The open reading frame of the full-length cDNA (GenBank: EU019537) comprised 1653 nucleotides; and the 5' UTR and 3' UTR comprised 28 nucleotides and 99 nucleotides, respectively. This was obtained by using primers based on a 421 bp opossum cDNA (GenBank: DR038241) showing sequence similarity to mammalian CES to amplify a full-length opossum CES cDNA from liver RNA. The best-scoring alignment in a BLAST analysis 3637 for the opossum cDNA sequence was with a human CES2 sequence, and the opossum cDNA is referred to as CES2.1 (gene designation: CES2.1). Using the opossum CES2 cDNA as the query in a BLAT search 3839, we found two additional CES gene regions and cDNAs which are referred to as CES2.2 (gene designation: CES2.2) and CES2.3 (gene designation: CES2.3) (Table 1). These latter genes were subjected to further analyses using RT-PCR and sequencing the product CES cDNAs.

Expression of the predicted CES genes in the liver and small intestine was analyzed by RT-PCR (Figure 1). Controls were also conducted to ensure that DNase-treated liver and intestine RNAs could not serve as templates for RT- PCR (See Additional File 1).

An opossum CES1 gene was identified by in silico methods using the human CES1 amino acid sequence to interrogate the opossum genome 40 in a BLAT analysis 373839. Primers based on the predicted opossum CES1 cDNA sequence amplified a 750 bp PCR product which was similar to the predicted PCR product (741 bp), indicating that the CES1 gene is expressed in liver and intestine.

Unlike the CES1 gene, the CES2.1 gene was expressed in the liver, but not in the small intestine. The CES2.2 and CES2.3 genes were also identified by in silico methods. Primers for the CES2.2 cDNA amplified an 800 bp PCR product, which was similar to the predicted PCR product (803 bp), indicating that the CES2.2 gene is expressed in liver and intestine. However, primers for CES2.3 cDNA amplified a PCR product of about 770 bp, which is larger than the 667 bp PCR product based on the predicted cDNA sequence. The CES2.3 cDNA was sequenced (GenBank: EU019539) and this showed that 99 nucleotides were missing from the predicted BLAT derived sequence. The PCR product for the CES2.3 gene should have been 766 bp. These results indicate that the CES2.3 gene is expressed in both liver and small intestine and that the BLAT derived sequence contained a deletion as compared with the actual sequence for the CES2.3 RT-PCR product. We also sequenced the CES2.2 cDNA (Accession No. EU019538), and found that 33 nucleotides in the predicted sequence were absent in the CES2.2 cDNA sequence. These RT-PCR studies of opossum CES2.2 and CES2.3 genes show that the BLAT software 3839 used to predict sequences for opossum CES gene products did not fully recognize all of the intron-exon junctions resulting in some deletions or insertions that are not found in the in vivo gene product. This emphasizes the importance of undertaking cDNA sequencing to obtain sequences for these and other cDNAs.

The deduced amino acid sequence for opossum CES2.1 is shown in Figure 2 with the previously reported sequence for human CES2 1819, together with the predicted protein sequences for opossum CES2.2 and CES2.3. The latter sequences were predicted from BLAT studies 3839 using the opossum CES2.1 sequence to interrogate the opossum genome but corrected following alignment with the CES2.2 (GenBank: EU019538) and CES2.3 (GenBank: EU019539) sequences. The three opossum CES2 like sequences were very similar, especially for the CES2.2 and CES2.3 sequences, which shared 94% sequence identity. In addition, the deduced amino acid sequences for opossum CES2.1, CES2.2 and CES2.3 showed 56–57% sequence identity with human CES2 and 44–47% identity with human CES1 (Figure 2; Table 2).

Several key residues in these enzymes were conserved, including (sequence numbers refer to opossum CES2) Cys94/Cys121 and Cys278/Cys289, the sites for disulfide bond formation in human CES1 315293041, and active site residues responsible for the formation of acyl-ester intermediates (Ser226) and the charge relay system in catalysis (Glu343) and His (452) 3. The deduced amino acid sequence for opossum CES2.1 was 9 residues shorter (550 residues) than for human CES2 (559 residues), and lacked the microsomal targeting sequence at the C-terminus end of the enzyme. The human CES2 C-terminal tetrapeptide sequence, His-Thr-Glu-Leu (HTEL), functions in protein retrieval from the Golgi apparatus and in CES retention in the ER lumen 12, but this is missing in opossum CES2.1 which may then influence the subcellular location and metabolic role for this enzyme in opossum liver. In contrast, opossum CES2.2 and CES2.3 sequences retain a homologous Arg-Met-Glu-Leu (RMEL) tetrapeptide sequence and these enzymes may then be localized within the liver endoplasmic reticulum. Opossum CES2.1, CES2.2 and CES2.3 sequences however share a homologous hydrophobic N-terminus signal peptide (residues 1–25, 1–24 and 1–24 respectively) with human CES2 (residues 1–26) 12.

The N-glycosylation site reported in human CES1 42 was also found in opossum CES2.1 (86Asn-Ala-Thr) with four additional potential N-glycosylation sites observed at 274Asn-Leu-Ser, 293Asn-Lys-Thr, 376Asn-Ile-Ser and 425Asn-Ser-Ser. In contrast, opossum CES2.2 and CES2.3 lacked the N-glycosylation site at the corresponding position (85Asp-X-Thr), as did human CES2 (87Asp-X-Thr), but retained another potential site (273Asn-Leu-Ser) in common with opossum CES2.1. The two charge clamps reported for human CES1 43 which contribute to the formation of trimers and hexamers for this enzyme are retained for one site on each of the opossum CES2 like gene products (85Arg predicted charge clamp188Glu) but not for the second site (86Asp...191Pro) in each case (Figure 2). The impact on the tertiary and potential quaternary structures for these opossum CES2 like enzymes remains to be determined, however it is relevant to note that human and baboon CES2 are monomers, presumably as a result of the lack of both charge clamps and the N-glycosylation site for sialic acid attachment, which contribute to subunit-subunit binding for human CES1 43.

Other key functional residues for human CES1 and CES2 have been retained by the opossum CES2 like enzymes, including the 'Z-site' (Gly356 for human CES1), which functions in cholesterol binding; the 'side door' residues at 424Val.Met425.Phe426; and the 'gate' residue 551Phe (for human CES1), both of which have either been retained or have undergone conservative substitution. These latter sequences apparently function in facilitating the release of fatty acyl or aromatic groups, respectively, following hydrolysis 3152930. In addition to the extensive sequence similarities observed for the human CES2 and opossum CES2Like products, there are strong similarities in predicted secondary structures for these enzymes suggesting that the αβ-hydrolase structure previously reported for human CES1 3152930 has been predominantly retained by these enzymes (see Figure 2). There were some differences observed however for predicted secondary structures for the opossum CES2 like enzymes near key residues and these may influence enzyme function. For example, predicted helical secondary structures near the 'Z-site' and 'side door' were of different lengths for these enzymes which may influence ligand binding and product release, respectively.

The predicted amino acid sequence for opossum CES1 is shown in Figure 2 together with previously reported sequences for human CES1 and other opossum and human CES sequences 1617181920212223242526. The predicted opossum CES1 amino acid sequence (derived from N-Scan ID 1.45.030 following BLAT analysis 3839 of the opossum genome using the human CES1 sequence) was longer (670 residues) than human CES1 (567 residues) with an additional 111 residues at the N-terminus end which did not align with the human CES1 sequence. Given the similarity of the predicted sequence for opossum CES1 with human CES1 at the N-terminus and our observations of incorrectly predicted splice sites, it is likely that opossum CES1 shares the same N-terminus start point with human CES1 and it is this sequence that is included in Figure 2.

Sequencing the RT-PCR product using primers derived from the predicted opossum CES1 sequence enabled confirmation of 750 bp of this sequence (GeneBank:EU07430).

The predicted amino acid sequence for opossum CES1 showed 63% identity with human CES1 and 46% identity with human CES2 supporting its designation within the CES1 family (Table 2). Opossum CES1 also shared several key residues with human CES1, including the active site 'triad', Ser221, Glu353 and His468 (residue numbers refer to the opossum CES1 sequence); the corresponding cysteine residues forming the disulfide bonds in human CES1 (Cys87/Cys116 and Cys273/Cys284); the microsomal C-terminus retention sequence His-Ile-Glu-Leu (HIEL); and the high-mannose N-linked glycosylation site at Asn190-X-Thr. Two other potential glycosylation sites (257Asn-Ser-Ser and 528Asn-Ile-Thr) were also observed for the opossum CES1 sequence (Fig. 3). The N-terminal microsomal signal peptide for human CES1, which retains the enzyme within the ER 12 was identical in sequence with the predicted opossum CES1 sequence, and both sequences retain 18 homologous residues in corresponding positions.

Subunit-subunit charge clamps reported for human CES1 43 are retained for the predicted opossum CES1 protein: 72Glu predicted charge clamp186Arg; and 78Lys predicted charge clamp183Glu, suggesting that this enzyme may share the trimer-hexamer quaternary structure. The 'Z-site' glycine residue, which is a site for binding cholesterol analogues in human CES1, has been substituted in the opossum CES1 sequence (355Ala) which may reflect a change in the ligand binding properties for this enzyme. The 'side door' and 'gate' sequences for human CES1 (424Val.425Met.426Phe and 551Phe, respectively), which function in facilitating the release of fatty acid and aromatic products respectively following hydrolysis of substrate 3152930, have also undergone substitutions for these key sites, involving two of three 'side door' residues (422Leu.423Ile.424Phe) and the 'gate' residue (549Met) for opossum CES1. In each case, however, the hydropathic nature of these sites has been retained, which may reflect retention of functions for these sites. The predicted secondary structure for opossum CES1 was similar with that reported for human CES1 3152930, although differences in helical lengths at the 'side door' and 'gate' sites were apparent (Figure 2).

Figure 2 and Table 2 show the amino acid sequence alignments, the predicted secondary structures and sequence identities for the predicted opossum CES1, multiple CES2Like and CES6Like proteins, as well as the six human CES gene products, previously described 1617181920212223242526 (RS Holmes, J Glenn, JL VandeBerg & LA Cox: Baboon carboxylesterases 1 and 2: sequences, structures and phylogenetic relationships with human and other primate carboxylesterases, unpublished). All of the opossum and human CESs examined showed similarities in sequences and predicted secondary structures, consistent with being members of the CES αβ-hydrolase family 315. Using the 3-dimensional structure reported for human CES1 as the basis for discussing structure-functional relationships 3152930, a number of common features were observed for most enzymes, including the N-terminal 'signal peptide', the cysteine residues involved in forming disulfide bridges, the active site serine, glutamate and histidine residues; and the hydrophobic regions of the 'side door' and 'gate'.

A number of differences were observed however between these family members, which have been previously reported for the human CES family members 161718192021222324. Human and opossum CES6 showed longer and distinct N-terminal sequences which may reflect differences in processing the signal peptide in the endoplasmic reticulum 12. Functional charge clamps that perform key roles in maintaining the trimeric-hexameric subunit structures for human CES1 33 were notably absent from human CES2 and CES7, while opossum CES2.2, CES2.3 and CES6, as well as human CES3 and CES6 had only one charge clamp. These may contribute to differences in the tertiary and quaternary structures for these enzymes, in particular the monomeric subunit structure for human and baboon CES2 27 [RS Holmes, J Glenn, JL VandeBerg & LA Cox: Baboon carboxylesterases 1 and 2: sequences, structures and phylogenetic relationships with human and other primate carboxylesterases, unpublished]. The tetrapeptide C-terminal sequence (HIEL for human CES1) which performs a key role of retaining CES1 within the lumen of the endoplasmic reticulum 44 was also found (with conservative substitutions) in human CES2 (HTEL) and CES3 (QEDL), and in opossum CES1 (HIEL), CES2.2 and CES2.3 (both RMEL) but was notably absent from opossum CES2.1 and from human CES6 and CES7 C-terminal sequences. It is likely that these enzymes have different subcellular distribution profiles as compared with other CES gene products. As mentioned previously, the N-glycosylation site reported for human CES1 (79–81: NAT) is retained for the opossum CES1 (NTT) and CES2.1 (NAT) sequences, but is absent from other opossum and human CES gene products, whereas opossum CES1 (2), CES2.1 (4), CES2.2 (1) and CES2.3 (1) have other potential glycosylation sites (Figure 3). Human CES3 (105Asn-Ser-Ser107), opossum CES6 (356Asn-Ser-Thr and 455Asn-Ile-Thr), human CES6 (37Asn-Pro-Ser, 318Asn-Val-Thr, 380Asn-Ser-Thr and 465Asn-Ile-Thr); and human CES7 (281Asn-Ala-Ser, 363Asn-Lys-Ser, 463Asn-Leu-Thr and 473Asn-Met-Thr) sequences also exhibit other potential glycosylation sites which may contribute to carbohydrate binding for these proteins.

Amino acid substitutions were observed for other key regions reported for human CES1, including the Z-site Gly354 (involved in cholesterol binding) 3, which was retained for human and opossum CES2Like gene products and human CES7, but replaced in opossum CES1, human CES3 and in opossum and human CES6 sequences; the human CES1 'side door' sequence at 423Val-Met-Phe425 (proposed to regulate fatty acid residue release following hydrolysis of fatty acyl ester linkages 3) and the 'aromatic-releasing' residue (the 'gate') at 551Phe have each undergone conservative substitutions for the opossum and human CES sequences analyzed. The impacts of these changes on the structures and functions for these enzymes remain to be determined.

Opossum CES2.1 cDNA [GenBank:EU019537] was cloned and sequenced. Using BLAT 3839 to align the cDNA sequence to the recently published genome 40 revealed that the gene is localized on chromosome 1. The identification of this gene as CES2.1 is based upon the identity of the corresponding predicted protein with the deduced amino acid sequence for opossum CES2.1 (Figure 2). The BLAT studies localized this gene on chromosome 1 at nucleotides 677,900,919–677,927,002 on the negative strand with 99.9% identity (Table 1; Figure 3). Two further CES2 'like' genes were identified on this chromosome at nucleotides 677,773,395–677,808,416 (CES2.2 gene) and 677,826,454–677,852,862 (CES2.3 gene), also on the negative strand. Consequently, all 3 CES2 like genes are localized within 155 kb of DNA on chromosome 1 (Figure 3). The predicted proteins from the genes (designated as CES2.2 and CES2.3) were highly similar with each other (94% identical) and with opossum CES2.1 (~70% identical) (Table 2).

The human CES1 amino acid sequence was used for BLAT interrogation of the opossum genome, generating an opossum CES1 homologue gene and a predicted amino acid sequence for this enzyme. The responsible gene (CES1) was located at nucleotides 446,224,784–446,256,371 on chromosome 1 with a span of 31588 nucleotides on the negative strand (Table 1; Figure 3). This protein was more similar with human CES1 (63% identity) than with human CES2 (46% identity) (Table 2). Another study was conducted using human CES6 for BLAT interrogation of the opossum genome and evidence obtained for a region of sequence similarity, located within 44 kb of the CES2.2 gene (Table 1; Figure 3). A predicted protein sequence for opossum CES6 was obtained and compared with other opossum and human CES gene product sequences in Figure 2. These results strongly suggest that there are three CES2 'like' genes which are 92 kb apart on opossum chromosome 1; a CES6 'like' gene, which is 44 kb closer to the pter region than the CES2 gene complex; and a CES1 'like' gene, which is further upstream on chromosome one (Figure 3). The RT-PCR studies reported earlier for opossum CES1, CES2, CES2.2 and CES2.3 genes confirm the existence, expression and sequences for CES coding regions for these opossum genes.

Human CES3 and CES7 amino acid sequences were also used to interrogate the opossum genome using the BLAT method 3839, however the results were inconclusive and further molecular genetic analyses will be required to establish the presence or otherwise of these CES genes on the opossum genome.

A phylogenetic tree (Figure 4) was estimated using a progressive alignment of 6 human CES amino acid sequences with the following opossum CES sequences: CES2.1 (derived from sequencing a full-length cDNA); CES1, CES2.2 and CES2.3 (derived from BLAT interrogations of the opossum genome and from sequencing cDNA clones of RT-PCR products for these genes); and CES6 (derived from blat interrogation of the opossum genome using human CES6) 3940. We also included other vertebrate CES homologues including chicken, frog and salmon, in addition to two divergent fly sequences as outgroup sequences.

The amino acid distance tree (Figure 4) shows a cluster of five main groups consistent with CES1, CES2, CES3, CES6, and CES7 gene families being products of ancestral gene duplication events, and indicates that all families arose prior to the divergence of therian mammals ~173–193 MYA 454647. This is consistent with a previous report for mammalian CES genes 28 and also with other studies that showed mammalian CES1, CES2 and CES7 are members of distinct but related gene families 27 [RS Holmes, J Glenn, JL VandeBerg & LA Cox: Baboon carboxylesterases 1 and 2: sequences, structures and phylogenetic relationships with human and other primate carboxylesterases, unpublished]. Phylogenetic trees based on maximum parsimony and Bayesian methods produce very similar results, with similarly high bootstrap and posterior probability support for the distinct CES clusters observed in the distance tree (Fig. 4). No strong support for early branching patterns emerged from any of the different phylogenetic analyses, suggesting the different gene families arose during a short period of time.

Opossum CES1 is an member of the gene CES1 family which has been previously shown to include other mammalian CES1 gene products from human, baboon, rat and mouse, as well as human CES4 227 [RS Holmes, J Glenn, JL VandeBerg & LA Cox: Baboon carboxylesterases 1 and 2: sequences, structures and phylogenetic relationships with human and other primate carboxylesterases, unpublished]. Opossum CES2.1, CES2.2 and CES2.3 gene product sequences, however, show phylogenetic association within the eutherian CES2 group, which includes multiple CES2-like genes from mouse and rat (mouse was included as a representative of this lineage in our analyses) 48495051. Eutherian CES3, CES6 and CES7 genes form distinct lineages, each originating during the early, pre-mammalian radiation of the CES gene family. Human CES3 shares 46% identity with human and opossum CES2 gene products (Figure 3 and Table 2). Human and opossum CES6 and human CES7 share an average 40% and 42% identity respectively, with other human and opossum CES products (Table 2). Our inclusion and phylogenetic placement of chicken, frog and salmon sequences (Fig. 4) is consistent with the view that the early CES divergence events occurred during the early stages of tetrapod evolution, but following divergence from bony fishes, which diverged much earlier from the common ancestor of all tetrapod CES sequences.

We estimated the times of divergence for the CES gene family and its members using a relaxed clock approach 5253, an amino acid alignment, and three major calibration points: mouse-human, eutherian-marsupial, and synapsid-diapsid divergence times estimated from the literature. Our results suggest that the earliest branching events that produced the extant mammalian CES gene families occurred between 378 and 328 MYA (Fig. 4). Divergence dates for the earliest splits coincide with the divergence time of tetrapods approximately 350–360 MYA based on both molecular and paloeontological data 455455 and are consistent with these events predating the tetrapod divergence and post-dating the tetrapod-bony fish split based on two observations: the placement of chicken and frog CES sequences in different CES gene families, and the early divergence of salmon CES-like sequence from the tetrapod sequences (Fig. 4). Based on these findings, it is proposed that several ancient CES gene duplication events took place prior to the origin of mammals, generating five ancestral genes for CES1, CES2, CES3, CES6 and CES7, which have all been retained over the past 180–230 million years of mammalian evolution. Subsequent gene duplications apparently took place prior to (~114 MYA) and during (~8 MYA) the evolutionary diversification of extant marsupials 55, generating three CES2-like genes in the opossum genome, in contrast to a single CES2 gene found in some eutherians.

A 421 bp opossum cDNA (GenBank: DR038241) which exhibits similarity to CES of other species is in the public database. To study opossum CES genes, primers were designed using this DNA sequence to amplify a full-length opossum CES cDNA from liver RNA by the RNA ligase-mediated rapid amplification of 5' and 3' cDNA ends (RACE) method 56. The GeneRacer kit (Invitrogen, Carlsbad, CA) was used for RACE according to the manufacturer's instructions. For 5' RACE, the GeneRacer 5' primer from the kit and a gene-specific primer (5'-acccaccgaagggcagccacctgat-3') were used in the first round of amplification, then the GeneRacer 5' nested primer from the kit and a gene-specific nested primer (5'-gcctgacgcatactcatccccagtgct-3') were used in the second round of amplification. For 3' RACE, one round of PCR with the GeneRacer 3' primer from the kit and a gene-specific primer (5'-tgtggccatccttcctggcatgctt-3') was sufficient to obtain a RACE product. RACE PCR products were cloned into the pCR4-TOPO vector (Invitrogen). Sequencing was performed using the BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA), and nucleotide sequences were analyzed on an ABI Prism 3100 DNA sequencer (Applied Biosystems).

Total RNA was isolated from livers and small intestines of opossums using the TRI Reagent (Molecular Research Center, Cincinnati, OH), and treated with DNase I from the Turbo DNA-free kit (Applied Biosystems) according to the manufacturer's protocol to remove contaminating DNA from the RNA preparations. DNase I-treated RNA was reverse transcribed into cDNA using a High Capacity cDNA Reverse Transcription kit (Applied Biosystems). Control RT-PCR reactions were conducted with DNase-treated liver and intestine RNA in the absence of reverse transcriptase. Liver and intestinal cDNAs were used as templates in RT-PCR to analyze expression of the predicted/cloned CES genes. Primers for the CES genes were as follows. For the CES1 gene, the forward primer was 5'-attcaggggaagcagtcctc-3' and the reverse primer was 5'-tgccatgatgctggaattgt-3'. For the CES2.1 gene, the forward primer was 5'-tgcctcgttgccaatctatctgcttgtg-3', and the reverse primer was 5'-tcagtagtcatgatctcccaatag-3'. For the CES2.2 gene, the forward primer was 5'-catttgtggctcagcttctgct-3' and the reverse primer was 5'-gggcaggaataacaaacatccag-3'. For the CES2.3 gene, the forward primer was 5'-ttgtctgcatcccagaatgtgata-3' and the reverse primer was 5'-gggcaggaatagcaaacatcaaa-3'. The CES2.2 and CES2.3 cDNAs are 97% identical to each other, therefore primers were selected such that mismatch at the 3' end of primers would allow only the target gene to be amplified. Primer sets for the four CES genes span introns, and the size of the PCR products and the absence of PCR products in the control studies demonstrates that they were not amplified from genomic DNA.

BLAT (BLAST-Like Alignment Tool) in silico studies were undertaken using the UC Santa Cruz web site with the default settings (Assembly:2006; Query: BLAT's guess; Sort Output: query score; Output type: hyperlink) 3839. Opossum CES2.2 and CES2.3 genomic alignments were confirmed using Ensembl (Ensembl release 45, MonDom5 assembly, June 2007) 5758. The following UniProtKB/Swiss-Prot Database sequences were used to interrogate the opossum genome sequence 38: human sequences CES1 (P23141), CES2 (O00748), CES3 (Q6UWW8), CES4 (Q8TDZ9); CES6 (Q6UX55) and CES7 (Q96DN9) (Table 1); as well as the opossum CES2 sequence reported in this paper. Genome locations and predicted protein sequences were obtained for each CES sequence used and the results for those regions showing high (>70%) levels of identity with the human CES gene products or with opossum CES2.1 (full identity) were examined and compared with opossum and human CES sequences using the SIM-Alignment tool for Protein Sequences 5758 (see Table 1). Protein sequences were generated in silico for opossum CES genes using the UC Santa Cruz Genome Browser 3839 and human CES1, CES2 and CES6 sequences in BLAT analyses of the opossum genome. Sequences for opossum CES1, CES2.2 and CES2.3 were aligned with the translated sequences for the corresponding RT-PCR products to ensure identity in each case.

Phylogenetic trees were constructed using an amino acid alignment from a ClustalW-derived alignment of mammalian, chicken, frog, salmon CES protein sequences, obtained with default settings 5960. Two fly (Drosophila melanogaster) CES sequences were included as outgroups. Alignment of ambiguous regions, including the amino and carboxyl termini, were excluded prior to phylogenetic analysis, yielding an alignment of 367 residues. Amino acid distance trees were built in PHYLIP (v. 3.57) using JTT+gamma corrected distances and the neighbor joining algorithm 61. Maximum parsimony trees were constructed in PAUP 4.0b 62 using heuristic searches (50 iterations with random addition of taxa). Bootstrap results for each method were based on 100 iterations under similar search criteria. A Bayesian analysis was performed in Mr.Bayes (vers. 3.1.2) 63 under the following search criteria: two independent runs of 1 million generations, with 4 independent chains sampled every 100 generations, under a mixed amino acid model, with a burnin set at 100,000 generations. The average standard deviation of split-frequencies between the two runs was <0.001.

Divergence dates were obtained with the programs ESTBRANCHES and MULTIDIVTIME 5051, assuming the NJ topology (Fig 4). The salmon CES sequence was included to root the tree. Branch lengths and variance-covariance matrices were estimated assuming the JTT model of amino acid replacement, and these branch lengths were used to estimate divergence times in MULTIDIVTIME. We used the following calibration points: (1) 84 and 99 MYA for the minimum and maximum ages, respectively, for each of the primate/rodent splits in Fig. 4 based on 95% confidence intervals of published molecular divergence estimates 64, (2) 173 and 193 MYA for the minimum and maximum ages, respectively, of each eutherian/metatherian split in Fig. 4 based on published molecular divergence estimates 454647, and (3) 300 and 320 MYA for the minimum and maximum ages, respectively, for the chicken/mammal CES1 divergence, based on a conservative 20 MYA window surrounding the well-constrained synapsid/diapsid split at 310 MYA 46. The prior for the root was set at 360 MYA, on the basis of previous paleoontological and molecular estimates for the age of tetrapods 4654.