Talpid Mole Phylogeny Unites Shrew Moles and IlluminatesOverlooked Cryptic Species Diversity

Kai HeDaggerdagger12 Akio Shinoharadagger3 Kristofer M Helgen4 Mark S Springer5 Xue-Long Jiang1 andKevin L Campbell2

1State Key Laboratory of Genetic Resources and Evolution Kunming Institute of Zoology Chinese Academy of Sciences KunmingChina2Department of Biological Sciences University of Manitoba Winnipeg MN Canada3Department of Bio-resources Division of Biotechnology Frontier Science Research Center University of Miyazaki Miyazaki Japan4National Museum of Natural History Smithsonian Institution Washington DC5Department of Biology University of California Riverside CADaggerPresent address The Kyoto University Museum Kyoto University Kyoto JapandaggerThese authors contributed equally to this work

Corresponding authors E-mails jiangxlmailkizaccn kevincampbellumanitobaca

Associate editor Emma Teeling

Abstract

The mammalian family Talpidae (moles shrew moles desmans) is characterized by diverse ecomorphologies associatedwith terrestrial semi-aquatic semi-fossorial fossorial and aquatic-fossorial lifestyles Prominent specializations involvedwith these different lifestyles and the transitions between them pose outstanding questions regarding the evolutionaryhistory within the family not only for living but also for fossil taxa Here we investigate the phylogenetic relationshipsdivergence times and biogeographic history of the family using 19 nuclear and 2 mitochondrial genes (16 kb) from60 of described species representing all 17 genera Our phylogenetic analyses help settle classical questions in theevolution of moles identify an ancient (mid-Miocene) split within the monotypic genus Scaptonyx and indicate thattalpid species richness may be nearly 30 higher than previously recognized Our results also uniformly support themonophyly of long-tailed moles with the two shrew mole tribes and confirm that the Gansu mole is the sole living Asianmember of an otherwise North American radiation Finally we provide evidence that aquatic specializations within thetribes Condylurini and Desmanini evolved along different morphological trajectories though we were unable to statis-tically reject monophyly of the strictly fossorial tribes Talpini and Scalopini

Key words Talpidae tree of life cryptic species aquatic fossorial

Introduction

Members of the mammalian family Talpidaemdashthe molesshrew moles and desmansmdashinclude some of the least studiedmammals on Earth The 43 recognized living species reflect asurprisingly diverse set of lifestyles including terrestrial semi-fossorial semi-aquatic fossorial and aquatic-fossorial (Nowak1999 Hutterer 2005) The evolution of adaptive specializationsto these diverse environments produced substantial morpho-logical variation for example in skeletal anatomy (Freeman1886) and sensory systems (Catania 2000) However increas-ingly derived ldquostepping-stonerdquo phenotypic adaptations can beidentified across a morphological gradient from the terrestrialshrew-like moles (Uropsilinae) to the semi-fossorial shrewmoleslong-tailed moles (Neurotrichini Urotrichini andScaptonychini) and further to the fossorial ldquotrue molesrdquo(Talpini and Scalopini) Conversely the more aquatic membersof the family the desmans (Desmanini) and the aquatic-fossorial star-nosed mole (Condylurini) are among the mostunique and morphologically isolated of living mammals

Prominent locomotory adaptations involved with the variedtalpid lifestyles have attracted considerable interest fromnatural historians and evolutionary biologists for morethan a century (Shimer 1903 Edwards 1937 Reed 1951Hutchison 1974 Yates and Moore 1990 Whidden 2000Piras et al 2012 Meier et al 2013) Biologists have especiallybeen fascinated by (i) whether the aquatic forms (Condyluriniand Desmanini) derived from semi-fossorialfully fossorialmoles or vice versa and (ii) whether the striking morpholog-ical similarity of the two strictly fossorial ldquotrue molerdquo tribes(Talpini and Scalopini) reflects synapomorphy or evolvedconvergently from an ancestral shrew-like morphology Thislatter dispute is particularly topical in light of a recent molec-ular hypothesis (Bannikova Zemlemerova Lebedev et al2015) that strongly refutes the sister-taxa association consis-tently recovered between the two fossorial tribes in anatom-ical studies (Whidden 2000 Motokawa 2004 Sanchez-Villagraet al 2006 Schwermann and Thompson 2015) Additionaloutstanding systematic questions that have attracted vigor-ous debate involve the phylogenetic placement of key

Article

The Author 2016 Published by Oxford University Press on behalf of the Society for Molecular Biology and EvolutionAll rights reserved For permissions please e-mail journalspermissionsoupcom

78 Mol Biol Evol 34(1)78ndash87 doi101093molbevmsw221 Advance Access publication October 30 2016

taxonomic and morphological lineages including whetherthe Chinese endemic Gansu mole (Scapanulus) is a shrewmole (Van Valen 1967) or a member of the North Americantrue mole tribe Scalopini (Hutchison 1968 Kawada Li et al2008 Bannikova Zemlemerova Lebedev et al 2015Schwermann and Thompson 2015) and whether theSoutheast Asian endemic long-tailed mole (Scaptonyx) ismore closely related to fossorial moles or to the trans-Pacific shrew mole tribes (Hutchison 1968) Numerous phy-logenetic hypotheses have been proposed for the familybased on different character sets including osteologicalmyological and molecular data (Hutchison 1976 Yatesand Moore 1990 Corbet and Hill 1992 Whidden 2000Motokawa 2004 Shinohara et al 2004 Cabria et al 2006Sanchez-Villagra et al 2006 Bannikova ZemlemerovaLebedev et al 2015 Schwermann and Thompson 2015)but results from these disparate comparisons have manyconflicts with subfamily and inter-tribal relationships in par-ticular lacking strong statistical support

Conflicting taxonomic classifications regarding the estab-lishment and composition of subfamilies and tribes have beenproposed for the talpid family (Hutchison 1968 Yates 1984McKenna and Bell 1997 Hutterer 2005) with substantial con-fusion also surrounding the phylogenetic placement and tax-onomic assignment of talpids in the fossil record (Ziegler2003 2012 Klietmann et al 2015) Because of the robustnessof the forelimb in some talpids relatively plentiful numbers ofhumeri are preserved as fossils for some lineages (eg TalpiniScalopini Urotrichini) and can provide insights into locomo-tory adaptations through geological time However the res-olution of locomotory adaptations through time requires awell-resolved phylogeny which is currently not available

Herein we investigated the phylogenetic relationships di-vergence times and biogeographic history of the familyTalpidae using comprehensive molecular approaches that in-corporated data from 19 nuclear (14 kb) and 2 mitochon-drial genes (2 kb) from 52 specimens representing all 17genera and 60 (25 of 43) of recognized talpid species(supplementary table S1 Supplementary Material online)To further access potential overlooked species diversity(Wan et al 2013 He et al 2014 Bannikova ZemlemerovaColangelo et al 2015 Feuda et al 2015) we also constructed amaximum-likelihood tree using an expanded supermatrixthat included published sequence data from 15 additionaldescribed talpid species Our results provide a robust scaffoldfor systematic studies of living and fossil species reveal thetempo of talpid diversification in the context of the Paleoceneextinction events identify up to 12 undescribed cryptic spe-cies across 6 genera and offer key insights to major questionsregarding the evolutionary history morphological transitionsand systematics of talpid moles

Results

Phylogenetic Relationships and Species DiversityOur concatenated gene tree (21 genes) indicates that thebasal split in Talpidae is between the subfamily Uropsilinaeand the seven recognized mole tribes (PPfrac14 10 BSfrac14 100)

(fig 1) Although the fossorial tribe Scalopini is strongly sup-ported as sister to the remaining six tribes in our Bayesiananalysis (PPfrac14 10) this relationship was only weakly sup-ported in our ML tree (BSfrac14 65) Similar discrepancies be-tween these analyses were recovered regarding themonophyly of the aquatic desmans (tribe Desmanini) withthe aquatic-fossorial Condylurini (PPfrac14 097 BSfrac14 25) andtheir grouping with the strictly fossorial tribe Talpini(PPfrac14 099 BSfrac14 47) While the latter clade was recoveredwhen the mitochondrial genes were removed from the data-set (PPfrac14 089) the monophyly of Condylurini and Desmaniniwas not supported (supplementary fig S1A SupplementaryMaterial online) Conversely all concatenated analyses pro-vided strong support (PPfrac14 10 BSfrac14 100) for (i) a sister-taxarelationship between the Gansu mole Scapanulus oweni andParascalops breweri within the North American Scalopini (ii)a monophyletic clade that unites long-tailed moles (tribeScaptonychini) with the shrew mole tribes Neurotrichiniand Urotrichini and (iii) the placement of Euroscaptor mizurasister to the genus Mogera Notably both analyses also iden-tified two long-tailed mole (Scaptonyx fusicaudus) lineageswith strikingly long branches (fig 1 Supplementary fig S1ASupplementary Material online)

The coalescent species tree estimations were highly con-gruent with the concatenated trees at the tribal level (fig 2and supplementary fig S1B and C Supplementary Materialonline) though most tribal interrelationships including thepositions of the star-nosed mole and the desmans were notresolved In addition while Scapanulus was unambiguouslyplaced within Scalopini its sister-relationship withParascalops only received modest support (BSfrac14 49ndash72)Conflicts among the coalescent trees included the interre-lationships among Southeast Asian Euroscaptor and be-tween long-tailed moles and the two shrew-mole tribesthough support scores for these alternative nodes werelow (BSfrac14 49ndash74) SVDquartets further placed Euroscaptorparvidens sister to a clade containing other Southeast AsianEuroscaptor plus Parascaptorthorn Scaptochirus (fig 2B) result-ing in a polyphyletic Euroscaptor (BSfrac14 31)

As bootstrap support scores for inter-tribal nodes wereoften lt50 we conducted three separate parsimony analy-ses to statistically test the monophyly versus diphyly ofScalopinithorn Talpini These tests revealed no significant differ-ences between the two hypotheses Six trees at 18707 stepswere recovered for the diphyly hypothesis and are only foursteps shorter than three trees at 18711 steps that were recov-ered when Scalopinithorn Talpini monophyly was enforced(Pgt 053 supplementary table S2 Supplementary Materialonline)

Species Diversification through Time and SpaceThe species delimitation analyses provided strong support forundescribed cryptic species in the genera UropsilusEuroscaptor and Parascaptor and identified 33 distinct line-ages (confidence interval [CI]frac14 30ndash37) in our concatenated21-gene dataset Moreover our analysis with an expandedsupermatrix which included93 of accepted species diver-sity suggests that talpid species richness may be nearly 30

Intended for the Discoveries Section of MBE doi101093molbevmsw221 MBE

79

higher than previously recognized (ie 55 vs 43 species) as 12uncharacterized lineages across 6 genera were identified (fig3 supplementary table S1 Supplementary Material online)

Divergence time estimates suggest that the most recentcommon ancestor of the family dates from the Eoceneboundary at ca 47 million years ago (Ma) with a 95 CI of57ndash42 Ma (fig 1) Tribal divergences commenced during thePriabonian Stage (395ndash339 Ma) of the late Eocene whenScalopini diverged from other tribes at 370 Ma (95 CI =395ndash353 Ma) and continued through the Rupelian (339ndash281 Ma) and Chattian (281ndash2303 Ma) stages of theOligocene when the three shrew mole tribes(Scaptonychini Urotrichini Neurotrichini) diverged fromeach other between 30 and 27 Ma (95 CI = 34ndash23 Ma)Divergences between Chinese and North American scalopinsbetween European and Asian talpins and between the twolong-tailed mole lineages all occurred in the mid-Miocene (ca16 Ma 95 CI = 22ndash10 Ma) Finally divergences between

sister genera generally occurred between 18 and 7 My (95CIfrac14 26ndash4 Ma) The lineage-through-time (LTT) plot (supplementary fig S2A Supplementary Material online) showed anincreasing diversification rate beginning ca 10 Ma Howeverthe Pybus and Harveyrsquos c-value was insignificantly negative(070 Pfrac14 024) while the likelihood-based test found pure-birth as the best constant rate model (AICfrac14 6867) and theyule-3-rate as the best rate variable model (AICfrac14 6794) sug-gesting that speciation rate did not shift significantly throughtime The fossil-diversity plot (supplementary fig S2BSupplementary Material online) demonstrated that the num-ber of semi-fossorial and nonfossorial species have substantiallydeclined in the last four million years

The results of biogeographic inference analyses with andwithout fossil taxa were discordant (fig 4A and B) For exam-ple the extant-species-only analysis supported an Asian orig-ination for Talpidae while the origin of the family is uncertainwhen fossil taxa are incorporated When incorporating fossils

FIG 1 Time calibrated Bayesian species tree of the family Talpidae based on a concatenated alignment of 19 nuclear and 2 mitochondrial genesBranch lengths represent time and node bars indicate the 95 CI for each clade age All relationships are highly supported (posterior probabilities[PP]frac14 100 and bootstrap values [BS] 95) otherwise support values are provided Inset Branch lengths represent substitutions per site

He et al doi101093molbevmsw221 MBE

80

the analysis supported a North American origin of Scalopiniand European origins of Talpini as well as the ancestor ofCondylurinithornDesmaninithorn Talpini The extant-species-onlyanalysis did not strongly support biogeographic origins forany of these clades The two analyses also supported differentorigination scenarios for the ancestor of the long-tailed moleand the shrew moles

Morphology-Based TreesWe also explored phylogenetic relationships based on 175morphological characters (Schwermann and Thompson2015) and recovered nine most parsimonious (MP) treeswith tree lengths of 569 steps (table 1 supplementary fileS1 Supplementary Material online) These trees broadly con-flicted with molecular phylogenetic relationships in that theyreject shrew mole monophyly place desmans as sister to theother non-Uropsilus tribes and recovered Condylura as thesister group of Talpinithorn Scalopini (supplementary fig S3Supplementary Material online) Eighteen transformation se-ries (TS) supported the sister relationship ofTalpinithorn Scalopini and another 12 TS supported the cladeuniting Condylurini with the fossorial moles with most per-taining to the clavicle humerus manubrium scapula andsternum (supplementary text S1 Supplementary Material on-line) When characters were optimized on our molecular

phylogenies single trees of 621ndash622 steps (fully constrained)or 5 trees of 581 steps (partly constrained) were recovered(table 1 supplementary file S1 Supplementary Material on-line) When the monophyly of CondylurathornDesmanini wasconstrained five MP trees with tree lengths of 582 steps wererecovered with only five TS supporting this semi-aquaticclade (supplementary text S2 Supplementary Material on-line) notably these characters do not appear to be speciali-zations for aquatic locomotion or lifestyles

DiscussionAccurate species trees are the lifeblood of many questions inevolutionary biology though have proven to be challenging toelucidate in practice due to ancient rapid radiation events andthe confounding effects of recombination hybridization andincomplete lineage sorting (ILS) While numerous multi-genephylogenetic reconstruction approaches (eg concatenationand coalescent estimation) have been developed to overcomethe influence of these forces the optimal analytical approachremains contested as each method is subject to differenttradeoffs drawbacks and biases (Chou et al 2015 Miraraband Warnow 2015 Springer and Gatesy 2016) Despite theirdifferent limitations concatenation ASTRAL-II andSVDquartets provide 100 bootstrap support for many ofthe same clades including a basal split between Uropsilinae

FIG 2 Bootstrap coalescent species trees of the family Talpidae estimated using (A) ASTRAL-II and (B) SVDquartets Support values are providedfor all non-highly supported nodes (posterior probabilities [PP]lt095 or bootstrap values [BS]lt90)

Intended for the Discoveries Section of MBE doi101093molbevmsw221 MBE

81

and other talpids inclusion of the Gansu mole (S oweni)within the otherwise New World tribe Scalopini monophylyof ScaptonychinithornUrotrichinithornNeurotrichini and mono-phyly of each of the remaining tribes that include morethan one species (Desmanini Talpini) While conflicts were

apparent among the species trees in all cases these differenceswere not robustly supported by two or more competingmethods

In general bootstrap support percentages are higher forconcatenation with 21 genes (mean bootstrap

FIG 3 Maximum likelihood tree (-ln likelihoodfrac14 138472) of the family Talpidae based on an expanded supermatrix for 19 nuclear and 2mitochondrial genes from 782 specimens (2836 gene sequences supplementary file S2 Supplementary Material online) Currently undescribedpotential cryptic species are numbered 1ndash12 (see supplementary table S1 Supplementary Material online) The tree was constructed usingRAxML with tree branch lengths representing substitutions per site

He et al doi101093molbevmsw221 MBE

82

supportfrac14 94) than for ASTRAL-II (90) and SVDquartets(87) This finding is consistent with other studies (eg Prumet al 2015) that have reported higher bootstrap percentagesfor methods when individual gene segments are not binnedthrough local or global concatalescence (Gatesy and Springer2013) to mitigate gene tree reconstruction errors Differencesbetween concatenation and coalescence species trees mayalso reflect the distinction between ILS ignorant (concatena-tion) and ILS aware (coalescence) methods but evaluations ofempirical datasets suggest that problems with gene tree inac-curacy are more problematic for shortcut coalescent methodsthan ILS is for concatenation for deep-level problems in theTree of Life (Gatesy and Springer 2014 Springer and Gatesy2016) Also species trees based on two coalescence methods(ASTRAL-II SVDquartets) exhibit topological differences witheach other but these differences cannot be explained by ILSThe strongly supported basal relationships among the fivenonuropsiline clades in the concatenated 21-gene Bayesiananalysis (PPgt 095 in all cases) is similarly suspect and mayreflect model mis-specification (Waddell and Shelley 2003) asplacements of these lineages received low support scores in all

other analyses conducted both with and without inclusion ofthe two mitochondrial loci The low branch supports we re-covered for the clade containing the six non-Scalopini tribes inour non-Bayesian analyses are particularly surprising giventhat this clade was highly supported in ML analyses(BSfrac14 92) conducted on a much smaller (six loci 58 kb)dataset (Bannikova Zemlemerova Lebedev et al 2015)Nonetheless our inability to statistically refuteTalpinithorn Scalopini monophyly re-opens the debate regardingthe single versus convergent evolution of fossorial adaptationsin these tribes This lack of resolution and the discordantplacement of tribes among our concatenated and coalescenttrees are consistent with a series of rapid radiation events andcalls for additional phylogenomic data and alternative analy-ses (eg retroposon insertions) to fully resolve difficult nodes

As noted above however even with these differencesthere are no topological incongruences that are strongly sup-ported by both concatenation and coalescence species treesthat is all methods resulted in a congruent set of stronglysupported clades and thus close several century-long debatesabout the taxonomy and phylogeny of moles For example all

FIG 4 Biogeographical reconstruction of ancestral ranges of the family Talpidae estimated using the DECthornJ model with only living species (A)and with 14 fossil taxa included (B) Pie charts represent the probabilities of the likely ancestral ranges at each node Map insets show the most likelytrans-continental migration scenarios at ancestral nodes or for each tribe Crosses indicate extinction after migration

Table 1 Summary of the Results of Morphological Analyses

of best trees steps CI RI RC

Maximum parsimony treesa 9 569 0424 0676 0286Full constraint (21 genesconcatenation)b 1 621 0388 0625 0242Full constraint (21 genesASTRAL-II) 1 622 0387 0624 0242Full constraint (21 genesSVDquartets) 1 622 0387 0624 0242Partial constraint (molecular consensus)c 1 581 0415 0664 0275Condylura thorn Desmanini constrained 5 582 0414 0663 0275

NOTEmdashall input datasets and output results are given in the supplementary file S1 Supplementary Material onlineaProvided in supplementary fig S3 Supplementary Material onlinebThe maximum clade credibility tree estimated using the concatenated 21-gene dataset implemented in BEASTcOnly those relationships that were strongly supported (PP095 and BS 85) in concatenated and coalescent species trees using 21 genes were constrained

Intended for the Discoveries Section of MBE doi101093molbevmsw221 MBE

83

analyses strongly supported inclusion of the long-tailed moles(Scaptonychini) into a monophyletic clade containing thetwo shrew mole tribes a point that has previously generatedconsiderable debate mostly disputing their monophyly (VanValen 1967 Ziegler 1971 Hutchison 1976 Yates and Moore1990 Corbet and Hill 1992 McKenna and Bell 1997Shinohara et al 2003 Motokawa 2004 Cabria et al 2006Sanchez-Villagra et al 2006 Kawada Li et al 2008Schwermann and Thompson 2015) The Gansu mole(S oweni) is also unambiguously placed with the NorthAmerican scalopins (Hutchison 1968 BannikovaZemlemerova Lebedev et al 2015 Schwermann andThompson 2015) rather than with Asian scaptonychins(Van Valen 1967) Moreover crown Scalopini is equally orslightly older than crown Talpini (fig 1) suggesting that theGansu mole is a relict scalopin in Asia that has survived a longhistory of climatic changes and competition with radiationsof Old World moles as well as various subterranean rodentlineages (Nevo 1979) In addition the surprisingly ancientdivergence (18 Ma) within the monotypic long-tailedmole genus Scaptonyx and paraphyletic or polyphyleticEuroscaptor indicate the presence of uncharacterized genus-level talpid lineages with strong evidence for cryptic speciesalso found in Euroscaptor Parascaptor and Uropsilus as sug-gested in previous studies (Wan et al 2013 He et al 2014)Remarkably our expanded supermatrix analysis indicates thatup to 12 genetically distinct talpid lineages remain to becharacterized (fig 3) indicating that talpid diversity may benearly a third higher than previously recognized (ie 55 vs 43species) Most of these putative species are from the moun-tains of southwest China implying that the ldquosky islandrdquo land-scape has had significant effects on speciation (Lei 2012 Heand Jiang 2014) Extensive sampling in this and adjacent areasis likely to reveal additional undescribed lineages and is re-quired to elucidate the full species diversity of the region

Timing and Tempo of Talpid DiversificationDivergence time estimates indicate that all tribes originatedduring the EocenendashOligocene transition which may be at-tributed to global cooling and desiccation and the conse-quent development of non-forest flora (Nevo 1979)Although most extant talpids have very limited distributionsand dispersal abilities there is strong evidence for trans-continental and oversea dispersals following these radiationevents (fig 4 supplementary text S3 Supplementary Materialonline) For example the fossil history of Uropsilinae (includ-ing Asthenoscapter and Desmanella) extends much furtherback in Europe than it does in Asia (Qiu and Storch 2005)Similarly members of the North American tribesNeurotrichini Condylurini and Scalopini were also distrib-uted in Europe during the Oligocene to Pleistocene (supplementary fig S2C Supplementary Material online) but havesince been extirpated (van den Hoek Ostende et al 2005)This decline of non-fossorial talpids was clearly evident onthe fossil-diversity plot (supplementary fig S2BSupplementary Material online) but not accounted for onthe LTT plot using living species only (supplementary fig S2ASupplementary Material online) Consequently

biogeographical reconstructions in the absence of fossil evi-dence (see fig 4) may be highly misleading though it is im-portant to note that they may also be strongly biased by bothmissing and incorrectly placed fossil taxa as well as by un-certain phylogenetic relationships among living taxa Theseconstraints call for studies addressing the definitive phyloge-netic placements of all known fossil species to better assessthe biogeographic origins of the various major lineages withinTalpidae and for the family itself as morphological charactermatrices are currently only available for a few fossil Europeanspecies (Schwermann and Thompson 2015 Hooker 2016)

Ecomorphological EvolutionThe adaptation of talpid moles to fossorial and aquatic en-vironments has generated substantial research attentionMapping morphological characters on our molecular phy-logenetic trees resulted in an increase in tree length from569 to at least 581 steps (table 1) When morphologicalcharacters primarily associated with digging are mappedon the most parsimonious morphological tree a patternof increasing robustness of the clavicle manubrium scapulasternum and humerus are apparent from the shrew molesthrough to the star-nosed moles and finally to the true(Scalopinithorn Talpini) mole tribes (supplementary fig S3and supplementary text S1 Supplementary Material online)As a sister-taxa relationship between these latter two tribeswas neither supported nor rejected by molecular evidencehowever questions regarding evolution of their fossorialspecializations remain unresolved (Schwermann andThompson 2015) Conversely the five uniquely shared ana-tomical characters in the tribes Condylurini and Desmaniniare unlinked to semi-aquatic lifestyles (supplementary textS2 Supplementary Material online) supporting the conten-tion that star-nosed moles and desmans evolved aquaticspecializations along independent morphological trajecto-ries from a common shrew-mole-like ancestor

ConclusionsWe provide a phylogenetic hypothesis that robustly uniteslong-tailed moles with shrew moles identifies a deep (mid-Miocene) split within the monotypic genus Scaptonyx andilluminates parallel adaptive routes to semi-aquatic habitsMoreover our phylogeny challenges numerous taxonomicclassifications pertaining to the mole family and reopensthe Talpinithorn Scalopini monophyly versus diphyly debateFinally our findings suggest that talpid species richness isunderestimated by nearly 30 with most of these unde-scribed lineages inhabiting the meagerly sampled mountainsof southwest Chinamdasha known biodiversity hotspot that islikely to harbor additional taxonomic diversity

Materials and Methods

Taxon Sampling and Data CollectionForty-three talpid species including four recently describedspecies (Kawada et al 2007 Kawada Yasuda et al 2008Kawada Son et al 2012 Liu et al 2013) are recognized world-wide Our taxon sampling included representatives from 25

He et al doi101093molbevmsw221 MBE

84

accepted species and 5 putative cryptic lineages identified inprevious studies (Wan et al 2013 He et al 2014) from all 17talpid genera (52 specimens in total) together with 3 shrewspecies (supplementary table S3 Supplementary Material on-line) Standard phenolndashchloroform extraction protocol orQiagen DNeasy Blood and Tissue kits were used to extracttotal DNA

Two mitochondrial (12S CYTB) and segments of 19 nu-clear genes (ADORA3 ADRB2 APOB APP ATP7A BCHEBDNF BMI1 BRCA1 BRCA2 CREM DMP1 ENAM GHRPLCB4 RAG1 RAG2 TTN2 VWF) were amplified following(Meredith et al 2011) Primer sequences are provided insupplementary table S4 Supplementary Material onlineCorresponding sequences of two moles two shrews onehedgehog one gymnure and one solenodon were down-loaded from GenBank and included in our datasetSequences from 54 ingroups and 7 outgroups were assembledusing SeqMan (Lasergene v71) and aligned using MUSCLE(Edgar 2004) Patterns of variation including mean geneticdistances and informative sites for each gene are given insupplementary table S5 Supplementary Material online

Gene Trees and Divergence TimeWe first constructed 20 maximum likelihood gene trees (onefor each nuclear locus plus a combined mitochondrial genetree) rooted with Solenodon and enforcing the monophyly ofeach family As preliminary analyses using RAxML revealedhigh percentages of arbitrarily resolved relationships withbranch lengthslt1 108 for all loci we instead imple-mented the default settings in GARLI v201 (Bazinet et al2014) and collapsed branches with lengthslt1 108 intopolytomies The partitioning schemes and evolutionary mod-els of each genecodon position were simultaneously esti-mated using PartitionFinder v111 (Lanfear et al 2012)(supplementary table S6 Supplementary Material online)GARLI was run twice for each gene alignment to estimate100 bootstrap trees and to determine the best scoring ML tree

Two concatenated datasets were compiled one solelybased on the 19 nuclear markers (13992 bp) and the otherincorporating both mitochondrial and nuclear loci(15986 bp) Bayesian and ML gene trees were estimatedfrom these datasets using BEAST v180 (Drummond et al2012) and RAxML v82 (Stamatakis 2014) respectively Thepartitioning schemes and evolutionary models were esti-mated using PartitionFinder (supplementary tables S7 andS8 Supplementary Material online) The concatenatedBEAST analyses consisted of 100 million generations withchains being sampled every 10000 generations and the nu-clear genes and mitochondrial genes given independent re-laxed molecular clocks Each analysis used a random startingtree birthndashdeath tree prior Ten calibrations were used fordivergence time estimation (supplementary text S4Supplementary Material online) Convergence of parameterswas accessed using Tracer v16 We ran RAxML analyses of thepartitioned data with random starting trees with GTRthornCmodel performed on CIPRES (Miller et al 2015) Supports forrelationships were estimated with a rapid Bootstrap

algorithm and 500 replicates were used to estimate theBootstrap supports

Species Delimitation Species Tree Reconstructionand Biogeographic ReconstructionSpecies delimitation analyses based on the phylogenetic spe-cies concept were calculated using a generalized mixed Yule-coalescent model (GMYC) implemented with the R packagesplits (Pons et al 2006)

As our dataset contained gene trees with bifurcationsconnected by branches with effective lengths of 0 (seeabove) species trees were estimated using a recently devisedand statistically consistent summary coalescent approachmdashASTRAL-IImdashthat accommodates polytomies (Mirarab andWarnow 2015) In brief we ran ASTRAL-II using the bestscoring ML trees estimated by GARLI with 100 bootstraptrees of each gene used for multi-locus bootstrapping Wealso estimated species trees with SVDquartets as imple-mented in PAUP 40a This SNP summary approach infersquartet trees for all subsets of four species from single-sitedata and then combines the set of quartet trees into a spe-cies tree thereby avoiding potential problems arising fromgene tree discordance (Chifman and Kubatko 2014) Weevaluated 100000 random quartets with 100 bootstrap rep-licates For both the ASTRAL-II and SVDquartet analysesspecies trees were estimated from both the nuclear and com-bined nuclearmitochondrial datasets

Because not all phylogenetic relationships in our concaten-ated tree were well supported by species tree estimations wefurther tested the monophyly versus diphyly ofScalopinithorn Talpini using three parsimony-based methodsmdashKishinondashHasegawa (Kishino and Hasegawa 1989) Templetonand winning site tests (Templeton 1983)mdashas implemented inPAUP 40a and CONSEL v01 respectively

We estimated ancestral distributions by applying a mod-ified likelihood-based DispersalndashExtinctionndashCladogenesis(DEC) modelmdashDECthornJ (Massana et al 2015)mdashwhich al-lowed dispersal events (J) and prohibited transitions intothe null range in the transition rate matrix () We used our21-gene ultrametric tree to assign living species to their con-temporary distributions The talpid distributions were dividedby continents Asia Europe and North America and we al-lowed a maximum of two areas at each node assuming suchsmall animals could not be distributed through three conti-nents at the same time Analyses were implemented in the Rpackage BioGeoBEARS (Matzke 2014) We also included fossiltaxa though adopted a more conservative approach thanPiras et al (2012) In brief we placed 14 fossil taxa whosephylogenetic positions have been strongly inferred (see sup-plementary text S5 Supplementary Material online for justi-fications) in previous studies on our concatenated gene treeand re-ran the analysis All fossil taxa were treated as terminaltaxa instead of direct ancestors thus accounting for dispersalextinction and cladogenesis in the analyses

Expanded Supermatrix AnalysisTo assess potential overlooked genetic diversity in the talpidfamily we assembled an expanded molecular supermatrix

Intended for the Discoveries Section of MBE doi101093molbevmsw221 MBE

85

(supplementary file S2 Supplementary Material online) con-taining 2856 sequences from 782 specimens (238 mbps)that included representatives from 40 of the 43 recognizedtalpid species plus 7 outgroup species using SequenceMatrixv17 (Vaidya et al 2011) In brief we searched GenBank forsequences that were orthologous to those in our 21-genedatasetgt300 bp in length and that included voucher or iso-late numbers New sequences were aligned with the above21-gene dataset using MUSCLE Phylogenetic relationshipswere estimated using RAxML v82 as this program is ableto handle short sequences and place them on the tree usingan MP-based evolutionary placement algorithm (Berger et al2011) The alignment was partitioned by genes and codonpositions We employed a GTR model of sequence evolutionfor searching the best tree and GTRthornC model for boot-strapping The mitochondrial CYT B sequences were sepa-rated into three partitions by codon positions while allnuclear coding genes were separated into two partitions(1stthorn 2nd codon positions were in the same partition)

Tracing Morphological EvolutionWe removed fossil talpid species from a recently publishedmorphological matrix containing 175 characters(Schwermann and Thompson 2015) and used this datasetto construct the most parsimonious (MP) trees using PAUP40a (Swofford 2002) We performed heuristic searches with10000 random addition replicates using the TBR branch-swapping algorithm The characters were optimized usingldquodelayed transformationrdquo on the trees in memory Treelengths were obtained and apomorphy lists were generatedWe then re-ran this analysis (i) using constrained topologiesaccording to our concatenated and coalescent species trees(ii) using a consensus tree whereby only those relationshipsthat were strongly supported (ie PP095 and BS 85) inour concatenated and coalescent species trees estimated us-ing 21 genes were constrained and (iii) with only the sister-relationship of Condylura and Desmanini constrained

Supplementary MaterialSupplementary tables S1ndashS8 figures S1ndashS3 supplementarytexts and files S1 and S2 are available at Molecular Biologyand Evolution online

AcknowledgmentsWe appreciate the three anonymous reviewers for constructivecomments and suggestions and thank Achim Schwermannand Marcelo Sanchez-Villagra for their advice on the morpho-logical transition series We also thank Sharon Birks at the BurkeMuseum of Natural History and Culture University ofWashington and Patricia Gegick at the New Mexico Museumof Natural History for providing tissue samples New sequencesobtained in this study were submitted to GenBank (accessionnumbers KX754466-KX755245) We thank Nancy Halliday forallowing us to use her illustration of Scalopus aquaticus Thisstudy was supported by the National Natural ScienceFoundation of China (no 31301869 to KH) by the Ministryof Education Culture Sports Science and Technology Japan

(no 15770060 to AS) by Natural Sciences and EngineeringResearch Council of Canada Discovery and DiscoveryAccelerator Supplement grants (NSERC 238838 and 412336respectively to KLC) by the National Science Foundation(NSF DEB-1457735 to M S) and by Smithsonian Institution(to KMH)

ReferencesBannikova AA Zemlemerova ED Colangelo P Sozen M Sevindik M

Kidov AA Dzuev RI Krystufek B Lebedev VS 2015 An under-ground burst of diversitymdasha new look at the phylogeny andtaxonomy of the genus Talpa Linnaeus 1758 (MammaliaTalpidae) as revealed by nuclear and mitochondrial genesZool J Linnean Soc 175930ndash948

Bannikova AA Zemlemerova ED Lebedev VS Aleksandrov DY Fang YSheftel BI 2015 Phylogenetic position of the Gansu mole Scapanulusoweni Thomas 1912 and the relationships between strictly fossorialtribes of the family talpidae Dokl Biol Sci 464230ndash234

Bazinet AL Zwickl DJ Cummings MP 2014 A gateway for phylogeneticanalysis powered by grid computing featuring GARLI 20 Syst Biol63812ndash818

Berger SA Krompass D Stamatakis A 2011 Performance accuracy andweb server for evolutionary placement of short sequence reads un-der maximum likelihood Syst Biol 60291ndash302

Cabria MT Rubines J Gomez-Moliner B Zardoya R 2006 On the phy-logenetic position of a rare Iberian endemic mammal the Pyreneandesman (Galemys pyrenaicus) Gene 3751ndash13

Catania KC 2000 Epidermal sensory organs of moles shrew-moles anddesmans a study of the family talpidae with comments on the func-tion and evolution of Eimerrsquos organ Brain Behav Evol 56146ndash174

Chifman J Kubatko L 2014 Quartet Inference from SNP data under thecoalescent model Bioinformatics 303317ndash3324

Chou J Gupta A Yaduvanshi S Davidson R Nute M Mirarab S WarnowT 2015 A comparative study of SVD quartets and other coalescent-based species tree estimation methods BMC Genomics 16S2

Corbet GB Hill JE 1992 The mammals of the Indomalayan region asystematic review London Natural History Museum

Drummond AJ Suchard MA Xie D Rambaut A 2012 Bayesian phyloge-netics with BEAUti and the BEAST 17 Mol Biol Evol 291969ndash1973

Edgar RC 2004 MUSCLE multiple sequence alignment with high accu-racy and high throughput Nucleic Acids Res 321792ndash1797

Edwards LF 1937 Morphology of the forelimb of the mole (Scalopsaquaticus L) in relation to its fossorial habits Ohio J Sci 3720ndash41

Feuda R Bannikova AA Zemlemerova ED Di Febbraro M Loy AHutterer R Aloise G Zykov AE Annesi F Colangelo P 2015Tracing the evolutionary history of the mole Talpa europaeathrough mitochondrial DNA phylogeography and species distribu-tion modelling Biol J Linnean Soc 114495ndash512

Freeman RA 1886 The anatomy of the shoulder and upper arm of themole (Talpa europaea) J Anat Physiol 20201ndash219

Gatesy J Springer MS 2013 Concatenation versus coalescence ver-sus ldquoconcatalescencerdquo Proc Natl Acad Sci U S A110E1179ndashE1179

Gatesy J Springer MS 2014 Phylogenetic analysis at deep timescalesunreliable gene trees bypassed hidden support and the coalescenceconcatalescence conundrum Mol Phylogenet Evol 80231ndash266

He K Jiang X 2014 Sky islands of southwest China I an overview ofphylogeographic patterns Chinese Science Bulletin 59585ndash597

He K Shinohara A Jiang X-L Campbell KL 2014 Multilocus phylogeny oftalpine moles (Talpini Talpidae Eulipotyphla) and its implicationsfor systematics Mol Phylogenet Evol 70513ndash521

Hooker JJ 2016 Skeletal adaptations and phylogeny of the oldest moleEotalpa (Talpidae Lipotyphla Mammalia) from the UK Eocene thebeginning of fossoriality in moles Palaeontology 59195ndash216

Hutchison JH 1968 Fossil talpidae (Insectivora Mammalia) from thelater Tertiary of Oregon Bull Mus Nat Hist Univ Oregon111ndash117

He et al doi101093molbevmsw221 MBE

86

Hutchison JH 1974 Notes on type specimens of European MioceneTalpidae and a tentative classification of old world TertiaryTalpidae (Insectivora Mammalia) Geobios 7211ndash256

Hutchison JH 1976 The Talpidae (Insectivora Mammalia) evolutionphylogeny and classification [PhD dissertation] University ofCalifornia Berkeley

Hutterer R 2005 Order Soricomorpha In Wilson DE Reeder DA edi-tors Mammal species of the world a taxonomic and geographicreference Baltimore John Hopkins University Press p 220ndash311

Kawada S-I Li S Wang Y-X Mock OB Oda S-I Campbell KL 2008Karyotype evolution of shrew moles (Soricomorpha Talpidae) JMammal 891428ndash1434

Kawada S-I Son NT Ngoc Can D 2012 A new species of mole of thegenus Euroscaptor (Soricomorpha Talpidae) from northernVietnam J Mammal 93839ndash850

Kawada S-I Yasuda M Shinohara A Lim BL 2008 Redescription of theMalaysian mole as to be a true species Euroscaptor malayana(Insectivora Talpidae) Mem Natl Mus Nat Sci 4565ndash74

Kawada S Shinohara A Kobayashi S Harada M Oda S Lin LK 2007Revision of the mole genus Mogera (Mammalia LipotyphlaTalpidae) from Taiwan Syst Biodivers 5223ndash240

Kishino H Hasegawa M 1989 Evaluation of the maximum-likelihood estimate of the evolutionary tree topologies fromDNA-sequence data and the branching order in Hominoidea JMol Evol 29170ndash179

Klietmann J Nagel D Rummel M van den Hoek Ostende LW 2015 Agap in digging the Talpidae of Petersbuch 28 (Germany EarlyMiocene) Paleuroaontologische Zeitschrift 89563ndash592

Lanfear R Calcott B Ho SYW Guindon S 2012 PartitionFinder com-bined selection of partitioning schemes and substitution models forphylogenetic analyses Mol Biol Evol 291695ndash1701

Lei FM 2012 Global endemism needs spatial integration Science335284ndash285

Liu Y Liu SY Sun ZY Guo P Fan ZX Murphy RW 2013 A new species ofUropsilus (Talpidae Uropsilinae) from Sichuan China ActaTheriologica Sinica 33113ndash122

Massana KA Beaulieu JM Matzke NJ OrsquoMeara BC 2015 Non-null effects ofthe null range in biogeographic models exploring parameter estima-tion in the DEC Model bioRxiv doi httpdxdoiorg101101026914

Matzke NJ 2014 Model selection in historical biogeography reveals thatfounder-event speciation is a crucial process in Island Clades SystBiol 63951ndash970

McKenna MC Bell SK 1997 Classification of mammals above the spe-cies level New York Columbia University Press

Meier PS Bickelmann C Scheyer TM Koyabu D Sanchez-Villagra MR2013 Evolution of bone compactness in extant and extinct moles(Talpidae) exploring humeral microstructure in small fossorialmammals BMC Evol Biol 1310

Meredith RW Janecka JE Gatesy J Ryder OA Fisher CA Teeling ECGoodbla A Eizirik E Simao TL Stadler T et al 2011 Impacts of theCretaceous terrestrial revolution and KPg extinction on mammaldiversification Science 334521ndash524

Miller MA Schwartz T Pickett BE He S Klem EB Scheuermann RHPassarotti M Kaufman S OrsquoLeary MA 2015 A RESTful API forAccess to Phylogenetic Tools via the CIPRES science gateway EvolBioinform 1143ndash48

Mirarab S Warnow T 2015 ASTRAL-II coalescent-based species treeestimation with many hundreds of taxa and thousands of genesBioinformatics 3144ndash52

Motokawa M 2004 Phylogenetic relationships within the familyTalpidae (Mammalia Insectivora) J Zool 263147ndash157

Nevo E 1979 Adaptive convergence and divergence of subterraneanmammals Annu Rev Ecol Syst 10269ndash308

Nowak RM 1999 Walkerrsquos mammals of the world Baltimore JohnsHopkins University Press

Piras P Sansalone G Teresi L Kotsakis T Colangelo P Loy A 2012Testing convergent and parallel adaptations in talpids humeral me-chanical performance by means of geometric morphometrics andfinite element analysis J Morphol 273696ndash711

Pons J Barraclough TG Gomez-Zurita J Cardoso A Duran DP Hazell SKamoun S Sumlin WD Vogler AP 2006 Sequence-based speciesdelimitation for the DNA taxonomy of undescribed insects Syst Biol55595ndash609

Prum RO Berv JS Dornburg A Field DJ Townsend JP Lemmon EMLemmon AR 2015 A comprehensive phylogeny of birds (Aves)using targeted next-generation DNA sequencing Nature526569ndash573

Qiu Z Storch G 2005 The fossil record of the Eurasian Neogene insec-tivores (Erinaceomorpha Soricomorpha Mammalia) Part I ChinaScripta Geologica Special Issue 537ndash50

Reed CA 1951 Locomotion and appendicular anatomy in three soricoidinsectivores Am Midland Nat 45513ndash671

Sanchez-Villagra MR Horovitz I Motokawa M 2006 A comprehensivemorphological analysis of talpid moles (Mammalia) phylogeneticrelationships Cladistics 2259ndash88

Schwermann AH Thompson RS 2015 Extraordinarily preserved talpids(Mammalia Lipotyphla) and the evolution of fossoriality J VertPaleontol 35e934828

Shimer HW 1903 Adaptations to aquatic arboreal fossorial andcursorial habits in Mammals III Fossorial adaptations Am Nat37819ndash825

Shinohara A Campbell KL Suzuki H 2003 Molecular phylogenetic re-lationships of moles shrew moles and desmans from the new andold worlds Mol Phylogenet Evol 27247ndash258

Shinohara A Suzuki H Tsuchiya K Zhang YP Luo J Jiang XL WangYX Campbell KL 2004 Evolution and biogeography of talpidmoles from continental East Asia and the Japanese islandsinferred from mitochondrial and nuclear gene sequences ZoolSci 211177ndash1185

Springer MS Gatesy J 2016 The gene tree delusion Mol Phylogenet Evol941ndash33

Stamatakis A 2014 RAxML version 8 a tool for phylogenetic analysisand post-analysis of large phylogenies Bioinformatics 301312ndash1313

Swofford DL 2002 PAUP phylogenetic analysis using parsimony ( andother methods) Version 40b

Templeton AR 1983 Phylogenetic inference from restriction endonu-clease cleavage site maps with particular reference to the evolutionof humans and the apes Evolution 37221ndash244

Vaidya G Lohman DJ Meier R 2011 SequenceMatrix concatenationsoftware for the fast assembly of multi-gene datasets with characterset and codon information Cladistics 27171ndash180

van den Hoek Ostende LW Doukas CS Reumer JWF 2005 The fossilrecord of the Eurasian Neogene insectivores (ErinaceomorphaSoricomorpha Mammalia) Part 1 Scripta Geologica Special Issue51ndash300

Van Valen LM 1967 New Paleocene insectivores and insectivore classi-fication Bull Am Mus Nat Hist 135217ndash284

Waddell PJ Shelley S 2003 Evaluating placental inter-ordinal phyloge-nies with novel sequences including RAG1 gamma-fibrinogen ND6and mt-tRNA plus MCMC-driven nucleotide amino acid and co-don models Mol Phylogenet Evol 28197ndash224

Wan T He K Jiang X-L 2013 Multilocus phylogeny and cryptic diversityin Asian shrew-like moles (Uropsilus Talpidae) implications for tax-onomy and conservation BMC Evol Biol 13232

Whidden HP 2000 Comparative myology of moles and the phy-logeny of the Talpidae (Mammalia Lipotyphla) Am Mus Novit32941ndash53

Yates TL 1984 Insectivores elephant shrews tree shrews and dermop-terans In Anderson S Jones JK Jr editors Orders and families ofrecent mammals of the world New York Wiley p 117ndash144

Yates TL Moore DW 1990 Speciation and evolution in the familyTalpidae (Mammalia Insectivora) Prog Clin Biol Res 3351ndash22

Ziegler AC 1971 Dental homologies and possible relationships of recentTalpidae J Mammal 5250ndash68

Ziegler R 2003 Moles (Talpidae) from the late Middle Miocene of SouthGermany Acta Palaeontologica Polonica 48617ndash648

Ziegler R 2012 Moles (Talpidae Mammalia) from Early Oligocene kars-tic fissure fillings in South Germany Geobios 45501ndash513

Intended for the Discoveries Section of MBE doi101093molbevmsw221 MBE

87

• msw221-TF1
• msw221-TF2
• msw221-TF3
• msw221-TF4