NMR analysis of the interaction of picornaviral proteinases Lb and 2A with their substrate eukaryotic initiation factor 4GII

Abstract Messenger RNA is recruited to the eukaryotic ribosome by a complex including the eukaryotic initiation factor (eIF) 4E (the cap‐binding protein), the scaffold protein eIF4G and the RNA helicase eIF4A. To shut‐off host–cell protein synthesis, eIF4G is cleaved during picornaviral infection by a virally encoded proteinase; the structural basis of this reaction and its stimulation by eIF4E is unclear. We have structurally and biochemically investigated the interaction of purified foot‐and‐mouth disease virus (FMDV) leader proteinase (Lbpro), human rhinovirus 2 (HRV2) 2A proteinase (2Apro) and coxsackievirus B4 (CVB4) 2Apro with purified eIF4GII, eIF4E and the eIF4GII/eIF4E complex. Using nuclear magnetic resonance (NMR), we completed 13C/15N sequential backbone assignment of human eIF4GII residues 551–745 and examined their binding to murine eIF4E. eIF4GII551–745 is intrinsically unstructured and remains so when bound to eIF4E. NMR and biophysical techniques for determining stoichiometry and binding constants revealed that the papain‐like Lbpro only forms a stable complex with eIF4GII551–745 in the presence of eIF4E, with K D values in the low nanomolar range; Lbpro contacts both eIF4GII and eIF4E. Furthermore, the unrelated chymotrypsin‐like 2Apro from HRV2 and CVB4 also build a stable complex with eIF4GII/eIF4E, but with K D values in the low micromolar range. The HRV2 enzyme also forms a stable complex with eIF4E; however, none of the proteinases tested complex stably with eIF4GII alone. Thus, these three picornaviral proteinases have independently evolved to establish distinct triangular heterotrimeric protein complexes that may actively target ribosomes involved in mRNA recruitment to ensure efficient host cell shut‐off.


Introduction
Successful viral replication requires rapid modulation of the infected cell's physiology. Interference with the host cell's machinery for protein synthesis is a mechanism favoured by positive strand RNA viruses to increase the efficiency of viral mRNA translation. The shut-off of host cell protein synthesis by picornaviruses, first observed in 1963 by Penman et al., 1 is a classic example. The effect has since been observed during replication of poliovirus (PV), coxsackievirus (CV) and human rhinovirus (HRV) (members of the enterovirus genus) and foot-andmouth disease virus (FMDV, a member of the aphthovirus genus). 2,3 The shut-off occurs one to two hours post-infection and before replication of the RNA genome takes place; thus, the reaction is efficiently carried out by proteinases translated from the RNA genome of the infecting particle. 4 Cleavage of the two isoforms of the cellular eukaryotic initiation factor (eIF) 4G, eIF4GI and eIF4GII [ Fig. 1(A)], observed in all enteroviruses and aphthoviruses so far examined, separates the domain that binds the polyA binding protein (PABP) and eIF4E (the capbinding protein) from that binding eIF4A (an ATPdependent RNA helicase), the protein complex eIF3 that binds to the 40S ribosomal subunit and the growth factor sensitive Mnk-1 kinase. 5,6 Consequently, the host cell fails to recruit capped mRNA to the ribosome; in contrast, viral RNA translation is initiated through the internal ribosome entry site (IRES) at the RNA's 5' end. 7 In FMDV, the shut-off of host cell translation through cleavage of eIF4G prevents synthesis of interferons and related cytokines, thus impairing host innate immunity. 8 Cleavage of eIF4G isoforms in FMDV is performed by the Leader proteinase (Lb pro ), a papainlike cysteine proteinase, with a globular domain of 155 amino acids and a C-terminal extension (CTE) of 18 amino acids. In enteroviruses, the cleavage is performed by the chymotrypsin-like 2A proteinase (2A pro ). The outcome of the cleavage of the two isoforms of eIF4G is the same; nevertheless, the cleavage sites and the mechanisms leading to the cleavage are not identical. Thus, on eIF4GI, the cleavages sites are located 7 amino acids apart whereas in eIF4GII, they are adjacent [ Fig. 1(A)]. 9 Prior to cleavage of the eIF4G isoforms, Lb pro binds to an exosite on the eIF4G isoforms located between 30 and 50 amino acids N-terminal to the cleavage site [ Fig. 1(A)]; Lb pro residues required for this bind-ing include C133 as well as 183-195 of the CTE. In contrast, the 2A pro of HRV2 binds to an ill-defined exosite region on eIF4G between 20 and 80 residues N-terminal to the cleavage site [ Fig. 1(A)]. The HRV2 2A pro residues involved in binding include residues L17-Y32 but have not been fully explored. Exosite binding by both Lb pro and HRV2 2A pro in GST pull-down experiments using full-length eIF4GI from rabbit reticulocyte lysates (RRLs) requires the presence of eIF4E; 10,11 furthermore, cleavage of both full-length and suitable fragments of eIF4G isoforms by both enzymes is stimulated in the presence of eIF4E in RRLs and in human 293 cells. 9,12,13 Indeed, with the Lb pro , cleavage of full-length eIF4GI in RRLs could be completely abrogated by addition of eIF4E-binding protein to complex the eIF4E. 13 The stoichiometry and structural basis of the stimulation of eIF4G isoform cleavage by eIF4E are unknown.
Structural information on the two proteinases as well as yeast and mammalian eIF4E is available. [14][15][16][17][18] For mammalian eIF4G, structural information [19][20][21] is limited to regions C-terminal to the cleavage sites of the picornaviral proteinases [ Fig.  1(A)]; nevertheless, the interaction of human eIF4E with human eIF4GI residues 557-646 has been characterised by isothermal titration calorimetry (ITC). 22 For yeast, the solution structure of residues 393-490 bound to yeast eIF4E has been determined, with residues 393-490 forming a ring around the N-terminus of eIF4E. 23 However, the yeast eIF4G is not cleaved by the HRV2 2A pro or the CVB4 2A pro24 and the sequence C-terminal to the eIF4E binding site shows low similarity to that found in mammalian eIF4G, putting into question the relevance of this structure for this region of the mammalian protein. Supporting Information Figure  S1 aligns the amino acid sequence of yeast eIF4G with those of human eIF4GI and eIF4GII (Supporting Information Fig. S1). The identity of the eIF4-GII 551-745 fragment to the corresponding regions of human eIF4GI and yeast eIF4GI is 48% and 20%, respectively.
To investigate the interaction of FMDV Lb pro as well as HRV2 and CVB4 2A pro with mammalian eIF4G and the eIF4G/eIF4E complex and the mechanism of stimulation of cleavage by eIF4E, we biochemically and biophysically characterised the formation of protein complexes between three picornaviral proteinases and the mammalian translation initiation factors and structurally characterised changes in eIF4GII and Lb pro and HRV2 2A pro by NMR. The eIF4GII fragment examined remained intrinsically unstructured even when bound to eIF4E. Lb pro or 2A pro from either HRV2 or CVB4 both form a ternary complex with eIF4GII and eIF4E; Lb pro and HRV2 2A pro recognise both eIF4G and eIF4E in their respective complexes. However, sLb pro , HRV2 2A pro and CVB4 2A pro interact differently with eIF4GII 551-745 and eIF4E.
Results eIF4GII 551-745 , eIF4E and sLb pro form a tight heterotrimeric complex To obtain structural information on the interaction of eIF4G with the picornaviral proteinases, we overexpressed and purified a fragment comprising residues 551-745 (eIF4GII 551-745 ) of the 1585 amino acid human eIF4GII protein. This fragment contains the eIF4E binding site 6,25 as well as the Lb pro and HRV 2A pro binding and cleavage sites [ Fig. 1(A)]. To ensure that this fragment behaves as the corresponding region of the full-length protein, we investigated its cleavage by purified recombinant picornaviral proteinases. For Lb pro , we used a naturally occurring shortened form (termed sLb pro ) lacking six amino acids at the C-terminus. 16,26 This deletion precludes protease dimerisation via interactions of the Cterminus of one molecule with the substrate binding site of a second and vice-versa, 27,28 thus simplifying complex formation and structural studies thereof. 29 For the 2A pro , we used wild-type HRV2 and CVB4 2A pro .
First, we wished to quantify the stimulation of picornaviral cleavage of eIF4GII 551-745 using the purified proteins; no data on the exact extent of stimulation are available. Thus, in an activity assay, cleavage of 1 mg (5 mM) of eIF4GII 551-745 was faintly visible in the presence of 1 ng sLb pro (concentration 0.125 mg/mL; 7 nM) following 90 min of incubation [ Fig. 1(B)]. However, the presence of 1 mg (5 mM) of recombinant full-length murine eIF4E purified from supernatants of E.coli lysates, cleavage of eIF4-GII 551-745 was already detectable after 10 min and 50% cleavage was reached between 20 and 30 minutes [ Fig. 1(B)]. N-terminal sequencing and mutational analysis both confirmed that cleavage of sLb pro on eIF4GII 551-745 was between residues G700 and S701, as determined previously. 9 To obtain cleavage of eIF4GII 551-745 in the absence of eIF4E, incubation with 40 ng (260 nM) of sLb pro for 60 minutes was required [Supporting Information Fig. S2(A)], whereas only 1 ng (7 nM) was required in the presence of eIF4E [ Fig. 1(B)]. The N-terminal cleavage product (cp N ) was however always only poorly visible. Subsequent experiments using residues eIF4GII 551-700 (corresponding to the N-terminal cleavage fragment) as substrate identified a further cleavage site for sLb pro in the presence of eIF4E (data not shown). To eliminate the possibility that the simple presence of eIF4E was affecting sLb pro cleavage, we constructed a version of eIF4-GII 551-745 bearing the mutations L629A and L630A (eIF4GII 551-745 L629A/L630A) in the eIF4E binding site. eIF4E failed both to form a complex with eIF4-GII 551-745 L629A/L630A (data not shown) and to stimulate cleavage of the mutated eIF4GII fragment, even when eIF4E was added in excess [ Fig. 1(C)]. Furthermore, we also examined the cleavage of the eIF4GII 653-745 fragment which lacks the binding site for eIF4E. No cleavage by sLb pro at 7 nM in the presence or absence of eIF4E was observed [ Fig.  1(D)]. When a 10-fold molar excess of sLb pro [Supporting Information Fig. S2(B)] was added, cleavage of this fragment could be observed; however, this level of cleavage was not increased by the addition of eIF4E [ Fig. S2(B)]. Together, these experiments show that formation of the eIF4GII:eIF4E complex stimulates sLb pro cleavage of eIF4GII 551-745 between 4 and 40 fold. This level was independent of the presence or absence of 5 mM of the cap analogue m7G(5')ppp(5')G (data not shown).
The eIF4GII 551-745 fragment was also cleaved by HRV2 2A pro [ Fig. 1(E)] and CVB4 2A pro (data not shown). However, 10 ng (80 nM) of enzyme were required in the absence of eIF4E to obtain detectable cleavage fragments after 60-90 minutes. The presence of an equimolar amount of eIF4E also stimulated HRV2 and CVB4 2A pro cleavage; however, the stimulation was less efficient than that seen with sLb pro , with 50% cleavage only being observed after 60 minutes [ Fig. 1(E) and data not shown].
In contrast to the stimulation of picornaviral proteinase cleavage, the presence of eIF4E reduced unspecific digestion of the eIF4GII 551-745 fragment by broad specificity proteinases, such as trypsin and elastase (Supporting Information Fig. S3), indicating that the stimulation of cleavage by eIF4E was a specific effect of viral enzymes.
Why is picornaviral cleavage of eIF4GII 551-745 in the presence of eIF4E more efficient? To answer this question, we first characterised the complexes formed by sLb pro with eIF4GII 551-745 and eIF4E. We used the proteolytically inactive variant sLb pro C51A (alanine in place of the catalytic cysteine) 28 Fig. S4). This value is 10 fold higher than that reported for the interaction of human eIF4E with human eIF4GI (residues 557-646) and yeast eIF4E with eIF4GI (residues 393-490). 22,23 This may be due to differences in experimental procedures or may reflect variations in the amino acid sequences of the two isoforms. Size-exclusion chromatography (SEC) analysis showed that sLb pro C51A forms a ternary complex with eIF4GII 551-745 /eIF4E [ Fig. 2(A)] with a K D between 20-40 nM as measured by ITC (Table I). sLb pro C51A, in contrast, does not form a complex with either eIF4GII 551-745 or eIF4E that is stable enough to be detected by SEC (Fig. 2, B and C) or ITC (data not shown).
Human eIF4GII 551-745 remains intrinsically disordered when bound to eIF4E and sLb pro To examine how eIF4GII 551-745 , eIF4E and sLb pro proteins interact, we tried to grow crystals of the eIF4GII 551-745 /eIF4E/sLb pro C51A complex for X-ray diffraction; to date, our efforts have remained unsuccessful, leading us to employ nuclear magnetic resonance (NMR). Accordingly, eIF4GII 551-745 was isotope labelled with nitrogen ( 15 N) or doubly labelled with nitrogen and carbon ( 15 N/ 13 C) expressed in E. coli and purified. High quality NMR spectra of eIF4GII 551-745 , with good signal to noise ratio, were observed at 48C ( Analysis of secondary shifts by SSP 30 and CSI 31 software confirms that eIF4GII 551-745 is indeed an IDP largely devoid of regular secondary structure. Furthermore, exploratory CPMG relaxation dispersion experiments of eIF4GII 551-745 do not reveal any clear dispersion profiles excluding the presence of either a global folded-unfolded equilibrium or substantial contributions from global conformational dynamics on a [ls-ms] timescale. In addition, certain residues in the eIF4E binding site (residues 620-626) are broadened beyond detection in eIF4GII 551-745 ; this may be indicative of [ms-ms] conformational exchange due to transient local secondary structure formation and/or fluctuating long-range contacts.
Very few ( 1 H, 15 N) chemical shift perturbations were observed between the spectra of eIF4GII 551-745 and the eIF4GII 551-745 /eIF4E complex, indicating that the binding of eIF4E did not induce any readily observable structural changes in the conformation of eIF4GII 551-745 . Instead, binding of eIF4E to eIF4-GII 551-745 markedly broadens signals of 72 residues (606-677) and attenuates, albeit to a lesser extent, the signals of residues from 678-735 (further away from the eIF4E binding site).
Global changes in the signal intensities in 15 N heteronuclear single quantum coherence (HSQC) spectroscopy spectra of eIF4GII 551-745 induced by complex formation are compared in Figures 3 and 4(A,B). The reasons for the signal broadening in the eIF4GII 551-745 /eIF4E are diverse, including conformational exchange and dynamics as well as some increase in the molecule mass of the resulting particle. The dynamic behavior of the intrinsically disordered eIF4GII 551-745 is complex and still under investigation. Conformational dynamics and exchange between different states of eIF4GII 551-745 can occur within the free or the bound form or between the two forms as has been observed for other IDPs such as the eIF4E-binding protein 4E-BP2 that binds to eIF4E through an extended binding interface. 32 The lack of observable 15 N relaxation dispersion effects would suggest that dynamics in free eIF4GII 551-745 are probably fast on an NMR time scale. In contrast, the low lM affinity of the eIF4GII 551-745 /eIF4E complex implies that conformational exchange between free and eIF4E bound eIF4GII is likely to be slow on the NMR time scale, as indicated by the complete absence of a second set of peaks for the bound form, again as was observed for 4E-BP2. 32 In summary, the direct observation of the bound eIF4GII 551-745 species is precluded by its dynamics and the conformational heterogeneity of a 'fuzzy' and extended ensemble, most likely interfacing with a large region on the surface of eIF4E. Nevertheless, we were able in the present study to use signal attenuation as a qualitative indication of the binding interaction. The average and standard deviation were calculated from three experiments. From the above NMR experiments, we conclude that human eIF4GII 551-745 remained intrinsically unstructured in complex with murine eIF4E and the resulting complex can be described as 'fuzzy' complex. This contrasts with the situation with the yeast fragment eIF4GI 393-490 , for which the NMR data indicates that this region of eIF4GI forms an ordered ring around the N-terminus of yeast eIF4E. 23 In the crystal structure of eIF4E with an eIF4GII peptide 14 corresponding to residues 622-633, a short a-helix was observed. In HNCA experiments (data not shown) using eIF4E and eIF4GII 551-700 , no such a-helix formation was noted for the observable residues (627-631), indicating a further difference to the yeast system. Interestingly, in our system, this region which comprises the canonical Y(X)4LU binding motif displays some consistent, but small secondary shifts, indicative of transient a-helix formation. SSP 30 software reveals a small (ca. 0.18) helical propensity for this region in an otherwise unstructured protein; this is consistent with a nascent a-helix.
In the eIF4GII 551-745 /eIF4E/sLb pro C51A complex, signals between residues 606-735 were further reduced in intensity; in addition, signals from 735 onwards were attenuated (Fig. 4, compare A and B). This again indicates complex formation, with the binding of sLb pro to the eIF4GII 551-745 /eIF4E complex affecting residues in both its binding and cleavage sites. Again, no evidence for discrete conformational changes through ( 1 H, 15 N) chemical shifts perturbations were observed. The specific changes in the eIF4-GII 551-745 /eIF4E/sLb pro C51A complex induced by sLb pro binding are illustrated by signals from six glycine backbone amides (Fig. 3, upper panel). The signal intensities of G580, G595 and G601 remain largely unchanged. However, in the eIF4GII 551-745 /eIF4E complex, the signal of G700 is reduced by about 50% and that of G674 is reduced beyond detection. In the eIF4GII 551-745 /eIF4E/sLb pro C51A complex, the signal of G700, the residue just prior to the sLb pro scissile bond, is now additionally absent, indicating an interaction with this residue. Together, these selected spectral changes indicate highly specific interactions between eIF4GII 551-745, eIF4E and sLb pro .
Reductions in signal intensities in the spectrum of eIF4GII 551-745 were also induced by the addition of the sLb pro C51A alone [ Fig. 4(C)]. They were, however, much less focussed and less pronounced than those seen with eIF4E, presumably as a consequence of the lower affinity between eIF4GII and sLb pro . Regions affected included 615-630 (the eIF4E binding site), 655-670 (the sLb pro binding site) and residues 680-702 (including the sLb pro cleavage site).
HRV2 2A pro and CVB4 2A pro form a stable complex with eIF4GII 551-745 /eIF4E We now examined whether the proteolytically inactive HRV2 2A pro C106S 24   [ Fig. 5(B)]. Examination of the complexes formed by 2A pro C106S or sLb pro C51A by NMR and SEC revealed appreciable differences. Thus, in the absence of eIF4E, addition of the HRV2 2A pro C106S to eIF4GII 551-745 [ Fig. 4(E)] broadens the signals of residues 638-658 (part of the 2A pro binding site) and the C-terminal region of eIF4GII 551-745 (residues 735-745) to a greater extent than sLb pro . This is in line with the larger molecular weight of HRV2 2A pro (it behaves as a dimer of 34 kDa, Table II) 34 observed in SEC-MALLS and its apparently higher affinity to eIF4GII 551-745 ; a K D of 86 mM was measured by ITC (Table I). In contrast, eIF4GII 551-745 /eIF4E/2A pro C106S complex formation with the 2A pro C106S led to a less dramatic loss of signal intensities of eIF4GII than that observed with sLb pro C51A.
Two further differences between the complexes observed with 2A pro C106S and sLb pro C51A were also observed. First, 2A pro C106S also formed a stable eIF4E/ HRV2 2A pro complex with eIF4E alone (K D 8 mM as measured by ITC, Table I), but not with eIF4GII 551-745 [ Fig. 5(C)]. Second, the affinity for the interaction of eIF4GII 551-745 /eIF4E with 2A pro C106S was approximately 100 times lower than that determined for the sLb pro C51A (Table I). These differences are reflected in the slower rates in vitro [ Fig. 1(B,E)] and in vivo of cleavage of eIF4GI observed for HRV2 2A pro than for Lb pro . 3,35,36 The inactive CVB4 2A pro C110A could also form a complex with the eIF4GII 551-745 /eIF4E complex [ Fig. 5(D)]. However, in contrast to the HRV2 enzyme, the CVB4 2A pro behaves as a monomer (Table II) and does not form a stable complex with either eIF4E or eIF4GII 551-745 . [Fig. 5(E,F)]. Furthermore, no interaction between eIF4E and CVB4 2A pro C110A could be measured by ITC. In contrast, the CVB4 2A pro interacts in ITC with the eIF4-GII 551-745 alone (Table I).
sLb pro also interacts directly with eIF4E The sLb pro does not form a stable complex with eIF4E; nevertheless, in view of the results with the HRV2 2A pro , we decided to search for similar interactions in the complex eIF4GII 551-745 /eIF4E/sLb pro . Accordingly, we examined the interaction of uniformly 15 N labelled sLb pro C51A/C133S/Q185R/E186K with eIF4GII 551-745 alone. This mutant, which fails to form a tight complex with eIF4G/eIF4E and whose eIF4G cleavage reaction is not stimulated by eIF4E [Supporting Information Fig. S5(C,F)], makes the same interactions with eIF4GII 551-745 as the wild-type (Supporting Information Fig. S6), indicating that the CTE is not required for interaction with eIF4GII 551-745 . The C-terminus of sLb pro (residues 183-195) is flexible and protrudes from the globular domain 27 ; therefore, we hypothesised that this region may be interacting with eIF4E. To examine this notion, we measured the 15 N-HSQC spectra of uniformly 15 N labelled sLb pro C51A in the presence of 1:1 molar ratios of eIF4GII 551-745 , eIF4E and as part of the eIF4GII 551-745 /eIF4E/sLb pro C51A complex (Fig. 6). In all experiments, the signals of 15 N sLb pro C51A were dramatically attenuated, indicating the formation of higher molecular weight complexes. This extensive line broadening precluded precise measurement of relaxation rates and quantification of exchange contributions.
With just eIF4GII 551-745 added, strong signal attenuation was observed across the entire sLb pro globular domain, but not for C-terminal residues 186-195. The line broadening effect is global, indiscriminate and diffuse, which resembles the behaviour of proteins with long and unstructured extensions. 37 With just eIF4E added, the line broadening is more pronounced; thus, intensities across the entire protein, including the C-terminal residues 183-191 are attenuated, indicating an interaction with low specificity, although heterogeneity and exchange dynamics may also contribute to this effect. In the eIF4GII 551-745 /eIF4E/sLb pro C51A complex, essentially all amide signals of sLb pro C51A were lost, except for the very C-terminal residues 194 and 195. As the effect is widespread across the whole surface of sLb pro C51A, it probably originates from the increase in particle size as well as including a contribution from exchange dynamics.
These results provide evidence for weak, low-to intermediate affinity interactions (too low to be accessible by ITC) between sLb pro C51A and eIF4E that include its C-terminal residues with K D values in the mid to high micromolar range (estimated by NMR, as the complex appears to still be in 'fast exchange' on the NMR timescale); these interactions help form and stabilise the triangular eIF4GII 551-745 /eIF4E/sLb pro C51A complex, that is one with pairwise contacts between all three pairs of proteins as defined by Popov. 38

C-terminal residues of sLb pro interact with eIF4E
Overall, the above NMR data indicate differing structural contexts and environments for eIF4GII 551-745 residues 551-605 and 606-745 as well as the close proximity and direct contacts of the C-terminal sLb pro residues to the eIF4E in the eIF4GII 551-745 /eIF4E/ sLb pro C51A complex. To investigate these insights further, we used paramagnetic relaxation enhancement (PRE) to sensitively probe local compaction by intramolecular PREs and transient tertiary contacts by intermolecular PREs. A paramagnetic spin-label (S-(1-oxyl-2,2,5,5-tetramethyl-2,5-dihydro-1H-pyrrol-3-yl)methyl methanesulfonothioate, MTSL), reaching out up to approximately 20 Å , was first attached to residue C638 of eIF4GII 551-745 and the NMR spectra of this protein alone were recorded. Most signals from residues 606-715 (which is also the region of interaction with eIF4E and sLb pro ) were lost, with those from 715-745 being affected to a lesser degree [Supporting Information Fig. S7(A)]. Thus, the residues whose NMR relaxation properties are affected by eIF4E or sLb pro are indeed located in a fairly compact globular core of approximately 20 Å which is still partially unstructured and located in an otherwise even more unstructured protein. The spin-label was then attached to residue C133 of sLb pro C51A/C125S/ C153S and the NMR spectra of eIF4GII 551-745 were measured in the presence of the spin-labelled sLb pro C51A/C125S/C153S [ Fig. S7(B)]. A strong attenuation of signals in the eIF4E binding site (628-631) of eIF4GII 551-745 and a partial signal attenuation of residues (638-655) between the eIF4E binding site and the Lb pro binding site of 50% was observed [ Fig. S7(B)], again indicating that these residues are within 20 Å of residue C133 of sLb pro .
We also wished to investigate whether the presence of the eIF4E binding site would be required for cleavage by sLb pro at the wild-type eIF4G cleavage site. For this, we replaced 10 residues of the cleavage site between Lb pro and VP4 with those of the eIF4GI cleavage site. 16 This site is cleaved by fulllength Lb pro but not by sLb pro when the encoded pro-tein is synthesised in RRLs. However, when residues 599-678 of eIF4GI N-terminal of the cleavage site are included in the substrate, sLb pro was also able to cleave this substrate. 16 To examine whether sLb pro could cleave when the eIF4E binding site was Figure 6. Overlaid spectra from 2D 15 N HSQC spectroscopy of native sLb pro C51A (grey), 1:1 sLb pro C51A and eIF4GII 551-745 (light blue), 1:1 sLb pro C51A and eIF4E (dark blue) and 1:1:1 sLb pro C51A, eIF4GII 551-745 and eIF4E (red). The eIF4GII 551-745 / eIF4E/sLb pro C51A complex was purified using SEC and concentrated, whereas for binary complex formation the individual proteins were titrated into 15 N sLb pro C51A. The assignments of the 15 N signals have been assigned previously 27 (BMRB Entry 15278). As a result of complex formation most residues of sLb pro C51A disappear upon the addition of the eIF4GII 551-745 or the eIF4E, since a globular, folded protein is more strongly affected by binding to an unstructured protein with a comparatively 'puffed up' shape. Due the resulting increased hydrodynamic drag of the complex the relaxation rates of the residues involved in the binding event are globally enhanced.
removed, we constructed a DNA fragment containing residues 29-195 of sLb pro C51A, residues 618-678 of eIF4GI (i.e., lacking the eIF4E binding site but maintaining the sLb pro binding site), 5-85 of VP4 and 1-23 of VP2. RNA from this construct was translated in RRLs in the presence of 35 S-methionine for 20 minutes followed by addition of unlabelled methionine and RNA for sLb pro . No cleavage of the synthetic polyprotein/eIF4GI substrate by sLb pro was observed over a period of 120 minutes [ Fig. 7(A), upper panel]. The synthesis of active sLb pro in RRLs was demonstrated by the cleavage of endogenous eIF4GI detected by immunoblotting of the samples [Fig. 7(A), lower panel]. As a further control for sLb pro activity, we tested the 599-678 substrate containing the eIF4E binding site reported in Steinberger et al. 16 and showed that it could be cleaved by sLb pro [Fig. 7(B), upper panel]. To control that the eIF4GI construct bearing residues 619-678 of the eIF4GI sequence could serve as a substrate, we incubated it with Lb pro and showed that it could be cleaved to the same extent as the 599-678 substrate [ Fig. 7(C,D), upper panels]. In all experiments, the synthesis of active Lb pro and sLb pro was confirmed by endogenous eIF4GI cleavage (Fig. 7, lower panels).

Discussion
The ability of enteroviruses and aphthoviruses to replicate successfully depends on the ability of the virus to shut down the synthesis of host proteins. FMDV lacking Lb pro fail to spread in the infected Uncleaved precursors Lb pro C51A eIF4GI 599-678 VP4/VP2 or Lb pro C51A eIF4GI 619-678 VP4/VP2 as well as cleavage products Lb pro C51A eIF4GI 599-674 , Lb pro C51A eIF4GI 675-678 VP4 5-85 /VP2  or Lb pro C51A eIF4GI 619-674 , Lb pro C51A eIF4GI 675-678 VP4 5-85 / VP2 1-23 are indicated. To observe cleavage on endogenous eIF4GI in the RRLs, 10 mL aliquots were analysed on 6% SDS gels followed by Western blotting using serum to detect the N-terminal part of the eIF4GI protein (lower pictures). Uncleaved eIF4GI and cleavage products (CP N ) are indicated. Negative controls without any RNA (2sub, 2prot) or comprising only RNA of the substrate protein (1sub, 2prot) are shown at the right of each gel. Protein standards are shown on the left. animal 8 whereas enteroviruses with mutations in 2A pro have small plaque phenotypes in cell culture. 39,40 It has long been recognised that, although the host cell shut-off achieved through the cleavage by Lb pro and 2A pro is the same, the variations in proteinase structure, locations of the cleavage sites and the times of onset of cleavage 3,[41][42][43][44] imply that the interactions with the eIF4GII/eIF4E complex differ considerably.
Several observations documented that the eIF4-GII 551-745 fragment analysed here behaved as the full-length eIF4G isoforms. First, complex formation between eIF4GII and eIF4E was measured by four techniques (SEC, SEC-MALLS, ITC (K D 40 nM) and NMR); mutation of the binding site for eIF4E on eIF4GII 551-745 eliminated complex formation. The observed K D of 40 nM compares well with those (around 4 nM) obtained for similar, but shorter, fragments of yeast and human eIF4GI. 22,23 These differences may reflect amino acid differences between the two isoforms. Second, conditions were found in which cleavage of eIF4GII 551-745 by sLb pro were observed only in the presence of eIF4E, as previously observed. 13 Third, a stimulation of eIF4GII 551-745 by eIF4E was observed by both sLb pro and HRV2 2A pro ; the results shown here extend this observation to the CVB4 2A pro .
The major hurdle to the study of the nature of the complexes formed by the proteinases with the initiation factors was the assignment of the NMR signals of the human eIF4GII 551-745 . This was achieved despite the intrinsically unstructed nature of the protein. In contrast to the yeast eIF4GI 393-490 , 23 human 4GII 551-745 remains intrinsically unstructured when bound to eIF4E. Indeed, neither discrete conformational changes nor a defined tertiary structure were observed on binding of murine eIF4E to the human eIF4GII 551-745 . Nevertheless, the NMR data for all complexes measured are consistent with the notion that eIF4G and eIF4E form a tight complex in which eIF4E contacts between eIF4GII residues 606 and 678. Whatever the nature of the interactions, this system should prove useful in screening for compounds that interrupt this interaction, a central control point in mammalian protein synthesis.
Given the lack of defined structure observed in eIF4GII 551-745 in the presence or absence of eIF4E, it is of interest to note that a similar phenomenon was previously observed in the binding of both human 4E-BP1 and 4E-BP2 to eIF4E. 32,45 Both 4E-BP1 and 4E-BP2 are IDPs that remain unstructured when bound to eIF4E. 32,45,46 The binding of the 4E-BP to eIF4E represents an important control point as it prevents eIF4G isoforms from binding to eIF4E or even displaces bound eIF4G from eIF4E. 47 Studies of the interaction between eIF4E and 4E-BP have revealed that both human 4E-BP2 and 4E-BP from D. melanogaster bind to two sites (i.e., in a bipartite manner) on eIF4E 32,47 ; for the D. melanogaster 4E-BP, one site is that at which the eIF4G isoforms bind whereas the other is on the lateral face comprising the C-terminus of helix a1 and loop 2. 47 It will be of interest to see whether the picornaviral proteinases also use the lateral face for their eIF4E interactions or a different part of the protein.
Together, the above experiments indicate that all three proteins in the ternary complexes of eIF4GII 551-745 with either sLb pro C51A, HRV2 2A pro C106S or CVB4 2A pro C110A interact with each other, forming heterotrimeric complexes that are also termed triangular complexes. 38,48 The triangular protein-protein interactions involved in sLb pro host cell shut-off are summarised in a model (Fig. 8). Residues 606-705 of eIF4GII 551-745 are shown as a relatively compact, preformed domain to which eIF4E binds. In the absence of structural evidence on the parts of murine eIF4E that interact with eIF4GII, we have drawn the model with eIF4GII contacting the N-terminus of eIF4E as in the yeast system. 23 Triangular ternary complex formation is stabilised by the interaction of the globular sLb pro domain with its binding site on eIF4GII and by interactions of residues C133, Q185 and E186 of the C-terminus (residues marked with side-chains) with eIF4E. Further support for the importance of this reaction was provided by the experiment in which cleavage by sLb pro at the eIF4GI cleavage site could be made dependent on the presence of the eIF4E binding site on eIF4GI (Fig. 7). Furthermore, the model explains why the presence of eIF4E can drive the processivity of eIF4G isoform cleavage as documented for Lb pro by Ohlmann et al. 13 in RRLs. Further evidence for the importance of eIF4E for the processivity is that, in actively translating ribosomes in RRLs, eIF4GI is efficiently cleaved during the synthesis of both L pro (50% cleavage in 4 min.) and HRV2 2A pro (50% cleavage in 15 min.). 49 Rau et al. 50 showed that 80% of eIF4GI is present in on the ribosome under such conditions and bound to eIF4E. This suggests that that cleavage of the eIF4G isoforms begins on the ribosome during synthesis of proteins from the RNA of the genome of the infecting particle, thus explaining the efficient cleavage of the isoforms in vivo. Indeed, cleavage of the eIF4G isoforms has been observed during PV under conditions that prevented the replication of viral RNA. 4 The HRV2 2A pro C106S also formed a triangular heterotrimeric complex with eIF4GII and eIF4E. However, the affinity of this 2A pro to the eIF4E/eIF4-GII complex was lower than that of sLb pro . Furthermore, the stoichiometry of the HRV2 2A pro C106S complex remained unclear as SLS measurements indicated an apparent molecular mass of 104-109 kDa (Table II). This corresponds neither to a 1:1:1 complex nor to any simple multiples of the components. Given this uncertain stoichiometry and the ill-defined nature of the binding site for HRV2 2A pro , 11 we cannot yet provide a model for HRV2 2A pro analogous to that for sLb pro shown in Figure 8.
A further interesting observation was the formation of a stable complex between HRV2 2A pro C106S and eIF4E, as evidenced by SEC, SLS and ITC measurements. This interaction helps to explain why previous attempts to exactly define the binding site of HRV2 2A pro on eIF4GI (600-674) or eIF4GII [613-685, Fig. 1(A)] met with little success. 11 In these experiments, performed in the presence of eIF4E, binding of HRV2 2A pro to eIF4G was lost upon deletion of the eIF4E binding site. Thus, it would appear that eIF4E is acting as a bridge to anchor the HRV2 2A pro to the eIF4G isoforms, indicating a fundamental difference to the model for sLb pro shown in Figure 8.
The apparent molecular mass determined by SLS on eIF4E/2A pro C106S complexes formed overnight was 38-40 kDa (Table II), indicating a 1:1 complex between 2A pro C106S and eIF4E. This would require dissociation of the 2A pro dimer observed in solution 34 (Fig. 5). Evidence for unusual molecular rearrangements in the formation of the eIF4E/ 2A pro C106S complex came when the SLS measurements were made after 10 min of complex formation, with an apparent molecular mass of over 300 kDa being observed (data not shown), compared to the 38-40 kDa observed on overnight complex formation. The dimerization interface for HRV2 2A pro has been predicted to involve amongst others residue C138, involved in a disulphide bridge between monomers in the crystal structure of HRV2 2A pro . 15 Investigation of a mutant HRV2A pro bearing the mutations C106S and C138S showed that the 1:1 complex with eIF4E was formed in this case after 10 min, suggesting that a reduction in the stability of the HRV2 2A dimer favours complex formation with eIF4E. Together, these results suggest that the monomer-dimer equilibrium of HRV2 2A pro C106S, together with known propensity of eIF4E to aggregate, 18 complicates the determination of the stoichiometry of the complex. Whatever the mechanism of complex formation, it will be of interest to investigate the interaction of the proteinases with eIF4E in transfected cells, using for instance fluorescently tagged proteins as has been done to investigate the dimerization state of a vaccinia virus protein. 51 The CVB4 2A pro C110A complex with eIF4-GII 551-745 /eIF4E clearly differed from that of HRV2 2A pro C106S. The stoichiometry was 1:1:1, as was that of the sLb pro . The CVB4 2A pro C110A is monomeric, as previously observed by Liebig et al.; 34 the protein did not form a complex with eIF4E alone. In addition, it was the only protein to form a measurable complex with eIF4GII 551-745 (Table I). CVB4 2A pro shares the 1:1:1 stoichiometry of binding with sLb pro . However, as we have no information on the precise interactions of CVB4 2A pro with eIF4GII 551-745 and eIF4E, the validity of the model in Figure 8 for CVB4 2A pro remains to be determined. HRV2 and CVB4 2A pro examined have the same fold 15,33 ; nevertheless, they share an amino acid identity of about only 40% (Supporting Information Fig. S8), with the majority of the identity in the C-terminal domain. One might therefore predict that the CVB4 and HRV2 2A pro behave differently toward eIF4E and the eIF4GII/eIF4E complex itself. These mechanistic differences may also explain why HRV2 cleaves both isoforms at the same rate in infected HeLa cells whereas in HRV14 and PV (both of which are more closely related to CVB4 than HRV2) infected HeLa cells, the eIF4GI isoforms are cleaved more rapidly than the eIF4GII isoforms. 35,41,42 Investigation of the importance of eIF4E interactions on the cleavage of full-length eIF4G isoforms during infection awaits the development of cell lines over-expressing recombinant eIF4G that can be infected by picornaviruses.
In summary, the data presented here show that FMDV sLb pro , HRV2 2A pro and CVB4 2A pro have each evolved different interactions to interact with the eIF4GII 551-745 /eIF4E complex, even though the viruses are all from the same family and that two of the proteinases share the same reduced chymotrypsin-fold.

Materials and Methods
Materials 13 C 6 -D-glucose and 15 N-ammonium chloride were purchased from Sigma-Aldrich, Germany and Euriso-Top, France, respectively. For over-expression of recombinant proteins in E. coli, the following vector systems were used: pET11d (NovagenV R ) was used for all eIF4GII constructs, active sLb pro and sLb pro C51A, pET3d (NovagenV R , formally known as pET8c) for HRV2 2A pro , HRV2 2A pro C106S, and CVB4 2A pro C110A VP1 8 and p proExHTA (InvitrogenV R ) for eIF4E.

Protein expression and purification
Fragments of eIF4GII (all having a C-terminal His-Tag) were expressed in BL21(DE3) grown in LB medium for biochemical analysis or minimal medium (M9) for NMR studies containing 13 C 6 -Dglucose as sole carbon source and 15 N-ammonium chloride as sole nitrogen source. Briefly, transformed bacteria were grown in 100 mL M9 medium overnight. 2 L expression cultures were then inoculated with the overnight culture and protein expression was induced at an OD 600 between 0.5 and 0.6 for 5-8 h at 378C. Following sonication, the sample was loaded on a 5 mL HisTrap HP (GE Healthcare), equilibrated with 50 mM Tris/HCl, 50 mM NaCl and 20 mM imidazole, washed with five column volumes and eluted with increasing imidazole concentration (500 mM). Pooled fractions were dialyzed into assay buffer (20 mM Hepes/KOH pH 7.4, 150 mM KCl, 1 mM EDTA and 5 mM DTT). For preparative SEC, a HiLoad 26/60 Superdex 75 pg column was used (GE Healthcare), equilibrated with assay buffer. The samples were concentrated to about 0.5-1 mM using Amicon Ultra Centrifugal Devices with a 10 kDa cut-off (Millipore). For expression of unlabelled eIF4-GII, expression was in LB medium and was terminated after 5 h at 378C.
Expression of sLb pro C51A (FMDV residues 29-195) was described previously. 28 PCR mutagenesis was used to restore the active cysteine residue C51. The expression and purification protocol of active sLb pro and mutants thereof was as described, 27 with the exception that plasmids were transformed in BL21(DE3)plysE and that assay buffer was used in the last step of purification.
Expression and purification of active HRV2 2A pro , HRV2 2A pro C106S as well as active CVB4 2A pro and CVB4 2A pro C110A VP1 8 were performed as described previously 33,34 with the exception that assay buffer was used in the final SEC purification step .
Recombinant murine eIF4E (N-terminal HisTag) was purified from strain T7 Express (New England Biolabs). Briefly, cells from 4 L were resuspended in 40 mL 50 mM Tris/HCl pH 8, 50 mM NaCl, 1 mM EDTA and 5% glycerol, lysed by sonication and centrifuged at 18,000 rpm for 30 min at 48C. The supernatant was then loaded onto a 5 mL HisTrap, equilibrated with 50 mM Tris/HCl, 50m M NaCl and 20 mM imidazole, washed with five column volumes and eluted with increasing imidazole concentration (500 mM). Pooled fractions were loaded onto a MonoQ HR 10/100 (GE Healthcare) column, equilibrated with buffer containing 50 mM Tris/HCl pH 8.0, 50 mM NaCl, 1 mM EDTA, 5 mM DTT and 5% glycerol, washed with five column volumes and eluted with rising NaCl concentrations (1M). Pooled fractions were dialyzed into assay buffer, followed by SEC using a HiLoadV R 26/60 SuperdexV R 75 pg (GE Healthcare).

In vitro cleavage assays
Cleavage assays were performed using purified recombinant eIF4GII 551-745 , murine eIF4E and FMDV sLb pro wt or mutants. 1 mg of substrate protein (equimolar eIF4GII 551-745 and eIF4GII 551-745 or eIF4E alone) were incubated with 1 ng of wt sLb pro or mutant sLb pro in assay buffer at 378C. After 0, 10, 20, 30, 60, and 90 minutes, reactions were stopped by the addition of SDS-PAGE 53 loading buffer and heated at 958C for 5 minutes. Reactions were visualized on a 17.5% SDS-PAGE gel stained with Coomassie brilliant blue. Gels were captured by a scanner (Canon) and uniform adjustments were performed using Adobe Photoshop. Densitometric analysis was performed by the ImageJ software.

NMR spectroscopy
Two and three-dimensional 1 H 15 N ( 13 C)-NMR experiments were performed at 48C on Varian Inova 500 MHz and 800 MHz and Varian Direct Drive 600 MHz spectrometers equipped with 5-mm triple resonance probes and pulsed field gradients. Samples containing 0.6-1 mM 15 N/( 13 C) eIF4GII 551-745 or other fragments of eIF4GII in assay buffer and 5-10% (v/v) D 2 O for field/frequency lock were used for the NMR measurements. For triple resonance experiments 0.1-0.2% (w/v) NaN 3 was added to the samples to inhibit bacterial growth. When possible, NMR measurements were performed on binary and ternary complexes that had been purified by SEC. The spectra were processed with NMRPipe 52 and analysed with Sparky (Goddard, unpublished).
The nitroxide spin-label MTSL was introduced by adding two to three times excess of MTSL to the protein, which was prior separated from DTT by a PD-10 desalting column (GE Healthcare). After incubation for 2-3 h, the protein was concentrated to 500 mL. Intramolecular paramagnetic relaxation enhancements (PREs) by nitroxide spin-label MTSL in single-cysteine C638.MTSL 15 N eIF4GII 551-745 or intermolecular PREs caused by unlabelled C133.MTSL sLb pro C51A/C125S/C153S were measured as intensity ratios of cross-peaks in 2D 15 N HSQC NMR spectra between oxidized and reduced forms. DTT was used to reduce the paramagnetic spin label. Individual cross-peak intensities were normalized using the most intense peaks in each data set.

Limited proteolysis
20 mg of eIF4GII 551-745 was subjected in the presence or absence of 20 mg eIF4E to limited proteolysis in assay buffer at 378C by the addition of a-chymotrypsin, trypsin, elastase, papain, subtilisin or endoproteinase Glu-C to final concentrations of 50, 5, and 0.5 mg/mL. The reaction was stopped after 60 minutes by adding 53 SDS sample buffer and immediately heated at 958C for 5 min. The samples were analysed on 17.5% polyacrylamide SDS-PAGE gels.

Analytical size-exclusion chromatography
Size-exclusion chromatography was performed with a HiLoadV R 16/60 SuperdexV R 200 pg (GE Healthcare) column, equilibrated with a assay buffer. 0.5 mg of pure protein or complexed proteins (10 minutes at 48C under gentle mixing) were injected together with 1 mg of aprotinin (6.5 kDa) using a 2 mL loop and eluted with 1 mL/min flow rate. Protein elution was monitored by following the absorption at 280 nm (mAU).

Static light scattering
For assessing the molecular weight and oligomeric state of eIF4GII 551-745 , eIF4E, sLb pro C51A and 2A pro C106S, a Superdex 200 10/300 column was equilibrated with assay buffer. The miniDAWN Trista light scattering instrument (Wyatt Technology, Santa Barbara, CA) was connected and 100 mg sample was injected on the gel-filtration column. Data analysis was performed using the manufacturer's software ASTRA.
Studying protein-protein interactions using isothermal titration calorimetry (ITC) ITC experiments were performed on a MicroCal ITC microcalorimeter. Proteins were dialyzed before the measurement in a buffer containing 20 mM Hepes/ KOH pH 7.4, 150 mM KCl and 1 mM EDTA filtered through a 0.22 mm filter. Following thermal equilibration at 258C, an initial delay of 60 seconds and a single 0.5 mL titrant injection, 20 serial injections of 2 mL of the titrant was added at an interval of 180 seconds into the stirred sample cell (200 mL) at a stirring rate of 750 rpm at 258C. The protein concentration in the cell was about 103 lower than protein titrant. Data analysis was performed using the Origin software package MicroCal.

In vitro transcription and translation
In vitro transcription reactions were performed as described. 16,54 In vitro translation reactions were performed as described in 16,27 with an RNA concentration of 14ng/mL.