WO2022195469A1

WO2022195469A1 - Predictive marker for sensitivity to immune checkpoint blockade in prostate cancer and other cancer types

Info

Publication number: WO2022195469A1
Application number: PCT/IB2022/052315
Authority: WO
Inventors: Ajinkya REVANDKAR; Marco BOLIS; Arianna VALLERGA; Daniela BOSSI; Jean-Philippe THEURILLAT
Original assignee: Fondazione Per L’Istituto Oncologico Di Ricerca (Ior)
Priority date: 2021-03-18
Filing date: 2022-03-15
Publication date: 2022-09-22

Abstract

The present invention relates to an in vitro method for identifying and/or selecting those subjects suffering from cancer who are affected by microsatellite instability-like (MSI-like) cancer and/or are responsive immune checkpoint blockade (ICB) therapy, as well as to a kit and computer program enabling the above identification and/or selection. The method involves determining the expression levels of the signature markers CD11b (ITGAM) and/or CD11c (ITGAX).

Description

Predictive Marker for Sensitivity to Immune Checkpoint Blockade in Prostate Cancer and

Other Cancer Types

The present invention relates to an in vitro method for identifying and/or selecting those subjects suffering from cancer who are affected by microsatellite instability-like (MSI-like) cancer and/or are responsive to immune checkpoint blockade (ICB) therapy, as well as to kits, a computer- implemented method and computer program implementing said method.

STATE OF THE ART

Cancer is the second leading cause of death globally, accounting for an estimated 9.6 million deaths, or one in six deaths, in 2018.

Notably, prostate cancer is the most commonly non-skin cancer diagnosed in men, representing one of the leading causes of cancer death. Survival rates are significantly low for prostate cancers that advance to metastatic castration-resistant disease, and unfortunately, despite recent advances and a range of therapeutic options, outcomes are varied, and clinicians are not able to predict patients’ response to the available therapies.

Immune checkpoint blockade (ICB) has shown remarkable clinical efficacy in several cancer types (Robert, 2020). However, prostate cancers are mostly immunologically cold tumors, and thus only a smaller fraction of patients have shown responses in clinical trials. However, these patients have shown also durable responses (Hussain et al., 2018; Kwon et al., 2014; Lu et al., 2017; Szymaniak et al., 2020; Tao et al., 2017). Among the responders to immune checkpoint blockade (ICB) a small subpopulation of prostate cancer patients with microsatellite instability (MSI) has been identified. Hence, patients with prostate cancer do not receive immune checkpoint inhibitors unless their tumor are proven to be positive for microsatellite instability. That being said, other prostate cancer patients and more generally patients that are currently not qualifying for immunotherapy in any tumor type might benefit from ICB therapy as well.

However, up to date no reliable and accurate tests exist that could be used to predict the response to ICB therapy of patients suffering from cancer, such as prostate cancer or other types of cancer. In this context, there is therefore an urgent need for predictive biomarkers that could enable tailoring the ICB treatment of cancer to individual patients. SUMMARY OF THE INVENTION

By generating a prostate cancer transcriptome atlas enabling the identification of roadmaps to tumor progression, the authors of the present invention have discovered a subpopulation of prostate cancers characterized by MSI-like features that could benefit from immune checkpoint blockade therapy (ICB). In contrast to MSI, these MSI-like prostate tumors do not exhibit the typical increase of somatic mutation burden and test negative for MSI. Notably, the inventors found that these MSI-like prostate tumors are infiltrated by inflammatory cells and show an upregulation of immune checkpoint ligands. Taken together, these findings strongly suggest that patients whose tumors show MSI-like characteristics and express consistently higher levels of immune checkpoint ligands can also benefit from ICB.

The inventors have also identified a panel of transcriptional signature markers for MSI-like prostate cancer in clinical samples. Surprisingly, the authors found that determining and/or quantifying the expression levels of the above signature markers in biological samples obtained from subjects suffering from cancer, enables an accurate identification of those subjects having MSI-like tumors (for example, around 10% of prostate cancer patients) and/or subjects who could benefit from ICB therapy.

Hence, a first object of the present invention refers to an in vitro method for identifying and/or selecting those subjects suffering from cancer who are affected by microsatellite instability-like (MSI- like) cancer and/or are responsive to immune checkpoint blockade (ICB) therapy, said method comprising: a. determining and/or quantifying the expression levels of the signature markers CD11b

(ITGAM) and/or CD11c (ITGAX) in a biological sample isolated from said subjects.

Another object of the present invention refers to a kit for identifying and/or selecting those subjects suffering from cancer who are affected by microsatellite instability-like (MSI-like) cancer and/or are responsive to immune checkpoint blockade (ICB) therapy, said kit comprising one or more agents for determining and/or quantifying the expression levels of the signature markers CD11b (ITGAM) and/or CD11c (ITGAX) in a biological sample isolated from said subjects.

A further object of the invention is a computer-implemented method for identifying and/or selecting those subjects suffering from cancer who are affected by microsatellite instability-like (MSI- like) cancer and/or are responsive to immune checkpoint blockade (ICB) therapy, the method comprising: a. receiving at, at least, one processor, input data representing the expression levels of the signature markers CD11b (ITGAM) and/or CD11c (ITGAX) in a biological sample isolated from said subjects; b. computing at, at least, one processor, a score using said input data.

A further object of the present invention is an immune checkpoint inhibitor for use in the treatment of a subject suffering from microsatellite instability-like (MSI-like) cancer; and an immune checkpoint inhibitor for use in a method of treatment of a subject suffering from cancer, wherein said method comprises the following steps:

(i) determining whether a biological sample isolated from said subject expresses the signature markers CD11b (ITGAM) and/or CD11c (ITGAX); and

(ii) if, in step (i), said biological sample expresses said CD11b (ITGAM) and/or CD11c (ITGAX) markers, administering to said subject an effective amount of said immune checkpoint inhibitor.

DETAILED DESCRIPTION OF THE FIGURES

Figure 1. Main Trajectory to Prostate Cancer Progression. (A) Principal component analysis (PCA) of pan-prostate cancer transcriptomes obtained from the indicated studies of normal, primary, castration-resistant (CRPC), and neuroendocrine prostate cancer (NEPC). (B) Unbiased trajectory analysis identifies the main path to disease progression. Quantification of the path is indicated by inferred pseudo-time. (C) Plot representing the correlation between mRNAs and pseudo-time inferred along the main trajectory. Polycomb repressive complex related genes highlighted in orange, cell cycle-related genes in green, immune response in blue, and AR signaling in red. X-axis: Pearson’s correlation coefficient between mRNAs and pseudo-time; Y-axis: The associated significance adjusted for False Discovery Rate (FDR) and expressed in form of -10xlog10(FDR). (D) Scatterplot revealing correlation between mRNAs and protein abundances, expressed in form of fold-change (log-scale) between CRPCs and Primary tumors. (E) Computed Pearson’s correlation between samples’ numeric copy number status (-2: homozygous deletion; -1 : heterozygous deletion; 0: wild-type; 1:gain; 2:amplification) and inferred pseudo-time, stratified for primary and metastatic tumors (CRPC, NEPC). (F) Boxplots representing different pseudo-time distributions for RBI- specific copy number alterations (homozygous, heterozygous, wild-type, gains). (G) Corresponding PCA plot highlighting RB1 copy-number status across samples.

Figure 2. Alternative Trajectory Linked to MSI-like Features and Viral Mimicry. (A) Hierarchical clustering of CRPC and NEPC samples identifies metastatic prostate cancer subtypes with different AR levels: AR-HIGH, AR-LOW, AR-negative NEPC, double-negative prostate cancers (DNPC), and prostate cancers with transcriptional features of microsatellite instability (MSI-like). The MSI-like samples are located in the center of the PCA plot and connect primary prostate cancer to AR- negative prostate cancers in a straight fashion.AR negative prostate cancers including both NEPC and DNPC were clustered into a single group due to overlapping PCA-positioning. (B) The relative percentage of patients’ predicted MSI-status in the corresponding subgroups performed using PreMSIm (Li et al., 2020). (C) Expression of RNAs derived from DNA satellites, quantified in AR- HIGH, MSI-like, and NE/DN-PC subgroups, shows higher levels in MSI-like tumors. (D) MSI-like tumors show a lower mutation load (ML) compared to the other metastatic prostate cancer subgroups. ML represents the total number of somatic mutations in each sample. (E) Quantification of total immune infiltrate is higher in MSI-like compared to AR-HIGH and NE/DN-PC subgroups. The analysis was performed using CibersortX (Steen et al., 2020). (F) The density plot of macrophage polarization index (MPI) reveals a significant shift toward M1 -like polarization in MSI-like tumors. MPI was determined using MacSpectrum (Li et al., 2019) based on gene-expression levels. (G) Boxplots representing mRNA expression of the indicated immune checkpoint ligands show a significant upregulation of the latter in MSI-like tumors. (H) Single-sample GSEA (ssGSEA) scores for indicated inflammatory-related pathways show increased activity in MSI-like tumors. Gene-sets were retrieved from the Hallmark collection curated from MSigDB. (I) Quantification of endogenous retroviral elements reveals high expression in MSI-like tumors. Left: Repeated sequences are quantified together as a single entity, and their amount is normalized to the coding fraction of the transcriptome. Right: Separated quantification of RNAs transcribed from Alu-sequences, L1/L2 LINE, and LTR/ERV1 elements. Abundance is expressed in the form of their relative percentage to the total number of repeated sequences. P-values for all panels were computed using Wilcoxon sum-rank test and adjusted for multiple testing using the false discovery rate (FDR). Significance level (FDR): ^* < 0.05, ^** < 0.01, ^*** < 0.001.

Figure 3. Features of MSI-like Primary Prostate Cancers. (A) Protein-protein interaction network representing genes whose expression is enriched in MSI-like tumors identifies the myeloid marker CD11b (ITGAX) and M1 -macrophage marker CD11c (ITGAM) as the most interconnected elements. Both proteins interact with CD18 (ITGB2) to build up complement receptor 3 and 4, respectively, as indicated by the crystal structure. (B) PCA plot highlighting the MSI-like similarity score computed in primary tumors using a 4-tiered scoring system. (TOP50: The top 50 patients characterized by the highest score; UP-INT/LW-INT: patients with upper and lower intermediate scores, composed of 198 and 199 samples respectively. BTM50: The 50 patients characterized by the lowest MSI-like similarity score). (C) PCA plot integrating a separate cohort (Yun et al., 2017) of samples of benign prostate hyperplasia (BPH), primary tumors, and CRPCs. Included are 4 patients with matched samples derived from the primary and corresponding castration-resistant disease. Their progression is indicated by an arrow. MSI-like similarity score was assessed using ssGSEA, and the top 33% of samples endowed with the highest values were predicted as MSI-like (highlighted in yellow). Notably, while patients 1 , 3, and 4 are positioned along the main trajectory, patient 2, characterized by a high MSI-like similarity score, positions to the alternative trajectory. (D) Boxplots representing mRNA expression of the indicated immune checkpoint ligands in MSI-like and non-MSI-like tumors. Checkpoint ligands are significantly up-regulated in MSI-like tumors. P-values were assessed using Wilcoxon sum-rank test and adjusted for multiple testing using the false discovery rate (FDR). Significance level (FDR): ^* < 0.05, ^** < 0.01 , ^*** < 0.001. (E) Frequency of the most occurring somatic point mutations in patients with the highest MSI-like similarity score (TOP50) compared to the other subgroups (UP/LW-INT and BTM50). Significant enrichments (q < 0.05) are highlighted by Asterisk. (F) Barplots representing the enrichment of Gene Ontologies (GO), based on over-representation analysis performed on the most frequently mutated genes within each subgroup. For each subgroup, we selected genes harboring somatic point mutations in more than 5% of patients. Enrichments were performed using ClusterProfile (G) Kaplan Meier estimates of disease-free survival of primary prostate cancers stratified by MSI-like scores. The analysis was performed on the TCGA cohort. (H) Forest plot representing a multivariate analysis of established predictors of disease recurrence (T- stage, Gleason grade), MSI-like similarity score, and pseudo-time associated with the main trajectory. (I) Representative pictures of tumors with low and high expression of CD11c by immunohistochemistry in the stroma and corresponding HE stainings. (J) Kaplan Meier estimates PSA-free survival of primary prostate cancer tissues stratified by CD11 c quantification as determined by IHC. (CD11c low, less than 1.5% of positive cells; CD11c high, more than 1.5% of positive cells). (K) Corresponding forest plot representing multivariate analysis of established predictors of disease recurrence (T-stage, Gleason grade), and CD11c quantification (high/low). Significance level (q- values): ^* < 0.05, ^** < 0.01, ^*** < 0.001.

Figure 4. Characterization of Molecular Features Related to the Main Trajectory. (A) Graphical representation of the RNA sequencing cohorts, their accession numbers, the total number of samples in each dataset, and tumor stages as indicated. (B) Position of individual tumors in the PCA after re-processing of the raw data by selecting the top 2000 most variable genes. Hybrid capture- based RNA sequencing samples derived from CPRC highlighted in light blue show a marked but consistent shift in the PC1 and PC2. No significant differences are observed in the first two principal components for TotaIRNA when compared to PolyA+ samples. (C) Gene-sets enrichments performed using Camera algorithm on genes ranked according to their relative contribution (coefficient) to the positioning of samples along the PC1 axis. The analysis performed on Hallmark gene sets reveals an increase of cell cycle-related gene sets along PC1. (D) Corresponding analysis performed on genes ranked according to their contribution to PC2 shows a decrease in androgen- responsive genes along this axis. (E) PCA plot representing the PC1/PC3 pane can be used to discern SPOP/FOXA1 mutant prostate cancers from those harboring gene fusions involving ETS transcription factors. (F) Gene set enrichment analysis performed on genes ranked for their Pearson’s coefficient as determined by the correlation between mRNA expression and pseudo-time inferred from the main trajectory. Increasing pseudo-time results in an increase of cell cycle-related genes and concomitant down-regulation of androgen-responsive genes. (G) EZH2 mRNA expression increases gradually along the main trajectory. Expression levels of each sample are reported within the PCA plot representing the PC1/PC2 pane. Gene expression levels are scaled between -1 and 1 and are represented in a three-color scale (blue: lowest value; white: median value; red: highest value). (H) Schematic representation of gene expression changes in AR-regulated target genes related to cell differentiation and proliferation and PRC2 components along the main trajectory. Correlation coefficients between mRNA expression and pseudo-time are depicted in a three-color scale (blue: -1 ; white: 0; red: +1 ). (I) Schematic representation of expression changes in genes linked to the tumor environment. Transcripts specific to M1 -macrophages decrease along the main trajectory while those of the M2 counterpart increase. (J) Histograms depicting the correlation between the inferred abundance of the indicated immune cell populations (as determined by Cibersortx) and pseudo-time. P-values associated with Pearson’s correlation coefficients were adjusted for multiple testing using the false discovery rate (FDR). (K) IHC analysis reveals upregulation of EZH2 in CRPC tumors compared to the matched primary tumors. Left: Quantification of EZH2 positive cells, Right: IHC images of a primary and its corresponding CRPC counterpart. (L) Pearson’s coefficients, as determined from the correlation between somatic mutations (0: wild-type; 1:non-synonymous mutation) and inferred pseudo-time along the main trajectory. To dissect the relative impact on disease progression at different stages, coefficients were computed separately in primary and CRPC/NEPC samples. The analysis was performed only for genes mutated at least in 6 individuals. Significance level (p-values): ^* < 0.05, ^** < 0.01 , ^*** < 0.001.

Figure 5. Molecular Characterization of Metastatic Prostate Cancers. (A) Hierarchical clustering of CRPC and NEPC samples identifies metastatic prostate cancer subtypes characterized by different AR activity levels: AR-HIGH, AR-LOW, AR-negative NEPC, double-negative prostate cancers (DNPC), and prostate cancers with transcriptional features of microsatellite instability (MSI-like). AR negative Prostate cancers, including both NEPC and DNPC were clustered into a single group due to over-imposable PCA-positioning. Hierarchical clustering was performed using pvclust. (B) AR mRNA expression level of each sample is reported within the PCA plot representing the PC1/PC2 pane. Gene expression levels are scaled between -1 and 1 and are represented in a three-color scale (blue: lowest value; white: median value; red: highest value). (C) Neuroendocrine enrichment score (NE-Score (Bluemn et a!., 2017)) computed with ssGSEA and depicted within the PCA plot (PC1/PC2). Signature enrichment levels are scaled between -1 and 1 and are represented in a three- color scale (green: lowest value; white: median value; violet: highest value). (D) Volcano plot showing mRNA expression changes comparing MSI-like to the AR-HIGH subgroup. Several key genes related to inflammation and androgen signaling are highlighted. (E) Corresponding pathway analysis performed on hallmark gene set collection, computed with Camera (pre-ranked). (F) Stacked bar charts representing patients and their number of metastases across the 3 identified CRPC tumor subgroups (AR-HIGH; MSI-like;NE/DN-PC). Samples were retrieved from the WCM-cohort, containing individuals with multiple metastatic sites. A larger number of metastasis has been observed for MSI-like patients. (G) PCA plot highlighting the biopsy/resection site for each tissue sample. (H) Stacked bar chart representing the distribution of metastatic sites across the different CRPC subgroups. (I) Absolute quantification of the indicated immune cell population using Cibersortx stratified by CRPC subgroups. P-values were determined using the Wilcoxon sum-rank test and adjusted for multiple comparisons using the false discovery rate (FDR). (J) MSI-like tumors express higher levels of the indicated endogenous retroviral elements when compared to primary and CRPC samples with high mutation burden (i.e. mutation count > 100). Left: Repeated sequences are quantified together as a single entity, and their amount is normalized to the coding fraction of the transcriptome. Right: Separated quantification of RNAs transcribed from Alu-sequences, L1/L2 LINE, and LTR/ERV1 elements. Abundance is expressed in the form of their relative percentage to the total number of repeated sequences. (K) MSI-like tumors express higher levels of the indicated immune checkpoint ligands when compared to primary and CRPC samples with high mutation burden (i.e. mutation count > 100 Significance was assessed using Wilcoxon sum-rank test and p- values were adjusted for multiple comparisons using false discovery rate (FDR): ^* < 0.05, ^** < 0.01 , ^*** < 0.001.

Figure 6. Molecular Features of MSI-like Prostate Cancers. (A) Primary prostate cancers were stratified for MSI-like similarity score (i.e. CD11b/c expression) using a 4-tiered scoring system. (TOP50: The top 50 patients characterized by the highest score; UP-INT/LW-INT : patients with upper and lower intermediate scores, composed of 198 and 199 samples respectively. BTM50: The 50 patients characterized by the lowest MSI-like similarity score). (B) Correlation between single sample GSEA scores computed for Hallmark gene sets and MSI-like similarity scores across primary tumors. Gene-sets are ranked according to their correlation coefficient to MSI-like similarity scores. (Left: direct correlation; Right: inverse correlation). Mostly correlated are inflammatory-related gene sets. (C) Boxplots depicting AR protein abundances across primary tumors stratified according to their MSI-like similarity score. The top 50 primary tumors (TOP50) endowed with higher scores show slightly lower levels of AR protein abundance. (D) Boxplots show no differences between the mutational load of primary tumors when stratified for MSI-like similarity scores. (E) Frequency of somatic point mutations across primary tumors occurring in SPOP, FOXA1 , CDK12, BRCA2, TP53, RB1, and PTEN, stratified by MSI-like similarity score. (F) Right: MSI-like tumors show significant enrichment for KMD6A amplifications (Amp). Left: Among deep gene deletions, multiple genes related to 10q11.21 reach significance. (G) MSI-like CRPC samples show higher mRNA levels of the H3K27 demethylase KMD6A when compared to other CRPC subtypes. (H) Corresponding analysis for the H3K27 methyltransferase EZH2 reveals lower levels in MSI-like CRPC samples. (I) RNA interference-mediated silencing of KMT2D in PC3 prostate cancer cells upregulates the expression of endogenous retroviral elements. Left: Quantification of all endogenous retroviral elements; Right: Quantification of LTR/ERV1 subfamily only. (J) Kaplan-Meier estimates for disease- free survival stratified for pseudo-time using a 4-tiered scoring system in analogy to the MSI-like score (TOP50; UP/LW-INT; BTM50). Curves reveal a significant association of higher pseudo-time with poor disease-free survival. (K) Higher magnification of HE and CD11c IHC (bar represents 20 mM) shows positive immune infiltrate in the stroma while tumor cells (indicated by arrowhead) remain negative. Significance was assessed using the Wilcoxon sum-rank test and p-values were adjusted for multiple comparisons using false discovery rate (FDR): ^* < 0.05, ^** < 0.01 , ^*** < 0.001.

Figure 7A-C. Prostate Cancer Transcriptome Atlas identifies a subgroup of MSI-like CRPCs that connect primary to end-stage metastatic NEPC/AR-negative prostate cancers in an alternative way, (samples highlighted by circle in the different panel figures). Figure 8. Transcriptional analysis of paraffin-embedded tissue samples of a patient who responded for 3 years to pembrolizumab (ICB). The tumor samples of the primary tumor correspond to an MSI- like prostate cancer that follows the alternative trajectory. A tumor nodule known to be resistant to ICB was successfully treated by radiotherapy. Biopsies from this latter nodule cluster with the tumor samples of the main trajectory (see Fig. 7). The data is in line with the notion that ICB responders follow the alternative trajectory of MSI-like prostate cancers, while tumors within the main trajectory (Fig. 7B) do not.

GLOSSARY

In the context of the present description, the term “microsatellite instability” (MSI) refers to a condition characterized by mutations in repetitive DNA sequence tracts, caused by a failure of the DNA mismatch repair system to correct these errors. Deficient DNA mismatch repair (dMMR) results from the bi-allelic mutational inactivation or epigenetic silencing of any of the genes in the MMR pathway (most commonly MSH2, MSH6, MLH1 , and PMS2).

In the context of the present description, the term “microsatellite instability-like cancer” (MSI-like cancer) refers to cancers that are characterized by a number of MSI-like features. Examples of these MSI-like features may include one or more, preferably all the following features:

Cardinal feature: positioning of the transcriptional output (based on the 2000 most robust and variable expressed genes) within the alternative trajectory in the center of the PCA plot and clustering of the output within the group of MSI-like tumors (see Figs. 2A, 7); infiltration by inflammatory cells; upregulation of immune checkpoint ligands; significant increase in the expression of DNA microsatellites and immune pathways; pronounced upregulation of viral-related pathways such as interferon-gamma and TNF-alpha signaling and expression of endogenous retroviral elements; high expression of CD11 b (ITGAM) and/or CD11c (ITGAX) mRNA and protein levels; high mRNA expression of a larger 36-gene set signature (see below); negative DNA test for MSI (absence of real MSI);

Enrichment of loss-of-fu notion mutations in chromatin-modifying enzymes (KMT2C, KMT2D) and gene amplifications in KMD6A;

Low expression EZH2 and DNMT3A & DNMT3B.

In the context of the present description, the terms “upregulation”, “significant increase”, “enrichment”, “high expression” and “low expression” as mentioned above, are used to indicate the expression, in a subject affected by MSI-like cancer, of the specific markers and/or biomolecules to which they refer with respect to the expression of the same markers and/or biomolecules in a healthy subject, unless specified otherwise.

In the context of the present invention, the expression “responsive to ICB therapy”, is used to indicate those subjects suffering from any type of cancer who are susceptible (i.e. not resistant) to a therapeutic treatment comprising the administration of at least one immune checkpoint inhibitor.

In the context of the present description, “about” refers to the experimental error that can occur during conventional measurements. More particularly, when referring to a value it indicates ± 5% of the indicated value.

DETAILED DESCRIPTION OF THE INVENTION

The authors of the present invention have identified a panel of signature markers that enable surprisingly accurate and reliable identification of those subjects suffering from cancer who are affected by microsatellite instability-like prostate (MSI-like) cancer and/or can benefit from ICB therapy, without the need for invasive techniques.

Hence, the present invention refers to an in vitro method for identifying and/or selecting those subjects suffering from cancer who are affected by microsatellite instability-like cancer and/or are responsive to immune checkpoint blockade therapy, said method comprising: a. determining and/or quantifying the expression levels of the signature markers CD11 b (ITGAM) and/or CD11c (ITGAX) in a biological sample isolated from said subjects.

In one embodiment, the present invention specifically refers to an in vitro method for identifying and/or selecting those subjects suffering from cancer who are responsive to immune checkpoint blockade therapy, said method comprising: a. determining whether said subjects are affected by microsatellite instability-like cancer (MSI-like cancer).

According to the present invention, the biomarkers allowing an identification of MSI-like tumors include CD11 b (ITGAM) e CD11c (ITGAX). Alternatively, a larger 36-gene mRNA signature may be used as well to determine whether a subject’s tumor is characterized by MSI-like features by using any of the methods known in the art, as will be explained below.

Notably, determining and/or quantifying the expression levels of the 36-gene mRNA signature represents a simplified method to detect the cardinal feature MSI-like tumors, which is the positioning of the transcriptional output measured by RNA sequencing (based on the 2000 most robust and variable expressed genes) within the alternative trajectory in the center of the PCA plot (see Figs. 2A, 7). Thus, the most accurate method to determine the MSI-like status of a given tumor is to perform RNA sequencing followed by integration of the sample into the PCA plot (see Figs. 2A, 7).

As clearly shown in the experimental section of the present specification, the inventors have found that genes encoding for proteins CD11 b (ITGAM gene) and CD11 c (ITGAX gene) are uniquely upregulated within castration-resistant MSI-like tumors. The acronyms CD11 b and CD11 c, as used in the present invention, refer to surface makers on various cell types, including but not restricted to: dendritic cells, monocytes, macrophages, granulocytes, T cells, NK cells, and B cells. CD11b is an integrin alpha M protein (ITGAM), while CD11c is an integrin alpha X chain protein (ITGAX). Integrins are heterodimeric integral membrane proteins composed of an alpha chain and a beta chain. Both CD11b and CD11c proteins interact with integrin subunit beta 2, ITGB2, (CD18) to build up the complement receptors 3 and 4.

In one embodiment, said CD11 b marker has the aminoacidic sequence set forth in SEQ ID N. 1 and the corresponding ITGAM gene encoding for CD11 b protein has the nucleotide sequence set forth in the Esembl reference sequence ENSG00000169896, and/or said CD11 c marker has the aminoacidic sequence set forth in SEQ ID N. 2 and the corresponding ITGAX gene encoding for CD11c protein has the nucleotide sequence set forth in the Esembl reference sequence ENSG00000140678.

In one embodiment, said CD11c marker is the integrin alpha-X isoform 1 precursor having the NCBI reference sequence N P_001273304.1 , and/or is the integrin alpha-X isoform 2 precursor having the NCBI reference sequence NP_000878. In one embodiment, said CD11b marker is the integrin alpha-M isoform 1 precursor having the NCBI reference sequence NP_001139280 and/or is the integrin alpha-M isoform 2 precursor having the NCBI Reference Sequence NP_000623.2.

The method according to the present invention may also include the determination and/or quantification of the expression levels of additional signature markers, which were identified by the inventors as being upregulated in MSI-like tumors.

Hence, in one embodiment, step a. of the above method further comprises determining and/or quantifying the expression levels of one or more, preferably all the signature markers (previously referred as 36-gene signature) selected from the group consisting of: BIRC3, C5AR1 , CD300LB, CEACAM3, CIITA, CLEC4D, CLEC5A, CSF3R, DOCK2, FGR,

FLT3, FOLR3, GPR84, HCK, IL18RAP, IL24, IL7R, ITGAM, ITGAX, ITK, JAK3, KLRD1 , LY9, MMP25, MNDA, NCF2, NLRC4, NLRP3, PGLYRP1 , PSTPIP1 , RAB44, SIGLEC14, SIGLEC9, SLA, STAT4, TNFAIP3.

In some embodiments, the above-mentioned markers have nucleotide sequences as set forth in the following Ensembl reference sequences: BIRC3: ENSG00000023445; C5AR1 : ENSG00000197405; CD300LB: ENSG00000178789; CEACAM3: ENSG00000170956; CIITA: ENSG00000179583; CLEC4D: ENSG00000166527; CLEC5A: ENSG00000258227; CSF3R: ENSG00000119535; DOCK2: ENSG00000134516; FGR: ENSG00000000938; FLT3:

ENSG00000122025; FOLR3: ENSG00000110203; GPR84: ENSG00000139572; HCK: ENSG00000101336; IL18RAP: ENSG00000115607; IL24: ENSG00000162892; IL7R:

ENSG00000168685; ITGAM: ENSG00000169896; ITGAX: ENSG00000140678; ITK:

ENSG00000113263; JAK3: ENSG00000105639; KLRD1 : ENSG00000134539; LY9:

ENSG00000122224; MMP25: ENSG00000008516; MNDA: ENSG00000163563; NCF2:

ENSG00000116701 ; NLRC4: ENSG00000091106; NLRP3: ENSG00000162711; PGLYRP1 : ENSG00000008438; PSTPIP1: ENSG00000140368; RAB44: ENSG00000255587; SIGLEC14:

ENSG00000254415; SIGLEC9: ENSG00000129450; SLA: ENSG00000155926; STAT4:

ENSG00000138378; TNFAIP3: ENSG00000118503.

In one particular embodiment, step a. of the above method comprises determining and/or quantifying the mRNA expression levels of one or more of the above-defined 36 signature markers by using any methods known in the art for the assessment of mRNA expression. In one embodiment, step a. of any of the methods of the present invention may further include the detection of one or more MSI-like specific features as previously mentioned in the present specification, such as: the expression of immune checkpoint ligands, endogenous retroviral elements, DNA satellites in the absence of MSI, and/or MSI-like specific mutations in genes like KMT2C, KMT2D, KMD6A.

In one aspect, the in vitro method of the invention or step a. of the method according to any of the embodiments disclosed herein further comprises one or more of the following steps:

Determining and/or quantifying the levels of infiltration of the biological sample by inflammatory cells; and optionally comparing said levels with at least one reference value; wherein said subjects are identified and/or selected as subjects who are affected by MSI-like cancer and/or are responsive to ICB therapy when said levels are higher than said at least one reference value.

Merely by way of example, quantification of immune cells infiltrates can be inferred from transcriptomic data derived from the biological sample obtained from the subject to be selected or identified, using any analytical tool available in the art, such as, for example, CibersortX.

Determining and/or quantifying the expression levels of one or more immune checkpoint ligands selected from the group consisting of PD-1 ligands and CTLA-4 ligands; and optionally comparing said expression levels with at least one reference value; wherein said subjects are identified and/or selected as subjects who are affected by MSI-like cancer and/or are responsive to ICB therapy when said expression levels are higher than said at least one reference value.

In one embodiment, PD-1 ligands comprise PDL-1 and PDL-2 ligands, while CTLA-4 ligands comprise CD80 and CD86 ligands.

Determining and/or quantifying the expression levels of DNA microsatellites; and optionally comparing said expression levels with at least one reference value; wherein said subjects are identified and/or selected as subjects who are affected by MSI-like cancer and/or are responsive to ICB therapy when said expression levels are higher than said at least one reference value.

Determining and/or quantifying the expression levels of endogenous retroviral elements; and optionally comparing said expression levels with at least one reference value; wherein said subjects are identified and/or selected as subjects who are affected by MSI-like cancer and/or are responsive to ICB therapy when said expression levels are higher than said at least one reference value.

Determining whether said subject tests negative for microsatellite instability. In one embodiment, such step comprises extracting DNA from the biological sample of the subject to be selected or identified followed by DNA sequencing and analysis to assess somatic mutations (using a biological sample from a healthy subject as a comparator) and mutational burden, as well as MSI-specific mutations in microsatellites.

Determining the levels of loss-of-function mutations in genes encoding for chromatin interacting proteins or chromatin-modifying enzymes, in particular the histone methyltransferases KMT2C and KMT2D; and optionally comparing said levels with a reference value, wherein said subjects are identified and/or selected as subjects who are affected by MSI-like cancer and/or are responsive to ICB therapy when said levels are higher than said at least one reference value.

Determining the levels of amplifications in KDM6A; and optionally comparing said levels with a reference value, wherein said subjects are identified and/or selected as subjects who are affected by MSI-like cancer and/or are responsive to ICB therapy when said levels are higher than said at least one reference value.

Determining and/or quantifying the expression levels of EZH4, DNMT3A and DNMT3B; and optionally comparing said levels with a reference value, wherein said subjects are identified and/or selected as subjects who are affected by MSI-like cancer and/or are responsive to ICB therapy when said levels are lower than said at least one reference value.

In one preferred embodiment, the in vitro method according to the present invention comprises all the above-mentioned additional steps.

Suitable biological samples that could be used to determine and/or quantify the expression levels of any of the signature markers mentioned above are biological samples isolated from subjects suffering from cancer. Said subjects may include subjects that have been diagnosed with cancer; in one preferred embodiment, said subjects are subjects suffering from cancer, such as prostate cancer patients, preferably that have not been diagnosed with MSI.

Hence, the methods, uses, and/or kits according to any of the embodiments as described in the present specification and in the claims may be advantageously used to identify and/or select those subjects suffering from any type of cancer who are affected by MSI-like cancer and/or are responsive to ICB therapy.

Preferably, the methods, uses, and/or kits of the present invention may be applied to subjects suffering from hormone-related cancer types, such as breast, endometrium, ovary, prostate, testis, thyroid cancer and/or osteosarcoma. In one preferred embodiment, the methods, uses and/or kits of the present invention may be applied to subjects suffering from prostate cancer. Non-limiting examples of suitable biological samples include body fluid samples or tissue samples, preferably a biopsy from the tumor tissue of the subject such for example a tumor tissue or cells related to tumor tissue. In one preferred embodiment, said biological sample is a tissue sample obtained from a resected tumor, preferably a tumor tissue of primary, recurrent or metastatic origin. For mRNA expression analysis of any of the 36 genes mentioned above, a fresh frozen or OCT (optimal cutting temperature compound) embedded tissue sample is preferred. Total RNA or PolyA enriched RNA is the preferred input material for subsequent analysis of gene expression by any of the methods known in the art (e.g. quantitative PCR, nanostring technology, RNA sequencing using next-generation technologies of any kind). Formalin-fixed, paraffin-embedded tissue material that is routinely harvested in the clinic may be used to assess the corresponding protein expression (most notably CD11 b/c), for example by means of immunohistochemistry.

Other suitable biological samples that can be used in any of the methods of the present invention include samples of exosomes isolated from subjects’ biofluids such as blood. Exosomes are a subtype of extracellular vesicles that range in size from approximately 40 to 160 nm in diameter, which has been found to be robustly produced and secreted by tumor cells. Exosomes carry membranous and cytoplasmic substances derived from their parental cells, such as proteins, messenger RNAs, microRNAs, long non-coding RNAs, lipids, metabolites, and even DNA fragments.

Exosomes may be found in multiple bodily fluids, including blood, lymph, urine, cerebrospinal fluid, bile, saliva, and milk (among other). Exosomes can be isolated from non-exosomal components using a number of techniques known in the field, such as ultracentrifugation (UC), filtration, size- exclusion chromatography (SEC), immunoaffinity capture, and microchip-based techniques. Furthermore, many kinds of commercial kits are available for exosome isolation.

In the context of the present invention, the wording “determining and/or quantifying the expression levels” refers to detecting the presence and/or measuring the expression levels, in either absolute or relative terms, of any of the markers or signature markers as defined in the present description and in the claims, according to any of the methods available to the skilled in the art. In one embodiment, in any of the methods described herewith the determination and/or quantification of the expression levels of any of said markers or signature markers can be performed by means of an in vitro test. Non limiting examples of in vitro tests suitable for determining and/or quantifying the expression levels of any of the above-mentioned markers include an immunological assay, an aptamer-based assay, a histological or cytological assay, an RNA expression levels assay or a combination thereof. All tests indicated above are known to a person skilled in the art who, knowing which signature marker has to be determined and/or quantified and the type of biological sample used, will be able to select the most suitable protocol.

In one embodiment, the methods according to any of the embodiments of the present invention comprise determining and/or quantifying any of the above signature markers by means of mRNA expression assays, which can be performed using any of the techniques known in the art, and/or a protein expression assay performed by means of immunohistochemistry. In one preferred embodiment, determining and/or quantifying the expression levels of any of the above-mentioned markers can be specifically performed by means of immunohistochemistry using agents that are capable of selectively targeting any of the above signature markers, such as, for example, an antibody. Merely by way of example and not for limitative purposes, the expression levels of any of the above markers may be determined and/or quantified in a tissue sample of prostate tumor obtained from a subject to be analyzed by means of immunohistochemistry, wherein said prostate tumor tissue is incubated with an antibody targeting any of the above markers.

In one embodiment, the method according to the present invention may further comprise the following steps: b. comparing the expression levels as determined and/or quantified in step a. with at least one reference value; and c. identifying and/or selecting said subjects based on said comparison.

Suitable reference values that can be used when implementing any of the methods according to the present invention comprise a reference value corresponding to the expression levels of the marker or signature marker to which it is to be compared, as determined and/or quantified in a biological sample isolated from subjects having microsatellite-instability (MSI)-like cancer, particularly subjects that have already been diagnosed as being affected by MSI-like cancer.

An alternative suitable reference value can be a reference value corresponding to the expression levels of the marker or signature marker to which it is to be compared, as determined and/or quantified in a biological sample isolated from subjects who are not responsive to ICB therapy, for example subjects suffering from primary, recurrent or metastatic tumors that are not responsive to ICB therapy.

In one preferred embodiment, step b. of said methods comprises comparing the expression levels as determined and/or quantified in step a. with (i) a first reference value corresponding to the expression levels of said signature marker CD11b (ITGAM) and/or with (ii) a second reference value corresponding to the expression levels of said signature marker CD11c (ITGAX), wherein said expression levels (i) and/or (ii) are determined and/or quantified in a biological sample isolated from subjects suffering from MSI-like cancer (e.g. subjects that have already been diagnosed as being affected by MSI-like cancer).

As previously mentioned, any of the methods according to the present invention enable an accurate and reliable identification and/or selection of those subjects suffering from cancer who are affected by MSI-like tumors and/or are responsive to ICB therapy.

In one embodiment, in step c. of any of the methods described in the present specification, said subjects are identified and/or selected as subjects suffering from cancer with MSI-like features and/or subjects who are responsive to ICB therapy when said expression levels as determined and/or quantified in step a. are about equal or higher than said at least one reference value.

An alternative suitable reference value that can be used when implementing any of the methods according to the present invention is a reference value corresponding to the expression levels of the signature marker to which it is to be compared, as determined and/or quantified in a biological sample isolated from a healthy subject. In this case, subjects can be identified and/or selected as subjects suffering from cancer with MSI-like features and/or subjects who are responsive to ICB therapy when the expression levels of the signature markers as determined and/or quantified in step a. of any of the above methods are higher than said at least one reference value obtained from a healthy subject.

According to a further embodiment of the invention, step a. of the above method further comprises conducting a complete transcriptome analysis, in particular a complete transcriptome sequencing, on said biological sample. As used herein, the term “transcriptome” is referred to the set of all RNA transcripts, including coding and non-coding, in a biological sample obtained from said subject according to any of the embodiments disclosed herein.

In one aspect of the invention, the in vitro method according to any of the embodiments disclosed in the present specification and in the claims further includes any of the following steps, preferably all the following steps:

- analyzing transcriptome data derived from a plurality of healthy individuals as well as from individuals suffering from cancer at different stages of disease progression, particularly from primary, castration-resistant and neuroendocrine prostate cancer, so as to generate or compute a principal component analysis (PCA) plot such as that represented in Figure 1A or 7A; performing trajectory and/or pseudotime inference analysis so as to identify a main trajectory and at least one alternative trajectory with respect to disease progression based on at least one part of said transcriptome data; determining, based on at least the expression levels or on the transcriptome analysis as determined and/or quantified, or conducted in step a. of any of the methods disclosed herein, whether the biological sample obtained from the subjects to be selected and/or identified localize within said alternative trajectory; and identifying and/or selecting said subjects as subjects who are responsive to ICB therapy and/or are affected from MSI-like cancer, when the biological sample obtained from the subjects to be selected and/or identified localizes within said alternative trajectory.

According to a preferred aspect of the invention, transcriptome analysis, the generation of the PCA plot, and trajectory and/or pseudotime analysis can be performed by means of any of the methodologies disclosed in the experimental section of the present specification, particularly as disclosed in the “method details” section.

In one preferred embodiment, transcriptome analysis, the generation of the PCA plot, trajectory and/or pseudotime analysis can be performed by means of the methodologies disclosed in the scientific publication of M. Bolis et al. “Dynamic prostate cancer transcriptome analysis delineates the trajectory to disease progression”, Nat. Comm. (2021) 12:7033, herein incorporated by reference.

A further object of the present invention relates to an in vitro method according to any one of the embodiments as previously described, further comprising the following steps: b’. calculating a score from the expression levels as determined and/or quantified in step a.; and c’. identifying and/or selecting said subjects based on the score calculated in step b’.

In one embodiment, said score can be calculated either by averaging the expression of ITGAX and ITGAM to consider said averaged value as a metagene, or by performing single-sample gene-set enrichment analysis. In this case, ITGAX and ITGAM can be used as signatures whose activity needs to be quantified. Said process can be obtained in R statistical environment using gsva package (Hanzelmann S, Castelo R, Guinney J (2013). “GSVA: gene set variation analysis for microarray and RNA-Seq data.” BMC Bioinformatics, 14, 7). The same criteria can be applied to determine the score from the extended 36-gene signature.

Merely by way of example and not for limitation purposes, said score can be determined from RNA-Seq data, provided that gene-expression are adjusted for library size and normalized with an appropriate algorithm such as variance stabilizing transformation or a simple logarithmic transformation. Said score can be computed either by using a single-sample gene-set enrichment analysis approach, or by averaging the expression values of either the 2 or 36 gene signatures into a single meta-gene. In one embodiment, in said step c’., said subjects are identified and/or selected as subjects affected by MSI-like cancer and/or subjects who are responsive to ICB therapy when said score/expression value is typically within the top 10% of a heterogenous larger-sized data set comprising at least 50 or at least 100 patients or more. Of note, an absolute value might be dataset and methodology dependent. That said, such a value may be provided after the establishment and validation of marketable test kit.

Another aspect of the invention refers to an in vitro method according to any one of the previous embodiments, said method comprising the following steps: a. determining the expression levels of the signature marker CD11c (ITGAM) and/or CD11b (ITGAX) in a resected tumor tissue isolated from said subjects by means of immunohistochemical staining; b’. calculating a score from the expression levels as determined in step a., c’. identifying and/or selecting said subjects based on the score calculated in step b’. One preferred embodiment refers to an in vitro method according to any one of the previous embodiments, said method comprising the following steps: a. determining the expression levels of the signature marker CD11c (ITGAM) and/or CD11b(ITGAX) in a resected tumor tissue isolated from said subjects by means of immunohistochemical staining; this can be performed regardless of the specific cell type in which the markers are expressed; b’. calculating a score from the expression levels as determined in step a., c’. identifying and/or selecting said subjects based on the score calculated in step b’, wherein subjects whose tumor tissues typically exhibit at least three percent of CD11c-positive stromal cells are identified and/or selected as subjects affected by MSI-like prostate cancer and/or subjects who are responsive to immune checkpoint blockade (ICB) therapy. Of note, the cut-off of three percent mentioned above may vary slightly and depend on the specific tissue preservation conditions (e.g. formalin fixation, cold and hot ischemia time) and the specific setting of the immunohistochemistry (e.g. antigen retrieval, buffer compositions, antibody type and concentration, detection methodology).

Any of the methods as described in the present specification may also include, when said subjects are identified and/or selected as subjects affected by MSI-like cancer and/or subjects who are responsive to immune checkpoint blockade therapy, a further step consisting in prescribing a treatment with an immune blockade inhibitor to said subjects.

Forms part of the present invention also the in vitro use of the signature markers CD11b (ITGAM) and/or CD11c (ITGAX) according to any of the embodiments as defined in the present specification, for identifying and/or selecting those subjects suffering from cancer who are affected by MSI-like cancer and/or are responsive to immune checkpoint blockade (ICB) therapy.

The present invention also relates to a kit for identifying and/or selecting those subjects suffering from cancer who are affected by microsatellite instability-like (MSI-like) cancer and/or are responsive to immune checkpoint blockade (ICB) therapy, said kit comprising one or more agents for determining and/or quantifying the expression levels of the signature markers CD11b (ITGAX) and/or CD11c (ITGAM) in a biological sample isolated from said subjects.

According to one embodiment, said kit might further comprise one or more agents for determining and/or quantifying the expression levels of any of the additional signature markers as previously defined in the present specification. Moreover, said kit can comprise one or more control reagents together with a list of instructions.

The invention further relates to a computer-implemented method for identifying and/or selecting those subjects suffering from cancer who are affected by microsatellite instability-like (MSI- like) cancer and/or are responsive to immune checkpoint blockade (ICB) therapy, the method comprising: a. receiving at, at least, one processor, input data representing the expression levels of the signature markers CD11 b (ITGAM) and/or CD11 c (ITGAX) in a biological sample isolated from said subjects; b. computing at, at least, one processor, a score using said input data.

The computer-implemented method according to the present invention can further comprise identifying and/or selecting said subjects based on said score. The expression levels of the signature markers CD11b (ITGAM) and/or CD11c (ITGAX) to be implemented in the above method, can be obtained according to any one of methods and using any of the biological samples as previously described in the present specification.

In one embodiment, said input data can further include data representing the expression levels of one or more signature markers selected from the group consisting of: BIRC3, C5AR1 , CD300LB, CEACAM3, CIITA, CLEC4D, CLEC5A, CSF3R, DOCK2, FGR, FLT3, FOLR3, GPR84, HCK, IL18RAP, IL24, IL7R, ITGAM, ITGAX, ITK, JAK3, KLRD1 , LY9, MMP25, MNDA, NCF2, NLRC4, NLRP3, PGLYRP1 , PSTPIP1 , RAB44, SIGLEC14, SIGLEC9, SLA, STAT4, TNFAIP3. In one preferred embodiment, said CD11b (ITGAM) marker has the aminoacidic sequence set forth in SEQ ID N. 1 and/or said CD11 c (ITGAX) marker has the aminoacidic sequence set forth in SEQ ID N. 2. According to a further embodiment, the input data can also include data representing the levels or expression levels of any of the additional markers as disclosed in the present specification and in the claims.

In one embodiment, the expression levels of any of the signature markers to be implemented in the computer-implemented method according to the present invention, are determined and/or quantified in a biological sample isolated from said subjects by means of the execution of an in vitro test selected from the group consisting of: an immunological assay, an aptamer-based assay, a histological or cytological assay, an RNA expression levels assay or a combination thereof. As previously mentioned, in one preferred embodiment said subjects suffering from cancer are subjects suffering from hormone-related cancer, more preferably are subjects suffering from prostate cancer. The present invention further relates to a computer program comprising instructions which, when the program is executed by a computer, cause the computer to carry out the steps of the computer-implemented method as described in the present specification.

Another object of the present invention is referred to an immune checkpoint inhibitor for use in the treatment of a subject suffering from microsatellite instability-like (MSI-like) cancer. One preferred embodiment of the present invention is referred to an immune checkpoint inhibitor for use in the treatment of a subject suffering from microsatellite instability-like (MSI-like) cancer, wherein said immune checkpoint inhibitor is selected from the group of chemical or molecular/biological approaches consisting of targeting: PD1 and its ligands (PDL1/PDL2) and CTLA-4 and its ligands (CD80/CD86). Currently available immune checkpoint inhibitors on the market are Ipilimumab (Bristol-Myers Squibb), Tremelimumab (Astra Zeneca), Nivolumab (Bristol-Myers Squibb), Pembrolizumab (Merck), Cemiplimab (Sanofi), Spartalizumab (Novartis), Atezolizumab (Roche), Durvalumab (Astra Zeneca), and Avelumab (Merck).

Finally, the present invention also relates to an immune checkpoint inhibitor for use in a method of treating a subject suffering from MSI-like cancer, and preferably prostate cancer. In one embodiment of the present invention the method comprises: (i) determining whether said subject is affected by MSI-like cancer by using any of the in vitro methods as previously described in the present specification;

(ii) if, in step (i), said subject is determined as being affected by MSI-like cancer, administering to the subject an effective amount of said immune checkpoint inhibitor. In one preferred embodiment, step (i) comprises determining the levels or the expression levels of any of the markers as disclosed in the present specification and in the claims. In particular, step (i) can include determining whether said subject tests negative for microsatellite instability by carrying out, for example, any of the methodologies disclosed in the present specification.

Another aspect of the present invention is referred to an immune checkpoint inhibitor for use in a method of treatment of a subject suffering from cancer, in particular prostate cancer, wherein said method comprises the following steps:

One preferred embodiment of the present invention specifically refers to a method of treatment of a subject suffering from cancer, said method comprising the following steps:

(i’) determining whether a biological sample isolated from said subject exhibits an increased expression of the signature markers CD11b (ITGAM) and/or CD11c (ITGAX) with respect to a sample of a healthy subject; and

(ii’) if, in step (i’), said biological sample exhibits such increased expression, administering to said subject an effective amount of said immune checkpoint inhibitor.

Another object of the present invention relates to an immune checkpoint inhibitor for use in a method of treatment of a cancer, in particular prostate cancer, expressing the signature markers CD11 b (ITGAM) and/or CD11 c (ITGAX), more preferably of a cancer exhibiting an up-regulation of CD11 b (ITGAM) and/or CD11 c (ITGAX).

Any of the methods of treatment mentioned above may further include additional steps comprising determining whether said subjects suffering from cancer exhibits one or more additional MSI-like features such as those described in the present specification. In one embodiment, any of the methods of treatment disclosed herein may further include additional steps comprising determining and/or quantifying the levels or expression levels of any of the markers as disclosed in the present specification and in the claims. Said method of treatment can particularly include determining whether said subject tests negative for microsatellite instability by carrying out, for example, any of the methodologies disclosed in the present specification.

According to one preferred aspect, any of the methods of treatment as disclosed in the present specification and in the claims includes one additional step, prior to administration, comprising determining whether the biological sample isolated from said subject localizes within the alternative trajectory as determined according to any of the methodologies disclosed in the present specification.

The immune checkpoint inhibitors according to any of the embodiments described in the present specification may be used in a method of treatment of a subject suffering from hormone- related cancer, particularly of a subject suffering from prostate cancer, breast cancer, endometrium cancer, ovary cancer, testis cancer, thyroid cancer and/or osteosarcoma, particularly wherein any of the above cancer types is characterized by one or more MSI-like features.

One further aspect of the present invention is referred to a pharmaceutical composition comprising a checkpoint inhibitor for use in any of the methods of treatment herein disclosed. Said pharmaceutical composition may further comprise a pharmaceutically acceptable carrier and/or excipient.

Examples are reported below which have the purpose of better illustrating the methodologies disclosed in the present description, such examples are in no way to be considered as a limitation of the previous description and the subsequent claims.

EXAMPLES

Example 1 - Generation of the Prostate Cancer Transcriptome Atlas

To nominate gene expression changes related to disease progression, RNA sequencing (RNAseq) data from a wide collection of thirteen different sequencing studies were combined into a large pan-prostate cancer transcriptome meta-analysis (Fig. 4A) (Abida et al., 2019b; Beltran et al., 2016; Consortium, 2013; Kumar et al., 2016; Labrecque et al., 2019; Lapuk et al., 2012; Robinson et al., 2015; Sharp et al., 2019; Stelloo et al., 2018; Suntsova et al., 2019). To this end, the raw data of 174 normal prostate tissue samples, 714 primary tumors, 316 CRPC, and 19 neuroendocrine prostate cancers (NEPC) were reprocessed and merged. As evidenced by an unbiased principal component analysis (PCA) an appreciable transcriptional “batch effect” related to hybrid capture sequencing was detected and subsequently corrected (Fig. 4B, see method for details).

The resulting PCA showed that samples’ position at a given disease stage largely overlapped with another regardless of their origin, while samples from different disease stages differed in localization (Fig. 1A). Gene set enrichment analysis (GSEA) of the first two principal components (PC) revealed that the PC1 strongly correlated with enhanced proliferation (i.e. E2F-targets, G2M- checkpoint, mitotic spindle) while PC2 mainly anti-correlated with canonical AR signaling (Fig. 4C & D). Moreover, PC3 separated cancers harboring recurrent truncal mutations in SPOP & FOXA1 from the ones harboring gene fusions involving ETS family transcription factors, most notably ERG (Fig. 4E). The findings are in line with previous reports highlighting major genetic and biological differences between these prostate cancer subtypes (Barbieri et al., 2012) (Cancer Genome Atlas Research, 2015) (Shoag et al., 2018) (Bernasocchi et al. 2021, biorxiv doi: 10.1101/2020.07.08.193581). PC4 and the following principal components represented a minor amount of the total variance (less than 4 %) and could not be associated with any major disease- related clinical or genetic features (data not shown).

Example 2 - Trajectory analysis identifies the main path to disease progression

In order to mine the prostate cancer transcriptome atlas, a framework was developed, termed prostate cancer profiler (PCaProfiler, https://www.pcaprofiler.com), and trajectory analysis was applied to the atlas to identify and quantify the roadmap to disease progression. The analysis identified the main trajectory and allowed the assignment of a pseudo-time that describes the advancement along this path. The latter indicated that the majority of cancers derive from normal tissue by gradually increasing AR signaling (PC2) and by augmenting expression of cell cycle genes (PC1), then eventually progress to CRPC by further increasing cell cycle genes and finally reach dedifferentiation with and without neuroendocrine trans-differentiation (NEPC) by a subsequent reduction in AR signaling (PC2) (Fig. 1B, 4F). Among the most up-regulated genes along this main trajectory, there were multiple genes related to the polycomb repressive complex 2 (PRC2) - an important complex that mediates gene silencing during development (Gaytan de Ayala Alonso et al., 2007). Most importantly, EZH2 was the top upregulated gene in line with multiple earlier reports, suggesting an important role in disease progression (Fig. 1C, 4G, H) (Varambally et al., 2002; Xu et al., 2012; Yu et al., 2010). Additional members of the polycomb complex (i.e. EED, SUZ12) were also upregulated along with the DNA methylating enzymes DNMT 1 and DNMT3A/B (Fig. 1 C), suggesting a fundamental play of epigenetic regulation.

AR signaling promotes under physiological settings, both cell differentiation and cell proliferation During prostate cancer tumorigenesis and disease progression, AR preferentially binds to genes related to cell cycle progression (Pomerantz et al., 2015; Pomerantz et al., 2020; Wang et al., 2009). Indeed, known AR-regulated genes that promote G2-M transition were among the top up-regulated genes, while canonical AR-target genes related to cellular differentiation were downregulated (Fig. 1C, 4H). It has been widely appreciated during recent years that cancer growth is supported by changes in the tumor microenvironment, such as the polarization of macrophages from an M1- towards M2-like phenotype (Di Mitri et al., 2019; Kowal et al., 2019). Indeed, along the main trajectory, a downregulation of M1 markers and an upregulation of markers of M2 polarization and associated pro-tumorigenic effectors such as PDGFB, VEGF, and MMP8 was observed (Fig. 1C, 4I, J). Interestingly, CD24 - a potent “don’t eat me” signal for M1 macrophages on tumor cells was upregulated as well (Barkal et al., 2019).

Subsequently, in order to investigate whether transcriptional changes along the trajectory would also translate into differences in protein levels, transcriptional differences between primary prostate cancers and CRPC in the atlas were correlated with corresponding protein expression changes of an independent cohort of primary and CRPC samples (Iglesias-Gato et al., 2018). There was a highly significant correlation between mRNA and protein level changes of these robustly expressed genes, even though the transcriptional and protein expression data were generated in different patient samples (Fig. 1D). Because EZH2 was not measured in the proteomic data set due to technical reasons, an upregulation of EZH2 protein expression with disease progression was ascertained on a tissue microarray of 33 primary and matched CRPC samples (Fig. 4K) (Federer-Gsponer et al., 2020).

Finally, it was asked whether genomic alterations in driver genes correlated with transcriptional features related to disease progression. In primary tumors, a positive correlation with MYC copy number status and an inverse correlation with deletions of RB1 , PTEN, and TP53, were noted as expected. In contrast, in CRPC/NEPC samples, only RB1 loss seemed to correlate well with increased transcriptional progression (Fig. 1 E-G). The latter finding is in line with a recent report showing that RB1 alterations correlate with poor survival (Abida et al., 2019a). Furthermore, a significant correlation of point mutations in PIK3CA, TP53, FOXA1 , KMT2C, and PTEN with transcriptional progression in primary tumors was noted as well (Fig. 4L). The findings underscore the interconnection of genetic alterations in cancer driver genes and the transcriptional changes associated with the main trajectory of disease progression.

Example 3 - Alternative trajectory characterized by MSI-like features

While the above unbiased analysis revealed the main path towards disease progression, some CRPC samples were noticed to be positioned distant from the main trajectory in the inner part of the PCA plot with a loose connection to the samples following the main circular path (Fig. 1A and Fig. 7A). Hierarchical clustering of CRPC and NEPC samples revealed different subgroups that differed in AR expression levels, as previously reported (Beltran et al., 2016; Bluemn et al., 2017; Kumar et al., 2016; Labrecque et al., 2019). The largest subgroup consists of tumors that had adapted to the low availability of androgens by a compensatory up-regulation of the AR itself and was positioned in the early phase of the main trajectory (termed AR-HIGH, Fig. 2A, 5A). Tumors with low (termed AR- LOW) and subsequently loss of AR expression with (NEPC) and without features of neuroendocrine trans-differentiation (termed double negative prostate cancer, DNPC) followed the progression trajectory, as expected (Fig. 2A & 5A-C). Surprisingly, the tumors positioned in the inner part of the circular trajectory consisted of a distinct subgroup of tumors that connect primary and NEPC/DNPC samples in a straight fashion (Fig. 1A, 2A). In other terms, these tumors aggregated into a distinct alternative trajectory that connects primary and NEPC samples in a straight fashion (see also Fig. 7A-C).

These tumors were characterized by transcriptional features reminiscent of deficiency in the DNA mismatch repair system and microsatellite instability (herein referred to as MSI-like, Fig. 2B). Indeed, the latter also included a significant increase in the expression of DNA microsatellites and immune pathways (Fig. 2C & 5D, E). Notably, previous studies reported prognostic value for several mismatch repair genes involved in MSI (Prtilo et al., 2005). However, the mutational load appeared to be significantly reduced compared to the remaining metastatic samples, excluding the possibility of real MSI (Fig. 2D). Patients within this alternative trajectory tend to have a more wide-spread metastatic disease with a slight preference towards visceral organs (Fig. 5F-H). That said, all organ types were represented among the MSI-like samples, indicating that the central positioning within the PCA plot was not determined by organ- specific cells. On average, MSI-like tumors were more infiltrated by immune cells, and macrophages tend to be polarized towards M1 (Fig. 2E, F & 5I), which contrasts to the increased M2 polarization along the main trajectory. In line with the more pronounced immune infiltration, a significant increase in the expression of immune checkpoint ligands (i.e., PDL-1, PDL-2, CD80, CD86) was also noticed (Fig. 2G). The latter finding suggested that MSI-like tumors could respond equally well to immune checkpoint inhibitors as prostate cancers with confirmed MSI (Abida et al., 2019a).

It was hypothesized that the increase of DNA microsatellites might result from increased expression of endogenous retroviral elements - a process that is known to induce an immune response referred to as viral mimicry (Deblois et al., 2020; Roulois et al., 2015). Indeed, a pronounced upregulation of viral-related pathways such as interferon-gamma and TNF-alpha signaling and expression of endogenous retroviral elements was observed in MSI-like tumors (Fig. 2H, I). The increased expression of endogenous retroviral elements and checkpoint ligands remained apparent even compared to primary and non-MSI-like CRPC samples with high mutation load (i.e., > 100 mutations, Fig. 5J, K). In aggregate, the data suggests that a subset of MSI-like prostate cancers with features of viral mimicry and higher expression of checkpoint ligands take a transcriptionally distinct path towards disease progression.

Example 4 - MSI-like primary cancers are rapidly progressing and frequently harbor mutations in chromatin-modifying enzymes

In order to identify primary cancers that could serve as a potential starting point for the alternative trajectory, a transcriptional signature was derived that was specific to this trajectory by searching for genes that are uniquely upregulated in castration-resistant MSI-like tumors when compared to primary tumors, AR-HIGH, and NEPC/DNPCs (see Method part for detailed description.) To uncover key biological features within the signature, common protein-protein interactions were also interrogated for. Interconnected, differentially expressed genes, belonged mainly to a network of proteins related to the innate immune response (Fig. 3A & 6B). Strikingly, the top enriched transcripts within the network were encoding for the myeloid marker CD11 b (ITGAM) and the M1 -macrophage marker CD11c (ITGAX). Both proteins interact with CD18 to build up the complement receptors 3 and 4, respectively (Fig. 3B).

Thus, in order to determine which primary prostate cancers were resembling the most to the castration-resistant MSI-like tumors, the joint expression of CD11 b and CD11 c (herein referred to as MSI-like similarity score) was evaluated using a four-tiered scoring system (Fig. 6A). As expected, tumors with the highest score were enriched in signatures related to innate and viral immunity pathways and complement activation (Fig. 6B). Besides, a slightly lower AR protein abundance was observed in line with the lower expression of AR target genes observed in MSI-like metastatic tumors and no increase in mutation burden (Fig.6C, 6D).

Given these similarities, it was asked if primary tumors characterized by a high MSI-like similarity score would give rise to the previously described MSI-like metastatic counterparts. In support, this subset of primary tumors clustered and partially overlapped with metastatic MSI-like tumors at the starting point of the alternative trajectory (Fig. 3B). A unique dataset was subsequently interrogated for MSI-like features, which also included some rare longitudinal RNA sequencing of primary tumors and their matched CPRC samples (Yun et al., 2017). All samples with low MSI-like scores were positioned along the main trajectory, and the matched primary/CRPC tumor samples of patients 1 , 3, 4 also progressed following this path (Fig.3C). In contrast, primary, metastatic, and tumor samples with high MSI-like similarity scores, as expected, were positioned in the inner part of the PCA plot and displayed an increased expression of immune checkpoint ligands (Fig. 3C, D). The trend observed for samples from patient 2 supported the notion that MSI-like tumors progress from primary to metastatic prostate cancer along an alternative trajectory.

The authors wondered if primary tumors characterized by high MSI-like similarity scores are enriched for specific genetic alterations that could explain viral mimicry. It was not found a significant enrichment in point-mutations or copy number variations of well-established driver genes linked to disease progression such as TP53, PTEN, or RB1. In contrast, several genes encoding for chromatin-interacting proteins were enriched in MSI-like resembling primary tumors (Fig. 3E, F & 6E). Among the copy number changes, a significant enrichment of gains for KMD6A, an H3K27 histone demethylase enzyme involved in chromatin remodeling was found (Fig. 6F). Interestingly, KMD6A mRNA expression was also elevated in metastatic MSI-like samples, while its molecular counteractor EZH2 (an H3K27 methyltransferase) was significantly decreased in the same sample set (Fig. 6G, H). This deeply differs from the behavior observed for the main trajectory, where EZH2 levels continuously increase during progression.

The most striking enrichment for point-mutations was found for KMT2C and KMT2D, two related histone methyltransferases, and candidate tumor suppressor genes in prostate cancer (Fig. 3E) (Lv et al., 2018) (Cancer Genome Atlas Research, 2015). A recent report indeed suggests that loss-of- function of KMT2D results in chromatin remodeling, activation of endogenous retroviral elements, and increased immune infiltration in cancer (Wang et al., 2020). In line with this, published RNA sequencing data revealed that the knockdown of KMT2D in PC3 prostate cancer cells leads to an upregulation of various types of endogenous retroviral elements, suggesting a functional link between genetic alterations in KMT2D, activation of retroviral elements, and inflammation (Fig. 61) (Lv et al., 2018). Despite this evidence, we cannot exclude that the transcriptional program characterizing MSI-like tumors might result from additional genetic/epigenetic alterations or a combination of events.

Given the notion that MSI-like samples seem to have a more wide-spread metastatic burden at the time-point of biopsy, the inventors wondered if MSI-like resembling primary tumors may relapse faster after the initial surgery. Indeed, univariate analysis revealed that the disease-free survival was significantly shorter with increasing MSI-like similarity score (Fig. 3G). The inventors noted that this score correlated only weakly with the Gleason grade and the tumor stage - the two most important predictors of disease progression in primary tumors (data not shown). In line with this, the MSI-like similarity score emerged as an independent prognostic marker in the multivariate analysis (Fig. 3H). In contrast, the pseudo-time inferred from the main trajectory was more tightly related to Gleason grade and tumor stage and thus scored only in the univariate but not in the multivariate analysis (Fig. 3H & Fig. 6J).

Next, the inventors further validated their results on tissue microarrays of different patients (n=482). For this purpose, immunohistochemistry for CD11 c was performed because its corresponding mRNA emerged as the most specific marker for MSI-like tumors. Using a two-tiered scoring system, primary tumors with abundant CD11c expression in immune/stromal cells relapsed significantly faster after surgery (Fig. 3I, J and Fig.6K). The effect was also independent of Gleason grade and tumor stage (Fig. 3K). Altogether, the data suggest that a subset of MSI-like tumors follows an alternative, more rapidly progressing trajectory compared to other tumors that are positioned along the main path.

METHOD DETAILS Immunohistochemically staining

For the assessment of CD11 c protein expression in primary prostate cancer and the correlation with PSA-recurrence patients were selected from two previously characterized tissue microarrays cohorts constructed in Zurich and Bern ((Cyrta et al., 2020; Groner et al., 2016; Spahn et al., 2010). Due to tissue loss, a common problem associated with TMA technology, a total of 482 high-quality tissue samples of primary tumor remained after sectioning (n = 272 from Zurich and n = 210 from Bern). In each case, the local scientific ethics committees approved (StV-Nr. 25/2007 and StV-Nr. 25-2008) and informed consent was obtained from all patients. Recurrence-free survival curves were calculated using the Kaplan-Meier method. Patients were censored at the time of their last tumor- free clinical follow-up visit. Time to PSA recurrence was selected as the clinical endpoint. Only patients undergoing radical prostatectomy were used for survival analysis.

For CD11c IHC, slides were analyzed with the Bond-Ill automated staining system (Leica) using manufactured reagents for the entire procedure. For antigen retrieval, slides were incubated for 20 min in Citrate buffer at pH6 at 98°C. Thereafter, slides were incubated with a rabbit anti-CD11c antibody targeting the C-terminus (ab52632) at the dilution of 1:1000 for one hour at room temperature. Detections were performed using the detection refine DAB kit (Leica). Immunohistochemical staining was evaluated with the automated Aperio ImageScope (Leica) image quantification system using a two-tiered score, i.e. tumor spots with at least three percent of CD11 c- positive cells were classified as CD11c high, while the remaining cases were classified as CD11c low. In the case of the Bern cohort, multiple spots per tumor were available and the percentage of CD11 c-positive tumor cells was established based on the average of two spots displaying the highest Gleason pattern. RNA Extraction for RNA-seq analysis

According to the manufacturer’s guidelines, the RNA extraction was performed from PDXs frozen fragment (25-30 mg) of cellular pellet using RNeasy kit (74106 Qiagen). The RNAs were processed using the NEB Next Ultra II Directional Library prep Kit for lllumina (E7765 NEB) and sequenced on the lllumina NextSeq500 with single-end, 75 base pair long reads.

Prostate Cancer Transcriptome Atlas

To build an integrated resource of transcriptional features representing all stages of prostate cancer progression, we collected raw sequencing data from a large panel of independent datasets. We gathered raw data for 1223 clinical samples (1104 excluding technical replicates, 1044 excluding multiple metastatic sites derived from the same individual). The resulting integrated cohort is representative of various stages of disease progression, namely, normal prostate specimens (n=174), primary tumors (n=714), castration-resistant prostate cancers (n=316), and castration- resistant prostate cancers showing features of neuroendocrine trans-differentiation (n=19). Raw sequencing files were retrieved from following sources: 1) Gene Tissue Expression Database (GTEX); 2) The Cancer Genome Atlas (TCGA); 3) Atlas of RNA sequencing profiles of normal human tissues (GSE120795); 4) Integrative epigenetic taxonomy of primary prostate cancer (GSE120741); 5) Prognostic markers in locally advanced lymph node-negative prostate cancer (PRJNA477449); 6) The Long Noncoding RNA Landscape of Neuroendocrine Prostate Cancer and its Clinical

Implications (PRJEB21092); 7) Integrative Clinical Sequencing Analysis of Metastatic Castration Resistant Prostate Cancer Reveals a High Frequency of Clinical Actionability (PRJNA283922; dbGaP: phs000915); 8) CSER - Exploring Precision Cancer Medicine for Sarcoma and Rare Cancers (PRJNA223419; dbGaP: phs000673); 9) Molecular Basis of Neuroendocrine Prostate Cancer (PRJNA282856; dbGaP: phs000909); 10) Heterogeneity of Androgen Receptor Splice Variant-7 (AR-V7) Protein Expression and Response to Therapy in Castration Resistant Prostate Cancer (CRPC) (GSE118435); 11) Molecular profiling stratifies diverse phenotypes of treatment- refractory metastatic castration-resistant prostate cancer (PRJNA520923; GEO: GSE126078). Depending on the specific dataset considered, fastq files were downloaded either by using gdc-client (TCGA) or sra-toolkit (SRA, dbGaP). RNA-seq data processing of clinical samples

The overall quality of sequencing reads was evaluated using FastQC (Andrews S., 2010). Sequence alignments to the reference human genome (GRCh38) were performed using STAR (v.2.6.1c) in two-pass mode. Gene-expression was quantified at the gene level by using the comprehensive annotations made available by Gencode (v29 GTF-File). Strand specific information were not maintained to avoid technical differences between stranded and unstranded libraries. Samples were adjusted for library size and normalized with the variance stabilizing transformation (vst) in the R statistical environment using DESeq2 (v1.28.1) pipeline. When performing differential expression analysis between groups we applied the embedded IndependentFiltering procedure to exclude genes that were not expressed at appreciable levels in most of the samples considered. If not otherwise specified, all gene set enrichment analyses were performed using the limma package (Camera, use. ranks set to TRUE) (Wu and Smyth, 2012). Gene-Sets collections were retrieved either from the Molecular Signature Database (MSigDB), or from previous publications (AR/NE- Score) (Bluemn et al., 2017).

Batch effects correction and Principal Component Analysis

In the processes of integrating different datasets from a variety of sources, we verified that batch effects did not overwhelm the biological signal. Batch effects may derive not only from differences across datasets, but also may be consequent of a different sequencing technique (PolyA+; TotaIRNA; Hybrid Capture Sequencing) or originate from other unknown sources. Principal component analysis (PCA), by identifying the transcriptional features endowed with the highest variance across samples, is a very useful tool to detect relevant batch effects. When the latter are overwhelming, they are likely to appear among the top principal components and cluster together samples sharing the same batch effect-related features. A PCA analysis performed on the complete set of 1223 samples (Figure 4B) showed that the largest source of batch effects was associated with the Hybrid Capture Sequencing technique (HCS), while no relevant differences could be clearly associated with the dataset of origin. Only two of the CRPC datasets (phs000915, phs000673) contained samples sequenced using HCS, and for several of these, matched technical replicates sequenced using PolyA+ technology were also available. This allowed us to assess and remove technology associated bias in gene expression (ComBat, PolyA+ samples set as reference batch). We further reduced the possibility of confounding biological with technical variation by generating a training-subset of our data, consisting of 883 PolyA+ samples (52 Normal prostate, 620 Primary tumors, 193 CRPCs, 19 NEPCs) and determined the top 2000 genes showing the highest amount of variation within the PolyA+ training set only. This way, for PCA representation we avoid the selection of genes that are possibly affected by the sequencing technique, despite the correction we had already performed on the data.

Hence, we used the same 2000 genes to generate a PCA plot computed on the extended set of samples. The results depicted in the main PCA plot shown in Figurel A clearly show that positioning of tumors at the same stages of cancer progression overlap to each other irrespectively of the dataset of origin and of the sequencing technology. This indicates that the different positioning of normal prostate, primary tumors, CRPCs, and NEPCs is due to a real biological signal and not consequent to an unwanted dataset-specific batch effect.

Integration and validation of additional bulk RNA-Seq samples and pseudo-time inference

We developed a method to include new prostate tumor samples in our current analysis by starting from raw counts, which allows the computation of pseudo-time and Principal components without modifying the original data and plots. Ideally, RNA-Seq should be quantified using the sample genome (hg38) and references used for the current study (Gencode V29). Predictions can be performed sequentially, one sample at a time. For each new sample of interest, raw counts will be merged with the ones composing our full set. The obtained numeric matrix (the original matrix + 1 extra sample of interest) undergoes the same normalization and processing steps up to the computation of the PCA. Here, coordinates may slightly differ from the original ones, due to the adding of a new sample which might exert a small effect on the global re-normalization of all samples. To address this behavior, we apply a machine learning-based approach that generates at runtime three elastic net models, one for each of the top 3 principal components, and train them to predict the error between the original coordinates and ones that are recomputed following the addition of the extra sample of interest. Hence, we apply these models to adjust the computed PC1 , PC2 and PC3 coordinates of the extra sample which can now be added to the main PCA plot and pseudo time can be determined using slingshot. Trajectory analysis

Trajectory and pseudo-time inference are frequently used in single-cell RNA sequencing data analysis to model developmental trajectories through smooth curves following dimensionality reduction and clustering. Here we applied one of these tools, slingshot (v1.6.0), to infer progression- associated trajectory and pseudo-time from our integrated set of bulk-RNA sequencing samples. We selected slingshot because of its capability to also determine branches along the trajectory if any. PCA positioning (PC1-PC2) of the individual samples was used as input for slingshot, along with the information that the computed trajectory had to start from the Normal tissue cluster. The analysis was performed using 1106 samples, discarding all technical replicates, in order not to overweight some samples and influence the computation of the trajectory. Metastatic lesions from the same individual but localized in different organs were admitted for this analysis. Subsequently, we could associate a pseudo-time for each sample, ranging from 0 to 250 (Figure 1B). Correlation of genes and pathways to pseudo-time

Having defined a unique pseudo-time value for each sample, we computed the correlation between pseudo-time and mRNA expression for each gene. For this purpose, we used Pearson’s correlation over Spearman’s because we aimed at identifying the strength of the linear relationship between gene expression and pseudo-time. However, to be more robust to outliers, we opted for 10 times repeated leave one third out procedure. Precisely, we randomly selected 10 subsets composed of 66% of the samples and computed correlation coefficients between pseudo-time and expression of each gene in all subsets. Finally, we averaged these values and ranked them according to their correlation coefficient to pseudo-time. Subsequently, using this ranking we applied Camera to perform gene-set enrichment analysis procedure (use. ranks = TRUE) and determined which gene- set were mostly directly or inversely associated with pseudo-time (Figure 4F).

Correlation of mRNA expression and protein abundances

Proteomics data were retrieved from the Proteomics Identifier Database (PRIDE: projects PXD009868, PXD003430, PXD003452, PXD003515, PXD004132, PXD003615, PXD003636). The dataset includes 28 gland confined prostate tumors and 8 adjacent non-malignant prostate tissue obtained from radical prostatectomy procedures, plus 22 bone metastatic prostate tumors obtained from patients operated to relieve spinal cord compression. To compute the correlation between mRNA expression and protein abundance we first computed, for each gene, the average Fold- change (log2) between CRPC and PRIMARY tumors based on mRNA expression. Then the same was applied to the proteomics data to obtain for each protein a log fold change representing differential abundance between CRPCs and primary tumors. For protein/mRNA correlation purposes, we discarded all genes that had not been evaluated in the proteomic data. Finally, we used Pearson’s method to evaluate the strength of correlation and the associated statistical significance.

Retrieval of genetic information and correlation with progression

Matched genetic information respective to mutations and copy number status could be retrieved for 763 samples through cBioportal. To determine associations between mutations and tumor progression, for each gene we compared the pseudo-time of mutant vs wild-type samples, by performing statistical testing using the Wilcoxon-sum rank test. Mutations were ordered according to their False Discovery Rate adjusted P-values and analyses were performed separately in PRIMARY and CRPC+NEPC tumors, to determine the relative contribution of mutations at various stages of disease progression. We only screened for genes being mutated in more than 5 individuals (Figure 4L). To determine associations between copy-number alterations and tumor progression, we associated for each gene a value of either -2 (homozygous deletion), -1 (heterozygous deletion), 0 (Wild-Type), 1 (Gain), 2(Amplification) and subsequently computed Pearson’s correlation between these values and pseudo-time. We restricted this last analysis to genes being frequently deleted or amplified in prostate tumors, namely, MYC, AR, RB1 , PTEN, and TP53 (Figure 1 E). The above- described analyses were performed discarding technical replicates. Metastatic lesions from the same individual but localized in different organs were admitted for this analysis.

Quantification of immune infiltrates and correlation with progression

Quantification of immune infiltrates for all samples in our cohort was inferred from transcriptomic data using CibersortX (Steen et al., 2020) by using the default signature matrix "LM22" to deconvolve 22 immune cell subsets from bulk RNA-Seq (Absolute quantification mode). The abundance of inferred immune populations was correlated to pseudo-time using the same strategy applied to correlate gene-expression and pseudo-time. We opted for 10 times repeated leave one third out procedure. Precisely, we randomly selected 10 subsets composed of 66% of the samples and computed correlation coefficients between pseudo-time and each immune population in all subsets. Finally, we averaged these values and ranked them according to their correlation coefficient to pseudo-time. Pearson’s correlation associated P-Values were corrected for multiple testing using the False Discovery Rate (FDR).

Macrophage Polarization Index

The Macrophage Polarization Index, indicating polarization towards M1 or M2 phenotypes was computed for all bulk-RNA samples in our cohort using MacSpectrum (Li et al., 2019). Quantification of retroviral transcripts

Expression of endogenous retroviral elements is not usually quantified in conventional RNA-Seq analysis, as their genomic loci are frequently located outside of coding exons and are repeatedly distributed over the entire genome. To quantify the abundance of these repeated sequences from RNA-Seq, we developed a custom pipeline. First, we retrieved their genomic annotations and their respective positioning from the RepeatMasker[ref] database. Using tools-intersect (v.2.29, -v flag), we discarded all repeated sequences that may have overlapped to known exons or UTRs. Subsequently, we generated a custom GTF file containing annotations and genomic coordinates for exactly 5350312 genomic loci. The most overrepresented families of repeats resulted to be SINE/Alu (n= 1205088); LINE/L1(n= 982906); SINE/MIR (n= 581903); LINE/L2(n= 461231); LTR/ERVL-MaLR (n= 348724); DNA/hAT-Charlie (n= 256265); LTR/ERV1(n= 175723); LTR/ERVL(n= 163031).

Sequencing reads were aligned to hg38 reference genome using STAR (v2.6.1c) by applying the following flags for the alignment procedure: (-outFilterMultimapNmax 100 winAnchorMultimapNmax 100 --alignlntronMax 1 --aligned type EndToEnd outFilterMismatchNmax 3). Subsequently, all reads mapping to any of the repeated features were quantified using featureCounts (v2.0.1, subread package) and for downstream analysis, we either summed all counts originating from any of the repeats or stratified them into the respective families. Repeats were normalized for the library size of each sample and were expressed in form of a ratio between the number of reads mapping to repeats and the number of reads mapping to protein- coding genes.

Hierarchical clustering of CRPC samples and definition of subgroups

Unsupervised hierarchical clustering was performed in the R statistical environment, using Euclidean distance measure and average agglomerative method. Input matrix consisted of vst-normalized (DESeq2) expression values of CRPC and NEPC samples. P-values for clusters in the dendrogram were assessed with pvclust (R package, https://cran.r-project.org/web/packages/pvclust/index.html), with 1000 bootstrap replications for resampling. We identified 5 clusters of samples which are represented in Figure 5 A. Four of these are positioned along the main trajectory of the PCA plot (Figure2 A) and can be associated with increasing pseudo-time. The largest cluster (AR-HIGH) is composed of samples showing lower pseudo-time and higher AR activity. The second cluster (AR- LOW) shows intermediate pseudo-time along the main trajectory, with low levels of AR pathway activity. Subsequently, proceeding along the main trajectory, we find two clusters that are both characterized by extremely low AR signaling which is typical of double negative prostate cancer (DNPC) and neuroendocrine tumors. Indeed, one out of the two groups is particularly enriched for NEPCs. These two clusters share a similar localization on the PCA plot, are located at the end of the main trajectory, and are thus characterized with similar pseudo-time. Hence, we grouped them into a single group (NEPC/DNPC). Finally, there is a remaining cluster that does not fit well to the main trajectory and appears to be composed of samples that localize within the center of the PCA plot, ranging from primary to neuroendocrine tumors. MSI-status predictions performed using PreMSIm (Li et al., 2020), classified the majority of samples composing this last cluster as MSI-High. However, due to the lack of the characteristic high mutational load, typical of microsatellite unstable tumors, we classified this particular group of prostate cancers as MSI-like.

Comparisons between CRPC subgroups Differential expression between clusters was performed using DESeq2. P-values were adjusted for False Discovery Rate and genes expressed at very low levels were discarded through the IndependentFiltering procedure provided by the algorithm. Expression of Immune checkpoints and their ligands as depicted in boxplots was obtained from vst-normalized data (DESeq2). Pathway activity levels for immune-related genesets (Interferon Gamma Response, TNF-Alpha Signaling, Inflammatory Response) were computed using single sample gene-set enrichment analysis (GSVA vl.36.2). Statistical significance between groups was assessed using the Wilcoxon sum rank test. P-Values were adjusted for multiple comparisons using the False Discovery Rate (FDR). Absolute levels of immune infiltrate were quantified using CibersortX (Absolute Score) and compared between clusters. The associated statistical significance was assessed using the Wilcoxon sum rank test. P- Values were adjusted for multiple comparisons using the False Discovery Rate (FDR). Repeated sequences were quantified as previously described and expressed in form of repeats/coding ratio if not otherwise specified. Statistical significance was assessed using the Wilcoxon sum rank test. P- Values were adjusted for multiple comparisons using the False Discovery Rate (FDR).

Definition of MSI-like marker genes and computation of a similarity score in primary tumors

To identify genes overexpressed specifically in MSI-like tumors, we selected features that resulted to be differentially expressed with a log2Fold-Change > 1 in all the following comparisons: MSI-like vs Normal; MSI-like vs Primary Tumors; MSI-like vs AR-HIGH; MSI-like vs DN/NEPC. The analysis resulted in the selection of 52 genes. Subsequently, we used these genes to generate a protein- protein interaction network on Cytoscape using StringDB. We removed nodes showing no interconnections to other genes in the network and reduced the list to 36 elements. The two most interconnected genes in the network were ITGAX (CD11c) and ITGAM (CD11b), which complex with ITGB2(CD18) to form respectively complement receptor 4 and 3. Expression of these two genes was used then to generate an MSI-like similarity score. To this purpose, we used a single sample gene set enrichment analysis to assess the combined expression of these two marker genes, and then correlate them to disease-free survival (DFS) in primary tumors.

Survival analysis of RNA-Seq data Survival analysis was performed on primary tumors for whom this type of information was available (TCGA-cohort). Disease-free survival (DFS) was used as a clinical outcome. Kaplan-Meier curves were generated in the R statistical environment (R packages: survival v3.2.3; survminer vO. 4.7). Primary tumor samples were stratified based on MSI-like similarity score levels as described in the figure legends. The multivariate analysis was performed using the cox-proportional hazard model.

Quantification of gene expression

Fastq files were generated by demultiplexing raw data using cellranger mkfastq (v3.1.0) To make single-cell gene-expression quantification more comparable to those of bulk RNA-Seq, we generated a custom genome with cellranger more, using the very same reference (GRCh38.p12) and annotations (encode v29).used for STAR when performing bulk RNA-Sequencing analysis. To discriminate between human and murine cells that may infiltrate the tumors in the in vivo setting, we created a Mouse-Human reference, by creating a hybrid genome (GRCh38.p12+GRCm38.p6) and hybrid gene-annotations (gencode v29 and M25, for human and mouse genes respectively). To avoid conflicts, mouse genomic coordinates were preceded by a prefix (i.e. mm_chr1 , mm_chr2, etc.). Subsequently, cellranger count was used to quantifying gene-expression in form of an h5 filtered matrix where Ensembl gene IDs are used as identifiers.

Data filtering and clustering Expression quantification files were imported in R statistical environment using Seurat (v3.1.5) package. We discarded individual cells from our data matrix by using two filtering procedures: first, we aimed at detecting transcriptional outliers, second, we looked for putative doublets, which we also discarded. Briefly, we computed per-cell quality control metrics using scatter (v1.16.1). The total amount of mitochondrial and ribosomal gene expression was quantified for both human and mouse cells. The number of genes being detected per cell, the total amount of reads per cell, and the mitochondrial and ribosomal fraction of the transcriptome were used to determine the skewness- adjusted multivariate outlyingness for each cell (robustbase vO.93-6). Outliers were detected by median absolute deviation (MAD) and removed at both tails. Counts were then normalized (Seurat::NormalizeData, method = LogNormalize, scale.factor = 1000) and the top 2000 most variable features were selected (Seurat::FindVariableFeatures, method = vst). Data were then scaled (Seurat::ScaleData) and principal component analysis was performed up to the top 50 components (Seurat::RunPCA). Subsequently, we identified and eliminated putative doublets using DoubletFinder (v2.0.3). Having identified outliers and doublets, we removed them from the original count data and went through the pre-processing step again (i.e. normalization, scaling, and pca- reduction). We proceeded to the determination of the k-nearest neighbors of each cell and the construction of a Shared Nearest Neighbor (SNN) Graph (Suerat::FindNeighbors), then we identified clusters using the shared nearest neighbor (SNN) modularity optimization based clustering algorithm (Seurat:: FindClusters, resolution = 0.5). Finally, we performed Umap dimensionality reduction on the first 10 Principal Components, annotated the previously identified clusters, and generated plots accordingly.

Macrophage Polarization Index of macrophages

The Macrophage Polarization Index, indicating polarization towards M1 or M2 phenotypes was computed for all cells being identified as macrophages from SingleR analysis (https://macspectrum.uconn.edu).

QUANTIFICATION AND STATISTICAL ANALYSIS Quantification methods and statistical analysis methods for ^** were mainly described and referenced in the respective Method Details subsection. If not otherwise specified, all statistical tests were corrected for multiple comparisons using the false discovery rate (FDR) correction method.

Example 5 - MSI-like cancer patient responds to immunotherapy Most patients with metastatic prostate cancer do not respond to ICB. To investigate if MSI-like features may predict response to immunotherapy, tissue samples collected from MSI-negative prostate cancer who responded to ICB have been characterized. These patients are relatively rare because they do not qualify for ICB treatment in daily clinical practice. Tissue samples from one patient who responded for 3 years to ICB were analyzed by RNA sequencing. While the primary tumor clustered clearly within the alternative pathway (Fig. 7, dots marked as “MSI-like primary tumor”) and displayed MSI-like transcriptional features, biopsies of a tumor mass that progressed under ICB (but responded to radiotherapy, dots marked as “recurrent local tumor”) positioned to the main trajectory. The preliminary data suggests indeed that ICB could be an effective treatment option for MSI-like tumors within the alternative trajectory but not tumors of the main trajectory.

Sequence listing part of the description:

SEQ ID N. 1 - Aminoacidic sequence of the signature marker CD11 b

MALRVLLLTALTLCHGFNLDTENAMTFQENARGFGQSWQLQGSRWVGAPQEIVAANQR

GSLYQCDYSTGSCEPIRLQVPVEAVNMSLGLSLAATTSPPQLLACGPTVHQTCSENTYVK

GLCFLFGSNLRQQPQKFPEALRGCPQEDSDIAFLIDGSGSIIPHDFRRMKEFVSTVMEQL

KKSKTLFSLMQYSEEFRIHFTFKEFQNNPNPRSLVKPITQLLGRTHTATGIRKVVRELFN

ITNGARKNAFKILWITDGEKFGDPLGYEDVIPEADREGVIRYVIGVGDAFRSEKSRQEL

NTIASKPPRDHVFQVNNFEALKTIQNQLREKIFAIEGTQTGSSSSFEHEMSQEGFSAAIT

SNGPLLSTVGSYDWAGGVFLYTSKEKSTFINMTRVDSDMNDAYLGYAAAIILRNRVQSLV

LGAPRYQHIGLVAMFRQNTGMWESNANVKGTQIGAYFGASLCSVDVDSNGSTDLVLIGAP

HYYEQTRGGQVSVCPLPRGRARWQCDAVLYGEQGQPWGRFGAALTVLGDVNGDKLTDVAI

GAPGEEDNRGAVYLFHGTSGSGISPSHSQRIAGSKLSPRLQYFGQSLSGGQDLTMDGLVD

LTVGAQGHVLLLRSQPVLRVKAIMEFNPREVARNVFECNDQWKGKEAGEVRVCLHVQKS

TRDRLREGQIQSVVTYDLALDSGRPHSRAVFNETKNSTRRQTQVLGLTQTCETLKLQLPN

CIEDPVSPIVLRLNFSLVGTPLSAFGNLRPVLAEDAQRLFTALFPFEKNCGNDNICQDDL

SITFSFMSLDCLVVGGPREFNVTVTVRNDGEDSYRTQVTFFFPLDLSYRKVSTLQNQRSQ

RSWRLACESASSTEVSGALKSTSCSINHPIFPENSEVTFNITFDVDSKASLGNKLLLKAN

VTSENNMPRTNKTEFQLELPVKYAVYMWTSHGVSTKYLNFTASENTSRVMQHQYQVSNL

GQRSLPISLVFLVPVRLNQTVIWDRPQVTFSENLSSTCHTKERLPSHSDFLAELRKAPW

NCSIAVCQRIQCDIPFFGIQEEFNATLKGNLSFDWYIKTSHNHLLIVSTAEILFNDSVFT

LLPGQGAFVRSQTETKVEPFEVPNPLPLIVGSSVGGLLLLALITAALYKLGFFKRQYKDM

MSEGGPPGAEPQ

SEQ ID N. 2 - Aminoacidic sequence of the signature marker CD11c

MTRTRAALLLFTALATSLGFNLDTEELTAFRVDSAGFGDSVVQYANSWVWGAPQKITAA

NQTGGLYQCGYSTGACEPIGLQVPPEAVNMSLGLSLASTTSPSQLLACGPTVHHECGRNM

YLTGLCFLLGPTQLTQRLPVSRQECPRQEQDIVFLIDGSGSISSRNFATMMNFVRAVISQ

FQRPSTQFSLMQFSNKFQTHFTFEEFRRSSNPLSLLASVHQLQGFTYTATAIQNVVHRLF

HASYGARRDAAKILIVITDGKKEGDSLDYKDVIPMADAAGIIRYAIGVGLAFQNRNSWKE

LNDIASKPSQEHIFKVEDFDALKDIQNQLKEKIFAIEGTETTSSSSFELEMAQEGFSAVF

TPDGPVLGAVGSFTWSGGAFLYPPNMSPTFINMSQENVDMRDSYLGYSTELALWKGVQSL

VLGAPRYQHTGKAVIFTQVSRQWRMKAEVTGTQIGSYFGASLCSVDVDSDGSTDLVLIGA

PHYYEQTRGGQVSVCPLPRGWRRWWCDAVLYGEQGHPWGRFGAALTVLGDVNGDKLTDVV

IGAPGEEENRGAVYLFHGVLGPSISPSHSQRIAGSQLSSRLQYFGQALSGGQDLTQDGLV

DLAVGARGQVLLLRTRPVLWVGVSMQFIPAEIPRSAFECREQVVSEQTLVQSNICLYIDK

RSKNLLGSRDLQSSVTLDLALDPGRLSPRATFQETKNRSLSRVRVLGLKAHCENFNLLLP

SCVEDSVTPITLRLNFTLVGKPLLAFRNLRPMLAADAQRYFTASLPFEKNCGADHICQDN

LGISFSFPGLKSLLVGSNLELNAEVMVWNDGEDSYGTTITFSHPAGLSYRYVAEGQKQGQ

LRSLHLTCDSAPVGSQGTWSTSCRINHLIFRGGAQITFLATFDVSPKAVLGDRLLLTANV

SSENNTPRTSKTTFQLELPVKYAVYTWSSHEQFTKYLNFSESEEKESHVAMHRYQVNNL

GQRDLPVSINFWVPVELNQEAVWMDVEVSHPQNPSLRCSSEKIAPPASDFLAHIQKNPVL

DCSIAGCLRFRCDVPSFSVQEELDFTLKGNLSFGWVRQILQKKVSWSVAEITFDTSVYS

QLPGQEAFMRAQTTTVLEKYKVHNPTPLIVGSSIGGLLLLALITAVLYKVGFFKRQYKEM

MEEANGQIAPENGTQTPSPPSEK

Claims

1. An in vitro method for identifying and/or selecting those subjects suffering from cancer who are responsive to immune checkpoint blockade (ICB) therapy, said method comprising: a. determining and/or quantifying the expression levels of the signature markers CD11b (ITGAM) and CD11c (ITGAX) in a biological sample isolated from said subjects; b. comparing the expression levels as determined and/or quantified in step a. with at least one reference value; and c. identifying and/or selecting said subjects based on said comparison; wherein said subjects are identified and/or selected as subjects who are responsive to ICB therapy when said expression levels as determined and/or quantified in step a. are higher than said at least one reference value.

2. The in vitro method according to claim 1 , wherein said step a. further comprises determining and/or quantifying the expression levels of one or more signature markers selected from the group consisting of: BIRC3, C5AR1 , CD300LB, CEACAM3, CIITA, CLEC4D, CLEC5A, CSF3R, DOCK2, FGR, FLT3, FOLR3, GPR84, HCK, IL18RAP, IL24, IL7R, ITGAM, ITGAX, ITK, JAK3, KLRD1 , LY9, MMP25, MNDA, NCF2, NLRC4, NLRP3, PGLYRP1, PSTPIP1 , RAB44, SIGLEC14, SIGLEC9, SLA, STAT4, TNFAIP3.

3. The in vitro method according to any one of claims 1 or 2, wherein said biological sample is selected from a body fluid sample or a tissue sample, preferably is a tissue sample obtained from a resected tumor of said subject.

4. The in vitro method according to any one of claims 1 to 3, wherein said step a. comprises the execution of an in vitro test selected from the group consisting of: an immunological assay, an aptamer-based assay, a histological or cytological assay, a RNA expression levels assay or a combination thereof.

5. The in vitro method according to any one of claims 1 to 4, wherein said at least one reference value corresponds to the expression levels of the signature marker to which it is to be compared, as determined and/or quantified in a biological sample isolated from subjects that have already been diagnosed as being affected by MSI-like cancer and/or from healthy subjects.

6. The in vitro method according to any one of claims 1 to 5, wherein step a. further comprising one or more of the following steps: determining and/or quantifying the levels of infiltration of the biological sample by inflammatory cells; determining and/or quantifying the expression levels of one or more immune checkpoint ligands selected from the group consisting of PD-1 ligands and CTLA-4 ligands; determining and/or quantifying the expression levels of DNA microsatellites; determining and/or quantifying the expression levels of endogenous retroviral elements; determining whether said subject tests negative for microsatellite instability; determining the levels of loss-of-fu notion mutations in genes encoding for chromatin interacting proteins or chromatin-modifying enzymes; determining the levels of amplifications in KDM6A; determining and/or quantifying the expression levels of EZH4, DNMT3A and DNMT3B.

7. The in vitro method according to claims 1 to 6, wherein said step a. further comprises conducting transcriptome analysis on said biological sample.

8. The in vitro method according to claims 1 to 7, wherein said method further comprises the following steps:

- analyzing transcriptome data derived from a plurality of healthy individuals as well as from individuals suffering from cancer at different stages of disease progression, particularly from primary, castration-resistant and neuroendocrine prostate cancer, so as to generate or compute a principal component analysis (PCA) plot;

- performing trajectory and/or pseudotime inference analysis so as to identify a main trajectory and at least one alternative trajectory with respect to disease progression based on at least one part of said transcriptome data;

- determining, based on at least the expression levels as determined and/or quantified in said step a., whether said biological sample localizes within said alternative trajectory; and

- identifying and/or selecting said subjects as subjects who are responsive to ICB therapy, when said biological sample localizes within said alternative trajectory.

9. The in vitro method according to any one of claims 1 to 8, further comprising the following steps: b’. calculating a score from the expression levels as determined and/or quantified in step a.; and c’. identifying and/or selecting said subjects based on the score calculated in step b’.

10. The in vitro method according to claim 9, wherein said score is calculated by averaging the expression levels of ITGAX and ITGAM and/or of any of said 36 signature markers so as to consider the obtained average expression as a metagene, and/or by performing single sample gene-set enrichment analysis.

11. The in vitro method according to claims 9 or 10, wherein, in said step c’., said subjects are identified and/or selected as subjects who are responsive to ICB therapy when said score corresponds to a value within the top 10% of a heterogenous larger-sized data set comprising at least 100 subjects.

12. The in vitro method according to any one of claims 1 to 11 , said method comprising the following steps: a. determining the expression levels of the signature marker CD11c (ITGAX) in a resected tumor tissue isolated from said subjects by means of immunohistochemical staining; b’. calculating a score from the expression levels as determined in step a.; c’. identifying and/or selecting said subjects based on the score calculated in step b’, wherein subjects whose tumor tissues exhibit at least three percent of CD11c-positive cells are identified and/or selected as subjects who are responsive to immune checkpoint blockade (ICB) therapy.

13. The in vitro method according to any one of claims 1 to 12, wherein said subjects are subjects suffering from hormone-related cancer, preferably are subjects suffering from prostate cancer, breast cancer, endometrium cancer, ovary cancer, testis cancer, thyroid cancer and/or osteosarcoma.

14. Kit for identifying and/or selecting those subjects suffering from cancer who are responsive to immune checkpoint blockade (ICB) therapy, said kit comprising one or more agents for determining and/or quantifying the expression levels of the signature markers CD11b (ITGAM) and CD11c (ITGAX) in a biological sample isolated from said subjects.

15. Computer-implemented method for identifying and/or selecting those subjects suffering from cancer who are responsive to immune checkpoint blockade (ICB) therapy, the method comprising: a. receiving at, at least, one processor, input data representing the expression levels of the signature markers CD11b (ITGAM) and CD11c (ITGAX) in a biological sample isolated from said subjects; b. computing at, at least, one processor, a score using said input data.

16. Immune checkpoint inhibitor for use in the treatment of a subject suffering from microsatellite instability-like (MSI-like) cancer.

17. Immune checkpoint inhibitor for use in a method of treatment of a cancer expressing the signature markers CD11b (ITGAM) and CD11c (ITGAX), more preferably of a cancer exhibiting an up-regulation of CD11b (ITGAM) and CD11c (ITGAX) with respect to a reference sample.

18. Immune checkpoint inhibitor for use in a method of treatment of a cancer in a subject, wherein said method comprises determining if said subject is responsive to immune checkpoint blockade (ICB) by carrying out an in vitro method according to any one of claims 1 to 13.

19. The immune checkpoint inhibitor according to claims 16 to 18, wherein said subject is suffering from hormone-related cancer, preferably is a subject suffering from prostate cancer, breast cancer, endometrium cancer, ovary cancer, testis cancer, thyroid cancer and/or osteosarcoma.