Farnesoid X Receptor Constructs an Immunosuppressive Microenvironment and Sensitizes FXR^highPD-L1^low NSCLC to Anti–PD-1 Immunotherapy

Non–small cell lung cancer (NSCLC), which constitutes about 85% of all lung cancers, remains a leading cause of cancer mortality worldwide (1). The identification of tumor oncogenic gene alterations has transformed the management of NSCLC, leading to responses in selected patients treated with matched tyrosine-kinase inhibitors (2). Moreover, the tumor microenvironment (TME) emerged as a target for anticancer therapy development, because the interaction between tumor cells and stromal tumor–promoting immune cells also plays a role in NSCLC progression (3). Progress in immunotherapy, targeting the tumor-promoting immunosuppressive microenvironment, has been promising for treatment of cancers including NSCLC. However, many issues remain to be addressed in immunotherapy, especially in broadening the range of patients who can benefit from immunotherapies.

Farnesoid X receptor (FXR) is a member of the nuclear receptor superfamily and is expressed in various tissues including liver, intestine, kidneys, and adrenal gland (4). As a ligand-activated transcription factor, FXR regulates expression of target genes involved in enterohepatic circulation and lipid homeostasis (5). However, emerging evidence has indicated the relevance of FXR in tumorigenesis. FXR deficiency in mice leads to increased colon cell proliferation and spontaneous liver tumors (6, 7). On the other hand, FXR has a causative role in esophageal cancer, breast cancer, and pancreatic cancer (8–10). We previously reported that FXR is upregulated in NSCLC, compared with pericarcinous lung tissues (11). Our data also showed that FXR contributes to NSCLC cell proliferation via transactivating CCND1, suggesting an oncogenic role for FXR in NSCLC progression.

FXR has been implicated as a regulator of inflammation and immune responses in a subset of immune-mediated disorders. For example, activation of FXR reduced immune cell infiltration and expression of inflammatory mediators (e.g., MCP-1, IL1β, IL6, and IFNγ) in animal models of nonalcoholic steatohepatitis or colitis (12, 13). Zhang and colleagues showed that expression of FXR repressed proinflammatory genes and improved lung permeability in FXR^−/− mouse acute lung injury models (14), implying that FXR might contribute to NSCLC progression by modulating the immune microenvironment as well. In the same set of NSCLC samples as used in our previous work (11), we observed a negative correlation between FXR expression and expression of the checkpoint PD-L1, which orchestrates immunosuppression in cancer. We identified a subgroup of FXR^highPD-L1^low NSCLC patients (Supplementary Table S1). In this study, we sought to determine whether FXR affects the tumor immune microenvironment of NSCLC, especially in the context of FXR^highPD-L1^low tumors. Our study provides insights into the immunosuppressive role for FXR in the FXR^highPD-L1^low NSCLC subtype. We recommend FXR^highPD-L1^low status as a biomarker for guiding anti–PD-1 immunotherapy.

Human NSCLC cell lines NCI-H1299, HCC4006, A549, and NCI-H1975, mouse Lewis lung carcinoma (LLC) cell line, mouse colon carcinoma (MC38) cell line, and mouse melonoma (S91) cell line were obtained from the ATCC in 2003 (LLC), 2011 (A549), 2013 (NCI-H1299 and HCC4006), and 2015 (MC38, NCI-H1975, and S91). HCC827 cells were obtained from the Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences in 2010. All cell lines used in this study were cultured according to the suppliers' instructions. Cells were checked to be Mycoplasma-free and passaged no more than 25 to 30 times after thawing. Cell lines were characterized by Genesky Biopharm Technology using short tandem repeat markers (latest tested in 2017).

From 2008 to 2010, 156 NSCLC samples and matched normal lung tissues with complete clinical and follow-up information were consecutively collected from the Department of Pathology in Ren Ji Hospital (Shanghai, China). Another 74 NSCLC samples without clinicopathologic or survival information were also collected from Ren Ji Hospital. Tissue microarrays were constructed as in our previous study (11). Two cores with a 1.6-mm diameter were obtained from the original paraffin block of each sample. Patients who had received neoadjuvant chemotherapy or radiotherapy were excluded. Tissue microarrays containing 178 NSCLC samples and matched normal lung tissues were obtained from Outdo Biotech Co. Ltd (product ID: HLugA180Su03 and HLugS150Su02) and ZuoChengBio Co. Ltd (product ID: LUC1505). A total of 408 NSCLC cases were enrolled. This study was approved by the Ethics Committee of Ren Ji Hospital. All experiments were conducted according to the Declaration of Helsinki principles, as well as the approved guidelines of School of Medical Graduate Shanghai Jiao Tong University. Written informed consent was obtained from all subjects.

The tissue microarrays and paraffin-embedded mouse tumors were subjected to IHC staining as previously described (15). For human and mouse FXR detection, anti-NR1H4 (Sigma-Aldrich) was applied at 1:100 and 1:200, respectively. For human and mouse PD-L1 detection, anti–PD-L1 (Zeta Life) was applied at 1:50. Isotype controls were conducted simultaneously using concentration-matched nonspecific rabbit IgG. The IHC staining results were independently evaluated by two trained pathologists. Staining intensity of tumor cells was scored 0 through 3 as negative (0), weak (1), moderate (2), and intense (3). The percentage of positive cells was binned as 0 through 4 as 0% (0), 1% to 25% (1), 26% to 50% (2), 51% to 75% (3), and 76% to 100% (4). The final IHC score was the multiplication of these two scores (range, 0–12). Human FXR expression in tissue microarrays was defined as low (score 0–4) or high (score 6–12), and human PD-L1 in tissue microarrays defined as low (score 0–2) or high (score 3–12), respectively (11, 16).

Transfection was carried out using Lipofectamine RNAiMAX reagent (Invitrogen) according to the manufacturer's protocol. The target sequences of double-stranded nucleotides used in siRNA transfection were 5′-GGACCATGAAGACCAGATT-3′ (FXRsiRNA-1) and 5′-GACCTCGACAACAAAGTCA-3′ (FXRsiRNA-2) for FXR knockdown, 5′-TCGCCCTATCATTGGAGAT-3′ (SHPsiRNA-1), 5′-CCAGCTATGTGCACCTCAT-3′ (SHPsiRNA-2) and 5′-CCAAGATGCTGTGACCTTT-3′ (SHPsiRNA-3) for small heterodimer partner (SHP) knockdown in NSCLC cells; 5′-CCAAGAACGCCGTGTACAA-3′ (mouse FXRsiRNA-1) and 5′-GCGTGATGGACATGTACAT-3′ (mouse FXRsiRNA-2) for FXR knockdown in mouse S91 cells. The target sequence of negative control (NC) was 5′-TTCTCCGAACGTGTCACGT-3′ (RiboBio). To inhibit the activity of phospho-EGFR, H1975 and HCC827 cells were incubated with afatinib (0.8 μmol/L) or gefitinib (0.02 μmol/L, Selleck Chemicals) 24 hours after transfection, respectively.

Lentiviral vectors were generated with OBiO Biotechnology, using pCMV-NR1H4-PGK-PuroR or pCMV-mNR1H4-PGK-PuroR plasmids, which carry the full-length human NR1H4 (GenBank accession NM_005123.3) or mouse NR1H4 (GenBank accession NM_009108.2) coding sequence, respectively. Cells were infected with lentiviral vectors as in our previous study (11).

A549 and LLC cells overexpressing FXR were seeded into 96-well plates (8 × 10³/well) overnight. Human peripheral blood mononuclear cells (PBMC; HemaCare) were used for CD8⁺ T-cell sorting via magnetic sorting techniques (Miltenyi Biotec). Mouse splenocytes from 7-week-old C57BL/6 female mice were also used for CD8⁺ T-cell sorting magnetically (Stemcell Technologies). Human PBMCs, mouse splenocytes, and corresponding CD8⁺ T cells were activated by plate-bound anti-CD3 (OKT3; 2 μg/mL, BioLegend), soluble anti-CD28 (1 μg/mL, BioLegend) mAbs and IL2 (5 ng/mL, PeproTech), and then added to the attached A549 or LLC in triplicate at an effector-to-target ratio of 25:1. After 3 days, PBMCs, splenocytes, and CD8⁺ T cells were harvested from the coculture system, stained with CD8-APC-H7 (BD Biosciences), IFNγ-APC and TNFα-PE (eBioscience), and analyzed by flow cytometry. The proliferation of cocultured CD8⁺ T cells was examined using the CellTrace Violet Cell Proliferation Kit (Invitrogen Corporation).

For in vitro analysis, cells were harvested and incubated with fluorochrome-linked antibodies for 30 minutes at 4°C. Matched nonspecific mouse/rat IgG served as an isotype control. After being washed two times, cells were analyzed using a Calibur flow cytometer (BD Biosciences) equipped with FlowJo version 10 software. For in vivo studies, mouse tumors were minced and digested using the Mouse Tumor Dissociation Kit (Miltenyi). Cells were passed through a 70-μm cell strainer, and single-cell suspensions were then analyzed by Aria III flow cytometry (BD Biosciences) as described above, with Fixable Viability Stain 510 (BD Biosciences) to discriminate viable and dead cells. The following fluorochrome-linked antibodies were used in this study. Anti-human: PD-L1–PE (BioLegend). Anti-mouse: CD45-PerCP-Cy5.5, CD45-APC-Cy7, Granzyme B-PE-Cy7, TNFα-PE, IFNγ-APC, PD1-PE, Lag3-APC, F4/80-BV421, CD206-APC, CD3e-FITC, CD4-PE-Cy7, CD25-APC, Foxp3-PE (eBioscience), Fixable Viability Stain 510, CD8-APC-H7, CD49b-BV421, CD69-BV711, CD11b-FITC, CD11c-BV650, CD86-PE, MHC II-BV650, PD-L1-BV786 (BD Biosciences), and CD11c-PE-Cy7, Gr-1-PE, MHC II-PerCP-Cy5.5 (BioLegend). Fixation/permeabilization buffers (eBioscience) were used for intracellular staining.

Western blot analysis was performed as previously described (11). Anti-human bile acid receptor (NR1H4; Abcam), anti-SHP (Santa Cruz Biotechnology), anti-phospho-EGF receptor (Tyr1068), anti-EGF receptor, anti-phospho-Akt (Ser473), anti-Akt, anti-phospho-p44/p42 MAPK (Erk1/2; Thr202/Tyr204), anti-p44/p42 MAPK (Erk1/2), anti–β-actin (Cell Signaling Technology), and anti-mouse bile acid receptor (NR1H4; Abcam) were used according to the manufacturer's recommended dilutions.

Quantitative RT-PCR was carried out as previously described (11). β-Actin mRNA was amplified as an internal control. Primers were designed as follows: human FXR forward, 5′-GATTGCTTTGCTGAAAGGGTC-3′; reverse, 5′-CAGAATGCCCAGACGGAAG-3′; SHP forward, 5′-AGGCCTCCAAGCCGCCTCCCACATTGGGC-3′; reverse, 5′-GCAGGCTGGTCGGAAACTTGAGGGT-3′; PD-L1 forward, 5′-CAATGTGACCAGCACACTGAGAA-3′;reverse, 5′-GGCATAATAAGATGGCTCCCAGAA-3′; β-actin forward, 5′-TTGCTGATCCACATCTGCT-3′; reverse, 5′-GACAGGATGCAGAAGGAGAT-3′; mouse FXR forward, 5′-TGTGAGGGCTGCAAAGGTTT-3′; reverse, 5′-ACATCCCCATCTCTCTGCAC-3′; PD-L1 forward, 5′-GACCAGCTTTTGAAGGGAAATG-3′; reverse, 5′-CTGGTTGATTTTGCGGTATGG-3′; β-actin forward, 5′-GGCATTGTTACCAACTGGGACGAC-3′; reverse, 5′-CCAGAGGCATACAGGGACAGCACAG-3′.

The ChIP assays were conducted using a SimpleChIP Plus Enzymatic Chromatin IP Kit (Cell Signaling Technology) as in our previous study (11). ChIP primers for PD-L1 promoter were designed as follows: forward 5′-CACCCATACACACACACACAC-3′; reverse 5′-ACAACAAGCCAACATCTGAAC-3′.

Luciferase reporter plasmids carrying either wild-type or FXR-responsive element (FXRE)-deleted PD-L1 promoter sequence were constructed by OBiO Biotechnology. The luciferase reporter assays were performed using a Dual-Luciferase Reporter Assay System (Promega) as our previous study (11). The firefly luciferase value was normalized to the corresponding Renilla luciferase value in order to compare the transfection efficiency.

Female C57BL/6 mice (4–6-week-old) were housed and maintained under specific pathogen-free conditions. To generate subcutaneous xenograft tumors, LLC-Mock or LLC-FXR cells (1 × 10⁶) were suspended in 100 μL of phosphate-buffered saline and inoculated subcutaneously (s.c.) into the right flanks of C57BL/6 mice. Subsequently, the well-developed tumors were cut into 1 mm³ fragments and transplanted s.c. into the right flanks of C57BL/6 mice using a trocar. When the tumor volume reached 100 mm³, the mice were randomly assigned into a control group and a treatment group. The control group was given isotype control rat IgG2a or IgG2b, and the treatment group received either anti-mouse PD-1 InVivoMAb (clone RMP1-14, 200 μg per mouse, Bio X Cell) or anti-mouse PD-L1 InVivoMAb (clone 10F.9G2, 200 μg per mouse, Bio X Cell) intraperitoneally (i.p.) every 3 days for 11 to 15 days. The size of tumors was measured twice per week using a microcaliper. Tumor volume = (length × width²)/2. Animal studies using mouse models were approved by the Institutional Animal Care and Use Committee at Shanghai Institute of Materia Medica (approval no.2018-03-GMY-03).

Statistical analysis was carried out with SPSS 17.0 (SPSS Inc.) and Prism 6.0 (GraphPad). Quantitative data were presented as means ± standard deviation (SD), except for mice tumor volume presented as means ± SEM. Comparisons between two groups were performed using Student t test, Mann–Whitney U test, or Wilcoxon rank-sum test. Survival data were analyzed by the Kaplan–Meier method (log-rank test). Correlational analysis was conducted using the χ² test. All tests were two-sided and a P value of < 0.05 was considered statistically significant.

First, we sought to confirm the negative correlation between FXR and PD-L1. For this, we extended the cohort of NSCLC specimens with 408 cases in total to evaluate FXR and PD-L1 expression by IHC staining. Due to their subcellular location for normal functions, FXR nuclear staining and PD-L1 membrane staining in tumor cells were scored for further analysis (4, 17). Consistent with previous reports (18), PD-L1 expression was significantly higher in NSCLC than in pericarcinous lung tissues (Supplementary Fig. S1A and S1B). Moreover, PD-L1 was significantly lower in “FXR high” tumors than in “FXR low” tumors (Fig. 1A and B). χ² analysis revealed that there was a statistically significant inverse correlation between FXR and PD-L1 in NSCLC samples (Table 1). We found that the proportion of FXR^highPD-L1^low tumors was up to 42.6% in all NSCLC samples. The survival curves for all four subgroups (FXR^highPD-L1^low, FXR^lowPD-L1^high, FXR^highPD-L1^high, and FXR^lowPD-L1^low) have been presented in Fig. 1C. Here, due to the inverse correlation between FXR and PD-L1 in NSCLC specimens, we focused mainly on the FXR^highPD-L1^low and FXR^lowPD-L1^high subgroups and found that the FXR^highPD-L1^low subgroup had a significantly worse overall survival as compared with FXR^lowPD-L1^high NSCLC patients (Fig. 1C).

We asked whether the inverse correlation between FXR and membrane PD-L1 in NSCLC specimens indicated a negative regulatory interaction. We therefore measured the amount of PD-L1 in FXR-silenced or -overexpressed NSCLC cells. SHP, a downstream gene target of FXR, was used to track alterations of FXR activity in experiments (19). Our data showed that FXR knockdown resulted in increased surface expression of PD-L1 in H1299, H1975, and HCC827 cells (Fig. 1D; Supplementary Fig. S2A and S2B). By contrast, stable FXR overexpression in A549 and HCC4006 cells significantly suppressed cell-surface PD-L1 expression compared with Mock groups (Fig. 1E; Supplementary Fig. S2C). Similar changes of PD-L1 mRNA were also observed in these cells upon FXR silencing or overexpression (Fig. 1D and E; Supplementary Fig. S2A–S2C). We measured expression of PD-L1 in FXR-silenced or -overexpressed murine cancer cells as well. Accordingly, FXR knockdown in S91 significantly increased PD-L1 surface expression and mRNA expression (Supplementary Fig. S3A), whereas stable FXR overexpression in LLC and MC38 cells significantly repressed PD-L1 expression either as protein at the cell surface or as mRNA (Fig. 1F; Supplementary Fig. S3B).

The molecular mechanisms by which FXR downregulates PD-L1 expression in NSCLC were subsequently investigated. As a transcription factor, FXR binds to FXRE in the DNA to regulate transcription of target genes (20). Gene sequence analysis revealed that the −1906/+94 region of the human PD-L1 promoter contains a putative FXRE sequence (TGTTCAGTCACCT) at nucleotide −378 from the transcriptional initiation site. Our ChIP assays detected enriched bands from chromatin immunoprecipitated with anti-FXR in control H1975 and H1299 cells, compared with that of chromatin immunoprecipitated with isotype IgG (Fig. 2A). Moreover, FXR knockdown significantly abrogated the recruitment of FXR to the putative FXRE motif in H1975 and H1299 (Fig. 2A–C), whereas enforced FXR expression led to significant enhancement of FXR binding to PD-L1 promoter in HCC4006 cells (Fig. 2D and E). To expand upon these findings, luciferase reporter assays were carried out using pGL3 luciferase constructs harboring the PD-L1 promoter with or without deleted FXRE element. In both H1975 and H1299 cells, FXR knockdown significantly increased the wild-type PD-L1 promoter activity without apparently affecting the activity of the FXRE-deleted PD-L1 promoter (Fig. 2F and G). In contrast, enforced FXR expression significantly reduced the wild-type PD-L1 promoter activity in HCC4006 cells, whereas the FXRE-deleted luciferase reporter was excluded from this reduction (Fig. 2H). The above data demonstrated that FXR can directly bind to the putative FXRE motif in PD-L1 promoter and repress its transcription.

We then asked if the effect of FXR on PD-L1 expression depends on its downstream effector, SHP (19). SHP silencing led to significantly increased PD-L1 expression in H1299 cells (Fig. 2I). SHP depletion reversed the downregulation of PD-L1 induced by ectopic FXR overexpression in A549 cells (Fig. 2J), suggesting that SHP is involved in FXR-induced PD-L1 downregulation in NSCLC cells.

In addition, we found that phosphorylated EGFR increased in EGFR-mutated FXR-silenced H1975 and HCC827 cells (Supplementary Fig. S4A and S4B), but not in EGFR-wild-type FXR-silenced H1299 or FXR-overexpressed A549 cells (Supplementary Fig. S4C). Subsequent experiments showed that afatinib and gefitinib suppressed this increase in phosphorylated EGFR, as well as impaired the upregulation of PD-L1 either as protein at the cell surface or as mRNA induced by FXR knockdown in H1975 and HCC827 cells (Supplementary Fig. S4A, S4B, S4D, and S4E), implying a functional role for EGFR signals in FXRsiRNA-induced PD-L1 upregulation in EGFR-addicted NSCLC cells. Collectively, these results indicated that FXR downregulates PD-L1 via transrepression and signaling through SHP and EGFR signals in NSCLC cells. On this basis, we defined the FXR^highPD-L1^low NSCLC subtype.

Thus, we found that FXR and PD-L1 expression were inversely correlated in NSCLC specimens. We identified a subgroup of NSCLC patients harboring FXR^highPD-L1^low tumor cells, in which FXR downregulated PD-L1 expression.

Next, we examined the effects of FXR on the tumor immune microenvironment in patients harboring FXR^highPD-L1^low NSCLC cells. Because CD8⁺ T cells represent the main target of the PD-L1/PD-1 checkpoint pathway in the TME (21), we used in vitro coculture of FXR^highPD-L1^low NSCLC cell lines with CD8⁺ T cells or PBMCs or mouse splenocytes. We tested CD8⁺ T-cell function and proliferation. In addition to the depressed PD-L1 expression in FXR-overexpressing A549 cells, cells cocultured with purified CD8⁺ T cells induced a decreased ratio of activated CD8⁺ T cells to total CD8⁺ T cells, evident from a decrease in CD8⁺ T cells positive for either IFNγ or TNFα compared with when CD8⁺ T cells were cocultured with mock A549 cells (Fig. 3A; Supplementary Fig. S5A). Similar results were obtained from PBMCs cocultured with FXR-overexpressing A549 cells (Fig. 3B). Under the same coculture conditions, CD8⁺ T-cell proliferation was inhibited upon FXR high expression (Supplementary Fig. S5B). We observed similar effects of FXR in the FXR^highPD-L1^low mouse cancer cell line. There was a nonsignificant trend toward a reduced ratio of IFNγ⁺ CD8⁺ T cells or TNFα⁺ CD8⁺ T cells following coculture of mouse CD8⁺ T cells with FXR-overexpressing LLC cells as compared with the mock LLC control (Fig. 3C; Supplementary Fig. S5C). Accordingly, the proportion of CD8⁺ T cells positive for either IFNγ or TNFα was significantly decreased when mouse splenocytes were cocultured with FXR-overexpressed LLC cells instead of the mock cells (Fig. 3D). These data showed that in FXR^highPD-L1^low NSCLC cells, FXR represses antitumor CD8⁺ T-cell proliferation and function in vitro.

To determine whether the immunosuppressive effects of FXR in FXR^highPD-L1^low NSCLC cells could be recapitulated in vivo, C57BL/6 mice were subcutaneously inoculated with LLC cells with or without ectopic FXR, and the tumor-infiltrating immune cells were characterized. Consistent with our previous study (11), enforced FXR expression in LLC led to a significantly increased tumor growth in this murine syngeneic tumor model (Fig. 4A). IHC analysis demonstrated downregulation of membrane PD-L1 in mouse LLC-FXR tumors, indicating an FXR^highPD-L1^low phenotype, as compared with LLC-Mock tumors (Fig. 4B). Results showed that although the frequency of total hematopoietic cells (CD45⁺) did not differ between the two groups, LLC-FXR tumors exhibited significantly increased infiltration of myeloid cell populations, such as dendritic cells (DCs) and myeloid-derived suppressor cells (MDSCs), and significantly decreased infiltration of lymphoid cell populations, such as CD8⁺ cytotoxic T cells, natural killer (NK) cells, and CD4⁺ effector T cells, as compared with LLC-Mock tumors (Fig. 4C). We found a lower ratio of CD8⁺ T cells to regulatory T cells (Treg) and decreased DC activation in the LLC-FXR group as compared with the LLC-Mock group (Fig. 4C). We further interrogated molecular features of these immune cells and found significantly increased PD-1 and increased, but not significant, exhaustion marker Lag-3 expression on NK cells in LLC-FXR tumors than in LLC-Mock tumors (Fig. 4C). CD8⁺ T cells from FXR-overexpressed tumors exhibited an inactivated and exhausted phenotype, shown as significantly decreased TNFα⁺ CD8⁺ T cells as well as increased Lag-3 expression on CD8⁺ T cells relative to mock tumors (Fig. 4C). These findings were consistent with our observations in vitro, showing that FXR suppresses tumor-infiltrating immune cells and thus constructs an immunosuppressive microenvironment in FXR^highPD-L1^low mouse LLC tumors.

We then sought translational implications for the FXR^highPDL1^low NSCLC subtype. Thus, anti–PD-1 was used. Anti–PD-L1 was included as a NC. Our results showed that anti–PD-1 led to significant tumor inhibition in FXR-overexpressed LLC tumors, but caused almost no change in mock LLC tumors (Fig. 5A). In parallel, anti–PD-L1 induced no significant change in terms of tumor volume in mouse LLC-FXR tumors (Supplementary Fig. S6A).

The tumor-infiltrating immune cells were then analyzed using flow cytometry. We found that the frequency of T cells expressing CD69, IFNγ, or TNFα was significantly higher in anti–PD-1-treated LLC-FXR tumors than in control LLC-FXR tumors (Fig. 5B). In contrast, the infiltration of Tregs within mouse LLC-FXR tumors was significantly reduced by PD-1 blockade (Fig. 5C). As expected, the frequency of T cells expressing CD69, IFNγ, TNFα, or Granzyme B did not differ significantly between anti–PD-L1-treated LLC-FXR tumors and control tumors, even though decreased Treg infiltration was observed within LLC-FXR tumors treated with anti–PD-L1 (Supplementary Fig. S6B and S6C). In total, these findings demonstrated that anti–PD-1 immunotherapy was effective against FXR^highPD-L1^low mouse LLC tumors, which were associated with a reactivation of antitumor immunity.

The TME, especially its immunologic attributes, plays a role in NSCLC progression (3). FXR has been implicated as a regulator of inflammation and immune responses in various diseases (12, 13). Zhang and colleagues found a protective and anti-inflammatory role of FXR in a lipopolysaccharide-induced mouse model of acute lung injury (14). But the relevance of FXR to the immune microenvironment of NSCLC has been unclear. Here, we found an inverse correlation between FXR and PD-L1 expression in a cohort of 408 samples from patients with NSCLC. We identified an FXR^highPD-L1^low NSCLC subgroup. Indeed, PD-L1 was downregulated by FXR in NSCLC cells. We showed in vitro and in an animal model that FXR constructs an immunosuppressive microenvironment, characterized by inactivated and exhausted CD8⁺ T cells, that enables anti–PD-1 therapy in FXR^highPD-L1^low mouse LLC tumors. The present study extends FXR function to an immunosuppressive role in FXR^highPD-L1^low NSCLC subtype, and recommends FXR^highPD-L1^low as a biomarker to guide anti–PD-1 immunotherapy.

We observed an inverse correlation between FXR and PD-L1 in NSCLC samples, pointing to a subgroup of FXR^highPD-L1^low NSCLC patients. In support of this definition, we found that FXR knockdown led to increased PD-L1 expression, whereas FXR overexpression induced PD-L1 downregulation either in NSCLC cells or in murine cancer cells. These results, which characterize a functional role for FXR in PD-L1 regulation in cancer, are consistent with suggestions from a previous study, showing that FXR agonists' regimen could reduce the expression of PD-L1 and PD-1 in spleen CD4⁺ T cells and CD19⁺ B cells in an experimental autoimmune encephalomyelitis mouse model (22). Mechanistically, our data showed that FXR can repress PD-L1 transcription by binding to the putative FXRE element in the PD-L1 promoter. In addition, SHP and EGFR signals are involved in FXR-induced PD-L1 downregulation in NSCLC cells.

We previously found that FXR contributes to tumor growth through increasing cyclin D1 transcription in NSCLC (11). In this study, the tumor-promoting effect of FXR was verified in an immune-competent murine LLC tumor model. FXR was found to repress effector function and proliferation of CD8⁺ T cells, which were cocultured in vitro with FXR^highPD-L1^low NSCLC cells. In the context of FXR^highPD-L1^low mouse LLC tumors, FXR remodeled an immunosuppressive microenvironment, as manifested by increased infiltration of myeloid cell populations, such as MDSCs, and decreased infiltration of antitumor lymphoid cell populations, such as CD8⁺ cytotoxic T cells, NK cells, and CD4⁺ effector T cells. CD8⁺ T cells from FXR-overexpressed tumors exhibited an inactivated and exhausted phenotype (shown as decreased TNFα⁺CD8⁺ T cells as well as increased Lag-3 expression on CD8⁺ T cells). These results are in accordance with prior studies showing the requirement for FXR activation in inducing anti-inflammatory macrophages and MDSCs as well as in suppressing T lymphocyte responses in mouse models of autoimmune diseases (22–24). Based on the change of multiple immune cells in FXR-overexpressed LLC tumors, we reasoned that FXR should influence different cell types within the tumor immune microenvironment. In agreement with this notion, our in vivo data showed that forced expression of FXR in mouse LLC tumors contributed to decreased PD-L1 expression, not only on tumor cells but also on infiltrating immune cells. However, our study has some limitations. The detailed mechanisms for FXR-modulated immunosuppressive microenvironment in FXR^highPD-L1^low NSCLC subtype have not been illustrated in the present work. Further studies are warranted to uncover it.

Immunotherapy has changed the landscape of lung cancer treatment. Immune-checkpoint inhibitors, especially for PD-L1/PD-1-blocking antibodies, have improved treatments for patients with advanced NSCLC (25–27). Preliminary studies suggested that PD-L1 expression in tumors is often associated with a higher likelihood of response to PD-L1/PD-1 inhibitors (28). However, some patients with low or negative PD-L1 expression could also benefit from PD-L1/PD-1 blockade (26, 29). Our in vivo treatment experiments showed an increased susceptibility to anti–PD-1 in FXR^highPD-L1^low mouse LLC tumors than in mock LLC tumors. Given the relatively intact antitumor immune responses preexisting in mock tumors, the impaired immune milieu, especially the exhausted CD8⁺ T phenotype, is likely to render mouse models more susceptible to anti–PD-1 in FXR-overexpressed LLC tumors, because it is readily reactivated after immune-checkpoint inhibition. In addition to tumor reduction, we detected increased proportions of CD69⁺, IFNγ⁺, or TNFα⁺CD8⁺ T cells in FXR-overexpressed tumors after anti–PD-1 treatments, suggesting a restoration of antitumor immunity. On the other hand, anti–PD-L1 was ineffective in treating FXR^highPD-L1^low mouse LLC tumors, perhaps due to the absence of targetable PD-L1 on tumor cells. The tumor immune microenvironment determines the response to anti–PD-1/anti–PD-L1 immunotherapy (30, 31). Our findings support FXR^high as a sensitive biomarker for anti–PD-1 immunotherapy in PD-L1 low/negative NSCLC patients. Further pilot studies of FXR^highPD-L1^low NSCLC specimens treated with anti–PD-1 are needed to verify the value of FXR^high status in predicting immunotherapy responses when PD-L1 is low. Due to the immunosuppressive status of the TME (characterized by increased CD206 on macrophages, decreased MHC II and CD86 on DCs, and increased Lag-3 on CD8⁺ T cells in FXR-overexpressed LLC tumors), we propose that the FXR^highPD-L1^low NSCLC patients might be cured by other immunotherapies, such as tumor-associated macrophage blockade or DC-based therapy, as previously reported (32, 33). Further studies are needed to analyze this hypothesis.

In summary, we have identified a subgroup of NSCLC patients characterized by FXR^highPD-L1^low tumor cells, in which FXR downregulated PD-L1 expression. Our findings have uncovered an immunosuppressive role for FXR in the FXR^highPD-L1^low NSCLC subtype. Our study provides translational insights into the therapeutic activity in PD-L1^low NSCLC patients treated with anti–PD-1. We suggest FXR^highPD-L1^low be considered as a biomarker for anti–PD-1 immunotherapy in the future.

No potential conflicts of interest were disclosed.

Conception and design: J. Ai, H. Jiang

Development of methodology: X. Liu, S. Xue, B. Chen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): W. You, L. Li, D. Sun, Z. Xia

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): W. You, L. Li, D. Sun, Z. Xia, J. Ai, H. Jiang

Writing, review, and/or revision of the manuscript: W. You, J. Ai, H. Jiang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B. Chen, H. Qin

Study supervision: J. Ai, H. Jiang

This work was supported in part by grants 91629104, 81874314, 81821005, and 81773762 from the National Natural Science Foundation of China and grants XDA12020000, XDA12020103 from the Personalized Medicines - Molecular Signature-based Drug Discovery and Development Strategic Priority Research Program of the Chinese Academy of Sciences.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.