USP35 expressed highly in PRAD samples that correlates with inferior prognosis
First of all, we queried the expression data of USP35 in GDS2545 derived from the Gene Expression Omnibus (GEO) database, observing that USP35 levels were higher in 65 tumor samples as compared to 63 adjacent normal tissues (Fig. 1A). The clinical information of PRAD samples was summarized in Table S1. In line with the findings, we also analyzed that USP35 expressed highly in PRAD samples in TCGA-PRAD and Oncomine datasets, respectively (Fig. 1B, C). Besides, we collected the information of clinical characteristics for the PRAD patients to conduct the correlation analysis. As expected, high USP35 expressions were observed in patients with advanced T or N stages (Fig. 1D, E). Also, patients with a definite biochemical recurrence or ≥8 Gleason scores showed elevated levels of USP35 (Fig. 1F, G). Moreover, Kaplan–Meier survival curves analysis revealed that patients with high USP35 levels bear worse disease-free survival (DFS) in TCGA-PRAD cohort (p < 0.001), worse overall survival (OS) in GSE70769 (p = 0.021), as well as worse progression-free survival (PFS) in GSE1169181 (p = 0.017) (Fig. 1H–J). Lastly, we also conducted the multi-variate Cox regression analysis by integrating several hazard clinical variables in TCGA-PRAD cohort. Compared with other variables, like age, TN stages, or Gleason scores, USP35 is an independent prognostic factor for predicting DFS of PRAD patients (Fig. 1K). The time-dependent receiver operating characteristic curve (ROC) further demonstrated that combination of USP35 and other clinical variables could reach higher predictive efficiency, as compared to USP35 or clinical variables respectively (Fig. 1L). In conclusion, our bioinformatic analysis suggested that USP35 is up-regulated in PRAD samples and possesses a tight correlation with a worse prognosis.
USP35 enhances tumorigenesis and stemness potentiality of PRAD cells
Based on the above bioinformatic findings, we intended to assess the USP35 functions in prostate cancer. First of all, we deleted USP35 in two prostate cancer cell lines C4-2b and PC-3 via sgRNA-mediated CRISPR/Cas9 KO technology (Fig. 2A). In contrast, we generated stable USP35-overexpressing cell lines via lentivirus infection (Fig. 2B). Next, we started to determine the roles of USP35 in PRAD proliferation. The colony formation assay exhibited that USP35 overexpression notably promoted the number and sizes of PRAD cell colonies in two cell lines (Fig. 2C). USP35 depletion could decrease PRAD cell proliferation rates relative to parental control cells that could be restored by USP35 overexpression, as suggested by the cell viability assays (Fig. 2D). Meanwhile, we also found that USP35 could further enhance stem cell-like properties in PRAD cells, as evidenced by the sphere formation assays (Fig. 2E). The wound-healing assay also indicated that USP35 deletion resulted in decreased migration rates of cells relative to parental control cells (Fig. 2F). Conversely, ectopic expression of USP35 robustly potentiated the migration efficacy of cells as compared to cells transfected with vector (Fig. 2G). Transwell assays also indicated that ectopic expression of USP35 in USP35-deficient cells could completely rescue the restricted migration abilities (Fig. 2H). Last of all, to further evaluate USP35 roles in prostate tumorigenesis, we generated an orthotopic prostate tumor model in which C4-2b cells were injected into the prostate gland of nude mice. The in vivo bioluminescence (BIL) signals were used to indicate the growth of orthotopic prostate tumors. In accordance to our speculations, USP35 overexpression could result in the formation of larger prostate tumors relative to those in control mice, suggesting that USP35 significantly promoted tumorigenesis (Fig. 2I). Collectively, these data indicated that USP35 promotes the proliferation, migration, and stemness properties of PRAD cells both in vitro and in vivo.
USP35 interacts with and deubiquitinates BRPF1 in PRAD cells
To clarify the downstream targets that exert the oncogenic functions of USP35 in PRAD, we generated the stable C4-2b cells with double-tagged FLAG-HA-USP35. Next, the USP35-containing protein complex was purified and isolated via Tandem Affinity Purification (TAP) method. We thus identified a list of peptides in the complex, including FZR1, TNIP2, MAVS, or BRPF1 (Fig. 3A). Considering that epigenetic deregulation contributes to PRAD progression and BRPF1 is less reported in PRAD, we thus decided to confirm the associations between USP35 and BRPF1. First of all, the co-immunoprecipitation (co-IP) analysis with an anti-USP35 indicated that USP35 could directly interact with BRPF1, implicating the endogenous interactions (Fig. 3B). Besides, we also detected that the BRPF1 proteins were decreased in USP35-deleted C4-2b cells relative to parental control cells (Fig. 3C). However, USP35 loss could not induce alterations of BRPF1 mRNA levels (Fig. 3C). Next, we observed an increased levels of BRPF1 proteins when C42-B cells were transfected with elevated doses of Myc-USP35 plasmids, suggesting that USP35 promotes BRPF1 levels in a dose-dependent manner (Fig. 3D). Intriguingly, high USP35 did not alter BRPF1 mRNA levels (Fig. 3D). Consistently, the cycloheximide (CHX) assays also indicated that USP35 deletion resulted in shorter half-life of BRPF1 proteins relative to those in parental control C4–2B cells (Fig. 3E). In contrast, USP35 overexpression significantly prolonged the half-life of BRPF1 proteins in C4-2B cells (Fig. 3F). As shown in Fig. 3G, only wild-type USP35, but not the C450A mutant, could sufficiently catalyze deubiquitination of BRPF1. Accordingly, only wild-type USP35, but not the C450A mutant, could promote the accumulations of BRPF1 proteins, not altering the corresponding mRNA levels of BRPF1 (Fig. 3G). Taken together, these data indicated that USP35 can function as a deubiquitinase for BRPF1.
USP35 depends on BRPF1 to exert oncogenic roles in prostate tumorigenesis
To further determine the biological roles in PRAD, we knocked down BRPF1 in C4-2b and PC-3 cells through lentiviral transduction method using three different BRPF1 short hairpin RNAs (Fig. 4A). The MTT assay showed that BRPF1 knockdown remarkably inhibited prostate cancer cell growth (Fig. 4B, C). However, BRPF1 overexpression could notably enhance cell colony formation ability of PRAD cells (Fig. 4D). In addition, bioinformatic analysis in TCGA-PRAD samples also suggested that high BRPF1 correlated highly with hazard clinical factors, like advanced T stages, high Gleason scores, as well as biological recurrence (Fig. 4E–G). Kaplan-Meier analysis suggested that patients with high BPRF1 levels had shorter DFS months as compared to those with low BRPF1 levels, as indicated by the log-rank test (Fig. 4H). Given that we have already generated USP35-overexpressing PRAD cell lines, we further knocked down BRPF1 in these cells. In line with the previous findings, USP35 is required for the proliferation, migration, and stem cell-like properties of PRAD cells, which could be largely abolished by BRPF1 KD (Fig. 4I–K). Collectively, these data implicated that USP35 depends on BRPF1 to enhance PRAD malignant progression.
USP35/BRPF1 axis modulates mevalonate (MVA) metabolism via epigenetically inducing SREBP2 levels
Given that BRPF1 is an epigenetic regulator that could induce specific genes expressions, we thus conducted the bioinformatic analysis based on sequencing expression data derived from TCGA-PRAD samples. Gene Set Enrichment Analysis (GSEA) implicated the tight interplay between BRPF1 and mevalonate (MVA) crosstalk in PRAD (Fig. 5A). Given that SREBP2 is the master regulator of MVA crosstalk, we thus intended to figure out the relationships between BRPF1 and SREBP2. Interestingly, BRPF1 KD could significantly reduce SREBP2 mRNA levels in PRAD cells (Fig. 5B). Besides, BRPF1 overexpression could promote SREBP2 mRNA levels (Fig. 5C). Meanwhile, we cloned a list of fragments of the SREBP2 promoter, and BRPF1 increased the luciferase activities of promoter fragments of P3, P4, and P5, but not the P1 or P2 (Fig. 5D). Thus, the region, ranging from −250 to −140 in the SREBP2 promoter, is the essential sequence binded and regulated by BRPF1. We also conducted the ChIP-qPCR assay to confirm that BRPF1 could directly bind to the SREBP2 promoter region to sustain the transcriptional activity, as implicated by the active H3K4me3 modification markers (Fig. 5E). Conversely, BRPF1 KD reduced the transcriptional activity at the SREBP2 promoter, as indicated by decreased H3K4me3 enrichment (Fig. 5F). To exclude the possibilities that BRPF1 could regulate other MVA transcriptional factors (TFs) like SREBP2, we detected that BRPF1 could not induce SREBP1a/c levels in PRAD cells (Fig. 5G). Along with these findings, we found that BRPF1 could not alter SREBP1a/c transcriptional activities (Fig. 5H). Given that USP35 could stabilize BRPF1, we thus wondered whether USP35 could rely on BRPF1 to modulate SREBP2 expressions. Apparently, we found that USP35 overexpression could enhance BRPF1 binding enrichment on the SREBP2 promoter region, which were notably suppressed with BRPF1 KD (Fig. 5I). In line with the results, we found that USP35 could enhance SREBP2 levels, and the increase could be completely abolished by BRPF1 KD (Fig. 5J). Conversely, USP35 deletion could suppress SREBP2 mRNA levels, which could be completely restored by BRPF1 overexpression in C4-2b and PC-3 cells (Fig. 5K). Taken together, these data implicated that USP35/BRPF1 axis could modulate MVA crosstalk in PRAD in a SREBP2-dependent manner.
Targeting BRPF1-induced MVA crosstalk endows a therapeutical vulnerability for USP35high PRAD
Given that SREBP2 regulates a series of enzymes that sustain MVA metabolism activities, we observed that USP35 depletion could down-regulate the mRNA levels of SREBP2 downstream targets, named as MVA signature, including HMGCR, FDFT1, SQLE, MSMO1, FDPS and LSS (Fig. 6A). However, BRPF1 overexpression could completely restore the expressions of MVA signature (Fig. 6A). Meanwhile, USP35 could activate MVA signature and this effect could be completely abolished by BRPF1 KD (Fig. 6B). Consistently, we detected that the free cholesterol content in USP35-OE cells, was about 50% higher than that in controls, and the increase could be completely abolished upon BRPF1 KD (Fig. 6C). In contrast, USP35 loss suppressed the free cholesterol content and the decrease could be largely restored with BRPF1 overexpression (Fig. 6D). In addition, we utilized the atorvastatin to inhibit the MVA pathway and found that atorvastatin alone could restrict the growth of PRAD cells relative DMSO treatment (Fig. 6E). Meanwhile, in line with our biological findings, atorvastatin could further largely suppress the growth of USP35-OE PRAD cells (Fig. 6E). Given that we have found the USP35/BRPF1/SREBP2 axis in regulating MVA pathway during prostate tumorigenesis, we thus intended to further test the translational significance for treating PRAD. We generated the C4-2b-derived tumor model and found that targeting BRPF1 could significantly suppress USP35high in vivo PRAD growth, as shown by tumor volumes and weight (Fig. 6F–H). Lastly, we also found that atorvastatin could also suppress the in vivo growth of tumors derived from USP35high PRAD cells, as compared to those derived from control cells (Fig. 6I). Collectively, these results implicated that suppressing BRPF1-induced MVA crosstalk provides a therapeutical vulnerability for USP35high PRAD.
Last of all, we collected some PRAD samples from our center and divided the samples into USP35high and USP35low groups. As shown in Fig. 7A, the positive associations among USP35, BRPF1, and SREBP2 were demonstrated via IHC method. We also assessed the positive relationships between USP35 and MVA signature in TCGA-PRAD samples (Fig. 7B). These data suggested that USP35 is an indicator for clinical PRAD samples with high MVA metabolism activity. We further illustrated the USP35/BRPF1/MVA axis in prostate cancer cells in Fig. 8.