Org Divers Evol (2012) 12:133–143 DOI 10.1007/s13127-012-0089-z

ORIGINAL ARTICLE

Refugia and geographic barriers of populations of the desert , fumariifolia ()

Eduardo Ruiz-Sanchez & Flor Rodriguez-Gomez & Victoria Sosa

Received: 18 July 2011 /Accepted: 22 March 2012 /Published online: 15 April 2012 # Gesellschaft für Biologische Systematik 2012

Abstract Phylogeographic data and divergence estimation Oriental and the Trans-Mexican Volcanic Belt. Furthermore, times as well as current and past ecological niche modeling all of these processes may have resulted in the patchy for the Mexican poppy, Hunnemannia fumariifolia distribution of suitable microhabitats for H. fumariifolia in Sweet, were combined in order to understand its biogeo- its geographical range. Ecological niche models constructed graphic history. Divergence times were estimated to deter- using the MIROC3 model indicated that populations did not mine if divergence occurred during the . move to the north but rather that they had suitable ecological Ecological niche modelling was used to determine if the last habitats in the Chihuahuan Desert, which harbored Pinus- glacial maximum (LGM) was responsible for the southward Juniperus forests during that period. movement of poppy populations into the Tehuacán-Cuicat- lán Valley. Analyses were performed to detect any geo- Keywords change . Ecological niche models . Last graphical barriers that might have caused genetic glacial maximum . Chihuahuan Desert . Sierra Madre discontinuities among populations across the entire range Oriental . Trans-Mexican Volcanic Belt. of distribution. Current and Pleistocene ecological niche models were created for H. fumariifolia using eight envi- ronmental variables derived from temperature and precipi- Introduction tation. The evidence shows that divergence of the three main clades in H. fumariifolia occurred from the Early Pleisto- An increasing number of biogeographic studies dealing with cene to Mid-Miocene. It was also found that gene flow a diverse array of from North America have identi- between the populations of H. fumariifolia could have been fied areas that served as refugia during the last glacial limited by the LGM, by during the Quater- maximum (LGM) in the Late Pleistocene. Studies combine nary, and by the complex topography of the Sierra Madre ecological niche modelling and molecular evidence to esti- mate the time of origin, divergence of populations, abiotic E. Ruiz-Sanchez (*) factors influencing to habitats and the historical University of , Berkeley, and Microbial , events that have had an effect on the evolutionary processes 431 Koshland Hall, of these taxa (e.g., Riddle et al. 2000; Carstens and Richards Berkeley, CA 94270, USA e-mail: [email protected] 2007; Castoe et al. 2007; McGuire et al. 2007; Knowles et : al. 2007; Waltari et al. 2007; Morris et al. 2008, 2010; F. Rodriguez-Gomez V. Sosa Cavender-Bares et al. 2011; Chan et al. 2011; Cosacov et Instituto de Ecología, A. C., Biología Evolutiva, al. 2010; Désamoré et al. 2011; Ornelas et al. 2010; Reber- Apartado Postal 63, 91000, Xalapa, Veracruz, México nig et al. 2010a). The utilization of palaeoclimatic models and ecological Present Address: niche models projected onto historical landscapes provides a E. Ruiz-Sanchez spatial context for phylogeographic analyses (Carstens and Instituto de Ecología, A. C., Centro Regional de Bajío, Av. Lázaro Cárdenas 253, Richards 2007; Waltari et al. 2007). These tools have been 61600, Pátzcuaro, Michoacán, México used widely to identify refugia during the Late Pleistocene 134 E. Ruiz-Sanchez et al.

LGM (Hugall et al. 2002; Carstens and Richards 2007; causing vicariance events that separated, for example, her- Marske et al. 2011). Neogene vicariance, due largely to petofauna, rodents and some plant populations from both orogenesis, and climate change have been postu- deserts (Riddle et al. 2000; Jaeger et al. 2005; Riddle and lated as drivers of evolutionary diversification in western North Hafner 2006; Castoe et al. 2007; Leaché and Mulcahy 2007; America (Riddle and Hafner 2006) and Mexico (Bryson et al. Bryson et al. 2010a; Rebernig et al. 2010a). The Chihua- 2010b). huan Desert also acted as a barrier between the Sierra Madre To understand the biogeographic history of the Mexican Occidental and Oriental for some gymnosperms, such as tulip poppy Hunnemannia fumariifolia, we combined diver- Pinus (Moreno-Letelier and Piñero 2009). gence time estimation and palaeoclimatic models to inves- The distribution of the Mexican tulip poppy is complex tigate if Neogene orogenesis and Quaternary climatic due to the presence of three mountain ranges: the Trans- change has influenced the distribution patterns of popula- Mexican Volcanic Belt in the south, the Sierra Madre Occi- tions of the Mexican tulip poppy. Our previous phylogeo- dental to the northwest and the in graphic study found that allopatric fragmentation had an north-eastern Mexico. In the middle of these mountains, the effect on genetic divergence in populations of the tulip Chihuahuan Desert is located on the . The poppy in the Sierra Madre Oriental, and that this divergence Sierra Madre Oriental has the most complicated geological may be a reflection of the complex of the area over history of the three formations, originating in the Laramide which this species is distributed (Sosa et al. 2009). More- formation during the Late Cretaceous to the Palaeogene over, our results suggested that the areas located in the north (80–55 Ma). This formation resulted from an orogenic event of the Sierra Madre Oriental acted as post-glacial refugia for that gave rise to the Rocky Mountain fold, the thrust belt in some populations of H. fumariifolia (Sosa et al. 2009). Canada, the Sierra Madre Oriental fold and the thrust belt in However, divergence time was not estimated for popula- Mexico (English et al. 2003). Erosion in the foothills of this tions, nor were further analyses conducted to confirm these mountain range occurred during the Palaeogene-Eocene and refugia or to discover the geographic barriers that prevented subsequently either during the Oligocene or the Neogene gene flow. (Roure et al. 2009). In contrast, the The tulip poppy is an herbaceous perennial, growing in resulted from volcano-tectonic events that occurred after the xerophytic habitats at middle elevations in Mexico. In the end of the Laramide orogeny and before the episode peaks north, populations are distributed in the Chihuahuan Desert of the Sierra Madre Occidental volcanic events in the Oli- and the Sierra Madre Oriental and, crossing the Trans- gocene. The volcano-tectonic peaks occurred in three main Mexican Volcanic Belt, there are populations in the south, episodes from the mid-late Oligocene to the early Miocene, in the Tehuacán-Cuicatlán Valley (Sosa et al. 2009). Hun- at about 32–30 Ma, 30–28 Ma and 26–25 Ma (Tristán- nemannia forms part of a North American clade in the González et al. 2009). The western area of the Trans- Papaveraceae, together with and Dendrome- Mexican Volcanic Belt originated during the Miocene and con (Hoot et al. 1997). the eastern area during the (Ferrari et al. 2000; An analysis of the biogeographic history of the populations García-Polomo et al. 2002). Palaeorecords from the Mio- of the Mexican tulip poppy may help to understand past cene and Pliocene at the end of the Tertiary (2–20 Ma) changes in plant distribution in the deserts of Mexico. It has indicate that plant communities reached elevated complexity been suggested that the habitats of the western highlands of the at that time, reflecting a warmer, more humid, and relatively Sierra Madre Oriental and the Chihuahuan Desert fluctuated stable climate compared to that of the Quaternary (Tausch et dramatically during the Pleistocene, and that this has resulted al. 1993). Furthermore, it has been postulated that the oro- in the cyclical downward displacement and retraction of the genesis that occurred from the Palaeogene to the Neogene pine-oak-juniper forests (Van Devender 1990; Metcalfe et al. and the Quaternary climate change were the drivers of plant 2000; Metcalfe 2006). As a result of forest shifts, populations diversification in North America (Bryson et al., 2010a). expanded their range during glacial periods and remained Our study focuses on several aspects of the evolutionary isolated in refugia at high elevations during interglacial peri- history of Hunnemannia fumariifolia populations: (1) esti- ods. Moreover, these events caused subsequent postglacial mating the time of divergence of its populations to deter- fragmentation that prevented gene flow (Bryson et al. 2010a). mine if divergence occurred during the Pleistocene; (2) During the Early Pliocene the Chihuahuan Desert, along establishing whether the LGM was responsible for the with the other North American deserts, attained its maxi- southward movement of populations into the Tehuacán- mum area but it decreased during the moist Late Pliocene Cuicatlán Valley; (3) detecting the geographical barriers that and during the Pleistocene pluvial intervals (Riddle and caused genetic discontinuities among populations across Hafner 2006). As a result of the uplift of the Sierra Madre this species’ entire range of distribution; and (4) identifying Occidental and the Mexican Plateau, the Chihuahuan Desert the geographic areas that served as refugia for its was separated from the Sonoran and Mojave Deserts, populations. Refugia and geographic barriers in Hunnemannia fumariifolia 135

Methods 2007), and evolutionary models were determined with jMo- delTest 0.1.1 (Posada 2008). Two estimates of divergence Sampling and DNA sequences time were worked out. First, divergence was assessed at the family level. The evolutionary model GTR + I + G was The DNA sequences of three plastid spacers: trnH-psbA selected based on the AICc result from jModelTest 0.1.1 (EF464658–EF464664), rpl32-trnL (UAG) (EU169024– (Posada 2008) for the combined chloroplast matrices (atpB EU169030) and ndhF-rpl32 (EU169018–EU169023) pub- and rbcL) and analyses were run under an uncorrelated lished previously by Sosa et al. (2009) corresponding to 17 lognormal relaxed clock model. The Yule speciation process populations with a total of 85 individuals were included in the was used as a prior to model the tree. We used five calibra- analyses. In addition, for molecular dating, DNA sequences tion points for this level, we treated all calibration points as from two plastid genes: atpB (U86384, U86386-U86401, minimum age constraints. Three secondary points derived AF293860, AF092116, AF092115, AF093396, AF093384, from the Bell et al. (2010) analysis were utilized, with a FJ026454, FJ026397, AF093375, DQ359689, AF093382, normal distribution. For the root node we used a mean age AF093393 and D8955) and rbcL (86621-86632; L01943, of 136 Ma, SD 2.8 (130–142 Ma), for the Eucotyledoneae a L01951, L08764, L12645, AF197599, AF093720, mean of 129 Ma, SD 3.2 (123–134 Ma), and for Ranuncu- AF093719, AF093731, AF093726, HQ260807, L37920, lales a mean of 100 Ma, SD 7.5 (85–115 Ma). We con- AF093723, DQ359689, L75849, AF093730 and strained Proteales with the fossil of Platanocarpus GQ997596) from 15 representative taxa of Papaveraceae were brookensis (98 Ma) (Bell et al. 2010) modelled with a mean used as the ingroup based on Hoot et al. (1997). For the of 1 and an offset of 98 (hard bound constraint), which outgroup, we chose representative taxa of Berberidaceae, equalled the minimum age of the fossil (Ho and Philips Circaeasteraceae, Eupetalaceae, Lardizabalaceae, Menisper- 2009). We then constrained the divergence of Menisperma- maceae and Ranunculaceae of Order , in addi- ceae from Ranunculaceae/Berberidaceae with the fossil of tion to Platanus occidentalis and Nelumbo nucifera of Prototinomiscium vangerowii (91 Ma) (Anderson et al. Proteales and Ceratophyllum submersum, based on a previous 2005) with a mean of 1 and an offset of 91. The second phylogenetic study by Bell et al. (2010). estimate of divergence time was conducted at the population level in Hunnemannia fumariifolia. We used the HKY + G Phylogenetic analyses model of sequence evolution and the three chloroplast spacers [trnH-psbA, rpl32-trnL(UAG) and ndhF-rpl32], un- Bayesian inference (BI) was conducted with MRBAYES v. der an uncorrelated lognormal relaxed clock model and 3.1.2 (Ronquist & Huelsenbeck 2003). The software jMo- coalescent model assuming exponential population growth. delTest 0.1.1 (Posada 2008) was run to determine the model To calibrate the root, we used the results of the first diver- of evolution that best fit using the AICc values for Papaver- gence time analysis. We used the mean (16.03 Ma; 95 % HDP aceae and the outgroup taxa for the combined matrices 0 4.6–30) divergence time of the separation between Hunne- (atpB, rbcL) was GTR + I + G and for each one of the three mannia-Eschscholzia. We used a lognormal distribution chloroplast markers (trnH-psbA 0 F81; rpl32-trnL (UAG) 0 with a mean of 2.78, SD 0.3, zero offset; range of 4.6-30. TIM1; ndhF-rpl32 0 TVM) for analyses at the population For both estimates four independent 107 generation runs level. Two different analyses were performed—the first for were performed with random starting trees, sampling every the family level based on the combined data matrix and the 1,000 generations. TRACER v. 1.5 (http://tree.bio.ed.ac.uk/ second based on the concatenated data matrix of H. fumar- software/tracer/) was used to assess convergence and effec- iifolia. Two independent runs were conducted to assess the tive sample sizes (ESS) for all parameters and also for repeatability of stationarity between runs for each analysis. combining tree files from the four runs performed with For each run, one cold and three heated chains were set for BEAST. Results were summarized in a single tree visualized 10,000,000 runs, sampling one tree every 1,000 generations. with FIGTREE v. 1.5.4 (http://tree.bio.ed.ac.uk/software/ Stationarity was determined based on the likelihood scores figtree/). for time to convergence and sample points collected prior to stationarity were eliminated (10 %). Posterior probabilities Geographic barriers for supported clades were determined by a 50 % majority- rule consensus of the trees retained after burn-in. BARRIER 2.2 (Manni et al. 2004) based on Monmonier’s algorithm (Monmonier 1973) was used to determine geo- Molecular dating graphic barriers within the Hunnemannia fumariifolia local- ities. These barriers represent zones of abrupt changes in the Divergence time was estimated with a Bayesian approach as pattern of genetic variation among sample populations in the implemented in BEAST v. 1.5.4 (Drummond and Rambaut presence of isolating factors () is likely to weaken 136 E. Ruiz-Sanchez et al. the gene flow by increasing the chances of finding signifi- r<0.7 based on all sample locations) (Peterson 2007;Naka- cant barriers (Manni et al. 2004). First the combined cpDNA zato et al. 2010). Two general atmospheric circulation models matrix was transformed into a distance matrix using the F84 (GCM) were used to generate past climate scenarios for LGM: model of nucleotide substitution implemented in DNADIST the Community Climate System Model (CCSM) and the and then SEQBOOT was used to generate 100 bootstrapped Model for Interdisciplinary Research on Climate (MIROC3). distance matrices from DNA sequences to evaluate support The original GCM data were downloaded from the for the observed barriers, both programs are included in the PMIP2 website (http:// www.pmip2.cnrs-gif.fr/)(Braconnot PHYLYP 3.69 package (Felsenstein 1989). et al. 2007). One Last Inter-Glacial (LIG; 120,000– 140,000 years BP) was used to generate past climate scenarios Ecological niche modelling for LIG; the data was downloaded from WorldClim (http:// www.worldclim.org/past). The models were run in Maxent Current and Pleistocene ENMs were created for Hunneman- 3.3.2 (Phillips et al. 2006; http://www.cs.princeton. edu/ nia fumariifolia using eight environmental variables derived ~schapire/maxent/). Maxent employs a maximum likelihood from temperature and precipitation data obtained from World- method that estimates the species’ distribution that has max- Clim 1.4 (Hijmans et al. 2005) (Mean diurnal range, Temper- imum entropy, subject to the constraint that the environmental ature seasonality, Temperature annual range, Mean variables for the predicted distribution must match the empir- temperature of warmest quarter, Mean temperature of coldest ical average (Elith et al. 2006; Phillips et al. 2006). quarter, Precipitation seasonality, Precipitation of wettest A total of 65 unique records with georeferences were quarter and Precipitation of coldest quarter) with a resolution compiled. Georeferenced data obtained from the previous of 1 km2. These variables are not highly correlated (pairwise phylogeographic study by Sosa et al. (2009) and data from

125 100 75 50 25 0Ma

Paleocene Eocene Oligocene Miocene Pli Ple Cretaceous Paleogene Neogene Menispermun candense 1.0 Tinospora esiangkara Ranunculus macranthus 0.96 1.0 Coptis trifolia 1.0 Hydrastis candensis

1.0 1.0 Glaucidium palmatum 1.0 Nandina domestica Sargentodoxa cuneata 0.78 0.67 Kingdonia uniflora

SELALUCLUNAR

EAENODELYTOCUE 1.0 Circaeaster agrestis Euptelea polyandra

Hunnemannia fumariifolia E/H 1.0 1.0 Dendromecon rigidum

Macleaya cordata EAECAREVAPAP Sanguinaria candensis 1.0 1.0 Glaucium flavum 1.0 1.0 Dicranostigma franchetiana 1.0 Stylophorum diphyllum coulteri Papaver orientalis 0.84 1.0 0.78 Platystemon californicus Argemone mexicana 1.0 Corydalis nobilis 0.98 Hypecoum imberba Pteridophyllum racemosum Platanus occidentalis PROTEALES 1.0 Nelumbo nucifera Ceratophyllum submersum

Paleogene Neogene Cretaceous Paleocene Eocene Oligocene Miocene Pli Ple 125 100 75 50 25 0Ma

Fig. 1 Chronogram of the Papaveraceae family based on a Bayesian Escholscholzia clade. Black stars secondary calibration points, black approach. Gray bars 95 % confidence intervals for node age estimates. crosses fossil calibration points. Numbers below the branches Bayesian Brackets identify Eucotyledoneae clade, Ranunculales clade, Proteales posterior probabilities (PP). Pli Pliocene, Ple Pleistocene, Ma Million clade, Papaveracae family clade and the H/E Hunnemannia/ years Refugia and geographic barriers in Hunnemannia fumariifolia 137 the "National Biodiversity Information Network" (REMIB; Eschscholzia sp B http://www.conabio.gob.mx/remib_ingles/doctos/remib_ Arteaga-14 Cd. Maiz-13 B ing.html; accessed May 2010). The Maxent logistic model C

Bonanza-2 C-T/HC output for a given species is a continuous surface of values 0.94 0.99 Real de Catorce-4 C ranging from 0 to 1, where high values indicate a higher Coixtlahuaca-6 N suitability for a given species. Maxent was run with default 0.99 N 1.0 Tequixtepec-17 settings, ensuring that we had only one locality per grid cell. Real de Catorce-4 O To evaluate model performance, we set aside a random Cd. Maiz-13 E subset of 25 % of the total unique records and mea- Cerro Tahti-5 D sured the area under the curve (AUC) of the receiver 0.99 Cienega-9 H operating characteristic, a threshold-independent measure Cd. Maiz-13 F J of performance. 0.97 Escondida-3 J 0.97 Zaragoza-12

1OMS Escondida-3 I 0.97 Mezquititlan-16 L Results 0.94 Tolantongo-15 L Phylogenetic analyses Cienega-9 G La Luz-8 G 1.0 San Isidro-7 G Results of the family level phylogenetic analysis are shown San Isidro-7 Q in Fig. 1. This chronogram is similar to the 50 % majority Mezquititlan-16 M consensus tree that resulted from the Bayesian analysis Altares-10 A

(posterior probabilities, PP are indicated). In this tree, Papa- Galeana-1 A 2OMS veraceae is depicted as monophyletic (1.0 PP); it also shows 0.95 Galeana-1 K that Hunnemannia is closely related to Eschscholzia (1.0 A 0.87 Zaragoza-12 PP). Papaveraceae is the sister group to the rest of the Rio San Jose-11 P families in the order Ranunculales (1.0 PP), and the repre- sentative taxa of Proteales was found to have an affinity Fig. 2 Bayesian 50 % majority-rule consensus tree based on concat- enated chloroplast [trnH-psbA, rpl32-trnL(UAG) and ndhF-rpl32] with Ranunculales (1.0 PP); representative taxa in Eucoty- spacer sequence data, showing the relationships among haplotypes ledoneae were the sister group to Ceratophyllum submersum from the populations of Hunnemannia fumariifolia. The three main without support. clades are identified in brackets: CH/T-C, Chihuahuan Desert/Tehua- The results of intra-specific phylogenetic relationships cán Cuicatlán Valley; SMO 1, Sierra Madre Oriental 1 and SMO 2, Sierra Madre Oriental 2. Circled letters at the end of the rows are the show the relationship between haplotypes of Hunnemannia haplotypes indicated in 1 of Sosa et al. (2009). Numbers below fumariifolia and are depicted in the 50 % majority consen- the branches Bayesian posterior probabilities (PP) sus tree shown in Fig. 2. Haplotypes are grouped into three main clades that received good support in an unresolved general topology. These clades correspond to the three areas of its distribution: Chihuahuan Desert, Tehuacán-Cuicatlán Papaveraceae is estimated to have diverged in the Mid-to- Valley and Sierra Madre Oriental. Within the Sierra Madre Late Cretaceous (mean096 Ma; 95 % HDP072–116). The clade Oriental two sub-clades were retrieved (SMO 1; 0.94 PP and formed by Hunnemannia-Eschscholzia is estimated to have SMO 2; 0.87 PP) and a clade grouped the haplotypes from diverged from the Early Oligocene to the Pliocene the Chihuahuan Desert and Tehuacán-Cuicatlán Valley (16.03 Ma; 95 % HDP04.6–30 MA) (Fig. 1). Divergence esti- CH/T-C; 1.0 PP). The only exception is a single haplotype mates for Hunnemannia fumariifolia using a coalescent from the Ciudad del Maíz (Chihuahuan Desert) nested with- model and assuming exponential population growth are in the SMO 1 clade, which shares haplotypes with the SMO shown in Fig. 4. The divergence of Hunnemannia fumarii- 1 and CH/T-V clades (Figs. 2, 3). The SMO 1 clade grouped folia occurred during the Early to Late Miocene (12.8 Ma; populations from the southern and northernmost localities of 95 % HDP06-20 Ma) (Fig. 4). Within the tulip poppy, the the Sierra Madre Oriental, while the SMO 2 clade has oldest clade is SMO 1 with a divergence time of 6.5 Ma populations only from the northern part of the SMO. (95 % HDP01.8-12.3), during the Mid-Pleistocene to Mid- Miocene. It was followed by the CH/T-V clade (5.5 Ma; Divergence times 95 % HDP01.2-11.1) and a similar divergence was found for the SMO and SMO 2 clades, which diverged during the Divergence time estimates at the family level, calculated Early Pleistocene to Late Miocene (4 Ma; 95 % HDP00.5- with the Yule speciation process, are shown in Fig. 1. 8.4) (Fig. 4). 138 E. Ruiz-Sanchez et al.

31o N Chihuahuan Desert 29o

S

27o

o 8 25 latn e dic c O7 erd9 a M arreiS 10 2 latn1 eir O erd a M arrei 11 3 SMO 2 4 12 23o 13 CH/T-V SMO 1

o 14 21 5 Transvolcanic Belt 15 16

19o

17 6 17o Tehuacán-Cuicatlán Valley

o 400500 200300 100 0Km 500 15

115o 113o 111o 109o 107o 105o 103o 101o 99o 97o 95o 93o 91o 89o

Fig. 3 Geographical distribution of the population of H. fumariifolia. Population numbers correspond to those in Sosa et al. (2009). Enclosed areas correspond to clades in Fig. 1

Geographic barriers Mexican Volcanic Belt (Fig. 6) where have never been collected. However, when the models were projected The geographic boundaries revealed by the program Barrier onto past climate (21 K) layers, two different scenarios were 2.2 are displayed in Fig. 5. There are several boundaries retrieved (Fig. 6). For the climate layers based on MIROC3, with the greatest bootstrap support (thick bars). Among a large area of suitable habitats in the central and northern them, a barrier separated the populations of Tehuacán-Cui- parts of the SMO and the central area of the Chihuahuan catlán from the rest of the populations in the SMO and CH Desert was predicted, with no connection to the Tehuacán- (Fig. 5) while another barrier separated the southern SMO Cuicatlán Valley. Predictions based on CCSM suggest a populations from those of the CH (Fig. 5). In addition, a slight connection between the Chihuahuan Desert-Sierra complex pattern of boundaries separated the rest of the Madre Oriental and the Tehuacán-Cuicatlán Valley, and also populations of the SMO from the populations in CH predicted an area of suitable habitat in the northernmost part (Fig. 5). Fragmentation was particularly evident in the of the Chihuahuan Desert. The model projected onto 120- SMO, with several barriers separating populations from 140 K revealed a highly fragmented scenario, including those in the southern, central and northernmost parts of the areas of suitable habitats in the Trans-Mexican Volcanic SMO (Fig. 5). Belt and in the Tehuacán-Cuicatlán Valley; however, no areas were predicted in the SMO or in the Chihuahuan Ecological Niche Modelling Desert.

The area under the curve (AUC) was 0.983, and the mod- elled distribution corresponds to the known distribution of Discussion H. fumariifolia. Ecological niche modelling (ENM) for the current climate variables predicted an accurate distribution Our phylogenetic reconstruction for Papaveraceae and relat- of H. fumariifolia, with the exception of areas in the north- ed families based on two chloroplast genes retrieved a well eastern SMO and one in the eastern part of the Trans- resolved tree, with a topology similar to analyses based on a Refugia and geographic barriers in Hunnemannia fumariifolia 139

20 17.5 15 12.5 10 7.5 5 2.5 0Ma Miocene Pliocene Pleistocene Hol Cd. Maiz-13 B Arteaga-14 B Bonanza-2 C Real de Catorce-4 C C-T/HC Real de Catorce-4 O Coixtlahuaca-6 N Tequixtepec-17 N Cd. Maiz-13 E Cerro Tahti-5 D Cienega-9 H Mezquititlan-16 M Cienega-9 G San Isidro-7 G

1OMS La Luz-8 G San Isidro-7 Q Mezquititlan -16 L Tolantongo-15 L Escondida-3 I Zaragoza-12 J Escondida-3 J Cd. Maiz-13 F Galeana-1 A

Galeana-1 K 2OMS Zaragoza-12 A Altares-10 A Rio San Jose-11 P

Miocene Pliocene Pleistocene Hol 20 17.5 15 12.5 10 7.5 5 2.5 0Ma

Fig. 4 Chronogram based on Bayesian approach of the H. fumariifolia the end of the rows are the haplotypes indicated in Table 1 of Sosa et al. phylogeography. Gray bars 95 % confidence intervals for node age (2009). Hol Holocene, Ma Million years estimates. Brackets identify clades as given in Fig 2. Circled letters at larger number of molecular markers (Wang et al. 2009; Bell Our estimate of divergence times for the Papaveraceae et al. 2010). based on five calibration points (three secondary points and The Mexican tulip poppy populations were grouped into two fossils) indicate a mean age of 96 Ma (95 % HDP072–116). three main clades that clearly share geographical areas The estimations of Anderson et al. (2005) that included this (Fig. 3). The southernmost populations were grouped in the group in their analyses of using penalized likeli- Sierra Madre Oriental (Clade 1), while the northern popula- hood and nonparametric rate smoothing found a mean age tions of the Sierra were in Clade 2, and the Chihuahuan Desert for the stem and crown groups of 114–106 Ma and 121–119, and Tehuacán-Cuicatlán Valley populations formed part of a respectively. Furthermore, Bell et al. (2010) found a mean more inclusive clade (CH/T-C). The linkage between the ageof88Ma(95 % HDP070–106 Ma) using a Bayesian populations of the Chihuahuan Desert and the northern pop- method (BEAST). Our results are more similar to those of ulations of the Sierra Madre Oriental corresponded to a hap- Bell et al. (2010). However, as stressed by Graur and Martin lotype from Ciudad del Maíz (13). These results coincide with (2004), the associated error must be considered when using previous findings for taxa from North American deserts (Baja secondary calibration and fossil points, as well as the error California, , and the Chihuahuan, Mojave bars around each node. Our estimate of the divergence time and Sonoran deserts). Those studies found patterns in the between Hunnemannia and Eschscholzia indicates that it distribution of the phylogroups: Eastern (Chihuahuan Desert occurred during the Early Oligocene to the Pliocene and Western of the Sierra Madre Oriental) vs Eastern (Baja (16.03 Ma; 95 % HDP04.6–30 Ma), coinciding with the last California, Colorado Plateau, Mojave and Sonoran deserts) volcanic-tectonic peak in the Sierra Madre Occidental, 26– groups (Riddle et al. 2000;Jaegeretal.2005; Leaché and 25 Ma (Tristán-González et al. 2009). The rise of the Sierra Mulcahy 2007;Castoeetal.2007; Moore and Jansen 2007; Madre Occidental could be responsible for the separation of Rebernig et al. 2010a, b). populations of Hunnemannia in the Chihuahuan Desert and 140 E. Ruiz-Sanchez et al.

(Sosa et al. 2009). The ENM reconstructions of the Pleisto- cene LGM based on two different climate scenarios (CCSM and MIROC3) performed for this study with populations of the Mexican tulip poppy, agree with the refugia identified by molecular data. However, the results of the MIROC3 model latn eir O erd a M arreiS suggested that populations had suitable ecological habitats

8 in the Chihuahuan Desert, where there were Pinus-Juniperus 7 9 Chihuahuan Desert 10 forests during that period, whereas the latter were not located 2 1 11 Bootstrap values in the north (Metcalfe et al. 2000, 2002; Lozano-Garcia et al. 3 91-100 4 12 2002; Bryson et al. 2010a, b). Recent phylogeographic studies 81-90 70-80 with desert plants (e.g., Euphorbia lomelii and 40-60 13 leucanthum) based on molecular data and ENM (Garrick et al. 2009; Rebernig et al. 2010a), demonstrated the movement of 14 populations from northern to southern refugia, contrary to our 5 15 tle B cin a clo v s n16 arT results, which showed that H. fumariifolia populations moved to western and northern refugia. Other phylogeographic stud- ies of trees and vertebrates in North America, however, show a common pattern of movement from north to south (Carstens and Richards 2007; Waltari et al. 2007;Morrisetal.2010; Tehuacán-Cuicatlán Valley 17 6 Cavender-Bares et al. 2011). Recent studies based on fossil and molecular data of Pseudotsuga menziesii indicate vertical migration in the southern Rocky Mountains and California (Gugger and Sugita 2010;Guggeretal.2010). It is possible Fig. 5 Geographic breaks identified in the distribution of H. fumar- iifolia using Monmonier’s algorithm. The sampling populations of H. that the pattern of migration of the tulip poppy populations fumariifolia are outlined by filled circles, with breaks as recovered by were similar in the northern SMO and in the west in the CH; the software program Barrier 2.2 (Manni et al., 2004) indicated with however, fossils are needed to test this hypothesis. black bars. The confidence level of the barrier is indicated by the Sosa et al. (2009) found that the historical processes that weight of the line, with heavy lines indicating the best-supported breaks as determined by analyses run on boot-strapped distance matri- influence the geographical pattern of genetic variation in H. ces. Population numbers correspond to those in Sosa et al. (2009) fumariifolia could be result of the complex geologic history of the SMO, while the refugia could be a consequence of climate change in the Pleistocene. Here, we used the same Sierra Madre Oriental from Eschscholzia in the Sonoran data and combined divergence estimates and palaeoclimatic Desert and the California Province where most of the species models to test these hypotheses, and found that Neogene of this are found (Hoot et al. 1997). The mean divergence orogenesis in the SMO could be the historical process time for each of the three main clades in H. fumariifolia responsible for the geographical pattern. Furthermore, Qua- indicates divergence occurred from the Early Pleistocene to ternary climate change might have had an influence on Mid-Miocene. This coincides with the Quaternary glacial- populations of H. fumariifolia, which remained in in-situ interglacial climate cycles and with the contraction and expan- refugia in the north of the SMO. Using molecular dating and sion of the pine-oak forest in the Sierra Madre Oriental, the ecological niche modelling with the Crotalus triseriatus Sierra Madre Occidental and the Chihuahuan Desert (Metcalfe group, Bryson et al. (2010b) found similar results, indicating et al. 2000, 2002; Lozano-Garcia et al. 2002;Brysonetal. that the Neogene orogenesis and Quaternary climatic 2010a, b). However, there are other geological/climatic events change drove the evolutionary diversification of this group that could explain the H. fumariifolia divergence and the sep- of snakes. aration of Hunnemannia-Eschscholzia, between them is the Results of the Barrier Analysis indicated that the Trans- erosion of the foothills of the SMOr during the Oligocene to Mexican Volcanic Belt was the geographic barrier that sep- the Neogene (Roure et al. 2009), the aridification during the arated the populations of the Tehuacán-Cuicatlán Valley Pliocene and the major temperature decline in the mid-Miocene from those in the SMO and the Chihuahuan Desert. Two (Braun 1950;Graham1999). This same period also coincides additional geographic barriers were detected in the SMO: with the divergence of other desert and highland taxa in North the Río Pánuco River Basin and the Cerritos-Arista and America (Riddle et al. 2000; Leaché and Mulcahy 2007;Castoe Saladan Filter Barriers. These same three geographical bar- et al. 2007;Brysonetal.2010a, b;Rebernigetal.2010a, b). riers were also identified for populations of the snake Cro- Our previous results with populations of H. fumariifolia talus triseriatus (Bryson et al. 2010a). The results of found molecular evidence of northern refugia in the SMO Monmonier’s algorithm for populations of the Mexican tulip Refugia and geographic barriers in Hunnemannia fumariifolia 141

Fig. 6 a–d. Ecological niche models for H. fumariifolia populations. a past climate conditions (LGM; MIROC3). d Prediction projected onto Prediction of suitable habitat in the current environment. b Prediction past climatic layers during an interglacial period (120–140 ka). Filled projected onto past climatic layers (LGM; CCSM). c Prediction under circles sampling sites poppy detected even more geographic barriers preventing all of these processes may have resulted in the patchy gene flow. A barrier separated Bonanza-2 populations from distribution of suitable microhabitats for H. fumariifolia. the Real de Catorce-4 population in the Chihuahuan Desert. ENM with the MIROC3 model indicated that populations Another separated populations in the central area from the did not move to the north but rather that they had suitable northern area of the SMO, and a barrier coinciding with the ecological habitats in the Chihuahuan Desert, where Pinus- southernmost part of the Sierra Gorda separated the southern Juniperus forests existed during that period. populations of the SMO from those of the Chihuahuan Desert. For example, the uplift of the Cascades and the Acknowledgments We are particularly grateful to Sasa Stefanovic Sierra Nevada caused the vicariance between the Rocky and two anonymous reviewers, whose comments improved this man- Mountain and the west coast populations of Pseudotsuga uscript significantly. We thank Bianca Delfosse for revising the English menziesii (Gugger et al. 2010). Thus, we suggest that the and Tania Hernandez for help with the molecular dating analyses. complex orogeny of the Sierra Madre Oriental that arose at different periods (English et al. 2003; Roure et al. 2009) gave rise to a diverse array of geographical barriers prevent- References ing gene flow in populations of H. fumariifolia (Fig. 5). Anderson, C. L., Bremer, K., & Friis, E. M. (2005). Dating phyloge- netically basal eudicots using rbcL sequences and multiple fossil reference points. American Journal of , 92, 1737–1748. Conclusions Bell, C. D., Soltis, D. E., & Soltis, P. S. (2010). The age and diversi- fication of the angiosperm re-revisited. American Journal of Bot- – In this study we found evidence that the divergence of the any, 97, 1296 1303. Braconnot, P., Otto-Bliesner, B., Harrison, S., Joussaume, S., three main clades of populations in H. fumariifolia occurred Peterchmitt, J.-Y., Abe-Ouchi, A., Crucifix, M., Driesschaert, from Early Pleistocene to Mid-Miocene. Gene flow between E., Fichefet, T., Hewitt, C. D., et al. (2007). Results of the populations of H. fumariifolia was also found to be PMIP2 coupled simulations of the Mid-Holocene and Last — limited by the LGM, climate change during the Quaternary, Glacial Maximum Part 1: experiments and large-scale fea- tures. Climate of the Past, 3,261–277. the complex topography caused by the Neogene orogenesis Braun, E. L. (1950). Deciduous forests of Eastern North America. of the SMO and the Trans-Mexican Volcanic Belt, and that Philadelphia: Blakiston. 142 E. Ruiz-Sanchez et al.

Bryson, R. W., Nieto-Montes, A., Jaeger, J. R., & Riddle, B. R. (2010). Graur, D., & Martin, W. (2004). Reading the entrails of chickens: Elucidation of cryptic diversity in a widespread Nearctic treefrog molecular timescales of evolution and the illusion of precision. reveals episodes of mitochondrial gene capture as frogs diversi- Trends in Genetics, 20,80–86. fied across a dynamic landscape. Evolution, 64, 2315–2330. Gugger, P. F., & Sugita, S. (2010). Glacial populations and postglacial Bryson, R. W., Jr., Murphy, R. W., Lathrop, A., & Lazcano-Villareal, migration of Douglas-fir based on fossil pollen and macrofossil D. (2010). Evolutionary drivers of phylogeographical diversity in evidence. Quaternary Science Reviews, 29, 2052–2070. the highlands of Mexico: a case study of the Crotalus triseriatus Gugger, P. F., Sugita, S., & Cavender-Bares, J. (2010). Phylogeogra- species group of montane rattlesnakes. Journal of , phy of Douglas-fir based on mitochondrial and chloroplast DNA 38, 697–710. sequences: testing hypotheses from the fossil record. Molecular Carstens, B. C., & Richards, C. L. (2007). Integrating coalescent and Ecology, 19, 1877–1897. ecological niche modeling in comparative phylogeography. Evo- Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones, P. G., & Jarvis, A. lution, 61, 1439–1454. (2005). Very high resolution interpolated climate surfaces for Castoe, T. A., Spencer, C. L., & Parkinson, C. L. (2007). Phylogeo- global land areas. International Journal of , 25, graphic structure and historical demography of the western dia- 1965–1978. mondback rattlesnake (Crotalus atrox): a perspective on North Ho, S. Y. W., & Philips, M. J. (2009). Accounting for calibration American desert biogeography. Molecular Phylogenetics and uncertainty in phylogenetic estimation of evolutionary divergence Evolution, 42, 193–212. times. Systematic Biology, 58, 367–380. Cavender-Bares, J., Gonzalez-Rodriguez, A., Pahlich, A., Koehler, K., Hoot, S. B., Kadereit, J. W., Blattner, F. B., Jork, K. B., Schwarzbach, & Deacon, N. (2011). Phylogeography and climatic niche evolu- A. E., & Crane, P. R. (1997). Data congruence and phylogeny of tion in live oaks (Quercus series Virentes) from the tropics to the the Papaveraceae s.l., based on four data sets: atpB and rbcL temperate zone. Journal of Biogeography, 38, 962–981. sequences, trnK restriction sites, and morphological characters. Chan, L. M., Brown, J. L., & Yoder, A. D. (2011). Integrating statis- Systematic Botany, 22, 575–590. tical genetic and geospatial methods brings new power to phylo- Hugall, A., Moritz, C., Moussalli, A., & Stanisic, J. (2002). Reconcil- geography. Molecular Phylogenetics and Evolution, 59, 523–537. ing paleodistribution models and comparative phylogeography in Cosacov, A., Sérsic, A. N., Johnson, L., Sosa, V., & Cocucci, A. A. the wet tropics rainforest land snail Gnarosophia bellendenker- (2010). Multiple periglacial refugia in the Patagonian steppe and ensis. Proceedings of the National Academy of Sciences USA, 99, post-glacial colonization of the Andes: the phylogeography of 6112–6117. Calceolaria polyrhiza. Journal of Biogeography, 37, 1463–1477. Jaeger, J. R., Riddle, B. R., & Bradford, D. F. (2005). Cryptic Neogene Désamoré, A., Laenen, B., Devos, N., Popp, M., González-Mancebo, vicariance and Quaternary dispersal of the red-spotted toad (Bufo J. M., Carine, M., & Vanderpoorten, A. (2011). Out of Africa: punctatus): insights on the evolution of North American warm north-westwards Pleistocene expansions of the heather Erica desert biotas. Molecular Ecology, 14, 3033–3048. arborea. Journal of Biogeography, 38, 164–176. Knowles, L. L., Carstens, B. C., & Keat, M. L. (2007). Coupling genetic Drummond, A. J., & Rambaut, A. (2007). BEAST: Bayesian evolu- and ecological-niche models to examine how past population dis- tionary analysis by sampling trees. BMC Evolutionary Biology, 7, tributions contribute to divergence. Current Biology, 17,1–7. 214. Leaché, A. D., & Mulcahy, D. G. (2007). Phylogeny, divergence times Elith, J., Graham, C. H., Anderson, R. P., Dudik, M., Ferrier, S., and species limits of spiny lizards (Sceloporus magister species Guisan, A., Hijmans, R. J., Huettmann, F., Leathwick, J. R., group) in western North American deserts and Baja California. Lehmann, A., Li, J., Lohmann, L. G., Loiselle, B. A., Manion, Molecular Ecology, 16, 5216–5233. G., Moritz, C., Nakamura, M., Nakazawa, Y., Overton, J. M., Lozano-García, S., Ortega-Guerrero, B., & Sosa-Nájera, S. (2002). Peterson, A. T., Phillips, S. J., Richardson, K., Scachetti-Pereira, Mid- to Late-Wisconsin pollen record of San Felipe Basin, Baja R., Schapire, R. E., Soberon, J., Williams, S., Wisz, M. S., California. Quaternary Research, 58,84–92. Zimmermann, N. E., et al. (2006). Novel methods improve pre- Manni, F., Guérard, E., & Heyer, E. (2004). Geographic patterns of diction of species’ distributions from occurrence data. Ecography, (genetic, morphologic, linguistic) variation: how barriers can be 29, 129–151. detected by using Monmonier’s algorithm. Human Biology, 76, English, J. M., Johnston, S. T., & Wang, K. (2003). Thermal modeling 173–190. of the Laramide orogeny: Testing the flat-slab subduction hypoth- Marske, K. A., Leschen, R. A. B., & Buckley, T. R. (2011). Reconcil- esis. Earth and Planetary Science Letters, 214, 619–632. ing phylogeography and ecological niche models for New Zea- Felsenstein, J. (1989). PHYLIP – Phylogeny Inference Package (Ver- land beetles: looking beyond glacial refugia. Molecular sion 3.2). Cladistics, 5, 164–166. Phylogenetics and Evolution, 59,89–102. Ferrari, L., Conticelli, S., Vaggelli, C., Petrone, C., & Manetti, P. McGuire, J., Linkem, C., Koo, M., Hutchison, D., Lappin, K., Orange, (2000). Late Miocene mafic volcanism and intra-arc tectonics D., Lemos-Espinal, J., Riddle, B., & Jaeger, J. (2007). Mitochon- during the early development of the Trans-Mexican Volcanic Belt. drial introgression and incomplete lineage sorting through space Tectonophysics, 318, 161–185. and time: phylogenetics of crotaphytid lizards. Evolution, 61, García-Palomo, A., Macías, J. L., Tolson, G., Valdez, G., & Mora, J. C. 2879–2897. (2002). Volcanic stratigraphy and geological evolution of the Metcalfe, S. E. (2006). Late Quaternary environments of the northern Apan region, east-central sector of the Trans-Mexican Volcanic deserts and central Transvolcanic Belt of Mexico. Annals of the Belt. Geofísica Internacional, 41, 133–150. Missouri Botanical Garden, 93, 258–273. Garrick, R. C., Nason, J. D., Meadows, C. A., & Dyer, R. J. (2009). Metcalfe, S. E., O’Hara, S. L., Caballero, M., & Davies, S. J. (2000). Not just vicariance: phylogeography of a Sonoran Desert euphorb Records of Late Pleistocene–Holocene climatic change in Mexico— indicates a major role of range expansion along the Baja penin- areview.Quaternary Science Reviews, 19,699–721. sula. Molecular Ecology, 18, 1916–1931. Metcalfe,S.E.,Say,A.,Black,S.,McCulloch,R.,&O’Hara, S. Graham, A. (1999). Studies in Neotropical paleobotany. XIII. An (2002). Wet conditions during the last glaciation in the Chi- Oligo-Miocene palynoflora from Simojovel (Chiapas, Mexico). huahuan Desert, Alta Babicora Basin, Mexico. Quaternary American Journal of Botany, 86,17–31. Research, 57,91–101. Refugia and geographic barriers in Hunnemannia fumariifolia 143

Monmonier, M. S. (1973). Maximum-difference barriers: an alternative perspectives on the history of a core North American warm numerical regionalization method. Geographical Analysis, 5, 245–261. deserts biota. Journal of Arid Environments, 66, 435–461. Moore, M. J., & Jansen, R. K. (2007). Origins and biogeography of Riddle, B. R., Hafner, D. J., & Alexander, L. F. (2000). Comparative gypsophily in the Chihuahuan Desert plant group Tiquilia Subg. phylogeography of Baileys’ pocket mouse (Chaetodipus bailey) Eddya (Boraginaceae). Systematic Botany, 32, 392–414. and the Peromyscus eremicus species group: historical vicariance Moreno-Letelier, A., & Piñero, D. (2009). Phylogeographic structure of the Baja California Peninsular Desert. Molecular Phylogenetics of Pinus strobiformis Engelm. across the Chihuahuan Desert and Evolution, 17, 161–172. filter-barrier. Journal of Biogeography, 36, 121–131. Ronquist, F., & Huelsenbeck, J. P. (2003). MrBayes 3: Bayesian Morris, A. B., Ickert, S. M., Brunson, B. D., Soltis, D. E., & Soltis, P. phylogenetic inference under mixed models. Bioinformatics, 19, S. (2008). Phylogeographical structure and temporal complexity 1572–1574. in American sweetgum (Liquidambar styraciflua; Altingiaceae). Roure, F., Alzaga-Ruiz, H., Callot, J. P., Ferket, H., Granjeon, D., Molecular Ecology, 17, 3889–3900. Gonzalez-Mercado, G. E., Guilhaumou, N., Lopez, M., Mougin, Morris, A. M., Graham, C. H., Soltis, D. E., & Soltis, P. S. (2010). P., Ortuño-Arzate, S., & Séranne, M. (2009). Long lasting inter- Reassessment of phylogeographical structure in an eastern North actions between tectonic loading, unroofing, post-rift thermal American tree using Monmonier’s algorithm and ecological niche subsidence and sedimentary transfer along the eastern margin of modelling. Journal of Biogeography, 37, 167–1667. the Gulf of Mexico: Some insights from integrated quantitative Nakazato, T., Warren, D. L., & Moyle, L. C. (2010). Ecological and studies. Tectonophysics, 475, 169–189. geographic modes of species divergence in wild tomatoes. Amer- Sosa, V., Ruiz-Sanchez, E., & Rodriguez-Gomez, F. C. (2009). Hidden ican Journal of Botany, 97, 680–693. phylogeographic complexity in the Sierra Madre Oriental: the Ornelas, J. F., Ruiz-Sanchez, E., & Sosa, V. (2010). Phylogeography of case of the Mexican tulip poppy Hunnemannia fumariifolia Podocarpus matudae (Podocarpaceae): pre-Quaternary relicts in (Papaveraceae). Journal of Biogeography, 36,18–27. northern Mesoamerican cloud forests. Journal of Biogeography, Tausch, R. J., Wigand, P. E., & Burkhardt, J. W. (1993). Viewpoint: 37, 2384–2396. Plant community thresholds, multiples steady states successional Peterson, A. T. (2007). Why not WhyWhere: the need for more pathways: legacy of the Quaternary? Journal of Range Manage- complex models of simpler environmental spaces. Ecological ment, 46, 439–447. Modelling, 203, 527–530. Tristán-González, M., Aguirre-Díaz, G. J., Labarthe-Hernández, G., Phillips, S. J., Anderson, R. P., & Schapire, R. E. (2006). Maximum Torres-Hernández, J. R., & Hervé Bellon, H. (2009). Post- entropy modeling of species geographic distributions. Ecological Laramide and pre-Basin and Range deformation and implications Modelling, 190, 231–259. for Paleogene (55–25 Ma) volcanism in central Mexico: A geo- Posada, D. (2008). jModelTest: Phylogenetic model averaging. Molec- logical basis for a volcano-tectonic stress model. Tectonophysics, ular Biology and Evolution, 25, 1253–1256. 471, 136–152. Rebernig, C. A., Schneeweiss, G. M., Bardy, K. E., Schönswetter, P., Van Devender, T. R. (1990). Late Quaternary vegetation and climate of Villaseñor, J. L., Obermayer, R., Stuessy, T. F., & Weiss- the Chihuahuan Desert, United States and Mexico. In Betancourt Schneeweiss, H. (2010a). Multiple Pleistocene refugia and Holo- JL, T. R. Van Devender, & P. S. Martin (Eds.), Packrat Middens: cene range expansion of an abundant south-western American The last 40,000 years of biotic change (pp. 104–133). Tucson: desert plant species (Melampodium leucanthum, ). Mo- University of Arizona. lecular Ecology, 19, 3421–3443. Waltari, E., Hijmans, R. J., Peterson, A. T., Nyari, A. S., Perkins, S. L., Rebernig, C. A., Weiss-Schneeweiss, H., Schneeweiss, G., Schönswetter, & Guralnick, R. P. (2007). Locating Pleistocene refugia: compar- P., Obermayer, R., Villaseñor, J. L., & Stuessy, T. F. (2010b). ing phylogeographic and ecological niche model predictions. Quaternary range dynamics and polyploid evolution in an arid PLoS ONE, 2, e563. brushland plant species (Melampodium cinereum, Asteraceae). Mo- Wang, W., Lu, A. M., Ren, Y., Endress, M. E., & Chen, Z. D. (2009). lecular Phylogenetics and Evolution, 54,594–606. Phylogeny and classification of Ranunculales: evidence from four Riddle, B. R., & Hafner, D. J. (2006). A step-wise approach to molecular loci and morphological data. Perspectives in Plant integrating phylogeographic and phylogenetic biogeographic Ecology, Evolution and Systematics, 11,81–110.