Caspase Inhibition by Baculovirus P35 Requires Interaction between the Reactive Site Loop and the β-Sheet Core*

Baculovirus P35 is a universal substrate-inhibitor of the death caspases. Stoichiometric inhibition by P35 is correlated with cleavage of its reactive site loop (RSL) and formation of a stable P35·caspase complex through a novel but undefined mechanism. The P35 crystal structure predicts that the RSL associates with the β-sheet core of P35 positioning the caspase cleavage site at the loop’s apex. Here we demonstrate that proper interaction between the RSL and the β-sheet core is critical for caspase inhibition, but not cleavage. Disruption of RSL interaction with the β-sheet by substituting hydrophobic residues of the RSL’s transverse helix α1 with destabilizing charged residues caused loss of caspase inhibition, without affecting P35 cleavage. Restabilization of the helix/sheet interaction by charge compensation from within the β-sheet partially restored anti-caspase potency. Mutational effects on P35 helix/sheet interactions were confirmed by measuring intermolecular helix/sheet association with the yeast two-hybrid system. Moreover, the identification of P35 oligomers in baculovirus-infected cells suggested that similar P35 interactions occur in vivo. These findings indicate that P35’s anti-caspase potency depends on a distinct conformation of the RSL which is required for events that promote stable, post-cleavage interactions and inhibition of the target caspase.

Baculovirus P35 is a universal substrate-inhibitor of the death caspases. Stoichiometric inhibition by P35 is correlated with cleavage of its reactive site loop (RSL) and formation of a stable P35⅐caspase complex through a novel but undefined mechanism. The P35 crystal structure predicts that the RSL associates with the ␤-sheet core of P35 positioning the caspase cleavage site at the loop's apex. Here we demonstrate that proper interaction between the RSL and the ␤-sheet core is critical for caspase inhibition, but not cleavage. Disruption of RSL interaction with the ␤-sheet by substituting hydrophobic residues of the RSL's transverse helix ␣1 with destabilizing charged residues caused loss of caspase inhibition, without affecting P35 cleavage. Restabilization of the helix/sheet interaction by charge compensation from within the ␤-sheet partially restored anti-caspase potency. Mutational effects on P35 helix/sheet interactions were confirmed by measuring intermolecular helix/sheet association with the yeast two-hybrid system.

Moreover, the identification of P35 oligomers in baculovirus-infected cells suggested that similar P35 interactions occur in vivo.
These findings indicate that P35's anti-caspase potency depends on a distinct conformation of the RSL which is required for events that promote stable, post-cleavage interactions and inhibition of the target caspase.
Baculovirus P35 is a potent inhibitor of the death proteases known as caspases. These aspartate-specific cysteinyl proteases are critical components of the cell death machinery and thus are targets in anti-apoptotic strategies (for reviews, see Refs. [1][2][3][4]. During baculovirus infection, P35 inhibits the activity of virus-induced cellular caspases, thereby preventing premature host cell death and promoting virus multiplication (5)(6)(7)(8). The anti-caspase activity of P35 accounts for its effectiveness in blocking programmed cell death in phylogenetically diverse organisms (9 -16) and forms the basis of its potential use in anti-apoptosis therapies (17,18). In vitro studies have demonstrated that P35 inhibits Group I, II, and III caspases (7, 13, 19 -21), despite the different substrate specificities of each protease group (22). P35 is a substrate-inhibitor in which cleav-age at Asp 87 and the formation of a stoichiometric complex with the target caspase are correlated with protease inhibition. However, the molecular mechanism of caspase inactivation and the basis for P35's potency among diverse caspases is unknown.
The 2.2-Å crystal structure of P35 (299 residues) has provided insight into its anti-caspase mechanism (23). The main core of P35 is an eight-stranded ␤-sheet which serves as a scaffold for adjacent functional domains (Fig. 1). The first half of P35 constitutes a ␤-barrel with two insertions. The largest insertion forms a solvent-exposed loop (residues 59 to 99), designated the reactive site loop (RSL), 1 which protrudes above the central ␤-sheet and contains the caspase recognition site ( 84 DQMD2G 88 ). The RSL begins with an amphipathic ␣-helix (␣1) that traverses the ␤-sheet. After ␣1, the RSL turns upward, placing the cleavage residue Asp 87 at the apex, and then extends downward to rejoin the ␤-barrel. The crystal structure predicts that the spatial orientation of the RSL is determined in part by its interactions with the notably flat ␤-sheet and the juxtaposed hairpin loop (Fig. 1).
Current evidence indicates that the RSL plays a critical role in P35 function. Substitution of RSL residue Asp 87 abrogated P35 cleavage and caused loss of caspase inhibition (7,13), thereby demonstrating the requirement of the P 1 residue for cleavage and anti-caspase activity. Two codon Ala-Ser insertions that altered the size and conformation of the RSL also disrupted P35 anti-apoptotic activity (7). Moreover, targeted disruption of the RSL helix ␣1 by substitution V71P caused loss of caspase-3 inhibition in vitro, but did not prevent cleavage by the target caspase (23). Thus, mutagenic distortion of the RSL converted P35 from a substrate-inhibitor to an efficient, but non-inhibitory substrate. Collectively, these findings suggested that the P35 RSL participates in both pre-and post-cleavage inhibitory events.
To determine the contribution of the RSL to anti-caspase activity and thereby define the molecular mechanism of P35, we investigated the functional significance of predicted interactions between the RSL and the ␤-sheet core. Our genetic and biochemical evidence indicated that RSL helix ␣1 interacts with the main ␤-sheet of P35. Disruption of this interaction abrogated caspase inhibition, but not P35 cleavage. Restoration of interaction increased anti-caspase potency. Thus, proper association of the RSL with the main core of P35 is required for caspase inhibition. We also report for the first time that RSL interaction with the P35 core can occur by intra-and intermolecular mechanisms and likely accounts for the existence of P35 oligomers in vivo. Collectively, our data support a model wherein the RSL adopts a distinct spatial configuration that promotes post-cleavage events that stabilize P35-caspase association and inactivate the target caspase.
Cells and Viruses-Spodoptera frugiperda cell line IPLB-SF21 (26) was propagated at 27°C in TC100 growth medium that contained 2.6% tryptose broth and 10% fetal bovine serum. For infections, cell monolayers were inoculated with the indicated plaque-forming units per cell. After 1 h, the inoculum was replaced with fresh growth medium, and the cells were incubated at 27°C.
AcMNPV (L1 strain) recombinant vP35 HA -His carrying an influenza hemagglutinin (HA) epitope-and C-terminal His 6 -tagged p35 inserted under control of the AcMNPV p10 promoter was generated by standard gene replacement procedures (6). A transplacement vector was constructed by inserting a 92-base pair BglII fragment containing the p10 minimal promoter into the BglII site downstream from the p35 promoter of plasmid p35KORF-hr5 ϩ ⌬TAAG (24). Subsequently, the HA epitope (YPYDVPDYA, inserted after P35 residue 205) and the Cterminal His 6 tag were inserted into p35 from plasmids described previously (7). The resultant transplacement vector was used to insert p35 HA -His into the p35 locus of recombinant virus v⌬35K/lacZ (6) by homologous recombination. The identity of recombinant vP35 HA -His was verified by restriction analyses of polymerase chain reaction-amplified DNA of plaque-purified virus. The virus vP35 (previously called v⌬35K/poly-p35) expressed p35 from the polyhedrin promoter and was described elsewhere (24).
Protein Purification-Recombinant C-terminal, His 6 -tagged human caspase 3 (23) and baculovirus P35-His 6 (wild-type or mutated) were overexpressed in Escherichia coli strain BL21(DE3) by using the pET 22(ϩ)b vector (Novagen). Proteins were purified by nickel (Ni 2ϩ ) affinity chromatography as described previously (7,23). Coomassie-stained polyacrylamide gel analysis indicated that the proteins, except V71K-P35-His 6 , were Ͼ95% pure. Protein concentrations were measured by using the Bio-Rad Protein Assay with bovine IgG (Sigma) as a protein standard. Impurities that affected V71K-P35-His 6 concentration determination were corrected by normalizing mutated P35 proteins to wildtype P35 by staining polyacrylamide gels with SYPRO Orange fluorimetric dye (Molecular Probes) and quantitation of protein bands with the Fluorimager 575 (Molecular Dynamics).
Caspase-3 Assays-For inhibition assays, increasing concentrations of wild-type or mutated P35-His 6 were mixed with purified caspase-3-His 6 (200 fmol) using reaction conditions described previously (7,23). After a 30-min incubation at room temperature, the tetrapeptide substrate Ac-DEVD-AMC (Peptides International) was added to 10 M and the accumulation of fluorescent product (excitation, 360 nm; emission, 465 nm) was monitored using a Biolumin 960 Kinetic Fluorescence/ Absorbance microplate reader (Molecular Dynamics). Values are the averages of duplicate or triplicate assays and are reported as the rate of product formation obtained from the linear portion of the reaction curves within the first 10% of substrate depletion. For cleavage assays, purified P35-His 6 (80 or 160 pmol) and caspase-3 (20 pmol) were mixed in 100 mM HEPES (pH 7.5), 0.1% CHAPS, and 10% sucrose and incubated at 27°C. Reaction products were analyzed by electrophoresis using SDS, 10 -20% polyacrylamide gradient gels followed by staining with colloidal Coomassie G-250 (Zaxis).
To determine the intracellular stability of mutated DBD-P35, transformed yeast were grown in selective medium, collected during log phase by centrifugation (4,000 ϫ g), and suspended in 2.5% 2-mercaptoethanol and 1% SDS. The cells were vortexed with 1 volume of glass beads (425-600 mm, Sigma) in the presence of protease inhibitors (Roche Molecular Biochemicals), boiled, and clarified by centrifugation (4,000 ϫ g). Equivalent amounts of total protein, as determined using the Bio-Rad Protein Assay, were subjected to immunoblot analysis with anti-P35 (␣-P35NF) serum (see below).
Immunoblot Analysis-Proteins were subjected to SDS-polyacrylamide gel electrophoresis and transferred to nylon membranes (Ni-troME, Micro Separations Inc.). The blots were blocked in 5% non-fat milk and incubated 1 h with a 1:10,000 dilution of ␣-P35NF serum (24) followed by 1 h with a 1:10,000 dilution of goat anti-rabbit immunoglobulin G (IgG) (Pierce) conjugated to alkaline phosphatase. P35 HA His was detected by using a 1:10,000 dilution of anti-HA serum (BabCo) followed by a 1:10,000 dilution of alkaline phosphatase-conjugated goat anti-mouse IgG (Jackson ImmunoResearch Laboratories, Inc.). Proteins were imaged by using colorimetric substrates as described previously (24).
Affinity Purification of Intracellular P35-SF21 cells were harvested 43 h after infection with recombinant baculoviruses vP35 HA -His and vP35 and ruptured by repetitive freeze-thaw cycles. Clarified lysates (16,000 ϫ g) were mixed with Ni 2ϩ -conjugated agarose beads (Novagen) in binding buffer (20 mM Tris, pH 7.9, 5 mM imidazole, 0.5 M NaCl) for 3 h at 16°C. After washing with binding buffer containing 30 mM imidazole, bound proteins were eluted with binding buffer containing 1 M imidazole and subjected to immunoblot analysis by using ␣-P35NF and ␣-HA sera.
For cross-linking, affinity purified P35 HA -His or chicken egg albumin (Sigma) were mixed with glutaraldehyde (Acros) to give final glutaraldehyde concentrations of 0.01, 0.05, and 0.1%. After 30 min at room temperature, the reactions were quenched with 350 mM Tris (pH 7.8). The products were subjected to electrophoresis on SDS, 4 -20% polyacrylamide gels and analyzed by immunoblotting with ␣-P35NF or by staining with Coomassie Brilliant Blue.
Image Processing-Stained gels and immunoblots were scanned at a resolution of 300 dots/inch using a Hewlett Packard ScanJetIIcx. The resulting files were printed from Adobe Photoshop 3.0 and Illustrator 7.0 using a Tektronics Phaser 450 dye sublimation printer.

Disruption of RSL Interactions Causes Loss of P35 Function-Stoichiometric inhibition of caspases by P35 includes
cleavage of the RSL at Asp 87 , the P 1 residue located at the apex of the loop. The transverse helix ␣1 (residues 63 to 75) is an essential component of the RSL, and may contribute to proper positioning of the cleavage site. The P35 crystal structure (23) suggests that ␣1 interacts with the ␤-sheet core (Fig. 1). The nonpolar residues on the bottom of ␣1 are accommodated by the hydrophobic environment provided by nonpolar residues of the ␤-sheet. Conversely, the charged or polar residues comprising the top side of ␣1 are solvent exposed ( Fig. 2A). To investigate the significance of the interaction between ␣1 and the ␤-sheet, we tested the effect of site-specific mutations within these domains on P35 anti-caspase activity.
To disrupt ␣1/␤-sheet association, hydrophobic residues Ile 67 and Val 71 comprising the underlying face of ␣1 ( Fig. 2A) were substituted with either lysine or tyrosine. The anti-apoptotic activity of I67K-, I67Y-, or V71K-mutated P35 was first determined by marker rescue. In this assay, anti-apoptotic function is measured by the capacity of the mutated P35 to block apoptosis and thereby restore replication of a p35-deletion mutant virus in cultured, apoptosis-sensitive SF21 cells (6,7). Neither I67K-P35 nor V71K-P35 exhibited anti-apoptotic activity in vivo (Fig. 2B). In contrast, the function of I67Y-P35 was comparable to that of wild-type P35. Thus, substitution of either Ile 67 or Val 71 with a positively charged residue disrupted P35 function, whereas substitution with an aromatic residue had no effect. These data suggested that the hydrophobic interaction between ␣1 and the ␤-sheet is required for P35 function. Consistent with this interpretation, charged-to-alanine substitutions ( Fig. 2B) of ␣1 solvent-exposed residues had no effect on P35 function in vivo (7).
␣1 Helix Mutations Disrupt Caspase Inhibition, Not P35 Cleavage-To determine whether loss of P35 function was due to loss of caspase inhibition, we measured the capacity of I67K-P35 and V71K-P35 to inhibit purified recombinant human caspase-3 (CPP32). Each mutated P35 was purified as a Cterminal His 6 fusion protein after overexpression in E. coli and tested in dose-dependent caspase inhibition assays that used the tetrapeptide DEVD-amc as substrate (Fig. 3A). As expected (7,23), wild-type P35 stoichiometrically inhibited caspase-3, whereas D87A-P35 which lacks the requisite cleavage residue Asp 87 failed to affect caspase-3 activity at all concentrations. V71K-P35 was as ineffective as D87A-P35 in inhibiting caspase-3 (Fig. 3A). Likewise, I67K-P35 had significantly reduced (Ͼ10-fold) anti-caspase activity when compared with wild-type P35. In contrast, caspase-3 inhibition by I67Y-P35 was comparable to that of wild-type P35, a finding consistent with its normal anti-apoptotic function in vivo. Thus, the lossof-function phenotype of P35 carrying the ␣1 lysine substitutions was correlated with loss of anti-caspase activity.
To determine whether I67K-P35 and V71K-P35 were recognized as caspase substrates, we examined their susceptibility to cleavage by caspase-3. After mixing caspase-3 with a 4-fold molar excess of purified wild-type or mutated P35-His 6 , the reaction products were analyzed by SDS-polyacrylamide gel electrophoresis (Fig. 3B). With the exception of D87A-P35, all forms of P35 were cleaved efficiently. ␣1-Mutated I67K-and V71K-P35 were cleaved to completion, as judged by loss of full-length protein and accumulation of their 25-and 10-kDa cleavage fragments (Fig. 3B, lanes 5 and 9). In contrast, wildtype and I67Y-P35 cleavage was limited as a direct result of caspase inhibition (lanes 3 and 7). Time course analysis (Fig.  3C) indicated that I67K-P35 cleavage was rapid. I67K-P35 cleavage fragments were detected within seconds of caspase-3 addition; by 6 h Ͼ95% of the 4-fold excess I67K-P35 was cleaved. In comparison, only limited cleavage of wild-type P35 occurred during the same period (Fig. 3C). Cleavage site substitution D87A-P35 was a poor substrate since cleavage was not detected. The finding that both I67K-P35 and V71K-P35 were efficient substrates, but ineffective caspase inhibitors, suggested that the ␣1/␤-sheet association is required for P35 anti-caspase activity, but not cleavage.
␤-Sheet Charge Compensation Increases P35 Anti-caspase Potency-We predicted that electrostatic stabilization of the weakened interaction between mutated ␣1 and the ␤-sheet would increase the potency of I67K-P35. In the P35 crystal structure, ␤-sheet residue Ile 15 is proximal to ␣1 residue Ile 67 (Fig. 1). We therefore substituted Ile 15 with Asp to neutralize the nearby positive charge of I67K. Caspase-3 inhibition by double-mutated I15D:I67K-P35 was then measured by using in residues. B, anti-apoptotic function of ␣1 mutations. Wild-type (wt) residues and amino acid substitutions within ␣1 are shown; nonpolar wild-type residues are highlighted. P35 anti-apoptotic activity was determined by marker rescue assay in which replication of p35-deletion virus v⌬35K/lacZ is restored in proportion to the functionality of P35 acquired by integration of transfected plasmid. Rescued virus yields were determined by plaque assay using apoptosis-sensitive SF21 cells. Percent anti-apoptotic activity is reported as the ratio of non-apoptotic lacZ-expressing plaques produced by mutated P35 to that of wild-type P35. Values shown are the average Ϯ standard deviation of triplicate transfections. *, charged to alanine (ca) mutations within ␣1 were as previously reported (7).
vitro dose-dependent assays (Fig. 4A). I15D:I67K-P35 was a more potent inhibitor than I67K-P35 at all concentrations tested, but less potent than single-mutated I15D-P35. On the basis of multiple experiments, the IC 50 of I15D:I67K-P35 ranged from 3.0 to 5.5 nM, compared with 9.5 to 13 nM for I67K-P35. Thus, with respect to caspase inhibition, the two mutations were compensatory, not additive, when present in the same P35 molecule. Extended incubations (Fig. 4B) demonstrated that inhibition was achieved in less than 30 min, since anti-caspase activity of I67K-and I15D:I67K-P35 was constant over a 6-h period. Thus, each mutated P35 did not exhibit delayed caspase inhibition. The increased anti-caspase potency of I15D:I67K-P35 compared with that of I67K-P35 suggested that the negative charge of substitution I15D stabilized interaction with I67K-mutated ␣1. These data confirmed the necessity of proper interaction between ␣1 and the ␤-sheet core for caspase inhibition.
P35 Oligomerizes in the Yeast Two-hybrid Assay-Previous studies have indicated that the yeast two-hybrid system (31) can detect intramolecular associations through bimolecular interaction of two independently synthesized protein domains (32)(33)(34). We predicted that the RSL of P35 could associate with the ␤-sheet of a different P35 molecule and that this intermolecular interaction could therefore be investigated using the two-hybrid system. To this end, we fused full-length P35 (residues 1 to 299) to the C terminus of the Gal4 DNA-binding domain (DBD) and the Gal4 activation domain (AD) to generate DBD-P35 and AD-P35, respectively (Fig. 5A). Co-transformation of S. cerevisiae with plasmids encoding DBD-P35 and AD-P35 induced strong Gal4-dependent lacZ expression as determined by filter lift assay and thus suggested stable interaction between both proteins (Fig. 5C). Neither plasmid alone induced lacZ expression in yeast (data not shown). The DBD-P35 and AD-P35 interaction was confirmed by growth of cotransformants on medium lacking histidine (see below, Fig. 7). The specificity of P35-P35 interaction in yeast was verified by mating strain Y190 (MATa) containing DBD-P35 with strain Y187 (MAT␣) containing unrelated proteins SNF4, lamin, CDK2, or p53 fused to the Gal4 activation domain. The resulting diploids failed to express lacZ as determined by filter lift FIG. 3. Effects of ␣1 mutations on caspase inhibition by P35. A, in vitro assay of caspase-3 activity. Purified human caspase-3 (200 fmol) was incubated with increasing amounts of purified wild-type (wt) P35-His 6 or P35-His 6 containing the indicated mutations. After 30 min, residual protease activity was measured in fluorometric assays by using the tetrapeptide DEVD-amc as substrate. Plotted values are the averages Ϯ S.D. of multiple determinations and are expressed as a percentage of uninhibited caspase activity. A representative experiment of three trials is shown. B, in vitro cleavage of P35 by caspase-3. P35-His 6 (80 pmol) containing the indicated mutations was incubated with (ϩ) or without (Ϫ) caspase-3 (20 pmol) for 15 h. The reaction products were subjected to SDS-polyacrylamide electrophoresis and stained with Coomassie Brilliant Blue. The 25-kDa (*) and 10-kDa (*Ј) cleavage fragments of P35 are indicated on the left. The reduced purity of V71Kmutated P35-His 6 was due to low protein yields from E. coli. C, time course of I67K-P35 cleavage. Purified I67K-mutated P35-His 6 (160 pmol) was mixed with purified caspase-3 (20 pmol). Samples were removed immediately after mixing (time zero) or the indicated times (hours) and subjected to electrophoretic analysis as described in panel B. assays (data not shown). Collectively, these data indicated that specific P35 interactions can be readily detected by using the two-hybrid system.
Helix ␣1 and ␤-Sheet Mutations Disrupt P35 Interaction-To identify residues involved in P35 interaction and to assess the contribution of helix/sheet association to P35 oligomerization, we generated a series of mutations (insertions and substitu-tions) in DBD-P35 and tested their effects on interaction with AD-P35. Immunoblot analysis of yeast lysates containing the DBD-P35 mutations confirmed that each fusion protein was stably synthesized (Fig. 5B) and ruled out the possibility that effects on P35 interaction were due to loss of protein stability. The capacity of each mutated P35 to interact in the two-hybrid system was subsequently compared with its anti-apoptotic activity in vivo (Fig. 5C).
Within the P35 RSL, only those loss-of-function mutations within the underlying nonpolar face of ␣1 caused loss of P35 interaction (Fig. 6). Mutations I67K and V71K failed to interact with AD-P35 (Fig. 5C). Similarly, Ala-Ser insertion in74, which disrupted the amphipathicity of ␣1, caused loss of interaction and loss of function. Conversely, I67Y-P35 which was fully functional for caspase inhibition interacted normally with AD-P35. In addition, charged to Ala substitutions of residues comprising the solvent-exposed portion of ␣1 had no effect on P35-P35 interaction and were fully functional for apoptotic suppression (Fig. 5C and Fig. 6). Although the P 4 and P 1 cleavage site substitutions D84A and D87A caused loss of antiapoptotic function, both mutated forms of P35 interacted normally with AD-P35, as did all mutations of RSL residues other than those altering the hydrophobic face of ␣1.
Mutations within the ␤-sheet (I15D, ca17, ca26, in52, and in278) also disrupted P35 interaction and caused loss of function ( Fig. 5C and Fig. 6). Conversely, mutations in loops that connected various ␤-strands (in10, ca22, ca41, ca143, and in273) had no effect on two-hybrid interactions. In addition, charged to Ala mutations (ca112 and ca126) within helix ␣2 and ␣3 had no effect. The pattern of P35 interaction was identical in reciprocal experiments where mutations were introduced into AD-P35 and tested for interaction with wild-type DBD-P35 (data not shown). Collectively, these data suggested that the residues involved in intramolecular ␣1/␤-sheet association also participate in two-hybrid interactions.
Compensatory P35 Mutations I15D and I67K Interact-Since the ␤-sheet substitution I15D partially restored anti-caspase activity to ␣1-mutated I67K-P35 (Fig. 4), presumably by stabilizing ␣1/␤-sheet interaction, we predicted that I15D would also restore intermolecular association with I67K-P35. To test FIG. 5. Mutagenic effects on two-hybrid P35 interactions. A, Gal4-P35 fusions. DBD-P35 and AD-P35 contain P35 residues 1 to 299 fused to the DBD (residues 1 to 147) and AD (residues 768 to 881) of Gal4, respectively. B, stability of DBD-P35 mutations. Steady state levels of DBD-P35 containing wild-type (WT) P35 or the indicated P35 mutations were determined by immunoblot analysis of S. cerevisiae cell lysates using anti-P35 serum. C, correlation between P35 interaction and anti-apoptotic function. Two-hybrid assays were performed using plasmids encoding wild-type AD-P35 and DBD-P35 containing the indicated P35 mutations. Protein interaction was assayed by transferring yeast (strain Y190) co-transformants grown on selective medium (Trp Ϫ Leu Ϫ ) to filter paper and measuring lacZ expression by 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside (X-gal) staining. The reported data is based on two to five independent transformations, each yielding hundreds of colonies, in which unambiguous and reproducible blue color was scored as a positive interaction (ϩ), whereas little or no blue color was scored as loss of interaction (Ϫ). When assayed in liquid culture, positive (ϩ) and negative (Ϫ) interaction transformants typically produced from 150 to 200 units and 20 to 50 units of ␤-galactosidase, respectively. Anti-apoptotic function of the indicated P35 mutations was determined by marker rescue assay as described in the legend to Fig. 2. P35 mutations which yielded Ͼ65,000 nonapoptotic lacZ-expressing (blue) plaque forming units/ml were scored as functional (ϩ), whereas mutations that yielded Ͻ4,000 blue plaque forming/units/ml were scored as loss of function (Ϫ). Wild-type P35 routinely yielded 65,000 to 80,000 blue plaque forming units/ml. *, anti-apoptotic function determined in this report. Other data was as previously reported (7).

FIG. 6. Location of residues required for P35 RSL interactions.
The position of mutations evaluated by the two-hybrid assay (Fig. 5C) for interaction (E) or loss of interaction (q) are indicated within the three-dimensional monomeric structure of P35 (23). Mutations include Ala-Ser insertions (in), charged to Ala (ca) substitutions of multiple residues, and single residue substitutions which are designated by lowercase letters as listed in Fig. 5C. * , substitution of residue Ile 67 (position c) with Lys (I67K) or Tyr (I67Y) caused loss of interaction or had no effect on interaction, respectively. this possibility, we assayed for two-hybrid interaction between I15D-P35 and I67K-P35 by using filter lifts and growth of yeast transformants on histidine-deficient medium (Fig. 7A). Although I15D-mutated and I67K-mutated DBD-P35 failed to interact with wild-type AD-P35, interaction between I15D-mutated DBD-P35 and I67K-mutated AD-P35 was strong. This finding was confirmed in reciprocal experiments where interaction between I67K-mutated DBD-P35 and I15D-mutated AD-P35 was as strong as that between wild-type P35s (Fig. 7A). These data indicated that intermolecular ␣1/␤-sheet association can occur in which the RSL of one P35 molecule interacts with the ␤-sheet of another (Fig. 7B). Thus, intramolecular interactions of the RSL with the ␤-sheet core within a P35 monomer also mediate intermolecular association of separate P35 molecules.
P35 Oligomerizes in Vivo-The intermolecular interaction of P35 within the two-hybrid system suggested that P35 also multimerizes in vivo. To investigate this possibility, electrophoretically distinct forms of P35 were co-synthesized in cultured SF21 cells and tested for interaction. By using recombinant baculoviruses, wild-type (untagged) P35 and HA epitope-tagged P35 HA -His 6 were synthesized separately or together in these cells. Immunoblot analysis of total cell lysates with P35and HA-specific antisera demonstrated that both forms of P35 were produced and were readily distinguished (Fig. 8). Metal (Ni 2ϩ ) affinity chromatography of cell extracts containing only P35 HA -His readily isolated this single protein (lane 10). Affinity purification of extracts containing cosynthesized P35 HA -His 6 and untagged P35 isolated both proteins (lane 12). Untagged P35 was not detected upon affinity purification of extracts containing untagged P35 alone (lane 11). Thus, co-purification of untagged P35 required the presence P35 HA -His. Stained gels failed to detect proteins other than uncleaved P35 HA -His and untagged P35, arguing against the presence of bridging molecules such as caspases (data not shown). Thus, the association of P35 HA -His 6 and untagged P35 was due to direct P35-P35 interaction.
To verify P35 oligomerization, metal affinity-purified P35 HA -His from SF21 cells was cross-linked with glutaraldehyde. Immunoblot analysis demonstrated a loss of P35 with a mass of 35 kDa and an increase in P35 multimers (Fig. 9, lanes 1-4). Under these conditions, purified monomeric ovalbumin failed to cross-link as expected (Fig. 9, lanes 5-8). Thus, the ability to cross-link purified P35 complexes confirmed in vivo P35 oligomerization. In the absence of cross-linker, a P35-specific protein with a size (70 kDa) expected of dimeric P35 was routinely detected by immunoblot analysis (Fig. 9, lane 1). The stability of the dimer-like P35 was unaffected by reducing agents, including dithiothreitol or 2-mercaptoethanol (data not shown). Thus, the detection of oligomeric P35 even after detergent treatment (1% SDS) suggested that hydrophobic interactions contribute to P35 oligomerization, a finding consistent with the hydrophobic association of helix ␣1 with the ␤-sheet core of a different P35 molecule. DISCUSSION P35 stoichiometrically inhibits caspases through a multistep mechanism involving slow binding, cleavage, and stable association with the target protease (7,19,20,23). Cleavage of the P35 RSL is correlated with formation of a stable complex with the target caspase, suggesting that cleavage is required for caspase inhibition. However, since cleavage is not sufficient for inhibition, additional undefined events are required (23). Our studies here demonstrate that regions or domains outside the P35 RSL cleavage site contribute to these events, either by direct participation or through the stabilization of required interactions between the RSL and the main core of P35.
Interactions between the RSL and P35 ␤-Sheet Core-The P35 crystal structure predicted that transverse helix ␣1, com-FIG. 7. Restoration of P35 ␣1/␤-sheet interactions. A, two-hybrid assays. Plasmids containing the indicated ␣1 and ␤-sheet mutations in DBD-P35 and AD-P35, respectively, were co-transformed into S. cerevisiae strain Y190. After growth on Trp Ϫ Leu Ϫ medium, recovered colonies were streaked onto His Ϫ Trp Ϫ Leu Ϫ plates. Growth on histidine-deficient medium indicated P35-P35 interaction since his3 expression is Gal4 responsive. Interaction was independently assessed by filter lift assays of hundreds of colonies from three independent transformations. Positive (ϩ) or negative (Ϫ) interaction was scored by unambiguous staining with 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside (X-gal). B, model of the P35 swap-mer. Within the swap-mer (top), the RSL of one molecule associates with the ␤-sheet of another through the same interactions that occur intramolecularly between ␣1 and the ␤-sheet of a P35 monomer (bottom) .   FIG. 8. In vivo P35 oligomerization. Cultured SF21 cells were infected with recombinant baculoviruses vP35 HA -His and vP35 (Ͼ10 plaque forming units/cell) which produced HA epitope-tagged P35 HA -His 6 or untagged wild-type P35, respectively. Freeze-thaw cell lysates prepared 43 h after infection were subjected to Ni 2ϩ affinity chromatography. Samples of total protein prior to purification (lanes 1-8) or affinity purified protein (lanes 9 -12) were subjected to immunoblot analysis by using HA-specific (anti-HA) or P35-specific (anti-P35) antiserum as indicated. Molecular mass standards (size in kilodaltons) are indicated on the left.
prising the base of the RSL, interacted with the ␤-sheet core (Fig. 1). Substitution of hydrophobic residues (Ile 67 and Val 71 ) within the underlying face of ␣1 with a positively charged lysine caused loss of P35 anti-apoptotic function in vivo (Fig.  2B) and reduced or eliminated P35 anti-caspase activity in vitro (Fig. 3A). The decreased anti-caspase potency of I67Kand V71K-mutated P35 was not due to reduced caspase recognition since both proteins were cleaved efficiently in vitro. These findings suggested that the charged lysine destabilized interactions between the RSL and the ␤-sheet, which in turn caused loss of caspase inhibition without affecting P35 cleavage.
Disruption of RSL interactions by these mutations was verified by using the yeast two-hybrid assay, which demonstrated bimolecular association of the RSL with the ␤-sheet core of P35. Although wild-type P35 fusion proteins interacted strongly with each other in this system, mutations I67K and V71K within ␣1 disrupted this interaction (Figs. 5 and 7). Likewise, substitution I15D within the ␤-sheet near ␣1 residue Ile 67 also caused loss of interaction with wild-type P35. However, I67Kmutated P35 interacted strongly with I15D-mutated P35 (Fig.  7). Thus, intermolecular interaction was restored by the introduction of a compensating charge at proximal positions within either ␣1 or the ␤-sheet. Reassociation of the RSL and the ␤-sheet core by charge compensation within the same P35 molecule increased anti-caspase efficacy, as demonstrated by the increased anti-caspase potency of doubly mutated I15D: I67K-P35 compared with singly mutated I67K-P35 (Fig. 4). We concluded that intramolecular interactions between the RSL and the ␤-sheet core are necessary for P35 anti-caspase activity. Supporting this conclusion, all mutations within helix ␣1 or the ␤-sheet that caused loss of interaction in the yeast twohybrid system were incapable of in vivo suppression of apoptosis (Fig. 5C).
Role of the RSL in P35 Anti-caspase Activity-The requirement for proper RSL interactions with the P35 core indicated that besides supplying the caspase recognition and cleavage site ( 84 DQMD2G 88 ), the RSL has an additional function(s) in caspase inhibition. Disruption of RSL interaction allowed postcleavage release of P35 without inactivating the target caspase, as demonstrated, for example, by the time-dependent turnover of I67K-mutated P35 (Fig. 3C). This behavior is analogous to that of V71P-mutated P35 in which mutagenic distortion of ␣1 abrogated the ability to complex stably with caspase-3 but did not affect P35 cleavage (23). These data suggest that ␣1 anchors the RSL to the P35 core, thereby confining the RSL to a specific conformation or position that is required for stabilization of P35-caspase interaction. Consistent with the critical nature of RSL conformation, two codon Ala-Ser insertions that altered the size or orientation of the RSL caused loss of P35 function (7). In contrast, P35 anti-apoptotic activity was not affected by charged-to-alanine substitutions of solvent-exposed residues of the RSL (Arg 64 , Asp 65 , Arg 66 , Lys 70 , Asp 72 , Glu 73 , His 90 , Asp 91 , Lys 97 , Asp 98 , and Glu 99 ), excluding the P 4 to P 1 cleavage residues. Thus, conformation and caspase recognition residues are paramount to RSL function.
Current evidence is consistent with a model in which the RSL participates in a multistep mechanism involving a cleavage-triggered event that stabilizes P35 association with the target caspase. If P35 is analogous to the serpins (reviewed by Ref. 35), this undefined event may be a conformational change that includes a structural rearrangement of the RSL to increase P35-caspase contacts. Proper interaction of the RSL with the main ␤-core of P35 may store the energy necessary to potentiate such a conformational change. Alternatively, the RSL may function to properly orient the target caspase for post-cleavage interactions with domains outside the RSL. In a different model, the RSL may restrict the caspase recognition residues ( 84 DQMD2G 88 ) to a specific conformation which when bound to the active site is sufficient for inhibition. Preliminary experiments have suggested that exogenously cleaved P35 is not inhibitory, consistent with the requirement for a distinct RSL precleavage conformation in P35 anti-caspase function. 2 Role of the RSL in P35 Oligomerization-Although the crystal structure (23) indicated that P35 can exist as a monomer, our study here demonstrates that P35 can also form oligomers. Upon affinity purification from infected insect cells, P35 was detected in an oligomeric complex that was readily cross-linked (Figs. 8 and 9). It is unlikely that this in vivo P35 multimerization was mediated by an interacting caspase since only full-length, uncleaved P35 was detected in these complexes (Fig. 8). Our analyses using the yeast two-hybrid system demonstrated that P35 oligomerization can occur through direct P35-P35 interaction in which the RSL of one molecule interacts with the ␤-sheet core of another. This RSL swapping generates a novel dimeric structure which we have designated the "swapmer" (Fig. 7B). It is unclear what fraction of in vivo P35 oligomers consist of swap-mers. In insect cells, I67K-and V71Kmutated P35s were poorly synthesized, complicating efforts to test interaction with I15D-mutated P35 in swap-mers. It is noteworthy that when independently synthesized P35 molecules were mixed in vitro, hetero-oligomerization was not detected. 2 Thus, dissociation and reassociation of P35 multimers may be unfavorable, a property consistent with strong hydrophobic interactions occurring within the swap-mer.
Oligomeric P35 may have multiple functions in vivo. Since caspases possess two catalytic sites that are independently inhibited by P35 (20,(37)(38)(39)(40), oligomers may increase the local concentration of P35 and thereby accelerate inhibition at both active sites. Due to structural constraints, it seems unlikely that the P35 swap-mer or other dimeric forms could interact simultaneously with both active sites of a single caspase. Nonetheless, simultaneous interaction of a P35 swap-mer with the active site of separate caspases is possible, raising the intriguing possibility that P35 is a multivalent protease inhibitor. P35 multimerization may also contribute to in vivo protein stability. Consistent with this role, mutated P35s that failed to interact in the two-hybrid assay (I67K, V71K, V71P, I15D, ca17, and ca26) were stably produced as fusion proteins in yeast, but were poorly produced in E. coli and cultured insect cells. 2 S. J. Zoog, unpublished data.
Truncation p35 1-76 dominantly interferes with the capacity of wild-type p35 to block apoptosis (8,41). It was originally speculated that hetero-oligomerization with P35 accounted for the observed dominant inhibition by p35  . However, by using the yeast two-hybrid assay, we found no evidence for interaction of P35 1-76 with wild-type P35 (data not shown). Moreover, mutations that caused loss of two-hybrid P35 interaction had no effect on p35 1-76 -mediated inhibition of wild-type p35 in vivo. Thus, despite the presence of ␣1 residues in P35 1-76 , it is unlikely that dominant inhibition by p35 1-76 is a consequence of interaction via the ␣1 helix. Nonetheless, investigation of other dominant-negative mutations of P35 may provide insight into the biological role of multimerization for P35 anti-apoptotic activity.