Hyperconservation of the N-Formyl Peptide Binding Site of M3: Evidence that M3 Is an Old Eutherian Molecule with Conserved Recognition of a Pathogen-Associated Molecular Pattern

Genomic DNA templates for PCR

Genomic DNA was isolated from tissue samples or cell lines. Cell lines from Mus dunni, Mus macedonicus and Mus spicilegus, Mus Coelomys pahari, M. Nannomys minutoides and M. Nannomys setulosus, M. Pyromys platythrix and M. Pyromys shortridgei were from Dr. S. Chattopadhyay (National Institutes of Health, Bethesda, MD). Cell lines of Mus praetextus (mice from R. Sage; Ref.25) and B10.CAS2 (bearing H2-M3^b) were made as described (26, 27). Rattus norvegicus DNA was from outbred Holtzman Sprague-Dawley (Harlan Breeders, Indianapolis, IN) and outbred Wistar rats (Harlan Breeders) and from the Fischer strain CREF cell line (28). DNA from African murine species, Hybomys uvivittatus (Eastern Black-striped field mouse), Mastomys (Praomys) natalensis (African multimammate rat) and Nannomys gratus were provided by P. d’Eustachio (New York University School of Medicine, New York, NY) (29). Murine species Apodemus agrarius (Asian Black-striped field mouse), Diplothrix legata (Ryukyu long-tailed giant rat), Rattus tanezumi (Sladen’s rat), and Tokudaia osimensis (Ryukyu spiny rat) DNA were from H. Suzuki (Hokkaido University, Sapporo, Japan; Ref.30). Peromyscus attwateri, Peromyscus leucopus, and Peromyscus pectoralis DNA were from R. Pfau (Tarleton State University, Stephenville, TX). A Sigmodon hispidus tissue sample was from P. Wyde (Baylor College of Medicine, Houston, TX). We authenticated every sample by sequencing cytochrome b (cytb) and phylogenetic comparison with published sequences. cytb primers were 5′-TYT YCW TYT TNG GTT TAC AAR AC (forward) (where Y = C + T, W = A + T, N = A + T + G + C, R = A + G) and 5′-TGA AAA AYC ATC GTT GT (reverse). These modifications of Ref.31 were suggested by S. J. Steppan (Florida State University, Tallahassee, FL).

High fidelity PCR

High fidelity PCR used the Invitrogen (Carlsbad, CA) Platinum Pfx kit and manufacturer’s protocol for 35 cycles with a 30 s melt at 94°C, 30 s of annealing with optimized temperature (usually 55°C), and 1 min per kilobase for extension time at 68°C and MgSO₄ usually at 1 mM. Most primers were within introns. Primers for exons 2 and 3 introduced SalI and HindIII for 5′ and 3′ ends, respectively (exon 2) and HindIII and XhoI (exon 3). Mus-specific intron primers were based on H2-M3^b (32). Exon 2 primers were 5′-GTC GAC CAA TGC TTG TTC ACT GGC CC (forward) and 5′-AAG CTT TGG ACC TAA ACT GAA AGT GA (reverse); exon 3 primers were 5′-AAG CTT TCA CTT TCA GTT TAG GTC CA (forward) and 5′-CTC GAG TGG TTC CTA GTT GTT CCT CA (reverse); rat-specific primers were based on putative rat M3 intron sequences (National Center for Biotechnology Information database, Accession no. AC112568). Rat-specific primers for exon 2 were 5′-GTC GAC GGT TAT CAG TGA AGG GTT (forward) and 5′-AAG CTT GGC TAA TCT AGC TTA GCA GTA (reverse); exon 3 primers were 5′-AAG CTT TGG TTT CAC TTT CAG TTT (forward) and 5′-CTC GAG CCC AGA CAA CAA GCC TCA CTT (reverse); M3 sequences from non-Mus species were obtained initially using a degenerate forward primer based on an alignment of exon 1 in Mus and Rattus M3: 5′-GGT CKC TYT GGC TGT TA. The reverse primer was based on a similar region at the beginning of exon 4: 5′-CAC ATG TGC CTT TGG GGG AT. The S. hispidus sequence allowed us to design intronic primers to clone S. hispidus M3 into our expression vector. The intronic primers to amplify exon 2 were 5′-GTC GAC GCC CAG GTT CTT GGA GGA A (forward) and 5′-AAG CTT GGA CAT GTG GGA GTT (reverse); exon 3 primers were 5′-AAG CTT GAT CCA AAC CTG GCA GAT (forward) and 5′-CTC GAG CCT AAG GTT GAG GGA TTT (reverse). PCR using these primers discovered S. hispidus M3 not found with the original primer set.

Cloning PCR products

PCR products of the expected size were gel-purified using the Qiagen (Valencia, CA) Qiaquick Gel Extraction kit and eluted products cloned with the Zero Blunt TOPO PCR Cloning kit (Invitrogen). Plasmids were sequenced by Lone Star Labs (Houston, TX) or the DNA sequencing core facility (Baylor College of Medicine) using an Applied Biosystems (Foster City, CA) ABI PRISM 377 DNA Sequencer. Most sequences were confirmed from independent clones and/or sequencing in the reverse direction. M3 and cytb sequences have been deposited at GenBank with Accession nos. AY263509-AY263623.

Phylogenetic trees

A well-aligned database of exons 2 and 3 of 160 class I genes, including the new M3 sequences, sampled four murid subfamilies, diverse mammalian orders, and five vertebrate classes. The transitional/transversional bias was 0.7 in the exon 2 and 3 data set. It was aligned using the Clustal method (33) by MegAlign (DNAstar, Madison, WI) and adjusted manually with BioEdit (34) to maintain codon alignments. Distant class I sequences such as CD1 and FcRN were excluded as they were difficult to align and often generated “long branch” effects. None of these exclusions affected the conclusions described. The databases included the new M3 genes and five M3 sequences from the National Center for Biotechnology Information database: M3^a (C57BL/6, U18797); M3^b (Mus musculus castaneous, M62844); M3^f (B10.SHH, L36074); M3^sp (Mus spretus, L36072), and RT1-M3 (Wistar, AJ249342). Trees were built in Clustal X (35) and MEGA2 version 2.1 (http://watson.hgen.pitt.edu) (36) using neighbor-joining (NJ), minimum evolution, and unweighted pair-group method with arithmetic means (UPGMA) algorithms, with significance estimated from 1000 bootstrap trials. Maximum parsimony methods were run with 100 bootstraps. Using the Tamura-Nei substitution model, the γ parameter was varied from 0.2 to 3.5 (37) without affecting our conclusions. The inclusion of long branch murid class I-b genes, such as M10, Qa-1, and TL, had no effect on the relevant conclusions. Divergence times were estimated by linear regression of genetic distance vs multiple reference divergence times.

Comparisons of genetic distances

Unless otherwise noted, genetic distances used the Tamura-Nei NJ method, γ parameter = 2. We tested a variety of distances estimating algorithms and parameters. To avoid overweighting heavily represented rodent genera in calculating genetic distances using monosubstitution models (e.g., Jukes-Cantor), between-group distances were averaged from between-genera distances. For example, Mus M3 vs nonmurid eutherian I-a distances were averaged as a single observation. Assuming constant rates of evolution, we used Tajima’s (38) relative rate test to assess which of three sequences was the outgroup. Thirty-five sets of the sequences were selected randomly and analyzed in MEGA2. χ² values were summed and significance was tested with 35 degrees of freedom.

Nonsynonymous and synonymous substitution rates

Pair-wise nonsynonymous (d_N) and synonymous (d_S) substitution rates (39) in the ARS and NARS were calculated using SNAP (http://hiv-web.lanl.govl) (40). The ARS, based on HLA-A2 (41), was residues 5, 7, 9, 22, 24, 26, 57–59, 61–77, 80–82, 84, 95, 97, 99, 114, 116, 143, 145–147, 149–152, 154–159, 161–163, 165–167, 169, 171 where residue 1 is the first Gly of α1. Class I-a controls were from pairwise comparison of D^b, D^d, D^f, D^k, D^p, D^s, D^x, K^d, K^dv, K^f, K^j, K^k, K^s, K^u, RT1.A1°, RT1.A1^b, RT1.A1^c, RT1.A1^f, RT1.A1^h, RT1.A1^k, RT1.A1ⁿ, RT1.A1^q, RT1.A2°, RT1.A2^b, RT1.A2^c, RT1.A2ⁿ, and RT1.A2^q (all available from the National Center for Biotechnology Information).

When multiple taxa are studied, variation in time of divergence has a heavy impact on variation in substitution rates (42). Therefore, we regressed (one parameter) the rate of nonsynonymous substitution (d_N) on the rate of synonymous substitution (d_S) and compared the slopes (43) to the value of 1 expected under neutral selection using where m is the slope, μ = 1, s is the SD of the one-parameter slope and n is the number of sequences (not the number of comparisons).

To test for hyperconservation of the 10 residues coordinating N-formyl binding, the d_N/d_S ratio of those residues was compared to the non-ARS (NARS) ratio using and R is the ratio, E_R its relative error and ς_N and ς_S are the SD of d_N and d_S, respectively.

Construction of M3-L^d chimeric expression vectors

Correct ligation of exons 2 and 3 of M3 in an H2-L^dα3 expression vector (44), which carries the epitope for mAb 28-14-8S (HB-27; American Type Culture Collection, Manassas, VA), was confirmed by sequencing. Because the null CTL activity of M3^b mapped to a Leu⁹⁵Gln substitution (12), only exon 3 of M3^b was inserted into pM3^aL^d to test for M3^b N-formyl specificity.

Functional assay of N-formyl peptide specificity

DAP-3 cells (45), cotransfected with pSVneo and pM3-L^d in Lipofectamine Plus (Invitrogen), were selected in 750 μg/ml geneticin (Invitrogen) and sorted with a Beckman-Coulter (Fullerton, CA) Altra flow cytometer after overnight culture at 37°C with N-formyl-MLF-benzylamide (Sigma-Aldrich, St. Louis, MO) and staining with mAb 28-14-8S. pM3L^d transfectants were cultured overnight with 20 μM peptide or DMSO vehicle before staining with 28-14-8S and FITC-goat anti-mouse Ig (Baxter, Mundelein, IL) (46, 47). Cells fixed in 1% paraformaldehyde were analyzed on an EPICS XL-MCL flow cytometer (Beckman-Coulter, Fullerton, CA) with Beckman-Coulter System II version 3.0 software. Peptides (Baylor College of Medicine Protein Core Facility) from L. monocytogenes (fMIVIL), or influenza hemagglutinin (fHA; fMLIIW) (48) and their nonformyl forms were dissolved in DMSO.

Confirmation of sample identity using cytb

To confirm sample identity, we constructed phylogenetic trees from sample cytb sequences (Fig. 1⇓) and published sequences. cytb sequences for M. Nannomys gratus and R. tanezumi were not published. As expected, M. N. gratus and R. tanezumi cytb clustered significantly (with high bootstrap values) with other Mus Nannomys and Rattus sequences, respectively. M. P. shortridgei clustered with M. C. pahari (98% of bootstraps) rather than with other Pyromys (49), consistent with our results for cytb and TL (19), suggesting Pyromys might be polyphyletic. Rattus and Diplothrix sequences formed a single cluster with 100% bootstrap values. All samples of subfamilies Murinae and Sigmodontinae clustered appropriately.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

Phylogeny of cytb. ∗∗, Six cytb sequences previously reported by this laboratory. Underlined sequences were determined for this study. The accession number indicates GenBank sequences. Percent (%) values represent the percentage of bootstrap trials supporting the branch; only values >50% are shown.

Sequence analysis of M3 in murid rodents

From 22 murid species we sequenced 43 distinct genes with high homology to H2-M3 in exons 2 and 3. These genes were not well-differentiated from class I-a-like sequences in exon 4 (data not shown). To confirm orthology with M3, we constructed phylogenetic trees of exons 2 and 3 with candidate sequences, five known M3, and over 100 other class I genes. All M3 candidates were isolated together (>98% of bootstraps) in all tree-forming methods used (Fig. 2⇓, A and B), justifying their classification as “M3”. The M3 species tree (Fig. 2⇓C) resembled that of cytb, except for T. osimensis, which has one copy of M3 very similar to that of A. agrarius, as expected (30), and one highly divergent copy.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2.

Phylogeny of M3. Sequences for exons 2 and 3 were aligned as described in Materials and Methods. A, Representative “close” phylogeny of M3 sequences. ∗, M3 sequences previously published. B, Large-scale tree representative of non-UPGMA methods. Most branches are collapsed for clarity. M3 clusters with eutherian I-a genes, but not with murid I-a, suggesting an origin at or prior to the eutherian radiation. C, UPGMA methods place M3 outside the eutherian cluster (86%), suggesting an origin prior to the eutherian radiation.

Large-scale trees (Fig. 2⇑, B and C) revealed three notable features regardless of the algorithm used. First, the M3 branch never clustered significantly with other murid class I genes even when (as in most NJ algorithms) it did cluster with eutherian class I-a genes in general (Fig. 2⇑B). Second, M3 genes were on a “long branch,” indicating a larger genetic distance between M3 and class I-a genes, than between class I-a genes of different mammalian orders. Third, in trees using an algorithm (UGPMA) that assumes constant rates of substitution, M3 appeared to have evolved before the radiation of placental mammals (Fig. 2⇑C). Thus, all tree-building methods suggested that M3 evolved before the eutherian radiation, or evolved unusually rapidly.

Thirty-nine of 43 new M3 sequences were easily aligned with class I-a genes in exons 2 and 3. Four sequences (marked insertion and/or deletion (indel) in Fig. 2⇑A) from M. C. pahari and M. P. shortridgei are altered near the 5′ end of exon 2 (Fig. 3⇓). In each of these, a 9-bp tandem duplication encoding the N-formyl interaction residue His⁹ is closely followed by an “indel”, predicting a polypeptide with a net gain of one amino acid without a frameshift. The indel⁺ M. C. pahari sequences have premature stop codons and exhibit a nonfunctional Asn Pro Ser glycosylation motif (50) at positions 86–88, a site critical for peptide binding by H2-L^d (51). In contrast, indel⁺ M3 from M. P. shortridgei are otherwise intact and might encode functional proteins.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3.

Amino acid alignment of M3 genes. Nucleotide sequences predicting identical protein sequences are listed together to the left of the sequence. A, α1, B, α2. Key:—identity with M3^a; ∗ stop codon; large arrows: N-formyl coordinating residues; #, deletion; @, insertion of HYFH; &, insertion of HFFH; $, insertion of RAPWMETR (a near tandem duplication).

Multiple copies of M3 have not been reported in laboratory mice. However, R. norvegicus has at least three copies (52). We isolated four sequences from M. C. pahari, three from M. P. shortridgei, and five from M. N. gratus. Assuming monomorphism, these sequences represent distinct loci. Two distinct M3 sequences were found in T. osimensis, each with multiple defects. We cannot rule out the possibility that this species carries another copy of M3 that is functional.

Conservation of the N-formyl coordination residues

Ten residues of M3^a (Y7, H9, Y22, S24, L63, K66, V67, I70, V99, Y159) coordinate N-formyl binding (53) of which four (Y7, Y22, S24, Y159) are frequent in other class I molecules. Excepting the six indel⁺ sequences, there were only three variants in these 10 residues among 42 sequences (99% identity), suggesting hyperconservation (Table I⇓). All three variants, at residues 7 and 159, in two genes from S. hispidus, are at sites highly conserved among most MHC molecules; two were identical conservative Tyr¹⁵⁹Phe substitutions. There was complete codon conservation for His⁹, Leu⁶³, and Val⁹⁹, and only synonymous substitutions at Lys⁶⁶, Val⁶⁷, and Ile⁷⁰ sites (Table I⇓). Thus, the N-formyl-coordinating residues are extremely well-conserved and the N-formyl coordinating residues should be under negative selection.

Synonymous and nonsynonymous rates of substitution in M3

To test whether the ARS of M3 has been under positive (diversifying) or negative (purifying) selection, we regressed nonsynonymous substitution rates against synonymous rates (Fig. 4⇓). As expected, mouse and rat I-a ARS were positively selected (1.48 ± 0.03), but the M3 ARS was under strong negative selection (slope = 0.236 ± 0.002, p < 0.001 versus a slope of 1). As expected, the slope of the M3 NARS, 0.162 ± 0.002, was also negatively selected (p < 0.001; Fig. 4⇓B); this slope is similar to that of many murine non-MHC genes (54). Negative selection of murine class I-a NARS sequences (0.59 ± 0.01) was consistent with other reports for class I-a genes (15). The 10 N-formyl-coordinating residues were hyperconserved (d_N/d_S = 0.025 ± 0.003; p < 0.001) compared to the NARS (19).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4.

ARS and NARS residues of M3 are under negative selection. Pairwise rates of nonsynonymous (d_N) substitution are plotted against synonymous rates (d_S). •, M3; □, mouse H2-K and H2-D and rat RT1-A. A, ARS. The slope of M3 was 0.236 ± 0.002; the slope of I-a was 1.48 ± 0.027. B, NARS. The slope of M3 was 0.162 ± 0.002; of I-a, 0.589 ± 0.006. All four slopes differed from neutrality (dashed line, slope = 1, p < 0.01).

Conservation of N-formyl peptide binding specificity

Hyperconservation of N-formyl coordination residues suggested conservation of N-formyl binding specificity. To test this, the α1α2 domains of “null” H2-M3^b and both rat and S. hispidus M3 were expressed in an L^dα3 expression vector and tested for peptide-induced surface expression in transfected cell lines (46, 47). fMIVIL and fHA but not nonformylated peptides induced surface expression of M3^a, M3^b, RT1-M3, and Sihi-M303 (Fig. 5⇓A), demonstrating conservation of N-formyl specificity.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 5.

Rat, cotton rat, and the null H2-M3^b proteins are N-formyl specific. Cells transfected with indicated M3L^d vectors were incubated with 20 μM of the indicated peptide derived from L. monocytogenes (MIVIL) or influenza hemagglutinin (MLIIW) and stained the next day for expression of M3L^d. Cells transfected with a neoplasmid served as a control. Values are the average and SD of three independent experiments.

The indel⁺ M. C. pahari sequences had multiple debilitating mutations in exon 2 but the indel⁺ Mush-M302 was intact except for two nonconservative substitutions in the N-formyl coordinating site (Fig. 3⇑A). This suggests the indel was acquired in the ancestral species before other mutations accumulated in M. C. pahari. Null duplicates can be fixed in a population but are also rapidly lost (42); the presence of indel⁺ M3 in two species suggests it may retain some function, and allowed us to test experimentally and statistically whether the indel event neutralized M3.

To test whether indel⁺ M3 retains N-formyl specificity, we expressed the α1 and α2 domains of Mush-M302 with in the L^d vector. In contrast to our results using M3^b, previously considered a “null” allele, the indel⁺ M3 was not induced by any tested peptide (not shown). N-formyl peptides did induce a chimeric molecule with M3^a exon 2 and exon 3 from the indel⁺ gene (data not shown). This negative experimental result suggests the indel of exon 2 destroyed N-formyl specificity but cannot rule out the possibility that the indel simply altered peptide specificity. Compared with all other M3, the d_N/d_S slopes of both the ARS (0.27 ± 0.01) and NARS (0.40 ± 0.02) of the indel⁺ genes were below 1, but significantly higher than the ARS (0.21 ± 0.01) and NARS (0.20 ± 0.01) of indel-negative genes of the same two species, consistent with a relaxation of negative selection after indel acquisition. Moreover, in direct comparisons (n = 2) between exons 2 and 3 of M3 of M. C. pahari and M. P. shortridgei, the d_N/d_S ratio of indel⁺ M3 was 1.4 ± 0.22 compared to 0.27 ± 0.01 for indel-negative M3. Clearly, only the indel-negative genes have been under negative selection. The experimental and statistical results are consistent with the interpretation that the indel in exon 2 inactivated M3.

Three models of M3 evolution before the sigmodontine/murine divergence

Despite the long branch length of M3 genes (Fig. 2⇑, A and B), the d_N/d_S ratios and functional data indicated M3 evolved conservatively under negative selection since the murine/sigmodontine split ∼65 MYA. Phylogenetic trees failed to join M3 to a murid branch, and UPGMA trees put M3 completely outside eutherian I-a genes. However, UPGMA models assume constant evolutionary rates, and all models assume a uniform or monophasic γ distribution of rates at different sites along the gene. Therefore, the origins of M3 within murid I-a genes (Fig. 6⇓A, model I) might be obscured by nonuniformity or inconstancy of substitution rates.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 6.

Evolution of NARS residues in M3 and I-a genes. A, Three models of M3 evolution. I. M3 was a recent offshoot of murid I-a genes. II. M3 arose before the eutherian radiation. III. Murid rodents were an early offshoot of eutherian mammals. B, M3 and I-a genes have evolved linearly since the mammal/bird divergence. The average between-group protein p-distances were plotted against estimated divergence time for different taxa. “Young” estimates (<∼65 MYA) were obtained from between-taxa comparisons of M3 or of murid I-a genes; older estimates used marsupial and birds as outgroups. For example, the average pairwise distance between Peromyscus and Mus M3 was plotted against the divergence time (50–65 MYA) of Murinae and Sigmodontinae. “Old” estimates were computed from genetic distances between M3 (or I-a) and marsupial or bird class I genes, the latter treated as outgroups. •, M3 distance = 0.043 + 0.0024/MY, r² = 0.98. □, I-a, distance = 0.068 + 0.0019/MY, r² = 0.96. This plot is representative of plots produced using protein and nucleotide data, and using Poisson or Tamura-Nei corrections for multiple substitutions. Substitution rates for I-a related genes ranged from 0.0013 to 0.0019/MY (r² from 0.80 to 0.98). Substitution rates for M3 were 10–30% higher, and ranged from 0.0016 to 0.0022/MY (r² from 0.97 to 0.99). Δ, Distances between M3 and murid and eutherian I-a genes (0.335 ± 0.004) are plotted at the time ∼75 MYA estimated for the murid radiation under model I or at 131 MYA predicted as the M3/I-a divergence (assuming linear evolution under model II). Values using other distance measures and divergence times ranged from 100 to 200 MYA, but were linear in all the models. A saltation is indicated by the gray line leaving the I-a line after the eutherian radiation, and returning to the M3 line by the time of the murine/sigmodontine divergence. Transitional forms would be predicted to fall along this line. C, No evidence for positive selection in the NARS driving rapid evolution of M3 from murid I-a genes. Nonsynonymous substitution rates are plotted against synonymous rates. •, M3 pairs, slope = 0.17; □, I-a pairs; slope = 0.6; Δ, M3-I-a pairs, slope = 0.20. Dashed line indicates a slope of 1 (neutral selection).

Because ARS and NARS residues can evolve at different rates in class I genes, we analyzed NARS residues, which are under negative selection and should exhibit more uniform evolutionary rates. As expected, genetic distances between M3 and murine I-a genes were less using the NARS alone, compared with ARS + NARS. However, this shortening also affected distances between I-a genes as well, such that NARS genetic distances between M3 and murid I-a-like genes were significantly greater than between murid and eutherian I-a-like genes (data not shown). Moreover, as in Fig. 2⇑, the NARS of M3 genes remained on a long branch (not shown) and did not cluster with other murid I-a genes.

To test whether M3 and/or I-a evolution rates were constant over time, we plotted NARS genetic distance against estimated taxonomic divergence dates (Fig. 6⇑B). Using a variety of distance-estimating algorithms, and the range of published rat/mouse, murine/sigmodontine, eutherian/marsupial, and mammal/bird divergence times, both M3 and I-a genetic distances were linear since the mammal/bird divergence. Moreover, M3 and I-a rates were very similar, though rates estimated for M3 ranged from 0 to 30% faster than those estimated for I-a genes.

Three models might explain why M3 looks old (Fig. 6⇑A). In model I, exons 2 and 3 of M3 arose from murid I-a genes, through a “leap” (saltation) involving rapid divergence after gene duplication, during which M3 evolved N-formyl specificity and, after which, M3 evolution became conservative. The saltation generated a large genetic distance, giving M3 the appearance of old age. This model predicts M3 will be found only in murids, that transitional forms should occur in other murids, and the genetic distance between M3 genes and such transitional forms should follow the gray line indicated in Fig. 6⇑B. In model II, M3 looks older because it is older, arising before the eutherian radiation. This model predicts M3 orthologs should be found in other rodent or mammalian groups. Model III combines the first two: M3 arose in murids, but murids are older than nonmurid mammals (55, 56).

Model I predicts M3 and murid I-a genes are most closely related, while model II predicts that murid I-a and eutherian I-a sequences are more closely related. However, the distance between M3 and murid I-a genes (0.23 ± 0.02, SEM; n groups = 36) was not different from that between M3 and nonmurid eutherian I-a genes (0.21 ± 0.01; n = 18), and longer than the distance between murid and nonmurid eutherian I-a genes (0.14 ± 0.004; n = 8). Assuming equal substitution rates (Fig. 6⇑B), we tested whether eutherian I-a genes are a likely outgroup for M3 and murid I-a genes (model I) or if M3 is the outgroup (model II). We used Tajima’s (38) relative rate test to assess these two models using randomly selected sequence sets from our database. The results were significantly different from model I (p = 2 × 10⁻¹⁵) but not with model II (p = 0.58). Thus, NARS genetic distances and Tajima’s test both favor model II. It might be argued that the NARS of M3 expresses an altered function, and that positive selection for new functionality (17) drove rapid evolution of M3 ARS. Thus, both models I and III predict an elevated d_N/d_S slope in comparisons of M3 NARS with murid I-a-like genes. However, this slope was very low (Fig. 6⇑C), consistent with model II.