Phylogenetic relationships in tribe Spiraeeae (Rosaceae) inferred from nucleotide sequence data

Phylogenetic relationships in tribe Spiraeeae (Rosaceae) inferred

from nucleotide sequence data

D. Potter1, S. M. Still

1, T. Grebenc

2, D. Ballian

3, G. Bozic

2, J. Franjiæ

4, and H. Kraigher

2

1Department of Plant Sciences, University of California, Davis, California, USA2Department for Forest Physiology and Genetics & Research / Program Group: Forest Biology, Ecologyand Technology, Slovenian Forestry Institute, Ljubljana, Slovenia3Faculty of Forestry, University of Sarajevo, Sarajevo, Bosnia and Herzegovina4Faculty of Forestry, University of Zagreb, Zagreb, Croatia

Received March 13, 2006; accepted August 18, 2006Published online: June 28, 2007� Springer-Verlag 2007

Abstract. Tribe Spiraeeae has generally beendefined to include Aruncus, Kelseya, Luetkea,Pentactina, Petrophyton, Sibiraea, and Spiraea.Recent phylogenetic analyses have supportedinclusion of Holodiscus in this group. Spiraea,with 50-80 species distributed throughout thenorth temperate regions of the world, is by farthe largest and most widespread genus in the tribe;the remaining genera have one to several specieseach. Phylogenetic analyses of nuclear ITS andchloroplast trnL-trnF nucleotide sequences for 33species representing seven of the aforementionedgenera plus Xerospiraea divided the tribe into twowell supported clades, one including Aruncus,Luetkea, Holodiscus, and Xerospiraea, the secondincluding the other genera. Within Spiraea, noneof the three sections recognized by Rehder basedon inflorescence morphology is supported asmonophyletic. Our analyses suggest a westernNorth American origin for the tribe, with severalbiogeographic events involving vicariance or dis-persal between the Old and New Worlds havingoccurred within this group.

Key words: Biogeography, Spiraeoideae.

Introduction

The roughly 100 genera and 3,000 speciescurrently accepted as belonging to Rosaceaehave been classified in four (Schulze-Menz1964) to 12 (Takhtajan 1997) subfamilies,many of which have been further subdividedinto tribes, or alternatively, in 17 tribes whichare not grouped in subfamilies (Hutchinson1964). All of these classifications have recog-nized Tribe Spiraeeae (Table 1), originallydescribed (as Spiraeaceae) by de Candolle(1825), whose circumscription of the tribewas even broader than more recent conceptsof subfamily Spiraeoideae (e.g. Schulze-Menz1964), including, as it did, species that are nowclassified in the genera Purshia DC., KerriaDC., GilleniaMoench., NeilliaDon, KageneckiaRuiz & Pav., Quillaja Molina, VauqueliniaCorrrea ex Humb. & Bonpl., Lindleya H. B. &K., and Spiraea L. The last genus was alsobroadly circumscribed, as it was by Linnaeus(1753), so as to include species now assigned toPhysocarpus (Cambess.) Raf., Sorbaria A.

Pl. Syst. Evol. 266: 105–118 (2007)DOI 10.1007/s00606-007-0544-zPrinted in The Netherlands

Plant Systematicsand Evolution

Braun, Aruncus Adans., and Filipendula Mill.in addition to Spiraea s. str.

More recent treatments have adopted con-siderably narrower concepts for both the genusand the tribe. As noted above, many of thespecies included in Spiraea by Linnaeus have,over the last 250 years, been transferred toother genera, many of which are now consid-ered quite distantly related according to bothmorphologically based taxonomy (e.g. Hutch-inson 1964) and recent phylogenetic analysesof molecular data (e.g. Potter et al. 2007). Allof the genera other than Spiraea that arecurrently recognized within Spiraeeae(Table 1) also had their origins in the typegenus for the tribe. The taxonomic history ofthe North American genera was thoroughlyreviewed by Henrickson (1985), who alsoprovided the most recent generic addition tothe group when he placed the two Mexicanspecies of Spiraea, S. hartwegiana Rydb. andS. northcraftii I. M. Johnston, in synonymyand transferred them to a new genus, Xero-spiraea (asX. hartwegiana (Rydb.)Henrickson).

Schulze-Menz (1964) included Spiraea,Sibiraea, and Aruncus in Spiraeeae and char-acterized members of the tribe as shrubs orperennial herbs lacking stipules with (2)–5–(8)free carpels, membranous seed coats, andendosperm scant or lacking. Both Hutchinson(1964) and Takhtajan (1997) included in Spir-

aeeae those three genera plus Kelseya, Luetkea,and Petrophyton. Takhtajan (1997) also listedPentactina, considered a synonym of Spiraeaby Hutchinson (1964), while the latter authorincluded Apopetalum Pax, now considered asynonym of Brunellia Ruiz & Pav. (Brunellia-ceae), in the tribe.

Spiraea has been variously divided bydifferent authors into subgenera, sections,series, and cycles (e.g. Poyarkova 1939). Inflo-rescence morphology has been emphasized inmost of these groupings, as reflected in thewidely accepted classification followed byRehder (1940), which recognizes three sections(Table 1): Spiraria Ser. (=Spiraea), with pan-icles, Calospira, with compound corymbs, andChamaedryon with simple corymbiform orumbellate inflorescences.

Beginning with Morgan et al.’s (1994)study of relationships across Rosaceae basedon sequences of the chloroplast gene rbcL,several molecular phylogenetic analyses havesupported the monophyly of Spiraea andAruncus plus Holodiscus Maxim., formerlyclassified in tribe Holodisceae due to thedifferent fruit type (achenes in the latter groupvs. follicles in the others). The most recentmolecular phylogenetic study of Rosaceae(Potter et al. 2007), based on multiple nuclearand chloroplast genes, strongly supported theinclusion of Kelseya, Luetkea, and Petrophyton

Table 1. Characteristics of genera of Spiraeeae

Genus and

Section

Number

of Species

Habit Leaves Inflorescence Distribution

Aruncus Adans. 1 perennial herb 2-3 pinnate panicle n tempHolodiscus Maxim. 5 erect shrub simple serrate panicle w N Am - n S AmKelseya Rydb. 1 cushion plant simple entire solitary MT, WYLuetkea Bong. 1 trailing subshrub biternate raceme w N AmPentactina Nakai 1 erect shrub simple serrate raceme KoreaPetrophyton Rydb. 4 prostrate shrub simple entire raceme w N AmSibiraea Maxim. 5 erect shrub simple entire panicle se Eu - w AsiaSpiraea L. 50-80 erect shrub simple serrate variable n tempSection Spiraea 10-20 erect shrub simple serrate panicle e/w N Am, Eu, AsiaCalospira K. Koch 20-30 erect shrub simple serrate corymb Eu, Asia, e/w N AmChamaedryon Ser. 20-30 erect shrub simple serrate umbel Eu, AsiaXerospiraea Henr. 1 erect shrub simple entire rac./pan. Mexico

106 D. Potter et al.: Phylogeny of Spiraeeae

in Spiraeeae. The remaining genera have notbeen included in any previously publishedmolecular phylogenetic study of the family.

We undertook molecular phylogeneticanalyses of this group using chloroplast trnL-trnF and nuclear rDNA ITS (including ITS1,5.8S rRNA gene, and ITS2) sequences forrepresentatives of eight genera of Spiraeeae,including 24 species of Spiraea, two specieseach of Petrophyton and Sibiraea, and oneeach of Aruncus, Holodiscus, Kelseya, Luetkea,and Xerospiraea, plus two outgroups selectedbased on results of family-level phylogeneticanalyses (Potter et al. 2007; Table 2). Thesequences were analyzed phylogenetically inorder to address the following questions:

1) Is monophyly of Spiraeeae including theseeight genera supported?

2) What are the relationships among generawithin the tribe?

3) Is Spiraea, the largest and most variablegenus in the tribe, with 50–80 speciesdistributed throughout the north temperateregions of the world, supported as amonophyletic group?

4) Is Rehder’s (1940) division of Spiraea intothree sections based upon inflorescence type(Table 1) supported?

5) What do the phylogenies based on molec-ular data suggest about historical biogeog-raphy and morphological evolution withinthe group?

Materials and methods

Thirty-eight accessions, representing 24 species ofSpiraea, nine species of other genera of Spiraeeae,and two outgroups, were sampled for this study(Table 2). Specimens were collected from the wildor provided by botanical gardens; species identifi-cations were verified by reference to publisheddescriptions (e.g. Rehder 1940). Fresh material wasused in all cases except Xerospiraea hartwegiana,for which DNA was extracted from an herbariumspecimen with permission of the curator of TEX.Voucher specimens for other taxa are deposited atDAV.

Total DNA was extracted from one to threeaccessions of each of the species examined using amodified CTAB protocol (Doyle and Doyle 1987)in which the RNase step was omitted. The nuclearITS region was amplified using primers ITS6(5’tcgtaacaaggtttccgtaggtga3’) and ITS9(5’ccgcttattgatatgcttaaac3’) designed by Sang-HunOh and published here for the first time. Thechloroplast trnL-trnF region was amplified usingprimers trnc and trnf (Taberlet et al. 1991). PCRamplification and bidirectional sequencing, usingthe same primers, were performed as previouslydescribed (Bortiri et al. 2001); in a few cases, oneor more of the internal primers ITS2, ITS3 (Whiteet al. 1990), trnd and trne (Taberlet et al. 1991)were used to obtain clear full-length sequencedata.

Sequences were edited with SequencherTM

(Gene Codes Corporation) and aligned in Clu-stalX (Thompson et al. 1997); alignments wereadjusted manually. Phylogenetic analyses basedon maximum parsimony were implemented inPAUP* (Swofford 2002). All positions wereweighted equally; gaps were treated as missingvalues except for several phylogenetically infor-mative indels in the trnL-trnF data, which werecoded as binary characters and added to the datamatrix. The partition homogeneity test, imple-mented in PAUP* with 1000 test replicates,maxtrees set to 100, and heuristic searches usingthe TBR branch-swapping algorithm and 10random taxon addition replicates per test repli-cate, was used to test for significant conflictbetween ITS and trnL-trnF data. The combineddata set was analyzed using the same searchalgorithm but with 1000 replicates of randomtaxon addition and maxtrees allowed to increaseautomatically as necessary. In the three cases forwhich we were unable to obtain sequence datafor one of the regions for a particular accession(Table 2), missing values were coded for thattaxon for that region (ITS for Kelseya uniflora,trnL-trnF for Sibiraea croatica and Xerospiraeahartwegiana). Relative support for clades wasassessed using 1000 bootstrap replicates with 10random taxon addition replicates per bootstrapreplicate and maxtrees set at 100.

Bayesian analyses, using a model of sequenceevolution selected in MrAIC (Nylander 2005),and with binary indel characters excluded, wereimplemented in MrBayes 3.1.1 (Huelsenbeck and

D. Potter et al.: Phylogeny of Spiraeeae 107

Table

2.Accessionsincluded

inthisstudy

GenusandSpecies

Source

Acc./Coll.

Number

Section

(Rehder)

Distribution

ITSGenbank

Accession#

trnGenbank

Accession#

[Outgroups:]

Adenostomafasciculatum

Hook.&

Arn.

Yolo

Co.,CA

S.Oh970424-01na

CA,Baja

CA

DQ88358

AF348535

Gillenia

stipulata

(Muhl.

exWilld.)Baillon

Berkeley

Bot.Gard.

92.0438

na

eN

Am

DQ811763

AF348554

[Spiraeeae:]

Aruncusdioicus(W

alter)

Fern.

Berkeley

Bot.Gard.

83.0466

na

NAm,Eur

DQ897602

AF196868

Holodiscusmicrophyllus

Rydb.

EldoradoCo.,CA

D.Potter

060711-01

na

wN

Am

DQ897603

DQ897573

Kelseyauniflora

(Wats.)Rydb.

Lew

is&

Clark

Co.,

MT

D.Barton2218

na

MT,WY

na

DQ851232

Luetkea

pectinata

(Pursh)

Kuntze

Whatcom

Co.,WA

D.Morgan2284na

wN

Am

DQ851235

DQ851233

Petrophytoncaespitosum

(Nutt.)Rydb.

SheridanCo.WY

D.Potter

020906-02

na

wN

Am

DQ851236

DQ851234

Petrophytonhendersonii

(Canby)Rydb.

Olympic

Mts.,WA

M.Loper,s.n.

na

WA

DQ897604

DQ897574

Sibiraea

croatica

Degen

Velebit,Croatia

SIB

11/1

na

eEur-w

As

AJ876553

na

Sibiraea

laevigata

(L.)Maxim

.Roy.Bot.Gard.

Edinburgh

19910654

na

eEur-w

As

DQ897605

DQ897575

Xerospiraea

hartwegiana

Henrickson

Puebla,Mexico

J.L.Panero

5805

na

Mexico

DQ897606

na

Spiraea

blumei

Don.

Arnold

Arboretum

748–94

ChamaedryonJapan,Korea

DQ897607

DQ897576

Spiraea

canescensD.Don.

Arnold

Arboretum

1229–85

Calospira

Him

alaya

DQ897608

DQ897577

Spiraea

cantoniensisLour.

UC

DavisArboretum

ChamaedryonChina,Japan

DQ897609

DQ897578

Spiraea

crenata

L.

Arnold

Arboretum

1398–80

Chamaedryonse

Eur-cAs

DQ897610

DQ897579

Spiraea

decumbensW.Koch.

‘WhiteLace’

Holden

Arboretum

96–270

Calospira

sEur

DQ897611

DQ897580

Spiraea

densiflora

Nutt.

PlacerCo.,CA

D.Potter

970619-02

Calospira

BC

-MT,WY,

&OR

DQ88362

AF348571

Spiraea

douglasiiHook.

Arnold

Arboretum

319–85

Spiraria

BC

–CA

DQ897612

DQ897581

108 D. Potter et al.: Phylogeny of Spiraeeae

Spiraea

form

osanaHayata

Arnold

Arboretum

261–98

Calospira

Taiwan

DQ897613

DQ897582

Spiraea

fritschianaSchneid.

Arnold

Arboretum

307–69

Calospira

cChina–Korea

DQ897614

DQ897583

Spiraea

hypericifoliaL.

Arnold

Arboretum

165–95

Chamaedryonse

Eur-cAs

DQ897615

DQ897584

Spiraea

japonicaL.f.1

Arnold

Arboretum

249–2000

Calospira

Japan

DQ897616

DQ897585

Spiraea

japonica2‘N

eonFlash’Holden

Arboretum

84–139

Calospira

Japan

DQ897617

DQ897586

Spiraea

lasiocarpaKar.&

Kir.

Arnold

Arboretum

841–90

ChamaedryonRussia

DQ897618

DQ897587

Spiraea

latifolia(A

it.)Borkh.

Arnold

Arboretum

116–79

Spiraria

neCanada–NC

DQ897619

DQ897588

Spiraea

longigem

misMaxim

.Arnold

Arboretum

291–84

Calospira

nw

China

DQ897620

DQ897589

Spiraea

miyabei

Koidz.

Holden

Arboretum

85–428

Calospira

cChina

DQ897621

DQ897590

Spiraea

nipponicaMaxim

.Arnold

Arboretum

1024–86

ChamaedryonJapan

DQ897622

DQ897591

Spiraea

prunifoliaSieb.&

Zucc.Holden

Arboretum

80–652

ChamaedryonKorea,China,Taiwan

DQ897623

DQ897592

Spiraea

pubescensTurcz.

Arnold

Arboretum

541–83

ChamaedryonnChina

DQ897624

DQ897593

Spiraea

salicifoliaL.

Arnold

Arboretum

525–89

Spiraria

seEur-neAsia&

JapanDQ897625

DQ897594

Spiraea

thunbergiiSieb.

Arnold

Arboretum

1226–85

ChamaedryonJapan,China

DQ897626

DQ897595

Spiraea

trichocarpaNakai

Holden

Arboretum

85–127

ChamaedryonKorea

DQ897627

DQ897596

Spiraea

trilobata

L.

Arnold

Arboretum

1052–59

ChamaedryonnChina–cAs

DQ897628

DQ897597

Spiraea

veitchiiHem

sl.

Arnold

Arboretum

1767–80

Calospira

c&

wChina

DQ897629

DQ897598

Spiraea

virginianaBrit.1

Arnold

Arboretum

414–93

Calospira

VA

-NC

&TN

DQ897630

DQ897599

Spiraea

virginiana2

Arnold

Arboretum

415–93

Calospira

VA

-NC

&TN

DQ897631

DQ897600

Spiraea

virginiana3

Arnold

Arboretum

416–93

Calospira

VA

-NC

&TN

DQ897632

DQ897601

D. Potter et al.: Phylogeny of Spiraeeae 109

Ronquist 2001). Double analyses were run withfour chains for 4,000,000 generations, samplingevery 10 generations. Burn-in was set to 1,000,000generations. The sampled trees from both analy-ses were pooled and the majority-rule consensustree was constructed from the resulting 600,000trees to estimate Bayesian posterior probabilityvalues.

Inflorescence types and geographic distribu-tions based on published reports (Rehder 1940,Hutchinson 1964; Tables 1 and 2) were scored asmultistate characters and MacClade 3.08 (Madd-ison and Maddison 1999) was used to mapcharacter states onto several of the most parsi-monious trees (see Results). Due to weak resolu-tion among tribes of Rosaceae obtained in otheranalyses (e.g. Potter et al. 2007) and resultinguncertainty about the precise phylogenetic posi-tion of Spiraeeae in the family, the outgroups weretreated in two ways - first with missing values andthen with their actual states - for these recon-structions.

Results

The ITS alignment included 778 characters, ofwhich 421 were constant, 121 variable butuninformative, and 236 were phylogeneticallyinformative. The trnL-trnF alignment included1,040 sites, to which 5 indel characters wereadded, yielding a dataset of 1,045 characters,of which 876 were constant, 102 variable butuninformative, and 67 were phylogeneticallyinformative. The partition homogeneity testsrevealed no significant conflict between ITSand trnL-trnF data (p = 0.567). The JC69model, determined by MrAIC to be the bestmodel for both partitions by all criteria (AIC,AICc, and BIC) was used for Bayesian anal-yses of the combined data set. The averagestandard deviation of split frequencies hadfallen below 0.01 by generation 91,000, indi-cating that the runs had reached stationarityand that the burn-in value of 1,000,000generations was more than adequate; thisconclusion was also supported by inspectionof the log-likelihood values of the cold chains,which showed fluctuations within a stablerange near the maximum values by 9,000generations in both runs.

Phylogenetic analysis of the combineddata set of 1,823 characters produced 192equally parsimonious trees. The strict con-sensus tree (Fig. 1) showed that, as expected,the eight genera of Spiraeeae form a stronglysupported monophyletic group, although ourpower to test the monophyly of the tribe wasadmittedly limited, with only two outgroupsincluded (see Discussion). Strong supportwas also found for the monophyly of eachof the following genera (number of speciessampled/total number) Spiraea (24/50–80),Petrophyton (2/4), and Sibiraea (2/5). Sup-port was weak to moderate from bootstrapanalysis and strong from Bayesian analysisfor the sister relationship between Aruncusand Luetkea and for that between Holodiscusand Xerospiraea; both analyses providedstrong support for monophyly of the cladeincluding these four genera and for its sisterclade including the other four. Within thelatter clade, Sibiraea diverged first, a positionwith weak bootstrap and strong Bayesiansupport, followed by Petrophyton; Kelseyaand Spiraea were weakly supported as sistertaxa.

Within Spiraea, S. decumbens was resolvedas sister to the rest of the genus; the mono-phyly of the remaining species was supportedweakly (49%) by bootstrap analysis butstrongly (posterior probability 99) by Bayes-ian analysis. Each of the following clades wassupported with 70% or better bootstrapsupport (except number 3) and 95% or betterBayesian posterior probability: 1) S. blumei,S. cantoniensis, S. pubescens, and S. trilobata;2) S. canescens, S. crenata, and S. longigem-mis; 3) 2 plus S. lasiocarpa (only 65%bootstrap support); 4) S. prunifolia plusS. thunbergii; 5) 4 plus S. trichocarpa; 6) 3plus 5 plus S. veitchii; 7) S. hypericifolia plusS. nipponica; 8) 6 plus 7; 9) 1 plus 6 plus 7;10) S. densiflora and S. douglasii; 11)S. formosana, S. fritschiana, S. japonica, andS. miyabei; 12) S. latifolia and S. salicifolia;13) three accessions of S. virginiana. Somerelationships resolved in some of the mostparsimonious trees (e.g. Figs. 2 and 3) were

110 D. Potter et al.: Phylogeny of Spiraeeae

not well supported by the bootstrap and/orthe Bayesian analyses. For example, a sisterrelationship between clades 12 and 13 (as inFig. 2) was weakly supported (49%) bybootstrap analysis but was not consistentwith the majority-rule consensus tree fromthe Bayesian analysis, while a sister relation-

ship between S. virginiana (clade 13) andclade 9 (as in Fig. 3) was moderately sup-ported (posterior probability 93) by Bayesiananalysis but was not consistent with themajority-rule consensus tree from the boot-strap analysis. Finally, a sister relationshipbetween clades 10 and 12 (as in Fig. 3) was

Fig. 1. Strict consensus of 192 most parsimonious trees (l=1,007, ci excluding autapomorphies=.64, ri=.78)from phylogenetic analysis of nuclear ITS and chloroplast trnL-trnF sequences from representative species ofSpiraeeae. Parsimony bootstrap and Bayesian posterior probability support values are shown above and belowbranches, respectively

D. Potter et al.: Phylogeny of Spiraeeae 111

strongly supported (posterior probability 98)by Bayesian analysis but was not consistentwith the majority-rule consensus tree from thebootstrap analysis.

Geographic distribution and inflorescencetypes were coded as multistate characters andmapped onto representatives of the mostparsimonious trees in order to explore how

differences in topology among those treeswould affect optimization of the ancestralstates for Spiraeeae and Spiraea (Figs. 2 and3). All topologies resolved western NorthAmerica as the ancestral area for the tribe,and this was true regardless of how theoutgroups were coded (see Materials andmethods). All topologies also required multi-

Fig. 2. One of 192 most parsimonious trees (l=1,007, ci excluding autapomorphies=.64, ri=.78) fromphylogenetic analysis of nuclear ITS and chloroplast trnL-trnF sequences from representative species ofSpiraeeae. Geographic distribution was coded as an unordered multistate character and optimized on the treeusing MacClade (see text)

112 D. Potter et al.: Phylogeny of Spiraeeae

ple independent vicariant events involving theOld and New Worlds, and several indepen-dent migrations between Europe, western/central Asia, and eastern Asia, within thetribe, with the possibility that the differentevents may have proceeded in different direc-tions. The ancestral area for Spiraea wasresolved as western North America in sometopologies (e.g. Fig. 2) but was equivocal inothers (e.g. Fig. 3; optimization not shown).Within Spiraea, all trees supported thehypothesis of a single migration into easternAsia followed by several independent migra-tions into western Asia and/or Europe. Alltrees also supported the sister relationship ofthe S. latifolia, from eastern North America,and S. salicifolia, widespread in Europe andAsia. Some topologies (e.g. Fig. 2) supportedan eastern North American origin for thispair of species, while in other topologies theoptimization of distribution at the nodejoining them was equivocal. Moreover,variation in the position of those two taxaand that of S. virginiana (compare Figs. 2 and3) resulted in some topologies suggesting thatthe most recent common ancestor of botheastern North American species sampled hereoccurred in western North America (e.g.Fig. 2), others suggesting that S. virginianawas derived from an Asian ancestor(e.g. Fig. 3; optimization not shown), andstill others in which the ancestral area for thetwo eastern North American species plus S.salicifolia was equivocal (not shown). Thesecond of these topologies was favored by theBayesian analysis, which, as mentionedabove, placed S. virginiana in the positionshown in Fig. 3.

While some of the clades mentioned aboveconsisted of members of just one of thesections recognized by Rehder (1940), noneof those three sections was supported asmonophyletic (Fig. 3). Reconstruction of theancestral inflorescence type of Spiraeeae wassensitive to outgroup coding; when outgroupswere coded as missing values for this character,raceme was the ancestral state for the tribe(e.g. Fig. 3), but when the outgroups were

coded as having panicles, the true conditionfor both Adenostoma and Gillenia, that wasoptimized as the ancestral condition in Spira-eeae. The ancestral state for Spiraea wasoptimized as a compound corymb for all treesusing the first outgroup coding and for sometrees using the second; in the remaining trees,the ancestral state for the genus was equivocal.In all cases, all three of the inflorscence typesfound in Spiraea (compound corymb, panicle,and simple umbel) had to be gained at leasttwice or lost at least once within in the genus(Fig. 3).

Discussion

This is the most inclusive molecular phyloge-netic study of Spiraeeae to date, includingrepresentatives of eight genera and multiplespecies of several of them, thereby allowing usto assess simultaneously the monophyly of thetribe and of each genus, the relationshipsamong the genera, and relationships withinSpiraea. Although the inclusion of only twooutgroups here allowed only a limited test ofthe first of these hypotheses based on thisstudy alone, our results, in combination withthose of previous phylogenetic analyses ofrelationships across Rosaceae (e.g. Potteret al. 2007) provide strong support formonophyly of Spiraeeae, including Aruncus,Holodiscus, Kelseya, Luetkea, Petrophyton,Sibiraea, Spiraea, and Xerospiraea. Our resultsalso support recognition of all eight of these asdistinct genera: Spiraea, from which we sam-pled 24 species representing all three ofRehder’s (1940) sections and the full geo-graphic range of the genus, was stronglysupported as monophyletic, and none of theother genera was nested within it.

The circumscription of Spiraeeae sup-ported here is not a surprising result. All ofthese genera, with one exception, have beenincluded in the tribe in recent infrafamilialclassifications of Rosaceae (Hutchinson 1964,Takhtajan 1997). The single exception isHolodiscus, strongly supported as part ofthis group by all molecular phylogenetic

D. Potter et al.: Phylogeny of Spiraeeae 113

studies (e.g. Morgan et al. 1994, Potter et al.2002, Potter et al. 2007) but classified in itsown tribe in most treatments due primarilyto its indehiscent fruits. A relationshipbetween Holodiscus and Spiraeeae was sug-gested by some previous authors, however.Watson (1890a) favored maintaining Holo-

discus as a section within Spiraea rather thantreating it as a separate genus, although hedid recognize the distinctness of the othergenera of the tribe. He noted that theachenes of Holodiscus are unlike those ofmost genera of Rosoideae because theydevelop from carpels with two ovules rather

Fig. 3. One of 192 most parsimonious trees (l=1,007, ci excluding autapomorphies=.64, ri=.78) fromphylogenetic analysis of nuclear ITS and chloroplast trnL-trnF sequences from representative species ofSpiraeeae. Inflorescence type was coded as an unordered multistate character and optimized on the tree usingMacClade (see text). Assignments of species of Spiraea to sections (Rehder 1940) are indicated

114 D. Potter et al.: Phylogeny of Spiraeeae

than one and because the fruits sometimesdehisce tardily and in any case open easilyalong the ventral suture when dissected. Thelatter observation was also made by Schulze-Menz (1964), who placed tribe Holodisceae,including only Holodiscus, near Spiraeeae inSpiraeoideae. Molecular phylogenetic analy-ses of Rosaceae, beginning with that ofMorgan et al. (1994) have repeatedly shownthat fruit type alone is not a reliable indica-tor of relationship in the family.

Material of Pentactina was not availablefor inclusion in this study. Hutchinson (1964)considered this monotypic Korean genus to bea synonym of Spiraea, but Schulze-Menz(1964) and Takhtajan (1997) recognized it asdistinct. Including this species in future studieswill be important to establish with certaintythe number of genera that should be recog-nized in Spiraeeae.

What, then, are potential non-molecularsynapomorphies for Spiraeeae? Two condi-tions characterize all members of the tribe andare not found in any of the putatively closelyrelated clades in the family: lack of stipules andunitegmic ovules (R. Evans, pers. comm.).Both of these conditions are found in otherclades of Rosaceae as well, but they appear tobe independently derived. Other characterssuggested as potential synapomorphies byEvans and Dickinson (1999), such as multiplepistils and apical epitropic ovules, are notresolved as such based on the most recentphylogenetic analyses of the family (Potteret al. 2007).

Henrickson (1985) conducted a cladisticanalysis of morphological characters for sixgenera of Spiraeeae, and obtained results quitedifferent from those presented here. His anal-ysis suggested that Spiraea is a paraphyleticgroup with subgenus Spiraea (containing sec-tions Spiraea and Calospira) sister to Sibiraeaand subgenus Metaspiraea Nakai (containingsection Chamaedryon) sister to a clade includ-ing Kelseya, Luetkea, Petrophyton, and Xero-spiraea. In our analyses (Figs. 1 and 2), theeight genera were divided into two wellsupported clades, one including Aruncus,

Luetkea, Holodiscus, and Xerospiraea; thesecond including Sibiraea, Petrophyton, Kel-seya, and Spiraea. We attribute the differencesbetween our results and those of Henrickson(1985) to the limited number of morphologicalcharacters that exhibit potentially informativepatterns of variation among genera of Spira-eeae and to homoplasy in at least some of thecharacters (e.g. inflorescence type, growth ha-bit) that are important in their classification.The challenge now before us is to identify non-molecular characters that support the relation-ships resolved by nucleotide sequence data.

We are not aware of any morphological,anatomical, or biochemical characters thatsupport the division of the tribe into twoclades of four genera each. Within the first ofthese clades, two subclades were resolved, oneincluding Holodiscus and Xerospiraea, thesecond including Aruncus and Luetkea. A closerelationship between Holodiscus and Xerospi-raea was suggested by Watson (1890a) when hestated that the Mexican Spiraea parvifolia(a synonym of X. hartwegiana) was moreappropriately placed in Spiraea section Holo-discus than in section Petrophytum, where ithad been placed by Maximowicz (1879).

The leaves are simple in all members of thetribe except Luetkea, where they are twiceternately dissected, and Aruncus, where theyare 2–3 times pinnately compound. Thus,divided leaves may be a synapomorphy ofthese two genera. Recent advances in under-standing of the genetics of leaf development(Bharathan et al. 2002, Kim et al. 2003) mayprovide exciting opportunities to test thishypothesis.

Reduced growth habits characterize mem-bers of three genera in Spiraeeae, Kelseya,Luetkea, and Petrophyton, all of which arefound only in western North America (Ta-ble 1). Watson, (1890a, b) classified these taxain three sections, Eriogynia, Kelseya, andPetrophytum of the genus Eriogynia Hook.,first established (Hooker 1834) to accommo-date just one species, E. pectinata (Pursh)Hook. The latter species, however, was laterdetermined by Kuntze (1891) to be synony-

D. Potter et al.: Phylogeny of Spiraeeae 115

mous with the earlier-named Luetkea sibbal-dioides Bongard; thus, the correct name forthe taxon is Luetkea pectinata Kuntze. Ryd-berg (1900) elevated Watson’s other twosections of Eriogynia to generic level (butwith a change in spelling in one case), therebyrecognizing the currently accepted generaPetrophyton and Kelseya, in addition toLuetkea.

Our analyses support the separation ofthe three genera and suggest that evolutionof a reduced growth habit has occurred atleast twice within Spiraeeae, with one eventproducing trailing subshrubs in Luetkea, andat least one resulting in the rosette-formingshrublets found in Petrophyton and Kelseya.The tree topologies recovered in our analysessuggest that the reduced habit may have beensecondarily lost in Spiraea (this could be asynapomorphy for the genus), but relation-ships among Kelseya, Petrophyton, Sibiraea,and Spiraea are not well supported andadditional studies may reveal a sister rela-tionship between Kelseya and Petrophyton, aswas found in Potter et al.’s (2007) multigeneanalysis of Rosaceae (which, however, wasbased on more limited sampling withinSpiraeeae).

Parsimony-based character reconstructionssuggested that the common ancestor of Spir-aeeae occurred in western North America,with independent migrations to the Old Worldoccurring in Aruncus, Sibiraea, and Spiraea.The ancestral area for Spiraea could not bereconstructed unequivocally based on ourresults, but a complex biogeographic historyof the genus, involving multiple dispersaland/or vicariant events between the Old andNew Worlds, and several independent migra-tions between Europe, western/central Asia,and eastern Asia, with the possibility that thedifferent events may have proceeded in differ-ent directions, are suggested.

Inflorescence type, the basis for the rec-ognition of three sections within Spiraea(Rehder 1940) does not appear to be areliable indicator of relationship within thegenus (Fig. 3), but some correlations were

nonetheless observed. While compound cor-ymbs, characteristic of section Calospira, arefound in S. decumbens, here resolved as sisterto the rest of Spiraea and in several otherclades within the genus, simple umbels (sec-tion Chamaedryon) are restricted to onestrongly supported clades and panicles (sec-tion Spiraea) are found in one clade (stronglysupported by Bayesian analysis) in some ofthe most parsimonious trees and two inothers. This suggests that compound corymbsmay be the ancestral inflorescence type for thegenus, and that each of the other types mayhave evolved once or twice, with, correspond-ingly, one to several reversals to the ancestralstate. This hypothesis is supported by char-acter state mapping for the tribe (Fig. 3).

Our results also suggest that, while inflo-rescence type alone may not predict relation-ship, many of the species of Spiraea that havepreviously been classified as close to oneanother based on morphology are in factclosely related. Examples include the groupof S. blumei, S. cantoniensis, S. pubescens, andS. trilobata, which appear together withinsection Chamaedryon in Rehder’s (1940) treat-ment and in Lingdi and Alexander’s (2003)treatment for the Flora of China, which doesnot divide the genus into sections. Besidesumbellate inflorescences, these four species,plus several others not sampled in this study,share the condition of stamens shorter than orsubequaling the petals (Rehder 1940, Lingdiand Alexander 2003). Spiraea prunifloia and S.thunbergii, here resolved as sister taxa, alsoappear together in both of the aforementionedtreatments, based on the combination of sessileumbels, serrate-dentate leaves, and stamens 1/3 –1/2 as long as the petals. In addition, Lingdiand Alexander (2003) list S. formosana as closeto S. japonica; the two species differ only indetails of pubescence and serration of theleaves, and the former has sometimes beentreated as a variety of the latter (Masamune1932).

Due to limited taxon sampling withinSpiraea (24 out of an estimated 50–80 species)and the weak support for resolution of rela-

116 D. Potter et al.: Phylogeny of Spiraeeae

tionships among major clades of species in ouranalyses, it is difficult to draw strong conclu-sions about phylogeny and character evolutionin the genus. We offer the foregoing aspreliminary hypotheses which should be testedwith additional sampling of both taxa andcharacters. Based on the results of phyloge-netic analyses of such expanded data sets, anew infrageneric classification for Spiraeashould eventually be proposed.

We thank Drake Barton and Kathy Lloyd(Montana), Tom Ward and Irina Kadis (ArnoldArboretum) and Ethan Johnson (Holden Arbore-tum), and the curator and staff of TEX forproviding plant material, and Seema Doshi,Michael Steinwand, and the staff of the UC DavisDBS DNA Sequencing Facility (Sheryl Bernauer,Kerry Cloud, and Shelley Williams), for technicalassistance. We gratefully acknowledge financialsupport from the Systematic Biology program ofthe National Science Foundation (Award No.DEB-0089662 to DP).

References

Bharathan G., Goliber T. E., Moore C., KesslerS., Pham T., Sinha N. R. (2002) Homologies inleaf form inferred from KNOXI gene expres-sion during development. Science 296: 1858–1860.

Bortiri E., Oh S., Jiang J., Baggett S., Granger A.,Weeks C., Buckingham M., Potter D., Parfitt D.(2001) Phylogeny and systematics of Prunus(Rosaceae) as determined by sequence analysisof ITS and the chloroplast trnL-trnF spacerDNA. Syst. Bot. 26: 797–807.

de Candolle A. P. (1825) Prodromus systematisnaturalis regni vegetabilis, sive, Enumeratiocontracta ordinum generum specierumque plan-tarum huc usque cognitarium, juxta methodinaturalis, normas digesta. 2: 541. Treuttel etWurtz, Paris.

Doyle J. J., Doyle J. L. (1987) A rapid DNAisolation procedure for small quantities of freshleaf tissue. Phytochem. Bull. 19: 11–15.

Evans R. C., Dickinson T. A. (1999) Floral ontog-eny and morphology in subfamily SpiraeoideaeEndl. (Rosaceae). Int. J. Pl. Sci. 160: 981–1012.

Henrickson, J. (1985) Xerospiraea, a generic segre-gate of Spiraea (Rosaceae) from Mexico. Aliso11: 199–211.

Hooker, W. J. (1834) Flora Boreali-Americana. I.H. G. Bohn, London.

Huelsenbeck J. P., Ronquist F. (2001) MRBAYES:Bayesian inference of phylogenetic trees. Bioin-formatics 17: 754–755.

Hutchinson J. (1964) The genera of floweringplants, vol. 1, Dicotyledons. Clarendon Press,Oxford.

Kim M., Pham T., Hamidi A., McCormick S.,Kuzoff R. K., Sinha N. (2003) Reduced leafcomplexity in tomato wiry mutants suggests arole for PHAN and KNOX genes in generatingcompound leaves. Development 130: 4405–4415.

Kuntze, O. (1891) Revisio Generum Plantarum 1:1–375.

Lingdi L., Alexander C. (2003) Spiraea. In:Zhengyi W., Raven P. H., Deyuan H. (eds.)Flora of China, vol. 9. Missouri BotanicalGarden Press, St. Louis, pp. 47–73.

Linnaeus C. (1753) Species Plantarum I, 1st ed.Stockholm.

Maddison W. P., Maddison D. R. (1999) MacClade,version 3.08. Analysis of phylogeny and charac-ter evolution. Sinauer Associates, Inc., Sunder-land, Massachusetts.

Masamune G. (1932) Genera plantarum formos-amarum. Annual Rep. Taihoku Bot. Gard. 2:123.

Maximowicz C. J. (1879) Adnotationes de spirae-aces.Trudy Imp. S.-Peterburgsk. Bot. Sada 6:105–261.

Morgan D. R., Soltis D. E., Robertson K. R.(1994) Systematic and evolutionary implicationsof rbcL sequence variation in Rosaceae. Amer. J.Bot. 81: 890–903.

Nylander J. A. A. (2005) MrAIC, version 1.4.,available at http://www.abc.se/�nylander/.

Potter D., Eriksson T., Evans R. C., Oh S.,Smedmark J., Morgan D. R., Kerr M., Robert-son K. R., Arsenault M., Campbell C. S. (2007)Phylogeny and classification of Rosaceae. Pl.Syst. Evol. 266: 5–43.

Potter D., Gao F., Bortiri P. E., Oh S., Baggett S.(2002) Phylogenetic relationships in Rosaceaeinferred from chloroplast matK and trnL-trnFnucleotide sequence data. Pl. Syst. Evol. 231:77–89.

D. Potter et al.: Phylogeny of Spiraeeae 117

Poyarkova A. I. (1939) Spiraeoideae. In: BorisovaA. G., Komarov V. L., Krishtofovich A. N.,Lozina-Lozinskaya A. S., Maleev V. P., PalibinI. V., Poyarkova A. I., Tsinzerling Yu. D.,Yuzepchuk S. V. (eds.) Flora of the U.S.S.R.Izdatel-stvo Akademii Nauk SSSR, Moscow, pp.216–245.

Rehder A. (1940) Manual of cultivated trees andshrubs. Dioscorides Press, Portland.

Rydberg P. A. (1900) Catalogue of the flora ofMontana. Mem. New York Bot. Gard. 1: 1–492.

Schulze-Menz G. K. (1964) Rosaceae. In: MelchiorH. (ed.) Engler’s Syllabus der PflanzenfamilienII, 12th ed. Gebruder Borntraeger, Berlin,pp. 209–218.

Swofford D. L. (2002) PAUP* Phylogenetic Anal-ysis Using Parsimony (* and Other Methods)Version 4. Sinauer Associates, Sunderland, Mas-sachusetts.

Taberlet P., Gielly L., Pautou G., Bouvet J. (1991)Universal primers for amplification of three non-coding regions of chloroplast DNA. Pl. Molec.Biol. 17(5): 1105–1109.

Takhtajan A. (1997) Diversity and classification offlowering plants. Columbia University Press,New York.

Thompson J. D., Gibson T. J., Plewniak F., Jean-mougin F., Higgins D. G. (1997) The CLU-STALX windows interface: flexible strategies formultiple sequence alignment aided by qualityanalysis tools. Nucl. Acids Res. 25: 4876–4882.

Watson S. (1890a) Contributions to Americanbotany. IX. Proc. Am. Acad. Arts 25: 124–163.

Watson, S. (1890b) On the genus Eriogynia Bot.Gaz. 15: 241–242.

White T. J., Bruns T., Lee S., Taylor J. (1990)Amplification and direct sequencing of fungalribosomal RNA genes for phylogenetics. In:Innis M. A., Gelfand D. H., Sninsky J. J., WhiteT. J. (eds.) PCR protocols: a guide to methodsand applications. Academic Press, San Diego,pp. 315–322.

Addresses of the authors: Daniel Potter (e-mail:[email protected]) and Shannon M. Still,Department of Plant Sciences, Mail Stop 2, Uni-versity of California, One Shields Avenue, Davis,California, 95616, USA. Tine Grebenc, GregorBozic, and Hojka Kraigher, Department for ForestPhysiology and Genetics & Research/ProgramGroup: Forest Biology, Ecology and Technology,Slovenian Forestry Institute, Vecna pot 2, 1000Ljubljana, Slovenia. Dalibor Ballian, Faculty ofForestry, University of Sarajevo, Zagrebacka 20,71000 Sarajevo, Bosnia and Herzegovina. JosipFranjiæ, Faculty Of Forestry, University ofZagreb, Svetosimunska 25, p. p. 422, 10002 Zagreb,Croatia.

118 D. Potter et al.: Phylogeny of Spiraeeae