Plant Syst. Evol. 231: 77±89 32002)

Phylogenetic relationships in inferred from chloroplast matK and trnL-trnF nucleotide sequence data

D. Potter1, F. Gao1, P. Esteban Bortiri1, S.-H. Oh1, and S. Baggett2

1Department of Pomology, University of , Davis, USA 2Ph.D. Program Biology, Lehman College, City University of New York, New York, USA

Received February 27, 2001 Accepted October 11, 2001

Abstract. Phylogenetic relationships in Rosaceae economically important of temperate were studied using parsimony analysis of nucleo- regions is produced by members of this family, tide sequence data from two regions of the including of 3apples), Pyrus chloroplast genome, the matK gene and the trnL- 3pears), 3plums, , , trnF region. As in a previously published phylog- , and ), 3raspberries), eny of Rosaceae based upon rbcL sequences, and 3strawberries). The family also monophyletic groups were resolved that corre- includes many ornamentals, e.g., species of spond, with some modi®cations, to subfamilies and , but were Rosa 3roses), 3cinquefoil), and polyphyletic. Three main lineages appear to have 3mountain ash). A variety of growth diverged early in the evolution of the family: 1) habits, types, and chromosome numbers Rosoideae sensu stricto, including taxa with a base is found within the family 3Robertson 1974), chromosome number of 7 3occasionally 8); 2) which is traditionally divided into four sub- actinorhizal Rosaceae, a group of taxa that engage families on the basis of fruit type 3e.g., Schulze- in symbiotic nitrogen ®xation; and 3) the rest of the Menz 1964). Spiraeoideae are characterized by family. The spiraeoid , not included follicles or capsules, Rosoideae by , in the rbcL study, was strongly supported as the 3Prunoideae) by , and sister taxon to Maloideae sensu lato. A New World Maloideae by . In the traditionally origin of Maloideae is suggested. The position of circumscribed Rosoideae, the base chromo- the economically important genus Prunus and the some number is x ˆ 7, 8, or 9; in Amygdaloi- status of subfamily Amygdaloideae remain unre- solved. deae, x ˆ 8; in Spiraeoideae, x ˆ 9 315, 17); and in Maloideae, x ˆ 17. The division of the Key words: Rosaceae, Gillenia, Maloideae, family into four subfamilies has not been Amygdaloideae, Spiraeoideae, phylogeny, matK, followed universally. For example, Hutchinson trnL-trnF. 31964) recognized 20 tribes and did not group these into subfamilies. The large and economically important angio- Because of their economic importance and sperm family Rosaceae has a worldwide dis- diversity, the Rosaceae have been the subject tribution and includes over 3000 species in 122 of numerous taxonomic and evolutionary genera 3Heywood 1993). The vast majority of studies. The family is generally considered to 78 D. Potter et al.: Chloroplast DNA Phylogeny of Rosaceae form a natural group united by ¯oral charac- , , and . The teristics 3Robertson 1974, Dickson et al. 1992). results suggested that, since some fruit types Kalkman 31988) suggested that the presence of have evolved several times within the family, a may be the only morphological they are not as reliable as indicators of synapomorphy for the group, but the uncer- relationship as chromosome numbers. tainty of the relationship of Rosaceae to other Takhtajan 31997) incorporated some results families casts some doubt on this conclusion. from recent phylogenetic studies, such as that of This, as well as the relationships among Morgan et al. 31994), into his classi®cation of subfamilies, genera, and species, are subjects the family, in which he recognized twelve of considerable discussion and investigation subfamilies. Exochorda, previously classi®ed in 3summarized in Morgan et al. 1994, Phipps Spiraeoideae, was included in Amygdaloideae, et al. 1991). Recent phylogenetic analyses have Maloideae were expanded to include the tradi- employed a variety of data, including vegeta- tionally spiraeoid genera , Vauqueli- tive, ¯oral, and fruit morphology 3Kalkman nia, and , and both Rosoideae and 1988; Phipps et al. 1991; Rohrer et al. 1991, Spiraeoideae were subdivided. 1994), ¯oral ontogeny 3Evans and Dickinson A long-standing question concerns the or- 1999a, b), wood anatomy 3Zhang 1992), pollen igin of Maloideae. Because of a base chromo- morphology 3Hebda and Chinnappa 1994), some number of 17, it is generally accepted that chloroplast DNA sequences 3Morgan et al. the subfamily originated either via polyploidi- 1994), nuclear gene sequences 3Potter 1997, zation of a spiraeoid ancestor with x ˆ 9 3e.g. Evans et al. 2000), and combined data from Gladkova 1972) or by hybridization between multiple sources 3Evans and Dickinson 1997, two lineages, most likely a spiraeoid with x ˆ 9 1999c), to address questions about the place- and an amygdaloid with x ˆ 8, followed by ment of problematic genera and the relation- polyploidization 3e.g. Sax 1933). The exact ships of the subfamilies. Sequences of the identity of the putative parental lineage3s) internal transcribed spacer 3ITS) regions of remains controversial, however. The rbcL nuclear ribosomal DNA have been used in analysis by Morgan et al. 31994) indicated that phylogenetic studies of two of the subfamilies members of the spiraeoid tribe Sorbarieae 3plus 3Maloideae, Campbell et al. 1995, Rosoideae, the traditionally rosoid genus ) are Eriksson et al. 1998). These studies have done sister to Maloideae sensu lato, a relationship much to improve our understanding of the consistent with data from carpel anatomy anities of particular taxa and the evolution 3Sterling 1966), but this relationship was weak- of speci®c characters in Rosaceae, but a variety ly supported. Morgan et al. 31994) pointed out of questions remain, and all of the cited papers that, since rbcL is a chloroplast-encoded gene, point out areas in which further resolution is phylogenetic hypotheses based upon it repre- needed. sent phylogenies of maternal lineages only. In a phylogenetic study of rbcL gene Thus, the results are not inconsistent with an sequence variation across the family 3Morgan ancestral member of Amygdaloideae with x ˆ 8 et al. 1994), monophyletic groups of genera having been one of the parents involved in an were identi®ed that corresponded, with some ancient hybridization that led to the origin of modi®cations, to all of the subfamilies except Maloideae 3Sax 1933, Phipps et al. 1991). Spiraeoideae, which was shown to be grossly Recent phylogenetic studies of the nuclear gene polyphyletic. The data strongly supported waxy by Evans et al. 32000), however, have not recognition of Rosoideae sensu stricto, exclud- provided any evidence for an amygdaloid ing several taxa with x ˆ 9, and of Maloideae ancestor having been involved in the origin of sensu lato, including several spiraeoid taxa Maloideae; nor have phylogenetic studies of with x ˆ 15 or 17. Weak support was found for several nuclear genes in our laboratory 3e.g. Amygdaloideae sensu lato, including Prunus, Potter et al. 1999). D. Potter et al.: Chloroplast DNA Phylogeny of Rosaceae 79

Recent molecular phylogenetic analyses exist 3Taberlet et al. 1991) and which has been 3Chase et al. 1993, Morgan et al. 1994, Soltis useful in phylogenetic studies of angiosperm et al. 1995, KaÈ llersjoÈ et al. 1998, Soltis et al. groups at various taxonomic levels 3e.g. Gielly 2000) place Rosaceae in a clade with Mora- and Taberlet 1996, Karol et al. 2000, Potter ceae, Rhamnaceae, Urticaceae, and several et al. 2000, Scott et al. 2000, Bortiri et al. in others, which together have been designated press). as order 3Angiosperm Phylogeny Group 1998). Data from several genes further support a sister relationship of Rosaceae to the Materials and methods rest of the families in the order 3Evans and The species sampled for this study are listed in Campbell 2000, Soltis et al. 2000). Table 1. Voucher specimens are deposited in the The goal of this study was to assess U. C. Davis Herbarium 3DAV). Total DNA was phylogenetic relationships across Rosaceae extracted from fresh material using the CTAB using new chloroplast DNA data, with a method 3Doyle and Doyle 1987) or modi®cations primary emphasis on the mostly woody taxa thereof. In the case of two species 3Fragaria vesca not included in Rosoideae sensu stricto, i.e., and Oemleria cerasiformis), the matK and trnL- genera traditionally classi®ed in Spiraeoideae, trnF sequences were determined from di€erent Maloideae, and Amygdaloideae, and taxa of accessions. Primers for PCR and sequencing were pur- Rosoideae with x ˆ 9. We wanted to test the chased from Genosys Biotechnologies, Inc. PCR relationships identi®ed by Morgan et al. 31994) ampli®cations were carried out using the Perkin- and we were especially interested in seeing Elmer GeneAmp II kit. For matK, a 1.5±1.9 kb whether or not we could obtain better resolu- fragment was ampli®ed using trnK685F and tion than was provided by the rbcL data in the trnK2R as primers as described by Hu et al. following areas: 1) the deep branches within 32000). The forward PCR primer, trnK685F, is the Rosaceae phylogenetic ; 2) the anities located at 685 bp upstream of the trnK50 exon of Prunus and other taxa traditionally classi- relative to the Pisum sequence 3Boyer and Mullet ®ed in Amygdaoideae; and 3) the relationships 1988). The reverse PCR primer, trnK2R, occurs at 0 of Maloideae to spiraeoid and/or amygdaloid site 2475 on the trnK3 exon. The trnL-trnF region taxa. In addition, we included Gillenia of tribe was ampli®ed using primers c and f 3Taberlet et al. Gillenieae 3Spiraeoideae; Schulze-Menz 1964), 1991). For both regions, PCR conditions were as follows: 1 minute at 95 °C; 40 cycles of 30 seconds a group not represented in Morgan et al.'s at 95 °C, 1 minute at 55 °C, and 2 minutes at 31994) rbcL study. Several members of other 72 °C; 7 minutes at 72 °C. families of Rosales sensu Angiosperm Phylog- PCR products were puri®ed from agarose gels eny Group 31998) were included as outgroups. with the QIAquick Gel Extraction Kit 3Qiagen We examined nucleotide sequences from Inc.). PCR products and sequencing primers were two regions of the chloroplast genome. The submitted to one of two sequencing facilities on the ®rst of these, the matK gene, has provided U. C. Davis campus, each of which uses an ABI/ useful information in phylogenetic studies at Prism 377 automated sequencer. For matK, the two the intrafamilial level in a number of angio- PCR primers, plus two internal primers, matK3F 0 0 sperm groups 3e.g., Xiang et al. 1998, Hu et al. 35 -TCCCTCTTCTTTGCATTTATTACG-3 ) and matK4R 350-GCGTTACAAAATTTCACTT- 2000). The maturase-encoding gene occurs as a 0 1.5 kb region embedded within a 2.5 kb intron TAGCC-3 ), were used for sequencing. For trnL- trnF, primers c, d, e, and/or f 3Taberlet et al. 1991) that interrupts the two trnK exons 3Sugita were used, in various combinations, because some et al. 1985). The second region of the chloro- primers worked better than others for sequencing plast genome we have examined is a fragment some templates. of about 1 kb, including the trnL intron and Sequences were edited with SEQUEN- the trnL-trnF spacer. This is a noncoding CHERTM 3.1.1 3Gene Codes Corporation). region of cpDNA for which universal primers Boundaries of the matK coding regions were 80 D. Potter et al.: Chloroplast DNA Phylogeny of Rosaceae

Table 1. Accessions included in this study, arranged by subfamilies as treated by Schulze-Menz 31964) Species Origin and Collection Number GenBank Accession Numbers: matK/ trnL-trnF

Maloideae: pannosa Franchet Santa Cruz County, CA DXP 033 AF288098/ AF348540 monogyna Jacq. Guemes Island, WA DP 970517-08 AF288099/ AF348541 serrulata Lindl. U. C. Davis campus DP 970911-01 AF288111/ AF348552 Pyrus caucasica Fed. UCDAa A69.0879 AF288120/ AF348564 E. Greene Placer County, CA DXP 073 AF288126/ AF348570 Amygdaloideae *Prunoideae): Oemleria cerasiformis Point Reyes, CA DXP 059 AF288110/ Ð 3Hook. & Arn.) J. W. Landon Oemleria cerasiformis Guemes Island, WA DXP 069 Ð/ AF348551 3Hook. & Arn.) J. W. Landon Prinsepia sinensis 3Oliv.) Harvard University DXP 191 AF288114/ AF348558 Oliv. ex Bean L. UCDAa EB 88 AF288116/ AF348559 Prunus persica 3L.) SAb AF288117/ AF348560 Sieb. & Zucc. ``54P 455'' L. NCGRc DPRU 393 AF288118/ AF348561 Rosoideae: Winters, CA SO 970424-01 AF288093/ AF348535 Hook. & Arn. betuloides UCDAa DXP 037 AF288095/ AF348537 Torrey & A. Gray foliolosa Benth. Yosemite, CA DP 970427-02 AF288096/ AF348538 paradoxa UCDAa DXP 610 AF288101/ AF348543 3D. Don) Endl. L. San Mateo, CA DXP 026 AF288102/ Ð Fragaria vesca L. Yosemite, CA DP 970427-01 Ð/ AF348545 alabamensis A. Gray BBGd 93.0973 AF288109/ AF348550 Potentilla anserina L. Montara, CA DP 970309-02 AF288113/ AF348556 tridentata 3Pursh) DC. Luther Pass, CA DP 970831-02 AF288119/ AF348562 scandens BBGd 86.0616 AF288122/ AF348566 3Thunb.) Mak. Cham. UCDAa T00292 AF288123/ AF348567 & Schldl. Cham. & Schldl. San Mateo, CA DXP 027 AF288124/ AF348568 Spiraeoideae: dioicus 3Walter) . BBGd 83.0466 AF288094/ AF348536 millefolium UCDAa A74.0245 AF288097/ AF348539 3Torr.) Maxim. Henan, China DXP 613 AF288100/ AF348542 3Lindl.) Rehder 3Muhl. ex Willd.) BBGd 92.0438 AF288103/ AF348554 Baillon 3L.) Moench Cornell University DXP 192 AF288104/ AF348555 D. Potter et al.: Chloroplast DNA Phylogeny of Rosaceae 81

Table 1 3continued) Species Origin and Collection Number GenBank Accession Numbers: matK/ trnL-trnF

Holodiscus discolor 3Pursh) Guemes Island, WA DXP 070 AF288105/ AF348546 Maxim. Ruiz & Pav. BBGd 88.0176 AF288106/ AF348547 ¯oribundus UCDAa A84.0082 AF288107/ AF348548 A. Gray thyrsi¯ora D. Don RBGEe 19841790 AF288108/ AF348549 capitatus UCDAa DP 970702-01 AF288112/ AF348553 3Pursh) Kuntze sorbifolia 3L.) A. Braun BBGd 83.0529 AF288125/ AF348569 densi¯ora Torr. Tahoe N.F., CA DP AF288127/ AF348571 & A. Gray 970619-02 Stephanandra chinensis Hance USNAf 59954 AF288128/ AF348572 californica 3Torr.) UCDAa A77.0200 AF288129/ AF348573 Sarg. Outgroups: Rhamnus californica Eschsch. UCDAa A93.0177 AF288121/ AF348565 Morus alba L. U. C. Davis campus DXP 100 AF400590/ AF400592 Ulmus procera Salisb. U. C. Davis campus DXP 347 AF400591/ AF400593 a UCDA = U. C. Davis Arboretum b SA = DNA kindly provided by S. Arulsekar, U. C. Davis c NCGR = National Clonal Germplasm Repository, Davis d BBG = Berkeley Botanic Garden e RBGE = Royal Botanic Garden, Edinburgh f USNA = U. S. National Arboretum

determined by comparison to that of Pisum order to determine whether or not there was 3Boyer and Mullet 1988). Only the coding regions signi®cant con¯ict between the data from the two were included in the alignment. The matK regions, the incongruence length di€erence 3ILD) sequences were aligned by eye; the trnL-trnF test 3Farris et al. 1994) was used, with heuristic sequences were aligned using ClustalX 3Thomp- searches, as described above, and 100 replicates. son et al. 1997) and re®ned by eye. Sequences Relative support for clades was assessed using and alignments were submitted to GenBank phylogenetic bootstrapping 3Felsenstein 1985), 3Table 1); these are also available, upon request, with 1000 replicates, and by decay indices 3Mishler from the ®rst author. et al. 1991). The latter were determined as follows: All sequence comparisons and phylogenetic After the initial heuristic search was completed and analyses were carried out using PAUP* 3Swo€ord the most parsimonious and their strict con- 2000). Pairwise divergences 3Jukes-Cantor distanc- sensus tree had been examined, the process was es) among sequences were calculated. Phylogenetic repeated eight times, each time increasing the analysis of the data employing maximum parsimo- maximum length of trees to be saved by one step ny was implemented using heuristic searches and over the previous search, beginning with one step 100 replicates of random taxon addition with TBR longer than the most parsimonious trees identi®ed branch-swapping and MULPARS in e€ect. All in the initial analysis. The strict consensus tree from positions were weighted equally; gaps were treated each of these searches was then compared to the as missing values. The matK and trnL-trnF data initial consensus tree to see which branches had were analyzed separately and in combination. In collapsed. 82 D. Potter et al.: Chloroplast DNA Phylogeny of Rosaceae

Results hia, also traditionally placed in Rosoideae but with x=9, varied among the most parsimoni- The matK sequences 3coding region) ranged ous trees. In two trees, this clade was sister to from 1503 to 1521 bp long, and the ®nal Rosoideae sensu stricto 3hereafter, topology alignment included 1551 sites. The trnL-trnF ``A''); in the other two 3e.g., Fig. 1), it was region sequences ranged from 930 to 1083 bp sister to the clade including all other members long, and the ®nal alignment included 1295 of the family except Rosoideae sensu stricto sites. The ®nal combined data matrix therefore 3hereafter, topology ``B''). These results were contained 2846 characters, of which 1592 were sensitive to both the outgroups and the char- constant, 540 were phylogenetically uninfor- acters included in the analysis. When the mative, and 714 3385 from matK and 329 from combined data were analyzed with just Rham- trnL-trnF) were phylogenetically informative. nus or just Morus and Ulmus as outgroups, The ILD test indicated that there was no only trees of topology ``B'' were obtained; any signi®cant incongruence 3 p ˆ 0.30) between other combination of outgroups produced at these two regions of the chloroplast genome. least some trees of topology ``A''. When the For comparisons between and trnL-trnF data were analyzed alone, only trees ingroup taxa, pairwise sequence divergence of topology ``B'' were obtained, regardless of values ranged from 0.1173 3Rhamnus califor- the outgroup3s) included. In contrast, when the nica and Lyonothamnus ¯oribundus) to 0.1846 matK data were analyzed alone, only trees of 3Ulmus procera and Potentilla anserina). topology ``A'' were obtained with all combi- Among the ingroup taxa, these values ranged nations of outgroups except Rhamnus alone, from 0.0024 3between Stephanandra chinensis which produced only trees of topology ``B''. and Neillia thyrsi¯ora) to 0.1461 3between The traditional Spiraeoideae and Amygda- Potentilla anserina and ). loideae were shown to be polyphyletic. Mono- Phylogenetic analysis of the combined data phyly of the traditional Maloideae, all of matrix produced four most parsimonious trees which produce pomes, was strongly supported, 3Length ˆ 2472, CI excluding autapomor- as was that of Maloideae sensu lato, including phies ˆ 0.58, RI ˆ 0.71), one of which is shown 3Spiraeoideae, x ˆ 15) in Fig. 1. Separate analyses of the matK and and Kageneckia oblonga 3Spiraeoideae, x ˆ 17) trnL-trnF data produced trees 3not shown) plus the -producing taxa. Gillenia was that were generally similar, in terms of levels of strongly supported as the sister group to homoplasy and clades recovered, to those from Maloideae sensu lato. Speci®c similarities and the combined analysis. An exception is dis- di€erences between our analysis and that of cussed below. Morgan et al. 31994) are discussed below. In terms of intergeneric relationships in Rosaceae, our results were generally consistent with those from rbcL data 3Morgan et al. 1994). Discussion As in that study, a number of clades were strongly supported, but support for relation- Phylogenetic resolution. Clades within Rosaceae ships among those groups was generally weak. that were strongly supported in our analysis The taxa of Rosoideae sensu stricto 3x ˆ 7) 380% or greater bootstrap support and decay formed a well-supported clade, as did most of index values of 5 or greater in Fig. 1) and that the remaining taxa in the family, with x ˆ 8or of Morgan et al. 31994; decay index values above. Adenostoma, Neviusia, and Rhodotypos, of 4 or greater in their Fig. 2) include: 1) traditionally classi®ed in Rosoideae but with Rosoideae sensu stricto, including members of x ˆ 9, all appeared within the latter clade. The the subfamily with x ˆ 7; our sampling within position of the clade including three actinorhi- this group was admittedly limited; 2) the zal taxa, Cercocarpus, Chamaebatia, and Purs- actinorhizal Rosaceae, including Cercocarpus, D. Potter et al.: Chloroplast DNA Phylogeny of Rosaceae 83

Purshia, and Chamaebatia 3the last not includ- andra 3the last not included in the rbcL study); ed in the rbcL study), a group of taxa dis- 6) Maloideae sensu lato, including the pome- tributed in western that form bearing Maloideae plus Kageneckia and Vau- symbiotic relationships with nitrogen-®xing quelinia;7)Spiraea, Aruncus, and ; actinomycetes of the genus ;3) 8) Adenostoma, Chamaebatiaria, and Sorbaria. Rhodotypos and Neviusia;4)Exochorda, Several di€erences between our results and Oemleria, and Prinsepia; 5) tribe Neillieae, those of Morgan et al. 31994) merit discussion. comprising Physocarpus, Neillia, and Stephan- The ®rst concerns the deep branches

Fig. 1. One of four most parsimonious trees 3Length ˆ 2472, CI excluding autapomorphies ˆ 0.58, RI ˆ 0.71) from phylogenetic analysis of matKandtrnL-trnF sequences of selected Rosaceae and outgroups. Bootstrap values appear above or to the left of, and decay index values below or to the right of, the branches. Branches with decay index values of 0 are also marked with asterisks; they were not present in the strict consensus tree. Fruit types 3ach ˆ ; cap ˆ ; dru ˆ ; drt ˆ drupelet; fol ˆ ; pom ˆ pome) and base chromosome numbers are indicated after species names. The branches along which it is inferred that the ability to accumulate and to form symbiotic relationships with nitrogen-®xing bacteria arose are indicated 84 D. Potter et al.: Chloroplast DNA Phylogeny of Rosaceae within Rosaceae. The rbcL data provided weak the base of the spiraeoid/amygdaoid/maloid support 3decay index 2) for monophyly of a clade makes sense in this regard. This view is group comprised of all taxa except Rosoideae consistent with the classi®cation system of sensu stricto, i.e., including taxa traditionally Takhtajan 31997), who placed the genus in its classi®ed in Spiraeoideae, Amygdaloideae, and own subfamily Lyonothamnoideae. Maloideae, plus several traditionally classi®ed The sister relationship of Sorbaria and in Rosoideae but with x ˆ 9. Most of the deep Chamaebatiaria, and that of Adenostoma to branches in that group were weakly supported. those two genera, were strongly supported in Nested within this large group was a weakly our analysis, whereas relationships among those supported 3decay index 1) clade comprised of three genera were unresolved in the rbcL study. Lyonothamnus and the actinorhizal group. In Our result is consistent with tribal classi®cations contrast, in our study, the position of the 3Hutchinson 1964, Takhtajan 1997) that unite actinorhizal clade varied among the most Sorbaria and Chamaebatiaria, whose members parsimonious trees, appearing as either sister produce follicles, in tribe Sorbarieae, and place to Rosoideae sensu stricto or as sister to all Adenostoma, which produces achenes, in its own members of the family except Rosoideae sensu tribe, Adenostomateae. stricto. The former topology was recovered in Finally, the relative positions of Prunus some or all trees resulting from analyses that and the other drupe-producing genera in the included the matK data 3alone or in combina- family, and of Exochorda, di€er between our tion with the trnL-trnF data) and certain analysis and that of Morgan et al. 31994). Our outgroup combinations. Although this rela- results, like those of Morgan et al.'s 31994) tionship is consistent with the classi®cation of rbcL study and Lee and Wen's 32001) phylo- these taxa in Rosoideae 3Schulze-Menz 1964) genetic analysis of ITS sequences of Amygda- based on fruit type and ¯oral characters, a loideae, strongly supported a clade including variety of data, including trnL-trnF and rbcL Oemleria, Prinsepia, and Exochorda. The rbcL sequences and several non-molecular charac- data rather weakly supported a sister relation- ters 3Morgan et al. 1994; see below) support ship between that clade and Prunus, a rela- the inclusion of the actinorhizal clade in tionship supported also by their shared base the spiraeoid/amygdaloid/maloid group. We chromosome number 3x ˆ 8; Goldblatt 1976) therefore favor the topology shown in Fig. 1, and wood anatomy 3Zhang 1992). Because Lee in which the actinorhizal clade appears as the and Wen's 32001) ITS analysis included only ®rst branch in the spiraeoid/amygdaloid/ taxa traditionally classi®ed in Amygdaloideae maloid clade, but we acknowledge that this plus Exochorda and one ougroup 3Lyonotham- requires further study. In any case, our data nus), it did not provide a test of the monophyly strongly suggest that the actinorhizal group of the group including Prunus plus Oemleria/ diverged from the other lineages early in the Prinsepia/Exochorda. The results of our anal- evolution of the family. yses place the latter clade sister to Rhodotypos We did ®nd strong support for monophyly and Neviusia, with moderate support. The of a group including all taxa in the family position of Prunus varied among the four except Rosoideae sensu stricto and the actino- most parsimonious trees in our study, appear- rhizal clade, and for the sister position of ing either as sister to the clade including Lyonothamnus to all other taxa within that Maloideae sensu lato 3Fig. 1) or as sister to group. Neither of these relationships was Adenostoma/Chamaebatiaria/Sorbaria, but we recovered in the rbcL analysis. We concur found no support at all for a close relationship with Morgan et al. 31994) that Lyonothamnus, between Prunus and Oemleria/Prinsepia/ a monotypic genus found only on the Channel Exochorda. Islands of California, appears to be an isolated Evolution of particular characters. Soltis taxon within the family and its divergence near et al. 31995) suggested that the predisposition D. Potter et al.: Chloroplast DNA Phylogeny of Rosaceae 85 for nitrogen ®xation may have been present in has not been reported from any other family in the common ancestor of all members of the Rosales 3Zimmerman and Ziegler 1975). As Eurosid I 3Angiosperm Phylogeny Group explained above, we favor the position of the 1998) clade, although members of only ten actinorhizal clade illustrated in Fig. 1, which is extant families actually form these relation- consistent with the hypothesis that sorbitol ships. Four of these families 3Rosaceae, synthesis evolved once within the family Rhamnaceae, Elaeagnaceae, Celtidaceae) fall 3Fig. 1). within the clade now designated as Rosales Morgan et al. 31994) stated that their rbcL- 3Angiosperm Phylogeny Group 1998); in all of based phylogeny was consistent with achenes, these except Celtidaceae 3in which Parasponia follicles, or drupes having been the ancestral spp. are nodulated by Rhizobium spp.), nitro- fruit type for the family, with each of these gen ®xation occurs in root nodules as a result other types having evolved at least twice. In of a symbiosis involving the host and Fig. 1, it is most parsimonious to consider members of the actinomycete genus Frankia. either achenes or follicles as ancestral; an Our results strongly support the monophyly of additional step would be required if any other the actinorhizal Rosaceae. Although we in- fruit type were ancestral. This conclusion is cluded only three of the ®ve genera involved, based on the assumption that transitions this conclusion was also strongly supported in among all of these fruit types are equally phylogenetic analyses of rbcL sequences by likely. Regardless of which fruit type was Swensen and Mullin 31997), who included the ancestral, our trees suggest that the achenes other two 3Cowania and ). Thus, our of some taxa traditionally classi®ed in Rosoi- phylogenetic trees suggest that symbiotic ni- deae 3e.g., Adenostoma) evolved independently trogen ®xation evolved only once in Rosaceae 3perhaps as a reversal) of those in Rosoideae 3Fig. 1); however, given the lack of resolution, sensu stricto. Similarly, capsules and follicles discussed above, among the actinorhizal clade, each may have evolved two or more times, and Rosoideae sensu stricto, and the rest of the our results suggest that the drupes of Oemleria family, we cannot dismiss the possibility that and Prinsepia evolved independently of those the ability to form this symbiosis was present in Prunus. In contrast, the pome, the unique in the common ancestor of Rosaceae and lost fruit type that characterizes Maloideae sensu early in the evolution of the family. More stricto, appears to have evolved just once thorough studies of relationships among ac- within the family. Our data support the tinorhizal and non-actinorhizal members of hypothesis 3Morgan et al. 1994) that the pome the entire order Rosales should help resolve was derived from a spiraeoid follicle 3as in this question. Gillenia). Because there is considerable diver- As discussed above, Morgan et al. 31994) sity, among pomes, in the degree of carpel found weak support 3decay index 2), for connation and of adnation to the monophyly of a group including all taxa hypanthium 3Rohrer et al. 1991, 1994), deter- except Rosoideae sensu stricto; this group mination of the evolutionary sequence in was supported in two of the four most parsi- which carpel fusion and development of the monious trees in our analysis and received ¯eshy hypanthium occurred, and of whether or 58% bootstrap support. This group is distin- not one or both of those occurred indepen- guished not only by the higher base chromo- dently in di€erent lineages, require more thor- some numbers, but also by a number of ough sampling of Maloideae than is presented ecological and biochemical characters 3Mor- here. gan et al. 1994), such as the ability to accu- Origin of Maloideae. Our data are consis- mulate and transport the sugar alcohol tent with rbcL sequences in suggesting that sorbitol 3Zimmermann and Ziegler 1975, Wal- taxa with x ˆ 9 traditionally classi®ed in laart 1980). Accumulation of sugar alcohols Spiraeoideae are sister to Maloideae plus 86 D. Potter et al.: Chloroplast DNA Phylogeny of Rosaceae

Spiraeoideae with x ˆ 15 and 17. The partic- Lyonothamnoideae. On the other hand, he ular taxa involved are di€erent, however. maintained all of the following in Spiraeoi- Whereas Morgan et al. 31994) found that deae: Physocarpus, Neillia, Stephanandra the sister group to the Maloideae sensu lato 3together forming tribe Neillieae), Spiraea, was the clade including Sorbaria, Chamaeba- Aruncus 3both in tribe Spiraeeae), Holodiscus tiaria, and Adenostoma, our analysis strongly 3alone in tribe Holodisceae), Sorbaria, Cham- supported a sister relationship between Gille- aebatiaria 3tribe Sorbarieae), Adenostoma nia and Maloideae sensu lato, with Adenos- 3alone in tribe Adenostomateae) and Gillenia toma, Sorbaria, and Chamaebatiaria more 3in tribe Gillenieae). Both our results and those distantly related. Gillenia was not included of Morgan et al. 31994) suggest that this in the rbcL study, but the similarity in carpel circumscription of the subfamily, while less anatomy of members of that genus to Maloi- diverse than the traditional Spiraeoideae, is deae 3and, interestingly, also to Sorbaria) has still a polyphyletic assemblage. Our results also been noted 3Saunders 1927, Sterling 1966). A indicate that Maloideae could be further sister relationship between Gillenia and Ma- expanded to include Gillenia, a circumscription loideae sensu lato was also found by Evans that we support since it would maintain and Dickinson 31999c) in an analysis of monophyletic taxa without requiring classi®- combined molecular and non-molecular data. cation of Gillenia alone in its own subfamily. Since the results presented here, like those of Our sampling within Rosoideae sensu Morgan et al. 31994), are based on maternally stricto was not thorough enough to test Takh- inherited markers, they are not inconsistent tajan's 31997) taxonomic treatment of most with the hypothesis that the polyploidization of the genera in that group. As discussed event that gave rise to Maloideae sensu lato above, some of our trees supported a sister involved hybridization between two evolutio- relationship of a clade comprised of the narily distinct lineages. We can, however, actinorhizal Rosaceae to all members of conclude that Gillenia is the closest extant Rosoideae sensu stricto examined here 3one descendant of at least one of the lineages species each of Fallugia, Fragaria, Potentilla, involved in that event. Rosa, and Rubus). Our data, however, provide Figure 1 suggests a New World origin of no support for Takhtajan's 31997) classi®ca- Maloideae. Gillenia and Vauquelinia are both tion of the actinorhizal genera 3Cercocarpus, distributed within North America, while Kage- Chamaebatia, Cowania, Dryas, and Purshia) neckia comprises three Chilean species. The within Potentilloideae, since his circumscrip- two species of Lindleya, another genus tradi- tion of that subfamily includes Potentilla, tionally placed in Spiraeoideae with a base Fragaria, and Fallugia, but not Rubus and chromosome number of 17, which was not Rosa. included in this study but appeared sister to As discussed above, the positions of Prunus Vauquelinia plus Maloideae sensu stricto in the and other genera with x ˆ 8 have varied among rbcL analysis 3Morgan et al. 1994), are found analyses to date. Recognition of a taxonomic in Mexico. group including Oemleria, Exochorda, and Taxonomic Implications. Some of the Prinsepia, each of which is placed in its own clades resolved in this study and that of tribe by Takhtajan 31997), seems warranted. Morgan et al. 31994) are equivalent to taxo- Such a group has not been recognized in any nomic groups recognized in the classi®cation existing treatment but is strongly supported by of the family proposed by Takhtajan 31997). molecular data. Whether or not that group For example, he included Kageneckia, Vauque- should be united with Prunus 3as in Takhta- linia, and Lindleya in subfamily Pyroideae jan's Amygdaloideae) requires further investi- 3Maloideae), Kerria, Neviusia, and Rhodotypos gation, but it appears that there is little in Kerrioideae, and Lyonothamnus alone in support from molecular data for such a taxon. D. Potter et al.: Chloroplast DNA Phylogeny of Rosaceae 87

We reached a similar conclusion using ITS and support was provided by the Agricultural Experi- trnL-trnF data from a subset of the taxa ment Station of the University of California. included here 3Bortiri et al. 2001). The weak support for some relationships indicates that additional evidence is required References before ®nal decisions can be made about Angiosperm Phylogeny Group 31998) An ordinal higher-level classi®cation of the family. Some, classi®cation for the families of ¯owering . but not all, of the subfamilies and/or tribes Ann. Missouri Bot. Gard. 85: 531±553. recognized by Takhtajan 31997) appear to Bortiri E., Oh S., Jiang J., Baggett S., Granger A., represent monophyletic groups, based on Weeks C., Buckingham M., Potter D., Par®tt D. both the current study and results from 32001) Phylogeny and systematics of the genus analysis of rbcL sequences 3Morgan et al. Prunus L. 3Rosaceae) as determined by sequence 1994). The issue of how many subfamilial analysis of ITS and the chloroplast trnL-trnF ranks should be recognized will be in¯uenced spacer DNA. Syst. Bot. 26: 797±807. by the degree to which resolution among Boyer S. K., Mullet J. E. 31988) Pea chloroplast tRNA-Lys 3UUU) gene: transcription and anal- strongly supported clades can be improved in ysis of an intron-containing gene. Photosyn. Res. future analyses. The branching order near the 17: 7±22. base of our trees was, in general, more Campbell C. S., Donoghue M. J., Baldwin B. G., strongly supported than in the rbcL study, Wojciechowski M. F. 31995) Phylogenetic rela- but the relationships among several of the tionships in Maloideae 3Rosaceae): evidence well-supported clades remain poorly resolved. from sequences of the internal transcribed spac- The low support values on the internal ers of nuclear ribosomal DNA and its congru- branches are due to a combination of ho- ence with morphology. Amer. J. Bot. 82: moplasy in the data, which might be attrib- 903±918. utable to a complex history perhaps involving Chase M. W., Soltis D. E., Olmstead R. G., reticulation among lineages, and relatively Morgan D., Les D. H., Mishler B. D., Duvall short branch lengths, which may re¯ect rapid M. R., Price R. A., Hillis H. G., Qiu Y.-L., Kron K. A., Rettig J. H., Conti E., Palmer J. D., radiation of major lineages within the family. Manhart J. R., Sytsma K. J., Michaels H. J., If resolution remains poor, it may be most Kress W. J., Karol K. G., Clark W. D., Hedren appropriate to recognize only one supergen- M., Gaut B. S., Jansen R. K., Kim K.-J., eric rank 3e.g., tribes, as in Hutchinson 31964)), Wimpee C. F., Smith J. F., Furnier G. R., which would best re¯ect the fact that there are a Strauss S. H., Xiang Q.-Y., Plunkett G. M., number of well-supported groups but the rela- Soltis P. S., Swensen S., Williams S. E., Gadek P. tionships among them are not well understood. A., Quinn C. J., Eguiarte L. E., Golenberg E., On the other hand, if strong resolution among Learn G. H., Jr., Graham S. W., Barrett S. C., clades is eventually obtained by integrating Dayanandan S., Albert V. A. 31993) Phyloge- multiple data sets, this resolution should be netics of plants: an analysis of nucleotide incorporated into the hierarchical structure of sequences from the gene rbcL. Ann. the of the family. Missouri Bot. Gard. 80: 528±580. Dickson E. E., Arumuganathan K., Kresovich S., Technical assistance from Patrick Cheung, Doyle J. J. 31992) Nuclear DNA content varia- Dean Lavelle, Millie Lee, Teri Nguyen, and Mindy tion within the Rosaceae. Amer. J. Bot. 79: Sin is gratefully acknowledged. The following 1081±1086. individuals assisted us in obtaining plant material: Doyle J. J., Doyle J. L. 31987) A rapid DNA S. Arulsekar, Phillip Ashby, Wei Cao, Holly isolation procedure for small quantities of fresh Forbes, Mary Hirshfeld, Jim Luby, and Warren leaf tissue. Phytochem. Bull. 19: 11±15. Roberts. Kenneth Sytsma, Kenneth Robertson, Eriksson T., Donoghue M. J., Vretblad M. 31998) and one anonymous reviewer provided many Phylogenetic analysis of Potentilla using DNA helpful comments on the manuscript. Financial sequences of nuclear ribosomal internal tran- 88 D. Potter et al.: Chloroplast DNA Phylogeny of Rosaceae

scribed spacers 3ITS), implications for the clas- the tribe Millettieae 3Leguminosae) based on si®caiton of Rosoideae 3Rosaceae). Plant. Syst. chloroplast trnK/matK sequences and its impli- Evol. 211: 155±179. cations for evolutionary patterns in Papilionoi- Evans R. C., Alice L. A., Campbell C. S., Kellogg deae. Amer. J. Bot. 87: 418±430. E. A., Dickinson T. A. 32000) The granule- Hutchinson J. 31964) The genera of ¯owering bound starch synthase 3GBSSI) gene in the plants, vol. 1. Clarendon Press, Oxford, U.K. Rosaceae: multiple loci and phylogenetic utility. Kalkman C. 31988) The phylogeny of the Rosa- Mol. Phylogenet. Evol. 17: 388±400. ceae. Bot. J. Linn. Soc. 98: 37±59. Evans R. C., Campbell C. S. 32000) [abstract] A KaÈ llersjoÈ M., Farris J. S., Chase M. W., Bremer B., multigene tree of the Rosales. Amer. J. Bot. 87 Fay M. F., Humphries C. J., Petersen G., Sebert 36 Suppl.): 125±126. O., Bremer K. 31998) Simultaneous parsimony Evans R. C., Dickinson T. A. 31997) [abstract] jackknife analysis of 2358 rbcL DNA sequences Phylogenetic relationships in the Rosaceae: a reveals support for major clades of green plants, total evidence approach. Amer. J. Bot. 84 land plants, seed plants and ¯owering plants. 36 suppl.): 190. Plant. Syst. Evol. 213: 259±287. Evans R. C., Dickinson T. A. 31999a) Floral Karol K. G., Suh Y., Schatz G. E., Zimmer E. A. ontogeny and morphology in subfamily Amyg- 32000) Molecular evidence for the phylogenetic daloideae T & G 3Rosaceae). Int. J. Pl. Sci. 160: position of Takhtajania in the Winteraceae: 955±979. inference from nuclear ribosomal and chloro- Evans R. C., Dickinson T. A. 31999b) Floral plast gene spacer sequences. Ann. Missouri Bot. ontogeny and morphology in subfamily Spirae- Gard. 87: 414±432. oideae Endl. 3Rosaceae). Int. J. Pl. Sci. 160: 981± Lee, S., Wen. J. 32001) A phylogenetic analysis of 1012. Prunus and the Amygdaloideae 3Rosaceae) using Evans R. C., Dickinson T. A. 31999c) [abstract] ITS sequences of nuclear ribosomal DNA. Rosaceae ¯oral evolution: hypothesis from anal- Amer. J. Bot. 88: 150±160. ysis of combined molecular and non-molecular Mishler B. D., Donoghue M. J., Albert V. A. data. Abstracts from the XVI International 31991) [abstract] The decay index as a measure of Botanical Congress, August 1±7, 1999, St. Louis, relative robustness within a cladogram. Tenth MO, p. 39. Meeting of the Willi Hennig Society, Toronto. Farris J. S., KaÈ llersjoÈ M., Kluge A. G., Bult C. Morgan D. R., Soltis D. E., Robertson K. R. 31994) Testing signi®cance of incongruence. 31994) Systematic and evolutionary implications Cladistics 10: 315±319. of rbcL sequence variation in Rosaceae. Amer. J. Felsenstein J. 31985) Con®dence limits on phylog- Bot. 81: 890±903. enies: an approach using the bootstrap. Evolu- Phipps J. B., Robertson K. R., Rohrer J. R., Smith tion 39: 783±791. P. G. 31991) Origins and evolution of subfam. Gielly L., Taberlet P. 31996) A phylogeny of the Maloideae 3Rosaceae). Syst. Bot. 16: 303±332. European gentians inferred from chloroplast Potter D. 31997) [abstract] Variation and phyloge- trnL 3UAA) intron sequences. Bot. J. Linn. netic relationships of nucleotide sequences for Soc. 120: 57±75. putative genes encoding polygalacturonase in- Gladkova V. N. 31972) On the origin of subfamily hibitor protein 3PGIPs) among taxa of Rosa- Maloideae. Bot. Zhurn. 57: 42±49. ceae. Amer. J. Bot. 84 36 Suppl.): 224. Goldblatt P. 31976) Cytotaxonomic studies in the Potter D., Gao F., Oh S., Baggett S. 31999) tribe Quillajeae 3Rosaceae). Ann. Missouri Bot. [abstract] Molecular phylogenetic studies in Gard. 63: 200±206. Rosaceae. Abstracts from the XVI International Hebda R. J., Chinnappa C. C. 31994) Studies on Botanical Congress, August 1±7, 1999, St. Louis, pollen morphology of Rosaceae. Acta Bot. MO, p. 39. Gallica 141: 183±193. Potter D., Luby J. J., Harrison R. E. 32000) Heywood V. H. 31993) Flowering plants of the Phylogenetic relationships among species of world. Oxford University Press, New York. Fragaria 3Rosaceae) inferred from non-coding Hu J.-M., Lavin M., Wojciechowski M. F., San- nuclear and chloroplast DNA sequences. Syst. derson M. J. 32000) Phylogenetic systematics of Bot. 25: 337±348. D. Potter et al.: Chloroplast DNA Phylogeny of Rosaceae 89

Robertson K. R. 31974) The genera of the Rosa- Swensen S. M., Mullin B. C. 31997) The impact of ceae in the southeastern . J. Arnold molecular systematics on hypotheses for the Arbor. 55: 303±332, 344±401, 611±662. evolution of symbioses and impli- Rohrer J. R., Robertson K. R., Phipps J. B. 31991) cations for expanding symbioses to new host Variations in structure among fruits of Maloi- plant genera. Pl. & Soil 194: 185±192. deae 3Rosaceae). Amer. J. Bot. 78: 1617±1635. Swo€ord D. L. 32000) ``PAUP*. Phylogenetic Rohrer J. R., Robertson K. R., Phipps J. B. 31994) Analysis Using Parsimony 3* and Other Meth- Floral morphology of Maloideae 3Rosaceae) and ods),'' Version 4. Sinauer Associates, Sunder- its systematic relevance. Amer. J. Bot. 81: 574± land, MA. 581. Taberlet P., Gielly L., Pautou G., Bouvet J. 31991) Saunders E. R. 31927) On carpel polymorphism. II. Universal primers for ampli®cation of three non- Ann. Bot. 41: 569±627. coding regions of chloroplast DNA. Pl. Mol. Sax K. 31933) The origin of the Pomoideae. Proc. Biol. 17: 1105±1109. Amer. Soc. Hort. Sci. 30: 147±150. Takhtajan A. 31997) Diversity and classi®cation of Schulze-Menz G. K. 31964) Rosaceae. In: Melchior ¯owering plants. Columbia Univ. Press, New H. 3ed.) Engler's Syllabus der P¯anzenfamilien York. II. 12th edn. GebruÈ der Borntraeger, Berlin, Thompson J. D., Gibson T. J., Plewniak F., pp. 209±218. Jeanmougin F., Higgins D. G. 31997) The Scott K. D., McIntyre C. L., Playford J. 32000) ClustalX windows interface: ¯exible strategies Molecular analyses suggest a need for a signif- for multiple sequence alignment aided by quality icant rearrangement of Rutaceae subfamilies and analysis tools. Nucl. Acids Res. 25: 4876±4882. a minor reassessment of species relationships Wallaart R. A. M. 31980) Distribution of sorbitol within Flindersia. Plant. Syst. Evol. 223: 15±27. in Rosaceae. Phytochemistry 19: 2603±2610. Soltis D. E., Soltis P. S., Morgan D. R., Swensen S. Xiang Q.-Y., Soltis D. E., Soltis P. S. 31998) M., Mullin B. C., Dowd J. M., Martin P. G. Phylogenetic relationships of Cornaceae and 31995) Chloroplast gene sequence data suggest a close relatives inferred from matK and rbcL single origin of the predisposition for symbiotic sequences. Amer. J. Bot. 85: 285±297. nitrogen ®xation in angiosperms. Proc. Natl. Zhang S.-Y. 31992) Systematic wood anatomy of Acad. Sci. USA 92: 2647±2651. the Rosaceae. Blumea 37: 81±158. Soltis D. E., Soltis P. S., Chase M. W., Mort M. E., Zimmermann M. H., Ziegler H. 31975) List of Albach D. C., Zanis M., Savolainen V., Hahn sugars and sugar alcohols in sieve-tube exudates. W. H., Hoot S. B., Fay M. F., Axtell M., In: Zimmermann, M. H., Milburn, J. A. 3eds.) Swensen S. M., Prince L. M., Kress W. J., Nixon Transport in plants. I. phloem transport. Spring- K. C., Farris J. S. 32000) Angiosperm phylogeny er, Heidelberg, pp. 480±502. inferred from 18S rDNA, rbcL, and atpB sequences. Bot. J. Linn. Soc. 133: 381±461. Sterling C. 31966) Comparative morphology of the Addresses of the authors: Daniel Potter 3e-mail: carpel in the Rosaceae. VIII. Spiraeoideae: Quil- [email protected]), Fangyou Gao, Pedro Este- lajeae, Sorbarieae. Amer. J. Bot. 53: 951±960. ban Bortiri, Sang-Hun Oh, Department of Pomol- Sugita M., Shinozaki K., Sugura M. 31985) Tobac- ogy, 1045 Wickson Hall, University of California, co chloroplast tRNALys 3UUU) gene contains a One Shields Avenue, Davis, California 95616, 2.5-kilobase-pair intron: an open reading frame USA. Scott Baggett, Ph.D. Program Biology, and a conserved boundary sequence in the Lehman College, City University of New York, intron. Proc. Natl. Acad. Sci. U.S.A. 82: 3557± 250 Bedford Park Boulevard West, Bronx, New 3561. York 10458, USA.