PGC1α suppresses prostate cancer cell invasion 1 through ERRα transcriptional control 2

23 The peroxisome proliferator-activated receptor gamma coactivator 1 alpha (PGC1α) is a 24 prostate tumor suppressor that controls the balance between anabolism and catabolism. 25 PGC1A downregulation in prostate cancer is causally associated with the development of 26 metastasis. Here we show that the transcriptional complex formed by PGC1α and 27 Estrogen related receptor 1 alpha (ERRα) controls the aggressive properties of prostate 28 cancer cells. PGC1α expression significantly decreased migration and invasion of various 29 prostate cancer cell lines. This phenotype was consistent with remarkable cytoskeletal 30 remodeling and inhibition of integrin alpha 1 and beta 4 expression both in vitro and in 31 vivo. CRISPR/Cas9-based deletion of ERRα suppressed PGC1α regulation of cytoskeletal 32 organization and invasiveness. Mechanistically, PGC1α expression decreased MYC levels 33 and activity prior to inhibition of invasiveness. In addition, PGC1α and ERRα associated at 34 the MYC promoter, supporting the inhibitory activity PGC1α. The inverse correlation 35 between PGC1α-ERRα activity and MYC levels was corroborated in multiple prostate 36 cancer datasets. Altogether, these results support that PGC1α-ERRα functions as a tumor 37 suppressive transcriptional complex through the regulation of metabolic and signaling 38 events. 39 how downregulation of the prostate tumor suppressor PGC1 drives invasiveness and migration of prostate cancer cells. are two key signaling pathways that regulate cytoskeletal Similar


Abstract 23
The peroxisome proliferator-activated receptor gamma coactivator 1 alpha (PGC1α) is a 24 prostate tumor suppressor that controls the balance between anabolism and catabolism. The process of cellular transformation stems from the acquisition of genomic aberrations 44 that altogether change the response of normal cells and enable them with hallmarks of 45 cancer (1,2). The mutational landscape changes within and among tumors and along time 46 following evolutionary principles (3). In addition, non-genomic alterations harness great 47 relevance in the process of cancer progression. Indeed, transcriptional regulation in cancer 48 is an emerging aspect that provides a feasible explanation to the rapid adaptation of 49 transformed cells to hostile environments (4). Yet, the control of oncogenic and tumor 50 suppressive transcriptional programs remains poorly characterized. 51 Transcriptional co-regulators encompass a family of versatile modulators of gene 52 expression (5). These proteins harbor the capacity of controlling distinct transcriptional 53 programs based on their partner transcription factors. In turn, transcriptional co-regulators 54 operate in a tissue and context-specific manner, thus revealing them as major players in 55 cell and organismal homeostasis. Among this family of genes, the Peroxisome proliferator-56 activator receptor (PPAR) gamma co-activator 1 alpha (PGC1α) controls biological 57 responses in health and disease (6,7). PGC1α is a tightly regulated protein that interacts 58 with a variety of transcription factors, including Estrogen related receptor 1 alpha (ERRα), 59 renal and prostate carcinoma, as well as in metastatic melanoma, where it opposes the 67 acquisition of aggressive features (15)(16)(17). The predominant mechanism of action of 68 PGC1α in cancer biology is ascribed to the regulation of metabolism. This co-regulator 69 promotes the expression of genes that mediate mitochondrial biogenesis, oxidative 70 metabolism and the production of glutathione. In turn, PGC1α enhances the oxidative 71 utilization of nutrients and antioxidant production. However, emerging data suggest that a 72 fraction of the activities of PGC1α neither relies on the regulation of metabolism nor on its 73 main partner, ERRα (16). 74 In prostate cancer (PCa), PGC1α suppresses cell proliferation, anchorage-independent 75 growth, tumor burden and metastasis (17). This co-regulator is profoundly downregulated 76 in localized PCa, with a further decrease in metastatic specimens (17). Moreover, reduced 77 PGC1α expression is associated to shorter time to biochemical recurrence after surgery, 78 pointing at the relevance of this gene in the control of PCa aggressiveness. 79 Mechanistically, we previously showed that PGC1α requires the presence of ERRα to 80 suppress PCa cell proliferation and metastatic outgrowth, which was consistent with the 81 reduction of biosynthetic capacity of PGC1α re-expressing cells and the elevation of 82 nutrient catabolism (17). Moreover, a recent study revealed that the metabolic control of 83 polyamine synthesis underlies the regulation of prostate cancer aggressiveness by this co-84 activator (18). 85 The metastatic process requires the acquisition of discreet capacities beyond cell 86 proliferation. Specifically, the motility and invasive capacity of cancer cells is paramount for 87 the achievement of metastasis (19). Stemming from this notion, in this study we evaluated 88 the contribution of PGC1α to the acquisition of these features in PCa cells. Our analysis 89 uncovers an ERRα-dependent activity of the co-activator that suppresses the acquisition of 90 invasive properties required for PCa aggressiveness. 91 formalin and stained with crystal violet as previously described (17). 148

Materials and Methods
Transwell invasion assay was carried out using matrigel-coated chambers (BD CioCoatTM 149 #354480). Cells (50,000 cells/well) were re-suspended in 0.1 % FBS DMEM and seeded in 150 the upper part of the chamber. In the bottom part of the well 1.4 mL of complete DMEM 151 were added. Plates were maintained at 37 ºC and 5 % CO 2 for 48 hours. Invasion was 152 stopped washing the well twice with PBS and using a cotton bud to remove the remaining 153 cell of the upper part of the membrane, being careful not to compromise the matrigel. The 154 membrane was fixed with 10 % formalin (15 minutes at 4 ºC) and stained with crystal violet 155 (Sigma C3886; 0.1% crystal violet in 20% methanol). Cells were counted under the 156 microscope. For transwell migration, chambers with membranes of 8 m pores (BD Falcon 157 351185) were used. Cell plating as well as washing and fixation conditions were the same 158 as in the invasion assay, but cells were fixed after 24 hours. 159 Spheroid cell culture and 3D invasion assays were performed as previously described 160 (23). Briefly, cells (700 cells/drop) were maintained in drops (25 L/drop) with DMEM and 161 6 % methylcellulose (Sigma M0387) on the cover of a 100 mm culture plate. Drops were 162 incubated at 37 ºC and 5 % CO 2 for 48 hours. Once formed, spheroids were collected, 163 resuspended in collagen I solution (Adanced BioMatrix PureCol®) and added to 12-well 164 plates. After 4h complete media was then added on top of the well and day 0 pictures were 165 taken. For invasive growth quantification, increase in area occupied by the spheroids 166 between day 0 and day 2 was calculated using FiJi software. For 3D invasion assays, cells 167 were resuspended in a FBS-free bovine collagen I solution at 2.3 mg/mL in a 1:1 168 proportion, to a final concentration of 15000 cells per 100 µL of matrix and spin down in a 169 96-well plate. After matrix polymerization, 10% FBS-containing media was added on top. Western blot was performed as previously described (9). Briefly, cells were seeded on 6-175 well plates and 4 days after seeding cell lysates were prepared with RIPA buffer (50mM 176 TrisHCl pH 7.5, 150 mM NaCl, 1mM EDTA, 0.1 % SDS, 1 % Nonidet P40, 1 % sodium 177 deoxycholate, 1 mM Sodium Fluoride, 1 mM sodium orthovanadate, 1 mM beta- and Western blotting techniques, proteins were visualized using the ECL system in the 187 iBright TM FL1000 Imaging System. 188 The cytoskeleton phospho-antibody array was performed following Tebu-bio protocol 189 (https://www.tebu-bio.com). Briefly, 5 million induced and non-induced cells were collected 190 and the cell pellet was frozen for further analysis by Tebu-bio services. Pgc1a cells per immunoprecipitation were grown in 150 mm dishes either with or without 206 0.5 µg mL -1 doxycycline during 16 hours. Cells were cross-linked with 37 % formaldehyde 207 for 10 min at room temperature. Glycine was added to dishes, and cells incubated for 5 208 min at room temperature. Cells were then washed twice with ice-cold PBS and scraped 209 into PBS+PIC. Pelleted cells were lysed and nuclei were harvested following the 210 manufacturer's instructions. Nuclear lysates were digested with micrococcal nuclease for 211 20 min at 37 °C and then sonicated in 500 μl aliquots on ice for 6 pulses of 20 s using a 212 on March 7, 2020. © 2019 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Author Manuscript Published OnlineFirst on October 8, 2019; DOI: 10.1158/0008-5472. CAN-19-1231 Branson sonicator. Cells were held on ice for at least 20 s between sonications. Lysates 213 were clarified at 11000 × g for 10 min at 4 °C, and chromatin was stored at -80 °C. HA-Tag 214 polyclonal antibody (CST #3724), anti-ERRα antibody (CST #13826) and IgG antibody 215 (CST #2729), were incubated overnight (4 ºC) with rotation and protein G magnetic beads 216 were incubated 2 hours (4 ºC). Washes and elution of chromatin were performed following 217 manufacturer's instructions. DNA quantification was carried out using a Viia7 Real-Time 218 PCR System (Applied Biosystems) with SybrGreen reagents and primers that amplify a 219 has been used for these calculations, together with ggplot2 package (https://cran.r-231 project.org/web/packages/ggplot2) in order to perform the corresponding graphs. 232 Individual gene expression patters in Patient dataset, as well as pairwise correlation 233 information can be visualized in the Cancertool interface. 234 The differential gene expression analysis driven by PGC1α in PC3 cells can be obtained 235 from GEO with reference GSE75193. 236 In addition, pathway and network enrichment analyses of the significantly regulated genes 237 from GSE75193 (Supplementary Table S3)

252
In order to address the role of PGC1α in the regulation of PCa features beyond 253 proliferation (17), we carried out a comprehensive evaluation of phenotypes associated to 254 cancer aggressiveness, based on an inducible system previously reported (17). 255 Interestingly, Pgc1α expression elicited a remarkable reduction in the migratory capacity of 256 PC3 and DU145 PCa cells in transwell assays ( Fig. 1A; Supplementary Fig. 1A). A 257 similar effect was achieved in matrigel-coated transwell assays as a measure of invasion 258 We next focused on the molecular alterations underlying the activity of PGC1α. In a 279 previous study, we analyzed a gene expression analysis in PC3 cells upon induction of 280 Pgc1α ( Fig. 1) (17) (GSE75193). We sought to extend the analysis of this microarray by 281 taking advantage of bioinformatic tools, such as Metacore 282 (https://clarivate.com/products/metacore/) and Cancertool (25) that enable cancer 283 researchers to perform various functional enrichment analyses. Since functional 284 enrichment allows the integration of larger sets of data in order to identify underlying 285 molecular and functional alterations, we focused our analyses on all genes whose 286 expression was altered with a significant p-value in the transcriptomics analysis 287 (regardless of the adjusted p-value). This led to 1347 upregulated and 990 downregulated 288 unique gene IDs (Supplementary Table S3). Strikingly, functional enrichment of the 289 downregulated genes revealed a significant alteration in cytoskeleton organization, 290 migration, adhesion and integrin and Rho signaling ( Fig. 2A; Supplementary Fig. 2A; 291 Supplementary Table S4-S5). Of note, we also identified other pathways with reported 292 activities in the regulation of invasion, such as p27, FAS and RAC, although their 293 prevalence in the analysis and their documented association to this phenotype were minor 294 (26-29). In line with our previous study (17), the enrichment analysis of the genes 295 upregulated upon Pgc1α expression confirmed a significant alteration of catabolic 296 Table S6 properties. In order to ascertain which signaling pathways were modulated and affecting 307 cytoskeleton organization upon Pgc1α expression, we carried out a cytoskeleton phospho-308 antibody array (Supplementary Table S7). The phosphorylation of Src protein was among 309 the most prominently reduced in the analysis (Supplementary Fig. 2B). We confirmed this 310 result by western blot analysis, both in vitro and in vivo, together with the reduction in 311

pathways (Supplementary
Cofilin phosphorylation, the final effector of actin filament polymerization downstream Src 312 (Fig 2C, D, Supplementary Fig. 2C, D). 313 Integrins are upstream regulators of the cytoskeleton with well-documented involvement in 314 cancer aggressiveness (19,31,32). The bioinformatics analysis of PGC1α downregulated 315 genes indicated an altered integrin signaling ( Fig. 2A, Supplementary Fig. 2A), which 316 would be consistent with the reduction in Src, MLC2 and Cofilin phosphorylation. This, 317 together with the fact that PGC1α controls integrin expression in melanoma (16), prompted 318 us to evaluate integrin expression in our experimental systems. Interestingly, the levels of 319 various integrins and caveolin-1 (CAV1, but not CAV2) were robustly reduced at protein 320 and mRNA levels upon Pgc1α induction, an event that was not influenced by doxycycline 321 treatment ( Fig. 2E; Supplementary Fig. 2E-I). Next, we analyzed extracts from 322 xenografts in which Pgc1α expression was activated (Fig. 1G). The western blot and 323 quantitative qRTPCR analysis corroborated the alterations elicited by the co-activator in 324 vivo ( Fig. 2F; Supplementary Fig. 2J, K). Our results suggest that PGC1α controls a 325 transcriptional program that results in the alteration of cytoskeleton organization with the 326 concomitant reduction in integrin expression, an event that is consistent with the observed 327 reduction in migratory and invasive properties of PCa cells. 328 We then asked which effector of PGC1α could contribute to the negative regulation of 329 invasive properties. Inhibitors of differentiation (ID) are responsible for integrin repression 330 in melanoma (16). We ruled out the potential contribution of ID2-4 to our phenotype, since 331 their expression was not upregulated upon induction of the co-activator (Supplementary 332  (Fig. 3C). We took advantage of our Pgc1α-inducible xenograft 342 analysis to further demonstrate that the reduction in MYC expression and function was not 343 an artifact of in vitro assays (Fig. 3D, E; Supplementary Fig. 3D). These results suggest 344 that MYC repression is upstream of the molecular and cellular alterations elicited by 345 PGC1α associated to PCa invasion. We validated this notion by two different means. On 346 the one hand, a time course experiment upon PGC1α induction showed that MYC 347 repression is prior to the reduction of its targets and integrin gene expression (Fig. 3F, 348 Supplementary Fig. 3E-G). On the other hand, MYC silencing with a validated shRNA 349 (33,34) recapitulated the phenotype of Pgc1a expression in cell area, p-MLC2 and 350 invasive growth (Supplementary Fig. 3H-L). 351 on March 7, 2020. © 2019 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Author Manuscript Published OnlineFirst on October 8, 2019; DOI: 10.1158/0008-5472. CAN-19-1231 The rapid repression in MYC mRNA levels prompted us to evaluate whether PGC1α could 352 exert a direct action on MYC promoter. To this end, we performed chromatin 353 immunoprecipitation (ChIP) analysis in Pgc1α-inducible PC3 cells with anti-HA antibody in 354 order to immunoprecipitate ectopic tagged Pgc1α. The ChIP analysis confirmed that the 355 co-regulator is bound to MYC promoter (Fig. 3G), thus suggesting that PGC1α represses 356 MYC expression in PCa. We next sought to ascertain whether the unprecedented 357 regulation of MYC by PGC1α in PCa could be recapitulated in human specimens. We 358 interrogated 5 PCa datasets (25,(35)(36)(37) and, in agreement with our molecular and 359 mechanistic data, PGC1A expression was inversely correlated with MYC mRNA levels in 360 primary tumors from the majority (4 out of 5) of datasets analyzed ( Fig. 3H and 361 Supplementary Fig. 3M). 362 Our previous studies demonstrated that the anti-proliferative activity of PGC1α in PCa is 363 dependent on its interaction with ERRα (17). In order to ascertain the requirement of 364 ERRα for the anti-invasive activity of PGC1α, we engineered Pgc1α-inducible PCa cells in 365 which ESRRA was deleted using CRISPR/Cas9. ERRα expression was undetectable in 366 PC3 cells in which ESRRA was deleted with two independent short guide RNAs 367 (sgERRα#1, sgERRα#2) (Fig. 4A). ESRRA deletion abolished the induction of target 368 genes of the transcription factor upon induction of Pgc1α, corroborating the functionality of 369 the genetic system (Supplementary Fig. 4A). Of note, we did not recapitulate the 370 regulation of ESRRA by PGC1A observed in vitro (Fig. 4A) in correlative human 371 transcriptomics analyses, suggesting that more complex ERRα-regulatory cues might 372 operate in human disease (Supplementary Fig. 4B). In line with our previous study (17), 373 ESRRA deletion hampered the growth suppressive activity of Pgc1α, rendering PC3 cells 374 insensitive to the action of the co-regulator (Fig. 4B) Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Author Manuscript Published OnlineFirst on October 8, 2019; DOI: 10.1158/0008-5472. CAN-19-1231 prior to the reduction in cell proliferation, thus demonstrating that the regulation of invasion 377 by the co-regulator is exquisitely dependent upon its interaction with ERRα (Fig. 4C, D  378 and Supplementary Fig. 4C, D). The morphological changes and growth suppressive 379 phenotype elicited by Pgc1α were also absent in tumors in which ESRRA was deleted 380 (Fig. 4E, Supplementary Fig. 4E, F, G). It is worth noting that despite the requirement of 381 ERRα for the tumor suppressive activity of PGC1α, deletion of the nuclear receptor alone 382 negatively influenced the establishment of tumors, suggesting that additional functions of 383 ERRα may be required for the first stages of tumor establishment (Supplementary Fig.  384

4H). 385
We next extended our analysis of ERRα dependency to the reported molecular alterations. 386 Our results showed that ESRRA deletion abrogated the reduction in protein and/or mRNA 387 levels of MYC, MYC targets, integrins, CAV1 as well as the reduced phosphorylation of 388 Src and Cofilin (Fig. 5A, B; Supplementary Fig. 5A, B). Moreover, ESRRA-ablated 389 tumors exhibited unperturbed MYC, integrin and CAV1 expression, as well as unchanged 390 Src and Cofilin phosphorylation upon Pgc1α expression (Fig. 5C, D; Supplementary Fig.  391 5C). All these data are in line with the association of ERRα to MYC promoter in Pgc1α-392 expressing PC3 cells (Supplementary Fig. 5D). 393 Since we have observed a robust inverse correlation between PGC1A and MYC 394 expression in various PCa datasets, we asked whether the dependency on ERRα could be 395 recapitulated in this setting. To this end, we carried out two independent approaches in 396 PCa patient datasets. On the one hand, we inferred ERRα canonical activity based on the 397 equilibrium between its main co-activators (PGC1A) and co-repressors (NRIP1). We 398 calculated the ratio of abundance of PGC1A and NRIP1 transcript (PGC1A/NRIP1), which 399 provided an estimation of ERRα canonical activity towards its targets, as confirmed 400 through the analysis of ACACB and LAMB2 expression (Supplementary Fig. 6A) with our mechanistic analysis, ERRα activity, but nor ERRα itself was consistently and 402 inversely correlated with MYC in various PCa datasets (Supplementary Fig. 6B, C). On 403 the other hand, we took advantage from a prognostic PGC1α-ERRα signature that we 404 generated previously (17). This signature was composed of 10 genes that were i) 405 regulated by PGC1α in vitro, ii) predicted to be ERRα targets and iii) correlated with 406 PGC1A in PCa datasets. In full support of our data, this PGC1α-ERRα activity signature 407 was inversely correlated with MYC expression in various PCa patient datasets ( Fig. 5E; 408 Supplementary Fig. 6D). Metabolic deregulation is a hallmark of cancer (2), and encompasses a variety of 413 biochemical routes, which must be coordinated in order to result in a phenotypic change. 414 We postulated in the past that this strict requirement for coordination could unveil novel 415 cancer genes. By focusing on transcriptional co-regulators that control the expression of 416 an ample set of metabolic genes, we discovered the predominant perturbation of PGC1α 417 in PCa (7,17). This metabolic regulator orchestrates the activation of catabolic and 418 antioxidant pathways, at the expense of anabolism (8). Interestingly, the contribution of 419 PGC1α to cancer biology is complex. Elegant studies have reported a role of this co-420 regulator: i) promoting aggressiveness of breast, pancreatic, gastric tumors, 421 cholangiocarcinoma, glioma and melanoma (10-14,38-40), and ii) suppressing cancer 422 aggressiveness in prostate, kidney tumors and melanoma (9,15-18). Moreover, the 423 expression of this co-regulator is associated to the efficacy of anticancer therapies 424 (10,11,14,15,41,42). 425 PGC1α exhibits a tumor type-dependent activity, ranging from tumor suppressor to 426 advantageous for cancer cells (7). This co-activator is required for the activity of pancreatic 427 cancer stem cells (13) and for the survival of breast cancer cells in circulation (12). In 428 melanoma, the metabolic activity of PGC1α promotes cell proliferation, whereas the non-429 metabolic function opposes metastatic dissemination (10,11,16).This study together with 430

reports by us and others demonstrate that PGC1α suppresses proliferation and invasion in 431
PCa through presumably distinct molecular pathways emanating from the regulation of 432 ERRα, consistent with its tumor and metastasis suppressive function (17,18) (Fig. 6). Our 433 results mirror the anti-invasive activity of the co-regulator in melanoma, whereas 434 proliferation is regulated in opposite sense in both tumor types. This apparent discrepancy 435 on March 7, 2020. © 2019 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Author Manuscript Published OnlineFirst on October 8, 2019; DOI: 10.1158/0008-5472. CAN-19-1231 could be associated to the tissue-specific molecular cues that drive these tumors or the 436 distinct nutrient and metabolic pathways that sustain their growth. 437 Cancer cell proliferation imposes tremendous pressure to meet the bioenergetics demands 438 and to generate sufficient biomolecules to build new cells. We now possess a more limited. An exciting possibility stems from the notion that factors that control metabolic 447 programs would also regulate molecular cues associated to cancer cell dissemination. 448 Little is known about the activities of PGC1α in cancer beyond proliferation. This co-449 regulator inhibits dissemination in melanoma through the regulation of ID2-TCF4-Integrins 450 (16). In gastric cancer, a recent report suggests that PGC1α upregulation supports 451 metastasis though the regulation of SNAI1 (38). Interestingly, none of these effects are 452 ascribed to the regulation of its main transcriptional partner, ERRα. Instead, we 453 demonstrate that the PGC1α-ERRα transcriptional axis in PCa accounts for the invasive 454 phenotype. We demonstrate that PGC1α/ERRα status influences signaling pathways that 455 are important for the regulation of cytoskeletal remodeling. In turn, changes in pathways 456 related to integrin and ROCK signaling provide a feasible explanation for the anti-invasive 457 effects of the co-regulator. Interestingly, the set of genes inhibited in PGC1α-expressing 458 cells that relate to cytoskeletal remodeling are enriched in MYC promoter binding sites. 459 This data is consistent with the notion that PGC1α/ERRα represses MYC expression and 460 that silencing of this transcription factor partly phenocopies the effect of PGC1α (18). 461 Similar to PGC1α, ERRα has opposing effects in different tumor types (7). Interestingly, 462 we show that this nuclear receptor is required for the tumor suppressive activity of PGC1α, 463 whereas its deletion delays tumor onset in immunocompromised mice independently of the 464 induction of PGC1α. Our results could be explained by the differential requirement of basal 465 In summary, our study together with recent reports (18)    Whitney U test (G). p, p-value. *p < 0.05, **p < 0.01, ***p < 0.001. 644        Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) . http://cancerres.aacrjournals.org/content/early/2019/10/08/0008-5472. CAN-19-1231 To request permission to re-use all or part of this article, use this link on March 7, 2020. © 2019 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.