REVIEW ARTICLE
Biosynthesis of D-arabinose in mycobacteria – a novel
bacterial pathway with implications for antimycobacterial
therapy
Beata A. Wolucka
Laboratory of Mycobacterial Biochemistry, Institute of Public Health, Brussels, Belgium
Keywords
cell wall biosynthesis; D-ribose; ethambutol;
Mycobacterium tuberculosis; mycolic acid;
polyisoprenoid glycolipid; review
Correspondence
B. A. Wolucka, Laboratory of Mycobacterial
Biochemistry, Institute of Public Health, 642
Engeland Street, B-1180 Brussels, Belgium
Fax: +32 2 373 3282
Tel: +32 2 373 3100
E-mail: bwolucka@pasteur.be
(Received 8 February 2008, revised 6 March
2008, accepted 12 March 2008)
doi:10.1111/j.1742-4658.2008.06395.x
Decaprenyl-phospho-arabinose (b-d-arabinofuranosyl-1-O-monophospho-
decaprenol), the only known donor of d-arabinose in bacteria, and its
precursor, decaprenyl-phospho-ribose (b-d-ribofuranosyl-1-O-monophospho-
decaprenol), were first described in 1992. En route to d-arabinofuranose, the
decaprenyl-phospho-ribose 2¢-epimerase converts decaprenyl-phospho-ribose
to decaprenyl-phospho-arabinose, which is a substrate for arabinosyltransfe-
rases in the synthesis of the cell-wall arabinogalactan and lipoarabinomannan
polysaccharides of mycobacteria. The first step of the proposed decaprenyl-
phospho-arabinose biosynthesis pathway in Mycobacterium tuberculosis and
related actinobacteria is the formation of d-ribose 5-phosphate from sedohep-
tulose 7-phosphate, catalysed by the Rv1449 transketolase, and⁄or the isom-
erization of d-ribulose 5-phosphate, catalysed by the Rv2465 d-ribose
5-phosphate isomerase. d-Ribose 5-phosphate is a substrate for the Rv1017
phosphoribosyl pyrophosphate synthetase which forms 5-phosphoribosyl
1-pyrophosphate (PRPP). The activated 5-phosphoribofuranosyl residue of
PRPP is transferred by the Rv3806 5-phosphoribosyltransferase to decaprenyl
phosphate, thus forming 5¢-phosphoribosyl-monophospho-decaprenol. The
dephosphorylation of 5¢-phosphoribosyl-monophospho-decaprenol to deca-
prenyl-phospho-ribose by the putative Rv3807 phospholipid phosphatase is
the committed step of the pathway. A subsequent 2¢-epimerization of decapre-
nyl-phospho-ribose by the heteromeric Rv3790 ⁄Rv3791 2¢-epimerase leads to
the formation of the decaprenyl-phospho-arabinose precursor for the synthe-
sis of the cell-wall arabinans in Actinomycetales. The mycobacterial 2¢-epimer-
ase Rv3790 subunit is similar to the fungal d-arabinono-1,4-lactone oxidase,
the last enzyme in the biosynthesis of d-erythroascorbic acid, thus pointing to
an evolutionary link between the d-arabinofuranose- and l-ascorbic acid-
related pathways. Decaprenyl-phospho-arabinose has been a lead compound
for the chemical synthesis of substrates for mycobacterial arabinosyltransfe-
rases and of new inhibitors and potential antituberculosis drugs. The peculiar
(x,mono-E,octa-Z) configuration of decaprenol has yielded insights into lipid
biosynthesis, and has led to the identification of the novel Z-polyprenyl
diphosphate synthases of mycobacteria. Mass spectrometric methods were
developed for the analysis of anomeric linkages and of dolichol phosphate-
related lipids. In the field of immunology, the renaissance in mycobacterial
polyisoprenoid research has led to the identification of mimetic mannosyl-b-
1-phosphomycoketides of pathogenic mycobacteria as potent lipid antigens
presented by CD1c proteins to human T cells.
Abbreviations
ALO, D-arabinono-1,4-lactone oxidase; Araf,D-arabinofuranose; GLO, L-gulono-1,4-lactone oxidase; PRPP, 5-phosphoribosyl 1-pyrophosphate.
FEBS Journal 275 (2008) 2691–2711 ª2008 The Author Journal compilation ª2008 FEBS 2691
The family of mycobacteria comprises about 100
species, several of which are pathogens of humans
and ⁄or animals, including Mycobacterium tuberculosis,
M. bovis,M. leprae,M. avium-intracellulare,M. ulcerans
and M. marinum. The pathogenic mycobacteria are
inherently resistant to many antibacterial drugs and
can persist for years inside infected cells. Mycobacte-
rium tuberculosis, the aetiological agent of tuberculosis,
kills about 1.7 million people per year [1] and, accord-
ing to World Health Organization estimations, is pres-
ent in a latent form in about one-third of the world’s
population (http://www.who.int/tb/en). A combination
of several factors, such as the requirement of long-term
multidrug therapy for the treatment of tuberculosis,
the synergy between M. tuberculosis and human immu-
nodeficiency virus infections [2], the emergence of mul-
tidrug-resistant strains and, in particular, the recent
outbreaks of extensively drug-resistant tuberculosis
[3,4], has contributed to the persistence of tuberculosis
as a global public health problem.
Several existing antituberculosis drugs, including the
first-line drugs isoniazid and ethambutol, act at the
level of the cell wall. This vital structure plays a crucial
role in the virulence and pathogenicity of M. tuberculo-
sis. Mycobacteria possess a thick, highly impermeable
hydrophobic cell wall composed of a thin layer of pep-
tidoglycan, d-arabinofuranose (Araf)-containing arabi-
nogalactan and arabinomannan polysaccharides,
mannans, glucans, long-chain (C
70
–C
90
)a-branched,
b-hydroxy fatty acids (mycolic acids) and other lipids,
glycolipids, poly-l-glutamate–glutamine polymers,
enzymes and other proteins. Like teichoic acids in
other Gram-positive bacteria [5], arabinogalactan is
covalently attached to peptidoglycan by a phosphodi-
ester linkage. The arabinan part of arabinogalactan is,
in turn, esterified to mycolic acids, thus forming a pep-
tidoglycan–arabinogalactan–mycolic acid skeleton
(reviewed in [6]). This rigid model of the mycobacterial
cell wall is now being replaced by a more dynamic pic-
ture, in which the cell wall undergoes substantial modi-
fications in response to changing growth conditions, as
may occur in host cells, for example, after the
proposed transfer from phagosomal to cytosolic com-
partments [7]. The plasma membrane-anchored lipo-
arabinomannans and lipomannans, reminiscent of
lipoteichoic acid, are probably translocated to the
outer layer of the cell wall and processed to lipid-free
arabinomannans and mannans [8]. The presence, at
least transient, of different proteins and enzymes in the
M. tuberculosis cell wall, such as the porins that are
involved in the transport of hydrophilic molecules
[9,10], the catalase-peroxidase katG [11,12], the heat
shock protein 60 chaperones (GroEL1) that assist lipid
traffic [13,14], the antigen 85 mycolyltransferases [15]
complexed with a histone-like protein [16], the gluta-
mine synthetase involved in the synthesis of poly-l-
glutamate–glutamine polymers [17], serine ⁄threonine
protein kinases [18–20] and other virulence factors
[21,22], points to the dynamic structure, and suggests
an active role of the organelle in host–pathogen inter-
actions. Indeed, profound alterations of the cell-wall
composition are thought to occur that could lead to
antigenic variation [23] and isoniazid resistance [24] of
non-replicating, dormant M. tuberculosis found in per-
sistent infections. Moreover, during human infection,
the pathogen elaborates new macromolecular struc-
tures at the cell surface: pili, putative host colonization
factors [25].
d-Arabinose occurs rarely in nature. In contrast with
d-arabinopyranose, which is found in some eukaryotes,
such as trypanosomatids and plants, Arafis confined
to the prokaryotic world, where it is a constituent of
cell-surface polymers and glycolipids. In mycobacteria
and related Actinomycetales species, Arafis a compo-
nent of the arabinan parts of the arabinogalactan and
(lipo)arabinomannan polymers of the cell wall and of
some glycerol-based glycolipids [26]. The branched
arabinan chains of the arabinogalactan are attached to
the linear galactan backbone. The arabinan consists of
an inner linear region of Araf-(1 fi5)-a-Arafand of
branched non-reducing terminal Ara6 motifs: Arafb1
fi2Arafa1fi5(Arafb1fi2Arafa1fi3)Arafa1fi
5Arafa1. About two-thirds of the terminal b-Araf
and the penultimate 2-a-Arafserve as attachment
sites for mycolic acids (reviewed in [6]).
The arabinan part of the M. tuberculosis lipoara-
binomannan consists of linear segments of Araf-
(1 fi5)-a-Arafwith some a(1 fi3) branching. The
non- reducing termini are composed of two distinct
motifs: the Ara6 motif similar to that present in arabi-
nogalactan, and a simplified linear Ara4 motif: Ara-
fbfi2Arafa1fi5Arafa1fi5Arafa1. Some of the
non-reducing arabinofuranose termini are capped with
short chains of a(1 fi2) d-mannose [27].
The physiological role of arabinans was thought to
be exclusively structural and of similar importance
within Corynebacterineae (the mycobacteria ⁄nocar-
dia ⁄corynebacteria group); however, recent studies have
challenged this simplistic view. For example, arabinan-
devoid mutants of corynebacteria can be obtained
[28,29], whereas abrogation of arabinan synthesis is
lethal in mycobacteria. In addition, the complex regula-
tion [30] and functions [31] of arabinan-assembling
Emb proteins suggest that this polymer could play a
role in sensing mechanisms and possibly other pro-
cesses, in particular in pathogenic mycobacteria.
A role for the D-arabinose lipid carrier B. A. Wolucka
2692 FEBS Journal 275 (2008) 2691–2711 ª2008 The Author Journal compilation ª2008 FEBS
Despite the efforts of many research groups, the bio-
synthesis of d-arabinose in mycobacteria was an
enigma for many years until the isolation of decaprenyl
-phospho-arabinose and its decaprenyl-phospho-ribose
precursor in 1990, and the proposal of the last step of
d-arabinose synthesis catalysed by a 2¢-epimerase
(Scheme 1) [32]. The subsequent structural charac-
terization of both the b-d-arabinofuranosyl-1-
monophosphodecaprenol (Fig. 1B) [33] and the
b-d-ribofuranosyl-1-monophosphodecaprenol (Fig. 1C)
[34] allowed the biological origins of bacterial Arafto
be deciphered, and a new era in the study of cell-wall
biosynthesis in mycobacteria to be started.
The discovery: decaprenyl-phospho-
arabinose, decaprenyl-phospho-ribose
and other endogenous lipid-linked
sugars of mycobacteria
In spite of several claims of the existence of activated
nucleotide and 1-phosphate derivatives of d-arabinose
[35–37], water-soluble activated forms of d-arabinose,
Scheme 1. The original scheme of biosynthesis of D-arabinofuranosyl residues of the cell-wall arabinogalactan and lipoarabinomannan in
mycobacteria, including a feedback mechanism and possible sites of action of ethambutol, an antituberculosis drug [32]. Two possible sites
of ethambutol are indicated: 1, inhibition of arabinosyltransferase activity; 2, inhibition of certain step(s) in the biosynthesis of the acceptor X,
where X may be a polyprenyl-pyrophosphoryl-oligosaccharide or a growing polymer chain. Note that option 2, namely the inhibition of arab-
inan synthase activity (Emb), was demonstrated later by others (see text and Fig. 2). Ara
f
,D-arabinofuranose; Rib
f
,D-ribofuranose.
Fig. 1. Decaprenyl phosphate and decapre-
nyl-phospho-monosaccharides of myco-
bacteria. (A) The mycobacterial lipid carrier
C
50
-decaprenyl phosphate has a unique
stereoconfiguration and contains only one
trans (E)-isoprene residue at its x-end [33]
(see Fig. 3). (B) Decaprenyl-phospho-arabi-
nose, the only known D-arabinose donor for
the synthesis of the cell-wall arabinogalactan
and lipoarabinomannan in mycobacteria
[32,33]. (C) Decaprenyl-phospho-ribose, the
direct precursor of the b-D-arabinofuranosyl-
monophosphodecaprenol donor (B) and the
major form of the naturally occurring deca-
prenyl-phospho-sugars of mycobacteria
[32,34]. (D) The mycobacterial decaprenyl-
phospho-mannose, a minor component
[107].
B. A. Wolucka A role for the D-arabinose lipid carrier
FEBS Journal 275 (2008) 2691–2711 ª2008 The Author Journal compilation ª2008 FEBS 2693
such as d-arabinose phosphates and d-arabinose nucle-
otides, have never been demonstrated in mycobacteria.
Exogenously added d-arabinose is catabolized by a
spontaneous M. smegmatis mutant via an inducible,
fungal-like pathway [32,38,39] that converts an aldo-
pentose into a ketopentose [40] (Fig. 2). In the myco-
bacterial pathway, d-arabinose is reduced by a
NADPH-dependent d-arabinose dehydrogenase to
d-arabinitol, and the latter compound is oxidized to
d-xylulose by a NAD-dependent d-arabinitol dehydro-
genase. d-Xylulose can then be phosphorylated to
d-xylulose 5-phosphate and enter the pentose phos-
phate cycle [32,39]. In contrast with mycobacteria, the
majority of bacteria use either isomerase ⁄kinase or
oxidation pathways for the utilization of pentoses
[41,42]. Interestingly, the oxidation of d-arabinose to
A role for the D-arabinose lipid carrier B. A. Wolucka
2694 FEBS Journal 275 (2008) 2691–2711 ª2008 The Author Journal compilation ª2008 FEBS
d-arabinono-1,4-lactone does not occur in mycobacte-
ria [32,39], but in fungi, where it has been believed, at
least until recently [43], to be involved in the biosyn-
thesis of d-erythroascorbic acid [44].
After a fruitless search for water-soluble intermedi-
ates of d-arabinose, we looked for lipid-linked pyro-
phospho-oligosaccharides similar to the dolichol-linked
oligosaccharides of archaebacteria [45]. Indeed, gradi-
ent-eluted DEAE-cellulose fractions of organic extracts
from M. smegmatis contained lipid-linked galactose-
oligosaccharides, but also large amounts of mono-
charged, acid-labile arabinose, ribose and mannose
linked to phosphorylated isoprenoid lipids, although
some mycolic acids could be detected as well. Subse-
quent analysis of the monocharged glycolipids by fast-
atom bombardment mass spectrometry demonstrated
the presence of decaprenyl-phospho-pentoses and deca-
prenyl phosphate ions at m⁄z909 and m⁄z777, respec-
tively [32]. This was the beginning of a fruitful search
that has led to the identification of the d-arabinose
pathway, and to a better understanding of cell-wall bio-
synthesis and of the mechanism of action of ethambutol
in mycobacteria. In particular, we discovered that eth-
ambutol does not interfere with decaprenyl-phospho-
arabinose synthesis, and that the site of action of the
drug is downstream in the arabinan pathway [32].
Accordingly, it was proposed that: (a) decaprenyl-phos-
pho-arabinose is synthesized via a 2¢-epimerization of
decaprenyl-phospho-ribose, and serves as the donor of
d-arabinofuranosyl residues in the biosynthesis of the
cell-wall arabinogalactan and (lipo)arabinomann; (b)
ethambutol inhibits an arabinosyltransferase or an
arabinan-forming enzyme, and this inhibition results in
the accumulation of decaprenyl-phospho-arabinose in
mycobacteria; (c) the synthesis of the decaprenyl-phos-
pho-ribose precursor is controlled by a feedback mecha-
nism (Scheme 1) [32]. These conclusions have proven to
be correct and have served as the basis for further
research.
The details of the decaprenyl-phospho-arabinose
structure, including the determination of the absolute
configuration, anomeric linkage and ring form of the
d-arabinosyl residue, were solved later using combined
proton-NMR spectroscopy, gas chromatography and
mass spectrometry (Fig. 1B) [33]. NMR analysis also
allowed the determination of the particular structure
of the mycobacterial decaprenol with important impli-
cations regarding its biosynthesis (Figs 1A and 3). It
was a big surprise for us to find that what is lacking in
the 10 isoprene unit-containing C
50
-decaprenol of
mycobacteria is not a cis (Z)-unit, but one of the two
trans (E)-isoprene units that are localized at the x-end
of the known polyisoprenyl lipid carriers, including the
common bacterial undecaprenol. The proposed
x,mono-E,octa-Zconfiguration of the mycobacterial
decaprenol [33] was, in fact, the first hint of the exis-
tence of unusual Z-prenyl diphosphate synthases in
mycobacteria: a Z-farnesyl diphosphate synthase that
would provide an x,E,Z-farnesyl diphosphate for a
subsequent specific enzyme, a Z-decaprenyl diphos-
phate synthase. These unusual enzymes have been
identified recently (see below).
The structure of the endogenous b-d-arabinofurano-
syl-1-monophosphodecaprenol of mycobacteria was
solved (Fig. 1B). This was unprecedented because,
until that time, no other natural lipid-linked sugar iso-
lated from an organism had been fully structurally
characterized [46,47].
The next step was the structural elucidation of deca-
prenyl-phospho-ribose (Fig. 1C) [34]. The presence of
substantial amounts of decaprenyl-phospho-ribose was
puzzling because no ribose-containing polymers have
ever been described in mycobacteria. We proposed that
decaprenyl-phospho-ribose is converted to decaprenyl-
phospho-d-arabinose by a novel 2¢-epimerase of
mycobacteria (Scheme 1) [32]. The decaprenyl-
phospho-ribose 2¢-epimerase has been identified
recently.
Fig. 2. The metabolism of D-arabinose in mycobacteria. The fungal-like assimilation pathway for D-arabinose of Mycobacterium smegmatis
[32,39] is shown (top reactions). Decaprenyl-phospho-D-arabinose, the only known D-arabinofuranose donor, and decaprenyl-phospho-ribose
(in rectangles), were isolated from M. smegmatis [32] and structurally characterized (see Fig. 1). Decaprenyl-phospho-arabinose was pro-
posed to be synthesized via a 2¢-epimerization of decaprenyl-phospho-ribose, and to control the synthesis of the latter compound by a feed-
back mechanism. The heteromeric decaprenyl-phospho-ribose 2¢-epimerase (Rv3790 ⁄Rv3791) was identified recently. Ethambutol, a first-line
drug for the treatment of tuberculosis, inhibits the utilization of decaprenyl-phospho-arabinose [32,33] at the level of the Emb proteins that
are involved in the formation of arabinans [75,88]. The enzymatic steps leading from the well-known 5-phosphoribosyl 1-pyrophosphate
(PRPP) intermediate to the formation of decaprenyl-phospho-ribose were identified later by in vitro assays. D-Ribose 5-phosphate, the direct
precursor of PRPP, is proposed to be synthesized mainly by an essential transketolase (Rv1449) of the non-oxidative pentose phosphate
pathway. A possible involvement of a non-essential ribose 5-phosphate isomerase (Rv2465) and of the oxidative pentose phosphate pathway
enzymes is also shown. Intermediates of the fungal-like catabolic pathway are shown in green; the non-oxidative and oxidative parts of the
pentose phosphate pathway are shown in blue and violet, respectively; the decaprenyl-phospho-arabinose pathway is shown in red. Essen-
tial genes of M. tuberculosis, as determined by Himar1-based transposon mutagenesis [52,133], are indicated in bold, and cloned genes are
underlined.
B. A. Wolucka A role for the D-arabinose lipid carrier
FEBS Journal 275 (2008) 2691–2711 ª2008 The Author Journal compilation ª2008 FEBS 2695
