doi:10.1016/j.jmb.2010.03.034 J. Mol. Biol.

(2010) 398, 497–506

Available online at www.sciencedirect.com

Crystal Structures of Human HMG-CoA Synthase

Isoforms Provide Insights into Inherited Ketogenesis
Disorders and Inhibitor Design
Naeem Shafqat 1 , Andrew Turnbull 1 , Johannes Zschocke 2 ,
Udo Oppermann 1,3 ⁎ and Wyatt W. Yue 1 ⁎
1
Structural Genomics 3-Hydroxy-3-methylglutaryl coenzyme A (CoA) synthase (HMGCS)
Consortium, University of catalyzes the condensation of acetyl-CoA and acetoacetyl-CoA into 3-
Oxford, Oxford OX3 7DU, UK hydroxy-3-methylglutaryl CoA. It is ubiquitous across the phylogenetic tree
2 and is broadly classified into three classes. The prokaryotic isoform is
Divisions of Human Genetics essential in Gram-positive bacteria for isoprenoid synthesis via the
and Clinical Genetics, Medical mevalonate pathway. The eukaryotic cytosolic isoform also participates in
University Innsbruck, the mevalonate pathway but its end product is cholesterol. Mammals also
A-6020 Innsbruck, Austria contain a mitochondrial isoform; its deficiency results in an inherited
3
Botnar Research Centre, disorder of ketone body formation. Here, we report high-resolution crystal
Oxford Biomedical Research structures of the human cytosolic (hHMGCS1) and mitochondrial
Unit, Oxford, OX3 7LD, UK (hHMGCS2) isoforms in binary product complexes. Our data represent
the first structures solved for human HMGCS and the mitochondrial
Received 25 January 2010; isoform, allowing for the first time structural comparison among the three
received in revised form isoforms. This serves as a starting point for the development of isoform-
16 March 2010; specific inhibitors that have potential cholesterol-reducing and antibiotic
accepted 18 March 2010 applications. In addition, missense mutations that cause mitochondrial
Available online HMGCS deficiency have been mapped onto the hHMGCS2 structure to
25 March 2010 rationalize the structural basis for the disease pathology.
© 2010 Elsevier Ltd. All rights reserved.
Keywords: mevalonate; ketone bodies; HMG-CoA; condensation; mitochon-
Edited by G. Schulz drial HMGCS deficiency

Introduction to form HMG-CoA and CoA. 2,3 The enzyme

The compartmentalization mechanism of an
enzymatic activity into different cellular locations
is common in eukaryotes.1 In many cases, isozymes
catalyzing the same chemical reaction are encoded
by multiple homologous genes localized in differ-
ent subcellular locations and involved in diverse
physiological functions. 3-Hydroxy-3-methylglu-
taryl coenzyme A (HMG-CoA) synthase (HMGCS;
EC 2.3.3.10) catalyzes the condensation of acetoa-
cetyl-CoA (AcAc-CoA) and acetyl-CoA (Ac-CoA)
*Corresponding authors. E-mail addresses:
udo.oppermann@sgc.ox.ac.uk; wyatt.yue@sgc.ox.ac.uk.
Abbreviations used: HMG-CoA, 3-hydroxy-3-
methylglutaryl coenzyme A; Ac-CoA, acetyl-CoA; AcAc-
CoA, acetoacetyl-CoA; HMGCS, HMG-CoA synthase;
saHMGCS, S. aureus HMGCS; efHMGCS, E. faecalis
HMGCS; bjHMGCS, plant HMGCS from B. juncea.
belongs to the family of acyl-condensing enzymes
that include thiolases, β-keto-acyl carrier protein
synthases, and polyketide synthases. While eubac-
teria and lower eukaryotes encode a single cyto-
solic HMGCS, two isoforms are found in higher
eukaryotes: the cytosolic HMGCS1 and mitochon-
drial HMGCS2.
The two mammalian isoforms participate in
distinct metabolic pathways. Cytosolic HMGCS1 is
under the control of the sterol regulatory binding
elements and catalyzes the first committed step in
the mevalonate pathway.4,5 The HMG-CoA pro-
duced is converted by HMG-CoA reductase into
mevalonate, a precursor for cholesterol synthesis.6
Therefore, HMGCS1 is a potential drug target for
regulating serum cholesterol level and for develop-
ing novel antibiotics against certain pathogenic
Gram-positive bacteria (e.g., Enterococcus, Staphylo-
coccus, and Streptococcus spp.), which depend on the
mevalonate pathway for survival.7 Mitochondrial
HMGCS2 is under the control of the peroxisome

0022-2836/$ - see front matter © 2010 Elsevier Ltd. All rights reserved.
498 Crystal Structures of Human HMGCS1 and HMGCS2

Fig. 1. Crystal structure of human HMGCS mitochondrial isoform (hHMGCS2). (a) Proposed reaction mechanism of
HMGCS. The Cys–His–Glu catalytic triad is labeled. (b) Topology of hHMGCS2 forming two structural regions. The
upper region consists of the thiolase fold (cyan) and the eukaryotic-specific helical extension (magenta). The lower region
contains the HMGCS-specific fold (yellow). (c) Dimeric arrangement of HMGCS. The entrance to the active-site tunnel is
indicated by an arrow.

proliferator regulatory elements8 and is involved in
hepatic ketogenesis.9 The HMG-CoA produced is
converted into ketone bodies, which are important
fuels during periods of fasting and starvation.
HMGCS2 deficiency (OMIM 600234) is an inherited
metabolic disorder that causes hypoketotic hypogly-
caemia after prolonged fasting.10 To date, no muta-
tions have been identified in the human cytosolic
HMGCS1.
HMGCS catalyzes a three-step ping-pong reaction
involving a conserved Cys-His-Glu catalytic triad
and an acyl-enzyme intermediate (Fig. 1a).2,11–13 In
the first step (acylation), His264/301† is thought to
act as a catalytic base or H-bond donor12,14–16 for the
nucleophilic Cys129/166, which attacks the carbonyl
carbon of Ac-CoA (first substrate), thereby transfer-
ring the acetyl group to the Cys Sγ atom and
releasing the reduced CoA. In the second step
(condensation), the methyl group of acetyl-Cys is
deprotonated by the general base Glu95/132 to form
a carbanion, which attacks the distal (γ-) carbonyl of
the incoming AcAc-CoA (second substrate).13 In the
† Residue numbering hereinafter denotes human
HMGCS1/HMGCS2.
final step (hydrolysis), the resultant enzyme–HMG-
CoA intermediate is hydrolyzed to release the
product HMG-CoA and regenerate the reduced
cysteine.
HMGCS from all phyla can be broadly classified
into the bacterial isoform, the eukaryotic cytosolic
isoform, and the mammal-specific mitochondrial
isoform. The reported HMGCS structures from
bacterial pathogens Staphylococcus aureus (saHMGCS)
and Enterococcus faecalis (efHMGCS)15,17–19 and of the
cytosolic isoform from the plant Brassica juncea
(bjHMGCS) 14 have shown that despite limited
sequence identity (Supplementary Fig. S1), they
share the conserved thiolase fold20 and catalytic
machinery. Nevertheless, structural information is
lacking for a mammalian orthologue. The catalytic
domains of mammalian HMGCS1 and HMGCS2
share ∼60% sequence identity, although HMGCS1
has an ∼50-residue C-terminal extension of unknown
function, whereas HMGCS2 has an ∼ 40-residue
N-terminal extension encoding the mitochondrial
targeting sequence. We report the crystal structures
of the catalytic domains of human HMGCS1 and
HMGCS2 in distinct product complexes: hHMGCS1
bound with CoA (product of the first, acylation step),
and hHMGCS2 bound with HMG-CoA (product of
the second, condensation step).
Crystal Structures of Human HMGCS1 and HMGCS2 499

Results and Discussion previous observations that HMGCS enzymes from

diverse species are dimeric in solution.21–23

Overall structure HMGCS–product complexes elucidate catalytic

mechanism
We have determined the crystal structures of
hHMGCS1 and hHMGCS2 catalytic domains to a The catalytic domains of hHMGCS1/2 were
resolution of 2.00 and 1.70 Å, respectively (Table 1). co-crystallized with substrates to trap reaction inter-
The two structures share a near-identical topology mediates or products. The structure of hHMGCS1
(rmsd of 0.447 Å for all Cα atoms) and are described co-crystallized with Ac-CoA revealed a bound CoA
together as hHMGCS1/2 unless stated separately (by-product of the acylation step) and an acetylated-
otherwise. hHMGCS1/2 adopts a common architec- Cys129/166 adduct at the active site. On the contrary,
ture composed of ‘upper’ and ‘lower’ regions co-crystallizing hHMGCS2 with both Ac-CoA and
(Fig. 1b). The larger upper region features the thiolase AcAc-CoA resulted in trapping HMG-CoA (final
fold conserved among acyl-condensing enzymes, product of catalysis). The identification of bound
built upon a compact five-layered αβαβα core with a products at the active sites, as revealed by well-
2-fold internal symmetry.20 The smaller lower region defined electron density (Fig. 2a and b and Supple-
is unique among HMGCS enzymes and consists of mentary Fig. S2), suggests that our hHMGCS1/2
several α-helices and β-strands. The interface be- constructs are active in situ. However, the dissocia-
tween the upper and lower regions defines a tunnel tion of reaction products from the enzymes was
that forms the active site. The structures of prevented presumably by crystal-packing contacts
hHMGCS1/2 reveal a common homodimeric assem- involving the surface entrance of the active site.
bly where the active-site tunnels of the two mono- The active site of hHMGCS1/2 (Fig. 2a and b)
mers are oriented away from each other (Fig. 1c). This consists of an extended 15 Å-deep tunnel that
dimeric arrangement, similar to the reported bacte- accommodates the substrates Ac-CoA and AcAc-
rial and plant HMGCS structures, buries ∼15% of CoA. The nucleotide moiety of the substrates is
total molecular surface. The dimer interface consists exposed at the tunnel entrance, while the pan-
of conserved interactions and is consistent with tetheine moiety fits into a hydrophobic sleeve at the
tunnel interior and directs the CoA thiol toward a
catalytic pocket containing the invariant triad
Table 1. Summary of diffraction and refinement statistics Cys129/166, His264/301, and Glu95/132. The
architecture of the active site is highly similar to
hHMGCS1 hHMGCS2 the reported bacterial and plant structures (compare
Data collection a and b with c and d, Fig. 2). In hHMGCS1/2, the
Wavelength (Å) 0.9999 0.9998 products CoA and HMG-CoA are anchored to the
Space group P43212 P1 active site in a similar configuration, via electrostatic
a, b, c (Å) 92.76, 92.76, 65.41, 83.51, 101.44
235.53 and polar interactions conserved across different
α, β, γ (°) 90.00, 90.00, 90.00 100.00, 108.08, 96.30 phyla. These include Asp43/Glu80 interacting with
Resolution (Å)a 2.00–50.00 1.70–35.00 the adenine ring, as well as Lys46/83, Lys269/306,
(2.00–2.05) (1.70–1.84) and Lys273/310 with the 3′,5′-ADP phosphates
Rmerge 0.128 (0.722) 0.08 (0.38)
I/σI 10.7 (2.6) 12.7 (2.1)
(Supplementary Fig. S2). In addition, hHMGCS1/2
Completeness (%) 94.4 (99.7) 97.2 (95.9) contributes a nonconserved residue, Asn167/204
Redundancy 5.0 (6.1) 3.8 (3.9) (equivalent to Pro in bjHMGCS and Gly in
Refinement saHMGCS/efHMGCS), whose side-chain hydrogen
No. of reflections 66,668 212,827 bonds with the diphosphates. The pantetheine
Rwork/Rfree 0.159/0.202 0.163/0.188
No. of atoms moiety is tethered by van der Waals and nonpolar
Protein 7150 14,379 interactions from hydrophobic residues lining the
Ligand/ion 96 232 tunnel sleeve.
Water 751 1463 The complex of CoA with hHMGCS1 acetylated at
B-factors (Å2)
Main chain 29.7 16.8
Cys129/166 (Ac-Cys129/166) represents the state of
Side chain and water 33.3 23.5 the enzyme after the acylation step, prior to release
Ligands 36.7b 32.0c of the CoA by-product. Ac-Cys129/166 is confor-
rmsd. mationally locked by a hydrogen bond between its
Bond lengths (Å) 0.013 0.010 carbonyl oxygen and the backbone amide of
Bond angles (°) 1.345 1.440
Ramachandran plot Ser377/414 (Fig. 2a). This juxtaposes Ac-Cys129/
Favoured (%) 97.9 98.4 166 for an optimal attack on its C2 methyl carbon by
Allowed (%) 2.1 1.6 the carboxyl oxygen of the general base Glu95/132
a
Numbers in parentheses represent data in the highest-
(the C–O distance is 2.6 Å), resulting in proton
resolution shell. abstraction from Ac-Cys129/166 and generation of a
b
The averaged B-factors for the 3′,5′-ADP and pantetheine carbanion intermediate. At this stage, the enzyme is
moieties are 39 and 32 Å2. primed for the condensation reaction involving the
c
The averaged B-factors for the 3′,5′-ADP and pantetheine
+ HMG moieties are 46 and 20 Å2.
incoming second substrate AcAc-CoA and the
carbanion intermediate.
500 Crystal Structures of Human HMGCS1 and HMGCS2

Fig. 2. hHMGCS1/2 active site. Stereo view of the catalytic pockets in (a) hHMGCS1 and (b) hHMGCS2, bound with
CoA and HMG-CoA, respectively. The unbiased Fo − Fc difference maps (contoured at 3.0σ) shown for the ligands were
computed with the PHENIX ‘kick map’ subroutine,24 using model phases with the respective ligand (CoA and HMG-
CoA) omitted. In hHMGCS1, the catalytic nucleophile Cys129/166 is acetylated (Ac-C129/166). For comparison, the
active sites of saHMGCS bound with AcAc-CoA (c) (PDB code 1XPK) and bjHMGCS bound with HMG-CoA (d) (PDB
code 2FA0) are shown. The protein chains are coloured in the same scheme as Fig. 1.

On the other hand, the complex of HMG-CoA
with hHMGCS2 represents the enzyme after the
condensation and hydrolysis steps (Fig. 2b). The
HMG moiety of HMG-CoA occupies the catalytic
pocket such that its β-hydroxyl group hydrogen-
bonds to His264/301 and Asn343/380, while the
terminal carboxyl interacts with Glu95/132 and the
backbone amide of Ser377/414. This is in agreement
with the previously reported HMG-CoA-bound
structures (Fig. 2d), which proposed a role for
these residues in positioning the HMG moiety of
the enzyme–HMG-CoA adduct in a proper orienta-
tion for the hydrolytic release of HMG-CoA.18
Although no water molecules were observed in the
vicinity of the HMG moiety, the HMG carboxylate
oxygen, at a short distance (2.3 Å) from the Glu95/
Crystal Structures of Human HMGCS1 and HMGCS2 501

Fig. 3. Structural comparison of hHMGCS1 with bacterial and plant homologues. (a) Cα superposition of hHMGCS1
(blue) with plant bjHMGCS (pink; PDB code 2F82) and bacterial saHMGCS structures (yellow; PDB code 1TVZ),
highlighting two regions of structural difference. (b) Surface representation of active-site entrance of hHMGCS1, coloured
according to sequence conservation (magenta, most conserved; cyan, most variable). The bound CoA and residues that
exhibit sequence variation across the phyla are shown as sticks. The potential pocket for inhibitor binding is indicated by a
dotted circle. (c) Sequence alignment of the six nonconserved residues illustrated in (b). The HMGCS sequences are
extracted from human (hHMGCS1 and hHMGCS2), plant, [B. juncea (bjHMGCS)], and bacteria [S. aureus (saHMGCS) and
E. faecalis (efHMGCS)]. (d) Surface representation of active-site entrance of bjHMGCS from the enzyme–inhibitor complex
(PDB code 2F9A), where the helical subdomain is disordered. Superimposed on the enzyme surface are the bound CoA
molecule (red) from the hHMGCS1 structure and the helical subdomain (yellow) from the apo-bjHMGCS structure (PDB
code 2F82).

132 carboxylate oxygen, may represent the site of a
water molecule ideally positioned to undergo
deprotonation by Glu95/132 and nucleophilic at-
tack on the enzyme–HMG-CoA adduct.
Exploiting structural differences across the phyla
Our data provide a first structural glimpse of the
mammalian isoforms, which resemble the reported
saHMGCS/efHMGCS (rmsd 2.2 Å) and bjHMGCS
(rmsd 1.2 Å) structures, but display significant
differences in two regions (Fig. 3a). The most
distinguishing feature from bacterial HMGCS is a
∼ 40-residue subdomain in the upper region of
hHMGCS1/2, inserted between helices α8 and α9
of bacterial orthologues (region 1 in Fig. 3a).15 This
consists of two α- and two 310-helices and forms a
periphery ‘arm’ close to the entrance of the active-site
tunnel. This helical subdomain is present in the
previously determined bjHMGCS structure14 and is
conserved only among eukaryotes (Supplementary
Fig. S1). However, sequence conservation within the
subdomain is low between bjHMGCS and
hHMGCS1/2 (27% identity; Supplementary Fig. S1)
and there are significant backbone differences (rmsd
1.85 Å). A second region of structural difference
(region 2 in Fig. 3a) is the surface loop connecting
helix α7 and strand β12 of the upper region, which
502
exhibits diversity in both sequence and length across
the fungi, metazoan, and plant kingdoms. The
hHMGCS1/2 loop is more extended due to an
insertion of 5 and 11 residues relative to the plant
and bacterial sequences, respectively.
Our data, together with previous structures,14,15,17–19
complete the structural coverage for each of the
three HMGCS classes from human to plant to
bacteria. This enables cross-species structural com-
parison to identify nonconserved regions that aid
the development of isoform-specific inhibitors, for
example, targeting the bacterial isoform for use as
antibiotics, or the eukaryotic cytosolic isoform for
use as cholesterol-lowering drugs. While there is
strict sequence conservation at the bottom of the
active-site tunnel, our analysis highlights a number
of nonconserved residues at the active-site entrance
of hHMGCS1 (Fig. 3b), which can potentially
modulate different chemical properties on the
pharmacophores.
The first group of residues (Asn167/204, Ile222/
259, and Lys219/256) occupies gatekeeper positions
at the tunnel entrance. Asn167/204 is hydrogen-
bonded to the CoA diphosphates (Fig. 2), an
interaction not present in bjHMGCS (Asn is replaced
by Pro) or saHMGCS/efHMGCS (Asn is replaced by
Gly) (Fig. 3c). Ile222/259 interacts with the substrate
pantotheine moiety via van der Waals and hydro-
phobic interactions and is replaced by polar residues
in bjHMGCS (Gln) and saHMGCS/efHMGCS (Lys
and Asn). Lys219/256 is conserved in bjHMGCS,
but is substituted with the less bulky Ala in
saHMGCS and Pro in efHMGCS.
The second group of residues (Phe311/Thr348,
Leu270/Met307 and Tyr267/Phe304) forms a hy-
drophobic pocket (circled in Fig. 3b) near the tunnel
entrance that is not occupied by reaction substrates.
The upper ridge of this pocket is formed by the side
chain of Phe311/Thr348 from the eukaryotic-specif-
ic helical subdomain. This residue is replaced by the
hydrophilic Thr in hHMGCS2 and Gln in bjHMGCS,
and is absent in the bacterial structures. The floor of
this pocket is formed by Leu270/Met307 and
Tyr267/Phe304, which are moderately conserved
across species both in sequence and polarity
(Fig. 3c). This pocket can therefore be exploited to
bind substituent groups from inhibitor scaffolds to
improve selectivity and potency. Structure-based
drug design efforts have been reported for
bjHMGCS, where binding of the polyketide inhib-
itor F-244 resulted in backbone displacement of helix
α7 close to the equivalent hydrophobic pocket.14 A
similar conformational rearrangement could occur
in the human enzymes on binding the inhibitor.
Interestingly, F-244 was bound with its acyl tail
protruding into the hydrophobic pocket of the
bjHMGCS–inhibitor structure [Protein Data Bank
PDB code 2F9A] (Fig. 3d). However, interpretation
of its binding mode was hindered because part of
the acyl tail was disordered.14 Furthermore, much of
the helical subdomain (contributing to the upper
ridge of the hydrophobic pocket in hHMGCS1/2;
yellow helices in Fig. 3d) was also not modelled in
Crystal Structures of Human HMGCS1 and HMGCS2
the structure. Therefore, further structure determi-
nation of enzyme–inhibitor complexes for both
human and plant HMGCS will help the next
generation of inhibitor development.
Structural basis of HMGCS2 deficiency
To establish a structural basis for mitochondrial
HMGCS2 deficiency, a metabolic disorder in ketone
bodies formation, we have mapped the causative
mutations on the hHMGCS2 structure. Two non-
sense and 15 missense mutations have been identi-
fied in the hmgcs2 gene of HMGCS2-deficient
patients (Table 2).25,27–29 The nonsense mutation is
expected to result in nontranslation due to non-
sense-mediated decay. On a protein level, it may
lead to a premature termination at Lys387/Arg424
with an 84-residue C-terminal truncation, which
encompasses the hHMGCS2 lower region and forms
part of the active site. This mutation, when
translated, would likely generate a protein fragment
with a misformed active site and devoid of activity.
The 15 missense mutations are distributed across
the entire hHMGCS2 polypeptide, with a majority in
the upper region (Supplementary Fig. S3). The first
group of mutations affects amino acids in the
protein core or conformationally restricted areas.
The side chains of Ile19/56, Leu229/266, Leu270/
Met307, and Ile370/407 are positioned within the
hydrophobic core in close proximity (b 4 Å) to
nonpolar residues. Introduction of I19/56N, L229/
266S, L270/M307T, and I370/407T mutations will
unfavourably place polar side chains into a highly
hydrophobic region. Val17/54, Gly132/169,
Gly175/212, and Gly351/388 are tightly packed in
the thiolase fold. Mutations to bulky or charged
residues (V17/54M, G132/169D, G175/212R, and
G351/388R) might result in steric and electrostatic
clashes. Thr196/233 is located in the core β-sheet of
the thiolase-fold, and its side-chain hydroxyl group
contributes a hydrogen bond to Thr382/Ser419 on
the neighbouring β-strand. The T196/233A muta-
tion would remove this interaction and potentially
destabilize the fold.
A second group of mutations involves the substi-
tution of arginine and lysine residues near the
dimeric interface (R151/188H, K206/243E, R463/
500H, and R468/505Q). However, their side chains
are not involved in intermolecule hydrogen bonds.
The positively charged side chains of Lys206/243,
Arg463/500, and Arg468/505 contribute a salt
bridge each with the acidic residues Glu95/132,
Asp459/496, and Glu78/115, respectively. On the
other hand, Arg151/188 is involved in charge
neutralization with a cluster of acidic residues
(Asp85/122, Asp119/156, Asp149/186, and
Asn117/154). Therefore, substitution of the strongly
positive charges (Arg/Lys side chain) with a weakly
positive charge (His side chain) or negative charge
(Gln/Glu side chain) will compromise the charge-
neutralization effect.
Y130/167C is the only reported mutation at
the active site. The Cβ atom of Tyr130/167 is
Crystal Structures of Human HMGCS1 and HMGCS2 503

Table 2. Summary of clinical mutations for hHMGCS2 deficiency

Alignmenta Consequence of mutation
Mutations hHMGCS1 bjHMGCS efHMGCS Protein stability Enzyme activity Predicted structural effectsb Reference
Missense
25
V54M V17 V I n.r. n.r. Steric and electrostatic clash
c
I56N I19 I I n.r. n.r. Protein misfold
25
Y167C Y130 Y Y n.r. n.r. Dimerization?
26
G169D G132 G A n.r. n.r. Steric and electrostatic clash
F174L F137 L Q Labile 104-fold decreased Steric clash 27
28
R188H R151 R K n.r. n.r. Change point charges
23
G212R G175 A G n.r. Not active Steric and electrostatic clash
c
T233A T196 S M n.r. n.r. Protein misfold
c
K243E K206 K R n.r. n.r. Change point charges
26
L266S L229 L F n.r. n.r. Protein misfold
28
M307T L270 L M n.r. n.r. Protein misfold
26
G388R G351 A L n.r. n.r. Steric and electrostatic clash
26
I407T I370 V I n.r. n.r. Protein misfold
23
R500H R463 R R n.r. Not active Change point charges
R505Q R468 K — n.r. n.r. Change point charges 26

Nonsense
c
Y84X Y47 Y F n.a. n.a. Premature termination
27
R424X K387 R E n.a. n.a. Premature termination
n.r. not reported; n.a. not applicable because truncated protein is predicted to be nonfunctional.
a
Equivalent residues in hHMGCS1, bjHMGCS, and efHMGCS based on sequence alignment.
b
This study.
c
Unpublished mutations; patient samples provided by O. Elpeleg (I56N), M. Leichsenring (T233A, K243E), and P. Clayton (Y84X).
d
Enzymatic characterization of the hHMGCS2 F174L mutation was performed on an hHMGCS1 construct with the analogous
mutation, due to the observed in vitro instability of the hHMGCS2 protein. For all mutations, functional effects were proven by finding
typical fasting intolerance and pathognomonic metabolic profiles in patients with homozygous and compound heterozygous mutant
genotypes.

approximately b5 Å from the thiol group of the
preceding residue, the catalytic nucleophile Cys129/
166. The Y130/167C mutation would place two thiol
groups in close proximity. Inspection of the structure
shows that formation of a vicinal disulfide bond is
stereochemically unfavourable and seems unlikely.
However, the mutation may affect dimerization,
since the Tyr130/167 side-chain hydroxyl points
toward the dimeric interface, packing against Ile123/
160′ (∼ 4 Å), and forms a water-mediated hydrogen
bond with the Gly122/159 amide nitrogen of the
dimeric partner (Fig. 4a).
The F137/174L mutation, which reduced catalytic
efficiency by 10,000-fold,29 also resides in helix α6
that accommodates the nucleophile Cys129/166. In
hHMGCS2, Phe137/174 forms stacking interactions

Fig. 4. Structural basis of HMGCS2 deficiency. (a) A close-up view showing the local environment of Tyr130/167 in
hHMGCS2 at the dimer interface between two subunits (coloured cyan and grey). The water-mediated hydrogen bond is
shown in dashed lines. (b) Local environment of Phe137/174 in hHMGCS2 (backbone and side chains coloured cyan),
superimposed with the equivalent region in bjHMGCS (backbone and side chains coloured pink). In bjHMGCS, a Leu
residue is in the place of Phe137/174 in hHMGCS2.
504
with an invariant Arg194/231 from the neighbour-
ing “R228GLR231” loop (interplanar distance between
Arg194/231 and Phe137/174 side chains is ∼ 3.8 Å)
(Fig. 4b). Intriguingly, wild-type HMGCS from
lower eukaryotes such as B. juncea and Caenorhabditis
elegans have Leu in place of Phe.30 Superposition of
the hHMGCS2 and bjHMGCS structures shows that
the Cα position of Arg194/231 in hHMGCS2 is
displaced by 2.5 Å toward helix α6 compared to the
bjHMGCS counterpart, due to steric hindrance by
Tyr236/273 on the other side of the RGLR loop (in
bjHMGCS Tyr is replaced by the less bulky Leu). This
displacement of Arg194/231 in hHMGCS2 means
that a F137/174L mutation in the human enzyme
would induce steric clash with Arg194/231 (by
reducing the interplanar distance to ∼ 3.0 Å). To
avoid this, the mutant protein could either displace
Arg194/231 away, which is unlikely due to the
limited space with the opposing Tyr236/273, or
more likely cause a displacement of helix α6, which
may force the Cys129/166 nucleophile away from its
native conformation for catalysis.
To conclude, we report here the first structure
determination of mammalian HMGCS. Our struc-
tures provide snapshots along the reaction path and
confirm the roles of conserved active-site residues in
the catalytic mechanism of the human enzymes. They
also reveal nonconserved regions near the active site,
which may facilitate the development of species-
specific and isoform-specific inhibitors. In addition,
the hHMGCS2 structure provides a framework to
rationalize human mutations found in patients
suffering from mitochondrial HMGCS2 deficiency.
Materials and Methods
Expression, purification, and crystallization
DNA fragments encoding hHMGCS1 (aa 16–507;
GenBank entry 39930537) or hHMGCS2 (aa 51–508;
GenBank entry 5031751) were subcloned into the
pNIC28-Bsa4 and pNH-TrxT vectors, respectively, which
incorporate an N-terminal tobacco etch virus (TEV)-
cleavable hexahistidine (His6)-tag (pNIC28-Bsa4) 'or a
His6-thioredoxin tag (pNH-TrxT). The corresponding
plasmid was transformed into Escherichia coli BL21(DE3)-
R3, cultured in 1 L of Terrific Broth at 37 °C to OD600 ∼ 1.5,
and induced with 0.5 mM IPTG overnight at 18 °C.
Harvested cells were homogenized in buffer A [50 mM
Hepes (pH 7.5), 500 mM NaCl, 5% glycerol, 20 mM
imidazole] and centrifuged. The supernatant was applied
to Ni-Sepharose resin preequilibrated with buffer A and
eluted with buffer B (buffer A + 230 mM imidazole). The
eluant was applied to a Superdex S200 column preequili-
brated with buffer GF [10 mM Hepes (pH 7.5), 500 mM
NaCl, 5% (w/v) glycerol, 0.5 mM tris(2-carboxyethyl)
phosphine (TCEP)]. To remove fusion tag, purified protein
was treated with His-tagged TEV protease overnight at
4 °C and passed over Ni-Sepharose resin equilibrated with
buffer GF. Protein was concentrated to 15 mg/ml
(hHMGCS1) and 12 mg/ml (hHMGCS2). The final protein
yield was 6 mg (hHMGCS1) and 4 mg (hHMGCS2) per
litre of culture.
Crystal Structures of Human HMGCS1 and HMGCS2
Crystals were grown by the vapour-diffusion method.
For hHMGCS1, protein was preincubated with 5 mM Ac-
CoA prior to crystallization. Sitting drops containing
50 nL of protein and 100 nL of reservoir solution [0.1 M
Bis-Tris (pH 6.5), 25% (w/v) polyethylene glycol (PEG)
3350, 0.2 M ammonium sulfate] were equilibrated at 4 °C.
For hHMGCS2, protein was preincubated with 5 mM Ac-
CoA and 5 mM AcAc-CoA prior to crystallization. Sitting
drops containing 100 nL of protein and 50 nL of reservoir
solution [0.2 M ammonium sulfate, 0.1 M Bis-Tris (pH 6.5),
25% (w/v) PEG 3350] were equilibrated at 20 °C. Crystals
were cryoprotected using 25% glycerol and flash-cooled in
liquid nitrogen. Diffraction data were collected at the SLS
beamline X10S and processed with the CCP4 program
suite.31
Structure determination
hHMGCS1 crystallized in the tetragonal space group
P43212 with one dimer in the asymmetric unit (rmsd
between subunits, ∼ 0.10 Å). The structure was solved by
molecular replacement using the program PHASER32 and
the B. juncea structure (PDB code 2F82) as search model.
Automated model building was performed with ARP/
wARP,33 followed by iterative cycles of restrained
refinement with medium NCS restraints and model
building using COOT34 and REFMAC5.35 Difference
Fourier map revealed well-defined electron density for
CoA in both subunits of the asymmetric unit, where they
were modelled unambiguously.
hHMGCS2 crystallized in the primitive space group P1
with two dimers in the asymmetric unit (rmsd between
subunits, ∼ 0.07 Å). The structure was solved by molecular
replacement using the refined hHMGCS1 structure as
search model. Automated model building was performed
with ARP/wARP,33 followed by iterative cycles of
restrained refinement with medium NCS restraints and
model building using COOT34 and PHENIX.36 Difference
Fourier map revealed well-defined electron density for
HMG-CoA in all four subunits of the asymmetric unit,
where they were modelled unambiguously. TLS refine-
ment was applied to the final hHMGCS1/2 models with
TLS groups recommended by the TLS server‡. Figures
were prepared with PYMOL§. The CONSURF server37
was used to compute sequence conservation across
HMGCS sequences. Structure-based sequence alignment
was performed with SEQUOIA.38
PDB accession numbers
Atomic coordinates and structure factors have been
deposited in the PDB with the accession numbers 2P8U
(hHMGCS1) and 2WYA (hHMGCS2).
Acknowledgements
We are grateful to the SGC crystallography group
for synchrotron data collection and the biotechnol-
ogy group for construct cloning. We greatly appre-
ciate sharing of unpublished data and permission to
‡ http://skuld.bmsc.washington.edu/~tlsmd
§ www.pymol.org
Crystal Structures of Human HMGCS1 and HMGCS2
describe structural effects by James Pitt and collea-
gues. The Structural Genomics Consortium is a
registered charity (No. 1097737) funded by the
Canadian Institutes for Health Research, the Cana-
dian Foundation for Innovation, Genome Canada
through the Ontario Genomics Institute, GlaxoS-
mithKline, Karolinska Institutet, the Knut and Alice
Wallenberg Foundation, the Ontario Innovation
Trust, the Ontario Ministry for Research and
Innovation, Merck and Co., Inc., the Novartis
Research Foundation, the Swedish Agency for
Innovation Systems, the Swedish Foundation for
Strategic Research, and the Wellcome Trust.
Supplementary Data
Supplementary data associated with this article
can be found, in the online version, at doi:10.1016/
j.jmb.2010.03.034
References
1. Danpure, C. J. (1995). How can the products of a single
gene be localized to more than one intracellular
compartment? Trends Cell Biol. 5, 230–238.
2. Middleton, B. (1972). The kinetic mechanism of
3-hydroxy-3-methylglutaryl-coenzyme A synthase
from baker's yeast. Biochem. J. 126, 35–47.
3. Miziorko, H. M. & Lane, M. D. (1977). 3-Hydroxy-
3-methylgutaryl-CoA synthase. Participation of
acetyl-S-enzyme and enzyme-S-hydroxymethylgu-
taryl-SCoA intermediates in the reaction. J. Biol.
Chem. 252, 1414–1420.
4. Gil, G., Goldstein, J. L., Slaughter, C. A. & Brown,
M. S. (1986). Cytoplasmic 3-hydroxy-3-methylglutaryl
coenzyme A synthase from the hamster. I. Isolation
and sequencing of a full-length cDNA. J. Biol. Chem.
261, 3710–3716.
5. Yokoyama, C., Wang, X., Briggs, M. R., Admon, A.,
Wu, J., Hua, X. et al. (1993). SREBP-1, a basic–helix–
loop–helix–leucine zipper protein that controls tran-
scription of the low density lipoprotein receptor gene.
Cell, 75, 187–197.
6. Magnuson, K., Jackowski, S., Rock, C. O. & Cronan,
J. E., Jr. (1993). Regulation of fatty acid biosynthesis
in Escherichia coli. Microbiol. Rev. 57, 522–542.
7. Wilding, E. I., Brown, J. R., Bryant, A. P., Chalker,
A. F., Holmes, D. J., Ingraham, K. A. et al. (2000).
Identification, evolution, and essentiality of the
mevalonate pathway for isopentenyl diphosphate
biosynthesis in gram-positive cocci. J. Bacteriol. 182,
4319–4327.
8. Rodriguez, J. C., Gil-Gomez, G., Hegardt, F. G. &
Haro, D. (1994). Peroxisome proliferator-activated
receptor mediates induction of the mitochondrial
3-hydroxy-3-methylglutaryl-CoA synthase gene by
fatty acids. J. Biol. Chem. 269, 18767–18772.
9. Hegardt, F. G. (1999). Mitochondrial 3-hydroxy-3-
methylglutaryl-CoA synthase: a control enzyme in
ketogenesis. Biochem. J. 338(Pt 3), 569–582.
10. Morris, A. A., Lascelles, C. V., Olpin, S. E., Lake, B. D.,
Leonard, J. V. & Quant, P. A. (1998). Hepatic
mitochondrial 3-hydroxy-3-methylglutaryl-coenzyme
a synthase deficiency. Pediatr. Res. 44, 392–396.
505
11. Miziorko, H. M., Clinkenbeard, K. D., Reed, W. D. &
Lane, M. D. (1975). 3-Hydroxy-3-methylglutaryl
coenzyme A synthase. Evidence for an acetyl-S-
enzyme intermediate and identification of a cysteinyl
sulfhydryl as the site of acetylation. J. Biol. Chem. 250,
5768–5773.
12. Misra, I. & Miziorko, H. M. (1996). Evidence for
the interaction of avian 3-hydroxy-3-methylglutaryl-
CoA synthase histidine 264 with acetoacetyl-CoA.
Biochemistry, 35, 9610–9616.
13. Chun, K. Y., Vinarov, D. A., Zajicek, J. & Miziorko,
H. M. (2000). 3-Hydroxy-3-methylglutaryl-CoA
synthase. A role for glutamate 95 in general acid/
base catalysis of C–C bond formation. J. Biol. Chem.
275, 17946–17953.
14. Pojer, F., Ferrer, J. L., Richard, S. B., Nagegowda,
D. A., Chye, M. L., Bach, T. J. & Noel, J. P. (2006).
Structural basis for the design of potent and species-
specific inhibitors of 3-hydroxy-3-methylglutaryl CoA
synthases. Proc. Natl Acad. Sci. USA, 103, 11491–11496.
15. Campobasso, N., Patel, M., Wilding, I. E., Kallender,
H., Rosenberg, M. & Gwynn, M. N. (2004).
Staphylococcus aureus 3-hydroxy-3-methylglutaryl-
CoA synthase: crystal structure and mechanism.
J. Biol. Chem. 279, 44883–44888.
16. Jez, J. M. & Noel, J. P. (2000). Mechanism of chalcone
synthase. pKa of the catalytic cysteine and the role of
the conserved histidine in a plant polyketide synthase.
J. Biol. Chem. 275, 39640–39646.
17. Steussy, C. N., Vartia, A. A., Burgner, J. W., II,
Sutherlin, A., Rodwell, V. W. & Stauffacher, C. V.
(2005). X-ray crystal structures of HMG-CoA synthase
from Enterococcus faecalis and a complex with its second
substrate/inhibitor acetoacetyl-CoA. Biochemistry, 44,
14256–14267.
18. Theisen, M. J., Misra, I., Saadat, D., Campobasso, N.,
Miziorko, H. M. & Harrison, D. H. (2004). 3-Hydroxy-
3-methylglutaryl-CoA synthase intermediate complex
observed in “real-time”. Proc. Natl Acad. Sci. USA, 101,
16442–16447.
19. Steussy, C. N., Robison, A. D., Tetrick, A. M., Knight,
J. T., Rodwell, V. W., Stauffacher, C. V. & Sutherlin,
A. L. (2006). A structural limitation on enzyme
activity: the case of HMG-CoA synthase. Biochemistry,
45, 14407–14414.
20. Huang, W., Jia, J., Edwards, P., Dehesh, K., Schneider,
G. & Lindqvist, Y. (1998). Crystal structure of beta-
ketoacyl-acyl carrier protein synthase II from E. coli
reveals the molecular architecture of condensing
enzymes. EMBO J. 17, 1183–1191.
21. Reed, W. D., Clinkenbeard, D. & Lane, M. D. (1975).
Molecular and catalytic properties of mitochondrial
(ketogenic) 3-hydroxy-3-methylglutaryl coenzyme A
synthase of liver. J. Biol. Chem. 250, 3117–3123.
22. Rokosz, L. L., Boulton, D. A., Butkiewicz, E. A.,
Sanyal, G., Cueto, M. A., Lachance, P. A. & Hermes,
J. D. (1994). Human cytoplasmic 3-hydroxy-3-methyl-
glutaryl coenzyme A synthase: expression, purifica-
tion, and characterization of recombinant wild-type
and Cys129 mutant enzymes. Arch. Biochem. Biophys.
312, 1–13.
23. Lowe, D. M. & Tubbs, P. K. (1985). 3-Hydroxy-3-
methylglutaryl-coenzyme A synthase from ox liver.
Purification, molecular and catalytic properties.
Biochem. J. 227, 591–599.
24. Praaenikar, J., Afonine, P. V., Guncar, G., Adams, P. D.
& Turk, D. (2009). Averaged kick maps: less noise,
more signal… and probably less bias. Acta Crystallogr.,
Sect. D: Biol. Crystallogr. 65, 921–931.
506
25. Aledo, R., Zschocke, J., Pie, J., Mir, C., Fiesel, S.,
Mayatepek, E. et al. (2001). Genetic basis of mitochon-
drial HMG-CoA synthase deficiency. Hum. Genet. 109,
19–23.
26. Pitt, J., Peters, H., Mishra, A., Eggington, M., Johnson,
D., Hinderhofer, K. et al. (2009). Abstract at 11th
International Congress on Inborn Errors of Metabo-
lism, San Diego, CA.
27. Aledo, R., Mir, C., Dalton, R. N., Turner, C., Pie, J.,
Hegardt, F. G. et al. (2006). Refining the diagnosis
of mitochondrial HMG-CoA synthase deficiency.
J. Inherit. Metab. Dis. 29, 207–211.
28. Wolf, N. I., Rahman, S., Clayton, P. T. & Zschocke, J.
(2003). Mitochondrial HMG-CoA synthase deficiency:
identification of two further patients carrying two
novel mutations. Eur. J. Pediatr. 162, 279–280.
29. Bouchard, L., Robert, M. F., Vinarov, D., Stanley,
C. A., Thompson, G. N., Morris, A. et al. (2001).
Mitochondrial 3-hydroxy-3-methylglutaryl-CoA
synthase deficiency: clinical course and description
of causal mutations in two patients. Pediatr. Res. 49,
326–331.
30. Nagegowda, D. A., Bach, T. J. & Chye, M. L. (2004).
Brassica juncea 3-hydroxy-3-methylglutaryl (HMG)-
CoA synthase 1: expression and characterization of
recombinant wild-type and mutant enzymes. Biochem.
J. 383, 517–527.
31. CCP4 (1994). The CCP4 suite: programs for protein
crystallography. Acta Crystallogr., Sect. D: Biol. Crystallogr.
50, 760–763.
Crystal Structures of Human HMGCS1 and HMGCS2
32. McCoy, A. J., Grosse-Kunstleve, R. W., Storoni, L. C. &
Read, R. J. (2005). Likelihood-enhanced fast translation
functions. Acta Crystallogr., Sect. D: Biol. Crystallogr. 61,
458–464.
33. Perrakis, A., Morris, R. & Lamzin, V. S. (1999).
Automated protein model building combined with
iterative structure refinement. Nat. Struct. Biol. 6,
458–463.
34. Emsley, P. & Cowtan, K. (2004). Coot: model-building
tools for molecular graphics. Acta Crystallogr., Sect. D:
Biol. Crystallogr. 60, 2126–2132.
35. Murshudov, G. N., Vagin, A. A. & Dodson, E. J.
(1997). Refinement of macromolecular structures by
the maximum-likelihood method. Acta Crystallogr.,
Sect. D: Biol. Crystallogr. 53, 240–255.
36. Adams, P. D., Grosse-Kunstleve, R. W., Hung, L. W.,
Ioerger, T. R., McCoy, A. J., Moriarty, N. W. et al.
(2002). PHENIX: building new software for automat-
ed crystallographic structure determination. Acta
Crystallogr., Sect. D: Biol. Crystallogr. 58, 1948–1954.
37. Landau, M., Mayrose, I., Rosenberg, Y., Glaser, F.,
Martz, E., Pupko, T. & Ben-Tal, N. (2005). ConSurf
2005: the projection of evolutionary conservation
scores of residues on protein structures. Nucleic
Acids Res. 33, W299–W302.
38. Bruns, C. M., Hubatsch, I., Ridderstrom, M., Mannervik,
B. & Tainer, J. A. (1999). Human glutathione transfer-
ase A4-4 crystal structures and mutagenesis reveal the
basis of high catalytic efficiency with toxic lipid
peroxidation products. J. Mol. Biol. 288, 427–439.