Predicting cancer metastasis

Abstract

Methods of predicting the development of a metastasis in a subject suffering from cancer wherein a sentinel lymph node with respect to the cancer does not comprise cancer cells are provided. Method of predicting the absence of development of a metastasis in a subject suffering from cancer which a sentinel lymph node with respect to the cancer does comprise cancer cells are also provided.

Description

PREDICTING CANCER METASTASISCROSS REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of priority of Israeli Patent Application No. 317177, filed November 21, 2024, the contents of which are all incorporated herein by reference in their entirety.FIELD OF INVENTION

[0002] The present invention is in the field of cancer diagnostics.BACKGROUND OF THE INVENTION

[0003] Malignant melanoma ranks as the fifth most prevalent cancer in the Western world, with its frequency steadily increasing. Approximately 85% of melanomas are identified at an early stage amenable to surgical resection, resulting in a favorable prognosis with a five- year survival rate exceeding 98%. However, metastatic melanoma is deadly, responsible for over 80% of fatalities associated with all skin cancers. This discrepancy in prognosis between localized and disseminated disease presents a significant clinical dilemma. Given that primary melanomas are often visible while recurrent metastases are not, one could advocate for treating primary tumors aggressively, as if they were metastatic, utilizing approaches such as immunotherapy. Conversely, such a strategy risks overtreating numerous patients, leading to increased morbidity and mortality.

[0004] The current clinical workflow to establish if the cancer has spread is to focus on the sentinel lymph nodes (sLNs). The sLN is the first lymph node (or group of nodes) connected to a primary tumor via lymphatic vessels, and thus the one to which cancer cells are most likely to spread first. During cancer staging, particularly in cancers like breast cancer and melanoma, the sLN is identified, removed, and examined to determine whether cancer cells are present. If metastases are found in the sLN, the cancer will be classified as stage III and the patient will likely be treated systemically, usually with targeted therapy or immunotherapy. Conversely, negative sLNs confer a stage II tumor, directing patients to surveillance. Although sLN involvement is a helpful prognostic factor and part of the routine guidelines for melanoma staging, 44% of patients with positive sLNs will not experiencedisease relapse. Moreover, 10-25% of patients with negative sLNs will proceed to develop recurrent metastases within three years of diagnosis. As the indications for systemic treatment following local excision are being expanded to include patients with stage II in addition to stage III disease, tools to better identify the subgroups of patients who stand to benefit from these potentially toxic and expensive therapies are needed.

[0005] LNs are not simply waystations for transiting tumor cells but serve as hubs of antitumor immunity. The sLN, as the first responding LN to the tumor, is shaped by tumors, and reciprocally targets the tumor. In the course of the tumor immunity cycle, antigens from the tumor are trafficked to the LNs by dendritic cells (DCs) and are presented to T cells to elicit an anti -tumor immune response. Presentation of tumor antigens by DCs in the LNs has been shown to drive anti-tumor immunity, and the success of immunotherapy has been shown to require activation of immune responses in LNs in addition to the immediate tumor microenvironment. However, recent studies have shown that the presence of metastases in tumor-draining LNs are associated with immunosuppressive cellular niches and impaired anti-tumor responses. Moreover, work in mice has suggested that the presence of metastases in tumor-draining LNs, in addition to being an expression of local immune tolerance, may facilitate recurrent metastatic spread through an inhibitory influence on the systemic immune response. Thus, murine models have shown that LNs may harbor conflicting roles, both suppressing and enabling tumor spreading. Knowledge of LN involvement in the metastatic process in human melanoma is even more limited. A new accurate test for assessing a patient’s risk of developing metastasis when there is no LN involvement is greatly needed.SUMMARY OF THE INVENTION

[0006] The present invention provides methods of predicting the development of a metastasis in a subject suffering from cancer wherein a sentinel lymph node with respect to the cancer does not comprise cancer cells are provided. Method of predicting the absence of development of a metastasis in a subject suffering from cancer which a sentinel lymph node with respect to the cancer does comprise cancer cells are also provided.

[0007] According to a first aspect, there is provided a method of predicting the development of a metastasis in a subject suffering from cancer and wherein a sentinel lymph node (sLN) with respect to the cancer does not comprise cancer cells, the method comprising:a. receiving from the subject a measure of at least one parameter from the sLN, wherein the at least one parameter is selected from: i. abundance of CCR7+ CD4 T cells; ii. abundance of PD-L1+ dendritic cells (DCs); iii. abundance of CCR7+ DCs; iv. abundance of CD4 T regulatory cells (Tregs) in the T cell enriched zone (T-zone) of the sLN; v. abundance of germinal center cells, in the follicle of the sLN; vi. abundance of CD45RA+ CD4 T cells; and vii. abundance of CD206+ macrophages in the sLN; and b. applying a trained machine learning algorithm to the received at least one parameter, wherein the trained machine learning algorithm is trained on a training set comprising the at least one parameter in subjects suffering from cancer with an sLN that did not contain cancer cells and who did not develop a metastasis, and subjects suffering from cancer with an sLN that did not contain cancer cells and who did develop a metastasis, and wherein the trained machine learning algorithm outputs a prediction of metastasis developing in the subject or a prediction of metastasis not developing in the subject; thereby predicting the development of a metastasis in a subject.

[0008] According to another aspect, there is provided a method of predicting the development of a metastasis in a subject suffering from cancer and wherein a sentinel lymph node (sLN) with respect to the cancer does not comprise cancer cells, the method comprising: a. receiving from the subject a measure of at least one parameter from the sLN, wherein the at least one parameter is selected from: i. abundance of CCR7+ CD4 T cells;ii. abundance of CCR7+ DCs; iii. abundance of CD4 T regulatory cells (Tregs) in the T cell enriched zone (T-zone) of the sLN; iv. abundance of germinal center cells, in the follicle of the sLN; v. abundance of CD45RA+ CD4 T cells; and vi. abundance of CD206+ macrophages in the sLN; and b. applying a trained machine learning algorithm to the received at least one parameter, wherein the trained machine learning algorithm is trained on a training set comprising the at least one parameter in subjects suffering from cancer with an sLN that did not contain cancer cells and who did not develop a metastasis, and subjects suffering from cancer with an sLN that did not contain cancer cells and who did develop a metastasis, and wherein the trained machine learning algorithm outputs a prediction of metastasis developing in the subject or a prediction of metastasis not developing in the subject; thereby predicting the development of a metastasis in a subject.

[0009] According to another aspect, there is provided a method of predicting the absence of development of a metastasis in a subject suffering from cancer and wherein a sentinel lymph node (sLN) with respect to the cancer does comprise cancer cells, the method comprising: a. receiving from the subject a measure of at least one parameter from the sLN, wherein the at least one parameter is selected from: i. abundance of PD-L1+ dendritic cells (DCs); ii. abundance of CD45RO+ CD4 T cells; iii. abundance of CD69+ DCs; iv. abundance of Temra GZMB+ CD8 T cells in the T-zone of the sLN; v. abundance of CD45RA+ CD4 T cells; andvi. abundance of CD45RA+ CD4+ central memory T cells (Tcm) in the T-zone; and b. applying a trained machine learning algorithm to the received at least one parameter, wherein the trained machine learning algorithm is trained on a training set comprising the at least one parameter in subjects suffering from cancer with an sLN that did contain cancer cells and who did not develop a metastasis, and subjects suffering from cancer with an sLN that did contain cancer cells and who did develop a metastasis, and wherein the trained machine learning algorithm outputs a prediction of metastasis developing in the subject or a prediction of metastasis not developing in the subject; thereby predicting the absence of development of a metastasis in a subject.

[0010] According to another aspect, there is provided a method of predicting the absence of development of a metastasis in a subject suffering from cancer and wherein a sentinel lymph node (sLN) with respect to the cancer does comprise cancer cells, the method comprising: a. receiving from the subject a measure of at least one parameter from the sLN, wherein the at least one parameter is selected from: i. abundance of CD45RO+ CD4 T cells; ii. abundance of CD69+ DCs; iii. abundance of Temra GZMB+ CD8 T cells in the T-zone of the sLN; iv. abundance of CD45RA+ CD4 T cells; and v. abundance of CD45RA+ CD4+ central memory T cells (Tcm) in the T-zone; and b. applying a trained machine learning algorithm to the received at least one parameter, wherein the trained machine learning algorithm is trained on a training set comprising the at least one parameter in subjects suffering from cancer with an sLN that did contain cancer cells and who did not develop a metastasis, and subjects suffering from cancer with an sLN that did contain cancer cells and who did develop a metastasis, and wherein thetrained machine learning algorithm outputs a prediction of metastasis developing in the subject or a prediction of metastasis not developing in the subject; thereby predicting the absence of development of a metastasis in a subject.[Oi l] According to some embodiments, the germinal center cells express CD20 and CD21.

[0012] According to some embodiments, the CD206+ macrophages define the lining of a sinus in the sLN.

[0013] According to some embodiments, the sinus is the medullary sinus.

[0014] According to some embodiments, the at least one parameter is a plurality of parameters.

[0015] According to some embodiments, the at least one parameter is all seven parameters.

[0016] According to some embodiments, the cancer is a solid cancer.

[0017] According to some embodiments, the solid cancer is melanoma.

[0018] According to some embodiments, the sLN is the lymph node into which the cancer drains.

[0019] According to some embodiments, the method further comprises performing high- resolution spatial multiplexed imaging of a section of the sLN to produce the parameter.

[0020] According to some embodiments, the parameter is determined based on protein expression in the sLN, mRNA expression in the sLN or both.

[0021] According to some embodiments, spatial protein expression in the sLN is determined by performing Multiplexed Ion Beam Imaging by Time of Flight (MIBI-TOF).

[0022] According to some embodiments, spatial mRNA expression in the sLN is determined by performing spatial transcriptomics molecular imaging, optionally wherein the spatial transcriptomics comprises CosMX spatial molecular imaging.

[0023] According to some embodiments, the method further comprises systemic administration of an anticancer therapy to a subject predicted to develop a metastasis.

[0024] According to some embodiments, the anticancer therapy is selected from an immunotherapy and targeted therapy.

[0025] According to some embodiments, the immunotherapy is selected from PD- 1 , PD-L 1 , LAG3, CTLA4 and TIGIT blockade.

[0026] According to some embodiments, the anticancer therapy is an anti-PD-1 or PD-L1 blocking antibody selected from Pembrolizumab, Nivolumab, Durvalumab, Atezolizumab, Retifanlimab, Dostarlimab, Pidilizumab, Cemiplimab and Avelumab or an anti-CTLA4 antibody selected from Ipilimumab and Tremelimumab.

[0027] According to some embodiments, the anticancer therapy is a BRAF inhibitor selected from vemurafenib and dabrafenib, a MEK inhibitor selected from trametinib and cobimetinib or a combination thereof.

[0028] According to some embodiments, the method further comprises surgically removing the cancer without systemic administration of an anticancer therapy to a subject predicted to not develop a metastasis.

[0029] Further embodiments and the full scope of applicability of the present invention will become apparent from the detailed description given hereinafter. However, it should be understood that the detailed description and specific examples, while indicating preferred embodiments of the invention, are given by way of illustration only, since various changes and modifications within the spirit and scope of the invention will become apparent to those skilled in the art from this detailed description.BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figures 1A-1H: Spatial profiling of proteins and mRNAs in Melanoma sentinel lymph nodes (sLNs). (1A) The inventors assembled retrospective sLN specimens from 69 melanoma patients. Patients were divided into four groups based on the metastatic status of their nodes (Positive / Negative) and whether they will develop metastases (Positive / Negative). (IB) sLN tissues underwent spatial profiling of 39 proteins using MIBI- TOF (left) and 960 mRNAs using CosMx (right). (1C) Multiplexed protein images underwent cell segmentation and classification. Shown is a representative image colored by protein expression (left) and cell type classification (right) (ID) Same as (1C) forthe mRNA images. (IE) Mean protein expression (x-axis) for each cell type (y-axis) in the protein data. (IF) Normalized protein expression (y-axis) for each CD8 T cell phenotype (x-axis) (left) and CD4 T cell phenotypes (x-axis) (right). Phenotypes are grouped together to functional states including Naive, T central memory (Tcm), T effector memory (Tem), CD45RA+ T effector memory (Temra), Progenitor exhausted T cells (Tpex), Exhausted T cells (Tex), and T resident cell (Trm). (1G) Normalized mRNA expression (x-axis) for each cell type (y- axis) in the mRNA data. Heatmap displays the top 3 differentially expressed mRNAs foreach cell type. (1H) Normalized composition of macrophages (y-axis) per patient (x-axis), cell types were derived from the protein (top). Normalized composition of CD8 T cells phenotypes (y-axis) per patient (x-axis), cell types were derived from the protein (bottom).

[0031] Figures 2A-2F: Vision transformer identifies micro and macroenvironments in sLNs. (2A) Images of classified cells were fed into DINO, a self-supervised vision transformer. DINO was trained by performing augmentations on images and learning to project them in proximity in the embeddings space. After training on sliding windows, the embeddings space is clustered where each cluster represents a microenvironment. Microenvironments with similar cell type compositions were manually grouped to macroenvironments . (2B) Normalized cell type composition (x-axis) for each microenvironment (y-axis) for the mRNA data. Rows are colored by the macroenvironment (first column) and the microenvironment (second column) . (2C) Same as (2B) for the protein images. (2D) Example image colored by cell type classification (left) and microenvironment classification (right), emphasizing a microenvironment (brown) colocalizing IDO1+ DCs (pink) with Tregs (brown) and CD8 T cell (red). (2E) Example image colored by cell type classification (left) and macroenvironment classification (right). Presenting the LN compartments. (2F) Normalized composition of macroenvironment (y-axis) per patient (x- axis) (top). Normalized composition of the follicle microenvironment (y-axis) per patient (x- axis) (bottom).

[0032] Figures 3A-3M: Metastases remodeling of the lymph node. (3A) Illustration, inside and outside the tumor. (3B) Paired normalized cell type composition (x-axis) per patient (y-axis) inside and outside tumor macroenvironment. Only regions with more than 100 cells were included in the analysis. (3C) Volcano plot of cell types abundant inside versus outside the metastasis region. Only regions with more than 100 cells were included in the analysis. (3D) Example image colored by CD31 (green) and PNAd (pink) expression inside the metastasis, marked by SOX10 (light purple). CD31 only expression represents blood vessels. Overlap of CD31 and PNAd represent HEVs. (3E) Comparison of CD4 and CD8 T cell phenotypes within the metastasis versus the T-zone. The X-axis shows the fraction of cells in each lineage, and the Y-axis denotes the cell phenotype. P-values were calculated across patients. Lines represent the 95% confidence intervals, and dots indicate the mean values across patients. Metastases regions are represented in blue, and the T-zone (outside the metastasis) is in orange. Only regions with more than 100 cells and at least 10 CD4 or CD8 T cells were included in the analysis. (3F) Representative image of Tex CD8 T cells within the tumor. Tumor cells are marked by SOX10 (light purple), the yellow linemarks the boundary between the tumor and T-zone. The overlap of CD8 (red) and PD-1 (green) indicated Tex CD8 T cells. (3G) Correlation between fraction of HLA1+ tumor cells (x-axis) and CD 8 T cell infdtration inside the tumor macroenvironment (y-axis). Only regions with more than 100 cells were included in the analysis. (3H) Representative image of HLA1+ Tumor with CD8 infiltration (left) vs low HLA1+ Tumor without CD8 T cell infiltration (right). (31) Correlation between fraction of HLA1+ tumors (x-axis) and fraction of Tex CD8 T cell infiltration within the tumor macroenvironment (y-axis). Only regions with more than 100 cells were included in the analysis. (3J) Illustration of clean (NX) vs metastases (PX) lymph node. Comparing only the non-metastases regions. (3K) Same as (3E) for comparison of CD4 T cell, CD8 T cell and macrophages phenotypes within the T- zone macroenvironment of NX and PX patients. NX are represented in blue, and PX are in orange. (3L) Fraction of Ki67+ B cells out of all B cells in the follicles of PX and NX, calculated for each patient with a minimum of 10 B cells. Each dot represents an individual patient. PX is shown in orange, and NX in blue. (3M) same as (3L) for HLA 11+ B cells. All p-values were calculated using Mann-Whitney statistic test.

[0033] Figures 4A-4P: Prediction of development of distance metastases in positive sLNs. (4A) Illustration of metastases lymph node; PP had recurrence; PN didn't had recurrence. (4B) Comparison of CD8 T cell phenotypes within the metastasis. The X-axis shows the fraction of cells in each lineage, and the Y-axis denotes the cell phenotype. P- values were calculated across patients. Lines represent the 95% confidence intervals, and dots indicate the mean values across patients. PN regions are represented in blue, and PP in orange. Only regions with more than 100 cells and at least 10 CD8 T cells were included in the analysis. (4C) On the left, fraction of CD45RO+ CD8 T cells out of all CD8 T cell in the tumor of PN vs PP, calculated for each patient with a minimum of 5 CD8 T cells. Each dot represents an individual patient. On the right, example protein images colored by CD45RO / GZMB (green) overlapping with CD8 (red) amongst tumor cells (SOX10; light purple). (4D) Comparison of DCs and Macrophages proteins positivity within the metastasis. The X-axis shows the fraction of positive cells in each lineage, and the Y -axis denotes the proteins. P-values were calculated across patients. Lines represent the 95% confidence intervals, and dots indicate the mean values across patients. PN regions are represented in blue, and PP in orange. Only regions with more than 100 cells and a minimum of 10 DCs / Macrophages cells were included in the analysis. (4E) Fraction of CFD+ macrophages in the metastasis region of PN and PP, calculated for each patient with a minimum of 5 macrophages cells. Each dot represents an individual patient. (4F) Fraction of CD8 T celland T regulatory cells in the follicles. The y-axis shows the fraction of each cell, and the x- axis denotes the cell type. P-values were calculated across patients. Lines represent the 95% confidence intervals, and dots indicate the mean values across patients. PN regions are represented in blue, and PP in orange. Only follicles with more than 100 cells were included in the analysis. (4G) Representative image of follicles in PN (left) and PP (right). FOXP3 (green) is dilated for emphasis. The follicle macroenvironment is outlined, and CD20 expression is shown in light blue, with CD20 appearing blurred. (4H) Same as (4B) in the T-zone macroenvironment. The Tem and Temra CD8 phenotypes were grouped in this analysis. (41) Same as (4D) in the T-zone macroenvironment. (4J) Same as (4D), comparing fraction of proteins positivity of tumor cells within the metastasis of PN and PP patients. PN regions are represented in blue, and PP in orange. Only regions with more than 100 cells and a minimum of 10 tumor cells were included in the analysis. (4K) Representative images capturing the high SOX10 PN tumor (left) and low SOX10 PP tumor (right). (4L) Normalized mRNA expression of tumor cells per patient. The Y-axis lists the patients, and the X-axis displays the top differentially expressed genes between PN and PP. Only regions with over 100 cells and at least 10 tumor cells were included, filtering out dispersed metastases. (4M) Go enrichment analysis of mRNA differentially expressed on PN tumors. X-axis display the -log(p-values), and the Y -axis lists the significant GO terms enriched in PN tumor as opposed to PP tumors. (4N) Illustration summarizing the key findings characterizing PN and PP. (40) Left: PCA of PN and PP patients based on 6 selected features. The X-axis represents the first PCA component, and the Y-axis represents the second PCA component, with each dot representing an individual patient (PN in blue, PP in orange). The separation line was determined using a support vector machine (SVM). Right: PCA loadings for the first component. (4P) ROC curve of a leave-one-out analysis over the patients. All p-values were calculated using Mann-Whitney statistic test.

[0034] Figures 5A-5P: predictive of development of distance metastases in negative sLNs. (5A) Illustration of clean lymph node; NP had recurrence; NN didn't had recurrence (5B) Comparison of CD4 T cells and CD8 T cells phenotypes within the T-zone. The X-axis shows the fraction of cells in each lineage, and the Y-axis denotes the cell phenotype. P- values were calculated across patients. Lines represent the 95% confidence intervals, and dots indicate the mean values across patients. NN are represented in blue, and NP in orange. Only regions with more than 100 cells and at least 10 CD4 / CD8 T cells were included in the analysis. (5C) Left: fraction of CCD7+ CD4 T cells and DCs relative to the CD4 T cell and DCs lineage in the T-zone regions of NN and NP patients. Only regions with more than 100cells and a minimum of 10 CD4 T cells / DCs were included in the analysis. Right, representative images of CCR7+ T cells in the T-zone of NN and NP patient. (5D) Left: fraction of PDL1+ DCs out of all DCs in the T-zone of NN and NP patients. Each dot represents an individual patient. Only regions with more than 100 cells and a minimum of 10 DCs were included in the analysis. Right: Representative image of PDL1+ DCs in the T- zone NP patient, marked by overlap of CD11c (red) and PDL1 (green). (5E) Fraction of the medullary sinus macroenvironment, with each dot representing a patient. Left: Spatial proteomics data. Right: Spatial transcriptomics data. Right panel: Representative images of medullary sinus expansion. Top row: images colored by cell type classification. Bottom row: images colored by macroenvironment regions. (5F) Volcano plot of mRNA expression in lymphatic endothelial cells within the T-zone macroenvironment. Growth factors are marked with red. P-values were calculated across NN and NP patients. The X-axis shows the Log2 fold change, and the Y-axis displays the -loglO(P-value). Dotted lines indicate the P-value threshold of 0.05. (5G) Representative images of proximity between medullary sinus macrophages and plasmablast cells. (5H) Correlation of the macroenvironment in the same field of view (FOV). Medullary sinus macroenvironment and plasmablast macroenvironment marked with yellow boxes. (51) Fraction of Mast cells in the Medullary Sinus regions of NN and NP patients. Each dot represents an individual patient. Only regions with more than 100 cells were included in the analysis. (5 J) Correlation between fraction of Follicular Germinal Center in Follicle (x-axis) and fraction of Medullary Sinus macroenvironment (y-axis). Each dot represents an individual patient. Only follicles with more than 100 cells were included in the analysis (5K) Fraction of IL12A+ DCs among the total DC population within the follicle. (5L) Fraction of TNFSF13B (TACI)+ B cells among the total B cell population within the Plasmablast macroenvironment. (5M) Fraction of CD40+ B cells among the total B cell population within the follicle macroenvironment. (5N) Illustration summarizing the key findings characterizing NN and NP. (50) Left: PCA of NN and NP patients based on 7 selected features. The X-axis represents the first PCA component, and the Y-axis represents the second PCA component, with each dot representing an individual patient (NN in blue, NP in orange). The separation line was determined using a support vector machine (SVM). Right: PCA loadings for the first component. (5P) ROC curve of a leave-one-out analysis over the patients. All p-values were calculated using Mann- Whitney statistic test.

[0035] Figures 6A-6D: (6A) Recurrent-free survival over time for each patient group. (6B) Left: Age distribution (Y -axis) across patient groups (X-axis). Right: Breslow distribution,same as left. (6C) Representative protein expression from several lymph node tissues. Each image is colored by a different protein. Images were processed with 99th percentile value clipping and Gaussian blurring. (6D) Normalized lineage cell type composition (y-axis) for each patient (x-axis).

[0036] Figures 7A-7F: (7A) Fraction of HEVs out of all blood vessels inside and outside the metastases. (7B) Left: image of PDL1 expressed on the border of the tumor. Right: Fraction of PDL1+ cells on the border of the tumor (y-axis). X-axis represents the distance in pixels from the border of the tumor macroenvironment. (7C) Correlation between fraction of HLA1+ tumors (x-axis) and CD8 T cell infiltration inside the tumor macroenvironment (y-axis). Only regions with more than 100 cells were included in the analysis PN patients colored in blue, PP patients colored in orange. (7D) Major axis of follicles in NX and PX in pixels. (7E) Fraction of CD1 lc+ B cells among all B cells in the follicle macroenvironment. Only follicles macroenvironment with more than 100 cells were included in the analysis. (7F) Fraction of PDL1+ DCs among all DCs in the follicle macroenvironment. Only follicles macroenvironment with more than 100 cells were included in the analysis. All p-values were calculated using Mann-Whitney statistic test.

[0037] Figures 8A-8E: (8A) Cell type composition inside the tumor macroenvironment for PN and PP patients. (8B) Fraction of vessels in the follicles of PN and PP. Only follicles macroenvironment with more than 100 cells were included in the analysis. (8C) Fraction of B-T Treg microenvironment in the follicles and B-T zone. Only follicles macroenvironment with more than 100 cells were included in the analysis. (8D) Fraction of patients who were treated before relapse. For NN and PN presented is fraction of patients who were treated after diagnosis. (8E) Baseline ROC curve for leave-one-out analysis across patients, using age and Breslow to predict recurrence in positive lymph nodes. All p-values were calculated using Mann-Whitney statistic test.

[0038] Figures 9A-9F: (9 A) Volcano plot of mRNA expression in DCs within the T-zone macroenvironment. P-values were calculated across NN and NP patients. The X-axis shows the Log2 fold change, and the Y-axis displays the -log 10(P -value). Dotted lines indicate the P-value threshold of 0.01 and 0.05. (9B) Correlation between fraction of blood vessels in the T-zone (x-axis) and the medullary sinus microenvironment (y-axis). Each dot represents a patient. (9C) Correlation between plasmablast macroenvironment and Medullary sinus macroenvironment. NN patients colored with blue, NP patients colored with orange. Each dot represents one field of view (FOV). (9D) Correlation between CD4 Treg percentage out of all CD4 T cells in the T-zone (y-axis) and age (x-axis). Each dot represents a patient. (9E)Correlation between fraction of medullary sinus macroenvironment (x-axis) and age (y- axis). Each dot represents a patient. (9F) Baseline ROC curve for leave-one-out analysis across patients, using age and Breslow to predict recurrence in negative lymph nodes.DETAILED DESCRIPTION OF THE INVENTION

[0039] The present invention, in some embodiments, provides methods of predicting the development of a metastasis in a subject suffering from cancer wherein a sentinel lymph node with respect to the cancer does not comprise cancer cells. Method of predicting the absence of development of a metastasis in a subject suffering from cancer which a sentinel lymph node with respect to the cancer does comprise cancer cells are also provided. Methods of treating a subject predicted to develop or not develop metastasis are also provided.

[0040] It is hypothesized that sLNs may play both a pro-tumorigenic and an anti- tumorigenic role, and that they may assume different roles in different patients and along different stages of the disease. Moreover, it is hypothesized that this role may influence systemic immunity and therefore be predictive for the development of recurrent metastases, thus addressing an urgent clinical need. To this end, the inventors assembled a retrospective cohort of negative and metastatic sLNs from 69 melanoma patients, ensuring that approximately half of the individuals in both groups developed recurrent metastases. Using high-resolution spatial multiplexed imaging, the inventors analyzed 39 proteins with MIBI- TOF and 960 mRNAs with CosMX to create an atlas of cell types and states within sLNs. The inventors developed a deep learning method to map unique microenvironments in these nodes, revealing both known LN structures and novel microenvironments including a colocalization of Tregs, CD8 T cells, and IDO1+ dendritic cells. The invention is based, at least in part, on the surprising finding that although tumor cells colonizing LNs are in immune hubs, they foster a local microenvironment that is enriched with neutrophils, macrophages and exhausted CD8 T cells. In patients with metastatic sLNs, not developing recurrent metastases correlated with increased activation of CD8 T cells in the metastasis, Tregs in the follicles, and tumor-cell apoptosis, potentially indicating a robust immune response. In patients with negative sLNs, two distinct patterns emerged: those who did not develop recurrent metastases exhibited an expanded medullary sinus with colocalizing plasmablasts, while those who developed recurrent metastases had increased Tregs and an influx of suppressive CCR7+ immune cells in the T zone, potentially promoting tolerance. By integrating cell type composition, phenotype, and microenvironmental data, the inventorsachieved an area under the curve (AUC) of 79% and 91% in predicting recurrent metastases for negative and metastatic LNs, respectively. The findings underscore the complex interplay between pro- and anti-tumorigenic immune responses in sLNs, occurring even prior to LN colonization, and highlight their potential for identifying patients at risk for metastatic disease.

[0041] By a first aspect, there is provided a method of predicting the development of a metastasis in a subject, the method comprising: a. receiving from the subject a measure of at least one parameter from a sentinel lymph node; and b. applying a trained machine learning algorithm to the received at least one parameter; thereby predicting the development of a metastasis in a subject.

[0042] In some embodiments, the method is an in vitro method. In some embodiments, the method is an ex vivo method. In some embodiments, the method is an in vivo method. In some embodiments, the method is a diagnostic method. In some embodiments, the method is a prognostic method. In some embodiments, the method is a method of treatment. In some embodiments, the method is a method for determining treatment.

[0043] In some embodiments, the subject in a mammal. In some embodiments, the mammal is human. In some embodiments, the subject suffers from cancer. In some embodiments, the cancer is a solid cancer. In some embodiments, the cancer is a tumor. In some embodiments, the cancer is skin cancer. In some embodiments, the skin cancer is melanoma. In some embodiments, the cancer is selected from cancer is selected from breast cancer, uterine cancer, head and neck cancer, brain cancer, prostate cancer, lung cancer, thyroid cancer, skin cancer, stomach cancer, bladder cancer, urothelial cancer, colon cancer, liver cancer, ovarian cancer, kidney cancer, cervical cancer, bone cancer, connective tissue cancer, esophageal cancer, pancreatic cancer, adrenal cancer, neuroendocrine cancer, rectal cancer, testicular cancer, uveal cancer, and bile duct cancer.

[0044] As used herein, the term “metastasis” refers to the spread of cancer cells from the primary site of origin to other parts of the body and forming new tumors. In some embodiments, metastasis is to a site other than the lymph nodes. In some embodiments, metastasis is to a secondary tissue or site. In some embodiments, the cancer is at a primary site or tissue and the metastasis is to a secondary site or tissue. In some embodiments,metastasis comprises the formation of a secondary tumor. In some embodiments, metastasis is after treatment. In some embodiments, treatment is treatment of the primary cancer. In some embodiments, the metastasis is a distant metastasis.

[0045] In some embodiments, the cancer comprises a sentinel lymph node (sLN) that does not comprises cancer cells. In some embodiments, the subject comprises a sentinel lymph node (sLN) that does not comprises cancer cells. In some embodiments, the cancer comprises a sentinel lymph node (sLN) that does comprises cancer cells. In some embodiments, the subject comprises a sentinel lymph node (sLN) that does comprises cancer cells. As used herein the term “sentinel lymph node” refers to the lymph node (or group of nodes) closest to the cancer. In some embodiments, the sLN is the closest lymph node with respect to the cancer. In some embodiments, the sLN is the lymph node into which the cancer drains. Methods of identifying the sLN are well known and can be performed by a skilled artisan. For example, a tracer (radioactive, dye, etc.) can be injected in or near the primary cancer site and then its spread to the lymphatic system can be monitored. The first node to receive the tracer is the sLN.

[0046] In some embodiments, the method comprises receiving at least one parameter from the sLN. In some embodiments, the method comprises receiving measurements from the sLN. In some embodiments, the method comprises excising the sLN. In some embodiments, the measurement or parameter is from an excised sLN. In some embodiments, excised is removed. In some embodiments, an excised sLN is a biopsy from the sLN. In some embodiments, an excised sLN is a section from the sLN. In some embodiments, an analysis is performed on the sLN. In some embodiments, the analysis ex vivo. In some embodiments, the analysis is an imaging of the sLN. In some embodiments, the imaging is high-resolution spatial multiplexed imaging. In some embodiments, the imagining is of a section of the sLN. In some embodiments, the assay produces the parameter. In some embodiments, the parameter is extracted from the imaging analysis. For example, the imaging may identify various cell types and then the cell type of interest (the one whose abundance is the parameter) can be quantified and the amount, abundance, frequency or percentage of cells that are from the cell type of interest can be determined. In some embodiments, the parameter is determined based on protein expression in the sLN. In some embodiments, the parameter is determined based on mRNA expression in the sLN. In some embodiments, expression is cellular expression. In some embodiments, expression is spatially resolved expression. In some embodiments, spatial protein expression in the sLN is determined by Multiplexed Ion Beam Imaging by Time of Flight (MIBI-TOF). In some embodiments, MIBI-TOF isperformed. In some embodiments, spatial mRNA expression in the sLN is determined by spatial transcriptomics analysis. In some embodiments, the spatial transcriptomics is spatial transcriptomics molecular imaging. In some embodiments, spatial transcriptomics molecular imaging is performed. In some embodiments, the spatial transcriptomics molecular imaging comprises CosMX spatial molecular imaging.

[0047] In some embodiments, the parameter is abundance of C-C chemokine receptor type 7 (CCR7) positive T cells. In some embodiments, abundance is number. In some embodiments, abundance is relative abundance. In some embodiments, abundance is percent of all cells. In some embodiments, all cells is all cells in the sLN. In some embodiments, all cells is all cells measured. In some embodiments, all cells is all T cells. In some embodiments, T cells are CD3 positive cells. In some embodiments, T cells are CD4 positive T cells. In some embodiments, T cells are CD4 T cells. In some embodiments, the parameter is abundance of CCR7+ CD4+ T cells. In some embodiments, abundance of CCR7+ CD4+ T cells is a parameter for predicting development of metastasis in a subject with an sLN that does not comprise cancer cells.

[0048] In some embodiments, the parameter is abundance of Programmed death-ligand 1 (PD-L1) positive dendritic cells (DCs). In some embodiments, the parameter is abundance of CCR7 positive DCs. In some embodiments, abundance is number. In some embodiments, abundance is relative abundance. In some embodiments, abundance is percent of all cells. In some embodiments, all cells is all cells in the sLN. In some embodiments, all cells is all cells measured. In some embodiments, all cells is all DCs. In some embodiments, DCs are CD11c positive cells. In some embodiments, DCs are negative for CD3, CD 14, CD 19, CD56 and CD66b. In some embodiments, DCs are CD 11b positive cells. In some embodiments, DCs are CD 14 positive cells. In some embodiments, DCs are CDlc positive cells. In some embodiments, DCs are CD 141 positive cells. In some embodiments, abundance of PD-L1+ DCs is a parameter for predicting development of metastasis in a subject with an sLN that does not comprise cancer cells. In some embodiments, abundance of CCRD7+ DCs is a parameter for predicting development of metastasis in a subject with an sLN that does not comprise cancer cells.

[0049] In some embodiments, the parameter is abundance of T regulatory cells (Tregs) in a T cell enriched zone (T-zone) of the sLN. In some embodiments, abundance is number. In some embodiments, abundance is relative abundance. In some embodiments, abundance is percent of all cells. In some embodiments, “all cells” is all cells in the sLN. In some embodiments, “all cells” is all cells measured. In some embodiments, “all cells” is all T cells.In some embodiments, abundance is number. In some embodiments, abundance is relative abundance. In some embodiments, all cells is all cells in the sLN. In some embodiments, all cells is all cells measured. In some embodiments, “all cells” is all cells in the T-zone. In some embodiments, Tregs are CD4 positive T cells. In some embodiments, Tregs are CD25 positive. In some embodiments, CD25 positive is CD25 highly positive. In some embodiments, Tregs are FOXP3 positive. In some embodiments, Tregs are CTLA-4 positive. In some embodiments, Tregs are CD25 high and CD 127 low. In some embodiments, Tregs are CD45RA negative Tregs. In some embodiments, Tregs are CD45RO positive. In some embodiments, abundance of CD4 Tregs in the T-zone is a parameter for predicting development of metastasis in a subject with an sLN that does not comprise cancer cells.

[0050] In some embodiments, the parameter is abundance of CD45RA positive T cells. In some embodiments, abundance is number. In some embodiments, abundance is relative abundance. In some embodiments, abundance is percent of all cells. In some embodiments, all cells is all cells in the sLN. In some embodiments, all cells is all cells measured. In some embodiments, all cells is all T cells. In some embodiments, T cells are CD3 positive cells. In some embodiments, T cells are CD4 positive T cells. In some embodiments, T cells are CD4 T cells. In some embodiments, the parameter is abundance of CD45RA+ CD4+ T cells. In some embodiments, CD45RA T cells are naive Tregs. In some embodiments, abundance of CD45RA+ CD4+ T cells is a parameter for predicting development of metastasis in a subject with an sLN that does not comprise cancer cells.

[0051] In some embodiments, the parameter is abundance of germinal center cells in the sLN. In some embodiments, the parameter is abundance of germinal center cells in the follicle of the sLN. In some embodiments, abundance is number. In some embodiments, abundance is relative abundance. In some embodiments, abundance is percent of all cells. In some embodiments, all cells is all cells in the sLN. In some embodiments, all cells is all cells measured. In some embodiments, all cells is all cells in the follicle. In some embodiments, the follicle is the primary follicle. In some embodiments, the follicle is the secondary follicle. In some embodiments, the follicle is the primary and secondary follicles. In some embodiments, the germinal center cells are CD20 positive. In some embodiments, the germinal center cells are CD21 positive. In some embodiments, the germinal center cells are CD20 and CD21 positive. In some embodiments, abundance of germinal center cells in the follicle is a parameter for predicting development of metastasis in a subject with an sLN that does not comprise cancer cells.

[0052] In some embodiments, the parameter is abundance of CD206 positive macrophages. In some embodiments, abundance is number. In some embodiments, abundance is relative abundance. In some embodiments, abundance is percent of all cells. In some embodiments, all cells is all cells in the sLN. In some embodiments, all cells is all cells measured. In some embodiments, all cells is all macrophages. In some embodiments, macrophages are CD68 positive cells. In some embodiments, macrophages are CD 163 positive macrophages. In some embodiments, macrophages are CD 14 positive macrophages. In some embodiments, the parameter is abundance of CD206+ CD4+ macrophages. In some embodiments, CD206+ macrophages define the lining of a sinus in the sLN. In some embodiments, the sinus is the medullary sinus. In some embodiments, abundance of CD206+ macrophages is a parameter for predicting development of metastasis in a subject with an sLN that does not comprise cancer cells.

[0053] In some embodiments, the parameter is abundance of PD-L1 positive DCs. In some embodiments, the parameter is abundance of CD69 positive DCs. In some embodiments, abundance is number. In some embodiments, abundance is relative abundance. In some embodiments, abundance is percent of all cells. In some embodiments, all cells is all cells in the sLN. In some embodiments, all cells is all cells measured. In some embodiments, all cells is all DCs. In some embodiments, DCs are CD11c positive cells. In some embodiments, DCs are negative for CD3, CD14, CD19, CD56 and CD66b. In some embodiments, DCs are CD 11b positive cells. In some embodiments, DCs are CD 14 positive cells. In some embodiments, DCs are CDlc positive cells. In some embodiments, DCs are CD141 positive cells. In some embodiments, abundance of PD-L1+ DCs is a parameter for predicting absence of development of metastasis in a subject with an sLN that does comprise cancer cells. In some embodiments, abundance of CD69+ DCs is a parameter for predicting absence of development of metastasis in a subject with an sLN that does comprise cancer cells.

[0054] In some embodiments, the parameter is abundance of CD45RO positive T cells. In some embodiments, abundance is number. In some embodiments, abundance is relative abundance. In some embodiments, abundance is percent of all cells. In some embodiments, all cells is all cells in the sLN. In some embodiments, all cells is all cells measured. In some embodiments, all cells is all T cells. In some embodiments, T cells are CD3 positive cells. In some embodiments, T cells are CD4 positive T cells. In some embodiments, T cells are CD4 T cells. In some embodiments, the parameter is abundance of CD45RO+ CD4+ T cells. In some embodiments, CD45RO+ T cells are memory Tregs. In some embodiments,abundance of CD45RO+ T cells is a parameter for predicting absence of development of metastasis in a subject with an sLN that does comprise cancer cells.

[0055] In some embodiments, the parameter is abundance of CD45RA positive T cells. In some embodiments, abundance is number. In some embodiments, abundance is relative abundance. In some embodiments, abundance is percent of all cells. In some embodiments, all cells is all cells in the sLN. In some embodiments, all cells is all cells measured. In some embodiments, all cells is all T cells. In some embodiments, T cells are CD3 positive cells. In some embodiments, T cells are CD4 positive T cells. In some embodiments, T cells are CD4 T cells. In some embodiments, the T cells are central memory T cells (Tcm). In some embodiments, Tcm cells are CD45RO T cells. In some embodiments, Tcm cells are CCR7 positive cells. In some embodiments, Tcm cells are CD62L+, CD27+, CD28+, CD127+, CD95+ or a combination thereof cells. Each possibility represents a separate embodiment of the invention. In some embodiments, Tcm cells are CD45RO, CCR7, and CD62L positive cells. In some embodiments, the parameter is abundance of CD45RA+ CD4+ T cells. In some embodiments, CD45RA T cells are naive Tregs. In some embodiments, the parameter is abundance of CD45RA+ CD4+ Tcm cells. In some embodiments, abundance of CD45RA+ CD4+ T cells is a parameter for predicting absence of development of metastasis in a subject with an sLN that does comprise cancer cells. In some embodiments, abundance of CD45RA+ CD4+ Tcm cells is a parameter for predicting absence of development of metastasis in a subject with an sLN that does comprise cancer cells.

[0056] In some embodiments, the parameter is abundance of effector memory T cells (Temra) in a T cell enriched zone (T-zone) of the sLN. In some embodiments, abundance is number. In some embodiments, abundance is relative abundance. In some embodiments, abundance is percent of all cells. In some embodiments, all cells is all cells in the sLN. In some embodiments, all cells is all cells measured. In some embodiments, all cells is all T cells. In some embodiments, T cells are T memory cells. In some embodiments, T cells are CD8+ T cells. In some embodiments, abundance is number. In some embodiments, abundance is relative abundance. In some embodiments, abundance is percent of all cells. In some embodiments, all cells is all cells in the T-zone. In some embodiments, Temra are CD8 positive T cells. In some embodiments, Temra are CD45RA positive. In some embodiments, Temra are CD45RO negative. In some embodiments, Temra are negative for CCR7, CD62L, CD27, CD28 or a combination thereof. Each possibility represents a separate embodiment of the invention. In some embodiments, Tregs are CD57 positive. In some embodiments, the Temra cells are granzyme B (GZMB) positive Temra cells . In some embodiments, the Temracells are CD8 and GZMB positive. In some embodiments, abundance of GZMB+, CD8+ Temra cells in the T-zone is a parameter for predicting absence of development of metastasis in a subject with an sLN that does comprise cancer cells.

[0057] In some embodiments, at least one parameter is a plurality of parameters. In some embodiments, at least one parameter is at least 2 parameters. In some embodiments, at least one parameter is at least 3 parameters. In some embodiments, at least one parameter is at least 4 parameters. In some embodiments, at least one parameter is at least 5 parameters. In some embodiments, at least one parameter is at least 6 parameters. In some embodiments, at least one parameter is at least 7 parameters. In some embodiments, at least one parameter is all the parameters.

[0058] In some embodiments, the at least one parameter is all of the following parameters: i. abundance of CCR7+ CD4 T cells; ii. abundance of PD-L1+ dendritic cells (DCs); iii. abundance of CCR7+ DCs; iv. abundance of CD4 T regulatory cells (Tregs) in the T cell enriched zone (T-zone) of said sLN; v. abundance of germinal center cells, in the follicle of said sLN; vi. abundance of CD45RA+ CD4 T cells; and vii. abundance of CD206+ macrophages in said sLN.

[0059] In some embodiments, the at least one parameter is all of the following parameters: i. abundance of PD-L1+ dendritic cells (DCs); ii. abundance of CD45RO+ CD4 T cells; iii. abundance of CD69+ DCs; iv. abundance of Temra GZMB+ CD8 T cells in the T-zone of said sLN; v. abundance of CD45RA+ CD4 T cells; andvi. abundance of CD45RA+ CD4+ central memory T cells (Tcm) in the T-zone.

[0060] In some embodiments, a trained machine learning (ML) algorithm is applied. In some embodiments, the ML algorithm outputs a prediction. In some embodiments, the prediction is of metastasis developing in the subject. In some embodiments, the prediction is of metastasis not developing in the subject. In some embodiments, the prediction is likelihood of metastasis developing or not developing. In some embodiments, the ML algorithm outputs a likelihood or percent chance of metastasis developing or not developing. In some embodiments, developing is in the subject. In some embodiments, the ML algorithm outputs a metastasis score. In some embodiments, a score above a predetermined threshold indicates a metastasis is likely to occur. In some embodiments, a score below a predetermined threshold indicates a metastasis is unlikely to occur. In some embodiments, unlikely to occur is absent. In some embodiments, the predetermined threshold is the score in control subjects. In some embodiments, the control subjects are subjects with a sLN without cancer cells and who did not develop metastasis. In some embodiments, the control subjects are subjects with a sLN with cancer cells and who did not develop metastasis.

[0061] In some embodiments, the ML algorithm is trained on a training set. In some embodiments, the training set comprises the at least one parameter in subjects suffering from cancer with an sLN that did not contain cancer cells and who did not develop metastasis and subjects suffering from cancer with an sLN that did not contain cancer cells and who did develop metastasis. In some embodiments, the training set further comprises labels indicating whether the at least one parameter was from a subject that developed metastasis or did not develop metastasis. In some embodiments, the training set comprises the at least one parameter in subjects suffering from cancer with an sLN that did contain cancer cells and who did not develop metastasis and subjects suffering from cancer with an sLN that did contain cancer cells and who did develop metastasis. In some embodiments, the training set further comprises labels indicating whether the at least one parameter was from a subject that developed metastasis or did not develop metastasis.

[0062] By another aspect there is provided, a computer program product comprising a non- transitory computer-readable storage medium having program code embodied thereon, the program code executable by at least one hardware processor to perform a method of the invention.

[0063] In some embodiments, any suitable machine learning algorithm or combination of methods may be employed, including, but not limited to:• Support Vector Machine (SVM): A nonparametric model which finds the optimal separating hyperplane that discriminates between different classes. It can perform linear or non-linear classification.• Penalized Logistic Regression (PLR) - a logistic model for regression that imposes a penalty to reduce the impact of certain features.• Generalized linear model (GLM): a generalization of linear regression that unifies statistical models such as linear regression, logistic regression and Poisson regression. GLM extends linear regression by (1) supporting response variables with error distributions other than the normal distribution (2) a non-linear relationship between the predictors and the response variable.• Random forest (RF): involved in the generation of multiple decision trees that consist of sequences of decision rules for protein expression values. To avoid over-fitting, these trees may be pruned. Each tree is constructed by randomly selecting different samples.• extreme Gradient Boosting (XGB): a gradient boosted decision trees-based classification and regression algorithm. The decision trees are built one at a time, and each new tree corrects the error of the previously trained decision tree.

[0064] The present invention may be a system, a method, and / or a computer program product. The computer program product may include a computer readable storage medium (or media) having computer readable program instructions thereon for causing a processor to carry out aspects of the present invention.

[0065] The computer readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device. The computer readable storage medium may be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. A non- exhaustive list of more specific examples of the computer readable storage medium includes the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a static random access memory (SRAM), a portable compact disc read-onlymemory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device having instructions recorded thereon, and any suitable combination of the foregoing. A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire . Rather, the computer readable storage medium is a non-transient (i.e., not-volatile) medium.

[0066] Computer readable program instructions described herein can be downloaded to respective computing / processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and / or a wireless network. The network may comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and / or edge servers. A network adapter card or network interface in each computing / processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing / processing device.

[0067] Computer readable program instructions for carrying out operations of the present invention may be assembler instructions, instruction-set-architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, or either source code or object code written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like, and conventional procedural programming languages, such as the "C" programming language, R, Python or other programming languages. The computer readable program instructions may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet Service Provider). In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate arrays (FPGA), or programmable logic arrays (PLA) may execute the computer readable program instructionsby utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present invention.

[0068] These computer readable program instructions may be provided to a processor of a general-purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks. These computer readable program instructions may also be stored in a computer readable storage medium that can direct a computer, a programmable data processing apparatus, and / or other devices to function in a particular manner, such that the computer readable storage medium having instructions stored therein comprises an article of manufacture including instructions which implement aspects of the function / act specified in the flowchart and / or block diagram block or blocks.

[0069] The computer readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process, such that the instructions which execute on the computer, other programmable apparatus, or other device implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0070] In some embodiments, the method further comprises administering an anticancer therapy to the subject. In some embodiments, the method further comprises administering an anticancer therapy to a subject predicted to develop a metastasis. In some embodiments, the administering is systemic administering. In some embodiments, a systemic anticancer therapy is administered to a subject predicted to develop a metastasis. In some embodiments, the anticancer therapy is an immunotherapy. In some embodiments, the anticancer therapy is a targeted therapy.

[0071] As used herein, the terms “administering,” “administration,” and like terms refer to any method which, in sound medical practice, delivers a composition containing an active agent to a subject in such a manner as to provide a therapeutic effect. One aspect of the present subject matter provides for oral administration of a therapeutically effective amount of a composition of the present subject matter to a patient in need thereof. Other suitable routes of administration can include parenteral, subcutaneous, intravenous, intramuscular, or intraperitoneal.

[0072] The dosage administered will be dependent upon the age, health, and weight of the recipient, kind of concurrent treatment, if any, frequency of treatment, and the nature of the effect desired.

[0073] In some embodiments, the immunotherapy is blockade of an immune checkpoint protein. In some embodiments, the immunotherapy is a blocking antibody that binds to an immune checkpoint protein. In some embodiments, binds to is specifically binds to. In some embodiments, the antibody blocks binding of the immune checkpoint protein to its ligand. In some embodiments, the antibody blocks signaling by the immune checkpoint protein. In some embodiments, the immune checkpoint protein is selected from Programmed cell death protein 1 (PD-1), PD-L1, Lymphocyte-activation gene 3 (LAG3), Cytotoxic T-lymphocyte associated protein 4 (CTLA4) and T cell immunoreceptor with Ig and ITIM domains (TIGIT). In some embodiments, the immune checkpoint protein is selected from PD-1, PD- Ll, LAG3 and TIGIT.

[0074] In some embodiments, the anticancer therapy is an anti-PD-1 or PD-L1 blocking antibody. In some embodiment, the anti-PD-1 or PD-L1 blocking antibody is selected from Pembrolizumab, Nivolumab, Durvalumab, Atezolizumab, Retifanlimab, Dostarlimab, Pidilizumab, Cemiplimab and Avelumab. In some embodiments, the anticancer therapy is an anti-PD-1 antibody. In some embodiment, the anti-PD-1 blocking antibody is selected from Pembrolizumab, Nivolumab, Retifanlimab, Dostarlimab, Pidilizumab, and Cemiplimab. In some embodiment, the anti-PD-Ll blocking antibody is selected from Durvalumab, Atezolizumab and Avelumab. In some embodiments, the anticancer therapy is an anti-CTLA4 antibody selected from Ipilimumab and Tremelimumab.

[0075] In some embodiments, the anticancer therapy is a BRAF inhibitor. In some embodiments, the BRAF inhibitor is selected from vemurafenib and dabrafenib. In some embodiments, the anticancer therapy is a mitogen-activated protein kinase kinase (MEK) inhibitor. In some embodiments, the MEK inhibitor is selected from trametinib and cobimetinib. In some embodiments, the anticancer therapy is a combination of therapies.

[0076] In some embodiments, the anticancer therapy is chemotherapy. In some embodiments, the anticancer therapy is radiation therapy.

[0077] In some embodiments, the method further comprises surgically removing the cancer in a subject predicted not to develop a metastasis. In some embodiments, surgically removing the cancer is tumor resection. In some embodiments, the surgery is without systemic treatment. In some embodiments, the surgery is without systemic administration of ananticancer therapy. It will thus be understood that a subject predicted to develop metastasis is given a treatment that is systemic so as to kill any cancer cells that have moved away from the main cancer, but a subject that is not predicted to develop metastasis, even one with cancer cells in the sLN, can have only surgery to directly remove the cancer. This is greatly beneficial as the multitude of side effects associated with systemic anticancer therapy can be avoided.

[0078] As used herein, the term "about" when combined with a value refers to plus and minus 10% of the reference value. For example, a length of about 1000 nanometers (run) refers to a length of 1000 nm+- 100 run.

[0079] It is noted that as used herein and in the appended claims, the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. Thus, for example, reference to "a polynucleotide" includes a plurality of such polynucleotides and reference to "the polypeptide" includes reference to one or more polypeptides and equivalents thereof known to those skilled in the art, and so forth. It is further noted that the claims may be drafted to exclude any optional element. As such, this statement is intended to serve as antecedent basis for use of such exclusive terminology as "solely," "only" and the like in connection with the recitation of claim elements, or use of a "negative" limitation.

[0080] In those instances where a convention analogous to "at least one of A, B, and C, etc." is used, in general such a construction is intended in the sense one having skill in the art would understand the convention (e.g., "a system having at least one of A, B, and C" would include but not be limited to systems that have A alone, B alone, C alone, A and B together, A and C together, B and C together, and / or A, B, and C together, etc.). It will be further understood by those within the art that virtually any disjunctive word and / or phrase presenting two or more alternative terms, whether in the description, claims, or drawings, should be understood to contemplate the possibilities of including one of the terms, either of the terms, or both terms. For example, the phrase "A or B" will be understood to include the possibilities of "A" or "B" or "A and B."

[0081] It is appreciated that certain features of the invention, which are, for clarity, described in the context of separate embodiments, may also be provided in combination in a single embodiment. Conversely, various features of the invention, which are, for brevity, described in the context of a single embodiment, may also be provided separately or in any suitable sub-combination. All combinations of the embodiments pertaining to the invention are specifically embraced by the present invention and are disclosed herein just as if each andevery combination was individually and explicitly disclosed. In addition, all subcombinations of the various embodiments and elements thereof are also specifically embraced by the present invention and are disclosed herein just as if each and every such sub-combination was individually and explicitly disclosed herein.

[0082] As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” include plural referents, unless the context clearly dictates otherwise. The terms “a” (or “an”) as well as the terms “one or more” and “at least one” can be used interchangeably.

[0083] Furthermore, “and / or” is to be taken as specific disclosure of each of the two specified features or components with or without the other. Thus, the term “and / or” as used in a phrase such as “A and / or B” is intended to include A and B, A or B, A (alone), and B (alone). Likewise, the term “and / or” as used in a phrase such as “A, B, and / or C” is intended to include A, B, and C; A, B, or C; A or B; A or C; B or C; A and B; A and C; B and C; A (alone); B (alone); and C (alone).

[0084] Wherever embodiments are described with the language “comprising,” otherwise analogous embodiments described in terms of “consisting of’ and / or “consisting essentially of’ are included.

[0085] Additional objects, advantages, and novel features of the present invention will become apparent to one ordinarily skilled in the art upon examination of the following examples, which are not intended to be limiting. Additionally, each of the various embodiments and aspects of the present invention as delineated hereinabove and as claimed in the claims section below finds experimental support in the following examples.

[0086] Various embodiments and aspects of the present invention as delineated hereinabove and as claimed in the claims section below find experimental support in the following examples.EXAMPLES

[0087] Generally, the nomenclature used herein and the laboratory procedures utilized in the present invention include molecular, biochemical, microbiological and recombinant DNA techniques. Such techniques are thoroughly explained in the literature. See, for example, "Molecular Cloning: A laboratory Manual" Sambrook et al., (1989); "Current Protocols in Molecular Biology" Volumes I-III Ausubel, R. M., ed. (1994); Ausubel et al., "CurrentProtocols in Molecular Biology", John Wiley and Sons, Baltimore, Maryland (1989); Perbal, "A Practical Guide to Molecular Cloning", John Wiley & Sons, New York (1988); Watson et al., "Recombinant DNA", Scientific American Books, New York; Birren et al. (eds) "Genome Analysis: A Laboratory Manual Series", Vols. 1-4, Cold Spring Harbor Laboratory Press, New York (1998); methodologies as set forth in U.S. Pat. Nos. 4,666,828; 4,683,202; 4,801,531; 5,192,659 and 5,272,057; "Cell Biology: A Laboratory Handbook", Volumes I- III Cellis, J. E., ed. (1994); "Culture of Animal Cells - A Manual of Basic Technique" by Freshney, Wiley-Liss, N. Y. (1994), Third Edition; "Current Protocols in Immunology" Volumes I-III Coligan J. E., ed. (1994); Stites et al. (eds), "Basic and Clinical Immunology" (8th Edition), Appleton & Lange, Norwalk, CT (1994); Mishell and Shiigi (eds), "Strategies for Protein Purification and Characterization - A Laboratory Course Manual" CSHL Press (1996); all of which are incorporated by reference. Other general references are provided throughout this document.Example 1: Spatial profiling of proteins and mRNAs in Melanoma sLNs

[0088] To construct a spatial atlas of sLNs in melanoma, the inventors assembled a retrospective cohort of 69 melanoma patients who underwent a sLN biopsy as part of their routine clinical diagnosis of the primary tumor. Of these, 39 had negative LNs whereas 30 had metastases in their LNs. Patients were selected for the study such that in both groups roughly half of the patients developed recurrent metastases within a follow up of at least five years (Fig. 1A). The inventors named the patient groups using a two-letter code: the first describes whether or not metastases were evident in their sLNs at diagnosis (P-positive / N- negative; P / N), and the second whether the patient proceeded to develop recurrent metastases (P / N). As expected, evidence of metastases in the LN were correlated with progression-free survival (Fig. 6A), as were the tumor’s Breslow score and the patient’s age (Fig. 6B).

[0089] To study the immune composition and organization in the sLNs using multiplexed proteomics imaging, the inventors developed a panel of 39 antibodies (Fig. IB, left). The panel was designed to identify melanoma cells, distinct immune and stromal cell populations as well as various immune cell states. The specificity and sensitivity of all antibodies was validated on control tissues (Fig. 6C). The inventors applied this antibody panel to stain formalin fixed paraffin embedded (FFPE) tissue specimens from the cohort and imaged them using MIBI-TOF. The inventors acquired 2-3 fields of view (FOVs) per patient, resulting in 177 high-dimensional images, each depicting the spatial expression of 39 markers (38 proteins and dsDNA) in situ (Fig. 1C). The inventors used Mesmer to segment individual cells in the images and CellSighter to classify the cells to 26 cell types (Fig. IE). Asexpected, the majority of the cells within the LN are lymphocytes, and only patients from the PP and PN groups had tumor cells, as opposed to patients from the NP and NN groups (Fig. 6D). The inventors found several types of B cells composing the follicle and the germinal center, CD4 and CD8 T cells composing the T-zone accompanied by DCs and macrophages, and CD206+ CD209+ macrophages, mostly lining the LN sinuses, specifically the medullary sinus.

[0090] Next, the inventors clustered the T cell populations based on expression of designated functional proteins including TCF1-7, PD-1, CD45RO and more. The inventors distinguished distinct CD8 T cell states, including naive cells which have not yet encountered specific antigen (CD45RA+); central memory T cells (Tcm: CD45RO+, CCR7+), which are antigen-experienced long-lived memory cells that reside in lymphoid tissues; effector memory T cells (Tem: CD45RO+) and resident memory cells (Trm: CD45RO+, CD103+), which are usually present in the circulation and non-lymphoid peripheral tissues; exhausted T cells (Tex: PD-1+ / TOX+ / TIM3+ / LAG3+), which have been continuously exposed to antigen and precursor-exhausted T cells (Tpex: PD-1 low TCF1+ TOX+), which retain high proliferative capacities, and co-express exhaustion proteins along with self-renewal proteins such as TCF-1. The inventors observed similar phenotypes for CD4 T cells, including precursor exhausted CD4 T cells and potentially cytotoxic GZMB+ CD4 T cells (Fig. IF).

[0091] In addition to the protein images, for 33 patients the inventors also performed spatial measurements of mRNA using Nanostring’s CosMx platform, imaging the expression of 960 genes (Fig. IB, right). The inventors used CellPose to segment individual cells and InSituType to classify the cells into 20 cell types and states (Fig. ID, 1G). The data acquired using both modalities yielded an overall similar composition and spatial organization for the major cell lineages, including T cells, B cells and tumor cells (Fig. IE, 1G). The technologies were complimentary, with the protein data providing higher confidence in the phenotype of any particular cell and the mRNA data adding additional cell types, not included in the protein panel including plasmablasts, mast cells and plasmacytoid DCs (pDCs). Moreover, the mRNA data exposed a more comprehensive phenotypic landscape within the macrophage and DC populations. For instance, while both the protein and mRNA data identified CD206+ CD209+ macrophages, the mRNA data further revealed that they also express MARCO, CXCL2 and CXCL5, (Fig. IE, 1G). Comparison of the cell type composition among the different groups in the cohort revealed distinct differences in macrophage and T cell phenotypes, with higher abundance of sinus macrophages (CD206+CD209+ macrophages) (Fig. 1H, top) in NN patients and increased CD8 T central memory in NP patients. (Fig. 1H, bottom).

[0092] To identify microenvironments in the sLNs, the inventors developed a novel self- supervised-based approach to cluster patches in the images according to their cellular composition and protein expression. Briefly, images depicting the outputs of the cell classification were fed into DINO, a self-supervised vision transformer, which clusters together images with similar semantic features (Fig. 2A). The model passes two different random transformations of an input image to a student and teacher network. It is trained to position these transformations in proximity in the final embedding space, which the inventors then clustered using PhenoGraph to produce the final classifications.

[0093] The inventors identified 20 microenvironments in the mRNA data and 28 microenvironments in the protein data (Fig. 2B-2C). The inventors annotated each microenvironment according to its main cell types. Although not all cell types were identical between the mRNA and protein data, the inventors could identify recurring motifs. For example, the medullary sinus (MS) was enriched for CD206+CD209+ macrophages in the protein data and CD206+ CD209+ MARCO+ CXCL2+ macrophages in the mRNA data. Both data types also identified a microenvironment enriched for interactions of DCs and Tregs, defined by colocalization of LAMP3+ DCs with Tregs in the mRNA data and an equivalent microenvironment co-localizing IDO1+ DCs and Tregs in the protein data (Fig. 2D) The inventors then manually grouped microenvironments into macroenvironments describing known microanatomical structures of the LN such as the B cell follicles, interfollicular T cell zones (T-zone) and the medullary sinus (MS) (Fig. 2E). The inventors found that the micro-environments associated with B cell follicles depicted different layers of the germinal center. For example, the microenvironment in the core of the germinal center was enriched with follicular germinal cells while the outer layer of the germinal center was enriched for naive B cells. The border of the follicle was delineated by a B-T zone (Fig. 2E).

[0094] While the different layers of the germinal center have been previously described, much less is known regarding the spatial organization of the T-zone in the lymph node. The inventors found that the T-zone was also composed of different micro-environments characterized by local enrichments of CD4 T cells, CD8 T cells, Tregs, DCs and APCs. Comparing the microenvironment and macroenvironment composition between the different groups revealed a higher fraction of the medullary sinus macroenvironment in NN patients (Fig. 2F, top). In the follicles, the inventors found an enrichment of the germinal center microenvironment in NP patients compared to NN, and an increased abundance of T regenriched B-T zone in PN patients compared to PP patients (Fig. 2F, bottom). Overall, the multiplexed proteomics and transcriptomics datasets allowed us to extract features on cell types, cell states and cell niches in sLNs.Example 2: Tumor seeding associates with immune remodeling in the lymph node

[0095] To understand the microenvironment of melanoma metastases in the sLN, the inventors first focused on patients who had positive sLNs (PN and PP, subsequentially referred to as PX). The inventors analyzed the metastatic regions and compared the composition and phenotype of non-tumor cells inside and outside of the metastases (Fig. 3A). This analysis revealed a unique intra-metastasis microenvironment, associated with a preferential immune infiltration of macrophages, DCs, neutrophils and CD8 T cells (Fig. 3B-C). This immune composition was very different than the LN composition outside the metastases, which was enriched for CD4 T cells and B cells. The inventors also observed an increase of stromal cells inside the metastasis and replacement of the high endothelial venules (HEVs) characteristic of the lymph node parenchyma with metastasis-associated blood vessels (Fig. 3C-D, 7A). These results suggest that even when metastases are seeded in immune hubs, they drive a distinct local immune microenvironment.

[0096] Next, the inventors focused on the phenotypes of CD8 and CD4 T cells inside and outside the metastasis regions. For each patient the inventors calculated the fraction of each distinct immune phenotype out of the cells of the same type. For example, the inventors calculated the fraction of exhausted CD8 T cells out of the CD8 T cells, either inside or outside the metastases’ microenvironments. Across patients, the inventors observed more effector, exhausted and T regs out of all the CD4 T cells inside the metastases (Fig. 3E). Similarly, for CD8 T cells the inventors observed more resident and exhausted phenotypes inside the metastasis. This difference diminishes when the inventors compare the precursor exhausted (Tpex) population, suggesting that the progression of exhaustion occurs in the context of the metastasis. Interestingly, for some patients the inventors observe a sharp boundary, in which T cells within the metastasis express exhaustion proteins, such as PD-1, whereas T cells outside do not (Fig. 3F). This behavior is accompanied by PD-L1 expression in the metastasis microenvironment (Fig. 7B). Together, these observations suggest that the exhausted T cells remain within the metastasis and there may be little efflux of T cells from the metastasis microenvironment to the metastasis-adjacent regions.

[0097] The analysis of CD8 T cell infiltration into the metastases revealed variability across patients, spanning over a 14-fold difference in abundance (Fig. 3B, 3G). To explore potentialmechanisms underlying this variance, the inventors compared the degree of CD8 T cell infiltration to the phenotypes of tumor cells. The inventors found that expression of HLA-1 on the tumor cells was positively correlated with CD8 T cell infiltration into the metastasis (R=0.59 p=0.005, Fig. 3G-3H). Furthermore, the inventors found that HLA-1 expression on tumor cells was correlated with an exhausted phenotype of the infiltrated CD 8 T cells (R=0.58 p=0.006, Fig. 31). These results suggest that continuous stimulation by antigens in the metastasis microenvironment drives T cell exhaustion, and that downregulation of HLA-I by tumor cells is an effective mechanism to evade predation by CD8 T cells, as previously suggested. While downregulation of HLA-1 was previously suggested as a mechanism for acquired resistance following treatment, here, in pre-treatment LNs, HLA-1 expression by tumor cells did not correlate with recurrence (Fig. 7C).

[0098] The results thus far have shown that metastases remodel their immediate surroundings. Next, the inventors evaluated whether the presence of metastases in the LN correlates with immune remodeling outside the metastases. To this end, the inventors compared the non-metastatic regions of the sLN between patients who had metastases in their LN (noted by PX) to patients who did not have metastases in their LNs (noted by NX) (Fig. 3J). When comparing the distribution of phenotypes of cells in the T-zone, the inventors observed an increase in the populations of resident T cells and exhausted T cells in patients with metastases in their LNs (Fig. 3K), indicating higher activation of CD4 and CD 8 T cells in metastatic lymph nodes. Comparing the phenotypes of macrophages revealed a more naive-like population of monocytes expressing HLA-DR-DP-DQ (Mono CD 14 DR) in the negative LNs in comparison with CD 163+ M2 -like macrophages in metastatic LNs (Fig. 3K).

[0099] Next, the inventors compared the follicle macroenvironment. The inventors observed an increase of Ki-67 on B cells in the follicle (Fig. 3L), suggesting more activated germinal centers in LNs that harbor metastases. Accordingly, the inventors found that metastatic sLNs had slightly larger follicles compared to negative sLNs (Fig. 7D, ns). The inventors also found altered phenotypes in B cells in metastatic LNs, which had a higher expression of classII antigen presentation (Fig. 3M), and a higher fraction of B cells expressing CD11c (median 7.5% vs. 5%, Fig. 7E). This subset may represent atypical B cells, previously connected to chronic infection, systemic autoimmunity and memory formation. In contrast, negative LNs had more suppressive PD-L1+ DCs (Fig. 7F).

[0100] Overall, metastatic LNs showed signs of immune remodeling, most prominently within the metastases. Within the metastases, the inventors observed a drastically alteredmicroenvironment enriched with macrophages and exhausted T cells. Comparing the non- metastatic regions revealed increases in T cell exhaustion and B cell proliferation in metastatic lymph nodes, suggesting that the LN is globally shaped by the presence of the metastases.Example 3: Immune organization in patients with LN metastases is predictive of development of distance metastases

[0101] To assess if immune responses in the LN associate with metastatic recurrence, the inventors focused on the patients who had metastatic sLNs and compared the immune landscape between the patients who proceeded to either develop recurrent metastases (PP) or not (PN) (Fig. 4A). While the cell lineage composition was similar in PN and PP patients (Fig. 8A), analysis of the phenotypes of the CD8 T cells in the metastatic regions revealed that in PN patients they are characterized by a more stem-like memory profde co-expressing TCF, CD45RO and granzyme B (Fig. 4B-4C). Accordingly, the metastatic regions of PN patients had more CD103+ (p=0.015) and TCF+ DCs (p=0.014), which could serve as local drivers of T cell activation (Fig. 4D). In contrast, the inventors found more M2-like macrophages expressing CD209 (p=0.047) and CD 163 (p=0.07) in the metastatic regions of PP patients, previously shown to have an immunosuppressive effect (Fig. 4D). Analysis of the mRNA data revealed higher expression of Complement factor D (CFD) in macrophages in the metastatic regions of PP patients (Fig. 4E), indicating alternative activation of the complement system, and previously shown to correlate with poor prognosis. Altogether, the inventors observed activation of T cells, possibly by local DCs, in the metastatic regions of PN patients, in contrast to potentially suppressive macrophages in PP patients.

[0102] Subsequently, the inventors analyzed the cellular composition outside the metastatic regions to identify systemic differences in the sLNs of PN and PP patients. Comparing the follicle regions, the inventors identified an increased vasculature in PP patients (Fig. 8B) and an increase in Tregs and CD8 T cells in the follicles of PN patients (Fig. 4F-4G). It was also manifested by an increase in the microenvironment that colocalized Tregs, CD8 T cells and IDO1 DCs in the B-T cell border of the follicle of PN patients (Fig. 8C). Next, the inventors compared the T-zones, and in accordance with the results in the metastases, the inventors observed that PN patients had more effector memory CD8 T cells (p=0.021, Fig. 4H). Next, the inventors analyzed DCs and macrophages and found high PD-L1 and CD69 expression in PN patients (Fig. 41). Those PD-L1 expressing cells might participate in driving the exhaustion of the CD8 T cells. Overall, analysis of immune phenotypes both inthe metastatic region and the non-colonized LN compartments, suggests distinct states in patients that eventually developed recurrent metastases and those that remained metastasis free. The inventors hypothesize that some tumors do not trigger a strong T cell response, whereas others do, but the response is impaired by T regulatory cells and T cell exhaustion. These dynamics are then predictive of whether the patient will further develop metastases or not, where the former are more likely to metastasize.

[0103] Next, the inventors explored properties of the tumor cells, to associate them with the observed immune organization and development of metastases. In the protein data, PN metastases had increased expression of CD56 (p=0.014, Fig. 4J), which indicates neuroendocrine differentiation. They also exhibited reduced expression of HLA-I, potentially as a mechanism of acquired resistance to predation by T cells (p=0.02, Fig. 4J). Interestingly, a subset of PP patients showed reduced expression of SOX 10, a transcription factor driving neural crest differentiation, supporting previous results that loss of SOXIO associates with a more invasive phenotype (Fig. 4J-4K). Analysis of the mRNA data indicated higher expression of APOD in tumor cells in some PP patients compared with PN patients, which is known to be associated with invasive melanoma (Fig. 4L). Gene Ontology (GO) enrichment analysis identified increased activation of cell-cycle arrest and apoptosis in PN patients (Fig. 4L-4M). Together, these results suggest that the metastases of PN patients may be more apoptotic, either driving immune activation, or mediated by immune activation, whereas the metastases of PP patients exhibit a less differentiated and more invasive phenotype, and associate with potentially suppressive macrophages, overall driving weaker lymphocyte activation.

[0104] The analysis suggested several features in metastatic sLNs that differ between patients that proceed to develop recurrent metastases and those who do not. The inventors therefore examined the predictive power of applying multiplexed imaging to predict development of recurrent metastases in these patients. Currently, patients with metastases in the sLN are treated based on clinical considerations as well as a pathological report, taking into consideration parameters such as the Breslow thickness of the primary tumor, the size of the metastases in the LN and other relevant clinical factors. To examine potential imagingbased biomarkers for predicting whether a patient will develop recurrent metastases, the inventors selected six features that were independently predictive of metastatic recurrence, and could be obtained using low-plex imaging, and preformed principal component analysis (PCA) to extract correlated components that describe the data (Fig. 40). The inventors then repeated this pipeline, each time leaving out one of the patients. In each iteration theinventors performed principal component analysis and trained a Support Vector Machine (SVM) classifier with the first two PCs as features. The inventors proceeded to test the classifier on the held-out patient. Overall, the approach resulted in an AUC of 91% (Fig. 4P), compared with an AUC of 55% when using age and Breslow thickness (Fig. 8E). In the cohort, 50% of the PP patients were not treated with systemic therapy, such as immune checkpoint inhibitors (Fig. 8D). Using the classifier, 76% of them would be classified as likely to develop recurrent metastases. While the size of the cohort is not sufficient to draw extensive conclusions, the results suggest that multiplexed imaging of the sUNs could be used to identify patients with a high chance of developing recurrent metastases, who could potentially benefit from systemic therapy.Example 4: Immune organization in patients with negative sLNs is predictive of development of metastases

[0105] Next, the inventors focused on patients with negative lymph nodes, who present a challenging treatment dilemma. The inventors compared the patients who proceeded to either develop recurrent metastases (NP) or not (NN) (Fig. 5A). The inventors analyzed the composition of microenvironments as well as the composition of cell types and phenotypes in each macroenvironment. The inventors found that overall NP patients were characterized by a T-cell mediated immune response in the UN compared to NN patients. NN patients displayed more naive T cell phenotypes, for both CD4 and CD8 T cells (Fig. 5B). In contrast, NP patients had more central memory CD8 and CD4 T cell phenotypes (Fig. 5B), overall suggesting the presence of antigen-experienced circulating immune cells in the UN. Accordingly, the inventors observed an increase in CCR7 expression in the T-zone for both CD4 T cells and DCs (Fig. 5C), suggesting that they entered the UN recently. In addition, LNs from NP patients had several indications for immunosuppression, such as more T regulatory cells (Tregs), which have been shown to have a suppressive effect on the immune response in the dUN48 (Fig. 5B), and more DCs expressing PD-LI, known to induce a suppressive microenvironment, in the T-zone (p=0.056, Fig. 5D). Collectively, these results could suggest that suppressive DCs from the primary tumor are seeding the UN and inducing a tolerogenic immune environment in the T zone (Fig. 5N). In contrast, the mRNA data revealed that DCs in NN exhibited higher levels of CUEC10A and CEEC12A (Fig. 9A), resembling the expression profile of DCs of subset 2, which are more associate with CD4 T cell response.

[0106] In contrast to NP patients, NN patients had an increase in the macroenvironment of the medullary sinus (Fig. 5E), also manifested by a global increase in CD206+ macrophages (Fig. 1H). This expansion was highly correlated with an increase in the abundance of blood vessels in the T zone (R=0.79 p<0.001, Fig. 9B). To verify this observation, the inventors turned to the mRNA data, which showed increased expression of the vascular growth factor VEGF on the lymphatic endothelial cells in the T-zone in NN patients (Fig. 5F). Together, these results suggest that the sLNs of NN patients are preferentially undergoing an expansion of the Medullary sinus and lymphangiogenesis, which have been shown to follow an immune response to inflammation and accompany the egress of lymphocytes from the LN. Accordingly, in the mRNA images, the inventors observed a high correlation between the expansion of the Medullary sinus and a microenvironment that was enriched with plasmablasts (R=0.75, p<0.01, Fig. 5G-5H, 9C). These plasmablasts were located in close proximity to the Medullary sinus (Fig. 5G-5H), possibly indicating their imminent exit from the LN. The Medullary sinus in NN patients was also enriched for mast cells (Fig. 51), which have previously been associated with good prognosis, in both primary tumors and LN metastases. Taken together, these findings indicate an immune response in NN patients that involves an expansion of the Medullary sinus and likely the egress of plasmablasts from the LN.

[0107] The spatial organization of the LNs in NN patients was reminiscent of an extrafollicular response (EFR), in which plasmablasts are produced from the extrafollicular zone with or without a limited germinal center response, usually by T cell independent activation of B cells. In support of this hypothesis, the inventors found a negative correlation between the expansion of the medullary sinus and the abundance of follicular germinal center cells in the follicle (R=-0.35, p=0.038, Fig. 5J). The higher abundance of mast cells in the medulla sinus of NN patients (Fig. 51) further supports this hypothesis as mast cells have been demonstrated to enhance the differentiation of naive B cells into plasmablasts. To further explore potential mechanisms that could drive an extrafollicular response, the inventors turned to the mRNA data. The inventors found that DCs in the follicles of NN patients expressed more IL-12 (Fig. 5K), which was shown to suppress GC formation, but not extrafollicular responses, and that NN patients had a higher expression of TNFSF13B (TACI) in the plasmablast zone (p=0.13, Fig. 5L), which was shown to regulate a T cell independent response of extrafollicular B cells. Conversely, the inventors observed an increased presence of CD40+ B cells in the follicle of NN patients (Fig. 5M), which couldsuggest that that there is T cell-dependent stimulation of B cells in the follicles of NN patients.

[0108] To conclude, the inventors suggest that the lymph nodes of NP patients exhibit increased influx of DCs and T cells, whereas NN patients show expansion of the medullary sinus and may be enriched for extrafollicular responses (Fig. 5N). The inventors note that in accordance with previous work in this cohort the average age of NP patients (62.6 y / o) was higher than the average age of NN patients (53.8 y / o) (Fig. 6A, left). As such, age may be associated with some of the differences between NP and NN. For example, previous studies have shown an increase in the accumulation of Tregs in the LNs with age 60, and the inventors also found a correlation between age and accumulation of Tregs in the LN (R=0.38, p=0.018, Fig. 9D). In contrast, expansion ofthe medullary sinus was not correlated with patient age (Fig. 9E), suggesting it as an age-independent differentiating factor between NN and NP patients.

[0109] Finally, the inventors examined whether the spatial changes between NN and NP patients could predict whether a patient with negative sLNs will develop recurrent metastases. For this analysis the inventors focused on the MIBI data, for which the inventors had more samples. The inventors selected 7 features that were independently predictive of metastatic recurrence, and could be obtained using low-plex imaging, and preformed PCA to extract correlated components that describe the data (Fig. 50). The first component is mostly comprised of CD4 T central memory cells and the expansion of the medullary sinus (MS). The inventors then repeated this pipeline, each time leaving out one of the patients. In each iteration the inventors performed feature selection, PCA and trained a Support Vector Machine (SVM) classifier. The inventors then tested the classifier on the held-out patient. Overall, the approach resulted in an area under the curve (AUC) of 79% (Fig. 5P) compared with an AUC of 74% when using Age and Breslow depth (Fig. 9F). Under standard of care, patients with negative LNs are generally not treated with systemic therapy, such as immune checkpoint inhibitors, but recent clinical trials have indicated that stage Ilb / c melanoma patients could benefit from immunotherapy. The results suggest that in-depth analysis of the sLNs could be used to identify patients with a high chance of developing recurrent metastases, who could potentially benefit from systemic therapy.

[0110] Although the invention has been described in conjunction with specific embodiments thereof, it is evident that many alternatives, modifications and variations will be apparent to those skilled in the art. Accordingly, it is intended to embrace all such alternatives,modifications and variations that fall within the spirit and broad scope of the appended claims.

Claims

CLAIMS:

1. A method of predicting the development of a metastasis in a subject suffering from cancer and wherein a sentinel lymph node (sLN) with respect to said cancer does not comprise cancer cells, the method comprising: a. receiving from said subject a measure of at least one parameter from said sLN, wherein said at least one parameter is selected from: i. abundance of CCR7+ CD4 T cells; ii. abundance of PD-L1+ dendritic cells (DCs); iii. abundance of CCR7+ DCs; iv. abundance of CD4 T regulatory cells (Tregs) in the T cell enriched zone (T-zone) of said sLN; v. abundance of germinal center cells, in the follicle of said sLN; vi. abundance of CD45RA+ CD4 T cells; and vii. abundance of CD206+ macrophages in said sLN; and b. applying a trained machine learning algorithm to said received at least one parameter, wherein said trained machine learning algorithm is trained on a training set comprising said at least one parameter in subjects suffering from cancer with an sLN that did not contain cancer cells and who did not develop a metastasis, and subjects suffering from cancer with an sLN that did not contain cancer cells and who did develop a metastasis, and wherein said trained machine learning algorithm outputs a prediction of metastasis developing in said subject or a prediction of metastasis not developing in said subject; thereby predicting the development of a metastasis in a subject.

2. The method of claim 1, wherein said germinal center cells express CD20 and CD21.

3. The method of claim 1 or 2, wherein said CD206+ macrophages define the lining of a sinus in the sLN.

4. The method of claim 3, wherein said sinus is the medullary sinus.

5. A method of predicting the absence of development of a metastasis in a subject suffering from cancer and wherein a sentinel lymph node (sLN) with respect to said cancer comprises cancer cells, the method comprising: a. receiving from said subject a measure of at least one parameter from said sLN, wherein said at least one parameter is selected from: i. abundance of Temra GZMB+ CD8 T cells in the T-zone of said sLN; ii. abundance of PD-L1+ dendritic cells (DCs); iii. abundance of CD45RO+ CD4 T cells; iv. abundance of CD69+ DCs; v. abundance of CD45RA+ CD4 T cells; and vi. abundance of CD45RA+ CD4+ central memory T cells (Tcm) in the T- zone; and b. applying a trained machine learning algorithm to said received at least one parameter, wherein said trained machine learning algorithm is trained on a training set comprising said at least one parameter in subjects suffering from cancer with an sLN that did contain cancer cells and who did not develop a metastasis, and subjects suffering from cancer with an sLN that did contain cancer cells and who did develop a metastasis, and wherein said trained machine learning algorithm outputs a prediction of metastasis developing in said subject or a prediction of metastasis not developing in said subject; thereby predicting the absence of development of a metastasis in a subject.

6. The method of any one of claims 1 to 5, wherein said at least one parameter is a plurality of parameters.

7. The method of any one of claims 1 to 6, wherein said at least one parameter is all the parameters.

8. The method of any one of claims 1 to 7, wherein said cancer is a solid cancer.

9. The method of claim 8, wherein said solid cancer is melanoma.

10. The method of any one of claims 1 to 9, wherein said sLN is the lymph node into which said cancer drains.

11. The method of any one of claims 1 to 10, further comprising performing high- resolution spatial multiplexed imaging of a section of said sLN to produce said parameter.

12. The method of any one of claims 1 to 11, wherein said parameter is determined based on protein expression in said sLN, mRNA expression in said sLN or both.

13. The method of claim 11 or 12, wherein spatial protein expression in said sLN is determined by performing Multiplexed Ion Beam Imaging by Time of Flight (MIBI- TOF).

14. The method of claim 11 or 12, wherein spatial mRNA expression in said sLN is determined by performing spatial transcriptomics molecular imaging, optionally wherein said spatial transcriptomics comprises CosMX spatial molecular imaging.

15. The method of any one of claims 1 to 14, further comprising systemic administration of an anticancer therapy to a subject predicted to develop a metastasis.

16. The method of claim 15, wherein said anticancer therapy is selected from an immunotherapy and targeted therapy.

17. The method of claim 16, wherein said immunotherapy is selected from PD-1, PD-L1, LAG3, CTLA4 and TIGIT blockade.

18. The method of claim 16 or 17, wherein said anticancer therapy is an anti-PD-1 or PD- L1 blocking antibody selected from Pembrolizumab, Nivolumab, Durvalumab, Atezolizumab, Retifanlimab, Dostarlimab, Pidilizumab, Cemiplimab and Avelumab or an anti-CTLA4 antibody selected from Ipilimumab and Tremelimumab.

19. The method of claim 16, wherein said anticancer therapy is a BRAF inhibitor selected from vemurafenib and dabrafenib, a MEK inhibitor selected from trametinib and cobimetinib or a combination thereof.

0. The method of any one of claims 1 to 19, further comprising surgically removing said cancer without systemic administration of an anticancer therapy to a subject predicted to not develop a metastasis.

Images