SystematicBotany (2004),29(1): pp. 147– 164 q Copyright 2004by the AmericanSociety of PlantTaxonomists

Paraphyly in Tribe Onagreae: Insights into Phylogenetic Relationships of Onagraceae Basedon Nuclear and Chloroplast Sequence Data

RACHEL A. LEVIN,1,7 WARREN L. WAGNER,1 PETER C. HOCH,2 WILLIAM J. HAHN,3 AARON RODRIGUEZ,4 DAVID A. BAUM,5 LILIANA KATINAS,6 ELIZABETH A. ZIMMER,1 and KENNETH J. SYTSMA5 1DepartmentofSystematic Biology,Botany,MRC 166,Smithsonian Institution,P.O.Box 37012,W ashington, District ofColumbia 20013-7012; 2Missouri BotanicalGarden, P.O.Box 299,St. Louis, Missouri 63166-0299; 3108White-Grav enor,Box 571003,Georgetow nUniversity,Washington,District ofColumbia, 20057-1003; 4Departamento de Botan ´‡ca y Zoolog´‡a,Apartado Postal 139, 45101 Zapopan, Jalisco, M exico; 5Department ofBotan y,University ofW isconsin, 430Lincoln Drive,Madison, Wisconsin 53706; 6DepartamentoCientiŽ co de PlantasV asculares,M useo de Ciencias Naturales,Paseo del Bosques/ n, 1900La Plata, Provincia de Buenos Aires, Argentina 7Author forcorrespond ence ([email protected])

CommunicatingEditor: Thomas G .Lammers

ABSTRACT. Onagraceaeare afamilyof 17 genera insev en tribes,with the majorityof species intribes Onagreae and Epilobieae.Despite the species-richnessof these twotribes,to da te no phylogenetic studyhas been done withsufŽ cient taxon samplingto examine relationshipsbetw een and withinthese tribes. In this study ,we used DNAsequence data from one nuclear region(ITS) and twochloroplastregions ( trnL-trnF and rps16)toinfer phylogene tic relationshipsamong93 taxa across the family,withconcentra ted samplingin the large tribeOnagreae .Results stronglysuggest tha ttribeGongylocarpeae issister to tribes Epilobiea e 1 Onagreae,bothof which are monophyletic. WithinOnagreae , Camissonia seemsto be broadly paraphyletic, and Oenothera isalso paraphyletic. In Oenothera thereappear tobe twolineages, one ofwhich has Gaura 1 Stenosiphon nested withinit. Atthebase ofthe Onagracea ephylogeny,wehave clariŽed previousconfusion regarding conicting placemen tsof Hauya and Lopezia based on nuclear versuschloropla stda ta. Results ofthese analyses are supported bymorpholo gyand suggestthe need fornew taxonomicdelimitations,which are forthcoming.

The plantfamily Onagraceae(Evening-p rimroses) Clarkia (Sytsmaan dSmith 1988,1992; Sytsma et al. comprises ca.655 species across17 genera (Levin et al. 1990;Gottlieb and Ford 1996;F ord and Gottlieb2003; 2003),with atleast twothirds ofthe species occurring W.J.Hahn et al.,in mss.), Epilobium and Chamerion in tribes Onagreae (8genera, 262spp .)and Epilobieae (Baumet al.1994), and Gaura (Hoggard et al.,2004). (2genera, 172spp .).Onagraceae have aworld-wide However, no such study has focusedon relationships distribution, with the majority ofspecies concentrated among tribes Onagreae and Epilobieae.Furthermore, in the New World, especially western North America. within Onagreae there have been no molecularphy - Over the pastfew decades, the family has developed logenetic studies ofthe species-richgenera Camissonia asa model system forstudying plante volution.Com- (62spp .;w estern North America,1 sp.in SouthAm er- parativestudies ofcytology ,embryology,palynology, ica) and Oenothera (120spp .;Americas, the majority of anatomy,morphology,reproductive biology,and species in western North America). chemistry have allbeen completed forv ariousgroups Using chloroplast rbcL and ndhF sequence data,Le v- within the family (reviewedin Raven 1988).U nfortu- in et al.(2003) showed thatthe small genus Gongylo- nately,alimitation ofthese previous studies has been carpus (2spp .),previous ly included in tribe Onagreae the absenceof a robustph ylogeneticframeworkwithin (Raven 1964,1979; M unz1965), is strongly supported which toexamine the evolution ofthese traits. assister tothe rest ofOnagreae 1 Epilobieae,and Todate there have been several molecular (Martin should beplaced in its own tribe,Gongylocarpeae. and Dowd 1986;Crisci et al.1990; Sytsma et al.1991b; That analysis alsosuggested thatneither Camissonia Bult and Zimmer 1993;Conti et al.1993) an dmorpho- nor Oenothera is monophyletic,although sampling logical (Hochet al.1993) ph ylogeneticstudies ofthe within these genera waslimited. Camissonia appears family,although only recently has there been amolec- tolack an ymorphologicalsynapomor phies (Raven ularstudy thatincluded members ofall Onagraceae 1969;H ochet al.1993), and the only characteruniting genera (Levin et al.2003). There have alsobeen various Oenothera (stigma with 4linear elongate non-commis- phylogenetic studies ofindividual genera within the sural lobes)also characterizes Stenosiphon and Gaura family,including Fuchsia (Sytsmaand Smith 1988, (Hochet al.1993; H oggard et al.,2004); howe ver, Sten- 1992;Sytsma et al.1991a; P .Berry et al.,U .Wisconsin- osiphon and Gaura differ because ofthe presence ofan Madison, in mss.), Lopezia (O’Kane and Schaal1998), indusium atthe baseof the stigma lobes.

147 148 SYSTEMATIC BOTANY [Volume 29

Thus, amajor goalof the present study is toun- inour study,and instead we included bothsubspecies of G. hex- andra (sect. Pterogaura).In the other six tribes, tw otaxa were sam- derstand relationships between and within tribes On- pled from Ludwigia (tribeJussiaeea e)to serv eas amonophyletic agreae and Epilobieae,with aparticularemphasis on outgroup forphylogen etic analyses, given previousstudies that evaluatingthe monophyly ofthe large and diverse unambiguouslyplace thisgenus sisterto the restof Onagraceae genera Camissonia and Oenothera.Aphylogenetic (e.g., Levin et al. 2003). Onespecies each fromtribes Ha uyeae, Fuchsieae,Circaeeae,and Gongylocarpeae was also included. In frameworkwill facilitatecomparativ eanalyses ofchro- orderto more precisely determinethe relationship ofthe newly mosomal evolution and pollination biology ofthese di- described monotypicgenus Megacorax to Lopezia (tribeLopezieae ), verse groups, aswell asbiogeograph icalstudies ofthe wesampled four Lopezia species fromv arious sectionsplus Me- gacorax gracielanus . The cp trnL-trnF and nuclear ITSregions w ere radiation ofthese tribes in southwestern North Amer- sequenced froma total of93 taxa. The cp rps16 regionw as se- ica(Katinas et al.2004). quenced froma subsetof 75 species focused mainlyin Onagreae , While the main focusof this study is on Onagreae inorder to improve resolutionwithin this species-rich tribe.All and Epilobieae,wehaveincluded sampling from mem- taxa included inthis study are listedin T able 1withvouche rin- formation. bers ofall Onagraceae genera. This strategy is not only DNAExtraction, AmpliŽcation, and Sequenc ing. Total genomic important forexamining relationships among tribes DNAforthe majority of taxa was provided byKJS (see protocols Onagreae and Epilobieae,butinclusion ofDN Ase- inConti et al. 1996; Sytsmaet al. 2002). However, several taxa were quence datafrom bothn uclearand chloroplast regions extracted bythe seniora uthorfrom either silica gel-driedor her- barium material usingthe QiagenDneasy y kit(Qiagen Inc., Va- allows examinationofprevious conict among evolu- lencia, CA).DN Asof Lopezia lopezioides , L. racemosa, and L. lang- tionary reconstructions based on these twogenomes maniae were provided byS. O’ Kane (Univ.NorthernIo wa), and and on morphology,especially aspertains tothe place- DNAs of Oenothera deltoides and O. pallida were provided byM. Evans (Univ.Arizona). ment of Hauya and Lopezia (Bultand Zimmer 1993; ITS.AmpliŽ ca tionof the internal transcribedspacer (ITS)re- Conti et al.1993; H ochet al.1993; Levin et al.2003). gionof nuclear ribosomalDN A, composed ofITS1, the5.8S gene, The recently described genus Megacorax (Gonza´lez Eli- and ITS2(Baldwin 1992; Baldwin et al. 1995) was mainlycon- zondoet al.2002) may bevital todiscerning relation- ducted byWJH using primers ITS4 (5 9-TCCTCC GCT TATTGA TAT GC-39;Whiteet al. 1990) and ITS5HP (5 9-GGA AGG AGA ships of Hauya and Lopezia tothe rest ofthe family,as AGT CGT AACAAGG-3 9;Hershkovitz and Zimmer1996); these Levin et al.(2003) found that Megacorax is sister to Lo- primersw ere also used forthose ampliŽ ca tionsdone bythe senior pezia.Because sampling of Lopezia species waslimited author.Standard PCR conditions were used, although Ready-to- goPCR beads (AmershamPharmacia Biotech Inc.)w ere employed in thatstudy ,it wasunclear whether Megacorax should fora fewtaxa that were difŽcult toamplify. PCR products were beplaced within Lopezia.Thus, the present study in- cleaned usingP EG precipitation and ethanol cleaning (Morgan cludes additional sampling from varioussections of and Soltis1993). Cycle sequencingused ABIBig Dye chemistry Lopezia (Plitmann et al.1973; O’ Kane and Schaal1998). (Applied Biosystems,Foster City ,CA),and was done inboth di- rectionsusing the same primersas forampliŽ cation. Additional In this paperw eendeavor to:1) examine relation- sequencingprimersw ere used byWJH, including ITS2(5 9-CGT ships between and within tribes Onagreae and Epilo- AGC TACTTC TTG CATCG-3 9;Whiteet al. 1990), ITS3B(5 9-GCA bieae,2)test the monophyly of Camissonia, Oenothera, TCG ATG AAGAACGTAGC-3 9;Whiteet al. 1990), and C5.8S 9 9 and Gaura,3)compare signal from nuclearvs. chlo- (5 -TGC GTT CAAA GACTC GAT-3 ;Suhet al. 1993). ITSse- quences for Lopezia lopezioides , L. racemosa, and L. langmaniae were roplastdata, especially asit relates toearlier conict provided byS. O’Kane (Univ.NorthernIow a), and thesequences regarding relationships of Hauya and Lopezia, and 4) for Chamerionangustifolium , all Epilobium species, and Clarkia bottae further examine the sister taxonrelationsh ip previous- were previouslyp ublished byBa umet al. (1994) (seeGenBank accession numbersin T able 1). ly reported between and . To accom- Megacorax Lopezia TRNL-TRNF.AmpliŽca tionof the trnL intron, trnL 39 exon, and plish these goals, weused DNAsequence datafrom trnL-trnF intergenicspacer used primers‘ ‘c’’ (5 9-CGA AAT CGG one nuclear region (ITS)and twochloroplast regions, TAGACG CTACG-3 9)and ‘‘f’’(5 9-ATT TGA ACT GGT GACACG the trnL-trnF region (Taberlet et al.1991) and the rps16 AG-39)ofT aberlet et al. (1991). PCR products were cleaned as described above. Cycle sequencingused ABIBig Dye chemistry , intron (Oxelman et al.1997; Popp and Oxelman 2001). and was done inboth directions using the same primersas for These gene regions evolvemore rapidly than the pro- ampliŽcation. Afewtaxa havesequences withlong repeats, re- tein-coding ndhF and rbcL genes used in ourearlier sultingin incomplet ecycle sequence products. Forthese taxa, cycle sequencingwas conducted withadditiona linternal primers(d: 5 9- study (Levin et al.2003). GGG GATAGAGGG ACT TGAAC-3 9, e: 59-GGT TCAAGT CCC TCT ATC CC-39;Taberlet et al. 1991). MATERIALSAND METHODS RPS16.AmpliŽca tionof the rps16 group IIintronused thefol- lowingprimers adapted fromOxelman et al. (1997) and Popp and TaxonSampling. Thisstudy includes samplingfrom all eight Oxelman(2001): forward primerP1840 (5 9-GTG GTAAAAA GC tribesand 17 genera ofOnagraceae ,witha concentrationon On- AACGCG CGACTT-3 9;similarto rpsF) and reverseprimer P1839 agreae and Epilobieae(Table 1). Withinthese two tribesw ein- (59-TCG GGA TCG CACATC AATTGC AAC-3 9;similarto rpsR2). cluded atleast one individual per section, subsection, orseries, PCR products were cleaned as previouslydescribed. Sequencing depending on current circumscriptions(T able 1). However, we did used ABIBig Dye chemistry ,and was done inboth directions us- notsample from Chamerion sect. Rosmarinifolium , as Chamerion has ingthe same primersas forampliŽ cation. Due tothe same cycle previouslybeen shownto be stronglymonophyl etic (Baumet al. sequencingproblemwith long repea tsmentioned abov e, twoad- 1994). Wealso did notinclude all ofthe subsections of Clarkia, as ditional internal primersw ere used insequencin gsometaxa: for- theyare the subject ofanother analysis (W.J.Hahn et al., inmss.), ward primerP1895 (5 9-GTG TATCGT GCG GGAA-3 9) and re- and wediscoveredlate inthe analysis thatour onlysample of verse primerP1896 (5 9-GTATTC TCATAACTC A-3 9). Gaura sect. Campogaura was misidentiŽed. Thus, that section isnot Cycle sequence products forall regionsw ere precipitated and 2004] LEVINET AL.: RELATIONSHIPSIN ONAGRACEAE 149

TABLE 1. Taxa, vouchers, localities, and Genbank accession numbersfor all sequences included inthis study (* Megacorax iscurrently notplaced inan ytribe;best afŽ nity with Lopezieae ).All tribes except Epilobieaeand Onagreae contain asinglegenus; forthese tribes, listedare thetotal numberof species inthat tribeand thenumber of sections (if relev ant) currentlycircumscribe dfortha ttribe’sgenus. Fortribes Epilobiea eand Onagreae,total numberof species per genus and sectionsper genusare indicated, as are total numberof species per section [except Chamerion sect. Rosmarinifolium (4 spp.) and Gaura sect. Campogaura (1 sp.)].Sectional informationand total species numbersare based on Raven (1969); Raven and Gregory(1972); Tobe et al. (1987); Baumet al. (1994); O’Kane and Schaal (1998); Levin et al. (2003); and Wagner et al. (inmss.).

OUTGROUP TribeJussiaeeae (81 spp., 23 sects.) Ludwigia peploides (Kunth)P .H.Raven—Alameda Co., CA, Sytsma 5010 (WIS);nrITS A Y271517, trnL-trnF AY264494, rps16 AY267386. Ludwigia ravenii C.Peng—Berkeley Co., SC, Peng 4402 (MO);nrITS A Y271518, trnL-trnF AY264495. INGROUP TribeHauyeae (2 spp.) Hauya elegans DC.—Esteli, Nicaragua, Moreno11352 (MO);nrITS A Y271519, trnL-trnF AY264496. TribeFuchsieae (105 spp., 10 sects.) Fuchsia cyrtandroides J.W.Moore—Tahiti, SocietyIslands (Fr .), Berry etal. 4618 (MO);nrITS A Y271520, trnL-trnF AY264497. TribeCircaeeae (7spp.) Circaea alpina L.—Lincoln Co., WI, Smith 1052 (WIS);nrITS A Y271521, trnL-trnF AY264498. TribeLopezieae (22 spp.,6sects.) Lopezia Sect. Pelozia Lopezia laciniata (Rose)Plitm., Raven &Breedl.—Durango, Mexico, O’Kane 3341 (MO);nrITS A Y271522, trnL-trnF AY264499. Sect. Jehlia Lopezia langmaniae Miranda—Chiapas, Mexico, Breedlove 32300 (CAS);nrITS A Y271523, trnL-trnF AY264500. Sect. Lopezia Lopezia racemosa Cav.—Queretaro, Mexico, O’Kane 3374 (MO);nrITS A Y271525, trnL-trnF AY264502. Sect. Diplandra Lopezia lopezioides (Hook. &Arn.)Plitm., Raven &Breedl.—Nayarit, Mexico, O’Kane 3389 (MO);nrITS A Y271524, trnL-trnF AY264501. *Megacorax gracielanus Gonza´lez &Wagner—Durango, Mexico, Acevedo etal. 1352 (US);nrITS A Y271526, trnL-trnF AY264503, rps16 AY267387. TribeGongylocarpeae (2 spp.) Gongylocarpus fruticulosus (Benth.)Brandegee— Michoaca ´n, Mexico, Rzedowski 44253 (IEB);nrITS A Y271527, trnL-trnF AY264504, rps16 AY267388. TribeEpilobieae Chamerion (Raf.) Raf. (8 spp., 2sects.) Sect. Chamerion (4 spp.) Chamerion angustifolium (L.)Holub— Barron Co., WI, Sytsma 5500 (WIS);nrITS L28011. King Co., WA, Wagner 6917 (US); trnL- trnF AY264505, rps16 AY267389. Epilobium L.(164 spp.,7sects.) Sect. Epilobium (ca. 150 spp.) E. ciliatum Raf.—Del Norte Co., CA, Hoch 3487 (MO);nrITS L28015, trnL-trnF AY264508. E.obcordatum A.Gray—Harney Co., OR, Seavey 1151 (MO);nrITS L28027. Lake Co., OR, Ertter15067 (JEPS); trnL-trnF AY264507. E. rigidum Hausskn.—Del Norte Co., CA, Wiens 6797 (MO);nrITS L28030, trnL-trnF AY264506, rps16 AY267390. Sect. Xerolobium (1 sp.) E.brachycarpum Presl—Yolo Co., CA, Sytsma s.n. (WIS);nrITS L28012, trnL-trnF AY264509. Sect. Crossostigma (2 spp.) E. minutum Lindl. ex Lehm.—Curry Co., OR, Chambers 4847 (MO);nrITS L28025, trnL-trnF AY264510. Sect. Cordylophorum (3 spp.) E.nevadense Munz—Clark Co., NV, Hoch 3440 (MO);nrITS L28026, trnL-trnF AY264511. Sect. Currania (2 spp.) E. pygmaeum (Speg.)Hoch &P.H.Raven—Butte Co., CA, Broyles 1090 (MO);nrITS L28029, trnL-trnF AY264512. Sect. Boisduvalia (4 spp.) E.densiorum (Lindl.)Hoch &P.H.Raven—Butte Co., CA, Oswald 794 (CHSC);nrITS L28019, trnL-trnF AY264513. Sect. Zauschneria (2 spp.) E. canum (Greene)P .H.Raven—Los AngelesCo., CA, Cult. UCBot. Gard. 59.1378; seed from Beard &Beard, coll. 1959 (UC); nrITSL28013, trnL-trnF AY264514, rps16 AY267391. TribeOnagreae Xylonagra Donn. Smith& Rose (1sp.) Xylonagra arborea (Kellogg)Donn. Sm.& Rose—Baja California, Mexico, Warshall s.n. (MO);nrITS A Y271528, trnL-trnF AY264515, rps16 AY267392. Clarkia Pursh (42 spp.,11 sects.) Sect. Myxocarpa (7 spp.) C.mildrediae (A.Heller)F .H.Lewis &M.R.Lewis—Butte Co., CA, Weeden 50 (DAV);nrITSA Y271529, trnL-trnF AY264516, rps16 AY267393. 150 SYSTEMATIC BOTANY [Volume 29

TABLE 1. Continued.

Sect. Rhodanthos (6 spp.) C.franciscana F.H.Lewis &P.H.Raven—San Francisco Co., CA, GottliebF28-2-2 (DAV);nrITSA Y271530, trnL-trnF AY264517, rps16 AY267394. Sect. Clarkia (1 sp.) C.pulchella Pursh—Grant Co., OR, Ford 8357 (DAV);nrITSA Y271531, trnL-trnF AY264518, rps16 AY267395. Sect. Eucharidium (2 spp.) C. concinna (Fischer& Meyer)Greene— Marin Co., CA, Weeden 146-16-3 (DAV);nrITSA Y271532, trnL-trnF AY264519. Sect. Godetia (7 spp.) C.imbricata F.H.Lewis &M.R.Lewis—Sonoma Co., CA, GottliebPG-1 (DAV);nrITSA Y271533, trnL-trnF AY264520, rps16 AY267396. Sect. Fibula (2 spp.) C. bottae (Spach) F.H.Lewis &M.R.Lewis—Los AngelesCo., CA, Weeden 35-4 (DAV);nrITSL28016, trnL-trnF AY264521. Sect. Phaeostoma (5 spp.) C. xantiana A.Gray—Tulare Co., CA, Gottlieb7436 (DAV);nrITS A Y271534, trnL-trnF AY264522, rps16 AY267397. Sect. Sympherica (9 spp.) C. rostrata W.Davis—Mariposa Co., CA, Weeden 97a (DAV);nrITSA Y271535, trnL-trnF AY264523, rps16 AY267398. Sect. Biortis (1 sp.) C. afŽnis F.H.Lewis &M.R.Lewis—Solano Co., CA, Weeden 79b (DAV);nrITSA Y271536, trnL-trnF AY264524, rps16 AY267399. Sect. Connubium (1 sp.) C. delicata (Abrams)Nelson &Macbride—San DiegoCo., CA, Lewis 1461 (LA);nrITS A Y271537, trnL-trnF AY264525, rps16 AY267400. Sect. Heterogaura (1 sp.) C.heterandra (Torrey)F .H.Lewis &P.H.Raven—Tuolumne Co., CA, Weeden 6 (DAV);nrITSA Y271538, trnL-trnF AY264526, rps16 AY267401. Gayophytum A.Juss.(9 spp.) G.heterozygum F.H.Lewis &Szweyk.—Shasta Co., CA, Baldwin 923 (MO);nrITS A Y271539, trnL-trnF AY264527, rps16 AY267402. Camissonia Link(62 spp., 9sects). Sect. Eulobus (4 spp.) C.californica (Nutt. ex Torr.&A.Gray) P.H.Raven—Pima Co., AZ, Schmidt &Merello 2581 (MO);nrITS A Y271597, trnL-trnF AY264585, rps16 AY267459. C.crassifolia (Greene)P .H.Raven—Baja California, Mexico, RSAseed coll. 16695; nrITS AY271540, trnL-trnF AY264528, rps16 AY267403. Sect. Chylismia (14 spp.) C.claviformis (Torr.&Fre´m.)P .H.Raven—Kern Co., CA, RSAseed coll. 16710; nrITSA Y271541, trnL-trnF AY264529, rps16 AY267404. Sect. Lignothera (2 spp.) C. arenaria (A.Nelson)P .H.Raven—Yuma Co., AZ, Raguso 98-22 (ARIZ);nrITS A Y271543, trnL-trnF AY264531, rps16 AY267406. Sect. Tetrapteron (6 spp.) C. ovata (Nutt. ex Torr.&A.Gray) P.H.Raven—Alameda Co., CA, Ertter13924 (JEPS);nrITS A Y271544, trnL-trnF AY264532, rps16 AY267407. C.subacaulis (Pursh)P .H.Raven—Adams Co., ID, Smith 2808 (MO);nrITS A Y271545, trnL-trnF AY264533, rps16 AY267408. C.tanacetifolia (Torr.&A.Gray) P.H.Raven—Washoe Co., NV, Tiehm 4528 (MO);nrITS A Y271546, trnL- trnF AY264534, rps16 AY267409. C.graciliora (Hook. &Arn.)P .H.Raven—Riverside Co., CA, Boyd 6162 (RSA);nrITS AY271547. Los AngelesCo., CA, Boyd etal. 10095 (US); trnL-trnF AY264535, rps16 AY267410. Sect. Holostigma (14 spp.) C.cheiranthifolia (Hornem.ex Spreng.)Raimann— Baja California, Mexico, Raguso RAR98-16 (ARIZ);nrITS A Y271548, trnL- trnF AY264536, rps16 AY267411. Sect. Camissonia (12 spp.) C.kernensis (Munz)P .H.Raven—Kern Co., CA, Howell &True47888 (MO);nrITS A Y271549, trnL-trnF AY264537, rps16 AY267412. C.campestris (E.Greene) P.H.Raven—Kern Co., CA, RSAseed coll. 16706; nrITSA Y271550, trnL-trnF AY264538, rps16 AY267413. Sect. Eremothera (7 spp.) C. refracta (S.Watson)P .H.Raven—Riverside Co., CA,RSAseed coll. 17552; nrITSA Y271551, trnL-trnF AY264539, rps16 AY267414. C. boothii (Dougl.)P .H.Raven—Ventura Co., CA, RSAseed coll. 17783; nrITSA Y271542, trnL-trnF AY264530, rps16 AY267405. C.nevadensis (Kell)P .H.Raven—Washoe Co., NV, Tiehm 11971 (MO);nrITS A Y271552, trnL-trnF AY264540, rps16 AY267415. C. minor (A.Nels.) P.H.Raven—Modoc Co., CA, Bartholomew 6623 (MO);nrITS A Y271553, trnL-trnF AY264541, rps16 AY267416. Sect. Chylismiella (1 sp.) C.pterosperma (S.Watson)P .H.Raven—Inyo Co., CA, MoreŽeld &McCarty 3364 (MO);nrITS A Y271554. Tooele Co., UT, Windham 93-32 (MO); trnL-trnF AY264542, rps16 AY267417. Sect. Nematocaulis (2 spp.) C. andina (Nutt.)P .H.Raven—Washoe Co., NV, Tiehm 8089 (MO);nrITS A Y271555, trnL-trnF AY264543, rps16 AY267418. 2004] LEVINET AL.: RELATIONSHIPSIN ONAGRACEAE 151

TABLE 1. Continued.

Oenothera L.(120 spp., 14 sects.) Sect. Oenothera (71 spp.) O.organensis Munz—Don ˜aAna Co., NM,Cult. DUSS76-0334 ( Emersons.n. , MO);nrITS A Y271556, trnL-trnF AY264544, rps16 AY267419. O.maysillesii Munz—Durango, Mexico, Cult. DUSS81-195 ( Breedlove 18812, MO);nrITS A Y271557, trnL-trnF AY264545, rps16 AY267420. O.macrosceles A.Gray—Coahuila, Mexico, Cult. DUSS197 ( Wagner etal. 4096, MO); nrITS AY271558, trnL-trnF AY264546, rps16 AY267421. O. stubbei W.Dietr.,W.L.Wagner &P.H.Raven—Nuevo Leo´n, Mexico, Cult. DUSS791 ( Sanders etal. 1203, MO);nrITS A Y271559, trnL-trnF AY264547, rps16 AY267422. O.heterophylla Spach— Houston Co., TX, Wagner 6916 (US);nrITS A Y271560, trnL-trnF AY264548, rps16 AY267423. O.laciniata Hill—St. Francis Co., AR, Hecht 21 (MO);nrITS A Y271561, trnL-trnF AY264549, rps16 AY267424. O.pubescens Willd. ex Spreng.—Michoacan, Mexico, Grown fromseeds ( Rzedowski s.n., 25 Aug1986, no voucher); nrITSA Y271562, trnL-trnF AY264550, rps16 AY267425. O. afŽnis Cambess—Buenos Aires,Argentina, Cult. DUSS82-603 ( Hecht 125, MO);nrITS A Y271563, trnL-trnF AY264551, rps16 AY267426. O. elata Kunth—San Mateo Co., CA, Cult. DUSS89-72 ( Cleland s.n., MO);nrITS A Y271564, trnL-trnF AY264552, rps16 AY267427. O. biennis L.—New Brunswick, Canada, Cult. DUSS91-313 ( Cleland s.n., MO);nrITS A Y271565, trnL-trnF AY264553, rps16 AY267428. Sect. Kleinia (2 spp.) O.albicaulis Pursh—Cochise Co., AZ, Raguso RAR98-52 (ARIZ);nrITS A Y271566, trnL-trnF AY264554, rps16 AY267429. Sect. Ravenia (3 spp.) O. tubifera Ser.—Durango, Mexico, Cult. DUSS0305, Stubbes.n. seeds ( Breedlove 14321, MO);nrITS A Y271567, trnL-trnF AY264555, rps16 AY267430. Sect. Eremia (1 sp.) O.primiveris A.Gray—Maricopa Co., AZ, Wagner &Mill 4565 (MO);nrITS A Y271568, trnL-trnF AY264556, rps16 AY267431. Sect. Contortae (1 sp.) O.xylocarpa Coville—Mono Co., CA,Not vouchered, fromsame population as DeDecker s.n. (MO);nrITS A Y271569, trnL-trnF AY264557, rps16 AY267432. Sect. Pachylophus (5 spp.) O.caespitosa Nutt.—Ada Co., ID, Wagner 6267, no voucher; nrITSA Y271570, trnL-trnF AY264558, rps16 AY267433. O.psammophila (A.Nels. &J.F.Macbr.)W.L.Wagner, Stockhouse &Klein—Fremont Co., ID, Raguso RAR01-56 (US);nrITS A Y271571, trnL- trnF AY264559, rps16 AY267434. Sect. Megapterium (4 spp.) O.brachycarpa A.Gray—Grant Co., NM, Wagner 3811 (MO);nrITS A Y271572, trnL-trnF AY264560, rps16 AY267435. Sect. Paradoxus (1 sp.) O. havardii S.Watson—Brewster Co., TX, Powell 6175 (MO);nrITS A Y271573, trnL-trnF AY264561, rps16 AY267436. Sect. Lavauxia (5 spp.) O. ava (A.Nelson)Garrett— Apache Co., AZ, Wagner 3796 (MO);nrITS A Y271574, trnL-trnF AY264562, rps16 AY267437. O. acutissima W.L.Wagner—Daggett Co., UT, Raguso RAR01-65 (US);nrITS A Y271575, trnL-trnF AY264563. Sect. Gauropsis (2 spp.) O.canescens Torr.&Fre´m.—Lubbock Co., TX, Robbins1820 (MO) (Sytsma 5030, WIS);nrITS A Y271576, trnL-trnF AY264564, rps16 AY267438. Sect. Xylopleurum (1 sp.) O. speciosa Nutt.—East Baton Rouge Parish, LA, Zimmer 48-86 (LSU) (Sytsma 5024, WIS);nrITS A Y271577, trnL-trnF AY264565, rps16 AY267439. Sect. Hartmannia (10 spp.) O. rosea L’He ´r.ex Ait.—Durango, Mexico, Wagner &Brown3960 (MO);nrITS A Y271578, trnL-trnF AY264566, rps16 AY267440. O.tetraptera Cav.—Me´xico, Mexico, Rzedowski s.n. in1986, no voucher; nrITSA Y271579, trnL-trnF AY264567, rps16 AY267441. O.multicaulis Ru´‡z&Pavon—Prov .Loja, Ecuador, Ellemann 66724 (MO);nrITS A Y271580, trnL-trnF AY264568, rps16 AY267442. Sect. KneifŽa (5 spp.) O.fruticosa L.—Dane Co., WI,Cult. U. WIBot. Gard., Sytsma 5025 (WIS);nrITS A Y271581, trnL-trnF AY264569, rps16 AY267443. O.linifolia Nutt.—St. Francois Co., MO, Solomon 21279 (MO);nrITS A Y271598, trnL-trnF AY264586, rps16 AY267460. Sect. Anogra (9 spp.) O.neomexicana (Small)Munz— Graham Co., AZ, Raguso RAR98-167 (ARIZ);nrITS A Y271582, trnL-trnF AY264570, rps16 AY267444. O. pallida Lindl.—Coconino Co., AZ, Raguso RAR96-05 (ARIZ);nrITS A Y271583, trnL-trnF AY264571, rps16 AY267445. O.deltoides Torr.&Frem.—Yuma Co., AZ, Raguso RAR99-01 (ARIZ);nrITS A Y271584, trnL-trnF AY264572, rps16 AY267446. Calylophus Spach (6 spp., 2sects) Sect. Salpingia (4 spp.) C.hartwegii (Benth.)P .H.Raven—Lubbock Co., TX, Robbinss.n. (MO);nrITS A Y271585, trnL-trnF AY264573, rps16 AY267447. Sect. Calylophus (2 spp.) C.berlandieri Spach—Lubbock Co., TX, Robbinss.n. (MO) (Sytsma 5021, WIS);nrITS A Y271586, trnL-trnF AY264574, rps16 AY267448. Stenosiphon Spach (1sp.) S. linifolius (Nutt.)Heynh.— Pottawatomie Co., KS, Barkley s.n. (KSC);nrITS A Y271587, trnL-trnF AY264575, rps16 AY267449. Gaura L.(21 spp.,8sects.) Sect. Gauridium (1 sp.) G.mutabilis Cav.—Hidalgo, Mexico, Rzedowski s.n. (MO);nrITS A Y271588, trnL-trnF AY264576, rps16 AY267450. 152 SYSTEMATIC BOTANY [Volume 29

TABLE 1. Continued.

Sect. Schizocarya (1 sp.) G.parvi ora Douglas ex Lehm.—Jeff Davis Co., TX, Clinebell 2017 (MO);nrITS A Y271589, trnL-trnF AY264577, rps16 AY267451. Sect. Xerogaura (2 spp.) G.macrocarpa Rothr.—BrewsterCo., TX, Clinebell 3077 (MO);nrITS A Y271590, trnL-trnF AY264578, rps16 AY267452. G. boquil- lensis P.H.Raven &Gregory—Brewster Co., TX, Clinebell 3074 (MO);nrITS A Y271591, trnL-trnF AY264579, rps16 AY267453. Sect. Pterogaura (4 spp.) G. hexandra Go´mez Ortegasubsp. gracilis—BrewsterCo., TX, Clinebell 2023 (MO);nrITS A Y271592, trnL-trnF AY264580, rps16 AY267454. G. hexandra Go´mez Ortegasubsp. hexandra—Durango, Mexico, Clinebell 3031 (MO);nrITS A Y271594, trnL-trnF AY264582, rps16 AY267456. Sect. Gaura (6 spp.) G. demareei P.H.Raven &Gregory—Garland Co., AR, Cult. St.Louis, Hoch 3574 (MO);seed fromClinebell s.n.; nrITS AY271593, trnL-trnF AY264581, rps16 AY267455. Sect. Xenogaura (1 sp.) G.drummondii (Spach) Torr.&A.Gray—Archer Co., TX, Hoggard 409 (OKL);nrITS A Y271595, trnL-trnF AY264583, rps16 AY267457. Sect. Stipogaura (5 spp.) G. villosa Torr.—UnionCo., NM, Clinebell 2052 (MO);nrITS A Y271596, trnL-trnF AY264584, rps16 AY267458.

cleaned withisopropan ol beforesequencin gon an ABI377 auto- thisŽ le was executed inP AUP* usingthe heuristic search option mated sequencer.ITScycle sequence products thatwere notgen- with100 randomaddition sequence replicatesand the Multrees erated bythe senior a uthorw ere cleaned usingCentri-Sep col- option turned off.The decay index foreach branch isthe difference umns(Princeton Separatio ns, Adelphi, NJ). inlength between the shortesttrees consistent with each con- Sequence Alignment. Sequenceswere edited inEditvie wver- straintand theglobally shortest trees. sion1.0.1 (Applied Biosystems,1996), and thesequences fromall NUCLEAR ITS.P arsimonyanalysis was conducted as above, primerswere aligned and edited usingA utoassembler y DNA Se- withtw ospecies of Ludwigia deŽned as amonophyletic outgroup quence AssemblySoftware v ersion1.4.0 (Applied Biosystems, (Levin et al. 2003). 1989–95) toconstruct aconsensus sequence foreach species [Se- CHLOROPLAST TRNL-TRNF AND RPS16. Parsimonyanalyses of quencherversion 3.0 (Gene Codes Corp., AnnArbor, Michigan) each ofthese twodata setsw ere conducted as described above. was used formost ITS sequences ].Species sequenceswere then The two species of Ludwigia were deŽned as amonophyletic out- aligned manually inSeqApp (Gilbert1993). These alignments group forthe trnL-trnF analysis; because onlya singlespecies of were importedinto MacClade 4.0 (Maddisonand Maddison 2000) Ludwigia, L. peploides ,was sampled for rps16, thisspecies was de- and executed inP AUP*4.0b10 (Swofford2002). The alignmentsof Žned as theoutgroup forthe rps16 analysis. Congruence ofthe 75 all threegene regionsare available on TreeBASE(study accession taxon trnL-trnF and rps16 data setsw as testedusing a partition numberS941, matrixaccession numbersM1559 and M1560). homogeneitytest (ILD; Farris et al. 1994, 1995) as implemented in Phylogenetic Analyses. The threeda ta setsw ere analyzed sep- PAUP*. Onehundred heuristic partition homogeneityreplica tes arately (Table 2), and inv arious combinationswith other data sets were completed,each with10 randomaddition sequence repli- (seebelow). Parsimonyanalyses were conducted inP AUP* using cates, TBR branch-swapping, and gaps treated as missingdata. heuristic searches with100 randomaddition sequence replicates, Constant characterswere excluded, and theM ulTrees option was TBR branch-swapping, and steepest descent. Constant characters notin effect. Parsimonyanalysis was also conducted withthe trnL- were excluded, and gaps were treated as missingdata. Following trnF and rps16 data setscombined, with L. peploides deŽned as the the analysis protocol ofZimmer et al. (2002), each addition rep- outgroup. licate was limitedto 200 treestha twere greaterthan orequal to NUCLEAR AND CHLOROPLAST DATA.Totestfor congruenc e theshortest trees for each replicate.Thisw as necessary due to amongthe nrITSdata setand the chloroplastda ta sets, PAUP* was large numbersof equal lengthtrees. The strengthof support for used toconduct pairwise ILDtests of the 93 taxon ITSand trnL- individual treebranches was estimated usingbootstrap values trnF data setsand the75 taxon ITSand rps16 data sets,and a (BS)(F elsenstein1985) and decay indices (DI)(Bremer 1988; Don- simultaneous ILDtest with the 75 taxon data setsfor all three oghue et al. 1992). Bootstrap values were from500 fullheuristic regions(settings for ILD tests are as above). Parsimonyanalyses bootstrap replicates, each with10 randomaddition sequence rep- were thenconducted usinga combinedda ta setof ITS and trnL- licates. The MulTrees option was notin effect, and constant char- trnF sequence data, as well as adata setincluding all threege- acters were excluded.Decayvalues foreach branch were deter- nomicregions. Beca use asubsetof taxa were sampled for rps16, minedby Ž rstusing the P AUP decay index command Žle in fewertaxa are included inthe analysis ofall data setsthan inthe MacClade toprepare asetof trees each witha singlebranch re- analysis ofthe ITS 1 trnL-trnF data sets.Two species of Ludwigia solved. ToŽnd theshortest trees consistent with each constraint, were deŽned as amonophyletic outgroup forthe ITS 1 trnL-trnF

TABLE 2. Comparisonofthe 75 taxa data setsfor the nrITSand two cp regions.P arsimony-informative 5 PI; consistencyindex 5 CI (RC 5 rescaled CI);retention index 5 RI.

nrITS trnL-trnF rps16 Range ofraw length 572–606 bp 743–957 bp 790–838 bp Alignedlength 663 bp 1174 bp 1009 bp Variable sites(proportion) 312 (0.47) 347 (0.30) 318 (0.32) PIsites (proportion) 176 (0.27) 153 (0.13) 163 (0.16) Pairwise distance ranges 0–0.26 0–0.077 0–0.092 CI (RC); RI 0.56 (0.37); 0.66 0.73 (0.59); 0.80 0.72 (0.58); 0.81 2004] LEVINET AL.: RELATIONSHIPSIN ONAGRACEAE 153 analysis. Forthe analysis ofall threedata setscombined, Ludwigia Gongylocarpeae 1 Epilobieae 1 Onagreae (BS 594; peploides was deŽned as theoutgroup . DI54).A dditionally,Onagreae 1 Epilobieae are well INDELS.Usingthe combineddata setof all threeregions, indels greaterthan 1bp were coded as separate binarycharacter s. Only supported (BS 597; DI55),and Epilobieae is strongly indelstha twere identical inlength and bases were included. Par- supported asm onophyletic (BS 598; DI57)including simonyanalysis was conducted on all threedata setscombined, Chamerionangustifolium ,which appearssister toa withindels included as additional characters. MAXIMUM LIKELIHOOD.Ananalysis usinga maximumlikeli- monophyletic Epilobium (BS597; DI55).A dditionally, hood (ML)model w as conducted withall threedata setscom- trnL-trnF yielded better resolutionwithin Onagreae bined. MLmodel parameters were determinedby using M odeltest than did ITS(see details from combined analysis be- v.3.06 (Posada and Crandall 1998). Thisprogram tests the Žtof 56 substitutionmodels to the data; based on ahierarchical likeli- low).Results ofan ILD test forcongruence between hood ratio test, amodeltha tbestŽ tsthe data isidentiŽ ed. The the ITS and trnL-trnF datasets strongly suggest that bestmodel w as used ina MLanalysis inP AUP*, usingthe heu- the datasets are congruent ( P50.33).Thus, the two risticsearch option, startingtree determined by neighbor-joi ning, TBR branch-swapping, and MulTrees option ineffect. Asabov e, datasets were combined (Fig. 2). Ludwigia peploides was deŽned as theoutgroup . NuclearITS andcp trnL-trnF. In general, analysis ALTERNATIVE TOPOLOGIES.Constrainttrees were constructed ofthe twodatasets combined (Fig. 1)yielded more inMacClade totest alternativ ephylogenetic hypotheses,including robustsupport and greater claderesolution than the themonophy lyof each ofthe follo wingthree genera: Camissonia, Oenothera, and Gaura.These treesw ere loaded intoP AUP*, and analyses ofthe separatedata sets. There is moderate heuristic searches were conducted toŽ nd the shortesttrees con- support for Hauya assister toall Onagraceae except sistentwith eac hconstraint. The numberof additional steps re- Ludwigia andforthe monophyly ofthe rest ofOnagra- quiredfor a given constraint isthe differencebetween theshortest treesconsistent with a particular constraint and the globallyshort- ceae above Ludwigia and Hauya (BS573; DI55). As esttrees. Further,one-tailed non-parametric Shimodaira-Hasega- with the analyses ofthe separatedata sets, the sister wa tests(S-H test; Shimodaira and Hasegawa 1999; see also Gold- relationship between Fuchsia and Circaea is well sup- man et al. 2000) were conducted inP AUP* toassess thestatistical 5 5 support forthese constraints, usingthe same MLparameters out- ported (BS 100; DI 19).Although Megacoraxgracie- lined above. Inthis procedure ,the likelihoodsof all the shortest lanus clearly forms acladewith the genus Lopezia treesconstrained tocontain aparticular lineage ofinterest w ere (BS599; DI514),it is apparentthat M.gracielan us is compared withthe likelihoodof a randommost-parsimo nious sister toa monophyletic Lopezia (BS5100; DI516). (MP)tree from the unconstrain ed analysis. The timeefŽ cient RELL methodw as used, with1000 bootstrap replicates. However, there is only limited supportfor the mono- phyly ofOnagraceae above the early diverging tribes ofJussiaeeae, Hauyeae, Circaeeae ,and Fuchsieae RESULTS (BS557; DI52). NuclearITS . ITSsequences for93 taxa ranged in Tribes Gongylocarpeae 1 Epilobieae 1 Onagreae length from 572–617 bp, with analigned length of680 are well supported asa monophyletic group (Node A; characters,including ITS1,the 5.8SrRN Agene,and BS592; DI55), with Gongylocarpus sister toEpilobieae ITS2.Of these 680characters, 243 were parsimony-in- 1 Onagreae (BS 598; DI59).W ithin this lineage,the formative acrossall 93 taxa. Among these species, % monophyly oftribe Epilobieae is strongly supported missing dataranged from 0–13.1%, with amean of (BS597; DI58), and Chamerionangustifolium is sister to 0.5%and amedian of0%. Although anumber of amonophyletic Epilobium (BS5100; DI521). In the clades near the tips are well supported, overall there phylogenyinferred from ITSdata alone ,the relation- wasnot strong signal (i.e.,low CI/RI)in the ITSdata ship of Chamerion to Epilobium wasnot resolved. set (CI50.51, RI50.73, RC50.37),resulting in ageneral Onagreae is alsosupported asmonoph yletic (Node lackof resolution and low supportfor man ynodes B; BS582; DI55),with very weak supportfor Xylon- (tree not shown). However, Fuchsia 1 Circaea (BS5100; agra assister tothe rest ofOnagreae (BS ,50; DI51). DI514) and Lopezia 1 Megacorax (BS581; DI54) are Within Onagreae relationships are somewhatequivo- well supported clades. Further, the genus Epilobium is cal,but a number oflineages are well supported. Clark- strongly monophyletic (BS 5100; DI513), as is Clarkia ia is clearly monophyletic (BS 5100; DI521), as is Cam- (BS5100; DI512).In general, the topology is asin Fig. issonia sect. Tetrapteron excluding C. graciliora (C. ovata 1 (ITS 1 trnL-trnF, see below),but with less support 1 C. subacaulis 1 C. tanacetifolia ; BS5100; DI514). Re- forrelationsh ips. lationships among these three Camissonia species and Chloroplast trnL-trnF. trnL-trnF sequencesfor93 Clarkia, Gayophytumheterozygum , and Camissoniapter- taxaranged in length from 729–957 bp ,with an osperma remain unclear (BS ,50, DI51).Am ong the aligned length of1204 characters. Of these characters, rest of the Camissonia species sampled, themonophyly 218w ere parsimony-informative.Taxonsequences con- of sections Eremothera (C. boothii 1 C. minor 1 C. nev- tained no missing data,except for one with 0.2%and adensis 1 C. refracta; BS5100; DI513) and Camissonia another with 0.6%missing data.Com pared tothe phy- (C. kernensis 1 C. campestris ; BS5100; DI511)is strong- logenyinferred from the nrITS dataalone ,the phylog- ly supported. Further, Camissoniagracili ora (sect. Te- enyinferred from trnL-trnF datais somewhat more re- trapteron) and C. cheiranthifolia (sect. Holostigma) form a solved (tree not shown). There is strong supportfor weakly supported clade(BS 564; DI51).Although 154 SYSTEMATIC BOTANY [Volume 29

FIG.1. Strictconsensus of13275 MPtrees(TL 51912, rescaled CI 50.44, RI50.76) fromthe combinedanalysis ofITS 1 trnL- trnF data (93 taxa). Nodes with bootstrap values (BS) .50% and decay indices (DI) .0are indicated, withBS listedabove the node and DIbelow .Tribesto which the taxa belongare listedto the right,and speciŽc nodes ofinterest are labeled with letters;see textfor discussion. 2004] LEVINET AL.: RELATIONSHIPSIN ONAGRACEAE 155

FIG.2. Strictconsensus of18200 MPtrees(TL 52018, rescaled CI 50.47, RI50.73) fromthe combinedanalysis ofnrITS and cp trnL-trnF and rps16 data (75 taxa). Nodes with bootstrap values (BS) .50% and decay indices (DI) .0are indicated, with BSlistedabove the node and DIbelow .Currentsectional afŽliations are listedto the rightfor all Camissonia and Oenothera species. The two main lineages of Oenothera species are indicated as ‘‘A’’and ‘‘B’’;see textfor discussion. 156 SYSTEMATIC BOTANY [Volume 29 manyofthe sections within Camissonia are strongly Chloroplast trnL-trnF and rps16. There is strong supported, there is little resolutionforrelationsh ips agreement between the topologies inferred from the among these lineages. nrITS and trnL-trnF datasets combined (Fig. 1)and However, there is moderate supportfor a cladecom- the trnL-trnF and rps16 datasets combined (tree not prising the remainderofOnagreae ,including fourspe- shown), butnote thatthe former includes 93taxa and cies of Camissonia 1 Oenothera 1 Calylophus 1 Gaura 1 the latter 75taxa. Difference sin the twotopologies are Stenosiphon (Node C; BS572; DI52).W ithin this line- mainly due todiffering levels ofsupport for nodes, age the monophyletic Camissonia sect. Eulobus (C. cras- with clades inferred from the trnL-trnF 1 rps16 data sifolia 1 C. californica ; BS5100; DI512)appears sister set often having higher bootstrapsupport than those tothe rest ofthe lineage (BS 582; DI52). The next inferred from the ITS 1 trnL-trnF analysis. The major- branchwithin this cladeconsists of Camissoniacla vifor- ity ofcon icts, especially those relating to Camissonia, mis (sect. Chylismia) 1 C. arenaria (sect. Lignothera) are associatedwith low resolutionin atleast one ofthe (BS589; DI52),which is strongly supported assister twocombined analyses. However, within Clarkia the to Oenothera 1 Calylophus 1 Gaura 1 Stenosiphon nuclearand cpdata support different topologies.For (BS593; DI54).In this well supported clade, Oenothera example,the ITSdata strongly support Clarkiarostrata is paraphyletic relativetothe other three genera. A as sister to C. afŽnis (BS597; DI53),whereas the cp lineage composed of Oenothera sects. Oenothera, Kleinia, analysis suggests strong supportfor the sister relation- Anogra, Ravenia, Eremia, and Contortae has moderate ships of C. rostrata 1 C. delicata (BS591; DI53) and C. support(N ode D; BS 584; DI52),as does anested imbricata 1 C. afŽnis (BS599; DI58). cladeof sects. Oenothera 1 Kleinia 1 Anogra (Node E; AllData SetsCo mbined. Results ofan ILD test BS589; DI53).Further, sects. Kleinia 1 Anogra (O. al- comparing allthree datasets simultaneously suggest bicaulis 1 O.neomexicana 1 O. pallida 1 O.deltoides ; signiŽcant incongruen ce( P50.01).H owever,there is BS595; DI54)are clearly sister toa monophyletic sect. evidence thata valuebetw een 0.01and 0.001(rath er Oenothera (O.organensis through O.pubescens ; BS5100; than the traditional P50.05)is the more appropriate DI56). criticalv aluefor incongrue nce (Cunningham 1997). The monophyly of Oenothera sects. Pachylophus (O. Otherwise asigniŽcant ILD test may simply reect dif- caespitosa 1 O.psammophila ; BS5100; DI59) and La- fering amounts ofsignal among datasets (Davis et al. vauxia (O.acutissima 1 O. ava; BS5100; DI516) is 1998;Y oder et al.2001; but see Hippet al.in press), a strongly supported, buttheir relationships tothe rest difference thatis alsofound in ourdata sets (Table2). ofthe genus are obscure(Fig. 1).Similarly ,the small When the pairwise comparisons are done acrossall the genus Calylophus appearsstrongly monophyletic datasets, it is clearthat the incongruent datasets are (BS599; DI56),but relations hips ofthis genus tospe- ITS and rps16 (P50.01).The above-mentioned conict ciŽc sections of Oenothera are uncertain. However,there between the placement of Clarkia species in the ITS- is strong supportfor a lineage composed ofthe rest of only and cp-onlytopologies suggests thatthese taxa Oenothera 1 Gaura 1 Stenosiphon (Node F; BS596; are the source ofthe dataset incongruence. Thus, ad- DI54).Among the Oenothera species in this group ditional ILD tests were conducted using the ITSand there is limited resolution,although there is strong rps16 datasets excludingboth Clarkiarostra ta and C. support for O.canescens 1 O. speciosa 1 O. rosea afŽnis and excludingeachof these taxaseparately .All (BS592; DI53)and fora sister taxonrelationsh ip be- three ofthese additional analyses yielded insigniŽcant tween O. speciosa and O. rosea (BS588; DI53). Oenothera P-values (bothtaxa exclude d, P50.53; C. rostrata ex- tetraptera 1 O.multicaulis are alsow ell supported as cluded, P50.07; C. afŽnis excluded, P50.39);how ever, sister taxa(BS 5100; DI58).Further ,there is moderate it appearsthat the allopolyploid C. afŽnis is the great- supportfor the monophyly of Gaura 1 Stenosiphonlin- est source ofincongrue nce. Therefore,the datasets are ifolius (Node G; BS574; DI52). not completelyincongruent,only the nrITS and rps16 Chloroplast rps16. Sequences of rps16 for 75 taxa sequence datafor a few taxa.Conseque ntly,webelieve ranged in length from 790–838 bp ,with analigned thatit is usefulto analyze all data sets together.The length of1009 bp (T able2). Of these characters,163 combined dataset included 2893characters for 75 taxa, were parsimony-informative.Within this dataset, per- ofwhich 492were parsimony-informative. cent missing dataper species ranged from 0–2.6%, Overall, the phylogeny inferred from the combined with amean of0.1% and amedian of0%. The strict analysis ofall data sets (Fig. 2)is very similar tothat consensus ofmost-parsim onious trees differs from Fig. shown in Fig. 1,although generally with greater (oc- 2(alldata sets combined, see below)only in the level casionally less) supportfor man ybranches. There is ofsupport and resolution ofa few nodes; thus, the strong supportfor a cladeof Onagreae 1 Epilobieae topology is not shown. Results ofan ILD test con- (BS5100; DI519), with Gongylocarpus sister tothis ducted with trnL-trnF and rps16 datasuggest thatthe clade.Tribe Epilobieae is alsomonoph yletic (BS 5100; twocpdata sets are highly congruent ( P50.63). DI513),and Onagreae is supported asmonoph yletic 2004] LEVINET AL.: RELATIONSHIPSIN ONAGRACEAE 157

(BS592; DI55),with moderate supportfor Xylonagra creased resolution,anexpected result ofthe addition assister tothe rest ofOnagreae (BS 585; DI55). Within ofindels asseparate characters. F or example,twoin- this claderelations hips are less well resolved.How- dels supportthe monophyly oftribe Epilobieae,Žve ever, Clarkia is clearly monophyletic (BS 5100; DI527), indels supportth emonophyly of Clarkia,and three in- and there is moderate supportfor Gayophytum 1 Cam- dels supportthe monophylyof Camissonia sect. Cam- issoniapterosperma (BS579; DI54)as sister to Clarkia issonia. (BS574; DI53). Maximum Likelihood. Maximum likelihood (ML) In accordwith the other analyses ofthe datasets analysis ofthe three region combined dataset wascon- individually andin combination, the majority of Cam- ducted using parameters estimated from the dataset issonia species do not appearto comprise amonophy- with Modeltest v.3.06(Posada and Crandall 1998). letic group;rather species groups form monophyletic This procedure indicated thatthe GTR 1 G 1 I model lineages (oftencorrespond ing torecognized sections) best Žtthe data.The ML model parameters included thatare basalto the remainderofthe taxain the tribe anucleotide frequency ofA 50.3215, C50.1824, (Fig. 2).Thus, there is moderate supportfor a clade G50.2023,and T 50.2938;substitution rate matrix ofA composed of Camissoniacla viformis 1 C. arenaria 1 toC: 1.3557, A toG: 1.07, A toT: 0.3806,C toG: 0.6649, Oenothera 1 Calylophus 1 Gaura 1 Stenosiphon (BS577; CtoT: 1.7271,and GtoT: 1.000;proportion ofin- DI51).W ithin this cladethere is strong supportfor variant sites 50.289;an dagamma rate distributionat Oenothera 1 Calylophus 1 Gaura 1 Stenosiphon (BS599; variablesites with shape (alpha) 50.795.U sing this DI58), but Oenothera is paraphyletic.A cladecom- model, the analysis conducted 20,097rearrangem ents posed of Oenothera sects. Oenothera, Ravenia, Kleinia, Er- and wasstopped beforecompletio n, afterretaining a emia, Anogra, and Contortae is well supported (Ain Fig. tree with the same–ln valuefor 72 hours. This analysis 2; BS588; DI53).In this lineage, O.primiveris 1 O. xy- yielded one tree with –ln 516321.085(Fig. 3). locarpa (BS5100; DI58)are sister tothe rest ofthe Generally,the ML tree has asimilar topology tothat clade (BS586; DI52), and O. tubifera is sister tosects. inferred using parsimony.Anydifferencesare due to Oenothera 1 Kleinia 1 Anogra (BS597; DI54).Sections the increased resolutionofthe ML tree (Fig. 3)com- Kleinia and Anogra form amonophyletic lineage pared toFig. 2;this is not surprising asFig. 2is astrict (BS5100; DI59)that is sister tothe strongly supported consensus ofman yMPtrees, whereas the ML algo- sect. Oenothera (BS5100; DI513). rithm generally yields asingle tree.For example,the Sister tolineage Ais aweakly supported group ML analysis yields atopology with variousclades of composed ofthe other Oenothera sections 1 Calylophus Camissonia species forming agrade atthe baseof a 1 Gaura 1 Stenosiphon (BS552; DI51).Although Oen- lineage of Oenothera 1 Calylophus 1 Gaura 1 Stenosi- othera sect. Pachylophus is strongly supported asm ono- phon,and there is asister taxonrelationsh ip between phyletic (BS 5100; DI511), as is Calylophus (BS5100; Calylophus and Oenothera sect. Pachylophus (i.e., O. caes- DI512),relationsh ips ofthese clades tothe other taxa pitosa and O.psammophila ).In general, the internal in this group remain equivocal.Lik ewise,the relation- branches are muchshorter in the Onagreae above ship of Oenothera ava (sect. Lavauxia)toother taxain Clarkia 1 Gayophytum 1 Camissoniapterosperma , likely this group is uncertain. contributing tothe low resolution among the various Alineage including the rest of Oenothera 1 Gaura 1 Camissonia clades in Figs. 1,2. Further, within this Stenosiphon is well supported (Bin Fig. 2;BS 5100; Camissonia 1 Oenothera 1 Calylophus 1 Gaura 1 Sten- DI57),although there is generally limited resolution osiphon clade,branchlengths are especially short atthe among the Oenothera species in this clade.Strongly tips of many Oenothera and Gaura species. supported groups include O.canescens (sect. Gauropsis) 1 O. speciosa (sect. Xylopleurum ) 1 O. rosea (purple- DISCUSSION owered sect. Hartmannia) (BS5100; DI510), O. specio- Relationshipswithin Epilobieae. In agreement sa 1 O. rosea (BS591; DI54), and O.tetraptera (white- with the nrITS analysis ofBaum et al.(1994), this tribe owered sect. Hartmannia) 1 O.multicaulis (yellow- is well supported asa monophyletic lineage, with owered sect. Hartmannia) (BS5100; DI512).The lin- Chamerion sister toa monophyletic Epilobium (Figs. 1– eage comprised of Gaura 1 Stenosiphon has moderate 3).Relationsh ips within Epilobium (Fig. 1)are generally support (BS585; DI52). consistent with the earlier analysis ofBaum et al. Indels. Atotalof 24 indels wasidentiŽ ed acrossall (1994),with strong supportfor a cladeof all sections three datasets; no indels were coded in the nrITS data, excluding sect. Epilobium (i.e., E. ciliatum, E. obcordatum , 13indels were identiŽed in the trnL-trnF data, and 11 and E. rigidum) (BS5100; DI511)and fora lineage indels were coded in the rps16 data.P arsimonyanal- composed ofsects. Currania (E. pygmaeum) 1 Boisdu- ysis ofthe combined dataset plus indels yielded ato- valia (E. densiorum ) 1 Zauschneria (E. canum) (BS591; pology (notshown) thatis nearly identical tothat DI53).In addition, the present study concurswith shown in Fig. 2.The only differencesare due toin- Baumet al.(1994) in supporting asister relationship 158 SYSTEMATIC BOTANY [Volume 29

FIG.3. Phylogramfrom the MLanalysis ofthe combinednrITS and cp trnL-trnF and rps16 data (75 taxa; 2ln516321.085). 2004] LEVINET AL.: RELATIONSHIPSIN ONAGRACEAE 159 between sects. Boisduvalia and Zauschneria (BS597; C. heterandra (sect. Heterogaura ; ’’Sympherica’’ clade of DI54).Interesting ly,although analysis ofthe ITSdata Ford and Gottlieb2003) in the ITSanalysis and with alone suggests thatsect. Epilobium may bemonoph y- C. rostrata (sect. Sympherica)in the cpanalyses. Ofthe letic (this study and Baumet al.1994), support for the species included in the present study, C. heterandra and inclusion ofthe enigmatic species E. rigidum (Raven C. rostrata are the closest tothe parental species (Ford 1976)is weak.A ddition ofcp trnL-trnF dataprovides and Gottlieb2003), as C. delicata is presumed (Lewis moderate supportfor the placement ofthis taxonout- and Lewis 1955)to ha ve arisen asan allopolypl oid ( n side ofsect. Epilobium and sister toall sections except 5 18) between C. unguiculata (n 5 9; sect. Phaeostoma; sect. Epilobium (BS577; DI52;Fig. 1),a position sup- ’’Sympherica’’cladeof F ord and Gottlieb2003) an d C. ported byseed morphology (Seaveyet al.1977). epilobioides (n 5 9; sect. Sympherica). Relationships within Onagreae. The genus Xylon- Our results show moderate supportfor Gayophytum agra has not often been included in phylogenetic anal- 1 Camissoniapterosperma as sister to Clarkia. The ap- yses ofOnagraceae .The morphologicalanalysis ofthe parently close relationship between Gayophytum and family (Hochet al.1993) placed Xylonagra in a basal Camissoniapterosperma (sect. Chylismiella ) supports polytomy in tribe Onagreae.Recently,afamily-wide speculationby Raven (1962,1969) and Lewis and study ofOnagraceae based on acombined analysis of Szweykowski (1964),but neither suggested aclose re- cp rbcL and ndhF sequence data(Levin et al.2003) pro- lationship ofthese species with Clarkia.This is the Žrst vided weak supportfor Xylonagra as sister to Gayoph- phylogenetic study toinclude allof these taxa.Al- ytum, with Clarkia sister to Xylonagra 1 Gayophytum. In though anearlier morphologicalph ylogeny(Hochet the present study,there is moderate supportfor Xylon- al.1993) show ed Gayophytum assister toEpilobieae , agra assister tothe rest ofOnagreae (BS 585; DI55; Levin et al.(2003), using similar taxonsam pling tothe Fig. 2). Xylonagra is clearly morphologically distinct, Hochet al.(1993) study and the more slowly-evolving with tubularred owers, woody capsules,and asym- ndhF and rbcL genes, showed aclose relationship metrically winged seeds. These charactersare homo- among Xylonagra, Gayophytum and Clarkia. In the pre- plasticwithin the family,butare undoubtedlyinde- sent study,the nrITS datado not suggest aclose re- pendentlyderived and autapomorphic in Xylonagra. lationship between Gayophytum and C. pterosperma , al- Within the rest ofOnagreae , Clarkia is strongly though these relationships were poorly resolved.By monophyletic. Previous studies ofthis genus used a contrast,both chloropla st datasets strongly supporta variety ofoutgroups, including single species of Epi- sister relationship between Gayophytum and C. pteros- lobium (Sytsmaet al.1990), Oenothera (Gottlieband perma.Raven (1962,1969) suggested the possibility of Ford 1996)and Gayophytum (W.J.Hahn et al.,in mss.). aclose relationship between these taxabased on sim- The monophylyof Clarkia is supported bya PgiC gene ilarities in habitand unique white petals with ayellow duplication(Gottlieband Ford 1996;F ord and Gottlieb base.However, he alsonoted thatthese taxawere quite 2003)and the shared presence ofunicellular papillae distinct, with C. pterosperma marked bystrongly auta- on the stigmatic surface(H eslop-Harrison 1990;H och pomorphic seeds with thick papillatewings, and Gay- et al.1993). Relationsh ips among the sections of Clarkia ophytum by2-loculed ovaries. generally agree with other more detailed analyses Relationships among the rest ofthe Camissonia spe- (Sytsmaet al.1990; Gottlieb and Ford 1996;F ord and cies included in this study are less clear,but, as pre- Gottlieb2003; W .J.Hahn et al.,in mss.); the fewdif- viously suggested byLe vin et al.(2003), there seems ferences between ourresults and previous studies little doubtthat the genus is not monophyletic. Con- were not strongly supported in one or bothanalyses straining Camissonia tobe monoph yletic requires eight being compared.H owever, ourresults show incongru- more steps (0.4%longer trees), and topologies with ence between the nrITS and rps16 data for C. afŽnis Camissonia constrained tobe monoph yletic have lower and C. rostrata.This appearsto be the result ofan al- likelihood values,although most comparisonswere not lopolyploid origin of C. afŽnis (n 5 26),which accord- signiŽcant after Bonferron icorrection (one-tailed S-H ing toLewis and Lewis (1955)arose viah ybridization test; values varied from P50.064 to P50.015across all between acommon ancestor of C. davyi and C. tenella constrained topologies, Bonferroni-corrected P50.017). (sect. Godetia; n 5 17)and aspecies ofsect. Sympherica Instead, anumber ofw ell supported clades of Camis- (n 5 9),and the resulting divergencebetween nuclear sonia species appearto form abroadly paraphyletic DNAand the maternally inherited plastid DNA.The grade within which is nested amonophyletic lineage other twoallopolyploids in ourstudy , C. delicata (sect. composed of Oenothera 1 Calylophus 1 Gaura 1 Sten- Connubium) and C. pulchella (sect. Clarkia),associate osiphon.All butone ofthese monophyletic groups of with the sectional representative ofon eoftheir pre- Camissonia species correspond tosections or groups of sumed parents determined from more detailed studies sections asdelimited byRav en (1969);only C. gracili- (Lewis and Lewis 1955;Gottlieb and Ford 1996;F ord ora,placedin sect. Tetrapteron byRaven (1969),ap- and Gottlieb2003). Thus, Clarkiadelicata groups with pears outside thatsection. Sections Eremothera and 160 SYSTEMATIC BOTANY [Volume 29

Camissonia are strongly monophyletic in allof the anal- thers. However,measurements taken byWL Wfrom yses (Figs. 1–3), and together they form acladewith one species eachof Clarkia, Gayophytum,and every sec- limited support(Fig. 2).These twosections are dis- tion of Camissonia indicate thatthe situation is not tinctive (Raven 1969);all of the species ofsect. Ere- quite so simple,with anthers exhibiting arange ofat- mothera have white petals,while those ofsect. Camis- tachment points. Nevertheless, thatanalysis veriŽes sonia haveyellow ones. The twosections do share sev- thatall taxa examined of Clarkia and Camissonia sect. eral characters(enumera ted below),but these are also Tetrapteron (including C. graciliora ), as well as Gayoph- shared with some ofthe other Camissonia species. ytum and C. pterosperma ,have anther attachments clos- Section Tetrapteron excluding C. graciliora also ap- er tothe basethan do other taxaof Camissonia. Thus, pears monophyletic. Itshould not betoo surprising this anther charactergenerally supportsthe relation- that C. graciliora does not,in fact,belong within sect. ships inferred from the DNAsequence data,except Tetrapteron ascircumscribed byRaven (1969);this spe- thatit suggests acloser relationship of C. graciliora to cies differs from allother members ofits section except the rest ofsect. Tetrapteron than do the DNAdata. for C. palmeri (notincluded in this study) byhaving an Our results suggest astrongly monophyletic Cam- annualhabit and winged capsules(all others are pe- issonia sect. Eulobus (C. californica 1 C. crassifolia ) that rennials and lackwings; Raven 1969).H owever, the is moderately supported assister to C. claviformis 1 C. section asdelimited byRaven is characterizedbya arenaria 1 Oenothera 1 Calylophus 1 Gaura 1 Stenosi- unique sterile projection ofthe ovary,elevating the phon. Species of Camissonia sect. Eulobus are character- owers abovethe leaves and connecting toa very short ized bydeeply pinnatiŽd leaves thatare mostly re- oraltube with aeshydisk atthe tubeentrance . stricted tothe baseof the plant,a oraltube closed by Close re-examination (byWL W)of this sterile ovary aeshydisk, petals Žnely ecked with red near the projection has shown thatin the perennial members of base,seeds brown with maroon dots, virgate inores- this section the projection appearscontinuous with cences, and usually pubescentanthers. The lastthree boththe short oraltube and the fertile partof the charactersmay besynapom orphies, butthe others ap- ovary,whereas in the twoannualspecies there is a pearto be homoplas tic.A dditionally,ourresults sup- clearabscission line atboth of these junctures. There- porta sister taxonrelationsh ip between C. claviformis fore,this characterappears hom oplasticas originally (sect. Chylismia) and C. arenaria (sect. Lignothera). Raven described. Raven (1969)suggested ane volutionary link (1969)suggested thatsect. Chylismia is closely related between these twoannuals and sect. Holostigma based to sect. Lignothera,based on the shared presence of on the similarity ofthe entire leaves in bothgroups. pedicellate,noncontorted capsules,seeds in tworows Our results supportthis proposed relationship, as per locule,and relativelybroad,primarily basalleaves. Camissoniagracili ora (sect. Tetrapteron ) 1 C. cheiranthi- Phylogeneticrelationships within the monophyletic folia (sect. Holostigma) 1 C. andina (sect. Nematocaulis ) group of Oenothera 1 Calylophus 1 Gaura 1 Stenosiphon form amoderately supported clade(Fig. 2).This clade show that Oenothera is not monophyletic ascurrently then forms aweakly supported lineage with sects. circumscribed.When Oenothera is constrained tobe Camissonia and Eremothera.There is additional mor- monophyletic, there is acostof 16 steps (trees 0.8% phological supportfor this relationship:all of these longer), and such topologies havesigniŽcantly lower taxahave sessile capsulesthat are often coiled or con- likelihood values (one-tailed S-Htest; signiŽcance v al- torted, seeds in asingle row per locule,simple sub- ues from P,0.001 to P50.003across all constrained entire leaves, no eshy disk atthe oraltube entrance , trees, Bonferroni-corrected P50.017).N evertheless, as and the absenceof non-ultra violet-reective areason in Camissonia,there is strong supportfor man ymono- the petals.A dditionally,allbut one species in this phyletic groups within Oenothera.These groups largely group has anann ualhabit, although this lastcharac- correspond torecognized sections, which together teristic seems tohave evolved several times indepen- form agrade within which Calylophus, Gaura, and Sten- dently in adjacentclades. This entire group (i.e.,sects. osiphon are nested (Figs. 1–3). Le vin et al.(2003) Ž rst Eremothera, Camissonia, Holostigma, Nematocaulis , and C. suggested these relationships, although taxonsam- graciliora )may best berecognized asa more narrowly pling waslimited. This broad’ ’Oenothera clade’’ ap- delimited genus Camissonia. pears tocom prise twoprimary lineages thatgenerally Two ofthe analyses presented here weakly support correspond tothe twogroups deŽned byT obeet al. the position of Camissonia sect. Tetrapteron s.s. (i.e.,ex- (1987),based on capsuleand seed coatanatomy. The cluding C. graciliora )assister to Clarkia 1 Gayophytum group containing O.fruticosa and O.brachycarpa is more 1 Camissoniapterosperma (Figs. 1,2), but clearly more closely related to Gaura 1 Stenosiphon (lineage B; Fig. dataare needed. Unlike most taxaof Onagreae ,allspe- 2)than toother Oenothera species thatcomprise lineage cies of Camissonia sect. Tetrapteron , including C. gracil- A(Fig. 2).H owever, unlike results from Levin et al. iora and C. palmeri (Raven 1969),as w ell asall species (2003), Calylophus appearsmore closely related tolin- of Clarkia (Lewis and Lewis 1955),ha ve basiŽxed an- eage Bthan tolineage A;in the earlier study Calylophus 2004] LEVINET AL.: RELATIONSHIPSIN ONAGRACEAE 161 hartwegii is closer to Oenotheraelata ,which is in lineage Stenosiphon )costsonly anadditional 4steps (0.2%lon- Ain the present study. Neither ofthese studies pro- ger), and, not surprisingly,the likelihood ofthe con- vides strong supportfor the exactplacem ent of Caly- strained topologies is not statistically lower (S-Htest; lophus,although it is clearly within the ’’Oenothera P-values from 0.264to 0.735). H owever, in addition to clade’’.Together with Oenothera sects. Pachylophus and the molecular support,morphol ogy supportsa close Lavauxia, Calylophus forms agrade atth ebaseof the relationship between Gaura and Stenosiphon . Stenosi- strongly monophyletic lineage B(Fig. 2). phon has ahabitand inorescence structure very sim- Analyses ofthe combined DNAsequence datasets ilar to Gauraparvi ora ,and wasonly separated from (Figs. 1–3) strongly supportthe monophyly of Oenoth- Gaura based on the autapomorphy offruits having one era lineage Aand relationships within this group. loculeand fourovules, with only one maturing (Spach There is alsomorphol ogicalsupport for this clade,but 1835;Johansen 1931;Rav en 1964).F ourmorphol ogical it is complicatedbythe position of Oenothera sect. Pa- characterssupport the Stenosiphon 1 Gaura lineage: 1) chylophus and bysom eapparently homoplasticchar- fruits condensed,indehiscent; 2)ovule number re- acters. Oenothera lineage A 1 sect. Pachylophus share a duced (1–8) from the muchgreater numbers in Oen- uniquethick seed endotesta (Tobeet al.1987). These othera;3)septa fragile ,incomplete and absent atma- twogroups alsoshare areticulate topapillate seed turity or wholly absent;and 4)presence ofan indusi- surface(also in sect. Megapterium ),exotestacells irreg- um atthe baseof the stigma lobes.In addition, Sten- ularly swollen or collapsed (alsoin sect. Megapterium ), osiphon 1 allspecies of Gaura except G.mutabilis share seed mesotesta with 1–3 cell layers (alsoin O. havardii), clawed, white topink petals.Consider ing the present and seed mesotesta cells crushed (alsoin O. havardii support for Stenosiphon nested within Gaura and sim- and sect. Megapterium )(Tobeet al.1987). In theanal- ilar results from Hoggard et al.(2004), it may bebest yses presented here,there is limited resolution of to include Stenosiphon in ataxonomicgroup with Gau- whether sect. Pachylophus is more closely related tolin- ra,and bothof these genera in anexpanded Oenothera. eages Aor B; ML analysis placesthis section assister Within Gaura,the monophyly of G. hexandra is not to Calylophus,and together they are sister tolineage B, supported, aŽnding consistent with results ofG .Hog- although the branchlengths involved are very short gard, U.Oklahoma (unpubl.data). The twosubspecies (Fig. 3).In addition, it costsonly one additional step of Gaurahexandra havebeen treated asdistinct species for section Pachylophus tobe constrained to Oenothera in the past(e .g.,M unz1965), in partbecause they dif- lineage A.In view ofthe equivocalnature ofthese fer consistentlyin having 3vs. 4-merous owers. By results, additional dataare needed toclarify the rela- contrast,the ywere grouped together byRaven and tionships of Calylophus and Oenothera sects. Pachylophus Gregory (1972)because of their shared predominant and Lavauxia toone another and tothe rest of Oenoth- autogamy and general similarity. However, bothsub- era. species share aunique 21bp insertion in rps16; there The well supported lineage Bis composed of Oen- is alsoa 5bpdeletion in trnL-trnF shared bythem and othera sects. Megapterium , Paradoxus, Gauropsis, Xylopleu- G. demareei,aspecies thatappears closely related in rum, Hartmannia, and KneifŽa,asw ell asth emonophy- ouranalyses and in Hoggard et al.(2004). letic Gaura 1 Stenosiphon clade,with these taxaall shar- Relationships Among Hauya, Circaea, Fuchsia, Lo- ing condensed,winged or angled capsules(T obeet al. pezia, and Megacorax. The placement ofthese taxa 1987).Am ong these Oenothera taxa, sect. Hartmannia within the family has varied considerably among pre- does not appearto be monoph yletic ascurrently cir- vious phylogeneticstudies ofOnagraceae (see review cumscribed, with ower colorapparently usefulin di- in Levin et al.2003). Strong consensus has emerged for viding the section. Oenotherarosea (sect. Hartmannia asister relationship between Fuchsia and Circaea (Syts- with rose-purple petals)is sister to O. speciosa (sect. maet al.1991b; Bult and Zimmer 1993;Conti et al. Xylopleurum with white topink owers), and together 1993;Le vin et al.2003; also observed in the present they are sister to O.canescens (sect. Gauropsis with study). Further, based on ourresults and those ofLev- white petals with pink ecks) in astrongly supported in et al.(2003), Megacorax is sister toa monophyletic clade(Figs. 1–3). The other twospecies sampled from Lopezia.However, conicts remain regarding the place- sect. Hartmannia, O.tetraptera (white petals)and O. ment of Hauya and the Lopezia lineage.Crisci et al. multicaulis (yellow petals or yellow with red center), (1990;nrD NArestriction sites), Sytsma atal. (1991b; comprise aseparatem onophyletic lineage.However, cpDNArestriction sites), and Bult and Zimmer (1993; relationships ofthese twoclades toeach other and to nrRNAsequence data)placed Hauya as sister to Circaea the rest oflineage Bare equivocal,suggesting the need 1 Fuchsia,whereas Martin and Dowd (1986;amino formore data. acidseq uence data)and Hochet al.(1993; morpholo gy) Nested among these Oenothera species is themod- placed Hauya in acladewith tribes Onagreae and Epi- erately supported cladeof Gaura 1 Stenosiphon . Con- lobieae.In contrast,Conti et al.(1993; rbcL sequence straining Gaura tobe monoph yletic (i.e.,excluding data)and Levin et al.(2003; rbcL and ndhF sequence 162 SYSTEMATIC BOTANY [Volume 29 data)placed Hauya assister toall Onagraceae exclud- Several conclusionscanbe drawn from this study. ing Ludwigia.Our combined ITS 1 trnL-trnF analysis In general, nuclearand chloroplast DNAsequence data (Fig. 1)provides moderate supportfor this latter place- sets were congruent,particularlyregarding placement ment of Hauya;the ITSdata alone alsoyield this rela- of Hauya and Lopezia. Further, Megacorax deŽnitely tionship,butwith limited support. does not belong within Lopezia,butit is strongly sup- Similarly,someprevious studies (nrRNAsequence ported assister toit. Tribes Epilobieae and Onagreae data:Bult and Zimmer, 1993;and aminoaciddata: are monophyletic and are supported assister taxa. Martin and Dowd 1986)placed Lopezia assister toall Within Onagreae current generic delimitations are in Onagraceaeexcept Ludwigia,whereas others (cpDNA: question,though Clarkia is well supported asa mono- Sytsma et al.1991b; Conti et al.1993) placed Lopezia phyletic lineage. Neither Camissonia nor Oenothera are assister only totribes Epilobieae and Onagreae.Anal- monophyletic ascurrently circumscribed,with the for- yses including both Lopezia and Megacorax (cpDNA se- mer broadly paraphyletic and comprising agrade of quence data:Levin et al.2003) concurred with this lat- monophyletic lineages thatcorrespond totraditional ter placement. Our results, bothfrom ITSalone and sections or groups ofsections. Further dataare needed the combined analysis ofnrITS and cp trnL-trnF (Fig. toclarify relationships of Camissonia lineages relative 1),conŽ rm these relationships. These differencesin to- toeach other and to Clarkia, Gayophytum, and Oenothera. pology prompted us toexplore whether the addition Oenothera is alsoparaphylet ic,but becomes monophy- ofne wly discovered Megacoraxgracielanus is the key to letic if Gaura, Stenosiphon , and Calylophus are included. resolving these relationships. Parsimony analyses were This broadly deŽned Oenothera would then containall conducted using the ITS-only dataset with: 1) Mega- ofthe species thathav eauniquestigma with fournon- corax excluded,and 2) Megacorax and all Lopezia species commissural lobes (further modiŽed in the peltate except L. langmaniae excluded (most previous studies stigma of Calylophus).W ithin this group, Gaura is haveonly included one Lopezia species; more taxaw ere monophyletic only if Stenosiphonlinifolius is included, included in the present study totest the relationship aresult consistent with Hoggard et al.(2004). Pending of Megacorax to Lopezia). When only Megacorax was ex- amore thorough review ofm orphologicalcharacters, cluded, relationships did not change; however,when the taxonomy ofthese genera clearly need tobe re- the additional Lopezia species were alsoexclude d, the evaluated,and revised circumscriptions are forthcom- placement of Hauya and Lopezia switched. Thus, it ap- ing (Wagner et al.,in prep.). pears thatthe previouslyreported incongruence be- tween nuclear and chloroplastdatais likely due toa ACKNOWLEDGEMENTS .The authorsthank SteveO’Kane for lackof sampling within the Lopezia 1 Megacorax line- generouslyproviding Lopezia genomicDN Asand ITSsequences , Carrie McCrackenforsequencin ghelp,MargaretEv ans forsharing age. genomicDN As, twoanonymous reviewerswhose comments From the present analyses it is now clearthat H au- greatlyimprov ed themanuscript ,and gratefullyacknowle dgethe yeae are sister toall Onagraceae min us Jussiaeeae.The SmithsonianInstitution Andrew W .MellonF ellowships inStruc- Fuchsieae 1 Circaeeaelineage wasnext todiv erge,and ture and EvolutionofT errestrialEcosystems and the National Sci- Lopezieae 1 Megacorax are sister toGongylocarpe ae 1 ence FoundationD EB 9407270 toKJS. Epilobieae 1 Onagreae.In accordwith Levin et al. (2003),there is strong supportfor the placement of LITERATURE CITED Gongylocarpus in its own tribe Gongylocarpeae(Smith and Rose 1913;Levin et al.2003) and sister toa re- BALDWIN,B. G.1992. Phylogenetic utilityof the internal tran- scribedspacers ofnuclear ribosomalDN Ainplants: an ex- deŽned Onagreae 1 Epilobieae. ample fromthe Compositae . Molecular Phylogenetics and Evo- MolecularEv olution. Comparison ofthe nrITS re- lution 1: 3–16. gion and the twocpregions (Table2) shows thatthe ———, M. J. SANDERSON, J. M. PORTER, M. F. WOJCIECHOWSKI , S. ITSdata have the highest proportion ofparsim ony-in- S. CAMPBELL, and M. J. DONOGHUE.1995. The ITSregion of nuclear ribosomalD NA: avaluable source ofevidence on formative(PI)characters, with the trnL-trnF data hav- angiospermphylogen y. Annalsof theMissouri Botanical Garden ing the lowest, butfairly similar to rps16.Although ITS 82: 247–277. has the highest pairwise distance between sequences, BAUM,D.A., K. J.S YTSMA, and P. C. HOCH.1994. Aphylogenetic yielding more PIcharacters, there is greater conict analysis of Epilobium (Onagraceae)based on nuclear ribosom- among these charactersthan in the cpDNA,as exem- al DNAsequences. Systematic Botany 19: 363–388. BREMER,K. 1988. The limitsof amino acid sequence data inan- pliŽed bythe low CIand RIv alues forITS compared giospermphylogene tic reconstruction. Evolution 42: 795–803. to trnL-trnF and rps16 (Table2). It is interesting tocon- BULT,C. J.and E. A. Z IMMER.1993. Nuclear ribosomalRN Ase- sider why the ITSdata ha ve such alow CI;possibly in quences forinferring tribal relationsh ips withinOnagraceae . this faste volving region there has been toom uchdi- Systematic Botany 18: 48–63. CONTI, E., A. FISCHBACH, and K. J. SYTSMA.1993. Tribal relation- vergence,resulting in multiple changes per site.Alter- ships inOnagracea e: implicationsfrom rbcL sequence data. natively,it may bedue toincomplete concerted evo- Annalsof the Missouri Botanical Garden 80: 672–685. lution ofthe multiple ITScopies in the nucleargenome . ———, A. LITT, and K. J. SYTSMA.1996. Circumscriptionof M yr- 2004] LEVINET AL.: RELATIONSHIPSIN ONAGRACEAE 163

tales and theirrela tionshipsto other rosids: evidence from ——— and J. SZWEYKOWSKI.1964. The genus Gayophytum (Ona- rbcL sequence data. American Journalof Botany 83: 221–233. graceae). Brittonia 16: 343–391. CRISCI,J.V ., E. A. Z IMMER, P. C. HOCH, G. B. JOHNSON, C. MUDD, MADDISON,W.P.and D.R. M ADDISON.2000. MacClade 4: analysis and N. PAN.1990. Phylogenetic implicationsof ribosomal ofphylogen yand character evolution. Sunderland,Massa- DNArestrictionsite v ariationin the plant familyOnagracea e. chusetts: Sinauer Associates. Annalsof theMissouri Botanical Garden 77: 523–538. MARTIN,P.G.and J.M. D OWD.1986. Phylogenetic studies using CUNNINGHAM,C. W.1997. Can threeincongruen ce testspredict protein sequences withinthe order Myrtales. Annalsof the when data should be combined? Molecular Biology and Evo- Missouri Botanical Garden 73: 442–448. lution 14: 733–740. MORGAN,D.R. and D.E. S OLTIS.1993. Phylogenetic relationships DAVIS,J.I., M.P .S IMMONS, D. W. STEVENSON, and J. F. WENDEL. amongmembers of Saxifragace ae sensu lato based on rbcL 1998. Data decisiveness, data quality,and incongruence in sequence data. Annalsof the Missouri Botanical Garden 80: 631– phylogenetic analysis: an example fromthe monocotyle dons 660. usingmitochond rial atp Asequences. Systematic Biology 47: MUNZ,P.A. 1965. Onagraceae. North American Flora II 5: 1–278. 282–310. O’KANE, S. L., JR. and B. A. SCHAAL.1998. Phylogeneticsof Lopezia DONOGHUE,M.J., R. G.O LMSTEAD, J. F. SMITH, and J. D. PALMER. (Onagraceae): evidence fromchloropla stD NArestriction 1992. Phylogenetic relationships ofDipsacales based on rbcL sites. Systematic Botany 23: 5–20. sequences. Annalsof theMissouri Botanical Garden 79: 333–345. OXELMAN, B., M. LIDE´ N, and D. BERGLUND.1997. Chloroplast rps16 FARRIS, J. S., M. KA¨ LLERSJO¨ , A. G. KLUGE, and C. BULT.1994. Test- intronph ylogenyof the tribe Sileneae (Caryophyllaceae ). Plant ingsigniŽ cance ofincongruen ce. Cladistics 10: 315–319. Systematics and Evolution 206: 393–410. ———,———,———,and ———.1995. ConstructingasigniŽ- PLITMANN, U., P. H. RAVEN, and D. E. BREEDLOVE.1973. The sys- cance testfor incongruen ce. Systematic Biology 44: 570–572. tematics ofLopezieae (Onagraceae). Annalsof theMissouri Bo- FELSENSTEIN,J.1985. ConŽdence limitson phylogenies: an ap- tanical Garden 60: 478–563. proach usingthe bootstrap . Evolution 39: 783–791. POPP, M. and B. OXELMAN.2001. Inferringthe history of the poly- FORD,V.S.and L. D.G OTTLIEB.2003. Reassessmentof phyloge- ploid Silene aegaea (Caryophyllaceae) usingplastid and ho- netic relationshipsin Clarkia sect. Sympherica. American Journal moeologousnuclear DNAsequences. Molecular Phylogenetics of Botany 90: 284–292. and Evolution 20: 474–481. GILBERT,D.G.1993. SeqApp: Abiosequence editorand analysis POSADA,D.and K. A. C RANDALL.1998. Modeltest: testingthe application. ftp://iubio.bio.indiana.edu/molbio/seqapp/. modelof D NAsubstitution. Bioinformatics 14: 817–818. AVEN GOLDMAN, N., J. P. ANDERSON, and A. G. RODRIGO.2000. Likeli- R ,P.H. 1962. The systematicsof Oenothera subgenus Chylis- hood-based testsof topologies inphylogen etics. Systematic mia. University of California Publicationsin Botany 34: 1–122. Biology 49: 652–670. ———.1964. The genericsubdivision ofOnagraceae ,tribeOna- greae. Brittonia 16: 276–288. GONZA´ LEZ ELIZONDO,M.S., I.L. L O´ PEZ ENRIQUEZ, and W. L. ———.1969. Arevision ofthe genus Camissonia (Onagraceae). WAGNER. 2002. Megacorax gracielanus (Onagraceae), anew ge- Contributionsfrom theUnited StatesN ational Herbarium 37: 161– nus and species fromDurango, Mexico. Novon 12: 360–365. 396. GOTTLIEB,L. D.and V.S.F ORD.1996. Phylogenetic relationships ———.1976. Generic and sectional delimitationin Onagracea e, amongthe sections of Clarkia (Onagraceae)inferred from the tribeEpilobiea e. Annalsof theMissouri Botanical Garden 63: nucleotidesequencesof PgiC. Systematic Botany 20: 45–62. 326–340. HERSHKOVITZ,M.A. and E. A. Z IMMER.1996. Conservation pat- ———.1979. Asurvey ofreproductiv ebiologyin Onagracea e. ternsin angiosperm rD NAITS2sequences . Nucleic Acids Re- NewZealand Journalof Botany 17: 575–593. search 24: 2857–2867. ———.1988. Onagraceae as amodelof plant evolution. Pp.85– HESLOP-HARRISON,Y.1990. Stigmaform and surface inrelation to 107 in Plant evolutionary biology: asymposium honoring G.Led- self-incompatibilityin the Onagraceae. Nordic Journalof Botany yard Stebbins, eds. L. D.Gottlieb and S.K. Jain. London: Chap- 10: 1–19. man and Hall. HIPP,A. L., J.C. H ALL, and K. J. SYTSMA.Inpress. Phylogenetic ———and D.P.G REGORY.1972. Arevision ofthe genus Gaura accuracy and the ILD. Systematic Biology . (Onagraceae). Memoirs of the Torrey Botanical Club 23: 1–96. OCH RISCI OBE ERRY H ,P.C., J.V .C , H. T , and P. E. B .1993. Acladistic SEAVEY,S.R., R. E. M AGILL, and P. H. RAVEN.1977. Evolutionof analysis ofthe plant familyOnagracea e. Systematic Botany 18: seed size, shape, and surface architecture inthe tribe Epilo- 31–47. bieae (Onagraceae). Annalsof the Missouri Botanical Garden 64: HOGGARD,G.D., P.J.K ORES, M. MOLVRAY, and R. K. HOGGARD. 18–47. 2004. The phylogenyof Gaura (Onagraceae)based on ITS, SHIMODAIRA, H. and M. HASEGAWA.1999. Multiple comparisons ETS, and trnL-F sequence data. American Journalof Botany 91: oflog-likeliho odswith applica tionsto phylogen etic inference. 139–148. Molecular Biology and Evolution 16: 1114–1116. JOHANSEN,D.A. 1931. Studieson themorpholog yofOnagraceae SMITH,J.D .and J.N .R OSE.1913. Amonograph ofthe Hauyeae IV. Stenosiphon linifolium . Bulletin of the Torrey Botanical Club and Gongylocarpeae, tribesof the Onagraceae . Contributions 57: 315–326. from the United StatesN ational Herbarium 16: 287–298. KATINAS, L., J. CRISCI, W. L. WAGNER, and P. C. HOCH. 2004. Geo- SPACH,E. 1835. Monographia Onagrearum. Nouvelles Annalesd u graphical DiversiŽcation ofTribes Epilobiea e, Gongylocar- Muse´umd’ Histoire Naturelle, Paris 4: 321–407. peae,and Onagreae (Onagraceae)inN orthAmerica, based SUH, Y., L. B. THIEN, H. E. REEVES, and E. A. ZIMMER. 1993. Mo- on ParsimonyAnalysis of Endemicity and Track Compati- lecular evolution and phylogenetic implicationsof internal bilityAnalysis. Annalsof theMissouri Botanical Garden 91: 159– transcribedspacer sequences ofribosomal D NAinW intera- 185. ceae. American Journalof Botany 80: 1042–1055. LEVIN,R. A., W.L. W AGNER, P. C. HOCH, M. NEPOKROEFF, J. C. SWOFFORD,D.L. 2002. PAUP*. Phylogenetic analysis usingparsi- PIRES, E. A. ZIMMER, and K. J. SYTSMA.2003. Family-level mony(*and othermethods). V ersion4. Sunderland,Massa- relationshipsof Onagraceae based on chloroplast rbcL and chusetts: Sinauer Associates. ndhF data. American Journalof Botany 90: 107–115. SYTSMA,K. J.and J.F .S MITH.1988. DNAand morphology: com- LEWIS,H. and M.E. L EWIS.1955. The genus Clarkia. University of parisons inthe Onagraceae. Annalsof theMissouri Botanical California Publicationsin Botany 20: 241–392. Garden 75: 1217–1237. 164 SYSTEMATIC BOTANY [Volume 29

———and ———.1992. Molecular systematicsof Onagracea e: TABERLET, P., L. GIELLY, G. PAUTOU, and J. BOUVET.1991. Universal examples from Clarkia and Fuchsia.Pp.295–323 in Molecular primersfor ampliŽ cation ofthree non-coding regions of chlo- systematics of plants, eds. P.S.Soltis,D .E. Soltis,and J.J. roplast DNA. Plant Molecular Biology 17: 1105–1109. Doyle. New York:Chapman and Hall. TOBE, H., W. L. WAGNER, and H.-C. CHIN.1987. Asystematic and ———, J. MORAWETZ, J. C. PIRES, M. NEPOKROEFF, E. CONTI, M. evolutionarystudyof Oenothera (Onagraceae): seed coatanat- ZJHRA, J. C. HALL, and M. W. CHASE.2002. Urticalean rosids: omy. Botanical Gazette 148: 235–257. circumscription, rosidancestry ,and phylogenetics based on WHITE, T. J., T. BRUNS, S. LEE, and J. TAYLOR.1990. AmpliŽca tion rbcL,trnL -F, and ndhF sequences. American Journalof Botany and direct sequencingoffungal ribosomalRN Agenesfor 89: 1531–1546. phylogenetics. Pp.315–322 in PCR protocols: aguide tomethods ———, J. F. SMITH, and P. E. BERRY.1991a. The use ofchloropla st and applications, eds. M.Innis, D .Gelfand, J.Sninsky ,and T. DNAtoassess biogeographyand evolution ofmorpholo gy, White.San Diego: Academic Press. breedingsystems, and avonoids in Fuchsia sect. Skinnera YODER,A. D., J.A. I RWIN, and B. A. PAYSEUR.2001. Failure ofthe (Onagraceae). Systematic Botany 16: 257–269. ILDto determine da ta combinabilityfor slow lorisphyloge- ———,———,and L. D.G OTTLIEB.1990. Phylogeneticsin Clarkia ny. Systematic Biology 50: 408–424. (Onagraceae): restrictionsite mapping ofchloropla stDN A. ZIMMER,E. A., E. H. R OALSON, L. E. SKOG, J. K. BOGGAN, and A. Systematic Botany 15: 280–295. IDNURM.2002. Phylogenetic relationships inthe Gesnerioi- ———,———,and P.C. H OCH.1991b. AchloroplastD NAanal- deae (Gesneriaceae) based on nrDNAand cpDNA trnL-trnF ysisof tribal and genericrela tionshipswithin Onagraceae . and trnE-T spacer regionsequence s. American Journal of Bot- American Journalof Botany 78: 222 (Abstract). any 89: 296–311.