Phylogeny and Adaptation in the Encelia Alliance (Asteraceae: Helliantheae)

The three related genera Encelia, Enceliopsis, and Geraea comprise the alliance. The first consists primarily of shrubs and the latter two of herbaceous perennials and an annual. With the exception of two Encelia species of arid South America, all inhabit southwestern North America. Enceliopsis and Geraea are sister groups, and together form the sister group to Encelia, which includes two major clades. Especially in Encelia, there are diverse morphologies and a variety of ecological strategies marked by differences in habitat, vestiture, water balance, and photosynthetic parameters. The North American species of all three genera are obligate outcrossers, all with n = 18 chromosomes. Although intergeneric hybrids are largely sterile, interspecific hybrids in Encelia are fertile in the wild and in cultivation. Hybrids in the wild are largely restricted to F ,s, except in areas of human disturbance. Two true-breeding species are of homoploid hybrid origin, and are evidently isolated from the parent species through external ecological barriers involving selection against backcross progeny. Studies of the chloroplast genome and the intercistronic transcribed spacer (ITS) of nrDNA show clear differentiation of the genera, but much less variation within Encelia, even between phenotypically disparate species, suggesting recent divergence. Because the species are interfertile, it will be possible to study the genetics of the traits that distinguish the species and contribute to their differences.


INTRODUCTION
Section 9.3 of Ehleringer and Clark (1987) is titled "Encelia: A model system for the study of adaptation." Fortunately, perhaps, there are no criteria for "model systems." In many cases, a model system is simply one that has been well studied, but most model systems also have features that lend them to certain types of research. The lab mouse, Mus musculus, is arguably a case of the former, its primary advantage being that it is amenable to captivity. Drosophila, Caenorhabditis elegans, and Arabidopsis thaliana are perhaps in the latter category; each has aspects of its biology that strongly favor certain types of studies. In the study of evolutionary biology of plants, Clarkia certainly stands out.
Encelia and its relatives certainly aren't the "new Drosophila" or "new Clarkia," but there are features of the group that support studies of the phylogeny of ecological adaptation (as explored by Ehleringer and Clark 1987), the nature of hybrid speciation, the role of breeding barriers in speciation, and the inheritance , Address for correspondence. of species-diagnostic characters. The purpose of this paper is to outline the state of current knowledge of the genus and point t6 directions for future research.
The "Encelia alliance," as described here, consists of the genera Encelia, Enceliopsis, and Geraea. They are mostly perennials, all with n = 18 chromosomes, and all inhabitants of arid regions, mainly in southwestern North America. PHYLOGENY Clark (1986) first presented a phylogeny for Encelia, and substantially the same tree was published by Ehleringer and Clark (1987). The relationships between Encelia, Enceliopsis, and Geraea were explored by Sanders and Clark (1987), Nishida and Clark (1988), and Nishida (1988). All these studies were based on phenotypic features: capitulum characters, including UV reflectance (Clark and Sanders 1986), trichome type and distribution (Clark et al. 1980;Clark and Clark 1984;Charest-Clark 1984;Charest 1988), and secondary chemistry and associated anatomical features (Budzikiewicz et al. 1984, 1986Proksch et al. 1988). More recently, Clark (1995) provided preliminary phylogenetic data from the internal transcribed spacer of nuclear ribosomal DNA (ITS).
Although these phylogenetic studies have seemed a "work in progress, " since full character support for a specific tree has never been published, they agree in a number of major features. First, they strongly support the monophyly of Encelia (diagnosed by a constricted apical notch of the achene and the general absence of achene awns; the clade has occurred in every tree in every analysis, with 100% bootstrap and jackknife in combined morphology and ITS [Clark 1995]), Enceliopsis sensu stricto (scapose habit and large capitula [Sanders and Clark 1987]; again occurring in all trees), and a sister-group relationship between Enceliopsis and Geraea (thickened unpigmented crown of the achene [Sanders and Clark 1987]; occurring in all trees). They weakly support the monophyly of Geraea and a sister-group relationship between Enceliopsis + Geraea and Encelia.
Within Encelia, the presence of two major clades (hereafter called the Jrutescens clade and the californica clade) is well supported. Encelia nutans Eastwood cannot be clearly assigned to either (ITS data are not yet available), and although phenotypic data clearly assign E. ravenii Wiggins  Flourensia has been traditionally viewed as an outgroup to the Encelia alliance (Clark 1986), and ITS data appear to support this. Preliminary sequences suggest that two "misfits" in Encelia, E. stenophylla E. L. Greene and E. scaposa (A. Gray) A. Gray, are more closely related to Flourensia than to the Encelia alliance. Morphology and biogeography provide some support to this view (Clark unpubl.)

Geraea
This genus consists of two species, G. viscida (A. Gray) S. F. Blake, an eradiate herbaceous perennial with sessile glandular leaves, and Geraea canescens Torr. & Gray, a radiate annual with alate-petioled pilose leaves (Fig. 1). Even though they have long been congeners, they are superficially dissimilar. In her study of the genus, Nishida (1988) demonstrated additional differences, but suggested that the genus is monophyletic, diagnosed by strongly obcuneate achenes (common in the Heliantheae, but otherwise unknown in the Encelia alliance) and a tapered disk UTAH ARIZONA corolla throat (as contrasted with the bulbous throat of the relatives). Although Proksch et al. (1986) showed phytochemical similarities between G. canescens and Enceliopsis, they found a different set of chemical constituents in G. viscida. ITS sequences are preliminary, but the clade occurred in 72% of 799 most parsimonious trees in an analysis of ITS 1 (Clark unpubl.).
Geraea canescens hybridizes in the wild with E. Jarinosa (Kyhos 1967) and E. Jrutescens A. Gray (Clark unpubl.). Although this has suggested close relationship, it is important to realize that these two species are the only members of the alliance with which G. canescens is sympatric. In cultivation, G. canescens has been successfully crossed with several other Encelia species (Clark unpubl.). Two important facts emerge from this study. First, all hybrids between G. canescens and other species are sterile (Kyhos 1967, found asynapsis in G. canescens X E. Jarinosa hybrids). Second, G. canescens is always the pollen parent; no other species will successfully pollinate G. canescens. This unilateral incompatibility is not unexpected, since G. canescens is an annual (thus highly disadvantaged by loss of eggs to sterile hybrid progeny) and all the other species are perennials.
Of course, the G. canescens X G. viscida hybrid would be very interesting. Is it sterile? How do the Arizona contrasting features of the parents sort out? Clark (unpubl.) and Nishida (1988) made numerous attempts to form this hybrid. As expected, all attempts using G. canescens as the ovulate parent resulted in no fruit set. When G. viscida was the ovulate parent, all the progeny were indistinguishable from G. viscida; evidently the G. canescens pollen had overwhelmed the selfincompatibility system of G. viscida, allowing selfpollination. Attempts to circumvent this by excising G. viscida anthers resulted in ovule abortion. Thus, no hybrids were ever formed.

Enceliopsis
Enceliopsis consists of three species of suffrutescent, generally scap9se perennials with large capitula (Encelia nutans spent most of its nomenclatural history in Enceliopsis, but it shares the achene synapomorphy with Encelia and no synapomorphies with Enceliopsis, so it is considered here in Encelia, where it was originally described). As mentioned above, the monophyly of the genus is well established. One species, E. nudicaulis (A. Gray) A. Nels., is widespread across the Great Basin, whereas the other two are restricted endemics, E. argophylla (D.C. Eaton) A. Nels. occurring on gypsum soils around Lake Mead in Nevada, and E. covillei (A. Nels.) S. E Blake being found only in a few canyons on the west slope of the Panamint Mountains west of Death Valley, California (Fig. 2). The latter two are similar in appearance, were once considered conspecific, and share features that can best be interpreted as apomorphies (Sanders and Clark 1987). However, preliminary ITS data seem to ally E. covillei and E. nudicaulis. Sanders and Clark (unpubl.) formed hybrids, E. nudicaulis X E. covillei and E. nudicaulis X E. argophylla, but neither survived long enough to examine their fertility (all members of the genus are difficult in cultivation). Like Geraea and almost all Encelia, the species of Enceliopsis seem to be self-incompatible.

Baja California
Nevada Utah Baja California Sur

Encelia
Encelia comprises 15 species, the aforementioned E. nutans an herbaceous perennial, and the rest shrubs. Two are South American: E. hispida Anderss. from some of the Galapagos Islands and E. canescens from Peru, Chile, and Argentina. The rest inhabit southwestern North America.
The genus contains two well-marked clades. The californica clade (Fig. 3, 4) (Proksch and Clark 1986) and ultraviolet-reflecting ray corollas (Clark and Sanders 1986). All but E. densifolia also share brown disk corollas. The jrutescens clade (Fig. 5), comprising E. actoni Elmer, E. jrutescens, E. resinifera C. Clark, and E. virginensis A. Nels., is marked by few or no benzopyrans or benzofurans , a lack of resin ducts (Proksch and Clark 1986), and erect fruiting heads with expanded paleae (Clark 1986). Encelia ravenii (Fig. 6) shares these features, but ITS sequences ally it instead with the californica clade; both data sets need to be re-examined.
Perhaps because of its unusual habit, Encelia nutans

Nevada Utah
Baja California ( Fig. 6) has no clear-cut synapomorphies tying it to either clade, and no molecular data are yet available. It perennates as a thick underground semisucculent rootstock, emerging during the winter and spring to flower and fruit, with the above-ground parts withering in the summer. Its biology is in need of detailed study. Because it is difficult in cultivation, no data pertaining to self-incompatibility or hybridization are available.

California
With the exception of E. canescens, and the probable exception of E. hispida, all the species are selfincompatible. Even E. can esc ens does not ordinarily spontaneously self. Pollination is generalist, involving butterflies, solitary bees, occasional honeybees, and beetles (Clark, unpubl. observations).
All the species of Encelia are interfertile in cultivation, and their Fls are fertile, as well as all studied F 2 s and backcrosses. Spontaneous natural hybrids are common in areas of sympatry. E. actoni X E. Jrutes-cens, E. Jarinosa X E. asperijolia, E. Jarinosa X E. Jrutescens, E. Jarinosa X E. halimijolia, E. Jarinosa X E. palmeri, E. virginensis X E. Jrutescens, E. vento rum X E. asperijolia, and E. vento rum X E. palmeri have been documented.
Because E. vento rum has dissected leaves with linear lobes, natural hybrids between it and E. palmeri are distinctive, and were first described as a separate species (E. Xlaciniata Vasey & Rose). Kyhos et al. (1981) carried out a detailed study of those hybrids, and more limited studies of other hybrid combinations confirm a general pattern in the genus. In the natural environment, hybrids are ordinarily of F J phenotype (by observation it is of course impossible to tell if they are actually FJs). To the extent that the parent species show a clear separation of habitat, the hybrids are found in areas of intermediate habitat. In areas of human disturbance, and occasionally in areas of natural disturbance, F2 and backcross plants may be found, but these are otherwise rare. Progeny tests show, however, that these recombinant forms are produced far in excess of their appearance in the habitat; Kyhos et al. (1981) showed that most seeds produced by E. Xlaciniata at a certain locality were backcrosses to one parent species or the other (as expected, since the FJs were in much lower frequency than the parents), but not a single backcross individual was seen in the population.
All this suggests that, whereas F J hybrids may in some cases have adaptive features allowing them to persist, backcrosses and F 2 s (except perhaps those that resemble FJs) are at a severe selective disadvantage, in many cases being totally eliminated from the population. The cause of this harsh selection against non-F J progeny is unknown, but the effect is clear-cut.

SPECIES OF HYBRID ORIGIN
Factors affecting interspecific hybridization in Encelia become all the more important, because two species show clear evidence of having originated from hybrids (Clark and Kyhos 1979;Clark et al. 1980;Allan et al. 1993Allan et al. , 1997Clark and Allan 1997 Although Riesberg and Ellstrand (1993) have pointed out the difficulties in inferring hybrid origin, there are nevertheless three somewhat independent lines of evidence that in combination can support such hypotheses. First, species of hybrid origin may be intermediate between their parent species (although in a number of documented cases they are not, and species that appear intermediate have not infrequently proven to be not of hybrid origin). Second, species of hybrid origin may agree in phenotype with F J hybrids. This may seem at first to be the same as intermediacy, but F Js themselves are not always intermediate, or are intermediate in specific ways. Third, species of hybrid origin may share features of the parent species that otherwise would be autapomorphies of those species, thus producing a pattern of shared characters that could not be easily explained by divergent evolution. Even in combination, these criteria can never prove that a species is of hybrid origin, but they can provide a weight of evidence.
Encelia virginensis A. Nels. is generally intermediate between its parents. Measurements were made of pedicel width, number of rays, ray corolla length, leaf length, leaf width, petiole length, capitulum height, and capitulum width of 119 E. virginensis, E. Jrutescens, E. actoni, and Jrutescens X actoni F J hybrids. These measurements were analyzed by principal coordinate analysis; both the E. Jrutescens plants and F J hybrids appear between the putative parents in a plot of the first two coordinate axes (Fig. 7).
Length and width of ten achenes each from five populations of E. actoni and E. Jrutescens and two populations of E. virginensis were measured. Encelia virginensis was intermediate to the parent species in both length and width; it differed significantly (t-test, P < .05) from both parents in width, but not in length (Fig. 8). Encelia virginensis also inhabits a climate somewhat intermediate to the parents (Fig. 9), but with a slightly greater annual range of temperature, consistent with its more interior distribution (Fig. 10).
Plants of E. virginensis are normally indistinguish-  able from F\ hybrids between E. actoni and E. jrutescens. (Prior to molecular studies, only the presence of populations of similar and apparently true-breeding E. virginensis plants hundreds of kilometers from either parent supported the idea that E. virginensis was a species, rather than a named hybrid.) Figure 11 shows the leaf trichomes of all three species and the F\.
Encelia virginensis shares apomorphies with both parents. With E. jrutescens it shares broad multicellular-based uniseriate leaf hairs (Clark et al. 1980, Ehleringer andCook 1986 (Allan et al. 1993(Allan et al. , 1997. In addition, the internal transcribed spacers of nuclear rONA (ITS) of several individual E. virginensis combine the sequences of both parents in a manner strongly suggestive of hybrid origin. The parent species differ by 9 bases; at each site, every sampled E. virginensis has either the E. actoni base, the E. jrutescens base, or a polymorphism consisting of both bases. Furthermore, few of the individuals are identical to each other, suggesting a "sorting out" of parental lineages (Clark in prep.).
Encelia asperifolia is less precisely intermediate to its parents, and in fact was originally described as a subspecies of E. californica (Clark and Kyhos 1980). Its achenes are smaller in both length and width than either parent (Fig. 12). It has capitula similar to those of E. californica, with long ray florets (as contrasted with the eradiate condition of E. jrutescens), but leaves more like those of E. jrutescens, so much so that sterile specimens are occasionally identified as that species, leading to the incorrect distribution of E. jrutescens in Baja California provided by Wiggins (1980). Unlike E. virginensis, E. asperifolia is allopatric to one of its parents (E. jrutescens) and only parapatric to the other (Fig. 14). It occupies climates, however, that are intermediate between the climates of the parent species (Fig. 13). Encelia asperifolia plants are not especially similar to F\ hybrids, the latter being more or less intermediate to the parents.
Like E. virginensis, the single sequenced E. asperifolia appears to have chimeric ITS. The parent species differ by 21 bases. Encelia asperifolia has E. calif or-. nica bases at eight sites, E. Jrutescens at seven, unique bases at four sites (including a site for which the parents are identical), and ambiguous bases at three. Its unique ITS mutations, its lack of precise intermediacy, and its disjunct distribution suggest that E. asperifolia is an older species than E. virginensis.
Theoretical Mechanisms Riesberg (1991) and Riesberg et al. (1990Riesberg et al. ( , 1995 have outlined a mechanism for hybrid speciation in Helianthus that agrees with the "recombinational speciation" model proposed by Grant (1981). In these examples from Helianthus, F1s have reduced fertility. The F 2 s are often more fertile than backcrosses, and repeated crossing among F2 and later generations restores fertility in the new hybrid species through recombination of chromosome segments (in essence, the entire genome becomes chimeric). Thus, the newly forming species is isolated from the parents by internal reproductive barriers. Other documented cases of hy-brid speciation (Arnold 1993;Gallez and Gottlieb 1982) involve similar mechanisms.
Hybrid speciation in Encelia differs from this model in many respects. In Encelia, F1s are fully fertile, and there are few or no chromosomal differences between parent species. Backcrosses are formed as readily as F 2 s, but are strongly selected against, and seldom occur as mature plants in the wild. This corresponds to another model proposed by Grant (1981), hybrid speciation with external barriers. Although he included such mechanisms as pollinator fidelity in "external barriers," the strong selection against backcross progeny serves the same purpose.

XEROPHYTIC ADAPTATIONS
All species of the three genera Encelia, Enceliopsis, and Geraea inhabit arid regions, and all have adaptations to the seasonal lack of water that characterizes them. These adaptations have been extensively studied by Ehleringer and his colleagues (summarized in Ehleringer and Clark 1987). Four different mechanisms serve different species in the group.
Geraea canescens is an annual, and thus a droughtavoider. Although annuals are very common in the regions inhabited by members of the Encelia alliance, they are uncommon among the group of Heliantheae to which it belongs.
Encelia Jarinosa exemplifies leaf shading by thick reflective pubescence. This shading reduces leaf temperatures, preventing leaf damage even when there is inadequate water for transpirational cooling. Encelia palmeri, E. can esc ens, E. actoni, E. raven ii, and the three species of Enceliopsis all have reflective pubescent leaves, and although they have not been studied as extensively as E. farinosa, the same mechanism can be inferred. Encelia densifolia is especially interesting in that its pubescence is wettable, and is much more Fig. 14. Encelia asperifolia,E. californica,and E. frutescens subsp. glandulosa,geographic distribution . reflective when dry than when wet with fog (Harrington and Clark 1989).

Baja California
Within the alliance, reflective pubescent leaves would seem to be ancestral, but the outgroup Flourensia has generally glabrous glutinous leaves, a condition not found in the Encelia alliance. An understanding of the relationships of these two groups among the rest of the Heliantheae will clarify the ancestral condition in the Encelia alliance .
Encelia californica is an example of a drought-deciduous species; it loses its glabrous leaves in the dry season. It inhabits coastal climates that are wetter than the typical desert haunts of the other species, but its range is characterized by no precipitation during the six warm months. Geraea viscida and Encelia nutans are also drought-deciduous, in both cases losing the above-ground parts of the plant. Encelia halimifolia is nearly as glabrous as E. californica, but the nature of its adaptation is not well understood.
Encelia frutescens exhibits the other adaptation used by glabrous-leaved species: transpirational cooling. It lives primarily in desert washes and other areas with water at depth, and thus can continue to transpire into the dry season. (It, too, loses its leaves when the water runs out, but that does not always happen.) Encelia ventorum lives on coastal sand dunes, another habitat with deep water, and most likely uses the same mechanism.

ECOTYPES OF ENCELIA FARlNOSA
Encelia farinosa is the most widespread species in the genus (Fig. 3). It contains three infraspecific taxa, two (f. farinosa and f. phenicodonta S. F. Blake) differing only by disk color (Kyhos 1971), and the third (var. radians Brandegee) lacking the characteristic leaf pubescence.
Other variation is not taxonomically distinguished. The plants of cismontane southern California (in western San Bernardino and Riverside counties) differ in vegetative appearance from those of the adjacent Colorado Desert, being taller and less "dome-shaped. " These " ecotypes" meet and to some extent intergrade in the San Gorgonio Pass . Miller (1988) grew plants from both areas in common gardens in both areas. He found that the leaves of desert-origin plants were significantly more reflective (around 50% reflectance) at 670 nm (the red absorption peak of the photosynthetic action spectrum) than cismontane-origin plants (30-40%) in both test gardens. Cismontane-origin plants had significantly longer peduncles (20-22 cm) than desert-origin plants (10-15 cm) in both test gardens. These results imply a genetic difference between the two "ecotypes" for these traits. However, even though cismontane plants growing in habitat had a significantly greater height! width ratio than desert plants (a measure of the "dome" -shape of the latter), this difference did not persist in the test gardens, suggesting phenotypic plasticity.
The maintenance of these differences across a narrow geographic region suggests substantial selection pressures, and indeed the contact between the two forms in San Gorgonio Pass also marks a transition from a generally coastal flora to a desert flora . This region most likely represents a secondary contact between the forms. The cismontane form is otherwise isolated from the bulk of the species.

EVOLUTION OF ENCELIA
Although the species show some striking differences in morphology and ecology, the low levels of variation in ITS and the chloroplast genome and the ability of all the species to interbreed suggest a recent origin, especially as compared to the sister group comprising Enceliopsis and Geraea. Despite the recent origin, and despite the potential for hybridization, the species are easily distinguished. Most taxonomic confusion has resulted from the overrepresentation of hybrids in herbarium collections (the, "1 don 't recognize that plant so I'll put one in the press," syndrome).
More important, the basic pattern of evolution in the group appears to be divergent. Grant (1981) characterized similar situations in other genera as "syngamea" ; by this reckoning, the genus Encelia would be a syngameon, equivalent to a single " biological" species, and the individual taxonomic species would be "semispecies." However, both morphological and mo-lecular evidence clearly support a branching cladogram. There is no evidence of introgression in the genus, and excepting the two species of hybrid origin, no evidence of reticulation. Clearly the mechanisms that restrict backcrosses in the natural environment thus restrict gene flow between the species.
This divergent evolution has led to clear morphological and ecological differentiation among the species. Because the species are so easily grown and hybridized in cultivation, Encelia provides a great opportunity to directly study the inheritance of traits that distinguish species.