The L-LDH alignment used in phylogenetic analyses includes 315 amino acids, with 68 constant characters and 214 parsimony-informative characters. Maximum-parsimony (MP) analysis of 49 L-LDH sequences using the tree-bisection-reconnection (TBR) algorithm found two most parsimonious trees of 1982 steps long with a consistency index (CI) = 0.48, a homoplasy index (HI) = 0.51 and a retention index (RI) = 0.63. The uncorrected number of amino acid differences per site between invertebrate and chordate groups is 0.38 ± 0.019.

Phylogenetic trees generated by MP and Neighbor-joining (NJ, Figure 1) and Bayesian Inference (BI, Figure 2) all support a deuterostome cluster. The only exception is the echinoderm, Strongylocentrotus purpuratus, which groups with nematodes, although bootstrap support for this phylogenetic relationship is extremely low (Figure 1). The vertebrate sequences form a well-supported cluster with the LDH-A and LDH-B isozymes separating into distinct groups. Even so, there are examples of species whose A, B and C isozymes cluster with one another (e.g. Xenopus laevis).

The arthropod LDH sequences form a well-resolved cluster, as do the insects within it. Two copies of Ldh were found in the Daphnia pulex genome, and three copies in the Helobdella robusta genome. As with other cases of gene duplication outside chordates, the suffixes A, B and C do not denote orthology to the vertebrate A, B and C isozymes. The predicted protein sequences from the two Daphnia paralogs show 0.17 ± 0.012 amino acid divergence (p-distance) and 0.224 ± 0.014 nucleotide divergence (p-distance) in the coding regions. They cluster with one another indicating that this gene duplication occurred after the divergence of Daphnia from the other crustaceans, the decapods Carcinus maenas and Petrolisthes cinctipes (Figure 1 and 2). The insects and crustaceans are reciprocally monophyletic in the MP and NJ trees (Figure 1), although bootstrap support for the crustacean node is very low. Conversely, very strong support for an arthropod clade in which the crustaceans are paraphyletic relative to the insects was obtained in the BI tree (Figure 2).

The level of amino acid divergence between the three Helobdella robusta LDH proteins (0.32 ± 0.026, 0.46 ± 0.028, 0.52 ± 0.028) is higher than that between the Daphnia copies or Fundulus heteroclitus (LDH-B and LDH-C, 0.21 ± 0.028) and much higher than that between the recently diverged Xenopus laevis copies (0.019 ± 0.008, 0.049 ± 0.013, 0.063 ± 0.014). Moreover, the relationships among the three genes differs among the phylogenetic trees. Only two of the three H. robusta genes (Ldh-A and Ldh-B) are clearly paralogous in the MP/NJ tree, and duplicated after the divergence of H. robusta from the other annelid in the analysis (the polychaete, Capitella capitata). However, the annelid cluster also contains the trematode flatworm, Clonorchis sinensis, although bootstrap support for these relationships is low (Figure 1). The annelid/trematode clade is also recovered in the BI tree (Figure 2), although in this case, C. capitata is the sister group to the three leech and the flatworm sequences.

The D-LDH alignment included 486 amino acids with 73 conserved sites and 319 parsimony informative sites. The uncorrected number of amino acid differences per site averaged over all sequence pairs between chordates and invertebrates is 0.41 ± 0.015.

In general, the topology of the NJ tree generated from these sequences (Figure 3) shows that the deuterostomes form a distinct cluster relative to the other animals, as expected. For example, there is strong bootstrap support for a vertebrate cluster, and the non-vertebrate deuterostomes (C. intestinalis and S. purpuratus) cluster with them, although unexpectedly, C. intestinalis (a tunicate) clusters with S. purpuratus (an echinoderm) instead of the vertebrates. Relationships among the other invertebrates are not well resolved. For example, the annelids, H. robusta and C. capitata, do not cluster with each other, and the protostomes themselves do not form a monophyletic group (H. robusta is the sister taxon to all the other animals except C. elegans), but bootstrap support for several of the invertebrate nodes is low. Overall, this tree strongly suggests that D-Ldh was present in the common ancestor of animals.

The majority of taxa of jawed vertebrates contain three isozymes of L-LDH (LDH-A, LDH-B and LDH-C) encoded by three loci. The M form (LDH-A) is found predominantly in white skeletal muscle (fast twitch glycolytic fibers) and is best suited for pyruvate reduction in anaerobic conditions, while the H form (LDH-B) is found in more aerobic tissues such as heart and brain and is most efficient for lactate oxidation. The X form (LDH-C) is found in various tissues such as the spermatozoa of mammals and birds, the eye lenses of birds and crocodilian and the liver and eye of teleosts 89.

It is commonly accepted that new metabolic capacities of L-LDH enzymes have often arisen by gene duplications in addition to more orthodox evolutionary changes in existing genes. For this reason the Ldh gene family has been used as a model for gene duplication in vertebrate evolution 810. It is generally accepted that the Ldh genes of jawed vertebrates arose as a series of gene duplications in early vertebrate evolution, after the divergence of vertebrates from tunicates. However, the succession of these gene duplication events is not well understood 91112. Two main evolutionary scenarios have been proposed. The classical scenario involves the duplication of an Ldh-A-like locus in Agnatha (lampreys have a single LDH form) that gave rise to Ldh-A and Ldh-B. A second round of gene duplication involved Ldh-B and gave rise to Ldh-B and Ldh-C. Several phylogenetic studies support this hypothesis of an original Ldh-A and Ldh-B gene duplication followed by more recent and independent origins of the Ldh-C genes in teleost fish, Xenopus laevis, pigeon and mammals 71013. A second school of thought suggests that the primordial vertebrate LDH was an LDH-C like enzyme. This hypothesis emerged from phylogenetic reconstructions and from the observation that the single LDH isozyme of the primitive agnathan, the sea lamprey, is immunologically more similar to LDH-C in teleost fish than to LDH-A or LDH-B 141516.

Our phylogenetic analyses of L-LDH isozymes are consistent with the general view of a major gene duplication event near the origin of vertebrates. Both NJ and MP phylogenies suggest that the vertebrate isozymes LDH-A and LDH-B evolved through a series of gene duplications soon after the vertebrates diverged from tunicates. The main difference between the MP, NJ and BI reconstructions is that MP and BI supports a sister group relationship between LDH in Petromyson marinus, the only chordate with a single LDH locus, and all of the other vertebrate LDHs. Conversely, this sequence clusters with fish LDH-A in the NJ tree. Although the controversial phylogenetic position of the LDH-C isozymes cannot be easily settled, it is clear from our results that Ldh-C in mammals, X. laevis and Fundulus have independent, derived origins from either Ldh-A or Ldh-B-like ancestors.

The relationship of invertebrate and protochordate Ldh to vertebrate Ldh is not well understood. Although most invertebrates seem to possess one copy of L-Ldh 171819, several events of gene duplication involving nonvertebrates have been reported in crustaceans, barley 20 and in the psychrophilic bacterium, Bacillus psychrosaccharolyticus 21. In studies of crustacean LDH, protein electrophoresis has detected the presence of two Ldh loci in the northern krill Meganyctiphanes norvegica, the Antarctic krill Euphausia superba 22, the lobster Homarus americanus 23, the snow crab Chionoecetes opilio 24, the amphipod Hyallela azteca (J. Witt personal communication), the cladoceran Daphnia magna 25 and possibly Daphnia carinata 26 and Daphnia cephalata 27. All of these crustaceans except Daphnia belong to the class, Malacostraca. Further work will be required to determine if a gene duplication occurred early in this crustacean lineage (as occurred in vertebrates), or if these duplicated genes have independent origins within the class. The two copies of L-Ldh in D. pulex group with one another in our phylogenetic analyses, suggesting that they do not represent an ancient crustacean duplication, but analysis of other branchiopods will be required to determine when this duplication occurred. Moreover, the MP and NJ trees suggest that insects and crustaceans are reciprocally monophyletic, but the crustaceans are paraphyletic relative to the insects in the BI tree, with strong support. The possibility of crustacean paraphyly has been suggested in other phylogenetic studies involving a variety of markers 282930, but these relationships have not been definitively resolved.

The only other invertebrate in our analysis with multiple copies of L-Ldh is the leech, H. robusta, with three copies. Two of the copies appear to be unique to this taxon, but our sample of sequences is not sufficient to determine when this duplication occurred. A very long branch connects Ldh-C in H. robusta with the flatworm instead of the other annelids in the NJ tree, while Ldh-B clusters with the flatworm in the BI tree. Additional annelid and flatworm taxa must be analyzed to "break up" these branches, and to examine the relationship between these two phyla. What is clear from our phylogenetic analyses is that the LDH genes of Arthropods are significantly distinct from the LDH genes of other invertebrates, including annelid and nematode worms.

LDH allozyme polymorphism has been intensely studied in Daphnia including species with strong habitat specificity. For example,D. pulex inhabits mainly freshwater ponds throughout North America and Europe that lack fish, while its closest relative, Daphnia pulicaria, inhabits lakes and is able to coexist with fish, which are efficient predators of these limnetic zooplankters. The two ecological species can be distinguished based on a diagnostic LDH allozyme polymorphism with a "slow" (S) allele in D. pulex and a "fast" (F) allele in D. pulicaria 31. However, the situation is complicated by hybridization and transitions from cyclical to obligate parthenogenesis 3233. Although there are distinct mitochondrial lineages in the two species in North America 34, there are many lake populations of Daphnia that have a D. pulicaria LDH profile (F), but a D. pulex-like mitochondrial DNA (T. Crease, unpublished data).

The maintenance of the LDH-F allele in lake populations of Daphnia, regardless of maternal origin, suggests that LDH genotypes differ in physiological performance, which may affect fitness. This is the case in the teleost, Fundulus heteroclitus 35, where the frequency of two LDH-B alleles shows clinal variation with latitude and environmental temperature. In addition, the fish allozymes show differences in kinetics related to temperature suggesting selection has favored a particular form of the LDH protein. Further studies on Ldh genes from populations of both the lake and pond Daphnia species are necessary to determine which of the two L-Ldh loci is responsible for the S/F polymorphism, and to provide evidence for the presence or absence of selection on the polymorphic locus. This work will be complemented by the examination of the enzyme kinetics of the allozymes themselves.

Enzymes that oxidize D-lactic acid have been mainly identified in lower organisms such as prokaryotes and fungi (e.g., Lactobacillus, Escherichia coli, yeast) in which they play an important role in anaerobic energy metabolism. While most invertebrates preferentially oxidize L-lactic acid, several species of mollusks (the oyster Crassostrea virginica, the mussel Mytilus edulis, the limpet Acmaea unicolor, the chiton Middendorffia caprearum, the octopus Eledone cirrosa), a few arthropods (horseshoe crab, spiders, scorpions) and one polychaete, Nereis sp., were found to have exclusively D-LDH enzymatic activity 3637. Therefore, it had been suggested that the L-LDH and D-LDH enzymes are mutually exclusive. However, Flick and Konieczny 5 identified and characterized a putative mammalian D-LDH enzyme that shows substantial similarity to three yeast proteins with D-LDH activity. This suggests that mammals have the two naturally occurring forms (D- and L-) of LDH. Moreover, we identified a D-Ldh gene in the genomes of several taxa that also possess L-Ldh including D. pulex, several non-mammalian vertebrates, the urchin, S. purpuratus, and C. elegans, suggesting that the possession of both forms of this enzyme is phylogenetically widespread. Further research will be required to determine the phylogenetic distribution of D-Ldh in animals, to understand why it has been retained in some groups but not others, and to determine the different roles that L-LDH and D-LDH play in such diverse animals. Comparisons of taxa with both types of enzymes to close relatives with only one type would be informative in this regard. In addition, it will be interesting to determine whether the "missing" enzyme in such cases is the result of gene inactivation, degradation into a pseudogene, deletion, or is expressed at such low levels that the enzyme is not detected in typical assays.

LDH sequences from vertebrates, invertebrates and plants were obtained from the GenBank and Swiss-Prot/EMBL data bases (Table 1) using a combination of queries based on the term "Lactate dehydrogenase" and BLAST searches. In the latter case, several well described vertebrate and invertebrate sequences were used as the query. Blast searches were also performed on invertebrate genome projects from which a few unannotated sequences were retrieved. The deduced amino acid sequence for the porcelain crab, Petrolisthes cinctipes was obtained from a cloned EST that contained the full length cDNA sequence 38. The alignment of amino acid sequences was conducted using CLUSTAL W 39 with a gap opening penalty of 10 and a gap length penalty of 0.1. Minor adjustments to the alignments were made manually. To avoid ambiguity due to extensive sequence variability and length variability in the amino terminal arm, alignment positions 1–30 were removed from the L-LDH alignment in all analyses. Without this ambiguous region, the alignment of the ingroup taxa included six gaps, while the addition of plant sequences to the alignment resulted in four additional gaps. Phylogenetic analyses were inferred in MEGA4 40, PAUP, version 4.0 41 and MrBayes V3.1.2. 42 using three analytical approaches: Neighbor-joining (NJ), Maximum parsimony (MP) and Bayesian inference (BI). NJ trees were constructed from pairwise amino acid distances estimated using a Poisson correction. MP trees were estimated using a heuristic search algorithm with 100 replicates, sequences added at random and tree bisection-reconnection branch swapping. Amino acids were treated as unordered characters with equal weight and gaps were treated as "missing". The stability of both phylogenetic hypotheses was assessed with bootstrap analyses (100 replicates for MP and 1000 replicates for NJ). Analysis of the amino acid data was also conducted using the Bayesian inference (BI) method with a fixed rate model of amino acid substitution. The fixed rate model WAG was estimated by allowing "model jumping" between nine fixed-rate amino acid models. Runs of 1,000,000 generations were executed, with a sampling frequency of 10, a burn-in parameter of 25,000. Stability of the likelihood scores was assessed in preliminary trials before setting the burn-in parameter. To confirm that the results converged to the same topology, we repeated the analysis three times.