Androgen deprivation upregulates SPINK1 expression and potentiates cellular plasticity in prostate cancer

The Serine Peptidase Inhibitor, Kazal type 1 (SPINK1) overexpression represents ~10-25% of the prostate cancer (PCa) cases associated with shorter recurrence-free survival and poor prognosis. Nonetheless, androgen-deprivation therapy (ADT) remains the mainstay treatment for locally advanced and metastatic PCa patients. However, majority of these individuals eventually progress to castration-resistant stage, and a subset of these patients develop ADT-induced neuroendocrine PCa. Despite adverse effects of ADT, possible role of androgen signaling in SPINK1-mediated prostate oncogenesis remains unexplored. Here, we show that androgen receptor (AR) and its corepressor, the RE1-silencing transcription factor (REST), occupy SPINK1 promoter and functions as a direct transcriptional repressor of SPINK1, thus blocking AR signaling via ADT relieves its repression, leading to SPINK1 upregulation. In agreement, an inverse association between SPINK1 levels and AR expression was observed across multiple PCa cohorts, and in neuroendocrine differentiated cells. While, lineage reprogramming factor SOX2 in turn binds to SPINK1 promoter leading to its transactivation in androgen-deprived conditions with concomitant increase in neuroendocrine markers. Additionally, we also confirm the role of SPINK1 in epithelial-mesenchymal transition, drug resistance, stemness and cellular plasticity. Moreover, we show that Casein Kinase 1 inhibitor stabilizes the REST levels, which in cooperation with AR, conjures transcriptional repression of SPINK1 expression, and impedes SPINK1-mediated oncogenesis. Collectively, our findings provide a plausible explanation to the paradoxical clinical outcomes of ADT, possibly due to increased SPINK1 levels. This study highlights the need to take a well-informed decision prior to ADT and develop alternative therapeutic strategies for castrate-resistant PCa patients.


Introduction
Genetic rearrangement involving androgen-driven promoter region of the serine protease gene, TMPRSS2 and the coding region of ERG, a member of ETS (E26 transformationspecific) transcription factor family occurs in almost half of the prostate cancer (PCa) cases (Tomlins et al, 2005). Subsequently, with the technological advances in genomics, numerous other molecular subtypes such as, fusion involving other members of ETS family, (ETV1, ETV4, FLI1 and NDRG1); RAF kinase rearrangements; SPOP/CHD1 alterations; mutations in FOXA1 and IDH1 have also been discovered (Abeshouse et al, 2015;Tomlins et al, 2007;Tomlins et al, 2006). While TMPRSS2-ERG fusion is the most common subtype, overexpression of SPINK1 constitutes a substantial ~10-25% of the total PCa cases exclusively in ETS-fusion negative subtype (Huang et al, 2016;Tomlins et al, 2008).
However, several independent studies show expression of SPINK1 and ERG in two distinct foci within a prostate gland, indicating that these two events are either independent or SPINK1 overexpression to be a sub-clonal event after TMPRSS2-ERG genetic rearrangement (Brooks et al, 2015;Flavin et al, 2014;Huang et al, 2016). Notably, SPINK1-positive patients show rapid progression to biochemical recurrence as compared to ETS-fusion positive (Leinonen et al, 2010;Tomlins et al, 2008). In a Finnish PCa cohort, SPINK1-positive patients exhibit an association with an early progression to castration resistance (Leinonen et al, 2010). Further, intermediate or high risk localized PCa patients who endured radical prostatectomy show positive correlation between SPINK1 and biochemical recurrence subsequently leading to disease-specific mortality (Johnson et al, 2016).
Androgen receptor (AR) signaling axis plays critical role in the PCa pathogenesis and progression. Hence, androgen deprivation therapy (ADT) remains the basis for PCa treatment, however the disease often relapses to an androgen-independent advanced stage known as castrate-resistant prostate cancer (CRPC), often associated with poor patient 4 prognosis (Karantanos et al, 2013;Lonergan & Tindall, 2011;Sun et al, 2012b). Numerous mechanisms that restore androgen signaling in CRPC individuals have been proposed, such as mutations in the AR ligand binding domain (F877L and T878A), constitutively active variants of AR (AR-V7 and ARv567es), AR amplification, or steroid-inducible glucocorticoid receptor that activates AR target genes (Antonarakis et al, 2014;Arora et al, 2013;Chen et al, 2004;Taplin et al, 1995), suggesting that these alteration in AR or ARsignaling ensue as an adaptive response to ADT. Current treatment regimen for CRPC patients include FDA-approved second generation anti-androgens such as enzalutamide or MDV3100 (blocks the nuclear translocation of AR or binding to its genomic sites) and abiraterone acetate (irreversible steroidal CYP17A1 inhibitor, targets adrenal or intratumoral androgen biosynthesis) (Attard et al, 2005;de Bono et al, 2011;Scher et al, 2010;Tran et al, 2009). Although, these anti-androgens are known to prolong the overall survival of PCa patients, but the response is temporary, and the disease eventually progresses. A subset of CRPC patients (~20% of advanced drug-resistant cases) elude selective pressure of ADT by minimizing its dependency on the AR signaling are identified as neuroendocrine (NE) PCa, often associated with poor prognosis and patient outcome (Rickman et al, 2017). NEPC exhibits a distinct phenotype characterized by reduced or no expression of AR and ARregulated genes, and increased expression of NEPC markers such as Synaptophysin (SYP), Chromogranin A (CHGA), and Enolase 2 (ENO2) (Beltran et al, 2014). Several molecular mechanisms have been ascribed to the transdifferentiation of CRPC to NEPC, for instance, frequent genetic alterations involving the TP53 and the retinoblastoma-1-encoding gene RB1 are associated with poorly differentiated NEPC tumors (Beltran et al, 2016;Tan et al, 2014).
Moreover, N-Myc amplification, BRN2 upregulation, mitotic deregulation via Aurora kinase A (AURKA), alternative splicing by serine/arginine repetitive matrix4 (SRRM4), and loss of repressor element-1 (RE-1) silencing transcription factor (REST) expression 5 are known to have a role in NE transdifferentiation (Beltran et al, 2011;Bishop et al, 2017;Dardenne et al, 2016;Li et al, 2017;Svensson et al, 2014). Mounting evidence suggests that REST expression is downregulated in relapsed PCa (Svensson et al, 2014), which has also been attributed to NE differentiation of PCa cells (Lin et al, 2016;Svensson et al, 2014;Zhu et al, 2014). Interestingly, REST expression is positively regulated by androgen signaling, and it serves as a transcriptional co-regulator of AR (Li et al, 2017;Svensson et al, 2014).
Although, SPINK1 overexpression has been associated with rapid progression to biochemical recurrence and aggressive stage of PCa, nonetheless the regulatory mechanism involved in SPINK1 upregulation and its functional significance in the advancement of disease is largely undefined. In this study we have explored the mechanism involved in SPINK1 overexpression and its functional implication in PCa tumorigenesis. Importantly, we show that AR antagonists unexpectedly have a positive effect on the transcriptional regulation of SPINK1, and consecutive increase in SPINK1 level, further positively associates with NE-phenotype. Our findings demonstrate that SPINK1 is transcriptionally suppressed by AR and its co-repressor REST, a negative master regulator of neurogenesis, and suggests a possible role of SPINK1 in NEPC transdifferentiation. Collectively, our findings alert against the widespread application of potent AR antagonist used for PCa treatment, and highlight expected clinical complication associated with increased expression of SPINK1 as well as therapy-induced NEPC.

Expression of SPINK1 and Androgen Receptor is inversely correlated in prostate
cancer patients 6 Altered AR signaling and AR-binding profile has been extensively studied in localized PCa and CRPC (Sharma et al, 2013). It has been known that AR binds with other cofactors, such as GATA2, octamer transcription factor 1 (Oct1), Forkhead box A1 (FoxA1) and nuclear factor 1 (NF-1) to mediate cooperative transcriptional activity of the target genes (Wang et al, 2007). Thus, we sought to discover the possible link between SPINK1 and AR expression in PCa patients, and stratified patients available at TCGA-PRAD (The Cancer Genome Atlas Prostate Adenocarcinoma) cohort based on high and low expression of AR. The patients with higher expression of AR showed a significantly lower expression of SPINK1 and contrariwise ( Fig 1A). To further confirm this association, we performed immunohistochemical (IHC) analysis for the expression of SPINK1 and AR on tissue microarrays (TMA) comprising of PCa patient specimens (n=237). Important to note that all of these cases underwent radical prostatectomy without any hormone or radiation therapy. In concordance with TCGA data analysis, our IHC findings reveal that SPINK1positive patients exhibit low or negative staining for AR expression, while SPINK1-negative patients show higher or medium AR staining ( Fig 1B and Supplementary Fig S1A).
Importantly, about ~67% of the SPINK1+ patients (34 out of 51) demonstrate either low or negative staining for AR expression (Fisher's exact test, P<0.0004) (Fig 1C and D). Based on our findings, we conjecture that SPINK1 is one of the AR repressed genes, hence we next examined the expression of other members of AR repressor complex (NCOR1, NCOR2 and NRIP1) including AR using TCGA-PRAD cohort, and the patients were sorted based on SPINK1 high and low expression by employing quartile-based normalization (Dillies et al, 2013). Interestingly, we found that SPINK1 expression is also negatively associated with other AR repressive complex members (Supplementary Fig S1B). Additionally, we investigated the correlation of SPINK1 and AR signaling score using transcriptomic data from two independent PCa cohorts, Memorial Sloan Kettering Cancer Center (MSKCC) and 7 TCGA-PRAD. As expected, lower AR signaling score in SPINK1-positive patients was recorded as compared to the SPINK1-negative patients ( Supplementary Fig S1C). Taken together, our findings show an inverse association between SPINK1 expression and AR signaling in PCa patients, indicating that overexpression of SPINK1 in SPINK1-postive subtype is owing to the loss of AR-mediated repression during prostate cancer progression.

AR antagonists trigger SPINK1 upregulation by relieving AR-mediated repression
Since, an inverse association between SPINK1 expression and AR signaling was observed in three independent PCa cohorts (TCGA-PRAD, MSKCC) including ours (Fig 1), thus we examined the role of active AR signaling in the regulation of SPINK1 using PCa cell lines, 22RV1 (endogenously SPINK1-positive) and androgen responsive VCaP cells  Fig S2B). Similarly, VCaP cells stimulated with R1881 (10nM) show a significant decline in the expression of SPINK1 both at transcript and protein levels, while an increase in the expression of AR target genes (KLK3, TMPRSS2 and FKBP5) was noticed   Supplementary Fig S2C). Further, we also analyzed the publicly available datasets (GSE71797 and GSE51872), wherein 22RV1 and VCaP cells stimulated with R1881 and dihydrotestosterone (DHT) respectively, exhibits reduced expression of several 8 previously known AR repressed genes, namely DDC, OPRK1, NOV and SERPIN1 (Wu et al, 2014;Zhao et al, 2012) besides SPINK1 (Fig 2G and H).
Non-steroidal pharmacological inhibitors for AR, namely bicalutamide (Bic) and enzalutamide (Enza) have been widely used for the treatment of locally advanced nonmetastatic as well as metastatic prostate cancer (Chen et al, 2009;Scher et al, 2010;Tran et al, 2009), therefore we determined the effect of these anti-androgens on the expression of SPINK1. To antagonize AR signaling, VCaP cells were treated with Enza and SPINK1 expression was examined, notably a remarkable increase in the SPINK1 both at transcript (~4-fold) and protein levels was observed, accompanied with reduced expression of androgen driven-genes such as KLK3 and ERG (Fig 2I-K). To further corroborate these findings, we treated VCaP cells with Bic (25 and 50µM) and found a significant increase in the SPINK1 expression ( Supplementary Fig S2D-F). Next, to evaluate any change in the oncogenic properties of the R1881-stimulated and/or Bic or Enza treated VCaP cells, Transwell migration assay was performed. A significant increase in the migratory properties of the androgen-stimulated VCaP cells treated with Bic or Enza was observed (Supplementary Fig   S2G). Notably, 22RV1 cells are not much responsive to androgen as VCaP cells, therefore we used a strategy to modulate the AR signaling via priming 22RV1 cells either with R1881 or Enza for 3 days, followed by Enza treatment or R1881 stimulation for the next 3 consecutive days. As anticipated, blocking androgen signaling with Enza in the androgenprimed 22RV1 cells result in significant increase in SPINK1 expression, while Enza-treated 22RV1 cells stimulated with R1881 show a significant repression of SPINK1 transcript ( Fig   2L). To examine the effect of long-term DHT treatment on SPINK1 expression, 22RV1 cells were cultured in DHT (8nM) for 2 months, which resulted in more than ~80% reduction in SPINK1 expression (Supplementary Fig S2H). Conversely, long-term blockade of androgen signaling in 22RV1 cells using Bic (5µM) led to significant increase (~1.5 folds) in the SPINK1 expression ( Supplementary Fig S2I). Similar results were obtained in androgenregulated CWR22Pc cells, a derivative cell line of CWR22 xenograft subjected to long-term Bic treatment (Supplementary Fig S2J).
Alternative to pharmacological inhibition of AR signaling, we used small interfering RNA (siRNA) approach to abolish AR expression in 22RV1 and VCaP cells and examine any change in SPINK1 levels. Similar to the small molecule inhibition of AR signaling, siRNA-mediated AR-silenced 22RV1 cells exhibit moderate increase in the expression of SPINK1 ( Supplementary Fig S2L-N), while a robust increase (~3-fold) in the SPINK1 transcript and protein was observed in AR-silenced VCaP cells (Fig 2M and N and Supplementary Fig S2K). Taken together, our findings demonstrate that AR signaling negatively regulates SPINK1 expression and draws attention to AR antagonists mediated upregulation of SPINK1 in prostate cancer.

AR directly binds to SPINK1 promoter and regulates its expression.
The role of AR has been extensively characterized both as a global transcriptional activator as well as repressor (Hu & Lazar, 2000;Zhao et al, 2012). To examine whether AR directly regulates SPINK1 transcription we looked for putative AR binding sites in the SPINK1 promoter region, and scanned the SPINK1 promoter for the presence of androgen response elements (AREs) by employing publicly available transcription factor binding prediction software, JASPAR (Khan et al, 2018) and MatInspector (Cartharius et al, 2005).
Several putative AREs within the ~5kb region upstream of transcription start site (TSS) of the SPINK1 gene were identified ( Fig 3A). Further, analysis of the publicly available Chromatin Immunoprecipitation-Sequencing (ChIP-Seq) dataset for AR binding in androgen stimulated VCaP cells (GSE8428) revealed another putative ARE on the SPINK1 promoter ( Fig 3B).

0
To confirm AR binding on the SPINK1 promoter, we performed ChIP-quantitative PCR (ChIP-qPCR) for AR in R1881-stimulated 22RV1 cells. A significant enrichment for AR-binding at three distinct binding sites (ARE-1, ARE-2 and ARE-3) was observed in androgen-stimulated 22RV1 cells with respect to EtOH control ( Fig 3C and Supplementary   Fig S3A). Promoters for KLK3 and NOV were used as controls for the AR binding (Wu et al, 2014). Next, we determined the transcription activity of SPINK1 by performing ChIP-qPCR for RNA polymerase II (RNA Pol II), interestingly a significant decrease in the occupancy of RNA Pol II on SPINK1 promoter was noticed in androgen stimulated 22RV1 cells ( Fig 3D).
Further, a decrease in the recruitment of RNA Pol II on the NOV promoter, while no significant change on the KLK3 promoter was observed upon androgen stimulation in 22RV1 Fig S3B). Moreover, ChIP-qPCR for the H3K9Ac, a histone mark for transcriptionally active gene promoters was also performed, and a significant reduction in the enrichment of H3K9Ac activation marks on the SPINK1 promoter was observed in R1881stimulated 22RV1 cells, thus confirming its transcriptionally repressed state ( Fig 3D).

cells (Supplementary
Similarly, a significant increase in the AR-occupancy was observed at all three ARE binding sites on the SPINK1 promoter in androgen-stimulated VCaP cells, while a remarkable decrease in the androgen-induced AR recruitment was noted in the presence of Enza, indicating impaired AR-binding to these ARE sites in the presence of a potent AR antagonist ( Fig 3E). While, no change in the RNA Pol II occupancy on the SPINK1 promoter was found in androgen stimulated VCaP cells ( Supplementary Fig S3D), indicative of its poised transcriptional state.
To further confirm the AR signaling-mediated transcriptional regulation of SPINK1, we constructed luciferase promoter reporter vectors by cloning the ARE containing proximal and distal promoter regions of SPINK1 (SPINK1-PP and SPINK1-DP, respectively) and performed luciferase reporter assay using 22RV1 cells. Upon androgen stimulation, a 1 1 significant decrease in the luciferase activity was observed in 22RV1 cells transfected with the SPINK1-PP ( Fig 3F). While, an increase in the luciferase activity was observed in androgen-stimulated 22RV1 cells transfected with PSA promoter construct, used as a positive control for androgen stimulation. Furthermore, siRNA mediated knockdown of AR also led to significant increase in the reporter activity of the SPINK1-PP transfected 22RV1 cells ( Supplementary Fig S3E). To identify the critical AR binding site for the transcriptional regulation of SPINK1, AREs were mutated (ARE MT) in both SPINK1-PP and SPINK1-DP and luciferase assay was performed. A significant decrease in the luciferase activity was recorded in the 22RV1 cells transfected with ARE WT reporter construct upon androgen stimulation, while mutation of the ARE in the SPINK1-DP showed no change in the luciferase activity ( Fig 3G). Conversely, overexpression of wild type AR results in significant decrease in the luciferase activity of both the SPINK1-PP and SPINK1-DP constructs. No significant change in the luciferase activity was observed when AR mutants (AR-ΔNLS and ARV581F) were overexpressed ( Fig 3G). In conclusion, our findings revealed that upon androgen stimulation, AR binds to the SPINK1 promoter and regulates its transcriptional activity by halting the recruitment of RNA Pol II and decreasing the active transcriptional marks (H3K27Ac) on the SPINK1 promoter. Together these findings support the conclusion that AR functions as a direct transcriptional repressor of the SPINK1, and attenuating AR signaling via potent AR-antagonists, such as enzalutamide relieves SPINK1 transcriptional repression resulting in SPINK1 upregulation ( Fig 3H).

Increased SPINK1 level promotes epithelial-mesenchymal transition (EMT) and stemness in prostate cancer
To identify the biological processes governed by SPINK1 in PCa cells, we established stable SPINK1 silenced 22RV1 cells and performed global gene expression profiling. Stable 1 2 22RV1 cells transduced with lentivirus-based short hairpin RNAs (shRNAs) (shSPINK1-1, shSPINK1-2 and shSPINK1-3) showed more than ~85% knockdown of SPINK1 as compared to the control shScramble (shSCRM) cells ( Supplementary Fig S4A and B). To further elucidate the functional roles of differentially expressed genes in shSPINK1-1, shSPINK1-2 and shSCRM cells, we performed pathway enrichment analysis using DAVID (Database for Annotation, Visualization and Integrated Discovery). Notably, genes down-regulated upon SPINK1 knockdown were associated with critical pathways such as, nervous system development, regulation of transcription, and stem cell population maintenance (Fig 4A; Supplementary Table 1). Further, to confirm the role of SPINK1 in EMT, we performed immunostaining for established EMT markers such as E-cadherin (epithelial marker) and Vimentin (mesenchymal marker) in shSCRM and shSPINK1-1 cells. Intriguingly, SPINK1 silenced cells show a significant increase in the expression of E-cadherin, while a decrease in the expression of Vimentin was observed, indicating that loss of SPINK1 leads to a decrease in mesenchymal marks in PCa cells ( Fig 4B). Previously, we demonstrated the role of SPINK1 in imparting chemoresistance to colorectal cancer cells (Tiwari et al, 2015), thus we investigated whether overexpression of SPINK1 governs the similar attribute in PCa cells. As expected, a significant increase in the chemosensitivity towards well-known chemotherapeutic drugs such as doxorubicin, cisplatin and 5-fluorouracil were recorded in shSPINK1-1 cells as compared to shSCRM cells ( Supplementary Fig S4C-E).
Since, one of the GO terms that showed significant enrichment was maintenance of stem cell population (P<0.01) (Fig 4A), thus we performed prostatosphere assay as a readout to test the self-renewal ability of SPINK1 silenced cells. Noticeably, a significant decrease in the number and size of the prostatospheres was observed in shSPINK1-1, shSPINK1-2 cells as compared to the control cells ( Fig 4C). Furthermore, we also performed the side population (SP) assay by evaluating the efflux of Hoechst dye via ABC-transporters in the 1 3 absence or presence of verapamil, a competitive inhibitor for ABC transporters (Zhou et al, 2001). As anticipated, loss of SPINK1 led to a significant decrease (~29% and ~47%) in the SP in the shSPINK1 cells as compared to control ( Fig 4D). Since, aldehyde dehydrogenase (ALDH) activity is crucial for promoting stemness and chemoresistance in cancer stem cells (Burger et al, 2009;Le Magnen et al, 2013), we found a significant decrease in the percent ALDH activity of the SPINK1-silenced 22RV1 cells as compared to the control ( Fig 4E).
Finally, taking a lead from our most significantly enriched GO term underscoring nervous system development (P< 0.001) (Fig 4A), we next investigated any alterations in the expression of neuroendocrine prostate cancer (NEPC) markers such as SYP, CHGA, ENO2 in SPINK1-silenced 22RV1 cells, notably a significant decrease in the expression of SYP (Supplementary Fig S4F) was observed in shSPINK1 cells relative to control, although no change was observed in the CHGA and ENO2 levels (data not shown). Nevertheless, 22RV1 cells with transient SPINK1 knockdown show a significant reduction in the surface expression of the neural cell adhesion molecule-1 (NCAM1), an established marker of neural lineage and known to induce neurite outgrowth ( Fig 4F). Taken together, our findings highlight the predominant role of SPINK1 in mediating EMT, stemness and promoting drug resistance in prostate cancer.

Androgen deprivation-mediated SPINK1 upregulation is associated with neuroendocrine-like (NE-like) phenotype in prostate cancer.
To understand the effect of long-term androgen deprivation on SPINK1 expression, we investigated publicly available gene expression profiling dataset (GSE8702), wherein LNCaP cells (SPINK1-negative) were androgen deprived for 12 months. Remarkably, with prolonged androgen deprivation, a robust increase in the SPINK1 expression was noticed ( Fig   5A). Further, Gene Set Enrichment Analysis (GSEA) revealed that with long-term androgen deprivation, there was a significant decrease in the expression of androgen-signaling To investigate the significance of SPINK1 in governing the cellular plasticity in context of AR signaling, we used LNCaP-derived CRPC cell line, namely 16D CRPC , and its derivative 42D ENZR and 42F ENZR cell lines established via multiple serial transplantation of 1 5 the enzalutamide resistant tumors in male athymic mice (Bishop et al, 2017). These enzalutamide-resistant cell lines harbor reduced AR activity as depicted by the minimal expression level of PSA as compared to parental 16D CRPC cells ( Fig 5F). Further, GSEA plots using the RNA-seq data of 16D CRPC and 42D ENZR cells reveal reduced expression of genes associated with AR signaling, with concomitant increase in the expression of neuronal makers and genes-associated with neurogenesis ( Supplementary Fig S5H). Moreover, these enzalutamide resistant cell lines, 42D ENZR and 42F ENZR also show higher expression of NEPC markers such as SYP, CHGA and ENO2 relative to the parental line (Supplementary Fig S5I).
Notably, 42D ENZR and 42F ENZR cells exhibit increased expression of SPINK1 both at transcript and protein levels as compared to the 16D CRPC cells ( To further examine the effect of ADT on the SPINK1 expression in clinical samples, we examined the expression of SPINK1 in a TMA comprising of PCa patient specimens (n=88) by performing IHC staining, wherein 55 out of 88 patients were given neoadjuvant hormone therapy (NHT) for 3 months. In concordance with our in vitro findings including enzalutamide-resistant tumors derived cells, about ~38% (21 out of 55) patients who were administered NHT exhibits SPINK1 positive status compared to only ~24% (8 out of 33) in the untreated group ( Fig 5K). Although, ADT or NHT-mediated SPINK1 upregulation and associated risk factors need to be tested in a larger PCa patient cohort. Collectively, our finding shows a trend that androgen-deprivation therapies may have an adverse effect, and the benefits must be weighed against treatment. Conclusively, we also show that elevated SPINK1 levels during NE-transdifferentiation strongly emphasizes the potential role of SPINK1 in governing stemness and cellular plasticity in prostate cancer.

SPINK1 expression is modulated by reprogramming factor SOX2 and AR transcriptional co-repressor REST
The role of SRY (sex determining region Y)-box 2 (SOX2) has been implicated in neuroendocrine differentiation and reprogramming/lineage plasticity in RB1 and TP53 deficient prostate cancer (Mu et al, 2017;Russo et al, 2016). Moreover, SOX2 has also been known as an androgen repressed gene (Kregel et al, 2013). Since, our data also showed a similar trend of decrease in SOX2 expression as SPINK1 in androgen-stimulated 22RV1 cells ( Supplementary Fig S6A), we sought to examine SOX2 mediated regulation of SPINK1 expression. We scanned the SPINK1 promoter for the SOX2 binding motif using MatInspector (Cartharius et al, 2005), and identified three putative binding sites (S1, S2 and S3) adjacent to the TSS (Fig 6A). To confirm SOX2 binding on the SPINK1 promoter, we performed ChIP-qPCR in 22RV1 cells, an endogenous SOX2 positive cell line, interestingly a significant enrichment for SOX2-binding at all the three distinct binding sites was observed 1 7 (S1, S2 and S3) (Fig 6B). To ascertain the transcriptional significance of SOX2 binding, we also looked for RNA Pol II binding on these sites and found an enrichment in the occupancy of RNA Pol II on the SPINK1 promoter ( Fig 6C). However, no change in SPINK1 expression was observed upon siRNA-mediated knockdown of SOX2, suggesting that other regulatory factor(s) might be involved in governing SPINK1 expression (Supplementary Fig S6B). To further investigate whether increase in the SPINK1 level in LNCaP cells during NE-like transdifferentiation is transcriptionally regulated by SOX2, we examined for SOX2 recruitment on the SPINK1 promoter region using LNCaP cells cultured in androgen deprived condition (LNCaP-AI) for 15 days. Interestingly, we found a remarkable enrichment of SOX2 on the SPINK1 promoter in LNCaP-AI cells as compared to the normal LNCaP cells grown in regular media ( Fig 6D). In addition, an increase in the occupancy of RNA Pol II was also noticed on the SPINK1 promoter in LNCaP-AI cells (Fig 6E), signifying the increased transcriptional activity of SPINK1 promoter. Next, to further confirm that SOX2 positively regulates SPINK1, we overexpressed SOX2 in LNCaP cells (SPINK1-negative), and examined for the expression of SPINK1, in agreement, a significant increase in the SPINK1 expression was observed in SOX2 overexpressing LNCaP cells relative to the control cells ( Fig 6F). Moreover, a significant increase in the luciferase activity of SPINK1-DP promoter was observed in the SOX2 overexpressing LNCaP cells ( Fig 6G).
Downregulation of transcriptional repressor REST has been well-established phenomenon involved in the transdifferentiation of CRPC to NEPC phenotype (Lapuk et al, 2012). On the other hand, REST is also known to act as a transcriptional co-repressor for AR for a subset of genes involved in promoting NE-like phenotype (Svensson et al, 2014). Since, we have already established that the AR signaling plays critical role in transcriptional repression of SPINK1, and upregulation of SPINK1 positively correlates with NE-like phenotype, thus we next examined the plausible association of SPINK1 with REST and with 1 8 other members of REST complex using TCGA-PRAD dataset. Quartile-based normalization method was used to stratify the patients based on high and low SPINK1 expression, notably SPINK1-high patients (SPINK1-positive) show inverse correlation between expression of SPINK1 and REST as well as other members of REST complex such as RCOR1, SIN3A, HDAC1 ( Supplementary Fig S6C and D). Next, we sought to examine the role of REST in AR-mediated transcriptional repression of SPINK1 in PCa cell lines. Notably, androgen stimulation in three different PCa cell lines, 22RV1, LNCaP and VCaP results in a significant increase in the REST expression ( Supplementary Fig S6E), while abrogating the AR signaling using AR-antagonists in VCaP cells resulted in significant decrease in REST expression ( Supplementary Fig S6F). To investigate whether REST is acting as a transcriptional co-repressor of AR in the regulation of SPINK1, we examined SPINK1 promoter for the REST binding motif using MatInspector (Cartharius et al, 2005), and checked for the recruitment of both AR and REST within ~5Kb region of the TSS of the SPINK1 gene ( Fig 6H). As expected, a robust enrichment of AR at all the three distinct ARE binding sites (ARE-1, ARE-2 and ARE-3) was observed in androgen-stimulated LNCaP cells with respect to EtOH ( Fig 6I); intriguingly, a remarkable recruitment of the REST was also observed at the three distinct RE1 sites (R1, R2 and R3) adjacent to the AR occupied ARE sites ( Fig 6I and Supplementary Fig S6G). Moreover, REST is post-translationally regulated significant increase in REST levels was observed, subsequently resulting in more than ~70% decrease in the SPINK1 expression at the highest concentration of iCK1 (Fig 6J), along with a concomitant decrease in SYP expression (Fig 6J). To examine the functional relevance of CK1 inhibition, we treated 22RV1 cells with a range of iCK1 concentrations, and a significant reduction in the cell viability was observed at higher concentration (10µM and 20µM) of iCK1 (Fig 6K). Similarly, iCK1 treatment also diminished the foci forming ability of the 22RV1 cells (Fig 6L). To evaluate its effect on the neoplastic transformation ability of 22RV1 (SPINK1+) cells, we also performed three-dimensional tumor spheroid assay, intriguingly a significant reduction in the number and size of the tumor spheroids was observed with the highest concentration of iCK1 ( Fig 6M). In summary, our findings suggest that the iCK1 stabilizes the REST levels, which in cooperation with AR, elicits transcriptional repression of SPINK1 and inhibits SPINK1-mediated oncogenic properties.
Collectively, we have shown the direct role of SOX2 in the transcriptional regulation of SPINK1 in prostate cancer. We also establish that REST acts as a transcriptional corepressor of AR in modulating SPINK1 expression, thus a cease in the AR signaling during NE-transdifferentiation results in SPINK1 upregulation, and its overexpression positively associates with NE-like phenotype (Fig 7).

Discussion
Overexpression of SPINK1 in prostate cancer has been associated with poor  (Paju et al, 2007). They also showed elevated SPINK1 level to be associated with higher Gleason grade, and higher expression of neuroendocrine marker, CHGA in a subset of cells (Paju et al, 2007). Taken together, these independent reports further strengthen our findings that SPINK1 is an androgen-repressed and SPAK represents important features of AR signaling in prostate, while same set of genes also play critical role in diverse physiological processes in the neoplastic progression (Nelson et al, 2002 Moreover, REST is also known to mediate gene repression by recruiting CoREST and SIN3A by binding to the RE1 elements of the target genes, which in turn recruits histone deacetylases, HDAC-1/-2, thereby governing epigenetic reprogramming (Noh et al, 2012).
Therefore, it is likely that AR and REST repressive complex may also involve other cofactors such as CoREST in repressing the SPINK1 expression. In another recent study, REST is shown to be a downstream effector of PI3K/AKT signaling, and inhibitors targeting this Furthermore, REST also suppresses EMT and stemness by repressing the expression of genes, namely CD44 and TWIST1 in 22RV1 cell line which is known to exhibit NE-like phenotype (Chang et al, 2017a). In the present study, we have identified REST to be a transcriptional corepressor of AR, which negatively regulates SPINK1. Moreover, we also found inverse association between SPINK1 and REST as well as other members of the REST interacting complex such as RCOR1, SIN3A and HDAC1 in PCa patients (Supplementary Fig   S6C). Importantly, we show that targeting the ubiquitination-dependent REST degradation could be attributed to decrease in the side-population as well as ALDH activity (Fig 4E), thus indicating the role of SPINK1 in imparting chemoresistance, stemness and cellular plasticity.
Recently, we have shown that SPINK1 expression positively correlates with EZH2, a member of Polycomb repressive complex 2, known to induce pluripotency and stemness (Bhatia et al, 2018). Furthermore, SOX2 has been implicated as a key regulator in governing pluripotency, neural differentiation (Zhang & Cui, 2014), and being an AR repressed gene, it is also known to drives NE-transdifferentiation (Mu et al, 2017). Notably, knockdown of SOX2 in mouse embryonic cells result in down-regulation of Spink3 (Sharov et al, 2008 REST levels relieve the repression of the SPINK1 promoter, subsequently SOX2 gets recruited onto SPINK1 promoter resulting in its enhanced transcriptional activity. Conclusively, our findings emphasize that administering PCa patients with androgen deprivation therapies, may result in increased SPINK1 levels accompanied by upregulation of NE markers, thus escalating ADT-associated NEPC incidence (Fig 7). Although, androgen ablation therapy is well-established for the treatment of PCa patients, its long-term benefits are still debatable, but several studies have identified EMT and metastasis as potential disregarded consequences of the potent AR-antagonists used in the clinics. Thus, there is an obligation to take a well-informed decision and the ADT benefits must be weighed before administering anti-androgens, or drug-regimens involving AR-antagonists.

Human Prostate Cancer Specimens
Tissue microarrays (TMA) with prostate cancer (PCa) specimens were obtained from

Conflict of interest
The authors declare no conflict of interest.     In panels (C), (E) and (F) biologically independent samples were used (n=3); data represents mean ± SEM. * P≤ 0.05 and * * P≤ 0.001 using two-tailed unpaired Student's t test. (H) Immunostaining for SPINK1 using same LNCaP xenografts derivatives as in (F).
(I) Heatmap representing fold increase in SPINK1 transcript versus AR target genes in 42D ENZR cells compared to 16D CRPC . Data is plotted as reads per million.

Figure 2
A Relative MFI/Area (AU)