expression in healthy tissue of the same patient: gene VEGFA increased 3.81 times (log fold change) gene UBALD1 increased 3.50 times (log fold change) gene CLCN2 increased 3.91 times (log fold change) gene ZNF771 increased 4 times (log fold change) gene FAM13A increased 4.60 times (log fold change) gene ABCC5 increased 2.47 times (log fold change) gene IKZF2 increased 3.78 times (log fold change) gene ZBTB43 increased 2.87 times (log fold change) gene IRS1 increased 2.44 times (log fold change) gene BIVM increased 3.73 times (log fold change) gene PIK3C2B increased 2.95 times (log fold change) gene LRIG2 increased 2.467 times (log fold change) gene MLLT3 increased 3.13 times (log fold change) gene PLD6 increased 2.88 times (log fold change) gene P4HA1 increased 2.23 times (log fold change) gene GDF15 increased 3.26 times (log fold change) gene LIPT1 increased 2.13 times (log fold change) gene MACC1 increased 2.90 times (log fold change) gene PPP1 R3E increased 2.78 times (log fold change) gene EPHX2 increased 3.18 times (log fold change) gene POFUT1 increased 2.14 times (log fold change) gene NIPAL1 increased 2.97 times (log fold change) gene CCNG2 increased 2.96 times (log fold change) gene HIF1 A increased 1 .05 times (log fold change) gene CTSS increased 2.27 times (log fold change).

Items

1 . An agent for use in prevention or treatment of metastasis, wherein said agent is selected from a pharmaceutical drug and a polypeptide, and wherein said agent promotes angiogenesis to achieve vascular normalization.

2. The agent for use according to item 1 , wherein said agent is an agonistic ligand specifically binding to the EphB4 receptor, wherein said agonistic ligand induces EphB4 activation, particularly EphB4 phosphorylation.

3. The agent for use according to item 1 or 2, wherein said agent is a polypeptide comprising or essentially consisting of

- EphrinB2 (UniProtKB - P52799 (EFNB2JHUMAN) SEQ ID NO 001), or a polypeptide which is at least 90% identical to EphrinB2 and has at least 80% of the biological activity of EphrinB2.

4. The agent for use in prevention or treatment of metastasis according to item 3, wherein said agent additionally comprises a fragment crystallizable (Fc) region.

5. The agent for use in prevention or treatment of metastasis according to item 1 or 2, wherein said agent is selected from a peptide, an antibody, an antibody fragment, an aptamer or an antibody-like molecule and the agent is capable of specifically binding to EphB4 receptor.

6. The agent for use in prevention or treatment of metastasis according to any one of the preceding items, wherein said agent is administered in combination with at least one nonagonist ligand, wherein each non-agonist ligand is specifically binding to one of VCAM1 (CD106), CD49d, CD29, IL6 receptor (CD126), IL6, IL1 b receptor (I L1 R1), IL1 b, F11 R,

ICAM1 , or ITGB2.

7. The agent for use in prevention or treatment of metastasis according to any one of the preceding items, wherein said agent is administered in combination with at least one inhibitor nucleic acid sequence, wherein each inhibitor nucleic acid sequence is capable of downregulating or inhibiting expression of a target nucleic acid sequence encoding a protein selected from VCAM1 (CD106), CD49d, CD29, IL6 receptor (CD126), IL6, IL1 b receptor (IL1 R1), IL1 b, F11 R, ICAM1 , and ITGB2.:

8. The agent for use in prevention or treatment of metastasis according to any one of the preceding items, wherein the prevention or treatment of metastasis is for use in treatment of breast cancer.

9. The agent for use in prevention or treatment of metastasis according to item 8, wherein the agent is administered to a breast cancer patient who is characterized by having a high risk of developing metastasis, defined by a tumor being estrogen receptor-negative and/or progesterone receptor-negative, and/or a tumor being HER-2-positive, and/or a tumor grade according to the WHO classification of G2, G3, or G4, particularly G3 or G4, and/or a tumor size of more than 1 .5 cm in diameter, particularly more than 2 cm in diameter, and/or a tumor mitotic fraction of at least 15%, particularly at least 20%, and/or tumor cells found in the nearby lymph nodes.

10. The agent for use in treatment or prevention of metastasis according to any one of the preceding items, wherein a metastatic cancer cells’ capability of extravasation, infiltration into stroma of organs, survival or growth at distant sites is diminished by the agent.

11 . The agent for use in treatment or prevention of metastasis according to any one of the preceding items, for use in a patient who is characterized by the presence of CTC-clusters, particularly by the presence of CTC-WBC-clusters, more particularly by the presence of CTC- neutrophil-clusters.

12. A nucleic acid molecule encoding the agent according to any one of the preceding items for use in treatment or prevention of metastasis.

13. A method for assigning a likelihood of developing metastasis to a patient, wherein a high likelihood of developing metastasis is assigned if a. at least 17 genes of the following list of genes show a single-cell gene expression value above the detection threshold:

VEGFA, UBALD1 , CLCN2, ZNF771 , FAM13A, ABCC5, IKZF2, ZBTB43, IRS1 , BIVM, PIK3C2B, LRIG2, MLLT3, PLD6, P4HA1 , GDF15, LIPT1 , MACC1 , PPP1 R3E, EPHX2, POFUT1 , NIPAL1 , CCNG2, HIF1A, and CTSS particularly wherein the detection threshold is defined as above 2 transcripts per million; or b. the expression of at least 17 genes of the following list of genes is increased inside the tumor compared to their expression in healthy tissue of the same patient: gene VEGFA increased 3.81 times (log fold change) gene UBALD1 increased 3.50 times (log fold change) gene CLCN2 increased 3.91 times (log fold change) gene ZNF771 increased 4 times (log fold change) gene FAM13A increased 4.60 times (log fold change) gene ABCC5 increased 2.47 times (log fold change) gene IKZF2 increased 3.78 times (log fold change) gene ZBTB43 increased 2.87 times (log fold change) gene IRS1 increased 2.44 times (log fold change) gene BIVM increased 3.73 times (log fold change) gene PIK3C2B increased 2.95 times (log fold change) gene LRIG2 increased 2.467 times (log fold change) gene MLLT3 increased 3.13 times (log fold change) gene PLD6 increased 2.88 times (log fold change) gene P4HA1 increased 2.23 times (log fold change) gene GDF15 increased 3.26 times (log fold change) gene LIPT1 increased 2.13 times (log fold change) gene MACC1 increased 2.90 times (log fold change) gene PPP1 R3E increased 2.78 times (log fold change) gene EPHX2 increased 3.18 times (log fold change) gene POFUT1 increased 2.14 times (log fold change) gene NIPAL1 increased 2.97 times (log fold change) gene CCNG2 increased 2.96 times (log fold change) gene HIF1 A increased 1.05 times (log fold change) gene CTSS increased 2.27 times (log fold change).

14. The agent for use in treatment or prevention of metastasis in a patient according to any one of items 1 to 11 , or the nucleic acid molecule according to item 12, wherein a high likelihood of developing metastasis is assigned to said patient according to the method of item 13.

Wherever alternatives for single separable features such as, for example, a ligand type or medical indication are laid out herein as “embodiments” or “items”, it is to be understood that such alternatives may be combined freely to form discrete embodiments of the invention disclosed herein. Thus, any of the alternative embodiments for a ligand may be combined with any of the alternative embodiments any medical indication or diagnostic method mentioned herein.

The invention is further illustrated by the following examples and figures, from which further embodiments and advantages can be drawn. These examples are meant to illustrate the invention but not to limit its scope.

Description of the Figures Fig. 1 Dynamic labeling of hypoxic CTCs and assessment of their metastatic potential. (A)

Schematic representation of the HIF1a activity reporter. The Hypoxia-Response Element (HRE) is repeated five times before a CMV minimal promoter (CMVmp), driving the expression of enhanced Yellow Fluorescent Protein (eYFP) upon HIF1a protein stabilization and activation. (B) Stacked bar graph showing the percent of CD31 positive (+) cells within the hypoxic or normoxic volume of the tumor in NSG-LM2-HIF1a reporter mice, defined by HIF1a reporter ( n = 4) or Pimonidazole staining (Pimo; n = 5). Error bars represents s.e.m. (C) Representative pictures of NSG-LM2-HIF1a reporter tumors stained for eYFP, CD31 and Dextran (Dex) (left) or with Pimo, CD31 and Dex {right). (D) Representative pictures of CTC clusters (fop) and single CTCs {bottom) from NSG-LM2-HIF1a reporter mice, positive or negative for eYFP expression. (E) The scatter dot plot shows the mean percentage of eYFP positive (+) single CTCs {n = 3) and CTC clusters {n = 3) from NSG-LM2-HIF1a reporter mice. Error bars represents s.e.m.; P value by two-tailed unpaired Student’s f-test is shown. (F) Schematic of the experimental design. From left to right, CTCs from donor NSG-LM2-HIF1a reporter mice were enriched from the whole blood. CTC clusters were then micromanipulated and separated into eYFP positive (+) or eYFP negative (-). Subsequently, equal number of cells from eYFP positive (+) or eYFP negative (-) clusters were intravenously injected in recipient NSG mice to assess their metastatic ability. (G) Kaplan-Meier survival analysis of NSG mice injected with eYFP (+) (n 4) or eYFP (-) {n = 3) CTC clusters. P value by log-rank test is shown.

Fig. 2 Hypoxic CTC clusters express a gene signature that is associated with a poor prognosis of Stage I breast cancer patients. (A) Schematic of the experimental design. From left to right, after sampling, blood specimens are processed with the Parsortix device to enrich for CTCs. Staining for CTC markers (EpCAM, HER2 and EGFR) and the HypoxiaRed dye is performed incassette. Subsequently, stained CTCs are released for micromanipulation, and isolated for single cell- resolution RNA-sequencing. (B) Representative pictures of CTC clusters from NSG-LM2-GFP/Luc and NSG-BR16 xenografts, as well as BR61 patient stained with HypoxiaRed and processed for RNA- sequencing. (C) Heat map showing differentially expressed genes between hypoxic (n = 14) and normoxic (n = 17) CTC clusters from NSG-LM2, NSG-BR16 and BR61 (False Discovery Rate; FDR < 0.25). (D) Kaplan-Meier curve showing overall survival rate of Stage I breast cancer patients expressing in their primary tumor high (quantile 4; Q4) or low (quantile 1 ; Q1) levels of genes upregulated in hypoxic CTC clusters (top). The number of patients that progressed at each time point is shown (bottom). P value by two-sided log-rank test is shown. (E) Receiver Operating Characteristic (ROC) curves for signatures of hypoxic CTC clusters (Donato-CTC; orange) and four other hypoxia signatures on overall survival. The Area Under the Curve (AUC) is shown in the legend.

Fig. 3 Knockdown of VEGFA increases CTC cluster shedding and metastasis formation. (A)

Tumor growth curves showing the mean tumor volume of NSG mice injected with LM2-mCherry/Luc cells and expressing a control shRNA (control), or h VEGFA shRNAs (h VEGFA sh-1 and sh-2) (n = 7). Error bars represents s.e.m.; P values by two-tailed paired Student’s f-test are shown. (B) The scatter plot shows the mean percentage of CD31 positive (+) cells within the primary tumor of NSG-LM2- control, NSG-LM2-h VEGFA sh-1 or sh-2 mice (n = 2). Error bars represents s.e.m.; P values by two- tailed unpaired Student’s f-test are shown. (C) Pie charts displaying the mean percentage of single CTCs and CTC clusters in NSG-LM2-control, NSG-LM2-h VEGFA sh-1 or sh-2 mice. The number of independent biological replicates (n) is shown for each condition. (D) The scatter dot plot shows the fold change of single CTCs and CTC clusters between NSG-LM2-control (n = 5), NSG-LM2-h VEGFA sh-1 (n = 7) and sh-2 (n = 7) mice. Error bars represents s.e.m.; P values by two-way Anova are shown. (E) The scatter dot plot shows the metastatic index of NSG-LM2-control (n = 7), NSG-LM2- h VEGFA sh-1 (n = 9) and sh-2 (n = 8) mice. Error bars represents s.e.m.; P values by two-tailed unpaired Student’s f-test are shown. (F) Representative bioluminescence images of lungs from NSG- LM2-control, NSG-LM2-h VEGFA sh-1 and sh-2 mice. Scale is expressed in Ph/s/cm ² /sr.

Fig. 4 Pro-angiogenic therapy reduces intra-tumor hypoxia and suppresses the formation of CTC clusters and metastasis. (A) Schematic of the experimental design. From left to right, cells are stably transduced with mVEGFAi6 ₄ -IRES-CD8aTr (mVIC) and injected into the mammary fat pad of NSG mice. During tumor growth, mice are treated biweekly with i.p. injections of EphrinB2 (EpB2). (B) Tumor growth curves showing the mean tumor volume of NSG mice injected with LM2-mCherry/Luc cells expressing mVIC or control CD8aTr (mC), treated with either control FC fragments (FC) or EpB2. The number of independent biological replicates (n) is shown for each condition. Error bars represents s.e.m.; P values by two-tailed unpaired Student’s f-test are shown. (C) Pie charts displaying the mean percentage of single CTCs and CTC clusters in NSG mice injected with LM2-mVIC or LM2-mC, treated with either FC or EpB2. The number of independent biological replicates (n) is shown for each condition. (D) The scatter dot plot shows the fold change of single CTCs and CTC clusters between NSG-LM2-mVIC and NSG-LM2-mC, treated with either FC or EpB2. The number of independent biological replicates (n) is shown for each condition. Error bars represents s.e.m.; P values by two-way Anova are shown. (E) The scatter dot plot shows the metastatic index of NSG-LM2-mVIC or LM2-mC mice, treated with either FC or EpB2. Error bars represents s.e.m.; P values by two-tailed unpaired Student’s f-test are shown. (F) Representative bioluminescence images of metastatic lungs from NSG- LM2-mVIC or NSG-LM2-mC mice, treated with either FC (top) or EpB2 (bottom). The scale is expressed in Ph/s/cm ² /sr. (G) Tumor growth curves showing the mean tumor volume of NSG-BR16- mCherry/Luc mice treated with either FC ( n = 5) or EpB2 (n = 5). Error bars represents s.e.m.; P values by two-tailed unpaired Student’s t- test are shown. (H) The scatter plot shows the fold change of single CTCs and CTC clusters between NSG-BR16-mCherry/Luc treated with either FC ( n = 4) or EpB2 ( n = 6). Error bars represents s.e.m.; P values by two-tailed unpaired Student’s f-test are shown. (I) The scatter plot shows the metastatic index of NSG-BR16-mCherry/Luc mice treated with either FC (n = 3) or EpB2 ( n = 6). Error bars represents s.e.m.; P values by two-tailed unpaired Student’s t- test are shown. (J) Representative bioluminescence images of metastatic lungs from NSG-BR16- mCherry/Luc treated with either FC (left) or EpB2 (right). The scale is expressed in Ph/s/cm ² /sr.

Fig. 5 In vitro characterization of the HIF1a activity reporter. (A-B) The plots show the activation of the HIF1a activity reporter in LM2-mCherry/Luc (A) and BR16-mCherry/Luc (B) cells, respectively, upon Deferoxamine (DFO) induction. DFO stimulation is performed for 4 or 15 hours in LM2- mCherry/Luc (A) and BR16-mCherry/Luc (B) cells, respectively ( n = 2). Error bars represents s.e.m.; P value by linear regression is shown. (C-D) The plots show the activation of the HIF1a activity reporter in LM2-mCherry/Luc (C) and BR16-mCherry/Luc (D) cells, respectively, upon incubation in 0.1% 02 (hypoxia) and subsequent incubation in 20% 02 (hypoxia). Hypoxia stimulation is performed for 4 days in LM2-mCherry/Luc (C) and BR16- mCherry/Luc (D) cells (n = 2). Error bars represents s.e.m.;

P value by linear regression is shown. (E) Representative pictures showing the induction of the HIF1a activity reporter upon incubation at 0.1% 02 of LM2-mCherry/Luc (left) or BR16-mCherry/Luc (right) cells. Pictures were taken at time (t) 0, t= 4 days of induction and after incubation in normoxia for 4 additional days (t= 8 days). (F-G) Representative western blot images showing HIF1a protein upon DFO induction for 4 hours in LM2-mCherry/Luc cells (F) or for 15 hours in BR16-mCherry/Luc cells (G). (H-l) Representative western blot images showing HIF1a protein upon 0.1 % 02 induction for 8 hours in LM2-mCherry/Luc (H) or in BR16-mCherry/Luc (I) cells. (J) Representative western blot image showing HIF1a protein upon DFO induction for 4 hours in LM2-mCherry/Luc cells, and residual HIF1a protein post-DFO treatment. (K) The plot shows eYFP expression from the HIF1a activity reporter in LM2-mCherry/Luc cells transduced with doxinducible HIF1 a shRNAs (sh-1 and sh-2) or sh- control (control) upon DFP stimulation. DFO stimulation was performed for 4 hours upon 5 days treatment with dox (n = 2). Error bars represents s.e.m.; P value by linear regression is shown.

Fig. 6 In vivo analysis of vascularized hypoxic regions of breast tumors. (A) Representative confocal images of NSG-LM2-mCherry/Luc (left) and NSG-BR16 (right) expressing the HIF1 a activity reporter and stained for human cytokeratin (hCK; cancer cells), CD31 , dextran (dex) and pimonidazole (pimo). (B) Plots showing the mean percentage of eYFP positive (+) cells within the tumor of LM2 (left; n = 3) and BR16 (right; n = 4) models. Error bars represents s.e.m. (C) Plots showing the mean percentage of pimo positive (+) cells within the tumor of LM2 (left; n = 5) and BR16 (right; n = 5) models. Error bars represents s.e.m. (D) Plots showing the mean percentage of colocalizing pimo and eYFP signals (cells) within the tumor of LM2 (left; n = 5) and BR16 (right; n = 5) models. Error bars represents s.e.m. (E) Plots showing the mean percentage of CD31 positive (+) cells within the tumor of LM2 (left; n = 5) and BR16 (right; n = 4) models. Error bars represents s.e.m. (F) Plots showing the mean percentage of CD31 (+) cells within the hypoxic volumes of the tumor in NSG-BR16 mice. Hypoxic volume is inferred from the eYFP signal of the HIF1a activity reporter (HIF1a) or with pimo staining. Error bars represents s.e.m (n = 4). (G) Representative confocal images of NSG-BR16 tumors expressing the HIF1a activity reporter and stained for human cytokeratin (hCK; cancer cells), CD31 and dextran (dex). The proximity of CD31-dex positive (+) vessels to eYFP (+) regions or pimo (+) regions is shown on the left and right panels, respectively. (H) Plots showing the mean percentage of CD31-dex (+) cells across the tumor volume of NSG-LM2 (left; n = 3) or NSGBR16 (right; n = 5) mice, respectively. Error bars represents s.e.m. (I) Tumor growth curves showing the mean tumor volume of NSG-LM2-mCherry/Luc (left) or NSG-BR16 (right) and expressing either the HIF1a activity reporter (HIF1a; n = 4 for LM2 and n = 7 for BR16) or empty vector (control; n = 4 for LM2 and n = 6 for BR16). Error bars represents s.e.m. ns = not significant by two-tailed unpaired Student’s f-test. (J) Scatter dot plots showing the logarithm to the base 10 (Log10) of total CTC numbers per ml of blood from NSG-LM2-mCherry/Luc (left) or NSG-BR16 (right) mice, expressing either the HIF1a activity reporter (HIF1a; n = 3 for LM2 and n = 4 for BR16) or empty vector (control; n = 3 for LM2 and BR16). Error bars represents s.e.m.; ns = not significant by two-way Anova. (K) Kaplan-Meier curves showing the overall survival rates of NSG-LM2-mCherry/Luc (left) or NSG-BR16 (right) mice, expressing either the HIF1a activity reporter (HIF1a; n = 4 for LM2 and n = 7 for BR16) or empty vector (control; n = 4 for LM2 and n = 6 for BR16). Error bars represents s.e.m.; ns = not significant by two-sided log-rank test. (L) Representative pictures showing CTC clusters (left) and single CTCs (right) from NSG-BR16 mice, stained for EpCAM and positive or negative for the HIF1a activity reporter expression (eYFP positive, top and eYFP negative, bottom). (M) The scatter dot plot shows the mean percentage of eYFP positive (+) single CTCs (n = 3) and CTC clusters (n = 3) from NSG-BR16 mice. Error bars represents s.e.m.; P value by two-tailed unpaired Student’s f-test is shown.

Fig. 7 Analysis of eYFP expression in single CTCs and CTC clusters. (A) Flow cytometry analysis showing CTCs from NSG-LM2-mCherry/Luc mice expressing the HIF1a reporter, gated for the Iog10 intensity of mCherry ( axis) versus the aspect ratio (Y axis). Aspect ratio represents the ratio between the Forward Side Scatter width (FSC-W) and the FSC-height (FSCH). (B) Flow cytometry analysis showing single CTCs and CTC clusters from NSG-LM2-mCherry/Luc mice expressing the HIF1a reporter, gated for the Iog10 intensity of mCherry (Xaxis) versus the Iog10 of the area in pixels per squared millimeter (px/mm2) (Y axis). (C) Flow cytometry analysis showing CTC clusters (top) and single CTCs (bottom) from NSG-LM2-mCherry/Luc mice expressing the HIF1a reporter, gated for the Iog10 intensity of eYFP (Xaxis). (D) Pie charts showing the percentage of eYFP positive cells within clusters of NSG-LM2 (top) or NSG-BR16 (bottom) mice expressing the HIF1a reporter. The number of independent biological replicates (n) is shown for each condition.

Fig. 8 Metastatic potential of hypoxic CTC clusters. (A) The scatter plot shows the number of cells per normoxic or hypoxic CTC cluster from NSG-LM2 mice expressing the HIF1a reporter. Error bars represents s.e.m.; P value by two-tailed unpaired Student’s f-test is shown. (B) Representative pictures of eYFP-negative (top) and eYFP-positive (bottom) CTC- White Blood Cells (WBC) clusters of NSG-LM2 mice expressing the HIF1a reporter. mCherry/Luc stains LM2 cells and CD45 stains WBCs. (C) Scatter plot showing the number of normoxic or hypoxic CTC-WBC clusters per ml of blood of NSG-LM2 mice expressing the HIF1a reporter. Error bars represents s.e.m.; P value by two-tailed unpaired Student’s f-test is shown. (D) The scatter plot shows the number of cells per hypoxic CTC- WBC clusters of NSG-LM2 mice expressing the HIF1a reporter. Error bars represents s.e.m. (E) Kaplan-Meier survival analysis of NSG mice injected with eYFP (+) CTC-WBC clusters ( n = 3). (F) Schematic of the experimental design. From left to right, BR16 cells expressing the HIF1a reporter were incubated in hypoxia (0.1% 02) or in normoxia (20% 02) for 4 days. Cell clusters were then collected and sorted into eYFP positive (+) or eYFP negative (-). Subsequently, an equal number of cells from eYFP positive (+) or eYFP negative (-) clusters were intravenously injected in recipient NSG mice to assess their metastatic ability. (G) The scatter plot shows the number of BR16 cells per normoxic or hypoxic cluster. Error bars represents s.e.m.; P value by two-tailed unpaired Student’s f- test is shown. (H) Flow cytometric analysis showing BR16 cells expressing the HIF1a reporter, gated for the Forward Side Scatter-Height (FSC-H; X axis) versus the Side scatter (SSC; Y axis). Cells are gated excluding debris. (I) Flow cytometric analysis showing BR16 cells expressing the HIF1a reporter, gated for the FSC-width (FSC-W; X axis) versus the SSC ( Y axis ). Cells are separated in singlets and clusters by assessing shape. (J) Flow cytometric analysis showing BR16 clusters expressing the HIF1a reporter, gated for the mCherry intensity (X axis) versus the SSC (Y axis). Clusters are separated by mCherry-positivity. (K) From left to right, shown are representative pictures of eYFP-positive (top) and eYFP-negative ( bottom ) clusters of BR16 cells expressing the HIF1a reporter. mCherry/Luc stains BR16 cells. Clusters of BR16-HIF1a are finally gated in eYFP-positive or eYFP-negative populations based on the expression of eYFP and also based on the results of the control BR16 (no eYFP) cell sort. (L) On the left, representative picture of lung metastasis in NSG mice injected with eYFP-positive (+) BR16 cells expressing the HIF1a reporter. BR16 cells are stained for human Cytokeratin (hCK). On the right, scatter plot shows the quantification of metastatic foci of NSG mice injected with eYFPpositive (+) or eYFP-negative (-) clusters of BR16 cells expressing the HIF1a reporter ( n = 3). Error bars represents s.e.m.; P value by two-tailed unpaired Student’s f-test is shown.

Fig. 9 Knockdown of HIF1a does not affect the metastatic potential of LM2 and BR16 cells. (A)

Representative western blot showing HIF1a protein in LM2-GFP/Luc (top) and BR16-GFP/Luc (bottom) cells expressing a control shRNA (control), h HIF1a sh-1 or h HIF1a sh-2 (sh-1 and sh-2).

DFO is used to induce HIF1a stabilization and Dox is used to activate the doxinducible knockdown. GAPDH is used as loading control. (B) Representative pictures showing LM2-GFP/Luc upon 5 days of Dox treatment. Dox-inducible shRNA constructs express TurboRFP together with the shRNA. (C) Representative pictures showing tumor (top) and metastatic lungs (bottom) of NSG-LM2-GFP/Luc mice expressing h HIF1a shRNAs. (D) Tumor growth curves showing the mean tumor volume of NSG mice injected with LM2-GFP/Luc (left) or BR16-GFP/Luc (right) and expressing a control shRNA (control), or h HIF1a shRNAs (h HIF1a sh-1 and sh-2). The number of independent biological replicates (n) is shown for each condition. Error bars represents s.e.m.; P values by two-tailed unpaired Student’s f-test are shown. (E) Scatter plots showing the logarithm to the base 10 (Log10) of total CTCs number per ml of blood of NSG-LM2-GFP/Luc (left) or NSG-BR16-GFP/Luc (right) mice expressing a control shRNA (control), or h HIF1a shRNAs ( UHIFIa sh-1 and sh-2). The number of independent biological replicates (n) is shown for each condition. Error bars represents s.e.m.; P values by two-way Anova are shown. (F) The scatter plots show the percentage of CTC clusters from NSG-LM2-GFP/Luc {left} or NSG-BR16-GFP/Luc {right) mice expressing a control shRNA (control), or h HIF1a shRNAs (h HIF1a sh-1 and sh-2). The number of independent biological replicates (n) is shown for each condition. Error bars represents s.e.m.; P values by two-tailed unpaired Student’s f- test are shown. (G) The scatter plot shows the metastatic index of NSG-LM2-GFP/Luc ( left) or NSG- BR16-GFP/Luc {right) mice expressing a control shRNA (control), or h HIF1a shRNAs (hH/Ffa sh-1 and sh-2). The number of independent biological replicates (n) is shown for each condition. Error bars represents s.e.m.; P values by two-tailed unpaired Student’s f-test are shown. (H) Kaplan-Meier curve showing overall survival rates of NSG-LM2-GFP/Luc {left) or NSG-BR16-GFP/Luc {right) mice expressing a control shRNA (control), or h HIF1a shRNAs (hH/Ffa sh-1 and sh-2). The number of independent biological replicates (n) is shown for each condition. P value by two-sided log-rank test is shown. (I) The scatter plot shows the mean percentage of pimonidazole (pimo) positive (+) cells colocalizing with the primary tumor cells of NSG-LM2-GFP/Luc {left) or NSG-BR16-GFP/Luc {right) mice expressing a control shRNA (control), or h HIF1a shRNAs (h HIF1a sh-1 and sh-2). The number of independent biological replicates (n) is shown for each condition. Error bars represents s.e.m.; P values by two-tailed unpaired Student’s f-test are shown.

Fig. 10 HypoxiaRed correlates with HIF1a activity reporter and reacts with live hypoxic CTC clusters. (A) The scatter plots show the percentage of HypoxiaRed-positive cells from LM2-GFP/Luc or BR16-GFP/Luc cells incubated in hypoxic conditions (0.1% 02) for 8 hours or 15 hours, respectively {n = 3). Error bars represents s.e.m.; P values by two-tailed unpaired Student’s f-test are shown. (B) Flow cytometric analysis showing LM2 cells expressing the HIF1a reporter, gated for the gradient RMS (X axis) versus the aspect ratio (Y axis). Gradient RMS represents the sharpness of the event acquired in brightfield. Aspect ratio represents the ratio between the Forward Side Scatter (FSC) width (FSC-W) and the FSC-height (FSC-H). (C) Flow cytometric analysis showing the correlation between the expression of eYFP (in Log10 intensity; X axis) and HypoxiaRed staining intensity (in Log10 intensity; Y axis) in LM2 {left) and BR16 {right) cells expressing the HIF1a reporter. Controls (uninduced and induced with hypoxia 0.1% 02 for 4 days) are shown for both models. (D) Representative pictures showing consistency between the expression of eYFP and HypoxiaRed staining intensity in LM2 cells expressing the HIF1a reporter. (E) On the top panel, flow cytometric analysis showing CTCs from NSG-LM2 mice cells expressing the HIF1a reporter, gated for the Iog10 intensity of GFP (X axis) versus the aspect ratio (Y axis). On the bottom panel, flow cytometric analysis showing single CTCs and CTC clusters from NSG-LM2 mice cells expressing the HIF1a reporter, gated for the Iog10 intensity of GFP (X axis) versus the Iog10 of the area in pixels per squared millimeter (px/mm2) (Y axis). (F) On the top panel, flow cytometric analysis showing CTC clusters ( left) and single CTCs {right) from NSG-LM2-GFP/Luc mice, gated for the Iog10 intensity of HypoxiaRed (Xax/s). On the bottom panel, representative pictures showing HypoxiaRed-positive or - negative CTC clusters ( left) and single CTCs {right). (G) The scatter plot shows the mean percentage of HypoxiaRed positive (+) single CTCs and CTC clusters from NSG-LM2-GFP/Luc {n = 4) and NSG- BR16-GFP/Luc {n = 3) mice, as well as from BR61 patient {n = 5). Error bars represents s.e.m.; P value by two-tailed unpaired Student’s f- test is shown. (H) The scatter plot shows the number of cells per CTC cluster in those clusters that are positive (+) or negative (-) for HypoxiaRed staining in NSG- LM2-GFP/Luc ( n = 3) and NSGBR16-GFP/Luc ( n = 4) mice as well as in BR61 patient ( n = 3). Error bars represents s.e.m.; P value by two-tailed unpaired Student’s f-test is shown.

Fig. 11 Single-cell resolution RNA-sequencing analysis of hypoxic and normoxic CTC clusters.

(A) Heat map showing the sample categories based on HypoxiaRed staining, expression score (based on HIF1a and VEGFA expression: negative for quantiles Q1 and Q2, positive for quantiles Q3 and Q4) and quantile score. The expression score is calculated by the mean expression values of HIF1a and VEGFA genes and by ranking samples by their score. (B) Scatter plot showing the number of detected genes in normoxic versus hypoxic CTC clusters. The lower and upper hinges of the boxplot correspond to the 25th and 75th percentiles and whiskers are extended to the most extreme data points. P value by two-sided Wilcoxon rank sum test is shown. (C) Scatter plot showing the cell cycle phase of normoxic versus hypoxic CTC clusters. (D) Heat map showing the expression of Epithelial- to-Mesenchymal transition (EMT) genes in normoxic versus hypoxic CTC clusters. Colors are based on normalized length-scaled transcript per million (TPM) values. (E) Kaplan-Meier curves showing overall survival rate of Stage I breast cancer patients expressing in their primary tumor high (quantile 4; Q4) or low (quantile 1 ; Q1) levels of genes upregulated in the hypoxic CTC clusters signature (Donato; orange) versus four other hypoxia signatures (grey). P value by two-sided log-rank test is shown.

Fig. 12 Knockdown of VEGFA increases CTC cluster shedding and metastasis formation. (A)

Bar plot showing the mean fold change of expression (qPCR) of h VEGFA or m Vegfa in LM2- mCherry/Luc (left) or 4T1-mCherry/Luc (right) cells, respectively (n = 3). DFO is used to induce HIF1a stabilization and Dox is used to activate the dox-inducible knockdown. GAPDH is used as normalization control. Error bars represents s.e.m.; P value by two-tailed unpaired Student’s f-test is shown. On the right side, representative pictures of LM2-mCherry/Luc cells upon 5 days of Dox treatments. Dox-inducible shRNA constructs express TurboGFP together with the shRNA. (B) Representative pictures of the primary tumor (top) and metastatic lungs (bottom) of NSG-LM2- mCherry/Luc mice expressing h HIF1a shRNAs. (C) Tumor growth curves showing the mean tumor volume of NSG mice injected with 4T1-mCherry/Luc cells expressing a control shRNA (control) or m Vegfa shRNAs (m Vegfa A sh-1 and sh-2;). The number of independent biological replicates (n) is shown for each condition. Bars represents s.e.m.; P values by two-tailed unpaired Student’s f-test are shown. (D) The scatter plot shows the mean percentage of CD31 positive (+) cells within the primary tumor of NSG-4T1-mCherry/Luc mice expressing a control shRNA, m Vegfa sh-1 or m Vegfa sh-2. The number of independent biological replicates (n) is shown for each condition. Error bars represents s.e.m.; P values by two-tailed unpaired Student’s f-test are shown. (E) The scatter plots show the mean percentage of pimonidazole positive (+) cells colocalizing with primary tumor cells of NSG-LM2- mCherry/Luc (left) or NSG-4T1-mCherry/Luc (right) mice expressing a control or VEGFA knockdown. The number of independent biological replicates (n) is shown for each condition. Error bars represents s.e.m.; P values by two-tailed unpaired Student’s f-test are shown. (F) Scatter plots showing the logarithm to the base 10 (Log10) of total CTC counts per ml of blood in NSG-LM2-mCherry/Luc (left) or NSG-4T1-mCherry/Luc ( right) mice expressing a control shRNA (control), or VEGFA knockdown. The number of independent biological replicates (n) is shown for each condition. Error bars represents s.e.m.; P values by two-way Anova are shown. (G) Pie charts displaying the mean percentage of single CTCs and CTC clusters in NSG-4T1-mCherry/Luc mice expressing a control shRNA, m Vegfa sh-1 or m Vegfa sh-2. The number of independent biological replicates (n) is shown for each condition. (H) The scatter plot shows the mean fold change of single CTCs and CTC clusters in NSG-4T1- mCherry/Luc-h VEGFA sh-1 ( n = 6) and NSG-4T1-mCherry/Luc-h VEGFA sh-2 ( n = 7) mice, compared to NSG-4T1- mCherry/Luc-control ( n = 5). Error bars represents s.e.m.; P values by two-way Anova are shown. (I) The scatter plot shows the metastatic index of NSG-4T1-mCherry/Luc-control ( n = 6), NSG-4T 1 -mCherry/Luc-h VEGFA sh-1 (n = 7) and NSG-4T 1 -mCherry/Luc-h VEGFA sh-2 (n = 7) mice. Error bars represents s.e.m.; P values by two-tailed unpaired Student’s f-test are shown. (J) Representative bioluminescence images of metastatic lungs of NSG-4T1-mCherry/Luc-control, NSG- 4T1 -mCherry/Luc-h VEGFA sh-1 and NSG-4T1 -mCherry/Luc-h VEGFA sh-2 mice. Scale is expressed in Ph/s/cm ² /sr.

Fig. 13 Pro-angiogenic therapy reduces CTC clusters shedding and metastasis formation in fast-growing tumor models. (A) On the left, schematic representation of the control mCD8aTr vector (mC; top) and the mVEGFA164-IRES-mCD8aTr vector (mVIC; bottom) for the overexpression of mVEGFA164. On the right, representative flow cytometry analysis showing CD8a expression, visualized through staining with anti-CD8a-APC antibodies in control, mC and mVIC 4T1-mCherry/Luc cells (quadrant II). (B) Representative flow cytometry analysis showing selected mC and mVIC clones from LM2-mCherry (left) or 4T1-mCherry (right) expressing CD8a (X axis). (C) The scatter plot shows the mean percentage of CD31 positive (+) cells within the primary tumor of NSG-LM2-mCherry/Luc mice expressing mVIC or mC, treated with either control FC fragments (FC) or EphrinB2 (EpB2). The number of independent biological replicates (n) is shown for each condition. Error bars represents s.e.m.; P values by two-tailed unpaired Student’s f-test are shown. (D) The scatter plot shows the mean percentage of pimonidazole positive (+) cells colocalizing with the primary tumor of NSG-LM2- mCherry/Luc mice expressing mVIC or mC, treated with either FC or EpB2. The number of independent biological replicates (n) is shown for each condition. Error bars represents s.e.m.; P values by two-tailed unpaired Student’s f-test are shown. (E) Scatter dot plots showing the logarithm to the base 10 (Log10) of total CTC counts per ml of blood of NSG-LM2-mCherry/Luc mice expressing mVIC or mC, and treated with either FC or EpB2. The number of independent biological replicates (n) is shown for each condition. Error bars represents s.e.m.; P values by two-way Anova are shown. (F) Kaplan-Meier curve showing overall survival rates of NSG-LM2-mCherry/Luc mice expressing mVIC or mC, and treated with either FC or EpB2. The number of independent biological replicates (n) is shown for each condition. P value by two-sided log-rank test is shown. (G) Tumor growth curves showing the mean tumor volume of NSG mice injected with 4T1-mCherry/Luc cells expressing mVIC or mC, and treated with either FC or EpB2. The number of independent biological replicates (n) is shown for each condition. Error bars represents s.e.m.; P values by two-tailed unpaired Student’s f-test are shown. (H) Kaplan-Meier curve showing overall survival rates of NSG-4T1-mCherry/Luc mice expressing mVIC or mC, and treated with either FC or EpB2. The number of independent biological replicates (n) is shown for each condition. P value by two-sided log-rank test is shown. Fig. 14 Pro-angiogenic therapy reduces CTC clusters shedding and metastasis formation in the slow-growing tumor model BR16. (A) The scatter plot shows the mean percentage of CD31 positive (+) cells within the primary tumor of NSG-BR16-mCherry/Luc mice treated with either control FC fragments (Control; n = 10) or EphrinB2 (EpB2; n = 6). Error bars represents s.e.m.; P values by two- tailed unpaired Student’s f-test are shown. (B) The scatter plot shows the mean percentage of pimonidazole positive (+) cells colocalizing with the primary tumor of NSGBR16- mCherry/Luc mice treated with either control ( n = 5) or EpB2 ( n = 9). Error bars represents s.e.m.; P values by two-tailed unpaired Student’s f-test are shown. (C) Scatter dot plots showing the logarithm to the base 10 (Log10) of total CTCs number per ml of blood of NSG-BR16-mCherry/Luc mice treated with either control ( n = 5) or EpB2 ( n = 6). Error bars represents s.e.m.; P values by two-way Anova are shown.

Examples Example 1:

The inventors first sought to dynamically trace spontaneous hypoxic events by generating an activity reporter vector for HIF1 a (HIF1 a reporter), the master regulator of hypoxia, designed to express enhanced yellow fluorescent protein (eYFP) under the control of hypoxia-response elements (HRE) repeats (Fig. 1A). Upon transduction of the HIF1a reporter in breast CTC-derived cells (BR16) as well as in established metastatic breast cancer cells (MDA-MB-231 lung metastatic variant, referred to as “LM2”) already expressing mCherry-Luciferase (mCherry/Luc), the inventors confirmed elevated eYFP levels both upon treatment with the HIF1a-stabilizer deferoxamine (DFO) and also as a consequence to incubation in hypoxic conditions (0.1% O2), compared to control cells (Fig. 5A-J). As a further control, the inventors confirmed that HIF1a knockdown completely abolished the ability of transduced cells to express eYFP (Fig. 5K), confirming the specificity of the reporter system for HIF1a. The inventors then injected BR16- and LM2-HIF1a reporter cells into the mammary fat pad of immunocompromised (NSG) mice and monitored spontaneous tumor progression. The inventors observed the presence of eYFP-positive regions in distinct areas of the primary tumor, largely overlapping with pimonidazole staining, a well-established marker of hypoxia (Fig. 6A-D). Of note, hypoxic regions within the primary tumor were also found in proximity to functional blood vessels, highlighting a potential accessibility route of hypoxic cancer cells to the vasculature (Fig. 1B,C and 6E-H). The inventors next assessed tumor progression as well as the number and composition of spontaneously-generated CTCs. The expression of the HIF1a reporter did not alter primary tumor growth, total CTC numbers nor time to metastasis formation (Fig. 6I-K). Strikingly however, the inventors found that while the majority of single CTCs are normoxic (i.e. eYFP-negative), CTC clusters are largely hypoxic in both models (Fig. 1D,E and 6L,M), with the majority of the cells in each cluster being eYFP-positive (Fig. 7A-D). Of note, despite the fact that in the slow-growing BR16 model only 5.6% of primary tumor cells were eYFP-positive, the inventors found 80.6% of CTC clusters to be positive for eYFP (Fig. 6B,M), suggesting their origin from hypoxic tumor areas. To assess whether hypoxic CTC clusters are endowed with a greater metastatic potential compared to their normoxic counterpart, the inventors first injected LM2-HIF1a reporter cells in the mammary fat pad of NSG mice and upon tumor development, spontaneously-generated CTC clusters were individually micromanipulated and separated into eYFP-positive or eYFP-negative (Fig. 1F). Importantly, while hypoxic CTC clusters were generally found to contain a higher number of cells than the normoxic ones (a mean of 5.3 cells per hypoxic CTC cluster versus a mean of 2.82 cells per normoxic CTC cluster; P<0.001) (Fig. 8A), the inventors intravenously injected a total of 100 cells per recipient tumor-free mouse for all groups, without disrupting the multicellular structure of CTC clusters (Fig. 1F). Mice injected with hypoxic CTC clusters developed metastasis earlier and survived for a shorter time than those injected with normoxic CTC clusters, highlighting the higher metastasis-seeding ability of hypoxic CTC clusters (Fig. 1G). Of note, the inventors also realized that virtually all CTC-WBC clusters from this model were also hypoxic (Fig. 8B-D), and as expected, their direct metastatic ability exceeded that of hypoxic CTC clusters that were not associated to WBCs (Fig. 8E). To gain further insights into the direct contribution of hypoxia in increasing the metastatic potential of CTCs, the inventors incubated BR16-HIF1a reporter cells in either normoxic (20% O2) or hypoxic (0.1% O2) conditions for four days and then isolated eYFP-positive or eYFP-negative cell clusters for intravenous injection into tumor-free recipient NSG mice (Fig. 8F-K). Also in this case, mice injected with hypoxic clusters developed a higher number of metastatic foci compared to mice injected with normoxic clusters (Fig. 8L), confirming the higher metastasis-seeding ability of hypoxic clusters. Thus, clustered cancer cells display marked hypoxic conditions in circulation, while single CTCs are generally normoxic. Hypoxic CTC clusters also display a higher ability to seed metastasis compared to their normoxic counterpart, highlighting that hypoxia may be both a trigger to CTC cluster generation and an enhancer of CTCs’ metastatic potential.

Example 2

The inventors next thought of testing whether HIF1a, beyond its role as an established hypoxia- associated transcription factor, is also directly involved in the mechanisms that promote CTC clusters generation and their higher ability to metastasize. To this end, the inventors generated inducible HIFIot knockdown in LM2 and BR16 cells, resulting in HIF1a suppression upon treatment with Doxycycline (Dox) (Fig. 9A,B). The inventors then injected these cells in the mammary fat pad of NSG mice and monitored primary tumor growth, CTC generation, and spontaneous metastasis formation upon Dox treatment. Unexpectedly, while Dox treatment successfully enabled the expression of HIF1a shRNAs throughout the experiment in vivo (Fig. 9C), the inventors did not observe any difference in primary tumor size, CTC composition or metastasis formation between HIF1a knockdown and control mice (Fig. 9D-H). Of note, pimonidazole staining also highlighted that HIF1a knockdown did not decrease the overall levels of intra-tumor hypoxia, but it rather increased them (Fig. 9I). Thus, while HIFIot is expressed by hypoxic cells, it does not appear to be directly required for the realization of intratumor hypoxia, nor for the generation of CTC clusters or metastasis formation in the tested models.

Example 3

Given that HIFIot knockdown did not prevent the generation of CTC clusters, the inventors sought to take an unbiased approach to interrogate the gene expression profile of hypoxic clusters, aiming to identify genes that are not only upregulated as a consequence to hypoxia, but that may also play a pivotal role in promoting CTC clusters generation and metastasis. To this end, the inventors isolated live CTCs from a breast cancer patient (BR61) and two breast cancer xenografts (BR16 and LM2) and labeled them with HypoxiaRed, a cell-permeable dye that tags hypoxic cells based on their nitroreductase activity, allowing to compare the gene expression profile of hypoxic versus normoxic CTC clusters (Fig. 2A). Of note, in control experiments the inventors demonstrate that HypoxiaRed- positivity increases in hypoxic conditions (0.1% O2) as well as it correlates with eYFP expression in both BR16- and LM2-HIF1a reporter cells upon hypoxia induction (Fig. 10A-D). Following CTCs isolation and HypoxiaRed staining, as expected, the inventors observed a higher HypoxiaRed- positivity in CTC clusters compared to single CTCs (mainly visible in xenografts, where the highest number of CTCs was retrieved; Fig. 10E-G). The inventors then individually micromanipulated a total of 28 HypoxiaRed-positive versus 33 HypoxiaRed-negative CTC clusters and processed them for RNA sequencing (Fig. 2B). Of note, since hypoxic CTC clusters generally contain more cells than their normoxic counterpart (Fig. 10H), typically resulting in a higher number of genes detected ( data not shown), to avoid technical biases the inventors only considered 2- and 3-cell clusters for the RNA sequencing analysis. Differential expression analysis highlighted that hypoxic CTC clusters (as defined by HypoxiaRed-positivity as well as expression of HIF1a and VEGFA; Fig. 11 A) differ in the expression of 32 genes compared to their normoxic counterpart (as defined by HypoxiaRed-negativity and absence of expression of HIF1a and VEGFA; Fig. 11 A), of which 25 upregulated and 7 downregulated (Fig2C and Table S1,2). In contrast, no changes were observed between hypoxic and normoxic clusters in terms of total number of detected genes, nor expression of genes related to cell cycle or epithelial-to-mesenchymal transition (Fig. 11B-D). Strikingly, the expression of the 25-gene signature found upregulated in hypoxic CTC clusters strongly predicts the clinical outcome of patients with early breast cancer (Stage 1 disease, all subtypes), with patients expressing high levels of the signature genes displaying a significantly shorter progression-free survival (P=0.031) and overall survival (P= 0.037) compared to patients with lower expression of the same genes (Fig. 2D and S7E). Of note, the inventors observe that this predictive value is largely superior when compared to previous hypoxia-related signatures (mostly obtained from in vitro data and bulk analyses) (Fig. 2E and 11F,G). Thus, hypoxia triggers the expression of a defined gene set in CTC clusters in vivo, highly associated to poor prognosis in breast cancer patients with early disease, and displaying a better predictive value compared to previously-identified hypoxia signatures.

Example 4

The inventors next asked whether the expression of VEGFA itself in cancer cells (as part of the inventors’ hypoxic CTC clusters signature but also as a master angiogenesis regulator) could play a role in promoting CTC clusters generation and metastasis. To this end, the inventors used Dox- inducible vectors expressing GFP along with shRNAs targeting the human or mouse VEGFA transcript and transduced them in LM2 cells or the mouse breast cancer 4T1 cells, respectively. Upon Dox stimulation, the inventors confirmed both the knockdown of VEGFA using two independent shRNAs as well as the expression of GFP (Fig. 12A). The inventors then injected LM2- and 4T1-shVEGFA cells in the mammary fat pad of NSG mice and monitored tumor progression. As expected, tumors expressing VEGFA shRNAs retained shRNA expression in vivo, grew slower and presented with a decreased percent of CD31 -positive cells relative to the total tumor area (i.e. less blood vessels), along with a higher positivity for pimonidazole (Fig. 3A,B and 12B-E). Strikingly however, despite the slower growth rate of VEGFA knockdown tumors, the inventors observed a marked increase in overall CTC number and a shift towards CTC clusters production compared to control tumors of the same size (T000 mm ³ for LM2 and 700mm ³ for 4T1) in both models (from 11.4% to 25.6-24.5% of CTC clusters in the LM2 model and from 25.8% to 32-40.6% of CTC clusters in the 4T1 model, respectively) (Fig. 3C,D and 12F-H). This increased CTC cluster ratio and overall CTC numbers also led to increased metastasis formation in animals bearing a VEGFA knockdown tumor (Fig. 3E,F and 121, J). Together, the inventors’ results suggest that VEGFA suppression leads to tumor shrinkage, slower growth rate and reduced vascularization but it also promotes intra-tumor hypoxia, leading to increased CTC clusters shedding and accelerated metastasis formation.

Example 5

The inventors then sought to address whether the opposite scenario, i.e. an increased tumor vascularization, could serve as a strategy to prevent the generation of CTC clusters and delay metastasis formation. The inventors first tested their hypothesis in two fast-growing breast cancer models, i.e. LM2 and 4T1 injected in NSG mice. As a first step, the inventors transduced both cell lines with a bicistronic construct expressing the mouse form of VEGFA ( _m VEGFAi6 ₄ ) along with the truncated form of mouse CD8a transmembrane protein (mCD8aTr), and then they selected clones with similar levels of _m VEGFAi6 ₄ expression, prospectively inferred through anti-mCD8aTr live staining (Fig. 13A,B). The inventors then injected two LM2- _m VEGFAi6 ₄ -IRES-mCD8aTr clones (LM2-mVIC) and a control LM2-mCD8aTr clone (LM2-mC) in the mammary fat pad of NSG mice, simultaneously treated with either EphrinB2 Fc chimera protein (previously shown to activate EphB4 signaling and to ensure normal and functional angiogenesis along with elevated VEGFA levels) (26, 27) or with Fc fragments as controls (Fig. 4A). The inventors observed that while EphrinB2 or mVIC expression alone did not dramatically alter primary tumor growth rate, the simultaneous expression of mVIC and EphrinB2 treatment led to the formation of tumors characterized by a similar growth rate, yet able to reach the maximum allowed size in our license (2’800 mm ³ ) without causing any sign of distress in the tumor-bearing mice (Fig. 4B). Primary tumor analysis revealed that LM2-mVIC tumors treated with EphrinB2 also displayed an increased CD31 -positivity and decreased pimonidazole reactivity (Fig. 13C,D), consistent with reduced intra-tumor hypoxia. Most importantly, despite having significantly larger tumors, mice with LM2-mVIC tumors treated with EphrinB2 generated less CTCs and displayed a reduced CTC clusters ratio (37.3% of CTC clusters in LM2-mC mice treated with EphrinB2 versus 24.2-24.9% of CTC clusters in LM2-mVIC mice treated with EphrinB2) compared to control animals (Fig. 4C,D and 13E), leading to a marked reduction in spontaneous metastasis formation and longer overall survival (Fig.4E,F and 13F). As a further confirmation in an independent model, the inventors also observed a higher tumor growth rate associated to a longer overall survival in mice carrying a 4T1-mVIC tumor and treated with EphrinB2 (Fig. 13G,H). Lastly, the inventors asked whether these findings were reproducible in CTC-derived BR16 breast cancer cells, inherently characterized by the ability to form slow-growing tumors and displaying a higher number of functional vessels and lower intra-tumor hypoxia compared to the LM2 model (Fig. 6B,H). In this case, given the above, the inventors tested whether the administration of EphrinB2 alone (i.e. without mVIC expression) would be sufficient to recapitulate the effects observed in the LM2 and 4T1 models. Treatment of BR16 xenografts with EphrinB2 led to the formation of significantly larger tumors (Fig. 4G) characterized by higher CD31 -positivity and reduced reactivity to pimonidazole (Fig. 14A,B). Strikingly, EphrinB2- treated BR16 xenografts failed to generate CTC clusters (20.4% of CTC clusters for controls and 0% of CTC clusters for EphrinB2) and displayed overall reduced CTC shedding (Fig. 4H and 14C), leading to the suppression of spontaneous metastasis formation (Fig. 4I,J). Thus, these results highlight that a pro-angiogenic approach results in the formation of larger tumors, yet it reduces intratumor hypoxia and suppresses the shedding of CTC clusters, leading to impaired metastasis formation.

Discussion

Together, the inventors’ findings suggest that intra-tumor hypoxia is a main trigger to CTC clusters intravasation, and that hypoxic CTC clusters are more efficient metastatic precursors compared to normoxic CTCs. As a consequence, VEGFA-targeting results in tumor shrinkage but at the same time it induces increased hypoxia levels, CTC clusters generation and metastasis. Conversely, the inventors propose that a pro-angiogenic approach may on the one hand increase tumor growth rate, yet suppress intra-tumor hypoxia and intravasation of CTC clusters, leading to a reduction in metastasis formation.

Materials and methods

Patients

Patient blood specimens were obtained at the University Hospital Basel through the study protocols (EKNZ BASEC 2016-00067 and EK 321/10), approved by the local ethics committee (EKNZ, Ethics Committee northwest/central Switzerland). The patients involved were characterized by having invasive breast cancer, high tumor load and progressive disease. In particular, breast cancer patient BR61 was characterized by having a ER-positive, PR-negative and HER2-negative disease at primary tumor diagnosis, and later developed bone, lymph node, soft tissue, brain, adrenal gland and pancreatic metastases at the time of CTC isolation. BR61 donated 7.5 -15ml blood in EDTA vacutainers during multiple timepoints at disease progression, upon written consent.

Cell lines and culture conditions

MDA-MB-231-LM2 (LM2) human triple negative breast cancer cell line was obtained from Dr. Joan Massague, MSKCC, NY, USA. CTC-derived BR16 cells were generated and cultured from the corresponding patient as previously described (Alix-Panabieres & Pantel, Nature reviews. Cancer 14, 623-631 (2014)). 4T1 murine breast cancer cells were purchased from ATCC (4T1 ATCC® CRL- 2539™). HEK293T Phoenix packaging cells were kindly donated by Dr. Andrea Banfi, University Hospital Basel, Switzerland. LM2, 4T1 , and HEK293T cells were grown in DMEM F-12 high glucose (Gibco, 11330-057) supplemented with 10% heat-inactivated FBS (Gibco, 10500064) and 1% antimycotic/antibiotic (Gibco, 15240-062) in a humidified incubator at 37 °C with 20% 02 and 5% C02. BR16 cells were established as previously described (Alix-Panabieres & Pantel, Nature reviews. Cancer 14, 623-631 (2014)) and grown as suspension cultures in RPMI medium (Gibco, 52400-025) supplemented with 1X B27 (Gibco, 17504-044), 1% antimycotic/antibiotic, 20 ng/ml human recombinant Fibroblast Growth Factor (FGF; Peprotech, 100-18B) and 20 ng/ml human recombinant Epidermal Growth Factor (EGF; Invitrogen, PHG0313) in a humidified incubator at 37 °C with 5% 02 and 5% C02 , using ultra-low attachment plates (Sarstedt, 83.3920.500). LM2, 4T1 and BR16 cells were stably transduced with lentiviral vectors expressing UBC_GFP-T2A-Firefly Luciferase (GFP/Luc) (System Biosciences, BLIV200PA-1-SBI) or ready-to-use virus EF1a_Firefly Luciferase-T2A-mCherry (mCherry/Luc) (Biosettia, GlowCell-15-10).

HIF1a activity reporter

The HIF1a activity reporter was purchased from Genecopoeia upon providing the exact nucleotide sequence. The human hypoxia response element (HRE) of the human VEGFA gene (SEQ ID NO 002: “5’ - CCACAGTGCATACGTGGGCTCCAACAGGTCCTCTT -3”’) (2) is followed by a CMV minimal promoter (CMVmp) (3) and by an enhanced yellow fluorescent protein (eYFP) sequence within a lentiviral vector. Transduced cells were selected with 0.5 pg/ml Puromycin (Invitrogen, ant-pr-1) for five days (LM2) or 15 days (BR16), respectively. Treatment with Deferoxamine (DFO; Sigma, D9533) 500 mM was used to induce the stabilization of HIF1a in LM2 and BR16 cells, for 4 and 15 hours, respectively. Alternatively, HIF1a induction was achieved using the humidified hypoxia chamber (Biospherix, ProOx 110) at 0.1% 02. Anti- HIF1a (Novus, NB100-449) antibodies were used to confirm HIF1a induction through western blot, with anti-GAPDH antibody (Cell Signaling, 2118S) as loading control.

Live imaging HIF1a activity reporter

LM2 and BR16 mCherry/Luc cells expressing the HIF1a activity reporter were seeded into coated (LM2) or uncoated (BR16) imaging chambers (Ibidi, 80826 and 80821), respectively. Following treatment with DFO, cells within chambers were cultured under the humidified live imaging box of the microscope Leica DMi8, at 37°C and stable 20% 02. For live imaging experiments requiring hypoxia, cells within chambers were cultured at 0.1% 02. hHIFIa, hVEGFA and mVegfa knockdown

LM2 and BR16 cells were stably transduced with doxycycline (Dox)-inducible shRNAs, targeting the Open Reading Frame (ORF) of human HIFIct (SEQ ID NO 003: 5’ - AAAGATATGATTGTGTCTC - 3’and SEQ ID NO 004: 5’ - T G CAT CTCG AG ACTTTT CT - 3’), (Dharmacon, TRIPZ®). LM2 and 4T1 cells were stably transduced with Dox-inducible shRNAs targeting ORF of human VEGFA (SEQ ID NO 005: 5’-CAGGGTCTCGATTGGATGG - 3’, SEQ ID NO 006: 5’AGTAGCTGCGCTGATAGAC - 3’), or mouse Vegfa (SEQ ID NO 007: 5’- ACCGCCTTGGCTTGTCACA - 3’, SEQ ID NO 008: 5’ - ACCGCCTTGGCTTGTCACA - 3’) (Dharmacon, SMART®), respectively. The transduced cells were selected using 0.05 - 0.2 pg/ml puromycin and subsequently sorted upon 2 days of 0.1 pg/ml doxycycline treatment (Dox; Sigma, D9891) for the highest expression of the shRNA-coupled fluorophore (TurboGFP or TurboRFP). \r\HIF1ct knockdown was measured by western blot, as described above h VEGFA and m Vegfa knockdown was measured by qPCR using previously described primers (4, 5). h GAPDH (forward primer: SEQ ID NO 009 5' - GAAGGTGAAGGTCGGAGTCAAC - 3', reverse primer: SEQ ID NO 010: 5' - CAG AGTT AAAAG CAGCCCTGGT - 3') or m Gapdh (forward primer: SEQ ID NO 011 : 5' AATGGTGAAGGTCGGTGTG - 3', and reverse primer: SEQ ID NO 012: 5' - GTGGAGTCATACTGGAACATGTAG - 3') were used as load controls. Treatment with DFO 500mM was used to induce the stabilization of HIF1a in LM2, 4T1 and BR16 cells, for 4, 8 and 15 hours, respectively, upon 5 days of treatment with 0.1 pg/ml dox. mVEGFAi64-tCD8a overexpression mVEGFAi6 ₄ -mCD8aTr and mCD8aTr only were introduced in LM2 and 4T1 mCherry/Luc as previously described (Aceto et al., Cell 158, 1110-1122 (2014)). Clonal populations were derived from single cells, obtained through single-cell sorting with BD FACS ARIA in 96-well plates. Successfully growing clones were expanded and analyzed for CD8aTr expression at the CytoFLEX (Beckman Coulter Life sciences, V-B-R series) upon staining with anti-CD8aTr APC (Biolegend, 100712) or isotype control Rat lgG2a (Biolegend, 400511) as previously described (Cheung & Ewald, Science 352, 167-169 (2016)). Clones were further selected based on morphology and stable expression of CD8aTr over multiple in vitro culture passages.

Mouse Experiments

All mouse experiments were carried out in compliance with institutional and cantonal guidelines (approved mouse protocol 2781 , cantonal veterinary office of Basel-City). NOD/scid GAMMA (NSG) mice were purchased from Jackson Laboratory and bred in pathogen-free conditions specified by the University of Basel and cantonal veterinary office of Basel-city. Orthotopic injection was performed between the second and third mammary gland of adult female mice (age range 8-12 weeks) with either 1x10 ⁶ LM2, 1x10 ⁶ BR16 or 0.25x10 ^® 4T1 cells, expressing the fluorescent construct GFP/Luc or mCherry/Luc. Cells were inoculated in 50% Cultrex Path Clear Reduced Growth Factor Basement Membrane Extract (R&D Biosystems, 3533-010-02) and 50% PBS. Mice injected with cells carrying a dox-inducible construct, water containing 0.5 mg/ml Dox (Sigma, D9891-25G) and 5% sucrose (Sigma, S9378) was administered 3 times a week upon tumor formation and for a maximum of 3 months. Injection of 0.02 mg in PBS of recombinant mEphrin-B2-hFC chimera (R&D Biosystem, 496- EB-200) or ChromPure IgG hFC fragment (Jackson Immuno research, 009-000-008) was performed intra-peritoneal (i.p.) and with a frequency of twice per week.

Metastatic index and organ fixation

Mice bearing GFP/Luc or mCherry/Luc tumors were subcutaneously (s.c.) injected with 3 mg DFirefly- Luciferin (Gold Bio, LUCK-5G). After 10 minutes, bioluminescent images of the full mouse were taken at I VIS Lumina LT (Perkin Elmer). After euthanasia and within 20 minutes from the injection of Luciferin, primary tumor and metastatic organs were imaged separately. Metastatic index was calculated as the ratio of the total flux in photons per second per square centimeter per steradian (Ph/s/cm2/sr) of the metastatic organ over the primary tumor. Sample exclusion is applied to metastatic index greater than 1 .3, mostly due to imprecise measurement as a consequence to high primary tumor necrosis. Primary tumors and metastatic organs were fixed in PFA-Lysine-Phosphate buffer (4% PFA, 0.2 M L-Lysine, 0.2% Nal03 and 0.1 M Phosphate buffer, pH 7.4 - 0.2 M NaH2P04 and 0.2 M H2HP04) O/N at 4 °C. Subsequently, organs were incubated in 30% sucrose for 6 hours before O.C.T. embedding.

CTCs capture and quantification

Patient-derived CTCs were enriched on the Parsortix Cell Separation Cassette (GEN3D6.5, ANGLE) within 1 hour of blood draw. In-cassette staining was performed with the antibody cocktail for antihuman EpCAM-AF488 (Cell Signaling, CST5198), anti-human HER2-AF488 (BioLegend, 324410) and anti-human EGFR-FITC (GeneTex, GTX11400). For mouse derived CTCs capture, mice were anaesthetized using isoflurane and blood was drawn from the central circulation through cardiac puncture. Blood was processed immediately on the Parsortix system for CTCs enrichment. For all xenograft models with GFP/Luc, mCherry/Luc or dox-inducible sh-RNA reporters, CTCs were directly quantified in cassette using their fluorescence signal. In-cassette staining for hypoxic status was performed with HypoxiaRed (Enzo Life technologies, ENZ-51042-K500) in the presence of 1% BSA (Sigma, A8412) in PBS. For the HIF1a activity reporter/HypoxiaRed correlation analysis and the in vitro validation of the dye, LM2 or BR16 cells were stained for the HypoxiaRed according to the manufacturer protocol. Quantification of mouse-derived CTCs with HIF1a activity reporter or HypoxiaRed-stained CTCs was achieved by releasing the CTCs from the Parsortix system into a PBS solution and by analyzing the cell suspension through ImageStreamX Mark II (Amnis, Luminex). In particular, all the events between 13-100 pm diameter were analyzed with 40x objective and at slow flow rate for the acquisition of images. The 405, 488, 561 and side scatter (SSC) lasers were used. GFP/Luc or eYFP were acquired on Channel 2 (532/56), mCherry/Luc and HypoxiaDye on channel 4 (628/69). Analysis was performed at the IDEASa software (Luminex, v6.0). Final graphs were created with FlowJo v10.

3D volumes and blood vessels functionality analysis

Mice bearing LM2 and BR16 mCherry/Luc tumors expressing the HIF1a activity reporter were sacrificed at week 5 and month 6 respectively, immediately after intra-peritoneal (i.p.) injection of 1 .2 mg of Pimonidazole (Hypoxyprobe, HP-500mg) and intra-venous (i.v.) injection of 1 mg of Dextran- Biotin 70 kDa (Thermo Fisher, D1957), 1 hour and 15 minutes before sacrifice, respectively. Tissue sections were prepared, stained and imaged as previously described (Cheung et ai., Proceedings of the National Academy of Sciences of the United States of America 113, E854-863 (2016)). Primary tumors were fixed for 24 hours in 4% PFA at 4 °C. Derived tissues were embedded in 4% low-gelling temperature agarose (Sigma, A9414) and subsequently sectioned (50-100 micrometer thick sections) using the Leica VT1200 S vibratome. For the IF staining, all protocol steps were performed at room temperature (RT) with permeabilization for a minimum of 2 hours followed by an O/N incubation with primary antibodies against GFP (Novus Biologicals, NB600-308), Pimonidazole-Red549 (Hypoxyprobe, Red549-Mab), human pan-Cytokeratin (7, 8, 18, 19) (Miltenyi Biotec, 130-112-743), and mouse CD31 (R&D, AF3628). Secondary antibodies against Goat lgG-CF405 (Biotium, 20416), Goat lgG-AF488 (Life technologies, A- 11055), Goat IgG-DyLight 549 (Abeam, ab96933), Rabbit IgG- CF405 (Biotium, 20420), Mouse lgG-AF647 (Life Technologies, A-31571), Human lgG-AF488 (Jackson Immuno Research, 709-545-149), Human lgG-Cy3 (Jackson Immuno Research, 709-165- 149), Streptavidin-AF555 (Life Technologies, S32355), Streptavidin-AF549 (Life Technologies, S32356) were incubated for 2 hours after extensive washings. 3D volumes were constructed using Imaris (Bitplane, v9). Surface rendering was created for all the channels individually (mCherry or hCK, Pimonidazole, eYFP, CD31 , Dextran). Area and volume of the individual surfaces were calculated with the Imaris “Measurement Pro” package. Channels were masked for “voxels out equal to 0” for colocalizing voxels of the respective channels, and with “voxels in equal to 0” for noncolocalizing voxels of the channels. Surface rendering of the masked channels was constructed to further calculate the area or volume of colocalizing channels.

Assessment of metastatic potential of hypoxic and normoxic CTC clusters

CTCs from mice bearing LM2-mCherry/Luc tumors and expressing the HIF1a activity reporter were enriched and later released in a PBS solution, as described above. The CTC suspension was then micromanipulated using CellCelector® (ALS) and a 50 pm glass capillary was used to isolate CTC clusters from the CTC suspension. The total number of cells (forming clusters) was counted and injected through the tail vein of NSG tumor-free female recipients. BR16-mCherry/Luc cells expressing the HIF1a activity reporter were cultured in a humidified hypoxia chamber at 0.1% 02 for four days before sorting. A control dish was cultured in a humidified incubator at 20% 02 for four days before sorting. At day four, cells were collected and sorted at the BD Influx sorter at five pounds per square inch (psi) and with a 200 pm nozzle to preserve the integrity of single CTCs and CTC clusters. Equal numbers of eYFPpositive or eYFP-negative cells (in a cluster form) were injected through the tail vein of NSG tumor-free female recipients. I.v. injected mice were monitored weekly through non-invasive bioluminescence imaging and sacrificed when showing signs of distress.

CTCs single-cell isolation and RNA Sequencing

Single cells or CTC clusters were isolated using CellCelector® based on the color combination of interest and deposited into individual tubes (Corning Axygen®, PCR-02-L-C) containing 2.5 pi RLT Plus lysis buffer (Qiagen, 1053393) and 1 U/pl SUPERase® In RNase Inhibitor (Invitrogen, AM2694). Samples were immediately frozen on dry ice and kept at -80°C until further processing. Following previously published protocol for parallel DNA and RNA sequencing from individual cells (Bos et al., Nature 459, 1005-1009 (2009)), transcriptomes of lysed cells were separated and amplified according to the Smart-Seq2 (Kang et al., Cancer cell 3, 537-549 (2003)). Subsequently, libraries were prepared with Nextera XT (lllumina) and sequenced on NextSeq75 single read for RNA.

Single-cell RNA-seq data processing

Quality assessment of RNA-seq data was performed using FastQC (v0.11 .4) (https://www.bioinformatics.babraham.ac.uk/projects/fastqc), FastQ Screen (vO.11.4) (https://www.bioinformatics.babraham.ac.uk/projects/fastq_sc reen), Kraken (v1.1) and visualized with MultiQC (v0.8). Reads were quality trimmed with Trim Galore! (vO.4.2, https://www.bioinformatics.babraham.ac.uk/projects/trim_galo re/). Trimmed reads derived from xenograft models were aligned to human (GRCh38) and mouse (GRCm38) genomes using STAR (v2.5.2a) and assigned to either the human or mouse using disambiguate (v 1.0.0). Transcript-level expression of transcripts obtained from Ensembl release 89 was quantified using Salmon (v0.11 .3, parameters --seqBias and --gcBias). Gene-level expression was obtained by aggregating transcript- level abundances using tximport. Quality control of processed data was performed with the scater package. Samples with at least 500,000 counts from endogenous genes, 8,000 features detected (threshold > 1 count) and showing less than 50% of counts from the 100 most expressed genes were retained for further analysis. Cell cycle was assigned to each sample using Seurat.

Differential expression

Differential expression (DE) between normoxic and hypoxic CTC clusters was computed with the likelihood ratio test method in the edgeR package (v3.20.1) and using the rounded lengthscaled TPM as input. Genes detected in less than 25% of the samples (threshold 1 TPM) were removed prior to the DE analysis. To define hypoxia, the inventors used a combined criteria defined by HypoxiaRed staining and hypoxia scoring based on gene expression. Hypoxia score was generated independently in each model (NSG-LM2, NSG-BR16 and BR61) by ranking samples according to their mean expression of VEGFA and HIF1A transcripts and calculating the fractional rank normalized between 0 and 1 . Scores above the median were considered as positive. Hypoxic CTC-clusters (n=14) were defined as positive for both hypoxia score and HypoxiaRed. On the contrary, normoxic CTC-clusters (n=17) were defined as negative for both hypoxia score and HypoxiaRed. Samples with discordant results for both criteria were not considered for DE analysis.

Software specification

Data analysis of RNA-seq data after quantification, differential expression and survival analysis was run in R v3.5 and bioconductor v3.8. Data visualization and statistical analyses were performed in GraphPad Prism v7 (GraphPad Software, San Diego, CA), R, ggplot2 and ComplexHeatmap.

Survival analysis using TCGA data

Harmonized gene expression quantification data of Breast Invasive Carcinoma Stage I samples of the Cancer Genome Atlas (TCGA-BRCA) was downloaded from the Genomic Data Commons Data Portal (GDC) using the TCGAbiolinks package. The expression matrix was constructed using the Fragments Per Kilobase of transcript per Million mapped reads normalized using upper quartile (FPKM-UQ) for each sample as obtained with the HTSeq workflow. Clinical data was obtained from the TCGA Pan- Cancer Clinical Data Resource (TCGA-CDR) and overall survival was defined as death from any cause. Hypoxia score (HS) on TCGA-BRCA data was constructed by calculating the mean of the gene-level standardized expression (zscores) across the 25 genes found upregulated in hypoxic CTC clusters and the signatures developed by Buffa (Massague & Obenauf, Nature 529, 298-306 (2016)), Winter (Minn et ai., Nature 436, 518-524 (2005)), Ragnum (Esposito et al., Cold Spring Harb Perspect Med 8 (2018)) and Elvidge (Peinado et ai., Nature reviews. Cancer 17, 302-317 (2017)). HS was then divided by quantiles and the overall survival of patients from Q1 and Q4 was compared using the Kaplan-Meier method using the survival package. The significance between both groups was assessed using the log-rank test. Time-dependent receiver operator curves (ROC) using a predictive time of 10 years were computed using the Nearest Neighbor Estimation (NNE) method implemented in the survivalROC package. IF staining

7pm-thick frozen slices were blocked for 30 minutes in 0.1 % Gelatin buffer (Sigma, G9391) for LM2 and 4T1 , or 10 % Donkey serum buffer (Millipore, S30) for BR16. Primary antibodies for mouse CD31 (R&D, AF3628), Pimonidazole (Hypoxyprobe, Red549-Mab or FITC-Mab), and human pan- Cytokeratin (7, 8, 18, 19) (Miltenyi Biotec, 130-112-743) were incubated O/N at 4 °C. Secondary antibodies against Rabbit lgG-AF647 (Invitrogen, A31573), anti-FITC lgGCF633 (Scientific Laboratory supply, SAB 4600145), and Goat lgG-AF633 (Thermo Fischer, A-21082) were incubated, after washing in PBS, for 1 hour at RT. Slides were mounted with Vectashield Hard set with Dapi (Vectashield, VC-H-1400-L010). Slides were scanned at the Zeiss Axio Imager Z2 with a 20x dry objective. CD31 quantification was performed with Fiji (v2) using the plugin “color pixel counter” of the CD31 over the total tumor background color area (e.g. mCherry or GFP). Pimonidazole quantification was performed with Fiji using the “colocalization threshold” analysis tool of the total tumor background color over the pimonidazole.