Modular folding and evidence for phosphorylation-induced stabilization of an hsp90-dependent kinase.

The de novo folding of the individual domains of the src family kinase p56(lck) was examined within the context of full-length p56(lck) molecules produced in rabbit reticulocyte lysate containing active chaperone machinery. The catalytic domain required geldanamycin-inhibitable heat shock protein 90 (hsp90) function to achieve its active protease-resistant conformation, but the src homology 2 (SH2) domain acquired phosphopeptide-binding competence independently of hsp90 function. The SH2 domain of hsp90-bound p56(lck) was folded and functional. In addition to the facilitation by hsp90 of kinase biogenesis, a conditional role in maintenance folding could be demonstrated; although wild type p56(lck) molecules with a negative-regulatory C-terminal tyrosine matured to a nearly hsp90-independent state, p56(lck) molecules with a mutated C-terminal tyrosine continued to require hsp90-mediated maintenance. De novo folding could be distinguished from maintenance folding on the basis of proteolytic fingerprints and the effects of different temperatures on folding behavior. Results indicate that during p56(lck) biogenesis, the SH2 domain rapidly folds independently of hsp90 function, followed by the slower hsp90-dependent folding of the catalytic domain and suggest the final stabilization of p56(lck) structure by phosphorylation-mediated interdomain interactions.

Protein domains are discrete units of compact globular structure. They frequently retain structure and function when separated from the whole protein by proteolysis or by protein engineering and are thus often defined as independently folding units. This independent folding of isolated domains may also reflect folding behavior occurring within the context of the whole protein. In vitro studies of protein renaturation support a model in which renaturation of many large, multidomain proteins proceeds through two stages: an initial rapid independent folding of individual protein domains, followed by a slower "domain pairing" stage in which folded domains coalesce to form the active multidomain protein (reviewed in Ref. 1).
However, protein folding events occurring within the concentrated environment of the cytosol are facilitated by molecular chaperones, a diverse group of proteins that prevent or reverse deleterious aggregation of protein folding intermediates (reviewed in Ref. 2). Protein folding in vivo is also facilitated by proteins that catalyze potentially rate-limiting steps of in vitro protein folding, the isomerization of peptidyl-prolyl bonds or of disulfide bonds (reviewed in Ref. 3). By these mechanisms, chaperones are thought to both facilitate the de novo folding of newly synthesized proteins and to allow renaturation of partially denatured proteins (reviewed in Refs. 4 and 5). Although our understanding of protein folding in vitro and of chaperone function in vivo have both advanced dramatically in recent years, the influences of molecular chaperones on individual protein folding pathways and processes are not well understood.
The multidomain protein tyrosine kinase p56 lck provides a unique opportunity to examine the chaperone-mediated folding of individual protein domains within the context of the whole protein. As a typical member of the src family (reviewed in Ref. 6), this nonreceptor protein tyrosine kinase is organized into several discrete domains. Proceeding from the N terminus, these domains include 1) a unique domain that appears to mediate interactions with membranes and integral membrane partner proteins (e.g. CD4; see Ref. 7), 2) an SH3 1 domain mediating intermolecular or intramolecular (8,9) interactions with proline-rich type II helices, 3) an SH2 domain mediating intermolecular or intramolecular interactions with specific phosphotyrosine motifs, 4) a highly conserved catalytic domain responsible for phosphotransfer reactions, and 5) a C-terminal tail thought to regulate the kinase via its intramolecular interaction with the SH2 domain.
For p56 lck and the related kinase viral p60 src chaperone machinery containing the 90-kDa heat shock protein (hsp90) appears to mediate protein biogenesis (10 -14). In addition to de novo folding, hsp90 also appears to supply conditional maintenance of kinase function (10,15). Because hsp90 function can be specifically inhibited by the benzoquinone anasamycin geldanamycin (16), hsp90-dependent protein folding pathways can be dissected in unfractionated rabbit reticulocyte lysates (RRL), containing active chaperone machinery (10,15,(17)(18)(19). In the work described here, we utilized this approach to test the hypothesis that chaperone-mediated folding of the full-length p56 lck molecule occurs in a modular fashion following its de novo synthesis. Results indicate a novel chaperone-mediated protein folding pathway in which the SH2 domain rapidly folds independently of hsp90, followed by the slower, hsp90-dependent folding of the catalytic domain. Additionally, we characterized the effects of interactions between the SH2 domain and the kinase regulatory tail and suggest that such interactions may serve as a linchpin, holding the folded kinase in a stable, hsp90-independent conformation.

EXPERIMENTAL PROCEDURES
Synthesis and Immobilization of Peptides-A high affinity SH2-binding motif derived from the hamster polyoma virus middle T antigen (20) was used as a core sequence for synthesis of four peptides (referred to subsequently as mTa peptides) by procedures described previously (21,22). The four peptides shared the common sequence GGGGEPQYEEIPI but differed with regard to N-terminal biotinylation and phosphorylation of tyrosine. The peptides were characterized by high performance liquid chromatography and mass spectrometry and were found to be greater than 90% pure. Biotinylated peptides were immobilized on avidin resins (Promega SoftLink Avidin). Nonbiotinylated peptides were used as soluble competitors or were covalently attached to agarose resins (p-nitrophenyl chloroformate-activated agarose) using aqueous coupling protocols recommended by the supplier (Sigma).
Solid Phase Peptide Binding Assays-The functional status of p56 lck SH2 domains was determined by assaying the ability of p56 lck folding intermediates (23) to specifically bind to immobilized mTa peptides (24). For these assays, gradient fractions containing either hsp90-bound or free monomeric [ 35 S]p56 lck were generated as described previously (23), adjusted to contain 1 mM sodium vanadate, and preincubated for 1 h at 4°C with or without 1.5 mol (2.7 mM) of soluble mTa peptide competitors (lacking biotin). Subsequently, binding reactions were incubated for 1 h on ice with avidin resins to which 10 nmol of the indicated biotinylated mTa peptide had been previously immobilized. Binding reactions were then washed twice with TBS/det, once with this buffer containing 0.5 M NaCl, and twice more with TBS/det. Bound materials were analyzed by SDS-PAGE, electrotransfer to PVDF, and autoradiography.
To assess binding of p56 lck and p56 lck folding intermediates directly from RRL translation reactions, reactions were chilled after p56 lck synthesis, treated with 1 mM N-ethylmaleimide to inactivate phosphatases and chaperone machinery, diluted into 200 l of binding buffer (10 mM sodium phosphate, 120 mM NaCl, 1% (w/v) Triton X-100, 1.0 mM sodium vanadate, 1.0 mM dithiothreitol) containing or lacking 1.2 mM soluble competitor peptides, and incubated for 1 h at 4°C. Subsequently, reactions were incubated for 1 h at 4°C with agarosepeptide resins. After binding, reactions were washed four times with binding buffer. Resins also were subjected to a stringency wash containing 0.5-2.0 M NaCl, as indicated. Coadsorption of the p56 lck ⅐hsp90 complex was detected by SDS-PAGE of resin-bound materials and Western blotting with 0.5 g/ml rabbit anti-hsp90 antibodies (PA3-013, Affinity BioReagents). The folding status of soluble and peptidebound p56 lck molecules was analyzed by mild proteolytic nicking assays as described previously (10,23) and by the addition of 10 l of control naive RRL per 250 l of dilute protease solution prior to its addition to resin-bound protein molecules, respectively. Kinase activities of antibodyor peptide-captured p56 lck molecules were assayed (25), with modifications as described previously (10). Immunoadsorptions of p56 lck were performed using polyclonal antibodies raised in rabbits repeatedly immunized with full-length recombinant p56 lck . Geldanamycin was provided by the Drug Synthesis and Chemistry Branch, National Cancer Institute.
Assays of hsp90 Dependence after de Novo Folding-Maturation of wild type p56 lck and of the mutant form p56 lck F505 (in which the regulatory tyrosine at position 505 was replaced with phenylalanine (26)) was assessed following translation and de novo folding in Promega TnT reticulocyte lysates. To establish the time required for p56 lck folding to reach an equilibrium, protein synthesis in the presence of [ 35 S]Met was carried out at 37°C for an initial 20 min, after which, reinitiation of protein synthesis was inhibited by 75 M aurintricarboxylic acid. Subsequently, 45-l aliquots of the translation reaction were transferred to truncated microtubes, allowing incubation of reactions at 37°C without significant evaporation or loss of chaperone vigor. Reactions were incubated for the indicated times at 37°C prior to flash-freezing in liquid nitrogen. After all samples representing the individual time points were collected, samples were thawed, immunoadsorbed with anti-p56 lck antibodies, and assayed for p56 lck activity. Protein synthesis was assayed by liquid scintillation counting of Requirements for hsp90-mediated maintenance of p56 lck activity were assessed using the techniques described above. Protein synthesis reactions programmed with wild type p56 lck or with the F505 mutation of p56 lck were assembled and incubated at 37°C for 20 min, after which, initiation of protein synthesis was arrested. Reactions were further incubated for 60 min to allow posttranslational folding to reach an equilibrium, after which geldanamycin or Me 2 SO (vehicle control) were added. Reactions were further incubated at either 30°C or 37°C for 1 h. Alternatively, geldanamycin was added prior to the initiation of p56 lck translation to assess its effects on de novo folding of p56 lck . p56 lck activity and structure were assessed as described previously (10).

RESULTS
To assess the status of SH2 domains on p56 lck folding intermediates, monomeric and hsp90-bound [ 35 S]p56 lck populations were produced by de novo synthesis in rabbit reticulocyte lysate translation reactions and separated on glycerol gradients. For each population thus produced, the functional status of its SH2 domain was analyzed by assessing the ability to bind to immobilized phosphotyrosine peptides. Monomeric [ 35 S]p56 lck showed low levels of nonspecific binding to avidin resins retaining the control unphosphorylated mTa peptide (Fig. 1A). In contrast to this nonspecific binding, resins retaining the phosphotyrosine mTa peptide bound 3-fold more monomeric [ 35 S]p56 lck than did control resins (Fig. 1A). Specificity of binding was confirmed by preincubation of monomeric [ 35 S]p56 lck with soluble competitor phosphopeptide; soluble competitor phosphopeptide reduced [ 35 S]p56 lck binding to background levels (Fig. 1A). In contrast to the phosphorylated soluble competitor peptide (pY), nonphosphorylated competitor peptide (Y) did not inhibit binding of [ 35 S]p56 lck to phosphopeptide resins (Fig.  1A). Using these criteria for specificity, phosphopeptide resins specifically retained approximately 15-50% of the input mono- After translation at 30°C, reactions were chilled, diluted, and incubated with phosphotyrosine peptide resin. Subsequently, resins were washed, eluted, and analyzed by SDS-PAGE, electrotransfer to PVDF, and Western blotting with anti-hsp90 antibodies. hsp90 recovered from binding reactions or applied as standard via an aliquot of RRL is indicated. Migrations and molecular masses of standards (kDa) are indicated along the right side of the panel. occurring in the p56 lck ⅐hsp90 complex was bound by phosphorylated mTa peptide resins in a specific fashion (Fig. 1A). This binding was qualitatively and quantitatively equivalent to that observed for the monomer similarly produced. These results indicated that the SH2 domain of hsp90-bound p56 lck was folded and functional, or was capable of becoming so during binding assays.
To determine whether hsp90 remained bound to p56 lck during mTa binding, covalently cross-linked mTa phosphopeptide resins were used to adsorb p56 lck folding intermediates directly from RRL translation reactions. To potentiate the efficiency of mTa binding, these assays utilized a point-mutant of p56 lck in which the regulatory tyrosine at position 505 was mutated to phenylalanine (subsequently called p56 lck F505). After binding and washing, proteins bound to the mTa peptides were analyzed by SDS-PAGE and Western blotting with anti-hsp90 antiserum. hsp90 showed no affinity for mTa peptide resin in the absence of p56 lck (Fig. 1B, lck template, Ϫ). In contrast, hsp90 could be coadsorbed from translation reactions following a brief (30 min at 30°C) synthesis of p56 lck F505 (Fig. 1B, lck template, ϩ). This specific coadsorption of the hsp90⅐p56 lck F505 complex was stable under washing with buffers containing up to 2.0 M NaCl (not shown). These results, in agreement with those presented in Fig. 1A, demonstrated that the SH2 domains of hsp90-bound p56 lck molecules were folded and functional.
However, structure/function relationships for the steroid hormone receptor-hsp90 heterocomplex suggested that although hsp90-bound p56 lck folding intermediates might contain folded and functional SH2 domains, hsp90 might nonetheless be required for SH2 domain folding. For steroid hormone receptors, the molybdate-stabilized hsp90⅐receptor complex is capable of high affinity steroid hormone binding (see, for example, Refs. [27][28][29]. However, hsp90-dependent hormone receptors require hsp90 function to acquire and/or maintain this high affinity hormone-binding conformation (18,30). To determine whether hsp90 function was necessary for efficient folding of p56 lck F505 SH2 domains within the context of the whole p56 lck molecule, we utilized the hsp90 inhibitor geldanamycin (16), which inhibits hsp90-mediated folding processes by markedly decreasing the K app of hsp90 machinery for ATP (19). Geldanamycin (5 g/ml) and the drug vehicle Me 2 SO (0.5%, v/v) had no effect on the rates or magnitude of [ 35 (Fig. 2, dmso). Binding of [ 35 S]p56 lck F505 molecules to unphosphorylated mTa peptide was not observed, nor was binding observed following preincubation with soluble phosphopeptide competitor, indicating that the recovery of [ 35 S]p56 lck F505 was specific (Fig. 2). Similarly, synthesis of the [ 35 S]p56 lck F505 population in the presence of the hsp90 inhibitor geldanamycin resulted in a population that was competent to specifically bind mTa peptide (Fig. 2,  geldanamycin). In this work, as in our previous work (10), geldanamycin-mediated inhibition of kinase folding was not coupled to enhanced degradation of p56 lck by endogenous RRL proteases. These experiments demonstrated that geldanamycin-inhibitable hsp90 function was not necessary for the efficient de novo folding and/or stabilization of the p56 lck SH2 domain within the context of the whole p56 lck protein.
Although the folding of the p56 lck SH2 domain was independent of hsp90 function (Fig. 2), we have previously observed that geldanamycin-mediated inhibition of hsp90 function results in the production of p56 lck molecules that are deficient in kinase activity and that have compromised integrity in subsequent protease nicking assays (10). To determine whether SH2-folded and kinase-deficient p56 lck molecules represented two separate populations of p56 lck , [ 35 S]p56 lck F505 molecules with folded and functional SH2 domains were recovered from control or geldanamycin-treated protein synthesis reactions. p56 lck F505 was the only tyrosine kinase captured by mTa adsorptions from RRL translations (Fig. 3A). Recovery of [ 35 S]p56 lck F505 and p56 lck F505 kinase activity by mTa resins was specific, in that its recovery could be prevented by soluble competitor peptide (Fig. 3B, top panel, (Fig. 3B, top panel). However, [ 35 S]p56 lck F505 molecules synthesized in geldanamycin-treated protein synthesis reactions had no detectable kinase activity (Fig. 3B, bottom  panel). Inhibition of hsp90 function by geldanamycin resulted in the isolation of a population of p56 lck molecules that had folded and functional SH2 domains but lacked kinase activity.
In vitro proteolytic nicking assays of mTa-captured [ 35 (Fig. 4, dmso) that differed from that previously observed for the free soluble kinase (10). This difference did not reflect differing structures of F505 versus wild type p56 lck (not shown), but instead, it likely reflected allosteric changes in the kinase domain produced upon binding of the SH2 domain to the high affinity mTa peptide resin (33). When synthesized in the presence of geldanamycin, molecules of mTa-bound [ 35 S]p56 lck F505 were hypersensitive to proteolytic nicking relative to those produced in the absence of the hsp90 inhibitor (Fig. 4, dmso versus  geldanamycin). Although their SH2 domains were folded and functional, p56 lck F505 molecules produced in geldanamycintreated translation reactions were deficient in the folded tertiary structure necessary to protect their catalytic domains from mild proteolytic nicking.
Although hsp90 facilitates de novo folding of p56 lck and also provides prolonged reiterative support of p56 lck F505 kinase function in the RRL model system, we have speculated that activities or events specific to T cells may stabilize p56 lck structure to an hsp90-independent state following hsp90-dependent de novo folding (10). This speculation was based on the observation that hsp90-bound p56 lck was detected only in the cytoplasmic fractions of T-cell lysates and appeared to represent a small portion of the total p56 lck population in vivo. One such potential stabilizing event was the phosphorylation of Y505 within the p56 lck C-terminal regulatory tail and its subsequent interaction with the SH2 domain.
Before assessing the maintenance of p56 lck structure and function by hsp90 machinery, the kinetics of p56 lck F505 folding were assessed. For such assessment, p56 lck F505 was produced in protein synthesis/folding reactions containing active hsp90 machinery. Then, reinitiation of protein synthesis was arrested and the "synchronized" kinase population was subsequently allowed to fold. Incorporation of [ 35 S]Met into newly synthesized p56 lck molecules terminated 5-10 min after inhibition of initiation of translation. Similarly, increases in p56 lck F505 kinase activity reached a plateau within 25 min of inhibition of initiation. Thus, the de novo folding of the synchronized p56 lck F505 population reached an equilibrium within 25 min of arrest of initiation of protein synthesis under these conditions. Using this established time frame, hsp90-mediated maintenance of p56 lck F505 function was assayed by adding geldanamycin well after de novo folding had reached an equilibrium. Nonetheless, late addition of geldanamycin caused p56 lck F505specific phosphotransferase activity to decline to 36 Ϯ 8% (Ϯ 1 S.D.; n ϭ 5) of that seen relative to Me 2 SO controls (e.g. Fig. 5A,  F505). In contrast, wild type p56 lck molecules were nearly independent of geldanamycin-inhibitable hsp90 function, retaining 88 Ϯ 9% (Ϯ 1 S.D.; n ϭ 5) of their phosphotransferase activity following a 1-h incubation at 37°C in the presence of geldanamycin (e.g. Fig. 5A, wild type). In contrast to the hsp90 dependence shown by p56 lck F505 at 37°C, neither kinase population showed significant requirements for hsp90-mediated maintenance at the lower temperature of 30°C (Fig. 5B). Additionally, as wild type kinase molecules acquired hsp90 independence following synthesis and maturation at 37°C, the FIG. 4. Structure of peptide-captured p56 lck produced in the presence of geldanamycin. [ 35 S]p56 lck F505 was produced at 30°C in rabbit reticulocyte lysate translation reactions lacking (dmso) or containing (geldanamycin) 5 g/ml geldanamycin. After translation and maturation, [ 35 S]p56lck was adsorbed to mTa phosphopeptide resins. Additionally, an aliquot of each translation reaction was preincubated with soluble competitor phosphopeptide to assess specificity of phosphopeptide binding (C). After binding and washing, peptide-captured materials were subjected to mild proteolytic nicking under nondenaturing conditions and analyzed by SDS-PAGE, electrotransfer to PVDF, and autoradiography. Migrations and molecular masses of standards (kDa) are indicated along the right side of the panel.

FIG. 5.
Conditional requirements for hsp90-mediated maintenance of p56 lck kinase activity. Wild type or F505 mutant [ 35 S]p56 lck molecules were produced in RRL translation/folding reactions and allowed to mature to equilibrium at 37°C. After maturation, either drug vehicle (geldanamycin Ϫ) or 5 g/ml geldanamycin (geldanamycin ϩ) were added, and reactions were further incubated for 1 h at 37°C. Subsequently, reactions were immunoadsorbed with anti-p56 lck antibodies, an aliquot of the reaction was analyzed by SDS-PAGE and autoradiography ([ 35 S]p56 lck ), and p56 lck kinase activity was assessed using the balance of each immunoadsorption. Autophosphorylated products ([ 32 P]p56 lck ) and phosphorylated exogenous substrate ([ 32 P] enolase) were analyzed by SDS-PAGE and detected by autoradiography through an intervening matrix to quench 35 (Fig. 6B). In contrast, the amount of hsp90 detected following coimmunoadsorption of p56 lck F505 declined only slightly upon chase incubation (not shown).
Four lines of evidence suggested that the differing requirements of wild type p56 lck versus mutant p56 lck F505 for hsp90mediated support resulted from regulatory phosphorylation of the tyrosine present at position 505 of wild type p56 lck molecules. 1) RRL contains csk protein (which is responsible for phosphorylation of the regulatory C-terminal tyrosine of p56 lck and cellular p60 src ) detectable by Western blotting (not shown); 2) in the absence of geldanamycin, the kinase activity of wild type p56 lck molecules was repressed relative to that of p56 lck F505 (see, for example, Fig. 5A, F505 versus wild type in geldanamycin-free reactions), consistent with the csk-mediated phosphorylation of Y505 and subsequent repression of p56 lck activity; 3) wild type p56 lck became incompetent to bind to mTa peptides following maturation and repression in RRL (Fig. 6A); and 4) although the amount of geldanamycin-mediated loss in p56 lck F505 kinase activity was highly reproducible among in-dividual lots of RRL used for these experiments, variation was noted for one lot of RRL. This lot synthesized both lck isoforms poorly, and the resulting lck kinase populations had nearly equivalent kinase activity for both wild type and F505 p56 lck . For this lysate, geldanamycin treatment led to losses in kinase activity for both wild type and F505 p56 lck molecules. However, supplementation of this lysate with purified csk restored repression of wild type p56 lck kinase activity and restored hsp90 independence for wild type p56 lck populations but did not affect the hsp90 dependence of p56 lck F505 (not shown). Together, the above observations suggested that C-terminal phosphorylation of p56 lck occurred in RRL and correlated with stabilization of p56 lck kinase molecules to an hsp90-independent state. However, due to technical limitations, we have not yet directly confirmed that the negative regulatory tyrosine of p56 lck (position 505) is phosphorylated in RRL.
To further characterize these hsp90-dependent folding events, proteolytic nicking assays of the native structure of soluble p56 lck molecules were used to compare the consequences of inhibiting de novo folding with the consequences of inhibiting hsp90-mediated maintenance folding. As previously shown (10), geldanamycin-mediated inhibition of hsp90 prior to p56 lck synthesis resulted in kinase molecules that were highly vulnerable to mild proteolysis relative to the properly folded kinase (Fig. 7A, de novo). When geldanamycin was applied subsequent to equilibrium folding of wild type p56 lck , geldanamycin-mediated inhibition of hsp90 did not result in the proteolytic sensitivity characteristic of inhibited de novo folding (Fig. 7A, de novo versus matured). This result was consistent with the finding that wild type p56 lck had only modest requirements for hsp90-mediated maintenance (Fig. 5A). Surprisingly, the protease sensitivity of p56 lck F505 was not detectably altered following post-maturational incubation at 37°C in folding reactions containing geldanamycin (Fig. 7B). For an aliquot of the specific population of p56 lck F505 molecules fingerprinted in Fig. 7B, immunoadsorption and assay of kinase activity confirmed the typical 65% loss of kinase function induced by geldanamycin (Fig. 7C), despite the lack of detectable change in the protease nicking fingerprint (Fig. 7B). Although disactivation of the kinase in the absence of hsp90 function led to losses in kinase function, these losses were not accompanied by the dramatic loss of tertiary structure that characterized the inhibition of de novo folding. Thus, protease hypersensitivity was a specific consequence of inhibition of de novo hsp90-mediated protein folding, but it did not accompany loss of kinase function per se.

DISCUSSION
The cellular nonreceptor tyrosine kinase p56 lck is a multidomain regulatory protein that requires geldanamycin-inhibitable hsp90 function to demonstrate a protease-resistant state (10). Thus, we were prompted to ask whether inhibition of hsp90 function by geldanamycin resulted in global misfolding of the whole p56 lck molecule or only in the misfolding of specific individual domain(s). Results presented here indicate that inhibition of hsp90 results in deficiencies in p56 lck folding that are specific to the catalytic domain, because specific SH2-mediated recognition of phosphotyrosine-containing peptides is not impaired (Fig. 2) for molecules of p56 lck in which the folding of the catalytic domain is clearly compromised (Figs. 3 and 4). Consistent with this finding, hsp90 associates with p56 lck molecules containing folded and functional p56 lck SH2 domains (Fig. 1). The interaction of hsp90 with p56 lck is stable in the presence of up to 2.0 M NaCl, suggesting that hsp90 is binding to p56 lck via direct hydrophobic interactions rather than associating electrostatically with kinase-bound hsp90 cohorts. Together, these observations suggest that hsp90 acts upon the FIG. 6. Effects of maturation incubations on intra-and intermolecular interactions of p56 lck . A, RRL protein synthesis/folding reactions were (lck template ϩ) or were not (lck template Ϫ) programmed to synthesize wild type [ 35 S]p56 lck and were incubated for 20 min at 37°C. Subsequently, initiation of protein synthesis was arrested, polyribosomes were allowed to run off for 7 min, and aliquots of the synthesis reactions were immediately chilled and incubated with phosphotyrosine phosphopeptide (chase time 0). Alternatively, the balance of the protein synthesis/folding reaction was matured for an additional 60 min at 37°C prior to chilling and incubation with phosphopeptide resin (chase time 60). After each reaction had been incubated with phosphopeptide resin for 3 h, resins were washed and eluted, and p56 lck was detected by Western blotting with anti-p56 lck antibodies (p56 lck ). Additionally, an aliquot of the mock translation reaction was applied directly to the SDS-PAGE gel to provide a control for antibody specificity (RRL std). Migrations and molecular masses of standards (kDa) are indicated along the right side of the panel. B, wild type [ 35 S]p56 lck was synthesized and matured in the presence of [ 35 S]Met as described for A, above. After the indicated minutes of maturation (chase time), p56 lck was immunoadsorbed from synthesis/folding reactions with anti-p56 lck antibodies. Immunoadsorbed [ 35 S]p56 lck was detected by SDS-PAGE/ autoradiography of an aliquot of each immunoadsorption reaction (top panel), whereas the balance of the adsorption reaction was analyzed by Western blotting with anti-hsp90 antibodies (bottom panel). Additionally, an aliquot of reticulocyte lysate was applied directly to the SDS-PAGE gel to supply a standard for the detection of hsp90 (RRL std). p56 lck , hsp90, and the migrations and molecular masses of standards in kDa are indicated.
hsp90, p56 lck , and Folding of Individual Protein Domains catalytic domain of p56 lck rather than the whole kinase molecule or other domains within p56 lck . Consistent with this functional analysis, the catalytic domains of viral p60 src , fes, and raf represent minimal domains required for physical interactions between these kinases and hsp90 (34 -36).
Based on the functional status of SH2 domains occurring on hsp90-bound p56 lck molecules and on the consequences of inhibiting hsp90 function, we postulate the existence of p56 lck molecules with folded and functional SH2 domains and unfolded, incompletely folded, and/or misfolded p56 lck catalytic domains (SH2 f -cat u ). These SH2 f -cat u molecules of p56 lck , then, appear to be substrates of hsp90 and, by inference, certain hsp90 cohort(s). The postulated interaction of hsp90 with partially folded p56 lck is consistent with our current understanding of hsp90-mediated folding/renaturation of other proteins. hsp90-bound steroid hormone receptors are capable of simultaneously binding hormone and hsp90 in the presence of the stabilizing agent molybdate (27)(28)(29). Also consistent with this understanding is the suggestion that hsp90 recognizes highly structured dimeric intermediates produced during the thermal denaturation of citrate synthetase (37). Additionally, the hsp90 homolog GRP94 (the 94-kDa glucose-regulated protein occurring in the endoplasmic reticulum) interacts with immunoglobulin folding intermediates in the endoplasmic reticulum after oxidation of the immunoglobulin subunits, but prior to dimerization of these subunits (38,39). Furthermore, the existence of individual folded protein domains on a chaperone-bound protein has precedence in that tubulin folding intermediates that are bound to the chaperonin GroEL have been reported to exhibit GTP binding activity (40).
These SH2 f -cat u molecules of p56 lck appear to represent one or more intermediates produced as part of the de novo folding pathway of p56 lck molecules. The placement of SH2 f -cat u on the de novo folding pathway rather than on a maintenance or "disactivation" pathway is based on several observations. 1) Newly synthesized p56 lck molecules associate with hsp90 in a very rapid and nearly quantitative fashion immediately after their biogenesis in RRL (10); this finding is consistent with pulse-chase analyses of viral p60 src maturation in vivo (12,13).
2) hsp90-mediated folding of the heme-regulated eIF-2␣ kinase is accompanied by cotranslational association of hsp90 with polyribosomes synthesizing nascent kinase molecules (15). 3) At 30°C, both wild type and p56 lck F505 require hsp90 to attain active conformations, but neither kinase population requires hsp90 function to maintain its activity at this temperature (Fig. 5). The p56 lck folding intermediates characterized in phosphopeptide binding assays (Figs. 1-4) were generated under conditions designed to yield de novo folding intermediates (brief syntheses at 30°C) rather than those conditions (1-h incubations at 37°C) required for partial disactivation of p56 lck F505 (see Fig. 5, B versus A). 4) Wild type p56 lck does not require significant hsp90-mediated maintenance in RRL, yet can be isolated as an abundant hsp90-bound entity immediately after its synthesis (23). 5) The magnitude of the requirements of p56 lck F505 for hsp90 function are much greater during or immediately after kinase synthesis relative to similar requirements seen after kinase maturation (e.g. Fig. 3B versus Fig. 5A).
Based on these results, we propose that the biogenesis of the multidomain protein tyrosine kinase p56 lck occurs via a chaperone-mediated protein folding pathway in which the p56 lck SH2 domain folds rapidly, independently of hsp90 function, followed by the slower, hsp90-dependent folding of the C-terminal catalytic domain. This model for the de novo folding of p56 lck is consistent with the renaturation pathways for chemically denatured multidomain proteins (reviewed in Ref. 1). The proposed model for p56 lck folding is also consistent with postulations that cotranslational folding of individual protein domains may be directed by ribosomal pausing (41) and is consistent with findings that isolated SH2 and SH3 domains fold and function efficiently in context-independent fashions (42)(43)(44). Additionally, our data support a previous speculation that hsp90 function facilitates interdomain packing of viral p60 src (45).
In addition to the role of hsp90 in de novo folding, hsp90 is necessary in RRL to maintain the function of p56 lck molecules with a single amino acid change in the C-terminal regulatory tail (10) (Fig. 5A). However, although hsp90 is necessary to maintain the active conformation of p56 lck F505 in RRL, hsp90bound p56 lck represents a small portion of the total kinase population in vivo (10). Based on this observation, we have previously suggested a model for kinase maturation in which hsp90 both facilitates kinase biogenesis and maintains kinase structure prior to specific in vivo stabilizing events (10). This model for hsp90 function also nicely describes the maturation of an unrelated kinase, HRI (the heme-regulated kinase re- FIG. 7. Effects of timing of geldanamycin application on p56 lck protease sensitivity. Either drug vehicle (Me 2 SO, dmso) or geldanamycin was added to RRL protein synthesis/folding reactions prior to synthesis of [ 35 S]p56 lck at 37°C (de novo). Alternatively, drug additions were performed after p56 lck folding had reached an equilibrium in RRL protein synthesis/folding reactions, and the reactions were further incubated for 1 h at 37°C (matured). A and B, after incubation, synthesis/ folding reactions were chilled on ice and subjected to mild proteolysis under nondenaturing conditions. Shown is an autoradiogram of 35 S proteolysis products. The full-length lck translation product (*) and migrations and molecular masses of standards in kDa are indicated. A, wild type p56 lck ; B, p56 lck F05. In C, aliquots of the reactions characterized in B (application of either Me 2 SO (D) or geldanamycin (G) after prior maturational incubations of p56 lck F505) were analyzed by immunoadsorption with anti-p56 lck antibodies and assay of kinase activity. [  hsp90, p56 lck , and Folding of Individual Protein Domains sponsible for phosphorylating the ␣ subunit of eucaryotic initiation factor 2), and its transformation from an hsp90-dependent state to an hsp90-independent state (15). For HRI, transformation to an hsp90-independent state requires autophosphorylation and perhaps other uncharacterized phosphorylation events. Data presented here suggest that regulatory phosphorylation of tyrosine 505 of p56 lck may represent a similar kinase stabilizing event, because mutant p56 lck molecules that cannot be phosphorylated at tyrosine 505 appear to have greater requirements for hsp90-mediated maintenance that do wild type molecules with the theoretical capacity for such phosphorylations (Figs. 5 and 6).
However, it remains to be determined whether stabilization of p56 lck structure via regulatory C-terminal phosphorylations is relevant to p56 lck maturation in vivo. We have not directly documented correlations between hsp90-p56 lck associations and phosphorylation of the negative regulatory tyrosine of p56 lck . Furthermore, in vivo, the interaction of hsp90 with newly synthesized cytoplasmic molecules of viral p60 src is transient, despite the lack of an analogous regulatory phosphorylation for this kinase (12,13). Additionally, other phosphorylation events or other postranslational modifications of p56 lck may influence its thermal stability. However, preliminary evidence presented here suggests that regulatory phosphorylations may facilitate the production the mature hsp90-free p56 lck molecules detected at the plasma membrane (10). Such phosphorylations may stabilize the catalytic core via enhancing associations between this core and the SH3 and SH2 domains, because deletion of the SH3 and SH2 domains of viral p60 src exaggerate its association with hsp90 in vivo (for review, see Ref. 34).
However, we do not suggest that hsp90 holds p56 lck in a specific static, or "poised," conformation until regulatory phosphorylation of the kinase triggers physical release from hsp90. Rather, we suggest that hsp90 machinery functions in reiterative cycles of binding and release typical of other chaperones. According to this interpretation, hsp90 releases p56 lck molecules irrespective of their phosphorylation status. After release, kinase molecules may experience loss of tertiary structure and may thus require repair by hsp90 machinery, i.e. hsp90 maintains the kinase population via protein refolding. According to this model, the effects of regulatory phosphorylations are to stabilize kinase structure such that this loss of structure does not occur, thus freeing kinase molecules from their dependence on hsp90-mediated refolding and terminating their physical associations with hsp90 machinery.
Consistent with our hypothesis that hsp90 has roles in both the de novo folding and in the conditional maintenance of kinases, newly synthesized p56 lck F505 is not equivalent to matured p56 lck F505 with regards to the consequences of hsp90 inhibition. For newly synthesized p56 lck F505, geldanamycinmediated inhibition of hsp90 function produces dramatic hypersensitivity to protease (Fig. 7B, de novo). For "matured" p56 lck F505, geldanamycin-mediated inhibition of hsp90 function results in significant loss of kinase specific activity (Fig.  7C), but this loss is not accompanied by detectable changes in protease hypersensitivity (Fig. 7B, matured). Thus, the aberrant structure underlying loss of mature p56 lck F505 function is too subtle to be detected by proteolytic nicking assays. Because these proteolytic fingerprints differ depending on whether geldanamycin is applied before or after de novo folding, the tertiary structure of p56 lck F505 must develop in an hsp90-dependent fashion during de novo folding. This interpretation is supported by the temperature dependence of the effects of geldanamycin on p56lckF505 "disactivation" versus the temperature-independent effects of geldanamycin on de novo fold-ing (Fig. 5). Furthermore, these results suggest that hsp90 may maintain p56 lck F505 structure without reiterating the full de novo folding pathway.
In conclusion, data reported here indicate that hsp90-mediated de novo folding of p56 lck proceeds via folding pathways similar to those utilized during in vitro renaturation of purified multidomain proteins in the absence of chaperones. Specifically, p56 lck biogenesis appears to occur via a novel chaperonemediated folding pathway in which the p56 lck SH2 domain rapidly folds independently of hsp90 function, followed by the slower, hsp90-dependent folding and maintenance of the catalytic domain prior to the final stabilization of mature protein structure via interdomain interactions. These results suggest that hsp90 machineries may function similarly in other scenarios, folding and refolding substrate proteins prior to the stabilization of their structures by the regulated establishment of intra-or intermolecular interactions between domains. Consistent with this possibility, the physical association of hsp90 with certain transcription factors (46) or kinases (47) can be specific to those forms of these proteins lacking essential partner proteins. Thus, we favor models for the biogenesis of certain regulatory proteins in which pathways of de novo folding and maintenance folding overlap with pathways of regulation.