Structural studies on folding intermediates of serine hydroxymethyltransferase using fluorescence resonance energy transfer.

Previous studies have demonstrated that the in vitro folding pathway of Escherichia coli serine hydroxymethyltransferase has both monomer and dimer intermediates that are stable for periods of minutes to hours at 4°C (Cai K., Schirch, D., and Schirch, V. (1995) J. Biol. Chem. 270, 19294-19299). Single Trp mutant enzymes were constructed and used in combination with other methods to show that on the folding pathway of this enzyme two domains rapidly fold to form a monomer in which the amino-terminal 55 amino acid residues and a segment around the active site region of Lys229 remain in a largely disordered form. This partially folded enzyme can form dimers and slowly undergoes a rate-determining conformational change in which the unstructured segments assume their native state (Cai, K., and Schirch, V. (1996) J. Biol. Chem. 271, 2987-2994). To further assess the kinetics and structural details of the intermediates during folding, fluorescence energy transfer and fluorescence anisotropy measurements were made of the three Trp residues and pyridoxal 5′-phosphate, attached covalently to the active site by reduction to a secondary amine by sodium cyanoborohydride. These studies confirmed that the basic kinetic folding pathway remained the same in the reduced enzyme as compared to the earlier studies with the apoenzyme. Both equilibrium and kinetic intermediates were identified and their structural characteristics determined. The results show that the active site Lys229-bound pyridoxyl 5′-phosphate remains more than 50 angstroms from any Trp residues until the final rate-determining conformational change when it approaches each Trp residue at the same rate. The environment of each Trp residue and the pyridoxyl phosphate in both an equilibrium folding intermediate and a kinetic folding intermediate are described.

We have demonstrated in previous studies that eSHMT that had been unfolded in 8 M urea can be completely refolded after a 10-fold dilution in Tris-HCl buffer at neutral pH (3,4). There was a significant temperature dependence of the folding rate, in which complete folding took less than 10 min at 30°C but required 15-20 h at 4°C. Both kinetic and equilibrium folding intermediates were identified at 4°C, including a dimer that did not bind pyridoxal-P (3). In another study, two of the three Trp residues in the protein were changed to Phe by site-directed mutagenesis to generate a set of three single Trp mutant enzymes (4). These mutant enzymes were purified and characterized and shown to retain essentially all of the properties of the wild-type enzyme. The folding process was then studied at 4°C by Trp fluorescence, circular dichroism, and limited proteolysis. The results demonstrated that the refolding enzyme forms two domains (domain 1 is from Tyr 55 to Thr 224 and domain 2 is from Leu 276 to Ala 417 ) in less than 10 s at 4°C. These two domains are resistant to subtilisin digestion and form the monomer M in Equation 1. The amino terminus and a central loop region between the two domains remain in a form susceptible to protease digestion. M contains essentially all of the secondary structure of the native enzyme. M then forms a dimer (DЈ) in the next 20 min. DЈ does not bind PLP but undergoes a slow rate-determining conformational change taking 15-20 h to form a structure characteristic of the native apoenzyme (apoD). PLP binds only to this apparent apoD to form holoD, which is the native and catalytically active enzyme (Equation 1). Fluorescence studies during refolding of the three single Trp mutants clearly identified the changes in environment occurring for Trp 16 in the amino-terminal segment, Trp 183 in domain 1, and Trp 385 in domain 2. The region between the two domains (residues 225-276) appeared to remain disordered in both intermediates M and DЈ as judged by susceptibility to protease digestion. However, the segment between the two domains has no fluorescent probe to monitor what is occurring during refolding.
In addition to these kinetic studies, the refolding of apo-eSHMT was also studied by equilibrium methods in increasing concentrations of urea. An  In the present work, we used NaCNBH 3 to reduce the bound PLP of holo-eSHMT to form a stable pyridoxyl-P secondary amine at Lys 229 , namely PyP-eSHMT. The reduced coenzyme remains attached in the protein in the unfolded state. This procedure places a fluorescent probe in the region of the protein between the two domains. Reduction of the PLP aldimine had an unexpected dramatic effect on the stability of the enzyme. This necessitated an investigation of the refolding pathway of this reduced eSHMT to determine whether a change in folding mechanism had occurred.
Fluorescence resonance energy transfer (FRET) between energy donors and acceptors has been widely used to determine distance relationships in macromolecules (5)(6)(7). The efficiency of this transfer depends on the inverse sixth power of the distance between the energy donor and the acceptor (8). The practical difficulty of applying this method to protein folding studies is that it is rare for a protein to contain both an intrinsic energy donor and acceptor (6). Blond and Goldberg (9) have shown that reduction of the aldimine between PLP and the active site Lys of the ␤ 2 subunit of tryptophan synthase permits Trp to transfer energy to the reduced PyP moiety. The reduction of the PLP aldimine bond in eSHMT results in an absorption maximum at 335 nm that overlaps the emission spectrum of Trp. Hence, the PyP can be used as an energy acceptor of Trp fluorescence to determine distance relationships in each folding intermediate of PyP-eSHMT. Also, fluorescence anisotropy measurements were used to monitor the rotational flexibility of the PyP moiety and Trp residues in each intermediate. These two techniques, together with the single Trp mutants constructed previously, provide important information about the structure of both an equilibrium intermediate (I e ) and a kinetic intermediate (I k ) in the refolding pathway of eSHMT.

EXPERIMENTAL PROCEDURES
Materials-t-Butoxycarbonyl-Lys was purchased from Advanced Chemtech (Louisville, KY). All other coenzymes, amino acids, buffers, and reagents were obtained from Sigma and were of the highest grade available.
Single Trp mutant eSHMTs were constructed by site-directed mutagenesis as described previously (4). Wild-type and single Trp mutant eSHMTs were overexpressed in E. coli and purified to near homogeneity (4). Thermograms for the denaturation of each protein were obtained with an MC-2 scanning calorimeter as described previously in 20 mM potassium phosphate (potassium P i ) buffer, pH 7.3, containing 5 mM 2-mercaptoethanol (4).
Synthesis of PyP-eSHMT-Protein samples (1 mg/ml) were dialyzed in 20 mM potassium P i , 5 mM 2-mercaptoethanol, pH 7.0, with 0.1 mM PLP. Each enzyme was then dialyzed against the potassium P i buffer without PLP for 4 h to remove excess PLP. The reduction of the PLP aldimine was achieved by dialyzing each protein in potassium P i buffer with 0.1% NaCNBH 3 for 5 h. This was followed by exhaustive dialysis in 20 mM potassium P i buffer overnight. Reduction of PLP was verified by observing the spectral shift of bound PLP from 422 to 335 nm and the loss of more then 95% of its catalytic activity. Treatment of the apoenzyme by the same reducing conditions resulted in fully active eSHMT after removal of NaCNBH 3 and reconstitution with PLP. The stoichiometry of the reduced eSHMTs was determined to be 1 PyP in each subunit in all cases, using an ⑀ 335 ϭ 8700 cm Ϫ1 M Ϫ1 to determine the concentration of PyP (10). Protein concentration was determined from its A 278 nm as described previously (4).
⑀-Lys-PyP was synthesized by the following procedure. One ml of 2 mM ␣-t-butoxycarbonyl-Lys was mixed with 1 ml of 1 mM PLP and 2 ml of 0.5% NaCNBH 3 . The resulting solution was incubated at 23°C for 2 h. During this period, the 390 nm peak of PLP was completely shifted to 325 nm. The t-butoxycarbonyl-Lys-PyP was then deblocked by incubation in 30% trifluoroacetic acid for 15 min. The resulting sample was neutralized by NaOH and chromatographed on a 1.5 ϫ 100-cm Bio-Rad P2 column equilibrated with water. The fractions containing material having an absorbance maximum at 325 nm and fluorescence emission at 386 nm (excitation at 335 nm) were pooled and used as the free Lys-PyP control in fluorescence anisotropy studies.
Fluorescence Measurements-Fluorescence spectra were taken on a Shimadzu 5000 fluorimeter with 5-nm slits for both excitation and emission. The concentration of eSHMT was 0.15 mg/ml for all experiments unless noted otherwise. For kinetic studies, the initiation of refolding by manual mixing of samples took about 9 s. The scanning of spectra from 300 to 450 nm took 6 s. A circulation bath was used to maintain the desired temperature.
Fluorescence anisotropy measurements were performed on a SLM AB-2 Fluorescence Polarimeter at the Center of Fluorescence Spectroscopy, University of Maryland at Baltimore. Fluorescence lifetime measurements were also performed on a frequency-domain instrument in the above center as described previously (11). Briefly, a rhodamine 6G dye laser was used as the excitation source at 290 nm, and the modulated emission was detected with a Hamamatsu R1564U microchannel plate photomultiplier tube with a 340-nm interference filter and 10-nm bandwidth. The lifetimes of Trp residues were resolved by a computer program for multiexponential intensity decay analysis developed by Dr. M. L. Johnson (University of Virginia, Charlottesville, VA).
Distance Calculations between Trp Residues and PyP-Since there is no available Förster distance (R 0 ) for Trp and PyP, the R 0 was determined as follows. The deconvoluted absorbance spectrum of each single Trp mutant PyP-eSHMT between 300 and 400 nm was integrated with 2-nm intervals by a spreadsheet computer program (Cricket Graph III, Version 1.5.1) and converted to an acceptor extinction coefficient as a function of wavelength (⑀ A ()). The donor fluorescence (F D ()) was obtained by the same integration method on the fluorescence emission spectrum of each single Trp mutant apo-eSHMT. The spectral overlap integral (J) of donor fluorescence and acceptor absorption was calculated by Equation 2 (6).
Since J had units of M Ϫ1 cm Ϫ1 nm 4 , R 0 was determined for each mutant eSHMT by using Equation 3, where R 0 is the Förster critical distance (in Å) at which transfer efficiency (E) is 50% and n is the refractive index (n ϭ 1.4 for water) (6).
Donor quantum yield in the absence of acceptor ( D ) was determined on single Trp mutant apo-eSHMTs based on a value of 0.14 for tryptophan at 25°C (12). A value of two-thirds was used for the orientation factor ( 2 ) by assuming that the Trp and PyP groups rotate in a shorter time relative to the excited state lifetime of the donor (7). In addition, the minimum and maximum values of 2  and single Trp mutant apo-eSHMTs were used to obtain D . The apparent distance (R) between Trp and enzyme-bound PyP was then calculated for each single Trp mutant eSHMT (6).
Equilibrium Unfolding-Refolding-Twenty mM Tris-HCl, pH 7.5, 5 mM 2-mercaptoethanol, and 1 mM EDTA buffer was used throughout the unfolding-refolding experiments and will be referred to as the Tris buffer. Equilibrium unfolding-refolding was performed by adding a concentrated solution of either wild-type or mutant PyP-eSHMTs to a series of urea concentrations in the Tris buffer (4). These solutions were incubated for 5 h at 30°C before analysis by fluorescence spectrometry. The fluorescence equilibrium unfolding-refolding data were converted to the apparent fraction of native protein (F app ) and plotted versus urea concentration, as defined by Equation 4.
Y obs is the observed value of the fluorescence signal at a defined wavelength and a particular urea concentration in the region where the protein is unfolding. The fluorescence values for Y N (native PyP-eSHMT) and Y U (unfolded PyP-eSHMT) were obtained by linear extrapolation of the base lines for native and unfolded protein into the region where the protein was unfolding (14,15).
The equilibrium unfolding-refolding plots were fit to either a twostate or a three-state model. In a three-state model, an equilibrium intermediate (I e ) accumulates as shown in Equation 5. For a two-state model, only U and N are present. The equations used to fit each model are as described by Matthews and Crisanti (15).
From each curve-fit for the equilibrium unfolding-refolding studies, the value of ⌬G°for unfolding, extrapolated to zero urea concentration, was calculated using the relationships described in Matthews and Crisanti (15), Pace (16), and Nath and Udgaonkar (17).
Kinetic Refolding Properties of PyP-eSHMT-Unfolding of eSHMT was performed by diluting 5-fold a concentrated enzyme solution (7.5 mg/ml) into 10 M urea. Refolding experiments were initiated by diluting the unfolded enzyme 10-fold into Tris buffer (3,4). Native controls were made in 0.8 M urea with Tris buffer, and the final protein concentration was exactly the same as the samples used in these refolding experiments. The unfolded control was made in the same solution as the native control except that it contained 8 M urea. The unfolded and native controls were used to define the 0 and 100% fractions folded, respectively.
Limited Proteolysis-Subtilisin digestion of apo-and PyP-eSHMT during refolding was the same as described previously (4), except that the protease concentration was reduced to 0.5 mg/ml. Subtilisin was added to each refolding sample for 2 min at 4°C after various periods of refolding. The digestion was stopped by the addition of phenylmethylsulfonyl fluoride before running on an SDS-PAGE gel.
Urea Gradient and Native Gel Electrophoresis-A 0 -8 M linear urea polyacrylamide gel was made as described previously (4). Seventy-five g of each form of PyP-eSHMT, in Tris buffer, pH 8.0, was layered on top of the urea gradient gel in a 6-cm slot. Electrophoresis was performed at 300 V for 2 h at room temperature, and the gel was then stained with Coomassie Blue R250.
Native acrylamide gel electrophoresis was performed in a 5% gel in 35 mM N- [2-hydroxyethyl]piperazine-NЈ-[2-ethanesulfonic acid] that had been adjusted to pH 7.0 with a 1 M Tris solution. Samples were loaded and electrophoresis was performed at 12 mA for 1 h.

RESULTS
Thermal Stability of PyP-eSHMT-Reduction of the PLP internal aldimine with NaCNBH 3 is expected to be a mild chemical procedure with a minimal effect on enzyme structure. However, this minimal effect does not appear to be true. Previous studies have shown that differential scanning calorimetry can be used to monitor the conformational state of eSHMT between putative open and closed forms (18). Formation of the external aldimine by saturation with serine causes a significant increase of 8°C in the T m of the protein and, with other evidence, suggests that the eSHMT⅐serine complex is in the more stable closed form (18). Fig. 1 shows the thermograms for holo-eSHMT, eSHMT⅐serine complex, and wild-type PyP-eSHMT obtained by differential scanning calorimetry. Reduction of the internal aldimine increases the T m (82.4°C) by 12.6°C compared to the unreduced enzyme, resulting in an enzyme that is more thermally stable than the eSHMT⅐serine complex. This large increase in T m suggests that reduction of the internal aldimine may have resulted in a conformational change in the protein, which results in an altered folding pathway. Similar 12-13°C increases in T m were observed for each of the single Trp mutant PyP-eSHMTs. Previously, Chaffotte and Goldberg had shown that the ␤ 2 subunit of tryptophan synthase undergoes a conformational change upon reduction of the bound PLP by sodium borohydride (19). The possibility that a conformational change results in an alternative folding pathway is addressed in the following experiments.
Protease Digestion of Folding Intermediates-Previous studies on the refolding of apo-eSHMT at 4°C showed that subtilisin digested the enzyme into two large fragments during the initial stages of refolding. The 23-kDa domain covered residues 55-224, and the 17-kDa domain covered from residue 276 to the carboxyl terminus residue at 417 (a 20-kDa fragment was also present during digestion during the first 2 min of refolding). The native enzyme was completely resistant to subtilisin digestion under the conditions used during refolding. With apo-eSHMT, protease-resistant forms characteristic of native enzyme did not begin to appear until after 15 min. These protease digestion experiments were interpreted to show that the two domains of apo-eSHMT fold rapidly to protease resistant forms (4).
To test the refolding mechanism of the PyP-eSHMT, these studies with subtilisin were repeated. The protease digestion pattern, as analyzed by SDS-PAGE, is shown in Fig. 2. As observed before, if subtilisin is added to the initial refolding buffer and digestion allowed to proceed for 2 min, there are three bands that migrate at 23, 20, and 17 kDa, respectively (lane 1, apo-eSHMT; lane 2, PyP-eSHMT). Lanes 3, 4, 5, and 6 show the digestion pattern of PyP-eSHMT when subtilisin was added 0.25, 1.7, 17, and 60 min after initiation of refolding, respectively. Lane 7 shows the digestion pattern for PyP-eSHMT after refolding for 60 min at 4°C and then for 10 min at 30°C before addition of the protease. Lane 8 is the native PyP-eSHMT incubated with subtilisin for 2 min. Lane 9 is native PyP-eSHMT without subtilisin, and lane 10 is molecular mass standards.
The similarity between protein bands on SDS-PAGE following digestion with subtilisin of PyP-eSHMT and those previously observed for apo-eSHMT provides support for a similar mechanism of domain folding being retained in the reduced enzyme. One difference is that a trace of the PyP-eSHMT that is resistant to subtilisin digestion starts to appear during the 1.7-3.7 min digestion period, which is sooner than observed with the apoenzyme, suggesting that in the PyP-eSHMT, the rate-determining step may be slightly faster then with apo-eSHMT.
Spectral Properties of PyP-eSHMTs-Treatment of wild-type holo-eSHMT with NaCNBH 3 results in the reduction of the internal aldimine bond between PLP and the ⑀-amino group of Lys 229 to form a stable secondary amine that does not break during unfolding of the enzyme. The reduction also results in a shift in the absorption maximum of bound PLP from 422 to 335 nm (Fig. 3). Treatment of apo-eSHMT with NaCNBH 3 does not alter its ability to form active holo-eSHMT in the presence of PLP, showing that the procedure does not alter protein structure in the absence of the bound PLP. Shifting the absorption maximum from 422 to 335 nm results in considerable overlap with the fluorescence emission spectrum of Trp residues in apo-eSHMT (Fig. 3, inset).
Evidence for FRET in PyP-SHMT-The fluorescence emission spectra of the reduced forms of wild-type and the three single Trp mutant PyP-SHMTs excited at 290 nm are recorded in Fig. 4. For each native PyP-eSHMT, there are two emission maxima (solid lines), one around 323-335 nm (Trp fluorescence) and the other at 380 nm (PyP emission resulting from FRET). Fig. 4 also shows the emission spectrum of the unfolded enzymes in 8 M urea with a single emission maximum at 352 nm (dotted lines). The absence of an emission maximum at 380 nm in the unfolded enzymes show that there is no FRET between any of the Trp residues and the bound PyP. The last panel in Fig. 4 contains the spectra of native and unfolded Trp Ϫ PyP-eSHMT (contains no Trp residues). The small amount of emission confirms that the major portion of fluorescence for each mutant is from the single Trp residue. In the native state, each of the three single Trp PyP-eSHMTs exhibits FRET, as determined by the emission at 380 nm, showing that each Trp residue is located less than 50 Å from the PyP in the native state (Table I). The small amount of emission at 380 nm in Trp Ϫ PyP-eSHMT may arise from FRET from a Tyr residue(s).
Since eSHMT is a dimer, FRET could occur between Trp residues in one subunit and the PyP in the other subunit. To determine whether this was the case, a 1:1 mixture of the Trp Ϫ mutant PyP-eSHMT and wild-type apo-eSHMT were refolded in the same solution to form three possible types of dimers, i.e. homodimers of both wild-type apo-eSHMT and Trp Ϫ PyP-eSHMT and a hybrid dimer containing a subunit of wild-type apo-eSHMT and a subunit of Trp Ϫ PyP-eSHMT. Neither the wild-type apo-eSHMT nor the Trp Ϫ PyP-eSHMT homodimers will exhibit FRET since neither contains both the Trp residue as energy donor and the PyP as the energy acceptor. The hybrid dimer will exhibit FRET only if the Trp residues in the wildtype apo-eSHMT subunit can transfer energy to the PyP moiety in the Trp Ϫ subunit. To determine whether hybrid dimers had formed the refolded mixture of wild-type apo-eSHMT and Trp Ϫ PyP-eSHMT was analyzed by native gel electrophoresis (see ''Experimental Procedures''). The homodimer of Trp Ϫ PyP-   eSHMT migrated faster than the homodimer of wild-type apo-eSHMT because of the extra negative charge on the PLP phosphate group. In the refolded mixture of apo-eSHMT and PyP-eSHMT, a band that ran intermediate to the two homodimers was observed in native gel electrophoresis, showing that hybrid dimers had formed (data not shown). No FRET was observed in the solution containing hybrid dimers. This suggests that all FRET observed in the samples recorded in Fig. 4 occurs between the Trp and PyP residues in the same subunit.
The distance between each Trp residue and the active site PyP can be determined from both fluorescent lifetime measurements of Trp residues and by donor quenching (5)(6)(7)(8). The fluorescent lifetimes for each single Trp mutant eSHMT were determined in the apoenzyme form ( D ) and in PyP-eSHMT ( DA ). The changes in lifetime were used to calculate the energy transfer efficiency between each Trp residue and the enzymebound PyP. The transfer efficiencies were also calculated by the amount of donor (Trp) fluorescence quenching as described under ''Experimental Procedures.'' Distance calculations for Trp 16 and Trp 385 PyP-eSHMTs were in good agreement by these two methods. However, for Trp 183 PyP-eSHMT, the fluorescence lifetime measurements have a large error because of the very low signal as a result of the quenching of this Trp fluorescence by the bound PyP, making a distance measurement difficult. Therefore, this method was not used for this mutant. Table I records the distance calculations only by the donor quenching method. The results show that Trp 183 is the closest to the active site PyP and Trp 385 is the most distant. These distances (R 2/3 ) have an error due to the lack of knowledge about the orientations of Trp and PyP in the native enzymes. The results of anisotropy measurement were used to calculate the maximum and minimum distances between donor (Trp) and receptor (PyP) to account for this uncertainty of the orientation factor (13). Table I records the maximum and minimum distance between each Trp residue and the PyP. However, the accuracy of the absolute distance is not so important in this study.
Equilibrium Unfolding-Refolding Studies-Previous studies monitored the environment of each Trp residue for the apo-eSHMTs under equilibrium unfolding-refolding conditions (4). It was observed that Trp 16 underwent local unfolding at a lower concentration of urea than either Trp 183 or Trp 385 , which unfolded at similar urea concentrations (shown as dotted lines in Fig. 5). Evidence was obtained for a three state model (Equation 5) for both Trp 16 and wild-type apo-eSHMTs, in which the intermediate I e was the dominant species at 2.1 M urea (4). In this intermediate, Trp 16 was largely solvent-exposed, whereas Trp 183 and Trp 385 were still mostly buried. The transitions for wild-type apo-eSHMT of N to I e and I e to U had similar ⌬G°v alues of 4.1 and 4.4 kcal/mol, respectively (4). Trp 183 and Trp 385 eSHMTs had equilibrium profiles that were fit by a two-state model since both of these Trp residues had similar fluorescent properties in the intermediate and the native states. A urea gradient gel of apo-eSHMT showed that the unfolded enzyme in 8 M urea migrated more slowly than the native dimer. Between 1.5 and 3 M urea, a second band that moved considerably more slowly was observed where I e was most stable. The urea gradient gel and size exclusion chromatography results were interpreted to show that in I e , the enzyme was still a partially unfolded dimer with an increased Stokes radius. No evidence for a compact monomeric form of apo-eSHMT could be seen. In this study, both the equilibrium experiments and the urea gradient gel were repeated with the PyP-eSHMTs. Both the environment of each Trp residue and its proximity to the PyP group can be monitored by observing the fluorescence emission at 335 (Trp) and 380 (PyP) nm, respectively. As shown in Fig.  5, Trp 16 PyP-eSHMT undergoes equilibrium unfolding-refolding with the changes in fluorescence at 335 and 380 nm paralleling each other and a midpoint in the 3-3.2 M range of urea. This is compared to a midpoint at about 1.6 M for the Trp 16 apoenzyme (dotted line, Fig. 5) (4). The transitions for the fluorescence changes at both 335 and 380 nm fit a two-state model as described by the solid lines. The values for ⌬G°c alculated from the two-state fit for the 335 and 380 nm curves gave 3.3 and 3.7 kcal/mol for unfolding, respectively (Table II). Likewise, the fluorescence changes at 335 and 380 nm for Trp 385 PyP-eSHMT parallel each other and fit a two-state transition model with a midpoint at 3.5 M urea. The ⌬G°values are slightly higher than for Trp 16 PyP-eSHMT, with values of 4.2 and 4.0 kcal/mol for the transitions monitored at 335 and 380 nm, respectively (Table II). These results suggest that in PyP-eSHMT, Trp 16 and Trp 385 become solvent-exposed (fluorescence at 335 nm) under the same conditions as they are  PyP-eSHMT, the reduced PyP quenches the fluorescence of Trp 183 , resulting in no increase of fluorescence in the unfolded state compared to the native state at its native fluorescence maximum of 323 nm (Fig. 4). Even at 335 nm, the difference in fluorescence emission between the unfolded and native state is small, with the unfolded state having the greater value. This small change in signal between the unfolded and native states makes it difficult to calculate the fraction folded from the fluorescence of Trp 183 at 335 nm, so only the FRET fluorescence at 380 nm is shown in Fig. 5 (open circles). The fraction folded as a function of urea concentration is shown in the upper right panel in Fig. 5 with the solid line being the fit for a three-state model, suggesting that the equilibrium involves N, I e , and U states (Equation 5). The ⌬G°value for N to I e is 3.6 kcal/mol, and the value for the I e to U transition is 14.3 kcal/mol (Table  II). Fluorescence emission as a function of urea concentration for the wild-type PyP-eSHMT is shown in the last panel in Fig.  5. The fluorescence emission pattern at 335 nm follows a different path from that of fluorescence at 380 nm. However, both are fit by a three-state model, as shown by the solid lines. The N to I e transitions have ⌬G°values of 6.9 and 7.0 kcal/mol as determined from the fluorescence at 335 and 380 nm, respectively (Table II). The I e to U transitions have ⌬G°values of 14 and 13.3 kcal/mol at 335 and 380 nm, respectively (Table II).
Using the equilibrium unfolding pattern of fluorescence at 380 nm for wild-type PyP-eSHMT, a graph of how the relative concentration of N, I e , and U vary with urea concentration was constructed (Fig. 6). The results show that at 5.5 M urea I e is almost the only species present. In I e , both Trp 16 and Trp 385 are solvent-exposed and separated by more than 50 Å from the PyP, but domain 1 containing Trp 183 remains folded and in close contact with the Lys 229 -PyP group. The closeness of Trp 183 and the PyP in I e is also supported by the changes in fluorescence at 335 nm for Trp 183 PyP-eSHMT with increasing urea. As shown in Fig. 4 for Trp 183 , the unfolded state has slightly greater fluorescence at 335 nm than the native enzyme. This is the result of two opposing factors. Unfolding results in a decrease in the Trp fluorescence as the buried Trp residue is exposed to a hydrophilic solvent. In opposition, unfolding increases fluorescence by decreasing its quenching as the PyP group becomes more distant. There is essentially no change in either the fluorescence quenching of Trp 183 or exposure to solvent during the transition of N to I e (from 0 to 4.5 M urea). Both the decrease in quenching and increased exposure to a hydrophilic environment occur with the same magnitude during the I e to U transition at 323 nm (4.5-8 M urea) (data not shown).
The equilibrium unfolding-refolding curve, monitored by both FRET at 380 nm and Trp fluorescence at 335 nm (Fig. 5), was performed between 0.6 and 6 M enzyme. No concentration effect was observed in these curves for any of the mutant enzymes or the wild-type enzyme (data not shown). As shown by a large increase in mobility of PyP-eSHMT in a urea gradient gel (upper panel in Fig. 6), it is likely that a monomer-dimer equilibrium exists in the 3-5 M urea range. The fact that no concentration effect was observed in either Trp fluorescence or FRET in the equilibrium unfolding studies suggests that FRET is the same in both the monomer and dimer forms of the enzyme. If some FRET had occurred between subunits, then one should observe increased FRET with increasing enzyme concentration. This is further support for the conclusion that all FRET occurs between the three Trp residues and PyP in the same subunit.
The urea gradient gel for PyP-eSHMT (Fig. 6, upper panel) is very different from the one previously shown for apo-eSHMT (4). In Fig. 6, the gel shows that in the region where N is starting to form I e , the enzyme migrates significantly faster than the native state. This suggests that the enzyme has dissociated into a monomer but is still compact. As the amino terminus and domain 2 unfold, this trend is reversed, and the protein migrates more slowly, indicating a larger Stokes radius. At higher urea concentrations, corresponding to the I e to U transition, the protein migrates more slowly, reaching the fully unfolded state at 8 M urea. The two bands seen in the 5-7 M urea range suggest that the equilibrium between the fully unfolded state and other partially unfolded states is slower than the 2 h required to run the gel.
Kinetics of Refolding-The rates of change in Trp fluorescence and FRET were monitored with time after a 10-fold dilution of unfolded PyP-eSHMTs in 8 M urea into Tris buffer, pH 7.5. Fig. 7 shows a time course of the fluorescence emission spectra at 4°C for the Trp 183 PyP-eSHMT. The results show that within the first 15 s after initiating refolding, there is a blue shift in the emission maximum from 352 to 323 nm and a large increase in fluorescence intensity at this wavelength (Fig.  7, dashed lines). The spectrum of this 15 s curve is essentially the same as the spectrum of the native apo-eSHMT as reported previously (4). There is very little increase in fluorescence at 380 nm at 15 s, suggesting that there is little FRET between Trp 183 and Lys 229 -PyP in the initially formed intermediate. With time, the fluorescence at 335 nm decreases, with a concomitant increase in the fluorescence peak at 380 nm associated with FRET (Fig. 7). Note that the fluorescence at 323 nm of the native and unfolded states is the same for Trp 183 PyP-eSHMT. The experiment with Trp 385 PyP-eSHMT gave similar results (data not shown). These results are in accord with our previous proposal that domains 1 and 2 fold rapidly to bury Trp 183 and Trp 385 .
A different pattern was observed for Trp 16 PyP-eSHMT at 4°C. There was an immediate blue shift after initiation of refolding but not a large increase in fluorescence (data not shown). The fluorescence intensity returned more slowly and did not reach 50% of its native value until about 1600 s. This was also observed in studies on the rate of refolding of Trp 16 apo-eSHMT (4). These results show that the amino-terminal residues are not in their native environment in the first stable kinetic intermediate. A model of this first kinetic intermediate is referred to as I k in Scheme 1.
The measurement of FRET at 380 nm during refolding was repeated for each single Trp mutant and wild-type PyP-eSHMT at 30°C. The appearance of FRET at 380 nm is plotted versus time in Fig. 8, as is the decrease in fluorescence for Trp 183 PyP-eSHMT at 323 nm (quenching). Each curve is fit by a single exponential rate (solid line), and the first order rate constants are given on each curve. The results show that all three Trp residues approach the active site PyP at the same rate. The first order nature of each curve suggests that there is a single structure for the kinetic intermediate (I k ) that goes to the native state (as defined by FRET) by a single first order process. This two-state conversion of I k to the native state is also supported by the isosbestic point at 355 nm, as recorded in the spectra in Fig. 7.
The rate of appearance of FRET at 380 nm was studied with Trp 183 PyP-eSHMT at temperatures between 4 and 30°C. The ln of the first order rate constant is plotted against the reciprocal temperature (Arrhenius plot) in degrees Kelvin in Fig. 9. The slope of the straight line was used to calculate a ⌬H ‡ of 22.5 kcal/mol. At 25°C, T⌬S ‡ is 2 kcal/mol.
Fluorescence Anisotropy of Trp Residues and PyP during Refolding-The kinetic studies on the rate of appearance of Trp fluorescence at 335 nm suggests that a kinetic intermediate (I k ) FIG. 8. Rate of appearance of FRET during refolding for single Trp mutants and wild-type PyP-eSHMTs at 30°C. The experiments were performed as described in Fig. 7 except that the temperature was at 30°C. The fluorescence intensity at 380 nm represents the amount of FRET, and the fluorescence at 323 nm represents the quenching of Trp 183 . The solid lines are the fit to a first order reaction. A zero time control could not be determined because of the large and rapid blue shift in fluorescence that occurs within a few seconds when unfolded enzyme is diluted 10-fold. The numbers on each curve are the first order rate constants in reciprocal seconds. is formed in which both Trp 183 and Trp 385 are buried in a hydrophobic environment. Trp 16 is also in a relatively hydrophobic environment, but the intensity suggests that it is not in its native state. The results of kinetic refolding monitored by FRET (Figs. 7 and 8) also show that all three Trp residues are more than 50 Å from the PyP attached to Lys 229 after the initial hydrophobic collapse. This suggests that the peptide chain between domains 1 and 2 that contains Lys 229 -PyP may still be free and disordered in solvent. To determine the nature of the environment of the PyP in I k , we measured its anisotropy every 30 s during refolding at 4°C (Fig. 10). The anisotropy of the PyP in the wild-type PyP-eSHMT in 0.8 M urea is 0.351, and in the unfolded state in 8 M urea it is 0.064. This change in values between the native and unfolded states represents the change in rotational freedom of the Lys 229 -PyP and not a solvent effect since free Lys-PyP has anisotropy values of 0.031 and 0.036 in 0.8 and 8 M urea, respectively. The anisotropy results show that after initiating refolding by the 10-fold dilution of unfolded enzyme in 8 M urea, about 70% of the difference in unfolded and folded state anisotropy has been regained in the 15 s it takes to record the first value (Fig. 10). During the next 1000 s, more than 95% of the anisotropy of the native state has been regained. The regain of the last 30% of anisotropy is not first order or second order, but is best fit by the equation for two parallel first order reactions (solid line) with rate constants of 0.017 and 0.0024 s Ϫ1 and relative amplitudes of 0.050 and 0.064, respectively. This suggests that there are at least two forms of the enzyme involved in this kinetic process.
During the initial few min after initiating folding, little FRET has occurred (Fig. 7), but more than 70% of the native anisotropy of PyP has been regained (Fig. 10). These results show that during the formation of the kinetic intermediate I k , the PyP attached to Lys 229 has become partially immobilized, but is still not in its native environment, since the PyP remains distant from any of the Trp residues.
Fluorescence anisotropy was also determined for each Trp residue in the single Trp apo-eSHMTs and single Trp PyP-eSHMTs. The anisotropy values changed from 0.03-0.04 in the unfolded enzymes to 0.11-0.14 in their native states. For each Trp, including Trp 16 , anisotropy had returned to the native state value within 30 s after initiation of refolding (data not shown). (Equation 1) from previous studies on apo-eSHMT (3,4) is shown in Scheme 1. This model showed that domains 1 and 2 fold in less than 10 s at 4°C, but the amino-terminal 55 residues and residues between 225 and 276 remain disordered and sensitive to digestion with subtilisin. The previous studies also suggested that M could form a dimer (DЈ) that cannot bind PLP. Because there was no fluo-rescent probe in the section between residues 225 and 276, the only available evidence as to its structure in M and DЈ was its sensitivity to digestion with subtilisin. This segment of the amino acid sequence is important because it contains not only the active site Lys 229 , which binds PLP, but several other residues that have been shown to be involved in catalysis (20,21). Reducing the enzyme-bound PLP with NaCNBH 3 forms a stable secondary amine that absorbs at 335 nm. This mild procedure now places a fluorescent probe in the critical active site region and serves as a fluorescent acceptor to receive energy transferred from Trp residues in FRET experiments. Unexpectedly, the reduction of the PLP aldimine resulted in a significant increase in stability of the enzyme, suggesting that the PyP-eSHMT may have a different conformation. A similar conclusion was reached with the ␤ 2 subunit of E. coli tryptophan synthase, where the reduction of the coenzyme resulted in a significant decrease in immunoreactivity (19).

Kinetic Properties of Refolding of PyP-eSHMT-A model of a kinetic folding intermediate M
Reduction of the coenzyme did not significantly change the basic folding mechanism as determined by the kinetic properties during refolding of PyP-eSHMT. The protease digestion pattern remains the same (Fig. 2). Trp 183 and Trp 385 are rapidly buried in a hydrophobic environment, and Trp 16 does not reach its native state of fluorescence until the final step in the folding pathway. After the first few seconds of refolding, the intermediate I k of PyP-eSHMT has the same structural properties as the kinetic intermediate proposed for apo-eSHMT (Scheme 1) (4). The evidence supporting this view is summarized in Table III, where the kinetic properties (time to reach 50% of the value between unfolded and native state) were monitored by Trp fluorescence, FRET between Trp residues and Lys 229 -bound PyP, anisotropy, and regain of the proteaseresistant form during the refolding of PyP-eSHMTs.
This study thus adds to our understanding of the structure of M and DЈ (Equation 1). Fluorescence anisotropy measurements show that Trp 16 is partially immobilized but not completely buried in a native-like environment during the first few seconds of refolding. Also, the anisotropy studies showed that the PyP group attached to Lys 229 is mostly immobilized in the first few seconds and approaches the rotational freedom of the native state at 4°C in a 20-min period. Fully folded enzyme is not present after this 20 min period as indicated by only a small amount of protease-resistant form of the enzyme (Fig. 2) and low FRET at 380 nm (Fig. 7).
Most importantly, the studies by FRET at 380 nm show that in M, and probably also in DЈ, all three Trp residues are more than 50 Å from the Lys 229 -PyP moiety. This provides direct evidence that the rate-determining step of folding brings a loop region into proximity of all three Trp residues, as found in the native state. In the previous study, we used size exclusion chromatography to determine the oligomeric state during refolding at 4°C (3). A brief study was done with the PyP- eSHMT, and it gave similar results, suggesting that a structure equivalent to DЈ is formed. However, the rate of forming a dimer may be too fast to be determined by the slow chromatography method. We are currently trying other methods to determine the rate of dimer formation. The refolding studies also show that the rate of the final folding step for PyP-eSHMT is much slower at 4°C than at 30°C, as observed with apo-eSHMT. The studies at both temperatures suggested a common rate-determining step in folding. With PyP-eSHMT, this rate appears to be slightly faster than with the apo-eSHMT. A major question is why the formation of the active site is so slow. From the temperature dependence on the rate of appearance of FRET (Fig. 9), it is concluded that this step has an energy of activation of 22.5 kcal/mol. This is not much different from the 20 kcal/mol determined for the isomerization of proline residues from the trans to cis configurations (22). As we noted previously, the segment from residues 225 to 276 contains eight Pro residues. Each boundary of this segment contains two or three proline residues. If one or more of these Pro residues is in the cis configuration in the native state it could account for the rate-determining step and the temperature dependence. Using FRET between Trp 177 and the reduced PyP on Lys 87 in tryptophan synthase, Blond and Goldberg (9) showed that the energy of activation for bringing these two groups together was also about 20 kcal/mol. Leistler et al. (23) have shown that the refolding of aspartate aminotransferase is characterized by an energy of activation in the 20 kcal/mol range.
Equilibrium Unfolding-Refolding-The equilibrium unfolding studies on PyP-eSHMT (Figs. 5 and 6) again suggest the presence of an intermediate that we refer to as I e . However, I e has quite different structural features compared to the equilibrium intermediate previously observed with apo-eSHMT (4). With wild-type apo-eSHMT, the intermediate was the dominant species at 2.1 M urea, but with the PyP-eSHMTs it is the dominant species at 5.5 M urea (Fig. 6). In I e of PyP-eSHMT, Trp 16 and Trp 385 are both solvent-exposed and greater than 50 Å from PyP, whereas Trp 183 retains its native state fluorescence and the PyP is still close to it. These results suggest that in the I e of PyP-eSHMT, the first 55 residues and the carboxylterminal domain have both unfolded, whereas domain 1 and the active site region remain compact and native-like. An additional support for this view is that the quenching of Trp 183 remains high until domain 1 unfolds. The structure of I e of PyP-eSHMT, as depicted in Scheme 2, differs mostly from the structure of the equilibrium intermediate of apo-eSHMT (same as I k in Scheme 1) by the Lys 229 -PyP group remaining in close contact with domain 1. There is also a significant increase in the stability of domain 1. The stability of domain 2 and the amino-terminal residues are not significantly changed in I e of PyP-eSHMT compared to the equilibrium intermediate of apo-eSHMT. The midpoint for unfolding with urea concentration has increased for the Trp 16 and Trp 385 PyP-eSHMTs compared to their counterparts with apo-eSHMT (Fig. 5, dashed lines), but when the data are extrapolated to zero urea concentration, both the amino-terminal residues, as monitored by Trp 16 , and domain 2, as monitored by Trp 385 , have only slightly lower stability in the PyP-eSHMTs compared to their apoenzyme counterparts (Table II in this study and Table II in Ref. 4). For each of these two regions, the ⌬G°has decreased by 1 to 2 kcal/mol in PyP-eSHMT. However, the midpoint for unfolding of domain 1, as determined by the fluorescent properties of Trp 183 PyP-eSHMT, has increased from about 2.5 M urea for Trp 183 apo-eSHMT to about 6 M urea for the PyP-eSHMT. The ⌬G°for unfolding of domain 1 and the active site PyP is about 14 kcal/mol compared to about 5 kcal/mol for domain 1 in the apoenzyme (Table II in this paper and Table II in Ref. 4). Apparently the reduced PLP has strong interactions with residues in domain 1. In our previous study it was concluded that in the equilibrium denaturation of the holo-enzyme, the first event was the loss of the PLP to form apoenzyme between 1 and 2 M urea (3). By reduction of the PLP, we have lost this first step and the PyP remains bound at the active site. This provides a possible explanation of why domain 1 was stabilized in PyP-eSHMT.
In the study of apo-eSHMT, it was concluded that I e is a dimer. This was supported not only by size exclusion chromatography experiments but by a urea gradient gel that showed decreased mobility for the apoenzyme in the 2-3 M urea range that was interpreted as being caused by a partially unfolded dimer (4). The urea gradient gel for PyP-eSHMT shows very different results (Fig. 6) compared to the urea gradient gel from the apoenzyme studies. It appears that I e in the 4 M urea range has significantly greater mobility then the native enzyme, even though both the amino-terminal region and domain 2 have been partially unfolded. This suggests that the enzyme is becoming a monomer and that at 5.5 M urea the I e of PyP-eSHMT is a monomer. The I e of PyP-eSHMT is not observed on the kinetic folding pathway. Since PLP adds to the refolding enzyme only after it has passed through the rate-determining step, it is unlikely that the structure shown in Scheme 2 is on the kinetic pathway for folding of this enzyme. Herold et al. (24) have characterized a monomeric fragment of aspartate aminotransferase that folds independently and binds PLP. This fragment has some sequence identity and is of similar size to domain 1 plus the active site region of eSHMT, shown as I e in Scheme 2 (25,26). It will be of interest to determine whether the domain shown as I e in Scheme 2 can also fold independently and bind PLP as a monomer as observed with aspartate aminotransferase.