Tumors induce de novo steroid biosynthesis in T cells to evade immunity

Tumors subvert immune cell function to evade immune responses, yet the complex mechanisms driving immune evasion remain poorly understood. Here we show that tumors induce de novo steroidogenesis in T lymphocytes to evade anti-tumor immunity. Using a transgenic steroidogenesis-reporter mouse line we identify and characterize de novo steroidogenic immune cells, defining the global gene expression identity of these steroid-producing immune cells and gene regulatory networks by using single-cell transcriptomics. Genetic ablation of T cell steroidogenesis restricts primary tumor growth and metastatic dissemination in mouse models. Steroidogenic T cells dysregulate anti-tumor immunity, and inhibition of the steroidogenesis pathway is sufficient to restore anti-tumor immunity. This study demonstrates T cell de novo steroidogenesis as a mechanism of anti-tumor immunosuppression and a potential druggable target.

S teroidogenesis is a metabolic process by which cholesterol is converted to steroids 1 . The biosynthesis of steroids starting from cholesterol is often termed de novo steroidogenesis 1 . Cytoplasmic cholesterol is transported into the mitochondria, where the rate-limiting enzyme CYP11A1 (also known as P450 side chain cleavage enzyme) converts cholesterol to pregnenolone. Pregnenolone is the first bioactive steroid of the pathway, and the precursor of all other steroids (Fig. 1a) 1,2 . The steroidogenesis pathway has been extensively studied in adrenal gland, gonads, and placenta. De novo steroidogenesis by other tissues, known as extra-glandular steroidogenesis, in brain 1,3 , skin 4 , thymus 5 , and adipose tissues 6 has also been reported. Steroid production as a result of immune response in the mucosal tissues, such as in the lung and intestine, has been shown to play a tolerogenic role to maintain tissue homeostasis 7,8 . However the physiological and pathological role of extra-glandular steroidogenesis remains largely unknown 2 .
Steroid hormones are known immunosuppressive biomolecules 26,27 . We recently reported that type 2 CD4 + T cells (Th2 lymphocytes) induce de novo steroidogenesis to restore immune homeostasis by limiting the immune response against a worm parasite 28 . Thus we sought to determine whether type 2 T cell-mediated steroidogenesis contributes to the generation of a suppressive niche in the TME.
To study de novo steroidogenesis we generated two transgenic mouse lines: a fluorescent reporter mouse line (Cyp11a1-mCherry) and a conditional Cyp11a1 (floxed) knockout mouse line. We show the presence of de novo steroidogenesis by tumorinfiltrating T lymphocytes, but not in unchallenged animals or draining lymph nodes. Genetic ablation of Cyp11a1 in T cells restricts experimental primary tumor growth and lung metastasis. Mechanistically, we find that intratumoral T cell steroidogenesis dysregulates anti-tumor immunity that could be restored by inhibiting the steroidogenesis pathway pharmacologically. This study therefore demonstrates that T cell de novo setroidogenesis is a cause of anti-tumor immunosuppression and a potential drug target for cancer immunotherapy.

Results
Generation of Cyp11a1 reporter and conditional knockout mice. Cyp11a1 is the first and rate-limiting enzyme during steroid production. The expression of Cyp11a1 is therefore also a faithful biomarker of de novo steroidogenesis 1 . Therefore, we generated a reporter mouse line to identify Cyp11a1-expressing steroidogenic cells definitively (Fig. 1b, c, Supplementary Fig. 1a-d). As expected, mCherry expression was detected in single-cell suspensions of testis and adrenal glands but negligible to no expression in the spleen (Fig. 1c) or other tissues including lung, kidney, blood, liver, bone marrow, lymph nodes, and thymus ( Supplementary Fig. 1b). However, Cyp11a1-mCherry signal was detected specifically in activated type-2 CD4 + T helper cells (Th2 cells) upon activation in vitro ( Supplementary Fig. 1c), as reported previously 28 . Cyp11a1 expression was detectable only in mCherry-expressing T helper cells ( Supplementary Fig. 1d).
To determine the functional consequences of cell-type-specific steroidogenesis we created a Cyp11a1 floxed (Cyp11a1 fl/fl ) mouse following EUCOMM/WSI conditional gene targeting strategy 29 . Briefly, a knockout-first (tm1a) mouse line was created using a promoter-driven targeting cassette (Fig. 1d). The tm1a mouse was then crossed with Flp-deleter mice (FlpO) to remove the LacZ and Neo cassette, and generate a tm1c allele (i.e. Cyp11a1 fl/fl ). When crossed with a Cre-driver, the Crerecombinase removes exon 3 of Cyp11a1 gene and creates a frameshift mutation (Fig. 1d). Because we had initially detected Cyp11a1 expression in Th2 cells 28 , we crossed the Cyp11a1 fl/fl line with a Cd4-driven Crerecombinase to delete Cyp11a1 and prevent de novo steroidogenesis in all T cells (Fig. 1e). Deletion efficiency of Crerecombinase in the Cyp11a1 cKO (Cd4-Cre;Cyp11a1 fl/fl ) mice was nearly complete in Th2 cells (Fig. 1f). Cyp11a1 cKO mice showed normal thymic development of T cells, and a normal distribution in the peripheral tissues ( Fig. 1g-i).
In vitro analysis of Cyp11a1 expression in T cells. Exploiting our Cyp11a1-mCherry reporter line, we assayed a panel of cytokines commonly found in inflammatory settings, including tumors, for their ability to induce steroidogenesis in CD4 + T cells. IL4, IL6, IL13, and TSLP induced a strong Cyp11a1-mCherry signal in in vitro activated CD4 + T cells (Fig. 2a, Supplementary Fig. 2a, b). In contrast, IL12 had an inhibitory effect on Cyp11a1-mCherry expression (Fig. 2b, Supplementary  Fig. 2c). This indicated that not only Th2 lymphocytes, but also other T cell types, are capable of de novo steroidogenesis, at least in vitro. To test this, we differentiated naïve CD4 + or CD8 + T cells into Th1, Th2, Th9, Th17, Tfh, Treg, Tc1, and Tc2 subsets in vitro. All subsets examined, with the exception of Th1 and Tc1 cells, exhibited Cyp11a1-mCherry expression when activated, consistent with an inhibitory role for IL12 in Cyp11a1 regulation (Fig. 2c, Supplementary Fig. 2d). These data suggest that Cyp11a1 expression, and thereby steroidogenesis, is a default process during T cell activation, and is inhibited by the presence of IL12. Overall, of all T cell subtypes generated in vitro, Th2 cells exhibited the highest percentage of Cyp11a1 expression and Th1 cells do not express Cyp11a1.
To determine the requirement of Cyp11a1 activity for T helper cell proliferation and differentiation, we purified naïve splenic T cells from Cyp11a1 cKO and control mice. We activated the cells in vitro to generate different subclasses of T helper cells, and analyzed signature cytokine expression by flow cytometry. In the absence of Cyp11a1, T cells proliferate normally (Fig. 2d). Cyp11a1 expression was not required for the differentiation of any T helper cell type tested as determined by signature cytokine expression (Fig. 2e, f, Supplementary Fig. 2e). We observed that deletion of Cyp11a1 in T cells does not interfere with the plasticity of T helper cells (Fig. 2g, Supplementary Fig. 2f, g).
As a next step, Cyp11a1 induction in T cells was investigated in vivo.

Tumors induce functional Cyp11a1 expression in T cells.
Tumor-infiltrating T cells are key fate determinants within a tumor, but are often suppressed 30 . The steroidogenesis-inducing type-2 cytokines such as IL4 are also often present in the TME 31,32 , thus we next sought to examine the steroidogenic capacity (i.e. Cyp11a1 induction) of T cells infiltrating tumors, and their impact on tumor development. To explore Cyp11a1 expression in vivo, we utilized the well-established B16-F10 melanoma model [33][34][35] and implanted tumors subcutaneously in Cyp11a1-mCherry reporter mice.
Cyp11a1 expression was detected in immune cells of established primary tumors, but not in tumor-draining brachial lymph nodes (LN) or blood (Fig. 3a), indicating that stimulation occurs in situ. In support of this, stimulation of splenic T cells did not induce Cyp11a1 expression ex vivo (Fig. 3b, Supplementary  Fig. 3a). The dominant Cyp11a1 + tumor-infiltrating immune cells were identified as T cells, predominantly CD4 + (helper T cells, Fig. 3a).
To determine whether Cyp11a1 induction in T cells is conserved in other tumor types we analyzed an EO771 orthotopic model of breast cancer [36][37][38] (Fig. 3e-g, Supplementary Fig. 3d). Again, we found that Cyp11a1 was upregulated in tumor-infiltrating T cells ( Fig. 3e-g), but not in tumor-draining lymph node or spleen. We also detected minor and rare populations of other Cyp11a1 + type-2 immune cells, such as mast cells and basophils ( Supplementary Fig. 3d).
Next, we sought to measure the functional output of Cyp11a1 expression. Significant concentrations of the steroid pregnenolone were detected exclusively in immune cells isolated from tumors, with negligible levels detected in cells from the spleen (Fig. 3h). Using the B16-F10 model of experimental metastatic dissemination 39 , we determined that lungs with metastatic nodules, but not control lungs without metastatic nodules, had elevated levels of pregnenolone (Fig. 3i). The presence of tumor cells did not induce or augment the Cyp11a1 expression in T cells in vitro ( Supplementary Fig. 3e). This indicates that type-2 (Th2) cytokines within the TME induce Cyp11a1 expression by T cells and this cannot readily be mimicked in vitro.
Having observed steroidogenic T cells in murine melanoma, we turned to publicly available transcriptomic data sets to verify our findings and ascertain relevance in the human setting. We identified CYP11A1 mRNA expression, and thus steroidogenic potential, in a range of cancer types including liver, breast, prostate, lung, kidney, sarcoma, glioma, uterine, cervical, lymphoma, and melanoma ( Supplementary Fig. 3f, g) 32 . Human melanoma tissues represented a prominent steroidogenic tumor type, expressing CYP11A1, HSD3B1, HSD3B2, CYP17A1, CYP21A1, and CYP11B1 and not expressing CYP11B2 (Fig. 3j, k). CYP11B2 catalyzes the aldosterone synthesis from corticosteroid precursor (Fig. 3k). Together this was indicative of melanoma-driven production of glucocorticoids (Fig. 3k, Supplementary Fig. 3h).
Interestingly, in melanoma, steroidogenic gene expression was correlated with IL4 expression (Fig. 3k, Supplementary Fig. 3j), a key inducer of T cell steroidogenesis 28 . Moreover, analysis of human tumor-infiltrating CD4 + T cell transcriptomes, confirmed CYP11A1 expression (Fig. 3l) implying that CD4 + T cells are a source of steroids in tumors, mirroring the murine setting. Collectively these data indicate that in both human and mice, TILs produce steroids within the tumor.
Gene expression identity of Cyp11a1 + immunocytes in melanoma. To reveal the gene expression identity of intratumoral steroidogenic immune cells and patterns of gene expression at single-cell resolution, we inoculated B16-F10 subcutaneous tumor in Cyp11a1-mCherry reporter mice, enriched and purified intratumoral Cyp11a1-mCherry + and Cyp11a1-mCherry − cells by cell sorting into 96-well plates [with a ratio of 79:15 (mCherry + : mCherry − ) cells per plate] and performed single-cell RNA sequencing (scRNA-seq) using the SMART-Seq2 platform 40 . The majority of cells,~3000, passed quality control. Unsupervised hierarchical clustering and visualization with reduced dimensionality using UMAP (uniform manifold approximation and projection) 41 identified 12 clusters of cells ( Supplementary Fig. 4a). There was no batch effect between samples ( Supplementary Fig. 4b) or impact of sex ( Supplementary  Fig. 4c) on clustering, and Cyp11a1-mCherry protein expression accurately reported Cyp11a1 mRNA expression ( Supplementary  Fig. 4d, e). We annotated the clusters based on classical lineage markers ( Supplementary Fig. 4a, f) and their signature gene expression ( Supplementary Fig. 4g). The majority of the Cyp11a1-expressing cells are T cells as suggested by Cd3e expression (Supplementary Fig. 4f). T cells constitute the clusters 0, 1, 2, 3, and 8 with cluster-1 displaying genes of CD8 + Tc2 cells; Cd8a, Cd3e, Gata3, and type 2 cytokines Il4, Il5, and Il13. Clusters 0, 2, and 3 were CD4 + T helper cells (Cd4 + Cd3e + ) as they expressed Gata3, Il4, and Il13 ( Supplementary Fig. 4f). As T cells were the dominant Cyp11a1-expressing population, we reanalyzed the data taking only the T cells and displayed in Fig. 4a-e. The three clusters of Th2 cells represent three closely related subpopulations (Fig. 4a, b, Supplementary Fig. 4a). The comparative gene expression between these clusters is shown in Supplementary Fig. 4i. In Supplementary Fig. 4a, Cluster-4 represents myeloid derived cells, macrophages, and dendritic cells, but mostly they constitute Cyp11a1-mCherry − cells, though a very small number of Cyp11a1-mCherry + cells have been found in this cluster. Cluster-5 contains B16-F10 cells and melanocytes from the host mice, and they were Cyp11a1 − . Cluster-6 and 9 constitute two different states of the Cyp11a1 + mast cells, because they express Kit (the gene that encodes cKit) and Fcer1 and other mast-cell-specific proteases such as Mcpt4 and Cma1. The detailed differences between these two clusters have been shown in Supplementary Fig. 4j. Cyp11a1 + basophils and eosinophils (cluster-7) fall in the same cluster. Cluster-8 is a mixed population of CD4 + and CD8 + T cells. The number of cells in these two clusters are not sufficient to make a conclusion of their identity. Clusters 4, 5, 8, 11, and 10 are constituted by both Cyp11a1-mCherry + and Cyp11a1-mCherry − cells. Cyp11a1-mCherry − cells representing the cells that are most frequently available cells within B16-F10 tumor, namely B16-F10 cells and macrophages.
Taking Cyp11a1-correlated genes we ran pySCENIC to identify potential transcription factors regulating Cyp11a1 expression and found that the GATA family of transcription factors (Gata1, 2, 3) were predicted to be the controller of the Cyp11a1 locus in these steroidogenic immune cells ( Supplementary Fig. 4h). Gata3 is the lineage-defining TF of Th2 cells. To validate the prediction that Gata3 controls Cyp11a1 expression in T helper cells, first we performed ATAC-seq to identify the open chromatin and potential locus control region (Fig. 4f). Next, we analyzed Gata3 ChIP-Seq data to confirm that Gata3 is indeed a potential regulator of the Cyp11a1 locus in Th2 cells (Fig. 4f). The putative locus control region (blue highlighted region in Fig. 4f) was found in a closed-chromatin state in naïve and Th1 cells but was open in Fig. 2 Analysis of Cyp11a1-expressing T cells using Cyp11a1 reporter and cKO mice. a Splenic naïve CD4 + T cells from Cyp11a1-mCherry reporter mice were purified by negative selection; activated in the anti-CD3e/anti-CD28 antibody coated plates in the presence of cytokines for 3 days, rested for 2 days, restimulated 6 h, and Cyp11a1-mCherry expression was analyzed by flow cytometry. N = 3 (except IL33 where N = 2). b IL12 inhibits Cyp11a1 expression. Splenic naïve CD4 + T cells from Cyp11a1-mCherry reporter mice were activated as mentioned above (a) for 5 days and Cyp11a1-mCherry expression was analyzed by flow cytometry. N = 6. c Splenic naïve CD4 + and CD8 + T cells from Cyp11a1-mCherry reporter mice were activated in vitro under Th1, Th2, Th9, Th17, Tfh, Treg, Tc1, and Tc2 differentiation conditions (activation 3 days, resting 2 days), and Cyp11a1-mCherry expression was analyzed by flow cytometry. N = 6 (Tc1, Tc2), N = 4 (Th1, Th9, Tfh), and N = 5 (Treg, Th17). d Splenic naïve CD4 + T cells were purified from Cyp11a1 cKO, Cd4-Cre, and Cyp11a1 fl/fl mice, stained with CellTrace Violet, activated in vitro, and cell proliferation was determined by a flow cytometric dye decay assay. Representative cell proliferation profiles are shown in the left panel and a comparison of the cell division index is shown in right panel. N = 9. e Splenic naïve CD4 + T cells were activated in vitro in the absence of any exogenous cytokine or cytokine-neutralizing antibody, and cytokine expression was determined by flow cytometry. N = 6. f Splenic naïve CD4 + T cells were activated in vitro under Th17 or Th2 differentiation conditions, and cell typespecific signature cytokine expression was determined by flow cytometry. N = 3 (Th17), N = 6 (Th2). g Splenic naïve CD4 + T cells were activated in vitro under Th1 or Th2 differentiation condition. After 3 days Th1 cells were allowed to differentiate under Th2-polarizing conditions (Th1 > Th2), and Th2 cells were allowed to differentiate under Th1-polarizing conditions (Th2 > Th1). Th1 or Th2-specific cytokine expression was determined by flow cytometry. N = 6. All error bars in this figure represent mean with s.d. P-value was calculated using unpaired two-tailed t-test. Representative of three independent experiments. N represents biologically independent animals. NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17339-6 ARTICLE NATURE COMMUNICATIONS | (2020) 11:3588 | https://doi.org/10.1038/s41467-020-17339-6 | www.nature.com/naturecommunications Th2 and Th17 cells and was occupied by Gata3 in Th2 cells (Fig. 4f).
The prediction of glucocorticoids in the tumor prompted us to check the cognate receptors in the tumor-infiltrating immune cells. We found that the glucocorticoid receptor Nr3c1 was expressed in all tumor-infiltrating immune cells ( Supplementary  Fig. 4k). T cells significantly restricted primary tumor growth rates and final volumes (Fig. 5a). Similarly, in the experimental metastasis model, impaired lung colonization was observed in the absence of T cell-expressed Cyp11a1. There was a significant reduction in number of metastatic foci in lungs of Cyp11a1 cKO mice compared to the control mice (Fig. 5b). In the B16-F10 subcutaneous melanoma model, topical application of exogenous pregnenolone at the primary tumor site was sufficient to compensate for the Cyp11a1 deficiency, restoring tumor growth to levels comparable with control mice (Fig. 5c). This suggests that the tumor restriction was a consequence of the absence of pregnenolone synthesis. Similar results were obtained using the EO771 breast cancer orthotopic tumor model: deletion of Cyp11a1 in T cells restricted dissemination to and nodule Fig. 3 Tumors induce Cyp11a1 expression in T cells in vivo. a B16-F10 cells were injected subcutaneously into the Cyp11a1-reporter mice. After 12 days brachial lymph node (LN), blood, and tumor tissues were analyzed by flow cytometry. Gating strategy: Singlets>Live cells>CD45,Cyp11a1-mCherry. N = 5. CD45 + Cyp11a1-mCherry + cells were further gated to show T helper cell (CD4 + CD3e + ) expression of Cyp11a1. b Splenic cells were purified from tumorbearing Cyp11a1-reporter mice, and restimulated in vitro using PMA/ionomycin and analyzed by flow cytometry. N = 9. c, d B16-F10 cells were injected subcutaneously into the Cyp11a1-mCherry reporter mice. After 5, 7, and 12 days tumor tissues were analyzed by flow cytometry to detect the CD4 + T cells (c), CD8 + T cells (d) N = 4. All cell>Singlets>Live cell>CD45 + >CD4 + CD3e + or CD8 + CD3e + . e-g EO771 cells were injected into the mammary fat pad of Cyp11a1-reporter mice. After 15 days, tumor tissues and tumor-draining LN were analyzed by flow cytometry to detect the Cyp11a1-mCherry expression in CD4 + T cells (e, f) CD8 + T cells (g). Gating: All cell>Singlets>Live cell>CD45 + >CD4 + TCRb + or CD8 + TCRb + . e Representative FACS profile of Cyp11a1 + CD4 + T cells. f, g Representative graphical presentation of one experiment showing Cyp11a1-expressing CD4 and CD8 T cells. N = 4. h B16-F10 tumorinfiltrating leukocytes (TIL) and splenocytes were purified from tumor-bearing mice on post-inoculation day-12, cultured for 48 h, and the supernatant was analyzed by ELISA to measure pregnenolone. N = 12, pooled analysis of three independent experiments. i Metastasized lungs were harvested 10 days post-B16-F10 intravenous injection in C57BL/6 mice, dissociated into single-cell suspension, cultured for 48 h, and the supernatant was analyzed by ELISA. Naïve uninfected lungs (normal lung) were used as control. N = 6, pooled analysis of two independent experiments. j Hierarchical clustering of steroidogenic genes and IL4 mRNA expression across 44 melanoma patient samples (GEO: GSE19234   formation in the lung (Fig. 5d). Importantly, in both tumor models, pharmacological inhibition of Cyp11a1 by aminoglutethimide (AG) was sufficient to recapitulate the tumor restriction phenotype resulting from Cyp11a1 genetic ablation (Fig. 5e, f). Together, these data indicate that T-cell-derived steroids can support tumor growth and both genetic and pharmacological interference with the pathway can restrict tumor growth.

Inhibition of T cell steroidogenesis boosts tumor immunity.
Steroid hormones can induce an immunosuppressive M2-like phenotype in macrophages 26,42,43 cell death and anergy in T cells 26,[44][45][46] , tolerance in dendritic cells 26,[47][48][49] and increase the frequency of Treg cells [50][51][52] . Therefore, we tested whether steroidogenic T cells support tumor growth through the induction of immunosuppressive phenotypes in infiltrating immune cells. To determine whether intratumoral macrophages were M1 or M2 type, we purified tumor-infiltrating macrophages and analyzed mRNA expression of the M2-macrophage signature genes Arg1 and Tgfβ1. In tumor-infiltrating macrophages from Cyp11a1 cKO mice, Arg1 and Tgfβ1 mRNA expression was significantly reduced compared to the control mice, indicating fewer of the tumorsupporting M2 macrophages in the T cell-specific Cyp11a1 knockout mice (Fig. 6a). We confirmed changes in macrophage characteristics at the protein level, showing a significant decrease in Arg1 + macrophages (M2) with coincident increase in tumorinfiltrating iNOS + macrophages (M1) upon T cell-specific Cyp11a1 deletion (Fig. 6b, c). Hence, depletion of Cyp11a1 in T cells increases the M1/M2 macrophage ratio in the TME. Consistent with enhanced tumor immunity, we found significantly higher levels of the inflammatory cytokines Ifnγ and Tnfα expression in CD8 + T cells following Cyp11a1 ablation (Fig. 6d). Moreover the expression of the co-inhibitory  (Supplementary Fig. 5a). The data indicate greater T cell functionality and less exhaustion in the Cyp11a1 cKO mice.
To test cytotoxic capacity of T and NK cell populations, we examined the degranulation response of these cells by analyzing cell surface expression of CD107a/LAMP1 in tumor-infiltrating T and NK cells. We observed a significantly increased proportion of degranulating CD107a + CD8 + T cells in Cyp11a1 cKO mice compared to control mice (Fig. 6e). The degranulation response in NK cells and CD4 + T cells was also enhanced (Fig. 6f, g).
Finally, we observed that the proportion of intratumoral Tregs was decreased (Fig. 6h). Altogether these data suggest that inhibition of T cell steroidogenesis by genetic deletion of Cyp11a1 changes immune cell composition in the tumor microenvironment in favor of anti-tumor immunity.
Similar to the genetic deletion of Cyp11a1, application of the Cyp11a1 inhibitor aminoglutethimide significantly increased frequencies of M1 macrophages and degranulating T and NK cells ( Fig. 6i-l, Supplementary Fig. 5b), recapitulating anti-tumor phenotypes. Cyp11a1 inhibition also decreased the number of M2 macrophages in the tumor (Fig. 6i). Together, these data demonstrate the potential to stimulate anti-tumor immunity through pharmacological suppression of Cyp11a1-dependent steroidogenesis pathways.

Discussion
The endocrine importance of systemic steroid hormones is well documented in regulating cell metabolism and immune cell function but the intracrine, autocrine, and paracrine role of local cell-type-specific steroidogenesis is less clear, particularly in pathologies such as cancer 1,2 . This is in part due to the lack of tools to study steroidogenesis in a tissue-specific manner in vivo. To overcome this, we generated a Cyp11a1-mCherry reporter and a conditional Cyp11a1 knockout mouse strain to identify de novo steroidogenic cells and study their role in vivo. Using these discovery tools, complemented by pharmacological intervention, we uncovered an anti-tumor immune suppression mechanism that may be exploited clinically to boost the anti-tumor immunity (Fig. 7).
The TME harbors innate immune cells (such as macrophages, neutrophils, eosinophils, mast cells, myeloid derived suppressor cells, dendritic cells, and natural killer cells), adaptive immune cells (such as T and B cells), stromal cells (such as fibroblasts, endothelial cells, pericytes, mesenchymal cells), and cancer cells.
The presence of immune cells in the tumor microenvironment raises a long-standing question: How tumors evade the immune surveillance and anti-tumor immunity? The present understanding suggests that the balance between protective antitumoral immune cells (and other protective factors) and protumoral immunosuppressive cells (and other immunosuppressive factors) determines the fate. Anti-tumoral immune cell subsets includes Th1, Tc1 (CD8 + CTL), M1 macrophage, DC1 which are antagonized by Th2, M2, tolerogenic DC, immature DC, DC2, Treg, and myeloid-derived suppressor cells. Established tumors often foster immunosuppressive and tolerogenic immune cells. In this report, the discovery of immune cell-mediated steroid biosynthesis in the TME explains how an immunosuppressive TME can be achieved. Previously it has been shown that steroids induce anergy and cell death in T cells [44][45][46]54 , M2 phenotype in macrophages 42,43 , tolerance in DCs [47][48][49] , and increase the frequency of Treg cells [50][51][52] . Here we demonstrate that tumorinfiltrating T cells can produce steroids and that the inhibition of T cell-mediated steroid biosynthesis tips the balance toward protective immunity. We have shown that ablation of intratumoral steroids by deleting Cyp11a1 in T cells reduced the number of M2 macrophages and Tregs, and increased the number of M1 macrophages, functional T cells, and NK cells. The results suggest that the secreted steroid works in a paracrine manner that effect on other infiltrating immune cells as expected. All these cell types are known to be crucial in regulating anti-tumor immunity and immunosuppression. Therefore, in the absence of intratumoral steroids, anti-tumor immunity is more effective and restricts tumor progression. Importantly, this pathway can be targeted pharmacologically. This may be exploited therapeutically in three ways: (1) steroidogenic gene expression or the presence of steroids might be used to stratify the patients, (2) use of pharmacologic drug (i.e. anti-steroidogenic) can convert the TME into protective type-1 inflammatory condition by reducing the steroid level, and (3) immunotherapy should work better in absence of immunosuppressive steroids.
Single-cell level transcriptomic analysis revealed important findings pertaining to tumor immune infiltrates and provide a rich resource for cancer immunologists. We identified three distinct classes of immune cells with steroidogenic potential, T cells (Th2 and Tc2, major population), mast cells (minor population), basophils (minor population), and a few monocyte/macrophages. GATA factors are revealed as key drivers of Cyp11a1 expression in immune cells. In Th2 cells Gata3 occupies the Cyp11a1 locus. Cyp11a1 + Th2 cells outnumbered other types of Cyp11a1expressing cells. Hence deletion of Cyp11a1 in T cells was Fig. 6 Inhibition of T cell steroidogenesis stimulates anti-tumor immunity. a-h Comparing Cyp11a1 cKO and Cd4-Cre control mice with B16-F10 cells injected subcutaneously. Representative of three independent experiments. a Tumor-infiltrating macrophages (Lin − CD11b + ) were purified by cell sorting at day 12. Arg1 and Tgfβ1 mRNA expression was quantified by RT-qPCR, with mRNA expression level normalized by Gapdh mRNA expression. N = 6. b, c Tumor-infiltrating macrophages (Lin − CD11b + F4/80 + ) analyzed by flow cytometry at day 12 to examine iNOS and Arg1 expression. Representative FACS profile of the expression (b). Representative graphical representation of one experiment (c). N = 5. d Tumor-infiltrating CD3e + CD8 + T cells purified by cell sorting at day 12, reactivated ex vivo, and Ifnγ and Tnfα mRNA expression quantified by RT-qPCR, with mRNA expression level normalized by Rplp0 expression. N = 6. e Cytotoxic T-lymphocyte degranulation assay. CD107a/LAMP1 expression on tumor-infiltrating CD8 + T cells was analyzed by flow cytometry after 12 days post-inoculation of B16-F10 cells. N = 8 (ctrl), 9 (cKO) Gating: All cells>singlets>live cells>CD8 + T cell>CD107a. f NK cell degranulation assay by measuring CD107a/LAMP1 expression on tumor-infiltrating NK cells. N = 8 (ctrl), 9 (cKO). Gating: All cells>singlets>live cells>NK cells>CD107a. g CD4 + T cell degranulation assay by measuring CD107a/LAMP1 expression on CD4 + T cells. N = 8. Gating: All cells>singlets>live cells>CD4 + T cell>CD107a. h Flow cytometric analysis to show the changes in B16-F10 subcutaneous tumor-infiltrating Treg (CD4 + CD3e + FoxP3 + ) populations upon Cyp11a1 deletion. N = 9. i-l Comparing immunophenotype in Cyp11a1 inhibitor, AG, treated and untreated mice with B16-F10 cells injected subcutaneously. All are representative of two independent experiments. i Representative FACS profile to show iNOS and Arg1 expression in tumorinfiltrating CD45 + Lin − CD11b + F4/80 + macrophages (left panel). Graphical presentation the iNOS and Arg1 expression of a representative experiment (right panel). N = 5. j-l CD107a/LAMP1 expression on CD4 + T cells (j), CD8 + T cells (k) and NK cells (l) was analyzed by flow cytometry after 12 days post-inoculation of B16-F10 cells. N = 4 (ctrl), 3 (cKO). Gating: All cells>singlets>live cells>CD4 + or CD8 + T or NK cells>CD107a. In this figure, all error bars represent mean with s.d. P-value was calculated by unpaired two-tailed t-test. N represents biologically independent animals. sufficient to remodel the TME for more effective anti-tumor immune responses. In the in vitro setting, we observed that Tregs and other T helper cell types (except Th1) can upregulate Cyp11a1 expression, but in the in vivo setting we noticed that only Th2 cells were Cyp11a1 + . We did not find Cyp11a1expressing Tregs or any other T helper cell types in the B16-F10 TME. The presence of Cyp11a1 + non-T cells indicates that in a different context, such as in different tumor types, the steroidogenic population may vary and other hematopoietic cells may also produce steroids. Future studies are essential to gain detailed insight of immune cell-mediated intratumoral steroidogenesis in different tumor type. It would be interesting to know why CD4 + T cells upregulate Cyp11a1 expression specifically in the TME, and not in the secondary lymphoid organs. This could be a result of the fact that Th2 cells are numerous in tumors but not in the spleen and lymph nodes. Therefore, further studies are needed to delineate whether this is a result of an increased Th2 cell abundance in tumors, and/or an enhanced Cyp11a1 expression per individual Th2 cell.
The induction of type-2 immune response in tumors such as breast, pancreatic, melanoma, glioma, cervical, lung, and bladder cancer has been previously reported, both in humans and mice, and its roles in the promotion of cancer by immunosuppression are known [12][13][14][15][16][17][18][19][20][21][22][23][24][25] . As expected, reanalysis of publicly available transcriptomic datasets showed evidence of many human tumor types expressing steroidogenic genes. In melanoma, we show evidence of steroidogenic gene expression being correlated with type-2 immune signature genes such as IL4. The limitation of this analysis is that it does not confirm the cellular source of CYP11A1 expression because the transcriptomes were obtained from bulk RNAseq of whole tumors. In the future, single-cell RNA sequencing of tumor-infiltrating Th2 cells from patients may provide direct evidence and further insights. Analysis of publicly available single-cell RNA sequencing datasets of human tumorinfiltrating T cells revealed that all the available datasets are from tumors that predominantly raised Th1-mediated immune responses. Therefore, as expected, we observed no or only a few CYP11A1-expressing cells in these data sets (data not shown). This is consistent with our hypothesis in the sense that we expect CYP11A1 expression predominantly in tumors exhibiting type-2 immune responses.
The study adds to the concern raised by recent studies about proper use of steroids, particularly glucocorticoids, in solid tumors as a part of the disease management, specifically when we seek to boost anti-tumor immune response [55][56][57][58][59][60][61][62] . Glucocorticoids have been used in clinical oncology for over a half a century, and at present these are routinely used to alleviate edema in patients with intracranial lesions and are first-line agents to suppress immune-related adverse events that arise with the immunotherapies. Therefore, further studies are required to evaluate the use of synthetic steroids if the tumors itself induces glucocorticoid synthesis. Reduction of intratumoral steroidogenesis may open a window for immunotherapies to work better at lower doses.
Similar experimental approaches can in future provide in depth mechanistic insights into other extraglandular (local) steroidogenic cell types such as adipose cells, neuron, osteoblasts, astrocytes, microglia, skin, trophoblast, and thymic epithelial cells. Our Cyp11a1-mCherry reporter mouse line can be used as a discovery tool to identify undiscovered steroidogenic cell types in tolerogenic physiological conditions such as pregnancy, mucosal tolerance, and inflammatory and immunopathological conditions. Their functional role can be dissected by using Cyp11a1 cKO mice using tissue-specific Cre-drivers. Nevertheless, cancer is a proven context where tumors hijack this pathway for immune evasion. Further studies in diverse physiological scenarios would be of great interest to more broadly understand the role of immune cell de novo steroidogenesis in regulating inflammation and immunity. Ethical Review Bodies (AWERB). Sample sizes were determined based on previous experience and a priori power analysis (G* Power). Housing conditions at the Sanger Institute: all the mice used in this study, were maintained in specific pathogen-free unit on a 12-h light and 12-h dark cycle (lights off at 7:30 pm and no twilight period). The ambient temperature was 21°C with a maximum variation of ±2°C. The humidity was 55 ± 10%. Mice were housed for phenotyping with a stocking density of 3-5 mice/cage (dimensions of caging: (L × W × H) 365 × 207 × 140 mm, floor area 530 cm 2 ) in individually ventilated caging (Tecniplast Seal Safe1284L) receiving 60 air changes/h. In addition to Aspen bedding substrate, we provide standard environmental enrichment of two nestlets, a cardboard Fun Tunnel and three wooden chew blocks. Given water and diet were available all the times. Mice were fed on Mouse Breeders Diet (Lab Diets, 5021-3). Animals recruited to studies at ARES Medical Research Council animal facility remained socially housed in individually ventilated cages, at ambient temperature and with cage enrichment. Animals of each genotype were randomly assigned to experimental groups, and where possible, technicians performing the experiment were blinded to experimental groups and treatments.

Methods
Generation of a Cyp11a1-mCherry reporter mouse line. Using CRISPR_Cas9 technology we introduced double-strand DNA breaks 5′ and 3′ adjacent to the Cyp11a1 termination codon in exon 9 to facilitate the introduction of our targeting construct. The 5′ and 3′ arms of homology were designed to remove the Cyp11a1 termination codon and 100 bp of the 3′ UTR immediately downstream and replace it with a minimal T2a self-cleavage peptide followed by the fluorescent marker mCherry. Using the web-based tool designed by Hodgkins et al. 63 , two sgRNAs were identified 5′ and 3′ adjacent to the Cyp11a1 termination codon. The guide sequences were ordered from Sigma Genosys as sense and antisense oligonucleotides, and annealed before individually cloning into a human U6 (hU6) expression plasmid (Addgene # 41824). Next, we generated targeted ES cell (ESCs) through nucleofection of 3 × 10 6 JM8 ESCs with 2 μg Cyp11a1 circular targeting construct, 1.5 μg of each hU6_sgRNAs, and 3 μg a plasmid expressing human codonoptimized CAS9 driven by with the CMV promoter (Addgene # 41815). 48 h post transfection the media was changed for G418 selection media (165 ug/ul). Confirmation of the correct targeting events were confirmed by quantitative PCR (qPCR) for loss of heterozygosity (LOH) and the presence of the selectable marker, and long range (LR) PCR and Sanger sequencing. Following blastocyst injection and chimaera breeding three F1 Cyp11a1 +/mCherry-Neo mice were bred to pCAGGSs-Flpo 64 mice to remove the neomycin selectable marker to generate Cyp11a1 +/mCherry mice. Genotyping primer sequence ( Supplementary Fig. 1a Lung colonization of EO771 cells: 4 × 10 5 cells were injected intravenously in to the tail vein. After 10 days (±1 day) the mice were killed via cervical dislocation, and their lungs were removed, rinsed in PBS and fixed in 10% neutral buffered formalin. The fixed lungs were paraffin embedded and sections stained with Hematoxylin and Eosin using routine histology protocols. The microscopic images were captured in Olympus with ×10 magnification and the number of metastatic foci were counted. Treatment with aminoglutethimide (AG): for both the B16-F10 and E0771 models 200 μl of a 2-mg/ml AG or vehicle were administered via the intraperitoneal route. The first dose was administered concurrently with cell implantation and repeated every 2 days thereafter.
Tumor tissue processing. Tumors were mechanically dissociated and digested in 1 mg/ml collagenase D (Roche), 1 mg/ml collagenase A (Roche), and 0.4 mg/ml DNase I (Sigma) in IMDM media containing 10% FBS, at 37°C for 40 mins. EDTA was added to all samples to neutralize collagenase activity (final concentration 5 mM) and digested tissues were passed through 70μm filters (Falcon).
Cell sorting. Once processed, single-cell suspension tumor samples were incubated with a fixable fluorescent viability stain (Life Technologies) for 20 mins (diluted 1:1000 in PBS) prior to incubation with conjugated primary antibodies for 30 mins at 4°C. Antibodies were diluted in PBS with 0.5% bovine serum albumin (BSA). Stained samples were sorted, using the MoFlo XDP or BD Influx cytometer system.
Single-cell RNA sequencing. Single cells were isolated from processed B16-F10 subcutaneous tumors using a fluorescence-activated cell sorter. Once processed, tumor tissues were incubated with a fixable fluorescent viability stain (Live-Dead Violet, Life Technologies) for 15 mins (diluted 1:1000 in PBS) prior to incubation with conjugated primary antibodies for 30 mins at 4°C in the dark. Antibodies were diluted in flow cytometry staining buffer (eBioscience/Thermo Fisher Scientific). After staining, samples were sorted using the BD influx flow cytometer. Single-cells were harvested in 2 μl of Lysis Buffer (1:20 solution with RNase Inhibitor (Clontech, cat. no. 2313 A) in 0.2% v/v Triton X-100 (Sigma-Aldrich, cat. no. T9284)) in 96-well plates. After sorting the plates were spun down quickly and kept frozen at −80°C. Then we performed reverse transcription (RT) and cDNA pre-amplification following SmartSeq2 protocol to obtain mRNA libraries from single-cells. Oligo-dT primer, dNTPs (ThermoFisher, cat. no. 10319879) were then added to the samples. RT-PCR were performed using 50U of SMARTScribe™ Reverse Transcriptase (Clontech, cat. no. 639538). cDNA libraries were prepared using the Nextera XT DNA Sample Preparation Kit (Illumina, cat. no. FC-131-1096), following the protocol supplied by Fluidigm (PN 100 − 5950 B1). Single-cell libraries were pooled, purified using AMPure XP beads (Beckman Coulter) and sequenced on an Illumina HiSeq 2500 aiming for an average depth of 1 million reads/cell (paired-end 100-bp reads).
Single-cell RNA sequencing analysis. The sequencing data was mapped using STAR 2.5.1b and quantified using htseq 0.9.0. The data was subsequently processed in Scanpy 1.3.6 with the standard Seurat-inspired workflow. Cells with above 20% mitochondrial reads were removed, and the data was transformed to log(CPM/100 + 1) form. Highly variable genes were identified via normalized dispersion, with the top 10% of genes reported as variable. The data was filtered to highly variable genes, scaled and used for PCA. The top 20 principal components were used to infer a neighbor graph, which served as the basis of UMAP dimensionality reduction and Leiden clustering with a 0.4 resolution value. Per-cluster markers were identified using default Scanpy settings, using log(CPM/100 + 1) data on input. Genes correlated with Cyp11a1 were identified with the Pearson Correlation Coefficient. Potential Cyp11a1 regulators were detected via pySCENIC, replicating the demonstration notebook.
ATAC-seq. Naïve T cells isolation and in vitro T helper cell differentiation were carried out as described above. After 72 h of continuous activation/differentiation, Th1, Th2, and Th17 cells were harvested. The ATAC-seq experiments were performed following the published protocol with some modification 66 . Briefly, 50,000 cells were washed with ice-cold 1X Dulbecco's PBS (twice), and resuspended in a sucrose swelling buffer (0.32 M sucrose, 10 mM Tris.Cl, pH 7.5, 3 mM CaCl 2 , 2 mM MgCl 2 , 10% glycerol). The cell suspension was incubated for 10 m on ice. In next, a final concentration of 0.5% NP-40 was added, and the cells suspension was vortexed for 10 s and incubated for 10 m on ice. Nuclei were pelleted at 500 × g at 4°C for 10 m. Then, the nuclei pellets were washed once with 1X TD buffer (from Nextera DNA Library  ATAC-seq data processing. ATAC-seq reads were aligned using Bowtie 2 67 with the parameter -X 2000 and the mouse genome mm10. This was followed by peak calling on each replicate individually using MACS2 68 with the function callpeak and the parameters -B -SPMR -shift -100 -extsize 200. The peaks obtained were kept if they overlapped a peak from the other replicate of the same T helper cell types. The resulting bedGraph files generated by MACS2 were converted to bigWig format for the visualization on UCSC genome browser 69 . ChIP-seq data processing. GATA3 ChIP-seq data sets from different Tn subtypes were downloaded from the NCBI GEO database (GSE20898) 70 . The SRA files were converted to fastq files using the SRA toolkits (http://ncbi.github.io/sra-tools/). Reads were aligned to the mm10 genome using Bowtie 2 67 with default parameters. Peak calling was performed by using MACS2 68 with the function callpeak and the parameters -B -SPMR -nomodel -extsize 200. The resulting bedGraph files generated by MACS2 were converted to bigWig format for the visualization on UCSC genome browser 69 .
B16-F10 and T cell co-culture assay. Negatively purified splenic naïve CD4 + T cells from Cyp11a1-mCherry mice were cultured in presence of B16-F10 cell with or without TCR activation and analyzed by flow cytometry to detect Cyp11a1-mCherry expression. Naïve T cell purification and activation condition is described in the T helper cell culture section above. Briefly, naïve splenic cells were activated in anti-CD3e and anti-CD28 coated plate in presence of IL2 but absence anti-IL4 neutralizing antibody.
Quantitative PCR. Tumor-infiltrating macrophages (Lin − CD11b + ) and CD8 + T cells were purified by cell sorting. We used the Cells-to-C T kit (Invitrogen/ Thermofisher Scientific) and followed SYBR Green format according to manufacturer's instructions. In all, 2 μl of cDNA was used in 12 μl qPCR reactions with appropriate primers and SYBR Green PCR Master Mix (Applied Biosystems). Data were analyzed by ddCT method. Experiments were performed three times and data represent mean values ± standard deviation. The primer list is provided below: Arg1: F-ATGGAAGAGACCTTCAGCTAC R-GCTGTCTTCCCAAGAGTTGGG Tgfβ1: F-TGACGTCACTGGAGTTGTACGG R-GGTTCATGTCATGGATGGTGC Ifnγ: F-ACAATGAACGCTACACACTGC Quantitative ELISA. CD45 + leukocytes were purified from B16-F10 tumor masses and lungs, of mice that had been tail vein administered B16-F0 cells, and seeded at equal density in IMDM medium supplemented with 10% charcoal stripped FBS (Life Technologies, Invitrogen) for 24 h. Pregnenolone concentrations of the culture supernatants were quantified using pregnenolone ELISA kit (Abnova) and corticosteroids ELISA (Thermofisher) kit following the manufacturers' instruction. Absorbance was measured at 450 nm, and data were analyzed in GraphPad Prism 6.
T cell proliferation assay. Negatively purified splenic naïve CD4 + T cells were purified from Cyp11a1 cKO, Cd4-Cre and Cyp11a1 fl/fl mice, stained with CellTrace Violet following the CellTrace Violet Cell Proliferation Kit (Invitrogen) protocol, activated in vitro, and the cell proliferation profile was captured by flow cytometry based dye decay assay on BD Fortessa 65 . Data were analyzed in FlowJo v9.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.
Supporting data for the Supplementary Fig. 4j-i can be found in the Supplementary Data 2. All remaining relevant data are available in the article, supplementary information, or from the corresponding author upon reasonable request.