Adipocyte-driven unfolded protein response is a shared transcriptomic signature of metastatic prostate carcinoma cells

Abstract

A critical unknown in the field of skeletal metastases is how cancer cells find a way to thrive under harsh conditions, as exemplified by metastatic colonization of adipocyte-rich bone marrow by prostate carcinoma cells. To begin understanding molecular processes that enable tumor cells to survive and progress in difficult microenvironments such as bone, we performed unbiased examination of the transcriptome of two different prostate cancer cell lines in the absence or presence of bone marrow adipocytes. Our RNAseq analyses and subsequent quantitative PCR and protein-based assays reveal that upregulation of Endoplasmic Reticulum (ER) stress and Unfolded Protein Response (UPR) genes is a shared signature between two metastatic prostate carcinoma cell lines of different origin. Pathway analyses and pharmacological examinations highlight the ER chaperone BIP as an upstream coordinator of this transcriptomic signature. Additional patient-based data support our overall conclusion that ER stress and UPR induction are shared, important factors in the response and adaptation of metastatic tumor cells to their micro-environment. Our studies pave the way for additional mechanistic investigations and offer new clues towards effective therapeutic interventions in metastatic disease.

INTRODUCTION

The functional cross-talk of metastatic tumor cells with their surrounding bone marrow microenvironment is a critical component of metastatic progression and therapy evasion. Bone-trophic tumors, such as prostate cancers (PCa), adapt to harsh skeletal microenvironments through mechanisms that are not clearly understood and investigation of which is likely to yield new clues towards effective therapeutic interventions. Hypoxia, oxidative stress and pro-inflammatory events in cancer are tightly linked to the unfolded protein response (UPR), a protective mechanism that maintains homeostasis and strengthens the ability of protein folding in the endoplasmic reticulum (ER), ultimately increasing a cell’s chances of survival¹.

The three mechanistic arms of UPR are initiated by three ER stress sensors: Inositol Requiring Enzyme 1α (IRE1), Protein kinase RNA- activated- like ER Kinase (PERK) and Activating Transcription Factor 6 (ATF6)². Under normal conditions, each of the stress sensors is maintained in the inactive state by the ER chaperone BIP (also known as HSPA5 and GRP78)³. During ER stress, BIP coordinates an adaptive response that comprises reduced protein translation, increased capacity for protein trafficking, and the elimination of misfolded proteins via proteasome-dependent ER- associated degradation (ERAD), as well as autophagy⁴. BIP expression in clinical samples from several types of tumors, including PCa, is linked to increased cancer cell survival and chemoresistance^5,6.

Our previous studies showed that increased marrow adiposity -- a condition associated with advanced age and obesity that exacerbates the stressful environment in the bone -- activates adaptive signaling in metastatic tumor cells that effectively enables their survival^7–10. Marrow fat cells induce the expression of lipid transporters and promote lipid uptake by tumor cells, which has profound effects on the biology of prostate cancer as well as other cancers that reside in bone^7,9,11–15. Specifically, we have shown previously that uptake of adipocyte-supplied lipids increases prostate tumor cell proliferation and invasion¹², induces Warburg phenotype¹⁵ and augments expression and secretion of interleukin 1L1β, leading to an enhancement in lipolysis and prostaglandin production⁹. We have also demonstrated that adipocytes promote pseudohypoxia and activate HIF-1α signaling in metastatic tumor cells, causing metabolic adaptation and supporting progression^7,15. Exposure to adipocyte-rich bone marrow enhances expression and activity of the anti-oxidant enzyme Heme Oxygenase 1 (HO-1), arming the tumor with coping mechanisms against oxidative stress and supporting survival¹⁰. We further reported that levels of ER chaperone BIP are augmented with adipocyte-induced oxidative stress, which coincides with increases in spliced XBP1, a substrate of ER-localized transmembrane sensor IRE1¹⁰. This finding suggested potential adipocyte-driven activation of UPR in metastatic tumor cells, which has not been previously reported.

Likely contributors to ER stress and UPR in bone marrow metastatic niche are adipocyte-supplied lipids. Lipids are stored as triglycerides and broken down through lipolysis into glycerol and free fatty acids (FFA)¹⁶. We and others have shown that prostate and leukemia cells in bone induce adipocyte lipolysis and this process is driven by bi-directional interaction between the tumor and the marrow fat cells^9,13. The avidity of tumor cells for lipids has been documented by multiple studies^{9,11–13,17,18} and the importance of lipid metabolism has been especially demonstrated in prostate cancer development and progression^19–23. Lipids -- particularly saturated fatty acids -- induce ER stress²⁴. Palmitic acid, one of the products of adipocyte lipolysis²⁵, promotes cellular stress signaling in a variety of tissues, including tumor cells^26,27, and high-fat diet drives UPR and pathological ER stress signaling in mice^28,29.

Several ER stress-related proteins are regulated by lipids or play a role in lipid homeostasis. Expression of synoviolin (SYVN1), an ER-resident E3 ubiquitin ligase implicated in the regulation of body weight and energy expenditure, is increased in response to lipid overload³⁰. DNA Damage Inducible Transcript (DDIT3, also referred to as CHOP), which is conventionally known as a transcriptional regulator of pro-and anti-apoptotic genes and a mediator of cell’s life vs. death decisions^31,32, regulates lipid droplet biogenesis and storage of cholesterol esters and triglycerides.³³ Pseudokinase Tribbles Homolog 3 (TRIB3) is linked to lipolysis and fatty acid oxidation.³⁴ Lastly, XBP1 regulates the transcription of lipid metabolism genes to alleviate lipid-induced ER stress.³⁵ Whether any of these ER stress-associated factors is regulated in the tumor by bone marrow adipose tissue is unknown.

The goal of the present study was to examine the impact of bone marrow adipocytes on the transcriptome of bone-metastatic prostate carcinoma cells. Through RNAseq analyses, we demonstrate that exposure to adipocytes drives ER stress/UPR signature in metastatic prostate tumor cells of different origin, in a manner coordinated at least in part by BIP/HSPA5. Our findings also reveal that several of the ER-stress associated genes are overexpressed in metastatic samples from prostate cancer patients. Our collective observations point to shared mechanistic signatures at the transcriptomic level that enable metastatic cancer cells to thrive in previously inhospitable environments and pave the way for novel therapeutic interventions.

MATERIALS AND METHODS

Materials

DMEM, RPMI-1640, MEMα, insulin, and other chemicals, unless otherwise stated, were obtained from Sigma-Aldrich (St. Louis, MO). HyClone FBS, TaqMan reagents, and RNAiMAX were from ThermoFisher Scientific (Waltham, MA). Trypsin-EDTA was from Invitrogen (Carlsbad, CA). PureCol® collagen type I was from Advanced Biomatrix (San Diego, CA). Transwell cell-support systems were from Corning (Corning, NY). β-tubulin (#E7-C) antibody was from Developmental Studies Hybridoma Bank (Iowa City, IA). Antibodies to SYVN1 (#14773), XBP-1 (#12782), HSPA5 (BIP; #3177), DDIT3 (CHOP; #2895), ATF4 (#11815), HERPUD1 (#26730), ASNS (#20843), and E2F1 (#3742) were from Cell Signaling Technology (Danvers, MA). StemXVivo Adipogenic Suppliment, was from R&D Systems (Minneapolis, MN). RNeasy Mini Kits were from Qiagen (Valencia, CA). QuantSeq 3’ mRNA-Seq Library Prep Kit FWD from Lexogen (Vienna, Austria). Luminata Forte Western HRP substrate was from MilliporeSigma (Burlington, MA). Rosiglitazone was from Cayman Chemical (Ann Arbor, MI).

Cell Lines

PC3 cells were purchased from ATCC (Manassas, VA, USA). ARCaP(M) cells were purchased from Novicure Biotechnology (Birmingham, AL, USA). PC3 cells were cultured in DMEM with 10% FBS and ARCaP(M) cells were cultured in RPMI-1640 medium with 5% FBS. All media were supplemented with 25mM HEPES, and 100U/ml penicillin-streptomycin. Primary mouse bone marrow stromal cells (mBMSC) were isolated from tibiae and femurs of 6- to 8-week old FVB/N mice according to previously established protocols^12,36. To induce bone marrow adipocyte differentiation, mBMSCs were treated with adipogenic cocktail (30% StemXVivo Adipogenic Supplement, 1μM insulin, 2μM Rosiglitazone) in DMEM with 10% FBS for 8–10 days as previously described¹². Human cell lines used in this study have been authenticated by the WSU Genomics facility. All cell lines are routinely tested for mycoplasma using MycoFluor Mycoplasma Detection Kit (Thermo Fisher) and LookOut Mycoplasma PCR Detection Kit (Sigma). Cells were used within 10–12 passages from thawing. All cells were maintained in a 37°C humidified incubator ventilated with 5% CO_2.

Transwell Co-culture of tumor cells with bone marrow adipocytes

The mBMSC cells were embedded in Collagen I, plated in 6-well plates and differentiated into adipocytes according to our previously published protocols^12,15. Briefly, collagen-embedded mBMSCs were grown to confluency for 48–72 hours and treated with adipogenic cocktail for 8–10 days. After mature adipocyte cultures were established (with minimum of 70% BMSCs differentiated into fat cells), tumor cells were seeded on top of a Transwell filter (0.2μm pore size) to allow sharing of soluble factors between the two cell types. After 48- to 72-hour co-culture, cells were washed with PBS and collected for RNA and protein analyses. For RNA extraction, tumor cells were trypsinized and collected into RLT Plus buffer. RNA was extracted following the protocol in the RNeasy Plus Mini Kit. For protein collection, cells were washed with PBS, trypsinized, and lysates re-suspended in SME buffer with protease (MBL International, Woburn, MA) and phosphatase (Thermo Scientific, Waltham, MA) inhibitors. RNA quantity was measured with NanoDrop 1000 Spectrophotometer (Thermo Scientific, Waltham, MA).

HA15 Inhibitor Treatments

PC3 or ARCaP(M) cells were seeded in 96-well plates and cultured overnight, followed by treatment with BIP inhibitor, HA15 (MedChemExpress) for 48 hrs (range of concentrations from 0 to 50 μM). Effects of HA15 on viability were determined using Calcein A/M assay per the manufacturer’s instructions (Tocris Bioscience, Minneapolis, MN). Fluorescence was read using Infinite F200 PRO (Tecan, Männedorf, Switzerland) plate reader at 490nm excitation and 520nm emission. 2.5 μM concentration was chosen as the effective dose for the examination of effects on tumor-adipocyte co-cultures. Briefly, tumor cells were seeded in 6-well plates or in Transwell filters and cultured overnight. HA15 treatment was applied next day upon merging adipocyte and tumor cell cultures into the Transwell system. After 48 hours of co-culture, tumor cells were collected and processed for RNA extraction and Taqman RT PCR analyses as described above.

siRNA- mediated knockdown

Tumor cells were pre-plated in 6-well plates or on Transwell filters and grown overnight. The following day, a unique 27mer siRNA duplex targeting BIP/HSPA5 transcripts (OriGene-SR320562) or Trilencer-27 Universal scrambled negative control (Origene-SR30004) were added using RNAiMAX transfection reagent at a final concentration of 20 μM (based on manufacturer’s protocol). After 6 hours, media was replaced and Transwell filters containing transfected tumor cells were merged with bone marrow adipocytes. After 72 hours, cells were collected and processed for RNA analyses as described above. Two unique 27mer siRNA duplexes that efficiently knocked down BIP/HSPA5 transcripts were used (OriGene; SR320562).

RNAseq analyses

3’ RNA-seq (QuantSeq 3′ mRNA) was performed at the Wayne State University Applied Genomics Technology Center. RNA from four biological replicates of PC3 and ARCaP(M) cells cultured alone or in Transwell with adipocytes was collected as described above, and run on an Agilent TapeStation 2200 (Agilent Technologies, Santa Clara, CA) for quality control. Lexogen’s QuantSeq 3’mRNA-seq Library Prep Kit (FWD for Illumina) was utilized for building RNAseq libraries. The barcoded libraries were multiplexed at equimolar concentrations and sequenced with 50 bp single reads on an Illumina HiSeq-2500 run in rapid mode. Data were demultiplexed using Illumina’s CASAVA 1.8.2 software and reads were aligned to the human genomes³⁷. Differential gene expression analysis was used to detect adipocyte-mediated transcriptome changes³⁸.

Ingenuity Pathway and DAVID analyses

Significantly altered genes (log fold change ≥ 2; FDR ≤ 0.01) were used to identify enriched gene ontology terms and affected signaling pathways using Qiagen Ingenuity Pathway Analysis³⁹. For DAVID analyses, the list of upregulated transcripts for each condition was uploaded into the Functional Annotation Tool provided by DAVID (https://david.ncifcrf.gov, v. 6.8) and submitted using the official gene symbol as identifier and Homo sapiens as background. Charts were created from Biological Process (BP_DIRECT). Terms were ordered based on −log(P-value). Gene set enrichment analysis (GSEA) using GSEA tools (java GSEA Desktop Application version 3.0; http://software.broadinstitute.org/gsea/downloads.jsp) was used for further examination of GO BP pathways.

Taqman RT PCR

The cDNA was prepared from 1–2 μg of total RNA using High-Capacity cDNA Reverse Transcription kit (Applied Biosystems). Gene expression analyses were performed using TaqMan^® Individual Gene Expression assays for Human SYVN1 (HS00381211), XBP1 (Hs00231936), HSPA5 (BIP; Hs00946087), DDIT3 (CHOP; Hs01090850), ATF4 (Hs00909569), HERPUD1 (Hs01124265), ASNS (Hs04186191), E2F1 (Hs00153451), SESN2 (Hs00230241), GDF15 (Hs00171132), ATF6 (Hs00232586), DNAJB11 (Hs00212527), PCK2 (HS01091129), TRIB3 (Hs01082394), CHAC1 (Hs03043929). Assays were conducted on at least three biological replicates using TaqMan^® Fast Universal PCR Master Mix and 50 ng of cDNA/well and all reactions were run on an Applied Biosystems StepOnePlus^™ system. All genes were normalized to hypoxanthine phosphoribosyltransferase (HPRT1; Hs02800695). DataAssist^™ Software (Applied Biosystems) was used for all analyses. Data were presented as mean of at least 3 experiments ± SD and statistically analyzed using unpaired student T-test.

Immunoblotting

Lysate and media samples were loaded based on DNA/protein concentrations and the corresponding lysates were electrophoresed on Novex WedgeWell 4–20% Tris-Glycine Gels (Invitrogen), transferred to PVDF membranes and immunoblotted for SYVN1, XBP-1, HSPA5 (BIP), DDIT3 (CHOP), ATF4, HERPUD1, ASNS, E2F1, and Tubulin, all at 1:1000. All horseradish peroxidase-labeled secondary antibodies were used at 1:10,000. Quantification and analyses of bands were performed using a ChemiDoc Imaging System (BioRad).

Oncomine and cBioPortal analyses

The Oncomine database (Oncomine^™ v4.5: 729 datasets, 91,866 samples) was used for the analysis of primary (P) vs. metastatic (M) tumors by employing filters for selection of conditions and genes of interest (prostate cancer; metastasis vs. primary; genes). Data were ordered by ‘overexpression’ and the threshold was adjusted to P-value < 1E⁻4; fold change, 2; and gene rank, top 10%. For each database, only genes that met the criteria for significance were reported.

The cBioPortal for Cancer Genomics⁴⁰ was utilized to assess mRNA expression of 12 genes (HSPA5, ATF3, ATF6, DDIT3, ASNS, SESN2, SYVN1, TRIB3, DNAJB11, GDF15, HERPUD1 and XBP1) in prostate cancer tissues using the Metastatic Prostate Cancer, SU2C/PCF Dream Team cohort.⁴¹ The mRNA expression of each gene was plotted against the clinical attributes “Sample Type” and “Tumor Site”, allowing us to examine the levels of each gene in various metastatic sites, including bone (X Axis). mRNA in this cohort was isolated by PolyA selection and data are shown as log2 scaled analysis of RNA Seq Reads per Kilobase Million (RPKM).

Statistical analyses

For RNAseq analyses, differential gene expression between the control and exposure treatment was evaluated using DEseq2 (available through GenePattern; Broad Institute, Cambridge, Massachusetts). Genes with significant changes in expression, as defined by absolute log fold change ≥ 2; FDR ≤ 0.01, were uploaded into Ingenuity Pathway Analysis software (IPA; QIAGEN Bioinformatics, Redwood City, CA) for analysis using RefSeq IDs as identifiers. IPA-generated pathways were analyzed with a Right-Tailed Fisher’s Exact Test. For qRT-PCR and Western blotting analyses, data were statistically analyzed using unpaired student T-test and presented as mean of 3 or more independent experiments ± standard deviation (SD).

RESULTS

Adipocyte-mediated differential gene expression in PCa cells

We sought to gain a comprehensive and systems-wide understanding of how adipocyte-rich bone marrow impacts the transcriptional landscape of a tumor to begin exploring mechanisms through which metastatic cells respond to their microenvironment and survive inhospitable regions. We co-cultured metastatic prostate carcinoma cells with bone marrow-derived adipocytes using our well-established Transwell system^9,10,12,15. Two different prostate carcinoma cell lines were used: PC3 cells, originally derived from the vertebral metastatic prostate tumor⁴², are androgen receptor (AR)- and prostate specific antigen (PSA)-negative and grow independently of androgens⁴³; ARCaP(M) cells, the mesenchymal clone of ARCaP cells derived from ascites fluid from a man with metastatic prostate cancer⁴⁴, express low levels of AR and PSA and their growth in vivo is accelerated in castrated animals⁴³. Total RNA was extracted from each cell line after 48 hours of co-culture, and RNA QuantSeq analysis was performed (Figure 1; NCBI GEO, in process). A significantly higher number of genes differentially expressed upon adipocyte exposure was found in PC3 cells as compared to ARCaP(M) cells (1743 and 179, respectively). We found eighty-three commonly altered transcripts in ARCaP(M) and PC3 cells upon co-culture with adipocytes. Sixty-seven transcripts were upregulated and twelve were commonly downregulated (Figure 2A).

Figure 1. Diagram of Experimental Design.

Bone marrow stromal cells were isolated from FVB mice and stimulated to become adipocytes in vitro. Differentiated adipocytes were cultured with prostate carcinoma (PCa) cells using a Transwell co-culture system. Cells were collected 48 hours later, and RNA was extracted for RNAseq analyses.

Figure 2. Examination of Differentially Expressed Genes (DEG) reveals Unfolded Protein Response and ER stress in PCa cells exposed to adipocytes.

A: Venn Diagrams (http://bioinformatics.psb.ugent.be/webtools/Venn) show overlap in number of differentially expressed genes (DEG) between PC3 and ARCaP(M) cells cultured alone vs Transwell co-culture with adipocytes;as determined using the DESeq2 package (available through GenePattern; Broad Institute, Cambridge, Massachusetts); log fold change ≥ 2; FDR ≤ 0.01; 4 biological replicates for each condition; all genes (top panel), upregulated genes only (middle panel), downregulated genes only (bottom panel). B: Top 5 canonical pathways enriched in PC3 (top) and ARCaP(M) (bottom) cells upon exposure to bone marrow adipocytes (log fold change ≥ 2; FDR ≤ 0.01 threshold); Right-Tailed Fisher’s Exact Test; C: Enriched biological process Gene Ontology Terms as determined by DAVID (Database for Annotation, Visualization and Integrated Discovery) analysis; differentially expressed (log fold change ≥ 2; FDR ≤ 0.01) were sorted into ‘upregulated’ and ‘downregulated’ categories and the ‘upregulated’ genes were uploaded into DAVID’s functional annotation tool (https://david.ncifcrf.gov v.6.8). The term BP_DIRECT was selected for chart creation. Biological Processes are shown based on −log(p-value).

Using Ingenuity Pathway Analysis (IPA) we examined canonical pathways enriched in PC3 and ARCaP(M) cells upon exposure to bone marrow adipocytes by applying log2FC >1, FDR <0.01 threshold. ‘Unfolded Protein Response’ (UPR) emerged as a common, significantly enriched pathway among top 5 canonical pathways for each cell line (Figure 2B). Furthermore, ‘Cell Death and Survival’ was among top 5 molecular and cellular functions for both cell lines in addition to ‘Cellular Growth and Proliferation’ for PC3 cells and ‘Cellular Function and Maintenance’ for ARCaP(M) cells (Supplementary Table 1). Additional functional enrichment analysis of differentially expressed genes (DEGs) using DAVID (Figure 2C) and Gene Ontology (Supplementary Figure 1) revealed that, indeed, several biological processes related to UPR and ER stress are activated in both cell lines in response to Transwell co-culture with adipocytes. These pathways and processes suggest a critical influence from marrow adipocytes on stress-mediated life/death decisions for metastatic cancer cells.

We were intrigued by the shared signature of UPR activation between PC3 and ARCaP(M) cells when co-cultured with adipocytes. Therefore, we next visualized the effect of adipocytes on stress response in the tumor cells by generating the ‘UPR pathway’ in IPA, and overlaying FDR values and FC values calculated from comparing gene expression levels in PC3 and ARCaP(M) cells grown alone or in Transwell co-culture with adipocytes (Figure 3). This functional network analysis predicted that the upstream regulator of the observed ER stress and UPR response is BIP/HSPA5, a highly conserved molecular chaperone with pivotal roles in protein folding and UPR⁴⁵. A total of 24 adipocyte-induced genes with known involvement in ER stress, UPR and protein quality control pathways (including BIP/HSPA5) were identified (Figure 4). We subsequently used a combination of RT PCR and Western blotting analyses to confirm the RNA-seq data for UPR and ER stress hits and to further explore this transcriptional “signature” triggered on cancer cells by adipocytes.

Figure 3. Unfolded Protein Response Pathway in PC3 (A) and ARCaP(M) (B) cells.

‘Unfolded protein response pathway’ was generated in IPA, and overlaid with FDR values and FC values calculated from comparing gene expression levels in PC3 and ARCaP(M) cells grown alone or in Transwell co-culture with adipocytes. Analyses were performed with Ingenuity Pathway Analysis software (IPA; QIAGEN Bioinformatics, Redwood City, CA) using RefSeq IDs as identifiers. IPA-generated pathways were analyzed with a Right-Tailed Fisher’s Exact Test. Genes with an outer pink border are significantly regulated in our dataset, filled-in shape indicates up-regulated genes (red) and down-regulated genes (green) with darker colors being higher expression ratios. Shapes represent the category of the pathway components (complex, enzyme kinase, etc.) and lines represent direct interaction (solid line) and indirect interaction (dashed line) between molecules.

Figure 4. ER Stress/UPR-associated genes induced in prostate carcinoma cells by adipocytes.

Differentially expressed ER Stress/UPR-associated genes in PC3 and ARCaP(M) cells culture in Transwell with bone marrow adipocytes as determined by DEseq2 (available through GenePattern; Broad Institute, Cambridge, Massachusetts); log fold change ≥ 2; FDR ≤ 0.01; 4 biological replicates for each condition. Genes are ordered alphabetically and LogFC and P-values for each gene are shown. Association of each gene with stress response pathways is indicated by specific cell colors.

Validation of adipocyte-induced stress response genes

We performed Taqman RT PCR and western blotting analyses of PC3 and ARCaP(M) cells grown alone or in Transwell co-culture with adipocytes for 8 factors selected for validation at the gene and protein level. The selected hits included BIP/HSPA5, ATF4, XBP1, SYVN1, DDIT3, HERPUD1, ASNS and E2F1 (Figure 5). In agreement with the RNAseq results, the message and protein levels of BIP/HSPA5, ATF4, XBP1, SYVN1, DDIT3, HERPUD1 and ASNS were significantly upregulated in both cell lines, demonstrating adipocyte-induced UPR. Both transcript and protein levels of E2F1 were reduced upon exposure to adipocytes, indicating inhibition of protein translation in response to ER stress. Further Taqman RT PCR analyses were performed to examine the expression of an additional 7 UPR-related genes: SESN2, GDF15, ATF6, DNAJB11, PCK2, TRIB3 and CHAC1. All 7 genes were significantly induced by Transwell co-culture in both cell lines (Figure 6), validating the RNAseq results and further supporting the notion of adipocyte contribution to UPR induction in tumor cells as a shared response signature. This signature coincided with adipocyte-driven sustained proliferation and survival, as we have previously demonstrated by increased clonogenic growth, augmented levels of cyclin D1 and reduced docetaxel response^9,10. It is noteworthy that, with exception of SYVN1, baseline levels of several ER stress markers were significantly lower in PC3 cells as compared to ARCaP(M) cells, which could account for a stronger ER stress/UPR response in this cell line (Supplementary Figure 2). Notably, co-culture of undifferentiated BMSCs with PC3 cells had no major effect on ER stress response in PC3 cells. Although changes in the expression of 2 genes -- BIP/HSPA5 and DDIT3 -- were significant, their levels of induction were an order of magnitude lower than those observed in the presence of adipocytes, further underscoring the role of adipocytes in driving the ER stress /UPR response in tumor cells.

Figure 5. RT PCR and western blot validation of select RNAseq hits.

A: HSPA5 (BIP); B: ATF4; C: XBP1; D: SYVN1, E: DDIT3, F: HERPUD1, G: ASNS; and H: E2F1. Top panels (A-H): Taqman RT PCR results depicting expression of each gene in PC3 and ARCaP(M) cells cultures alone or in Transwell with adipocytes. Bottom panels (A-H): western blot analyses depicting expression of each protein in PC3 and ARCaP(M) cells cultures alone or in Transwell with adipocytes; densitometric analyses of bands normalized to either tubulin or actin are shown above each blot. Taqman RT PCR and western blot data are shown as individual values from at least 3 independent biological replicate experiments; statistical values were determined by student T-test * p<0.05; ** p < 0.01; ***p < 0.001; ****p<0.0001.

Figure 6. RT PCR validation for select RNAseq hits.

A:SESN2; B:GDF15; C:ATF6; D:CHAC1, E:PCK2, F:TRIB3, G:DNAJB11. Taqman RT PCR results depicting expression of each gene in PC3 and ARCaP(M) cells cultures alone or in Transwell with adipocytes. H: Taqman RT PCR results depicting expression of ASNS, ATF4, BIP, SYVN1, GDF15, HERPUD1, PCK2, SESN2, DDIT3 and XBP1 in PC3 cells cultures in the presence or absence of BMSCs. Data are shown as individual values from at least 3 independent biological replicate experiments; statistical values were determined by student T-test * p<0.05; ** p < 0.01; ***p < 0.001; ****p<0.0001.

Inhibition of adipocyte-induced UPR

Our IPA analysis predicted that the adipocyte-driven stress response in tumor cells is largely orchestrated by BIP/HSPA5 (Figure 3). We also showed previously that BIP/HSPA5 overexpression upon adipocyte exposure in vitro and in mice with diet-induced marrow adiposity coincides with oxidative stress response and the activation of pro-survival pathways¹⁰. Therefore, we utilized a novel, selective inhibitor of BIP/HSPA5, the thiazole benzensulfonamide compound HA15⁴⁶, to examine whether the targeting of BIP/HSPA5 activity has an impact on the adipocyte-induced stress response.

HA15 binds to and inhibits the ATPase activity of BIP/HSPA5, dissociating it from PERK, IRE1α, and ATF6, which can ultimately lead to ER stress-mediated cell death, as previously demonstrated in melanoma cells⁴⁶. Indeed, Calcein A/M assays revealed that, in our hands, HA15 concentrations above 5 μM for PC3 cells and 10 μM for ARCaP(M) cells significantly reduced viability (Figure 7A). Importantly, treatment of Transwell co-cultures with a lower, 2.5 μM concentration of HA15, that does not significantly impact viability, reduced mRNA levels of BIP/HSPA5, HERPUD1 and SYVN1 but did not significantly impact other adipocyte-induced UPR-related genes, including ATF4, SESN2, ASNS, GDF15 and PCK2 (Figure 7B). To determine if genetically targeting BIP/HSPA5 also impacts UPR genes in co-cultures, we knocked it down with siRNA in ARCaP(M) cells grown alone or in Transwell co-culture with adipocytes. Silencing BIP/HSPA5 in cells grown in the absence of adipocytes had no significant effect on the expression of most of the UPR genes tested, with the exception of increases in transcript levels of XBP1 and DDIT3. Knockdown of BIP/HSPA5 in cells grown in Transwell co-culture with adipocytes, however, significantly induced SYVN1, HERPUD1, ASNS, SESN2, XBP1, DDIT3 and GDF15 (Figure 7C). Taken together, these data indicate BIP/HSPA5 as a key player of adipocyte-driven stress response. The difference in direction of change between BIP/HSPA5 inhibitor and its genetic targeting is likely due to the method of interference (inhibition of complex formation, without depleting BIP/HSPA5 protein vs. reduction of overall BIP/HSPA5), and highlights the need to further investigate the molecular role of BIP/HSPA5 in this context. We propose that adipocyte-induced BIP/HSPA5 expression is an adaptive response of tumor cells to lipid-rich microenvironment; silencing of this important UPR gene might invoke an integrated stress response and reduced potential for survival, which requires further examination.

Figure 7. Effects of targeting activity and gene expression of BIP/HSPA5 on transcription of ER stress/UPR genes.

A: Calcein A/M viability assay demonstrating effects of increasing concentrations of HA15 on the viability of PC3 (top panel) and ARCaP(M) (bottom panel) cells; Taqman RT PCR results depicting expression of HSPA5 (BIP), SYVN1, HERPUD1, ATF4, SESN2, GDF15 and PCK2 in PC3 cells cultures alone or in Transwell with adipocytes in the absence or presence of 2.5 μM HA15 (B) or upon silencing of BIP/HSPA5 with non-overlapping siRNAs (C). Data are shown as individual values from 3 independent experiments; statistical values were determined by student T-test * p<0.05; **p < 0.01; ***p<0.001; n.s. – not significant.

In silico analysis of ER stress genes in metastatic samples from prostate cancer patients

The data we collected so far highlight an important role for UPR and ER stress response for cancer cells to accommodate to the harsh environment created by adipose cells. To extrapolate our findings to the human condition, we next performed Oncomine analysis of primary and metastatic prostate tumors and compared mRNA expression of several of UPR and ER stress genes that we found to be modulated by exposure to adipocytes.

Examination of the Grasso prostate dataset, which contains a significant number of metastatic samples compared to other available datasets, revealed statistically significant increases in the expression of ASNS, ATF4, SESN2, TRIB3, PCK2, SYVN1, DNAJB11 and DDIT3 in metastatic as compared to primary tissues (Figure 8A). Overexpression of the majority of these genes was confirmed in the Chandran prostate dataset (Figure 8B).

Figure 8. Expression of ER stress/UPR genes in tissue samples from prostate cancer patients.

Oncomine gene analysis of Grasso (A) and Chandran (B) prostate databases comparing the expression of ASNS, ATF4, SESN2, TRIB3, PCK2, SYVN1, DNAJB11 and DDIT3 in patient samples collected from metastatic or primary sites. Data were ordered by ‘overexpression’ and the threshold was adjusted to P-value <1E⁻4; fold change, 2 and gene rank, top 10%.

Several other Oncomine datasets containing metastatic tissues were interrogated, each demonstrating increased levels of at least two of the above genes (Supplementary Table 2). ASNS was found to be a most frequently altered gene in metastatic tissues among the 8 genes surveyed (7 datasets). Additional cBioPortal polyA transcriptome analyses of RNA-seq filtered for mRNA from the Metastatic Prostate Cancer, SU2C/PCF Dream Team cohort showed that expression of the 8 genes identified by our Oncomine analyses is indeed augmented in metastatic sites, including bone (Supplementary Figure 3). Furthermore, 6 additional genes revealed by our transcriptome analyses to be regulated by adipocytes (BIP/HSPA5, ATF3, ATF6, GDF15, HERPUD1 and XBP1) are also preferentially expressed in metastatic tissues as opposed to other tumor sites (Supplementary Figure 4). Collectively, these data underscore the significance of a UPR/ER stress-related transcriptomic signature in metastatic disease as a potential key player in the cancer cell survival in harsh microenvironments.

DISCUSSION

Despite its complexity and uniquely harsh environmental conditions, bone marrow is a major site of metastasis from several tumor types including prostate cancer.^47,48 In order to grow and thrive within the bone marrow space, metastatic tumor cells develop interactions with various cell types, including osteoclasts, osteoblasts, endothelial cells, hematopoietic stem cells, immune cells and adipocytes, and they transform the metastatic niche for their own benefit.^8,49,50 Tumors colonizing bone marrow disrupt normal bone remodeling, alter coupling between osteoclasts and osteoblasts, stimulate the release of growth factors and cytokines^7,51,52, and shape the metabolism of their microenvironment to assure sufficient supply of energy needed for metastatic progression.^7,51–54. Exact mechanisms that can be targeted therapeutically for the treatment of metastatic disease, however, remain unclear.

Hypoxia, nutrient deprivation, metabolic and oxidative stresses are typical conditions that tumor cells encounter and adapt to within the bone marrow. Cancer cells respond to low oxygen levels by augmenting Hypoxia Inducible Factors (HIF)-1 and -2.^55,56 HIF-1 signaling is important for the initial colonization of tumor cells in bone via Lysyl Oxidase (LOX)⁵⁷. HIF-1 is also a critical determinant of tumor growth and progression in bone⁵⁵. HIF-1α acts as a transcriptional activator of a number of glycolytic genes⁵⁸. Accordingly, HIF-1α activation in metastatic prostate carcinoma cells drives their metabolic re-programming and promotes Warburg phenotype¹⁵. Increased HIF activity also alters mitochondrial function and reduces Oxidative Phosphorylation activity in the tumor, providing a survival mechanism under hypoxic stress.⁵⁹ Hypoxia is one of the key inducers of oxidative stress, which arises from the imbalance between ROS production and the cell’s ability to repair the damage^60,61. Persistently high ROS levels, combined with the compensatory increases in antioxidant enzymes, promote tumor progression^62,63. This is in line with our previous work showing that metastatic PCa cells upregulate the oxidative stress enzyme heme oxygenase 1 (HO-1) as the means of creating an anti-oxidant system needed for survival under oxidative stress conditions¹⁰. HO-1 levels are highly induced in metastatic tissues from prostate cancer patients, and forced expression of HO-1 in prostate tumor cells promotes tumor growth and invasion¹⁰.

Both hypoxia and HO-1 upregulation are linked to UPR^64,65, and UPR activation and disruption of protein folding is a pro-survival mechanism for tumor cells growing in the hypoxic, nutrient-deficient bone marrow space^66,67. Hypoxia contributes to phosphorylation of eIF2A and global dampening of protein synthesis⁶⁸. Transcriptional induction of ATF4 by eIF2A regulates GADD34, a driver of recovery from ER stress, and DDIT3, a pro-apoptotic gene involved in life vs death decisions⁶⁹. Another key effect of hypoxia is to activate the IRE1–XBP1 arm of the UPR, which has been shown to contribute to the progression of triple negative breast cancers⁷⁰. Hypoxia and protein misfolding also result in activation and cleavage of ATF6, which increases the transcription of genes that expand ER capacity and ameliorate ER stress^71,72. We previously showed that adipocytes promote oxygen-independent HIF-1α activation in prostate cancer cells¹⁵, and induce expression of the ER stress chaperone BIP¹⁰; but, how marrow adiposity affects UPR in a context of metastatic progression had not been previously explored.

Adipocytes, which are abundantly present in adult bone marrow, secrete lipids, hormones, adipokines, and growth factors, and their tumor-promoting and supportive functions have been demonstrated in a number of cancers^7,73,74. No studies to date, however, have examined the impact of adipocyte-rich bone marrow, on the transcriptional landscape of metastatic tumor cells. Therefore, we performed RNAseq analysis of two prostate carcinoma cells lines of different origin cultured alone or in the Transwell system with bone marrow adipocytes and identified UPR and ER stress as pathways commonly regulated by adipocyte exposure. A set of genes emerged that are either directly involved in UPR and ER stress responses, or are downstream metabolic and oxidative stress genes associated with UPR and ER stress. We also confirmed that several of those genes are overexpressed in metastatic tissues from prostate cancer patients.

Our pathway analyses highlighted the ER chaperone BIP as an important player in tumor cell response to adipocytes, whose activity in prostate tumor cells might be important for regulating two UPR genes -- SYVN1 and HERPUD1. SYVN1 is an E3 ubiquitin ligase involved in ERAD, specifically in the recognition, ubiquitination and disposal of harmful proteins to reduce ER damage and inhibit apoptosis.⁷⁵ HERPUD1 is an ER protein with cytoprotective functions against ER and oxidative stress.⁷⁶ HERPUD1 inhibits several caspases, serves as a negative regulator of autophagy and maintains cellular homeostasis under stress conditions^76–78. The role of neither of these proteins has been explored in metastatic prostate cancer. We showed previously that hypoxia and abundant fatty acids in adipocyte-rich bone marrow modulate the local redox environment and promote tumor survival.¹⁰ Accordingly, overexpression of HERPUD1 upon adipocyte exposure might be a cancer cell defense mechanism to survive under high oxidative stress conditions. It is also not surprising that we observed increases in tumor SYVN1 expression upon co-culture with adipocytes, as this protein can be induced by lipid overload³⁰. It is, however, intriguing that SYVN1 reportedly regulates energy expenditure and mitochondrial oxidative metabolism via the degradation of peroxisome proliferator-activated receptor gamma co-activator 1 beta (PGC-1β)³⁰. It has been reported that expression of PGC-1β and its relative PGC-1α is inversely correlated with prostate cancer progression⁷⁹. Since SYVN1 levels are increased in metastatic tissues from prostate cancer patients and in tumor cells exposed to marrow adipocytes, it is plausible that this ligase plays pleiotropic roles in metastatic disease, warranting further investigation.

Another important observation from our study is that exposure of prostate carcinoma cells to adipocytes in vitro results in augmented expression of several genes associated with mitochondrial stress response including ASNS, SESN2 and PCK2. ASNS, an enzyme which catalyzes the synthesis of asparagine and glutamate⁸⁰, is implicated in late-stage castrate-resistant prostate cancer⁸¹. SESN2 protects carcinoma cells from oxidative stress and damage⁸². Augmented expression of PCK2 correlates with aggressiveness and lower survival rates in prostate cancer⁸³, as well as adaptive response and redox balance maintenance in lung cancer⁸⁴. Intriguingly, all three genes are transcriptional targets of ATF4^85–89 and our data showed that all three, along with ATF4, are overexpressed in patient metastatic tissues. Our study is the first to indicate a potential impact of adipocyte-rich bone microenvironment on the expression of ATF4 and its target mitochondrial stress genes at the metastatic site. ATF4 expression was recently reported to be higher in prostate tumor tissues as compared to benign samples, and silencing of ATF4 profoundly inhibited tumor growth in preclinical models.⁹⁰ However, the molecular mechanisms behind the action of this powerful transcription factor in adaptive behavior of metastatic tumors in bone are not known and need further investigation.

Adaptation to the hostile metastatic environment is a prerequisite for tumor survival and therapy evasion. Although significant progress has been made in identifying the mechanisms responsible for tumor adaptation in bone, metastatic prostate cancer remains incurable. Here, through unbiased transcriptomic analyses we identified UPR as a common adipocyte induced pathway in metastatic prostate carcinoma cells and uncovered its involvement in adipocyte-driven regulation of pro-survival signaling in skeletal tumors. To our knowledge, this is the first reported systemics approach examining the impact of adipocyte-rich bone marrow on the transcriptional landscape of metastatic tumors. Our findings -- pointing to specific stress response pathways in adipocyte-mediated tumor survival -- have clinical implications. UPR is a complex and dynamic signaling framework integrated with various cellular processes critical to tumor growth and progression, including lipid metabolism, redox balance, inflammation, autophagy and integrated stress response. Various inhibitors of specific components of UPR are being developed and tested in preclinical and clinical studies^91,92. It is also becoming increasingly evident that the interactions of metabolic and stress pathways contribute to tumor progression and metastasis. As a result of our current work, future investigations are well positioned to provide a comprehensive framework of molecular mechanisms that enable cancer cells to survive and thrive in difficult micro-environments and to co-opt such pathways for clinical intervention.

HIGHLIGHTS.

• Bone marrow adipocytes induce Unfolded Protein response (UPR) and ER stress gene signature in metastatic prostate carcinoma cells

• ER chaperone BIP is an upstream regulator of adipocyte-driven UPR response

• UPR and ER stress-associated genes are upregulated in metastatic tissues from prostate carcinoma patients

• Enhanced UPR and ER stress response represent a potential mechanism of tumor adaptation to adipocyte-rich bone marrow niche