LIMK2-NKX3.1 Engagement Promotes Castration-Resistant Prostate Cancer

Simple Summary Prostate cancer is the principal cause of cancer-related mortality in men. While localized tumors can be successfully treated by orchiectomy or medical castration, most of the patients ultimately progress to the castration-resistant prostate cancer (CRPC) stage, which is incurable at present. Thus, uncovering the underlying mechanisms that cause CRPC could result in promising therapeutics. Our laboratory has identified LIMK2 kinase as an actionable target for CRPC. LIMK2 is vastly expressed in CRPC but minimally in normal prostates. LIMK2 knockout mice are healthy, indicating that LIMK2 inhibition should have minimal toxicity. LIMK2 is also expressed in other aggressive cancers; however, the molecular mechanisms leading to malignancy remain mostly unknown. This study identified that LIMK2 downregulates a prostate-specific tumor suppressor protein-NKX3.1 using two mechanisms. NKX3.1 loss is strongly associated with prostate cancer. Thus, LIMK2 inhibitor provides a powerful opportunity to rescue NKX3.1 loss, thereby preventing and/or delaying prostate cancer progression. Abstract NKX3.1’s downregulation is strongly associated with prostate cancer (PCa) initiation, progression, and CRPC development. Nevertheless, a clear disagreement exists between NKX3.1 protein and mRNA levels in PCa tissues, indicating that its regulation at a post-translational level plays a vital role. This study identified a strong negative relationship between NKX3.1 and LIMK2, which is critical in CRPC pathogenesis. We identified that NKX3.1 degradation by direct phosphorylation by LIMK2 is crucial for promoting oncogenicity in CRPC cells and in vivo. LIMK2 also downregulates NKX3.1 mRNA levels. In return, NKX3.1 promotes LIMK2’s ubiquitylation. Thus, the negative crosstalk between LIMK2-NKX3.1 regulates AR, ARv7, and AKT signaling, promoting aggressive phenotypes. We also provide a new link between NKX3.1 and PTEN, both of which are downregulated by LIMK2. PTEN loss is strongly linked with NKX3.1 downregulation. As NKX3.1 is a prostate-specific tumor suppressor, preserving its levels by LIMK2 inhibition provides a tremendous opportunity for developing targeted therapy in CRPC. Further, as NKX3.1 downregulates AR transcription and inhibits AKT signaling, restoring its levels by inhibiting LIMK2 is expected to be especially beneficial by co-targeting two driver pathways in tandem, a highly desirable requisite for developing effective PCa therapeutics.


Introduction
Prostate cancer (PCa) is among the most prevalent noncutaneous cancers and one of the leading causes of cancer-related death in men in the western world [1,2]. Androgen deprivation therapy (ADT), which inhibits the androgen receptor (AR) signaling, is considered the gold standard treatment for PCa. However, ADT has a temporary effect and most of the patients ultimately develop castration-resistant prostate cancer (CRPC) [3]. CRPC is a heterogeneous cancer that has a poor prognosis, with 9-36 months of median survival. The lower panel shows a gel stained with Coomassie blue. In lane 1, LIMK2 was incubated with radiolabeled ATP and without NKX3.1. In lane 2, NKX3.1 was incubated with radiolabeled ATP and without LIMK2. In lane 3, NKX3.1 was incubated with radiolabeled ATP and LIMK2. The in vitro kinase assay was repeated three times; (B) LIMK2 depletion does not impact the subcellular localization of NKX3.1 in C4-2 cells. Immunofluorescence micrographs of C4-2 cells infected with scrambled and LIMK2 shRNA were probed with NKX3.1 antibody (red). Nuclear counterstain is represented by DAPI (blue).
[Scale bar = 20 µm]; (C) Quantification of subcellular localization based on analysis of micrographs in B. In total, 100 cells were counted from 20 different frames; (D) Treatment of C4-2 cells with LIMK2 shRNA lentivirus did not change NKX3.1 localization. C4-2 cells were treated with LIMK2 shRNA lentivirus, and NKX3.1 location was analyzed using fractionation. Actin and lamin A were used as cytoplasmic and nuclear controls, respectively. The experiment was performed three times; (E) LIMK2 depletion does not impact the subcellular localization of NKX3.1 in 22Rv1 cells. Immunofluorescence analysis of subcellular localization of NKX3.1 (red) in 22Rv1 cells with and without LIMK2 knockdown. DAPI (blue) is used for nuclear counterstain. To probe the relationship between LIMK2 and NKX3.1, two CRPC cell lines, 22Rv1 and C4-2, were utilized. NKX3.1 was predominantly nuclear as expected ( Figure 1B,C). LIMK2 depletion showed no change in NKX3.1 localization in C4-2 cells ( Figure 1B,C). To confirm this result, we also performed nuclear and cytoplasmic fractionation of scrambled shRNA-treated and LIMK2-knocked down cells, which showed minimal change in NKX3.1 localization upon LIMK2 depletion ( Figure 1D). We obtained the same results in 22Rv1 cells, confirming that LIMK2 does not regulate NKX3.1 localization in CRPC cells ( Figure 1E-G). We further confirmed these results using an LIMK2 inhibitor-N-benzyl-Nethyl-4-(N-phenylsulfamoyl)benzamide (LI) [25]. The IC 50 of LI was found to be 2.07 and 2.17 µM in C4-2 and 22Rv1 cells, respectively, when treated for 48 h ( Figure S5). Thus, LIMK2 was inhibited for 12 h at the 10 µM concentration and its impact on the subcellular location of NKX3.1 analyzed. Similar to the results obtained upon LIMK2 silencing, LIMK2 inhibition did not change the nuclear localization of NKX3.1 in both C4-2 and 22Rv1 cells ( Figure 1H-K). Together, these data show that LIMK2 does not regulate the subcellular residence of NKX3.1.

LIMK2 Negatively Regulates NKX3.1
NKX3.1 is downregulated upon castration in mice, which is consistent with its origin as an androgen-responsive gene [26]. We showed that LIMK2 levels increase upon castration due to increased hypoxia [4]. Therefore, we suspected that LIMK2 may also be responsible for NKX3.1 downregulation. To check this hypothesis, we infected C4-2 cells with LIMK2 retrovirus transiently, which caused significant downregulation of NKX3.1 (Figure 2A,B). Consistent with this finding, overexpression of LIMK2 reduced expression of NKX3.1 in 22Rv1 cells as well ( Figure 2C,D). Besides, LIMK2 knockdown using the corresponding shRNA in C4-2 cells enhanced NKX3.1 levels ( Figure 2E,F). To confirm specificity, we also used additional LIMK2 shRNA to investigate its impact on NKX3.1 levels, which confirmed that LIMK2 negatively regulates NKX3.1 ( Figure 2G). Analogous results were obtained in 22Rv1 cells ( Figure 2H-J). Thus, these results indicate that LIMK2 negatively impacts NKX3.1.   To uncover the mechanistic basis of NKX3.1 downregulation by LIMK2, we first analyzed the mRNA levels of NKX3.1 level in response to LIMK2 overexpression in C4-2 and 22Rv1 cells by qRT-PCR, which revealed massive downregulation of NKX3.1 mRNAs in both cell lines ( Figure 2K,L). This result was confirmed by silencing LIMK2 in C4-2 and 22Rv1 cells, which resulted in about 3-fold increase in NKX3.1 mRNA levels ( Figure 2M,N). As NKX3.1 directly represses RAMP1 transcription [27], RAMP1 mRNA levels were used as control, which increased when NKX3.1 mRNA levels decreased and vice versa. These results confirm that LIMK2 downregulates NKX3.1 mRNA levels.
As LIMK2 directly phosphorylates NKX3.1, we hypothesized that LIMK2 may also regulate the protein stability of NKX3.1. C4-2 and LIMK2-depleted C4-2 cells were exposed to cycloheximide to prevent protein synthesis, and the NKX3.1 degradation profile compared, which revealed a robust increase in NKX3.1 stability when LIMK2 was reduced ( Figure 2O,P). LIMK2 knockdown in 22Rv1 cells also increased NKX3.1 stability, thereby confirming that LIMK2-mediated NKX3.1 downregulation is a common mechanism in CRPC cells ( Figure 2Q,R). We next examined whether LIMK2-mediated NKX3.1 degradation was dependent or independent of ubiquitin. LIMK2 overexpression indeed resulted in enhanced ubiquitylation of NKX3.1 in C4-2 and 22Rv1 cells ( Figure 2S,T), thereby confirming that LIMK2 regulates NKX3.1 both at transcriptional and post-translational stages. Figure S2 shows raw data for Figure 2A [4,9,10]. Therefore, we investigated whether such a loop exists between NKX3.1 and LIMK2. We overexpressed NKX3.1 in C4-2 and 22Rv1 cells, which caused robust downregulation of LIMK2 in both cell types ( Figure 3A-D). Similarly, when NKX3.1 was knocked-down in CRPC cells, LIMK2 levels increased substantially, confirming the reciprocal cross talk ( Figure 3E-H).   As NKX3.1 is a transcription factor, we examined whether it upregulates LIMK2 at the transcriptional stage. C4-2 and 22Rv1 cell lines were infected with NKX3.1 retrovirus. Overexpression of NKX3.1 induced a robust reduction in LIMK2 mRNA levels in both C4-2 and 22Rv1 cells ( Figure 3I,J, respectively). RAMP1 was used as a control, since it shows an inverse correlation with NKX3.1 mRNA. Similarly, NKX3.1 knockdown led to a more than 2-fold increase in both cell types ( Figure 3K,L). These results show that NKX3.1 downregulates LIMK2 mRNA levels.

LIMK2 Phosphorylates NKX3.1 at S185 Triggering Its Degradation
To probe the mechanism further, we analyzed LIMK2-mediated phosphorylation sites on NKX3.1. LIMK2 is a dual-specificity kinase and thus can phosphorylate S, T, or Y. Nevertheless, the only sites reported to be phosphorylated by LIMK2 are serine. The optimal sequence that LIMK2 may prefer also remains unidentified. Based on the cofilin phosphorylation site, we previously predicted that LIMK2 prefers serine residues, which are followed by alanine or glycine, which led to the identification of LIMK2-mediated sites on TWIST1, PTEN, and SPOP in vitro and in cells. Based on these criteria, we identified S185 as the potential site for NKX3.1. We generated S185A phospho-dead mutant and exposed it to a kinase assay with LIMK2. While WT-NKX3.1 was efficiently phosphorylated, the S185A mutant showed no phosphorylation ( Figure 4A), confirming that LIMK2 only phosphorylates the S185 site in NKX3.1.
To validate whether the S185 site is indeed phosphorylated by LIMK2 in cells, we isolated HA-tagged WT and S185A-NKX3.1 proteins from the corresponding WT and S185A-NKX3.1-expressing C4-2 cells using HA antibody and analyzed their phosphoserine (phospho-Ser) levels using a phospho-serine antibody. Parental C4-2 cells were used as a negative control. To ensure that equal amounts of WT and mutant protein were analyzed, excess cell lysate was used with a limiting amount of HA antibody. As indicated in Figure 4B,C, WT-NKX3.1 showed >3-fold higher phospho-Ser levels, as compared to the S185A mutant. Importantly, to examine the role of LIMK2 in this process, it was inhibited in both C4-2 and 22Rv1 cells, which caused phospho-Ser levels of NKX3.1 to plummet significantly, whereas phospho-Ser levels of S185A remained unchanged. This occurrence was confirmed in 22Rv1 cells, which too showed that LIMK2 regulates the phosphorylation of NKX3.1 at the S185 site in CRPC cells ( Figure 4D,E).
To probe the consequences of NKX3.1 phosphorylation by LIMK2, we ectopically expressed WT and S185A mutant in C4-2 cells, which revealed increased expression of the mutant as compared to WT ( Figure 4F,G), indicating that phospho-resistant mutant is more stable than WT. Furthermore, increased NKX3.1 levels were associated with lower levels of LIMK2, confirming the negative loop. Analogous results were observed in 22Rv1 cells ( Figure 4H,I).
As S185A-NKX3.1 is more stable than the WT, an equivalent decrease in LIMK2 levels was observed in corresponding C4-2 and 22Rv1 cells ( Figure 4F-I). Therefore, we examined the ubiquitylation of LIMK2 in WT and S185A-NKX.1-expressing C4-2 and 22Rv1 cells. As expected, S185A mutant-expressing CRPC cells showed increased ubiquitylation of LIMK2 as compared to WT-NKX3.1-expressing cells, thereby confirming the feedback loop between LIMK2 and NKX3.1 ( Figure 4N,O). Figure S4 shows raw data for Figure 4B NKX3.1 is a tumor suppressor and inhibits cell proliferation [19]. As phospho-resistant S185A-NKX3.1 is more stable, we reasoned that it should enhance the inhibitory effect of NKX3.1 on cell growth. To test this hypothesis, we measured the relative growth rate of C4-2, NKX3.1-C4-2, and S185A-NKX3.1-C4-2 cells. As expected, S185A expression was more effective in reducing cell proliferation as compared to WT, whereas parental C4-2 cells displayed the maximal growth rate ( Figure 5A). Ectopic expression of WT and S185A-NKX3.1 showed a similar pattern in 22Rv1 cells as well, indicating that degradation of NKX3.1 by LIMK2-mediated phosphorylation may be a key step in promoting oncogenic phenotypes ( Figure 5B). To investigate this further, we ectopically expressed LIMK2 in C4-2, NKX3.1-C4-2, and S185A-NKX3.1-C4-2 stable cells, and examined the change in cell proliferation rates as compared to the corresponding parental cell lines. LIMK2 overexpression increased the cellular growth in all cell lines, albeit least in S185A-expressing cells, indicating that LIMK2 uses multiple pathways to promote oncogenicity, including NKX3.1 phosphorylation and subsequent degradation ( Figure 5C). for 18 and 36 h. Proliferative activity was measured by MTT assay. The absorbance was measured at 570 nm. The results are plotted as the means ± SD of three independent experiments. * p < 0.05; (C) C4-2 cells stably expressing NKX3.1 and S185-NKX3.1 were infected by LIMK2 retrovirus. Proliferative activity was measured by MTT assay. All data are from three independent experiments. **** p < 0.0001; (D) NKX3.1 suppressed cell migration in C4-2 cells. Target cells were infected by NKX3.1 shRNA lentivirus, LIMK2 sRNA lentivirus, NKX3.1 retrovirus, and S185A-NKX3.1 retrovirus. The cells were starved in serum-free media for 12 h and the migration assay was performed using Boyden chambers; (E) The results are plotted as the means ± SD of three independent experiments as described in (D). * p < 0.05, ** p < 0.01; (F) NKX3.1 suppressed cell migration in 22Rv1 cells; (G) The results are plotted as the means ± SD of three independent experiments as described in (F). * p < 0.05, ** p < 0.01; (H) Migration assays of C4-2 cells that stably expressed NKX3.1 and S185A-NKX3.1 infected by LIMK2 retrovirus are shown. C4-2 cells were used as control; (I) The results are presented as the mean ± SD of three independent experiments. The graph represents the data analysis as compared to untreated C4-2 cells. ** p < 0.01, *** p < 0.001; (J) NKX3.1 and S185A-NKX3.1 inhibit colony formation ability of C4-2 cells; (K) Quantitative data analysis of the soft-agar experiment as shown in J. All data were from n = 3. * p < 0.05, ** p < 0.01. (L) LIMK2-mediated phosphorylation of NKX3.1 does not impact the subcellular localization of NKX3.1, as both the WT and S185A mutant show nuclear residence. Immunofluorescence micrographs of C4-2 cells infected with WT or S185A mutant and probed with HA antibody (green). We next investigated the effect of NKX3.1 phosphorylation on chemotaxis. As indicated, NKX3.1 knockdown greatly increased cell migration ( Figure 5D,E), whereas LIMK2 silencing had the opposite effect. In vitro migration assays in C4-2, NKX3.1-C4-2, and S185A-NKX3.1-C4-2 cells displayed a similar trend, where C4-2 cells showed robust migration, where WT-NKX3.1-and S185A-NKX3.1-expressing cells were impaired in cell migration ( Figure 5D,E). We observed an analogous pattern of cellular migration in 22Rv1 cells, where LIMK2 depletion decreased, and NKX3.1 silencing increased cell migration ( Figure 5F,G). Consequently, ectopic expression of NKX3.1 (WT and S185A mutant) in 22Rv1 cells significantly inhibited cell migration. We next examined whether LIMK2 overexpression rescues WT and S185A-NKX3.1-mediated suppression of motility in C4-2 cells. While LIMK2 overexpression considerably increased cell motility in C4-2 cells, both WT and S185A-expressing cells showed moderate enhancement, suggesting that NKX3.1 degradation by LIMK2 is an important step in increasing cell migration ( Figure 5H,I).
Although our IF data showed that LIMK2 does not regulate the subcellular localization of NKX3.1, nevertheless, we investigated whether phosphorylation-resistant HA-tagged NKX3.1 shows similar localization in C4-2 cells. HA-tagged WT NKX3.1 was used as the control. As shown in Figure 5L,M, both WT and S185A-NKX3.1 showed nuclear localization, confirming that LIMK2 does not affect the nuclear residence of NKX3.1.
2.9. The Phosphomimetic S185D-NKX3.1 Promotes Cell Proliferation, Colony Formation, and Chemotaxis As our data showed that the S185D mutant is expressed at a lower level as compared to WT-NKX3.1, we measured its impact on cell proliferation and chemotaxis. In accordance with its decreased stability, S185D-NKX3.1-expressing C4-2 cells showed increased proliferative capacity, compared to parental cells, whereas WT-expressing cells showed a reduced proliferation rate ( Figure 6J). We observed a similar trend in 22Rv1 cells upon ectopic expression of S185D mutant ( Figure 6K). Similarly, S185D mutant promoted colony formation, whereas WT inhibited it ( Figure 6L). Likewise, ectopic expression of NKX3.1 fully inhibited chemotaxis in both C4-2 and 22Rv1 cells, whereas S185D expression pro-moted cell motility, which was higher than the corresponding parental cells (Figure 6M-P). These results are consistent with the increased stability of LIMK2 in S185D-expressing CRPC cells ( Figure 6E-H), which promotes oncogenic phenotypes. Figure S6 shows raw data for Figure 6A,C,E,G,I.

NKX3.1 Downregulates AKT Activation and Decreases AR and ARv7 Levels in CRPC Cells
As NKX3.1 is known to inhibit AKT activation, we investigated whether LIMK2mediated phosphorylation of NKX3.1 regulates AKT activation. While parental C4-2 cells showed robust AKT activation as visualized by S473 and T308 phosphorylation, ectopic expression of WT or S185A mutant fully suppressed it, whereas total AKT levels remained unchanged in both C4-2 and 22Rv1 cells ( Figure 7A-D). AR levels were next analyzed in C4-2, NKX3.1, and S185A-NKX3.1 cells, which also showed severely reduced AR levels upon WT or S185A expression ( Figure 7E,F). In 22Rv1 cells, both AR and ARv7 levels decreased drastically upon WT and mutant NKX3.1 expression (Figure 7G,H). Notably, in each case, S185A-NKX3.1 was expressed at higher levels as compared to WT; nevertheless, we did not observe any differences in AR and ARv7 levels between the two cell types. It was presumably because WT NKX3.1 expression decreased AR and ARv7 to almost basal levels, therefore a relatively higher expression of mutant NKX3.1 could not show any additional impact. Figure S7 shows raw data for Figure 7A,C,E,G.   The data from three independent experiments were plotted with ** p < 0.01 compared to vector-expressing cells; (G) Ectopic expression of NKX3.1 and S185A-NKX3.1 decreases AR and ARv7 protein levels in 22Rv1 cells. Vector-expressing control, wild-type NKX3.1, and phospho-resistant NKX3.1-overexpressing 22Rv1 cells were lysed and probed for AR and ARv7 protein levels using Western blot analysis; (H) The bar graph shows changes in AR and ARv7 protein levels in 22Rv1, NKX3.1-22Rv1, and S185A-NKX3.1-22Rv1 cells. The data from three independent experiments were plotted as mean ± SEM, ** p < 0.01 vs. 22Rv1 vector-expressing control cells; (I) Schematic model portraying the plausible role of LIMK2-NKX3.1 signaling in CRPC pathogenesis.

NKX3.1 Suppresses Tumor Growth In Vivo
Our results indicated that both WT and S185A-NKX3.1 suppress aggressive phenotypes in cells, but the latter is more effective, leading us to hypothesize that while WT NKX3.1 may partially inhibit tumorigenesis in mice, S185A mutant should fully inhibit it. Therefore, we investigated the tumor-suppressing potential of C4-2 and WT NKX3.1-C4-2 cells in nude mice. C4-2 and NKX3.1-C4-2 cells were injected in the left and right flanks, respectively. We measured tumor size every alternative day. As expected, animals receiving C4-2 developed robust tumors. By contrast, WT-NKX3.1 expression fully suppressed tumorigenesis (Material Figure S8A,B). As WT-NKX3.1 completely suppressed tumorigenesis, S185A mutant was not analyzed in xenograft models. These results thus highlight the tumor-suppressive impact of NKX3.1 in CRPC.

Discussion
NKX3.1, an androgen-regulated homeobox transcription factor, was the first known prostate epithelium-specific marker, which is indispensable for normal prostate and testes development [26]. NKX3.1 promotes prostatic epithelial specification and differentiation and is essential for maintaining luminal stem cells [14,28]. NKX3.1 is also an established marker for PCa. NKX3.1 is located on the short arm of chromosome 8p21. Loss of heterozygosity (LOH) at this locus occurs in up to 89% of high-grade PIN lesions and up to 86% of prostate adenocarcinomas [29]. Thus, NKX3.1 is present in primary prostate adenocarcinomas, although the levels decline in higher-grade lesions but are still not lost completely, consistent with its haploid status [18,30]. Nevertheless, the loss or reduction in NKX3.1 levels is a key initiating event in PCa [31,32]. Germline loss-of-function of Nkx3.1 in mice causes pre-malignant lesions similar to PIN. NKX3.1 gene copy loss is significantly more frequent in CRPC than in localized disease, indicating that a loss or reduction in NKX3.1 levels is critical for PCa progression, including in CRPC.
Although NKX3.1 transcriptional regulation certainly reduces its levels during PCa progression in some cases, there is a clear disagreement between NKX3.1 protein and mRNA levels in normal and PCa clinical tissues [30]. NKX3.1 protein was reduced in four out of six specimens, which had either normal or increased mRNA levels [30]. Similarly, NKX3.1 copy number and NKX3.1 mRNA levels do not show a correlation [33][34][35]. These findings sparked intense interest in examining its regulation at a post-translational level, which revealed it to be a target of several different kinases.
Our recent studies uncovered LIMK2 as a clinical target for CRPC [4]. We showed that LIMK2 stabilizes TWIST1 by direct phosphorylation at four sites, which leads to EMT and cancer stem cell phenotypes in CRPC [4]. LIMK2 degrades tumor suppressor SPOP by direct phosphorylation, which in turn stabilizes AR, ARv7, and c-Myc, promoting CRPC. Phospho-dead SPOP fully abrogates tumorigenesis in mice, demonstrating that SPOP downregulation by LIMK2 is a key event in CRPC [9]. More recently, we have shown that LIMK2 degrades PTEN and inhibits its phosphatase activity by direct phosphorylation at five sites [10]. PTEN destabilizes LIMK2 in response, and thus LIMK2 levels were significantly more in PTEN −/− prostates as compared to prostates from PTEN +/+ mice. PTEN loss often occurs in response to hypoxia. We further observed that hypoxia increases LIMK2 levels, which in turn results in PTEN degradation in CRPC.
Importantly, PTEN and NKX3.1 are intricately linked to PCa progression. While NKX3.1 LOH occurs at an early stage of PCa progression, which leads to PIN [43], PTEN loss occurs at a relatively advanced stage. Nevertheless, loss of Nkx3.1 and Pten acts synergistically, leading to increased Akt signaling and high-grade PIN lesions [44]. More recently, PTEN was shown to dephosphorylate NKX3.1 at S185, which stabilizes it. Thus, PTEN loss also promotes a decrease in NKX3.1 levels [45].
This study uncovered a new mechanism of NKX3.1 regulation by LIMK2 in CRPC pathogenesis. LIMK2 directly phosphorylates NKX3.1, which results in its degradation. Consequently, S185A-NKX3.1 is more effective in resisting cell growth, cell migration, and anchorage-independent growth. Interestingly, LIMK2 is often regulated by its substrates, which prompted us to investigate whether NKX3.1 controls LIMK2. Ectopic expression of WT or S185A mutant decreased LIMK2 levels by increased ubiquitylation, confirming the feedback loop. Importantly, LIMK2 also robustly downregulates NKX3.1 mRNA levels, although the mechanism remains unclear.
As many studies have documented that PTEN loss causes NKX3.1 downregulation, our findings provide a new link between NKX3.1 and PTEN via LIMK2. We show that LIMK2 downregulates both PTEN and NKX3.1 by direct phosphorylation. Thus, if NKX3.1 is genomically lost, it increases LIMK2 levels due to increased stabilization, causing PTEN downregulation by LIMK2. PTEN loss also increases LIMK2 levels, which in turn can degrade NKX3.1 ( Figure 7I).

Cell Lines, Antibodies, and Chemicals
C4-2, HEK293T, 22Rv1, and Phoenix cells were purchased from ATCC. Each antibody was used at 1-1000 dilution. Antibodies details, including their vendors and RRID number, are provided in Table S1. Cycloheximide, MG132, and polybrene were purchased from Sigma (Sigma-Aldrich, St. Louis, MO, USA). LIMK2 inhibitor was synthesized according to the published protocol.

In Vitro Kinase Assays
Recombinant LIMK2 on Ni-NTA beads was treated with 1 mM ATP in kinase buffer for 2 h as reported earlier [49]. Subsequently, the beads were washed 3 times with kinase buffer to remove all excess ATP. LIMK2 was eluted using imidazole buffer and incubated with recombinant NKX3.1 (WT and mutant) in kinase buffer along with and 1-5 µCi [γ-32 P-ATP] for 20-30 min at RT. The reaction was stopped by the addition of SDS-PAGE dye followed by SDS-PAGE gel. The resulting radioactivity was determined by autoradiography.

Transfection and Retroviral Infection
To generate WT or S185A-NKX3.1 retrovirus, the corresponding plasmids were transfected into Phoenix cells using calcium phosphate [12]. The supernatants containing virus particles were collected 48 h post-transfection and used to infect the target cells (C4-2 and 22Rv1) in the presence of polybrene. Protein expression was analyzed 32 h post-infection. Stable cells were generated using puromycin.

Western Blotting
First, 2 × 10 6 cells were plated in 100-mm dishes. After 12 h, they were infected with retroviruses or shRNA lentiviruses. After 32-36 h post-infection, cells were rinsed twice with chilled phosphate-buffered saline and lysed in modified RIPA (1% NP-40, 20 mM Tris, pH 7.5, 150 mM NaCl, 2 mM EDTA, 0.25% sodium deoxycholate) buffer, supplemented with protease inhibitors. After 30 min on ice, cell lysate was centrifuged at 10,000 rpm for 10 min at 4 • C. Bradford assay was used for protein quantification. The proteins were separated using SDS-PAGE and transferred to a PVDF membrane. The blocking was done using 5% skim milk in TBST (20 mM Tris, pH 7.4, 150 mM NaCl, and 0.1% Tween 20). The membrane was incubated with a primary antibody overnight followed by 3-4 h treatment with the secondary antibody. The protein was detected using chemiluminescence-based detection. Image J program was used for densitometry analysis.
To separate the nuclear fraction, the pellet was resuspended in buffer B (5 mM Tris, pH 7.9, 1.5 mM MgCl 2 , 0.2 mM EDTA, 0.5 mM DTT, 26% glycerol (v/v), 300 mM NaCl, and 1 mM PMSF), and homogenized by passing 10 times through a 27 1 2 -gauge needle. The lysates were incubated on ice for 30 min and the nuclear fraction was separated by centrifugation at 24,000× g at 4 • C for 20 min. The cytosolic and nuclear extracts were further analyzed by Western blotting [50].

Cell Viability Assay
Cells were seeded in triplicate in a 24-well plate at 1000 cells per well. After 12 h, the cells were infected with the virus. To detect cell viability, 25 µL of 3-(4,5-dimethylthiazol-2yl)-2,5-diphenyl tetrazolium bromide (MTT), at a final concentration of 0.5 mg/m, were added to each well at indicated times. After 4 h of incubation at 37 • C, the medium was replaced with 500 µL of DMSO. Absorbance was read at 570 nm using a microplate reader (Tecan Spark multimode).

Dose-Response Assay with LIMK2 Inhibitor
C4-2 and 22Rv1 cells were seeded into 96-well plates at 10 4 cells/well concentration. Then, 12 h after seeding, LIMK2 inhibitor was added to the cells at different concentrations (10 nM, 100 nM, 1 µM, 10 µM, and 100 µM) to determine the IC 50 value. An equal volume of DMSO was used as the control. After 48 h of incubation, cell viability was determined using MTT assay. Data were analyzed by GraphPad Prism 5 Software (Graphpad, CA, USA). All experiments were conducted in triplicate, two independent times.

Clonogenic Assay
Soft agar assay was conducted as before (21). Briefly, 0.5% agar was made by mixing an equal amount of Noble agar (1%; DNA grade) and 2 × RPMI (20% FBS) and plated in a 24-well plate. Cells (1250 per well) were mixed with 500 µL of 0.7% agarose in 2 × RPMI 1640 (20% FBS) and added on top of solidified base agar. In total, 500 µL of growth medium were added per well. The cells were fed every 3 days with a culture medium. After 3-4 weeks of incubation at 37 • C, colonies were stained with 0.01% crystal violet and incubated at RT for 45 min. Colonies were detected by a light phase-contrast microscope.

Migration Assay
Migration assay was conducted as we reported before (22). CRPC cells were serumstarved for 12 h in serum-free media and recovered by limited trypsin digestion. Then, 10 5 cells were suspended in 300 µL of DMEM supplemented with 0.5% bovine serum albumin and added to the top compartment of a Boyden chamber, and 400 µL of full growth media were added to the bottom compartment. After incubating for 4 h at 37 • C, chambers were taken out and the inner surface of the membrane was wiped with a cotton applicator. Chambers were washed once with PBS, and cells were counted under a phasecontrast microscope in 10 random fields at 100× magnification. The assays were performed in triplicate, four times. To allow for comparison between multiple assays, the data were normalized and expressed as a percentage of the number of cells present on the membrane.
4.14. qPCR Assay qPCR assays were conducted according to our previously published methods [4]. The primer sequences used for qPCR are included in Table S3.

C4-2 Xenografts in Nude Mice
Male NCRNU-M athymic nude mice (4 weeks old) were obtained from Taconic Laboratories. All mice were housed at the Purdue animal facility, which provided husbandry and clinical care. Tumor injections were done as reported previously [4]. Briefly, 1 × 10 6 cells were mixed with 50% Matrigel and injected subcutaneously in nude mice. Tumor measurements were done every alternative day. Tumor diameter was measured using vernier calipers and volume was calculated based on the formula for spheroid (volume = 4πr 3 /3, where r is the radius).

Statistical Analysis
All data are presented as the mean and error bars represent standard deviation (SD) from 3 biological replicates. Statistical analyses were performed using OriginPro 2019. The significance of difference was determined by one-way analysis of variance followed by Bonferroni's post hoc test. p < 0.05 was considered statistically significant.

Conclusions
While NKX3.1 homozygously deleted tumors can only be treated by gene therapy, preserving the levels of NKX3.1 in heterozygous tumors by LIMK2 inhibition provides a tremendous opportunity for developing targeted therapy in CRPC. Further, as NKX3.1 downregulates AR transcription and inhibits AKT signaling, restoring its levels by inhibiting LIMK2 is expected to be especially beneficial by co-targeting two driver pathways in tandem, a highly desirable requisite for developing effective PCa therapeutics.