Vitamin D–VDR Signaling Inhibits Wnt/B-Catenin– Mediated Melanoma Progression and Promotes Antitumor Immunity

Published OnlineFirst November 5, 2019; DOI: 10.1158/0008-5472.CAN-18-3927

Cancer Tumor Biology and Immunology Research

Vitamin D–VDR Signaling Inhibits Wnt/b-Catenin– Mediated Melanoma Progression and Promotes Antitumor Immunity Sathya Muralidhar1,2, Anastasia Filia3,Jer emie Nsengimana1, Joanna Pozniak 1,4,5, Sally J. O'Shea1,6,7, Joey M. Diaz1, Mark Harland1, Juliette A. Randerson-Moor1, Jorg€ Reichrath8, Jonathan P. Laye1, Louise van der Weyden9, David J. Adams9, D.T. Bishop1, and Julia Newton-Bishop1

Abstract

1a,25-Dihydroxyvitamin D3 signals via the vitamin D and gene deletion in metastases. Vitamin D deficiency (<25 receptor (VDR). Higher serum vitamin D is associated with nmol/L 10 ng/mL) shortened survival in primary mela- thinner primary melanoma andbetteroutcome,althougha noma in a VDR-dependent manner. In vitro functional causal mechanism has not been established. As patients validation studies showed that elevated vitamin D–VDR with melanoma commonly avoid sun exposure, and con- signaling inhibited Wnt/b-catenin signaling genes. Murine sequent vitamin D deficiency might worsen outcomes, we melanoma cells overexpressing VDR produced fewer pul- interrogated 703 primary melanoma transcriptomes to monary metastases than controls in tail-vein metastasis understand the role of vitamin D–VDR signaling and rep- assays.Insummary,vitaminD–VDR signaling contributes licated the findings in The Cancer Genome Atlas metastases. to controlling pro-proliferative/immunosuppressive Wnt/ VDR expression was independently protective for melano- b-catenin signaling in melanoma and this is associated with ma-relateddeathinbothprimaryandmetastaticdisease. less metastatic disease and stronger host immune responses. High tumor VDR expression was associated with upregula- This is evidence of a causal relationship between vitamin D– tion of pathways mediating antitumor immunity and cor- VDR signaling and melanoma survival, which should be responding with higher imputed immune cell scores and explored as a therapeutic target in primary resistance to histologically detected tumor-infiltrating lymphocytes. High checkpoint blockade. VDR–expressing tumors had downregulation of prolifer- Significance: VDR expression could potentially be used as ative pathways, notably Wnt/b-catenin signaling. Deleteri- a biomarker to stratify patients with melanoma that may ous low VDR levels resulted from promoter methylation respond better to immunotherapy.

Introduction antiproliferative genes in prostate (5), breast (6), colon (7), squamous cell carcinoma, and leukaemia cells (8, 9). 1a,25-Dihydroxyvitamin D is the ligand for the dimeric 3 We have previously reported that higher serum vitamin D levels vitamin D receptor (VDR) and retinoid X receptor (RXR): at recruitment were associated with lower American Joint Com- ligand–receptor binding facilitates transcription of target genes mittee on Cancer (AJCC) stage and better melanoma-specific containing the vitamin D response element (1). The physio- survival (MSS) in the Leeds Melanoma Cohort (LMC; ref. 10). logiceffectofthevitaminD–VDR signalingaxisisoftentarget This was subsequently verified in five additional studies (11–15), tissue specific(2). which collectively indicate a significant role for vitamin D–VDR The association of low serum 25-hydroxyvitamin D levels 2/3 signaling in melanoma progression. (henceforth referred to as vitamin D) with higher cancer incidence The unique dataset used in this study was derived from 703 has been reported (3), but the significance of the association has formalin-fixed, paraffin-embedded (FFPE) primary melanomas, been debated (4). Extensive in vitro evidence however indicates an from the LMC, a population-based and extensively annotated antiproliferative role of vitamin D, with 1,25(OH) vitamin D 2 3 cohort with a long follow-up (10). Tumor-derived transcriptomic treatment shown to induce expression of proapoptotic genes and data, clinical, histopathologic, and whole-genome copy number

1University of Leeds School of Medicine, Leeds, United Kingdom. 2Division of Note: Supplementary data for this article are available at Cancer Research Molecular Pathology, The Institute of Cancer Research, London, United King- Online (http://cancerres.aacrjournals.org/). dom. 3Centre for Translational Research, Biomedical Research Foundation of the Corresponding Author: Julia Newton-Bishop, University of Leeds, St. James's Academy of Athens, Athens, Greece. 4Laboratory for Molecular Cancer Biology, University Hospital, Beckett Street, Leeds LS9 7TF, United Kingdom. VIB Center for Cancer Biology, KU Leuven, Leuven, Belgium. 5Department of Phone: 4401-1320-64668; E-mail: [email protected] Oncology, KU Leuven, Leuven, Belgium. 6Faculty of Medicine and Health, University College Cork, Cork, Ireland. 7Mater Private Hospital Cork, Citygate, Cancer Res 2019;79:5986–98 Mahon, Cork, Ireland. 8Center for Clinical and Experimental Photodermatology, doi: 10.1158/0008-5472.CAN-18-3927 The Saarland University Hospital, Homburg, Germany. 9Wellcome Sanger Insti- tute, Wellcome Genome Campus, Hinxton, Cambridge, United Kingdom. 2019 American Association for Cancer Research.

5986 Cancer Res; 79(23) December 1, 2019

Vitamin D–VDR Signaling in Primary Melanoma Prognosis

alteration (CNA) data were jointly analyzed to assess the pan- the expression of each gene and VDR in the tumors. Genes with genome effects of vitamin D–VDR signaling and to determine the FDR < 0.05 and |Reg Coef| > 0.2 (regression coefficient; Reg Coef) processes most associated with this pathway. Importantly, these were plotted in a volcano plot (R function: "plot") and used for findings were replicated in The Cancer Genome Atlas (TCGA) functional enrichment analysis. metastatic melanomas and then functionally validated using Because VDR is expressed by keratinocytes (18), a FLG2- in vitro and in vivo experiments. adjusted whole-transcriptome correlation with VDR was addi- tionally performed [Fillagrin family member 2 (FLG2) being a maker of keratinocyte differentiation). This sensitivity analysis Materials and Methods was conducted to account for any bias in VDR expression, which The LMC tumor transcriptome might have originated from keratin-rich melanoma subsets, LMC tumor transcriptome processing was described previously which have been reported previously (19–21). (accession no. EGAS00001002922; ref. 16). Briefly, tumor samples were taken from FFPE primary melanomas and RNA was Functional enrichment analyses extracted to generate whole-genome gene expression data (Illu- The "Gene set/mutation analysis" feature of ReactomeFIViz mina DASL HT12.4 array). Background correction and quantile (22) was used to identify significantly enriched pathways (Ben- normalization were applied; singular value decomposition was jamini–Hochberg FDR < 0.05, from hypergeometric test) from a used to assess the confounding factors that were subsequently given input gene list. adjusted out. Participants in the LMC gave written informed consent; the study was conducted in accordance to international VDR-binding regions ethical guidelines (Declaration of Helsinki) and was approved by The genomic regions identified as having VDR-binding peaks the national ethics committee (MREC 1/03/57 and PIAG3-09(d)/ across six tissue types (23) were downloaded as BED files. In 2003). addition, genomic regions known to contain the VDR-binding motif were downloaded from Motifmap (as BED files). In both Measurement of serum vitamin D at diagnosis cases, genes associated with genomic regions were identified using For 554 of the 703 participants with transcriptomic data, 25- GREAT 3.0.0 (24): "basal plus extension" approach was used with hydroxyvitamin vitamin D2 and D3 (nmol/L) was measured as Human GRCh37 assembly, whole genome as background described previously (10) and adjusted for season (see Supple- regions, and the gene regulatory domain set to 20 Kb upstream mentary Materials and Methods). and 400 Kb distal. The genes that mapped to these regions ("region–gene associations") were exported and their overlap TCGA melanoma dataset with VDR-correlated genes in the LMC (at FDR < 0.05) was TCGA metastatic melanoma data (n ¼ 365 samples) such as assessed. transcriptomic (RNA-seq), clinical, methylation, and copy number data were downloaded from cBioPortal (http://www.cbiopor VDR expression across Lund and TCGA molecular phenotypes tal.org/). The same statistical tests and software/packages were Nearest centroid method (21) was used to classify the LMC used to analyze Leeds and TCGA data. tumors into TCGA and Lund molecular melanoma phenotypes (see Supplementary Materials and Methods). Differential VDR Statistical analyses expression across these subgroups was tested with the Mann– Association of VDR expression with clinical variables. Univariate Whitney test. and multivariate linear regression were used to test the association between VDR and AJCC stage, mitotic number, tumor site, age, Imputed immune scores and sex. As we described previously (25), briefly, 28 immune cell scores were calculated as mean expression of genes pertaining to an Association of VDR expression with MSS. Univariate and multivar- immune cell type, after deducting genes identified as potentially iate Cox proportional hazards models were used to test the nonimmune cell specific, from the initial immune gene signature prognostic effect of VDR expression level after adjusting for AJCC described by Angelova and colleagues (26). Correlation analysis stage (7th Edition), tumor mitotic number, tumor site, and of each immune cell score with VDR expression was conducted. tumor-infiltrating lymphocytes (TIL). For this analysis VDR expression was on the continuous log2 scale, meaning that HR Expression of VDR and response to checkpoint blockade per VDR unit corresponds to change in hazard when expression is VDR expression was compared between responders and non- doubled. responders in two published albeit small transcriptomic data sets: (i) 38 patients treated with PD-1 blockade (27) and (ii) 40 VDR copy number changes in the LMC tumors patients treated with CTLA4 blockade (28). Both studies used Copy number profiles were estimated from next-generation pretreatment biopsies. Fold change was computed as ratio of sequencing output from sequenced DNA samples from a subset of mean VDR expression in responders to nonresponders. LMC tumors (n ¼ 276, 39%) as described previously (17). Gistic2.0 was used to identify VDR copy number estimates (see Vitamin D-VDR subgroup analysis Supplementary Materials and Methods). X-tile (29) was used to identify patient subgroups with the most contrasted survival profiles (melanoma specific) based on their Whole-transcriptome correlation with VDR tumor VDR expression in the LMC. This approach was trained in Linear regression analysis and Benjamini–Hochberg multiple randomly selected two-thirds of the samples and validated in the testing correction (FDR) was used to test the correlation between remaining third. The cut-off points defining patient subgroups in

www.aacrjournals.org Cancer Res; 79(23) December 1, 2019 5987

Muralidhar et al.

terms of percentiles were applied to the TCGA metastatic mela- 5-mm sections were cut and used for hematoxylin and eosin noma dataset for replication. (H&E) and CD3 staining. H&E sections were digitally scanned The identified VDR groups were further stratified on the basis of (Leica Aperio AT2) and total metastatic area ("met area" mm2) was the participants' serum vitamin D levels at recruitment (season calculated as the total area of all metastatic deposits, across all five þ adjusted) in the LMC. Clinically defined, vitamin D deficiency is lung lobes using Aperio Imagescope (Leica Biosystems). CD3 generally considered to be <25 nmol/L (30), with recent evidence lymphocytes were estimated as described in Supplementary Mate- defining vitamin D deficiency able to compromise bone health as rials and Methods. 30 nmol/L (31). Thus, in the LMC dataset, serum vitamin D levels < 25 nmol/L and 25 nmol/L were classified as "deficient" Wnt/b-catenin signaling in VDR-B16BL6 and control-B16BL6 and "sufficient." cells cDNA from V1, V2, C1, and C2 cells were analyzed using a Vitamin D treatment of human melanoma cells RT-PCR array of 84 mouse Wnt/b-catenin pathway genes (see SK-MEL-28 and MeWo cells (courtesy of Professor Alan Supplementary Materials and Methods). Relative expression was Melcher, Leeds CRUK cell line bank) were authenticated using calculated using the DDCt method, normalized to average Ct of the PowerPlex 16 System (Promega), cultured, treated with 1,25 the five housekeeping genes provided in the array. Fold change (OH)2 vitamin D3 (Supplementary Materials and Methods), (FC) of the VDR-B16BL6 clones relative to control-B16BL6 DDC collected at different time points, RNA extracted, and used to tV1ðorÞV2 clones was calculated as follows: FCV1ðorÞV2 ¼ 2 , where generate gene expression profiles (HG-U133 plus 2.0 array, Affy- DDCt ¼ DCt DCt : metrix). Genes differentially expressed in vitamin D–treated ver- V1ðorÞV2 V1ðorÞV2 avgðC1 & C2Þ sus control cells were identified at each timepoint (FDR < 0.10) and were used for functional enrichment using Reactome FIViz. Results Generation of VDR-B16BL6 and control-B16BL6 cells VDR expression is independently protective for melanoma B16BL6 cells (chosen owing to lowest endogenous VDR expres- death in the LMC primary and TCGA metastatic melanomas sion among B16 strains, estimated from transcriptomic data by VDR expression was significantly lower in tumors of higher Dr. Martin del Castillo, Wellcome Sanger Institute, Hinxton, AJCC stage, higher mitotic rate, and tumors on the trunk and sun United Kingdom, personal communication) were purchased protected sites (compared with tumors on the head; Univariate from the M.D. Anderson Cancer Center Cell Line Core facility. analysis; Table 1) in the 703 LMC primaries. In multivariate Cells were screened for the presence of Mycoplasma and mouse analysis, lower VDR expression was independently associated P ¼ P ¼ pathogens (at Charles River Laboratories) before culturing. Early with higher mitotic rate ( 0.001) and tumor site ( passage B16BL6 cells were cultured and transfected with murine 0.001 for tumors of trunk compared with those on head), with fi P VDR cDNA (synthesized by GeneArt) or empty backbone plasmid borderline signi cance for higher stage ( < 0.06; Table 1). to generate the VDR-B16BL6 (clones V1 and V2) and control- Higher tumor VDR expression was protective for melanoma- B16BL6 (clones C1 and C2) cells, respectively (Supplementary related death independent of stage, tumor mitotic rate, tumor site, Materials and Methods), verified by Western blot analysis and and histologically reported TILs (HR for melanoma-related ¼ P ¼ VDR qRT-PCR. death 0.80; 0.008; Table 1). Although expression correlated significantly with other members of the NR1lL family In vivo tail-vein metastasis assay such as LXRB, FXR1, FXR2, and PXR expressions (Supplementary , fi Although subcutaneous B16 models have been used to dem- Table S1) it remained signi cantly prognostic after adjusting for onstrate the role of natural killer (NK) cells in host responses to the expression of those genes (adjusted HR for melanoma-related ¼ P ¼ melanoma, we adopted the tail-vein model based on successful death 0.77; 0.001). demonstration by the group of a role for effector T cells and NK Because VDR forms heterodimers with RXR and RXRg signaling – cells in B16 pulmonary metastases studies (32) and other reports has been reported to drive epithelial mesenchymal transition of a role for T cells in B16 melanoma cells (33–35). The care and and invasion in melanoma, we tested the prognostic effect of the VDR use of all mice in this study were in accordance with the United expression of genes coding for RXR receptors. In the LMC, did Kingdom Animals in Science Regulation Unit's Code of Practice not correlate significantly with RXRg (Supplementary Table S1) fi VDR for the Housing and Care of Animals Bred, or Used for Scientific and the prognostic signi cance of expression was indepen- Purposes, the Animals (Scientific Procedures) Act 1986 Amend- dent of the expression of RXRa, RXRb, and RXRg (adjusted HR for ¼ P ¼ ment Regulations 2012, and all procedures were performed under melanoma-related death 0.75; 0.0001). a United Kingdom Home Office Project license, which was IHC of LMC primary melanomas revealed that VDR expression reviewed and approved by the Sanger Institute's Animal Welfare was predominantly in tumor (rather than immune) cells (repre- and Ethical Review Body. Housing and husbandry conditions sentative images in Supplementary Fig. S1A–S1C, quantified in were as described previously (32) and mice were maintained on Supplementary Table S2). Mouse Breeders' Diet (Laboratory Diets, 5021-3) throughout the In the TCGA metastatic melanomas (n ¼ 353), higher VDR ¼ study. V1, V2, C1, and C2 cells (detailed above) were tail-vein expression was protective for death (overall survival HR 0.82; administered to 6- to 10-week-old sex-matched wild-type P ¼ 0.03). (C57BL/6NTac) mice (104 cells in 0.1 mL PBS). After 21 days, mice were humanely sacrificed and their lungs macroscopically Tumor VDR expression is associated variably with CNAs and examined to determine the number of metastatic deposits in all promoter methylation in primary and metastatic melanomas five lobes ("met count"). Lungs were formalin-fixed, processed, In the LMC primary melanomas, VDR expression showed a and embedded in paraffin wax blocks, from which consecutive weak positive correlation with VDR copy number, which failed to

5988 Cancer Res; 79(23) December 1, 2019 Cancer Research

Vitamin D–VDR Signaling in Primary Melanoma Prognosis

Table 1. Association of tumor VDR expression with clinicopathologic variables and MSS in the 703 LMC participants Univariate Multivariate Correlation of tumor VDR expression with Regression coefficient SE P Regression coefficient SE P Age (per year) 0.005 0.002 0.04 0.003 0.002 0.14 Sex Femalesa Males 0.18 0.06 0.003 0.09 0.06 0.16 AJCC stage Stage Ia Stage II 0.17 0.07 0.012 0.12 0.07 0.08 Stage III 0.25 0.09 0.009 0.18 0.09 0.06 Mitotic rate <1 mitoses/mm2 tumora 1 mitoses/mm2 tumor 0.23 0.06 0.0004 0.20 0.06 0.001 Tumor site Heada Limbs 0.04 0.10 0.63 0.11 0.10 0.28 Trunk 0.35 0.10 0.001 0.36 0.10 0.001 Rare (sun-protected sites) 0.44 0.12 0.001 0.38 0.13 0.003 Univariate Multivariateb HR SE P HR SE P Effect of tumor VDR expression on MSS (per unit expression) 0.75 0.05 0.0001 0.80 0.06 0.008 aBaseline group used for comparison with relevant groups. bMultivariate survival analysis was adjusted for AJCC stage, mitotic rate, tumor site, and tumor immune infiltrate. Regression coefficient and P value from linear regression.

reach statistical significance (P ¼ 0.12; Fig. 1A). However, in the associated with 73 genes (GREAT, Supplementary Materials and TCGA metastatic melanomas, VDR expression was lowest in Methods), of which, 43 genes (58%) were among the significant tumors with hemizygous deletion compared with those with "no VDR-correlated genes in the LMC. Alternatively, 60% of genes changes/neutral" (P ¼ 0.01) and was significantly higher in mapped to genomic regions containing the VDR-binding motif tumors with "gain" (P ¼ 0.007) and "high-level amplifications" (identified by Motifmap, see Materials and Methods) also corre- (P < 0.00001; Fig. 1B). VDR copy number was more frequently lated significantly with VDR in the LMC. reduced in distant metastases compared with regional lymph In the TCGA metastatic melanomas, VDR correlated negatively node metastases (P ¼ 0.015; Fig. 1C). Concomitantly, the TCGA with genes enriched for: ECM organization, cadherin signaling, data also showed progressively reduced VDR expression from eukaryotic translation initiation, TGFb, and VEGFR1 signaling; primary to lymph node than distant metastases (Fig. 1D). VDR and positively with: NF-kB, TNF, IFNa/b, IFNg, IL12-mediated, expression was significantly and inversely correlated with VDR T-cell receptor, and chemokine signaling in na€ve CD4 T-cells promoter methylation (P ¼ 0.0001; Fig. 1E) in TCGA metastases. pathways (Supplementary Tables S6 and S7). The majority over- lapped with those observed in LMC primaries. VDR correlates positively with genes enriched for immune- related pathways and negatively with proliferation-related Tumor VDR expression is associated with reported melanoma pathways in LMC primary and TCGA metastases phenotypes, imputed immune cell scores, and reduced Wnt/ In the LMC, an agnostic whole-transcriptome correlation anal- b-catenin signaling ysis identified genes positively (n ¼ 2,025) and negatively (n ¼ The LMC primaries were classified on the basis of previously 3,408) correlated with VDR (Fig. 2). The negatively correlated described melanoma molecular phenotypes (Supplementary genes were enriched for proliferation-related pathways such as Materials and Methods; refs. 20, 36). The TCGA signature (20) mitotic prometaphase, Wnt signaling, mitochondrial translation, classified the 703 LMC melanomas into immune (n ¼ 192), tricarboxylic acid cycle, and cadherin signaling (Fig. 2; keratin (n ¼ 247), and MITF-low (n ¼ 150) subtypes. VDR Supplementary Table S3). In contrast, the positively correlated expression was significantly higher in the keratin and immune genes were enriched for immune-related pathways such as cyto- subtypes compared with the MITF-low subtype (P ¼ 1.1 kine–cytokine receptor interaction, TNF, IFNg, IL12-mediated, 10 6; Fig. 3A). The Lund signature (36) classified tumors into NF-kB, and chemokine signaling (Fig. 2; Supplementary high-immune (n ¼ 173), normal-like (n ¼ 198), pigmentation Table S4). VDR-correlated genes remained largely unchanged (n ¼ 222), and proliferative (n ¼ 83) subtypes. VDR expression after FLG2 adjustment (Materials and Methods; Supplementary was significantly higher in high-immune subtype compared with Table S5), indicating that confounding from epidermal sampling the poorer prognosis proliferative (P ¼ 7.5 10 8) and pigmen- is unlikely. tation subtypes (P ¼ 6 10 13; Fig. 3B). We assessed whether the VDR-correlated genes in the LMC were VDR expression correlated positively with 25 of the 26 immune known to have a VDR-binding site. Tuoresmaki and colleagues cell scores (Supplementary Table S8), of which, the strongest previously reported 54 nonoverlapping genomic VDR-binding correlation (correlation coefficient > 0.30) was with dendritic regions recurrent in six tissue types, based on meta-analysis of cells, myeloid-derived suppressor cells, neutrophils, central mem- VDR chromatin immunoprecipitation-sequencing (ChIP-seq) ory CD4, NK, Th1, Th2, and T cells (Fig. 3C). Concordantly, VDR data (23). We mapped the 54 genomic-binding regions to be expression was significantly lower in tumors with "absent"

www.aacrjournals.org Cancer Res; 79(23) December 1, 2019 5989

Muralidhar et al.

A B P < 0.00001 P = 0.007

P = 0.01

)

expression VDR

expression (log expression VDR

Corr. coef = 0.09, P = 0.12

4 4 5 6 7 8 9 2 4 6 8 10 10 8 6 4 2 −1 −0.5 0 0.5 1 0 1 2 3 LMC- VDR copy number TCGA- VDR copy number (GISTIC measures)

CDP = 0.001 P = 0.008

P = 0.015

4 6 8

expression VDR

TCGA- VDR copy number copy VDR TCGA-

0 0 2

1 1 0 1 2 − Regional lymph Distant Distant Regional lymph Primary node metastases Metastases Metastases node metastases tumors

)

expression (log expression VDR

Corr. coef = –0.3, P = 0.0001 2 4 6 8 10 12 10 8 6 4 2 0 0.2 0.4 0.6 0.8 1 VDR methylation beta value

Figure 1. Association of VDR expression with copy number and methylation. A, Correlation of VDR expression with copy number in 276 LMC primaries. B, VDR expression across categorical copy number estimates from TCGA melanoma metastases: 0, no change; 1, gain; 2, high-level ampliﬁcation. C, VDR median copy number in regional lymph node and distant metastases in the TCGA melanomas. D, VDR expression in distant, regional lymph node metastases, and primary tumors in the TCGA dataset. E, VDR expression and promoter methylation in TCGA metastases. Corr. coef, correlation coefﬁcient.

immune infiltrate compared with tumors with "non-brisk" (P ¼ coding for checkpoint molecules (16). VDR expression was lowest 0.02) and "brisk" immune infiltrate (P ¼ 0.004; Fig. 3D), accord- in that tumor subtype (Fig. 3E), which was concordant with the ing to histopathologic scores that were available for 601 (86%) of agnostic correlation analysis identifying Wnt/b-catenin as the the LMC melanomas. pathway most strongly associated with low immune signals in We previously reported an immunologically "cold" tumor sub- both LMC primary and TCGA metastatic melanomas. type in the LMC (Consensus Immunome Cluster 4) with increased To compare VDR expression with response to immunotherapy, b-catenin signaling, reduced imputed immune scores for cytotoxic, we used previously published therapy-response datasets (27, 28). T cell, and activated dendritic cells (aDC), and expression of genes VDR expression did not vary significantly between responders

5990 Cancer Res; 79(23) December 1, 2019 Cancer Research

Vitamin D–VDR Signaling in Primary Melanoma Prognosis

Figure 2. Whole-transcriptome correlations with VDR expression. Volcano plot of genes correlated significantly (FDR < 0.05), positively (regression coefficient > 0.2, green dots), and negatively (regression coefficient < 0.2, red dots) with VDR expression in the LMC data; top 25 pathways enriched for each gene list are listed in adjoining tables (left, negative correlates; right, positive correlates). FDR, Benjamini–Hochberg FDR.

(n ¼ 15) and nonresponders (n ¼ 13) to anti-PD-1 therapy (P ¼ predicted differential survival identified the following groups: 0.27; Supplementary Fig. S2B), nor to anti-CTLA4 treatment (P ¼ 17% with the lowest VDR expression (low-VDR group, n ¼ 119), 0.12) in a dataset of responders and (n ¼ 13) nonresponders (n ¼ 17% of tumors with highest VDR expression (high-VDR group, 22; Supplementary Fig. S2A), although in the latter, VDR expres- n ¼ 119), and middle 66% (intermediate-VDR group, n ¼ 465) sion was 1.35-times higher in responders compared with non- having the worst, best, and intermediate survival, respectively responders (FC ¼ 1.35). (P ¼ 5.2 10 8; Fig. 4A and B). This was replicated in the TCGA metastatic melanomas (n ¼ 353) using the same VDR expression Deficient levels of serum vitamin D are associated with more percentiles (P ¼ 0.03; Fig. 4C and D). melanoma-related deaths within the context of VDR expression Among the three VDR groups: deficient serum vitamin D levels The 703 LMC primary melanomas were stratified into three (<25 nmol/L) were associated with poorer prognosis compared groups using a survival-based stratification approach (X-tile, see with sufficient vitamin D levels (25 nmol/L) in the intermedi- Materials and Methods). VDR expression thresholds that best ate-VDR group (HR ¼ 1.73; P ¼ 0.02), but not in the low-VDR

www.aacrjournals.org Cancer Res; 79(23) December 1, 2019 5991

Muralidhar et al.

P = 7.5×10−8 P −6 A = 1.07×10 B P = 6×10−13 P

= 0.10 P = 0.0003

)

expression (log expression VDR

2 4 6 8 8 4 6 2 10 2 4 6 8 8 4 10 6 2

Immune Keratin MITF-low High-immune Normal-like Pigmentation Proliferative N N N N N = 192 N = 247 N = 150 = 173 = 198 = 222 = 83

Figure 3. C D P = 0.004 P = 0.02 Association of VDR expression with

measures of immune response. )

2 A and B, Comparison of VDR expression between LMC samples

classiﬁed into Lund and TCGA )

2 Dendritic Central memory NK cells prognostic molecular phenotypes, cells CD4 cells C, Correlation of VDR expression with imputed immune cell scores

4 6 8 8 4 10 6 whose correlation coefﬁcient > 0.3.

VDR expression (log expression VDR D, expression across

pathologist-graded TILs in the 2 2 whole-tumor FFPE section. T cells Th1 cells

MDSCs Absent Non-brisk Brisk P t expression (log expression VDR values are from the Student test. N = 45 N = 492 N = 64 E, Heatmap depicting expression of Tumor immune infiltrate VDR expression (top), cytotoxic, T-cell and aDC scores, expression of genes coding for checkpoint molecules, and expression of Th2 cells Neutrophils Wnt/b-catenin signaling genes across the six consensus E High VDR (17%) immunome clusters (CIC) Intermediate VDR (66%) reported previously (16). Low VDR (17%) Cytotoxic T-cell aDC PD1 PDL1 PDL2 CTLA4 1.00 CTNNB1 0.50 cMYC 0.00 SOX11 –0.50 SOX2 –1.00 EFNB3 APC APC2 DKK2 DKK3

(P ¼ 0.66) or high-VDR (P ¼ 0.55) groups (Fig. 4E). The vitamin D sufficiency: NK-cell–mediated cell killing (P ¼ 0.02), deleterious association with vitamin D deficiency was therefore IL12 (P ¼ 0.03), and T-cell receptor signaling on na€ve CD4 and apparent within the context of VDR expression. Intermediate- CD8 signaling (P ¼ 0.05). VDR group participants with deficient vitamin D were characterized by higher Breslow thickness (P ¼ 0.02), higher frequency of Vitamin D treatment and increased VDR expression inhibit pathologist-reported vascular invasion (P ¼ 0.01), and AJCC stage Wnt/b-catenin pathway and melanoma cell growth in vitro and II tumors (compared with stage I: P ¼ 0.01), when compared with in vivo those with sufficient vitamin D. An agnostic whole-transcriptome We functionally validated our findings from the LMC primary analysis identified no gene significantly differentially expressed and TCGA metastatic melanomas, where high VDR-expressing (at FDR < 0.10) between participants with deficient or sufficient tumors (with active vitamin D–VDR signaling) had better survival vitamin D in the intermediate-VDR group. However, among the and reduced expression of proliferative pathways, in particular the pathways that correlated significantly with VDR (Fig. 2), the Wnt/b-catenin signaling. The human melanoma cell lines, SK- following were significantly underexpressed in participants with MEL-28 and MeWo, treated with 1,25 (OH)2 vitamin D3 showed

5992 Cancer Res; 79(23) December 1, 2019 Cancer Research

Vitamin D–VDR Signaling in Primary Melanoma Prognosis

A B 1.00

Intermediate-VDR N = 463

High-VDR N = 118

Number of participants of Number Low-VDR

N = 119 P = 5.2 × 10−8

Melanoma-specific survival (%) survival Melanoma-specific

0 50 100 100 50 0 150

0.75 0.50 0.25 0.00 0 5 10 15 0 2 4 6 8 10 Number at risk VDR Low-VDR expression (log2) 116 52 19 0 440 268 104 0 Intermediate-VDR 110 84 36 0 High-VDR Years after primary diagnosis

C D

Intermediate- VDR

N = 270

0.50 0.75 0.75 0.50 1.00

Low-VDR N = 71 High-VDR

N = 69

Overall survival (%) survival Overall

0.25 Number of participants of Number

P = 0.03

0 10 10 20 0 30 40 0.00 0.00 0 2 4 6 8 0 5 10 15 Number at risk VDR expression (log ) 2 76 300Low-VDR 282 25 4 0 Intermediate-VDR 71 8 1 0 High-VDR Survival time (years) E

n = 77 n = 15

n = 303

n = 77 n = 16

n = 58

0.25 0.50 0.75 0.75 1.00 0.50 0.25

Melanoma-specific survival (%) survival Melanoma-specific 0.00 0.00 0 5 10 15 Number at risk 16 8 30Low-VDR, low vitD 75 38 13 0 Low-VDR, high vitD 53 28 6 0 Intermediate-VDR, low vitD 292 210 86 0 Intermediate-VDR, high vitD 14 13 7 0 High-VDR, low vitD 73 60 24 0 High-VDR, high vitD Years after primary diagnosis

Figure 4. Vitamin D and VDR subgroup analysis in the LMC dataset. A, Prognosis-based stratification of 703 LMC tumors into low-, intermediate-, and high-VDR tumor groups identified using X-tile. B, Kaplan–Meier curves showing the differential melanoma-specific survival in those groups. C and D, Replication of the three VDR- groups and their association with overall survival in TCGA metastases. E, Melanoma-specific survival of six vitamin D-VDR subgroups produced by stratification of three VDR groups by low (<25 nmol/L) or high (25 nmol/L) serum vitamin D levels for winter season. All P values are from the Cox model. vitD, vitamin D.

reduced proliferation posttreatment (Fig. 5A). At 24 and 48 hours In a second functional validation model, the in vivo metastatic posttreatment, VDR expression was signiﬁcantly upregulated, potential of murine melanoma B16BL6 cells overexpressing VDR while Wnt signaling and ECM organization genes were among ("VDR-B16BL6 cells") was compared with control cells expressing the top downregulated pathways in both cell lines. The pathways low/no VDR ("control-B16BL6 cells"; Supplementary Fig. S3A upregulated in both cell lines at both time points were: MAPK, and S3B). Mice injected with VDR-B16BL6 cells developed sig- IFNa/b, TGFb, and TLR signaling (Fig. 5B). niﬁcantly lower pulmonary metastatic load (metastatic area

www.aacrjournals.org Cancer Res; 79(23) December 1, 2019 5993

Muralidhar et al.

A 140.0 120.0

100.0

80.0

60.0

40.0

Relative proliferation Relative 20.0

0.0 24 h 48 h 72 h 144 h

B MeWo SkMel28 24 h 48 h 24 h 48 h

PI3K-Akt signaling pathway Wnt signaling pathway

Extracellular matrix organization Signaling by VEGF MAPK signaling pathway Interferon alpha/beta signaling

Osteoclast differentiation Natural killer cell mediated cytotoxicity Interferon gamma signaling MHC class II antigen presentation TCR signaling Cytokine-cytokine receptor interaction Toll-like receptor signaling pathway(K) Jak-STAT signaling pathway(K) TGF-beta signaling pathway(K) Downreg p38 MAPK signaling pathway(N) mTOR signaling pathway(K) p38 MAPK signaling pathway(N) Upreg

C P = 0.04 C1 D C2 Wnt16 5e+06 Wnt2 V1 Wif1 4e+06 Fosl1 Wnt10a V2 Fzd8 Wnt8a Foxn1 Wnt8b

3e+06 Frat1 Wnt3 2 2 2.5 Nkd1 Wnt3a 2e+06 Lef1 Wnt4

Met area Met Mmp7 Wnt5a 1e+06 Strp1 Wnt5b Strp2 Wnt7a 0 Wnt11 Wnt7b

P = 0.0006

V1: FC- relative to control cells control to relative FC- V1:

0 0.5 1 1.5 1 0.5 0 40 0 0.5 1 1.5 2 V2: FC- relative to control cells

20 Met count Met

Control-B16BL6 VDR-B16BL6 n = 19 n = 15

Figure 5. Functional validation using in vitro and in vivo models. A, Relative proliferation of vitamin D–treated human melanoma cell lines SK-MEL-28 and MeWo: 24, 48, 72, and 144 hours after treatment, compared with ethanol-treated cells (control). B, Pathways enriched in genes upregulated (Upreg, green bars) or downregulated (Downreg, red bars) in response to vitamin D treatment of SK-MEL-2828 and MeWo cell lines at 24 and 48 hours posttreatment. C, Pulmonary metastatic loads in control-B16BL6 (clones C1 and C2) and VDR-B16BL6 (clones V1 and V2) obtained from the tail vein metastasis assay. Met count, macroscopic counts of metastases; met area, total microscopic metastatic area (mm2) in mice lung sections; N, number of mice/group. Inset, representative H&E–stained images of metastases-harboring pulmonary lobes. D, qRT-PCR array-derived expression of Wnt/b-catenin pathway genes in VDR-B16BL6 cells (V1 and V2; FC, fold change relative to control-B16BL6 cells). Red, genes with a FC < 0.5 for both VDR-B16BL6 clones V1 and V2.

and metastases count) compared with those dosed with control- experiments represented in Fig. 5C). Differential expression of B16BL6 cells (P < 0.04), with comparable results from both clones Wnt/b-catenin pathway genes between VDR-B16BL6 and control- in two independent experiments (pooled analyses from two B16BL6 was compared using a preformatted qRT-PCR–based

5994 Cancer Res; 79(23) December 1, 2019 Cancer Research

Vitamin D–VDR Signaling in Primary Melanoma Prognosis

array. Of the 84 Wnt/b-catenin genes tested, 62 genes had lower analysis adjusting for FLG2 expression produced no significant expression (FC < 1, of which, 26 had FC 0.5) in both VDR clones changes in the correlated pathways, indicating that artefact from compared with control clones (Fig. 5D). Twelve genes had keratinocyte-rich tumor populations was unlikely. VDR expres- increased expression (FC > 1, none with FC 2) in both VDR sion was higher in "brisk" immune-infiltrated primaries com- clones. pared with tumors with no immune infiltrate, which is in agree- In comparing the tumor immune infiltrate, although the num- ment with a previous report of a smaller IHC study (43) and is an þ ber of CD3 cells/100 mm2 met area was not statistically signif- independent validation of the transcriptomic imputation of icantly higher in mice injected with VDR-B16BL6 cells (P ¼ 0.11, immune infiltration. Fifty-eight percent of the genes within compared with control-B16BL6 cells), there was a trend for reported genomic VDR-binding regions (identified from VDR þ increased CD3 immune infiltrate (Supplementary Fig. S4). ChIP-seq data as well as VDR-binding motif) were found to Collectively, both models provide causal evidence that elevated correlate with tumor VDR expression in the LMC. This is consis- vitamin D–VDR signaling in melanomas cells inhibits Wnt/ tent with direct transcriptional control by the VDR transcription b-catenin signaling and tumor proliferation. factor for a proportion of differentially expressed genes. Validation using published prognostic melanoma molecular phenotypes (16, 36) revealed that VDR expression was signifi- Discussion cantly higher in high-immune subtypes compared with prolifer- Melanoma is one of the most immunogenic malignancies, ative subtypes, consistent with the view that the prognostic increased lymphocytic infiltration in both primaries (37, 38) and significance of VDR is associated with increased immune and metastases (39) is associated with improved outcomes (16, 21) decreased proliferative signaling. We also assessed whether VDR and melanoma responses to immune checkpoint blockade are preferentially correlated with a particular immune cell type, which high (40, 41). However, dampened immune responses and was previously uncharacterized in primary melanomas. However, therapeutic resistance attributed to oncogenic pathways such as VDR was strongly positively correlated with all imputed immune Wnt/b-catenin signaling (16, 42) mean that only 58% of patients cell scores, and we have previously reported simultaneous upre- with stage IV melanoma have significant benefit. Identification of gulation of adaptive and innate immunity in good prognosis factors that boost antitumor immunity is required to improve primary melanomas (16). Furthermore, the proimmune effect of outcomes. vitamin D–VDR signaling was supported by strong positive Low vitamin D levels are associated with thicker, poorer- correlation of VDR with genes involved in pathways such as prognosis primary melanomas (11), and lower VDR expression extracellular matrix organization, TNFa, NF-kB, IFNg, and is associated with melanoma progression (43, 44), but neither IL12-mediated signaling. causality nor the mechanistic basis has been established. In this In comparing pretreatment gene expression between respon- article, we demonstrate that vitamin D–VDR signaling is protec- ders and nonresponders to immunotherapy, VDR expression was tive for melanoma-related death at least, in part, through inhi- higher in responders to anti-CTLA4 (FC ¼ 1.35), albeit not bition of Wnt/b-catenin signaling, impacting on melanoma pro- statistically significant (P ¼ 0.12) in this very small dataset. liferation and antitumor immune response. Although data from these immunotherapy studies did not sup- We report that VDR expression was significantly lower in port VDR expression as a biomarker of response, we posit that the advanced tumors in the LMC primary melanomas. Importantly, datasets available to explore VDR as a biomarker were insufficient high VDR expression was independently protective of melanoma- to properly explore this possibility. related death after adjusting for AJCC stage, mitotic rate, and TILs. Wnt/b-catenin signaling was among the top negatively corre- Survival benefit was replicated in the TCGA metastatic tumors, lated pathways with VDR in the LMC and TCGA melanomas, highlighting the significance of vitamin D–VDR signaling in both concordant with reports of vitamin D-VDR–mediated inhibition primary and metastatic melanoma progression. This protective of Wnt/b-catenin signaling in colon cancer (47). We further effect of VDR was independent of the expression of other NR1L explored this relationship and report that VDR expression was family genes, despite reports of integrated activity between nucle- lowest in previously reported subset of tumors characterized by ar receptors (45). In assessing factors that could control VDR high Wnt/b-catenin expression, reduced immune infiltrate, and expression, VDR copy number was not significantly associated high mortality (16). Collectively, our findings support the with expression in LMC primaries. However, distant metastases hypothesis that, as in colon cancer, some of the effects of vitamin (which have worse prognosis) had lower VDR copy number D–VDR signaling in melanoma are mediated by inhibition of compared with regional lymph node metastases in the TCGA, Wnt/b-catenin signaling. suggesting a progression-associated genomic loss of VDR. Low Vitamin D deficiency (25 nmol/L, 10 ng/mL) was asso- VDR–expressing metastatic tumors from TCGA were also more ciated with worse prognosis only in participants with inter- likely to be hypermethylated, consistent with previous reports of mediate-VDR expression (albeit the majority). Although an the epigenetic control of VDR expression (46). Despite the pro- agnostic analysis identified no significant transcriptomic dif- gressive deletion of the VDR locus with tumor progression in ferences associated with vitamin D in this subgroup, there was metastatic disease, the data suggest that therapeutic manipulation some evidence for paradoxically reduced expression of of vitamin D–VDR signaling could have adjuvant therapeutic immune-associated pathways with higher vitamin D levels in benefit in primary disease where we saw little evidence of VDR a candidate gene expression analysis. The reported associations deletion. between vitamin D and the immune system are numerous and An agnostic correlation analysis revealed that VDR expression complex. The findings of this study are therefore not incon- was strongly positively correlated with immune-related pathways sistent with the view that vitamin D deficiency should be and negatively correlated with proliferation-associated pathways avoided but that high levels would not necessarily be bene- in LMC primaries and TCGA metastases. An additional sensitivity ficial to all patients.

www.aacrjournals.org Cancer Res; 79(23) December 1, 2019 5995

Muralidhar et al.

The lack of a protective effect of higher vitamin D levels in Disclosure of Potential Conflicts of Interest participants with low-VDR tumors was not unexpected as low S.J. O'Shea has received speakers bureau honoraria from L'Oreal Ltd UK, is an receptor expression could preclude effective signaling despite unpaid consultant/advisory board member for British Association of Derma- ligand sufficiency. A lack of benefit in the high-VDR tumors was tologists Research Subcommittee, and has provided expert testimony for Janssen Ltd (sponsorship for travel and subsistence to attend the EADV meeting more surprising and we postulate that this could reflect receptor in 2018, the BAD meeting in 2019, and EADV meeting in 2018) and Pierre Fabre saturation as reported in other NHR family receptors (48) or a Ltd (sponsorship for travel and subsistence to attend the Avene hydrotherapy ligand-independent effect of VDR, which has been described in centre in 2018). D.J. Adams is a consultant (paid consultant) for Microbiotica other cancers (49, 50). and reports receiving a commercial research grant from BMS. No potential In functional validation, the observed vitamin D-treatment– conflicts of interest were disclosed by the other authors. induced reduction in cell proliferation is concordant with previous reports (51). However, the pan-transcriptome–based findings Authors' Contributions that vitamin D treatment of human melanoma cell lines promotes Conception and design: S. Muralidhar, J. Reichrath, D.T. Bishop, J. Newton- VDR expression and inhibition of protumor pathways including Bishop VDR Development of methodology: S. Muralidhar, D.T. Bishop, J. Newton-Bishop the Wnt/b-catenin pathway, is novel. Furthermore, elevated Acquisition of data (provided animals, acquired and managed patients, expression in murine melanoma cells decreased their in vivo provided facilities, etc.): A. Filia, J. Nsengimana, S.J. O'Shea, M. Harland, metastatic potential, the expression of key Wnt/b-catenin genes, L. van der Weyden, D.J. Adams, D.T. Bishop, J. Newton-Bishop some of which (Dkk1 and Sfrp2) have previously been shown to Analysis and interpretation of data (e.g., statistical analysis, biostatistics, be inhibited by VDR (52), but not in melanomas. Interestingly, computational analysis): S. Muralidhar, A. Filia, J. Nsengimana, J. Pozniak, the "classic" noncanonical Wnt ligands Wnt5a, Wnt5b, Wnt10a, J.M. Diaz, M. Harland, L. van der Weyden, D.J. Adams, D.T. Bishop, J. Newton- Wnt7 Wnt11 fi Bishop , and were also downregulated by VDR. This nding Writing, review, and/or revision of the manuscript: S. Muralidhar, A. Filia, is of significance because Wnt5a (and some other noncanonical J. Nsengimana, J. Pozniak, S.J. O'Shea, J.M. Diaz, M. Harland, J.A. Randerson- Wnt ligands) affect cell motility and invasion and is implicated in Moor, J. Reichrath, J.P. Laye, L. van der Weyden, D.J. Adams, D.T. Bishop, worse melanoma prognosis (53). Thus, the findings functionally J. Newton-Bishop validate the transcriptome-derived evidence for the inverse asso- Administrative, technical, or material support (i.e., reporting or organizing ciation between VDR and Wnt/b-catenin signaling. data, constructing databases): J.A. Randerson-Moor, D.T. Bishop, J. Newton- fi Bishop We report a trend toward increased T-cell in ltration of lung Study supervision: J.P. Laye, D.T. Bishop, J. Newton-Bishop metastases produced by VDR-expressing B16BL6 melanoma cells, although this did not reach statistical significance. Rejection of Acknowledgments B16-derivative melanoma cell lines in vivo requires both NK and T This work was funded by Cancer Research UK C588/A19167, C8216/A6129, cells, and in the pulmonary metastasis model, NK-cell–derived and C588/A10721 and NIH CA83115. J. Pozniak, J.M. Diaz, and S. Muralidhar IFNg is important but T cells also exert important antitumor were funded by Horizon 2020 Research and Innovation Programme no. 641458 activity (33–35, 54–57). The lack of statistically significant reduc- (MELGEN). Our grateful thanks to the participants who gave of their time and their blood samples. Also to the research nurses and technicians who collected tion in T cell numbers does not exclude the possibility that T cells the data over many years: Susan Leake, Susan Haynes, Birute Karpavicius, Paul are more active and/or that NK-cell recruitment and activity is also Affleck, Kairen Kukalizch, Linda Whitaker, Sharon Jackson, Edwina Gerry, modified. To this effect, further analyses of T-cell and NK-cell Elaine Fitzgibbon, Clarissa Nolan, Saila Waseem, Yvonne Taylor, Pauline recruitment and activity following VDR manipulation is war- Brunyee, Paul King, Tracy Lee, Samira Lobo, and Minttu Polso. To Jo Gascoyne ranted using both the subcutaneous and metastatic B16 models. and May Chan who provided critical support in managing the complex dataset. This study reports that vitamin D–VDR signaling bestows a We thank Dr. Robert Salmond, University of Leeds (Leeds, UK), for his guidance fi and access to cell culture facilities and Professor Graham Cook for immunologic prognostic bene t for patients with melanoma by inhibiting Wnt/ insights. We acknowledge the advice and assistance of Kenneth MacLennan, fi b-catenin signaling and increasing immune cell in ltration. These Consultant Professor of Tumour Pathology at Leeds Teaching Hospitals NHS findings also suggest that activating vitamin D–VDR signaling has Trust, in reviewing IHC sections. We thank Dr. James Hewinson (Experimental the potential to enhance antitumor immunity in an adjuvant Cancer Genetics group, Wellcome Sanger Institute, Cambridge, UK) for his setting. Notably, our findings suggest that vitamin D deficiency valuable assistance to the in vitro transfection experiments. We thank Dr. Martin in vitro (<25 nmol/L) is deleterious for melanoma survival rather than del Castillo for aiding in the choice of appropriate cell line for the and in vivo experiments. that high levels are protective. As melanoma is causally related to intense sun burn (58), sun avoidance is frequently recommended The costs of publication of this article were defrayed in part by the payment of to patients in follow-up. Our data suggest a causal relationship page charges. This article must therefore be hereby marked advertisement in between reduced vitamin D–VDR signaling and therefore, as sun accordance with 18 U.S.C. Section 1734 solely to indicate this fact. exposure is the dominant vitamin D source in most populations, simultaneous avoidance of vitamin D deficiency is important Received December 17, 2018; revised February 12, 2019; accepted October 1, health advice. 2019; published first November 5, 2019.

References 1. Haussler MR, Jurutka PW, Mizwicki M, Norman AW. Vitamin D receptor 3. Garland CF, Garland FC, Gorham ED, Lipkin M, Newmark H, Mohr SB, (VDR)-mediated actions of 1a,25(OH)(2)vitamin D(3): genomic and et al. The role of vitamin D in cancer prevention. Am J Public Health 2006; non-genomic mechanisms. Best Pract Res Clin Endocrinol Metab 2011; 96:252–61. 25:543–59. 4. Autier P, Boniol M, Pizot C, Mullie P. Vitamin D status and ill health: a 2. Campbell FC, Xu H, El-Tanani M, Crowe P, Bingham V. The yin and yang of systematic review. Lancet Diabetes Endocrinol 2014;2:76–89. vitamin D receptor (VDR) signaling in neoplastic progression: operational 5. Guzey M, Luo J, Getzenberg RH. Vitamin D3 modulated gene expression networks and tissue-speciﬁc growth control. Biochem Pharmacol 2010;79: patterns in human primary normal and cancer prostate cells. J Cell Bio- 1–9. chem 2004;93:271–85.

5996 Cancer Res; 79(23) December 1, 2019 Cancer Research

Vitamin D–VDR Signaling in Primary Melanoma Prognosis

6. Swami S, Raghavachari N, Muller UR, Bao YP, Feldman D. Vitamin D 29. Camp RL, Dolled-Filhart M, Rimm DL. X-tile: a new bio-informatics tool growth inhibition of breast cancer cells: gene expression patterns for biomarker assessment and outcome-based cut-point optimization. assessed by cDNA microarray. Breast Cancer Res Treat 2003;80: Clin Cancer Res 2004;10:7252–9. 49–62. 30. SACN TSACoN. SACN recommendations on vitamin D. Available from: 7. Wood RJ, Tchack L, Angelo G, Pratt RE, Sonna LA. DNA microarray analysis https://www.gov.uk/government/publications/sacn-vitamin-d-and-health- of vitamin D-induced gene expression in a human colon carcinoma cell report. line. Physiol Genomics 2004;17:122–9. 31. Shah S, Chiang C, Sikaris K, Lu Z, Bui M, Zebaze R, et al. Serum 25- 8. Suzuki T, Tazoe H, Taguchi K, Koyama Y, Ichikawa H, Hayakawa S, et al. hydroxyvitamin D insufficiency in search of a bone disease. J Clin Endo- DNA microarray analysis of changes in gene expression induced by 1,25- crinol Metab 2017;102:2321–8. dihydroxyvitamin D3 in human promyelocytic leukemia HL-60 cells. 32. van der Weyden L, Arends MJ, Campbell AD, Bald T, Wardle- Biomed Res 2006;27:99–109. JonesH,GriggsN,etal.Genome-widein vivo screen identifies 9. Wang TT, Tavera-Mendoza LE, Laperriere D, Libby E, MacLeod NB, novel host regulators of metastatic colonization. Nature 2017;541: Nagai Y, et al. Large-scale in silico and microarray-based identification of 233–6. direct 1,25-dihydroxyvitamin D3 target genes. Mol Endocrinol 2005;19: 33. Curran MA, Montalvo W, Yagita H, Allison JP. PD-1 and CTLA-4 combi- 2685–95. nation blockade expands infiltrating T cells and reduces regulatory T and 10. Newton-Bishop JA, Beswick S, Randerson-Moor J, Chang YM, Affleck P, myeloid cells within B16 melanoma tumors. Proc Natl Acad Sci U S A 2010; Elliott F, et al. Serum 25-hydroxyvitamin D3 levels are associated with 107:4275–80. Breslow thickness at presentation and survival from melanoma. J Clin 34. Ryu MS, Woo MY, Kwon D, Hong AE, Song KY, Park S, et al. Accumulation Oncol 2009;27:5439–44. of cytolytic CD8þ T cells in B16-melanoma and proliferation of mature T 11. Timerman D, McEnery-Stonelake M, Joyce CJ, Nambudiri VE, Hodi FS, cells in TIS21-knockout mice after T cell receptor stimulation. Exp Cell Res Claus EB, et al. Vitamin D deficiency is associated with a worse prognosis in 2014;327:209–21. metastatic melanoma. Oncotarget 2017;8:6873–82. 35. Xu D, Gu P, Pan PY, Li Q, Sato AI, Chen SH. NK and CD8þ T cell-mediated 12. Fang S, Sui D, Wang Y, Liu H, Chiang YJ, Ross MI, et al. Association of eradication of poorly immunogenic B16-F10 melanoma by the combined vitamin D levels with outcome in patients with melanoma after adjust- action of IL-12 gene therapy and 4-1BB costimulation. Int J Cancer 2004; ment for C-reactive protein. J Clin Oncol 2016;34:1741–7. 109:499–506. 13. Wyatt C, Lucas RM, Hurst C, Kimlin MG. Vitamin D deficiency at mela- 36. Jonsson G, Busch C, Knappskog S, Geisler J, Miletic H, Ringner M, et al. noma diagnosis is associated with higher Breslow thickness. PLoS One Gene expression profiling-based identification of molecular subtypes in 2015;10:e0126394. stage IV melanomas with different clinical outcome. Clin Cancer Res 2010; 14. Gambichler T, Bindsteiner M, Hoxtermann S, Kreuter A. Serum 25-hydro- 16:3356–67. xyvitamin D serum levels in a large German cohort of patients with 37. Oble DA, Loewe R, Yu P, Mihm MC Jr. Focus on TILs: prognostic signif- melanoma. Br J Dermatol 2013;168:625–8. icance of tumor infiltrating lymphocytes in human melanoma. 15. Saiag P, Aegerter P, Vitoux D, Lebbe C, Wolkenstein P, Dupin N, et al. Cancer Immun 2009;9:3. Prognostic value of 25-hydroxyvitamin D3 levels at diagnosis and during 38. Clemente CG, Mihm MC Jr, Bufalino R, Zurrida S, Collini P, Cascinelli follow-up in melanoma patients. J Natl Cancer Inst 2015;107:djv264. N. Prognostic value of tumor infiltrating lymphocytes in the vertical 16. Nsengimana J, Laye J, Filia A, O'Shea S, Muralidhar S, Pozniak J, et al. growth phase of primary cutaneous melanoma. Cancer 1996;77: b-Catenin-mediated immune evasion pathway frequently operates in 1303–10. primary cutaneous melanomas. J Clin Invest 2018;128:2048–63. 39. Erdag G, Schaefer JT, Smolkin ME, Deacon DH, Shea SM, Dengel LT, et al. 17. Filia A, Droop A, Harland M, Thygesen H, Randerson-Moor J, Snowden H, Immunotype and immunohistologic characteristics of tumor-infiltrating et al. High-resolution copy number patterns from clinically relevant FFPE immune cells are associated with clinical outcome in metastatic melano- material. Sci Rep 2019;9:8908. ma. Cancer Res 2012;72:1070–80. 18. Bikle DD. Vitamin D and the skin: physiology and pathophysiology. 40. Haanen JB. Immunotherapy of melanoma. EJC Suppl 2013;11: Rev Endocr Metab Disord 2012;13:3–19. 97–105. 19. Lauss M, Nsengimana J, Staaf J, Newton-Bishop J, Jonsson G. Consensus of 41. Wolchok JD, Chiarion-Sileni V, Gonzalez R, Rutkowski P, Grob JJ, Cowey melanoma gene expression subtypes converges on biological entities. CL, et al. Overall survival with combined nivolumab and ipilimumab in J Invest Dermatol 2016;136:2502–5. advanced melanoma. N Engl J Med 2017;377:1345–56. 20. The Cancer Genome Atlas Network. Genomic classification of cutaneous 42. Spranger S, Bao R, Gajewski TF. Melanoma-intrinsic beta-catenin signalling melanoma. Cell 2015;161:1681–96. prevents anti-tumour immunity. Nature 2015;523:231–5. 21. Nsengimana J, Laye J, Filia A, Walker C, Jewell R, Van den Oord JJ, et al. 43. Brozyna AA, Jozwicki W, Slominski AT. Decreased VDR expression in Independent replication of a melanoma subtype gene signature and cutaneous melanomas as marker of tumor progression: new data and evaluation of its prognostic value and biological correlates in a population analyses. Anticancer Res 2014;34:2735–43. cohort. Oncotarget 2015;6:11683–93. 44. Slominski AT, Brozyna AA, Zmijewski MA, Jozwicki W, Jetten AM, Mason 22. Wu G, Stein L. A network module-based method for identifying cancer RS, et al. Vitamin D signaling and melanoma: role of vitamin D and its prognostic signatures. Genome Biol 2012;13:R112. receptors in melanoma progression and management. Lab Invest 2017;97: 23. Tuoresmaki P, Vaisanen S, Neme A, Heikkinen S, Carlberg C. Patterns of 706–24. genome-wide VDR locations. PLoS One 2014;9:e96105. 45. Long MD, Campbell MJ. Pan-cancer analyses of the nuclear receptor 24. McLean CY, Bristor D, Hiller M, Clarke SL, Schaar BT, Lowe CB, et al. GREAT superfamily. Nucl Receptor Res 2015;2. doi: 10.11131/2015/101182. improves functional interpretation of cis-regulatory regions. Nat Biotech- 46. Saccone D, Asani F, Bornman L. Regulation of the vitamin D receptor nol 2010;28:495–501. gene by environment, genetics and epigenetics. Gene 2015;561: 25. Pozniak J, Nsengimana J, Laye JP, O'Shea SJ, Diaz JMS, Droop AP, et al. 171–80. Genetic and environmental determinants of immune response to cutane- 47. Aguilera O, Pena C, Garcia JM, Larriba MJ, Ordonez-Moran P, Navarro D, ous melanoma. Cancer Res 2019;79:2684–96. et al. The Wnt antagonist DICKKOPF-1 gene is induced by 1alpha,25- 26. Angelova M, Charoentong P, Hackl H, Fischer ML, Snajder R, Krogsdam dihydroxyvitamin D3 associated to the differentiation of human colon AM, et al. Characterization of the immunophenotypes and antigenomes of cancer cells. Carcinogenesis 2007;28:1877–84. colorectal cancers reveals distinct tumor escape mechanisms and novel 48. Bennesch MA, Picard D. Minireview: tipping the balance: ligand- targets for immunotherapy. Genome Biol 2015;16:64. independent activation of steroid receptors. Mol Endocrinol 2015;29: 27. Hugo W, Zaretsky JM, Sun L, Song C, Moreno BH, Hu-Lieskovan S, et al. 349–63. Genomic and transcriptomic features of response to anti-PD-1 therapy in 49. Skorija K, Cox M, Sisk JM, Dowd DR, MacDonald PN, Thompson CC, et al. metastatic melanoma. Cell 2017;168:542. Ligand-independent actions of the vitamin D receptor maintain hair 28. Van Allen EM, Miao D, Schilling B, Shukla SA, Blank C, Zimmer L, et al. follicle homeostasis. Mol Endocrinol 2005;19:855–62. Genomic correlates of response to CTLA-4 blockade in metastatic mela- 50. Trivedi T, Zheng Y, Fournier PGJ, Murthy S, John S, Schillo S, et al. The noma. Science 2015;350:207–11. vitamin D receptor is involved in the regulation of human breast cancer cell

www.aacrjournals.org Cancer Res; 79(23) December 1, 2019 5997

Muralidhar et al.

growth via a ligand-independent function in cytoplasm. Oncotarget 2017; 55. van der Weyden L, Karp NA, Swiatkowska A, Adams DJ, Speak AO. Genome 8:26687–701. wide in vivo mouse screen data from studies to assess host regulation of 51. Spath L, Ulivieri A, Lavra L, Fidanza L, Carlesimo M, Giubettini M, et al. metastatic colonisation. Sci Data 2017;4:170129. Antiproliferative effects of 1alpha-OH-vitD3 in malignant melanoma: 56. Umeshappa CS, Zhu Y, Bhanumathy KK, Omabe M, Chibbar R, Xiang J. potential therapeutic implications. Sci Rep 2017;7:40370. Innate and adoptive immune cells contribute to natural resistance to 52. Cianferotti L, Demay MB. VDR-mediated inhibition of DKK1 and SFRP2 systemic metastasis of B16 melanoma. Cancer Biother Radiopharm suppresses adipogenic differentiation of murine bone marrow stromal 2015;30:72–8. cells. J Cell Biochem 2007;101:80–8. 57. Taylor A, Rothstein D, Rudd CE. Small-molecule inhibition of PD-1 53. Katoh M. WNT/PCP signaling pathway and human cancer (review). transcription is an effective alternative to antibody blockade in cancer Oncol Rep 2005;14:1583–8. therapy. Cancer Res 2018;78:706–17. 54. Takeda K, Nakayama M, Sakaki M, Hayakawa Y, Imawari M, Ogasawara K, 58. Chang YM, Barrett JH, Bishop DT, Armstrong BK, Bataille V, Bergman et al. IFN-g production by lung NK cells is critical for the natural resistance W, et al. Sun exposure and melanoma risk at different latitudes: a to pulmonary metastasis of B16 melanoma in mice. J Leukoc Biol 2011;90: pooled analysis of 5700 cases and 7216 controls. Int J Epidemiol 2009; 777–85. 38:814–30.

5998 Cancer Res; 79(23) December 1, 2019 Cancer Research

Vitamin D−VDR Signaling Inhibits Wnt/β-Catenin−Mediated Melanoma Progression and Promotes Antitumor Immunity

Sathya Muralidhar, Anastasia Filia, Jérémie Nsengimana, et al.

Cancer Res 2019;79:5986-5998. Published OnlineFirst November 5, 2019.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-18-3927

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2019/11/05/0008-5472.CAN-18-3927.DC1

Cited articles This article cites 56 articles, 11 of which you can access for free at: http://cancerres.aacrjournals.org/content/79/23/5986.full#ref-list-1

Citing articles This article has been cited by 1 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/79/23/5986.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/79/23/5986. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.