Leucine-rich repeat–containing G protein–coupled receptor 4 (Lgr4) is necessary for prostate cancer metastasis via epithelial–mesenchymal transition

Prostate cancer is a highly penetrant disease among men in industrialized societies, but the factors regulating the transition from indolent to aggressive and metastatic cancer remain poorly understood. We found that men with prostate cancers expressing high levels of the G protein–coupled receptor LGR4 had a significantly shorter recurrence-free survival compared with patients with cancers having low LGR4 expression. LGR4 expression was elevated in human prostate cancer cell lines with metastatic potential. We therefore generated a novel transgenic adenocarcinoma of the mouse prostate (TRAMP) mouse model to investigate the role of Lgr4 in prostate cancer development and metastasis in vivo. TRAMP Lgr4−/− mice exhibited an initial delay in prostate intraepithelial neoplasia formation, but the frequency of tumor formation was equivalent between TRAMP and TRAMP Lgr4−/− mice by 12 weeks. The loss of Lgr4 significantly improved TRAMP mouse survival and dramatically reduced the occurrence of lung metastases. LGR4 knockdown impaired the migration, invasion, and colony formation of DU145 cells and reversed epithelial–mesenchymal transition (EMT), as demonstrated by up-regulation of E-cadherin and decreased expression of the EMT transcription factors ZEB, Twist, and Snail. Overexpression of LGR4 in LNCaP cells had the opposite effects. Orthotopic injection of DU145 cells stably expressing shRNA targeting LGR4 resulted in decreased xenograft tumor size, reduced tumor EMT marker expression, and impaired metastasis, in accord with our findings in TRAMP Lgr4−/− mice. In conclusion, we propose that Lgr4 is a key protein necessary for prostate cancer EMT and metastasis.

Prostate cancer is the second most common cancer in American men (after non-melanoma skin cancer) and the second leading cause of cancer mortality (1), with metastasis accounting for most of the mortality. Epithelial-mesenchymal transition (EMT) 5 is a developmental process through which epithelial cells lose their epithelial identity, gain mesenchymal characteristics, and migrate to populate distant regions of the body. Tumors often reinitiate EMT to invade local tissue and intravasate into blood vessels early in metastatic progression (2). In prostate cancer, morphological evidence of EMT was observed within the Gleason grading system itself. As the Gleason grade increases, the epithelial glandular architecture is degraded, featuring loss of tight junctions and cell polarity and the appearance of invasive cells, cords, and sheets (3). Prostate cancer cells with mesenchymal characteristics are more invasive in vitro and produce more metastases in vivo (4). Key drivers of EMT (including Twist and Snail) have been proposed as targets for the development of novel anti-metastatic prostate cancer therapeutics (5,6); however, these transcription factors are challenging to target pharmacologically.
The Wnt signaling pathway has been implicated in EMT from the earliest stages of embryonic development to adult tissue repair (7,8). Wnt signaling also plays a key role in pathological EMT in prostate and other cancers (9). Wnt inhibition in PC-3 cells increased epithelial marker expression, decreased invasive ability, and attenuated SNAIL2 and TWIST activities (10,11), supporting a role for Wnt in regulating EMT in prostate cancer.
The TRAMP mouse is a widely used model for prostate cancer that well recapitulates the pathogenesis of human prostate cancer (28). Specific expression of SV40 T antigen in prostate epithelial cells driven by the probasin promoter is androgendriven and regulated by prostate development. Disease pathogenesis initiates from early PIN lesions between 8 and 12 weeks of age, evolving through well-differentiated (WD) adenocarcinomas at 12 weeks and moderately differentiated (MD) carcinoma by 18 weeks, progressing to poorly differentiated (PD) carcinomas by 24 to 30 weeks.
Accumulating evidence supports a key role for Lgr4 in cancers. Lgr4 expression is elevated in gastric cancers and correlated with nodal spread of tumors (29). Lgr4 is overexpressed in colorectal cancers and associated with metastasis, and high Lgr4 levels are a marker for poor prognosis (30). Overexpression of Lgr4 increased the invasive and metastatic potential of prostate and colon carcinoma cells and prostate cancer xenograft growth; conversely, Lgr4 silencing decreased prostate, cervical, and lung cancer cell invasiveness (31,32). Humans bearing a germ line nonsense mutation in Lgr4 curiously have an increased risk of skin squamous cell carcinoma and biliary tract cancer (26), indicating that Lgr4 has varying roles related to carcinogenesis in different tissues. However, at present, the role of Lgr4 in cancer and cancer stem cells remains unclear.
Here we sought to investigate the role of Lgr4 in prostate cancer. We found that LGR4 was overexpressed in human prostate tumors and in cancer cell lines.
LGR4 expression was negatively correlated with recurrence-free survival in cancer patients, suggesting an important role of Lgr4 in prostate cancer progression. In addition, we generated a new mouse prostate cancer model by crossing Lgr4 Ϫ/Ϫ mice with TRAMP mice to analyze the effects of Lgr4 loss on prostate tumor development. Systemic analysis of this in vivo model indicated that Lgr4 depletion did not affect tumor initiation but impaired prostate cancer metastasis by delaying EMT. Further studies on LGR4 knockdown DU145 cells revealed that LGR4 inactivation impedes EMT by up-regulating E-cadherin, possibly through attenuating Wnt/␤-catenin signaling and decreasing Snail expression levels. Therefore, Lgr4 plays an important role in regulating prostate cancer metastasis.

Elevated LGR4 expression in human prostate cancer is correlated with a shorter time to disease recurrence
To investigate LGR4 function in prostate cancer progression, we first studied the expression level of LGR4 in human prostate and prostate cancer cell lines. Limited LGR4 expression was found in the normal human prostate luminal epithelial cell line PNT1A (Fig. 1, A and B). An elevated LGR4 level was present in the primary human prostate cancer cell lines MDA-PCa-2a and MDA-PCa-2b. The androgen-dependent cell lines LNCaP, LaPC4, and VCaP also showed higher expression of LGR4, and LGR4 expression was highest in the androgen-independent cell line DU145 (Fig. 1, A and B). Notably, LGR4 expression was roughly correlated with the metastatic ability of these cell lines (DU145 and PC-3 Ͼ LNCaP), indicating a potential relationship between prostate cancer metastasis and LGR4 expression. Immunohistochemical analysis of normal human prostate specimens showed that LGR4 is normally expressed in basal epithelial cells of adult prostates, in accord with our previous report (27). High-grade prostatic intraepithelial neoplasia (PIN) or prostate cancers, however, showed a change in LGR4 expression pattern, with a more diffuse epithelial localization not just limited to basal cells, suggesting a role of LGR4 in prostate cancer development (supplemental Fig. S1A). In human prostate patient samples, LGR4 mRNA expression was significantly increased by 78% in prostate cancers compared with normal prostate epithelium (Fig. 1C) (GEO accession no. GSE6099) (33). Therefore, LGR4 expression was increased in human prostate cancers.
We then sought to determine whether tumor Lgr4 expression could predict prostate cancer recurrence by analyzing online microarray databases. According to the Lapointe database (GEO accession no. GSE3933) (34) (Fig. 1D), when dividing the 62 prostate cancer patients according to relative tumor LGR4 expression, cancer recurred within 1 year in 37% of patients with high LGR4 expression, in contrast to only 18% of patients with low LGR4 expression. To further validate this correlation, we compared 60 patients with the highest LGR4 expression with 59 patients with the lowest LGR4 expression in the Nakagawa database (GEO accession no. GSE10645) (35) (Fig. 1E). Although cancer recurrence in the LGR4 hi and LGR4 low groups was similar within 3 years, prostate cancer recurred in 45% of LGR4 hi patients from 3 to 8 years, significantly higher than the cancer recurrence rate of 23.7% in LGR4 low patients. Together, these data strongly suggested that LGR4 modulates prostate cancer progression and metastasis.

Lgr4 ablation in TRAMP mice repressed tumor metastasis and decreased mortality
To determine LGR4 function in prostate tumorigenesis in vivo, we generated a novel TRAMP/Lgr4 Ϫ/Ϫ mouse model by crossing TRAMP transgenic mice with Lgr4 ϩ/Ϫ mice (which have a gene trap cassette containing the ␤-gal coding sequence inserted after exon 1 of the Lgr4 gene). Lac Z staining of TRAMP/Lgr4 ϩ/Ϫ mouse prostates at various stages of cancer progression was examined to elucidate the Lgr4 expression pattern during prostate cancer progression (supplemental Fig.   Lgr4 loss blocks prostate cancer EMT and metastasis S1B). Consistent with our earlier description (27), in adult TRAMP mouse prostates at 8 weeks of age, Lgr4 is expressed in the basal epithelial layer (supplemental Fig. S1B, first column). Luminal epithelial expression of Lgr4 was first observed in the dorsolateral prostate (DLP) at the age of 16 weeks (supplemental Fig. S1B, second column) and in the anterior prostate (AP) from 20 weeks of age (supplemental Fig. S1B, third column). Before 12 weeks of age, Lgr4 ablation in the TRAMP model delayed PIN initiation ( Fig. 2A, first and second columns). We found a transient decrease in cell proliferation of prostates lacking Lgr4 at early ages that reached the level of wild-type mice by 12 weeks (Fig. 2, C and D) but no change in apoptosis (supplemental Fig. S2A) in mice lacking Lgr4. As suggested by our previous study (27), disruption of prostate luminal cell differentiation, function, and lumen secretion was found at early developmental stages in Lgr4 knockout mice. This may account for the temporary reduction in proliferation of Lgr4 Ϫ/Ϫ prostate cells. However, no significant difference in tumor formation, as examined by pathological analysis of PIN, WD/MD, and PD areas was found in TRAMP/Lgr4 Ϫ/Ϫ mice after 12 weeks (Fig. 2, A and B), suggesting that Lgr4 is important for early prostate development but does not affect tumor formation or cell proliferation at later ages.
Because metastasis is the predominant cause of prostate cancer mortality, we evaluated tumor metastasis in TRAMP mice. Mice were matched for primary tumor size and sacrificed at 25, 30, or 33 weeks to evaluate lung metastasis formation (n ϭ 6/group/time point). By 25 weeks, tumor cells had metastasized to the lungs in all TRAMP/Lgr4 ϩ/ϩ mice, whereas no evidence of tumor lung metastasis was found in TRAMP/Lgr4 Ϫ/Ϫ mice ( Fig. 3A, left column). At 30 and 33 weeks, macroscopic metastases were found in wild-type TRAMP lungs (24.3 Ϯ 12.3 metastases/mouse, n ϭ 10), whereas there was minimal metastasis formation in TRAMP/Lgr4 Ϫ/Ϫ lungs (10.7 Ϯ 7.4 metastases/mouse, n ϭ 10, mean Ϯ S.D., p Ͻ 0.01; Fig. 3, A and B). We performed cytokeratin 18 (CK18) immunofluorescence on 30and 33-week-old TRAMP mouse lungs to confirm the epithelial identity of metastatic tumor cells (Fig. 3C). CK18 ϩ adenocarcinoma cells were greatly reduced in 30-and 33-week-old TRAMP/Lgr4 Ϫ/Ϫ lungs. Finally, we evaluated the consequences of Lgr4 loss on TRAMP mouse survival. Lgr4 ablation in TRAMP mice significantly improved the survival rate by 40% at 1 year of age and prolonged the median life span from 36 weeks (TRAMP/Lgr4 ϩ/ϩ ) to 48 weeks (TRAMP/Lgr4 Ϫ/Ϫ ) (Fig. 3D). Lgr4 heterozygosity did not significantly affect TRAMP tumor formation or survival, indicating that a single allele of Lgr4 is sufficient to allow tumor development, whereas loss of both Lgr4 alleles significantly improved survival from prostate cancer. Together, these results indicate that Lgr4 is a crucial regulator of tumor metastasis.

LGR4 silencing in DU145 cells impairs cell migration and invasion without affecting cell proliferation
To explore the mechanism underlying Lgr4 regulation of cell metastasis and invasion, we established DU145 cell lines with stable Lgr4 knockdown using two different LGR4 shRNAs. Lgr4 shRNA-1 and -2 reduced the LGR4 mRNA level in DU145 cells by 75% and 60% and the protein level by 90% and 80%, respectively (Fig. 4A). To determine the effect of LGR4 on cell proliferation, we compared cell growth between non-silencing con- LGR4 is overexpressed in human prostate cancer and positively correlated with cancer recurrence. A and B, LGR4 expression was assessed by qPCR (A) and Western blotting (n ϭ 3) (B) in different human prostate cancer cell lines. C, the LGR4 mRNA level was analyzed using GEO accession no. GSE6099 (33). p ϭ 0.007. D, the Kaplan-Meier disease-free survival curve was generated using GEO accession no. GSE3933 (34). LGR4 low , n ϭ 28; LGR4 high , n ϭ 30. E, the Kaplan-Meier disease-free survival curve was generated by using GEO accession no. GSE10645 (35). p Ͻ 10EϪ6; LGR4 low , n ϭ 59; LGR4 high , n ϭ 60. The data in A are expressed as means Ϯ S.E. of three independent experiments (*, p Ͻ 0.05; **, p Ͻ 0.01; versus PNT1A control).

Lgr4 loss blocks prostate cancer EMT and metastasis
trol shRNA and Lgr4 shRNA knockdown DU145 cells. We found no significant effect of LGR4 down-regulation or knockout by CRISPR-Cas9 on DU145 cell proliferation ( Fig. 4B and supplemental Fig. S3A). When we analyzed DU145 cells by flow cytometry, we were surprised to observe a decreased forward scatter in cells expressing LGR4 shRNA, suggesting a potential role for LGR4 in cell size regulation (Fig. 4C). As expected from our in vivo data, LGR4 inactivation dramatically impaired DU145 cell migration and invasion.
LGR4 knockdown repressed DU145 cell migration by 70% (Fig. 4D) and trans-well invasion by 60 and 75% (Fig. 4E). Three-dimensional Matrigel colony growth was used previously to study prostate cancer invasion. Prostate cancer cell spheroids form either a round phenotype (poor invasiveness), a "mass" phenotype with occa-sional invasive filopodia, or invasive/stellate structures that express EMT-related mesenchymal markers (highly invasive). Thus, in this study, we used 3D Matrigel colony formation to study the morphological change and invasive potentials in the colonies (36). LGR4 silencing dramatically reduced the frequency of 3D Matrigel colony formation (from 18.7% Ϯ 2.2% to 5.5% Ϯ 1.5% (shRNA-1) and from 15.2% Ϯ 1.4% to 6.8% Ϯ 1.1% (shRNA-2) of total cells seeded) (Fig. 4F). Interestingly, we found that DU145 spheroids expressing control shRNA showed occasional invasive filopodia typical of the mass phenotype at the colony edge, whereas colonies containing LGR4 shRNA-1 maintained a less invasive "round" morphology, suggesting that LGR4 silencing might affect the invasive properties of DU145 cells.

Lgr4 loss blocks prostate cancer EMT and metastasis
We further assessed the impact of LGR4 on prostate cancer cell proliferation, migration, and invasion by ectopic expression of LGR4 in LNCaP cells. To generate LGR4-overexpressing LNCaP cells, we transduced LNCaP cells with a lentiviral expression vector containing CMV-promoter driven FLAGtagged LGR4 followed by IRES-GFP (where IRES is an internal ribosomal entry site) to visualize LGR4 overexpressing cells (Fig. 4G). Consistently, LGR4 overexpression did not affect LNCaP cell proliferation (data not shown) but increased LNCaP cell migration by 40%, cell invasion by 2-fold, and nearly doubled the frequency of 3D Matrigel colony formation (from 6.5% Ϯ 1.0% to 12.6% Ϯ 1.2%) (Fig. 4, G-J). Altogether, our in vivo and in vitro findings support a key role for Lgr4 in regulating human prostate cancer cell invasion and metastasis.

LGR4 regulates tumor metastasis through EMT
To elucidate candidate downstream effectors of LGR4 in DU145 cells, reverse-phase protein array analysis was carried out on DU145 cells stably expressing non-silencing control shRNA or Lgr4 shRNA-1. Proteins or phosphoproteins whose levels changed more than 50% in DU145-Lgr4 shRNA cells compared with the average of untransfected control and nonsilencing DU145 cells were considered potential targets of LGR4 and are listed in supplemental Table S1. In summary, LGR4 knockdown affected regulators of cell migration and invasion, including focal adhesion kinase (FAK) and PKC, the EMT modulators E-cadherin and snail, the Wnt pathway component ␤-catenin, caveolin-1, and c-Myc, supporting our find-ing that LGR4 down-regulation reduced DU145 cell migration and invasion.
LGR4 is known as a potentiator of Wnt pathway signaling. Not surprisingly, LGR4 silencing in DU145 cells attenuated the expression of Wnt target genes, including c-Myc, Cyclin D1, and MMP3 (Fig. 5A). The mRNA levels of the EMT-inducing transcription factors ZEB and Twist were also decreased by LGR4 inactivation (Fig. 5B). We also found attenuated Snail protein expression together with elevated E-cadherin, as expected for a decrease in EMT, in LGR4-silenced DU145 cells (Fig. 5C). Interestingly, the level of ␤-catenin was also elevated in DU145-Lgr4 shRNA cells. Immunofluorescence data showed that ␤-catenin and E-cadherin co-localized on the cell membrane in DU145-Lgr4 shRNA cells, indicating ␤-catenin sequestration in a complex with E-cadherin that forms cell-cell contacts and prevents ␤-catenin nuclear translocation and transcriptional activity (Fig. 5D).
p63 is a canonical prostate basal cell marker (37). p63-expressing cells have been shown to possess stem cell properties in that p63-positive basal cells are capable of luminal transdifferentiation (38). Poorly differentiated prostate cancer is marked by the loss of basal cells (39). 25-week-old TRAMP/ Lgr4 Ϫ/Ϫ mouse prostates had a higher proportion of p63-positive cells (Fig. 5E, top row) with distinct cell shapes from other basal cells, suggesting that TRAMP/Lgr4 Ϫ/Ϫ mouse prostates are more heterogeneous than their wild-type littermates. Ecadherin expression in the wild-type TRAMP prostate diminished as cancer progressed but remained high in TRAMP/ Lgr4 Ϫ/Ϫ prostates, (Fig. 5E, bottom row), indicating that Lgr4 disruption reduced tumor EMT. At the same time, as shown by ␣ smooth muscle actin (␣-SMA) immunohistochemistry staining, the smooth muscle layer of TRAMP/Lgr4 Ϫ/Ϫ prostates was relatively intact compared with wild-type TRAMP mice and, thus, potentially more effective at preventing cancer cell invasion outside of the prostate gland. Vimentin is another marker for prostate cancer cell EMT (40), so we examined the protein level of Vimentin upon LGR4 silencing. Fluorescent staining and Western blot analysis showed that LGR4 silencing led to decreased Vimentin expression (Fig. 5, F and G), although to a lesser extent than other EMT proteins. Moreover, a reduction of Snail and Vimentin and a dramatic increase of E-cadherin and ␤-catenin were consistently observed in LGR4 knockout DU145 cells (supplemental Fig. S4A). We next examined the level of EMT proteins in LGR4-overexpressing LNCaP cells. Upon LGR4 overexpression (indicated by green fluorescent cells marked with asterisks), both E-cadherin and ␤-catenin levels were reduced compared with WT/control cells or GFP-negative cells (Fig. 5, H and I). Notably, ␤-catenin in LNCaP cells predominantly localized at the plasma membrane, indicating the central role of ␤-catenin anchored to E-cadherin in cellcell adhesion. In contrast, LGR4 overexpression reduced E-cadherin, dissociating the E-cadherin-␤-catenin complex at the plasma membrane, leading to degradation of ␤-catenin (Fig. 5J). We also found elevated Snail and Vimentin expression in LGR4-overexpressing LNCaP cells (Fig. 5J). Together, these data strongly suggest that Lgr4 functions as a positive regulator of EMT in prostate cancer, and therefore loss of Lgr4 impaired tumor metastasis.

Lgr4 loss blocks prostate cancer EMT and metastasis
R-spondins 1-4 (Rspo1-4) have been identified as ligands of LGR4 -6 that potentiate Wnt pathway signaling (12,13). Rspo2 is highly expressed in the male mouse urogenital tract and regulates prostate development as well as prostate cancer progression (41). To investigate whether Lgr4 is a key mediator of Rspo2 to promote Wnt/␤-catenin signaling in prostate cancer cells, we treated DU145-Lgr4 shRNA or DU145-control shRNA cells with 10 ng/ml Rspo2. Rspo2 had minimal effects on control cell E-cadherin or Snail and modestly increased ␤-catenin protein levels, in accord with the known role of Rspo in potentiating Wnt signaling (Fig. 5K). Rspo2 was able to partially reverse the down-regulation of Snail observed in DU145-Lgr4 shRNA cells and reversed the increase in E-cadherin and membrane ␤-catenin because of LGR4 knockdown. Similarly, treatment with the GSK3-␤ inhibitor TWS119 (5 M) for 6 h significantly rescued the protein level of Snail in DU145-Lgr4 shRNA cells (Fig. 5L). E-cadherin levels decreased in TWS119-treated DU145-Lgr4 shRNA cells,

Lgr4 loss blocks prostate cancer EMT and metastasis
suggesting a restoration of Snail-mediated repression, and we also observed a decrease in ␤-catenin levels, consistent with increased ␤-catenin turnover following release from membrane-associated E-cadherin-␤-catenin-␣-catenin complexes in DU145-Lgr4 shRNA cells (42). Thus, Rspo2/LRG4/ Wnt signaling is important for regulating the E-cadherin-␤-catenin complex during EMT.

LGR4 silencing in DU145 cells delays tumor metastasis in a xenograft mouse model
To investigate in vivo LGR4 function in a human prostate cancer cell line, we injected 1 ϫ 10 5 DU145 cells expressing Lgr4 shRNA or control shRNA into NSG mouse prostates and measured the resulting xenograft formation. Twelve weeks
We did not observe significant differences in cell proliferation or apoptosis between control and LGR4 knockdown DU145 xenograft tumors (Fig. 6, E and F). More importantly, prostate

Lgr4 loss blocks prostate cancer EMT and metastasis
tumor metastasis to the lungs was greatly reduced by LGR4 inactivation (Fig. 6, G and H). The number of metastatic foci in the lungs of mice orthotopically injected with DU145 cells was reduced from 29 Ϯ 8.5 (control) to 3 Ϯ 1.4 (Lgr4 shRNA) (n ϭ 6, mean Ϯ S.D., Fig. 6G). Human prostate cancer cell metastasis and LGR4 knockdown was confirmed by staining with an antibody against human LGR4 in the lungs (Fig. 6H, right column). Cell proliferation in metastatic tumor foci was reduced by LGR4 inactivation (Fig. 6H, center column). Importantly, xenograft tumors of Lgr4 shRNA-expressing cells had high E-cadherin expression, increased membrane ␤-catenin, and low vimentin expression, suggesting that EMT is reduced upon LGR4 knockdown (Fig. 6I). Moreover, Lgr4 shRNA-expressing prostate tumor cells expressed the epithelial marker K18, further supporting reversal of EMT following LGR4 knockdown. Macrophage infiltration, shown by F4/80 staining, was also reduced because of LGR4 inactivation (Fig. 6I). Taken together, these data suggest that LGR4 modulates tumor metastasis by promoting EMT.

Discussion
Loss of Lgr4 had multiple effects on prostate cancer development. In young TRAMP mice, Lgr4 inactivation delayed prostatic epithelial hyperplasia and diminished cell proliferation in 6-week-old DLP and 8 week-old AP (Fig. 2, A-D) because of impaired luminal cell function at early stages of prostate development upon Lgr4 loss (27). However, in older mice, we observed similar lesion loads, suggesting that the cancer progression delay in TRAMP/Lgr4 Ϫ/Ϫ mice was transient (Fig. 2, A  and B). Although prostate cancer initiation was similar in TRAMP mice at 12 weeks, Lgr4 deletion delayed tumor metastasis (Fig. 3, A-C) and prolonged the life span (Fig. 3D). LGR4 expression in human prostate cell lines generally correlated with invasive and metastasis potential (Fig. 1, A and B), and LGR4 down-regulation inhibited DU145 migration and invasion (Fig. 4, C and D), whereas LGR4 overexpression promoted LNCaP migration and invasion (Fig. 4, H and I). Finally, LGR4 expression in human prostate cancers positively correlated with cancer recurrence (Fig. 1, D and E). Together, these results strongly indicated that Lgr4 may function in tumor invasion and metastasis, in accord with previous publications (29,31,32).
EMT is a driver of prostate cancer metastasis (2,5,6,43). Loss or aberrant expression of the epithelial marker E-cadherin, as occurs following EMT, is correlated with a poor prognosis in prostate cancer patients (43,44). Nuclear localization of the EMT-inducing transcription factor Snail correlated with a higher Gleason score, tumor stage, and likelihood of relapse (43,45,46). We found that loss of Lgr4 resulted in elevated E-cadherin levels in vitro and in vivo (Fig. 5, A-C, and supple-mental Table S1), suggesting a reversal of EMT. This was accompanied by decreased expression of the EMT-inducing transcription factors ZEB, Twist, and Snail and the EMT marker Vimentin. These results suggest that Lgr4 is a master regulator of prostatic adenocarcinoma EMT and, thus, regulates cancer metastasis.
Androgen signaling plays a key role in prostate cancer development, and targeting this pathway is the standard first-line therapy for prostate cancers. The relationship between LGR4 and androgen signaling is complex. Humans carrying a missense LGR4 allele have lower testosterone levels (26), and mouse Lgr4 knockout models have defects in the male reproductive system, including loss of androgen receptor (AR) expression in peritubular myoid cells (47). In the mouse prostate, Lgr4 is expressed in basal cells, whereas AR expression is restricted to the luminal epithelium (27). Lgr4 Ϫ/Ϫ mice had impaired AR ϩ luminal cell development; however, AR signaling was operational in adult knockout mice, as both WT and Lgr4 Ϫ/Ϫ mice exhibited prostate regression upon castration. Upon androgen replenishment, CK8 ϩ luminal cells failed to differentiate in Lgr4 Ϫ/Ϫ prostates, suggesting that Lgr4 might indirectly regulate AR (27). Curiously, forced LGR4 expression increased AR levels in prostate cancer cell lines in a manner dependent on the reported LGR4 target JMJD 2A (48,49). Here, we found decreased disease-free survival in prostate cancer patients with high LGR4 mRNA levels but no correlation between tumor LGR4 and AR mRNA levels (data not shown), suggesting that LGR4 functions in prostate cancer independent of AR status; however, our results do not rule out LGR4 regulation of AR in prostate cancer. Unraveling the connections between LGR4 and AR is an important area for further research.
Finally, our results lead to the speculation that Lgr4 may be involved in cancer stem cell (CSC) regulation. Prostate CSCs possess the capacity to self-renew and differentiate into the heterogeneous lineages of cancer cells within a tumor. Prostate CSCs employ the signaling pathways used for maintenance of normal prostate stem cells, such as Pou5F1 (Oct3/4), Nanog, Sox2, and c-kit. Furthermore, CSCs are proposed to be the origin of prostate cancer recurrence (50,51). Hedgehog signaling may induce a transitory differentiation of prostatic stem/progenitor cells into CD44 ϩ /p63 Ϫ/ϩ hyperplastic basal cells with an intermediate phenotype (CK8/14) (52), potentially implicating Hedgehog in prostate CSCs during early cancer development. We have previously reported impaired SHH (sonic hedgehog) signaling in prostate cells of Lgr4 Ϫ/Ϫ mice (27). Wnt signaling, another downstream target of Lgr4, can promote prostate cancer sphere self-renewal (53), suggesting that Wnt signaling is a driver of prostate CSCs. On the other hand, CARN

Lgr4 loss blocks prostate cancer EMT and metastasis
(castration-resistant Nkx3.1-expressing) cells, which have been shown to be the cells of origin of prostate cancer (54), are deficient in Lgr4 Ϫ/Ϫ prostates (27). The fact that Lgr4 inactivation compromises Wnt as well as SHH signaling and impedes p63 high cell differentiation into p63 Ϫ/low cells implies a potential role of Lgr4 in prostate CSCs and is an area worthy of further investigation. In conclusion, we provide evidence that Lgr4 regulates prostate cancer metastasis at least in part through modulating expression of the EMT transcription factor Snail (supplemental Fig. S5).

Animals
All experiments using mice were performed in accordance with a protocol approved by the Texas A&M Health Science Center Institutional Animal Care and Use Committee. Lgr4null mice were generated from an Lgr4 gene trap ES cell clone (LST020) from William Skarnes (Bay Genomics) as described previously (20). Original Lgr4 ϩ/Ϫ mice on a 129ϫC57/BL6 background were crossed with male TRAMP mice to generate TRAMP/Lgr4 ϩ/Ϫ mice, which were then backcrossed with Lgr4 ϩ/Ϫ mice to generate TRAMP/Lgr4 Ϫ/Ϫ . NSG (NOD Cg-Prkdc scid Il2rg tm1Wjl /SzJ) mice were purchased from The Jackson Laboratory (Bar Harbor, ME).

Lentivirus production and infection
The control and Lgr4 shRNAs were described previously (17). Bacteria containing the Lgr4 shRNA plasmid was purchased from Open Biosystems (Thermo Fisher Scientific Inc., Rockford, IL). Plasmid purification was conducted using the PowerPrep HP Plasmid Midiprep system (OriGene USA, Rockville, MD). We designed human LGR4 single guide RNA (sgRNA) using the sgRNA method described in Ref. 55 using an online CRISPR design tool (crispr.mit.edu) 6 by entering the targeted exon sequence. The LGR4 sgRNA with the highest score and lowest off-target effect (targeting sequence CTGCGACG-GCGACCGTCGGG) was cloned into the BsmB1 site of the lentiCRISPR vector containing Cas9-P2A-puromycin and was verified by sequencing analysis.
293FT cells (Life Technologies) were used to produce the lentivirus. Co-transfection of the lentiviral plasmids pMD2.G and psPAX2 (Addgene, Cambridge, MA) was performed using the CalPhos TM mammalian transfection kit (Clontech Laboratories, Inc., Mountain View, CA). 48 and 72 h after transfection, virus-containing cell culture supernatant was collected and passed through a 0.45-m filter (EMD Millipore Corp., Billerica, MA). For DU145 and LNCaP (ATCC), cells were seeded at 70% confluence 1 day before infection. 5 ml of virus-containing cell culture supernatant with 8 g/ml Polybrene (EMD Millipore Corp.) was added to each 10-cm dish for 3 h before adding 5 ml of complete minimal essential medium containing 10% FBS. Virus infection was repeated 24 h later. 48 -72 h after the first infection, cells were harvested and seeded in 96-well plates at 1 cell/well for single colony selection. Lgr4 knockdown efficiency was assessed by quantitative PCR and Western blot.

Proliferation assay
Cells were serum-starved overnight in MEM containing 0.5% FBS. Starved cells were seeded in 12-well plates at 1 ϫ 10 4 cells/well with complete MEM containing 10% FBS. Cells were counted every 24 h over the following 6 days, in triplicate. Complete MEM was changed every other day.
Wound-healing assay 1 ϫ 10 6 cells were plated and cultured in 6-well plates until they reached 100% confluence, followed by overnight starvation in medium containing 0.5% FBS. Starved cells were pretreated with mitomycin C (Sigma) at 20 g/ml for 4 h. A wound was made in confluent cells by vertical and horizontal scratching. Cells were then washed with PBS twice, and migration was induced by complete medium containing 10% FBS. Cell migration was checked every 12 h until the wound healed in one well. Cells were fixed in 4% paraformaldehyde.

Trans-well invasion assay
Growth factor-reduced Matrigel (BD Biosciences) was mixed with serum-free MEM at a ratio of 1:4. 100 l of the mixture was added to the bottom of cell culture inserts (8-m pores) (BD Biosciences). Inserts were incubated at 37°C for 1 h.
Cells were serum-starved overnight in MEM containing 0.5% FBS and then pretreated with mitomycin C (Sigma) at 20 g/ml for 4 h. Starved cells were harvested and resuspended at 1 ϫ 10 5 cells/ml in MEM containing 0.5% FBS. 100 l of cell suspension was seeded into Matrigel-coated cell culture inserts. 500 l of complete MEM containing 10% FBS (chemoattractant) was added to the lower chamber.
Cell invasion was checked every 12 h until a significant amount of cells invaded through the 8-m pores (48 h in most experiments). Non-invasive cells were removed, and invasive cells were fixed by immersing the cell culture inserts in 4% paraformaldehyde for 15 min at room temperature. Cell culture inserts were then washed in PBS. Invading cells were stained with 0.05% crystal violet (Sigma) in distilled water for 30 min. After two washes in PBS, invaded cells were visualized under an Olympus IX70 microscope, and three randomly chosen fields per membrane were photographed and quantitated using ImageJ software.

Soft agar assay and 3D culture
Cell transformation was determined by anchorage-independent growth in soft agar as described previously (57). Briefly, equal volumes of 1% agar melted in PBS and 2ϫ RPMI 1640 with 20% FBS at 40°C were mixed. 1 ml/well of the mixture (containing 0.5% agar) was added into 6-well plates as the base agar. Then, 5 ϫ 10 3 cells/well were suspended in 1 ml of Top Agar (a RPMI 1640 -agar mixture containing 0.35% agar) and added to each well. Top Agar was covered with 1 ml of culture medium, and the medium was replaced with fresh medium every 3 days. Plates were incubated at 37°C and 5% CO 2 for 4 weeks, and colonies were stained with 0.005% crystal violet, photographed, and quantified. 3D culture was done as described previously (58). Briefly, prechilled 24-well plates were coated with a thin layer of BD Matrigel TM basement membrane matrix and incubated for 30 min at 37°C. Cells were suspended in diluted Matrigel (1:4 dilution in RPMI 1640 medium with 10% FBS), and 2 ϫ 10 3 cells/well were added to each well and incubated for 30 min at 37°C. 500 l of culture medium was then added, and the culture was maintained for 10 days, with medium changed every 2 days. Colony formation was then photographed and quantified.

Quantitative real-time PCR (qPCR) analysis
Total RNA was isolated from cells or tissues, and first-strand cDNA was generated from total RNA using oligo(dT) primers and reverse transcriptase II (Life Technologies). qPCR was performed using specific primers and the ABI Prism 7000 analyzer (Applied Biosystems) with the SYBR Green qPCR Super Mix Universal kit (Life Technologies). Target gene expression values were normalized to human GAPDH or mouse Gapdh. The primers used for qPCR are listed in supplemental Table S2.

Xenograft
Orthotopic prostate injection of 8-week-old male NSG mice with DU145 cells was performed. Briefly, mice were anesthetized, and an abdominal incision was made to expose the prostate. 1 ϫ 10 5 cells in 50 l of RPMI 1640 with 10% FBS were mixed with an equal volume of pathogen-free BD Matrigel TM basement membrane matrix and injected into the prostate in each of two anterior lobes. Tumor growth was monitored by palpation. At 12 weeks, mice were sacrificed and necropsied to remove tumors and organs potentially containing metastatic foci (lymph nodes, lung, etc.) for formalin fixation, paraffinembedding, and tissue analysis.

Statistics
Unless otherwise specified, statistical analysis was performed using Student's t test or Holm's test where multiple groups were compared. A p value of 0.05 or less was considered significant.