POH1 deubiquitylates and stabilizes E2F1 to promote tumour formation

Hyperactivation of the transcriptional factor E2F1 occurs frequently in human cancers and contributes to malignant progression. E2F1 activity is regulated by proteolysis mediated by the ubiquitin–proteasome system. However, the deubiquitylase that controls E2F1 ubiquitylation and stability remains undefined. Here we demonstrate that the deubiquitylase POH1 stabilizes E2F1 protein through binding to and deubiquitylating E2F1. Conditional knockout of Poh1 alleles results in reduced E2F1 expression in primary mouse liver cells. The POH1-mediated regulation of E2F1 expression strengthens E2F1-downstream prosurvival signals, including upregulation of Survivin and FOXM1 protein levels, and efficiently facilitates tumour growth of liver cancer cells in nude mice. Importantly, human hepatocellular carcinomas (HCCs) recapitulate POH1 regulation of E2F1 expression, as nuclear abundance of POH1 is increased in HCCs and correlates with E2F1 overexpression and tumour growth. Thus, our study suggests that the hyperactivated POH1–E2F1 regulation may contribute to the development of liver cancer.

T he E2F transcription factor 1 (E2F1) is a master transcription factor that participates in numerous important biological processes 1 . Besides the clinical evidence that aberrant upregulation of E2F1 frequently occurs in various types of human cancer and correlates with malignant progression and poor survival prognosis [2][3][4] , the E2F1-driven onocgenic activity has been reinforced in different models based on E2F1 transgenic or knockout mice [5][6][7][8] . Although the observations showing that E2F1 is involved in cellular senescence and apoptotic response may suggest its dual role in tumorigenesis 9 , several events contributing to tumorigenesis may counterbalance the tumour suppressive effects of E2F1. For instance, cells with the deficiencies in p53 or p14(ARF) can escape from E2F1-mediated apoptosis or prosenescent effects 6,10 , and the protumorigenic signals generated by epidermal growth factor receptor and phosphatidylinositol 3-kinase (PI3K)/Akt are capable of inhibiting E2F1 apoptotic function 11,12 . In addition, E2F1 itself has been shown to promote tumour cell survival and EMT as well as angiogenesis in certain circumstances [13][14][15][16] . Therefore, the oncogenic activity of E2F1 is determined by the strength of prosurvival factors either downstream of this transcriptional factor or provided by other signals. E2F1 turnover is controlled by the ubiquitin-proteasome system [17][18][19][20][21] . Several factors responsible for the ubiquitination of E2F1 have been identified, including SKP2, APC/C and the ROC-cullin complex [17][18][19]21 . MDM2, which negatively regulates p53, directly interacts with and increases the half-life of E2F1 protein by displacing SCF 20 . Deubiquitination is considered a key process in the maintenance of proper cellular function and homeostasis. Numerous studies have established that the dysfunction of deubiquitinating enzymes is critical for tumour development, progression or chemosensitivity [22][23][24][25][26] . However, the contribution of deubiquitinating enzymes to the stabilization of E2F1 and its biological significance in carcinogenesis has not been determined. POH1/rpn11/PSMD14 is a deubiquitinating enzyme within the 19S particle lid that regulates proteasomal activities 27,28 . POH1 plays a 'proof-reading' role in controlling the fate of incoming substrates [27][28][29] . In mammalian cells, POH1 functions in various biological processes, including double-strand DNA break responses 30 , embryonic stem cell differentiation 31 , aggresome disassembly and clearance 32 , cellular viability 33,34 , multidrug resistance 35 and protein stability [36][37][38] . However, whether POH1 deregulation occurs in and contributes to the development of liver cancer has not been determined.
In this study, we identify POH1 as the deubiquitinating enzyme that stabilizes E2F1 and demonstrate that aberrant hyperactivity of POH1-E2F1 regulation promotes liver tumour formation. Our study therefore describes a previously unknown mechanism by which E2F1 expression is regulated as well as its implication in tumorigenesis.

Results
Identification of POH1 as a positive regulator of E2F1. To identify deubiquitinating enzymes (DUBs) with the capacity of regulating E2F1 expression, we initially screened 37 DUBs, expression of which are relatively high in human liver tissues based on in sillico EST profile analysis. For each DUB tested we employed a pool of three non-overlapping siRNA oligos for transfection experiments. The results of the relative quantification of E2F1 protein levels, as assembled in rank order, showed that knockdown of POH1 markedly repressed E2F1 expression (Fig. 1a). Representative images of the immunoblots are shown in Supplementary Fig. 1a. The mRNA levels of POH1 were increased in a number of hepatocellular carcinomas (HCCs) compared with that in the adjacent non-tumour specimens ( Supplementary Fig. 1b), suggesting a potential clinical relevance for POH1 in the development of HCC.
We further validated the downregulation of E2F1 expression by POH1 knowdown in different liver cancer cell lines ( Fig. 1b; Supplementary Fig. 1c). The decrease in E2F1 protein was not associated with changes in the E2F1 mRNA levels in SMMC-7721 and PLC/PRF/5 cells ( Supplementary Fig. 1d), indicating that a post-transcriptional mechanism is involved in POH1-mediated regulation of E2F1. Furthermore, in p53 À / À mouse liver progenitor cells transformed with myristoylated AKT (LPC-Akt) or c-Myc (LPC-c-Myc), and the control cells (LPC-Mig) 39 , POH1 depletion markedly inhibited E2F1 expression ( Fig. 1c; Supplementary Fig. 1e). The effects of POH1 on E2F1 expression might also apply to other types of cancer cells such as the colon cancer LoVo cells ( Supplementary Fig. 1f). Remarkably, the expression levels of E2F2 and E2F3, two E2F1 relatives 9 , were not significantly reduced ( Supplementary Fig. 1g), indicating that POH1-mediated regulation of E2F1 may be a selective event.
We examined the effect of gain of POH1 function on E2F1 activity in tumour cells. PLC/PRF/5, SMMC-7721 and SK-Hep1 cells stably expressing exogenous Flag-tagged POH1 had higher levels of E2F1 than did the control cells ( Fig. 1d; Supplementary  Fig. 1h). POH1 overexpression positively regulated E2F1 activity, as monitored by an E2F1-binding element-luciferase reporter, whereas POH1 deletion yielded an opposite effect (Fig. 1e). More importantly, we did not observe an appreciable change in proteasomal activity within the cells overexpressing POH1, as revealed by the levels of total ubiquitinated proteins Figure 1 | POH1 is a positive regulator of E2F1 stability. (a) For each DUB protein indicated, a pool of three independent siRNAs was transfected into PLC/PRF/5 cells. The protein levels of E2F1 were analysed by immunoblotting and normalized to that of the control GAPDH. The relative expression compared with that of the siRNA control is presented in rank order. (b,c) Immunoblot analysis of POH1 and E2F1 in human liver tumour cells (b) and mouse LPCs (c) transfected with the indicated siRNAs. Untransfected cells were shown as parental controls. (d) Immunoblot analysis of POH1 and E2F1 in human liver tumour cells with or without POH1-Flag expression. (e) Luciferase assay in HEK293T cells co-transfected with the reporter plasmids combined with the indicated siRNAs (left panel) or with POH1 and the vector control, respectively (right panel). Data are mean ± s.d. (by t-test analysis, P values are shown in the graph, n ¼ 3). (f) SMMC-7721 cells transfected with control siRNA or POH1 siRNA were cultured for 72 h. Cell lysates were immunoprecipitated with anti-E2F1 antibody, and the immunocomplexes were immunoblotted with antibodies against UB and E2F1. (g) Human liver tumour cells transfected with POH1-Flag were immunoprecipitated with Flag-M2 agarose beads. The eluates were immunoblotted for the detection of E2F1.
(h) HEK293T cells were co-transfected with E2F1-His and POH1-Flag followed by immunoprecipitation using anti-His antibody or IgG. The immunocomplexes were analysed by immunoblotting. (i) Endogenous E2F1 proteins were immunoprecipitated with anti-E2F1 antibody or IgG, and then analysed by immunoblotting. (j) PLC/PRF/5 cells with or without stable expression of POH1-Flag were treated with CHX (100 mg ml À 1 ) for the indicated time points. The cell lysates were examined by immunoblotting (left panel). A plot of normalized amount of E2F1 protein is shown (right panel). (k) SMMC-7721 cells were transfected with either control siRNA or POH1 siRNA for 48 h, followed by CHX(100 mg ml À 1 ) treatment for the indicated times. The cell extracts were analysed by immunoblotting (left panel). A plot of normalized amount of E2F1 protein is shown (right panel). Data in panels (j,k) are mean ± s.d. (by ANOVA analysis, P values are shown in the graphs, n ¼ 3).
( Supplementary Fig. 1i). These results indicate that POH1mediated upregulation of E2F1 is not paralleled by a change in the global activity of the proteasome. POH1 stabilizes E2F1. To investigate whether POH1 regulates E2F1 ubiquitination, we first examined the levels of polyubiquitin-modified E2F1 in tumour cell lines with or without POH1 knockdown. POH1 knockdown significantly enhanced the levels of E2F1 ubiquitination (Fig. 1f). This finding led us to hypothesize that POH1 may interact with and deubiquitinate E2F1. We therefore performed immunoprecipitation analyses, and the results revealed that E2F1 was co-immunoprecipitated with POH1 protein in cells transfected with Flag-tagged POH1 (Fig. 1g). The interaction between POH1 and E2F1 was verified by co-immunoprecipitation assays in HEK293T cells (Fig. 1h). Furthermore, the interaction between endogenous POH1 and E2F1 was demonstrated by co-immunoprecipitation P a r e n t a l C o n t r o l s iR    (Fig. 1i). We then generated a set of truncated mutants of POH1 and E2F1 to delineate the regions within these proteins essential for their interaction. Depletion of the COON-terminal domain (aa181-310) of POH1 significantly impaired the association between E2F1 and POH1, whereas the portion of E2F1 (aa201-437) displayed strong capability to interact with POH1 (supplementary Fig. 2a,b). The interaction between the N-terminal fragment (aa1-200) of E2F1 and POH1 was relatively weak, but certainly reproducible. To test whether POH1 stabilizes E2F1 protein, we examined the effect of both POH1 depletion and overexpression on the stability of endogenous E2F1 protein in the presence of the inhibitor of protein translation, cycloheximide (CHX). The results indicate that forced POH1 expression significantly potentiates E2F1 protein stability (Fig. 1j).
In contrast, E2F1 protein was degraded more rapidly in POH1-knockdown cells compared with control cells (Fig. 1k). These data collectively suggest that POH1 interacts with and stabilizes E2F1 in tumour cells.
E2F1 downregulation in mouse liver cells deficient in Poh1. To evaluate whether POH1-mediated regulation of E2F1 occurs in primary mouse liver cells deficient in POH1, we generated conditional Poh1 f/f mice by targeting Poh1 exon 6 ( Fig. 2a), the deletion of which led to loss of POH1 expression in the presence of Cre recombinase. The inducible Poh1 knockout mice (Mx-Cre þ , Poh1 f/f ), generated by crossing Poh1 f/f mice with Mx-Cre mice, were treated with three rounds of polyinosinic:polycytidylic acid (polyI:C) injection. Disruption of the floxed Poh1 allele was then achieved in liver tissues and accompanied by a substantial reduction of E2F1 expression (Fig. 2b,c). Immunochemistry staining verified the reduction in both POH1 and E2F1 expression in mouse hepatocytes (Fig. 2d). Consistent with results obtained in cultured cells, the levels of ubiquitin-modified E2F1 were substantially increased in the liver tissues of mice with poh1 ablation (Fig. 2e). Accordingly, depletion of Poh1 in poh1 f/f MEFs could cause an acceleration of E2F1 protein turnover that apparently was not due to the change in E2F1 mRNA levels ( Supplementary Fig. 3a-c). ubiquitin-HA), separately, were purified and subjected to immunoblotting analysis using an anti-HA antibody. POH1 overexpression substantially reduced K11-and K63-linked ubiquitination on E2F1 (Fig. 3a). Similar results were obtained in the experiments using the antibodies recognizing K48 or K63-linked polyubiquitin chains ( Supplementary Fig. 4a,b). The cleavage of K63-linked polyubiquitin chains on E2F1 by POH1 is consistent with the function of POH1 described as a K63-specific deubiquitinase in previous structural and biochemical studies 40,41 . Of note, K11-linked ubiquitination is related to E2F1 degradation by the proteasomal 19 . Overexpression of the K63-resistant form of ubiquitin (K63R) attenuated POH1 knockdown-induced E2F1 downregulation, indicating that K63-linked polyubiquitination is important for POH1-mediated E2F1 turnover (Fig. 3b). Although POH1 overexpression caused the deubiqutination of E2F1, the approach did not have an appreciable impact on levels of total polyubiquitin-modified proteins ( Supplementary Fig. 5a), suggesting that POH1-mediated deubiquitination of E2F1 can take place in cells in absence of a noticeable alteration in global proteasome activity. In addition, in MG132-treated cells, wherein the global proteasomal activity was inhibited, ectopic POH1 expression was able to deubiquitinate E2F1 ( Supplementary  Fig. 5b), suggesting that this effect is independent of the overall proteasome function.
Previous structural and functional studies indicate that the JAMM domain of POH1 and the C120 and H113 sites are important for the deubiquitinating enzymic activity of POH1 ( Fig. 3c) 36,37,42,43 . Accordingly, ectopic expression of the POH1 mutants C120S or H113Q failed to deubiquitinate of E2F1 proteins (Fig. 3d). Of note, the POH1 mutant with deletion of JAMM motif (deletion of 113-126aa) was capable of binding to E2F1 ( Supplementary Fig. 2a). Our in vitro analysis using 293T cell-derived recombinant POH1 further confirmed the capability of POH1 for cleaving K63-linked polyubiquitin chains from E2F1 and the importance of the functional domain and sites (Fig. 3e). Unexpectedly, E.coli-expressed POH1 did not display a deubiquitining activity on E2F1; that was consistent with the previous studies 28,36 , suggesting that the defect might be due to either inappropriate folding of prokaryotic recombinant POH1 or absence of POH1-partner proteins in reaction. Furthermore, we generated different siRNA-resistant (siR-R) POH1 constructs for rescue assays, and the results showed that restoration of POH1 expression with the C120S-, H113Q-and DJAMM-POH1 mutants did not attenuate the inhibition of E2F1 caused by the deletion of endogenous POH1 (Fig. 3f). Altogether, we have demonstrated that the deubiquitinating activity that POH1 possesses is fundamental for POH1 regulation of E2F1.
It should be noted that POH1 may function separately in regulating E2F1 abundance and proteasomal activities. C120S-POH1 was enzymatically dead and not effective in deubiquitinating E2F1 (Fig. 3d,e). As expected, ectopic expression of the siRNA-resistant C120S-POH1 did not restore E2F1 expression in cells where the endogenous POH1 was downregulated by siRNA ( Fig. 3g left). However, C120S-POH1 expression is sufficient in maintaining the global proteasomal activities, as revealed by the abundance of ubiquitin-modified proteins (Fig. 3g right). These results are consistent with the findings that the C120 mutation has a deleterious effect on POH1 deubiquitinating activities but does not abolish its proteasome activator function 31,44 . Altogether, these results indicate that POH1-mediated deubiquitination of E2F1 may be independent of the canonical function of POH1 in the proteasome.
POH1 has protumorigenic activities. We next investigated the effects of POH1 inhibition on the malignant growth potential of liver cancer cells. POH1 inhibition by siRNAs markedly suppressed the colony-forming ability of three human liver cancer cell lines and the oncogenes-transformed mouse LPCs and the control cells (Fig. 4a,b; Supplementary Fig. 6a). Consistently, MTT assays also revealed reduced viability of human liver cancer cells with POH1 knockdown (Supplementary Fig. 6b). In addition, POH1 knockdown resulted in increased percentage of apoptotic cells ( Supplementary Fig. 6c-e) and cell cycle arrest ( Supplementary Fig. 6f). Conversely, ectopic POH1 expression substantially potentiated tumour cell survival in detached culture conditions (Fig. 4c). Ectopic expression of C120S-POH1, the enzyme-dead mutant with the capacity of maintaining overall proteasomal activity, did not rescue the suppression of cell proliferation caused by endogenous POH1 deletion ( Supplementary  Fig. 6g,h).
To address the importance of POH1 in regulating the growth of liver cancer cells, we generated the SK-Hep1 cells with a Tet-On inducible shRNA cassette against POH1 (sh-POH1-Tet-on; Supplementary Fig. 6i).Compared with the control groups, mice with induced sh-POH1 expression in tumour cells displayed a significant inhibition of tumour growth (Po0.001, by analysis of variance (ANOVA) analysis) ( Fig. 4d-f). Immunohistochemical (IHC) staining revealed reduced E2F1 and Ki67 expression in the xenograft tumours generated from POH1knockdown cells ( Supplementary Fig. 7). On the contrary, POH1 but not the JAMM-deleted POH1 overexpression in the PLC/ PRF/5 liver cancer cells significantly increased the tumours' volumes ( Fig. 4g,h). Furthermore, E2F1 knockdown with stable shRNA expression was able to counteract the protumorigenic activities of POH1 (Fig. 4g,h). These results underscore an essential role of E2F1 in POH1-mediated protumorigenic activities.

POH1 activates Survivin and FOXM1 expression through E2F1.
To delineate the roles of POH1 in regulating tumour cell survival and proliferation, we examined genome-wide transcription profiles of HCC SMMC-7721 cells with or without POH1 knockdown by mRNA microarray. To identify the downstream effectors of the POH1-E2F1 pathway, we focused on the expression alterations of a set of known E2F1 target genes between the control and the POH1-depleted cells (Supplementary Fig. 8 and Supplementary Table 1). The results showed that the majority of E2F1 targets examined were downregulated in the POH1-depleted cells, including Survivin and FOXM1, both of which are critical for tumorigenesis [45][46][47][48] . The downregulation of Survivin and FOXM1 in the cells with POH1 deletion was validated by real-time RT-PCR (Fig. 5a). Accordingly, POH1 deletion substantially inhibited the expression of Survivin and FOXM1 proteins in human liver cancer cell lines ( Fig. 5b; Supplementary  Fig. 9a). Conversely, the forced expression of exogenous POH1 significantly elevated Survivin and FOXM1 levels ( Fig. 5c; Supplementary Fig. 9b), whereas the overrexpression of an enzymatic dead mutant of POH1 failed to up-regulate Survivin and FOXM1 ( Supplementary Fig. 10). Consistent with these findings, E2F1 knockdown robustly inhibited Survivin and FOXM1 expression and increased the frequency of apoptosis in SMMC-7721 cells ( Supplementary Fig. 11a,b). Furthermore, E2F1 inhibition by siRNAs almost entirely abolished POH1-mediated upregulation of Survivin and FOXM1 in liver cancer cells (Fig. 5d). The restoration of E2F1 or Survivin expression in liver cancer cells substantially counteracted POH1 deletioninduced cell growth inhibition and apoptotic response ( Fig. 5e-h; Supplementary Fig. 12). Altogether, these results have demonstrated that E2F1 is responsible for POH1-mediated activation of the protumorigenic factors.   Nuclear POH1 is essential for liver tumour cell growth. Given that POH1 may function in the nucleus 30,37,49 , we hypothesized that the nuclear localization of POH1 is critical for regulating E2F1 expression. We first isolated the nuclear and cytoplasmic fractions of SK-Hep1 and SMMC-7721 cells and measured the expression pattern of POH1. Indeed, both endogenous and exogenous POH1 proteins were detected in the nuclear and cytoplasmic compartments with antibodies against POH1 and Flag-tag, respectively (Fig. 6a,b). We further substantiated the observation by fluorescent immunostaining (Fig. 6c,d). In cells ARTICLE with POH1 knockdown, E2F1 expression in the nuclear fractions was substantially decreased (Fig. 6e). E2F1 protein was mainly localized in tumour cell nucleus and the co-immunoprecipitation between POH1 and E2F1 was primarily observed in the purified nuclear compartment of the cells (Supplementary Fig. 13). To investigate the biological function of nuclear POH1, we incorporated a nuclear export signal (NES) sequence into the NH 2 terminus of POH1-Flag. The POH1(NES) is prominently located in the cytoplasmic compartment (Fig. 6f). We therefore comparatively analysed the effectiveness of the siRNA-resistant forms of POH1 and POH1(NES) on restoring the levels of E2F1, Survivin and FOXM1 in cells with endogenous POH1 knockdown. In contrast with POH1 that entirely reinstated the expression of these proteins, POH1(NES) did not compensate for the loss of function of the endogenous POH1 (Fig. 6g). More importantly, cytoplasm-restricted expression of POH1 did not alleviate the inhibition of cell growth and the induction of cell apoptosis caused by endogenous POH1 deletion ( Fig. 6h and Supplementary Fig. 14a,b). In addition, the POH1 mutant protein (POH1(NL)-Flag) with a predominant distribution in cell nucleus efficiently upregulated the expression of E2F1 protein (Fig. 6i). Collectively, these data demonstrate that nuclear-localized POH1 is critical for E2F1 stabilization and the efficient growth of liver tumour cells.

POH1 is upregulated in HCCs and correlates with E2F1 levels.
To further delineate whether POH1-mediated regulation of E2F1 expression is clinically relevant to human HCC development, we characterized the expression patterns of POH1 and E2F1 and assessed their association in human HCC samples. Western blot analyses showed elevated POH1 and E2F1 expression in 10 HCCs in comparison with their adjacent non-tumoral tissues (Fig. 7a). To examine POH1 and E2F1 expression in a relatively large pool of HCC samples, we employed tissue microarrays containing 154 matched HCCs and the adjacent non-malignant liver tissues for IHC staining. Cytoplasmic immunoreactivity was observed in both hepatocellular carcinoma cells and normal hepatocytes. In total, 41.6% (64/154) of the non-tumour liver tissues exhibited clear nuclear staining, whereas in the HCCs, 72.1% (111/154) of the samples displayed nuclear staining (Fig. 7b, Supplementary  Fig. 15a). IHC staining for POH1 and E2F1 was scored according to the intensity (ranging from 1 to 4) and the percentage of positive cells. Given that nuclear POH1 mainly contributed to E2F1 regulation, the staining location was also added to the scores. The overall score was then determined by multiplying the intensity score, the percentage score, and the location score ( Supplementary Fig. 15b). The increases in the expression of POH1 and E2F1 in HCCs were statistically significant (Fig. 7c). Moreover, we detected a significant correlation between POH1 staining scores and tumour stages of the samples examined with Kruskal-Wallis test (P ¼ 0.015). In addition, the multiple hypothesis test (Holm-Sidak's multiple comparisons test) was performed and these results showed that the protein levels of POH1 in stage T3/T4 tumours were significantly higher than those in tumours at earlier stages (T1 versus T2, P ¼ 0.845; T1 versus T3-T4, P ¼ 0.047; T2 versus T3-T4, P ¼ 0.005) (Fig. 7d). However, there was no correlation between POH1 levels with tumour grades (Supplementary Fig. 16). Importantly, a positive correlation was found between the staining scores of POH1 and E2F1 (Po0.001, by Spearman correlation test; Fig. 7e,f). Thus, these results reveal the clinical relevance of POH1-mediated regulation of E2F1 in HCC development.
To further corroborate the findings aforementioned, we analysed a publicly available transcriptome data set (GSE14520), which was collected from 241 non-tumoral liver tissues and 247 HCC samples. The transcript levels of POH1 and its downstream genes Survivin and FOXM1 in the tumour tissues were aberrantly upregulated in comparison with those in the nontumour samples (Fig. 8a). Moreover, the POH1 transcript abundance was positively correlated with the expression of Survivin and FOXM1 in the HCC samples (Survivin: Pearson correlation R ¼ 0.409, Po0.001; FOXM1: Pearson correlation R ¼ 0.346, Po0.001, Fig. 8b). To assess the correlation between POH1 expression and E2F1 target gene signature, we performed the gene set enrichment analysis (GSEA) of the dataset GSE14520 using a signature of E2F1-target genes from the Molecular Signature Database (http://www.broadinstitute.org/gsea/msigdb/ cards/E2F1_UP.V1_UP.) 50,51 . The positive enrichment score obtained from the analyses with the nominal P value ¼ 0.0175, false discovery rate Q value ¼ 0.0175 and FWER P value ¼ 0.009 (Fig. 8c), is indicative of a high enrichment in the expression of E2F1 target gene signature in the tumour subset with high POH1 expression.

Discussion
Emerging evidence indicates that POH1 plays roles in several biological processes, including DNA repair, cell differentiation and transcriptional control. Our present study provides evidence that POH1 deubiquitinates and stabilizes the master transcription factor E2F1 and functions as a tumour-promoting protein in HCCs. We demonstrated that POH1 efficiently deubiquitinates E2F1 by removing the K-63 polyubiquitin chains, and this observation is consistent with previous studies revealing that the JAMM domain within POH1 is responsible for removing K63linked ubiquitin chains 40,41,43 . The deubiquitination of E2F1-K63 polyubiquitin chains appears to be important for POH1-mediated stabilization of E2F1 because Ub-K63R overexpression showed a dominant-negative effect that rescued E2F1 expression in cells with POH1 inhibition (Fig. 3b). Although protein modification via K63-linked polyubiquitin chains is not generally considered a degradation signal, previous studies have suggested that K63-linked polyubiquitin chains serve as proteasome recognition tags by which substrates are efficiently processed in proteasome 52,53 . Of note, the proteasome indeed degrades  54 . The modification of K11-linked polyubiquitination on E2F1 is substantially downregulated by POH1 (Fig. 3a). The K11-linked polyubiquitin chains are involved in the proteasomal degradation of E2F1 protein 19 , raising the possibility that loss of the K11-linked polyubiquitin modification may contribute to POH1-mediated stabilization of E2F1. In addition, the possibility cannot be excluded that mutual regulation between K11-and K63modification on E2F1 may occur; if so, then the deubiquitination of the K63-modification by POH1 may cause a reduction in K11-linked polyubiquitination of E2F1 proteins. Furthermore, the formation of mixed K11/K63 chains on E2F1 may also need to be taken into consideration. Of note, although E2F1 deubiquitination by POH1 has been revealed to be important for E2F1 stabilization, the molecular details of the regulation remain to be defined. An approach using point mutants of these proteins that specifically disrupt the E2F1/POH1 interaction may deepen our understanding of POH regulation of E2F1. Designing such mutants is currently difficult owing to lack of structural insights into the interaction, but warrants our future studies.
The 19S regulatory particle localized in the nucleus may display non-proteolytic functions through directly binding the transcriptional regulatory elements in various conditions 55 . In primary liver tissues and cultured cell lines, we observed a considerable amount of POH1 in the nucleus. The existence of nuclear POH1 is important for regulating E2F1 because cytoplasmic-localized POH1 cannot stabilize E2F1 in cells with deletion of endogenous POH1. We propose that the deubiquitination of E2F1 by POH1 primarily occurs in the nucleus; as a consequence, the prosurvival genes, including Survivin and FOXM1, are transcriptionally activated. A previous study indicated that POH1 overexpression in HEK293 cells downregulates c-Jun ubiquitination, resulting in c-Jun accumulation and upregulation of AP1-mediated gene expression 36 . These observations suggest that transcriptional regulation by the 19S proteasome is not simply due to direct binding to and sequential activation of the targeted locus.Instead, deubiquitinating enzymes of the 19S particle may provide an important alternative by deubiquitinating the master transcriptional factors. Therefore, these findings shed light on the mechanism by which the 19S particle or its subunits play transcriptional regulatory roles in the nucleus.
Most importantly, our study has revealed a coordinated upregulation of POH1 and E2F1 expression in clinical HCC samples, implying a pathological significance for POH1 regulation of E2F1 in HCC development. E2F1 is thought to play dual roles in tumorigenesis 9 , and E2F1 downstream targets consist of both oncogenes and tumour suppressors. The paradoxical function of E2F1 is cell context dependent and determined by the output and nature of downstream targets, which are either beneficial or deleterious to cell proliferation and survival 12,56 . Some oncogenic factors activated by E2F1, including Akt, mTORC1 and EGR1, can override the tumour-suppressive networks downstream of E2F1, producing a growth advantage to the tumour cells 12,13,57 . Consistently, we demonstrated that the anti-apoptotic factors Survivin and FOXM1 are downstream of the POH1-E2F1 axis. Importantly, these prosurvival factors are involved in the development of various types of cancer, including liver cancer 47,48 . The results based on data mining reveal that the abundance of POH1 positively correlates with the expression of Survivin and FOXM1 in HCC tissues and demonstrate the activation of E2F1 target genes in the subset with high POH1 expression. In summary, we propose that the aberrant upregulation of nuclear POH1-mediated E2F1 stabilization is an important event in the development of liver cancer and that targeting POH1 might serve as a promising strategy for cancer treatment.

Methods
Cell lines and tissue specimens. PLC/PRF/5, SK-Hep1 and HEK293T cells were acquired from the American Type Culture Collection (ATCC, Manassas VA, USA). Authentication of these cell lines was performed using the GenePrint10 System (Promega Biotech Co.) and via comparisons to the ATCC STR database. All the STR profiles yielded 100% matches. The SMMC-7721 cell line was obtained from the cell bank of the Institute of Biochemistry and Cell Biology of the Chinese Academy of Sciences (Shanghai, China). Mouse LPC-Mig, LPC-Akt, and LPC-c-Myc cells were generated in our laboratory 39 . Primary poh1 w/w and poh1 f/f MEFs were obtained from embryos at embryonic day 13.5 from poh1 w/w and poh1 f/f mice. The HEK293T cells and the packaging plasmids were used for lentivirus production. Transduced cells were isolated by puromycin selection or FASC sorting. Cell lines were maintained at 37°C in 5% CO 2 in DMEM supplemented with 10% fetal bovine serum. Cell lines were tested routinely for mycoplasma before use in an experiment. Human HCC samples and matched non-tumoral liver tissues were obtained from the Renji Hospital, Shanghai Jiaotong University, China. The use of human material was approved by the ethical review committee of Renji Hospital and informed consent was obtained from these subjects. Three sets of commercially available tissue microarrays (TMA) containing 154 HCC and nontumoral tissue pairs were used for IHC staining.
Reagents and primary antibodies. Doxycycline (DOX), CHX and puromycin were from Sigma Aldrich. MG132 were from Millipore. Lipofectamine 2000 or   Lipofectamine 3000 Transfection Reagent was from Invitrogen. The following antibodies used for western blotting: POH1 (Proteintech, 12059-1-AP For the luciferase reporter assay, the pGME2F-luc plasmid was purchased from Genomeditech. The duplex siRNAs were chemically synthesized by Genepharma (Shanghai, China), and the sequences of the siRNAs were as follows: Co-immunoprecipitation. Cells were collected and lysed with an IP lysis buffer (Beyotime Institute of Biotechnology, P0013). Total protein (up to 5 mg) was incubated with 50 ml of Protein G-agarose suspension (Millipore, 16-266,) for 3 h at 4°C on a rocking platform to reduce non-specific binding. After removing the beads, the supernatant was supplemented with the primary antibodies followed by incubation for an additional 3 h at 4°C. A total of 100 ml of Protein G-agarose was then added to each immunoprecipitation mixture, and the incubation was continued overnight at 4°C on a rocking platform. The immunoprecipitates were collected by centrifugation and washed three times with the cold 1 Â TBS. After the loading buffer was added, the agarose was boiled and subjected to western blot analysis. The EasyBlot anti-mouse (GTX221667-01) or EasyBlot anti-rabbit (GTX221666-01) IgG HRP-conjugated secondary antibodies (Genetex) were employed to avoid the denatured heavy and light chains from antibodies used in immunoprecipitation assays. The Flag M2 Affinity Gel (Sigma, A2220) was used to immunoprecipited the Flag-tagged proteins, After three times' washes with cold 1 Â TBS the immunoprecipited protein complexes were eluted through 3 Â Flag peptides (Sigma, F4799). The co-immunoprecipated proteins were detected through western blot assay.
Immunohistochemistry. In the immunocytochemical assay, the slides were rehydrated and immersed in 3% hydrogen peroxide solution for 15 min; pretreated by microwave for 25 min in 0.01 mol l À 1 citrate buffer, pH 6.0, at 95°C; and cooled for 60 min at room temperature. In between each incubation step, the sections were washed with PBS, pH 7.4. The slides were blocked by 10% normal goat serum for 30 min at 37°C, washed, and then incubated overnight at 4°C with diluted antibody against each protein studied. After washing with PBS, the slides were visualized using GTVisionTMIII Detection System/Mo&Rb (GeneTech, GK500710) following the manufacturer's instructions. IHC staining for POH1 and E2F1 was cored according to the intensity (1: low staining; 2: moderate staining; 3: high staining; 4: extremely high staining), the percentage of positive cells (1: 0-25%; 2: 26-50%; 3: 51-75%; 4: 76-100%), and the location of staining (1: no nuclear staining or the intensity of nuclear staining was weaker than that of cytoplasm; 2: the intensity of nuclear staining was equal to that of cytoplasm; 3: the intensity of nuclear staining was higher than that of cytoplasm). The overall score ¼ intensity score Â percentage score Â location score.
Xenograft model. The right flanks of male BALB/c nude mice (5 weeks of age) were subcutaneously injected with SK-Hep1 cells (5 Â 10 6 in 0.2 ml PBS) infected with virus expressing either ptripz-sh-POH1or empty vector (n ¼ 12 per group). Each group of mice was then randomly divided into two groups (n ¼ 6 per group). One group was administered doxycycline (2 mg ml À 1 ) to induce to the POH1 shRNA expression and the other group was provided normal water. Tumour growth was monitored every 5 days. The tumour-bearing mice were sacrificed 40 days after inoculation, and the tumours were removed for further study (photographing, weighing, fixing and paraffin-embedding). The left and right flanks of male BALB/c nude mice (5 weeks of age) were subcutaneously injected with PLC/PRF/5 cells (6 Â 10 6 in 0.2 ml PBS). The tumour-bearing mice were killed 35 days after inoculation. All experiments were subject to approval by the Animal Care and Use Committee of Shanghai Cancer Institute.
Quantitative real-time RT-PCR. Total cellular RNA from tissues and cells were extracted by the RNAiso Plus kit (Takara Bio Inc.) and cDNA preparation was preformed according to standard procedures using primeScript RT Master kit (Takara Bio Inc.). Real-time RT-PCR was performed by SYBR gGreen quantitative PCR kit (Takara Bio Inc.) using the 7300 Real-Time PCR System or ViiA7 System (AB Applied Biosystems). The primers used in the mRNA levels detection were as follows: human POH1-F: 5 0 -TTGCTATGCCACAGTCAGGA-3 0 , human POH1-R: Generation of Poh1 conditional knockout mice. Poh1 knockout mouse model was created by Beijing Biocytogen. In brief, homology regions covering 8.0 kb upstream of Poh1 exon 6 and 5.7 kb downstream of exon 6 were subcloned from a BAC clone (RP23-1O11; Invitrogen) from C57BL/6J mouse genomic BAC library. FRT-flanked Neo resistance-positive selection cassette was inserted downstream of exon 6 and two loxP sites were introduced upstream of exon 6 and downstream of exon 6, respectively. After linearization, the targeting vector was transfected into C57BL/6J embryonic stem cells (Biocytogen) by electroporation. Seven positive clones were identified by Southern blotting with 5 0 -probe, 3 0 -probe and Neo probe. Two positive clones were injected into Balb/c blastocysts and implanted into pseudopregnant females. Chimeric mice were crossed with C57BL/6J mice to obtain F1 mice carrying the recombined allele containing the floxed Poh1 allele and Neo selection cassette. These mice were mated with Flp recombinase expressing C57BL/6J Flp mice to remove the Neo resistance cassette and generate a line of Neo-excised floxed mice. Heterozygous Neo-excised, Poh1-floxed mice were then crossed with C57BL/6J Mx-Cre mice to achieve the genomic deletion of Poh1 exon 6 from the floxed alleles. Pathogen-free mice were housed under specific pathogenfree conditions and fed food and water regularly. All of the experiments involving mice were performed according to the Institutional Animal Care and Use Committee of Shanghai and the National Research Council Guide for Care and Use of Laboratory Animals. Genotyping was performed by PCR using the proper primers. By crossing Mx-Cre mice with Poh1 f/f mice and then backcrossing, we generated interferon-inducible Poh1 knockout mice. Poh1 f/f , Mx-Cre þ mice (6-8 weeks old) were injected with a total of three injections of 5 mg g À 1 body weight of polyI:C to induce Poh1 deletion. The Poh1 WT, Mx-Cre þ mice (6-8 weeks old) were administered the same amount of polyI:C injections on the same dosing schedule; these mice were used as controls. The primers used to detect the deleted poh1 include F: 5 0 -TGTGATTGCTGTTTTATGAGGCA-3 0 and R: 5 0 -GAGAATGAC AACTATTGGGAGACTTAGC-3 0 .
Western blot. Cells or tissues were lysed with RIPA buffer (Thermo Fisher Scientific, 89901) containing protease inhibitors cocktail (Roche Diagnostics, 05892970001) and phosphatase inhibitor cocktail (Roche Diagnostics, 04906845001). The lysates were clarified by centrifugation at 13,000 g for 30 min at 4°C. The total protein concentration was estimated using a BCA protein assay kit (Thermo Fisher Scientific, 23225). Protein samples (50-150 mg) were loaded on to and separated using SDS/PAGE, transferred on to NC membranes (Pall Corporation) blocked and probed with the primary antibodies. After washing, the blots were incubated with goat anti-rabbit (Santa Cruz, sc-2004) or goat anti-mouse (Santa Cruz, sc-2005) HRP (horseradish peroxidase)-conjugated secondary antibodies (Santa Cruz Biotechnology) and visualized using the SuperSignal West Dura Extended Duration Substrate (Thermo Fisher Scientific, 34076). The uncropped versions of western blots are shown in Supplementary Fig. 17.
Immunofluorescence staining. Cells were seeded on coverslips and washed with PBS buffer. Cells were fixed with 4% paraformaldehyde at room temperature for 15 min and permeabilized with 0.1% Triton X-100 for 10 min. After washing with PBS buffer, cells were incubated with a blocking solution (5% goat serum) for 30 min at room temperature and then incubated with primary antibodies overnight at 4°C. Cells were then washed three times and incubated with the anti-mouse or anti-rabbit Alexa Fluor secondary antibodies (dilution 1:200, Invitrogen) for 30 min at room temperature. DNA was counterstained using 4, 6 -diamidino-2phenylinodole (DAPI). The slides were observed under a ZEISS Axiovert 200 fluorescence optics microscope (Zeiss Shanghai, China).
Cell viability analysis. Cells with different treatments were seeded at 1,500 cells in 200ml DMEM per well in 96-well culture plates. At the indicated time points, 20 ml of 0.5 mg ml À 1 MTT (Thiazolyl Blue Tetrazolium Bromide, M5655, Sigma) was added to each well and incubated at 37°C for 3 h. Then, the culture was replaced with 150 ml dimethyl sulphoxide (D8418, Sigma) to stop the reaction. The absorbance values (OD 590 nm) were measured using a spectrophotometer (Thermo Fisher Scientific).
Colony-formation assays. To assay the proliferation potential of cells, cells were seeded at 2,000 cells in 2 ml DMEM per well in the six-well culture plates. After 10 days' culture, cells were fixed with methanol and stained with crystal violet (Beyotime, C0121).
Apoptosis analysis. For apoptosis analysis, cells were placed in six-well plates with the confluence of 50% 24 h before analysis, cells were harvested by trypsin digestion, washed with binding buffer and stained with 5 ml annexin-V-PE (BD, 556422) or annexin-V-APC (ebioscience, BMS306APC-100) and 0.5 ml 7-AAD (7-aminoactinomycin D, A9400, Sigma) in 100 ml binding buffer for 30 min at room temperature in the dark. Apoptotic cells were detected by flow cytometry using a FACSCalibur flow cytometer (BD). For anoikis analysis, cells were cultured in the 6-well plates with polystyrene-coated low attachment surface (3471, Corning Incorporated) for 72 h, then cells were collected for annexin V/7-AAD double staining.
Cell cycle analysis. Cells were placed in six-well plates with the confluence of 50% 24 h before analysis; cells were collected by trypsin digestion, and resuspended in 75% ethanol at 4°C overnight. The cells were collected by centrifugation and washed with 1 Â PBS. Finally, the cells were resuspended in 1 Â PBS containing 100 g ml À 1 RNase A and 50 g ml À 1 propidium iodide (PI). After incubation for 30 min at 37°C in dark, samples were subjected to flow cytometry for cell cycle analysis.
Luciferase reporter assay. HEK293T cells were transiently transfected with luciferase reporter plasmid (300 ng), POH1 expression plasmid (1 mg) or POH1 siRNA (100 nM). To correct the transfection efficiency variations in each group of transfection, 20 ng of the Renilla luciferase plasmids were co-transfected in each experiment. Forty-eight hours post-transfection, the firefly and Renilla luciferase activities were monitored using the Dual-Luciferase Reporter Assay System (Promega). The data are shown as the ratio of firefly to Renilla luciferase activity.