The Major Head Protein of Bacteriophage T4 Binds Specifically to Elongation Factor Tu*

The Lit protease in Escherichia coliK-12 strains induces cell death in response to bacteriophage T4 infection by cleaving translation elongation factor (EF) Tu and shutting down translation. Suicide of the cell is timed to the appearance late in the maturation of the phage of a short peptide sequence in the major head protein, the Gol peptide, which activates proteolysis. In the present work we demonstrate that the Gol peptide binds specifically to domains II and III of EF-Tu, creating the unique substrate for the Lit protease, which then cleaves domain I, the guanine nucleotide binding domain. The conformation of EF-Tu is important for binding and Lit cleavage, because both are sensitive to the identity of the bound nucleotide, with GDP being preferred over GTP. We propose that association of the T4 coat protein with EF-Tu plays a role in phage head assembly but that this association marks infected cells for suicide when Lit is present. Based on these data and recent observations on human immunodeficiency virus type 1 maturation, we speculate that associations between host translation factors and coat proteins may be integral to viral assembly in both prokaryotes and eukaryotes.

The Lit protease in Escherichia coli K-12 strains induces cell death in response to bacteriophage T4 infection by cleaving translation elongation factor (EF) Tu and shutting down translation. Suicide of the cell is timed to the appearance late in the maturation of the phage of a short peptide sequence in the major head protein, the Gol peptide, which activates proteolysis. In the present work we demonstrate that the Gol peptide binds specifically to domains II and III of EF-Tu, creating the unique substrate for the Lit protease, which then cleaves domain I, the guanine nucleotide binding domain. The conformation of EF-Tu is important for binding and Lit cleavage, because both are sensitive to the identity of the bound nucleotide, with GDP being preferred over GTP. We propose that association of the T4 coat protein with EF-Tu plays a role in phage head assembly but that this association marks infected cells for suicide when Lit is present. Based on these data and recent observations on human immunodeficiency virus type 1 maturation, we speculate that associations between host translation factors and coat proteins may be integral to viral assembly in both prokaryotes and eukaryotes.
Elongation factor (EF) 1 Tu is the major host translation factor in Escherichia coli responsible for delivering charged tRNAs to the ribosome for protein synthesis. During its role in translation, EF-Tu interacts with a number of molecules, including the nucleotides GTP and GDP, aminoacylated tRNA molecules, mRNA-programmed ribosomes, and the nucleotide exchange factor EF-Ts (1). In addition to forming complexes with these molecules, EF-Tu is also known to serve as one of the subunits in the bacteriophage Q␤ replicase (2) and has recently been shown to have chaperone-like activity in vitro, promoting the renaturation of some denatured proteins (3). Hence, EF-Tu is capable of interacting with a variety of macromolecules and serving more than one biological function. In addition to these associations, EF-Tu also undergoes posttranslation modifications, including methylation at a specific lysine residue, which appears to attenuate its GTPase activity and is linked to the growth phase of the cells (4), and phosphorylation, although the physiological role of this modification is uncertain (5).
EF-Tu is also the target of a bacteriophage exclusion system (6). Bacteriophage exclusion is a defense mechanism in which bacteria commit altruistic suicide in response to infection, thereby preventing propagation of the phage (7,8). Although distinct to apoptosis in eukaryotes, similarities have been drawn between this form of suicide and programmed cell death events in multicellular organisms (9). Generally, these exclusion systems are mediated by the action of one or more nonessential proteins encoded by prophages, plasmids, or transposons. One of the best studied is that of E. coli K-12 strains, which exclude T4 bacteriophage as well as other T-even phages through the action of a metalloprotease called Lit (Late Inhibitor of T4), encoded by the defective prophage e14 (10). Following T4 phage infection of a Lit-containing cell, EF-Tu is specifically cleaved by Lit between Gly 59 and Ile 60 in the RGITI motif of the effector I region resulting in the inhibition of translation (6). This region, which is involved in co-ordinating the ␥-phosphate of GTP as well as the bound magnesium ion (11 and references cited therein), is conserved in other translation elongation factors such as EF-G and common to the homologous eukaryotic translation factors EF-1␣ and EF-2. The fact that neither EF-G nor heat-inactivated EF-Tu are cleaved by Lit suggests that the three-dimensional structure of EF-Tu, and not just the primary sequence of the proteolysis site, is required for cleavage and that the cleavage reaction is highly specific, implying the involvement of other regions of the translation factor (12).
Proteolysis of EF-Tu by Lit is activated by the appearance in the cell of a short peptide determinant approximately 29 amino acids long, beginning about 100 amino acids in from the N terminus of the unprocessed major head protein of the infecting T-even phage, gp23 (56 kDa). This short peptide sequence was named the Gol peptide sequence, because it was first identified by mutations in gp23 that allow the phage to Grow On Litproducing bacteria (13). The Gol peptide sequence is highly conserved in the head protein of all T-even phages that have been characterized (14). It has been possible to reproduce the entire phage exclusion system in a purified three-component system containing EF-Tu, Lit protease, and a chemically synthesized 29-amino acid peptide representing the minimal gol region sequence. In this system, Lit-mediated cleavage of the Gly 59 -Ile 60 bond of EF-Tu is dependent on the addition of the Gol peptide, with no additional viral or bacterial proteins being required (12).
The absolute dependence for proteolysis on the Gol peptide in the purified system raises the question of how the activation occurs. Few peptide-activated proteases have been described in the literature, although a well known example occurs in the maturation of adenovirus in which an 11-residue C-terminal peptide of the viral coat protein activates the thiol protease Ad2 through thiol-disulfide exchange to cleave the same coat protein (15). In contrast, the Gol peptide acts in trans, and, although the peptide contains a single cysteine, this residue is not required for activation of Lit-mediated cleavage of EF-Tu, indicating a mechanism distinct to that of Ad2 (12).
Here we report the results of a number of different experimental approaches to show that the phage-derived Gol peptide binds to EF-Tu. We also show that this complex forms an activated substrate for Lit and that this association is sensitive to the identity of the nucleotide bound to the translation factor. Because the association of viral capsid proteins with host translation factors has also been demonstrated in eukaryotic systems, we discuss how this type of association may be a hitherto unrecognized requirement in the viral infection of prokaryotic and eukaryotic cells.
Construction of the Gol Peptide S-Tag Fusion-The plasmid, pET30PZ1, which was used to synthesize Gol peptide in vivo, contains a 160-base pair PZ1 fragment extending from the natural PstI site in the T4 gp23 gene to an HindIII linker inserted at the end point of the ⌬1 deletion downstream of the gol region (17) and then ligated into the PstI and HindIII sites of pET30b (Novagen). Once induced, the construct directs the synthesis of a 12-kDa polypeptide containing the gol region from gp23 close to the C terminus and an S-Tag at the N terminus. Overexpression of active Gol peptide was confirmed by its inability to transform E. coli cells containing Lit and by the in vitro activation of EF-Tu cleavage by Lit.
Lit-mediated EF-Tu Cleavage Assays-The concentrations of Lit, EF-Tu⅐GDP (or EF-Tu⅐GTP) and Gol peptide in the cleavage assays were generally 0.2, 2, and 5 M, respectively, in a buffer containing 50 mM Tris-HCl, pH 8.0, 10 mM ␤-mercaptoethanol, 2 mM MgCl 2 , and 10 M GDP (or 10 M guanosine 5Ј-o-(3-thiotriphosphate)). All reactions were generally incubated at 30°C for up to 60 min. The reactions were initiated by the addition of Gol peptide and stopped by the addition of EDTA to a final concentration of 10 mM. The reaction products were analyzed by 10% SDS-polyacrylamide gel electrophoresis (PAGE), and EF-Tu cleavage was quantified by densitometry using a Bio-Rad GS690 imaging densitometer. The rates of EF-Tu cleavage were determined as a percentage of the starting EF-Tu using the software Multi-Analyst (Bio-Rad).
GSH Affinity Chromatography-Cell extracts containing either the S-Tagged Gol peptide or the GST-EF-Tu fusion protein were prepared separately from E. coli JM109DE3/pET30PZ1 or DH5/pGEX2T-tufA (18), respectively. Cultures (200 ml) were grown at 37°C to an A 625 nm of 0.4 before being transferred to 23°C for 30 min where protein expression was induced by the addition of isopropyl-1-thio-␤-D-galactopyranoside to a final concentration of 0.1 mM. The cultures were then grown for a further 3 h at 23°C before being harvested by centrifugation. Preparation of the cell extracts prior to loading onto the glutathione column involved the resuspension of the cell pellets in 5 ml of 50 mM Tris, pH 7.5, 150 mM KCl, 5 mM MgCl 2 followed by sonication and finally centrifugation at 25,000 ϫ g for 30 min to remove any particulate matter.
To detect interactions between the Gol peptide-S-Tag and GST-EF-Tu fusion proteins, the extracts were mixed and incubated at 30°C for 30 min. The combined extracts were filtered through a 0.45-m Gelman filter prior to layering onto a 2.5-ml glutathione Redipack column. Following application, the column was washed with 3 volumes of the above buffer before the retained proteins were eluted from the column with reduced glutathione according to the manufacturer's instructions. Fractions were collected and assayed for the Gol peptide-S-Tag fusion protein by spotting 3 l on a nitrocellulose filter and developing the filter with an S-Tag detection kit (Novagen). Elution of the GST-EF-Tu fusion protein was monitored by 10% SDS-PAGE.
EDC Cross-linking-All cross-linking reactions were performed in 50 mM MOPS, pH 7.9, 2 mM MgCl 2 , 10 M GDP (or 10 M guanosine 5Ј-o-(3-thiotriphosphate)) and generally contained 2 M EF-Tu with a 100-fold excess of Gol peptide. EF-Tu and Gol peptide were mixed and incubated at 18°C for 30 min before the addition of EDC to a final concentration of 2.5 mM. The reactions were then incubated at 18°C for an additional amount of time, as described in the figure legend, prior to their quenching by the addition of Tris-HCl, pH 8.0, to 100 mM final concentration. The samples were analyzed by 10% SDS-PAGE.
Tritiation of the Gol Peptide-The Gol peptide ( 1 AVMGMVRRAIPN-LIAFDICGVQPMNSPTG 29 ), corresponding to residues 93-122 of the full-length gp23 protein, contains a single cysteine residue (Cys 19  H]Gol were pooled, lyophilized, and reconstituted into 1 ml of 50 mM MOPS, pH 7.9. The concentration and, hence, the specific activity of the peptide (15.8 Ci/mol peptide) were determined. The efficiency of radiolabeling was estimated using 5,5Ј-dithiobis(2-nitrobenzoic acid) to quantify the amount of unmodified cysteine residues (19). Generally, the Gol peptide was greater than 80% labeled.
Tryptic Digests-Cross-linking reactions between the tritiated Gol peptide and EF-Tu and all the relevant controls (i.e. [ 3 H]Gol and EF-Tu alone) were performed as described previously except that the radiolabeled peptide was in a 20-fold molar excess to that of EF-Tu. Following gel-filtration of the cross-linked complex to remove excess peptide, appropriate fractions containing both radioactivity and EF-Tu were pooled and concentrated using an Amicon Centricon 10 microconcentrator. The specific activity of the [ 3 H]Gol⅐EF-Tu complex was determined (5.9 nCi/nmol complex). This material was then proteolytically digested by the addition of 2% trypsin followed by incubation at 37°C for 2 h. The proteolysis of bacterial EF-Tu by trypsin has been reported from a number of laboratories (20,21). The conditions used in our study were similar to these and ensured the complete digestion of EF-Tu and the generation of small peptide fragments. The digestions were stopped by the addition of soya bean trypsin inhibitor to a 4-fold molar excess of that of trypsin. Control reactions were also treated in a similar manner. The samples were analyzed by 10% SDS-PAGE and reverse-phase HPLC.
Peptide Mapping by HPLC-Peptides generated by trypsin digestion were HPLC-purified on a Vydac C8 reverse-phase column that had been equilibrated in 90% water/10% acetonitrile containing 0.1% trifluoroacetic acid. The column was developed using a gradient up to 65% of 0.1% trifluoroacetic acid in 95% acetonitrile/5% water at a flow rate of 1 ml/min over 35 min. For the digestion and separation of the [ 3 H]Gol⅐EF-Tu complex, three regions of the peptide map, at retention times 16.5, 18.5, and 20 min, respectively, were collected (pools A, B, and C) and rechromatographed on a Vydac C18 reverse-phase column equilibrated with 85% water/15% acetonitrile containing 0.1% trifluoroacetic acid. The column was developed with a gradient up to 30% of 0.1% trifluoroacetic acid in 95% acetonitrile/5% water at a flow rate of 1 ml/min over 35 min. Single radioactive peaks from all three pools were collected, lyophilized, and submitted for N-terminal sequencing. For all HPLC steps, the recovery of applied radioactivity was greater than 95%, and both the C8 and C18 columns were run at 18°C while monitoring the absorbance at 220 nm .
Conversion of EF-Tu⅐GDP into EF-Tu⅐GTP-To convert the EF-Tu⅐GDP form into EF-Tu⅐GTP, 20 M EF-Tu⅐GDP was incubated with 200 M guanosine 5Ј-o-(3-thiotriphosphate), 200 M phosphenolpyruvate, and 0.05 g/l pyruvate kinase for 30 min at 37°C, essentially as described by Fasano et al. (22). The reaction was quenched on ice.
EF-Tu GTPase Assays-Determination of the intrinsic GTPase ac-tivity of EF-Tu was based upon the measurement of ␥-32 P liberated from [␥-32 P]GTP upon its hydrolysis as described by Mesters et al. (23).
All assays were carried out in 64 mM Tris-HCl, pH 7.6, 80 mM NH 4 Cl, 10 mM MgCl 2 , 10 mM ␤-mercaptoethanol, 83 M phosphenolpyruvate, 17 g/ml pyruvate kinase, 1 M EF-Tu, and Gol peptide as indicated in the figure legend. For the rate experiments, a 270-l solution containing EF-Tu⅐GDP, phosphenolpyruvate, pyruvate kinase, and 500 M Gol peptide was preincubated at 37°C for 30 min. Then, [␥-32 P]GTP (1200 dpm/pmol), also preincubated with phosphenolpyruvate and pyruvate kinase, was added to give a final volume of 300 l and a final concentration of [␥-32 P]GTP of 5 M. Aliquots of 30 l were removed from the reaction at the times indicated in the figure legend and quenched by the addition of silicotungstate in 1 mM H 2 SO 4 to a final concentration of 3.7 mM. Following the addition of ammonium molybdate in 2 M H 2 SO 4 to a final concentration of 0.74% (w/v), the liberated ␥-32 P, as a dodecamolybdate complex, was extracted into a 1:1 toluene/butan-2-ol mixture (24) and counted on a scintillation counter. For the titration experiments, the concentration of Gol peptide in separate 30-l reactions was increased as indicated in the figure legend. The reactions were then preincubated at 37°C for 30 min before the addition of [␥-32 P]GTP, then they were incubated for a further 60 min before being quenched and analyzed as described above.

Detection of Gol Peptide-EF-Tu Complexes by Affinity
Chromatography-To monitor the association of EF-Tu with Gol peptide, separate extracts were prepared of bacterial cells containing a glutathione S-transferase-EF-Tu (GST-EF-Tu) fusion protein and a Gol peptide-S-Tag fusion (see "Experimental Procedures"). In this construct, a 60-amino acid peptide of the gp23 protein that spanned the Gol region was fused to an S-Tag. The extracts were mixed and loaded onto a glutathione-Sepharose column. Fig. 1A shows that the GST-EF-Tu fusion protein bound to the column along with some of the Gol peptide-S-Tag fusion, which could then be eluted when the GST-EF-Tu fusion protein was stripped from the column by the addition of excess glutathione to the wash buffer (lanes d, e, and f). Much less S-Tagged peptide was retained in identical control experiments in which the S-Tag was not fused to the Gol sequence (Fig. 1B) or if the GST protein was not fused to EF-Tu (Fig. 1C). These data demonstrate that the peptide was retained on the column due to a direct interaction between the Gol peptide and EF-Tu rather than through nonspecific interactions between, for example, GST and the S-Tag. Overall, these results indicate that the Gol peptide is retained on the column by virtue of its interaction with EF-Tu.
Attempts were made to recapitulate these chromatographic experiments using the chemically synthesized 29-residue Gol peptide. However, we were unable to detect retention of EF-Tu on a Gol peptide affinity column nor indeed retention of the peptide to an EF-Tu affinity column (data not shown). It is clear from estimates of the binding constant for the Gol peptide-EF-Tu complex (see below) that the synthetic peptide binds weakly to EF-Tu, which explains why complexes could not be detected chromatographically using the peptide. The fact that binding could be detected for the 60-mer version of the Gol peptide in the GST and S-Tag experiments implies that the extra flanking amino acids from gp23 likely increase the binding affinity for EF-Tu. Lastly, we could not detect Lit retention on the EF-Tu or Gol-peptide affinity columns. This implied that binary complexes between Lit and the other components did not occur or were too weak to be detected. 2 The Gol Peptide-EF-Tu Complex Is the Substrate for Lit Cleavage-The chromatographic experiments suggested that the Gol peptide formed a complex with EF-Tu. However, these experiments did not address the question of whether the resulting complex is relevant to the mechanism by which the Gol peptide activates proteolysis of EF-Tu by Lit. To investigate this, we employed chemical cross-linking, which is capable of capturing even weakly bound complexes, and, to simplify interpretation, we focused on the chemically synthesized Gol peptide and purified EF-Tu. Using the bifunctional, zero-length cross-linking reagent EDC, it was possible to specifically crosslink the synthetic 29-residue Gol peptide to EF-Tu bound with GDP (Fig. 2, lane 1) demonstrating that a complex is detectable. A cross-linked adduct was observed of ϳ47 kDa, corresponding to EF-Tu (44 kDa) covalently bound with a single 2 R. Bingham and C. Kleanthous, unpublished observations.

FIG. 1. A Gol peptide-S-Tag fusion protein can bind Sepharose
4B glutathione-immobilized GST-EF-Tu. Extracts were made of each construct as described under "Experimental Procedures," mixed, and applied to a Sepharose 4B glutathione column, and bound material was eluted with glutathione. For each experiment, lane a represents the applied extracts, lane b, the column flow-through, and lanes c-e, f, or g the first three, four, or five fractions eluted from the column. The results for the detection of the S-Tag using an S-Tag-specific binding assay are shown below each lane. A, S-Tag-Gol peptide fusion extract mixed with GST-EF-Tu fusion extract. A significant amount of S-Tag signal is observed in the dot blot. B and C, control experiments in which GST-EF-Tu was mixed with the S-Tag peptide (no Gol control) and GST was mixed with S-Tag-Gol peptide (no EF-Tu control) prior to their application to the column. In both cases, little or no S-Tag is retained, indicating the need for both EF-Tu and the Gol sequence for interaction on the column. copy of the 3-kDa Gol peptide (cross-linking efficiency ϳ20% by laser densitometry). Increasing concentrations of either the cross-linking reagent or Gol peptide did not yield species of higher molecular mass (data not shown), suggesting that a 1:1 complex was being cross-linked. Cross-linking was also used to probe for Gol peptide binding to Lit, but no cross-linked adducts could be detected (data not shown).
Having demonstrated cross-linking between EF-Tu and the Gol peptide, the functional relevance of this complex could then be tested by incubating the mixture of cross-linked and uncross-linked EF-Tu with the Lit protease after excess peptide had been removed by gel-filtration chromatography (see "Experimental Procedures"). Only the EF-Tu to which the Gol peptide was covalently attached was a substrate for Lit cleavage as deduced by the loss of the upper band at 47 kDa but not the lower band at 44 kDa (Fig. 2, lane 2). The specific cleavage of cross-linked EF-Tu is also apparent from the size of the cleavage fragments. Lit-mediated proteolysis of native EF-Tu ordinarily yields two cleavage fragments of approximately 37 kDa (amino acids 60 -393) and 7 kDa (amino acids 1-59), respectively, with only the larger fragment being visible on 10% SDS-polyacrylamide gels. Proteolysis of the cross-linked complex, however, gave rise to a larger cleavage fragment of ϳ40 kDa, corresponding to the 37-kDa fragment covalently bound to the Gol peptide. These results not only demonstrate the high specificity of cross-linking between the Gol peptide and EF-Tu, because the complex still acts as a substrate for Lit proteolysis, but also shows that the cross-linking between Gol peptide and EF-Tu is to amino acids C-terminal to the cleavage site (Gly 59 -Ile 60 ). To demonstrate that EDC does not inactivate the remaining native EF-Tu, excess Gol peptide was added along with Lit. Then the uncross-linked EF-Tu was also digested, generating the normal 37-kDa fragment in addition to the 40-kDa cross-linked fragment (Fig. 2, lane 3).
Gol Peptide Contacts Domains II and III of EF-Tu-Because the Gol peptide is specifically cross-linked to EF-Tu at a site where it can activate cleavage by Lit, mapping the sites of cross-linking on EF-Tu should begin to reveal where the Gol peptide binds. We therefore identified the positions of some of the cross-links by peptide mapping (see "Experimental Procedures"). These experiments capitalized on two observations. First, it was possible to specifically radiolabel the Gol peptide at its single cysteine residue (Cys 19 ) by chemical modification using tritiated N-ethylmaleimide ([ 3 H]NEM). This modification did not inhibit the ability of Gol to activate Lit-mediated cleavage of EF-Tu nor its cross-linking with EF-Tu (data not shown), in agreement with our previous data on the ability of a Cys 19 3 Ala mutant of the Gol peptide to sustain Lit cleavage of EF-Tu (12). Second, radioactive Gol peptide cross-linked to EF-Tu was not degraded by trypsin, which allowed the isolation of EF-Tu peptide fragments that retained Gol-associated radioactivity. The Gol peptide contains two arginines (at positions 8 and 9 in the 29-residue peptide; see "Experimental Procedures") both of which are readily digested by trypsin in the unbound peptide. However, no tryptic fragments corresponding to cleaved Gol peptide were observed for [ 3 H]Gol cross-linked to EF-Tu, implying protection of the peptide proteolysis sites when bound to the translation factor (data not shown).
Following cross-linking of [ 3 H]Gol to EF-Tu by EDC and the removal of excess peptide by gel-filtration, the pooled and concentrated [ 3 H]Gol peptide-EF-Tu cross-linked complex was digested with trypsin and fractionated on a reverse-phase C8 column developed with an acetonitrile gradient (described under "Experimental Procedures"). Peptides eluted between 5 and 30 min, with three distinct radioactive pools eluting at 16.5, 18.5, and 20 min (designated pools A-C, respectively) with little starting material remaining. The three pools were subjected to further purification on a reverse-phase C18 column and these data are shown in Fig. 3 (A-C). Homogeneous radiolabeled peptide fractions were isolated from each of these pools (S1 from pool A, S2 and S3 from pool B, and S4 and S5 from pool C) and were subjected to N-terminal sequencing during which the eluted radioactivity was monitored after each cycle.
Of the purified peptide fractions from the C18 column, S2 and S5 did not yield clear data and so were not studied further; S2 contained trypsin sequence as well as an unassigned sequence, whereas S5 gave multiple amino acids during each sequencing cycle that could not be assigned to EF-Tu, trypsin, or the Gol peptide. In both cases the radioactivity associated with the peptide eluted throughout all the sequencing cycles.
Interpretable sequencing data were obtained for the remaining three peptides S1, S3, and S4, and these data are shown in Table I. Peptide fraction S1, isolated in three independent experiments, corresponded to residues 326 -333 of E. coli EF-Tu from domain III, but no discernible Gol sequence could be identified. The radioactivity associated with each sequencing cycle of this peptide was only just above background (normally ϳ25 cpm) and eluted across all the residues, making sequence-specific assignment of the cross-linking site impossible. In contrast, peptide S3 was derived from residues 254 -261 of domain II and contained a single unassigned residue (Cys 256 ), which also corresponded to the elution position of the radioactivity ( Table I). The absence of Gol peptide sequence infers that, although Gol is bound (because radioactivity is retained), its N terminus is most likely blocked. Peptide S4 contained sequence corresponding to both EF-Tu (residues 271-278 of domain II) and Gol peptide (residues 1-3). After three cycles, however, the Gol sequence was lost and this corresponded to the appearance of a single unassigned residue in the EF-Tu sequence (Glu 273 ) and the elution of all the associated radioactivity ( Table I). The data indicate that Cys 256 and Glu 273 of domain II are sites of Gol-peptide cross-linking (both are consistent with the known chemistry of EDC) as well as residues 326 -333 of domain III, although no single residue can be identified in this domain. It is important to note that the three identified sequences all lie C-terminal to the Lit cleavage site at Gly 59 -Ile 60 , consistent with the original cross-linking data in Fig. 2.
Gol Peptide Inhibits the Intrinsic GTPase of EF-Tu-EF-Tu is a member of the GTPase family of proteins, and this enzymatic activity is central to its role as a translation factor. Consequently, molecules that affect its properties as a trans- lation factor generally also affect its GTPase activity (25,26). Because the Gol peptide binds EF-Tu, we analyzed its effect upon the intrinsic GTPase activity of EF-Tu. The inset to Fig. 4 shows that in the presence of 500 M Gol peptide the GTPase activity of EF-Tu is inhibited by Ͼ50%. Control experiments with 500 M bovine serum albumin or unrelated peptides did not inhibit the GTPase activity, showing that the effect induced by the Gol peptide was specific (data not shown). Fig. 4 also shows that this inhibition is dependent upon the concentration of Gol peptide and saturates at a concentration of Ͼ600 M. From these data it was possible to estimate a K i value of ϳ300 M, demonstrating a weak interaction between the 29-residue synthetic Gol peptide and EF-Tu, although as indicated previously this interaction may be stronger in the intact gp23 of the T4 phage.
EF-Tu⅐GDP is the Preferred Conformer for Lit Proteolysis-In the absence of other effector molecules, EF-Tu exists in two distinct conformations that are determined by the identity of bound nucleotide, GDP or GTP. Although structural data for E. coli EF-Tu have been published for both forms, only data for thermophilic EF-Tu factors (which share a high degree of sequence identity with the E. coli protein) are available in the structure data bases (27, 28) (Fig. 5). The GDP-bound form of the protein adopts an "open" conformation while in the GTP-bound form (in which the nonhydrolyzable analogue guanosine 5Ј-O-(␤,␥-imidotriphosphate (GDPNP) is used) EF-Tu is more compact with its three domains coming together to form the binding site for aminoacylated tRNA. The domain rearrangements are linked directly to the conformational changes that take place around the nucleotide binding site to accommodate the ␥-phosphate of the GTP, and it is within this site that Lit-mediated cleavage occurs (Fig. 5). Moreover, the accessibility of this cleavage site in response to nucleotide binding mirrors that of the molecule as a whole with the Gly 59 -Ile 60 bond exposed in the open GDP-bound form but sequestered in the "closed" GTP-bound form.
To gain further insight into the nature of the EF-Tu substrate for proteolysis by Lit, we tested the effect of nucleotides on the Lit-mediated cleavage of EF-Tu. Fig. 6 shows that GDPbound EF-Tu (EF-Tu⅐GDP) is cleaved at an approximate 2-fold faster rate than that for the GTP-bound form (EF-Tu⅐GTP). The Gol peptide also preferentially cross-linked to the GDPbound conformation of EF-Tu (data not shown). These data are consistent with the structural data on EF-Tu and show that the open conformation of EF-Tu⅐GDP, in which the Gly 59 -Ile 60 bond is more exposed, is the preferred conformation for Gol binding and Lit-mediated cleavage.

T4 Gol Peptide Binding to EF-Tu Targets the Cell for Litmediated Proteolysis-
The function of the Gol peptide from gp23 in the T4-directed phage exclusion response of E. coli could be to activate proteolysis either by contributing essential catalytic groups to Lit, activating the substrate EF-Tu, or both. Circumstantial evidence suggests that the peptide does not activate the protease directly. For example, if the peptide furnished essential catalytic groups for Lit, then it might be expected that some of the originally identified gol mutations in

FIG. 3. Peptide mapping of [ 3 H]Gol-EF-Tu cross-links.
Purified [ 3 H]Gol-EF-Tu cross-linked complex (ϳ4.5 nmol) was digested with 2% trypsin and fractionated on a C8 reverse-phase HPLC column (see "Experimental Procedures"). Three radioactive pools (A, B, and C), eluting at 16.5, 18.5, and 20 min, respectively, were collected. The figure shows the subsequent separations of each of these pools rechromatographed on a C18 reverse-phase column developed using an acetonitrile gradient, as described under "Experimental Procedures." Fractions containing 3 H-label are indicated by the shaded boxes, and the cpm values represent background subtracted counts per fraction. Single radioactive peptides were collected from these traces (S1, S2, S3, S4, and S5) and submitted for N-terminal sequencing. The traces were run at 18°C while the absorbance was monitored at 220nm . a Yield of each phenylthiohydantoin-derivative. b Total radioactivity determined for each fraction (background counts, ϳ25 cpm).
c The primary sequence superscripts refer to EF-Tu sequence; superscripts in italics refer to Gol sequence. the gp23 protein would not support cleavage of EF-Tu, as is the case for active site mutations of essential catalytic groups in enzymes. However, this is not the case, because overexpression of Lit renders bacteria susceptible to T4 gol mutants (17) and equivalent Gol mutant peptides in the in vitro assay can activate the cleavage of EF-Tu at elevated concentrations (12).
The present work shows for the first time that the Gol peptide binds to EF-Tu, to a site distinct to that cleaved by the enzyme, to form the specific substrate for Lit. There is still the formal possibility that the peptide furnishes amino acids for catalysis in the ternary complex, but this does not seem likely based on the effects of gol mutations discussed above. It is also possible that the peptide might bind to Lit in the absence of EF-Tu, although it has not been possible to detect stable binary complexes between the Gol peptide and Lit using column chromatography or chemical cross-linking experiments. The most plausible mechanism at the present time is that Lit only forms a stable complex with EF-Tu when the Gol peptide is bound, implying that the binary, Gol peptide-EF-Tu complex is an "activated" substrate for Lit. This type of substrate activation mechanism has similarities to the staphylokinase cofactor in the action of plasmin (29). The protein cofactor does not affect the geometry of the plasmin-active site but instead forms an additional docking site for the enhanced presentation of the plasminogen substrate to the enzyme. In this case, both the enzyme and the substrate are in contact with the activating  (27) and EF-Tu⅐GDP from T. aquaticus (28) prepared using Molscript (40). Both structures show the positions of the three ␤-strands from which peptide fractions S1, S3, and S4 were isolated (shown in red) and thereby define the Gol peptide binding site. The conversion of EF-Tu⅐GTP to EF-Tu⅐GDP, following the hydrolysis of bound GTP, causes the Lit cleavage site (in yellow) and the Gol binding site to become more exposed to solvent and ϳ15 Å closer to each other. An E. coli EF-Tu numbering system has been employed (percentage identity between T. aquaticus and E. coli EF-Tu is ϳ72%). However, it should be noted that Cys 256 is Val 266 and Glu 273 is Asp 284 in T. aquaticus EF-Tu. The three-domain structure of EF-Tu is also highlighted in the figure. The gray sphere represents the bound magnesium ion at the active site along with either GDP or the nonhydrolyzable analogue of GTP, GDPNP (both are shown in black ball-and-stick representation).
cofactor. In a similar manner, Gol peptide bound to EF-Tu could form additional contacts with Lit to enable proteolysis to occur at the Gly 59 -Ile 60 bond.
We have previously shown that Lit-mediated cleavage of EF-Tu is specific for the tertiary structure of the translation factor, which, in the context of the present work, is most likely explained by the specific contacts that are made between the Gol peptide across two domains of EF-Tu (Fig. 5). The chemical cross-linking data show that the Gol peptide contacts domains II (residues 271-278 and 254 -261) and III (residues 326 -333) of EF-Tu, and so its binding site likely extends ϳ30 Å across this domain-domain interface. (At the present time we cannot discount the possibility that domain I is also contacted by Gol, but no cross-links were identified in this domain from three independent experiments.) Both circular dichroism and NMR spectroscopy indicate that the isolated 29-residue Gol peptide does not have stable secondary structure in solution 3 but likely adopts a defined structure on binding to EF-Tu. Twenty-nine amino acids would be sufficient to span the domain II-domain III interface either as an extended strand or even as an ␣-helix.
High-resolution structural information is available for EF-Tu in both its nucleotide-bound conformations (EF-Tu⅐GDPNP and EF-Tu⅐GDP, respectively) (Fig. 5). From these crystal structures, it is apparent that, in the GDP-bound conformation, both the Lit cleavage site (Gly 59 -Ile 60 ) and the Gol binding sites are more solvent accessible. This could explain the preferential binding of the peptide and the faster rate of cleavage for this form of the translation factor. The interconversion between the GTP-and GDP-bound conformations of EF-Tu involves a large intramolecular movement of domain I (the GTPase domain), such that the distance between domains I and II increases by ϳ10 Å. This movement not only causes the cleft between the two domains to open but presents the Lit cleavage site at the base of this cleft close to the now exposed Gol binding site at the domain II/III interface (15-20 Å). Our data suggest that the Gol peptide binds preferentially to the open complex of EF-Tu bound with GDP across this interface placing the peptide in the vicinity of the cleavage site, thereby creating an additional docking surface for the Lit protease. Binding to the nucleotide product complex in preference to the closed substrate complex implies that the Gol peptide should be an inhibitor of the intrinsic GTPase of EF-Tu, and this is precisely what is observed (Fig. 4).
The Physiological Function of Gol Peptide Binding to EF-Tu-The results presented above make it clear that the Gol region of the major head protein gp23 of T4 binds to EF-Tu and this targets an infected cell for Lit-mediated cell death. The fact that this association must occur in bacteria whether or not they contain Lit implies that it has a physiological role. The interaction as measured in vitro with the 29-residue peptide is weak (K i ϳ0.3 mM) by comparison to other macromolecular associations such as antibody-antigen interactions (micromolar to nanomolar), although this affinity may well be higher for intact gp23. In this context it is interesting to note the relative concentrations of EF-Tu and gp23 in an infected E. coli cell. EF-Tu represents 5-10% of total cell protein (ϳ120,000 -240,000 copies, equivalent to a concentration of ϳ0.1-0.2 mM (30)) and that a similar number of gp23 molecules appear late in T4 infection, with 960 self-associating to form a single phage head particle (31,32). Because binding of gp23 to EF-Tu is localized to a small, 29-residue region of the major coat protein (as defined in vivo by the original gol mutants and in vitro by peptide binding), this association could occur as gp23 is being synthesized, suggesting a possible role in the assembly of the phage head itself. T4 bacteriophage head assembly is highly complex, involving up to 11 different proteins in the prohead and requiring both host and phage-encoded chaperones (GroEL and gp31, respectively) as well as scaffolding proteins (for a review see Black et al. (33)). With its high cellular concentration and its ability to bind a specific sequence in gp23, EF-Tu might be acting as an early chaperone or ancillary scaffolding protein in the assembly process. EF-1␣, the human equivalent of EF-Tu, has been reported to have a chaperone-like function by binding proteins and targeting them for degradation by the ubiquitin system, a role for which EF-Tu can substitute (34). The potential for a chaperonelike role for EF-Tu has been given further credence by recent in vitro work. EF-Tu can aid the renaturation of denatured proteins such as citrate synthase, ␣-glucosidase, and rhodanase (3,35); forms stable complexes with several unfolded proteins such as bovine pancreatic trypsin inhibitor and carboxymethyl ␣-lactalbumin; and protects some proteins from irreversible aggregation (3). We speculate that this chaperone activity may have been put to use by infecting bacteriophage in vivo to bind the gol region of gp23.
The fact that EF-Tu is a GTPase is also consistent with a putative chaperone role in phage head assembly, because the translation factor can switch between two distinct conformational states. Importantly, the folding properties of EF-Tu are enhanced in the open GDP-bound conformation (3), the conformation favored for Gol binding, and the binding of unfolded proteins inhibit its intrinsic GTPase activity (35), as does Gol peptide. Therefore, it seems possible that EF-Tu could have a chaperone-like role early in the synthesis/assembly of T4 simply by binding to the Gol sequence of gp23 in a nucleotide-dependent manner prior to it being shuttled to GroEL and gp31 (33).

Viral Protein Interactions with Host Translation Factors-
The association of EF-1␣ and the human immunodeficiency virus (HIV) type 1 gag polyprotein was reported recently by Cimarelli and Luban (36), but although this interaction seems to be essential for viral replication, its role remains unclear. The HIV-1 gag polyprotein is processed by an activated viral protease and so directs the formation and release of nascent virions. These proteolyzed products exhibit important functions during virion assembly, forming the capsid, the nucleocapsid, and the matrix of the virion while others, such as p6, play key roles in the early phases of the viral life cycle (for a review see Swanstrom and Wills (37)). Two of these domains (nucleocapsid and matrix) have been shown to interact with 3 R. Bingham and C. Kleanthous, unpublished results.
FIG. 6. EF-Tu⅐GDP is the preferred substrate for Lit-mediated proteolysis. The figure shows plots used to determine the first-orderrate constants for the cleavage of EF-Tu⅐GDP (q) and EF-Tu⅐GTP (E) by Lit in the presence of the Gol peptide (see "Experimental Procedures"). Fitting a number of data sets (n ϭ 4) for each gave mean observed rates of 0.044 and 0.024 min Ϫ1 , respectively (S.D., Ϯ0.001 and Ϯ0.002 min Ϫ1 , respectively). EF-1␣ in an RNA-dependent manner (36). In sharp contrast to the interaction between gp23 and bacterial EF-Tu, which occurs at the domain II/III interface, the interaction between matrix and EF-1␣ has been shown to take place with domain I at the N terminus of EF-1␣ (residues [1][2][3][4][5][6][7][8][9][10][11][12][13][14]. Furthermore, this interaction not only inhibits host protein translation but also directs the incorporation of a truncated form of EF-1␣ into the viral membrane. Translational inhibition has also been reported upon the interaction of the herpes simplex virus 1-infected cell protein 0 and EF-1␦ (38). In addition to these interactions, it has been shown that RNA polymerase from vesicular stomatitis virus specifically associates with EF-1␣ for its activity (39), which is analogous to the association of bacterial EF-Tu with the Q␤ RNA polymerase (2). Whatever their relationship to Gol binding, these other studies indicate that peptides other than Gol peptide can bind to translation factors such as EF-Tu. In view of our data on the binding of a bacteriophage head protein to EF-Tu and that of Cimarelli and Luban (36) on the complex formed between HIV-1 capsid proteins with EF-1␣, it is tempting to speculate that viral capsid protein association with host translation factors is the norm rather than the exception and that this association plays a pivotal role in viral maturation in prokaryotes and eukaryotes.