Antagonists Induce a Conformational Change in cIAP1 That Promotes Autoubiquitination

See allHide authors and affiliations

Science  21 Oct 2011:
Vol. 334, Issue 6054, pp. 376-380
DOI: 10.1126/science.1207862


Inhibitor of apoptosis (IAP) proteins are negative regulators of cell death. IAP family members contain RING domains that impart E3 ubiquitin ligase activity. Binding of endogenous or small-molecule antagonists to select baculovirus IAP repeat (BIR) domains within cellular IAP (cIAP) proteins promotes autoubiquitination and proteasomal degradation and so releases inhibition of apoptosis mediated by cIAP. Although the molecular details of antagonist–BIR domain interactions are well understood, it is not clear how this binding event influences the activity of the RING domain. Here biochemical and structural studies reveal that the unliganded, multidomain cIAP1 sequesters the RING domain within a compact, monomeric structure that prevents RING dimerization. Antagonist binding induces conformational rearrangements that enable RING dimerization and formation of the active E3 ligase.

Inhibitor of apoptosis (IAP) proteins are anti-apoptotic factors important in blocking programmed cell death, or apoptosis, in response to a variety of stimuli (1, 2). Whether initiated by external death signals transduced by specific cell surface receptors (extrinsic pathway) or by internal cues of compromised cellular integrity (intrinsic pathway), apoptotic signaling pathways converge in the activation of caspases (cysteine-dependent aspartyl-specific proteases), which effect widespread proteolytic damage and cell death (3). IAPs hold these cellular executioners in check, either through direct inhibitory interactions or by impeding upstream caspase activation pathways (4). Many cancer cells overexpress IAPs, which allows them to resist cytotoxic therapies (2, 5). Thus, IAPs are potentially important targets for cancer treatment (2, 5, 6).

IAP-targeting therapeutics designed to mimic the endogenous IAP antagonist, SMAC (second mitochondrial activator of caspases)–DIABLO (direct IAP-binding protein with low isoelectric point) (7, 8), have recently entered phase I clinical trials (9). The conserved amino-terminal tetra–amino acid peptide (Ala-Val-Pro-Ile or AVPI) of mature SMAC binds to a surface groove on baculovirus IAP repeat (BIR) domains (10, 11). SMAC mimetics show proapoptotic activity by preventing interactions between IAPs and active caspases (2, 4). Moreover, binding of these small-molecule antagonists to cellular IAP (cIAP) BIR domains leads to a second mechanism of antagonism—rapid proteasomal degradation of the cIAP proteins induced by autoubiquitination (1214). Although analyses of antagonist–BIR domain complexes have yielded detailed information regarding the molecular determinants of these interactions (1517), how antagonist binding induces cIAP RING E3 ligase activity is not understood.

We sought to define a minimal cIAP1 construct that would maintain key full-length cIAP1 properties and be suitable for biophysical and structural studies. As IAP antagonists bind preferentially to the BIR3 domain of cIAP1 (13), we focused on a truncated version of the protein that lacks the first two BIR domains (Fig. 1A and fig. S1A). Removal of the BIR1 and BIR2 domains, which include the tumor necrosis factor receptor–associated factor 2 (TRAF2)–interacting region of cIAP1 (1820), did not affect antagonist-induced degradation (fig. S1). The resulting cIAP1 construct, comprising the BIR3 domain followed by the ubiquitin-associated (UBA) domain, the caspase activation and recruitment domain (CARD), and the carboxy-terminal RING domain (BIR3-RING), recapitulates full-length cIAP1 behavior and demonstrates proteasomal degradation that is mediated by IAP antagonist-induced autoubiquitination activity (fig. S1).

Fig. 1

BIR3-binding IAP antagonists induce RING dimerization. (A) cIAP1 constructs. (B) Native gels show the relative mobility of BIR3-RING and BIR3-RINGΔC7 proteins (M, monomer; D, dimer) in the absence and presence of IAP antagonists. The differing effects of monovalent antagonists MV1 and AVPW on cIAP1 BIR3-RING protein mobility are consistent with their binding half-lives of 0.9 and 4.2 min, respectively, as determined by surface plasmon resonance (SPR) spectroscopy (fig. S2, A and B). For comparison, the half-life of BV6 binding to BIR3-RING is 10.6 min (fig. S2C).

RING E3 ligases often function as dimers (21, 22), and disruption of RING-mediated dimerization stabilizes IAPs against IAP antagonists (23). We thus assessed whether IAP antagonists affect the oligomerization state of the BIR3-RING construct. Monomeric unliganded BIR3-RING appears as a single, discrete band by native gel analysis (Fig. 1B). Addition of a bivalent antagonist (BV6) that can cross-link the protein through simultaneous binding of separate BIR3 domains (13) is a control for migration of the BIR3-RING dimer. Analytical size-exclusion chromatography and multiangle light scattering (SEC-MALS) results confirm that the native gel bands represent monomer and dimer BIR3-RING species (fig. S2D). Monovalent small-molecule (MV1) and peptide (AVPW) antagonists shift the mobility of the BIR3-RING band toward the BV6 dimer control. In contrast, a BIR3-RING construct lacking RING dimerization residues at the carboxy terminus (M266 to G611; BIR3-RINGΔC7) (18, 23) does not band-shift unless presented with the bivalent antagonist. Together, these findings suggest that binding of monovalent antagonists to the BIR3 domain induces RING dimerization through an intramolecular, allosteric mechanism.

It intrigued us that the mobility of the ligand-free “apo” BIR3-RINGΔC7 band is considerably retarded compared with apo BIR3-RING, despite modest differences in molecular weight and net charge for the two proteins (Fig. 1B). MV1- and AVPW-bound BIR3-RINGΔC7 demonstrate identical band migrations as the apo BIR3-RINGΔC7 sample. Further SEC-MALS analysis corroborates the distinct mobilities seen by native gel (fig. S2D). These results suggest that the dimerization-deficient mutant exists in a more extended conformation than the monomeric BIR3-RING protein and that the carboxy-terminal residues of the RING domain are critical for maintaining the more compact apo structure.

Crystallization required design of a variant of the BIR3-RING protein that eliminated two protease cleavage sites by deletion of a 10-residue section within the BIR3 to UBA linker (ΔS363 to D372) and a point mutation in which Tyr replaced Phe387 (F387Y) (BIR3-RINGxtal) (figs. S1 and S3). The structure of BIR3-RINGxtal was solved to 1.9 Å resolution by single-wavelength anomalous dispersion (SAD) by using three bound zinc ions, and the final model was refined to an Rfree of 23% (fig. S4 and table S1).

Consistent with an allosteric activation hypothesis, the apo monomer adopts a compact conformation, with the BIR3 domain, UBA domain, CARD, and intervening linkers combining to sequester a centralized RING domain (Fig. 2A). Multiple interfaces formed between the RING domain and the surrounding components together bury >4400 Å2 of surface area (table S2). Although portions of the RING domain remain accessible to solvent, key RING dimerization elements are buried or displaced in this autoinhibited state (23) (Fig. 2B). In contrast, the conformations of the BIR3 domain, UBA domain, and CARD are similar to homologous domain structures (fig. S6). The interdomain linkers also form well-ordered secondary structures that interact with the RING domain. Inspection of the interdomain interfaces reveals specific interactions that appear to contribute to the confinement of the RING domain and thus to suppression of RING:RING dimerization. The BIR3 domain forms an extensive interface with the RING domain (table S2), which is buttressed by the UBA domain and a section of the BIR3-UBA linker (BU-linker). The BIR3 residue W329, located in a hydrophobic pocket between these four modules, is positioned to interact favorably with the helix dipole of an adjacent UBA α helix (Fig. 3A). In addition to contributing to this pocket, residues P353 to T362 of the BU-linker form an amphipathic helix whose hydrophobic surface pairs with the hydrophobic face of the RING N helix and a nearby surface of the RING domain (Fig. 3B). These interactions pin the N helix in a bent conformation (relative to dimeric RING protomers) (Fig. 2B) by effectively substituting the N helix dimerization contacts with similar intramolecular interactions. On the opposite side of the RING domain, backbone carbonyl groups and residue R600 interact with R300 and R314, and N301 from the BIR3 domain, respectively (Fig. 3C). The remaining RING interactions are also polar, with the large CARD:RING interface (table S2) characterized by a hydrogen-bonding network that includes residues R465, E508, D511, and R569, as well as ordered water molecules (Fig. 3D). Additionally, the UBA-CARD linker (UC-linker) acts as a negatively charged helical “strut” that extends from the final UBA α helix. The strut flanks a positively charged surface of the RING domain (Figs. 3E and 2A) and thus positions the CARD for its interactions with the RING domain. Finally, the C strand of the RING associates with the edge of the BIR3 β sheet. Comparison with a structure of the BIR3 domain bound to the SMAC amino-terminal peptide reveals that the RING C strand partially occludes the antagonist-binding site (Fig. 3F) (24).

Fig. 2

Crystal structure of the apo cIAP1 monomer. (A) A ribbon diagram of BIR3-RINGxtal illustrates the global arrangement of the BIR3 domain, UBA domain, and CARD around a core RING domain. Inset shows surface views of the structure. Protein domains are shown in cyan, green, purple, orange, and yellow as indicated; zinc ions are shown as gray spheres. (B) Structural alignment with a dimeric cIAP2 RING protomer (red) illustrates how the RING dimerization elements are sequestered in the BIR3-RINGxtal structure. Specifically, the RING carboxy-terminal strand (C strand), which self-associates in the RING:RING interface, interacts with the edge strand of the BIR3 β sheet in the monomer structure; and the RING amino-terminal helix (N helix) is bent ~110° and partially unwound as compared with dimerized, active cIAP2 RING protomers and so affects the E2-binding surface (fig. S5). cIAP1 domains other than the RING domain are shown faded for visual clarity. Inset depicts the cIAP2 RING dimer structure (red and gray protomers, PDB code 3EB5).

Fig. 3

Domain-domain interactions. Detailed views of the (A) BIR3:BU-linker:UBA:RING, (B) BU-linker:RING, (C) BIR3:RING (D) CARD:RING, and (E) UC-linker:RING interfaces are shown with key residues highlighted as sticks. Color scheme is the same as that of Fig. 2A, with the exception of electrostatic surfaces shown for the UC-linker (top) and RING (bottom) in panel (E). (F) Overlay of the RING C strand (orange sticks) and SMAC-peptide antagonist (white sticks, PDB code 3D9U) within the BIR3 (cyan surface) peptide-binding groove. Unlike the peptide antagonist, which demonstrates good surface complementarity and makes numerous favorable interactions with the BIR3 domain, the RING C strand merely crosses over the peptide-binding groove, making a single hydrogen bond contact with G312 of the BIR3 domain via residue T612. (G) Native gel analysis reveals mutants of BIR3-RINGxtal that dimerize in the absence of antagonist (red labels). All gels aligned by standards shown at far left (stds). Positions of WT monomer (M) and dimer (D) are indicated. (H) The stability of full-length FLAG-tagged cIAP1 variants in 293T cells (±10 μM MG132 or 2 μM BV6 for 1 hour) was determined by evaluating cell lysates for cIAP1 constructs by Western blotting with anti-Flag antibody. Actin was probed as a loading control.

To evaluate the contributions of these and other specific interactions to the compact monomer state, we introduced mutations designed to disrupt these contacts into the BIR3-RINGxtal construct (fig. S7). Evaluating these variants with native gels revealed that mutations within the BIR3:RING and BU-linker:RING interfaces promote dimerization, whereas mutations targeting other interactions have little to no effect, compared with the near–wild-type (“WT”) BIR3-RINGxtal control (Fig. 3G). Note that the cIAP1 monomer was insensitive to perturbation of the highly polar CARD:RING and UC-linker:RING interfaces, consistent with previous gross deletions of regions encompassing the UBA domain or CARD that showed no observable effect on cIAP1 autoubiquitination activity (18). SEC-MALS analyses of the dimer-promoting mutants confirm the dimeric state of these variants in the absence of antagonist (fig. S8). Accordingly, addition of AVPW did not significantly alter the apparent molecular weight of these mutants, whereas WT and W329F monomer samples shifted to dimeric species under similar conditions.

We next assessed the effect of these interface mutations on the stability and, therefore, on the autoubiquitination activity of full-length cIAP1 expressed in human embryonic kidney 293T cells. As predicted, WT cIAP1 was considerably more stable than variants expected to dimerize on the basis of biophysical analysis of the BIR3-RING constructs (Fig. 3H). This was particularly evident for the W329A and T612R mutants; these could only be detected after treatment with the proteasome inhibitor, MG132, which suggests that their instability results from elevated autoubiquitination. Similar results were observed on expression of these variants of the shorter BIR3-RINGxtal construct (fig. S9).

To visualize activated, dimeric cIAP1, we collected small-angle x-ray scattering (SAXS) data for the AVPW-bound BIR3-RING construct. An averaged ab initio model of the dimer reveals an extended, two-fold symmetric complex that is compatible with placement of the RING:RING homodimer structure at its center (Fig. 4, A and B). Ab initio SAXS models generated for two other cIAP1 constructs, an N-terminally MBP-tagged BIR3-RING (MBP-BIR3-RING) dimer and a shorter UBA-RING (S385 to S618) dimer, support a model in which BIR3 domains are located at the periphery of an elongated, nearly planar BIR3-RING homodimer that forms around a central RING:RING interface (Fig. 4, A to C, and figs. S10 and S11). In agreement with the mutational analysis, UBA-RING purifies as a constitutive dimer, whereas MBP-BIR3-RING, like BIR3-RING, requires the addition of antagonist (AVPW) to dimerize (fig. S12).

Fig. 4

SAXS analysis of the cIAP1 dimer. (A) Proposed domain arrangements for MBP-BIR3-RING, BIR3-RING, and UBA-RING dimers and (B) two views of averaged ab initio SAXS envelopes (P2 symmetry) for the same constructs. All data were collected in the presence of 1 mM AVPW. (C) Pairwise distance-distribution functions of SAXS data for each dimer.

Taken together, these data suggest an E3 ligase activation model in which apo BIR3-RING exists predominantly in a closed, inactive monomer state. Antagonist binding to the BIR3 domain is incompatible with this conformation and stabilizes an open state in which the RING domain is exposed and free to dimerize. RING dimerization enables rapid cIAP1 autoubiquitination and subsequent degradation by the proteasome (23, 25). Given the high sequence conservation between cIAP1 and 2, the proposed model likely reflects a general mechanism of antagonist-induced cIAP activation, although recent work suggests that apo cIAP2 may have a greater propensity to dimerize than cIAP1 (25). In contrast, modeling suggests that X-linked AIP (XIAP), which lacks the CARD, could not adopt the same autoinhibited conformation as cIAP1. Nevertheless, the current work provides a molecular basis for understanding the mechanism of action of IAP antagonists and offers important insights into the regulation of cIAP E3 ligase activity.

Supporting Online Material

Materials and Methods

Figs. S1 to S12

Tables S1 and S2

References (2637)

References and Notes

  1. Acknowledgments: We thank C. Yu, P. Lupardus, Shamrock Structures LLC, and staff at the Advanced Light Source, the Stanford Synchrotron Radiation Lightsource (SSRL) and the Advanced Photon Source at Argonne National Laboratory for their assistance with sample handling and data collection. The Advanced Light Source is supported by the Director, Office of Science, Office of Basic Energy Sciences, of the U.S. Department of Energy (DOE) under contract no. DE-AC02-05CH11231. SSRL is a Directorate of SLAC National Accelerator Laboratory and an Office of Science User Facility operated for the DOE Office of Science by Stanford University. The SSRL Structural Molecular Biology Program is supported by the DOE Office of Biological and Environmental Research and by the Biomedical Technology Program, National Center for Research Resources, NIH (P41RR001209). Use of the Advanced Photon Source, a facility operated for DOE Office of Science by Argonne National Laboratory, was supported under contract no. DE-AC02-06CH11357. We also thank the Microchemistry and Proteomics Laboratory and the Oligo Synthesis and DNA Sequencing facilities at Genentech for their technical support. Atomic coordinates and structure factors have been deposited in the Protein DataBank with accession code 3T6P. Reagents are available from Genentech subject to a material transfer agreement.
View Abstract

Navigate This Article