Dedication: This work is dedicated to the memory of Professor Franco Tatò.
R. Contestabile, Dipartimento di Scienze Biochimiche, ‘A. Rossi Fanelli’, Università‘La Sapienza’, Piazzale Aldo Moro 5, 00185 Roma, Italy. Fax:+39 06 49917566, Tel.:+39 06 49917569, E-mail: firstname.lastname@example.org
Serine hydroxymethyltransferase (SHMT) is a member of the fold type I family of vitamin B6-dependent enzymes, a group of evolutionarily related proteins that share the same overall fold. The reaction catalysed by SHMT, the transfer of Cβ of serine to tetrahydropteroylglutamate (H4PteGlu), represents in the cell an important link between the breakdown of amino acids and the metabolism of folates. In the absence of H4PteGlu and when presented with appropriate substrate analogues, SHMT shows a broad range of reaction specificity, being able to catalyse at appreciable rates retroaldol cleavage, racemase, aminotransferase and decarboxylase reactions. This apparent lack of specificity is probably a consequence of the particular catalytic apparatus evolved by SHMT. An interesting question is whether other fold type I members that normally catalyse the reactions which for SHMT could be considered as ‘forced errors’, may be close relatives of this enzyme and have a catalytic apparatus with the same basic features. As shown in this study, l-threonine aldolase from Escherichia coli is able to catalyse the same range of reactions catalysed by SHMT, with the exception of the serine hydroxymethyltransferase reaction. This observation strongly suggests that SHMT and l-threonine aldolase are closely related enzymes specialized for different functions. An evolutionary analysis of the fold type I enzymes revealed that SHMT and l-threonine aldolase may actually belong to a subgroup of closely related proteins; fungal alanine racemase, an extremely close relative of l-threonine aldolase, also appears to be a member of the same subgroup. The construction of three-dimensional homology models of l-threonine aldolase from E. coli and alanine racemase from Cochliobolus carbonum, and their comparison with the SHMT crystal structure, indicated how the tetrahydrofolate binding site might have evolved and offered a starting point for further investigations.
During the past decade, the understanding of the evolutionary relationships amongst vitamin B6-dependent enzymes and of the structural basis of their mechanistic diversity and uniformity have increased markedly [1,2]. The progress in this field has been greatly aided by the raising numbers of sequences and three-dimensional structures that have become available. Although there are five evolutionarily unrelated families of B6 enzymes, each having a completely different fold [3–5], the whole range of diverse reaction specificity  is covered by the largest and best characterized family known as the α family , fold type I , or the aspartate aminotransferase family . This family therefore offers an opportunity to understand how distinct catalytic properties that exploit the chemical reactivity of a single coenzyme and a common protein scaffold have evolved. One member of this family, serine hydroxymethyltransferase (SHMT), whose crystallographic structure has been determined from several sources [7–10], shows a particularly broad reaction specificity. Although there is little doubt that the role of SHMT in vivo is to catalyse the reversible transfer of Cβ of serine to tetrahydropteroylglutamate (H4PteGlu) to form glycine and 5,10-methylene-H4PteGlu, in vitro and in the absence of H4PteGlu, SHMT also catalyses decarboxylation, transamination, retroaldol cleavage and racemization reactions (Table 1), at rates sometimes approaching and even exceeding those of the physiological reaction [11,12]. This makes SHMT and the related enzymes that catalyse similar reactions a good system for revealing the structural factors controlling reaction specificity. The particular reaction catalysed by SHMT is mainly determined by the structure of the amino-acid substrate. With the true substrates, serine or glycine, SHMT catalyses none of the alternate reactions. The currently accepted model attributes this reaction specificity to the existence of open and closed active site conformations. The physiological substrates generate the closed conformation, whereas alternate substrates react while the enzyme remains in open conformation, which permits reaction paths leading to decarboxylation, transamination and racemization .
Table 1. Reactions catalysed by serine hydroxymethyltransferase.
l-allo-threonine or l-threonine ⇔ glycine + acetaldehyde
l-erythro- or l-threo-β-phenylserine ⇔ glycine + benzaldehyde
[2-3H]glycine + H2O ⇔ glycine +3H2O
d- or l- alanine + PLP ⇔ pyruvate + PMP
aminomalonate ⇔ glycine + CO2
Threonine aldolase activity detected in cell extracts has long been attributed to SHMT because it had been observed when pure cytosolic SHMT from rabbit liver and threonine were mixed in vitro. Recently, it has been proved that in some microbial organisms threonine aldolase is a distinct enzyme . However, although SHMT from some mammalian livers slowly cleaves threonine, there is no threonine aldolase activity in rat liver  and mammals are believed to lack the genuine threonine aldolase enzyme. Several orthologous genes encoding threonine adolase from various microbial sources have been cloned and expressed . Their protein products are capable of cleaving both l-threonine and l-allo-threonine (and in this case are called low-specificity l-threonine aldolases; l-TAs) or are highly specific for l-allo-threonine (l-allo-TAs). It has been also recognized that l-TAs belong to the fold type I family [1,15]. Moreover, in recent years, a vitamin B6-dependent alanine racemase from two different fungal organisms was isolated that shows a strong sequence similarity to l-TAs from many species [17–19]. Fungal alanine racemases are unrelated at the level of the primary structure to any known bacterial alanine racemase (fold type III) . We became interested in l-TA and fungal alanine racemase because they normally catalyse reactions which for SHMT may be considered as ‘forced errors’. A comparison of these enzymes with SHMT may give important clues to the origin of the reaction and substrate specificity in fold type I enzymes. In the study reported here, we provide evidence that the E. colil-TA is a close relative of E. coli SHMT, exhibiting very similar spectral and catalytic properties. A theoretical study, based on sequence comparison, homology modelling and structural analysis, supports this interpretation and tentatively extends it to fungal alanine racemase.
It is clear that SHMT, alone amongst the members of the fold type I family, has acquired a folate binding site. Some discussion on this aspect has been made between Escherichia coli SHMT and aspartate aminotransferase . However, these two members of the fold type I family are too distantly related to offer a meaningful comparison. Our comparative study of SHMT and the closely related l-threonine aldolase and fungal alanine racemase was undertaken in the hope that it would reveal how evolution created the folate site and how alternative catalytic activities were repressed or enhanced.
Materials and methods
Ingredients for bacterial growth were from Difco. Chemicals for the purification of the enzymes were from BDH, DEAE-Sepharose and phenyl-Sepharose from Pharmacia Biotech. l-Lactic dehydrogenase, alcohol dehydrogenase, l-alanine dehydrogenase and d-amino-acid oxidase were from Sigma–Aldrich. SHMT from E. coli was expressed and purified as previously described . The oliglonucleotides primers used in the PCR reaction were from Integrated DNA Technologies, Inc. (Coralville, IA, USA). (6S)-H4PteGlu was a gift from A. G. Eprova, Schaffhausen, Switzerland. All other reagents were from Sigma–Aldrich.
Cloning of the gene encoding threonine aldolase
The coding sequence of the ltaE gene, encoding the E. coli low-specificity l-threonine aldolase (eTA) , was obtained from the genomic DNA of the E. coli K-12 strain by PCR amplification, using an Expand High Fidelity PCR System (Roche Diagnostics Corporation, Indianapolis, IN, USA) and the following oligonucleotides were used as primers: upstream primer, 5′-AGGACATCATATGATTGATTTACGCAGTGA-3′; downstream primer, 5′-CGTCTGAATTCTTAACGCGCCAGGAATG-3′. NdeI and EcoRI restriction sites, underlined in the sequences, were introduced to allow insertion into a pET22b(+) plasmid vector (Novagen, Inc., Madison, WI, USA). This construct was used to transform E. coli HMS174 (DE3) strain cells (Novagen, Inc.). The nucleotide sequence of the insert was determined to confirm that no mismatching had occurred during the PCR amplification.
Purification of threonine aldolase
An overnight culture (30 mL) of HMS174 (DE3) cells, transformed with the l-threonine aldolase overexpressing plasmid, was inoculated (1 : 200) into 6 L of Luria–Bertani broth containing ampicillin (100 mg·L−1) and vitamin B6 (30 mg·L−1). Bacteria were grown aerobically at 37 °C to exponential phase (D600 = 0.3–0.4), then expression of eTA was induced with 0.05 mm isopropyl thio-β-d-galactoside (IPTG). Bacteria were harvested after 20 h and suspended in 10 mm Tris/HCl, 1 mm NaCl, 1 mm EDTA, pH 7.6. Cell lysis was carried out by the addition of 1 mg of lysozyme per g of packed bacterial cells and incubation for 20 min at room temperature. Streptomycin sulfate (10 g·L−1) was added to precipitate DNA. The cell extract was then centrifuged at 15 000 g for 30 min and the pellet was discarded. Solid ammonium sulfate was added to the cell extract to 50% of saturation (313 g·L−1). The solution was centrifuged at 15 000 g for 20 min and the pellet was discarded. Ammonium sulfate was added to the supernatant to 75% saturation (176 g·L−1). After centrifugation, the precipitate was dissolved in 20 mm potassium phosphate, pH 7.5, and dialysed against two 3-L changes of the same buffer. The dialysed material was added to a DEAE-Sepharose column (5 × 15 cm) that had been equilibrated with 20 mm potassium phosphate, pH 7.5. The column was washed with 100 mL of the equilibrating buffer. The enzyme was then eluted with a 1-L linear gradient of 0–0.4 m NaCl in 20 mm phosphate buffer, pH 7.5. The fractions with the highest A420/A280 ratio were collected and the protein was precipitated by the addition of ammonium sulfate to 75% saturation. After centrifugation, the pellet was dissolved in an equal volume of 50 mm potassium phosphate, pH 7.0, and loaded on a phenyl-Sepharose column (4 × 15 cm) equilibrated with the same buffer containing ammonium sulfate to 30% saturation (176 g·L−1). Elution was carried out with a 600-mL linear gradient from the equilibrating buffer to 20 mm potassium phosphate, pH 7.0. The fractions showing a single band on SDS/PAGE were pooled and concentrated by ammonium sulfate precipitation, followed by dialysis in 20 mm potassium phosphate at pH 7.0.
Spectral and kinetic studies
With serine and H4PteGlu as substrates, the rate of 5,10-methylene-H4PteGlu production was determined by oxidizing this compound to 5,10-methenyl-H4PteGlu, using NADP+ and methylenetetrahydrofolate dehydrogenase. The initial rate of the reaction, carried out in 20 mm potassium phosphate pH 7.2 at 30 °C, was calculated from the absorbance change at 340 nm due to NADPH formation, using a value of ε340 = 7200 m−1·cm−1. The reaction mixture included either eSHMT (0.1 µm) or eTA (34 µm). The dependence of the initial velocity of the reaction on H4PteGlu concentration was determined maintaining l-serine at 30 mm and varying H4PteGlu concentration between 5 µm and 450 µm. The formaldehyde produced from the retroaldol cleavage of l-serine catalysed by eTA was measured incubating the enzyme with 150 mm l-serine for a definite period of time and then adding 500 µm H4PteGlu, 0.2 mm NADP+ and methylenetetrahydrofolate dehydrogenase (200 µg·mL−1) to the reaction mixture. The spontaneous condensation between formaldehyde and the oxidation of the product 5,10-CH2-H4PteGlu by NADP+, resulted in a burst of absorbance at 340 nm, due to formation of NADPH, from which the formaldehyde produced during the incubation with eTA was measured. Km and kcat values for l-serine cleavage were determined using the standard assay for SHMT, holding H4PteGlu concentration constant at 500 µm and varying l-serine concentration from 0.33 mm to 90 mm. Under these conditions, the spontaneous reaction of formaldehyde with H4PteGlu was not rate-limiting.
The rate of threonine cleavage was measured by coupling the reaction with reduction of the product acetaldehyde by NADH and alcohol dehydrogenase . The rate of the reaction was calculated from the rate of disappearance in absorbance at 340 nm, using a value of ε340=6220 cm−1·m−1. Benzaldehyde production from phenylserine cleavage was measured spectrophotometrically at 279 nm, employing a molar absorption value of ε279=1400 cm−1·m−1. Both retroaldol cleavage reactions were carried out in 20 mm potassium phosphate, pH 7.0, at 30 °C.
The rate of exchange of the α-protons of glycine with solvent protons was determined by incubating 0.03 mm[2-3H]glycine (2.2×106 c.p.m.) with eTA (5–30 µm) in 20 mm potassium phosphate, pH 7.2, at 30 °C and determining the amount of 3H2O formed in a 15-min incubation . The effect of H4PteGlu on the rate of exchange was determined by adding 0.5 mm H4PteGlu to the reaction solution.
The pseudo first-order rate constants of transamination of d- and l-alanine were determined by measuring the disappearance of absorbance at either 498 nm or 420 nm, during the conversion of the enzyme-bound pyridoxal-5′-phosphate (PLP) to pyridoxamine-5′-phosphate (PMP) . Each reaction was carried out in 50 mm sodium N,N-bis(2-hydroxyethyl)-2-aminoethanesulfonate (NaBES), pH 7.6, at 37 °C and contained 37 µmeTA and 210 mm alanine. After completion of the reaction with d-alanine, the enzyme was separated from the small molecules using a 10-kDa cut-off centrifugal concentrator (MicrosepTM, Filtron Technology Corporation, Northborough, MA, USA). An aliquot of the filtrate was used to measure the pyruvate produced in the transamination reaction by reducing it to lactate using NADH and lactate dehydrogenase. The pyruvate concentration was calculated from the change in absorbance at 340 nm. Other aliquots of the same sample were used for the spectrophotometric identification of the PMP produced in the transamination reaction and released from the enzyme. Absorption spectra were measured in either 0.1 m NaOH or 0.1 m HCl. The apparent Kd for both alanine enantiomers were determined by titrating enzyme with increasing concentrations of alanine and determining the maximum absorbance at 498 nm for the formation of the quinonoid intermediate. The Kd was determined from a best fit of the values of ΔA498 to Eqn (2).
Racemization reactions of d- and l-alanine were carried out in 50 mm NaBES, pH 7.6, at 37 °C. The reaction mixture contained 30 µm enzyme, 200 mm d- or l-alanine and 1 mm PLP in a volume of 0.5 mL. A control reaction contained no enzyme. At various time intervals, 45-µL aliquots of the reaction mixture were removed and the reaction was stopped by the addition of 160 mm HClO4. The solution was neutralized by adding an equivalent amount of KOH and centrifuged to remove the precipitated protein and KClO4. The sample was then assayed for either d- or l-alanine. The assay for l-alanine consisted of 10 mm NAD+, 0.2 m hydrazine, sample and 5 U of l-alanine dehydrogenase in 100 mm sodium borate pH 9.5. The change in absorbance at 340 nm due to reduction of NAD+ was used to calculate the concentration of l-alanine produced. For the l→d direction, d-amino-acid oxidase and lactate dehydrogenase were used as the coupling enzymes. In the assay, the sample deriving from the reaction mixture was mixed with 0.2 mm NADH and 5 U of lactate dehydrogenase in 20 mm NaBES pH 7.0. Under these conditions, the pyruvate produced from alanine in the transamination reaction catalysed by eTA was converted into lactate. Addition of 1.5 U of d-amino-acid oxidase then converted the d-alanine produced in the racemization reaction into pyruvate and the latter into lactate with the simultaneous consumption of NADH. The concentration of d-alanine in the sample was calculated from the decrease in absorbance at 340 nm observed after addition of d-amino-acid oxidase. All enzymes used in the assays were dialysed against either 100 mm sodium borate, pH 9.5, or 20 mm NaBES, pH 7.0, and mixed with 50% glycerol before use.
All spectral and kinetic studies were carried out on a Hewlett–Packard 8452 A diode array spectrophotometer. Kinetic data analysis, curve-fitting procedures and statistical analysis were performed using the data manipulation software of Scientist (micromath, Salt Lake City, USA). The following equations were used to fit the data:
Multiple sequence alignment and evolutionary analysis
Sixteen l-TA and two AlaRac sequences from various sources were extracted from the SWISS-PROT databank . A multiple sequence alignment was obtained employing the program clustalw. This alignment was then used for the secondary structure prediction (PhD server ), and for detection of evolutionary conserved residues . psi-blast was used to find distantly related sequences of known three-dimensional structure from the GenPept database , using the E. coli low-specificity threonine aldolase sequence as probe. Ten three-dimensional structures of fold type I B6 enzymes were selected and retrieved from the PDB Brookhaven databank (PDB codes: 1dfo, E. coli SHMT ; 1bs0, E. coli 8-amino-7-oxononanoate synthase ; 1b9h, Amycolatopsis mediterranei 3-amino-5-hydroxybenzoic acid synthase ; 1ars, E. coli aspartate aminotransferase [34,35]; 1cl1, E. coli cystathionine β-lyase ; 1qj5, E. coli diaminopelargonic acid synthase ; 2dkb, Pseudomonas cepacia dialkylglycine decarboxylase ; 4gsa, Synechococcus sp. glutamate-1-semialdehyde aminomutase ; 1bjn, E. coli phosphoserine aminotransferase ; 2tpl, Citrobacter freundii tyrosine-phenol lyase ). A multiple alignment of the monomeric three-dimensional structures was performed and the structurally conserved regions (SCRs) were identified. SCRs consisted of regions of secondary structure with similar local conformation, with a root mean square deviation (rmsd) of the equivalent α-carbon positions ≤ 3 Å and composed of at least three consecutive residues. The rmsd values were calculated on the multiple alignment using the homology package in insight ii (Biosym/MSI, 1995, San Diego, CA, USA). The final sequence alignment amongst the SCRs of eTA, Cochliobolus carbonum alanine racemase (cAlaRac) and the 10 fold type I enzymes was manually performed optimizing the matching of several characteristics including the observed and predicted secondary elements, the hydrophobic regions in the three-dimensional structures, the residue similarity measured by the PAM 250 mutational matrix, the structurally and functionally conserved residues, the SCRs, the insertion and deletions in the structures. The homology package in insight ii was used for the manipulation of alignments. A phylogenetic tree was constructed on the basis of the sequence similarity in the SCRs of the large domain of all proteins, using the modules protdist, kitsch and drawgram of the phylip program , available at the Pasteur Institute server (http://bioweb.pasteur.fr).
Model construction and evaluation
The choice of distant templates for a model construction produced by spatial restraint methods should be made carefully to obtain reliable results [43,44]. E. coli SHMT and 8-amino-7-oxononanoate synthase and Citrobacter freundii tyrosine-phenol lyase were chosen as templates to model eTA and Cochliobolus carbonum AlaRac. This choice was made on the basis of the structural and evolutionary similarity of the templates and the sequence similarity of the latter to the target sequences. Protein dimeric models were constructed using the modeller-4 package . Ten different models were built for each target protein and evaluated using several criteria. The model displaying the lowest objective function , which measures the extent of violation of constraints from the templates, was taken as the representative model. Superimposition and rmsd calculation of Cα traces of the 10 models were performed to detect the most variable and therefore less reliable modelled regions. These invariably corresponded to loop elements. procheck was used to monitor the stereochemical quality of the eTA and cAlaRac representative models, whereas prosaii was used to measure the overall protein quality in packing and solvent exposure. The geometrical properties of the representative models were within the accepted ranges. The prosaii method indicated consistent scores for almost all regions, except around sequence position 110 in eTA (large domain). This corresponds to a variable structural element amongst the templates used for modelling, represented by a solvent exposed loop region, distant from the active site. The low score could be a result of local misfolding of this loop.
Analysis of eSHMT substrate specificity
The PLP–glycine complex in the eSHMT three-dimensional structure (which is in the form of a PLP–glycine–5-formyl-H4PteGlu ternary complex ) was converted into l-allo-threonine, l-threonine and l-erythro-3-phenylserine–PLP complexes, using the builder package from insight ii. The χ1 angle of each substrate (the dihedral angle defined by the N, Cα, Cβ and Oγ1 atoms of the substrate) was then rotated by 360° in ten 36° steps, in order to explore the potential energy surface within the active site pocket. Each different conformation of the PLP–substrate–eSHMT complex was subjected to energy minimization and molecular dynamics. All eSHMT atoms were fixed, except those interacting with the substrate side chain as this was rotated (from Thr52′ to Glu69′, from Val256′ to Leu265′, Leu127 and Val133). The PLP–substrate complex was also subjected to a tethering force of 418.4 kJ·mol−1. A Cff91 force field, a distance-dependent dielectric constant, a cut-off distance of 40 Å and a 1-fs time-step were used during the simulation. An initial minimization was performed, for 500 steepest descent steps, followed by conjugated gradient minimization until the maximum energy derivative was less than 0.0418 kJ·mol−1. The complex was then equilibrated for 100 steps at 300°K and the dynamic simulation performed for 10 000 steps at the same temperature. The total energy of the system was monitored for the entire simulation. At the end of the simulation, the system was subjected to energy minimization using 500 steepest descent steps, followed by conjugated gradient minimization until the maximum energy derivative was less than 0.0418 kJ·mol−1. The final energy of the system, for each different conformation, was taken as the potential energy associated to the PLP–substrate–eSHMT complex conformation analysed. discover 2.9 and analysis packages from insight ii were used for minimization and molecular dynamics.
Properties of E. coli low specificity l-threonine aldolase
SHMT activity and retroaldol cleavage of l-serine
When eTA was assayed for serine hydroxymethyltransferase activity, mixing the enzyme at a relatively high concentration (34 µm, i.e. 300-fold higher than that of eSHMT in the standard assay) with 30 mm l-serine and 400 µm H4PteGlu, a measurable production of 5,10-methylene-H4PteGlu (5,10-CH2-H4PteGlu) was observed. The dependence of the initial rate of the reaction on H4PteGlu concentration apparently conformed to the Michaelis–Menten equation and proved to be quite different from that observed with eSHMT (Fig. 1). With eSHMT, the initial velocity of the reaction decreased at concentrations of H4PteGlu higher than 70 µm; these data fitted well to Eqn (1), which accounts for uncompetitive substrate inhibition, with best-fit values of kcat=574±100 min−1, Km=40±10 µm and Ki=83±22 µm, where Km is actually the apparent Michaelis–Menten constant measured at saturating l-serine concentration (30 mm) and Ki is the substrate inhibition constant. The inhibitory effect shown by H4PteGlu may be attributed to the formation of an inactive form of the enzyme due to the binding of this substrate to the SHMT-glycine binary complex.
The drastically different behaviour observed with the two enzymes rules out the hypothesis that the serine hydroxymethyltransferase activity detected with eTA is actually due to the presence of a small contaminant of eSHMT. However, we noticed that the apparent Km for H4PteGlu measured with eTA decreased systematically as the concentration of enzyme employed in the assays was reduced. This suggested that H4PteGlu was engaged in a reaction that did not take place on the enzyme. H4PteGlu is known to react spontaneously with formaldehyde to give 5,10-CH2-H4PteGlu . In our experiments, the rate of 5,10-CH2-H4PteGlu production was measured by oxidizing this compound to 5,10-methenyl-H4PteGlu, using NADP+ and methylenetetrahydrofolate dehydrogenase. This assay cannot discriminate between an enzyme-catalysed hydroxymethyltransferase reaction and an enzyme-catalysed retroaldol cleavage of l-serine followed by the spontaneous reaction of the product formaldehyde with H4PteGlu. An experiment conducted in the absence of H4PteGlu, in which the formaldehyde produced from the cleavage of serine was measured, demonstrated that 5,10-CH2-H4PteGlu formed spontaneously in the previous experiments and that H4PteGlu has no effect on the rate of retroaldol cleavage reaction (Fig. 1, inset). We were unable to measure any serine aldolase activity with eSHMT, employing the same assay conditions and methodology for detection of formaldehyde adopted for eTA (see methods section).
Retroaldol cleavage of 3-hydroxyamino acids
eTA is active towards both erythro and threo isomers of l-threonine and l-3-phenylserine . The effect of pH on the initial velocity of l-allo-threonine cleavage was examined between pH 5.5 and 9.0, at 30 °C and in the presence of 165 mm substrate. The maximum activity of the enzyme was around pH 7.0. The kinetic parameters of this and other retroaldol cleavage reactions catalysed by eTA were then calculated from experiments carried out at pH 7.0 (Table 2). eTA is 60- to 100-fold more efficient as a catalyst (in terms of kcat/Km) than eSHMT in the cleavage of the hydroxyamino substrates tested. Both eTA and eSHMT show a clear preference for the erythro isomer of threonine, being 330-fold and 200-fold, respectively, more efficient in the cleavage of l-allo-threonine.
Table 2. Kinetic parameters of the reactions catalysed by E. coli serine hydroxymethyltransferase and l-threonine aldolase. Kinetic constants are the average of at least three determinations. The range of values was always less than ± 5%. Numbers in parentheses indicate the reaction number, as defined in Table 1.
a ND, not determined. bSchirch et al. . cContestabile et al. . dShostak & Schirch . eApparent Km at saturating [H4PteGlu].
The addition of glycine to eTA resulted in the formation of three absorption bands with peaks at 342, 420 and 492 nm, respectively. As shown in Fig. 2, the relative intensities of these absorption bands are sensitive to pH. A similar observation was first reported for rabbit liver SHMT . With SHMT, the effect of pH suggested that the three bands correspond to three interconverting enzyme-glycine adducts at equilibrium: the geminal diamine of glycine and enzyme-bound PLP (342 nm), the protonated form of the Schiff base between glycine and PLP (420 nm) and a quinonoid structure (492 nm), in which one of the α-protons of glycine has been lost (intermediate IIa in Scheme 1). The value of Kd for glycine (20 mm) was calculated according to Eqn (2), from titration experiments in which the maximum intensity of the absorbance at 492 nm was measured upon addition of different concentrations of glycine.
Rate of exchange of glycine α-protons with solvent protons
In the retroaldol cleavage of 3-hydroxyamino acids, the hydroxyalkyl group of the substrate is replaced by the pro-2S proton of glycine (Scheme 1, path a). As expected and as previously observed with SHMT [50,51], eTA catalysed the exchange of the α-protons of glycine with the protons of the solvent. kcat/Km for this reaction was similar to that measured for eSHMT in the absence of H4PteGlu (Table 2). H4PteGlu binding to eSHMT is known to increase dramatically both the rate and the stereospecificity of the α-protons exchange , raising kcat/Km 3700-fold (Table 2). Inclusion of 0.5 mm H4PteGlu in the reaction catalysed by eTA did not affect the rate of proton exchange.
Transamination of d- and l-alanine
SHMT is slowly inactivated by both d- and l-alanine SHMT [12,53]. This inactivation is the result of transamination between alanine and PLP to give pyruvate and PMP, which is loosely bound to the enzyme and dissociates from it generating the apo-form. Addition of either d- or l-alanine to eTA resulted in the rapid formation of enzyme–substrate complexes absorbing at 420 and 498 nm, whose concentration decreased with time as a new compound absorbing at 324 nm appeared (Fig. 3). This compound, which was easily separated from the enzyme in nearly stoichiometric amounts using a centrifugal concentrator, had the same spectral characteristics of PMP in acid and base . Precipitation of the enzyme was observed as the transamination reaction proceeded. The precipitated enzyme had lost its yellow colour, suggesting it was apoenzyme. A nearly stoichiometric amount of pyruvate with respect to enzyme was formed when the reaction reached equilibrium. The pseudo-first-order rate constant of the reaction was measured by following the disappearance of absorbance at either 420 or 498 nm. Kd values for d- and l-alanine (Table 2) were calculated according to Eqn (2), from titration experiments in which the maximum intensity of the absorbance at 498 nm was measured upon addition of different concentrations of alanine. These values of Kd may be assumed to correspond to the Km values for the transamination and racemization reactions (see below). Although kcat and Km values for transamination differ somewhat between eTA and eSHMT, the values of kcat/Km are very similar.
Racemization of d- and l-alanine
As the quinonoid intermediate formed in the half-transamination reaction of d- and l-alanine (intermediates IIb and IIIb in Scheme 1) is symmetric with respect to both amino acid and cofactor, the racemization of alanine was expected to be catalysed by eTA. Indeed, initial rate studies showed that eTA does catalyse the racemization of alanine starting with either enantiomer (Table 2). kcat values for both reactions are similar to those measured with eSHMT  and, as observed with eSHMT, are faster than the catalytic constants for transamination. However, the lower affinity of eTA for both l- and d-alanine makes this enzyme somewhat less efficient in terms of kcat/KM than eSHMT in the racemization reaction.
L-Threonine aldolase, SHMT and fungal alanine racemase evolutionary context
A detailed, spatial analysis of the structural and possibly evolutionary relationships amongst fold type I enzymes with known three-dimensional structures, groups the members of this family into eight subclasses [5,37]. In the absence of three-dimensional structures for l-threonine aldolase and fungal alanine racemase (AlaRac), an evaluation of the evolutionary relationships between these enzymes and the fold type I members had to rely on sequence comparisons. However, it seemed appropriate to restrict this analysis to segments which have similar spatial conformation in the enzymes of known three-dimensional structure (structurally conserved regions, SCRs), excluding the intervening regions whose structure differs markedly in different enzymes. SCRs were evidently subjected to similar constraints during evolution, therefore the comparison of their sequences may represent a reliable measure of the evolutionary distance amongst different enzymes.
Amino-acid sequences of 16 l-TAs and two AlaRacs were aligned (Fig. 4). A multiple alignment of 10 three-dimensional structures of prokaryotic enzymes was also produced. These were selected so as to include at least one representative of each subclass with the exception of the prokaryotic ornithine decarboxylase subclass. This enzyme was excluded from the analysis because its monomer is composed of five sequential folding domains, only one of which is structurally similar to the major of the two domains present in all fold type I enzymes . The multiple alignment led to the identification of 24 SCRs (as defined in Materials and methods), most of which corresponded to secondary structure elements (Fig. 5A). The secondary structure prediction based on the alignment of l-TAs and AlaRacs showed a strong correspondence with the secondary structure elements of fold type I enzymes (Fig. 5A), confirming a similarity of tertiary fold. Sequence segments of eTA and Cochliobolus carbonum alanine racemase (cAlaRac) were then matched to the SCRs of the fold type I enzymes, taking into account their predicted secondary structure, the Dayhoff evolutionary mutation scoring matrix (PAM 250 matrix) and the position of the functionally and structurally conserved residues that characterize the fold type I enzymes [1,3,56], i.e. the aromatic residue that stacks to the re side of PLP and the aspartate that binds to the pyridine nitrogen of the cofactor (Fig. 5A). In our alignment, a glycine residue at the N-terminal region, positioned at the beginning of the fourth SCR, is also conserved and it is located in the so called glycine rich region of the fold type I enzymes, corresponding to a polypeptide loop which binds the cofactor phosphate . As the greatest similarity in topology was seen, as expected, in the large domain, only the corresponding SCRs of the alignment (the first 19 SCRs) were used to calculate a distance matrix from which a phylogenetic tree was built (Fig. 6). The evolutionary relationships emerging from our analysis coincide with the structural relationships proposed by Käck et al.  on the ground of a spatial alignment of the SCRs and calculated by an alignment score of root mean square deviation (rmsd) of the equivalent α-carbon positions. In addition, our analysis suggests a possible evolutionary route for SHMT, 8-amino-7-oxononanoate synthase (AONS, a member of the α-oxoamine synthases subgroup, enzymes which use derivatives of the CoA as second substrate), 3-amino-5-hydroxybenzoic acid synthase (AHBS), fungal alanine racemase and l-threonine aldolase. According to our analyses, the top seven taxa in Fig. 6 could be arbitrarily clustered in a subgroup including enzymes which catalyse α-replacement and ω-aminotransferase reactions. Within this subgroup, the enzyme that is most similar to l-TA and AlaRac, as indicated by the percentage of identity between l-TA and the other enzymes (Fig. 6), is SHMT. The hypothesis that SHMT, l-TA and AlaRac racemase belong to a subgroup of strictly related enzymes, suggested by the experimental evidence, is confirmed and reinforced by these considerations.
Construction of the three-dimensional homology models of l-threonine aldolase and fungal alanine racemase and comparison with E.coli SHMT overall structure
The rmsd value of the equivalent α-carbon positions in the first 19 SCRs of eSHMT and 8-amino-7-oxononanoate synthase (2.51 Å) is lower than that of eSHMT and glutamate-1-semialdehyde aminomutase (2.93 Å), eSHMT and dialkylglycine decarboxylase (3.48 Å) or eSHMT and diaminopelargonic acid synthase (3.05 Å). These considerations prompted the decision to build the eTA and cAlaRac homology models using E. coli SHMT and 8-amino-7-oxononanoate synthase crystal structures as templates. Moreover, the three-dimensional structure of Citrobacter freundii tyrosine-phenol lyase (TPL) was also used as a template because of its similarity with eSHMT and local sequence identity with eTA and cAlaRac. The very low sequence identity among eTA, cAlaRac and the templates (15.6% maximum, between eTA and eSHMT), required a detailed structural analysis for model construction, mostly based on the SCRs alignment previously obtained (Fig. 5B; see Materials and methods for details).
The homology models were very similar to each other, apart from minor differences in surface loops and in the identity of a few active site residues. Therefore, we will refer to eTA when comparing the models to the eSHMT structure, except when significant differences between eTA and cAlaRac are noteworthy. The superimposed Cα backbone traces of eSHMT and eTA are shown in Fig. 7. The main differences reside in the N- and C-terminal regions, which are shorter in eTA, and in some of the loops connecting the secondary structure elements.
Residues interacting with pyridoxal phosphate
A clear structural similarity between eSHMT and the models is seen in the portion of the active site which interacts with the PLP cofactor (Fig. 8A). Table 3 lists the ligand groups to PLP atoms in the active site of the models as compared to corresponding residues in the structural templates (eSHMT, AONS and TPL). The residue that stacks to the re face of PLP in most fold type I enzymes is either Tyr, Phe or Trp, while in SHMT and AONS it is histidine. In both models, a histidine residue occupies this position. This residue is invariant in the l-TA and AlaRac aligned sequences (Fig. 4), suggesting that the evolutionary proximity of these enzymes to SHMT and AONS may also appear in this feature. It is worth noting that in AONS, SHMT and probably also in l-TA and AlaRac, the histidine imidazole ring lies parallel to PLP, while in the class II aminotransferases (GSAM, DGD and DAPAS) the re face stacking residue packs nearly perpendicular to the cofactor . His129 in eSHMT, which seems to have an important function in stabilizing the interaction between Asp200 and the pyridinium nitrogen of PLP , is present in AlaRac, but it is absent in l-TA. However, in l-TA another histidine residue (His164 in eTA), although not structurally equivalent to His129 of eSHMT, may play the same role. Another residue involved in the stabilization of the same interaction in eSHMT, Asn102, is however, conserved throughout l-TA and AlaRac sequences. In eSHMT, the phenol oxygen, O-3′, of the PLP pyridine ring interacts with His203 and Ser175 residues. Similarly to TPL, these are conservatively replaced in l-TA and AlaRac by Arg and Asn or Gln, respectively.
Table 3. Residues interacting with the PLP cofactor in the structural templates and in the models.
Shiff base forming residue
Bond to pyridine nitrogen
Bond to aspartic acid
Ring stacking re side
Ring stacking si side
Bond to phenol oxygen
Phosphate binding residues
His123 His202? Asn98
Tyr71′ Gly99 Ser254
Residues interacting with the amino-acid substrate
The external aldimines of l-allo-threonine and l-alanine were modelled into the active sites of eTA and cAlaRac, using the PLP–glycine complex in the crystallographic structure of eSHMT as a template. The side chain of l-allo-threonine in eTA was oriented in the active site just as postulated for serine in SHMT (Fig. 8B), whose hydroxyl group, as suggested by modelling and site-directed mutagenesis studies [8,58], should be directed towards Glu57′ and Tyr65′ (the prime indicates that the residues are contributed from the other subunit).
As in eSHMT and several other fold type I enzymes, the α carboxyl group of the substrate in the models is ion-paired to an arginine residue, contributed from the minor domain (Arg308 in eTA, Arg354 in cAlaRac and Arg363 in eSHMT). The side chain of Ser35 in eSHMT hydrogen bonds the α-carboxylate of the substrate and is possibly implicated in the conformational change observed upon formation of the enzyme–glycine–H4PteGlu ternary complex . The models of eTA and cAlaRac have a corresponding serine residue (Ser6 and Ser42, respectively), which is conserved in all l-TAs and AlaRacs aligned but is absent in all other fold type I enzymes examined (see Figs 4 and 5).
In the three-dimensional structure of eSHMT, two residues, Glu57′ and Tyr65′, seem to occupy a crucial position and therefore play an important role in catalysis. It was proposed that one of the two residues, which are located on the re face of the cofactor, opposite to the PLP-binding lysine, may be the catalytic base required for the retroaldol cleavage mechanism (Scheme 1) . However, subsequent mutagenesis experiments suggested that neither Glu57′ nor Tyr65′ acts as a catalytic base. The carboxylate group of Glu57′ very likely interacts with the hydroxyl group of β-hydroxy amino-acid substrates and orients the side chain for the retroaldol cleavage . Moreover, this interaction is possibly one of the main requirements for the conformational transition that confers a definite reaction specificity to the enzyme . The role of Tyr65′ is unclear. It may be responsible for the stabilization of the quinonoid intermediate forming after cleavage of the Cα–Cβ bond  or for the transition from the closed to the open form . All attempts carried out so far to identify the catalytic base in SHMT have failed. It has been proposed that an activated water molecule at the active site may play this role . In the eTA model, a lysine residue (Lys222′) is located approximately in the same position occupied by Glu57′ in eSHMT. A lysine residue is invariantly found in this position in all l-TAs examined. On the other hand, there are no residues in the model corresponding to Tyr65′ in eSHMT, which is located in an insertion respect to eTA and cAlaRac. However, the regions flanking the insertion are highly conserved in l-TAs and do contain a tyrosine residue (Tyr30′), which seems to occupy a position homologous to that of Tyr55′ in eSHMT and of the acid catalyst Tyr71′ in TPL . Tyr30′ is present in all sequences examined except in that from Arabidopsis thaliana (T00716), whose putative identity has been attributed only on the basis of sequence homology. This residue and Asp33′, which is invariant but according to our model it is too far (more than 5 Å) from the hydroxyl group of substrates, are also likely candidates involved in catalysis.
Possible structural determinants of stereospecificity in eTA and eSHMT
eSHMT and eTA show a high degree of specificity toward the erythro isomer of threonine (Table 2). We carried out an investigation of the structural basis of stereospecificity in eSHMT with the purpose of extending the results obtained to eTA. The PLP–glycine complex in the eSHMT three-dimensional structure was used as a template for the construction of PLP–substrate complexes with l-threonine, l-allo-threonine and l-erythro-3-phenylserine. A number of different conformations of each PLP-substrate complex, obtained by applying a torsional force on the Cα–Cβ bond of the amino-acid substrate, were explored in order to detect the potential energy minima of the system analysed (as detailed in Materials and methods). In the case of l-allo-threonine, several minima were found, the lowest of which corresponded to the postulated productive substrate conformation, in which the hydroxyl group is at hydrogen bond distance to both Glu57′ and Tyr65′ (Fig. 9). In this conformation, the methyl group of the substrate is directed towards a hydrophobic pocket, mainly defined by His126, Leu127, Gly262′, Gly263′ and Tyr55′. During the molecular dynamics simulation, the dipeptide formed by Gly262′ and Gly263′ flexed in order to better accommodate the methyl group, as showed in Fig. 9. Analogous results were obtained with l-erythro-3-phenylserine, with the difference that fewer local energy minima were detected; the one corresponding to the productive substrate conformation was the absolute minimum. This latter conformation coincides with that predicted by Scarsdale et al.  for l-erythro-3-phenylserine in the rabbit cytosolic SHMT, solely on the basis of the three-dimensional structure of the enzyme. During the simulation, the bulky phenyl group caused a more extensive flexion of the Gly262′-Gly263′ dipeptide (Fig. 9). The lowest energy minimum detected for the PLP–l-threonine complex presumably corresponds to an unproductive (or less productive) conformation, in which the methyl group is again oriented towards the hydrophobic pocket but the hydroxyl group is directed away from Glu57′, interacting only with Tyr65′ (Fig. 9). The conformation in which the hydroxyl group interacted with both Glu57′ and Tyr65′ did not correspond to an energy minimum and had the methyl group pointing away from the hydrophobic pocket.
The tetrahydropteroylglutamate binding site
Analysing the structure of the active site entrance, a major difference is seen amongst eSHMT, eTA and cAR in the size and shape of three polypeptide loops (Fig. 10), which in eSHMT are larger and are involved in H4PteGlu binding . The different size of these loops is possibly due to insertions in eSHMT (Fig. 5B). The first of these insertions, extending from Asn345 to Gly361 in eSHMT, forms a loop that changes conformation upon formation of the enzyme–glycine–H4PteGlu ternary complex. Importantly, this loop contains Asn347, which binds to N1 and N8 of H4PteGlu and is probably the most significant structural element in the pteridine ring recognition. Other residues (Leu121 and Gly125), whose backbone carbonyls are involved in H4PteGlu binding, are placed in the second short insertion, extending from position 121–125. The third insertion in eSHMT (from Ala56′ to Val71′) makes important interactions with the p-aminobenzoic acid ring of H4PteGlu and possibly with its polyglutamate tail.
eSHMT and eTA have very similar catalytic properties, being able to catalyse the same set of reactions, i.e the retroaldol cleavage of several different l-3-hydroxyamino acids, the transamination and the racemizations of d- and l-alanine and the exchange of the α-protons of glycine with solvent. However, the highest specificity constants (kcat/Km) are measured when eTA and eSHMT are tested, respectively, for the retroaldol cleavage of l-allo-threonine and for the transfer of the Cβ of serine to H4PteGlu (kcat/Km≈ 2000 for both reactions; Table 2). The specificity constants measured for the retroaldol cleavage of l-threonine isomers and l-erythro-3-phenylserine are 60- to 100-fold higher with eTA. Transamination, racemization and exchange of the α-protons of glycine with solvent are catalysed much less efficiently but at similar rates by both enzymes (Table 2). All these data suggest that eSHMT and eTA are closely related enzymes specialized during evolution for different functions. The evolutionary analysis and the structural comparison between eSHMT and the eTA model carried out in this work agree with this interpretation.
The apparent lack of specificity of eSHMT and eTA and their similar catalytic properties are likely to derive from the particular catalytic apparatus of the common ancestor of these proteins that evolved to catalyse two different but strictly related reactions. The basic features of this apparatus are probably still present in the current enzymes and represent the foundation of their similar catalytic properties. An aldolase is thought to break the Cα–Cβ bond of 3-hydroxyamino acids, most probably through a retroaldol cleavage mechanism (Scheme 1), leaving a quinonoid intermediate (IIa) that, in SHMT, is then protonated from the same side of the cleaved bond (Scheme 1, reaction 2a), maintaining the configuration of the α-carbon [51,52]. The catalytic base responsible for the abstraction of a proton from the hydroxyl group of substrates and the protonation of the quinonoid (or the pro-2S proton abstraction from glycine in the reverse direction) must be located on the re face of the cofactor, opposite to the PLP-binding lysine (Lys229 in eSHMT, Lys197 in eTA and Lys235 in cAlaRac). Therefore, two basic residues should be present on the opposite faces of PLP: the PLP-binding lysine and the catalytic base. The presence of two distinct catalytic bases in SHMT has been demonstrated experimentally  and explains why the enzyme shows racemase (Scheme 1, reactions 1b–3b) and aminotransferase activities with both alanine enantiomers (Scheme 1, reactions 1c–2c). It is therefore possible that eSHMT, eTA and fungal AlaRac may act in the racemization of alanine through a similar two-base racemization mechanism. SHMT and eTA evolved so as to prevent as far as possible racemization and transamination of substrates. Analogously, subtle changes occurred during the evolution in fungal AlaRac, without substantially altering the basic structure of the active site, may have shifted the reaction specificity in favour of a racemase activity. The two sequences of fungal alanine racemases are only slightly more similar to each other (41.6% of identity) than to any l-TA examined (36.9% is the highest percentage of identity, between AlaRac from Tolypocladium niveum and l-TA from Candida albicans). Therefore, it does not seem possible to distinguish AlaRac from l-TA solely on the basis of sequence. All the invariant residues in l-TA sequences are also present in AlaRac, except Lys222′ in eTA, which is absent in AlaRac from T. niveum (Fig. 4). Moreover, eTA and cAlaRac three-dimensional models show very strong similarity in the structure of the active site and many of the possibly critical residues in substrate binding and catalysis listed above for eTA have indeed the same identity and location in AlaRac.
We have analysed eTA and AlaRac homology models with the aim of identifying possible candidates for the role of catalytic base in these enzymes. As reported previously by Käck et al. , the superposition of three-dimensional structures of fold type I members reveals that the pyridine ring of the cofactor has a similar orientation and interacts with a portion of the active site pocket whose structure is mainly conserved within this family of enzymes. In fact, SHMT, l-TA and AlaRac are very similar in this region. Fundamentally different structural areas of the active site of fold type I enzymes are two polypeptide loops (located, respectively, between the first and the second SCRs and between the seventeenth and the eighteenth SCRs; Fig. 5A) that are invariably involved in substrate binding and catalyses. The size, the amino-acid composition and the overall fold of these structural elements are highly diversified and appear to confer to the basic ensemble of the active site of fold type I enzymes the structural requirements necessary for substrate and reaction specificity. On the basis of these considerations, our research of the catalytic base was focused on highly conserved residues present on the ‘specificity loops’. Three are the most likely candidates in eTA: Lys222′, Tyr30′ and Asp33′. Only the aspartate residue is conserved in the two AlaRac sequences available and correspond to Asp69′ in cAlaRac (see Results for details).
In the retroaldol cleavage of threonine, the erythro form of the substrate (l-allo-threonine) is preferred by both eTA and eSHMT. In SHMT, the relative specificity constant in the cleavage of threonine isomers, although always in favour of l-allo-threonine, varies depending on the source of the enzyme. eSHMT shows a clear preference for l-allo-threonine, cleaving this substrate with a 200-fold higher specificity constant (Table 2). l-TAs are classified as low specificity l-TAs, which cleaves both threonine isomers, and as l-allo-TAs, absolutely specific for the erythro form . However, except for l-TA from Pseudomonas sp., which does not clearly select the isomers , l-allo-threonine is always the preferred substrate. For instance, eTA shows a 330-fold higher specificity constant for l-allo-threonine cleavage (Table 2). l-threo-3-Phenylserine is a good substrate for both enzymes and is cleaved with an efficiency similar to that of l-allo-threonine cleavage. Some specificity towards the configuration of the Cβ of the substrates with a phenyl group is observed with both enzymes [21,24]. The molecular dynamics simulations of PLP–substrate complexes carried out using the eSHMT crystal structure suggest that the main structural determinants of stereospecificity in this enzyme are the interactions of the substrate hydroxyl group with the side chains of E57′ and Y65′ and of threonine methyl group with the hydrophobic pocket defined by His126, Leu127, Gly262′, Gly263′ and Tyr55′. The flexion of the dipeptide formed by Gly262′ and Gly263′, observed during the simulations, appears to accommodate the Cβ substituent of the hydroxyamino-acid substrate and may account for the capability of SHMT to bind and cleave efficiently substrates with a bulky Cβ substituent such as the phenyl group of 3-phenylserine. The hydrophobic pocket that accommodates the Cβ substituents in eSHMT is similar in the eTA model and the residues that define it are mainly conserved in l-TAs (Figs 4 and 5). This is an interesting clue for the explanation of the observed preference of l-TAs for l-allo-threonine.
SHMT from different sources is known to catalyse, in the absence of H4PteGlu, although extremely slowly (kcat=3–6×10−3 min−1), the cleavage of l-serine to glycine and formaldehyde [52,62]. H4PteGlu accelerates the rate of cleavage up to 200 000-fold, although in the presence of this second substrate SHMT actually catalyses a hydroxymethyltransferase reaction. The kcat for the exchange of the pro-2S proton of glycine with solvent is also increased by 1100-fold when H4PteGlu is added to eSHMT (Table 2). All other reactions catalysed by SHMT are accelerated upon binding of H4PteGlu, although to a much lesser extent (twofold to fourfold [12,62]). eTA cleaves l-serine with a higher kcat (1.9 min−1) than that measured with SHMT. This was expected as eTA was found to be a more efficient catalyst in the cleavage of all other 3-hydroxyamino acids. However, H4PteGlu has no effect on the rate of this reaction, although it spontaneously combines with the product formaldehyde, forming 5,10-CH2-H4PteGlu. No effect of H4PteGlu was also observed on the rate of exchange of the α-protons of glycine with solvent. The contribution of H4PteGlu to catalyse and especially its involvement in the cleavage of serine are the only substantial differences in the properties of eSHMT and eTA. The most conspicuous difference in the structures of the two enzymes appears indeed to be related to the additional binding of H4PteGlu to SHMT. According to our analysis, the evolution of the folate site apparently consisted in the appropriate shaping of three polypeptide loops present at the active site entrance of SHMT, which left basically unaltered the structure of the rest of the active site (Fig. 10). One of these loops, extending for Ala56′ to Val71′, is a ‘specificity loop’ and contains residues that also interact with the amino-acid substrates.
An interesting question is which enzyme, SHMT or l-TA, more closely resembles the most recent common ancestor. This is equivalent to asking whether the capability of transferring the serine β-carbon to H4PteGlu evolved after that of catalysing the retroaldol cleavage or, alternatively, l-TA originated from a more complex enzyme, to perform a simpler, apparently minor metabolic function. Useful clues may come from structural and mechanistic considerations. It is reasonable to suppose that evolution of reaction specificity in B6-enzymes came before specialization for substrate specificity . Indeed, the adjustment of a catalytic apparatus required to change the reaction specificity presumably involves larger and more conditioning structural adaptations than those necessary for a change of substrate specificity . Thus, it seems plausible that the most recent common ancestor of l-TA and SHMT was able to catalyse the retroaldol cleavage of 3-hydroxyamino acids and then evolved the additional level of complexity that enabled it to bind H4PteGlu and catalyse the serine hydroxymethyltransferase reaction. This hypothesis is supported by the fact that serine cleavage is thermodynamically unfavourable compared to threonine cleavage [62,63]. In the absence of H4PteGlu, SHMT catalyses the serine cleavage, although extremely slowly, while it is able to cleave threonine and other 3-hydroxyamino acids quite rapidly. This difference in reactivity is attributed to the instability of the product formaldehyde. The presence of a stabilizing electron donor group at Cβ greatly accelerates the reaction [62,63]. The participation of H4PteGlu in SHMT-catalysed serine cleavage overcomes this energy barrier, driving the reaction through a different and more favourable path. Therefore, it seems plausible that when the SHMT ancestor evolved the capability to include H4PteGlu in the reaction, it already had to possess all the necessary requirements to be a generic 3-hydroxyamino-acid aldolase.
The comparative study between SHMT and l-TA initiated with the present work will be further developed and extended to fungal alanine racemase. A detailed structural and functional comparison requires the availability of the crystal structures of l-TA and fungal AlaRac. However, the three-dimensional homology models represent a good starting point for the design of site-directed mutagenesis experiments aimed at determining the identity of the crucial catalytic residues.
We thank Professors Günther Kreil, Verne Schirch and Robert A. John for their help during the writing of the manuscript and for helpful discussions. We are grateful to Eprova AG, Schaffhausen, Switzerland, for kindly providing us with pure (6S)-H4PteGlu. This work was supported by the Italian Ministero dell'Università e della Ricerca Scientifica e Tecnologica.