Unlike Diablo/smac, Grim Promotes Global Ubiquitination and Specific Degradation of X Chromosome-linked Inhibitor of Apoptosis (XIAP) and Neither Cause Apoptosis*

Grim is a Drosophila inhibitor of apoptosis (IAP) antagonist that directly interferes with inhibition of caspases by IAPs. Expression of Grim, or removal of DIAP1, is sufficient to activate apoptosis in fly cells. Transient expression of Grim in mammalian cells induces apoptosis, arguing for the conservation of apoptotic pathways, but cytoplasmic expression of the mammalian IAP antagonist Diablo/smac does not. To understand why, we compared Grim and Diablo. Although they have the same IAP binding specificity, only Grim promoted XIAP ubiquitination and degradation. Grim also synergized with XIAP to promote an increase in total cellular ubiquitination, whereas Diablo antagonized this activity. Surprisingly, Grim-in-duced ubiquitination of XIAP did not require the IAP RING finger. Analysis of a Grim mutant that promoted XIAP degradation, but was not cytotoxic, suggests that Grim killing in transient assays is due to a combination of IAP depletion, blocking of IAP-mediated caspase inhibition, and at least one other unidentified function. Unlike transiently transfected cells, inducible mammalian cell lines can sustain continuous expression of Grim and se-lective degradation of XIAP without undergoing apoptosis, demonstrating that down-regulation

The apoptotic program has been conserved in mammals and flies, with some interesting variations. The core molecules required for apoptosis, i.e. the adaptors required to activate caspases, the caspases themselves, and the inhibitors of apoptosis (IAPs), 1 are found in both organisms. However, although the adaptor molecule Apaf-1 (mammals) and DARK (Drosophila) are both implicated in developmental cell death, Apaf-1 requires activation by cytochrome c whereas DARK does not (1,2). Similarly, caspase-inhibitory IAPs have been identified in both organisms. Although in Drosophila, removal of Drosophila inhibitor of apoptosis (DIAP1) has profound effects resulting in large numbers of ectopic cell deaths (3,4), in mice, removal of X chromosome-linked inhibitor of apoptosis (XIAP) has no developmental or phenotypic consequence (5).
Proteins that bind to and antagonize IAPs have been identified in both flies and mammals. Five IAP antagonists have been identified in Drosophila: Hid, Grim, Reaper, Sickle, and Jafrac. Grim was described as a protein present in the genomic interval at 75C1 (6), a region that had previously been described to be required for programmed cell death in the fly (7). When ectopically expressed, Grim was able to induce cell death in flies and insect cells in tissue culture (6). It was subsequently shown to bind to and antagonize IAPs (8). Because depletion of DIAP1 by RNAi is sufficient to induce cell death in insect cell lines, these activities of Grim would also be sufficient to cause apoptosis (1,9).
The mammalian protein XIAP contains three BIR domains that interact with both caspases and IAP antagonists, as well as a RING finger domain. The RING finger in XIAP has been shown to function as a ubiquitin ligase (10 -12), as has that of DIAP1 (13,14). The RING E3 ligase activity can be directed toward the IAP itself, allowing IAPs to regulate their own levels (10,14). It may also promote ubiquitination of caspases, contributing to their inhibition (12,14), and may target IAP antagonists for degradation (15).
Ubiquitination and degradation of XIAP can be stimulated by Reaper (16), just as ubiquitination and degradation of Hid is stimulated by DIAP1 (17), in each case in a RING-dependent manner. Because XIAP is specifically degraded in thymocytes following an apoptotic stimulus (10), it seems likely that there is a mammalian IAP antagonist that promotes IAP ubiquitination and degradation in an analogous way.
It has been reported that ectopic expression of Grim causes apoptosis when expressed in mammalian cell lines (18,19). This is consistent with the notion that antagonism of IAPs in mammalian cells is also sufficient to initiate apoptosis. However, ectopic expression of the mammalian IAP antagonist Diablo/smac is not sufficient to cause apoptosis, even when the mature form is targeted to the cytoplasm (20). To investigate this discrepancy between fly and mammalian IAP antagonists we investigated whether Grim has other pro-apoptotic functions that are distinct from those of Diablo, and are in addition to IAP antagonism.
As recently described for Reaper and Hid (16,17), we show here that in addition to IAP antagonism, Grim promotes XIAP ubiquitination and destruction. However, in marked contrast to Reaper or Hid, Grim is able to promote ubiquitination of an XIAP lacking its RING domain. As well as targeting XIAP for degradation, Grim also promotes a massive increase in total cellular ubiquitination, whereas Diablo does neither.
In addition to promoting an increase in general protein ubiq-uitination, and simultaneously antagonizing and destabilizing IAPs, Grim must have a third pro-apoptotic function, because a mutant of Grim that promotes XIAP degradation and global ubiquitination does not induce cell death. Using stable cell lines in which XIAP expression can be regulated by tetracycline we confirm that Grim must interact with XIAP for Grim to induce global ubiquitination. In stable cell lines in which Tet can regulate Grim, we show that stably expressed Grim is able to promote degradation of endogenous XIAP. Surprisingly XIAP is specifically targeted by Grim, and levels of ciap-1 and ML-IAP remain constant. Thus it is still possible that mammalian cells like Drosophila cells are unable to tolerate reduction in all IAPs, but consistent with the XIAP knock-out we show they can tolerate expression of an IAP antagonist and reduction in XIAP.
Tagged XIAP and Grim, and paracaspase inserts were removed from pEF C-terminal FLAG-tagged vectors with BamHI NotI and cloned into pcDNA5 FRT TO (Invitrogen) digested with BamHI NotI. HA-tagged ubiquitin was amplified using primers 948 5Ј-cgggatccggtaccaccatggctagctacccttat-3Ј and 952 5Ј-cgcctcgaggctgcaataaacaaagttgggga-3Ј digested KpnI XhoI and inserted into pcDNA5 FRT TO digested KpnI XhoI. All PCR amplifications were performed using Pfu I polymerase (Promega) and all constructs were verified by digest and sequencing.
Stable Cell Lines-Stable cell lines were established by transiently transfecting the Flp-In TM T-REx TM 293 cell line (Invitrogen) with pcDNA5 FRT TO constructs (Invitrogen) with the recommended amount of pOG44. 24 h after transfection cells were split into 15 cm tissue culture plates and selected with 1.6 g/ml hygromycin (Invitrogen) in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (Invitrogen). After 1 week individual colonies were picked, then expanded and tested for tetracycline (Sigma) regulatable expression of the relevant construct, using protein separation on SDS-PAGE gels followed by Western analysis.
Immunoprecipitations (IP) and Western Blot Analysis-Cells were lysed in a Triton X-100-based lysis buffer, the so-called DISC Buffer (1% Triton X-100, 10% glycerol, 150 mM NaCl, 20 mM Tris, pH 7.5, 2 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, and 10 g/ml leupeptin) for 1 h at 4°C and the debris cleared by centrifugation. Immunoprecipitations were performed using FLAG-specific monoclonal antibody M2 covalently coupled agarose beads (Sigma). The immunoprecipitates were washed 4 times in DISC lysis buffer, and proteins were eluted with 100 mM glycine, pH 3, and then neutralized with 1/10th volume of 1 M Tris, pH 8. Proteins were separated by SDS-PAGE and immunoprecipitates were examined by Western blot analysis following transfer of proteins to Immobilon TM -P membranes (Millipore). Alternatively cells were lysed using a digitonin fractionation protocol; cells were harvested, washed in ice-cold phosphate-buffered saline, then lysed in 20 mM HEPES, pH 7.4, 5 mM MgCl 2 , 10 mM KCl, 1 mM EDTA, 1 mM EGTA, 250 mM sucrose, 0.025% digitonin (Calbiochem) for 5 min on ice. The supernatant from a 13,000 rpm centrifugation was retained as the cytoplasmic fraction. The pellet was resuspended in RIPA buffer (150 mM NaCl, 50 mM NaF, 2 mM EDTA, 1% Triton X-100, 10 mM sodium phosphate, pH 7), left on ice for 10 mins, centrifuged at 13,000 rpm, and the supernatant was retained as the membrane fraction. Immunoprecipitations from the cytoplasmic fraction were performed using FLAG-specific monoclonal antibody M2 covalently coupled agarose beads (Sigma) in exactly the same way as from lysates prepared using the DISC lysis buffer.
Apoptosis Assays-293T Cells were used in transient transfection analyses and transfected with test constructs and EGFP at a ratio of 20:1. Green cells were scored as either apoptotic or healthy in a double blind manner. For stable cell lines, cells were fixed in 70% ethanol, washed, and stained with propidium iodide. The cells were then run on a fluorescence-activated cell sorter and the proportion of cells with a sub 2N DNA content calculated.

RESULTS
Grim, Like Diablo, Binds to Residues Asp-214 and Glu-314 of XIAP-To test whether Grim binds to the same residues of XIAP as Diablo, we used a combination of XIAP mutants that have been previously described and extensively characterized (23). A co-transfection and immunoprecipitation analysis was performed with Grim and the XIAP mutants as was previously described for Diablo and XIAP (23). As with Diablo, binding by Grim was markedly dependent on XIAP residues Asp-214 and Glu-314 (compare IP: ␣-Flag probed with ␣-HA panel, lanes 2, 3, 6, 7, and 8 with lane 4, Fig. 1) and Grim only bound to BIR2 and BIR3, but not BIR1 (data not shown). The small residual amount of Grim binding to the IAP mutants may have been due to a small binding potential of Grim for the RING finger (data not shown). A comparison between Grim and Diablo is fair because when they are expressed with the same FLAG Tag very similar levels of expression were observed (see Supplemental Data).
Interestingly, we also noticed that the abundance of wild type XIAP was less than that of the mutants (in Fig. 1 compare lanes 6, 7, and 8 with lane 4 in the lysates probed with anti-FLAG). Because this was a consistent observation in several independent experiments, we wondered whether this was due to promotion of XIAP degradation by Grim.
Grim but Not Diablo Promotes Degradation of XIAP-Cotransfection of Grim and wild type XIAP resulted in a reduction in the total amount of XIAP present. To determine whether this was due to increased ubiquitination of XIAP we co-transfected them together with HA-tagged ubiquitin. Consistent with our previous observation, co-expression of XIAP and Grim reduced the amount of XIAP, but Grim co-expression did not affect levels of a control protein CrmA. This was in marked contrast to findings when we transfected the cells with cytoplasmic Diablo, which did not affect the levels of XIAP or CrmA ( Fig.  2A, compare anti-FLAG probed lysates, lanes 1 and 3).
To gauge the level of XIAP ubiquitination we probed the lysates and FLAG immunoprecipitates with anti-HA ( Fig. 2A). The specificity of the HA antibody for HA ubiquitin was confirmed by including a Grim control that was not co-transfected with HAtagged ubiquitin ( Fig. 2A, lane 5). Co-transfection with Grim increased the amount of ubiquitinated XIAP compared with cotransfection with Diablo, which is particularly striking when the total levels of XIAP are compared ( Fig. 2A, cf. lane 1 with 3). Furthermore, inspection of the lysates shows that Grim induced a global increase in the amount of ubiquitination in the lysates. This is even more striking when total protein present in the lysates is taken into consideration, as judged by the protein loading control blot performed with anti-Hsp70.
We performed co-transfection experiments with XIAP and a range of other proteins to determine whether Grim specifically targeted XIAP for ubiquitination and degradation, and whether the increase in total cellular ubiquitination was a specific effect. Total cellular ubiquitination was very uniform regardless of which protein was transfected, even when the other protein was either an unrelated E3 ligase such as SOCS1, or an E2 ubiquitin conjugating enzyme such as UbcH5.
Cytoplasmic Diablo (Fig. 2B, lane 6) and Grim (Fig. 2B, lane 5) make two notable exceptions. When cytoplasmic Diablo was cotransfected with XIAP there was a general suppression of total cellular ubiquitination by the transiently transfected XIAP (Fig.  2B, lane 6). The full-length Diablo control is particularly illuminating, because this protein is targeted to the mitochondrial inter-membrane space, where it is unable to interact with XIAP. In stark contrast, when Grim was co-transfected with XIAP there was a massive increase in total cellular ubiquitination (Fig. 2B, lane 5). Interestingly, there was no evidence for either Diablo or cytoplasmic Diablo being ubiquitinated by XIAP.
The total amount of XIAP was also very uniform except when Grim was transfected, causing a dramatic decrease in the amount of XIAP. As XIAP is a ubiquitin ligase, it is expected to have a certain basal level of ubiquitination, and, provided other elements within the ubiquitin/proteasome system remain constant, it is expected that the ratio of ubiquitinated to nonubiquitinated XIAP should remain constant. If the amount of ubiquitinated XIAP, most easily visualized in the immunoprecipitation samples, is compared with the total amount of XIAP, the amount of ubiquitinated XIAP does correlate well with the amount of XIAP particularly when the co-transfected protein does not have an obvious role in the ubiquitin/proteasome pathway cf Jak2 and paracaspase (lanes 3 and 4). Co-expression of proteins that do play a role in the ubiquitin/proteasome pathway, such as SOCS1 and UbcH5, does affect the ratio of ubiquitinated to non-ubiquitinated XIAP, but at least in the case of SOCS1 this is unlikely to be due to a specific effect on XIAP but rather a general nonspecific effect. Even when such perturbations on the ubiquitin/proteasome pathway are taken into account, Grim and Diablo produce the most profound disruption in the ratio of ubiquitinated to non-ubiquitinated XIAP. When cytoplasmic Diablo is transfected there is a large amount of XIAP but no ubiquitinated XIAP, and when Grim is transfected there is a greatly reduced level of XIAP but a very large amount of ubiquitinated XIAP. Due to the transient nature of the assay and over-expression of the proteins, an increased ratio of ubiquitinated to non-ubiquitinated XIAP does not always give rise to an unambiguous reduction in XIAP; however, prolonged expression in stable cell lines of proteins that alter this ratio, such as Grim, would be expected to, and does, result in such a reduction (see Fig. 6A).
Grim, Not Apoptosis, Promotes Ubiquitination and Degradation of XIAP-One possible confounding effect in the above analyses was that Grim but not Diablo promotes apoptosis when transiently transfected. Therefore it was possible that it was the induction of apoptosis that indirectly induced XIAP ubiquitination, rather than Grim itself. To test this we performed a similar co-transfection experiment, with two additional controls. First, caspase-9 auto-processes and causes apoptosis when over-expressed in cells (25,26), and as processed caspase-9, like Grim, binds to XIAP in a BIR-dependent manner, we reasoned that this was an excellent control to determine whether apoptosis induced XIAP ubiquitination. Second, we identified a double point mutant of Grim (L88A/L89A) that did not promote apoptosis when co-transfected in mammalian cells and that was expressed to the same levels as the wild type protein. A similar point mutant (L89E) has recently been described that destroys Grim's ability to induce cell death in Drosophila cells (27). A quantitation of the cell death induced by these constructs in 293T cells shows that Grim (L88A/L89A) is as non-toxic as Diablo (Fig. 3A).
From Fig. 3B it is apparent that apoptosis caused by caspase-9 over-expression does not result in XIAP ubiquitination (see lanes 1-3, ␣-Flag IP and ␣-HA panels). Furthermore there was no evidence for ubiquitination of caspase-9 itself by XIAP. The control blots show that equal levels of protein were loaded (lanes 1-3, ␣-Hsp70 lysate panel), that caspase-9 was processed (lanes 1-3, ␣-caspase 9 lysate panel), and that wild type XIAP was able to immunoprecipitate processed caspase-9, but that the mutant XIAP D214S E314S was not (cf. lanes 1 and 3, IP: ␣-Flag ␣-caspase-9 panel), as expected from previous results (23,28). Consistent with the previous results, wild type Grim promoted both XIAP degradation and a global increase in ubiquitination when expressed with wild type XIAP, but caused significantly less degradation and global ubiquitination when expressed with a mutant XIAP (D214S/E314S) that is significantly impaired in its ability to bind Grim (cf. lanes 7 and 9, ␣-HA and ␣-Flag panels). As in Fig. 4, the ratio of ubiquitinated to non-ubiquitinated XIAP has been changed dramatically by Grim co-transfection, resulting in reduced levels of XIAP, but the ratio of ubiquitinated to non-ubiquitinated XIAP in a mutant that cannot interact with Grim remains the same, regardless of whether Grim is transfected or not.
The reduction in XIAP levels was not as pronounced in the samples transfected with Grim L88A/L89A, nevertheless comparison of the levels of XIAP in the lysates shows a significant reduction, and there is also an increase in global ubiquitination when XIAP mutants that are unable to bind Diablo are also significantly impaired in their ability to bind Grim. There is an inverse correlation between the extent of Grim interaction and the amount of XIAP in the lysates. 293T cells were transiently transfected with plasmids expressing FLAG-XIAP and mutants thereof, and C-terminal HA-tagged Grim. Cell lysates were immunoprecipitated with anti-FLAG beads, separated on SDS-PAGE gels, transferred to PVDF membranes, and immunoblotted with anti-HA (to reveal Grim) and anti-FLAG (to reveal XIAP mutants).
XIAP and Grim L88A/L89A are expressed. As in the case of wild type Grim, this required XIAP interaction because the mutant D214S/E314S XIAP was not degraded (Fig. 3B, lanes 6 and 9).
Grim Promotion of XIAP Ubiquitination Requires Grim XIAP Interaction-The obvious mechanism by which Grim might promote XIAP ubiquitination is by activating the RING finger of XIAP that has previously been shown to function as an E3 ligase. To determine whether the RING finger was required, and to confirm that direct Grim-XIAP interaction was necessary for XIAP ubiquitination, co-transfection analyses were performed with different XIAP mutants, C-terminally myctagged Grim and HA-tagged ubiquitin.
XIAPs co-transfected with an empty myc vector have an almost constant ratio of ubiquitinated to non-ubiquitinated protein regardless of the mutation. The one exception to this being the structurally catastrophic H467A mutant, which is clearly unstable. The ␦RING XIAPs have a greatly reduced level of ubiquitination consistent with the idea that the RING is responsible for IAP auto-ubiquitination. Wild type XIAP co-transfected with Grim shows a significant increase in ubiquitination and an equally significant decrease in protein when compared with the vector control (cf. lanes 1  and 9). However a mutant XIAP that is unable to bind Grim (D214S/W310A/E314S) maintains the same ratio of ubiquitinated to non-ubiquitinated in both the Grim and vector control transfections (Fig. 4, cf. lanes 4 and 12). This confirms that The FLAG reactive band with a slightly higher molecular weight than CrmA in lane 4 is a breakdown product of XIAP and is also visible in Fig.  3. B, Grim mediated degradation of XIAP and increase in total cellular ubiquitination is remarkably specific. 293T cells were transiently co-transfected with plasmids expressing wild type XIAP together with a selection of proteins, some of which, such as UbcH5 (lane 1) and SOCS1 (lane 2), function in other ubiquitination and protein degradation pathways. Cell lysates were immunoprecipitated with anti-FLAG beads, separated on SDS-PAGE gels, transferred to PVDF membranes, and immunoblotted with anti-HA (to reveal ubiquitination), anti-FLAG (to reveal XIAP), anti-Diablo, and anti-Hsp70 (to control loading of lysates).
XIAP interaction with Grim is required for XIAP ubiquitination and degradation.
Surprisingly, although removal of the RING finger impaired XIAP ubiquitination by Grim, it did not abolish it (Fig. 4, lane  2). The most marked difference in ubiquitination of XIAP was among XIAP mutants that are unable to bind Grim, and this effect was striking, in either the full-length molecule or in the ␦RING molecules (Fig. 4, cf. lanes 2 and 3). These results show that promotion of XIAP ubiquitination by Grim requires XIAP binding, but contrary to expectation the RING finger was not required for Grim-induced XIAP or cellular ubiquitination. Mutations just N-terminal to the RING finger in a region conserved among different mammalian IAPs (E438A/E439A and R442A/R443A) did not impair Grim-induced ubiquitination and degradation of XIAP and appeared to enhance it, presumably because they actually enhanced binding (cf. lanes 1, 7, and  8). Interestingly both these double mutations make XIAP more similar to DIAP1, probably accounting for the increased affinity of XIAP to Grim. Conversely a mutation in the RING finger itself (K451S/I452A) decreases the affinity of XIAP for Grim and as expected renders the mutant protein more resistant to Grim-induced ubiquitination and degradation.
Previous publications have described an H467A XIAP mutant that was unable to auto-ubiquitinate (10,12,16). However even in transient transfection experiments this structurally catastrophic mutant was considerably less stable than wild type XIAP (Fig. 4, lanes 6 and 14, ␣-HA and ␣-Flag panels), 2 and in isogenic stable cell lines this mutant was almost undetectable (data not shown), and therefore it is difficult to draw conclusions about its ability to auto-ubiquitinate in the presence of Grim.

Grim Promotes Ubiquitination in the Presence of Stably Expressed XIAP-Although
Grim was able to promote XIAP degradation when both were transiently transfected, it remained possible that this was merely an artifact of the transfection process. To address this issue we created isogenic stable cell lines bearing tetracycline regulatable constructs. Using the Flp-In TM T-REx TM 293 cell line (Invitrogen) we created FLAG XIAP and FLAG XIAP mutant cell lines. The isogenic nature of these lines is shown by the equal levels of expression of XIAP obtained from different cell lines (Fig. 5, cf. lanes 1, 5, and 7).
The stable cell lines were transiently transfected with FLAG Grim and HA-ubiquitin and treated with or without tetracycline. Presumably because of the low transfection efficiencies (ϳ15%) and the lack of T-antigen driven replication of the transfected plasmid in this 293 cell line, it was not possible to see a reduction in the amount of stable XIAP in these assays. Consistent with the transient transfection analyses, however, Grim induced an increase in ubiquitination of wild type XIAP in two different cell lines, but not of a mutant XIAP (D214S/ E314S) that is unable to bind Grim (Fig. 5, cf. lanes 2 or 7 with  lane 6). In contrast, Diablo did not induce an increase in XIAP 2 Unpublished observations. GFP was used as a marker to indicate transfected cells that were subsequently scored for viability by morphology. The mean Ϯ S.E. of 3 independent transfections are shown. B, Grim and Grim L88A/L89A promote a general increase in ubiquitination and increase in ubiquitination of XIAP. As for Fig. 2A, CrmA is included as a control for the specificity of XIAP degradation by Grim. cl. casp9, processed caspase-9 p37 subunit; cl. XIAP, XIAP processed by caspase-3 between the BIR2 and BIR3; pro casp9, zymogen full-length caspase-9. 293T cells were transiently transfected with caspase-9, FLAG-Grim, or FLAG-Grim LL88AA. Cell lysates were immunoprecipitated with anti-FLAG beads, separated on SDS-PAGE gels, transferred to PVDF membranes, and immunoblotted with anti-HA (to reveal ubiquitination), anti-FLAG (to reveal XIAP and Grim), anti-Diablo, anti-caspase-9, and anti-Hsp70 (to control loading of lysates) .   FIG. 3. Grim, but not apoptosis,  ubiquitination. The slight increase in ubiquitination caused by transient transfection of Grim apparent in lanes 4, 6, and 9 is probably due to activation of the E3 ligase activity of endogenous XIAP, because we were not able to show Grim promoting ubiquitination of either ML-IAP or ciap-1 in transient transfection assays (see Supplemental Data).
Expression of Grim in Stable Lines Promotes Degradation of XIAP-To investigate the ability of Grim to promote degradation of XIAP in stable lines, we used the Flp-In TM T-REx TM system to create cells that inducibly express either Grim or the Grim LL88AA mutant. Consistent with the transient transfection experiments, Grim promoted ubiquitination and degradation of transiently transfected wild type XIAP and ѨRING XIAP (data not shown).
Immunoprecipitations of lysates of the stable lines were performed to look at the ability of Grim to bind to endogenous XIAP and promote its degradation, but initially we were not able to immunoprecipitate endogenous XIAP with FLAG-Grim from lysates made from these cell lines, even though both FLAG-Grim and XIAP were present (Fig. 6). Because the lysis buffer used released endogenous IAP antagonists, such as Diablo and HtrA2/Omi from the mitochondria, it was possible that these were competing with Grim for binding to XIAP post lysis. We therefore induced expression of Grim in the stable cell lines for 48 h and then fractionated the cells with digitonin into membrane and cytoplasmic fractions.
The fractionations were performed successfully as both VDAC/porin and cytochrome c were present only in the mem-brane fractions (Fig. 6A). Both XIAP and Grim localized to both membrane and cytoplasmic fractions, with XIAP being more abundant in the cytoplasm. The ␣-XIAP panel in Fig. 6A shows that the level of endogenous XIAP was specifically reduced in both cytoplasmic and membrane fractions of the Grim stable cell line, because none of the other proteins examined (e.g. Hsp70, cytC, VDAC/Porin, caspase-9, or caspase-3) were reduced. Although only a comparison with a stably expressing paracaspase cell line is shown as a control, levels of endogenous XIAP in this cell line were identical with other stable cell lines tested (data not shown). Importantly, FLAG-Grim immunoprecipitation from the cytoplasmic fraction, which contained Grim but did not contain any mitochondrial proteins, was able to pull down endogenous XIAP (Fig. 6B). Although the Grim L88A/ L89A mutant was not able to immunoprecipitate endogenous XIAP from the cytoplasmic fraction, it is clear from an examination of the lysates and immunoprecipitations that there was less Grim LL88AA than wild type Grim in the cytoplasmic fraction, and it is therefore not possible to make a conclusion about the ability of Grim LL88AA to bind to XIAP. Nevertheless there was significantly less XIAP in the membrane fraction from the Grim LL88AA stable cell line, which is consistent with our previous observations that Grim L88A/L89A was able to promote XIAP ubiquitination, albeit with a slightly reduced activity, and consistent with the idea that XIAP can still bind Grim LL88AA.
Induction of Grim in Stable Cell Lines Does Not Induce Apoptosis and Does Not Greatly Sensitize Cells to Apoptotic  1-8) together with HA ubiquitin and C-terminally myc-tagged Grim or empty myc vector. Cell lysates were immunoprecipitated with anti-FLAG beads, separated on SDS-PAGE gels, transferred to PVDF membranes, and immunoblotted with anti-HA (to reveal ubiquitination) and anti-FLAG (to reveal XIAP and Grim).
Insults-As expected, the viability of all the stable cell lines used in this study was approximately equal in the absence of tetracycline (data not shown). Surprisingly, however, in the presence of tetracycline, all stable cell lines tested were equally viable, with the Grim lines showing more or less the same viability as cell lines over-expressing XIAP (Fig. 7). This was true whether the cells were exposed to a short pulse of tetracycline (Fig. 7) or prolonged induction with tetracycline over a month. This similar viability was reflected in very similar doubling times over prolonged periods of tetracycline induction with an average doubling time for all isogenic lines of 38 h. Furthermore after a month of tetracycline induction, all isogenic lines, including Grim, maintained expression of their respective transgene (data not shown). Thus continuous expression of Grim, despite reducing levels of endogenous XIAP, had little effect on normal cell viability.
To ascertain whether reduced levels of XIAP might sensitize cells to apoptotic stimuli we determined cellular viability following low doses of etoposide treatment (Fig. 7). Although cells over-expressing wild type XIAP showed decreased sensitivity toward etoposide, cells expressing Grim were only slightly more sensitive than isogenic cell lines expressing an inactive XIAP mutant D148A/D214S/E314S or GFP (Fig. 7). DISCUSSION Six recent publications reported the ability of Reaper, Grim and Hid to regulate DIAP1 levels in Drosophila (13,14,16,17,29,30). Hay and co-workers (17) reported that Hid caused a RING-dependent degradation of DIAP1, and that in contrast Reaper and Grim down-regulated DIAP1 levels in a less specific manner by interfering with general translation (17). The finding of RING-dependent degradation by Hid was consistent with other observations that a DIAP1 RING finger mutant was able to suppress Hid, but not Reaper and Grim, induced cell death in the eye (4,14,31). In contrast, Holley et al. (16) and Steller and co-workers (13) reported that Reaper was able to greatly decrease IAP abundance by increasing IAP degradation, a function that required the IAP RING finger, but in addition Holley et al. (16) also reported that Reaper repressed total protein translation. Cagan and co-workers (20) similarly reported that Grim and Reaper, but not Hid, promoted degradation of DIAP1 in vivo. Overall, therefore, although there was a broad area of agreement between the six studies that some of the Drosophila IAP antagonists can affect levels of DIAP1, they proposed different mechanisms and a consensus of how Drosophila IAP antagonists affect DIAP1 levels remains elusive.
The observation that DIAP1 became more multi-ubiquitinated in an IAP antagonist-dependent fashion prompted us to investigate the ability of the best characterized mammalian IAP antagonist, Diablo/smac, to promote degradation of mammalian IAPs. By using Grim as a positive control we have decisively shown that Diablo cannot promote ubiquitination of XIAP, and in fact actually has the opposite activity. However, our findings do show that a mechanism for "Grim-like" induction of IAP degradation exists in mammals. Consistent with observations for Reaper (16), we find that Grim induces a specific degradation of XIAP that occurs through an increase in XIAP ubiquitination. Consistent with several of the above studies we also see a decrease in total protein when Grim is transiently transfected into mammalian cells, which might result from a general inhibition of translation (16,17,30). This decrease in total protein is not caused by the cytotoxicity associated with Grim, as a non-toxic Grim mutant was as effective as wild type Grim in reducing total protein. An alternative explanation for the decrease in total protein would be that Grim induces an increase in protein turnover as evidenced by the increased ubiquitination present when Grim is transiently transfected. Possibly a combination of both these processes occur and Grim promotes the ubiquitination and degradation of an essential translational component. Whatever the hypothesis, it seems as though this effect is small when compared with the ability of Grim to promote ubiquitination and subsequent degradation of XIAP, because stably expressed Grim promoted destruction of endogenous XIAP, but did not noticeably affect the levels of any other proteins tested. Although our results are consistent with Holley et al. (16) they are not consistent with those from Hay and co-workers (17) who showed in vitro that Hid, but not Grim, was capable of inducing DIAP1 auto-ubiquitination. The two systems are obviously quite different and Hay and co-workers (17) cautioned that "lack of in vitro activity may well have represented adoption of unphysiological conformations of Grim"; however, it would be unusual if Grim could promote the ubiquitination of XIAP but not DIAP1.
Our findings also add to our understanding of the mechanism of Grim-induced IAP degradation. Firstly we find that XIAP ␦RING becomes ubiquitinated and degraded when coexpressed with Grim. This finding is incompatible with the proposal that Grim promotes the E3 ligase activity of IAPs. It is possible that Grim activates the E3 ligase activity of the endogenous XIAP, which then oligomerizes with and trans- FIG. 5. Grim promotes ubiquitination in the presence of stably expressed XIAP. Stable cell lines were generated using a commercially available Flp recombinase system to create isogenic stables that are tetracycline inducible. All constructs (FLAG-XIAP, FLAG-XIAP mut, and EGFP) are integrated into the same genomic locus under the same promoter. These stable cell lines were transiently transfected with Diablo (lanes 1, 3, 5, and 8) or FLAG-Grim (lanes 2, 4, 6, 7, and 9) together with HA ubiquitin. A second XIAP stable cell line was included (lane 7) to confirm that the XIAP effect was repeatable in more than one cell line. N-ethyl maleimide (10 mM) was added post lysis to prevent deubiquitination in the lysates. Cell lysates were immunoprecipitated with anti-FLAG beads, separated on SDS-PAGE gels, transferred to PVDF membranes, and immunoblotted with anti-HA (to reveal ubiquitination) and anti-FLAG (to reveal XIAP and Grim) and anti-XIAP.
ubiquitinates the delta RING construct; however, we do not favor this model for several reasons. First, ␦RING XIAP oligomerizes with full-length XIAP significantly less well than full-length XIAP with itself (23). Second, transiently transfected ␦RING XIAP became efficiently ubiquitinated even though amounts of this protein vastly exceed those of the endogenous full-length protein. Last, and most telling, ␦-RING XIAP must be able to bind to Grim to become ubiquitinated. We therefore propose that Grim actually serves as a surrogate E3 ligase to directly recruit and activate the E2 ubiquitin conjugating enzyme. Holley et al. (16) found that Reaper was able to stimulate XIAP degradation but not that of the RING finger H467A mutant. As our data were obtained with Grim, it is possible that Grim and Reaper function differently in promoting IAP degradation. However, we show that the H467A mutant is extremely unstable, as might be expected from such a structurally catastrophic mutant, and we believe, therefore, that it is difficult to draw conclusions regarding this particular mutant.
Although the mechanisms regulating levels of IAPs appear to have been conserved between mammals and flies, an important difference exists. XIAP with a RING finger that has had a Zn co-ordinating ligand disrupted (e.g. H467A) is significantly less stable than the wild type protein. It appears therefore that although RING mediated degradation of XIAP is important in determining the level of XIAP in the cell, other factors such as structural integrity are in fact more important than RING mediated degradation alone. This is in contrast to the reports that show that DIAP1 with a Zn co-ordinating ligand disrupted (i.e. C412Y) is in fact more stable than wild type DIAP1 (14).
We also observed a significant increase in total cellular ubiquitination when Grim was either transfected or induced in stable cell lines. This increase in total cellular ubiquitination was further augmented when XIAP was co-expressed with Grim. It is noteworthy that cytoplasmic Diablo decreased the levels of total cellular ubiquitination, arguing that the increase in ubiquitination is specific to XIAP. The increase in ubiquitination due to Grim could result from a general activation of XIAP's ubiquitinating activity, but might also occur if Grim activates other RING finger proteins, such as ciap-1 or ciap-2. We favor the hypothesis that the increase in total ubiquitination results predominantly from an activation of the E3 ligase activity of XIAP, because in transient transfection experiments we saw no significant increase in total cellular ubiquitination when Grim was co-transfected with ciap-1 or -2 (see Supplemental Figure). The large general increase in ubiquitination caused by XIAP indicates that there are a large number of XIAP ubiquitination substrates, and these substrates are un-FIG. 6. Expression of Grim in stable lines results in degradation of endogenous XIAP. Stable Grim cell lines were generated using a commercially available Flp recombinase system to create isogenic stables that are induced by tetracycline. A, XIAP and Grim localize to the cytoplasmic and membrane fractions and endogenous XIAP is specifically reduced in the Grim and Grim L88A/L89A mutant cell lines. Cells were induced with tetracycline for 48 h and lysed using a digitonin fractionation protocol. Cytoplasmic and membrane fractions were separated on SDS-PAGE gels, transferred to PVDF membranes, and immunoblotted with anti-XIAP, anti-VDAC/Porin, anti-Hsp70, anti-FLAG, anti-cytochrome c, anti-caspase-3, and anti-caspase-9. B, Grim localized in the cytoplasmic fraction is able to immunoprecipitate endogenous XIAP present in the cytoplasmic fraction of FLAG Grim stable cell lines. Cells were induced with tetracycline for 48 h and lysed using a digitonin fractionation protocol. The cytoplasmic fraction was immunoprecipitated with anti-FLAG beads, separated on SDS-PAGE gels, transferred to PVDF membranes, and immunoblotted with anti-XIAP and anti-FLAG.
likely to be only known XIAP interacting proteins because strikingly we saw no evidence for caspase-9 or Diablo ubiquitination by XIAP.
It has been previously reported (18,19) that Grim is able to induce apoptosis in mammalian cells. We confirmed these observations and also identified a mutant of Grim that was unable to promote cell death in transient transfection, i.e. LL88AA. We tested this mutant because it is part of a partially conserved helix that is present in Hid, Grim, Reaper, and Sickle. A recent study (27) has reported that this loosely conserved helix, or GH3, is required for Grim to induce cell death in Drosophila. A point mutant Grim (L89E) was also analyzed, and consistent with our observations it was unable to promote cell death in Drosophila. Our results support the idea that Grim has a pro-apoptotic function distinct from IAP antagonism, and additionally indicate that this pro-apoptotic function is distinct from IAP degradation as this mutant was almost as effective as wild type in promoting XIAP degradation. However, our results indicate that mammalian cells are able to tolerate lower levels of this cytotoxic function, because cells with stably integrated Grim were almost as healthy as cells stably transfected with GFP. Most likely the partner for the cytotoxic activity of Grim is not as well conserved in mammals as the IAPs are.
Stable cell lines containing an inducible Grim construct continuously induced with tetracycline were as viable as any other isogenic cell line we analyzed. This was despite a significant reduction in the levels of XIAP. This reduction in levels of XIAP was very specific as none of VDAC, cytochrome C, Hsp70 caspase-9, or caspase-3 were reduced. Although not all XIAP was lost from these cells, reduction of DIAP1 mRNA to one quarter of the initial amount by double-stranded RNA treatment was sufficient to induce cell death in Drosophila SL2 cells (1). This ability of mammalian cells to withstand a reduction in XIAP implies a significantly different role for IAPs in mammalian compared with Drosophila cells. This ability is also consistent with the fact that mammalian cells tolerate high levels of ectopic expression of Diablo in the cytoplasm, and the lack of a phenotype when XIAP is knocked out in mice.
In the "intrinsic" model of apoptosis cytochrome c released from mitochondria activates Apaf-1 that in turn activates caspase-9, which then activates caspase-3. Activated caspase-3 is able to feed back to process and activate caspase-9, amplifying the original signal. We have proposed that IAPs in mammalian cells function as a safety catch that prevents initiation of the caspase-9, caspase-3 amplification loop until IAP antagonists release them. Our data suggest that in mammalian cells activation of caspases is tightly regulated, and that in most normal circumstances there is no accidental activation of either caspase-3 or caspase-9. These results are also consistent with the lack of phenotype of XIAP knock-out mice. If this model is correct, the challenge is to find situations where the existence of the IAP safety catch is critical.