Genetic characterization of three varieties of Astragalus lentiginosus (Fabaceae)

Abstract Knaus, B. J. (Oregon State University, Department of Botany and Plant Pathology, 2082 Cordley Hall, Corvallis, OR, 97331-2902, U.S.A.; knausb@science.oregonstate.edu), R. C. Cronn (USDA Forest Service Pacific Northwest Research Station, 3200 SW Jefferson Way, Corvallis, OR, 97331, U.S.A.; rcronn@fs.fed.us) & A. Liston (Oregon State University, Department of Botany and Plant Pathology, 2082 Cordley Hall, Corvallis, OR, 97331-2902, U.S.A.; listona@science.oregonstate.edu). Genetic characterization of three varieties of Astragalus lentiginosus (Fabaceae). Brittonia 57: 334–344. 2005.—Astragalus lentiginosus is a polymorphic species that occurs in geologically young habitats and whose varietal circumscription implies active morphological and genetic differentiation. In this preliminary study, we evaluate the potential of amplified fragment length polymorphism (AFLP) markers to resolve infraspecific taxa in three varieties of Astragalus lentiginosus. Distance-based principle coordinate and neighbor-joining analyses result in clustering of individuals that is congruent with population origin and varietal circumscription. Analysis of molecular variance of two Oregon varieties demonstrates that varietal categories account for 11% of the total variance; in contrast, geographic proximity does not contribute to the total variance. AFLPs demonstrate an ability to discriminate varieties of A. lentiginosus despite a potentially confounding geographic pattern, and may prove effective at inferring relationships throughout the group.

Astragalus lentiginosus Dougl. ex Hook. is a polymorphic species consisting of 40 varieties (lsely, 1998) distributed through out the Intermountain and Desert Southwest regions of North America. The current va rietal circumscription reflects Barneby's (1945) view that A. lentiginosus comprises multiple lines of evolutionary radiation, as well as occasional reticulation, both of which may explain the morphological in termediates between varieties. Many of the habitats where extant populations occur (e.g., inland dune systems, desert seeps, mountain ridges) were profoundly different during the Pleistocene (Grayson, 1993). Oc currence on these geologically young hab itats suggests that the separation of current 1 To whom correspondence should be addressed. Brittonia, 57(4), 2005, pp. 334-344. populations and vanetles may have oc curred as recently as the late Pleistocene/ early Holocene period. This study of A. len tiginosus represents an investigation of a di verse group whose circumscription reflects dramatic morphological variation and im plies genetic differentiation. Jones (1923) provided the first revision of Astragalus during the twentieth century. Among his innovations was the reduction of the section Diphysi to a single species, Astragalus lentiginosus. He characterized this species as the "most variable of all As tragali" (Jones, 1923, p. 123). As defined by Jones (1923), this species consisted of 18 varieties, many of which were originally described as species. Within his varieties Jones included "forms" (never validly pub lished) that had originally been recognized as species or varieties. Applying a very different taxonomic ap proach, Rydberg (1929) split Astragalus into 28 genera (summarized by Barneby, 1964). In doing so, he divided A. lentigi nosus between the genera Cystium (33 spe cies, all currently known as A. lentiginosus) and Tium (39 species, only three of which are currently known as A. lentiginosus). This revision resulted in most of Jones's ( 1923) varieties (and forms) being raised to species (Barneby, 1945). Barneby's treatments (1945Barneby's treatments ( , 1964Barneby's treatments ( , 1989) reduced the group back to a single species, Astragalus lentiginosus, with ca. 40 varie ties. This classification came from the re duction of many of Rydberg's (1929) spe cies to varieties, or similarly elevating many of Jones's (1923) "forms" to varie ties. Since Barneby's 1945 treatment there has been a reduction in the original number of names accepted as varieties in part due to additional collecting that has blurred some intervarietal distinctions (Barneby, 1964(Barneby, , 1989. This trend has been partially offset by the description of new varieties since 1964 (e.g., Barneby, 1977;Welsh, 1981 ). While the currently recognized num ber of varieties is 40 (lsely, 1998) there is still an active debate as to what constitutes A. lentiginosus (Alexander, 2005) as well as to the validity of the varieties.
The varieties of Astragalus lentiginosus are morphologically distinct when observed at distant stations, yet when geographically proximal their distinctiveness may become obscured to the point where they are indis tinguishable (Barneby, 1964, p. 922). Most of the group is characterized by an inflated bilocular pod with a false septum that in trudes from the abaxial surface and is in complete in the beak. Barneby (1945Barneby ( , 1964Barneby ( , 1989 created major divisions within the group based on flower size (keel greater or less than 8.5 mm), raceme length (greater or less than 4 em), and flower color (purple or white).
Attempts to evaluate taxonomic interpre tations within Astragalus utilizing molecu lar methods have been complicated by low levels of divergence among the New World species. Chromosome number variation was identified as potentially useful at the infra generic level (Barneby, 1964;Spellenberg, 1976), but phylogenetic inference was lim ited by the large size of the group and ap parent homoplasy for this character. Molec ular phylogenetic studies of nuclear ribo somal DNA internal transcribed spacers (Wojciechowski et al., 1993), chloroplast DNA (Liston, 1992;Sanderson & Doyle, 1993), and com'bined nuclear and chloro plast DNA datasets (Wojciechowski et al., 1999) have demonstrated the monophyly of the New World aneuploid species of As tragalus. However, none of these studies has resolved interspecific relationships within this clade.
Amplified fragment length polymor phism (AFLP) analysis (Vos et al., 1995) is an anonymous genetic fingerprinting tech nique developed for plant breeding that has been adapted to studies of natural popula tions (Wolfe & Liston, 1998;Mueller & Wolfenbarger, 1999). The AFLP method has been used to infer interspecific (Abdel fattah et al., 2002;Beardsley et al., 2003) and intraspecific (Brouat et al., 2004;Juan et al., 2004;Travis et al., 1996) relation ships as well as population level dynamics (He et al., 2004). This method was applied by Travis et al. (1996) to evaluate popula tion-level differentiation of Astragalus cremnophylax var. cremnophylax occurring at the Grand Canyon, U.S.A. Results from that study showed strong differentiation of north and south rim populations, suggesting that AFLPs are sufficiently sensitive to dis cern genetic differences in recently di verged lineages, such as A. lentiginosus.
In this preliminary study we utilize AFLP analysis to test Barneby's (1945Barneby's ( , 1964Barneby's ( , 1989) taxonomic treatment of three Astragalus lentiginosus varieties. We focus on the question of whether the sampled populations of A. lentiginosus varieties dis play genetic relationships that can be attri buted to geographic proximity, or whether the varieties exhibit genetic cohesiveness.

PLANT MATERIAL
Plants included in this study were select ed to determine whether genetic differenti ation could be discerned across the latitu dinal extent of Astragalus lentiginosus.
[VOL. 57 the Columbia Basin to northern California and the northwestern Great Basin. Barneby (1945Barneby ( , 1964Barneby ( , 1989 placed this taxon among the A. lentiginosus possessing short flowers (keel length< 8.5 mm), short racemes (axis < 4 em long in fruit), and white flowers. Astragalus lentiginosus var. lentiginosus is distinctive among the varieties of A. lenti ginosus in having moderately inflated pods with a coarse texture (thick walls), de scribed as stiffly papery, leathery, or woody (as opposed to thinly papery or membra naceous).
Astragalus lentiginosus var. salinus (Howell) Barneby is a short-lived perennial consisting of diffuse and ascending stems occurring in sagebrush and alkaline flats from eastern Oregon to southern Idaho and throughout the northwestern Great Basin. This variety possesses short flowers (keel length< 8.5 mm long), short racemes (axis < 4 em long in fruit), and white flowers. Astragalus lentiginosus var. salinus is geo graphically proximal to A. l. var. lentigi nosus and can appear identical to the latter except for the thinly papery texture and in flation of the pods.
Astragalus lentiginosus var. variabilis Barneby is a monocarpic to short-lived pe rennial plant consisting of erect to ascend ing diffuse stems occurring on sandy flats, washes, desert playas, and sometimes on in land dunes throughout the Mojave Desert of southern California and southern Neva da. This variety possesses large flowers (keel length> 8.5 mm), long racemes (axis > 4 em long in fruit) and purple flowers. This variety was chosen because it is geo graphically disjunct from the above varie ties. In addition to these traits, A. l. var. var iabilis differs from the previous two varie ties in its robust and sometimes erect habit and its frequently cinereous to canescent vestiture.
Fragments were resolved on an Applied Biosystems Inc. 3100 capillary fragment analyzer (36 em capillary; POP4 polymer) using a 1 :2 dilution of "+ 3" product. Trace files were scored using ABI GeneMapper v3.0 following Rinehart (2004). To de crease the probability of including homo plastic AFLP bands we included only bands between 200-490 bp with peak heights above 200 relative fluorescent units (Koop man & Gort, 2004 ).

DATA ANALYSIS
AFLP products were scored as the pres ence (1) or absence (0) of bands. Population and variety level band frequencies were cal culated using GenAlEx v5.1 (Peakall & Smouse, 2001). Two distances were calcu lated from these data. Pairwise squared Eu clidian distances (E2i j = !k[Xki-X.�gF. where i and j represent a pair of objects, and k denotes cases where Yki and Y �g are both pre sent) were calculated using GenAlEx and used as input for analysis of molecular var iance (AMOVA; Excoffier et al., 1992). The Lynch shared band similarity index (Sx y = 2Nx/(Nx + N ) where Nx is the number y of bands in sample x, N Y is the number of bands in sample y, and Nx is the number y of bands shared between both samples (Lynch, 1990;Lamboy, 1994), as calculated using NTSYSpc version 2.1 (Rohlf, 2000) was used for neighbor-joining and ordina tion analyses. Because this index is equiv alent to one minus the Dice (1945) coinci dence index, we refer to this measure as "Dice distance." Principle coordinates analysis (PCoA) of Dice distances was used to evaluate inter individual distances in low dimension space to determine whether a priori population and varietal designations matched patterns of genetic variation. Dice distances were evaluated using the neighbor-joining algo rithm (NJ; Saitou & Nei, 1987). Both PCoA and NJ were implemented in NTSYS-pc (Rohlf, 2000). Partitioning of genetic vari ation between populations nested within va rieties was evaluated by AMOV A using the approach of Excoffier et al. (1992), as im plemented with GenAlEx v5.1 (Peakall & Smouse, 2001). For variance partitioning, we looked at two groupings: at populations within varieties or regions (<I>PR), and among varieties or regions (<I>RT). Signifi cance was measured with 999 permutations of individuals among hierarchical levels and recalculating null distributions of the test statistic (Excoffier et al., 1992).

Results
Two primer pairs resulted in 184 loci scored between 200 and 490 bp with an av erage of 36.5 fragments per individual. Ten loci were monomorphic and 174 were poly morphic. Seventy three of the bands ( 40%) had a frequency of less than 0.05. Number of loci and private alleles per population are summarized in Table II. Ordination by principle coordinates anal ysis (PCoA; Fig. 2) shows varieties to oc cupy distinct regions of coordinate space. The first axis accounted for 22.8% of the variation, the second accounted for 16.5%, and the third accounted for 10.3% of the variation (49.6% of the total variation). Populations of Astragalus lentiginosus var. variabilis formed a single discrete cluster while most individuals of A. l. var. lentigi nosus also formed a cluster. In contrast, the two populations of A. l. var. salinus clus tered separately from each other. Individu als lentiginosus_1-5, lentiginosus_1-6, len tiginosus_1-9, lentiginosus_2-1, and sali nus_1-11 (A, B, C, D, E, respectively; Fig.  2) fall in a region of intermediacy between populations salinus_1 and lentiginosus_l. However, individual salinus_1-11 (E) is separated from the others on the third axis by its strongly negative eigenvector ( -0.0446) as opposed to the positive values on the third axis of individuals lentigino sus_1-5, lentiginosus_1-6, and lentigino sus_1-9 (0.0166, 0.0176 and 0.0212 respec tively). Individual lentiginosus-2-1 (D) is notable in its intermediate third axis eigen vector ( -0.0230) and its position as closer to the origin than the other members of its population. Close inspection of AFLP trace files show that these were high quality am plifications, so intermediacy is not the result of aberrant amplification.
Cluster analysis utilizing the neighbor joining method (Fig. 3) resulted in a den drogram that reflects ordination results with each variety forming a distinct cluster. In dividual lentiginosus_2-1 (D; Fig. 2) was the sole exception in that it neither clustered with its source population or variety. Indi viduals showing intermediacy on PCoA ( Fig. 2; lentiginosus_1-5, lentiginosus_1-6, lentiginosus_1-9 and salinus_1-11; A, B, C and D respectively) cluster with their re spective populations (Fig. 3). Dice genetic distances show a general pattern where within population genetic dis tance was lowest (Table III; mean .0823 = ± 0.0202, n 154), within variety genetic = distance was slightly higher (mean = 0.1055 ± 0.0201, n = 176), and between variety genetic distances were greatest (mean 0.1216 ± 0.0176, n 705).

= =
Among-population genetic distances for As tragalus lentiginosus var. salinus (0.119) is comparable to the average among popula tion divergence. Analysis of molecular variance revealed significant structure within the sampled va rieties and populations across the range of sampled sites. Analyses were conducted three ways to evaluate diversity at different scales. Analysis #1 included all varieties and populations to address the question of whether there was structure to the sampled populations and varieties of Astragalus len tiginosus throughout its range. Variance de composition showed that most of the vari ation in A. lentiginosus was contained with in populations (67.3%, p < 0.001; Table  IV), due in large part to the abundance of bands restricted to one or a few individuals within a population. The remaining 32.7% of the total variance was apportioned among populations, with 13.4% of the var  iation distributed among vaneties (p < 0.001) and 19.3% distributed among pop ulations within varieties (p < 0.001).
Because Astragalus lentiginosus var. var iabilis is several hundred kilometers dis junct (Table III; Fig. 1 ), it may have con tributed disproportionately to the varietal component of the variance. Analysis #2 was performed to determine whether geograph ically proximal populations retained the va rietal signal exhibited in the full dataset. Removing A. l. var. variabilis populations from the analysis resulted in little change in the among-variety variance component (<I>RT = 10.9%, p < 0.001; Table IV). The third analysis was conducted to test whether geographic proximity accounted for more of the variance than varietal cir cumscription. The Oregon populations were grouped based on geographic proximity (re gion) instead of taxonomic variety. Popu lations lentiginosus..1 and salinus_1 were grouped into one region while populations lentiginosus_2 and salinus_2 were grouped into the second region. This grouping ac counted for none of the variance at the re = gional level of hierarchy (<I>RT 0, p > 0. 999; Table IV). These results indicate that the variation revealed by AFLP shows a strong within-variety signal, and that geo graphic proximity does not contribute to the measured variance. 's (1945, 1964, 1989) circum scription of Astragalus lentiginosus repre sents an exceptionally detailed account of a group of organisms that appear too different to be considered the same taxon but also exhibit intergradation that precludes one from delimiting species; hence he chose the Taxonomic circumscription within this group was explored with ordination by principle coordinates analysis (PCoA), which resulted in each variety occupying distinct coordinate space. Cluster analysis using the neighbor-joining method largely concurred with ordination results. A series of AMOV As were then employed to pro vide statistical significance to varietal cir cumscription. These data suggest that inter breeding occurs preferentially within the sampled varieties as opposed to between in tervarietal populations in close geographic proximity.

Bameby
The comparison of Dice genetic distance with geographic distance (Table III) shows interpopulation genetic distance within As tragalus lentiginosus var. variabilis and A. l. var. lentiginosus to be relatively small. In contrast, the two populations of A. l. var salinus, supported in both PCoA and clus tering, show considerably greater diver gence than the other two varieties.
One of these populations, salinus_1, was collected near Whitehorse Ranch, a site mentioned by Bameby (1964, p. 922) as possessing individuals that defy classifica tion (due to intermediate morphologies with Astragalus lentiginosus var. jloribundus). Genetic divergence between the sampled populations of A. l. var. salinus may be due to factors such as introgression or hybrid ization with other varieties (e.g., A. l. var. J!oribundus; not included in this study), or It may simply represent a greater molecular diversity in this variety. The inclusion of more varieties (such as A. l. var. jloribun dus) in future studies may provide addition al evidence for this observation.

INTER-VARIETAL INTERMEDIACY WITHIN A. LENTIGINOSUS
Four individuals of Astragalus lentigi nosus var. lentiginosus and one of A. l. var. salinus were positioned near the origin in the PCoA analysis (Fig. 2). The difference between varieties is resolved on the third axis where they are clustered according to variety (except individual lentiginosus_2-1) and in the neighbor-joining dendrogram, suggesting differentiation among the varie ties. The position of these individuals, par ticularly lentiginosus-2-1, near the origin, may indicate limited intergradation among the varieties. Due to the inclusion of only three varieties in this analysis, inferences about the cause of this finding cannot yet be made.
COMPARISON WITH A. CREMNOPHYLAX VAR. CREMNOPHYLAX Travis et al. (1996) used AFLPs to dis tinguish populations of Astragalus crem nophylax var. cremnophylax on the north and south rim of the Grand Canyon. AMO V A results partitioned 63% of the variance amongst the north and south rim popula tions, as contrasted to 32% variance among populations of A. lentiginosus reported here. Principle coordinate analysis showed discrete clusters in the A. cremnophylax data as compared to the clusters that in cluded intermediate individuals reported in the current study. This suggests that popu lations of A. cremnophylax var. cremnophy lax are more structured and divergent than the varieties included in this current study despite our sampling of A. lentiginosus over a much larger geographic range (1000 km). Recent allozyme data (Allphin et al., in press) suggest the north and south rim pop ulations warrant specific rank. This inter pretation seems to be in accordance with the amount of AFLP divergence found be tween varieties in this study.

CONCLUDING REMARKS
In this pilot study we have demonstrated genetic differentiation between varieties of Astragalus lentiginosus. If populations of A. lentiginosus represented a simple pattern of isolation by distance one would expect a cline of molecular diversity across the tax on's range. Populations collected at distant sites would appear distinct, yet intermediate samples would fall along a continuum, both morphologically and in terms of the molec ular data. However, in the present study va rietal distinction has been demonstrated in both disjunct and geographically proximal varieties suggesting that simple isolation by distance may not be responsible for all of the differentiation among varieties. While inferences about mechanisms are premature at this time, these data do not contradict Bameby's circumscription of these three varieties.
This study accompanied with that of Travis et al. (1996) show AFLPs to be a promising tool in the study of Astragalus species. Here we have demonstrated the ability of AFLPs to discriminate infraspe cific relationships within A. lentiginosus de spite a potentially confounding geographic pattern. This suggests their use may prove effective at inferring relationships through out the group.