Specific Requirement for Bax, Not Bak, in Myc-induced Apoptosis and Tumor Suppression in Vivo*

Bax and Bak comprise the mitochondrial gateway for apoptosis induced by diverse stimuli. Loss of both bax and bak is necessary to block cell death induced by such stimuli, indicating a great degree of functional overlap between Bax and Bak. Apoptosis is the major intrinsic pathway that limits the oncogenic potential of Myc. Using a switchable mouse model of Myc-induced apoptosis in pancreatic β cells, we have shown that Myc induces apoptosis in vivo exclusively through Bax but not Bak. Furthermore, blockade of Myc-induced apoptosis by the inactivation of Bax, but not Bak, eliminates all restraints to the oncogenic potential of Myc, allowing the rapid and synchronous progression of invasive, angiogenic tumors.

lesions that suppress apoptosis and/or enhance proliferative rate over that of apoptosis are sufficient to expose the oncogenic potential of Myc (12)(13)(14)(15). The pivotal role in tumor suppression played by the innate apoptotic activity of Myc is evident in the well described pIns-MycER-TAM model in which a switchable form of the Myc oncoprotein, Myc-ER TAM , is transgenically targeted to pancreatic ␤ cells via the insulin promoter. Acute activation of MycER TAM triggers abrupt entry of ␤ cells into cycle, but this is accompanied by overwhelming apoptosis that rapidly leads to ␤ cell involution and onset of diabetes. In contrast, when Myc-induced apoptosis in ␤ cells is blocked by co-expression of the apoptosis suppressor Bcl-x L , rapid and synchronous tumorigenesis occurs in all islets (14). Not only does this demonstrate that apoptosis is a critical restraint to the otherwise potent and pleiotropic oncogenic potential of Myc, but the dramatic anti-apoptotic influence of Bcl-x L also indicates that Myc-induced apoptosis proceeds exclusively through the mitochondrial Bax/Bak pathway, presumably through the activation of specific BH3 proteins that are antagonized by the anti-apoptotic Bcl-2/Bcl-x L proteins. Consistent with this, in vitro experiments indicate that Bax is important in mediating Myc-induced apoptosis (at least in fibroblasts) and that Bax may contribute to limiting the oncogenic potential of Myc (16). By contrast, the role of Bak in such tumor-suppressive processes remains undefined.
The advantage of the acutely switchable pIns-MycER TAM model is that it allows direct observation of the acute consequences of Myc activation for proliferation and apoptosis in an orthotopic in vivo setting as well as how such acute effects elaborate into tumorigenesis. We have therefore used pIns-MycER TAM mice to assess the relative roles of Bax and Bak in mediating Myc-induced apoptosis and the suppression of Myc-induced tumorigenesis. We show that Myc-induced apoptosis is mediated solely through Bax and not Bak. Moreover, blocking apoptosis by deleting bax unleashes the full potential of Myc to drive the rapid and synchronous formation of angiogenic, invasive ␤ cell tumors. TAM and pIns-MycER TAM ;RIP7-Bcl-x L mice have been described previously (14). bax Ϫ/Ϫ mice were purchased from Jackson Laboratories (Bar Harbor, ME) (stock number 002994), whereas bak Ϫ/Ϫ mice (5, 6) were provided by Dr. Craig Thompson (Department of Molecular Genetics and Microbiology, State University of New York at Stony Brook, NY). Strains were crossed to generate requisite genotypes, which were then verified by PCR analysis on genomic DNA isolated from tail snips. Tamoxifen (1 mg/20 g bodyweight, dissolved in peanut oil) (Sigma) or vehicle control (peanut oil) (Sigma) administration was initiated at 6 -8 weeks of age and repeated for up to 12 consecutive days by daily intraperitoneal injection. For each bax or bak genotype, 12 mice were treated: 3 pIns-MycER TAM and three non-transgenic mice with tamoxifen and 3 pIns-MycER TAM and three non-transgenic mice with oil control. Blood glucose (using Accu-Chek strips, Roche Applied Science) and . The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 To whom correspondence should be addressed. animal weight were monitored before treatment and after 6 and 12 days of tamoxifen administration. All mouse experiments were approved by the University of California at San Francisco Institutional Animal Care and Use Committee.

Manipulation of Transgenic Animals-Both pIns-MycER
Harvesting of Pancreata-Mice were injected with 150 l of 10 mg/ml BrdUrd (Sigma) solution dissolved in saline 3.5 h before sacrifice, euthanized with CO 2 and perfused through the heart with 6 ml of phosphate-buffered saline followed by 6 ml of zinc-buffered formaldehyde (Anatech, Battle Creek, MI). Pancreata were harvested and fixed overnight at 4°C, dehydrated, processed, embedded in paraffin wax, and sectioned.
Histochemistry-Immunohistological detection of insulin, glucagon, PECAM-1 (CD31), as well as hematoxylin and eosin staining, were as described previously (14). BrdUrd incorporation and TUNEL 2 staining were assayed using, respectively, a Roche Applied Science BrdUrd labeling and detection kit and the Chemicon (Temecula, CA) Apoptag peroxidase kit, according to the manufacturers' instructions, with the modification that sections were then counterstained with Gill's hematoxylin.
Confocal Laser Scanning Microscopy-Primary antibodies against insulin, Bax (Cell Signaling, Beverly, MA; 1:50) or Bak (Cell Signaling; 1:200) were applied as described for insulin (14). Anti-guinea pig-Alexa488 or anti-rabbit-Alexa568 (Molecular Probes, Leiden, The Netherlands) were used as secondary antibodies, as appropriate. Sections were embedded in DAKO fluorescent mounting medium containing 1 mg/ml Hoechst 33258 to stain nuclear DNA. Images were taken with a Zeiss LSM Meta confocal microscope.
Quantification of Apoptosis and Proliferation-Mice were treated with a single injection of tamoxifen and sacrificed 32 h later. TUNEL and BrdUrd staining on pancreas sections was done as above. Apoptotic cells and BrdUrd-positive cells were counted in at least 6 islets/mouse, with a minimum total of 500 ␤ cells. Negligible apoptosis was observed in non-␤ cells using a double stain for TUNEL and insulin (data not shown).

RESULTS
Both Bax and Bak Are Expressed in Pancreatic ␤ Cells-To assess the relative influence of Bax and Bak on Myc-induced tumorigenesis in pancreatic ␤ cells in vivo, we first established that both Bax and Bak proteins are expressed in pancreatic ␤ cells. Pancreata were harvested from bax ϩ/ϩ ;bak ϩ/ϩ mice, fixed, paraffin-embedded, sectioned, and stained with antibodies specific for either Bax or Bak proteins. Abundant and comparable staining of both Bax and Bak was observed in ␤ cells of bax ϩ/ϩ ;bak ϩ/ϩ mice (Fig. 1). Bax was expressed at similar levels in both exocrine and endocrine tissue (Fig. 1A), whereas Bak expression was more prominent in the endocrine compartment (Fig. 1B). Analogous staining was performed on pancreas sections from either bax Ϫ/Ϫ or bak Ϫ/Ϫ animals, and this confirmed the specificity of the signal of each antibody (Fig. 1).
Absence of Bax, but Not Bak, Incapacitates Myc-induced ␤ Cell Apoptosis and Allows Rapid ␤ Cell Expansion-MycER TAM is functionally active only in the presence of its activating ligand 4-hydroxytamoxifen (4-OHT). MycER TAM was activated in mice in vivo by intraperitoneal administration of tamoxifen, which is then rapidly converted into 4-OHT by the liver. Acute activation of MycER TAM in pancreatic ␤ cells induces both cell cycle entry but also overwhelming apoptosis leading to net ␤ cell involution. As described previously, sustained activation of MycER TAM for 12 days by daily tamoxifen injection induced complete and universal ablation of islet ␤ cells ( Fig. 2A) (14). To dissect out the respective roles of Bax and Bak in Myc-induced ␤ cell apoptosis, switchable pIns-MycER TAM transgenic mice were crossed into either heterozygous or homozygous bax or bak knock-out backgrounds. MycER TAM was then continuously activated for 12 days, pancreata harvested, and sections stained for insulin to identify ␤ cells in the islets. ␤ cells of animals lacking one or both copies of bak behaved exactly like wild-type ␤ cells, undergoing rapid involution upon activation of Myc. In complete contrast, ␤ cells lacking both copies of bax did not involute and, instead, underwent rapid, synchronous, and progressive expansion, similar to that observed in ␤ cells expressing the apoptosis inhibitor Bcl-x L (14). ␤ cells in animals lacking only one copy of bax behaved like wild types, undergoing rapid involution ( Fig. 2A). Tamoxifen treatment of mice lacking the MycER TAM transgene were unaffected by tamoxifen irrespective of Bax or Bak status, neither involuting nor expanding (Fig. 2B).
To ascertain that ␤ cell expansion in the absence of Bax is specifically because of a blockade of Myc-induced apoptosis, we used TUNEL staining to identify apoptotic cells in pancreas sections from mice in which MycER TAM had been continuously activated for 12 days. In each case, sections were co-stained for glucagon to mark the positions of hypoand hyperplastic islets; glucagon is a marker for insulin-negative islet ␣ cells that do not express MycER TAM and, consequently, do not undergo apoptosis upon tamoxifen treatment. TUNEL-positive cells were evident in the shrunken involuted islets in all tamoxifen-treated pIns-MycER TAM mice in which at least one copy of bax was present (Fig. 3, A Sections were co-stained with Hoechst (nuclei, blue) and anti-insulin antibody (red) to identify ␤ cells in the overlays. In non-transgenic wild-type animals, both Bax (A, upper panels) and Bak (B, upper panels) are clearly detectable in ␤ cells. However, Bax antibody fails to stain any tissues in bax Ϫ/Ϫ mice (A, lower panels), whereas Bak antibody fails to stain tissue from bak Ϫ/Ϫ animals (B, lower panels). Note that erythrocyte autofluorescence is in the green channel but can be distinguished from signals in other cells by lack of a Hoechst-staining nucleus. and B). In such animals, islet remnants consisted predominantly of glucagon-positive ␣ cells. In normal islets, glucagon-expressing cells represent no more than 10% of the cells in each islet, with the bulk of the remaining cells being ␤ cells. Thus, the great preponderance of ␣ cells in the remnants of involuted islets attests to the extensive apoptotic attrition that Myc induces within the ␤ cell compartment. In contrast, islets from homozygous bax-deficient animals were all markedly hyperplastic with no discernible TUNEL-positive staining. Moreover, such hyperplastic bax Ϫ/Ϫ islets exhibited a vast excess of ␤ cells over ␣ cells, indicative of selective ␤ cell hyperplasia (Fig. 3B). Of note, islet remnants were never observed in any of the bax Ϫ/Ϫ mice, indicating that none of the islets had involuted and, consequently, that the hyperplastic islets do not represent a select subset of ␤ cells that expand after Myc activation. Hetero-or homozygous loss of bak exerted no detectable suppressive influence on Myc-induced apoptosis, consistent with the notion that Bak plays no significant role in Myc-dependent apoptosis. No TUNEL staining was evident in control tamoxifen-treated, non-transgenic pancreatic ␤ cells of any genotype (Fig. 3A) nor in pIns-MycER TAM animals of any bax or bak genotype treated with oil carrier instead of tamoxifen. Islets in such animals remained normal throughout (data not shown).
To exclude the possibility that tamoxifen treatment of bax Ϫ/Ϫ pIns-MycER TAM mice somehow leads secondarily to loss of Bak in the islets, we examined Bax and Bak expression after 12 days of MycER TAM acti-vation. ␤ cells in hyperplastic islets of bax Ϫ/Ϫ pIns-MycER TAM mice maintained Bak expression (compare Figs. 4A and 1B), confirming that ␤ cells in bax Ϫ/Ϫ pIns-MycER TAM islets fail to undergo apoptosis upon MycER TAM activation, despite the continued presence of Bak. Of note, the few ␤ cells that remained in involuted bak Ϫ/Ϫ pIns-MycER TAM islets showed clear evidence of Bax expression (Fig. 4B).
It nonetheless remained possible that Myc-induced apoptosis, although not completely blocked, was reduced in bak-deficient ␤ cells, but that such partial protection was overwhelmed by 12 days of sustained Myc activation. We therefore also examined the extent of ␤ cell apoptosis in mice in which MycER TAM had been activated for only 32 h. Once again, we discerned no suppression of apoptosis in animals lacking one or both copies of bak. By contrast, homozygous bax-deficient ␤ cells exhibited complete resistance to Myc-induced apoptosis. Again, even a single copy of bax was enough to confer complete sensitivity to Myc-induced apoptosis in the absence of bak (Fig. 5). Proliferative rates upon activation of c-MycER TAM were not affected by Bax and Bak status, and the percentage of BrdUrd-positive cells was ϳ4% in all cases where the pIns-MycER TAM transgene product had been activated (data not shown).
tained activation of MycER TAM in ␤ cells co-expressing Bcl-x L leads to synchronous and progressive ␤ cell expansion that is accompanied by lock-step co-expansion of the adjacent endothelial compartment (angiogenesis) together with local invasion of surrounding exocrine pancreatic tissue (14). Concomitantly, the affected ␤ cells undergo partial loss of differentiation, down-regulate surface E-cadherin, and lose the intimate cell-cell contacts that characterize the normal islet. The rapidity and synchrony with which such neoplastic attributes arise indicate that these disparate pleiotropic changes are all instructed by the action of Myc, consistent with the normal physiological role of Myc as a coordinator of cell and tissue expansion and reorganization. Importantly, however, we previously demonstrated that these diverse oncogenic actions of Myc are manifest only when apoptosis is suppressed by co-expression of Bcl-x L , which permits Myc-expressing cells to survive FIGURE 3. Apoptosis and glucagon staining of ␤ cells after 12 days of sustained MycER TAM activation. Paraffin-embedded sections of pancreata from pIns-MycER TAM mice treated for 12 days with tamoxifen were stained for glucagon to identify ␣ cells, whereas TUNEL labeling was used to identify apoptotic ␤ cells.
Dashed lines indicate the margins of islets or islet remnants. A, after 12 days of tamoxifen administration, islets appear normal in non-transgenic bax;bak wild-type mice, as indicated by the small relative proportion of ␣ cells, which are largely confined to the islet periphery, and the lack of TUNEL-positive apoptotic cells. B, in pIns-MycER-TAM transgenic bax ϩ/ϩ ;bak ϩ/ϩ mice, however, only remnants of the islet remain, mostly comprising ␣ cells, and most of the few remaining ␤ cells are TUNEL-positive. Essentially the same is seen in islets from bax ϩ/ϩ ;bak ϩ/Ϫ pIns-MycER TAM , bax ϩ/ϩ ; bak Ϫ/Ϫ pIns-MycER TAM , bax ϩ/Ϫ ;bak ϩ/ϩ pIns-MycER TAM , and bax ϩ/Ϫ ;bak Ϫ/Ϫ pIns-MycER TAM mice. In contrast, bax Ϫ/Ϫ ;bak ϩ/ϩ pIns-MycER TAM islets are hyperplastic, have only a low proportion of glucagon-positive cells, and TUNEL-positive cells are largely absent.  Bax (green, B) antibodies, exactly as for the untreated mice in Fig. 1. A, Bak expression is still clearly evident in hyperplastic islets from bax Ϫ/Ϫ ;bak ϩ/ϩ pIns-MycER TAM animals, indicating that neither the expression nor stability of Bak is dependent upon the presence of Bax. B, Bax expression is also maintained in the few ␤ cells remaining in involuted bax ϩ/ϩ ;bak Ϫ/Ϫ pIns-MycER TAM islets after 12 days of MycER TAM activation.

FIGURE 5. Early stage Myc-induced apoptosis is not suppressed by the absence of Bak.
To assess the influence of bax and bak status on ␤ cell apoptosis at an early stage, prior to complete ␤ cell ablation, the incidence of TUNEL-positive ␤ cells was determined after only 32 h of MycER TAM activation when ␤ cell involution was incomplete. Shown are TUNEL staining of representative islets from bax ϩ/ϩ ;bak ϩ/ϩ pIns-MycER TAM (A) and bax Ϫ/Ϫ ;bak ϩ/ϩ pIns-MycER TAM (B) mice. No TUNEL stain is evident in islets from non-transgenic animals (see also Fig. 3A) or in oil-treated transgenic pIns-MycER TAM controls (data not shown). C, quantitation of the percentage of apoptotic ␤ cells per total islet in various genetic backgrounds shows that only loss of both bax copies blocks Myc-induced apoptosis. In all other genotypes, the extent of TUNEL-positive cells was ϳ7.5%. Error bars indicate standard deviations. Proliferative rates, as measured by BrdUrd incorporation, were not affected by Bax or Bak status (data not shown). and thereby exert their neoplastic influence on neighboring cells and stroma. To determine whether the absence of Bax, but not Bak, confers a similar capacity on Myc to drive the formation of complex tumors, we examined the morphology, architecture, and dynamics of ␤ cell hyperplasias arising from the activation of MycER TAM in ␤ cells of bax-deficient mice and compared it to that in equivalently treated pIns-MycER TAM mice co-expressing Bcl-xL in their ␤ cells (pIns-MycER TAM ;RIP7-Bcl-x L ).
Hematoxylin and eosin staining of hyperplastic islets in bax Ϫ/Ϫ mice in which MycER TAM had been activated for 12 days showed ␤ cells to have enlarged nuclei containing dark chromatin, exactly as in analogously treated pIns-MycER TAM ;RIP7-Bcl-x L animals (Fig. 6A, right panels). Multiple mitotic figures could be found in these islets. Systemic BrdUrd labeling (3.5 h intraperitoneal bolus prior to sacrifice) indicated that ␤ cells remain in cycle after 12 days sustained Myc activity, exactly as in pIns-MycER TAM ;RIP7-Bcl-xL mice. Indeed, even the proportions of BrdUrd-positive cells are similar in islets from each of these two backgrounds. In contrast, no BrdUrd-positive cells were visible in islets of pIns-MycER TAM transgenic mice in which MycER TAM had never been activated (Fig. 6A, left panels), irrespective of their Bax or Bcl-x L status. ␤ cells in bax Ϫ/Ϫ ;bak ϩ/ϩ pIns-MycER TAM mice showed dense insulin staining before MycER TAM activation. However, after 12 days of sustained Myc activity, insulin staining appeared patchy, and the mice exhibited acute diabetes, indicative of a lack of circulating insulin. Thus, as with pIns-MycER TAM ;RIP7-Bcl-x L mice, even though ␤ cell numbers progressively increased, the dominant dedifferentiating influence of Myc was sufficient to curtail insulin production. By 12 days of Myc activation, bax Ϫ/Ϫ ;bak ϩ/ϩ pIns-MycER TAM islets also exhibited invasive intrusions into the exocrine tissue, identical to those in pIns-MycER TAM ;RIP7-Bcl-x L mice. Finally, Myc-induced ␤ cell hyperplasias in bax Ϫ/Ϫ ;bak ϩ/ϩ pIns-MycER TAM mice were highly vascularized as compared with non-tumorigenic islets from the same genetic background that were treated with the vehicle control rather than tamoxifen (Fig. 6B). This is consistent with earlier observations in pIns-MycER TAM ; RIP7-Bcl-x L mice that Myc is potently angiogenic. Thus, by multiple independent parameters, Myc-induced tumors arising in bax-deficient animals resemble those in pIns-MycER TAM ;RIP7-Bcl-x L mice in being dedifferentiated, invasive, and angiogenic. We conclude that loss of bax, but not bak, is sufficient not only to suppress Myc-induced apoptosis in vivo but also to unfetter the full oncogenic potential of Myc.

DISCUSSION
Previous in vitro studies have shown that Myc-dependent apoptosis in serum-deprived fibroblasts is dependent upon the expression of Bax but not Bak (16,17). Our data confirm for the first time that the specific relationship between Myc and Bax holds in other cell types in a physiological setting in vivo. Of note, Myc does not induce expression of the bax gene but rather appears to trigger activation of the Bax protein (16,10). Thus, its peculiarly private relationship with Bax implies the existence of some Bax-specific intermediate downstream of Myc, most likely a member of the BH3 family. Recently, we 3 and others (18) have identified Bim as a potential mediator of Myc-dependent activation of the mitochondrial pathway, which is intriguing given the fact that Bim is reported to possess a greater affinity for Bax versus Bak (19). Nonetheless, studies show that Bim can still interact with Bak, and consistent with this, we showed earlier that Bak could restore Myc-induced apoptosis in bax-deficient fibroblasts if ectopically overexpressed at sufficiently elevated levels (16). Taken together, our data suggest that the private relationship between Myc and Bax so evident in physiological circumstances is, at the molecular level, relative rather than absolute. Indeed, bax;bim (but not bak;bim) double knock-out mice have impaired developmental apoptosis in the interdigital space (20), similar to that observed in the bax;bak double knock-out mouse, whereas the bax single knock-out does not. This suggests that Bim is capable of efficiently triggering apoptosis through Bak when Bax is absent, at least in this specific developmental instance. Another recent study showed that the pro-apoptotic Bcl-2 family member Bcl-x S selectively induces apoptosis through Bak and not Bax (21). Finally, evidence is mounting that the various anti-apoptotic members of the Bcl-2/Bcl-x L family selectively target different pro-apoptotic members (22,23), indicating that the ability of any given trigger to activate Bax or Bak will be contingent upon the complement of anti-apoptotic proteins expressed in that cell type. In summary, all of these studies strengthen the notion that the roles of Bax and Bak are not fully interchangeable in the mediation of apoptosis.
Our data demonstrate that, by blocking Myc-dependent apoptosis, bax inactivation is sufficient to expose the diverse oncogenic potential of Myc, driving the synchronous progression of angiogenic, invasive ␤ cell carcinomas. This is consistent with the established notion that apoptosis is the principal intrinsic mechanism limiting Myc-induced tumorigenesis (24). Recently, however, it was reported that although loss of Bax inhibits apoptosis in a classical, constitutive Myc-transgenic mouse model for mammary tumorigenesis (25), paradoxically such suppression of apoptosis did not appreciably promote Myc-induced tumorigenesis. This inconsistency may indicate that apoptosis suppresses tumorigenesis to different extents in differing tissues. On the other hand, in classical, constitutive transgenic tumor models, it is extremely difficult to dissect out the Myc-dependent component of apoptosis from that induced by any of the many other insults sustained by established tumors (e.g. DNA damage, hypoxia, and survival factor deprivation), any or all of which may contribute to limiting tumorigenesis. Because most of such triggers can route through either Bax or Bak (3, 6), they would not be blocked appreciably by the inactivation of Bax alone. By contrast, acutely switchable Myc oncogenesis models have the advantage of allowing direct inspection, in real time, of Myc-induced apoptosis in vivo and its influence on tumorigenesis. On the other hand, our model is limited by the fact that tamoxifen must be administered daily by intraperitoneal injection to keep the transgene active. As with all mouse models, we also realize that our model has its limitations, and although the rapid switchability of our model allows us to inspect the roles of Bax and Bak in the prevention of early tumor formation in a unique way, the need for daily tamoxifen injections to keep MycER TAM active precludes examination of long term tumor maintenance. We therefore cannot exclude the possibility that loss of Bak could play a role in the maintenance of islet tumors that form in the bax Ϫ/Ϫ ;bak ϩ/ϩ pIns-MycER TAM mice at a later stage.
It remains to be seen whether Bak, in the absence of Bax, can mediate apoptosis in pancreatic ␤ cells when it is induced by triggers other than Myc. It is known that ␤ cell attrition can be induced in bax Ϫ/Ϫ ;bak ϩ/ϩ mice, for example, by treatment with streptozotocin (26), although it has not been determined whether Bak is required to mediate this death. Ultimately, defining precisely when, and in response to what signals, Bax and Bak differentially contribute to apoptosis in tumors will be important in establishing if and when each can function to limit the emergence of tumors in vivo.