PRL1 is overexpressed in GBM tissues and cell lines
PRL1 expression levels were first analyzed in 163 GBM tissues and 207 NBTs using GEPIA (http://gepia.cancer-pku.cn/). As shown in Fig. 1a, PRL1 was significantly upregulated in the GBM specimens compared to NBTs. To determine whether PRL1 plays a role in the invasiveness and progression of GBM, we analyzed PRL1 protein levels in 10 NBTs and 26 grade IV GBM samples using Western blot, and detected markedly high levels in the tumors in a subset of NBT/GBM tissues (Fig. 1b, quantification not shown). In a cohort of 62 glioma specimens, PRL1 protein expression levels were overall positively correlated with the clinical grading (Fig. 1c, d). Consistent with these findings, PRL1 protein was also found significantly upregulated in multiple GBM cell lines (U87MG, U251, LN229 and T98G) compared to the normal human astrocytes (NHAs) and grade III glioma cell lines (HS683 and SW1783) (Fig. 1e). Taken together, these data suggest that PRL1 likely plays an oncogenic role in GBM.
PRL1 increased glioma cell invasion, migration and tumor formation by promoting EMT
PRL1 plays a vital role in the invasion and metastasis of various tumors [18]. To determine whether PRL1 overexpression in glioma is correlated to their invasiveness and migration ability, we ectopically expressed this protein in HS683 and SW1783 cells (Supplementary Fig. S1). As shown in Fig. 2, overexpression of PRL1 significantly enhanced their invasion through Matrigel (Fig. 2a and 2b), and increased in vitro migration and wound coverage (Fig. 2c and 2d). EMT is a known driver of cancer cell invasion and migration [36], and is characterized by the upregulation of epithelial markers and downregulation of mesenchymal proteins. We found that overexpression of PRL1 significantly increased N-cadherin and vimentin levels, and decreased the level of E-cadherin (Fig. 2e). Consistent with these in vitro findings, PRL1-overexpressing HS683 cells grew more rapidly in nude mice and formed significantly larger tumors compared to the control cells (Fig. 2f-h). Furthermore, PRL1 overexpression increased the levels of N-cadherin and vimentin, and decreased that of E-cadherin in the tumor sections (Fig. 2i and 2j). Taken together, these data indicate that PRL1 may promote the invasion and migration of glioma cells by accelerating EMT.
PRL1 knockdown in GBM cells blocked EMT and inhibited the malignant phenotype
To further explore the mechanistic role of PRL1 in regulating GBM cell invasion and migration, we knocked down PRL1 in U87MG and U251 cells (Fig. 3a). PRL1 silencing significantly decreased N-cadherin and vimentin levels, and increased that of E-cadherin (Fig. 3a). The transwell-Matrigel invasion assay (Fig. 3b and 3c) and wound healing assay (Fig. 3d and 3e) further confirmed that PRL1 knockdown significantly reduced the invasiveness and migration capacity of the glioma cells respectively. To assess the effect of PRL1 downregulation in vivo, luciferase-expressing U87MG cells were transfected with a lentiviral vector containing small hairpin interference RNA against PRL1 (shPRL1) or control scrambled sequence (shCtrl) and intracranially injected into nude mice. Compared to the control group, knocking down PRL1 significantly inhibited tumor growth and prolonged the survival of tumor-bearing mice (Fig. 3f-h). Furthermore, in-situ N-cadherin and vimentin expression levels were markedly lower, and E-cadherin was upregulated in the PRL1-knockdown xenografts compared to the control (Fig. 3i and 3j). These results strongly suggest that loss of PRL1 may impair EMT in glioma cells.
PRL1 knockdown increased Snail2 polyubiquitination and proteasomal degradation
EMT is regulated at the molecular level by several transcription factors such as Snail1, Snail2, Twist1, Twist2, ZEB1 and β-catenin [37]. As shown in Fig. 4a, we found that only Snail2 levels were markedly decreased in the PRL1-kncokdown U87MG and U251 cells lines, whereas the other EMT-related transcription factors were unaffected. As such, we hypothesized that PRL1 regulates EMT in the GBM cells by activating Snail2. However, no significant differences were seen in the levels of Snail2 mRNA (Fig. 4b and 4c) between the control and PRL1-kncokdown cells in both U87MG and U251 lines, which suggests that PRL1 is less likely to regulate Snail2 at the transcriptional level. We next treated PRL1-knockdown or shCtrl U87MG and U251 cells with the proteasome inhibitor MG132, and found that the latter restored the levels of Snail2 protein in cells lacking PRL1 (Fig. 4d).
To further ascertain whether PRL1 affects the stability of Snail2 protein, we treated the control and PRL1-knockdown U87MG/U251 cells with the protein synthesis inhibitor cycloheximide (CHX, 100µg/ml). As shown in Fig. 4e, the half-life of Snail2 was significantly shortened in cells lacking PRL1, indicating that loss of PRL1 may accelerate the degradation of Snail2. Indeed, the combination of PRL1 knockdown and CHX treatment rapidly decreased Snail2 protein levels (Fig. 4e and 4f). It has been previously shown that the ubiquitin proteasome system (UPS) controls the degradation and turnover of multiple target proteins including those of the Snail family [1, 38]. To determine whether PRL1 mediates Snail2 proteolysis via the UPS pathway, we co-transfected the PRL1-knockdown or control U87MG cells with plasmids encoding HA-tagged ubiquitin and His-tagged Snail2, and treated them with 20 µM MG132. Silencing PRL1 in both cell lines significantly decreased Snail2 protein levels and enhanced its ubiquitination compared to control cells (Fig. 4g). Taken together, these findings indicate that PRL1 stabilizes Snail2 in GBM cells through polyubiquitination and targeted proteasome degradation.
PRL1 stabilizes Snail2 by activating USP36
Deubiquitinating enzymes (DUBs) are a broad group of proteases that directly remove ubiquitin groups from proteins, thereby regulating ubiquitin-dependent signaling pathways [30, 38]. To identify potential DUBs involved in Snail2 protein degradation, we silenced the existing 98 DUBs in HEK293T cells using the DUB siGENOME RTF library siRNAs. Knocking down USP3, USP20, USP36, USP50 and USP52 significantly reduced Snail2 protein levels by more than 2-fold (Fig. 5a; Supplementary Fig. S2). However, only USP36 mRNA levels were significantly downregulated in the PRL1-knockdown cells (Fig. 5b) and conversely upregulated in the PRL1-overexpressing cells (Fig. 5c), whereas the other DUB mRNAs were unaltered in either conditions. Previous studies have shown that USP5, USP10 and USP20 can stabilize Snail2 expression levels in various solid tumors [26, 38, 39]. In contrast, USP5 and USP10 mRNA levels were unaffected by changes in PRL1 expression (Supplementary Fig. S3a and 3b). Based on these results, we hypothesized that PRL1 protects Snail2 from ubiquitin-mediated degradation by activating USP36. We transfected HEK293 cells with Flag-tagged wild-type (WT) or the catalytically-inactive C131A mutant USP36, and found that overexpression of USP36-WT, but not USP36-C131A, elevated Snail2 protein in a dose-dependent manner (Fig. 5d). Furthermore, USP36 depletion significantly decreased Snail2 protein levels in the U87MG and U251 cells, which was restored by overexpression of USP36-WT but not USP36-C131A (Fig. 5e). To determine whether USP36 directly interacts with Snail2, HEK293T cells were co-transfected with Flag-tagged USP36-WT or USP36-C131A along with His-tagged Snail2. Co-immunoprecipitation (Co-IP) confirmed that both wild type and mutant USP36 bound to Snail2 (Fig. 5f), indicating that the DUB activity of USP36 was independent of its interaction with Snail2. Furthermore, Co-IP also demonstrated a direct physical interaction between endogenous USP36 and Snail2 proteins in U87MG and U251 cells (Fig. 5g and 5h). Taken together, these data suggest that PRL1 stabilizes Snail2 protein in GBM cells by activating USP36.
USP36 stabilizes Snail2 through deubiquitination
To determine whether USP36 directly deubiquitinates Snail2, we co-transfected HEK293 cells with His-Snail2, HA-ubiquitin and Flag-USP36 (WT or C131A). IP of MG132-treated cells with the anti-Snail2 antibody indicated heavy ubiquitination of Snail2, which was almost completely abolished by overexpression of USP36-WT but not USP36-C131A (Fig. 6a). On the other hand, knocking down USP36 significantly increased Snail2 polyubiquitylation in U87MG and U251 cells (Fig. 6b). Lys48- or Lys63-linked chains are the two major forms of polyubiquitin chains. Lys48-linked ubiquitin chains serve as the main targeting signals for the proteasome, whereas lys63-linked ubiquitin chains are involved in the endosomal/lysosomal-dependent degradation pathway [40, 41]. We found that overexpression of USP36 effectively disassembled Lys48-linked polyubiquitylation of Snail2 in U87MG and U251 cells, but had no significant effect on Lys63-linked polyubiquitylation (Fig. 6c and 6d). In addition, forced expression of a Lys48-resistant (Lys48R) form of ubiquitin in USP36-knockdown U87MG and U251 cells restored Snail2 levels (Fig. 6e and 6f). Taken together, Lys48-linked polyubiquitination plays a vital role in USP36-mediated de-ubiquitination of Snail2.
The oncogenic function of PRL1 in GBM cells is mediated by Snail2
To further clarify the role of Snail2 in PRL1-induced GBM progression, Snail2 was overexpressed in U87MG and U251 cells co-transfected with shPRL1. As shown in Fig. 7a, Snail2 overexpression reversed changes in EMT biomarkers induced by PRL1 knockdown. Furthermore, the inhibitory effects of PRL1 silencing on in vitro invasion (Fig. 7b and 7c) and migration (Fig. 7d and 7e) of the GBM cell lines were also abrogated by the overexpression of Snail2. Consistent with these observations, the loss of in vivo tumorigenicity of PRL1-knockdown cells was recovered in the presence of ectopic Snail2, as measured in terms of orthotopic xenograft growth and the survival of tumor-bearing mice (Fig. 7f-h). Finally, N-cadherin and vimentin proteins were upregulated, while E-cadherin was downregulated in tumors overexpressing Snail2 in the absence of endogenous PRL1 (Fig. 7i). Taken together, these data suggest that Snail2 is a key mediator of the oncogenic effects of PRL1 in GBM.
PRL1 is positively correlated with Snail2 and predicts poor outcome of GBM
To assess the clinical relevance of our findings, 26 glioblastoma patients’ specimens were immune-stained for PRL1, USP36 and Snail2. There was a strong positive correlation between the in-situ expression levels of PRL1, USP36 and Snail2 in these tissues (Fig. 8a-c), which was further corroborated by analyzing the total protein extracts of 26 freshly collected GBM tissue samples (Fig. 8d and 8e). In addition, Kaplan-Meier survival analysis indicated that patients with high expression of PRL1 (n = 13) had worse disease-free and overall survival compared to the PRL1low group (n = 13) (Fig. 8f and 8g). These results confirm the clinical relevance of the PRL1/USP36/Snail2 axis, and establish PRL1 as a potential prognostic marker for GBM patients.