Crystal structure of cystathionine β -synthase from honeybee Apis mellifera

Cystathionine β -synthase (CBS), the key enzyme in the transsulfuration pathway, links methionine metabolism to the biosynthesis of cellular redox controlling molecules. CBS catalyzes the pyridoxal-5 ′ -phosphate-dependent condensation of serine and homocysteine to form cystathionine, which is subsequently converted into cysteine. Besides maintaining cellular sulfur amino acid homeostasis, CBS also catalyzes multiple hydrogen sul ﬁ de-gen-erating reactions using cysteine and homocysteine as substrates. In mammals, CBS is activated by S-adeno- sylmethionine (AdoMet), where it can adopt two di ﬀ erent conformations (basal and activated), but exists as a unique highly active species in fruit ﬂ y Drosophila melanogaster . Here we present the crystal structure of CBS from honeybey Apis mellifera , which shows a constitutively active dimeric species and let explain why the enzyme is not allosterically regulated by AdoMet. In addition, comparison of available CBS structures unveils a substrate- induced closure of the catalytic cavity, which in humans is a ﬀ ected by the AdoMet-dependent regulation and likely impaired by the homocystinuria causing mutation T191M.


Introduction
Transsulfuration is an ancient metabolic process that allows the interconversion of methionine (Met) and cysteine (Cys) through the common intermediates homocysteine (Hcy) and cystathionine (Cth) (Brosnan and Brosnan, 2006) (Fig. 1). In evolutionary terms, the transsulfuration consists in two routes, the "reverse" and the "forward" pathways (Carmel and Jacobsen, 2001). The reverse transsulfuration is found in a wide range of species, such as mammals and yeast, and converts Met into Cys (Brosnan and Brosnan, 2006) (Fig. 1). Some organisms, including enteric bacteria (Kredich, 1996;Auger et al., 2002), plants (Macnicol et al., 1981) and yeast (Cherest and Surdin-Kerjan, 1992), also possess the forward transsulfuration route that enables the formation of Met from Cys (Brosnan and Brosnan, 2006) (Fig. 1). Importantly, the presence or absence of these routes place different metabolic constraints on different organisms. For example, yeast can utilize either methionine or cysteine as a sulfur source, whereas humans are auxotrophic for Met, but are not for Cys.
Cystathionine β-synthase (CBS, EC 4.2.1.22), is the key enzyme in the reverse transsulfuration pathway (Mudd et al., 1965), and catalyzes the pyridoxal-5′-phosphate (PLP)-dependent condensation of serine and Hcy to form Cth and H 2 O (Fig. 1). The following second step in the route is mediated by another PLP-requiring enzyme, the cystathionineγ-lyase (CGL), that cleaves Cth into Cys, 2-oxobutyrate and ammonia (NH 3 ) (Banerjee et al., 2003;Miles and Kraus, 2004). The resulting Cys can either be used in protein synthesis or for the biosynthesis of glutathione (GSH, a potent antioxidant) (Beatty and Reed, 1980), taurine (an organic acid widely distributed in animal tissues and a major constituent of bile) (Stipanuk, 1986) or can be further catabolized into sulphate ( Fig. 1) (Brosnan and Brosnan, 2006;Prudova et al., 2006). Thus, CBS links methionine metabolism to the biosynthesis of cellular redox controlling molecules (Mudd et al., 1982;Welch and Loscalzo, 1998;Seshadri et al., 2002;Meier et al., 2003;Beyer et al., 2004;Mudd, 2011). Because of its pivotal role in the transsulfuration pathway, lack of CBS activity leads to classical homocystinuria (CBSDH, Online Mendelian Inheritance in Man (OMIM no. 236200)), an autosomal recessive inborn error of sulfur amino acid metabolism characterized by increased levels of Hcy in plasma and urine. Clinical symptoms of CBSDH manifest as a combination of connective tissue defects, skeletal deformities, vascular thrombosis, and mental retardation (Mudd et al., 2001). Remarkably, increased plasma Hcy concentrations are considered as a risk factor for dementia and Alzheimer's disease (Seshadri et al., 2002).
Besides maintaining cellular Hcy homeostasis, CBS also catalyzes alternative hydrogen sulfide (H 2 S)-generating reactions using Cys and Hcy as substrates (Fig. S1) , what converts this enzyme in the major physiological source of hydrogen sulfide. H 2 S plays a relevant role in the cardiovascular and nervous systems (Yadav and Banerjee, 2012;Paul and Snyder, 2012), induces smooth muscle relaxation, and has antiinflammatory and cytoprotective effects on cells (Szabó, 2007). Noteworthy, alterations of the H 2 S metabolism are linked with human diseases: in the brains of Alzheimer's disease patients H 2 S synthesis is decreased (Eto et al., 2002), whereas in Down syndrome patients H 2 S synthesis is increased due to the overexpression of CBS (Kamoun, 2004;Kabil and Banerjee, 2010). In turn, transsulfuration pathway-dependent H 2 S production was found related to dietary restriction-mediated longevity in yeast, worm, fruit fly, and rodent models, providing an interesting explanation for the long-sought relationship between slimness and longevity (Hine et al., 2015).  1. The transsulfuration pathway. The transsulfuration pathway, the metabolic route that allows the conversion of homocysteine into cysteine, is connected to the methionine cycle. Cystathionine β-synthase (CBS) is the first enzyme in the reverse transsulfuration pathway (in black arrows), playing pivotal role in deciding the fate of homocysteine. In some organisms, such as bacteria and yeast, cysteine can be converted into homocysteine by the forward transsulfuration pathway (grey dashed arrows), which is mediated by the cystathionine γ-synthase (CGS) and cystathionine βlyase (CBL) enzymes.
The H 2 S-production ability is not exclusive of eukaryotes but extends to bacteria as well, where the CBS and CGL genes are found clustered together (Matoba et al., 2017). Importantly, the presence of these genes is crucial for survival, as deletion of the CBS/CGL gene cluster or chemical inhibition of the encoded enzymes render pathogens such as Bacillus anthracis, Lactobacillus plantarum, Helicobacter pylori or Pseudomonas aeruginosa highly sensitive to a multitude of antibiotics (Shatalin et al., 2011).
The domain organization, quaternary structure and regulatory mechanism of CBS enzymes vary among species (Fig. 2). While most of the CBS enzymes form homotetramers like in humans (Ereño-Orbea et al., 2013a), rodents and yeasts (Jhee et al., 2000), we find homodimers in insects like fruit fly (Koutmos et al., 2010) or honeybee (Oyenarte et al., 2012), and monomers in worms (Vozdek et al., 2012) (Fig. 2). In higher eukaryotes, the N-terminal region includes a heme-binding domain ( Fig. 2) that is thought to function in redox sensing and/or enzyme folding (Janosík et al., 2001b;Singh et al., 2007;Majtan et al., 2010). The heme-binding domain is followed by a conserved catalytic core with the fold of the type II family of PLP-dependent enzymes (Christen and Mehta, 2001;Meier et al., 2001). Finally, the C-terminal region, also known as "Bateman module", consists of two consecutive "CBS domains" (Fig. 2) (Bateman, 1997;Baykov et al., 2011;Ereño-Orbea et al., 2013b;Anashkin et al., 2017) and exhibits the highest degree of sequence variability in CBS primary structures (Vozdek et al., 2012). Strikingly, some organisms like C. elegans lack this module (Fig. 2), which plays a key role in regulating the activity and oligomerization degree of many CBS enzymes. Interestingly, the presence of missense mutations or the artificial removal of this region not only activates the human (Kery et al., 1998;Janosík et al., 2001a), and yeast  CBS enzymes, but causes a disassembly of protein tetramers into homodimers (Meier et al., 2003;Kery et al., 1998). The molecular basis for all these observations has historically been delayed by the scarce availability of structural data, which until 2010 was limited to the catalytic core of human CBS (Meier et al., 2001; and to the full-length enzyme from Drosophila melanogaster (Koutmos et al., 2010). The long-sought crystal structure of hCBS (Ereño-Orbea et al., 2013aMcCorvie et al., 2014) recently opened a new scenario and showed how, in the lack of the allosteric regulator S-adenosylmethionine (AdoMet), the Bateman module occludes the entrance to the catalytic cavity, thus maintaining the enzyme in a basal, low activity state ( Fig. S2) (Ereño-Orbea et al., 2013a). It additionally revealed that binding of AdoMet to the Bateman module causes a relative rotation of its two CBS motifs that weakens their interaction with the loops configuring the entrance to the catalytic cavity, thus leading to the activated conformation of the enzyme (Fig. S2) (Ereño-Orbea et al., 2014). At the same time, Bateman modules from complementary subunits associate into an AdoMet-bound disk-like structure designated as CBS module that stabilizes an activated state (Ereño-Orbea et al., 2014) (Fig. S2). Such an activated state is structurally similar to that found in the fruit fly enzyme (Koutmos et al., 2010). Strikingly, the allosteric mechanism involving two different conformations (basal and activated) occurs only in mammals. CBS enzymes from less evolved eukaryotes, such as Drosophila melanogaster, only exist in a constitutively activated conformation ensuring a permanent access of substrates into the catalytic cavity (Koutmos et al., 2010).
Aimed to reduce the current structural gap existing in the CBS field, we describe herein the crystal structure of full-length CBS from honeybee Apis mellifera (AmCBS) at 3.2 Å resolution. These data provide new insights for understanding the molecular mechanisms involved in catalysis and allosteric regulation of CBSs, and may help to develop drugs to modulate CBS activity.

The heme domain
The heme binding domain of AmCBS is ten and forty residues shorter than the equivalent region in dCBS (Koutmos et al., 2010) and hCBS (Meier et al., 2001;) (Ereño-Orbea et al., 2013aMcCorvie et al., 2014), respectively (Fig. 4). It lacks secondary elements and embraces three helices of the catalytic core (α3, residues 77-92; α7, residues 185-202 and α8, residues 218-230) (Fig. S3). Its function remains enigmatic but the sequence and structural similarity with hCBS suggest that it likely fulfills a structural and/or a regulatory role Janosík et al., 2001a;Majtan et al., 2008;Weeks et al., 2009). The heme group is relatively surface exposed and is nested in a hydrophobic pocket formed by residues 7-24, helices α7 and α8 and the loop following the strand β6 (Figs. 4, S3). The iron in heme is axially coordinated by the sulfhydryl group of Cys12 and the N ε2 atom of His23. In turn, the sulfhydryl group of Cys12 forms additional polar interactions with the side chain of Arg225 and the main chain nitrogen of Trp14 (Fig. S3). The heme carboxylate groups are partially solvent accessible and participate in polar interactions with other residues like Arg8 or Tyr11.

The catalytic core
The central catalytic core of AmCBS (Fig. 3) is structurally similar to that found in the human (Meier et al., 2001;Ereño-Orbea et al., 2013aMcCorvie et al., 2014) and in the fruit fly (Koutmos et al., 2010) CBSs, and shows the overall fold of PLP-dependent enzymes (Fig. S4). Interestingly, the comparative analysis of all these enzymes revealed that this region is in turn composed by two distinguishable blocks: (i) a large static subdomain that in AmCBS includes amino acid residues 1-76 and 184-342 ( Fig. 4) (the equivalent residues in hCBS are 1-116 and 226-384, respectively; see Fig. S5) and (ii) a small moveable subdomain, which is intercalated in the larger block and includes residues 77-183 and 117-225 in AmCBS and hCBS, respectively (Fig. 4, S5). Both subdomains present an α/β fold and are linked in AmCBS by two loops formed by residues 70-77, that link strand β2 and helix α3, and 279-284, that are located between strand β6 and helix α7 (Fig. 4). The crevice formed between the static and the moveable subdomains (Fig. S4) accommodates the PLP cofactor, which is deeply buried in the cavity and resides as an internal aldimine, where the ε-amino group of Lys78 forms a Schiff base with aldehyde of PLP (Fig. S3). There are other hydrogen bonds between the nitrogen of the pyridine ring and the O γ of Ser307, and between the 3′-hydroxyl group of PLP and the N δ2 of Asn108. A highly conserved phosphate binding loop known to participate in catalysis and composed by residues Gly215, Thr216, Gly217, Gly218 and Thr219 in AmCBS, is located between strand β7 and helix α8 (Figs. S3 and S5). In AmCBS, the Fig. 3. Structure of the AmCBS protomer. The N-terminal domain (in green) with heme cofactor (spheres) precedes the catalytic core (in blue) that contains the PLP molecule (balls and sticks) at the catalytic site. The C-terminal Bateman module (in yellow) includes two CBS motifs (CBS1, CBS2) and is linked to the core through a long linker (in red). Two main (empty) cavities, S1 and S2, are formed between the central β-sheets of the two CBS motifs.  entrance of the catalytic cavity is defined by four loops that include residues 104-107, 128-134, 151-161 and 253-274. The first three loops are located in the moveable subdomain, while the fourth loop belongs to the larger static subdomain. In our AmCBS crystals, residues 252-254 of the fourth loop are not visible in the electron density map, suggesting a high mobility of this zone in the absence of bound substrates, as it has previously been described in both hCBS (Meier et al., 2001) and dCBS (Koutmos et al., 2010).

The Bateman module of AmCBS does not host AdoMet
The Bateman module within the C-terminal domain is tethered to the catalytic core by a long linker (residues 341-370) (Figs. 3 and 5) and consists of two CBS motifs (CBS1: 369-430; CBS2: 437-504) that exhibit a α13-α14-β11-β12-α15 and a α16-β13-α17-β14-β15-α18 fold, respectively (Figs. 3-5). Each short N-terminal helix (α13 or α16) forms an integral part of the other CBS motif by antiparallel packing between its C-terminal β-strand (β12 or β15) and the α-helix (α18 or α15), so that both CBS motifs form a nested structure with pseudo-C 2 symmetry (Fig. 5). The two CBS motifs interact with each other via their two-or three-stranded β-sheets, and both long edges of this bilayer interface form two major cavities (designated S1 and S2) (Figs. 4 and 5). Importantly, the chemical-physical properties of sites S1 and S2 lack key features to host nucleotides thus explaining why, in contrast with mammals, insect CBS enzymes do not bid and are not regulated by AdoMet. Among these features is, for example, the lack of a conserved aspartate at the first turn of the α-helix following the last β-strand of each CBS domain (Figs. 6, S6), which is crucial to stabilize the orientation of the ribose ring of the nucleotide within the cavity through the interaction with its hydroxyl groups (Ereño-Orbea et al., 2013a,b, 2014. In AmCBS, the position of this aspartate is occupied by a lysine (K422) or by a histidine (H487) in sites S1 and S2, respectively (Figs. 6, S6). In addition, the hydrophobic cage required to accommodate the adenine ring of the nucleotide (Ereño-Orbea et al., 2013b;Baykov et al., 2011;Anashkin et al., 2017) is only partially present in site S1 (residues Y467, V443, V447, V468) and is completely absent in cavity S2, which Fig. 5. Structure of the AmCBS dimer. AmCBS (left) exists as a tight dimer in which the two protomers interact through residues located at both, the catalytic and the regulatory region similarly to dCBS (right) (Koutmos et al., 2010). In both proteins, the Bateman modules (Bat-1 and Bat-2) from complementary subunits associate into a disk-like CBS module. In this conformation, the entrance to the catalytic cavity of each subunit is open and accessible, thus yielding a highly active species. Each Bateman module contains two cavities (S1 and S2) that, in contrast with the human enzyme, are not able to bind AdoMet. Heme and PLP are depicted in spheres and sticks, respectively. is occupied by polar residues (N484, N399, D401, S402, Q403) (Fig. 6). Similar characteristics can be observed in dCBS (Koutmos et al., 2010) (Fig. 6), which has very high basal activity (Fig. S7) (i.e. is constitutively active) and does not bind nor is regulated by AdoMet .
Of note, the Bateman module of AmCBS does not contact the catalytic core except via the connecting linker (Figs. 3 and 5, Movie S1). This arrangement helps maintaining a concrete distance between the CBS2 motif of the Bateman module and the loops defining the entrance of the active site cavity. The CBS1 motif also remains far apart from the protein core with no elements in between (Fig. 3). Among the main interactions between the linker and the CBS2 domain is a salt link between residues E350 (at helix α11) and R460 (at helix α17). The position of the α-helical region of the linker is supported by hydrophobic interactions between Y347 (α11) and the alkyl chains of residues R460 (at helix α17) and K464. The linker maintains several hydrophobic interactions with the catalytic core through residues M349 (α11), I336, Y339, F343, V344, L355, and R294 (α9), I297 (α9) and L303. Fig. 6. Sites S1 and S2 in AmCBS, hCBS and dCBS. The figure shows the main residues located at (A) site S1 and (B) site S2 of AmCBS (left, in yellow), hCBS (middle, in grey) and dCBS (right, in orange), respectively. AdoMet at site S2 of hCBS is represented in orange sticks. Residues from complementary subunits are in blue sticks, indicated with #. The presence of an aspartate residue as well as of a threonine at the equivalent position of residues D538 and T535 of hCBS (marked in panels B and C with red and black asterisks, respectively), is a known key feature to host adenosine derivatives in the canonical cavities of CBS domains (Baykov et al., 2011;Ereño-Orbea et al., 2013b) (Supp. Fig. S6). (C) Sequence alignment of the main amino acid residues configuring the walls of sites S1 and S2 in AmCBS, hCBS and dCBS. The nucleotide binding motif G-h-h-T/S-x-x-D/N usually found in CBS domains that host adenosine derivatives (where ''h'' is hydrophobic, ''x'' is any residue, T/S is a threonine or a serine residue and D/N is an aspartate /asparagine residue) (Ereño-Orbea et al., 2013b) (see also Supp. Fig. S6), is indicated underneath the third block of aligned residues. The secondary elements that contain the corresponding residues in AmCBS are indicated above the alignment. Movie S1. Similarly to dCBS (Koutmos et al., 2010), AmCBS associates into tight dimers that represent the functional biological unit. Each subunit shares a large interface (3282 Å 2 ) with the complementary subunit with extensive contributions from the central core (1861 Ǻ 2 ) and the Bateman module (1316 Ǻ 2 ) (Fig. 5). This interface is mainly hydrophobic with hydrogen bonds and no salt bridges between the two protomers. A pair of four-helix bundles forms the interface (α14 and α15 from CBS1 of protomer A interact with α17 and α18 from CBS2 of protomer B, and α17 and α18 from CBS2 of protomer A with α14 and α15 from CBS1 of protomer B) (Fig. 5). Remarkably, the two Bateman modules from the complementary subunits associate through their helix bundles to configure an antiparallel disk-like CBS module (Fig. 5) (Baykov et al., 2011;Ereño-Orbea et al., 2013b;Anashkin et al., 2017). Such arrangement is rare among CBS domain proteins as Bateman modules usually associate into parallel CBS modules; however, it is observed in structures of all full-length CBS enzymes solved so far (Koutmos et al., 2010;Ereño-Orbea et al., 2014). It imposes a physical separation between the Bateman module and the entrance of the catalytic cavity and permits free access of substrates into the catalytic site (Fig. 5). Thus, our crystals contain constitutively active dimers of AmCBS.

Discussion
The crystal structure of AmCBS described herein provides the third three-dimensional structure of a full-length CBS enzyme (second from an insect) containing a regulatory Bateman domain available to date. Two additional structures of full-length CBS enzymes from Lactobacillus plantarum (PDB code 5BIH) (Matoba et al., 2017) and from Bacillus anthracis (PDB code 5XW3) (Devi et al., 2017) have been deposited recently, although the corresponding protomers do not include a Bateman module in their amino acid sequences (Fig. 2). The species present in our crystals correspond to highly active dimers (likely constitutively activated), (Fig. 5, Movie S1). Similar conformation and consequences have also been observed for dCBS (Koutmos et al., 2010;Majtan et al., 2014). The main cavities (S1 and S2) in the Bateman module of AmCBS lack key residues and characteristics usually required to host nucleotides or their structural analogs, as shown for the human enzyme (Ereño-Orbea et al., 2014;McCorvie et al., 2014) and other CBS domain proteins of unrelated function (Baykov et al., 2011;Ereño-Orbea et al., 2013b) (Figs. 6, S6). Therefore, AdoMet, the allosteric activator of the mammalian enzyme (Ereño-Orbea et al., 2014;McCorvie et al., 2014), cannot bind and consequently does not regulate the AmCBS activity (Fig. S7). It seems clear that the capability of CBS to adopt two different conformations, the basal (of low activity) and the activated, is exclusive to mammals and appeared later in evolution (Kabil et al., 2011).
The structural data on CBS enzymes obtained during the last decade revealed a significant resemblance between the catalytic core of CBSs and the β-family of PLP-dependent enzymes (Fig. S4). However, the difficulties found to crystallize full-length CBS enzymes in the absence and in the presence of their multiple ligands have prevented to prove with certainty whether CBSs suffer substrate-induce conformational changes analogous to those reported for the related PLP-dependent enzymes (Raj et al., 2013). For example, binding of methionine to conserved residues surrounding the active site of O-acetyl serine sulfhydrylase (OASS) (evolutionary the most closely related PLP-dependent enzyme to CBS) results in the movement of the N-terminal domain and the closure of the active site (Raj et al., 2013). Similar changes were observed in threonine deaminase (TD) (Hyde et al., 1988) or tryptophan synthase (TS) (Rhee et al., 1996). By analyzing all the available structural information on CBS enzymes, we found that the moveable subdomain of CBS catalytic core participates in such substrate-induced structural change (Fig. 7). Of note, in OASS the majority of the substrate-to-protein hydrogen bonding interactions affect the residues located in two conserved loops: the "Asn loop" (85-TSGNT-89) from the N-terminal domain and the "Gly loop" (236-GIGA-239) from the Cterminal domain (marked with asterisks in Fig. S5). In this protein, the largest conformational change observed in the substrate-bound state is represented by residue S86 (equivalent to S106 in AmCBS and S147 in hCBS), which shifts around 6 Å to make contacts with the substrate methionine in the active site (Raj et al., 2013). Although it has not been credited as important as the Asn loop, some additional elements including strands β4 to β7, helices α6, α7 and loops 85-88 and 130-133 of OASS (all belonging to the small subdomain), modify their conformation concomitantly (Raj et al., 2013). Our comparative analysis ( ) also revealed that strands β4 to β7, loops L171-174 and L191-202, as well as helices α6 and α7 (comprising a major part of the moveable subdomain), vary their orientation in the activated state with respect to the basal conformation (Fig. 7). Of note, we have noticed that helices α4 and α5 of hCBS remain unaltered and anchor the moveable motif to the static subdomain. These observations indicate that the inhibitory effect exerted by the regulatory Bateman module of hCBS in the basal state (Ereño-Orbea et al., 2013a) is not determined solely by a closure of the loops defining the entrance of the catalytic cavity, as we initially thought (Ereño-Orbea et al., 2014), but by the compression of a major part of the moveable subdomain of the protein core that behaves as rigid body. Moreover, an equivalent whole-motif displacement is observed in dCBS when the structure of the native protein (PDB ID 3PC3) is superimposed with its corresponding substrates-bound complexes (PDB codes 3PC3, 3PC4) (Koutmos et al., 2010). As shown in Figs. 7 and 8, binding of substrates into the catalytic cavity of dCBS promotes the movement of the entire moveable motif, and not of just the entrance loops, as was formerly proposed (Koutmos et al., 2010). In the same way, it can be shown that the effect of substrate binding in OASS is not limited to the displacement of a single loop (Raj et al., 2013), but involves a shift of a region equivalent to the moveable motif of hCBS (Fig. 7). Based on these observations and despite no crystal structure of hCBS in complex with its substrates is available so far, it is reasonable to postulate that there are two circumstances that trigger a displacement of the moveable motif and the consequent closure of the catalytic cavity in the human enzyme: (i) the presence of the Bateman module above the catalytic cavity (as seen in the basal state) and (ii) the presence of bound substrates at the PLP site. Interestingly, in constitutively active CBS enzymes, such as dCBS or AmCBS, where the Bateman module never interacts with the catalytic core, the closure of the moveable motif appears to exclusively dependent on the presence of bound substrates inside the catalytic cavity. In agreement with this hypothesis, the moveable motif of AmCBS shows an open state in our crystals equivalent to that found in apo-dCBS (PDB code 3PC2) (Koutmos et al., 2010) (Fig. 8).
Interestingly, twelve of the 160 pathogenic mutations described in homocystinutic patients (http://cbs.lf1.cuni.cz/mutations.php) affect residues that are located in the moveable submotif (Fig. 9). This group includes the mutation T191M that is prevalent in the Iberian Peninsula and South America, (Urreizti et al., 2003(Urreizti et al., , 2006aDe Lucca and Casique, 2004;Porto et al., 2005;Bermúdez et al., 2006;Hnízda et al., 2012;Alcaide et al., 2015). The T191M variant is structurally unstable and shows decreased catalytic activity and higher susceptibility to an accelerated proteasome-dependent degradation . Several explanations have been proposed over the years for the effect of the T191M mutation on the hCBS activity (Katsushima et al., 2006;Urreizti et al., 2003Urreizti et al., , 2006a. Urreizti et al. speculated that mutation T191M might interfered with the normal substrate-induced mobility of the region 186-222 making it impossible for the hCBS to retain PLP within the catalytic cavity (Urreizti et al., 2003(Urreizti et al., , 2006a. In light of our recent structural data (Ereño-Orbea et al., 2013a, it is reasonable to think that this mutation likely imposes a steric hindrance that severely distorts its environment (Fig. 9), thus impairing the entire three-dimensional fold of the moveable subdomain and consequently the conformational change associated with the aperture of the catalytic cavity. This would explain the structural instability and extensive unfolding caused by the mutation T191M , in both the basal and activated states of hCBS. We hypothesize that a similar scenario might occur in mutants V168M, I143M and E144K, which are also located in this región (Fig. 9).
The structural data presented herein represents another step towards understanding the molecular mechanism underlying the catalysis and regulation of the CBS enzymes. Together with previously elucidated molecular mechanism of allosteric regulation of CBS by AdoMet (Ereño-Orbea et al., 2013aMcCorvie et al., 2014), herein described substrate-induced closure of the catalytic site broadens our knowledge and both will be instrumental in the rational design of drugs modulating CBS activity.  Structure of the basal (grey, PDB ID 4LOD) and activated (pink, PDB ID 4PCU) conformation of the catalytic core of hCBS. In the basal state, the moveable submotif remains in a closed conformation (in grey) due to the presence of the Bateman module (not shown) above the catalytic cavity. Binding of AdoMet at the Bateman module triggers a migration of the latter from atop the catalytic cavity, thus allowing the displacement of the moveable motif towards an open conformation (pink). The shift is indicated with a blue arrow. Although not represented, the artificial removal of the Bateman module (Meier et al., 2001) exerts a similar effect in hCBS, and facilitates the aperture of the moveable motif. (B, C) In the absence of substrates in the catalytic cavity, the moveable subdomain of (B) dCBS; (PDB ID 3PC2) and (C) EhOASS (PDB ID 2PQM), adopts an open conformation (in yellow and blue slate, respectively) that evolves towards a closed state (PDB IDs 3PC4 in blue marine and 3BM5 in cyan, in B and C, respectively), when the substrates enter the PLP cavity.

Expression and purification of AmCBS
The pET-28a-C-AmCBS expression construct was prepared as described previously (Oyenarte et al., 2012). Full-length AmCBS was expressed and purified following the protocols that were developed for other CBS enzymes (Oyenarte et al., 2012).
The AmCBS structure was determined by molecular replacement with the program PHENIX (Adams et al., 2010), using the crystal  Structural superimposition of the catalytic core of AmCBS (red) with the catalytic core of (A) apo-dCBS (yellow, PDB ID 3PC2) and (B) dCBS with bound aminoacrylate (blue PDB ID 3PC3) or with serine (3PC4, not represented). The loops (and residues) involved in configuring the entrance to the catalytic cavity are indicated with arrows. The moveable submotif is enhanced in solid ribbons, whereas the static domain is in transparent cartoon. structure of the dCBS (PDB 3PC2) as the initial search model. Crystallographic refinement was carried out with PHENIX (Adams et al., 2010) and REFMAC5 (Winn et al., 2003;Murshudov et al., 2011). Ramachandran statistics for the refined coordinates (residues in favored region (%), number of outliers) were (97.27%, 0.11) for AmCBS. . The final refinement statistics are summarized in Table 1.
The structural analysis of all enzymes discussed in this manuscript was done using The PyMOL Molecular Graphics System (http://www. pymol.org) and Coot (Emsley et al., 2010). Calculation of surfaces was done with the PISA server (Krissinel and Henrick, 2007). The figures showing three-dimensional protein structures were prepared with PyMOL and CHIMERA (http://www.rbvi.ucsf.edu/chimera) (Pettersen et al., 2004). Sequence alignments were done with Clustal W (Larkin et al., 2007) and represented with CINEMA (Parry-Smith et al., 1998).

CBS specific activity measurements
The CBS activity in the classical reaction was determined by a radioisotope assay using ( 14 C(U)) L-serine as the labeled substrate, essentially as described previously (Majtan et al., 2010).

Accession numbers
The atomic coordinates of AmCBS, and structure factors reported in this paper have been deposited in the Protein Data Bank database, under PDB ID code 5OHX.