Download Rampant horizontal gene transfer and phospho

Gene 318 (2003) 185 – 191
www.elsevier.com/locate/gene
Rampant horizontal gene transfer and phospho-donor change in the
evolution of the phosphofructokinase
Eric Bapteste a,*, David Moreira b, Hervé Philippe a,1
a
Equipe Phylogénie, Bioinformatique et Génome, UMR CNRS 7622, Université Pierre et Marie Curie, 9 quai St. Bernard, 75005 Paris, France
b
Unité d’Ecologie, Systématique et Evolution, UMR CNRS 8079, Université Paris-Sud, 91405 Orsay Cedex, France
Received 1 April 2003; received in revised form 2 June 2003; accepted 24 June 2003
Received by A. Roger
Abstract
Previous work on the evolution of the phosphofructokinase (PFK) has shown that this key regulatory enzyme of glycolysis has undergone
an intricate evolutionary history. Here, we have used a comprehensive data set to address the taxonomic distribution of the different types of
PFK (ATP-dependent and PPi-dependent ones) and to estimate the frequency of horizontal gene transfer (HGT) events. Numerous HGT
events appear to have occurred. In addition, we focused on the analysis of sites 104 and 124 (usually Gly104 + Gly124 or Asp104 + Lys124),
known to be involved in catalysis (J. Biol. Chem. 275 (2000) 35677). It revealed the existence of numerous sequences from distantly related
species carrying atypical combinations of amino acids. Several adaptive changes of phospho-donors, probably requiring a single mutation at
position 104, have likely occurred independently in many lineages. The analysis of this gene suggests the existence of a high rate of both
HGT and substitution in its active sites. These rampant HGT events and flexibility in phospho-donor use illustrate the importance of tinkering
in molecular evolution.
D 2003 Published by Elsevier B.V.
Keywords: Phosphofructokinase; Phylogeny; Evolution; Horizontal gene transfer; Adaptability
1. Introduction
Energy metabolism of many organisms is based on
glycolysis. The classical glycolytic pathway (the Embden –Meyerhof pathway) is regulated by the phosphofructokinase (PFK). However, at least two alternative pathways
to degrade glucose also exist. One of them, the EntnerDouderoff pathway, does not require PFK and is broadly
distributed in bacteria (Conway, 1992). The other one is
present only in some Archaea and requires an enzyme called
ADP-PFK, which does not seem to be homologous to PFK.
Abbreviations: EST, expressed sequence tag; GLK, glucokinase; HGT,
horizontal gene transfer; KDPG, 2-keto-3-deoxy-6-phosphogluconate; ML,
maximum likelihood; NJ, neighbour joining; PFK, phosphofructokinase;
PFP, PPi-phosphofructokinase.
* Corresponding author. Tel.: +33-1-44-27-34-70; fax: +33-1-44-2734-45.
E-mail address: [email protected] (E. Bapteste).
1
Present address: Canadian Institute for Advanced Research. Département de Biochimie, Université Montréal, C.P. 6128 Succursale CentreVille. Montréal, QC. Canada H3C 3J7.
0378-1119/$ - see front matter D 2003 Published by Elsevier B.V.
doi:10.1016/S0378-1119(03)00797-2
It belongs instead to the glucokinase (GLK) family of
kinases (Verhees et al., 2001).
PFK is central to the classical glycolytic pathway, which
is present in all three domains of life (Siebers et al., 1998),
so that one could assume that the evolution of this enzyme is
very constrained. However, this does not seem to be the
case, since various types of PFK exist. They use distinct
energy phospho-donors, such as ATP and inorganic pyrophosphate (PPi), and fulfil different tasks in the cell. The
widely distributed ATP-PFK catalyses an irreversible catabolic reaction, the phosphorylation of fructose-6-phosphate
to fructose-1,6-bisphosphate, while the PPi-PFK, also called
PPi-phosphofructokinase (PFP), catalyses the same reaction
in a reversible way and can thus function both in glycolysis
and gluconeogenesis. Furthermore, in many cases, the
physiological role of PFP is not obvious (Siebers et al.,
1998). More precisely, the simultaneous presence of PFP
and ATP-PFK (i.e., of a reversible and an irreversible
enzyme) in a single organism, even if rarely reported, has
suggested that PFP may perform an alternative unknown
function (Alves et al., 2001; Van Praag, 1997).
186
E. Bapteste et al. / Gene 318 (2003) 185–191
Biochemical approaches, including functional, structural,
and mutational analyses, have led to the determination of
several amino acids essential for ATP-PFK and PFP functions (Chi and Kemp, 2000; Moore et al., 2002), in
particular, the positions 104 and 124 (according to the
numbering of Escherichia coli ATP-PFK) (Chi and Kemp,
2000). First, it has been demonstrated that Gly at position
104, present in all ATP-PFK, is essential for the use of ATP
(Chi and Kemp, 2000; Moore et al., 2002; Van Praag, 1997).
By contrast, the Asp residue present at this position in PFP
prevents ATP from binding by sterical hindrance (Moore et
al., 2002). Second, when the position 124 in E. coli PFK is
occupied by Gly, the absence of a lateral chain allows room
for the alpha-phosphate of ATP. This is not the case when a
larger amino acid, Lys, is present at this position, as it occurs
in all characterised PFP (Hinds et al., 1998). In fact, this Lys
plays an important role in the recognition of PPi, which has
been biochemically tested in Entamoeba histolytica and in
several other organisms (Chi and Kemp, 2000; Hinds et al.,
1998; Lopez et al., 2002; Moore et al., 2002).
In summary, PFK working with ATP harbours Gly at
positions 104 and 124, while PFP has an Asp and a Lys (Chi
and Kemp, 2000; Claustre et al., 2002; Moore et al., 2002).
However, the atypical amino-acidic combination Gly104 and
Lys124 has been reported in E. histolytica (Chi and Kemp,
2000), Trypanosoma brucei (Claustre et al., 2002), Leishmania donovani (Lopez et al., 2002), and Chlamydia
trachomatis (Moore et al., 2002). For the three first cases,
biochemical characterisation indicates that these PFK use
ATP as a phospho-donor, and not PPi. Moore et al. (2002)
argued that this same phospho-donor is likely used also in
C. trachomatis.
ATP-PFK and PFP share a common ancestry, but phylogenies show a very complex evolutionary pattern (Müller
et al., 2001; Siebers et al., 1998). An ancient duplication has
been supposed to give birth to two groups of PFK, using PPi
and ATP as phospho-donor, respectively (Alves et al.,
1996). However, their monophyly is questioned (Müller et
al., 2001; Siebers et al., 1998) as well as the validity of the
amino acid content in sites 104 and 124 as a phylogenetic
signature. In fact, the evolutionary history of PFK does not
coincide in several points with accepted notions of organismic relationships. On the contrary, several duplications
and horizontal gene transfer (HGT) blur the phylogeny of
PFK (Müller et al., 2001). Some recent duplications were
followed by the differentiation of catalytic and regulatory
subunits in animals, fungi, and plants (Heinisch et al., 1989;
Kemp and Gunasekera, 2002; Poorman et al., 1984; Van
Praag, 1997). In addition, the fusion of the catalytic and
regulatory subunits in fungi, animals, and Dictyostelium led
to huge variation in sequence size of PFK. All these
duplications and HGT events can eventually lead to multiple
divergent copies of PFK in a single species.
In this work, we have analysed a large data set of ATPPFK and PFP sequences to study the distribution of the
different types of PFK (ATP-PFK, PFP, and those with
atypical active sites) and to estimate the frequency of HGT
events. We include a new sequence from the choanoflagellate Monosiga ovata providing an additional example of
HGT affecting eukaryotic species. We also looked carefully
at positions 104 and 124 and detected several atypical active
sites (Gly104 + Lys124) in distantly related species. Hence,
we suggest that adaptive changes of phospho-donors, requiring a mutation at position 104, have occurred independently in many lineages. The activity of these particular
proteins should be biochemically tested, and if they possess
the expected ATP-PFK activity, species owning simultaneously ATP-PFK and PFP in their genomes would be in
fact quite common.
2. Materials and methods
2.1. Sequencing
The sequence of the pfk gene from M. ovata was
obtained from random sequencing of a cDNA library
(collaboration with Dr. P. Holland, to be published elsewhere). Both 5Vand 3Vends of the pfk clone were sequenced,
allowing and overlapping of around 500 nucleotides for a
total cDNA length of 897 nucleotides. The sequence has
been submitted to GenBank under the accession number
AY291291.
2.2. Sequence recovery and alignment
Most PFK protein sequences were retrieved from GenBank using the program ALIBABA (Philippe Lopez, unpublished work). The sequences from Dictyostelium
discoideum, fungi, and animals (formed by the fusion of
the catalytic and regulatory subunits) were split in two parts
to align separately the catalytic and the regulatory regions.
The sequences were aligned with CLUSTAL W (Thompson
et al., 1994) and the alignment was manually refined with
the program ED of the MUST package (Philippe, 1993).
To construct a more comprehensive data set, we also
included sequences obtained from ongoing expressed sequence tag (EST) and genome projects. PFK homologues
were detected by TBLASTN search (Altschul et al., 1997).
All of the high-scoring segments with a BLAST score below
10 10 were retained and incorporated into the alignment.
Only a few regions could be aligned without ambiguity
for the entire data set, notably those around the active site of
the enzyme. Our complete data set contained 227 sequences.
The alignment is available upon request.
2.3. Phylogenetic analysis
The complete PFK data set was initially analysed by
using the neighbour joining (NJ) method (Saitou and Nei,
1987). Partial and/or phylogenetically very closely related
sequences were discarded, yielding a final alignment that
E. Bapteste et al. / Gene 318 (2003) 185–191
contained 152 sequences and 153 unambiguously aligned
positions.
A phylogenetic tree with the 152 representative sequences was then reconstructed using TREE-PUZZLE 5.0
(Schmidt et al., 2002) and Neighbor (Felsenstein, 1999).
To handle rate variation among sites, a maximum likelihood
(ML) distance matrix with a G law model (eight discrete
classes) was computed and then used to reconstruct the tree
by the NJ method. Bootstrap values were computed upon
1000 replicates using PUZZLEBOOT (www.tree-puzzle.de/
puzzleboot.sh). An extended majority rule consensus tree
from the replicates was inferred by CONSENSE from the
PHYLIP package (Felsenstein, 1999). This NJ approach
allows to work on a vast number of sequences using a
complex model of sequence evolution, which is impossible
by a standard maximum likelihood (ML) analysis.
Statistical comparisons of alternative tree topologies
were carried out by applying the Shimodaira’s (2002)
approximately unbiased (AU) test implemented in the
program CONSEL (Shimodaira and Hasegawa, 2001).
3. Results
3.1. An odd taxonomic distribution
Our taxonomic sample contained almost exclusively
bacteria and eukaryotes, with the exception of the sequence
of the archaeon Thermoproteus tenax, most likely acquired
by HGT (Siebers et al., 1998). All completely sequenced
eukaryotic genomes contain a pfk gene, whereas it is missing
in some bacterial and in all archaeal completely sequenced
genomes. A BLAST search (using standard parameter values) shows that PFK is not present (with an expectation
value threshold at 10 30) in the complete genome sequences
of some alpha-proteobacteria (Brucella melitensis, Caulobacter crescentus, Rickettsia conorii, Rickettsia prowazekii),
some beta-proteobacteria (Neisseria meningitidis, Ralstonia
solanacearum), some gamma-proteobacteria (Pseudomonas
aeruginosa, Xanthomonas campestris, Xanthomonas citri),
epsilon-proteobacteria (Campylobacter jejuni, Helicobacter
pylori), the Gram-positive Oceanobacillus iheyensis, the
Green sulfur bacterium Chlorobium tepidum, the fusobacterium Fusobacterium nucleatum, and the cyanobacterium
Thermosynechococcus elongatus. Several of these species
lacking PFK (H. pylori, N. meningitidis, R. solanacearum,
S. typhimurium, and X. citri) possess a KDPG aldolase,
central in the Entner-Douderoff pathway.
3.2. Numerous HGT events in the PFK tree
A comprehensive data set (152 sequences, 153 unambiguously aligned positions) was used to construct a phylogenetic tree (Fig. 1). It allowed the definition of eight
monophyletic groups, five of them containing both bacterial
and eukaryotic sequences. Five of these groups (X, P,
187
SHORT, LONG, and III) have been previously named by
Siebers et al. (1998) and Müller et al. (2001). In this work,
we name three additional groups: B1 (containing only
bacteria), B2 (containing Clostridium spp. and several
proteobacteria), and E (containing mainly eukaryotes). In
addition, four sequences (Aquifex aeolicus, Dictyoglomus
thermophilum, T. tenax, and one sequence from Thermotoga
maritima) do not belong to any of the eight monophyletic
groups. The phylogenetic tree is very complex, suggesting
an intricate evolutionary history involving duplications and
HGT in addition to vertical inheritance.
The finding of several monophyletic groups with a wide
and phylogenetically coherent taxonomic composition
(e.g., the bacterial clade B1 and the eukaryotic clade E)
is in agreement with an extensive vertical inheritance.
Moreover, phylogenetic relationships within these groups
are coherent with accepted phylogenetic groups based on
ribosomal RNA and protein markers (Baldauf et al., 2000;
Brochier et al., 2002; Embley et al., 1992; Van de Peer et
al., 2000). For instance, metazoa are sister group to fungi
in group E, and gamma proteobacteria emerge within a
well-supported clade. Within group B1, we found both the
monophyly of Thermus + Deinococcus and that of Cytophaga + Bacteroides.
Yet, other regions of the tree are more puzzling. Some
species that are thought to be closely related are separated
among several groups (e.g., the alpha-proteobacteria in
groups B1, B2, P, and III). In other cases, some groupings
were recovered with unusual relationships inside a group
(e.g., spirochetes forming a monophyletic group with the
choanoflagellate M. ovata and plants in group X). This may
have resulted from ancient duplication events followed by
differential gene losses. However, the number of duplications and losses needed to explain that the complete phylogeny would probably be very large, so that at least for a
number of cases, independent HGT events seem a parsimonious explanation. One of the clearest examples of HGT
concerns the choanoflagellate M. ovata, expected to branch
in clade E as sister of metazoa (Lang et al., 2002) but which
emerges as relative of spirochetes in clade X (Fig. 1).
HGT may also concern species with several pfk copies
(see numbers on the right of the species names in Fig. 1).
For instance, a delta proteobacterium Desulfitobacterium
hafniense and a low GC Gram-positive bacterium Clostridium perfringens harbours three gene copies, branching in
groups B1, B2, and III. Some eukaryotes also have multiple
pfk gene copies. However, in some cases, they arose from
duplication events, such as in Dictyostelium, fungi, and
metazoa forming two paralogous clusters within the group
E, as supported by the tree topology. In other cases, one of
the copies has probably been acquired by HGT (e.g., E.
histolytica, present in groups X and LONG). More complex
scenarios simultaneously involving HGT and gene duplications are also found. For instance, in group LONG, plants,
apicomplexa (Plasmodium and Cryptosporidium), and chlamydiales have at least two copies. One possible interpreta-
188
E. Bapteste et al. / Gene 318 (2003) 185–191
51
Aquifex aeolicus
**G*C**
Magnetococcus sp.
2 **G*G**
Bacteroides fragilis
2 **G*G**
Cytophaga hutchinsonii
**G*G**
58
Thermus thermophilus
**G*G**
Deinococcus radiodurans
**G*G**
Desulfitobacterium hafniense
3 **G*G**
2 **G*G**
Clostridium difficile
Paenibacillus macquariensis
**G*G**
54
Clostridium perfringens
3 **G*G**
Clostridium botulinum
**G*G**
93
Clostridium acetobutylicum
**G*G**
Vibrio cholerae
**G*G**
Buchnera sp.
**G*G**
**G*G**
Yersinia pestis
Yersinia enterocolitica
**G*G**
100 Escherichia coli
**G*G**
91
Salmonella typhi
**G*G**
Klebsiella
pneumoniae
**G*G**
98
100
Enterobacter
cloacae
**G*G**
58
Haemophilus ducreyi
**G*G**
Pasteurella multocida
**G*G**
Haemophilus influenzae
**G*G**
82
54
100
Actinobacillus actinomycetemcomitans **G*G**
**G*G**
Staphylococcus aureus
**G*G**
Bacillus sphaericus
**G*G**
Bacillus halodurans
97
**G*G**
Bacillus subtilis
**G*G** ATP
Geobacillus stearothermophilus
86
60
**G*G**
Enterococcus faecium
70 Listeria innocua
**G*G**
**G*G**
Lactobacillus delbrueckii
**G*G**
Streptococcus pneumoniae
**G*G**
Streptococcus pyogenes
**G*G** ATP
80 54 Lactococcus lactis
Thermotoga maritima 2 **G*A** ATP
Dictyostelium discoideum
**G*G**
2 **G*G**
Chloroflexus aurantiacus
Schistosoma mansoni
**G*G**
57
Homo sapiens
**G*G** ATP
100
Homo sapiens
**G*G**
81 62 Homo sapiens
**G*G**
99
Drosophila melanogaster
**G*G** ATP
Caenorhabditis elegans
**G*G**
Haemonchus
contortus
**G*G** ATP
99
Aspergillus oryzae
**G*G**
Aspergillus oryzae
**G*G**
78
Pichia pastoris
**G*G**
54 53 Candida albicans
**G*G**
Saccharomyces
cerevisiae
96
**G*G** ATP
52 Kluyveromyces lactis
**G*G**
Schizosaccharomyces pombe **G*G**
Neurospora crassa
**G*G**
56
Pichia pastoris
**G*G**
Candida albicans
**G*G**
86
Saccharomyces cerevisiae **G*G**
Kluyveromyces lactis
**G*G**
100
86
B1
E
Dictyostelium discoideum **N*A** reg
100
94
Magnetospirillum magnetotacticum 2 **G*K**
Nostoc sp.
**G*K**
Synechocystis sp.
**G*K**
Synechocystis sp.
**G*K**
90
Nostoc sp.
**G*K**
96 99 Nostoc punctiforme
**G*K**
2 **G*K**
Chloroflexus aurantiacus
3 **G*K**
Clostridium perfringens
Desulfitobacterium hafniense 3 **G*K**
Myxococcus xanthus
**G*K**
III
66
**D*K**
Streptomyces coelicolor
**D*K**
Streptomyces coelicolor
95
**D*K**
Thermobifida fusca
56
**G*K** ATP
Streptomyces coelicolor
Amycolatopsis mediterranei
**D*K**
100
Amycolatopsis methanolica
2 **D*K** PPi
81
Corynebacterium diphtheriae
**G*K**
Mycobacterium leprae
**G*K**
91
**G*K**
Mycobacterium tuberculosis
100 Dictyoglomus thermophilum
**G*K**
Thermoproteus tenax **D*K** PPi
54
Propionibacterium freudenreichii **D*K** PPi
**D*K** PPi
Mastigamoeba balamuthi
P
Sinorhizobium meliloti
**D*K**
100
**D*K**
Agrobacterium
tumefaciens
97
Desulfitobacterium hafniense
3 **D*K**
Mesorhizobium loti
**D*K**
Magnetospirillum magnetotacticum 2 **D*K**
99
100
Magnetococcus sp.
2 **D*K**
67
Xylella
fastidiosa
**D*K**
100
Nitrosomonas europaea **D*K**
100
87
2 **D*K**
Clostridium difficile
54
Clostridium perfringens
3 **D*K**
Thermotoga maritima
**D*K** PPi
Naegleria fowleri
**D*K** PPi
SHORT
99
Trichomonas vaginalis
**D*K** PPi
80
Cryptosporidium parvum
Plasmodium falciparum
87
Chlamydia muridarum
Chlamydophila pneumoniae
100
96
Chlamydia trachomatis
98
Borrelia burgdorferi
Spirochaeta thermophila
Treponema pallidum
Entamoeba histolytica
Porphyromonas gingivalis
Bacteroides fragilis
91
Cryptosporidium parvum
Solanum tuberosum
Arabidopsis thaliana
96
Hexamita inflata
Giardia intestinalis
90
Chlamydophila pneumoniae
Chlamydophila psittaci
99
Chlamydia muridarum
90
58
93
Opisthokonta **E*A** reg
85
82
92
90
0.1
Entamoeba histolytica
Amycolatopsis methanolica
Trypanoplasma borreli
Trypanosoma brucei
94
Leishmania donovani
98
Monosiga ovata
Borrelia burgdorferi
Treponema denticola
75
Treponema pallidum
76
Arabidopsis thaliana
Oryza sativa
Oryza sativa
77
Arabidopsis thaliana
99
Arabidopsis thaliana
Arabidopsis thaliana
Arabidopsis thaliana
2 **G*K**
2 **G*K**
**G*K**
**G*K**
**G*K**
**G*K**
2 **G*K**
**G*K**
2 **G*K**
2 **G*K**
**G*K**
**G*K**
2 **G*K**
2 **G*K**
2 **G*K**
2 **G*K**
ATP
ATP
ATP
ATP
X
Zea mays 100
Solanum tuberosum
90
Arabidopsis thaliana
68
B2
**G*K**
**G*K**
**G*K**
**G*K**
**G*K**
2 **D*K**
**D*K**
2 **D*K**
2 **D*K**
**D*K**
2 **D*K**
**D*K**
**D*K**
2 **D*K**
**D*K**
**D*K**
**N*K**
**N*K**
**N*K**
C. parvum **F*A**
P. falciparum **I*L**
**T*V**
**T*V**
**T*V**
PPi
PPi
PPi
PPi
reg?
reg?
reg
reg
reg
LONG
E. Bapteste et al. / Gene 318 (2003) 185–191
tion is to invoke an ancient duplication in the ancestor of
plants and apicomplexa (or two independent duplications in
these two lineages) and the subsequent acquisition of the
two paralogous copies by the chlamydiales. Finally, species
for which a single copy has been identified are also involved
in HGT, such as the cases of T. tenax and Mastigamoeba
balamuthi (Müller et al., 2001; Siebers et al., 1998).
Our phylogenetic tree suggests several HGT events.
However, some of them have to be tested because of the
lack of resolution of certain regions in the tree. We have
therefore analysed alternative tree topologies that minimise
the number of HGT events. These topologies concerned
nodes supported by bootstrap values < 90%, since we have
assumed that nodes with superior bootstrap values were
correctly inferred. In particular, we have analysed three
alternative topologies affecting the groups that show a
mixture of eukaryotic and prokaryotic sequences, namely,
groups E, LONG, and X. In group E, we have constrained
the position of Chloroflexus aurantiacus at the base of the
group; in group LONG, we have constrained the monophyly
of the eukaryotic sequences, and in group X, we have
constrained the monophyly of the bacterial sequences.
Shimodaira’s approximately unbiased test confirmed that
these constrained topologies minimising HGT were significantly worse than the preferred tree. Therefore, it appears
that HGT events are a likely explanation for the complex
phylogeny of PFK.
3.3. A complex evolution of PFK function
This complicated PFK phylogeny that seems largely
blurred by HGT events, gene duplications, and gene losses
provides the framework to understand the relationships
between ATP-PFK and PFP as well as phospho-donor
transitions in PFK. Sensu stricto ATP-PFK, classically
defined on the base of their sequence conservation, notably
by the possession of two glycines in the active site, is
monophyletic. They were all included into the groups B1
and E. Similarly, four other monophyletic groups in the tree
(B2, P, SHORT, and X) showed a homogeneous amino acid
composition at the active site. Groups P, B2, and SHORT
included only Asp104 + Lys124 PFP, while the group X was
constituted exclusively by Gly104 + Lys124 PFK using ATP.
However, groups III and LONG contain a mixture of sequences with atypical amino acid combinations (Gly104 + Lys124
and Asp104 + Lys124), meaning that enzymes using ATP and
those using PPi are mixed within these two groups. Therefore, Gly104 + Lys124 PFK (see the species marked with
189
**G*K** in Fig. 1) and PFK using PPi probably appeared
several times in evolution. Consequently, these amino acid
combinations and, likewise, the use of ATP or PPi as
phospho-donor do not define reliable phylogenetic signatures. It is also possible to envisage that the transition from
one form to the other has occurred several times.
Nevertheless, amino acids of these two sites display very
little polymorphism, and their evolutionary conservation,
covariation, and involvement in phospho-donor recognition
have been discussed in several works (Chi and Kemp, 2000;
Claustre et al., 2002; Lopez et al., 2002; Moore et al., 2002).
In most catalytic PFK sequences, the position 104 is either a
Gly or an Asp, while position 124 is either a Gly or a Lys.
The exceptions were rare (see Fig. 1). In fact, in our data set,
the three sequences of the regulatory alpha subunits of
plants (Van Praag, 1997) and the two closely related
sequences from apicomplexa contained other amino acids,
which would be coherent with the fact that the alphasubunits do not link to phospho-donors. The other exceptions were the sequences from Chlamydophila psitacci, C.
pneumoniae, and Chlamydia muridarum (Asn at position
104), from A. aeolicus (Cys at position 124), and from T.
maritima (Ala at position 124).
4. Discussion
Despite being a key enzyme involved in glycolysis
regulation, PFK has a very complex evolutionary history.
The presence of this protein in several bacterial species
whose genomes have been completely sequenced but its
absence in closely related species (e.g., the alpha-proteobacteria) strongly suggests that PFK can be lost. However, the
loss of PFK is not synonymous of the loss of glycolysis (for
instance, in the species harbouring a KDPG aldolase, central
in the Entner-Douderoff pathway). Hence, since PFK can be
lost, it is not surprising to find a phylogenetic tree with odd
relationships. Such independent losses make the task of
identifying paralogous more difficult and could allow subsequent LGT of PFK in the species secondarily devoid of
PFK.
Moreover, flexibility in the use of PFK also resulted from
independent acquisitions of new PFKs by HGT. Our tree
illustrates that a key enzyme can be obtained from another
species. Indeed, HGT led to a complex phylogeny with
groups at odds with the classical relationships (Müller et al.,
2001). This could occur if the native PFK is replaced by a
new one or if the original copy is conserved together with
Fig. 1. Unrooted NJ tree for 152 PFK sequences and 153 amino acid positions based on distances calculated with a G law correction. Eukaryotic species are in
bold, while prokaryotic species are in italic. Numbers at nodes are bootstrap values (only values >50% are shown). Monophyletic groups are named according
to Müller et al. (2001) and this work. Solid circles indicate the most parsimonious distribution of G104 to D104 mutations on the tree. Numbers after species
names indicate species with multiple PFK copies in distantly related groups. The various amino acid combinations at positions 104 and 124 are reported after
species name, notably **G*G** for the typical ATP-PFK, **D*K** for the typical PFP, and **G*K** for putative atypical PFK using ATP. If the sequence
has been biochemically characterised, its phospho-donor is mentioned. Confirmed and putative regulatory subunits are indicated by reg and reg?, respectively.
The triangle corresponds to fungi and metazoa regulatory sequences. The scale bar corresponds to the number of substitutions per site.
190
E. Bapteste et al. / Gene 318 (2003) 185–191
the new copy, leading to species harbouring distantly related
PFK (e.g., in Amycolatopsis methanolica). In many cases,
conservation of two distantly related PFK enzymes in a
single species could be further explained in terms of
adaptability, each copy allowing the use of either a phospho-donor or the other, potentially enhancing the fitness of
the species. Such species have a PFK using ATP and a PFK
using PPi, one of which may have been obtained by HGT. In
fact, species with both enzymes would be or would have
been able to initiate glycolysis either with ATP or with PPi,
which may be adaptive, according to the relative concentrations of these metabolites in the cell and its environment.
For example, parasites may be advantaged by a PFK using
PPi to economise their limited stock of ATP (Moore et al.,
2002). Moreover, supplementary PFK copies may be
recruited in alternative metabolic pathways. For instance,
ATP-PFK participate in the RuMP cycle in Amycolatopsis
(Alves et al., 2001).
As a single-point mutation can induce a change of the
phospho-donor (Chi and Kemp, 2000), one can imagine that
the opposite mutation should revert it, allowing the PFK to
switch from the use of PPi to the use of ATP and vice versa.
Our phylogenetic tree suggests that these changes of phospho-donor concerned several species. One example can be
found within group III, containing a monophyletic group of
Gram-positive bacteria showing a mixture of sequences with
either Gly104 or Asp104. Moreover, the direction of these
changes can be suggested. It is most parsimonious to
postulate that Gly104 + Lys124 PFK using ATP was ancestral
and that PFK using PPi evolved more recently, in seven
independent occasions (i.e., once in T. tenax, once in group
LONG before the emergence of Borrelia burgdorferi, once
in the ancestor of group SHORT, once in the ancestor of
groups B2 and P, and three times in the group III, in the
ancestor of Thermobifida fusca, Streptomyces coelicolor,
and Amycolatopsis mediterranei). This suggests that at least
seven species have evolved from an irreversible glycolysis
towards a reversible one, whereas the opposite appears to
have never occurred in nature (Fig. 1). Yet, this result is very
sensitive to our still limited taxonomic sample and would
need additional confirmation.
We show here that the presence of multiple PFK copies
with different active sites is a common situation in nature. It
is notably the case for all the plants, some apicomplexa
(Cryptosporidium parvum, Plasmodium falciparum), some
amoeba (E. histolytica), some high GC Gram-positive
bacteria (S. coelicolor), some alpha-proteobacteria (Magneetococcus sp., Magnetospirillum magnetotacticum), some
low GC Gram-positive bacteria (D. hafniense), some spirochetes (B. burgdorferi, Treponema pallidum), and green
sulfur bacteria (C. aurantiacus). Thus, in these species,
multiple PFK copies may be involved in different metabolic
functions.
The complex phylogeny of PFK illustrates that tinkering
in evolution can concern fundamental molecular processes.
Our study emphasises the potential adaptability of this key
enzyme—lost, gained, and modifiable by single-point mutations—in living beings. Finally, our results complicates the
debate on the nature of the ancestral phospho-donor used by
the PFK, which opposes the advocates of the emergence of
metabolism in an organic or an inorganic context (see
Siebers et al., 1998; Chi and Kemp, 2000, for two opposed
opinions). We suggest that in several cases, the use of PPi as
phospho-donor may be derived. However, in addition to
possible convergent adaptive mutations, HGT events, including those of mutated sequences, further complicate the
distribution of these enzymes. Hence, it is not possible with
this tree to conclude which phospho-donor, ATP, or PPi was
ancestrally used by PFK.
Acknowledgements
We thank Simonetta Gribaldo and Miklós Müller for
critical reading of the manuscript, and Céline Brochier for
advice on bacterial phylogeny. We thank Peter Holland and
Elizabeth Snell for the M. ovata PFK sequence. Genome
and EST sequence data of the species E. histolytica, Oryza
sativa, Zea mays, C. parvum, and P. falciparum were obtained from databases deposited at GenBank (www.ncbi.
nlm.nih.gov/dbGSS/index.html and www.ncbi.nlm.nih.gov/
dbEST/index.html). Neurospora crassa EST data were obtained from the Center for Genome Research (www-genome.
wi.mit.edu/annotation/fungi/neurospora). Bacteroides fragilis, Clostridium difficile, Clostridium botulinum, Staphylococcus aureus, Streptococcus equi, Yersinia enterocolitica, and Corynebacterium diphtheriae sequence data
were obtained from the Sanger Institute (www.sanger.ac.uk/
DataSearch/). Treponema denticola and Chlamydophila
psittaci sequence data were obtained from the Institute for
Genomic Research (www.tigr.org/tdb/mdb/mdbcomplete.
html). Cytophaga hutchinsonii, D. hafniense, C. aurantiacus,
M. magnetotacticum, Nostoc punctiforme, and T. fusca
sequence data were obtained from the DOE Joint Genome
Institute (www.science.doe.gov/ober/EPR/mig_cont.html).
References
Altschul, S.F., Madden, T.L., Schaffer, A.A., Zhang, J., Zhang, Z., Miller,
W., Lipman, D.J., 1997. Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res. 25,
3389 – 3402.
Alves, A.M., Meijer, W.G., Vrijbloed, J.W., Dijkhuizen, L., 1996. Characterization and phylogeny of the pfp gene of Amycolatopsis methanolica encoding PPi-dependent phosphofructokinase. J. Bacteriol.
178, 149 – 155.
Alves, A.M., Euverink, G.J., Santos, H., Dijkhuizen, L., 2001. Different
physiological roles of ATP- and PP(i)-dependent phosphofructokinase
isoenzymes in the methylotrophic actinomycete Amycolatopsis methanolica. J. Bacteriol. 183, 7231 – 7240.
Baldauf, S.L., Roger, A.J., Wenk-Siefert, I., Doolittle, W.F., 2000. A kingdom-level phylogeny of eukaryotes based on combined protein data.
Science 290, 972 – 977.
E. Bapteste et al. / Gene 318 (2003) 185–191
Brochier, C., Bapteste, E., Moreira, D., Philippe, H., 2002. Eubacterial phylogeny based on translational apparatus proteins. Trends Genet. 18, 1 – 5.
Chi, A., Kemp, R.G., 2000. The primordial high energy compound: ATP or
inorganic pyrophosphate? J. Biol. Chem. 275, 35677 – 35679.
Claustre, S., Denier, C., Lakhdar-Ghazal, F., Lougare, A., Lopez, C., Chevalier, N., Michels, P.A., Perie, J., Willson, M., 2002. Exploring the active
site of Trypanosoma brucei phosphofructokinase by inhibition studies:
specific irreversible inhibition. Biochemistry 41, 10183 – 10193.
Conway, T., 1992. The Entner-Douderoff pathway: history, physiology and
molecular biology. FEMS Microbiol. Rev. 9, 1 – 27.
Embley, T.M., Thomas, R.H., Wlliams, R.A.D., 1992. Reduced thermophilic bias in the 16S rDNA sequence from Thermus ruber provides
further support for a relationship between Thermus and Deinococcus.
Syst. Appl. Microbiol. 16, 25 – 29.
Felsenstein, J., 1999. PHYLIP—Phylogeny Inference Package. University
of Washington, Seattle, WA.
Heinisch, J., Ritzel, R.G., von Borstel, R.C., Aguilera, A., Rodicio, R.,
Zimmermann, F.K., 1989. The phosphofructokinase genes of yeast
evolved from two duplication events. Gene 78, 309 – 321.
Hinds, R.M., Xu, J., Walters, D.E., Kemp, R.G., 1998. The active site of
pyrophosphate-dependent phosphofructo-1-kinase based on site-directed mutagenesis and molecular modeling. Arch. Biochem. Biophys.
349, 47 – 52.
Kemp, R.G., Gunasekera, D., 2002. Evolution of the allosteric ligand sites
of mammalian phosphofructo-1-kinase. Biochemistry 41, 9426 – 9430.
Lang, B.F., O’Kelly, C., Nerad, T., Gray, M.W., Burger, G., 2002. The
closest unicellular relatives of animals. Curr. Biol. 12, 1773 – 1778.
Lopez, C., Chevalier, N., Hannaert, V., Rigden, D.J., Michels, P.A., Ramirez, J.L., 2002. Leishmania donovani phosphofructokinase. Gene characterization, biochemical properties and structure-modeling studies.
Eur. J. Biochem. 269, 3978 – 3989.
Moore, S.A., Ronimus, R.S., Roberson, R.S., Morgan, H.W., 2002. The
structure of a pyrophosphate-dependent phosphofructokinase from the
Lyme disease spirochete Borrelia burgdorferi. Structure (Camb.) 10,
659 – 671.
Müller, M., Lee, J.A., Gordon, P., Gaasterland, T., Sensen, C.W., 2001.
191
Presence of prokaryotic and eukaryotic species in all subgroups of the
pp(i)-dependent group ii phosphofructokinase protein family. J. Bacteriol. 183, 6714 – 6716.
Philippe, H., 1993. MUST, a computer package of Management Utilities
for Sequences and Trees. Nucleic Acids Res. 21, 5264 – 5272.
Poorman, R.A., Randolph, A., Kemp, R.G., Heinrikson, R.L., 1984. Evolution of phosphofructokinase-gene duplication and creation of new
effector sites. Nature 309, 467 – 469.
Saitou, N., Nei, M., 1987. The neighbor-joining method: a new method for
reconstructing phylogenetic trees. Mol. Biol. Evol. 4, 406 – 425.
Schmidt, H., Strimmer, K., Vingron, M., von Haeseler, A., 2002. TREEPUZZLE: maximum likelihood phylogenetic analysis using quartets
and parallel computing. Bioinformatics 18, 502 – 504.
Shimodaira, H., 2002. An approximately unbiased test of phylogenetic tree
selection. Syst. Biol. 51, 492 – 508.
Shimodaira, H., Hasegawa, M., 2001. CONSEL: for assessing the confidence of phylogenetic tree selection. Bioinformatics 17, 1246 – 1247.
Siebers, B., Klenk, H.P., Hensel, R., 1998. PPi-dependent phosphofructokinase from Thermoproteus tenax, an archaeal descendant of an ancient
line in phosphofructokinase evolution. J. Bacteriol. 180, 2137 – 2143.
Thompson, J.D., Higgins, D.G., Gibson, T.J., 1994. CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through
sequence weighting, position-specific gap penalties and weight matrix
choice. Nucleic Acids Res. 22, 4673 – 4680.
Van de Peer, Y., Baldauf, S.L., Doolittle, W.F., Meyer, A., 2000. An updated and comprehensive rRNA phylogeny of (Crown) eukaryotes
based on rate-calibrated evolutionary distances. J. Mol. Evol. 51,
565 – 576.
Van Praag, E., 1997. Use of 3-D computer modelling and kinetic studies to
analyse grapefruit pyrophosphate-dependent phosphofructokinase. Int.
J. Biol. Macromol. 21, 307 – 317.
Verhees, C.H., Tuininga, J.E., Kengen, S.W., Stams, A.J., van der Oost, J.,
de Vos, W.M., 2001. ADP-dependent phosphofructokinases in mesophilic and thermophilic methanogenic archaea. J. Bacteriol. 183,
7145 – 7153.