BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors, nonprofit publishers, academic institutions, researchlibraries, and research funders in the common goal of maximizing access to critical research.

Molecular Phylogeny of Crested Iris Based on Five Plastid Markers (Iridaceae)Author(s): Jinyan Guo and Carol A. WilsonSource: Systematic Botany, 38(4):987-995. 2013.Published By: The American Society of Plant TaxonomistsURL: http://www.bioone.org/doi/full/10.1600/036364413X674724

BioOne (www.bioone.org) is a nonprofit, online aggregation of core research in the biological, ecological, andenvironmental sciences. BioOne provides a sustainable online platform for over 170 journals and books publishedby nonprofit societies, associations, museums, institutions, and presses.

Your use of this PDF, the BioOne Web site, and all posted and associated content indicates your acceptance ofBioOne’s Terms of Use, available at www.bioone.org/page/terms_of_use.

Usage of BioOne content is strictly limited to personal, educational, and non-commercial use. Commercial inquiriesor rights and permissions requests should be directed to the individual publisher as copyright holder.

Systematic Botany (2013), 38(4): pp. 987–995© Copyright 2013 by the American Society of Plant TaxonomistsDOI 10.1600/036364413X674724

Molecular Phylogeny of Crested Iris Based on Five Plastid Markers (Iridaceae)

Jinyan Guo1,2 and Carol A. Wilson1

1Rancho Santa Ana Botanic Garden, 1500 North College Avenue, Claremont, California 91711, U. S. A.2Author for correspondence (j.yanguo@gmail.com)

Communicating Editor: Chrissen E. C. Gemmill

Abstract—About one hundred years ago, Dykes noticed an innovative feature in Iris, a raised central ridge along the midvein of the sepal,that he called a crest. Molecular phylogenetic and ancestral state reconstruction studies suggested that the sepal crest is a homoplasticcharacter, even though the majority of the crested species form a monophyletic group. We investigated the putative multiple origins of sepalcrests in Iris and relationships among crested species via comprehensive sampling of crested species in Iris using five plastid markers. Weemployed maximum parsimony, maximum likelihood, and Bayesian inference in reconstructing relationships. Our data analyses resolved alarge core-crested clade along with four other independent lineages that also have crested species. Within the core-crested clade, four highlysupported major clades were identified comprising species from subgen. Nepalensis and Scorpiris and some species from subgenus Limnirissection Lophiris sensu Mathew. However, relationships among these four major clades were not fully resolved. Six species from sectionLophiris represent four additional origins of the sepal crest.

Keywords—Sepal crest, subgenus Crossiris, subgenus Limniris, subgenus Nepalensis, subgenus Scorpiris.

Iris L., with about 300 taxa distributed throughout theNorthern Hemisphere, is the largest genus of Iridaceae,which is estimated to consist of more than 2,000 species(Goldblatt and Manning 2008). Iris is unique within Iridaceaebecause it has features such as petaloid styles, a floral tubewith basal nectary tissue, and distinct perianth whorls ofpetals and sepals (Wilson 2004). Although the overall floralmorphology is generally conserved across lineages in Iris(except I. domestica (L.) Goldblatt & Mabb., a species that hastwo perianth whorls and lacks obvious petaloid stylebranches), the sepals can be variously elaborated and typicallyhave raised midveins, beards (multicellular hairs), crests(central ridges or laminar outgrowths), signal patches, and/or pollination guides that maybe reproductively significant.

Species that are described as bearing sepal crests, whetherthese crests are subtle or obvious, are referred to as crestedspecies in this study. Sepal crests, although not common acrossangiosperms, occur in several subgenera in Iris (Table 1). Thesubgenera with crested species are morphologically diversewith geophytic organs that are rhizomes, bulbs, or tubers,leaves that are unifacial or dorsiventral, and petals that arefully developed or greatly reduced. Iris sepal crests are alsomorphologically diverse, ranging from a central ridge to anelaborately colored and/or fringed perpendicular outgrowth(Guo and Wilson unpubl. data).

Sepal crests have played an important role in the taxo-nomic delineation of higher ranks in Iris (Table 1). The firstcomprehensive systematic study of Iris (Dykes 1913) assignedthe approximately 140 described Iris species to 12 sections,with crested species placed in three sections. Six rhizomatousspecies were included in sect. Evansia, which Dykes desig-nated as the crested section. He assigned crested species withfleshy rootstocks to sect. Nepalensis, and crested and non-crested species with bulbs, dorsiventral leaves, and reducedpetals to section Juno. Lawrence (1953) subordinated sect.Evansia to the subsectional level and elevated Dykes’ sect.Nepalensis to subgen. Nepalensis. Lawrence (1953) also ele-vated sect. Juno to the subgeneric rank and used the nameScorpiris. While keeping subgen. Nepalensis at the same rankas Lawrence, Rodionenko (1984) elevated subgen. Scorpiris togenus Juno (using Dykes’ earlier name) and elevated subsect.Evansia to subgen. Crossiris. Mathew (1989) adopted Lawrence’s(1953) taxonomic circumscriptions of subgen. Nepalensis and

Scorpiris and elevated Lawrence’s subsect. Evansia to the sec-tional level, placing the section within subgen. Limniris(Tausch) Spach (rhizomatous, beardless species), and usedthe name sect. Lophiris. Mathew (1989) considered sect.Lophiris the crested group based on the shared character of asepal crest, although he recognized that some species insubgen. Nepalensis and most species in subgen. Scorpiris alsohave sepal crests (Table 1). In this study, we used Mathew’s(1989) classification while incorporating Wilson’s (2011) tax-onomic suggestions where new subgeneric names apply toclades recovered.Previous studies using morphological and molecular data

suggested that the sepal crest is homoplastic in Iris (Wilson2004, 2006, 2011) because sect. Lophiris sensu Mathewwas notresolved as monophyletic. Subgen. Nepalensis (Wilson 2011)and Scorpiris (Wilson 2011; Ikinci et al. 2011) were resolved asmonophyletic and species from these two subgenera andmost of the Asian species from sect. Lophiris were resolvedin one clade (Wilson 2011). Each of the previous studies sam-pled a limited number of crested species from sect. Lophirisand subgen. Nepalensis.By sampling crested species in subgen. Nepalensis and

Scorpiris, sect. Lophiris, and species representing other majorclades across Iris, we investigated the following questions:1) What are the phylogenetic relationships among speciesfrom subgen. Nepalensis, subgen. Scorpiris, and the Asian spe-cies from sect. Lophiris? 2) Are Asian species from sect.Lophiris monophyletic? 3) How many times has the sepalcrest originated in Iris?

Materials and Methods

Taxon Sampling—Taxon sampling focused on the crested species inIris. We sampled all the crested species in subgen. Nepalensis and sect.Lophiris, and approximately 50% of the species in the monophyleticsubgen. Scorpiris. We did not comprehensively sample subgen. Scorpirisbecause a phylogenetic study of the subgenus was underway (M. Chase,pers. comm.) and has subsequently been published (Ikinci et al. 2011).Subgen. Crossiris sensu Wilson, Nepalensis, and Scorpiris (Table 1) wereresolved in one clade in the Wilson (2011) study that we considered thecore-crested clade and the ingroup for this study (36 taxa sampled). The17 outgroup taxa chosen for this study represent each of the major cladesresolved in Iris (Wilson 2011) with an emphasis on crested species thatwere resolved outside of the core-crested clade in the Wilson study.Our study added 22 ingroup taxa (expected to be resolved within the

987

core-crested clade) and two outgroup taxa that were not included in theWilson (2011) study. Sampled species, voucher information, and GenBankaccession numbers are listed in Appendix 1.

DNA Extraction, Amplification, and Sequencing—Genomic DNA wasextracted from silica-dried leaf or bract material using a modified CTABmethod (Doyle and Doyle 1987) with modifications including a RNasetreatment and an ethanol precipitation with ammonium acetate after theisopropanol precipitation.

Five plastid markers were used. Sequences of the matK gene with itsflanking trnK introns (matK/trnK) and trnL–trnF intergenic spacer regionwere generated for all taxa except 28 where the matK/trnK and 10 wherethe trnL–trnF regions were previously sequenced by Wilson (2009, 2011).Each of these two regions was amplified in 25 ml PCR using a PTC-100thermal cycler (MJ Research Inc., Watertown, Massachusetts). The follow-ing reaction mix for the matK/trnK region was used: 1.56 ml dH2O, 1.25 ml50% glycerol, 2.5 ml 25 mM MgCl2, 2.5 ml buffer, 2.5 ml 10 mM dNTP, 0.9 ml10 mM of each primer, and 2.25 units of GoTaq Flexi DNA polymerase(Promega, Madison, Wisconsin). The reaction conditions for the matK/trnK region were: 97�C for 1 min; 40 cycles of 97�C for 10 s, 50�C for 1 min,72�C for 20 s; 72�C for 4 min. The matK/trnK region was amplified in twoseparate reactions. The 50 segment was amplified using the primer pair3914m (Johnson and Soltis 1994) and 1360ir (Wilson 2009). The 30 segmentwas amplified using the primer pair 1176i (Wilson 2004) and trnK2r(Johnson and Soltis 1994). The reaction mix for the trnL–trnF region wassimilar to that of the matK/trnK region except that 1.25 ml DMSO wasadded to the reaction mix, the amounts of dH2O, glycerol, and bufferwere reduced to 1.41 ml, 0.5 ml, and 2.1 ml, respectively, and the amountof Taq was increased to 3 units. The reaction conditions for the trnL–trnFregion were identical to that of the matK/trnK region except that anannealing temperature of 48�C was used. The trnL–trnF region wasamplified using the primer pair c (Taberlet et al. 1991) and trnf-10 (Wilsonand Calvin 2006).

In addition to the matK/trnK and trnL–trnF regions, psbJ–petA, rpl32–trnL, and rpoB–trnC were amplified for all study taxa using the primersdeveloped by Shaw et al. (2005, 2007). Due to high poly-A and -T sequenceregions that interfered with sequencing, the rpoB–trnC region was ampli-fied in two separate reactions. Two internal primers were developed:trnC800r (50- tcgctcttatgcacatatcc -30) and rpoB600f (50- tcgagttatccgtatatcatacc -30). The 50 segment of the rpoB–trnC region was amplified usingthe primer pair of rpoB and trnC800r and the 30 segment was amplifiedusing the primer pair of rpoB600f and trnC. The reaction mix for the psbJ–petA, rpl32–trnL, and rpoB–trnC regions was the same as that used for thetrnL–trnF region except MgCl2 was decreased to 2 ml and dH2O wasincreased to 1.91 ml. The cycling conditions for all three regions were thesame as those used for the trnL–trnF region.

The resulting amplification products were purified using polyethyleneglycol (PEG) precipitation followed by BigDye Terminator (v. 3.1 AppliedBiosystems, Foster City, California) cycle sequencing. The following reac-tion mix was used: 12.4 ml dH2O, 1 ml DMSO, 1.5 ml BigDye, 3.5 ml BigDyebuffer, approximately 60 ng template, and 1.6 ml primer (1 mm). Cyclesequencing products were purified using Sephadex columns (AmershamBiosciences, Piscataway, New Jersey) and sequenced on an AppliedBiosystems 3130 XL automated sequencer at the Rancho Santa AnaBotanic Garden. The matK/trnK region was sequenced using internal andamplification primers from Wilson (2011). Sequencing reactions for thetrnL–trnF, psbJ–petA, rpl32–trnL, and rpoB–trnC regions used the ampli-fication primers. Sequences were edited and assembled in Sequencherv. 3.1.1 (Gene Codes Corporation, Ann Arbor, Michigan) and the datamatrixes were manually aligned using Se-Al v. 2.0a11 (Rambaut 2002).

Data Analyses—Sequence data of the five plastid markers were ana-lyzed using maximum parsimony (MP; Swofford et al. 1996), maximum

likelihood (ML; Felsenstein 1981), and Bayesian inference (BI; Huelsenbeckand Ronquist 2001). Separate and combined datasets of 53-taxa (36 ingrouptaxa and 17 outgroup taxa) without coded indels and 40-taxa (36 ingrouptaxa and four outgroup taxa from the sister clade) with coded indels(nucleotide insertions or deletions) were assembled. Indels were not codedin the 53-taxa dataset because of ambiguities in gap length and locationamong some outgroup taxa. PAUP* v. 4.0b10 (Swofford 2002) was used tocalculate the number of variable sites and parsimony informative charac-ters, consistency index (CI), retention index (RI), and rescaled consistencyindex (RC) for each plastid region and the combined datasets (Table 2).

ML and BI analyses were used for the 53-taxa combined dataset of fiveplastid regions. This dataset was analyzed using PhyML 3.0 (Guindonet al. 2010) and the general time reversible (GTR) substitution model. Theproportion of invariable sites (0.3649) and gamma shape parameter (0.8785)were set to fixed according to the values suggested by Akaike InformationCriterion (AIC) in MODELTEST v. 3.06 (Posada and Crandall 1998). Bothnearest neighbor interchanges (NNIs) and subtree pruning and regrafting(SPR) were used for tree improvement, the number of random startingtrees was set to 10, and both topology and branch lengths were optimized.One hundred bootstrap (BS) replicates (Felsenstein 1985) were performed.BS values ³ 90 and ³ 80 were considered high and moderate, respectively.

Partitioned BI analyses were performed inMrBayes v. 3.1.2 (Huelsenbeckand Ronquist 2003). The 53-taxa combined dataset was partitioned corre-sponding to the five plastid markers and all five partitions were allowed toevolve at different rates. For BI analyses, the number of substitution rateswas set to six, gamma distribution was approximated using four rate cate-gories, and a proportion of sites were assumed to be invariable. Ten millionMarkov Chain Monte Carlo (MCMC) generations were performed usingfour Markov chains with a sampling frequency of every 2,000 generations.The final average standard deviation of split frequencies for each analysiswas < 0.004. Of the 5,000 tree samples, 1,250 were discarded as burn-in foreach run to calculate the posterior probability distribution. Bayesian poste-rior probabilities (PP) £ 0.8 were considered non-significant; PP ³ 0.99 and³ 0.9 were considered high and moderate, respectively.

Table 2. Characteristics of each dataset including aligned length(L), percentage of missing data (MD), number of indels coded (IND),number and percentage of variable sites (Var), and results of MP analysesincluding number and percentage of parsimony informative characters(PI), consistency index (CI), retention index (RI), and rescaled consistencyindex (RC). One 14 bp reversal that occurred in the psbJ–petA region wastreated as an indel.

L %MD IND Var (%) PI (%) CI RI RC

53-taxamatK/trnK 2,685 0.21 0 727 (27.0) 305 (11.4) 0.88 0.84 0.74trnL–trnF 912 0.04 0 160 (17.5) 80 (8.8) 0.83 0.86 0.71rpl32–trnL 1,456 0.02 0 320 (22.0) 156 (10.7) 0.79 0.84 0.66psbJ–petA 914 2.54 0 863 (94.4) 111 (12.1) 0.98 0.85 0.83rpoB–trnC 1,343 3.29 0 958 (71.3) 170 (12.7) 0.98 0.87 0.86Combined 7,310 1.01 0 3,028 (41.4) 822 (11.2) 0.95 0.84 0.8040-taxamatK/trnK 2,626 0.24 11 430 (16.4) 176 (6.7) 0.95 0.92 0.88trnL–trnF 853 0.05 8 83 (9.7) 45 (5.3) 0.90 0.93 0.84rpl32–trnL 1,352 0.02 19 171 (12.6) 100 (7.4) 0.84 0.90 0.75psbJ–petA 878 1.36 11 443 (50.5) 80 (9.1) 0.98 0.93 0.92rpoB–trnC 1,247 2.86 12 890 (71.4) 106 (8.5) 0.99 0.94 0.94Combined 6,956 0.79 61 2,017 (29.0) 507 (7.3) 0.97 0.91 0.89

Table 1. Taxonomic history, morphology, and distribution in Asia and North America (N Am) of Iris groups that have species with sepal crests.The number of species with sepal crests and the total number of species in each group are given in parentheses.

Rhizomes, unifacial leaves;Asia, N Am; (13/13)

Fleshy rootstocks, unifacialleaves; Asia; (3/4)

Bulbs, dorsiventral leaves,reduced petals; Asia; (40/55)

Dykes (1913) sect. Evansia Benth. sect. Nepalensis Dykes sect. Juno (Tratt.) Benth.Lawrence (1953) subgen. Limniris subsect. Evansia (Benth.)

G. H. M. Lawr.subgen. Nepalensis (Dykes)

G. H. M. Lawr.subgen. Scorpiris Spach

Rodionenko (1984) subgen. Crossiris Spach subgen. Nepalensis genus Juno Tratt.Mathew (1989) subgen. Limniris sect. Lophiris Tausch subgen. Nepalensis subgen. ScorpirisWilson (2011) subgen. Crossiris, N Am; subgen. Lophiris (Tausch)

C. A. Wilson, Asia; unamed clade, N Am, Asiasubgen. Nepalensis subgen. Scorpiris

988 SYSTEMATIC BOTANY [Volume 38

The MP and BI analyses were used for the 40-taxa combined dataset.The 40-taxa combined dataset included indel in addition to sequence datato further resolve phylogenetic relationships within the core-crested clade.Unambiguous indels shared by two or more species were recorded aspresent or absent (1 or 0). Poly-A, T, C, and G regions and autapomorphieswere ignored. We found a 14 reverse-complement from the 5593–5620 bpregion in 21 species, which was coded as an additional character (Table 1).The MP analyses of the 40-taxa combined dataset were performed usingPAUP* v. 4.0b10 (Swofford 2002). Characters were treated as unorderedand were equally weighted. Gaps were treated as missing data. Themost parsimonious trees were obtained through heuristic searches using1,000 random stepwise-addition replicates with four trees held at eachstep. Tree-bisection-reconnection (TBR) was used as the branch-swappingalgorithm and the option of saving multiple trees (MULTREES) wasemployed. When multiple starting trees existed, only the best tree wasswapped on. BS analyses (Felsenstein 1985) were also performed in PAUP*using the full heuristic mode. For the BS consensus tree, only groups withfrequency ³ 50% were retained. The same search settings used in the MPanalyses were used in BS analyses and 1,000 BS replicates were performed.BS values ³ 90 and ³ 70 were considered high and moderate, respectively.

The 40-taxa combined dataset was also analyzed using partitioned BI.The dataset was partitioned into six regions representing sequence datafrom the five plastid regions and the indel data. Parameters used were thesame as those used for the 53-taxa dataset except the number of genera-tions was reduced to eight million because convergence was reached(average standard deviation of split frequencies for each analysis was< 0.009). Of the 4,000 tree samples, 1,000 were discarded as burn-in foreach run. Bayesian PP £ 0.8 were considered non-significant; PP ³ 0.99and ³ 0.9 were considered high and moderate, respectively.

Three replicates each of the MP, ML, and BI analyses using the com-bined 53- and 40-taxa datasets were performed to ensure topologies andsupport values were consistent.

Results

Data and Tree Characteristics—The 53-taxa combineddataset includes the entire matK/trnK, trnL–trnF, and rpl32–trnL regions for all accessions and the entire psbJ–petA andrpoB–trnC regions for 51 and 45 taxa, respectively. The 40-taxacombined dataset includes the entire psbJ–petA and rpoB–trnC regions for 39 and 34 taxa, respectively. In both datasetsincomplete sequences of the psbJ–petA and rpoB–trnC regionswere included for the remaining taxa. Characteristics of theindividual partitions and combined data for the 53- and40-taxa datasets, including the % of missing data, are givenin Table 2. Among the five partitions, psbJ–petA and rpoB–trnC have the highest percentages of variable sites, slightlyhigher percentages of parsimony informative characters, andhigher CI and RC values.

For the BI analyses of the 53-taxa combined dataset, theproportion of invariable sites varies between 0.227553 and0.557924 and the gamma distribution shape parameter variesbetween 0.053286 and 1.295313 among the five partitions.The final ML analyses found an optimal tree with a log like-lihood (ln) –24,804. The 53-taxa phylogeny inferred by MLwas congruent with the BI consensus tree for all branchesthat had moderate to high support. Replicates of the MLanalysis resolved the same tree topology as did replicates ofthe BI analysis. The ML tree is given in Fig. 1 with BS and PPvalues from the final analyses shown above branches.

MP analyses of the 40-taxa combined dataset resulted innine most parsimonious trees each with a length of 1,131steps. One of the most parsimonious trees is given in Fig. 2with BS and PP values from the final analyses shown abovebranches. The topology of the MP tree in Fig. 2 is congruentwith the BI consensus tree inferred from the 40-taxa com-bined dataset and the other eight most parsimonious treesexcept for relationships among terminal branches within sub-

gen. Scorpiris that have low to moderate support. Replicatesof the MP analysis resolved the same tree topology as didreplicates of the BI analysis.Phylogenetic Relationships of Crested Species—For the

53-taxa phylogeny, a clade (PP = 1.00, BS = 100) was resolved(Fig. 1), that includes seven of the 13 species from sect.Lophiris sensu Mathew and all sampled taxa from subgen.Nepalensis and Scorpiris. Because most of the species in thisclade have a sepal crest and this clade includes all except sixof the crested species in Iris, we consider this the core-crestedclade. Six crested taxa from sect. Lophiris were resolved inclades outside of the core-crested clade and represent fouradditional independent origins of sepal crests (I. gracilipes +I. tenuis, I. cristata + I. lacustris, I. speculatrix, and I. proanthavar. valida). Because the character state of the common ances-tor of the core-crested clade is unknown, the core-crestedclade represents at least one gain of a sepal crest in Iris. Thus,based on our findings there are at least five independentorigins of the sepal crest in Iris.Phylogenetic Relationships within the Core-crested Clade—

Relationships within the core-crested clade are mostly con-gruent for all analyses of both datasets. Both MP and BIanalyses of the 40-taxa combined dataset resolved four highlysupported clades within the core-crested clade (Fig. 2);subgen. Scorpiris (PP = 1.00, BS = 100) comprising the 24 sam-pled taxa from this subgenus; the Japonica clade (PP = 1.00,BS = 100) including five Asian species from sect. Lophiris(I. confusa, I. formosana, I. japonica, I. milesii, and I. wattii);subgen. Nepalensis (PP = 1.00, BS = 100) comprising the threesampled species from this subgenus (I. barbatula, I. collettii,and I. decora) and one Asian species from sect. Lophiris(I. latistyla); and the Tectorum clade (PP = 1.00, BS = 100)comprising one Asian species from sect. Lophiris (I. tectorumand I. t. f. alba). The Japonica clade is sister to subgen.Nepalensis with high (PP = 1.00) to moderate (BS = 86) sup-port while together they are sister to subgen. Scorpiris withlow support (PP = 0.63 and BS = 51). The placement of theTectorum clade was unresolved in the BI consensus treewhile resolved as sister to all other species in the core-crestedclade in the MP tree (Fig. 2).For the 40-taxa combined dataset using MP and BI ana-

lyses, relationships among taxa within the Japonica clade,subgen. Nepalensis, and the Tectorum clade were completelyresolved while relationships among taxa within subgen.Scorpiriswere not (Fig. 2). Two major subclades were resolvedin the Japonica clade. Iris formosana and I. japonica wereresolved as sister (PP = 1.00, BS = 100), and this clade is sisterto I. confusa (PP = 1.00, BS = 96). The clade with these threespecies was sister to a second clade (PP = 1.00, BS = 100)comprising I. milesii and I. wattii (PP = 1.00, BS = 100). Insubgen. Nepalensis, two subclades were also resolved withhigh support (PP = 1.00, BS =100) for their sister relationship.Iris barbatula and I. collettii were resolved in one of the clades(PP = 1.00, BS = 100) and I. decora was resolved as sister toI. latistyla (PP = 1.00, BS = 100) in the other clade. In subgen.Scorpiris, four major subclades were resolved (names aregiven to facilitate discussion): 1) the Rosenbachiana clade,comprising I. rosenbachiana, is sister to all other species insubgen. Scorpiris; 2) the Bucharica clade (PP = 1.00, BS = 100)comprising I. bucharica and I. vicaria; 3) the Aucheri clade(PP = 0.99, BS = 74) comprising 10 taxa; and 4) the Magnificaclade (PP = 1.00, BS = 89) comprising 10 species. The Auchericlade was resolved as sister to the Magnifica clade (PP = 1.00,

2013] GUO AND WILSON: CRESTED IRIS PHYLOGENY 989

Fig. 1. Maximum likelihood tree of Iris species based onmatK/trnK, trnL–trnF, rpl32–trnL, psbJ–petA, and rpoB–trnC sequence data using the 53-taxacombined dataset. Numbers above branches are bootstrap/Bayesian posterior probability values. The core-crested clade is within the grey box. Initialsafter the species names are used where multiple accessions were sampled for one species. Rhizomatous crested species in sect. Lophiris sensu Mathew areindicated by arrowheads and those with distributions in North America (N Am) are also indicated.

990 SYSTEMATIC BOTANY [Volume 38

Fig. 2. One of nine most parsimonious trees of Iris species based on matK/trnK, trnL–trnF, rpl32–trnL, psbJ–petA, and rpoB–trnC sequence and indeldata using the 40-taxa combined dataset. Numbers above branches are bootstrap/Bayesian posterior probability values. Major clades are within boxesand subclades in subgen. Scorpiris are indicated on the right. Initials after the species names were used where multiple accessions were sampled for onespecies. Subgen. Crossiris (Wilson 2011) is indicated.

2013] GUO AND WILSON: CRESTED IRIS PHYLOGENY 991

BS = 98), with the Bucharica (PP = 1.00, BS = 100) and theRosenbachiana (PP = 1.00, BS = 100) clades as successivesisters. These results are congruent with relationships resultingfrom ML and BI analyses of the 53-taxa combined datasetexcept for the position of the Tectorum clade (Figs. 1, 2).Within the core-crested clade, two accessions each were

included for four species, I. tectorum, I. japonica, I. bucharica,and I. caucasica (Fig. 2). Iris tectorum and I. japonica wereresolved as monophyletic in all analyses. The monophyly ofI. bucharica was not supported because its two accessionswere in a clade with I. vicarica without resolution of relation-ships among the three accessions. In both analyses, I. caucasicawas not monophyletic.

Discussion

More than 70% of species from subgen. Scorpiris, a groupcharacterized by bulbs, and 80% of species from subgen.Nepalensis, a group characterized by tuberous roots, and allof the rhizomatous species historically placed in sect. Lophirissensu Mathew have sepal crests. Previous phylogenetic stud-ies suggested that sect. Lophiris is polyphyletic while subgen.Nepalensis and Scorpiris are monophyletic (Wilson 2004,2009, 2011; Ikinci et al. 2011) and that species from the Asiansubgen. Nepalensis and Scorpiris, and some Asian species fromsect. Lophiris form a clade (Wilson 2011). Through extensivesampling of crested species, we resolve a core-crested cladecomprising subgen. Nepalensis and Scorpiris, and seven of the13 Asian rhizomatous crested species and for the first timeidentify the species comprising this clade.Rhizomatous Crested Species—Our results show that

approximately half of the rhizomatous crested species are inthe core-crested clade while other species are in at leastfour independent lineages outside of the core-crested clade(Fig. 1). These results differ from traditional classifications ofrhizomatous crested species. Mathew (1989) placed the rhi-zomatous crested species known at that time into one groupsect. Lophiris while Rodionenko (1984) assigned all of thespecies except I. tenuis to subgen. Crossiris. It is likely thatRodionenko was not aware of the presence of a sepal crestin I. tenuis.In our study, the three North American species (I. cristata,

I. lacustris, and I. tenuis) and three Asian crested species(I. gracilipes, I. speculatrix, and I. proantha var. valida) areresolved outside of the core-crested clade (Fig. 1), which iscongruent with the Wilson (2011) study except that our studyincluded I. speculatrix, a species not included in Wilson’sstudy. Our study resolves seven Asian rhizomatous crestedspecies within the core-crested clade (Fig. 1). Relationshipsamong these species are discussed below.Phylogenetic Relationships Among Major Lineages within

the Core-Crested Clade—Four major clades, subgen.Nepalensis,subgen. Scorpiris, the Japonica clade, and the Tectorum clade,form the core-crested clade although relationships amongthese four lineages are not fully resolved (Figs. 1, 2). The closerelationships among these clades are a new finding that hasnot been suggested in previous classifications (Dykes 1913;Lawrence 1953; Rodionenko 1984; Mathew 1989). Our ana-lyses suggest that I. tectorum is sister to all other species in thecore-crested clade, although this relationship is not sup-ported by PP or BS (Fig. 2). Rodionenko (1984) assignedI. tectorum (Tectorum clade) and I. milesii (Japonica clade) toser. Tectores Rodion. Waddick and Zhao (1992) agreed with

Rodionenko and cited their shared flower shape as a char-acter uniting the two species. Species in the Tectorum andJaponica clades share other morphological characters suchas an inflorescence with several to many branches and ellip-tical pollen with a wide longitudinal furrow and reticulateexine (Guo and Wilson unpubl. data). However, the exinepattern of the pollen grain of I. japonica is comparativelysmall-celled and uneven while that of I. tectorum is large-celled and smooth with nipple-like projections (Guo andWilson unpubl. data).

Iris tectorum has several morphological characters thatare similar to species in a clade comprising subgen. Irisand Pardanthopsis. Our study included I. dichotoma (subgen.Pardanthopsis) and I. germanica var. florentina, I. subbiflora, andI. sari (subgen. Iris) and these species were resolved as thesister lineage to the core-crested clade (Fig. 1). Iris tectorumand I. dichotoma both have pollen grains with two longitudi-nal folds (Rodionenko 1984). Furthermore, the stout horizon-tal rhizome that is formed near the soil surface of I. tectorumis shared with most species in subgen. Iris and Pardanthopsisand is rare within the core-crested clade (Guo and Wilsonunpubl. data). Finally, I. tectorum has sepal crests that aredistinct from other sepal crests in the core-crested clade. Thesepal crests of this species are prominent, dissected, and haveunique anatomical features (Guo and Wilson unpubl. data).The morphology of I. tectorum is congruent with the unsup-ported results of this study that suggest the species may beearly diverging in the core-crested clade (Fig. 2).

Dykes (1913) suggested that species in subgen. Nepalensismight be related to species in subgen. Scorpiris based ontheir shared character of thick roots during the restingstate. In contrast, Rodionenko (1984) recognized subgen.Scorpiris at the generic level using the name Juno whileleaving subgen. Nepalensis in Iris. Our study resolves bothsubgen. Nepalensis and Scorpiris in the core-crested cladebut not as sister clades (Figs. 1, 2). With moderate support(PP = 0.94, BP = 0.84), Wilson (2011) suggested a sisterrelationship for subgen. Crossiris (which corresponds toour Japonica clade) and Scorpiris. Our analyses insteadresolve, with higher support (PP = 1.0, BS = 0.86), the sisterrelationship of subgen. Nepalensis and the Japonica clade.Our data resolves the Japonica clade + subgen. Nepalensisclade as sister to subgen. Scorpiris although support for thisrelationship is low.

Indel information in the MP and BI analyses for the 40-taxacombined dataset improved resolution among species in thecore-crested clade compared to the analyses for the 53-taxacombined dataset that lacked indel information (Figs. 1, 2).Even though not all relationships within the core-crestedclade have support, the resolution of four highly supportedclades and the well supported sister relationship of subgen.Nepalensis and the Japonica clade are important steps towardsunderstanding phylogenetic relationships within the core-crested clade.

Phylogenetic Relationships within Major Clades of theCore-Crested Clade—Two rhizomatous taxa, I. tectorum andI. t. f. alba, form the Tectorum clade (Fig. 2), which corre-sponds to Rodionenko’s (1984) subgen. Crossiris sect. Crossirisser. Tectores except Rodionenko included I. milesii within ser.Tectores. This study resolves I. milesii in the Japonica clade(Fig. 2). Iris tectorum f. alba, is similar to the typical formexcept that it is less robust, has white flowers instead ofviolet, and has yellow-mottled sepals. Iris tectorum f. alba and

992 SYSTEMATIC BOTANY [Volume 38

the typical form have similar pollen morphologies, rhizomes,and sepal crests (Guo and Wilson unpubl. data).

Subgenus Nepalensis is characterized by a geophytic organthat lacks rhizomes and bulbs, and instead has fleshy spin-dle-shaped tuberous roots, a character that separates thisgroup from all other irises (Dykes 1913; Lawrence 1953;Rodionenko 1984; Mathew 1989) (Table 1). Subgen.Nepalensishas two well-supported clades, I. barbatula + I. collettii andI. latistyla + I. decora (Fig. 2). Iris collettii has been placed insynonomy with I. decora by some researchers (Dykes 1913;Lawrence 1953). Mathew (1989) considered I. collettii andI. decora similar but distinct because I. collettii has a stemlesshabit whereas I. decora has a floral stem, and, the former hassmaller flowers. In our study, these two species are in twodifferent clades within subgen. Nepalensis. Iris latistyla, a spe-cies formerly included in the rhizomatous crested group(Waddick and Zhao 1992; BIS 1997), is nested in subgen.Nepalensis (Fig. 2). Zhao (1980) described I. latistyla and placedthe species in Rodionenko’s subgen. Crossiris (Waddick andZhao 1992) because although it has a sepal crest it lacks theshort tuberous roots that are found in subgen. Nepalensisspecies and the bulbs present in subgen. Scorpiris species.However, Zhao (1980) noted that I. latistyla is similar toI. decora except for its root morphology. The roots of I. latistylaare described as fat and fleshy (Waddick and Zhao 1992),which is different from the slender roots of all other rhizoma-tous crested species (Guo andWilson unpubl. data). We haveobserved that the relatively long and radially enlarged rootsof I. latistyla are intermediate in morphology between thetuberous roots found in subgen. Nepalensis and the long slen-der roots found in species of the Tectorum and Japonicaclades. The flowers of I. collettii, I. decora, and I. latistylaclosely resemble each other in coloration patterns and sepalcrest morphology (Guo andWilson unpubl. data). Iris staintonii,a Nepalese species included in subgen. Nepalensis by Mathew(1989), was not sampled in our study. This species is interest-ing because it has a few small tuberous roots similar to otherspecies in subgen. Nepalensis but also a short rhizome andthin fibrous roots. Descriptions of this species also indicatethat it lacks a sepal crest (Mathew 1989).

Among the seven Asian core-crested species, five of themare resolved in the highly supported Japonica clade (Fig. 2).Within the Japonica clade, we recognize two subclades, onecomprising I. confusa, I. formosana, and I. japonica, and theother I. milesii and I. wattii. Rodionenko (1984) groupedI. japonica, I. wattii, and I. speculatrix in subgen. Crossiris sect.Crossiris ser. Japonicae, and I. tectorum and I. milesii in ser.Tectores (I. confusa and I. formosana were not treated). He alsonoted that I. speculatrix, a species resolved in a lineage out-side of the core-crested clade (Fig. 1), was relatively unknownand acknowledged that it might be premature to place thespecies in ser. Japonicae. Mathew (1989) considered I. formosanaa hybrid or a form of I. japonica, species that are sister in ourstudy. Waddick and Zhao (1992) indicated that I. confusa issimilar to I. japonica in floral and fruit morphologies and toI. wattii in vegetative morphology. The sister relationship ofI. milesii and I. wattii has not been proposed by previous authors.

We consider I. wattii and I. milesii distinct yet closelyrelated species, although we may not have captured all ofthe diversity represented by these two species. We sampledI. wattii for this study from a clone collected near the Chinese-Burmese border and established at the University of CaliforniaBotanical Garden, Berkeley. Because of the geographical dis-

tance (about 500 km) between the collection sites of thisclone and the type specimen of I. wattii, the lack of informa-tion in the original description, and the incomplete typespecimen, we are not certain that our accession representsthe type of I. wattii.The morphology of our I. milesii accession from southwest-

ern China differs somewhat from the description of the typespecimen that was collected from the Kulu or Parbutta val-leys of India (Foster 1883). Our unpublished data indicatethat our accession of I. milesii might represent a new species.Wild collections of I. wattii and I. milesii from northeasternIndia and southwestern China, morphological studies of col-lections from the eastern and western Himalayan regions,and the inclusion of additional accessions in future studiesare needed to clarify the identities and diversity present inboth species.Seven of the 10 Asian rhizomatous species are in the core-

crested clade and are found in three of the four major core-crested clades, subgen. Nepalensis (one species), the Japonicaclade (five species), and the Tectorum clade (one species).Thus, our study shows a much more complicated evolution-ary history of rhizomatous crested species compared to mostprevious hypotheses of relatedness.The monophyly of subgen. Scorpiriswas predicted by early

workers because species in this subgenus have unique fea-tures such as distinctive spherical pollen (Rodionenko 1984),bulbs, dorsiventral leaves, and reduced petals (Table 1). Withinsubgen. Scorpiris, we recover four of the five major cladessuggested by a recent molecular phylogenetic study on thesubgenus (Ikinci et al. 2011). The fifth clade identified byIkinci et al. (2011) has only one species, I. microglossa, whichis not included in our study. The phylogenetic relationshipsamong subgen. Scorpiris species suggested by our study aremostly incongruent with the classification of the subgenus byRodionenko (1984, 1994), where four sections (Physocaulon(Rodion.) B.Mathew & Wendelbo, Acanthospora Rodion., Juno(Tratt.) Benth., and Wendelboa Rodion.) were recognized.Our Rosenbachiana clade corresponds to clade A of Ikinciet al. (2011) and comprises species from Rodionenko’s sect.Physocaulon. Our study resolved I. bucharica and I. vicariatogether in the Bucharica clade and separate from other spe-cies in Rodionenko’s sect. Juno. Our study confirmed thefindings of Ikinci et al. (2011) that sect. Juno is polyphyletic.In our study, the monophyly of I. bucharica is unknown(Fig. 2). Our Aucheri clade corresponds to clade C of Ikinciet al. (2011) and includes I. cycloglossa from Rodionenko’smonotypic sect. Wendelboa and several species from twoother sections, Acanthospora and Juno. One species in thisclade, I. caucasica is not monophyletic because the two sub-species are not sister (Fig. 2). Our Magnifica clade correspondsto clade D of Ikinci et al. (2011) and includes the remainingspecies from Rodionenko’s sect. Juno. Of the four sectionscurrently recognized in subgen. Scorpiris (Rodionenko 1984,1994), only sect. Physocaulon is monophyletic.Although our results are similar to those of Ikinci et al.

(2011), our phylogeny is more resolved and our supportvalues are higher overall. In Ikinci et al. (2011), I. microglossa,clade C (our Aucheri clade), and clade D (our Magnificaclade) formed a trichotomy. Our data strongly support (PP =1.0, BS = 98) the Aucheri and Magnifica clades as sister,which together are sister to the Bucharica clade (PP = 1.0,BS = 100) (Fig. 2). Our study includes a greater number of bpand about twice as many parsimony informative characters

2013] GUO AND WILSON: CRESTED IRIS PHYLOGENY 993

when compared to the Ikinci et al. (2011) study and resolvesall species relationships in subgen. Scorpiris except one tri-chotomy involving I. willmottiana + I. zenaidae, I. kuschakewczii,and I. pseudocapnoides (Fig. 2). However, subgen. Scorpiris isnot comprehensively sampled in our study and some brancheslack support. Further studies should include additional spe-cies, especially I. microglossa, a species resolved on a longbranch in the Ikinci et al. (2011) study, and markers such aslow-copy nuclear regions that may increase resolution atinternal nodes.Crested Lineages Outside of the Core-Crested Clade—Our

study indicates that there are at least five independent ori-gins of sepal crests in Iris (Fig. 1). The core-crested clade isour focus, but we also identify six rhizomatous species rep-resenting four additional independent lineages outside of thecore-crested clade that have been traditionally placed in sect.Lophiris sensu Mathew.Other than a sepal crest, the characteristics of these six

species are so diverse it is not surprising they are in indepen-dent lineages. Two crested species, I. gracilipes and I. tenuis,are sister, although each has a long branch indicating thatmore comprehensive sampling may reveal a more compli-cated relationship. Lawrence (1953) assigned I. gracilipes tosubsect. Evansia while placing I. tenuis in subsect. Apogon.However, Lenz (1959) transferred I. tenuis to subsect. Evansiabecause he noticed the morphological and ecological similar-ities among I. tenuis, I. cristata, and I. gracilipes. Lenz’s (1959)observation on the similarities between the North AmericanI. tenuis and the eastern Asian I. gracilipes, and the results inWilson (2011) that resolved I. tenuis and I. gracilipes as sisterspecies in her subgen. Limniris III clade, are supported by ourstudy (Fig. 1).Iris cristata and I. lacustris are sister in a second lineage

outside of the core-crested clade (Fig. 1). Dykes (1913) pro-posed I. lacustris as a local form of I. cristatawhile later authorsconsidered them as distinct yet closely related (Rodionenko1984; Mathew 1989). Wilson (2011) suggested that these sis-ter species might represent an early diverging lineage in Iris.The sister relationship is supported by their shared distinc-tive sepal crest morphology of a central crest flanked oneach side by a row of irregular outgrowths (Guo and Wilsonunpubl. data).Iris speculatrix and I. proantha var. valida are resolved in

separate clades outside of the core-crested clade even thoughMathew (1989) placed both in sect. Lophiris. Wilson (2011)included I. proantha var. valida in her study and it wasresolved in a subgen. Limniris s.s. clade with rhizomatousnon-crested species from ser. Chinenses (Diels) G. H. M. Lawr.Our study with its limited sampling of non-crested speciessupports the results of Wilson (2011).Anatomical studies of leaves across Iris indicated that

I. speculatrix should also be included in ser. Chinenses (Wuand Cutler 1985). Mathew (1989) although placing I. speculatrixin sect. Lophiris, questioned its position with other crestedspecies because its crest was only a slightly raised ridge. Inour phylogeny, I. speculatrix is not resolved with othercrested species nor with ser. Chinenses species but instead isresolved as sister to I. winogradowii, a bulbous species fromsubgen. Hermodactyloides Spach. Further studies are neededto confidently resolve the placement of I. speculatrix.In conclusion, we find that sepal crests have evolved at

least five times in Iris. Most of the crested species areincluded in a core-crested clade in which four well-defined

lineages are resolved. This core-crested clade includes spe-cies with bulbs, rhizomes, and tuberous roots, characters thathistorically were used to delineate subgenera in Iris. Six crestedspecies that are not in the core-crested clade are in four otherindependently evolved lineages. Thus, by extensive sam-pling of the crested species, our study determines whichspecies are included in the core-crested clade, and indicatesa complex evolutionary history for the sepal crest, a morpho-logically unique yet diverse character of Iris.

Acknowledgments. We thank Lei Shi, Guofeng Sun, Dongqi Wu,and Jinzheng Zhang from the Beijing Botanical Garden of the ChineseAcademy of Sciences; Cuifen Wu and Qianyu Zuo from the CaojianForest Institute of Yunlong County; Jianfeng Du from the HangzhouBotanical Garden; Heng Li, Yunguang Shen, and Weibang Sun from theKunming Institute of Botany; Kuo-Fang Chung from the National TaiwanUniversity; Mingshui Zhao from the TianMu Mountain Institute; HollyForbes from the University of California Botanical Garden at Berkeley;personnel from the Betty Ford Alpine Garden, Juniper Level BotanicGarden, and ShawNature Preserve; and JamesWaddick, Kenneth Walker,and Harvey and Irene Wrightman, for providing plant materials and/orassisting with field trips. We thank J. Travis Columbus and J. Mark Porterfor helpful comments on an earlier version of this manuscript, and twoanonymous reviewers and Chrissen Gemmill for comments that improvedthis final manuscript. This research was supported by grants from theGraduate Program at the Rancho Santa Ana Botanic Garden, ClaremontGraduate University (JG), American Iris Society Foundation (JG andCAW) and NSF: DEB–1011731 (JG and CAW), DEB–0601068 (CAW),and DEB–1020826 (CAW).

Literature Cited

British Iris Society [BIS]. 1997. A guide to species irises: their identificationand cultivation, eds. The Species Group of the British Iris Society.Cambridge: Cambridge University Press.

Doyle, J. J. and J. L. Doyle. 1987. A rapid DNA isolation procedure forsmall quantities of fresh leaf tissue. Phytochemical Bulletin 19: 11–15.

Dykes, W. R. 1913. The genus Iris. New York: Dover Publications.Felsenstein, J. 1981. Evolutionary trees from DNA sequences: a maximum

likelihood approach. Journal of Molecular Evolution 17: 368–376.Felsenstein, J. 1985. Confidence limits on phylogenies: an approach using

the bootstrap. Evolution 39: 783–791.Foster, M. 1883. Notes on irises. The Gardeners’. The Chronicle 20: 231–232.Goldblatt, P. and J. C. Manning. 2008. The Iris family: natural history and

classification. Portland: Timber Press.Guindon, S., J. F. Dufayard, V. Lefort, M. Anisimova, W. Hordijk, and O.

Gascuel. 2010. New algorithms and methods to estimate maximum-likelihood phylogenies: assessing the performance of PhyML 3.0.Systematic Biology 59: 307–321.

Huelsenbeck, J. P. and F. Ronquist. 2001. MRBAYES: Bayesian inferenceof phylogeny. Bioinformatics 17: 754–755.

Huelsenbeck, J. P. and F. Ronquist. 2003. MRBAYES 3: Bayesian phylo-genetic inference under mixed models. Bioinformatics 19: 1572–1574.

Ikinci, N., T. Hall, M. D. Lledo, J. J. Clarkson, N. Tillie, A. Seisums, T.Saito, M. Harley, and M. W. Chase. 2011. Molecular phylogenetics ofthe juno irises, Iris subgenus Scorpiris (Iridaceae), based on six plas-tid markers. Botanical Journal of the Linnean Society 167: 281–300.

Johnson, L. A. and D. E. Soltis. 1994. MatK DNA sequences and phylo-genetic reconstruction in Saxifragaceae s. str. Systematic Botany19: 143–156.

Lawrence, G. H. M. 1953. A reclassification of the genus Iris. GentesHerbarum 8: 346–371.

Lenz, L. W. 1959. Iris tenuis S. Wats., a new transfer to the subsectionEvansia. Aliso 4: 311–319.

Mathew, B. 1989. The Iris. London: Batsford.Posada, D. and K. A. Crandall. 1998. MODELTEST: testing the model of

DNA substitution. Bioinformatics 14: 817–818.Rambaut, A. 2002. Se-Al, sequence alignment editor, v. 2.0a11. Institute of

Evolutionary Biology, Edinburgh, U. K. Available at http://tree.bio.ed.ac.uk/software/seal/.

Rodionenko, G. I. 1984 [1987 translation]. The genus Iris L. London: TheBritish Iris Society.

Rodionenko, G. I. 1994. Rod Juno (Iridaceae). Botanicheskii Zhurnal,Moscow & Leningrad 79: 100–108.

994 SYSTEMATIC BOTANY [Volume 38

Shaw, J., E. B. Lickey, J. T. Beck, S. B. Farmer, W. Liu, J. Miller, K. C.Siripun, C. T. Winder, E. E. Schilling, and R. L. Small. 2005. Thetortoise and the hare II: relative utility of 21 noncoding chloroplastDNA sequences for phylogenetic analysis. American Journal of Botany92: 142–166.

Shaw, J., E. B. Lickey, E. E. Schilling, and R. L. Small. 2007. Comparison ofwhole chloroplast genome sequences to choose noncoding regionsfor phylogenetic studies in angiosperms: the tortoise and the hare III.American Journal of Botany 94: 275–288.

Swofford, D. L. 2002. PAUP*: Phylogenetic analysis using parsimony(*and other methods), v. 4.0 beta 10. Sunderland: Sinauer.

Swofford, D. L., G. J. Olsen, P. J. Waddell, and D. M. Hillis. 1996.Phylogenetic inference. Pp. 407–514 in Molecular systematics ed. 2,eds. D. M. Hillis, C. Moritz, and B. K. Mable. Sunderland: Sinauerand Associates.

Taberlet, P., L. Gielly, G. Pautou, and J. Bouvet. 1991. Universal primersfor amplification of three non-coding regions of chloroplast DNA.Plant Molecular Biology 17: 1105–1109.

Waddick, J. W. and Y. T. Zhao. 1992. Iris of China: Chinese Iris in the wildand in the garden. Portland: Timber Press.

Wilson, C. A. 2004. Phylogeny of Iris based on chloroplast matK geneand trnK intron sequence data.Molecular Phylogenetics and Evolution33: 402–412.

Wilson, C. A. 2006. Patterns of evolution in characters that define Irissubgenera and sections. Aliso 22: 425–433.

Wilson, C. A. 2009. Phylegenetic relationships among the recognizedseries in Iris section Limniris. Systematic Botany 34: 277–284.

Wilson, C. A. 2011. Subgeneric classification in Iris re-examined usingchloroplast sequence data. Taxon 60: 27–35.

Wilson, C. A. and C. L. Calvin. 2006. An origin of aerial branch parasit-ism in the mistletoe family, Loranthaceae. American Journal of Bot-any 93: 787–796.

Wu, Q. G. and D. F. Cutler. 1985. Taxonomic, evolutionary and ecologicalimplications of the leaf anatomy of rhizomatous Iris species. Botani-cal Journal of the Linnean Society 90: 253–303.

Zhao, Y. T. 1980. Some notes on the genus Iris of China.Acta PhytotaxonomicaSinica 18: 53–62.

Appendix 1—List of Iris taxa included in this study with collectionand voucher information, geographic origin, and GenBank accessions(matK/trnK, trnL–trnF, rpl32–trnL, psbJ–petA, rpoB–trnC). GenBank num-bers for newly gathered sequence data are in bold and matK/trnK datarepresented by more than one GenBank number are given in parentheses.The psbJ–petA sequence (< 200 bp) for I. tenuis lacks a GenBank number.

Subgen. Scorpiris Spach: I. albomarginata (B. Fedtsch.) R. C. Foster.UZBEKISTAN. (cult.), Wilson WM10-01 (RSA), KC510964, KC510921,KC510763, KC510816, KC510868. I. aucheri (Baker) Sealy. SYRIA. WilsonS06-39 (RSA), (HM574525, HM574652, HM574591),KC510922,KC510764,KC510817, KC510869. I. bucharica Foster. RUSSIA. Wilson JM06-02(RSA), (HM574529, HM574656, HM574595), KC510923, KC510765,KC510818, KC510870; I. bucharica Foster. TAJIKISTAN. (cult.), GuoKW11-34 (RSA), KC510965, KC510924, KC510766, KC510819, KC510871.I. caucasica subsp. caucasica Hoffm. GEORGIA. M. Mosulishvili G00-03(RSA), (HM574526, AY596668), KC510925, KC510767, KC510820,KC510872. I. caucasica subsp. turcica B. Mathew. TURKEY. Wilson T06-63 (RSA), (HM574550, HM574675, HM574614), KC510926, KC510768,KC510821, KC510873. I. cycloglossa Wendelbo. AFGHANISTAN. WilsonBFA07-19 (RSA), (HM574527, HM574654, HM574593),KC510927,KC510769,KC510822, KC510874. I. galatica Siehe. TURKEY. Wilson T04-01 (RSA),(FJ156302, FJ197279, FJ197232), EU939473, KC510770, KC510823, KC510875.I. kuschakewiczii B. Fedtsch. KAZAKHSTAN. (cult.), Wilson WM10-03(RSA), KC510966, KC510928, KC510771, KC510824, KC510876. I. linifoliaO. Fedtsch. UZBEKISTAN. Wilson UZ11-27 (RSA), KC510967, KC510929,KC510772, KC510825, KC510877. I. magnifica Vved. RUSSIA. WilsonJM06-05 (RSA), (HM574528, HM574655, HM574594), KC510930, KC510773,KC510826, KC510878. I. maracandica Vved. UZBEKISTAN. Wilson UZ11-24 (RSA), KC510968, KC510931, KC510774, KC510827, KC510879.I. narbutii O. Fedtsch. UZBEKISTAN, Wilson UZ11-28 (RSA), KC510969,KC510932, KC510775, KC510828, KC510880. I. nezahatiae Guner &H. Duman. TURKEY. Wilson T11-15 (RSA), KC510970, KC510933,

KC510776, KC510829, KC510881. I. persica L. TURKEY. Usta T02-15(RSA), (HM574530, AY596667), KC510934, KC510777, KC510830, KC510882.I. planifolia Fiori & Paol. SPAIN. Wilson SP07-07 (RSA), (HM574531,HM574658, HM574597), KC510935, KC510778, KC510831, KC510883.I. pseudocapnoides Ruksans. UZBEKISTAN. Wilson UZ11-29 (RSA),KC510971,KC510936,KC510779,KC510832,KC510884. I. pseudocaucasicaGrossh. TURKEY. Wilson T11-06 (RSA), KC510972, KC510937, KC510780,KC510833, KC510885. I. rosenbachiana Regel. TADZHIKISTAN. (cult.),Wilson WM10-02 (RSA), KC510973, KC510938, KC510781, KC510834,KC510886. I. stenophylla Hausskn. ex Baker. TURKEY. Usta T03-03 (RSA),(FJ156332, AY596666), EU939503, KC510782, KC510835, KC510887. I.vicaria Vved. UZBEKISTAN. Wilson UZ11-26 (RSA), KC510974, KC510939,KC510783, KC510836, KC510888. I. warleyensis Foster. TAJIKISTAN.Wilson JM06-14 (RSA), KC510975, KC510940, KC510784, KC510837,KC510889. I. willmottiana Foster. TAJIKISTAN. Wilson JM06-07 (RSA),KC510976, KC510941, KC510785, KC510838, KC510890. I. zenaidaeVved. KYRGYZSTAN. (cult.), Wilson WM10-04 (RSA), KC510977,KC510942, KC510786, KC510839, KC510891.

Subgen. Limniris sect. Lophiris Tausch: I. confusa Sealy. CHINA. GuoCH09-24 (RSA), KC510978, KC510943, KC510787, KC510840, KC510892.I. cristata Sol. Ex Aiton. U. S. A. Karst US09-27 (RSA), KC510979,KC510944,KC510788,KC510841,KC510893. I. formosanaOhwi. TAIWAN.Chung 1928 (HAST), KC510980, KC510945, KC510789, KC510842,KC510894. I. gracilipes A. Gray. JAPAN. Wilson JPW08-32 (RSA),(HM574566, HM574691, HM574630), KC510946, KC510790, KC510843,KC510895; I. japonica Thunb. CHINA. Guo BJBG07-30 (RSA),(HM574563, HM574688, HM574627), KC510947, KC510791, KC510844,KC510896. I. japonica Thunb. CHINA. Guo CH08-12 (RSA), KC510981,KC510948, KC510792, KC510845, KC510897. I. lacustris Nutt. U. S. A.Wilson JPW08-30 (RSA), (HM574567, HM574692, HM574631), KC510949,KC510793, KC510846, KC510898. I. latistyla Y. T. Zhao. CHINA. GuoBJBG07-32 (RSA), KC510982, KC510950, KC510794, KC510847,KC510899. I. milesii Baker ex Foster. CHINA. Guo CH10-30 (RSA),KC510983, KC510951, KC510795, KC510848, KC510900. I. proantha var.valida (S. S. Chien) Y. T. Zhao. China. Guo CH08-11 (RSA), (HM574565,HM574690, HM574629), KC510952, KC510796, KC510849, KC510901. I.speculatrix Hance. CHINA. Guo CH08-10 (RSA), KC510984, KC510953,KC510797, KC510850, KC510902. I. tectorumMaxim. CHINA. Guo CH10-18 (RSA), KC510985, KC510954, KC510798, KC510851, KC510903. I.tectorum f. alba Makino. CHINA. Guo BJBG07-28 (RSA), KC510986,KC510955, KC510799, KC510852, KC510904. I. tenuis S. Watson. U. S. A.Wilson US01-12 (RSA), (FJ156337, AY596638), EU939508, KC510800,KC510905. I. wattii Baker ex Hook. f. CHINA. UCBG 92.1515 (UC),(HM574536, AY596640), KC510956, KC510801, KC510853, KC510906.

Subgen. Nepalensis Dykes: I. barbatula Noltie & K. Y. Guan. CHINA.Wilson PH04-11 (RSA), (HM574520, HM574647, HM574586), KC510957,KC510802, KC510854, KC510907. I. collettii Hook. f. CHINA. UCBG03-16 (UC), (HM574521, AY596664), KC510958, KC510803, KC510855,KC510908. I. decora Wall. CHINA. UCBG 89.0164 (UC), (FJ156294,FJ197273, FJ197226), EU939465, KC510804, KC510856, KC510909.

Subgen. Limniris sect. Limniris Tausch: I. douglasiana Herb. U. S. A.Wilson US92-56 (RSA), (FJ156296, FJ197275, FJ197228), EU939467,KC510805, KC510857, KC510910. I. henryi Baker. CHINA. Guo BJ08-17(RSA), (HM574504, HM584200, HM574569), KC510959, KC510806,KC510858, KC510911; I. henryi Baker. CHINA. Guo JW11-35 (RSA),KC510987, KC510960, KC510807, KC510859, KC510912. I. lazica Albov.TURKEY. Usta T03-12 (RSA), (HM574548, AY596626), EU939482,KC510808, KC510860, KC510913. I. setosa Pall. ex Link. U. S. A. PfauthUS02-09 (RSA), (FJ156327, AY596644), EU939498, KC510809, KC510861,KC510914. I. unguicularis Poir. TURKEY. Usta T03-07 (RSA), (FJ156344,AY596625), EU939515, KC510810, KC510862, KC510915.

Subgen. Iris L.: I. germanica L. ITALY. Wilson DBG05-23 (RSA),(HM574509, HM574636, HM574575), KC510961, KC510811, KC510863,KC510916. I. sari Schott ex Baker. TURKEY. Usta T02-11 (RSA),(FJ156326, AY596659), EU939497, KC510813, KC510865, KC510918. I.subbiflora Brot. SPAIN. Wilson SP07-10 (RSA), (FJ156333, FJ197301,FJ197254), EU939504, KC510812, KC510864, KC510917.

Subgen. Hermodactyloides Spach: I. winogradowii Fomin. Georgia.Wilson JM06-12 (RSA), (HM574535, HM574662, HM574601), KC510962,KC510814, KC510866, KC510919.

Subgen. Pardanthopsis (Hance) Baker: I. dichotoma Pall. CHINA. GuoCH10-29 (RSA), KC510988, KC510963, KC510815, KC510867, KC510920.

2013] GUO AND WILSON: CRESTED IRIS PHYLOGENY 995