Human Rhinovirus 2 2Apro Recognition of Eukaryotic Initiation Factor 4GI

The 2A proteinase (2Apro) of human rhinovirus 2 is a cysteine proteinase with a unique chymotrypsin-like fold. During viral replication, 2Apro performs self-processing by cleaving between its own N terminus and the C terminus of the preceding protein, VP1. Subsequently, 2Apro cleaves the two isoforms of the cellular protein, eukaryotic initiation factor (eIF) 4G. We have previously shown that HRV2 2Apro can directly bind to eIF4G isoforms. Here we demonstrate using deletion mutants of eIF4GI that HRV2 2Apro requires eIF4GI amino acids 600–674 for binding; however, the amino acids at the cleavage site, Arg681 ↓ Gly, are not required. The HRV2 2Apro binding domain for eIF4GI was identified by site-directed mutagenesis. Specifically, mutations Leu17 → Arg and Asp35 → Glu severely impaired HRV2 2Apro binding and thus processing of eIF4GI in rabbit reticulocyte lysates; self-processing, however, was not affected. Alanine scanning analysis further identified the loop containing residues Tyr32, Ser33, and Ser34 as important for eIF4GI binding. Although Asp35 is part of the catalytic triad, most of the eIF4GI binding domain lies in a unique exosite structure absent from other chymotrypsin-like enzymes and is distinct from the substrate binding cleft. The exosite represents a novel virulence determinant that may allow the development of specific inhibitors for HRV2 2Apro.

The 2A proteinase (2A pro ) of human rhinovirus 2 is a cysteine proteinase with a unique chymotrypsin-like fold. During viral replication, 2A pro performs self-processing by cleaving between its own N terminus and the C terminus of the preceding protein, VP1. Subsequently, 2A pro cleaves the two isoforms of the cellular protein, eukaryotic initiation factor (eIF) 4G. We have previously shown that HRV2 2A pro can directly bind to eIF4G isoforms. Here we demonstrate using deletion mutants of eIF4GI that HRV2 2A pro requires eIF4GI amino acids 600 -674 for binding; however, the amino acids at the cleavage site, Arg 681 2 Gly, are not required. The HRV2 2A pro binding domain for eIF4GI was identified by sitedirected mutagenesis. Specifically, mutations Leu 17 3 Arg and Asp 35 3 Glu severely impaired HRV2 2A pro binding and thus processing of eIF4GI in rabbit reticulocyte lysates; self-processing, however, was not affected. Alanine scanning analysis further identified the loop containing residues Tyr 32 , Ser 33 , and Ser 34 as important for eIF4GI binding. Although Asp 35 is part of the catalytic triad, most of the eIF4GI binding domain lies in a unique exosite structure absent from other chymotrypsin-like enzymes and is distinct from the substrate binding cleft. The exosite represents a novel virulence determinant that may allow the development of specific inhibitors for HRV2 2A pro .
Many viruses encode proteins that tailor the host cell physiology to their own needs (1). Modification of the cellular machinery of protein synthesis is a favored strategy of RNA viruses, because this can directly increase the efficiency of the expression of their genetic information. Almost all picornaviruses, a family of positive strand RNA viruses, including FMDV, 1 enteroviruses (such as PV), and HRVs, specifically prevent recruitment of capped cellular mRNAs to the 40 S ribosomal subunit (2,3). This is achieved by virally encoded proteinases, which cleave the two isoforms of the eukaryotic initiation factor (eIF) 4G, eIF4GI, and eIF4GII (4 -6). eIF4G isoforms are scaffolding proteins that have binding sites for eIF4E, the protein recognizing the 5Ј mRNA cap, and eIF3, a complex of about 10 proteins, which in turn binds to the 40 S ribosomal subunit (7). The virally induced cleavage separates these two binding domains, preventing association between the cellular mRNA and the 40 S ribosomal subunit. Initiation of translation from viral mRNA is unaffected, because it is not cap-dependent but initiates internally from an internal ribosomal entry site (8,9).
In enteroviruses and HRVs, a cysteine proteinase termed 2A pro is responsible for the cleavage of eIF4G isoforms. Viruses possessing certain specific mutations in 2A pro have a small plaque phenotype, as exemplified by PV (10) or SVDV (11,12). In addition, such SVDV mutations show an attenuated phenotype (11,12). A similar attenuated phenotype is also observed in FMDV when the proteinase responsible for eIF4G cleavage, the Leader proteinase, is deleted (13). Thus, the eIF4G cleavage reaction is an important determination of virulence in many picornaviruses.
The mechanism of cleavage of eIF4G isoforms by the picornaviral proteinases has not been completely resolved. Evidence for direct cleavage by both the 2A pro and the L pro has been presented (14,15). However, activities in the PV-infected cell, distinct from the 2A pro , have been described that are able to perform cleavage of eIF4G isoforms (3). A further unresolved observation is that the cleavage of eIF4GI and eIF4GII occurs concomitantly in HeLa cells infected either with HRV serotypes 2 or 16 (16,17). In contrast, in cells infected with PV or HRV14, cleavage of eIF4GI temporally precedes that of eIF4GII (6,18). Interestingly, HRV2 and HRV16 are members of the genetic A group of HRVs, whereas HRV14 belongs to the B group (19). Further experiments will be required to show whether this difference in cleavage of the eIF4G isoforms extends to other representatives of the two genetic groups.
Recently, we have shown that the Leader proteinase of FMDV, a cysteine proteinase with a papain-like fold, can bind directly to the isoforms of eIF4G (20); the region recognized on eIF4GI was mapped to amino acids 640 -669 (numbering according to Byrd et al. (21)) but did not include the amino acids at which cleavage occurs (Gly 674 2 Arg). The eIF4G binding domain of the L pro was shown to comprise the 18 amino acids of the C-terminal extension as well as the residue Cys 133 , which is adjacent to the CTE in the tertiary structure (22). A similar interaction between the 2A pro of HRV2 and eIF4G isoforms was also observed (20). However, the binding domain on the HRV2 2A pro , which is a cysteine proteinase with a chymotrypsin-like fold (23, 24), was not identified. Furthermore, although the site of interaction on eIF4GI was shown to lie N-terminal to the HRV2 2A pro cleavage site (Arg 681 2 Gly), it was not defined further (20).
Here, we use site-directed mutagenesis to identify certain residues between amino acids 17 and 35 of HRV2 2A pro as being responsible for the binding of eIF4GI. The site recognized on eIF4GI was defined as being within amino acids 600 -674.

EXPERIMENTAL PROCEDURES
Reagents-pHRV2 VP1-2A pro contains the HRV2 nucleotides 2318 -3586, encoding all of VP1 except the first two amino acids, followed by all of 2A pro and two stop codons (25). Mutations in 2A pro were generated by standard PCR mutagenesis and introduced as EcoRI/XhoI fragments into the plasmid pGEX5X (Amersham Biosciences) as required (20). Fragments of eIF4GI for in vitro translation were amplified from plasmid pSKHC1 (26), which contains the entire eIF4GI cDNA, and cloned as EcoRI/HincII fragments into pBluescriptKS (Stratagene).
Purification of GST Fusion Proteins-Escherichia coli BL21(DE3)LysS (Novagen) cells were transformed with plasmids encoding the GST-2A pro fusion proteins or GST alone. For expression, overnight cultures were diluted 1:10 in 250 ml of medium, and the cells were incubated at 37°C to an A 600 of 0.8 and induced for 3 h at 18°C by addition of isopropyl-1-thio-␤-Dgalactopyranoside to a final concentration of 0.3 mM (20). The fusion protein was purified on glutathione-agarose resin (Amersham Biosciences) using standard techniques.
GST Pull-down Assays-Glutathione-Sepharose beads coated with GST fusion proteins were incubated in binding buffer (50 mM Tris-Cl, pH 7.4, 10 mM EDTA, 150 mM NaCl) with either an aliquot (8 l) of RRL (Promega) or with radiolabeled in vitro translated proteins for 2 h at 4°C. After three washes with binding buffer, bound proteins were eluted by boiling in SDS-PAGE loading buffer, resolved by SDS-PAGE, and visualized by Western blotting (using enhanced chemiluminescence system of Pierce for detection) or fluorography (20).
In Vitro Translation-In vitro translations in RRLs to examine 2A pro self-processing and eIF4GI cleavage were performed using in vitro transcribed RNAs as described (25,27). In vitro expression of radiolabeled proteins for GST pull-down assays was performed in RRLs (Quick Coupled Transcription/Translation system; Promega) in the presence of [ 35 S]methionine (20 Ci per reaction, Hartmann Analytic) (20). Labeled proteins were resolved by SDS-PAGE and gels were dried and exposed to x-ray films.

Mutations in HRV2 2A pro Specifically Affecting eIF4GI
Cleavage but Not Self-processing-A literature search revealed three mutations in picornaviral 2A pro reported to delay or inhibit eIF4GI cleavage without affecting self-processing (Table  I). The poliovirus mutants were identified in an analysis of PV 2A pro by site-directed mutagenesis, whereas the SVDV mutant was determined by the study of virulent and attenuated SVDV strains.
In PV 2A pro , Asp 38 is the third member of the catalytic triad, orienting the catalytic histidine (23,28). The phenotype of the PV 2A pro D38E mutant (normal self-processing but an inability to process eIF4GI; Table I) implied, however, that this residue had a role in recognizing eIF4GI in addition to its function in the catalytic triad. To investigate whether a similar phenotype could be observed in HRV2 2A pro , we examined self-processing and eIF4GI cleavage activities of a mutant protein bearing the equivalent mutation D35E. For this, we used our previously described assay for measuring HRV2 2A pro self-processing and eIF4GI cleavage in RRLs (25,27). In this system, an in vitro transcribed mRNA encoding the HRV2 capsid protein VP1 and the subsequent 2A pro was used to program protein synthesis. As can be seen in Fig. 1A, self-processing can be monitored by examining the status of the newly synthesized radiolabeled protein. The cleavage of eIF4GI, which is abundant in RRLs and migrates as a series of bands with a relative molecular mass of about 220kDa, is investigated by immunoblotting aliquots of the same samples with an antiserum against the N terminus of eIF4GI (Fig. 1B). When an mRNA encoding the wild-type 2A pro enzyme is translated, 50% of self-processing is observed 20 min after initiation of protein synthesis (Fig. 1A, left panel). Similarly, 50% eIF4GI cleavage is also seen after 20 min (Fig. 1B, left panel). To investigate the behavior of the D35E HRV2 2A pro mutant protein, mRNA was transcribed from an appropriately mutated plasmid and used to program RRLs. As can be seen in Fig. 1A (middle panel), self-processing occurred at wild-type rates. However, processing of eIF4GI was severely delayed in this mutant; a significant degree of cleavage was first observed after 180 min of incubation ( Fig. 1B, middle panel). Thus, the replacement of Asp 35 by glutamate in the HRV2 2A pro produces a similar phenotype as the equivalent residue in PV 2A pro . Using a bacterial expression system, Sommergruber et al. (29) also investigated the behavior of the HRV2 2A pro D35E mutant but found only 16% self-processing activity. Because these experiments required induction at 42°C for 2 h in bacteria, we believe that our assay system in the RRL gives a more authentic picture of the behavior of the D35E mutant.
Yu and Lloyd (28) identified a second mutation in PV 2A pro , Y89L (Table I), which severely affected eIF4GI cleavage without affecting self-processing. The introduction of the equivalent Y86L mutation into HRV2 2A pro did not essentially affect self-processing (Fig. 1A, right panel). The onset of eIF4GI cleavage was, however, delayed (Fig. 1B, right panel), but cleavage was essentially complete after 90 min. Thus, information derived from one picornaviral proteinase cannot always be directly applied to the corresponding proteinase in another picornavirus.
This notion was further strengthened when a residue in HRV2 2A pro was replaced with one identified as being important in SVDV 2A pro for cleavage of eIF4GI. In this enzyme, the presence of isoleucine at residue 20 is severely detrimental to eIF4GI cleavage (Table I). Replacement of the equivalent residue in HRV2 2A pro , Leu 17 , with isoleucine did not affect either self-processing or eIF4GI cleavage in any detectable way in HRV2 2A pro (Fig. 2, left panel). However, the presence of Arg at this position (the virulent wild-type residue in SVDV 2A pro ) dramatically inhibited eIF4GI cleavage without affecting selfprocessing (Fig. 2, right panel). This result is completely the opposite of that observed in SVDV; here the 2A pro from the virulent strain had arginine at the equivalent position, whereas the avirulent strain, in which eIF4GI processing was hampered, possessed Ile (Table I) (11,12). These results suggest that the recognition of eIF4GI by 2A pro may be different between HRV2 and SVDV.
HRV2 2A pro Mutant Proteins D35E and L17R Fail to Bind eIF4GI-The ability of the above mutants to perform selfprocessing at similar levels to the wild-type indicated that the catalytic property of the enzyme had not been reduced by the various mutations. We therefore investigated whether the mutants were defective in their ability to bind eIF4GI, because we have recently shown that a GST-HRV2 2A pro C106A fusion protein can bind to the eIF4G isoforms (20). The mutation C106A inactivates the enzyme by replacing the active site cysteine by alanine, preventing any possible cleavage of eIF4GI during the experiment. We constructed GST-HRV2 2A pro C106A fusion proteins bearing the respective mutations and expressed them in E. coli (Fig. 3A). The mutant proteins were bound to glutathione-coupled Sepharose and incubated with RRLs as a source of eIF4GI. After washing, proteins bound to the respective HRV2 2A pro C106A fusion proteins were resolved by SDS-PAGE and detected by immunoblotting. eIF4GI is retained by HRV2 2A pro C106A and HRV2 2A pro C106A-Y86L (Fig. 3B, lanes 2 and 4). In contrast, no binding was observed with the mutant proteins HRV2 2A pro C106A-D35E and -L17R (Fig. 3B, lanes 3 and 5), suggesting that the reduction in their cleavage of eIF4GI was due to an impairment of binding.
The Activity of the HRV2 2A pro L17R Mutant Protein Can Be Partially Rescued by a Second Site Mutation-To identify further residues involved in the HRV2 2A pro ⅐eIF4GI interaction, we concentrated on residues interacting with Leu 17 , which were identified using the three-dimensional structure of HRV2 2A pro . Although the mutation D35E was detrimental to eIF4GI cleavage, we were concerned that substitution of its neighbors, with which it forms an extensive hydrogen-bond network, would be detrimental to catalysis in general (see Fig. 5 in Ref. 24). These residues were therefore not investigated further.
The structure of HRV2 2A pro is shown in Fig. 4. HRV2 2A pro has a chymotrypsin-related fold comprising a unique fourstranded ␤ sheet as the N-terminal domain and a six-stranded ␤ barrel as the C-terminal domain (24). The asymmetric unit contained two HRV2 2A pro molecules (designated A and B). Detailed analysis of the structure revealed that Leu 17 (posi-  Table II; the equivalent residues in SVDV 2A pro are indicated). In contrast, two other residues, Asn 21 and Met 24 , are positioned about 4 Å from Leu 17 in molecule A, but are much more distant in molecule B (Table II). It is worth noting that the loop containing residues 17-30 shows the greatest conformational difference between the two molecules (24). We selected Val 30 and Tyr 32 for further investigation because of their consistent proximity. Of the two, Val 30 appeared of greater importance because it was closer and also faces Leu 17 in both molecules. In addition, in the SVDV 2A pro of the virulent strains, the equivalent residues to Leu 17 and Val 30 are Arg 20 and Glu 33 ; thus, in this SVDV strain, Arg 20 and Glu 33 could form a salt bridge, providing that they are oriented in a similar fashion. This positive salt bridge would be lost in the avirulent strain when isoleucine replaces Arg 20 . Interestingly, the equivalent residues at these positions in several coxsackieviruses (24) are also arginine and glutamate, strengthening the idea of an important ionic interaction.
To test this hypothesis, we attempted to generate this salt bridge in HRV2 2A pro . We introduced the mutation V30E into HRV2 2A pro L17R and measured the ability of this double mutant to cleave eIF4GI compared with HRV2 2A pro L17R. As a further control, we also investigated the activity of the single mutant HRV2 2A pro V30E. Fig. 5 (left panel) shows that self-processing by the mutant protein HRV2 2A pro L17R-V30E occurred at wild-type rates; furthermore, the processing of eIF4GI by this double mutant was significantly more efficient than that of the original HRV2 2A pro L17R mutant protein. With HRV2 2A pro L17R-V30E, 50% cleavage of eIF4GI was observed after 90 min (Fig. 5B, left panel) compared with more than 180 min with the single L17R mutant protein (Fig. 2B, right panel) and 20 min with the wild-type HRV2 2A pro (Fig. 1B, left panel). This suggests that the interaction of Leu 17 and Val 30 is important in eIF4GI cleavage by HRV2 2A pro .
The positive effect of the presence of glutamate at the residue 30 in the HRV2 2A pro L17R supported the idea of an interaction between residues 17 and 30. This would imply that the single mutant HRV2 2A pro V30E would also be handicapped in eIF4GI cleavage. However, examination of the mutant HRV2 2A pro V30E revealed only a minimal effect on self-processing and eIF4GI cleavage (Fig. 5, right panel). This result argues against the importance of interaction between Leu 17 and Val 30 and stresses, rather, that the occupancy at residue 17 is more critical.
To investigate whether the introduction of the mutation V30E had also restored the ability of the L17R mutant protein to bind to eIF4GI, we expressed the fusion protein GST-HRV2 2A pro C106A-L17R-V30E. Fig. 5D (lane 3) shows that the binding of the HRV2 2A pro C106A-L17R-V30E to eIF4GI is improved compared with that of the L17R single mutant (Fig. 3B,  lane 5). Nevertheless, wild type binding was not restored.
We also investigated the role of a third residue interacting with Leu 17 , methionine 24 (Table II). The equivalent residue in SVDV 2A pro is Trp 27 and may be in a position to cover and thus stabilize the putative salt-bridge between Arg 20 and Glu 33 . Accordingly, the triple mutant HRV2 2A pro L17R-V30E-M24W was constructed; however, it showed no increase in activity on eIF4GI when compared with the double mutant (data not shown).
Alanine Scanning Reveals the Involvement of Residues Tyr 32 , Ser 33 , and Ser 34 in eIF4GI Recognition-To investigate the importance of other residues bordering on strand eI2 (Fig. 4) in eIF4GI binding, we carried out alanine scanning analysis of amino acids lying close to Leu 17 . This method allows the removal of side-chain interactions without inducing extreme perturbations into the overall fold of the molecule. Two sets of three amino acids, Phe 20 -Asn 21 -Ser 22   other group A human rhinoviruses (24). Furthermore, Tyr 32 was of interest because of its proximity to Leu 17 ( Fig. 4 and Table II). Following construction of the respective plasmids, RNAs were prepared in vitro and used to program RRLs. Selfprocessing of HRV2 2A pro was essentially unaffected by the substitution of alanine residues at either of the two positions (Fig. 6A). In contrast, an effect on eIF4GI cleavage was observed with both mutant proteins (Fig. 6B). With the Phe 20 -Asn 21 -Ser 22 alanine-scanning mutant, 50% cleavage of eIF4GI was obtained at 90 min; cleavage was complete at 180 min (Fig.  6B, left panel). The Tyr 32 -Ser 33 -Ser 34 alanine-scanning mutant was, however, more seriously handicapped in eIF4GI cleavage, with 50% cleavage not occurring during the time frame observed (Fig. 6B, right panel).
Examination of the eIF4GI binding by the corresponding GST fusion proteins (Fig. 6D) once again revealed that the significant delay in eIF4GI cleavage correlated with the binding of eIF4GI. Thus, the Phe 20 -Asn 21 -Ser 22 HRV2 2A pro alanine-scanning mutant was still able to bind eIF4GI (Fig. 6D,  lane 3), whereas almost no binding was observed with the Tyr 32 -Ser 33 -Ser 34 (Fig. 6D, lane 4) alanine-scanning mutant.
Thus, this experiment further defines the loop region Tyr 32 -Ser 33 -Ser 34 as part of the HRV2 2A pro binding domain, while pointing to only a minor role for Phe 20 -Asn 21 -Ser 22 .
Definition of the Amino Acids Recognized on eIF4GI by HRV2 2A pro -We showed previously (20) that HRV2 2A pro binds to the N-terminal cleavage fragment of eIF4GI generated by the L pro (residues 1-674). Furthermore, HRV2 2A pro also bound a fragment of eIF4GI comprising amino acids 220 -674 but did not bind one comprising amino acids 220 -669. The N-terminal boundary was, however, not defined. We therefore prepared suitable eIF4GI fragments (Fig. 7A), expressed the encoded proteins in RRLs using a coupled transcription/translation system, and investigated whether the GST-HRV2 2A pro C106A fusion protein could recognize them in a binding assay. Bound proteins were resolved by SDS-PAGE, and the gels were subjected to fluorography.
As a control, Fig. 7B shows the binding of the GST-2A pro C106A fusion protein to the in vitro translated eIF4GI fragment spanning amino acids 260 -674. In addition, the eIF4GI fragment 600 -739 was also recognized (Fig. 7C); the fragment was extended relative to the 260 -674 fragment to make it more easily detectable on PAGE. Extension of the deletion analysis was complicated by the presence of the binding site for eIF4E (eIF4GI amino acids 609 -623 (30)). The presence of eIF4E greatly accelerates the cleavage of eIF4G isomers by HRV2 2A pro (31), suggesting that its presence may also be required for HRV2 2A pro binding. Thus, the fragment comprising eIF4GI amino acids 640 -820 was only poorly bound, if at all (Fig. 7D). Although the interpretation of this last experiment is complicated by some binding to the GST protein alone (Fig. 7D, lane 2), it suggests that the presence of eIF4E bound to eIF4G significantly increases the binding by HRV2 2A pro and that the eIF4E binding site defines the Nterminal boundary of the HRV2 2A pro binding domain. DISCUSSION The interaction of human entero-and rhinoviral 2A pro with the eIF4G isoforms has, despite a great deal of research, remained controversial. Mutations that affect this reaction result in a reduction of virulence in these viruses; thus, a thorough understanding of the mechanism of this interaction is of fundamental importance.
The above experiments define the regions of the HRV2 2A pro and eIF4GI required for their interaction. In HRV2 2A pro , this domain involves some of the residues 17-35, which form a loop-strand-loop structure at the edge of the N-terminal domain ( Fig. 4 (24)). Of these residues, Leu 17 , Asp 35 , and together Tyr 32 , Ser 33 , and Ser 34 were shown to be crucial for binding eIF4GI. The role of Asp 35 is especially noteworthy, because this residue forms part of the active site. Because none of the mutations examined here, not even the D35E mutation, had a significant effect on self-processing, this indicates that the interaction with eIF4GI is much more sensitive to perturbation.
Comparison of the eIF4GI binding site on HRV2 2A pro with that on the FMDV L pro , which we determined earlier (20), reveals no similarity in the binding sites. In the FMDV L pro , the eIF4GI binding domain does not involve the residues from the catalytic triad. Instead, it is located more than 25 Å away from the active site and comprises part of the flexible CTE. Thus, the enzymes have evolved different structural domains to recognize the same cellular target.
Given the different structure of the binding domains, it is not surprising that the two enzymes bind to somewhat different fragments of eIF4GI (600 -674 for HRV2 2A pro and 640 -669 for FMDV L pro ). Nevertheless, two similarities can be found. First, binding by both enzymes is enhanced by the presence of  2 and 3); bound eIF4GI was detected by immunoblotting using an anti-eIF4GI antiserum. the eIF4E binding site on eIF4GI, with the HRV2 2A pro enzyme being more sensitive to the presence of eIF4E. It seems likely that the enzymes bind to the conformation generated when eIF4E binds to eIF4GI (32). Second, the C-terminal boundaries of the binding sites are separated by at least five amino acids, a similar number to the seven found between the cleavage sites of the two proteinases on eIF4GI.
Is the HRV2 2A pro binding domain for eIF4GI likely to be present in 2A pro encoded by other human rhino-or enteroviruses? Analysis of amino acid sequences shows that the equivalent residues to HRV2 2A pro 17-35 are highly conserved in A group human rhinoviruses. As mentioned in the results, the residues 20 -22 and 32-34 were chosen for alanine scanning analysis partly for this reason (24). However, little conservation between residues 17-35 of the HRV2 2A pro and HRV14 2A pro (the only representative of the B group human rhinoviruses sequenced to date) and between the HRV2 2A pro and those of enteroviruses is evident. Furthermore, little identity between the 2A pro of enteroviruses themselves is found in this region. Indeed, it was not possible to replace residues Leu 17 and Val 30 of HRV2 2A pro with those (arginine and glutamate) found in SVDV and most coxsackieviruses (Fig. 5, HRV2 2A pro L17R-V30E). This illustrates the complexity of this part of the picornaviral 2A pro and indicates that the binding abilities of other rhino-and enterovirus 2A pro will have to be determined experimentally. Possibly, differences in binding behavior may explain why the eIF4G isoforms are cleaved simultaneously in cells infected by HRV2 and HRV16 and at different times during infection by HRV14 and poliovirus (6, 16, 18).
Residues 19 -29 of HRV2 2A pro forming the loop between the catalytic His 18 and strand eI2 (Fig. 4), previously attracted attention during the structure determination (24). Indeed, this loop is found in two almost unrelated conformations in the two molecules in the asymmetric unit of the HRV2 2A pro crystal. A further indication of the singularity of the loop region 19 -29 can be seen on comparison with the Streptomyces griseus B proteinase and the 3C pro of PV (24,33). These proteinases both possess a minimal chymotrypsin-like fold and are structurally the closest relatives of 2A pro . Despite their minimal fold, however, they both possess two ␤-strands (dI and eI1) between the catalytic histidine and ␤-strand eI2. Strands dI and eI1 are lacking in HRV2 2A pro . Furthermore, the comparisons show that the ␤-strand eI2 of HRV2 2A pro is substantially truncated compared with that in S. griseus B proteinase and PV 3C pro . Thus, in these two enzymes, an equivalent structure to that adopted by residues 19 -29 of HRV2 2A pro is not possible. Furthermore, the superimposition of the residues 31-34, which in HRV2 2A pro comprise strand eI2, with the equivalent residues of the S. griseus and PV enzymes is very poor.
The ability of the HRV2 2A pro to recognize a substrate using a domain outside the active site is, however, not unique among chymotrypsin-like enzymes. At least two members of this family, thrombin and factor VII, can interact with substrates or inhibitors at sites away from the classic substrate binding site (34). Indeed, thrombin has two exosites, which mediate intermolecular interactions. Both are positively charged; exosite I recognizes fibrinogen, exosite II recognizes heparin (35,36). Interestingly, small two-domain protein inhibitors of thrombin have been shown to occupy both the active site and exosite I (e.g. hirudin (37)) or the active site and exosite II (e.g. hemadin (38)). These examples suggest that potent specific inhibitors of  35 Slabeled proteins translated in vitro from the indicated cDNA fragments of eIF4GI (input lanes 1, corresponding to a quarter of that used in each pull-down assay) were incubated with GST alone or GST-HRV2 2A pro C106A. Bound proteins were resolved by SDS-PAGE and detected by fluorography.
HRV2 2A pro might be designed that bind both to the active site and the eIF4GI binding exosite.
In summary, the N-terminal domain of HRV2 2A pro lacks 4 ␤-strands found in almost all other proteinases bearing a chymotrypsin fold. Nevertheless, we show here that the N-terminal domain has evolved to perform a vital interaction in the function of the enzyme and to make a significant contribution to determining virulence.