Gadd45b is a pro-survival factor associated with stress ... - Nature

22 downloads 98 Views 439KB Size Report
Sep 24, 2007 - constitute a severe problem in the context of cancer ... It is now widely accepted that most cancer ...... Chemotherapy and the war on cancer.
Oncogene (2008) 27, 1429–1438

& 2008 Nature Publishing Group All rights reserved 0950-9232/08 $30.00 www.nature.com/onc

ORIGINAL ARTICLE

Gadd45b is a pro-survival factor associated with stress-resistant tumors A Engelmann1, D Speidel1, GW Bornkamm2, W Deppert1 and C Stocking1 1

Heinrich-Pette-Institut, Hamburg, Germany and 2GSF-Forschungszentrum fu¨r Umwelt und Gesundheit, Institut fu¨r Klinische Molekularbiologie und Tumorgenetik, Munich, Germany

Tumors that acquire resistance against death stimuli constitute a severe problem in the context of cancer therapy. To determine genetic alterations that favor the development of stress-resistant tumors in vivo, we took advantage of polyclonal tumors generated after retroviral infection of newborn Ek-MYC mice, in which the retroviral integration acts as a mutagen to enhance tumor progression. Tumor cells were cultivated ex vivo and exposed to c-irradiation prior to their transplantation into syngenic recipients, thereby providing a strong selective pressure for pro-survival mutations. Secondary tumors developing from stress-resistant tumor stem cells were analysed for retroviral integration sites to reveal candidate genes whose dysregulation confer survival. In addition to the gene encoding the antiapoptotic Bcl-xL protein, we identified the gadd45b locus to be a novel common integration site in these tumors, leading to enhanced expression. In accord with a thus far undocumented role of Gadd45b in tumorigenesis, we showed that NIH3T3 cells overexpressing Gadd45b form tumors in NOD/SCID mice. Interestingly and differently to other known ‘classical’ antiapoptotic factors, high Gadd45b levels did not protect against MYC-, UV- or c-irradiation-induced apoptosis, but conferred a strong and specific survival advantage to serum withdrawal. Oncogene (2008) 27, 1429–1438; doi:10.1038/sj.onc.1210772; published online 24 September 2007 Keywords: B-cell lymphoma; MYC; p53; Bcl-xL; retroviral insertional mutagenesis; stress-resistance

Introduction Despite considerable progress over the past fifty years in the treatment of human malignancies, intrinsic and acquired resistance to chemotherapeutic agents and radiation remains a severe problem (Bernier et al., 2004; Chabner and Roberts, 2005). Our understanding of mechanisms leading to therapeutic resistance has been expanded considerably during this time, and the crucial role of stress resistance pathways is increasingly Correspondence: Dr C Stocking, Molecular Pathology, HeinrichPette-Institut, PO Box 201652, D-20206 Hamburg, Germany. E-mail: [email protected] Received 27 March 2007; revised 11 July 2007; accepted 6 August 2007; published online 24 September 2007

recognized. It is now widely accepted that most cancer therapies work by inducing the intrinsic apoptotic pathway; however other mechanisms may also be involved, including activation of senescence, necrosis, mitotic catastrophe or autophagic cell death (Johnstone et al., 2002; Okada and Mak, 2004). Altered expression or mutation of genes encoding key proteins involved in any of these processes can provide cancer cells with both an intrinsic survival advantage and inherent resistance to therapeutic-induced stress. To identify novel genes and thus mechanisms by which tumor cells escape stress response, we have developed an approach in which tumor growth is accelerated by retroviral insertional mutagenesis, and subclones that have acquired stress resistance are selected. This approach takes advantage of the fact that retroviral infection of transgenic mice expressing the human MYC oncogene under control of the Igl enhancers, mimicking human Burkitt’s lymphoma (Kovalchuk et al., 2000), leads to the rapid development of polyclonal tumors. Retrovirus infection accelerates tumor progression by integration in the vicinity of oncogenes and/or tumor suppressor genes and their subsequent activation or inactivation, conferring a selective advantage and subsequent outgrowth of the affected cell (Mikkers and Berns, 2003). In the approach developed here, the tumor cells are subsequently subjected to stress factors (ex vivo cultivation and g-radiation) prior to transplantation into recipient mice, thus providing a strong selective pressure for tumor cells with integrations that confer pro-survival mutations. The strength of this approach was demonstrated by identifying the known antiapoptotic factor Bcl-xL, but also identifying Gadd45b as a pro-survival factor in this system. Recent studies using gadd45b knockout mice have suggested an antiapoptotic function for Gadd45b (Gupta et al., 2005). Extending these studies, we provide evidence that normal levels of Gadd45b may be critical to prevent apoptosis, but increased levels of Gadd45b contribute to tumorigenesis and tumor resistance by a mechanism distinct to classical apoptosis. Results Mouse model for the identification of tumor-related pro-survival genes To define genetic alterations that contribute to development of stress-resistant hematopoietic tumors

Pro-survival function of Gadd45b in tumorigenicity A Engelmann et al

1430

in vivo, we extended the method of retroviral insertional mutagenesis in a mouse model for leukemia by adding an additional selection step. In the first step, we could

Oncogene

show that infection of newborn El-MYC mice with Moloney-murine leukemia virus (Mo-MuLV) enhanced tumor progression and decreased the median survival

Pro-survival function of Gadd45b in tumorigenicity A Engelmann et al

1431

time of the mice from 100 to 49 days (Figure 1a). Similarly to uninfected controls, the mice succumbed to an aggressive B-cell (B220 þ IgM þ ) or pro-B-cell (B220 þ IgM) neoplasia, characterized by lymphoma and hepatosplenomegaly, as well as high leukocyte counts in the peripheral blood (mean 114  106 cells ml1, n ¼ 44). The accelerated tumor progression can be attributed to additional oncogenic mutations caused by retrovirus integration. To specifically select for tumor subclones in which mutations imparted a pro-survival advantage to stress stimuli, we explanted the tumors, cultivated the tumor cells ex vivo and exposed them to 7 Gy g-radiation, prior to their transplantation into syngenic recipient animals. Such treatments combine several stress stimuli and eliminate the vast majority of tumor cells by apoptosis as indicated by fragmentation of DNA and the presence of active cleaved caspase-3 (Figures 1b and c). Although irradiation clearly provided the major death stimulus, ex vivo cultivation alone was sufficient to provoke a measurable apoptotic response in some tumor cells within a short time, indicating the sensitivity of primary tumor cells to non-genotoxic environmental changes. In concordance with the massive cell death upon irradiation, tumor-initiating cells giving rise to secondary tumors were detected at a frequency four orders of magnitude lower than that of non-irradiated controls, as determined by serial dilution (Figure 1d). Southern blot analysis performed with a Mo-MuLV env probe confirmed that the stress treatment induced a strong selection for specific clones; the viral integration pattern changed significantly from a predominantly diffuse picture in the parental primary tumor (P), indicative of a highly polyclonal tumor, to a more distinct band pattern in the different, primarily monoclonal, secondary tumors (S1–S9) (Figure 1e). Of note, some bands not prominently appearing in the parental tumor could be detected in several of the different independent secondary tumors, with specific differences between the sets of non-irradiated and irradiated secondary tumors. This finding confirms that both treatments, ex vivo cultivation alone and ex vivo cultivation plus g-irradiation, provide a distinct selection pressure, selecting for different tumor cell clones. A second round of radiation and re-transplantation of both irradiated and non-irradiated secondary tumors was performed to verify the selection for stress-resistant tumors (Figure 2a). Although irradiation still delayed the in vivo growth of tumors cells selected for stress

Figure 2 Secondary tumors arising from irradiated tumors have acquired mutations conferring radiation resistance. (a) Schematic representation of experimental design to test if tumors selected after radiation are less sensitive to stress signals. (b) Survival curves of mice transplanted with 5  105 secondary tumor cells that were either irradiated (lightning bolt and solid lines) or untreated (dashed line). Two sets of secondary tumors were tested: tumors generated from irradiated primary tumors (gray lines) or nonirradiated primary tumors (black lines). Results depicted are a compilation of five experiments with independent primary tumors, for which three to four mice were used for each experimental condition (total of 68 mice).

resistance (secondary tumors derived from irradiated primary cells), tumor growth was significantly faster and with a higher penetrance (95 versus 50%) than that of irradiated tumor cells derived from secondary tumors from non-irradiated primary cells (Figure 2b). These results demonstrate that tumor cells had been selected that were significantly less sensitive to irradiation treatment. To determine if spontaneous mutations within the Trp53 gene were responsible for the acquired stress resistance, sequence analysis was performed. Only 25% of the tumors carried point mutations in the p53 coding region, in agreement with the low incidence of p53 inactivation in Mo-MuLV-induced murine and

Figure 1 Mouse model to identify pro-survival genes. (a) Survival curves of Moloney-murine leukemia virus (Mo-MuLV) infected (n ¼ 65) and uninfected (n ¼ 47) El-MYC C57Bl/6 mice. Newborn mice infected with Mo-MuLV show an acceleration of disease as compared to uninfected control animals. (b and c) Apoptosis induction in tumor cells upon ex vivo cultivation and irradiation. (b) DNA fragmentation indicative of apoptosis-related nucleases can be observed in all tumors 3 h after irradiation with 7 Gy and to a lesser extent in non-irradiated tumors. (c) Western blot analysis of protein isolated from explanted tumor cells confirmed the presence of cleaved caspase-3, a marker of apoptosis, in tumors 3 h after irradiation and, to a lesser extent, in one non-irradiated tumor. (d) Survival curves of mice transplanted with limiting dilutions of either non-irradiated tumor cells (black lines) or 5  106 g-irradiated tumor cells (gray line). The kinetics of tumor induction allowed the calculation of cells in the tumor mass that were able to form tumors after irradiation. Shown are the results of one tumor, but similar results were obtained with independent tumors (n ¼ 3). (e) Detection of proviral integrations by Southern blot analysis of genomic DNA digested with BglII and hybridized with an env-probe. The arrow marks an endogenous retroviral fragment, also observed in DNA from uninfected lymph nodes. Lightning bolt indicates tumors arising from irradiated cells. P, parental tumor; S1–S9, secondary tumors. Oncogene

Pro-survival function of Gadd45b in tumorigenicity A Engelmann et al

1432 Table 1 Common integration sites found in secondary tumors Locus

Protein

Bmi1

Polycomb complex protein

Ahi1 Pim2 Bcl2l1 Mcl1 Runx3 Gadd45b

Abelson helper integration site 1 Serine/threonine kinase Apoptosis regulator Bcl-xL Apoptosis regulator EAT/MCL-1 Runt transcription factor 3 Growth arrest and DNA damage-inducible protein b

Chr

Hitsa

Mycb

2

4

+

10 X 2 3 4 10

3 3 3 2 2 2

+ +  + + 

Reference Haupt et al. (1991) and van Lohuizen et al. (1991) Poirier et al. (1988) van der Lugt et al. (1995) This article Mikkers et al. (2002) Stewart et al. (2002) This article

a

Number of independent integrations found in secondary tumors. bPreviously identified as a CIS in MYC-associated tumors.

human lymphopathic tumors (Baxter et al., 1996; Krug et al., 2002). Integrations into the Bcl2l1 locus as a proof of principle To identify genetic alterations responsible for a prosurvival advantage in stress-resistant tumor cells, we analysed secondary tumors derived from ex vivo cultivated and irradiated primary tumors for viral integration sites. In total, 76 viral integration sites of 21 secondary tumors from 10 independent tumor sets were determined, and their specific location in the cellular genome was mapped (Supplementary data). Proviral insertions occurring in several, independently derived tumors are classified as common integration sites (CISs) and mark genes contributing to tumorigenicity (and progression), based on the fact that the frequency of finding two independent integrations in the same locus without a selective advantage would be too low to be detected (Mikkers and Berns, 2003). The majority of CISs that we identified in secondary tumors have been described previously in MYC tumor mouse models (Table 1). However, in addition, we found two loci, bcl2l1 and gadd45b, which have not been identified as a CIS in these models, but which were associated with stress-treated secondary transplants in our analysis. Three independent integrations in two tumor sets were observed in the bcl2l1 locus, which codes for Bcl-xL. This is a bona fide antiapoptotic factor, being a member of the Bcl-2 family, which constitutes the master regulator controlling the intrinsic, mitochondria-mediated pathway of apoptosis (Tsujimoto, 2003). Southern blot analysis of DNA isolated from one tumor set showed that integration into the bcl211 locus was present in two of three irradiated secondary tumors, while it was absent in all tumors derived from non-irradiated tumor cells and in the primary tumor (Figure 3a). Although the secondary tumors were derived from the same primary tumor, the integrations were independent events—occurring approximately 1000 bp from another, within the second intron of the bcl211 gene. A third integration was also detected in a secondary tumor arising from an independent primary tumor after irradiation. In this case, integration had occurred 300 bp upstream of the first coding exon. Quantitative reverse transcriptase (RT)–PCR confirmed Oncogene

Figure 3 Integrations within the bcl2l1 locus are selectively found in secondary tumors derived from irradiated tumor cells and lead to upregulation of Bcl-xL expression. (a) Southern blot analysis of BglII digested genomic DNA. Proviral integrations were detected with a bcl2l1 probe. C, control DNA from uninfected lymph node; P, parental tumor; S1–S6, secondary tumors. Irradiation is indicated by lightning bolt. Asterisks denote rearranged bands in the tumors S4 and S5. (b) Quantitative reverse transcriptase (RT)– PCR of bcl211 mRNA levels relative to hprt controls isolated from the indicated tumors.

that retroviral insertion into the bcl2l1 locus resulted in ca. 10-fold overexpression of Bcl-xL mRNA as compared to parental tumors (compare tumor P with S3 and S4 in Figure 3b). However, upregulation was also observed in secondary tumors from the same set, in which no bcl211 integrations were detected (tumor S6). This may be due to integration in a region not detected by the probe, integration and dysregulation of a gene whose product regulates Bcl-xL expression, or to an unknown mechanism. Taken together, we found the Bcl2l1 locus to be a novel CIS characteristic of radiation-resistant tumors, leading to elevated levels of Bcl211 expression.

Pro-survival function of Gadd45b in tumorigenicity A Engelmann et al

1433

upregulation (see Discussion). Taken together, we found retroviral integrations into the gadd45b locus in two independent sets of tumors thereby clearly identifying this gene as a CIS. Significantly, integration and the resulting upregulation of Gadd45b expression were exclusively observed in secondary tumors, again indicative of a selection advantage that this integration confers to tumor cells that have been stressed by ex vivo cultivation with or without subsequent g-irradiation.

Figure 4 Integrations within the gadd45b locus are found in secondary tumors and lead to upregulation of gadd45b expression. (a and b) Southern blot analysis of HindIII (a) or BglII (b) digested genomic DNA hybridized to a locus-specific probe. C, control DNA from uninfected lymph node; P1 and P2, parental tumors; S1–S21, secondary tumors. Lightning bolts indicate tumors arising from irradiated cells. Arrows denote rearranged bands. (c) Northern blot analysis of RNA isolated from tumors and hybridized with a gadd45b probe.

The gadd45b locus is a novel CIS From integrations in two independent tumor sets, we identified the gadd45b gene locus to be a novel CIS. Gadd45b is a member of the growth arrest and DNA damage-inducible Gadd45 family that is upregulated by various stress stimuli (Liebermann and Hoffman, 2002). In one tumor set, several secondary tumors (four of five) from an irradiated primary tumor contained an integration 1.5 kb downstream of the gadd45b gene, which was not observed in five of seven secondary tumors that had developed from non-irradiated cells (Figure 4a). No integration in this locus was detectable in the parental tumor. In a second independent tumor set, integration was detected 6.4 kb downstream of the gadd45b gene. This integration was detectable in three of four tumors derived from irradiated tumor cells, but not in the parental primary tumor (Figure 4b). Northern blot and quantitative RT–PCR analyses showed that expression in tumors performing the integration was significantly elevated, with the exception of S19 (Figure 4c and data not shown). Also two tumors (S6, S7) without integration in this specific locus showed elevated gadd45b expression, suggesting alternative mechanisms for

Gadd45b is an oncogenic pro-survival factor The finding that tumor cells with elevated Gadd45b levels are enriched upon stress treatment prompted us to ask whether high intracellular concentrations of this factor actively support tumorigenesis. To study the consequence of high Gadd45b expression, we stably expressed Gadd45b in NIH3T3 cells using a retroviral vector co-expressing enhanced green fluorescent protein (eGFP). The transduced cell population was enriched (75–95%) by flow cytometric sorting of GFP positive cells. Expression of Gadd45b and GFP in these cells was verified by western blot analysis (data not shown). Subcutaneous injection of 1  106 Gadd45b expressing NIH3T3 cells reproducibly induced tumors in NOD/ SCID mice (n ¼ 10), which were clearly visible 15 days post injection (mean weight 0.5970.32 g; range 0.25– 1.28 g). At the same time and up to the end of the observation period (21 days post injection), animals that had received parental- (n ¼ 9) or vector-transduced (n ¼ 9) NIH3T3 cells did not show any tumors. Overexpression of Gadd45b thus transforms NIH3T3 mouse fibroblasts and supports tumorigenesis. Moreover, these findings indicate that Gadd45b has an oncogenic activity, which is independent of deregulated MYC expression. We speculated that the pro-tumorigenic potential of Gadd45b is due to inhibition of stress-induced cell death in a way similar to Bcl-xL. To address this, we analysed the impact of Gadd45b overexpression in several systems. DNA damage, as induced by high-dose irradiation, leads to apoptotic cell death by engaging the intrinsic mitochondria-mediated pathway. This can be measured via western blot analysis monitoring the presence of the cleaved and thereby active form of caspase-3 or assessing DNA fragmentation (% sub-G1) by fluorescence-activated cell sorting (FACS) analysis. NIH3T3 fibroblasts exposed to two different doses of UV irradiation show the characteristics of apoptosis to varying degrees (Figures 5a and b). As expected, overexpression of Bcl-xL completely blocked apoptosis. In contrast, Gadd45b overexpressing NIH3T3 cells underwent apoptosis upon UV treatment to the same extent as parental- or vector-transduced cells (Figures 5a and b). Similar results were obtained when murine hematopoietic Ba/F3 cells were subject to 6 Gy girradiation (Figure 5b). We conclude that elevated Gadd45b levels do not protect against DNA damageinduced apoptosis mediated by the intrinsic pathway. Another mechanism to induce cell death that might be relevant for our model system is activation of apoptosis Oncogene

Pro-survival function of Gadd45b in tumorigenicity A Engelmann et al

1434

as a direct consequence of MYC overexpression. Similar to DNA damage, deregulated MYC expression can also induce apoptosis via the intrinsic pathway, but the signaling upstream of the mitochondria is different.

Oncogene

To test whether overexpression of Gadd45b protects against MYC induced cell death, we took advantage of a murine hematopoietic cell line (ERP 15–41) that expresses a fusion protein coupling MYC to the

Pro-survival function of Gadd45b in tumorigenicity A Engelmann et al

1435

estrogen receptor (MYC-ER) (Blyth et al., 2000). Upon addition of 4-hydroxytamoxifen (4OHT), MYC-ER becomes activated and induces apoptosis as evidenced by DNA fragmentation and a high percentage of cells positive for propidium iodide (PI) staining (Figure 5c). Overexpression of Bcl-xL also blocked MYC-induced cell death in these cells. No protective effect upon MYC activation, however, was observed in Gadd45b overexpressing cells, which behaved as the control cells. Thus, elevated Gadd45b levels do not confer a pro-survival effect by means of inhibiting the intrinsic pathway of apoptosis induced by DNA damage or activation of MYC. Since integration in the gadd45b locus and elevated expression of its gene product was also observed in some secondary tumors that were not irradiated (see Figure 4), we reasoned that Gadd45b might help the tumor cells to better survive non-genotoxic stress as a result of the ex vivo cultivation. To mimic this particular type of stress in a cell culture setting, we cultivated the differently transduced NIH3T3 fibroblasts in medium without serum. Serum withdrawal efficiently led to cell death of the vector-transduced control cells as evidenced by PI staining (Figure 6a). Interestingly, overexpression of Bcl-xL provided only an initial, slightly protective effect against this stress stimulus, but could not prevent cell death. In striking contrast, Gadd45b expression prevented cell death under these conditions and maintained cell viability and growth (Figure 6b). As it has been reported that serum deprivation induces classical caspase-dependent apoptosis (Paddenberg et al., 2001), we analysed the presence of cleaved and thus active caspase-3 in these cells (Figure 6c). Cleaved caspase-3 was detectable in all serum-starved NIH3T3 cells, but, consistent with the ability of Bcl-xL but not Gadd45b to block the caspase pathway, levels were significantly reduced in cells expressing Bcl-xL (Figure 6c). However, since only a small proportion of Gadd45b overexpressing cells died upon serum withdrawal, and Bcl-xL overexpressing cells died despite suppression of caspase-3 activation, we conclude that the major death pathway in NIH3T3 cells induced by serum withdrawal differs from known caspase-induced apoptosis. This conclusion is supported by other studies investigating the death pathway initiated by serum withdrawal (Simm et al., 1997; Kues et al., 2002; Leicht et al., 2003). Gadd45b overexpression clearly protects against this yet to be identified death program. In summary, our data show that elevated levels of Gadd45b confer a potent and very specific protection against stress-induced cell death.

Discussion The goal of this study was to define molecular mechanisms by which hematopoietic tumors escape

Figure 6 Overexpression of Gadd45b protects NIH3T3 cells from cell death upon serum withdrawal. Transduced and sorted NIH3T3 cells (same cells as in Figure 5) were cultivated with 10% or no serum. (a) To determine cell viability, cells were harvested at the indicated time, stained with propidium iodide (PI) and analysed by flow cytometry. The percentage of dead cells upon serum withdrawal was calculated as % PI positive cells (0% serum)% PI positive cells (10% serum). Figure shows the result and s.d. of three independent experiments. (b) To determine cell proliferation rates, the number of viable cells was determined by Trypan blue exclusion at the indicated times. Figure shows the result and s.d. of two independent experiments. (c) Western blot analysis of NIH3T3 fibroblasts transduced with the retroviral vectors and subjected to serum withdrawal for 30 h. Proteins were detected using specific antibodies as indicated. The presence of cleaved caspase-3, the active form of the executioner caspase, was used as a marker for apoptosis.

Figure 5 Bcl-xL but not Gadd45b protects against DNA damage and MYC-induced apoptosis. (a) Western blot analysis of NIH3T3 fibroblasts transduced with the retroviral vectors and either mock-treated or subjected to UV irradiation at the indicated dose. Cells were harvested 42 h after irradiation. Proteins were detected using specific antibodies as indicated. The presence of cleaved caspase-3, the active form of the executioner caspase, was used as a marker for apoptosis. (b) Fluorescence-activated cell sorting (FACS) analysis of NIH3T3 cells, subjected to 50 J m2 UV-irradiation, and Ba/F3 cells, subjected to 6 Gy g-irradiation. Cells were fixed in ethanol and stained with propidium iodide (PI) 48 h after treatment. The percentage of sub-G1 fragments is a measure of apoptosis. (c) FACS analysis of transduced and sorted ERP15-41 cells after MYC-ER activation. Transduced ERP15-41 cells were treated with 4-hydroxytamoxifen (4OHT) to activate MYC-ER or with ethanol as a control. Cells were harvested at the indicated time points. To measure apoptosis or viability, ethanol-fixed (left) or non-fixed (right) cells, respectively, were stained with PI. Viability was calculated as % PI positive cells (4OHT treated)% PI positive cells (ethanol treated). Figure shows the result and s.d. of two independent experiments. Oncogene

Pro-survival function of Gadd45b in tumorigenicity A Engelmann et al

1436

stress-induced cell death. We established a powerful mouse model using retroviral insertional mutagenesis to identify genetic lesions that can abrogate cell death in lymphoid cells after induction of stress. One locus identified in this manner was mapped to the Bcl2l1 gene. This locus has been identified as a CIS in other mouse models (Packham et al., 1998; Hwang et al., 2002), but has not been identified as a CIS for MYC mouse models. The finding that Bcl-xL expressing cells are protected against apoptosis induced by irradiation was not surprising. The protein belonging to the Bcl-2 family is a well-known antiapoptotic molecule, which blocks apoptosis at the level of the mitochondria and inhibits the release of apoptogenic molecules from the intermembrane space of the organelle (Tsujimoto, 2003). Our findings are in accord with studies showing that elevated Bcl-xL levels correlate with therapy resistance of tumor cells from hematologic and solid tumors. The identification of the bcl2l1 locus as a CIS in secondary tumors after irradiation thus provides a satisfying ‘proof-ofprinciple’ to our approach. The second CIS identified in this study was mapped to the gadd45b locus. This locus is a new retroviral CIS, since integrations in this locus have, to our knowledge, never been found in any mouse model. Integrations were primarily but not exclusively found in secondary tumors derived from irradiated primary tumor cells, leading to an approximately fivefold increase in expression levels. High expression of gadd45b was also observed in two secondary tumors that did not have integrations within the gadd45b locus. Owing to the extremely high levels of expression in these tumors, the integration sites of these two tumors were also cloned and sequenced. Interestingly, this analysis revealed integration in the tpl2 locus. Tpl2 is an activator of the transcription factor NF-kB (Das et al., 2005), which is a well-known activator of gadd45b gene expression (De Smaele et al., 2001). Thus the induction of gadd45b expression by NF-kB could account for the high expression levels. In summary, there appears to be a high selective advantage for upregulation of gadd45b in the development of tumors after stress, from either in vitro cultivation or irradiation. One explanation for this is that Gadd45b acts as an oncogene during tumor progression, either alone or in combination with existing mutations. An oncogenic function was indeed demonstrated by assaying tumor development of fibroblasts engineered to overexpress Gadd45b in NOD/SCID mice. To determine whether it is also an antiapoptotic protein, the cells were also subjected to different apoptotic-inducing stimuli. Using two different cell lines, we could show that Bcl-xL but not Gadd45b overexpression protects cells from irradiation-induced apoptosis. Those findings were somewhat surprising, because, although Gadd45b was originally described as a pro-apoptotic protein in response to TGFb (Selvakumaran et al., 1994; Yoo et al., 2003), more recent results have revealed antiapoptotic functions for gadd45b. Importantly, however, these studies have only demonstrated an antiapoptotic function for Gadd45b in the extrinsic apoptotic pathway (induced by Oncogene

stimulation with TNFa or Fas ligand) (De Smaele et al., 2001; Zazzeroni et al., 2003) or have inferred an antiapoptotic function by evaluating genotoxic-stressinduced apoptosis in gadd45b/ bone marrow cells (Gupta et al., 2005, 2006). In accordance with our data, other work demonstrated that Gadd45b does not have an influence on apoptosis (Zerbini et al., 2004). Although high levels of Gadd45b were unable to protect cells from radiation or MYC-induced apoptosis, we were able to show that Gadd45b overexpression protected fibroblasts from cell death after serum withdrawal, showing that Gadd45b expression inhibits cell death under certain circumstances. Importantly, Gadd45b interferes with a cell death-inducing pathway that is different from the ‘classical’ caspase-3 driven apoptosis, which can be suppressed by Bcl-xL, indicating that the intrinsic pathway of apoptosis plays only a minor role upon serum withdrawal, and another, yet to be determined signaling cascade is the driving force to eliminate the cells under these conditions. Recent evidence suggests that besides apoptosis and senescence, therapeutic regimens can eliminate tumor cells by alternative pathways (Okada and Mak, 2004). For example, it has been shown that cells protected from apoptosis by Bax and Bak knockout undergo necrosis or autophagic cell death in response to otherwise apoptotic stimuli (Lindsten et al., 2003; Shimizu et al., 2004). It has also been suggested that the antitumor effect of mTOR inhibitors is mediated by their ability to induce autophagy, although it is unknown to what extent the autophagy-inducing effect of the mTOR inhibitors is involved in their antineoplastic activity (Kroemer and Jaattela, 2005). Many synergistic mechanisms may contribute to the elimination of tumor cells in the course of therapy and it is likely that inhibition of any one such mechanism results in escape of cell death. Our work shows for the first time a novel role for Gadd45b in the emergence of stress-resistant clones. The specificity of the Gadd45bmediated effect and its difference to Bcl-xL support the hypothesis that different mechanisms act in concert to protect tumor cells from stress signals. Understanding these alternative cell death pathways will be important to develop alternative therapies to eliminate apoptosisresistant tumors.

Materials and methods Mice, infections and tumor formation Mice carrying the human MYC oncogene under the control of the Igl regulatory elements were crossed with wild-type C57BL/6 mice, and newborns were infected intraperitoneally within 24 h after birth with 0.05 ml Mo-MuLV-containing supernatant with a virus titer of 1.6  106 infectivity units per milliliter. Non-transgenic littermates served as control in these experiments. Infected mice were monitored for lymphoma development, and killed when moribund. Cells dispersed from lymph nodes were transplanted (5  106 cells or 5  105 for the first and second transplantations, respectively, or after serial dilutions in RPMI1640) intraperitoneally into syngenic C57BL/6 recipients. To determine the tumorigenic potential

Pro-survival function of Gadd45b in tumorigenicity A Engelmann et al

1437 of Gadd45b, 1  106 transduced NIH3T3 cells in 100 ml PBS were injected subcutaneously into NOD/SCID mice. Southern blot analysis of genomic DNA and isolation of integration sites High-molecular-weight genomic DNA was extracted from frozen tissue samples and subjected to Southern blot analysis by standard methods. For the analysis of apoptosis, highmolecular-weight DNA was extracted from 1  107 tumor cells cultured for 3 h, as described (Heinrichs and Deppert, 2003). Retroviral integration sites were isolated using a modified protocol of a previously described method (Schmidt et al., 2001). Details are available upon request. Integrations were mapped to the murine genome using the BLAT and BLAST programs of the UCSC (http://genome.ucsc.edu/) and Ensembl (http://www.ensembl.org/) databases, respectively, and compared to known CISs in the RTCGD databank (http:// rtcgd.abcc.ncifcrf.gov/mm7/index.html). RNA analysis Total RNA was isolated from fresh or frozen tissue lysed in guanidine isothiocyanate and pelleted through a CsC1 cushion using the standard protocols. Northern blot analysis was performed using standard methods. To quantify expression levels, cDNA was produced with AMV reverse transcriptase (Promega, Mannheim, Germany) and subjected to real-time PCR using the LightCycler FastStart DNA Master SYBR Green I Kit (Roche, Penzberg, Germany). Conditions and primers are available upon request. Relative quantification of real-time PCR products was performed using the calibratornormalized method. Two independent experiments were performed in duplicates for each sample. Cell culture and assays Cell lines were cultured under standard conditions in either DMEM or RPMI1640 media supplemented with 10% FCS and 4 mM glutamine, or in the case of Ba/F3 cells with interleukin-3. Induction of MYC was achieved by adding 250 nM 4-OHT (Sigma). Cells were subjected to UV irradiation in a UV-crosslinker (1800, Stratagene, La Jolla, CA, USA). g-irradiation was performed using a Cs137 source. Viable cells were determined by PI exclusion and counted by FACS analysis. To determine sub-G1 cell fraction, cells were fixed in 80% ethanol (20 1C), stained with PI and analysed by FACS as described (Speidel et al., 2006).

Production of retroviral supernatant and infection of cells The coding regions of bcl2l1 and gadd45b were amplified via PCR (primers and conditions available upon request) and cloned into an FMEV-based retroviral vector (Schwieger et al., 2002) to generate pSF91-Bcl-xL (R1116) and pSF91-Gadd45b (R1117). Retroviral pseudo-types were produced as described (Schwieger et al., 2002). Transduced cells were sorted for GFP-expressing cells with a FACS Aria (BD Biosciences, Heidelberg, Germany). Western blot analysis Cells were washed with ice-cold PBS, harvested and stored at 70 1C. Cells were lysed in 50 mM HEPES, pH 7.5, 150 mM NaCl, 0.1% NP40, supplemented with a proteinase inhibitor cocktail (complete mini, Roche). Sixty micrograms of lysate was separated on 13% SDS–PAGE gels. For western blotting and ECL we used standard protocols. Antibodies against the following antigens were used, that is, cleaved caspase-3 (9661; Cell Signalling, Danvers, MA, USA), tubulin (CP06; Oncogene/Calbiochem, San Diego, CA, USA), Bcl-xL (sc8392; Santa Cruz Biotechnology, Santa Cruz, CA, USA) and Gadd45b (BC100; Novus Biologicals, Littleton, CO, USA).

Abbreviations 4OHT, 4-hydroxytamoxifen; CIS, common integration site; eGFP, enhanced green fluorescent protein; FACS, fluorescence-activated cell sorting; Mo-MuLV, Moloney-murine leukemia virus; PI, propidium iodide; RT, reverse transcriptase. Acknowledgements We thank Karen Blyth and Ewan Cameron for conditional MYC overexpressing cell lines, Arne Du¨sedau for providing FACS sorter support and other members of the Stocking laboratory for helpful discussions. This work was part of the doctoral thesis of A Engelmann, Department of Biology, University of Hamburg, Hamburg, Germany and was supported by a grant of the Deutsche Krebshilfe. The Heinrich-Pette-Institut is supported by the Freie und Hansestadt Hamburg and the German Ministry of Health and Social Safety.

References Baxter E, Blyth K, Donehower L, Cameron E, Onions D, Neil J. (1996). Moloney murine leukemia virus-induced lymphomas in p53-deficient mice: overlapping pathways in tumor development? J Virol 70: 2095–2100. Bernier J, Hall E, Giaccia A. (2004). Radiation oncology: a century of achievements. Nat Rev Cancer 4: 737–747. Blyth K, Stewart M, Bell M, James C, Evan G, Neil J et al. (2000). Sensitivity to myc-induced apoptosis is retained in spontaneous and transplanted lymphomas of CD2-mycER mice. Oncogene 19: 773–782. Chabner BA, Roberts T. (2005). Chemotherapy and the war on cancer. Nat Rev Cancer 5: 65–71. Das S, Cho J, Lambertz I, Kelliher MA, Eliopoulos AG, Du K et al. (2005). Tpl2/cot signals activate ERK, JNK, and NF-kappaB in a cell-type and stimulus-specific manner. J Biol Chem 280: 23748–23757. De Smaele E, Zazzeroni F, Papa S, Nguyen DU, Jin R, Jones J et al. (2001). Induction of gadd45beta by NF-kappaB downregulates pro-apoptotic JNK signalling. Nature 414: 308–313.

Gupta M, Gupta S, Hoffman B, Liebermann DA. (2006). GADD45A and GADD45B protect hematopoietic cells from UV induced apoptosis via distinct signaling pathways including P38 activation and JNK inhibition. J Biol Chem 281: 17552–17558. Gupta M, Gupta SK, Balliet AG, Hollander MC, Fornace AJ, Hoffman B et al. (2005). Hematopoietic cells from Gadd45a- and Gadd45b-deficient mice are sensitized to genotoxic-stress-induced apoptosis. Oncogene 24: 7170–7179. Haupt Y, Alexander WS, Barri G, Klinken SP, Adams JM. (1991). Novel zinc finger gene implicated as myc collaborator by retrovirally accelerated lymphomagenesis in E mu-myc transgenic mice. Cell 65: 753–763. Heinrichs S, Deppert W. (2003). Apoptosis or growth arrest: modulation of the cellular response to p53 by proliferative signals. Oncogene 22: 555–571. Hwang HC, Martins CP, Bronkhorst Y, Randel E, Berns A, Fero M et al. (2002). Identification of oncogenes collaborating with p27Kip1 loss by insertional mutagenesis and high-throughput insertion site analysis. Proc Natl Acad Sci USA 99: 11293–11298. Oncogene

Pro-survival function of Gadd45b in tumorigenicity A Engelmann et al

1438 Johnstone RW, Ruefli AA, Lowe SW. (2002). Apoptosis: a link between cancer genetics and chemotherapy. Cell 108: 153–164. Kovalchuk AL, Qi CF, Torrey TA, Taddesse-Heath L, Feigenbaum L, Park SS et al. (2000). Burkitt lymphoma in the mouse. J Exp Med 192: 1183–1190. Kroemer G, Jaattela M. (2005). Lysosomes and autophagy in cell death control. Nat Rev Cancer 5: 886–897. Krug U, Ganser A, Koeffler H. (2002). Tumor suppressor genes in normal and malignant hematopoiesis. Oncogene 21: 3475–3495. Kues WA, Carnwath JW, Paul D, Niemann H. (2002). Cell cycle synchronization of porcine fetal fibroblasts by serum deprivation initiates a nonconventional form of apoptosis. Cloning Stem Cells 4: 231–243. Leicht M, Marx G, Karbach D, Gekle M, Kohler T, Zimmer HG. (2003). Mechanism of cell death of rat cardiac fibroblasts induced by serum depletion. Mol Cell Biochem 251: 119–126. Liebermann DA, Hoffman B. (2002). Myeloid differentiation (MyD) primary response genes in hematopoiesis. Oncogene 21: 3391–3402. Lindsten T, Golden JA, Zong WX, Minarcik J, Harris MH, Thompson CB. (2003). The proapoptotic activities of Bax and Bak limit the size of the neural stem cell pool. J Neurosci 23: 11112–11119. Mikkers H, Allen J, Knipscheer P, Romeyn L, Hart A, Vink E et al. (2002). High-throughput retroviral tagging to identify components of specific signaling pathways in cancer. Nat Genet 32: 153–159. Mikkers H, Berns A. (2003). Retroviral insertional mutagenesis: tagging cancer pathways. Adv Cancer Res 88: 53–99. Okada H, Mak TW. (2004). Pathways of apoptotic and non-apoptotic death in tumour cells. Nat Rev Cancer 4: 592–603. Packham G, White EL, Eischen CM, Yang H, Parganas E, Ihle JN et al. (1998). Selective regulation of Bcl-XL by a Jak kinasedependent pathway is bypassed in murine hematopoietic malignancies. Genes Dev 12: 2475–2487. Paddenberg R, Loos S, Schoneberger HJ, Wulf S, Muller A, Iwig M et al. (2001). Serum withdrawal induces a redistribution of intracellular gelsolin towards F-actin in NIH 3T3 fibroblasts preceding apoptotic cell death. Eur J Cell Biol 80: 366–378. Poirier Y, Kozak C, Jolicoeur P. (1988). Identification of a common helper provirus integration site in Abelson murine leukemia virusinduced lymphoma DNA. J Virol 62: 3985–3992. Schmidt M, Hoffmann G, Wissler M, Lemke N, Mussig A, Glimm H et al. (2001). Detection and direct genomic sequencing of multiple rare unknown flanking DNA in highly complex samples. Hum Gene Ther 12: 743–749.

Schwieger M, Lohler J, Friel J, Scheller M, Horak I, Stocking C. (2002). AML1-ETO inhibits maturation of multiple lymphohematopoietic lineages and induces myeloblast transformation in synergy with ICSBP deficiency. J Exp Med 196: 1227–1240. Selvakumaran M, Lin HK, Sjin RT, Reed JC, Liebermann DA, Hoffman B. (1994). The novel primary response gene MyD118 and the proto-oncogenes myb, myc, and bcl-2 modulate transforming growth factor beta 1-induced apoptosis of myeloid leukemia cells. Mol Cell Biol 14: 2352–2360. Shimizu S, Kanaseki T, Mizushima N, Mizuta T, Arakawa-Kobayashi S, Thompson CB et al. (2004). Role of Bcl-2 family proteins in a nonapoptotic programmed cell death dependent on autophagy genes. Nat Cell Biol 6: 1221–1228. Simm A, Bertsch G, Frank H, Zimmermann U, Hoppe J. (1997). Cell death of AKR-2B fibroblasts after serum removal: a process between apoptosis and necrosis. J Cell Sci 110(Part 7): 819–828. Speidel D, Helmbold H, Deppert W. (2006). Dissection of transcriptional and non-transcriptional p53 activities in the response to genotoxic stress. Oncogene 25: 940–953. Stewart M, MacKay N, Cameron ER, Neil JC. (2002). The common retroviral insertion locus Dsi1 maps 30 kilobases upstream of the P1 promoter of the murine Runx3/Cbfa3/Aml2 gene. J Virol 76: 4364–4369. Tsujimoto Y. (2003). Cell death regulation by the Bcl-2 protein family in the mitochondria. J Cell Physiol 195: 158–167. van der Lugt NM, Domen J, Verhoeven E, Linders K, van der Gulden H, Allen J et al. (1995). Proviral tagging in E mu-myc transgenic mice lacking the Pim-1 proto-oncogene leads to compensatory activation of Pim-2. EMBO J 14: 2536–2544. van Lohuizen M, Verbeek S, Scheijen B, Wientjens E, van der Gulden H, Berns A. (1991). Identification of cooperating oncogenes in E mu-myc transgenic mice by provirus tagging. Cell 65: 737–752. Yoo J, Ghiassi M, Jirmanova L, Balliet AG, Hoffman B, Fornace Jr AJ et al. (2003). Transforming growth factor-beta-induced apoptosis is mediated by Smad-dependent expression of GADD45b through p38 activation. J Biol Chem 278: 43001–43007. Zazzeroni F, Papa S, Algeciras-Schimnich A, Alvarez K, Melis T, Bubici C et al. (2003). Gadd45 beta mediates the protective effects of CD40 costimulation against Fas-induced apoptosis. Blood 102: 3270–3279. Zerbini LF, Wang Y, Czibere A, Correa RG, Cho JY, Ijiri K et al. (2004). NF-kappa B-mediated repression of growth arrest- and DNA-damage-inducible proteins 45alpha and gamma is essential for cancer cell survival. Proc Natl Acad Sci USA 101: 13618–13623.

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

Oncogene