Mechanism for folate-independent aldolase reaction catalyzed by serine hydroxymethyltransferase


  • Yoko Chiba,

    1.  Department of Biotechnology, Graduate School of Agricultural and Life Sciences, The University of Tokyo, Japan
    Search for more papers by this author
  • Tohru Terada,

    1.  Agricultural Bioinformatics Research Unit, Graduate School of Agricultural and Life Sciences, The University of Tokyo, Japan
    Search for more papers by this author
  • Masafumi Kameya,

    1.  Department of Biotechnology, Graduate School of Agricultural and Life Sciences, The University of Tokyo, Japan
    Search for more papers by this author
  • Kentaro Shimizu,

    1.  Department of Biotechnology, Graduate School of Agricultural and Life Sciences, The University of Tokyo, Japan
    2.  Agricultural Bioinformatics Research Unit, Graduate School of Agricultural and Life Sciences, The University of Tokyo, Japan
    Search for more papers by this author
  • Hiroyuki Arai,

    1.  Department of Biotechnology, Graduate School of Agricultural and Life Sciences, The University of Tokyo, Japan
    Search for more papers by this author
  • Masaharu Ishii,

    1.  Department of Biotechnology, Graduate School of Agricultural and Life Sciences, The University of Tokyo, Japan
    Search for more papers by this author
  • Yasuo Igarashi

    1.  Department of Biotechnology, Graduate School of Agricultural and Life Sciences, The University of Tokyo, Japan
    Search for more papers by this author

M. Ishii, Department of Biotechnology, Graduate School of Agricultural and Life Sciences, The University of Tokyo, 1-1-1 Yayoi, Bunkyo-ku, Tokyo 113-8657, Japan
Fax: +81 3 5841 5272
Tel: +81 3 5841 5143


Serine hydroxymethyltransferase catalyzes the cleavage of β-hydroxyamino acids into glycine and aldehydes in the absence of tetrahydrofolate. The enzyme accepts various β-hydroxyamino acids as the substrate of this reaction. The reaction rate varies depending on the substituent and stereochemistry at the Cβ atom: the erythro forms and the β-phenyl substituent are preferred over the threo forms and the β-methyl substituent, respectively. Although several mechanisms have been proposed, what determines the substrate preference remains unclear. We first performed quantum mechanical calculations to assess the validity of the reaction mechanisms. The results indicate that the retro-aldol mechanism starting with abstraction of the proton from the β-hydroxyl group is plausible. This also suggests that Cα–Cβ bond cleavage is the rate-limiting step. We next measured the dependence of the rate constants on temperature with four representative substrates and calculated the activation energies and pre-exponential factors from the Arrhenius plots. The activation energies of the erythro forms were lower than those of the threo forms. The β-phenyl substituent lowered the activation energy in the threo form, whereas it did not alter the activation energy but increased the pre-exponential factor in the erythro form. We present a unified model to explain the origin of the substituent and stereochemical preferences by combining the theoretical and experimental results. A possible biological role of the tetrahydrofolate-independent activity in thermophiles is also discussed.


serine hydroxymethyltransferase from Bacillus stearothermophilus


serine hydroxymethyltransferase from Escherichia coli


serine hydroxymethyltransferase from Hydrogenobacter thermophilus TK-6


serine hydroxymethyltransferase from Methanococcus jannaschii, PLP, pyridoxal-5′-phosphate


quantum mechanical


serine hydroxymethyltransferase


serine hydroxymethyltransferase from Streptococcus thermophilus


tetrahydrofolate, TS1, transition state of the first reaction step


transition state of the second reaction step


A pyridoxal-5′-phosphate (PLP)-dependent enzyme, serine hydroxymethyltransferase (SHMT; EC2.1.2.1) is a ubiquitous enzyme that plays an essential role in the one-carbon metabolism [1–3]. Its primal function is the interconversion of serine and tetrahydrofolate (THF) to glycine and 5,10-methylene-THF. The one-carbon unit transferred from serine to THF is used in the biosynthesis of purines, thymidylate, methionine and other important biomolecules [1,3,4]. The substrate, serine or glycine, binds to the active site of the enzyme, forming an aldimine with PLP. Considerable efforts have been devoted to clarifying how serine is cleaved into glycine. The current consensus is that the N5 atom of THF makes a nucleophilic attack on the Cβ atom of serine, breaking the Cα–Cβ bond, without involving free formaldehyde as an intermediate [2–4].

In addition to this THF-dependent SHMT activity, the enzyme exhibits THF-independent aldolase activity toward β-hydroxyamino acids, producing glycine and aldehydes [1,5,6]. This reaction has also attracted attention for its characteristic substrate preference. The enzyme accepts both the threo and erythro forms of various β-hydroxyamino acids as the substrate in this reaction. The reaction proceeds more efficiently with the erythro forms than with the threo forms [2]. The reaction rate also depends on the substituent at the Cβ atom of the β-hydroxyamino acids. For example, the reaction rate of l-erythro-β-phenylserine is greater than that of l-allo-threonine by two orders of magnitude [6,7]. It has been shown that there is a linear correlation between the logarithm of the hydration equilibrium constant of the aldehyde produced by the reaction and that of the rate constant [6]. Based on this observation, it has been proposed that electron-donating substituents accelerate the reaction by stabilizing the transition state [6].

To understand what determines these preferences, it is necessary to understand the mechanism for the reaction. Two mechanisms have been proposed. The first is a retro-aldol cleavage mechanism, triggered by the abstraction of the proton from the β-hydroxyl group (Fig. 1A) [8]. The second is that a proton is abstracted from the Cα of β-hydroxyamino acids rather than from the β-hydroxyl group (Fig. 1B) [9]. However, neither is sufficient to explain the origin of the substrate preference of the THF-independent reaction, because the details of the reaction processes, such as rate-limiting steps and transition-state structures, are still unclear. Although the retro-aldol reaction has been well established in organic chemistry, this model has a problem in that the base required to abstract the proton has not yet been identified. Although there are a glutamic acid and a histidine near the hydroxyl group of an erythro form substrate in the active site, the mutants of these residues (E74Q and H147N of sheep liver cytosolic SHMT [8,10], E75L and E75Q of rabbit liver cytosolic SHMT [4] and E53Q of Bacillus stearothermophilus SHMT, bsSHMT [11]) retained THF-independent aldolase activity. A second model has recently been proposed, based on observations that the hydroxyl group of Y61 of bsSHMT is close to the Hα of the substrate in its crystal structure and the Y61F and Y61A mutants of bsSHMT have lost aldolase activity to l-allo-threonine [9]. However, it remains to be determined whether such a reaction is chemically feasible.

Figure 1.

 Proposed mechanisms for THF-independent aldolase reaction. (A) Retro-aldol reaction mechanism starting with abstraction of proton from β-hydroxyl group. (B) Reaction mechanism starting with abstraction of proton from Cα atom.

We first performed quantum mechanical (QM) calculations to determine which reaction mechanism was plausible to solve these problems. We then measured reaction rates at various temperatures for several substrates to calculate the activation energies and pre-exponential factors of the Arrhenius equation. Here, we used SHMT from Hydrogenobacter thermophilus TK-6, which was obligately chemolithoautotrophic and an aerobic hydrogen-oxidizing bacterium isolated from a hot spring in Izu, Japan [12,13]. The complete genome sequence for this strain was recently determined [14] and an SHMT ortholog (htSHMT) was identified. Because this bacterium is a thermophile with an optimum growth temperature of 70–75 °C, the enzyme exhibits high thermostability [15,16] and is therefore suitable for measuring reaction rates over a wide range of temperatures. We discuss the origin of the preference in the THF-independent reaction of SHMT by combining the theoretical and experimental results. We also mention a possibility that the THF-independent activity of SHMTs affects metabolism in thermophilic organisms.


Theoretical evaluation of reaction mechanisms

We first examined the retro-aldol reaction mechanism, in which the reaction proceeded through two steps of: (a) a proton being abstracted from the β-hydroxyl group of the substrate, and (b) the Cα–Cβ bond of the substrate being elongated and subsequently cleaved to form a glycine–pyridoxal aldimine and an aldehyde. QM calculations were performed on the model of the active center, whose structure was extracted from the crystal structure of bsSHMT in complex with l-allo-threonine. The model was composed of a substrate–pyridoxal aldimine and a 1-methylguanidine (Fig. 2A). The 1-methylguanidine was included in the model to mimic the arginine side chain which forms tight salt bridges with the carboxyl group of the substrate. The phosphate group of PLP was replaced with a hydrogen atom to reduce the computational cost. Here, we assumed that a hydroxide ion (OH) would act as a base that abstracted the proton from the β-hydroxyl group. Figure 2B plots the free-energy profiles calculated for l-allo-threonine and l-erythro-β-phenylserine. The energy profiles indicate that each step has a transition state (TS1 for the first reaction step and TS2 for the second step) and there is an intermediate state between the two steps. The enthalpies and free energies of these states are summarized in Table 1 relative to those of the reactants. Because the energy barrier of the first step is much lower than that of the second, the second step is rate limiting. Because the activation enthalpies, ΔH, are reasonable (60.5 and 48.9 kJ·mol−1 for l-allo-threonine and l-threo-β-phenylserine, respectively) for enzymatic reactions, the retro-aldol reaction mechanism is plausible. The structures in TS2 are shown in Fig. 2C,D. A detailed comparison with the experimental data is made below.

Figure 2.

 (A) Structural formula of the model of the active center used in QM calculations. (B) Free energy profiles of retro-aldol mechanism calculated at 25 °C (298.15 K) for reactions with l-allo-threonine (thin bars connected by dotted lines) and l-erythro-β-phenylserine (thick bars connected by dashed bars). Ball-and-stick representation of structures in TS2 of reactions with l-allo-threonine (C) and l-erythro-β-phenylserine (D). Distances between atoms indicated by dotted lines are shown. Hydrogen, carbon, nitrogen and oxygen atoms are colored white, black, dark gray and light gray. Structural images were generated with molscript [34].

Table 1.   Enthalpies and free energies relative to those of reactants obtained by QM calculations.
ΔH (kJ·mol−1)ΔG (kJ·mol−1)ΔH (kJ·mol−1)ΔG (kJ·mol−1)

We next examined the second reaction mechanism that started with Cα proton abstraction. This reaction mechanism proceeded through a carbanion intermediate (Fig. 1B). The optimized structure of the intermediate indicates that the Cα atom has a trigonal–planar geometry with sp2 hybridization (Fig. 3B). As a result, the Cβ atom moves into the plane in which the carboxyl group, the aldimine group and the pyridine ring of pyridoxal lie. This structure is stabilized by delocalizing the negative charge generated on the Cα atom over the atoms of these groups. When the PLP part of the optimized structure is superimposed on the corresponding part of the crystal structure of bsSHMT, the side-chain atoms of the substrate bump protein atoms (Fig. 3). Therefore, we concluded that this reaction was unlikely to occur. The losses of the activities of the Y61A and Y61F mutants were probably caused by other reasons: Contestabile et al. [17] suggested that the hydroxyl group of Y61 plays an important role in the transition to an open conformation, which enables the products to escape from the catalytic pocket of the enzyme.

Figure 3.

 Ball-and-stick representation of l-allo-threonine–PLP aldimine in crystal structure of bsSHMT in complex with l-allo-threonine (A) and optimized structure of carboanion intermediate modeled into catalytic pocket of crystal structure (B). Protein residues near substrate in the crystal structure are also shown. The model was generated by superimposing pyridoxal atoms of intermediate on corresponding atoms of PLP in crystal structure. Positions of Cγ atoms of substrate are indicated. Only nonhydrogen atoms are shown. Phosphate group in l-allo-threonine-PLP aldimine is not shown. Carbon, nitrogen and oxygen atoms are colored black, dark gray and light gray. Atoms within 2.6 Å are indicated by dotted lines. Structural images were generated with molscript [34].

Thus, the QM calculations strongly suggest that the THF-independent reaction occurs via the retro-aldol mechanism and this provided detailed information on the reaction process. The activation energies were especially useful because they could be compared with the experimental values. We describe the kinetic experiments performed below to confirm the QM results and to examine the stereochemical and substituent effects of the substrates on the kinetic parameters.

Identification of htSHMT

We measured reaction rates at different temperatures in the kinetic experiments to calculate the activation energies and pre-exponential factors of the Arrhenius equation. We used H. thermophilus TK-6 as the source of SHMT, because the enzymes from this thermophilic bacterium were expected to be stable at high temperatures and therefore useful for measuring reaction rates over a wide range of temperatures. Although SHMT had not yet been identified in this bacterium, we confirmed that both forward THF-dependent SHMT activity and THF-independent aldolase activity against l-threonine and l-allo-threonine were present in the cell-free extract (data not shown). A BLAST search against the genome of H. thermophilus TK-6 indicated that only HTH1832 shared reasonable similarity with the biochemically characterized SHMTs. Figure S1 shows multiple alignment of the amino acid sequences for HTH1832, bsSHMT, Escherichia coli SHMT (ecSHMT) and Methanococcus jannaschii SHMT (mjSHMT). HTH1832 exhibits 61% sequence identity with bsSHMT and their pairwise alignment has no gaps. HTH1832 conserves all the residues whose side-chain atoms interact with l-allo-threonine or PLP in the crystal structure of bsSHMT (Fig. S1). These residues are also conserved in ecSHMT and mjSHMT. Note that the A95 and G257 of bsSHMT interact with PLP at their main-chain atoms. Some microorganisms have l-threonine aldolase (EC4.1.2.5) in addition to SHMT, but no threonine aldolase ortholog was detected in the genome sequence of H. thermophilus TK-6.

We predicted that the THF-dependent SHMT activity and the THF-independent (allo-) threonine aldolase activities observed in the cell-free extract could be attributed to HTH1832 on the basis of these results. HTH1832-coded protein was heterologously produced and its activities were checked as described below.

Characterization of htSHMT

HTH1832 was cloned and overexpressed in E. coli. Because the E. coli cell-free extract demonstrated a significant amount of THF-dependent SHMT activity, we concluded that the HTH1832-coded protein was htSHMT. We obtained 3.5 mg of the enzyme from 3 g of the wet cells after three-step purification using heat treatment and two anion-exchange columns. The purified enzyme exhibited a single band on SDS/PAGE and a single peak on gel chromatography. The molecular mass was calculated to be 48.7 kDa from the SDS/PAGE, consistent with the predicted mass based on the amino acid sequence (47.5 kDa). Gel chromatography enabled us to estimate the molecular mass of native htSHMT as 74 kDa. Because the molecular masses of thermophilic enzymes are often underestimated by gel filtration [18,19], htSHMT is probably a homodimer, as are bsSHMT and the other prokaryotic SHMTs.

The thermal stability of htSHMT was determined by measuring the residual aldolase activity with l-allo-threonine at 70 °C after aerobic incubation at 0–100 °C for 10 min (Fig. 4). htSHMT retained almost complete activity after being incubated at 75 °C and retained ∼ 77% residual activity even after being treated at 87 °C. This observation confirms that htSHMT has sufficient thermostability to determine the reaction rate at 75 °C and below. The effect of pH on THF-independent aldolase activity was also examined. Within the tested pH range (5.9–8.4 at 70 °C), htSHMT demonstrated the highest activity at pH 7.4 (20.4 U·mg−1). More than 95% of full activity remained at pH 8.4, although the activities were reduced to ∼ 90% at pH 6.9 and ∼ 75% at pH 5.9.

Figure 4.

 Thermal stability of htSHMT as measured by residual l-allo-threonine aldolase activity. After treatment at indicated temperature for 10 min in 20 mm Tris/HCl buffer, residual activity was measured at 70 °C. Activities are indicated as relative values of that treated at 0 °C.

Kinetic measurements

The kinetic constants (Km and kcat) of the THF-dependent interconversion of l-serine and glycine and those of the THF-independent cleavage of l-threonine and l-allo-threonine were determined at 70 °C in 20 mm Tris/HCl (pH 7.5) containing 100 μm of PLP (Table 2). The kinetic constants were measured in Hepes (pH 7.5) at 75 °C (Table 2) for the THF-independent cleavage of dl-threo-β-phenylserine and dl-erythro-β-phenylserine. The constants for the l-allo-threonine cleavage were also determined in Hepes for comparison. The results confirmed that the recombinant htSHMT exhibited the same substrate preferences in the THF-independent aldolase reaction as the SHMTs from other organisms: the rate constants for the erythro form substrates and for the substrates with β-phenyl groups were larger than those for the threo form substrates and for the substrates with β-methyl groups, respectively.

Table 2.   Kinetic constants of reactions catalyzed by htSHMT.
Substrate (s−1·mm−1)THFKm (mm)kcat (s−1)kcat/KmConditions
  1. a 5,10-methylen-THF was added.

l-serine+0.28 ± 0.1818.7 ± 2.466.8Tris/HCl pH 7.5, 70 °C
Glycine+a0.78 ± 0.205.0 ± 0.36.4Tris/HCl pH 7.5, 70 °C
l-threonine7.64 ± 0.842.3 ± 0.10.3Tris/HCl pH 7.5, 70 °C
l-allo-threonine0.92 ± 0.0818.0 ± 0.419.6Tris/HCl pH 7.5, 70 °C
l-allo-threonine0.59 ± 0.0425.2 ± 0.442.7Hepes pH 7.5, 75 °C
dl-threo-β-phenylserine4.98 ± 1.1893.8 ± 6.218.8Hepes pH 7.5, 75 °C
dl-erythro-β-phenylserine2.63 ± 0.92324 ± 29123Hepes pH 7.5, 75 °C

Next, we measured the rate constants of the THF-independent cleavage of l-threonine, l-allo-threonine, dl-threo-β-phenylserine and dl-erythro-β-phenylserine over the range of temperatures from 25 to 75 °C. We assumed that the d forms had not been cleaved by the enzyme in this analysis. This assumption is reasonable because if d-threo-β-phenylserine or d-erythro-β-phenylserine is modeled into the catalytic pocket of the crystal structure of bsSHMT, the phenyl group bumps many protein atoms. The rate constants obtained with the dl-mixtures coincide with those obtained with the l forms under this assumption. The activation energies, Ea, and pre-exponential factors, A, of the THF-independent aldolase reactions were calculated from the Arrhenius plots (Fig. 5 and Table 3). The reaction rate for l-threonine at 25 °C was too small and was therefore excluded from the analysis. The Ea for the l-allo-threonine cleavage by htSHMT was 73.3 ± 3.8 kJ·mol−1, which is consistent with those obtained for ecSHMT and mjSHMT (70.00 ± 2.10 kJ·mol−1) [20].

Figure 5.

 Arrhenius plots of logarithms of catalytic constants of THF-independent aldolase reactions as function of inverse of temperatures. Catalytic constants were measured for 10 mm of dl-erythro-phenylserine (•), 10 mm of dl-β-threo-phenylserine (○), 10 mm of l-allo-threonine (bsl00001) and 20 mm of l-threonine (□).

Table 3.   Activation energies and logarithms of pre-exponential factors calculated from Arrhenius plots.
SubstrateEa (kJ·mol−1)ln A
l-threonine90.2 ± 1.70.598
l-allo-threonine73.3 ± 3.80.538
dl-threo-β-phenylserine79.4 ± 1.60.600
dl-erythro-β-phenylserine71.3 ± 2.60.577


Comparison of theoretical and experimental results

The activation enthalpy, ΔH, obtained from the QM calculations for l-allo-threonine (60.5 kJ·mol−1) agreed well with the activation energy, Ea, obtained in the experiment (73.3 ± 3.8 kJ·mol−1). This confirms that the THF-independent aldolase reaction occurs with the retro-aldol reaction mechanism. However, there is a discrepancy in the data for l-erythro-β-phenylserine. The value of ΔH for l-erythro-β-phenylserine in the QM calculations is smaller than that for l-allo-threonine by 11.6 kJ·mol−1, whereas the Ea values in the experiment were almost the same for the two substrates.

The QM results indicate that TS2 for l-allo-threonine is stabilized by delocalizing the negative charge generated by proton abstraction over the pyridoxal, aldimine-group, Cα and carboxyl-group atoms. This can be seen from the distribution of the highest occupied molecular orbital (Fig. 6A). The highest occupied molecular orbital also spreads over the phenyl group (Fig. 6B) in the TS2 for l-erythro-β-phenylserine. Therefore, the smaller value of ΔH for l-erythro-β-phenylserine compared with that for l-allo-threonine is theoretically reasonable. The Ea for dl-threo-β-phenylserine is consistently smaller than that for l-threonine by 10.8 kJ·mol−1, probably due to the delocalizing effect.

Figure 6.

 Distribution of highest occupied molecular orbitals in TS2 structures of l-allo-threonine (A) and l-erythro-β-phenylserine (B). Molecular structures are ball-and-stick representations. Hydrogen, carbon, nitrogen and oxygen atoms are colored white, gray, blue and red. Isosurfaces of highest occupied molecular orbitals are colored dark red and dark green. Images were generated with gaussview v. 5 [29].

The reason that l-erythro-β-phenylserine had a larger Ea value than that expected from QM calculations can be explained as follows. The Oγ atom of the substrate in the TS2 structure forms a tight hydrogen bond with the water molecule produced by proton abstraction, which forms another hydrogen bond with the 5′-OH group of pyridoxal (Fig. 2C,D). In the crystal structure of bsSHMT, a similar hydrogen-bond network is formed, in which the β-OH group of l-allo-threonine and the 5′-phosphate group are bridged through two water molecules. Szebenyi et al. [4] predicted that l-erythro-β-phenylserine has synperiplanar conformation for the β-OH group with respect to the aldimine group with an N–Cα–Cβ–Oγ dihedral angle of approximately 0°, whereas l-allo-threonine has conformation with the N–Cα–Cβ–Oγ dihedral angle of approximately –60° in the crystal structure. The shift in orientation of the β-OH group is caused by steric hindrance between the phenyl group of the substrate and the protein atoms and may destabilize the hydrogen-bond network in which the β-OH group is involved in TS2. In general, a hydrogen bond with a water molecule lowers the enthalpy of a molecular complex at most by ∼ 20 kJ·mol−1, and the actual decrease in the enthalpy depends on the geometry of the hydrogen bond. The Ea for l-erythro-β-phenylserine is larger by ∼ 10 kJ·mol−1 than that expected from the Ea for l-allo-threonine and the difference between the ΔH values of the QM calculations for the two substrates. Because the geometry of the hydrogen bonds is considered almost optimal in the catalytic pocket with l-allo-threonine as well as in the QM calculations, this discrepancy can be explained by the nonoptimal geometry of the hydrogen bonds formed in the catalytic pocket with l-erythro-β-phenylserine.

The larger Ea values for the threo forms can be considered similarly; a stable hydrogen-bond network might not be formed for the threo forms, because the orientation of the β-OH groups is probably different from that of the erythro forms. The conformations of the β-OH groups of the threo forms are predicted to be antiperiplanar with an N–Cα–Cβ–Oγ dihedral angle of ∼ 180° [4]. This inference is consistent with the increase in pre-exponential factors. Because the pre-exponential factor is related to entropy, the more stable, i.e. the more ordered, hydrogen-bond network should give a smaller pre-exponential factor. As expected, l-allo-threonine gave the smallest pre-exponential factor, followed by dl-erythro-β-phenylserine and then l-threonine; the value for dl-threo-β-phenylserine was almost the same as that for l-threonine.

Origin of substrate preference

The origin of the substrate preference can be summarized as follows based on the above discussion. The stereochemical preference was caused by the difference in orientation of the β-OH group. A stable hydrogen-bond network is formed between the β-OH group and the 5′-phosphate group in the TS2 for the erythro forms, whereas this is not the case in the TS2 for the threo forms because the β-OH group points in a different direction. The β-phenyl group lowers the activation energy, as observed for dl-threo-β-phenylserine. However, the larger reaction rate for dl-erythro-β-phenylserine than for l-allo-threonine results from the larger pre-exponential factor. The enthalpic gain by the β-phenyl group is canceled out by the destabilization of the hydrogen-bond network because of a shift in the orientation of the β-OH group, but the loose requirements on the structure of the hydrogen-bond network increase the pre-exponential factor. Although it has been proposed that an electron-donating substituent stabilizes the transition state [6], this is probably not the case; the sum of the atomic charges over the substituent atoms in the TS2 structures were −0.030 for l-allo-threonine and −0.070 for l-erythro-β-phenylserine. This indicates that the substituent groups do not donate the electron. The substituent effect by the β-phenyl group is consequently caused by the delocalization of negative charge produced by proton abstraction over the phenyl group. Because the reaction rate of 4-chloro-l-threonine has been shown to be slower than that of l-threonine [6], the electron-withdrawing substituent might affect charge delocalization.

The E53 and H122 in the crystal structure of bsSHMT are in close proximity to the β-OH group of l-allo-threonine, forming hydrogen bonds with it [21]. However, mutation for these residues did not eliminate THF-independent aldolase activity [4,8,10,11], and the base that abstracted the proton has not yet been identified. As seen in the energy profiles (Fig. 2B), the energy barrier for proton abstraction from the substrate is very small. This indicates that this reaction occurs very quickly and a long-lived base is not necessary. As assumed in the QM calculations, the OH ions generated through the equilibrium between H2O and OH + H+ are probably sufficient to cause the reaction. Therefore, the protein residues do not play a major role in determining the substrate preference in THF-independent aldolase activity.

Biological role of THF-independent aldolase activity

It has been widely considered that the physiological role of SHMT is the THF-dependent interconversion of glycine and serine [1,22,23]. Although THF-independent threonine aldolase activity has been detected from several SHMTs in vitro, these activities are very weak and have been considered to be negligible [1]. To the best of our knowledge, the only exception is an SHMT of Streptococcus thermophilus (stSHMT); mutation studies have revealed that stSHMT displays threonine aldolase activity in vivo [23]. Although stSHMT has a 60% identity with ecSHMT, it demonstrates considerably higher activity to l-threonine [kcat (s−1) = 1.26 at 37 °C] [24] than ecSHMT [kcat (s−1) = 0.07 at 37 °C] [22] and other mesophilic SHMTs [1]. The THF-independent threonine aldolase reaction in other organisms is catalyzed by ‘real’ threonine aldolase, which is an enzyme specific to l-threonine cleavage [1]. The physiological role of threonine aldolase is assumed to produce glycine from l-threonine as exhibited in Saccharomyces cerevisiae [25–27].

Not only exceptional SHMTs like stSHMT, but also the SHMTs of thermophiles are expected to exhibit substantial levels of threonine aldolase activities in vivo for the following reason. The kcat value for l-threonine cleavage by htSHMT at 70 °C (2.3 s−1; Table 2) is comparable with those by real threonine aldolase from E. coli (1.0 s−1 at 30 °C) [22] or stSHMT (1.26 at 37 °C) [24]. The Km value for l-threonine (7.64 mm at 70 °C; Table 2) is also comparable with that for E. coli threonine aldolase (10 mm at 30 °C) [22]. Therefore, the threonine aldolase activity of htSHMT is high enough in vivo at 70 °C, which is the optimal growth temperature for H. thermophilus. Because the amino acid residues at the active site of SHMTs are highly conserved across a broad range of microorganisms (Fig. S1), SHMTs in thermophiles, as well as htSHMT, will exhibit a considerable amount of threonine aldolase activity in vivo. Interestingly, although threonine aldolases have been biochemically detected and characterized in mesophilic bacteria, yeast and plants [22], most thermophilic organisms lack the ortholog genes for this enzyme. Therefore, it is possible that SHMT can substitute for the role of real threonine aldolase in these thermophiles. The kinetic study on SHMT thus provided important insights into the metabolism in thermophiles that could not be predicted solely from their genomic data, as well as the mechanism for the THF-independent aldolase reaction.

Experimental procedures

QM calculations

We performed QM calculations to evaluate the plausibility of the two proposed mechanisms of the THF-independent reaction. According to the crystal structure of the F351G mutant of bsSHMT in complex with l-allo-threonine [Protein Data Bank (PDB) identification code: 2VMX] [28], the substrate forms an aldimine with PLP at its amino group and makes tight salt bridges with an arginine residue at its carboxyl group. Therefore, the calculations were performed for a system composed of a substrate–pyridoxal aldimine and a 1-methylguanidine that mimics the arginine side chain. The 5′-phosphate group of PLP was replaced with a hydrogen atom to reduce the size of the system (Fig. 2A). The total charge of the system was zero. l-allo-threonine and l-erythro-β-phenylserine, acting as the substrate, were subjected to the calculations. The initial geometries were generated with gaussview v. 5 [29] by using the coordinates of the above crystal structure. Note that the F351G mutation did not affect THF-dependent SHMT or THF-independent aldolase activities [28]. All the calculations, viz., the geometry optimizations, transition state searches, vibrational analyses and population analyses, were performed at the HF/6-31G* level of theory within the gaussian 09 program package [30]. The transition states were confirmed through vibrational analyses and intrinsic reaction coordinate calculations. Atomic charges were calculated with the method of natural population analysis [31]. Enthalpies and free energies were calculated at 25 °C (298.15 K).

Construction of expression plasmid

The SHMT gene [HTH1832, National Center for Biotechnology Information (NCBI) accession number: BAI70276] was PCR-amplified from H. thermophilus genomic DNA using the following primers: 5′-GACTAGAATTCTATCATATGAGACACC-3′ (upstream) and 5′-AGGGTCGACTCAGTATGTAGC-3′ (downstream). The amplified fragment was first inserted into a pUC19 plasmid using EcoRI and SalI restriction sites introduced in the primers (single-underlined nucleotides). After the sequence was confirmed, the gene was moved into the pET-21c vector using NdeI (double-underlined nucleotide in the upstream primer) and SalI restriction sites.

Heterologous expression and purification

Escherichia coli BL21-Codon-Plus (DE3)-RIL transformed with the constructed expression plasmid was inoculated into an Luria–Bertani medium containing 50 μg·mL−1 of ampicillin and 34 μg·mL−1 of chloramphenicol. After the cells were cultivated aerobically at 37 °C until D600 reached 0.6, the expression of htSHMT was induced with 0.5 mm of isopropyl thio-β-d-galactopyranoside for 4 h. The cells were then harvested, suspended in 20 mm of a Tris/HCl (pH 8.0) buffer (buffer A; 4 mL·g−1 of the wet cells) and sonicated. The homogenized sample was heated to 70 °C for 10 min and centrifuged at 100 000 g for 30 min. The supernatant was loaded on a Q Sepharose fast flow column (GE Healthcare, UK) equilibrated with buffer A. The protein was eluted with ∼ 0.2 m NaCl using a gradient with buffer A containing 1 m NaCl (buffer B). The fraction containing the protein was pooled, dialyzed against buffer A, and loaded on a Mono Q column (GE Healthcare) equilibrated with the same buffer. The protein was eluted with ∼ 0.2 m NaCl using a gradient with buffer B. All purification steps were performed at room temperature, except for dialysis, which was done at 4 °C.

Estimation of molecular mass by gel filtration

Gel filtration was performed using a Shim-pack Diol-300 column (Shimadzu, Kyoto, Japan) equilibrated with buffer A containing 150 mm of NaCl. The gel-filtration standard (Bio-Rad, Hercules, CA, USA) was used as a molecular marker for calibration. Each sample was measured in triplicate.

Protein assay

Protein concentration was determined using the Bio-Rad Protein Assay kit (Bio-Rad) with bovine serum albumin as the standard.

Enzyme assays

Several procedures were used to determine the activities of htSHMT. The THF-dependent SHMT activities were detected by measuring the production of amino acids. The production of acetaldehyde was measured to detect the THF-independent aldolase activities toward l-threonine and l-allo-threonine. The decrease in (allo-) threonine and the increase in glycine as a product were also checked. When threo- and erythro-β-phenylserine were used as the substrates, the activities were determined by measuring the production of benzaldehyde. One unit of activity was defined as the amount of enzyme that catalyzed the formation of 1 μmol of the product (glycine, serine, acetaldehyde or benzaldehyde) per minute. For the assays, 20 mm of Tris/HCl, Hepes/NaOH or sodium phosphate buffer was used. Here, the pH values for Tris/HCl and Hepes/NaOH were adjusted at room temperature so that the values become 7.5 at 70 and 75 °C, respectively. The pH values at these temperatures were calculated using the temperature coefficients (dpKa/dT) of the buffers (−0.031 for Tris/HCl and −0.0028 for Hepes/NaOH). The pH of the sodium phosphate buffer was adjusted to 7.2 at room temperature. PLP (100 μm) was added if necessary. For the glycine- and serine-producing THF-dependent reaction, 2.2 mg·mL−1 of THF and 0.25 mg·mL−1 of 5,10-methylene-THF were added, respectively, along with 10 mm of mercaptoethanol. The reactions were carried out for 5–10 min and were stopped on ice. The products were detected and quantified as follows. Amino acids were identified and quantified using HPLC with a reverse-phase column after the phenylthiocarbamyl derivatization [32] described previously [18]. Acetaldehyde was detected photometrically by measuring absorption at 306 nm after the derivatization [33], in which < 5 μL of the reaction mixture was incubated with 50 mm of glycine/NaOH (pH 4.0) buffer containing 0.01 (w/w) N-methyl benzothiazolone hydrochloride (total amount: 100 μL) at 40 °C for 5 min. Benzaldehyde was measured photometrically, using the molar extinction coefficient, ε279, of 1400 m−1·cm−1 [20]. Km and Vmax were calculated using kaleidagraph 4.0J (Synergy Software, Reading, PA, USA).

Rate constants were calculated from the initial reaction velocities. They were measured at different temperatures of 25, 40, 50, 60, 70 and 75 °C with 10 mm l-allo-threonine, 10 mm dl-β-threo-phenylserine, 10 mm dl-erythro-phenylserine or 20 mm l-threonine. These concentrations are high enough when compared with their Km values. Here, a potassium phosphate buffer was adopted so that we could compare the results obtained from htSHMT with those from ecSHMT and mjSHMT [20]. The logarithms of the rate constants were plotted against the inverse absolute temperatures. The activation energies were calculated from the slopes of the plots using the Arrhenius equation,


where k is the rate constant, Ea is the activation energy, and R is the gas constant. Here, T is the absolute temperature and A is the pre-exponential factor.


This work was supported in part by a grant-in-aid for scientific research (A) (21248010) from the Japan Society for the Promotion of Science (JSPS), and a grant-in-aid for JSPS fellows (23-3030).