Introduction

Prostate cancer (PCa) is a commonly diagnosed malignancy and is the predominant cause of cancer-related mortality in men worldwide [1]. Epidemiological evidence has revealed that ethnicity-related heterogeneity at the genetic, epigenetic, and phenotypic level is directly related to incidence, prognosis, and the chemotherapy response of PCa [2,3,4,5]. Recent sequencing studies have revealed that the somatically acquired genetic, epigenetic, and transcriptional alterations in PCa, including oncogenic gene PIK3CA, ETS fusion gene, and tumor suppressor gene (TSG) PTEN, SPOP, CHD1 [6,7,8,9]. A further genotyping study has revealed that the frequency of PTEN loss is 29.7 vs. 5.4% in Caucasian and Chinese patients, respectively [3], thus suggesting that Caucasian patients with PCa have a poor prognosis and less sensitivity to abiraterone [4, 10]. Therefore, discovering and validating PCa-associated genomic aberrations in Chinese patients, especially prevalent significant genes, would provide potential therapeutic targets for PCa.

Histone lysine methyltransferase 2D (KMT2D/MLL2), an important epigenetic modifier that catalyzes the monomethylation of H3K4 (H3K4me1) [11], has been found to be mutated from 0–12% in prostate cancers [12]. KMT2D is enriched at enhancers of multiple genes and epigenetically stimulates gene expression in various pathways, including the p53 pathway, cholestasis signaling, and cAMP-mediated signaling [13]. In lymphoma, KMT2D, a strong TSG, effectively inhibits tumorigenesis and metastasis [14]. In solid tumors such as pancreatic adenocarcinoma, breast, and colorectal cancer, KMT2D is essential for tumor cell proliferation, and its depletion increases chemotherapeutic sensitivity [15, 16]. A recent study has revealed that the mutation of KMT2D is PCa-specific and might be involved in the malignant progression from high grade prostatic intraepithelial neoplasia to PCa [17]. However, whether KMT2D is an oncogene or TSG in PCa remains unclear, and its downstream pathways still must be comprehensively determined.

Herein, we present our discovery of a highly mutated and overexpressed epigenetic gene, KMT2D, where its expression was positively correlated with metastasis and predicts prognosis. We also identified putative KMT2D/PI3K and KMT2D/EMT oncogenic pathways that promote prostate cancer outgrowth and metastasis by epigenetically transcriptional activation of novel targets, namely LIFR and KLF4. These findings suggested epigenetic regulator KMT2D might have therapeutic efficacy for treating PCa.

Results

Identification of recurrent KMT2D alterations in PCa through disease-targeted sequencing

To identify the mutational spectrum of PCa pathogenesis-associated genes in Chinese patients, 32 significantly mutated genes were analyzed by using disease-targeted sequencing in 46 FFPE samples (patient information see Supplementary Table 1). Sequencing depths were 3527 × on average (range, 1156 × −5419 × ). Each specimen harbored an average of five mutations with an allele frequency ≥ 3% (range, 1–34). We identified a total of 245 somatic non-synonymous mutations, including 199 missense, 3 nonsense, 24 in-frame dels, 1 nonstop, and 18 frameshift in 27 genes (Supplementary Table 2). Among them, 99 mutations (40.57%) were expected to be harmful for the structure or function of the encoded proteins, according to a SIFT/PolyPhen analysis (Ion Reporter). In an independent platform (Ion Torrent PGM), 91.5% of these high-confidence mutations were validated (Supplementary Table 2).

The prevalence of mutations in high-ranking genes was different among the mutational spectrum in the Caucasian population and our cohort of Chinese patients (n = 46). Compared with the Caucasian population, the significantly mutated genes in our cohort were KMT2D (29/46, 63.04%), AKAP9 (15/46, 32.61%), GLI1 (15/46, 32.61%), THSD7B (9/46, 19.57%), CDK12 (7/46, 15.22%), KDM4B (7/46, 15.22%), MED12 (7/46, 15.22%), ZNF473 (6/46, 13.04%), PIK3CA (6/46, 13.04%), and NCOA2 (5/46, 10.87%) (Fisher’s test, P < 0.05; Fig. 1). In contrast, frequently mutated genes in the Caucasian population, such as PTEN, CDKN1B, CDKN2A, KLF6, MYC, and SYNE3, were rarely identified in the 46 Chinese patients (P < 0.05, Fisher’s test; Supplementary Fig. 1a). Furthermore, only two significantly mutated genes (TP53 and SPOP) overlapping between the Chinese cohort and three independent Caucasian populations, whereas 10 highly mutated genes were uniquely identified in Chinese patients (Supplementary Fig. 1b).

Fig. 1
figure 1

Integrative molecular profiling and major clinical characterization in prostate cancer (PCa) samples. A heatmap showing genes (rows) with somatic mutations in 46 PCa patients (columns). Clinicopathological variables are shown in the header, including tumor stage, Gleason score, and tumor purity estimated by HE staining. Variant classification is indicated by color coding. Right, mutation percentage in 46 PCa specimens

Full size image

Given the high mutation frequency of KMT2D, a correlation analysis was performed to determine whether functional interactions existed on the basis of the KMT2D variant frequency. We found that the allele frequency of KMT2D showed a positive trend toward gene abnormalities in PI3K pathway (P = 0.064; Supplementary Fig. 1c), thus suggesting that functional interplay is involved.

High expression of KMT2D in PCa is associated with recurrent mutations and leads to poor prognosis

In 29 out of 46 Chinese patients, 46 KMT2D mutations, including 23 missense, 18 in-frame del, and 5 frameshift, were found to be distributed over 23 loci (Fig. 2a). Owing to multiple subclones in the sequencing specimens and the long length of the KMT2D gene (~19 kb), more than one KMT2D mutation (maximum 5) was identified in 28.26% of our PCa samples (Supplementary Table 2).

Fig. 2
figure 2

High expression of KMT2D in PCa is associated with recurrent mutations and leads to poor prognosis. a Schematic representation of KMT2D mutations detected in 63.04% (29/46) of prostate tumors. b mRNA expression of KMT2D in KMT2D mutated, wild-type cases, and benign prostate hyperplasia. c Linkage disequilibrium plot of KMT2D mutations in the protein-coding region. The LD plot is based on pairwise r 2 value. d KMT2D gene expression classified by H3K4me1 protein levels in PCa patients. Scale bars: 50 μm. e Representative images of immunohistochemistry staining against KMT2D in benign prostate hyperplasia (n = 3) and PCa samples (n = 51). Scale bars: 50 μm. f KM survival analysis of PCa patients based on protein expression of KMT2D (n = 51). Values are shown as median ± quartile

Full size image

Compared to BPH, KMT2D was significantly overexpressed in PCa tissues in our cohort 1 and TCGA data (Supplementary Fig. 2a and 2b). Meanwhile, KMT2D transcription in mutation cases were 2.73-fold and 5.15-fold higher, respectively than that in wild-type and BPH (Fig. 2b). Linkage disequilibrium among variations is predicted to affect gene expression [18]. To determine the potential link between KMT2D mutations and its transcription, an allelic imbalance analysis was performed. Four strong linkage disequilibrium events among the KMT2D mutations were observed, including NM_003482:c.C13801A (p.P4601T) and NM_003482:c.G10640A (p.R3547H), NM_003482:c.C13801A (p.P4601T) and NM_003482:c.C2132T (p.P711L), NM_003482:c.G10640A (p.R3547H) and NM_003482:c.C2132T (p.P711L), as well as NM_003482:c.G10126C (p.V3376L) and NM_003482:c.A5975G (p.E1992G) (r 2 = 1; Fig. 2c). Whereas no significant correlation between KMT2D gene expression and copy number variation (CNV) was observed (Supplementary Fig. 2c). Besides, a weak negative correlation between DNA methylation and KMT2D transcription was found using TCGA data (r = −0.27, P < 0.001; Supplementary Fig. 2d), implying that rather than CNV, KMT2D variants, and epigenetic methylation may associate with its transcriptional regulation.

Furthermore, H3K4me1 abundance, marker for mirroring KMT2D activity [14, 19], was examined in PCa tissues with KMT2D mutations and wild-type. As shown in Fig. 2d, KMT2D transcription was positively correlated with H3K4me1 levels (r = 0.592, P < 0.001), suggesting KMT2D gene expression was consistent with its catalytic activity. Moreover, in KMT2D frameshift cases, the percentage of H3K4me1-positive cells was consistently decreased (n = 4, P = 0.085; Table 1), and compared to other KMT2D mutations, KMT2D catalytic activity in KMT2D frameshift cases were significantly reduced (P = 0.021; Table 1). These suggested that in our cohort 1, frameshift might probably be associated with the reduced KMT2D catalytic activity.

Table 1 H3K4me1 expression in different KMT2D mutated tissues (n = 35)
Full size table

In cohort 2, KMT2D was abundantly expressed in 66.67% (34/51) of PCa specimens but was not detected in any (0/3) of BPH tissues (P = 0.037; Fig. 2e and Supplementary Table 1). KMT2D was predominantly expressed in nuclei and exhibited a heterogeneous pattern within the same PCa specimens (Fig. 2e). Notably, higher KMT2D protein levels were correlated with poorer disease-free survival (DFS) time in 51 Chinese patients (log-rank P = 0.048; Fig. 2f). Similarly, a negative correlation was also observed between KMT2D transcription and prognosis in TCGA dataset (P = 0.031; Supplementary Fig. 2e). These findings indicated that high KMT2D expression may serve as a putative prognostic biomarker for Chinese PCa patients.

KMT2D depletion inhibits cell proliferation and induces cell apoptosis in PCa

To investigate KMT2D function in PCa tumorigenesis, cell proliferation and apoptosis were evaluated in KMT2D-depleted PCa cells. Since KMT2D abundance in LNCaP cells was relatively low, PC-3 and DU145 cells were selected for further knockdown (P < 0.001; Supplementary Fig. 3a). Both the transcriptional and translational levels of KMT2D were effectively silenced by two KMT2D-specific siRNAs (#1 and #2; Supplementary Fig. 3b and 3c).

As expected, siKMT2D transfections significantly suppressed cell viabilities (all P < 0.001; Fig. 3a). In KMT2D-depleted PC-3 cells, the percentage of EdU-positive cells was decreased by 59.67%, significant reduction was also observed in KMT2D-silenced DU145 cells (P = 0.005; Fig. 3b). Moreover, with KMT2D deficiency, the early and total apoptotic ratios were significantly increased (P < 0.05; Fig. 3c), and subsequently, the expression of anti-apoptotic proteins, Bcl-2 and Bcl-xl, were dramatically decreased in KMT2D-silenced PC-3 and DU145 cells (P < 0.001; Fig. 3d).

Fig. 3
figure 3

KMT2D is a prerequisite for tumor proliferation and induces cell apoptosis in PCa. a Cell viability analysis of PC-3 and DU145 cells containing siControl or siKMT2D at 48, 72, and 96 h post-transfection. Values represent the mean OD of triplicate measurements. ***P < 0.001, compared with siControl cells. b Representative images of an EdU incorporation assay (20 × objective) for PC-3 and DU145 cells with siControl or siKMT2D. The EdU positive percentage is shown as bar graphs. **P < 0.01, ***P < 0.001, compared with siControl cells. c Cell apoptosis was evaluated with flow cytometry using Annexin V-FITC/PI staining in PC-3 and DU145 cells with siControl or siKMT2D. Quantification is shown as a bar graph in the right panel (analyzed with FlowJo software). *P < 0.05, **P < 0.01, ***P < 0.001, compared with siControl cells. d Western blot analysis of the apoptosis-related proteins Bcl-2 and Bcl-xl after KMT2D knockdown in PC-3 and DU145 cells. e shControl and shKMT2D PC-3 cells were injected into the left flanks of mice (n = 5 each group), and tumor volumes were recorded and plotted for 35 days. f Images showing the size of the tumors that developed in mice of both groups; tumors were detected with IntegriSense 750 by IVIS at day 35. Color scale depicts fluorescence in units of radiance (photons/sec/cm2/sr/μM/cm2). All values are mean ± SD. Data are representative of at least three independent experiments. sr steradian radium

Full size image

To further verify the role of KMT2D in PCa development in vivo, KMT2D-depleted PC-3 xenograft model was developed (Supplementary Fig. 3d). Tumor growth was effectively inhibited, and tumor volume was remarkably suppressed by 92.21% (P < 0.001, n = 5; Figs. 3e, f) with KMT2D knockdown. Hence, KMT2D depletion restricted PCa progression, partially by inhibiting tumor cell proliferation and inducing cell apoptosis.

KMT2D knockdown attenuates tumor metastatic capacity in vitro

Considering approximately 90% of PCa patients develop metastasis [20], to further determine the role of KMT2D in tumor metastasis process, wound healing and transwell assays were performed with KMT2D-depleted cells. Our results showed that the migratory ratio was significantly decreased after KMT2D silencing (range, 23.07–54.21%, P < 0.01; Fig. 4a). In addition, KMT2D knockdown markedly inhibited the invasiveness of PC-3 and DU145 cells (range, 65.61–87.40%, P < 0.001; Fig. 4b). KMT2D depletion remarkably increased the expression of E-cadherin (epithelial marker [21]; Fig. 4c), while decreased that of N-cadherin (mesenchymal marker [21]; Fig. 4d) in PC-3 and DU145 cells.

Fig. 4
figure 4

Loss of KMT2D attenuates cell migration and invasion. a Cell migration capacity measured with a wound healing test in PC-3 and DU145 cells transfected with siControl or siKMT2D for 48 h (10× objective). The wound healing percentage is shown as a bar graph and was quantified with ImageJ software. **P < 0.01, ***P < 0.001, compared with siControl cells. b The invasive ability of cells, as measured with a transwell matrigel assay (10 × objective). Invading cells were stained with violet and dissolved in DMSO. The OD levels were quantified with microplate reader and are shown as a bar graph. ***P < 0.001, compared with siControl cells. Immunofluorescence staining analysis of E-cadherin and N-cadherin in PC-3 c and DU145 d cells treated with siControl or siKMT2D (scale bars: 50 μm). All values are mean ± SD. Data are representative of three independent experiments

Full size image

By analyzing transcriptional data in the MSKCC cohort [22], we found that the KMT2D mRNA level in metastatic tissues was significantly higher than that in primary tissues (P = 0.029; Supplementary Fig. 4a). Similarly, in our cohort 1, KMT2D gene expression also exhibited a positive trend relating to high aggression (Supplementary Fig. 4b and 4c).

KMT2D sustains the expression of a set of genes in PCa cells

To further determine the mechanism underlying the biological consequences of KMT2D knockdown, RNAseq were performed. After silencing KMT2D with siRNA, 1442 and 1310 differentially expressed genes (DEGs) were detected, and 928 overlapping DEGs were also found (Fig. 5a and Supplementary Table 3). Among the top 10% of the overlapping DEGs, 90.2% of the genes were decreased after KMT2D depletion. Through KEGG and GSEA analysis, the 928 overlapping DEGs were clustered into the PI3K/Akt pathway (Fisher’s test, P = 4.56E-06; Figs. 5b, c). In validating whether the PI3K/Akt pathway is involved in KMT2D function, we found a significant decrease in the phosphorylation of Akt at S473 and T308 in KMT2D-depleted DU145 and PC-3 cells (Fig. 5d). Moreover, two downstream targets of p-Akt, p-BRCA1 and p-CREB, were also markedly decreased, while Akt-inactivated p-GSK3β was increased in KMT2D-depleted cells (Fig. 5d). GSEA enrichment revealed that disrupted genes by KMT2D-knockdown were negatively correlated with prostate carcinogenesis and Akt upregulated genes, while positively correlated with “breast cancer metastasis downregulated” (FDR < 0.05; Fig. 5e). Hence, the subsequently altered genes enriched in the PI3K/Akt and metastasis pathways may partially contribute to the oncogenic role of KMT2D in tumor proliferation and metastasis.

Fig. 5
figure 5

RNAseq identifies transcriptional changes enriched in PI3K signaling following loss of KMT2D. a A Venn diagram depicting the overlapping DEGs in PC-3 cells transduced with two siRNAs against KMT2D vs. siControl-treated PC-3 cells (top) and heatmaps showing the top 10% of overlapping DEGs (bottom). b, c Pathway enrichment identified by KEGG b and GSEA c analysis of the DEGs between KMT2D-depleted PC-3 and siControl cells. d Western blot analysis of Akt and phospho-Akt (S473 and T308), p-BRCA1, p-CREB, and p-GSK3β after KMT2D knockdown in PC-3 and DU145 cells. e GSEA analysis of the DEGs enriched in prostate carcinogenesis, Akt upregulated, and breast cancer metastasis downregulated pathway

Full size image

KMT2D modulates activation of PI3K/Akt and EMT-associated pathways by epigenetically stimulating LIFR and KLF4

As an effective histone methyltransferase catalyzing monomethylation of H3K4, KMT2D plays an important role in cancer initiation, progression, and metastasis [14, 23]. To confirm the catalytic function of KMT2D, the abundance of H3K4me1 was evaluated with immunofluorescence staining. As expected, in KMT2D-silenced PC-3 and DU145 cells, the protein expression of H3K4me1 was markedly suppressed in the nucleus (Fig. 6a).

Fig. 6
figure 6

KMT2D up-regulates LIFR and KLF4 by methylating H3K4me1. a Immunofluorescence staining analysis using anti-H3K4me1 (red) in PC-3 and DU145 cells treated with siControl or siKMT2D (nuclear staining with DAPI; scale bars: 50 μm). b Genomic distribution of different peaks localized in the functional region (exon, intron, promoter, intergenic region) of nearest genes by ChIPseq using anti-H3K4me1 in PC-3 cells transfected with siControl or siKMT2D. c Normalized read density tracks of H3K4me1 ChIPseq peaks by UCSC in PC-3 cells transfected with siControl or siKMT2D. d The consensus motif of the binding sites loss upon KMT2D silencing: H3K4me1 (P = 1.56E-15). The alignment of LIFR and KLF4 gene sequences from Human (Homo sapiens), Chimpanzee (Pan troglodytes), and Rhesus (Macaca mulatta). e qRT-PCR confirmed the decreased gene expression of LIFR and KLF4 after KMT2D knockdown in PC-3 and DU145 cells. **P < 0. 01, compared with scramble control cells. f, g KMT2D mRNA levels positively correlates with LIFR f and KLF4 g gene expression in PCa tissues in our cohort 1 and TCGA dataset. All values are mean ± SD. Data are representative of three independent experiments

Full size image

To further investigate KMT2D-mediated downstream genes, ChIPseq was performed using the H3K4me1 antibody in PC-3 cells. We identified 33,527 and 28,466 H3K4me1-binding sites in siControl and siKMT2D transfections, respectively. In the analysis of 230 differential binding events, 45.7% were distributed in intergenic regions, whereas the others were located in introns (53.3%), exons (0.7%), or promoters (0.4%) of 115 protein-coding genes (Fig. 6b and Supplementary Table 4).

According to the clustered PI3K/Akt pathway and EMT in KMT2D-depleted PC-3 cells, LIFR and KLF4 were selected, owing to their significant discrepancy in H3K4me1 binding ability (Supplementary Fig. 5 and Supplementary Table 5). LIFR, an upstream gene in the PI3K/Akt pathway, is essential for tumor progression through regulating cell proliferation [24]. KLF4 is a crucial protein involved in PCa metastasis [25]. After KMT2D knockdown, we observed a substantial loss of H3K4me1-binding sites in the gene body of LIFR and downstream (~643 kb) of KLF4 (Fig. 6c). In addition, we observed that both LIFR and KLF4 genes contained the consensus motif that was identified in the H3K4me1 binding peaks in the setting of KMT2D depletion (Fig. 6d).

To validate whether LIFR and KLF4 were downstream target genes for KMT2D modification, qRT-PCR was conducted with KMT2D-depleted cells. We observed that the gene expression of LIFR and KLF4 was consistently decreased (range, 0.31–0.70-fold) after KMT2D silencing (Fig. 6e). Furthermore, using cohort 1 and TCGA dataset, we also found that the KMT2D mRNA level was positively correlated with that of LIFR (P < 0.001; Fig. 6f) and KLF4 (P < 0.001; Fig. 6g). These results suggested that the KMT2D-mediated alterations in the PI3K/Akt pathway and EMT may be ascribed to epigenetic activation of LIFR and KLF4.

Discussion

In this study, we profiled the mutational spectrum of 32 PCa-associated significantly mutated genes in Chinese PCa patients (cohort 1, n = 46) and found KMT2D was the most recurrently mutated gene (variant frequency 63.04%) and high transcription. In an independent cohort (n = 51), KMT2D expression was negatively correlated with PCa prognosis. By generating KMT2D-depleted PCa cells, KMT2D-mediated biological consequences and corresponding signaling pathways were determined. After KMT2D knockdown, tumor cell proliferation, invasiveness, and migratory capacities were suppressed, whereas apoptosis was enhanced. Combinatorial analysis of RNAseq and ChIPseq revealed that KMT2D markedly activated the PI3K/Akt pathway and EMT by epigenetically stimulating LIFR and KLF4, respectively (Fig. 7).

Fig. 7
figure 7

The proposed mechanistic model of KMT2D in prostate cancer. The KMT2D sustains cell H3K4me1 level in prostate cancer and facilitates the expression of a set genes, including LIFR and KLF4. LIFR upon activation of PI3K pathway, and activated AKT phosphorylates downstream target such as CREB and BRCA1, promoting tumor growth. KLF4 facilitates transcription of EMT pathways, leading to tumor metastasis

Full size image

Our study revealed the mutation profiling of KMT2D in Chinese PCa patients. The distinct mutational spectrum of 32 significant genes also laid a foundation for further and extensive discoveries regarding the ethnicity-specific mutational landscape. In contrast to frequently mutated genes in the Caucasian population, such as PTEN, CDKN1B, and CDKN2A [6], in Chinese patients, the high-ranking mutated genes (>20%) were KMT2D (63.04%), AKAP9 (32.61%), and GLI1 (32.61%). The variant frequency of KMT2D, top mutated gene in our study, was much higher than the frequencies in Michigan (10.1%) [8], Robinson (12%) [26] and TCGA (6%) [12]. Moreover, KMT2D transcription was higher in PCa specimens than that in BPH (P < 0.001), and KMT2D high expression was correlated with poor prognosis (P = 0.048). These suggested KMT2D could serve as an alternative prognostic biomarker for PCa patients and further indicated Chinese-specific molecular subclassification and targeted therapy are urgently needed for Chinese PCa patients.

We provided a comprehensive portrait for KMT2D at genomic, transcriptional, translational, and especially functional levels in PCa patients. Mutation within gene body (or exons) is thought to affect protein integrity and/or activity. While unexpectedly, in our study, KMT2D transcription was increased in PCa tissues harboring KMT2D mutations. Recent study revealed that for certain genes (RYR2, NRIP1, and IRF8), their transcription could be triggered by mutations in coding exon regions [27]. Meanwhile, KMT2D-silencing was reported to cause genomic instability [28]. And this kind of genomic instability, induced by duplication of chromosome harboring the mutated sites, might promote certain gene amplification through structural variation [29]. Our TCGA analysis showed that DNA methylation could also potentially contribute to KMT2D gene regulation (r = −0.27, P < 0.001), suggesting an alternative mechanism underlying KMT2D transcription might be involved. Frameshift and nonsense usually led to truncated protein and decreased KMT2D activity in small cell lung cancer [30]. In our study, KMT2D frameshift mutation may probably associate with the reduced KMT2D catalytic activity compared to other mutations such as missense and in-frame del (P = 0.0212, Table 1). But given the small size of frameshift cases (n = 4), unfortunately, no significance was observed compared that to wild-type cases. Additionally, positive correlation between KMT2D gene expression and H3K4me1 level was also observed in PCa tissues (r = 0.749, P < 0.001), suggesting the transcription of KMT2D was consistent with its catalytic activity.

In addition, our findings provide substantial evidence that KMT2D is a putative oncogene in PCa progression. In KMT2D-knockdown PCa cells, cell proliferation was reduced while mitochondria-dependent apoptosis was induced, and KMT2D deficiency dramatically restricted tumor growth in xenograft model. Furthermore, KMT2D silencing significantly decreased invasion and migration accompanied by the increase of E-cadherin and decrease of N-cadherin. In contrast to the role of KMT2D as a TSG in lymphoma [14], the potential oncogenic function of KMT2D in PCa illustrated that the biological consequences of KMT2D are dependent on cancer type. By addressing the distinct function of KMT2D in PCa progression and metastasis, our present study strengthens the current understanding of KMT2D-linked epigenetic alterations in tumorigenesis and further expands the list of promising chemotherapeutic targets [31].

The most interesting discovery in our study is the undescribed identification of certain signaling pathways linked to KMT2D, such as PI3K/Akt pathway and EMT. Furthermore, integrative analysis of RNAseq and ChIPseq identified novel targets (LIFR and KLF4) in each pathway. Our current study provides clues that complement the signaling pathways linked to H3K4 methyltransferase and insights improving understanding of the mechanism underlying KMT2D-induced tumorigenesis in PCa: (i) Since approximately 25% of PCa tumors harbor PI3K abnormalities, and nearly 90% of PCa patients develop metastasis [20, 32], KMT2D-mediated activation of PI3K/Akt pathway strongly suggests that development of KMT2D inhibitors is critical either for PCa progression, or metastasis. (ii) Emerging evidence suggests that KMT2D is closely linked with nuclear receptors (ERα, RXR) [33, 34], transcription factors (Hox and Sox2) [34, 35] and inflammation pathways (CD40, JAK-STAT, and Toll-like receptor) in distinct tissues [14]. By identifying two novel KMT2D downstream targets in PCa, LIFR and KLF4, our study indicates that multiple cancer-specific targets are involved in KMT2D-mediated tumorigenesis. (iii) Previous studies have revealed that LIFR, through binding with IL-6, mediates several critical signaling pathways that regulate cell proliferation and survival, including JAK/STAT3, PI3K/Akt, and ERK1/2 [24]. KLF4 overexpression is reported to determine migration and proliferation of PCa cells, whereas its depletion attenuates tumor metastasis [25]. In line with these studies, the KMT2D-mediated epigenetic modification of LIFR and KLF4 observed in our study further supports these observations.

In summary, our current study deepens understanding of KMT2D in following aspects: (i) The distinct mutational spectrum of significant genes in Chinese PCa patients. (ii) By identifying numerous KMT2D-regulated pathways and novel downstream targets, combined with functional studies, we robustly defined the oncogenic role of KMT2D in PCa. (iii) The study revealed a direct correlation between PI3K/Akt signaling, the predominant aberrant pathway in PCa, and KMT2D, thus establishing an extensive basis and providing a rationale for studying the role of KMT2D in PCa tumorigenesis. (iv) LIFR and KLF4 identification complemented KMT2D-associated downstream genes in tumorigenesis. Our study lays a foundation for further understanding of the role of KMT2D in PCa and provides evidence for epigenetic therapy in treating PCa.

Materials and methods

Patients

This study was approved by Medical Ethics Committee of Nanfang hospital, and the species were used with the informed written consent in accordance with the Declaration of Helsinki. The study comprised a total of 17 benign prostate hyperplasia samples and 97 subjects with PCa including 46 in target next generation sequencing group (cohort 1; Supplementary Table 1) and 51 subjects in immunohistochemical staining of KMT2D (cohort 2; Supplementary Table 1). All samples have been histologically confirmed as PCa by two independent pathologists. Cases received pre-operative hormone therapy, chemotherapy, and radiotherapy were excluded.

DNA sample preparation and disease-targeted sequencing

The samples of disease-targeted sequencing cohort were dissected from 4 μm formalin-fixed paraffin-embedded (FFPE) sections and genomic DNA was extracted by a Qiagen GeneRead DNA FFPE Tissue kit (Qiagen, Hilden, Germany) with Uracil-N-glycosylase according to the manufacturer’s protocol. A prostate specific panel was brought by Qiagen containing 32 common frequently mutated genes, including AKAP9, APC, AR, CDK12, CDKN1B, CDKN2A, GLI1, IKZF4, KDM4B, KLF6, KMT2D, MED12, MYC, NCOA2, NIPA2, NKX3-1, NRCAM, OR5L1, PDZRN3, PIK3CA, PTEN, RB1, SCN11A, SPOP, SYNE3, TBX20, TFG, THSD7B, TP53, ZFHX3, ZNF473, and ZNF595. The extracted DNA was quantified using a Qubit fluorimeter (Invitrogen, CA, USA) and QIAseq DNA QuantiMIZE Assay Kit (Qiagen, Hilden, Germany). Genomic DNA with a yield >40 ng without low fragmentation and damage was used for library preparation. Amplicon enrichment was performed using the prostate cancer panel within the GeneRead DNAseq Targeted Panels V2 (Qiagen, Hilden, Germany).

Library preparation was prepared with the GeneRead DNA Library Core Kit for Ion Torrent instrument (Qiagen, Hilden, Germany) and barcode with the Ion Xpress™ Barcode Adapters (Thermo Fisher, MA, USA). Final barcoded pools were sequenced using the Ion Torrent Proton. An average of 6,243,762 mapped reads were finally obtained for each sample. The sequencing reads were mapped to hg19 using TMAP algorithm. Single-nucleotide variants were detected using a combinational online software consisted of the Cloud-Based DNAseq Sequence Variant Analysis (Qiagen, Hilden, Germany), custom scripts of Qiagen, and Ion Reporter (Thermo Fisher, MA, USA). We validated the variants by sequencing genomic DNA in three PCa samples with sufficient amounts of DNA using Ion Torrent PGM platform. Significantly mutated genes were identified by MutSigCV software [36].

For CNVs prediction, peripheral blood mononuclear cells from seven independently healthy males were collected and sequenced for CNVs baseline detection. CNVs calling was performed by Cloud-Based DNAseq Sequence Variant Analysis (Qiagen, Hilden, Germany). For the targeted genes, CNVs were defined as clustered-amplicons with significant change of normalized log2-read-depth, as previously reported [37].

Immunohistochemical staining

Consecutive 4-μm sections were obtained from FFPE blocks of the immunohistochemically staining group (86 subjects with PCa and three BPH). Sections were immunostained by anti-KMT2D antibody (Santa Cruz, CA, USA, 1:200) or H3K4me1 (Abcam, #ab8895, 1:1000) using the immunoperoxidase technique as previous study [38]. All sections stained with target protein were assessed by two independent pathologists.

RNA interference

Two specific siRNA targeting KMT2D and scrambled siRNA were bought from GE Dharmacon (# LQ-004828-00-0010, sequences see Supplementary Table 6). For siRNA transfection, a final concentration of 100 pmol siRNA for each with 1 μl DharmaFECT 2 or DharmaFECT 1 Transfection Reagent (GE Dharmacon, Lafayette, USA) was used to transfect PC-3 and DU145 cells (1 × 105), respectively. Lentivirus expressing scrambled vector and KMT2D shRNA were packaged by Genepharma (Shanghai, China, sequences see Supplementary Table 6). PC-3 cells were infected with the lentivirus at 100 MOI by polybrene (Genepharma, Shanghai, China) at 5 μg/ml. After 2–5 days incubation, transfected cells were used for further analysis.

MTT and EdU cell proliferation assay

Cells (1 × 104/well) were seeded in 96-well plates. After transfection with siRNAs for 2–4 days, the alive cells were detected by MTT assay.

At 48 h post-transfection, cell proliferation was also assessed by EdU incorporation assay (Ribobio, Guangzhou, China).

Quantitative PCR

RNA from 46 PCa samples and 14 BPH tissues was isolated by a RNeasy FFPE Kit (Qiagen, Hilden, Germany) as the manufacturer’s protocol. RNA from cells was isolated by Trizol Reagent (Invitrogen, CA, USA). The extracted RNA was quantified by Qubit fluorimeter (Thermo Fisher, MA, USA). Reverse transcription (RT) was performed with 1 μg RNA using QuantiTect Reverse Transcription Kit (Qiagen, Hilden, Germany). The cDNA was amplified with gene-specific primers (Supplementary Table 6) and SYBR Premix Ex Taq II kit (TaKaRa, Shiga, Japan). Data were analyzed using a 2−ΔΔCt method [39].

Cell culture

Human PCa cell lines PC-3, DU145 and LNCaP were purchased from cellcook biological technology Co., Ltd. (Guangzhou, China). Cells were cultured in RPMI1640 medium supplemented with 10% FBS (Thermo Fisher Scientific, Waltham, MA, USA). All cell lines were authenticated by STR profiling and tested negative for mycoplasma contamination.

Immunoblotting

Cells (3 × 105/well) were plated in 6-well plates and transfected with siControl or siKMT2D for 72 h. Whole protein lysates were prepared by RIPA buffer containing 1% PMSF and phosphatase inhibitor cocktail (Roche). As describe previously [14], the blots were probed with primary antibodies to: KMT2D (Santa Cruz, # sc-68671, 1:200), Bcl2 (CST, #2872, 1:1000), Bcl-xL (CST, #2764, 1:1000), Akt (CST, #4691, 1:1000), p-Akt (Thr308, CST, #13038, 1:1000), p-Akt (Ser473, CST, #4060, 1:1000), p-GSK-3β (CST, #5558, 1:1000), p-BRCA1 (CST, #9009, 1:1000), p-CREB1 (Santa Cruz, # sc-7978,1:200), GAPDH (CST, #2118, 1:1000).

Flow cytometry analysis

Cells (3 × 105/well) were plated in 6-well plates and transfected with siRNA vector or siKMT2D. Cell cycle distribution was analyzed with PI staining (BD Biosciences, Auckland, New Zealand); For cell apoptotic assay, cells were transfected for 48 h and assessed by Annexin V- Propidium Iodide kit (BD Biosciences, Auckland, New Zealand). The stained cells were acquired by flow cytometry (BD Biosciences, San Diego, CA, USA) and analyzed by FlowJo v7.6 software.

Xenograft tumor model

All animal studies were approved by Animal Care and Use Committee (IACUC) in Guangzhou University of Chinese Medicine. Female Balb/c-nude mice (4–6 weeks, 18–20 g) were purchased from Laboratory Animal Center of Sun Yat-Sen University (Guangzhou, China) and maintained in specific pathogen free environment in Guangzhou University of Chinese Medicine. Animals were fed with food and water freely and housed with a 12-dark:12-light cycle. Mice were randomized by weight and sample sizes were estimated according to “Resource Equation”.

PC-3 cells transfected with shRNA expressing lentivirus (n = 5) or shKMT2D (n = 5) were injected subcutaneously (6 × 106 cells per mouse) into left flanks of mice. Tumors were measured with a caliper every 4 days and tumor volumes were calculated as formula: TV (mm3) = (L × W2)/2. When the tumor volume reached 800 mm3, mice were intravenously administered with IntegriSense750 (PerkinElmer, MA, USA), which specially labeled the tumors. Then the tumor load was evaluated by fluorescence imaging using IVIS (PerkinElmer, MA, USA) after 24 h.

Wound healing assay

Cells (3 × 105/well) were cultured in 6-well plate and transfected with siControl or siKMT2D. After growing up to 100% confluence, the cell monolayer was scratched with a 10 μl pipette and incubated with 1% serum-containing medium for another 24 h. Photomicrographs were taken by contrast phase microscope (Leica, Jena, Germany) and migratory percentage was analyzed by ImageJ software.

Transwell assay

Cells (1 × 104/well) transfected with siRNAs control or siKMT2D were seeded in upper chamber of transwell plate coated with Matrigel. The upper chamber contained 1% serum medium, while 10% serum medium was added to the low chamber. After 48 hrs of incubation, invasive cells were stained with 0.1% crystal violet, photographed and quantified by colorimetry.

Immunofluorescence staining

Cells (1 × 105 /well) were plated on sterile glass cover slips in 24-well plates. After transfected with siRNAs control or siKMT2D for 72 h, cells were fixed, permeabilized and blocked. Cells were stained overnight with primary antibodies to H3K4me1 (Abcam, #ab8895, 1:500), E-cadherin (Santa Cruz, # sc-7870, 1:50), N-cadherin (Santa Cruz, # sc-7939, 1:50). Subsequently, cells were incubated with secondary fluorescent antibody conjugated with PE (Santa Cruz, 1:200) and DAPI (5 μg/ml).

RNA-seq and sequencing analysis

PC-3 cells (3 × 105/well) were seeded in 6-well plates and transfected with siControl or siKMT2D#1 or siKMT2D#2. Cells were collected and total RNA for RNA sequencing (RNA-seq) was isolated using Trizol (Invitrogen, CA, USA). RNA quantification and integrity was measured by BioAnalyzer 2100 (Agilent, CA, USA). Oligo (dT) magnetic beads (Invitrogen, CA, USA) were used to enrich mRNA. Then mRNA was fragmented, reversed into cDNA and amplified by Phusion High Fidelity DNA polymerase as previously reported [40]. Each PCR products were single-end (50 bps) sequenced on a BGISEQ-500 sequencer (BGI, Shenzhen, China).

Average 658 M raw reads were aligned to hg19 human genome by Bowtie2 [41] and HISAT [42] algorithm. Gene expression level was quantified by RSEM software [43].

We used false-discovery rate (FDR, Benjamini Hochberg method) ≤0.001 and cut-off of Log2 fold change ≥1 as the threshold to judge the significance of differential gene expression between siKMT2Ds vs. siControl. The DAVID online software were used for KEGG pathway enrichment of the selected DEGs [44]. Gene set enrichment analysis (GSEA) was also used to evaluate the related pathway affected by leading-edge genes [45].

ChIP-seq and sequencing analysis

H3K4me1 ChIP were carried out using a Pierce ChIP assay protocol (Thermo Fisher Scientific, Waltham, MA). Briefly, PC-3 cells (2 × 105/well) were plated in 6-well plates and transfected with siRNA and siKMT2D. Cells were fixed with 1% formaldehyde, lysed and digested by micrococcal nuclease. Precleared DNA were immunoprecipitated with 4 μl of H3K4me1 antibody (Abcam, Cambridge, UK) overnight and eluted. ChIP DNA was then prepared for sequencing library with fragment ranging 100–300 bps. Finally, prepared libraries were sequenced on BGISEQ-500 sequencer as 50 bps single-end read runs.

The ChIP-seq data was mapped to the hg19 human genome by SOAP2 [46]. Peaks calling was performed by SICER [47] and differential peaks were identified between siControl and siKMT2D samples by MAnorm algorithm [48]. The significant regions were picked up if |M| ≥ 1 and p-value ≤10−5. The peak-related genes were annotated by ANNOVAR software using UCSC known gene database [49]. Sequences centered on 500 bps of peak region were submitted to MEME online software to identify the consensus motif enriched in H3K4me1 [50].

Statistical analysis

Kaplan–Meier (KM) analysis was performed as previously report [51]. Disease-free survival (DFS) was defined as time from diagnosis to recurrence or to last follow-up data. Log-rank test was used to calculate the P values of KM.

Other data were statistically calculated using Student’s t-test (two groups) or one-way ANOVA followed by the LSD post-hoc test (for more than two groups) or Mann–Whitney U test (nonparametric analysis). *P < 0.05, **P < 0.01, ***P < 0.001 are determined as significance. All the experiments were performed at least in triplicates. Value presented as the means ± standard deviation (SD) by GraphPad Prism software (GraphPad Software, CA, USA).

Data availability

RNA-seq (GSE94807) and ChIP-seq (GSE94817) are available in GEO dataset. Targeted sequencing data has been deposited into SRA dataset (SRA527454).