Role of proline residues in the folding of serine hydroxymethyltransferase.

Previous studies on the folding mechanism of Escherichia coli serine hydroxymethyltransferase (SHMT) showed that the final rate determining folding step was from an intermediate that contained two fully folded domains with N-terminal segments of approximately 55 residues and interdomain segments of approximately 50 residues that were still solvent exposed and subject to proteolysis. The interdomain segment contains 3 Pro residues near its N terminus and 2 Pro residues near its C terminus. The 5 Pro residues were each mutated to both a Gly and Ala residue, and each mutant SHMT was purified and characterized with respect to kinetic properties, stability, secondary structure, and folding mechanism. The results showed that Pro214 and Pro218 near the N terminus of the interdomain segment are not critical for folding, stability, or activity. The P216A mutant also retained most of the characteristics of the native enzyme, but its folding rate was altered. However, the P216G mutant was severely compromised in folding into a catalytically competent enzyme. Mutation of both Pro258 and Pro264 had altered folding kinetics and resulted in enzymes that expressed little catalytic activity. The Phe257-Pro258 bond is cis in its configuration, and the P258A mutant SHMT showed reduced thermal stability. Pro216, Pro258, and Pro264 are conserved in all 53 known sequences of this enzyme. The results are discussed in terms of the role of each Pro residue in maintaining the structure and function of SHMT and a possible role in pyridoxal 5'-phosphate addition to the apo-enzyme.

Previous studies on the folding mechanism of Escherichia coli serine hydroxymethyltransferase (SHMT) showed that the final rate determining folding step was from an intermediate that contained two fully folded domains with N-terminal segments of approximately 55 residues and interdomain segments of approximately 50 residues that were still solvent exposed and subject to proteolysis. The interdomain segment contains 3 Pro residues near its N terminus and 2 Pro residues near its C terminus. The 5 Pro residues were each mutated to both a Gly and Ala residue, and each mutant SHMT was purified and characterized with respect to kinetic properties, stability, secondary structure, and folding mechanism. The results showed that Pro 214 and Pro 218 near the N terminus of the interdomain segment are not critical for folding, stability, or activity. The P216A mutant also retained most of the characteristics of the native enzyme, but its folding rate was altered. However, the P216G mutant was severely compromised in folding into a catalytically competent enzyme. Mutation of both Pro 258 and Pro 264 had altered folding kinetics and resulted in enzymes that expressed little catalytic activity. The Phe 257 -Pro 258 bond is cis in its configuration, and the P258A mutant SHMT showed reduced thermal stability. Pro 216 , Pro 258 , and Pro 264 are conserved in all 53 known sequences of this enzyme. The results are discussed in terms of the role of each Pro residue in maintaining the structure and function of SHMT and a possible role in pyridoxal 5-phosphate addition to the apo-enzyme.
Serine hydroxymethyltransferase (SHMT) 1 catalyzes the reversible interconversion of serine and glycine with the onecarbon group being transferred to H 4 PteGlu as the 5,10-CH 2 adduct (1). This reaction is the major source of one-carbon groups in the cell that are required for thymidylate, purine, and methionine biosynthesis (2). SHMT is a member of the family of enzymes that contain tightly bound PLP as a cofactor. This family of enzymes is involved in many reactions in amino acid metabolism and in the biosynthesis of most neurotransmitters (3). When PLP is removed from these enzymes, they form the functionally inactive apo forms. Little is known about the in vivo mechanism of activation of the apo forms of B6 enzymes by PLP addition. Here we continue our study on understanding the mechanism of PLP addition and its relationship to the folding mechanism of eSHMT.
How and when PLP is added to apo-eSHMT from Escherichia coli during its biosynthesis and folding has been previously addressed (4 -6). Folding studies show that two domains fold rapidly and form a dimer without the involvement of PLP. An N-terminal segment of approximately 55 residues and an interdomain segment of approximately 50 amino acids remain sensitive to protease digestion until the final rate determining step in the formation of native apo-eSHMT. This 50-amino acid segment between the two domains not only contains Lys 229 , which forms the internal aldimine bond with the 4Ј aldehyde of PLP, but several other residues involved in PLP binding. PLP appears to bind only to the fully formed apo enzyme. Denaturation studies in urea show that release of PLP from the active site is one of the first events to occur in the unfolding process, which is in agreement with the addition of PLP during or after the final folding step. The final folding step is very temperature-sensitive, with a half-life of 40 min at 4°C and approximately 1 min at 30°C. The rate determining step exhibits an enthalpy of activation of 22 kcal mol Ϫ1 (6). At the boundaries of this interdomain segment are 5 Pro residues: 3 at the Nterminal end and 2 at the C-terminal end. The enthalpy of activation of converting a trans X-Pro peptide bond to the cis configuration is also approximately 20 kcal mol Ϫ1 (7). We proposed that one or more of these Pro residues would be in the cis configuration (6). Since those folding studies were published, the three-dimensional structure of eSHMT has been determined, and 1 of the 5 Pro residues is in the cis configuration.
In this study we report the effect of changing each of the 5 Pro residues that flank the interdomain stretch of 50 amino acids to both Ala and Gly and determine the effects of these mutations on catalytic activity, stability, and folding kinetics. We conclude that 2 of the Pro residues are not critical for the folding process but that the other 3 Pro residues are important in folding and maintaining catalytic activity. The Pro that is most critical is the one that is in the cis configuration. From the three-dimensional structure of the enzyme, we discuss the role of the 3 critical Pro residues in maintaining a competent active site. We also discuss the possible role of this 50-residue interdomain segment in the mechanism of converting apo-eSHMT into holo-eSHMT.

EXPERIMENTAL PROCEDURES
Materials-All coenzymes, buffers, amino acids, column chromatography supports, reagents for bacterial growth, and ultra pure urea were purchased from either Sigma-Aldrich or Fisher. Methylenetetrahydrofolate dehydrogenase was purified from an E. coli clone as previously described (8). Oligonucleotide primers for making mutant forms of eSHMT were obtained from Invitrogen. Mutant forms of eSHMT were made with the Transformer site-directed mutagenesis kit from Clon-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  tech (Palo Alto, CA). All of the restriction enzymes were purchased from New England Biolabs (Beverly, MA). The plasmid purification kit was purchased from Qiagen (Valencia, CA).
Site-directed Mutagenesis-The selection primer of the Transformer site-directed mutagenesis kit contained a mutation in the plasmid that converts a BglII site into a AflIII site. The mutation primers contained both the desired mutation as well as a silent mutation that either destroyed or introduced a restriction site in the eSHMT gene. The mutations for Pro 214 , Pro 216 , and Pro 218 were done in the W16F/W385F cDNA mutant of eSHMT, which contains only Trp 183 . The mutations for Pro 258 and Pro 264 were done in the W16F/W183F mutant cDNA of eSHMT, which contains only Trp 385 (5). The mutant oligonucleotide sequences used to mutate each Pro to an Ala and Gly contained the changes as follows: Pro 214 , CCG to GCG/GGG (Ala/Gly); Pro 216 , CCG to GCG/GGG (Ala/Gly); Pro 218 , CCT to GCG/GGG (Ala/Gly); Pro 258 , CCT to GCT/GGT (Ala/Gly); and Pro 264 , CCG to GCA/GGA (Ala/Gly). The mutations were verified by DNA sequencing of both strands of the coding region for eSHMT.
Purification of eSHMT Mutant Proteins-All mutant forms of eSHMT were expressed and purified as previously described for the wild-type enzyme (9 -11). In general, 6 liters of cells were grown in LB-ampicillin medium overnight and disrupted after harvesting in an Avestin Cell Disrupter (Ottawa, Canada). Each enzyme was purified by ammonium sulfate precipitation and DEAE Sephadex, phenyl-Sepharose, and hydroxylapetite chromatography. For some of the mutants a final purification was performed on a fast protein liquid chromatography MonoQ column (1.5 ϫ 6 cm) with a 30-min linear gradient from 20 mM potassium phosphate, pH 7.5, to 50 mM potassium phosphate, pH 7.3, containing 400 mM NaCl. Each mutant protein was greater than 95% pure as judged by SDS-PAGE patterns. P216G gave a very poor yield, and purity was only achieved after extensive purification on a sizing column and fast protein liquid chromatography. For this mutant we were able to only get approximately 2 mg of pure enzyme as compared with approximately 80 mg for all of the other Pro mutants.
Each mutant eSHMT contained tightly bound PLP. This was removed to prepare the apo enzymes by chromatography on a 0.7 ϫ 4-cm phenyl-Sepharose column equilibrated with 35% ammonium sulfate in 50 mM potassium phosphate, pH 7.3, and containing 100 mM L-cysteine. Under these high salt conditions the enzyme binds to the phenyl-Sepharose. The bound PLP reacts with L-cysteine to form a thiazolidine complex, which has reduced affinity for the enzyme and dissociates to leave column bound apo-eSHMT. After removing all of the PLP with the equilibration buffer, apo-eSHMT is eluted with 20 mM potassium phosphate, pH 7.3 containing 1 mM dithiothreitol.
Kinetic, Spectral, and Thermal Properties of Mutant Enzymes-K m and k cat values were determined for the conversion of serine and H 4 PteGlu to glycine and 5,10-CH 2 H 4 PteGlu by coupling the NADP dependent oxidation of the product CH 2 -H 4 PteGlu to CH ϩ -H 4 PteGlu and NADPH using methylenetetrahydrofolate dehydrogenase (10). Double reciprocal plots of initial velocity versus serine concentration was used to determine K m and k cat values with saturating levels of H 4 PteGlu (0.2 mM) used as the cosubstrate. The K d for H 4 PteGlu was determined by a spectrophotometric method involving the formation of an abortive complex with glycine that absorbs uniquely at 495 nm (10). Saturating levels of glycine (50 mM) were used, and the absorbance at 492 nm of the SHMT⅐Gly⅐H 4 PteGlu abortive complex was determined as a function of increasing concentrations of H 4 PteGlu.
Spectra of each mutant eSHMT were obtained with a HP 8450A spectrophotometer. Circular dichroism spectra from 260 to 184 nm were obtained with an OLIS recording CD spectrophotometer using an absorbance value of 0.22 at 278 nm for each enzyme in 20 mM potassium phosphate, pH 7.3, in a 1-mm-pathlength cell. Thermal transition values were obtained in a Nano II differential scanning calorimeter (Calorimetry Sciences, Corp., Salt Lake City, UT) at approximately 2 mg/ml concentration in the same buffer. Fluorescence emission spectra were obtained with an SLM Aminco Bowman Series 2 Luminescence Spectrometer (Rochester, NY) with a scan time of 5 nm/s at room temperature.
Equilibrium Unfolding of Mutant apo-eSHMTs-A concentrated solution of each apo-eSHMT was added to a series of urea solutions, in Tris-HCl, pH 7.5, and 5 mM 2-mercaptoethanol, ranging in concentration from 0.2 to 8.0 M as previously described (6). The final apo-eSHMT concentration was 0.15 mg/ml. The solutions were allowed to come to equilibrium by incubating for at least 2 h at 30°C or overnight at 4°C. The fluorescence spectrum of each mutant protein was taken with excitation at 290 nm and emission at 300 -410 nm. This procedure was repeated but starting with unfolded apo-eSHMT in Tris-buffered 8 M urea.
Determination of Refolding Rate Constants-Each mutant holo-eSHMT was reduced with NaCNBH 3 as previously described (6). The PLP aldimine bond of the holo enzyme is reduced by this procedure to a stable secondary amine of PyP and is referred to as eSHMT⅐PyP.
Reduction occurs with a spectral shift from 422 nm to 335 nm. The reduced mutant enzymes were unfolded in 8 M urea containing 5 mM 2-mercaptoethanol at a final concentration of 1.5 mg/ml of protein.
Refolding was initiated by a 10-fold dilution of this solution at 30°C in refolding buffer of 20 mM Tris-HCl, pH 7.5, and 5 mM 2-mercaptoethanol. The rate of refolding was followed by fluorescence energy transfer (FRET) between the excitation of a single Trp residue (the energy donor) at 290 nm and the fluorescence of the covalently attached PyP moiety (energy acceptor) at 384 nm (6). The rate constants and amplitudes for the appearance of fluorescence at 384 nm were determined by curve fitting routines in Sigma Plot.

RESULTS
Expression, Purification, and Activity of Pro Mutants of eSHMT-As reported previously the 3 Trp residues Trp 16 , Trp 183 , and Trp 385 in eSHMT can be replaced with Phe residues without significant alterations in the catalytic and stability properties of the enzyme (5). At the time when this work was published a three-dimensional structure of the E. coli enzyme was not available. From studies using FRET measurements of site mutants, proteolytic digestion studies during folding and fluorescence quenching of Trp residues, we proposed a detailed mechanism for the folding of eSHMT (5,6). A model of the final folding intermediate involved in the rate determining conversion to apo-eSHMT is shown in Fig. 1. Proteolytic digestion studies during folding showed that the active site Lys 229 , which forms the aldimine with PLP, is present on a sequence of approximately 50 amino acids between two rapidly folding domains that were protease resistant ( Fig. 1, Domains 1 and 2). In addition to this interdomain segment, the Nterminal 55 amino acid residues were also susceptible to protease digestion, suggesting that this region of the protein was also not in its native state. The protease-resistant domain 1 stretches from Tyr 55 to Thr 224 , and domain 2 stretches from Leu 276 to the C-terminal Ala 417 . The rate determining step in forming the fully protease-resistant enzyme was the "folding in" of the N-terminal 55 amino acids and the interdomain segment, which occur at the same rate. We noted that the boundaries of the interdomain segment are populated by 5 Pro residues (Fig. 1). Pro residues 214, 216, and 218 are near the C terminus of domain 1, and Pro residues 258 and 264 are in the interdomain segment near the N terminus of domain 2. To probe the function of these Pro residues on the folding mechanism of eSHMT, each was changed to an Ala and Gly residue. The mutations for Pro 214 , Pro 216 , and Pro 218 were done in the protein containing only the Trp at position 183, which is referred to as W183 eSHMT. The mutations of Pro 258 and Pro 264 were done in the construct containing only Trp 385 , which is referred to as W385 eSHMT. Having only one Trp residue permits an examination of the effect of these Pro residues on the folding of the domain with which they are most closely associated without interference of the fluorescence properties of the other two Trp residues.
Each mutant enzyme was tested for catalytic activity and affinity for serine and H 4 PteGlu (Table I). Only P216G, P258A, P258G, P264A, and P264G exhibited greatly reduced catalytic activity. With P258A and P258G the activity is so low that affinity for serine and H 4 PteGlu could not be determined. However, with P258A the addition of serine caused a spectral shift of the bound PLP from 406 to 416 nm. This spectral shift permitted the determination of a K d for L-serine of 8.0 mM. The K d value of 8 mM is 25-fold higher than the K m value for L-serine with the wild-type enzyme. The other Pro mutants exhibited K m values for serine that are not significantly different from wild-type eSHMT. Only P216G showed a significant change in the affinity for H 4 PteGlu, which is 5.5-fold higher than wild-type eSHMT.
Spectral Properties-With the exception of P258A and P258G, the spectral properties of each mutant eSHMT exhibits absorbance maxima at 418 and 322 nm. The wild-type enzyme shows a single absorbance maxima at 422 nm. The 322-nm peak was approximately 30 -50% of the height of the 418-nm peak and varied with the mutant and the age of the enzyme preparation. This suggests that some of the bound PLP is not in its catalytically active form. This probably accounts for many of the k cat values being only approximately 200 instead of 300 under our assay conditions. The importance of these spectral difference will be addressed under "Discussion." Each mutant protein was analyzed for secondary structure using circular dichroism spectroscopy. There was no significant difference in circular dichroic spectra between the mutant and the wild-type enzymes. Also, we had previously shown by circular dichroism that apparently complete secondary structure forms in less than a minute at 4°C, but the final folding step takes up to 15 h to complete at this temperature (6). We checked the circular dichroic spectrum of wild-type enzyme at 4°C during the first 10 min after refolding had been initiated by dilution of an 8 M urea unfolded solution. The sample was then warmed to 30°C, at which folding is completed in a few minutes. A second scan at this temperature showed no significant difference from the original spectrum taken at 4°C. We conclude that the secondary structural elements are essentially complete after a few minutes at 4°C when the enzyme is represented as the structure shown in Fig. 1.
Stability of Pro Mutants-The ability of each mutant to undergo a reversible unfolding-folding cycle in increasing concentrations of urea was determined (5). A plot of fluorescence emission at 330 nm versus urea concentration showed a sigmoidal decrease in fluorescence with increasing concentration of urea characteristic of a two-state folding curve. However, because each mutant enzyme contained only Trp 183 or Trp 385 , this method follows only the degree of folding of either domain 1 or domain 2. The midpoint of the curve for W183 eSHMT is 2.5 M urea. Mutants P214A, P214G, and P218 A also had midpoints at 2.5 M urea. Mutants P216A and P216G exhibited midpoints at 2.0 -2.2 M urea, showing that domain 1 in these mutants is slightly less stable. The urea midpoint of W385 eSHMT is 3.0 M urea (6). The midpoint for P258A and P258G were also at 3.0 M urea, but for P264A and P264G the midpoint was at 2.5 M. These results suggest that the mutation of the 5 proline residues has only marginal effects on stability of domains 1 and 2 as determined by unfolding in urea.
The thermal stability of each alanine mutant protein was determined by differential scanning calorimetry. The thermal denaturation of eSHMT is irreversible. An apparent T m (T m, app ) was determined from the peak in the thermogram for each Pro mutant in the absence and presence of 30 mM L-serine (Table II). Each Pro mutant showed similar heat stability (T m, app ϭ 63-67°C) and stabilization of approximately 5°C upon the addition of serine. The exception is the P258A mutant, which exhibits a double thermal transition that is considerably lower than wild-type enzyme (Fig. 2). For this mutant there is no increase in thermal stability upon binding serine, even though spectral studies showed that serine binds to the active site with a K d of 8 mM (Table I). Most of the denaturation traces exhibited nonsymmetrical curves. This has been shown before to result from different domains in the enzyme unfolding at different temperatures and is consistent with the urea denaturation curves showing that domain 2 is slightly more sta-  ble than domain 1 (12). Apo-W183 eSHMT denatures approximately 3-4 degrees lower than holo-W183, suggesting that PLP binding does not greatly increase thermal stability in eSHMT (Table II).
Kinetics of Refolding-The refolding rate was determined by using FRET, which occurs during the final rate determining step in folding when the interdomain segment and the Nterminal 55 residues collapse into the native structure. The fluorescence traces for P214A and P216A eSHMT are shown in Fig. 3. We had previously shown that this final step in folding follows a first order reaction (6). In these studies of Pro mutants, the traces are best fit by two parallel first order reactions. The two first order rate constants and their relative associated amplitudes (shown in parentheses) are recorded in Table I. The slower of the two rate constants is in the 0.008 s Ϫ1 range. These values are close to the values observed for wild-type eSHMT under these same conditions. The faster rate is not observed with wild-type eSHMT. Note that in the refolding of P214A, 90% of the amplitude follows a first order refolding path. With P216A 66% folds by the faster rate, and the 34% folding at the slower rate occurs approximately 2-fold slower than P214A. P216G, P258A, and P258G give no FRET signal, suggesting that in the final structure the PyP group is more than 40 Å distant from the Trp residue associated with its closest domain.
Three-dimensional Structure of eSHMT- Fig. 4 shows the location of each of the 5 Pro residues, with respect to the active site PLP (shown in green). Pro 214 , Pro 216 , and Pro 218 are part of a structural motif from Ala 210 through His 221 that forms a long loop between a ␤ strand (Val 222 -Thr 226 ) and an ␣ helix (Ala 205 -Ala 209 ). Although the loop is exposed to the solvent, the ␤ strand and the ␣ helix are buried and form a part of the  (bold dashed line). The y axis is the molar heat capacity. Because denaturation is not reversible and results in aggregation, the peaks are not really T m values, and the area under the curves not the actual heat released during denaturation, because aggregation is exothermic. We therefore define the peak as an apparent T m (T m, app ).

FIG. 3. FRET traces of the refolding of Pro mutants of eSHMT.
Holo enzyme that had been reduced and unfolded in urea was diluted in Tris-2-mercaptoethanol buffer, pH 7.5. The folding of the final step was monitored by FRET with excitation of Trp residues at 290 nm and emission of the PyP group at 384 nm. The dashed line is for the P214A mutant, and the solid line is for the P216A mutant. Rate constants and amplitudes were determined in Sigma Plot using the double exponential with five variables equation as the fitting routine. The square of the correlation coefficient (r 2 ) was better than 0.998 for the fitted curves. catalytic site. Some of the residues from the ␤ strand and ␣ helix and others immediately preceding or following these secondary structures are known to play significant roles in PLP binding and catalytic activity. These include Asp 200 , His 203 , His 228 , and Lys 229 (13). Interestingly, Pro 216 is located in the middle of the loop, and whereas the side chains for Pro 214 and Pro 218 face the bulk solvent, that of Pro 216 faces the interior of the protein and forms part of a hydrophobic patch The Pro 258 -Pro 264 motif is also a loop, with the two Pro residues located at the ends of the loop and initiating and terminating two ␣-helices (Glu 246 -Phe 257 and Leu 265 -Glu 278 , respectively). The loop region and part of the ␣-helices form a significant part of the catalytic site. The amide nitrogen of Gly 263 makes a short hydrogen bond interaction with the phosphate of the PLP in the other monomer. The Pro 258 -Glu 259 -Gly 260 -Gln 261 -Gly 262 -Gly 263 -Pro 264 motif is located at the interface of the two subunits and is involved in many direct hydrogen bond interactions across the subunit interface, including Asn 134 ND2 to Pro 258 O, Asn 134 ND2 to Gly 259 O, Phe 135 N to Gly 259 O, Gln 100 NE2 to Gly 260 O, and Ser 99 OG to Gly 262 N. There are a number of significant water-mediated interactions involving the loop region and residues from the other subunit and with PLP (Fig. 5). DISCUSSION Pro Residues and Folding-The role of Pro residues in protein folding has been widely discussed and studied (14,15). Pro residues often act to slow down the folding process creating intermediates on the folding pathway. Of particular interest is the role of cis Pro residues that often cap ␣-helices or are involved in ␤ turns. The reversible cis-trans interconversion of X-Pro peptide bonds exhibits a large enthalpy of activation of approximately 20 kcal mol Ϫ1 (15). In the unfolded state the equilibrium favors the more stable trans form by a 3 or 4 to 1 ratio. Most Pro residues in proteins are trans, but approximately 6% are cis (16). In small proteins and some larger ones Pro isomerization can be the rate determining step in folding and because Pro residues exist as a mixture of isomers in the unfolded state leads to multiple parallel pathways in folding (15,17). In trying to discern the role of Pro residues in the folding of larger multidomain proteins, it is helpful to have structural information about intermediates in which Pro residues play a role.
Structure of the Folding Intermediate in eSHMT-The folding mechanism of eSHMT has been elucidated previously, and by using a variety of methods we were able to define several steps and intermediates on the reversible folding-unfolding pathway (4 -6). For our purposes in this study we divide the folding mechanism into two phases: a relatively rapid phase lasting less than a minute and involving several intermediates in which two domains have folded into their native state and a dimer is formed and a slow final phase that involves the folding in of the N-terminal 55 amino acid residues and a 50-residue interdomain segment. The final rate determining step is described by a single first order rate as determined by using information from FRET studies between three widely spaced Trp residues and an active site PyP, suggesting that the first rapid steps in folding lead to a single structural intermediate. The slower final rate determining step of this multidomain homodimer is complete in approximately 5 min at 37°C and exhibits an energy of activation of 22 kcal mol Ϫ1 , leading to the prediction that the formation of a cis X-Pro peptide bond is involved.
Subsequent to these folding studies the three-dimensional structure of several SHMT proteins, including eSHMT, were determined (13, 18 -21). The SHMT structures from human, mouse, rabbit, E. coli, and Bacillus stearothermophilus all have a cis X-Pro bond equivalent to Pro 258 in eSHMT. These struc- tural studies also define the positions of all 5 Pro residues that were mutated in this study (Figs. 4 and 5). One question is how good were the previous folding studies in defining the structural properties of the intermediate shown in Fig. 1? One test of the model is the accuracy of the distances calculated between the 3 Trp residues and the active site PyP for the final structure. In the intermediate there was no FRET between these groups, but in the final structure we predicted that Trp 15 , Trp 183 , and Trp 385 were 27, 17, and 33 Å from the active site PLP (6). From the now known eSHMT structure, these distances are correct to within an angstrom. Another test of the model is from fluorescent quenching studies showing that Trp 16 is mostly solvent exposed, Trp 183 is buried in a hydrophobic core, and Trp 385 is partially solvent exposed (5). Again, the three-dimensional structure confirms these results. These comparisons help confirm that our conclusions about the structure of the intermediate shown in Fig. 1 have validity.
The Size of Domains 1 and 2-The intermediate exhibits subtilisin digestion sites at Tyr 55 , Thr 224 , and Leu 276 , all of which disappear in the folded enzyme. This tells us that these sites are solvent-exposed only in the intermediate, but because of protease specificity and steric effects it does not tell us the boundaries of domains 1 and 2. Noting that the collapse of the structure into its final protease-resistant form requires rotation around peptide bonds, we looked for logical places in the structure that could explain the results. We propose that domain 1 starts at the Pro 60 -Gly 61 peptide bond, which is on the surface of the protein in the final structure and could act as a hinge point for collapse of the N-terminal segment into the final structure. Likewise, we think that domain 1 ends at Gly 211 , which is part of a loop on the surface of the protein. The subtilisin digestion site at Thr 224 is buried in the native protein and is a part of a ␤-strand. It is unlikely to be a hinge point, but because it is exposed in the intermediate, Pascarella et al. (22) have defined a special algorithm for comparing sequence alignments of SHMT proteins. Using their alignments in 53 SHMT structures from a wide variety of organisms, we found that our putative first hinge point Gly 61 is conserved in 39 of 53 known structures of SHMT and its preceding Pro 60 is conserved in 49 of the 53 structures. The second predicted hinge point at Gly 211 is conserved in 45 of 53 SHMT structures, and those that do not contain Gly 211 have either a preceding Gly or Ala residue. The small loop structure at Ala 279 -Glu 283 is also highly conserved with 43 sequences having an equivalent Ala 212 and 42 sequences having either a Glu 283 or Asp 283 . Also, more than half of the sequences have a conserved Pro 282 that precedes the Glu 283 found in eSHMT. These putative hinge points will be tested in future studies.
Secondary Structure of the "Intermediate"-Fluorescence polarization studies showed that the PyP group attached to the ⑀-amino group of Lys 229 was rotating freely (6). However, we could make no statement about the secondary structural aspects of the N-terminal segment or the interdomain segment in our model shown in Fig. 1. We now know that both of these segments contain considerable ␣-helices and ␤-strands. We calculate that 28% of the ␣-helices and 25% of the ␤-strands are located in the N terminus and interdomain segments. However, the circular dichroic spectrum of the enzyme taken at 4°C after initiation of refolding does not change by more than 2 or 3% from the spectrum of the fully folded enzyme. Therefore, we conclude that the N-terminal segment and the interdomain segment contain most, if not all, of the secondary structural elements found in the fully folded enzyme. Role of Pro 214 and Pro 218 -Changing Pro 214 and Pro 218 to Ala or Gly has little effect on the catalytic activity or thermal stability of the enzyme. The decrease in activity is most likely related to the folded form where the PLP absorbs at 322 nm. The affinity for both L-serine and H 4 PteGlu remain largely unchanged for these two Pro mutants, suggesting that the active site of the enzyme is not greatly changed. The fact that both Pro 214 and Pro 218 are located on the surface of the enzyme suggests that these residues do not play a critical role in the final folding step. An examination of 53 sequences of SHMT from a wide spectrum of organisms shows that 36 retain a Pro equivalent to Pro 214 in eSHMT. Likewise, 23 SHMT sequences retain a Pro residue equivalent to Pro 218 in eSHMT. We conclude that although Pro 214 and Pro 218 may play some role in the final folding step, it is not a crucial role.
Role of Pro 216 -The P216A mutant exhibits catalytic and stability properties that resemble those of the wild-type enzyme. However, the P216G mutant has very low activity and could only be purified in poor yield. Pro 216 lies in the middle of a loop and points to the interior of the protein and is a part of a hydrophobic cluster that includes Phe 86 , Trp 183 , Val 204 , Tyr 213 , Val 199 , Met 201 , Val 208 , and Val 223 . Apparently substitution by an Ala maintains this hydrophobic patch, but substitution with Gly results in the loss of its integrity, leading to serious misfolding problems. Of interest is that all 53 SHMT sequences have a Pro at this position.
Roles of Pro 258 and Pro 264 -The most critical residues examined in this study are Pro 258 and Pro 264 . Pro 258 is in a cis configuration in the five known structures from mouse, human, rabbit, E. coli, and B. stearothermophilus. Our evidence that the secondary structure of the interdomain segment has already been formed in the intermediate would suggest that this Phe-Pro bond is already cis and would explain why the final folding step occurs as a single exponential reaction in wild-type eSHMT. Both Pro 258 and Pro 264 lie next to the PLP ring in the adjoining subunit (Fig. 5). Any changes in the orientation of these two residues can alter the orientation of the PLP ring. During a catalytic cycle the PLP rotates to accommodate the incoming amine substrate and the leaving of the ⑀-amino group of Lys 229 (18 -21). The Pro residues must maintain this rotational flexibility. When these two Pro residues are replaced by either Gly or Ala, they result in either inactive enzymes or ones with a very low turnover rate. However, the spectrum of these enzymes shows that PLP is still mostly bound as an aldimine, so they do not lack activity because PLP is not there. Also, P258A can still form an external aldimine with L-serine, as judged by the spectral change when serine is bound. The addition of glycine and H 4 PteGlu to every SHMT that has been studied results in an abortive complex that absorbs near 500 nm, which is characteristic of a quinonoid structure known to be on the catalytic pathway. However, the Pro 258 and Pro 264 mutants do not give this complex with glycine and H 4 PteGlu, suggesting that they are blocked in going from the external aldimine to the quinonoid intermediate. A previous study of sheep liver SHMT also showed that the P297R mutant (equivalent to Pro 258 ) resulted in the loss of 85% of its activity (23). All 53 sequences of SHMT have Pro residues equivalent in position to Pro 258 and Pro 264 .
Physiological Role of the Folding Mechanism- Fig. 6A shows a stereo ribbon diagram of the enzyme with the interdomain section between Val 212 and Ala 279 lifted out of the structure. The yellow ribbon represents one of the subunits in the dimer. The magenta, gray, cyan, and brown ribbons are the other subunit and represent domain 1, domain 2, the N-terminal segment, and the interdomain segment, respectively. The brown lift-out shows the positions of the 5 Pro residues (shown in black) as they will appear in the final structure. Fig. 6B shows how the interdomain section fits into the total structure. Pro 214 and Pro 218 are on the surface of the protein, whereas Pro 216 is buried. Also Pro 264 is buried with Pro 258 being partially buried. PLP is shown in dark blue, and 5-CHO-H 4 PteGlu is in red. Lys 229 on the lift-out interdomain segment (brown) forms the PLP aldimine that is on the backside of the structure as it is shown. In Fig. 6A the PLP is exposed, but in 6B it is clear that in the final folded structure the PLP is buried. The figure shows, however, that 5-CHO-H 4 PteGlu is bound largely at the surface of the enzyme.
We propose that the final folding step is involved in PLP binding. Removing the 50-residue segment (shown in Fig. 6A) exposes the PLP site to solvent. Our hypothesis is that in the absence of PLP the interdomain and N-terminal segments are in equilibrium with a more open structure that is similar to the intermediate shown in Fig. 1. In this more open structure PLP can bind at a pre-site and then is folded in to form the active site. The new contacts between the coenzyme and the protein stabilize this closed form, resulting in the tight binding properties seen in all PLP enzymes. Our previous studies showed that in denaturation at low urea concentrations, the first event is the release of PLP into solvent but that this event did not greatly alter the denaturation properties of domains 1 and 2. This is confirmed in this study where the thermal denaturation of apo-W183 is very similar to that of holo-W183 (Table II). When the PLP aldimine bond is reduced with NaCNBH 3 , it now requires a much higher concentration of urea (midpoint is 6 M urea) and temperature (T m of 82.4°C), because the first step in the unfolding pathway has been greatly stabilized (6).