A Pathway for Conformational Diversity in Proteins Mediated by Intramolecular Chaperones*

Conformational diversity within unique amino acid sequences is observed in diseases like scrapie and Alzheimer’s disease. The molecular basis of such diversity is unknown. Similar phenomena occur in subtilisin, a serine protease homologous with eukaryotic pro-hormone convertases. The subtilisin propeptide functions as an intramolecular chaperone (IMC) that imparts steric information during folding but is not required for enzymatic activity. Point mutations within IMCs alter folding, resulting in structural conformers that specifically interact with their cognate IMCs in a process termed “protein memory.” Here, we show a mechanism that mediates conformational diversity in subtilisin. During maturation, while the IMC is autocleaved and subsequently degraded by the active site of subtilisin, enzymatic properties of this site differ significantly before and after cleavage. Although subtilisin folded by Ile−48 → Thr IMC (IMCI-48T) acquires an “altered” enzymatically active conformation (SubI-48T) significantly different from wild-type subtilisin (SubWT), both precursors undergo autocleavage at similar rates. IMC cleavage initiates conformational changes during which the IMC continues its chaperoning function subsequent to its cleavage from subtilisin. Structural imprinting resulting in conformational diversity originates during this reorganization stage and is a late folding event catalyzed by autocleavage of the IMC.

Subtilisins constitute a family of serine proteases that have served as model systems for understanding the structural origin of protease function and specificity (1). Cloning of the gene for subtilisin E from Bacillus subtilis revealed that it is synthesized with an N-terminal precursor domain that is referred to as the propeptide (2). Although several biochemical studies on subtilisin have indicated that the 77-mer propeptide was not part of the 275-amino acid active protease (3,4), Escherichia coli expression experiments revealed that the propeptide of subtilisin was essential for proper activation and secretion of the protease in vivo (5)(6)(7). Soon thereafter, the propeptide of ␣-lytic protease (a serine protease not evolutionarily related to subtilisin) was also shown to be essential for proper activation of its protease domain (8). Although these results were surprising at the time, propeptides are now known to be required for production of a large number of proteases. Propeptidedependent systems have since then been identified both in prokaryotes and eukaryotes (9). Virtually all known extracellular bacterial proteases have propeptides (10). There are now examples of serine (5,11), aspartyl (12), cysteine (13), and metalloproteases (14) synthesized as precursor pro-proteases. Propeptides seem to perform two distinct functions: folding and inhibition of their protease domains (15). In doing so they act as single turnover enzymes due to their tight binding with folded proteases and their proteolytic sensitivity. As propeptides facilitate correct folding they were classified as molecular chaperones (16), but due to differences that exist between molecular chaperones, propeptides were also termed "intramolecular chaperones" (17). Although many proteases require their propeptides to facilitate their folding, these findings do not constitute a rule. In addition to proteases, proteins such as growth factors, neuropeptides, hormones, and plasma proteins also require their propeptides for correct folding (18 -21). Studies on the conformations of propeptides of subtilisin (22)(23)(24) and carboxypeptidase Y (25) have shown them to be randomly structured. In the vicinity of the prefolded protease domain, the propeptide of subtilisin adopts an ␣-␤ conformation (22)(23)(24).
Anfinsen demonstrated that all the information necessary for folding of a protein into an active conformation resides in the amino acid sequence of that protein (26). Based on Anfinsen's observation, the native state of a protein is generally believed to be the global free energy minimum (27). However, there is increasing evidence that kinetically selected states play a role in the biological function of some proteins (28). For example, the serpin plasminogen activator inhibitor 1 folds into an active structure and then converts slowly to a more stable, but low activity "latent" conformation (29). Thus, the folding of plasminogen activator inhibitor 1 is apparently under kinetic control. When ␣-lytic protease (11,30) and subtilisin (31,32) are folded in the absence of their propeptides, they fold into partially structured molten globule intermediates. These stable but inactive intermediates can convert into active conformations upon addition of their propeptides (30,31) and suggest that propeptides promote folding of their proteases by direct stabilization of the rate-limiting folding transition state (30,31). Propeptides seem to be essential only during late stages of the folding process because they help overcome a kinetic block in the folding pathway. A similar state is formed even in the presence of the propeptide (33), and autoprocessing may occur in this intermediate state.
Point mutations within the propeptides can affect their folding function (34,35). As a result, identical protein sequences can display altered, enzymatically active conformations with specificity toward the propeptide that mediates folding (36). This finding helped introduce the concept termed "protein memory" and suggests that propeptide domains function as steric chaperones (28,36). Altered protein folding also occurs in prion diseases wherein the protein can exist in two different conformations with invariant amino acid sequences, although this folding does not require propeptides (37). Therefore, proteins can display conformationally diverse biologically active states, and the native state of proteins may not necessarily reside at a thermodynamic minimum but may be ones that are most easily accessible (28,36).
In the present manuscript we describe a mechanism that mediates conformational diversity in subtilisin. Our results show that the propeptide continues its chaperoning function subsequent to its cleavage from subtilisin and that structural imprinting that results in conformational diversity originates during this reorganization stage.

MATERIALS AND METHODS
Site-directed Mutagenesis-The desired mutations within the propeptide will be introduced through site-directed mutagenesis using polymerase chain reaction (38). After site-directed mutagenesis the plasmids were sequenced using an automated sequencer (Applied Biosystems, Inc., model 310).
Protein Expression and Purification-The plasmid (pET11a) carrying mutant and wild-type proteins were expressed in E. coli strain BL21(DE) that is grown on M9 medium. Propeptides were isolated from inclusion bodies by solubilizing in 15-25 ml of 6 M guanidine HCl (39). After overnight incubation at 4°C, the insoluble materials were removed through centrifugation. The supernatants were then be dialyzed against 50 mM sodium phosphate buffer at pH 5.0. Purification was carried out using a cation-ion exchange column (CM-Sephadex-50) with a linear gradient of NaCl (0 -0.4 M) followed by high pressure liquid chromatography (C18-reverse phase) with a linear acetonitrile gradient (0 -65%). The propeptides elute in 43% acetonitrile and are lyophilized and then resuspended into required buffers for use in different experiments (39).
Estimation of Autoprocessing Activity-The autocleavage of the propeptide occurs under conditions in which the refolding buffer contains 1.0 M urea. Therefore, for estimating autocleavage efficiency as a function of pH, the precursors were dialyzed in a step-wise manner to a final concentration of 1.5 M urea. Proteins were dialyzed extensively, and no autocleavage occurs under these conditions. Aliquots of the protein were further dialyzed against refolding buffers at different pHs that contained no urea. After 24 h of dialysis, the proteins were subjected to SDS-polyacrylamide gel electrophoresis, and the extent of autocleavage was estimated densitometrically (40).
Estimation of Enzymatic Activity-An aliquot of the sample was incubated at 25°C in 200 l of 50 mM buffer at different pH levels, containing 1 mM CaCl 2 and 0.13 mM succinyl-Ala-Ala-Pro-Phe-p-nitroanilide (41). Enzymatic activity of subtilisin at different pH levels was estimated by monitoring the release of p-nitroaniline through changes in absorbance at 405 nm. Reading were collected at 15 s using a Bio-Rad UV microplate reader, and the velocity of reaction was estimated (36).
Slow Binding Kinetics-The propeptide of subtilisin behaves like a slow binding inhibitor. Competitive slow binding inhibition is usually described using two mechanisms (41).
Both mechanisms can be described by the same equation, where v represent enzymatic rates at time t, and v o and v s represent initial and steady state enzymatic rates, respectively.
The apparent rate constant k for a first order reaction has a different significance for a one-step process and is shown by Equation 4.
However, for a two-step process the rate constant is depicted using Equation 5.
Because the initial velocity (v o ) obtained from the on-set progress curves is independent of inhibitor concentration [I] for subtilisin E (35,36), the results confirm that inhibition of subtilisin by its propeptide follows Equation 1.
However, if the initial velocity (v o ) is a function of the inhibitor concentration [I], then the inhibition occurs by a two-step process (35) described by Equation 2.
Integration of Equation 1 at a constant substrate concentration [S] results in the following equation.
where A 0 and A depict the absorbance at 415 nm at time 0 and t, respectively. Hence v s and k can be graphically fitted using a nonlinear least square fitting of the absorbance A versus t.
Because the propeptide is also a substrate for the reaction, measurement of the inhibition constants is more complex. Therefore, the inhibition experiments were carried out under pseudo-first-order kinetics where the lowest propeptide concentrations were greater than 10-fold excess. Enzyme concentrations were set at suitably low levels to give measurable rates of substrate hydrolysis, with observable state of inhibitor binding over the steady state time scale. The propeptide concentrations vary between 1 and 10 M. Initial substrate concentration was 3 mM, and less than 10% of the substrate was degraded throughout the experiment. Reactions were initiated by addition of Sub WT or Sub I-48T. After thorough mixing, substrate cleavage was monitored by recording at 415 nm the release of p-nitroaniline with time (36,42). Data from each curve were fitted to Equation 8, and the K i is estimated by fitting a set of inhibitor concentrations Circular Dichroism Studies-CD measurements are performed on an automated AVIV 60DS spectrophotometer fitted with a thermostated cell holder that is controlled by an on-line temperature control unit. Quartz rectangular cells (Precision Cells, Hicksville, NY) with a path length of 1 mm are used. For CD studies, modified subtilisin in 50 mM Tris-HCl, pH 7.0, containing 0.5 M (NH 2 ) 4 SO 4 and 1 mM CaCl 2 will be used. Solutions are filtered through a 0.22-m filter before measurement. Scans were carried out at wavelengths between 260 and 190 nm in a cuvette with a 1-mm path length maintained at 10°C. In the thermal unfolding experiments, temperatures were increased from 10 to 90°C in 1°C intervals, with a 30-s equilibration at each temperature. Data were collected at each temperature for 10 s. Protein solutions contain 1 mM phenylmethylsulfonyl fluoride to prevent autolysis during this procedure (36).

RESULTS
The in vitro maturation of the 352-residue IMC-subtilisin E involves folding, autocleavage, and degradation of its IMC domain (77-residues) to give mature subtilisin (275-residue). The folding of pro-subtilisin occurs through a molten globule like intermediate that becomes compact during or after the autocleavage of the propeptide domain.
Refolding of Denatured Pro-subtilisin-Denatured precursors are refolded through a step-wise dialysis procedure that gradually reduces urea concentration (see "Materials and Methods"). Fig. 1a depicts CD spectra of denatured IMC WT -S221C-subtilisin. Extensive dialysis (48 h) against buffer that contains 1.5 M urea, pH 7.0, induces significant secondary structure within the uncleaved precursor (Fig. 1a). This precursor binds with ANS (Fig. 1b), a hydrophobic dye that fluoresces upon interacting with hydrophobic patches within proteins. Complete removal of urea through dialysis induces the subtilisin domain to cleave its IMC. The resulting autocleaved complex (Fig. 1c) was purified using gel filtration and displays a secondary structure pattern similar to native subtilisin and fluoresces approximately 8-fold less than the uncleaved precursor (Fig. 1b). Differences in ANS fluorescence intensities suggests that a structural reorganization that alters solvent accessible hydrophobic surfaces within the protein coincides with autocleavage.
Characterization of Autoprocessing and Degradation Activity of Pro-Ser 221 Cys-subtilisin-A Ser 221 3 Cys substitution at the active site of subtilisin blocks maturation after autocleavage, resulting in a stable stoichiometric IMC-S221C-subtilisin complex (43), whose x-ray structure was recently solved at 2 Å resolution (44,45). Although active site residues Asp 32 , His 64 , and Ser 221 within subtilisin mediate both autocleavage and degradation (5)(6)(7)(8), the Ser 221 3 Cys substitution completely abolishes degradation while allowing efficient autocleavage of its IMC domain. Fig. 1d displays both the autocatalytic cleavage of IMC WT -S221C-subtilisin and proteolytic activity of wild-type subtilisin as functions of pH. The procedure used to estimate the autoprocessing efficiencies is described under "Materials and Methods." Maximum autocleavage occurs at pH 7.0, whereas optimum proteolytic activity is observed at pH 8.5. Although IMC WT -subtilisin also displays an optimum pH of 7.0 (data not shown), IMC WT -S221C-subtilisin is used because it allows an accurate estimation of autocleavage in the absence of any degradation of the cleaved IMC. Substitutions at position 221 by Cys or Ala does not alter the K m of subtilisin (46,47). Because S221C-subtilisin only autocleaves but does not degrade its IMC domain, these results suggest that although mediated via the active site of subtilisin, autocleavage and degradation activities of this catalytic triad differ significantly before and after cleavage. Differences in optimum pH substantiate that autocleavage and degradation activities are different from each other.
Mapping the Effect of the Propeptide Mutation on the Maturation Pathway-IMC I-48V -subtilisin affects the folding process, and although the precursor undergoes maturation, the resulting conformation of the protease domain (Sub I-48V ) differs from that Sub WT . Because autocleavage and degradation represent two activities of the same active site, effects of IMC mutations on these activities may help elucidate the molecular basis of altered folding. Sub I-48V retains significant enzymatic activity, and hence the mutation Ile Ϫ48 3 Thr was isolated in an attempt to further reduce proteolytic activity. Crystallographic studies of the IMC-subtilisin complex (44,45) shows that Ile Ϫ48 is surrounded by hydrophobic side chains (Fig. 2a). Theronine, a polar residue, may disrupt the hydrophobic core and augment the phenomenon of altered folding. Upon renaturation, IMC I-48T -S221C-subtilisin cleaves its IMC domain (Fig. 2b) to form a complex that is further purified using gel filtration. Similarly, IMC I-48T -subtilisin undergoes maturation forming mature Sub I-48T (Fig. 3b) with a well defined secondary structure (Fig. 2c). The purified autocleaved complexes IMC WT and IMC I-48T -S221C-subtilisin interact differently with ANS, with IMC I-48T -S221C-subtilisin exhibiting a 4-fold greater fluorescence intensity than IMC WT -S221C-subtilisin complex. Similarly, Sub I-48T fluoresces with an magnitude approximately 2-fold of that of Sub WT (Fig. 2d). These results suggest that subtilisin domains folded by the IMC I-48T display exposed hydrophobic surfaces that are absent in domains folded using IMC WT . Far ultraviolet CD spectra indicate that Sub I-48T and Sub WT are well folded and have similar but not identical conformations (Fig. 2c). The pH profile of the Sub I-48T is depicted in Fig. 1d, and it displays maximum proteolytic activity at approximately pH 8.2, whereas autoprocessing is maximum at pH 7.0.
Thermostability of Autoprocessed Complexes-Thermosta- 1. a, folding of IMC WT -S221C-subtilisin denatured using 6 M guanidine HCl (green line). The precursor acquires significant secondary structure after extensive dialysis against 1.5 M urea (magenta line). Upon complete removal of urea, the precursor begins to cleave its IMC domain. Completely autocleaved complex (blue line) purified using gel filtration (43) shows a well defined secondary structure similar to mature subtilisin (black line). b, fluorescence due to ANS binding to the protein. IMC WT -S221C-subtilisin in 5 M urea does not bind ANS (green line). The autocleaved complex (blue line) binds to ANS approximately 2-fold greater than mature subtilisin (black), whereas the protein in 1.5 M urea (magenta) binds to ANS by a factor of 8-fold greater than the autocleaved complex (blue line). c, lefthand lane represents IMC WT -S221C-subtilisin in 5 M urea. The protein was dialyzed using a step-wise gradient. Extensive dialysis against 1.5 M urea (middle lane) does not result in autocleavage. After 12 h of dialysis against refolding buffer without urea, the precursor undergoes autocleavage to give IMC WT -S221C-subtilisin complex (righthand lane). d, activity as a function of pH. Open circles (autocleavage efficiency) and filled circles (proteolytic activity) represent subtilisin folded using IMC WT , whereas filled triangles (enzymatic activity) represent subtilisin folded using IMC I-48T . bilities of mature protease domains Sub WT and Sub I-48T and the corresponding autocleaved complexes were determined by monitoring changes in negative ellipticity at 222 nm (Fig. 3a). The autoprocessed complexes used in these experiments were separated from the unautoprocessed precursors using gel filtration chromatography as described earlier (41). Sub I-48T (T m 49.5°C) was found to be substantially more unstable than Sub WT (T m 58.5°C), whereas the IMC I-48T -S221C-subtilisin autocleaved complex (T m 46.8°C) was also more unstable than the IMC WT -S221C-subtilisin complex (T m 53.2°C). Both protease domains complexed with the IMC domains melt at lower temperatures than their corresponding protease domains. This suggests that the presence of IMC domains destabilize the corresponding complexes and conversely that the degradation of the IMC domains enhances stability of cognate protease domains. It is important to note that phenylmethylsulfonyl fluoride covalently interacts with the active site Ser 221 residue. Crystallographic results have shown that such covalent linkage may cause local perturbations but does not seem to affect the rest of the backbone. The introduction of an bulky residue into the active site may therefore slightly destabilize the mature protein, causing it to lower the T m of the mature protease. Therefore the difference between the T m of subtilisin and the complex with its propeptide may actually be greater.
Characterization of the Binding of Propeptide with Mature Subtilisin- Fig. 3b describes the IMC I-48T and IMC WT binding with Sub WT and Sub I-48T . From these curves it is evident that IMC I-48T binds to Sub I-48T more strongly (and is degraded more quickly) than with Sub WT , suggesting more specific recognition of Sub I-48T than Sub WT by IMC I-48T . Rates of autocleavage of the IMC I-48T and IMC WT -S221C-subtilisin precursors are estimated as described under "Materials and Methods." Relative rates of autocleavage of the two precursor proteins appear similar (Fig. 3c), whereas proteolytic activities of Sub WT and Sub I-48T toward a peptide substrate N-succinyl-A-A-P-F-p-nitroanilide differ significantly (Fig. 3d). Enzymatic activities of Sub WT and Sub I-48T were estimated as described earlier (12, 26), and Sub I-48T displays a K m of 0.24 ϫ 10 Ϫ3 M, whereas Sub WT displays a K m of 2.0 ϫ 10 Ϫ3 M. Because autocleavage occurs at similar rates whereas the proteolytic specific activity of Sub I-48T is only 9.0% of that of Sub WT , the results indicate that Ile Ϫ48 3 Thr substitution seems to selectively affect proteolytic activity (final conformation) but not autocleavage of the maturation intermediate.

DISCUSSION
Based on the above results a mechanism that leads to altered folding is proposed (Fig. 4). It is unclear how the IMC initiates this folding process, but it has been speculated that two helices between residues 100 and 144 within the protease domain may be involved (44,45). The IMC domain facilitates the precursors adopt an intermediate state that is capable of autocleavage, a process mediated by the active site of subtilisin (5)(6)(7)(8). The properties of this active site differ from that of the mature protease because (i) it displays an optimum pH that is 1.5 units less than the mature protein; (ii) IMC-S221C-subtilisin can only autocleave its IMC but cannot degrade it to release mature S221C-subtilisin; and (iii) both IMC I-48T and IMC WT uncleaved precursors bind strongly with ANS, whereas the autocleaved complexes do not. This is consistent with structural reorgani- zation that occurs after proteolytic cleavage (33). IMC I-48T and IMC WT S221C-subtilisin both cleave their IMC domains at similar rates, indicating that the Ile Ϫ48 3 Thr substitution has little effect on autocleavage. Because Sub WT and Sub I-48T differ in their enzymatic activity after cleavage and subsequent degradation, we conclude that altered folding and protein memory that result from IMC-mutations occurs after autocleavage, after which the IMC continues to functions as a chaperone. Therefore, autocleavage does not imply that the folding process has been completed and may represent the transition state of the folding reaction. Autocleavage and degradation are closely coupled in wild-type subtilisin, and occurrence of these two processes renders the folding process irreversible.
Although proteolytic activity of subtilisin at pH 8.5 is approximately 2-fold greater than pH 7.0 (Fig. 1d), incubation of the IMC WT -S221C-subtilisin complex at pH 8.5 does not facilitate degradation of the IMC domain (data not shown). Changes in the optimum pH alone cannot explain the absence of proteolytic activity toward the noncovalently bound IMC domain. The protease domain in the autocleaved complex is almost identical FIG. 3. a, thermal unfolding of precursor and mature proteins monitored using changes in negative ellipticity at 222 nm. Unfolding spectra of Sub WT (black line) and Sub I-48T (cyan line) displays higher melting temperatures (T m ) than autocleaved precursors IMC WT -S221C-subtilisin (blue line) and IMC I-48T -S221Csubtilisin (red line), respectively. b, comparison of the inhibition profiles of Sub WT and Sub I-48T by IMC WT and IMC I-48T . Concentrations of Sub WT and Sub I-48T were adjusted such that they degrade the substrates at similar velocities. Substrate concentration was maintained at 4 mM, whereas the IMC WT and IMC I-48T were maintained at 6.0 and 25 M, respectively. Sub WT and Sub I-48T alone are shown in black, Sub I-48T binding to IMC I-48T is depicted in red, and Sub WT interacting with IMC I-48T is displayed in green. Blue represents IMC WT interacting with Sub WT , and the gray line indicates IMC WT binding to Sub I-48T . c, a comparison of autocleavage of IMC WT -S221Csubtilisin (blue line) and IMC I-48T -S221Csubtilisin (red line) and a function of time. The amount of precursor autocleaved was estimated densitometrically from SDSpolyacrylamide gel. d, a comparison of Nsuccinyl-A-A-P-F-p-nitroanilide degradation monitored by changes in absorbance at 405 nm due to release of p-nitroaniline. Profiles due to Sub WT are shown by in blue, whereas Sub I-48T is shown in red.

FIG. 4. Schematic representation of folding mediated by the IMC domain.
Red depicts the IMC domain, whereas orange represents the protease domain. Green represents mutation within the IMC domain.
to the wild-type protease domain with the exception of the active site Ser 221 3 Cys substitution that causes a 10,000-fold drop in the proteolytic activity. Because uncleaved precursors display exposed hydrophobic surfaces relative to cleaved complexes, it is reasonable to speculate that distribution of charge on cleaved and uncleaved protein surfaces are different. It is known that altering surface charges through mutagenesis produces subtilisin with altered specificities and pH activity profiles with enhanced catalytic activities (48,49). Therefore, the active site within uncleaved precursor may also display enhanced autocleavage activity. After cleavage, the resulting conformational changes alter the catalytic properties of this active site. Because Ile Ϫ48 3 Thr substitution within the IMC does not affect cleavage but alters enzymatic activity of the protease domain, our results indicate that conformational diversity that leads to protein memory occurs late during the folding pathway and the IMC functions even after its autocleavage.
A sequence alignment of propeptides from the subtilisin family shows an interesting observation. Although the overall sequence identity is low, two regions dispersed over the N and C termini display significant sequence conservation (Fig. 5). These regions designated as motifs N1 and N2 contain the hydrophobic core residues (Val 12 , Phe 14 , Ile 30 , Val 37 , Leu 51 , Val 56 , Leu 59 , Val 65 , and Val 68 ) within the subtilisin propeptide, apart from Ile 30 . We speculate that motifs N1 and N2 may be critical for nucleation of the folding process, whereas the nonconserved segments between N1 and N2 may be crucial for specific interactions with their cognate protease domains, during the propagation of folding and subsequent inhibition of activity. This speculation is supported by four of our findings: (i) A substitution at position Ile 30 by a Val allows the propeptide to exert its chaperoning function but alters the final conformation of the protein (36). This Ile 30 residue is located in the nonconserved region flanked by motifs N1 and N2 (note that Ile 30 was previously termed Ile Ϫ48 ). (ii) By using an Ile 30 3 Thr substitution, we have shown that the structural imprinting, which results in protein memory, occurs late in the folding pathway, after the propeptide has been cleaved from the protease domain through autocatalysis. Therefore, a substitution at position 30 within the propeptide does not affect folding nucleation but dramatically alters the final conformation of the protein. (iii) Results of a random mutagenesis procedure that helped identify mutational hot spots within the propeptide show that substitutions that blocked folding of subtilisin are localized within motifs N1 and N2 (34). (iv) NMR spectroscopy on the propeptide of subtilisin E shows that residues located within motifs N1 and N2 display residue-specific conformational rigidity, 2 a criteria that classifies a polypeptide segment as a potential nucleation site in a protein folding reaction.