ARTICLE https://doi.org/10.1038/s41467-019-10369-9 OPEN LIF regulates CXCL9 in tumor-associated macrophages and prevents CD8+ T cell tumor- infiltration impairing anti-PD1 therapy Mónica Pascual-García 1,2,8, Ester Bonfill-Teixidor1,2,8, Ester Planas-Rigol1, Carlota Rubio-Perez1, Raffaella Iurlaro1, Alexandra Arias1, Isabel Cuartas1, Ada Sala-Hojman1, Laura Escudero 1, Francisco Martínez-Ricarte3,4, Isabel Huber-Ruano1, Paolo Nuciforo 1, Leire Pedrosa5, Carolina Marques 6, 1234567890():,; Irene Braña1, Elena Garralda1, María Vieito1, Massimo Squatrito 6, Estela Pineda5, Francesc Graus5, Carmen Espejo 3,4, Juan Sahuquillo 3,4, Josep Tabernero 1,2,4 & Joan Seoane1,2,4,7 Cancer response to immunotherapy depends on the infiltration of CD8+ T cells and the presence of tumor-associated macrophages within tumors. Still, little is known about the determinants of these factors. We show that LIF assumes a crucial role in the regulation of CD8+ T cell tumor infiltration, while promoting the presence of protumoral tumor-associated macrophages. We observe that the blockade of LIF in tumors expressing high levels of LIF decreases CD206, CD163 and CCL2 and induces CXCL9 expression in tumor-associated macrophages. The blockade of LIF releases the epigenetic silencing of CXCL9 triggering CD8 + T cell tumor infiltration. The combination of LIF neutralizing antibodies with the inhibition of the PD1 immune checkpoint promotes tumor regression, immunological memory and an increase in overall survival. 1 Vall d Hebron Institute of Oncology (VHIO), Vall d’Hebron University Hospital, 08035 Barcelona, Spain. 2 CIBERONC, 028029 Madrid, Spain. 3 Vall d’Hebron Institut de Recerca (VHIR), Vall d’Hebron University Hospital, 08035 Barcelona, Spain. 4 Universitat Autònoma de Barcelona, 08193 Cerdanyola del Vallès, Spain. 5 Hospital Clinic, University of Barcelona and Institut d’Investigació Biomèdica August Pi i Sunyer (IDIBAPS), 08036 Barcelona, Spain. 6 Seve Ballesteros Foundation Brain Tumor Group, Spanish National Cancer Research Center, CNIO, 28029 Madrid, Spain. 7 Institució Catalana de Recerca i Estudis Avançats (ICREA), 08010 Barcelona, Spain. 8These authors contributed equally: Mónica Pascual-García, Ester Bonfill-Teixidor. Correspondence and requests for materials should be addressed to J.S. (email: jseoane@vhio.net) NATURE COMMUNICATIONS | (2019)10:2416 | https://doi.org/10.1038/s41467-019-10369-9 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-10369-9 L IF is a pleiotropic cytokine with a critical dual function in modest increase in survival (Fig. 1c, e, h, i, l, n, q, Supplementary embryonic development1,2. LIF regulates embryonic stem Figs. 2b–f, 3e, f). The blockade of LIF in the GL261 tumor model, cell (ESC) self-renewal preventing early differentiation3 and a tumor that did not express LIF, did not inhibit tumor growth is required for blastocyst implantation in the uterus4. Specifically, (Supplementary Fig. 2a, g). LIF generates a local immunosuppressive microenvironment in Neutralizing antibodies against LIF induced a marked decrease order to protect the embryo from the mother’s immune system4–7. in p-STAT3 levels showing that in these animal models (selected Ablation of LIF in mice prevents embryo implantation, and based on high LIF expression) LIF was the main cytokine women with loss-of-function LIF mutations show infertility with inducing the JAK-STAT3 pathway (Fig. 1d, m). Moreover, while impairment of in vitro fertilization4–7. Thus, LIF is a master we did not observe a significant decrease in Ki67 positive cells, we regulator of embryo implantation through the regulation of the detected an increase in cleaved caspase 3 (CC3) indicating that immune tolerance in pregnancy. the blockade of LIF induced tumor cell death (Fig. 1d, m). LIF can act as an oncogenic factor through the induction of the In order to evaluate the role of the immune system in the self-renewal of cancer-initiating cells8, the regulation of cancer- response to anti-LIF treatment, we performed experiments using associated fibroblasts9, as well as promoting radioresistance10 and immunodeficient animals. Treatment of GL261N tumors in RAG chemoresistance11. Importantly, the expression of LIF in cancer is −/− or NOD SCID mice (both strains of mice lacking the adaptive widely heterogeneous and some tumors express aberrantly high immune response) with anti-LIF did not show a significant levels of LIF while others do not express LIF12. Non-small cell impact on tumor growth (Supplementary Fig. 2h). This indicated lung cancer, breast cancer, colorectal cancer, and nasopharyngeal that in our models the antitumor response to the blockade of LIF cancer patients with tumors expressing high levels of LIF exhibit a was mainly mediated by the adaptive immune response. shorter overall survival8–10 indicating that high LIF expression We decided to further investigate the molecular mechanisms might be involved in the oncogenic progression of these malig- involved in the immune response to anti-LIF treatment. We nancies. However, the molecular mechanisms underlying the role observed a decrease in the number of protumoral TAMs (Fig. 1f, of LIF in the immune response in cancer have not been j, o) and, importantly, a concomitant increase in tumor elucidated yet. infiltration of CD8+ T cells upon anti-LIF treatment (Fig. 1d, g, Here, we find that tumors expressing high levels of LIF tend to k, m, p). Natural killer (NK) and regulatory T (Treg) cell numbers be infiltrated with tumor-associated macrophages (TAMs). LIF increased and decreased upon treatment with anti-LIF, respec- regulates the expression of CD163 and CD206, as well as several tively (Supplementary Fig. 2i–l). Infiltrating CD8+ T cells chemokines including CXCL9 and CCL2 in TAMs. Specifically, expressed Granzyme A (GZMA) suggesting that they were LIF triggers the epigenetic silencing of the CXCL9 locus by the mediating the cytotoxic effect (Supplementary Fig. 3a). Moreover recruitment of EZH2 to the CXCL9 promoter. The blockade of a compartment of CD8+ T cells expressed PD1 (Supplementary LIF in tumors expressing high levels of LIF releases the CXCL9 Fig. 3b, c). TAMs derived from recruited monocytes (CD11b+ repression acting as a chemoattractant of cytotoxic CD8+ T cells. Ly6G− Ly6C− CD49d+)14 were decreased in response to anti-LIF LIF is then a crucial regulator of CD8+ T cell tumor infiltration. or LIF shRNA (Supplementary Fig. 3d–f) and no major effect was Importantly, the combination of the blockade of LIF with observed on the dendritic cell population (CD11b+, CD11c+, checkpoint inhibitors induces tumor regression, immune mem- MHCII+) (Supplementary Fig. 3g) nor on the levels of IL12 or ory, and an increase in overall survival. IL10 in the tissue (Supplementary Fig. 3h). We then assessed whether the LIF-mediated regulation of the Results tumor immune infiltrates was the cause or the consequence of the LIF expression correlates with TAM tumor infiltration. To antitumor response. To this end, we performed an acute study the effect of LIF on the cancer immune system, we first treatment experiment where we treated mice with established determined whether LIF expression correlated with the presence tumors with anti-LIF for 4 days. The 4 day-treatment did not of tumor immune cell infiltrates in human cancer. We observed a affect tumor growth (Supplementary Fig. 3i) but was enough to significant correlation between LIF and TAMs across several engage CD8+ T cell tumor infiltration (Supplementary Fig. 3j). tumor types of The Cancer Genome Atlas (TCGA) (Fig. 1a, b, This showed that CD8+ T cell infiltration was not the result of the Supplementary Data 1). Glioblastoma (GBM), prostate adeno- anti-tumor response to the blockade of LIF. carcinoma, thyroid cancer and ovarian cancer were the 4 tumor types exhibiting the highest correlations between LIF and TAMs, LIF regulates the expression of protumoral cytokines in TAMs. while showing a high LIF expression across tumor samples We isolated CD11b+ cells from the ID8 mouse model treated or (Fig. 1a, b). A wide range of LIF expression was observed in GBM untreated with anti-LIF antibodies and performed a tran- tumors being it expressed by tumor cells and the immune cell scriptomic analysis. Several genes related to an oncogenic phe- infiltrates (Supplementary Fig. 1). notype were downregulated (i.e., CCL2, CCL3, CCL7, PF4, CTSK, CD206, CD163) and, interestingly, CXCL9 was upregulated The blockade of LIF induces an antitumor immune response. (Fig. 2a, Supplementary Data 2). The aforementioned gene We decided to use immunocompetent mouse models of GBM responses were validated by qRT-PCR in the ID8 and GL261N and ovarian cancer (tumor types where LIF strongly correlates models (Fig. 2b). with TAMs) to study the potential immune-modulating role of CXCL9 and CCL2 stood out as chemokines described to be LIF in cancer. We identified the GBM cell line, GL261N critical for CD8+ T cell tumor infiltration, and the recruitment of (a derivative of the GL261 cell line), the GFAP-tv-a RCAS- TAMs and Tregs15–17, respectively. We confirmed the CXCL9 PDGFA, shp53, shNF1 (RCAS) transgenic model13 and the and CCL2 regulation in TAMs (CD11b+ Ly6G− Ly6C−) by the ovarian cancer cell line, ID8, that generated tumors in the brain neutralization of LIF or LIF shRNA (Fig. 2c, Supplementary (GL261N and RCAS) and peritoneum (ID8) of mice expressing Fig. 3k–m). Immunofluorescence (IF) and isolation of TAMs high levels of LIF (Supplementary Fig. 2a). showed that CXCL9, CCL2, CD206, and CD163 were mainly We repressed LIF function in GL261N, RCAS and ID8 models expressed in TAMs (Fig. 2d, Supplementary Fig. 4, 5a) and using neutralizing antibodies, CRISPR/CAS9 or RNA interference inhibition of LIF function regulated their expression (Fig. 2c, d, technologies and observed a decrease in tumor growth and a Supplementary Figs. 3k–m, 4). CXCR3 (CXCL9 receptor), CCR2 2 NATURE COMMUNICATIONS | (2019)10:2416 | https://doi.org/10.1038/s41467-019-10369-9 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-10369-9 ARTICLE a 15 b 12 GBM PRAD 2 R 2 = 0.43 LIF log2(RSEM) R = 0.54 LIF log2(RSEM) 10 8 10 8 6 6 5 4 4 0 0.25 0.50 0.75 1 0 0.25 0.50 0.75 1 0 THCA OV 10 10 TAMs 0.62 0.58 0.57 0.54 0.52 0.49 0.45 0.42 0.42 0.39 0.38 0.33 0.32 0.31 0.25 0.25 0.21 0.14 0.11 0.09 –0.15 2 R = 0.33 2 R = 0.42 LIF log2(RSEM) 9 9 2 PCPG UVM LGG GBM THCA LIHC PRAD KICH OV SARC BLCA HNSC BRCA SKCM LUSC MESO KIRC CHOL COADREAD UCEC STAD UCS KIRP ACC PAAD LUAD ESCA CESC R 8 8 –1 +1 7 7 6 6 ssGSEA 0 0.25 0.50 0.75 1 0 0.25 0.50 0.75 1 c 2.5 × 107 ** i 6 × 106 * l 1 × 104 *** Abdominal vol. (mm3) GL261N 2 × 107 RCAS ID8 8 × 103 Total flux (p/s) Total flux (p/s) IgG (control) IgG (control) 4 × 106 IgG (control) 1.5 × 107 6 × 103 α-LIF α-LIF α-LIF 1 × 107 4 × 103 2 × 106 5 × 106 2 × 103 Days Days 0 3 6 9 12 0 0 10 20 30 40 0 Days 14 20 27 33 40 0 d Control α-LIF % of + cells e Control α-LIF j m Control α-LIF % of + cells n Control α-LIF * * 15 *** 0.08 15 p-STAT3 p-STAT3 10 0.06 10 %TAMs 5 0.04 5 0 0.02 0 10 15 0 7.5 f * g * k ** 10 o ** p * Ki67 4 Ki67 0.4 6 0.06 3 5 5 %CD8 T cells 3 %CD8 T cells %CD8 T cells 2.5 0.3 4 0.04 2 %TAMs %TAMs 0 2 0 0.2 ** + *** + + 30 2 0.02 1 40 0.1 1 20 CC3 CC3 20 0 0 0 0 0 h 100 10 q 100 0 0 **** *** *** %Survival %Survival 4 6 3 50 50 CD8 CD8 4 2 ** 2 1 0 0 0 Time 0 10 20 30 40 0 Time 0 10 20 30 40 Fig. 1 LIF blockade decreases tumor growth and regulates immune cell infiltration. a Top-panel, distribution of LIF mRNA expression (log2 RSEM) across 28 distinct solid tumors (see Supplementary Data 1b) in boxplots (middle line depicts the median and the whiskers the interquartile range). Black line represents the estimated cut-off between low expression/background noise. Bottom panel, correlation values (Pearson R2 values) between LIF expression and the relative abundance of TAMs based on ssGSEA of the gene signature (see Supplementary Data 1a). Correlation values are only shown if the correlation is significant (adjusted P-value < 0.1) (see Supplementary Data 1c). Cancer types are sorted according correlation value between LIF and TAMs relative abundance. b Linear regression plots of LIF expression and relative abundance (ssGSEA rescaled from 0 to 1 for visualization purposes) of TAMs in GBM, prostate adenocarcinoma (PRAD), thyroid carcinoma (THCA) and ovarian carcinoma (OV) cohorts. Shade represents the confidence intervals of the regression estimate. c, i, l Tumor growth of GL261N (c), RCAS (i), and ID8 (l) models measured as total flux (p/s) or abdominal volume (mm3), respectively. Scheme representing the experimental procedure is shown. Anti-LIF or isotype control (IgG) treatment started on the day of surgery (GL261N and RCAS) or 14 days post-inoculation (dpi) (ID8). d, m Representative p-STAT3, Ki67, CC3, and CD8 IHC images and percentages of staining for GL261N (d) and ID8 (m) tumors. e, n Representative images of GL261N (e) and ID8 (n) tumor-bearing mice treated with anti-LIF or IgG. f–g, j–k, o–p, Percentages of CD11b+ F4/80+ CD163+ CD206+ MHCIIlow TAMs (f, o) or CD11b+ Ly6G− Ly6C− CD163+ CD206+ MHCIIlow (j) and CD8+ T cells (CD3+ CD8+) of GL261N (g), RCAS (k), and ID8 (p) tumors analysed by flow cytometry. h, q, Overall survival of GL261N (h) and ID8 (q) models treated with anti-LIF or IgG. Data are mean ± SEM. Statistical analyses by Mann–Whitney T-test and Log-rank test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 (CCL2 receptor) and LIFR were expressed in TAMs and CD8+ tumors in CXCL9−/− and CCL2−/− mice and hence the tumors T cells (Supplementary Fig. 5b). were infiltrated with CXCL9−/− or CCL2−/− TAMs. We then In order to address the relevance of the regulation of CXCL9 transplanted the tumors into wild type mice and assessed CD8+ T and CCL2 in the LIF oncogenic function, we used CXCL9 and cell tumor infiltration in response to the blockade of LIF. We CCL2 knockout (CXCL9−/−, CCL2−/−) mouse models, and observed that the TAM-specific deletion of CXCL9 reduced the blocking antibodies against CXCL9 and CCL2. Interestingly, the infiltration of CD8+ T cells in response to the blockade of LIF anti-tumor response to the inhibition of LIF was blunted in the (Supplementary Fig. 5c, d). In contrast, the TAM-specific deletion CXCL9−/− mice but not in the CCL2−/− mice (Fig. 2e). Similarly, of CCL2 did not impact on the CD8+ T cell tumor recruitment the CXCL9 neutralizing antibody but not the CCL2 antibody (Supplementary Fig. 5c, d). impaired the anticancer response to anti-LIF (Fig. 2e). These We went back to the analysis of TCGA datasets of human results indicated that the main mediator of the anti-LIF response GBM and ovarian cancer and observed a significantly positive was CXCL9. As expected, the blockade of CXCL9 prevented the correlation between LIF and CCL2, CD163, and CD206 in both increase of CD8+ T cell tumor infiltration in response to anti-LIF tumor types (Supplementary Fig. 6). No correlation was (Fig. 2f). We then investigated whether the TAM-specific observed between LIF and CXCL9. In order to validate these expression of CXCL9 was responsible for the regulation of the results at the protein level, we analysed a cohort of 20 GBM tumor infiltration of CD8+ T cells. We generated GL261N patients and performed LIF, CXCL9, CCL2, CD163, and NATURE COMMUNICATIONS | (2019)10:2416 | https://doi.org/10.1038/s41467-019-10369-9 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-10369-9 a PF4 Control α-LIF 2.5 CCL7 PF4 –4.14 CCL3 CCL7 –2.58 2 CCL2 CD163 –log10 adjusted P-value CD206 CCL3 –2.53 1.5 CTSK CXCL9 CTSK –2.4 CD206 –2.03 1 CD163 –1.93 0.5 CCL2 –1.64 CXCL9 3.1 0 Expression log2 FC –5 –4 –3 –2 –1 0 1 2 3 log2 fold change (FC) 3 6 10 b 80 ID8 5 GL261N Control α-LIF * **** 60 4 40 20 3 mRNA mRNA 3 *** *** *** *** *** *** *** **** **** **** **** **** * **** 2 2 1 1 0 0 CCL3 CCL7 CD206 PF4 CCL2 CD163 CTSK CXCL9 CCL3 CCL7 CD206 PF4 CCL2 CD163 CTSK CXCL9 c d DAPI/Iba1/CXCL9 DAPI/Iba1/CCL2 DAPI/Iba1/CD206 DAPI/Iba1/CD163 15 * 4 ** MFI CCL2 (TAMs) Control %TAMs CCL2+ 3 10 2 5 1 GL261N 0 0 α-LIF 15 * 4 * MFI CXCL9 (TAMs) %TAMs CXCL9+ 3 10 2 5 1 CXCL9+ CCL2+/Iba1+ CD206+/Iba1+ CD163+/Iba1+ 0 50 100 0 20 40 60 80 100 % of cells 0 10 20 30 40 50 0 20 40 60 80 100 0 0 *** *** *** *** Control α-LIF e * * * f 3 × 107 * 2.5 × 107 * 8 * * 2× 107 6 Total flux (p/s) FC CD8+ T cells 1.5 × 107 4 1 × 107 2 5 × 106 0 0 α-LIF – + – + – + – + – + α-LIF – + – + α-CCL2 CCL2–/– α-CXCL9 CXCL9–/– α-CXCL9 g LIF CCL2 CD163 CD206 CXCL9 Tumor high LIF Tumor low LIF h 250 R 2 = 0.76*** 2 250 R = 0.52** 2 250 R = 0.87*** 2 250 R = –0.46* 200 200 200 200 LIF H-score 150 150 150 150 100 100 100 100 50 50 50 50 0 0 0 0 50 100 150 200 50 100 150 200 50 100 150 200 50 100 150 CCL2 H-score CD163 H-score CD206 H-score CXCL9 H-score CD206 IHC of the tumors. We noted a strong positive LIF induces the epigenetic silencing of CXCL9. We decided to correlation between LIF and CCL2, CD163 and CD206 explore the molecular mechanisms involved in the regulation of (Fig. 2g, h). CXCL9 was expressed in isolated clusters of cells CXCL9 by LIF in macrophages. To this end, we studied the effect and its levels showed an inverse correlation with LIF in human of LIF on primary cultures of mouse bone marrow derived GBM (Fig. 2g, h). macrophages (BMDMs). LIF regulated the expression of several 4 NATURE COMMUNICATIONS | (2019)10:2416 | https://doi.org/10.1038/s41467-019-10369-9 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-10369-9 ARTICLE Fig. 2 LIF regulates CXCL9, CCL2, CD206, and CD163 in TAMs. a Differential expression analysis of isolated CD11b+ cells from anti-LIF treated ID8 mice vs. control. Volcano plot representing the genes significantly (Q-value < 0.1) overexpressed (brown) and significantly underexpressed (turquoise). Heatmap representing the expression values of the indicated genes, each column represents a sample and each row a gene. The last column represents the log2 fold change (log2 FC) of gene expression. b mRNA expression for the indicated genes in isolated CD11b+ cells from anti-LIF treated or untreated ID8 and GL261N tumors. c Percentage and mean fluorescence intensity (MFI) of CCL2 and CXCL9 in TAMs (CD11b+ Ly6G− Ly6C−) from anti-LIF treated or untreated GL261N tumors. d Representative IF images of Iba1 and the indicated markers stainings of GL261N tumors (see Supplementary Fig. 4). Scale bar, 20 μm. Bottom, percentage of double positive cells relative to the TAM marker positive cells. CXCL9 quantification is relative to the total number of cells. e Tumor growth of GL261N in WT, CXCL9−/−, and CCL2−/− mice or mice treated with the indicated antibodies is shown as total flux (p/s). f Fold change (FC) of tumor infiltrating CD8+ T cells in the indicated treatments. Data are mean ± SEM. g, h Representative IHC of the indicated markers from 20 GBM tumors. The degree of staining was quantified using H-score method. Correlations between LIF and CCL2, CD206, CD163, and CXCL9 with the R-squared coefficients (R2) were calculated (h). Statistical analyses by Mann–Whitney T-test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 M1-like and M2-like markers induced by IFNγ or IL4 in BMDMs this experimental model cannot be used for long experiments, (Fig. 3a). CXCL9 expression was not detected except when and hence survival analysis, due to graft-versus-host disease. BMDMs were treated with IFNγ. Recombinant LIF repressed the Interestingly, mice treated with anti-LIF showed an increase in induction of CXCL9 by IFNγ both at the mRNA or protein levels CD8+ T cell tumor infiltration and most of the infiltrating CD8+ (Fig. 3b, c). A similar result was observed when BMDMs were T cells expressed the CXCL9 receptor, CXCR3 (Fig. 4e, treated with LPS (Supplementary Fig. 7a). Importantly, CXCL9 Supplementary Fig. 10a, b). We also isolated CD3+ T cells from was also regulated by IFNγ and LIF in patient-derived TAMs PBMCs and infected them with lentiviruses encoding luciferase. (CD11b+ CD14+) obtained from fresh human GBM tumors Interestingly, most of the cells expressed CXCR3. We inoculated (Fig. 3d, Supplementary Fig. 7b, c). Thus, LIF acted as a repressor these cells in mice bearing patient-derived tumors and were able of CXCL9 induction. CXCL9 has been described to undergo to monitor the anti-LIF-mediated CD3+ T cell tumor infiltration epigenetic regulation18 and LIF is known to regulate the epige- by bioluminescence (Fig. 4f, Supplementary Fig. 10c, d) netic status of embryonic stem cells1,19. In line with this, we found that treatment with LIF increased the levels of H3 lysine 27 trimethylated (H3K27me3), decreased the levels of acetylated H4 LIF and PD1 blockade promotes tumor regression. So far, our (H4ac), and increased EZH2 binding to the CXCL9 promoter data indicated that the inhibition of LIF induced a change in the region, altogether showing that LIF was regulating CXCL9 phenotype of TAMs increasing the expression of CXCL9 and expression through epigenetic silencing (Fig. 3e). leading to the recruitment of CD8+ T cells to the tumor. As expected based on the literature21,22, we observed that a sub- LIF regulates CD8+ T cell tumor infiltration. To confirm that population of CD8+ T cells within the tumor in our mouse LIF regulates immune cell tumor infiltration through the models were PD1 positive (see Supplementary Fig. 3b, c). This repression of CXCL9 in tumors from actual cancer patients, we indicated that the blockade of PD1 in this context might increase generated organotypic tissue cultures of GBM specimens freshly the anti-tumor response to anti-LIF agents. We decided to use obtained from patients. These organotypic models allow for the our immunocompetent mouse models bearing overt tumors with short-term culture of tumor slices that maintain the tissue anti-LIF and anti-PD1 antibodies and observed that the combi- architecture and stroma (including immune cells) of the patient’s nation of the blockade of LIF and PD1 further decreased tumor tumor20. We obtained organotypic tissue cultures from three growth when compared to each individual treatment. Moreover patient-derived tumors expressing high levels of LIF (Fig. 3f, and importantly, the combined treatment with anti-LIF and anti- Supplementary Fig. 8). In all three cultures a large infiltration of PD1 increased overall survival and induced tumor regression TAMs was present as detected by the Iba1 marker and most of (Fig. 4g, h). the TAMs expressed CCL2, CD163, and CD206. Interestingly, a We collected the mice exhibiting complete tumor regression 3-day treatment of the organotypic culture with a neutralizing and reinoculated 3 × 105 tumor cells. No tumor appeared in these antibody against LIF promoted a decrease in CCL2, CD163, and mice while tumors rapidly grew in naive mice inoculated in CD206 and an increase in CXCL9 expression (Fig. 3f, Supple- parallel with the same number of cells (Fig. 4i). The result of this mentary Fig. 8). rechallenge experiment indicated that the combined treatment We then decided to test whether the regulation of LIF could with anti-LIF and anti-PD1 generated immunological memory. impact on the immune cell tumor infiltration. After anti-LIF treatment, we incubated the organotypic slices from three patient- derived tumors (expressing high levels of LIF) with peripheral Discussion blood mononuclear cells (PBMCs) from the same patient (Fig. 4a). CD8+ T tumor infiltration and TAMs determine among other Treatment with anti-LIF increased CXCL9 and decreased CCL2 things the response to cancer immunotherapy17,23–27. We have expression (Fig. 4b), while inducing immune cell infiltration into observed that LIF acts as an oncogenic factor through the reg- the Matrigel surrounding the tumor specimen (Fig. 4b). Notably, ulation of the immune system. LIF establishes a cross-talk CD8+ T cells were recruited to the tumor tissue upon LIF between the innate and the adaptive immune response since LIF blockade (Fig. 4b, c, Supplementary Fig. 9) and this effect was neutralization regulates the expression of a chemokine program dependent on CXCL9, since its neutralization prevented CD8+ T in TAMs, including CXCL9, leading to an increase in CD8+ T cell cell infiltration (Fig. 4d). tumor infiltration (Fig. 4j). This indicates that LIF is responsible We next decided to confirm this result in the context of an for CD8+ T cell tumor exclusion and that it could be involved in in vivo model. We inoculated tumor fragments from four resistance to immune checkpoint blockade. Indeed, we explored patients, whose tumors expressed high LIF levels, in NSG mice transcriptomic data of checkpoint inhibitor clinical trials28 and and treated these mice with the LIF neutralizing antibody for observed that LIF, and the genes that we found regulated by LIF, 5 days. Next, we inoculated each patient’s PBMCs in the mice and were associated to resistance to checkpoint inhibitor therapy analyzed the short-term tumor infiltration of T cells. Note that (Supplementary Fig. 11). NATURE COMMUNICATIONS | (2019)10:2416 | https://doi.org/10.1038/s41467-019-10369-9 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-10369-9 a 2.5 × 10 4 40 **** 800 **** 15 *** 2 × 104 **** 30 600 CXCL9 mRNA NOS2 mRNA TNFα mRNA CCL5 mRNA 10 1.5 × 104 20 400 4 1 × 10 5 10 200 5 × 103 0 0 0 0 LIF – + – + – + LIF – + – + – + LIF – + – + – + LIF – + – + – + IFNγ – – + + – – IFNγ – – + + – – IFNγ – – + + – – IFNγ – – + + – – IL4 – – – – + + IL4 – – – – + + IL4 – – – – + + IL4 – – – – + + 60 ** 2 × 104 ** 20 ** 20 *** CCL2 mRNA 4 15 CD206 mRNA 1.5 × 10 15 ARG1 mRNA IL10 mRNA 40 1 × 104 10 10 20 **** *** **** *** **** **** 3 5 × 10 5 5 0 0 0 0 LIF – + – + – + LIF – + – + – + LIF – + – + – + LIF – + – + – + IFNγ – – + + – – IFNγ – – + + – – IFNγ – – + + – – IFNγ – – + + – – IL4 – – – – + + IL4 – – – – + + IL4 – – – – + + IL4 – – – – + + b c d e CXCL9 promoter Control LIF 1 × 104 60 **** 15 200 M.m. *** ** –988 –458 8 × 103 40 *** H3K27me3 20 EZH2 2 H4ac CXCL9 mRNA 0.2 CXCL9 pg/ml 150 CXCL9 pg/ml 10 *** *** *** Relative enrichment 6 × 103 20 0.15 15 1.5 100 4 × 103 **** 5 0 50 0.1 10 1 2 × 103 *** 0 **** 0 0 0.05 5 0.5 LIF – + – + – + – + – + LIF – + – + LIF – + – + IFNγ – – IFNγ – – + + IFNγ – – + + 0 0 0 Promoter GAPDH CXCL9 GAPDH CXCL9 GAPDH CXCL9 f Patient 1 Patient 2 Patient 3 Control Control Control DAPI/Iba1/LIF Control α-LIF α-LIF α-LIF α-LIF % of cells 60 *** DAPI/Iba1/CXCL9 CXCL9+ 40 20 0 100 DAPI/Iba1/CD206 *** CD206+/Iba1+ 80 60 40 20 0 DAPI/Iba1/CD163 100 *** CD163+/Iba1+ 80 60 40 20 0 100 DAPI/Iba1/CCL2 ** 80 CCL2+/Iba1+ 60 40 20 0 Fig. 3 LIF represses CXCL9 through epigenetic silencing. a, b qRT-PCR analysis of the indicated genes in BMDMs. BMDMs were pre-incubated with 20 ng/ml LIF for 72 h and then stimulated with 5 ng/ml IFNγ or 10 μg/ml IL4 during 6 h (a) or with 0.1, 0.5, 1 and 5 ng/ml IFNγ for 24 h (b). c CXCL9 ELISA from BMDMs pre-incubated with 20 ng/ml LIF and then stimulated with 0.1 ng/ml IFNγ for 24 h. d CXCL9 ELISA from human CD11b+sorted cells (77% CD11b+ CD14+, see Supplementary Fig. 7b, c) from human GBM cultured with 20 ng/ml LIF for 72 h and then with 0.1 ng/ml IFNγ for 24 h. e ChIP of Tri-methyl- histone H3 (H3K27me3), EZH2 and acetyl-histone4 (H4ac) was performed in BMDMs treated with 20 ng/ml LIF for 72 h. Scheme shows the analysed CXCL9 promoter region. Representative data are presented as mean ± SD. f Representative images of IF of the indicated markers in human GBM organotypic slices (patients 1, 2, 3) incubated with 10 μg/ml anti-LIF for 3 days. Scale bar, 20 μm. (see Supplementary Fig. 8). Bottom, percentage of double positive cells relative to Iba1+ cells and percentage of CXCL9+ cells relative to the total number of cells. Data are mean of all patients ± SEM. Statistical analyses by Student’s t-test or Mann–Whitney T-test. *P < 0.05, **P < 0.01; ***P < 0.001; ****P < 0.0001 6 NATURE COMMUNICATIONS | (2019)10:2416 | https://doi.org/10.1038/s41467-019-10369-9 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-10369-9 ARTICLE a b DAPI/CXCL9 DAPI/CCL2 PBMCs DAPI/CD8 Tumor Control PBMCs Tumor α-LIF c e Control α-LIF 15 ** CXCL9 mRNA CCL2 mRNA # PBMCs in matrigel % CD8+ T cells FC CD8+ T cells 20 *** 0 2 4 6 0 0.5 1 1.5 0 100 200 300 0 1 2 3 4 5 10 ** * * * 15 Ratio CD8 Tcells 5 (Tumor/Blood) f 1.5 × 10 5 * + 0 10 Control d 15 * * 5 1 × 105 Total flux (p/s) FC CD8 T cells 10 0 + Pt7 Pt9 5 × 104 α-LIF 5 Pt8 Pt10 0 α-CXCL9 – + – + 0 g GL261N IgG (control) h 2309 1571 1220 2629 2484 857 337 292 278 936 436 322 245 991 888 814 709 209 268 α-LIF α-PD1 α-LIF + α-PD1 150 147 150 102 100 65 64 % Tumor growth Days 0 6 8 12 50 17 100 0 % Survival –1.2 –6.8 50 –50 –57 –65 * –84 –87 –100 0 Time 0 20 40 60 i j + GL261N IgG (control) CD8 α-LIF α-PD1 T cells α-LIF + α-PD1 CXCL9 Control Rechallenge Days 0 6 8 12 21 25 0 Control 2 × 107 *** LIF 1.5 × 107 Tumor Total flux (p/s) cells 1 × 107 Rechallenge 5 × 106 CD206 TAMs CD163 3 × 104 0 CCL2 Control Rechallenge The immunosuppressive response to LIF is not only seen in designed through evolution to protect the embryo from the cancer. LIF has been shown to inhibit the immune response immune system of the mother and prevent early differentiation of in organ transplantation29, autoimmune diseases30 as well as in ESCs. Tumors abduce this dual mechanism of action by secreting embryo implantation2,6. As in many other cases, cancer hijacks a aberrantly high levels of LIF. In this way, LIF protects tumors normal developmental program for its own gain. LIF has been from the host immune response and, as we and others have NATURE COMMUNICATIONS | (2019)10:2416 | https://doi.org/10.1038/s41467-019-10369-9 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-10369-9 Fig. 4 LIF blockade induces CD8+ T cell tumor infiltration and promotes tumor regression when combined with anti-PD1. a Schematic representation of GBM patient-derived xenografts and human organotypic models. b Organotypic specimens were treated with anti-LIF for 72 h and then cultured with PBMCs for 24 h. CXCL9 and CCL2 mRNA expression levels are shown. Representative images (patient 4, 5, 6) of CFSE-stained PBMCs into Matrigel containing GBM specimens and IF of the indicated factors in organotypic tissues are displayed. Bars represent quantification of five different fields of each condition. Data are presented as mean ± SD. Scale bar, 20 μm. c, d FC of CD8+ T infiltrating cells detected by flow cytometry in organotypic tissues treated with anti-LIF (c, d) and/or anti-CXCL9 (1.5 μg/ml) (d) for 72 h and then cultured with PBMCs for 48 h (patient 4, 5, 6). e CD8+ T infiltrating cells into subcutaneously engrafted GBM specimens in NSG mice. Bar graph represents the ratio of CD8+ T cells detected by flow cytometry in the tissue vs. CD8+ T cells detected in the blood of the same animal. Four patients (7, 8, 9, 10) with their corresponding PBMCs were evaluated and are represented with different colors. f Representative images of treated or untreated patient-derived xenografts (PDX) bearing mice 48 h after luciferase CD3+ T cells inoculation. T cells infiltration is measured as total flux (p/s) within the tumor. Data are mean ± SEM. g Scheme represents the experimental procedure of GL261N mice treated with anti-LIF, anti-PD1 or the combination. Overall survival determined by Kaplan-Meier curves is shown. h Tumor growth of the treated GL261N model represented as a fold change of tumor size between 13 and 6 dpi. Representative images are shown. i Scheme representing the experimental procedure for regression study. Images of mice and tumor size measured by total flux (p/s) at 13 dpi are shown. j Schematic representation shows the effect of LIF in CD8+ T cell tumor infiltration. Statistical analyses by Mann–Whitney T-test or Log-rank test. *P < 0.05; **P < 0.01; ***P < 0.001 previously reported8,31, prevents differentiation of cancer- national directives. Female C57BL/6 and NOD/SCID were purchased from Janvier, initiating cells. RAG−/−, CCL2−/− and CXCL9−/− from Jackson Laboratories and NOD scid gamma (NSG) from Charles River. For brain tumor models, 3 × 105 GL261N, Our work reveals a promising therapeutic target in cancer. The GL261 or RCAS cells, all of them with luciferase expression, were stereotactically blockade of LIF in tumors expressing high levels of LIF promotes inoculated into the corpus striatum of the right brain hemisphere (1 mm anterior effector T cell tumor infiltration transforming a T cell excluded and 1.8 mm lateral to the lambda; 2.5 mm intraparenchymal) of 8-week-old tumor into an inflamed tumor. This effect synergizes with C57BL/6 mice. For ovary tumor model, 5 × 106 ID8 ovarian cancer cells were checkpoint inhibitors inducing tumor regression in very aggres- intraperitoneally injected into 8-week-old C57BL/6 mice. A dose of 300 µg (ID8) or 600 µg (GL261N, GL261 and RCAS) of anti-LIF or a control IgG was administered sive tumors such as GBM. Therapies directed against LIF may intraperitoneally twice a week. Additionally, a dose of 200 µg of rat anti-mouse therefore have broad applications in the treatment of cancer. PD1 blocking antibody (anti-PD1, BioXCell; BE0146), anti-mouse/human/rat CCL2 antibody (MCP-1, BioXcell; BE0185) or 3 µg of anti-mouse CXCL9 antibody (R&D; AF-492-NA) was administered intraperitoneally twice a week. Tumor Methods progression was monitored by body weight and by abdominal girth (ID8), or Patients. Human GBM specimens were obtained from the Vall d’Hebron Uni- bioluminescence measurements using the Xenogen IVIS® Spectrum (GL261N, versity Hospital and Clinic Hospital. The clinical protocol was approved by the Vall GL261, and RCAS). Mice were euthanized when they exhibited clinical signs of d’Hebron Institutional Review Board and Clinic Hospital (CEIC), with informed disease or distress (i.e., cachexia, anorexia, or increased respiratory rates) or when consent obtained from all subjects. tumors began to interfere with normal body functions. For brain tumor transplantation, GL261N tumors were isolated from tumor- Human tumoroid and organotypic slice cultures. GBM tumoroids were gener- bearing mice, embedded in pieces into Matrigel and subcutaneously implanted in ated as follows8. Briefly, tumor samples were processed within 30 min after surgical new control or anti-LIF treated mice. resection. Minced pieces of human GBM samples were digested with 200 U/ml For regression study, 3 × 105 GL261N cells were inoculated in 6 tumor bearing- collagenase I (Sigma) and 500 U/ml DNase I (Sigma) in PBS for 1 h at 37 °C with mice with complete regression after anti-LIF, anti-PD1 combination treatment. constant vigorous agitation. The single-cell suspension was filtered through a 70 Ten naive mice were inoculated 3 × 105 GL261N cells in parallel. μm cell strainer (BD Falcon) and washed with PBS. Finally, cells were resuspended and subsequently cultured in GBM medium that consisted of Neurobasal medium supplemented with B27, penicillin/streptomycin (all from Life Technologies) and Primary cell cultures and cell lines. BMDMs were obtained from 6 to 10-week- growth factors (20 ng/ml EGF and 20 ng/ml FGF-2 (PeproTech)). GBM organo- old C57BL/6 mice32. Briefly, bone marrow precursors were cultured in DMEM typic slice cultures were generated as follows. After resection, surgical specimens (Life Technologies), supplemented with 20% heat inactivated FBS and 30% L-cell were cut with a scalpel into rectangular blocks of 5–10 mm length and 1–2 mm conditioned medium (cm) as a source of macrophage-colony stimulating factor. width and individually transferred into 0.4 µm membrane culture inserts (Milli- Differentiated macrophages were obtained after 6 days culture. L-cell cm was pore) within 6-well plates. Before placing the inserts into 6-well plates, 1.2 ml of obtained from L929 cells grown in DMEM supplemented with 10% heat- Neurobasal medium (Life Technologies) supplemented with B27 (Life Technolo- inactivated FBS (Life Technologies). Human macrophages were isolated from gies), penicillin/streptomycin (Life Technologies) and growth factors (20 ng/ml human GBM specimens. Briefly, tumor tissue was enzymatically digested with EGF and 20 ng/ml FGF-2) (PeproTech) were placed into each well. The cultures Tumor Dissociation kit and CD11b+ cells were isolated using CD11b magnetic were kept at 37 °C with constant humidity, 95% air and 5% CO2. After one day, beads and the MultiMACS Cell24 separator Plus (all from Miltenyi Biotec). CD11b + cells obtained were cultured in RPMI medium supplemented with 10% heat- slices were treated with 10 µg/ml of a rat anti-mouse/human LIF blocking antibody (referred to as anti-LIF) (developed in house) or with its corresponding normal IgG inactivated FBS (Life Technologies). Recombinant LIF, IFNγ, LPS, and IL4 were (10 µg/ml) for 3 days. For the blocking CXCL9 studies, a neutralizing mouse purchased from Millipore, R&D Systems, Sigma and Creative BioMart, monoclonal antibody against human CXCL9 (R&D Systems; MAB392) was added respectively. to the culture at 1.5 µg/ml. In some occasions, 0.1 ng/ml of human rIFNγ (R&D GL261N were derived from GL261 cells that were obtained from Charles River Systems) was added for 24 h. In parallel, peripheral blood mononuclear cells and cultured in RPMI (Life Technologies). ID8 was a generous gift from Dr. (PBMCs) were obtained from the whole blood of the same patient by centrifuge George Coukos, Ludwing Institute for Cancer Research, Lausanne, and cultured in density separation using Lymphosep (Biowest). PBMCs were cryopreserved in DMEM. The media for both cell lines were supplemented with 10% FBS and RPMI medium supplemented with 10% inactivated FBS and 10% DMSO until use. penicillin/streptomycin (Life Technologies). RCAS GFAP-tva PDGFA, shp53, For immune cell infiltration assays, control or anti-LIF slices were embedded shNF1 (RCAS) cells13 were cultured in Neurocult basal medium supplemented into growth factor deprived Matrigel (Corning) with subsequent addition of 1 × 106 with Neurocult proliferation supplement, heparin, FGF, and EGF (all from PBMCs into 24-well plate in complete (10% inactivated FBS) RPMI medium. In StemCell Technologies). addition, supernatants were collected and organotypic slices were recovered from Matrigel and further processed for IF and flow cytometry. In some conditions, GBM patient-derived in vivo models. Surgical human GBM specimens were cut PBMCs were resuspended with PBS at a concentration of 1 × 106 cells/ml and with a scalpel into rectangular blocks of 5–10 mm length and 1–2 mm width and incubated for 20 min with 5 µM Cell Trace CFSE (Invitrogen). After the incuba- transferred into a tube containing 100 µl of Matrigel (Corning) supplemented with tion, cells were washed with RPMI and added to the sections embedded into 5 ng/ml of rIFNγ (R&D Systems). Each specimen was introduced subcutaneously Matrigel. After 24 h, fluorescent PBMCs invasion into Matrigel was evaluated into one or two flanks of 4 to 5-weeks-old NSG mice. Half of the mice received under microscope by counting migrating cells in five different areas per each anti-LIF (600 µg) treatment every three days starting at the day of the implantation condition. of the samples. Minimum of three specimens per condition were implanted. After 5 days, 1 × 107 PBMCs obtained from the same patient, and previously stimulated Animals and in vivo experiments. All animal experiments were approved by and with 150 U/ml of rIL2 (R&D Systems) for 24 h, were intravenously inoculated into performed according to the guidelines of the Institutional Animal Care Committee the mice in exception of one (control without PBMCs). Forty-eight hours post- of the Vall d’Hebron Research Institute in agreement with the European Union and inoculation, engrafted tumors were collected and processed to determine CD8+ 8 NATURE COMMUNICATIONS | (2019)10:2416 | https://doi.org/10.1038/s41467-019-10369-9 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-10369-9 ARTICLE T cells infiltration by flow cytometry. Total blood of the mice was also processed Quantitative real-time PCR. Cells were lysed for mRNA extraction (RNeasy Mini using Ficoll-Paque Premium (Ge Healthcare) in order to obtain the PBMCs. or Micro Kit, Qiagen), retrotranscription (iScript Reverse Supermix from BioRad For immune cell infiltration assay, 1 × 107 of luciferase expressing CD3+ T cells for mRNA), and qRT-PCR was performed using Taqman probes from Applied were intravenously introduced into control or anti-LIF (600 µg) treated GBM Biosystems, according to manufacturer’s recommendations. For paraffin- patient-derived models. After 48 h CD3+ T cells recruitment into the tumor was embedded sections, RNA was obtained by using High Pure FFPET RNA isolation monitored by bioluminescence measurements using the Xenogen IVIS® Spectrum. kit (Roche) and following manufacturer instructions. Reactions were carried out in a CFX384 Touch™ Real-Time PCR Detection System (Bio-Rad) and results were expressed as fold change calculated by the Ct method relative to the control sample. Tumor digestion, flow cytometry, and cell isolation. Mice were euthanized and Murine or human ACTB or GAPDH were used as internal normalization controls. tumors were isolated. GL261N and RCAS tumors were enzymatically digested with Brain Tumor Dissociation kit and myelin was removed with Myelin Removal Beads II (all from Miltenyi Biotec). ID8 tumors were processed with Mouse Tumor Immunohistochemical and immunofluorescence staining. Immunohistochem- Dissociation kit (Miltenyi Biotec) and ascitic liquids were collected. Human GBM ical (IHC) staining was performed as follows. Slides were deparaffinized and specimens of the organotypic model and the patient-derived xenografts were hydrated. Antigen retrieval was performed using pH 6 or pH 9 Citrate Antigen enzymatically digested with Human Tumor Dissociation kit (Miltenyi Biotec). Retrieval Solution (DAKO), 10 min 10% peroxidase (H2O2) and blocking solution From GL261N cell suspension, CD11b+ cells isolation was performed using (2% BSA) for 1 h at room temperature. As a detection system, EnVision FLEX + anti-Ly6C-APC and anti-APC microbeads and anti Ly6G microbeads to deplete (DAKO) was used according to the manufacturer’s instructions, followed by Ly6G+ and Ly6C+ populations and then with CD11b magnetic beads. CD45+ cells counterstaining with hematoxilin, dehydration and mounting (DPX). The quan- isolation was performed with anti-mouse CD45 magnetic beads. From ID8 cell tification of LIF, CCL2, CD163, CD206, and CXCL9 staining in GBM tumors from suspension, CD11b+ cells were isolated using anti-CD11b magnetic beads. Finally, patients was expressed as H-score (3× percentage of strong staining + 2× per- from organotypic slices, CD45+ cells isolation was performed using anti-human centage of moderate staining + percentage of weak staining), giving a range of CD45 magnetic beads. All the isolation procedures were performed using the 0–300. Quantification of p-STAT3, Ki67, CC3, and CD8 was performed with MultiMACS Cell24 separator Plus following manufacturer instructions and ImageJ, counting the total number of cells of three different fields per mouse, five magnetic beads were purchased from Miltenyi Biotec. mice/group, and calculating the percentage of positive cells. Data in graphs are The murine antibodies against CD3 (17-0032-82; 1:40), CD4 (11-0042-81; presented as mean ± SEM. 1:250), CD335 (25-3351-82; 1:20), CD163 (12-1631-82; 1:40, MHC class II (47- IHC antibodies: human LIF (Atlas; HPA018844; 1:200), murine LIF (Atlas; 5321-82; 1:80), CXCR3 (12-1831-82; 1:40) from eBioscience, CD45 (BD-550994; HPA018844; 1:300), murine p-STAT3 (Cell Signaling; 9131; 1:50), murine Ki67 1:200), CD8 (BD-557654; 1:40), F4/80 (BD-565410; 1:25), CD11b (BD-553310; (AbCam; ab15580; 1:200), murine Cleaved-Caspase3 (CC3) (Cell Signaling; 9661; 1:100), CD11c (BD-550261; 1:100), CD206 (BD-565250; 1:30), CD49d (BD- 1:500), murine CD8 (Bioss; bs-0648R; 1:200), human/murine CCL2 (Novus 740012; 1:30) from BD Bioscience, LIFR (Novus FAB5990N; 1:20) from Novus Biologicals;NBP2-22115; 1:200), human CXCL9 (Thermo Fischer Scientific; Biologicals, Ly6G (127622; 1:200), Ly6C (128018; 1:150), CCR2 (150604; 1:50) and 701117; 1:100), human CD206 (Abcam; ab64693; 1:4000) and human CD163 PD1 (135208; 1:80) from Biolegend were used for flow cytometry. Intracellular (Leica Novacastra; NCL-L-CD163; 1:200). staining of FoxP3 (12-5773-82; 1:40), Granzyme A (GZMA) (48-5831-82; 1:80), Immunofluorescence (IF) staining was performed as follows. Nuclei were CXCL9 (12-3009-80; 1:40) from eBioscience and CCL2 (505910; 1:100; Biolegend) counterstained with DAPI and images were captured using a laser scanning were performed using a specific staining set (eBioscience). For flow human studies, confocal NIKON Eclipse Ti microscope. Quantification of immunofluorescence antibodies against CD11b (333142; 1:100), CD14 (555397; 1:100) from BD were performed with ImageJ, counting all or up to 100 cells positive for CD11b, Bioscience, CD45 (304032; 1:100), CD3 (344816; 1:100), CXCR3 (353708; 1:100) Iba1, or CD3 of 2–3 different fields of each mouse, 3–5 mice/group, and calculating from Biolegend and CD8 (560179; 1:75; BD Pharmigen) were used. In some the percentage of those cells positive for CCL2, CD206, and CD163 inside the Iba1 occasions samples were previously incubated with LIVE/DEAD fixable yellow dead positive population. For CXCL9, it was calculated the percentage of cells stain kit (L34959; 1:2000;Thermo Fisher Scientific) to determine viability. As a surrounded by the signal of this cytokine inside the total population of cells. For positive control of human GBM organotypic and patient derived xenografts, organotypic slices, 3–4 fields of each patient (n = 3) were quantified. For 1 × 106 PBMCs (named as “spike PBMCs”) was added into the control samples in organotypic tissue immunofluorescence, five different Z-stack images per condition order to establish the leukocyte populations. were processed with Fiji-Image J software. For CD8+ T cells, percentage of CD8+ Samples were acquired on a BD LSRFortessa™ cell analyser or Navios (Beckman T cells was calculated among the total population. Data in graphs are represented Coulter) and data were analysed with Flow Jo software. as mean ± SEM. IF antibodies: human/murine CCL2 (Novus Biologicals; NBP-22115; 1:200), human/murine Iba1 (Wako; 019-19741; 1:1000 and Abcam; ab5076; 1:50), human/ Lentiviral infection. 293T cells were transfected using lipofectamine 2000 (Invi- murine CD206 (Abcam; ab64693; 1:500), murine CD163 (Abcam; ab199402; trogen) transfection method with pMD2.G enveloping plasmid, psPAX.2 packa- 1:200), CXCL9 (murine R&D; AF-492-NA; 1:200; human Thermo Fischer ging plasmid, and either pLenti CMV Luciferase (v-168-1) (Addgene), pLKO.1 or Scientific; 701117; 1:200), human CD8 (DAKO; M7103; 1:200), human LIF (Atlas; pLKO.1-shLIF murine (Sigma). After 16 h, medium was replaced by fresh medium HPA018844; 1:200) and human CD163 (Leica Novocastra; NCL-L-CD163; 1:200). and lentivirus was harvested for an additional 24 h. For cell lines infection, medium containing recombinant lentivirus was added to cells with polybrene (Sigma) at a Microarray expression analysis. RNA was assayed on the Affymetrix microarray concentration of 8 µg/ml. Following 16 h of incubation, cells were washed and platform with the Mouse Gene 2.1 ST. Next, it was normalized based on a Robust- incubated in fresh medium previous to selection. Microarray Average (RMA)34 (by Progenika). We identified the genes differentially CD3+ T cells were negatively isolated from PBMCs using Dynabeads expressed in anti-LIF treated mice through a Bayesian linear regression considering Untouched Human T cell Kit (Thermo Fisher). Isolated CD3+ T cells were infected paired samples, using limma Bioconductor package35. with lentivirus vectors33. Briefly, a 24 well plate was pre-coated for 1 h at 37 °C with 1 µg/ml of anti-CD3 (clone OKT3, Thermo Fisher Scientific) and 5 µg/ml of CH- 296 (Retronectin recombinant fibronectin fragment, Takara). CD3+ T cells were Chromatin immunoprecipitation assays. Immunoprecipitation of chromatin was resuspended at concentration of 0.7 × 106 cells/ml with RPMI supplemented with performed according to the Upstate (Millipore) standard protocol. Briefly, 1.2 × 107 10% heat-inactivated FBS and 600 IU/ml of rIL2 (R&D systems). 0.4 ml of cell BMDMs were fixed using 1% formaldehyde for 10 min at 37 °C, harvested, and suspension was mixed with 1 ml of luciferase lentiviral vector (supplemented with sonicated to generate chromatin fragments of 200–500 bp. Then 20 µg of sheared 600 IU/ml of rIL2 and 8 µg/ml of polybrene) and seeded into anti-CD3/CH-296 chromatin was immunoprecipitated overnight with 2 µg of anti-Tri-Methyl- coated 24-well plate. Plate was centrifuged at 1000×g for 30 min at 32 °C. The day Histone H3 (Lys27) (Cell Signaling; 97335), anti-acetyl-H4 (Millipore; 06–866) or after, medium was replaced by fresh RPMI supplemented with 10% heat- anti-EZH2 (Millipore; CS203195) antibody. Immunocomplexes were recovered inactivated FBS and 600 IU/ml of rIL2. After 4 days cells were subcultured 1:2 into using 20 µl of protein G magnetic beads, washed, and eluted. Cross-linking was new anti-CD3/CH-296 pre-coated 24-well plate until used. At same time point, reversed at 65 °C 4 h and immunoprecipitated DNA was recovered using the PCR luciferase was determined by the luciferase assay system (Promega). purification kit from Qiagen. Genomic regions of interest were identified by real- time quantitative PCR (qPCR) using SYBR Green Master Mix (Invitrogen). Each value was corrected by the corresponding input chromatin sample. CXCL9 pro- CRISPR/LIF cell line generation. Cells were transfected with the commercial moter was amplified using “CCCCGTTGCAATACTTTCAT” (forward) and mouse LIF CRISPR/Cas9 KO plasmid (Santa Cruz Biotechnology) following “CCCCGTTGCAATACTTTCAT” (reverse) primers (EZH2, acetyl-H4) or manufacturer’s specifications. Green Fluorescent Protein (GFP) was used as a “TGCTGTTGAATGCCACTTTC” (forward) and “TCCCTGCTACCTTTTC- selective label. After 3 days, transfected cells were selected by sorting. CRISPR/LIF- CAGA” (reverse) primers (H3me3). Mouse GAPDH promoter (Diagenode) was KO positive cells were checked by qRT-PCR and ELISA. amplified as a control. ELISA. For the quantitative determination of CXCL9 and CCL2 protein levels TCGA data sets analysis. We downloaded RNA-seq data from The Cancer secreted to the media, and IL10 and IL12 protein levels from tumor tissue, we used Genome Atlas (TCGA) for 9,403 patients suffering from 28 distinct solid tumors the Mouse or Human Duo-Set ELISA kit (R&D systems) following manufacturer’s (see details on the tumor cohorts considered in Supplementary Data 1b) from specifications. Firebrowse server (http://firebrowse.org, version 2016_01_28). The RNA-seq NATURE COMMUNICATIONS | (2019)10:2416 | https://doi.org/10.1038/s41467-019-10369-9 | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-10369-9 normalized data (RSEM) was log2 transformed for all downstream analyses. Next, 11. Yu, H. et al. LIF negatively regulates tumour-suppressor p53 through Stat3/ we obtained the gene signatures of TAMs from 37 (see Supplementary Data 1a). ID1/MDM2 in colorectal cancers. Nat. Commun. 5, 5218 (2014). Based on previous work36–38, we inferred the relative abundance of the immune 12. Ye, F., Hu, Y., Lu, W., Zhou, C. & Xie, X. Expression of leukaemia inhibitory populations through a single sample gene set enrichment (ssGSEA)39 across all factor in epithelial ovarian carcinoma: correlation with clinical characteristics. patients. Thereafter we computed the correlation between LIF expression and the Histopathology 53, 224–228 (2008). ssGSEA scores of the four immune populations and the correlation between LIF 13. Ozawa, T. et al. Most human non-GCIMP glioblastoma subtypes evolve and a set of genes of interest (see Statistical tests). from a common proneural-like precursor glioma. Cancer Cell 26, 288–300 (2014). Transcriptome analysis of metastatic urothelial carcinoma anti-PD1 treated 14. Bowman, R. L. et al. Macrophage ontogeny underlies differences in tumor- patients analysis. We downloaded RNA-seq from the phase 2 clinical trial specific education in brain malignancies. Cell Rep. 17, 2445–2459 (2016). (IMvigor210) which investigates the clinical activity of PD-L1 blockade with ate- 15. Chang, A. L. et al. CCL2 produced by the glioma microenvironment is zolizumab in metastatic urothelial carcinoma28 We downloaded RNA-seq data essential for the recruitment of regulatory T cells and myeloid-derived (FASTQ files) and clinical annotations from the European Genome-Pheno Archive suppressor cells. Cancer Res. 76, 5671–5682 (2016). (EGAS00001002556). We normalized RNA-seq counts into Transcripts per million 16. Curiel, T. J. et al. Specific recruitment of regulatory T cells in ovarian (TPM) for Homo Sapiens GRCh38 using Kallisto40. Two groups of patients, as carcinoma fosters immune privilege and predicts reduced survival. Nat. Med. previously defined by the authors, were considered for downstream analyses based 10, 942–949 (2004). on their clinical response to atezolizumab: non-responders (patients with Pro- 17. Nagarsheth, N., Wicha, M. S. & Zou, W. Chemokines in the cancer gressive Disease and Sustained Disease after treatment) and responders (patients microenvironment and their relevance in cancer immunotherapy. Nat. Rev. with Partial and Complete response after treatment). To compare the expression Immunol. 17, 559–572 (2017). levels of LIF between both groups TPMs were log2 transformed. Next, we per- 18. Peng, D. et al. Epigenetic silencing of TH1-type chemokines shapes tumour formed an enrichment analysis of the anti-LIF down-regulated genes in anti-PD1 immunity and immunotherapy. Nature 527, 249–253 (2015). non-responder patients. We considered as anti-LIF down-regulated genes those 19. Tang, Y. et al. Jak/Stat3 signaling promotes somatic cell reprogramming by genes with a log fold change (FC) <−1.5 in anti-LIF treated ID8 mice (see epigenetic regulation. Stem Cells 30, 2645–2656 (2012). Microarray expression analysis above). Then, we performed a pre-ranked Gene Set 20. Merz, F. et al. Organotypic slice cultures of human glioblastoma reveal Enrichment Analysis41, using gseapy Python3 package, on the genes differentially different susceptibilities to treatments. Neuro-Oncology 15, 670–681 expressed between non-responder and responder patients (we considered the log (2013). FC as rank). Ranks were computed using rankdata Python3 SciPy package16. 21. Youngblood, B. et al. Chronic virus infection enforces demethylation of the locus that encodes PD-1 in antigen-specific CD8(+) T cells. Immunity 35, Statistical analysis. Data were analysed using GraphPad Prism 5.0 software. To 400–412 (2011). compare two different groups we calculated P-value (P) using Student’s T-test 22. Gros, A. et al. PD-1 identifies the patient-specific CD8(+) tumor-reactive (paired or unpaired) for parametric variables and Mann–Whitney test for non- repertoire infiltrating human tumors. J. Clin. Investig. 124, 2246–2259 parametric variables. Survival curves comparison was performed using Log-rank (2014). (Mantel-Cox) test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. TCGA data 23. Gajewski, T. F., Schreiber, H. & Fu, Y. X. Innate and adaptive immune cells in was analysed with Python3 SciPy library16, the correlation coefficient (R2) was the tumor microenvironment. Nat. Immunol. 14, 1014–1022 (2013). obtained using a Pearson’s correlation. P-values were adjusted for multiple testing 24. Sharma, P., Hu-Lieskovan, S., Wargo, J. A. & Ribas, A. Primary, adaptive, and correction based on a Benjamini-Hochberg False Discovery Rate. acquired resistance to cancer immunotherapy. Cell 168, 707–723 (2017). 25. Joyce, J. A. & Fearon, D. T. T cell exclusion, immune privilege, and the tumor Reporting summary. Further information on research design is available in microenvironment. Science 348, 74–80 (2015). the Nature Research Reporting Summary linked to this article. 26. Quail, D. F. & Joyce, J. A. Microenvironmental regulation of tumor progression and metastasis. Nat. Med. 19, 1423–1437 (2013). 27. Quail, D. F. & Joyce, J. A. The microenvironmental landscape of brain tumors. Data availability Cancer Cell 31, 326–341 (2017). The microarray data of the mice ovary tumor model has been deposited in the Gene 28. Mariathasan, S. et al. TGFbeta attenuates tumour response to PD-L1 blockade Expression Omnibus (GEO)/NCBI public database (accession no. GSE79852). by contributing to exclusion of T cells. Nature 554, 544–548 (2018). 29. Metcalfe, S. M., Watson, T. J., Shurey, S., Adams, E. & Green, C. J. Leukemia Received: 28 March 2019 Accepted: 8 May 2019 inhibitory factor is linked to regulatory transplantation tolerance. Transplantation 79, 726–730 (2005). 30. Janssens, K. et al. Leukemia inhibitory factor tips the immune balance towards regulatory T cells in multiple sclerosis. Brain Behav. Immun. 45, 180–188 (2015). 31. Edwards, L. A. et al. ZEB1 regulates glioma stemness through LIF repression. References Sci. Rep. 7, 69 (2017). 1. Mathieu, M. E. et al. LIF-dependent signaling: new pieces in the Lego. Stem 32. Valledor, A. F., Comalada, M., Xaus, J. & Celada, A. The differential time- Cell Rev. 8, 1–15 (2012). course of extracellular-regulated kinase activity correlates with the 2. Trouillas, M. et al. The LIF cytokine: towards adulthood. Eur. Cytokine Netw. macrophage response toward proliferation or activation. J. Biol. Chem. 275, 20, 51–62 (2009). 7403–7409 (2000). 3. Rathjen, P. D. et al. Developmentally programmed induction of differentiation 33. Chono, H. et al. Optimization of lentiviral vector transduction into peripheral inhibiting activity and the control of stem cell populations. Genes Dev. 4, blood mononuclear cells in combination with the fibronectin fragment CH- 2308–2318 (1990). 296 stimulation. J. Biochem. 149, 285–292 (2011). 4. Stewart, C. L. et al. Blastocyst implantation depends on maternal expression of 34. Irizarry, R. A. et al. Summaries of Affymetrix GeneChip probe level data. leukaemia inhibitory factor. Nature 359, 76–79 (1992). Nucleic Acids Res. 31, e15 (2003). 5. Kralickova, M. et al. Leukemia inhibitory factor gene mutations in the 35. Ritchie, M. E. et al. limma powers differential expression analyses for RNA- population of infertile women are not restricted to nulligravid patients. Eur. J. sequencing and microarray studies. Nucleic Acids Res. 43, e47 (2015). Obstet. Gynecol. Reprod. Biol. 127, 231–235 (2006). 36. Charoentong, P. et al. Pan-cancer immunogenomic analyses reveal genotype- 6. Schofield, G. & Kimber, S. J. Leukocyte subpopulations in the uteri of leukemia immunophenotype relationships and predictors of response to checkpoint inhibitory factor knockout mice during early pregnancy. Biol. Reprod. 72, blockade. Cell Rep. 18, 248–262 (2017). 872–878 (2005). 37. Angelova, M. et al. Characterization of the immunophenotypes and 7. Novotny, Z. et al. Leukaemia inhibitory factor (LIF) gene mutations in women antigenomes of colorectal cancers reveals distinct tumor escape mechanisms diagnosed with unexplained infertility and endometriosis have a negative and novel targets for immunotherapy. Genome Biol. 16, 64 (2015). impact on the IVF outcome. A pilot study. Folia Biol. 55, 92–97 (2009). 38. Bindea, G. et al. Spatiotemporal dynamics of intratumoral immune cells reveal 8. Penuelas, S. et al. TGF-beta increases glioma-initiating cell self-renewal the immune landscape in human cancer. Immunity 39, 782–795 (2013). through the induction of LIF in human glioblastoma. Cancer Cell 15, 315–327 39. Barbie, D. A. et al. Systematic RNA interference reveals that oncogenic KRAS- (2009). driven cancers require TBK1. Nature 462, 108–112 (2009). 9. Albrengues, J. et al. LIF mediates proinvasive activation of stromal fibroblasts 40. Bray, N. L., Pimentel, H., Melsted, P. & Pachter, L. Near-optimal probabilistic in cancer. Cell Rep. 7, 1664–1678 (2014). RNA-seq quantification. Nat. Biotechnol. 34, 525–527 (2016). 10. Liu, S. C. et al. Leukemia inhibitory factor promotes nasopharyngeal 41. Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based carcinoma progression and radioresistance. J. Clin. Investig. 123, 5269–5283 approach for interpreting genome-wide expression profiles. Proc. Natl Acad. (2013). Sci. USA 102, 15545–15550 (2005). 10 NATURE COMMUNICATIONS | (2019)10:2416 | https://doi.org/10.1038/s41467-019-10369-9 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-10369-9 ARTICLE Acknowledgements from Novartis. E.G. reports scientific consultancy role for Roche/Genentech, Hoffmann We want to thank the Medical Oncology department, the Neurosurgery department, and La Roche, Ellipses Pharma, Neomed Therapeutics, Boehringer Ingelheim, Janssen the Pathology department of the Vall d’Hebron Hospital and Clínic Hospital. We also Globlas Services, AstraZeneca. I.B. reports scientific consultancy role for Orion Pharma. want to thank Cristina Sánchez for technical support. This work has been funded by the P.N. reports scientific consultancy role for Bayer, Novartis and Merck Sharp and Dohme. European Research Council grant (ERC 205819), the Asociación Española contra el The remaining authors declare no competing interests. Cáncer (AECC), FIS - Instituto de Salud Carlos III grant (PI16/01278), MINECO- RETOS RTC-2015-3771-1, the FERO, laCaixa, BBVA (CAIMI) and Cellex foundations. Reprints and permission information is available online at http://npg.nature.com/ reprintsandpermissions/ Author contributions Journal peer review information: Nature Communications thanks the anonymous J.S. designed the study. M.P.G., E.B.T., E.P.R., C.R.P., R.I., A.A., I.C., A.S.H., L.E., F.M.R., reviewers for their contribution to the peer review of this work. I.H.R, P.N., L.P., C.M., M.S., C.E. generated data and developed methods. I.B., E.G., M.V., E.P., F.G., J.S., J.T. contributed through clinical aspects and intellectual inputs. J.S. Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in interpreted data and wrote the manuscript with assistance of other authors. All authors published maps and institutional affiliations. approved the final manuscript. Additional information Open Access This article is licensed under a Creative Commons Supplementary Information accompanies this paper at https://doi.org/10.1038/s41467- Attribution 4.0 International License, which permits use, sharing, 019-10369-9. adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Competing interests: J.S. is a founder of Mosaic Biomedicals and share holder Commons license, and indicate if changes were made. The images or other third party of Northern Biologics. J.S. received grant/research support from Mosaic material in this article are included in the article’s Creative Commons license, unless Biomedicals, Northern Biologics and Roche/Glycart. J.T. reports scientific consultancy indicated otherwise in a credit line to the material. If material is not included in the role for Array Biopharma, AstraZeneca, Bayer, BeiGene, Boehringer Ingelheim, Chugai, article’s Creative Commons license and your intended use is not permitted by statutory Genentech Inc., Genmab A/S, Halozyme, Imugene Limited, Inflection Biosciences regulation or exceeds the permitted use, you will need to obtain permission directly from Limited, Ipsen, Kura Oncology, Lilly, MSD, Menarini, Merck Serono, Merrimack, Merus, the copyright holder. To view a copy of this license, visit http://creativecommons.org/ Molecular Partners, Novartis, Peptomyc, Pfizer, Pharmacyclics, ProteoDesign SL, Rafael licenses/by/4.0/. Pharmaceuticals, F. Hoffmann–La Roche Ltd., Sanofi, SeaGen, Seattle Genetics, Servier, Symphogen, Taiho, VCN Biosciences, Biocartis, Foundation Medicine, HalioDX SAS and Roche Diagnostics. E.P. reports scientific consultancy role for Celgene. E.P. © The Author(s) 2019 reports scientific consultancy role for Celgene. E.G. received grant/research support NATURE COMMUNICATIONS | (2019)10:2416 | https://doi.org/10.1038/s41467-019-10369-9 | www.nature.com/naturecommunications 11