American Journal of Botany 101 ( 11) : 2 005– 2 016 , 2014. M ORPHOMETRICS OF DAUCUS (APIACEAE): A COUNTERPART TO A PHYLOGENOMIC STUDY 1 CARLOS ARBIZU 2 , K ATHLEEN R. REITSMA 3 , PHILIPP W. SIMON 2 , AND D AVID M . SPOONER 2,4 2 USDA-Agricultural Research Service, Vegetable Crops Research Unit, Department of Horticulture, University of Wisconsin- Madison, 1575 Linden Dr., Madison, Wisconsin 53706-1590 USA; and 3 Iowa State University, North Central Regional Plant Introduction Station, Ames, Iowa 50011-1170 USA • Premise of study: Molecular phylogenetics of genome-scale data sets (phylogenomics) often produces phylogenetic trees with unprecedented resolution. A companion phylogenomics analysis of Daucus using 94 conserved nuclear orthologs supported many of the traditional species but showed unexpected results that require morphological analyses to help interpret them in a practical taxonomic context. • Methods: We evaluated character state distributions, stepwise discriminant analyses, canonical variate analyses, and hierarchi- cal cluster analyses from 40 morphological characters from 81 accessions of 14 taxa of D aucus and eight species in related genera in an experimental plot. • K ey results: Most characters showed tremendous variation with character state overlap across many taxa. Multivariate analyses separated the outgroup taxa easily from the D aucus ingroup. Concordant with molecular analyses, most species form phenetic groups, except the same taxa that are problematical in the molecular results: (1) the subspecies of D . carota , (2) D . sahariensis and D . syrticus , and (3) D . broteri and D. guttatus . • Conclusions: Phenetic analyses, in combination with molecular data, support many D aucus species, but mostly by overlapping ranges of size and meristic variation. The subspecies of D . carota are poorly separated morphologically, are paraphyletic, and all could be recognized at the subspecies rank under D . carota. D aucus sahariensis and D. syrticus are so similar morphologi- cally that they could be placed in synonymy. Combined molecular and morphological data support three species in accessions previously identifi ed as D . broteri and D. guttatus . Molecular and morphological results support the new combination D aucus carota subsp. capillifolius . Key words: Apiaceae; D aucus ; germplasm; morphological phenetics; species boundaries; Umbelliferae. R ecent phylogenomic analyses are producing unprecedented D. guttatus (T able 1 lists authors of taxa investigated here). phylogenetic resolution (e.g., R okas et al., 2003; Burki et al., Distinguishing characters of D . broteri and D. guttatus are un- 2008 ; Christelová et al., 2011 ; D os Reis et al., 2012; E gan et al., clear from the taxonomic literature, and they fell into three 2012 ). An analysis of 92 accessions of 13 species and two sub- clades that were labeled as D . guttatus (the earliest name) 1, 2, species of D aucus and an additional 15 accessions of related and 3 (A rbizu et al., 2014 ) ( Fig. 1 ). R ouya polygama Coincy genera [A mmi L., Astrodaucus Drude, Caucalis L., Rouya Co- (2 n = 20) and Pseudorlaya pumila (2 n = 16) were resolved as incy (incorrectly identifi ed in our prior papers as M argotia ingroups to D aucus, sister to a 2 n = 18 clade composed of D . Boiss.), O enanthe L., Orlaya Hoffm., P seudorlaya (Murb.) carota, D . capillifolius, D. sahariensis, and D . syrticus (clade Murb., T orilis Adans., T urgenia Hoffm.] was examined with A ′ of F ig. 1 ), with the remaining species of Daucus being 2 n = 20, DNA sequences of 94 nuclear orthologs of average length of 22, and 44. Discordant or confusing phylogenetic results, no 1180 bp, with an aligned length of 111 166 bp ( Arbizu et al., matter how strongly supported by molecular methods, are of 2014) . It provided 100% bootstrap support for most of the external little value without corroborative studies, here examined using and many of the internal clades, grouped different accessions of morphological data. most of the species with strong support, but failed to support The necessity of corroborative phenotyping studies in phylo- others such as (1) the subspecies of D . carota and D . capillifo- genetics is analogous to its need in high-resolution linkage lius , (2) D . sahariensis and D . syrticus , and (3) D . broteri and mapping and genome-wide association studies, where “phenom- ics” is emerging as a time-consuming and expensive constraint needed with next-generation DNA sequencing data ( Cobb 1 Manuscript received 3 June 2014; revision accepted 10 September 2014. et al., 2013 ; D hondt et al., 2013 ; F iorani and Schurr, 2013 ). The This work was supported by the USDA-ARS National Plant Germplasm purpose of the present study was to phenotype the accessions System Horticultural Evaluation Grants to D.S., P.S., and K.R. The authors examined by Arbizu et al. (2014), to examine the support for thank Ms. Lucinda Clark and other North Central Regional Plant Intro- species in D aucus , to provide a morphological counterpart to duction Station (NCRPIS) staff for fi eld assistance and John Freudenstein, the phylogenomic results of Arbizu et al. (2014) , and to place Richard Jensen, and one unnamed reviewer for their reviews, and Fernando both sets of data in a practical taxonomic context. It is an exten- Martínez Flores for pointing out our misidentifi cation of R ouya polygama in our prior papers as the morphologically very similar M argotia gummifera sion of a similar study examining the taxonomic boundaries of (now T hapsia gummifera) . the subspecies of D aucus carota and D . capillifolius ( Spooner 4 Author for correspondence (e-mail: david.spooner@ars.usda.gov) et al., 2014 ). That morphological study suggested the failure to provide molecular support for the subspecies of D . carota was d oi:10.3732/ajb.1400252 partly a result of too many recognized subspecies and supported American Journal of Botany 1 01 ( 11 ): 2 005 – 2016 , 2 014; http://www.amjbot.org/ © 2014 B otanical Society of America 2005 2006 AMERICAN JOURNAL OF BOTANY [Vol. 101 TABLE 1. Accessions examined in this study. Tentative new Taxona identifi cations Accessionb Location or sourcec Ingroup Daucus aureus Desf. PI 295854 Israel. Wadi Rubin (HaMerkaz). D. aureus PI 319403 Israel. Mediterranean Region. D. aureus PI 478858 France. Dijon. D. broteri Ten. D . guttatus 1 PI 652233 Iran. Mazandaran: Dhalus Road, Dasht-e Nazir, Kandalus. D. broteri D. guttatus 2 PI 652329 Greece. Peloponnese: 4 km from Skoura, toward Leonidion, Laconia Prefecture. D. broteri D. guttatus 1 PI 652340 Syria. Kassab. D. broteri D. guttatus 3 PI 652367 Turkey. Mugla. D. capillifolius Gilli PI 279764 Libya. Near Jefren. D. capillifolius Ames 30198 Tunisia. Medenine. D. capillifolius Ames 30202 Tunisia. Medenine. D. capillifolius Ames 30207 Tunisia. Medenine. D. carota L. subsp. carota Ames 25017 Germany. Saxony-Anhalt. D. carota subsp. carota Ames 26393 Portugal. Castelo Branco. D. carota subsp. carota Ames 26394 Portugal. Portalegre near Monforte. D. carota subsp. carota Ames 26401 Portugal. Portalegre near Monforte. D. carota subsp. c arota Ames 27397 Uzbekistan. Between Yalangoch and Sobir Raximova. D. carota subsp. carota Ames 30250 Tunisia. Nabuel: along Route 28 at junction of road to Takelsa. D. carota subsp. carota Ames 30251 Tunisia. Nabuel: Route 26, between Takelsa and El Haouaria, 26 km from El Haouaria. D. carota subsp. c arota Ames 30252 Tunisia. Nabuel: Sidi Daoud, 1 km from Route 27. D. carota subsp. carota Ames 30253 Tunisia. Nabuel: between El Haouarcae and Dor Allouche. D. carota subsp. carota Ames 30254 Tunisia. Nabuel: between El Haouarcae and Dor Allouche. D. carota subsp. carota Ames 30255 Tunisia. Nabuel: along road between Korba and Beni Khalled. D. carota subsp. carota Ames 30259 Tunisia. Bizerte: south side of Ischkeul. D. carota subsp. carota Ames 30260 Tunisia. Bizerte: along Route 51, west of Ghzab. D. carota subsp. carota Ames 30261 Tunisia. Bizerte: grounds of Direction Regionale Mogods, Khroumerie Sejnane. D. carota subsp. carota Ames 30262 Tunisia. Beja: road from Route 7, just west of Sejnane to Cap Negro. D. carota subsp. carota PI 274297 Pakistan. Northern Areas. D. carota subsp. carota PI 279762 Source: Denmark. Copenhagen. D. carota subsp. carota PI 279775 Source: Hungary. Pest. Botanical Garden. D. carota subsp. carota PI 279777 Source: Egypt. Giza: Orman Botanic Garden. D. carota subsp. carota PI 279788 Austria. Vienna. D. carota subsp. carota PI 279798 Spain. Madrid. D. carota subsp. carota PI 295862 Spain. D. carota subsp. carota PI 390887 Israel. Central Israel: from Bet Elazari. D. carota subsp. carota PI 421301 USA. Kansas: Elk County. D. carota subsp. carota PI 430525 Afghanistan. Zardek. D. carota subsp. carota PI 478369 China. Xinjiang: near Chou En Lai Monument Stone River, Sinkiang. D. carota subsp. carota PI 478873 Italy. Sardinia: St. Elia Beach, 50 m from sea, Cagliari. D. carota subsp. carota PI 478881 USA. Oregon: roadside between Echo and Pendleton. D. carota subsp. carota PI 478884 Source: The Netherlands, South Holland: Botanical Garden, Leiden. D. carota subsp. carota PI 502244 Portugal. Coimbra: Lousa. D. carota subsp. carota PI 652225 Source: France. Collection site unknown. D. carota subsp. carota PI 652226 Greece. N. Khalkidiki: 10 km N of Kassandra on coast road. D. carota subsp. carota PI 652229 Source: Tunisia. D. carota subsp. carota PI 652230 Albania. Lushnje. D. carota subsp. carota PI 652393 Turkey. Konya: 10–15 km to Seydisehir, between Yarpuz and Konya. D. carota subsp. gummifer (Syme) Hook.f. Ames 7674 Source: Italy. Tuscany: Botanic Garden. D. carota subsp. gummifer Ames 26381 Portugal. Faro: Near Portunao. D. carota subsp. gummifer Ames 26382 Portugal. Faro: Near Sagres. D . carota subsp. gummifer Ames 26383 Portugal. Faro: Near Aljezur. D. carota subsp. gummifer Ames 26384 Portugal. Beja. D . carota subsp. g ummifer Ames 31193 France. D. carota subsp. gummifer PI 478883 France. Finistere: maritime turf, Le Conquet. D. carota subsp. gummifer PI 652411 France. Finistere: Pointe de Rospico, Navez. D. carota subsp. major (Vis.) Arcang. D . guttatus 1 Ames 25898 Turkey. Konya: Konya, toward Beysehir. D. carota D . guttatus 1 PI 286611 Source: Lebanon. Faculty of Agricultural Sciences. D. crinitus Desf. PI 652412 Portugal. Braganca: near Zava. D. crinitus PI 652413 Portugal. Guarda: near Barca de Alva. D. crinitus PI 652414 Portugal. Faro: near Bengado. November 2014] ARBIZU ET AL.—MORPHOMETRICS OF DAUCUS 2007 TABLE 1. Continued. Tentative new Taxona identifi cations Accessionb Location or sourcec D. glochidiatus (Labill.) Fisch., PI 285038 Source: CSIRO, Australia. Capital Territory. C.A.Mey. & Avé-Lall. D. guttatus Sibth. and Sm. D. guttatus 1 PI 279763 Source: Israel. Jerusalem Department of Botany. D. guttatus D. guttatus 2 PI 652331 Greece. Peloponnese: village of Loutra Agias Elenis, 17 km S of Korinthos, Korinthia Prefecture. D. guttatus D . guttatus 1 PI 652343 Syria. Halwah. D . guttatus D. guttatus 2 PI 652360 Turkey. Mugla: between Soke and Milas. D. involucratus Sm. PI 652332 Greece. Peloponnese: village of Loutra Agias Elenis, 17 km S of Korinthos, Korinthia Prefecture. D . involucratus PI 652350 Turkey. Izmir. D. involucratus PI 652355 Turkey. Izmir: 5 km N of Kusadasi. D. littoralis Sm. PI 295857 Israel. Beit Alpha. D. littoralis PI 341902 Israel. D. littoralis D . guttatus 3 PI 652375 Turkey. Mugla: between Dalaman-Gocik and Fethiye. D. muricatus L. Ames 25419 Portugal. Coimbra: Pitanca de Baixo-Condeixa. D. muricatus Ames 29090 Tunisia. South of Tunis along Hwy. 3 toward Zaghouan. D . muricatus PI 295863 Spain. Cordoba. From Villa del Rio (Cordoba). D. pusillus Michx. PI 349267 Uruguay. Montevideo. Near La Colorado Beach. D. sahariensis Murb. Ames 29096 Tunisia. between Tataouine and Bir Lahmer. D. sahariensis Ames 29097 Tunisia. between Tataouine and Remada. D. sahariensis Ames 29098 Tunisia. between Remada and Chenini. D. syrticus Murb. D. sahariensis Ames 29107 Tunisia. near Beni Kdache to the south. D. syrticus Ames 29108 Tunisia. between Medenine and Matmatas. D. syrticus Ames 29109 Tunisia. between Medenine and Matmatas. D. syrticus Ames 29110 Tunisia. between Matmatas and El Hamma, near the Gabes airport. Rouya polygama Coincy Ames 30292 Tunisia. Jendouba: road to Tabarka, near Tabarka airport. Outgroups A mmi visnaga (L.) Lam. Ames 30185 Tunisia. Bizerte: National Park Ischkeul on road to Eco Museum. A strodaucus littoralis Drude PI 277064 Source: Azerbaijan. Baku Botanical Garden. Caucalis platycarpos L. PI 649446 Germany. Saxony-Anhalt: Mannsdorf. Oenanthe virgata Poir. Ames 30293 Tunisia. Beja: Route 11, 41 km from Eudiana, 254 km from Beja. O rlaya daucoides (L.) Greuter PI 649477 Turkey. Aydin: Dilek Peninsula Reserve. O rlaya daucorlaya Murb. PI 649478 Greece. Epirus: 8 km from Aristi, toward Ioannina. T orilis leptophylla Rchb.f. Ames 25750 Syria. Salma. a These names correspond to those in the Germplasm Resources Information Network (GRIN) website (see methods section), except for the proposed new identifi cations of the subspecies of D . carota listed by Spooner et al. (2014) . b Plant Introduction (PI) numbers are permanent numbers assigned to germplasm accessions in the National Plant Germplasm System (NPGS). Germplasm centers in the NPGS assign temporary site-specifi c numbers to newly acquired germplasm (Ames numbers for carrots and other Apiaceae maintained at the North Central Regional Plant Introduction Station in Ames, Iowa, USA) until an accession’s passport data and taxonomy is verifi ed, it is determined not to be a duplicate accession, and it has been determined the accession can be successfully maintained. These accessions may or may not be assigned a PI number after the assessment period. c Location refers to where the germplasm was collected in the wild, while source refers to germplasm acquired through another entity such as a market vendor or genebank. only subsp. c arota and subsp. g ummifer, not the 9–12 subspe- Tunisia (L e Floc’h et al., 2010) , Palestine (Z ohary, 1972 ), Syria cies of D. carota recognized by other authors. (M outerde, 1986) , and Turkey and the East Aegean Islands Daucus is an economically important genus, but is in need of (C ullen, 1972 ). However, identifi cations in these taxonomic modern taxonomic and monographic studies. The genus in- treatments frequently use different characters in their taxo- cludes about 20 recognized species mostly centered in the Med- nomic keys and descriptions, have incomplete synonymies iterranean area in contrast to the widespread D aucus carota that which preclude comparison of their taxonomic concepts, often occurs on almost every continent. The haploid chromosome have little information about geographic ranges, and lack distri- number for D aucus ranges from n = 9 to n = 11. Most species bution maps. In addition, there has been no single compilation are diploids with 2n = 18, 20, and 22, but two polyploid species of type specimens and many of the types lack the full range of have been reported ( Grzebelus et al., 2011) . The latest taxo- plant parts necessary for unambiguous identifi cation. In sum- nomic monograph of D aucus by S áenz Laín (1981) lacks com- mary, there has been no accepted standard to quantify and de- plete synonymies, distribution maps and phylogenetic data and scribe the huge range of variation in D aucus , and identifi cations cites few specimens. Practical identifi cations have relied more are often problematic. on fl oristic treatments such as those from Algeria ( Quezel and The present study expands the morphological analysis of Santa, 1963 ), Europe ( Heywood, 1968) , the Iberian Peninsula S pooner et al. (2014) to include all Daucus species available as and Balearic Islands (P ujadas Salvà, 2003) , Libya ( Jafri and germplasm and used the same accessions examined by A rbizu El-Gadi, 1985 ), Morocco ( Jury, 2002 ; F aris and Ibn Tattou, 2007 ), et al. (2014) . In addition to phylogenetic insights needed for 2008 AMERICAN JOURNAL OF BOTANY [Vol. 101 Fig. 1. Maximum parsimony phylogeny of D aucus and outgroups based on 94 aligned nuclear orthologs of aligned length 111 166 bp ( Arbizu et al., 2014 ; correcting M argotia gummifera to Rouya polygama ). November 2014] ARBIZU ET AL.—MORPHOMETRICS OF D AUCUS 2009 crop improvement, these combined molecular and morphologi- with the aid of a dissecting microscope. As part of normal genebank operations cal analyses are needed to organize the world’s germplasm col- at the NCRPIS, electronic images of leaves were generated on a fl atbed scan- lections of D aucus . The US collection of D aucus is maintained ner; images of various plant parts were made from plants in the fi eld with a digi-tal camera; and images will be available on the GRIN website (http://www. at the North Central Regional Plant Introduction Station ars-grin.gov/). These serve as useful resources for others to conveniently check (NCRPIS) in Ames, Iowa. This genebank conserves 1381 ac- the morphology of our accessions and as supplements to the voucher speci- cessions of D aucus . Of these, 569 are classifi ed as landraces, mens. Herbarium vouchers collected for this morphological study are a subset cultivars, cultivated populations, or breeding lines. Improve- of the same accessions from A rbizu et al. (2014) but different specimens ( Table ment status for the remaining accessions includes 571 wild, 17 1) and are deposited at the herbarium of the Potato Introduction Station, Stur- uncertain, and 224 accessions with no status designated (though geon Bay, Wisconsin, USA. many of these most likely are cultivated). Taxonomically, there are 917 accessions identifi ed as D. carota, with 247 of these Analytical methods— T hirty-eight of the 40 characters were scored and ana-lyzed as continuous variables; the remaining two were treated as nominal vari- identifi ed as D . carota with a variety or subspecies designation ables ( Table 2 ). All analyses were conducted in JMP software version 10.0.0 (1164 D . carota total), leaving 217 accessions identifi ed as (SAS Institute, Cary, North Carolina, USA). To examine character state distri- other D aucus species. butions, we analyzed the accessions with the box plot or histogram functions of Graph Builder in JMP (Appendix S1; see Supplemental Data with the online version of this article). M ATERIALS AND METHODS For multivariate analyses, means were assessed for the continuous variables. We fi rst performed stepwise discriminant analyses (linear, common covari- S tudy species— We examined 81 accessions of 14 taxa of D aucus (D . au- ance) using all 38 continuous variables to obtain a model whose variables were reus, D. broteri , D . capillifolius , D . carota subsp. carota, D. carota subsp. signifi cant in identifying accession composition with characters removed one at gummifer , D. crinitus , D. glochidiatus , D . guttatus, D. involucratus , D. litto- a time until the model F test P value was ≤ 0.05. We then performed canonical ralis , D. muricatus , D. pusillus , D . sahariensis , D. syrticus ), seven species in variate analysis (CVA) and hierarchical cluster analyses (HCA) of (1) all taxa; other genera in the Apiaceae supported as outgroups (A mmi visnaga , Astro- (2) D aucus ingroup (all D aucus and R ouya polygama ); (3) D . capillifolius and daucus littoralis , Caucalis platycarpos , Oenanthe virgata , O rlaya daucoides , D. carota; (4) clade B species (see F ig. 1 that shows the species in different Orlaya daucorlaya , T orilis leptophylla ), and one species in another genus sup- clades) D . glochidiatus , D. guttatus (subsets 1, 2, 3), D . involucratus, D. litto- ported as an ingroup (R ouya polygama; Spalik and Downie, 2007 ; Spooner ralis , D . pusillus ; (5) D . guttatus (subsets 1, 2, 3); and (6) D . sahariensis and D. et al., 2013 ) ( Table 1 ). We did not examine P seudorlaya pumila, another in- syrticus . These analyses use only the characters identifi ed by stepwise discrimi- group species, because it died in the fi eld plot, and D . tenuisectus because it was nant analyses as signifi cant in the F test, P ≤ 0.05. The HCA uses standardized acquired too late for planting. On the basis of the morphological analysis of data and average similarity. Spooner et al. (2014) , we labeled all accessions of D . carota as either subsp. c arota or subsp. gummifer, not as their current listing in the Germplasm Re- sources Information Network (GRIN; http://www.ars-grin.gov/) as subsp. R ESULTS c arota , subsp. c ommutatus (Paol.) Thell., subsp. d repanensis (Arcang.) Hey- wood, subsp. f ontanesii Thell., subsp. g ummifer (Syme) Hook.f., subsp. h is- C haracter state distributions — Graphical analyses of all 40 panicus (Gouan) Thell., subsp. major (Vis.) Arcang., subsp. maritimus (Lam.) Batt., and subsp. m aximus (Desf.) Ball. This classifi cation of D . carota into two character state distributions are shown in Appendix S1. Step- subspecies is similar to that of Onno (1937) , who classifi ed the “subsp. g um- wise discriminant analyses of the 38 characters coded as con- mifer ” taxon as D. gingidium L., and the “subsp. carota” taxon as D. carota . It tinuously variable showed all but fi ve of them (stipule width, is also similar to the classifi cations of Small (1978) and Reduron (2007), who foliage color, bract length, length of longest peripheral ray, recognized two “species aggregates”, or “subgroups,” within the single species barbs at tips of spines) to be signifi cant in the F test, P ≤ 0.05, D. carota . However, these authors recognized more subspecies than our two. in at least one of the six analyses of different groups of species On the basis of A rbizu et al. (2014) , we labeled accessions formerly identifi ed as D. broteri or D. guttatus as D. guttatus 1, 2, or 3. (T able 2) . Of the two nominal characters (petal color, anther color), yellow petals are unique to D . capillifolius , and red an- Daucus observation plots— T o ensure suffi cient plant populations in the thers are unique to O enanthe virgata . Some characters showed observation plot, biennial and mixed life-cycle accessions were planted in the little variation and little to no overlap or ranges of character greenhouse in early November 2012. Seedlings were thinned to one per pot, states with many other characters, but most characters showed and plants were fertilized weekly with a commercial liquid fertilizer (NPK tremendous variation within some taxa and overlap of ranges 20–10–20). Roots were vernalized in the dark (4–5° C , 50–70% relative humid- across taxa. Examination of additional accessions, especially ity) for approximately 60 d beginning in February 2013. A fungicide spray for those species with only few available accessions (e.g., D . (Rubigan, DuPont, Wilmington, Delaware, USA) was applied at the beginning of vernalization and reapplied as necessary to prevent Botrytis blight. Roots glochidiatus , D. pusillus, Rouya polygama ) will be needed to were moved outside to a protected area in mid-April to allow them to develop make more defi nitive conclusions of character state variation. new foliage. Annual accessions were planted in the greenhouse in late February Figure 2 illustrates one of these 40 characters, plant height, 2013 and maintained using the same protocols as with the biennials without showing cases of both narrow and wide ranges of overlap vernalization. Twenty plants per accession were transplanted into 6-m rows, among characters. Despite character overlap, many characters one row per accession in each of two fi eld plots in late April. Harsh weather were useful to distinguish taxa, but often only in combination conditions (excessive rain, snow, and cold temperatures) following transplant- ing damaged or killed many of the annual accessions. As a result, additional with others. seeds of the affected accessions were direct seeded by hand into a parallel fur- row 30 cm from the transplanted row in late May. Field plots were maintained Multivariate analyses — We employed both CVA and HCA with small plot tillers and hand weeding. to analyze our data because both distinguish taxa using differ- ent methods and assumptions and both are useful to visualize Characters recorded —F orty characters were recorded from at least three results and infer group membership, here inferred to be poten- individuals per accession ( Table 2 ), and character sets were always recorded by tially valid taxa. The CVA is an ordination method that uses the same individual. These characters were chosen to represent all those used in assigned groups to derive a linear combination of the variables prior morphological analyses ( Small, 1978; Spooner et al., 2014) , the latest comprehensive monograph of D aucus (S áenz Laín, 1981) , and regional fl oras (morphological characters) that produces the greatest separa- outlined in the introduction. Size characters were recorded in the fi eld with a tion of the groups. The HCA, in contrast, makes no assumptions ruler or calipers, and fl oral and fruit characters were recorded in the laboratory about group membership; it produces trees based on average 2010 AMERICAN JOURNAL OF BOTANY [Vol. 101 TABLE 2. The 40 morphological characters recorded in this study, modeling type, and F - test P values of characters retained in a stepwise discriminant analysis for (1) all taxa; (2) D aucus ingroup (all D aucus and Rouya polygama ); (3) D. capillifolius and D. carota ; (4) D. glochidiatus , D. guttatus (subsets 1, 2, 3), D . involucratus , D . littoralis , D. pusillus ; (5) D . guttatus subsets 1, 2, 3; (6) D . sahariensis and D. syrticus . Character a Model type b P 1 P 2 P3 P 4 P 5 P 6 Plant Plant height (cm) C 0.0001 0.0001 0.0180 Stem diameter (mm) C 0.0022 0.0007 Leaf Leaf length (cm) C 0.0001 0.0008 Leaf width (cm) C 0.0053 0.0086 Stipule width (mm) C Petiole length (cm) C 0.0008 0.0033 Petiole diameter (mm) C 0.0003 Petiole shape (round, 1; semiround, 2; fl at, 3) C 0.0037 0.0028 Leaf type (celery, 1; normal, 2; parsley, 3; other, 4) C 0.0240 0.0324 Leaf and petiole pubescence (smooth, 1; intermediate, 2; C 0.0001 0.0001 very hairy, 3) Foliage color (light green, 1; medium green, 2; C gray green, 3; dark green, 4) Flower Peduncle pubescence (glabrous, 1; soft hairs, 2; C 0.0353 0.0001 0.0067 scabrous, 3; very scabrous, 4) Primary umbel shape, full bloom (convex, 1; fl at, 2; concave, 3) C 0.0009 0.0005 Primary umbel shape, mature seed (convex, 1; fl at, 2; concave, 3) C 0.0001 0.0010 Primary umbel height (cm) C 0.0142 Primary umbel diameter (cm) C 0.0001 0.0028 0.0004 Secondary umbel diameter (cm) C 0.0001 0.0083 Bract length (mm) C Bract width (mm) C 0.0204 Involucral bract posture (defl exed, not defl exed C 0.0001 0.0001 [outward or upward]) Number of bract lobe points C 0.0001 0.0001 0.0112 0.0001 0.0001 Number of bract lobe pairs C 0.0001 0.0001 0.0020 0.0005 0.0001 Number of umbel rays C 0.0005 0.0370 Pigmented central umbel (concolorous to outer C 0.0003 0.0100 0.0002 [uniform color], 1; differently pigmented, 2) Length of longest peripheral ray (cm) Length of shortest peripheral ray (cm) 0.0001 0.0392 Petal color (white, cream, yellow [only D . capillifolius ], pink) N Anther color (white, cream, yellow, pink, purple, brown) N Peripheral petal length (mm) C 0.0001 0.0021 0.0005 Central petal length (mm) C 0.0001 Stamen length (mm) C 0.0001 0.0001 0.0062 Style length (mm) 0.0001 0.0001 Stylopodium length (mm) 0.0015 0.0213 Stylopodium width (mm) 0.0001 0.0001 Seed Seed length (mm) C 0.0001 0.0023 0.0001 Seed width (mm) C 0.0001 0.0070 Confl uency of seed spines (separate, 1; little confl uency, 2; C 0.0030 much confl uency, 3) Barbs at tips of spines of secondary seed ribs (3, 2 barbs; 7, 4–8) C Number spines on the secondary seed ribs C 0.0001 0.0001 Length of secondary seed spines (mm) C 0.0001 0.0001 a Additional details on these descriptors can be found at the USDA, ARS, National Genetic Resources Program, Germplasm Resources Information Network (GRIN) website [online database], National Germplasm Resources Laboratory, Beltsville, Maryland, http://www.ars-grin.gov/cgi-bin/npgs/html/ crop.pl?70 [accessed 10 October 2014]. b N, nominal; C, continuous. similarity of all data. Because the clustering results differ with D . glochidiatus, D. guttatus [subsets 1, 2, 3], D. involucratus, different sets of accessions, we performed both analyses (and D. littoralis , D. pusillus ). stepwise discriminant analyses) with six different groups of pu- tatively related taxa. We present the F -test P values of charac- A ll taxa—T he sole purpose of this analysis was to see how ters retained in a stepwise discriminant analysis of all six well the D aucus outgroups (i.e., all non-D aucus species ex- analyses in T able 2 , and all 12 CVA and HCA in Figs. 3–6 or cept R ouya polygama) were separated from the ingroup. Step- Appendix S2 (see online Supplemental Data). For space con- wise discriminant analyses identifi ed 28 of the 38 continuous siderations, we here present only the CVA and HCA results characters as signifi cant discriminators within all taxa at the of analysis 2 (D aucus ingroup) and analyses 4 (clade B species: F -test P value ≤ 0.05 ( Table 2 ). The HCA separated all outgroups November 2014] ARBIZU ET AL.—MORPHOMETRICS OF DAUCUS 2011 Fig. 2. Box plot of plant height for all taxa examined in this study showing individual plant values for median, 25% and 75% quantiles, range, and outliers. except Orlaya daucoides , which grouped with D. muricatus . (Appendix S2) cluster well with the elimination of other taxa The CVA analysis, however, separated O rlaya daucoides except for one accession of subsp. g ummifer (Ames 26381) that from all D aucus species well, close to O enanthe virgata and clusters near D . carota subsp. c arota. Although not used in R ouya polygama (Appendix S2). Within the ingroup, R ouya these analyses, the yellow petal trait easily separates D . capil- polygama is distinct from other D aucus ingroups by both lifolius from D. carota, and when used in combination with the analyses. longer seeds of D . capillifolius ( Table 2 , Appendix S2), these two species are easily distinguished. D aucus ingroup (all Daucus and Rouya polygama)— Deletion of the outgroups signifi cantly changed the phenetic structure Clade B species— The HCA clusters D . guttatus 2, D . gutta- of the ingroup in both the HCA ( Fig. 3) and CVA (F ig. 4 ). tus 3, D . littoralis, and D . pusillus . D aucus glochidiatus and R ouya polygama and D . pusillus appear far from others in the D . involucratus cluster together, and D . guttatus 1 clusters with HCA, while Rouya polygama and D. muricatus are pheneti- D . guttatus 2 and D . guttatus 3 ( Fig. 5 ). The CVA clusters all cally most separate in the CVA (F ig. 4 ). Many ingroup taxa species separately, with D . muricatus being the phenetically cluster in these analyses, concordant to their grouping in the most distinctive species ( Fig. 4 ). multiple nuclear orthology phylogeny ( D. aureus, D . crinitus, D. involucratus , D. littoralis , D. muricatus ). As in the phe- D aucus guttatus subsets 1, 2, and 3— All three groups of netic analysis ( Spooner et al., 2014 ), D . carota subsp. carota D. guttatus are distinguished with the elimination of all other and D. carota subsp. gummifer , are diffi cult to separate in the species (Appendix S2). They are best distinguished by number HCA and CVA, D. capillifolius is similar in the CVA, and all of bract lobe pairs (highest number in D . guttatus 3), peduncle the accessions of D . capillifolius cluster together in the HVA. pubescence (harshest in D . guttatus 1) and peripheral petal length Daucus guttatus 2 and 3 cluster separately in the CVA, but (longest in D. guttatus 1) ( Table 2 , Appendix S1). none of the three forms do so consistently in HCA. D aucus sahariensis and D. syrticus form their own clusters near each D aucus sahariensis and D. syrticus— Daucus sahariensis other in CVA, but fail to form species-specifi c clusters in the and D . syrticus do not cluster in the HCA when analyzed sepa- HCA. rately, but do so in the CVA. The best characters separating these two species are stem diameter, primary umbel diameter, Daucus capillifolius, D. carota subsp. carota, and D. carota number of bract lobe points, central petal length, and stamen subsp. gummifer— The HCA and CVA analyses of D . capillifo- length (T able 2 ). However, all of these characters overlap con- lius , D . carota subsp. c arota , and D . carota subsp. g ummifer siderably in range (Appendix S1). 2012 AMERICAN JOURNAL OF BOTANY [Vol. 101 Fig. 3. Hierarchial cluster analyses of the D aucus ingroup including Rouya polygama . November 2014] ARBIZU ET AL.—MORPHOMETRICS OF DAUCUS 2013 Fig. 4. Canonical variate analysis of the D aucus ingroup including Rouya polygama ; (A) D. guttatus 1 and (B) D. capillifolius are highlighted in larger colored type solely to distinguish them from their intermingled D . carota subspecies. DISCUSSION Arbizu et al. (2014) confi rmed that R ouya polygama and P seu- dorlaya pumila are part of the Daucus clade, and our results Daucus taxonomy traditionally has been diffi cult, as we infer show that Rouya ( Pseudorlaya not examined here) was indeed from our continuing challenges with identifi cations at the NCRPIS distinct and reasonably excluded from D aucus on morphological and from different and often overlapping sets of character states criteria. Weitzel et al. (2014) recently showed, with ITS data, that provided in regional fl oras for the same species. Our raw data con- Rouya polygama (misidentifi ed as M argotia gummifera in the fi rm such a pattern of overlapping character states, traditionally studies of S pooner et al. [2013] and A rbizu et al. [2014] ) was a used as species identifi ers (F ig. 2, Appendix S1). Similarly, our Daucus ingroup. They made the transfer of this species to Thap- multivariate analyses ( Figs. 3–6 ; Appendix S2) show diffi culty in sia gummifera (Desf.) Spring. Problems in such paraphyletic distinguishing some taxa. Comparison of the multiple nuclear or- genera in the Apiaceae are common. The Apiaceae comprise tholog study of Arbizu et al. (2014 ; Fig. 1 of this paper) and the some 300–455 genera and 3000–3750 species ( Constance, 1971; analyses presented here shows both data sets are concordant re- Pimenov and Leonov, 1993 ). Many generic boundaries within garding problems with distinguishing the subspecies of D . carota , the Apiaceae are unnatural, as documented by molecular inves- D. sahariensis and D . syrticus , and D . broteri and D . guttatus , tigations based on DNA sequences from nuclear ribosomal here considered as three putatively different taxa and labeled as internal transcribed spacers, plastid r poC1 intron and r pl16 in- D. guttatus groups 1, 2, and 3. While some characters showed little tron sequences, plastid m atK coding sequences, plastid DNA variation and little to no overlap with many other characters, most restriction-site data, and DNA sequences from nuclear ortho- characters showed tremendous variation within some taxa and logs (P lunkett et al., 1996 ; Downie et al., 2000 ; Lee and Downie, overlap of ranges across taxa, demonstrating that most D aucus 2000 ; Spalik and Downie, 2007; S pooner et al., 2013 ). Generic species are distinguished by size and meristic variation, not the boundaries are particularly diffi cult in D aucus, as molecular data possession of unique traits (Appendix S1). from the above studies place species from the genera A grocharis , 2014 AMERICAN JOURNAL OF BOTANY [Vol. 101 Fig. 5. Hierarchial cluster analyses of clade B species D aucus glochidiatus, D . guttatus (subsets 1, 2, 3), D . involucratus , D. littoralis, D . pusillus. Athamanta , Cryptotaenia , Melanoselinum , Monizia , Pachycte- examining additional collections of D . guttatus 1, 2, and 3 with nium, P seudorlaya , Rouya, and Tornabenea within a monophy- these markers, in concert with an expanded morphological study. letic D aucus clade. However, we do not yet have access to living Iorizzo et al. (2013) demonstrated the utility of single nucleotide collections of many of these non-D aucus genera, which are nec- polymorphism (SNP) data to distinguish D. carota subsp. c arota essary for our nuclear ortholog studies. (wild) from D . carota subsp. s ativus (cultivated), and even major cultivar groups of the latter (eastern vs. western carrot). We are T axonomic concepts — While the identifi cation of most extending these analyses by generating single nucleotide poly- D aucus species relies on a variety of traits with overlapping morphisms (SNPs) via genotyping by sequencing to examine the ranges of values (polythetic support), in contrast to possessing subspecies of D . carota that cannot be distinguished by multiple unique traits ( Sokal and Sneath, 1963) , many species are sup- nuclear orthologs (A rbizu et al., 2014) . Longer term, we plan ported by molecular data and can be distinguished by morphol- morphological analyses of these subspecies in a fi eld trial in a ogy (e.g., D . aureus, D. capillifolius, D. crinitus, D. glochidiatus , maritime environment, where D . carota subsp. gummifer (sensu D . involucratus, D. littoralis, D . muricatus , D. pusillus, Rouya lato) grows, as an extension and comparison to the study by polygama ). However, for some species, the taxonomy of Spooner et al. (2014), using additional collections of these sub- Daucus remains complicated by the lack of suffi cient germ- species. Until we have access to these data, and because of our plasm for defi nitive morphological and molecular analyses, desire to make taxonomic decisions that are stable, we are reluc- lack of comprehensive herbarium studies to associate names to tant to make comprehensive taxonomic changes now, following type specimens, unsettled generic affi liations, and undefi ned our desire to follow a phylogenetic species concept. species boundaries. Our combined molecular and morphologi- One taxonomic relationship is well supported, however. cal studies indicate particular problems in the subspecies of D . D aucus capillifolius is morphologically distinct and diagnosable carota , species distinctions of D . sahariensis and D. syrticus ( Spooner et al., 2014 ) (Appendix S2), yet nested in a clade of D . and the subgroups of D . broteri / D. guttatus . carota (A rbizu et al., 2014; Iorizzo et al., 2013 ) ( Fig. 1 ). It shares W e are pursuing these remaining problems with additional the same number of chromosomes as all subspecies of D. carota fi eldwork; collaborations with other D aucus investigators to (2 n = 18) and is fully intercrossable with the other subspecies share germplasm, DNA, and herbarium data; and additional fo- ( McCollum, 1975 , 1 977 ), supporting its inclusion within D . cused morphological and molecular studies. Arbizu et el. (2014) carota . It has its own range, confi ned to western Libya and adja- identifi ed a subset of nuclear orthologs that give a topology cent northeastern Tunisia. Coauthors P. W. Simon and D. M. nearly identical to the use of 94 nuclear orthologs, and we are Spooner collected D. capillifolius throughout most of its range in November 2014] ARBIZU ET AL.—MORPHOMETRICS OF DAUCUS 2015 Fig. 6. Canonical variates analysis of clade B species D aucus glochidiatus , D. guttatus (subsets 1, 2, 3), D . involucratus , D . littoralis , D. pusillus . northern Tunisia and have good knowledge of its variation. We for enhancing our understanding of genotype–phenotype relation- here recognize it as a subspecies of D . carota. ships and its relevance to crop improvement. Theoretical and Applied D aucus carota subsp. capillifolius (A. Gilli) C. Arbizu, Genetics 126 : 867– 887 . comb. et stat. nov. D aucus capillifolius A. Gilli, Oester- CONSTANCE , L. 1971 . History of the classifi cation of Umbelliferae reichische Botanische Zeitschrift 104: 574. 1958. (Apiaceae) . I n V. H. Heywood [ed.], The biology and chemistry of the Umbelliferae, 1–8, Academic Press, New York, New York, USA. T YPE: Libya, Tripolitania, hills northwest of Jefren (Yaf- C ULLEN , J . 1972. Daucus . In P. H. Davis [ed.], Flora of Turkey and the ran), 24 Jul 1956, in sandy soil with F oeniculum vulgare , Echi- East Aegean Islands, vol. 4, Flora of Turkey, 531–536. Edinburgh nops spinosus , and A ristida pungens , E lfrid Gerhart, s.n. University Press, Edinburgh, UK. (holotype: W). D HONDT, S. , N. WUYTS , AND D. I NZÉ . 2 013 . Cell to whole-plant phenotyp- ing: The best is yet to come. T rends in Plant Science 1 8: 4 28 –4 39 . LITERATURE CITED D OS REIS , M ., J. INOUE , M. HASEGAWA , R. J. ASHER , P . C. D ONOGHUE , AND Z. Y ANG . 2 012. P hylogenomic datasets provide both precision and ARBIZU, C. , H. R UESS, D . SENALIK, P. W. S IMON, AND D . M. SPOONER . 2014. accuracy in estimating the timescale of placental mammal phylog- Phylogenomics of the carrot genus (D aucus, Apiaceae) . A merican eny. Proceedings of the Royal Society, B, Biological Sciences 2 79: Journal of Botany 101 : 1 666 – 1685 . 3491 –3 500 . BURKI, F ., K. S HALCHIAN-TABRIZI , AND J. PAWLOWSKI. 2 008. Phylogenomics DOWNIE , S. R., D. S. K ATZ-DOWNIE , AND M . F. W ATSON . 2000 . A phy- reveals a new ‘megagroup’ including most photosynthetic eukaryotes. logeny of the fl owering plant family Apiaceae based on chloroplast Biology Letters 4: 366 – 369. DNA rpl16 and rpoc1 intron sequences: Towards a suprageneric clas- C HRISTELOVÁ , P . , M. VAL ÁRIK , E. HŘ IBOVÁ  , E. DE LANGHE , AND J. DOLEŽEL. sifi cation of subfamily Apioideae. A merican Journal of Botany 87 : 2 011. A multi gene sequence-based phylogeny of the Musaceae (ba- 2 73– 2 92. nana) family. B MC Evolutionary Biology 11: 103. E GAN, A., J . SCHLUETER , AND D. M. SPOONER. 2012 . Applications of C OBB, J. N., G . DECLERCK , A . G REENBERG, R. CLARK , AND S. M CCOUCH. next-generation sequencing in plant biology. A merican Journal of 2013 . Next-generation phenotyping: Requirements and strategies Botany 99 : 175 –1 85 . 2016 AMERICAN JOURNAL OF BOTANY FARIS , F . E. , AND M . IBN TATTOU . 2007 . U mbelliferae . I n M. Fennane, M. Ibn P IMENOV , M. G., AND M . V. LEONOV. 1993. T he genera of Umbelliferae: A Tattou, A. Ouyahya, and J. El Oualidi [eds.], Flore Pratique du Maroc, vol. nomenclator . Royal Botanic Gardens, Kew, UK. 2, Angiospermae (Leguminosae–Lentibuliaceae), Série Botanique no. P LUNKETT, G. M. , D. E. SOLTIS , AND P. S. SOLTIS . 1996. Evolutionary 38, 281–336. L’Institut Scientifi que, Rabat, Morocco. patterns in Apiaceae: Inferences based on m atK sequence data. FIORANI , F. , AND U. SCHURR. 2 013 . F uture scenarios for plant phenotyping. S ystematic Botany 2 1: 4 77 – 495 . A nnual Review of Plant Biology 64 : 267 –2 91 . P UJADAS SALVÀ, A. J. 2003 . D aucus. In G. Nieto Feliner, S. L. Jury, and G RZEBELUS , D. , R. BARANSKI, K . SPALIK , C. ALLENDER , AND P. W. SIMON . A. Herrero [eds.], Flora iberica: Plantas vasculares de la península 2011. Daucus. In C. Kole [ed.], Wild crop relatives: Genomic and Ibérica e islas Baleares, vol. 10, Alaliaceae-Umbelliferae, 97–125. breeding resources, vol. 5, Vegetables, 129–216. Springer-Verlag, Real Jardín Botá n ico, CSIC, Madrid, Spain. Berlin, Germany. QUEZEL , P ., AND S. S ANTA . 1963. Daucus . In P. Quezel and S. Santa [eds.], H EYWOOD , V. H. 1968. Daucus . F eddes Repertorium 7 9: 1 – 68 . Nouvelle fl ore de L’Algérie et regions désertiques méridionales, IORIZZO, M . , D. S ENALIK , S. ELLISON , D. GRZEBELUS , P. CAVAGNARO, 659–663. Éditions du Centre National de la Recherche Scientifi que, C. A LLENDER, J . BRUNET , D. M. SPOONER , A . V AN DEYNZE , AND P. W. Montpellier, France. S IMON . 2 013. Genetic structure and domestication of carrot ( Daucus REDURON , J. P. 2007. Ombellifères de France, vol. 2. Société Botanique carota subsp. sativus L.) (Apiaceae). A merican Journal of Botany du Centre-Ouest, Nercillac, France. 1 00 : 9 30 – 938 . R OKAS , A., B. L. WILLIAMS , N . K ING, AND S. B. CARROLL . 2 003. G enome- J AFRI , S. M. H., AND A. EL-GADI. 1 985 . D aucus. In S. M. H. Jafri and scale approaches to resolving incongruence in molecular phylogenies. A. El-Gadi [eds.], Flora of Libya, vol. 117, Apiaceae, 130–144, Nature 425 : 798– 804 . Al Faatheh University, Faculty of Science, Department of Botany, S Á ENZ LAÍN , C. 1981. R esearch on Daucus L. (Umbelliferae). A nales del Tripoli, Libya. Jardín Bot ánico de Madrid 37: 481– 534. J URY , J. L. 2002 . Daucus. In B. Valdés, M. Rejdali, A. Achhal El SMALL , E. 1978. A numerical taxonomic analysis of the Daucus carota Kadmiri, J. L. Jury, and J. M. Montserrat [eds.], Catalogue des plan- complex. C anadian Journal of Botany 56 : 248 –2 76 . tes vasculaires du Nord de Maroc, incluant des clés d’identifi cation, S OKAL , R. R., AND R . R. S NEATH . 1 963 . Principles of numerical taxonomy . vol. 1, 467–469. Consejo Superior de Investigaciones Cientiìfi cas, W. H. Freeman, San Francisco, California, USA. Madri, Spain. SPALIK , K. , AND S. R. DOWNIE . 2 007. I ntercontinental disjunctions in L EE, B. Y. , AND S. R. D OWNIE . 2000 . Phylogenetic analysis of cpDNA re- Cryptotaenia (Apiaceae, Oenantheae): An appraisal using molecular striction sites and r psl6 intron sequences reveals relationships among data. Journal of Biogeography 34 : 2039 – 2054 . Apiaceae tribes Caucalideae, Scandiceae and related taxa. P lant S POONER, D. M. , P . R OJAS , M. BONIERBALE, L . A. M UELLER , M . S RIVASTAV, Systematics and Evolution 221: 35 – 60 . D . SENALIK , AND P . SIMON . 2013 . Molecular phylogeny of D aucus. LE FLOC’H, É . , L . B OULOS , AND E. VÉLA . 2010 . Catalogue synonymique S ystematic Botany 38 : 850– 857 . commenté de la fl ore de Tunisie . Ministère de l’Environnement de SPOONER, D . M., M . P. WIDRLECHNER , K. R. REITSMA , D. E. PALMQUIST , Développment Durable Banque Nationale de Gènes, Tunis, Tunisia. S . R OUZ, Z. G HRABI-GAMMAR, M . NEFFATI, ET AL . 2014. Reassess- MCCOLLUM , G . D. 1975. Interspecifi c hybrid D aucus carota × D . capil- ment of practical subspecies identifi cations of the USDA D aucus lifolius. Botanical Gazette 136: 201 – 206. carota germplasm collection: Morphological data. C rop Science 54 : M CCOLLUM , G. D. 1977. Hybrids of D aucus gingidium with cultivated 706– 718 . carrots (D . carota subsp. s ativus ) and D . capillifolius. B otanical WEITZEL, C . , N. RØNSTED , K. S PALIK, AND H. T. S IMONSEN . 2 014. Gazette 1 38 : 5 6– 63 . R esurrecting deadly carrots: Towards a revision of T hapsia (Apiaceae) MOUTERDE , P . 1986 . N ouvelle fl ore du Liban et de la Syrie, part 2 . Dar based on phylogenetic analysis on nrITS sequences and chemical pro- El-Machreq Editeurs, Beirut, Lebanon. fi les . B otanical Journal of the Linnean Society 174: 6 20– 6 36 . O NNO, M . 1937 . Die wildformen von Daucus sect. Carota . Beihefte zum ZOHARY , M. 1972 . Flora Palaestina, part 2 . Israel Academy of Sciences Botanischen Centralblatt 5 6( B): 83 –1 36. and Humanities, Jerusalem, Israel.