Both positional and chemical variables control in vitro proteolytic cleavage of a presenilin ortholog

Mechanistic details of intramembrane aspartyl protease (IAP) chemistry, which is central to many biological and pathogenic processes, remain largely obscure. Here, we investigated the in vitro kinetics of a microbial intramembrane aspartyl protease (mIAP) fortuitously acting on the renin substrate angiotensinogen and the C-terminal transmembrane segment of amyloid precursor protein (C100), which is cleaved by the presenilin subunit of γ-secretase, an Alzheimer disease (AD)-associated IAP. mIAP variants with substitutions in active-site and putative substrate-gating residues generally exhibit impaired, but not abolished, activity toward angiotensinogen and retain the predominant cleavage site (His–Thr). The aromatic ring, but not the hydroxyl substituent, within Tyr of the catalytic Tyr–Asp (YD) motif plays a catalytic role, and the hydrolysis reaction incorporates bulk water as in soluble aspartyl proteases. mIAP hydrolyzes the transmembrane region of C100 at two major presenilin cleavage sites, one corresponding to the AD-associated Aβ42 peptide (Ala–Thr) and the other to the non-pathogenic Aβ48 (Thr–Leu). For the former site, we observed more favorable kinetics in lipid bilayer–mimicking bicelles than in detergent solution, indicating that substrate–lipid and substrate–enzyme interactions both contribute to catalytic rates. High-resolution MS analyses across four substrates support a preference for threonine at the scissile bond. However, results from threonine-scanning mutagenesis of angiotensinogen demonstrate a competing positional preference for cleavage. Our results indicate that IAP cleavage is controlled by both positional and chemical factors, opening up new avenues for selective IAP inhibition for therapeutic interventions.

Intramembrane proteases (IPs) 4 cleave within a transmembrane (TM) helix of membrane-bound substrates to release cytoplasmic or extracellular proteins/peptides, which in turn translocate to different regions of the cell, where they elicit their corresponding biological response related to, for example, cell differentiation, development, and metabolism (1). Despite their broad biomedical reach, basic questions surrounding the structure of the active enzyme, how substrates are presented, and how hydrolysis chemistry occurs in an active site sequestered within the hydrophobic lipid membrane remain active areas of research.
The least biochemically understood IP is the intramembrane aspartyl protease (IAP) enzyme class, one of just three bona fide IP types that hydrolyze substrates within the hydrophobic lipid environment by using different catalytic nucleophiles (2). IAPs employ membrane-embedded aspartate residues found in the consensus sequence motif YD . . . GXGD (where X is any amino acid, typically Leu or Met) to catalyze peptide hydrolysis in the lipid membrane. The eukaryotic IAP signal peptide peptidase (SPP) acts on leader sequences found in nascent polypeptides (3,4). N-terminal signal sequences are first cleaved in the endoplasmic reticulum from a maturely folded protein by the soluble enzyme signal peptidase. The membrane-embedded signal peptide is then further processed by SPP to liberate shorter remnants from the membrane, which become signaling peptides. The known biological substrates of human SPP include the leader peptides of proteins involved in innate immunity (5), inflammatory response (6), and hepatitis C viral replication (7). SPP also shares key sequence similarity with presenilin, the catalytic component of the ␥-secretase complex involved in the production of amyloid-␤ (A␤) peptides implicated in Alzheimer disease (8). Presenilin and SPP both contain membrane-embedded active-site motifs (Fig. 1A), can cleave the same substrates, and are inhibited by the same active sitedirected small molecules (9). These structural and catalytic similarities extend to IAP orthologs across the domains of life (10 -12).
Despite intense work in the area, how IPs distinguish substrates from non-substrates and thus avoid deactivating a significant percentage of the proteome is unclear (13). There is no consensus motif across the ϳ90 different substrates reported for ␥-secretase or the ϳ30 for SPP (14,15). Substrate sequences exhibit high variability in amino acid sequence and length (16,17). Early studies suggested a requirement for helix-breaking residues within the TM substrate at or around the cleavage site. Such residues were proposed to destabilize the TM helix substrate and thus serve as a driving force for exposure of the scissile bond to the protease (18 -20). However, canonically helical residues (e.g. Leu) have been reported at the cut site of IAP substrates (10,11). In such cases, access to the scissile bond should be restricted, leaving the driving force for hydrolysis ambiguous. Conversely, residues other than aspartate within the IAP enzyme that affect catalysis (kinetics, cleavage specificity, substrate, and water entry) are largely obscure. For example, the specific roles of Tyr and Gly within the highly conserved YD . . . GXGD motif remain to be clarified. Finally, recent structures (21,22) do not reveal the site of substrate or water entry. The TM helices proximal to the proposed C-terminal substrate-gating motif PAL in presenilin (Fig. 1A) (23)(24)(25) need to undergo a substantial conformational change to enable entry of an exogenous TM substrate to the active site, a motion that is chemically incompatible with the potential presence of water in the active site.
Here, we compare in vitro kinetic parameters and cleavage sites of a microbial IAP ortholog (mIAP) from Methanoculleus marisnigri toward two substrates, one fortuitous (angiotensinogen cleaved by renin, a soluble aspartyl protease) (26) and one biological (isolated regions of amyloid precursor protein C-terminal 100 residues cleaved by ␥-secretase, C100). Our studies reveal a preference for a polar residue, Thr, at the scissile bond across all substrates. Using the angiotensinogen sequence, we examined biochemical properties of YD . . . GXGD and PAL motif variants. Kinetics are impaired except in the case of Y161F, and the predominant His-Thr cleavage site (26) is retained. mIAP cleaves the C100 sequence at Ala-Thr, the so-called ␥-secretase "␥-site" (27), leading to the formation of Alzheimer-associated amyloid-␤ species, A␤42 (28). Kinetics are more favorable in bicelles than in detergent solution, and both conditions lead to faster kinetics than the reaction with the angiotensinogen substrate. mIAP also cleaves the so-called ␥-secretase "⑀-site" corresponding to longer A␤ peptides (27), centered around Thr. However, Thr-scanning mutagenesis within the renin substrate reveals that cleavage is preferred at the original cut site, even absent a Thr side chain. Our system-atic biochemical study demonstrates an interplay between chemical and positional preferences for cleavage. Such results provide a foundation for elucidating substrate entry and other mechanistic details to better enable the development of selective inhibitors for diseases associated with IAP catalysis, such as Alzheimer disease and hepatitis C viral infection, without affecting processing of substrates for other biological processes.

Enzymatic analysis of mIAP mutants
To assess contributions of particular residues to catalysis ( Table 1 and Fig. 1A; see below), we employed a continuous kinetics Förster resonance energy transfer (FRET) peptide assay in combination with a discontinuous gel-based assay, methodology we reported previously (26). Kinetic assays with mIAP variants were conducted using the fortuitous renin FRET peptide (Ren390FRET, IHPFHLVIHT sequence flanked by Arg-Glu(5-[(2-aminoethyl)amino]naphthalene-1-sulfonate (EDANS fluorophore)-and -Lys([4-((4-(dimethylamino)phenyl)azo)benzoic acid (DABCYL quencher)-Arg). Due to the incompatibility of the FRET assay components with mass spectrometry analysis, we turned to a gel-based assay that uses a fusion protein (MRS WT ) in which the renin substrate sequence IHPFHLVIHT from angiotensinogen is flanked by E. coli maltose-binding protein (MBP) and yeast small ubiquitin-like modifier (SUMO). Cleavage profiles from a gel-based assay were analyzed by liquid chromatography tandem mass spectrometry (LC-MS) analysis (26). In aqueous solution, MRS WT adopts an elongated structure as visualized by the small-angle X-ray scattering (SAXS) envelope ( Fig. 1B and Fig. S1), rendering the desired 10-residue angiotensinogen sequence (IHPF-HLVIHT 10 ) accessible to detergents and mIAP. mIAP variants (Table 1) Y161A, Y161F, and G219A within the YD 162 . . . GXGD 220 motif and Q272A and L275F on the C-terminal helix near the presenilin PAL motif were purified and biophysically characterized as for wildtype (WT). Like WT, all mIAP mutants exhibit ␣-helical signatures (Fig. S2). Although the thermal melts did not yield a sigmoidal curve, both WT and mutants display similar curves, indicating that relative structural characteristics are retained upon introducing mutations.
Kinetic and cleavage site profiling reveal that mIAP is largely tolerant to mutation, with primary defects observed in catalytic efficiency and details of processivity. The only variant tested that exhibits Michaelis-Menten kinetic properties indistinguishable from WT is mIAP Y161F , representing the Tyr in the   Table 1. F, gel assay of mIAP mutants using MRS WT substrate. Negative control without enzyme is indicated by (Ϫ). The uncut substrate and cleavage products (indicated by an arrow) are detected via anti-MBP immunoblot. Molecular mass is indicated in kDa. G and H, cleavage profiles of MRS substrate were generated by Glu-C digestion of the N-terminal product, facilitated by a C-terminal Glu within MBP (black triangle). Major cleavage site is His 9 -Thr 10 (red triangle).
Representative LC-MS analysis of MRS WT cleavage sites are shown for all mutants tested. The relative abundance of each reporter peptide, compared with total peptide spectral matches of seven independent peptide products formed by proteolytic cleavage, is presented. Error bars, S.D. from LC-MS analytical replicates. All LC-MS data for mIAP variants are presented in Fig. S6 and Table S1. I, LC-MS spectrum of reporter peptide (z ϭ 3) displaying 350% more 18 O incorporation than 16 O. Left, the extracted ion current chromatogram for relative abundance of peptide (ALKDAQTNSIHPFHLVIH) with 16 O incorporation (middle) versus 18 O incorporation at the C terminus of the N-terminal product (right) from enzymatic cleavage. Approximately 11.8% of the peak at 681.7058 is part of the ion cluster for the 16 O-labeled peptide containing two 13 C isotopes.

IAP substrate specificity
detected reduced end product via immunoblot. Consistent with our measurements (Table 1), one additional study of the G384A presenilin variant reported 2-fold reduced V max compared with WT enzyme (33). Although rigidifying this region with the G219A substitution is detrimental to optimal orientation of Asp 220 for catalysis, it is noteworthy that the reaction still proceeds.
In terms of how substrate may gain entry into the active site, replacing the mIAP sequence AGL 275 with PAL 435 found in presenilin (23)(24)(25) reduces catalytic efficiency by ϳ80% compared with WT, indicating that PAL is not a favorable substitute ( Fig. 1D and Table 1). For mIAP L275F , corresponding to the presenilin L435F familial Alzheimer disease mutant (34 -36), catalytic efficiency is similar to WT, but this is due to the countering effect of a ϳ60% decrease in V max combined with 50% improvement in K m ( Fig. 1D and Table 1). More favorable substrate binding to the enzyme when the aromatic Phe is present appears to be unfavorable to some aspect of catalysis (see below). For mIAP Q272A , immediately N-terminal to AGL 275 , catalytic efficiency is half of WT due to decreased V max . K m remains near that of WT ( Fig. 1D and Table 1), so this residue appears not to modulate substrate binding and may instead play a cursory role in positioning other helices for catalysis.
Given the sensitivity of our FRET assay in detecting low levels of activity for the aforementioned variants, we tested the hypothesis that polar Asn in place of the negatively charged Asp residues supports catalysis. Each of the catalytic mutants, mIAP D162N , mIAP D220N , and the double mutant (DM) mIAP D162N/D220N , were found to be inactive (Fig. 1E). Thus, the specifically charged state(s) of the aspartic acid residues are critical for mIAP proteolysis.
Variants of mIAP exhibiting measurable activity in the FRET assay cleave MRS WT (Fig. 1F) with profiles that mirror WT mIAP, namely a predominant cut site at His 9 -Thr 10 ( Fig. 1 (G and H) and Table S1). However, subtle differences in overall cleavage profiles suggest that members of the YD . . . GXGD and PAL sequence motifs proximal to the catalytic aspartates ( Fig. 1A) influence processivity of IAP cleavage when presented with the renin substrate. For example, for WT-like mIAP Y161F , cleavage at Thr 10 -Met 11 , the second-most preferred site for WT and beyond the sequence harbored by Ren390FRET, is substantially diminished (Fig. 1G). In addition, mIAP L275F displayed higher cleavage site specificity than WT, with little to no product at positions His 5 -Leu 6 or Ile 8 -His 9 . Thus, tighter substrate binding by mIAP L275F , reflected in the lower K m value in the FRET assay, appears to be detrimental to processivity.

Role of bulk water in mIAP catalysis
In the case of soluble aspartyl proteases, water is activated for nucleophilic attack (37). To test for incorporation of bulk water in mIAP catalysis, we carried out the MRS WT cleavage reaction in the presence of 75% H 2 18 O. LC-MS analysis of the resulting cleavage products reveals robust 18 O incorporation at the C terminus of the N-terminal cleavage products, exceeding the observed 16 O incorporation in the expected proportion (Fig.  1I). To our knowledge, this isotope partitioning experiment is the first to provide direct evidence for a role for bulk water in an IAP.

mIAP cleavage of the presenilin C100 substrate in the "␥-site" region
Kinetic analysis of mIAP was expanded next to a second substrate, C100FRET, which contains 8 residues surrounding the "␥-site" from the Alzheimer-associated C100 substrate (GGV-VIATV flanked by N-methyl anthranilate (Nma) fluorophore and lysine-modified with 2-nitrophenol (DNP) quencher, followed by triple D-Arg), associated with A␤42 production. With this substrate, the initial velocity is sustained over the first 2 h of the incubation, and arbitrary fluorescence units were converted to product concentration with a calibration curve using free Nma (Fig. S3).
Reactions of mIAP and C100FRET were conducted in n-dodecyl-␤-D-maltoside (DDM) and bicelles composed of CHAPSO and 1,2-dimyristoyl-sn-glycero-3-phosphocholine (38). mIAP cleaves C100FRET with similar V max in detergent and in bicelles, but K m is notably more favorable in bicelles ( Fig.  2A and Table 2). As expected, catalytic mutants mIAP D162A , mIAP D220A , and DM mIAP D162A/D220A failed to cleave C100FRET ( Fig. 2A) and 1,3-di-(N-carboxybenzoyl-L-leucyl-Lleucyl)amino acetone ((ZLL) 2 ketone) is a competitive inhibitor of WT mIAP, with the K i value we reported before (26) (Fig.  2B). For reactions in DDM, V max is nearly 4-fold faster and K m is 20% lower for C100FRET compared with Renin390FRET in either DDM or bicelles, resulting in ϳ5-15-fold higher catalytic efficiency (Tables 1 and 2). The dramatic kinetic difference is consistent with the fact that C100FRET derives from a biological substrate, whereas Ren390FRET, although competent for catalysis, is not a biological IAP substrate. Overall, our K m is in the typical low micromolar range, similar to the soluble aspartyl protease renin (39), but mIAP is less efficient than renin in hydrolyzing substrates. Our data are consistent with other reports that IAP cleavage is sluggish, but it is challenging to further compare our study to analogous ones of ␥-secretase. For example, experiments using C100FRET to study ␥-secretase kinetics did not convert to an absolute scale (40,41). In other studies, where initial rates were evaluated via immunoblot quantitation after a set incubation time, our K m is either 10-fold weaker (33) or 30-fold tighter than that of ␥-secretase (42).
To identify the cleavage site within the C100FRET sequence, we adapted our gel assay by generating MCS10, an analogous 10-mer substrate in which the intervening sequence between MBP and SUMO in MRS WT was converted to 37 GGVVIA-TVIV 46 . MCS10 contains the octapeptide sequence of C100FRET plus IV 46 from C100. An increase in the N-terminal product was observed over the course of a 2-24-h reaction period in DDM ( Fig. 2C and Table S2). In parallel with results from C100FRET, the reaction between mIAP and MCS10 in bicelles was faster than in DDM, as product was readily visible within 30 min followed by a continued increase over the duration of the experiment (Fig. 2D). As expected, the reaction is dependent on mIAP concentration and is inhibited by (ZLL) 2 ketone, and catalytic mutants mIAP D162A , mIAP D220A , and DM mIAP D162A/D220A are unable to cleave MCS10 (Fig. 2,  E-G). LC-MS cleavage profiles for MCS10 reveal highly specific cut sites, Ala 42 -Thr 43 and Thr 43 -Val 44 , both in DDM and IAP substrate specificity bicelles (Fig. 2H). Isotope partitioning again demonstrates the use of bulk water in the mIAP reaction (Fig. S4). In sum, like results for mIAP cleavage of the MRS WT substrate, these data indicate the involvement of a site-adjacent Thr residue using a typical protease reaction mechanism.
The identified cleavage positions for mIAP reacted with MCS10 recapitulates the ␥-secretase ␥-site cut site, leading to Alzheimer-associated A␤42 and A␤43 production, but we did not observe even low-level background cleavage at Val-Ile corresponding to the ␥-secretase cleavage product A␤40 (27). On the basis of the MRS substrate (above and see below), the MBP fusion protein should not be restricting access to the A␤40 cut site. Our results differ from a previous study indicating that the same mIAP ortholog cleaves an alternately epitope-tagged C100 to release predominantly A␤38 and A␤40, with only a minor product of A␤42 (12). In that study, kinetics were not investigated, and detailed biochemical characterization of the substrate was not provided. Further work will be required to better assess why cleavage at sites generating A␤38 and A␤40 were not observed in our system.

IAP substrate specificity
and LC-MS analyses converge on cut sites at three consecutive peptide bonds. The predominant cut site is Thr 48 -Leu 49 (Fig. 3, C and D), which corresponds to the presenilin-associated product A␤48. Whereas the Thr residue of the MCS10 ␥-site and MCSGG ⑀-site are in register (Fig. 3C), MCSTV and MCSGG are offset (Fig. 3D). This result suggests that sequence preference for Thr may preside over positional preference relative to the membrane.

Probing substrate specificity by Thr scanning within the MRS substrate
To further address the interplay between chemical and positional preferences for mIAP cleavage, we introduced Thr substitutions adjacent to His residues in MRS WT (IHP 3 FHL 6 VIHT 10 ), namely replacing Pro 3 or Leu 6 with Thr and either retaining or removing the original His 9 -Thr 10 cut site (MRS P3T , MRS P3T/T10L , MRS L6T , MRS L6T/T10L ; Fig. 4A and Ref. 26). Because cleavage at His 5 -Leu 6 was of low abundance in MRS WT , we predicted that replacing Thr 10 in MRS WT with Leu (MRS T10L ) would disfavor this position as a cleavage site and serve as a control for positional preference (Fig. 4A). mIAP cleaves all five MRS Thr-scanning mutants, including MRS T10L where no His-Thr is present (Fig. 4 (B-E) and Table S2). The major cleavage position remains the same as for MRS WT and does not preferentially shift internally for MRS L6T , MRS L6T/T10L (Fig. 4D), MRS P3T , or MRS P3T/T10L (Fig. 4E). Thus, in contrast with C100 ␥and ⑀-site cleavages, positional preference presides over chemical preferences in this sequence.

Discussion
Intramembrane proteolysis must be highly regulated in the cell (13), yet contrary to intuition and knowledge of soluble proteases (44,45), data to date do not converge on recognition motifs for IPs within their TM substrate(s). In addition, cleavage of non-physiological substrates has been demonstrated for IAPs (10 -12, 46). Whereas helix-breaking residues within the TM segment were once thought to be important for cleavage (18,20,47), examples to the contrary appear in the literature (10,11). Confounding the puzzle for IAPs is that IAPs appear to cleave at multiple cut sites and/or trim TM helices to smaller segments that can eventually be released from the membrane (27, 48 -50). Systematic in vitro studies using purified model enzymes should help to clarify the physicochemical preferences of IAPs toward their substrates.
Overall, our results point to faster kinetics and a higher degree of specificity for IAP cleavage than previous studies have reported using other methods. Across the substrates examined, a chemical preference for Thr at the scissile bond emerged, regardless of enzyme or substrate variant used. None of the residues targeted for mutagenesis within mIAP substantially affected cleavage profiles, even where kinetic parameters that might be correlated with specificity, such as K m , were altered. Processivity, largely confined to residues adjacent to Thr on the substrate, was changed in some of the mIAP variants. In contrast to the fortuitous cleavage of the renin substrate, mIAP cleaved regions of the presenilin C100 substrate with a high degree of specificity around Thr. Finally, bulk water is incorporated into the product of the hydrolysis reaction, confirming that the chemical reaction is similar to that of soluble aspartyl proteases (37), but how water enters the active site, and the extent to which each Asp functions in a manner similar to those in soluble aspartyl proteases, remain open questions for future work.
Activity of mIAP toward C100FRET is 5-fold more efficient than the fortuitous Ren390FRET, reflected in faster V max and

IAP substrate specificity
lower K m . There is no difference in cleavage profiles of MCS10 in bicelles or DDM, indicating that the presentation of MCS10 to mIAP within DDM micelles and bicelles, and subsequent cleavage processes, are similar. However, the K m is significantly more favorable for the reaction in bicelles, which contains CHAPSO, a cholesterol mimic. Cholesterol is a modulator of both membranes (51,52) and C100 catalysis (53)(54)(55)(56)(57). Indeed, relevant regions of C100 are partitioned well inside the lipid bilayer and specifically bind cholesterol (53).
Our study delineates properties and/or roles for specific residues on the enzyme beyond the two catalytic Asps. The bulky phenyl ring, but not the hydroxyl group, in the YD 162 motif on TM helix 6 is the pertinent chemical property, perhaps to orient the substrate in the active site, to stabilize residues from another TM helix in the conformation that supports bound substrate. Reinforcing the need for an aromatic residue, but not a proton donor/acceptor, at this position, inspection of hundreds of presenilin-like protein sequences (PFAM PF01080 (58)) reveals that WD is a prevalent alternative to YD. Likewise, in the C-terminal TM helix 9, both mIAP A273P/G274A and mIAP L275F within the PAL motif exhibit slowed kinetics, but K m values are as favorable, or more favorable, than for WT. In our pseudo-first-order reaction setup, the K m reflects all of the binding events before catalysis, which is the slow step. Thus, because the substrate has higher affinity for the mutant enzyme than WT mIAP, impaired hydrolysis rates reflect a problem later in the process (e.g. enabling a conformational change or correctly orienting the substrate for optimal cleavage(s)).
The finding that mIAP generally exhibits a sequence preference for Thr at the scissile bond is consistent with numerous studies of ␥-secretase and C100 describing ␥and ⑀-site cleavages, but the specificity with which mIAP cleaved at these sites containing a Thr across several substrates was unexpected with respect to the general tendency IAPs exhibit toward pro-cessivity. Thr exhibits average helical propensity (59 -61), but intrahelical hydrogen bonds from the Thr side chain induces a more significant bend (62) and thus can displace remaining residues traversing the lipid membrane. Other residues capable of forming intrahelical hydrogen bonds include Ser and Cys (62). TM segments of SPP substrates (preprolactin, HLA-A*0301, and hepatitis C virus core protein) and ␥-secretase substrates (ErbB4, Notch-1, p75, Delta, Jagged, and STIM1 (5, 7, 63-68)) harbor Thr, Ser, and Cys residues, but most are not obviously in register with Thr residues in C100. As demonstrated by the Thr-scanning experiment, which failed to significantly shift to internal HT sites, positional preferences relative to the membrane are also relevant, perhaps due to substrate interactions with the membrane. Notably, in the case of Notch-1, the relative position of presenilin cleavage, VL (69), is similar to that of T 48 L from C100 (53) with respect to the membrane bilayer. Our data point to the possibility that different IAP substrates are governed by disparate physicochemical properties. Broadening such studies to other substrates and IAP family members should further clarify substrate preferences, ultimately leading to the ability to better target substrates selectively for therapeutic applications.

Expression and purification of enzyme variants and fusion substrates
Plasmids containing mIAP (WT and variants) were transformed into E. coli Rosetta 2 cells (Novagen). Large-scale growth of bacteria, membrane isolation, and protein purification steps were carried out as described previously (26), with the following modifications found to improve protein yield. Cells were induced with 0.2 mM isopropyl-␤-D-1-thiogalactopyranoside at A 600 nm ϭ 1.5, followed by growth overnight at 22°C. During nickel-affinity purification, weakly bound impurities were first removed with 5% Buffer B (50 mM Hepes (pH 7.5), 500 mM NaCl, 500 mM imidazole, 0.1% DDM) before the full gradient was applied. Protein purity after size-exclusion chromatography on Sephacryl S-300 (GE Healthcare) was assessed by SDS-PAGE (12% polyacrylamide) stained with Coomassie. Pure protein was concentrated and exchanged into phosphate-buffered saline (PBS: 10 mM sodium phosphate, pH 7.2, 150 mM NaCl) supplemented with 0.05% DDM using a 15-ml Amicon Ultra 50K MWCO concentrator (Millipore). The protein concentration was measured using a Thermo Scientific NanoDrop spectrophotometer. The calculated extinction coefficient, ⑀ ϭ 33,920 M Ϫ1 cm Ϫ1 , and molecular mass from the amino acid sequence were obtained by ExPASy Prot-Param (70).
Plasmids encoding substrates were transformed into E. coli BL21 DE3 (Novagen). A single colony was inoculated in 200 ml of Luria Bertani broth medium supplemented with 50 g ml Ϫ1 kanamycin except for MCSTV and MCSGG, which were supplemented with 60 g ml Ϫ1 ampicillin. Inoculum was grown overnight in an orbital shaker (225 rpm, 37°C). Large cultures (1 liter) were inoculated with 20 ml of overnight culture and grown until an A 600 nm ϭ 0.6 -0.8 was reached. After cooling the cultures to 18°C, protein expression was induced with an addition of 1 mM isopropyl-␤-D-1-thiogalactopyranoside. Cultures were grown overnight and harvested by centrifugation at 5,000 ϫ g, followed by flash-cooling with liquid nitrogen to store at Ϫ80°C. The purification of substrate proteins was carried out as described previously (26) except for MCS15, which was present in a mixture of aggregated and monomeric protein.

FRET assay and inhibitor studies
Assays with Ren390FRET (Anaspec) were conducted as described previously (26). For C100FRET (Millipore), lyophilized peptide was dissolved in DMSO to generate a stock solu-tion at 1.25 mM. A working stock solution (500 M) was prepared by diluting the original stock with 10 mM sodium phosphate, pH 7.2, 150 mM NaCl (PBS), 0.05% (w/v) DDM. The substrate was not reconstituted into bicelles due to phase incompatibility between liquid phases of bicelles and DMSO. The 500 M C100FRET working stock was sonicated (Branson Ultrasonics Corp.) at 50 -60 Hz, 117 V, 0.7 amp for 30 min before being diluted to 3-40 M solutions with PBS with 0.05% DDM and dispensed into black-bottomed, non-binding 96-well plates (Corning, Inc.). Plates were covered with optical adhesive film (Micro-Amp) and incubated at 37°C for 30 min, a process that resolved the initial high background from the C100FRET substrate. Freshly purified mIAP (0.5 M, in PBS, 0.05% DDM and reconstituted inside 5% (v/v) bicelles (26)) was then added to the substrate solution in the 96-well plate. The fluorescence readings were acquired every 2 min for 2.5 h at 37°C in a Synergy H4 plate reader (BioTek, ex ϭ 350 Ϯ 9 nm, em ϭ 440 Ϯ 9 nm). The initial velocity was determined from the fluorescence readings over the first 2 h. The background reading at each substrate concentration was subtracted, and the fluorescence reading was converted from arbitrary fluorescence units to concentration of product (nM) by using the standard calibration curve of free Nma (Fig. S3). The inner filter effect was not observed for the Nma-DNP donor-acceptor pair. GraphPad Prism 5 software was used for Michaelis-Menten kinetic analysis as described previously (26). Twelve replicates from independent batches of cell paste and enzyme purifications were used in data analysis.
For inhibitor studies, (ZLL) 2 ketone (Calbiochem) was dissolved in DMSO to prepare 1 mM stock. (ZLL) 2 ketone was diluted to a final 2-10-fold molar excess of the enzyme in the activity assay and preincubated with purified enzyme for 1 h at 37°C. C100FRET was then added, and the assay was conducted as above. Kinetics data of nine independent replicates were analyzed as described above.

Gel-based assay
Varying concentrations of mIAP (1-8 M), prepared in PBS with 0.05% DDM or 5% (v/v) bicelle, were incubated with MCS10 (5 M) at 37°C for 24 h. For time course experiments, 16 M mIAP was incubated with MCS10 (5 M), and the reaction was quenched at different time points by adding an equal volume of Laemmli SDS-PAGE sample-loading dye containing ␤-mercaptoethanol, followed by storage at Ϫ20°C. For experiments with (ZLL) 2 ketone, a range of inhibitor concentration (8 -32 M) was preincubated with 16 M IAP at 37°C before adding MCS10 (5 M). The cleavage reaction samples were separated by SDS-PAGE (12% (w/v) polyacrylamide) followed by transfer onto polyvinylidene difluoride membrane (MilliPore). Standard Western blot procedures were performed using MBP-probe mouse monoclonal IgG primary antibody (Santa Cruz Biotechnology, Inc.; 1:1000) and horseradish peroxidaseconjugated goat anti-mouse monoclonal IgG secondary antibody (KPL; 1:5000). The PVDF membrane was sprayed with HyGlo Quick Spray (Denville) and visualized on an Amersham Biosciences Imager 600 (GE Healthcare) except for the blot in Fig. 2C, which was exposed using autoradiography film and visualized on Konica SRX 101A film processor. The molecular IAP substrate specificity mass ruler was PageRuler Plus prestained protein ladder (Thermo Scientific).
For LC-MS analysis, the gel assay was set up as described above except that ice-cold acetone (6.5ϫ sample volume) was added to quench reaction at 24 h for samples containing MCS10 substrate and at 24 and 48 h for samples containing MRS substrate, followed by storage at Ϫ20°C. After decanting the acetone, samples were analyzed by SDS-PAGE. LC-MS analysis of samples was carried out as described previously (26) with the following modifications. An UltiMate TM 3000 RSLCnano System UPLC system (Dionex) was coupled to a Q Exactive Plus Mass Spectrometer (Thermo Scientific). The Mascot Search engine (version 2.6.0) was used with Proteome Discoverer version 2.1 (Thermo Scientific). Only peptide spectral matches with an expectation value of less than 0.01 ("High Confidence") were used.
For isotopic 18 O labeling, PBS with 0.05% DDM (75 l) was first lyophilized in a SpeedVac (Savant) for 30 min, after which 75 l of heavy water H 2 18 O (Cambridge Isotopes) was added. Purified mIAP (16 M) and either MRS WT or MCS10 (5 M) was then added to run the gel-based assay at 37°C for 24 h. LC-MS analysis of samples was carried out as described above.

Circular dichroism (CD)
CD spectra and thermal melts were acquired on a Jasco J-810 spectropolarimeter equipped with Neslab RTE 111 circulating water bath and a Jasco PTC-4245/15 temperature control system using a 0.1-cm cuvette. CD spectra of mIAP variants (5 M in 20 mM HEPES (pH 7.5), 250 mM NaCl, 0.05% DDM) were acquired from 300 to 200 nm at room temperature. Data were blank-subtracted and converted to molar ellipticity ϭ ( obs ϫ 10 6 )/(path length (mm) ϫ c ϫ n), where obs is the observed ellipticity (millidegrees) at wavelength ; c is the protein or peptide concentration (M); and n is the number of residues. CD thermal melts were performed with a 1°C min Ϫ1 increase in temperature from 5 to 90°C. Both CD spectra and melts were acquired with 15 averaged scans from 300 to 200 nm at a 200 nm min Ϫ1 scan rate. Data at 222 nm were blank-subtracted, converted to molar ellipticity, and plotted against temperature in GraphPad Prism 5.

SAXS and ab initio modeling
SAXS data for substrate solutions containing ϳ1 mg/ml of substrate in 20 mM Hepes (pH 7.5), 250 mM NaCl were collected at 12°C using a Rigaku BioSAXS-2000 instrument with 2D Kratky collimation and a rotating copper anode X-ray source ( ϭ 1.54187 Å). Each sample was exposed for a total of 3 h using multiple scans to confirm consistency between measurements and the absence of radiation damage. Matched buffers measured at an identical instrument configuration were used for background subtraction. Rigaku SAXSLab software was used for data reduction and yielded a 1D plot of the scattered intensities over the full range of momentum transfer 0.0104 Ͻ q Ͻ 0.6782 Å Ϫ1 (q ϭ 4 sin()/, 2 is the scattering angle, and is the X-ray wavelength). 3D reconstruction of the substrate structure from the SAXS data were performed using the ATSAS software suite (71). Briefly, the pair-distribution function P(r) (using 0.0104 Ͻ q Ͻ 0.3464 Å Ϫ1 ) was determined from an indirect Fourier transform of the scattering data for each substrate via GNOM. The DAMMIF tool was then used to rapidly generate 17 compact bead representations of the scattering particle. This initial collection of models was clustered using DAMCLUST, and the averaged model from the most populated cluster was used with DAMSTART to generate a fixed core (using a cut-off volume of one-half the excluded volume of the particle) for further refinement in DAMMIN. Ten refined models from DAMMIN (slow mode) were similarly clustered, and a representative model of the most populated cluster was compared with results from a rigid-body model reconstruction. CORAL was employed for rigid-body modeling, performing translation and rotation of high-resolution PDB structure fragments, and constructing dummy atom segments of defined lengths linking between domains and at the C terminus. MBP used for rigid-body modeling was excised from PDB 5CL1, chain A: residues 1-368. The structure for the SUMO domain was taken from the first of 20 conformers in the collection of NMR structures deposited under PDB 1L2N; the first 20 and C-terminal 5 amino acids are absent in the NMR structure. Dummy atom representations of the linker region containing the substrate target sequence, omitted residues of the SUMO domain, and C-terminal hexahistidine tag were all constructed using CORAL.