High Gli3 expression, especially t-Gli3, is associated with advanced prostate cancer

To investigate a potential key player among Gli-family proteins, which could serve as a promising candidate for targeting CRPC, we determined the expression of Gli-family proteins using human tissue microarray (TMA) comprising normal, normal adjacent tissue, PRAD (grade I, II, III, and IV), and malignant (unspecified stage). Interestingly, clinical PCa samples revealed significantly increased expression of Gli3 in all malignant stages compared to normal prostate tissue (Fig. 1A). While Gli1 exhibited high expression exclusively in stage II, Gli2 expression remained unchanged compared to normal prostate tissues. The characterization of Gli-family protein expression patterns in malignant clinical samples showed distinct localization of these proteins within the cytosolic and nuclear compartments. Gli1 predominantly resided in the cytosol, Gli2 showed nuclear accumulation, and Gli3 was observed in both cytosolic and nuclear compartments (Fig. 1A). The significantly elevated expression of Gli3 was noted in the nuclear region compared to the normal tissue. Conversely, the changes in Gli3 expression in the cytosolic region were non-significant compared to normal prostate tissues (Fig. S1A). The nuclear accumulation of Gli3 suggests a potential activation of Gli3-mediated signal transduction, contributing to the progression of PCa, especially in advanced stages. We further validated the functional significance of Gli3 localization (nuclear versus cytoplasmic) in human PCa using isotype controls alongside positive and negative tissue controls. We conducted an in silico analysis of Gli3 protein expression through the Human Protein Atlas (Fig. S1B, https://www.proteinatlas.org). Our IHC results revealed high nuclear Gli3 expression in malignant colon tissue, moderate nuclear Gli3 expression in malignant lung tissue, and very low nuclear Gli3 expression in malignant liver tissue (Figure S1C), reinforcing the findings from the Gli3 protein atlas. Consistently, PCa samples from the MYCOE PCa mouse model, which recapitulates the molecular features of human prostate tumors [25], also exhibited increased expression of Gli3, particularly in the nuclear compartment of both high-grade lesion and carcinoma prostatic tissues compared to age-matched normal prostatic tissues (Fig. 1B–D). Our IHC analysis showed mild Gli3 expression in the mouse prostate, higher levels in the spleen and kidney, and no expression in the pancreas and liver (Fig. S2A–B).

Fig. 1: Expression profile Gli-family proteins, with emphasis on t-Gli3 in advanced PCa.

A Quantitation-H score (left panel) and Representative images (right panel) from IHC showing Gli-family proteins, including Gli1, Gli2, and Gli3 in human TMA of PCa. B Schematic representation of MYCOE (Hi-MYC) mouse PCa model C IHC quantitation-H score (upper panel) and corresponding representative images demonstrate the expression of Gli3 in mouse MYCOE cancerous prostatic tissue, HGPIN lesions, and age-matched normal mouse prostatic tissue. D Representative IF images (left panel) and fluorescence intensity plot (right panel) of Gli3 in mouse MYCOE cancerous prostatic tissue and age-matched normal mouse prostatic tissue. E Gene expression of Gli-family proteins and regulator Sufu in human PCa cell lines using expression atlas (https://www.ebi.ac.uk/gxa/home). F Representative WB images show the expression of Gli proteins and Sufu invarious PCa cell lines. GAPDH was used as a loading control. G Representative IF images (left panel) and fluorescence intensity plot (right panel) of Gli1, Gli2, and Gli3 with α-tubulin in human CRPC cells (22Rv1). H The quantitation of intensity mean of nuclear vs. cytoplasmic expression was analyzed using ImageJ software.

Next, considering Sufu serves as the crucial interacting partner of Gli proteins, influencing their actions (GeneCards/STRING interaction network: Fig. S3A), we examined the expression of Gli proteins and Sufu in datasets from human PCa cell lines using an expression atlas. Intriguingly, Gli3 and Sufu exhibited higher expression levels than Gli1 and Gli2 in most PCa cell lines (Fig. 1E). To delve deeper, we looked at the protein expression of Gli1, Gli2, Gli3, and Sufu in various PCa cell lines (Fig. 1F). These PCa cells exhibit varying androgen responses, androgen receptor (AR) expression patterns, and distinct morphology (Fig. S3B). Gli1 and Gli2 expression levels were decreased, while Gli3 expression level was increased, and Sufu expression levels showed no difference in most of the PCa cell lines (Fig. 1F). Surprisingly, the expression of truncated-Gli3 (t-Gli3) was found to be increased in all PCa cell lines, with some showing abundant expression of both full-length (FL-Gli3~190 kDa) and t-Gli3 (~83 kDa) Gli3 proteins, such as PC3M and 22Rv1. Moving forward, experiments with mouse syngeneic cell lines demonstrated increased expression of both (full-length and truncated) forms of Gli3 in castrated PTEN conditional knockout mouse-derived cells (cE2) compared to non-castrated mouse-derived cells (E2) (Fig. S3C). These observations suggest the potential involvement of t-Gli3 in an advanced stage of PCa progression, including metastasis and CRPC.

The specific localization of Gli proteins in the sub-cellular compartment was confirmed using 22Rv1 human PCa cell line (resistant to enzalutamide and regarded as CRPC cell line [26, 27]) via immunofluorescence imaging with Gli1, Gli2, and Gli3 along with α-tubulin, a cytoskeletal protein (Fig. 1G). Nuclear localization of Gli proteins was represented via fluorescence intensity profile (Fig. 1G right panel). We also quantified the cytosolic vs. nuclear expression of Gli proteins (Fig. 1H). These cumulative results suggested that the cytosolic expression of Gli1, nuclear expression of Gli2, and Gli3 are consistent with the localization pattern observed in human TMA of PCa patients. Considering the expression pattern of Gli-family proteins, observation revealed that Gli3 might have a more potent role in the advanced PC context; however, further evidence is needed to validate this hypothesis.

Gli3 is modulated with androgen deprivation and treatment with AR agonist/antagonist

In the subsequent analysis, the investigation of sub-cellular localization of Hh signaling regulatory proteins, including Gli1, Gli2, Gli3, and Sufu, revealed increased expression of Gli3 (both cytosolic and nuclear) and Sufu (cytosolic) in androgen-independent (AI)/CRPC cell lines compared to androgen-dependent (AD) and normal prostate cell lines (Fig. 2A, B). Notably, Gli1 expression was reduced in AD cell lines but unchanged in AI/CRPC cell lines. In contrast, the Gli2 expression was increased in AD cell lines but decreased in AI/CRPC cell lines. To demonstrate the effect of androgen-deprived conditions on the Shh/Gli cascade, we utilized PCa cells expressing either AR-FL (C4-2B) or AR-FL/AR-V7 (22Rv1). Under the androgen-deprived condition, the expression of Gli1, Gli2, and Shh proteins showed no alterations, but AR and Gli3, exclusively t-Gli3 expression, were downregulated in 22Rv1 cells, but not in C4-2B cells (Fig. 2C). This indicates the involvement of the non-canonical Hh pathway, specifically ligand-independent Hh signaling, in response to ADT due to the presence of AR-V7 in 22Rv1. Further, the cytosolic and nuclear fractions revealed the diminished nuclear accumulation of AR-FL/AR-V7 (Fig. S4A). In contrast, under serum-free media (SFM) conditions, the upregulated expression of AR, Shh, Gli1, Gli2, and Gli3 was observed in 22Rv1 cells (Fig. 2C). These results support the involvement of the canonical Hh pathway, precisely the ligand-dependent Shh/Gli axis, in response to SFM-induced starvation. Considering the differential expression pattern of Gli3 in different AR-positive and negative PCa cell lines, we next investigated Gli3 activity in relation to AR signaling. We treated the normal prostatic epithelial cell line (RWPE1), the AR-FL-only expressing cell line (C4-2B), and the cell line co-expressing AR-FL and AR-V7 (22Rv1) with AR antagonists (ENZ-Enzalutamide), agonists (DHT - Dihydrotestosterone) or a combination of both and determined AR and Gli3 expression. We observed the upregulation of AR and Gli3 in RWPE1 and C4-2B cell lines treated with DHT, while treatment with ENZ or a combination led to their downregulation (Fig. S4B). No effect was seen in the 22Rv1 cell line, which is considered an ENZ-resistant cell line (Fig. S4B). Confocal microscopy provided additional evidence that aligned with the expression patterns seen in WB, reinforcing the notion that the presence of AR is crucial for sustaining Gli3 expression (Fig. S4C). Additionally, we observed increased Gli3 expression, along with downregulation of Gli1 and Gli2, and no changes in Shh in C4-2B ENZ-resistant cells compared to parental C4-2B cells, suggesting Gli3’s role in advanced disease progression (Fig. S4D).

Fig. 2: Sub-cellular expression profile of Gli-family proteins with the dependency of AR and ciliary localization of Hh receptors.

A, B Representative IF images for sub-cellular localization of Gli1/Gli3 and Gli2/ Sufu (left panel) and fluorescence intensity plot (right panel) in normal human prostate cells (RWPE1) and PCa cell lines (AD: LNCaP-35; AI: LNCaP-135 & 22Rv1). C Representative WB images depict the expression of Shh/Gli cascade proteins under ADT (C4-2B and 22Rv1) and SFM conditions (22Rv1). GAPDH was used as a loading control. D Representative IF images of Smo (left panel) and Ptch (right panel) in normal human prostate cells and PCa cell lines with the status of AD and AI. Arrowheads mark the localized proteins, red for nuclear and yellow for cytoplasmic area. E Representative WB images show the expression of Smo and Ptch receptors in normal human prostate cells RWPE1 and PCa cell lines (AD: LNCaP-35; AI: LNCaP-135 & 22Rv1). F A scatter plot depicting the correlation of Gli-family transcription factor (Gli1, Gli2, and Gli3), and Hh receptors (Smo and Ptch1) with AR in TCGA-PRAD patients using OncoDB. G Representative IF images for colocalization analysis of Smo and AR in PCa cell lines display AR presence (LNCaP-135 and 22Rv1) and absence (DU145), along with AI status. H Representative IF images for ciliary localization of Smo, Ptch, and Gli3, where cilia were examined using acetylated α-tubulin. Arrowheads mark the localization of proteins, whether colocalized or not.

Ciliary localization of Ptch1 is associated with Gli3 activation via a Smo-independent mechanism

Unraveling the canonical regulation of Gli3 protein, influenced by receptor-mediated activation of the Hh pathway, the expression level of Smo and Ptch in both AD and AI PCa cells along with normal prostate cells were examined. The predominantly membranous expression of Smo in PCa cells was observed in contrast to normal cells exhibiting Smo expression in both the cytosolic and nuclear regions (Fig. 2D). Intriguingly, Ptch displayed an inverse expression pattern localized in both cytosolic and nuclear regions in AD and AI PCa cells compared to normal prostate cells. As Ptch acts as a downstream target of Gli, the elevated expression of Ptch in PCa cells indicates that this overexpression could result from Gli3 activation. We observed a reduction in Smo expression and an elevation in Ptch expression in PCa cell lines compared to normal prostate cells (Fig. 2E). Furthermore, we performed a correlation analysis using OncoDB to examine the relationship between AR and Gli-family proteins (Gli1, Gli2, and Gli3) and receptors (Smo and Ptch1). Interestingly, we observed a positive correlation specifically between Gli3/AR (weak) and Ptch1/-AR (moderate) at the gene expression level of PRAD patients (Fig. 2F). In the same cohorts, no significant correlation was found between AR and Gli1 or Gli2 or Smo (Fig. 2F). This further emphasizes the potential relevance of Gli3 with functional activation of AR in the PCa. In support, our investigation into the association of Smo with AR also revealed no colocalization of Smo and AR in CRPC cells, suggesting a lack of direct association between these proteins (Fig. 2G). Notably, we found reduced Gli3 nuclear expression in DU145, which does not express AR, suggesting the nuclear translocation of Gli3 may be due to the presence of AR (Fig. 2G).

Further, we investigated the presence of Smo, Ptch, and Gli3 on cilia, which is a crucial determinant for receptor-mediated modulation of the Smo/Gli canonical axis. A subset of PCa cells exhibited cilia and the Ptch expression on the cilium. Strikingly, no expression of Smo and Gli3 on the cilium was observed (Fig. 2H). The absence of Smo on the cilium, coupled with the presence of Ptch, provides additional evidence supporting the involvement of a Smo-independent non-canonical mechanism. Additionally, the lack of Gli3 expression on the cilium and the presence in the nucleus suggest the role of Gli3 as a transcription factor activating Hh target proteins in CRPC cells (Fig. 2H).

Androgen receptor variant (AR-V7) is required for Gli3-mediated action

To understand the role of Gli3 for downstream target activation and its impact on CRPC, we analyzed the biochemical structures of Gli3 (FL-Gli3 and t-Gli3) and AR (AR-FL and AR-V7), which led to an intriguing observation (Fig. 3A). FL-Gli3 possesses a transactivation domain, whereas t-Gli3 lacks this domain. The elevated AR expression was colocalized with increased Gli3 expression (Fig. 3B) with a Pearson coefficient of 0.6 (Fig. 3C), indicating a high correlation between AR and Gli3 in MYCOE prostatic tissue (Fig. 3C). A prominent colocalization of AR and Gli3 was observed in the LNCaP-135 and 22Rv1 cell lines expressing both AR-FL and AR-V7 variants (Fig. 3D). However, there was no Gli3 localization in the nucleus of C4-2B cells, expressing only AR-FL and the DU145 cells with the lack of AR expression. In addition, our colocalization analysis inspected correlation coefficients in 22Rv1 (r = 0.6) and LNCaP-135 (r = 0.7) showed a strong correlation between both proteins (Fig. 3E). Moreover, analysis of quantitative mean fluorescence intensity suggested Gli3 expression is positively regulated by AR expression, where more pronounced AR and Gli3 expression in 22Rv1 and LNCaP-135 as compared to C4-2B, suggesting that AR-V7 might be a regulator of FL-Gli3 and t-Gli3 (Fig. 3F).

Fig. 3: Colocalization of AR and Gli3 and an association between AR-V7 and t-Gli3.

A Schematic representation of biochemical domains of FL-Gli3, t-Gli3, AR, and AR-V7 and hypothesized possible signaling mechanism. B Representative IF images (upper panel) and fluorescence intensity plot (lower panel) of Gli3 and AR in mouse MYCOE cancerous and normal prostatic tissues. C Colocalization analysis using ImageJ software using plugins (colocalization finder and JACoP). D Representative IF images (left panel) and fluorescence intensity plot (right panel) in normal and PCa cell lines, including LNCaP-135 and 22Rv1 expressing AR-FL/AR-V7, C4-2B expressing AR, and DU145 expressing no AR. E Colocalization analysis using ImageJ software using plugins (colocalization finder and JACoP). F The intensity mean of AR and Gli3 was quantified using ImageJ software. G Representative WB image of Gli1, Gli2, and Gli3 after AR enrichment via IP in LNCaP-135, 22Rv1, and C4-2B cells H Representative WB image of Gli3 and AR after Gli3 enrichment via IP in 22Rv1 and C4-2B cells.

To investigate the potential interaction between t-Gli3 and AR-V7, we performed a co-immunoprecipitation (co-IP) assay with AR enrichment using cell lines expressing both AR-FL and AR-V7 (LNCaP-135 and 22Rv1) and the AR-FL-only expressing cell line C4-2B (Fig. 3G). Our findings revealed an interaction between t-Gli3 and AR-V7 in LNCaP-135 and 22Rv1 cells (Fig. 3G left panel). Interestingly, no interaction was observed in C4-2B cells lacking AR-V7 (Fig. 3G right panel). This result suggests a selective interaction between t-Gli3 and AR-V7 in CRPC cells, further validated by performing a reverse co-IP approach (Fig. 3H). This additional evidence strengthens the observation of a specified interaction between t-Gli3 and AR-V7, reinforcing the significance of this interaction in the context of CRPC.

Gsk3β- mediated non-canonical mechanism involved in Gli3 processing

The biochemical structure of Gli3-FL reveals the presence of a Gsk3β binding site within the activator sequence (Fig. 4A). The expression of Gsk3β is significantly upregulated in PCa patients (GTEx and TCGA) (Fig. 4B). Moreover, a significant positive correlation was observed between Gsk3β/AR (strong) and Gsk3β/Gli3 (moderate) in TCGA-PRAD patients (Fig. 4C). We observed distinct expression patterns of Gsk3β and p-Gsk3β across different PCa cell lines based on their respective AR statuses (Fig. 4D). Furthermore, we investigated the sub-cellular expression of Gsk3β and p-Gsk3β in normal, AD, and AI PCa cell lines. Gsk3β was found to be localized in both cytosolic and nuclear regions of PCa cells (Fig. 4E left panel), whereas p-Gsk3β exhibited predominant nuclear localization (Fig. 4E right panel). In contrast, RWPE1 displayed both cytosolic and nuclear expression of Gsk3β and p-Gsk3β. Importantly, LNCaP-135 and 22Rv1 showed higher expression levels of Gsk3β and p-Gsk3β than LNCaP-35 and RWPE1 cells, suggesting their potential relevance in CRPC. Similarly, Gsk3β expression in human TMA significantly increased in PRAD patients (Fig. S5A).

Fig. 4: Gsk3β- mediated non-canonical mechanism for Gli3 processing.

A Schematic representation of biochemical domains of Gli3 consisting of the post-translational binding site and proposed mechanism. B Gsk3β expression level in primary tumor vs. normal prostate human samples (GTEx and TCGA). C A scatter plot of correlation analysis of Gsk3β/AR and Gsk3β/Gli3genes using in silico approach in TCGA-PRAD patients using OncoDB. D Representative WB images show the expression of Gsk3β and p-Gsk3β in normal human prostate cells and various PCa cell lines. GAPDH was used as a loading control. E Representative IF images showing the expression of Gsk3β (left panel) and p-Gsk3β (right panel) in normal human prostate cells RWPE1 and PCa cell lines (AD: LNCaP-35; AI: LNCaP-135 & 22Rv1). F Cell viability assay using Gsk3β-activator and inhibitors in AD and AI PCa cell lines. Values are expressed as mean ± SEM (n = 5). G Representative IF images (left panel) and fluorescence intensity plot (right panel) showing sub-cellular expression Gsk3β/Gli3 (upper panel) and p-Gsk3β/Gli3 (lower panel) in 22Rv1 cells after treating Gsk3β-activator. H Representative WB images show the expression of p-Gsk3β, Gsk3β, and Gli proteins in 22Rv1 cells treated with Gsk3β-activator. GAPDH was used as a loading control.

To explore the functional relevance of Gsk3β in CRPC, we employed a selective Gsk3β activator (LY-294002) and inhibitor (AR-A014418) and investigated the impact of Gsk3β on PCa cells viability (Fig. 4F). Interestingly, we observed a substantial decrease in cell survival when Gsk3β was inhibited. Surprisingly, activating Gsk3β in the AI cell lines (LNCaP-135 and 22Rv1) did not impact cell viability, whereas AD cells (LNCaP-35) showed effect after treatment. To explore the Gsk3β/Gli3 axis, using Gsk3β activator, we observed the significant effect on Gsk3β activation, leading to the dephosphorylation of p-Gsk3β at 2 h and 6 h time points with 10 µM concentration (Fig. S5B). Using confocal microscopy, we noticed the increased expression of both Gsk3β and Gli3, with a predominant nuclear translocation, in Gsk3β activator- treated cells (Fig. 4G upper panel). At the same time, p-Gsk3β expression was decreased and localized only in the nucleus (Fig. 4G lower panel). Next, we observed an increase in Gli3 expression, particularly t-Gli3, after 2 h, and both the full and truncated forms of Gli3 after 6 h (Fig. 4H). These findings suggest that the activity of Gsk3β contributes to the generation of the t-Gli3 protein.

Gsk3β-knockdown induces an anti-proliferative response in CRPC Cells

To investigate the Gsk3β mediated generation of t-Gli3, we first generated Gsk3β knockdown (KD)- 22Rv1 cells by utilizing a human shRNA lentiviral particle system (Fig. S6A and S6B) and determined the KD efficiency (Fig. S6C, D, and 5A). To elucidate the functional impact of the reduction in Gsk3β expression, we examined the cell viability assay (Fig. 5B) and noted a significant suppression in growth on the third day. This data was further validated by analyzing real-time growth kinetics using an incucyte system (Fig. S7A). Additionally, we examined the cellular morphology through fluorescence imaging and observed distinct morphological alterations, including merged and enlarged structures, compared to the scrambled group (Fig. S7B). Subsequently, we observed fewer colonies in Gsk3β KD cells than in the scrambled control and parental cells (Fig. 5C). Using a 3D/spheroids culture system, we observed a reduction in spheroid size with Gsk3β KD (Fig. 5D). We quantified the spheroid object area (Fig. 5D upper panel) and darkness (Fig. 5D lower panel) and found a decrease in the object area along with an increase in darkness, indicating both growth suppression and enhanced dead populations. However, we did not find any phase cell cycle arrest between scramble and Gsk3β KD (Fig. S7C). The Gsk3β KD significantly decreased tumor growth and volume in the PCa xenograft mice model (Fig. 5E). Additionally, Gene Set Enrichment Analysis (GSEA) was performed using the RNA-Seq of 22Rv1-Gsk3β KD vs. scramble cells and selected altered ten major upregulated and downregulated HALLMARK GSEAs out of the top 20 for each group, which are related to oncogenic and metabolomic shift (Fig. S7D). We noticed that PI3K/AKT/mTOR signaling pathway in the top 20 enriched GSEAs in scramble cells compared to Gsk3β KD cells (Fig. 5F). This finding was substantiated by the decrease in p-AKT protein expression observed after 72 h in Gsk3β KD cells, aligning with the observed inhibitory growth kinetics in Gsk3β KD cells compared to the scramble (Fig. 5G). In summary, the outcomes derived from diverse assays collectively emphasize the pivotal role of Gsk3β in regulating the cellular proliferation of CRPC cells.

Fig. 5: Knocking down of Gsk3β triggers an anti-proliferative response in CRPC Cells.

A Representative WB images show the expression of Gsk3β in 22Rv1- Gsk3β KD vs. scramble cells. GAPDH was used as a loading control. B Cell viability assay demonstrates the growth pattern of 22Rv1- Gsk3β KD vs. scramble cells in a time-dependent manner. Values are expressed as mean ± SEM (n = 5). Statistical significance was defined as P ≤ 0.05. For group comparisons, *P < 0.05, **<0.01, ***<0.001 and ****P < 0.0001 for Gsk3β KD (shRNA4) vs. scramble cells; #P < 0.05, ##P < 0.01, ###P < 0.001, and ####P < 0.0001 for Gsk3β KD (shRNA1) vs. scramble cells. C Evaluation of in vitro colony formation capabilities in 22Rv1- Gsk3β KD (shRNA1 and shRNA4 clone), scramble, and parental cells (upper panel). Quantitative analysis of these colonies is represented (lower panel), and values are expressed as mean ± standard error of the mean (SEM, n = 5).Statistical significance was considered at P ≤ 0.05. For between-group comparisons, ****P < 0.0001. D Representative images of the corresponding spheroids (left panel) were presented and collected with the Incucyte System. Incucyte-based real-time kinetics of 3D/matrix-multi spheroids growth presented as spheroids object area (upper panel) and spheroids darkness (lower panel) of 22Rv1- Gsk3β KD (shRNA4 clone) vs. scramble cells (right panel), One 4× image per well is taken every 6 h, and data are presented as mean ± SEM; n = 3. P ≤ 0.05 was considered to indicate statistical significance. **P < 0.01, *<0.05 compared with Scramble group. E Representative fluorescent images and picture of xenograft tumors (upper panel), quantitative analysis of tumor volume (lower panel) of scramble vs. Gsk3β KD. F Gene set enrichment analysis (GSEA) on RNA-seq data of Gsk3β KD vs. scramble cells. The enrichment plot represents PI3K_AKT_MTOR pathway analysis in these groups. G Representative WB images show the expression of p-AKT and AKT in 22Rv1- Gsk3β KD vs. scramble cells in time-dependent manners. GAPDH was used as a loading control. H The enrichment plot represents apoptosis pathway GSEA in Gsk3β KD vs. scramble cells. I Representative WB images show the expression of Bcl and Bax in 22Rv1- Gsk3β KD vs. scramble cells. GAPDH was used as a loading control. J Representative micrograph of apoptosis determined by flow cytometric analysis of annexin-V Cy-5/PI- dual stained cells (AV+/PI–intact cells; AV/PI+–nonviable/necrotic cells; AV+/PI and AV+/PI+–apoptotic cells) in 22Rv1- Gsk3β KD vs. scramble K Mitochondria fragmentation was determined by using Mitotracker Red (100 nm for 30 min) in 22Rv1- Gsk3β KD vs. scramble cells L Enrichment plot illustrating the GSEA of oxidative phosphorylation and glycolysis gene sets of Gsk3β KD vs. scramble cells. M Heatmap displays the top 10 significant upregulated genes related to RNA-seq analysis of Gsk3β KD and scramble cells. N ChIP assay illustrates the binding of AR to the endogenous NKX3-1 or PSA gene in Gsk3β KD vs. scramble cells.

Manifestations of apoptosis, mitochondrial dysfunction, and Metabolic reprogramming evident with Gsk3β inhibition

The GSEA analysis showed that apoptosis enriched out of the top 20 altered HALLMARK GSEAs in Gsk3β KD cells compared to scramble cells (Fig. 5H). In order to unravel the underlying apoptosis pathway, we analyzed the expression analysis of apoptotic proteins, such as Bcl, Bax, cleaved PARP, cleaved caspase-9, and cleaved caspase-3 and observed the downregulation of the anti-apoptotic Bcl protein and the upregulation of the pro-apoptotic Bax in Gsk3β KD cells at 48 h compared to scramble (Fig. 5I). Surprisingly, no significant changes in the expression of caspases were observed between the Gsk3β KD cells and the scramble condition (Fig. S7E). However, the FACS assay indicated an increased population of late apoptotic cells in Gsk3β KD cells compared to the scramble (Fig. 5J and S7F). Similarly, no substantial alterations were observed in the autophagic markers, such as Beclin-1 and SQSTM1/p62, between the Gsk3β KD cells and scramble, suggesting the involvement of pathway, independent of the caspase-regulated cell death (Fig. S7E). Considering the role of Gsk3β in mitochondrial respiration and association with apoptosis, the morphology determination of mitochondria using Mitotracker staining showed significant fragmentation of mitochondria in the Gsk3β KD cells (Fig. 5K). This finding suggests that lack of Gsk3β contributes to mitochondrial dysfunction, which is known to be associated with apoptotic cell death.

The GSEA revealed that Gsk3β KD cells displayed significant enrichment in the glycolysis pathway phenotype compared to scramble cells. Interestingly, in scramble cells, the oxidative phosphorylation phenotype was enriched in out of 20 hallmark GSEA (Fig. 5L). This metabolomic rewiring was confirmed by evaluating oxygen consumption rate/mitochondrial respiration. We observed the significant downregulation of the OCR rate (at the basal respiratory rate) in Gsk3β KD cells compared to scramble (Fig. S7G). This shift suggests that the Gsk3β might be a potential regulator for maintaining oxidative phosphorylation as an energy source for PCa cell survival.

Subsequent to this, in order to comprehend the effects of Gsk3β KD at the gene level, we ran differential expression analysis (DEA) to identify the top 10 up and down-regulated genes in Gsk3β KD vs. scrambled cells. Our DEA highlighted that NKX3-1 is the top upregulated gene (Fig. 5M). NKX3-1, a prostate-specific tumor suppressor gene, is transcriptionally regulated by AR and considered the downstream target of Shh in the prostate [28]. It operates by stabilizing P53 and inhibiting AKT activation, thus suppressing the initiation of PCa [29]. In addition, NKX3-1 was. In our gene set enrichment analysis, we observed the upregulation of P53 and the downregulation of the AKT pathway (Fig. S7D). To further corroborate these findings, we conducted chromatin immunoprecipitation (ChIP) assays to investigate the binding of AR to transcriptionally active genes, including NKX3-1 and the well-known PSA (Fig. 5N) In Gsk3β KD cells compared to scramble cells, we noted an augmentation in AR binding to NKX3-1, whereas there was a reduction in AR binding to PSA genes. However, further investigation is needed to examine whether the upregulation of NKX3-1 acts as a potential regulator in the PCa growth inhibition by Gsk3β KD, as suggested by the data.

Potential involvement of Gsk3β in the generation of t-Gli3 and facilitating interaction between t-Gli3 and AR-V7

Examining the Gsk3β-mediated signaling pathway regulation, we noted an enrichment of Hh signaling in Gsk3β KD cells and a significant enrichment of MYC targets in scramble cells (Fig. 6A). This observation further reinforces the notion that Gsk3β serves as a master regulator of the Hh pathway. Subsequently, we examined the effect of Gsk3β KD on Gli proteins and noted that the expression of Gli1, Gli2, and Sufu remained unchanged. However, a noteworthy upregulation was observed in the Gli3 expression, particularly the FL-Gli3, at both 72 h and 96 h (Fig. 6B). The observed increase in FL-Gli3 suggests that inhibition of Gsk3β may influence the processing of Gli3, potentially inhibiting the generation of t-Gli3. Furthermore, we performed co-expression analyses of Gli3/AR, Gli2/p-Gsk3β, and Gli1/Gsk3β using confocal microscopy (Fig. 6C–E) and fluorescence intensity was analyzed. We did notice a loss of nuclear expression of Gli3 in Gsk3β KD cells, indicating that inhibition of Gsk3β impedes the nuclear translocation of Gli3. Similar to the WB results, we did not observe any changes in the expression levels of Gli1 and Gli2 between Gsk3β KD and scramble cells. These observations suggest a potential regulatory role of Gsk3β for the nuclear translocation of t-Gli3 and subsequent modulation of Gli3-mediated signaling response. Taken together, gene set enrichment and protein expression analysis, we can again conclude that the presence of Gsk3β regulates Gli cascade at the protein level of Gli3 processing.

Fig. 6: Impact of Gsk3β knockdown on Hh/Gli cascade and processing of Gli3.

A Enrichment plot illustrating the GSEA analysis for the Hh pathway and MYC targets in Gsk3β KD cells compared to scramble cells. B Representative WB images show the expression of Gsk3β, Gli proteins, and Sufu in 22Rv1- Gsk3β KD vs. scramble cells for different time points. C–E Representative IF images (left panel) and fluorescence intensity plot (right panel) of co-expression of Gli3/AR, p-Gsk3β/Gli2, and Gsk3β/Gli1 in 22Rv1- Gsk3β KD vs. scramble cells. F Representative WB images show the expression of AR and Gli proteins following IP with AR in 22Rv1- Gsk3β KD vs. scramble cells (left panel), 22Rv1 cells treated with Gsk3β activator vs. untreated (control) cells (right panel). AR and Gli proteins were expressed following IP with AR in C4-2B cells treated with Gsk3β activator vs. untreated (control) cells (middle panel). Gli3 and AR proteins were expressed following IP with Gli3 in 22Rv1 cells treated with Gsk3β activator vs. untreated (control) cells (lower panel). G Representative WB images show the expression of Gli3 and AR in different treated groups. GAPDH was used as a loading control. H Representative IF images (left panel) and fluorescence intensity plot (right panel) of co-expression of Gli3/AR in indicated cells.

To substantiate the earlier observations regarding the reliance of t-Gli3 generation on the presence of Gsk3β and its specific interaction with AR-V7, we conducted experiments involving Gsk3β inhibition and Gsk3β activation in PCa cells through IP (Fig. 6F). In our investigation with Gsk3β KD-22Rv1 cells expressing both AR-FL and AR-V7, the reduction in t-Gli3 expression, compared to the scramble following AR enrichment, indicated a lack of interaction between AR and Gli-family proteins (Gli1, Gli2, and t-Gli3) after Gsk3β knockdown (Fig. 6F left panel). Conversely, upon Gsk3β activation using LY-294002 in 22Rv1 cells, a significant increase in both FL-Gli3 and t-Gli3 expression was observed, suggesting the facilitation of the interaction between Gli3 and AR in the presence of Gsk3β (Fig. 6F right panel). Furthermore, in C4-2B cells, which only express AR-FL, we did not observe any interaction between AR and Gli-family proteins (Fig. 6F middle panel). Moreover, in reverse co-IP experiments conducted in 22Rv1 cells with Gli3 enrichment, we noticed an interaction between Gli3 and AR (Fig. 6F lower panel).

Phosphorylated-Gli3 processing is triggered in a ubiquitin-dependent manner via the SCF-βTrCP1 complex [30]. To further corroborate these crucial findings, we employed the proteasome inhibitor MG132, Gsk3β activator, or a combination of both (Fig. 6G). When the Gsk3β activator was used alone, we observed an increase in the expression of both F-Gli3 and t-Gli3. Conversely, treatment with MG132 alone resulted in the inhibition of t-Gli3 formation. Interestingly, when the group pre-treated with MG132 was subsequently treated with the Gsk3β activator, we found a significant induction of t-Gli3 expression compared to the group treated with MG132 alone. We also examined the expression of AR in these experimental conditions and found that treatment with MG132 alone was able to inhibit AR-V7 expression. However, no significant changes in AR expression were observed in the groups treated with the Gsk3β activator alone or in combination with MG132. This suggests that Gsk3β-mediated processing is specific to Gli3 and does not affect AR expression. Indeed, the additional confocal imaging analysis allowed us to visually confirm the expression pattern of Gli3 and AR (Fig. 6H). The consistent results obtained from confocal microscopy and WB provide robust evidence supporting the conclusion that Gsk3β activator can override the inhibitory effect of MG132 on t-Gli3 formation, leading to an increase in t-Gli3 expression.

Concurrent inhibition of Gsk3β along with Smo or Gli1 leads to an improved cytotoxic effect in CRPC cells

We investigated the therapeutic potential of Smo/Gli cascade inhibition using Cyclopamine and its analog, Tomatidine. However, no cytotoxic effects were observed with these inhibitors (Fig. S8A). While Cyclopamine effectively downregulated Smo (48h, 72 h) and Gli1 (72h) over time (Fig. S8B), it did not significantly impact the cell viability in CRPC cells under serum starvation and androgen deprivation (Fig. S8C). However, preincubation with DTH led to growth inhibition upon Cyclopamine treatment (Fig. S8D). We assessed additional Hh inhibitors, GDC0149 (a Smo inhibitor) and GANT-58 (a Gli1 inhibitor), on PCa cell growth kinetics. Despite effectively targeting/modulating Smo and Gli1 expression (Fig. S8E), these inhibitors did not significantly impact cell viability and apoptosis (Fig. S5F, and S5G, H). In contrast to limited efficacy when inhibiting Smo or Gli1 alone, the combination of Smo/Gli1 and Gsk3β inhibitors demonstrated significant efficacy in inhibiting cellular viability (Fig. S9A,B). Following the intriguing results, we extended our investigation by employing a 3D culture system (spheroid assay). Similarly, in a 3D spheroid assay, the combination of GDC-0449 and AR-A014418, as well as GANT-58 and AR-A014418, resulted in reduced spheroid count compared to individual or untreated groups (Fig. 7A, B). To confirm the successful formation of spheroids using 22Rv1 cells, we performed fluorescence imaging utilizing Calcein AM and PI staining (Fig. 7C). These findings shed light on the potential synergistic effects of simultaneously targeting Gsk3β along with Smo or Gli1 in inhibiting cell viability, indicating a promising therapeutic approach.

Fig. 7: Cytotoxic effects of concurrent Gsk3β inhibition with either Smo or Gli1 or solely Gli3 inhibition.