The Transcription Factor PAX4 Acts As a Survival Gene in INS-1E Insulinoma Cells

Oncogene (2007) 26, 4261–4271 & 2007 Nature Publishing Group All rights reserved 0950-9232/07 $30.00 www.nature.com/onc ORIGINAL ARTICLE The transcription factor PAX4 acts as a survival gene in INS-1E insulinoma cells

T Brun, DL Duhamel, KH Hu He, CB Wollheim and BR Gauthier

Department of Cell Physiology and Metabolism, University Medical Center, Geneva 4, Switzerland

The paired/homeodomain transcription factor Pax4 is domain nuclear factor Pax4 is essential for b-cell essential for islet b-cell generation during pancreas maturation. Indeed, Pax4 mRNA is detected in the development and their survival in adulthood. High Pax4 mouse pancreatic bud at E9.5, becoming maximal at expression was reported in human insulinomas indicating E13.5–15.5 during the period of massive b-cell expan- that deregulation of the gene may be associated with sion and thereafter declining. In mature islets, low levels tumorigenesis. We report that rat insulinoma INS-1E of Pax4 expression are detected in human, mouse and cells express 25-fold higher Pax4 mRNA levels than rat rat b-cells (Zhang et al., 2001; Heremans et al., 2002; islets. In contrast to primary b-cells, activin A but not Kojima et al., 2003; Zalzman et al., 2003; Brun et al., betacellulin or glucose induced Pax4 mRNA levels 2004, 2005; Theis et al., 2004; Gunton et al., 2005). indicating dissociation of Pax4 expression from insulino- Consistent with its tissue-specific expression pattern, ma cell proliferation. Short hairpin RNA adenoviral targeted disruption of the pax4 gene in mice results constructs targeted to the paired domain or homeodomain in the absence of mature pancreatic b- and d-cells with (viPax4PD and viPax4HD) were generated. Pax4 mRNA a commensurate increase in the glucagon-producing levels were lowered by 73 and 50% in cells expressing a-cells (Sosa-Pineda et al., 1997; Wang et al., 2004). This either viPax4PD or viPax4HD. Transcript levels of the increase was attributed to the a-cell-specific transcrip- Pax4 target gene bcl-xl were reduced by 53 and 47%, tion factor Arx, which is repressed by Pax4 during whereas Pax6 and Pdx1mRNA levels were unchanged. development (Collombat et al., 2003). Interestingly, viPax4PD-infected cells displayed a twofold increase simultaneous inactivation of the arx and pax4 gene in spontaneous apoptosis and were more susceptible in mice islets promotes a somatostatin-producing dÀ to cytokine-induced cell death. In contrast, proliferation cell fate specification to the detriment of the a-and was unaltered. RNA interference-mediated repression of b-cell lineages, indicating that Pax4 is redundant for d- insulin had no adverse effects on either Pax4 or Pdx1 cell formation under certain conditions (Collombat expression as well as on cell replication or apoptosis. et al., 2005). These results indicate that Pax4 is redundant for Several studies in the Japanese population and one in proliferation of insulinoma cells, whereas it is essential Afro-Americans have associated mutations in the pax4 for survival through upregulation of the antiapoptotic gene gene to type II diabetes, whereas polymorphisms have bcl-xl. been linked to type I diabetes in Scandinavian, German Oncogene (2007) 26, 4261–4271; doi:10.1038/sj.onc.1210205; and Swiss families (Shimajiri et al., 2001, 2003; published online 29 January 2007 Kanatsuka et al., 2002; Holm et al., 2004; Mauvais- Jarvis et al., 2004; Biason-Lauber et al., 2005; Tokuya- Keywords: RNAi; insulin; Pax4; proliferation; apoptosis; ma et al., 2006). These genetic studies suggest an survival important role of this factor in regulating b-cell function and/or mass in adult islets. Accordingly, we have demonstrated that increased Pax4 mRNA levels in rat islets coincide with b-cell proliferation induced by the mitogens activin A and betacellulin. Overexpression of Introduction Pax4 in rat islets resulted in the induction of the c-myc/ Id2 proliferation pathway and of the antiapoptotic gene Normal islet b-cell mass is achieved through the inter- bcl-xl. The coordinated stimulation of these genes by play of numerous transcription factors during pancreas Pax4 prompted b-cell replication and conferred protec- organogenesis. In particular, the paired homeo- tion against cytokine-induced apoptosis in human islets (Brun et al., 2004). These results clearly show that Pax4 Correspondence: Dr BR Gauthier and T Brun, Department of Cell acts as a key regulator of b-cell plasticity in mature Physiology and Metabolism, University Medical Center, 1211 Geneva islets, substantiating the concept that terminally differen- 4, Switzerland. tiated b-cells retain a significant capacity to proliferate E-mails: [email protected] and thierry.brun@ medecine.unige.ch in response to physiological stimuli (Dor et al., 2004). Received 5 May 2006; revised 23 October 2006; accepted 9 November Human insulinomas were shown to contain aber- 2006; published online 29 January 2007 rantly high expression levels of Pax4 as compared to Pax4 and apoptosis T Brun et al 4262 normal islets (Miyamoto et al., 2001). Furthermore, aberrant DNA demethylation in the promoter region of the pax4 gene was recently associated with induced expression of the transcript and development of haematologic malignancies (Li et al., 2006). These results clearly implicate the transcription factor in tumorigenesis. Consistent with the latter, several members of the PAX family members have been associated with cancer; increased levels of Pax3 and Pax7 are observed in sarcomas and neural crest-derived human tumours whereas uncontrolled expression of Pax2 and Pax8 is associated with kidney, prostate, breast and ovary carcinomas. Pax5 was also identified as a key factor for the maintenance of the neuroblastoma phenotype (for review, see Robson et al., 2006). The aim of the current study was to determine whether sustained deregulated expression of Pax4 is permissive for insulinoma growth or is facilitating tumour cell survival. To address these questions, we took advantage of the well-characterized rat insulinoma cell line INS-1E (Merglen et al., 2004). This stable clone was derived from the parental INS-1 cell line and exhibits a high proliferation index (Asfari et al., 1992). Thus, INS-1E provides a suitable cell model to investigate the impact of Pax4 on cell proliferation and/or survival. INS-1E cells were found to contain 25-fold higher levels of Pax4 mRNA as compared to rat islets. Targeted suppression of Pax4 using RNA interference resulted in downregulation of the bcl-xl gene with a concomitant increase in cell death. In contrast, cell proliferation was unaltered, suggesting that Pax4 is crucial for survival but not replication of the insulinoma.

Results

Pax4 is highly expressed in INS-1E cells but does not correlate with the proliferative capacity of the insulinoma Figure 1 Pax4 is highly expressed in INS-1E cells but does not correlate with the proliferative capacity of the insulinoma. High levels of Pax4 mRNA have been previously shown (a) Higher levels of Pax4 and Bcl-xL transcripts are expressed in in human insulinomas and proposed to be a predisposi- the rat insulinoma INS-1E cell line as compared to adult rat islets. tion factor for tumour formation and/or progression QT-PCR using RNA purified from INS-1E cells and freshly (Miyamoto et al., 2001). Accordingly, we found that the isolated rat islets were performed on Pax4, Bcl-xL (left panel) and rat insulinoma INS-1E cell line contained 25-fold more Serca3 (right Panel). Data are presented as fold change mRNA levels as compared to primary cells and normalized to the Pax4 mRNA as compared to rat islets (Figure 1a). housekeeping gene transcript cyclophilin. Each value represents Interestingly, transcript levels for the anti-apoptotic the mean7s.e. of four independent experiments performed in gene bcl-xl, a downstream target of Pax4, were also duplicate. *Po0.05; **Po 0.01. (b) Pax4 mRNAs in INS-1E cells increased fivefold in INS-1E cells. In contrast, mRNA were treated with increasing doses of activin A and betacellulin as indicated on the graph. The results are normalized to cyclophilin levels for the unrelated serca3 gene were decreased (or Rsp291, see Supplementary Information) and are expressed as twofold as compared to islets (Figure 1a). To determine fold change of mRNA as compared to control. (c) INS-1E cells and whether a correlation could be established between Pax4 rat islets were incubated with increasing concentrations of glucose. expression levels and the capacity of INS-1E cells to Pax4 and c-myc transcript abundance levels were then estimated by proliferate, the cells were exposed to the mitogens QT-PCR. Statistical significances were carried out in between control INS-1E cells and islets incubated with the various growth activin A or betacellulin, which were shown to stimulate factors by Student’s t-test. *Po0.05; **Po0.01. Pax4 transcription as well as replication in primary rat b-cells (Brun et al., 2004). Cells treated with increasing concentrations of activin A for 24 h exhibited a dose- dependent stimulation of Pax4 expression, reaching constant when normalized to a second housekeeping maximal induction of 11-fold at 0.5 nM. Surprisingly, gene (RpS29) (Theander-Carrillo et al., 2006) indicating betacellulin provoked only a small nonsignificant that cyclophilin transcription is refractory to the effect increase in Pax4 mRNA levels (Figure 1b). The of mitogens (Supplementary data, panel A). Consistent reference gene cyclophilin expression levels remained with this premise, similar activin A-induced increases in

Oncogene Pax4 and apoptosis T Brun et al 4263 Pax4 transcript levels were observed when the expres- abrogated the shRNA-mediated repression of insulin sion profile was normalized with RpS29, whereas no mRNA levels. More importantly, downregulation of the alterations were detected in the presence of betacellulin insulin transcript had no effect on either Pdx1 or Pax4 (Supplementary data, panel B). The discrepancy ob- mRNA levels (Figure 3b). Consistent with QT-PCR served between the two mitogens prompted us to results, immunofluorescence experiments revealed that investigate whether glucose, a potent activator of INS- GFP-negative cells stained for insulin, whereas no 1 proliferation (Hugl et al., 1998), could induce Pax4 staining was detected in GFP þ /siINS6 cells. Cells expression. Similar to betacellulin, glucose failed to expressing siMut1 retained insulin staining (Figure 3c). increase Pax4 mRNA levels in INS-1E, whereas a robust Taken together, these results clearly demonstrate that 10-fold induction was detected in rat islets (Figure 1c). specific repression can be achieved through RNA The latter was also accompanied by a fourfold increase interference and that inhibition of Pax4 by siPax4PD in c-myc mRNA at 30 mM glucose, correlating with or siPax4HD but not by siINS6 resulted predominantly stimulation of cell replication (Brun et al., 2004). Taken in the suppression of Bcl-xL mRNA levels. together, although high levels of Pax4 are detected in INS-1E cells, further induction is not observed with Repression of Pax4 selectively induces apoptosis betacellulin and glucose, suggesting that mitogen- without altering proliferation mediated increases in INS-1E proliferation do not We next sought to determine the physiological con- correlate with modulation in Pax4 expression levels. sequences of shRNA-mediated repression of Pax4 in INS-1E cells. This was initially studied by transient RNA interference-mediated suppression of Pax4 transfection assays in which cells coexpressing a specifically inhibits mRNA levels of the antiapoptotic phogrin-GFP recombinant protein and either siPax4PD gene bcl-xl or siPax4HD were analysed for cell apoptosis by To determine the specific contribution of Pax4 to the terminal deoxynucleotidyl transferase-mediated dUTP insulinoma phenotype, we inhibited expression of the nick end labelling (TUNEL) assay (Figure 4a) or transcription factor by RNA interference. To this end, proliferation by 5-bromo-20-deoxy-uridine (BrdU) in- two regions of Pax4, the paired (PD) and the homeo corporation (Figure 4b). Increased colocalization of (HD) domains, were individually targeted to circumvent phogrin-GFP and TUNEL was observed by immu- potential nonspecific effects of a single short hairpin nofluorescence in cells expressing either siPax4PD or RNA (shRNA). As no satisfactory Pax4 antibodies are siPax4HD as compared to U6- or siScr-transfected cells. commercially available to detect the protein by immu- In contrast, repression of insulin did not induce nohistochemistry or by Western blotting, repression was apoptosis in INS-1E cells (Figure 4a). Consistent with monitored by quantitative real-time polymerase chain unaltered Id2 mRNA levels, cells retained their capacity reaction (QT-PCR). Pax4 steady-state mRNA levels to proliferate in the presence of siPax4PD and were reduced by 54 and 40% in INS-1E cells coexpres- siPax4HD (Figure 4b). sing GFP (used to enrich the transfected cell population) In order to measure accurately apoptosis and and either siPax4PD or siPax4HD, whereas no inhibi- proliferation, INS-1E cells were infected with adenoviral tion was detected with either empty vector (U6) or a constructs vU6, viScr, viPax4PD, viPax4HD or viINS6. scrambled shRNA (siScr) (Figure 2a). Repression was The latter approach ensures delivery of the shRNAs in specific to Pax4, as mRNA levels for the highly approximately 80% of cells (Brun et al., 2004). Gene homologous Pax family member Pax6, also expressed profiling revealed that Pax4 mRNA levels were dose in INS-1E cells, as well as the homeodomain containing dependently inhibited by either viPax4PD or viPax4HD transcription factor Pdx1 were unaltered (Figure 2b and (Figure 5a). Consistent with transfection studies, c). Furthermore, STAT1 mRNA levels were refractory viPax4PD was more effective than viPax4HD in to siPax4PD and siPax4HD, indicating that the inter- suppressing Pax4, reaching 73710% inhibition as feron system implicated in double-stranded RNA- compared to 5075%. A similar dose- and siRNA- mediated cell apoptosis was not activated by the levels dependent inhibition was observed for Bcl-xL mRNA of shRNAs being transfected (Figure 2d) (Sledz et al., levels (5376% for viPax4PD versus 4775% for 2003). In contrast, the transcript of the Pax4 target gene viPax4HD; Figure 5d). In contrast, transcript levels of bcl-xl (Brun et al., 2004) was inhibited by 55 and 30% in Pdx1 and Pax6 were unchanged (Figure 5b and c). siPax4PD- and siPax4HD-expressing cells, respectively. Concomitantly, TUNEL immunostaining increased in Surprisingly, Id2 mRNA levels were significantly viPax4PD- and viPax4HD-infected cells as compared to increased by approximately 75 and 50% in the presence viScr-, vU6- and viINS6-transduced cells (Figure 6a). of siPax4PD and siPax4HD, respectively, as compared Quantification revealed that the highest concentration of to cells transfected with U6 (Figure 2f). To further viPax4PD and viPax4HD induced apoptosis in approxi- substantiate the specificity of siPax4PD- and siPax4HD- mately 25.672 and 18.773% of INS-1E cells, respec- mediated repression of Pax4, we generated an shRNA tively, whereas control viScr-infected cells exhibited on targeted to insulin (siINS6) and evaluated its impact on average 1271% cell death (Figure 6b). Interestingly, insulin, Pdx1 and Pax4 transcripts (Figure 3). Insulin increasing the amount of either viPax4PD or viPax4HD steady-state mRNA levels were inhibited by 75% in resulted in massive apoptosis, indicating that further INS-IE cells coexpressing GFP and siINS6 (Figure 3a). repression of Pax4 was lethal. Consistent with the lack A single point mutation in siINS6 (siMut1) completely of viINS6 effect on Pax4 expression, no significant

Oncogene Pax4 and apoptosis T Brun et al 4264

Figure 2 Repression of Pax4 by RNAi speciﬁcally inhibits mRNA levels of Bcl-xL. Attempts to inhibit Pax4 mRNA were made by creating shRNAs targeted to either the paired domain (siPax4PD) or to the homeodomain (siPax4HD) of the transcription factor. The empty vector U6 and U6 containing a scrambled shRNA (siScr) served as controls. Subsequent to co-transfection with the appropriate shRNA- and GFP-expressing vectors, INS-1E cells were FACS puriﬁed into GFP- (white bars) and GFP þ (black bars) subgroups, RNA extracted and transcript levels for (a) Pax4, (b) Pdx1, (c) Pax6, (d) Stat1, (e) Bcl-xL and (f) Id2 were determined by QT-PCR and normalized to cyclophilin. Results are the mean of 4–9 independent experiments performed in duplicate and are expressed as % change of mRNA as compared to GFPÀ cells. *Po0.05 and **Po0.01.

difference was observed in cells expressing this shRNA Discussion as compared to viScr-transduced cells (Figure 6b). Addition of cytokines to the media provoked a further The present study establishes a causal association 1.7-fold increase in apoptosis, suggesting that INS-1E between Pax4 mRNA levels and susceptibility of INS- cells were more susceptible to apoptosis subsequent to 1E cells to apoptosis, suggesting an important phy- Pax4 repression (Figure 6c). In contrast, quantification siological function of the transcription factor in cell of BrdU staining confirmed the ability of INS-1E cells survival. Accordingly, suppression of Pax2, Pax7 or to sustain replication independent of the adenoviral Pax3 in tumour cell lines has been shown to increase cell construct applied (Figure 7a and b). To ensure that death (Margue et al., 2000; Muratovska et al., 2003; He uniformity of labelling was not due to prolonged et al., 2005), indicating that high expression levels of pax exposure to BrdU (72 h), the analogue was added 4 or genes are most likely essential for evading the apopto- 16 h before processing cells for immunofluorescence. tic programme, a hallmark of cancer (Hanahan and The latter revealed a partial BrdU labelling, which Weinberg, 2000). Interestingly, deletion of two Pax2/5/ increased with time (Figure 7c). However, no differences 8-related genes (egl-38 and pax2)inCaenorhabditis in proliferation were detected in viPax4PD-expressing elegans was shown to induce apoptosis in somatic and cells as compared to non-infected or viScr-transduced germline cells via the antiapoptotic gene bcl-2 (Park cells (Figure 7d). et al., 2006). Similarly, we have previously demonstrated

Oncogene Pax4 and apoptosis T Brun et al 4265

Figure 3 Specific downregulation of insulin mRNA levels in INS-1E cells expressing an insulin-targeted shRNA. (a) INS-1E cells were co-transfected with the empty vector (U6, control), the shRNA targeted to insulin (siINS6) or the siINS6 containing a single point mutation (siMut1) along with the reporter GFP-expressing vector and subsequently processed as in Figure 2. Transcript levels for (a) insulin, (b) Pdx1 and Pax4 were determined by QT-PCR and normalized to cyclophilin. Results are the mean of least three independent experiments performed in duplicate and are expressed as % change of mRNA as compared to GFPÀ cells. **Po0.01. (c) Immunofluorescent detection of GFP (green) and insulin (red) in INS-1E cells 72 h after transfection with the indicated plasmids. No staining was detected in GFP þ /siINS6 cells (arrows) whereas cells expressing the control vectors U6 or siMut1 retained insulin staining. that increased Pax4 expression not only protected cells. Interestingly, Pax6 is also expressed in endocrine against cytokine-induced apoptosis but also promoted cells as well as in INS-1E and was previously shown to proliferation of primary rat and human islet b-cells. induce pancreatic cystic adenoma in transgenic animals Taken together, these observations indicate that INS-1E bearing the transcription factor under the Pdx1 promo- cells, in contrast to mature b-cells, have selectively ter (Yamaoka et al., 2000). Pax6 could therefore exploited high expression levels of Pax4 to confer a potentially be involved in INS-1E cell proliferation by strong advantage to survival while potentially short- overriding the effect of Pax4. However, we have circuiting the impact of the transcription factor on cell previously demonstrated that in contrast to Pax4, proliferation. This premise is supported by our findings overexpression of Pax6 in islets did not promote that repression of Pax4 in INS-1E by RNA interference proliferation (Brun et al., 2004). Consistent with the did not inhibit proliferation. Furthermore, Id2, a latter, several studies have also shown that increased downstream target of the proto-oncogene c-myc regu- Pax6 expression correlates with tumour suppression lated by Pax4 in primary b-cells, was increased rather rather than stimulation of cell replication in human than decreased in Pax4-repressed INS-1E cells. More- cancers (Zhou et al., 2003, 2005). Therefore, these over, rat islets and INS-1E cells have similar Id2 mRNA studies do not support a fundamental role of Pax6 in levels (data not shown), suggesting that Id2 regulation is b-cell replication. Most likely, other adaptive mechan- not important for proliferation in these transformed isms such as upregulation of the telomerase gene prevail

Oncogene Pax4 and apoptosis T Brun et al 4266

Figure 4 Repression of Pax4 induces apoptosis without altering proliferation. INS-1E cells were co-transfected with the indicated vectors along with the reporter phogrin-GFP-expressing vector for 72 h as described in Materials and methods. Cell death (a) and cell proliferation (b) were measured in INS-1E cells by the TUNEL assay and by BrdU incorporation, respectively. For proliferation, cells were cultured in the presence of BrdU (10 mM) for 72 h. Representative composite images of INS-1E cells positive for TUNEL (a)or immunostained for BrdU (b) (red), GFP (green) and DAPI (nuclei, blue) are shown. Arrows depict INS1-E cells expressing siPax4PD or siPax4HD. Bars, 50 mM.

in the insulinoma to convey self sustained cell replica- incorporation in INS-1 cells (Huotari et al., 1998; tion (Blasco, 2005). Buteau et al., 2003). In agreement with a previous study Consistent with the potential dissociation of Pax4 (Ueda, 2000), we found that activin A stimulated Pax4 expression from proliferation, treatment of INS-1E expression in INS-1E cells. This growth factor mediates cells with betacellulin or glucose did not stimulate its effect through the TGF-b signalling pathway Pax4 expression, whereas the sugar evoked an 11-fold independent of the PI3-kinase (Feng and Derynck, induction of Pax4 mRNA with a concomitant increase 2005), which may explain its ability to induce Pax4 in c-myc gene expression in rat islets. Accordingly, we expression in INS-1E cells. Thus, regulation of the pax4 have also demonstrated that betacellulin induced Pax4 gene in INS-1E cells appears to be governed by the expression in rat b-cells (Brun et al., 2004). Further- action of several converging signal-transduction path- more, we found that Pax4 was stimulated in human ways but that intrinsic high levels of the transcription primary b-cells by both betacellulin and glucose (Brun factor do not correlate with cell proliferation. et al., 2005). These growth factors are known to mediate Consistent with this premise, we found that inhibition their mitogenic effects on INS-1 cells through the PI3- of Pax4 expression using RNA interference induced kinase signalling pathway (Barker et al., 2002; Buteau apoptosis in the insulinoma cells without any impact on et al., 2003). Paradoxically, betacellulin was originally proliferation. We also observed a concomitant decrease isolated from the conditioned medium of cell lines in the antiapoptotic gene bcl-xl, which was previously derived from mouse pancreatic b-cell tumours (Shing demonstrated to be transcriptionally regulated by Pax4 et al., 1993). Self-sufﬁciency in growth is a hallmark of (Brun et al., 2004). Accordingly, both Pax4 and Bcl-xL cancer in which tumour cells escape the requirement of mRNA levels were dose- dependently inhibited by exogenous derived signals in order to survive (Hanahan increasing amounts of viPax4PD and viPax4HD, and Weinberg, 2000). It is thus conceivable that de- resulting in a graded increase of apoptosis. Interestingly, regulated high Pax4 expression in INS-1E cells is a Henderson et al. (2005) showed that the moderate consequence of elevated endogenous betacellulin levels reduction in Bcl-xL expression observed in bcl-xl þ /À trans- and that further activation by the PI3-kinase pathway is genic mice conferred protection against tumorigenesis. inefﬁcient. However, this does not preclude the potential Conversely, high expression levels of Bcl-xL were detected action of betacellulin on other target genes involved in in rhabdomyosarcoma cell lines and conveyed resistance proliferation, as the mitogen stimulates thymidine to cell death (Margue et al., 2000). Remarkably, the

Oncogene Pax4 and apoptosis T Brun et al 4267

Figure 5 Dose-dependent decrease in both Pax4 and Bcl-xL transcripts in INS-1E cells transduced with viral constructs harbouring either shPax4PD or shPax4HD. INS-1E cells were exposed to either 1 Â 107 PFU/ml (white bars) or 2 Â 107 PFU/ml (black bars) of the indicated adenoviruses for 72 h and transcript levels were subsequently measured for (a) Pax4, (b) Pdx1 (c) Pax6 and (d) Bcl-xL. Expression levels were normalized to the reference gene cyclophilin. Results are the mean of four independent experiments performed in duplicate and are expressed as % change of mRNA as compared to vU6-transduced INS-1E cells. *Po0.05 and **Po0.01. latter anti-apoptotic effect was mediated through the instances, INS-1E cells were exposed to 0.1, 0.5 and 2 nM of direct transcriptional activation of Bcl-xL by Pax3. activin A and betacellulin (Sigma, Basel, Switzerland) for 24 h. Likewise, Bcl-xL mRNA levels were found to be fivefold In parallel, islets as well as INS-1E cells were also cultured in higher in INS-1E cells as compared to islets corroborat- the presence of 2.5, 10 and 30 mM glucose for 24 h. ing the increased Pax4 expression. Suppression of Pax4, and consequently Bcl-xl, leads to increased apoptosis, a RNA interference phenomenon observed with other pax genes (Bernasconi RNAi experiments were performed using a novel expression et al., 1996; Margue et al., 2000; Ostrom et al., 2000; vector that was created using the Adeno-X Tet-On expression Henderson et al., 2005). Taken together, these findings system from BD Clontech (Basel, Switzerland). Briefly, a PmeI strongly suggest that increased expression of Pax4 and primer was introduced into the NheI site of the pTRE-Shuttle2 the activation of its downstream target gene bcl-xl are vector. Subsequently, restriction digest of the vector was critical determinants in sustaining the insulinoma performed using MluI and PmeI to release a fragment phenotype. In support of this survival function, we find containing the Tet-responsive element as well as the CMV promoter. The MluI/PmeI vector backbone was then purified that subsequent to Pax4 inhibition, INS-1E cells are and ligated to a mouse U6 promoter MluI/PmeI fragment more susceptible to cytokine-induced apoptosis. In amplified by PCR from a construct kindly provided by Dr conclusion, we show that Pax4 acts as a survival gene Altman (Yale University, CN, USA). The resulting plasmid, in INS-1E cells, a process requiring upregulation of the denoted as pDLDU6 (U6), contains a novel PmeI site that antiapoptotic gene bcl-xl. allows for the insertion of hairpin siRNA sequences at the þ 1 position of the endogenous U6 transcript. Two 21-nucleotide Pax4 shRNA structures with a 6-nucleotide loop were Materials and methods synthesized as two DNA oligonucleotides and annealed and ligated to the PmeI and XbaI sites of U6. The targeted Cell culture sequences were; (1) siPax4PD, aimed at the sequence 50-GGC INS-1E cells (Merglen et al., 2004) were cultured in RPMI- TCG AAT TGC CCA GCT AAA-30 located in the paired 1640 (Invitrogen, Basel, CA, USA) supplemented with 10% domain and (2) siPax4HD, targeted to the sequence 50-GCA fetal calf serum (FCS; Brunschwig AG, Basel, Switzerland) GAC AAG AGA AGT TGA AAT-30 found in the homeo- and other additions as described previously (Asfari et al., domain of Pax4. To monitor specificity of Pax4 repression and 1992). Pancreatic islets were isolated from Wistar rats as to control against physiological artefacts induced by RNAi, an previously described (Gauthier et al., 2004). In several shRNA directed against insulin (siINS6; 50-CAG GCT TTT

Oncogene Pax4 and apoptosis T Brun et al 4268

Figure 6 Repression of Pax4 using adenoviral vectors induces cell death and increased susceptibility to cytokine-induced apoptosis. INS-1E cells were infected with the indicated adenoviral vectors and apoptosis was measured by the TUNEL assay 72 h posttransduction. (a) Representative immunofluorescent composite images of INS-1E cells revealing insulin- (green), DAPI- (nuclei, blue) and TUNEL (red)-positive cells. Arrows depict apoptotic INS1-E cells expressing viPax4PD or viPax4HD. (b) TUNEL-positive INS-1E cells were counted under a fluorescent microscope in five separate fields (each containing approximately 50 cells) of five independent experiments and graphical results are depicted as a percentage of TUNEL/insulin positive cells over the total amount of insulin positive cells. (c) Infected INS-1E cells were exposed for the last 24 h to IFN-g, IL-1b and TNF-a (2 ng/ml each) to induce apoptosis. Similar to (b), TUNEL-positive cells were counted and results are depicted as a percentage of TUNEL/insulin-positive cells over the total amount of insulin-positive cells. *Po0.05 and **Po0.01. Bars, 50 mM.

GTC AAA CAG CAC-30) and also a scrambled shRNA (siScr; monitored by real-time PCR. In parallel, cell proliferation as 50-CGG CGT TAG CGA TTA GAT GAT-30) were generated. well as cell death assays were performed on GFP-phogrin- Subsequently, the various U6/shRNAs cassettes were trans- co-transfected cells. The latter recombinant protein, targeted ferred into the promoter-deﬁcient Adeno-X viral DNA to to insulin granules, was used for these experiments as GFP was generate recombinant adenoviruses (viPax4HD, viPAx4PD, found to interfere with the immunohistochemical detection of viINS6 and viScr). apoptosis detected by the TUNEL assay (data not shown).

Cell transfection Adenoviral infection of INS-1E cells and cytokine treatment Constructs were transfected into INS-1E cells along with a Before infection (24 h), INS-1E cells were seeded at reporter GFP-expressing vector (ClonTech) using the Lipo- 3 Â 105 cells/ml in Falcon 24-well plates. Cells were then fectamine 2000 reagent according to the manufacturer’s exposed to 1 or 2 Â 107 PFU/ml of recombinant adenoviruses guidelines (Invitrogen). The ratio of shRNA plasmid to for 3 h, washed, replenished with fresh media and cultured for reporter GFP vector was maintained at 1:2. Subsequent to an additional 48–72 h. In some instances, infected cells were cell sorting using GFP (72 h post-transfection), the effect of incubated in the presence of IFN-g, IL-1b and TNF-a (2 ng/ siPax4PD, siPax4HD, siINS6 and siScr on selected genes was ml) for 24 h to induce apoptosis.

Oncogene Pax4 and apoptosis T Brun et al 4269

Figure 7 Proliferation is sustained in Pax4-repressed INS-1E cells. INS-1E cells were transduced with the indicated concentrations of adenoviruses for 72 h. Cell proliferation was then assessed by BrdU incorporation. Representative composite images of INS-1E cells immunostained for BrdU (red), insulin (green) and DAPI (blue) are shown subsequent to a 72 h (a), as well as a 4 and 16 h (c) BrdU treatment (BrdU, green; and insulin, red). BrdU-positive cells were counted under a ﬂuorescent microscope as described in Figure 6 and graphical results are depicted as a percentage of positive cells over the total amount of insulin-positive cells (b and d). Note that only a fraction of cells are labelled in (c) and that, similar to (b), no alterations in proliferation are detected among the various experimental groups. Data represent the mean7s.e. of at least three independent experiments.

Quantitative real-time PCR (QT-PCR) Immunohistochemistry Total RNA was extracted from INS-1E cells using the Qiagen INS-1E cells were cultured on polyornithine-treated glass RNeasy mini kit and 2 mg was reversed transcribed into cDNA coverslips for 3 days, washed with phosphate-buffered saline as described previously (Gauthier et al., 1999). Primers for (PBS) and fixed in 4% paraformaldehyde in PBS for 20 min at cyclophilin, Serca3, insulin, Pdx1, Id2, Bcl-xL, Pax4, Pax6, room temperature. Immunochemical detection of insulin was Stat1 and RpS29 were designed using the Primer Express performed as described previously (Ishihara et al., 2003). 0 Software (Applera Europe, Rotkreuz, Switzerland) and Nuclei were stained with 4 ,6-diamino-2-phenylindole (DAPI) sequences are shown in Table 1. QT-PCR was performed (10 mg/ml; Sigma). Coverslips were mounted using DAKO using an ABI 7000 Sequence Detection System (Applera fluorescent mounting medium and visualized using a Zeiss Europe) and PCR products were quantified using the SYBR Axiophot I. Green Core Reagent kit (Gauthier et al., 2004). Three distinct amplifications derived from 4–6 independent experiments were Cell proliferation and TUNEL assays performed in duplicate for each transcript and mean values For proliferation, transfected and/or infected INS-1E cells were normalized to the mean value of the reference mRNA were labelled with 10 mM BrdU for 4, 16 or 72 h. Proliferation cyclophilin or alternatively to the RpS29 transcript. Authen- was estimated using an immunohistochemical assay kit as ticity of each amplicon was verified by DNA sequencing. described by the manufacturer (BrdU labelling and detection

Oncogene Pax4 and apoptosis T Brun et al 4270 Table 1 Primers used for QT-PCR analysis cells) over the total amount of INS-1E cells (nuclei staining by DAPI). Primer Sequence name Statistical analysis Bcl-xL S: 50-GGA TGC GCG GGA GGT AA-30 Results are expressed as mean7s.e.m. Where indicated, the A: 50-TCA TCG CCA GCC TCT CTC A-30 statistical signiﬁcance of the differences between groups Cyclophilin S: 50-TCA CCA TCT CCG ACT GTG GA-30 was estimated by Student’s unpaired t-test. * and ** indicate 0 0 A: 5 -AAA TGC CCG CAA GTC AAA GA-3 statistical signiﬁcance with Po0.05 and Po0.01, respectively. Id2 S: 50-CCG ACT GTA GAA AGG GCA TTG-30 A: 50-GAT CAT CTC CCC CAG GTG TTC-30 Pax6 S: 50-CCA GCT TCA CCA TGG CAA A-30 A: 50-GCA GGA GTA CGA GGA GGT CTG A-30 Abbreviations Pax4 S: 50-TGG ACA CCC GAC AGC AGA T-30 0 0 A: 5 -CTT AAG GCT CCG TGA GAT GTC A-3 0 0 0 BrdU, 5-bromo-2 -deoxy-uridine; BSA, bovine serum albumin; Pdx1 S: 5 -CCG CGT TCA TCT CCC TTT C-3 0 A: 50-CTC CTG CCC ACT GGC TTT T-30 DAPI, 4 , 6-diamidino-2-phenylindole; FCS, fetal calf serum; RpS29 S: 50-GCT GAAA CAT GTG CCG ACA GT-30 shRNA, short hairpin RNA; PBS, phosphate-buffered saline; A: 50-GGT CGC TTA GTC CAA CTT AAT GAA G-30 TUNEL, terminal deoxynucleotidyl transferase (TdT)- Serca3 S: 50-ACC CTG TTG CTC CTT TG-30 mediated dUTP nick end labelling. A: 50-TCA CAC TGA CAG GCG CTT TC-30 0 0 Stat1 S: 5 -CCT GCT TTG CCT CTG GAA TG-3 Acknowledgements A: 50-CCT TGA GCA GAG CAC GTT CTC-30 We are grateful to Elodie Husi, Mathurin Baquie´ and Delphine Chesnel for their expert technical assistance. Kit, Roche Diagnostics, Rotkreuz, Switzerland). Cell death This work was supported by grants from the Swiss National was measured by the TUNEL assay (In Situ Cell Death Science Foundation (#3100A0-107682/1 to BRG) and from Detection Kit, Roche). Results are expressed as a percentage the Juvenile Diabetes Research Foundation (#7-2005-1158 of BrdU or TMR-red-labelled nuclei (TUNEL-positive to CBW).

References

Asfari M, Janjic D, Meda P, Li G, Halban PA, Wollheim CB. lineages in the mouse endocrine pancreas. Development 132: (1992). Establishment of 2-mercaptoethanol-dependent dif- 2969–2980. ferentiated insulin-secreting cell lines. Endocrinology 130: Collombat P, Mansouri A, Hecksher-Sorensen J, Serup P, 167–178. Krull J, Gradwohl G et al. (2003). Opposing actions of Arx Barker CJ, Leibiger IB, Leibiger B, Berggren PO. (2002). and Pax4 in endocrine pancreas development. Genes Dev 17: Phosphorylated inositol compounds in beta-cell stimulus– 2591–2603. response coupling. Am J Physiol Endocrinol Metab 283: Dor Y, Brown J, Martinez OI, Melton DA. (2004). Adult E1113–E1122. pancreatic beta-cells are formed by self-duplication rather Bernasconi M, Remppis A, Fredericks WJ, Rauscher III FJ, than stem-cell differentiation. Nature 429: 41–46. Schafer BW. (1996). Induction of apoptosis in rhabdomyo- Feng XH, Derynck R. (2005). Speciﬁcity and versatility in sarcoma cells through down-regulation of PAX proteins. tgf-beta signaling through Smads. Annu Rev Cell Dev Biol Proc Natl Acad Sci USA 93: 13164–13169. 21: 659–693. Biason-Lauber A, Boehm B, Lang-Muritano M, Gauthier BR, Gauthier B, Robb M, McPherson R. (1999). Cholesteryl ester Brun T, Wollheim CB et al. (2005). Association of childhood transfer protein gene expression during differentiation of diabetes mellitus with a genomic variant of Pax4: possible human preadipocytes to adipocytes in primary culture. link to beta cell regenerative capacity. Diabetologia 48: Atherosclerosis 142: 301–307. 900–905. Gauthier BR, Brun T, Sarret EJ, Ishihara H, Schaad O, Blasco MA. (2005). Telomeres and human disease: ageing, Descombes P et al. (2004). Oligonucleotide microarray cancer and beyond. Nat Rev Genet 6: 611–622. analysis reveals PDX1 as an essential regulator of mitochon- Brun T, Duhamel D, Sarret EJ, Bosco D, Wollheim CB, drial metabolism in rat islets. J Biol Chem 279: 31121–31130. Gauthier BR. (2005). The diabetes-linked transcription Gunton JE, Kulkarni RN, Yim S, Okada T, Hawthorne WJ, factor Pax4 is expressed in human pancreatic islets and is Tseng YH et al. (2005). Loss of ARNT/HIF1beta mediates activated by glucose activin A and betacellulin. Diabetologia altered gene expression and pancreatic-islet dysfunction in 48(Suppl 1): A66. human type 2 diabetes. Cell 122: 337–349. Brun T, Franklin I, St-Onge L, Biason-Lauber A, Schoenle E, Hanahan D, Weinberg RA. (2000). The hallmarks of cancer. Wollheim CB et al. (2004). The diabetes-linked transcription Cell 100: 57–70. factor Pax4 promotes beta-cell proliferation and survival in He SJ, Stevens G, Braithwaite AW, Eccles MR. (2005). rat and human islets. J Cell Biol 167: 1123–1135. Transfection of melanoma cells with antisense PAX3 Buteau J, Foisy S, Joly E, Prentki M. (2003). Glucagon-like oligonucleotides additively complements cisplatin-induced peptide 1 induces pancreatic beta-cell proliferation via cytotoxicity. Mol Cancer Ther 4: 996–1003. transactivation of the epidermal growth factor receptor. Henderson CC, Zhang Z, Manson SR, Riehm JJ, Kataoka M, Diabetes 52: 124–132. Flye MW et al. (2005). A moderate reduction of Bcl-x(L) Collombat P, Hecksher-Sorensen J, Broccoli V, Krull J, Ponte I, expression protects against tumorigenesis; however, it also Mundiger T et al. (2005). The simultaneous loss of Arx and increases susceptibility to tissue injury. Oncogene 24: 7120–7124. Pax4 genes promotes a somatostatin-producing cell fate Heremans Y, Van De Casteele M, in’t Veld P, Gradwohl G, speciﬁcation at the expense of the alpha- and beta-cell Serup P, Madsen O et al. (2002). Recapitulation of

Oncogene Pax4 and apoptosis T Brun et al 4271 embryonic neuroendocrine differentiation in adult human Robson EJ, He SJ, Eccles MR. (2006). A PANorama of PAX pancreatic duct cells expressing neurogenin 3. J Cell Biol genes in cancer and development. Nat Rev Cancer 6: 52–62. 159: 303–312. Shimajiri Y, Sanke T, Furuta H, Hanabusa T, Nakagawa T, Holm P, Rydlander B, Luthman H, Kockum I. (2004). Fujitani Y et al. (2001). A missense mutation of Pax4 gene Interaction and association analysis of a type 1 diabetes (R121W) is associated with type 2 diabetes in Japanese. susceptibility locus on chromosome 5q11–q13 and the 7q32 Diabetes 50: 2864–2869. chromosomal region in Scandinavian families. Diabetes 53: Shimajiri Y, Shimabukuro M, Tomoyose T, Yogi H, Komiya I, 1584–1591. Takasu N. (2003). PAX4 mutation (R121W) as a prodiabetic Hugl SR, White MF, Rhodes CJ. (1998). Insulin-like growth variant in Okinawans. Biochem Biophys Res Commun 302: factor I (IGF-I)-stimulated pancreatic beta-cell growth is 342–344. glucose-dependent. Synergistic activation of insulin receptor Shing Y, Christofori G, Hanahan D, Ono Y, Sasada R, substrate-mediated signal transduction pathways by glucose Igarashi K et al. (1993). Betacellulin: a mitogen from and IGF-I in INS-1 cells. J Biol Chem 273: 17771–17779. pancreatic beta cell tumors. Science 259: 1604–1607. Huotari MA, Palgi J, Otonkoski T. (1998). Growth factor- Sledz CA, Holko M, de Veer MJ, Silverman RH, Williams BR. mediated proliferation and differentiation of insulin-produ- (2003). Activation of the interferon system by short- cing INS-1 and RINm5F cells: identiﬁcation of betacellulin interfering RNAs. Nat Cell Biol 5: 834–839. as a novel beta-cell mitogen. Endocrinology 139: 1494–1499. Sosa-Pineda B, Chowdhury K, Torres M, Oliver G, Gruss P. Ishihara H, Maechler P, Gjinovci A, Herrera PL, Wollheim CB. (1997). The Pax4 gene is essential for differentiation of (2003). Islet beta-cell secretion determines glucagon release insulin-producing beta cells in the mammalian pancreas. from neighbouring alpha-cells. Nat Cell Biol 5: 330–335. Nature 386: 399–402. Kanatsuka A, Tokuyama Y, Nozaki O, Matsui K, Egashira T. Theander-Carrillo C, Wiedmer P, Cettour-Rose P, Nogueiras R, (2002). Beta-cell dysfunction in late-onset diabetic subjects Perez-Tilve D, Pﬂuger P et al. (2006). Ghrelin action in the carrying homozygous mutation in transcription factors brain controls adipocyte metabolism. J Clin Invest 116: NeuroD1 and Pax4. Metabolism 51: 1161–1165. 1983–1993. Kojima H, Fujimiya M, Matsumura K, Younan P, Imaeda H, Theis M, Mas C, Doring B, Degen J, Brink C, Caille D et al. Maeda M et al. (2003). NeuroD-betacellulin gene therapy (2004). Replacement by a lacZ reporter gene assigns mouse induces islet neogenesis in the liver and reverses diabetes in connexin36, 45 and 43 to distinct cell types in pancreatic mice. Nat Med 9: 596–603. islets. Exp Cell Res 294: 18–29. Li Y, Nagai H, Ohno T, Ohashi H, Murohara T, Saito H et al. Tokuyama Y, Matsui K, Ishizuka T, Egashira T, Kanatsuka A. (2006). Aberrant DNA demethylation in promoter region (2006). The Arg121Trp variant in PAX4 gene is associated and aberrant expression of mRNA of PAX4 gene in with beta-cell dysfunction in Japanese subjects with type 2 hematologic malignancies. Leuk Res 9: 1547–1553. diabetes mellitus. Metabolism 55: 213–216. Margue CM, Bernasconi M, Barr FG, Schafer BW. (2000). Ueda Y. (2000). Activin A increases Pax4 gene expression in Transcriptional modulation of the anti-apoptotic protein pancreatic beta cell lines. FEBS Lett 480: 101–105. BCL-XL by the paired box transcription factors PAX3 and Wang J, Elghazi L, Parker SE, Kizilocak H, Asano M, Sussel L PAX3/FKHR. Oncogene 19: 2921–2929. et al. (2004). The concerted activities of Pax4 and Nkx2.2 are Mauvais-Jarvis F, Smith SB, Le May C, Leal SM, Gautier JF, essential to initiate pancreatic beta-cell differentiation. Dev Molokhia M et al. (2004). PAX4 gene variations predispose Biol 266: 178–189. to ketosis-prone diabetes. Hum Mol Genet 13: 3151–3159. Yamaoka T, Yano M, Yamada T, Matsushita T, Moritani M, Merglen A, Theander S, Rubi B, Chaffard G, Wollheim CB, Ii S et al. (2000). Diabetes and pancreatic tumours in Maechler P. (2004). Glucose sensitivity and metabolism- transgenic mice expressing PaÂ6. Diabetologia 43: 332–339. secretion coupling studied during two-year continuous Zalzman M, Gupta S, Giri RK, Berkovich I, Sappal BS, culture in INS-1E insulinoma cells. Endocrinology 145: Karnieli O et al. (2003). Reversal of hyperglycemia in mice 667–678. by using human expandable insulin-producing cells differen- Miyamoto T, Kakizawa T, Ichikawa K, Nishio S, Kajikawa S, tiated from fetal liver progenitor cells. Proc Natl Acad Sci Hashizume K. (2001). Expression of dominant negative form USA 100: 2426–2431. of PAX4 in human insulinoma. Biochem Biophys Res Zhang YQ, Mashima H, Kojima I. (2001). Changes in the Commun 282: 34–40. expression of transcription factors in pancreatic AR42J cells Muratovska A, Zhou C, He S, Goodyer P, Eccles MR. (2003). during differentiation into insulin-producing cells. Diabetes Paired-Box genes are frequently expressed in cancer and often 50(Suppl 1): S10–14. required for cancer cell survival. Oncogene 22: 7989–7997. Zhou YH, Tan F, Hess KR, Yung WK. (2003). The expression Ostrom L, Tang MJ, Gruss P, Dressler GR. (2000). Reduced of PAX6, PTEN, vascular endothelial growth factor, and Pax2 gene dosage increases apoptosis and slows the epidermal growth factor receptor in gliomas: relationship to progression of renal cystic disease. Dev Biol 219: 250–258. tumor grade and survival. Clin Cancer Res 9: 3369–3375. Park D, Jia H, Rajakumar V, Chamberlin HM. (2006). Pax2/ Zhou YH, Wu X, Tan F, Shi YX, Glass T, Liu TJ et al. (2005). 5/8 proteins promote cell survival in C. elegans. Development PAX6 suppresses growth of human glioblastoma cells. 133: 4193–4202. J Neurooncol 71: 223–229.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc).

Oncogene