The present invention relates to the field of anticancer treatment. In particular, the present invention concerns the role of the gut microbiota in the efficacy of treatments comprising administration of an immune checkpoint inhibitor (ICI) and/or a tyrosine kinase inhibitor (TKI), in the treatment of cancer. The present invention provides "metagenomics-based gut oncomicrobiome signatures" (GOMS) at diagnosis prior to ICI and/or TKI administration and/or after initiation of the treatment, as novel predictors of response or resistance to the treatment. The present invention also provides theranostic methods to identify patients in need of a bacterial compensation treatment before receiving an ICI and/or TKI and/or during the therapy with such ICI and/or TKI, as well as novel bacterial species appropriate for such a bacterial compensation.

BACKGROUND AND PRIOR ART

Metastatic renal cell carcinoma (RCC) have long been considered as "immunogenic malignancies" susceptible to immunotherapies (Rosenberg et ai, 1993; Escudier et ai, 1994). In this tumor type, the prognostic role of the immune contexture was broadly heralded. In both primary and metastatic RCC, CD38+ tumor associated macrophages, immature dendritic cells (DC), the absence of tertiary lymphoid organs or overt expression of T cell inhibitory receptors and tumoral PD-L1 are associated with shorter overall survival (OS) in both primary and metastatic RCC (Ascierto et ai, 2016; Becht et ai, 2015, 2016; Chevrier et al. , 2017; Giraldo et ai, 2015, 2017). Despite the success seen with interleukin-2, an immuno-oncological revolution has been truly precipitated by the regulatory approval of immune checkpoint blockers, agents that release latent anticancer immunity. After positive trials in second line (2L) setting (Motzer et ai, 2015), the Checkmate 214 trial combining anti-PD-1 and anti-CTLA-4 (CICB) in first line (1L) metastatic RCC (Motzer et ai, 2018), new data from randomized Phase III trials (JAVELIN Renal 101, KEYNOTE-426, and IMmotion 151) provide evidence that immune-based combination therapy (anti-PD-(L)l and tyrosine kinase inhibitors (TKI)) is superior to standard care sunitinib (Motzer et ai, 2019; Porta and Rizzo, 2019; Rini et ai, 2019b, 2019a). In this rapidly expanding field, patients stratification is now required to predict tumor aggressiveness. Moreover, immune-related adverse events are common and lead to complex treatment paradigms. These obstacles can be overcome by exploring the impact of neo-angiogenesis/hypoxia patterns and Th1 geared- inflammatory profile to generate novel molecular classification of RCC (Beuselinck et ai, 2015; Casuscelli et ai, 2017). In addition, several arguments are currently in favor of the influence of the intestinal microbiome in oncogenesis and response to therapy, some establishing cause-effect relationships between the fecal composition and clinical outcome in mice and humans. First, distinct commensals exert protumorigenic effects, as observed in colon and pancreatic cancers (Kroemer and Zitvogel, 2018). Second, antibiotics (ATB) compromise the efficacy of (combined) immune checkpoint blockade (ICB), independently of the tumor histology (Derosa et ai, 2018; Elkrief et ai, 2019; Routy et ai, 2018). Third, microbiome profiling revealed different fingerprints between responders and non-responders to ICB across groups and countries (Gopalakrishnan et ai, 2018; Matson et ai , 2018; Routy et ai, 2018). Finally, selecting immuno-stimulatory bacteria species ( Akkermansia muciniphila (Routy et ai , 2018), Bifidobacterium longum (Matson et ai, 2018; Sivan et ai, 2015), Bacteroides fragilis (Vetizou et ai, 2015)) or strains ( Enterococcus hirae 13144 but not 13344 (Daillere et ai, 2016)) can elicit systemic immune responses and reprogram the tumor microenvironment (TME) in mouse tumors treated with anti-CTLA-4 and/or anti-PD-1 antibodies.

The results disclosed in the present application show that the composition of the microbiome is influenced by antibiotics, tyrosine kinase inhibitors (TKI) and immune checkpoint blockers (ICB), and that the composition of the microbiome has an impact on the success of immunotherapy by modulating the cancer-immune set point of the host and can be modified to increase the response to these treatments.

SUMMARY OF THE INVENTION

According to a first aspect, the present invention pertains to a composition comprising bacteria selected from the group consisting of Alistipes senegalensis, Dorea longicatena, Eubacterium siraeum and mixtures thereof, for use for treating a cancer, in combination with an immune checkpoint inhibitor (ICI)-based therapy and/or a tyrosine kinase inhibitor (TKI)-based therapy wherein said composition induces immunostimulation in a cancer patient.

The invention also pertains to a fecal microbial composition, for use in treating a cancer, in combination with an ICI-based therapy and/or a TKI-based therapy, wherein said composition has been enriched with a composition as above-described.

Method of in vitro determining if an individual having a cancer is likely to respond to a treatment with an ICI-based therapy and/or a TKI-based therapy are also part of the present invention. One is based on determining the relative abundances of Clostridium hathewayi, Clostridium clostridioforme and/or Clostridium boltae in a biological sample of said individual, wherein overrepresentation of at least one of these species indicates that the individual is likely to be a poor responder to said treatment.

Another method of in vitro determining if a cancer patient is likely to respond to an ICI-based and/or a TKI-based therapy comprises:

(i) determining the relative abundance of at least two immunotolerant species selected from the group consisting of Clostridium hathewayi, Clostridium clostridioforme and Clostridium boltae in the gut microbiota of said patient;

(ii) determining the relative abundance of at least two (e.g., 2, 3, 4 or 5) immunostimulatory species selected from the group consisting of Akkermansia muciniphila, Bacteroides salyersiae, Alistipes senegalensis, Dorea longicatena, and Eubacterium siraeum in the patient’s gut microbiota;

(iii) calculating the ratio of the relative abundance of the immunotolerant species of step (i) to the relative abundance of the immunostimulatory species of step (ii); wherein the lower the ratio calculated in step (iii), the higher the probability that the individual responds to the treatment.

Other methods for in vitro determining if a cancer patient is likely to respond to an ICI-based and/or a TKI-based therapy are based on assessing, in a blood sample from said patient, the presence of memory Th1 or Tc1 cells towards Alistipes senegalensis, Dorea longicatena and/or Eubacterium siraeum, and/or the presence of memory Tr1 cells towards Clostridium hathewayi, Clostridium clostridioforme and/or Clostridium boltae, wherein the presence of memory Th1 or Tc1 cells towards Alistipes senegalensis, Dorea longicatena and/or Eubacterium siraeum indicates that the patient is likely to be a good responder to said treatment and the presence of memory CD4+ Tr1 cells (IL-10 producing) or TH17 regulatory Rorct/foxp3 towards Clostridium hathewayi, Clostridium clostridioforme and/or Clostridium boltae indicates that the patient is likely to be a poor responder to said treatment.

Theranostic methods for determining whether an individual needs a bacterial compensation with a bacterial composition and/or by FMT before receiving an ICI-based therapy and/or a TKI-based treatment are also part of the invention.

The invention also pertains to the use of an endonuclease (e.g., a CRISPR/Cas9) capable of inducing a double-stranded break in a sequence specific for Clostridium hathewayi, Clostridium clostridioforme or Clostridium boltae, as a medication for treating cancer, in combination with a TKI and/or an ICB-based therapy. BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1: Antibiotics compromise the efficacy of PD-1 blockade and affect the intestinal composition of feces in advanced renal cell carcinoma patients. (A) Patients with advanced renal cell carcinoma (n=85) were evaluated for clinical outcomes and correlative fecal microbiota (n=69) analyses prior to and following initiation of anti-PD-1 blockade. Tumor response was assessed using the Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST v1.1).

(B) The best overall response was stratified by use of ATB (ATB=11, patients who took antibiotics; noATB=58, patients who did not take antibiotics). P value was obtained with two-tailed chi-squared test and Yates correction and significant p values are indicated with * (*p < 0.05, **p < 0.01, ***p < 0.001).

(C) Beta-diversity ordination plot based on Principal Coordinate Analysis (PCoA) of normalized and standardized data of fecal microbiota composition in pre-treatment (T0-T4) samples (n=69). Bacterial relative abundances were obtained with MetaOmineR package developed in ‘R’ by INRA. Percentage of variance embraced by each new coordinate is reported in percentages for each axis. Ellipses describing the 95% of confidence are even depicted for each cohort. ANOSIM metrics was implemented with 999 permutations to assess differences among ATB (gold) and noATB (blue) cohorts.

(D) LEfSe graph was implemented in Python v2.7 on bacterial species undergoing two-stages Benjamini-Hochberg False Detection Rate (FDR) at 10%, resulting in the identification of the most discriminant species for each cohort based on LDA score. Figure 2: Metagenomic analyses of fecal samples predict response of anti-PD-1 mAb in renal cell carcinoma patients.

(A) Shotgun sequencing of fecal microbiota in no-ATB basal (T0-T4) samples (n=58) with representation of gene richness and MGS count for all cancer patients according to clinical outcome (PFS at 3, 6, 9, 12 months). MeaniSEM of count are depicted for patients who experienced PFS more or less 3, 6, 9, 12 months. Of note gene richness and MGS count predict PFS at 12 months, while gene richness alone predicts PFS at 6 months.

(B) Beta-diversity ordination plot based on Principal Coordinate Analysis (PCoA) of normalized and standardized data of fecal microbiota composition in no-ATB basal (T0-T4) samples (n=58). Bacterial relative abundances were obtained with MetaOmineR package developed in ‘R’ by MetaGenoPolis (INRA). Percentage of variance embraced by each new coordinate is reported in percentages for each axis. Ellipses describing the 95% of confidence are even depicted for each cohort. ANOSIM and PERMANOVA metrics were implemented with 999 permutations to assess differences according to R (complete response or partial response or stable disease more than 6 months) and NR (death or progressive disease or stable disease less than 6 months).

(C) Variable Importance Plot (VIP) was implemented within Partial Least Square Discriminant Analysis (PLS-DA, inset differentiating NR and R), describing the 35 most discriminant species in descending order of importance. Each bar reports the following information: i) length, VIP score; ii) bar color, cohort in which the species has the highest mean relative abundance (high); iii) edge color, cohort in which the species has the lowest mean relative abundance (low); iv) thickness, Fold Ratio (FR) among high and low; v) significance of Mann-Whitney U test among high and low (*p < 0.05, **p < 0.01, ***p < 0.001).

(D) Barplots of relative abundances (within the 0-1 interval) and prevalence of selected species (A. muciniphila and B. salyersiae). P values for relative abundances were obtained after two-tailed Mann-Whitney U test, while P values for prevalence were retrieved by chi-square test.

Figure 3: Oral gavage with immunostimulatory commensals or feces from responders-RCC patients rescues the primary resistance in RCC tumor bearing mice.

(A) Experimental setting: Fecal microbial transplantation (FMT) was performed following 3 days of ATB in specific pathogen free (SPF) BALB/c mice. Two weeks later, luciferase engineered RENal cell CArcinoma (RENCA) were orthotopically inoculated and anti-PD-1 plus anti-CTLA-4 mAbs (CICB) or isotype control mAb (Ctrl) were inoculated intraperitoneally every 4 days starting from day-7. ATB induced dysbiosis were restored by oral administration of commensals A. muciniphila (Am), B. salyersiae (Bs), control bacteria B. xylanosolvens (Bx) or feces from responder patients (R) to recipient mice receiving CICB.

(B) Proportion of 15 FMT donors feces (human-responders, HR; human-non-responders, HNR) reflected in BALB/c mice (mice-responders, MR; mice- non-responders, MNR), as described in Table 7.

(C-D-E) Monitoring of RENCA progression using bioluminescence imaging of luciferase activity (C, E) or tumor weight (C-D) in ATB-treated mice post FMT with feces from 5 R and 10 NR RCC patients, treated with CICB, and compensated by oral administration of commensals A. muciniphila (Am), B. salyersiae (Bs) or feces from responder patients (R). All experiments were composed of 5-7 mice/group and were performed at least twice in similar conditions yielding similar results. ANOVA & Student T test statistical analyses of meansiSEM: (*p < 0.05, **p < 0.01, ***p < 0.001). Dx: last I VIS measurement, DO day of randomization.

Figure 4: The gut microbiota influences the systemic and local immune tonus in RCC tumor bearing mice. (A) Correlations of splenocyte profiles with selected bacterial consortium in isotype control (Ctrl) treatment group. Standardized relative abundances of species selected in Tables 6 and 7 were correlated with splenocyte profiles following a spearman correlation method and Benjamini-Hochberg correction. Splenocyte profiles were obtained by flow cytometry analyses at 48h after 2 ^nd injection of Ctrl in mice bearing orthotopic RENCA post NR FMT. Significant p values <0.05 are indicated with star.

(B) Flow cytometry analyses of CD103+CD11 b- DC in CD45 measured in the spleen of Ctrl treatment group in orthotopic RENCA tumor-bearer post NR FMT mice. Differences between each group were assessed by an ANOVA (ANalysis Of Variance) and significant p values are indicated with stars (*p < 0.05, **p < 0.01, ***p < 0.001).

(C) Correlations of Tumor infiltrated lymphocytes (TIL) profiles with selected bacterial consortium in Ctrl treatment group in orthotopic RENCA tumor-bearer post NR FMT mice. Standardized relative abundances of species selected in Tables 6 and 7 were correlated with TIL profiles following a spearman correlation method and Benjamini-Hochberg correction. TIL profiles were obtained by flow cytometry analyses in tumors at 48h after 2 ^nd injection of Ctrl in RENCA tumor-bearer post NR FMT mice. Significant p values <0.05 are indicated with a star.

(D) Flow cytometry analyses of TIL CXCR3+ CD4 in CD45 measured in the kidney of Ctrl treatment group in RENCA tumor-bearer post NR FMT mice. Differences between each group were assessed by an ANOVA (ANalysis Of Variance) and significant p values are indicated with * (*p < 0.05, **p < 0.01, ***p < 0.001).

Figure 5: Immuno-stimulatory versus -tolerant commensals govern the cancer-immune set point of tumor bearers.

(A) Correlations of TIL profiles with selected bacteria. Standardized relative abundances of species selected from Tables 6 and 7 were correlated to fold- ratio of CICB (anti-PD-1 & anti-CTLA-4 Abs) TIL divided by isotype control (Ctrl) TIL following a spearman correlation method and Benjamini-Hochberg correction. TIL profiles were obtained by flow cytometry analyses in tumors at 48h after 2 ^nd injection of CICB in RENCA tumor-bearer post NR FMT mice. Significant p values < 0.05 are indicated with a star.

(B) Percentages of TIL CXCR3+CD4+ in CD45 in CICB treatment group (B, left panel) and percentages of TIL CXCR3+CD8+ in CD45 in CICB treatment group (B, right panel) measured in the kidney of RENCA tumor-bearer post NR FMT mice. TIL profiles were obtained by flow cytometry analyses in tumors at 48h after 2 ^nd injection of CICB or Ctrl in RENCA tumor-bearer post NR FMT mice. Differences between each group were assessed by an ANOVA (ANalysis Of Variance) and significant p value are indicated with * (*p < 0.05, **p < 0.01, ***p < 0.001). (C-D) Linear correlation plots were performed on normalized and standardized relative abundances of selected bacterial species and normalized total flux (C) or splenocytes and TIL phenotypes (D) obtained by flow cytometry analyses (CICB+Ss (B. salyersiae) on CICB or CICB+Ss on Ctrl). Pearson coefficient and corresponding P values are reported within each graph as inset. (E) Clustermap of normalized total flux of RENCA tumor-bearer post

NR (7 donors) FMT mice and treated with CICB+Ss (B. salyersiae) or CICB or Ctrl. Logarithm in base 2 and a Bray-Curtis distance metrics were implemented.

(F) Kaplan-Meier curves showing progression-free survival of patients in relation to their microbial composition. Bs_Anr. with detectable B. salyersiae and A. muciniphila, Ss: with detectable iae, Am: with detectable A. muciniphila, NoBs_NoAm. without detectable B e and A. muciniphila in the upper panel; Ch : with detectable C. hathewayi without detectable C. hathewayi, in the middle panel; D/: with detectable D. longicatena and NoDl·. without detectable D. longicatena in the lower panel. Significant p values <0.05 are indicated with a star. Figure 6: Fecal microbiota differences in patients and mice treated with TKI.

Fecal microbiota compositional differences of patients who underwent first-line TKI treatment and control adults (A) and BALB/c mice (B) which underwent TKI treatment (axitinib, sunitinib, cabozantinib) were analyzed. LEfSe (Linear discriminant analysis Effect Size) and Partial Least Square Discriminant Analysis (PLS-DA) coupled to Variable Importance Plot (VIP) were implemented for humans and mice, respectively, in order to describe the most discriminant species in descending order of importance. In humans we considered first-line TKI treatment compared to literature-based controls (A), while in mice we considered the mean VIP score taken from the combined TKI. Briefly, VIP scores of all bacterial species which were present in at least two mice VIP plots were averaged and classified in descending order according to the species belonging to TKI or control cohort (B). Arrows highlight relevant bacterial species. Relative abundance and prevalence of the most discriminant species for TKI group, Alistipes senegalensis and Akkermansia muciniphila were reported (C) for the three different treatments (axitinib, sunitinib, cabozantinib), and a Mann-Whitney U test was used to assess statistical differences (*p < 0.05, **p < 0.01, ***p < 0.001). Figure 7: Antiangiogenic tyrosine kinase inhibitors induce an immuno-stimulatory intestinal microbiome shift.

(A) Variable Importance Plot (VIP) was implemented to describe the 35 most discriminant species in descending order of importance among BALB/c and C57BL6 mice treated with axitinib and sunitinib. Each bar reports the following information: i) length, VIP score; ii) face color, cohort in which the species has the highest mean relative abundance (high); iii) edge color, cohort in which the species has the lowest mean relative abundance (low); iv) thickness, Fold Ratio (FR) among high and low; v) significance of Mann-Whitney U test among high and low (*p < 0.05, **p < 0.01, ***p < 0.001).

(B) Monitoring of RENCA progression using bioluminescence imaging of luciferase activity in ATB-treated mice post FMT with feces from 1 NR RCC patients and treated with CICB or CICB with oral administration of B. salyersiae (Bs) or ICB with oral administration of axitinib or Ctrl. (C) Survival curves of RENCA bearing mice treated with CICB or Ctrl or ICB with oral administration of axitinib with or without oral gavage with Akkermansia muciniphila. Each line represents one survival curve for each group of 5 mice from 2 independent experiment. Log-rank (mantel-Cox) statistical analyses: (*p < 0.05,

**p < 0.01, ***p < 0.001). All experiments were composed of 5-7 mice/group and were performed at least twice in similar conditions yielding similar results. ANOVA & Student T test statistical analyses of meansiSEM: (*p < 0.05, **p < 0.01, ***p < 0.001). Dx: last IVIS measurement, DO day of randomization.

Figure 8: Antibiotics compromise the efficacy of anti-PD-1 mAb in renal cell carcinoma patients.

Kaplan-Meier estimates for progression-free survival (PFS) or overall survival (OS) of renal cell carcinoma patients. P values are shown [log-rank (Mantel-Cox) analysis].

Figure 9: Metagenomic analyses (MetaPhlAn2 pipeline) of fecal samples predict response of anti-PD-1 mAb in renal cell carcinoma patients.

(A) Beta-diversity ordination plot based on Principal Coordinate Analysis (PCoA) of normalized and standardized data of fecal microbiota composition in no-ATB pre-treatment (T0-T4) samples (n=58). Bacterial relative abundances were obtained with MetaPhlAn2 package developed in Python 2.7 by Center for Integrative Biology (CIBIO). Percentage of variance embraced by each new coordinate is reported in percentages for each axis. Ellipses describing the 95% of confidence are even depicted for each cohort. ANOSIM and PERMANOVA metrics were implemented with 999 permutations to assess differences according to R (complete response or partial response or stable disease more than 6 months) and NR (death or progressive disease or stable disease less than 6 months).

(B) Variable Importance Plot (VIP) was implemented within Partial Least Square Discriminant Analysis (PLS-DA, inset differentiating NR and R), describing the 35 most discriminant species in descending order of importance. Arrows are depicted to highlight species of importance. Each bar reports the following information: i) length, VIP score; ii) bar color, cohort in which the species has the highest mean relative abundance (high); iii) edge color, cohort in which the species has the lowest mean relative abundance (low); iv) thickness, Fold Ratio (FR) among high and low; v) significance of Mann-Whitney U test among high and low (*p < 0.05, **p < 0.01,

***p < 0.001).

(C) Barplots of relative abundances (within the 0-1 interval) and prevalence of selected species (A. muciniphila and B. salyersiae). P values for relative abundances were obtained after two-tailed Mann-Whitney U test, while P values for prevalence were retrieved by chi-square test.

Figure 10: GC and MGS count varies longitudinally with time. Heat maps of Log2 fold ratio (FR) of R versus NR (left) and R(Tx) versus R(T0) (right) for outcome (A) and PFS12 (B). Both GC and MGS counts were considered for FR calculation in overall pre-treatment samples (n=69) and in no-ATB pre-treatment samples (n=58). Patients’ numbers are considered at TO, and significance was assessed by Mann-Whitney U test (*p < 0.05, **p < 0.01, ***p < 0.001).

Figure 11 : Bacterial network of RCC patients (regardless ATB usage). Network was created by co-occurrences of 124 bacterial species (the nodes) and concomitant significance of pair-wise Pearson correlation coefficient (the edges). In order to fulfil the formal requirements for patents figures, Figure 11A has been divided to be represented in three pages (23/39, 24/39 and 25/39), which can be put side-by-side to reconstitute the network. (A). Node properties are as follows: i) size, normalized and standardized bacterial relative abundances; ii) color, ‘guilds’ (GIG) retrieved by Blondel algorithm to detect bacterial communities; iii) name size, betweenness centrality (a measure of the importance within the network). Edge properties: i) thickness, proportional to P value of Pearson correlation coefficient divided into 8 categories from the most significant (thicker) to the lesser one (thinner); ii) color, red for positive and blue for negative Pearson correlation coefficient. Spp, number of species within each GIG. EdgesT, total number of edges. Edges+, percentage of edges with positive correlation. Edges-, percentage of edges with negative correlation. The same network underwent four different node coloring (panel B, sunitinib; panel C, OUTCOME2; panel D, ATB; panel E, axitinib) taking into account the cohort in which each species had the highest average relative abundance, while the node size derived from the importance for that species in dividing the cohorts following the random forest algorithm. Within panels B-D edge coloring was discarded to ameliorate node visualization and interpretation.

Figure 12: Analysis of the bacteria discriminating RCC from control adults The 69 RCC samples were analyzed and compared with 2994 control adults acquired from publicly available repositories and spanning multiple countries and lifestyles. This set of control samples was enlarged by 54 Italian samples newly acquired and sequenced in the current study. Bacterial species discriminating RCC from control adults were determined using LefSe. Figure 13: Local network of B. salyersiae

(A-B) Local network of B. salyersiae within 69 patients regardless ATB usage (A) and within 58 patients who did not take ATB (B). Network properties (nodes, edges) as per Figure S.1.

(C) Monitoring of RENCA progression using bioluminescence imaging of luciferase activity in ATB-treated mice post FMT with feces from 1 NR RCC patients and treated with CICB or CICB with oral administration of B. salyersiae (Bs) and Acidaminococcus intestini (Ai) or CICB with oral administration of B. salyersiae (Bs) and Sutterella Wadswothensis (Sw) or Ctrl.

Figure 14: The gut microbiota controls the cancer-immune set point in RCC tumor bearing mice

(A) Splenocyte profiles in isotype control (Ctrl) treatment group in RENCA tumor-bearer post-FMT mice.

(B) Tumor infiltrated lymphocyte (TIL) profiles in isotype control (Ctrl) treatment group in RENCA tumor-bearer post-FMT mice. Splenocyte or TIL profiles obtained by cytometry and standardized relative abundances of species selected in Tables 6 and 7 were clustered following a hierarchical clustering (Euclidean distance and complete method).

Figure 15: Fecal microbiota differences in C57BL6 mice treated with TKI. Fecal microbiota compositional differences of C57BL6 mice which underwent TKI treatment (sunitinib, panel A; axitinib, panel B; cabozantinib, panel C) were analyzed. In order to assess beta-diversity, Principal Coordinate Analysis (PcoA, insets) was implemented, while ANOSIM and PERMANOVA metrics were used with 999 permutations to assess differences among the patients’ cohorts. Variable Importance Plot (VIP) were generated by Partial Least Square Discriminant Analysis (PLS-DA) to describe the 35 most discriminant species in descending order of importance for each TKI treatment. Each bar reports the following information: i) length, VIP score; ii) face color, cohort in which the species has the highest mean relative abundance (high); iii) edge color, cohort in which the species has the lowest mean relative abundance (low); iv) thickness, Fold Ratio (FR) among high and low; v) significance of Mann-Whitney U test among high and low (*p < 0.05, **p < 0.01, ***p < 0.001).

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS In the present text, the following definitions are used:

- An “immune checkpoint inhibitor” (ICI) designates any drug, molecule or composition which blocks certain proteins made by some types of immune system cells, such as T cells, and some cancer cells. These proteins help keep immune responses in check and can keep T cells from killing cancer cells. When these proteins are blocked, the “brakes” on the immune system are released and T cells are able to kill cancer cells better. Examples of checkpoint proteins found on T cells or cancer cells include PD- 1/PD-L1 and CTLA-4/B7-1/B7-2. In particular, ICIs encompass anti-PD1 antibodies (such as Nivolumab or Pembrolizumab), anti-PD-L1 antibodies (such as Atezolizumab or Durvalumab), anti-CTLA-4 antibodies and anti-PD-L2 antibodies. In the scientific literature, ICIs are also designated as “drugs blocking an immune checkpoint”, or “immune checkpoint blockers” (ICB) or “immune checkpoint blockade drugs”.

- An “anti-PD1/PD-L1/PD-L2 Ab-based therapy” herein designates any therapy including the use of a drug that antagonizes PD1, PD-L1 or PD-L2. These include therapies mainly based on an ICI such as a drug antagonizing PD1 or PD-L1 or PD-L2, as well as combined therapies using several ICIs and/or additional anticancer drugs such as chemotherapeutic drugs. Non-limitative examples of combined therapies encompassed by the phrase “anti-PD1/PD-L1/PD-L2 Ab-based therapy” include anti- PD1 + anti-CTLA4, anti-PD1 + polychemotherapy (pemetrexed+ carboplatin), anti-Lag3 + anti-PD1, anti-NKG2A + anti-PD1, IDO inhibitor + anti-PD1 and anti-ICOS+anti-PD1. Although the currently used drugs antagonizing immune checkpoint proteins are monoclonal antibodies, other molecules specifically binding to PD1, PD-L1, PD-L2 or other proteins could be used for the development of future ICIs such as, for example, antibody fragments or specifically designed aptamers. Of course, the phrase “anti- PD1/PD-L1/PD-L2 Ab-based therapy” encompasses any therapy including an active molecule that antagonizes PD1 or PD-L1 or PD-L2.

- A “tyrosine kinase inhibitor” (TKI) designates any drug, molecule or composition which inhibits tyrosine kinases. Tyrosine kinases are enzymes responsible for the activation of many proteins by signal transduction cascades. The proteins are activated by adding a phosphate group to the protein (phosphorylation), a step that TKIs inhibit. Non-limitative examples of TKIs are listed below.

Table 1: tyrosine kinase inhibitors

- “NR” defines a non-responder status to PD-1/PDL-1/PDL-2 blockade

- “R” defines a responder status to PD-1/PDL-1/PDL-2 blockade - “NGS” defines any Next Generation Sequencing platform available in the past, present or in the future.

- In the present text, each "bacterial species" is defined by a Co- Abundance gene Group (“CAG”), which is a group of bacterial genes from the gut microbiome (i.e., the gene repertoire of the gut microbiota), which abundance level varies in the same proportion among different individual samples. In other words, a bacterial species according to the invention is defined by a cluster of bacterial gene sequences which abundance levels in samples from distinct subjects are statistically linked rather than being randomly distributed.

Most current approaches for analyzing metagenomic data rely on comparisons to reference genomes, but the human gut microbiota diversity extends beyond what is currently covered by reference databases. In the results disclosed herein, the inventors used a method based on binning co-abundant genes across a series of metagenomic samples, that enables comprehensive discovery of new microorganisms without the need for reference sequences. In what follows, some species identified as likely to play a role in the patients’ response to therapies based TKI or ICI may be newly- identified species, not yet precisely referenced in public databases. For each of the identified species (both newly-identified and species very close to already referenced species), the present application discloses a set of bacterial genes which are non- redundant sequences and can be used, alone or in combination, as tracer genes to assess the presence and relative abundance to the corresponding species. Of course, once the species are identified, either by the set of non-redundant genes disclosed herein, or later on by their further identification and/or inclusion into a data base, the skilled in the art can assess their relative abundance by any appropriate means, such as, for example, by measuring the copy number of another non-redundant gene that co varies with the sequences disclosed in the present application, or even by identifying a signature of this species at the protein level rather than in a nucleic acids sample. Hence, the present invention is not limited to the use of the disclosed sequences to measure the relative abundance of the corresponding species.

- The “relative abundance” of a definite bacterial is defined as the fraction of the entire bacterial ecosystem belonging to this bacterial species. Throughout the present text, all relative abundances are expressed within the closed interval [0 : 1], where 1 stands for the maximum fraction available for a single bacterial species (/.e., a bacterial species with a relative abundance equal to 1 means that 100% of the bacteria present in the sample are of the considered species). Using a NGS technique, the relative abundance of a bacterial species is considered as the number of reads of that selected species divided by the total number of reads representing the overall bacterial community. Using a qPCR technique, the relative abundance of a bacterial species is considered as the ACt value of that species X (amplified by a pair of primers specific for X) divided by the ACt value of the total bacteria (amplified by an universal primers pair able to catch all the eubacteria present in a sample, the pair consisting of primers PRK341F and PRK806R or the pair consisting of primers 27F and 1492R).

When necessary, other definitions are provided later in the present text.

According to a first aspect, the present invention concerns a composition comprising bacteria selected amongst Alistipes senegalensis, Dorea longicatena and Eubacterium siraeum, for use in treating a cancer, in combination with an immune checkpoint inhibitor (ICI)-based therapy and/or a tyrosine kinase inhibitor (TKI)-based therapy wherein said composition induces immunostimulation in a cancer patient. According to a particular embodiment, the composition comprises a mix of at least two species selected amongst Alistipes senegalensis, Dorea longicatena and Eubacterium siraeum.

Other immunostimulating bacterial compositions have already been described, for example in WO 2016/063263 and in WO 2018/115519. The bacterial compositions according to the present invention can also comprise one or several of the bacterial species of the previously described compositions, in order to combine the favorable effects of the bacterial species.

According to a particular embodiment, the composition according to the present invention further comprises bacteria of at least one species selected amongst Enterococcus hirae, Akkermansia muciniphila and Bacteroides salyersiae. For example, the composition can comprise a mix of Alistipes senegalensis and Bacteroides salyersiae, or a mix of Alistipes senegalensis and Akkermansia muciniphila.

Non-limitative additional examples of bacterial strains which can be included in the compositions according to the invention are: Blautia strains, Coprococcus comes strains, Alistipes shahii, other Alistipes species (e.g. Alistipes indistinctus and/or onderdonkii and/or finegoldii), Ruminococcacae, Clostridiales species, Bacteroidales species, Actinobacteria, Coriobacteriales species, Methanobrevibacter smithii, Burkholderia cepacia, Bacteroides fragilis, Actinotignum schaalii, as well as Clostridiales bacteria of the species Christensenella minuta ; Erysipelotrichia of the species Dielma fastidiosa or Erysipelatoclostridium ramosum\ Eubacterium iimosum\ Barnesiella intestinihominis ; Coriobacteriales bacteria of the species Collinsella intestinalis and/or Collinsella tanakaer, and Firmicutes bacteria of the species Flavonifractor plautii.

The present invention also pertains to a fecal microbial composition enriched with a bacterial composition as above-described, and to its use in treating a cancer, in combination with an ICI-based therapy and/or a TKI-based therapy. A fecal microbial composition is a composition of matter derived from one or several feces sample(s), preferably obtained (directly or indirectly) from a stool sample from (a) healthy individual(s) and/or from (a) responder(s) to a treatment with an ICI- and/or TKI-based therapy, or at least from an individual exhibiting a gut microbiota profile that identifies him/her as likely to respond to the envisioned treatment. The fact that the fecal microbial composition can be obtained indirectly from a healthy individual’s stool sample means that banks of fecal microbial material may be created, with possible mixes of stool samples, and possible creation of “standard healthy fecal microbial compositions”, possibly adapted to certain conditions requiring FMT (e.g., a fecal microbial composition for treating a Clostridium infection may be different from a fecal microbial composition for use in a cancer context) and/or to other characteristics of patients (age, ethnic origin, food regimen etc.). Several ways of conditioning fecal microbial material and conducting FMT have been described and are currently developed, and the skilled artisan is free to choose appropriate techniques for preparing the fecal microbial composition according to the invention, which can be freshly-prepared liquid, freeze-dried material or any other conditioning. In what follows, the word “composition(s)” indifferently designates bacterial compositions and fecal microbial compositions according to the invention.

The above compositions are particularly useful for inducing immunostimulation in patients having a cancer that can be treated with a TKI, such as (but not limited to) any of those listed in Table 1, especially breast cancer, chronic myeloid leukemia (CML), GIST and sarcoma, glioblastoma, thyroid cancers, (advanced) renal cell cancer (RCC) and non-small cell lung cancer (NSCLC).

According to another particular embodiment, the composition according to the invention is used in combination with an ICI-based therapy and a TKI-based therapy. The present invention also pertains to the use of the above bacterial compositions or fecal microbial compositions, as a medicament for compensating dysbiosis in a cancer patient. A “dysbiosis” can be defined as a disequilibrium between potentially “detrimental” and “beneficial” bacteria in the gut or any deviation to what is considered a healthy microbiota in terms of main bacterial groups composition and diversity. Dysbiosis may be linked to health problems, including cancer (as shown in WO 2018/115519). It can also be the consequence of a treatment, such as a cytotoxic treatment or an antibiotic treatment.

It is to be understood that when a composition according to the invention is used “in combination with” a TKI and/or an ICI-based therapy, the bacterial or fecal material composition and the TKI and/or ICI can be administered either concomitantly or sequentially. For example, the patient is first treated with the TKI (first-line therapy in RCC), followed by a second treatment sequence in which the patient receives a TKI and an ICB, as well as a bacterial composition comprising Alistipes senegalensis and/or Akkermansia muciniphila. According to another aspect, the present invention pertains to a protocol for treating a patient having a cancer (e.g., a RCC or another TKI-sensitive cancer such as those listed above), in which: (i) the patient receives a first-line TKI-based therapy, (ii) the patient’s microbiota is analyzed to assess whether an intestinal microbiome shift has occurred (compared to the intestinal microbiota before TKI uptake), and (iii) depending on the result of step (ii), the TKI-based therapy is maintained in combination with an ICI-based therapy, if necessary accompanied by administration of a compensating composition as the bacterial compositions and fecal material compositions described above. In particular, if the result of step (ii) shows that the relative abundance of Alistipes senegalensis has increased in the intestinal microbiota following TKI administration, the ICI-based immunotherapy comprising anti-PD1 Ab could be combined to TKI for the rest of the clinical management, or the combination of anti-CTLA4+anti-PD1 could be the main therapy. If the result of step (ii) shows that the relative abundance of Alistipes senegalensis or E. siraeum has not increased in the intestinal microbiota following TKI administration, the ICI-based immunotherapy comprising anti-PD1 Ab could be combined to FMT or administration of beneficial bacterial compositions (described above) for the rest of the clinical management, or the combination of anti-CTLA4+anti- PD1 together with the benefical bacteria. In the above method, the patient’s microbiota is analyzed in an appropriate sample from the patient, such as, for example, a feces sample, a biopsy from the patient’s ileum or colon mucosae or a tumor biopsy.

The present invention also pertains to a method of in vitro determining if an individual having a cancer is likely to respond to a treatment with an ICI-based therapy and/or a TKI-based therapy, comprising the following steps:

(i) determining the relative abundances of Clostridium hathewayi (previously known as Hungatelia hathewayi), Clostridium clostridioforme and/or Clostridium boltae in a biological sample of said individual, and

(ii) comparing each of the relative abundances measured in step (i) to a control value, wherein overrepresentation of at least one of Clostridium hathewayi, Clostridium clostridioforme and Clostridium boltae indicates that the individual is likely to be a poor responder to said treatment.

In the above method, step (i) can be performed by measuring, in an appropriate sample from the patient (as defined above), the relative abundances of Clostridium hathewayi, Clostridium clostridioforme and/or Clostridium boltae. The obtained values are then compared, in step (ii), to control values based on relative abundances of the same species in normal volunteers (healthy volunteers who did not take antibiotics recently). A given species is considered as “overrepresented” when its relative abundance in the sample from the patient is superior to the control value, it being understood that (a) for species that are normally not present in healthy volunteers (e.g., the value is zero in healthy volunteers in novel data bases such as MetaphLan or Meta HIT at the plateau of worldwide machine learning), the mere presence of the bacterium is considered of negative predictive value and (b) for species which are normally present in healthy volunteers, the control values are determined so that a relative abundance above this value is significantly superior (for a skilled person) to what is observed in healthy volunteers. Alternatively, the above method can be performed by determining, in the patient’s serum, IgG responses directed against Clostridium hathewayi, Clostridium clostridioforme and/or Clostridium boitae. These responses are then compared to control values, such as those observed in healthy volunteers. The present invention also pertains to a method for in vitro determining if an individual having a cancer is likely to respond to a treatment with an ICI-based therapy and/or a TKI-based therapy, comprising the following steps:

(i) from an appropriate biological sample of said individual, determining the relative abundance of at least two immunotolerant species selected from the group consisting of Clostridium hathewayi, Clostridium clostridioforme and

Clostridium boitae ;

(ii) from an appropriate biological sample of said individual, determining the relative abundance of at least three immunostimulatory species selected from the group consisting of Akkermansia muciniphila, Bacteroides salyersiae, Alistipes senegalensis, Dorea longicatena and Eubacterium siraeum;

(iii) calculating the ratio of the relative abundance of the immunotolerant species of step (i) to the relative abundance of the immunostimulatory species of step (ii); wherein the lower the ratio calculated in step (iii), the higher the probability that the individual responds to the treatment.

In the above method, the terms “immunostimulatory” and “immunotolerant” reflect the effects of the recited bacteria on the response or the resistance of the patient to the treatment with an ICI and/or TKI. Indeed, as shown in Example 4 below, some species, which disappear during cancer development or antibiotics uptake, appear associated with an efficient response to a treatment by, e.g., anti-PD-1 blockade. These species are defined as "immunostimulatory". Other species, on the contrary, are specifically selected following antibiotics administration and during cancer development, and may confer primary resistance to this therapy. Such species are referred to as "immunotolerant” herein. When performing the above method, the ratio obtained in step (iii) can be compared to one or several predetermined thresholds to obtain a probability score that the patient responds to the treatment. These thresholds can be calculated by the skilled person based on the results obtained in patients cohorts. One particular threshold is also calculated based on the relative abundancies observed in healthy volunteers, it being understood that if the ratio is inferior to this threshold, the patient is likely to be a good responder.

In particular, the above methods can be performed by using the following control values, obtained in healthy volunteers: • “immunotolerant” bacteria are considered as overrepresented when their relative abundances are:

>0.09% +/- 0.017% for C. clostridiofome,

>0.21% +/- 0.024% for C. boltae and >0.06% +/- 0.009% for C. hathewayi

• “immunstimulatory” bacteria are considered as overrepresented when their relative abundances are:

>0.855% +/- 0.022% for D. longicatena >1.11 % +/- 0.058% for E. siraeum, >1.89% +/- 0.10% for A. muciniphila,

>0.05% +/- 0.0022% for A. senegalensis, and >0.15% +/- 0.01% for B. salyersae

These data can be used also to calculate the control value for the ratio of the relative abundance of the immunotolerant species of step (i) to the relative abundance of the immunostimulatory species of step (ii) in the above method. For example, if this method is based on the measure of the relative abundances of C. boltae, C. hathewayi, D. longicatena and A. muciniphila, one can compare the ratio

[RA(C. boltae) + RA(C. hathewayi )] / [RA(D. longicatena) + RA(A muciniphila )] to 2 control values calculated as follows V1 = [(0.21+0.024) + (0.06+0.009)] / [(0.855 - 0.022) + (1.89% - 0.10)]

V2 = [(0.21-0.024) + (0.06-0.009)] / [(0.855 + 0.022) + (1.89% + 0.10)] and consider that if the measured ratio is superior to V1, the patient is likely to be a poor responder to the treatment, and/or if the measured ratio is inferior to V2, the patient is likely to respond to the treatment. When performing the methods according to the invention, the skilled person can use any technique to measure the relative abundances of the bacterial species, such as NGS (through any past or future NGS platform, from the first generation to the last available on the market and those in development, using any NGS output file provided as fastq, BAM, SAM, or other kind of files extensions) or any other technique such as, for example, qPCR (quantitative polymerase chain reaction) and microarrays to express the relative abundances of selected bacterial species.

Specific genome sequences and primer pairs are disclosed herein (Table 2), which can be used to detect the bacterial species mentioned above and measure their relative abundance according to the invention.

Table 2: genome sequences and primers specific for the recited bacterial species.

*: primer pairs for specifically amplifying fragments (of a length comprised between 70 and 350 pb) of the recited species are formed with primers having two consecutive numbers (SEQ ID No: 2n+1 and SEQ ID No: 2n+2, n being an integer)

Other methods for ex vivo determining whether a cancer patient is likely to benefit from a treatment with an ICI-based therapy and/or a TKI-based therapy are also part of the present invention, based on the analysis of memory immune responses directed against the immunostimulatory and/or immunotolerant bacterial species defined above.

Thus, the present invention pertains to a method for ex vivo determining whether a cancer patient is likely to benefit from a treatment with an ICI-based therapy and/or a TKI-based therapy, comprising assessing the presence of memory Th1 or Tc1 cells towards Alistipes senegalensis, Dorea longicatena and/or Eubacterium siraeum in a blood sample from said patient, wherein the presence of memory Th1 or Tc1 cells towards Alistipes senegalensis, Dorea longicatena and/or Eubacterium siraeum indicates that the patient is likely to be a good responder to said treatment.

Another method according to the invention for ex vivo determining whether a cancer patient is likely to benefit from a treatment with an ICI-based therapy and/or a TKI-based therapy, comprises assessing the presence of memory Tr1 cells towards Clostridium hathewayi, Clostridium clostridioforme and/or Clostridium boltae in a blood sample from said patient, wherein the presence of memory CD4+ Tr1 cells (IL-10 producing) or TH17 regulatory Rorct/foxp3 towards Clostridium hathewayi, Clostridium clostridioforme and/or Clostridium boltae indicates that the patient is likely to be a poor responder to said treatment.

The skilled person can of course decide to combine two of the above described methods, to better assess the patient’s profile. For example, the memory immune response against both the immunostimulatory bacterial species and the immunotolerant bacterial species can be assessed according to the invention. According to another example, method based on the measure of the relative abundance of immunotolerant bacterial species can be combined with that based on assessing the memory immune response against immunostimulatory bacterial species, etc. Such combined methods are also part of the present invention.

Alternatively or complementarily, one or several of the above methods is (are) combined with another method for determining, from a feces sample from a cancer patient, whether said patient is likely to be a good responder to a treatment with an ICI, based on an animal model. Such a method was already described in a previous application from the inventors’ team (WO2016/063263) and comprises the steps of (i) performing a fecal microbial transplantation (FMT) of a feces sample from the patient into germ free (GF) model animals (e.g., GF mice); (ii) at least 7 to 14 days after step (i), inoculating said mice with a transplantable tumor model; (iii) treating the inoculated mice with the ICI; and (iv) measuring the tumor size in the treated animals. The results of step (iv) are illustrative of the response that can be expected for said patient to said treatment.

A nucleic acid microarray designed to perform a method according to the invention is also part of the present invention. Such a nucleic acid microarray comprises nucleic acid probes specific for each of the microorganism species to be detected in said method. In a specific embodiment, the nucleic acid microarray is an oligonucleotide microarray comprising at least one oligonucleotide specific for at least one sequence selected from SEQ ID NOs: 1-72. For example, the said microarray comprises at least 6 oligonucleotides, each oligonucleotide being specific for one sequence of a distinct species. The microarray of the invention can of course comprise more oligonucleotides specific for sequences of SEQ ID NOs: 1-72. The microarray according to the invention may further comprise at least one oligonucleotide for detecting at least one gene of at least one control bacterial species. A convenient bacterial species may be e.g. a bacterial species the abundance of which does not vary between individuals having a R or a NR status. Preferably, the oligonucleotides are about 50 bases in length. Suitable microarray oligonucleotides specific for any gene of SEQ ID NOs: 1-72 may be designed, based on the genomic sequence of each gene, using any method of microarray oligonucleotide design known in the art. In particular, any available software developed for the design of microarray oligonucleotides may be used, such as, for instance, the OligoArray software, the GoArrays software, the Array Designer software, the Primer3 software, or the Promide software, all known by the skilled in the art.

The above methods can also be performed for determining if a cancer patient needs a bacterial compensation before or during administration of an ICB-based therapy and/or a TKI-based therapy. Indeed, if the patient is identified as likely to be a poor responder to the treatment, his/her situation can be improved by bacterial compensation. According to this aspect of the invention, the bacterial compensation can be done either by fecal microbiota transplant (FMT), using microbiota from one or several donors (for example, from responders to the treatment), or by administration of a fecal microbial composition or a bacterial composition as above-described. The inventors already described other bacterial compositions that can be used for such a compensation and restore the ability, for the patient, to respond to the treatment (e.g., in WO 2016/063263 and in WO 2018/115519). The present invention thus pertains to a theranostic method for determining if a cancer patient needs a bacterial compensation before or during administration of an ICB-based therapy and/or a TKI-based therapy, comprising assessing, by any method as above-described, whether the patient is likely to respond to the treatment, wherein if the patient is likely to be a poor responder to the treatment, he/she needs a bacterial compensation, for example with a composition according to the invention.

The above methods for determining whether a cancer patient is likely to benefit from a treatment with an ICI-based therapy and/or a TKI-based therapy, and/or whether this patient need a bacterial compensation are especially useful for patients having a breast cancer, chronic myeloid leukemia (CML), GIST and sarcoma, glioblastoma, thyroid cancers, (advanced) renal cell cancer (RCC) and non-small cell lung cancer (NSCLC).

Recently, attempts to directly manipulate the gut microbiome in a targeted manner in situ have been described, using gene editing tools such as the CRISPR / Cas9 system (Ramachandran and Bikard, 2019; Lee et al. , 2018). This strategy can be used to design "precision" antimicrobials that target immunotolerant bacterial species in a DNA sequence-specific manner.

The present invention thus pertains to the use of an endonuclease capable of inducing a double-stranded break in a sequence specific for Clostridium hathewayi, Clostridium clostridioforme or Clostridium boltae, as a medicament for treating cancer, in combination with a TKI and/or an ICB-based therapy. Several sequence-specific endonucleases useful for gene editing have been described, such as TALE nucleases (TALENs) or zinc-finger nucleases (ZFNs) and CRISPR/Cas systems. According a particular embodiment of the invention, the endonuclease is a CRISR/Cas9. The skilled person can choose any appropriate delivery methods for vectorising the endonuclease according to the invention, such as, for example, transduction (via a phage) or conjugation.

Specific endonucleases according to the invention target sequences listed in Table 3 below. In particular, when CRISPR/Cas is used with a guide RNA targeting a sequence disclosed in this table, the corresponding PAM sequence is indicated. The present invention thus pertains to an endonuclease which targets a sequence selected from the group consisting of SEQ ID Nos: 213 to 248.

Table 3: target sequences and corresponding PAM for CRISPR/Cas gRNA

Other characteristics of the invention will also become apparent in the course of the description which follows of the biological assays which have been performed in the framework of the invention and which provide it with the required experimental support, without limiting its scope.

EXAMPLES

In the experimental examples, the following abbreviations are used: 1L: first line therapy, 2L: second line therapy, ATB: antibiotics, “ATB”: patient who took antibiotics, CICB: combined immune checkpoint blockade using anti-PD-1/anti-CTLA-4 antibodies, Ctrl: isotype control, DC: dendritic cells, FMT: fecal microbial transplantation, GC: gene count, GIG: genome interaction group, GOMS: Gut OncoMicrobiome Signature, HV: healthy volunteers, ICB: immune checkpoint inhibitor anti-PD1 antibody, LEfSe: linear discriminant analysis of effect size, MGS: metagenomic species, “noATB”: patient who did not take antibiotics, NR: non-responders, OS: overall survival, PCoA: principal coordinate analyses, PD: progressive disease, PFS: progression free survival, R: responders, RCC: renal cell carcinoma, RENCA: renal cell carcinoma murine model, SD: stable disease, Tel: IFNy producing CD8+ T lymphocyte, TH1: IFNy producing CD4+ T lymphocyte, TIL: tumor infiltrating lymphocytes, TKI: tyrosine kinase inhibitors, TME: tumor microenvironment, VEGF: anti-vascular endothelial growth factor, WGS: wall genome sequencing.

Materials and Methods A. Patient characteristics and clinical study details:

Medical centers and regulatory approvals for translational research. The clinical study was conducted according to the ethical guidelines and approval of the local CCPRB. For feces collection, the study name was “Oncobiotics”, B2M ethics protocol number PP: 15-013. Written informed consent in accordance with the Declaration of Helsinki was obtained from all patients. Collection of patient feces.

The patients were included from Gustave Roussy Cancer Campus, France. Inclusion criteria were patients with stage IV clear cell or non-clear cell RCC histology and disease progression during or after ³1 prior anti-angiogenic therapy regimens who received nivolumab intravenously (i.v.) 3 mg/kg every 2 weeks until disease progression or intolerable toxicity in the NIVOREN GETUG-AFU 26 Phase II trial (EudraCT: 2015-004117-24) (Albiges et ai, 2018). Computer tomography (CT) scans were performed at baseline and every 8 to 12 weeks for the first year and then every 12 to 15 weeks until disease progression. Tumor response was assessed using the Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST v1.1) (Eisenhauer et ai, 2009). Data were collected from a case report form (CRF). All patents werefollowed- up until death or data lock (September 2018). We evaluated objective response rate defined as the number of patients with a complete response and a partial response. The best overall response was defined as the investigator-assessed best response (complete response, partial response, stable disease, or progressive disease) from the start date of nivolumab to objectively documented disease progression or subsequent therapy, whichever occurred first. Patient were divided into 2 groups: responders (those in complete response, partial response or stable disease for more than 6 months) and non responders (who either progressed or were in stable disease for less than 6 months or died). Progression-free survival (PFS) was defined as the time from the start date of nivolumab to first documented RECIST-defined tumor progression or death from any cause. Four median PFS values (3, 6, 9 and 12 months) were used to examine the impact of PFS time on metagenomics. Feces were collected according to International Human Microbiome Standards (IHMS) guidelines (SOP 03 V1) at different timepoints: before the first injection (0-1 month before) (TO), after the 2nd (T4- 4 weeks), after the 4th (T8- 8 weeks) and after the 12th (T24 - 24 weeks) injection (Figure 1). In brief, a collection kit including an anaerobic generator (Biomerieux) was given to patients. Samples were collected by patients at home, and frozen 4 to 24h later at -80°C at Gustave Roussy Cancer Campus in plastic tubes (Plastic vessel by 1000-Sarstedt) with or without BHI + 2 % glycerol. Forty patients were analyzed in addition to 60 non-small cell lung cancer patients in the Science 2018 paper (Routy et ai, 2018). Finally, from February 2016 to September 2018, a total of 85 patients with RCC were enrolled in the NIVOREN trial at Gustave Roussy and we collected T0-T4 feces of 69 patients. For the first time, RCC were analyzed as a single and independent cohort in this paper.

Metagenomic analysis of patient stools.

Total fecal DNA was extracted as described (Godon etai, 1997; Suau et ai, 1999) and sequenced using ion-proton technology (ThermoFisher) resulting in 22.7 ±0.9 million (meaniSD) single-end short reads of 150-base-long single-end reads as a mean. Reads were cleaned using (Criscuolo etai, 2013). AlienTrimmer in order (i) to remove resilient sequencing adapters and (ii) to trim low quality nucleotides at the 3’ side using a quality and length cut-off of 20 and 45 bp, respectively. Cleaned reads were subsequently filtered from human and other possible food contaminant DNA (using Human genome RCh37-p10, Bos Taurus and Arabidopsis thaliana and an identity score threshold of 97%). For the MetaOMineR analyses the gene abundance profiling was performed using the 9.9 million gene integrated reference catalog of the human microbiome (Li et ai, 2014). Filtered high-quality reads were mapped with an identity threshold of 95% to the 9.9 million-gene catalogue using (Langmead et Salzberg, 2012) Bowtie 2 included in METEOR software (Cotillard et ai, 2013). The gene abundance profiling table was generated by means of a two-step procedure using METEOR. The gene abundance table was processed for rarefaction and normalization and further analysis using the MetaOMineR (momr R) package (Le Chatelier et ai, 2013). The gene abundance table was rarefied to 13 million reads per sample (a threshold chosen to include all samples, but 1 with 12.5 million reads) by random sampling of 13 million mapped reads without replacement. The resulting rarefied gene abundance table was normalized according to the FPKM strategy (normalization by the gene size and the number of total mapped reads reported in frequency) to give the gene abundance profile table. Metagenomic species (MGS) are co-abundant gene groups with more than 500 genes corresponding to microbial species. 1436 MGS were clustered from 1267 human gut microbiome samples used to construct the 9.9 million-gene catalogue (Li et ai, 2014), as described (Nielsen et ai, 2014). Differentially abundant MGS between different patients’ groups were selected using the Wilcoxon test (p<0.05). Microbial gene richness (gene count) was calculated by counting the number of genes that were detected at least once in a given sample, using the average number of genes counted in 10 independent rarefaction experiments. MGS richness (MGS count) was calculated directly from the MGS abundance matrix. For the MetaPhlAn2 analyses fastq files were cleaned/filtered as described above and underwent an additional filtering for possible human contaminants (reference database GRCh37/hg19) and contextual quality control using KneadData. This wrapper entangles Bowtie2 ("--very-sensitive" and "--dovetail" settings) to rule out contaminant sequences and Trimmomatic (sliding window 20, min- length 50) to exclude low-quality reads. Filtered reads underwent MetaPhlAn2 pipeline (default settings) for unambiguous taxonomic classification and to generate a table of relative abundances for bacterial, archaeal, eukaryotic and viral species. Only taxa that were present in at least 20% of all samples were considered. Raw tabular data were firstly normalized then standardized using QuantileT ransformer and StandardScaler methods from Sci-Kit learn package v0.20.3. Normalization using the output_distribution='normal' option make each variable to strictly have a gaussian shape distribution, while the standardization makes each variable to have zero mean and unit variance. Measurements of a diversity (within sample diversity) such as observed_otus and Shannon index, were calculated at OTU level using the SciKit-learn package v.0.4.1. Exploratory analysis of b-diversity (between sample diversity) was calculated using the Bray-Curtis measure of dissimilarity and represented in Principal Coordinate Analyses (PCoA), while for Hierarchical Clustering Analysis (HCA) ‘Bray-Curtis’ metrics and ‘complete linkage’ method was implemented using custom scripts (Python v.2.7.11). We implemented Partial Least Square Discriminant Analysis (PLS-DA) and the subsequent Variable Importance Plot (VIP) as a supervised analysis in order to find out the most discriminant bacterial species. Mann-Whitney U and Kruskall-Wallis tests were employed to assess significance for pair-wise or multiple comparisons, respectively, taking into account a p-value £0.05 as significant. For the Network analysis cross correlation Pearson matrices for network analysis (metric = Bray-Curtis, method = complete linkage) were generated with in-house scripts (Python v.2.7) and visualized with Gephi v.0.9.2, considering species having a prevalence ³ 20% and a significant Pearson correlation coefficients divided into eight categories to define edge thickness (Li etai, 2008). A network analysis was performed on each dataset using co occurrences and concomitant significance of pair-wise Pearson correlation coefficient, taking care of an optimized visual representation as proposed by current guidelines (Merico et ai, 2009; Berry and Widder, 2014; Faust et ai, 2012a; Faust et ai, 2012b; Lozupone etai, 2012). The degree value, measuring the in/out number of edges linked to a node, and the betweenness centrality, measuring how often a node appears on the shortest paths between pairs of nodes in a network, were computed with Gephi v.0.9.2. Intranetwork communities (here called ‘guilds’) were retrieved using the Blondel community detection algorithm (Blondel et ai, 2008) by means of randomized composition and edge weights, with a resolution equal to 1 (Lambiotte etai, 2014).

Analysis of the impact of antibiotics or tyrosine kinase inhibitors (TKI) on anti-PD-1 mAh efficacy.

Patients who received any oral or intravenous antibiotics within 60 days before the first injection of nivolumab were defined “ATB” and compared to “noATB” patients. The class of antibiotics, the indication route of administration and the duration were collected. Best overall response differences (as defined before) between “ATB” and “noATB” patients were analyzed using Chi-squared test. Prior regimens (anti-angiogenic therapy - TKI, i.e, sunitinib, axitinib or other- or mTOR (mammalian target of rapamycin)) used before starting nivolumab were collected and underwent network analysis to highlight putative differences in the species distribution among ‘guilds’ previously found. The feature group belonging (e.g., TKI, ATB, mTOR) for each species was computed taking into account when the mean relative abundance was higher for that distinctive feature. Nodes within Networks where then re-colored according to the new classification and two different distribution criteria within each guild were computed: i) feature distribution; ii) taxonomical distribution (phylum, class, order level). Chi-square test with Yates correction was used to assess putative differences within the distributions, and a P value less than or equal to 0.05 was considered significant.

Analyses of the bacteria discriminating RCC from control adults (HV) The 69 RCC samples were analyzed and compared with 2994 control adults acquired from publicly available repositories and spanning multiple countries and lifestyles (Pasolli et al., 2017). This set of control samples was enlarged by 54 Italian samples newly acquired and sequenced in the current study. Bacterial species discriminating RCC from HV were determined using LefSe (Segata et al. , 2011).

Analyses of the bacteria discriminating RCC from lung cancer patients (NSCLC).

The 69 RCC samples were analyzed and compared with 118 NSCLC samples acquired and sequenced in the Oncobiotics’ study. Bacterial species discriminating NR from R in NSCLC cohort were determined using MetaOMineR analyses as previously described for RCC cohort.

B. Pre-clinical study details: Mice

All animal experiments were carried out in compliance with French and European laws and regulations. The local institutional animal ethics board and French Ministere de la Recherche approved all mouse experiments (permission numbers: 2016- 049-4646, 2018-078-17530). Experiments were performed in accordance with Government and institutional guidelines and regulations. Female BALB/c were purchased from Janvier (France). Mice were used between 7 and 12 weeks of age. All mouse experiments were performed at the animal facility in Gustave Roussy Cancer Campus where animals were housed in specific pathogen-free conditions. Cell culture, reagents and tumor cell line.

Luciferase-transfected RENCA cell lines (syngeneic for BALB/c mice, kindly provided by Transgene, lllkirch, France) were cultured at 37°C in the presence of 5% C02 in RPMI 1640 containing 10% FCS, 2 mM L-glutamine, 100 Ul/ml penicillin/streptomycin, 1 mM sodium pyruvate and MEM non-essential amino acids (henceforth referred to as complete RPMI 1640). All reagents were purchased from Gibco-lnvitrogen (Carlsbad, CA, USA). Renca cells were maintained in RPMI 1640 medium in the presence of 0.7mg/ml geneticin (G418).

Antibiotic treatments.

Mice were treated with an antibiotic solution (ATB) containing ampicillin (1 mg/ml), streptomycin (5 mg/ml), and colistin (1 mg/ml) (Sigma-Aldrich), with or without the addition vancomycin (0.25 mg/ml) added in the drinking water of mice. Antibiotic activity was confirmed by cultivating fecal pellets resuspended in BHI+15% glycerol at 0.1 g/ml on COS (Columbia Agar with 5% Sheep Blood) plates for 48 h at37°C in aerobic and anaerobic conditions. In brief, in the context of fecal microbial transplantation experiments, mice received 3 days of ATB before undergoing fecal microbial transplantation the next day by oral gavage using animal feeding needles.

Orthotopic lucif erase engineered-renal cell carcinoma (RENCA)

BALB/c mice were anesthetized with isoflurane. A lateral incision was made on the dorsolateral right flank of each mouse, 104 Renca-Luc cells in 30mI_ PBS were injected into the subcapsular space of the right kidney. The skin incision was then closed with surgical clips. Tumor growth was monitored once weekly on an I VIS Imaging System 50 Series (Analytic Jenap). Treatment began on day 7 after tumor inoculation. Mice were injected intraperitoneally 4 times every 4 days with anti-PD-1 (250 mg/mouse; clone RMPI-14) with anti-CTLA-4 mAbs (100 mg of clone 9D9) or anti-PD-1 mAb and axitinib or isotype control mAb (clone 2A3 and clone MPC11, respectively) with or without oral gavage of fecal samples from responding patients or of commensal species.

FMT experiments

Fecal microbiota transfer (FMT) was performed by thawing fecal material. Two hundred pl_ of the suspension was then transferred by oral gavage into ATB pre- treated recipient. In addition, another 100mI_ was applied on the fur of each animal. Two weeks after FMT, tumor cells were injected subcutaneously or orthotopically and mice were treated with anti-PD-1 and CTLA-4 mAbs or anti-PD-1 mAb and axitinib or isotype controls with or without oral gavage of fecal samples from responding patients or of commensal species, as mentioned above.

Gut colonization with commensal species.

A. muciniphila CSUR P2261 and A. indistinctus CSUR P723 were provided by the Institut hospitalo-universitaire Mediterranee Infection, Marseille, France. Bacteroides salyersiae was isolated from the feces of an RCC patient while Bacteroides xylanisolvens was isolated from the ileal mucosa of a colorectal cancer patient. Both patients responded to therapy. Sutterella wadsworthensis was isolated from the ileal mucosa of a non-responder colorectal cancer patient. A. muciniphila was grown on COS plates in an anaerobic atmosphere created using 3 anaerobic generators (Biomerieux) at 37°C for at least 72h. Alistipes indistinctus, Bacteroides salyersiae, Sutterella wadsworthensis and Bacteroides xylanisolvens were also grown on 5% sheep blood enriched Columbia agar (BioMerieux) in an anaerobic atmosphere created using a single anaerobic generator at 37°C for 48h. Bacteria were verified using a Matrix-Assisted Laser Desorption/Ionization Time of Flight (MALDI-TOF) mass spectrometer (Microflex LT analyser, Bruker Daltonics, Germany). Colonization of ATB pre-treated mice was performed by oral gavage with 100 pi of suspension containing 1 ^c 10 ⁸ bacteria. For bacterial gavage: suspensions of 10 ⁹ CFU/mL were obtained using a fluorescence spectrophotometer (Eppendorf) at an optical density of 600 nm in PBS. Five bacterial gavages were performed for each mouse, the first 24 h before the first injection of anti-PD-1 and CTLA-4 mAbs and subsequently four times on the same day anti-PD-1 and CTLA-4 mAbs injections.

Flow cytometry analyses. Tumor-bearer kidneys and spleens were harvested at different time points, 48h days after the second injection of anti-PD-1 + anti-CTLA-4 mAbs into mice bearing RENCA tumors. Excised tumors were cut into small pieces and digested in RPMI medium containing LiberaseTM at 25 pg/mL (Roche) and DNasel at 150 UI/mL (Roche) for 30 minutes at 37°C and then crushed and filtered twice using 70 pm cell strainers (Becton & Dickinson). Spleen were crushed in RPMI medium and subsequently filtered through a 100 pm cell strainer. Four million tumor cells or splenocytes were pre incubated with purified anti-mouse CD16/CD32 (clone 93; eBioscience) for 30 minutes at 4°C, before membrane staining. For intracellular staining, the Foxp3 staining kit (eBioscience) was used. Dead cells were excluded using the Live/Dead Fixable Aqua dead cell stain kit (Life Technologies). Anti-mouse antibodies for CD3 (145-2C11), CD4 CTLA-4 (CD152, UC10-4B9), CD86 (GL1) (BD, BioLegend, R&D and eBioscience) were used to stain cells. Stained samples were acquired on Cytoflex cytometer (Beckman Coulter) and analyses were performed with Kaluza software (Beckman Coulter). T central memory (TCM) gating: after gating on CD3+ alive, CD4+ or CD8+ then, TCM were identified as being CD62L+ and CD44+. Effector memory T (TEM) cells were selected as being CD62L- and CD44+. Treg were gated on CD45+ alive, CD3+, CD4+, CD25+, FoxP3+. Dendritic cells were gated on CD45+ alive, CD3-, Ly6G-, CD11chi, IA/IE+, F4/80-. Macrophages were gated on, CD45+ alive, CD3-, CD11b+ F4/80+. Myeloid-derived suppressor cells (MDSC) were gated after exclusion of

Macrophages, on CD45+ alive, CD3-, CD11b+, Ly6Clo Ly6G+ for G-MDSC (granulocytic) and Ly6Chi Ly6G- for M-MDSC (monocytic). T central memory (TCM) gating: after gating on CD3+ alive, CD4+ were selected excluding CD8+ and CD4+CD8+ then, TCM were identified as being either CD62L+ and CD44+ or CD45RB-. Effector memory T (TEM) cells were selected as being CD62L- and CD44+ or CD45RB-.

Mouse samples for TKI experiment

BALB/c and C57BL6 mice were treated with sunitinib (40mg/Kg/day) or axitinib (30mg/Kg/day) (Diaz-Montero et ai, 2016) or cabozantinib (60mg/Kg/day) (Doran et ai, 2014) or PBS by oral gavage. At least 5 longitudinal stool samples were collected from mice and stored at -80°C until DNA extraction. Preparation and sequencing of mouse fecal samples was performed at IHU Mediterranee Infection, Marseille, France. Briefly, DNA was extracted using two protocols. The first protocol consisted of physical and chemical lysis, using glass powder and proteinase K respectively, then processing using the Macherey-Nagel DNA Tissue extraction kit (Duren, Germany)(Dridi et ai, 2009). The second protocol was identical to the first protocol, with the addition of glycoprotein lysis and deglycosylation steps (Angelakis et ai, 2016). The resulting DNA was sequenced, targeting the V3-V4 regions of the 16S rRNA gene as previously described (Million et ai, 2016). Raw FASTQ files were analyzed with Mothur pipeline v.1.39.5 for quality check and filtering (sequencing errors, chimerae) on a Workstation DELL T7910 (Round Rock, Texas, United States). Raw reads (12692043 in total, on average 127k per sample) were filtered (2949373 in total, on average 30k per sample) and clustered into Operational Taxonomic Units (OTUs), followed by elimination of low-populated OTUs (till 5 reads) and by de novo OTU picking at 97% pair-wise identity using standardized parameters and SILVA rDNA Database v.1.19 for alignment. In all, considering BALB/c and C57BL6 samples, 188 bacterial species were identified. Sample coverage was computed with Mothur and resulted to be on average higher than 99% for all samples, thus meaning a suitable normalization procedure for subsequent analyses. Bioinform atic and statistical analyses on recognized OTUs were performed with Python v.2.7.11. The most representative and abundant read within each OTU (as evidenced in the previous step with Mothur v.1.39.5) underwent a nucleotide Blast using the National Center for Biotechnology Information (NCBI) Blast software (ncbi-blast-2.3.0) and the latest NCBI 16S Microbial 722 Database accessed at the end of April 2019 (ftp://ftp.ncbi.nlm.nih.gov/). A matrix of bacterial relative abundances was built at each taxon level (phylum, class, order, family, genus and species) for subsequent multivariate statistical analyses. Raw data were firstly normalized then standardized using QuantileTransformer and StandardScaler methods from Sci-Kit learn package v0.20.3. Normalization using the output_distribution='normal· option transforms each variable to a strictly Gaussian-shaped distribution, whilst the standardization results in each normalized variable having a mean of zero and variance of one. These two steps of normalization followed by standardization ensure the proper comparison of variables with different dynamic ranges, such as bacterial relative abundances, tumor size, or colonic infiltrate score. Measurements of a diversity (within sample diversity) such as observed_otus and Shannon index, were calculated at OTU level using the SciKit-learn package v.0.4.1. Exploratory analysis of b-diversity (between sample diversity) was calculated using the Bray-Curtis measure of dissimilarity calculated with Mothur and represented in Principal Coordinate Analyses (PCoA), while for Hierarchical Clustering Analysis (HCA) ‘Bray-Curtis’ metrics and ‘complete linkage’ method were implemented using custom scripts (Python v.2.7.11). We implemented Partial Least Square Discriminant Analysis (PLS-DA) and the subsequent Variable Importance Plot (VIP) as a supervised analysis in order to identify the most discriminant bacterial species among the different cohorts of mice treated or not. Where needed, univariate/multivariate statistics and correlation analysis were performed with Python v2.7 and related packages (Scipy, Scikit-learn).

Example 1 : Antibiotics compromise the efficacy of ICB

From February 2016 to September 2018, a total of 85 patients with RCC were enrolled in the NIVOREN trial (Albiges etal., 2018). We collected baseline (T0-T4) feces from 69 patients (Figure 1A). Results from 40 patients were previously reported in a pooled analysis with 60 NSCLC patients in the Science 2018 paper (Routy etal., 2018). Here, RCC have been analyzed for the first time as a single cohort after inclusion of additional patients. The demographic and clinical characteristics of the patients are illustrated in Table 4.

Table 4: Baseline characteristics of renal cell carcinoma patients.

IMDC, International Metastatic Renal Cell Carcinoma Database Consortium (includes: Karnofsky performance status, time from diagnosis to treatment, hemoglobin, serum calcium concentration, neutrophil and platelet counts); ATB, Antibiotics; TKI, tyrosine kinase inhibitor; mTOR, mammalian target of rapamycin

Tumor response was assessed using the Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST v1.1) (Eisenhaueref a/., 2009). Patients who received “ATB” (n=11, 16%) had a lower objective response rate (ORR, number of patients with a complete response and a partial response) compared to the noATB subgroup (9% versus 28%, p<0.03) (Figure 1A-B) and lower PFS and OS (Figure 8). Based on prior studies demonstrating a higher diversity of the gut microbiome in R melanoma patients to anti-PD-1 blockade (Gopalakrishnan et al., 2018) we first compared the median alpha diversity in noATB versus ATB, and observed no significant differences which was preserved across multiple diversity metrics (Shannon or observed OTUs or Simpson index, not shown). We then performed principal coordinate analyses (PCoA) for microbial beta diversity, which provides a measure of the overall relatedness (or lack thereof) between samples. Significant differences separated bacterial species from feces of ATB versus noATB individuals (ANOSIM=0.089; p<0.04) (Figure 1C). Using linear discriminant analysis of effect size (LEfSe) (Segata et al., 2011), coupled to a pairwise comparison of relative taxonomic abundances (for species having a prevalence equal or greater than 20%) within each level using bootstrapping of two-tailed Mann-Whitney U tests (with 1000 permutations and correction for continuity and ties), we concluded that selected bacterial taxa were overrepresented in noATB stools such as Eubacterium rectale (p-0.02) while others were overrepresented in “ATB” fecal materials such as Erysipelotrichaceae bacterium_2_2_44A (p-0.02) and Clostridium hathewayi (p<0.02) (Figure 1D). Altogether, we confirmed that ATB compromised the clinical efficacy of ICB in RCC patients and altered the taxonomic beta diversity and composition of intestinal microbiota.

Example 2: The intestinal microbiota composition predicts clinical outcome to ICB in the cohort that did not take antibiotics

Given the confounding factor of ATB uptake on microbiota composition, we firstly considered only noATB patients (n=58). We started analyzing whether metagenomic profiles of baseline stools (T0-T4) could predict PFS (at 3, 6, 9, 12 months).

The taxonomical annotation of each MGS was performed based on gene homology to previously sequenced organisms (using blastN against the nt and whole genome sequencing (WGS, Meta-Hit) (Li et al., 2014) as well as the MetaPhLan database and pipeline (Segata et al., 2011)). The higher richness (alpha-diversity) of the samples evaluated both at the gene richness count (GC) or metagenomic species (MGS) levels correlated with the clinical response defined by the absence of PD at 12 months after initiation of ICB (Figure 2A). Then, we performed the PCoA (beta-diversity) using a threshold of bacteria prevalence >20%. When segregating patients into non-responders (NR) (i.e., progressive disease (PD) or stable disease (SD) for less than 6 months) and responders (R) to nivolumab according to the best overall response (BOR, the investigator-assessed best response: complete response, partial response, stable disease, or progressive disease), we observed a significant bacterial composition contrasting R versus NR (Figure 2B), with an overrepresentation of distinct species including Akkermansia muciniphila (p<0.02), Bacteroides salyersiae (p-0.04), Eubacterium siraeum (p-0.01) and a trend towards Clostridium ramosum (ns) and Alistipes senegalensis (ns), in R, using both, the MetaPhlAn2 pipeline (Figure 9) or using the MetaOMiner pipeline (Figure 2C) and of Erysipelotrichaceae bacterium_2_2_44A (p<0.01) and Clostridium hathewayi (p<0.01) and Clostridium clostridioforme in NR (ANOSIM=0.727; p<0.0009, Figure 2C) as observed in ATB (Figure 1D). The prevalence and relative abundance of A. muciniphila and B. salyersiae were both higher in R versus NR in RCC patients’ stools, using either one of these catalogs (Figure 2D). Considering higher GC and MGS counts at baseline in patients with PFS longer than 12 months and significant beta-diversity between R and NR (BOR) (Figure 2A,C), we addressed whether paired metagenomic profiles could change over time under ICB therapy by performing a longitudinal analysis of stools (TO, T4, T8, T12) correlating with BOR or PFS at 12 months. When excluding ATB usage (n=58), MGS count was significantly higher in R compared to NR at TO and T4 (Figure 10A, left). At the same way, MGS count was significantly higher in patients with PFS longer than 12 months compared to those with PFS shorter than 12 months at TO and T8 (Figure 10B, left). Interestingly, we observed higher GC only in patients with PFS longer than 12 months compared to those with PFS shorter than 12 months at TO and T8 (Figure 10B, left).

Finally, to perform a robustness test across at least 3 clinical parameters (BOR (with SD > 6 months and partial response), PFS3, PFS6, PFS9, PFS12)), we took into consideration all 69 individuals and we found 27 reliable MGS (out of 1347) contrasting R (n=21) and NR (n=6) (based on the cliff delta for each MGS recovered in >50% tests). Four among these selected MGS are in common with NSCLC microbiome profiles (listed in Table 5), especially encompassing A. muciniphila associated with favorable outcome during anti-PD-1 blockade. Of note, the robustness of MGS for the prediction was superior in the long-term clinical readouts (not shown).

Table 5: Bacteria species associated with resistance to therapy in renal cell carcinoma patients (RCC) and in non-small cell lung cancer patients (NSCLC).

Summary of the contrasting species found in 100 robustness tests (100 Wilcoxon tests on 80% of the total RCC samples randomly sampled, N = 55 samples for each test) and comparison with the contrasting species of the NSCLC cohort (Routy et ai). MGS are selected based on the contrasting tests (MGS need to be significantly contrasted for more than 50% of the robustness tests, N = 98 MGS) and on the number of clinical parameter for which MGS is found contrasted (more than three clinical parameters over nine clinical parameters in total, N = 27 MGS out of the 98). Among these species, 4 also show a signal in NSCLC cohort for at least one tested clinical parameter.

Altogether, we conclude that the alpha and beta diversity of stool composition could be considered to stratify the RCC patient’s population in responder and non-responder and to predict patients with PFS longer than 12 months. Example 3: RCC-associated gut dysbiosis fingerprint

Given the commonalities observed between MGS resulting from ATB- induced dysbiosis and species associated with primary resistance to immunotherapy, and in order to better appreciate the magnitude of intestinal "dysbiosis" in NR (as opposed to R), we analyzed MGS discriminating RCC cancer patients from control adults (HV, n=2994). Significant differences in stool composition were observed between RCC and HV (PCOA not shown, p<0.00t, LEfSe Figure 11). Hence, by merging only significant species in each intersection (ATB yes/no, RCC yes/no, NR yes/no), we only found two distinct species shared between the fecal repertoires of diseased groups (ATB yes, RCC yes, NR yes) i.e C. hathewayi and C. clostridioforme. Conversely, there were no common species shared between the opposite groups. Interestingly, Alistipes senegalensis and C. ramosum were the only 2 common spp. between R and noATB subgroups, while Dorea longicatena, Dorea formicigenerans, Eubacterium rectale and Streptococcus salivarius were all shared between HV and noATB cancer patients (Tables 6-7, Figure 11).

Table 6: Bacteria species associated with resistance to therapy and ATB in renal cell carcinoma patients.

Refers to discriminant species taken from PLS-DA variable importance plot (VIP) drawn to differentiate NR (non-responder) and R (responder) to nivolumab using RECIST1.1 best overall response in the RCC patient cohort. Bacterial species which are in common among NR and “ATB” or among R and “noATB” are in bold. Bacterial species which are in common among NR, RCC and ATB or among Control adults and noATB (data from Tables 4-7 and Figure 9) are bold underlined. sensitivity (R>NR or HV> cancer) to therapy and cancer. Refers to discriminant species taken from PLS-DA variable importance plot (VIP) drawn to differentiate between NR (non-responder) and R (responder) (RCC patient cohort, best overall response) and between RCC and control adults (> 2500 control adults acquired from publicly available repositories and spanning multiple countries and lifestyles, enlarged by 54 Italian samples newly acquired). Bacterial species which are in common among NR and RCC or R and control adults are in bold.

Bacterial species which are in common among NR, RCC and ATB or Control adults and noATB (data from Tables 4-5-Q-7 and Figure 9) are bold underlined.

Example 4: Prior tyrosine kinase inhibitors (TKI) and ATB use are associated with distinct qut microbiota ‘guilds’ in RCC patients

The majority of RCC patients (n=55, 80%) received two previous lines of treatment for their advanced RCC before starting nivolumab (Table 4). Sunitinib (n=49, 71%) or axitinib (n=13, 19%) were the most frequent previous TKI. Co-occurrence network analysis revealed six ‘species interaction groups’ referred as ‘SIG’(Zhao et ai, 2018), highlighting that i) ATB and axitinib were the most powerful medications shifting fecal microbiota (using cross-validation model, predictive power for ATB=84%; for axitinib=81%; for sunitinib=69%); ii) defined bacterial species drove the stratification of the whole RCC network into ‘SIG’, such as A. muciniphila for R and Dorea formicigenerans for noATB (random forest analysis) (Figure 12).

Altogether, the stool composition of RCC patients is influenced by ATB and axitinib and distinct species, missing during cancer development or ATB uptake, appear associated with BOR and PFS during anti-PD-1 blockade ("immunostimulatory" D. longicatena) while others, specifically selected following ATB administration and the cancer status (C. hathewayi), may confer primary resistance to this therapy and will be referred to as "immunotolerant" henceforth.

Example 5: Oral gavage with immunostimulatory or beneficial commensals or feces from responding RCC patients rescues primary resistance in RCC tumor bearing mice

To further provide evidence of a cause-effect relationship between bacterial fecal composition and therapy outcome, we humanized BALB/c mice sterilized by ATB with RCC patient stools, 15 days prior to orthotopic inoculation of luciferase engineered-RENCA (Figure 3A). Transfer of 15 FMT (5R and 10NR) patient stools by oral gavage (referred to as "FMT" henceforth) in ATB-treated avatar mice that were subsequently implanted with RENCA induced significant responses (for FMT R) or resistance (for FMT NR) to CICB. It should be noted that we observed only 27% of exceptions of concordance between patient’s response and mouse recipient’s response to ICB: only 4 stools above 15 FMT used (Table 8, Figure 3A-B). However, compensation of NR-FMT (that did not contain A. muciniphila or B. salyersiae ) with oral administration of immunostimulatory A. muciniphila or B. salyersiae or R-FMT prior to each CICB cycle restored sensitivity to therapy, as evidenced by kidney weight at sacrifice (Figure 3D) and decreased luminescence (Figure 3E)). Despite strong co- occurrence of B. salyersiae with other commensal species (Figure 13A-B) varying in their identity in the R versus NR networks, the antitumor efficacy of the former bacterium was not boosted by coadministration of a neighboring species (Figure 13C).

In conclusion, bacteria contrasting R and NR in our 69 RCC cohort compensate the lack of responsiveness observed with NR-FMT in avatar mice, establishing cause-effect relationship between favorable bacterial composition of feces and clinical outcome.

Table 8: Patient stools mostly retain best overall response (BOR) in mice host after

FMT.

SPF BALB/c mice were gavaged with fecal material (FMT) from 15 patients: 5 responders (R) and 10 non-responders (NR) patient donors (RCC patient cohort, best overall response). We calculated fold ratio (FR) of total flux D15/D0 among mice treated with anti-PD1 plus anti-CTLA-4 (CICB) and control (Ctrl). Underlined the discrepancies between human outcome and mice outcome.

Example 6: The gut microbiota controls the cancer-immune set point in RCC tumor bearing mice

To analyze how NR FMT could influence the systemic (spleen) and tumor (RENCA) immune tonus or contexture, we performed multicolor flow cytometric analyses of splenocytes 48h after the second cycle of CICB in five NR FMT (from independent donors). Indeed, we observed major differences in the relative capacity of each FMT to influence the splenic residence of CD103 ⁺ XCR1 ⁺ cross-presenting DC and effector IFNy-producing CD4 ⁺ and CD8 ⁺ Th1 or Tc1 lymphocytes, according to patient's stool composition (Figure 4A-B). Indeed, donor stools containing immunotolerant spp. ( Clostridia ) clustered with CD103+ DC and anticorrelated with those containing immunostimulatory spp. (A. senegalensis and D. longicatena) inducing Ly6C ^hi9h myeloid cells and CD4+ T cells. Moreover, homeostatic bacteria (D. formicigenerans ) enriched feces resulted in the accumulation of splenic Th1 cells and Ly6C ^low macrophages, in contrast to stools containing E. rectale and S. salivarius clustering with high abundancy of Tc1 and CD103+CD11b+ DC (Figure 4A, Figure 14A). Similar influences were observed in the TME (Figure 4C-D, Figure 14B), stools containing immunostimulatory spp. (A. senegalensis and D. longicatena ) inducing Th1 tumor infiltrating lymphocytes (TIL), a phenomenon anti-correlated with the fecal presence of immunotolerant Clostridia spp. or S. salivarius in the donor material (Figure 4F-G-B-C). During CICB therapy, the induction of tumor immunosurveillance based on CD3+ and CD4+ accumulation in TIL was blunted after transfer of stools containing Clostridia spp. while presence of A. senegalensis and D. longicatena clustered together and were associated with increased CD3+, CD4+, CD8+ and Tc1 TIL accumulation (Figure 5A-B). This tumor contexture mirrored the strong association between stool immunostimulatory bacteria and Th1/Tc1 systemic immunity in contrast to immunotolerant commensals that favored the overrepresentation of myeloid cells (Figure 14).

To illustrate the pathophysiological relevance of the NR FMT RENCA avatar model system, we show first that the CICB/Ctrl ratio of the bioluminescence flux in the retroperitoneum of avatar mice was significantly proportional to the relative abundance of the immunotolerant versus immunostimulatory commensals; correlated and anticorrelated with C. hathewayi or C. clostridioforme versus D. longicatena respectively (Figure 5C). Secondly, oral gavage with B. salyersiae to compensate for FMT NR-mediated immunomodulation culminated in induction of splenic CD4+PD1+T cells and Tc1 TIL proportional to the relative abundance of A. senegalensis in donor stools while CD8+ TIL were correlated with the relative abundance of D. longicatena (Figure 5D). Finally, Kaplan Meier survival curves illustrated the clinical relevance of some of these commensals for PFS during anti-PD-1 blockade, namely D. longicatena associated with longer PFS, and presence of C. hathewayi or absence of both A. muciniphila+B. salyersiae for shorter PFS (Figure 5E).

Altogether, we infer from these findings that the relative abundance of immuno-stimulatory versus -tolerant commensals will govern the cancer-immune set point of tumor bearers, paving the way to CICB-induced tumor control.

Example 7: Antiangiogenic tyrosine kinase inhibitors (TKI) induce an immuno- stimulatory intestinal microbiome shift

Data from the co-occurrence network analysis revealed six species interaction groups called ‘SIG’ (Figure 12). Interestingly, axitinib (like ATB) appeared to markedly influence SIG distribution within network topology (RF importance), more specifically SIG2, centered by Odoribacter splanchnicus, belonging to the same community as Dorea longicatena (Figure 12). To assess the distinct bacteria related to TKI, we compared a subgroup of patient who taken TKI in 1L (within our 69 RCC patients’ stools, regardless of ATB) with HV. An overrepresentation of A. senegalensis and A. muciniphila induced by TKI (LEfSe, Figure 6A) was observed in these patients. LEfSe performed to assess distinct species associated with TKI versus mTOR inhibitors taken as 2L therapy in subgroup analysis within our 69 RCC patients’ stools (regardless of ATB) revealed significant beta diversities contrasting these 2 subgroups for fecal composition and a trend for an overrepresentation of A. senegalensis induced by TKI (not shown). Since we enrolled patients after failure of 1L (or more) TKI, feces collection preceding introduction of TKI were not available to uncouple the effects of tumor progression from that of TKI on the microbiome shift. To circumvent this limitation, we administered in two mouse genetic backgrounds a tumoricidal antiangiogenic dose of various TKI (sunitinib, axitinib, or cabozantinib) over 3 weeks and collected longitudinally stools. Strikingly, all three TKI markedly induced significant changes in the alpha and beta diversity of the microflora over time, in both BALB/c (Figure 6B) and C57BL/6 mice with a common dominant deviation of the microbiota composition (Figure 15). In BALB/c intestines, there was a prototypic TKI signature, with an over representation of Eubacterium coprostanoligenes, Vampirovibrio chlorellavorus, Longibaculum muris, Parabacteroides goldsteinii, Alistipes timonensis, Faecalicatena contorta, with a relative lower dominance of Neglecta timonensis, Adlercreutzia equolifaciens, and Bacteroides fragilis at 15 days of all three TKI uptake (mean VIP score). Importantly, sunitinib and cabozantinib favored a higher abundance of immunostimulatory Alistipes senegalensis as observed in humans (Figure 6A-B). Accordingly, in C57BL/6 intestines, there was an over representation of the immunostimulatory E. siraeum, among other species shared by all three TKI (Figure 15). Importantly, TKI favored a higher abundance of immunostimulatory A. senegalensis and A. muciniphila (Figure 6B), especially for cabozantinib. Overall, TKI induced a significant and prototypic microbiota shift including immunostimumatory commensals (such as E. siraeum, A. senegalensis, A. muciniphila) that could be harnessed to improve the efficacy of ICB in RCC patients.

In patients, axitinib and sunitinib-induced microbiome shifts could be contrasted, with axitinib favoring the immunogenic A. senegalensis and C. ramosum. (Figure 7A). Therefore, to circumvent resistance to ICB in RENCA (Routy et ai, 2018), we undertook experiments using axitinib alone or in combination with A. muciniphila. Indeed, we observed a markedly increased efficacy combining axitinib with ICB and A. muciniphila in tumor bearers with or without FMT NR (Figure 7B-C). Overall, TKI induced a significant and prototypic microbiome shift including immunostimumatory commensals that could be harnessed to improve the efficacy of ICB in RCC patients. Discussion

RCC encompasses a wide spectrum of morphologically and molecularly distinct cancer subtypes. The introduction of targeted therapies (inhibiting VEGF, PFGF, MET, AXL tyrosine kinases) and immune checkpoint inhibitors into clinical practice has markedly improved the median overall survival (OS) in clear cell RCC patients, the most common subtype. With 12 approved drugs acting through 6 different effective mechanisms, novel biomarkers are needed to stratify and simplify this therapeutic landscape, to improve efficacy and reduce side effects. Based on pan-omics approaches integrating genetics, transcriptomics and immunoscoring, molecular stratifications of RCC identified subgroups of patients with dismal prognosis that may benefit more specifically from antiangiogenic or immunotherapies (Casuscelli et ai, 2017). However, it appears that some tumors are a desert of immune reactivities while others are invaded with overt inflammatory and/or exhausted cell infiltrates that do not convey long term protection, suggesting that the immune tonus of RCC patients is not properly triggered or controlled.

Our study highlights the potential of harnessing the intestinal microbiome to better control the "cancer-immune set point" (Chen and Mellman, 2017), i.e. , the threshold beyond which ICB triggers a clinical benefit. Mapping the gut holobiont to identify a minimalist ecosystem governing the cancer-immune set point and more specifically immunogenic versus tolerogenic commensals and medications tilting their balance remains an open conundrum. By applying various bioinformatic and clinical subgroup analyses (LEfSe, PLS-DA VIP, networks), we identified a limited set of species (phylum Firmicutes, family Clostridiaceae, species C. clostridioforme, C. hathewayi) that were associated with primary resistance and enriched by ATB use and metastatic cancer status.

The “C. clostridioforme group” comprises three principal species that differ in virulence and antimicrobial susceptibility despite similar colony and microscopic morphology. C. bolteae and C. clostridioforme are observed with approximately equal frequency, but C. hathewayi is seen with much greater frequency (Dababneh etai, 2014; Finegold et ai., 2005). Infections with the “C. clostridioforme group” are the second most frequently isolated species of Clostridium, after Clostridium perfringens (Dababneh et ai, 2014; Finegold et ai, 2005). C. hathewayi has been reported to be part of the pathobionts associated with the diagnosis of colon cancers (Liang et at. , 2017) and could mitigate antigen-specific T cell responses in mice (Rossi et ai., 2016).

Conversely, we identified some commensals associated with favorable prognosis and the intestinal homeostatic status, which belong to Eubacteriaceae (E. rectale, E. siraeum), Lachnospiraceae ( Dorea longicatena ), Verrucomicrobioaceae (A. muciniphila) families and to the Bacteroidales order ( Rikenellaceae family/Alistipes/Alistipes senegalensis, Bacteroidaceae family/Bacteroides/ Bacteroides salyersiae). While A. senegalensis and A. muciniphila alone or together within minimalist communities were clearly associated with the elicitation of adaptive immune responses beneficial against murine cancers (Routy et ai, 2018; Tanoue et ai, 2019), Eubacteriaceae and Dorea longicatena have been described as pivotal to keep in check the homeostasis of the intestinal barrier (Kamo et ai, 2017).

Experiments initially conducted in mice showed that broad-spectrum ATB blunt the activity of ICB against a wide range of transplantable and orthotopic tumors, suggesting that a minimalist intestinal ecosystem is required for the function of the mammalian host immune system. These pioneering observations in preclinical models encouraged retrospective analyses in cancer patients to determine if premedication with ATB would influence the clinical response to ICB. In the literature, 11 retrospective analyses assessed the impact of ATB taken shortly before or after the initiation of ICB on clinical outcome of patients treated with ICB in several malignancies. Eleven out of the 12 analyses reported a negative impact of ATB uptake in PFS and/or OS, mirroring the murine data (Derosa et ai, 2018; Elkrief et ai., 2019; Routy et ai., 2018). However, the impact of these puzzling findings on the clinical management of cancer patients remains controversial. Here, we describe how ATB (mostly betalactams and quinolones) affect the intestinal composition of feces of 69 RCC patients. ATB markedly affected the beta diversity, leading to the underrepresentation of Eubacteriaceae family members as already described (Raymond et ai, 2016) (such as Eubacterium rectale) for the benefit of pathobiont species ( Erysipelotrichaceae bacterium_2_2_44A and Clostridium hathewayi). This microbiome shift is associated with reduced ORR during ICB therapy (73% of primary resistance in ATB versus 33% in the no ATB subgroups, p<0.03).

Given the incidence of gastrointestinal toxicity associated with TKI, pioneering studies investigated TKI-induced dysbiosis and the impact of ATB on diarrhea and survival. Pal et ai. evaluated a population of 20 RCC patients receiving VEGF-TKI and reported a positive and negative association between Bacteroides spp. and Prevotella spp. and diarrhea, respectively (Pal et ai., 2015). When comparing their TKI- RCC stool data with those from HV, they observed a relative loss of Bifidobacterium spp. Accordingly, Gong et at. followed up 5 RCC patients treated with TKI and showed that Bacteroides, Barnesiella and Phascolarctobacterium were elevated in responders while Bifidobacterium were elevated in non-responders (Gong et ai, 2019). However, in parallel, Hahn et at. showed that ATB targeting stool Bacteroides spp. improved PFS in patients receiving 1L VEGF-TKI in a duration-dependent manner (Hahn et ai, 2018). Our data fuel this hypothesis of an unconventional mode of action of VEGF-TKI whereby a treatment-induced prototypic gut microbiome fingerprint might influence therapeutic outcome. We observed a relative loss of Bifidobacterium and overrepresentation of distinct species of the Bacteroidales order (A. timonensis, P. goldsteinii) post-TKI in naive mice and showed that axitinib could compensate NR FMT induced dysbiosis and reduced responsiveness to ICB, in a microbiota-dependent manner.

Limitations of our study include that this conclusion relies on a single cohort of 69 RCC patients including only 11 cases who took ATB and in 2L therapy with the interference of many confounding factors (prior therapies, comedications, and other factors such as hemoglobin (Maier etal., 2018; Pasolli etai, 2019)). Prospective studies in 1L therapy should validate this fingerprint as a new predictor of primary resistance to ICB.

REFERENCES

Albiges L, Negrier S, Dalban C, Gravis G, Chevreau C, Oudard S, et al. Safety and efficacy of nivolumab in metastatic renal cell carcinoma (mRCC): Results from the NIVOREN GETUG-AFU 26 study. J Clin Oncol 2018;36:577-577. doi:10.1200/JC0.2018.36.6_suppl.577.

Angelakis E, Bachar D, Henrissat B, Armougom F, Audoly G, Lagier JC, Robert C, Raoult D (2016) Glycans affect DNA extraction and induce substantial differences in gut metagenomic studies. - Sci Rep. May 18;6:26276 Ascierto, M.L., McMiller, T.L., Berger, A.E., Danilova, L., Anders, R.A., Netto, G.J., Xu, H., Pritchard, T.S., Fan, J., Cheadle, C., etal. (2016). The Intratumoral Balance between Metabolic and Immunologic Gene Expression Is Associated with Anti-PD-1 Response in Patients with Renal Cell Carcinoma. Cancer Immunol. Res.

Becht, E., Giraldo, N.A., Beuselinck, B., Job, S., Marisa, L., Vano, Y., Oudard, S., Zucman-Rossi, J., Laurent-Puig, P., Sautes-Fridman, C., et al. (2015). Prognostic and theranostic impact of molecular subtypes and immune classifications in renal cell cancer (RCC) and colorectal cancer (CRC). Oncoimmunology 4, e1049804.

Becht, E., Giraldo, N.A., Lacroix, L., Buttard, B., Elarouci, N., Petitprez, F., Selves, J., Laurent-Puig, P., Sautes-Fridman, C., Fridman, W.H., et al. (2016). Estimating the population abundance of tissue-infiltrating immune and stromal cell populations using gene expression. Genome Biol. 17, 218.

Berry D, Widder S. Deciphering microbial interactions and detecting keystone species with co-occurrence networks. Front Microbiol 2014;5:219. doi:10.3389/fmicb.2014.00219. Beuselinck, B., Job, S., Becht, E., Karadimou, A., Verkarre, V., Couchy, G., Giraldo, N., Rioux-Leclercq, N., Molinie, V., Sibony, M., etal. (2015). Molecular subtypes of clear cell renal cell carcinoma are associated with sunitinib response in the metastatic setting. Clin. Cancer Res. Off. J. Am. Assoc. Cancer Res. 21, 1329-1339.

Blondel VD, Guillaume J-L, Lambiotte R, Lefebvre E. Fast unfolding of communities in large networks. J Stat Mech Theory Exp 2008;2008:P10008. doi: 10.1088/1742- 5468/2008/10/PI 0008.

Casuscelli, J., Vano, Y.-A., Fridman, W.H., and Hsieh, J.J. (2017). Molecular Classification of Renal Cell Carcinoma and Its Implication in Future Clinical Practice. Kidney Cancer 1, 3-13. Chen, D.S., and Mellman, I. (2017). Elements of cancer immunity and the cancer- immune set point. Nature 541, 321-330.

Chevrier, S., Levine, J.H., Zanotelli, V.R.T., Silina, K., Schulz, D., Bacac, M., Ries, C.H., Allies, L., Jewett, M.A.S., Moch, H., et al. (2017). An Immune Atlas of Clear Cell Renal Cell Carcinoma. Cell 169, 736-749. e18. Criscuolo A, Brisse S. AlienTrimmer: a tool to quickly and accurately trim off multiple short contaminant sequences from high-throughput sequencing reads. Genomics 2013;102:500-6. doi:10.1016/j.ygeno.2013.07.011. Cotillard A, Kennedy SP, Kong LC, Prifti E, Pons N, Le Chatelier E, et al. Dietary intervention impact on gut microbial gene richness. Nature 2013;500:585-8. doi: 10.1038/nature12480.

Dababneh, A.S., Nagpal, A., Palraj, B.R.V., and Sohail, M.R. (2014). Clostridium hathewayi bacteraemia and surgical site infection after uterine myomectomy. BMJ Case Rep. 2014.

Daillere, R., Vetizou, M., Waldschmitt, N., Yamazaki, T., Isnard, C., Poirier-Colame, V., Duong, C.P.M., Flament, C., Lepage, P., Roberti, M.P., etal. (2016). Enterococcus hirae and Barnesiella intestinihominis Facilitate Cyclophosphamide-Induced Therapeutic Immunomodulatory Effects. Immunity 45, 931-943.

Derosa, L., Hellmann, M.D., Spaziano, M., Halpenny, D., Fidelle, M., Rizvi, H., Long, N., Plodkowski, A.J., Arbour, K.C., Chaft, J.E., et al. (2018). Negative association of antibiotics on clinical activity of immune checkpoint inhibitors in patients with advanced renal cell and non-small-cell lung cancer. Ann. Oncol. Off. J. Eur. Soc. Med. Oncol. 29, 1437-1444.

Diaz-Montero CM, Mao FJ, Barnard J, Parker Y, Zamanian-Daryoush M, Pink JJ, et al. MEK inhibition abrogates sunitinib resistance in a renal cell carcinoma patient-derived xenograft model. Br J Cancer 2016;115:920-8. doi:10.1038/bjc.2016.263.

Doran MG, Spratt DE, Wongvipat J, Ulmert D, Carver BS, Sawyers CL, et al. Cabozantinib resolves bone scans in tumor-naive mice harboring skeletal injuries. Mol Imaging 2014;13. doi: 10.2310/7290.2014.00026.

Dridi B, Henry M, Khechine A, Raoult D, Drancourt M. (2009) High prevalence of Methanobrevibacter smithii and Methanosphaera stadtmanae detected in the human gut using an improved DNA detection protocol. PLoS One. 2009 Sep 17;4(9):e7063. doi: 10.1371/journal. pone.0007063

Eisenhauer, E.A., Therasse, P., Bogaerts, J., Schwartz, L.H., Sargent, D., Ford, R., Dancey, J., Arbuck, S., Gwyther, S., Mooney, M., etal. (2009). New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur. J. Cancer Oxf. Engl. 199045, 228-247. Elkrief, A., El Raichani, L., Richard, C., Messaoudene, M., Belkaid, W., Malo, J., Belanger, K., Miller, W., Jamal, R., Letarte, N., et al. (2019). Antibiotics are associated with decreased progression-free survival of advanced melanoma patients treated with immune checkpoint inhibitors. Oncoimmunology 8, e1568812.

Escudier, B., Farace, F., Angevin, E., Charpentier, F., Nitenberg, G., Triebel, F., and Hercend, T. (1994). Immunotherapy with interleukin-2 (IL2) and lymphokine-activated natural killer cells: improvement of clinical responses in metastatic renal cell carcinoma patients previously treated with IL2. Eur. J. Cancer Oxf. Engl. 199030A, 1078-1083.

Faust K, Sathirapongsasuti JF, Izard J, Segata N, Gevers D, Raes J, et al. Microbial Co occurrence Relationships in the Human Microbiome. PLOS Comput Biol 2012;8:e1002606. doi:10.1371/journal.pcbi.1002606.

Faust K, Raes J. Microbial interactions: from networks to models. Nat Rev Microbiol 2012; 10:538-50. doi: 10.1038/nrmicro2832. Finegold, S.M., Song, Y., Liu, C., Hecht, D.W., Summanen, P., Kononen, E., and Allen, S.D. (2005). Clostridium clostridioforme: a mixture of three clinically important species. Eur. J. Clin. Microbiol. Infect. Dis. 24, 319-324.

Giraldo, N.A., Becht, E., Pages, F., Skliris, G., Verkarre, V., Vano, Y., Mejean, A., Saint- Aubert, N., Lacroix, L., Natario, I., et al. (2015). Orchestration and Prognostic Significance of Immune Checkpoints in the Microenvironment of Primary and Metastatic Renal Cell Cancer. Clin. Cancer Res. Off. J. Am. Assoc. Cancer Res. 21, 3031-3040.

Giraldo, N.A., Becht, E., Vano, Y., Petitprez, F., Lacroix, L., Validire, P., Sanchez-Salas, R., Ingels, A., Oudard, S., Moatti, A., etal. (2017). Tumor-Infiltrating and Peripheral Blood T-cell Immunophenotypes Predict Early Relapse in Localized Clear Cell Renal Cell Carcinoma. Clin. Cancer Res. Off. J. Am. Assoc. Cancer Res. 23, 4416-4428.

Gong, J., Noel, S., Pluznick, J.L., Hamad, A.R.A., and Rabb, H. (2019). Gut Microbiota- Kidney Cross-Talk in Acute Kidney Injury. Semin. Nephrol. 39, 107-116.

Gopalakrishnan, V., Spencer, C.N., Nezi, L., Reuben, A., Andrews, M.C., Karpinets, T.V., Prieto, P.A., Vicente, D., Hoffman, K., Wei, S.C., etal. (2018). Gut microbiome modulates response to anti-PD-1 immunotherapy in melanoma patients. Science 359, 97-103.

Godon J J, Zumstein E, Dabert P, Habouzit F, Moletta R. Molecular microbial diversity of an anaerobic digestor as determined by small-subunit rDNA sequence analysis. Appl Environ Microbiol 1997;63:2802-13. Hahn, A.W., Froerer, C., VanAlstine, S., Rathi, N., Bailey, E.B., Stenehjem, D.D., and Agarwal, N. (2018). Targeting Bacteroides in Stool Microbiome and Response to Treatment With First-Line VEGF Tyrosine Kinase Inhibitors in Metastatic Renal-Cell Carcinoma. Clin. Genitourin. Cancer 16, 365-368.

Kamo, T., Akazawa, H., Suda, W., Saga-Kamo, A., Shimizu, Y., Yagi, H., Liu, Q., Nomura, S., Naito, A.T., Takeda, N., et al. (2017). Dysbiosis and compositional alterations with aging in the gut microbiota of patients with heart failure. PloS One 12, e0174099.

Kroemer, G., and Zitvogel, L. (2018). Cancer immunotherapy in 2017: The breakthrough of the microbiota. Nat. Rev. Immunol. 18, 87-88. Lambiotte R, Delvenne J-C, Barahona M. Random Walks, Markov Processes and the Multiscale Modular Organization of Complex Networks. IEEE Trans Netw Sci Eng 2014;1:76-90. doi:10.1109/TNSE.2015.2391998.

Langmead B, Salzberg SL. Fast gapped-read alignment with Bowtie 2. Nat Methods 2012;9:357-9. doi:10.1038/nmeth.1923. Le Chatelier E, Nielsen T, Qin J, Prifti E, Hildebrand F, Falony G, et al. Richness of human gut microbiome correlates with metabolic markers. Nature 2013;500:541-6. doi: 10.1038/nature12506.

Lee HL, Shen H, Hwang IY, Ling H, Yew WS, Lee YS, Chang MW. (2018) Targeted Approaches for In Situ Gut Microbiome Manipulation. Genes (Basel). Jul 12;9(7). Li M, Wang B, Zhang M, Rantalainen M, Wang S, Zhou H, etal. Symbiotic gut microbes modulate human metabolic phenotypes. Proc Natl Acad Sci U S A 2008;105:2117-22. doi: 10.1073/pnas.0712038105.

Li, J., Jia, H., Cai, X., Zhong, H., Feng, Q., Sunagawa, S., Arumugam, M., Kultima, J.R., Prifti, E., Nielsen, T., etal. (2014). An integrated catalog of reference genes in the human gut microbiome. Nat. Biotechnol. 32, 834-841.

Liang, Q., Chiu, J., Chen, Y., Huang, Y., Higashimori, A., Fang, J., Brim, H., Ashktorab, H., Ng, S.C., Ng, S.S.M., et al. (2017). Fecal Bacteria Act as Novel Biomarkers for Noninvasive Diagnosis of Colorectal Cancer. Clin. Cancer Res. Off. J. Am. Assoc. Cancer Res. 23, 2061-2070.

Lozupone CA, Stombaugh Jl, Gordon Jl, Jansson JK, Knight R. Diversity, stability and resilience of the human gut microbiota. Nature 2012;489:220-30. doi: 10.1038/naturel 1550.

Maier, L., Pruteanu, M., Kuhn, M., Zeller, G., Telzerow, A., Anderson, E.E., Brochado, A.R., Fernandez, K.C., Dose, H., Mori, H., et al. (2018). Extensive impact of non antibiotic drugs on human gut bacteria. Nature 555, 623-628.

Matson, V., Fessler, J., Bao, R., Chongsuwat, T., Zha, Y., Alegre, M.-L., Luke, J.J., and Gajewski, T.F. (2018). The commensal microbiome is associated with anti-PD-1 efficacy in metastatic melanoma patients. Science 359, 104-108. Merico D, Gfeller D, Bader GD. How to visually interpret biological data using networks. Nat Biotechnol 2009;27:921-4. doi: 10.1038/nbt.1567.

Million, M. et al. (2016) Increased Gut Redox and Depletion of Anaerobic and Methanogenic

Prokaryotes in Severe Acute Malnutrition. Sci. Rep. 6, 26051; doi: 10.1038/srep26051 Motzer, R.J., Escudier, B., McDermott, D.F., George, S., Hammers,

H.J., Srinivas, S., Tykodi, S.S., Sosman, J.A., Procopio, G., Plimack, E.R., etal. (2015). Nivolumab versus Everolimus in Advanced Renal-Cell Carcinoma. N. Engl. J. Med. 373, 1803-1813.

Motzer, R.J., Tannir, N.M., McDermott, D.F., Aren Frontera, O., Melichar, B., Choueiri, T.K., Plimack, E.R., Barthelemy, P., Porta, C., George, S., etal. (2018). Nivolumab plus

Ipilimumab versus Sunitinib in Advanced Renal-Cell Carcinoma. N. Engl. J. Med. 378, 1277-1290.

Motzer, R.J., Penkov, K., Haanen, J., Rini, B., Albiges, L., Campbell, M.T., Venugopal, B., Kollmannsberger, C., Negrier, S., Uemura, M., et al. (2019). Avelumab plus Axitinib versus Sunitinib for Advanced Renal-Cell Carcinoma. N. Engl. J. Med. 380, 1103-1115.

Nielsen HB, Almeida M, Juncker AS, Rasmussen S, Li J, Sunagawa S, et al. Identification and assembly of genomes and genetic elements in complex metagenomic samples without using reference genomes. Nat Biotechnol 2014;32:822-8. doi: 10.1038/nbt.2939. Pal, S.K., Li, S.M., Wu, X., Qin, H., Kortylewski, M., Hsu, J., Carmichael, C., and Frankel, P. (2015). Stool Bacteriomic Profiling in Patients with Metastatic Renal Cell Carcinoma Receiving Vascular Endothelial Growth Factor-Tyrosine Kinase Inhibitors. Clin. Cancer Res. Off. J. Am. Assoc. Cancer Res. 21, 5286-5293.

Pasolli E, Schiffer L, Manghi P, Renson A, Obenchain V, Truong DT, et al. Accessible, curated metagenomic data through ExperimentHub. Nat Methods 2017;14:1023-4. doi: 10.1038/nmeth.4468.

Pasolli, E., Asnicar, F., Manara, S., Zolfo, M., Karcher, N., Armanini, F., Beghini, F., Manghi, P., Tett, A., Ghensi, P., et al. (2019). Extensive Unexplored Human Microbiome Diversity Revealed by Over 150,000 Genomes from Metagenomes Spanning Age, Geography, and Lifestyle. Cell 176, 649-662. e20. Porta, C., and Rizzo, M. (2019). Immune-based combination therapy for metastatic kidney cancer. Nat. Rev. Nephrol. 15, 324-325.

Ramachandran G, Bikard D. (2019) Editing the microbiome the CRISPR way. Philos Trans R Soc Lond B Biol Sci. 2019 May 13;374(1772):20180103.

Raymond, F., Ouameur, A.A., Deraspe, M., Iqbal, N., Gingras, H., Dridi, B., Leprohon, P., Plante, P.-L., Giroux, R., Berube, E., et al. (2016). The initial state of the human gut microbiome determines its reshaping by antibiotics. ISME J. 10, 707-720.

Rini, B.I., Plimack, E.R., Stus, V., Gafanov, R., Hawkins, R., Nosov, D., Pouliot, F., Alekseev, B., Soulieres, D., Melichar, B., et al. (2019a). Pembrolizumab plus Axitinib versus Sunitinib for Advanced Renal-Cell Carcinoma. N. Engl. J. Med. 380, 1116-1127. Rini, B.I., Powles, T., Atkins, M.B., Escudier, B., McDermott, D.F., Suarez, C., Bracarda, S., Stadler, W.M., Donskov, F., Lee, J.L., etal. (2019b). Atezolizumab plus bevacizumab versus sunitinib in patients with previously untreated metastatic renal cell carcinoma (IMmotion151): a multicentre, open-label, phase 3, randomised controlled trial. Lancet Lond. Engl. Rosenberg, S.A., Lotze, M.T., Yang, J.C., Topalian, S.L., Chang, A.E., Schwartzentruber, D.J., Aebersold, P., Leitman, S., Linehan, W.M., Seipp, C.A., et al. (1993). Prospective Randomized Trial of High-Dose lnterleukin-2 Alone or in Conjunction With Lymphokine-Activated Killer Cells for the Treatment of Patients With Advanced Cancer. JNCI J. Natl. Cancer Inst. 85, 622-632. Rossi, O., van Berkel, L.A., Chain, F., Tanweer Khan, M., Taverne, N., Sokol, H., Duncan, S.H., Flint, H.J., Harmsen, H.J.M., Langella, P., etal. (2016). Faecalibacterium prausnitzii A2-165 has a high capacity to induce IL-10 in human and murine dendritic cells and modulates T cell responses. Sci. Rep. 6, 18507.

Routy, B., Le Chatelier, E., Derosa, L., Duong, C.P.M., Alou, M.T., Daillere, R., Fluckiger, A., Messaoudene, M., Rauber, C., Roberti, M.P., et al. (2018). Gut microbiome influences efficacy of PD-1 -based immunotherapy against epithelial tumors. Science 359, 91-97.

Segata, N., Izard, J., Waldron, L., Gevers, D., Miropolsky, L., Garrett, W.S., and Huttenhower, C. (2011). Metagenomic biomarker discovery and explanation. Genome Biol. 12, R60. Sivan, A., Corrales, L, Hubert, N., Williams, J.B., Aquino-Michaels, K., Earley, Z.M., Benyamin, F.W., Lei, Y.M., Jabri, B., Alegre, M.-L., et al. (2015). Commensal Bifidobacterium promotes antitumor immunity and facilitates anti-PD-L1 efficacy. Science 350, 1084-1089. Suau A, Bonnet R, Sutren M, Godon JJ, Gibson GR, Collins MD, etal. Direct analysis of genes encoding 16S rRNA from complex communities reveals many novel molecular species within the human gut. Appl Environ Microbiol 1999;65:4799-807.

Tanoue, T., Morita, S., Plichta, D.R., Skelly, A.N., Suda, W., Sugiura, Y., Narushima, S., Vlamakis, H., Motoo, I., Sugita, K., etal. (2019). A defined commensal consortium elicits CD8 T cells and anti-cancer immunity. Nature 565, 600.

Vetizou, M., Pitt, J.M., Daillere, R., Lepage, P., Waldschmitt, N., Flament, C., Rusakiewicz, S., Routy, B., Roberti, M.P., Duong, C.P.M., et al. (2015). Anticancer immunotherapy by CTLA-4 blockade relies on the gut microbiota. Science 350, 1079- 1084. Zhao, L, Zhang, F., Ding, X., Wu, G., Lam, Y.Y., Wang, X., Fu, H., Xue, X., Lu, C., Ma, J., et al. (2018). Gut bacteria selectively promoted by dietary fibers alleviate type 2 diabetes. Science 359, 1151-1156.