Functional genomics identified a novel protein tyrosine phosphatase receptor type f-mediated growth inhibition in hepatocarcinogenesis


  • Potential conflict of interest: Nothing to report.

  • Supported in part by a grant from the Chang Gung Memorial Hospital (CMRPG3C0021), a grant from the National Research Program for Biopharmaceuticals (100CA1026), and a grant from the National Health Research Institute (NHRI-EX102-10013BI).


It is unclear how proliferating cells elicit suppression on cell proliferation and how cancer cells evade this growth suppression. Using a loss-of-function screening of the human kinome and phosphatome to identify genes suppressing tumor initiation in human hepatocellular carcinoma (HCC), we identified 19 genes and characterized one of the top-scoring tumor suppressor candidates, protein tyrosine phosphatase receptor type F (PTPRF). We found that PTPRF was induced during cell proliferation by cell-cell contact. Ectopic expression of wild-type PTPRF, but not the phosphatase-inactive mutant, suppressed cell proliferation and colony formation in soft-agar assays. In contrast, PTPRF silencing led to cell hyperproliferation, enhanced tumor colony formation in soft agar, and increased xenograft tumor growth in nude mice. Mechanistically, PTPRF silencing showed aberrant ERK-dependent signaling including the phosphorylation/stabilization of v-myc avian myelocytomatosis viral oncogene homolog (MYC) through the direct activation of v-src avian sarcoma viral oncogene homolog (SRC) and suppression of PP2A. This PTPRF-mediated growth suppression during cell proliferation functioned independently of the Hippo-Yap pathway. Clinically, PTPRF was down-regulated in 42% HCC (37/89), 67% gastric cancer (27/40), and 100% colorectal cancer (40/40). PTPRF up-regulation was found in 24% HCC (21/89) and associated with better clinical outcomes. Conclusion: A novel PTPRF-mediated growth suppression pathway was identified by way of a functional genomics screening in human hepatoma cells. Induction of PTPRF by cell-cell contact during cell proliferation quenched the activated ERK-dependent proliferation signaling to prevent cell hyperproliferation and tumor initiation. PTPRF down-regulation in HCC facilitated tumor development. Our findings shed light on how cancer cells can evade growth suppression and open a new avenue for future development of anticancer therapies. (Hepatology 2014;59:2238–2250)


epithelial growth factor


EGF receptor


extracellular signal-regulated kinase1 and 2


hepatocellular carcinoma




v-myc avian myelocytomatosis viral oncogene homolog






protein phosphatase 2A


protein tyrosine phosphatase receptor type F


RNA interference


short-hairpin RNA


small interference RNA


v-src avian sarcoma viral oncogene homolog


tumor suppressor gene

Cell growth and growth suppression are tightly regulated to ensure tissue homeostasis and to maintain normal tissue function. Aberrant activation of growth signaling and/or loss of growth suppression can lead to hyperplasia and tumor transformation.[1] Although the dysregulation of growth signaling pathways in human cancers has been well studied, it is still unclear how proliferating cells elicit suppression signals or how cancer cells evade this growth suppression to achieve unrestricted proliferation, a process that directly contributes to tumor initiation.

Protein phosphorylation is tightly regulated by protein kinases and phosphatases. Aberrant protein phosphorylation or dephosphorylation has led to the development of many human diseases including cancers.[2] Many protein tyrosine kinases are proto-oncogenes and some protein phosphatases are thought to act as tumor suppressors in human cancers.[3] However, there are no systematic studies investigating the roles of all human kinases (kinome) and phosphatases (phosphatome) in human cancers, in particular hepatocellular carcinoma (HCC).

HCC is the fifth most common human cancer and the leading cause of cancer death in the world. Since most patients present with advanced HCC at the initial diagnosis, curative therapies are extremely limited and there are currently no effective chemotherapies or targeted regimens against advanced HCC. The 5-year survival rate of HCC is no higher than 50% and the disease-free survival rate is as low as 30%.[4, 5] Gaining a more comprehensive understanding of the tumorigenic process in HCC may therefore have a great impact on HCC treatment.[6, 7]

Herein, we report the human kinome/phosphatome RNA interference (RNAi) screen in human hepatoma cells and identify 19 genes that might suppress tumor initiation. We identified and characterized a novel protein tyrosine phosphatase receptor type F (PTPRF)-mediated growth suppression pathway during cell proliferation to prevent hyperproliferation and tumor formation in human hepatoma cells.

Materials and Methods

Screening for Tumor Suppressor Genes

A short-hairpin RNA (shRNA) library targeting human kinases and phosphatases originated from an international collaboration with the Massachusetts Institute of Technology and Broad Institute of Harvard University (the RNAi Consortium) was obtained from the National RNAi Core, Taiwan ([8] The contents of this library and the detailed procedures used to screen for tumor suppressor genes (TSGs) in the human nontumorigenic HepG2 cell line are detailed in the Supporting Information.

Hepatoma Cells, siRNAs, Plasmids, Growth Factors, and Antibodies

Please refer to the Supporting Information. In this study we used shRNA lentivirus at the multiplicity of infection (MOI) of 2 to transduce hepatoma cells. For gene silencing, 5-10 nM small interference RNA (siRNA) was used for transfection.

Cell Transfection, Cell Proliferation Assays, Anchorage-Independent Colony Formation in Soft Agar, and Xenograft Tumor Formation in Nude Mice

Cell transfection, cell proliferation assays, anchorage-independent colony formation in soft agar, and xenograft tumor formation in nude mice were performed using a modification of a previously described method.[9] The procedures are detailed in the Supporting Information.

Western Analysis, Immunofluorescence, In Situ Proximity Ligation Assays, and Reciprocal Immunoprecipitation Assays

Western analysis, immunofluorescence analysis, in situ proximity ligation assays, and reciprocal immunoprecipitation assays were performed using a modification of a previously described method.[9, 10] The procedures are described in the Supporting Information.

Patients, Tissue Arrays, and Immunohistochemistry

All specimen collection procedures were approved by the Internal Review Board for Medical Ethics of Chang Gung Memorial Hospital. Tissue arrays containing 45 cases of hepatitis B virus (HBV)-associated HCC (including tumor stages I, II, and III) and 45 cases of hepatitis C virus (HCV)-associated HCC (including the pathological TNM tumor stages I, II, and III) were obtained from the Taiwan Liver Cancer Network. Informed consents were collected at the same time as the tissue sample collection and recorded in the Taiwan Liver Cancer Network. These patients had received curative hepatectomy during the period from 2005 to 2008. The mean follow-up duration for these patients was 41 months, ranging from 2 to 106 months. All HCC tissues were histologically reviewed and the most representative areas of embedded tissue samples were carefully selected and sampled for the tissue microarray blocks using a 1.5 mm diameter stylet. Two core samples were selected from different areas of each HCC tissues. Sections of 4 μm thickness were cut from each tissue array block, deparaffinizied, and dehydrated. The immunohistochemistry (IHC) scores were determined by the intensity of the positive signal for PTPRF expression in hepatoma cells by IHC: score 0: negative; score 1: weakly positive; score 2: moderately positive; score 3: strongly positive. The IHC scores were determined by two independent observers (R.B. and S.Y.H.). The slides were reexamined when there was disagreement between the observers until a consensus was reached.

Statistical Analysis

The chi-square test was used for comparison between two categorical variables. HCC staging was according to the TNM/AJCC and BCLC staging systems.[11] Correlation of PTPRF expression with clinical manifestations were conduced by chi-square test or Fisher's exact test as the sample size was smaller than 10. For the continuous variables, we used a nonparametric Mann-Whitney test or Kruskal-Wallis test. Independent variables potentially related to tumor recurrence or death were first identified by univariate methods, followed by a multivariate analysis of the identified variables using the Cox proportional hazard regression model. Kaplan-Meier analysis and the log-rank test were used to illustrate the differences in time-to-recurrence (TTR) and overall survival probabilities for each potential risk factor after patients underwent primary curative hepatectomy. In our TTR analysis, we defined recurrence as the first event of treatment failure. The data for all other patients were censored at the date of the last follow-up visit and death from causes other than hepatoma and any subsequent recurrence of hepatoma. Data were analyzed from the date of surgery to the time of the first event or to the date when data were censored (according to the Kaplan-Meier method) and the curves were compared using the log-rank test. The log-rank test was used to compare the patients with and without PTPRF up-regulation at each observed event (tumor recurrence for TTR, or patient's death for overall survival) time, and the observed and expected number of events in each group at each observed event time was calculated and added to obtain an overall summary across all the points where there was an event.


Kinome/Phosphatome RNAi Screening for the Genes Suppressing Tumor Initiation in Human HCC

To identify novel TSGs in HCC, we conducted a human kinome and phosphatome RNAi screen using a lentivirus-based shRNA library. The library contained 6334 shRNA clones targeting 1,228 genes that encode almost all human kinases and phosphatases ( We used a nontumorigenic human hepatoma cell line, HepG2 (ATCC HB-8065), as an HCC model for synthetic lethality screening. HepG2 cells harbor the wild-type p53 and low PTEN expression[12, 13] and are not tumorigenic in nude mice or in soft-agar assays.[14] The human kinome and phosphatome shRNA library was divided into seven pools, which were separately transduced into HepG2 cells (Fig. 1A). Tumor formation was assayed by measuring colony formation in soft agar and xenograft-tumor development in athymic nude mice. We anticipated that the silencing of potential TSGs would lead to colony and tumor formation in soft-agars and nude mice respectively (Fig. 1B,C). Our controls, the parental HepG2 and luciferase shRNA-transformed HepG2 cells did not develop colonies in soft-agar assays or tumor in the xenografted nude mice, as expected (Fig. 1B,C). The shRNA sequences retrieved from these colonies and xenograft tumors were then used to identify their target genes. In the primary screening, 29 genes were identified as candidate TSGs (Supporting Table S1). Five different shRNA clones targeting each of these candidate TSGs were then used separately to verify their tumor suppressor activity and to rule out the off-target effects. Only those genes with at least two different shRNA clones showing positivity for tumors in both soft-agar and xenograft assays were regarded as TSGs. A total of 19 genes were identified (Fig. 1B,C; Supporting Table S2), including two known TSGs (PTPRO, CDKN2D), four candidate TSGs (PTPRF, PPPICB, MAP3K1, GUCY2F), two oncogenes (KIT, CHKA), four candidate oncogenes (ICK, MADD, TEP1, PKM2), and seven genes not previously reported as either TSGs or oncogenes (AMHR2, CAMK2N1, CDKL4, FIGN, MUSK, PKLR, PTPN22). The shRNA knockdown efficiencies in these TSGs were confirmed using quantitative reverse-transcription polymerase chain reaction (RT-PCR) (Supporting Fig. S1).

Figure 1.

Functional genomics approach for identifying TSGs in human hepatoma cells. (A) Schematic shows the loss-of-function screen. (B) Validation results of potential TSGs using soft-agar colony formation (>100 μm). Experiments were performed twice in duplicate. Error bars represent SD. Cells transduced with shRNAs targeting the luciferase gene (shLuc) were used as negative control. Insert images show colonies treated by shRNAs targeting luciferase gene (shLuc), PTPRO (shPTPRO), and CDKN2D (shCDKN2D) (×40 magnification). (C) TSGs validation by way of xenograft tumor development in nude mice. An equal number (2 × 106) of HepG2 cells expressing shRNAs targeting candidate TSGs and shLuc were inoculated into each nude mouse. Tumor volumes were measured weekly for up to 8 weeks and growth curves were generated.

To deduce common cellular pathways or functions shared in these identified TSGs, we conducted MetaCore GeneGO analysis. Biological functions involved in the regulation of the cell cycle, cell death, immune response-regulated signaling, development, differentiation, and carbohydrate metabolism were significantly enriched when these potential TSGs were silenced (Supporting Table S3). We also found that many of these identified TSGs were the targets of transcription factors E2F1 (CAMK2N1, KIT, FIGN, MADD, CHK1, PKM2, TEP1, CDKN2D) and MYC (FIGN, PTPRF, PKM2, TEP1, CHKA, CDKN2D, MADD, MAP3K1; Supporting Fig. S2), which highlighted the significant roles of MYC and Rb-E2F1 pathways in hepatocarcinogenesis.

Since KIT and MADD are well-known oncogenes in other human cancers,[15, 16] we confirmed their tumor suppressor roles in human HCC by transducing shRNAs targeting KIT and MADD into a low tumorigenic hepatoma cell line, Hep3B. We showed that KIT or MADD silencing significantly promoted colony formation in soft agar and xenograft tumor growth in nude mice (Supporting Fig. S3), suggesting that KIT and MADD function as TSGs in Hep3B as well as HepG2 cells. Array-based comparative genomic hybridization (CGH) revealed that KIT and MADD were deleted in 16% and 10% of human HCCs, respectively (Supporting Table S2), further supporting the hypothesis that KIT and MADD are TSGs in HCC.

Suppression of Cell Hyperproliferation and Tumor Initiation by PTPRF

Of the identified TSGs, we focused on PTPRF because it represented one of the top-scoring TSGs from our screening assays (Supporting Table S1). To validate the tumor suppressor activity of PTPRF, we ectopically expressed or silenced PTPRF expression in several human hepatoma cell lines and examined for cell proliferation by cell count and tumor formation ability in soft-agar assay. We also used a PTPRF phosphatase inactive mutant in our assays to test if the effect seen was due to PTPRF phosphatase activity. Ectopic expression of the wild-type PTPRF, but not the phosphatase inactive mutant, showed reduced cell proliferation in HepG2, Huh7, and SK-Hep1 cells (Fig. 2A) and tumor colony formation in Huh7 and SK-Hep1 cells (Fig. 2B; Supporting Fig. S4). Conversely, PTPRF silencing in HepG2 cells showed enhanced colony formation in the soft-agar assay and enhanced xenograft tumor development in nude mice compared to cells transduced with shRNA targeting the luciferase control, which showed no tumor formation, as expected (Fig. 2C,D; Supporting Fig. S4). In addition to facilitating cell proliferation, PTPRF silencing caused cell hyperproliferation in Hep3B, HepG2, Huh7, and SK-Hep1 cells, as these cells continued to proliferate after 100% confluency was reached in culture (Fig. 2E,F; Supporting Figs. S4, S5). Similar results were also obtained using two different siRNAs targeting different parts of the PTPRF mRNA (Supporting Fig. S6). These findings suggest that PTPRF is a tumor suppressor restricting cell hyperproliferation and its phosphatase activity is important for cell growth suppression. We thus proposed that PTPRF down-regulation leads to evasion of growth suppression, resulting in hyperproliferation and potential tumor initiation.

Figure 2.

PTPRF suppressed cell proliferation and tumor initiation. Human hepatoma cells were transfected with wild-type PTPRF (wd), phosphatase-inactive mutant PTPRF (mt), or empty vector. (A) Cell proliferation was measured using both water soluble tetrazolium (WST1) and BrdU incorporation assays. An OD440 reading was obtained and was normalized to total cell numbers at 120 hours after seeding (when proliferation reached the stationary phase). Experiments were conducted three times in triplicate (six experiments) and the relative cell number was graphed as mean ± SD (six experiments/group). (B) Soft-agar colony formation assays were performed on SK-Hep1 cells transfected with PTPRF-wd or PTPRF-mt, or vector. Experiment was conducted twice in duplicate (four experiments). (C) Xenograft tumor assay was performed using HepG2 cells transduced with PTPRF-shRNA (shPTPRF) or luciferase-shRNA (control or shLuc), which were then injected subcutaneously into the left and right dorsal flanks of each nude mouse, respectively. Each group contained 8 mice and the experiment was conducted twice (n = 16/each group). The tumor volumes were measured and graphed as mean tumor volume ± SD (n = 16/group). (D) Soft-agar colony formation assays of shPTPRF- or shLuc-treated HepG2 cells. Two different clones of shPTPRF were used. Experiment was conducted three times in duplicate (six experiments). The number of colonies was counted and represented as mean ± SD (six experiments/group). (E) Cell proliferation assay for hepatoma cells transfected with siRNA targeting PTPRF (siPTPRF) or siRNA containing scrambled sequences (siNS). Experiment was conducted twice in triplicate (six experiments). (F) Human hepatoma cells were transduced with two different shPTPRFs or shLuc control for proliferation assays. The total cell numbers at stationary growth phase (120 hours after seeding) are shown. Experiment was conducted twice in triplicate (six experiments). Western analysis of PTPRF expression in our ectopic or silencing experiments are shown in Supporting Fig. S4. Consistent results of (E,F) were obtained by using BrdU incorporation assay to measure the newly synthesized DNA of replicating cells (Supporting Fig. S5).

Induction of PTPRF by Cell-Cell Contact During Proliferation

Since PTPRF prevented cell hyperproliferation, we hypothesized that PTPRF controls contact inhibition during cell proliferation. We thus examined whether its expression is regulated by cell-cell contact. PTPRF expression was induced in Hep3B and HepG2 cells with increasing time in culture (Fig. 3A). Moreover, the level of PTPRF increased in various hepatoma cell lines as the number of cells seeded into each well (cell density) increased (Fig. 3B), indicating that the induction of PTPRF expression was due to cell density or cell-cell contact rather than cell cycle progression. We then added ethyleneglycol-bis (2-aminoethylether)-N,N,N′,N′-tetraacetic acid (EGTA), a calcium chelator, to disrupt the adherens junctions of cell-cell contact (Fig. 3C).[17, 18] PTPRF induction was reduced as adherens junctions were disrupted (Fig. 3D). Consistent results were also obtained by E-cadherin silencing to disrupt the adherens junctions (Supporting Fig. S7). Taken together, these data suggested that PTPRF was induced by cell-cell contact during cell proliferation and that PTPRF plays a role in growth suppression to prevent cell hyperproliferation.

Figure 3.

Induction of PTPRF by cell-cell contact. (A) Western analysis for PTPRF (top panel) was performed at 24, 48, and 72 hours after seeding. The lower panel shows the quantification results of relative PTPRF expression per cell (normalized by GAPDH and then compared to that at 24 hours) and relative viable cell numbers. (B) PTPRF protein level in relation to the increasing cell density. Cells were seeded at the indicated number per well. Relative PTPRF amount was determined 24 hours after seeding by normalization with GAPDH and then compared to that seeded with 0.5 × 105/well. (C) Representative phase contrast images of Huh7 cells growing in culture with or without 5 mM EGTA treatment. (D) Western analysis for PTPRF expression in different cell density with or without EGTA treatment. GAPDH was used as loading control. Densitometry was performed and PTPRF/GAPDG relative expression was graphed.

PTPRF Silencing Caused Aberrant Activation of Extracellular Signal-Regulated Kinase/Mitogen-Activated Protein Kinase (ERK/MAPK) Pathways

To elucidate how PTPRF suppresses cell proliferation, we examined if PTPRF regulates the activation of the receptor tyrosine kinases (RTKs), particularly members of the epithelial growth factor receptor (EGFR) family that play crucial roles in tumorigenesis including HCC[10] and insulin receptor, a known PTPRF target.[19] PTPRF silencing by siRNA (siPTPRF) showed unaltered phosphorylation of EGFR, HER2, or ERBB3, but a modest increase in phospho-insulin receptor and its downstream target, IRS1 (Fig. 4A), suggesting that PTPRF does not directly regulate the phosphorylation of EGFR, HER2, or ERBB3.

Figure 4.

The effect of PTPRF silencing on intracellular signaling. (A) Western analysis of PTPRF silencing effects on the phosphorylation of EGFR, ERBB3, HER2, and insulin receptor (IR). Human hepatoma cells were transfected with siPTPRF or siNS. Lower panel shows the effect of siPTPRF on the activation of IR and insulin receptor substrate 1 (IRS1) in insulin-treated cells. (B) PTPRF silencing showed enhanced phosphorylation of ERK1 at Y187 and ERK2 at Y204 (p-ERK1/2), but not other intracellular signaling molecules. (C) PTPRF silencing in HepG2 and Huh7 cells (upper and lower panel, respectively) showed enhanced p-ERK1/2 at 30 minutes (30m) and 2 hours (2h) in response to EGF, insulin, and NGR treatment. siRNA-N and siRNA-F: siRNAs containing scrambled sequences and targeting PTPRF, respectively. The ratios of the indicated targets to their corresponding controls are indicated.

Next, we examined the effects of PTPRF knockdown on the intracellular oncogenic pathways. PTPRF knockdown showed enhanced phosphorylation of ERK1 (Y204) and ERK2 (Y187)(ERK1/2), but not p38/MAPK, JNK/MAPK, AKT, mTOR, PTEN, STAT3, or IκBα signaling (Fig. 4B), suggesting that PTPRF negatively regulates ERK1/2 phosphorylaton (p-ERK1/2). To further confirm that enhanced p-ERK1/2 by PTPRF silencing is downstream to RTK activation, HepG2 cells were serum-starved for 24 hours followed by treatment with EGF, insulin (INS), or neuregulin (NRG) for 30 minutes and 2 hours, which showed RTK-dependent activation of p-ERK1/2 in controls. The p-ERK1/2 level was further enhanced when PTPRF was silenced (Fig. 4C, upper panel). This experiment was repeated in Huh7 cells and p-ERK1/2 was examined after 30 minutes of EGF treatment. Despite that EGFR phosphorylation was not changed, p-ERK1/2 was further enhanced when PTPRF expression was silenced, similar to that obtained using HepG2 cells (Fig. 4C, lower panel). These data indicate that PTPRF is a negative regulator of ERK1/2 phosphorylation downstream to the activated RTKs.

PTPRF-Mediated Suppression of ERK Signaling by Way of Interaction With SRC and PP2A

To investigate how PTPRF suppresses ERK1/2 activation, we used coimmunoprecipitation to identify PTPRF interacting partners. SRC, PP2A, and PP2AC (the catalytic subunit C of PP2A) (Fig. 5A), but not ERK1/2 or FYN (data not shown), were coprecipitated with PTPRF in HepG2 and Huh7 cells. Moreover, PTPRF was reciprocally captured using antibodies against endogenous SRC and PP2A, but not IgG control, confirming that the interactions between PTPRF-SRC and PTPRF-PP2A were indeed direct (Fig. 5A). The direct interactions between endogenous PTPRF-SRC and PTPR-PP2A were further confirmed by in situ proximity-ligation assays, which showed colocalization in the cytosol (Fig. 5B).

Figure 5.

PTPRF suppresses ERK1/2 activation by way of direct interaction with SRC and with PP2A. (A) Reciprocal coimmunoprecipitation between PTPRF and SRC (left panel) and between PTPRF and PP2A (right panel). IP: antibodies used for coimmunoprecipitation; IB: antibodies used for western analysis. IgG was used as a nonspecific control. (B) In situ proximity ligation assay was used to detect direct interaction between PTPRF and SRC (left panel), and between PTPRF and PP2A (right panel) in Huh7 cells. The middle panel represents a negative control using antibodies against PTPRF and β-actin to show the reaction background. Bright red dots indicate positive interactions. (C) PTPRF silencing enhanced the phosphorylation of MEK, RAF, ERK1/2, SRC at Y416, and PP2AC at Y307. Three days after the silencing of PTPRF, cells were harvested for western analysis. (D,E) Enhanced p-ERK1/2 by PTPRF knockdown was abrogated when SRC (D) or PP2A (E) were also silenced. (F) Schematic diagram shows the proposed mechanism whereby PTPRF suppresses ERK1/2 activation by way of SRC and PP2A. PTPRF suppresses SRC (dephosphorylation at Y416) and activates PP2A (dephosphorylation at Y307). It is known that SRC (p-Y416) activates the RAF-MEK1/2-ERK signaling cascade, whereas PP2A (dephosphorylation at Y307) suppresses the activation of MEK1/2 and ERK1/2. Silencing of PTPRF enhances SRC activation (p-Y416) and PP2A inactivation (p-Y307), thereby synergically activating the RAF-MEK1/2-ERK1/2 signaling. Red arrows indicate target activation; blue lines indicate target inhibition. The ratios of the indicated targets to their corresponding controls are indicated.

Furthermore, we examined PTPRF-knockdown effects on SRC and PP2A phosphorylation. As shown in Fig. 5C, PTPRF silencing showed enhanced phosphorylation of SRC at Y416 (but not Y527) and PP2AC at Y307. Also, PTPRF silencing increased the phosphorylation of RAF1 at Y341 and MEK at S217/221, consistent with the known roles of SRC and PP2A in the regulation of RAF and both MEK1/2 and ERK1/2, respectively[20, 21] (Fig. 5F). Therefore, PTPRF suppresses ERK signaling through SRC-RAF1, PP2A-MEK, and PP2A-ERK1/2 interactions (Fig. 5F). We then cosilenced the expression of PTPRF with SRC or PP2A and found that the enhancement of p-ERK1/2 by PTPRF silencing was abrogated in both conditions (Fig. 5D,E), indicating that PTPRF negatively regulates ERK1/2 phosphorylation through SRC and PP2A. Taken together, it is proposed here that PTPRF facilitates SRC and PP2AC-PP2A dephosphorylation, leading to SRC activity suppression and PP2A activity activation, which together suppress ERK1/2 activation (Fig. 5F).

Aberrant Activation of ERK-MYC Signaling by Silencing of PTPRF

We further examined PTPRF-mediated effects on the ERK signaling pathways. PTPRF silencing showed enhanced phosphorylation of all the examined effectors downstream to ERK, including MNK, MYC, RSK, and eIF4E (Fig. 6A). The transcription factor MYC drives quiescent cells to enter and progress through the cell cycle and is phosphorylated at serine 62 for stabilization and accumulation by ERK1/2.22 Therefore, we examined PTPRF-mediated MYC phosphorylation/stabilization under growth stimuli. In response to EGF, INS, NRG, and hepatocyte growth factor (HGF) treatments, ERK1/2 and MYC phosphorylation were induced in the control (Fig. 6B). The phosphorylation of ERK1/2 and MYC were further enhanced in the presence of growth stimuli when PTPRF was knocked down in hepatoma cells (Fig. 6B). To exclude the possibility of off-target effects, we repeated the experiment using two different siRNAs targeting different parts of the PTPRF mRNA, which showed similar results and confirmed our observation (Fig. 6C). Notably, total MYC was increased when PTPRF was silenced, which was consistent with our speculation that PTPRF silencing led to ERK-mediated phosphorylation, stabilization and accumulation of MYC (Fig. 6B,C). Furthermore, we examined the essential role of ERK1/2 activation in the phosphorylation/stabilization of MYC by PTPRF silencing. We found that simultaneous knockdown of PTPRF and ERK1/2 abrogated the phosphorylation/stabilization of MYC that was seen in PTPRF silencing alone (Fig. 6D). These data indicated that the total MYC protein levels, phosphorylation, and stabilization were ERK-dependent and downstream of PTPRF. We thus propose here that during proliferation, cell-cell contact occurs, which induces PTPRF to form a negative-feedback loop of the “ERK-MYC proliferative cascade” (Fig. 6E, left panel). The down-regulation or loss of function of PTPRF would disrupt this negative feedback loop, resulting in the aberrant activation of ERK-MYC signaling, leading to cell hyperproliferation and potential tumor initiation (Fig. 6E, right panel).

Figure 6.

PTPRF regulates ERK-dependent proliferation signaling. For gene silencing experiments, cells were harvested for western analyses 3 days after transduction with shRNA lentivirus. For growth factor treatment, cells were harvested 30 minutes after treatment. (A) PTPRF silencing augmented the phosphorylation of ERK1/2 (T202/Y204) and its downstream effectors including MNK, MYC, RSK, and eIF4E. (B) PTPRF silencing enhanced the phosphorylaton of ERK1/2 and MYC (S62) in response to treatment with different growth factors. (C) Two siRNAs clones targeting PTPRF (#1 and #2) were used to exclude off-target effects (right panel). (D) ERK1/2 activation was required for MYC phosphorylation/stabilization induced by PTPRF silencing. Notably, increased MYC phosphorylation (S62) by PTPRF silencing led to an increase in total MYC proteins in all the experiments demonstrated in (B,C) in this figure. (E) Schematic shows the proposed PTPRF-mediated negative feedback loop of the ERK-MYC proliferation pathway (left). The loss of PTPRF function can cause aberrant activation of ERK1/2 and stabilization/accumulation of MYC, leading to cell hyperproliferation and tumor initiation (right). Red arrows indicate induction or enhancement of the downstream effector activity; black lines, suppression of the downstream effector activity.

Frequent Down-Regulation of PTPRF in Human HCC, Gastric, and Colorectal Cancers

To examine PTPRF expression in human HCC, we performed IHC on tissue arrays, which contained 89 pairs of HCC and adjacent nontumor tissues (Supporting Table S4). PTPRF was down-regulated in 37/89 (42%), whereas it was up-regulated in 21/89 (24%) HCC cases (Fig. 7A; Supporting Table S5). The frequent down-regulation of PTPRF in HCC was further verified in randomly selected 10 cases by quantitative RT-PCR (Fig. 7B, upper panel; Pearson correlation coefficient = 0.340, P = 0.337) and immunoblots (Fig. 7B, lower panel; Pearson correlation coefficient = 0.953, P < 0.001). In correlating PTPRF expression with clinical manifestations, we found that PTPRF down-regulation in HCC was strongly associated with advanced tumor stages (stage I versus stage III, P = 0.0132, Fig. 7C). Kaplan-Meier analysis revealed that, in addition to the TNM/AJCC and BCLC staging in prediction of long-term prognosis, PTPRF up-regulation was significantly associated with a lower recurrence rate (P = 0.0395, log-rank test; Fig. 7D; Supporting Table S5; Supporting Fig. S8). Multivariate analysis further revealed that PTPRF was an independent factor for tumor recurrence after primary resection (hazard ratio [HR] = 0.441; 95% confidence interval [CI]: 0.185-1.052; P = 0.0651; Supporting Tabled S6 and S7).

Figure 7.

PTPRF down-regulated in HCC, gastric cancer, and colorectal cancer. (A) IHC of PTPRF on tissue arrays containing paired HCC (T) and para-tumor liver (N) tissues in duplicate from 89 cases with HCC. Representative results from two cases with HCC are shown (×5 and ×100 magnification left and right panels, respectively). (B) Semiquantitative RT-PCR (upper panel) and western analysis (lower panel) derived from randomly selected 10 pairs of tumor (T) and nontumor liver tissues (N) were performed. Only the PCR and western analysis results derived from eight and six pairs of the samples, respectively, are shown. A normal liver tissue (nl) was included. (C) Correlation of IHC scores to TNM tumor stages. (D) Kaplan-Meier analysis of the probability of patients remaining free of recurrence (left panel) and the probability of disease-related survival (right panel). The P values were calculated using the log-rank test. M: months. (E) IHC of PTPRF on tissue arrays containing paired tumor (T) and nontumor tissue sections (N) from 40 patients with gastric cancer (left panel) and 40 patients with colorectal cancer (right panel). All of the nontumor parts of the colorectal and gastric mucosal epithelial cells show high expression of PTPRF. All of the 40 CRC and 27/40 GC show decreased PTPRF in tumor. Magnification: ×20 and ×100. (F) Higher magnification of the PTPRF IHC (A,E) show the subcellular distribution of PTPRF in the liver, gastric mucosa, and colonic mucosa and their corresponding tumors. Scale bars = 25 μm.

In addition to HCC, we also examined PTPRF expression in gastric and colorectal cancers. Interestingly, high expression of PTPRF was observed in the cytoplasm of normal mucosal epithelial cells resided in the gastrointestinal (GI) tract including gastric and colonic mucosa (Fig. 7E,F). PTPRF down-regulation was noted in 67% (27/40) and 100% (40/40) of the gastric (GC) and colorectal cancers (CRC), respectively (Fig. 7E), whereas PTPRF up-regulation was not seen in GC or CRC. Of note, PTPRF mainly distributed in the cytoplasm of the hepatocytes and mucosal epithelial cells of the GI tract, and was frequently down-regulated in tumor cells (Fig. 7F). These clinical observations implicated PTPRF as an important tumor suppressor that is frequently down-regulated in human HCC, GC, and CRC.


Growth suppression elicited from proliferating cells is a crucial process that prevents cells from hyperproliferation. Loss of this mechanism can promote tumor development.[23] In this study, we conducted human kinome and phosphatome screening to identify TSGs in the nontumorigenic HepG2 cells and identified 19 TSGs. Of them, PTPRF was selected for further characterization. PTPRF is a receptor type of protein tyrosine phosphatase and reported in the regulation of neural, thymocyte, and urinary tract development.[24] We have shown here that PTPRF silencing led to cellular hyperproliferation and enhanced tumor formation, whereas ectopic expression of PTPRF inhibited cell proliferation and tumor development. Our data also demonstrated that PTPRF was induced by cell-cell contact, which was consistent with a previous report.[25] These data all support that PTPRF is a tumor suppressor and plays a role in growth suppression in liver cells. We confirmed that this growth suppression property of PTPRF was dependent on its phosphatase activity as cell growth suppression was abrogated when the phosphatase motif of PTPRF was mutated.

PTPRF negatively regulates insulin receptor signaling, which is related to insulin resistance in experimental animals.[26] PTPRF also suppresses c-MET activity in response to growth stimuli.[27] In this study, we have shown that PTPRF did not affect EGFR, HER2, and ERBB3 phosphorylation, but suppressed downstream ERK-mediated signaling. We found that PTPRF was induced by cell-cell contact, which then quenched the activated ERK-MYC signaling, thereby suppressing cell proliferation. Furthermore, we have shown that PTPRF suppressed ERK1/2-dependent signaling by directly interacting with SRC and PP2A, which led to the suppression of SRC and activation of PP2A. This speculation was further evidenced by the findings that as co-silencing of PTPRF with SRC or PP2A abrogated the enhanced ERK1/2 activation by PTPRF silencing alone. Previous studies have shown that SRC indirectly activates ERK1/2 through sequential phosphorylation (activation) of RAF and MEK1/2 in response to growth stimuli,[20] while PP2A dephosphorylates/inactivates both MEK1/2 and ERK1/2 signaling.[21] Simultaneous suppression of SRC and activation of PP2A by PTPRF synergically suppress ERK activation (Fig. 5F). In fact, SRC and PP2A have been reported to be oncogenic and tumor suppressive in many human cancers.[28, 29] Taken together, we conclude that the PTPRF-SRC/PP2A interactions form a negative feedback loop of the ERK-dependent cell proliferative signaling (Fig. 6D).

In addition, the negative regulation of ERK-dependent signaling by PTPRF is different from other phosphatases such as MAP kinase phosphatases, because the PTPRF-mediated negative regulation of ERK signaling is driven by cell-cell contact during cell proliferation. Interestingly, we found a discrepancy between the PTPRF protein and mRNA levels in some HCC samples, suggesting that a novel regulatory mechanism regulates the cell-cell contact induced PTPRF. Further studies are warranted. This PTPRF-mediated growth suppression is reminiscent of the Hippo-Yap pathway, which has been shown to play a key role in controlling organ size, primarily by inhibiting cell proliferation and promoting apoptosis.[30-32] Accumulating evidence suggests the implication of the Hippo-Yap pathway in tumorigenesis, cell stemness, tissue regeneration, and tumor metastasis.[33-35] In this study, we showed that YAP1 activity was unaffected by PTPRF, and also PTPRF expression and PTPRF-ERK signaling were not modulated by the Hippo-Yap pathway (Supporting Fig. S9). Thus, we conclude that the PTPRF-mediated growth suppression pathway is functionally independent of the Hippo-Yap pathway despite both pathways being triggered by cell-cell contact. Interestingly, PP2A are involved in both PTPRF-ERK and Hippo-Yap signaling pathways.[36, 37] It is most likely that both pathways are functioning in parallel. However, unlike the universal expression of YAP1 in most epithelial cells, neurons, mesenchymal cells, and myocytes, PTPRF is specifically expressed in the proliferating mucosal epithelial cells such as gastrointestinal, respiratory, mammary, testicular, prostate, and renal glandular/ductular epithelia (Human Protein Atlas,[38] Given that Ptprf knockout mice showed hyperneurogenesis and hyperplasia of the mouse brain during brain development[39] and intima hyperplasia of blood vessels in response to vascular injury,[40] we believed that PTPRF may play a critical role in maintaining proper proliferation, particularly during organogenesis or tissue repair.

Recently, somatic mutations of PTPRF were detected in human colorectal cancer and HCC.[41, 42] Here we observed that PTPRF down-regulation was detected in 42%, 67%, and 100% of HCC, GC, and CRC, respectively, and was strongly associated with higher histological grade of HCC and worse clinical outcomes in HCC patients. We propose here that PTPRF might be a prognostic marker for human cancers such as HCC, GC, and CRC (Supporting Fig. S8). We also found that the PTPRF gene was deleted in 20.5% of HCC (Supporting Fig. S10). Indeed, the PTPRF gene is located at chromosome 1p34. Deletion of chromosome 1p is frequently found in many human cancers, including HCC. Therefore, PTPRF is most likely a tumor suppressor gene located at chromosome 1p. We also speculate the implication of PTPRF inactivation in tumorigenesis of tissues with chronic inflammation, such as HCC patients with chronic viral hepatitis, GC patients with chronic gastritis due to Helicobacter pylori infection, and CRC patients with chronic colitis. Moreover, tissues with growth signaling dysregulation, for instance, aberrant activation of the HER2/ERBB3 signaling in HCC[10] and HER2 in breast cancer and melanoma,[43] may also be particularly susceptible to tumorigenesis initiated by the down-regulation or loss-of-function of PTPRF. Further studies on the role of PTPRF in carcinogenesis of tissues with chronic inflammatory are warranted. In addition, MYC is frequently amplified in many human cancers, including HCC. Therefore, it is most likely that the PTPRF-SRC/PP2A-ERK-MYC pathway is vital in suppressing MYC phosphorylation and activation during cell growth in order to prevent cell hyperproliferation and tumor initiation.

In summary, we have identified PTPRF as a novel tumor suppressor of human HCC by way of a loss-of-function screening of the human kinome and phosphatome. PTPRF is induced during cell proliferation by cell-cell contact, which then quenches the activated ERK-MYC cascade by simultaneously suppressing SRC and activating PP2A. This cell contact-induced PTPRF-SRC/PP2A-ERK pathway constitutes a negative feedback loop of the ERK-mediated proliferation signaling to prevent cell hyperproliferation and tumor initiation. Our discovery opens a new avenue for the future development of anticancer therapies.


The authors thank Professor Ruey-Hua Chen at the Academia Sinica for providing the wild-type and phosphatase-inactive mutant PTPRF cDNA clones, Professor Chee-Jen Chang at the Chang Gung University for biostatistical analysis, the Taiwan Liver Cancer Network for HCC tissue array, and the National RNAi Core of Taiwan for the lentivirus-based shRNA clones used in this study.