MED19 alters AR occupancy and gene expression in prostate cancer cells, driving MAOA expression and growth under low androgen

Androgen deprivation therapy (ADT) is a mainstay of prostate cancer treatment, given the dependence of prostate cells on androgen and the androgen receptor (AR). However, tumors become ADT-resistant, and there is a need to understand the mechanism. One possible mechanism is the upregulation of AR co-regulators, although only a handful have been definitively linked to disease. We previously identified the Mediator subunit MED19 as an AR co-regulator, and reported that MED19 depletion inhibits AR transcriptional activity and growth of androgen-insensitive LNCaP-abl cells. Therefore, we proposed that MED19 upregulation would promote AR activity and drive androgen-independent growth. Here, we show that stable overexpression of MED19 in androgen-dependent LNCaP cells promotes growth under conditions of androgen deprivation. To delineate the mechanism, we determined the MED19 and AR transcriptomes and cistromes in control and MED19 LNCaP cells. We also examined H3K27 acetylation genome-wide. MED19 overexpression selectively alters AR occupancy, H3K27 acetylation, and gene expression. Under conditions of androgen deprivation, genes regulated by MED19 and genomic sites occupied by MED19 and AR are enriched for ELK1, a transcription factor that binds the AR N-terminus to promote select AR-target gene expression. Strikingly, MED19 upregulates expression of monoamine oxidase A (MAOA), a factor that promotes prostate cancer growth. MAOA depletion reduces androgen-independent growth. MED19 and AR occupy the MAOA promoter, with MED19 overexpression enhancing AR occupancy and H3K27 acetylation. Furthermore, MED19 overexpression increases ELK1 occupancy at the MAOA promoter, and ELK1 depletion reduces MAOA expression and androgen-independent growth. This suggests that MED19 cooperates with ELK1 to regulate AR occupancy and H3K27 acetylation at MAOA, upregulating its expression and driving androgen independence in prostate cancer cells. This study provides important insight into the mechanisms of prostate cancer cell growth under low androgen, and underscores the importance of the MED19-MAOA axis in this process. Author summary Prostate cancer is one of the most common cancers worldwide, and androgen hormones are essential for prostate cancer growth. Androgens exert their effects through a protein called the androgen receptor (AR), which turns on and off genes that regulate prostate cancer growth. Powerful drugs that block AR action by lowering androgen levels – so-called androgen deprivation therapy - are used to treat prostate cancer patients, and these yield initial success in reducing tumor growth. However, over time, tumors circumvent androgen deprivation therapy and patients relapse; in many cases, this occurs because AR becomes re-activated. The factors responsible for re-activating AR and promoting growth under androgen deprivation are not well understood. Here, we demonstrate that a subunit of the Mediator transcriptional regulatory complex, called MED19, promotes growth of prostate cancer cells under low androgen conditions, mimicking the ability of tumors to grow under androgen deprivation in prostate cancer patients. MED19 promotes androgen-independent growth by working with a transcription factor that interacts with AR, called ELK1, to induce the expression of genes regulated by AR that promote prostate cancer growth. This study provides important insight into how prostate cancer cells can maintain growth under androgen deprivation through MED19.

141 overexpression compared to its control counterpart (Fig 2C and S2C Fig). This was 142 recapitulated in a xenograft model, where the MED19-overexpressing cells produced larger 143 tumors than controls (Fig 2D). This corroborates the growth advantage of MED19 found in 144 LNCaP cells. Together, this indicates a role for MED19 in conversion of early stage cells to 145 aggressive growth and androgen independence.
146 MED19 depends on AR activity for its growth advantage but does not alter AR expression 147 As AR amplification is a common mechanism to achieve androgen independence, we 148 evaluated the mRNA and protein levels of AR with MED19 overexpression in LNCaP cells. We

153
We then examined the reliance of MED19 LNCaP cells on AR for their androgen-154 independent growth by determining their sensitivity to enzalutamide, an AR antagonist that 155 reduces AR transcriptional activity in part by preventing AR nuclear accumulation [33].
156 Enzalutamide inhibited the proliferation of MED19 LNCaP cells both in the presence and 157 absence of androgens, indicating that the growth advantage conferred by MED19 requires AR 158 transcriptional activity (Fig 3C and 3D). We confirmed these results by siRNA depletion of AR, 159 which also reduced androgen-dependent and androgen-independent growth of MED19 LNCaP 172 Global AR occupancy under androgen deprivation, as measured by total number of and 173 individual level of occupancy at AR sites in published ChIP-seq studies, is low compared to 174 androgen treatment. We speculated that MED19 may alter AR activity under androgen 175 deprivation by modulating AR at low occupancy sites. Therefore, we included in our ChIP-seq 176 study all AR-specific sites, including those with low occupancy in androgen-deprived conditions.
177 We used rigorous quality controls to maximize capture of AR sites and ensure strict specificity to 178 AR occupancy (see detailed ChIP-seq section under Materials and Methods) (S6C Fig and S7 179 Table).

180
Under androgen deprivation, there was a striking and very selective change in gene 181 expression profile with MED19 overexpression, with a total of 151 genes altered (76 genes 182 upregulated and 75 genes downregulated, fold change≥1.5 and p-adj≤0.05) (Fig 4A and S1 183 Table). This was accompanied by a selective shift in the AR cistrome (~12% of total AR sites 184 are occupied only in control LNCaP cells or only in MED19 LNCaP cells), without a global 185 change in the total number of sites occupied by AR ( Fig 4A). As expected, with androgen 186 treatment, the total number of AR sites increased (Figs 5A and 5B). There was a selective shift 187 in the AR cistrome when MED19 is overexpressed in the presence of androgens as well (Fig   188 4B). There was also a shift in gene expression: 309 genes were differentially expressed with 189 MED19 overexpression in the presence of androgens (78 genes upregulated and 231 190 downregulated, fold change≥1.5 and p-adj≤0.05) (Fig 4B and S2 Table).

191
Of these 309 genes, 82 were also differentially expressed in the absence of androgens 192 (~50% of total genes differentially expressed in the absence of androgens) (S1 and S2 Tables).
193 This holds true for MED19 occupancy as well: the total number of MED19 sites increased with 194 androgen treatment, with ~50% of the sites occupied in the absence of androgens also occupied 195 in the presence of androgens ( Fig 5B). This indicates that there is partial overlap in MED19 196 regulation of gene expression and AR activity in the absence and presence of androgens, 197 consistent with the differential growth advantage in the absence and presence of androgens when 198 MED19 is overexpressed.

199
MED19 occupancy in the absence and presence of androgens corresponds almost entirely 200 with AR occupancy, with virtually every gene differentially expressed in MED19 LNCaP cells 201 occupied by AR, and many (the majority in the absence of androgens) also occupied by MED19, 202 indicating direct regulation by MED19 (Fig 4A and 4B, S6 Table). In fact, most of the MED19-203 regulated genes are androgen-responsive, and many have been reported as AR target genes (S1 204 and S2 Tables). AR was also the top predicted regulatory transcription factor candidate using 205 Chromatin immunoprecipitation Enrichment Analysis (ChEA) (Fig 4A and 4B, S1 and S2 206 Tables). This confirms that MED19 regulation of gene expression is driven by AR.

207
In response to androgen treatment, ~4500 genes in control LNCaP cells and ~5000 genes 208 in MED19 LNCaP cells were differentially expressed (≥1.5-fold, p-adj≤0.05), and, as expected, 209 AR was the top transcription factor from ChEA analysis for both ( Fig 5A and Table).

216
However, MED19 appears mainly to modulate the response of canonically androgen-  Table). However, the overall response to androgens does not appear to markedly 221 differ with MED19 overexpression, and differential gene expression with MED19

227
Although AR occupies unique sites in MED19 LNCaP cells, the majority of genes altered 228 with MED19 overexpression contain AR sites shared by control LNCaP cells and MED19 229 LNCaP cells (S6 Table). What we observed at a number of these sites was a change in the level 230 of AR occupancy and/or H3K27 acetylation beyond a "present/absent" or "on/off" binary. These 231 subtle changes in gene occupancy corresponded with MED19 activation or repression of genes 232 from the RNA-seq study, indicating that MED19 alters gene expression through small shifts in 233 AR occupancy and H3K27 acetylation. We also observed that the changes in AR occupancy and 234 H3K27 acetylation, like the changes in gene expression with MED19 overexpression, did not 235 simply mimic the changes that occurred with androgen treatment.

236
For example, LRRTM3 (Leucine Rich Repeat Transmembrane Neuronal 3) is one of the 237 most upregulated genes upon MED19 overexpression under androgen deprivation, while 238 androgen treatment suppresses LRRTM3 expression (S1-S4 Tables, Fig 6A). With MED19 239 overexpression, there is a clear increase in AR occupancy and H3K27 acetylation at several 240 regulatory intronic sites at LRRTM3, with MED19 occupancy at one of these sites ( Fig 6B, S6 241 Table). Conversely, androgen treatment reduces H3K27 acetylation and alters AR occupancy  Table). Thus, it appears that AR occupancy at specific targets is altered by MED19 249 overexpression. This may be responsible for the changes in gene expression and attendant 250 effects on cell proliferation.

251
We wanted to determine the specific gene targets altered by MED19 overexpression that 252 were responsible for promoting androgen-independent growth. We decided to focus on genes 253 upregulated by MED19 overexpression under androgen deprivation, which could be depleted to 254 inhibit androgen-independent growth. Given that LRRTM3 appears to be a direct target of 255 MED19, with changes in AR occupancy that correlated with a large upregulation in expression, 279 MAOA depletion reduced growth by ~50% ( Fig 7D). Interestingly, from the RNA-seq study,    328 rather, these were associated with SRF, which was not enriched in the ChIP-seq data (S12C Fig). 329 Given the strong connection to ELK1, we tested its functional role in gene expression and 332 ELK1 knockdown both greatly reduced expression of MAOA and inhibited androgen-333 independent growth (Fig 8C-8E).

334
To determine if ELK1 occupied MAOA at sites where AR and MED19 were present, we 335 performed ChIP-qPCR for ELK1 adjacent to the MAOA promoter, overlapping with the reported 336 ARE, where AR, MED19, and H3K27 acetylation were present (Fig 8F, S14 Fig). Indeed, we 337 found that ELK1 also occupied this region, with a trend toward increased ELK1 occupancy 338 under androgen deprivation when MED19 is overexpressed (Fig 8F). This indicates that MED19 339 and ELK1 cooperate to further AR occupancy and H3K27 acetylation, increase MAOA 340 expression, and promote androgen-independent growth.

342
Discussion 343 We have demonstrated that overexpression of MED19 in androgen-dependent LNCaP 344 cells provides a growth advantage in the absence and presence of androgens. This is mediated 345 by AR. The cells remain dependent on AR for growth under androgen deprivation, without 346 increasing full-length AR abundance or splice variant AR-V7 expression. Therefore, increased 347 expression of MED19 is sufficient to convert a cell that is androgen-dependent to one that is 348 androgen-independent for growth.

397
Another Mediator subunit, MED1 (also a middle subunit), has been described to promote