The Circadian Clock Component RORA Increases Immunosurveillance in Melanoma by Inhibiting PD-L1 Expression Open Access

Melanoma, a malignant tumor that originates from melanocytes, accounts for more than 80% of skin cancer deaths (1). Although most patients with primary melanoma are cured by local excision, patients with advanced metastatic melanoma often have a severe prognosis with long-term survival rates less than 10% due to the absence of effective treatment strategies (2). For a long time, treatments for advanced melanoma have been limited to chemotherapy and high-dose interleukin 2, which have minimal effects (3). Notably, over the past 10 years, novel treatment options involving targeted therapy (BRAF or MEK inhibitors) and immunotherapy (anti-CTLA4 or anti-PD-1) have revolutionized melanoma therapy (4, 5). Although immunotherapy has achieved promising outcomes in cancer treatment, complete pathologic response (pCR) rates are only 19% to 25% in melanoma patients treated with PD-1 blockade alone (6), suggesting that the complex immunosuppressive microenvironment of melanoma leads to immunotherapy evasion via unknown mechanisms. Elucidating the underlying mechanisms and developing novel strategies to overcome intrinsic or acquired resistance to immunotherapy are urgently needed.

The circadian clock consists of the core transcription factors that control 100 of genes, regulate 24-hour rhythms of multiple physiological functions and maintain homeostasis in normal cells and tissues (7–10). Clock oscillation is based on transcriptional feedback loops (11). In the negative feedback loop, CLOCK and BMAL1 form a transcriptional activator complex and periodically drive the transcription of period (PER) and cryptochrome (CRY) genes, which dimerize and inhibit the activity of the CLOCK–BMAL1 complex. In the positive loop, retinoic acid–related orphan receptors [ROR; ROR-α (RORA), RORB, and RORC] and REV-ERB maintain their circadian rhythm by competitively binding to the ROR response element promoter site and controlling the rate of BMAL1 transcription, which accelerates when RORs are bound and slows when REV-ERB is bound (12, 13).

However, clock disruption caused by night shift work or chronic jet lag can increase the risk of multiple diseases, including cancer (14–16). Human data from The Cancer Genome Atlas indicate a significant correlation between core clock gene downregulation and oncogenic driver pathways, tumor staging and tumor immune microenvironment in various cancer types (17, 18). Additionally, the aberrant expression of clock genes such as CRY1, PER1, PER2, and RORs has been shown to promote tumor progression (19–23) and regulate cancer-related processes, including cell proliferation (24, 25) and the DNA damage response (26, 27). Furthermore, circadian disruption caused by a light-dark shift in melanoma is associated with immune microenvironment remodeling (28). However, how tumor-derived clock disruption influences antitumor immunity and the precise molecular mechanisms underlying this connection remain unclear.

RORA, a potential tumor suppressor, has been reported to inhibit proliferation and tumorigenesis in glioma cell lines (29). However, its role and mechanisms in anticancer immunity remain unclear. Herein, we report that the clock gene RORA activates antitumor T-cell effects via corepressor complex–mediated PD-L1 inhibition. Hence, our study reveals a novel molecular mechanism for the clock component–mediated antitumor immune response against melanoma, highlighting RORA as a potential combination partner and the combined scores of the RORA complex as predictive biomarkers for melanoma immunotherapy.

C57BL/6 female mice were purchased from Hunan SJA Laboratory Animal Co., Ltd., and bred at the Experimental Animal Center, Central South University. The mice used in the experiments were specific-pathogen-free grade. All animal protocols were approved by the Animal Care and Use Committee of Xiangya Hospital (Central South University, license no. 2022020580). All animal experiments conducted in this study strictly followed the guidelines for experimental pain research in conscious animals to minimize animal suffering and improve animal welfare.

For in vivo studies, 1 × 10⁶ B16F10 vector, RORA-oe and B16F10 cells were injected subcutaneously at 6 to 8 weeks of age. As detailed in the figure legends, 5 to 7 mice were utilized to observe the tumor phenotypic changes, and another five mice were used for recording survival in each treatment group. Tumor size was measured every day for approximately 1 week after injection. At the endpoint, the mice were euthanized, and the tumors were harvested for weight measurement and further FACS analysis.

Blood biochemical analysis: Blood was collected from the tail vein of mice into anticoagulant tubes. Blood samples were collected for complete blood count (CBC, XN-1000-B1) analysis, including white blood cell count, red blood cell count, hemoglobin analysis, and platelet count. Blood was collected from the retroorbital venous plexus of the mice. Hematoxylin and eosin staining of kidney, liver, and spleen tissue was performed and analyzed by Servicebio.

The murine melanoma cancer cell line B16F10 and the human melanoma cell lines A375 and SK-MEL-28 were purchased from ATCC. B16F10 cells were cultured in RPMI 1640 medium (Gibco), whereas other cells were grown in DMEM (Gibco) supplemented with 10% FBS (Gibco) at 37°C and 5% CO₂. These cell lines were routinely tested to confirm they are free of Mycoplasma contamination by using a MycoAlert Mycoplasma Detection Kit (Lonza).

For generation of CRISPR-mediated knockout melanoma cell lines, sgRNAs targeting human RORA and DDX3X were subcloned and inserted into the pLenti-CRISPRV2 vector (Addgene) following the manufacturer’s instructions. The sgRNA sequences used for human RORA, HDAC3, and DDX3X knockout are listed in the key resources table. For the stable transfection of cell lines, a lentivirus-mediated shRNA targeting RORA was purchased from Shanghai Genechem.

The human RORA-ha-oe, DDX3X-gfp-oe, and mouse RORA-oe vectors were constructed by amplifying and cloning the corresponding cDNAs into the pLVX-IRES-Puro-3xFlag and pCDH-CMV-EGFP-MCS-EF1-Puron vectors, which were purchased from Unibio (VT8006 VT8070). cDNAs encoding the N-terminal domain of RORA (amino acids 1–139; amino acids 1–272) and the C-terminal domain of RORA (amino acids 139–523), as well as cDNA encoding the N-terminal domain of DDX3X (amino acids 1–410) and the C-terminal domain of DDX3X (amino acids 132–661; amino acids 410–661), were amplified, subcloned and inserted into the pCDH-CMV-EGFP-MCS-EF1-Puron vector. For transient reporter analysis, the pGL3 2 kb prom. CD274 promoter plasmid was obtained from Addgene (107003, Addgene). Then, 1-kb prom. CD274 cDNA was amplified from the pGL3 2 kb prom. CD274 promoter plasmid and cloned and inserted into the pGL3-Basic vector, and the pGL3 2-kb prom. CD274 mutant plasmid with a specific mutation in the putative binding site was generated via PCR and verified by sequencing. The sequences are listed in the key resources table.

For packaging of lentiviruses, HEK293T cells at 60% to 80% confluence was transfected using Turbofect (Thermo Fisher Scientific) in DMEM (Gibco) following the manufacturer’s instructions. After 48 and 72 hours of transfection, the virus-containing supernatant was collected and centrifuged at 2,000 rpm for 10 minutes. Cells (30%–50% confluency) were incubated in medium containing optimal dilutions of lentiviruses mixed with polybrene (5–10 µg/mL). The medium containing lentiviruses was changed every 24 hours posttransduction. After 48 hours of transfection, the cells were lysed or subjected to puromycin selection (1 µg/mL) for 3 days to obtain stably transfected cells. Finally, the transfection efficiency was determined by measuring the protein and mRNA levels after cell collection.

A total of 1,800 cells were seeded in a 96-well plate and treated with or without 50-µmol/L nobiletin for 4, 24, 48, and 72 hours. Then, 10 μL of Cell Counting Kit-8 solution was added to the cells of the plate with fresh medium. After 1 hour of incubation, cell viability was determined by measuring the absorbance at 450 nm.

Cells were seeded in six-well plates (500 or 2,000 cells per well) in growth medium for 1 week and, where applicable, subjected to nobiletin treatment for 3 days. When the colonies reached a considerable size, the cells were fixed in 4% paraformaldehyde for 10 minutes, followed by staining with crystal violet solution (Sigma–Aldrich) for another 15 minutes. The colonies were quantified by a spectrometer at an optical density of 570 nm after being dissolved in 30% glacial acetic acid.

Melanoma cells with RORA knockdown or overexpression were seeded in 12-well plates at 60% confluence and then transfected with Turbofect (Thermo Fisher Scientific) according to the manufacturer’s instructions. In detail, 1 µg of each experimental plasmid (pGL3 basic vector, pGL3-CD274 1 kb/2 kb or pGL3-CD274-RORA site 3/site 4 mutation vector) and 20 ng of the Renilla pRL-SV40P plasmid were used for normalization per well. After 24 hours in culture, the cells were collected and split, and the relative luciferase units (RLU) were measured using the Dual-Luciferase Reporter Assay System (Promega) according to the manufacturer’s instructions. RLUs from the firefly luciferase signal were normalized to RLUs from the Renilla signal.

Cells lysates were prepared with lysis buffer (2% SDS) supplemented with protease inhibitor. Lysates with 10 ng of protein were subjected to SDS-PAGE and subsequently transferred onto polyvinylidene difluoride membranes (Millipore). After the membranes were blocked in 3% to 5% nonfat milk in Dulbecco’s PBS plus  0.1% v/v polysorbate 80 for 1 hour at room temperature, they were incubated with the following primary antibodies at a dilution of 1:1,000 at 4°C overnight: anti-RORA, anti-PD-L1, anti-DDX3X, anti-HDAC3, anti-HA, anti-GFP, anti-CD276, anti-VISIA, anti-CD112, anti-CD155, anti-CD86, and anti-CD40. After incubation with peroxidase-conjugated secondary antibody at a dilution of 1:3,000 for 2 hours at room temperature, the blots were visualized by a chemical chemiluminescence imaging system. ImageLab (Bio-Rad) was used to process the images. For coimmunoprecipitation (co-IP), cell lysates (1 mg) were mixed with 2 mg of anti-HA, anti-GFP, anti-RORA, or anti-DDX3X at 4°C overnight and then pulled down with protein A/G magnetic beads (Biolinkedin) at room temperature for 2 hours. IP beads were then washed with lysis buffer three times, and the binding proteins were eluted with SDS loading buffer at 100°C for 5 minutes for SDS-PAGE and subsequent immunoblot analysis.

Human melanoma cells were collected using ethylenediaminetetraacetic acid and washed three times with cold Dulbecco's PBS. Fractionation was performed using the NE-PER Nuclear and Cytoplasmic Extraction Reagents (Thermo Fisher Scientific) according to the manufacturer’s manual.

For identification of RORA-binding proteins, human melanoma cells transfected with the HA-tagged RORA plasmid were lysed with IP buffer containing a protease inhibitor cocktail. Protein lysates were concentrated at 13,800 rpm for 15 minutes at 4°C, and the supernatant was transferred to a new tube for use. Lysates (1 mg) were immunoprecipitated with anti-HA magnetic beads at 4°C for 12 hours. The beads were then washed with IP buffer four times before being heated in SDS loading buffer at 100°C for 10 minutes. Samples were loaded onto SDS-PAGE gels for analysis by immunoblotting.

Total cellular RNA was extracted from the samples using an RNA isolater (Vazyme) and reverse transcribed into cDNA using HiScript II Q RT SuperMix for quantitative RT-PCR (qRT-PCR) analysis (Vazyme) according to the manufacturer’s guidelines. The resulting cDNA was subsequently used as a template for the amplification of target gene transcripts by real-time PCR using SYBR qRT-PCR Master Mix (Vazyme). Information about the primers is listed in the Key resources table.

Chromatin immunoprecipitation (ChIP) was performed with an anti-HA antibody in SK-MEL-28 and A375 cells stably expressing HA-tagged RORA using a ChIP kit (CST) according to the manufacturer’s protocol. Quantitative PCR analysis of the immunoprecipitated DNA was performed using SYBR qRT-PCR Master Mix (Vazyme). The human RORA primers used are listed in the Key resources table.

For activated CD8⁺ T cells, mononuclear cells were isolated from human peripheral blood according to the manufacturer’s protocol and cultured in CTS AIM V SFM (A3021002, Gibco) containing ImmunoCult Human CD3/CD28/CD2 T-cell activator (STEMCELL Technologies) and recombinant human IL2 (1,000 U/mL, R&D) for at least 1 week. Human melanoma cell lines were allowed to adhere to 24-well plates overnight and then incubated in the presence or absence of nobiletin (100 mmol/L) for 12 to 24 hours at 37°C. Subsequently, the cells were cocultured with activated T cells at a 1:2 or 1:3 cancer cell to CD8⁺ T-cell ratio for 24 to 36 hours. The T cells and cell debris were discarded by washing with PBS, and the living cancer cells were then stained with crystal violet and quantified by a spectrometer at an optical density of 570 nm.

For flow cytometric analysis of cancer cells and tumor tissue, single-cell suspensions were prepared from cultured cells and mouse tumors by fine cutting, physical grinding and filter filtration. Single-cell suspensions of spleens were processed as described above. The cells were blocked with an anti-CD16/32 antibody and stained with the following fluorochrome-conjugated antibodies or an isotype for 30 minutes at room temperature: human anti-PD-L1, mouse anti-PD-L1, mouse anti-CD45, mouse anti-CD3, mouse anti-CD8, and mouse anti-CD4. For intracellular staining, after staining for cell-surface markers, single cells were fixed and permeabilized with a Cyto-Fast Fix/Perm Buffer Set (426803, Biolegend) prior to staining with a mouse anti-GZMB antibody at room temperature for 1 to 2 hours. Samples were subjected to FACS with Dxp Athena and Aurora (Cytek), and the data were analyzed using FlowJo analysis software.

For treatment with the RORA agonist nobiletin, nobiletin (25 mg/kg; Topscience) and solvent were administered intraperitoneally daily for 9 days beginning 1 week after subcutaneous inoculation of the wild-type B16F10 cells. For antibody treatments, mice were treated with an anti-CTLA4 antibody (200 µg/per mouse) or an IgG control (BioXCell) at a dose of 200 µg/per mouse via intraperitoneal injection twice a week beginning 1 week after B16F10 cell inoculation. For combination treatments, nobiletin and an anti-CTLA4 antibody were administered at a dose of 25 mg/kg daily or 200 µg/per mouse twice per week for 9 days beginning on day 7 after tumor implantation. For the depletion of CD8⁺ T cells in vivo, mice were intraperitoneally injected with 200 µg of anti-CD8 antibody (BioXCell) on days 7, 10, and 13 after tumor inoculation. One group of mice treated with the IgG isotype (BioXCell) served as controls. Prior to treatment initiation, the mice were randomized into treatment and control groups with similar average tumor volumes.

For immunofluorescence (IF) staining, melanoma patient tissue microarrays and normal skin microarrays purchased from Bioaitech Co., Ltd., were baked for 2 hours at 60°C and then deparaffinized. Antigen was retrieved in ethylenediaminetetraacetic acid antigen retrieval buffer (pH 8.0), and the samples were maintained at a subboiling temperature for 8 minutes, followed by incubation for 8 minutes and then incubation at another subboiling temperature for 7 minutes. After spontaneous fluorescence quenching, the samples were blocked in 3% BSA supplemented with 0.1% Tween-20 in PBS for 60 minutes at room temperature. Each tissue sample was stained with primary antibodies targeting RORA overnight at 4°C. After extensive washing in PBS with 0.1% Tween-20, the secondary antibody Alexa Fluor^®488 donkey was added to the blocking solution and incubated for 120 minutes. After extensive washing in PBS-0.1% Tween-20, the microarrays were coverslipped with antifade mounting medium and kept in the dark after incubation with 4′,6-diamidino-2-phenylindole solution at room temperature for 10 minutes. Images were then obtained by fluorescence microscopy (Nikon, ECLIPSE Ts2R). For quantification of these IF results, the images were further processed by ImageJ.

Melanoma tissues were collected, and written informed consent was obtained from all 177 patients from our dataset. The inclusion of melanoma tissues was approved by the Ethics Committee of our hospital, Central South University. Additionally, we collected melanoma cohorts and immunotherapy cohorts from the GEO and the literature for external validation. The melanoma cohorts included the GSE65904, GSE54467, GSE22153, and GSE100797 datasets. The immunotherapy cohorts included PRJEB23709 (combined with ipilimumab and anti-PD-1 immunotherapy, patients, n = 41), Motzer NatMed 2020, and GSE135222.

The Maxstat package was used to determine the optimal cutoff points for overall survival (OS) and progression-free survival (PFS) in the survival analysis. Then, the log-rank test was used to calculate the significance of the differences. The cancer immune cycle gene set, immune pathway gene set of ImmPort Shared Data, and partial immune pathway gene sets of Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) were collected and analyzed using the single-cell gene set enrichment analysis (GSEA) algorithm. The Microenvironment Cell Populations counter algorithm and Estimation of STromal and Immune cells in MAlignant Tumors using Expression (ESTIMATE) data algorithm were applied to calculate the abundance of immune infiltrating cells and the ESTIMATE data score. For non-normally distributed variables, significant differences between groups were determined by the Wilcoxon test. The reciprocal relationships between proteins were obtained by the mean of rigid protein‒protein docking (ZDOCK). Finally, the protein structure was imported into PyMOL software for visualization.

GraphPad Prism 8.4.3 was used for statistical analysis. The data are presented as the means ± SD or means ± SEM. Student t test was performed to compare two groups of independent samples, and the ordinary one-way ANOVA was used to compare more than two groups of independent samples. A two-sided Wilcoxon rank-sum test was used for IF analysis. Kaplan–Meier survival curves were compared using the log-rank test.

The data anlyzed in this study are publicly available in GEO at GSE65904, GSE54467, GSE22153, GSE100797, GSE135222, in BioProject at PRJEB23709, and in the Motzer NatMed 2020 (30). All plasmids, cell lines, datasets, and other stable reagents generated within the article and its supplementary data files are available from the corresponding author upon request. Our analysis did not involve any special custom codes, and the visualization codes in this study has been uploaded to https://github.com/weijinchen01/Rhythm_analyze.

To explore the relationship between the tumor-derived clock and antitumor immunity, we collected core clock genes and clock control genes identified in previous studies for further analysis (8, 17). Single-sample GSEA was used to calculate a score representing the absolute enrichment degree of the clock gene set in each melanoma sample, and all samples were classified into two groups according to the median of the scores. GSEA revealed that 7 of 25 significantly increased hallmark pathways in the high-score group were associated with the immune response in our melanoma transcriptomic data (Fig. 1A). Moreover, correlation analysis of immune-related genes in the ImmPort database and multiple immune-related pathways in the GO and KEGG databases revealed that clock scores were significantly positively correlated with T-cell activation and B-cell differentiation (Fig. 1B), suggesting that tumor-derived clock genes are strongly related to antitumor immune signaling.

To further investigate the key clock molecule most likely involved in the clock-related antitumor immune response, we separately calculated the differential expression of these genes in melanoma samples compared with healthy skin samples and the effects of each molecule on OS and PFS after immunotherapy. Among them, RORA was the only gene that was significantly downregulated and positively correlated with patient prognosis (Fig. 1C; Supplementary Fig. S1A–S1C). Consistently, the expression of this gene was notably decreased in a panel of cancer cell lines compared with that in normal cells (HEK293T; Supplementary Fig. S1D) and further suppressed with the progression of melanoma (Fig. 1D). The positive correlation between RORA expression and cancer prognosis was also validated in multiple melanoma public datasets and immunotherapy cohorts (Fig. 1E; Supplementary Fig. S1E). To explore the clinical relevance of RORA, we compared its protein levels in human tissue microarrays containing 21 melanoma patient samples and 28 healthy skin tissues by IF and found that tumor samples exhibited weak positive/negative RORA staining, whereas most of the normal skin tissues displayed strong or extrastrong RORA staining (Fig. 1F). These results indicate that RORA may play a critical role in clock-related antitumor immunity in melanoma.

To better understand the role of RORA in the immune response against melanoma, we used the Microenvironment Cell Populations-counter to estimate the proportion of immune cell infiltration in the samples. RORA expression was notably correlated with the abundance of infiltrating immune cells, such as myeloid dendritic cells, T cells, and neutrophils (Fig. 1G). Further analysis revealed that this molecule showed the strongest positive correlation with T-cell recruitment in the cancer immune cycle (Fig. 1H). The cytolytic activity score (CYT) was developed as a quantitative means of assessing cytotoxic T-cell infiltration and activity. Here, RORA expression was prominently positively correlated with CYT (Supplementary Fig. S1F). Taken together, these observations suggest that the clock gene RORA might play an immunostimulatory role in melanoma.

Given the low expression of RORA in melanoma, to investigate its function in tumor growth, we overexpressed RORA in B16F10 murine melanoma cells using a plasmid encoding RORA (RORA-oe) with noncoding constructs as a control (vector). An increase in RORA was confirmed at the mRNA and protein levels (Supplementary Fig. S2A). To investigate the effect of nobiletin in antitumor immunity, we first compared the antitumor effects of RORA in vitro and in vivo. The transfected cells were then incubated subcutaneously in the flanks of immunocompetent C57BL/6 mice. All mice developed comparable-sized tumors. However, the RORA-overexpressing melanoma cells showed more significant tumor regression in vivo than in vitro (Fig. 2A–C; Supplementary Fig. S2B). A previous study revealed that nobiletin (5,6,7,8,3,4-hexamethoxyflavone) directly targets RORs, leading to their activation (31). Similarly, nobiletin treatment activated RORA by suppressing the mRNA expression of BMAL1 (Supplementary Fig. S2C). To further explore whether the tumor-suppressive effect described above was induced by RORA activation, we observed the antitumor effect of nobiletin in vitro and in vivo. Although nobiletin treatment inhibited melanoma cell proliferation and colony formation, immunocompetent C57BL/6 mice injected with B16F10 cells developed smaller tumors, suggesting that the antitumor effect of RORA might be closely related to the immune response (Fig. 2D–F; Supplementary Fig. S2D). RORA overexpression/activation effectively prolonged mouse survival but did not result in a significant change in mouse weight (Fig. 2G; Supplementary Fig. S2E). In addition, the administration of nobiletin did not result in significant changes in routine blood parameters; the tissue structure of the kidney, liver, or spleen of the mice, suggesting the limited toxicity of nobiletin treatment (Supplementary Fig. S2F and S2G).

Bioinformatics analysis revealed a positive correlation between RORA and T-cell recruitment and CYT, so we further analyzed CD4⁺ and CD8⁺ T-cell infiltration in melanoma, and the absolute numbers of the subclasses of tumor infiltrating immune cells are summarized in Supplementary Table S1. Absolute and relative immune cell composition analysis indicated an increase in the number and percentage of GZMB⁺ CD8⁺ T cells that infiltrated in RORA-overexpressing and nobiletin-treated tumors (Fig. 2H; Supplementary Fig. S2H; Supplementary Table S1). Therefore, these studies indicate that RORA regulates the activity and function of tumor infiltrating CD8⁺ T cells in vivo. Furthermore, we overexpressed RORA in SK-MEL-28 and A375 human melanoma cells, which was confirmed by qRT-PCR and immunoblotting analysis (Supplementary Fig. S2I). Additionally, we disrupted RORA in these two cell lines via CRISPR-Cas9-targeted knockout (sgRORA) with nontargeting constructs as a control (sgCont). A reduction in the mRNA expression of RORA was found (Supplementary Fig. S2J). Functionally, T-cell–mediated cytotoxicity assays showed that RORA overexpression or activation by nobiletin increased T-cell–mediated cancer cell death (Fig. 2I and J; Supplementary Fig. S2K), whereas downregulation of RORA reduced T-cell–mediated cytotoxicity in vitro (Fig. 2K; Supplementary Fig. S2L). Taken together, these data indicate that RORA increases cytotoxic CD8⁺ T-cell activity, which might confer tumor-suppressive effects.

To further confirm that T-cell populations mediate this potent rejection, we depleted CD8⁺ T cells in mice inoculated subcutaneously with B16F10 cells overexpressing RORA or treated with nobiletin. The depletion of CD8⁺ T cells in the mouse tumors was confirmed by flow cytometry (Supplementary Fig. S2M). Strikingly, a significant difference in tumor burden between the RORA-oe and vector tumors in C57BL/6 mice was not detected after treatment with a depleting antibody against CD8 (Fig. 2L). Consistently, depletion of CD8⁺ T cells reversed tumor growth inhibition in the mice treated with nobiletin (Fig. 2M). Collectively, these studies indicate that RORA controls CD8⁺ T-cell-dependent tumor rejection.

Because cancer cells exploit immune checkpoints to reduce CD8⁺ T-cell activity (32), we next detected changes in the expression of major immune checkpoint proteins after RORA overexpression via qRT-PCR and immunoblotting. The mRNA and protein levels of PD-L1 showed the most significant decreases, despite changes in the expression of other molecules, such as CD276 and VISTA, at the mRNA level (Fig. 3A and B). PD-L1, a ligand for PD-1 on immune T cells, can be highly induced in tumor cells by IFNγ stimulation, leading to T-cell exhaustion and immune escape (33). To confirm the effect of RORA on PD-L1 expression, we subsequently treated human melanoma cells under IFNγ exposure. Nobiletin-treated cells exhibited significant IFNγ-induced PD-L1 downregulation at both the mRNA and total protein levels (Fig. 3C). In contrast, PD-L1 expression was upregulated by RORA knockout and knockdown (Fig. 3D; Supplementary Fig. S3A). Consistently, surface-PD-L1 expression was notably inhibited by RORA overexpression or nobiletin treatment (Fig. 3E and F) but increased by RORA knockout or knockdown (Fig. 3G; Supplementary Fig. S3B). Pharmacological stimulation of RORs using the agonist SR1078 activates downstream target genes of RORA, whereas the specific antagonist SR3335 functions as a selective inhibitor of RORA (34). Like nobiletin, SR1078 induced concentration-dependent PD-L1 downregulation at both the mRNA and protein levels, whereas SR3335 treatment increased PD-L1 expression (Supplementary Fig. S3C and S3D). In vivo, PD-L1 expression in the tumors treated with nobiletin was much lower than that in the DMSO-treated tumors (Supplementary Fig. S3E). Taken together, these data indicate that RORA is a major negative regulator of PD-L1 expression in melanoma cells.

Finally, to establish a functional link between PD-L1 and in vitro T-cell–mediated cytotoxicity controlled by RORA activation, we disrupted PD-L1 expression in melanoma cells (PD-L1ko; Supplementary Fig. S3F). A reduction in PD-L1 expression led to a substantial increase in T lymphocyte cytotoxicity (Fig. 3H; Supplementary Fig. S3G). Importantly, compared with the control treatment, nobiletin treatment failed to increase T-cell–mediated cancer cell death after PD-L1 knockout (Fig. 3H; Supplementary Fig. S3G). Therefore, RORA activates cytotoxic T lymphocyte activity in vitro mainly by inhibiting PD-L1 expression.

As a transcription factor, RORA drives the expression of hundreds of target genes by recognizing and binding to specific DNA sequences, known as ROR response elements, on their promoters (34, 35). To determine whether RORA is directly responsible for PD-L1 regulation by binding to its promoter, we first constructed plasmids containing 1- and 2-kb PD-L1 promoter sequences (Supplementary Fig. S4A). Using a dual-luciferase reporter assay, we found that PD-L1 transcriptional activity was impaired by RORA overexpression in SK-MEL-28 and A375 cells, with almost the same effect on the activity of the 1 and 2 kb PD-L1 promoters (Fig. 4A). In contrast to RORA overexpression, RORA knockdown increased the transcriptional activity of the PD-L1 promoter (Fig. 4B), suggesting that RORA negatively mediates PD-L1 transcription.

Furthermore, by searching for potential RORA-binding sites in the human PD-L1 promoter through JASPAR core database analysis, we identified four potential binding sites (Fig. 4C). Subsequently, a chromatin immunoprecipitation assay was conducted in RORA-transduced human melanoma cells, and the results indicated that RORA bound to cis-acting element fragment 4 (ATAAAGGTTA, site 4) of the PD-L1 promoter (Fig. 4D). To determine whether the RORA-binding site is a transcriptionally active region, we generated promoter constructs containing mutations in the sites that cause RORA-binding deficiency, and dual-luciferase assays were performed to test the effect of RORA on promoter activity in the SK-MEL-28 and A375 cells transfected with the WT or mutant PD-L1 promoter (Fig. 4E). The results showed that nobiletin treatment significantly suppressed PD-L1 promoter activity, and this effect was significantly reversed by site 4 mutation (Fig. 4F) but not site 3 mutation (Supplementary Fig. S4B). Together, these data verify that RORA binds to a specific region on the PD-L1 promoter and transcriptionally downregulates its expression.

Previous studies have indicated that RORA regulates target gene transcription via interactions with cofactors (36, 37). RORA suppresses the expression of inflammatory genes or the transcriptional activity of peroxisome proliferator-activated receptor-γ by specifically interacting with HDAC3 (38, 39). By reducing histone acetylation and preventing recruitment of the bromodomain protein BRD4 to the promoter of PD-L1, HDAC3 acts as a crucial repressor of PD-L1 transcription in B-cell lymphoma and melanoma (40, 41). Next, we examined whether HDAC3 is required for RORA-mediated PD-L1 inhibition. We first verified the interaction between RORA and HDAC3 by co-IP in the RORA-transduced SK-MEL-28 and A375 cells (Fig. 4G). We generated various RORA deletion mutants, and a deletion mapping study showed that HDAC3 interacted with the C-terminal ligand binding domain (LBD) of RORA (Fig. 4H). Furthermore, by qRT-PCR and immunoblotting analysis, we found that HDAC3 knockout substantially increased PD-L1 expression (Fig. 4I) and partly reversed nobiletin-mediated PD-L1 inhibition at the mRNA and protein levels in human melanoma cells (Fig. 4J). In summary, these data indicate that HDAC3 may be involved in RORA-mediated PD-L1 inhibition by forming a coinhibitory complex with RORA.

Since HDAC3 knockout partly abolished RORA agonist-mediated PD-L1 repression, we speculate that other molecular mechanisms might be involved in the correlation between RORA and PD-L1, which needs to be further investigated. For this purpose, nuclear RORA in SK-MEL-28 and A375 cells with RORA-ha overexpression was separated and immunoprecipitated with an anti-HA antibody, and protein bands at approximately 37, 45, and 70  kDa were observed to be enriched (Fig. 5A; Supplementary Fig. S4C). Intriguingly, subsequent analysis of the bands by mass spectrometry revealed that among the top eight candidate binding proteins, DDX3X, a member of the DEAD-box helicase family, was the most strongly positively associated with PD-L1 expression in our dataset and the public melanoma cohort (GSE100797; Fig. 5B and C; Supplementary Fig. S4D). To investigate the physical link between RORA and DDX3X, using reciprocal co-IP, we first validated the interaction between HA-tagged RORA and GFP-tagged DDX3X in HEK293T cells (Fig. 5D). Next, the interaction between endogenous nuclear RORA and DDX3X was confirmed by nucleoprotein isolation and subsequent reciprocal co-IP in human melanoma cells (Fig. 5E–G).

Moreover, to identify the domains of RORA that are responsible for direct physical interactions with DDX3X, we used various RORA deletion mutants, and a deletion mapping study showed that, similar to HDAC3, DDX3X interacted with the LBD of RORA (Fig. 5H). To further demarcate the structural domain of DDX3X required for binding to RORA, we constructed a plasmid bearing the truncated forms of DDX3X, which contained a deletion of the first 132 amino acids at its N-terminus, the highly conserved helicase core domain 1 (1-441) and domain 2 (441-661). Coimmunoprecipitation showed that the C-terminus of 441-661 was unable to bind to RORA (Fig. 5I), indicating that the highly conserved helicase core domain 1 of DDX3X is essential for its interaction with RORA.

Since both DDX3X and HDAC3 interact with the same domain of RORA (LBD), we speculate that DDX3X and HDAC3 may interact competitively with RORA. To test this hypothesis, we performed proteomic bioinformatics analysis by using the crystal structures of RORA (Protein Data Bank: 1s0x), HDAC3 (Protein Data Bank: 4a69), and DDX3X (Protein Data Bank: 2jgn) for interactive docking prediction. Interestingly, the results predicted by ZDOCK showed that both the ARG-224 residue of HDAC3 and the PHE-429 residue of DDX3X form hydrogen bonds with the same residue of RORA (ASN-350; Fig. 5J; Supplementary Table S2). Furthermore, we observed via a competitive co-IP experiment that the overexpression of GFP-tagged DDX3X promoted the RORA-DDX3X interaction while weakening the binding of HDAC3 to RORA (Fig. 5K), which corroborated the competitive effects of DDX3X. Additionally, using co-IP, we found that HDAC3 and DDX3X did not exist in the same complex (Supplementary Fig. S4E). Functionally, a T-cell–mediated cytotoxicity assay revealed that DDX3X overexpression inhibited T-cell–mediated cancer cell death, and vice versa (Fig. 5L and M; Supplementary Fig. S4F). Taken together, these results suggest that DDX3X competes with HDAC3 for interaction with RORA and disrupts the activity of cytotoxic CD8⁺ T cells.

DDX3X exists in both the cytoplasm and nucleus and acts as a cofactor for transcription factors to synergistically activate the transcription of target genes (42). By investigating the role of DDX3X in PD-L1 transcription via immunoblotting and qRT-PCR analysis, we found that DDX3X overexpression increased the total protein, mRNA and cell-surface levels of PD-L1, whereas DDX3X knockout had the opposite effect on human melanoma cells (Fig. 6A–D). Consistently, dual-luciferase reporter assays showed that DDX3X knockout decreased PD-L1 transcriptional activity, whereas DDX3X overexpression increased PD-L1 transcriptional activity (Fig. 6E). Taken together, these results indicate that DDX3X increases the transcription and expression of PD-L1 in a manner opposite to that of RORA.

Given that DDX3X competes with the corepressor HDAC3 for interaction with RORA but that the two molecules do not exist in a complex and play opposite roles in PD-L1 transcription, we postulated that DDX3X activated PD-L1 expression by competitively binding to RORA and blocking it from the promoter of PD-L1. To this end, by overexpressing DDX3X in SK-MEL-28 and A375 cells, we found that the reduction in PD-L1 expression caused by RORA overexpression or nobiletin treatment was significantly attenuated at the protein and mRNA levels (Fig. 6F and G; Supplementary Fig. S4G). Similarly, the decrease in transcriptional activity induced by nobiletin treatment was reversed when DDX3X was overexpressed (Fig. 6H). Furthermore, to test whether the DDX3X-mediated transcriptional upregulation of PD-L1 depends on interaction with RORA, we overexpressed plasmids bearing the truncated forms of DDX3X containing the N-terminus 1-441 and C-terminus 441-661. The results indicated that 1-441, but not 441-661, was unable to bind to RORA and notably reversed the transcriptional activity of PD-L1 inhibited by nobiletin (Fig. 6I). Functionally, the increase in T-cell-mediated cancer cell death after nobiletin treatment was significantly repressed by DDX3X overexpression (Fig. 6J; Supplementary Fig. S4H). These results suggest that the function of DDX3X in PD-L1 transcriptional regulation is dependent on its interaction with RORA.

Furthermore, ChIP–RT-PCR revealed that RORA bound to the PD-L1 promoter, whereas DDX3X did not (Supplementary Fig. S4I). Importantly, the association of RORA with the PD-L1 promoter was significantly decreased by DDX3X overexpression (Fig. 6K), which corroborated our hypothesis that DDX3X interacts with RORA to prevent it from binding to the promoter of PD-L1. Salt bridges are the strongest noncovalent interactions in nature and are known to be involved in protein folding (43). The binding of DDX3X to RORA obviously reduces the number and changes the stability of salt bridges in its nucleic acid binding domain (Supplementary Fig. S4J), which may trigger conformational changes and reduce the DNA-binding affinity of this domain, thereby preventing it from associating with the promoter of PD-L1.

To explore the clinical relationships among the RORA, HDAC3 and DDX3X complexes and PD-L1, we first used the combined score of the three molecules as a measure of complex levels based on their expression levels, and the samples were then divided into the high-score array and low-score array groups according to the median score [score = (Exp (DDX3X)/Exp (HDAC3)) × Exp (RORA)]. In the high-score group, the RORA-DDX3X complex is dominant, whereas the RORA-HDAC3 complex is dominant in the low-score group. We subsequently analyzed the clinical correlation of the three molecules in our dataset, and the results showed that the expression level of PD-L1 was positively correlated with the score and that the proportion of responsive patients in the high-score array was significantly greater than that in the low-score array (Fig. 6L and M). Taken together, our findings indicate that the combined score of HDAC3, DDX3X, and RORA may be a potential indicator for predicting the clinical efficacy of antitumor immune checkpoint therapy in melanoma patients.

The immunosuppressive checkpoints PD-L1 and CTLA4 function through distinct mechanisms. Recently, the combination of PD-L1 and CTLA4 blockade has been shown to increase the antitumor response and significantly improve immunotherapy efficacy (44, 45). Building on our observation that nobiletin treatment suppresses PD-L1 expression and promotes T-cell-mediated antitumor immunity, we selected CTLA4 blockade and nobiletin for combination therapy in the B16F10 melanoma mouse model. To rule out the impact of tumor size on anti-CTLA4 and nobiletin therapy, we selected tumors of similar size to those treated with IgG for anti-CTLA4 and/or nobiletin treatment, and the mice were given nobiletin (25 mg kg⁻¹) daily and/or anti-CTLA4 (200 μg/mouse) once every 3 days. The tumor volume of each mouse was measured daily and combined with the invivoSyn method (46), the results showed that nobiletin and anti-CTLA4 combination therapy synergistically and most notably reduced tumor growth and tumor weight (Fig. 7A and B; Supplementary Fig. S5A and S5B) without significantly affecting mouse body weight, liver or spleen morphology (Supplementary Fig. S5C–S5E). In addition, the combination of nobiletin and an anti-CTLA4 antibody synergistically and obviously improved mouse survival (Fig. 7C). We further analyzed GZMB⁺ CD8⁺ T-cell infiltration in melanoma by flow cytometry, and the absolute numbers of tumor infiltrating immune cells are summarized in Supplementary Table S3. Notably, cotreatment with nobiletin and an anti-CTLA4 antibody also increased the percentage of GZMB⁺ CD8⁺ T cells that infiltrated the tumor (Fig. 7D–G). These results indicate that nobiletin can significantly inhibit tumor growth and may be a promising therapeutic for increasing the efficacy of anti-CTLA4 immunotherapy in melanoma.

Although circadian clock disruption is known to be associated with antitumor immunity (47), the precise molecular mechanisms underlying this association in melanoma are unclear. Here, we showed that RORA is the most likely key clock component associated with the tumor immune response and that its activation enhances T-cell cytotoxicity by repressing PD-L1 transcription. Moreover, we identified a novel RORA cofactor, DDX3X, and revealed its role in activating immunosuppressive effects by upregulating PD-L1. Furthermore, we identified the potential synergistic clock-targeting therapeutic and immunotherapeutic efficacy predictive strategy in melanoma patients (Fig. 7H). In summary, our data increase our understanding of the effects of clock component–regulated immunostimulants on melanoma with potential clinical implications.

Clock gene disruption facilitates carcinogenesis and cancer progression. The design of potential clock-oriented therapeutic strategies is attracting increasing attention. Through application of agonists, therapeutic targeting of clock components (NR1D1, NR1D2, and REV-ERBs) could be beneficial by stimulating apoptosis and impairing the proliferation of malignant cancer cells (48). Additionally, nobiletin is an activator of RORs (31) and exhibits antitumor effects on multiple cancer types by triggering tumor cell cycle arrest, apoptosis and ferroptosis (49–51); however, the effect of nobiletin on antitumor immune cell composition and immune responses is largely unknown. Here, we extend the effects of nobiletin, an antitumor immune activator, on melanoma. Nobiletin has been proposed to inhibit PD-L1 by suppressing EGFR/STAT3 signaling in NSCLC (52), suggesting that nobiletin may regulate the expression of PD-L1 in different cancer types through multiple mechanisms, so it would be interesting to further investigate the main mechanism behind its antitumor immunity. Although nobiletin is a very promising drug that can be applied in cancer treatment, there are currently no drugs that precisely target RORA. Therefore, the development of drugs specifically targeting RORA in the future will be a valuable achievement in cancer treatment, and in vivo and clinical trials will be needed to promote their clinical application.

As a nuclear receptor, RORA engages in the transcriptional inhibition of target genes by interacting with coregulators. RORA and HDAC3 have been shown to form coinhibitory complexes that are responsible for the transrepressive effect of NF-κB (39, 41). Our study showed that HDAC3 interacts with RORA and plays an important role in RORA-mediated PD-L1 suppression. Moreover, RORA was reported to increase the expression of target genes by interacting with coactivators, and GRIP-1 is the first proven coactivator of RORA in yeast and mammalian cells (37). DDX3X functions as a tumor suppressor and increases the transcription of target genes by interacting with transcription factors (53). However, the function of DDX3X in immunosurveillance and its relationship with RORA have not been reported. Our study revealed that DDX3X is a novel “coactivator” of RORA and highlighted its critical role in immune checkpoint modulation and cancer immunosuppression, adding another layer of complexity to its function in melanoma. Future investigations of the potentially important residues in the RORA protein that interact with DDX3X will be critical for the development of targeted therapeutic drugs.

The predictive value of clock components for immunotherapy response is another area of innovation. CLOCK–BMAL1 inhibition increases the antitumor responsiveness to anti-PD1 therapy in glioblastoma (54). We showed that RORA can form various coregulator complexes with distinct cofactors in melanoma cells and speculated that the ability of these cofactors to bind to RORA may depend on their expression levels. Therefore, the combined score of these three molecules was used as a measure of complex levels, and a higher score indicated a better therapeutic outcome and longer PFS in immune checkpoint blockade-treated melanoma patients. Thus, we identified a potential screening strategy for the assessment of therapeutic efficacy enabling more informed use of immunotherapy, but this method requires further validation with more patient tumor samples.

No disclosures were reported.

D. Liu: Conceptualization, data curation, writing—original draft, writing—review and editing. B. Wei: Validation, methodology. L. Liang: Validation, writing—review and editing. Y. Sheng: Validation, methodology. S. Sun: Validation, methodology. X. Sun: Funding acquisition, validation. M. Li: Validation, methodology. H. Li: Validation, methodology. C. Yang: Validation, methodology. Y. Peng: Funding acquisition, validation, methodology. Y. Xie: Funding acquisition, validation. C. Wen: Validation. L. Chen: Validation, methodology. X. Liu: Conceptualization, supervision. X. Chen: Conceptualization, supervision. H. Liu: Conceptualization, supervision, funding acquisition, writing—review and editing. J. Liu: Conceptualization, supervision, funding acquisition, writing—review and editing.

This work was supported by the grants from the National Key Research and Development Program of China (2019YFE0120800 and 2019YFA0111600 to H. Liu), Key Program of National Natural Science Foundation of China (U22A20329 to H. Liu), Natural Science Foundation of China for outstanding Young Scholars (82022060 to H. Liu), Science and Technology Innovation Program of Hunan Province (2022RC3004 to H. Liu), Central South University Research Program of Advanced Interdisciplinary Studies (2023QYJC004 to H. Liu), and The Scientific Research Program of FuRong Laboratory (2023SK2095 to H. Liu). J. Liu was supported by the National Natural Science Foundation of China (82270127 and 81920108004). Y. Xie was supported by the Central South University Fundamental Research Foundation (2023ZZTS0572). X. Sun was supported by the National Natural Science Foundation of China (82203880) and Fellowship of the China Postdoctoral Science Foundation (2022M720174 and 2023T160740). Y. Peng was supported by the Postdoctoral Fellowship Program of CPSF (GZC20233178).