Structural insight into the molecular mechanism ofallosteric activation of human cystathionineβ-synthase by S-adenosylmethionineJune Ereño-Orbeaa,1, Tomas Majtanb,1, Iker Oyenartea, Jan P. Krausb,2, and Luis Alfonso Martínez-Cruza,2

aStructural Biology Unit, Center for Cooperative Research in Biosciences (CIC bioGUNE), Technology Park of Bizkaia, 48160 Derio, Spain; and bDepartment ofPediatrics, University of Colorado School of Medicine, Aurora, CO 80045

Edited by Solomon H. Snyder, The Johns Hopkins University School of Medicine, Baltimore, MD, and approved August 12, 2014 (received for reviewJuly 30, 2014)

Cystathionine β-synthase (CBS) is a heme-dependent and pyridoxal-5′-phosphate–dependent protein that controls the flux of sulfur frommethionine to cysteine, a precursor of glutathione, taurine, and H2S.Deficiency of CBS activity causes homocystinuria, the most frequentdisorder of sulfur amino acid metabolism. In contrast to CBSs fromlower organisms, human CBS (hCBS) is allosterically activated byS-adenosylmethionine (AdoMet), which binds to the regulatory do-main and triggers a conformational change that allows the protein toprogress from the basal toward the activated state. The structuralbasis of the underlying molecular mechanism has remained elusiveso far. Here, we present the structure of hCBS with bound AdoMet,revealing the activated conformation of the human enzyme. Bindingof AdoMet triggers a conformational change in the Bateman moduleof the regulatory domain that favors its association with a Batemanmodule of the complementary subunit to form an antiparallel CBSmodule. Such an arrangement is very similar to that found in theconstitutively activated insect CBS. In the presence of AdoMet, theautoinhibition exerted by the regulatory region is eliminated, allow-ing for improved access of substrates to the catalytic pocket. Basedon the availability of both the basal and the activated structures, wediscuss the mechanism of hCBS activation by AdoMet and the prop-erties of the AdoMet binding site, as well as the responsiveness ofthe enzyme to its allosteric regulator. The structure describedherein paves the way for the rational design of compounds mod-ulating hCBS activity and thus transsulfuration, redox status, andH2S biogenesis.

CBS domain | hydrogen sulfide

Cystathionine β-synthase (CBS; EC 4.2.1.22) is a pyridoxal-5′-phosphate (PLP)-dependent enzyme that catalyzes the β-

replacement of the hydroxyl group of L-serine (Ser) by L-homo-cysteine (Hcy), yielding cystathionine (Cth) (1). A deficient activityof human CBS (hCBS) is the cause of classical homocystinuria[CBS-deficient homocystinuria (CBSDH); Online Mendelian In-heritance in Man (OMIM) no. 236200], an autosomal, recessiveinborn error of sulfur amino acid metabolism, characterized by in-creased levels of Hcy in plasma and urine. CBSDH manifests asa combination of connective tissue defects, skeletal deformities,vascular thrombosis, and mental retardation (2).The hCBS is a homotetrameric enzyme whose subunits are or-

ganized into three structural domains. The N-terminal regionbinds heme and is thought to function in redox sensing and/orenzyme folding (3, 4). The central catalytic core shows the fold ofthe type II family PLP-dependent enzymes (5, 6). Finally, theC-terminal region consists of a tandem pair of CBS motifs (7–9)that bind S-adenosylmethionine (AdoMet) and lead to an increasein catalytic activity by up to fivefold (10, 11). The CBS motif pair,commonly known as a “Bateman module” (12, 13), is responsiblefor CBS subunit tetramerization (14, 15). The presence of path-ogenic missense mutations in this region often does not impairenzyme activity but typically interferes with binding of AdoMetand/or the enzyme’s activation by AdoMet (15–17). Removal of

the regulatory region leads to a dimer with much increased activity(14, 15). Recently, we showed that removal of residues 516–525,forming a flexible loop of the CBS2 motif of hCBS, yields dimericspecies (hCBSΔ516–525) with intact AdoMet binding capacityand activity responsiveness to AdoMet similar to a native hCBSWT (18).hCBS is regulated by a complex molecular mechanism that

remains poorly understood. More than a decade ago, we andothers hypothesized that hCBS might exist in two different con-formations: a “basal” state with low activity, where the C-terminalregulatory domain would restrict the access of substrates into thecatalytic site, and an AdoMet-bound “activated” state, where theAdoMet-induced conformational change would allow for enzymeactivation (16, 19). Recently, we have unveiled the relative ori-entations of the regulatory and catalytic domains in hCBS (18),which were in a striking contrast to those of both the previous insilico models (20, 21) and the Drosophila melanogaster (dCBS)structure (22). Our data showed that, although the pairing modeand the orientation of catalytic cores are similar in both insectdCBS and hCBS, the position of their regulatory domains ismarkedly different (18). In the basal state, the Bateman modulesfrom each hCBS unit are far apart and do not interact with eachother, being placed just above the entrance of the catalytic siteof the complementary subunit, thus hampering the access of

Significance

Cystathionine β-synthase (CBS), the pivotal enzyme of thetranssulfuration pathway, regulates flux through the pathwayto yield compounds, such as cysteine, glutathione, taurine, andH2S, that control cellular redox status and signaling. Our crystalstructure of an engineered human CBS construct bound toS-adenosylmethionine (AdoMet) reveals the unique binding siteof the allosteric activator and the architecture of the human CBSenzyme in its activated conformation. Together with the basalconformation that we reported earlier, these structures unravelthe molecular mechanism of human CBS activation by AdoMet.Current knowledge will allow for modeling of numerous path-ogenic mutations causing inherited homocystinuria and for de-sign of compounds modulating CBS activity.

Author contributions: T.M., J.P.K., and L.A.M.-C. designed research; J.E.-O., T.M., and I.O.performed research; J.P.K. and L.A.M.-C. contributed new reagents/analytic tools; J.E.-O.,T.M., I.O., and L.A.M.-C. analyzed data; and J.E.-O., T.M., J.P.K., and L.A.M.-C. wrotethe paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Data deposition: The atomic coordinates and structure factors have been deposited in theProtein Data Bank, www.pdb.org (PDB ID code 4PCU).1J.E.-O. and T.M. contributed equally to this work.2To whom correspondencemay be addressed. Email: Jan.Kraus@ucdenver.edu or amartinez@cicbiogune.es.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1414545111/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1414545111 PNAS | Published online September 2, 2014 | E3845–E3852

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 19

, 202

0

substrates into this cavity. Our hCBSΔ516–525 structure addition-ally revealed the presence of two major cavities in the Batemanmodule, S1 and S2, one of which (S2) is solvent-exposed andprobably represents the primary binding site for AdoMet (18).These findings are in agreement with the much higher basalactivity of dCBS and its inability to bind or to be regulated byAdoMet (23, 24) and suggest that the structural basis underlyingthe regulation of the human enzyme markedly differs from CBSregulation in insects or yeast (24). Taken together, the availabledata indicate that binding of AdoMet to the Bateman moduleweakens the interaction between the regulatory domain and thecatalytic core although the mechanism and the magnitude of theunderlying structural effect are still under debate (16, 19, 25–27).To solve the molecular mechanism of hCBS regulation by Ado-

Met, we have analyzed the crystals of an engineered hCBSΔ516–525protein that bears the mutation E201S, which potentially weak-ens and/or disrupts the interaction between the Bateman moduleand the catalytic core (Fig. 1A), thus favoring the activation ofthe enzyme. The data presented here fill a long-sought structuralgap by unraveling the crystal structure of AdoMet-bound hCBS,thus providing the overall fold of the enzyme in its activatedconformation and the identity of the AdoMet binding sites.Comparison with the structures of hCBS in basal conformationand constitutively activated dCBS was instrumental in the un-derstanding of the regulatory role played by the C-terminaldomain as well as the effect of some of the pathogenic muta-tions in the activation and/or inhibition of this key molecule oftranssulfuration.

ResultsStructure of Activated hCBS. A careful analysis of the interfacebetween the Bateman module and the catalytic core inhCBSΔ516–525 revealed that, in the absence of the allostericregulator, the C-terminal regulatory domain interacts with the

catalytic core via (i) hydrophobic interactions involving residuesI537, L540, and A544 of the CBS2 motif and I166, V189, V206,L210, and I214 of the protein core (18), and (ii) H-bonds betweenresidues T460, N463, S466, and Y484 from the CBS1 motif andE201, N194, R196, and D198 from loop L191–202 at the entranceof the catalytic cavity (Fig. 1A). Based on this knowledge, we engi-neered a protein construct hCBSΔ516–525 E201S (hCBS E201S)that, besides lacking residues 516–525 to ensure the formation ofdimeric species, carries the artificial mutation E201S. In the basalconformation, residue E201 is key in maintaining the Batemanmodule anchored to the entrance of the catalytic site througha network of H-bonds involving residues T460, N463, and Y484(Fig. 1A). Confirming our expectations, the resulting hCBS E201Sprotein is highly active (i.e., as active as AdoMet-stimulated hCBSWT) and retains its capacity to bind AdoMet although it is notresponsive to this allosteric regulator (Fig. 1B). We attemptedto crystallize the hCBS E201S protein in the presence and absenceof AdoMet. Despite great experimental effort, all crystallizationtests performed in the absence of AdoMet were unsuccessful,probably caused by the structural instability induced by the engi-neered mutation over the orientation of the Bateman module.By contrast, cocrystallization of the protein with AdoMet yieldedfew crystals suitable for crystallographic studies (Table S1). First,we compared the structure of hCBS E201S with both thehCBSΔ516–525 protein (18) and the truncated 45-kDa hCBSlacking the regulatory domain (6) and found that, like theseproteins, it forms dimers in which complementary catalytic coresinteract tightly (Fig. 2). In hCBS E201S, the fold of the coremaintains its secondary elements basically unaltered with respect tothe structures mentioned above, being composed of 13 α-helicesand two β-sheets consisting of four (β4–β7) and six (β2, β3,and β8–β11) strands, respectively. Importantly, due to a differ-ent orientation of the Bateman module, the loops L145–148,L171–174, and L191–202, which are sandwiched between thecore and the Bateman module at the entrance of the catalytic sitein the basal state of hCBS (18), now show an open conformationthat allows free access of substrates into the PLP site (Fig. S1).Strikingly, this loop arrangement reproduces a similar arrange-ment found in the truncated 45-kDa hCBS (6, 28) and in dCBS(22). Except for these small although functionally importantfeatures, the overall fold of the catalytic core shows no majorstructural differences.A major contribution of our hCBS E201S structure is the

unveiling of the relative orientations of the regulatory and catalyticdomains in AdoMet-bound hCBS. In contrast to our previouslyproposed activation mechanism (18), hCBS E201S shows anoverall arrangement that resembles dCBS (Fig. 2 B–D). There, theBateman module does not interact with the catalytic core exceptvia the connecting linker and associates in a tight dimer through itsinterfacial α-helices, forming a head-to-tail–oriented disk-shapedstructure referred to as an “antiparallel CBS module” (Fig. 2 B–Dand Fig. S2) (22). Flat disk-like dimers represent the most fre-quent association manner of Bateman modules (12, 13). Weshould mention, however, that, in our crystals, the CBS module ofhCBS E201S appears slightly tilted (∼40 degrees) with respect tothe surface of the protein cores compared with dCBS (Fig. 2C).This deviation is caused by the presence of a symmetry-relatedmolecule in the crystal (Fig. S3) and reveals two important facts:first, the structural compactness of the disk-like CBS module withbound AdoMet, and second, the intrinsic flexibility of the con-necting linker between the catalytic core and the Bateman modulein hCBS. A straightforward structural alignment of the CBSmodule of hCBS E201S with that of dCBS (Fig. S2) easily revealsthe most probable orientation of the regulatory domain in theactivated hCBS in solution (Fig. 2D), where no crystal packingeffects exist.

Fig. 1. Interactions between protein domains in basal hCBS. (A) InhCBSΔ516–525, residues Y484, N463, and S466 anchor the Bateman module(blue) to the protein core (gray) through H-bonds with the residues E201and D198 from the loop L191–202, thus occluding the entrance to the cat-alytic pocket. (B) The CBS-specific activity of selected hCBS variants in theabsence (blue bars) and the presence (red bars) of 300 μM AdoMet. hCBSenzyme species marked with “Δ” lack residues 516–525 and form dimers.

E3846 | www.pnas.org/cgi/doi/10.1073/pnas.1414545111 Ereño-Orbea et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 19

, 202

0

The AdoMet Binding Site. Similarly to unrelated CBS domainproteins (12, 13), the Bateman module of hCBS contains twomajor cavities (designated as sites S1 and S2) at the β-sheet–linedcleft between the CBS1 and CBS2 motifs that represent twopossible AdoMet binding sites (18). In the basal hCBS con-formation, cavity S1 is occluded by the presence of structuralelements from the catalytic core and is occupied by bulkyhydrophobic residues (Fig. 3A). Conversely, site S2 is exposedand is presumably the primary binding site for AdoMet (Fig. 3C).In agreement with these observations, the Fo − Fc exp(iφc) mapsshowed residual electron density consistent with one molecule ofAdoMet bound to site S2 of hCBS E201S (Fig. 3D and Fig. S4).No further electron density was detected at site S1 despiteit being solvent-exposed in the crystals (Fig. 3B). At site S2,the adenine base is stabilized in a predominantly hydrophobicbinding pocket via stacking interactions with aliphatic residues,namely L419, L423, V425, I437, F443, A446, P447, V533, andV534 (Fig. 3D). The adenine ring is further stabilized by a net-work of hydrogen bonds between the 6’-amino exocyclic groupand the backbone carbonyl oxygen of residue L423 and A446,and also between the 1’-aza groups and the main-chain carbonylatoms of L423. In addition, there are hydrogen-bond interactionsbetween the N7 of the adenine ring and the main-chain carbonyloxygen of Q445. The ribose moiety is bound in a polar pocket inwhich one of the free hydroxyl groups hydrogen bonds the side-chain carboxyl of D538, as well as the hydroxyl of S420. The alkylchain of AdoMet is stabilized by I537 with its amino and car-boxylate groups pointing toward the hydrophilic pocket provided

by the side chain of residue T535 and Q445. Residue D444H-bonds the amino group of AdoMet and additionally compen-sates the positive charge of the sulphonium ion of AdoMet.

Allosteric Activation of hCBS by AdoMet. The comparison ofhCBSΔ516–525 (18) with hCBS E201S+AdoMet shows that theoverall fold of the catalytic core remains virtually unchangedwhereas the location of the Bateman module differs markedly(Fig. 2 A versus C and D). The structural alignment of the Bate-man modules from these two structures reveals a distinct orien-tation of the CBS motifs that, in hCBS E201S, appeared rotatedwith respect to each other due to the presence of the allostericregulator. (Fig. 4 and Movie S1). This torsion takes place aroundthe flexible loops connecting helices α17-α18 and helix α20 withstrand β14 without affecting secondary structure elements. Sucha structural change is clearly more dramatic than the previouslyobserved modest compression of the regulatory CBS domains inthe partially activated D444N mutant (Fig. 1B) (18).In hCBS E201S, the rotation of the CBS motifs results in

a concomitant approximation of helices α21 and α22 toward thelarge central cavity S2 that accommodates AdoMet (indicatedwith arrows in Fig. 4). The direct consequence of such rotation isa disruption of the interactions that anchor the Bateman moduleto the catalytic core in the basal form, thus alleviating theocclusion imposed by the regulatory domain (Fig. S5). Underthis scenario, the Bateman module is free to move away from theentrance of the catalytic cavity of the complementary monomer,favored by the intrinsic flexibility of the linker connecting the

Fig. 2. The 3D structure of hCBS and dCBS. Crystal structure of (A) hCBSΔ516–525, (B) dCBS, and (C) AdoMet-bound hCBS E201S. Similarly to dCBS, hCBSE201S associates in dimers in which the Bateman module from two subunits interacts to form an antiparallel CBS module. In hCBS E201S, the CBS module hoststwo molecules of AdoMet. The slightly tilted orientation (∼40 degrees) of the regulatory domain with respect to the catalytic core is due to the presence ofa symmetry-related molecule in the crystal (Fig. S3). (D) The high structural similarity existing between the CBS modules of dCBS and hCBS E201S (Fig. S2),allowed us to easily reorient the tilted CBS module to its most probable position in solution using the dCBS structure as template. The dash-dotted linesrepresent the planes containing the flat disk-like CBS modules in each protein.

Ereño-Orbea et al. PNAS | Published online September 2, 2014 | E3847

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 19

, 202

0

domains. Subsequently, the compatibility existing between the aminoacid residues of helices α18, α19, α21, and α22 facilitates the for-mation of a flat disk-like antiparallel CBS module (Fig. 2 and Fig.S2), thus turning the enzyme into its activated status (Fig. 2 C andD).To accomplish the full sequence of hCBS activation, the pro-

tein necessarily needs to meet certain requirements: (i) AdoMetshould remain bound to the S2 cavity to maintain the Batemanmodule in the proper conformation favoring the assembly of theantiparallel CBS module; (ii) the connecting linker has to besufficiently long and flexible to facilitate the displacement of theregulatory domain; (iii) the strength of the interactions betweenthe regulatory domain and the protein core should enable therelease of the Bateman module from its initial position above thecatalytic site upon binding of AdoMet at cavity S2; and finally (iv)the interactions between the interfacial α-helices in the CBSmodule should allow the return of each regulatory domain backto its initial position once AdoMet abandons the S2 cavity.The comparison of basal and activated hCBS E201S as well

as of the partially activated D444N mutant clearly shows thatmaintenance of AdoMet bound to cavity S2 is critical for theformation of the CBS module. In the absence of AdoMet, theCBS motifs would rotate back toward their basal conformation,

and that would promote disassembly of the CBS module (Fig. S6).Thus, release of AdoMet from S2 is most probably the drivingforce causing deactivation of the enzyme. In the same way, theabsence of AdoMet in the S2 cavity is an obligate requirement topreserve the interactions between the Bateman module and thecatalytic core in the basal state (Figs. S5 and S6). The super-imposition of the activated Bateman module onto the basalregulatory domain clearly shows that the main interactions main-taining both protein domains together would otherwise be im-paired (Fig. S5).Taken together, the structures currently available suggest that

mutations such as D444N induce only smooth structural changesin the Bateman module that justify higher basal hCBS activities(25) but that are not sufficient to mimic the effect of AdoMet,which involves formation of a CBS module. The linker (residues386–411) connecting the protein core with the regulatory domainis similar in length to the equivalent region in dCBS (Fig. 5) (24).In hCBS, the linker contains two α-helices (α15 and α16) and anunstructured loop that precedes the first α-helix (α17) of theBateman module. During hCBS activation, helix α15 remainsfixed in the same position, helped by hydrophobic and polarinteractions that hold it anchored to helix α12 of the protein core

Fig. 3. AdoMet binding site in basal and activated hCBS. (A) Site S1 in basal hCBSΔ516–525. The entrance to cavity S1 is sterically occluded by the presence ofstructural elements from the catalytic core of a complementary monomer in the dimer (cyan). Additionally, bulky hydrophobic residues occupy the cleft andimpede the binding of AdoMet at this site. (B) Site S1 in activated AdoMet-bound hCBS E201S. Despite the presence of AdoMet during the crystallization, siteS1 remained empty in our crystals. As shown, binding of AdoMet (in black lines) at site S1 would cause steric clashes within the cavity S1, even in the activatedconformation of hCBS. Note: The potential location of AdoMet at site S1 has been modeled by performing a structural alignment of site S2 and S1, both intheir activated conformations. (C) Site S2 in basal hCBSΔ516–525. The S2 cavity is fully solvent-exposed and is not blocked by bulky residues. In hCBS, AdoMetbinds at a previously proposed site S2 of the Bateman module and induces a relative rotation of the two CBS motifs that results in a slight reorientation of theresidues within the S2 cavity. In the absence of such structural change, accommodation of AdoMet within site S2 would be sterically impeded. (D) Site S2 inthe activated AdoMet-bound hCBS E201S. This cavity represents the unique AdoMet binding site. The S2 cavity shows a hydrophobic cage that hosts theadenine ring of AdoMet, conserved aspartate (D538), threonine (T535), and serine (S420) residues to stabilize the ribose ring, and a hydrophobic residue (I537)preceding D538 that accommodates the alkyl chain of AdoMet.

E3848 | www.pnas.org/cgi/doi/10.1073/pnas.1414545111 Ereño-Orbea et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 19

, 202

0

(Fig. 5C). However, the rest of the linker, including helix α16(residues 395–407), moves away from their original positionallowing the relocation of the Bateman module to its activatedposition (Movie S2). Interestingly, the amino acids forming helixα15 are more conserved across phyla than those of helix α16(Fig. 5E).In the light of the available structural data, the hCBS activa-

tion can be summarized as follows:

i) hCBS being in its basal conformation, one molecule of Ado-Met per subunit binds to the exposed site S2 of the Batemanmodule (Fig. 3 C and D and Fig. S4). Accommodation ofAdoMet within the S2 site triggers rotation of CBS1 andCBS2 motifs with respect to each other (Fig. 4, Fig. S5,and Movie S1). This torsion takes place around the flexibleloops connecting helices α17–α18 and helix α20 with strandβ14, without affecting the secondary structure elements(a complete assignment of all secondary elements can befound in ref. 18).

ii) As a consequence of the relative rotation of the CBS motifs,the interactions between the Bateman module and the cata-lytic core of a complementary subunit are weakened and/ordisrupted (Fig. S5). The Bateman module moves away fromthe core, helped by the intrinsic flexibility of the linker con-necting the core with the regulatory domain (Movie S2).Hydrophobic interactions between helix α15 and α12 main-tain the position of helix α15 basically unaltered during theactivation process (Fig. 5C). The physical-chemical featuresof the residues located at helices α18, α19, α21, and α22

further facilitate the formation of the antiparallel CBS mod-ule (Fig. S2). The entrance of the catalytic cavity is not oc-cluded in this scenario, and the loops L145–148, L171–174,and L191–202 can relax toward an open conformation, thusallowing for unrestricted access to the catalytic center (Fig. 2C and D and Fig. S1). Although the site S1 is now solventaccessible, AdoMet does not occupy this cavity (Fig. 3B).

iii) Release of AdoMet from site S2 would promote a conforma-tional change of the Bateman module back toward its basalconformation (Fig. 6). Loss of AdoMet would cause disas-sembly of the CBS module by impairing the interaction be-tween the interfacial α-helices of the two Bateman modules(Fig. S6), which would return back to their location abovethe catalytic cavity of the complementary monomer, formingthe basal conformation of the enzyme (Fig. 2A).

DiscussionDespite the intense research carried out over the last threedecades, the structural basis underlying the allosteric regulationof human cystathionine β-synthase has remained elusive. hCBSrepresents an extremely complex system involving three differentcofactors, at least two conformational stages and multiple olig-omeric species, of which the most abundant is a tetramer (1, 16,19, 29, 30). Currently, the overall folds of catalytic and regulatorydomains are known, and the interactions stabilizing the dimericenzyme, as well as the location of residues involved in catalysis,have been identified (6, 18, 28). More recently, the structure ofdCBS has provided new insights on the catalytic process by re-vealing that substrate binding induces conformational changesin the three loops delineating the entrance of the catalytic cavity,which determines the size of the entrance to this site (22). On theother hand, the structure of the dimeric species of hCBSΔ516–525(18), where the Bateman module of the first subunit occludes theentrance to the catalytic crevice of a complementary monomer(Fig. 2A), has explained how and why, in the absence of AdoMet,the regulatory domain blocks access to the catalytic site in hCBSwhereas dCBS remains constitutively activated (22). Moreimportantly, the structure of hCBSΔ516–525 allowed us todemonstrate why removal of the regulatory domain yieldshyperactivated dimers (14) and why deletion of the residues 516–525in hCBS inevitably disrupts the oligomeric equilibrium toward theformation of dimers, thus highlighting the fundamental role ofthe loop L513–529 protruding from the central β-strand of theCBS2 motif in stabilizing protein tetramers (18, 31). Finally, thehCBSΔ516–525 structure revealed one of the main cavities ofthe Bateman module (site S2) as the most probable site forAdoMet binding (18).We recently demonstrated that the structure of D444N, a

pathogenic mutant with increased basal activity and impairedresponse to AdoMet (25, 32), closely resembles that of basalhCBS, showing only small structural rearrangements limited toa slight displacement of helices α18, α19, and α22 within theregulatory domain (18). This arrangement, concordant with recentfindings indicating that activation of hCBS by AdoMet elapseswithout significant alteration of secondary structure elements(26, 27), seemed to contradict former data claiming the occur-rence of large conformational rearrangements to reach the acti-vated conformation (16, 19, 25). Nevertheless, the effect inducedby the D444N mutation allowed us to explain higher basal activityof this mutant, in which the loops at the entrance of the catalyticsite could find sufficient space to move freely (18).The herein presented structure of hCBS E201S with bound

AdoMet allowed for significant progress in understanding themechanism of the allosteric regulation of hCBS. Taking intoaccount (i) the orientation and association manner of theC-terminal domain in hCBS E201S protein dimer (Fig. 2 C andD), (ii) the open conformation of the loops that regulate accessof substrates into the catalytic PLP site (Fig. S1), (iii) the high

Fig. 4. Effect of AdoMet on the Bateman module of hCBS. Two differ-ent views of the structural superimposition of the Bateman modules ofhCBSΔ516–525 (gray; PDB ID code 4L3V) and hCBS E201S (blue; PDB ID code4PCU). The latter contains one molecule of AdoMet (sticks) bound at site S2.AdoMet induces a conformational change in the Bateman module thatconsists of a relative rotation of the two CBS motifs without altering theiroverall secondary and tertiary structure. The root-mean-square deviations(rmsd) between the CBS1-CBS1* and CBS2-CBS2* motifs of both proteins are1.211 and 1.017, respectively. The large structural difference betweenBateman modules in the basal and in the activated states is reflected in anrmsd of 2.647. Blue arrows indicate the direction of the rotation uponAdoMet binding. The location of sites S1 and S2 is also indicated.

Ereño-Orbea et al. PNAS | Published online September 2, 2014 | E3849

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 19

, 202

0

specific activity similar to AdoMet-stimulated hCBS WT and nofurther response to AdoMet (Fig. 1B), and (iv) the presence ofthe allosteric regulator in the C-terminal regulatory domain (Fig.3D and Fig. S4), we conclude that the solved structure reflectsthe activated conformation of the hCBS. Our data unequivocallyconfirm that, indeed, the S2 pocket represents the main bindingsite for AdoMet and call into question the ability of site S1 tobind this small molecule. Additionally, we show that the activa-tion of hCBS occurs primarily through movements of entireprotein domains, which behave like rigid bodies during theprocess, barely affecting the secondary structure elements (Fig. 4,Fig. S5, and Movies S1 and S2). Interestingly, AdoMet bindingpromotes a sharp rotation of the two CBS motifs constituting theregulatory domain, which, upon this structural change, migratesfrom its initial location above the catalytic cavity of the com-plementary monomer, toward the catalytic core of its own subunit,to form a disk-shaped antiparallel CBS module. This displacement

of the Bateman module grants freedom of movement to theloops that delineate the entrance into the catalytic site and leadsto enzyme activation. Our data provide, to our knowledge, thefirst experimental evidence of the overall architecture of acti-vated hCBS and reveal that it basically reproduces the structureof dCBS (22). Consequently, we can now state that the previouslysolved structure of a D444N hCBS mutant corresponds to apartially activated state that significantly differs from the AdoMet-induced hCBS activation (Fig. 6) (18). Thus, in contrast with ourinitial prediction, the new data support the idea of an evolution-arily conserved overall fold for the activated forms of CBSsacross phyla.In light of these findings, we were able to propose the se-

quence of steps that occurs during the activation/deactivation ofthe enzyme. Moreover, these findings may allow us to explainhow pathogenic missense mutations alter catalytic activity ofhCBS and its regulation by AdoMet and subsequently classify

Fig. 5. Interactions that stabilize the activated conformation of CBSs. Several hydrophobic interactions between helices α12 (core) and α15 (connective linker)stabilize the orientation of the helix α15 in both dCBS (A) and AdoMet-bound hCBS E201S (C). (B) Network of polar interactions involving the residues R310,D362, N363, E366, K371, R475, and E478 determine in dCBS the orientation of the α-helices equivalent to α12, α15, and α21 of hCBS. (D) On the other hand, inAdoMet-bound hCBS E201S, the presence of residues L392, L397, and H501 (in equivalent positions to E366, K371, and R475, respectively, in dCBS) weakensthe polar network between helices α15 and α21 and presumably facilitates hCBS activation and the displacement of the Bateman module. In the hCBS E201Scrystals, the region corresponding to residues 398–405 of hCBS (dark gray) is disordered. To facilitate the structural comparison, this region has been modeledusing the dCBS structure as template. The side chains of residues R389 and H501 are not represented because their orientation is not clear in the electrondensity maps. (E) Sequence alignment of dCBS and hCBS in these regions.

E3850 | www.pnas.org/cgi/doi/10.1073/pnas.1414545111 Ereño-Orbea et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 19

, 202

0

them according to the particular step(s) of the activation processthey affect. For example, one group would include mutationsthat weaken or disrupt the interaction between the Batemanmodule and the catalytic core without impairing the capability tobind AdoMet. These mutants, such as hCBS S466L as well as theartificial hCBS E201S, show very high basal activity similar toAdoMet-activated hCBS WT, but are not further stimulated byAdoMet (Fig. 1B), even though the AdoMet binding is not im-paired (16, 27). Unlike artificial E201S or pathogenic S466L,another pathogenic mutation, D198V, located in the loop L191–202, delineating the access to the catalytic site (Fig. 1A), does notaffect the basal activity or the enzyme’s response to AdoMet(Fig. 1B). Other mutations may impair the AdoMet bindingcapacity, the relative rotation of the CBS motifs, the flexibilityand/or the correct fold of the connecting linker, the free move-ment of the loops delineating the entrance of the catalytic cavity,or the formation and stability of the CBS module. Some muta-tions may act via multiple mechanisms and thus affect severalsteps of the activation process. As an example, D444N weakensthe interaction between protein domains in the basal confor-mation, thus partially activating the enzyme and, at the sametime, reducing the responsiveness to AdoMet (Fig. 1B) (18). Inthe activated conformation, however, the D444N mutation likelyimpairs binding of AdoMet at the site S2 by abolishing the

neutralizing effect exerted by aspartate 444 on the closely placedamino groups of the allosteric regulator (Fig. 3D). The absence ofbound AdoMet at any Bateman module within the dimer inevitablyentails disassembly of the CBS module (as shown in Fig. S6).Comparative analysis of the activated conformation of both

dCBS and hCBS indicates that the constitutive activation foundin insect CBS is mainly due to polar interactions (Fig. 5 A and B).The presence of hydrophobic residues in equivalent positions ofhCBS (Fig. 5D) impairs some of these contacts and presumablyallows the human enzyme to oscillate between (at least) twodifferent conformations (Figs. 2 and 6).In conclusion, in the absence of AdoMet, hCBSΔ516–525

forms symmetrical dimers in which the catalytic core of eachsubunit interacts with both the catalytic core and the regulatorydomain of the complementary subunit (Fig. 2A) and the C-ter-minal regulatory domain occludes the entrance to the catalyticsite of the complementary monomer (Fig. S1). In contrast, sim-ilar to dCBS (Fig. 2B), AdoMet-bound hCBSΔ516–525 E201Sassociates in dimers in which the CBS motifs from the twosubunits form an antiparallel CBS module (Fig. 2 C and D),leaving the catalytic site unobstructed (Fig. S1) and the enzymeactivity much increased (Fig. 1B). However, this new structuralinformation raises new questions, such as whether AdoMet playsa role in the oligomeric state of the enzyme or whether it altersthe overall structure of the tetrameric species. It would appearthat the transition from the basal to the activated state may besterically hindered in native tetramers and thus would involvedisassembly of the protein tetramer into dimers. Further studiesare needed to figure out how the length and flexibility of theconnecting linker allow such structural change.

Materials and MethodsPreparation and Biochemical Characterization of the Recombinant Protein. Thepreparation of a pGEX-6P1-hCBSOPTΔ516–525, pET28-C-hCBSOPTΔ516–525,and pET28-C-hCBSOPTΔ516–525 D444N expression constructs is described indetail elsewhere (18, 31, 33). The artificial E201S and pathogenic D198V andS466L mutations were introduced by using a QuikChangeII XL mutagenesiskit (Agilent) according to the manufacturer’s recommendations. The corre-sponding mutagenesis oligonucleotides were as follows (capital lettersdesignate the codon of the mutated residue): 5′-ccgcttcgacagtccgTCAtcc-catgtgggtgtt (FWD E201S), 5′-aacacccacatgggaTGAcggactgtcgaagcgg (REVE201S), 5′-gaatgcccgcttcGTCagtccggaatccc (FWD D198V), 5′-gggattccggact-GACgaagcgggcattc (REV D198V), 5′-cctgggtaacatgctgCTAtcctgctggcgggca(FWD S466L), and 5′-tgcccgccagcagggaTAGcagcatgttacccagg (REV S466L).Purification of recombinant proteins followed the protocols that we de-veloped for various hCBS constructs carrying either cleavable GST at theN terminus or a permanent 6xHis tag at the C terminus with a few mod-ifications (11, 31, 33). Protein concentration was determined by the Bradfordmethod (Thermo Pierce) using BSA as a standard according to the manu-facturer’s recommendations. The CBS activity in the classical reaction wasdetermined by a previously described radioisotope assay using [14C(U)] L-ser-ine as the labeled substrate, essentially as described elsewhere (11, 33, 34).

Crystallization and Data Collection. The crystals were grown by the hanging-dropvapor-diffusion method at 293 K in 24-well crystallization plates according to theprotocol described previously (31). Drops consisted of 200 nL of hCBS E201S so-lution mixed with 400 nL of precipitant solution [20% (vol/vol) polyethylene glycolmonomethyl ether 3350] and AdoMet (Sigma); the protein concentration was20 mg/mL. The nucleotide was added to the protein at a final concentration of0.34 mM, resulting in ∼1:1 protein-to-nucleotide ratio. All trials to scale up thesuccessful conditions in higher volumes did not yield crystals. Single crystals weretransferred to a cryoprotection solution [20% (vol/vol) polyethylene glycol mon-omethyl ether 3350 and 20% glycerol] and flash frozen in liquid nitrogen. Thediffraction properties of the crystals were examined on beamlines ID23.1, ID23.2,and ID29 of the European Synchrotron Radiation Facility (ESRF), Grenoble. Thedataset presented here was collected at ALBA Synchrotron, Barcelona, beamlineMX XALOC-BL13 (λ = 0.9793 Å) and was processed using HKL2000 or XDSsoftware (35, 36).

X-Ray Diffraction Data Collection, Phasing, and Refinement. The hCBSE201Sstructure was determined by molecular replacement with the PHENIX pro-

Fig. 6. A model summarizing the effect of AdoMet binding and/or patho-genic mutations on the structure of hCBS. The hCBSΔ516–525 associates indimers (the subunits are depicted in orange and blue, respectively). AdoMetbinding triggers a conformational change that makes the protein to prog-ress from a basal (A) toward an activated (B) state. In addition, the patho-genic mutation, such as D444N, may partially activate the enzyme by slightlymodifying the dimer structure that still retains the overall fold of the basalconformation (C). Under these circumstances, the loops delineating theentrance of the catalytic cavity (represented by red bars) find some morespace to move, thus contributing to increase in the basal activity of theenzyme. The progression of these mutants toward the activated state (B) inthe presence of AdoMet depends on the effect of a particular mutation onthe formation and/or stability of the CBS module. (D) Other mutations, suchas the S466L or the artificial E201S, which impair the interactions betweenthe Bateman module and the catalytic core (as illustrated in Fig. 1A), pro-mote a more significant displacement of the regulatory domain away fromthe catalytic cavity, without formation of the disk-like CBS module. If Ado-Met binding is not impaired, these mutants can easily progress toward theactivated conformation as found in hCBS E201S in the presence of AdoMet(B). Both states depicted in B and D allow for free movement of the entranceloops and subsequently unrestricted flow of substrates into the PLP-con-taining active site.

Ereño-Orbea et al. PNAS | Published online September 2, 2014 | E3851

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 19

, 202

0

gram (37) using the crystal structure of the truncated 45-kDa hCBS (PDB IDcode 1JBQ) or the structure of the catalytic core of protein hCBSΔ516–525(PDB ID codes 4L3V and 4L0D) as the initial search model. After several cyclesof refinement using PHENIX and REFMAC5 (37, 38), CBS domains were builtmanually using Coot (39). The Ramachandran statistics for the refinedcoordinates are 95% residues in favored region (0.1%, number of outliers).The final refinement statistics are summarized in Table S1. The atomiccoordinates and structure factors have been deposited in the PDB under PDBID code 4PCU.

ACKNOWLEDGMENTS. We thank the staff at beamlines ID23.1 and ID29 ofthe European Synchrotron Radiation Facility and the staff at XALOC beam-

line of ALBA synchrotron facility for valuable support and technical assis-tance during synchrotron data collection. We also thank Dr. Adriana Rojasfor maintenance of the in-house X-ray equipment. This work was supportedby Postdoctoral Fellowship 0920079G from the American Heart Association(to T.M.); by National Institutes of Health Grant HL065217, American HeartAssociation Grant-In-Aid 09GRNT2110159, and a grant from the JeromeLejeune Foundation (all to J.P.K.); and by grants from the Department ofEducation, Universities and Research of the Basque Government (PI2010-17),the Department of Industry of the Basque Government (ETORTEK Program IE05-147 and IE07-202), the Bizkaia County (Exp.7/13/08/2006/11 and 7/13/08/2005/14),and the Spanish Ministry of Economy and Innovation (BFU2010-17857 andBFU2013-47531-R and the Spanish Ion Channel Initiative Program SICI-CONSOLIDER CSD2008-00005) (all to L.A.M.-C.).

1. Miles EW, Kraus JP (2004) Cystathionine beta-synthase: Structure, function, regula-tion, and location of homocystinuria-causing mutations. J Biol Chem 279(29):29871–29874.

2. Mudd SH, Levy HL, Kraus JP (2001) Disorders of transsulfuration. The Metabolic andMolecular Bases of Inherited Disease, eds Scriver CR, et al. (McGraw-Hill, New York),8th Ed, pp 2007–2056.

3. Banerjee R, Evande R, Kabil O, Ojha S, Taoka S (2003) Reaction mechanism and reg-ulation of cystathionine beta-synthase. Biochim Biophys Acta 1647(1-2):30–35.

4. Majtan T, Singh LR, Wang L, Kruger WD, Kraus JP (2008) Active cystathionine beta-synthase can be expressed in heme-free systems in the presence of metal-substitutedporphyrins or a chemical chaperone. J Biol Chem 283(50):34588–34595.

5. Christen P, Mehta PK (2001) From cofactor to enzymes: The molecular evolution ofpyridoxal-5′-phosphate-dependent enzymes. Chem Rec 1(6):436–447.

6. Meier M, Janosik M, Kery V, Kraus JP, Burkhard P (2001) Structure of human cys-tathionine beta-synthase: A unique pyridoxal 5′-phosphate-dependent heme protein.EMBO J 20(15):3910–3916.

7. Bateman A (1997) The structure of a domain common to archaebacteria and thehomocystinuria disease protein. Trends Biochem Sci 22(1):12–13.

8. Kemp BE (2004) Bateman domains and adenosine derivatives form a binding contract.J Clin Invest 113(2):182–184.

9. Ignoul S, Eggermont J (2005) CBS domains: Structure, function, and pathology inhuman proteins. Am J Physiol Cell Physiol 289(6):C1369–C1378.

10. Shan X, Kruger WD (1998) Correction of disease-causing CBS mutations in yeast. NatGenet 19(1):91–93.

11. Majtan T, Kraus JP (2012) Folding and activity of mutant cystathionine β-synthasedepends on the position and nature of the purification tag: Characterization of theR266K CBS mutant. Protein Expr Purif 82(2):317–324.

12. Baykov AA, Tuominen HK, Lahti R (2011) The CBS domain: A protein module with anemerging prominent role in regulation. ACS Chem Biol 6(11):1156–1163.

13. Ereño-Orbea J, Oyenarte I, Martínez-Cruz LA (2013) CBS domains: Ligand binding sitesand conformational variability. Arch Biochem Biophys 540(1-2):70–81.

14. Kery V, Poneleit L, Kraus JP (1998) Trypsin cleavage of human cystathionine beta-synthase into an evolutionarily conserved active core: Structural and functionalconsequences. Arch Biochem Biophys 355(2):222–232.

15. Jhee KH, McPhie P, Miles EW (2000) Domain architecture of the heme-independentyeast cystathionine beta-synthase provides insights into mechanisms of catalysis andregulation. Biochemistry 39(34):10548–10556.

16. Janosík M, Kery V, Gaustadnes M, Maclean KN, Kraus JP (2001) Regulation of humancystathionine beta-synthase by S-adenosyl-L-methionine: Evidence for two catalyti-cally active conformations involving an autoinhibitory domain in the C-terminal re-gion. Biochemistry 40(35):10625–10633.

17. Maclean KN, et al. (2002) High homocysteine and thrombosis without connectivetissue disorders are associated with a novel class of cystathionine beta-synthase (CBS)mutations. Hum Mutat 19(6):641–655.

18. Ereño-Orbea J, Majtan T, Oyenarte I, Kraus JP, Martínez-Cruz LA (2013) Structuralbasis of regulation and oligomerization of human cystathionine β-synthase, thecentral enzyme of transsulfuration. Proc Natl Acad Sci USA 110(40):E3790–E3799.

19. Shan X, Dunbrack RL, Jr, Christopher SA, Kruger WD (2001) Mutations in the regu-latory domain of cystathionine beta synthase can functionally suppress patient-derived mutations in cis. Hum Mol Genet 10(6):635–643.

20. Sen S, Banerjee R (2007) A pathogenic linked mutation in the catalytic core of humancystathionine beta-synthase disrupts allosteric regulation and allows kinetic charac-terization of a full-length dimer. Biochemistry 46(13):4110–4116.

21. Hnízda A, et al. (2010) Cross-talk between the catalytic core and the regulatory do-main in cystathionine β-synthase: study by differential covalent labeling and com-putational modeling. Biochemistry 49(49):10526–10534.

22. Koutmos M, Kabil O, Smith JL, Banerjee R (2010) Structural basis for substrate acti-vation and regulation by cystathionine beta-synthase (CBS) domains in cystathioninebeta-synthase. Proc Natl Acad Sci USA 107(49):20958–20963.

23. Su Y, et al. (2013) Comparative study of enzyme activity and heme reactivity inDrosophila melanogaster and Homo sapiens cystathionine β-synthases. Biochemistry52(4):741–751.

24. Majtan T, et al. (2014) Domain organization, catalysis and regulation of eukaryoticcystathionine beta-synthases. PLoS ONE 9(8):e105290.

25. Evande R, Blom H, Boers GH, Banerjee R (2002) Alleviation of intrasteric inhibition bythe pathogenic activation domain mutation, D444N, in human cystathionine beta-synthase. Biochemistry 41(39):11832–11837.

26. Hnízda A, et al. (2012) Conformational properties of nine purified cystathionineβ-synthase mutants. Biochemistry 51(23):4755–4763.

27. Pey AL, Majtan T, Sanchez-Ruiz JM, Kraus JP (2013) Human cystathionine β-synthase(CBS) contains two classes of binding sites for S-adenosylmethionine (SAM): Complexregulation of CBS activity and stability by SAM. Biochem J 449(1):109–121.

28. Taoka S, et al. (2002) Human cystathionine beta-synthase is a heme sensor protein:Evidence that the redox sensor is heme and not the vicinal cysteines in the CXXC motifseen in the crystal structure of the truncated enzyme. Biochemistry 41(33):10454–10461.

29. Banerjee R, Zou CG (2005) Redox regulation and reaction mechanism of human cys-tathionine-beta-synthase: A PLP-dependent hemesensor protein. Arch Biochem Bio-phys 433(1):144–156.

30. Skovby F, Kraus JP, Rosenberg LE (1984) Biosynthesis and proteolytic activation ofcystathionine beta-synthase in rat liver. J Biol Chem 259(1):588–593.

31. Oyenarte I, et al. (2012) Purification, crystallization and preliminary crystallographicanalysis of human cystathionine β-synthase. Acta Crystallogr Sect F Struct Biol CrystCommun 68(Pt 11):1318–1322.

32. Kluijtmans LA, et al. (1996) Defective cystathionine beta-synthase regulation byS-adenosylmethionine in a partially pyridoxine responsive homocystinuria patient.J Clin Invest 98(2):285–289.

33. Majtan T, Liu L, Carpenter JF, Kraus JP (2010) Rescue of cystathionine beta-synthase(CBS) mutants with chemical chaperones: Purification and characterization of eightCBS mutant enzymes. J Biol Chem 285(21):15866–15873.

34. Kraus JP (1987) Cystathionine beta-synthase (human).Methods Enzymol 143:388–394.35. Otwinowski Z, Minor W (1997) Processing of X-ray diffraction data collected in

oscillation mode. Methods Enzymol 276:307–326.36. Kabsch W (2010) Xds. Acta Crystallogr D Biol Crystallogr 66(Pt 2):125–132.37. Adams PD, et al. (2011) The Phenix software for automated determination of mac-

romolecular structures. Methods 55(1):94–106.38. Murshudov GN, et al. (2011) REFMAC5 for the refinement of macromolecular crystal

structures. Acta Crystallogr D Biol Crystallogr 67(Pt 4):355–367.39. Emsley P, Lohkamp B, Scott WG, Cowtan K (2010) Features and development of Coot.

Acta Crystallogr D Biol Crystallogr 66(Pt 4):486–501.

E3852 | www.pnas.org/cgi/doi/10.1073/pnas.1414545111 Ereño-Orbea et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 19

, 202

0