Regulation of apoptosis by XIAP ubiquitin-ligase activity

The ΔRING mutation generates a truncated XIAP protein

We used gene targeting to create a ΔRING truncation by replacing the first two amino acids (C449 and K450) of the mouse XIAP RING domain with stop codons (Fig. 1A). Direct sequencing of the genomic locus verified the presence of stop codons in targeted ES cells (Fig. 1B), and chimeric mice transmitted the mutant allele through the germline (Fig. 1C). The mutant allele encoded a truncated protein of the expected size that was expressed at similar levels as full-length XIAP in ES cells (Fig. 1D). ΔRING mice were fertile and born at expected Mendelian ratios for gender and genotype. The histopathology of mutant mice was indistinguishable from wild-type littermates at different ages, and ΔRING mice have shown no overt susceptibility to disease in our pathogen-free mouse colony after >18 mo of age. These observations are consistent with those seen in XIAP-null mice, which develop mostly normally (Harlin et al. 2001; Olayioye et al. 2005).

The RING is the primary determinant of XIAP stability during thymocyte apoptosis

XIAP and cIAP1 RING domains can serve as E3 ubiquitin-ligases to promote autoubiquitination and IAP degradation during mouse thymocyte apoptosis (Yang et al. 2000). Additionally, ΔRING XIAP mutant protein is resistant to ubiquitination, stably expressed, and imparts resistance to apoptosis when overexpressed in a mouse T-cell hybridoma (Yang et al. 2000). The observation that efficient thymocyte apoptosis requires the proteasome (Grimm et al. 1996) led to a hypothesis that IAP RINGs may act proapoptotically in these cells by engaging the ubiquitin system to relieve a break on cell death pathways by degrading IAPs (Yang and Li 2000). If this hypothesis were correct, thymocytes from the ΔRING mouse should have elevated XIAP protein levels and should be more resistant toward apoptosis (Silke et al. 2005). However, we were unable to detect any impairment of thymocyte apoptosis in several paradigms (Fig. 2). Although ΔRING XIAP protein was more stable than full-length XIAP in cultured thymocytes upon treatment with either dexamethasone or etoposide (Fig. 2A), deletion of the RING did not inhibit apoptosis (Fig. 2C–E). The basal expression of ΔRING XIAP was greater, and it was cleaved over time during apoptosis. In addition, deletion of the RING did not alter the basal expression or down-regulation of cIAP1 during apoptosis (Fig. 2A). XIAP and cIAP1 can bind each other in a RING-dependent manner (Silke et al. 2005), but our results indicate that removing the XIAP RING does not influence the abundance of cIAP1 in thymocytes, similar to what has been reported for XIAP-null thymocytes (Conze et al. 2005). Furthermore, XIAP ΔRING thymocytes still degraded Smac/DIABLO by the ubiquitin system (Fig. 2A; data not shown). Consistent with prior reports (Yang et al. 2000), the proteasome inhibitor MG-132 blocked the decrease of full-length XIAP during apoptosis while the pan-caspase inhibitor zVAD-fmk had negligible effects (Fig. 2B). ΔRING XIAP expression was more sensitive to caspase inhibition, however, as zVAD-fmk fully prevented the decrease in protein levels and abolished the cleavage product, while MG-132 had a lesser effect by comparison (Fig. 2B). This supports the hypothesis that the RING engages the ubiquitin system to promote XIAP turnover in apoptotic thymocytes.

ΔRING XIAP efficiently blocks glucocorticoid-induced apoptosis when expressed in a T-cell hybridoma (Yang et al. 2000). We used ΔRING thymocytes to see if this also occurs in primary cells. We labeled the active site of caspase-3 with a fluorescent FAM-DEVD-FMK affinity tag to count the number of apoptotic wild-type and ΔRING thymocytes. Caspase-3 activation was unaltered in untreated and dexamethasone-treated ΔRING thymocytes (Fig. 2C). Caspase-3 enzyme activity (Fig. 2D) and apoptosis (Fig. 2E) were also comparable in dying cells of both genotypes. Total thymocyte counts were equal in 4- to 6-wk-old wild-type and ΔRING mice (Fig. 2F), as were the distributions of major cell populations in the thymus (Fig. 2G). These data imply that the RING was important for XIAP protein stability in thymocytes, but its removal affected neither the development of the thymus nor the apoptosis of thymocytes in response to several stimuli. In this cell type, removing XIAP RING function did not reveal any specific proapoptotic role for ubiquitin-mediated XIAP turnover during cell death, and other labile proteins like cIAP1 and Smac were not affected. Likewise, XIAP-null mice have no detectable defects in thymocyte apoptosis (Harlin et al. 2001).

Elevated caspase-3 activity and reduced caspase-3 polyubiquitination in cultured ΔRING embryonic cells during apoptosis

Next we investigated whether XIAP function is required for caspase regulation and apoptosis in other cell types. For this purpose, we focused on ES cells and primary fibroblasts (MEFs [mouse embryonic fibroblasts]). ΔRING ES cells and MEFs displayed significantly elevated caspase-3 enzyme when exposed to different apoptotic stimuli (Fig. 3A). However, the increased caspase-3 activity was not associated with increased death of embryonic cells (Fig. 3B–D). Also, in spite of the elevated caspase-3 enzyme activity in ΔRING cells, the native and cleaved forms of different caspases were expressed similarly in untreated and apoptotic ES cells (Fig. 3D) and MEFs (Fig. 3E). While XIAP, cIAP1, and Smac/DIABLO became labile during thymocyte apoptosis, the expression levels of these proteins did not change much during ES cell apoptosis (Fig. 3D). MEFs are primed to undergo apoptosis following oncogenic transformation (Harrington et al. 1994; White 2001). We generated polyclonal pools of transformed wild-type and ΔRING MEFs by retroviral transduction of RasV12 and E1A to see if the transformed state might sensitize ΔRING MEFs to apoptosis. Indeed, different apoptotic stimuli caused elevated caspase activity in transformed ΔRING MEFs, and apoptosis occurred at earlier time points and lower drug concentrations than with primary cells (Supplemental Fig. 1A). Deletion of the XIAP RING led to elevated spontaneous apoptosis of transformed MEFs, but not in response to etoposide or UVC (Supplemental Fig. 1B,D).

XIAP can polyubiquitinate caspase-3 when it is overexpressed (Suzuki et al. 2001), and the Diap1 RING has a critical function in ubiquitinating the caspase-9 ortholog Dronc in Drosophila (Wilson et al. 2002). Therefore, we asked if genetic deletion of the RING affected caspase-3 ubiquitination in apoptotic cells. For this purpose, we used an antibody that recognizes both full-length caspase-3 and the cleaved large subunit to immunoprecipitate the caspase from mock- and UVC-irradiated cells, followed by immunoblotting with either an antibody against ubiquitin or the large subunit of caspase-3 (Fig. 3F). A “smear” indicating polyubiquitination was detected when caspase-3 was immunoprecipitated from irradiated wild-tye MEFs, but caspase-3 polyubiquitination was severely reduced in irradiated ΔRING MEFs. Collectively, these experiments suggest that XIAP E3-ligase activity is required for ubiquitination of caspase-3 in response to apoptotic stimuli, and that ubiquitination of caspase-3 inhibits its protease activity.

Deletion of XIAP RING sensitizes MEFs to TNF-α apoptosis

The TNF-α cytokine simultaneously engages apoptotic JNK and NF-κB survival signaling pathways in MEFs, and apoptosis only occurs when the NF-κB pathway is inhibited by blocking protein synthesis (e.g., with cycloheximide [CHX]) or by inactivating specific components of the pathway (Beg and Baltimore 1996). Caspase-3 activity was greater in ΔRING cells when they were treated with CHX and TNF-α (Fig. 4A). Furthermore, treatment with TNF-α/CHX accelerated different markers of apoptosis in ΔRING MEFs. Cleaved initiator caspase-8 was detected by indirect immunofluorescence in greater percentages of ΔRING cells over time (Fig. 4B). Caspase-3 activation was accelerated (Fig. 4C), and ΔRING MEFs died faster than wild-type cells (Fig. 4D). We also examined MEFs from mice deficient for XIAP (Harlin et al. 2001) and found them to be equally sensitive to TNF-α/CHX-induced apoptosis (Supplemental Fig. 2A–D). Therefore, at least in this paradigm, deletion of the XIAP RING causes a similar sensitization toward apoptosis as a complete deletion of the XIAP gene. We also observed increased TNF-α apoptosis in ΔRING cells when the NF-κB pathway was inactivated specifically using RNAi against the RelA transcription factor (data not shown). This argues against nonspecific effects arising from general translational inhibition. The ΔRING mutation was not dominant in TNF-α/CHX apoptosis: Complementing ΔRING cells with ectopic wild-type XIAP restored caspase-3 activity to wild-type levels, while expression of ΔRING XIAP did not elevate caspase activity in wild-type cells (Supplemental Fig. 1E).

Both wild-type and ΔRING XIAP were cleaved during apoptosis, while cIAP1 and the BIR-containing protein Survivin were down-regulated similarly in both genotypes (Fig. 4E). The expression of Smac/DIABLO was similar in wild-type and ΔRING cells. Immunoblotting showed that enhanced caspase-3 cleavage was evident after 4 h of treatment with TNF-α/CHX, and expression of the native form of caspase-8 decreased after 8 h in ΔRING cells (Fig. 4E). The principal TNF receptor in MEFs, TNF-R1, was comparably expressed in cells of both genotypes, as were other key cytoplasmic components of the receptor complex like TRAF2, FADD, and c-FLIP, the large form of which was degraded during apoptosis (Fig. 4E). RIP kinase is a likely substrate for caspase-8 (Lin et al. 1999), and RIP was completely cleaved in ΔRING cells at a time (8 h) when most of the cells were apoptotic. JNK activation (as assessed by phosphorylation) was essentially identical in wild-type and ΔRING cells, even at a time (4 h) when there were differences in death between cells of the two genotypes (Fig. 4D,E). Moreover, JNK activation is normal in TNF-α-treated XIAP-null cells (Supplmental Fig. 2E; Harlin et al. 2001). These data suggest that the accelerated caspase activation in ΔRING and null cells is independent or downstream from signals that activate JNK. To further verify that apoptosis was genuinely accelerated in ΔRING cells, we used a FADD antibody to immunoprecipitate the complex that activates caspase-8 in the cytoplasm after TNF-α treatment (Micheau and Tschopp 2003). Unprocessed and cleaved caspase-8 were detected in complex with FADD after 4 h in ΔRING cells; this was evident after 6 h in wild-type cells upon longer exposure of the immunoblot (Fig. 4F). A higher molecular weight form of cIAP1 was recruited to the complex more quickly in ΔRING cells, while the short form of FLIP (FLIP-s) was bound at all times to FADD in cells of both genotypes. The increased susceptibility to death receptor apoptosis was not limited to TNF-α. We saw hypersensitivity to death receptor apoptosis when RasV12/E1A-transformed ΔRING MEFs were treated with the death ligand TRAIL: Cleaved caspase-3 and caspase-8 were detected in a greater number of cells, caspase-3 enzyme activity was elevated, and cells died more readily (Supplemental Fig. 1A–D). These experiments illustrated the general sensitivity to apoptosis caused by members of the TNF-α superfamily of death ligands in cells with XIAP loss-of-function mutations, and they complement genetic arguments that XIAP nonredundantly inhibits TRAIL-induced apoptosis (Cummins et al. 2004).

The ΔRING mutation causes increased apoptosis and increased organismal survival in the Eμ-Myc mouse lymphoma model

XIAP has been implicated in human tumorigenesis and lymphoproliferative diseases (Tamm et al. 2000; Nakagawa et al. 2005; Rigaud et al. 2006), and inhibiting XIAP and other IAPs with Smac/DIABLO-derived peptides (Fulda et al. 2002) or small molecule antagonists can reduce tumor xenografts in mice (Petersen et al. 2007). Although transgenic mice overexpressing human XIAP in the thymus have increased numbers of developing T cells and resistance to apoptosis (Conte et al. 2001), no published mouse models have shown a role for XIAP in tumorigenesis. Since genetic deletion of the XIAP RING domain appears to constitute a loss of function with respect to apoptosis inhibition in certain situations, we explored the effects of this mutation on an established mouse lymphoma and leukemia model. Transgenic mice develop lymphomas and lymphoblastic leukemias when c-Myc is expressed in the B-cell compartment by the Eμ immunoglobulin heavy chain enhancer (Adams et al. 1985). Eμ-Myc transgenic mice provide a sensitized genetic background for studying the phenotypes of apoptosis gene mutations in vivo because misregulated c-Myc expression is a potent inducer of apoptosis in the B-cell lineage (Strasser et al. 1990; Schmitt et al. 2002). By crossing the ΔRING allele onto the Eμ-Myc background and following the survival of wild-type and ΔRING mice over time, we found that ΔRING mice showed improved survival on this sensitized background (median survival times: wild type, 127 d; ΔRING: 246) (Fig. 5A). Peripheral white blood cell counts were consistently lower in ΔRING mice, while wild-type mice tended to show increases over time indicative of large blastic cell leukemias (Fig. 5B; data not shown). Proliferating B-cells in Eμ-Myc mice can be distinguished based on their increased size (Langdon et al. 1986; Sidman et al. 1993). We found decreased numbers of large proliferating B cells in the bone marrow of ΔRING Eμ-Myc mice (Fig. 5C). In order to examine whether this decrease could be due to increased death of B cells in ΔRING Eμ-Myc mice, we compared the rate of apoptosis in Eμ-Myc B cells from wild-type and ΔRING mice. Removal of the XIAP RING led to elevated rates of apoptosis (Fig. 5D), consistent with the idea that increased death of Eμ-Myc B cells accounts for the observed decrease of proliferating B cells in the ΔRING mutant. Because apoptotic cells are rapidly cleared by phagocytes in situ, it is difficult to quantify rates of cell death in the bone marrow (Jacobsen et al. 1994). In order to circumvent this problem, we also isolated pure populations of B cells from the spleens of Eμ-Myc mice. Prior to isolation, the major cell populations in the spleens of wild-type, Eμ-Myc, and Eμ-Myc ΔRING mice were distributed similarly (data not shown). After purifying B cells, the isolated cells were nearly pure populations of early B cells (>95% B220⁺, CD43⁻, IgM⁻) as assessed by flow cytometry. We deprived the purified Eμ-Myc B cells of serum because B cells that misexpress Myc are strongly sensitized to apoptosis caused by growth factor withdrawal (Milner et al. 1993; Cherney et al. 1994). B cells isolated from Eμ-Myc ΔRING mice showed significantly enhanced apoptosis (P < 0.01 by two-tailed Student’s t-test) (Fig. 5E). These results suggest that impaired XIAP E3 ubiquitin-ligase activity inhibits tumor formation and prolongs organismal survial by sensitizing premalignant B cells to apoptosis.

Discussion

Previous efforts to study the regulation of apoptosis by XIAP have largely focused on how BIR domains bind and inhibit caspases, often by relying on overexpression in cancer cell lines and on the biochemical and structural analyses of isolated domains (Hunter et al. 2007). These approaches have offered valuable insights into the molecular mechanisms by which XIAP may regulate caspases, and they also guided the design of small-molecule therapeutics that disrupt BIR/caspase complexes to promote apoptosis in tumors (Carter et al. 2005; Petersen et al. 2007; Varfolomeev et al. 2007; Vince et al. 2007). However, because no significant phenotypes for XIAP-null mice were previously reported, the physiological function of this protein in mammalian apoptosis has remained unclear. Here we show that XIAP function is required for negative regulation of caspases and apoptosis in certain murine cell types in situ, and that inactivation of the E3-ligase activity of XIAP inhibits tumor formation and prolongs survival in a mouse model of lymphoma. Much of the current study focuses on the importance of the XIAP RING domain, which bestows E3 ubiqutin-ligase activity on XIAP. However, we also investigated XIAP-null mice (Harlin et al. 2001) in several paradigms and obtained results virtually identical to the ones seen for XIAP ΔRING (Supplemental Fig. 2A; data not shown). One surprising conclusion from the current study is that deletion of the XIAP RING causes increased caspase activity and sensitivity toward TNF-α apoptosis comparable with complete inactivation of XIAP. This is surprising because the prevailing view, based on biochemical and structural biology data, has been that IAPs inhibit caspases via direct binding through their BIR domain (Salvesen and Duckett 2002; Eckelman and Salvesen 2006). Furthermore, a truncated XIAP protein containing only the BIR domains (BIR 1 + 2 + 3) inhibits caspases in vitro and apoptosis upon overexpression in cultured cells comparably with the full-length protein (Takahashi et al. 1998; Silke et al. 2002). Additionally, the different BIR domains fold correctly, bind caspases, and account for virtually all inhibition of caspases by XIAP in vitro when expressed individually (Takahashi et al. 1998; Sun et al. 1999, 2000; Chai et al. 2001; Riedl et al. 2001; Shiozaki et al. 2003; Scott et al. 2005). Therefore, based on these studies, the increased amounts of “BIR-only” XIAP protein in our XIAP ΔRING mice would have been expected to cause reduced caspase activity and apoptosis, and not increased activity as we found. In addition, the ΔRING mutation does not appear to exert a dominant effect. Although we were not able to directly study heterozygous combinations of wild-type and ΔRING alleles because of X-chromosome inactivation, we found that elevated caspase activity was rescued in ΔRING cells after stably reintroducing a full-length XIAP construct containing the RING (Supplemental Fig. 1C). Finally, expression of a ΔRING XIAP construct did not sensitize wild-type cells to TNF-α/CHX apoptosis, as might have been expected from a dominant-negative mutant.

Our results establish an anti-apoptotic function of the XIAP RING domain for certain cell types in vivo. However, it is important to note that different cell types displayed different requirements for XIAP function. For example, consistent with previous studies, caspase activity and apoptosis of thymocytes was not significantly affected by inactivation of XIAP or loss of XIAP RING function (Fig. 2; Harlin et al. 2001). In addition, the sensitivity to death receptor apoptosis was not common to all ΔRING cell types: Hepatocellular apoptosis occurred normally in vivo following intravenous administration of agonistic Fas antibody, or TNF-α together with the liver-selective transcriptional inhibitor GalN (Supplemental Fig. 2F; data not shown). A key difference between death receptor apoptosis in MEFs and hepatocytes is the obligate requirement for mitochondrial factors in hepatocyte apoptosis (Yin et al. 1999), and it is possible that one of these proteins modulates the effect of the ΔRING mutation. In any case, the present work illustrates the difficulties in drawing generalized conclusions regarding physiological function from in vitro studies and limited analysis of select cell types.

How RING ubiquitin-ligase activity regulates apoptosis in mammalian cells remains largely unknown. In Drosophila, the E3 ubiquitin-ligase activity of Diap1 is required for ubiquitin-mediated degradation of the caspase Dronc and to prevent unwanted apoptosis (Wilson et al. 2002; Ryoo et al. 2004; Steller 2008). In addition, Reaper family IAP antagonists can stimulate Diap1 autoubiquitination during apoptosis that leads to Diap1 degradation, and a RING mutant is stable in the presence of Reaper (Ryoo et al. 2002). Inactivation of the mouse XIAP RING by gene targeting revealed parallels with fruit fly Diap1 RING functions, and differences as well. While Diap1 RING mutants die as advanced embryos, ΔRING mice are viable on mixed and congenic C57BL/6 backgrounds. This may reflect different regulatory strategies between insects and mammals, or alternatively greater functional redundancy of IAP family members in mammals (Steller 2008). Nonetheless, like in Drosophila, XIAP RING function influences caspase activity and ubiquitination in embryonic cells: Caspase-3 enzyme activity was elevated in ΔRING cells in response to diverse apoptotic stimuli and irradiated MEFs were defective in caspase subunit ubiquitination (Fig. 3). However, caspase subunit abundance was not dramatically altered in apoptotic ΔRING cells, indicating that polyubiquitination may not lead to degradation of caspases. In principle the ubiquitination of active caspase-3 by a functional XIAP RING could impair the enzymatic activity of a caspase or its ability to form functional dimers, although ΔRING cells could tolerate higher caspase activity without increased death when cells were killed through the mitochondrial pathway of apoptosis (Fig. 3C).

IAPs have diverse regulatory roles in death receptor signaling pathways (Li et al. 2004; Varfolomeev et al. 2007; Vince et al. 2007). cIAP1 is a key negative regulator of an NF-κB-dependent, autocrine TNF-α apoptosis program that is revealed in the presence of an IAP antagonist, and immortalized cIAP1-null cells are killed by TNF-α alone (Vince et al. 2007). The observation that XIAP-null cancer cells are highly susceptible to TRAIL offered the first genetic evidence that XIAP may be a nonredundant regulator of death receptor apoptosis (Cummins et al. 2004). We established a clear role for the XIAP RING in negatively regulating TNF-α/CHX and TRAIL apoptosis. This effect is probably independent of signals that activate JNK, since the phosphorylation events that activate JNK occurred normally in ΔRING cells (Fig. 4E) and XIAP-null MEFs (Supplemental Fig. 2E; Harlin et al. 2001). Moreover, the JNK phosphorylation state remained transient in ΔRING cells when TNF-α was added alone in the presence of intact NF-κB signaling (data not shown). TNF-α/CHX apoptosis was genuinely accelerated, because the caspase-8 activating complex was assembled more quickly in mutant MEFs (Fig. 4F). RING ubiquitin-ligases usually function to regulate the abundance of their binding partners, but the differential sensitivity between wild-type and ΔRING cells was not obvious from the protein expression levels of key components in the TNF effector pathways.

Although ΔRING mice appeared as healthy as wild-type littermates, we found an anti-apoptotic role for RING function after crossing mutant mice onto the Eμ-Myc transgenic background. The importance of apoptosis in modulating cancer progression in Eμ-Myc mice is evident from animals carrying mutations in cell death genes (Strasser et al. 1990; Schmitt et al. 2002). We found small but reproducible increases in apoptosis in the population of proliferating, premalignant ΔRING B cells, and a general decrease in their relative abundance (Fig. 5D,E). Our genetic evidence supports the long-standing notion that XIAP antagonism can lead to the death of cancer cells (Hunter et al. 2007). Further support for an anti-apoptotic role for XIAP in lymphoid disorders comes from a positional cloning effort that identified XIAP deficiency as a genetic basis for the human X-linked lymphoproliferative syndrome (Rigaud et al. 2006). Lymphocytes in afflicted patients are sensitized to TRAIL and Fas death receptor apoptosis, and natural killer T lymphocytes are depleted. Collectively, our data demonstrate that XIAP, and RING function in particular, play a far greater role in regulating apoptosis in vivo than was previously recognized.

Materials and methods