Modifications to a common phosphorylation network provide individualized control in caspases

Caspase-3 activation and function have been well-defined during programmed cell death, but caspase activity, at low levels, is also required for developmental processes such as lymphoid proliferation and erythroid differentiation. Post-translational modification of caspase-3 is one method used by cells to fine-tune activity below the threshold required for apoptosis, but the allosteric mechanism that reduces activity is unknown. Phosphorylation of caspase-3 at a conserved allosteric site by p38-MAPK (mitogen-activated protein kinase) promotes survival in human neutrophils, and the modification of the loop is thought to be a key regulator in many developmental processes. We utilized phylogenetic, structural, and biophysical studies to define the interaction networks that facilitate the allosteric mechanism in caspase-3. We show that, within the modified loop, Ser150 evolved with the apoptotic caspases, whereas Thr152 is a more recent evolutionary event in mammalian caspase-3. Substitutions at Ser150 result in a pH-dependent decrease in dimer stability, and localized changes in the modified loop propagate to the active site of the same protomer through a connecting surface helix. Likewise, a cluster of hydrophobic amino acids connects the conserved loop to the active site of the second protomer. The presence of Thr152 in the conserved loop introduces a “kill switch” in mammalian caspase-3, whereas the more ancient Ser150 reduces without abolishing enzyme activity. These data reveal how evolutionary changes in a conserved allosteric site result in a common pathway for lowering activity during development or a more recent cluster-specific switch to abolish activity.

Caspase-3 activation and function have been well-defined during programmed cell death, but caspase activity, at low levels, is also required for developmental processes such as lymphoid proliferation and erythroid differentiation. Post-translational modification of caspase-3 is one method used by cells to fine-tune activity below the threshold required for apoptosis, but the allosteric mechanism that reduces activity is unknown. Phosphorylation of caspase-3 at a conserved allosteric site by p38-MAPK (mitogen-activated protein kinase) promotes survival in human neutrophils, and the modification of the loop is thought to be a key regulator in many developmental processes. We utilized phylogenetic, structural, and biophysical studies to define the interaction networks that facilitate the allosteric mechanism in caspase-3. We show that, within the modified loop, Ser 150 evolved with the apoptotic caspases, whereas Thr 152 is a more recent evolutionary event in mammalian caspase-3. Substitutions at Ser 150 result in a pH-dependent decrease in dimer stability, and localized changes in the modified loop propagate to the active site of the same protomer through a connecting surface helix. Likewise, a cluster of hydrophobic amino acids connects the conserved loop to the active site of the second protomer. The presence of Thr 152 in the conserved loop introduces a "kill switch" in mammalian caspase-3, whereas the more ancient Ser 150 reduces without abolishing enzyme activity. These data reveal how evolutionary changes in a conserved allosteric site result in a common pathway for lowering activity during development or a more recent cluster-specific switch to abolish activity.
Caspases (cysteinyl-aspartate-specific proteases) are an ancient class of cysteinyl proteases that maintain cellular homeostasis by controlling the cell death program, apoptosis, when activated at high levels, and they are critical to cell development when activated at low levels (1,2). Caspases most likely evolved in metazoans from ancestral functions used in immune responses, and due to gene duplication events followed by neofunctionalization, human cells contain 11 caspases that are critical to the immune response or to apoptosis (3). Within the apoptotic caspases, two subclasses are further described based on entry into the apoptotic cascade. Initiator caspase-8, -9, and -10 are responsive to external or internal signals and initiate apoptosis by activating the downstream effector caspase-3, -6, and -7, where caspase-3 is the primary executioner of the cell death program (4). Activation of the presynthesized zymogen results in an overwhelming response to stimuli, leading to the dismantling of the cell without eliciting an immune response. Although the caspase degradome is a current topic of study, the role of caspases in apoptosis is reasonably well-understood (3,5). Hallmarks of apoptosis are largely observed through the caspase-3 cleavage of Rho effector protein (ROCK 1), apoptotic chromatin condensation inducer in the nucleus (acinus), nuclear lamin proteins, and caspase-activated DNase (CAD) as well as many other proteins, which leads to membrane blebbing, nuclear condensation, and DNA fragmentation (6,7).
Whereas caspase-3 is a well-known executioner of apoptosis, it is also a key protease in many non-apoptotic processes (8). For example, caspase-3 activity is important for remodeling the cytoplasm and in the development of the eye lens and inner ear (9,10). Caspase-3 is also involved in differentiation of many cell types, such as erythroblasts, keratinocytes, macrophages, lens epithelial cells, sperm cells, skeletal muscles, embryonic stem cells, osteoblasts, and placental trophoblasts (8,(11)(12)(13). Importantly, caspase-3 is required for terminal erythroid differentiation, starting from the mature BFU-E stage onward, and either inhibition or knockout of caspase-3 causes a decrease in proliferation due to halted cell-cycle progression through the G 2 /M phase and prevention of terminal erythroid maturation (14,15). Substrate cleavage data suggest that cells contain regulatory mechanisms to direct caspase-3 activity, although, at present, it is not clear how cells obtain sufficient activity for developmental responses yet maintain activity levels below the threshold required for apoptosis (8).
There are three general post-translational mechanisms that affect the population of active caspase-3 in cells: changes in zymogen maturation, active site-directed inhibitors, and allosteric inhibition through post-translational modifications or metal binding (Fig. 1A). Maturation of the caspase-3 zymogen, for example, is inhibited by protein kinase CK2, which phosphorylates Thr 174 and Ser 176 in the intersubunit linker (IL) 2 (see Fig. 1). Phosphorylation of the IL prevents initiator caspases from cleaving caspase-3 at Asp 175 , thus inhibiting maturation (16). In the mature caspase-3, X-linked inhibitor of apoptosis (XIAP) competitively inhibits the enzyme by binding to the active site and targeting the caspase for proteasomal degradation (17). In addition, the catalytic cysteine is sensitive to nitrosylation, although the modification may affect only a subset of caspase-3 localized to the mitochondria (18).
Caspase-3 is also allosterically regulated through post-translational modifications, including glutathionylation and phos-phorylation (19 -21). Although systematic global studies of caspase phosphorylation have not been done, several phosphorylation sites have been mapped on individual caspases (20,22,23). Based on the current data, phosphorylation sites of mature caspases can be characterized in two categories: those located near the active site that probably prevent substrate binding and those located some distance away from the active site that affect caspase activity allosterically. One allosteric site of interest, Ser 150 , is located in a loop at the C-terminal end of helix-3 near the dimer interface, and this serine residue (or threonine Figure 1. A, model of caspase-3 activation, regulation by XIAP binding, phosphorylation, and degradation. The caspase protomer is represented by one LS and one SS. In the zymogen, the subunits in the protomer are covalently connected by the IL. In the mature caspase, the IL is cleaved to yield the mature caspase protomer. Following cleavage, the IL provides two of the five active site loops, called L2 and L2Ј (see B and Fig. 2). Both the zymogen and the mature caspase-3 are considered a dimer of protomers, but the IL is cleaved in the mature enzyme. Phosphorylation of the zymogen inhibits maturation, whereas phosphorylation of the mature caspase inhibits enzyme activity. Binding of XIAP inhibits the mature caspase and leads to proteasomal degradation. B, secondary structural elements mapped onto the caspase-3 sequence. This figure was generated using Polyview-2D (67) and the structure of human caspase-3 (PDB entry 2J30). ␤-Strands 1-6 (green), ␣-helices 1-5 (red; H1-H5), and active site loops (blue) L1-L4 and L2Ј are indicated (see also Fig. 2). The arrow indicates the site of cleavage (Asp 175 ) of the IL to yield loops L2 and L2Ј of the mature protomer.

Common phosphorylation networks control caspase activity
in caspase-7) is conserved in all human caspases except caspase-10 and -14 (4). Helix-3 is an important regulator of effector caspase activity because fluctuations in the N-terminal region of the helix disrupt conserved water networks in caspase-3 and reposition the catalytic cysteine and histidine (24,25). In addition, the same region of helix-3 and the adjoining ␤-strands undergo a coil-to-helix transition in caspase-6, which also disrupts the catalytic residues by extending helix-3 (22).
Phosphorylation of Ser 150 by p38 MAPK was reported to promote survival in human neutrophils (26). Introducing a phospho-null mutant, S150A, into cells rendered Fas-induced apoptosis in neutrophils insensitive to p38 MAPK inhibition, suggesting that phosphorylation of Ser 150 inhibits catalytic activity (26). In addition to effector caspases, the initiator caspase-8 is phosphorylated by p38 MAPK at the same site (Ser-347), suggesting a common mechanism for controlling caspases by allosterically inhibiting the enzyme. The Ser 150 allosteric loop is ϳ33 Å from either active site of the caspase-3 dimer (intraprotomer or interprotomer) (Fig. 2), so it is not clear how the signal from phosphorylation of Ser 150 would propagate to either active site. Although the prevailing view is that phosphorylation of the loop is a common mechanism for inhibiting caspase activity, the previous in cellulo data for caspase-3 also suggested that phosphorylation at Ser 150 may decrease caspase stability, resulting in increased turnover in the cell (26).
We investigated the allosteric mechanism conferred by modifying the loop containing Ser 150 in caspase-3, and we show Figure 2. A, structure of caspase-3 (PDB entry 2J30). The LS and SS (see Fig. 1) fold into a single domain with a central six-stranded ␤-sheet core (␤1-␤4 contributed by the LS and ␤5 and ␤6 contributed by the SS; labeled in protomer 1) and five external ␣-helices (H1-H3 contributed by the LS and H4 and H5 contributed by the SS). Five loops comprise the active site of each protomer, where L1, L2, L3, and L4 are contributed by one protomer and L2Ј is contributed by the second protomer of the dimer (see Fig. 1). The loop bundle refers to interactions between L4, L2, and L2Ј, which stabilize the active site. B, interactions in the H3CL (helix-3 C-terminal loop, green) with helix-2 (H2, blue/brown) and the loop bundle (L2, red; L2Ј, cyan). Hydrogen bonds from Ser 150 and Thr 152 are shown as dashed lines. C, active-site loop L4 (brown) is stabilized, in part, by hydrogen bonds from Ser 249 , which contributes to substrate binding through interactions with Phe 250 and the aspartate in the P4 position of the substrate. Inhibitor refers to DEVD-chloromethylketone used during crystallization of the wildtype enzyme. Note that, for clarity, the active site in C is rotated 180°relative to the orientation in A.

Common phosphorylation networks control caspase activity
that, whereas there is no change in activity against small peptide substrates, the substitutions affect catalytic efficiency against a protein substrate. The effect is due to a pH-dependent decrease in dimer stability. Structural studies and molecular dynamics simulations show that localized changes in the loop propagate through helix-3 and affect the connected ␤-strand containing the catalytic histidine. Likewise, the modifications disrupt the nearby "loop bundle" that stabilizes the active site of the second protomer (Fig. 2), so modifications of the loop affect both active sites. We show that evolutionary changes within the loop of mammalian caspase-3 resulted in the introduction of a "kill switch" in the enzyme where loop modifications abolish, rather than diminish, activity. The evolutionary changes add an additional level of control to the allosteric mechanism in mammalian caspase-3 such that modifications in the common phosphorylation network propagate to both active sites. We note that, in terms of defining the caspase structural units, both the zymogen and the mature caspase-3 are considered a dimer of protomers. Each caspase protomer contains one large (LS) and one small (SS) subunit ( Fig. 1A) that fold into a single domain containing a six-stranded ␤-sheet core with five ␣-helices on the surface (Fig. 2). In the zymogen, the IL covalently connects the subunits in the protomer (Figs. 1 and 2), whereas the IL is cleaved in the mature enzyme. Cleavage of the IL (Fig. 1A, Maturation) results in active site loop rearrangements and formation of the mature enzyme active site (27,28).

Evolutionary conservation of helix-3 C-terminal loop
The level of conservation of the loop containing Ser 150 , herein called the helix-3 C-terminal loop (H3CL), is not known, aside from its conservation in human caspases. Because current models suggest that modification of Ser 150 represents a common mechanism for inhibiting caspases, we first examined the conservation of the H3CL to discern whether the studies of human caspases are representative of other species. To this end, we conducted a comprehensive phylogenetic analysis to probe the level of conservation and functional importance of the H3CL. Our analysis used 1,325 caspase sequences in the NCBI Protein Data bank, including caspase-1, -2, -3, -4, -6, -7, -8, -9, -10, and -14. The incorporated species ranged from fish to mammals and span ϳ500 million years of evolution (29). We calculated the amino acid usage frequency for the residue 149 -153 region, which contains Ser 150 and Thr 152 (caspase-3 numbering) ( Fig. 3 and Table S1). For position 150, the data show that a Ser/Thr was observed in Ͼ92% of effector caspases (caspase-3, -6, and -7) and in initiator caspase-8 and -9. In contrast, a lower frequency of Ser/Thr was observed in inflammatory caspases: caspase-4 (63.3%) and caspase-1 and -2 (40 and 44%, respectively). Very low frequencies (Ͻ5%) were observed in caspase-10 and -14, demonstrating that Ser/Thr is not conserved in the H3CL of those caspases (Table S1). We also examined the neighboring Thr 152 because, from previous reports, Thr 152 may also be phosphorylated along with Ser 150 (26). For position 152, the data show that hydrophobic amino acids (Ala, Val, Ile, Leu, and Gly) are present in Ͼ90% of caspase-3, -6, -7, -8, and -10, whereas lower frequencies are observed in caspase-9 (51%), and charged amino acids are common in caspase-1, -2, -4, and -14 ( Fig. 3 and Fig. S3A). Together, the data suggest that position 150 probably evolved as a phosphorylation site after the apoptotic caspases diverged from the inflammatory caspases. The data also show that position 152 is evolutionarily constrained in the apoptotic caspases to require a hydrophobic amino acid.
The analysis of amino acid frequency for the H3CL does not reveal when in evolutionary time the extant residues arose, so we generated a phylogenetic tree for each family member using the maximum likelihood method (30). Neofunctionalization is associated with site-specific rate variation over time and is more readily observed when a phylogenetic tree is parsed into clusters, as shown in Fig. 3 (31). For residue 150, the phylogenetic trees support the amino acid frequency data and show that the Ser/Thr residue evolved early in apoptotic caspases and only later in mammals of the inflammatory caspases (caspase-1 and -2 in particular) ( Fig. 3A and Fig. S3A). In contrast, Thr 152 evolved in early mammals, whereas prior evolutionary species utilized alanine or valine (ϳ98%) (Table S1). We note that mammalian caspase-3 is the only caspase to utilize a residue at this position with the potential to be phosphorylated while maintaining hydrophobicity in the native state, due to the presence of both the hydroxyl and methyl groups in the side chain. Together, the data suggest that phosphorylation of Ser 150 is an evolutionarily conserved mechanism in apoptotic caspases, whereas modification of Thr 152 may represent an allosteric control mechanism that only recently evolved in caspase-3 of mammals.

Phosphorylation of Ser 150 is predicted to modulate activity, whereas modification of Thr 152 is predicted to abolish activity
To examine the effects of phosphorylating the H3CL, we made several point mutations to mimic phosphorylation at Ser 150 and Thr 152 (Table 1). We introduced phosphomimetics (Asp and Glu) or phospho-null side chains (Ala). We also introduced a large polar side chain (Tyr) at position 150, and we increased hydrophobicity at position 152 by introducing valine. For each of the variants, we determined the steady-state parameters, k cat and K m , as well as the specificity constant, k cat /K m , and the results are shown in Table 1. Unlike the previous suggestions, our data show that modifications of Ser 150 (Asp, Glu, Ala, or Tyr) have no effect on enzyme activity, even when a large amino acid (Tyr) is introduced at the site (Table 1) (26). In all cases, the specificity constant was within about 3-fold that of wildtype caspase-3. We note, however, that in contrast to previous reports, our data were measured against tetrapeptide substrates using in vitro enzymatic assays rather than in cellulo apoptosis assays (26).
In contrast to the Ser 150 variants, the T152D substitution abolished activity (Table 1). To determine whether the decreased activity was due to the loss of the hydrophobic pocket or to the loss of H-bonds, we made two point mutants (T152A and T152V) to restore local hydrophobicity. In these mutants, the substitution of Thr 152 with alanine removes the hydrogen bonds observed with the hydroxyl group and decreases the number of hydrophobic groups in the side chain, whereas the isosteric valine simply replaces the hydroxyl group with a Common phosphorylation networks control caspase activity methyl group. The results show that both variants (T152A and T152V) retain wildtype levels of activity (Table 1).
In addition to the modification of the mature caspase-3, phosphorylation of Ser 150 was previously shown to occur in the zymogen before maturation (26). Procaspase-3 is not fully active because of the covalent connection between active site loops L2 and L2Ј (the intersubunit linker), so we wanted to determine whether the phosphorylation of the H3CL affected the activity of the zymogen. To this aim, we introduced the same mutations within the background of the full-length uncleavable zymogen (D9A,D28A,D275A, called D 3 A) and determined changes in enzymatic activity. As with the mature caspase-3, only the substitution of Thr 152 with aspartate abolished activity (Table 1), so the effects on the zymogen appear to be the same as those on the mature caspase-3.
We showed previously that the pro-domain increases the stability of caspase-3 (32). Amino acids 1-28 of the pro-domain are intrinsically disordered, but they bind to the protease domain in an extended conformation (33). To determine whether the pro-domain modulates the effects due to phosphorylation of Ser 150 or Thr 152 , we introduced each of the mutations into the background of mature caspase-3 (cleaved at  Fig. S3B. C, multiple-sequence alignment for each caspase using every known caspase sequence from NCBI using WebLogo (68). Numbers below each position refer to caspase-3 numbering. D, salt bridge between Glu 176 and Arg 271 in caspase-7 that spans across the dimer interface.

Common phosphorylation networks control caspase activity
Asp 175 ) with an uncleavable pro-domain (D9A,D28A, called D 2 A) (32). The data show that the presence of the pro-domain does not change the effects of phosphorylation at either Ser 150 or Thr 152 . Like the fully mature enzyme, only the substitution of Thr 152 with aspartate resulted in a decrease in activity (Table 1). Overall, the data show that phosphorylation of Ser 150 has no effect on caspase-3 activity when measured against peptide substrates. In contrast, the phosphorylation of Thr 152 is sufficient to abolish activity. The results were not dependent on whether the intersubunit linker was uncleaved (zymogen) or cleaved (mature) or whether the pro-domain was present.
Our phylogenetic analysis, described above, revealed that the hydrophobic pocket of the H3CL is highly conserved and that Thr 152 is a more recent evolutionary event in mammalian caspase-3 (Fig. 3). Based on the consensus sequences generated from the phylogenetic analysis, one observes that the effector caspases contain a polar group at position 149, which, in caspase-3, -6, and -7, H-bonds with the backbone of active site loop 2 across the dimer interface (Fig. 2). All caspases contain leucine at position 151, and the hydrophobic side chain anchors the loop to the protein core (Fig. 3C). Caspase-7 contains a charged group at position 153 and is more similar to the inflammatory caspases than to the effector or initiator caspases, which contain glycine at that position. In caspase-7, Glu 176 (equivalent to Gly 153 in caspase-3) forms a salt bridge with Arg 271 Ј (equivalent to Thr 245 in caspase-3), which is located at the C terminus of helix 5 and at the base of active-site loop 4 (Figs. 1B and 3 (C and D)). Thus, in caspase-7, the Glu 176 -Arg 271 Ј interaction provides a direct connection between the H3CL of one protomer and the active site of the second protomer.
Because Ser 150 is conserved in caspase-6 and -7, we made point mutations in caspase-6 (S150D) and caspase-7 (T173D) to determine the effects, if any, of phosphomimetics on enzyme activity. The data reveal that, like caspase-3, the mutations at this conserved site in caspase-6 and -7 also had no significant effect on the activity of these enzymes (Table 1). We note that a recent study by Hardy and co-workers (34) showed similar results for a comparable variant of caspase-7, that of T173E. In addition to introducing phosphomimetic mutations, we removed the salt bridge between Glu 176 and Arg 271 Ј in caspase-7 by replacing Glu 176 with glycine, either individually or in combination with the T173D phosphomimetic. Both variants showed specificity constants similar to wildtype caspase-7 (Table 1), so removing the salt bridge connecting the two protomers had no effect on activity. Overall, the results for caspase-3, -6, and -7 show that phosphorylation of Ser/Thr 150 , as assessed through phosphomimetic substitutions, had no effect on enzyme activity in our in vitro assays. Likewise, variations of the consensus sequence for the H3CL observed in caspase-7, which contains an additional salt bridge across the dimer interface, do not affect activity. Therefore, phosphorylation of the highly conserved Ser/Thr 150 in the H3CL does not decrease activity in vitro, but disrupting the hydrophobic cluster of the H3CL with the loop bundle abolishes activity.
Finally, to examine putative interaction networks that propagate allosteric signals from the conserved loop to one or both active sites, we used NetPhos 3.1 and identified two possible sites: Thr 245 , which is located at the C terminus of helix-5, and Ser 249 , which is located in active-site loop 4 ( Fig. 2C) (35). This region is also part of the loop bundle, on the side opposite from the H3CL (Fig. 2A). Interestingly, Thr 245 is in the same position as Arg 271 Ј in caspase-7, but the side chain, while solvent-exposed, is too short to interact across the dimer interface with the H3CL. In contrast, Ser 249 is located in active site loop 4 and forms several H-bonds with the C-terminal end of L4 to stabilize the loop conformation. The N-terminal region of L4, specifically the amide nitrogen of Phe 250 , H-bonds with the P4 side chain of the substrate, and the C-terminal end of L4 forms part of the hydrophobic S2 binding pocket (Fig. 2C). To examine the effects of phosphorylation of either site, we substituted each amino acid with aspartate (Table 1). Consistent with the loss of the salt bridge in caspase-7, the substitution of Thr 245 with aspartate had no effect on enzymatic activity. In contrast, substitution of Ser 249 with aspartate abolished activity, and this was true also for the double mutant, T245D,S249D (Table 1). Overall, the data show that phosphorylation of sites that flank the loop bundle (Thr 152 or Ser 249 ) abolishes activity and suggest that changes in the H3CL propagate across the dimer interface, through the loop bundle, which consists of active-site loops of the second protomer.

Modifications of the H3CL affect dimer stability
To examine whether modifications of the H3CL affect activity for a larger substrate, we examined the cleavage of caspase-7 by the caspase-3 variants. As part of the maturation process for caspase-7, the pro-domain is removed by caspase-3 at 20 DSVD 23 , and the protomer is cleaved in the intersubunit linker at 203 NDTD 206 to generate the large and small subunits.

Common phosphorylation networks control caspase activity
Cleavage of larger protein substrates can distinguish effects that may not be observed with small peptides that bind only to the active site (36,37). We also examined cleavage at two pH levels (7.5 and 6) because the pH of the cytoplasm is known to decrease during apoptosis (38). The data show that the pro-peptide of caspase-7 is efficiently removed by wildtype caspase-3 and that the intersubunit linker is readily cleaved, at pH 7.5 and pH 6 ( Fig. 4A, lanes 4 and 5). In contrast, the variants showed comparable cleavage of the pro-domain but not of the intersubunit linker. For example, the S150A variant cleaved the intersubunit linker comparably with wildtype at pH 7.5 but showed reduced cleavage at pH 6 ( Fig. 4A, lanes 6 and 7). The S150D variant showed reduced cleavage at both pH levels (Fig.  4A, lanes 8 and 9). Consistent with the activity assays described

Common phosphorylation networks control caspase activity
above for small peptides, the T152V variant efficiently cleaved both the pro-peptide and the intersubunit linker of caspase-7 (Fig. 4A, lanes 10 and 11), whereas the T152D variant showed greatly diminished activity (Fig. 4A, lanes 12 and 13).
Previous studies in cellulo suggested that phosphorylation of Ser 150 either reduced enzyme activity or decreased protein stability, and either mechanism could result in lower apoptosis (26). Our data show that changes in activity are observed in large protein substrates but not in tetrapeptide substrates, when Ser 150 is modified. To further examine the differential effects of pH on the variants, we measured the pH-dependent assembly of the caspase dimer. Our previous studies showed that the caspase-3 dimer dissociates to the two protomers below pH ϳ5.5, and the dimer reversibly assembles above pH ϳ4 (Fig. 4B) (32, 39 -41). In those studies, we utilized several biochemical and biophysical assays to monitor caspase-3 dimer dissociation versus pH. Changes in fluorescence emission, as monitored by the average emission wavelength (AEW), and size-exclusion chromatography showed a blue shift in fluorescence emission between pH 5.5 and pH 4 that corresponds to dissociation of the dimer to the two protomers (40,41). Below pH ϳ4, the subunits within the protomer unfold, which is accompanied by a red shift in fluorescence emission. In our previous studies, we confirmed that the dimer of wildtype caspase-3 dissociates between pH 5.5 and pH 4 by extensive pH experiments in which we monitored changes in fluorescence emission and circular dichroism as a function of urea over a broad pH range (41) as well as protein concentration and salt dependence (32). Finally, using a combination of fluorescence emission and circular dichroism combined with quenching by iodide, cesium, and acrylamide and limited proteolysis by trypsin and V8 proteases, we showed that the protomer remains well-folded at pH 4 and that the tryptophan residues in the protomer are not exposed to solvent (39,40,42). Together, our previous studies establish that the caspase-3 dimer dissociates between pH 5.5 and pH 4 and that dimer dissociation is accompanied by a blue shift in fluorescence emission (Fig. 4B). The minimum in AEW at pH ϳ4 and the red shift in AEW at pH Ͻ4 indicate the unfolding of the protomer.
To examine changes in the dimer of the H3CL variants, we first characterized the fluorescence emission of each protein.
Caspase-3 has two tryptophan residues, and both are located in the active site near the surface of the substrate-binding loop (L3). The tryptophan fluorescence emission, therefore, indicates possible conformational changes to the active site. The results show that, at pH 7.5, the AEW of wildtype caspase-3 is ϳ350 nm, as expected for tryptophans that are solvent-accessible (Fig. 4C). There was no significant difference in the AEW for the S150A/D/E or T152A/V variants, all of which showed AEW of ϳ350 nm (Fig. 4C). In contrast, the T152D variant demonstrated a significant blue shift in AEW to 346 nm (Fig.  4C). When unfolded in 8 M urea-containing buffer, the AEW of all proteins was ϳ350 nm, so the blue shift observed for the T152D variant shows that the active-site tryptophans are in a less solvent-exposed environment compared with those of the wildtype enzyme at pH 7.5 (Fig. 4D). As noted below, all proteins were dimeric at pH 7.5, so the altered fluorescence emis-sion for the T152D variant suggests that the mutation affects the conformation of the active-site tryptophans.
The activity observed for a caspase is dependent on dimerization because loops from each protomer comprise the loop bundle (see Fig. 2A), and one measure of dimer stability in caspases is the pH-dependent dissociation of the dimer (Fig. 4B) (32,41). To examine the effects of mutations in the H3CL, we incubated the phosphomutants in buffer between pH 3 and pH 7.5, as described previously, and we monitored enzyme activity after returning the protein to pH 7.5 (Fig. 4E) and allowing the sample to equilibrate (14 h minimum) (32). Relative to controls for each protein, which were incubated for equivalent times at pH 7.5, the major loss of activity occurred when protein was incubated between pH 6 and pH 4 before returning the protein to pH 7.5. Furthermore, the wildtype caspase-3 dimer reassembled reversibly when incubated at pH 6, as shown by the complete recovery of activity upon return to pH 7.5. However, in the Ser 150 and Thr 152 variants, only 65-85% activity was recovered in protein incubated at pH 6.
To further examine dissociation of the dimer and changes resulting from mutations in the H3CL, we determined the AEW for each protein over the pH range of 3-7.5. For wildtype caspase-3 (Fig. 4F), the data show one primary transition where the AEW for tryptophan fluorescence decreases between pH 6 and 4, with a midpoint for the transition of ϳ5. As described above, the fluorescence emission of the protomer is blueshifted compared with the dimer, so the major transition in AEW between pH 6 and pH 4 is due to dimer dissociation. The results were independent of whether we monitored only the tryptophans (Fig. 4F) or all aromatic residues (Fig. S1). The variants (S150A and S150D) demonstrated an increase in the transition midpoint to pH ϳ5. 2-5.4. Below pH 4, the protomers dissociate to the large and small subunits, which is accompanied by an increase in AEW due to exposure of the tryptophan residues. For the S150A and T152V variants, the pH-dependent minimum in AEW increased when compared with that of wildtype caspase-3 from pH 4 to pH Ͼ4.5. Together, the fluorescence data are consistent with a destabilized dimer and protomer in the H3CL variants (with the exception of T152V). Both the midpoint for dimer dissociation and the minimum in AEW are shifted to higher pH values. The lower recovery of activity versus pH for the mutants (Fig. 4E) may be due to the destabilized protomer, because protomer unfolding is not reversible. Finally, we note that all proteins were dimers at pH 7.5, as determined by size-exclusion chromatography (Fig. S2), so the blue shift in fluorescence emission and loss of activity observed for the T152D variant at pH 7.5 were not due to dimer dissociation.

High-resolution structures of H3CL variants reveal localized changes in hydrogen bonding and hydrophobic interactions
We examined changes in the H3CL due to mutations at Ser 150 or Thr 152 by X-ray crystallography, and each mutant was solved to 2.1 Å or higher resolution (Table S2). We note that the presence of the pro-domain (in the D 2 A variants) did not affect the changes caused by mutations at Ser 150 or Thr 152 . As expected, the pro-domain was disordered in the crystal. In general, the presence of the pro-domain appeared to increase the Common phosphorylation networks control caspase activity flexibility of the N terminus because several residues were disordered in the D 2 A variants that were ordered in the fully mature variants, namely residues 29 -34. As Asn 35 is the first residue to contact the dimer, the effects of the pro-domain are considered inconsequential to the conclusions described below.
The data show no gross structural changes in the protein resulting from the mutations at Ser 150 or Thr 152 , as the root mean square deviation compared with wildtype caspase-3 was generally Ͻ0.3 Å (Fig. 5). Not surprisingly, the two H-bonds contributed by the Ser 150 side chain were lost in the Ser 150 variants (Fig. 5, A-D). Although the hydroxyl of Ser 150 is partially buried, replacement with larger side chains (aspartate, glutamate, tyrosine) is accommodated by rotation of the side chain toward solvent. The S150D variant shows a net increase in H-bonds with helix-2 and -3, whereas S150A has a net loss of three H-bonds (Fig. 5, A and B). In addition, the larger side chains caused a displacement of 1-1.5 Å between the H3CL and the neighboring loop connecting helix-2 to the protein core (Fig. 5, C and D).
In the Thr 152 variants, removal of the hydroxyl group of Thr 152 and replacement of the side chain with alanine or valine results in disruption of the H-bond between Arg 149 and Glu 173 Ј in the loop bundle as well as loss of the H-bond with the amide nitrogen of Gly 145 (Fig. 5, E-G). In the case of Arg 149 , the side chain is observed in two rotamers, where one rotamer is rotated toward solvent. In addition, the shorter alanine side chain results in changes to the hydrophobic cluster of Ala 152 , Ile 172 Ј, and Ile 187 Ј, where the distance of the Ala 152 methyl group from that of Ile 187 increases by ϳ2 Å compared with that of Thr 152 . The changes in the hydrophobic cluster result in a partially disordered side chain for Glu 173 Ј in the loop bundle. The hydrophobic pocket is restored in the T152V mutant, although as noted above, the H-bonds from the Thr 152 hydroxyl are lost. Notably, L2Ј of the loop bundle is disordered beyond His 185 . In wildtype caspase-3, residues 179 -184 of one protomer interact with active site loop L4 and the S4 substrate-binding pocket of the second protomer, and those interactions are disrupted in the H3CL variants (43). We note that the residues of L2 and L2Ј that form the hydrophobic pocket between the H3CL and the loop bundle (Ile 172 Ј/Ile 187 Ј) are also highly conserved in effector caspases (Fig. S3B). Thus, changes at Thr 152 affect both the hydrophobic interactions and the H-bonding in the loop bundle as well as the flexibility of the termini of loops L2 and L2Ј.
We next characterized the structures of variants in the two sites that are thought to be part of the H3CL interaction network: Thr 245 and Ser 249 near active-site loop L4. The T245D variant, with or without the pro-domain (D 2 A,T245D), had no significant structural changes. Indeed, the H3CL contacts were the same as for wildtype caspase-3 at Ser 150 and Thr 152 . Closer to the mutation site, the Asp 245 side chain forms a new salt bridge with Arg 241 in helix 5, causing the Arg 241 side chain to Figure 5. A-G, hydrogen bonds and electrostatic interactions in the H3CL of WT (gray) superimposed to each caspase-3 mutant (green). Blue spheres, conserved waters as described under "Experimental procedures" (24). A, S150A; B, S150D; C, S150E; D, S150Y; E, T152V; F, T152A; G, T152D. For each panel, black dashes represent hydrogen bonds in WT caspase-3, and red dashes represent hydrogen bonds in the mutant.

Common phosphorylation networks control caspase activity
adopt another rotamer to form the interaction (Fig. S4). In wildtype caspase-3, Arg 241 H-bonds with the backbone carbonyl of Thr 270 Ј and with the side chain of Asn 35 Ј across the dimer interface, so these interactions are lost in the mutant, but the repositioned Arg 241 forms new interactions with the side chain of Asp 34 Ј.
We were unable to crystallize the S249D single mutant, but we were able to determine the structure of the double mutant, T245D,S249D. In this case, the changes noted above for T245D were also observed in the double mutant. Contrary to our expectation that active-site loop L4 would be disordered in the variant, the Asp 249 residue forms new H-bonds near the active site (Fig. S4).

Molecular dynamics simulations show increased fluctuations in active site loops
We conducted MD simulations for 50 ns on each variant to determine whether changes to the H3CL affected fluctuations in the active-site loops. To set a baseline with which to compare the H3CL variants, we performed three independent MD simulations on wildtype caspase-3 (dimer of PDB entry 2J30) and used the average data in further analyses (Figs. S5 and S6). A comparison of the average structures from the three MD simulations of wildtype caspase-3 with the crystal structure is shown in Fig. S5 (A and B). For wildtype caspase-3, the MD simulations show very little change in the substrate-binding pocket (active site loop L3) or in L2. In contrast, active site loops L1 and L4 show the most flexibility, with a root mean square fluctuation (RMSF) for each loop of ϳ3 Å (Fig. S6). In the crystal structure, L2Ј is bound near the S4-binding pocket and forms through-water H-bonds with L4 and the substrate. The simulations, however, show that the loop does not remain bound near the S4 pocket, but rather rotates toward the H3CL and L2 (Fig. S5B). As expected, the N terminus is flexible from Ser 29 to Asp 34 but has lower fluctuations at Asn 35 , where the N terminus contacts the protein core. In protomer B, the elbow loop (Ala 200 -Tyr 203 ) and turn above (called the 124-loop) are displaced compared with the crystal structure and show fluctuations of ϳ1.5-2 Å (Figs. S5 and S6). Together, the simulations show that there are few to no changes in the H3CL over the course of the simulations. In the three simulations of wildtype caspase-3, the protein core shows very similar fluctuations, and the largest variations among the three simulations occur at the chain termini (Fig. S6).
Representative data for the H3CL variants are shown in Fig.  S5 (C and D) as the average structures compared with that of wildtype caspase-3. The RMSF for each amino acid is presented in Figs. S7-S13 for all variants. The RMSF data are shown as the difference between the mutant and wildtype caspase-3 (⌬RMSF). In this case, values above zero indicate increased fluctuations in the mutant, whereas values below zero indicate decreased fluctuations in the mutant compared with wildtype caspase-3. The average structures for the variants were similar overall to that of wildtype caspase-3 in that the active-site loops L1 and L4 showed the largest fluctuations, as did the terminus of L2Ј. There were small fluctuations in the H3CL, but the mutations resulted in repositioning helix-3 as well as the loop connecting the surface strands ␤ 1 and ␤ 2 (the 124-loop), which showed increased fluctuations of ϳ2 Å in some cases (Fig. S5, C  and D). The increased motion in the surface loop was also observed in the ⌬RMSF analysis (Figs. S7-S13). In the Ser 150 and Thr 152 variants, there was little additional fluctuation in active site loops L1 or L4 compared with wildtype caspase-3. In other words, the ⌬RMSF values for those regions were generally Ͻ1 Å. The loop 4 mutants (T245D and T245D,S249D) showed increased mobility in the 124-loop, similar to the H3CL variants, but the double mutant also resulted in increased mobility in active site loop 4 compared with wildtype caspase-3 (Fig.  S13). In some cases (e.g. S150E), the displacement of 1-1.5 Å observed in the crystal structure between the H3CL and the neighboring helix-2 loop was observed to propagate along helix-2 and -3 to the short surface ␤-strand, called ␤ 1 -␤ 3 (Fig.  6A). The short surface strands not only contain the catalytic histidine (His 121 ), but they also contribute to the oxyanion hole as well as stabilize active site residues (28). The increased mobility in the 124-loop (connecting ␤ 1 and ␤ 2 ) resulted in transient interactions between Glu 123 and His 121 , where Glu 123 was observed to move to within ϳ3 Å of His 121 , from ϳ11 Å in wildtype caspase-3 (Fig. 6B). A similar displacement of helix-2 and -3 and ␤ 1 is observed in the T152D mutant (Fig. S5D). Our phylogenetic analysis of the 124-loop shows that Glu 123 is Figure 6. A, representative frame of a molecular dynamics simulation of S150E (green) compared with the X-ray crystal structure of wildtype caspase-3 (PDB code 2J30) showing movement of surface strands ␤ 1 and ␤ 2 toward the catalytic residue His 121 . Black dashes, initial distance between the side chain of Glu 123 and the catalytic histidine (ϳ11 Å); red dashes, distance between Glu 123 and His 121 in the mutant (ϳ3.5 Å). B, positions of the catalytic dyad (His 121 and Cys 163 ) and of Glu 123 shown as 200 frames from the 50-ns MD simulation.

Common phosphorylation networks control caspase activity
highly conserved in apoptotic caspases (Fig. S14). We note that changes to fluctuations in the hydrophobic cluster of the H3CL and active-site loop 2 were not observed in the MD simulations, but the chain termini of L2 and L2Ј are among the most mobile in the protein, so it is difficult to determine whether the mutations further increased mobility of the termini.

Discussion
The regulation of caspase activity is an essential determinant to cell survival, and post-translational modifications provide reversible mechanisms to control caspase activity below the threshold of activity required for cell death (28). Although poorly understood, the phosphorylation of caspase-3 in the C-terminal loop of helix-3 is of particular interest because it is thought to represent a common allosteric mechanism in all caspases (25,26). In addition, Thr 152 is part of a hydrophobic cluster of amino acids that stabilize the active site loops across the dimer interface.
Our data show that there are two allosteric networks in caspase-3 that facilitate signal propagation following phosphorylation of the H3CL. On one side of the loop, Ser 150 is conserved in most vertebrate species, and the residue evolved after the apoptotic caspases split from inflammatory caspases. The phylogenetic data support the hypothesis that phosphorylation at Ser 150 represents an allosteric mechanism that has been conserved in vertebrates. For Thr 152 , however, small hydrophobic residues are common, which probably stabilize the hydrophobic cluster in the loop bundle. The use of threonine at position 152 arose in mammalian caspase-3, and as our results show, Thr 152 provides a unique mechanism to abolish activity of the enzyme. The backbone carbonyl of Thr 152 also makes through-water H-bonds with the N terminus of the protein, so the H3CL not only bridges the dimer interface, but it also bridges the N and C termini of the protein. Importantly, this region of the N terminus has been shown to be phosphorylated in other caspases (44). Thus, interactions in the H3CL form extensive H-bonding and hydrophobic contacts that contribute to the stability of helix-2, helix-3, the loop bundle of L2 and L2Ј, and possibly the N and C termini of the protein.
We show that the allosteric mechanism of the H3CL is propagated through both intra-and interprotomer contacts. In the intraprotomer allosteric network, small changes in positioning the neighboring loop near helix-3 propagate along helix-3 and affect the adjoining surface ␤-sheet of ␤ 1 -␤ 3 . The surface ␤-sheet not only contains the catalytic His 121 , but it also forms stabilizing interactions with active site loop L2 in the dimer interface (28). Molecular dynamics simulations also demonstrated a repositioning of helix-3 and increased fluctuations in the ␤ 1 -␤ 3 sheet. At the end of the allosteric network, the conserved Glu 123 transiently interacts with the catalytic His 121 , and the interactions probably decrease the rate of proton transfer during the catalytic reaction. The conservation of Ser 150 in the H3CL as well as a charged residue at position 123 suggest a common allosteric mechanism in the apoptotic caspases (caspase-3, -6, -7, -8, and -10). Interestingly, the surface ␤-sheet undergoes a coil-to-helix transition in caspase-6, so fluctuations in helix-3/␤ 1 -␤ 3 may reflect an ancient allosteric mechanism coupled to Ser 150 (22). The allosteric network in caspase-6 may have further evolved to allow the additional coil-to-helix transition, providing a unique control mechanism from the common allosteric network.
The inclusion of Thr 152 is a more recent evolutionary event in mammalian caspase-3 and provides an added control mechanism by engaging an interprotomer allosteric network that connects the H3CL to the second active site via the loop bundle. The loss of activity due to modifications of the H3CL may be due to the increased flexibility of active site loop L4, which is stabilized through interactions with loops L2 and L2Ј, and the hydrophobic cluster that connects L4 to the H3CL is less ordered in the T152D variant. Our results identify Thr 152 as an allosteric kill switch that evolved in mammalian caspase-3. Whereas phosphorylation of Ser 150 decreases enzyme activity, the phosphorylation of Thr 152 is sufficient to abolish activity.
The pH of the cell decreases during early apoptosis from ϳ7.4 to ϳ6.8, so phosphorylation of the H3CL may couple changes in cellular pH with caspase dimer stability (38). The wildtype caspase-3 dimer dissociates to the inactive protomer in a pH-dependent manner, and the Ser 150 variants appear to affect the midpoint of the transition, resulting in transition to the inactive state at higher pH. The midpoint for the pH-dependent transition of ϳ5-5.5 is consistent with the titration of one or more histidine residues. Including the catalytic His 121 , there are eight histidine residues in the caspase-3 protomer (see Figs. 1B and 7A). Close to the site of modification, His 108 is located in a turn between helix 2 and ␤-strand 3, and the side chain of Ser 150 H-bonds with the backbone amide of His 108 (Fig. 7B). In Figure 7. A, location of histidine residues in caspase-3. Active-site loops L1, L4, and L2Ј each contain one histidine (His 56 , His 257 , and His 185 , respectively). Two histidine residues at the C terminus (His 277 and His 278 ) are not labeled. B, near the H3CL, Ser 150 hydrogen-bonds to His 108 . C, in wildtype caspase-3, the dimer is stabilized by electrostatic interactions between helix-5 and helix-5Ј across the dimer interface, facilitated by Glu 231 , His 234 , and Glu 272 of each protomer. The prime symbol indicates amino acids of the second protomer.

Common phosphorylation networks control caspase activity
addition, active site loops 1, 4, and L2Ј each contain one histidine. Although it is not clear at present how modifications in the H3CL may affect the pK a of the histidine residues, the most likely candidate for the pH-dependent effect on dimer stability is His 234 (Fig. 7C). Helix 5 makes several chargecharge interactions across the dimer interface with helix 5Ј of the second protomer, and His 234 , Glu 231 , and Glu 272 form salt bridges with Glu 231 Ј, Glu 272 Ј, and His 234 Ј in the interface (Fig. 7C). The H3CL is ϳ27 Å from the salt bridges, but the changes may propagate through interactions with the N terminus. Because the N terminus is phosphorylated in other caspases, the H3CL-N terminus-helix 5 interactions appear to represent a common network for signal propagation due to phosphorylation.
Targeting the dimerization of caspase-3 via phosphorylation is a novel allosteric mechanism and explains the previous suggestion that modification of the H3CL destabilizes the enzyme and leads to survival of the cell (26). Because apoptotic cells initially undergo an acidification process, the H3CL modification might help cells maintain a level of activity that allows for continued progression to cell death or toward "incomplete apoptosis" for processes such as erythroid or monocyte differentiation (45,46). Taken together, our results reveal the structural interaction network of allostery that connects evolutionarily conserved and novel phosphorylation sites directly to the active site.

Cloning, protein expression, and purification
All mutants were generated by PCR site-directed mutagenesis using primers containing the mutation (see supporting data) and pET21b expression plasmid that contained either wildtype caspase-3, caspase-3 D9A,D29A (called D 2 A), or caspase-3 D9A,D29A,D175A (called D 3 A) with a C-terminal His 6 tag (33,42). Mutants of caspase-6 and caspase-7 were generated by PCR site-directed mutagenesis using primers containing a mutant of interest (see supporting data) in pET21b and pET23b expression plasmid that contained wildtype caspase-6 or caspase-7 with a C-terminal His 6 tag (47,48). Mutations were confirmed by sequencing both DNA strands. Plasmids were transformed into Escherichia coli BL21(DE3) pLysS, and proteins were expressed as described (33,49).

Phylogenetic analysis
A fasta file for each caspase family member was compiled, and a preliminary alignment of the sequences was performed with MUSCLE (50). Sequences that opened gaps or had large deletions were removed from the analysis to avoid noise or bias. We used 1,325 sequences for the analysis, and the number of sequences for each individual caspase is shown in the leftmost column of Table S1. A multiple-sequence alignment (MSA) was generated for each family member with PROMALS3D (51). The data in Table S1 were generated by analyzing the amino acid frequency at each position with ProtParam on the ExPASy server (52). The phylogenetic tree for each caspase family member was created by using the maximum likelihood method in MEGA 7 (30,31). One hundred bootstraps were performed as a test of confidence for the tree topology (53). Each MSA was analyzed on the ProtTest3 server, and the Jones-Taylor Thornton substitution model of protein evolution plus ␥ distribution was selected to construct the trees (54,55). The trees are drawn to scale, with the branch lengths measured in the number of substitutions per site. Positions in the MSA with less than 80% site coverage were eliminated. Fewer than 20% alignment gaps, missing data, and ambiguous bases were allowed at any position. In each figure, the tree with the highest log likelihood from each family member is shown. The positions corresponding to Ser 150 and Thr 152 in caspase-3 are displayed on the branches of currently existing species, and the most likely ancestors are displayed at the nodes for each caspase family member. The bootstrap results are shown next to the branches and represent the percentage of trees in which the associated taxa are clustered together.

Enzyme activity assays
Enzyme activity was determined in a buffer of 150 mM Tris-HCl, pH 7.5, 50 mM NaCl, 10 mM DTT, 1% sucrose, 0.1% CHAPS (assay buffer) at 25°C, as described previously (56). The total reaction volume was 200 l, and the final enzyme concentration was 10 nM. Following the addition of substrate, the samples were excited at 400 nm, and emission was monitored at 505 nm for 60 s. The steady-state parameters, K m and k cat , were determined from plots of initial velocity versus substrate concentration and are presented in Table 1.

Crystallization and data collection
Caspase-3 variants were crystallized as described previously for wildtype caspase-3 (43). Briefly, each protein was dialyzed in a buffer of 10 mM Tris-HCl, pH 8.5, and 1 mM DTT and concentrated to ϳ7 mg/ml. Inhibitor, Ac-DEVD-chloromethylketone (reconstituted in DMSO), was added at a 5:1 (w/w) inhibitor/protein ratio, and DTT and NaN 3 were added to final concentrations of 10 and 100 mM, respectively. Samples were incubated for 1 h in the dark on ice. Crystals were obtained at 18°C by the hanging-drop vapor diffusion method using 4-l drops that contained equal volumes of protein and reservoir solutions over a 0.5-ml solution of 100 mM sodium citrate, pH 4.9 -5.2, 8 -18% PEG 6000 (w/v), 10 mM DTT, and 3 mM NaN 3 . Crystals appeared within 3-5 days and were briefly immersed in a cryogenic solution containing 20% 2-methylpentane-2,4diol and 90% reservoir solution. Crystals were stored in liquid nitrogen. Data sets were collected at 100 K at the SER-CAT synchrotron beamline (Advance Photon Source, Argonne National Laboratory, Argonne, IL). Each data set contained 180 frames at 1°rotation. The proteins crystallized in either the monoclinic space group C2 or the orthorhombic space group I222 and were phased with a previously published human CASP3 structure (PDB entry 2J30) using Phaser-MR in the Phenix software (57). Data reduction and model refinements were done using HKL2000, COOT, and Phenix, and a summary of the data collection and refinement statistics is shown in Table  S2 (57-60).

Molecular dynamics simulations
Molecular dynamics (MD) simulations were performed as described previously with GROMACS 2016, using the

Common phosphorylation networks control caspase activity
Amber99 force field and the TIP3P water model (61)(62)(63)(64)(65). All simulations started with the structure obtained from X-ray crystallography, as described above, and the inhibitor was removed from the structure file. As described previously for human caspase-3, the proteins were solvated in a periodic box of 62 ϫ 48 ϫ 66 Å, with ϳ13,500 water molecules (65). Sodium or chloride ions were added as required to neutralize the charge on the system. The system was first minimized using steepest descent, and then the waters were relaxed during a 20-ps MD simulation with positional restraints on the protein. Simulations of 50 ns were then run for each protein under constant pressure and temperature (300 K). A time step of 2 fs was used, and coordinates were saved every 5 ps. In each simulation, the protein was equilibrated within 500 ps.

pH studies and fluorescence emission
Proteins were incubated in a buffer of 10 mM Tris-HCl, pH 8.0, 100 mM NaCl, 1 mM DTT or in 8 M urea-containing buffer. Samples (1 M) were excited either at 280 nm or at 295 nm, and emission scans were collected between 300 and 400 nm (PTI C61 spectrofluorometer). The average emission wavelength (AEW or ͗͘) was determined for each protein as follows, where I represents fluorescence intensity and is wavelength, as described previously (66). We performed two experiments to examine pH-induced dimer dissociation and reassembly. First, proteins were incubated over a pH range of 3.0 -7.5, and the return of enzyme activity was determined after the proteins were re-equilibrated at pH 7.5. In those experiments, proteins were incubated for 4 h at 25°C over a pH range of 3.0 -7.5 using either 50 mM citrate buffer (pH 3-6), or 50 mM Tris-HCl, pH 7.5, containing 2 mM DTT, as described previously (32,40). Following the initial incubation, samples were then dialyzed in a buffer of 50 mM Tris-HCl, pH 7.5, 2 mM DTT for 14 h at 25°C. Reassembly of the active dimer was monitored by measuring enzyme activity after incubation at pH 7.5 for a minimum of 14 h. The data were compared with samples that were incubated at pH 7.5 for 18 h (25°C). Second, we monitored the intrinsic fluorescence emission of each protein (in the absence of substrate) as a function of pH. In those experiments, samples were incubated for a minimum of 12 h at 25°C over a pH range of 3.0 -7.5 using either 50 mM citrate buffer (pH 3-6) or 50 mM Tris-HCl, pH Ͼ6, containing 2 mM DTT. Samples (at each pH) were then excited at 280 or 295 nm, and the fluorescence emission was measured from 300 to 400 nm. All data were corrected for background signal, and the (or ͗͘) was calculated as described above (66).

Whole protein digest
The caspase-7(C186S) inactive variant was used as a substrate for caspase-3 cleavage assays, and the protein was diluted into activity buffer (pH 7.5 or 6) to a final concentration of 75 M. Caspase-3 (wildtype or variant) was then added to a final concentration of 10 M, and the samples were incubated for 20 min at 37°C. The cleavage reaction was stopped by adding 6ϫ SDS (1ϫ final) followed by boiling. Cleavage products were analyzed by 12.5% SDS-polyacrylamide gel.

Size-exclusion chromatography
Protein oligomeric state was examined using a Superdex75 Increase 10/300GL column (AKTA-FPLC). Proteins were dialyzed in a buffer of 10 mM phosphate, pH 7.5, and the column was equilibrated with the same buffer. Protein (100 l) at a concentration of 1-5 mg/ml was loaded onto the column at a flow rate of 0.8 ml/min. The column was calibrated using the gel filtration LMW calibration kit (GE Health Sciences, 28-4038-41), following the manufacturer's instructions.