Evolution, 48(6), 1994, pp. 1986-2001

A SPECIATIONAL HISTORY OF "LIVING FOSSILS": MOLECULAR EVOLUTIONARY PATTERNS IN HORSESHOE CRABS

JOHN C. AVISE', WILLIAM S. NELSON', AND HIROAKI SUGITA2 'Department of Genetics, University of Georgia, Athens, Georgia, 30602 2Institute of BiologicalSciences, University of Tsukuba, Ibaraki 305, Japan address correspondence and reprint requests to John C. Avise

Abstract.-Horseshoe crabs' exceptional morphological conservatism over the past 150 My has led to their reputation as "living fossils," but also has served to obscure phylogeneticrelationships within the complex. Here we employ nucleotide sequences from two mitochondrial genes to assess molecular evolutionary rates and patterns among all extant horseshoe crab species. The American species Limulus proved to be the sister taxon to a clade composed of the Asiatic species Tachypleus gigas, T. tridentatus, and Carcinoscorpius rotundicauda, whose relationships inter se were not resolved definitively. Both absolute and relative rate tests suggest a moderate slowdown in sequence evolution in horseshoe crabs. Nonetheless, dates of the lineage separations remain uncertain primarily because of reservations about molecular-clock calibrations resulting from large rate variances at examined loci across and other lineages,as inferred in this and prior studies. Thus, ironically, separation dates as estimated by molecular evidence in general may remain most insecure in taxonomic groups for which such information is needed most-those lacking strong biogeographic or fossil benchmarks for internal-clock calibrations. In any event, the current results show that large numbers of molecular characters distinguish even these most morphologically conservative of organisms. Furthermore, comparisons against previ­ ously published mitochondrial sequence data in the morphologically dynamic -king crab complex demonstrates that striking heterogeneity in levels of morphotypic differentiation can characterize lineages at similar magnitudes of molecular divergence.

Key words.-Mitochondrial DNA, molecular clocks, phylogeny.

Received September 14, 1993. Accepted March 30, 1994.

Evolutionary stasis, the near absence of de­ crabs (Arthropoda; class Merostomata; subclass tectable evolutionary change in some lineages Xiphosura) have been considered the "arche­ over long periods of geological time, has been types of bradytely [extremely slow evolution]" perceived as one of the central challenges for and "a classic example of arrested evolution" modem evolutionary theory (Gould and Eld­ (Fisher 1984). The fossil record indicates that redge 1977; Williams 1992). In some cases, stasis xiphosurans arose and radiated in the early to with regard to morphological appearance also middle Paleozoic, and that by the mid-Mesozoic may extend across speciation events, as in the some taxa had attained a morphological ap­ origin of sibling species. In such lineages and pearance strikingly similar to that ofpresent-day clades, what processes confine long-term mor­ species (Stermer 1955). Thus, with regard to ex­ phological differentiation within boundaries that ternal morphology, these xiphosuran lineages are far narrower than they theoretically could be, have remained remarkably unchanged over 150 given the rapid rates of short-term phenotypic My or more. As a result, the extant horseshoe change commonly observed in most species un­ crabs often are discussed as paradigm examples derartificial ornatural selection? (Williams 1992). of "living fossils" (Eldredge and Stanley 1984) Potential explanations range from molecular(e.g., or "phylogenetic relics" (Selander et al. 1970). low mutation rate; paucity of additive genetic At the molecular level, however, living horse­ variation), to ontogenetic (e.g., developmental shoe crabs appear to be unexceptional with re­ constraint or coherence producing resistance to gard to intraspecific genetic variation and pat­ selection), to ecological (e.g., long-term stability terns ofpopulation differentiation. Thus, in Li­ in environmental selection pressures). Here we mulus polyphemus, the level ofallozyme hetero­ examine patterns of molecular evolution in an zygosity (H = 0.057) proved similar to mean evolutionary clade renowned for long-term stasis estimates for many other (Selander et with regard to external morphology. al. 1970); and, levels and patterns ofintraspecific Because of a reputation for extreme conser­ differentiation in mitochondrial DNA (mtDNA) vatism in morphotypic evolution, horseshoe were similar to those of several other inverte- 1986

© 1994 The Society for the Study of Evolution. All rights reserved. PHYLOGENY OF LIVING FOSSILS 1987 brate and vertebrate species inhabiting the same Mitochondrial DNA was isolated from the coastal range in the southeastern United States fresh gill and muscle tissues ofL. polyphemus by (Saunders et al. 1986; Avise 1992). This "nor­ CsCI gradient centrifugation. For the other spe­ malcy" ofgenetic variability within and among cies, phenolic extractions ofgenomic DNA were populations ofL. polyphemus suggests that con­ employed. Mitochondrial sequences were am­ sensus patterns ofmolecular versus morphotypic plified via the polymerase chain reaction (Innis evolution in horseshoe crabs may be conspicu­ et al. 1990; Saiki et al. 1988), using primer pairs ously decoupled (but see Riska [1981] and Shus­ 16sar-L and 16sbr-H for the l6S rRNA gene, ter [1982], who documented variability and geo­ and COlf-L and COla-H for the cytochrome graphic differentiation in the morphological traits oxidase I gene (Palumbi et al. 1991). These pro­ ofL. polyphemus as well). duced fragments ofabout 525 bp and 650 bp in Actually, four extant species ofhorseshoe crabs length, respectively. Fragments were checked for are recognized: L. polyphemus in eastern North correct size on 1% agarose gels and then sepa­ America, and Carcinoscorpius rotundicauda, rated from excess primers and dNTPs with use Tachypleus tridentatus, and T. gigas in Southeast of the Magic PCR Preps system from Promega. Asia. Despite their placement in three genera, to Direct sequencing of heat-denatured, double­ the untrained eye, the differences in morphology stranded amplification products was performed are subtle indeed, consisting ofsuch distinctions either in our laboratory by dideoxy chain ter­ as whether the cross-section of the telson (tail) mination using T7 DNA polymerase and 35S ra­ is subtriangular (C rotundicauda) versus trian­ dioactive labeling (Sanger et al. 1977), or by flu­ gular (other species), and whether the number of orescent-dye sequencing conducted by the Mo­ immovable spines on the midposterior margin lecular Genetics Instrumentation Facility at the of the opisthosomatic carapace is three (T. tri­ University of Georgia. Correctness of DNA se­ dentatus) versus one (other species). As noted by quences was checked by several procedures: first, Selander et al. (1970), generic separation among various portions ofthe l6S rRNA sequence were various living and extinct horseshoe crabs "is a assayed both in our laboratory and in the DNA reflection of the morphological differences be­ sequencing facility and results compared; sec­ tween them, relative to the collection ofall forms ond, both heavy and light strands were se­ in the Xiphosura....since taxonomic lines tend quenced from each individual; third, two indi­ to be drawn relative to the range of variation viduals (A and B) from each species ofhorseshoe specific to the taxon underconsideration." In any crab were sequenced independently. event, one consequence of the general morpho­ The l6S rRNA gene sequences were aligned logical conservatism ofhorseshoe crabs over the using the computer program GeneWorks 2.1.1 past 150 My is that considerable uncertainty ex­ (Intelligenetics), with assigned penalties of 10 and ists overthe evolutionary histories ofthe lineages 4 for opening and extending a gap, respectively. leading to the extant forms. This irresolution of Cytochrome oxidase alignments were unambig­ phylogeny as a result of extreme morphological uous and done by eye. Estimates of genetic di­ stasis, and questions concerning the magnitude vergence were calculated as direct counts ofnu­ ofmolecular differentiation in a clade renowned cleotide sequence differences, and by the "two­ for slow morphological evolution, prompted the parameter" method of correction for multiple current assessment ofmolecular-level evolution­ substitutions at a site (Kimura 1980). Distance ary patterns among all extant horseshoe crab spe­ matrices were clustered by the unweighted pair­ cies. group method with arithmetic means (UPGMA; Sneath and Sokal 1973) and by the neighbor­ MATERIALS AND METHODS joining procedure (Saitou and Nei 1987). Se­ Specimens of Tachypleus gigas and Carcinos­ quences also were analyzed by maximum par­ corpius rotundicauda were collected in the vicin­ simony methods as applied to information coded ity of Bangsaen (Gulf of Siam), Thailand. Spec­ as: (1) nucleotide sequences themselves; (2) pu­ imens of T. tridentatus came from the Bay of rines versus pyrimidines (such that only trans­ Hakata, Fukuoka, Japan. As elaborated below, versions were considered); and (3) for the cyto­ the two samples of L. polyphemus came from chrome oxidase locus, first and second positions genetically distinctive populations along the At­ ofcodons only (such that silent substitutions at lantic and GulfofMexico coastlines in the south­ third positions were disregarded). The latter two eastern United States. approaches permit focus on conservative char- 1988 JOHN C. AVISE ET AL. acter-state changes most likely to be phyloge­ representation ofG at the third positions of co­ netically informative at deeper evolutionary lev­ dons, a phenomenon noted previously for meta­ els. Both exhaustive and branch-and-bound op­ zoan mtDNA in general (see Kornegay et al. tions were employed in the parsimony searches, 1993). Indeed, base frequencies at the various and bootstraps were conducted across batches of codon positions in horseshoe crabs, as well as 100-1000 replicates. The distance-based and the magnitudes ofbias in base composition, are parsimony analyses were performed using com­ remarkably similar to those previously reported, puter programs PHYLIP (Felsenstein 1991) and for example, at the cytochrome b gene in mam­ PAUP (Swofford 1993), respectively. malian species (table 3; Irwin et al. 1991). In the For outgroup, we generated and employed an 16S rRNA gene sequence, there was a noticeable homologous 16S rRNA gene sequence from the shortage ofCs, but this pattern again was shared scorpion Vejovis carolinian us (the cytochrome by all assayed taxa (and also by the king crabs oxidase primers did not successfully amplify from and hermit crabs previously assayed) (table 3). this species). We also used a GenBank 16SrRNA These observations suggest that the comparative gene sequence from the Artemia evolutionary patterns and phylogenetic conclu­ salina. The scorpion is an arachnid placed in sions drawn in thecurrent study will notbe gross­ Chelicerata, the subphylum to which the mer­ ly affected by differential patterns of base-com­ ostomatidhorseshoe crabs belong (Barnes 1963), positional bias across the surveyed species. whereas the brine shrimp is a con­ ventionally classified in the Mandibulata. Phylogeny Mitochondrial genes were employed because 168Ribosomal RNA Locus.-At the intraspe­ explicit evolutionary rate calibrations for partic­ cific level, no sequence differences were observed ular mtDNA loci recently have appeared (Lynch between the specimens ofeither Tachypleus tri­ 1993), and because comparative data on 16S dentatus or Carcinoscorpius rotundicauda; a sin­ rRNA gene sequences are available for another gle transition distinguished the two specimens of group of marine crabs (king and hermit crabs; T. gigas; and six substitutions (three transitions Crustacea: Anomura) that in striking contrast to and three transversions; genetic distance 0.013) the horseshoe crabs has undergone exceptionally distinguished Limuluspolyphemus A from B. The rapid morphological evolution (Cunningham et two specimens ofL. polyphemus came from ge­ al. 1992; see Discussion). Sequence analyses were netically distinctive Atlantic and GulfofMexico employed because mtDNA restriction digestion populations in the southeastern United States, profiles proved too divergent to permit mean­ which previously were estimated by mtDNA re­ ingful phylogenetic comparisons among the striction site analyses to differ at a mean genetic horseshoe crab species (Sugawara et al. 1988), distance ofabout 0.02 (Saunders et al. 1986). and because in general sequence analyses reveal Sequence identities (one minus sequence dif­ finer details about molecular-level changes. ferences) at the 16S gene ranged from 82%-100% among the various horseshoe crabs (table 4), and REsULTS thus, fell within the range of similarities (70%­ Newly obtained sequences for the 16S rRNA 100%) that Hillis and Dixon (1991) recommend locus and the cytochrome oxidase I locus are as optimal for phylogenetic analyses based on presented in tables 1 and 2, respectively. At the rRNA loci. However, mean sequence identities 16S rRNA gene, more than 480 nucleotide po­ with the outgroup taxa were only 0.64 and 0.65 sitions were sequenced per individual, ofwhich and are thus in a range where phylogenetic anal­ slightly more than 200 were variable among the yses normally become somewhat ambiguous be­ nine newly assayed specimens. At the cyto­ cause ofalignment difficulties and increased like­ chrome oxidase locus, a total of 582 nucleotide lihoods of multiple substitutions at a site. In positions perindividual was scored, ofwhich 135 comparisons among the various horseshoe crab were variable in the survey. Most ofthese latter species, transition: transversion ratios ranged substitutions were synonymous, with only 15 from about 4: 1 to 1:1; use ofa 3:1 ratio produced (11%) ofthe variable positions producing amino a genetic distance matrix summarized in table 4. acid substitutions in the encoded protein. In an exhaustive parsimony search based on Patterns of base-compositional bias are sum­ the 16S data set (includingArtemia and Vejovis), marized in table 3. At the cytochrome oxidase two minimum-length networks of equal total gene, most noteworthy is a pronounced under- length (358 mutational steps) were identified. The PHYLOGENY OF LIVING FOSSILS 1989

165 cyt. ox. :) T. gigas A. ) Tschypleus glgss (:2JOO c. rotundicauda TlIchypltlus :) A. ) (: 100 .010 Irldsntatus .054 A 7' 98 A) Csrclnoscorplus .125 T. tridentatus B rotundfc8urlll ( B 100 :) .012 100 A) Llmulus B polyphemus (: :) L. polyphemus Artemis salina .012 , , VB/ovls caro/ln/anus I' I ,, I 0.20 0.10 0.00 FIG. I. Parsimony networks for horseshoe crabs based on exhaustive searches oftransversional substitutions Genetic distance only. Left, l6S rRNA locus: minimum length phylog­ FIG. 2. UPGMA dendrogram for horseshoe crabs eny (180 steps total), with Artemia salina (brine shrimp) based on cytochrome oxidase I gene sequences from and Vejoviscarolinianus (scorpion) as outgroups. Right, mtDNA. Thegenetic distance scale (method ofKimura cytochrome oxidase locus: strict consensus oftwo min­ 1980) at the bottom of the figure is drawn to indicate imum-length networks (58 steps each), one of which mean clustering levels (average sum ofbranch lengths grouped Tachypleus gigas with T. tridentatus as sister connectingrelevant nodesand extanttaxa). Also shown taxa, and the otherlinked T. gigas with Carcinoscorpius on the tree are individual branch lengths from a neigh­ rotundicauda. Also shown are levels ofstatistical sup­ bor-joining analysis that produced the same branching port based on 1000 bootstrap replicates. structure as the UPGMA dendrogram. strict consensus of these shows the American ranged from 83% to 95%, and ratios of transi­ species (L. polyphemus) as the sister taxon to a tion: transversion ranged from about 3:1 to 1:1 clade composedofthe three Asiatic species, whose (values ofl3:1 and 1:1 characterized the two sets relationships to one another remained unre­ of conspecific samples that were polymorphic, solved. In most other analyses, including con­ but numbers of variable sites were small). Use sensus parsimony searches as applied to trans­ ofa 3:1 ratio yielded the genetic distance matrix versional substitutions only (fig. 1),and UPGMA in table 4. and neighbor-joining methods as applied to the In an exhaustive parsimony search ofthe total distance matrices (see fig. 4, presented later), a cytochrome oxidase data, one minimum-length branching structure with T. tridentatus as a sister network (total 172 steps, seven fewer than its taxon to T. gigas was favored slightly. nearest competitor) was identified. The structure Uncorrected genetic distances of horseshoe ofthis network is identical to that shown on the crabs to Artemia and to Vejovis appeared both right side of figure 1, except that the "trichoto­ large, and nearly identical (table 4). Thus, the my" was resolved in favor of a sister-group re­ 168 sequences ofhorseshoe crabs may have ap­ lationship between T. gigas and C. rotundicauda. proached saturation with respect to base substi­ Parsimony analyses of first and second codon tutional differences from the outgroups, and for positions only, as well as UPGMA and neighbor­ this reason, we prefer to reserve judgment about joining analyses of these data (fig. 2), also pro­ the position of horseshoe crabs within Arthro­ duced a network of this latter structure. How­ poda from ourdata. The arrangement ofLimulus ever, parsimony analyses based solely on trans­ polyphemus within the broader radiation ofAr­ versions left unresolved the exact branching or­ thropoda (and otherinvertebrate phyla) has been der between T. gigas, T. tridentatus, and C. ro­ addressed elsewhere using nucleotide sequences tundicauda (fig. 1). from conservative nuclear rRNA loci (Field et Combined Data Sets. - The 168 and cyto­ al. 1988; Turbeville et al. 1991). chrome oxidase data also were combined for fur­ Cytochrome Oxidase Locus. -No nucleotide ther parsimony analyses. When all data were sequence differences were observed between the considered, the aforementioned "trichotomy" two specimens ofT. tridentatus, nor between the among Asiatic species was resolved in favor of two T. gigas, whereas 14 substitutions each (ge­ aT. gigas-Co rotundicauda clade (with bootstrap netic distance 0.024) distinguished the conspe­ support 100%). However, based on transversions cific samples of C. rotundicauda, and L. poly­ alone, T. gigas grouped with T. tridentatus (at phemus. Between species, sequence identities bootstrap level 76%). TABLE 1. Aligned sequences in horseshoe crabs of the mitochondrial gene encoding 16S rRNA. Dots indicate nucleotide identity to Tachypleus tridentatus (top row), and dashes indicate gaps. The gaps at positions 300 and 394 arose because previously published sequences from an outgroup, Artemia salina (Cunningham et al. 1992) were also included in the alignment.

2 4 6 8 1 1 1 1 1 T. tridentatus A, B aaaaatetgatetgeeeaat gatgaaaaattaaatggeeg eggtatattgaeegtgegaa ggtageataateatttgett tttaattgaaggetggaatg T. gigas A ...... g . ... aggt . . g a . T. gigas B ...... g . ... aggt . . g a . C. rotundicauda A, B ...... a . ... t .. g g . . g . L. polyphemus A ...... ga.e . · . g .. g-- t . . g a. t . L. polyphemus B ...... g .. e . .. g .. g-- t . . g a. t . (V. carolinianus) · .e.g .. aage .. tgt e ae.aegg--- a . · ta a .. a.. . te ...... g.a .. a .

1 1 1 1 1 o 2 4 6 8 1 1 1 1 1 T. tridentatus A, B aatggttggaegaaaaaa-g tetgtettagttttagtttt taaagtttaetttttagtga aaaggetaaaattttaetaa gggaegagaagaeeetatta T. gigas A ... . a g -...... a aa . .... t ...... te .. T. gigas B ... . a g - ...... a aa . .... t ...... te .. C. rotundicauda A, B · ... a .. t g- g...... a ag . · t . g g te .. L. polyphemus A ...... t aa · g.aaa a . .g .. t t . . .. a g . L. polyphemus B ...... t aa ...... g.aaa a . .g .. t t . . .. a g . (V. carolinianus) · .a .... a ... -t. tgt .. t .tg taa.a.ta . gg.. a ta a . .. . a aagtt . a.aega e.

2 2 2 2 2 o 2 4 6 8 1 1 1 1 1 T. tridentatus A, B agetttatttttgt-ta-tt tt------ttattatattt tteattataaaaaataagag gaattttaetggggeggtag agaaagaaaaaaatetttt- T. gigas A ...... -. tg .. ·.------g . ... g gga .. . g.g - T. gigas B ...... -. tg .. ·. ------g . ... g gga .. . g.g - C. rotundicauda A, B ...... a.tatt .. . g------g.t g . · .at e gt.a a ...... g g - L. polyphemus A · e t. ggtt .. ·.------at . eggt.ggg gge - .. g ...... ttt - L. polyphemus B · c t. ggtt .. .'. ------at . .ggt.ggg gge - .. ga . .." ..... ett a.- (V. carolinianus) · .t eaa.aa.tate .. a.ttteaaa g.. ag . a.tg .. tggetgggge .. ea a.taaa .. t.ttttatt.ta .t.ttatgg .. t .. ggeee- TABLE 1. Continued.

3 3 3 3 3 o 2 4 6 8 1 1 1 1 1 T. tridentatus A, B ---tttagaagttaaaaeat ttatgtaaatttttgateea tttttaatgataataagaaa gagttaetatagggataaea gegtaatttttte-ggagag T. gigas A --- a ...... a g g T. gigas B --- a ...... a g g a . C. rotundicauda A, B --- a .. a ...... t a a g a . L. polyphemus A --- tt.taa t .. t. a.t g e c g .. g g .. g. a g .. L. polyphemus B --- tt.taa t .. t. a. t g e. .. c g .. g g .. gt a g -. (V. carolinian us) --- .. g.a .. a.e a .. . ·t.------g g a .. aa g.t gg a t-a .

4 4 4 4 4 o 2 4 6 8 1 1 1 1 1 T. tridentatus A, B tteatateggegaagaagtt tgeg-aeetegatgttgaat taaagagteaatagggegga gaagttetaaatgagggtet gtte T. gigas A ... ga . T. gigas B ... ga . C. rotundicauda A, B ...... g .. a ...... t .. t . L. polyphemus A ... t ...... g ...... tg .. ga . L. polyphemus B ... t - ...... g ...... tg .. ga . (V. carolinian us) ga.t.e .. ae.eeee a t.t e g .. egg g .. ga------TABLE 2. Aligned sequences of the cytochrome oxidase gene in horseshoe crabs. Dots indicate nucleotide identity to Tachypleus tridentatus (top row).

3 6 1 0 0 T. tridentatus A, B ggo oao oot gaa gtt tat att tta att oto ooa gga ttt gga ata att tot oat att att ago oao oaa aoa gga aaa aaa T. gigas A, B .. t ...... o. t .. 0 .• t ...... · . t C. rotundicauda A .. a ...... 0 .•. ••• 0 .• .• 0 .•...... 0 ...... · . t C. rotundicauda B .. a ...... 0 ...... 0 .• .• 0 .•...... 0 •.• . .. · . t L. polyphemus A .. g .. t ...... 0 .. 0 ...... t .. t ...... g ... .. 0 ...... " ...... · .g L. polyphemus B .. g ...... •. 0 .• 0 ...... g.. .. t .. t .. g ...... 0 ...... " ...... · .g

1 1 9 2 5 0 0 0 T. tridentatus A, B gaa oot ttg gga aot ota gga ata att tao got ata tta goo att gga att tta gga ttt ata gtt tga got oat oat ata T. gigas A, B ... .• 0 •• t ...... t ...... t ...... 0 .. .. a ...... • •• 0 •• ...... g.. . .. " . ..0 .. 0 C. rotundicauda A .. . .• 0 .• t ... t .. a ...... t ...... t ...... • 0 .• ...... g.. ... " . ..0 •. 0 .• 0 C. rotundicauda B .. . .. 0 .. t ...... t ...... t ...... a ...... • •• 0 •• ...... g.. . .. " . ..0 •• 0 •• 0 L. polyphemus A ...... 0 .. g ... .. t ...... 0 ...... t ... .. t ... o. t .. t ...... a '" . .. •• 0 L. polyphemus B ...... • 0 .• g ... .. t ...... t .. 0 .•...... t ... ..0 ... o. t .. t ...... a '" . .. . .0

1 2 2 8 1 4 0 0 0 T. tridentatus A, B ttt aoa gta ggo ata gat gtt gat aoa oga got tao ttt aot goa got aot ata att ate got gtt ooa aoo ggt att aaa T. gigas A, B •. 0 ...... t .. 0 .• a ...... 0 .. t ... ..a '" .. t C. rotundicauda A •. 0 •.• ...... •. 0 .. t .. 0 .• a ...... •• 0 •.• ..0 .. t ... · .0 C. rotundicauda B .. 0 •.• ...... •• 0 •. t .. 0 .. a ...... •. 0 •.• ..0 •. t ... · .0 L. polyphemus A ...... g .. a ... •• 0 •• a .. 0 '" .. . ..a ... •• 0 •• a ...... a ... .. 0 .. t ...... 0 •. a .. a L. polyphemus B ...... g .. a ... .. 0 .• a .. 0 '" .. . .. a ... .. 0 •• a ...... a ... .. 0 .. t ...... 0 •• a .. a

2 3 7 0 0 0 T. tridentatus A, B att ttt aga tga ota got aot tta oat ggo tot oaa ato tot tat gaa oot oot tta tta tga goo tta gga ttt gta ttt T. gigas A, B .• 0 ••...... 0 ...... t ...... o. t ...... · . t '" ...... g . · .0 C. rotundicauda A .. 0 .• 0 .•...... • •. 0 ...... a o. t ...... " ...... · .0 C. rotundicauda B .• 0 .• 0 ...... a 0'0 .0. ooa 00' 0.0 oot 00. o.. 000 0.0 ooa c . too. . .. oot 00 • 0.0 o.. .. . 0.0

L. polyphemus A 00000' ...... 0 000 000 000 • 0 • ooa ooa 00. 000 ooa 000 000 ooa 00. c i t 000 ..0 ooto oo oog 00 • oot 000 L. polyphemus B 00000' .. . 0.0 o.. .. . 000000 ... ooa ooa . 00 000 ooa 000 000 ooa '00 Oot 000 . .. 000000 oog 00' oot 000 TABLE 2. Continued.

3 3 3 3 6 9 0 0 0 T. tridentatus A, B cta ttt act att gga gga tta aca ggt gtt gtc cta gca aac tct tcc att gac att att ctt cat gac aca tat tac gtt T. gigas A, B t ...... a ... · . t t .. .. c ... .. c .. t ...... c ...... · . t C. rotundicauda A t .. .. c ...... · . t t .. .. c ... . . c .. t ...... · .a C. rotundicauda B t .. .. c ...... · . t t .. ..c ... . . c .. t ...... · .a L. polyphemus A t ...... a ...... c .. a .. a a ...... t ... .. c ...... t ... ..c .. c ...... c L. polyphemus B t ...... a .. c ...... t .. a .. a a ...... t ... .. c ...... t ... .. c .. c ...... c ... . . c

4 4 4 2 5 8 0 0 0 T. tridentatus A, B gta gca cat ttt cat tac gtt tta tca ata gga .gca gta ttt gca att tta gca gga gtt acc cat tga ttt cca ctc ttt T. gigas A, B .. . .. c ...... c .. t .. c c ...... a ... .. c ...... t C. rotundicauda A .. g ...... c .. c .. t ... c ...... t ...... c ...... a C. rotundicauda B .. g .. c ... .. c .. c .. t .. cc ...... t ...... c ...... a L. polyphemus A .. . .. c ...... c .. t .. cc.C ...... c ...... g .. c .. t ...... t.a .. c L. polyphemus B .. . .. c ...... c .. t .. c c.c ...... c ...... g .. c .. t ...... t.a .. c

5 5 1 4 0 0 T. tridentatus A, B ttt gga act tct ata aat ccc aaa tga cta aaa att cat ttt cta att ata ttt att gga gta aat acc act ttt ttc cct T. gigas A, B ...... c ...... t ...... ac ...... t .. c ... · .c C. rotundicauda A ...... a .. c ...... c ...... c ... g ...... t .. c ... · .c C. rotundicauda B ...... c .. c ...... c...... c ... ac ...... t .. t .. c ... · .c L. polyphemus A .. . .. g . ta g.a ... .. c ...... c ... t .. g ...... g ... .. c gt ...... c L. polyphemus B ...... ta g.a ... .. c .. a ...... c ... t .. g ...... c gt ...... c

5 7 0 T. tridentatus A, B caa cac ttt tta ggt T. gigas A, B ... .. t ... c .. .. a C. rotundicauda A ... .. t C. rotundicauda B .. . .. t ... c .. L. polyphemus A .. . .. t .. c c .. L. polyphemus B .. . .. t .. c c .. TABLE 3. Base composition (percentage) at the first, second, and third positions of codons in the cytochrome oxidase gene, and in the overall sequence at the 168 rRNA gene.

First Second Third Species G AT C G A TC G A TC Cytochrome oxidase locus Tachypleus tridentatus 28.9 28.9 24.7 17.5 14.4 19.1 43.3 23.2 0.5 40.7 44.8 13.9 Tachypleus gigas 29.4 28.9 23.7 18.0 14.9 19.1 42.3 23.7 0.0 41.2 43.3 15.5 Carcinoscorpius rotundicauda 29.9 28.4 23.7 18.0 14.9 19.1 43.3 22.7 0.5 41.2 37.1 21.1 Limulus polyphemus 27.8 29.0 27.3 15.0 14.4 19.1 44.3 22.2 4.6 38.7 31.4 25.2 Mean 29.0 29.0 24.9 17.1 14.6 19.1 43.3 23.0 1.4 40.5 39.2 18.9 Bias· 0.107 0.244 0.396 (mammals, cyt b)t Mean 21.6 29.4 22.7 26.3 13.6 20.2 41.6 24.7 3.6 42.7 16.3 37.4 Bias· 0.076 0.221 0.401

G AT C 168 rRNA locus Tachypleus tridentatus 21.0 33.8 34.7 10.5 Tachypleus gigas 22.2 32.7 34.6 10.5 Carcinoscorpius rotundicauda 21.3 33.2 35.3 10.2 Limulus polyphemus 24.2 29.1 35.3 11.3 Vejovis carolinianus 18.6 36.3 31.8 13.2 Mean 21.5 33.0 34.3 11.1 Bias· 0.231 (king and hermit crabs):\: Mean 17.0 37.4 34.3 11.3 Bias· 0.289

* Bias in base composition, calculated as C = '/3 ~ 1ci - 0.251, where Ci is the frequency of the ith base. t Shown, for comparison, are mean values for the mitochondrial cytochrome b gene in 20 mammalian species (from Irwin et al. 1991). *Shown, for comparison, are mean values for the 16S rRNA gene in ten species of king crabs and hermit crabs (from data in Cunningham et al. 1992). TABLE 4. Genetic distances (top line, uncorrected for multiple hits; second line, Kimura's two-parameter correction) at the 16S rRNA (above diagonal) and cytochrome oxidase I (below diagonal) gene sequences for horseshoe crabs and outgroups. In parentheses are observed numbers of transversional differences.

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (I) Tachypleus tridentatus A 0.000 0.000 0.062 0.064 0.092 0.092 0.140 0.142 0.356 0.344 0.000 0.065 0.067 0.100 0.100 0.163 0.167 0.663 0.486 (0) (6) (6) (17) (17) (34) (37) (86) (69) (2) T. tridentatus B 0.000 0.000 0.062 0.064 0.092 0.092 0.140 0.142 0.356 0.344 0.000 0.065 0.067 0.100 0.100 0.163 0.167 0.663 0.486 (0) (6) (6) (17) (17) (34) (37) (86) (69) (3) T. gigas A 0.101 0.101 0.000 0.002 0.088 0.088 0.167 0.167 0.351 0.354 0.110 0.110 0.002 0.095 0.095 0.199 0.200 0.613 0.475 (14) (14) (0) (16) (16) (34) (37) (86) (67) (4) T. gigas B 0.101 0.101 0.000 0.000 0.085 0.085 0.165 0.165 0.349 0.352 0.110 0.110 0.000 0.093 0.093 0.196 0.198 0.607 0.470 (14) (14) (0) (16) (16) (34) (37) (86) (67) (5) Carcinoscorpius rotundicauda A 0.100 0.100 0.067 0.067 0.000 0.000 0.176 0.180 0.364 0.328 0.108 0.108 0.071 0.071 0.000 0.213 0.220 0.656 0.456 (14) (14) (16) (16) (0) (40) (43) (90) (66) (6) C. rotundicauda B 0.113 0.113 0.053 0.053 0.024 0.000 0.176 0.180 0.364 0.328 0.125 0.125 0.056 0.056 0.024 0.219 0.220 0.656 0.456 (16) (16) (14) (14) (8) (40) (43) (90) (66) (7) Limulus polyphemus A 0.163 0.163 0.167 0.167 0.170 0.167 0.000 0.013 0.354 0.367 0.192 0.192 0.197 0.197 0.200 0.196 0.013 0.657 0.572 (35) (35) (37) (37) (35) (35) (3) (96) (77) (8) L. polyphemus B 0.163 0.163 0.167 0.167 0.167 0.168 0.024 0.000 0.359 0.359 0.192 0.192 0.197 0.197 0.196 0.199 0.024 0.660 0.552 (36) (36) (38) (38) (36) (36) (I) (97) (74) (9) Vejovis carolinianus ------0.000 0.427 0.897 (96) (10) Artemia salina ------0.000 1996 JOHN C. AVISE ET AL.

DISCUSSION Nei 1988; Tajima 1983; Takahata 1989). Under Cladogenetic Branching Order this line of argumentation, weak statistical sup­ port in thegenealogies for particularwell-assayed The mtDNA sequence data are consistentwith loci, as well as "inconsistencies" across indepen­ the view that living horseshoe crabs constitute a dent gene genealogies, themselves become prima rather closely knit assemblage relative to the out­ facie criteria for recognizing approximate mul­ groups, and that the North American species (Li­ tichotomies in an organismal phylogeny. Both of muluspolyphemus) is the sistertaxon to the three these criteriawould appearto apply to theAsiatic Asian species. These results agree with previous horseshoe crabs. conclusions based on serological studies (Shuster In the case of the mtDNA data (and perhaps 1962), amino acid sequence analyses of a fibri­ unlike the situation for morphological charac­ nopeptide-like protein (Shishikura et al. 1982), ters), failure to resolve the phylogenetic branch­ immunological comparisons of hemocyanins ing order ofthe Asiatic species cannot be attrib­ (Sugita 1988), two-dimensional electrophoresis uted solely to a paucity ofvariable characters for ofgeneral proteins (Miyazaki et al. 1987), results analysis. In the combined 16S and cytochrome of interspecific hybridization experiments (Sek­ oxidase data sets for all horseshoe crabs assayed, iguchi and Sugita 1980), and cladistic appraisals more than 230 nucleotide positions exhibited ofmorphological characters (Fisher 1984). These variation, and more than 130 varied within the conclusions also are in accord with current tax­ Asiatic clade alone. Thus, the lack of phyloge­ onomy, which places Limulus in the subfamily netic resolution appears attributable more to Limulinae, the three other extant horseshoe crabs evolutionary "noise" at the level of individual in Tachypleinae, and all four species of Limu­ nucleotide positions (i.e., homoplasy and/or re­ lidae as sole living representatives of the Mer­ tentions of plesiomorphic characters) than to a ostomata (Yamasaki 1988). lack of molecular genetic diversity. Within the clade composed ofthe Asiatic spe­ cies, a conservative interpretation is that the Molecular Rates and Nodal Dates mtDNA data leave unresolved the phylogenetic Apart from cladistic assessments per se, what branching order for Tachypleus gigas, T. triden­ are the approximate dates of the speciational tatus, and Carcinoscorpius rotundicauda. Con­ nodes to which the extant horseshoe crab lin­ flicts between networks based on alternative eages trace? Morphological and biogeographic mtDNA data bases, and between alternative pro­ evidence have provided only the crudest ofclues. cedures of data analysis, suggest that the three According to Fisher (1984), "there is no satis­ Asiatic species probably stem from two lineage factory control on the age of the most recent bifurcations relatively close in time. This con­ common ancestorofanyofthethreeIndo-Pacific clusion also can be interpreted as consistent with species," and, thus, most speculation has cen­ other lines of evidence: cladistic assessments of tered on possible separation dates between the morphological traits tend to link T. gigas with lineages leading to extant Limulinae (North T. tridentatus (Fisher 1984), but immunological America) versus Tachypleinae (Asia). From pro­ comparisons and analyses of fibrinopeptide se­ visional generic assignments offossils ofknown quences suggest that C. rotundicauda and T. tri­ age (in particular, theextinct species"Tachypleus dentatus form a clade. In theory, when phylo­ decheni" and "Limulus coffini"), and from bio­ genetic nodes are spaced closely in evolutionary geographic arguments involving the opening of time (relative to the effective size of the popu­ the North Atlantic Ocean, rough dates ofabout lations that traversed these speciation events), 75 Mya and 90 Mya, respectively, have been the lineages of independent loci and the traits postulated for the Limulinae-Tachypleinae sep­ they encode may truly sort into descendant taxa aration (Fisher 1984). In studies of fibrinopep­ in such a way as to lead to apparentdiscordancies tideaminoacid sequences, Shishikuraet al. (1982) in structure among separate gene genealogies postulated that the Limulinae-Tachypleinae split (such as those providedby unlinked nucleargenes, occurredabout 135 Mya, butthis appears to have or between nuclear genes and mtDNA), as well been based more on fossil and/or biogeographic as to occasional discrepancies between the "con­ inference than on independent time appraisals sensus" organismal phylogeny and the structure from the molecular data themselves. ofany single gene tree (such as that provided by From analyses of published DNA sequences mtDNA) (Neigel and Avise 1986; Pamilo and as interpreted against fossil evidence or biogeo- PHYLOGENY OF LIVING FOSSILS 1997

TABLE 5. Separation dates among horseshoe crab lineages as estimated from previous clock calibrations for the mitochondrial 16S rRNA and cytochrome oxidase I loci.*

Millions of years ago, from clock calibrations Cunningham et al. (1992)t Lynch (1993):1:

Node leading to 16S rRNA (95%CL) 16S rRNA cyt, ox. I Tachypleus gigas vs. T. tridentata 17 (6-28) o o Carcinoscorpius vs. Tachypleus 27 (12-41) o 15 Limulinae vs. Tachypleinae 52 (27-77) 46 59 * If rates of mtDNA evolution in horseshoe crabs are lower than these conventional estimates (as is suggested provisionally by both absolute and relative rate comparisons-see text), then divergence times would have to be adjusted backward (to older dates) accordingly. t Confidence intervals suggested by Cunningham (pers. comm. 1994) based on a correction for nonindependence of pairwise distances for king crab-hermitcrab species, and are slightly wider than the uncorrected limits implied by the original Cunningham et al. (1992) treatment. :I: Based on equation (13) in Lynch (1993), using the following parameters as suggested by the author: for nucleotide sequences at the 16S rRNA locus, I~ (asymptotic genetic identity as time approaches infinity) = 0.30, a(probability ofa substitution per site per billion years) = 0.44, and Ho (nucleotide diversity in the ancestral population) = 0.13; for amino acid sequences at the cytochrome oxidase locus, I~ = 0.08, a= 0.30, and Ho = 0.022. graphically inferred dates of evolutionary sepa­ graphic evidence noted above. Does this imply ration in other animal groups, two independent that mtDNA sequence evolution in horseshoe clock calibrations recently have been proposed crabs has been slower than underthe calibrations for the mtDNA loci surveyed in this report. An ofLynch and Cunningham et al.? In the absence empirical rate estimate for the l6S rRNA gene of reliable independent information (fossil or for (including king crabs and hermit biogeographic) on separation dates for horseshoe crabs) was proposed by Cunningham et al. (1992), crab lineages, an alternative approach involves and calibrations for both the l6S rRNA and cy­ relative rate tests against outgroups. These eval­ tochrome oxidase I loci across a much broader uations too pose several difficulties, stemming diversity ofvertebrate and invertebrate taxa were from: (1) absence of clear independent knowl­ proposed by Lynch (1993). These calibrations as edge about which Arthropod or othertaxa would applied to the horseshoe crab data suggest that constitute an appropriate and usable outgroup; the three Asiatic species separated from one an­ (2) the possibility that phylogenetic connections other between 0 and 30 Mya, and that their most of any outgroups to horseshoe crabs might be recent shared ancestor with L. polyphemus dates quite ancient relative to separation dates among to approximately 45-60 Mya (table 5). ingroup members; and (3) the likelihood that the Some points should be addressed about these dynamics by which mutational differences ac­ estimates. First, within the Asiatic clade, species cumulate at mtDNA (or other) loci could be divergence times under the Lynch (1993) ap­ strongly nonlinear over the potentially long ti­ proach are somewhat more recent than those de­ mescales involved (caused by, e.g., mutational rived from Cunningham et al.'s (1992) calibra­ saturation at potentially variable nucleotide po­ tion. One factor contributing to this outcome sitions, or to rate heterogeneities along extended involves Lynch's correction for mean baseline branches). With these caveats in mind, provi­ nucleotide diversity ("intraspecific" variationjust sional tests of relative rate nonetheless can be prior to the time that populations would be rec­ conducted with available data. ognizable as species). This adjustment accounts Assume that current correctly re­ for why some of the divergence-time estimates flects phylogeny, such that the scorpion Vejovis are 0 Mya underthe Lynch method. A take-home truly is a sister taxon to the xiphosurans (relative message is that divergence times inferred for re­ to the brine shrimp), and that Anemia is there­ cently separated species can be highly sensitive fore an appropriate outgroup. Then, following to levels of genetic variability assumed for the the logic offigure 3, evolution ofthe 16S rRNA ancestral taxa. gene in the Vejovis lineage appears to have been A second point is that the molecular-based about four times greater, on average, than that estimates ofdivergence time for the Limulinae­ leading to the extant horseshoe crabs. Alterna­ Tachypleinae split are somewhat more recent tively, if Vejovis is assumed to be an outgroup than suspected from the fossil and/or biogeo- to a sister-taxon clade composed ofArtemia and 1998 JOHN C. AVISE ET AL.

Xiphosura Vejovis Artemia Xiphosura Artemia Vejovis

observations: a + b = 0.647 observations: a + b = 0.494 b + c = 0.897 b + c = 0.897 a + c = 0.494 a + c = 0.647

therefore, 2a + b + C = 1.141 therefore, 2a + b + C = 1.141

and subtracting b + c = 0.897 and subtracting b + c = 0.897 yields 2a = 0.244 yields 2a = 0.244

so: a = 0.122 so: a = 0.122 b = 0.525 b = 0.372 c = 0.372 c = 0.525

FIG. 3. Relative rate tests for 16S rRNA gene sequences in horseshoe crabs (Xiphosura) under alternative assumptions that (A) Vejovis is a sister taxon to the exclusion of Artemia; and (B)Artemia is a sister taxon to the exclusion of Vejovis. Note that in either case, the inferred branch length "a" leading to the Xiphosurans is shorter than the inferred branch length "b" leading to the sister taxon. the xiphosurans, then by similar calculations the ofmolecular evolution-secure molecular infer­ rate of 16S gene evolution in the Artemia lineage ences about phylogenetic nodes in sidereal time has been about three times greater than thatlead­ will be most difficult to obtain precisely for those ing to the horseshoe crabs. Similar tests of rel­ evolutionary groups in which they are most ativerate as applied to countednumbers of'trans­ needed; that is, those groups for which there is versions (rather than adjusted distance esti­ no independent fossil or biogeographic evidence mates) yields apparent slowdowns in horsehose against which to calibrate internally the molec­ crab lineages by 1.2-1.9-fold. Thus, these several ular timepieces used in the dating exercise. analyses suggest that there may have been a slow­ down in the evolutionary rate ofthe 16S rRNA gene in horseshoe crab lineages relative to rates Comparative Morphological and in these other Arthropods. If so, estimated sep­ Molecular Evolution aration dates in the xiphosuran phylogeny would Regardless of the particular phylogenetic in­ have to be adjusted backward in time accord­ ferences possible from nucleotide sequences of ingly, relative to the dates presented in table 5. the mtDNA genes from extant horseshoe crabs, In summary ofthis section, despite the avail­ the data also are interesting when interpreted in ability of considerable molecular sequence in­ a comparative context against genetic distances formation, we appear to have gained relatively from homologous sequences in king crabs and little firm knowledge about the absolute dates of hermit crabs. King crabs are extremely large branch points in the horseshoe crab phylogeny. crustaceans with a typical crablike morphology Such results imply a general irony for the field and a strongly calcified exoskeleton, whereas the ·002 T. gigas A

B

tridentatus A, B

C. rotundicauda A, B .057

.006 L. poypI hemus A

.410 .116 .OOSL. polyphemus B

.....------Artemia salina

.5 .4 .3 .2 .1 0

1""""------Artemia salina

r------CI. vittatus

P. longicarpus ,/' (76 mya) <.:> Pa. camtschatica (37 mya ) La. splendescens

P. acadianus

( ),/ P bernhardus 2.8 mya . ------.5 .4 .3 .2 .1 0 Genetic distance

FiG. 4. Comparative UPGMA dendrograms for horseshoe crabs (above) and selected king crabs and hermit crabs (below) based on 16S rRNA gene sequences from mtDNA. The genetic distance scale (method ofKimura 1980) is the same for both phenograms, and is drawn to indicate mean clustering levels (average sum ofbranch lengths connecting relevant nodes and extant taxa). Also shown for the horseshoe crab lineages are individual branch lengths in a neighbor-joining tree that produced the same branching structureas the UPGMA dendrogram. Sequence data underlying the recalculated king crab/hermit crab tree are from Cunningham et al. (1992); the indicated dates for particular nodes (Mya, millions ofyears ago) stem from vicariant geologic evidence, and also are from Cunningham et al. (1992). T., Tachypleus; C, Carcinoscorpius; L., Limulus; Cl., Clibanarius; P., Pagurus; Pa., Paralithodes; La., Labidochirus. The brine shrimp, Artemia, is an outgroup common to both studies. 2000 JOHN C. AVISE ET AL. much-smaller hermit crabs have a decalcified long-term evolution proved to be sharply ar­ asymmetrical abdomen that the animalscoil into rested at the level ofoverall nucleotide sequence. adoptedgastropod shells for protection. As shown by Cunningham et al. (1992) in their analysis of ACKNOWLEDGMENTS 165 rRNA gene sequences as interpreted against Research was supported by a National Science a fossil-biogeographic record, the loss of shell­ Foundation grant, by funds from the University living habit and the complete carcinization of ofGeorgia and by a Sloan Foundation sabbatical king crabs from hermitlike ancestors appears to award to J.c.A. We thank J. Westphaling for have taken place over a relatively "short" time providing specimens of Vejovis. M. Lynch and period of13-25 My. This"rapid" morphological R. Harrison provided useful comments and dis­ shift (perhaps involving heterochrony, an evo­ cussion on the analysis. We would also like to lutionary change in the timing of development; thank D. Powers for providing laboratory and see also Gould 1992) has been accompanied by office space at the Hopkins Marine Station of molecular differentiation at the 16S rRNA gene Stanford University, where much of this work that ifanything is somewhat lower in magnitude was conducted. than that exhibited by the morphologically con­ servative horseshoe crabs (fig. 4). In this impor­ tant sense, levels ofmorphological and molecular LITERATURE CITED divergence are conspicuously decoupled among Avise, J. C. 1992. Molecularpopulation structure and these arthropod lineages, a recurring theme long the biogeographic history ofa regional fauna: a case history with lessons for conservation biology. Oikos emphasized by evolutionists for various other 63:62-76. organisms (e.g., King and Wilson 1975; Cherry Bames, R. D. 1963. Invertebrate zoology. W. B. et al. 1978). Saunders, Philadelphia. Thus, even among the horseshoe crabs (pro­ Cherry, L. M., S. M. Case, and A. C. Wilson. 1978. Frog perspective on the morphological divergence verbial epitomes of morphological conserva­ between humans and chimpanzees. Science 200: tism), multitudinous nucleotide differences have 209-211. accumulatedamongevolutionarylineages (albeit Cunningham, C. W., N. W. Blackstone, and L. W. Buss. at an uncertain exact pace). Consider, for ex­ 1992. Evolution of king crabs from hermit crab ample, the American versus the Asiatic horse­ ancestors. Nature 355:539-542. Eldredge, N., and S. M. Stanley, eds. 1984. Living shoe crab species, which at face value (uncor­ fossils. Springer, New York. rected for multiple hits) differed at about 16% of Felsenstein, J. 1991. PHYLIP: phylogeny inference the total of 1066 mtDNA nucleotide positions package, Version 3. 4. DepartmentofGenetics, SK­ assayed. The nuclear genome ofhorseshoe crabs 50, University ofWashington, Seattle. consists Field, K. G., G. J. Olsen, D. J. Lane, S. J. Giovannoni, of perhaps some 2.5 billion nucleotide M. T. Ghiselin, E. C. Raff, N. R. Pace, and R. A. pairs (see Galau et al. 1976). Even if nuclear Raff. 1988. Molecular phylogeny of the animal genomic divergence on a per-nucleotide basis is kingdom. Science 239:748-753. only one-tenth that of mtDNA, the American Fisher, D. C. 1984. The Xiphosurida: archetypes of and Asian horseshoe crabs would nonetheless bradytely? Pp. 196-213 in N. Eldredge and S. M. Stanley, eds. Living fossils, Springer, New York. differ at several tens of millions of nucleotide Galau, G. A., M. E. Chamberlin, B. R. Hough, R. J. sites, on average. Contrast this with the small Britten, and E. H. Davidson. 1976. Evolution of handful of morphological characters known to repetitive and nonrepetitive DNA. Pp. 200-224 in distinguish these species. F. J. Ayala, ed. Molecular evolution. Sinauer, Sun­ derland, Mass. Our conclusions are consistent with those Gould, S. J. 1992. We are all monkeys' uncles. Nat­ reached a decade ago by Schopf (1984): "Other ural History 101(6):14-21. 'living fossils,' where examined, exhibited no Gould, S. J., and N. Eldredge. 1977. Punctuatedequi­ biochemical traits indicative ofa species 'where libria: the tempo and mode ofevolution revisited. time has stood still.' There is nothing special Paleobiology 6:115-151. Hackman, R. H., and M. Goldberg. 1975. Peripatus: distinguishing the DNA or RNA ofLingula (Shi-' its affinities and its cuticle. Science 190:582-583. mizu and Miura 1971a,b; Shimizu 1971), the Hillis, D. M., and M. T. Dixon. 1991. Ribosomal cytochrome C ofGinkgo (Ramshaw et al. 1971), DNA: molecular evolution and phylogenetic infer­ or the cuticle ofPeripatus (Hackman and Gold­ ence. Quarterly Review of Biology 66:411-453. Innis, M. A., D. H. Gelfand, J. J. Sninsky, and T. J. ing 1975)." Although molecular rates may vary White. 1990. PCR protocols: a guide to methods widely, a revolutionary finding would now be an and applications. Academic Press, New York. example from any taxonomic group in which Irwin, D. M., T. D. Kocher, and A. C. Wilson. 1991. PHYLOGENY OF LIVING FOSSILS 2001

Evolution of the cytochrome b gene of mammals. Shimizu, N. 1971. Studies on nucleic acids ofliving Journal of Molecular Evolution 32:128-144. fossils. III. A classification of transfer ribonucleic Kimura, M. 1980. A simple method for estimating acids by elution profiles on gel filtration and sedi­ evolutionary rate of base substitutions through mentation profiles on sucrose density gradient. comparative studies ofnucleotide sequences. Jour­ Journal of Biochemistry 69:761-770. nal of Molecular Evolution 16:111-l20. Shimizu, N., and K.-I. Miura. 1971a. Studies on nu­ King, M.-C., and A. C. Wilson. 1975. Evolution at cleic acids of living fossils. I. Isolation and char­ two levels in humans and chimpanzees. Science acterization ofDNA and some RNA components 188:107-116. from the brachiopod Lingula. Biochimica et Bio­ Kornegay, J. R., T. D. Kocher, L. A. Williams, and A. physica Acta 232:271-277. C. Wilson. 1993. Pathways oflysozyme evolution ---. 1971b. Studies on nucleic acids ofliving fos­ inferred from the sequences of cytochrome b in sils. II. Transfer RNA from the brachiopod Lin­ birds. Journal ofMolecular Evolution 37:367-379. gula. Biochimica et Biophysica Acta 232:278-288. Lynch, M. 1993. A method for calibrating molecular Shishikura, F., S. Nakamura, K. Takahashi, and K. clocks and its application to animal mitochondrial Sekiguchi. 1982. Horseshoe crab phylogeny based DNA. Genetics 135:1197-1208. on amino acid sequences ofthe fibrino-peptide-like Miyazaki, J., K. Sekiguchi, and T. Hirabayashi. 1987. peptide C. Journal of Experimental Zoology 223: Application of an improved method of two-di­ 89-91. mensional electrophoresis to the systematic study Shuster, C. N., Jr. 1962. Serological correspondence ofhorseshoe crabs. Biological Bulletin 172:212-224. among horseshoe "crabs" (Limulidae). Zoologica Neigel, J. E., and J. C. Avise. 1986. Phylogenetic 47:1-8. relationships ofmitochondrial DNA under various ---. 1982. A pictorial review ofthe naturalhistory demographic models ofspeciation. Pp. 515-534 in and ecology of the horseshoe crab Limulus poly­ E. Nevo and S. Karlin, eds. Evolutionary processes phemus. with reference to other Limulidae. Pp, 1­ and theory. Academic Press, New York. 52 in J. Bonaventura, C. Bonaventura, and S. Tesh, Palumbi, S. R., A. Martin, S. Romano, W. O. Mc­ eds. Physiology and biology of horseshoe crabs: Millan, L. Stice, and G. Grabowski. 1991. The studies on normal and environmentallystressed an­ simple fool's guide to PCR, Version 2. University imals. Alan R. Liss, New York of Hawaii Zoology Department, Honolulu. Sneath, P. H. A. ,and R. R. Sokal. 1973. Numerical Pamilo, P., and M. Nei. 1988. Relationships between taxonomy. W. H. Freeman, San Francisco. gene trees and species trees. Molecular Biology and Stermer, L. 1955. Chelicerata, Part P. Pp. 4-41 in Evolution 5:568-583. R. C. Moore, ed.Treatise on invertebrate paleon­ Ramshaw, J. A. M., R. Richardson, and D. Boulter. tology. University of Kansas Press, Lawrence. 1971. The amino-acid sequence ofthe cytochrome Sugawara, K., H. Yonekawa, Y. Tagashira, and K. Sek­ c of Ginkgo bi/oba L. European Journal of Bio­ iguchi. 1988. Mitochondrial DNA polymor­ chemistry 23:475-483. phisms. Pp. 375-382 in K. Sekiguchi, ed. Biology Riska, B. 1981. Morphological variation in the horse­ of horseshoe crabs. Science House, Tokyo. shoe crab, Limuluspolyphemus. Evolution 35:647­ Sugita, H. 1988. Immunological comparisons of he­ 658. mocyanins and their phylogenetic implications. Pp. Saiki, R. K., D. H. Gelfand, S. Stoffel, S. J. Scharf, R. 315-334 in K. Sekiguchi, ed. Biology ofhorseshoe Higuchi, G. T. Horn, K. B. Mullis, and H. A. Erlich. crabs. Science House, Tokyo. 1'988. Primer-directed enzymatic amplification of Swofford, D. L. 1993. PAUP: Phylogenetic Analysis DNA with a thennostabile DNA polymerase. Sci­ using Parsimony, Version 3. I. Computer Program ence 239:487-491. distributed by the Illinois Natural History Survey, Saitou, N., and M. Nei. 1987. The neighbor-joining Champaign, Ill. method: a new method for reconstructing phylo­ Takahata, N. 1989. Gene genealogy in three related genetic trees. Molecular Biology and Evolution populations: consistency probability between gene 4:406-425. and population trees. Genetics 122:957-966. Sanger, F., S. Nickien, and A. R. Coulson. 1977. DNA Tajima, F. 1983. Evolutionary relationships ofDNA sequencing with chain-terminating inhibitors. Pro­ sequences in finite populations. Genetics 105:437­ ceedings ofthe National Academy ofSciences, USA 460. 74:6463-5467. Turbeville, J. M., D. M. Pfeifer, K. G. Field, and R. Saunders, N. c., L. G. Kessler, andJ. C. Avise. 1986. A. Ralf. 1991. The phylogenetic status ofarthro­ Genetic variation and geographic differentiation in pods, as inferred from 18S rRNA sequences. Mo­ mitochondrial DNAofthe horseshoe crab, Limulus lecular Biology and Evolution 8:669-686. polyphemus. Genetics 112:613-627. Williams, G. C. 1992. Natural Selection: Domains, Schopf, T. J. M. 1984. Rates of evolution and the Levels and Challenges. Oxford University Press, notion of"living fossils. " Annual Review ofEarth New York. and Planetary Science 12:245-292. Yamasaki, T. 1988. Taxonomy. Pp. 10-21 in K. Sek­ Sekiguchi, K., and H. Sugita. 1980. Systematics and iguchi, ed. Biology of horseshoe crabs. Science hybridization in the four living species ofhorseshoe House, Tokyo. crabs. Evolution 34:712-718. Selander, R. K., S. Y. Yang, R. C. Lewontin, and W. E. Johnson. 1970. Genetic variation in the horse­ shoe crab (Limulus polyphemus), a phylogenetic "relic". Evolution 24:402-414. Corresponding Editor: R. Harrison