PLEK2 was upregulated in CRC

To explore the potential role of PLEK2 in cancers, we analyzed the mRNA levels of PLEK2 with the published next-generation sequencing data from Oncomine comprising major types of human cancer with respective normal tissues [31]. Notably, PLEK2 is highly expressed in multiple types of cancer including CRC (Fig. S1A). Individual CRC datasets from Gene Expression Omnibus (GEO) databases also showed that PLEK2 was significantly upregulated in primary tumor tissues of CRC compared with adjacent normal tissues (Fig. 1A). In addition, CRC cells exhibited the increased expression of PLEK2 (Fig. S1B). In line with these findings, our immunohistochemical staining revealed that PLEK2 protein levels were also elevated in CRC tissues compared with normal colorectal tissues (Fig. 1B, C). Further analysis of CRC datasets from GEO database showed that patients with high expression of PLEK2 suffered significantly worse overall survival (Fig. 1D). These data demonstrate that PLEK2 is upregulated in CRC and indicate its potential role in the progress of CRC.

Fig. 1

PLEK2 is a potential unfavorable prognostic marker in CRC. (A) RNA-seq expression profiles of PLEK2 in colorectal cancer and paired para cancerous tissues from public databases (GSE8671, GSE117606, GSE106582, GSE44076 and GSE41657). (B-C) IHC analysis of PLEK2 in the tissues from CRC patients’ tissues microarray. Representative images of IHC staining (B) and the statistical analysis (C) is shown. (D) Kaplan–Meier plot of overall Survival of CRC patients was stratified by PLEK2 expression level (GSE33114). See also Fig. S1

Silencing of PLEK2 inhibited the CRC cell proliferation via the cellular senescence

To determine the role of PLEK2 in CRC, we performed cell proliferation assays in CRC cells HCT116 and HT29 transduced with lentivirus encoding PLEK2 shRNA (Fig. 2A). PLEK2 knockdown led to the substantial inhibition of two CRC cells proliferation (Fig. S2A). This was further confirmed by the reduced colony formation in HCT116 and HT29 cells with PLEK2 silencing (Fig. 2B). To clarify the effects of PLEK2 on the CRC cell growth, we examined the cell cycle and cell viability and found that silencing of PLEK2 led to a significant increase of cell frequency at G0/G1 phase but not obvious cell death (Fig. S2B-S2C and date not shown). Moreover, the MKi67 mRNA level was also reduced after PLEK2 knockdown (Fig. 2C). EdU assay demonstrated that PLEK2 deficiency impaired the DNA replication in HCT116 and HT29 cells (Fig. 2D, E).

Fig. 2

Silencing PLEK2 inhibited the proliferation of CRC cells. (A) Immunoblotting analysis of PLEK2 in HCT116 and HT29 cells transduced with retroviruses encoding indicated shRNAs. GAPDH was used as a loading control. shNC represents a non-targeting shRNA. (B) Representative images of colony formation assay of HCT116 and HT29 cells transduced with retroviruses encoding indicated shRNAs. shNC represents a non-targeting shRNA. (C) Quantitative PCR analysis of MKI67 mRNA level in HCT116 and HT29 cells transduced with retroviruses encoding indicated shRNAs. shNC represents a non-targeting shRNA. Data were presented as mean ± SD from three independent experiments. (D) Representative images of Edu staining in HCT116 and HT29 cells transduced with retroviruses encoding indicated shRNAs. shNC represents a non-targeting shRNA. Scale bars, 20 µm. (E) Quantification of EdU positive cells in D. Data were presented as mean ± SD from three independent experiments. (F) Representative images of SA-β-gal staining in HCT116 cells transduced with indicated retroviruses treated in the presence of Doxorubicin (0.3 μM) for 24 h. Red arrows indicated senescent cells. OE-C represents overexpression of blank vector, and shNC represents a non-targeting shRNA. Scale bars, 25 µm. (G) Quantification of SA-β-gal positive cells in F. Data were presented as mean ± SD from three independent experiments. (H) HCT116 cells transduced with retroviruses encoding indicated shRNAs, were subcutaneously transplanted into nude mice. Tumor volumes were measured every 3 days. Data were presented as mean ± SD. N = 6 mice for each group. P value was determined by two-way ANOVA. shNC represents a non-targeting shRNA. (I) Representative images of tumors from H. (J-K) Quantification of tumor sizes from the mice as in H on day 22. Each dot represents one mouse. Data were presented as mean ± SD. See also Fig. S2

Considering that cell cycle arrest and proliferative inhibition are the typical characteristics of cellular senescence, we analyzed whether PLEK2 knockdown trigger the cellular senescence in CRC cells. Senescence-associated-β-galactosidase (SA-β-gal, a biomarker of senescence) staining showed that the percentage and strength of SA-β-gal-positive cells were remarkably increased in HCT116 cells with PLEK2 knockdown (Fig. 2F, G). On the contrary, PLEK2 overexpression provided an obvious protective benefit for HCT116 cells against the doxorubicin-induced senescence (Fig. 2F, G). Similar findings were also observed in HT29 cells (Fig. S2D-S2E). In addition, PLEK2 knockdown significantly increased the sensitivity of 5-Fu on HCT116 cells (Fig. S2F).

We next used the xenograft mice model to confirm the role of PLEK2 in CRC cell proliferation in vivo. As expected, PLEK2 knockdown led to the marked reduction in tumor growth of HCT116 cells (Fig. 2H). Correspondingly, tumor volumes and weight in shPlek2 group were significantly reduced compared with that of control group (Fig. 2 2I-K and Fig. 2G).

PLEK2 is required for CRC cell migration, invasion and stemness-like properties

To further investigate the functional role of PLEK2 in CRC, we evaluated the effect of PLEK2 knockdown on the CRC cell metastasis. Wound healing assays demonstrated that PLEK2 knockdown led to the substantial decrease of the migration in HCT116 and HT29 cells (Fig. 3A, B). Furthermore, we performed the transwell assays and found that PLEK2 deficiency impaired the invasion of these cells (Fig. 3C, D).

Fig. 3

PLEK2 promotes CRC cell migration, invasion and stemness-like properties. (A) In vitro wound-healing/scratch assays with HCT116 and HT29 cells transduced with retroviruses encoding indicated shRNAs. shNC represents a non-targeting shRNA. (B) Quantitative analysis of cells migration in A. Recovered area with cells were calculated and normalized to the corresponding control group. Data were presented as mean ± SD from three independent experiments. (C-D) Invasion assays of HCT116 and HT29 cells transduced with retroviruses encoding indicated shRNAs. Representative microscopic fields of invasion cells on the bottom of transwell inserts were shown (C). Quantitation of cell invasion by counting invaded cells from five microscopic fields (D). shNC represents a non-targeting shRNA. (E) Immunoblotting analysis of PLEK2 in indicated cells. Diff represents the differentiated colorectal cancer stem cells (CCSCs) generated from CCSCs by culturing in 3% serum medium for 48 h. (F) Sphere formation activity of CCSCs transduced with retroviruses encoding indicated shRNAs. Data were presented as mean ± SD from three independent experiments. shNC represents a non-targeting shRNA. Scale bars, 100 µm (G) Immunoblotting analysis of stemness-associated proteins in HCT116 cells transduced with retroviruses encoding indicated shRNAs. shNC represents a non-targeting shRNA. (H) Representative images of CCSCs metastasis in NXG mice. CCSCs stably expressed luciferase were transduced with retroviruses encoding PLEK2 shRNAs or shNC (1 × 106/mouse), and then injected into NXG mice via the tail vein. CCSCs metastasis were quantitated using the bioluminescence imaging of luciferase after 7 weeks of injection (right). Each dot represents one mouse. Data were presented as mean ± SD. shNC represents a non-targeting shRNA. Data were presented as mean ± SD from three independent experiments. P value was determined by two-way ANOVA. See also Fig. S3

CRC stem cells (CCSCs) are a group of cells in tumor tissues that have the characteristics of stem cells such as self-renewal, multi-directional differentiation and tumorigenesis, which account for the higher metastasis and invasion of CRC [32, 33]. Notably, PLEK2 was highly expressed in colorectal CCSCs compared to the differentiated CRC cells (Fig. 3E). Silencing of PLEK2 significantly reduced the sphear number and size derived from colorectal CCSCs, indicating that PLEK2 knockdown impaired the proliferation of colorectal CCSCs (Fig. 3F). This was reversely confirmed by the increased sphear number in PLEK2-overexpression colorectal CCSCs compared with the control group (Fig. S3A-S3B). Moreover, CCSCs specific markers were also markedly down-regulated after PLEK2 knockdown, including CD44 and CD133 (Fig. 3G).

Considering that cancer stem cells are responsible for the tumor metastasis, we further evaluated the effect of PLEK2 on the CRC metastasis driven by CCSCs in vivo. CCSCs stably expressed luciferase were transduced with retroviruses encoding PLEK2 shRNA or control shRNAs, and injected into NXG mice via the tail vein (Fig. S3C). We found that PLEK2 knockdown in CCSCs significantly decreased metastatic potential to kidney, lung and liver in xenograft models after 7 weeks of injection, as demonstrated by decreased number of metastatic nodules (Fig. 3H and Fig. S3D). Histological features of metastatic tumor CCSCs in engraftment organs were characterized by HE staining (Fig. S3E). These results indicated that PLEK2 play an important role in the self-renewal and metastatic capacity of CCSCs.

PLEK2 regulated the expression of TYMS in CRC cells

To understand the underlying basis of the impaired CRC cell proliferation induced by PLEK2 deficiency, we performed bulk RNA sequencing of HCT116 cells with or without PLEK2 knockdown. A total of 306 differentially expressed genes (DEG) were identified (≥ 1.5-fold, P < 0.05) (Table S1), including genes related to cell cycle and cell senescence (Fig. 4A, B). In parallel, we also performed the quantitative proteomic analysis and found that p21 and TYMS were the most changed proteins after PLEK2 knockdown (Fig. S4A). In line with this, p21 and TYMS mRNA levels were significantly altered in HCT116 cells with PLEK2 deficiency (Fig. 4C).

Fig. 4

PLEK2 knockdown led to cell senescence via the regulation of TYMS. (A) Volcano plot for differentially expressed genes in HCT116 cells transduced with retroviruses encoding indicated shRNAs. shNC represents a non-targeting shRNA. |Fold Change|> 1.5, P < 0.05. (B) Gene ontology enrichments of differentially expressed genes in A using DAVID bioinformatics database. (C) Quantitative PCR analysis of p21 and TYMS mRNA levels in HCT116 and HT29 cells transduced with retroviruses encoding indicated shRNAs. Data were presented as mean ± SD from three independent experiments. shNC represents a non-targeting shRNA. (D) Western blotting analysis of indicate proteins in HCT116 cells transduced with retroviruses encoding indicated shRNAs. GAPDH was used as a loading control. shNC represents a non-targeting shRNA. (E) Western blotting analysis of indicate proteins in HCT116 cells with retroviruses encoding TYMS shRNAs. β-Tubulin was used as the loading control. shNC represents a non-targeting shRNA. (F) Immunoblotting analysis of indicated proteins in HCT116 cells with PLEK2 shRNA and the presence of TYMS overexpression. β-Tubulin was used as the loading control. OE-C represents overexpression of blank vector. shNC represents a non-targeting shRNA. (G) Representative images and quantification of SA-β-gal staining positive cells in HCT116 cells. Scale bar, 25 μm. (H) Quantification of cell proliferation in HCT116 cells as in F. P value was determined by two-way ANOVA. See also Fig. S4

KEGG enrichment analysis revealed that p53 signaling pathway was substantially enriched in the PLEK2-shRNA group (Fig. S4B). Indeed, PLEK2 knockdown led to the increased phosphorylation of ATM and p53 and subsequent p21 expression (Fig. 4D). To determine whether PLEK2 regulated p21 and TYMS expression in a p53-independent manner, we examined the effect of PLEK2 on the HCT116-p53−/− cells, a p53-null originating from HCT116 cells. Notably, p53 deletion largely abolished the upregulation of p21 but not the reduced TYMS induced by PLEK2 knockdown (Fig. S4C-S4D). Indeed, we found that TYMS knockdown led to the obvious upregulation of p21 and increase of SA-β-gal-positive cells in HCT116 cells (Fig. 4E and Fig. S4E).

To confirm the correlation between PLEK2 and TYMS, we analyzed their mRNA levels with the published next-generation sequencing data from CRC patient samples. PLEK2 expression was significantly correlated with the highly-exressed TYMS in human samples (Fig. S4F). Furthermore, ectopic expression of TYMS successfully reverted the upregulation of p21 and cell senescence in HCT116 cells with PLEK2 silencing (Fig. 4F, G). TYMS overexpression also significantly restored the impaired HCT116 cell proliferation induced by the PLEK2 knockdown (Fig. 4H). These data indicate that PLEK2 regulated cell proliferation and senescence of HCT116 via the expression of TYMS.

PLEK2 interacted with YTHDF2 to regulate the TYMS mRNA stability

To determine how PLEK2 regulates TYMS expression, we first examined the effects of PLEK2 on the TYMS protein stability. PLEK2 knockdown exhibited a comparable degradation of TYMS upon the cycloheximide (CHX) treatment in the CRC cells (Fig. S5A). The reduced TYMS proteins induced by PELI1 knockdown was also not reversed by the pre-treatment of MG132, a proteasome inhibitor (Fig. S5B).

We then evaluate its mRNA expression upon the treatment of actinomycin D. TYMS showed a faster time-dependent decay in HCT116 cells transduced with PLEK2 shRNA than that of control groups (Fig. 5A). Considering that PLEK2 has been identified as a membrane and cytoskeletal protein[23], we speculated that other proteins interacts with PLEK2 to regulate TYMS mRNA stability. To this end, we performed immunoprecipitation of PLEK2 followed by mass spectrometry in HCT116 cells (Fig. 5B and Table S2). Among these identified proteins, YTHN6-Methyladenosine RNA Binding Protein 2 (YTHDF2) is an attractive candidate due to its role in the regulation of RNA stability in a m6A-dependent manner[34]. PLEK2 and YTHDF2 were reciprocally coimmunoprecipitated with each other in HEK293T cells with their overexpression (Fig. S5C). This was further confirmed by the co-IP assays of endogenous PLEK2 and YTHDF2 in HCT116 cells (Fig. 5C). Furthermore, GST pull-down assays showed that YTHDF2 were specifically retained in the presence of GST-PLEK2, confirming their physical interaction (Fig. 5D).

Fig. 5

PLEK2-YTHDF2 complex promoted the TYMS transcript stability in a m6A‐dependent manner. (A) Quantitative PCR analysis of TYMS mRNA levels in HCT116 cells with PLEK2 knockdown upon the treatment of actinomycin D (10 μg/ml). Data were presented as mean ± SD from three independent experiments. P value was determined by two-way ANOVA. shNC represents a non-targeting shRNA. (B) Proteomic study of PLEK2 interacting proteins in HCT116 cells transduced with retroviral constructs encoding HA-PLEK2 (OE-PLEK2) or empty vector (OE-C). Proteins immunoprecipitated using anti-HA were resolved by SDS-PAGE and visualized by silver staining followed by mass spectrometry analysis. (C) Co-IP analysis of endogenous YTHDF2 binding to PLEK2 in HCT116 cells. (D) GST pull-down assay of the physical interaction between PLEK2 (GST-Tagged) and YTHDF2 (His-Tagged) in vitro. (E) Quantitative PCR analysis of TYMS transcript in HCT116 cells transduced with retroviruses encoding indicated shRNAs. Data were presented as mean ± SD from three independent experiments. shNC represents a non-targeting shRNA. (F) Pie chart depicting the region distribution of PLEK2-binding sites identified by RIP-Seq. (G) Schematic representation depicting the Luciferase reporters herbing indicated regions of TYMS mRNA based on the pCDH plasmid. 5’UTR: 5’-untranslated regions, CDS: coding sequence, 3’UTR: 3’-untranslated regions. (H) EMSA assay showed direct binding of PLEK2 to the CDS region of TYMS transcripts as in G. (I) RIP analysis the 3’UTR of TYMS in HCT116 cells using YTHDF2 antibodies. Data were presented as mean ± SD from three independent experiments. (J) Immunoblotting of indicated proteins in HCT116 cells after RNA pull-down assay using single-stranded TYMS RNA with methylated (ss-A) or unmethylated adenosine (ss-M6A). #1-#6 indicated different sites of TYMS RNA as in Fig. S5J. (K) Quantitative PCR analysis of TYMS transcript in m.6A-MeRIP from HCT116 cells transduced with retroviruses encoding indicated shRNAs. Data were presented as mean ± SD from three independent experiments. shNC represents a non-targeting shRNA. (L) HCT116 cells carrying Luciferase-fusion CDS and 3’UTR regions of TYMS as in F were transduced with indicated lentivirus. Luciferase mRNA levels were analyzed after 48 h. OE-C represents overexpression of blank vector. Data were presented as mean ± SD from three independent experiments. See also Fig. S5

Indeed, YTHDF2 is also highly expressed in CRC (Fig. S5D). Silencing of YTHDF2 led to the downregulation of TYMS via mRNA stability (Fig. 5E and Fig. S5E) and even the inhibitory proliferation of HCT116 cells (Fig. S5F), which phenocopied the effects of PLEK2 knockdown on the HCT116 cells. However, TYMS did not show the further reduce in HCT116 cells transduced with PLEK2 shRNAs upon the YTHDF2 knockdown (Fig. 5E). Similar findings were also observed in the proliferation assays of HCT116 cells (Fig. S5F-G). Moreover, we found that TYMS RNA stability were also impaired by the silencing of METTL14, one of critical components of RNA m6A modification as the “writer” (Fig. S5H). These data suggest that PLEK2 interacted with YTHDF2 to promote TYMS stability and proliferation of CRC cells in a m6A-dependent manner.

PLEK2 and YTHDF2 cooperated to promote the TYMS mRNA stability in a m6A-dependent manner

Based on the above findings, we investigated whether PLEK2 function as an RNA-binding protein in CRC cells. We performed RNA immunoprecipitation sequencing using HA antibody in HCT116 cells with overexpression of HA-tagged PLEK2, and identified 1516 potential PLEK2-binding targets (Table S3). These RNA targets were substantially enriched in RNA transport/degradation and cell cycle pathways (Fig. S6A). Moreover, most of the PLEK2-binding sites were highly enriched on the CDS region of RNA targets including TYMS (Fig. 5F and Fig. S6B). To further confirm the direct binding of PLEK2 with TYMS mRNA, we constructed pCDH-Luciferase-based reporters bearing 5’-untranslated regions (5’UTR), coding sequence (CDS) and 3’UTR of TYMS mRNA respectively and incubated with PLEK2 protein in vitro (Fig. 5G). As expected, electrophoretic mobility shift assay results revealed that PLEK2 preferentially bound to the CDS region of TYMS mRNA (Fig. 5H).

We then analyzed the published RIP-seq data of YTHDF2 in HeLa cells [34], and found that 35.4% (537 out of 1,516) of the RIP targets of PLEK2 overlapped with that of YTHDF2 (Fig. S6C). Although TYMS is the common target of both proteins, YTHDF2 showed exclusive binding to the 3’UTR of TYMS mRNA that was demonstrated by RIP-qPCR analysis in HCT116 cells (Fig. 5I). Indeed, several potential m6A sites were observed in the TYMS mRNA containing the “DRACH” (D: A\G\U, R = A\G, H = A\U) consensus sequence [35] (Fig. S6D). To confirm that the binding of YTHDF2 to TYMS mRNA is m6A-dependent, we designed the methylated single-stranded RNA bait (ss-m6A) or unmethylated control RNA (ss-A) according to these potential m6A sites on the TYMS mRNA. RNA pull-down assay demonstrated that YTHDF2 specifically bound to two ss-m6A but not the corresponding unmethylated controls (ss-A) in the 3’UTR region of TYMS (Fig. 5J). Furthermore, we also performed m6A-MeRIP combined with METTL14 knockdown. Quantitative PCR analysis of TYMS with primers targeting YTHDF2-binding sites showed that the enrichments of m6A sites were significantly reduced by METTL14-silencing in HCT116 cells (Fig. 5K). These data demonstrated that PLEK2/YTHDF2 binds to TYMS mRNA via an m6A-dependent manner.

To further confirm that the binding of PLEK2/YTHDF2 with TYMS mRNA mediates its stability, we constructed the pCDH-Luciferase-based reporters bearing CDS and 3’UTR of TYMS mRNA (pCDH-Luci-CDS-3'UTR, Fig. 5G). Single overexpression of PLEK2 or YTHDF2 mildly increased the luciferase activity in HCT116 cells transduced with pCDH-Luci-CDS-3'UTR, which was further enhanced by their combination (Fig. 5L). By contrast, similar approaches failed to detect any increase of luciferase activity in HCT116 cells transduced with pCDH-Luci-CDS or pCDH-Luci- 3'UTR (data not shown).

Loss of PLEK2 inhibited AOM/DSS-induced colonic tumorigenesis in vivo

To clarify the functional roles of PLEK2 in the tumorigenesis and development of CRC, we crossed Plek2fl/fl mice with Vil1-cre transgenic mouse to generate an intestinal-specific knockout mouse model (Fig. S7A-D, referred to as CKO mice). Plek2 deletion did not affect the intestinal function characterized by the comparable body and intestine weigh in CKO mice as WT mice (data not shown). We then utilized an AOM/DSS-induced colorectal tumor model, in which CKO or wild-type mice were given a single dose of the carcinogen azoxymethane (AOM) plus 3 cycles of 2.5% dextran sulfate sodium (DSS) (Fig. 6A). As expected, the body weight was reduced after DSS administration. However, CKO mice showed less susceptible to chronic inflammation, which was induced by AOM and DSS treatment, as compared to that of WT mice (Fig. 6B). Mice were sacrificed 72 days after AOM-DSS administration to analyze colonic tumor incidence and burden. Although there was no significant difference in colon length (Fig. 6C, D), the number of tumors was significantly reduced in CKO mice than that in WT mice (Fig. 6E, F). These results suggested that PLEK2 deficiency suppresses the colonic tumorigenesis in the colon.

Fig. 6

Plek2 deletion reduced the intestinal tumorigenesis. (A) Schematic representation of inflammatory colorectal cancer mouse model with wild type (WT, N = 6) or Plek2 conditional knockout (CKO, N = 4). (B) Statistical analysis of body weight from the mice as in A. (C) Representative image of colon in A. (D) Statistical analysis of intestines weight from the mice as in C. Each dot represents one mouse. Data were presented as mean ± SD. (E) Methylene blue staining of the colorectum from wild type (WT) and Plek2-CKO mice. (F) Statistical analysis of the numbers of intestinal polyps from the mice as in E. Each dot represents one mouse. Data were presented as mean ± SD. See also Fig. S6