Predictive markers for epithelial cancers

Abstract

The present invention relates to a method of a predicting clinical outcome for a subject diagnosed with epithelial cancer, such as oral squamous cell carcinoma, as well as to biomarkers for use in the method.

Description

[0001] PREDICTIVE MARKERS FOR EPITHELIAL CANCERS FIELD OF THE INVENTION The present invention belongs to the field of biomedicine. More specifically, the invention relates to biomarkers for use in a predicting clinical outcome for a subject diagnosed with epithelial cancer, such as squamous cell carcinoma, especially oral squamous cell carcinoma. BACKGROUND OF THE INVENTION Epithelial tissues (epithelia) are composed of closely apposed epithelial cells attached to an under- lying non-cellular basement membrane. Epithelial cells can be squamous, cuboidal or columnar in shape and they can form one-cell thick simple epithelia or stratified epithelia with multiple cell layers. Epithelia cover external body surfaces (e.g., skin, alveoli of lungs and digestive tract), line internal cavities (e.g peritoneal cavity) and form key components of many internal organs (e.g. kidney, bladder, mammary and other glands, uterus). Epithelia function as selective barriers be- tween the exterior and underlying tissue compartments in an organ-specific manner. In addition to the protec- tive and structural functions, the epithelial cells of- ten also have secretory and / or absorptive functions. The majority of surface epithelial cells are continuously renewing and exposed to environmental carcinogens, which increases their risk of accumulating cancer-driving mu- tations over time. Epithelial cancer (carcinoma) is the most common form of human solid malignant tumors, ac- counting for 80-90% of all cancer cases. Squamous cell carcinoma is a common type of epithelial cancer, orig- inating from squamous epithelial cells found, for exam- ple, on the surface of the skin and in the oral cavity. Cuboidal and columnar epithelial cells can also undergo malignant transformation and form other common epithe- lial cancers (such as adenocarcinomas of lung, colon, kidney and breast). Oral squamous cell carcinoma (OSCC) is the most common form of head and neck cancer, presenting a sig- nificant health concern worldwide. Recent global reports have estimated 377,713 new cases and 177,757 deaths in the year 2020. Among the subsites of OSCC, squamous cell carcinoma of the mobile tongue (OTSCC) stands out as the most prevalent, making it the most common type of cancer within the oral cavity. Notably, OTSCC also demonstrates a rising incidence in young adults and carries the worst prognosis among OSCC cases, further underscoring its clinical significance. Even with early-stage OTSCC (cT1- cT2), close to 20% of the patients will eventually face relapse and cancer-related death. Currently, there are no effective methods to recognize these high-risk pa- tients. Novel prognostic biomarkers are urgently needed to guide appropriate treatment strategies for these pa- tients. SUMMARY The invention relates to a method of predicting clinical outcome for a subject diagnosed with epithelial cancer as defined in independent claim 1 and to a kit for use in said method as defined in independent claim 14. Some embodiments of the invention are set forth in the dependent claims. Further embodiments and aspects become apparent from the detailed description, the fig- ures and the examples. BRIEF DESCRIPTION OF THE DRAWINGS The accompanying drawings, which are included to provide a further understanding of the invention and constitute a part of this specification, illustrate em- bodiments of the invention and together with the de- scription help to explain the principles of the inven- tion. In the drawings: Figure 1 is a schematic flow chart summarizing the study protocol which led to the present invention. Figure 2 illustrates PROX1, KI-67 and E-cad- herin expression in a representative OSCC tumor. Images a-c show IMC expression data of E-cadherin, KI-67 and PROX1 respectively. Dotted squares in images a-c demon- strate the cropped area magnified in images d-f. Cell segmentation masks of Prox1-positive cells are overlayed on images d-f. Arrows in e-f indicate KI-67 and PROX1 double positive cells. Figure 3, LEC proliferation: survival analysis and biomarker testing with ROC. Plots a-c show Kaplan- Meier curves of recurrence-free (a), disease-specific (b) and overall (c) survival in two patient groups di- vided by median (3.4%) LEC proliferation value. Patients with total LEC count below 15 are excluded. Low group represents patients with LEC proliferation value below median. Percentages displayed in the plots a-c show mean survival probabilities in 3-year follow-up time. Plots d-f show time-dependent receiver operating char- acteristic (ROC) estimation from the 3-year recurrence- free (d), disease-specific (e) and overall (f) survival data. Area under the ROC curve (AUC) values are shown in the plots. Figure 4, PROX1 and KI-67: Recurrence-free sur- vival analysis and ROC assessment. Plots a-b show Kaplan-Meier curves of recurrence-free survival in low and high patient groups divided by the median value of LEC count (a) and Total proliferation value (b). LEC count value represents the total amount of PROX1-posi- tive cells in a sample. Total proliferation value rep- resents the percentage of KI-67 positive cells in a sample. Plots c-d show time-dependent receiver operating characteristic (ROC) estimation from the 3-year recur- rence survival data for LEC count (Prox1) and Total proliferation (KI-67) respectively. Area under the ROC curve (AUC) values are shown in the plots. Figure 5A shows representative IMC images of Podoplanin (Pdpn), Prox1, and Ki-67 expression in OSCC tumors of patients A-C. Scale bar: 100 µm. Figure 5B shows zoomed-in regions (images i- iii) of the Figure 5A images. Dotted lines highlight representative Ki-67 positive LECs. Scale bar: 33 µm. Figure 6 shows representative images of Pdpn, Prox1, and Ki-67 expression in lymphatic vessels near OSCC tumors of patients D-F. Dotted lines indicate Pdpn (2nd column) or Prox1 (4th column) signal outlines. Scale bars: 20 µm for each row. Figure 7 shows scatterplots of Podoplanin vs. Prox1 expression in a random sample of 70,000 segmented cells, divided by cell identity. Vertical and horizontal dotted lines indicate Podoplanin and Prox1 positivity cutoffs, calculated as mean - 1SD in LECs. Figure 8 shows Kaplan-Meier survival plots of LEC proliferation (LECp %) using Podoplanin and Ki-67 (left column) or Prox1 and Ki-67 (right column). Cutoff value for Ki-67 positivity is 2. Figure 9A shows a diagram in which the patient cohort described in Example 1 is classified into dif- ferent groups on the basis of their combined risk scores described in Example 3. Figure 9B shows Kaplan-Meier survival plots for the patient groups of Figure 9A. Figure 9C shows Kaplan-Meier survival plots for the patient groups having a combined risk score below or above 2. Figure 10, LEC proliferation: Survival analy- sis by lymph node status (pN). Kaplan-Meier plots a-b show recurrence free survival in patients with (a) one or more metastatic regional lymph node (pN > 0) or (b) no evidence of cancer in regional lymph nodes (pN = 0). Patients are divided into low and high lec proliferation groups based on the median value (3.4%). Patients with total LEC count below 15 are excluded. Low group repre- sents patients with LEC proliferation value below me- dian. DEFINITIONS Before the invention is described, it is to be understood that this disclosure is not strictly limited to any particular compositions, reagents, antibodies, devices, protocols or methodology described herein, as such may vary. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only, and is not intended to be limiting, since the scope of the present invention will be limited only by the appended claims. It is also to be noted that, unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It is further to be noted that certain features of the disclosure, which are, for clarity, described in the context of separate embodiments, can also be pro- vided in combination in a single embodiment. Conversely, various features of the disclosure, which are, for brev- ity, described in the context of a single embodiment, can also be provided separately or in any suitable sub- combination. Moreover, any features, details or embod- iments disclosed in the context of a method provided herein apply also to a kit provided herein, and vice versa, even if not repeated. Herein, the meaning of a singular noun includes that of a plural noun and thus a singular term, unless otherwise specified, may also carry the meaning of its plural form. In other words, the term “a”, “an” or “the” may mean one or more. The term “and / or” in a phase such as “X and / or Y” shall be understood to mean either “X and Y” or “X or Y” and shall be taken to provide explicit support for both meanings or for either meaning. As used herein, the term "carcinoma” refers to cancer that forms in epithelial tissue, and is therefore used interchangeably with the term “epithelial cancer”. Different types of carcinomas include squamous cell car- cinoma, adenocarcinoma, transitional cell carcinoma, and basal cell carcinoma. Epithelial tissue is found throughout the body, forming the covering and lining of all body surfaces and internal passageways. Most cancers affecting the skin, breast, kidney, liver, lung, pancreas, gastrointestinal tract, prostate gland, oral cavity, pharynx and larynx are carcinomas. Indeed, carcinomas account for 80 to 90 percent of all cancer cases. As used herein, the term “oral squamous cell carcinoma (OSCC)” refers to a malignant tumor that may occur anywhere within the oral cavity, including malig- nant tumours in the cheek, tongue, base of the tongue, gums, floor of the mouth, and hard palate. Among the OSCC locations, squamous cell carcinoma of the oral tongue (OTSCC) is the most common type of cancer in the oral cavity. As used herein, the term "predicting" refers to determining the likelihood that a patient will have a particular clinical outcome, whether positive or neg- ative, following surgical removal of the primary tumor. The predictive methods of the present invention can be used clinically to make treatment decisions by choosing the most appropriate treatment modalities for any par- ticular patient, including non-treatment. The predic- tive methods of the present invention are valuable tools in predicting if a patient is likely to benefit from a treatment modality, such as chemotherapy and / or radia- tion therapy. As used herein, the term "positive clinical outcome" means an improvement in any measure of patient status, including those measures ordinarily used in the art, such as an increase in the duration of Recurrence- Free Survival (RFS), an increase in the time of Disease- Specific Survival (DSS), an increase in the time of Overall Survival (OS), and the like. An increase in the likelihood of positive clinical outcome corresponds to a decrease in the likelihood of cancer recurrence. A decrease in the likelihood of positive clinical outcome corresponds to an increase in the likelihood of cancer recurrence. Herein, the term “positive clinical outcome” is interchangeable with the term “good prognosis”. In other words, “positive clinical outcome” indicates that the subject will have a survival time that will be longer than the median (or mean) observed in the general population of subjects suffering from the same disease. Inversely, the expression "negative clin- ical outcome" indicates that the subject will have a survival time that will be shorter than the median (or mean) observed in the general population of subjects suffering from the same disease. Herein, the term “neg- ative clinical outcome” is interchangeable with the term “poor prognosis”. As used herein, the term "recurrence" refers to a return of the cancer, either locally (e.g., where it used to be before therapy) or distally (e.g., metas- tasis). As used herein, the term "Recurrence-Free Sur- vival (RFS)" refers to the length of time from date of curative surgery to the time of cancer recurrence or death from any cause. As used herein, the term “Disease-Specific Sur- vival DSS)” refers to the length of time from date of curative surgery to the time of cancer recurrence or death from a cause other than the cancer in question. As used herein, the term “Overall Survival (OS)” refers to the length of time from date of curative surgery to death from any cause. As used herein, the term "risk classification" refers to the level of risk or the prediction that a subject will experience a particular clinical outcome. A subject may be classified into a risk group based on the predictive methods of the present invention. A "risk group" is a group of subjects or individuals with a similar level of risk for a particular clinical outcome. As used herein, the term “predetermined refer- ence value" refers to a threshold value or a cut-off value established based upon comparative measurements between cancer patients who have positive clinical out- come and cancer patients who have negative clinical out- come. For example, the predetermined reference value may refer to a LEC proliferation value below which it has been found that the subject is likely to have positive clinical outcome and / or above which the subject is un- likely to have positive clinical outcome. In some in- stances, the predetermined reference value may also re- fer to the expression level of a selected marker below which it has been found that the subject is likely to have positive clinical outcome and / or above which the subject is unlikely to have positive clinical outcome, or vice versa. Statistical methods for determining ap- propriate reference values will be readily apparent to those of ordinary skill in the art. As used herein, the term “LEC proliferation value” refers to the percentage of proliferating lym- phatic endothelial cells (LECs) among all lymphatic en- dothelial cells in a region of interest (ROI). Generally, the threshold value has to be de- termined in order to obtain the optimal sensitivity and specificity according to the function of the test and the benefit / risk balance (clinical consequences of false positive and false negative). Typically, the optimal sensitivity and specificity (and so the threshold value) can be determined using a Receiver Operating Character- istic (ROC) curve based on experimental data. For exam- ple, after determining the LEC proliferation value and / or the expression level of a selected marker in a group of reference, one can use algorithmic analysis for the statistic treatment of the proliferation value and / or the expression levels determined in samples to be tested, and thus obtain a classification standard having significance for sample classification, for ex- ample into “high risk” and “low risk” groups. As used herein, the term "expression level” refers to a relative quantity of the biomarker in ques- tion in a cell or tissue sample. As used herein, the terms “subject”, “patient”, and “individual” are used interchangeably, unless oth- erwise indicated, and they refer to a mammal, particu- larly to a human. As used herein, the term “sample” refers to a tissue sample obtained from a subject. Specific examples of tissue samples include, but are not limited to, solid tissue samples, such as fresh, frozen and / or preserved organ or tissue samples, and biopsy specimens, as well as sections or smears prepared from any of these sources. Preferably, the tissue sample results from the resected tumor and encompasses the center of the tumor (or core), and the tissue directly surrounding the tumor (“invasive border”). Generally, obtaining the biologi- cal sample to be analyzed from the subject is not part of the method of the invention. The term "sample" also includes samples that have been manipulated or treated in any appropriate way after their procurement, including but not limited to washing, reagent treatment, fixation (e.g., formalin fixation), or freezing or embedding in semi-solid or solid matrices for sectioning purposes. For the purposes of the present disclosure, a "section" of a tissue sam- ple means a single part or piece of a tissue sample, for example, a thin slice of tissue or cells cut from a tissue sample. As used herein, the term “invasive border” re- fers to a region on each side of the border between malignant cells and normal host tissue. The term is interchangeable with the term “invasive margin”. As used herein, the terms “biomarker” and “marker” are interchangeable and refer to a naturally occurring molecule which is an objective, quantifiable indicator of a particular characteristic of its source. For example, a biomarker for predicting clin- ical outcome for a subject diagnosed with epithelial cancer refers to a molecule which is differentially pre- sent in a biological sample taken from a subject with a certain cancer, such as OSCC, as compared to a compara- ble sample taken from a control subject, such as a sub- ject without said cancer (i.e., an apparently healthy subject) or a subject with said cancer but having a different probable course of or recovery from said can- cer. Thus, the present biomarkers provide information regarding a probable outcome of said cancer (e.g. OSCC) or recovery from said cancer (e.g. OSCC) and associate either quantitatively or qualitatively with the progno- sis of said cancer (e.g. OSCC). Accordingly, the term “lymphatic endothelial cell marker” refers to a marker that identifies a cer- tain cell as a lymphatic endothelial cell. By a biomarker that is “specific” for a certain target is meant a biomarker that is present in the target without being substantially present in a non-target. For example, a biomarker specific for lymphatic endothelial cells refers to a biomarker that is not substantially present in cells other than lymphatic endothelial cells. As used herein, the term “nuclear biomarker” refers to a biomarker that localizes to a nucleus. As used herein, the term “binding partner” re- fers broadly to any molecule capable of specific binding to their target antigens, such as biomarkers. Non-lim- iting examples of binding partners include antibodies, antibody mimetics, and oligonucleotide and peptide ap- tamers. As used herein, the term “specific binding” refers to binding where a molecule binds to a particular polypeptide or epitope on a particular polypeptide with- out substantially binding to any other polypeptide or polypeptide epitope. As used herein, the term "antibody" refers gen- erally to an immunoglobulin structure comprising two heavy chains and two light chains interconnected by di- sulfide bonds, and includes both monoclonal antibodies and polyclonal antibodies. Antibodies can exist as in- tact immunoglobulins or as any of a number of well- characterized antigen-binding fragments or single chain variants thereof, all of which are herein included in the term "antibody". Non-limiting examples of said an- tigen-binding fragments include Fab fragments, F(ab)2fragments, Fab' fragments, F(ab')2fragments, Fd frag- ments, Fd' fragments, Fv fragments and scFv. Said frag- ments and variants may be produced by recombinant DNA techniques, or by enzymatic or chemical separation of intact immunoglobulins as is well known in the art. As used herein, the term “treating”, and the like, refer to the administration of a pharmaceutical drug, such a chemotherapeutic agent, or other treatment modalities, such as radiation therapy, to a subject in need thereof for purposes which may include ameliorat- ing, lessening, inhibiting, or curing the epithelial cancer in question, such as OSCC. Amounts and regimens for said administration may be determined readily by those with ordinary skill in the clinical art of treat- ing cancers depending on different variables. As used herein, the term "effective amount" refers to an amount of pharmaceutical drug, such as a chemotherapeutic agent, by which harmful effects of the epithelial cancer in question, such as OSCC, are, at a minimum, ameliorated. DETAILED DESCRIPTION Based on evidence of differential prolifera- tion of lymphatic endothelial cells in the invasive bor- der of OSCC tumors, the present invention provides means and methods for predicting the clinical outcome of ep- ithelial cancers, such as OSCC. Predictive markers and associated information provided by the present invention allow physicians to make more intelligent treatment de- cisions, and to customize the treatment of said cancer to the needs of individual patients, thereby maximizing the benefit of treatment and minimizing the exposure of patients to unnecessary treatments, which do not provide any significant benefits and often carry serious risks due to adverse effects. In one aspect, the present invention provides a method of predicting the clinical outcome for a sub- ject diagnosed with epithelial cancer, especially after surgical resection of said cancer. In essence, the pre- diction is based on the proportion of proliferating lym- phatic endothelial cells in the invasive border of the cancerous tissue. If said proportion is increased as compared to a predetermined reference value, the like- lihood of a positive clinical outcome is decreased. On the other hand, if said proportion is not increased or is below a predetermined reference value, the likelihood for a negative clinical outcome is decreased. In an embodiment, the present invention pro- vides a method of predicting the clinical outcome for a subject diagnosed with OSCC, especially after surgical resection of said OSCC. In essence, the prediction is based on the proportion of proliferating lymphatic en- dothelial cells in the invasive border of the OSCC. If said proportion is increased as compared to a predeter- mined reference value, the likelihood of a positive clinical outcome is decreased. On the other hand, if said proportion is not increased or is below a prede- termined reference value, the likelihood for a negative clinical outcome is decreased. The proportion of proliferating lymphatic en- dothelial cells in the area of interest, such as the invasive border of the OSCC, can be determined by di- viding the number of proliferating lymphatic endothelial cells by the total number of lymphatic endothelial cells. Lymphatic endothelial cells can be identified, and thus their presence and number in a given tissue sample be determined, in different ways. Typically, but without limitation thereto, one or more biomarkers of lymphatic endothelial cells may be employed. For exam- ple, PROX1 (Prospero Homeobox 1) is a protein biomarker specific for lymphatic endothelial cells, expressed in the nucleus of both quiescent and proliferating lym- phatic endothelial cells. Accordingly, in some embodiments, the number of lymphatic endothelial cells in the invasive border of a tissue sample comprising cancerous cells is deter- mining by detecting and counting the number of cells that express PROX1. This may be achieved by contacting the tissue sample, preferably a histological tissue sec- tion, with a binding partner, such as an antibody, ca- pable of selectively interacting with the PROX1 protein present in the tissue sample, and by detecting any bind- ing reaction between the PROX1 protein and said binding partner, thereby enabling counting of lymphatic endo- thelial cells, i.e., PROX1 positive cells, in said in- vasive border. Moreover, lymphatic endothelial cells may be detected by employing other biomarkers of said cells. Such biomarkers include, but are not limited to, CD31, Lyve-1 and Podoplanin, which biomarkers may be used in any combination with PROX1, although each and every com- bination is not listed herein. None of said CD31, Lyve- 1 and Podoplanin are nuclear biomarkers. One challenge associated with Podoplanin is that its expression is not limited to lymphatic endo- thelial cells. As demonstrated in the experimental part, although most lymphatic endothelial cells express Podoplanin, overall Podoplanin stains mostly cancer cells. It follows that it is not possible to distinguish lymphatic endothelial cells from other cells near lym- phatic vessels by Podoplanin only. Therefore, it is en- visaged that PROX1 could be replaced by Podoplanin only if it was possible to make sure that Podoplanin-positive cancer or other non-endothelial cells were not misin- terpreted as lymphatic endothelial cells. In some embodiments, lymphatic endothelial cells in a tissue sample, or a certain area thereof, are detected by contacting the tissue sample, preferably a histological tissue section, with a binding partner, such as an antibody, capable of specific interaction with a nuclear biomarker of lymphatic endothelial cells, such as PROX1, and by detecting the interaction between the nuclear biomarker and said binding partner, if any. The number of proliferating lymphatic endo- thelial cells in a tissue sample can be determined, for example, by detecting and counting the number of pro- liferating cells that express PROX1 or another nuclear biomarker of lymphatic endothelial cells, optionally together with one or more other biomarkers of lymphatic endothelial cells, such as CD31, Lyve-1 and / or Podoplanin. Proliferation of cells can be determined by different ways, as is well known to those skilled in the art. In some embodiments, one or more proliferation markers may be employed, including but not limited to KI-67, a nuclear protein known to be associated with cellular proliferation. KI-67 is present during all ac- tive phases of the cell cycle but is absent from quies- cent cells. Since KI-67 protein expression is required for progression through the cell-division cycle, it is an excellent marker of the proliferation status of a given cell population. Therefore, in some embodiments, KI-67 is used as a proliferation marker for lymphatic endothelial cells, either alone or in combination with one or more other proliferation markers. In some embodiments, proliferating cells in a tissue sample, or a certain area thereof, are detected by contacting the tissue sample, preferably a histolog- ical tissue section, with a binding partner, such as an antibody, capable of selectively interacting with the KI-67 protein expressed by proliferating cells present in the tissue sample, and by detecting the interaction between KI-67 and said binding partner, if any. In accordance with the above, the number of proliferating lymphatic endothelial cells present in the tissue sample may be determined on the basis of co- expression of a nuclear marker of lymphatic endothelial cells, such as PROX1, and a proliferation marker, such as KI-67. In some embodiments, said co-expression may be determined by contacting a tissue sample, preferably a histological tissue section, with a first binding partner specific for PROX1 and a second binding partner specific for KI-67, and by determining whether binding reactions between i) PROX1 and the first binding part- ner, and ii) KI-67 and the second binding partner occur in the same cells. The binding reactions result in for- mation of a first complex, i.e., that of the first bind- ing partner and PROX1, and a second complex, i.e., that of the second binding partner and KI-67, respectively. In embodiments involving histological tissue sections, said first and second binding partners may be contacted with a same tissue section, either simultaneously or sequentially. Also, different tissue sections, prefer- ably successive tissue sections, may be employed. Use of a nuclear biomarker for lymphatic endo- thelial cells, such as PROX1, and a nuclear prolifera- tion marker, such as KI-67, ensures correct identifica- tion of proliferating lymphatic endothelial cells. If for example a cytoplasmic biomarker of lymphatic endo- thelial cells was used instead of a nuclear biomarker, it would be challenging to make sure that in all in- stances the signal originating from the proliferation marker and the signal originating from the lymphatic endothelial cell marker are from the same cells. The clinical outcome of the method of the in- vention may be expressed, for example, in terms of Re- currence-Free Survival (RFS), Disease-Specific Survival (DSS), Overall Survival (OS), or the like. Indeed, as demonstrated in the examples, in- creased number of proliferating lymphatic endothelial cells in the invasive border of OSCC is indicative of decreased likelihood of a positive outcome as determined by decreased RFS, DSS as well as OS. For example, three- year RFS rate was 96% for the patients with a low LEC proliferation value versus 64% for the patients with a high LEC proliferation value, when a median LEC prolif- eration percentage of 3.4 was used as a cut-off for stratifying each patient as having either a low LEC proliferation value or a high LEC proliferation value. The following formula was used for calculating the LEC proliferation values: LEC proliferation value (%) = n(KI67+PROX1 positive cells ) / n(PROX1 positive cells) Using the same cut-off value, three-year DSS rate was 100% for the OSCC patients with a low LEC proliferation value versus 73% for the patients with a high LEC proliferation value. Three-year OS rate, in turn, was 86% for the patients with a low LEC prolifer- ation value versus 63% for the patients with a high LEC proliferation value. In some embodiments, use of additional bi- omarkers may provide greater predictive value than PROX1 and KI-67 alone. Thus, the detection of one or more further biomarkers in a sample increases the percentage of true positive and true negative predictions and would decrease the percentage of false positive or false neg- ative predictions. Thus, the methods of the present in- vention can include the measurement of more than one additional biomarker, including one or more biomarkers listed in Table 2. In such methods, PROX1 can be replaced or complemented by using one or more other potential nuclear biomarkers specific for lymphatic endothelial cells, and / or KI-67 can be replaced or complemented by using one or more other proliferation markers. In some embodiments, the method of the present invention can include the detection of one or more bi- omarkers selected from the biomarkers of Table 2 in lymphatic endothelial cells, preferably identified as such on the basis of their PROX1 expression or on the basis of another nuclear marker specific for lymphatic endothelial cells. Generally, determining the level of a biomarker expression at protein level comprises contacting a sam- ple obtained from a subject in need of said determina- tion with a binding partner, such as an antibody, spe- cifically recognizing polypeptide in question under con- ditions wherein the binding partner specifically inter- acts with the biomarker; and detecting said interaction (if any); wherein the presence or degree of said inter- action correlates with the presence of the biomarker or the level of the biomarker expression in said sample. Binding partners alternative to antibodies and fragments thereof suitable for determining biomarker expression at protein level include but are not limited to oligo- nucleotide or peptide aptamers. Immunohistochemistry (IHC) is a preferred method for determining the proportion of proliferating lymphatic endothelial cells in tissue sample and / or for determining the expression level of a selected biomarker in cells. IHC specifically provides a method of detect- ing targets in a sample or tissue specimen in situ. The overall cellular integrity of the sample is maintained in IHC, thus allowing detection of both the presence and location of the targets of interest. Typically, IHC includes the following steps i) fixing a tissue sample with formalin, ii) embedding the sample in paraffin, iii) cutting the sample into sec- tions for staining, iv) incubating said sections with a detectably labelled binding partner specific for the selected marker, v) rinsing said sections, and vi) de- tecting the marker-binding partner complex by an appro- priate technique, depending on the detectable label em- ployed. Multiple binding partners for different selected markers may be used simultaneously, each binding partner labelled with a different detectable label. If desired, counterstaining may also be used, e.g. Hematoxylin & Eosin, DAPI, Hoechst. For IHC, a binding partner may be detectably labelled either directly or indirectly, e.g. through a secondary antibody-based labelling, thereby enabling detection of the target protein (i.e. the selected marker). Exemplary labels include radioactive isotopes (e.g.3H,14C,32P,35S or125I), non-radioactive heavy metal isotopes (e.g. those listed in Table 2), fluorescent dyes (e.g. fluorescein, rhodamine, phycoerythrin, flu- orescamine), chromophoric dyes (e.g. rhodopsin), chem- iluminescent agents (e.g. luminal, imidazole), enzymes (e.g. horseradish peroxidase, alkaline phosphatase, beta-lactamase), ligands, bioluminescent proteins (e.g. luciferin, luciferase), haptens (e.g. biotin), parti- cles (e.g. gold) and combinations thereof. In some embodiments, the resulting stained sam- ples are each imaged using a system for viewing the detectable label and acquiring an image, such as a dig- ital image of the staining. Methods for image acquisi- tion are well known to one of skill in the art. For example, once the sample has been stained, any optical or non-optical imaging device can be used to detect the stain or biomarker label, such as, for example, upright or inverted optical microscopes, scanning confocal mi- croscopes, mass cytometric imaging devices, cameras, scanning or tunneling electron microscopes, canning probe microscopes and imaging infrared detectors. In some examples, the image can be captured digitally. The obtained images can then be used for quantitatively or semi -quantitatively determining the amount of the marker of interest in the sample, or the absolute number of cells positive for the maker of interest, or the surface of cells positive for the maker of interest. Various automated sample processing, scanning and analysis systems suitable for use with IHC are also available in the art. Determining the proportion of proliferating lymphatic endothelial cells in a tissue sample, as well as determining the expression level of a selected bi- omarker may also be achieved using techniques other than IHC. For example, tumor-derived cell suspensions can be subjected to single-cell RNA sequencing for identifying PROX1 mRNA expressing cells which express also KI-67 mRNA (MKI67). Also spatial transcriptomics, including methods based on single-cell RNA sequencing, and fluo- rescence in situ hybridization (FISH)-methods, can be applied. Spatial methods are preferred because they en- able recognition of the tumor margin and allow directing the analysis to the cells located therein. In some embodiments, the present method may further include analysis of known prognostic factors to improve the method’s prognostic power. Such factors in- clude, but are not limited to, one or more of the fol- lowing: nodal involvement (pN), tumor size (pT), inva- sion depth, tumor growth pattern (budding), the number of tumor-infiltrating lymphocytes, and T-cell activa- tion markers (e.g., Granzyme B). A combined risk score can be calculated by integrating multiple prognostic factors with the LEC proliferation percentage. Factors with a significant, independent correlation to high re- currence-free survival (hazard ratio < 1) decrease the combined risk score, whereas factors with a significant, independent correlation to low recurrence-free survival (hazard ratio > 1) increase the combined risk score. Patients with a moderate to high total combined risk score are at an increased risk of cancer recurrence. Combining multiple prognostic factors with LEC prolif- eration assessment may provide a more accurate predic- tion of cancer progression. In an embodiment, the subject with epithelial cancer to whom the clinical outcome is to be predicted has already experienced lymph node metastasis. When those patients who had pathologically verified sentinel lymph node metastasis (N1, N2 or N3) were analyzed as a separate cohort, the patients with high numbers of pro- liferating LECs still had worse outcome than the ones with low numbers of proliferating LECs (Fig. 10, panel a). These findings have important clinical value, since they show that patients who already have local metasta- sis in the lymph node (an unfavorable prognostic factor by itself) can still be divided into good and poor prog- nosis subgroups based on the number of proliferating LECs. In an embodiment, the subject with epithelial cancer to whom the clinical outcome is to be predicted does not have lymph node metastasis. When those patients who had no pathologically verified sentinel lymph node metastasis (N0) were analyzed as a separate cohort, the patients with high numbers of proliferating LECs had worse outcome than the ones with low numbers of prolif- erating LECs (Fig. 10, panel b). These analyses further demonstrate that the predictive value of proliferating LECs is not dependent on the lymph node status of the patients. Moreover, they have important clinical value, since they show that patients without any local metas- tasis in the lymph node (and thus typically representing early, low risk cases) can already be divided into good and poor prognosis groups based on the number of pro- liferating LECs. In some implementations, the present method of predicting the clinical outcome for a subject with ep- ithelial cancer, such as OSCC, may further include ther- apeutic intervention. Once an individual is identified to have a decreased likelihood for a positive clinical outcome, he / she may be subjected to any appropriate treatment modality known to those skilled in the art, including but not limited to, chemotherapy, radiation therapy, immunotherapy, and / or targeted therapy. The present invention provides advantages not only for individual patient care but also for better selection and stratification of patients for clinical trials. For example, OSCC patients or patients with other epithelial cancers grouped as high-risk patients could be recruited to clinical studies with the aim of developing efficient new personalized therapeutic tools such as new medical procedures or drugs. In one aspect, the present invention also pro- vides a kit for predicting the clinical outcome for a subject diagnosed with epithelial cancer, such as OSCC. In an embodiment, the kit comprises a reagent capable of indicating the presence of a nuclear bi- omarker specific for lymphatic endothelial cells and a reagent capable of indicating cell proliferation. In an embodiment, the kit comprises a reagent capable of indicating the presence PROX1 and / or a rea- gent capable of indicating the presence of KI-67. In an embodiment, the kir comprises a first binding partner capable of specific binding to PROX1 (as a reagent capable of indicating the presence PROX1) and / or a second binding partner capable of specific binding to KI-67 (as a reagent capable of indicating the presence of KI-67). In any of the embodiments, the kit may further comprise one or more additional reagents capable indi- cating the presence of a marker listed in Table 2 and / or one or more binding partners capable of specific binding to a protein marker listed in Table 2. In any of the embodiments, one or more of the binding partners may be detectably labelled, inde- pendently from each other, either directly or indi- rectly. In some embodiments, the kit comprises one or more binding partners capable of specifically binding to or otherwise detecting one or more of the biomarkers of the invention which associate with a probable outcome of said epithelial cancer (e.g. OSCC). In some embodi- ments, the kit may also comprise at least one detecting reagent or a detecting apparatus capable of indicating binding of the one or more binding partners to said one or more biomarkers or otherwise capable of indicating the presence or the level of said one or more biomarkers in a sample obtained from a subject whose probable out- come of epithelial cancer (e.g. OSCC) is to be deter- mined or predicted. In some embodiments, the kit further comprises positive and / or negative control samples or predeter- mined reference values that can be assayed or used for comparing to the patient sample. It is to be understood that the contents of the kit may vary depending on the assay technique to be employed, as is readily apparent to those skilled in the art. Although the description above and the exper- imental part focus on OSCC, it is envisaged that the present method applies to other carcinomas as well. Non- limiting examples of such other carcinomas include car- cinomas of the gastrointestinal tract, pancreas, the lungs, breast, uterus, prostate, kidney, bladder, or the skin. It is obvious to a person skilled in the art that with the advancement of technology, the basic idea of the invention may be implemented in various ways. The invention and its embodiments are thus not limited to the examples described above, instead they may vary within the scope of the claims. EXPERIMENTAL PART EXAMPLE 1. Identification of prognostic factors in OSCC Samples and preparation Formalin-fixed paraffin-embedded (FFPE) whole tumor sections of early-stage (cT1-T2) oral cavity squa- mous cell carcinomas (n = 95, Table 1) were used as study samples. Corresponding hematoxylin-eosin (H&E) - stained sections were used to identify a region of in- terest (ROI) from each tumor sample. Criteria for ROI selection (1 = most important, 3 = least important) were: 1. Intact tissue 2. Margin area between invasive tumor and sur- rounding stroma. 3. Lymphocyte-rich area In other words, the ROI was a good-quality, immune-cell infiltrated invasive margin of the tumor containing both malignant cancer cells and normal cells. Table 1. Patient cohort; clinical and pathological char- acteristics Characteristic N = 951Age 67 (58, 74) Gender Female 46 (48%) Male 49 (52%) Smoking status Non-smoker 42 (45%) Active smoker 26 (28%) Ex-smoker 25 (27%) Hospital HUS 29 (31%) KYS 20 (21%) OYS 15 (16%) TAYS 16 (17%) TYKS 15 (16%) pT 1 42 (44%) 2 47 (49%) 3 6 (6.3%) pN 0 76 (81%) 1 8 (8.5%) 2a 1 (1.1%) 2b 3 (3.2%) 3b 6 (6.4%) OSCC subsite (ICD-10) Tongue (C02) 73 (77%) Floor of mouth (C04) 11 (12%) Gum (C03) 6 (6.3%) Other and unspecified parts of 5 (5.3%) mouth (C06) Surgery marginal (mm) 3.50 (2.00, 5.50) Re-operation 15 (16%) Post-operative treatment None 68 (72%) Radiotherapy 19 (20%) Chemoradiotherapy 8 (8.4%) Follow-up time (months) 29 (16, 37) IMC images (n) 1 90 (95%) 2 1 (1.1%) 0* 4 (4.2%)1Median (interquartile range); n (%) *IMC staining unsuccessful, sample missing or no tumor left in the section. HUS = University Hospital of Helsinki; KYS = Univer- sity Hospital of Kuopio; OYS = University Hospital of Oulu; TAYS = University Hospital of Tampere; TYKS = University Hospital of Turku Imaging mass cytometry (IMC) staining protocol Immunostainings of the samples were carried out according to the following protocol: 1. Deparaffinization and hydration; 2. Heat-induced epitope retrieval (Aptum 2100- retriever) in pH 6.0 citrate-buffer (Dako target re- trieval); 3. A circular area (diameter = 6mm) surrounding the pre-determined ROI was trimmed from the tissue sec- tion and circled with hydrophobic IHC pen; 4. Blocking with 5% BSA (bovine serum albumin) in PBS (phosphate-buffered saline), 45min in room tem- perature; 5. A combined mix of 25 commercial metal-con- jugated or in-house conjugated antibodies (dilution- range 1:50 – 1:400) (see Table 2) was prepared in PBS containing 0.5% BSA; 6. The antibody mix was applied on each trimmed tumor section and incubated overnight in +4°C; 7. Cell-ID intercalator-Ir (Ir191 and Ir193) diluted 1:200 in PBS and incubated on the sections 30min in room temperature (to detect nuclei of the cells); and 8. Air-drying and storing samples in room tem- perature. Table 2. Antibody panel for IMC Cancer cells Marker Metal tag Commercial Antibody antibody clone Anti-E-Cad- 158Gd Fluidigm 24E10 herin 3158029D Immune suppression markers Marker Metal tag Commercial Antibody antibody clone CD274 / PD-L1 150Nd Fluidigm E1L3N 3150031D CD276 / B7-H3 173Yb Fluidigm Polyclonal 3173014D CD73 165Ho Cell Signal- D7F9A ing 87661SF IDO1 171Yb Ionpath EPR20374 717101 LAG-3 153Eu Fluidigm D2G40 3153028D PD-1 148Nd R&D AF1086 Polyclonal PD-L2 172Yb Fluidigm D7U8C 3172031D TIM-3 154Sm Fluidigm D5D5R 3154024D VISTA 160Gd Fluidigm D1L2G 3160025D Extracellular matrix / Cell cytoskeleton Marker Metal tag Commercial Antibody antibody clone Anti-Colla- 169Tm Fluidigm Polyclonal gen Type I 3169023D Anti-Vi- 143Nd Fluidigm D21H3 mentin 3143027D aSM-actin 141Pr Fluidigm 1A4 3141017D Leukocytes Marker Metal tag Commercial Antibody antibody clone CD206 174Yb Ionpath E2L9N 717402 CD3 170Er Fluidigm Polyclonal 3170019D CD45 175Lu Ionpath 2B11 & 717501 PD7 / 26 CD56 151Eu Ionpath MRQ-42 715101 CD8a 162Dy Fluidigm C8 / 144B 3162034D FoxP3 155Gd Fluidigm PCH101 3155018D Granzyme B 167Er Fluidigm EPR20129-217 3167021D Lymphatics Marker Metal tag Commercial Antibody antibody clone CD31 152Sm Ionpath EP3095 715201 Lyve-1 163Dy Abcam EPR21857 ab232935 Podoplanin 156Gd BioLegend D2-40 916606 Prox1 161Dy ab236026 EPR19273 Cell proliferation Marker Metal tag Commercial Antibody antibody clone Anti-Ki-67 168Er Fluidigm B56 3168022D It is to be noted that basically all the mark- ers employed can stain several cell types and could be listed under more than one category (shown here is a simple classification used for the present panel gener- ation). IMC Imaging Imaging of the immunostained samples was car- ried out with Hyperion™ imaging system and CyTOF soft- ware. The pre-determined ROI was identified for each sample from an optical panorama image generated during the Hyperion preparation steps by utilizing information from the corresponding H&E-section. IMC data (27 chan- nels) from an area of 1mm x 1mm was acquired (one ROI / sample), and data obtained stored in .mcd files. Data preparation and cell segmentation Data preparation and cell segmentation steps were performed in accordance with IMC Segmentation Pipe- line developed by Vito Riccardo Tomaso Zanotelli and Bernd Bodenmiller (ImcSegmentationPipeline: A pixel- classification based multiplexed image segmentation pipeline. Zenodo. [Online] 2022). Randomly cropped image stacks of each sample were used in generating Ilastik pixel-classification training set. Image channels for nuclear markers (Ir193, Ir193), E-cadherin, Podoplanin, Vimentin, CD31, TIM-3, Prox1, CD8a, Granzyme B, CD3, IDO1, CD206 and CD45 were used in the training set for manual pixel labeling. Three compartments were labeled: nucleus, cytoplasm and background. Final pixel probabilities produced by the trained Ilastik pixel-classification algorithms were exported as rgb images and used to make cell segmenta- tion masks in CellProfiler. Single-cell and image features were measured in CellProfiler and used to generate a spatial single- cell dataset in R for downstream analysis. Data analysis Downstream analyses were performed in accord- ance with IMC analysis workflow by Windhager et al. (An end-to-end workflow for multiplexed image processing and analysis. 2020, bioRxiv.). Prox1-positive cells (named as LECs, n = 4773) were gated from the single-cell data to be analyzed separately. Gating was confirmed successful by visual- izing gated cells on IMC images. Batch effect correction was performed with mu- tual nearest neighbor (MNN) method. Principal component analysis (PCA) was performed with corrected expression data. Unsupervised clustering was performed using PhenoGraph-clustering algorithm (Levine et al. Data- Driven Phenotypic Dissection of AML Reveals Progenitor- like Cells that Correlate with Prognosis. 2015, Cell, 162(1), 184–197). A distinct LEC cluster (named as LECk) charac- terized with high KI-67 expression was identified, and LEC proliferation value (%) for each sample was calcu- lated dividing LECk cell amount by total LEC amount in the sample. Proliferation value was calculated for each sample containing at least 15 LECs. Patients were divided to two groups based on LEC proliferation value. Median (3.4%) cut-off was used to define the Low and High groups. Log-rank test was used to compare the survival distributions of the two patient groups, and time-dependent ROC curve estimation from the censored survival data was performed using Kaplan-Meier method by Heagerty, Lumley & Pepe (Time- dependent ROC curves for censored survival data and a diagnostic marker. 2000, Biometrics, 56(2), 337–344.4). Results Kaplan-Meier estimates were used to analyze the Recurrence-free survival (RFS), Disease-specific sur- vival (DSS) and Overall survival (OS) for the patients in the study cohort (Figures 3A-3C). The patients were categorized into two groups based on their LEC prolif- eration (LECp) values. The low group comprised those with values below the median, while the high group in- cluded those with values above the median. The high- group exhibited a three-year RFS probability of 64%, whereas the low-group demonstrated a probability of 96%. Similar trends were observed in DSS and OS probabili- ties. The low-group displayed a DSS of 73% compared to 100% in the high-group, while the OS was 63% in the low- group and 86% in the high-group. When patients with sentinel lymph node-positive (pN+) status were analyzed separately, the high LECp group exhibited a three-year RFS of 0%, compared to 100% in the low LECp group (Figure 10, panel a). Similarly, among patients with sentinel lymph node-negative (pN0) status, the three-year RFS was 75% in the high LECp group and 95% in the low LECp group (Figure 10, panel b). The LEC proliferation value was further eval- uated in the whole patient cohort as a potential bi- omarker using time-dependent receiver operating charac- teristic (ROC) analysis (Figures 3D-3F). In three-year follow-up time, LEC proliferation value demonstrated an AUC value of 0.863 for predicting RFS, 0.984 for DSS, and 0.684 for OS. Prox1 and Ki-67 were also evaluated individu- ally by calculating the total number of Prox1-positive cells (LEC count) and the total percentage of Ki-67- positive cells (total proliferation). Kaplan-Meier analysis (Figures 4A and 4B) was performed by dividing the patients into high and low groups based on the median values for both markers. When used alone neither marker showed any significant difference in survival (RFS, DSS or OS). ROC analysis of the three-year RFS (Figures 4C and 4D) gave an AUC-value of 0.618 for LEC count and 0.572 for total proliferation value further demonstrat- ing that neither marker individually holds prognostic value within the context of this study. Some of the results are summarized in Table 3 below. In the Table 3, LEC proliferation refers to cells expressing both PROX1 and KI-67, while LEC count refers to cells expressing only PROX1. Table 3. Recurrence-free Disease-specific Overall 3-year p- 3-year p- 3-year p- survival value survival value survival value LEC pro- 0.002 0.005 0.035 liferation high 64 73 63 (56-88) (55-95) (47-86) low 96 100 86 (89-100) (100-100) (74-100) LEC count 0.33 0.99 0.78 high 78 88 77 (93-66) (100-76) (92-64) low 87 87 77 (99-76) (100-76) (93-64)1Kaplan-Meier % (95 % CI),2Log-rank test EXAMPLE 2. Prognostic value of Podoplanin Podoplanin expression in the OSCC tissue sam- ples described in Example 1 was evaluated for its po- tential prognostic value. Manually gated training data and random forest classification were used to identify all LECs and tumor cells from the segmented IMC data described in Example 1. The identification was visually confirmed from the tumor sections. Cutoff expression values were calculated from the identified LECs as the mean - 1 standard devi- ation (SD) for PROX1 (0.9) and PDPN (2.1). The cutoff value for KI-67 positivity (2) was determined through visual inspection. These cutoff values were then used to gate PROX1+, PDPN+, and KI-67+ cells from the IMC data. In addition to visual inspection, expression data from a random sample of 70,000 cells was analyzed to evaluate LEC and tumor cell expression profiles in terms of PROX1 and PDPN. The proportion of KI-67 posi- tivity in PROX1+ cells and PDPN+ cells was calculated for each sample separately to compare the prognostic value of both marker combinations. According to the results, Podoplanin is ex- pressed at high levels by OSCC tumor cells, whereas Prox1 remains specific to LECs, even in highly tumor- rich regions (Figures 5A and 5B). These results demon- strate that Podoplanin cannot distinguish lymphatic vas- culature (LECs) near the OSCC tumor cells. Moreover, although some lymphatic vessels can be visualized with Podoplanin, individual endothelial cell nuclei cannot be separated from nuclei of other cells (e.g., lympho- cytes and fibroblasts) closely associated with the lym- phatic vessels. This leads to overestimation of LEC pro- liferation, since Ki-67 can only be expressed in the nucleus (Figure 6, Table 4). Collectively these data show that the majority of Podoplanin1+Ki67+ are cells other than LECs. Most LECs express both Prox1 and Podoplanin, but the vast majority of Podoplanin+ cells are tumor cells or other non-LEC cell types (Figure 7). Higher- than-median Ki-67 positivity in Podoplanin+ cells does not significantly predict overall or recurrence-free survival. Conversely, higher-than-median Ki-67 positiv- ity in Prox1+ cells significantly correlates with de- creased overall survival and recurrence-free survival (Figure 8, Table 4).

[0002] EXAMPLE 3. Combined risk score The patient cohort and methods described in Example 1 were used to analyze various cell populations from the IMC data to evaluate their individual and com- bined prognostic value. Random forest classification and manually gated training data were utilized to identify predefined cell populations, including LECs (Prox1+), T-cells (CD3+, CD45+), CD3- leukocytes (CD3-, CD45+), and tumor cells (E-cadherin+). These cell populations were further analyzed using the marker-based clustering method de- scribed in Example 1. Combined with the spatial infor- mation, the analyses revealed several cell phenotypes with significant independent prognostic value: LECp (Prox1+, Ki-67+), Tc Granzyme B+ (CD3+, CD45+, CD8a+, Granzyme B+), and Tc TIL (CD3+, CD45+, CD8a+, distance to nearest tumor <0 µm). For each subpopulation, the amount in a sample was calculated as a percentage of the corresponding parent cell type. Higher than median val- ues are regarded as “high” and lower or equal to median values are regarded as “low.” Multivariate hazard ratios (HR) were calcu- lated for each clinical parameter and cell phenotype using the Cox proportional hazards model. The multivar- iate model included known prognostic parameters: pN, pT, and age. Factors with significant independent prognostic value were chosen for combined risk analysis (Table 5). Factors with an HR less than 0 decrease the combined risk score by 1 point, and factors with an HR greater than 0 increase the combined risk score by 1 point. Finally, patients were grouped by their total risk score (Figure 9A). Patients with a high risk score (2 and 3) show significantly worse recurrence-free and overall survival compared to low risk score groups (-1, 0, and 1) (Figure 9B). High risk score patients have an estimated mean recurrence-free survival of 52%, compared to 92% for low risk score patients (Figure 9C). These results demonstrate the prognostic potential of using the LEC proliferation measurement in combination with other prognostic parameters. Table 5. Hazard ratio Risk score Recurrence Overall survival pN % 0 or unknown - - 0 > 0 6.04 (p = 3.17 (p = +1 0.005) 0.043) Tc TIL % Low or unknown - - 0 High 0.21 (p = 0.11 (p = -1 0.036) 0.003) Tc Granzyme B % Low or unknown - - 0 High 9.38 (p = 3.35 (p = +1 0.003) 0.044) LECp % Low or unknown - - 0 High 9.47 (p = 5.65 (p = +1 0.005) 0.01)1HR adjusted by Age, pT and pN

Claims

CLAIMS 1. A method of predicting clinical outcome for a subject diagnosed with epithelial cancer, comprising: i) determining the proportion of proliferating lymphatic endothelial cells in an invasive border of a cancerous tissue sample obtained from said subject, wherein the lymphatic endothelial cells are identified by a nuclear biomarker specific for lymphatic endothe- lial cells; ii) comparing the proportion determined at step i) with a predetermined reference value; and ii) concluding that the subject has decreased likelihood of a positive clinical outcome, when the pro- portion determined at step i) is higher than the prede- termined reference value; or concluding that the subject has increased likelihood of a positive clinical outcome when the proportion determined at step i) is lower than the predetermined reference value.

2. The method according to claim 1, wherein the proportion of proliferating lymphatic endothelial cells is calculated by dividing the number of proliferating lymphatic endothelial cells with the total number of lymphatic endothelial cells in an area of interest in the invasive border of the cancerous tissue sample.

3. The method according to claim 1 or 2, wherein the nuclear biomarker is Prospero homeobox pro- tein 1 (PROX1).

4. The method according to any one of claims 1-3 wherein proliferating cells are identified on the basis of a proliferation marker KI-67.

5. The method according to any one of claims 1-4, wherein the proliferating lymphatic endothelial cells are identified on the basis of co-expression of a proliferation marker KI-67 and a lymphatic endothelial cell marker Prospero homeobox protein 1 (PROX1).

6. The method according to any one of claims 1-3, wherein the tissue sample is a histological tissue sample.

7. The method according to claim 6, wherein the co-expression of KI-67 and PROX1 is detected by con- tacting the histological tissue sample with a detectably labelled anti-KI-67 antibody and a detectably labelled anti-PROX1 antibody.

8. The method according to any one of claims 1-5, wherein the expression of KI-67 and PROX1 is de- tected by single-cell RNA sequencing or by spatial tran- scriptomics.

9. The method according to any one of claims 1-8, further comprising determining the expression level one or more biomarkers listed in Table 2, and comparing the determined expression levels to respective prede- termined reference values.

10. The method according to claim 9, wherein said expression level is determined by a technique se- lected from the group consisting of immunohistochemis- try, single-cell RNA sequencing and spatial tran- scriptomics.

11. The method according to any one of claims 1-10, wherein the epithelial cancer is squamous cell carcinoma, adenocarcinoma, transitional cell carcinoma or basal cell carcinoma.

12. The method according claims 11, wherein the squamous cell carcinoma is oral squamous cell carcinoma (OSCC), preferably selected from the group consisting of OSCC of the tongue, OSCC of the base of the tongue, OSCC of the cheek, OSCC of the gums, OSCC of the floor of the mouth, and OSCC of the hard palate.

13. The method according to any one claims 1- 12, wherein said clinical outcome is expressed in terms of Recurrence-free survival, Disease-specific survival, or Overall survival.

14. A kit for use in the method according to any one of claims 1-13, comprising a reagent capable of indicating the presence of a nuclear biomarker specific for lymphatic endothelial cells and a reagent capable of indicating cell proliferation.

15. The kit according to claim 14, comprising at least one of a reagent capable of indicating the presence PROX1 and a reagent capable of indicating the presence of KI-67.

16. The kit according to claim 15, comprising at least one of a first binding partner capable of spe- cific binding to PROX1 and a second binding partner capable of specific binding to KI-67.

17. The kit according to any one of claims 14- 16, further comprising one or more additional reagents capable indicating the presence of a marker listed in Table 2 and / or one or more binding partners capable of specific binding to a protein marker listed in Table 2.

18. The kit according to claim 16 or 17, wherein one or more of the binding partners are detect- ably labelled, independently from each other, either directly or indirectly.

Images