Biochemical Journal

Research article

Mutational analysis of residues in the regulatory CBS domains of Moorella thermoacetica pyrophosphatase corresponding to disease-related residues of human proteins

Joonas Jämsen, Heidi Tuominen, Alexander A. Baykov, Reijo Lahti


mtCBS-PPase [CBS (cystathionine β-synthase) domain-containing pyrophosphatase from Moorella thermoacetica] contains a pair of CBS domains that strongly bind adenine nucleotides, thereby regulating enzyme activity. Eight residues associated with the CBS domains of mtCBS-PPase were screened to explore possible associations with regulation of enzyme activity. The majority of the substitutions (V99A, R168A, Y169A, Y169F, Y188A and H189A) enhanced the catalytic activity of mtCBS-PPase, two substitutions (R170A and R187G) decreased activity, and one substitution (K100G) had no effect. AMP-binding affinity was markedly decreased in the V99A, R168A and Y169A mutant proteins, and elevated in the R187G and H189A mutant proteins. Remarkably, the R168A and Y169A substitutions changed the effect of AMP from inhibition to activation. The stoichiometry of AMP binding increased from one to two AMP molecules per CBS domain pair in the Y169F, R170A, R187G and Y188A variants. The ADP-binding affinity decreased in three and increased in four mutant proteins. These findings identify residues determining the strength and selectivity of nucleotide binding, as well as the direction (inhibition or activation) of the subsequent effect. The data suggest that mutations in human CBS domain-containing proteins can be translated into a bacterial context. Furthermore, our data support the hypothesis that the CBS domains act as an ‘internal inhibitor’ of mtCBS-PPase.

  • adenine nucleotide
  • allosteric regulation
  • cystathionine β-synthase domain (CBS domain)
  • inorganic pyrophosphatase
  • Moorella thermoacetica
  • mutagenesis


In Bacilli, Clostridia and other bacterial lineages, including several human pathogens, the intracellular PPi level is controlled by family II PPases (pyrophosphatases) (EC that convert PPi into Pi [1,2]. Approx. one-quarter (67) of known family II PPase sequences, mostly from Clostridia and sulfate-reducing δ-proteobacteria, contain a large insert (~100–125 residues) composed of two CBS (cystathionine β-synthase) domains that form a Bateman fold [3]. This structure, named after cystathionine β-synthase from where it was identified originally [3], binds regulatory adenine nucleotides and S-adenosylmethionine and is present in a variety of proteins [4,5]. We recently expressed the CBS domain-containing PPase of the bacterium Moorella thermoacetica (mtCBS-PPase) for the first time, and analysed enzyme regulation by kinetic and binding measurements [6,7]. The enzyme binds adenine nucleotides in the submicromolar range, is activated by ATP and is inhibited by AMP and ADP [6]. Stopped-flow kinetic measurements using a fluorescent AMP analogue revealed that nucleotide binding to mtCBS-PPase occurs in three steps with relaxation times from 0.01 to 100 s, implying conformational changes in the complex [7]. Nucleotide binding was modulated by the substrate (PPi), which additionally induced enzyme conversion into a more active form within a timescale of minutes [7]. Recently, we expressed the CBS domain-containing PPase of Clostridium perfringens (cpCBS-PPase) and determined the structure of the regulatory region comprising two CBS domains [8].

According to the Pfam database [9], CBS domains have been identified in nearly 30000 protein sequences from all kingdoms of life. In addition to the role played by CBS domains in cystathionine β-synthase [3], such domains are implicated in the control of intracellular trafficking by the chloride channel CLC-5 [10]; gating of the osmoregulatory transporter OpuA [11,12]; protein phosphorylation by AMP-dependent protein kinase [4]; Mg2+ transfer by the transporter MgtE [13]; plant vacuolar nitrate transport by the transporter AtCLCa [14]; and adenylate nucleotide synthesis (IMP dehydrogenase) [15]. Single mutations in the CBS domains of these proteins are linked to a variety of human hereditary disorders, including homocystinuria, Wolff–Parkinson–White syndrome, retinitis pigmentosa and Bartter syndrome [4,5]. However, little is known about the mechanism underlying the transduction of regulatory inhibition or activation signals from CBS domains to the substrate-binding regions of these proteins. Most crystal structures of CBS domains have been obtained for small proteins of unknown function containing only CBS domains, or separate regulatory regions of multidomain proteins, which reveal no significant changes upon binding of nucleotides. Although the crystal structures of several complex proteins containing regulatory CBS domains are available [1618], pairs of structures for activated/inhibited or nucleotide-bound/nucleotide-free forms have not been solved for any protein to date. Large structural differences in CBS domains have been observed in only two instances, dependent on the bound nucleotide: with the MJ0100 protein of unknown function from Methanocaldococcus jannaschii [19] and the regulatory part of cpCBS-PPase [8]. The cpCBS-PPase structures display marked conformational changes upon binding of activator (diadenosine tetraphosphate) compared with inhibitor (AMP) to the regulatory CBS domains, indicating that the CBS-PPase (CBS domain-containing PPase) is a promising model in which to explore the mechanism of regulation of CBS proteins.

In the present paper, we report the effects of point mutations in the CBS domains of mtCBS-PPase on interactions with adenine nucleotides, with a view to identifying residues that determine the strength and selectivity of nucleotide binding as well as the direction of the effect (inhibition or activation). Additionally, we have characterized the effects of substitutions on the substrate- and nucleotide-induced activity transitions observed in previous studies [7]. Residues were selected for substitution based on approximate correspondence with physiologically disruptive mutations in other CBS domains and the X-ray structure of homologous cpCBS-PPase [8].



Restriction enzymes and T4 DNA ligase were obtained from Fermentas, and Pfu Turbo DNA polymerase was purchased from Stratagene. Oligonucleotides were purchased from MWG Biotech. AMP, ADP and ATP were from Fluka. [14C]AMP was acquired from Moravek Biochemicals and Radiochemicals. All other reagents and kits were obtained from Sigma.

Site-directed mutagenesis, protein expression and purification

Mutagenesis of the mtCBS-PPase gene was performed in pBluescript SK (Stratagene) using a QuikChange kit (Stratagene). Primer sequences used are shown in Supplementary Table S1 at Variant genes were cloned for expression into the pET36b+ plasmid (Novagen) using NdeI and XhoI restriction sites. Gene integrity and presence of the desired point mutations were confirmed via DNA sequencing. Expression and purification of mutant proteins was performed as described for wild-type mtCBS-PPase [6]. Protein concentrations were determined from absorbance at 280 nm (ϵ=22920 M−1·cm−1, calculated using ProtParam [20]). Protein expression level and purity of the final preparation were analysed by SDS/PAGE (8–25% gels) using the Phast system (GE Healthcare).

Activity measurements

Unless stated otherwise, activity measurements were performed at 25 °C in 25–40 ml of 100 mM Mops/KOH buffer (pH 7.2), including 0.1 mM CoCl2, 5 mM MgCl2 and 0.16 mM PPi. PPi hydrolysis was initiated by adding an aliquot (5–200 μl) of diluted enzyme solution to obtain 0.005–5 μg/ml final enzyme concentration. Pi formation was monitored for 20–25 min using a semi-automatic Pi analyser [21]. Stock nucleotide solutions used in kinetic measurements were prepared immediately before use and kept on ice.

Membrane filtration

[14C]AMP binding to mtCBS-PPase was determined with a filtration method [7] using Multiscreen 96-well opaque filtration plates with an Immobilon®-P membrane (Millipore). The membrane was washed with 20% ethanol using buffer supplemented with 1 mM unlabelled AMP, and twice with buffer without AMP. Next, 50 μl of 10 μM mtCBS-PPase solution was added and incubated for 10 min. Finally, 100 μl of 14C-labelled AMP (200–5000 c.p.m.) was added, the complete mixture was incubated for 10 min, and the wells were emptied using a vacuum pump (taking ~1 s to empty). Wells were washed twice with 0.1 ml of buffer, 0.2 ml of Ultima Gold XR scintillant was added, and the radioactivity adsorbed on the membrane was determined with a Chameleon V multimode microplate reader (Hidex). BSA was added instead of enzyme to estimate non-specific [14C]AMP binding.

Data analyses

All non-linear least-squares fittings were performed with the program Scientist (MicroMath). Dependence of activity (A) on nucleotide concentration ([L]) at a fixed substrate concentration was fitted to eqn (1), where A+L and A−L represent activities with and without bound nucleotide respectively, and Ki,app is the apparent nucleotide binding constant: Embedded Image

Time courses of PPi hydrolysis were fitted to eqns (2) and (3) [7], where A0, A and Alim represent velocities at time zero, t and infinity respectively, [P] is the product (Pi) concentration, [E]0 is the total enzyme concentration, and Km is the Michaelis constant. Eqn (3) assumes that activity increases with a first-order rate constant, k, to a constant level due to enzyme transition to a more active form during substrate conversion, and decreases upon substrate depletion. A0, Alim and k values were treated as adjustable parameters. The Km value, initially set at 10 μM at the initial round of calculation, was subsequently estimated from the calculated dependences of Alim on [S]0 (the initial substrate concentration), and the calculation cycle was repeated with the refined Km value until the results converged. Embedded Image Embedded Image


Selection and production of enzyme variants

Eight residues were selected for substitution in the two CBS domains of mtCBS-PPase, on the basis of (i) the positions of disease-causing mutations in CBS proteins (Figure 1A and see Supplementary Table S2 at, (ii) the identity of the ligands binding to nucleotides in cpCBS-PPase (Figure 1B), and (iii) the residue conservation profiles of CBS-PPases and other CBS proteins (Figure 1A). Residues corresponding to Lys100, Tyr169 and Arg170 are nucleotide ligands in the cpCBS-PPase structure (lysine, tyrosine and serine respectively) (Figure 1) and are well conserved (61%, 69% and 87% respectively) in 67 putative CBS-PPase sequences. Furthermore, residues at positions 168–170 form part of the RYRN (Arg-Tyr-Arg-Asn) sequence corresponding to the RYSN (Arg-Tyr-Ser-Asn) loop of cpCBS-PPase, which undergoes the most significant conformational change when the activator- and inhibitor-bound forms of cpCBS-PPase are compared [8]. Residues 187–189 are part of a conserved CBS domain motif. Finally, Val99 is located at the same relative position in the second CBS domain of mtCBS-PPase compared with Tyr169 of the first domain (Figure 1).

Figure 1 Selection of residues for mutation

(A) Sequence alignment of CBS domains 1 and 2 of two characterized CBS-PPases and CBS proteins with disease-causing mutations. Also shown are consensus sequences of CBS-PPases (based on 67 proteins) with 30% and 85% identity. Residue numbering is for mtCBS-PPase. Some sequences contain inserts compared with mtCBS-PPase; the number of residues in the insert is indicated in parentheses. Residues mutated in mtCBS-PPase are marked with asterisks, AMP-binding residues (side-chain contacts only) in cpCBS-PPase [8] are shown in italics, and disease-associated residues are shaded in grey. The box indicates a conserved CBS domain motif. Non-PPase CBS protein designations and UniProt identifiers: hCBS, human cystathionine β-synthase, P35520; hClC-7, human chloride channel ClC-7, P51798; hIMPDH1, human IMP dehydrogenase 1, P20839; hAMPKγ2, human AMP-activated protein kinase γ2, Q9UGJ0; hAMPKγ3, human AMP-activated protein kinase γ3, Q9UGI9; pigAMPKγ3, pig AMP-activated protein kinase γ3, Q9MYP4. UniProt identifiers for mtCBS-PPase and cpCBS-PPase sequences were Q2RFU3 and Q8XIQ9 respectively. For AMPKs containing two pairs of CBS domains per subunit, CBS domain order, beginning from the N-termini, are presented after protein names. (B) Stereo view of the nucleotide-binding site of cpCBS-PPase with bound AMP (PDB code 3L31) [8] showing the residues that were selected for mutation in the present study. Residue numbering is for cpCBS-PPase/mtCBS-PPase. The primed Lys100 belongs to the other subunit and forms an ion pair with Pα of AMP. Ser279/Arg170 is expected to interact with Pβ of ADP, and Tyr278/Tyr169 forms a stacking interaction with the adenine moiety [8].

Keeping in mind that replacement of disease-causing residues yields various effects in different proteins (Figure 1A and Supplementary Table S2), alanine is commonly used as a substituting residue to eliminate any effect of polar side-chain groups. In addition, Tyr169 was conservatively replaced with phenylalanine. Two arginine residues were mutated to glycine, because such substitutions are known to affect the function of pig AMP-dependent protein kinase γ3 linked to CBS domains [22]. For replacing Arg168, the choice between alanine and glycine was made on the basis of protein solubility, as the R168G variant showed a tendency to aggregate during purification. Expression levels for all mtCBS-PPase variants (~10% of total cell protein) were similar to that for wild-type enzyme, and variants were easily separated from other proteins, including Escherichia coli PPase. The final preparations were at least 90% pure as estimated by SDS/PAGE analysis (see Supplementary Figure S1 at

Catalytic activity

Earlier studies on mtCBS-PPase revealed ~2-fold activation by substrate, with a half-time of approx. 40 s, during PPi hydrolysis [7]. This effect was preserved in all variants studied, except for the Y169A and R170A proteins (Figure 2). Accordingly, the steady-state rate was attained after several minutes of reaction. The corresponding activity (Alim) was physiologically more relevant than was initial activity in estimation of the catalytic constants (kcat,lim values) and Michaelis constants (Km values) for variant enzymes. On the basis of the effect of mutations on kcat,lim, the variant enzymes were divided into three groups (Table 1): (i) those displaying kcat,lim values similar to that of the wild-type enzyme (K100G), (ii) more active enzymes (V99A, R168A, Y169A, Y169F, Y188A and H189A), and (iii) less active enzymes (R170A and R187G). The effects of mutation on Km were smaller and, in most instances, did not exceed the error of determination.

Figure 2 Time courses of PPi hydrolysis by wild-type mtCBS-PPase (WT) and selected variants

The assay medium contained 160 μM PPi. The reaction was initiated by adding an aliquot of enzyme solution pre-equilibrated at 25 °C. Curve labels refer to variant identity and final enzyme concentration in ng/ml.

View this table:
Table 1 Catalytic and nucleotide-binding parameters for wild-type and variant mtCBS-PPases

Values showing large deviations from wild-type enzyme are in bold. Parameter values for catalytic activity and nucleotide inhibition refer to substrate-activated enzyme. n.d., not determined.

Nucleotide inhibition

The effects of substitutions on activity regulation by AMP and ADP were significant (Figure 3). Substrate (PPi) concentration was fixed at 160 μM for these measurements, ensuring at least 85% saturation of enzyme active sites, based on the Km values listed in Table 1. Thus the observed effects of nucleotides on activity mainly reflect an influence on kcat,lim.

Figure 3 Effects of AMP (A) and ADP (B) on the equilibrium activity of wild-type mtCBS-PPase (WT) and selected variants

The assay medium contained 160 μM PPi. The lines show the best fit to eqn (1). The broken line, representing the effect of AMP on wild-type enzyme, is from Jämsen et al. [7].

Remarkably, AMP inhibited wild-type and the majority of variant enzymes, but activated the R168A and Y169A mutants (Figure 3, Table 1). Activation was particularly marked (2.8-fold) for the R168A protein. In keeping with these observations, the maximal degree of inhibition, characterized by the residual activity of an enzyme–AMP complex, was increased in the Y188A and H189A variants compared with wild-type enzyme (Table 1). The AMP-binding affinity, estimated using Ki,app, was not significantly affected in the Y169F and R170A mutants; somewhat decreased in the K100G and Y188A substituents; markedly less in R168A, Y169A and, particularly, V99A; and significantly elevated in the R187G and H189A mutant proteins.

The effects of ADP were more uniform in that all variants were inhibited more than 5-fold (Table 1). However, the Ki,app values showed more variation. Specifically, increases were seen for the V99A, Y169A and R170A variants, and falls were noted for other variants, except for R168A and Y188A. In most instances, the effects of ADP on Ki,app paralleled those of AMP.

Equilibrium AMP binding

Direct measurements of [14C]AMP binding to mtCBS-PPase and variants thereof in the absence of substrate were performed using the membrane filtration technique (Figure 4). In an earlier study, we presented evidence that, in wild-type mtCBS-PPase, AMP binding to a dimeric enzyme is characterized by negative co-operativity [7]. Unfortunately, the binding curves obtained from filtration assays showed some scatter, precluding accurate analysis of curve form, which was subsequently approximated as a simple two-parameter hyperbola. Furthermore, the membrane filtration assay may underestimate the binding affinity because of partial dissociation of the complex during the removal of unbound AMP, which occurs in ~1 s. This may explain why the Kd values obtained were markedly larger than those of Ki,app derived from kinetic assays. However, regardless of the actual significance of filtration-derived Kd values, a marked increase upon residue substitution, as observed with V99A, K100G, Y169A, Y188A and H189A proteins (Table 1), is clearly indicative of altered (probably weakened) AMP binding. Another important finding is that the stoichiometry of AMP binding increased from 1 to 2 mol/mol monomer in the Y169F, R170A, R187G and Y188A variants, implying occupancy of both tandem CBS domains.

Figure 4 [14C]AMP binding to wild-type mtCBS-PPase (WT) and selected variants, determined using membrane filtration

The abscissa depicts the concentration of free unbound AMP, calculated as the difference between total and bound AMP concentrations. The lines represent the best fit to eqn (1) with AL set to zero.

Activity transition

In the absence of substrate, wild-type mtCBS-PPase exists as a mixture of two conformations with different activities, and is converted into the more active conformation on the timescale of the enzyme assay upon addition of substrate [7]. Nucleotides bound to wild-type mtCBS-PPase slowed this transition. Similar analyses were performed for the variant enzymes. Fitting of eqns (2) and (3) to the time courses of PPi hydrolysis (Figure 2) yielded both a ratio of initial to final activities (kcat,0/kcat,lim) (Table 2) and a rate constant for the substrate-induced transition (ktr). For most variants, the kcat,0/kcat,lim ratio was not significantly different from that of wild-type. The rate constant for transition in the absence of nucleotides was 1.5–3-fold less compared with wild-type enzyme, and nucleotide addition further decelerated the transition (Figure 5), analogous to what was seen with wild-type PPase. Conversely, the Y169A and R170A variants exhibited no activity transition in the absence of nucleotides (kcat,0/kcat,lim=1). However, this transition was clearly accelerated in the presence of AMP or ADP. Another interesting finding was that ADP accelerated transition in the Y188A variant. The lack of effect of ADP on the ktr,lim value of the V99A variant appears to result from weak nucleotide binding (Table 1).

Figure 5 Effects of AMP (A) and ADP (B) on the rate constant for substrate-induced activity transition in wild-type mtCBS-PPase (WT) and selected variants

The assay medium contained 160 μM PPi. The lines represent the best fit to eqn (1).

View this table:
Table 2 Parameters describing activity transition in wild-type and variant mtCBS-PPases

Values showing large deviations from wild-type enzyme are presented in bold.


CBS domain pair as an internal inhibitor in CBS-PPase

The specific activities of two CBS-PPases characterized to date, those of M. thermoacetica [6] and C. perfringens [8], are two to three orders of magnitude lower than those reported for common family II PPases lacking CBS domains [1,2,23]. Previously, we hypothesized that CBS domains act as ‘intrinsic inhibitors’ in CBS-PPases [6], supported by the observation that the activity of CBS-PPase and other CBS proteins may either decrease or rise, depending on the nature of the bound nucleotide. These allosteric effects are clearly mediated by conformational changes in CBS domains, and it is possible that other factors causing similar structural changes in CBS domains also affect PPase activity. The activation effect is particularly interesting, as it may provide relief against the ‘inhibition’ induced by CBS domains.

The present results provide further support for the above theory. First, five of nine substitutions in CBS domains resulted in enhanced activity (as measured by kcat,lim values; Table 1). Although functional improvements in proteins after artificial mutation have been documented in the literature, function-disrupting mutations are far more common. This observation reflects the general dogma that functions tailored by nature are difficult to improve upon, but easy to damage. Our findings are in line with this belief if the apparent ‘improvement’ in function (activation) actually results in a disruption of the autoinhibition mechanism underlying the regulatory activity of the CBS domains.

Secondly, the ‘internal inhibitor’ mechanism is supported by reversal of the AMP effect from inhibition to activation in the R168A and Y169A substitutions. To the best of our knowledge, no change in the direction of a nucleotide effect after amino acid substitution has ever been reported for any CBS protein. The possibility of such reversal implies high plasticity of the regulatory mechanism. Importantly, in the Y169A variant, the basal kcat,lim value in the absence of nucleotides (5.0 s−1) significantly exceeded that of wild-type mtCBS-PPase (2.2 s−1). The combined effect of mutation and AMP binding led to a 2.8–3.5-fold increase in kcat,lim values in the R168A and Y169A variants compared with the wild-type basal value. Most constructed mutations thus widen the range of activities attainable in the absence and presence of nucleotides, fully consistent with the ‘internal inhibitor’ hypothesis.

Nucleotide-binding stoichiometry

Each CBS domain of the Bateman fold contains a potential binding site for nucleotides. However, cpCBS-PPase structures [8], most other reported structures [2426] and those found only in the PDB (3LFR, 3FWR, 3FWS, 3FNA, 2YZQ and 3DDJ) contain only one nucleotide molecule per two CBS domains. However, both sites are occupied by nucleotides in a few reported protein structures [19,2729].

Wild-type mtCBS-PPase is similar to cpCBS-PPase in that the enzyme binds only one nucleotide per two CBS domains (Table 1). Remarkably, four substitutions in CBS2 elevated the AMP-binding stoichiometry of mtCBS-PPase to two nucleotides per CBS domain pair (Table 1). Two of the residues replaced (Tyr169 and Arg170) correspond to cpCBS-PPase AMP ligands, and two (Arg187 and Tyr188) belong to a conservative CBS motif. One possible explanation is that mutations in CBS2 induce conformational changes in this domain, inhibiting interactions with the CBS1-bound nucleotide and facilitating intrinsic nucleotide binding. Interestingly, the Ki,app value was markedly decreased in the R187G variant, indicating tighter binding. This may indicate that, at least in this variant, the Ki,app value is referable to the CBS2-bound nucleotide. Gómez-García et al. [29] recently reported unusual binding of AMP to the linker sequence connecting the two CBS domains in a CBS protein from M. jannaschii. However, this type of binding is unlikely in mtCBS-PPase, because the linker sequence is relatively short.

Possible roles of mutated residues in regulation

The effects of substitutions on nucleotide binding affinity are easily explainable in terms of changes in nucleotide contact (group 1 residues), whereas effects on activity may be attributable to closure/opening of the CBS domain interface (group 2 residues), as seen in cpCBS-PPase structures [8]. Group 1 residues include Lys100, Tyr169 and Arg170. The finding that AMP binding is weakened and ADP binding strengthened upon K100G substitution is consistent with the theory that Lys100 interacts with P1 of the nucleotides [8]. Tyr169 stacks on to the adenine ring in cpCBS-PPase structures [8], and alteration to alanine is thus expected to weaken binding, leading to malpositioning of the bound nucleotide, in turn facilitating opening of the CBS domain interface with concomitant enzyme activation. An opposite effect was predicted and indeed confirmed for the Y169F mutant. On the basis of the cpCBS-PPase structure [8], Arg170 is expected to play a role in ADP P2 binding. Indeed, the Ki,app values obtained for the R170A substitution indicate that this residue is clearly important for ADP, but not AMP, binding.

Group 2 residues, which do not interact with bound nucleotide in the cpCBS-PPase structure [8], include Arg168, Arg187 and Tyr188. Arg168 is located at the CBS domain interface within the RYRN sequence. The loop adopts markedly different conformations in activated and inhibited cpCBS-PPase, and may therefore play a central role in transmitting the regulatory signal to catalytic domains. This hypothesis is confirmed by the finding that mutation of both residues of this loop region reversed the effect of AMP from inhibition to activation. The importance of Arg187, a highly conserved (90%) residue in CBS-PPase, is explained by the interactions of this residue with the RYRN loop. Interestingly, the conformations adopted by Arg187 in inhibited and activated cpCBS-PPase structures [8] are markedly different. Similar data were obtained for Tyr188, which is displaced from the domain interface to a hydrophobic cavity when a complex is formed with activating nucleotide. After Y188A substitution, the domain interface is expected to open, resulting in increased activity of the variant enzyme.

In summary, six CBS-domain residues, including four corresponding to disease-related residues of human proteins, are important in adenine nucleotide regulation of CBS-PPase. Together with the previously determined structures of the inhibitor- and activator-bound forms of the regulatory domain of cpCBS-PPase, our results significantly clarify the mechanism underlying the regulatory effect of the CBS domain. Elucidation of the detailed mechanism of the cooperation between regulatory and catalytic domains is the next challenging step, and the results of the present study will serve as a basis for this work.


Joonas Jämsen and Heidi Tuominen designed and conducted experiments. Alexander Baykov and Reijo Lahti supervised the study and co-wrote the paper.


This work was supported by the Academy of Finland [grant number 114706], the Russian Foundation for Basic Research [grant number 09-04-00869], and the Ministry of Education and the Academy of Finland (for the National Graduate School in Informational and Structural Biology).


We acknowledge Dr Georgiy Belogurov for mutant design and the initial concept as well as Juho Vuononvirta and Minna Peippo for expression and purification of the mutant proteins.

Abbreviations: CBS, cystathionine β-synthase; CBS-PPase, CBS domain-containing pyrophosphatase; cpCBS-PPase, CBS-PPase from Clostridium perfringens; mtCBS-PPase, CBS-PPase from Moorella thermoacetica; PPase, pyrophosphatase


View Abstract