The crystal structure of human
a
-amino-
b
-carboxymuconate-
e
-semialdehyde decarboxylase in complex with
1,3-dihydroxyacetonephosphate suggests a regulatory
link between NAD synthesis and glycolysis
Silvia Garavaglia
1
, Silvia Perozzi
1
, Luca Galeazzi
2
, Nadia Raffaelli
2
and Menico Rizzi
1
1 DiSCAFF Dipartimento di Scienze Chimiche, Alimentari, Farmaceutiche e Farmacologiche, University of Piemonte Orientale ‘A. Avogadro’,
Novara, Italy
2 Department of Molecular Pathology and Innovative Therapies, Section of Biochemistry, Universita
`
Politecnica delle Marche, Ancona, Italy
Introduction
In humans, tryptophan at a level that exceeds the basal
requirements for protein and serotonin synthesis is oxi-
datively degraded through the kynurenine pathway,
producing the highly unstable intermediate a-amino-
b-carboxymuconate-e-semialdehyde (ACMS) [1]. As
shown in Fig. 1, ACMS can be either non-enzymati-

physiological and pathological conditions, mainly affecting the central ner-
vous system. As their relative concentrations must be tightly controlled,
modulation of ACMSD activity appears to be a promising prospect for the
treatment of neurological disorders, including cerebral malaria. Here we
report the 2.0 A
˚
resolution crystal structure of human ACMSD in complex
with the glycolytic intermediate 1,3-dihydroxyacetonephosphate (DHAP),
refined to an R-factor of 0.19. DHAP, which we discovered to be a potent
enzyme inhibitor, resides in the ligand binding pocket with its phosphate
moiety contacting the catalytically essential zinc ion through mediation of
a solvent molecule. Arg47, Asp291 and Trp191 appear to be the key resi-
dues for DHAP recognition in human ACMSD. Ligand binding induces a
significant conformational change affecting a strictly conserved Trp–Met
couple, and we propose that these residues are involved in controlling
ligand admission into ACMSD. Our data may be used for the design of
inhibitors with potential medical interest, and suggest a regulatory link
between de novo NAD biosynthesis and glycolysis.
Abbreviations
ACMS, a-amino-b-carboxymuconate-e-semialdehyde; ACMSD, a-amino-b-carboxymuconate-e-semialdehyde decarboxylase; DHAP,
1,3-dihydroxyacetonephosphate; PA, picolinic acid; QA, quinolinic acid.
FEBS Journal 276 (2009) 6615–6623 ª 2009 The Authors Journal compilation ª 2009 FEBS 6615
acid-e-semialdehyde, which possibly collapses to picoli-
nic acid (PA) [2,3]. Therefore, by competing with the
non-enzymatic synthesis of QA, ACMSD ultimately
controls the metabolic fate of tryptophan catabolism
along the kynurenine pathway, and is a medically rele-
vant enzyme in light of the important roles played by
QA and PA in physiological and pathological condi-
tions. Indeed, QA is not only a key precursor of

eukaryotes, but also in some micro-organisms, in
which the enzyme plays a key role in both the trypto-
phan to QA transformation and catabolism of 2-nitro-
benzoic acid [14,15]. Extensive biochemical and
structural characterizations have been carried out on
Pseudomonas fluorescens ACMSD (PfACMSD), lead-
ing to the discovery that the enzyme is a member of
the metal-dependent amidohydrolase superfamily fea-
turing an (a ⁄ b)
8
TIM barrel fold [16–18]. Biochemical
and structural analysis of PfACMSD led to proposal
of a non-oxidative decarboxylation catalytic mecha-
nism, unprecedented amongst known decarboxylases
[18,19]. The gene encoding human ACMSD (hA-
CMSD) was identified few years ago [20], and very
recently the existence of two isoforms originating by
alternative splicing was demonstrated; although
comparably expressed in various organs, only the
hACMSD I isoform was reported to be enzymatically
active and extensively characterized [3]. hACMSD
shares a high degree of sequence identity with
PfACMSD (38%), with strict conservation of all resi-
dues that are proposed to play a key role in catalysis
and are involved in co-ordination of the catalytically
essential zinc ion. As no ACMSD structure with
bound substrate or inhibitor has been reported so far
from any source, understanding of the ACMSD
catalytic mechanism is still incomplete. In light of the
reported ACMSD upregulation in the liver of

COOH
Quinolinic acid
non enzymatic
NAD
Glutaryl CoA
CO2 + ATP
non enzymatic
N
COOH
Picolinic acid
Fig. 1. The reaction catalyzed by hACMSD in a metabolic context.
ACMS is derived from tryptophan degradation through the kynure-
ine pathway, and, depending on hACMSD activity, has various
metabolic destinies.
Crystal structure of human ACMSD S. Garavaglia et al.
6616 FEBS Journal 276 (2009) 6615–6623 ª 2009 The Authors Journal compilation ª 2009 FEBS
glycolytic intermediates on the enzyme activity.
Interestingly, several phosphorylated glycolytic
intermediates were found to be strong inhibitors of
hACMSD, of which 1,3-dihydroxyacetonephosphate
(DHAP) was the most potent and was therefore
selected for our structural investigation. Our results
provide the first structural image of an ACMSD in a
ligand-bound form, and may be used to assist the struc-
ture-based rational design of enzyme inhibitors with
potential medical interest.
Results and Discussion
Overall quality of the model
The three-dimensional structure of hACMSD in com-
plex with DHAP was solved by molecular replacement

8
barrel
domain and a C-terminal extension that comprises two
short a-helices. Functional hACMSD was previously
reported to be a monomer in solution [3]. Consistently,
one molecule is present in the asymmetric unit in our
crystal, although a dimer can be observed in the crys-
tal lattice by applying the crystallographic two-fold
axis. PfACMSD was reported to be a dimer, with
subunits related by a dyad axis in the crystal, and a
mixture of monomeric and dimeric forms in solution
[18]. Therefore, the available structural data suggest
that the minimal functional unit in the ACMSD
enzyme is a monomer. and the biological significance,
if any, of the loose dimer observed in the crystalline
state remains to be established. The overall structural
organization observed in hACMSD confirms the previ-
ous assignment of the enzyme to the metal-dependent
hydrolase superfamily [17,18], whose members feature
by a structurally conserved TIM a ⁄ b barrel fold [24].
The significant structural conservation observed
between hACMSD and PfACMSD extends to the
peculiar small insertion domain, which may be consid-
ered a unique trait of ACMSDs.
The metal centre and the ligand binding site
The hACMSD active site is located in a crevice on the
protein surface at the C-terminal opening of the b-bar-
rel (Fig. 2), with a Zn
2+
ion occupying the metal

centre into a small pocket delimited by residues
Asp291, Trp191, Met195, Arg47 and the Phe294-
Pro295-Leu296 amino acid stretch (Fig. 3A,B). The
ligand binds in an extended conformation, with its
phosphate moiety located in the proximity of the zinc
ion and with the hydroxymethylene group pointing
toward Pro295. DHAP interacts with the catalytically
essential Zn
2+
through mediation of the metal-coordi-
nating solvent molecule w1, which establishes two
strong hydrogen bonds with the ligand O1 and O1P
atoms at 2.5 and 2.8 A
˚
, respectively. Moreover, the
ligand is engaged in a number of stabilizing interac-
tions with the protein milieu by contacting both pro-
tein residues and solvent molecules. In particular, the
DHAP phosphate moiety establishes a salt bridge with
Arg47 (distance of 3.0 A
˚
), an electrostatic interaction
with the Zn
2+
ion (closest distance of 4.1 A
˚
) and an
extensive network of hydrogen bonds involving a set
of well-ordered solvent molecules. O1P contacts the
solvent molecule w268 (2.8 A

tural and spectroscopic investigations, mainly carried
out on PfACMSD, led to the proposal of two possible
alternative catalytic mechanisms [18], whose common
feature involves formation of a tetrahedral intermedi-
ate resulting from the nucleophilic attack of the metal-
bound hydroxyl group onto the substrate, as observed
in other members of the amidohydrolase superfamily
[24,25]. However, as no structure of complexes with
either the substrate, product or inhibitors had been
reported, precise identification of the protein residues
involved in ligand recognition and catalysis remained
elusive.
Although our structural data do not allow us to
discriminate between the two alternative catalytic
His 6
His 8
His 174
Asp 291
w 1
DHAP
Arg 47
Leu 296
Trp 191
Pro 295
w 1
DHAP
w 268
w 119
w 51
Met 195

CMSD:DHAP complex represents the first structure of
an ACMSD in a ligand-bound form and can be used
to provide insights into catalysis. In particular, the
residues involved in inhibitor binding can be suggested
to be important players for recognition of the physio-
logical substrate ACMS. On the basis of our structure,
we propose that the strictly conserved Asp291 is a fun-
damental residue for catalysis, not only because it con-
tributes to Zn
2+
coordination but also because of its
direct involvement in substrate binding (Fig. 3A,B).
Indeed, as detailed above, Asp291 stabilizes DHAP
through interaction with the inhibitor carbonyl group,
and is therefore likely to demonstrate an equivalent
role on ACMS by possibly recognizing its aldehyde
group (Fig. 1). Our observation is in agreement with
what was observed for PfACMSD, where a significant
increase in the K
M
value was observed when the equiv-
alent residue (Asp294) was mutated to alanine [16].
Another residue emerging as a key molecular determi-
nant for ligand recognition in hACMSD is Arg47
(Fig. 3B). Its guanidinium group contacts both the
phosphate and the hydroxyl moieties present on
DHAP, and appears to be a major contributor to
ACMS recognition by stabilizing interactions with the
negatively charged carboxylic groups and with the
aldehydic portion of the physiological substrate. A

hACMSD appears to be an important enzyme control-
ling the cellular levels of QA, PA and NAD. As a
consequence, modulation of hACMSD activity could
be of considerable relevance in certain therapeutic con-
texts [2,26–30]. The various neurological disorders in
which severe imbalance in the kynurenyne pathway is
seen include cerebral malaria [31], where an elevated
level of the pro-inflammatory PA has been proposed to
contribute to the development of this frequently fatal
clinical manifestations of the disease [12,32]. Moreover,
this disease also features a significant depletion of the
NAD+ level [31,33]. We propose that hACMSD inhi-
bition could result in alleviation of cerebral malaria
symptoms by controlling both PA and NAD levels
[7,34–36].
Recently, a direct link between NAD synthesis and
diabetes has been reported [37], suggesting that an
increased NAD level is a desirable condition to combat
the disease. Intriguingly, ACMSD was reported to be
overexpressed in streptozotocin-induced diabetic rats
Trp 191
w 1
DHAP
Met 195
Fig. 4. Conformational changes affecting ACMSD upon ligand bind-
ing. The image was obtained by optimal superposition of the hA-
CMSD:DHAP and PfACMSD:ligand-free structures. Protein residues
are colored in white for hACMSD and in green for PfACMSD, with
the DHAP ligand in yellow; the protein portion shown by a ribbon
representation refers to hACMSD, as do the amino acid numbers.

interest to reduce life-threatening complications of
cerebral malaria, and as an important tool in validat-
ing our proposal of hACMSD as a novel drug target
for the treatment of diabetes and to investigate its
proposed novel regulatory role.
Experimental procedures
Enzyme expression, purification and inhibition
studies
The expression vector pHIL-D2-ACMSDI constructed pre-
viously [3] was used as a template to amplify the ACMSD
gene, resulting in a C-terminal (His6) fusion protein. The
amplicon was cloned into the pHIL-D2 vector and the
NotI-digested construct was used to transform Pichia pas-
toris GS115 cells [3]. Expression of the recombinant pro-
tein was achieved as described previously [3]. Purification
was performed as described previously [3] with the follow-
ing modifications. The hydroxylapatite column was washed
with 130 mm potassium phosphate buffer pH 7.0, 50 mm
NaCl, and the recombinant protein was eluted with
300 mm potassium phosphate buffer pH 7.0, 50 mm NaCl.
The active fractions were pooled and directly applied to a
HisTrap HP column equilibrated in 10 mm potassium
phosphate buffer pH 7.0, 100 mm NaCl. After extensive
washing with the equilibration buffer, elution was per-
formed with a linear gradient of imidazole from 0–0.3 m
in the same buffer. The active fractions were pooled and
diluted 10-fold with 50 mm potassium phosphate buffer,
concentrated by ultrafiltration with a YM30 membrane
(Millipore SpA, Milan, Italy) and used for the crystalliza-
tion trials. Using 400 mL of yeast culture, approximately

France). An X-ray fluorescence scan performed on the hA-
CMSD crystals using the same beamline, clearly indicated
the presence of a Zn metal ion bound to the enzyme. Anal-
ysis of the diffraction data set allowed us to assign the
crystal to the trigonal P3
2
21 or P3
1
21 space group, with
cell dimensions a = b = 86.27 A
˚
and c = 92.84 A
˚
,
containing one molecule per asymmetric unit with a
Table 1. Effect of glycolytic intermediates on the activity of hA-
CMSD. Inhibition values refer to the percentage inhibition exerted
by the indicated metabolite, relative to a reaction carried out in the
absence of any inhibitor (control). Experiments were performed at
37 °C in triplicate in the presence of 5 l
M ACMS substrate.
Metabolite
Metabolite
concentration (m
M)
Inhibition
(%)
Dihydroxyacetonephosphate 1
0.5
0.1

21 and P3
1
21 space groups. A clear solu-
tion was obtained for the former only, allowing us to
unambiguously assign P3
2
21 as the correct space
group. The initial model was subjected to iterative
cycles of crystallographic refinement using the pro-
gram refmac [41], alternated with manual graphic
sessions for model building using the program o [42].
Approximately 7% of the randomly chosen reflections
were excluded from refinement of the structure and
used for the free R-factor calculation [43]. The pro-
gram arp ⁄ warp [44] was used for adding solvent mol-
ecules. When the R-factor decreased to a value of 0.27
at 2.0 A
˚
resolution, inspection of the electron density
map in the enzyme active site clearly revealed the pres-
ence of one molecule of DHAP that was consequently
manually modelled based on both the 2F
o
)F
c
and
F
o
)F
c

tryptophan catabolism. FEBS J 274, 827–840.
4 Schwarcz R, Whetsell WO Jr & Mangano RM (1983)
Quinolinic acid: an endogenous metabolite that
produces axon-sparing lesions in rat brain. Science 219,
316–318.
5 Stone TW, Mackay GM, Forrest CM, Clark CJ &
Darlington LG (2003) Tryptophan metabolites and
brain disorders. Clin Chem Lab Med 41, 852–859.
6 Beninger RJ, Colton AM, Ingles JL, Jhamandas K &
Boegman RJ (1994) Picolinic acid blocks the neurotoxic
but not the neuroexcitant properties of quinolinic acid
in the rat brain: evidence from turning behaviour and
tyrosine hydroxylase immunohistochemistry. Neurosci-
ence 61, 603–612.
7 Guillemin GJ, Cullen KM, Lim CK, Smythe GA,
Garner B, Kapoor V, Takikawa O & Brew BJ (2007)
Characterization of the kynurenine pathway in human
neurons. J Neurosci 27, 12884–12892.
8 Shin M, Kim I, Inoue Y, Kimura S & Gonzalez FJ
(2006) Regulation of mouse hepatic a-amino-b-carbo-
xymuconate-e-semialdehyde decarboxylase, a key
enzyme in the tryptophan–nicotinamide adenine dinu-
cleotide pathway, by hepatocyte nuclear factor 4a and
peroxisome proliferator-activated receptor a. Mol
Pharmacol 70, 1281–1290.
9 Ogata S, Takeuchi M, Fujita H, Shibata K, Okumura K
& Taguchi H (2000) Apoptosis induced by niacin-related
Table 2. Data collection and refinement statistics.
Parameter Value
Resolution (A

˚
2
) 25.6
S. Garavaglia et al. Crystal structure of human ACMSD
FEBS Journal 276 (2009) 6615–6623 ª 2009 The Authors Journal compilation ª 2009 FEBS 6621
compounds in K562 cells but not in normal human
lymphocytes. Biosci Biotechnol Biochem 64, 1142–1146.
10 Fernandez-Pol JA, Klos DJ & Hamilton PD (2001)
Antiviral, cytotoxic and apoptotic activities of picolinic
acid on human immunodeficiency virus-1 and human
herpes simplex virus-2 infected cell. Anticancer Res 21,
3773–3776.
11 Bosco MC, Rapisarda A, Reffo G, Massazza S,
Pastorino S & Varesio L (2003) Macrophage activating
properties of the tryptophan catabolyte picolinic acid.
In The Developments in Tryptophan and Serotonin
Metabolism (Allegri G, Costa CVL, Ragazzi E,
Steinhart H & Varesio L eds), pp 55–65. Kluwer
Academic ⁄ Plenum Publishers, New York.
12 Miu J, Ball HJ, Mellor AL & Hunt NH (2009) Effect
of indoleamine dioxygenase-1 deficiency and kynurenine
pathway inhibition on murine cerebral malaria. Int J
Parasitol 39, 363–370.
13 Clark CJ, Mackay GM, Smythe GA, Bustamante S,
Stone TW & Phillips RS (2005) Prolonged survival of a
murine model of cerebral malaria by kynurenine path-
way inhibition. Infect Immun 73, 5249–5251.
14 Colabroy KL & Begley TP (2005) Tryptophan cata-
bolism: identification and characterization of a new
degradative pathway. J Bacteriol 187, 7866–7869.

a-amino-b
-carboxymuconate-e-semialdehyde decarbox-
ylase (ACMSD). J Biol Chem 277, 35162–35167.
21 Tanabe A, Egashira Y, Fukuoka SI, Shibata K &
Sanada H (2002) Expression of rat hepatic 2-amino-
3-carboxymuconate-6-semialdehyde decarboxylase is
affected by a high protein diet and by streptozotocin-
induced diabetes. J Nutr 132, 1153–1159.
22 Sasaki N, Egashira Y & Sanada H (2009) Production
of l-tryptophan-derived catabolites in hepatocytes from
streptozotocin-induced diabetic rats. Eur J Nutr 48,
145–153.
23 Laskowski RA, MacArthur MW, Moss DS & Thornton
JM (1993) PROCHECK: a program to check the stereo-
chemical quality of protein structures. J Appl Crystal-
logr 26, 283–291.
24 Seibert CM & Raushel FM (2005) Structural and cata-
lytic diversity within the amidohydrolase superfamily.
Biochemistry 44, 6383–6391.
25 Liu A & Zhang H (2006) Transition metal-catalyzed
nonoxidative decarboxylation reactions. Biochemistry
45, 10407–10411.
26 Berger F, Ramı
´
rez-Herna
´
ndez MH & Ziegler M (2004)
The new life of a centenarian: signalling functions of
NAD(P). Trends Biochem Sci 29, 111–118.
27 Belenky P, Bogan KL & Brenner C (2007) NAD+

Crystal structure of human ACMSD S. Garavaglia et al.
6622 FEBS Journal 276 (2009) 6615–6623 ª 2009 The Authors Journal compilation ª 2009 FEBS
metabolic fluxes of tryptophan in hepatocytes isolated
from rats fed a clofibrate-containing diet. Biochem
Pharmacol 52, 247–252.
36 Shin M, Ohnishi M, Iguchi S, Sano K & Umezawa C
(1999) Peroxisome-proliferator regulates key enzymes of
the tryptophan–NAD+ pathway. Toxicol Appl Pharma-
col 158, 71–80.
37 Revollo JR, Ko
¨
rner A, Mills KF, Satoh A, Wang T,
Garten A, Dasgupta B, Sasaki Y, Wolberger C, Town-
send RR et al. (2007) Nampt ⁄ PBEF ⁄ Visfatin regulates
insulin secretion in beta cells as a systemic NAD
biosynthetic enzyme. Cell Metab 6, 363–367.
38 Leslie AG (1999) Integration of macromolecular diffrac-
tion data. Acta Crystallogr D 55, 1696–1702.
39 Collaborative Computational Project Number 4 (1994)
The CCP4 suite: programs for protein crystallography.
Acta Crystallogr D 50, 760–763.
40 Navaza J (2001) Implementation of molecular replace-
ment in AMoRe. Acta Crystallogr D 57, 1367–1372.
41 Murshudov GN, Vagin AA & Dodson EJ (1997) Refine-
ment of macromolecular structures by the maximum-
likelihood method. Acta Crystallogr D 53, 240–255.
42 Jones TA, Zou JY, Cowan SW & Kjeldgaard M (1991)
Improved methods for building protein models in elec-
tron density maps and the location of errors in these
models. Acta Crystallogr A 47, 110–119.