Signal Sequence Conservation and Mature Peptide Divergence WithinSubgroups of the Murine b-Defensin Gene Family

Gillian M. Morrison, Colin A. M. Semple, Fiona M. Kilanowski,Robert E. Hill, and Julia R. DorinMRC Human Genetics Unit, Western General Hospital, Edinburgh, Scotland

b-defensins are two exon genes which encode broad spectrum antimicrobial cationic peptides. We have analyzed thelargest murine cluster of these genes which localizes to chromosome 8. Using hidden Markov models, we identified sixb-defensin exon 2–like sequences and subsequently found full-length expressed transcripts for these novel genes.Expression was high in brain and reproductive tissues. Eleven b-defensins could be grouped into two clear subgroups byvirtue of their position and high signal sequence (exon 1 encoded) identity. In contrast, however, there was a very lowlevel of sequence conservation in the exon 2 region encoding the mature antimicrobial peptide. Examination of the genesequences of orthologs in other rodents also revealed an excess of nucleotide changes that altered amino acids in themature peptide region. Evolutionary analysis revealed strong evidence that following gene duplication, exon 1 andsurrounding noncoding DNA show little divergence within subgroups. The focus for rapid sequence divergence islocalized in the DNA encoding the mature peptide and this is driven by accelerated positive selection. This mechanism ofevolution is consistent with the role of this gene family as defense against bacterial pathogens and the sequence changeshave implications for novel antibiotic design.

Introduction

Defensins are antimicrobial cationic peptides ofwhich three groups have been described in vertebrates.The a and b forms are distinguishable by the spacing andconnectivity of the conserved cysteine residues within themature peptide (Ganz and Lehrer 1994). The b-defensinsare two exon genes which code for a prepropeptide thatincludes a signal sequence, a short propiece, and a maturecarboxy-terminal peptide, which is liberated by proteolyticcleavage (Lehrer and Ganz 2002). The signal sequence isexon 1 encoded and the pro and mature peptides are codedfor by the second exon. Recent studies have shown b-defensins to be present in the airway epithelia of severalspecies, including humans, rodents, and cattle where theycontribute to the host defense system by the eradication ofpathogens at the mucosal surface.

Four human b-defensins have been functionallyanalyzed, and all of them have displayed antimicrobialactivity in vitro (Lehrer and Ganz 2002). Interestingly,several human b-defensins also display chemoattractantactivity for immature dendritic cells and memory T-cells,suggesting that these peptides may also function as animportant link between the innate and adaptive immunesystem (Yang et al. 1999; Garcia et al. 2001). Untilrecently, only eight murine b-defensin genes had beenidentified, all mapping to a region of conserved syntenywith the human locus on chromosome 8 (Huttner, Kozak,and Bevins 1997; Bals, Goldman, and Wilson 1998;Morrison et al. 1998; Morrison, Davidson, and Dorin1999; Jia et al. 2000; Bauer et al. 2001; Yamaguchi et al.2001). Synthetic murine b-defensin peptides also possessantimicrobial activity (Bals, Goldman, and Wilson 1998;Morrison et al. 1998; Bals et al. 1999; Bauer et al. 2001;Yamaguchi et al. 2001; Morrison et al. 2002) demonstrat-ing the functional conservation of these peptides between

mouse and man. However, dramatic differences areevident within the murine b-defensin gene family in termsof their tissue distribution and responsiveness to in-flammatory stimuli (Lehrer and Ganz 2002). We havealso reported functional differences within members of themurine b-defensin gene family in terms of their ability tokill specific pathogens (Morrison et al. 1998, 2002). Theseresults suggest that the murine b-defensins may haveevolved under positive selection to have different profilesof antimicrobial activity and thus maximize the host’sability to fight opportunistic infections.

Here we report the further characterization of themurine b-defensin gene cluster and investigate the molec-ular mechanisms responsible for the divergence betweenthese molecules.

Materials and MethodsIsolation of BACs Containing b-Defensins

A murine C57BL/6J BAC library (RPCI-23) wasscreened using radiolabeled double-stranded 40merscomprising two 24mers with a complementary 8-bpoverlap (overgos) specifically designed to the exon 1regions of the previously identified murine b-defensins(http://genome.wusl.edu/gsc/overgo). The positive BACswere then screened for the presence of the individualdefensin genes by PCR and hybridization with radio-labeled oligonucleotides. The sequences of the PCRprimers and hybridization oligonucleotides are detailedin supplementary information.

BAC Sequencing

BACs were sequenced as part of the UK mousesequencing program (http://mrcseq.har.mrc.ac.uk), and thesequence is freely available.

BAC Sequence Analysis

Sequence generated from the BAC 353A15 and357B4 was analyzed using NIX, a Web tool designed to

Key words: adaptive evolution, b-defensin, positive selection, Musmusculus.

E-mail: julia@hgu.mrc.ac.uk.

Mol. Biol. Evol. 20(3):460–470. 2003DOI: 10.1093/molbev/msg060� 2003 by the Society for Molecular Biology and Evolution. ISSN: 0737-4038

460

allow the viewing of multiple DNA analysis programs(http://www.hgmp.mrc.ac.uk). The sequence was alsoanalyzed for the presence of novel b-defensins usinga calibrated HMMER model (version 2.1.1). The HMMERmodel was constructed from a ClustalW (version 1.82 withdefault settings) (Higgins, Thompson, and Gibson 1996)multiple sequence alignment of b-defensin sequencesretrieved from GenBank.

Tissue Expression of Novel Murine b-Defensin Genes

The RACE protocol was adapted from (Townleyet al. 1997), as previously described (Morrison, Davidson,and Dorin 1999), and RT-PCR was carried out asdescribed (Morrison, Davidson, and Dorin 1999). Primersequences are available in supplementary information.

Phylogenetic Inference and Evolutionary Analysis

Sequences were aligned using ClustalW (version1.82) (Higgins, Thompson, and Gibson 1996), and gappedpositions were omitted from subsequent analyses. Thephylogenetic tree was constructed by neighbor-joining(Saitou and Nei 1987) based on the proportion of differentamino acid sites and was rooted with chicken Gallinacin 1;its reliability was assessed with 1,000 bootstrap replica-tions using MEGA (version 2.1) (Kumar et al. 2001). Inpairwise comparisons between nucleotide sequenceswithin subgroups, the number of synonymous substitu-tions per synonymous site (dS) and the number ofnonsynonymous substitutions per nonsynonymous site(dN) were estimated using the Nei-Gojobori test. Standarderrors for dS and dN were calculated using 1,000 bootstrapreplicates. In addition, the Jukes-Cantor correction (Jukesand Cantor 1969) was applied to account for multiplesubstitutions at the same site. A one-tailed Fisher’s exacttest was used to test the null hypothesis that theproportions of synonymous and nonsynonymous differ-ences are the same (Zhang, Kumar, and Nei 1997). Alldistance calculations and codon-based tests of selectionwere carried out using MEGA (version 2.1). Estimates ofthe proportions of radical and conservative nonsynon-ymous substitutions were made according to the methodsof Zhang (2000) using the HON-NEW program. Theradical or conservative nature of nonsynonymous sub-stitutions was assessed with respect to charge and to thepolarity and volume of the amino acids (the Miyata-Yasunaga amino acid classification) as defined by Miyata,Miyazawa, and Yasunaga (1979). Intron sequences forDefb11, Defb10, and Defb2 were aligned using DIALIGN(version 2.1) (Morgenstern 1999). The numbers ofsubstitutions between the introns were estimated usingKimura’s two-parameter method (Kimura 1980).

Sequence Analysis of Exon 2 Within the Rodentia Genus

The exon 2 sequences of nine murine b-defensingenes, representing members of both subgroups, wereamplified from genomic DNA from the inbred Musmusculus mouse lab stains C57Bl/6N, 129/Ola, CBA,and BALB/cJ; from the Mus musculus species Musmusculus domesticus, Mus musculus musculus, Mus

musculus castaneous, and Mus musculus hortulanus; fromthe Mus subspecies Mus spretus, Mus caroli, and Muspahari; and from the non-Mus species Apodemus sylva-ticus and Rattus norvegicus. DNA was purchased from theJackson Laboratories. The primer sequences are detailed insupplementary information. Genomic samples that failedto generate a PCR product were amplified with at least oneset of alternative primers designed specifically towards themurine gene exon 2 sequence. PCR products weresequenced as previously described (Morrison, Davidson,and Dorin 1999) and the resulting sequence compared withthat obtained from C57Bl/6 DNA. Isoelectric points (pI)were calculated from predicted mature peptide proteinprimary sequence using the ExPasy compute pI tool (http://us.expasy.org/tools/pi_tool.html).

ResultsConstruction of Two BAC Based Contigs ContainingSix Novel and Eight Known b-Defensin Genes

We have isolated a series of BACs containing themurine chromosome 8 b-defensin gene family froma C57BL/6J library using overlapping oligonucleotidesdesigned on the previously identified murine b-defensins.Data from a BAC end sequence database (www.tigr.org),together with the results of the hybridization experimentsand PCR, allowed the BACs to be ordered into two contigsseparated by approximately 850 kb (fig. 1A). The firstcontig contained five previously described genes (Defb7[Bauer et al. 2001] Defb3 [Bals et al. 1999], Defb5[GenBank accession number AF318068], Defb6 [Yama-guchi et al. 2001], and Defb 4 [Jia et al. 2000]). Anadditional gene, Defr1, was identified from this contig. Ithad antimicrobial activity and clear b-defensin homology,although it was lacking one of the canonical cysteines inthe mature peptide (GenBank accession numberAJ344114). This gene is 98% identical to Defb8 (Baueret al. 2001) but has significant changes in the exon 2sequence, resulting in a tyrosine residue rather thana cysteine residue (Morrison et al. 2002).

The second contig consisted of two overlapping BACs,353A15 and 357B4, which contained Defb1 and Defb2,respectively (Huttner, Kozak, and Bevins 1997; Bals,Goldman, andWilson 1998;Morrison et al. 1998;Morrison,Davidson, and Dorin 1999). These BACs were chosen forsequencing because of reduced stringency hybridizationexperiments that had indicated they contained multiple b-defensin related sequences (data not shown). Completesequence was generated from BACs 353A15 and 357B4 bythe UK MRC mouse genome sequencing program, fundedby the UK Medical Research Council and can be obtainedfrom ftp://ftp.sanger.ac.uk/pub/mouse/ and analyzed usingNIX (http://www.hgmp.mrc.ac.uk). This analysis identifiedthe known b-defensins Defb1 and Defb2 and multiple a-defensin genes (cryptdins) as expected, in addition to theCu21 transporting, b-polypeptide gene, Atp7b, which is notrelated to themurine defensins. Analysis using theHMMERmodel described in the methods section resulted in the iden-tification of six novel b-defensin exon 2–like sequences.Following translation, all of these sequences were foundto contain the six canonical cysteine residues within their

Murine b-Defensin Gene Evolution 461

predictedmature peptide sequence.A recent study describedthe identification of multiple putative murine b-defensinsequences also using a computational search strategy(Schutte et al. 2002). The six b-defensin genes describedhere were also identified as gene fragments in theirpublication, and the names given to them were Defb7,Defb11, Defb9, Defb15, Defb13, and Defb35. We namethem here following the recently approved HUGO nomen-clature for the b-defensin gene family, respectively,Defb10,Defb15, Defb9, Defb7, Defb13, and Defb35 (fig. 1B). 59RACE sequence analysis and examination of EST librariesallowed us to identify full-length transcripts for all of thesesequences and verify that they are expressed transcriptsindicative of functional genes.

Expression Profiles of the Novel Murine b-DefensinGenes Defb10, Defb11, Defb9, Defb15, Defb13, andDefb35

Full-length transcripts for two of the six sequenceswere present in the murine EST database. Defb11 waspresent in a neonate cerebellum and an adult testisexpression library and Defb9 was present in an adulthippocampus library. 59 RACE analysis from trachea,testis, and brain tissue resulted in the identification of full-length transcripts for four of these sequences, Defb10,Defb11, Defb15, and Defb13.

Defb10 and Defb11 were found to be expressed inboth adult and neonate brain (fig. 2) with weak expressionbeing detected from kidneys, epididymis, and testis afterincreased numbers of cycles of PCR (data not shown).Defb15 and Defb13 are expressed in the testis and to

a lesser extent in the epididymis (fig. 2). Weak expressionof Defb15 and Defb13 was detected in the kidneys, andDefb15 was also found to be expressed at low levels in thecolon (data not shown). Expression of Defb9 was onlydetected from brain cDNA after 35 cycles of RT-PCRfollowed by hybridization of the PCR product witha radiolabeled internal oligonucleotide probe. RT-PCRanalysis of Defb35 using primers specific to the putativeexon 2 sequence indicated Defb35 to be expressed in thetestis and epididymis (fig. 2). RT-PCR using primers fromexon 1 to 2 indicated expression also in kidney andneonatal and adult brain (data not shown). The predictedpeptide sequences of the genes contained in the twoseparate contigs are shown in figure 3.

Evolutionary Analysis of the Murine b-Defensin GeneFamily

A phylogenetic tree was constructed with the 14validated and expressed murine b-defensin genes (fig.3B). Two clusters of genes are evident with Defb6, Defb4,Defb5, Defb3, Defb7, and r1 in one subgroup supported bya 99% bootstrap value and Defb9, Defb2, Defb10, andDefb11 in another subgroup with a statistically significantbootstrap value of 99%. Defb1 probably also falls into thisgroup with a bootstrap value of 94%. We refer to thesegroups as subgroup 1 and 2, respectively. Defb35, Defb15,and Defb13 are not so strongly supported by bootstrapvalues as being subgroup members. The separation of thegenes into subgroups 1 and 2 also reflects the spatialarrangement of the genes along chromosome 8 (fig. 4), that

FIG. 1.—Genomic organization of the murine defensin genes. (A) Diagrammatic representation of mouse chromosome 8 region A3 showingthe relative positions of murine b-defensin genes. The b-defensin gene names are abbreviated as follows: 75 Defb7, R15 Defr1, 35 Defb3, 55Defb5, 6 5 Defb6, and 4 5 Defb4. Arrowheads indicate gene orientation. The two BACs selected for sequence analysis, 353A15 and 357B4, areshown. (B) Relative positions of the new b-defensin genes identified within BAC 357B4. The gene Atp7b, which is not a b-defensin gene, is alsoshown.

462 Morrison et al.

is, closely related genes being adjacent to each other. Inaddition the genomic organization (exon 1 length) is thesame within subgroup members (fig. 3A).

Evidence of Positive Selection Within the Murineb-Defensin Gene Family

The mean identities of the exon 2 sequences ofsubgroup 1 and subgroup 2 were significantly lower thanthe mean identities of the first exons at 65 % (standarddeviation [SD] 5 10.54) and 62 % (SD 5 6.08), re-spectively (fig. 4). This difference between identities inthe different regions of these genes was even morepronounced when the predicted peptide sequences wereanalyzed. The prepro part of the peptide has meansequence identities of 86 % (SD 5 6.44) and 92 % (SD 56.6) in subgroup 1 and subgroup 2, respectively, but thepredicted mature peptide had mean identities of only 45%(SD 5 7.94) and 42 % (SD 5 3.6) in subgroup 1 andsubgroup 2, respectively. The identity between familymembers appears to be lost very rapidly after the firstcysteine in the mature peptide. In order to characterize thedriving force underlying the functional diversificationof the murine b-defensin genes, we tested the role of posi-tive Darwinian selection. Figure 5 shows that when thesynonymous (dS) and nonsynonymous (dN) nucleotidesubstitutions per site are calculated for the exon 2sequences of paralogous genes within the subfamilies ofMus musculus C57Bl/6, it is clear that for subfamily 1 allexcept one pairwise comparison (Defr1 and Defb3) givesa dN value greater than the dS value. This implies positiveselection, and using the Nei and Gojobori (1986) method,we can show that the excess of nonsynonymous changes issignificant (using the Fisher’s exact test) between Defb5versus Defb7 and Defr1 versus Defb7. In subfamily 2,

pairwise comparisons of family members with both Defb1and Defb 9 do not show an excess of nonsynonymouschanges. However, in diverging from Defb10 and Defb11,the evolution of the sequence encoding the Defb2 maturepeptide has involved an excess of dN relative to dS. Wealso subjected this observation to rigorous analysis usingthe Nei and Gojobori (1986) method. The excess ofnonsynonymous changes in both subfamilies are signifi-cant by the conservative Fisher’s exact test of the nullhypothesis that the proportions of synonymous andnonsynonymous differences are the same (Zhang, Rosen-berg, and Nei 1998) (table 1).

In order to investigate what type of amino acidchanges are favored in these genes, we calculated theproportions of radical versus conservative amino acidsubstitution. The averages for pR (radical nonsynonymoussubstitution)/pC (conservative nonsynonymous substitu-tion) calculated over 47 mammalian genes were 0.81 forcharge and 0.49 for polarity and volume (Miyata-Yasunaga amino acid classification) (Zhang 2000). Theequivalent values (shown in table 2) for comparisonsbetween the second exons Defb7 versus Defb5 and Defr1are greater than 1 with respect to charge but not withrespect to polarity and volume. That is, of the non-synonymous changes observed, most have tended tochange the charges of the residues encoded but havetended to conserve the polarities and volumes of thoseresidues. For the Defb2, Defb11, and Defb10 secondexons, comparisons the effect is weaker but pR/ pC is stillgreater for charge than for polarity and volume (table 2).

FIG. 3.—Peptide alignments phylogenetic tree mouse b-defensins.(A) Alignment of Mus musculus b-defensin genes from chromosome 8A3. Subgroups 1 and 2 are indicated. The splice between first and secondexons are indicated with an arrowhead, a new arrowhead indicatinga change for the sequences below. The signal region and mature peptideregions of Defb1 according to the Swissprot annotation are indicated.Residues conserved in all but one of the b-defensins are shaded in black.Residues conserved in 80% of subfamily members are shaded in grey. (B)Relationship between mouse b-defensins only showing clades supportedby bootstrap values of 70% and over. The numbers 1 and 2 refer to thesubgroups to which the genes have been assigned in the text.

FIG. 2.—Tissue expression of Defb10, Defb11, Defb15, Defb13, andDefb35. RT-PCR of the new murine b-defensin genes and Hprt (control)from various mouse tissues. All PCR products were verified by sequenceanalysis. Minus reverse transcriptase control reactions were performed inparallel for all samples and produced no PCR products (data not shown).

Murine b-Defensin Gene Evolution 463

The pI (isoelectric point) of Defb2, Defb10, andDefb11mature peptides are calculated as 8.66, 9.13, and 7.8,respectively. Thus the change in amino acid content changesthe charge quite significantly. This change in chargepresumably reflects change in function of these cationicpeptides. Only Defb2 peptide has been subjected tofunctional analysis and it has clear antibacterial function(data not shown). The pI of Defb7 mature peptide iscalculated to be 9.59. For thematureDefr1 the pI is 9.26, andfor the mature Defb5 it is 9.0. So the charge change is againevident between genes showing positive selection. Theantimicrobial activity of Defr1 peptide, despite havinga tyrosine rather than the canonical first cysteine, is stillpotent and is the onlymurine defensin described to date to beactive against Burkholderia cepacia (Morrison et al. 2002).

We aligned the introns of the genes, which showedthe evidence for positive selection (see SupplementaryMaterial). The rates of substitution between the intronsfrom Defb2 and Defb11 and between those from Defb2and Defb10 were estimated as 0.111 6 0.013 and 0.090 60.012, respectively. The rates of substitution between theintrons from Defb7 and Defb5, introns from Defb7 andDefr1, and introns from Defb5 and Defr1 were estimatedas 0.104 6 0.009, 0.240 6 0.015, and 0.237 6 0.014,respectively. These values (except for Defr1 and Defb7comparison) are not significantly different from the ratesof synonymous substitution estimated between the secondexons of the same pairs of genes given in table 1. Thus thehigh dN/dS ratio observed between these pairs of exons isnot attributable to artificially low dS estimates as a result ofsampling error, but is caused by a real increase in dNrelative to dS, which is the pattern expected to be generatedby positive selection.

Gene Alignments Reveal That the Region of AdaptiveChange Is Confined to a Small Portion of Exon 2

ClustalW lineups of Defb2, Defb10, and Defb11reveal that exon 1 only has one nucleotide change betweenthe three sequences, and the first 200 bp of the intronsequence 39 to exon 1 has 95% identity where all threesequences are present (see Supplementary Material). Thereare many examples of small deletions from the sequences,and Defb11 and Defb2 have independent deletions relativeto Defb10. All three gene have small regions ofdinucleotide expansions. Defb10 and Defb11 are recentduplications as shown by the peptide sequence identity andcladogram in figure 3. Within the intron, Defb10 andDefb11 have 98% identity where Defb10 has sequence tocompare. In the coding sequence, there are only two aminoacid changes in the first 44 residues of the peptide. This isup to the second cysteine of the mature peptide. From thatpoint, however, to the end of the peptide, the DNA identitydrops from 90% to only 30%. The amino acid identitybetween the two genes drops to only 50% in the regionthat encodes the 15 amino acids between cysteine 2 and 4in the mature peptides. The dN/dS ratio of the second exonof Defb10 and Defb11 is 1.76 but does not reachsignificance for positive selection in Fisher’s exact test.The intron sequence just 59 to exon 2, where all three

FIG. 4.—Comparison of the mean identity of distinct regions of theb-defensin genes and peptide sequences within subgroups 1 and 2.Standard deviations are shown for all values.

FIG. 5.—Synonymous versus nonsynonymous substitutions withinthe mouse subgroups 1 and 2. The number of synonymous substitutionsper nonsynonymous site (dN) plotted against number of synonymoussubstitutions per synonymous site (dS) for pairwise comparisons of geneswithin subgroups 1 and 2. A 45-degree line is drawn so a point above theline indicates dN . dS.

Table 1Estimated Distances Between the Second Exons of Paralogous Genes from Subfamilies 1 and 2

Defb7 vs. Defb5 Defb7 vs. Defr1 Defb2 vs. Defb11 Defb2 vs. Defb10

S 21.833 6 1.479 23.250 6 1.776 33.917 6 1.971 33.583 6 1.804N 74.167 6 1.467 72,750 6 1.737 119.083 6 1.985 119.417 6 1.828s 3.833 6 1.778 1.333 6 1.121 5.667 6 1.875 6.000 6 2.031n 32.167 6 4.268 20.667 6 3.847 45.000 6 5.882 48.333 6 6.721Fisher’s exact test 0.027 0.040 0.027 0.011

NOTE.—Statistically significant evidence for positive selection in Mus musculus inbred strain C57Bl/6 is shown. S indicates

synonymous sites, N indicates nonsynonymous sites, s indicates synonymous substitutions, and n indicates nonsynonymous

substitutions. Fisher’s exact test is used as the test of significance.

464 Morrison et al.

sequences align, has 85% identity. Within the first 50nucleotides of exon 2, the identity across the three genesfalls to 66%, (although Defb10 and Defb11 are still 98%identical at the nucleotide level). Thus, the high similarityof these genes in both exon 1, the intron, and 39and 59 ofthe first and second exons shows that these genes haverecently duplicated.

The ClustalW lineups of Defr1, Defb5, and Defb7 aregiven in supplementary information and again it is clearthat these genes have arisen by recent duplication. TheDNA surrounding the exons and exon 1 are 85% identicalbetween the three genes, whereas the exon 2 DNA are only54% identical.

The b-Defensin Genes Have Evolved At an AcceleratedRate Since Mouse/Rat Divergence

Southern blot hybridization experiments with mouseand rat genomic DNA indicate the existence of an excessof mouse b-defensin–like sequences over rat sequences(data not shown), indicative of a high rate of species spe-cific divergence. In order to examine b-defensin evolutionwithin Rodentia in more detail, we used PCR primersdesigned to intron sequence and specific for the murineb-defensin genes Defb1 to Defb6, Defb10, Defb11, and r1to amplify the corresponding genomic regions fromanimals of the Mus genus and the non-Mus species(table 3). A negative signal implies the gene is either notpresent or too divergent to detect using this assay. Basedon exon 2 sequence, it appears that Def b1 and Defb4 arosebefore rat/mouse divergence. The subgroup 1 and 2 genesthen arose from these as a consequence of gene duplicationin an ancestral species. The genes in subgroup 2, Defb10and Defb11 arose before Mus divergence from an ancestralspecies around 8 to10 MYA and Defb2 from a duplicationof one of these genes in the common ancestor of M.hortulanus and M. musculus. In subfamily 1, Defb3,Defb4, Defb5, and Defb6 arose by dramatic duplication ofan ancestral Defb4 gene before the divergence of M. caroliand other Mus species 4 to 6 MYA. The inability to detectgenes in more ancient species—for example, Defb10 beingpresent in M. pahari but not M. caroli—is either due totheir divergence rendering them undetectable by PCR ordue to the loss of a functional gene in a particular species.

Table 3 shows there are virtually no nucleotidevariation between orthologs of the Mus musculus species.The other species tested displayed the highest frequency ofcoding changes within the mature peptide region. Inparticular, the Defb10 and Defb11 sequence identified inM. pahari had diverged from the Defb10 and Defb11sequence present in Mus musculus by 17 and 13 nu-cleotides, respectively, leading to 13 and nine amino acidsubstitutions. Table 4 shows that the majority of aminoacid changes are located in the mature peptide regions.Full nucleotide sequences are given in supplementaryinformation.

Table 5 shows that using the same codon-based testfor selection as before, further evidence for the role ofpositive selection in the evolution of b-defensins wasfound. In this case for subfamily 2, positive selection wasobserved between the second exons of Defb2 sequenceT

able

2Estim

ated

Rates

ofRad

ical

andCon

servativeCha

nges

forNon

syno

nymou

sSu

bstitution

sBetweentheSecond

Exo

nsof

Genes

Display

ingPositiveSelection

Defb7vs.Defb5

Defb7vs.Defr1

Defb2vs.Defb11

Defb2vs.Defb10

pC

pR

pR/p

CpC

pR

pR/p

CpC

pR

pR/p

CpC

pR

pR/p

C

Charge

0.3346

0.069

0.5876

0.104

1.76

0.2696

0.080

0.3306

0.113

1.23

0.3956

0.064

0.4306

0.068

1.09

0.4656

0.080

0.3666

0.080

0.79

Miyata-Yasunaga

0.5226

0.141

0.4116

0.072

0.79

0.3116

0.104

0.2906

0.079

0.93

0.5746

0.091

0.3536

0.059

0.61

0.6276

0.112

0.3576

0.065

0.57

Murine b-Defensin Gene Evolution 465

found in the lab mouse, M. musculus and M. domesticusversus Defb11 in M. spretus, M. m. hortulanus, and M.caroli. Positive selection was also seen to be actingbetween Defb2 versus Defb10 in Mus musculus sp., M.spretus, and M. pahari. Defb2 hortulanus showed positiveselection against the same genes as Defb2 from the labmouse. Defb10 M. pahari demonstrated positive selectionagainst Defb11 Mus musculus sp. and M. spretus, inaddition to Defb10 Mus musculus sp. and Mus spretus. Insubfamily 1, positive selection was found for Defb7 Musmusculus C57Bl/6 versus Defb5 M. spretus and M. caroli.The pI values of the mature peptides are given in table 4.The changes in amino acids results in charge changes inthe mature peptides, and Defb10 M. pahari (witha calculated pI of 9.10) shows positive selection betweenDefb11 (pI of 7.8), Defb11M. spretus (pI of 8.32), and M.m. hortulanus (pI of 7.8). These charge changes aredramatic, but Defb10 M. pahari also shows positiveselection versus Defb10 Mus musculus with a very similarpI of 9.13. Conversely, Defb11 has a pI of 7.8 but does notshow evidence of positive selection versus Defb11 spretus,although the pI change is to 8.92. It is thus unclear whethercharge is the only important feature in changing functionin these peptides.

The only two known rat b-defensins (Rbd-1 and Rbd-2) were amplified by the primers specific for the mousegenes Defb1 and Defb4, respectively, supporting our beliefthat Defb1 is orthologous to Rbd-1, and Defb4 isorthologous to Rbd-2. No other rat sequences wereamplified in the PCR reactions specific for the other genes.

We recently identified and characterized a murine b-defensin–like gene named Defensin-related 1 (Defr1)because it does not contain the first of the canonical sixcysteine residues characteristic of the b-defensin family(GenBank accession number AJ344114), however it doesdisplay very potent antimicrobial activity in vitro (Morri-son et al. 2002). PCR analysis amplified Defr1 exon 2sequence from C57BL/6 genomic DNA, but no signal wasobtained from any of the other genomic samples. The lackof amplification in any other of the DNAs except C57B/6was also obtained using two other sets of primers designedtowards the Defr1 sequence. A product was also produced

from C57BL/6 DNA using primers that had 100% identityto the very similar (98% identity) Defb8 gene described byBauer et al. (2001) but contained mismatches to the Defr1sequence. However, when this product was sequenced, ithad 100% identity to Defr1 and not Defb8. We are forcedto conclude that this novel gene arose by gene duplicationvery recently within the C57Bl/6 inbred strain, afterduplication from an ancestral Defb7 gene.

Discussion

We report here the characterization of six new murineb-defensin genes identified from a BAC containing thepreviously identified b-defensin Defb2. The expressionprofile of these novel genes was surprising in that thehighest expression of two of the genes, Defb10 andDefb11, was found in adult and neonatal brain. The Defb9sequence was found in an adult hippocampal EST library,and we found Defb9 to be expressed in adult and neonatalbrain tissue but only after an increased number of cycles ofPCR, indicating its expression to be weak. It is unclearwhat functions antimicrobial peptides might have in thebrain. The human peptides DEFB1 and DEFB2 have beendetected in brain tissue and DEFB2 can be induced bylipopolysaccharide in astrocytes (Hao et al. 2001), whichsuggests the possibility that b-defensins may play a role inhost defense against bacterial central nervous systempathogenesis where infection would be catastrophic.

The high similarity these peptides share over theirfirst exons (particularly within subgroups 1 and 2), anddissimilarity in their second exons, prompted us toinvestigate the sequence similarities and disparities ofthese peptides in more detail.

A phylogenetic tree constructed using all of theknown mammalian b-defensin genes showed speciesspecific clustering (Hughes 1999). Many second exonfragments from known and novel mouse b-defensins wererecently reported by Schutte et al. (2002), but from thesedata it is impossible to tell which fragments are pseu-dogenes and which are functioning genes. Althoughwe acknowledge the value of examining pseudogenesequence in determining ‘‘birth and death’’ mechanisms of

Table 3Genomic PCR Analysis of b-Defensin Exon 2 Sequences in Rodentia

Defb1 Defb2 Defb11 Defb10 Defb3 Defb4 Defb5 Defb6 Defr1

C57Bl/6 ’’ ’’ ’’ ’’ ’’ ’’ ’’ ’’ ’’129/Ola ’’ ’’ ’’ ’’ ’’ ’’ ’’ ’’ xCBA ’’ ’’ ’’ ’’ ’’ ’’ ’’ ’’ xBalbC ’’ ’’ ’’ ’’ ’’ ’’ ’’ ’’ xM. m. domesticus ’’ ’’ ’’ ’’ ’’ ’’ ’’ ’’ xM. m. musculus ’’ ’’ ’’ ’’ ’’ ’’ ’’ ’’ xM. m. castaneus ’’ ’’ (1) ’’ ’’ ’’ ’’ ’’ ’’ xM. m. hortulanus ’’ (1) 1 (2) 2 (3) ’’ ’’ (1) 1 (3) 1 (2) ’’ xM. spretus ’’ (1) x 3 (4) ’’ ’’ (1) ’’ 1 (3) x xM. caroli 1 (1) x 3 (6) x 1 (2) 4 (4) 4 (4) 8 (9) xM. pahari 1 (3) x 9 (13) 13 (17) x x x x xA. sylvaticus 1 (5) x x x x x x x xR.norvegicus 5 (12) x x x x 14 (24) x x x

NOTE.—The number of amino acid changes detected in the exon 2 sequence are indicated with the number of nucleotide

differences shown in parentheses. The symbol ’’ indicates the predicted amino acid sequence had 100 % identity to that found in

C57Bl/6, and x indicates that no sequence was detected after PCR.

466 Morrison et al.

evolution (Nei, Gu, and Sitnikova 1997), in the presentstudy (because we are primarily interested in function), werestricted our attention to complete gene sequences. Theseare either previously known or determined from theanalysis of the BAC clone based map of the region andverified as encoding real transcripts by RNA analysis.

The existence of independent, species-specific geneclusters suggests that multiple duplications and diversifi-cation events have occurred at these loci after mammalianradiation. The increased number of mouse hybridizingfragments compared with rat under reduced stringencyconditions implies that gene duplications took place aftermouse and rat divergence approximately 40 MYA(Hedges et al. 1996). Our cross-species PCRs indicatesthat only Defb4 and Defb1 can be found in the rat. Theconstruction of a phylogenetic tree with only the Musmusculus C57Bl/6J sequences separated the genes intosubgroups, and it became apparent that within subgroups 1and 2, there were distinct regional constraints upon thedivergence of these sequences. Within subgroups 1 and 2,the genes appear to have risen by recent duplication andthe extremely high level of sequence identity in the firstexons may be due to this and possibly some level of geneconversion after duplication. However, in marked contrastto exon 1, the exon 2 sequences show rapid divergence asa result of positive selection. Although statistical signif-icance is not achieved for all of the genes showing dN .dS, only Defb1 and Defb9 show dS . dN. Intriguingly,primate b-defensin 1 orthologs (Del Pero et al. 2002), incommon with the mouse Defb1 gene, also demonstrateneutral rather than adaptive evolution.

In addition to the clear evidence we have for positiveDarwinian selection in the M. musculus b-defensin genes,

there is also evidence that the subgroup evolved relativelyrecently by a ‘‘birth and death’’ mechanism (Nei, Gu, andSitnikova 1997). Rapid ‘‘birth and death’’ and ‘‘genesorting’’ has recently been described for the rodenteosinophil-associated RNase family. ‘‘Gene sorting’’ isdefined as the process leading to differential retention ofancestral genes or gene lineages in different species, where‘‘birth’’ implies gene duplication and ‘‘death’’ refers togene deactivation (Zhang, Dyer, and Rosenberg 2000). Wepostulate that the inability to detect particular genes incertain species is as a result of this process. For example,Defb10 and Defb11 arose by duplication of an ancestralgene in an ancestor of M. pahari. Defb10 was then lostfrom M. caroli, but in other Mus species, the gene wasfunctionally important and was retained. In addition, itappears that once a gene arises by duplication, rapidchange occurs until optimal function arises and is selectedso the gene sequence is stabilized and adaptive change isrelaxed.

Hill and Hastie (1987) first demonstrated that certainregions of genes can demonstrate an excess of non-synonymous nucleotide change. Subsequent research hasshown that gene duplication and specialization by positiveDarwinian selection is a hallmark of vertebrate host-defense, although a few other genes, including thoseinvolved in reproduction (Swanson and Vacquier 2002),are also known to be under strong positive selection.Interestingly, the subgroup 1 genes Defb15, Defb13, andDefb35 are strongly expressed in the testis and epididymis,and indeed expression in male testis/epididymis is seen forall the novel genes described here. Recently, a rat geneBin1b that is exclusively expressed in the caput region ofthe rat epididymis, and which is responsible for sperm

Table 4Amino Acid Changes in the Mature Peptide Sequence of b-Defensin Rodentia Orthologs and theConsequent Change in the pI

NOTE.—The peptide sequences deduced from exon 2 PCRs are lined up to show changes from the C57Bl/6 sequence. Periods

indicate identity with the C57Bl/6 ortholog. Arrows indicate the presumed start of the mature peptides. The calculated pI (isoelectric

focusing point) of the mature peptide is given.

Murine b-Defensin Gene Evolution 467

maturation, storage, and protection, was described. Bin1bexhibits structural characteristics and antimicrobial activitysimilar to that of b-defensins. Bin1b appears to be a naturalepididymis-specific antimicrobial peptide that plays a rolein reproductive tract host defense and male fertility, andthe mouse defensins we describe may also be involved inthese processes (Li et al. 2001).

Interestingly, for functional studies of the nonsynon-ymous changes seen for subgroup 2, there has beena tendency to change the charge but not the polarity andvolume of the amino acids. However, this is not so strongfor subgroup 1, where positive selection is observedbetween Defb7 (pI 5 9.59), Defr1 (pI 5 9.26), and Defb5(pI 5 9.03). In this subgroup, the b-defensins appear toremain very cationic, but in subgroup 2, the charge variesfrom 7.8 to 9.13. Again, the signal peptide remainsvirtually unchanged between the paralogs. Inefficientsecretion may be too damaging to the organism to allowamino acid change. However, we see no evidence fora coordinated change of charge in the prepropeptide, suchas is observed with the a-defensins (Hughes and Yeager1997). The propeptide region is very small and does notbalance the change in charge observed in the maturepeptides. In the cross-species analysis, Defb11 shows veryrapid accumulation of amino acid changes by positiveselection. Independent changes occur within differentspecies. Interestingly, one amino acid (leucine), which isnonpolar, is present in the Defb11 M. pahari sequence andin M. caroli and M. m. hortulanus but is phenylalanine(also nonpolar) in M. spretus and the other strains moreclosely related to the inbred lines. This amino acid is codedfor by three different codons in the three species that havea leucine rather than a phenylalanine. The phenylalaninecodon is TTT, and the leucine codons are TTA, TTG, andCTT, respectively, in M. m. hortulanus, M. caroli, and M.pahari. We conclude from these data that this amino acidchange must be highly advantageous for functionalchange, as its change has been selected three independenttimes.

How have these genes evolved to rapidly accumulatenucleotide changes in the mature region of the peptide andto remain relatively unchanged in the signal peptide andintron? The rapid, accelerated evolution of the exon 2domain is achieved by positive selection, and the strikingsequence similarity between genes adjacent on thechromosome implies recent duplication. After duplication,both relaxation of purifying and the existence of strongpositive selection may have had complementary roles inallowing the mature peptide region of duplicate genes tofunctionally diverge, as recently postulated for theevolution of the ribonuclease gene in a leaf-eating monkey(Zhang, Zhang, and Rosenberg 2002). Our cross-speciesanalysis demonstrates that the duplications of all genesexcept Defb4 and Defb1 occurred after the divergence ofMus. Thus, the amino acid changes have occurred overa small time scale and imply a strong selective pressure.Bauer et al. (2001) recently showed that Defb7 and Defb8(Defr1 variant) mature peptides are similar in tertiarystructure, despite a high level of sequence change (58%identity). A high level of sequence change must bepermissible without loosing antimicrobial function, andT

able

5PositiveSelectionin

Exo

n2BetweenParalog

uesan

dOrtho

logs

WithinRod

entia

Subgroup1

Subgroup2

Defb7

Defb2

Defb2

M.m.ho

rtulan

usDefb1

0M.pahari

Defb5

M.spretus

38.750(1.750)4.250(1.876)0.030

——

—Defb5

M.caroli

41.417(4.700)3.583(1.754)0.003

——

—Defb11

—43.000(5.039)6.000(2.191)0.027

42.000(5.015)6.000(2.194)0.024

23.000(4.243)2.000(1.365)0.044

Defb11M.spretus

—45.000(5.068)6.000(2.191)0.012

44.000(5.047)6.000(2.195)0.015

23.000(4.242)2.000(1.365)0.044

Defb11M.m.ho

rtulan

us—

44.000(5.053)7.000(2.319)0.035

43.000(5.031)7.000(2.322)0.043

23.000(4.242)2.000(1.365)0.044

Defb11M.caroli

—44.167(5.056)6.833(2.299)0.035

43.167(5.0336.833(2.303)0.043

23.000(4.242)3.000(1.641)0.036

Defb10

—44.333(5.079)5.667(2.135)0.030

43.333(5.056)5.667(2.138)0.027

13.000(3.379)0.000(0.000)0.036

Defb10M.pa

hari

—41.333(5.009)5.667(2.143)0.040

40.333(4.982)5.667(2.145)0.037

—

NOTE.—

Pairw

isecomparisonsbetweensequences,whichgiveevidence

ofpositiveselectionusingFisher’sexacttest,areshown.Thenumber

ofnonsynonymousdifferences(w

ithstandarddeviation[SD]in

parentheses),followed

by

thenumber

ofsynonymousdifferences(SDin

parentheses),followed

bythePvalues,arelisted.Wherenospeciesnam

eisgiven,thesequence

isthesameas

thatfoundin

M.m.musculusandM.m.domesticusandin

theinbredstrains(see

table

3).Comparisonswerenotmadebetweengenes

indifferentsubgroups.

468 Morrison et al.

change is advantageous to confer pathogen specifity. Therapid evolution of these mouse-specific clades is reminis-cent of the morpheus gene family recently described byJohnson et al. (2001), which arose after chromosomalduplication and positive selection during the emergence ofhumans and African apes.

Functional differences between murine b-defensinsidentified to date have been observed (Lehrer and Ganz2002). Our own data have demonstrated a species-specificantimicrobial action of Defb1 (Morrison et al. 1998;Morrison et al. 2002). We demonstrate here that positiveselection acting on b-defensins can account for thediversity observed between the mature peptide regions ofparalogous and orthologous genes. Selection appears topromote changes in the charges of the residues encoded.This rapid adaptive evolution is in keeping with theirproposed specialized role in innate immunity diversity anda necessity to respond to pathogen diversity.

Supplementary Material

The novel genes in this paper have been submitted tothe EMBL Nucleotide Sequence Database and assignedthe following accession numbers: AJ437646 (Defb11 gene);AJ437647 (Defb9 gene); AJ437648 (Defb15 gene);AJ437649 (Defb13 gene); AJ437650 (Defb35 gene);and AJ437645 (Defb10 gene). These sequence data wereproduced by the UK MRC mouse genome sequencingprogram, funded by the UK Medical Research Council andcan be obtained from ftp://ftp.sanger.ac.uk/pub/mouse/.

Intron alignments, gene alignments, primer se-quences, and nucleotide sequences from the crossRodentia gene analysis are included in supplementary in-formation available online only at the journal Web site.

Acknowledgments

We would like to thank the UK mouse sequencingprogram. This work was supported by The Cystic FibrosisTrust UK, and the Medical Research Council, UK. Thanksto Paul Sharp for initial advice and to Nick Hastie for hissupport and encouragement.

Literature Cited

Bals, R., M. J. Goldman, and J. M. Wilson. 1998. Mouse b-defensin 1 is a salt-sensitive antimicrobial peptide present inepithelia of the lung and urogenital tract. Infect. Immun.66:1225–1232.

Bals, R., X. Wang, R. L. Meegalla, S. Wattler, D. J. Weiner, M.C. Nehls, and J. M. Wilson. 1999. Mouse b-defensin 3 is aninducible antimicrobial peptide expressed in the epithelia ofmultiple organs. Infect. Immun. 67:3542–3547.

Bauer, F., K. Schweimer, E. Kluver, J. R. Conejo-Garcia, W. G.Forssmann, P. Rosch, K. Adermann, and H. Sticht. 2001.Structure determination of human and murine b-defensinsreveals structural conservation in the absence of significantsequence similarity. Protein Sci. 10:2470–2479.

Del Pero, M., M. Boniotto, D. Zuccon, P. Cervella, A. Spano, A.Amoroso, and S. Crovella. 2002. b-defensin 1 gene variabilityamong non-human primates. Immunogenetics 53:907–913.

Ganz, T., and R. I. Lehrer. 1994. Defensins. Curr. Opin. Im-munol. 6:584–589.

Garcia, J. R., F. Jaumann, S. Schulz et al. (13 co-authors). 2001.Identification of a novel, multifunctional b-defensin (humanb-defensin 3) with specific antimicrobial activity. Its in-teraction with plasma membranes of Xenopus oocytes and theinduction of macrophage chemoattraction. Cell Tissue Res.306:257–264.

Hao, H. N., J. Zhao, G. Lotoczky, W. E. Grever, and W. D.Lyman. 2001. Induction of human b-defensin-2 expression inhuman astrocytes by lipopolysaccharide and cytokines. J.Neurochem. 77:1027–1035.

Hedges, S. B., P. H. Parker, C. G. Sibley, and S. Kumar. 1996.Continental breakup and the ordinal diversification of birdsand mammals. Nature 381:226–229.

Higgins, D. G., J. D. Thompson, and T. J. Gibson. 1996. UsingClustal for multiple sequence alignments. Methods Enzymol.266:383–402.

Hill R. E., and N. D. Hastie. 1987. Accelerated evolution in thereactive centre regions of serine protease inhibitors. Nature326:96–99.

Hughes, A. L. 1999. Evolutionary diversification of the mam-malian defensins. Cell Mol. Life Sci. 56:94–103.

Hughes, A. L., and M. Yeager. 1997. Coordinated amino acidchanges in the evolution of mammalian defensins. J. Mol.Evol. 44:675–682.

Huttner, K. M., C. A. Kozak, and C. L. Bevins. 1997. The mousegenome encodes a single homolog of the antimicrobialpeptide human b-defensin 1. FEBS Lett. 413:45–49.

Jia, H. P., S. A. Wowk, B. C. Schutte, S. K. Lee, A. Vivado, B. F.Tack, C. L. Bevins, and P. B. McCray, Jr. 2000. A novelmurine b-defensin expressed in tongue, esophagus, andtrachea. J. Biol. Chem. 275:33314–33320.

Johnson M., L. Viggiano, J. Bailey, M. Abdul-Rauf, G.Goodwin, M. Rocchi, and E. Eichler. 2001. Positive selectionof a gene family during the emergence of humans and Africanapes. Nature 413:514–519.

Jukes T. H., and C. R. Cantor. 1969. Evolution of proteinmolecules. Pp. 21–132 in H. N. Munro, ed. Mammalianprotein metabolism. Academic Press, New York.

Kimura, M. 1980. A simple method for estimating evolutionaryrates of base substitutions through comparative studies ofnucleotide sequences. J. Mol. Evol. 16:111–120.

Kumar, S., K. Tamura, I. B. Jakobsen, and M. Nei. 2001.MEGA2: molecular evolutionary genetics analysis software.Bioinformatics 17:1244–1245.

Lehrer, R. I., and T. Ganz. 2002. Defensins of vertebrate animals.Curr. Opin. Immunol. 14:96–102.

Li, P., H. Chan, B. He, S. So, Y. Chung, Q. Shang, Y. Zhang, andY. Zhang. 2001. An antimicrobial peptide found in the malereproductive system of rats. Science 291:1783–1785.

Miyata, T., S. Miyazawa, and T. Yasunaga. 1979. Two types ofamino acid substitutions in protein evolution. J. Mol. Evol.12:219–236.

Morgenstern, B. 1999. DIALIGN 2: improvement of thesegment-to-segment approach to multiple sequence alignment.Bioinformatics 15:211–218.

Morrison, G. M., D. J. Davidson, and J. R. Dorin. 1999. A novelmouse b defensin, Defb2, which is upregulated in the airwaysby lipopolysaccharide. FEBS Lett. 442:112–116.

Morrison, G. M., D. J. Davidson, F. M. Kilanowski, D. W.Borthwick, K. Crook, A. I. Maxwell, J. R. Govan, and J. R.Dorin. 1998. Mouse b defensin-1 is a functional homolog ofhuman b defensin-1. Mamm. Genome 9:453–457.

Morrison G., M. Rolfe, F. Kilanowski, S. Cross, and J. R. Dorin.2002. Identification and characterization of a novel murine b-defensin related gene. Mamm. Genome 13:445–451.

Murine b-Defensin Gene Evolution 469

Nei, M., and T. Gojobori. 1986. Simple methods for estimatingthe numbers of synonymous and nonsynonymous nucleotidesubstitutions. Mol. Biol. Evol. 3:418–426.

Nei, M., X. Gu, and T. Sitnikova. 1997. Evolution by the birth-and-death process in multigene families of the vertebrateimmune system. Proc. Natl. Acad. Sci. USA 94:7799–7806.

Saitou, N., and M. Nei. 1987. The neighbor-joining method:a new method for reconstructing phylogenetic trees. Mol.Biol. Evol. 4:406–425.

Schutte, B. C., J. P. Mitros, J. A. Bartlett, J. D. Walters, H. P. Jia,M. J. Welsh, T. L. Casavant, and P. B. McCray Jr. 2002.Discovery of five conserved b-defensin gene clusters usinga computational search strategy. Proc. Natl. Acad. Sci. USA99:2129–2133.

Swanson W., and V. Vacquier. 2002. The rapid evolution ofreproductive proteins. Nat. Rev. Genet. 3:137–144.

Townley D. J., B. Avery, B. Rosen, and W. C. Skarnes. 1997.Rapid Sequence analysis of gene trap integrations to generatea resource of insertional mutations in mice. Genome Res.7:293–298.

Yamaguchi, Y., S. Fukuhara, T. Nagase, T. Tomita, S. Hitomi, S.Kimura, H. Kurihara, and Y. Ouchi. 2001. A novel mouse b-defensin, mBD-6, predominantly expressed in skeletalmuscle. J. Biol. Chem. 276:31510–31514.

Yang, D., O. Chertov, S. N. Bykovskaia et al. (11 co-authors).1999. b-defensins: linking innate and adaptive immunitythrough dendritic and T cell CCR6. Science 286:525–528.

Zhang, J. 2000. Rates of conservative and radical nonsynon-ymous nucleotide substitutions in mammalian nuclear genes.J. Mol. Evol. 50:56–68.

Zhang, J., K. D. Dyer, and H. F. Rosenberg. 2000. Evolution ofthe rodent eosinophil-associated RNase gene family by rapidgene sorting and positive selection. Proc. Natl. Acad. Sci.USA 97:4701–4706.

Zhang, J., S. Kumar, and M. Nei. 1997. Small-sample tests ofepisodic adaptive evolution: a case study of primatelysozymes. Mol. Biol. Evol. 14:1335–1338.

Zhang, J., H. F. Rosenberg, and M. Nei. 1998. PositiveDarwinian selection after gene duplication in primate ribo-nuclease genes. Proc. Natl. Acad. Sci. USA 95:3708–3713.

Zhang J., Y. Zhang, and H. Rosenberg. 2002. Adaptive evolutionof a duplicated pancreatic ribonuclease gene in a leaf-eatingmonkey. Nat. Genet. 30:411–415.

Naruya Saitou, Associate Editor

Accepted December 3, 2002

470 Morrison et al.