Knaus, Cronn and Liston 19

Knaus, Cronn and Liston 19

Knaus et al.

A fistful of Astragalus

A fistful of Astragalus: the morphometric architecture of an infra-specific group.1

Brian J. Knaus2, 4

Rich Cronn3

Aaron Liston2

1Manuscript received ______; revision accepted ______.

2Department of Botany and Plant Pathaology, Oregon State University, Corvallis, OR 97331 USA.

3Pacific Northwest Research Station, USDA Forest Service, 3200 SW Jefferson Way, Corvallis, OR 97331 USA.

4Author for correspondence (e-mail: ).

BJK thanks Richard Halse (OSC) for arranging herbarium loans and support in the herbarium. Nancy Mandel and Randy Johnson (USDA FS, PNW) provided help with statistical analyses. Peter Dolan (OSU) helped fit sine waves to monthly data. Chris Poklemba (USDA FS, PNW) helped with propagation of A. lentiginosus at the Corvallis FSL. Dana York, Kathy Davis, Dell Heter, and Patrick and Christine Whitmarsh provided inspiration, locality information, and housing during collections. Lisa Graumlich (Montana State University), John King (Lone Pine Research, MT), Mark Fishbein (Portland State University), Lucinda McDade (Rancho Santa Ana Botanical Garden), Constance Millar and Bob Westfall (USDA FS, PSW) provided inspiration to BJK to pursue this study as a graduate program. The title is a tribute to Rupert Barneby who titled a series of his papers “Pugillus Astragalorum.”

Grants!!! NPSO, NNPS, Hardman Award
ABSTRACT

The study of infra-taxa has historically been considered the study of incipient species. The species Astragalus lentiginosus (Fabaceae) is the most taxonomically complex species in the United States flora. The implausible amount of diversity within A. lentiginosus is reflected by its taxonomic history. Morphometric data presented here indicate that the varieties lack clear regions of distinction, which is congruent with their circumscription as infra-taxa. Significant correlations to climatic parameters suggest that the great diversity within A. lentiginosus may be due to local adaptation. Existing infra-specific circumscription is surprisingly similar to statistical optimization. K-means clustering was employed to determine the number of groups but failed to result in an optimal number of groups, suggesting that the varieties are clinal and can be divided into an arbitrary number of infra-taxa. The bewildering amount of diversity contained within the species Astragalus lentiginosus begs for decomposition yet its clinal nature precludes it from division into discrete groups. The use of infra-taxa in this species appears useful in that it divides this bewildering diversity, however there does not appear to be an optimal method based on morphology to decompose this species. Varieties of A. lentiginosus are interpreted as important documentation of infraspecific diversity, however the authors wish to stress that these groups should be interpreted as clinal in nature and not discrete.

Keywords: Astragalus; clines; Fabaceae; Great Basin; Mojave Desert; morphometrics; speciation.

(up to 8)

INTRODUCTION

The ‘species’ is considered the fundamental unit of biology (Stebbins, 1950; Mayr and Ashlock, 1969; Raven and Johnson, 2002; Coyne and Orr, 2004); but see (Bachmann, 1998) Campbell and Reese, 200X. While the PLANTS database of United States plants (USDA, 2006) includes 33,383 species it also includes 3,853 taxa at infraspecific ranks (table 1), indicating that around 11% of the species in the United States flora include infra-specifics (taxa recognized at the ranks of subspecies or variety). If the species is the fundamental unit of biology then of what value is the infra-specific rank and why do we have so many of them?

A unifying theme among species concepts is that the species is somehow ‘discrete’ (Mayden, 1997; Coyne and Orr, 2004) even though the metric is debatable (e.g., significant morphological distinctiveness, reproductive isolation, reciprocal monophyly, etc.). For example, the Biological Species Concept (Mayr and Ashlock, 1969) indicates that a species is a group of entities that are reproductively isolated from other ‘species.’ This is philosophically attractive because it implies that these entities (the species) no longer share a common evolutionary path due to an inability to share genetic material. However, theory indicates that adaptive divergence, another important concept in evolutionary biology, can occur in spite of gene flow (Wu, 2001; Via, 2002). This indicates that ‘groups’ of organisms can diverge to occupy different adaptive peaks while reproductive barriers may be incomplete or nonexistent. Therefore an important ‘unit’ of evolution may not necessarily require reproductive isolation. However, as long as the transfer of genetic material is possible, there is the possibility of intermediates which may represent poorly adapted individuals or individuals that are adapted to selective forces that are intermediate to the ends of the spectrum. We believe Infra-taxa may represent these entities which represent intermediates along a continuum that is too great to be considered as one.

If a species this is discrete than how does one delineate infraspecies, which therefore must be somehow non-discrete? Here we employ the most taxonomically complicated species in the United States Flora, Astragalus lentiginosus Dougl. ex Hook. (Fabaceae; USDA, NRCS 2007, table 1), to explore the value and delineation of infraspecies.

A Brief History of Infra-Taxa— Linnaeus is credited with providing the modern system of binomenal nomenclature however he also employed the trinomial at the rank of ‘variety’ (Linnaeus, 1753). Linnaeus considered the species to be the product of creation while the variety to be variation that has arisen since the creation (Stearn, 1957). Modern nomenclature has adapted an evolutionary system, however, the systems share the concept that infraspecies are recently derived.

One path to speciation described by Darwin (Darwin, 1859) is the increase in variation to a point where the magnitude of variation is no longer maintainable, resulting in divergence that ends in distinct species. Subsequently he relied on the multitude of artificial selection experiments performed informally by breeders (Darwin, 18??; animals under domestication) to demonstrate this increase in variation, and how this variation can accumulate in a relatively short amount of time. These ideas were somewhat formalized by Fisher (Fisher, 1958) who described the idea of ‘steady states’ and their maintenance, which would break down at large magnitudes.

A slightly different perspective was presented by Huxley (Huxley, 1938, 1939) through the study of clines. Huxley addressed the study of large amounts of discontinuous variation and sought to classify it. Huxley tried to decompose the problem of clines, in part, by proposing the idea of stepped clines, clines in which different groups possess a shallower slope than the entire group. This usually requires the a priori determination of groups, a move that is as contentious as the argument between lumpers and splitters.

Wilson and Brown (Wilson and Brown, 1953) criticized the subspecific rank. Among their points were that the naming of these groups detracted attention from the species and implied a discrete nature to these infraspecific groups. This is misleading because it is the ‘species’ that is supposedly discrete. This leaves the subspecies as a group of entities whose divisions appear arbitrary and therefore may have little value. In a rebuttal to these criticisms Mayr (1953) agreed that the infraspecific rank confused the importance of the species (which should be the focus of biology) but defended the infraspecies as an important record of infraspecific diversity.

The controversy surrounding infraspecies continues in the literature. Perhaps the most recent critic being Zink (Zink, 2004) who presents mitochondrial data as refuting the evolutionary relavence of the infraspecies. A shortcoming of Zink’s argument is that it doesn’t address current research topics such as the coalescence (Hudson, 1991; Nordborg, 2001; Hudson and Coyne, 2002; Roseberg and Nordborg, 2002) or adaptive divergence (Wu, 2001; Via, 2002; Dieckmann et al., 2004). Haig et al. (Haig et al., 2006) have provided a recent review of the infraspecies in the context of the Endangered Species Act and biological conservation. They supported recognition of the infraspecies in part on grounds that in the United States legal protection is applied only to named groups of organisms (particularly in plants) which puts an emphasis on recognizing polymorphism, even if it may be geographical.

What’s wrong with geographical variation???

Much of the theoretical discussion of the speciation process (Hudson, 1991; Nordborg, 2001; Wu, 2001; Roseberg and Nordborg, 2002; Via, 2002; Dieckmann et al., 2004) employs explicitly genetic models of evolution. While the discrete character of molecular genetic data (e.g., A, T, G or C) promises a discrete answer these authors present theoretical rationale for the existence of genetic intermediates. Here we choose to focus on the morphometrics of a varietal complex. The phenotype has many obvious relations to the genotype (Falconer and Mackay, 1996; Waitt and Levin, 1997; Walsh, 2001) and is of great relevance to the species problem (Rieseberg, Wood, and Baack, 2006). The vast majority of plant taxa have been circumscribed based on the Linnaean Species Concept, or Morphological Species Concept (Mayden, 1997), based on its ease of application and relatively long history. The quantification of the morphological aspects of a taxon of evolutionary interest is therefore a logical first step in gaining inference into processes that may be active within the group.

The Most Taxonomically Complex Species in the United States Flora— Astragalus lentiginosus Dougl. ex Hook. (Fabaceae) is the most taxonomically complex species in the North American Flora (USDA, NRCS, 2007, table 1). The species is distributed throughout the arid regions of western North America (Fig. 1) where it frequently occupies disturbed, saline, or otherwise marginal habitats. Many of the varieties were originally described as species (Hooker, 1833; Gray, 1856, 1863; Sheldon, 1894). As collections increased intermediate forms became apparent which led to their reduction as varieties (Jones, 1895, 1923). Rydberg (Rydberg, 1929a, b) employed a very different species concept, elevating the varieties of A. lentiginosus to species in the genera Cystium (inflated pods) and Tium (slightly inflated pods). Within the genus Cystium he included the subgenera Lentiginosa ,Coulteriana, and Diphysa which were separated based on inflorescence length, flower size, and flower color. This grouping is no longer formally recognized with a name but is reflected in the modern keys to the group (Barneby, 1964, 1989; Spellenberg, 1993; Isely, 1998; Welsh et al., 2003) which largely follow the treatments of Barneby. Barneby (Barneby, 1945) returned the group to a single species with numerous varieties. Through time several varieties have been reduced to synonymy (Barneby, 1964, 1989) while new varieties have also been described (Barneby, 1977; Welsh, 1981; Welsh and Barneby, 1981; Welsh and Atwood, 2001). As many as 40 varieties have been recognized at once (Barneby, 1945; Isely, 1998), currently we recognize 35 (USDA, NRCS, 2007, table 1).

MATERIALS AND METHODS

Morphometric measurements— Specimens from major western herbaria were measured including BRY, JEPS, NESH, NY, ORE, OSC, POM, RENO, RM, RSA, UC and WILLU. A goal of 20 specimens, possessing fruit and flower, were attained for the common varieties of A. lentiginosus (table 2). Sampling was focused on the widespread varieties for the practical reason that these taxa have been most abundantly collected. Some endemic taxa only occur at a few localities (e.g., A. l. vars. albifolius, sesquimetralis, and piscinensis). Sampling of endemic taxa, even if there were sufficient specimens, would have confounded the sampling of single populations versus the range of widespread taxa.

Fourteen linear morphometric characters were chosen from the keys of Barneby (Barneby, 1945, 1964, 1989) and measured with a ruler, electronic caliper, or ocular micrometer (table 3). Three measurements were made of each structure whenever possible and a mean of these values was recorded. Measurements were made from different parts of the plant (e.g., different stems or racemes) whenever possible or from different plants when more than one was on a sheet. The fourteen characters were: stem internode length, leaf rachis length, leaf petiole length, leaflet number, leaflet width, leaflet length, peduncle length, floral axis in fruit, keel length, calyx tooth length, calyx tube length, pod length, pod height, pod valve thickness, beak length.

Assesment of infraspecific structure— All data were examined for univariate normality and heteroscedacity using histograms and scatterplots (using the generic functions ‘hist’ and ‘plot’, R package ‘graphics’)(R Development Core Team, 2007). The characters ‘floral axis in fruit’ and ‘pod valve thickness’ were natural log transformed to improve normality. Principle components analysis was performed (using the function ‘princomp’ in the R package ‘stats’) on a matrix of correlations to explore patterns of structure in the group. Principle components analysis is an eigen analysis used to explore data without the a priori assignment of groups (Everitt, 2005; Tabachnick and Fidell, 2007). A matrix of correlations was chosen to give each character equal weighting in the analysis.

Assesment of varieties— Discriminant function analysis was performed (using the function ‘lda’ in the R package ‘MASS’) to explore structure given the a priori grouping as varieties (Barneby, 1964, 1989) both with and without the use of latitude and longitude as additional explanatory variables (fig. 3). All characters were standardized by standard deviations in order to equalize the magnitude of each character. Discriminant function analysis seeks to build multivariate functions that best discriminate among a priori groups (Everitt, 2005; Tabachnick and Fidell, 2007). These functions can then be plotted in ordination space.

In order to asses the optimal number of groups k-means clustering was performed on the data (using the function kmeans in the R package ‘stats’). K-means analysis uses a predefined number of groups and utilizes an optimality criterion to fit the data within these groups (Everitt, 2005; Tabachnick and Fidell, 2007). A sum of squares can then be calculated to assess the fit. Note that as the number of groups increases the sum of squares is expected to decrease, therefore researchers usually examine plots for a breakpoint in the data where additional groups no longer appear to dramatically decrease the sum of squares. Standardization by standard deviation was performed to equalize the contribution of each trait. In order to explore the sensitivity of the data to the optimality criterion several methods were employed (Hartigan-Wong, Lloyd, Forgy, and MacQueen).

Phenological standardization— In order to explore climatic trends in morphology the PRISM dataset (Daly et al., 2002) was used. Specimens were assigned a latitude and longitude by referencing label information to a place-name database (topozone.com) or converted from township and range when available this data was available (www.esg.montana.edu/gl/trs-data.html). The spatial join command in ESRI’s ArcView Spatial Analyst (Redlands, California) was used to extract elevation, monthly minimum and maximum temperature and monthly precipitation from the PRISM dataset.