Introduction

One of the main questions emerging from evolutionary studies is how to understand the proximal mechanisms that define the observed variations of life history traits in nature (Zera and Harshman 2001). Explaining why rates of senescence can be so different among animals has been one of the most fascinating question in the last few years (Kirkwood and Austad 2000; Mair et al. 2005). Several aging mechanisms have been described and, among them, oxidative stress resistance is likely to be one of the most important (Lithgow and Kirkwood 1996). Reactive oxygen species are formed during mitochondrial respiration (Beckman and Ames 1998) and mitochondrial electron transport has been demonstrated to be a key determinant for life span (Feng et al. 2001). Decreasing ROS production may be one solution that could extend lifespan (Brand 2000), and uncoupling mitochondrial respiration via uncoupling proteins (UCPs) may lead to such a result (Fridell et al. 2005).

UCPs are transporters present in the mitochondrial inner membrane. UCP1 is known to dissipate, as heat, the proton gradient generated by the mitochondrial respiratory chain (Ricquier and Bouillaud 2000). Several UCP1 homologues have been found, such as UCP2 and UCP3 (Fleury et al. 1997; Vidal-Puig et al. 1997), which share about 60% of their protein sequences with UCP1. Other proteins, named brain and kidney mitochondrial uncoupling proteins (BMCP and KMCP, respectively) (Haguenauer et al. 2005) and UCP4 (Mao et al. 1999) share only around 30% of amino acid sequence with UCP1 (Bouillaud et al. 2001), the lowest degree of similarity among all mitochondrial carriers, and are not considered to belong to the same subfamily as UCP1/2/3 (Ricquier and Bouillaud 2000). However, a complete phylogenetic analysis on UCPs has not been published to date.

Because UCP1 gene expression was first believed to be restricted to mammals, UCP1 was considered as the most recent UCP homologue, perhaps resulting from transposition of the closely located UCP2 and UCP3, which themselves are believed to stem from a duplication event (Pecqueur et al. 1999). Uncoupling activity was then believed to be a late evolutionary acquisition related to thermogenesis and endothermy (Klingenberg 1990). However, the discovery of UCPs in many different eukaryotes (for a review see Criscuolo et al. 2005a) and the characterization of UCP1 in fish (Jastroch et al. 2005) have recently challenged our view of the evolutionary history of the physiological uncoupling of oxidative phosphorylation, which seems to be a common strategy developed early in evolution. A key question is, therefore, which taxa actually possess the UCP1/2/3 and for what biological purposes?

Intriguingly, whereas the UCP1, UCP2 and UCP3 have been characterized in fish (Jastroch et al. 2005) and mammals (Bouillaud et al. 2001), only one uncoupling homologue (avian UCP) has been cloned in the avian class, which presents around 70% of sequence similarity with mammalian UCP2 and UCP3 (Raimbault et al. 2001; Vianna et al. 2001). Gene expression of avian UCP is principally found in skeletal muscle (Raimbault et al. 2001), but has also been detected in heart, liver, lung and kidney (Vianna et al. 2001). Avian UCP tissue distribution seems more ubiquitous, like mammalian UCP2, depending on the avian species (Evock-Clover et al. 2002; Dridi et al. 2005). In addition to its hypothetical thermogenic function, several recent studies have also highlighted the involvement of avian UCP in ROS metabolism in relation to its uncoupling activity (Talbot et al. 2004; Talbot et al. 2003, Abe 2006), which is an interesting hypothesis since birds have surprising low rates of senescence (Holmes and Austad 1995). However, when expressed in yeast, avian UCP protects the mitochondria from ROS damage without any detectable uncoupling activity, despite its capacity to behave as a true uncoupler when activated by retinoic acid (Criscuolo et al. 2005b). All of these characteristics make the avian model particularly interesting since we can hypothesize that avian UCP can fulfil the activity of UCP1 and UCP2/3. Therefore, based on an in silico analysis, we attempted to gather genetic evidence for the presence (or not) of UCP1 in the chicken genome, and we have tried to build a phylogenetic history of UCPs by describing more precisely the relationship between uncoupling protein sequences among different taxa.

Materials and Methods

Phylogenetic Inference and Comparative Genomics

We used two strategies to identify UCP1, UCP2 or UCP3 genes in the different taxa:

  1. 1.

    We conducted a similarity search using BLAST (Altschul et al. 1990) with the full-length length protein sequences of UCP1 of zebra fish (Danio rerio) (Ensembl Genome Browser, http://www.ensembl.org) against the genomes of the chicken (Gallus gallus), Xenopus (Xenopus tropicalis), a nematode (Caenorhabditis elegans) and three insects, the fruit fly (Drosophila melanogaster), the bee (Apis mellifera) and the mosquito (Anopheles gambiae).

  2. 2.

    As described previously (Jastroch et al. 2005), highly conserved genes surrounding UCP1 and UCP2/3 were sought in the genomes concerned, to define areas of interest where UCP paralogues may be found, but also to determine whether the sequences can be defined as UCP1-like or UCP2/3-like genes (http://www.ensembl.org). Therefore, using conserved neighbouring genes surrounding UCP1 or UCP2/3 in mammalian lineage, we defined regions of conserved synteny in genomes of other taxa and we looked at whether it enclosed a UCP-like gene.

Additionally, we determined the membership of the new sequences by comparing the aligned predicted protein sequences with the amino acid sequences of UCPs available in public database. The 75 protein sequences were aligned using MUSCLE with the default settings (Edgar 2004). Regions of uncertain amino acid alignment homology were removed using Gblocks 0.91b (Castresana 2000) with the following parameters: the minimum number of sequences for a conserved position and a flanking region was set to 38, the maximum number of non-conserved positions was set to 2, with half the sequences allowed as gaps. Phyml v2.4.4 (Guindon and Gascuel 2003) was used with the online web server (Guindon et al. 2005) for maximum likelihood (ML) analyses with 100 bootstraps, using the WAG substitution model selected with ModelGenerator (Keane et al. 2006). We accounted for the among-site rate variation using a gamma distribution and a proportion of invariant sites (pInvar). Bayesian analyses were also conducted using the latter model with MrBayes v3.1.2 (Huelsenbeck and Ronquist 2001). Nodal support was assessed as the posterior probability from two independent runs each with four chains of 1,000,000 generations in the Markov chain Monte Carlo procedure (the first 500,000 generations were discarded as “burn-in”). The tree was rooted with an out-group containing one plant Adenine Nucleotide Translocase (ANT) (see Figs. 2 and 3 legends for further details). Sequences of new UCP homologues were aligned (ClustalX) and illustrated with Genedoc 2.602 (Fig. 5) http://www.psc.edu/biomed/genedoc). All access numbers of sequences characterized in the present study are given in Table 1 and the aligned matrix is available as supplementary material.

Table 1 Listing of protein sequences of mitochondrial carriers characterized in this study. Accession numbers are given to allow access to the entire amino acid sequence on the website http://www.ensembl.org
Full size table

Results

Genomic Search for UCP1

Both highly conserved genes A (NM_178736) and B (NM_027758) of Mus musculus were used as neighbour genes of UCP1-like proteins in different genomes (Fig. 1A). For each one, peptides best matching known mammalian UCP1 sequences were found between these flanked genes. No UCP1 sequence was found in the genome of D. melanogaster or C. elegans. In A. mellifera and A. gambiae genomes, we found two genes bracketed by genes A or B (Fig. 1A) that presented amino acid identities of 53% (UCP1_Ame) and 48% (UCP1_Aga1) with fish UCP1. However, these insect UCP1-like proteins are not monophyletic with the mammalian UCP1 or even with the UCP2/3 (Fig. 2). In X. tropicalis, we characterized a region containing an UCP1-like protein exhibiting 80% similarity with fish UCP1 on scaffold 429 (Fig. 1A). In contrast, no UCP1-like protein was found in the chicken genome, and the region bounded by chicken genes A and B did not show predicted peptide sequences that look like an UCP or more generally like a mitochondrial carrier (Fig. 1A).

Fig. 1
figure 1

Schematic gene maps of the conserved syntenic regions of the uncoupling proteins 1 (UCP1), 2 (UCP2) and 3 (UCP3) genes (A) The protein identity of UCP1 neighbours was calculated as a percentage of the murine ortholog. Orthologs of gene A in Mus musculus (NM_178736) and Danio rerio (ENSDART00000026397). Orthologs of gene B in Mus musculus (NM_027758) and Danio rerio (ENSDART00000034884). Orthologs of murine genes A and B in Apis mellifera, Anophele gambiae, Xenopus tropicalis (ENSXETP00000032628; ENSXETP00000032669), and Gallus gallus (XP_417165.1; XP_420416.1) were obtained by genome blasting (http://www.ensembl.org). The cross indicates that there is no UCP1-like sequence between gene A and B in the chicken genome (B) The protein identity of UCP2 and UCP3 neighbors was calculated as a percentage of the human orthologs for UCP2 and UCP3. TSAR6 orthologs in Homo sapiens, Xenopus tropicalis, and Gallus gallus. Orthologs of human genes C (Q8BXE5) and TSAR6 in Xenopus tropicalis (ENSXETP00000027466; ENSXETP00000027442) and Gallus gallus (ENSGALP00000027466; ENSGALP00000027933) were obtained by genome blasting (http://www.ensembl.org)

Full size image
Fig. 2
figure 2

Bayesian tree of the uncoupling protein (UCP) core family based on 209 amino acids. Values at the branches are support values (Bayesian posterior probabilities above the branches and maximum-likelihood bootstrap support below). Information on the position of the UCP gene with respect to the neighboring genes are shown as boxes with either A and B representing genes A and B, or with 6 and C representing TASR6 and gene C, respectively (see Fig. 1)

Full size image

UCP2 and UCP3 in Different Taxa

The genomes were also used to search for the C (Q8NDH7) and TSAR6 genes, which define the region containing UCP2 and UCP3 in mammals and fish (Fig. 1B). Again, proteins showing the best matching scores were found in this region for all the species tested. We were unsuccessful in the characterization of UCP2 and UCP3 proteins in the insect and C. elegans genomes. In Xenopus, a UCP2/3-like protein (409 aa) was located on the scaffold 1234 which presents 38% similarity with UCP1, 50% with UCP2, and 49% with UCP3 of H. sapiens (Fig. 1B). This peptide, composed of the 100-amino acid tripartite structure that specifically characterizes mitochondrial carriers and UCPs, is ended by a 103 aa-long C-terminal (C-ter) sequence that does not contain specific protein motifs (protein kinase C or casein phosphorylation sites PS00005 / 6, N-myrisoylation site PS 00008). Once this C-ter sequence is deleted, the new UCP2/3-like protein shows 49% similarity with UCP1, 66% with UCP2 and 64% with UCP3. Instead of finding UCP2 and UCP3 between chicken C and TSAR6 genes located on chromosome 1, we only characterized the already known avian UCP gene (Fig. 1B), thereby confirming that birds have lost UCP1 and UCP2 or UCP3 already present in fish.

Phylogenetic Analysis

Tree topologies generated by Bayesian inference and maximum likelihood only differed in the position of the UCP4 genes. In the maximum-likelihood analysis, UCP4 was monophyletic with the other UCPs (bootstrap support of 72 not shown), whereas in the Bayesian inference UCP4 was monophyletic with BMCP (posterior probability of 0.72). Both methods strongly support the monophyly of BMCP, UCP4, vertebrate UCP2, mammalian UCP3 and avian UCP. The monophyly of vertebrate UCP1 is only supported by the Bayesian posterior probabilities, although there is strong support for the monophyly of mammalian UCP1, and only the posterior probability supports UCP1 as sister clade to UCP2. The monophyly of UCP1, UCP2 and UCP3 of all metazoan is strongly supported. Within each monophyletic group, the phylogenetic relationships are broadly in agreement with the widely accepted phylogeny of vertebrates although both the monophyletic UCP1 and UCP2 lack bird representative genes. Additionally, the phylogenetic analysis suggests that UCP2 from Pagrus major is more closely related to UCP1 from Danio rerio and to other UCP1 than to UCP2 genes (Figs. 2 and 3). Because the whole genome of Pagrus major is not available, it was not possible to determine whether UCP2 Pagrus major is located in the vicinity of genes A and B, like other UCP1. However, we suggest that this sequence has been misannotated.

Fig. 3
figure 3

Amino acid sequence alignments of uncoupling proteins (UCPs). Sequence from Danio rerio (Dre), Pagrus major (Pma), Xenopus tropicalis (Xtr), Homo sapiens (Hsa), Apis mellifera (Ame), and Anopheles gambiae (Aga) were aligned using ClustalX 1.8 (ftp://www.ftp-igbmx.u-strasbg.fr/pub/ClustalX) and represented with Genedoc 2.5 (http://www.psc.edu/biomed/genedoc). The darkness of the background is representative of the amino acid conservation. The black bar above the alignment indicates the mitochondrial carrier protein signature

Full size image

Characterization of BMCPs and UCP4 Homologues

We next investigated, in the different taxa, the presence of mitochondrial carriers such as BMCPs and UCP4 suspected to be related to UCPs. Using the known UCP sequences, we discovered several peptides containing the mitochondrial carrier signature [PROSITE PS00215, PX(D/E)X(LIVAT)(K/R)X(LRH)(LIVMFY)] in A. gambiae and D. melanogaster. These proteins are more related to BMCP (BMCP-like_Aga) and to UCP4 (UCP4-like Aga, UCP4-like DmeA and B) (Fig. 2). BMCP- and UCP4-like genes have also been found in X. tropicalis genome (Fig. 2). Also, two additional avian BMCP-like proteins were located on chromosomes 4 (BMCP1_Gga) and 1 (KMCP_Gga) (Fig. 4). Human BMCP1 (O95258) has an amino acid sequence identity of 78% with BMCP1_Gga, but only 19% with KMCP_Gga. When looking at the alignment with other UCPs and BMCPs (Fig. 4), only the third tripartite section of KMCP_Gga, which contained the specific signature of mitochondrial carrier (PROSITE PS00215), was aligned. Interestingly, KMCP_Gga starts with a non-specific sequence of 250 aa, followed by 80 non-aligned amino acids that contain an additional UCP signature (PIDLTKTR) (Fig. 4). By deleting the first nonaligned 250 aa segment that does not contain any specific protein motif (PS00001 / 2 / 4 / 5 / 6 / 8 / 9), we obtained a new alignment confirming the presence of the thirds numbered 1 and 3, which are highly homologous to human KMCP1 (88% and 66%, respectively) and BMCP1 (79% and 58%, respectively), whereas this value was reduced to 25–44% when compared to other UCPs (Fig. 5). Phylogenic inferences showed that KMCP_Gga may be a truncated KMCP1 homologue that has lost its second third and gained an initial 250 aa-long undefined sequence.

Fig. 4
figure 4

Comparison of kidney mitochondrial uncoupling protein(KMCP)_Gga amino acid sequence with uncoupling proteins (UCPs) and brain mitochondrial uncoupling proteins (BMCPs). Sequences from human UCP1, UCP2, UCP3, KMCP1, BMCP1, avian UCP and KMCP_Gga were aligned using ClustalX 1.8 (ftp://www.ftp-igbmx.u-strasbg.fr/pub/ClustalX) and represented with Genedoc 2.5 (http://www.psc.edu/biomed/genedoc). The darkness of the background is representative of the amino acid conservation. The black boxes indicate the common mitochondrial carrier protein signature. The gray box indicates a nonaligned mitochondrial carrier protein signature of KMCP_Gga

Full size image
Fig. 5
figure 5

Identification of kidney mitochondrial uncoupling protein(KMCP)_Gga as a human KMCP1 ortholog. Sequences from human uncoupling proteins (UCPs) 1, 2 and 3, KMCP1, brain mitochondrial uncoupling protein 1 (BMCP1), avian UCP and the truncated KMCP_Gga were aligned using ClustalX 1.8 (ftp://www.ftp-igbmx.u-strasbg.fr/pub/ClustalX) and represented with Genedoc 2.5 (http://www.psc.edu/biomed/genedoc). The darkness of the background is representative of the amino acid conservation. The black boxes indicate the common mitochondrial carrier protein signature. The mitochondrial signature that was nonaligned in Fig. 4 (gray box) is now aligned with the first third of UCPs and BMCPs. Inset: schematic representation of the tripartite structure of UCP protein (A). The alignment with KCMP_Gga (B) reveals the presence of third third. When KCMP_Gga was cleared of the first 250 amino acids (C), we obtained KMCP Fig. 5 (D), which, after realignment, also reveals the presence of the first third (E)

Full size image

Two additional peptides (PepI_Gga and PepL_Gga) were found by matching chicken genome with either avian UCP, or mammalian and fish UCPs. These peptides do not group with other UCPs and thus should be rather considered as undefined mitochondrial carriers (Fig. 2).

Discussion

A Phylogenic View of the UCP Family

The idea that UCP1 is restricted to the thermogenic brown adipose tissue of mammals and derived from UCP2 or UCP3 has recently been turned upside down (Jastroch et al. 2005). The authors found that fish, like the common carp (Cyprinus carpio), possess the three uncoupling proteins UCP1, UCP2 and UCP3. We extended the localization of UCP1-like proteins to insect and to amphibian classes as shown by synteny (Fig. 1) and the phylogeny (Fig. 2), confirming that UCP1 probably appeared early in evolution.

Firstly, although, the maximum-likelihood analysis supports the monophyly of UCP4 with the other UCPs, the Bayesian topology (Fig. 2) supports the view that UCP4 is more closely related to BMCP genes, as previously suggested (Ricquier and Bouillaud 2000). Secondly, the fact that UCP1, UCP4 and BMCP are found in Protostomia and Deuterostomia but not UCP2 and UCP3 provides support for the emergence of UCP2 and UCP3 after the Protostomia–Deuterostomia split, i.e. around 500–550 million years ago. Thirdly, UCP2 and UCP1 seem to be more closely related to each other than to UCP3, in contrast to generally suggested hypotheses (Pecqueur et al. 1999).

Can Insect and Fish UCP1 Proteins be Comparable to Mammalian UCP1? Phylogenetic and Biochemical Evidence

Should we, however, consider that all UCP1s exhibit a comparable mitochondrial function (i.e., an uncoupling effect inducing thermogenesis) to the mammalian UCP1? The monophyly of UCP1/2/3 for all metazoan, added to the fact that the insect UCP1 genes are neighboring the conserved genes A (in Apis mellifera) and B (in Anopheles gambiae) in a similar manner to other vertebrate UCP1 genes being syntenic to gene A and gene B, strongly suggests that UCP1s in the class Insecta are orthologs to the vertebrate UCP1s. Therefore, it appears that the thermogenic function for these insect, fish and amphibian UCP1s, based on their chromosomal localization (Fig. 1) (Jastroch et al. 2005) and on the phylogeny (Fig. 2), could be supported.

However, biochemical data are lacking to support the notion that nonshivering thermogenesis (heat production due to biochemical waste of energy) has evolved concomitantly with the appearance of the UCP1 gene in poikilotherm animals. It has not been proven that uncoupling activity is involved in the nonshivering thermogenesis observed in fish and marsupial mammals (Block 1994; Rose et al. 1999). In addition to scant biochemical, quantitative and histological evidence for a physiological uncoupling activity of fish UCP1 (Jastroch et al. 2005), one can question the significance of the presence of UCP1-like proteins in insects and amphibians, in which thermogenesis via mitochondrial uncoupling is unexpected. One model of respiration uncoupling by UCP1 utilizes fatty acid cycling (Garlid et al. 2000). In this model, protonated fatty acid diffuses through the inner membrane and release a proton in the matrix. Then UCP translocates the anion form of fatty acid through the membrane. So one hypothesis explaining the early appearance of UCP1 is that it first evolved a mitochondrial transport mechanism, different from respiration uncoupling, combined to proton import into the mitochondrial matrix, like fatty acid import for metabolic purposes, or others. Indeed, UCP1 can also ensure chlorine transport (Klingenberg 1990) and has been involved in the control of pyruvate flux through the mitochondrial membrane (Jezek and borecky 1998, see also Mozo et al. 2006).

UCP2 and UCP3

To date, UCP1 was thought to derive from UCP2 or UCP3 (Cannon and Nedergaard 2004). However, the presence of a UCP1-like protein” and the absence of UCP2 and UCP3 in the insect class support the idea that UCP1 appeared first, from the ANT and BMCP-UCP4 lineage, as previously suggested (Hanàk and Jesek 2001; Aquila et al. 1985) (Fig. 4), and through two rounds of genome duplications in early vertebrates evolution that gave rise to UCP3 and UCP2 in vertebrates. The increase in complexity during the evolution of vertebrate lineage was accompanied by duplication of genes or genomes early in the vertebrate evolution (Hokamp et al. 2003), which can explain the increased number of UCP paralogues in mammals compared to insects (Fig. 2). The relationships of the monophyletic avian UCP with UCP3 from amphibians and the fish are unresolved. Moreover, the lack of phylogenetic resolution for the relationships between UCP3 and UCP1 or UCP2, as well as the lack of monophyly for the UCP3 genes, hinders unambiguous conclusions regarding the evolution of UCP genes. Further taxonomic sampling might help to resolve these relationships in the future. However, the lack of avian UCP2 and avian UCP1 within the monophyletic UCP2 and UCP1 clades, which include both representatives from the fish and the amphibian, clearly suggests that birds have lost both UCP2 and UCP1 after the two genome duplications during early vertebrate evolution. In this context, our hypothesis is that the remaining avian UCP picked up new functions (neofunctionalization) (Sidow 1996) that may compensate for the loss of other UCP homologues.

Avian UCP: A Multifunctional Protein?

The literature reports a variety of avian UCP activities: UCP1-like uncoupling activity leading to thermogenesis or decreases in mitochondrial ROS production (Talbot et al. 2004), a preventive effect against free radical damage, without any detectable uncoupling activity (Criscuolo et al. 2005a), or control of lipid metabolism (Collin et al. 2003).

Three main hypotheses can be proposed to explain this heterogeneity. Firstly, as for mammals, the avian genome may contain several uncoupling homologues, each devoted to one specific function. However, we failed to characterize any UCP1 or UCP2 proteins in birds (Fig. 1A).

Secondly, thermogenic, antioxidant and metabolic functions may also be ensured by other mitochondrial transporters, for example, in the cold-adapted king penguin (Aptenodytes patagonicus), at least one part of the induced uncoupling activity is due to the avian ANT (Talbot et al. 2004). It is known that ANT can transport protons and uncouple mitochondrial respiration when exposed to high levels of free fatty acids (FFA) (Andreyev et al. 1989), and glucagon-induced cold acclimation in birds occurs concomitantly with an increase in FFA availability (Benistant et al. 1998).

Thermogenesis (Enerback et al. 1997), protection against free-radical production (Criscuolo et al. 2005b), signalling (Emre et al. 2007), and fatty acid and pyruvate transport (Garlid et al. 2000, Mozo et al. 2006) are all described functions of UCPs. However there is no consensus that all UCPs ensure the same function, or even that they are regulated in the same manner (Echtay et al. 1999). Nevertheless, one can imagine that the particular role of avian UCP is to combine all or some of these functions, for example, it has been successfully proposed that avian UCP has an uncoupling activity potentially related to heat production in penguins (Talbot et al. 2004) or to a decrease in free-radical production in chicken (Criscuolo et al. 2005, Abe et al. 2006). Additionally, avian UCP was associated with the regulation of lipid utilisation (Collin et al. 2003) and we have recently reported an K+/H+ activity for avian UCP (Criscuolo et al. 2006). Such an antiport activity can be linked to the ability of avian UCP to protect yeast mitochondria against oxidative stress, since in addition to its effect on mitochondrial volume homeostasis and respiration (Garlid and Paucek 2003; Nicholls et al. 1972), the ATP-sensitive K+ channel modifies mitochondrial free radical production (Garlid et al. 2003).

In conclusion, avian UCP activities are probably multiple, and the uncoupling and potentially thermogenic effects are probably neither its main or initial role. In the future, cloning and expression in artificial systems devoted to the fine study of the transport activities of avian UCP may be of prime importance in resolving this controversial question.