American Journal of 98(1): 109–121. 2011.

I NFLUENCES OF LANDSCAPE AND POLLINATORS ON POPULATION GENETIC STRUCTURE: EXAMPLES FROM THREE PENSTEMON (PLANTAGINACEAE) IN THE GREAT BASIN 1

Andrea T. Kramer2,3,4,5 , Jeremie B. Fant 3 , and Mary V. Ashley4

2 Botanic Gardens Conservation International (U.S.), 1000 Lake Cook Road, Glencoe, Illinois 60022 USA; 3 Chicago Botanic Garden, 1000 Lake Cook Road, Glencoe, Illinois 60022 USA; and 4 University of Illinois at Chicago, 845 West Taylor Street, M/C 066, Chicago, Illinois 60607 USA

• Premise of the study : Despite rapid growth in the fi eld of landscape genetics, our understanding of how landscape features in- teract with life history traits to infl uence population genetic structure in species remains limited. Here, we identify popula- tion genetic divergence in three species of Penstemon (Plantaginaceae) similarly distributed throughout the Great Basin region of the western United States but with different pollination syndromes (bee and hummingbird). The Great Basin’ s mountainous landscape provides an ideal setting to compare the interaction of landscape and dispersal ability in isolating populations of dif- ferent species. • Methods : We used eight highly polymorphic microsatellite loci to identify neutral population genetic structure between popula- tions within and among mountain ranges for eight populations of P. deustus, 10 populations of P. pachyphyllus, and 10 popula- tions of P. rostrifl orus . We applied traditional population genetics approaches as well as spatial and landscape genetics approaches to infer genetic structure and discontinuities among populations. • Key results : A ll three species had signifi cant genetic structure and exhibited isolation by distance, ranging from high structure

and low inferred gene fl ow in the bee-pollinated species P. deustus (F ST = 0.1330, R ST = 0.4076, seven genetic clusters identi- fi ed) and P. pachyphyllus ( F ST = 0.1896, R ST = 0.2531, four genetic clusters identifi ed) to much lower structure and higher in- ferred gene fl ow in the hummingbird-pollinated P. rostrifl orus ( F ST = 0.0638, R ST = 0.1116, three genetic clusters identifi ed). • Conclusions : These three Penstemon species have signifi cant yet strikingly different patterns of population genetic structure, fi ndings consistent with different interactions between landscape features and the dispersal capabilities of their pollinators.

Key words: gene fl ow; landscape genetics; microsatellite; Penstemon ; pollination syndrome; population genetic structure.

Landscape features interact with life history traits to either Eucalyptus globulus (Mimura et al., 2009), loss of habitat con- enhance or truncate gene fl ow in ways that are not always pre- nectivity unexpectedly enhanced gene fl ow, because primary dictable. The pattern and degree of population divergence will pollinators traveled greater distances between fragmented habi- depend largely on realized gene fl ow ( Slatkin, 1985 ), which in tats. Signifi cant differences in gene fl ow, and corresponding is determined by the composition, confi guration, and ma- effects on genetic divergence, can therefore exist among popu- trix quality of the landscapes they inhabit (Manel et al., 2003; lations of the same species occupying different landscapes as Storfer et al., 2007 ; Holderegger and Wagner, 2008 ), as well as well as among different species occupying the same landscape. life history traits such as pollination system and dispersal sys- For most species, the interacting effects of landscape features tem ( Hamrick and Godt, 1996 ; Richards, 1997 ; Duminil et al., and life history traits on population genetic structure are poorly 2007 ). In animal-pollinated plants, landscape connectivity and understood, but the growing fi eld of landscape genetics is be- pollinator movement can affect population genetic structure in ginning to address this gap ( Storfer et al., 2010 ). unpredictable ways. For example, in two herbaceous species, The use of neutral genetic markers and traditional estimates Lantana camara (butterfl y pollinated) and Rudbeckia hirta (hy- of genetic structure like FST ( Weir and Cockerham, 1984 ) and menoptera pollinated), greater habitat connectivity predictably Nei’ s genetic distance (Nei, 1978) provide insight into popula- enhanced gene fl ow by facilitating greater pollinator movement tion genetic divergence, as do more recent Bayesian clustering among populations, decreasing population genetic divergence methods, which do not require a priori assignment of individu- (Townsend and Levey, 2005). However, in the tropical forest als to populations and thus allow cryptic population genetic excelsa ( Dick et al., 2003 ) and temperate forest tree structure to be identifi ed (Pritchard et al., 2000). Genetic dis- tance between populations can be graphically represented via 1 Manuscript received 24 June 2010; revision accepted 22 November 2010. hierarchical genetic clustering methods such as UPGMA The authors thank K. Havens and P. Olwell for project support, and ( Sneath and Sokal, 1973 ), or it can be combined with geo- C. Newton, R. Tonietto, C. Flower, L. Jefferson, S. Karumuthil-Melethil, graphic distance information to detect patterns of isolation by E. Lukina, E. Sirkin, and J. Keller for fi eld and laboratory assistance. They distance. In landscapes where topographic features such as also thank P. Wilson and an anonymous reviewer for comments on an mountain ranges and arid valleys likely present obstacles to earlier version of this manuscript. This research was supported by the gene fl ow even for neighboring populations, newly developed Bureau of Land Management, Department of the Interior (Assistance methods allow genetic discontinuities between adjacent popu- Agreement PAA-01 – 7035), and an EPA STAR Fellowship to A.T.K. 5 Author for correspondence (e-mail: [email protected]) lations to be identifi ed (e.g., BARRIER software, Manni et al., 2004). Inference of gene fl ow patterns from population struc- doi:10.3732/ajb.1000229 ture can be further enhanced by combining molecular maker

American Journal of Botany 98(1): 109–121, 2011; http://www.amjbot.org/ © 2011 Botanical Society of America 109 110 American Journal of Botany [Vol. 98 data directly with information on geographic landscape features Here, we focus on three Penstemon species that are similarly ( Manel et al., 2003 ; Holderegger et al., 2006 ; Storfer et al., distributed throughout the Great Basin ’ s mountainous land- 2007 ) and new statistical methods in spatial genetics ( Guillot scape and ask how landscape features affect their population et al., 2009). Interpreting genetic discontinuities in the context genetic structure. We use microsatellite markers and multiple of landscape features provides a powerful approach for under- analyses to identify population genetic differentiation in these standing the interaction of plant dispersal systems, landscape, species and to better understand how this structure relates to and microevolutionary processes, yielding insights relevant to landscape features. Because the species chosen for study have evolutionary biology ( Kay and Sargent, 2009 ). different pollination syndromes but otherwise similar life his- Genetically isolated populations may follow different evolu- tory characteristics, we also identify how the population genetic tionary trajectories because of a combination of mutation, ge- structure of each species may be differently affected by the in- netic drift, and/or natural selection. Extreme examples come teraction of landscape features with characteristics related to from high-elevation mountaintops like the Andes ( Hughes and their dispersal ability. We expect that genetic divergence will Eastwood, 2006 ), rock outcrop “ inselbergs ” ( Barbara et al., be greater among mountain ranges than within mountain ranges 2007), and the edaphically diverse habitat of southern Africa’ s for all species but that the degree and structure of population Karoo region ( Ellis et al., 2006 ). With more than 100 separate genetic differentiation may vary for species with different pri- mountain ranges isolated by arid basins, the Great Basin region mary pollinators. Given previous fi ndings for greater long-dis- of the western United States provides the opportunity to deter- tance foraging in hummingbirds vs. bees, we expect that the mine how a complex topography infl uences gene fl ow in spe- hummingbird-pollinated species will have lower population ge- cies with different dispersal capabilities. Previous studies on netic divergence than the bee-pollinated species. plants distributed throughout the region have shown that coni- fer species that rely on wind for pollen movement and birds for seed dispersal display little population divergence (Johnson, MATERIALS AND METHODS 1975 ; Wells, 1983 ; Hamrick and Godt, 1996 ; Jorgensen et al., 2002 ), whereas populations of terrestrial animals are more ge- Study species— This study focuses on three species: Penstemon deustus netically isolated by the regions ’ landscape ( Floyd et al., 2005 ). Douglas ex Lindl. var. pedicellatus M. E. Jones, P. pachyphyllus A. Gray ex For animal-pollinated plants, pollinator movement is likely af- Rydb. var. congestus (M. E. Jones) N. H. Holmgren, and P. rostrifl orus Kel- fected by the region’ s steep elevational gradients, arid valleys, logg. All three species are considered common and widespread throughout sagebrush– steppe habitat in the Great Basin (Kartesz, 1999). This habitat is and patchy distribution of plant communities (West, 1988). found primarily on the region’ s sky islands at mid to high elevations, and within Pollinators may navigate this landscape in different ways, with it the distribution of each species is quite patchy: P. deustus and P. rostrifl orus varying impacts on the population genetic structure of the spe- often occur on rocky slopes, whereas P. pachyphyllus is primarily found on cies they pollinate. For example, birds might be more effective sandy or rocky plateaus. Of the three species, P. deustus has the largest distribu- than bees at connecting distant populations via successful long- tion throughout the Great Basin, whereas P. rostrifl orus is found primarily distance pollination ( Graves and Schrader, 2008 ). throughout the southern half of the region and P. pachyphyllus the southeastern half. Study species are found on many but not all mountain ranges within their One of the most species-rich plant genera in the Great Basin distribution, so where their ranges overlap, two or even all three study species is Penstemon (Plantaginaceae). It is also North America ’ s larg- can occur on the same mountain range. Although populations of each species est endemic genus, with over 270 species. The great diversity in occasionally occur near each other, they do not grow at the same site. Within this genus is hypothesized to be the result of a recent, rapid each mountain range, as few as one and as many as fi ve or more populations evolutionary radiation centered in the western United States, may exist at distances ranging from 1 km to more than 30 km apart. Population including the Great Basin region (Wolfe et al., 2006). Most sizes and densities are similar for all three species, ranging from 100 to 500 or more fl owering plants per population. Penstemon species are long lived, primarily outcrossing peren- These three species share most life history traits thought to affect population nial forbs with gravity-dispersed seeds, though wind might aid genetic structure. They are all long-lived, herbaceous, perennial forbs that pro- dispersal along the ground (Fuller and del Moral, 2003). Real- duce numerous, structurally similar protandrous fl owers borne on multiple ized gene fl ow is thus believed to be largely a function of pol- fl owering stalks. They have overlapping bloom times, but P. deustus and P. linator movement. Primary pollinators for Penstemon species pachyphyllus generally bloom earlier in the season than P. rostrifl orus . The include an array of insects (primarily bees) and hummingbirds. species are not known to reproduce clonally and have mixed-mating systems with at least some degree of self-compatibility ( Kramer, 2008 ). Seeds produced Pollinators generally can be predicted by the pollination syn- by Penstemon species are dispersed primarily by gravity but may tumble along drome (the suite of fl oral traits that have evolved in response to the ground with the wind until trapped (Fuller and del Moral, 2003). Seeds of particular pollinators) displayed by each species (Thomson et the three study species differ in size: P. deustus = 0.130 mg ± 0.028 mg; al., 2000 ; Wilson et al., 2004 ; Wilson et al., 2006 ). Penstemon P. pachyphyllus = 1.411 mg ± 0.472 mg; P. rostrifl orus = 0.343 mg ± 0.095 mg has been at the center of many studies on the validity and ( Kramer, 2008 ). application of pollination syndromes ( Thomson et al., 2000 ; The most notable difference between the three study species is in their fl ower morphologies and predominant visitors. Both P. deustus and P. pachy- Castellanos et al., 2003; Castellanos et al., 2004; Wilson et al., phyllus have fl oral traits associated with bee syndromes ( Thomson et al., 2000 ; 2004 ; Castellanos et al., 2006 ; Wilson et al., 2006 ; Wilson Fenster et al., 2004; Wilson et al., 2004) : P. deustus produces small, pale fl ow- et al., 2007 ; Thomson and Wilson, 2008 ). In particular, humming- ers, which are visited predominately by small bees (many Osmia spp.) but also bird pollination has arisen independently in the genus at least bumblebees (Bombus spp.) at certain high-elevation sites (A. Kramer, personal 10 times from bee pollination and possibly more than 20 times observations), and P. pachyphyllus produces larger purple fl owers, which are (Wilson et al., 2007). Interestingly, this shift seems to occur in generally visited by larger bees, including bumblebees at most sites (A. Kramer, personal observations). The third species, P. rostrifl orus , has a hummingbird only one direction; no occurrences have been reported of a shift syndrome, producing red, tubular fl owers, which are visited primarily by many from hummingbird to other pollination syndromes nor of rever- of the hummingbird species found in the western United States (Thomson et al., sions back to the original bee-pollination syndrome (Wilson et 2000 ; Wilson et al., 2004 ). al., 2007 ). Similar biases have been detected in numerous other genera: Mimulus (2 times; Beardsley et al., 2003), Erythrina (4 Study sites— Herbarium records were used to identify study sites on moun- times; Bruneau, 1997 ), and Costus (7 times; Kay et al., 2005 ). tain ranges throughout the distribution of each study species in the Great Basin January 2011] Kramer et al. —Q1 Landscape genetics of Great Basin Penstemon 111

fl oristic region (Cronquist et al., 1972). Many of the larger mountain ranges in admixture proportions (Q) were recorded for each study population by identi- the region contained at least one and often many herbarium records for each fi ed genetic cluster for the selected value of K. Finally, to provide a graphical species. Site selection for each species was targeted to mountain ranges with at representation of genetic distance data and relations within and among moun- least two herbarium collections from different locations, spanning their distri- tain ranges for each species, Nei’ s unbiased estimate of minimum genetic dis- bution throughout the Great Basin. In the summer of 2003, six populations of P. tance ( Nei, 1978 ) was used for unweighted pair-group clustering based on deustus and eight populations each of P. pachyphyllus and P. rostrifl orus were arithmetic averages (UPGMA), performed in TFPGA 1.3 ( Miller, 1997 ). identifi ed on four to six mountain ranges per species ( Fig. 1 , Table 1 ). Each For each species, genetic and spatial data were used to investigate isolation population contained more than 100 adult plants, and when possible, two popu- by distance. First, pairwise genetic distance among populations was estimated lations were located on each mountain range (separated by at least 2 km). At by calculating two measures in SPAGeDi ( Hardy and Vekemans, 2002 ): (1) each study site, 5 – 10 g of fresh leaf tissue was collected from at least 32 hap- Rousset ’ s linearized F ST (F ST /(1 – FST ) ( Weir and Cockerham, 1984 ; Rousset, hazardly located individuals, spanning the range of the population and avoiding 1997 ) and (2) unbiased R ST . Isolation by distance was then tested in each spe- sampling from adjacent plants. All leaf collections were dried in silica gel for cies by regressing the resulting pairwise population comparisons of genetic later DNA extraction, while GPS coordinates were recorded and voucher her- distance (F ST and R ST ) on spatial distance (ln km, or distance between popula- barium specimens collected for all species and sites. Vouchers were later de- tions), and the signifi cance of these relations was evaluated with Mantel (1967) posited at the Nancy Poole Rich Herbarium (Chicago Botanic Garden) and the tests (10 3 permutations) in GENALEX ( Peakall and Smouse, 2006 ). Great Basin Herbarium (Utah State University, see Appendix 1). In 2006, col- We used BARRIER 2.2 (Manni et al., 2004) to identify potential barriers to lections were made following the same protocols at two additional populations gene fl ow between geographically adjacent populations. With the use of latitu- for P. deustus, while in 2008 collections were made at two additional popula- dinal and longitudinal coordinates and voronoi tessellation, BARRIER imple- tions each for P. rostrifl orus and P. pachyphyllus to increase the number of ments Monmonier ’ s algorithm to generate a neighborhood polygon around all sampled populations (see Table 1 ). populations. Using pairwise matrices of Nei’ s genetic distance, generated in GENALEX for each locus ( Peakall and Smouse, 2006 ), BARRIER assigns Molecular data —Total genomic DNA was extracted from silica-dried leaf each polygon edge a value on the basis of the genetic distance between the two material following a CTAB method modifi ed from Doyle and Doyle (1987). neighboring populations sharing it, and it separately ranks them from maximum Genotypes were obtained for eight polymorphic nuclear (dinucleotide repeat) to least genetic difference for each locus. Starting at the polygon edge with the microsatellite loci developed from P. rostrifl orus (Pen02, Pen04, Pen05, Pen18, highest pairwise difference between two neighboring populations, BARRIER Pen23, Pen24, Pen25; detailed in Kramer and Fant, 2007, as well as an eighth draws a line along the edges of the polygons, taking the path of highest genetic locus, Pen06, developed following the same protocol; see Appendix 2 for de- distance until it transverses the sampling area, then it repeats this process with tails). DNA from all individuals was amplifi ed with polymerase chain reaction the next-highest pairwise distance until all have been used. We restricted (PCR) by using fl uorescently tagged forward primers (WellRed D2, D3, or D4; BARRIER to draw only the three greatest disparities in genetic distance for Sigma-Proligo, St. Louis, Missouri, USA) following methods described in each locus across the sampling area. By repeating this process for each locus, Kramer and Fant (2007) . Genotypes were scored with a CEQ 8000 Genetic we identifi ed which areas in the sampling range consistently showed high ge- Analysis System and CEQ FRAGMENT ANALYSIS software (Beckman netic distance across multiple loci, thereby identifying areas with a potential Coulter, Fullerton, California, USA). barrier to gene fl ow. If genetic distance is associated purely with geographic distances, then the two adjoining populations that are the farthest apart would Statistical analysis— Microsatellite genotype data were formatted for analy- be expected to show the most barriers across all loci. However, if adjoining sis with CREATE software ( Coombs et al., 2008 ). Descriptive parameters were population pairs have different origins, if there is strong genetic structure, or if calculated in GDA ( Lewis and Zaykin, 2001 ), including the following: P, pro- there is another physical or temporal barrier to gene-fl ow separating popula- portion of polymorphic loci; n, mean sample size; A, mean number of alleles tions, these population pairs would also reveal a high number of barriers. per locus; Ap, total number of private alleles; H E , expected heterozygosity; H O , observed heterozygosity; and Weir and Cockerham’ s (1984) estimates of RESULTS Wright ’ s FIS (f; within-population inbreeding coeffi cient). Departures from Hardy – Weinberg equilibrium (HWE) were tested for each species by using ex- act tests in GENEPOP (Raymond and Rousset, 1995) for each locus and popu- Descriptive statistics of loci— All eight microsatellite primer lation, as well as globally. pairs consistently amplifi ed products and were highly polymor- Population differentiation for each species was calculated by using Weir and phic in all study species, with two exceptions: Pen25 did not Cockerham ’ s (1984) estimate of F ST (theta) in TFPGA 1.3 ( Miller, 1997 ) and consistently amplify in P. deustus and P. pachyphyllus , so it Slatkin’ s unbiased estimator of R ST using RST_CALC (Goodman, 1997). For F , signifi cance of values was determined by analyzing jackknife support over was not used in the analysis for these two species, whereas ST Pen24 did not consistently amplify in P. rostrifl orus, so it was all loci in TFPGA 1.3, and for R ST , signifi cance of values was tested across all loci by permutation tests and bootstrapping to provide 95% confi dence intervals not used in the analysis for this species. Therefore, all analyses in RST_CALC. These two measures were chosen because F ST assumes each were conducted on seven loci for each species. In all three spe- mutation can produce an allele of any size, and hence differences between pop- cies, these seven remaining loci were highly polymorphic; the ulations are driven primarily by drift, whereas unbiased R ST is thought to better number of alleles ranged from 23 to 30 in P. deustus, seven to refl ect differentiation at microsatellite data because it assumes stepwise muta- tion such that each allele mutates to one of the immediately neighboring alleles 26 in P. pachyphyllus , and 18 to 34 in P. rostrifl orus. with equal probability, incorporating mutation and drift ( Olivier et al., 2003 ). Insight into the drivers of population differentiation can then be gained by com- Descriptive statistics of species and populations— When all paring values for F ST and R ST : values will be similar if genetic drift is the pri- loci were combined, each species displayed a different pattern mary cause of population differentiation, whereas R ST will be larger if of within-population diversity and heterozygosity ( Table 2 ). differentiation is also driven by stepwise mutation ( Hardy et al., 2003 ). Both gene diversity (H e ) and mean alleles per loci were lowest The Bayesian clustering analysis software STRUCTURE v. 2.2 ( Pritchard et al., in P. pachyphyllus, followed by P. deustus and P. rostrifl orus . 2000 ; Falush et al., 2007 ) was used to provide insight into patterns of gene fl ow This may be the result of ascertainment bias (Hutter et al., 1998) (admixture, Q) and population subdivision (number of genetic clusters, K) in each study species. This software uses individual multilocus genotypes to test because the microsatellite library was developed from P. rostri- for the presence of population structure without a priori assignment of individ- fl orus. Populations of all three species harbored between zero ual plants to populations. It does so by introducing population structure and and eight private alleles, with the exception of one P. deustus fi nding population groupings in the least possible disequilibrium (HWE and population, Pd-SM1, which contained 23 private alleles. Mea- linkage disequilibrium) by using a Markov chain Monte Carlo method. For sures of the population-level inbreeding coeffi cient (f) was sig- each species, we carried out 20 independent runs per K using a burn-in period nifi cant in four P. deustus populations, three P. rostrifl orus of 10 000 and collected data for 10 000 iterations for K = 1 to 16 (described in Evanno et al., 2005). The most likely value of K was assessed by using the rate populations, and two P. pachyphyllus populations. The average of change in the log probability of data between corresponding K values ( Δ K), inbreeding coeffi cient across all sampled populations was low- as detailed in Evanno et al. (2005). For each species, average and individual est in P. rostrifl orus (0.015) and highest in P. deustus (0.06). 112 American Journal of Botany [Vol. 98

Fig. 1. Study locations within the Great Basin region of the western United States, centered on the states of Nevada and Utah. Mountain ranges and arid valleys are outlined and shaded in gray. Between 8 and 10 study populations (white circles; see Table 1 for additional information) were identifi ed for each of three Penstemon species: (A) P. deustus , (B) P. pachyphyllus , and (C) P. rostrifl orus . Lines depict genetic discontinuities identifi ed in BARRIER ( Manni et al., 2004 ).

Analysis of population genetic structure— All three study had the greatest difference between measures, suggesting a species showed greater population differentiation with R ST than strong role for mutation in driving divergence, with an interme- FST measures, suggesting mutation and genetic drift together diate FST value (0.1330 ± 0.0270) and the highest RST value are driving population differentiation. However, because the among the three species (0.4076 ± 0.0.063). change in value between the two measures was different for Bayesian analysis performed in STRUCTURE was used to each species, their relative ranks changed depending on mea- infer spatial population structure and estimate the number of sure used. Of the three study species, P. rostrifl orus showed genetic clusters (K), or populations, into which the genotypic both the lowest population differentiation and the least change data could be grouped. STRUCTURE results confi rmed pro- in measures, with FST = 0.0639 ± 0.0125 and RST = 0.1116 ± nounced genetic structure in all three species, with generally 0.0219. For F ST measures, P. pachyphyllus showed the greatest more structure in P. deustus and P. pachyphyllus than in P. ros- population differentiation among the three study species (0.1896 trifl orus. The modal value of the distribution of the true K iden- ± 0.0437), but it was only intermediate between the three spe- tifi ed a peak at Δ K = 7 for P. deustus and at Δ K = 4 for P. cies when R ST was used (0.2531 ± 0.0250). Penstemon deustus pachyphyllus. In P. rostrifl orus, the modal peak was at Δ K = 3, January 2011] Kramer et al. —Q1 Landscape genetics of Great Basin Penstemon 113

Table 1. Detailed study site information for all three Penstemon study species collected throughout the Great Basin region of the western United States.

Population Mountain range State Latitude Longitude Elevation (m) Population size

Penstemon deustus Pd-DM1 Desatoya Mountains NV 39.25 − 117.68 2025 100 – 150 Pd-DM2a Desatoya Mountains NV 39.24 − 117.78 1909 400 – 500 Pd-SCR1 Schell Creek Range NV 39.56 − 114.64 2649 200 – 300 Pd-SCR2 a Schell Creek Range NV 39.57 − 114.59 2022 200 – 300 Pd-SM1 Steens Mountains OR 42.63 − 118.53 1793 200 – 300 Pd-SM2 Steens Mountains OR 42.05 − 118.62 1368 200 – 300 Pd-PNM1 Pine Nut Mountains NV 39.18 − 119.53 1861 150 – 200 Pd-PNM2 Pine Nut Mountains NV 39.12 − 119.42 1834 150 – 200 Penstemon pachyphyllus Pp-MP1 Markagunt Plateau UT 37.34 − 113.08 2122 300 – 400 Pp-MP2 Markagunt Plateau UT 37.17 − 113.08 1119 300 – 400 Pp-WWM1 Wah Wah Mountains UT 38.33 − 113.59 2560 150 – 200 Pp-WWM2 Wah Wah Mountains UT 38.34 − 113.61 2216 300 – 400 Pp-WWM3b Wah Wah Mountains UT 38.25 − 113.58 2430 150 – 200 Pp-SR1 Snake Range NV 39.11 − 114.35 2323 1000+ Pp-SR2 Snake Range NV 39.15 − 114.33 2227 300 – 400 Pp-SCR1 Schell Creek Range NV 39.19 − 114.63 2530 1000+ Pp-SCR2 Schell Creek Range NV 39.46 − 114.67 2103 500+ Pp-AR1 Antelope Range NV 40.04 − 114.51 1995 300 – 400 Penstemon rostrifl orus Pr-MP1 Markagunt Plateau UT 37.35 − 113.08 2092 200 – 300 Pr-MP2 Markagunt Plateau UT 37.29 − 113.10 1632 100 – 150 Pr-MP3 c Markagunt Plateau UT 37.89 − 112.73 2037 500+ Pr-WWM1 Wah Wah Mountains UT 38.35 − 113.61 2510 100 – 150 Pr-WWM2 Wah Wah Mountains UT 38.26 − 113.58 2455 150 – 200 Pr-SR1 Snake Range NV 39.02 − 114.27 2768 150 – 200 Pr-SR2 Snake Range NV 38.99 − 114.22 2147 100 – 150 Pr-SCR1 Schell Creek Range NV 39.19 − 114.63 2530 100 Pr-PM1 Pilot Mountains NV 38.39 − 118.03 1919 100 – 150 Pr-PNM1 Pine Nut Mountains NV 38.85 − 119.44 1678 200 – 300 Note: All populations were collected in 2003 unless otherwise indicated, and population size is shown as estimated number of fl owering and nonfl owering plants at the time of collection. For state abbreviations, NV = Nevada, UT = Utah, and OR = Oregon. a Collected in 2006. b Collected in 2007. c Collected in 2008. which is at the lower limits at which STRUCTURE can detect species, Markagunt Plateau populations exhibited the greatest clustering using Δ K, however the value of Δ K = 3 was sup- genetic differentiation from each other and all other popula- ported by large shifts in L(K) and Ln ’ (K) at K = 3 to K = 4 as- tions (Pp-MP1 and MP2). Genetic differentiation among P. de- sociated with true value of K, as described in Evanno et al. ustus populations was generally similar to P. pachyphyllus . (2005). For P. deustus and P. pachyphyllus populations, the av- Populations of P. deustus on the same mountain range grouped erage admixture proportion among identifi ed “ genetic clusters ” together, with the exception of Steens Mountain populations ranged from 80% to 98% assignment to a single cluster ( Fig. 2 ). (Ps-SM1 and SM2), which were located 65 km apart. The UP- Across the eight P. deustus populations, STRUCTURE iden- GMA tree for P. rostrifl orus was the shallowest, a result of rela- tifi ed strong genetic subdivision with limited admixture between tively low genetic distances between all population pairs. populations both within a mountain range and among mountain ranges ( Fig. 2 ). The exceptions were the two populations from Effect of landscape on population genetic structure— Man- Schell Creek Range (Pd-SCR1 and SCR2), which clustered into tel tests revealed signifi cant isolation by distance for all three a single grouping. In P. pachyphyllus, STRUCTURE identifi ed study species and measures of genetic differentiation used: strong genetic subdivision and limited admixture between Penstemon pachyphyllus ( FST P = .01; RST P = .05), P. deustus mountain ranges but no genetic structure between populations ( FST P = .02; RST P = .01) and P. rostrifl orus ( F ST P = .01; RST within the same mountain range. In P. rostrifl orus , K did not P = .03). The relation between different measures of genetic correspond to number of mountain ranges nor number of popu- distance and geographic distance varied by species and genetic lations sampled, and all populations had considerably more ad- measures used ( Fig. 4 ). When genetic distance was calculated as mixture among genetic clusters than the other two species. FST , P. pachyphyllus showed the greatest increase in genetic Results from UPGMA analysis (Fig. 3) provide additional distance with geographic distance (y = 0.095 × + 0.10, r 2 = 0.47), resolution regarding genetic distances between populations on followed by P. deustus ( y = 0.095 × + 0.02, r 2 = 0.30) and P. rostri- the same mountain range and on increasingly distant mountain fl orus (y = 0.032 × + 0.01, r 2 = 0.38). However, when genetic ranges. Penstemon pachyphyllus collectively had the largest distance was calculated as R ST (incorporating stepwise mutation overall genetic distances between populations, yet populations as well as drift), P. deustus showed a slightly greater increase in on the same mountain range always grouped together. In this genetic distance with geographic distance (y = 0.06 × + 0.12, 114 American Journal of Botany [Vol. 98

Table 2. Summary statistics for seven microsatellite loci shown by population for each Penstemon study species.

a b c Population Number of plants Mean sample size Mean alleles per locus Private allelesH e H O f Penstemon deustus Pd-DM1 32 31.9 14.7 8 0.807 0.736 0.101*** Pd-DM2 32 31.9 10.7 4 0.730 0.736 – 0.009 Pd-SCR1 32 32.0 7.9 1 0.701 0.674 0.039 Pd-SCR2 32 31.9 9.6 5 0.694 0.693 0.001 Pd-SM1 32 31.7 14.7 23 0.812 0.744 0.085*** Pd-SM2 32 31.9 11.0 8 0.771 0.740 0.041 Pd-PNM1 32 31.9 12.4 1 0.847 0.758 0.107*** Pd-PNM2 32 32.0 11.3 3 0.795 0.723 0.092*** Overall 256 31.9 11.5 7 0.771 0.726 0.060 Penstemon pachyphyllus Pp-SR1 32 32.0 7.6 0 0.530 0.518 0.023 Pp-SR2 32 31.7 8.6 5 0.519 0.474 0.087** Pp-SCR1 31 30.6 8.6 5 0.613 0.598 0.025 Pp-SCR2 32 30.3 9.3 7 0.650 0.675 – 0.040 Pp-AR1 32 31.7 6.3 0 0.608 0.563 0.076 Pp-WWM1 32 31.0 8.3 4 0.617 0.560 0.094*** Pp-WWM2 32 31.9 9.7 7 0.642 0.620 0.034 Pp-WWM3 32 30.7 8.2 4 0.640 0.635 0.007 Pp-MP2 32 31.7 9.4 6 0.772 0.761 0.015 Pp-MP1 32 31.9 7.3 7 0.612 0.647 – 0.058 Overall 319 31.7 8.2 5 0.620 0.605 0.025 Penstemon rostrifl orus Pr-SR1 32 31.9 11.3 2 0.797 0.816 – 0.025 Pr-SR2 32 31.3 13.0 1 0.825 0.804 0.026 Pr-SCR1 29 28.6 13.9 3 0.835 0.801 0.041*** Pr-WWM1 32 32.0 16.1 3 0.866 0.835 0.036* Pr-WWM2 32 32.0 14.6 0 0.870 0.879 – 0.012 Pr-MP1 32 31.9 16.4 6 0.906 0.870 0.041* Pr-MP2 32 31.9 12.3 0 0.799 0.776 0.030 Pr-MP3 32 31.7 17.6 7 0.913 0.887 0.036 Pr-PM1 32 30.7 12.6 4 0.779 0.786 0.009 Pr-PNM1 32 31.6 7.9 2 0.665 0.645 0.030 Overall 317 31.6 13.0 3 0.813 0.801 0.015 a Expected heterozygosity. b Observed heterozygosity. c Weir and Cockerham ’ s (1984) estimate of the within-population inbreeding coeffi cient. * P < .05. ** P < .01. *** P < .001.

r2 = 0.38) than P. pachyphyllus ( y = 0.02 × + 0.10, r 2 = 0.19), (Pr-MP1, MP2, and MP3), and between the Markagunt Plateau whereas P. rostrifl orus again had the smallest increase in ge- populations and all other populations ( Fig. 1 ). The genetic dis- netic distance with increasing distance (y = 0.01 × + 0.01, tance between two populations on the Markagunt Plateau (Pr- r2 = 0.25). MP1 and MP3) was comparable to the distance between BARRIER was used to identify the largest difference in ge- population pairs on other mountain ranges, whereas the largest netic distances between two adjoining populations, indicative difference was identifi ed between MP2 and the other two popu- of potential barriers to gene fl ow ( Fig. 1 ). In P. deustus, the lations on the Markagunt Plateau (Pr-MP1 and MP3). adjoining regions that produced the highest differences in ge- netic distance occurred between northern populations and all others, and western populations and all others, suggesting lim- DISCUSSION ited gene fl ow north – south and east – west. Interestingly, there was also a signifi cant disparity in genetic distances between the Gene fl ow maintains species cohesion. In its absence, popu- two populations in the Pine Nut Mountains (Pd-PNM1 and lations isolated from one another will follow different evolu- PNM2) and between populations in the Steens Mountains (Pd- tionary trajectories ( Ellstrand, 1992 ; Morjan and Rieseberg, SM1 and SM2). For P. pachyphyllus, the most signifi cant 2004 ). For plants, gene fl ow via the dispersal of seeds and pol- boundaries were identifi ed between mountain ranges. The only len is therefore a critical determinant of current and future pop- other signifi cant boundary identifi ed for this species was be- ulation genetic divergence ( Kramer et al., 2008 ). Both physical tween the two populations from the Markagunt Plateau (Pp- distance and landscape effects can interact with dispersal abil- MP1 and MP2), located on the same mountain range but at very ity to enhance or deter gene fl ow. Our results show that the different elevations ( Table 1 and Fig. 1 ). For P. rostrifl orus , the Great Basin ’ s landscape signifi cantly isolates the plant popula- only boundaries resulting from high genetic distance in three or tions that occupy its sky islands. Further, results of our exami- more loci were between populations on the Markagunt Plateau nation of three otherwise-similar Penstemon species with January 2011] Kramer et al. —Q1 Landscape genetics of Great Basin Penstemon 115

Fig. 2. Identifi ed genetic clusters (I – VII, as shown at the bottom of the fi gure) and Bayesian admixture proportions depicted for individual plants and populations of all three Penstemon species (P. deustus K = 7, P. pachyphyllus K = 4, and P. rostrifl orus K = 3). Squares outline individual samples for each population. Population names correspond to study site information ( Table 1 ), and populations on the same mountain range are similarly named, e.g., P. deustus populations DM1 and DM2 both occur in the Desatoya Mountains. different pollination syndromes suggest that pollinators do not ers to gene fl ow via BARRIER, likewise revealed signifi cant interact with landscape features in the same ways, creating dif- genetic structure largely partitioned among sampled mountain ferent patterns and degrees of population genetic structure in ranges. It is likely that bees either avoid crossing the Great Ba- the species they pollinate. sin ’ s arid valley fl oors or, if they do cross it, are ineffective at We identifi ed signifi cant population genetic differentiation in transferring pollen across these expanses. Many bees groom all three Penstemon species using R ST, with greatest divergence pollen from their bodies at regular intervals, so even if they fl y in P. deustus (R ST = 0.4076) and lowest divergence in P. rostri- long distances, they may not effect long-distance pollinations fl orus (R ST = 0.1116). This divergence is likely driven by ge- ( Wilson et al., 2004 ). netic drift and mutation, as R ST values were higher than F ST With far less genetic structure than its bee-syndrome coun- values for all three species (e.g., FST = 0.1896 and R ST = 0.2531 terparts, results from P. rostrifl orus suggest that hummingbirds for P. pachyphyllus ). In general, population genetic divergence are better than bees at maintaining cohesion between popula- in these three species was greater than that reported for other tions separated over much greater distances. This has often co-occurring plant species in the Great Basin region. For ex- been suggested but rarely substantiated ( Graves and Schrader, ample, populations of eight wind-pollinated, bird-dispersed co- 2008). Bird-mediated long-distance (> 5 km) gene fl ow between nifer species had an average G ST ranging from 0.033 for pinyon populations in fragmented agricultural habitat was recently pine to 0.169 for bristlecone pine (Hamrick et al., 1994; documented in the sunbird-pollinated Calothamnus quadrifi dus Jorgensen et al., 2002). A comparison of genetic differentiation (Byrne et al., 2007), but the specifi c distances and terrains that among our study species based on pollination syndrome revealed avian pollinators are capable of bridging are largely unknown. greater structure in the bee-pollinated species, P. pachyphyllus Additionally, the territoriality of some bird species (humming- and P. deustus , compared with the hummingbird-pollinated birds) may actually limit long-distance pollen movement (Parra P. rostrifl orus (Figs. 2– 4). For both bee-pollinated species, et al., 1993 ). Nine hummingbird species are recorded in the Bayesian clustering analysis independently identifi ed each Great Basin region, including Broad-billed ( Cynanthus latiro- sampled mountain range as a separate genetic cluster, with little stris ), Magnifi cent (Eugenes fulgens), Black-chinned ( Ar- or no admixture between mountain ranges (Fig. 2). Other analy- chilochus alexandri ), Anna ’ s ( Calypte anna ), Costa ’ s (C. ses, including UPGMA analyses and the identifi cation of barri- costae), Calliope ( Stellula calliope ), Broad-tailed (Selasphorus 116 American Journal of Botany [Vol. 98

Fig. 3. UPGMA clustering using Nei ’ s unbiased estimate of minimum genetic distance (1978) for each species showed generally similar levels of genetic distance between populations in the bee-pollinated Penstemon deustus and P. pachyphyllus , with much lower genetic distances between populations in the hummingbird-pollinated P. rostrifl orus. platycercus ), Rufous (S. rufus ), and Allen ’ s (S. sasin ) hum- Our results highlight an important but often overlooked inter- mingbirds (Johnsgard, 1983). Our study reveals that at least action between pollinators and landscape features like topography some of these species appear capable of moving pollen over and distance on the evolutionary trajectory of a species. The three large distances across the Great Basin. Results of Bayesian Penstemon species studied share most life history traits, and it is cluster analyses (Fig. 2) indicate hummingbird-assisted admix- unlikely their gravity-dispersed seeds contribute much to gene ture between populations separated by at least 19 km within fl ow among populations on different mountain ranges. Therefore, mountain ranges (e.g., Pr-WWM1 and WWM2) and over 100 differences in primary pollinators are a likely explanation for the km between mountain ranges, particularly when populations differences in population structure among species. Our results were at high elevations (e.g., Pr-MP1 and WWM1). A study of suggest that pollination syndromes do not just summarize the fl o- two Streptocarpus species in South Africa also reported signifi - ral architecture and functional pollinator group of a species cantly lower genetic differentiation in a primarily sunbird-pol- (Thomson et al., 2000) but also have important impacts on popu- linated species than in a fl y-pollinated species (Hughes et al., lation structure and genetic isolation of populations. As Kay and 2007 ), providing another case in which bird pollination main- Sargent (2009) recognized, “ fl oral isolation alone is rarely a tains greater population cohesion than insect pollination. complete barrier (Chari and Wilson, 2001; Ramsey et al., 2003; January 2011] Kramer et al. —Q1 Landscape genetics of Great Basin Penstemon 117

Fig. 4. Relations between different measures of pairwise genetic distance (F ST and R ST, , both linearized) and geographic distance (ln km) for each species. Although signifi cant isolation by distance was detected with Mantel tests for each study species and genetic distance measure, the degree of isola- tion by distance varied. In (A), when only genetic drift was incorporated (F ST ), Penstemon pachyphyllus had a steeper slope than P. deustus However, in (B), that pattern was reversed with R ST (incorporating drift and stepwise mutation), and P. deustus had a greater slope than P. pachyphyllus. Regardless of mea- sure used, P. rostrifl orus had a smaller slope than either of the two species.

Kay, 2006) but can act in concert with isolating factors to reduce for unidirectional shifts from bee- to bird-pollination syn- the homogenizing effects of gene fl ow, allowing divergent line- dromes, particularly in areas with high topographic and edaphic ages to persist and further diverge. ” Although our results agree variation (Kay and Sargent, 2009). Our results suggest that in with predictions from previous reports of pollinator dispersal dis- bird-pollinated species, greater functional connectivity exists tances, additional studies that compare species with different dis- among distant populations, and this greater gene fl ow likely persal agents in differing landscapes are needed to clarify the role constrains additional local adaptive divergence (Hendry et al., of pollinators in shaping the genetic structure of plant populations. 2002 ; Hendry and Taylor, 2004 ; Moore et al., 2007 ; R ä s ä nen Our fi ndings are of particular relevance to recent interest in and Hendry, 2008). This constraining effect may explain why evolutionary shifts between bee- and bird-pollination syn- reversion back to a bee syndrome or toward another syndrome dromes (Cronk and Ojeda, 2008; Thomson and Wilson, 2008; has not been documented once a bird-pollination syndrome has Kay and Sargent, 2009), as they provide a possible explanation arisen in Penstemon species ( Wolfe et al., 2006 ). 118 American Journal of Botany [Vol. 98

An alternative hypothesis for the differing levels of popula- peak fl owering time of 1– 2 mo (Kramer, unpublished data). tion structure in these three species might involve the timing of These results suggest that, regardless of pollination syndrome, speciation and the origin of each species. For example, lower phenological differences caused by topography are also im- levels of genetic structure in P. rostrifl orus might refl ect a more portant in driving genetic divergence between populations. recent origin than those of P. deustus and P. pachyphyllus . Phy- Our results can be used to guide ecological restoration efforts logenetic work on the genus ( Wolfe et al., 2006 ) might allow for our study species and those that share similar characteristics the relative timing of origins to be inferred, as all three study found throughout the unique landscape of the Great Basin region. species were included in an analysis using nucleotide sequence Findings of high genetic diversity and signifi cant population ge- data from ITS (nuclear DNA) as well as trnC – D and trnT – L netic structure in all three species support ongoing efforts to bank (chloroplast DNA) regions. Results show the phylogenetic seeds of multiple populations to store genetic diversity for future placement and branch length of each study species varies de- restoration and research efforts ( DeBolt and Spurrier, 2004 ). Our pending on the region analyzed. With ITS, the bird-pollinated fi ndings of signifi cant genetic structure also caution against P. rostrifl orus appears to be more recently derived than the bee- broad-scale movement and mixing of populations for restoration pollinated P. deustus but not as recently derived as P. pachy- purposes. Given the greater among-population gene fl ow identi- phyllus, and with trnC – D / T – L, the bee-pollinated species fi ed in the hummingbird-pollinated P. rostrifl orus , the large-scale actually appear to be more recently derived than the bird-polli- movement of seeds to restore populations of hummingbird-polli- nated P. rostrifl orus. Thus, phylogenetic work to date does not nated species may pose less risk to the success of a restoration support a recent origin for P. rostrifl orus as an explanation for than in bee-pollinated species. However, we recommend addi- its low levels of population differentiation. tional research to investigate adaptive divergence in quantitative Whereas the bees pollinating our study species appear to be traits, which can occur even in the presence of gene fl ow ( Endler, less effective than hummingbirds at mediating long-distance gene 1973 ) and may infl uence the success or failure of a restoration. fl ow between Great Basin mountain ranges, our cluster analysis also indicated that bees pollinating P. pachyphyllus are more ef- LITERATURE CITED fective at moving pollen between populations on the same mountain range than are bees pollinating P. deustus ( Fig. 2 ). Barbara , T. , G. Martinelli , M. F. Fay , S. J. Mayo , and C. Lexer . Numerous bee visitors to P. deustus have been reported, including 2007 . Population differentiation and species cohesion in two closely Osmia species, Anthophora and similar small nectar-collecting related plants adapted to neotropical high-altitude “ inselbergs, ” bees, as well as larger nectar-collecting Bombus species (Wilson Alcantarea imperialis and Alcantarea geniculata (Bromeliaceae). P. pachyphyllus Molecular Ecology 16 : 1981 – 1992 . et al., 2004 ). Visitors to fl owers have not been Beardsley , P. M. , A. Yen , and R. G. Olmstead . 2003 . AFLP phylog- well documented, but fl owers are similar to those of the well- eny of Mimulus section Erythranthe and the evolution of hummingbird studied P. strictus, so we would expect a similar range of visitors. pollination. Evolution; International Journal of Organic Evolution 5 7 : This includes the same functional groups as in P. deustus but ex- 1397 – 1410 . tends to a broader range of generally larger bee and wasp visi- Bruneau , A. 1997 . Evolution and homology of bird pollination syn- tors (Wilson et al., 2004). The different pollinators observed for dromes in Erythrina (Leguminosae). American Journal of Botany each species may affect long-distance pollen fl ow over differ- 84 : 54 – 71 . ent scales for various reasons, including (1) larger bee pollina- Byrne , M. , C. P. Elliott , C. Yates , and D. J. Coates . 2007 . Extensive tors of P. pachyphyllus may move pollen more effectively pollen dispersal in a bird-pollinated shrub, Calothamnus quadrifi dus , in between distant populations on the same mountain range than a fragmented landscape. Molecular Ecology 16 : 1303 – 1314 . Castellanos , M. C. , P. Wilson , S. J. Keller , A. D. Wolfe , and J. D. smaller bee pollinators of P. deustus, or (2) larger bees have Thomson . 2006 . Anther evolution: Pollen presentation strategies greater energy requirements than smaller bees ( Westphal et al., when pollinators differ. American Naturalist 167 : 288 – 296 . 2006), requiring that they use a broader foraging area and lead- Castellanos , M. C. , P. Wilson , and J. D. Thomson . 2003 . Pollen trans- ing to greater long-distance gene fl ow. However, the higher fer by hummingbirds and bumblebees, and the divergence of pollina- level of admixture for P. pachyphyllus compared with P. de- tion modes in Penstemon. Evolution; International Journal of Organic ustus for populations on the same mountain range ( Fig. 2 ) may Evolution 57 : 2742 – 2752 . be in part a refl ection of our study design. The two within- Castellanos , M. C. , P. Wilson , and J. D. Thomson . 2004 . ‘ Anti-bee ’ mountain populations of P. pachyphyllus were separated by and ‘ pro-bird ’ changes during the evolution of hummingbird pollination less than 5 km, whereas all comparisons for P. deustus were in Penstemon fl owers. Journal of Evolutionary Biology 17 : 876 – 885 . Chari , J. , and P. Wilson . 2001 . Factors limiting hybridization between separated by more than 5 km. Resolving this question would Penstemon spectabilis and Penstemon centranthifolius. Canadian require incorporating additional populations on each mountain Journal of Botany 79 : 1439 – 1448 . range at a variety of geographic distances. Coombs , J. A. , B. H. Letcher , and K. H. Nislow . 2008 . Create: A soft- One exception exists to our fi nding of higher connectivity ware to create input fi les from diploid genotypic data for 52 genetic between populations within mountain ranges than among them, software programs. Molecular Ecology Resources 8 : 578 – 580 . and it occurred in the same location for two of our study spe- Cronk , Q. , and I. Ojeda. 2008 . Bird-pollinated fl owers in an evo- cies. BARRIER identifi ed strong landscape-level barriers to gene lutionary and molecular context. Journal of Experimental Botany fl ow between population pairs on the Markagunt Plateau for both 59 : 715 – 727 . P. pachyphyllus (bee pollinated) and P. rostrifl orus (humming- Cronquist , A. , A. H. Holmgren , N. H. Holmgren , and J. L. Reveal . bird pollinated). Despite the small geographic distance separat- 1972 . Intermountain fl ora. Vascular plants of the Intermountain West, USA. Hafner, New York, USA. ing these populations (each less than 1 km apart), there was DeBolt , A. , and C. S. Spurrier. 2004 . Seeds of success and the relatively high genetic distance. This may be attributable to large Millennium Seed Bank Project. In A. L. Hild, N. L. Shaw, S. E. Meyer, elevational differences between sites: nearly 500 m in elevation D. T. Booth, and E. D. McArthur [eds.], RMRS-P-31: Seed and soil for P. rostrifl orus and more than 1000 m in elevation for P. pac- dynamics in shrubland ecosystems: Proceedings, Laramie, Wyoming, hyphyllus . For P. rostrifl orus , this elevation difference, as well as USA, 2002, 100 – 108. US Department of Agriculture, Forest Service, differences in aspect, resulted in population-level differences in Rocky Mountain Research Station, Fort Collins, Colorado, USA. January 2011] Kramer et al. —Q1 Landscape genetics of Great Basin Penstemon 119

Dick , C. W. , G. Etchelecu , and F. Austerlitz . 2003 . Pollen disper- Holderegger , R. , and H. H. Wagner . 2008 . Landscape genetics. sal of tropical ( Dinizia excelsa: ) by native insects and Bioscience 58 : 199 – 207 . African honeybees in pristine and fragmented Amazonian . Hughes , C. , and R. Eastwood . 2006 . Island radiation on a continen- Molecular Ecology 12 : 753 – 764 . tal scale: Exceptional rates of plant diversifi cation after uplift of the Doyle , J. J. , and J. L. Doyle . 1987 . A rapid DNA isolation procedure Andes. Proceedings of the National Academy of Sciences, USA 103 : for small quantities of fresh leaf tissue. Phytochemical Bulletin 19 : 10334 – 10339 . 11 – 15 . Hughes , M. , M. Moller , T. J. Edwards , D. U. Bellstedt , and M. de Duminil , J. , S. Fineschi , A. Hampe , P. Jordano , D. Salvini , G. G. Villiers . 2007 . The impact of pollination syndrome and habitat on Vendramin , and R. J. Petit . 2007 . Can population genetic struc- gene fl ow: A comparative study of two Streptocarpus (Gesneriaceae) ture be predicted from life-history traits? American Naturalist 169 : species. American Journal of Botany 94 : 1688 – 1695 . 662 – 672 . Hutter , C. M. , M. D. Schug , and C. F. Aquadr . 1998 . Microsatellite Ellis , A. G. , A. E. Weis , and B. S. Gaut . 2006 . Evolutionary radiation variation in Drosophila melanogaster and Drosophila simulans: A re- of “ stone plants ” in the genus Argyroderma (Aizoaceae): Unraveling ciprocal test of the ascertainment bias hypothesis. Molecular Biology the effects of landscape, habitat, and fl owering time. Evolution; and Evolution 15 : 1620 – 1636 . International Journal of Organic Evolution 60 : 39 – 55 . Johnsgard , P. A. 1983 . The hummingbirds of North America. Smithsonian Ellstrand , N. C. 1992 . Gene fl ow by pollen: Implications for plant con- Institution Press, Washington, DC, USA. servation genetics. Oikos 63 : 77 – 86 . Johnson , N. K. 1975 . Controls of number of bird species on montane is- Endler , J. A. 1973 . Gene fl ow and population differentiation. Science lands in the Great Basin. Evolution; International Journal of Organic 179 : 243 – 250 . Evolution 29 : 545 – 567 . Evanno , G. , S. Regnaut , and J. Goudet . 2005 . Detecting the number Jorgensen , S. , J. L. Hamrick , and P. V. Wells . 2002 . Regional pat- of clusters of individuals using the software structure: A simulation terns of genetic diversity in Pinus fl exilis (Pinaceae) reveal complex study. Molecular Ecology 14 : 2611 – 2620 . species history. American Journal of Botany 89 : 792 – 800 . Falush , D. , M. Stephens , and J. K. Pritchard . 2007 . Inference of Kartesz , J. T. 1999 . A synonymized checklist of the vascular fl ora of population structure using multilocus genotype data: Dominant mark- the U.S., Canada, and Greenland. In J. T. Kartesz and C. A. Meacham ers and null alleles. Molecular Ecology Notes 7 : 574 – 578 . [eds.], Synthesis of the North American fl ora, version 1.0. North Fenster , C. B. , W. S. Armbruster , P. Wilson , M. R. Dudash , and J. D. Carolina Botanical Garden, Chapel Hill, North Carolina, USA. Thomson . 2004 . Pollination syndromes and fl oral specializations. Kay , K. M. 2006 . Reproductive isolation between two closely related Annual Review of Ecology Evolution and Systematics 35 : 375 – 403 . hummingbird-pollinated neotropical gingers. Evolution; International Floyd , C. H. , D. H. Van Vuren , and B. May . 2005 . Marmots on Great Journal of Organic Evolution 60 : 538 – 552 . Basin mountaintops: Using genetics to test a biogeographic paradigm. Kay , K. M. , P. A. Reeves , R. G. Olmstead , and D. W. Schemske . 2005 . Ecology 86 : 2145 – 2153 . Rapid speciation and the evolution of hummingbird pollination in Fuller , R. N. , and R. del Moral . 2003 . The role of refugia and disper- neotropical Costus subgenus Costus (Costaceae): Evidence from sal in primary succession on Mount St. Helens, Washington. Journal nrDNA ITS and ETS sequences 1. American Journal of Botany 92 : of Vegetation Science 14 : 637 – 644 . 1899 – 1910 . Goodman , S. J. 1997 . RST CALC: A collection of computer programs Kay , K. M. , and R. D. Sargent . 2009 . The role of animal pollination in for calculating unbiased estimates of genetic differentiation and deter- plant speciation: Integrating ecology, geography, and genetics. Annual mining their signifi cance for microsatellite data. Molecular Ecology Review of Ecology Evolution and Systematics 40 : 637 – 656 . 6 : 881 – 885 . Kramer , A. T. 2008 . Ecological genetics of Penstemon in the Great Graves , W. R. , and J. A. Schrader . 2008 . At the interface of phyloge- Basin, USA. Ph.D. thesis, University of Illinois at Chicago, Chicago, netics and population genetics, the phylogeography of Dirca occiden- Illinois, USA. talis (Thymelaeaceae). American Journal of Botany 95 : 1454 – 1465 . Kramer , A. T. , and J. B. Fant . 2007 . Isolation and characterization of Guillot , G. , R. Leblois , A. Coulon , and A. C. Frantz . 2009 . Statistical microsatellite loci in Penstemon rostrifl orus (Plantaginaceae) and methods in spatial genetics. Molecular Ecology 18 : 4734 – 4756 . cross-species amplifi cation. Molecular Ecology Notes 7 : 998 – 1001 . Hamrick , J. L. , and M. J. W. Godt. 1996 . Effects of life history traits Kramer , A. T. , J. L. Ison , M. V. Ashley , and H. F. Howe . 2008 . The on genetic diversity in plant species. Philosophical Transactions paradox of forest fragmentation genetics. Conservation Biology 22 : of the Royal Society of London, B, Biological Sciences 351 : 878 – 885 . 1291 – 1298 . Lewis , P. O. , and D. Zaykin . 2001 . Genetic Data Analysis: Computer Hamrick , J. L. , A. F. Schnabel , and P. V. Wells . 1994 . Distribution of program for the analysis of allelic data, version 1.0 (d16c). Free pro- genetic diversity within and among populations of Great Basin co- gram distributed by the authors, website: http://hydrodictyon.eeb. nifers. In K. T. Harper, L. L. St. Clair, K. H. Thorne, and W. W. uconn.edu/people/plewis/software.php . Hess [eds.], Natural history of the Colorado Plateau and Great Basin, Manel , S. , M. K. Schwartz , G. Luikart , and P. Taberlet . 2003 . 147 – 162. University Press of Colorado, Boulder, Colorado, USA. Landscape genetics: Combining landscape ecology and population Hardy , O. J. , N. Charbonnel , H. Freville , and M. Heuertz . 2003 . genetics. Trends in Ecology & Evolution 18 : 189 – 197 . Microsatellite allele sizes: A simple test to assess their signifi cance on Manni , F. , E. Guerard , and E. Heyer . 2004 . Geographic patterns of genetic differentiation. Genetics 163 : 1467 – 1482 . (genetic, morphological, linguistic) variation: How barriers can be de- Hardy , O. J. , and X. Vekemans . 2002 . SPAGeDi: A versatile computer tected by “ Monmonier ’ s algorithm. ” Human Biology 76 : 173 – 190 . program to analyse spatial genetic structure at the individual or popu- Mantel , N. 1967 . The detection of disease clustering and a generalized lation levels. Molecular Ecology Notes 2 : 618 – 620 . regression approach. Cancer Research 27 : 209 – 220 . Hendry , A. P. , and E. B. Taylor . 2004 . How much of the variation in Miller , M. P. 1997 . Tools for population genetic analyses (TFPGA) 1.3: adaptive divergence can be explained by gene fl ow? An evaluation A Windows program for the analysis of allozyme and molecular pop- using lake– stream stickleback pairs. Evolution; International Journal ulation genetic data. Computer software distributed by author, website of Organic Evolution 58 : 2319 – 2331 . http://www.marksgeneticsoftware.net/tfpga.htm . Hendry , A. P. , E. B. Taylor , and J. D. McPhail . 2002 . Adaptive Mimura , M. , R. C. Barbour , B. M. Potts , R. E. Vaillancourt , and K. divergence and the balance between selection and gene fl ow: Lake N. Watanabe . 2009 . Comparison of contemporary mating patterns and stream stickleback in the misty system. Evolution; International in continuous and fragmented Eucalyptus globulus native forests. Journal of Organic Evolution 56 : 1199 – 1216 . Molecular Ecology 18 : 4180 – 4192 . Holderegger , R. , U. Kamm , and F. Gugerli . 2006 . Adaptive vs. neu- Moore , J.-S. , J. L. Gow , E. B. Taylor , A. P. Hendry , and M. Peterson . tral genetic diversity: Implications for landscape genetics. Landscape 2007 . Quantifying the constraining infl uence of gene fl ow on Ecology 21 : 797 – 807 . adaptive divergence in the lake– stream threespine stickleback sys- 120 American Journal of Botany [Vol. 98

tem. Evolution; International Journal of Organic Evolution 61 : Storfer , A. , M. A. Murphy , S. F. Spear , R. Holderegger , and L. P. 2015 – 2026 . Waits . 2010 . Landscape genetics: Where are we now? Molecular Morjan , C. L. , and L. H. Rieseberg . 2004 . How species evolve col- Ecology 19 : 3496 – 3514 . lectively: Implications of gene fl ow and selection for the spread of Thomson , J. D. , and P. Wilson . 2008 . Explaining evolutionary shifts advantageous alleles. Molecular Ecology 13 : 1341 – 1356 . between bee and hummingbird pollination: Convergence, divergence, Nei , M. 1978 . Estimation of average heterozygosity and genetic distance and directionality. International Journal of Plant Sciences 169 : from a small number of individuals. Genetics 89 : 583 – 590 . 23 – 38 . Olivier , J. H. , N. Charbonnel , H. Fr é ville , and M. Heuertz . 2003 . Thomson , J. D. , P. Wilson , M. Valenzuela , and M. Malzone . 2000 . Microsatellite allele sizes: A simple test to assess their signifi cance on Pollen presentation and pollination syndromes, with special reference genetic differentiation . Genetics 163 : 1467 – 1482 . to Penstemon. Plant Species Biology 15 : 11 – 29 . Parra , V. , C. F. Vargas , and L. E. Eguiarte . 1993 . Reproductive- Townsend , P. A. , and D. J. Levey . 2005 . An experimental test of whether biology, pollen and seed dispersal, and neighborhood size in the hum- habitat corridors affect pollen transfer. Ecology 86 : 466 – 475 . mingbird-pollinated gibbifl ora (). American Weir , B. S. , and C. C. Cockerham . 1984 . Estimating F-statistics for the Journal of Botany 80 : 153 – 159 . analysis of population structure. Evolution; International Journal of Peakall , R. , and P. E. Smouse . 2006 . GENALEX 6: Genetic analy- Organic Evolution 38 : 1358 – 1370 . sis in Excel. Population genetic software for teaching and research. Wells , P. V. 1983 . Paleobiogeography of Montane Islands in the Great Molecular Ecology Notes 6 : 288 – 295 . Basin since the Last Glaciopluvial. Ecological Monographs 5 3 : Pritchard , J. K. , M. Stephens , and P. Donnely. 2000 . Inference of popula- 341 – 382 . tion structure using multilocus genotype data. Genetics 155 : 945 – 959 . West , N. E. 1988 . Intermountain deserts, shrub steppes and woodlands. Ramsey , J. , H. D. Bradshaw , and D. W. Schemske . 2003 . Components In M. G. Barbour and W. D. Billings [eds.], North American terrestrial of reproductive isolation between the monkeyfl owers Mimulus lewisii vegetation, 209 – 230. Cambridge University Press, Cambridge, UK. and M. cardinalis (Phrymaceae). Evolution; International Journal of Westphal , C. , I. Steffan-Dewenter , and T. Tscharntke . 2006 . Organic Evolution 57 : 1520 – 1534 . Bumblebees experience landscapes at different spatial scales: Possible R ä s ä nen , K. , and A. P. Hendry . 2008 . Disentangling interactions be- implications for coexistence. Oecologia 149 : 289 – 300 . tween adaptive divergence and gene fl ow when ecology drives diver- Wilson , P. , M. C. Castellanos , J. N. Hogue , J. D. Thomson , and W. S. sifi cation. Ecology Letters 11 : 624 – 636 . Armbruster . 2004 . A multivariate search for pollination syndromes Raymond , M. , and F. Rousset . 1995 . GENEPOP (version 1.2): Popu- among penstemons. Oikos 104 : 345 – 361 . lation genetics software for exact tests and ecumenicism. Journal of Wilson , P. , M. C. Castellanos , A. D. Wolfe , and J. D. Thomson . Heredity 86 : 248 – 249 . 2006 . Shifts between bee and bird pollination in Penstemons. In N. Richards , A. J. 1997 . Plant breeding systems. Chapman & Hall, London, M. Waser and J. Ollerton [eds.], Plant– pollinator interactions, from UK. specialization to generalization, 47 – 68. University of Chicago Press, Rousset , F. 1997 . Genetic differentiation and estimation of gene fl ow from Chicago, Illinois, USA. F-statistics under isolation by distance. Genetics 145 : 1219 – 1228 . Wilson , P. , A. D. Wolfe , W. S. Armbruster , and J. D. Thomson . Slatkin , M. 1985 . Gene fl ow in natural populations. Annual Review of 2007 . Constrained lability in fl oral evolution: Counting convergent Ecology and Systematics 16 : 393 – 430 . origins of hummingbird pollination in Penstemon and Keckiella. New Sneath , P. H. A. , and R. R. Sokal . 1973 . Numerical . W. H. Phytologist 176 : 883 – 890 . Freeman, San Francisco, California, USA. Wolfe , A. D. , C. P. Randle , S. L. Datwyler , J. J. Morawetz , N. Storfer , A. , M. A. Murphy , J. S. Evans , C. S. Goldberg , S. Robinson , Arguedas , and J. Diaz . 2006 . Phylogeny, taxonomic affi nities, and S. F. Spear , R. Dezzani , et al. 2007 . Putting the “ landscape ” in biogeography of Penstemon (Plantaginaceae) based on ITS and cp- landscape genetics. Heredity 98 : 128 – 142 . DNA sequence data. American Journal of Botany 93 : 1699 – 1713 .

Appendix 1.

Taxon — Population name, Voucher specimen , Collection locale; Herbarium. Fant et al. s.n., UT; CHIC. Pp-SR1, Fant et al. s.n., NV; CHIC, UTC. Pp-SR2, Fant et al. s.n., NV; CHIC, UTC. Pp-SCR1, Fant et al. s.n. , NV; Penstemon deustus Douglas ex Lindl. var. pedicellatus M.E. Jones — Pd-DM1, Tietmeyer et al. s.n., NV; CHIC, UTC. Pd-DM2, Fant et al. s.n. , NV; CHIC. Pp-SCR2, Fant et al. s.n., NV; CHIC. Pp-AR1, Fant et al. s.n., NV; CHIC. Pd-SCR1, Tietmeyer et al. s.n. , NV; CHIC, UTC. Pd-SCR2, Fant CHIC, UTC. et al. s.n., NV; CHIC. Pd-SM1, Fant et al. s.n., OR; CHIC, UTC. Pd-SM2, Penstemon rostrifl orus Kellogg — Pr-MP1, Tietmeyer et al. s.n., UT; CHIC, Fant et al. s.n., OR; CHIC, UTC. Pd-PDNM1, Tietmeyer et al. s.n. , NV; UTC. Pr-MP2, Fant et al. s.n., UT; CHIC, UTC. Pr-MP3, Fant et al. s.n., CHIC, UTC. Pd-PNM2, Tietmeyer et al. s.n., NV; CHIC, UTC. UT; CHIC. Pr-WWM1, Tietmeyer et al. s.n., UT; CHIC. Pr-WWM2, Penstemon pachyphyllus A. Gray ex Rydb. var. congestus (M.E. Jones) Tietmeyer et al. s.n., UT; CHIC, UTC. Pr-SR1, Tietmeyer et al. s.n., NV; N.H. Holmgren — Pp-MP1, Fant et al. s.n., UT; CHIC, UTC. Pp-MP2, CHIC, UTC. Pr-SR2, Tietmeyer et al. s.n., NV; CHIC, UTC. Pr-SCR1, Tietmeyer et al. s.n., UT; CHIC, UTC. Pp-WWM1, Tietmeyer et al. s.n. , Fant et al. s.n., NV; CHIC. Pr-PM1, Fant et al. s.n., NV; CHIC, UTC. Pr- UT; CHIC, UTC. Pp-WWM2, Tietmeyer et al. s.n., UT; CHIC, Pp-WWM3, PNM1, Tietmeyer et al. s.n., NV; CHIC, UTC. January 2011] Kramer et al. —Q1 Landscape genetics of Great Basin Penstemon 121

Appendix 2. Locus name, repeat type, GenBank accession number, primer sequence, and size range (bp) for eight Penstemon microsatellite loci.

Locus Repeat GenBank accession number Primer sequence (5 ′ – 3 ′ ) Size range (bp)

Pen02 (TC)14 (CA)13 DQ917423 F: TTCTATGCTTCGTTAACCCAAAA 163 – 245 R: GGTCGTATTGGTCCTTTCCA

Pen04 (TC)22 DQ917425 F: GATGGAAAATGTGCCAGGAC 209 – 287 R: CTCTGCGGTGCATGAAAGTA

Pen05 (TC)25 DQ917426 F: CAGATAGGGTGGAGGGGCTA 159 – 245 R: CAACCCAATCTGGTCGATCT

Pen06 (TG)9 (GA)12 GU902974 F: TGTTGACAGTTTTAATTGAAAGGAA 185 – 253 R: GAGGCCAGAAATGTTCCAAA

Pen18 (CT)20 (CA)20 DQ917428 F: CTCATGATGATTGTGCGGATA 530 – 616 R: ACAACTCTCGCACTCTCACG

Pen23 (GA)21 DQ917430 F: TGGTCTGATTTCAGGAAAAGC 148 – 206 R: TGCTCAAGACGATAATAAAAGTGC

Pen24 (GT)9 (GA)22 EF203408 F: TCAAATTGAGAAAATGAGTGAAAGTC 145 – 225 R: ATATGGTGGGACCTTTCGTG

Pen25 (CT)29 DQ917431 F: GATGATCACCCAAGTTGCTT 120 – 176 R: CCTAATGCACGAGGCAAACT