Selective in Vivo Rescue by GroEL/ES of Thermolabile Folding Intermediates to Phage P22 Structural Proteins*

The in vivo conformational substrates of the GroE chaperonins have been difficult to identify, in part be- cause of limited information on in uiuo polypeptide chain folding pathways. Temperature-sensitive folding (tsf) mutants have been characterized for the coat pro- tein and tailspike protein of phage P22. These mutations block intracellular folding at restrictive temperature by increasing the lability of folding intermediates without impairing the stability or function of the native state. Overexpression of GroEL/ES suppressed the defects of tsf mutants at 17 sites in the coat protein, by improving folding efficiency rather than assembly efficiency or protein stability. Immunoprecipitation experiments demonstrated that GroEL interacted transiently with newly synthesized wild-type coat protein and that this interaction was prolonged by the tsf mutations. Folding defects of the tailspike polypeptide chains were not sup- pressed. A fraction of the tsf mutant tailspike chains bound to GroEL but were inefficiently discharged. The results suggest that 1) thermolabile folding intermedi- ates are natural substrates of GroELIES; 2) although GroEL may bind such intermediates for many proteins, the chaperoning function is limited to a subset of substrate proteins; and 3) a key reason for the heat-shock response may be to stabilize thermolabile folding intermediates at elevated temperatures.

heptameric rings, while HsplO is a single ring of 7 subunits (Hendrix, 1979;Zwickl et al., 1990;Hartman et al., 1992). GroELlES are essential for E. coli viability at all temperatures (Fayet et al., 1989). Hsp60/Hsp10 have been shown to mediate the folding of a number of proteins, by preventing aggregation arising from incompletely folded molecules (Viitanen et al., 1991;Gething and Sambrook, 1992;Hendrick and Hartl, 1993). In the case of oligomeric proteins, it is not known if GroEL/ES also mediate protein assembly.
Hsp6O can bind to a diverse set of partially folded protein substrates but not to their native states (Martin et al., 1991; al., 1992;. Folding probably occurs in the central cavity of Hsp60, so that discharged substrate molecules are released in a conformation different (often committed to correct folding) from that of incoming substrate molecules (Martin et al., 1991;Langer et al., 1992b;Braig et al., 1993). Folding generally occurs in an ATP-dependent manner and requires cycles of binding and release of HsplO (Martin et al., 1993a). Molecular chaperone proteins can function cooperatively, with Hsp7O proteins recognizing extended polypeptides which are then transferred to Hsp6O (Cheng et al., 1989;Kang et al., 1990;Manning-Krieg et al., 1991;Langer et al., 1992a).
Hsp6O has been shown to reduce the extent of denaturation of native proteins (Martin et al., 1992). However, Honvich et al. (1993) have reported that a temperature-sensitive lethal mutation in GroEL did not cause destabilization of prefolded proteins. It has been difficult to establish whether or not folding intermediates, which may be particularly sensitive to elevated temperatures, are the major substrates of Hsp6O in vivo. Van Dyk et al. (1989) have reported that overexpression of GroEL/ES can rescue a subset of temperature-sensitive mutations in Salmonella and P22. Here we have taken advantage of the availability of well defined sets of temperature-sensitive folding mutants of two phage P22 structural proteins to probe chaperonin function in vivo. The tsf' mutations act by further destabilizing an intracellular folding intermediate, without altering the stability of the native state (Goldenberg et al., 1983;Sturtevant et al., 1989).
The P22 tailspike protein, encoded by gene 9 , is a trimer, each chain of which is 666 amino acids and has a molecular mass of 72 kDa (Sauer et al., 1982). The tailspike is a structural protein of the phage head which is involved in host recognition. Its structure is dominated by a P-helix motif (Steinbacher et al., 1994). Temperature-sensitive folding mutants of the tailspike protein destabilize a single chain folding intermediate without destabilizing the native folded trimer (Goldenberg and King, 1981;Goldenberg et al., 1983;Sturtevant et al., 1989). At elevated temperatures, single chain intermediates transiently accumulate and then aggregate into inclusion bodies (Haase-Pettingell and .
The P22 coat protein, encoded by gene 5, is composed of 430 amino acids and has a molecular mass of 47 kDa . 420 icosahedrally arranged coat protein molecules are the major protein constitutents of phage heads, within which is located P22 DNA (King and Casjens, 1974). A set of 18 temperature-sensitive mutants, at 17 different sites in the coat protein, have been characterized. Phage particles and precursor phage particles (procapsids) carrying these mutations were not destabilized at restrictive temperature in vivo (Gordon and King, 1993). In vitro, the stabilities of folded but unpolymerized mutant chains, and of mutant procapsids, were similar to their wild-type counterparts.' In vivo, mutant coat protein synthesized at restrictive temperature was not degraded and instead accumulated as inclusion bodies (Gordon and King, 1993). The mutants chains are recessive to wild-type, do not exhibit intragenic complementation, and cannot be rescued by wild-type subunits. These genetic properties are also consistent with their being defective in the folding process (Gordon and King, 1994). These results indicate that these temperature-sensitive mutants of the coat protein, like the tailspike mutants, are defective in folding.
The heat-shock response, in which the synthesis of specific proteins is increased in response to heat or other stresses, has been documented in many organisms, but its biological reason and mechanism of induction are less clear (Lindquist and Craig, 1988;Ang et al., 1991;Craig and Gross, 1991). The tsf mutants allow one to generate intracellular partially disordered folding intermediates a t temperatures optimal for cell growth. The results reported here suggest that a class of physiological substrates of heat-shock proteins are thermolabile folding intermediates of cytoplasmic proteins.
Bacteriophage-All phage strains used in these experiments carried the cl-7 allele, which prevents lysogeny. The strains carrying ts mutations in gene 5 (coat) (Jarvik and Botstein, 1975;Casjens et al., 1991;King, 1993,1994) and gene 9 (tailspike) (Smith et al., 1980) were isolated as described. Other mutations carried by some phage used in these experiments are gene 13 amHlO1, which delays lysis, and gene 3 amN6, which prevents DNA packaging (King et al., 1973). Multiple mutant strains were constructed by crossing, and phage stocks were prepared (Gordon and King, 1993). Freshly prepared stocks grown in LB/Amp were used for the experiments of Table 111.
Growth Curues-Exponentially growing DB7136/pBR322 or DB71361 pOF39 cells in LB/Amp, a t a concentration of 4 x 108/ml, were infected with phage a t a multiplicity of infection of 10, at 39 or 37 "C. Infections were diluted 1000-fold after 12 min and half was shifted down to 30 or M. L. Galisteo, C. L. Gordon, and J. King, unpublished results.
28 "C. At the indicated times, aliquots were removed and lysed by diluting into dilution fluid saturated with CHCl,, followed by titering for viable phage.
Radiolabeling Experiment-Overnight cultures of DB7136/pBR322 and DB7136/pOF39 growing in MinimaUAmp were diluted 1:50 in fresh media and grown to a concentration of 1 x 108/ml at 32 "C. Cells were pelleted, resuspended in fresh media to a concentration of 4 x 108/ml and placed on ice. Cells were infected with phage at a multiplicity of 7 at 29 "C (time 0). At 10 min the infections were split to 29, 33, 35, and 39 "C. Infected cultures were incubated with I4C-amino-acids at a final concentration of 2 pci/ml. The 39 "C cultures were labeled at 45 min, the 35 "C cultures were labeled a t 53 min, the 33 "C cultures were labeled at 61 min, and the 29 "C cultures were labeled a t 71 min. 15 min after labeling infections were chased with casamino acids a t a final concentration of 2%. Infections were terminated by being placed on ice 35 min after labeling. Cells were pelleted and resuspended in 1/10 volume of M9/Mg2'. To lyse the infected cells, samples were frozen at -20 "C, thawed, frozen again in a dry-ice/ethanol bath, and thawed. 25 pl of each lysate sample was centrifuged for 3 min in a microcentrifuge. The supernatants were removed and the pellets were washed in 25 pl of M9/Mg2' and recentrifuged. These supernatants were removed and combined with the corresponding supernatants from the first centrifugation. The pellets were resuspended in 50 pl of M9/". Samples from the supernatant, pellet, and original lysates were mixed 1:2 with 3 x SDS sample buffer (0.1875 M Tris, 6% SDS, 15% p-mercaptoethanol, and 30% glycerol) and electrophoresed through a n SDS gel, which was subject to autoradiography.
Immunoprecipitations-pOF39 cells in MinimdAmp were grown overnight to stationary phase at 30 "C. Cells were diluted 150 in fresh media, grown to -1 x 108/ml, and concentrated to 4 x 108/ml. 1.5 ml cells were incubated with 1.5 ml phage, a t a multiplicity of infection of 7 phagehell, at 37 "C with aeration. 50 min after infection, cells were labeled with 20 pci/ml I4C-amino-acids (to 3 pci/ml). Infections were chased a t 50.5 rnin after infection with 10% casamino acids (to 3%). At the indicated times, 700 pl of infected cells were added to 78 pl of buffer A (500 mM Tris-HC1, 1 M NaCI, 100 mM CDTA, pH 7.6) and immediately frozen in dry ice/ethanol. Cells were thawed a t 4 "C, where all subsequent manipulations were performed.
Thawed infection samples were centrifuged in a microfuge for 5 min.
tants were mixed 3:2 with the beadslantibodies, followed by rocking for Supernatants were carefully removed and respun. Resultant superna-1 h. Samples were centrifuged for 30 s and washed in one-tenth buffer A. Pelleted beads were boiled in SDS sample buffer, electrophoresed through SDS gels, and exposed to film. PhosphorZmager Quantifications-Coat protein, tailspike protein, GroEL, DnaK, and gpl were quantified by exposing gels to Phos-phorImager screens and using the ImageQuant software (Molecular Dynamics). Bands were circled using the "region" tool at x4 magnification. Background was defined as counts/pixel in each lane immediately above DnaK, where no protein bands were visible.

RESULTS
Effect of GroELIES Overexpression on Plating Eficiencies of tsf M u t a n t P22 Phage-The E . coli GroE operon, consisting of the natural GroE promoter followed by the G r o E S and GroEL genes, was cloned into a pBR322-derivative plasmid (Fayet et al., 1986). This plasmid, called pOF39, was transferred into s. typhimurium, the host for phage P22. The ability of pOF39 to rescue temperature-sensitive folding alleles of the P22 coat protein and tailspike protein genes was examined.
Phage were plated with Salmonella cells carrying either the pOF39 plasmid or a control plasmid. For each phage strain, the ratio of plaques seen at several test temperatures to plaques seen at 24 "C with control cells was determined. Wild-type phage exhibited almost no decrease in plating efficiency, as Phage carrying the tsf coat protein and tailspike alleles exhibit sharp reductions in plating efficiencies at elevated temperatures on wild-type cells (Smith et al., 1980;King, 1993, 1994). At their minimum restrictive temperatures of growth, most of these strains have reductions in titer of about lo6, which is in the range of that expected for reversion events.
The minimum restrictive growth temperature was significantly increased for each of the 18 tsf coat protein strains when plated on the pOF39 cells (Fig. 1B). For example, the extremely temperature-sensitive strains 5lk(Pro31OAla) and 52WThr294ZZe) were rescued at 33 "C, while the least temperature-sensitive strains 5Ts(!Pyr411His), 5Ts(Dp48Gln), and 5Ts(Gly403Asp) were rescued at 41 "C. Plaque counts with pOF39 cells were increased to 24 "C levels and were about lo6 higher than with control cells. At temperatures below the minimum restrictive growth temperature, plaques were larger on the pOF39 cells than on control cells.
The relative plating efficiencies of 15 tsf tailspike protein strains were also determined ( Fig. 1C). In marked contrast to the results for the coat protein strains, overexpression of the GroE operon had no effect on the plating efficiency of any of the temperature-sensitive tailspike strains examined.

Effect of GroEL / E S Overexpression in Liquid Culture-
The kinetics and extent of phage propagation was quantitatively examined by performing one-step growth curve experiments with and without GroEUES overexpression.
Growth curves were performed at 28 and 37 "C with wildtype phage and three tsf coat protein strains ( Fig. 2 A ) . These temperature-sensitive strains are restricted for growth at -37 "C, as assayed by plating. Wild-type phage had a burst of -270 phagekell at 28 "C, on both pOF39 cells and control cells, and a burst of -350 phagekell at 37 "C with both cell types. At 28 "C the three tsf coat strains had bursts which were about 75% of the wild-type burst. At 37 "C all three ts coat strains had bursts of less than one phagekell on control cells. Bursts at 37 "C for the coat ts mutants incubated with the pOF39 cells were -500 times higher than with the control cells. The pOF39 37 "C burst was about 20% of the corresponding wild-type burst for 5l'b(Ser262Phe), 35% for 5Zb(Aspl74Asn), and 60% for

5!&(Ser223Phe).
At 37 "C the kinetics of phage growth on the pOF39 cells were slightly delayed in the mutants as compared with wild-type phage.
Growth curves performed at 30 and 39 "C with wild-type phage, the coat protein mutant 5WPro418Ser), and the two tailspike mutants 9TWThr307Ala) and 9WGly244Arg) are presented in Fig. 2 B . On plates, 5!&(Pro418Ser) was restricted for growth at -37 "C, and the two tailspike mutants were partially restricted at 39 "C and completely restricted at 41 "C. Wildtype phage yielded -350 phagelcell at 30 "C and -400 phage/ cell at 39 "C. At 39 "C 9Tk(Thr307Ala) yielded -one phagelcell on both cell types and 9WGly244Arg) yielded -three phage/ cell. At 39 "C, the coat tsf mutant 5lWPro418Ser) yielded less than one phagekell on control cells, while it had achieved about 25 phagdcell on the pOF39-carrying cells 120 min after infection. This coat ts rescue was less than that demonstrated in Fig. 2A because of the higher temperature. Even though 39 "C is -2 degrees above the restrictive growth temperature for 5!&(Pro418Ser), the ts coat protein mutant was significantly rescued by the GroE overexpression while the tailspike mutants exhibited no improvement in propagation eflkiency.

Requirement for GroES Overexpression-Is overexpression of
both GroE proteins required for rescue of the coat protein mutants? For these experiments Salmonella cells carrying pTG10derivative plasmids were used. pLS carried the GroE operon cloned directly from pOF39 into pTG10. A frameshift mutation in the GroES gene was introduced to generate pL. It has been reported that this mutation was not polar on expression of the downstream GroEL gene (Goloubinoff et al., 1989). pS carries the natural GroE promoter, followed by GroES and only the first 30% of the GroEL coding sequence (Van Dyk et Fayet et al., 1986). Control cells carried only the pTGlO plasmid without the GroE operon. The efficiency of plating of the tsf coat protein and tsf tailspike protein strains was examined on these four cell types at permissive temperature and at each strain's particular minimum restrictive temperature.
Wild-type phage exhibited no reduction in plating ability at 30 "C and 41 "C on all four cell types ( Table I ) . All 18 tsf coat protein strains formed plaques with high efficiency at 30 "C on all four cell types. At their specific restrictive temperatures, the efficiency of plating on pTGl0 cells, pS cells, and pL cells was sharply reduced. In contrast, on the pLS cells, plating eficiency increased to near permissive levels. These results demonstrated that overexpression of both GroEL and GroES was required for rescue of the ts coat protein strains. The results also illustrated that the rescue could be attained with two different plasmid systems, as long as the GroE operon was present. Though still substantial, the rescue achieved with pLS was reproducibly slightly less than that achieved with pOF39. This is consistent with the expected lower copy number of the pLS plasmid (Sambrook et al., 1989). On all four cell types, the ts tailspike strains plated with high efficiency at permissive temperature and low efficiency at restrictive temperature (Table 11).
Effect of GroELIES Overexpression on Protein Folding-Newly synthesized tsf coat protein and tailspike protein accumulate in inclusion bodies at restrictive temperature (Haase-Pettingell and Gordon and King, 1993). To follow the fate of newly synthesized protein in the presence or absence of GroE overexpression, pOF39 cells and pBR322 control cells grown in minimal media were infected at temperatures ranging from 29 to 39 "C, with wild-type phage, with 5n(Phe353Leu), or with gD(Gly244Arg). In this experiment phage strains also carried a gene 3 amber mutation, which blocks productive phage assembly at the procapsid stage . Infected cells were pulsed with [14C]amino acids and incubated for 35 min. Cultures were subjected to freezing followed by thawing at room temperature, twice, which gives efficient cell lysis (Haase-Pettingell and Gordon and King, 1993).
A pellethpernatant separation was then performed. Inclusion bodies are pelleted in this procedure, while procapsid particles and native tailspike trimers remain in the supernatant (Haase-Pettingell and Gordon and King, 1993). Wild-type coat protein remained largely in the supernatant, presumably assembled into procapsid particles, at all temperatures tested on both cell types (Fig. 3A). In contrast, the 5!MPhe353Leu) coat polypeptide chains were largely pelleted at the higher temperatures when control cells were infected. When the infection of 5Zb(Phe353Leu) phage was performed on pOF39 cells, the percentage of pelleted coat protein was sharply reduced at all temperatures. These results establish that GroE overexpression is improving coat protein folding efficiency by significantly reducing the levels of coat protein inclusion body formation.
Tailspike folding efficiency was assayed by determining the distribution of tailspike between native trimers and incompletely folded species. Native trimers are resistant to solubilization by SDS, while inclusion body aggregates and partially folded species are solubilized by SDS (Haase-Pettingell and . As shown in Fig. 3B, wild-type tailspike folded and assembled into trimers with high efficiency at temperatures up to 35 "C on both cell types examined. At 39 "C the efficiency of trimerization was reduced, and this was not improved by overexpression of GroE. The tsf tailspike mutant 9WGZy244Arg) exhibited reduced folding efficiency at temperatures of 33 "C and higher, which was not improved by GroE overexpression.
Induction of the Heat-shock Response by tsf Mutants-The levels of GroEL and DnaK, which are major prokaryotic heatshock proteins, were quantified from SDS gel lanes corresponding to the 39 "C lysates described above (Fig. 4). These samples had been labeled with 14C-amino-acids 45 min after infection. On the pOF39 cells, the levels of GroEL synthesis in the infections with the coat and tailspike tsf mutants were -2-fold higher than in the wild-type infection (Fig. 48). GroEL synthesis was also induced in the control cells. In addition, DnaK synthesis was induced on both cell types by the ts mutants (Fig.  4C). Apparently, the incorrectly folded coat and tailspike protein were able to serve as signals for augmenting the heat shock response. Fig. 4 also demonstrates that the pOF39 plasmid, which rescued the coat tsf mutants, generated a -9-fold increase in the rate of GroEL synthesis at 39 "C, when infected with 5!iWPhe353Leu) or 9fi(Gly244Arg), as compared with control cells.

Kinetics of in Vivo
Interaction with GroEL-Immunoprecipitation of phage-infected cell extracts with anti-GroEL antibodies was performed to study the in vivo interaction between GroEL and the coat and tailspike proteins. pOF39 cells were grown in minimal media and infected at 37 "C with wildtype phage, 5!&(Ser223Phe), or 91s(Thr235Zle). 50 min after infection, cultures were pulse-labeled with 14C-amino-acids and chased 30 s later. Samples were withdrawn and immunoprecipitated with anti-GroEL antibodies or control serum at various times after labeling. Anti-GroEL precipitates and control samples were electrophoresed through SDS gels and exposed to  Permissive temperature used was 30 "C for all strains except 9Ts(amR232), for which a permissive temperature of 24 "C was used.  Fig. 5

, D-I?
With wild-type phage (Fig. 5 0 ) , 0.75 min after labeling, 25% of coat protein was bound to GroEL, and 2 min after labeling, 5% was bound. For about 6 min after labeling, -7% of tailspike protein was bound. This suggested that the majority of tailspike protein did not bind GroEL, and once formed, the complex decayed slowly.
With 5WSer223Phe) phage (Fig. 5E), 0.75 min after labeling 43% of ts coat protein was bound to GroEL, and 2 min after labeling 21% of the coat protein was bound. 5Ts(Ser223Phe) coat protein transiently interacted with GroEL and was discharged at a substantially slower rate than wild-type coat protein. The amount of bound tailspike protein in the ts coat protein infection was reproducibly reduced as compared with tailspike bound in the wild-type phage infection, even though the levels of GroEL were -2-fold higher, as if ts mutant coat protein were saturating available GroEL.
With 9iWThr235ZZe) phage (Fig. 5C), the (wild-type) coat protein had similar binding behavior to GroEL as seen with wild-type phage. At early times after labeling, 30% of ts mutant tailspike protein was bound to GroEL, and this complex decayed slowly, with an apparent half-life of -35 min. Extrapolating toward early times based on the slow rate of decay suggests that most ts mutant tailspike protein did not interact with GroEL.
By mass, -15 times more coat protein was synthesized than tailspike protein. This binding of ts tailspike protein to GroEL may be counterbalanced by the -2-fold increase in the rate of GroEL synthesis, so that similar amounts of GroEL remained free for binding to coat protein.
At the 2-min time points, there was -50% more labeled coat protein in the lysates than at the 0.75-min time points, pre-A the two strains 5lWfip48Gln) and 5l3 (Gly403Asp)  two strains grown on cells overexpressing GroELlES were prepared. As shown in Table 111, these phage exhibited the same reduction in titer when incubated in medium lacking M e as phage grown on control cells. As shown above, GroELlES overexpression could improve folding efficiency, but in this case Interestingly, these two mutants are among the least restricted for growth in vivo, and yet the other coat protein ts phage (such as 523(Thr294Ile), which have lower restrictive bilized under these in vitro conditions. Thus, in agreement with fore of the folded coat protein molecule, is not the key factor in determining the in vivo temperature sensitivity. "he mutations appear to differentially destabilize folding intermediates and the folded protein (Gordon and King, 1994).

DISCUSSION
Folding intermediates are generally less structured than their own native state and, therefore, particularly susceptible to destabilization (Haase-Pettingell and Mitraki and King, 1989;Viitanen et al., 1990;Goldberg et al., 1991;van der Vies et al., 1992). Cells infected a t restrictive temperature accumulate thermolabile folding intermediates (Goldenberg et al., 1983;Gordon and King, 1993;Danner and Seckler, 1993). A key reason for the heat-shock response may be to stabilize folding intermediates at elevated temperatures. stabilization of the native state (Martin et al., 1992). SDS gel, which was exposed to film. Equal amounts (10 of each major species being synthesized in the infected fiesum- gpl as a normalization standard has been performed previously (King sumably due to completion of labeled nascent chains after addition of cold amino acids. This suggests that the initial rate of coat protein-GroEL complex discharge is faster than that plotted and that most newly synthesized coat protein chains may have undergone a brief interaction with GroEL. Can GroELIES Overexpression during Particle Formation Improve Stability?-Phage particles carrying the tsf coat protein mutations are not destabilized a t restrictive temperature in vivo (Gordon and King, 1993). However, phage particles from intermediate which is recognized by GroEL or lead to an altered conformation, due to destabilization of a wild-type folding intermediate, which is recognized by GroEL (Haase-Pettingell and . A model for the action of temperature-sensitive folding mutations posits an equilibrium between the productive folding intermediate and an off-pathway species (1-1' ). Elevated temperatures result in increased levels of I*, which is the species that has the property of aggregating (Haase-Pettin-gel1 and Mitraki and King, 1989). "sf substitutions further shift the equilibrium to I*. If chaperonins recognize I*, they could block aggregation, in this case without participating in the productive pathway. The mutants may disrupt a critical conformation yielding a common off-pathway intermediate (I*). Exponentially growing pOF39 cells were infected a t 37 "C. Infections were labeled with "C-amino-acids and chased 30 s later. Samples were removed at the indicated times after labeling. These were mixed with buffer A, frozen in dry icdethanol, and thawed at 4 "C, where subsequent manipulations were performed. 14 pl of each sample were removed (lysate; lane 1 ). The remainders were subject to two prespins. Pellets, representing un-lysed cells and aggregated material, were resuspended in SDS sample buffer to the volume of the supernatants, and 14 pl was loaded on the gels (prespin pellet; lane 2 ) . 14 pl of supernatant fractions were removed (prespin supernatant; lane 3). 300 pl of each buffer, boiled, electrophoresed through 10% SDS gels, and exposed to films. Quantification of the amount of coat protein and tailspike protein bound Recognition of such a common intermediate would rescue the "sf mutants of the tailspike protein were not rescued by diverse starting mutants, which were present a t 17 different GroE overexpression. These mutations generate thermolabile sites (Gordon and King, 1993).
intermediates in the tailspike folding pathway (Smith et al., 1980; Haase-Pettingell and . Apparently, GroELrES cannot rescue folding defects in all proteins. This reveals an aspect of the selectivity of chaperonin function. Our results suggest that the ability of GroEL to mediate folding is proteinspecific rather than allele-specific. The tailspike is a large trimeric protein whose secondary structure is dominated by the recently discovered parallel p-coil motif (Steinbacher et al., 1994). The folding and assembly of the coat protein have been studied in vitro (Prevelige et al., 1988;Teschke and King, 1993;Teschke et al., 19931, although its structure is not yet known. Different chaperones may be specialized for classes of protein conformations. Alternatively, the 72-kDa tailspike protein may be too large to permit productive folding within the GroEL ring (Langer et al., 1992b). Similarly, folding steps occurring after the assembly of protein oligomers, which may be too large for GroEL, may not be well chaperoned.
Interaction ofFolding Intermediates with GroEL-The recovered complexes of protein bound to GroEL were stable for several hours without the use of cross-linking agents. The CDTAin the lysis buffer was expected to chelate Mg2' ions necessary for ATP hydrolysis (Martin et al., 1991). In addition, the effective concentrations of GroES and ATP, which are generally involved in complex discharge, were reduced upon cell lysis. At the earliest time point, a significant fraction of the labeled wild-type coat chains were immunoprecipitated with anti-GroEL. However, from the data reported here, it is not clear if this interaction was essential for productive folding. It appears that nearly all 5!l'sfSer223Phe) coat protein transiently interacted with GroEL. When bound to GroEL, 5ZkfSer223Phe) coat protein required more time to achieve the conformation permitting final release and presumably committed to assembly. It is possible that 511s(Ser223Phej coat protein cycled between a bound state and a prematurely released state. This reduced rate of GroEL turnover may explain the requirement for the higher concentrations of GroEL/ES necessary to mediate the folding of the ts mutant coat protein. Less than 10% of wild-type tailspike protein was recovered bound to GroEL, and this bound protein had a slow rate of discharge. Apparently, most wild-type tailspike protein does not pass through a conformation which is recognized by GroEL while folding in vivo; and the tailspike protein that does bind is released extremely slowly. 30% of 91b(Thr23511ej tailspike was recovered bound to GroEL at early times, and this complex was discharged with a half-time of -35 min, apparently without reaching the native state. The tailspike protein could also have undergone periods of release followed by rebinding. It is possible that upon temperature reduction in vivo, tsf mutant tail-rnctions in Vivo 27949 spike protein which had bound to GroEL at high temperature could be released in a conformation permitting correct folding (Holl-Neugebauer et al., 1991). In this experiment, -70% of ts mutant tailspike protein aggregated without being bound by GroEL. Van Dyk et al. (1989) reported that several temperaturesensitive proteins in s. typhimurium, E. coli, and the phage P22 tailspike protein could be suppressed by overexpressing GroELrES. Such suppression was reported to occur for four of nine tsf tailspike mutants e~amined.~ In their experiments on 9!L%(Gly244Arg), GroEL/ES overexpression increased plating efficiency at 39 "C from less than to 0.1 and bursts from 0.2 to 3% of the permissive value (Van Dyk et al., 1989).
Comparison with the much greater increases in yields seen for the coat tsf mutants suggests that these responses do not represent significant rescue. The results reported in our work confirm the low level of rescue of the tsf tailspike mutants, as seen by several assays. We conclude that overexpression of GroE cannot suppress any of the ts tailspike mutants. Brunschier et al. (1993) have examined the interaction between purified wild-type tailspike protein and GroELrES in vitro. They found that GroEMS could not significantly improve the folding efficiency of the wild-type tailspike protein, after denaturation by acid urea. GroEL formed a complex with refolding wild-type tailspike protein at high temperature (above 30 "C). This complex could be discharged with addition of ATP, but at high temperature the tailspike protein was discharged in a conformation still susceptible to aggregation. The amount of bound tailspike was reduced on subjection of a preformed complex to temperatures below 25 "C, suggesting that the tailspike was not tightly bound.
Uncoupling of GroEL Binding from Chaperoning of Folding-The results reported here, together with those of Brunschier et al. (1993), demonstrate that GroEL can bind tailspike folding intermediates without efficiently chaperoning folding, and thus that substrate binding and chaperoning of folding are separate aspects of GroEL function. These results indicate that GroEL/ES do not mediate folding only through binding to and sequestering substrates. These results are consistent with previous evidence that, in productive interactions, substrates fold to a non-aggregating conformation while complexed with GroEL (Martin et al., 1991).
Our results support the idea that HsplO is essential in those cases where large increases in folding yields are obtained. Apparently, the level of endogenous Salmonella GroES is insufficient for catalyzing productive release of tsf mutant coat protein bound to overexpressed GroEL. Chaperones and Phage Assembly-The GroE genes were originally discovered as a result of their requirement for bacteriophage morphogenesis. The "E" refers to gene E of phage A, encoding the major coat protein, in which suppressors of GroE mutations were recovered (Georgopoulos et al., 1973). On GroE mutant strains, bacteriophages T4 and A form aberrant head structures (Takano and Kakefuda, 1972;Georgopoulos et al., 1972Georgopoulos et al., , 1973Sternberg, 1973). Viral second-site suppressors of these GroE mutants have been recovered in the T4 coat protein (Revel et al., 19801, T4 gene product 31, which appears to function as a coat-specific chaperone (Coppo et al., 1973;Revel et al., 1980), the A coat protein, and the A portal protein (Georgopou-10s et al., 1973). Van der Vies et al. (1994) have recently shown that the T4 gene 31 product is a GroES homolog.
These observations and the data reported in this paper suggest that GroEL may link coat protein folding to shell assembly. The P22 coat subunits assemble into a T = 7 icosahedral shell requiring seven different conformations (Prasad et al., 1993). Shell assembly probably proceeds from incompletely folded molecules, whose final conformations are dependent on intersubunit interactions (Harrison et al., 1978;Liddington et al., 1991). GroEL may stabilize an assembly-competent conformation prior to shell assembly. In this case the absence of GroE function could cause defects in folding and assembly.
Induction of the Heat-shock Response-Parse11 and Sauer (1989) have shown that unstable variants of A repressor can induce the heat-shock response in E. coli. Infection with the P22 folding mutants appeared to induce a similar response in 5 ' . typhimurium. Our results are consistent with the accumulation of thermally perturbed folding intermediates (I*) being a regulator of the heat-shock response.
As temperature increases, perturbed folding intermediates will accumulate before perturbed forms of the native state (Haase-Pettingell and Viitanen et al., 1990;van der Vies et al., 1992). Accumulation of these partially folded intermediates may lead to induction of the heat-shock response, perhaps mediated by binding of folding intermediates to heat-shock proteins (Craig and Gross, 1991).