Transcriptional perturbation of protein arginine methyltransferase-5 exhibits MTAP-selective oncosuppression

We hypothesized that small molecule transcriptional perturbation could be harnessed to target a cellular dependency involving protein arginine methyltransferase 5 (PRMT5) in the context of methylthioadenosine phosphorylase (MTAP) deletion, seen frequently in malignant pleural mesothelioma (MPM). Here we show, that MTAP deletion is negatively prognostic in MPM. In vitro, the off-patent antibiotic Quinacrine efficiently suppressed PRMT5 transcription, causing chromatin remodelling with reduced global histone H4 symmetrical demethylation. Quinacrine phenocopied PRMT5 RNA interference and small molecule PRMT5 inhibition, reducing clonogenicity in an MTAP-dependent manner. This activity required a functional PRMT5 methyltransferase as MTAP negative cells were rescued by exogenous wild type PRMT5, but not a PRMT5E444Q methyltransferase-dead mutant. We identified c-jun as an essential PRMT5 transcription factor and a probable target for Quinacrine. Our results therefore suggest that small molecule-based transcriptional perturbation of PRMT5 can leverage a mutation-selective vulnerability, that is therapeutically tractable, and has relevance to 9p21 deleted cancers including MPM.

PRMT5 silencing mediates growth arrest in MTAP negative mesothelioma. To determine whether MTAP negative MPM cells were dependent on PRMT5, we silenced PRMT5 expression by RNA interference in both MTAP wild-type and negative MPM cell lines (Fig. 1C). Reduced clonogenic growth was selective for MTAP negative cell lines, with concurrent reduction in symmetrical di-methylation of Histone H4 arginine 3 (H4R3me2S). Silencing of the PRMT5 interactor WDR77 phenocopied PRMT5 silencing in MTAP negative cells, leading to a reduced clonogenic activity and a reduced H4R3me2S (Fig. 1D). This effect was not phenocopied by siRNA targeting the PRMT5 interactor RIO1 kinase ( Supplementary Fig. 1A).
To investigate the kinetics of growth arrest, we conducted real time analysis following PRMT5 RNAi which showed a growth arrest after 120 h ( Fig. 2A). Neither apoptosis nor cell cycle perturbation were observed by flow cytometry after PRMT5 silencing ( Supplementary Fig. 1B).
To determine whether or not PRMT5 silencing-induced chromatin remodelling would upregulate tumour suppressor pathways, we conducted gene expression, examining canonical pathways linked to upregulated genes. Overexpression of tumour suppressors, such as EIF3F, FOXP4, ZBTB4, GANAB, TMEM141 was observed. Gene set enrichment analysis revealed a significant enrichment of the EZH2 target gene (Fig. 2B,C, Supplementary Table 2).
In common with PRMT5 siRNA, the small molecule EPZ015666 reduced clonogenic growth (Supplementary Fig. 1C). Neither apoptosis nor cell cycle perturbation were observed by flow cytometry after inhibition by EPZ015666 ( Supplementary Fig. 1D).
Quinacrine hydrochloride is a PRMT5 perturbagen. To fully harness the potential for MTAP selective activity via PRMT5 inhibition, we utilised the connectivity map 8 to identify novel transcriptional suppressors of PRMT5 (Supplementary Table 3, Fig. 3A). Among the top 5 predicted molecules, Quinacrine Hydrochloride led to significant suppression of both PRMT5 mRNA levels ( Fig. 3B) and protein with consequent reduction of H4R3me2S and growth arrest in MTAP negative but not MTAP positive cells (Fig. 3C). Quinacrine Hydrochloride directly inhibited PRMT5 promoter activity as confirmed by luciferase reporter assay (Fig. 3D), but did not have any direct effect on PRMT5 enzymatic activity (Fig. 3E).
Neither apoptosis nor cell cycle perturbation were observed after treatment with Quinacrine by flow cytometry, in common with both PRMT RNA interference or small molecule inhibitor (Supplementary Fig. 2A). We used a functional genetic approach to establish the extent to which the phenotypic effects of Quinacrine in MTAP negative cells were reliant upon PRMT5. Overexpression of wild-type PRMT5 rescued cells from treatment with Quinacrine, however this was not observed with transfection of the methyltransferase dead mutant PRMT5 E444Q compared to empty vector control. This is consistent with modulation of endogenous PRMT5 transcription as an essential mechanism underpinning the effect of Quinacrine in MTAP negative context (Fig. 3F).

Quinacrine transcriptionally regulates PRMT5 via c-JUN.
To further explore the possible mechanism of PRMT5 transcriptional perturbation by Quinacrine, we used RNA interference to screen for putative transcription factors (TFs) implicated by PROMO 9 . These TFs included CEBP1, c-JUN and NF-YA were predicted to bind to the PRMT5 promoter. RNAi mediated silencing of c-JUN, but not of CEBP1 or NF-YA, resulted in a significant reduction of PRMT5 mRNA levels in MTAP negative cells and this was comparable to that achieved with Quinacrine (Fig. 4A). c-JUN silencing led to reduced clonogenic growth with concurrent downregulation of both PRMT5 and H4R3me2S in MTAP negative cells (Fig. 4B), but no apoptosis or cell cycle perturbation ( Supplementary Fig. 2B). Quinacrine suppressed c-JUN mRNA suggesting that it targets PRMT5 transcription indirectly via this transcription factor (Fig. 4C). The effects of c-JUN silencing were MTAP-selective, as MTAP positive cells failed to reduce PRMT5 mRNA (Fig. 4D), without evidence of impaired clonogenic growth or reduced H4R3me2S (Fig. 4E). Interestingly, Quinacrine did not suppress c-JUN mRNA in MTAP positive cells (Fig. 4F). www.nature.com/scientificreports/

Discussion
Copy number loss of MTAP is one of the most frequent events in MPM. Our data confirmed a marked negative prognostic effect, which warrants novel targeted therapy. It should be noted that given the co-deletion of MTAP with CDKN2A, which has been also shown to be negatively prognostic, it was not possible to deconvolute the impact of MTAP in isolation from our data involving chromosome 9p21.3 deleted cel lines. PRMT5 plays a key role in the regulation of several pathways including DNA damage response, apoptosis, inhibition of tumour suppressors, and activation of survival pathways 10 and has been reported to be a dependency in MTAP negative cells 11 , which we have verified in MPM. We confirmed global epigenetic modification associated with reduced H4R3me2S and re-expression of tumour suppressors, such as EIF3F, FOXP4, ZBTB4, GANAB, TMEM141, in association with loss of clonogenicity. SAM competitive PRMT5 inhibitors, such as EPZ015666 and GSK3326595, have shown limitations in recapitulating the vulnerability of MTAP negative cells in response to PRMT5 inhibition 10 . To address this limitation, we used the connectivity map approach 8 to screen for small molecules with PRMT5 downregulating activity and identified the off-patent small molecule Quinacrine, which was primarily used as an antimalarial drug as well as an intrapleural sclerosing treatment for malignant pleural effusions with an excellent safety profile 12 . Quinacrinemediated growth arrest was PRMT5 dependent as confirmed by a methyltransferase dead PRMT5 mutant.
Apoptosis was not observed as a mechanism of reduced clonogenicity; the phenotype was cytostatic, but not restricted to any phase of the cell cycle. Recent studies suggest that small molecule PRMT5 inhibition, either directly or indirectly through inhibition of cyclin dependent kinases 4 and 6, alters RNA splicing, leading to an MDM4 dependent activation of p53/p21 13,14 . Whether this axis is exploited by Quinacrine requires further exploration.
Quinacrine has been reported to regulate c-jun phosphorylation at positions 349 to 340 and 266 to 257 15 . We showed that c-jun is essential in driving PRMT5 expression and mediating Quinacrine-induced apoptosis. This study therefore implicates c-jun or SAPK pathway regulation as a mechanism underlying PRMT5 transcriptional repression by Quinacrine.
Based on early pharmacokinetic studies conducted with Quinacrine in the 1940′s, plasma concentration was measured as 225 nM at a dose of 100 mg, 3 times a day. However, hepatic and leucocyte distributions are high, with high plasma binding. This agent can however be safely instilled into the pleural cavity at a much higher dose of 5 mM, at which dose it acts as a sclerosant. This implies that lower micromole doses are achievable and could be locally administered, achieving localised exposure to mesothelioma at concentrations capable of suppressing PRMT5 expression.
Our results provide a proof of concept to support the use of small molecule transcriptional perturbation to leverage a somatic-mutation based vulnerability, suggesting a repurposing potential that warrants further study.  Supplementary Table I. Overall survival was calculate from the date of surgery. A separate cohort of 100 patients 16 was used for validation and clinico-pathological characteristics are described in Supplementary Table I. Informed consent to provide research samples was obtained from all patients. All methods were carried out in accordance with local guidelines and regulations.

Material and methods
Oncoscan analysis. DNA was extracted with the GeneRead DNA FFPE kit (Qiagen, Manchester, UK). 80 ng of gDNA were analysed using the OncoScan FFPE Assay Kit (Affymetrix, Wooburn Green High Wycombe, UK), which utilizes molecular inversion probe (MIPs) technology 17 . The BioDiscovery Nexus Express 10.0 for OncoScan software was then used to define copy number alterations and loss of heterozygosity as previously described 18 . Clonogenic assays. 5000 cells per well were seeded in 12 well plates and left untreated or treated with Quinacrine (500 nM, 1 µM, 2 µM). Cells were fixed between days five and seven (once enough colonies had formed in the control) on ice in methanol for 10 min. Cells were then stained with crystal violet (Sigma, Gillingham, UK) for 20 min. Colonies were dissolved in 30% acetic acid to allow quantification 19  Real time proliferation assay. The xCELLigence RTCA DP instrument (Acea Bioscience, San Diego, CA) was used as described in the manufacturer´s instruction manual. Cells (5,000 cells/ well) were seeded in E16-Plates. The cell indices were measured every 15 min for 120 h. Each treatment condition was measured in triplicate.

Connectivity mapping.
A PRMT5-centred gene signature was created from co-expression analyses of 9 independent Gene Expression Omnibus (GEO) datasets GSE37745, GSE50081, GSE28571, GSE77803, GSE43580, GSE19804, GSE18842, GSE10245, and GSE19188. Within each dataset, PRMT5 was used as the seed gene, with which the gene expression correlation coefficients for other genes (probes) were calculated. All genes were then ranked based on the magnitude and statistical significance of their correlations with the seed, following the ranking method described in 18 . The genes' ranks were then combined across these datasets to obtain an overall rank for each gene to determine its inclusion to the PRMT5 gene signature for subsequent connectivity mapping analysis. A gene signature progression approach 19 determined that an 8-gene signature was the optimal length including PRMT5 and its 7 strongest co-expression correlates PSMB5, HNRNPC, APEX1, HNRNPC, IPO4, TOX4, and TUBB. This 8-gene signature was used as an input to query a collection of 83,939 reference drug gene expression profiles 20 covering 1353 FDA approved drugs (http:// www. lincs cloud. org). This connectivity mapping analysis was conducted in the framework of sscMap 21,22 . In our analysis all the individual reference profiles with the same drug formed a reference set. A set score was then calculated between the gene signature and each reference set, and the associated p-value was estimated by generating a large number of random gene signatures of the same length. Any signature-drug connections with a p-value no greater than a pre-set threshold (1/1353 = 7.4e−4) were declared as statistically significant. Additionally, gene signature perturbation analysis 23 was performed to obtain the robustness (perturbation stability) of the significant signature-drug connections.
Only the significant drugs that had 100% perturbation stability were selected for further consideration. Finally, significant drugs were ranked by the absolute value of their connection z-score to the PRMT5 gene signature.
PRMT5 enzymatic activity. PRMT5 chemoluminescent assay was purchased from AMS Biotechnology (Europe) Ltd (Abingdon, UK). Quinacrine or EPZ015666 was added to the plate pre-coated with histone H4 peptide substrate. PRMT5 enzymatic activity was measured after reaction with the antibody against methylated arginine3 residue of Histone H4, the secondary HRP-labeled antibody, S-adenosylmethionine, methyltransferase assay buffer, and purified PRMT5 enzyme, according to the manufacturer instructions.
Protein extraction and immunoblotting. Seventy-two hours after treatment cells were lysed in RIPA buffer containing protease inhibitors (Roche, Burgess Hill, UK). Lysates were clarified by centrifugation at 4 °C at 13,000 rpm for 10 min. 40 µg of total cell lysates were loaded on SDS-PAGE gels. Signal detection was performed with ECL-plus chemiluminescent system (GE Healthcare, Little Chalfont, UK).