(Received for publication, December 16, 1996, and in revised form, February 17, 1997)
From BASF Bioresearch Corp., Worcester, Massachusetts 01605
The caspase family represents a new class of
intracellular cysteine proteases with known or suspected roles in
cytokine maturation and apoptosis. These enzymes display a preference
for Asp in the P1 position of substrates. To clarify differences in the
biological roles of the interleukin-1 converting enzyme (ICE) family
proteases, we have examined in detail the specificities beyond the P1
position of caspase-1, -2, -3, -4, -6, and -7 toward minimal length
peptide substrates in vitro. We find differences and
similarities between the enzymes that suggest a functional subgrouping
of the family different from that based on overall sequence alignment.
The primary specificities of ICE homologs explain many observed enzyme
preferences for macromolecular substrates and can be used to support
predictions of their natural function(s). The results also suggest the
design of optimal peptidic substrates and inhibitors.
A growing body of evidence supports important roles for the
interleukin-1 converting enzyme (ICE)1
(1, 2) and its homologs (recently renamed caspases (3)) in cytokine
maturation and apoptosis. The caspase gene family, defined by protein
sequence homology but also characterized by conservation of key
catalytic and substrate-recognition amino acids, includes caspase-2
(4), caspase-3 (5-7), caspase-4 (8-10), caspase-5 (10), caspase-6
(11), caspase-7 (12-14), caspase-8 (15-17), caspase-9 (18, 19), and
caspase-10 (17). Each is an intracellular cysteine protease that shares
with the serine protease granzyme B specificity for Asp in the P1
position of substrates. The specific biological roles and
interrelationships of these enzymes are for the most part unknown and
are areas of active investigation in many laboratories.
A role for caspase-1 in inflammation is supported by several lines of
evidence. Caspase-1-deficient mice, and cells derived from those
animals, are deficient in IL-1 maturation and are resistant to
endotoxic shock (20, 21). Peptidic inhibitors of caspase-1 can be
effective in blocking maturation and release of IL-1
by cultured
cells (1) and in whole animals (22, 23) and of inflammation in animal
models (24, 25). The selectivity of the inhibitors employed in these
studies among the caspases has not been demonstrated, and so the
precise role of each caspase in inflammation is uncertain. Nevertheless
the results uphold the promise of caspase-1 and/or its homologs as
targets for anti-inflammatory drug discovery.
Caspases play important roles in apoptosis signaling and effector
mechanisms. Sequence alignments reveal homology with Ced-3 (26), a
nematode cysteine protease (27, 28) that is required for cell death.
The viral proteins CrmA and p35 are antiapoptotic and act by inhibition
of caspases (29, 30). A bacterial invasin induces apoptosis by binding
to and activating caspase-1 specifically (31). Caspase-3 is necessary
and sufficient for apoptosis in one acellular model (6); however, in
mice the essential function of this enzyme is limited to apoptosis in
the brain (32). A hallmark of apoptosis is the proteolytic inactivation
of poly(ADP-ribose) polymerase (33), and several caspases can catalyze
that cleavage (7, 11-13, 18, 34, 35). Lamin A cleavage during
apoptosis is catalyzed by caspase-6 (34, 36). Cleavage of other
proteins at Asp residues, including DNA-protein kinases C, protein
kinase C-, and Gas2, also accompanies apoptosis (37-41), although
in most cases the enzyme(s) responsible for those reactions have not
been identified.
Caspases can transduce or amplify signals by mutual activation. The "death domain" motifs of caspase-8 (15-17, 19) and -10 (17, 19) and their association with Fas or tumor necrosis factor receptors via interaction with FADD, suggests that they are upstream activators that proteolytically mature other caspases. Fas-induced apoptosis is characterized by an early, transient "caspase-1-like" protease activity followed by a "caspase-3-like" activity (42), suggesting an ordered activation cascade. Reconstitution experiments suggest that both caspase-3 and -7 are activated by caspase-6 (43) and caspase-10 (17). Little is known regarding caspase regulation or inactivation. Caspase-1 contains a site that is rapidly autodegraded in vitro, and cleavage at this site may represent a physiological mechanism of down-regulation (44).
To facilitate studies of the roles of caspase family proteases in inflammation and apoptosis, and to help elucidate protease-mediated mechanisms of signal transduction and regulation, we have characterized in detail the substrate preferences of caspase-1 and five of its homologs. We present here our findings of the specificities of each enzyme, which suggest the design of novel specific peptidic substrates and inhibitors, and help to predict the biological roles of these enzymes.
Underivatized peptides (Quality Controlled
Biochemicals, Hopkinton, MA) were prepared by standard solid-phase
methods, purified to 95% by HPLC, and confirmed by low resolution
mass spectrometry. Concentrations of stock solutions in
Me2SO were determined in duplicate as described (45).
Chromogenic and fluorogenic peptides were from California Peptide
Research (Napa, CA). Concentrations of chromogenic peptide stock
solutions were determined in duplicate as described (46).
Concentrations of fluorogenic peptide stocks were determined in
duplicate by UV absorbance using
325 = 15,990 in aqueous
buffer at pH 7.5.
Caspase-1, -2, -3, and
-4, containing N-terminal His tags to facilitate purification, were
prepared as described (9, 30, 47). Caspase-1 contained the
stability-enhancing point mutation D381E (47). Caspase-6 (Mch2) (11)
beginning at Ala-2, or caspase-7 beginning at Ala-23, with
amino-terminal polyhistidine tags (MGHHHHHHGSG), were subcloned into a
pBluescript II KS(+) (Stratagene, La Jolla, CA) derivative under
control of the Lambda PL promoter. Amino-terminal coding
sequences were modified to reflect Escherichia coli codon
preferences. To increase soluble expression the proteases were
coexpressed with the bacterial chaperones groES and groEL (48). The
groESL operon, under control of the isopropyl-1-thio-
-D-galactopyranoside-inducible
Ptac promoter, was cloned into a pACYC177 (49) derivative
encoding LacI and the temperature-sensitive repressor
cI857 inserted into the
-lactamase gene. Both plasmids
were stably maintained in the same bacterial cell. Simultaneous
expression of the chaperones and proteases were induced by addition of
isopropyl-1-thio-
-D-galactopyranoside to 2 mM and temperature shift from 30 °C to 40 °C at
A600 = 0.6. The bacteria were harvested at
t = 5 h, and the proteases were purified as
described (47).
Peptide cleavage assays were modeled on those of Howard et al. (50). Substrate concentration depletion during the reaction is described by Equation 1.
![]() |
(Eq. 1) |
![]() |
(Eq. 2) |
![]() |
(Eq. 3) |
Assays of chromogenic or fluorogenic substrate cleavage contained in 100 µl: 100 mM HEPES (pH 7.5), 20% (v/v) glycerol, 5 mM dithiothreitol, 0.5 mM EDTA, 0.1% (w/v) bovine serum albumin, and 0.3 (fluorogenic substrates) or 3.0-200 (chromogenic substrates) ng of purified enzymes. After a 30-min preincubation at 30 °C, substrates in Me2SO were added to various concentrations. For chromogenic substrates, enzyme-catalyzed release of p-nitroanilide was monitored at 405 nm in a microtiter plate reader (Molecular Devices, Menlo Park, CA). For fluorogenic substrates, Amc release was monitored at 460 nm using 385 nm excitation in a Labsystems (Needham Heights, MA) Fluoroskan Ascent fluorescence plate reader. Km and kcat values were determined from plots of activity versus substrate concentration. Absolute kcat values for chromogenic substrates were calculated using a standard curve determined with p-nitroanilide and are uncorrected for enzyme purity and percent activity.
Assays of inhibitor potency were performed with fluorogenic substrates using nominal enzyme concentrations of 0.1 nM or less. Apparent Ki values were calculated by dividing IC50 values by (1 + [S]/Km).
Building on previous work
(1, 50, 51) we studied the substrate specificity of several caspases by
measuring relative Vmax/Km
values, a measure of enzymatic specificity (52), toward a series of
defined peptide sequence variants. We chose this rather than a
combinatorial approach to obtain quantitative data on both optimal and
suboptimal substrate sequences. Peptides were generally of minimal
length, contained C-terminal Trp residues to facilitate detection and
concentration determination, and were acetylated and amidated to avoid
introduction of nonnatural charges. Sequence variants were selected to
test hypotheses of substrate recognition and to explore the structural
range of natural amino acids. Application of Equation 3 requires that
[S]0 Km. Km
values of each enzyme toward preferred peptide substrates were
determined from the initial rates of peptide hydrolysis as a function
of peptide concentration (Table I). All observed values were at least 10-fold greater than the peptide concentration utilized in the specificity studies (10 µM), confirming that the
use of Equation 3 is appropriate.
|
Caspase-1 and -4 specificities
were tested using variants of Ac-YVADGW-Am, a preferred peptide
substrate (1) based on the caspase-1 cleavage site within IL-1
(YVHD117A) (53, 54). In P4, caspase-1 and -4 prefer
hydrophobic amino acids (Fig. 1), with the former
preferring aromatics (consistent with IL-1
cleavage), and the
latter, aliphatics. In P3, both enzymes prefer Glu. The crystal
structure of caspase-3 in complex with Ac-DEVD-CHO (55) shows that the
P3 Glu of the peptidic ligand makes an ionic interaction with the side
chain of the conserved residue Arg-207. Ligands containing P3 Glu
probably make analogous favorable contacts with caspase-1 and -4. In
P2, a broad range of amino acids are tolerated by both enzymes,
consistent with the observation from the caspase-1 and -3 crystal
structures that side chains of amino acids in this position are
solvent-exposed. As expected (1, 51), caspase-1 and also caspase-4
prefer Gly in the P1
position of substrates. In addition, we find that all of the aromatic and the thiol- and hydroxyl-containing amino acids
were also efficiently cleaved.
As predicted from the peptide cleavage results, the chromogenic tetrapeptide p-nitroanilides Ac-YEVD-pNA and Ac-LEVD-pNA were preferentially cleaved by caspase-1 and -4, respectively, with kcat/Km values considerably higher than related peptides, including in the former case the widely used substrate Ac-YVAD-pNA (Table II).
|
In a survey of peptide substrates spanning known or suspected cleavage sites within caspases and their substrates, we found that only peptides containing VDQQD (caspase-2 residues 312-316) and LDVVD (caspase-6 residues 175-179) were cleaved by caspase-2, but peptides Ac-DQQDGW-Am and Ac-DVVDGW-Am were not. We examined the length dependence of peptide cleavage by caspase-2 (Table III) and found that, in contrast to caspase-1 (1), caspase-2 requires a P5 residue in peptide substrates for efficient cleavage. Like caspase-1 (1), efficient cleavage of peptides by caspase-2 did not require prime-side substrate residues (Table III). We therefore examined the substrate specificity of caspase-2 using sequence variants of Ac-VDQQDGW-Am (Fig. 2). Hydrophobic residues were preferred in the P5 position. Like caspase-3 and -7 (see below), caspase-2 displayed a strong preference for Asp in P4. In P3 and P2, Val and Ala, respectively, were preferred over Gln, but like caspase-1 and -4, Glu was a favored P3 substrate residue and in P2 a structurally wide range of amino acids were tolerated.
|
The results suggested that optimal caspase-2 peptidic substrates and inhibitors would contain the sequence VDVAD, and consistent with that, Ac-VDVADGW-Am was cleaved at a correspondingly greater efficiency than either of the singly substituted peptides (not shown). We tested Ac-VDVAD-pNA as a caspase-2 substrate and found it was preferred by caspase-2, cleaved with a kcat/Km value similar to those of other enzymes and their preferred substrates (Table II). Ac-DEVD-pNA, widely used as a caspase-3 substrate, was poorly cleaved by caspase-2. Using the fluorogenic substrate Ac-VDVAD-Amc (Km = 80 µM), the pentapeptide aldehyde Ac-VDVAD-CHO inhibited caspase-2 with a Ki value of 3.5 nM. In contrast, the Ki of Ac-DEVD-CHO was 1.75 µM, again supporting the observed in vitro requirement of caspase-2 for a P5 peptide residue.
Caspase-3 and -7 SpecificityCaspase-3 and -7 were probed in P4-P2 using the peptide variants designed for caspase-2. Length dependence experiments showed that for these enzymes a P5 residue was unnecessary for efficient cleavage, and the presence of the P5 Val residue in particular did not significantly affect cleavage rates.2 The two enzymes displayed many similarities in their primary specificities. The P4 Asp preference of caspase-3 was almost absolute (Fig. 2), with peptides containing charge-conserving (Glu) or isosteric (Asn) substitutions cleaved poorly. Caspase-7 also displayed a strong preference for Asp in P4. Like the other caspases, both enzymes tolerated a broad range of amino acids in P3 but preferred Glu, again consistent with the structure of caspase-3 bound to Ac-DEVD-CHO (55). In P2, caspase-3 and -7 were also similar to other caspases in their preferences for hydrophobic residues, but for these enzymes the preference was stronger. The favorable interaction of caspase-3 for Met in P2 probably contributes to its potent inhibition by p35, which inactivates caspase-3 by binding at a site containing DQMD (30, 56).
Consistent with the peptide cleavage assay results, the preferred chromogenic substrate for caspase-7 was Ac-DEVD-pNA (Table II). Caspase-3 also efficiently cleaved this peptide, and, as expected, cleaved the p35-based substrate Ac-DQMD-pNA with greater efficiency. The corresponding fluorogenic peptide Ac-DEVD-Amc displayed Km values of 10 and 11 µM for caspase-3 and -7, respectively. With this substrate, Ac-DEVD-CHO inhibited both enzymes potently, with apparent Ki values of 0.2 and 0.3 nM, respectively. The caspase-2-designed inhibitor Ac-VDVAD-CHO also potently inhibited caspase-3 and -7, with apparent Ki values of 1.0 and 7.5 nM, respectively.
Caspase-6 SpecificitySequence variants of a peptide based on
the caspase-6 cleavage site in lamin A (34, 36) were used in a limited
study of the substrate preference of caspase-6 (Fig. 3).
Supporting its proposed role in cleavage of lamin A at
VEID230N during apoptosis, the results show a favorable
interaction of caspase-6 with peptides containing Val in P4. The
similar preference for Thr in P4 suggests that, for this enzyme, it is
not necessarily hydrophobics that are preferred in P4, but rather the
-branched amino acids. Unexpectedly, the observed preference in
peptide substrates for Gln and Val in P3 and P2, respectively (Fig. 3), was not reflected in the corresponding chromogenic peptides, where Ac-VEID-pNA displayed a kcat/Km value
more than 10 times that of Ac-VQVD-pNA (Table II).
We are interested in the development of caspase-1 inhibitors as anti-inflammatory drugs. We are studying the biological roles of caspases to predict the consequences of cross-inhibition by our compounds and to reveal new promising drug discovery targets. The present study resulted in new research tools with which to probe the structure and function of caspases, explanations for observed activities toward specific natural substrates, and information that can be used to predict activation pathways and which enzyme(s) are likely or not to cleave newly discovered substrates.
Consistent with previous studies, we find the most significant
differences in caspase specificities at the P4 positions. P3 specificities are similar between these enzymes, and in P2 a wide range
of amino acids is tolerated. These observations are consistent with the
structures of caspase-1 and -3 (55, 57, 58), which show that the P3 and
P2 side chains of peptidic inhibitors are relatively solvent-exposed,
and that the P4 side chains occupy defined pockets that vary
significantly between those enzymes. The P4 preferences can be
categorized as hydrophobic (caspase-1, -4, and -6) or Asp (caspase-2,
-3, and -7). This categorization is at odds with sequence alignment
(3), which predicts that caspase-6 is more closely related to caspase-3
and -7 than to caspase-1, and that caspase-2 is more distantly related.
The caspase-3 structure (55) shows that the P4 Asp side chain of a
cocrystallized peptidic inhibitor makes polar contacts with the
backbone amide nitrogen of Phe-250 (referred to as Phe-381B in Ref. 55
by alignment with caspase-1) and the N2 of Asn-208 (referred to as
Asn-342). By alignment, the analogous residue in caspase-6 is Glu-221,
which might make an unfavorable ionic interaction with Asp in P4 of a
substrate.
Caspase-1 and -4 display very similar specificities in each substrate
position including P1, and preliminary results suggest that caspase-5
is also similar.2 The P4 specificity of caspase-1 suits it
to activation of pro-IL-1
, and the defect in pro-IL-1
maturation
in caspase-1-deficient mice (20, 21) suggests that little or no
pro-IL-1
is matured by other proteases. This does not rule out a
requirement for a second protease in IL-1
maturation, for example by
regulating maturation of caspase-1. Murine Ich-3 stimulates pro-IL-1
maturation by ICE (59), and caspase-4 and -5, for which clear roles
have not been established, may function analogously. The P4 hydrophobic preference of caspase-4 is consistent with activation of caspase-1, where cleavages occur at AVQD119N, WFKD297S,
and FEDD316A. Although caspase-1 is capable of
autoactivation in vitro (60), it exists largely as its
inactive precursor in vivo (61) and therefore may require
activation by another enzyme in cells. Since caspase-1 knockout mice
display normal apoptosis, the transient caspase-1-like activity
(defined as cleavage activity toward a P4 Tyr-containing peptide
substrate), that precedes caspase-3-like activity and is required for
Fas-induced apoptosis in mouse W4 cells (42) might include caspase-4 or
-5.
The primary specificity of caspase-6 suits it well to cleavage of one
known natural substrate, lamin A (34, 36). The caspase-6 preference for
the -branched amino acids in P4 distinguishes it from caspase-1 and
-4 and suggests cleavage of other substrates with sites containing
-branched P4 residues, including maturation sites within caspase-6
(19) (TETD23A and TEVD193A) and caspase-3
(IETD175S). The latter is consistent with proposals based
on cotransfection (43) and in vitro maturation experiments
(19, 43). Intolerance for P4 Tyr (Fig. 3) would prevent caspase-6
maturation of IL-1
.
Caspase-2, -3, and -7 each display similar specificities, which suggests that their roles in cells, if not completely redundant, are at least overlapping. They share a strong requirement for Asp in P4, which qualifies them for cleavage of several known P4 Asp-containing apoptosis substrates. The relationships between these enzymes are largely unknown. In one model system, caspase-3 is required for apoptosis (6), but other enzymes may participate in other tissues or after different apoptotic stimuli. Despite the in vitro cleavage of a peptide spanning a known maturation site of caspase-6, the intact protein is not a caspase-2 substrate,3 demonstrating that the structural context of a protein sequence also contributes to recognition as a caspase substrate. Notably, the caspase-3-like proteolytic activity often measured in apoptotic cells is usually defined by cleavage of Ac-DEVD-pNA or similar substrates. Caspase-2, if active, would not be observed using that substrate, but could be observed along with caspase-3 and -7 by using Ac-VDVAD-pNA.
The design of efficiently recognized novel peptidic substrates and inhibitors based on in vitro specificity studies was straightforward and successful, and several such compounds are described here. However, optimized peptides were recognized similarly by enzymes with similar specificities, limiting their usefulness as specific in vivo probes of these enzymes. The requirement of caspase-2 for a P5 residue allows for inhibitors and substrates (such as Ac-DEVD-CHO and Ac-DEVD-pNA) that are not well recognized by caspase-2. Unfortunately the presence of the P5 residue, while not required, was not detrimental to recognition by other enzymes such as caspase-3 and -7. It may be possible to design new peptidic probes that display greater specificity by using residues that are suboptimal for target enzymes and not tolerated by others.
The caspase proteases have overlapping substrate specificities that suggest at least partially overlapping functions. Despite evidence for a central role for caspase-3 in apoptosis, mice deficient in this enzyme display a defect in apoptosis that is limited to the brain (32), suggesting that in most tissues other enzymes such as caspase-1 and -7 function redundantly with it. Partial functional overlap may allow cells to modulate the degree or nature of the response to different proinflammatory or proapoptotic stimuli by differential regulation of the caspases. Alternatively, functional overlap might simply provide backup mechanisms for critical processes such as viral or tumor clearance. Some caspase functions suggested by in vitro studies may not occur because of tissue or subcellular compartmentalization or regulation of activation or expression during cell cycle or development. It should not be surprising that the systems that execute key steps in the processes of cytokine maturation and apoptosis are capable of both fine regulation and redundant function.