TFEB Mediates Immune Evasion and Resistance to mTOR Inhibition of Renal Cell Carcinoma via Induction of PD-L1

Renal cell carcinoma (RCC) encompasses a heterogeneous group of cancers derived from renal tubular epithelial cells (1). Patients with localized RCC after partial or radical nephrectomy often go on to develop metastatic disease, which requires systemic therapies that are rarely curative (2, 3). The mTOR is a serine/threonine kinase that forms two complexes mTORC1 and mTORC2 (4, 5). These sense the availability of nutrition, growth factors, and energy levels to regulate cell growth, proliferation, and differentiation. Dysregulation of mTOR pathways and mutations of mTOR pathway–related genes are often associated with tumor growth in a variety of cancers (6, 7). Inhibitors of mTOR, everolimus, and temsirolimus that are derived from rapamycin have been approved by the FDA to treat advanced metastatic renal cancers (8, 9). Despite initial excitement of mTOR inhibitors for the treatment of RCC, mTOR inhibitors rarely achieve complete responses and most patients ultimately develop resistance to mTOR inhibitor therapy (10, 11). However, the underlying mechanisms by which the RCC resists mTOR inhibition are elusive.

The transcription factor EB (TFEB) is a member of the microphthalmia family of basic helix-loop-helix-leucine-zipper (bHLH-Zip) transcription factors (MiT family), which binds to the coordinated lysosomal expression and regulation (CLEAR) consensus motif and plays important functions in regulation of lysosome biogenesis, autophagy, and metabolism (12, 13). TFEB can be phosphorylated by mTORC1 and GSK3β at serine 211; serine phosphorylated TFEB interacts with 14-3-3 and is sequestered within cytoplasm (14, 15). Mutation of serine 211 to alanine (S211A) results in a loss of phosphorylation of TFEB by mTORC1 and leads to its constitutively nuclear localization (14). TFEB can bind to the Tfeb promoter and induces its own expression (16). The levels of the TFEB and its activity are elevated in multiple types of human cancers and associated with proliferation and motility of cancer cells (17). Furthermore, TFEB fusion and overexpression caused by chromosomal translocation events are linked with a poor prognosis in a subset of patients with RCC with elevated recurrence and metastasis (18, 19).

Cancer cells are subject to immune surveillance by which cytotoxic T cells are able to recognize and kill tumor cells. RCC tumors are characterized by an inflammatory infiltrate and the T-cell growth cytokine IL2 has long been used to treat RCC (20). To evade the immune system, tumor cells express coinhibitory molecules, of which PD-L1 has attracted great interest (21). PD-L1 expressed by the tumor cells binds to the inhibitory coreceptor PD-1 on the surface of tumor-infiltrating T cells to suppress T-cell function. Blockade of the PD-L1/PD-1 axis has been of benefit in the treatment of many different types of cancers (21, 22). Currently, there are about 20 clinical trials for RCC at different stages targeting PD-1, PD-L1, or both (23).

It has been shown that PD-L1 expression in tumor cells is regulated by many transcriptional factors including HIF-1α, STAT3, and NF-κB (24). Given the critical role of mTOR in the cross-regulation of tumor cells and functions of immune cells, in combination with the fact that TFEB is associated with RCC, we speculated that RCC might acquire the resistance to mTOR inhibition by upregulation of PD-L1 expression to promote immune evasion via TFEB.

In this study, we found that TFEB directly regulates PD-L1 expression in RCC cell lines and primary human RCC cells, despite no effect on tumor cell biology. Overexpression of a constitutively active form of TFEB (S211A) in RCC promotes tumor growth in immune competent but not nude mice by inhibiting antineoplastic CD8⁺ T-cell function. Inhibition of mTOR enhances TFEB nuclear translocation and PD-L1 expression in RCC cell lines and human primary renal cancers. Combination of mTORi with anti-PD-L1 enhances the therapeutic efficacy in a mouse RCC xenograft model. Thus, our data provide evidence and rationale to support the combination of mTORi and PD-L1 blockade as a potential therapeutic approach to treat RCC.

Cells were cultured at 37°C and 5% CO₂ in a humidified incubator. 786-O, ACHN, H1975, A549, H2126, HCT116, LoVo, SW480, and Renca cells were from ATCC; 769-P, OS-RC-2, and Caki-1 cells were from and authenticated by Cell Repository, Chinese Academy of Sciences (Shanghai, China). 786-O, ACHN, OS-RC-2, 769-P, and Caki-1 cells were authenticated by STR profiling and authentication of other cell lines was not routinely performed. All of the cell lines were passaged less than 2 months after each thaw and tested to be free of mycoplasma contamination (D101, Vazyme).

769-P, HEK 293T, and Renca cells were transfected with either pcDNA3.1 control plasmid (EV) or pcDNA3.1 encoding the constitutively active form of TFEB S211A (TFEB) using lipofectamine 2000 (11668019, Invitrogen) as described by the manufacturer. TFEB mutant of TFEB-S211A was generated by site-directed point mutagenesis using MutanBEST kit (D401, Takara). Transfected 769-P and Renca cells were selected with G418 (A2513, APExBIO) for 3 weeks and subcloned by dilution at 1 cell/well in 96-well microliter plates. 786-O cells were transduced with control lentivirus expressing a scrambled shRNA or lentiviral particles encoding two TFEB short hairpin RNAs (shRNAs) as follows. ShTFEB-1 (forward): 5′-CCGGCCCACTTTGGTGCTAATAGCTCTCGAGAGCTATTAGCACCAAAGTGGGTTTTTG-3′, ShTFEB-1 (reverse): 5′-AATTCAAAAACCCACTTTGGTGCT AATAGCTCTCGAGAGCTATTAGCACCAAAGTGGG-3′. ShTFEB-2 (forward): 5′-CCGGCGATGTCCTTGGCTACATCAACTCGAGTTGATGTAGCCAAGGACATCGTTTTTG-3′, ShTFEB-2 (reverse): 5′-AATTCAAAAACGATGTCCTTGG CTACATCAACTCGAGTTGATGTAGCCAAGGACATCG-3′. Cells were selected with puromycin (HY-B1743, MCE) for 3 weeks and subcloned by dilution at 1 cell/well in 96-well microtiter plates.

Immunoblotting was performed according to standard method with primary antibodies against TFEB (ab2636, Abcam), TFE3 (SAB4503154, Sigma), PD-L1 (17952-1-AP, ProteinTech), GAPDH (5174, Cell Signaling Technology), PCNA (AV03018, Sigma), BAX (2772, Cell Signaling Technology), phospho-S6 (4851, Cell Signaling Technology), phospho-4EBP1 (2855, Cell Signaling Technology), Histone H3 (17168-1-AP, ProteinTech), HIF-1α (14179S, Cell Signaling Technology), p65 (8242, Cell Signaling Technology), phospho-p65 (3033, Cell Signaling Technology), STAT3 (12640, Cell Signaling Technology), and phospho-STAT3 (9145, Cell Signaling Technology), followed by incubation with appropriate secondary horseradish peroxidase–conjugated antibodies, then evaluated with ECL (RPN2232, GE Healthcare).

The different experimental groups of 786-O and 769-P cells were plated in 96-well plates at 1 × 10³ cells per well and cultured for 5 days or 7 days, respectively. Cell proliferation was determined by Cell Counting Kit-8 (CK04, Dojindo) every 24 hours according to the manufacturer's instructions. The incorporation of 5-ethynyl-2′-deoxyuridine (EdU) was stained with the EdU Cell Proliferation Assay Kit (Q10310-1, RiboBio) according to the manufacturer's protocol. The percentage of EdU⁺ for each field of view captured was recorded and quantified.

786-O and 769-P cells were pretreated with DMSO or paclitaxel (10 μmol/L) for 12 hours and then digested with 0.25% Trypsin (25200-114, Invitrogen) and washed twice with ice-cold PBS. Samples were stained with Annexin V-FITC (640906, BioLegend) and propidium iodide (P4170, Sigma) in Annexin V binding buffer according to the manufacturer's specifications. Samples were analyzed on a BD Verse flow cytometer and data were analyzed using FlowJo software.

The transwell migration assay was performed with standard method. In brief, the bottom chambers were filled with 600 μL of DMEM medium containing 10% FBS. In total, 2.5 × 10⁴ cells were suspended in 100 μL of DMEM containing 1% FBS and seeded in the top chamber. After 24 hours, nonmigrated cells were removed and migrated cells were fixed with 4% paraformaldehyde and stained with 1% crystal violet. Images were taken using Olympus model IX83 inverted fluorescence microscope and the percentages of migrated cell areas in the total areas were quantified.

Wild-type BALB/c and BALB/c nude mice (7–8 weeks of age, Huafukang) were housed in a specific-pathogen-free facility at the Tongji Medical College, HUST. All experiments were performed according to the guidelines of the Institutional Animal Care and Use Committee of Tongji Medical College, HUST.

In total, 1 × 10⁶ Renca cells that were stably transfected with either an empty vector or TFEB-S211A pcDNA3.1 plasmids were injected subcutaneously on the back of WT BALB/c mice. For the experiments performed with nude mice, 5 × 10⁵ (EV or TFEB) Renca cells were injected. For the combinational therapy, WT BALB/c mice were subcutaneously injected with 1 × 10⁶ Renca cells. Once tumor volumes reached 50 to 100 mm³, mice were injected intraperitoneally with either 10 mg/kg temsirolimus daily, 200-μg anti-mouse PD-L1 antibody (10F.9G2, BioXCell) every 3 days, combination of both, or vehicle together with control rat IgG2b (LTF-2, BioXCell). Tumor volumes were measured along major axis (a) and minor axis (b) daily and were calculated using the formula: V = ab²/2. Mice were sacrificed and tumors were excised and weighted.

For the analysis of PD-L1 and PD-L2 expression in human tumor cell lines or primary RCC cells, cells were harvested under normal condition or after administration of rapamycin (A8167, APExBIO), Torin-1 (A8312, APExBIO), and EBSS (E2888, Sigma) for the indicated times. Cells were stained with antibodies against phycoerythrin (PE)-conjugated anti-human PD-L1 antibody (329706, BioLegend) and APC-conjugated antihuman PD-L2 antibody (345507, BioLegend). Mouse RCC cells were stained with PE-conjugated antimouse PD-L1 antibody (124308, BioLegend) or FITC-conjugated antihuman/mouse Ki67 antibody (11-5698-82, Invitrogen). The antibodies used to stain tumor-infiltrating lymphocytes (TIL) were listed as followed: anti-CD45-APC/Cy7 (103115, BioLegend), anti-CD8-Percp/Cy5.5 (100734, BioLegend), anti-CD107a-APC (121613, BioLegend), anti-GZMB-FITC (515403, BioLegend), anti-IL2-BV421 (503825, BioLegend), anti-TNFα-PE/Cy7 (557644, BD), and anti-IFNγ-APC (554413, BD Biosciences). Samples were collected on a BD Verse Flow cytometer and data were analyzed using FlowJo software.

RCC tissue specimens were isolated after surgery, formalin fixed and paraffin embedded, and stained with hematoxylin and eosin. IHC was performed as standard protocol with antibodies against TFEB (ab2636, Abcam), PD-L1 (clone 22c3, Dako), and carbonic anhydrase IX (CAIX) (TA336805, ZSGB-BIO, Beijing). Slides containing both PD-L1 positive and negative areas were taken for analysis of TFEB expression. Intensities of PD-L1 were determined according to the clinical score guideline for PD-L1 staining (clone 22c3, Dako). The 42 cases were divided into PD-L1⁻ group (<1%), PD-L1^low group (1%–49%), and PD-L1^high group (≥50%) with (≥1+) PD-L1 cellular and membrane staining of viable tumor cells. The tumor grades were assigned to the highest (≥5%) within the tumors if the tumors were heterogeneous. IHC reactivity of cytoplasmic or nuclear TFEB was scored as follows: multiplication of the intensity of immunostaining (1, weak; 2, moderate; and 3, strong) and the percentage of positive tumor cells, which resulted in a score of 0 to 300. The total TFEB expression was evaluated as the mean of cytoplasmic and nuclear score. A score of less than 10 was considered as 0, a score of 10 to 40 was considered as 1+, 41 to 140 as 2+, and 141 to 300 as 3+. IHC data were evaluated by two independent pathologists.

Studies with human RCC specimens have been approved by the Ethics Committee of Tongji Hospital of HUST (Wuhan, China), and signed informed consents were obtained from all patients' family. The demographic and clinical characteristics of the enrolled patients are presented in Supplementary Table S1.

Cells were harvested followed by cross-linking for 10 minutes with 1% (vol/vol) formaldehyde. Afterward, cells were lysed by sonication. The cell lysates were immunoprecipitated with anti-TFEB (ab2636, Abcam) overnight at 4°C. After washing and elution, cross-links were reversed for 4 hours at 65°C. The eluted DNA was purified and analyzed by qPCR using a Bio-Rad SYBR Green intercalating fluorophore system with PD-L1 primers: (forward): 5′-AGTTTATGTGGC TGTGGGCA-3′; PD-L1 primers: (reverse): 5′-GGATATTTGCTGTCTTTATATTC -3′. The C_t value of each sample was normalized to corresponding input value.

The PD-L1 promoter sequence (−281 bp to +43 bp) relative to the transcription start site was amplified by PCR from human peripheral blood mononuclear cells and inserted into the pGL3-basic vector (E1751, Promega). The primers used for cloning the PD-L1 promoter are: forward: 5′-CGGCTAGCTGGGCAGATTTTTTTC-3′ and reverse: 5′-ATCTCGAGG CAAATGCCAGTAGG-3′. HEK 293T cell was cotransfected with pRL-TK (E2241, Promega), pGL3-PD-L1 or pGL3-basic, empty pcDNA3.1 vector, or TFEB-S211A pcDNA3.1 plasmids in 24-well plates with Lipofectamine 2000. After 48 hours, firefly and Renilla luciferase activities were measured using the Dual-Luciferase Reporter Assay Kit (E1901, Promega) with microplate reader (Synergy H1, Bio-Tek) and the ratio of firefly/Renilla luciferase was determined.

Tumor specimens were gently minced into small pieces and then digested with 6 mL PBS containing 50 μL 25 mg/mL collagenase IV (17104019, Invitrogen) and 25 μL 10 mg/mL DNase I (10104159001, Roche) for 1 hour at 37°C. Cell suspensions were filtered twice and centrifuged at 1,500 rpm for 5 minutes. Tumor cells and TILs were enriched and harvested separately by Percoll gradient (17-0891-01, GE Healthcare) following the manufacturer's protocol.

Primary RCC cells were seeded on glass slides and incubated overnight for proper attachment. Cells were stimulated with vehicle or Torin-1 (500 nmol/L) for 3 hours and then washed three times with PBS and fixed with 4% paraformaldehyde for 30 minutes, permeabilized with 0.05% Triton X-100 for 30 minutes, and blocked in 5% BSA for 1 hour. Cells were incubated with anti-TFEB (ab2636, Abcam) and anti-PD-L1 (clone 22c3, Dako) overnight at 4°C. Secondary antibodies labeled with FITC or Cy3 (Life Technologies or Jackson Laboratories) were added for 1 hour at room temperature, and DAPI was used for nuclear counterstaining for 10 minutes. Samples were imaged with an Leica TCS SP5 confocal microscope in 24 hours after mounting.

Statistical analysis was performed using Prism (GraphPad, San Diego) software. Statistical significance was determined by Student t test or for variances by ANOVA. P values less than 0.05 were considered significant.

TFEB expression has been linked with both occurrence and a poor prognosis in RCC (19). To determine whether TFEB affects the proliferative, apoptotic, and metastatic capacities of RCCs, we took the advantage of the differential expression of TFEB in human 786-O and 769-P RCC cell lines (Fig. 1A). Knockdown of TFEB in 786-O cells did not affect expression of PCNA, a marker for cell proliferation, as well as proapoptotic protein BAX expression (Fig. 1B). Consistent with these, knockdown of TFEB expression did not affect cell proliferation (Fig. 1C). These results were further confirmed with EdU staining (Fig. 1D and E). Next, we looked at cell survival; knockdown of TFEB in 786-O cells had no effect on apoptosis in untreated cells or cells treated with paclitaxel (Fig. 1F and G). Finally, using an in vitro transwell assay, we found that downregulation of TFEB in 786-O cells had no impact on cell migration compared with cells transduced with scrambled shRNA lentivirus (Fig. 1H and I).

To confirm these conclusions, we overexpressed a constitutive active form of TFEB mutant, TFEB-S211A, in 769-P cells. Consistent with the knockdown data, enhanced TFEB expression did not affect 769-P cell proliferation, apoptosis, and migration in vitro (Supplementary Fig. S1A–S1E).

We next explored the function of TFEB in an in vivo mouse model of RCC. To this end, we first generated a mouse-derived RCC Renca cell line that overexpressed TFEB-S211A mutant (Fig. 2A). Then we hypodermically inoculated control Renca-EV cells or Renca-TFEB (S211A) cells to wild-type BALB/c mice. We found that mice that received cells overexpressing TFEB-S211A showed significantly greater tumor burdens compared with control group (Fig. 2B). Greater tumor burdens were associated with significantly reduced frequencies of CD107a, GZMB, IL2, and TNFα-producing CD8⁺ cytotoxic cells (CTL) within tumors from mice that received Renca cells overexpressing TFEB (Fig. 2C–F), compared with mice that received control cells. There was a similar reduction in the percentages of IFN γ-expressing CTL in TFEB-S211A Renca tumors, but this was not significant (Supplementary Fig. S2A). In contrast, the percentages of CD107a⁺ NK cells and presence of tumor-associated macrophages and myeloid-derived suppressive cells within TFEB-S211A Renca tumors were unaltered compared with control Renca tumors (data not shown).

To resolve our contrasting findings between in vitro and in vivo experiments, we repeated the same experiment with BALB/c nude mice, which lack T cells. In contrast with the experiments performed on wild-type BALB/c host animals, there was no significant difference in tumor growth (Supplementary Fig. S2B and S2C). Moreover, the percentage of Annexin V⁺ as well as Ki-67⁺ tumor cells in TFEB-S211A groups isolated from nude mice was comparable to the control group (Supplementary Fig. S2D and S2E), suggesting that the effect of TFEB is dependent on the presence of tumor-infiltrating T cells rather than intrinsic differences within the cells themselves.

This led us to investigate the effect of TFEB on expression of inhibitory coreceptor ligands. We found that forced expression of TFEB in Renca cells enhanced PD-L1 expression within tumor tissues (Fig. 2G and H). In addition, the PD-L1 expression levels within tumors were positively associated with tumor weights (Fig. 2I). Taken together, TFEB mediated immune evasion of RCC via suppressing the cytotoxic function of CD8⁺ T cells.

We next evaluated whether TFEB correlated with PD-L1 expression in primary RCC cells from patients. Within individual tumors, PD-L1 staining showed heterogeneous expression, which can be readily differentiated into PD-L1⁻ and PD-L1⁺ areas. The PD-L1⁺ regions had higher expression and enhanced nuclear localizations of TFEB (Fig. 3A and B); the representative images of differential intensities of cytoplasmic or nuclear TFEB staining were shown in Supplementary Fig. S3A. To explore possible link between PD-L1 expression and tumor progression, we extended the study to a cohort of 42 patients with clear cell renal cell carcinoma (ccRCC). On the basis of the percentages of PD-L1 expression in viable tumor cells, patients are divided into three groups: PD-L1⁻ (<1%), PD-L1^low (1%–49%), and PD-L1^high groups (≥50%). PD-L1 high-expression patients had higher tumor grades and more advanced tumor stages (Fig. 3C and D). Next, we compared surface PD-L1 expression in a panel of human RCC cell lines. Although ACHN cells had the highest expression of PD-L1, which is followed by 786-O, Caki-1, and OS-RC-2 cells, 769-P cells had the lowest surface PD-L1 expression (Fig. 3E). We obtained similar results for PD-L1 expression using immunoblotting analysis; furthermore, expression of TFEB was positively correlated with the levels of PD-L1 (Fig. 3F). In contrast, expression of TFE3, another member of MITF, was not associated with PD-L1 expression in RCC cell lines (Fig. 3F; Supplementary Fig. S3B).

Knockdown of TFEB expression led to reduced PD-L1 expression in 786-O cells (Fig. 3G and H), which was accompanied with reduced TFEB binding to the PD-L1 promoter (Fig. 3I). Conversely, overexpression of TFEB in the 769-P cells significantly enhanced TFEB binding to the PD-L1 promoter and PD-L1 expression (Supplementary Fig. S3C and S3D). Furthermore, transient expression of TFEB significantly enhanced luciferase activity driven by the PD-L1 promoter (Fig. 3J). Together, these findings demonstrate that TFEB directly binds to the PD-L1 promoter and positively regulates PD-L1 expression.

Given that TFEB is a major target of mTORC1 (14), we asked whether inhibition of mTOR could lead to changed expression of PD-L1. Inhibition of mTOR pathway, either directly by rapamycin and Torin-1 or indirectly by cell starvation, significantly enhanced TFEB expression and its nuclear accumulation in 786-O cells (Fig. 4A). Furthermore, this was associated with enhanced PD-L1 expression (Fig. 4A and B). HIF-1α, STAT3, and p65 have all been implicated as regulators of PD-L1 expression. In our hands, inhibition of mTOR did not affect the phosphorylation of STAT3 and p65 but significantly reduced HIF-1α expression (Supplementary Fig. S4A and S4B). We found similar results in 769-P cells (Fig. 4C and D; Supplementary Fig. S4C and S4D). The negative regulation of mTOR on PD-L1 seemed to be RCC specific, as inhibition of mTOR resulted in reduced expression of PD-L1 in lung carcinoma H1975 cells (Supplementary Fig. S4E) and no change of expression in A549 and H2126 lung carcinoma cells, as well as HCT116, LoVo, and SW480 colon cancer cells (Supplementary Fig. S4E and S4F).

Torin-1 treatment enhanced TFEB binding to the PD-L1 promoter in both 786-O and 769-P cells, compared with control groups (Fig. 4E). To further substantiate that PD-L1 induction by mTOR inhibition was via activation of TFEB, we knocked down TFEB expression by shRNA in 786-O cells and looked at the effect of mTOR inhibition on PD-L1 expression. Knockdown of TFEB abolished the enhanced PD-L1 expression upon inhibition of mTOR by rapamycin, Torin-1, or starvation (Fig. 4F–H).

Together, these data demonstrate that inhibition of mTOR specifically promoted PD-L1 expression in RCC via TFEB activation.

Next, we asked if the inhibition of mTOR enhanced PD-L1 expression in human patients with renal cancer. To this end, we isolated primary renal cancer cells from freshly surgically removed tumor tissues. IHC staining showed that tumor tissues had a CAIX positive staining, a transmembrane protein and marker for ccRCC (Fig. 5A). The isolated primary ccRCC cells from patients were almost all CAIX-positive, indicating the purity of primary cells (Fig. 5B). Torin-1 induced rapid drop in the cytoplasmic concentrations of TFEB at 30 minutes, followed by a slower increase in the nuclear concentration of TFEB (Fig. 5C). In line with this, prolonged mTOR inhibition led to enhanced PD-L1 expression measured by immunoblot analysis and flow cytometry (Fig. 5D and E). In contrast, PD-L2 expression was not affected by mTOR inhibition (Fig. 5F). These results were confirmed using immunofluorescence staining of TFEB and PD-L1 (Fig. 5G). Together, these data demonstrate that mTOR inhibition induces PD-L1 expression via TFEB activation in human primary renal cancer cells.

We then asked whether combination of antibody against PD-L1 could potentiate the efficacy of mTOR inhibition by temsirolimus on ccRCC growth in a xenograft mouse model. When tumor volume reached 50 mm³, mice were treated with either temsirolimus (10 mg/kg, i.p.) daily, anti-PD-L1 (200 μg per mouse, i.p.) four times over 12 days, combination of both, or vehicle plus control IgG (Fig. 6A). There was a reduction in tumor growth in mice treated with either anti-PD-L1 alone or temsirolimus alone inhibited tumor growth compared with the control group, but this was not significant (Fig. 6B and C). In contrast, the combination of temsirolimus and anti-PD-L1 therapy resulted in a significant reduction in tumor size compared with all other groups and complete disappearance of tumors in 3 mice after 23 days (Fig. 6B and C). Temsirolimus suppressed mTORC1 activation in tumor cells as measured by ribosome S6 phosphorylation and enhanced TFEB and PD-L1 expression within tumor tissues (Fig. 6D and E; Supplementary Fig. S5A), compared with control group, indicating the in vivo significance of TFEB–PD-L1 axis during tumor growth. Although temsirolimus did not significantly inhibit tumor cell proliferation assessed by Ki67 IHC staining, anti-PD-L1 led to a small but significant reduction in Ki67 staining and combination of PD-L1 and temsirolimus had a synergistic effect on suppression of Ki67 staining (Fig. 6E; Supplementary Fig. S5B).

Next, we tested the effect of mTOR and PD-L1 inhibition on cytotoxicity in tumor-infiltrating CD8⁺ T cells. Temsirolimus treatment suppressed CD107a and GZMB expression in CTL; anti-PD-L1 treatment had little effect but the combination of anti-PD-L1 and temsirolimus significantly enhanced their expression (Fig. 6F). The combination of the two significantly enhanced IFN γ expression compared with the CTL from untreated animals, although IFNγ was not inhibited by temsirolimus treatment alone and was enhanced by anti-PD-L1 alone (Supplementary Fig. S5C). Together, these data demonstrate that only the combination of inhibiting mTOR and PD-L1 signaling significantly induced the expression of all three markers of T-cell cytotoxicity.

Identifying the mechanism by which tumors are resistant to targeted mTOR inhibitors has largely focused on analysis of intracellular signaling pathways with limited focus on the immune microenvironment of the tumors (25, 26). In this study, we demonstrated that inhibition of mTOR in RCC cell lines and human primary RCC cells leads to enhanced translocation and expression of TFEB, which subsequently induces PD-L1 expression. Furthermore, combination of mTOR inhibition and anti-PD-L1 enhanced the cytotoxic functions of tumor-infiltrating CTL and therapeutic efficacy in a mouse RCC xenograft model.

TFEB has been linked with many aspects of cellular events including proliferation, metabolism, and autophagy (12, 13). Yet in our hands, alteration of TFEB expression has little intrinsic effects on in vitro RCC tumor biology including cell proliferation, survival, and migration. Only in the presence of T cells did we see the effect of TFEB. Identification of the TFEB target genes in RCC cells may provide better understanding of the functions of TFEB in regulation of RCC tumorigenesis and the interaction between RCC cells and the immune microenvironment.

PD-L1 expression in cancer cells is regulated by a variety of transcriptional factors including HIF-1α, NF-κB, STAT1, and STAT3 (24). Our studies revealed a strong association between PD-L1 protein and TFEB expression in RCC cells, in which TFEB directly binds to the PD-L1 promoter. Of note, the induction of PD-L1 in RCCs was irrespective of the VHL status, as the tested RCCs contain both VHL-negative and VHL-positive cells (27). Consistent with this, PD-L1 expression was enhanced concurrent with enhanced TFEB expression, despite reduced HIF-1α expression in both 786-O and 769-P cells upon mTOR inhibition. Furthermore, knockdown of TFEB in RCC rendered the cells less responsive to PD-L1 induction upon mTOR inhibition, indicating a critical role of TFEB in regulation of PD-L1 expression in RCC cells.

The induction of TFEB and PD-L1 by mTOR inhibition seems to be RCC specific, as PD-L1 expression in colon cancer and lung carcinoma cells was not enhanced by mTOR inhibitors. The activity of TFEB is tightly regulated by protein phosphorylation at multiple serine sites, which can be mTOR-dependent (S122) or -independent (S138 and S134; refs. 28, 29). It is conceivable that the activity of TFEB is independent of mTOR regulation in non-RCCs. In line with this, the phosphorylation and nuclear translocation of TFEB can be regulated by GSK3β in breast cancer cells (15). A PARP inhibitor that inactivated GSK3β in breast cancer cells can enhance PD-L1 expression and cancer-associated immunosuppression (30). It is tempting to speculate that PARPi induces PD-L1 expression via activation of TFEB.

In contrast to the positive role of PI3K-AKT-mTOR in the regulation of PD-L1 expression in lung carcinoma cells (31), mTOR inhibition led to enhanced PD-L1 expression in RCC cells in our hands, which is consistent with a previous study (32). PI3K-AKT-mTOR can regulate PD-L1 expression following growth factors or inflammatory stimuli in both an IFN γ-dependent and -independent manner in non–small cell lung cancer (NSCLC), glioma, breast cancer, and melanoma cells (24). Together, these data highlight the contextual roles of PI3K-AKT-mTOR in regulation of PD-L1 expression in tumor cells.

Although checkpoint and mTOR inhibitors have been successful as cancer therapies, as monotherapies these drugs seem to be insufficient to fully block cancer progression (33, 34). Consistent with our findings, targeting mTOR and PD-1/PD-L1 axis simultaneously has improved efficacies in treatment of oral cavity cancer and hepatocellular carcinoma (35, 36). Although the enhanced tumor control with combination of mTOR and PD-L1 targeting depends on CTL but not NK cells (35), some of the mechanisms may be different. mTOR inhibition leads to enhanced MHC-I expression in oral cavity tumors; in HCC, PD-1 promotes tumor growth via enhancing the phosphorylation of 4-EBP1 and ribosomal protein S6 (36).

In summary, our data demonstrated that TFEB mediates PD-L1 upregulation by mTOR inhibitors, which can attenuate mTORi therapeutic efficacy via tumor-associated immune suppression. These data provide strong scientific rationale for the combination of mTOR-targeted therapy and anti-PD-L1 immunotherapy, which may benefit patients with RCC.

No potential conflicts of interest were disclosed.

Conception and design: C. Zhang, Y. Duan, J. Wu, X-P. Yang

Development of methodology: C. Zhang, Y. Duan, M. Xia, Y. Dong, Y. Chen, L. Zheng, X-P. Yang

Acquisition of data (provided animals, acquired and managed patients, provided facilities etc.): S. Chai, Q. Zhang, Z. Wei, N. Liu, J. Wang, C. Sun, Z. Tang, F. Zheng, G. Wang, B. Li

Analysis and interpretation of data (e.g. statistical analysis, biostatistics, computational analysis): C. Zhang, Y. Duan, M. Xia, X-P. Yang

Writing, review, and/or revision of the manuscript: C. Zhang, Y. Duan, M. Xia, A. Laurence, X-P. Yang, X. Cheng

Administrative, technical, or material support (i.e. reporting or organizing data, constructing databases): Y. Duan, S. Chai, G. Wang

Study supervision: X-P. Yang

This work was supported by grants from the National Scientific Foundation of China (#31870892, #81671539, and #31470851, to X-P. Yang), 81873870 (to Z. Tang), and Integrated Innovative Team for Major Human Diseases Program of Tongji Medical College, HUST (to X-P. Yang). A. Laurence is supported by the Crohn's & Colitis Foundation of America.

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.