Abstract The gut microbiota and tumor-associated macrophages (TAMs) affect tumor responses to anti–programmed cell death protein 1 (PD-1) immune checkpoint blockade. Reprogramming TAM by either blocking or deleting the macrophage receptor triggering receptor on myeloid cells 2 (TREM2) attenuates tumor growth, and lack of functional TREM2 enhances tumor elimination by anti–PD-1. Here, we found that anti–PD-1 treatment combined with TREM2 deficiency in mice induces proinflammatory programs in intestinal macrophages and a concomitant expansion of Ruminococcus gnavus in the gut microbiota. Gavage of wild-type mice with R. gnavus enhanced anti–PD-1–mediated tumor elimination, recapitulating the effect occurring in the absence of TREM2. A proinflammatory intestinal environment coincided with expansion, increased circulation, and migration of TNF-producing CD4^+ T cells to the tumor bed. Thus, TREM2 remotely controls anti–PD-1 immune checkpoint blockade through modulation of the intestinal immune environment and microbiota, with R. gnavus emerging as a potential probiotic agent for increasing responsiveness to anti-PD-1. INTRODUCTION Immune checkpoint inhibitors (CPIs) that block programmed cell death protein 1 (PD-1), programmed death-ligand 1 (PD-L1), and cytotoxic T lymphocyte–associated protein–4 (CTLA-4) are successfully used to activate antitumor T cell responses in the treatment of several types of cancers ([62]1–[63]3). However, although many patients respond durably to CPIs, a substantial number remain unresponsive or undergo recurrence, prompting the investigation of complementary therapeutic avenues to improve CPIs ([64]3). Tumor-associated macrophages (TAMs) can support tumor cell survival and proliferation and angiogenesis and suppress the immune response, which ultimately hampers CPI efficacy ([65]4, [66]5). Strategies targeting TAMs increase sensitivity to CPIs in mouse models and are currently being tested in clinical trials ([67]4). Genetic deletion or antibody-mediated blockade of triggering receptor on myeloid cells 2 (TREM2), a receptor broadly expressed on TAMs, attenuates TAM immunosuppressive function ([68]6–[69]9) and augments antitumor T cell responses induced by anti–PD-1 ([70]6). Another strategy to improve CPI therapy relies on targeting the intestinal microbiota ([71]10–[72]13). Many studies have reported that the composition of the microbiota correlates with CPI efficacy in preclinical models and patients with cancer ([73]10–[74]12). Fecal microbiota transplant (FMT) or oral gavage with specific bacteria can increase sensitivity to CPIs in mouse models. Human studies found associations between the gut microbiota and effectiveness of CPIs in patients with cancer ([75]10–[76]12, [77]14) and more recently provided evidence of a causal relationship ([78]15, [79]16). Specific members of the gut microbiota have been identified as probiotic strains that can boost antitumor T cell responses and alleviate TAM immunosuppression during CPI therapy ([80]11, [81]12, [82]14). Here, we found that the enhanced responsiveness of tumors to anti–PD-1 in TREM2-deficient mice is dependent on the microbiota. Administration of anti–PD-1 in Trem2^−/− mice bearing subcutaneous tumors induced an expansion of Ruminococcus gnavus that was associated with remodeling of intestinal macrophages toward a proinflammatory phenotype and accumulation of tumor necrosis factor (TNF)–producing CD4^+ T cells. These changes in the intestinal microenvironment corresponded with the migration of TNF-producing CD4^+ T cells from the gut and systemic circulation into tumor-draining lymph nodes (dLNs), as well as improved control of tumor growth. We conclude that anti–PD-1 together with TREM2 deficiency causes a distinct response in the intestinal microenvironment that affects CPI therapy with a beneficial effect on tumor rejection. RESULTS TREM2’s impact on tumor response to anti–PD-1 is influenced by the intestinal microbiota We sought to determine whether the intestinal microbiota affects the beneficial effect of TREM2 deficiency in tempering tumor growth. Eight-week-old conventionally raised Trem2^+/+ and Trem2^−/− mice co-housed from birth were inoculated subcutaneously with the methylcholanthrene (MCA)/1956 sarcoma cell line and immediately divided into two groups: In one group, Trem2^+/+ and Trem2^−/− mice were kept co-housed (Coh) throughout the experiment (Trem2^+/+-Coh and Trem2^−/−-Coh); in another group, Trem2^+/+ and Trem2^−/− mice were separated (Sep) after tumor injection (day 0) (Trem2^+/+-Sep and Trem2^−/−-Sep) ([83]Fig. 1A). Litters from different genotypes derived from co-housed pregnant mothers were cross-fostered to avoid a potential effect of the maternal microbiota. All animals consumed a standard polysaccharide-rich mouse chow ad libitum. In both groups, Trem2^−/− mice had a mild reduction of tumor burden compared with Trem2^+/+ mice, which was similar to that previously reported ([84]fig. S1, [85]A and [86]B; n = 5 or 6 animals per group) ([87]6). Because TREM2 deficiency also enhances the efficacy of anti–PD-1 treatment ([88]6) and given the contribution of the gut microbiota to CPI therapy ([89]6, [90]13), we examined the impact of housing conditions in these settings. Trem2^+/+ and Trem2^−/− mice were inoculated with MCA/1956, divided into Coh and Sep groups, and then treated with anti–PD-1, starting at day 9 ([91]Fig. 1A). Trem2^−/−-Sep had an enhanced response to anti–PD-1 compared with Trem2^+/+-Sep ([92]Fig. 1B and [93]fig. S1C). However, Trem2^−/−-Coh and Trem2^+/+-Coh mice responded similarly to anti–PD-1 ([94]Fig.1B and [95]fig. S1C). Fig. 1. Co-housing of TREM2-deficient and wild-type mice abolishes the improved tumor response to anti–PD-1 observed in the absence of TREM2. Fig. 1. [96]Open in a new tab (A) Experimental schemes. Mice were co-housed from birth and kept co-housed during the experiment (Coh) or separated at day 0 (Sep). Pregnant mothers of different genotypes were co-housed and litters were cross-fostered to avoid a potential effect driven by the maternal microbiota. (B) MCA/1956 tumor growth in the indicated conditions: wild-type (Trem2^+/+) or TREM2-deficient (Trem2^−/−) Sep Coh. All groups were treated with anti–PD-1 (n = 4 to 6). (C) Experimental schemes. Mice were generated from separated breeders and therefore separated at birth (Sep-b). (D) MCA/1956 tumor growth in the indicated conditions: Trem2^+/+ or Trem2^−/− Sep or Sep-b (n = 4 or 5). (E) T cell frequency in the tumor at day 15 after tumor injection in the indicated groups. (B and D) All groups were treated with anti–PD-1. Results are representative of at least two independent experiments and are presented as means ± SEM. Two-way ANOVA followed by post hoc comparisons (Tukey’s multiple comparisons test) of time points was used to compare tumor growth across the indicated conditions; statistical significance is indicated for the last time point of the curve. (E) Mann-Whitney U test was used to compare two groups (***P < 0.001; *P < 0.05). The results from manipulating housing conditions raised the possibility that the effect of TREM2 deficiency in enhancing anti–PD-1–mediated tumor control could be related to the gut microbiota. Because Trem2^−/−-Sep and Trem2^+/+-Sep mice were separated just at the time of tumor inoculation, we determined whether the enhancing effect of TREM2 deficiency on anti–PD-1 could be further amplified by having Trem2^−/− and Trem2^+/+ mice bred and hence born separately (Trem2^−/−–Sep-b; Trem2^+/+–Sep-b) rather than segregating them at the time of tumor inoculation. However, an equivalent increase in anti–PD-1 efficacy was evident in Trem2^−/− mice that were separated from Trem2^+/+ mice either from birth or at the time of tumor inoculation ([97]Fig. 1, [98]C and [99]D, and [100]fig. S1D). Additionally, separation of Trem2^−/− and Trem2^+/+ mice from birth did not amplify the difference in tumor growth between Trem2^−/− and Trem2^+/+ mice in the absence of anti–PD-1 treatment ([101]fig. S1E; n = 4 or 5 animals per group). Because deletion or blockade of TREM2 together with anti–PD-1 enhances the accumulation of T cells in the MCA/1956 tumor model ([102]6), we sought to determine whether T cell responses elicited by anti–PD-1 in Trem2^−/− and Trem2^+/+ mice were also affected by housing conditions. Trem2^−/−-Sep mice injected with anti–PD-1 displayed an increase in intratumoral T cell frequency at two time points in comparison with Trem2^+/+-Sep mice, whereas no differences were noted between Trem2^−/−-Coh and Trem2^+/+-Coh mice ([103]Fig. 1E and [104]fig. S1F). The proportion of intratumoral CD4^+/CD8^+ cells was similar in all conditions tested ([105]fig. S1F). Additionally, T cell frequency and CD4^+/CD8^+ proportion did not change significantly in tumor dLNs ([106]fig. S1F). Together, these data raised the possibility that anti–PD-1 induces changes in the microbiota of Trem2^−/− mice that promote T cell enrichment within the tumor and augment anti–PD-1 efficacy. To corroborate that anti–PD-1 induces de novo changes in the microbiota of Trem2^−/− mice that enhance anti–PD-1 response, we performed microbiota transplantation experiments. Cecal contents of Trem2^−/−-Sep and Trem2^+/+-Sep mice were collected at the time of tumor inoculation [microbiota-d0 (mbt-d0)] and during tumor growth 3 days after initiation of PD-1 treatment [microbiota-d12 (mbt-d12)]. Cecal contents were introduced by oral gavage into either Trem2^+/+ or Trem2^−/− mice that had been pretreated for 3 weeks with the VNAM (vancomycin, neomycin, ampicillin, and metronidazole) antibiotic cocktail. Seven days after the cecal microbial transplant (CMT), animals were subsequently inoculated with MCA/1956 and treated with anti–PD-1 ([107]Fig. 2A). Trem2^+/+ recipients of mbt-d12 from Trem2^−/− mice controlled tumor growth better than Trem2^+/+ mice that received mbt-d0 from Trem2^+/+ or Trem2^−/− mice ([108]Fig. 2B). Conversely, reconstitution of Trem2^−/− mice with mbt-d0 or mbt-d12 from Trem2^+/+ mice delayed or completely abrogated their capacity to control tumor growth in comparison with Trem2^−/− mice reconstituted with mbt-d0 from Trem2^−/− mice. Overall, these results indicate that anti–PD-1 treatment in TREM2-deficient mice induces changes in intestinal microbiota that support tumor rejection. Fig. 2. Gut microbiota affects tumor response to anti–PD-1 in the absence of TREM2. Fig. 2. [109]Open in a new tab (A) Schematic representation of CMT experiment. Trem2^+/+ and Trem2^−/− mice were treated with a cocktail of antibiotics (VNAM) for 3 weeks before colonization with cecal contents previously collected from either Trem2^+/+ or Trem2^−/− at steady state (mbt-d0) or 3 days after the anti–PD-1 injection (mbt-d12). (B) MCA/1956 tumor growth curve of Trem2^+/+ mice (left) colonized with cecal contents previously collected from either Trem2^+/+-mbt-d0, Trem2^−/−-mbt-d0, or Trem2^−/−-mbt-d12; tumor growth curve in Trem2^−/− mice (right) colonized with cecal contents collected from either Trem2^+/+-mbt-d0, Trem2^−/−-mbt-d0, or Trem2^+/+-mbt-d12. All mice were treated with anti–PD-1 (n = 5 to 7). (C) Schematic representation of antibiotic treatment experiment in Trem2^+/+ and Trem2^−/− during tumor growth. Trem2^+/+ and Trem2^−/− mice were treated with the indicated antibiotics for 3 weeks before tumor injection and throughout the experiment. (D) Tumor growth curve of Trem2^+/+ or Trem2^−/− treated with either vehicle, ampicillin, or vancomycin. All mice were treated with anti–PD-1 (n = 4 or 5). (B and D) Results are representative of at least two independent experiments and are presented as means ± SEM. (B) Two-way ANOVA followed by post hoc comparisons (Tukey’s multiple comparisons test) of time points was used to compare tumor growth across the indicated conditions; statistical significance is indicated for the last time point of the curve. (D) Two-way ANOVA followed by post hoc comparisons (Šidák’s multiple comparisons test) of time points was used to compare tumor growth across the indicated conditions; statistical significance is indicated for the last time point of the curve (***P < 0.0001; *P < 0.05; ns = not statistically significant). Enrichment of R. gnavus is associated with TREM2 deficiency and improved tumor response to anti–PD-1 To identify key features in microbiota induced by anti–PD-1 in the TREM2-deficient background, we treated Trem2^−/−-Sep and Trem2^+/+-Sep with various antibiotics alone or in combination before tumor inoculation, separation, and anti–PD-1 treatment ([110]Fig. 2C). Individually, ampicillin and vancomycin impaired the ability of Trem2^−/−-Sep mice to control tumor growth in response to anti–PD-1. Conversely, vehicle, metronidazole, or neomycin did not affect tumor control by Trem2^−/−-Sep mice ([111]Fig. 2D and [112]fig. S2). A mix of all four antibiotics (VNAM) completely abrogated the response to anti–PD-1 in both Trem2^−/−-Sep and Trem2^+/+-Sep mice ([113]fig. S2); in fact, tumor growth was more robust than in any other condition, in agreement with the described requirement of the microbiota for CPIs to succeed ([114]13, [115]17). To further characterize the changes induced by TREM2 deficiency and anti–PD-1, we characterized mbt-d0 and mbt-d12 from Trem2^+/+-Sep, Trem2^−/−-Sep, Trem2^+/+-Coh, and Trem2^−/−-Coh tumor-bearing mice by variable region 4 (V4)–16S rRNA sequencing ([116]data file S1). Principal coordinate analysis (PCoA) showed segregation of mbt-d12 collected from Trem2^−/−-Sep animals compared with those collected from Trem2^+/+-Sep, Trem2^−/−-Coh, and Trem2^+/+-Coh mice ([117]Fig. 3A). PCoA plots of mbt-d12 and mbt-d0 communities from Trem2^−/−-Coh and Trem2^+/+-Coh mice showed differences that were significantly influenced by the treatment with anti–PD-1 (permutational multivariate analysis of variance, PERMANOVA: P = 0.001) but not by the genotype (PERMANOVA: P = 0.238) ([118]fig. S3A). In contrast, differences between mbt-d12 and mbt-d0 among Trem2^−/−-Sep and Trem2^+/+-Sep or Trem2^−/−–Sep-b and Trem2^+/+–Sep-b were influenced by both the lack of TREM2 (PERMANOVA: P = 0.001) and the anti–PD-1 treatment (PERMANOVA: P = 0.001) ([119]fig. S3, [120]B and [121]C). These findings demonstrate that Trem2^−/− mice treated with anti–PD-1 develop a change in their gut microbiota that disappears when Trem2^−/− and Trem2^+/+ mice are co-housed together. Fig. 3. Increased relative abundance of R. gnavus is specifically associated with TREM2 deficiency upon anti–PD-1 treatment. Fig. 3. [122]Open in a new tab (A) PCoA plots showing beta-diversity analysis performed on the microbiota of Trem2^+/+-Sep, Trem2^−/−-Sep, Trem2^+/+-Coh, and Trem2^−/−-Coh mice. (B) Heatmap showing the indicator species analysis performed to compare the fecal microbiota composition of Trem2^+/+-Sep and Trem2^−/−-Sep. (C) Heatmap showing the indicator species analysis performed to compare the fecal microbiota composition of Trem2^−/−-Sep and Trem2^−/−-Coh. (D) Heatmap showing the indicator species analysis performed to compare the fecal microbiota composition of Trem2^+/+ and Trem2^−/− mice treated with metronidazole (−Metro). (B to D) Each column represents a different mouse, and each row denotes a different ASV significantly associated with either group. (A to D) All groups were treated with anti–PD-1 and analyzed 3 days after anti–PD-1 injection (mbt-d12). To identify genotype and treatment discriminatory taxa, we performed an indicator species analysis of mbt-d12 collected from Trem2^+/+-Sep, Trem2^−/−-Sep, Trem2^−/−-Coh, and Trem2^+/+-Coh mice. An initial comparison between Trem2^+/+-Sep and Trem2^−/−-Sep microbiota highlighted 39 amplicon sequence variants (ASVs) that were significantly associated with Trem2^+/+-Sep and 23 ASVs that were significantly associated with Trem2^−/−-Sep mice ([123]Fig. 3B; see [124]data file S2 for statistical test results). Fewer ASVs were significantly associated with either Trem2^+/+-Coh or Trem2^−/−-Coh fecal microbiota (1 and 5, respectively; [125]data file S3). To narrow down the number of ASVs in the Trem2^−/−-Sep microbiota that may enhance the anti–PD-1 response, we next compared ASVs in Trem2^−/−-Sep mice with those in Trem2^−/−-Coh animals where the enhanced response to anti–PD-1 was abolished. Eighteen ASVs were identified as selectively associated with Trem2^−/−-Sep; 10 of these were also significantly enriched in Trem2^−/−-Sep compared with Trem2^+/+-Sep mice ([126]Fig. 3, [127]B and [128]C, and [129]data file S4A) and represent taxa that belong to two distinct phyla (see [130]data file S4B for taxonomy). We further identified indicator species in mbt-d12 collected from Trem2^−/−-Sep mice treated with metronidazole, which does not diminish the enhanced response to anti–PD-1 (i.e., responses are comparable in metronidazole-treated and untreated Trem2^−/−-Sep mice; [131]Fig. 3D and [132]fig. S2A). Of the 10 ASVs enriched in Trem2^−/−-Sep versus Trem2^+/+-Sep and Trem2^−/−-Coh, only 1 ASV was significantly enriched in Trem2^−/−-Sep mice treated with metronidazole; this ASV was assigned to the anaerobic Gram-positive bacterium R. gnavus ([133]Fig. 3D and [134]data file S5). It should be noted that although Trem2^+/+-Sep microbiota contained R. gnavus, its mean percent relative abundance in Trem2^−/−-Sep mice was significantly higher (ASV.144: 0.68 ± 0.43%; ASV.102: 0.56 ± 0.27%) than the mean percentage detected in their Trem2^+/+-Sep counterparts (ASV.184: 0.27 ± 0.1%) after anti–PD-1 ([135]fig. S3D). Moreover, when we compared the relative abundance of ASV.102 and ASV.144 associated with Trem2^−/− mice between mbt-d0 and mbt-d12 ([136]fig. S4, [137]A to [138]D), both ASVs were significantly increased after anti–PD-1 treatment (mbt-d12), supporting the idea that the presence of R. gnavus is not sufficient per se to promote the control of tumor growth. Instead, the relative abundance of R. gnavus in the microbial community must be considered. Together, these results demonstrate that anti–PD-1 treatment in TREM2-deficient mice biases microbiota composition toward an enrichment of specific species, with R. gnavus being closely associated with better control of tumor growth. Given the growing interest in TREM2 blockade in oncological patients, we then asked whether anti-TREM2 treatment with a monoclonal antibody (mAb) in combination with anti–PD-1 in wild-type mice drove a similar effect to TREM2 deficiency in terms of modulation of the microbial communities. Anti-TREM2 treatment was associated with changes in the microbiota communities at several time points (days 0, 9, 12, and 18 after tumor injection) compared with mice treated with a control mAb ([139]data file S6). However, we did not observe R. gnavus enrichment in any of the analyzed time points, suggesting that the contribution of the microbiome to the combined effect of anti-TREM2 and anti–PD-1 acute blockade is not dependent on R. gnavus but might potentially rely on different species that act similarly to R. gnavus. TREM2 deficiency in combination with anti–PD-1 blockade shifts the gut environment toward a proinflammatory state Because an intestinal bloom of R. gnavus has been observed during active phases of inflammatory bowel disease ([140]18), we investigated whether anti–PD-1 treatment in Trem2^−/− mice may be associated with colon inflammation. Initially, we assessed weight loss and fecal lipocalin levels and found no significant differences between Trem2^+/+ and Trem2^−/− mice whether or not they were treated with anti–PD-1 ([141]fig. S5A). This suggested no clinical evidence of overt colitis in our model. To investigate potential subclinical inflammation, we conducted bulk RNA sequencing of colonic tissue and real-time polymerase chain reaction (PCR) analysis of selected cytokines. Principal components analysis (PCA) plots of sequenced samples showed that TREM2-deficient samples treated with anti–PD-1 clustered separately from the wild-type counterpart ([142]fig. S5B), exhibiting an enrichment in pathways associated with cell activation, immune cell adhesion, T cell activation, and antigen presentation ([143]fig. S5, [144]C and [145]D). Cytokine transcripts were below the detection limit, with the exception of Tnf, which was increased in the Trem2^−/− group treated with anti–PD-1 ([146]fig. S5E). Analysis of cytokine expression by real-time PCR showed a trend toward increased TNF, an increase in Il17a, and a decrease in Il10 ([147]fig. S5F), corroborating the presence of a more inflammatory environment in Trem2^−/− mice treated with anti–PD-1. Together, these data suggest that anti–PD-1 treatment in Trem2^−/− mice is associated with intestinal subclinical inflammation, in line with a previous report ([148]19). Considering the expression of TREM2 in macrophages, particularly in colon macrophages ([149]fig. S6A), we investigated whether colon Trem2^−/− macrophages contributed to subclinical inflammation. Flow cytometry analysis of colonic myeloid cells isolated at day 12 from Trem2^−/−-Sep and Trem2^+/+-Sep mice inoculated with MCA/1956 tumors and treated or not treated with anti–PD-1 revealed no statistically significant differences in the frequency of total myeloid cells, macrophages, monocytes, and neutrophils across all conditions ([150]fig. S6, [151]B and [152]C). To improve the sensitivity of myeloid cell analysis under the same conditions, we performed single-cell RNA sequencing (scRNA-seq) of CD11b^+ cells sorted from the colon lamina propria (live/CD45^+Ly6G^−SiglecF^−/CD11b^+). After quality controls, 16,807 cells were partitioned in 10 clusters ([153]Fig. 4A). The major cell types were manually identified based on cluster markers and previous studies ([154]6, [155]20). These included (i) macrophages with features of activation and inflammation (cluster 0—Ccl3, Ccl4, Rgs1, and Atf3; cluster 2—H2-Ab, H2-M2, Il1b, Nlrp3, and Cd80); (ii) macrophages with markers of immunosuppression (cluster 1—Mrc1, Stab1, Cd163, Cd36, Fcrls, Pf4, and F13a1); (iii) less differentiated macrophages expressing Ccr2 (cluster 6—Ccr2, Mrc1, and Fcrls); (iv) monocytes (cluster 4—Ly6i and Ccr2); (v) macrophages with a type I interferon (IFN) signature (cluster 9—Ifit3 and Ifit2); (vi) CD11b^+ conventional dendritic cells 2 (cDC2) (cluster 3—Zbtb46 and Sirpa) and cDC1 (cluster 8—Zbtb46 and Xcr1); (vii) proliferating monocytes—cDCs (cluster 7—Mki67); and (viii) contaminating CD11b-expressing B cells (cluster 5—Jchain) ([156]Fig. 4B). TREM2 was expressed in all monocyte-macrophage subsets (clusters 0, 1, 2, 4, 6, and 9 and part of cluster 7) ([157]Fig. 4C). Analysis of cluster abundance across the four conditions (Trem2^+/+, Trem2^−/−, Trem2^+/+ anti–PD-1, and Trem2^−/− anti–PD-1) revealed that anti–PD-1 treatment of Trem2^−/− mice was associated with a limited enrichment of macrophages with immunostimulatory and inflammatory features (clusters 0 and 2), whereas immunosuppressive macrophages (cluster 1) were slightly reduced ([158]Fig. 4A). Moreover, both TNF and a type I IFN signature were induced in cluster 0 ([159]fig. S6D). Conversely, anti–PD-1 treatment in Trem2^+/+ mice was not associated with obvious changes in macrophage populations ([160]Fig. 4A). Negligible differences were also detected in untreated Trem2^−/−-Sep and Trem2^+/+-Sep mice ([161]Fig. 4A). Together, these results suggest that anti–PD-1 treatment in Trem2^−/− mice biases intestinal macrophages toward a proinflammatorym phenotype that may affect the colonic environment together with the expansion of R. gnavus. Fig. 4. TREM2 deficiency is associated with a more inflamed colonic environment after anti–PD-1 treatment. Fig. 4. [162]Open in a new tab (A) UMAP plot showing scRNA-seq of colonic myeloid cell clusters from merged conditions: Trem2^+/+ or Trem2^−/− treated or not with anti–PD-1 (left). Cells were sorted as live/CD45^+Ly6G^−SiglecF^−/CD11b^+. Cluster proportions in each condition: Trem2^+/+ and Trem2^−/− treated or not with anti–PD-1 (bottom right panel) (n = 2 mice per group). (B) UMAP plots of selected genes from merged conditions. (C) UMAP plots showing Trem2 expression in Trem2^+/+ and Trem2^−/− mice from merged conditions. (D) T cell frequency in the colon lamina propria in Trem2^+/+ and Trem2^−/− mice, separated at day 0 (Sep) or co-housed (Coh). All groups were treated with anti–PD-1. (E) TNF protein content in Trem2^+/+ and Trem2^−/− CD4^+ T cells from mice treated or not with anti–PD-1. (A to E) Samples were analyzed 3 days after anti–PD-1 treatment, 12 days after tumor injection (day 12). (A to C) Each group is representative of two mice that were sequenced separately. (D and E) Results are presented as means ± SEM. Mann-Whitney U test was used to compare two groups (*P < 0.05). Next, we examined whether T cells isolated from the colon of tumor-bearing Trem2^−/− mice treated with anti–PD-1 also exhibited a proinflammatory profile. Flow cytometric analysis documented increased frequency of total CD3^+ T cells in the colon lamina propria of Trem2^−/−-Sep mice compared with Trem2^+/+-Sep controls at day 12 after tumor injection ([163]Fig. 4D); this difference was abrogated in co-housed mice. The frequency of colonic CD4^+ and CD8^+ T cells, as well as PD-1^+ T cells, was similar among all experimental groups ([164]fig. S6E); the proportions of T cells in mesenteric lymph nodes (mLNs) were also comparable ([165]fig. S6F). In anti–PD-1–treated tumor-bearing Trem2^−/− mice, higher frequencies of colonic TNF-producing CD4^+ T cells were observed compared with anti–PD-1–treated Trem2^+/+ mice and all untreated mice ([166]Fig. 4E). We conclude that microbiota alteration in anti–PD-1–treated Trem2^−/− mice is associated with an expansion of TNF-producing CD4^+ T cells in the colon. Microbiota derived from TREM2-deficient mice promotes gut T cell migration to the tumor bed We determined how remodeling of the gut microbiota and immune landscapes may affect the efficacy of anti–PD-1 in controlling tumor growth at a distant site. We tested the hypothesis that imprinting of T cells by the subclinical inflammation induced by anti–PD-1 treatment in Trem2^−/− mice can facilitate their migration to the tumor bed. To track gut-imprinted T cells in the MCA/1956 tumor model, we used mice globally expressing Kikume Green-Red (KikGR), a green-to-red photoconvertible fluorescent protein ([167]21, [168]22). Application of an infrared laser light to surgically exposed colon or mLNs converted cells from KikG^+ to KikR^+; photoconversion was complete in mLNs (>99%) but partial in the colon (55% to 80%). KikR^+ cells were detected in both tumors and dLNs starting from 48 hours up to 4 days after photoconversion and were 31.32 ± 11.30 more enriched in dLNs than tumors ([169]Fig. 5 and [170]fig. S7A). Because the frequency of KikR^+ cells in dLNs and tumors was higher after photoconversion in mLNs than in colons, we only applied the laser light to mLNs in all subsequent experiments. Fig. 5. Microbiota from TREM2-deficient mice promotes T cell activation in the tumor bed. Fig. 5. [171]Open in a new tab (A) Experimental scheme of CMT in KikGR mice. Mice were treated with a cocktail of antibiotics (VNAM) for 3 weeks before colonization with the cecal content previously collected from either Trem2^+/+ or Trem2^−/− mice at 3 days after the anti–PD-1 injection (mbt-d12). Mice were treated with anti–PD-1 as indicated and underwent surgery and in situ photoconversion (PC) of mLNs 10 days after tumor injection. Mice were analyzed on day 14 after tumor injection. (B) TNF production in PC and not PC CD4^+ T cells, upon ex vivo restimulation with PMA-ionomycin. Cells were sorted from dLNs. Each dot represents the pool of two mice. (C) Flow cytometry analysis of PC T cells in the indicated organs [dLN, tumor, spleen, and control lymph node (cLN)] in the two groups: KikGR mice transplanted with Trem2^+/+ mbt-d12 or Trem2^−/− mbt-d12. (D) Representative flow plots of PC T cells (KikR^+) in dLNs, in the indicated groups. For the control mice (CTRL), mLNs were not photoconverted. (E) Flow cytometry analysis of PC and not PC T cells in the dLNs in the indicated groups. (B, C, and E) Results are representative of at least two independent experiments and presented as means ± SEM. Mann-Whitney U test was used to compare two groups (**P < 0.01; *P < 0.05). Next, two groups of KikGR mice were reconstituted with mbt-d12 obtained from either Trem2^+/+-Sep or Trem2^−/−-Sep mice inoculated with MCA/1956 and treated with anti–PD-1 ([172]Fig. 5A). After CMT, KikGR mice were inoculated with MCA/1956 at day 0, treated with anti–PD-1 at day 9, exposed to infrared laser light in the mLNs at day 10, treated again with anti–PD-1 at day 12, and euthanized for collection of tumors and tumor dLNs at day 14 ([173]Fig. 5A). KikR^+ and KikG^+ T cells were sorted from tumor dLNs and tested for intracellular cytokine content after in vitro stimulation with phorbol 12-myristate 13-acetate (PMA)–ionomycin. KikR^+ CD4^+ T cells derived from mice reconstituted with mbt-d12 of Trem2^−/−-Sep mice produced significantly more TNF than KikG^+ cells from the same mice and KikR^+ CD4^+ T cells from mice that had received mbt-d12 of Trem2^+/+-Sep mice ([174]Fig. 5B). No statistically significant differences were observed in KikR^+ CD8^+ T cells ([175]fig. S7B). Thus, gut CD4^+ T cells imprinted by the mild inflammatory environment in anti–PD-1–treated TREM2-deficient mice are predisposed to produce TNF upon reaching the tumor dLNs. Additionally, the frequency of KikR^+ T cells in tumor dLNs was significantly higher in KikGR mice reconstituted with mbt-d12 from Trem2^−/−-Sep mice than in counterparts receiving mbt-d12 from Trem2^+/+-Sep mice, although the frequencies of KikR^+ T cells in tumors, contralateral LNs, and spleens were similar ([176]Fig. 5, [177]C and [178]D). Notably, the frequency of KikG^+ T cells in tumor dLNs was also higher in KikGR mice transplanted with mbt-d12 from Trem2^−/−-Sep mice ([179]Fig. 5E), suggesting that mbt-d12 from Trem2^−/−-Sep mice promoted not only T cell migration from the gut but also either in situ expansion or recruitment from other sites. Hence, apart from its local effect, mbt-d12 may also exert a systemic influence, further affecting T cell migration into the tumor bed. Oral gavage of R. gnavus enhances tumor response to anti–PD-1 From a translational perspective, we aimed to directly investigate the specific impact of R. gnavus on the efficacy of anti–PD-1, irrespective of TREM2 deficiency. To determine whether the enrichment of R. gnavus observed in Trem2^−/−-Sep mice played a causal role in improving tumor growth control, we administered R. gnavus [1.5 × 10^6 colony-forming units (CFU) per mouse] to wild-type mice by daily oral gavage for 1 week starting at day 10 after subcutaneous tumor inoculation, in addition to anti–PD-1, and monitored tumor growth ([180]Fig. 6A). Wild-type mice that received R. gnavus controlled tumor growth substantially better than their counterparts ([181]Fig. 6B). In control experiments, Clostridium scindens was chosen as the closest taxon assigned to R. gnavus V4–16S sequence, but its administration did not affect tumor growth ([182]fig. S8A). Moreover, heat inactivation (100°C) of R. gnavus impaired the capacity of the bacterium to provide protection, suggesting that the exposure to the live microorganism or certain heat-sensitive components are necessary to drive tumor control ([183]fig. S8B). Fig. 6. R. gnavus exposure promotes anti–PD-1 response and drives a remodeling of the tumor immune environment. Fig. 6. [184]Open in a new tab (A) Experimental scheme. Wild-type mice were gavaged with R. gnavus (ATCC 29149) or vehicle at the indicated time points. All mice were treated with anti–PD-1. (B) MCA/1956 tumor growth in the indicated conditions: wild-type mice treated with vehicle (CTRL) or R. gnavus. All mice were treated with anti–PD-1 (n = 5). (C) UMAP plot showing scRNA-seq of CD45^+ cells from tumors from merged conditions obtained. Mice were euthanized on day 14 (n = 4 mice per group). (D) UMAP plots of selected genes (cluster markers) in CD45^+ cell clusters in merged conditions. (E) UMAP plots of CD45^+ cells in each condition: mice treated with vehicle (CTRL) or R. gnavus. (F) Cluster proportions of CD45^+ cells in the indicated conditions (bar plot and pie charts). (G) Bar plot showing the proportion of each group per cluster among the CD45^+ cell clusters. (H) Heatmap showing the selected genes in the indicated T cell clusters. (A to C) Groups are representative of four mice that were sequenced together and deconvoluted on the basis of a unique hashtag antibody. (B) Results are representative of at least two independent experiments and presented as means ± SEM. Two-way ANOVA followed by post hoc comparisons (Šidák’s multiple comparisons test) of time points was used to compare tumor growth across the indicated conditions; statistical significance is indicated for the last time point of the curve (***P < 0.001). To assess the influence of R. gnavus on the tumor’s immune environment, we conducted scRNA-seq on sorted CD45^+ cells isolated from MCA/1956 tumors in mice that received anti–PD-1 treatment and were either gavaged with R. gnavus or received a vehicle. We obtained a total of 25,095 cells from the two groups after quality controls, which were then projected onto a uniform manifold approximation and projection (UMAP) encompassing five major populations ([185]Fig. 6C). Macrophages, monocytes, and dendritic cells expressing Fcgr1, Ly6c2, and Zbtb46, respectively, encompassed clusters 0, 1, 2, 5, 9, 11, 14, and 18. T cells comprised TCRαβ+ CD4^+ (clusters 6 and 8—Trac and Cd4) and CD8^+ (clusters 4 and 12 and a portion of cluster 6—Trac and Cd8) T cells, as well as a small proportion of TCRγδ+ T cells (cluster 15—Trdc). NK cells corresponded to clusters 3 and 7, with cluster 3 cells showing chemokine production (Ccl5) and cluster 7 exhibiting cytotoxicity (Gzmc). Neutrophils corresponded to cluster 10 (S100a9), whereas B cells were found in cluster 13 (Cd19) ([186]Fig. 6, [187]C and [188]D). After exposure to R. gnavus, we observed substantial remodeling of the macrophage landscape and an increased frequency in all TCRαβ+ CD4^+ and CD8^+ T cell clusters, including activated CD8^+ T cells (cluster 4), resting CD4^+ and CD8^+ T cells (cluster 6), and cycling CD8^+ T cells (cluster 12). Regulatory T cells (cluster 8) remained unchanged, whereas TCRγδ+ T cells (cluster 15) showed a slight decrease ([189]Fig. 6, [190]E to [191]H). To further examine the complexity of macrophage changes, we reclustered myeloid cell populations, resulting in 14 clusters ([192]Fig. 7A and [193]fig. S9, [194]A and [195]B). Notably, in R. gnavus–treated mice, clusters 0 (Il1r2 Macs) and 3 (Trem2^high Macs) decreased in size, whereas clusters 2 (Cd300e Macs), 4 (Fabp5 Macs), and 5 (Tmem176 Macs) expanded ([196]Fig. 7, [197]B and [198]C, and [199]fig. S9C). Cluster 0 showed heterogeneous features: It was enriched for IL-1 family molecules, such as Il1r2 and Il1b, with the latter being associated with pathogenic inflammation in cancer ([200]23). At the same time, cluster 0 displayed immunostimulatory properties, as indicated by high Cxcl9 expression ([201]Fig. 7, [202]B and [203]C, and [204]fig. S9A). The impact of R. gnavus treatment on the macrophage compartment was particularly evident in changes to the relative abundance of clusters 3 and 4. Both of these clusters expressed markers associated with immunosuppression, such as Trem2, Cx3cr1, Arg1, Stab1, and Mrc1. However, cluster 3 (Trem2^high Macs), which had higher expression of immunosuppressive molecules like Trem2, Cx3cr1, Spp1, Fn1, and Ldha, was decreased in response to R. gnavus. In contrast, cluster 4 (Fabp5 Macs), with higher expression of inflammatory molecules such as Ccl5, increased after R. gnavus treatment ([205]fig. S9D). Furthermore, R. gnavus treatment led to reduced expression of the TGFβ receptor Tgfbr1 (P = 0.0286), whereas the expression of the IFN-γ receptor Ifngr1 trended toward an increase at least in one macrophage population (cluster 5, P = 0.0571) ([206]Fig. 7D). We then generated a manually curated module based on immunosuppressive genes from clusters 3 and 4, which included Trem2, Cx3cr1, Arg1, Stab1, Mrc1, Fn1, and Ldha. We projected this module onto the UMAPs of both groups. This signature was enriched in the control group (P = 0.000186) when compared with R. gnavus–treated mice ([207]Fig. 7E). Fig. 7. R. gnavus exposure promotes anti–PD-1 response and drives remodeling of the macrophage compartment in the tumor. Fig. 7. [208]Open in a new tab (A) UMAP plot showing scRNA-seq of myeloid cell clusters after reclustering in merged conditions. (B) UMAP plots of myeloid cell clusters in each condition: mice treated with vehicle (CTRL) or R. gnavus. (C) Cluster proportions of myeloid cell clusters in the indicated conditions (bar plot and pie charts). (D) UMAP plots of selected genes (Tgfbr1 and Ifngr1) in myeloid cell clusters in the indicated conditions. (E) UMAP plots of manually curated module of the indicated genes associated with immunosuppression in the indicated conditions. (F) Frequency of macrophages in the tumors in the indicated groups. (G) Frequency of T cells and migratory dendritic cells (mig cDCs) in mLNs in the indicated groups. (H) Frequency of T cells, T cell subsets, and macrophages in colons in the indicated groups. (F to H) Results are presented as means ± SEM. Mann-Whitney U test was used to compare two groups (**P < 0.01; *P < 0.05). (E) One-way ANOVA test was used (P = 1.2 × 10^−41). T[regs], regulatory T cells. We further examined the immune cell infiltration in tumors, mLNs, and colons in controls and R. gnavus–treated mice by flow cytometry. We observed that R. gnavus treatment induced the enrichment of MHCII^+ macrophages, confirming the enrichment of less suppressive myeloid cells in the tumors ([209]Fig. 7F and [210]fig. S10A). We also observed enrichment of migratory conventional dendritic cells (cDCs) and total T cells in mLNs ([211]Fig. 7G and [212]fig. S10B), as well as macrophages and activated T cells in the colon after R. gnavus exposure ([213]Fig. 7H), corroborating that R. gnavus also drives immunechanges in the gut. DISCUSSION We previously found that deficiency of TREM2 amplifies the antitumor efficacy of anti–PD-1 treatment ([214]6). Here, our study demonstrates that this amplification depends on a complex interaction among T cells, TREM2-deficient intestinal macrophages, and intestinal microbiota that creates a subclinical inflammatory environment in the intestine. We believe that it is unlikely that TREM2 deficiency alone is sufficient to induce a proinflammatory polarization of macrophages and the R. gnavus expansion, because macrophage signatures and R. gnavus abundance were similar in Trem2^−/− and Trem2^+/+ mice before anti–PD-1 treatment. We envision that anti–PD-1 activates CD4^+ T cells in the gut, which engages macrophages through cell surface molecules, such as CD40L, and/or proinflammatory cytokines, such as IFN-γ ([215]24). Previous work showed that TREM2 attenuates Toll-like receptor signaling, specifically mitogen-activated protein kinase (MAPK) activation ([216]25, [217]26). Thus, TREM2 deficiency may lead to unhindered macrophage responses to both T cell– and microbial-induced stimuli. Such response might generate a subclinical inflammatory and/or hypoxic state that facilitates the expansion of the anaerobe R. gnavus, among other species. This phenomenon mirrors observations in Crohn’s disease, where R. gnavus colonizing the intestinal mucosal surface blooms during flare-ups of symptoms ([218]18). The specific effect of TREM2 deletion and PD-1 blockade on the gut permeability and integrity, which might enhance the subclinical inflammation state, as well as the mechanism by which R. gnavus promotes antitumor responses at a distant site, will require further analysis. R. gnavus produces glucorhamnan, a polysaccharide that stimulates the production of TNF by cDCs through TLR4 in vitro ([219]18). Glucorhamnan or other bacterial components that elicit proinflammatory responses may induce TNF-producing T cells and inflammatory macrophages. Moreover, R. gnavus–derived metabolites might act locally in the gut environment to promote T cell activation and migration and systemically to directly facilitate the antitumor immune response at distant sites. Previous analyses of the fecal microbiota delineated the relative abundance of certain bacteria, including R. gnavus, in the intestines of mice and patients with cancer responding to PD–1 blockade and showed a causal relationship ([220]10, [221]11, [222]27). Our results extend these studies, suggesting that the enrichment of microbiota species associated with heightened response to anti–PD-1 therapy is also influenced by intestinal macrophages. Notably, anti-TREM2 treatment, although enhancing anti–PD-1 control of tumor growth similar to TREM2 deficiency, was not associated with an enrichment of R. gnavus. Yet, anti-TREM2 treatment did induce significant changes in the microbiota communities compared with mice treated with a control mAb. These data suggest that the contribution of the microbiome to the efficacy of anti–PD-1 in TREM2 deficiency or blockade does not always depend on R. gnavus but might potentially rely on the expansion of different species that enact mechanisms similar to those of R. gnavus. Because the abundance of R. gnavus in the intestine is relatively low, the impact of these species may depend on their relative abundance before antibody treatment. It will be important to validate this concept in the human population by searching for correlations among responsiveness to anti–PD-1, enrichment of certain microbiota species, and macrophage responsiveness to environmental stimuli. In light of the ongoing clinical trial ([223]NCT04691375), it will be informative to evaluate the effect of TREM2 and PD-1 blockade on gut microbial composition and any relationship to the response to treatment. MATERIALS AND METHODS Study design The goal of this study was to evaluate the link between the gut microbiota and the antitumor effect of TREM2 deletion after anti–PD-1 treatment. We used the syngeneic sarcoma tumor cell line (MCA/1956) that we subcutaneously implanted into Trem2^+/+ and Trem2^−/− B6 mice. We used different housing conditions to evaluate the effect of the microbiota on anti–PD-1 response. To further corroborate the impact of Trem2^−/−-associated species on the response to anti–PD-1, we performed microbiota transplantation experiments and treated mice with different combinations of antibiotics. We identified the bacterial species associated with TREM2 deficiency through V4–16S rRNA sequencing analysis; among those, R. gnavus was enriched in anti–PD-1 responder groups. To track the migration of gut-imprinted T cells into the tumor bed, we performed CMT in KikGR mice. We photoconverted the mLNs and analyzed dLNs, tumors, and spleens. To examine the downstream effect of R. gnavus on the antitumor immune response, we orally gavaged wild-type (WT) mice with the American Type Culture Collection (ATCC) R. gnavus strain. For all experiments, we monitored tumor growth and analyzed the percentage and functions of immune cells in tumors, dLNs, colons, and mLNs by flow cytometry and scRNA-seq. Mice were used for tumor growth experiments and harvesting of mouse tissue, with approved humane end points to terminate in vivo experiments. Sample size was determined by preliminary experiments. For 16S and R. gnavus gavage experiments, groups were randomized before treatment. Animals that failed to develop tumors were excluded from the analysis. The number of mice and statistical methods used are described in the figure legends. Animals All mice used in this study were bred under specific pathogen–free conditions at Washington University School of Medicine animal facility. Mice within experiments were age- and sex-matched, and both sexes were used. Pregnant mothers of different genotypes were co-housed, and litters were cross-fostered, to avoid a potential effect driven by the maternal microbiota. All mice were bred in the same room (SRF-E-228) and were transferred and kept in the same room for experiments (CSRB-742D). All animals consumed a standard polysaccharide-rich mouse chow ad libitum (PicoLab Rodent Diet 20). All mice were housed in standard LD 12:12 conditions (light on from 6:00 a.m. to 6:00 p.m.). Housing conditions before and after weaning and during the experiments are specified in the [224]Results section of the text. Mice from different genotypes were either co-housed from birth and kept co-housed or separated during the experiment (the day of tumor injection). Mice used for the experiment with anti-TREM2 were purchased from the Jackson Laboratory (C57BL/6J). Mice used in this study include WT C57BL/6J, Trem2^−/−, and KikGR [Tg(CAG-KikGR)33Hadj/J—Jax 013753]. KikGR mice were transferred to us by G. Randolph (Washington University School of Medicine). Mice did not undergo any procedures before their stated use. Mice were humanely euthanized at different time points, as indicated. All animals were backcrossed until at least >98% C57BL/6J confirmed by genotype wide microsatellite typing. For tumor models, animals were injected at 8 to 10 weeks of age. All studies performed on mice were done in accordance with the Institutional Animal Care and Use Committee at Washington University in St. Louis, which approved all protocols used in this study (protocol nos. 19–0981, 22–0433, and 22–0274). Tumor models The MCA-induced sarcoma cell line (MCA/1956) was generated in female C57BL/6 WT mice in R.D.S’s laboratory. Cell batches were frozen as low-passage tumor cells and injected after a maximum of three passages. Mice were shaved on the flank, and MCA/1956 cells were injected subcutaneously [10^6 cells per mouse in 100 μl of phosphate-buffered saline (PBS)]. Mice were monitored every day, and tumors were measured by caliper every other day. Mice were euthanized on the indicated days or when tumors reached 1.5 cm in diameter. Flow cytometry Single-cell suspensions were prepared from the collected tissues upon sacrifice. Tumors were minced and digested with collagenase IV (Sigma) for 30 min at 37°C. Cells were filtered through 70-μm strainers, washed with PBS, and stained for flow cytometry. Lymph nodes were collected and mashed on a 70-μm strainer. Small intestine or colon were collected, flushed with HBSS to remove luminal contents, and opened lengthwise. Peyer’s patches were removed. The tissue was gently agitated for 20 min in HBSS with Hepes, BCS, and EDTA. The tissue was rinsed and agitated for a second round of 20 min. After rinsing, the tissue was digested with collagenase IV (Sigma) in complete RPMI (Gibco) for 40 min at 37°C under agitation, as previously described ([225]28). Digested tissue was filtered through a 100-μm strainer, collected into a 50-ml tube, and washed. The following antibodies were used: CD45-AlexaFluor700 or -BUV563 (clone 30-F11); CD11b-PerCPCy5.5, -PECy7, -APC, or -BV421 (clone M1/70); I-A/I-E-BV650 or -BUV805 (clone M5/114.15.2); Ly6C-BV421, -APC, -PerCPCy5.5, or -PE (clone HK1.4); Ly6G-AlexaFluor700 or -APC (clone 1A8); CD64-APC or -BV605 (clone X54–5/7.1); TCRβ-PE (clone H57–597); CD8-BV785 (clone 53–6.7); CD4-PECF594 (clone RM4–5); PD-1-FITC or -APC (clone J43); IFN-γ–APC (clone XMG1.2); TNFa-eFluor 450 (clone MP6-XT22); XCR1-BV605 (ZET); LAG3-PE-Cy7 (C9B7W); TCRgd-BV421 (GL3); CD11c-APC/Cy7 (N418); and FoxP3-AF647 (clone 150D). For cell hashing, the following hashtag antibodies were used: Total-Seq-B0301, B0302, B303, and B304 (clone M1/42; 30-F11). Cells were incubated with Fc block before staining. Cell viability was determined by Aqua LIVE/Dead-405 nm staining (Invitrogen); negative cells were considered viable. Foxp3/Transcription Factor Staining Buffer Set (eBioscience) was used for intracellular staining. Cells were analyzed on BD FACSymphony (BD Bioscience). An AriaII (BD Bioscience) was used for sorting. Data were analyzed with FlowJo software (Treestar). In vivo treatments Mice were treated intraperitoneally with anti–PD-1 antibody (200 μg per mouse, clone RMP1–14, Leinco P372) every 3 days, starting at day 9 after tumor injection, as specified above. Anti-TREM2 (178) and control (135.5) antibody (200 μg per mouse) were produced in our laboratory as previously described ([226]6) and were given intraperitoneally every 5 days, starting at day 2 after tumor injection. T cell stimulation Single-cell suspensions were prepared from tumors upon sacrifice and leukocytes were enriched with a PercollTM (GE Healthcare) gradient. Obtained cells were cultured in complete RPMI with or without PMA (10^–7 M, Sigma)–ionomycin (500 ng/ml, Sigma) and Protein Transport Inhibitor Cocktail (eBioscience, 500×) for 4 hours. Cells were collected and stained for flow cytometry analysis. Real-time PCR Total RNA was extracted from snap-frozen tissue using the RNeasy Mini Kit (Qiagen Inc.), following the supplier’s instructions. RNA was converted to cDNA using the iScript cDNA Synthesis Kit (Bio-Rad), and PCR was performed using iTaq Universal SYBR Green Supermix (Bio-Rad). Gene expression analysis was normalized using the 2^−ΔΔCt method with Gapdh as a reference gene. Primers are listed in [227]table S1. RNA-seq and analysis Colonic RNA was extracted using an RNeasy mini kit according to the manufacturer’s protocol (QIAGEN). Total RNA integrity was determined using Agilent Bioanalyzer or 4200 Tapestation. Library preparation was performed with 5 to 10 μg of total RNA with a Bioanalyzer RIN score greater than 8.0. Ribosomal RNA was removed by poly-A selection using Oligo-dT beads (mRNA Direct kit, Life Technologies). mRNA was then fragmented in reverse transcriptase buffer and heated to 94°C for 8 min. mRNA was reverse-transcribed to yield cDNA using SuperScript III RT enzyme (Life Technologies, per the manufacturer’s instructions) and random hexamers. A second strand reaction was performed to yield ds-cDNA. cDNA was blunt ended, had an A base added to the 3′ ends, and then had Illumina sequencing adapters ligated to the ends. Ligated fragments were then amplified for 12 to 15 cycles using primers incorporating unique dual index tags. Fragments were sequenced on an Illumina NovaSeq-6000 using paired-end reads extending 150 bases. RNA-seq reads were then aligned and quantitated to the Ensembl release 101 primary assembly with an Illumina DRAGEN Bio-IT on-premise server running version 4.0.3–8 software. Read counts were used as input and normalized using built-in algorithms in DESeq2 (1.38.3). Pairwise comparisons among the four conditions were done on all genes. The resulting P values were adjusted using Benjamini and Hochberg’s approach for controlling the false discovery rate. Genes with an adjusted P < 0.05 were assigned as differentially expressed. For pathway enrichment analysis, gene set enrichment analysis (GSEA) was performed using the hallmark pathways from the Molecular Signature Database. Single-cell RNA sequencing Sorted cells from the colon lamina propria were washed in PBS + 0.04% BSA at a final concentration of ~1000 cells/μl ([228]Fig. 4). Cells were extracted from tumor ([229]Fig. 6), and before sorting, each sample was labeled with a unique hashtag antibody (TotalSeq B). Sorted cells were resuspended in PBS with 0.04% BSA at a final concentration of approximately 1200 cells/μl, and samples were pooled and loaded on a 10x Genomics Chromium Single-Cell Controller at the McDonnell Genome Institute in Washington University in St. Louis. The 10x Genomics Next GEM Single-Cell 3′ Reagents Kit was used for library generation on the 10x Genomics platform. This process used the Chromium Next GEM Single Cell 3′ Kit v3.1 (16 reactions), Chromium Next GEM Chip G Single Cell Kit (48 reactions), and Dual Index Kit TT Set A (96 reactions). In brief, up to 23,000 cells were partitioned into nanoliter gel-bead-in-emulsion (GEM) droplets, each containing retro-transcriptase reaction mix. Single-cell cDNA received a unique 12-nucleotide barcode and a unique molecular identifier (UMI). Subsequently, GEM cDNA underwent 11 cycles of amplification before purification using SPRIselect beads. The generation of gene-expression (GEX) and hashtag-oligo (HTO) libraries followed the recommendations outlined in the 10x Genomics Chromium Single-Cell 3′ Reagent Kits User Guide (v3.1 Chemistry Dual Index), with modifications to PCR cycles based on the calculated cDNA concentration. The concentration of each library was determined using qPCR with the KAPA library Quantification Kit following the manufacturer’s protocol (KAPA Biosys tems/Roche). The sequencing of normalized cDNA libraries was performed on a NovaSeq6000 S4 Flow Cell using the XP workflow and a 28 × 10 × 10 × 150 sequencing recipe as per the manufacturer’s protocol (Illumina). A median sequencing depth of 50,000 and 5000 reads per cell was targeted for the GEX and HTO library, respectively. For gene expression comparisons ([230]Fig. 7, [231]D and [232]E), we performed a pseudobulk analysis and extracted log-transformed counts. For [233]Fig. 7E, we then performed DESeq2 analysis and pathway enrichment analysis for selected genes. scRNA-seq analysis The Cell Ranger Software Suite (v6.0.0) by 10x Genomics was used for sample demultiplexing, barcode processing, and single-cell counting. Subsequent analysis was conducted using the Seurat package (v4.1.1) in R. The HTODemux function from the Seurat package was used for demultiplexing. Quality control entailed the removal of cells with mitochondrial content exceeding 5%, as well as those containing fewer than 1000 or more than 7000 genes per cell. Raw UMIs in each cell were both scaled and normalized via the SCTransform algorithm, regressing out unwanted source of variation arising from the percentage of mitochondrial genes. Highly variable genes were identified and selected for PCA reduction of high-dimensional data. To reduce dimensionality, the RunUMAP, Find-Neighbors, and FindClusters functions were used including the top 20 principal components. All visualization was done with BioTuring version 3 and BBrowserX (BioTuring). In situ photoconversion mLN photoconversion was performed as previously reported ([234]22). In brief, mice were anesthetized with 2% isoflurane during the procedure. mLNs and colons were exposed by an abdominal wall incision. Opaque, nonreflective material was used to shield the mouse from light exposure, except at the desired site. Photoconversion was performed using a low-power 80-mW LED-violet laser (Laser-land) in three cycles of 10 s powered on followed by 30 s powered off, keeping the light source approximately 2 cm away from the tissue. mLNs were photoconverted with three consecutive spots, comprising the whole mLN chain. Colons were photoconverted with six consecutives spots in the proximal colon, labeling approximately 3 cm of colon length. Next, mice were sutured and allowed to recover from anesthesia on a heating pad until responsive. After surgery, mice were given long-acting buprenorphine for analgesia. KikR^+ cells present outside the photoconverted area were quantified by flow cytometry to establish emigrated cells. Antibiotic treatment Mice were treated for 6 weeks with the following antibiotics in their drinking water: ampicillin (Sigma-Aldrich) 1 g/liter, vancomycin (Sigma-Aldrich) 0.5 g/liter, neomycin (Sigma-Aldrich) 1 g/liter, and metronidazole (Sigma-Aldrich) 1 g/liter, together (VNAM) or separately as indicated. Grape Kool-Aid (Kraft Foods; 20 g/liter) was used as vehicle. In the microbiota transplantation experiment, recipient mice were treated for 3 weeks with the VNAM antibiotic cocktail before recolonization. Microbiota transfer Trem2^+/+ and Trem2^−/− recipient mice (8 weeks old) were treated for 3 weeks with the following antibiotics in the drinking water: ampicillin (Sigma-Aldrich) 1 g/liter, vancomycin (Sigma-Aldrich) 0.5 g/liter, neomycin (Sigma-Aldrich) 1 g/liter, and metronidazole (Sigma-Aldrich) 1 g/liter. Two days after the antibiotic treatment was stopped, mice were gavaged with microbiota harvested from the ceca of different donors as indicated (n = 3 mice per donor genotype/treatment group). Briefly, each donor microbiota sample was weighed and brought into the Coy chamber. Three animals were euthanized for each donor genotype (Trem2^+/+ or Trem2^−/−) at day 0 (before tumor injection; referred to as mbt-d0) or at day 12 (48 hours after the first anti–PD-1 injection; referred to as mbt-d12). Cecal contents isolated from the three animals belonging to the same donor genotype/treatment group were combined and diluted 1:10 (wt/vol) with reduced PBS (PBS/0.05% l-cysteine–HCl) in 50-ml conical plastic tubes. Tubes were gently vortexed, and the resulting slurry was passed through a 100-μm–pore diameter nylon cell strainer (BD Falcon). The clarified cecal sample was then combined with an equal volume of a solution of PBS/0.05% l-cysteine–HCl/30% glycerol and aliquoted into 1.8-ml glass vials (E-Z vials, Wheaton). Tubes were crimped with covers containing a polytetrafluoroethylene (PTFE)/gray butyl liner (Wheaton) and stored at −80°C until use ([235]29). Each recipient mouse received 150 μl of the clarified cecal content one time 2 days after the end of the antibiotic treatment, as described above. V4–16S rRNA amplicon sequencing DNA was isolated from mouse fecal samples by bead beating in phenol:chloroform and purified on QiaQuick plates (QIAGEN). PCR amplification of V4 of bacterial 16S rRNA genes present in fecal samples was performed using published primers and cycling conditions ([236]29). Amplicons were sequenced using an Illumina MiSeq instrument (2 × 250 nucleotide reads; 970,038 ± 75,587 reads per sample). Amplicon sequences were oriented, trimmed of adaptor and primer sequences, and paired using the bbtools (37.02) software package ([237]https://sourceforge.net/projects/bbmap/“bbduk.sh” and “repair.sh” tools). DADA2 (1.8.0) was used to remove chimeric sequences and to identify ASVs ([238]30). Representative sequences for ASVs were then used for taxonomic profiling using the following multi-taxonomic assignment (MTA) approach. Each representative sequence was aligned using NCBI BLAST toolkit version 2.10.0 to the 16S rRNA gene reference database compiled by joining unique sequences from two databases: Ribosomal Database Project version 11.5 and the NCBI 16S ribosomal RNA Project. In vivo treatment with selected bacterial species R. gnavus VPI c7–9 (ATCC; ref. no. 29149) was grown at 37°C in Wilkins-Chalgren anaerobe broth (Oxoid; ref. no. CM0643) in a Coy chamber (atmosphere: 75% N[2], 20% CO[2], and 5% H[2]) without shaking. The overnight R. gnavus culture was mixed with filter-sterilized PBS solution containing 30% glycerol and 0.05% cysteine in a 1:1 (v/v) ratio (freezing medium). The resulting bacterial stocks were preserved at −80°C before experiments. The number of viable R. gnavus in the stocks was quantified by the CFU counting method. C. scindens was used as a control because it represents the next closest taxon assigned to the V4–16S sequence associated to R. gnavus against the NCBI 16S rDNA database. C. scindens was obtained from ATCC (35704) and was grown and prepared for gavage as specified above for R. gnavus. For the heat inactivation experiment, R. gnavus was kept at 100°C for 1 hour before freezing. Mice were injected with MCA/1956 tumors and treated with anti–PD-1 as previously described. Each mouse was treated with 150 μl of the stock via oral gavage, containing 1.5 × 10^7 viable R. gnavus. The oral gavage with the bacterium was performed daily for 5 days, starting at day 7 after tumor injection. Controls were treated with a vehicle (freezing medium). Anti–PD-1 was injected starting at day 9, and tumors were monitored daily as specified. Statistical analysis Data are shown as mean ± SEM. Two-way analysis of variance (ANOVA) was used to model data generated from factorial design with the combination of two factors, and two-way ANOVA for repeated-measures was used to model longitudinal tumor growth between treatments followed by post hoc comparisons on treatment difference at time points (Tukey’s or Šidák’s multiple comparisons test was used as indicated in figure legends). The Mann-Whitney U test was used to compare two groups. Statistics were calculated with GraphPad Prism 6 (GraphPad Software). PCoAs were performed using the R packages “ape” ([239]31) and “vegan” ([240]32). Indicator species analysis was conducted using the package “indicspecies” ([241]33). PERMANOVA analyses were performed using 1000 permutations to estimate the statistical significance (P value) of the comparison between the groups in the PCA. For gene expression comparison analysis ([242]Fig. 7D), Mann-Whitney U test was used for log-transformed counts. Supplementary Material S Table and Figures [243]NIHMS2009314-supplement-S_Table_and_Figures.pdf^ (2.1MB, pdf) Datafile S2 [244]NIHMS2009314-supplement-Datafile_S2.xlsx^ (20.7KB, xlsx) Datafile S3 [245]NIHMS2009314-supplement-Datafile_S3.xlsx^ (10.9KB, xlsx) Datafile S1 [246]NIHMS2009314-supplement-Datafile_S1.xlsx^ (87.6KB, xlsx) Datafile S6 [247]NIHMS2009314-supplement-Datafile_S6.xlsx^ (14.6KB, xlsx) Datafile S7 [248]NIHMS2009314-supplement-Datafile_S7.xlsx^ (191.7KB, xlsx) Datafile S5 [249]NIHMS2009314-supplement-Datafile_S5.xlsx^ (11.9KB, xlsx) Datafile S4 [250]NIHMS2009314-supplement-Datafile_S4.xlsx^ (16.3KB, xlsx) MDAR Checklist [251]NIHMS2009314-supplement-MDAR_Checklist.pdf^ (460.5KB, pdf) Acknowledgments: