Upregulation of PD-L1 via HMGB1-Activated IRF3 and NF-κB Contributes to UV Radiation-Induced Immune Suppression

Solar ultraviolet radiation (UVR) is a key epidemiologic factor causing skin cancers, such as cutaneous melanoma (1). As an environmental genotoxic stressor, UVR induces DNA damage, elicits inflammation, as well as alters genome structure and function in skin cells, which all contribute to the development of skin cancers and aging. Within the solar UV spectrum, UVB and UVA are of major environmental significance to skin carcinogenesis, because UVC is mostly absorbed by ozone in the earth's atmosphere. UVB can penetrate into the dermis papillary area and induce DNA damage in skin-residing keratinocytes, melanocytes, and dendritic cells, resulting in its much higher carcinogenicity than UVA (2). The influence of UVR in oncogenic mutation of melanoma was further supported by the TCGA melanoma study, which identified the UVR-associated mutation signature in 76% of primary tumors and 84% of metastatic samples in patients with melanoma (3). Besides leading to genomic mutation, UVR could suppress the local immune response through damaging and expelling skin Langerin⁺ antigen-presenting dendritic (Langerhans) cells. In addition, UVR attenuates systemic immunity by inhibiting effector and memory T cells while activating regulatory T and B cells (4). The resulting immunosuppressive microenvironment of UVR-exposed skin enables premalignant skin cells and tumor cells to escape immune surveillance and facilitates cutaneous melanoma initiation and progression. Consistently, increased risk of invasive melanoma was observed in organ transplant patients who normally underwent medical immunosuppression to prevent graft rejection (5). Therefore, reinvigorating the immunosuppressive microenvironment of the skin after UVR could play a pivotal role in reducing incidence and progression of invasive melanoma.

Recent advances in understanding the critical role of immune checkpoints in regulating tumor-infiltrating T-cell activity has led to a radical shift in cancer immunotherapy and remarkable success in treating patients with invasive melanoma with immune checkpoint blockers, such as humanized antibodies antagonizing cytotoxic T lymphocyte antigen 4 (CTLA4, CD152), programed death-1 (PD-1, CD279), or its ligand (PD-L1, CD274; ref. 6). Naïve T-cell activation requires T-cell receptor (TCR) activation by recognition of specific antigen presented by antigen-presenting cells (APC), and costimulatory or coinhibitory signals to further modulate T-cell activation (7). Costimulatory signals, such as CD28 ligation with B7-1/CD80 or B7-2/CD86, are required for effective activation of T-cell immunity. On the contrary, coinhibitory signals, such as CTLA4 binding with B7-1/B7-2 and PD-1/PD-L1 ligation, function as immune checkpoints to prevent tissue damage from overactivated T-cell immunity and maintain peripheral immune tolerance. Tumor cells can exploit the immune checkpoints by expressing increased ligands for coinhibitory receptors, such as PD-L1 and PD-L2, and induce an immunosuppressive tumor microenvironment, thereby escaping from antitumor immunity (8). Thus, blocking immune checkpoint signals mediated by CTLA4 and PD-1/PD-L1 significantly enhances antitumor immunity and has shown durable efficacy in treating various types of cancer, including invasive melanoma.

Although the immune-suppressive effect of UVR has been well established, whether immune checkpoint activation is involved in the UVR-dependent immune suppression is not completely understood. Gene expression profiling using neonatal melanocytes from mouse skin exposed to UVR revealed an IFN response signature that includes CTLA4 induction (9). This increased CTLA4 transcription is likely dependent on macrophage-produced IFNγ within the skin microenvironment (9, 10). Here, we show that UVB also induces PD-L1 upregulation in melanocytes and melanoma cells, which is independent of IFN signaling. Instead, UVR induces HMGB1 release from skin cells, which engages the receptor for advanced glycation endproducts (RAGE) receptor and activates the NF-κB/IRF3 transcriptional complex in melanocytes. The NF-κB/IRF3 complex was enriched on the PD-L1 promoter upon UVR and was responsible for transcriptional upregulation of PD-L1. Consistently, PD-L1 levels were significantly correlated with activation of NF-κB and IRF3 gene signature in melanoma patient samples. Moreover, blocking the HMGB1/RAGE/NF-κB/IRF3 signaling cascade or using PD-1/PD-L1 checkpoint blocker dramatically enhanced the susceptibility of melanoma cells to CD8⁺ T-cell-mediated cytotoxicity after UVR exposure. Overall, our findings support a critical role of UVR-induced PD-L1 upregulation in promoting an immunosuppressive microenvironment in the skin after UVR, which facilitates the evasion of premalignant melanocytes and melanoma cells from tumor immune surveillance, leading to melanoma initiation and progression.

The SK-Mel-28, SK-Mel-2, MeWo, A375, HaCat, and B16 cells were obtained from ATCC, and maintained in DMEM or RPMI1640 media supplemented with 10% FBS following standard culture conditions. Human primary epidermal melanocytes (HEM) were purchased from Lifeline and cultured with DermaLife Melanocyte Complete Medium following the manufacturer's protocol. The WM35 and WM1341 cell lines were generous gifts from Dr. Meenhard Herlyn (Wistar Institute, Philadelphia, PA), and were cultured in DMEM supplemented with 10% FBS. ATCC cell lines were characterized by short tandem repeat profiling, and cells resuscitated from original passage were used up to 20 passages within 6 months in all experiments. Mycoplasma contamination was routinely tested by DAPI staining. SK-Mel-28 KO lines (TBK1 KO, HMGB1 KO, RAGE KO, and IRF3 KO) were generated with CRISPR/Cas9. lentiCRISPR v2 was a gift from Feng Zhang (MIT McGovern Institute, Cambridge, MA; Addgene plasmid # 52961). The following sgRNA sequences were used: sg1TBK1 CAT AAG CTT CCT TCG TCC AG and sg2TBK1 GAA GAA CCT TCT AAT GCC TA; sg1HMGB1 GGA GAT CCT AAG AAG CCG AG and sg2HMGB1 GAG AAG TTG ACT GAA GCA TC; sg1RAGE TAT CTC CCA GGG GCA GTA GT and sg2RAGE GTG GCT CAC CCC ACA GAC TG; sg1IRF3 TTT AGC AGA GGA CCG GAG CA and sg2IRF3 ATC TAC GAG TTT GTG AAC TC.

Briefly, 10 μg of total cellular extract was incubated with ³²P-labeled double-stranded oligonucleotides (NF-κB: 5′-TCA ACA GAG GGG ACT TTC CGA GAG GCC-3′, or OCT-1: 5′-TGT CGA ATG CAA ATC ACT AGA A-3′) in electrophoretic mobility super-shift assay (EMSA) reaction buffer for 15 minutes at room temperature. The reaction was separated on a 4% nondenaturing polyacrylamide gel, which was then dried and analyzed with a Cyclone phosphoimager (PerkinElmer Life Sciences).

Briefly, cells were lysed in 10% PBS and 90% IP lysis buffer (20 mmol/L Tris (pH 7.0), 250 mmol/L NaCl, 3 mmol/L EDTA, 3 mmol/L EGTA, 0.5% Nonidet P-40, 2 mmol/L DTT, 0.5 mmol/L phenylmethylsulfonylfluoride, 20 mmol/L β-glycerol phosphate, 1 mmol/L sodium orthovanadate, 1 μg/mL leupeptin, 1 μg/mL aprotinin, 10 mmol/L p-nitrophenyl phosphate, and 10 mmol/L sodium fluoride). Five percent of total lysates were used as input for each sample. The remaining lysate was incubated with 1 μg of primary antibody (or control IgG) on the rotator at 4°C overnight. Protein G sepharose was then added and incubated for another 4 hours at 4°C. Protein G sepharose–enriched complexes were resolved on SDS-PAGE gels and transferred onto polyvinylidene difluoride membranes. Protein signals were detected by specific antibodies and visualized by chemiluminescence.

In brief, cells were cross-linked with 1% formaldehyde, sonicated, and then immunoprecipitated with antibodies against p65 or IRF3 in dilution buffer (1% Triton X-100, 2 mmol/L EDTA, 150 mmol/L NaCl, 20 mmol/L Tris-HCl, pH 8.1). The chromatin immunoprecipitation (ChIP)-qPCR primers were designed to amplify the promoter regions containing respective binding sites in PD-L1 promoter. For ChIP-reChIP experiments, the p65 immunopreciptate-containing beads were washed and resuspended in reChIP buffer (dilution buffer containing 10 mmol/L DTT) and incubated at 37°C for 50 minutes. The sample was then diluted 40 times in dilution buffer, and subjected to secondary immunoprecipitation with anti-IRF3 antibody. Enriched chromatin was analyzed by qPCR with respective primer sets.

Centricon-10 (Amicon) was used to concentrate the extracellular HMGB1 protein. Cell culture media (10 mL) were collected after respective treatments and applied to the concentrators followed with centrifugation at 5,000 × g for 2 hours. The media fraction containing molecules larger than 10 kDa (HMGB1; 25 kDa) were enriched in a volume around 50 μL. The extracellular HMGB1 level was examined by Western blot analysis.

To measure cell surface PD-L1 protein level, cultured cells were washed with PBS and incubated with recombinant human PD-1 Fc protein (R&D Systems) at room temperature for 1 hour. The Alexa Fluor 488-conjugated anti-human (Life Technologies) was used as a secondary antibody. After mounting, the cells were visualized using an EVOS fluorescence microscope (Thermo Fisher Scientific). To determine the PD-L1 level in cells by flow cytometry, unfixed cells were incubated with Alexa Fluor 488-conjugated PD-1 Fc protein on ice for 30 minutes and analyzed with Accuri 6 flow cytometer (BD Biosciences).

To prepare tumor cell suspension for flow cytometry analysis, xenograft tumors were harvested, washed extensively in PBS, and minced before digestion with type IV collagenase (2 mg/mL), hyaluronidase (0.01%), and DNAse I (20 μg/mL; Sigma-Aldrich) in Hank balanced salt solution at 37°C for 2 hours. The cell suspensions were filtered through a cell strainer (70 μm) and washed in PBS containing 2 mmol/L EDTA and 2% FBS. Cell suspensions were then incubated with anti-CD8 (53-6.7)-FITC and anti-granzyme B [anti-GZMB; (GB11)-PE (all from Invitrogen)] for 1 hour at 4°C and analyzed with Accuri 6 flow cytometer.

Human peripheral blood mononuclear cells (PBMC) were isolated from apheresis rings obtained from the Blood Donor Center at St. Jude Children's Hospital (Memphis, Tennessee). Written informed consent was obtained from blood donors for their discarded blood products to be used for research purposes. This consent form has been approved by the Institutional Review Board at St. Jude Children's Research Hospital (Memphis, Tennessee) and is in accordance with the Declaration of Helsinki. The PBMCs were isolated using lymphocyte separation medium and density centrifugation, and the naïve CD8⁺ T cells were purified by magnetic beads (Miltenyi Biotec, catalog no. 130-093-244) following the manufacturer's instructions. Polyclonal effector CD8⁺ T cells were generated by stimulating cells with immobilized anti-CD3 (1 μg/mL; clone OKT3, Bio X Cell) and soluble anti-CD28 (2 μg/mL; clone CD28.2, BioLegend) antibodies in the presence of 100 U/mL recombinant human IL2 for two days in complete RPMI medium containing 10% FBS and penicillin, streptomycin, and l-glutamine. Live cells were isolated using lymphocyte separation medium and density centrifugation, followed by expansion in complete medium containing IL7 (10 ng/mL) and IL15 (10 ng/mL) for five days.

To generate tumor-specific effector CD8⁺ T cells (TA-CTL), PBMCs were incubated in tissue culture-treated plates for 2 hours at 37°C in complete RPMI medium. Naïve CD8⁺ T cells were purified from the suspension cell fraction as above and rested in complete RPMI medium containing IL7 and IL15. The suspension cell fraction that was depleted of naïve CD8⁺ T cells was then incubated with soluble anti-CD3 antibody (10 μg/mL) overnight to induce both T and B-cell activation, which allows for B cells to serve as antigen-presenting cells (11, 12). The cells were then washed and incubated for 24 hours with irradiated SK-Mel-28 melanoma cells at a 3:1 tumor cell-to-immune cells ratio. The cultures with then fixed with 1% PFA prepared in PBS for 10 minutes at 37°C, followed by washing with complete RPMI medium. Naïve CD8⁺ cells and the fixed, tumor-pulsed immune cells were then mixed at a 2:1 ratio and incubated for two days in complete RPMI1640 medium containing soluble anti-CD28 antibody (1 μg/mL) and 100 U/mL IL2. The live CD8⁺ T cells were isolated using lymphocyte separation medium and expanded in completed medium containing IL7 and IL15 for five days.

To obtain CD8⁺ T cells from OT-I mice, the spleen and peripheral lymph nodes (pLN) were harvested from female OT-I mice (The Jackson Laboratory), which contain T cells whose TCRs are specific for ovalbumin peptide residues 257–264 (OVA257–264). After manual disruption and ACK-mediated lysis of red blood cells, CD8⁺ T cells were positively selected using CD8 microbeads (Ly-2, Miltenyi Biotec) as per the manufacturer's instructions. Naïve CD8⁺CD25⁻CD44⁻CD62L⁺ OT-I cells were then sorted using a Sony Synergy. Naïve OT-I cells were suspended in complete Click's medium and activated with anti-CD3 (clone 2C11, 5 μg/mL) and anti-CD28 (5 μg/mL) antibodies in the presence of IL2 (100 U/mL) for two days, removed from the antibodies, and expanded in IL2-containing medium for three additional days. The cells were maintained in IL7 and IL15-containing medium prior to use in cytotoxicity assays or tumor experiments.

All animal studies were conducted in accordance with NIH animal use guidelines and following protocols approved by Institutional Animal Care and Use Committee at University of Tennessee Health Science Center (UTHSC; Memphis, TN). NOD.Cg Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice (age 6 weeks) were maintained in the UTHSC animal facility. SK-Mel-28 cells (Luciferase-labeled) were treated with UVB (50 mJ/cm²) or mock treated. After 24 hours, 1 × 10⁶ SK-Mel-28 cells and 2.5 × 10⁶ TA-CTLs were subcutaneously coinjected into the flanks of these mice (2 per mouse, n = 5 per group) following a cotransplantation model reported previously (13). Tumor volume was assessed by caliper measurement using the formula (width² × length/2). Mice were also imaged by Xenogen IVIS system (Perkin Elmer). The mice were treated with anti-PD-1 mAb (BioXCell, clone J116, 100 μg/mouse) or isotype on day 4 after transplantation then every other day for two weeks as indicated. At the endpoint, the mice were sacrificed and primary tumors were isolated for further analyses.

For syngeneic mouse melanoma xenograft model, C57/BL6J mice (age 7 weeks) were maintained in the UTHSC animal facility. B16-OVA cells were treated with UVB (50 mJ/cm²) or mock treated. After 24 hours, 5 × 10⁵ B16-OVA cells and 1 × 10⁶ OTI-CD8⁺ T cells were subcutaneously coinjected into the flanks of C57/BL5 mice (2 per mouse, 4 male + 3 female mice per group). The mice were treated with anti-PD-L1 mAb (BioXCell, clone 10F.9G2, 100 μg/mouse) or isotype on day 1 after transplantation then every other day for two weeks as indicated. Tumor growth was monitored by caliper measurement and harvested tumors were analyzed as above.

UVB radiation enhances the expression of the immune checkpoint molecule CTLA4 in melanocytes in neonatal mouse skin, which is dependent on macrophage-secreted IFNγ within the skin microenvironment upon UVR (9, 10). We further analyzed the melanocytic gene expression profile and revealed that another immune checkpoint molecule PD-1 ligand (PD-L1/CD274) was also significantly upregulated following UVB exposure (Supplementary Fig. S1A). To determine whether UVR influenced the expression of PD-L1 in melanoma cells, we treated SK-Mel-28 cells with UVB. Expression of PD-L1 was significantly increased at both the mRNA and protein levels in response to UVB treatment (Fig. 1A). UVB exposure increased PD-L1 levels in SK-Mel-28 cells in a time-dependent manner (Supplementary Fig. S1B). Consistently, PD-L1 level on cell surface also was substantially increased in SK-Mel-28 cells upon UVB exposure (Fig. 1B and C). UVB exposure also increased the expression of PD-L1 in a panel of human melanoma cell lines and mouse B16 melanoma cells (Fig. 1D), suggesting that PD-L1 induction is a conserved stress response to UV exposure in melanocytes and melanoma cells.

In the tumor microenvironment, PD-L1 expression can be induced by inflammatory cytokines such as IFNγ (6), which activates STAT1 and increases PD-L1 transcription (14). We found PD-L1 induction by UVB was not affected by momelotinib (CYT387, JAK1/2 inhibitor) or WP1066 (JAK2 inhibitor), although UVB-induced activation of STAT1 and STAT3 was blocked (Supplementary Fig. S1C), suggesting IFNγ likely is dispensable for UVB-induced PD-L1 in melanoma cells. UVR also activates the transcription regulator NF-κB, which was suggested to be a downstream mediator of IFNγ-induced PD-L1 induction (15). We observed that UVB treatment induced NF-κB activation in melanoma cells (SK-Mel-28, WM35, and WM1341), melanocytes (HEM), and keratinocytes (HaCaT) in a dose-dependent manner (Fig. 1E; Supplementary Fig. S1D and S1E). NF-κB has been reported to upregulate PD-L1 transcription in ovarian, lung, and breast cancer cells (16–18). Indeed, we found that UVB-induced PD-L1 upregulation in melanoma cells was blocked by an IKK2 inhibitor TPCA-1 (Supplementary Fig. S1F). Moreover, depletion of RelA/p65 of NF-κB family remarkably attenuated PD-L1 induction in SK-Mel-28 cells exposed to UVB (Fig. 1F and G, compare blue and green lines). All these data suggested that UVB-induced NF-κB activation plays a critical role in mediating PD-L1 induction in melanoma cells.

NF-κB activation by UVR, especially by UVC, depends on casein kinase 2 (CK2)-mediated p38α activation, which, in turn, phosphorylates IκBα and promotes its degradation (19). However, inhibiting either p38 or CK2 minimally affected NF-κB activation in UVB-treated melanoma cells (Supplementary Fig. S2A). We previously reported that UVC-induced NF-κB activation was regulated by DNA damage response kinases ATM and ATR (20, 21). Nevertheless, inhibiting ATM, ATR, or DNA-PK with specific inhibitors or by genetic deletion did not substantially affect UVB-induced NF-κB activation (Supplementary Fig. S2B–S2D). IκBα kinases (IKK) were shown to be required for UV-induced NF-κB activation (20, 22, 23), and we found IKK inhibition, by either a pharmacologic inhibitor or genetic deletion, abrogated NF-κB activation by UVB (Fig. 2A and B), supporting an essential role of IKKs in mediating UVB-induced NF-κB signaling. Supershift analysis further revealed that the p65/p50 heterodimer was the major NF-κB complex activated by UVB in SK-Mel-28 melanoma cells (Supplementary Fig. S2E). Because IKK activity is required for UVB-induced NF-κB activation, we examined whether the IKK upstream kinase TAK1 is required for this signaling cascade. Surprisingly, depletion of TAK1 barely affected UVB-induced phosphorylation of IκBα and p65, suggesting that TAK1 is not essential for IKK activation in response to UVB (Supplementary Fig. S2F).

TBK1 was originally identified as a TRAF2/TANK–associated kinase activating NF-κB (24). We found inhibiting TBK1 abolished UVB-induced NF-κB activation, which was comparable to that by inhibiting IKK (Fig. 2A). TBK1-dependent phosphorylation of IKKβ was reported to activate canonical IKK activity by directly phosphorylating IKKβ (25). In accordance, we found that UVB-activated TBK1 phosphorylates IKKβ in vitro, which depends on TBK1 kinase activity (Fig. 2C). UVB treatment induced a dose-dependent activation of TBK1 and phosphorylation of its canonical downstream substrate IRF-3 (Supplementary Fig. S2G). Moreover, UVB-induced p65 phosphorylation was abolished in TBK1-knock out (KO) SK-Mel-28 cells (Fig. 2D). PD-L1 induction by UVB was also inhibited in IKKα/β^−/− MEFs, TBK1-KO cells, and cells cotreated with inhibitors of IKK or TBK1 (Fig. 2A–D), supporting that IKK and TBK1 are essential for UVB-induced NF-κB activation and subsequent PD-L1 induction.

We next asked whether TBK1/NF-κB activation by UVB depends on a signal from the extracellular milieu after UVB treatment. We found conditioned media (CM) from UVB-treated SK-Mel-28 cells could activate NF-κB, which was mediated by TBK1 and IKK, but not CK2 (Fig. 3A). CM from UVB-treated primary melanocytes also effectively induced NF-κB activation in cells that were not exposed to UVB (Fig. 3B), suggesting that secreted molecule(s) in CM from UVB-treated cells is sufficient for inducing NF-κB activation. Moreover, this activity was sensitive to heat inactivation, but not nuclease treatment, suggesting that secreted proteins are essential for NF-κB activation upon UVB (Fig. 3B).

Secretion of the nuclear protein HMGB1, as a DAMP (danger-associated molecular patterns) molecule, from keratinocytes has been linked with skin inflammation upon UVR (26, 27). We found HMGB1 was released from melanocytes and melanoma cells as early as 0.5 hours after UVB exposure (Fig. 3C; Supplementary Fig. S2H). The increased extracellular HMGB1 was likely due to active secretion instead of passive release from necrotic cells, because little cell necrosis was observed until 2 hours after UVB radiation. The subcellular translocation and release of HMGB1 was further confirmed by immunofluorescence showing a substantial decrease of nuclear HMGB1 in response to UVB (Fig. 3D). To determine whether HMGB1 is essential for UVB-induced N-κB activation in melanoma cells, we generated HMGB1-KO SK-Mel-28 cells. Deletion of HMGB1 abolished the activation of TBK1 and phosphorylation of p65 and IRF-3 in response to UVB (Fig. 3E). Moreover, CM from UVB-treated HMGB1-KO SK-Mel-28 cells failed to activate NF-κB in parental SK-Mel-28 cells (Fig. 3F). These results strongly support that HMGB1 secretion upon UVB exposure plays a critical role in mediating NF-κB activation in an autocrine and/or paracrine fashion.

To determine whether HMGB1 is sufficient for activating NF-κB in melanoma cells, we treated SK-Mel-28 cells with recombinant human HMGB1 (rHMGB1). NF-κB activation was substantially increased in SK-Mel-28 cells exposed to rHMGB1 (Fig. 3G). Moreover, treatment with recombinant HMGB1-BoxA fragment, which serves as an inactivating competitor for the HMGB1 receptor (28), decreased UVB-induced NF-κB activation in melanoma cells in a dose-dependent manner (Fig. 3H). Importantly, PD-L1 induction was detected in melanoma cells exposed to UVB CM from wild-type (WT) SK-Mel-28 cells, but not from HMGB1-KO SK-Mel-28 cells, likely due to the inability to induce NF-κB activation (Fig. 3I).

Active secretion of HMGB1 has been associated with increased HMGB1 acetylation at its nuclear localization signals, which blocks the re-entry of shuttling HMGB1 into the nucleus (29). We found UVB treatment significantly increased HMGB1 acetylation (Supplementary Fig. S3A). Moreover, depletion of the deacetylase SIRT1 further enhanced UVB-induced HMGB1 acetylation (Supplementary Fig. S3B), which is consistent with previous reports showing that SIRT1-dependent deacetylation inhibits HMGB1 secretion (30, 31). Accordingly, deletion of SIRT1 further enhanced UVB-induced TBK1 activation and subsequent phosphorylation of IRF3 and p65 (Supplementary Fig. S3C). We found the interaction between HMGB1 and SIRT1 was decreased in response to UVB treatment (Supplementary Fig. S3D). Inhibiting SIRT1 activity by nicotinamide (NAM) promoted HMGB1 secretion from SK-Mel-28 cells, comparable with that by UVB treatment (Supplementary Fig. S3E). Moreover, WT SIRT1, but not the catalytic-inactive H355A mutant, suppressed UVB-induced TBK1/NF-κB activation in melanoma cells (Supplementary Fig. S3F). We also found treatment with the SIRT1-activating agent, resveratrol, inhibited HMGB1 secretion by UVB (Supplementary Fig. S3G), which was correlated with decreased TBK1 activation and reduced phosphorylation of IRF3 and p65 in cells exposed to UVB (Supplementary Fig. S3H). SIRT1 is a NAD-dependent deacetylase and decreased NAD⁺/NADH ratio suppresses SIRT1 function (32). We found that the intracellular NAD⁺/NADH ratio was significantly decreased in SK-Mel-28 cells in response to UVB (Supplementary Fig. S3I), which may be responsible for decreased SIRT1 activity upon UVB exposure. All these results indicate that UVR may promote HMGB1 acetylation and secretion by suppressing SIRT1 in melanoma cells, which results in increased activation of NF-κB/IRF3.

To identify the receptor responsible for HMGB1-induced NF-κB activation upon UVB exposure, we treated SK-Mel-28 cells with UVB in the presence of inhibitors of Toll-like receptor 4 (TLR4) or RAGE (receptor for advanced glycation endproducts). Inhibiting RAGE with FPS-ZM1 abolished UVB-induced NF-κB activation to a similar extent as inhibition of either IKKβ or TBK1, while treatment with TLR4 inhibitor had little effect (Fig. 4A). Inhibiting RAGE also abrogated UVB-induced TBK1/NF-κB activation and IRF3 phosphorylation in another melanoma cell line MeWo (Fig. 4B). UVB-induced TBK1/NF-κB activation was absent in RAGE-KO SK-Mel-28 cells (Fig. 4C). Inhibiting RAGE also blocked NF-κB activation by conditioned medium (CM) from UVB-treated melanoma cells (Fig. 4D). Moreover, the CM from UVB-treated melanoma cells failed to induce TBK1/NF-κB activation in RAGE-KO cells, although the same CM was able to induce TBK1/NF-κB activation in HMGB1-KO cells (Fig. 4E). In accordance, UVB-induced PD-L1 upregulation was also diminished by RAGE inhibition (Fig. 4B and F) or in RAGE-KO SK-Mel-28 cells (Supplementary Fig. S4A). These data support an essential role of RAGE in mediating the HMGB1-induced activation of NF-κB and IRF3.

It was recently reported that NF-κB activation by TNFα could increase PD-L1 levels by stabilizing PD-L1 protein, which depends on upregulation of the deubiquitinase CSN5 and subsequent deubiquitination of PD-L1 (33). PD-L1 mRNA levels were not significantly changed in breast cancer cells treated with TNFα. In contrast, we found UVB exposure significantly increased PD-L1 mRNA expression (Fig. 1A), which was attenuated by inhibiting RAGE, IKK, or TBK1 (Fig. 5A). The increased PD-L1 transcription by UVB was also diminished in HMGB1-KO cells (Fig. 5B), suggesting that UVB/HMGB1-activated NF-κB directly enhanced PD-L1 transcription. Accordingly, we detected a substantial enrichment of RelA/p65 on the PD-L1 promoter region upon UVB exposure (Fig. 5C), which supports a role of NF-κB in upregulating PD-L1 transcription in melanoma cells after UVR.

Interestingly, we found that UVB-induced PD-L1 upregulation was also decreased in IRF3-KO melanoma cells (Fig. 5D) or cells transfected with IRF3-targeting siRNAs, in which UVB-induced TBK1/NF-κB activation was not affected (Supplementary Fig. S4B). This observation indicates that UVB-activated NF-κB may not be sufficient for upregulating PD-L1 in melanoma cells, and IRF3 also plays an important role in mediating PD-L1 induction by UVB. Scanning the promoter sequence of PD-L1 revealed two putative IRF-binding sites (ISRE) along with the NF-κB–binding site within the proximal region of PD-L1 promoter. We also detected significantly increased IRF3 binding at the distal ISRE2 site within the PD-L1 promoter upon UVB (Fig. 5E). Unexpectedly, UVB-induced IRF3 enrichment was also detected at NF-κB–binding site in the promoter, while p65 recruitment was specifically at NF-κB–binding site upon UVB exposure. To determine the importance of ISRE2 and NF-κB–binding site in regulating PD-L1 transcription upon UVB, we generated a luciferase reporter whose transcription was controlled by the PD-L1 promoter. We found deletion of ISRE2 did not significantly reduce the reporter activity in response to UVB, whereas mutating NF-κB–binding site diminished UVB-induced transactivation (Fig. 5F), supporting an essential role of the NF-κB-binding site in enhancing PD-L1 transcription upon UVB. IRF3 was previously showed to interact with NF-κB/p65, and the IRF3–p65 complex was required for transactivation of IRF3-target genes such as IFN induction by LPS (34). We confirmed that IRF3 and p65 formed a complex in melanoma cells treated by UVB (Fig. 5G). Moreover, we found the p65–IRF3 complex was enriched on the NF-κB–binding site in the PD-L1 promoter in response to UVB exposure (Fig. 5H), suggesting that UVB-induced transactivation of PD-L1 was regulated by p65–IRF3 regulatory complex.

Coinhibitory signals from PD-1/PD-L1 keep T cells' activity in check and attenuate cytotoxic CD8⁺ T-cell (CTL)-mediated tumoricidal effects (6). We found that UVB treatment significantly reduced the susceptibility of SK-Mel-28 (Fig. 6A) or B16 melanoma cells (Supplementary Fig. S5A) to CTL-dependent cytotoxicity. The decreased CTL susceptibility was observed in SK-Mel-28 cells exposed to CTL cells activated by either anti-CD3 or CD28 antibodies (Supplementary Fig. S5B, AA T cells, black lines) or tumor-specific antigens from IR-inactivated melanoma cells (Supplementary Fig. S5B, TA T cells, red lines). Inhibiting RAGE or TBK significantly increased the specific lysis of UVB-exposed melanoma cells by CTLs (Fig. 6B). Moreover, depletion of IRF3 also enhanced the susceptibility of melanoma cells to CTL-mediated cytotoxicity after UVB exposure (Fig. 6C). Blocking PD-L1/PD-1 signaling with an anti-PD-1 antibody significantly decreased the survival of UVB-treated SK-Mel-28 cells cocultured with activated CTLs (Fig. 6D), which was associated with dramatically increased CTL-dependent cytolysis in UVB-reposed melanoma cells (Fig. 6E). These findings suggest that UVB-induced activation of TBK1/IRF3/NF-κB promotes melanoma cell escape from CTL-mediated antimelanoma immunity by upregulating PD-L1.

To determine whether the level of PD-L1 correlates with NFκB activation and/or IRF3 activity in melanoma patient samples, we analyzed the genomic data from the TCGA cutaneous melanoma study (3). We found that the PD-L1/CD274 expression level is significantly correlated with NF-κB-p65-pS536 levels, a surrogate of NF-κB activity, determined by reverse-phase protein analysis (RPPA) in patients with melanoma (Fig. 6F; Supplementary Fig. S5C). Because IRF3 was not included in the RPPA analyses, we used an IRF3-upregulated gene signature (35) as a readout for measuring IRF3 activity in the patient samples. High IRF3 activity in melanoma patient samples was also correlated with PD-L1/CD247 transcription level (Fig. 6F and G), supporting a critical role of IRF3 activation in promoting PD-L1 expression.

To determine whether UVB exposure promotes melanoma growth by escaping antimelanoma immunity, we stimulated human CD8⁺ T cells with inactivated SK-Mel-28 cells. UVB treatment of tumor antigen–activated CTLs had little effect on their cytotoxicity toward SK-Mel-28 cells in vitro, while UVB exposure of the SK-Mel-28 cells decreased their cytotoxicity (Supplementary Fig. S6A). After a cotransplantation animal model using melanoma cells and human tumor–reactive T cells (13), we coinjected SK-Mel-28 cells, exposed to UVB or mock-treated, with or without activated CTLs into the flanks of immunodeficient NSG mice. As expected, activated CTLs dramatically suppressed SK-Mel-28 xenograft tumor growth (Fig. 7A–C). Consistent with our in vitro results, UVB exposure substantially enhanced SK-Mel-28 xenograft growth in the presence of tumor-reactive CTLs, which was attenuated by anti-PD1 treatment (Fig. 7A–C). These data suggested that upregulation of PD-L1 in melanoma cells by UVB could suppress the tumoricidal activity of CTLs to promote melanoma progression. Indeed, we detected substantially increased expression of PD-L1 in xenografts from UVB-treated melanoma (Fig. 7D and E), which correlated with increased TBK1/IRF3/NF-κB activation in these tumors (Fig. 7D). CTL cytotoxicity against melanoma cells, visualized by granzyme B staining and quantified by percentage of GZMB⁺ cells in the CD8⁺ population isolated from tumors, was significantly suppressed in SK-Mel-28 tumors exposed to UVB, which was reinvigorated by anti-PD1 treatment (Supplementary Fig. S6B; Fig. 7F). Moreover, injection of PD-1 antibody did not suppress TBK1/IRF3/NF-κB activation by UVB and consequent PD-L1 induction (Fig. 7D and E), suggesting that the improved tumor regression by anti-PD1 was mediated by improved CTL activity. Consistent results were also observed in a syngeneic B16 melanoma xenograft model, in which UVB exposure substantially decreased the susceptibility of B16-OVA melanoma to activated OVA-specific OTI-CTLs (Supplementary Fig. S6C). In accordance to SK-Mel-28 xenograft model, treatment with anti-PD-L1 antibody significantly enhanced CTL-dependent antitumor immunity while minimally affecting UVB-induced TBK1/IRF3/NF-κB signaling (Supplementary Fig. S6D and S6E) in B16-OVA tumors.

We also compared the growth of xenograft tumors from UVB-treated WT or IRF3-KO SK-Mel-28 cells cotransplanted with tumor-reactive CTLs. Although both of these cells were exposed to UVB, IRF-3 deficiency significantly reduced tumor growth (Supplementary Fig. S6F), likely owing to increased melanoma sensitivity to CTL-mediated antimelanoma immunity. Consistent with this notion, PD-L1 levels in IRF3-KO SK-Mel-28 xenograft tumors were substantially decreased compared with that in wild-type xenografts (Supplementary Fig. S6G and S6H). Taken together, our results support that UVB-induced IRF3/NF-κB activation promotes melanoma cells' evasion from CTL-mediated antitumor immunity, which is dependent on the upregulation of PD-L1.

UVR-induced immunosuppression plays critical roles in promoting skin cancer (36). UVB-dependent expulsion of Langerhans cells from the skin microenvironment substantially reduces the antigen-presenting capacity of the skin-resident dendritic cells, which facilitates the premalignant cells and early-stage melanoma cells to escape from tumor immune surveillance (36, 37). UVB also induces immunosuppressive cytokine production from keratinocytes, such as IL4, IL10, and prostaglandin (PGE2). Moreover, exposure of UVB enhances migration of UVR-damaged Langerhans cells into the regional lymph nodes, which induces the expansion of regulatory T cells and reduces effector and memory T cells, resulting in systemic inhibition of adaptive immunity (4, 37). Here we showed that UVB also increased expression of PD-L1 in melanocytes and melanoma cells, which could facilitate premalignant and melanoma cells to inhibit effector T-cell activity, thereby reducing antitumor immunity in the skin microenvironment. Moreover, PD-L1 has been shown to be present on the surface of the exosomes secreted from tumor cells, which are found in the circulation and are expected to suppress T-cell activity systemically (38). Therefore, PD-L1 induction in skin cells may participate in local and systemic immune suppression by UVR, and could play an important role in immune evasion of melanoma cells from tumor immune surveillance (Fig. 7G). Targeting UVB-induced HMGB1/RAGE/TBK1 signaling could abrogate PD-L1 induction in melanocytes and melanoma cells exposed to UVB, which may serve as potential pharmacologic targets for mitigating immune escape of the premalignant melanocytes from immune surveillance.

Our previous studies have shown that transcription factor NF-κB can be activated by a variety of genotoxic agents in human cancer cells (39, 40). As a DNA-damaging agent, UVC has been shown to activate NF-κB in a p38 and CK2-dependent manner (19). Moreover, we found that UVC-induced NF-κB activation was also regulated by the DNA damage response kinases ATM and ATR (20, 21). Inhibition of ATM or ATR could suppress or augment UVC-induced NF-κB activation, respectively. In contract to UVC, our data indicated that UVB-induced NF-κB activation in human keratinocytes and melanoma cells was not affected by inhibiting p38/CK2 or by ATM/ATR deletion. Moreover, another upstream kinase TAK1, which we have previously shown to be essential for DNA damage–induced NF-κB signaling (41), was dispensable for UVB-induced NF-κB activation in skin cells. All these data suggest that UVB could activate NF-κB in skin cells through a distinct signaling cascade. Indeed, we found HMGB1, secreted from skin cells as a stress response to UVR, was essential for NF-κB activation by UVB exposure. UVB-enhanced HMGB1 acetylation, which was mediated by inactivation of SIRT1, may play an essential role in promoting HMGB1 secretion from the skin cells. In the skin microenvironment, HMGB1 binds to the DAMP receptor RAGE and leads to downstream kinase TBK1 activation, which is required for UVB-induced NF-κB activation. Consistent with previous studies (22, 23, 42), IKKβ was required for UVR-induced NF-κB activation. There was strong evidence that IKKs are indispensable for UVR-induced IκB proteolysis, as UV-induced IκBα degradation and NF-κB activation were not detected in IKK-KO cells, particularly in IKKβ-deficient (IKKβ^−/−) cells. Besides phosphorylating IκBα in cytoplasm, the nuclear IKKβ also functions as an adaptor protein for IκBα association with β-TrCP in the nucleus, which results in IκBα degradation and subsequent NF-κB activation in response to UVR (42). Furthermore, computational modeling also suggested that constitutive phosphorylation of IκBα by basal IKK activity, which affects the turnover rate of IκBα, and IκBα synthesis inhibition are crucial for UV-induced NF-κB activation (23, 43). We found UVB-activated TBK1 could directly phosphorylate IKKβ, which may promote IKK activation. TBK1-dependent phosphorylation of IKKβ was reported to either activate or inhibit canonical IKK activity depending on the upstream stimuli (25, 44). In skin cells exposed to UVB, TBK1 activity is essential for IKK activation and subsequent NF-κB activation. It remains to be further delineated how RAGE promotes TBK1 activation in cells exposed to UVB. Nevertheless, our findings demonstrated that UVB activates NF-κB in skin cells through a signaling pathway initiated from the cell surface receptor RAGE and HMGB1 stimulation from the skin microenvironment, which is distinct from the atypical nucleus-to-cytoplasm NF-κB activation signaling utilized by other genotoxic agents, including UVC.

A number of transcriptional regulators have been shown to modulate PD-L1 expression. In the tumor microenvironment, PD-L1 can be induced by inflammatory cytokines such as IFNγ, which induces activation of STAT1 and increased PD-L1 transcription (6, 14). Other transcriptional regulators, including HIF1, Myc, STAT3, AP-1, and NF-κB, have also been shown to enhance PD-L1 transcription in response to a variety of upstream signals (45). In addition, NRF2 was recently reported to be required for UVR-induced PD-L1 upregulation in primary melanocytes and keratinocytes isolated from the human skin (46). Our data indicated that PD-L1 induction by UVB was minimally affected in NRF2-depleted human melanoma cells examined, suggesting a potential cell-type–specific mechanism for regulating PD-L1 transcription in primary cells. We found NF-κB, in association with IRF3, plays an essential role in enhancing PD-L1 transcription in melanoma cells upon UVB. Also, NF-κB and IRF3 were indispensable for PD-L1 upregulation by UVB in primary melanocytes, in which they may mediate PD-L1 transcription with NRF2 alternatively and/or collaboratively. Besides directly promoting PD-L1 gene transcription, NF-κB has been shown to indirectly stabilize PD-L1 by upregulating the deubiquitinase CSN5, which inhibits the ubiquitination and proteasomal degradation of PD-L1 in breast cancer cells (33). We showed that, in melanoma cells, both the mRNA and protein levels of PD-L1 were increased after UVR. Moreover, the increased PD-L1 transcription by UVR was attenuated by inhibiting NF-κB upstream kinases IKKβ or TBK1. We further confirmed the enrichment of NF-κB on the PD-L1 promoter in melanoma cells after UVR. All these findings support that UVB-induced NF-κB activation directly increases PD-L1 transcription in melanocytes and melanoma cells.

Intriguingly, we found that IRF3 also plays a critical role in promoting PD-L1 induction by UVR by forming a transcriptional complex with NF-κB/p65. To our knowledge, the transcriptional regulatory role for IRF3 in PD-L1 expression has not been reported. A recent study showed that DNA damage–induced PD-L1 upregulation in breast cancer cells was regulated in a STING-dependent manner (47). As the primary downstream transcriptional regulator of STING, IRF3 likely plays a role in regulating PD-L1 transcription in breast cancer cells treated with chemotherapeutic drugs. Furthermore, our previous studies demonstrated that NF-κB is activated upon chemotherapeutic treatment in breast cancer cells (48). It is tempting to speculate that NF-κB may also collaborate with IRF3 in promoting PD-L1 transcription in breast cancer cells treated with chemotherapeutics. IRF3 was previously showed to interact with NF-κB/p65, and the IRF3–p65 complex was required for transactivation of IRF3-target genes in response to LPS (33). In UVB-treated melanocytes, IRF3 was recruited onto the NF-κB–binding site within PD-L1 promoter through association with p65, which coordinately promoted PD-L1 transcription. The finding that the upstream kinase TBK1 is responsible for activation of both NF-κB and IRF3 upon UVR may have provided an attractive pharmacologic target for antagonizing PD-L1 induction by UVR by simultaneously blocking the activation of NF-κB and IRF3.

Cancer immune surveillance plays an essential role in preventing tumor incidence and effectively eliminates oncogenic cells at precancerous stages (8). Therefore, the occurrence of many cancers can be considered a breach of immune control, which is fueled by additional oncogenic mutations. The immune-suppressive landscape caused by UVR provides a permissive environment for premalignant cells to escape from tumor immunity and promote skin cancer initiation and progression. In addition to the well-established local and systemic immunosuppressive mechanisms by UVR (4), our studies indicated that activating PD-1/PD-L1 immune checkpoints by upregulating PD-L1 in melanocytes and melanoma cells also play an essential role in facilitating immune evasion of malignant skin cells from tumor immunity. Genome-wide examination of neonatal melanocytes after UVR identified an interferon gene signature, which promotes melanocytic cell survival and immune evasion (9). Secretion of HMGB1 from skin cells in response to UVR may promote an immunosuppressive microenvironment by activating PD-1/PD-L1 checkpoint, along with other previously reported immunosuppressive cytokines, such as IL4, IL10, PGE2, and TGFβ. Our data indicate that blocking PD-1/PD-L1 significantly enhanced CD8⁺ T-cell–mediated antitumor immunity against melanoma cells after UVB exposure, suggesting that immune checkpoint blockers may also mitigate UVR-promoted immune evasion during melanoma development. Treatment with anti-PD-1 blocking antibody has been shown to increase the number of CD8⁺ memory T cells in the tumor microenvironment (49). Using small molecules, such as BMS-202 (50), which inhibit the PD-1/PD-L1 interaction, may serve as an immune prevention approach to mitigate immune suppression by UVR, increase effector and memory T-cell skin infiltration, and antagonize oncogenic cell immune evasion, thereby preventing the initiation and progression of invasive melanoma after UVR.

No potential conflicts of interest were disclosed.

Conception and design: W. Wang, N.M. Chapman, R.N. Laribee, L.M. Pfeffer, Z.-H. Wu

Development of methodology: W. Wang, N.M. Chapman, Z.-H. Wu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): W. Wang, N.M. Chapman, B. Zhang, Z.-H. Wu

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): W. Wang, N.M. Chapman, M. Fan, L.M. Pfeffer, Z.-H. Wu

Writing, review, and/or revision of the manuscript: W. Wang, N.M. Chapman, B. Zhang, M. Fan, R.N. Laribee, M.R. Zaidi, L.M. Pfeffer, H. Chi, Z.-H. Wu

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Li, L.M. Pfeffer, Z.-H. Wu

Study supervision: Z.-H. Wu

We thank Dr. Francesca-Fang Liao for providing SIRT1 plasmids. This work was supported in part by the National Cancer Institute (R01 CA149251), American Cancer Society (RSG-13-186-01-CSM), and NGS and CORNET awards from UTHSC.

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.