Kelly P. Steele 2,6 , Stefanie M. Ickert-Bond 3,5 , Shahin Zarre 4 , and Martin F. Wojciechowski 5

American Journal of Botany 97(7): 1142–1155. 2010.

P HYLOGENY AND CHARACTER EVOLUTION IN MEDICAGO (LEGUMINOSAE): EVIDENCE FROM ANALYSES OF PLASTID TRN K/MATK AND NUCLEAR GA3OX1 SEQUENCES 1

Kelly P. Steele 2,6 , Stefanie M. Ickert-Bond3,5 , Shahin Zarre4 , and Martin F. Wojciechowski5

2 Arizona State University, Department of Applied Sciences and Mathematics, Polytechnic Campus, 6098 Backus Mall, Mesa, Arizona 85212 USA; 3 University of Alaska, Department of Biology and Wildlife, Institute of Arctic Biology and University of Alaska Museum of the North Herbarium, 907 Yukon Drive, Fairbanks, Alaska 99775-6960 USA; 4 Tehran University, Department of Biology, Tehran, Iran; and 5 Arizona State University, School of Life Sciences, Tempe, Arizona 85287-4501 USA

• Premise of the study: The genus Medicago, with about 87 species, includes the model legume species M . truncatula, and a number of important forage species such as M . sativa (alfalfa), M . scutellata (snail medic), and M . lupulina (black medic). Relationships within the genus are not yet suffi ciently resolved, contributing to diffi culty in understanding the evolution of a number of distinguishing characteristics such as aneuploidy and polyploidy, life history, structure of cotyledons, and number of seeds per fruit. • Methods : Phylogenetic relationships of 70 – 73 species of Medicago and its sister genus Trigonella (including Melilotus ) were reconstructed from nucleotide sequences of the plastid trnK/matK region and the nuclear-encoded GA3ox1 gene (gibberellin 3- β -hydroxylase) using maximum parsimony and Bayesian inference methods. • Key results: Our results support certain currently recognized taxonomic groups, e.g., sect. Medicago (with M . sativa) and sect. Buceras . However, other strongly supported clades — the “ reduced subsection Leptospireae clade” that includes M . lupulina , the “ polymorpha clade” that includes M . murex and M . polymorpha and the “ subsection Pachyspireae clade” that includes M . truncatula— each of which includes species presently in different subsections of sect. Spirocarpos , contradict the current classifi cation. • Conclusions : These results support the hypothesis that some characters considered important in existing taxonomies, for example, single-seeded fruits that have arisen more than once in both Medicago and Trigonella , are indeed homoplastic. Others, such as the 2n = 14 chromosome number, have also arisen independently within the genus. In addition, we demonstrate support for the utility of GA3ox1 sequences for phylogenetic analysis among and within closely related genera of legumes.

Key words: aneuploidy; GA3ox1 ; Leguminosae; Medicago ; Trigonella ; trnK/matK ; single-seeded fruits.

Medicago L., a genus of approximately 87 species of herbs not accepted by most taxonomists. Instead, most authors recog- and shrubs widespread from the Mediterranean to central Asia nized the tribe Trifolieae, which included these three genera (Small and Jomphe, 1989b; Lewis et al., 2005; Small, 2010), and Trifolium L. (e.g., Rechinger, 1984 ). Some authors have includes the widely cultivated major forage crop and weedy also included Ononis L. and Parochetus Buch.-Ham. ex D. Don species M. sativa L. (alfalfa, lucerne) and the legume model (e.g., Heyn, 1981 ). More recently, Trigonellinae were recog- species M. truncatula Gaertn. (Cannon et al., 2006). Taxonomi- nized as a subtribe of Trifolieae by Small (1987b), who noted cally, Medicago along with Melilotus Mill. (sweetclovers) and that the species in this group share morphological character Trigonella L. were included in the tribe Trigonellinae, fi rst rec- states including leaves that are digitately trifoliate with stipules ognized by Schultz (1901), but as circumscribed this tribe was adnate to the stem, but not encircling it entirely. Delimitation of the three genera in the subtribe has been problematic, particularly between Medicago and Trigonella. Nevertheless, using 1 Manuscript received 6 January 2010; revision accepted 14 May 2010. The authors especially thank E. Small (DAO) for seeds, literature, and several fl oral features associated with the explosive pollination encouragement in early stages of the research and the curators of the syndrome, Small et al. (1987 ) transferred 23 species of Trigo- following herbaria for specimen loans and leaf material of species that nella (the so-called “ medicagoid Trigonella ” ) to the genus were used in this study: Arizona State University (ASU), Royal Botanic Medicago, which currently comprises sections Buceras and Gardens at Kew (K), and Tehran University (TUH). This work was supported Lunatae. The three genera of Trigonellinae can be further dis- by grants from Arizona State University and National Science Foundation tinguished using biochemical characteristics, specifi cally the grants DEB-9707571 and DEB-0041311 to K.P.S. and DEB-0542958 to type of phytoalexins produced after fungal infection. While M.F.W. Fieldwork in Iran was supported by grant 7452-03 from the species of Medicago accumulate vesitol and sativan, these sub- National Geographic Society Foundation. The authors acknowledge the stances are absent from Trigonella and Melilotus ( Ingham and signifi cant contributions of D. Chin, M. Sabir, B. Udpa, and L. Yang, Harborne, 1976). Furthermore, species of Medicago contain students at California State University, East Bay, to the laboratory work. The authors also thank M. Lavin, M. Simmons, and an anonymous reviewer hemolytic saponins, which are not found in Trigonella or for many helpful comments on the manuscript. Melilotus ( Jurzysta et al., 1988 ). Medicago and Trigonella , as 6 Author for correspondence (e-mail: [email protected]) delimited by Small and Jomphe (1989b) , have always been strongly supported as sister genera based upon analyses of both doi:10.3732/ajb.1000009 the nuclear ribosomal internal transcribed spacer region (nrDNA

ITS) and the fl anking external transcribed spacer region (nrDNA molecular phylogenetic analyses of Medicago was not an eval- ETS) (Bena, 2001), as well as the plastid-encoded matK gene uation of morphology, chromosome number, biochemical char- ( Steele and Wojciechowski, 2003 ; Wojciechowski et al., 2004 ). acters, nor taxonomic revision (Downie et al., 1998; Bena, Thus the monophyly of Medicago as delimited by Small et al. 2001; Maureira-Butler et al., 2008). Thus, one of our goals was (1987 ) is not in question. to summarize their fi ndings and to use results from all mo- Interspecifi c relationships within Medicago are still in need lecular markers analyzed to date to carefully consider the evo- of study although they have been investigated. Small (1987b) lution of certain morphological characters traditionally presented a tentative phylogeny of sections of Medicago using considered taxonomically important, including numbers of just six morphological characters that were previously used to seeds per pod, variation in cotyledon pulvini, chromosome circumscribe the genus. Phylogenetic analyses utilizing nucle- number, and growth forms. It is particularly useful to take into otide sequence data have begun to provide some resolution. account the results derived from all of these studies, because Sixty-four species of Medicago were sampled using nrDNA none of the previous studies sampled all of the principal lin- ITS sequences in a study by Downie et al. (1998), while 61 spe- eages within Medicago ; although a total of 75 species were cies were sampled by Bena (2001), who used both nrDNA ITS represented in these previous studies, the largest number of and ETS sequences. More recently, Maureira-Butler et al. species sampled by any one study was 61. In several cases, (2008) sampled 56 species of Medicago for the nuclear-encoded consideration of all results allows conclusions to be drawn genes CNCG5 and β cop and the mitochondrial gene rpS14-cob . that could not be drawn on the basis of results from just one or Several species groups were well supported in phylogenetic re- two molecular markers. constructions in these studies, but numerous polytomies were also present. Thus, a number of questions regarding relationships among Medicago species remain to be resolved. MATERIALS AND METHODS Maureira-Butler et al. (2008) found incongruency among results based on analyses of these two nuclear-encoded genes and Taxon sampling— For the genus Medicago , a total of 49 species (of 87 in to a lesser extent, the mitochondrial gene, and suggested that genus) were sampled for this study and include representatives from all sections this incongruency casts some doubt on the use of phylogenetic and from nearly all subsections; only section Buceras subsections Defl exae and trees to estimate sister group relationships in Medicago or to Isthmocarpae were not sampled (infrageneric delimitation following Small and investigate character evolution. While that study focused on Jomphe, 1989b ) ( Table 1 ) : seeds were not available from the U. S. Department of Agriculture (USDA) for M . retrosa (monotypic subsection Defl exae ), native possible explanations of incongruence and concluded that their to Afghanistan, nor for the two species M . isthmocarpa and M . rhytidocarpa data favored a hypothesis of past hybridization, rather than in- (subsection Isthomocarpae ), each of which has a limited distribution in Turkey. complete linkage assortment, it is nevertheless important to Multiple accessions were sampled for three taxa, M . italica (2), M . lupulina (2), note that these authors also found nine well-supported clades and M . monantha (3). In addition, we sampled 17 species of Trigonella (55 common to both nuclear genes. Based on a consideration of the species in genus) representing eight of 12 sections, and up to four species from results of the previous study of Maureira-Butler et al. (2008) each of the two subgenera of Melilotus (20 species in genus) (Table 1), from subtribe Trigonellinae. For a few species, sequences were obtained from Gen- and those of Bena (2001) with new results presented in this Bank (http://www.ncbi.nlm.nih.gov/Genbank) ( Table 1 ). We performed phylo- paper, here we discuss clades with a similar species composi- genetic analyses on two sets of data; trnK/matK sequences and GA3ox1 tion that have been identifi ed by analyses of fi ve different mo- sequences. For analyses of trnK/matK sequences, Pisum sativum , Ononis bi- lecular markers, representing both the plastid and nuclear fl ora Desf., and O . natrix L. were used as outgroup taxa based on results of our genomes. In spite of incongruency among gene trees, the pres- earlier work ( Wojciechowski et al., 2000 ; Steele and Wojciechowski, 2003 ). ence of similar, strongly supported groups in results of analyses Initial analyses of GA3ox1 sequences used an exon-only data set (introns are diffi cult to align outside closely related genera) from 92 species of Trifolieae of most, if not all, of these markers strongly suggests an under- and Fabeae that corroborated the monophyly of subtribe Trigonellinae and the lying biological reality that we believe is useful to consider. sister group relationship of Medicago and Trigonella (Appendix S1, see Sup- Increased resolution of relationships within Medicago will plemental Data with the online version of this article). Analyses of GA3ox1 allow greater understanding of the evolution of morphological, sequences from Medicago , Trigonella , and Melilotus included both exon and molecular, and biochemical characters in this genus. Medicago intron data, and Trigonella plus Melilotus were designated as the outgroup to has been one of the most widely studied legumes because of Medicago based on results from analyses of the larger exon-only data set. numerous agriculturally important and domesticated species ( Small, 2010 ). In addition to alfalfa, which is the most widely Molecular methods— Fresh leaf material was obtained from plants col- lected in California by the fi rst author or from plants grown from seeds obtained cultivated legume and the most important forage crop in the from E. Small or the USDA Plant Introduction Program (http://www.ars-grin. world; estimated world acreage of alfalfa is 32 million ha (Mi- gov/npgs/). Total genomic DNA from these taxa was isolated following Doyle chaud et al., 1988 ), other species are used as medicine, human and Doyle (1987) , while genomic DNA was isolated from samples of herbar- food (honey, sprouts), green manure, sources of industrial en- ium specimens following methods described by Wojciechowski et al. (2004) . zymes in biotechnology (Lewis et al., 2005), model genomic We used two molecular markers, the plastid-encoded trnK intron- matK gene species ( Cannon et al., 2006 ), and model systems for the study region, widely used in phylogenetic analyses of legumes (Hu et al., 2000; Steele and Wojciechowski, 2003; Wojciechowski et al., 2004; Bruneau et al., 2008) of nitrogen fi xation (e.g., Bailly et al., 2007 ). Lesins and Lesins and the novel nuclear-encoded protein-coding gene GA3ox1 (Steele et al., (1979) and numerous papers by Small and colleagues (summa- 1999 ). The mat K gene consists of 1500 – 1525 bp in most species, while the 5 ′ rized in Small and Jomphe, 1989b ) discussed a variety of char- and 3 ′ trnK intron regions that fl ank the mat K gene add approximately another acters such as chromosome number, presence of woody tissue, 1000 bp of sequence ( Steele and Wojciechowski, 2003 ). Sequences of primers and cotyledon structure that support recognition of infrageneric used were as described in Steele and Wojciechowski (2003) (Table 2). In addi- taxa and the delimitation of species within the genus Medicago . tion, we also used both coding and noncoding regions (approximately 1718 bp) of the single-copy, nuclear-encoded gene GA3ox1 , that codes for gibberellin The results of phylogenetic analyses based on nucleotide se- 3- β -hydroxylase. In Pisum , this gene has been referred to as “ Mendel ’ s stem quence data that we present here allow us to consider hypothe- length gene ” (Le ) after having been independently discovered to be the basis of ses of evolution of these and other features and to consider their the normal height vs. dwarf peas originally studied by Mendel, by researchers taxonomic implications as well. The primary goal of previous in the United States ( Martin et al., 1997 ) and New Zealand ( Lester et al., 1997 ). 1144 American Journal of Botany [Vol. 97

Table 1. Taxa sampled, source of DNA, chromosome number of taxon from the literature, and GenBank accessions.

GenBank accessiond Present taxonomya Sourceb 2 n c trnK/matK GA3ox1

Medicago Section Heynianae M. heyniana Greuter E. Small M839 (=PI 537135); Greece (KPS 47) 16 AF522093 HM211113 Section Orbiculares M. orbicularis (L.) Bartal. E. Small M316; BG (KPS 28) 16 AF522101 HM211130 Section Hymenocarpos M. radiata L. PI 459140; Turkey (KPS 87) N/A AF522106 HM211138 Section Platycarpae M. platycarpa (L.) Trautv. PI 257499; Siberia (KPS 74) 16 AF522102 HM211133 M. ruthenica (L.) Ledebour E. Small T263; BG (KPS 56) N/A AF522107 HM211140 Section Buceras Subsection Erectae M. fi scheriana (Ser.) Trautv. PI 464826; Turkey (KPS 88) N/A HM159562 HM211111 M. medicaginoides (Retz.) E. Small E. Small T264; Uzbekistan (KPS 66) 16 AF522097 HM211124 M. monantha (C. A. Meyer) Trautv. E. Small T251; Uzbekistan (KPS 36) 44 HM159572 N/A M. monantha (C. A. Meyer) Trautv. TUH 34031; Iran (KPS 183) HM159573 HM211127 M. monantha (C. A. Meyer) Trautv. PI 352711; Turkey (KPS 35) AF522098 HM211126 M. polyceratia (L.) Trautv. E. Small T254; ZGK (KPS 44) 28, 44 AF522103 HM211134 Subsection Refl exae M. monspeliaca (L.) Trautv. E. Small T252; BG (KPS 60) 16 AF522099 N/A Section Lunatae M. brachycarpa M. Bieb. PI 244326; Spain BG (KPS 72) N/A AF522092 HM211106 M. bifl ora (Griesb.) E. Small PI 464827; Turkey (KPS 70) N/A AF522091 HM211104 Section Spirocarpos Subsection Leptospireae M. arabica (L.) Huds. PI 233253; Israel (KPS 79) 16 HM159554 HM211102 M. lanigera Winkl. & Fedtsch. E. Small 1579; NE Afghanistan (KPS 30) 16 AF522096 HM211118 M. minima (L.) Bart. W6 8305; Uzbekistan (KPS 83) 16 HM159571 HM211125 M. polymorpha L. KPS 2; USA 14 AF522104 HM211135 M. praecox DC PI 495434; France (KPS 80) 14 HM159577 HM211136 M. tenoreana Ser. PI 499156; Italy (KPS 97) 16 HM159589 HM211150 Subsection Rotatae M. bonarotiana Arcang. PI 495222; Lebanon (KPS 77) 16 HM159556 HM211105 M. noeana Boiss. PI 495404; Turkey (KPS 90) 16 AF522100 HM211129 M. scutellata (L.) Miller PI 505433; Spain (KPS 124) 30 HM159584 N/A M. shepardii Post PI 495592; Turkey (KPS 107) 16 HM159586 HM211147 Subsection Intertextae M. ciliaris (L.) Krocker PI 535598; Tunisia (KPS 119) 16 HM159559 HM211108 M. granatensis Willd. PI 498810; Israel (KPS 85) 16 HM159563 HM211112 M. intertexta (L.) Miller PI 498826; United Kingdom BG (KPS 73) 16 HM159565 HM211116 Subsection Pachyspireae M. constricta Durieu PI 516644; Morocco (KPS 86) 14 HM159560 HM211109 M. italica (Mill.) Fiori PI 385014; Tunisia (KPS 78) 16 HM159566 N/A M. italica W6 5954; Malta (KPS 133) 16 AF522095 HM211117 M. littoralis Rohde E. Small 111; Israel (KPS 45) 16 HM159567 HM211119 M. littoralis E. Small 64; Israel (KPS 63) HM159568 HM211120 M. murex Willd. PI 535624; Tunisia (KPS 84) 14 HM159574 HM211128 M. rigidula (L.) All PI 459173; Turkey (KPS 132) 14 HM159579 HM211139 M. soleirolii Duby PI 537240; Algeria (KPS 92) 16 HM159587 HM211148 M. turbinata (L.) All PI 577606; Lebanon (KPS 91) 16 HM159590 HM211151 M. truncatula Gaertn. Australia (cultivar: Jemalong) 16 NC_003119.6 AC208096.4 Section Lupularia M. lupulina L. KPS 4; USA 16 HM159569 HM211121 M. lupulina L. TUH 34031; Iran (KPS 185) 16 N/A HM211122 M. secundifl ora Durieu PI 537238; France (KPS 118) 16 HM159585 HM211146 Section Dendrotelis M. arborea L. PI 504540; Aegean Islands, Greece (KPS 81) 32, 48 HM159555 HM211103 Section Medicago M. cancellata M. Bieb PI 440493; Russia (KPS 76) 48 HM159557 HM211107 M. daghestanica Rupr. #22590 DNABank (K); Russia 16 HM159561 HM211110 M. hybrida (Pourr.) Trautv. PI 538998; Russian Federation (KPS 134) 16 HM159564 HM211114 M. marina L. A. Della 2252 (#22591 DNABank) (K); Cyprus 16 HM159570 HM211123 M. papillosa Boiss. PI 464699; Turkey (KPS 99) 32 HM159575 HM211131 M. pironae Vis. PI 253450; Slovenia (KPS 112) 16 HM159576 HM211132 M. prostrata Jacq. PI 577453; Greece (KPS 120) 16, 32 AF522105 HM211137 M. rhodopea Velen. Kojoucharov & Kouzmanov s.n. (K); Bulgaria 16 HM159578 N/A M. saxatilis M. Bieb. W6 5898; France (KPS 102) 48 HM159583 HM211145 M. sativa L. subsp. sativa KPS 14; USA 16, 32 AF522108 HM211141 July 2010] Steele et al. — Phylogenetic relationships in MEDICAGO 1145

Table 1. Continued.

GenBank accessiond Present taxonomya Sourceb 2 n c trnK/matK GA3ox1 M. sativa × varia (Martyn) Arcangeli PI 346899; Georgia (KPS 106) 16, 32 HM159582 HM211144 M. sativa subsp. caerulea (Less. ex Ledeb.) Schmalh. PI 464721; Turkey (KPS 135) 16 HM159580 HM211142 M. sativa subsp. caerulea unknown N/A EF121426.1 M. sativa subsp. falcata (L.) Arcangeli PI 631619; Nepal (KPS 130) 16, 32 HM159581 HM211143 M. sativa subsp. falcata unknown N/A EF121424.1 M. suffruticosa subsp. suffruticosa Raymond ex DC PI 632023; Spain (KPS 105) 16 HM159588 HM211149 Section Geocarpae M. hypogaea E. Small E. Small F177; Israel (KPS 27) 14 AF522094 HM211115 Section Cartiensae M. carstiensis Wulf. PI 641414; Russian Federation 16 HM159558 N/A Melilotus Subgenus Micromelilotus M. indicus (L.) All. M F Wojciechowski 540 (ARIZ); USA 16 AF522111 HM211153 M. segetalis Ser. F. Sales & I. Hedge 94/7 (E); Portugal 16 HM159591 N/A M. sulcatus Desf. J. R. Edmondson & M. A. S. McClintock 2897 (E); 16 HM159592 N/A Cyprus Subgenus Melilotus M. albus Medik. KPS 21; USA 16 AF522110 HM211152 Ononis O. bifl ora Desf. PI 244319 (KPS 41) 32 AF522113 N/A O. natrix L. PI 246743 (KPS 17) 32 AF522114 N/A Pisum P. sativum L. M F Wojciechowski 1014 (ASU) 14 AY386961 N/A Trigonella incertae sedis T. bicolor Boiss. & Balansa E. Small L309; Turkey (KPS 19) N/A AF522141 HM211157 Section Ellipticae T. elliptica Boiss. TUH 33808; Iran (KPS 170) N/A HM159594 HM211163 T. species novum TUH 33879; Iran (KPS 179) N/A HM159596 HM211169 Section Falcatulae T. anguina Delile PI 517185; Morocco (KPS 5) N/A HM159593 HM211154 T. balansae Boiss & Reut. PI 222211; Afghanistan (KPS 6) 16 AF522140 HM211156 T. corniculata L. PI 220123; Afghanistan (KPS 31) 16 AF522145 HM211161 T. hamosa L. E. Small T273; Saudi Arabia (KPS 65) 16 HM159595 HM211166 Section Callicerates T. calliceras Fischer ex M. Bieb E. Small T239; BG (KPS 40) 16 AF522142 HM211158 Section Cylindricae T. kotschyi Fenzl ex Boiss. PI 206775; Turkey (KPS 64) N/A AF522149 HM211167 T. spruneriana (Boiss.) Ponert PI 352710; Turkey (KPS 8) N/A AF522151 HM211170 Section Samaroideae T. cretica (L.) Boiss PI 415833; Switzerland (KPS 18) 16 AF522146 HM211162 Section Capitatae T. caerulea (L.) Ser. E. Small T270; United Kingdom BG (KPS 48) N/A AF522143 HM211159 Section Pectinateae T. arabica Delile PI 194476; Israel (KPS 69) 2n-16 AF522139 HM211155 Section Foenum-graecum T. coerulescens Halacsy PI 314398; Uzbekistan (KPS 7) 16 AF522144 HM211160 T. foenum-graecum L. PI 567879; Turkey (KPS 16) 16 AF522147 HM211164 T. gladiata L. PI 203474; Turkey (KPS 42) 16 AF522148 HM211165 T. macrorrhyncha Boiss. PI 222232; Iran (KPS 43) 16 AF522150 HM211168 a Medicago: follows Small and Jomphe, 1989b; Melilotus : follows Stevenson, 1969; Trigonella : follows Huber and Morath, 1970; Rechinger, 1984; Zohary, 1972 b E. Small collections, seeds donated to fi rst author; USDA accessions: PI or W6; K: DNA obtained from specimens at Royal Botanic Garden, Kew, UK; ZGK: Zentralinstitut für Genetik und Kulturpfl anzenforschung der Akademie der Wissenschaften der DDR; KPS: collections of the fi rst author; BG: seeds sent to E. Small from various botanic gardens, but collection locations for seeds unknown. c Medicago , Melilotus , and Trigonella chromosome numbers from Lesins and Lesins, 1979; Small and Brooks, 1984; Index to Plant Chromosome Numbers (IPCN); Dundas et al., 2006; N/A: chromosome number not available d GenBank accessions for sequences. N/A: sequence data not available

Recently, a gene referred to as MsDWF1 was identifi ed from diploid Medicago and L. Yang, unpublished results) support the hypothesis that GA3ox1 is a sin- sativa by Dalmadi et al. (2008) who hypothesized that the gene is orthologous gle-copy gene. to the Le gene based on syntenic map position and similarity of the mutant The GA3ox1 gene consists of two exons with a central intron; in Pisum sa- plants. GA3ox1 has not previously been used for phylogenetic purposes, al- tivum, the exons together are 1125 bp long, and the intron sequence is 544 bp though preliminary results suggest that it has potential use ( Steele et al., 1999 ). for a total of 1669 bp (Lester et al., 1997). Initially, we used primers slightly Southern blot analyses in Pisum (Lester et al., 1997) and Ononis (K. P. Steele modifi ed from those designed by Lester et al. (1997) for amplifi cation and 1146 American Journal of Botany [Vol. 97

Table 2. Amplifi cation and sequencing primers. replicates per partition homogeneity replicate (with a maximum of 500 trees per replicate) and by comparing bootstrap consensus trees derived from the indi- Primer name 5 ′ to 3 ′ sequence vidual data partitions. A general time reversible model with gamma shape parameter and propor- trnK1L CTCAATGGTAGAGTACTCG tion of invariant sites was selected as the best model for both the GA3ox1 and trnK 685F GTATCGCACTATGTATCATTTGA trnK/matK data sets based on the Akaike information criterion (Akaike, 1974) trnK 708R TCAAATGATACATAGTGCGATAC in the program MrModeltest (version 2; Nylander 2004 ). As in the MP analy- matK 3R CCTGTTGTCGAGATCTAGCTCG ses, multiple Bayesian analyses of both data sets, performed using the program matK 4L CTTCGCTACTGGGTGAAAGATG MrBayes v.3.1.1 (Ronquist and Huelsenbeck, 2003), included two separate matK 4R CATCTTTCACCCAGTAGCGAAG runs (4 chains per) of 2 – 5 × 106 generations, estimating branch length, substitu- matK 1932R CAGACCGGCTTACTAATGGG tion parameters, and topology and using uniform (default) priors, with the ex- matK 1932L CCCATTAGTAAGCCGGTCTG ception of the rate parameters (ratepr = variable), and sampling either every trnK 2R AACTAGTCGGATGGAGTAG 4000 or 10 000 generations. Stationarity of the Bayesian Markov chain Monte Carlo (MCMC) runs were based on the following criteria: (1) convergence to a GA3ox 18F CACCCTGATTTCAACTCAC stable value of the log likelihood score in separate runs, (2) a value less than GA3ox 3F CTCCTCCTTCTTCCCCAAACTCA 0.01 for the average standard deviation of split frequencies between two runs, GA3ox 5R AATGTTGAGTCCGTGTGCGGGGC and (3) a value approaching 1.0 for the potential scale reduction factor (PSRF) GA3ox 7R CTCTGACGGGTTCGGTTCAC for each parameter in the model. Trees sampled prior to stationarity were ex- GA3ox 20R GTGCCAAGGTACTCATTCC cluded by “ burnin ” (50% of samples) and the 100 – 250 remaining trees were used to construct a majority rule consensus tree with clade credibility values (posterior probabilities; PP). sequencing of pea DNA. Using those slightly modifi ed primers, we obtained sequences from several species including Lathyrus sativus , Medicago hypogaea , Trigonella calliceras, and Vicia villosa (K. P. Steele, unpublished data), RESULTS those sequences were then used to design primers suitable for amplifi cation of DNA from a larger number of species in tribes Trifolieae and Fabeae including Molecular markers plastid (trnK/matK) and nuclear DNA those in subtribe Trigonellinae ( Table 2 ). (GA3ox1)— Sequences of the trnK intron- matK gene region Protocols for PCR amplifi cation of both molecular markers were similar to those described previously ( Steele and Wojciechowski, 2003 ). Purifi ed PCR were obtained from 77 specimens representing 70 taxa (Table products were sequenced at the Arizona State University DNA Laboratory and 1). Characteristics of the sequences obtained and results of par- Davis Sequencing (http://www.davissequencing.com). For some taxa, PCR simony analyses are given in Table 3 . The matK portion of the products of GA3ox1 were cloned using the TOPO TA Cloning Kit (Invitrogen, sequences could easily be unambiguously aligned, but short Carlsbad, California, USA), following the manufacturer ’ s instructions, and a sections of the 5 ′ and 3 ′ portions of the trnK intron could not minimum of 20 putative recombinant clones per product were analyzed for and were excluded from analyses. Fourteen unambiguous gaps their inserts by restriction analysis and gel electrophoresis. The plasmids were purifi ed using QIAGEN Plasmid Mini Kit (Qiagen, Valencia, California, USA), in the trnK intron-matK gene region (110 total bases) were and clones of appropriate size were selected for sequencing. Preliminary phylo- coded as binary characters. The length of the matK gene ranged genetic analyses were carried out using sequences from all complete clones for from 1500 to 1536 bp, while the 5 ′ and 3 ′ portions of the trnK a particular species, but in the fi nal phylogenetic analyses presented here, one intron averaged approximately 744 bp and 257 bp, respectively complete clone was chosen at random and used to represent that species. Clone (both without insertions). Sequences of GA3ox1 were obtained sequences from a single individual plant were tested for the presence of recom- from 72 specimens representing 65 taxa (Table 1). Alignment bination using the Phi test (Bruen et al., 2006) as implemented in the program SplitsTree ( Huson and Bryant, 2006 ). of both intron and exon sequence required manual adjustment, Sequence fi les were assembled to produce contigs and edited, then assem- and some regions could not be unambiguously aligned and were bled into a data matrix using the program Sequencher versions 4.1– 4.7 omitted from the analyses. Thirty unambiguous gaps (345 total (GeneCodes, Ann Arbor, Michigan, USA). For the GA3ox1 sequences, initial bases) were coded as binary characters. Sequences of the alignments of both the exon and intron sequences were performed using ClustalX GA3ox1 gene ranged from 1400 to 1530 bp with an average ap- (Thompson et al., 1997), but manual adjustment was required. For the trnK/ proximate length of 1467 bp (without insertions) of which, on matK sequences, new sequences were added to our previous data set ( Steele and Wojciechowski, 2003 ), with relatively minor adjustments used for the 5 ′ and 3 ′ average, approximately 904 bp was exon sequence and 578 bp intron regions of trnK . All new sequences have been deposited in GenBank (see was intron sequence. The length of the intron sequence varied Table 1 ), and the fi nal data matrices have been deposited in the database Tree- among individuals sampled from 501 to 642 bp (without inser- BASE (http://www.treebase.org/treebase-web/home.html) under study acces- tions). Note that variation in exon length can be intrinsic or due sion S10406. to variation in sequencing techniques, e.g., how close to the Phylogenetic analyses used maximum parsimony (MP) and Bayesian infer- primer readable sequence is obtained, but length of the central ence methods. All parsimony and bootstrap analyses were conducted using heuristic search strategies, as implemented in PAUP* 4.0b10 (Altivec) intron is strictly intrinsic. (Swof ford, 2002), that included the following options: SIMPLE and CLOSEST Cloning was necessary to obtain readable sequence from the addition sequences, with tree-bisection-reconnection (TBR) branch swapping, following species with the number of cloned sequences indicated and retention of all multiple parsimonious trees. Multiple tree searches were in parentheses: M . cancellata (one complete clone), M. medi- conducted to increase the detection of globally optimal solutions (convergence caginoides (two complete clones, three partial), M . polyceratia to trees of shortest length found); for example, searches with MAXTREES set (fi ve complete clones), M . prostrata (one complete clone and to 10 000 with an automatic increase in the number of trees to be retained while branch swapping, with and without invoking steepest descent. All characters four partial), M . sativa (fi ve complete clones), and Trigonella used were unweighted and unordered. Relative levels of support for each clade anguina (two complete clones and three partial). A few differ- were estimated by nonparametric bootstrap analysis ( Felsenstein, 1985 ) as im- ences were observed among cloned sequences, but not more plemented in PAUP* 4.0b10; 500 replicate samples were obtained using identi- than could be reasonably accounted for by the presence of two cal heuristic searches. For each replicate, 500 trees were retained for bootstrap different alleles at one locus or random PCR/bacterial replication analyses ( Table 3 ). Confl ict between the data sets (based on a combined data set errors except in the case of M . polyceratia . There was no sup- with 69 taxa) was evaluated by the incongruence length difference test (Farris et al., 1995), performed in PAUP * 4.0b10 using searches as described for the port for the presence of recombination among cloned sequences maximum parsimony analyses (with a maximum of 500 trees and excluding within a single species for the six species that were cloned. uninformative characters) as well as by using two random addition sequence For two of the fi ve species tested, there were no informative July 2010] Steele et al. — Phylogenetic relationships in MEDICAGO 1147

Table 3. Summary of molecular data sets and their phylogenetic Comparison of GA3ox1 and trnK/matK sequences— There analyses. is a greater proportion of informative characters (24%) for the GA3ox1 data set compared to the trnK/matK data set (15%) Result trnK/matK GA3ox1 ( Table 3 ). In addition, there are more phylogenetically informa- Number of species 73 64 tive indels for the GA3ox1 data set (30) than for the trnK/matK Number of sequences 80 72 data set (14) even though the minimum aligned length of the Sequence characteristics GA3ox1 data set (1509 bp) is about 62% of the length of the Length of complete sequence (range) 2297 – 2448 1321 – 1509 trnK/matK Aligned length (maximum) 2833 1715 sequence (2448 bp). If the phylogenetic reconstruc- Variable sites: coding region 446 302 tions are compared, there are 36 groups with bootstrap values Variable sites: noncoding region 281 281 greater than 70% based on the trnK/matK data set for 80 taxa Pairwise differences % (intergeneric) 2.2 – 5.3 6.3 – 11.0 (Fig. 1), compared to 30 groups with the GA3ox1 data set for 72 (range) taxa ( Fig. 2 ). Pairwise differences % Medicago 0.00 – 4.0 1.0 – 6.6 (range) Use of GA3ox1 for phylogenetic analyses— Our study is the Pairwise differences % Trigonella 0.08 – 2.7 1.0 – 6.4 (range) fi rst published investigation of sequence variation in the nu- % Informative characters in data set 375/2448 × 100 366/1509 × 100 clear-encoded GA3ox1 gene for phylogenetic analyses, and our = 15.3 = 24.2 results suggest that this gene offers a reasonable number of in- Parsimony analyses formative characters to help resolve relationships among and No. of excluded characters 411 287 within genera in legumes (Table 3). With the exception of one Informative characters (coding/ 375 (224/151) 366 (201/165) species, M . polyceratia, only one gene is amplifi ed with our noncoding) by codon position 1st, 32, 21, 47 21.4, 16.4, 62.2 2nd, 3rd primers. Our analyses provide strong support for the orthology Gaps coded as binary characters 14 30 of Msdwf1 recently identifi ed by Dalmadi et al. (2008) and the No. of trees 21860 1156 GA3ox1 gene that we sampled; sequences of Msdwf1 from their No. of steps, excluding uninformative 1247 (861) 1530 (1285) study from M . sativa subsp. falcata and M . sativa subsp. caeru- characters lea and our GA3ox1 sequences from the same taxa are all found CI, excl. uninformative characters 0.7081 (0.5772) 0.5340 (0.4451) nested within the sect. Medicago clade with sequences from RI 0.8148 0.7693 RC 0.473 0.3425 other taxa within M . sativa. The presence of intraspecifi c varia- No. of groups with bootstrap 36 30 tion supports the hypothesis that GA3ox1 sequences could be support > 70% valuable for phylogenetic reconstruction among closely related species. Note that analyses of GA3ox1 sequences indicate that Notes: CI = consistency index, RI = retention index, RC = rescaled Medicago consistency index within sect. fi ve groups are resolved with greater than 60% bootstrap support ( Fig. 2 ), but only three groups are resolved within that same larger group in analyses of trnK/matK sequences. characters, while for the other three the Phi test did not fi nd statistically signifi cant evidence for recombination. One species Phylogenetic analyses— Results of maximum parsimony had a single complete cloned sequence and could not be tested. and Bayesian inference analyses of the trnK/matK sequences In preliminary phylogenetic analyses, sequences from all are shown in Fig. 1 , while results of identical analyses of complete clones for each species were used. In all cases, the GA3ox1 sequences are shown in Fig. 2. Character states for cloned GA3ox1 sequences from a single sample form a mono- specifi c characters, that were not included in the analyses, pe- phyletic group, except with M . polyceratia where two sequences rennial or shrub habit, chromosome number to indicate aneu- form a single group that is itself sister to a group in which M . ploid reduction (2n = 14) and polyploidy, as well as loss of the fi scheriana is sister to three additional sequences from M . rpoC1 intron are indicated in Fig. 1 . Roman numerals indicate polyceratia (results not shown). Sequence differences among the presence of well-supported clades that are further consid- most cloned sequences derived from one taxon range from 0.1 – ered in the discussion (online Appendix S2). 1.0%. Slightly higher values are found among clones from The results of an incongruence length difference test sug- Trigonella anguina and range from 0.4 to 1.8%. To put these gested signifi cant confl ict between the plastid trnK/matK and values in perspective, note that in Pisum sativum the difference nuclear GA3ox1 gene data sets (all, P < 0.01). Comparison of between the “ wild-type ” allele (GenBank accession U93210) clade support values derived from maximum parsimony analy- and a dwarf mutant allele, le , (GenBank accession AF004730) ses and Baysian analyses showed that, although many of the is 0.24%. On the other hand, in M . polyceratia , differences same well-supported clades were found using both data sets, among clones in one clade range from 0.26 to 0.72%, similar to some taxa were in modestly to well-supported, but alternative the values obtained for clones from other species as indicated phylogenetic positions within trees derived from the two data above, but differences between clones in the two clades range sets ( Figs. 1, 2 ). from 2.7 to 3%. That value is similar to the differences between Analyses of trnK/matK sequence data confi rm the mono- those clades and M . fi scheriana , 2.5 to 4%. phyly of Medicago and Trigonella ( Fig. 1 ) with each genus sup- There is also intraspecifi c variation among GA3ox1 se- ported by bootstrap values of 100%. Species in section Buceras quences in species of Medicago . For three species, we have two (four species sampled) form a well-supported monophyletic different accessions; there are six base substitutions between group, and species of section Platycarpae , M . ruthenica and M . accessions of M . monantha (accession numbers 35 and 183), platycarpa, are always sister taxa in analyses of both markers although these two are not sister taxa (Figs. 1, 2), two base sub- (clades I and II, respectively, in Figs. 1, 2 ). Representatives of stitutions between accessions of M . lupulina and three substitu- section Lunatae , M . bifl ora and M . brachycarpa, form a mono- tions and two small indels between accessions of M . littoralis . phyletic group based on trnK/matK sequence data ( Fig. 1 ). 1148 American Journal of Botany [Vol. 97

Species in section Spirocarpos do not form a monophyletic ( Small and Jomphe, 1989b ; Small, 2010 ; http://plants.usda. group in any of our analyses, although species in subsection gov/), but in the trnK/matK analyses our four samples of M . Intertextae form a very strongly supported group with analyses sativa do not form a monophyletic group ( Fig. 1 ). Analyses of using both markers (clade VI in Figs. 1, 2). In addition, two the GA3ox1 data provide evidence for a weakly supported group species from subsection Rotatae , M . shepardii and M . bonaro- that includes two hexaploid species, M . cancellata and M . saxa- tiana form a well-supported group, hereafter referred to as the tilis in addition to six samples of M . sativa (including two se- “ reduced subsection Rotatae clade ” (clade IV in Figs. 1, 2 ). quences from GenBank, Table 1 ) ( Fig. 1 ). Within section Spirocarpos, results based on analyses of the Three species, M . arabica (sect. Spirocarpos ), M . suffruti- trnK/matK data indicate a modestly supported clade of 10 spe- cosa subsp. suffruticosa (sect. Medicago ), and M . orbicularis cies from three subsections; seven species from subsection (monotypic sect. Orbiculares), are not part of any well-sup- Pachyspireae including M. truncatula , one species from sub- ported group using either trnK/matK or GA3ox1 sequence data section Leptospireae , M . praecox, and two from subsection Ro- and are referred to hereafter as “ orphan species ” (indicated in tatae , M . noena and M . scutellata, the latter a polyploid species. Figs. 1 and 2 with double asterisks). This clade of 10 species will be referred to hereafter as the The four sampled species of Melilotus form a weakly sup- “ subsection Pachyspireae clade ” (clade VII in Figs. 1, 2 ). ported group with Trigonella cretica based on analyses of trnK/ Within this clade, six species, all in subsection Pachyspireae , matK data (Fig. 1). Only two species of Melilotus were sampled comprise a very well-supported group. The only other sampled for GA3ox1 ; analyses of those data show those two species as species in the subsection, M . rigidula, is present in the larger part of a basal polytomy within the clade of all Trigonella spe- clade, but not in this smaller well-supported group. However, cies (Fig. 2). Trigonella bicolor, previously included in Melilo- based on GA3ox1 sequence data alone fi ve of the 10 species of tus is clearly nested within Trigonella , the strongly supported the “ subsection Pachyspireae clade ” form a strongly supported sister group to the two sampled species of sect. Cylindricae , T . group, the remaining species are part of a polytomy with a vari- kotschyi and T . spruneriana ( Figs. 1, 2 ). ety of species of Medicago ( Fig. 2 ). Based on analyses of sequences of both genes, the model legume, M . truncatula , is included within the “ subsection Pachyspireae clade ” and forms DISCUSSION a very strongly supported group with M . littoralis and M . italica; the three species form a polytomy (Figs. 1, 2). In all, we Clades supported by most or all molecular markers— Clades sampled eight species from subsection Pachyspireae : seven of found in analyses of at least four of the fi ve molecular markers which are found in the subsection Pachyspireae clade men- used in this study or others ( Bena, 2001 ; Maureira-Butler et al., tioned above, M . constricta , M. turbinata , M . soleirolii , M . 2008) are indicated in Figs. 1 and 2 with roman numerals. On- truncatula , M . littoralis , M . italica, and M . rigidula . Medicago line Appendix S2 provides a comparison of the well-supported murex is an important exception; it is the strongly supported clades found in analyses of each of the fi ve molecular markers. sister species of M . polymorpha from subsection Leptospireae in all analyses (clade V in Figs. 1, 2). The group formed by Evolution of chromosome number— Both polyploids and these two species will be referred to hereafter as the “ polymor- aneuploids are found within Medicago ( Lesins and Lesins, pha clade” in the discussion. Medicago heyniana is sister to 1979); both are concentrated within some clades and not found these two species, both 2n = 14, with moderate support in the in others. For example, several polyploids, including M . sativa trnK/matK analyses, but not in analyses of GA3ox1 sequences. (alfalfa), are in section Medicago , but no aneuploids are found An interesting, well-supported clade, not predicted by the in that section. The base number in the genus is thought to be x present taxonomy, is that formed by M . lupulina (sect. Lupu- = 8, although some annual species have x = 7 (Goldblatt, 1981); laria ), M . tenoreana (sect. Spirocarpos , subsection Leptospi- this is the base number in its sister group Trigonella , in other reae ) and their sister species, M . minima (also subsection genera in the tribe Trifolieae, and is the most common number Leptospireae) (clade III in Figs. 1, 2). This group, hereafter re- in the genus Medicago . Within sect. Spirocarpos (all annuals), ferred to as the “ reduced subsection Leptospireae clade ” , never eight species are aneuploids and have a 2n = 14 chromosome includes M . secundifl ora, the only other species in sect. number. Aneuploid reduction, whereby chromosome material Lupularia . from one small chromosome is added to another chromosome Medicago sativa and its relatives in section Medicago form a to form one larger chromosome instead of two smaller ones, has weakly to strongly supported monophyletic group, hereafter re- been hypothesized to result in the change from 2 n = 16 to 2 n = ferred to as the “ section Medicago clade” , that includes nearly 14 ( Lesins and Lesins 1979 ). One question is of particular inter- all sampled species from this section; however, there are differ- est; how many times has 2 n = 14 arisen within the genus Medi- ences in species included depending on the marker used (clade cago ? According to the current taxonomy ( Table 1 ), species VIII in Figs. 1, 2). Two species, M . marina and M . prostrata , with 2 n = 14 are found in two subsections, Pachyspireae and are included in the “ section Medicago clade ” based on analyses Leptospireae , and in sect. Geocarpae , thus 2n = 14 is most of the trnK/matK data (Fig. 1), but are not included in this clade likely to have arisen at least three times. But examination of based on analyses of GA3ox1 data ( Fig. 2 ). Members of the chromosome number and results of phylogenetic analyses of “ section Medicago clade” share a 19-bp deletion in their both markers indicate at least three origins in two clades, the GA3ox1 intron sequences. Medicago marina and M . prostrata “ polymorpha clade ” (V) and the “ subsection Pachyspireae have those 19 bases as do all other species of Medicago . Medi- clade ” (VII) and additionally, in the monotypic sect. Geocar- cago arborea (section Dendrotelis ) is always resolved within pae ( Fig. 1 ). the clade with section Medicago ( Figs. 1, 2 ); it shares the 19-bp The “ polymorpha clade ” (V) includes two species, M. murex deletion with other species in section Medicago . Medicago sa- and M . polymorpha , each 2n = 14, that are strongly supported tiva is often defi ned broadly and includes three to fi ve subspe- sister species in analyses with both markers (Figs. 1, 2). Other cies, for example, M . sativa L. subsp. falcata (L.) Archang. workers also found a group that includes these two species; July 2010] Steele et al. — Phylogenetic relationships in MEDICAGO 1149

Fig. 1. Phylogenetic relationships of Medicago based on maximum parsimony analyses of plastid trnK/matK sequences. Tree shown is strict consensus of 21 860 most parsimonious trees (length = 1247 steps; CI = 0.7081; RI = 0.8148) from heuristic search analyses of 80 trnK/matK sequences. Numbers along branches indicate bootstrap percentages above 50%. Numbers below branches are Bayesian posterior probabilities, indicated for those clades found in both parsimony and Bayesian trees. Thicker branches indicate bootstrap level support greater than 80%. Sections of Medicago and the monophyletic subsect. Intertextae of sect. Spirocarpos are indicated on the left. Colors indicate three polyphyletic subsections of sect. Spirocarpos . Informal groups discussed in the text are indicated on the left in quotations. Clades identiﬁ ed as a result of analyses of other molecular markers are indicated with roman numerals. Information on habit, ploidy level, and loss of the rpoC1 intron is shown and symbols for those characters are explained at the bottom of ﬁ gure. Taxa with both diploid and polyploid populations are indicated as polyploid (see Table 1 ). “ M ” indicates monotypic sections. 1150 American Journal of Botany [Vol. 97

Bena (2001) identifi ed a moderately supported clade that includes M . cancellata hypothesized to be M . rupestris and M . sativa M . murex as sister to M . polymorpha with M . syriaca (n = 8) as ( Lesins and Lesins, 1979 ). Our phylogenetic reconstructions pro- sister to them, while Downie et al. (1998) found that M . polymor- vide some support for that hypothesis. Results of analyses based on pha , M . syriaca, M. lesinii, and M. laxispira form a moderately GA3ox1 sequences show M . cancellata as the sister species to supported group, but without resolution among the species. Mau- M. sativa subsp. falcata (the Msdwf1 sequence), although analy- reira-Butler et al. (2008) found that M . murex , M . polymorpha , ses of trnK/matK data show both species unresolved as part of a M. lesinii , and M . laxispira form a well-supported group based polytomy with other species in sect. Medicago . The putative on analyses of CNGC5 and β cop, and most interestingly M . parents of M . saxatilis have not been hypothesized, but the spe- murex is sister to M . lesinii (n = 8), and M . polymorpha is sister cies can be crossed with M. sativa and M. cancellata , suggesting to M . laxispira ( n = 8). If we only considered our results and the two hexaploid species have genomes in common. Our results those of Bena (2001) , it would appear that the n = 7 chromosome support that hypothesis as M . saxatilis is always nested in the number is a synapomorphy for M . murex and M . polymorpha . clade with infraspecifi c taxa of M. sativa . Results based on anal- However, based on all analyses to date it appears the n = 7 condi- yses of ITS/ETS by Bena (2001) show M . cancellata as part of tion has arisen twice among the fi ve species within this clade. a polytomy along with species from sect. Medicago and M. sax- Gillies (1977 , published in 2006, p. 2) noted that M . murex and atilis as sister to M . sativa subsp. falcata . Note that all species in M . polymorpha “ have almost identical pachytene ideograms.” the section are either herbaceous perennials or shrubs, and poly- One can consider the similarity in karyotype to have arisen inde- ploids are more common in taxa with this life form than are an- pendently, perhaps supporting the hypothesis that a similar pro- nuals (Grant, 1981). cess was involved and possibly that there is a predilection for Relatively few species in sections Lunatae and Buceras of aneuploid reduction to occur among the species in the “ poly- Medicago have had their chromosome number determined, al- morpha clade” . Genomic in situ hybridization was used to test though a few species in section Buceras , such as M . medicagi- the hypothesis that M . murex is closely related to M . lesinsii noides (as Trigonella arcuata ) have 2n = 16 ( Goldblatt and (Falistocco et al., 2002) and extensive cross-hybridization sup- Johnson, 1979). But there are a few records of unusual chromo- ported that hypothesis. In addition, using similar analyses of both some numbers from species in sect. Buceras, which need to be M . murex and M . lesinsii with M . littoralis (as a closely related further investigated if we are to more fully understand chromo- species in the same subsection), they found much less cross- some evolution within Medicago . Medicago monantha , M . or- hybridization ( Falistocco et al., 2002 ). This latter result is con- thoceras , and M . polyceratia all have records of n = 22 or 2n = sistent with our results showing M . littoralis in a different 44, but M . monantha also has a record of n = 24 and M . polyc- well-supported group from M . murex . eratia also has a record of 2n = 28 ( Goldblatt and Johnson, The “ subsection Pachyspireae clade” (VII) includes four 1979 ). Results of cloning and sequencing of GA3ox1 PCR species with 2 n = 14, M . constricta , M . rigidula , rigiduloides , products support the hypothesis that M . polyceratia is a poly- and M . sinksiae that form a monophyletic group when results ploid or at least has a duplicated GA3ox1 gene (see Results). from all workers are considered, and thus the chromosome Unfortunately, little is known about changes in chromosome number reduction from 2n = 16 to 2n = 14 may have taken place number within the sister group of Medicago , i.e., the genus only once. Medicago praecox , another n = 7 species, is part of Trigonella (including Melilotus ). Records are available for only the “ subsection Pachyspireae clade ” based on some analyses 19 of the approximately 55 species ( Goldblatt and Johnson, ( trnK/matK and CNCG5), but not with others (ITS/ETS, 1979 ), and all report counts of 2 n = 16 except for single records GA3ox1, and β cop). In any case, this species is never part of the for two species. Nearly all records for Melilotus are 2 n = 16 smaller group with the other n = 7 species, so the n = 7 chromo- ( Goldblatt and Johnson, 1979 ). Thus, in Trigonella (including some number in M. praecox must represent an independent Melilotus), there is little or no aneuploidy and very little poly- aneuploid reduction. ploidy, in contrast to the variation in chromosome number in its Within the “ subsection Pachyspireae clade ” , we also fi nd a sister group, Medicago . polyploid species, M. scutellata of subsection Pachyspireae , that is 2n = 30 ( Fig. 1 ) ( Bauchan and Elgin, 1984 ). Medicago Life history and habit — Most species in subtribe Trigonel- scutellata is thought to be a polyploid derivative of a hybrid linae are annuals (about 80% of species of Medicago , about between a 2n = 16 species and a 2n = 14 species (Lesins and 50% species of Trigonella and all species of Melilotus ). On- Lesins, 1979 ; Bauchan and Elgin, 1984 ), but the diploid species onis , the sister group to the subtribe ( Steele and Wojciechowski, involved have not been identifi ed. Medicago scutellata was also 2003 ), also includes both annuals and perennials ( Lewis et al., sampled by Bena (2001) who also found it nested in a group 2005 ). Medicago arborea and its close relatives are the largest similar to our “ subsection Pachyspireae clade” . These results woody species, and some perennial Medicago and Trigonella have suggest that one or both parental species of M . scutellata are aboveground woody tissue (e.g., M . cretacea and T . elliptica ). likely to be found within the “ subsection Pachyspireae clade” . Results of our phylogenetic analyses of subtribe Trigonellinae Genomic in situ hybridization was used to test the hypothesis allow us to consider two questions on the evolution of life his- that M . murex (2 n = 14) was one of the parents of M . scutellata tory within those taxa. First, what is the character state of the ( Falistocco et al., 2002 ), but their results did not support that early diverging groups within the two major clades, the genus hypothesis consistent with the placement of M . murex in the Medicago and the genus Trigonella , and second, where are pe- strongly supported “ polymorpha clade” as discussed above. rennial and/or woody taxa found within the tree? Section Medicago has no aneuploid species, but it does have Unfortunately, there is no consensus on the earliest-branch- several polyploid taxa including M . cancellata and M . saxatilis ing lineage within Medicago . We sampled similar perennial (both hexaploids), and few to many autotetraploid individuals species of Medicago from sect. Platycarpae and annual Trigo- and/or populations within some taxa, e.g., M . papillosa, M . nella as those sampled by other authors ( Bena, 2001 ; Maureira- prostrata, and M . sativa (Small, 1986). The two hexaploid spe- Butler et al., 2008 ), but results of all studies are as yet cies are both thought to be allopolyploids, with the parents of inconclusive. Based on analyses of GA3ox1 sequences, there is July 2010] Steele et al. — Phylogenetic relationships in MEDICAGO 1151

Fig. 2. Phylogenetic relationships of Medicago based on maximum parsimony analyses of nuclear GA3ox1 sequences. Tree shown is strict consensus of 1156 most parsimonious trees (length = 1530 steps; CI = 0.5340; RI = 0.7693) from heuristic search analyses of 72 GA3ox1 sequences. Numbers along branches indicate bootstrap percentages above 50%. Numbers below branches are Bayesian posterior probabilities, indicated for those clades found in both parsimony and Bayesian trees. Thicker branches indicate bootstrap level support greater than 80%. Sections of Medicago and the monophyletic subsect. Intertextae of sect. Spirocarpos are indicated on the left. Colors indicate three polyphyletic subsections of sect. Spirocarpos. Informal groups discussed in the text are indicated on the left in quotations. Clades seen in results of analyses of other molecular markers are indicated with roman numerals. “ M ” indicates monotypic sections. “ Gb ” indicates sequences obtained from GenBank (accession numbers given in Table 1 ). 1152 American Journal of Botany [Vol. 97 weak support for a basalmost node with the two perennial spe- seeded species in Medicago sect. Lupularia , (summarized cies of sect. Platycarpae , M . platycarpa and M . ruthenica ( Fig. in Small, 1987b). However, species of Melilotus are always 2 ), while for analyses of trnK/matK there is a similar degree of found nested within Trigonella in analyses of molecular weak support for a basal node leading to sect. Buceras (consist- data (Wojciechowski et al., 2000; Bena, 2001; Steele and ing only of annual species) plus M . radiata as the sister group Wojciechowski, 2003 ). Our results suggest that reduction of the to the rest of the genus, while the two species in sect. Platycar- number of seeds in a legume is homoplastic, having occurred pae are included within a strongly supported group that includes independently in more than one lineage within Trigonella , and most species of Medicago one node deeper in the tree (Fig. 1). that one of those lineages includes all species of Melilotus plus This is one of the most noticeable discordant aspects of the tree T . cretica. topology when results of the two analyses are compared. Al- Within Medicago , the only two species with single-seeded, though the character state(s) of the most recent common ances- indehiscent fruits, M . lupulina and M . secundifl ora , are pres- tor of most species of Medicago cannot be established at this ently placed in sect. Lupularia ( Small and Jomphe, 1989b ). time, the results of previous molecular analyses (Downie et al., However, there are numerous morphological characters that are 1998 ; Bena, 2001 ; Maureira-Butler et al., 2008 ) and ours, sup- different between the two species such as infl orescence type port the hypothesis that the perennials of sect. Medicago are not (compact racemes of 14– 24 fl owers in M . lupulina and secund among the earliest-branching lineages and are most likely to be racemes of 3 – 10 fl owers in M . secundifl ora ), pattern of vena- derived from an annual ancestor. This pattern of perennial taxa tion on the mature legumes, pollen morphology, position of that appear to be derived from annual ancestors is also observed radicle on the seed, and number of ovules per ovary (two in M . in Trigonella ( Fig. 1 ). lupulina and one in M . secundifl ora ) ( Small et al., 1981 ; Small, The only shrubby species within subtribe Trigonellinae are 1988; Small and Jomphe, 1989b). Furthermore, comparison of in Medicago, sect. Dendrotelis. Two of the species, M. arborea the karyotypes and restriction endonuclease fragment patterns and M . strasseri, are tetraploids, while M . citrina is hexaploid, of the chloroplast genome of M . lupulina and M . secundifl ora although the origin of the polyploidy, whether auto- or allop- found a much greater degree of difference between them than loidy is unknown in each case. If allopolyploid, the putative was found between two closely related species in subsection parents of the three species are unknown ( Lesins and Lesins, Intertextae (Schlarbaum et al., 1989). E. Small (Agriculture and 1979 ; Small and Jomphe, 1989b ; Rosato et al., 2008 ). Based on Agri-Food Canada, personal communication) indicates that analyses of all markers, M . arborea is part of a group with other changes in fruit morphology such as the lack of spines, reduc- species of sect. Medicago , although it is often unresolved as tion in venation, and coiling associated with reduction in seed part of a basal polytomy within that group. It is likely that the number greatly reduces the number of taxonomically useful common ancestor of these shrubby, polyploid species is an her- characters that can be used to help resolve relationships be- baceous perennial in sect. Medicago , thus woodiness is a de- tween these and other species. rived character state in these species. Medicago lupulina and M . secundifl ora are each found in separate strongly supported groups in our analyses ( Figs. 1, 2 ) and Cotyledonary pulvini— Species in sections Buceras and Lu- those of other researchers (Downie et al., 1998; Bena, 2001; natae all have pulvinate cotyledons as do all species of Trigo- Maureira-Butler et al., 2008 ); it is most likely that the single- nella and Melilotus , whereas all other species of Medicago lack seeded, indehiscent fruit found in the two species arose indepen- cotyledonary pulvini ( Small and Brookes, 1984 ; Small, 1987a ). dently. Medicago lupulina is the strongly supported sister species This characteristic was originally used to support the inclusion of M . tenoreana (sect. Spirocarpos , subsection Leptospireae ). of sections Buceras and Lunatae in Trigonella (summarized in With their sister species, M . minima (also subsect. Leptospireae ), Small, 1987b). The presence of pulvini in sections Buceras and they form a well-supported group based on analyses of most mo- Lunatae supports the hypothesis that the species in sections lecular markers (Figs. 1, 2 (clade III); Downie et al., 1998; Bena Buceras and Lunatae comprise the earliest-diverging branch in 2001; Maureira-Butler et al., 2008). These three species all lack the genus and that pulvini have been lost only once. Medicago the rpoC1 intron, whereas M . secundifl ora has this intron edgeworthii is currently in sect. Platycarpae (a group that lacks ( Downie et al., 1998 ). Bena (2001) showed that two other spe- pulvini), but Maureira-Butler et al. (2008) found it in a strongly cies, M . disciformis and M . coronata (in subsect. Leptospireae ), supported group with two species in sect. Lunatae , M . brachy- which also lack the rpoC1 intron were included in a clade with carpa and M . huberi. Interestingly, Small (2010) indicates that M . lupulina . Inclusion of M . lupulina in this group of species M . edgeworthii, although previously reported to lack pulvini from subsect. Leptospireae is surprising given that it had not (e.g., Small and Jomphe, 1989b), has now been shown to have been predicted based on morphological characters. Medicago lu- pulvini and has been transferred to sect. Lunatae . Cotyledonary pulina has some value for forage as a pasture legume ( Rum- pulvini in sect. Lunatae may be a plesiomorphic trait because baugh, 1990 ; Hanelt 2001 ) and has been tested for its ability to species of Trigonella have cotelydonary pulvini, but the sister hybridize with M . secundifl ora and with species in sect. Medi- group of subtribe Trigonellinae, Ononis, lacks pulvini in the cago, but no defi nitive hybrids were formed with any of the latter eight species sampled by Small and Brookes (1984) . species, (summarized in Lesins and Lesins, 1979). If M . lupulina were to be crossed with M . tenoreana and/or M . minima , perhaps Number of seeds per fruit— The hypothesized reduction in it would be able to hybridize with M . tenoreana and/or M . min- number of seeds per fruit has been an important character to ima and gain additional agriculturally important characteristics. support particular hypotheses of relationship within the subtribe Trigonellinae (Small, 1987b). For example, recognition of the Sister group of Medicago truncatula— The model genomic genus Melilotus is based in part on its single-seeded, indehis- species M . truncatula is resolved in a very well-supported clade cent fruits and very small fl owers in racemes, which differ from with two other species, M . italica and M. littoralis in our analyses the character states in most species of Trigonella . Melilotus has (Figs. 1, 2) and previous studies (Bena, 2001; Maureira-Butler been considered to be closely related to Trigonella or to single- et al., 2008). However, the relationship among the three species July 2010] Steele et al. — Phylogenetic relationships in MEDICAGO 1153 is not consistently resolved in the molecular analyses. Based on and GA3ox1 (the present study). Maureira-Butler et al. (2008) our trnK/matK and GA3ox1 results, the three species form a found a strongly supported group with M . edgeworthii from sect. polytomy (Figs. 1, 2). On the other hand, based on ITS/ETS Platycarpae and two species from sect. Lunatae, M. brachy- sequences, M . truncatula is sister to a group formed by M . ital- carpa and M . huberi . Examination of herbarium specimens of ica (as M . tornata) and M . littoralis (Bena, 2001), while with those three species (K. P. Steele, unpublished data) indicates that CNGC 5 the three species form a polytomy and with β cop , M . all have a fairly distinctive dense orange-brown pubescence. truncatula and M . littoralis are sister species with M . italica Small (2010) indicates that M . edgeworthii has been transferred sister to those two species (Maureira-Butler et al., 2008). Both to sect. Lunatae . Furthermore, it is interesting to note that M . Small and Brookes (1990a, b ) and Lesins and Lesins (1979) cretacea , previously in sect. Platycarpae has been transferred to consider M . littoralis and M . truncatula more closely related to sect. Medicago by Small, (2010). Both Bena (2001) and Maurei- each other than either is to M . italica, although neither stated ra-Butler and colleagues (2008) found M . cretacea to be part of that explicitly. Medicago littoralis is thought to hybridize with a clade comprised of species from sect. Medicago , although this M . truncatula in nature ( Small and Brookes, 1990a , b ); how- result was not discussed by either group. The relationship of ever, there is a breeding barrier between M. italica and M . trun- Medicago cretacea, which is endemic to the Crimea and Cauca- catula (Lesins and Lesins, 1979), and while M . italica and M . sus Mountains, to taxa in sect. Medicago merits further investi- littoralis can be crossed, some specifi c strains will not (Lesins gation (K. P. Steele et al., unpublished manuscript). and Lesins, 1979 ; Small and Brookes, 1990b ). The following three groups represent strongly supported hypotheses that could be tested with expanded sampling both of Taxonomic implications— Certainly there is no longer any taxa and molecular markers: (1) A “ reduced subsection Lep- question about the delimitation of the two genera, Trigonella tospireae clade” (clade III) that includes M . lupulina in addition and Medicago (including all of the former “ medicagoid ” Trigo- to M . coronata , M . disciformis , M . minima , and M . tenoreana . nella), as analyses of all molecular markers indicate each is a This group is supported by results of analyses of several mo- very well-supported monophyletic group. However, when only lecular markers and loss of the rpoC1 intron. (2) The “ polymor- morphological characteristics are considered placement of pha clade ” (clade V), which includes M. laxispira and M . some species into either genus can be diffi cult ( Small, 1987b ). polymorpha (subsection Leptospireae), as well as M . lesinsii , Additionally, analyses of molecular data clearly show Melilo- M . murex , M . sphaerocarpos), and M . syriaca, which are cur- tus nested within Trigonella ( Figs. 1, 2 ). Trigonella has nomen- rently in subsection Pachyspireae. (3) A “ subsection Pachyspi- clatural priority because it was published in Species Plantarum reae clade” (clade VII), that includes M . constricta , M. in 1753 ( Linnaeus, 1753 ), while Melilotus was published in turbinata , M . soleirolii , M . truncatula , M . littoralis , M . italica , 1754 (Miller, 1754). Historically, a few species have been and M . rigidula from that subsection and one species from sub- transferred from one genus to the other. For example, M . bisection Leptospireae , M . praecox, and three from subsection color Boiss. & Bal. was transferred to Trigonella as T . bicolor Rotatae , M . noena and the two hexaploid species, M . rugosa (Boiss. & Bal.) Lassen ( Small and Jomphe, 1989a ) based on and M . scutellata . Highlighting the placement of these species overall similarity to species of Trigonella . The inclusion of Me- in one group could help narrow the search for the parents of the lilotus species in Trigonella will necessitate name changes for two hexaploid species. Morphological and biochemical simi- some important weeds and forage crops, but this taxonomic re- larities among the species in each of these three hypothesized vision would more accurately refl ect the close relationship groups should be further investigated. among species currently recognized as two genera. Other well-supported changes of taxonomic signifi cance in- Taxa of special interest including “ orphan” taxa— The phy- clude the following: (1) the shrubby species M. arborea, M. cit- logenetic relationships of three species, M . hypogaea , M . lanig- rina , and M . strasseri (currently sect. Dendrotelis ) should be era, and M . radiata , are of particular interest for a number of included within sect. Medicago where their placement will high- reasons. All three are very distinctive morphologically, two had light the phylogenetic relationship and morphological similari- been placed in segregate genera (as discussed later) and there is ties among those species. Hybridization is diffi cult to accomplish no agreement on their closest relatives. Medicago hypogaea E. between species in sect. Medicago , especially those that differ in Small, fi rst described as Trigonella aschersoniana by Urban in ploidy level ( Lesins and Lesins, 1979 ; McCoy and Bingham, 1882, was segregated into a monotypic genus, Factoryovskya 1988 ; Bauchan and Hossain, 1999 ); however, female fertile hy- by Eig in 1927, was then placed into Medicago by Small and brids between M . sativa and M . arborea have been produced Brookes (1984). Earlier analyses of sequence data (Downie et ( Bingham, 2005 ; Armour et al., 2008 ). (2) Section Lupularia, al., 1998 ; Steele and Wojciechowski, 2003 ) provided very presently containing the two species M . lupulina and M . secun- strong support for its inclusion within Medicago, but there was difl ora, should no longer be recognized. The two species are only weak support for its relationship to other species of Medi- never resolved as sister taxa and M . lupulina is nested in a sepa- cago . Small and Brookes (1984) hypothesized that M . hypo- rate, well-supported clade (reduced subsection Leptospireae ) gaea and M . lanigera were sister species; they have similar and based on all markers sampled. (3) Section Platycarpae is un- unique (within Medicago) cottony hairs on the fruit. However, likely to be monophyletic based on results from analyses of all M . lanigera has the plesiomorphic 2 n = 16 chromosome num- molecular markers and thus should not be recognized in its cur- ber of subtribe Trigonellinae and M . hypogaea has 2n = 14. rent form. However, three groups of species within the section Both species lack the plastid rpoC1 intron, although Downie et are part of strongly supported groups. For example, three species al. (1998) hypothesized that this loss in these species represents from this section, M . platycarpa, M. ruthenica, and M. popovii, two of the minimal three independent losses in the genus. Our form a consistently strongly supported clade using CNGC 5 and results do not provide strong support for a sister group relation- β cop (Maureira-Butler et al., 2008); studies using other markers ships between these two species nor for any other particular only sampled two of the these three species, but those pairs are topology ( Figs. 1, 2 ). Downie et al. (1998) , the only other work- also strongly supported by ITS/ETS ( Bena, 2001 ), trnK/matK , ers to sample both species, also did not fi nd either species in a 1154 American Journal of Botany [Vol. 97 well-supported group. Phylogenetic reconstructions based on LITERATURE CITED trnK/matK and GA3ox1 indicate that M . lanigera and M . hypogaea (particularly the latter) are on relatively long branches Akaike , H. 1974 . A new look at the statistical model identifi cation. IEEE (see online Appendices S3, S4), and this may make it diffi cult Transactions on Automatic Control 19 : 716 – 723 . Armour , D. J. , J. M. Mackie , J. M. Jusial , and J. A. G. Irwin . 2008 . to accurately determine the relationships of these species. Transfer of anthracnose resistance and pod coiling traits from Medicago radiata L. has been ( Medikus, 1789 ) and contin- Medicago arborea to M. sativa by sexual reproduction. Theoretical ues to be recognized as the monotypic genus Radiata by some and Applied Genetics 117 : 149 – 156 . authors (e.g., Yakovlev et al., 1996 ). However, results based on Bailly , X. , I. Olivieri , B. Brunel , J.-C. Cleyet-Marel , and G. Bena . molecular data strongly support its placement in Medicago 2007 . Horizontal gene transfer and homologous recombination drive ( Downie et al., 1998 ; Bena, 2001 ; Maureira-Butler et al., 2008 ) the evolution of the nitrogen-fi xing symbionts of Medicago species. with some consensus that it diverged relatively early and is Journal of Bacteriology 189 : 5223 – 5236 . more closely related to species in sections Buceras , Lunatae , Bauchan , G. R. , and J. H. Elgin . 1984 . A new chromosome number for and Platycarpa than to other species of Medicago . Our results the genus Medicago. Crop Science 24 : 193 – 195 . indicate that this species is weakly supported as a sister to all Bauchan , G. R. , and M. A. Hossain. 1999 . Advances in alfalfa cytogenet- species of Medicago except those of sect. Buceras ( Fig. 1 ) or ics [online]. In Symposium proceedings of The Alfalfa Genome, 1999, Madison, Wisconsin, USA. Available at http://www.naaic.org/TAG/ that it is part of a polytomy within a large group of species of TAGpapers/Bauchan/advcytog.html [accessed 7 June 2009.] North Medicago except sect. Platycarpae (Fig. 2). Downie et al. American Alfalfa Improvement Conference, Manhattan, Kansas, USA. (1998) found M . radiata to be within a moderately supported Bena , G. 2001 . Molecular phylogeny supports the morphologically based group that includes species in sections Buceras , Lunatae , and taxonomic transfer of the “ medicagoid ” Trigonella species to the genus Platycarpae , while Bena (2001) resolved Medicago radiata as Medicago L. Plant Systematics and Evolution 229 : 217 – 236 . the strongly supported sister to sections Buceras and Lunatae . Bingham , E. T. 2005 . Medicago arborea project at University of Maureira-Butler et al. (2008) found M . radiata nested within a Wisconsin. Medicago Genetic Reports 5 : 1 – 7 [online]. Website http:// group with species from sections Buceras , Lunatae , and Platy- www.medicago-reports.org/index.html [accessed 13 June 2008]. carpae based on analyses of CNCG 5, but in an entirely differ- Bruen , T. C. , H. Philippe , and D. Bryant . 2006 . A simple and robust ent group based on analyses of β cop sequences. Thus, while statistical test for detecting the presence of recombination. Genetics 172 : 2665 – 2681 . there is strong support for the inclusion of M . radiata within Bruneau , A. , M. Mercure , G. P. Lewis , and P. S. Herendeen . 2008 . Medicago , currently analyses of molecular data show discor- Phylogenetic patterns and diversifi cation in the caesalpinioid legumes. dant results with regard to its relationships within the genus. Botany 86 : 697 – 718 . There are three so-called “ orphan ” species that are not con- Cannon , S. B. , L. Sterck , S. Rombauts , S. Sato , F. Cheung , J. Gouzy , sistently resolved as part of any well-supported group based on X. Wang, et al . 2006 . Legume genome evolution viewed through analyses of either trnK/matK or GA3ox1 sequence data, M . ara- the Medicago truncatula and Lotus japonicus genomes. Proceedings bica , M . orbicularis, and M . suffruticosa subsp. suffruticosa of the National Academy of Sciences, USA 103 : 14959 – 14964 . ( Figs. 1, 2 ). There is some precedent for a lack of hypotheses of Dalmadi , A. , P. Kalo , J. Jakab , A. Saskoi , T. Petrovics , G. Deak , and relationships for one of these three species based on morpho- G. B. Kiss . 2008 . Dwarf plants of diploid Medicago sativa carry a logical data; M . orbicularis is in a monotypic section, and no mutation in the gibberellin 3- β -hydroxylase gene. Plant Cell Reports species has been suggested as being closely related to it (Lesins 27 : 1271 – 1279 . Downie , S. R. , D. S. Katz-Downie , E. J. Rogers , H. L. Zujewski , and E. and Lesins, 1979 ; Small and Jomphe, 1989b ). However, M. Small. 1998 . Multiple independent losses of the plastid rpo C1 intron arabica and M . orbicularis are strongly supported sister species in Medicago (Fabaceae) as inferred from phylogenetic analyses of nu- when using ITS/ETS sequence data (Downie et al., 1998; Bena, clear ribosomal DNA internal transcribed spacer sequences. Canadian 2001 ) and in the same monophyletic group using β cop data Journal of Botany 76 : 791 – 803 . ( Maureira-Butler et al., 2008 ). But in analyses using CNGC 5 Doyle , J. J. , and J. Doyle . 1987 . A rapid DNA isolation procedure for small data, M . orbicularis and M . arabica are resolved in two differ- quantities of fresh leaf tissue. Phytochemical Bulletin 19 : 11 – 15 . ent and well-supported groups. Dundas , I. S. , R. M. Nair , and D. C. Verlin . 2006 . First report of Medicago suffruticosa (sect. Medicago ), is an orphan taxon meiotic chromosome number and karyotype analysis of an acces- in our analyses, but is the strongly supported sister species to M . sion of Trigonella balansae (Leguminosae). New Zealand Journal of hybrida in analyses of ETS/ITS data ( Downie et al., 1998 ; Agricultural Research 49 : 55 – 58 . Farris , J. S. , M. K ä llersj ö , A. G. Kluge , and C. Bult . 1995 . Testing Bena, 2001) and part of a polytomy with M . sativa based on signifi cance of incongruence. Cladistics 10 : 315 – 319 . CNGC 5 data, but in another group entirely based on β cop Falistocco , E. , R. Torricelli , and M. Falcinelli . 2002 . Genomic re- analyses (M . hybrida not sampled for either CNGC 5 or β cop ) lationships between Medicago murex Willd. and Medicago lesinsii E. (Maureira-Butler et al., 2008). Medicago suffruticosa and M . Small investigated by in situ hybridization. Theoretical and Applied hybrida produce fertile hybrids and have very similar karyo- Genetics 105 : 829 – 833 . types ( Gillies, 1972 ). Felsenstein , J. 1985 . Confi dence limits on phylogenies: An approach using the bootstrap. Evolution 38 : 783 – 791 . Conclusions— We have shown GA3ox1 to be an informative Gillies , C. B. 1972 . Pachytene chromosomes of perennial Medicago spe- nuclear-encoded molecular marker for phylogenetic analyses cies. III. Unique karyotypes of M. hybrida Trautv. and M. suffruticosa and to provide strong support for a number of groups within Ramond. Hereditas 72 : 309 – 310 . Medicago. Our results suggest a reduction in chromosome Gillies , C. B. 2006 . Pachytene chromosomes of annual Medicago species. number from 2n = 16 to 2n = 14 is likely to have occurred at Medicago Genetic Reports 6: 1 – 15; http:// www.medicago-reports. org/volumes06.html [accessed 7 July 2009]. least four times within the annual species of Medicago sect. Goldblatt , P. 1981 . Cytology and the phylogeny of Leguminosae. In R. Spirocarpos. Reduction in seed number per fruit has also oc- M. Polhill and P. H. Raven [eds.], Advances in legume systematics, curred independently both within Medicago and Trigonella and vol. 2, 427– 463. Royal Botanic Gardens, Kew, UK. thus is not a useful character to defi ne Medicago sect. Lupularia Goldblatt , P. , and D. E. Johnson [eds.]. 1979 . Index to plant chromosome nor the genus Melilotus as a taxon separate from Trigonella . numbers. Missouri Botanical Garden Press, St. Louis, Missouri, USA. July 2010] Steele et al. — Phylogenetic relationships in MEDICAGO 1155

Grant , V. 1981 . Plant speciation. Columbia University Press, New York, Schlarbaum , S. E. , R. J. Rose , E. Small , and L. B. Johnson. 1989 . New York, USA. Karyotypic and chloroplast genomic diversity in Medicago, section Hanelt , P. [ed.]. 2001 . Mansfeld ’ s encyclopedia of agricultural and horti- Lupularia (Leguminosae). Plant Systematics and Evolution 166 : 69 – 78 . cultural crops, vol. 5, 879– 880. Springer-Verlag, Berlin, Germany. Schultz , O. F. 1901 . Monographie der Gattung Melilotus. Botanische Heyn , C. C. 1981 . Trifolieae. In R. M. Polhill and P. H. Raven [eds.], Jahrbücher für Systematik, Pfl anzengeschichte und Pfl anzengeog- Advances in legume systematics, vol. 1, 383– 385. Royal Botanic raphie 29 : 660 – 735 . Gardens, Kew, UK. Small , E. 1986 . A clarifi cation of the Medicago papillosa complex. Hu , J.-M. , M. Lavin , M. F. Wojciechowski , and M. J. Sanderson . 2000 . Canadian Journal of Botany 64 : 2800 – 2806 . Phylogenetic systematics of the tribe Millettieae (Leguminosae) based Small , E. 1987a . A taxonomic study of the “ medicagoid ” Trigonella on chloroplast trnK /matK sequences and its implications for evolu- (Leguminosae). Canadian Journal of Botany 65 : 1199 – 1211 . tionary patterns in Papilionoideae. American Journal of Botany 8 7 : Small , E. 1987b . Generic changes in Trifolieae subtribe Trigonellinae. 418 – 430 . In C. H. Stirton [ed.], Advances in legume systematics, vol. 3, 169– Huber-Morath , A. 1970 . Trigonella . In P. H. Davis [eds.], Flora of 181. Royal Botanical Gardens, Kew, UK. Turkey, vol. 3, 452 – 482. Edinburgh University Press, Edinburgh, UK. Small , E. 1988 . Pollen-ovule patterns in tribe Trifolieae (Leguminosae). Huson , D. H. , and D. Bryant . 2006 . Application of phylogenetic net- Plant Systematics and Evolution 160 : 195 – 205 . works in evolutionary studies. Molecular Biology and Evolution 23 : Small , E. 2010 . Alfalfa and relatives: Evolution and classifi cation of 254 – 267 . Medicago . NRC Research Press, Ottawa, Ontario, Canada. Ingham , J. L. , and J. B. Harborne . 1976 . Phytoalexin induction as a Small , E. , I. J. Bassett , and C. W. Crompton . 1981 . Pollen varia- new dynamic approach to the study of systematic relationships among tion in tribe Trigonelleae (Leguminosae) with special reference to higher plants. Nature 260 : 241 – 243 . Medicago. Pollen et Spores 23 : 295 – 320 . Jurzysta , M. , E. Small , and C. Nozzolillo . 1988 . Hemolysis, Small , E. , and B. S. Brookes . 1984 . Reduction of the geocarpic a synapomorphic discriminator of an expanded genus Medicago Factorovskya to Medicago. Taxon 33 : 622 – 635 . (Leguminosae). Taxon 37 : 354 – 363 . Small , E. , and B. S. Brookes . 1990a . A numerical taxonomic analysis Lesins , K. A. , and I. Lesins . 1979 . Genus Medicago (Leguminosae): A of the Medicago littoralis -M. truncatula complex. Canadian Journal taxogenetic study. Dr. W. Junk, Boston, Massachusetts, USA. of Botany 68 : 1667 – 1674 . Lester , D. L. , J. J. Ross , P. J. Davies , and J. B. Reid . 1997 . Mendel ’ s Small , E. , and B. S. Brookes . 1990b . A taxonomic simplifi cation of stem length gene ( Le ) encodes a gibberellin 3 β -hydroxlyase. Plant Medicago italica. Canadian Journal of Botany 68 : 2103 – 2111 . Cell 9 : 1435 – 1443 . Small , E. , and M. Jomphe . 1989a . A note on Trigonella ( Melilotus ) Lewis , G. , B. Schrire , B. Mackinder , and M. Lock . 2005 . Legumes of bicolor. Canadian Journal of Botany 67 : 1604 – 1606 . the world. Royal Botanic Gardens, Kew, UK. Small , E. , and M. Jomphe . 1989b . A synopsis of the genus Medicago Linnaeus , C. 1753 . Species plantarum Impensis Laurentii Salvii, (Leguminosae). Canadian Journal of Botany 67 : 3260 – 3294 . Stockholm, Sweden. , vol. 2 , 778 – 781 . Small , E. , P. Lassen , and B. Brookes . 1987 . An expanded circum- Martin , D. N. , W. M. Proebsting , and P. Hedden . 1997 . Mendel ’ s scription of Medicago (Leguminosae, Trifolieae) based on explosive dwarfi ng gene: cDNAs from the Le alleles and function of the ex- fl ower tripping. Willdenowia 16 : 415 – 437 . pressed proteins. Proceedings of the National Academy of Sciences, Steele , K. P. , and M. F. Wojciechowski . 2003 . Phylogenetic analyses USA 94 : 8907 – 8911 . of tribes Trifolieae and Vicieae, based on sequences of the plastid Maureira-Butler , I. J. , B. E. Pfeil , A. Muangprom , T. C. Osborn , and gene, matK (Papilionoideae: Leguminosae). In B. B. Klitgaard and J. J. Doyle . 2008 . The reticulate history of Medicago (Fabaceae). A. Bruneau [eds.] Advances in legume systematics, part 10, 355– 370. Systematic Biology 57 : 466 – 482 . Royal Botanic Gardens, Kew, UK. McCoy , T. J. , and E. T. Bingham . 1988 . Cytology and cytogenetics of Steele , K. P. , B. Udpa , D. Chinn , O. Curameng , D. Throckmorton , alfalfa. In A. A. Hanson, D. K. Barnes, and R.R. Hill [eds.], Alfalfa and M. F. Wojciechowski. 1999 . Phylogenetic relationships of tribes and alfalfa improvement. Agronomy Monograph 29: 737 – 776. Trifolieae and Vicieae (Fabaceae). In Proceedings of XVI International Medikus , F. K. 1789 . Philosophische Botanik , vol. 1 , 208 . Erstes Heft. Botanical Congress, St. Louis, Missouri, USA [abstract, page no. unknown]. Von den mannigfaltigen Umh ü llungen der Samen. Neue Hof- und Stevenson , G. A. 1969 . An agronomic and taxonomic review of the Akademische Buchhandling, Mannheim, Germany. genus Melilotus Mill. Canadian Journal of Botany 49 : 1 – 20 . Michaud , R. , W. F. Lehman , and M. D. Rumbaugh . 1988 . World dis- Swofford , D. 2002 . PAUP * : Phylogenetic analysis using parsi- tribution and historical development. In A. A. Hanson, D. K. Barnes, mony (*and other methods), version 4.0b10. Sinauer, Sunderland, and R. R. Hill [eds.], Alfalfa and alfalfa improvement. Agronomy Massachusetts, USA. Monograph 29: 25 – 91. Thompson , J. D. , T. J. Gibson , F. Plewniak , F. Jeanmougin , and D. Miller , P. 1754 . The gardener’ s dictionary, abridged, 4th ed., no. 2. John G. Higgins . 1997 . The ClustalX windows interface: Flexible strate- and James Rivington, London, UK. gies for multiple sequence alignment aided by quality analysis tools. Nylander , J. A. A. 2004 . MrModeltest v2. Computer program distributed Nucleic Acids Research 25 : 4876 – 4882 . by the author, website http://www.abc.se/~nylander/ . Evolutionary Wojciechowski , M. F. , M. Lavin , and M. J. Sanderson . 2004 . A Biology Centre, Uppsala University, Uppsala, Sweden. phylogeny of legumes (Leguminosae) based on analysis of the plas- Rechinger , K. H. 1984 . Trigonella. In K. H. Rechinger [ed.], Flora tid matK resolves many well-supported subclades within the family. iranica, Papilionaceae, vol. 2, 207– 253. Akademishce Druch-und American Journal of Botany 91 : 1846 – 1862 . Verlagsanstalt Dr. P. Struzl, Graz, Austria. Wojciechowski , M. F. , M. J. Sanderson , K. P. Steele , and A. Liston . Ronquist , F. , and J. P. Huelsenbeck . 2003 . MrBayes 3.0: Bayesian phy- 2000 . Molecular phylogeny of the ‘ ‘ temperate herbaceous tribes ’ ’ of logenetic inference using mixed models. Bioinformatics 19 : 1572 – 1574 . papilionoid legumes: A supertree approach. In P. S. Herendeen and Rosato , M. A. , M. Castro , and J. A. Rosselo . 2008 . Relationships of A. Bruneau [eds.], Advances in legume systematics, part 9, 277– 298. the woody Medicago species (section Dendrotelis ) assessed by mo- Royal Botanic Garden, Kew, UK. lecular cytogenetic analyses. Annals of Botany 102 : 15 – 22 . Yakovlev , G. P., A. K. Sytin A, and Y. R. Roskov. 1996 . Legumes of Rumbaugh , M. D. 1990 . Special purpose forage legumes. In J. Janick Northern Eurasia. A checklist. Royal Botanic Gardens, Kew, UK. and J. E. Simon [eds.], Advances in new crops, 183– 190. Timber Zohary , M. 1972 . Trigonella. In Flora palestina, part 2, 123 – 136. Israel Press, Portland, Oregon, USA. Academy of Sciences and Humanities, Jerusalem, Israel.