Refolding Intermediates of Acid-unfolded Mitochondrial Aspartate Aminotransferase Bind to hsp70*

The cytosolic (cAAT) and mitochondrial (mAAT) isozymes of eukaryotic aspartate aminotransferase share a high degree of sequence identity and almost identical three-dimensional structure. The rat liver proteins can be refolded and reassembled into active dimers after unfolding at low pH. However, refolding of the mitochondrial form after unfolding at pH 2.0 is arrested in the presence of hsp70, whereas this chaperone does not affect the refolding of the cytosolic isozyme unfolded under similar conditions. To elucidate the nature of the differential interaction between hsp70 and the two transaminase forms, we have characterized their refolding from their acid-unfolded states. The recovery of activity of the cytosolic enzyme is monophasic and can be adequately described by a single first-order reaction. By contrast, two sequential first-order rate-limiting steps can be detected for the refolding and reactivation of the mitochondrial protein. The overall refolding pathway of mAAT includes a very fast collapse to an intermediate with 80% of the secondary structure of the active dimer. This is followed by a slow isomerization to form assembly-competent monomers that rapidly associate to form an inactive dimer and a final structural rearrangement of the dimer to the native conformation. Analysis of the interaction of hsp70 with intermediates along the folding pathway of mAAT shows that the polypeptide loses its ability to bind to the chaperone after it has proceeded through the first isomerization/fast dimerization steps. Thus it appears that only the first collapsed intermediate states in the folding of mAAT bind hsp70. By contrast a faster refolding of cAAT from this collapsed state could explain, at least in part, the inability of hsp70 to bind this isozyme.

Proteins that are synthesized in the cytoplasm but reside in other cellular locations must traverse membranes to reach their final destination. In the case of nuclear-encoded mitochondrial proteins, most are synthesized as precursors with N-terminal presequences that target them to mitochondria (1). In addition, a partially unfolded conformation of the precursor is a common requirement for efficient translocation (2)(3)(4). Various cytosolic proteins, including members of the hsp70 1 stress protein family, have been implicated in the binding of these incompletely folded precursors to presumably prevent their aggregation while maintaining them in a loose, import-competent conformation (5)(6)(7)(8), that is folding to the native conformation is delayed until the protein reaches its final destination. Several molecular chaperones present in the mitochondrial matrix appear to participate in both the translocation itself and the subsequent folding of the imported protein (9 -11).
The hsp70 family of molecular chaperones is present in the cells of all organisms. Eukaryotes contain multiple hsp70s some of which are expressed constitutively while others are induced in response to metabolic stress. They participate in a variety of cellular processes, including folding of nascent polypeptides, assembly and disassembly of protein complexes, translocation of proteins across membranes, and protein degradation (12,13). All of these activities involve binding and release of unfolded polypeptides which are regulated by the binding and hydrolysis of ATP (14,15). This ATPase activity resides in the highly conserved N-terminal domain of hsp70, whereas the peptide binding site is localized in the less well conserved C-terminal region (16 -18). The three-dimensional structure of each isolated domain has been determined (17)(18)(19). Analysis of the specificity of hsp70 for binding of synthetic peptides indicates that exposed hydrophobic residues are generally required (20,21). This supports the initial suggestion that hsp70 function involves the binding of aggregation-prone hydrophobic regions that may become exposed in the surface of proteins under conditions of stress (22) or during nascent chain elongation (23). However, it is still not clear whether hsp70 is required for the correct folding of all nascent chains in the cytoplasm. Previous studies in HeLa cells suggested that most newly synthesized proteins interact with cytosolic hsp70 (24). However, while this interaction was short-lived for some proteins, others remained complexed with hsp70 for considerably longer periods of time. The latter group might include proteins that require additional time to reach their native conformation, such as multimeric proteins or proteins that are post-translationally translocated to organelles. Others still might be able to fold without the collaboration of hsp70. The ability of hsp70 to recognize not just unfolded from folded states of a given protein but between unfolded proteins might contribute to the selective control of the folding of proteins in the cytosol. Unfortunately, very little attention has been directed at this aspect of the hsp70 function.
Naturally occurring isozymes such as the cytosolic (cAAT) and mitochondrial (mAAT) forms of aspartate aminotransferase found in eukaryotes provide unique systems to address this question. A wealth of structural and functional information is available for several vertebrate aspartate aminotrans-ferases (25). These homodimeric, pyridoxal 5Ј-phosphate (PLP)dependent enzymes share a high degree of sequence similarity (about 63%) and an almost identical crystal structure (25,26). Each of the two identical active sites is composed of residues from both subunits; therefore, only the dimeric state shows catalytic activity. The two AAT isozymes are encoded by the nuclear genome and synthesized in free polysomes in the cytoplasm. After synthesis, cAAT remains in the cytoplasm while mAAT is post-translationally transported into the mitochondrial matrix. As most translocated mitochondrial proteins, rat liver mAAT is synthesized as a precursor (pmAAT) with a 29-residue targeting presequence at its N-terminal end (27). We have previously reported that the two isozymes show remarkably different folding behavior when expressed in in vitro translation systems (28). After translation in rabbit reticulocyte lysate, newly synthesized cAAT folds within the dead time of the experiment (in Ͻ5 min), whereas pmAAT folding proceeds at a very slow rate (t1 ⁄2 Ͼ1 h at 15°C) (28). Furthermore, nascent pmAAT synthesized in rabbit reticulocyte lysate was found transiently associated with hsp70 (29). Although hsp70 has affinity for binding to mitochondrial presequences (29 -31), its interaction with pmAAT was independent of the presence of the presequence suggesting that hsp70 also binds to unfolded regions of the mature part of the precursor. By contrast, no cAAT could be detected in complex with hsp70 under identical conditions. These observations suggest that the folding of the two isozymes in the cytosol is selectively regulated. The folding of the translocated mitochondrial enzyme may be delayed to facilitate its efficient uptake by mitochondria while such a delay in folding would not be necessary for cAAT since it remains in the cytosol after translation.
Several factors can be considered as responsible for the selective interaction of hsp70 with only the mitochondrial isozyme of AAT. These include the intrinsic folding rate of the polypeptide chain as proposed for the Escherichia coli chaperone SecB, which is involved in protein export (32), or the conformational properties of certain intermediates along the folding pathway. To address this issue, we have now analyzed the interaction of the AAT isozymes with the constitutive cytosolic hsp70 (or hsc70) purified from bovine brain during their refolding in vitro from their acid-unfolded state. Only a single first-order rate-limiting step could be detected for the reactivation of cAAT. On the other hand, refolding of the mitochondrial enzyme is a multistep process that includes a non-native dimeric intermediate state. When present in the refolding medium, hsp70 binds to intermediate(s) along the folding pathway of mAAT but fails to interact with acid-unfolded cAAT. Binding of hsp70 to pmAAT is lost after the refolding protein forms assembly-competent monomers. These results indicate that, as observed for the proteins newly synthesized in cell-free extracts, hsp70 seems to have a preference for binding to the unfolded state of only the mitochondrial form of aspartate aminotransferase. The distinct refolding kinetics of the isozymes may contribute to the selective recognition of the mitochondrial protein by hsp70.

EXPERIMENTAL PROCEDURES
Materials and Protein Purification-Aspartic acid, cysteine sulfinic acid, and ␣-ketoglutarate were purchased from Sigma. All other reagents were of the highest purity available. Published procedures were followed for the purification of pmAAT and cAAT and for the preparation of mature mAAT (28) and of the apoenzyme form (33). Stock solutions of the enzymes (0.2-0.4 mM) were prepared in 2 mM Tris-HCl, pH 7.5, by ultrafiltration (Centricon, 30,000 M r cut-off). Protein concentrations were estimated from the absorbance at 356 nm of the pyridoxal 5Ј-phosphate cofactor using a ⑀ ϭ 8,500 M Ϫ1 cm Ϫ1 and M r ϭ 46,597 for pmAAT and M r ϭ 46,399 for cAAT. The enzyme activity was measured at 37°C in a coupled assay with malate dehydrogenase, using L-aspartate and ␣-ketoglutarate as substrates as described previously (34). hsp70 was purified from bovine brain following published procedures with minor modifications (35). Briefly, bovine brains were homogenized in a Waring blender with 2 volumes of buffer A (10 mM Tris acetate, 10 mM NaCl, 0.1 mM EDTA, 10 mM ␤-mercaptoethanol, pH 7.5). Following centrifugation at 15,000 rpm for 30 min in a 19Ti rotor, the supernatant was chromatographed on a DE52 column using 150 mM NaCl in buffer A (buffer B) as elution buffer. The eluate was loaded directly on a hydroxylapatite column equilibrated with buffer A, and the proteins were eluted with 150 mM potassium phosphate in buffer A. The fractions containing hsp70 were identified by SDS-PAGE analysis under reducing conditions. After dialysis against buffer A, 5 mM magnesium acetate was added before loading the pooled fractions on an ATPagarose column. After washing the column with 10 volumes of 1 M NaCl in buffer A followed by 3 volumes of buffer A, hsp70 was eluted with 3 mM ATP in buffer B. The excess of ATP was removed by precipitation with 80% ammonium sulfate. The pellet was resuspended in the minimum volume of buffer A and dialyzed extensively against the same buffer. The hsp70 preparation was over 95% homogeneous according to SDS-PAGE analysis. The functional state of hsp70 was assessed by testing its ability to arrest refolding of pmAAT (see "Results"). Protein concentration was measured using an ⑀ 280 ϭ 47,800 M Ϫ1 cm Ϫ1 (15). Stock solutions of hsp70 were kept at 4°C in 20 mM Tris-HCl, 10 M ␤-mercaptoethanol, pH 7.5, and remained active for up to 12 months. When a decrease in activity was observed upon storage, a partial recovery could often be achieved by addition of 10 M ␤-mercaptoethanol.
Unfolding and Refolding of AAT-Reversible guanidine hydrochloride (GdnHCl) or acid unfolding of the proteins was performed according to published procedures (33,36). Briefly, a stock solution of the enzymes in 2 mM Tris-HCl, pH 7.5, was denatured by addition of either GdnHCl to achieve a final concentration of 4 M or of diluted HCl to pH 2.0. The final protein concentration was 8.7 M. The samples were then incubated for 90 min at 25°C, conditions that were established previously to achieve maximum unfolding (36). Before initiation of refolding, unfolded protein samples were preincubated at 10°C for 15 min. Refolding and reactivation of AAT was performed by dilution of the unfolded protein at 10°C in refolding buffer (40 mM Hepes, 0.1 mM EDTA, 1 mM dithiothreitol, pH 7.5) to a final protein concentration of 0.1-1.8 M. The dead time of the manual mixing was 15-20 s. When necessary, the refolding buffer contained 10 M PLP or pyridoxamine 5Ј-phosphate (PMP). The recovery of activity was followed by measuring the enzyme activity of aliquots taken at different times after the initiation of refolding. When monitoring reactivation of the apoenzyme in the absence of coenzyme, the assay mixture was supplemented with 10 M PLP. The enzymatic activity was followed during less than 30 s to ensure that no significant increase in activity occurs during the determination. Reactivation data are expressed as percentage relative to the activity of a sample of native enzyme maintained under identical conditions. To study the effect of hsp70 on the refolding of the enzymes, different concentrations of hsp70 were added to the refolding mixture either previous to the addition of the unfolded proteins or at different times after the initiation of refolding.
Dimerization-A method based on the formation of heterodimers between pmAAT and mAAT (36) was used to monitor dimer formation during refolding. Refolding reactions containing 1.8 M pmAAT or mAAT were started in refolding buffer at 10°C as described above. At different times after starting the reactions, equal volumes of the refolding samples were mixed and incubated for an additional 120 min at 10°C to allow complete refolding. After removing the small amount of aggregated material by centrifugation, an equal volume of a 1,2-dipalmitoyl-sn-glycerophosphoglycerol:1,2-dipalmitoyl-sn-glycero-3-phosphocholine (3:1) liposome preparation was added to the supernatant and incubated at room temperature to allow binding of the precursorcontaining dimers. The bound protein was then separated by centrifugation on a Beckman airfuge equipped with an A-100 30°rotor, at 60 p.s.i. for 20 min. The pellet, containing the hybrid dimer and precursor homodimer, was analyzed by 12% SDS-PAGE. After staining with Coomassie Blue, the intensity of the protein bands present in the pellets was quantified with a Personal Densitometer Scanner (Molecular Dynamics). Control experiments performed using different known concentrations of mAAT indicated that the amounts of protein loaded on the SDS-PAGE gels (0.5-1.5 g) were well within the linear range for both stain binding and instrument response (at least 0.1-2.5 g). The supernatant, containing the mature homodimers, was assayed for enzymatic activity. Liposomes were prepared in refolding buffer containing 0.2 M NaCl as already described (36).
Refolding Monitored by Intrinsic Fluorescence-The recovery of native fluorescence during refolding was followed in a SLM 8000 C Aminco spectrofluorometer, using photon counting mode, by monitoring the intrinsic fluorescence emission at 338 nm after excitation at 280 nm. Usually 4-nm slits were used for both excitation and emission.
Refolding Monitored by CD Spectroscopy-Refolding was also followed by monitoring the appearance of specific dichroic signals that are characteristic of the native protein but are absent in the acid-unfolded species. All measurements were made in a Jasco J-720 spectropolarimeter, using either 1-or 0.1-cm cuvettes thermostated at 10°C. The acquisition of native secondary structure was monitored at 222 nm, and the binding of PLP to the active site during refolding was followed by monitoring the appearance of a strong Cotton effect due to the PLP centered at 356 nm (37). The values of molar ellipticity at 222 nm () were obtained by using the expression () ϭ (Ј/10)m/lc, where m is the mean residue molecular weight (114), l is the path length of the sample solution, c is the protein concentration in g/ml, and Ј is the direct value from the instrument. The units of () are degree cm 2 dmol Ϫ1 .
Analysis of the Complex hsp70⅐pmAAT-The complex between hsp70 and refolding pmAAT was isolated by centrifugation on a top-bench centrifuge (Marathon 14K/M, Fisher Instruments) at 10,000 rpm for 20 min, and the pellet and supernatant were analyzed on 12% SDS-PAGE gels. After staining with Coomassie Blue, the fraction of protein present in the pellet was estimated from the intensity of the protein bands determined using a Molecular Dynamics Scanner. While no hsp70 precipitated in the absence of pmAAT, a small fraction of pmAAT (ϳ10%) was recovered in the pellet of refolding reactions containing no hsp70. The intensity of the pmAAT band coprecipitating with hsp70 was in each case corrected for this hsp70-independent aggregation.
Data Analysis-Reactivation and CD 356 nm data were fit to the following sequential first-order reaction (Reaction 1), where A is the initial state, B an intermediate, and C the final state. The concentration of A, B, and C will change with time according to Equations 1-3, in which R is the total change of the variable being measured, and [A] t , [B] t , and [C] t are the concentrations of A, B, and C at time t, respectively. Nonlinear least squares fitting of the data to Equation 3 was performed using the Levenberg-Marquardt algorithm in the commercial program SigmaPlot 2.0 (Jandel Corp.) to estimate the parameters R, k 1 , and k 2 . The initial estimates of the rate constants could be varied by as much as 100%, either simultaneously or independently, and the fitting routine still converged on identical values of the kinetic parameters. Thus, these parameters represent a global minimum in the fitting procedure. A time course of the kinetic mechanism shown in Reaction 2 (see "Discussion") was simulated using the KINSIM program (38), and the rate constants for reactivation of the PLP form of pmAAT are shown in Table I, assuming diffusion controlled rates for the bimolecular steps and that activity is recovered only with the final phase of folding. The kinetic trace obtained coincided with the continuous line shown in Fig.  1A for the reactivation of pmAAT which represents the best fit of a single set of experimental data to Equation 3. The other kinetic data presented in this work were fit to a single exponential curve using the SigmaPlot program. The quality of the fit was judged by visual analysis of scatter diagrams of the residuals (difference between the experimental and fitted values) and by calculating the 2 statistic (39) using the Minitab program (Minitab Inc.) to verify whether there are significant differences between the experimental and calculated data.

RESULTS
Reactivation of Acid-unfolded AAT Isozymes-Incubation of pmAAT (36) or cAAT at pH 2.0 results in an almost instantaneous loss of activity. When the pH of a sample of acid-unfolded enzyme is returned to neutrality by rapidly diluting an aliquot of the unfolded enzyme in refolding buffer at 10°C, much of the original catalytic activity is recovered (65-85%). The catalytic properties as well as the absorbance, fluorescence, and circular dichroism spectra (far-UV and visible) of the refolded proteins are identical to those of the native form (data not shown). The yield of reactivation of either isozyme is dependent on the final protein concentration. At 10°C the optimal protein concentration ranges from 0.6 to 1.8 M, with typical yields of about 65-75% of the original activity for the mitochondrial enzyme and slightly higher for the cytosolic isozyme. At higher protein concentrations the yield of reactivation decreases, most likely due to increased aggregation of the protein. The yield of reactivation is also considerably reduced for the apoenzyme or the PMP forms of the enzyme. However, since the yield of reactivation is significant even in the absence of PLP (50 -60%), the presence of this cofactor is not required for the correct refolding of the protein.
The kinetics of reactivation of both isozymes after acid denaturation was monitored by assaying aliquots at different time intervals after initiation of refolding. The reactivation of the mitochondrial isozyme at 10°C after unfolding at pH 2.0 shows a sigmoidal behavior with a characteristic initial lag period of about 4 min (Fig. 1A). The sigmoidal kinetic trace indicates that the reactivation process cannot be determined by . The continuous lines represent best fits of the data to an irreversible two-step consecutive uni-unimolecular mechanism with rate constants of 0.4 min Ϫ1 and 0.038 min Ϫ1 for pmAAT and to a single exponential raise for cAAT (k ϭ 0.2 min Ϫ1 ). The lower panel shows the plot of the residuals for the reactivation of pmAAT in the absence of hsp70 (open circles). B shows a plot of the maximum activity recovered after incubation for 120 min at 10°C as a function of the hsp70/pmAAT ratio. only one single rate-limiting step. In addition, because the reactivation is independent of the protein concentration in the range 0.1-1.8 M (data not shown), a bimolecular rate-limiting step can be excluded. Therefore, the minimum kinetic model describing the reactivation of pmAAT must include two consecutive first-order reactions. Indeed, the sigmoidal reactivation kinetics can be easily fit by an irreversible two-step consecutive uni-unimolecular mechanism according to Reaction 1 (Fig. 1A). We are assuming that activity is recovered only in the final phase of folding which agrees with the fact that only the fully folded native dimer of pmAAT is active (25,40). In the mathematical treatment used to analyze Reaction 1 (41), irreversibility only implies that the reverse reactions have rate constants much smaller than those for the forward reactions. The continuous line in Fig. 1A for the reactivation of the PLP form of pmAAT represents best fit of the data to such a sequential model by using Equation 3 and rate constants of 0.4 min Ϫ1 and 0.038 min Ϫ1 . Within the limits of error, the same set of rate constants describes the kinetics of reactivation of the apoenzyme and PMP forms of pmAAT (Table I). The assignment of the two rate constant values to k 1 and k 2 as defined in Reaction 1 (Table I) is tentative since in a two-step consecutive reaction the order of the two rate constants cannot be established without independent information on the concentration of intermediate.
In contrast to the sigmoidal kinetic behavior of pmAAT, no lag phase was detected for the reactivation of the cytosolic enzyme unfolded at pH 2.0 under identical experimental conditions, including the protein concentration in the refolding reaction (0.1-1.8 M). For this enzyme, the overall recovery of activity is faster and can be adequately described by a single first order reaction mechanism with a rate constant of 0.2 min Ϫ1 (Fig. 1A).
Refolding Monitored by Fluorescence and CD Spectroscopy-The mechanism proposed above represents the minimal set of steps compatible with the reactivation kinetics of pmAAT. To test this model and further characterize the folding process, we analyzed the refolding kinetics of pmAAT unfolded at pH 2.0 by monitoring the changes in intrinsic fluorescence at 338 nm after excitation at 280 nm. Rat liver pmAAT contains seven tryptophanyl residues (27), one of which (Trp-140) is located near the coenzyme PLP at the active site (25). Due to quenching of the fluorescence of this residue upon binding of coen-zyme, the fluorescence intensity of native holoenzyme is lower than that of native apoenzyme (traces 1 and 2, Fig. 2B). A slight blue-shift in em (ϳ2 nm) is also observed upon binding of PLP to native apoenzyme. Unfolding of the protein at pH 2.0 results in a red shift of the emission maximum to 352 nm (trace 3 in Fig. 2B; 36) and a fluorescence intensity that is intermediate between those of the native apo-and holoenzyme forms. These changes reflect the combined effects of two opposing factors. Unfolding results in a decrease in fluorescence due to the higher accessibility of tryptophan side chains to the aqueous solvent but also in an increase in fluorescence caused by the relief of the quenching effect of PLP as the coenzyme is released from the active site.
Immediately after initiation of refolding by dilution and neutralization of the acid-unfolded protein, there is a shift in the emission maximum from the unfolded value of 352 nm to the native value of 338 nm (data not shown). This shift is complete within the dead time of the experiment (15-20 s). Since the unfolded protein has lower intrinsic fluorescence than the native apoenzyme (Fig. 2B), one would expect an increase in fluorescence at 338 nm during refolding of the enzyme in the absence of coenzyme. Yet, a decrease in fluorescence is observed instead ( Fig. 2A). This fluorescence decay can only be explained by assuming the rapid formation of a transient intermediate having higher fluorescence than either the unfolded and native apoenzyme during the dead time of the manual mixing. The rapid changes in the emission maximum and quantum yield of the intrinsic fluorescence are consistent with the sequestering of tryptophan side chains in the interior of the collapsed intermediate.
When refolding is performed in the absence of coenzyme (apoenzyme), the subsequent decrease in fluorescence of this transient intermediate is monophasic (Fig. 2A) with a rate   (2) forms of pmAAT was performed by diluting the unfolded protein directly in a fluorescence cuvette at 10°C at a final protein concentration of 40 g/ml after denaturation for 30 min at pH 2.0. All other conditions were as described in Fig. 1. Fluorescence was measured at 338 nm with ex ϭ 280 nm. Forty-five minutes after initiation of refolding of the apoenzyme in the absence of PLP, 5 M PLP was added to the sample (arrow), and fluorescence changes were monitored for an additional 45 min. Solid lines represent best fit of the data to a single exponential decay. The lower panel shows the plots of the residuals for the apo-(1) and holoenzyme (2). The fluorescence intensity at time 0 could not be determined because of the rapid change in em and intensity that occurs within a few seconds of diluting the unfolded proteins in refolding buffer. B, fluorescence emission spectra of native apo-(1) and holoenzyme (2) in 50 mM Tris, pH 7.5, and of an equivalent concentration (100 g/ml) of pmAAT unfolded at pH 2.0 (3). Fluorescence spectra were recorded at 10°C using ex ϭ 280 nm. constant similar to that of the faster step of the biphasic reactivation reaction (k 1 in Table I). The amplitude of this decay represents about 15% of the total decrease in fluorescence observed for the holoenzyme. However, since we cannot estimate the initial fluorescence of this exponential decay due to the long dead time of the manual mixing, its actual amplitude might be considerably higher. The decrease in fluorescence observed during refolding of the holoenzyme in the presence of PLP can also be best fit by a single exponential ( Fig. 2A) with a first-order rate constant close to that found for the slower stage of the reactivation curve (k 2 in Table I). Fitting of the data to a function with two exponential terms results in significant deviations from randomness in the residuals. The low amplitude of the apoenzyme fluorescence decay and its faster rate, together with the long dead time of the experiment, might explain our inability to detect this phase when following refolding of the holoenzyme. The amplitude of the fluorescence change for the holoenzyme is rather large (about 80% of that expected for the binding of PLP to the native apoenzyme) and most likely represents fluorescence quenching by PLP binding. However, the folding event underlying this fluorescence change can occur in the absence of coenzyme. Addition of PLP 45 min after initiation of refolding of the apoenzyme produces a rapid decrease in fluorescence ( Fig. 2A) which is identical to that observed for the reconstitution of native apoenzyme with PLP (data not shown). This is followed by a more gradual decay that parallels the decrease obtained for refolding holoenzyme.
Binding of the coenzyme to refolding pmAAT can also be followed directly by monitoring changes in the visible region of the CD spectrum. Binding of PLP to native mAAT elicits a strong Cotton effect centered at 356 nm (absorption maximum of bound PLP at pH Ն 7.0) due to the asymmetric environment of the coenzyme in the active site (37). The recovery of ellipticity at 356 nm during refolding obeys sigmoidal kinetics and parallels closely the recovery of activity (data not shown). Fitting of the data to the same function used to analyze the reactivation data gives rate constants that are identical, within experimental error, to those determined by following the recovery of activity (Table I). Thus, within the precision of our study, the binding site for PLP and the catalytic activity of pmAAT appear simultaneously during refolding.
The formation of secondary structure was studied by monitoring changes in ellipticity at 222 nm during refolding. As observed for many other proteins (43), a substantial fraction (about 80%) of the signal was regained within the dead time of the experiment (15-20 s) (Fig. 3). The remaining 20% of the signal was recovered much more slowly according to a single exponential with a rate constant of 0.07 min Ϫ1 . The overall recovery of secondary structure is about 60% that expected for a similar concentration of native enzyme which correlates with the yield of recovery of activity obtained at the end of refolding (50 -60% after incubation for 120 min at 10°C). Interestingly, pmAAT refolding from its GdnHCl-unfolded state recovers all of the ellipticity at 222 nm during the dead time of the experiment. The absence of a slow phase under these conditions is probably due to the presence of residual GdnHCl (0.1 M) in the refolding reaction. Indeed, no slow phase was observed for the refolding of acid-denatured protein in refolding buffer supplemented with 80 mM NaCl (Fig. 3) or 0.1 M GdnHCl. On the other hand, the recovery of CD signal at 222 nm by the cytosolic enzyme cAAT during refolding is complete during the dead time of the experiment (33), regardless of the method of unfolding used or the ionic strength of the refolding reaction (data not shown).
Dimerization of Acid-unfolded pmAAT-The dimerization step must be a relatively fast process since it is not ratelimiting for reactivation under the conditions used. To determine where along the folding pathway dimerization occurs relative to the two first-order steps described above, we studied the time course of appearance of pmAAT dimers during refolding. Several methods have been used to follow oligomerization of complex proteins including fluorescence polarization or energy transfer between subunits (44), cross-linking reactions to trap reassembled oligomers, or formation of hybrid heterodimers between chemically modified monomers (45). Some of these methods have been used with AAT for different purposes. Using the cross-linking approach, Leistler et al. (46) found that reactivation of a related AAT from E. coli (eAAT) coincided with the formation of dimers suggesting that inactive dimers do not accumulate during refolding. A method for the formation of heterodimers between native subunits and chemically modified subunits was developed to assess the catalytic independence of the two active sites of AAT from pig heart (40). More recently, a similar approach based on the formation of hybrids between the mature and precursor forms of the mitochondrial isozyme allowed us to follow the pH-dependent monomerization of pmAAT during acid unfolding of the protein (36).
We have now adapted the latter method to study the kinetics of dimerization of pmAAT during refolding in vitro. At different times after the initiation of refolding, equal volumes of separate reaction mixtures containing identical concentrations of either mAAT or pmAAT were mixed and allowed to complete refolding by incubating at 10°C for 120 min. If at the time of mixing there were still free monomers present in the refolding mixture, formation of heterodimers could be expected. As shown earlier, the kinetics of refolding of pmAAT and mAAT are identical (33), and it appears that there is no preference to form homodimers (36). Therefore, the disappearance of the monomers from the refolding reaction (i.e. association into dimers) can be followed by monitoring the decrease in the fraction of heterodimers with time. Since only precursor-containing dimers bind to anionic phospholipid vesicles through the presequence peptide (47), precursor homodimers and heterodimers can be resolved from the mature homodimers by binding to liposomes and centrifugation as indicated under "Experimental Procedures." The relative amount of maturesized subunits in the lipid-bound protein is a direct measure of the amount of heterodimers in the sample, and this in turn depends on the fraction of protein remaining as monomers at each time point. Fig. 4 shows the time dependence of the disappearance of monomers from the refolding reaction esti- mated from the decrease in the fraction of mature subunits associated with vesicles or from the increase in the enzyme activity (mature homodimers) recovered in the supernatant as the refolding reaction progresses. Once dimerization is complete at the time the individual refolding reactions are mixed only precursor and mature homodimers are present and thus no mature subunits pellet with the lipid vesicles (120-min time point in Fig. 4). The disappearance of monomers from the reaction (i.e. dimerization) follows first-order kinetics with a rate constant of about 0.3 min Ϫ1 (Table I), close to the faster of the two slow rate-limiting steps detected by other methods. This shows that the rate-limiting step is an isomerization and not the dimerization. Fig. 4 also shows that formation of the dimer seems to occur well before the recovery of catalytic activity.
Effect of hsp70 on the Spontaneous Refolding of pmAAT and cAAT-The presence of stoichiometric amounts of hsp70 in the refolding buffer markedly decreases the yield of reactivation of pmAAT unfolded at pH 2.0 (Fig. 1A) without affecting the rate of recovery of activity. These results suggest that the percentage of activity recovered in the presence of hsp70 corresponds to the fraction of protein molecules that avoids binding to hsp70 and continues folding. The extent of inhibition of productive folding increases as the concentration of hsp70 is raised. Inhibition is almost complete at a hsp70 to pmAAT monomer ratio of about 4 (Fig. 1B). The mature enzyme, lacking the presequence peptide, exhibits the same behavior indicating that the interaction of pmAAT with hsp70 is independent of the presence of the presequence peptide. hsp70 has no effect on the activity of native (folded) pmAAT (data not shown). The interaction between hsp70 and unfolded pmAAT seems to be specific, since addition of high concentrations (1 mg/ml) of other non-related native proteins, such as carbonic anhydrase, ovalbumin, bovine albumin, malic dehydrogenase, or hemoglobin, did not affect the rate or yield of reactivation of pmAAT (data not shown).
Inhibition of pmAAT refolding by hsp70 is accompanied by the formation of large insoluble pmAAT aggregates. To analyze the nature of these aggregates, refolding reactions containing increasing hsp70 concentrations were fractionated by centrifugation into insoluble pellets (P) and soluble supernatants (S). These pellets and supernatants were analyzed by SDS-PAGE (Fig. 5A). Quantitative analysis of the Coomassie Blue-stained gels showed that increasing the hsp70 concentration results in a progressive disappearance of both proteins from the supernatant and a concomitant recovery in the pellet. At a 4 -5 molar excess of hsp70, all pmAAT is quantitatively found in the pellet (lane 9 in Figs. 5, A and B). At all ratios tested (up to 50-fold molar excess of hsp70), the initial interaction between hsp70 and pmAAT results in the formation of insoluble aggregates that contain both proteins at approximately 1:1 molar ratio (Fig. 5B), suggesting that hsp70 predominantly binds to a single binding site in the transaminase polypeptide chain. In the absence of hsp70, about 90% pmAAT is recovered in the supernatant (Fig. 5A, lanes 1 and 2). Likewise, no hsp70 precipitates in the absence of pmAAT (data not shown).
In contrast to the behavior described above for pmAAT unfolded at pH 2.0, hsp70 is unable to interfere with the reactivation of pmAAT denatured in 4 M GdnHCl. This is most likely due to the presence of residual GdnHCl (0.1 M) in the refolding reaction of GdnHCl-unfolded protein which might interfere with the interaction between the chaperone and its substrate. After incubation for 120 min at 10°C, heterodimers and precursor homodimers were resolved from mature homodimers by binding to anionic liposomes as described under "Experimental Procedures." The time course of dimer formation during refolding was followed by estimating either the amount of mature-size subunit associated with liposomes (q) (i.e. heterodimer content) or the enzymatic activity remaining in the supernatant (Ç) that corresponds to species unable to bind to liposomes, i.e. mature homodimers. Reactivation data are expressed relative to that of a mixture of native enzymes maintained under identical conditions. Continuous lines represent best fit of the data to a single exponential with rate constant of 0.3 min Ϫ1 . Dotted line represents the time course of reactivation of pmAAT (see Fig. 1A) included for comparison. pmAAT (data not shown). Stoichiometric amounts of hsp70 have no effect either on the reactivation of the homologous isozyme cAAT after unfolding at pH 2.0 (Fig. 1A).
pmAAT Intermediates Competent for Binding to hsp70 -To analyze which of the pmAAT folding intermediates detected in the kinetic studies is competent for binding to hsp70, we examined the ability of hsp70 to block reactivation of pmAAT when it was added at different times after initiation of refolding. As shown in Fig. 6, when hsp70 is present in the refolding buffer at the time of dilution of the unfolded protein, the yield of reactivation decreases from 75% to about 20% of the total protein present. If addition of the chaperone is delayed, the fraction of activity recovered increases until it reaches the value obtained in the absence of hsp70 when the chaperone is added 15 min or more after starting the reaction (Fig. 6). These results show that as refolding of the unfolded protein progresses, it loses its ability to bind to hsp70. The time course of this process can be described by a single exponential with a rate constant of about 0.3 min Ϫ1 , close to the faster of the two kinetically resolved steps detected by following either activity, intrinsic fluorescence or the assembly of dimers. It appears that pmAAT loses its ability to bind hsp70 after it has undergone the first slow isomerization to the assembly-competent monomer.

DISCUSSION
One of the many questions that remain open regarding the mechanism of action of molecular chaperones concerns the factors that determine the binding of a chaperone to a particular protein substrate. The selectivity of E. coli SecB for its ligands has been proposed to be determined, at least in part, by the rate at which the protein folds (48). Such a kinetic partitioning, however, seems to be less important than the stability of intermediates in the folding pathway in determining the affinity of a polypeptide for binding to GroEL (49). On the other hand, hsp70 is known to favor binding of highly unstructured regions of unfolded polypeptides (50) and to be able to bind short peptides containing at least seven residues in an extended conformation (17,21). hsp70 also appears to show preference for binding peptides containing predominantly hydrophobic residues (20,21) that would usually be buried in the interior of a folded protein. This is consistent with the ability of these chaperones to discriminate between folded and unfolded structures (15,51). Thus, hsp70 could discriminate between potential substrates according to their rate of folding to an intermediate it can no longer recognize.
We have previously shown that only the mitochondrial form of AAT binds to cytosolic hsp70 early during translation in a cell-free extract; no association of cAAT with hsp70 could be detected (29). However, given the considerable dead time of the analytical method used in these studies (co-immunoprecipitation of translation product with hsp70 antibodies) and the rapid folding of the cAAT translation product (28), the possibility of a short-lived transient interaction of hsp70 with cAAT at the early stages of protein synthesis could not be discarded. Furthermore, due to the complex nature of the cellular extract (rabbit reticulocyte lysate) in which these experiments were carried out, it is difficult to reach a definitive conclusion about the lack of affinity of hsp70 for cAAT. In addition to hsp70, a number of other components have been found to associate with nascent chains in the cytosol of eukaryotes (13). The interaction of cAAT with some of these cytosolic chaperones might facilitate or even accelerate the folding of the chain thereby precluding binding to hsp70. It was therefore of interest to examine the ability of hsp70 to recognize the isozymes of AAT under well defined in vitro conditions using chemically unfolded polypeptides and purified cytosolic hsp70.
We show now that indeed hsp70 arrests refolding of mAAT that had been unfolded at pH 2.0 but not cAAT. The reasons for this selectivity are not clear. The affinity of mAAT and cAAT for binding to the chaperonin GroEL parallels the behavior described here with hsp70. GroEL completely stops the refolding of mAAT but only slows down the reaction for cAAT (52). It is interesting to point out that whereas a substantial fraction of GdnHCl-unfolded cAAT refolds at physiological temperature without assistance, no mAAT refolds unless GroEL/GroES are present (52). It appears that these two isozymes have evolved to display remarkably different behavior with regard to both spontaneous and assisted folding. A correlation between the overall hydrophobicity of the protein and the strength of its interaction with GroEL was proposed to explain similar findings for the cytosolic and mitochondrial isozymes of malate dehydrogenase (53). However, mAAT and cAAT contain about equal amounts of hydrophobic amino acids very similarly distributed along their primary structure (28). Thus this feature alone cannot explain the different affinities of these enzymes for at least two types of chaperones.
The folding mechanism of each isozyme could provide an alternative explanation for these differences. In a previous report, we proposed a minimum folding pathway for the renaturation of mAAT from its GdnHCl-unfolded state (33). However, hsp70 did not bind to mAAT refolding under these conditions. The most likely explanation for this is that the residual 0.1 M GdnHCl carried over during dilution of the unfolded protein in refolding buffer interferes with binding to hsp70. Similar concentrations of GdnHCl have been found to affect the activity of the chaperonin GroEL (54). For this reason, in the present study we chose to use protein unfolded at pH 2.0 as the starting material for the analysis of the refolding reaction. Although acidification at this pH results in almost instantaneous loss of activity, the unfolded protein contains detectable amounts of residual secondary structure indicating that unfolding may not be as extensive as in the presence of high concentrations of GdnHCl (36). Nevertheless, the acid-unfolded protein refolded with even higher efficiency than from the GdnHCl-unfolded state and with comparable kinetics, at least with regard to the slow kinetic phases resolved with the manual mixing method used in this study (dead time of about 15-20 s).
The results described in this paper are consistent with the presence of several kinetic phases in the refolding pathway of pmAAT. The initial step is the rapid collapse of the unfolded polypeptide (U) to an intermediate (MЈ) containing a substan- tial amount (about 80%) of the secondary structure of the native dimer. This appears to be a common feature in the refolding of many proteins. Stopped-flow CD studies in the far-UV have shown that formation of elements of secondary structure can occur within the low millisecond time scale after initiation of refolding (55,56). This structure is presumably rapidly fluctuating and may not correspond precisely to the native secondary structure of the protein (56). The origin of the subsequent slow phase in the recovery of the CD 222 nm signal which accounts for the remaining 20% of the total change is obscure. This slow component is not observed during refolding from GdnHCl or upon increasing the ionic strength of refolding reactions of acid-unfolded protein (Fig. 3). This effect could be traced to the anion component of the neutral salts. Due to the limited information available we can only speculate that perhaps this effect is caused by the shielding of positive charges by anion binding which would favor the establishment of hydrophobic interactions and therefore the acquisition of structure. Furthermore, the rapid blue-shift in em and increase in fluorescence intensity that takes place during the dead time of manual mixing are consistent with the switch of aromatic residues to a more hydrophobic environment in MЈ. This intermediate MЈ therefore displays properties that resemble those ascribed to the molten globule state (42,57). Hence, although we have no information on the early steps of the refolding process, a fast conversion of the unfolded species U to an intermediate or population of intermediates MЈ must be included in the reaction scheme to accommodate the above observations.
The sigmoidal reactivation kinetics observed for the apoenzyme and holoenzyme forms of pmAAT may be adequately described by a mechanism involving at least two slow ratelimiting steps in the conversion of MЈ to the native dimer N. These rate-limiting steps appear to be first-order because the kinetic parameters are independent of the protein concentration. Apparently dimerization is not rate-limiting, and therefore, a mechanism of the uni-bimolecular type can be discarded. A similar sigmoidal kinetics was observed for the acquisition by the folding protein of the ability to bind PLP at its active site (CD 356 nm in Table I). Hence the binding site for PLP and catalytic activity appear simultaneously during refolding. These unimolecular steps probably represent isomerizations of folding intermediates along the folding pathway. The nature of the conformational rearrangements associated with the faster step, characterized by k 1 (0.4 min Ϫ1 ), is not clear. Fluorescence-detected refolding of the apoenzyme occurs in a single exponential phase with a rate constant similar to k 1 for reactivation and a small amplitude. This phase involves the quenching of fluorescence and could therefore involve the formation of tertiary contacts during the packing of the tertiary structure of the protein. The fluorescence quenching might result from the appearance of other aromatic residues (aromatic interactions) or ionized side chains in the vicinity of the aromatic fluorophores. These conformational changes may be part of a general hydrophobic collapse since previous studies with the GdnHCl-unfolded protein showed that the conversion of MЈ to M is characterized by a complete loss of the affinity for 1-anilino-8-naphthalene sulfonate suggesting that M has no significant hydrophobic surface exposed (33). The slower kinetic step characterized by k 2 (0.038 min Ϫ1 ) is associated with a fluorescence decrease observed only in the presence of the coenzyme PLP. It probably represents the quenching of the fluorescence of Trp-140 at the active site upon binding of PLP. This interpretation is supported by the almost instantaneous decrease in fluorescence observed when PLP is added to a sample of apoenzyme refolded in the absence of coenzyme ( Fig.   2A). As reported earlier (33), the magnitude of this decay increases with the time of addition of PLP during refolding which indicates that the formation of a functional active site occurs in the absence of PLP, but it does not translate in a change in fluorescence unless PLP binds. Thus, the slower phase of the sigmoidal reactivation most likely represents the formation of a functional active site able to bind PLP.
First order rate-limiting steps comparable to those delineated above had been previously detected for the refolding of pmAAT from GdnHCl (33). The present paper also provides information on the dimerization step. Denaturation at pH 2.0 causes dissociation of the native dimer and unfolding of the free subunits (36), and therefore, during refolding the monomers must reassociate to give the native dimer. As shown in Fig. 4, the disappearance of monomers from the reaction seems to obey first-order kinetics. Thus, an isomerization and not dimerization is rate-limiting. The rate constant of dimer formation (0.3 min Ϫ1 ) is close to the faster of the two rate-limiting steps of the reactivation kinetics. This shows that, even though dimerization itself may be fast, the relatively slow isomerization step characterized by k 1 precedes and limits dimerization during renaturation of pmAAT. On the other hand, the results presented in Fig. 4 clearly show that the dimerization step occurs before the appearance of catalytic activity. Thus, in the refolding of pmAAT, dimerization indeed occurs before the rate-limiting isomerization responsible for the recovery of the functional properties of the enzyme (k 2 ). The experimental evidence presented in this paper also supports the conclusion that productive binding of PLP represents the last step in the reactivation process. We cannot, however, discard completely the possibility that PLP might bind to an earlier monomeric intermediate but remain spectroscopically silent until the native architecture of the active site is acquired. The various steps of the refolding pathway of pmAAT that have been identified to date can be summarized by the following tentative scheme (Reaction 2). A very rapid folding of each monomer (U 3 MЈ) as evidenced by the acquisition of secondary structure and the burying of aromatic residues is followed by a slow isomerization (MЈ 3 M) that precedes the formation of the dimer. Then, a rapid dimerization (2M 3 DЈ) is followed by an isomerization of the already assembled dimer (DЈ 3 D) and binding of PLP (N active ). This refolding pathway includes a partially folded, non-active dimer that subsequently undergoes a final isomerization leading to the formation of a functional active site able to accept the coenzyme PLP and thereby become active. Apparently after the assembly step, each subunit undergoes a conformational change brought about by their association. Examination of the three-dimensional structure of mAAT (60) reveals that although the active site of each subunit is constructed essentially from peptide segments of irregular structure, it is a well-ordered region of the protein. Two of the critical residues found at the active site are contributed by the other subunit, the substrate binding Arg-292 and the cofactor binding residue Tyr-70 which forms a hydrogen bond with the phosphate group of bound PLP. In addition, Lys-258, to which the cofactor PLP is covalently attached, is part of the subunit interface (60). Therefore, it is not surprising that conformational changes mediated by intersubunit interactions are required to position properly the residues comprising the active site.
Formation of partially folded dimers has also been observed for the refolding of another dimeric PLP-dependent enzyme, the ␤ 2 subunit of tryptophan synthetase from E. coli (44). On the other hand, earlier reports on the refolding of AAT from E. coli (eAAT) implied that dimerization of that enzyme occurs by association of fully folded monomers without accumulation of partially unfolded inactive dimers (46). We should point out that pmAAT and eAAT also differ in their unfolding processes. Equilibrium unfolding studies of eAAT showed that the first transition at low concentrations of GdnHCl involved the formation of an inactive but compact monomeric intermediate (58). No such compact monomeric state could be detected in unfolding studies of mAAT. At low GdnHCl concentrations we observed instead the formation of a partially unfolded dimeric equilibrium intermediate (59). It appears that, despite the similar overall shape of the subunit and dimer structures of pmAAT (60) and eAAT (61), the relative contribution of the quaternary (intersubunit) interaction to the thermodynamic stability of the native protein is quite different for these two homologous proteins. eAAT seems to be stabilized predominantly by secondary and tertiary interactions at the monomer level, whereas quaternary interactions may instead contribute to a greater extent to the stabilization of pmAAT. In fact, about 18.9% of the total area of the two subunits is buried at the subunit interface of mAAT (60). This value is higher than the averages (about 15%) reported for dimeric proteins (62) and is consistent with the high stability of the dimeric structure of mAAT (63,64).
How many of the folding intermediates included in Reaction 2 bind to hsp70? The ability of the unfolded protein to bind to hsp70 is lost as refolding progresses in a process that is monophasic and has a rate constant similar to that for the isomerization of MЈ to the assembly-competent M. We can therefore conclude that hsp70 can bind to folding intermediates MЈ corresponding to a collapsed state, although it is not clear whether binding to hsp70 requires the conversion of U to MЈ or the chaperone is able to bind to intermediates preceding MЈ in the kinetic pathway. The transformation of MЈ to M, which rapidly dimerizes, precludes binding of hsp70 to pmAAT. Given the lack of information on the kinetics of the dimerization reaction itself, it is not clear whether the loss of binding ability is due to the formation of tertiary interactions characteristic of the MЈ to M isomerization or to the hiding of potential binding sites in M as a result of its association into homodimers.
In any case, hsp70 seems to bind only a population of folding intermediates of pmAAT lacking extensive tertiary interactions. This agrees with the apparent preference of hsp70 for binding to substrates, even when they are large proteins, in an extensively unfolded state (50) and may explain its inability to arrest refolding of the homologous cAAT. The single rate-limiting step detected for the reactivation of cAAT (k ϭ 0.2 min Ϫ1 ) most likely corresponds to the conversion between intermediates analogous to DЈ and D in Reaction 2 (k 2 ) for pmAAT since it roughly coincides with the PLP-dependent decay in intrinsic fluorescence (k ϭ 0.15 min Ϫ1 ). Assuming that the refolding pathway of cAAT includes an isomerization step similar to that proposed between monomeric states MЈ and M in Equation 4 and that the relative values of k 1 and k 2 are comparable, k 1 for the refolding of cAAT would be in the order of 1-2 min Ϫ1 , too fast to be resolved kinetically with the methods used in this study. This isomerization to an intermediate no longer recognized by hsp70 might contribute to the lack of interaction between hsp70 and cAAT. Yet, a more thorough characterization of the refolding pathway of cAAT using rapid mixing techniques is needed to properly answer this question.
The proposed interaction of hsp70 with largely unfolded fold-ing intermediates also agrees with a previous observation that binding of hsp70 to nascent pmAAT chains synthesized in cell-free extracts seems to be transient as the amount of hsp70 found associated with the nascent polypeptide decreases rapidly as translation and folding progresses (29). Other molecular chaperones may take over and continue the job initiated by hsp70 in preventing the premature folding and/or aggregation of the newly synthesized polypeptide until it can be delivered to the import machinery on mitochondria. This scenario may also help explain the extensive aggregation of the pmAAT⅐hsp70 complex observed during refolding of pmAAT in the presence of hsp70 in vitro. It has been proposed that hsp70 can prevent aggregation of heat-denatured or damaged proteins (22) as well as of newly synthesized proteins (7) probably by binding to transiently exposed hydrophobic regions. However, hsp70 appears to act as an anti-chaperone, promoting aggregation of refolding pmAAT. A possible explanation for this behavior is that, under in vitro conditions and in the presence of hsp70 alone, binding of pmAAT folding intermediate(s) to hsp70 may interfere with subsequent folding steps thus leaving homologous hydrophobic surfaces exposed to the solvent long enough to allow the formation of self-aggregates. By contrast, during protein translation in the cell, the vectorial elongation of the polypeptide chain on the ribosome may allow the binding of additional hsp70 molecules or, alternatively, the association of other molecular chaperones present in the cytosol (13). A similar anti-chaperone activity has been described for BiP, an endoplasmic reticulum hsp70, with refolding lysozyme (65). This anti-chaperone activity of BiP has been proposed to play a role in retaining unfolded or misfolded proteins in the endoplasmic reticulum. The fact that these hsp70 proteins bind denatured proteins but do not promote correct folding even in the presence of ATP suggests that the role of these chaperones, at least with certain substrates, might be the stabilization of the unfolded structure of newly synthesized proteins rather than the assistance in productive folding. The irreversible aggregation step is probably not a physiological event but rather the unfortunate consequence of an incomplete experimental system. Additional components of the cytosol (or the lumen of the endoplasmic reticulum in the case of BiP) may be required to replicate the direct involvement of hsp70 in protein folding or delivery to a membrane system such as mitochondria.
It is difficult to reproduce in the test tube the action of molecular chaperones mediating protein folding of large proteins in vivo. Among these difficulties are the simultaneous display of multiple binding sites in the protein substrate during refolding in vitro as opposed to the gradual elongation of the polypeptide chain during synthesis in the cell, and the absence of the complete set of cytosolic components that may collaborate in the process. Nevertheless, studies in vitro allow us to address under well defined conditions issues that are difficult to analyze using more complex experimental systems such as crude cellular extracts. The nature of the factors responsible for the recognition by molecular chaperones of large proteins as unfolded substrates is one of these pending questions. This work represents an attempt at improving our understanding of the basis for the selectivity that hsp70 displays toward only one of the two AAT isozymes. Although the different kinetics of refolding of the two isozymes may very well represent one of the factors determining this selectivity, we cannot exclude that there are linear or three-dimensional motifs in folding intermediates of pmAAT that are recognized by hsp70 and are absent in cAAT. A detailed exploration of the presence of sequential binding signals for hsp70 within the pmAAT polypeptide is in progress. the statistical analyses and John Bollin for the purification of cAAT and pmAAT.