African American Prostate Cancer Displays Quantitatively Distinct Vitamin D Receptor Cistrome-transcriptome Relationships Regulated by BAZ1A

African American (AA) prostate cancer associates with vitamin D3 deficiency, but vitamin D receptor (VDR) genomic actions have not been investigated in this context. We undertook VDR proteogenomic analyses in European American (EA) and AA prostate cell lines and four clinical cohorts. Rapid immunoprecipitation mass spectrometry of endogenous protein (RIME) analyses revealed that nonmalignant AA RC43N prostate cells displayed the greatest dynamic protein content in the VDR complex. Likewise, in AA cells, Assay for Transposase-Accessible Chromatin using sequencing established greater 1α,25(OH)2D3-regulated chromatin accessibility, chromatin immunoprecipitation sequencing revealed significant enhancer-enriched VDR cistrome, and RNA sequencing identified the largest 1α,25(OH)2D3-dependent transcriptome. These VDR functions were significantly corrupted in the isogenic AA RC43T prostate cancer cells, and significantly distinct from EA cell models. We identified reduced expression of the chromatin remodeler, BAZ1A, in three AA prostate cancer cohorts as well as RC43T compared with RC43N. Restored BAZ1A expression significantly increased 1α,25(OH)2D3-regulated VDR-dependent gene expression in RC43T, but not HPr1AR or LNCaP cells. The clinical impact of VDR cistrome-transcriptome relationships were tested in three different clinical prostate cancer cohorts. Strikingly, only in AA patients with prostate cancer, the genes bound by VDR and/or associated with 1α,25(OH)2D3-dependent open chromatin (i) predicted progression from high-grade prostatic intraepithelial neoplasia to prostate cancer; (ii) responded to vitamin D3 supplementation in prostate cancer tumors; (iii) differentially responded to 25(OH)D3 serum levels. Finally, partial correlation analyses established that BAZ1A and components of the VDR complex identified by RIME significantly strengthened the correlation between VDR and target genes in AA prostate cancer only. Therefore, VDR transcriptional control is most potent in AA prostate cells and distorted through a BAZ1A-dependent control of VDR function. Significance: Our study identified that genomic ancestry drives the VDR complex composition, genomic distribution, and transcriptional function, and is disrupted by BAZ1A and illustrates a novel driver for AA prostate cancer.

three different clinical prostate cancer cohorts. Strikingly, only in AA patients with prostate cancer, the genes bound by VDR and/or associated with 1α,25(OH) 2 D 3 -dependent open chromatin (i) predicted progression from high-grade prostatic intraepithelial neoplasia to prostate cancer; (ii) responded to vitamin D3 supplementation in prostate cancer tumors; (iii) differentially responded to 25(OH)D3 serum levels. Finally, partial correlation analyses established that BAZ1A and components of the VDR complex identified by RIME significantly strengthened the correlation between VDR and target genes in AA prostate cancer only. Therefore, VDR transcriptional control is most potent in AA prostate cells and distorted through a BAZ1A-dependent control of VDR function.
Significance: Our study identified that genomic ancestry drives the VDR complex composition, genomic distribution, and transcriptional function, and is disrupted by BAZ1A and illustrates a novel driver for AA prostate cancer.

Introduction
Among African American (AA) men, prostate cancer occurs in a more aggressive form, and at a younger age compared with European American (EA) counterparts (1). Genomic ancestry underpins this disparity whereby genetic (2,3) and epigenetic (4)(5)(6)(7) factors combine with biopsychosocial processes to drive AA prostate cancer. For example, the lower incidence of TMPRRS and ETS genetic fusion in AA prostate cancer (8), is just one common difference between EA and AA prostate cancer (9).
One potential driver of AA prostate cancer arises from altered vitamin D 3 signaling (10). Skin is the site where UVB radiation converts 7-dehydrocholesterol to vitamin D 3 , which is then metabolized further to form the 1α,25(OH) 2 D 3 . This secosteroid hormone is able to bind with high affinity to its target receptor, the vitamin D receptor (VDR) and thereby regulate gene expression (reviewed in ref. 11). UVB radiation also has the capacity to degrade folic acid in the bloodstream, and as a result skin pigmentation levels have modulated during adaptation to different environmental UVB exposure and given rise to a correlation between high UVB exposure and high skin pigmentation (12,13). Currently, however, many individuals live in UVB environments that differ from their ancestral ones and include AA men who as a result may experience vitamin D 3 deficiency. Supportively, there are significant associations between lower serum vitamin D 3 levels and the incidence of several cancers, including among AA men the incidence and progression risks of prostate cancer (10,(14)(15)(16)(17). Indeed, this relationship has been examined and the target of study in large-scale vitamin D 3 supplementation trials such as the VITAL randomized trial cohort (18). Although in this study, there was no overall impact on cancer incidence across the whole cohort, among the AA participants there was a 23% (P = 0.07) reduction in cancer risk, which is suggestive of a functional relationship and justification for increased AA participation in future studies (19).
The interaction of 1α,25(OH) 2 D 3 with VDR and the regulation of gene networks has been the subject of intensive investigation, and consistently highlighted relationships with genes that control cell-cycle progression, cell differentiation, immunomodulatory actions. As research on the genomic functions of the VDR has expanded, other regulatory actions have been identified including the regulation of the circadian rhythm (20,21). Again, supporting a role for the VDR function in prostate cancer health disparities, VDR transcriptional actions are significantly stronger in AA patients with prostate cancer compared with EA patients, with significantly more dynamic regulation of genes implicated in control of inflammation (15,22). Together, these data suggest that AA men are more acutely sensitive to low serum vitamin D 3 levels that leads to inadequate VDR signaling. Given that frequently that clinical trials in prostate cancer of vitamin D 3 analogs have often recruited largely from EA men, it is possible that this significant biological relationship has been overlooked amongst AA patients with prostate cancer (23).
In the current study, we aimed to establish VDR genomic functions in AA and EA prostate cancer with the goal to assess how this may contribute to health disparities. We utilized EA and AA nonmalignant prostate and prostate cancer cell models with confirmed genomic ancestry and defined the basal and 1α,25(OH) 2 D 3 -regulated VDR protein interactome (RIME), the VDR cistrome [Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq) and chromatin immunoprecipitation sequencing (ChIP-seq)], and the VDR transcriptome [RNA sequencing (RNA-seq)]. These interactome-cistrometranscriptome relationships were associated with outcomes in three clinical cohorts; (i) in AA and EA men with high-grade prostatic intraepithelial neoplasia (HGPIN) who progressed to prostate cancer; (ii) a prostate cancer chemoprevention trial where AA and EA men were supplemented with vitamin D 3 ; and (iii) a prostate cancer cohort of AA and EA men with gene expression data, clinical data, and measured serum vitamin D 3 levels. We also mined three publicly available datasets to identify and subsequently test a mechanism for BAZ1A, a member of the ATP-dependent ACF-1/5 ISWI chromatin remodeling complex, to suppress VDR signaling in AA prostate cancer; an overview of the workflow is shown in Supplementary Fig. S1. Together, these approaches revealed that VDR signaling qualitatively and quantitatively differed between AA and EA cells, and that BAZ1A expression in AA prostate cancer regulated the capacity of the VDR to control immunomodulatory and circadian signaling.

Cell Line Genetic Admixture Estimation
To ensure accurate ancestral group assignment, HaplotypeCaller and Admixture v1.3.0 were used to estimate ancestry proportions, based on reference populations from the 1000 Genomes Project phase III superpopulations in the AA cell lines ( Supplementary Fig. S1A).

qRT-PCR
Total RNA was isolated via TRIzol reagent (Thermo Fisher Scientific) for mRNA detection by the AllPrep DNA/RNA/miRNA Universal Kit (Qiagen).
cDNA was prepared using iScriptTM cDNA Synthesis Kit (Bio-Rad) and relative gene expression quantified via Applied Biosystems 7300 Real-Time PCR System (Applied Biosystems), for both TaqMan and SYBR Green (Thermo Fisher Scientific) applications. All SYBR Green primers were tested for specificity by melting curve analysis. All qRT-PCR experiments were performed in biological triplicates, with at least technical duplicates, and fold changes (FC) determined using the 2 − Ct method, as described previously (27).

Western Immunoblotting
Total cellular protein was harvested from exponentially growing cells, washed in ice-cold PBS. Cell lysis was in ice-cold RIPA buffer (50 mmol/L Tris-HCl pH 7.4, 150 mmol/L NaCl, 1% volume for volume Triton X-100, 1 mmol/L Ethylenediaminetetraacetic acid pH 8.0, 0.5% w/v sodium deoxychlorate, 0.1% w/v SDS) containing 1× cOmplete Mini Protease Inhibitor Tablets (Roche). Protein concentrations were quantified using DC Protein Assay (Bio-Rad). Equal amounts of proteins (30-60 μg) were resolved via SDS-PAGE using precast polyacrylamide gradient gels (Mini-Protean TGX, Bio-Rad) and transferred onto polyvinylidene fluoride membrane (Roche) for 30 V for 16 hours. Post transfer, membranes were blocked with 5% nonfat dry milk for 1 hour at room temperature. Blocked membranes were probed with primary antibody against BAZ1A, SMARCA5, VDR, GAPDH either overnight at 4°C or for 3 hours at room temperature. Primary antibody was detected with horseradish peroxidase-linked rabbit anti-mouse IgG (P0161, Dako) or goat anti-rabbit IgG (P0448, Dako) secondary antibody at room temperature using enhanced chemiluminescence Western Blotting substrate (Pierce). Signal quantification was performed using the ProteinSimple Fluorochem M Imager, as described previously (27).

Cell Viability
Bioluminescent detection of cellular ATP, as a measure of cell viability, was undertaken using CellTiter-Glo (Promega) reagents. Cells at optimal seeding density to ensure exponential growth were plated in 96-well, white-walled plates. Wells were dosed with agents to a final volume of 100 μL. Dosing occurred at the beginning of the experiment, and cells were incubated for up to 120 hours. Luminescence was detected with Synergy 2 multimode microplate reader (BioTek Instruments). Each experiment was performed in at least triplicate wells in triplicate experiments, as described previously (27).

Clonogenic Assays
Colony formation was undertaken with 1,000 cells plated in triplicates in a 6well plate and treated with 1α,25(OH) 2 D 3 every 3 days for a period of 14 days, and then washed and fixed with neutral buffered formalin and stained with crystal violet stain and quantified (28).

Transfection of BAZ1A
GFP-BAZ1A was purchase from Addgene (plasmid # 65371) and cells were stable transducted by selection and maintenance in media supplemented with puromycin (2 μg/mL). RIME RIME analyses were undertaken with antibody toward the VDR in cells treated with either vehicle or 1α,25(OH) 2 D 3 . A total of 20 × 10 6 cells were cross-linked with 1% formaldehyde solution, quenched with glycine (0.1 mol/L, pH 7.5), and harvested in cold PBS. Nuclei were separated (29) and sonicated (30 seconds on 30 seconds off cycles for 30 minutes) for genomic DNA fragmentation. A total of 30 μL of 10% triton-X was added and high-speed centrifugation was performed to separate the nuclear proteins. Furthermore, VDR and IgG antibody conjugated beads were incubated with nuclear lysates overnight and washed 10 times with RIPA buffer and two washes of Ambic solution as described previously (29). LC/MS-MS was performed over a 2-hour separation and mean spectral count results were analyzed with a generalized linear model workflow to identify differentially enriched proteins.

ChIP-seq
Cistromes were analyzed with Rsubread/csaw (33) along with TF motif analyses (MotifDb). To find potential transcription factor (TF) binding enrichment within cistromes, GIGGLE was utilized to query the complete human TF ChIP-seq dataset collection [10,361 and 10,031 datasets across 1,111 TFs and 75 histone marks (HM), respectively] in Cistrome DB (34). Prostate-specific filtering limited analysis to 681 datasets across 74 TFs and 238 datasets across 19 HMs. For each query dataset, the overlap with each experimental cistrome was determined. Putative coenriched factors were identified by assessment of the number of time a given factor was observed in the top 200 most enriched datasets relative to the total number of datasets for that factor in the complete Cistrome DB (>1.2 FC enrichment over background). For prostate-specific analysis, overlaps across datasets were averaged for each factor.

RNA-seq
RNA was extracted from cells in the presence of 1α,25(OH) 2 D 3 (100 nmol/L, 8 hours) or EtOH in biological triplicate samples and analyzed by RNA-seq. Sequencing libraries prepared with the TruSeq Stranded Total RNA kit (Illumina Inc), from 1 μg total RNA. Alignment of raw sequence reads to the human transcriptome (hg38) was performed via Rsubread (35) and transcript abundance estimates were normalized and differentially expressed genes (DEG) identified using a standard edgeR pipeline. Functional annotation of gene sets: Pathway enrichment analysis and gene set enrichment analysis (GSEA) were performed using gene sets from the Molecular signatures database (MSigDB). For transcript-aware analyses, the FASTQ files were aligned with salmon (36) and differentially enriched transcripts were identified using DRIMSeq (37) in a similar workflow to edgeR, as described previously (32).

Small RNA-seq
Cell lines were treated as RNA-seq and library preparation included ligation of 5 and 3 RNA adapters to the mature miRNAs 5 -phosphate and 3 -hydroxyl groups and 11-13 PCR cycles using a universal primer and a primer containing one of 48 index sequences, which allowed pooling of libraries and multiplex sequencing. Prior to pooling, each individual sample's amplified cDNA construct was visualized on a DNA-HS Bioanalyzer DNA chip (Agilent Technologies) for mature miRNA and other small RNA products (140-150 bp). Successful constructs were purified using a Pippen prep (Sage Inc.), using 125 to 160 bp product size settings with separation on a 3% agarose gel. The purified samples were validated for size, purity, and concentration using a DNA-HS Bioanalyzer chip. Validated libraries were pooled at equal molar to a final concentration of 10 nmol/L in Tris-HCI 10 mmol/L, pH 8.5, before 50 cycle sequencing on a MiSeq (Illumina, Inc.). FASTQ files were aligned to the genome (hg38) using Rsubread (with small RNA alignment options). Expression counts were called against the miRbase consensus miRnome using featureCounts and a standard edgeR pipeline determined differentially expressed miRNA as described previously (32).

Next-generation Sequencing
Sequencing was performed at the Nationwide Children's Hospital Institute for Genomic Medicine, Columbus, OH.

Determining How miRNA Expression in Serum Samples Associates with Progression to Prostate Cancer
In serum samples from SWOG S9917 (HGPIN to prostate cancer), nanostring PCR was used to identify miRNA associated with progression in AA patients, within race and across race by progression status. The data were processed with NanoStringDiff and significantly different miRNA identified (logPV > 1 and absFC > 0.58).

Identifying VDR Cistrome Genes in EA and AA Prostate Tumors Treated with Vitamin D 3
Transcriptomic data from tumors were obtained from a cohort of 7 AA patients with prostate cancer with confirmed African genomic ancestry and 16 EA patients with prostate cancer who were treated with vitamin D 3 (4,000 IU daily) prior to radical prostatectomy. Significantly differentially regulated genes in the AA prostate cancer group (there were no DEGs in the EA prostate cancer group) were overlapped with genes annotated to ATAC-seq or ChIP-seq regions within 100 kb. The percentage overlap of DEGs with the total number of the indicated cistrome genes was calculated.

Identifying VDR Cistrome Genes in EA and AA Prostate Tumors and Associations with Serum Vitamin D 3
Transcriptomic data from were obtained from a cohort of 57 AA and 18 EA patients with prostate cancer who underwent radical prostatectomy.

Data Analyses and Integration
All analyses were undertaken using the R platform for statistical computing (R version 4.1.3) and the indicated library packages implemented in Bioconductor.

Ethics Approval and Consent to Participate
In accordance with the U.S. Common Rule, the archived samples used in this study were obtained from patients with written informed consent, and reviewed and approved by the Institutional Review Boards (IRB) of their respective clinical institutions (6,38,39). The serum samples from the Southwest Oncology Group (SWOG) clinical trial (SWOG S9917) were collected in accordance with recognized ethical guidelines and under local IRB approval (38). The prostate cancer samples from EA and AA patients who received vitamin D 3 prior to radical prostatectomy were collected in accordance with recognized ethical guidelines and under local IRB approval (6). The radical prostatectomy prostate cancer samples from EA and AA patients prior were collected in accordance with recognized ethical guidelines and under local IRB approval (39).

Availability of Data and Materials
The datasets generated and/or analyzed during the current study will be available on Gene Expression Omnibus.

AA Cell Line Genomic Ancestry and Relationship to Primary AA Prostate Samples
In the first instance, we confirmed that the RC43N, RC43T, RC77N, and RC77T cell lines all had more than 90% African genomic ancestry, and as expected LNCaP were predominantly European genomic ancestry (Supplementary Fig. S2A). Recently (40), shared androgen receptor (AR) binding has been established between primary EA prostate cancer samples and prostate cancer cell lines, such as LNCaP, supporting cell line research utility. To complement this, we reasoned that cell lines may reflect primary prostate tissue in an ancestry-dependent manner and therefore we measured the similarity between RC43N cells and nonmalignant AA prostate epithelium, by examining the genomic overlap between chromatin accessibility in cell lines and AR cistrome in primary tissues (41). These analyses demonstrated that the RC43N and RC43T very significantly overlapped with AA nonmalignant prostate and prostate cancer AR cistromes, but not the EA prostate cancer AR cistrome, and LNCaP most significantly overlapped with the EA prostate cancer AR cistrome ( Supplementary Fig. S2B).

The Composition of the VDR Complex Differs Significantly Between EA and AA Cells
VDR protein levels were detected in all cells, and generally were elevated following 1α,25(OH) 2 D 3 treatment, which was most pronounced in nonmalignant AA RC43N cells; RC77T cells did not change VDR expression in response to 1α,25(OH) 2 D 3 (Fig. 1A). RIME was used to measure VDR-interacting proteins, were established in RC43T compared with LNCaP, and RC43N with HPr1AR, and then the delta between these comparisons were identified and ranked.
To identify VDR-protein interactions that reflected genomic ancestry, the significant differences in proteins in the VDR complex were calculated in RC43N compared with HPr1AR, and RC43T compared with LNCaP. In S2D and S2E). Together, these results suggest the VDR complex differs significantly by genomic ancestry and transformation, and reflects resistance to 1α,25(OH) 2 D 3 -inhibited colony formation in RC43T cells.
The greatest 1α,25(OH) 2 D 3 -dependent impact on NF and mononucleosome (mono) regions was in RC43N and RC43T ( Fig. 2A,  LNCaP cells displayed the fewest significant associations although one of the most enriched motifs was the AR/Half-site (45), which was also only enriched in RC43T regions. Enrichment in mono regions included SP (e.g., SP2), C2H2-ZNF (e.g., KLF5), and ETS (e.g., ELK4) family members ( Fig. 2B; Supplementary Fig. S3A). Delta motif enrichment values were calculated by comparing each cell with HPr1AR cells (Fig. 2C). RC43N displayed the most striking gains in enrichment of multiple TF families including bHLH and FOX family members. The bHLH motifs included multiple TFs for circadian rhythm such as bMAL1 and CLOCK, which were either most clearly or exclusively enriched compared with other cells; for example, there was a striking loss of enrichment for these factors between RC43T compared with RC43N (Fig. 2D).
Similarly, FOX family members were highly enriched in RC43N, although it is notable that the FOXA1-AR motif enrichment was prominent in LNCaP cells ( Supplementary Fig. S3B). These results support the fact that the VDR complex has the greatest impact in AA cells and underscores the divergent response between the isogenic RC43N and RC43T cells.
Annotating with ChromHMM states revealed that VDR enrichment in transcribed regions and bivalent promoters were shard across cells and treatment.
Other enrichments were cell specific, for example, polycomb regions were significantly enriched only in RC43T and LNCaP (Supplementary Table S5). Compared with the ATAC-seq enrichments, which were frequent at Promoter regions, only basal VDR ChIP-seq in LNCaP was enriched at Promoter regions, and only modestly so. Interestingly, this annotation approach revealed the significant impact of 1α,25(OH) 2 D 3 treatment. For example, basal VDR enrichment in bivalent promoters was most significant in LNCaP and RC43N, and reduced by 1α,25(OH) 2 D 3 ; in LNCaP log 10 (P adj ) = 46, and reduced to 6 in the presence of 1α,25(OH) 2 D 3 . In contrast, the score in RC43N (∼20) was broadly equivalent in the basal and 1α,25(OH) 2 D 3 -stimulated states, suggesting one aspect of cancer cells, regardless of genomic ancestry is for 1α,25(OH) 2 D 3 to reduce VDR binding at these regions, whereas it is sustained in AA prostate cells. Binding at an enhancer region illustrates that RC43N binding is comparable in both basal and 1α,25(OH) 2 D 3 -stimulated treatment, but reduced in RC43T ( Supplementary Fig. S4D).  Fig. S5A). There were also significant overlaps with nuclear receptors, including for PPARs, RARs, and VDR, which were all most significant in RC43N ( Supplementary Fig. S5B Fig. S5C). Together, these data suggest that VDR is highly integrated in RC43N cells with multiple TFs including ETS, FOX, and nuclear receptors, and circadian rhythm ( Supplementary Fig. S5C, right); this reflected motif enrichment as well as in the ATAC-seq data. Frequently, these enrichments are missing or diminished in the other cells.
To illustrate the similarities and differences between the cell lines, we integrated different datasets. For example, RIME data ( Fig. 1) were related to VDR cistromes ( Fig. 2 and 3) within each cell. Specifically, VDR cistromes (ATAC-seq and ChIP-seq) were annotated to genes within a 100 kb window up and downstream, or within genes (46). From these cistrome-annotated genes we measured enrichment for RIME-identified VDR-interacting proteins in either the basal or 1α,25(OH) 2 D 3 -stimulated state within each cell (Supplementary Table S7). The 1α,25(OH) 2 D 3 -dependent NF regions in RC43N were enriched (logPV = 14.9) for 1α,25(OH) 2 D 3 -dependent VDR-interacting proteins identified in the same cell, including the VDR itself, the splicing factor PTBP1, and the DNA helicase XRCC6. The 1α,25(OH) 2 D 3 -dependent NF regions in RC43T annotated to multiple proteins that were enriched in the basal VDR complex (logPV = 26.8) including VDR and XRCC5 and given that these represent loss of NF regions it may reflect why the number of proteins in the 1α,25(OH) 2 D 3 -regulated VDR complex in RC43T is much reduced compared with the basal state (Fig. 1B).    Table  S9) also support AR and FOS being highly enriched in RC43T, and TAL1 and PIAS1 unique in RC43N and PPARγ in RC43T.

The 1α,25(OH) 2 D 3 -dependent Transcriptome is Larger in RC43N and RC43T Than HPr1AR and LNCaP Cells
Together, these findings reveal that 1α,25(OH) 2 D 3 transcriptional signaling is more impactful in AA compared with EA prostate models, and that within the isogenic AA cells, there were frequent divergent enrichment of pathways suggesting that 1α,25(OH) 2 D 3 signaling is significantly altered in RC43T compared with RC43N, which reflects the ATAC-and ChIP-seq.

Integration of VDR-dependent Cistrome and Transcriptome Data Reveals the Strongest 1α,25(OH) 2 D 3 -dependent Gene Regulation Responses Occur in RC43N Cells
In the first instance, we examined how the ATAC-seq and ChIP-seq cistrome data overlapped within each cell and treatment combination. Remarkably, a significant overlap (a minimum of 1 bp) between 1α,25(OH) 2 D 3 -stimulated VDR ChIP-seq and ATAC-seq data was infrequent. That is, individual genes had both a significant VDR binding site and a significant 1α,25(OH) 2 D 3stimulated NF region, but these sites did not frequently overlap. In contrast, in RC43N the basal and 1α,25(OH) 2 D 3 -stimulated VDR ChIP-seq data significantly overlapped with the basal NF regions (i.e., not the differentially enriched 1α,25(OH) 2 D 3 -dependent NF regions). Similarly, in LNCaP the basal VDR ChIP-seq overlapped with the basal NF regions (Supplementary Table S10). In these cells, VDR binding was significantly enriched in NF regions.
Next, relationships were identified between either VDR binding or 1α,25(OH) 2 D 3 -induced NF regions, and gene expression; namely, peak:gene relationships. To refine these relationships further, the VDR cistromes were annotated to within a 100 kb of the nearest gene, including those that were members of the VDR interactome as defined by BioGRID.
Naturally, the larger ATAC-seq cistromes annotated to more genes, and most clearly in RC43N cells (Supplementary Table S11). For example, the approximately 10,000 RC43N NF regions were significantly enriched in ChromHMM defined promoters, which collectively annotated to approximately 20,000 nonunique genes. This is broadly true for RC43T, except these NF regions were sites of significant loss in chromatin accessibility. It is also clear that VDR.biogrid genes including VDR itself and NCOR associated most frequently (n∼90) with the 1α,25(OH) 2 D 3 -stimulated NF regions in RC43N, but this was less frequent in the other cell lines. The basal and 1α,25(OH) 2 D 3 -stimulated VDR ChIP-seq data displayed a similar pattern (Supplementary Table S12).
Again, reflecting that 1α,25(OH) 2 D 3 treatment in RC43T cells reduces VDRgenome interactions, almost all peak:gene relationships were in the basal state and most were not annotated to ChromHMM regions. Finally, across the ChIP-seq data, very few VDR.biogrid genes were commonly identified, but did include LCOR in LNCaP, RC43N, and RC43T.
Next we filtered the ChromHMM-classified cistrome gene:peak relationships to 1α,25(OH) 2 D 3 -stimulated genes that were significantly different between AA and EA models [e.g., Supplementary Fig. S8 Fig. S10A; left). VDR ChIP-seq cistromes peak:gene relationships were analyzed in the same manner (Supplementary Fig. S10A; right), and again revealed the pronounced effect of 1α,25(OH) 2 D 3 in RC43N, but a minimal impact in RC43T, and a modest impact of 1α,25(OH) 2 D 3 in HPr1AR cells.
Next, we calculated the frequency of peak:gene relationships in 10 kb bins around from DEGs (ChIP-seq data; Supplementary Fig. S10B, ATAC-seq data). To test the significance of the VDR-dependent cistrome-transcriptome relationships (32), we applied the BETA method (48). Specifically, within each cell type, we summed significance of the peaks within 100 kb of each annotated DEG multiplied by the absolute FC for the same DEG and weighted by the peak distribution (proximal vs. distal), or unweighted. We defined this score as the weighted cistrome-transcriptome (wt-C-T). Using this approach, we tested how the cistrome data significantly related to gene expression changes cells treated with 1α,25(OH) 2 D 3 .
For the ATAC-seq cistrome, categorized into ChromHMM distributions, in most cases wt-C-T scores were significantly greatest in RC43N cells notably at Promoters, and Poised and Active enhancers ( Supplementary Fig. S10C, left).
VDR cistrome data were integrated with DEGs regardless of ChromHMM association and revealed that RC43N wt-C-T values were also generally greater than HPr1AR cells when treated with 1α,25(OH) 2 D 3 ( Supplementary Fig. S10C, right). The only exception was in RC43T cells considering genes bound by proximal basal VDR associated had a higher wt-C-T score than RC43N. Together these data support the concept that that VDR binding and 1α,25(OH) 2 D 3 associated NF regions in RC43N were most consistently associated with stronger patterns of gene expression than the other three cells.

The BAZ1A/SMARCA5 Complex Regulates 1α,25(OH) 2 D 3 -stimulated VDR Responses
The cistrome-transcriptome studies support the concept that nonmalignant RC43N prostate cells are significantly more sensitive to VDR-mediated gene regulation than either RC43T or the EA cells. The VDR binding in RC43N is co-incident with motifs for other nuclear receptors and other TFs that control circadian rhythm. In contrast to RC43N, the isogenic malignant counterpart RC43T, displays suppressed gene expression patterns that include disruption of numerous transcriptional programs. Therefore, we exploited clinical cohorts to determine whether there was evidence for suppressed VDR signaling in AA prostate cancer progression.
In the first instance, we screened a large panel of coregulators (42) in the DEGs between TMPRSS2 fusion positive and negative prostate cancer in EA and AA AACRJournals.org Cancer Res Commun; 3(4) April 2023 patients in the TCGA prostate cancer cohort (TCGA-PRAD). This identified 27 altered coregulators in AA patients, and from these, five were uniquely or more significantly altered in AA compared with EA TMPRSS2 fusion negative prostate cancer (Fig. 5A). The most altered coregulators included several known to interact with VDR signaling including BAZA (bromodomain adjacent to zinc finger domain 1A), and SMARCA/WSTF (SWI/SNF-related, matrix-associated, actin-dependent regulator of chromatin, subfamily A, member 5), which functions cooperatively with BAZA (49). In contrast, expressions of the AR and VDR were unchanged between EA and AA prostate cancer samples. Across the cell models, BAZ1A protein expression was highest in LNCaP and unchanged by 1α,25(OH) 2 D 3 , whereas it was strongly induced in RC43N, but strongly repressed in RC43T, reflecting the broader transcriptional patterns identified by RNA-seq. Expression responses were broadly the same for SMARCA5 (Fig. 5B). It is interesting to note that LNCaP has the highest expression of BAZ1A, whereas the other cell lines, which are immortalized with human papillomavirus, have lower expression, suggesting a link between RB status and BAZ1A expression. RC77N and RC77T cells displayed some similarities, with a modest 1α,25(OH) 2 D 3 -induced SMARCA5 repression in RC77N, but no repression in RC77T. Given that the majority of experiments were undertaken in RC43N and RC43T, we therefore pursued the BAZ1A/SMARCA5 relationships to RC43N/T cells.
To test the impact of altered BAZ1A/SMARCA5 on expression of VDR target genes, we examined the correlations between either BAZ1A or SMARCA5 and the 1α,25(OH) 2 D 3 -regulated genes from RC43N and RC43T in the AA TMPRSS2 fusion negative tumors from TCGA-PRAD cohort. From these correlations, we filtered those genes in Hallmarks_InflammatoryResponse and GO_Circadian Rhythm (Supplementary Fig. S11A). Supportively, BAZ1A and SMARCA5 correlations to these pathway genes were more pronounced for RC43N 1α,25(OH) 2 D 3 -regulated genes than those from RC43T, supporting a CoA role for BAZ1A/SMARCA5 regulation of genes associated with inflammation.
Next, we examined the genes in BAZ1A containing SWI/SNF complexes that were expressed in the cell lines and tumor cohorts. Expression patterns of the BAZ1A SWI/SNF complex genes significantly distinguished the AA from the EA cell line models ( Supplementary Fig. S11B). The most altered genes in this complex also significantly distinguished TMPRSS fusion positive from negative tumors in TCGA-PRAD cohort (χ 2 P = 0.002), suggesting that genomic ancestry impacted expression of these genes was most common in the absence of TMPRSS fusion (Supplementary Fig. S11C). Finally, we examined expression of all the genes in all four SWI/SNF complexes in the Rayford and colleagues cohort of AA tumors (n = 596) compared with EA tumors (n = 556; ref. 7). All detected genes from each of the four complexes were significantly downregulated in AA tumors compared with EA counterparts (Supplementary Fig. S11D).
We also analyzed the RNA-seq data from the Berchuck and colleagues cohort of AA and EA tumors (41), and identified approximately 3,600 DEGs in the AA patients compared with EA counterparts, and within these DEGs, BAZ1A and BAZ1B are significantly downregulated, and a hypergeometric test reveals that the members of the BAZ1A and BAZ1B complexes are significantly enriched in the DEGs (P < 0.001).
Next, we tested the impact of BAZ1A expression using RNA-seq after 1α,25(OH) 2 D 3 stimulation (100 nmol/L, 8 hours) in BAZ1A transfected cells compared with vector controls (Supplementary Fig. S12A; Fig. 5C). BAZ1A overexpressed cells displayed enhanced 1α,25(OH) 2 D 3 responses and was most pronounced in RC43N in terms of the number of genes with enhanced responses. Genes associated with a VDR ChIP-seq peak or NF regions was also most pronounced and significant RC43N and RC43T ( Fig. 5C; Table 1). However, it is made more striking by the fact that the same analyses of the parental cells (Fig. 4A) found no significant enrichment.
Analyses of the GSEA terms also supported a role for BAZ1A to impact gene expression. In RC43N, the most enriched terms were IFNγ and α, Inflammatory responses, IL6 signaling, and TNFA signaling suggesting an immunomodulatory phenotype (Fig. 5D). Focusing on the impact of BAZ1A on 1α,25(OH) 2 D 3 -induced expression changes in RC43N and RC43T, and calculating the enrichment terms that change the most, in either the same or a divergent manner, revealed that BAZ1A exerted a potent and cell-specific impact on gene expression patterns induced by 1α,25(OH) 2 D 3 . Indeed, GSEA terms are illustrated that changed to convergent from a divergent response in parental cells (e.g., Hallmarks_TGF_Beta; Fig. 5E, top) or the opposite (e.g., Hallmark_Interferon_Gamma; Fig. 5E, bottom). However, although circadian rhythm TFs and their motifs were enriched in VDR cistromes (e.g., Fig. 2A; Supplementary Fig. S5B), circadian rhythm transcriptomes were not enriched in a BAZ1A-dependent manner and suggests there are alternative factors combining with VDR to regulate this aspect.
More widely, we reasoned that the 1α,25(OH) 2 D 3 -regulated and BAZ1Adependent transcriptomes may reflect the ability of VDR to control prostate lineage and differentiation decisions. We therefore tested the significant over-  Second, we reexamined data from our earlier prostate cancer RNA-seq study from EA and AA patients who received vitamin D 3 (4,000 IU daily) prior to radical prostatectomy. Ancestry informative markers confirmed the African genomic ancestry of the AA patients. As we reported previously (6), the responses in EA patients were essentially null, whereas a strong vitamin D 3 transcriptional response was observed in the AA patients, and GSEA revealed these genes were enriched in immunomodulatory and prostate cancer-relevant pathways (Fig. 6A). The DEGs were again enriched for inflammatory signaling components ( Supplementary Fig. S13). Interestingly, 70% of the significantly   Table S15).
In a third cohort of AA and EA patients VDR cistrome genes were examined in which DEG analyses in tumor and contralateral normal material was available (39). We therefore measured how gene expression was impacted by either deficient serum vitamin D 3 levels or obesity and enriched for VDR cistrome genes.
Reflecting, the radical prostatectomy samples (Supplementary Table S16), significant serum 25(OH)D 3 -dependent DEGs were only identified in the AA patients. Furthermore, the impact of obesity was more profound in the AA patients (2,415 DEGs). In both cases, these DEGs were enriched for VDR ChIP-seq or 1α,25(OH) 2 D 3 -stimulated NF ATAC-seq genes ( Fig. 6B; Table 2).
Finally, we used partial correlation analyses to define how genomic ancestry and oncogenic transformation enriched components of the VDR complex (Fig. 1C) impacted the strength of the correlations between VDR and AA VDR ChIP-seq annotated genes in AA tumors, or EA VDR ChIP-seq annotated genes in EA tumors. In this manner, we were able to test how components of the VDR complex identified in cell lines were plausibly impacting VDR-target gene relationships in prostate cancer in patients. This demonstrated that several of these coregulators, including SAFB, PARP1, HDAC2, NONO, BAZ1A, and SMARCA5 significantly and positively impacted the strength of the correlations between VDR and AA VDR ChIP-seq genes but not the corollary relationships of the strength between VDR and EA VDR ChIP-seq genes in EA tumors (Fig. 6C).

Discussion
The current study aimed to define VDR genomic signaling in the context of the racial health disparities in prostate cancer, given that the AA patient group appears to be most acutely vulnerable to low serum vitamin D 3 . Therefore, we integrated genomic, transcriptomic, and proteomic datasets, combined with clinical genomic data to reveal how genomic ancestry may impact VDR signaling in prostate cancer.
There are several challenges in health disparities research not least of which are confirming genomic ancestry of biological materials and then defining acceptable parameters for comparison of disease status across samples in a meaningful manner. In prostate cancer research, compared with some other solid tumors such as breast cancer, there are fewer cell lines established from AA patients, and to date even fewer patient-derived xenograft models. Therefore, caution always needs to be taken when attempting to make direct comparisons between cancer cell lines of different genomic ancestry. In the current study, the genomic ancestry of the AA cells was confirmed, and the chromatin accessible regions of  (Supplementary Fig. S2C); both cells gained mononucleosome sites in response to 1α,25(OH) 2 D 3 ( Fig. 2A) whereas HPr1AR and LNCaP did not; both cells displayed a loss of VDR binding in response to 1α,25(OH) 2 D 3 ( Supplementary Fig. S4A); both cells displayed highly significant basal enrichment of VDR in bivalent promoters, and polycomb regions (Supplementary Table S5); nuclear receptor motif enrichment was also comparable LNCaP and RC43T ( Fig. 3B; Supplementary We sought to identify mechanisms that may drive the divergent responses between EA and AA cells, and between RC43N and RC43T. TCGA analyses identified a potential role for altered expression of the BAZ1A/SMARCA5 SWI/SNF complex, and this was replicated in another AA EA prostate cancer cohort (41). These data prompt the question of how is BAZ1A expression downregulated to suppress 1α,25(OH) 2 D 3 signaling in RC43T cells compared with RC43N cells. 1α,25(OH) 2 D 3 treatment induced chromatin accessibility around the BAZ1A locus only in RC43N cells, and 1α,25(OH) 2 D 3 treatment upregulated BAZ1A and SMARCA5 in RC43N, but downregulated these proteins in RC43T. Transcript-aware analyses revealed that 1α,25(OH) 2 D 3 induced different transcripts of SMARCA4 only in RC43N; the BAZ1A complex contains multiple SMARCA components. Reflecting this, RIME analyses revealed that SMARCA5 was only significantly enriched in the VDR complex in RC43N and RC43T, but significantly reduced by 1α,25(OH) 2 D 3 -treatment in RC43T. It is noteworthy that in RC43N the VDR ChIP-seq significantly overlapped with publicly available SMARCA4 ChIP-seq from LNCaP cells. To test the clinical significance of these relationships, we used partial correlation analy-ses to demonstrate that BAZ1A and SMARCA5 expression, as well as other VDR-interacting proteins identified in cell lines by RIME, were able to significantly strengthen the correlations between VDR and AA cistrome genes in AA prostate cancer.
Our data perhaps suggest in RC43N, there is an 1α,25(OH) 2 D 3 -autoregulatory mechanism for BAZ1A/SMARCA5 expression and this function that is corrupted in RC43T, and AA prostate cancer. Finally, it is interesting to note that and genome-wide association study SNPs in BAZ1A associate significantly with heel bone strength, also supporting a role in regulating VDR function (51).
Restoring BAZ1A expression led to significantly enhanced 1α,25(OH) 2 D 3regulated transcriptome, but these genes were only significantly enriched for VDR bound and NF regions in RC43N and RC43T suggesting that BAZ1A expression had the most significant impact on VDR function in AA cells.
The most strongly upregulated gene in RC43N that is annotated to RC43N 1α,25(OH) 2 D 3 -regulated NF region was PIKR, which recently was identified as a novel antigen in a clinical immunotherapy trial in advanced prostate cancer (52).
Validation in three clinical cohorts revealed that the footprint of VDR signaling was most apparent in AA prostate cancer. For example, miRNA that predicted progression from HGPIN to prostate cancer in AA men were highly enriched for VDR cistrome data, as were genes that responded in prostate tumors from men receiving vitamin D3 supplementation prior to radical prostatectomy. This was also strikingly apparent in prostate tumors from men who had deficient serum 25(OH)D 3 levels, and indeed this interacted significantly with obesity (BMI > 30.0 kg/m 2 ) status. The strength of the correlation between VDR and AA ChIP-seq target genes was also significantly impacted by coregulators interacting with the VDR such as PARP1 and NONO, as well as BAZ1A.
Our data also contribute to the earlier analyses of VDR functions in the prostate (53) and also mechanisms which limits VDR control of proliferation. Observational evidence supports a cross-talk between VDR and components of the transcriptional network that regulates melatonin production, circadian rhythm, and sleep duration (20,54). VDR function can be corrupted through various mechanisms, either through changes in serum vitamin D 3 levels, changes in membrane transport (16) or disruption to the composition of the VDR complex (55,56). To this concept, we have added the mechanistic insight that the SWI/SNF complexes containing BAZ1A and SMARCA5 are distorted in a manner that reflects genomic ancestry, and further distorts the normal functions of the VDR and suggests enhanced sensitivity to DNA damage. It is also interesting to note that a new class of drugs has recently been developed to target SMARC-containing complexes in prostate cancer (57).
It is reasonable to suggest that the VDR stands at the crossroads of biopsychosocial signaling that impacts prostate cancer by a distinct pattern of VDR genomic binding in a manner that is governed by African genomic ancestry. Although guidance is available for serum 25(OH)D 3 levels required for bone health, it is far from clear what level is required either to promote cardiovascular health or to prevent autoimmune diseases. It is even less clear how 25(OH)D 3 deficiency among AAs, which is highly prevalent, impacts cancer and these other diseases.
Our genomic data suggest that the powerful example of changing melanin content in the skin through ancestral adaptation has occurred in parallel with a range of genomic mechanisms to govern VDR functions in noncalcemic tissues.
The current study suggests that the functions of the VDR may have adapted with significant distinctions between people of different genomic ancestry.
More specifically, we reason that adaptation to environments of lower UVB exposure that could potentially be associated with vitamin D insufficiency and alter the prominence of how VDR signaling occurs in a wide variety of tissues. Furthermore, we propose that the prostate is an important gland with which to test this possibility given it is the site of significant syndromes and diseases that are highly impactful on U.S. men, and furthermore many of these conditions display significant health disparities.