HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (17) Citing Articles via Google Scholar Google Scholar Articles by Morrell, PeterL. Articles by Rieseberg, LorenH. Search for Related Content PubMed Articles by Morrell, PeterL. Articles by Rieseberg, LorenH. Agricola Articles by Morrell, PeterL. Articles by Rieseberg, LorenH.

Molecular tests of the proposed diploid hybrid originof Gilia achilleifolia (Polemoniaceae)¹

^a Rancho Santa Ana Botanic Garden, 1500 NorthCollege Avenue, Claremont, California 91711-3157 ^b Department of Biology, Indiana University,Bloomington, Indiana 47405-6801

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Gilia achilleifolia is a putative diploid hybrid species. Hybrid origin was hypothesized based on traditional biosystematicevidence (i.e., morphological, cytological, and crossability data),which may be insufficient to establish genealogical history. Here,phylogenetic analysis of sequence data from the internal transcribedspacer (ITS) regions is used to examine the relationship between theputative hybrid species and its proposed parents. Isozyme variation isassayed to test for genetic additivity in the putative hybrid taxon andmorphological data are analyzed cladistically to evaluate the charactersthat led to the original hypothesis of hybrid origin. The ITS-basedgene tree placed G. achilleifolia in two divergent clades, eachsister to one of the putative parental lineages. Little isozymeadditivity was observed and G. achilleifolia possessed sixunique alleles among 42 alleles observed. However, ITS and isozymetrees differed in their placement of the two lineages of G.achilleifolia; both lineages are closer to a third putative parentin the isozyme tree. Also, G. achilleifolia is intermediate orpolymorphic for all nine morphological characteristics differentiatingthe parental species. Sorting of ancestral polymorphisms cannot easilyaccount for expression patterns of seven of these characters. In ourview, these results fail to distinguish between alternative hypothesesof ancient hybrid origin and divergent evolution, belying the difficultyof detecting ancient hybrids.

Key Words: geneticadditivity • hybridspeciation • isozymes • ITS (InternalTranscribedSpacer) • morphology • recombinationalspeciation • trnL

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Natural hybridization between closely related taxa results in theproduction of recombinant genotypes of novel genetic composition. Someof these new genetic combinations may be favorable (Arnold and Hodges, 1995; Rieseberg et al., 1996), implying thathybridization between lineages, along with mutation within a lineage,can provide the raw material for adaptive evolution (Anderson, 1949). Hybrid speciation canstabilize or "fix" these potentially adaptive genecombinations (Grant, 1981). This studyexamines a putative case of diploid hybrid speciation in the wild plantspecies Gilia achilleifolia (Polemoniaceae), thought to beexemplary of the "recombinational" model of hybridspeciation (Grant, 1981). The speciesis compared to diploid hybrid species confirmed in other studies.

The idea that diploid hybrid speciation can result from geneticexchange between semisterile parental lineages was first proposed byMüntzing (1930, 1934, 1938). The theory was extended by Stebbins (1942, 1945, 1950,1957a, b), Grant(1958, 1963, 1981), Templeton(1981), and McCarthy, Asmussen, andAnderson (1995) and termed "recombinationalspeciation" (Grant, 1958). Arecombinational species possesses a new homozygous (or true breeding)combination of the chromosomal or genic sterility factorsdifferentiating the parental species. Recombination betweenchromosomally divergent parental taxa may increase chromosomal breakagein the hybrid population, which could accelerate the speciation process(Holm, Fitz-Earle, and Sharp, 1980) andfacilitate the development of reproductive isolation (Templeton, 1981; Rieseberg, Van Fossen, and Desrochers,1995). Once formed, recombinational species are fertile inintraspecific crosses, but intersterile when crossed to eitherparent.

This genetic model of speciation has several additional implications. Given that a hybrid must either coexist with one or both parentalspecies or colonize a new habitat, new species that are ecologicallyisolated from the parent species have an increased likelihood ofestablishment and persistence (Templeton,1981). In the absence of ecological divergence, a fitnessadvantage in hybrids relative to their parents is critical to successfulestablishment (McCarthy, Asmussen, andAnderson, 1995). Recombinational speciation also may befacilitated by inbreeding, as this serves to increase the frequency ofthe union of balanced gametes (Grant,1981; McCarthy, Asmussen, andAnderson, 1995). Outcrossing retards the speciation process,but given a sufficiently high selective advantage for the stabilizedderivative, the process becomes feasible even under conditions ofobligate outcrossing (McCarthy, Asmussen, andAnderson, 1995). From a macroevolutionary perspective, it isnoteworthy that segregation of sterility factors from a single set ofparents may produce multiple reproductively isolated derivatives(Stebbins, 1957b; Grant, 1958; Rieseberg,1991).

The recombinational model has been verified by the artificialsynthesis of several hybrid neospecies (Gerassimova, 1939; Smith and Daly, 1959), including at least threeexperiments that test the model under conditions of strong sterilitybarriers (Stebbins, 1957b; Grant, 1966b; Rieseberg et al., 1996). In all threeexperiments, several generations of backcrossing and/or selfing led tothe recovery of fully fertile hybrid derivatives that were at leastpartially reproductively isolated from their parents.

Although experimental studies demonstrate the feasibility of thismode of speciation, its frequency in nature is still unclear. More than50 putative examples of homoploid hybrid species are described in thebotanical literature (reviewed in Rieseberg,1997), yet only 16 have been rigorously tested with molecularmarkers. In our judgment, hybrid speciation has been convincinglydocumented in eight of these cases: Helianthus anomalus, H.deserticola, and H. paradoxus (Rieseberg, Carter, and Zona, 1990; Rieseberg, 1991; Rieseberg, Van Fossen, and Desrochers, 1995,Rieseberg et al., 1996); Irisnelsonii (Arnold, Bennett, and Zimmer,1990; Arnold, 1993);Paeonia emodi, and Paeonia spp. (Sang, Crawford, and Stuessy, 1995); Pinusdensata (Wang et al., 1990; Wang and Szmidt, 1994); and Stephanomeriadiegensis (Gallez and Gottlieb,1982). In the eight remaining cases, hybrid speciation eitherwas not confirmed (Rieseberg, Soltis, andPalmer, 1988; Rieseberg, Carter, andZona, 1990; Spooner, Systma, andSmith, 1991; Wolfe and Elisens,1993, 1994, 1995) or the molecular marker data were ambiguouswith regard to hybrid speciation (Crawford andOrnduff, 1989; Wendel, Stewart, andRetting, 1991; Allan, Clark, andRieseberg, 1997).

Testing hybridorigin
Taxa of hybrid origin should be characterized by a distinctivecombination of parental morphological and molecular characters. Withrespect to molecular characters, a recently derived hybrid speciesshould combine the alleles of its parents, but show few if any uniquealleles (Gallez and Gottlieb, 1982;Rieseberg, Carter, and Zona, 1990;Wolfe and Elisens, 1995). However, dueto extinction of parental alleles following the initial hybridizationevent(s), additivity often can only be detected by the analysis ofmultiple loci (Gallez and Gottlieb,1982). Allele extinction is particularly likely forcytoplasmic loci, where the effective number of alleles is reducedrelative to nuclear genes. Thus, hybridization events are sometimesdetected by incongruence between gene trees of different genomic origin(e.g., Smith and Sytsma, 1990; Wendel, Stewart, and Retting, 1991). Inaddition, the accumulation of genetic changes in the hybrid and parentallineages following speciation can erase evidence of hybridity overtime.

Although analysis of multiple loci differentiating the putativeparental species is a prerequisite for the successful detection ofhybrid taxa, phylogenetic information can play an important role aswell. Genetic additivity for biparentally inherited characters (e.g.,isozymes), of itself is not sufficient to determine whether the proposedhybrid is ancestral to its putative parents or derived throughhybridization (Rieseberg, Carter, and Zona,1990). However, by estimating relationships with gene treedata, it may be possible to discriminate between these hypotheses (e.g.,Rieseberg, Carter, and Zona,1990).

From a morphological perspective, recent hybrids are predicted to bea mosaic of both parental and intermediate morphologicalcharacteristics. Both hybrids and their parental taxa are expected toaccumulate genetic changes that over time will obscure the mosaicpattern of inheritance and make the detection of ancient hybrid originmore difficult (Gallez and Gottlieb,1982; Rieseberg and Ellstrand,1993). Morphological characters appear more likely thanmolecular characters to diverge rapidly following speciation, and thus,are less likely to be useful for the diagnosis of hybrid taxa regardlessof method of analysis (Rieseberg andEllstrand, 1993; Rieseberg andMorefield, 1994).

Here, we employ a battery of loci to test the putative hybrid originof G. achilleifolia including the chloroplast locustrnL, the internal transcribed spacer (ITS) of nuclearribosomal genes, isozymes, and morphological characteristics. We alsocompare G. achilleifolia with well-documented"recombinational" species from other groups and discuss somebiological properties common to hybrid species and theirparents.

Gilia achilleifolia asa hybrid species
The California endemic Gilia achilleifolia is one of theearliest and most thoroughly characterized putative examples ofrecombinational speciation (Grant,1954a, 1981). Hybrid originwas originally proposed primarily because of morphological intermediacyin G. achilleifolia relative to related taxa in Giliasection Gilia (Grant, 1953). Table 1 lists characterssuggested to be intermediate between G. achilleifolia and itsputative parents (Grant, 1953). Fourpossible hybrid parents were identified. Either G. angelensisor G. tricolor were considered one possible parent. EitherG. capitata ssp. abrotanifolia or G. capitatassp. staminea were proposed as the other hybrid parent(Grant, 1953, 1954a). Grant(1954a) stated that morphological features of extant G.tricolor seem to preclude it as a hybrid parent but do not excludepast races as a potential hybrid parent. Later descriptions of theorigin of G. achilleifolia emphasize the role of G.angelensis as a hybrid parent (Grant,1981).

Gilia achilleifolia is proposed to be of relatively ancienthybrid origin (Grant, 1953, 1954a). Grant(1953) noted that it arose before the origin of theallotetraploid derivative G. clivorum, which now occupies itsown ecological niche over a longitudinal range of >640 km (Grant, 1954b).

Gilia achilleifolia ssp. achilleifolia is a robustplant, with loose heads of 8–25 flowers on pedicels 1–2 mmlong. Flower size varies a great deal in the taxon (10–20 mm)(Day, 1993). As the "sunrace" designation suggests, the plants occur primarily in openareas, generally in loose soil on dry slopes and canyon walls inchaparral borders, and at the edge of open oak woodlands. Giliaachilleifolia ssp. multicaulis is generally a small,slender, spreading plant with a loose cymose inflorescence of 1–7flowers on pedicels 1–30 mm long. Flowers also vary in size inthis taxon, but are generally much smaller (5–10 mm) than those ofG. achilleifolia ssp. achilleifolia (Day, 1993). Habitat is similar to that of G.achilleifolia ssp. achilleifolia, but G.achilleifolia ssp. multicaulis is more frequently found inthe shade of oak woodlands. Both subspecies are self-compatible(Grant, 1953; Grant and Grant, 1965), but outcrossing ratevaries tremendously among populations (t =0.15–0.96) (Schoen, 1982a). Bothsubspecies flower in mid-to-late spring.

The proposed hybrid parents were believed to be from divergentlineages within Gilia section Gilia, and are distinctmorphologically (Grant, 1953). Gilia angelensis and G. tricolor are relatively small,slender, spreading annuals with a loose cymose inflorescence of1–10 flowers. Gilia angelensis occurs primarily indisturbed areas of natural or human origin, including road cuts orroadsides. Gilia tricolor is found in this habitat and in opensandy fields. Gilia tricolor is primarily outcrossing butfacultatively autogamous, while G. angelensis is predominantlyselfing (Grant, 1952b; Grant and Grant, 1965). The species differprimarily in that G. tricolor has large, showy, multicoloredflowers, while G. angelensis has flowers much smaller and palerin color. Both species flower in the early spring. Giliacapitata ssp. abrotanifolia and G. capitata ssp.staminea are robust annuals with dense heads of flowers. Gilia capitata ssp. staminea has 50–100 flowersper head, while G. capitata ssp. abrotanifoliagenerally has a more diffuse inflorescence of 25–50 flowers(Grant, 1952a; Day,1993). Both taxa occur in disturbed areas, with G.capitata ssp. abrotanifolia being most abundant in loosesoil and steep slopes, especially following fires. Giliacapitata ssp. staminea occurs on sandy soil, often at theedge of cultivated fields and along roadsides. Both species flower inmid-to-late spring. Collection localities for Giliaachilliefolia and the putative parental taxa are shown in Fig. 1.

View larger version (17K): [in this window] [in a new window]   Fig. 1. Map of California showing sampled populations of Gilia achilleifolia ssp. achilleifolia, G. achilleifolia ssp. multicaulis, G. angelensis, G. capitata ssp. abrotanifolia, and G. capitata ssp. staminea.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Taxonomy
All sampled taxa were identified according to the taxonomictreatments of Grant (e.g., 1950,1954a, 1956) and Day(1968, 1993) and compared toannotated specimens at RSA-POM and other herbaria. The majority ofsampled populations conform to previous taxonomy. However, thesubspecific determination of six populations of G.achilleifolia used here (BC, CR, GR, PI, MH, and MH2; Table 2) differed from the taxonomicdefinitions and annotated voucher specimens of Grant (1954a, 1956; V. Grant, University of Texas, personalcommunication). Variation in the inflorescence type, flower color, andflower size, the primary characters used to differentiate these taxa,led to the decision to interpret the subspecific identity of thesepopulations differently. The implications of this difference intaxonomy are considered in the Results andDiscussion.

DNA was extracted from leaf tissue using a CTAB extraction (Saghai-Moroof et al., 1984; Doyle and Doyle, 1987) modified by the additionof 1% sodium metabisulfite to the CTAB, a second chloroformextraction, and a reduction in volume to allow for use with 1.5-mLmicrocentrifuge tubes. The use of liquid nitrogen allowed the planttissue to be ground to a fine powder prior to the introduction of hotCTAB. For DNA extractions taken from dried leaf material or herbariumcollections, a small amount of silica sand was added to the tube beforegrinding to improve cell disruption.

Nuclear ribosomal DNA including the internal transcribed spacers, the5.8s region, and flanking portions of the 18s and 26s regions wasamplified using ITS primers ITS5 and ITS4 (Baldwin, 1992). Chloroplast DNA from thetrnL (UAA) intron and the intergenic spacer between thetrnL (UAA) 3' exon and trnF (GAA) was amplifiedusing primers c and f (Taberlet et al.,1991). Initial surveys of the trnL region foundlimited variation within the study group. Therefore, a subset of thesamples included in the ITS survey were sequenced in the trnLstudy.

Initial sequencing of both ITS and trnL employed the Sangermethod as described in Swensen (1996). Sequencing employed primers ITS2, ITS3, ITS4I, and ITS5I (Porter, 1997) and trnL primers c, d, e,and f (Taberlet et al., 1991). Cyclesequencing was used to generate the majority of sequence data. Forcycle sequencing, PCR products were purified using MilliporeUltrafree–MC(TM) filters (Millipore Corporation Bedford,Massachusetts). Cycle sequencing reactions were purified usingCentri–Sep(TM) Columns (Princeton Separation, Inc. Adelphia,New Jersey) to remove unincorporated primers, dyes, enzymes, etc. AnApplied Biosystems 373A automated DNA sequencer and the PRISM(TM)DyeDeoxy(TM) Terminator Kit (Perkin Elmer, Foster City, California)were used following the manufacturers' instructions. DNA of alltaxa were sequenced 5' to 3' and 3' to 5' toverify accuracy of coding.

Isozymes
Ten populations representing the proposed hybrid taxon and three ofthe four possible parents were sampled for isozyme variation. Giliatricolor was excluded from the isozyme analysis based onpreliminary ITS data that eliminated it as a possible parent (seebelow). The populations chosen are allopatric with related species,although G. angelensis does occur in canyons adjacent to the WWpopulation of G. capitata ssp. abrotanifolia. Cellulose acetate electrophoresis was performed as described in Richardson, Baverstock, and Adams (1986). Twelve presumptive loci were polymorphic and produced consistentresults. They were alcohol dehydrogenase (Adh-1,2), glutamateoxaloacetate transaminase (Got-1,2), isocitrate dehydrogenase(Idh-1), malic enzyme (Me-1), menadione reductase(Mnr-1), phosphoglucomutase (Pgm-1),phosphoglucoisomerase (Pgi-1,2), and 6–phosphogluconatedehydrogenase (6Pgd-1,2). Gels were run at 115–140 V;times varied from 30 to 80 min.

Morphology
All morphological characters used in previous studies of the group(Mason and Grant, 1948; Grant, 1950, 1952a, b,1954a, b, c, 1965, 1966a;Schoen, 1977, 1982b; Steele,1986) were evaluated for cladistic analysis of Giliasection Gilia. Cladistic methodology was chosen so thatcharacter polarity and homology could be more rigorously assessed. Characters were selected that appeared to be developmentallyindependent, could be coded as discrete states, and tended not to varywithin populations. Some characters used in the original proposal ofhybrid origin did not meet one or more of these criteria. They wereeither scored differently than in previous studies or excluded from thephylogenetic analysis (see notes in Table 1). All populations werescored for all morphological characters. Scoring for each populationwas based on observations of herbarium specimens, greenhouse-grownplants, and/or individuals in natural populations. Pollen surfacefeatures were observed with scanning electron microscopy. All otherfeatures were scored with the light microscope or the unaided eye. Scores for each population were combined for use in taxon-levelanalysis. Scoring was compared to published reports and to additionalcollections available at RSA-POM to insure accuracy and to overcome theproblem of small sample size, especially for taxa not critical to thestudy.

Data analysis
PAUP 3.1.1 (Swofford, 1993) wasused for maximum parsimony analyses of sequence and morphological dataand to calculate genetic distances from the ITS data set. MacClade3.0.1 (Maddison and Maddison, 1992)was used to produce constraint trees, to test alternate phylogenetichypotheses, and to map morphological characters on to the ITS-basedtree.

Bases were coded as polymorphic only when two bases at a positionshowed equally strong signals. This minimized the number of polymorphiccells in the data matrix and reduced the potential for error due tointerpretation of sequencing artifacts as polymorphisms. Unambiguousalignment of ITS sequence was possible because only a single indel wasidentified in the section Gilia accessions. For thetrnL sequence, initial manual alignment was tested usingCLUSTALW (Thompson, Higgins, and Gibson,1994). Binary coding of insertions and deletions was not useddue to very limited indel variation in the ITS data set, and becauseoverlapping indels in the trnL data resulted in ambiguity inhomology assessment.

Morphological characters polymorphic within taxa were coded aspolymorphic for phylogenetic analyses. For character mapping,characters polymorphic within populations were scored as having multiplestates.

Heuristic searches were performed for maximum parsimony analysis ofthe ITS sequence data set. Initial searches were followed by 1000replicates, saving no more than one tree per replicate, to search foradditional islands of most parsimonious trees (Maddison, 1991). With the trnL andmorphological data sets, the branch and bound method was used. Treerobustness for all data sets was calculated with the bootstrap method(Felsenstein, 1985). Bootstrapanalysis involved 1000 replicates, with uninformative charactersexcluded (Harshman, 1994) to decreaseprocessing time, and saving no more than 100 trees per replicate.

Levels of genetic variation within and among populations wereestimated by computing parameters of population genetic variationincluding percentage of polymorphic loci (P), mean number of alleles perlocus (A), mean number of alleles per polymorphic locus(A_p), mean observed heterozygosity(H_o), and mean expected heterozygosity(H_e). All of the above were calculated usingGENESTAT-PC (Lewis and Whitkus, 1989)except H_o, which was calculated manually. Geneticidentities and distances among populations (Nei,1972) were calculated using PHYLIP (Felsenstein, 1995) and used to generate anunrooted tree by the Neighbor-Joining method (N-J) of Saitou and Nei (1987). The repeatability ofeach cluster in the N-J tree was tested using 1000 bootstrap replicatesas implemented by PHYLIP.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
ITSsequence data
Fifty-four variable nucleotide sites were identified within the studygroup, 33 of which were phylogenetically informative. The ITS and 5.8ssequence data set had an aligned length of 628 bases. No sequencevariation in the 5.8s region was identified. Additivity at individualnucleotides, which has been used to infer hybrid origin in several cases(e.g., Kim and Jansen, 1994; Sang, Crawford, and Stuessy, 1995), was not foundin G. achilleifolia or other diploid Gilia species,though it does occur in tetraploid Gilia species (Morrell, 1997).

Maximum parsimony analysis of the ITS sequence data produced a singletree (Length = 71, Consistency Index [CI] =0.803, Retention Index [RI] = 0.898; Fig. 2). Accessions of G.achilleifolia are found in two divergent clades. The most similaraccessions of G. achilleifolia from each clade are separated byten steps in the analysis, whereas the most divergent accessions of thespecies are separated by 20 steps. Sequence divergence values betweenG. achilleifolia accessions from the two clades range from 1.6to 2.9% (Table3).

View larger version (33K): [in this window] [in a new window]   Fig. 2. The single phylogenetic tree from maximum parsimony analysis of the ITS sequence data set (length = 71, CI = 0.803, RI = 0.898). Bootstrap values above 50% are shown. Gilia stellata was designated as the outgroup. Gilia achilleifolia samples are in boldface. Populations that correspond to the three compatibility groups identified by Grant (1954a) are designated N, C, and S for the northern, central, and southern compatibility groups. Population names followed by circles have "sun race" morphology those followed by triangles have "shade race" morphology representing the taxonomy of Grant (1954a) (vouchers at RSA-POM).

Adding the constraint of a single origin for G.achilleifolia increased the internal length of the tree (i.e.,excluding the outgroup) by five steps and resulted in changes of theconsistency index from 0.803 to 0.750 and in the retention index from0.898 to 0861. Tree number increased from one to 178.

Gilia capitata is placed at the base of the ITS tree, and isparaphyletic with regard to rest of section Gilia. The base ofthe ITS tree has limited resolution, but this gene tree suggests thatmultiple lineages, including the putative hybrid parents of G.achilleifolia, evolved from G. capitata-likeancestors.

trnL sequencedata
The trnL regions exhibited limited variation in the studygroup. Ten variable nucleotide sites were identified over 979 bases ofaligned sequence. Of these, only three sites were phylogeneticallyinformative. Variation in the form of insertions and deletions was alsopresent, but was not phylogenetically informative. Maximum parsimonyanalysis produced 450, nine step trees (not shown) with no homoplasy. Although the trnL trees were consistent with the ITS phylogeny,variation was insufficient to assess the relationship between theproposed hybrid and its putative parents.

Isozymes
Forty-two putative alleles were identified at 12 presumptive loci(Table 4). Allozymevariation for PGI could not be interpreted for population WW, apparentlydue to a complex pattern of locus duplication. Twenty alleles wereidentified which could potentially serve as "marker"alleles. That is, they were found in only one of the proposed hybridparents, either in Gilia angelensis or G. capitata,but not in both (Table 4). Of the marker alleles, five are the majority allele in the population inwhich they occur. Eight of the 20 marker alleles were found in at leastone of the putative hybrid populations, but none of these were found inall populations of the proposed hybrid. The proposed hybrid alsocontained six unique alleles (Table4). The six alleles occur among four populations, but allbut one are found in the MH2 population (Table 4). When broken down bysubspecies, there were no alleles unique to Gilia achilleifoliassp. achilleifolia, but three alleles found only in G.achilleifolia ssp. multicaulis (Table 5).

View larger version (12K): [in this window] [in a new window]   Fig 3. The unrooted N-J tree resulting from 1000 bootstrap replicates of the isozyme allele frequency data. Bootstrap values above 50% are shown. Gilia achilleifolia samples are in boldface. Population names followed by circles indicate "sun race" morphology; triangles indicate "shade race" morphology.

The isozyme-based N-J tree (Fig.3) placed the three populations of G. achilleifoliassp. achilleifolia in a discrete group, but populations ofG. achilleifolia ssp. multicaulis fell out separately. The observation that G. achilleifolia ssp.multicaulis populations have a greater genetic distance betweenthem than do the populations of G. achilleifolia ssp.achilleifolia is consistent with the greater level of ITSsequence variation between samples of G. achilleifolia ssp.multicaulis. Likewise, Schoen(1982b) found more divergence among northern G.achilleifolia ssp. multicaulis populations than amongsouthern G. achilleifolia ssp. achilleifoliapopulations. Although the N-J tree is unrooted, forced rooting usingingroup taxa does not change the basic topology.

Morphology
Twenty-three polymorphic morphological characters (Appendix) werescored for each population (Table6). Although we attempted to avoid characters that werepolymorphic within taxa, this was not always possible. The twosubspecies of Gilia achilleifolia were the most morphologicallyvariable taxa, both among and within populations—an observationconsistent with previous reports (Grant,1954a). Gilia achilleifolia had the same majoritycharacter state as both parents for 12 of the 23 characters (1, 2, 3, 5,7, 9, 12, 15, 16, 17, 20, 23; Appendix), was intermediate or polymorphicfor the majority parental character states for nine characters (4, 6, 8,10, 11, 13, 14, 18, 22; Appendix), and had a unique or novel phenotypefor two characters (19, 21; Appendix). If the subspecies of G.achilleifolia were analyzed separately, Giliaachilleifolia ssp. achilleifolia was intermediate orpolymorphic for the majority parental character states for eightcharacters, whereas G. achilleifolia ssp. multicauliswas intermediate or polymorphic for five characters. Seven of the nineintermediate or polymorphic characters found in G.achilleifolia (4, 6, 10, 11, 13, 14, 22; Appendix) were homoplasticon the ITS gene tree due to their expression in G.achilleifolia and thus could be explained by hybridization. Theother two characters were not homoplastic on the ITS gene tree due totheir expression in G. achilleifolia; intermediacy for thesecharacters may be due to the sorting of ancestral polymorphisms.

View larger version (23K): [in this window] [in a new window]   Fig 4. One of 460 most parsimonious trees of the morphological date set with polymorphic characters scored with multiple states (length = 28, CI = 0.821, RI = 0.914). Bootstrap values above 50% are shown. Gilia stellata was designated as the outgroup. Gilia achilleifolia samples are in boldface.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

On the other hand, an equally convincing argument could be made forhybrid origin. Gilia achilleifolia is intermediate orpolymorphic for all nine morphological characteristics differentiatingthe parental species, and expression patterns for seven of thesecharacters are not easily accounted for by the sorting of ancestralpolymorphisms. Although there are two unique characteristics,hybridization is known to generate evolutionary novelty (Rieseberg and Ellstrand, 1993), and somedivergent evolution is expected in ancient hybrid species. The averageisozyme genetic distance between the subspecies of G.achilleifolia and their putative parents (D = 0.33)is near half that between the most likely parental species (D =0.85), as would be predicted for hybrid origin. If the subspecies ofG. achilleifolia represented independent offshoots of eachparental species as suggested above, then isozyme divergence betweennonsister hybrid-parent pairs should be closer to 0.85. Moreover, theisozyme and ITS trees differ in their placement of the two subspecies ofG. achilleifolia, with the isozyme tree placing both subspeciescloser to a third putative parent, G. capitata ssp.staminea. Phylogenetic incongruence appears to be a hallmarkof hybridization (Rieseberg and Soltis,1991; Kellogg, Appels, andMason-Gamer, 1996). The lack of complete genetic additivityand presence of unique alleles in G. achilleifolia could be aby-product of the age of the hybrid speciation event, a partiallyselfing mating system, and sampling error. Moreover, many of the"marker" alleles are relatively rare in the parental speciesand might not have been represented in the ancestral parentalpopulations. Finally, the presence of two apparent G.achilleifolia lineages in the ITS and isozyme data sets is equallyconsistent with the sorting of ancestral polymorphisms following one ormore hybridization events, and the large amount of divergence betweenlineages could be explained by ancient hybrid origin, as the probabilityof the establishment of new alleles in a hybrid is expected to increasewith time (Gallez and Gottlieb,1982).

Phylogenetic analysis of the morphological data is not directlyrelevant to tests of the hybrid origin hypothesis, as the placement ofhybrids in phylogenetic trees does not appear to be predictable(McDade, 1990, 1992; Rieseberg andMorefield, 1994). Nonetheless, it is important to note thatunlike the ITS and allozyme trees, the two subspecies of G.achilleifolia are placed as a single, albeit weakly supported clade(Fig. 4). Two synapomorphiesthat support this clade are unique to G. achilleifolia save theoccurrence of the same pollen exine pattern in some individuals of theWW population of G. capitata ssp. abrotanifolia andlarge seed size in an unusual population of G. capitata ssp.pacifica, known as the Cape Medocino Race (see Grant, 1953). Unlike other characters common tothe two subspecies of G. achillefolia, such as inflorescencetype and filament length, it is difficult to explain these similaritiesas a product of mating system evolution. While it might be possible todetermine whether similarities in pollen exine pattern and seed size aredue to homologous genetic changes, it is likely to be very difficult todetermine whether hybrid origin is responsible for these homologies. Despite the fact that many of the morphological characters in G.achilleifolia appear highly homoplastic when mapped onto the ITStree, removal of G. achilleifolia from the morphologicalanalysis does not result in a large reduction in tree length ortopology. This appears to be due to the high level of polymorphism inthe species, and its placement at the base of the morphology-basedphylogenetic tree.

Even with more data, it may be difficult to distinguish between theindependent divergent origins and ancient hybrid origin hypotheses. Clearly, additional gene tree data might, under the right circumstances,allow us to rule out one of these hypotheses. For example, ifphylogenetic analyses of several independent genes generated treesidentical to those produced by ITS, the independent divergent originshypothesis would be supported. On the other hand, if populations sisterto one parent in the ITS tree were placed with the other parent in treesbased on one or more additional gene sequences, the hybrid originhypothesis would gain support. Additional allozyme data might alsoclarify the situation. For example, increased sampling of G.achilleifolia might increase the number of marker alleles detected,thus providing stronger support for hybrid origin. On the other hand,rejection of hybrid origin would be better supported if sampling ofadditional parental populations determined that unique alleles found inthe hybrid are indeed absent from the putative parental taxa.

Classification of populations according to Grant's (1954a, 1956) taxonomy might provide stronger support forthe hybrid origin hypothesis since the ITS lineages would includepopulations of both subspecies. However, the population designationsused here are based on the analysis of many morphological characters andaccord well with the ITS lineages and compatibility relationships. Thus, we are reluctant to use the discordance between Grant'sclassification and the ITS tree as evidence for hybrid origin.

Other explanations can also account for the patterns of variationobserved in G. achilleifolia and its close relatives, althoughthey seem less plausible. For example, G. achilleifolia couldrepresent a single polymorphic taxon that has arisen through divergentevolution. However, this hypothesis is in conflict with almost all ofthe available data. Grant (1954a)pointed out that the sun and shade races (G. achilleifolia ssp.achilleifolia and G. achilleifolia ssp.multicaulis) differ in shade tolerance, plant stature andarchitecture, size of the corolla, density of the inflorescence andlength of the pedicels, compatibility relationships, floral mechanisms,and breeding behavior. In addition to these differences, isozymestudies (Schoen, 1982a, and referencesherein) and ITS sequence data suggest that G. achilleifolia aspresently circumscribed consists of two geographically distinct andgenetically divergent species (Wyatt,1988). Furthermore, it appears that inflorescence type, theprimary character used to delineate the species, is highly variable andpolymorphic within G. achilleifolia.

It is also possible that G. achilleifolia represents asingle species that has experienced localized introgression from relatedspecies. Gilia achilleifolia has been proposed to be a part ofa "homogamic complex" of species connected by interspecificgene flow (Grant, 1953). However,extensive introgression into G. achilleifolia seems unlikelygiven the very limited cross-compatibility with related taxa (Grant, 1954c) and widespread introgression intothe species has not been proposed. Nonetheless, at least one populationhas been identified that appears to show signs of past introgressionfrom G. capitata staminea (Grant,1954a). Efforts to locate this population have not beensuccessful to date, and it may not be extant.

The biology of homoploid hybrid species
Because there appear to be only eight well-documented examples ofhomoploid hybrid speciation in the plant literature, it is difficult tomake valid generalizations about their biology. Nonetheless, somegeneral trends may be emerging. The eight confirmed hybrid speciesinclude one tree (Pinus densata), three perennial herbs(Iris nelsonii, Paeonia emodi, and Paeoniaspp.), and four annual herbs (Helianthus anomalus, H.deserticola, H. paradoxus, and Stephanomeriadiegensis). All have an outcrossing mating system and, exceptPinus densata, are pollinated by insects. The presence ofoutbreeding is unexpected, because hybrid speciation is predicted tooccur more readily in highly inbreeding populations (Grant, 1981; Templeton, 1981; McCarthy, Asmussen, and Anderson, 1995). Likewise, the high proportion of annual species of confirmed hybridorigin is unusual as hybridization may be more frequent in perennials(Ellstrand, Whitkus, and Rieseberg,1996). With a mixed-mating system, G. achilleifolia(if confirmed to be of hybrid origin) would be unique among homoploidhybrids, but would continue the apparent trend toward annual habit.

The eight confirmed hybrid species have diverged ecologically fromtheir parents. This is expected as basic ecological law does not allowtwo species to occupy the same niche. However, the habitat occupied bythe hybrid taxa often seems novel or extreme relative to that of theparental species rather than intermediate. For example, Pinusdensata occurs at higher altitudes than either parent (Wang et al., 1990) and three Helianthushybrid species are found in more xeric or marshy habitats than theirparents (Rieseberg, 1991). Fromthis perspective, it is noteworthy that Gilia achilleifoliaoccurs at higher latitude and in more mesic habitat than eitherparent.

Stebbins (1950) and Grant (1958) suggested that hybrid taxa might bemore genetically variable than their parents because they would combinethe alleles of both parents. Presumably, increased genetic variationwould enhance the evolutionary potential of the hybrid neospecies. Although plausible, this hypothesis has not been confirmed by empiricalevidence. The three Helianthus hybrid species actuallyexhibited lower levels of isozyme diversity than either parent asmeasured by estimates of percentage polymorphic loci, mean number ofalleles per locus, and mean heterozygosity (Rieseberg et al., 1991). Pinusdensata was slightly more variable genetically than either parent(Wang et al., 1990), whereasStephanomeria diegensis and Iris nelsonii were roughlyequivalent to their parents in terms of variability (Gallez and Gottlieb, 1982; Arnold, Bennett, and Zimmer, 1990). Bycontrast, our data indicate that in both levels of isozyme variation andin terms of ITS repeat type variation, G. achilleifolia isconsiderably more diverse genetically than either parent. Presumably,diversity levels in hybrid taxa will be strongly affected by severalfactors, including the number of parental individuals involved in theirorigin, the degree of genetic differentiation between the parentalspecies, species age, mating system, and historical contingency. Forexample, the lower than predicted levels of diversity in theHelianthus hybrids may indicate that a small number of parentalindividuals was involved in their origin, possibly via hybrid founderevents (Rieseberg, 1991).

Parental hybrid swarms have been identified forStephanomeria (Gallez and Gottlieb,1982), Helianthus (Heiser,1947), and Iris (Arnold,Bennett, and Zimmer, 1990; Arnold,Buckner, and Robinson, 1991), but are not known forPaeonia, Pinus, or among species in the G.angelensis and G. capitata lineages. Some populations ofG. angelensis and G. capitata abrotanifolia thatappear to have undergone past introgression with each other have beenidentified (Grant, 1952a, 1953, 1954a). However, the perceived effects of hybridization between these speciesare confined to a few restricted localities (Grant, 1952a). Finally, hybrid taxa inStephanomeria, Helianthus, and Iris appear to berestricted to a few localized populations (Arnold, Bennett, and Zimmer, 1990; Rieseberg, Carter, and Zona, 1990; Rieseberg, 1991) or limited geographic area(Gallez and Gottlieb, 1982), whereasG. achilleifolia, Pinus densata, Paeoniaemodi, and Paeonia spp. are much more widespread andabundant. Possibly, the differences in geographic ranges reflect timesince origin and/or the distribution of appropriatehabitat.

Taxonomicrevision
Based on the all available lines of evidence, G.achilleifolia now appears to be composed of two or more independentevolutionary lineages. This is exemplified in the ITS gene tree(Fig. 2), where all southernpopulations form a single clade and all northern populations are placedin a separate clade. This is concordant with compatibilityrelationships identified in G. achilleifolia (Grant, 1954a), with the northern, central, andsouthern compatibility groups all representing distinct clades (Fig. 2).

In some populations, the morphological characters typicallyconsidered diagnostic for the subspecies (e.g., corolla size and colorand the number of flowers per head; V. Grant, University of Texas,personal communication) are not consistent. GR and PI are examples, withblue-violet corollas (like ssp. achilleifolia) equal to or lessthan 8 mm long (like ssp. multicaulis) and clusters of fiveflowers per head or less (like ssp. multicaulis). For thesepopulations and others where diagnostic characters conflict, themorphological characters from this study that generally differ betweenthe two lineages (corolla tube color, inflorescence type, stigmaticsurface, calyx pubescence, calyx sinus width, and degree of stigmaexsertion (Table 6) may beuseful for identification.

In mapping the distribution of the sun and shade races (Grant, 1954a) only a single population of theshade race, later designated G. achilleifolia ssp.multicaulis (Grant, 1956), wasdepicted in the southern half of the species range. Plants in thispopulation, CR, both in their small size, and for the morphologicalcharacters scored here, are similar to northern populations of G.achilleifolia (Table6). However, ITS sequence from the population is identicalto that of the much more robust G. achilleifolia ssp.achilleifolia population CN and the population shows higherisozyme genetic identity values with the neighboring G.achilleifolia ssp. achilleifolia population AB (I =0.791) and with CN (I = 0.712) than with G.achilleifolia ssp. multicaulis populations GR (I =0.653) or MH2 (I = 0.688). This evidence suggests that while thesouthern population CR resembles the northern populationsmorphologically, it differs somewhat genetically, implying that thenorthern and southern populations have distinct evolutionary histories. Grant recognized that this population could be interpreted as eithertaxon; his 1958 annotation of a 1954 collection from this siteidentifies the specimens as G. achilleifolia ssp.achilleifolia (V. Grant #9374 at RSA). MH and MH2 have sunrace morphology, and the latter has levels of isozyme polymorphismsuggesting outcrossing, yet by other measures, they appear to be a partof the G. achilleifolia ssp. multicaulis lineage. Though it would seem improbable, it appears that there are sun and shadeand outcrossing and selfing races of both the northern and southernlineages. This agrees with previous experimental evidence, whichsuggests that the ecological races were not equivalent in all respects,as the northern sun races were capable of greater autogamous seed setthan southern sun races (Grant and Grant,1965b).

Available evidence suggests that regardless of whether hybridspeciation was involved, the two subspecies of G. achilleifoliashould be considered separate species. The recognition of separatespecies in this case has the advantage of being predictive ofrelationship for the majority of gene genealogies (see Avise and Wollenberg, 1997). This follows theoriginal taxonomy (Bentham, 1833),which recognized two species, G. achilleifolia and G.multicaulis. Despite taxonomic arrangement, it should berecognized that the primary unit of morphological variation in thesetaxa is among individual local races (Grant1954a, 1963; V. Grant,University of Texas, personal communication). Giliaachilleifolia has been widely cited in literature on the evolutionof selfing (e.g., Stebbins, 1957b;Schemske and Lande, 1985; Richards, 1986; Wyatt,1988). It has been suggested that based on isozyme divergenceestimates (Schoen, 1982b) the taxa arebest regarded as separate species (Wyatt,1988). It is clearly inappropriate for future discussions ofthe taxa to assume that the selfing races of G. achilleifoliawere derived directly from outcrossing races of the samespecies.

Summary
Rigorous molecular and morphological analyses of G.achilleifolia and related species have failed to discriminatebetween the hypotheses of hybrid origin and divergent evolution. Thisappears to be due to the age of the hybrid speciation event (if itoccurred) and the subsequent evolution of both the hybrid taxon and itsparents. Presumably, there is a relatively narrow window of time inwhich hybrid species can be identified, and detection of ancient hybridtaxa may be intractable to even modern molecularmethods.

	FOOTNOTES

 
¹ The authors thank M. Debacon for technical assistance, N. Takebayashi for assistance with isozymes, field work, and helpful discussion; and D. Krause, D. Wolf, K. Kendrick, A. Day, and S. Markos for assistance in the field. J. M. Porter generously supplied several DNA sequences. Special thanks to V. Grant whose thoughtful input greatly improved the final manuscript. E. Friar, S. Spencer, D. Wolf, R. Morrell, D. Krause, and L. Johnson provided helpful comments on earlier drafts. This research was supported by grants from the Mellon Foundation (through the Rancho Santa Ana Botanic Garden), The Claremont Graduate School, Sigma Xi, The California Native Plant Society, and the Indiana Academy of Science.

	REFERENCES

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Allan G. J., C. Clark, and L. H.Rieseberg. 1997 Distribution of parental DNA markers inEncelia virginensis (Asteraceae: Heliantheae), a diploidspecies of putative hybrid origin. Plant Systematics andEvolution 205: 205–221. [CrossRef]

Anderson,E. 1949 Introgressive hybridization. John Wiley &Sons, New York, NY.

Arnold, M. L. 1993 Irisnelsonii (Iridaceae): origin and genetic composition of a homoploidhybrid species. American Journal of Botany 80:577–583.[CrossRef][ISI]

———., B. D. Bennett, andE. A. Zimmer. 1990 Natural hybridization betweenIris fulva and I. hexagona: pattern of ribosomal DNAvariation. Evolution 44: 1512–1521. [CrossRef][ISI]

———, C. M. Buckner, and J.J. Robinson. 1991 Pollen-mediated introgression andhybrid speciation in Louisiana irises. Proceedings of the NationalAcademy of Sciences, USA 88: 1398–1402. [Abstract/Free Full Text]

———, and S. A.Hodges. 1995 Are natural hybrids fit or unfitrelative to their parents? Trends in Ecology and Evolution 10:67–71.

Avise, J. C., and K.Wollenberg. 1997 Phylogenetics and the origin ofspecies. Proceedings of the National Academy of Sciences, USA94: 7748–7755. [Abstract/Free Full Text]

Baldwin, B.G. 1992 Phylogenetic utility of the internaltranscribed spacers of nuclear ribosomal DNA in plants: an example fromthe Compositae. Molecular Phylogenetics and Evolution 1:3–16.[CrossRef][Medline]

Bentham,G. 1833 Polemoniaceae. Botanical Register 19:folio 1622.

Crawford, D.J. 1990 Plant molecular systematics: macromolecularapproaches. John Wiley&Sons, New York, NY.

———, and R.Ornduff. 1989 Enzyme electrophoresis and evolutionaryrelationships among three species of Lasthenia (Asteraceae:Heliantheae). American Journal of Botany 76:289–296.[CrossRef][ISI]

Day, A.G. 1968 Gilia. In P. A. Munz and D.D. Keck, A California flora and supplement, 478–481. Universityof California Press, Berkeley, CA.

———. 1993 Gilia. In J. C. Hickman [ed.], The Jepson manual: higherplants of California, 828–836. University of California Press,Berkeley, CA.

Doyle, J. J., and J. L.Doyle. 1987 A rapid DNA isolation procedure for smallquantities of fresh leaf tissue. Phytochemical Bulletin 19:11–15.

Ellstrand, N. C., R. Whitkus, and L. H.Rieseberg. 1996 Distribution of spontaneous planthybrids. Proceedings of the National Academy of Sciences, USA93: 5090–5093. [Abstract/Free Full Text]

Felsenstein,J. 1985 Confidence limits of phylogenies: an approachusing the bootstrap. Evolution 39: 783–791. [CrossRef][ISI]

———. 1995 PHYLIP,phylogeny inference package, version 3.57c. University of Washington,Seattle, WA.

Gallez, G. P., and L. D.Gottlieb. 1982 Genetic evidence for the hybrid originof the diploid plant Stephanomeria diegensis. Evolution 36: 1158–1167. [CrossRef][ISI]

Gerassimova,H. 1939 Chromosome alterations as a factor ofdivergence of forms. I. New experimentally produced strains ofC. tectorum which are physiologically isolated fromthe original form owing to reciprocal translocation. Comptes Rendusde l'Academie des Sciences, Union des républiqussocialispes soviétiques 25: 148–154.

Grant, V. 1950 Genetic andtaxonomic studies in Gilia I. Gilia capitata. Aliso 2: 239–316.

———. 1952a Geneticand taxonomic studies in Gilia II. Gilia capitataabrotanifolia. Aliso 2: 361–373.

———. 1952b Geneticand taxonomic studies in Gilia III. The Giliatricolor complex. Aliso 2: 375–388.

———. 1953 The role ofhybridization in the evolution of the leafy-stemmed gilias. Evolution 7: 51–64. [CrossRef][ISI]

———. 1954a Geneticand taxonomic studies in Gilia IV. Giliaachilleaefolia. Aliso 3: 1–18.

———. 1954b Geneticand taxonomic studies in Gilia V. Gilia clivorum. Aliso 3: 19–34.

———. 1954c Geneticand taxonomic studies in Gilia VI. Interspecific relationshipsin the leafy-stemmed gilias. Aliso 3: 35–49.

———. 1956 Genetic andtaxonomic studies in Gilia X. Conspectus of the subgenusGilia. Aliso 3: 297–300.

———. 1958 Theregulation of recombination in plants. Cold Springs Harbor Symposiain Quantative Biology 23: 337–363.

———. 1959 Naturalhistory of the Phlox family. Martinus Nijhoff, The Hague.

———. 1963 The originof adaptations. Columbia University Press, New York, NY.

———. 1965 Specieshybrids and spontaneous amphiploids in the Gilia laciniatagroup. Heredity 20: 537–550. [ISI]

———. 1966a Geneticand taxonomic studies in Gilia XIII. The Gilialaciniata group. Aliso 6: 67–80.

———. 1966b Theorigin of a new species of Gilia in a hybridization experiment. Genetics 54: 1189–1199. [Free Full Text]

———. 1981 Plantspeciation. Columbia University Press, New York, NY.

———, and K.Grant. 1965 Flower pollination in the phlox family. Columbia University Press, New York, NY.

Harshman, J. 1994 Theeffect of irrelevant characters on bootstrap values. SystematicBiology 43: 419–424.

Heiser, C.B. 1947 Hybridization between the sunflower speciesHelianthus annuus and H. petiolaris. Evolution 1: 249–262. [CrossRef][ISI]

Holm, D. G., M. Fitz-Earle, and C. B.Sharp. 1980 Chromosome replacement in mixed populationsof compound-2L; free-2R and standard strains of Drosophilamelanogaster. Theoretical and Applied Genetics 57:247–255.[ISI]

Johnson, L.A., J. L. Schultz, D. E. Soltis,and P. S. Soltis. 1996 Monophyly and genericrelationships of Polemoniaceae based on matK sequences. American Journal of Botany 83: 1207–1224. [CrossRef][ISI]

———, and D. E.Soltis. 1995 Phylogenetic inference in Saxifragaceaesensu stricto and Gilia (Polemoniaceae) using matKsequences. Annals of the Missouri Botanical Garden 82:149–175.[CrossRef][ISI]

———, ———,and P. S. Soltis. 1994 Phylogenetic affinities inGilia section Kelloggia (Polemoniaceae): inferencesfrom chloroplast DNA matK sequences, nuclear ITS sequences, andmorphology. American Journal of Botany 81: 163(Abstract)

Kellogg, E. A., R. Appels, and R. J.Mason-Gamer. 1996 When genes tell different stories:the diploid genera of Triticeae (Gramineae). Systematic Botany21: 1–27.

Kim, K. -J., and R. K.Jansen. 1994 Comparisons of phylogenetic hypothesesamong different data sets in dwarf dandelions (Krigia,Asteraceae): additional information from internal transcribed spacersequences of nuclear ribosomal DNA. Plant Systematics andEvolution 190: 157–185. [CrossRef]

Lewis, P., and R.Whitkus. 1989 GENESTAT for microcomputers,(Genestat-PC), version 3.31. ASPT Newsletter 2:15–16.

Maddison, D. R. 1991 Thediscovery and importance of multiple islands of most-parsimonious trees. Systematic Zoology 40: 315–328. [CrossRef]

Maddison, W. P., and D. R.Maddison. 1992 MacClade, version 3.01. Sinauer,Sunderland, MA.

Marticorena,C. 1961 Morfología de los granos de polen de lasPolemoniaceae Chilenas. Gayana 2: 5–22.

Mason, H. L., and A. D.Grant. 1948 Some problems in the genus Gilia. Madroño 9: 201–220.

McCarthy, E. M., M. A. Asmussen, and W. W.Anderson. 1995 A theoretical assessment ofrecombinational speciation. Heredity 74:502–509.[ISI]

McDade, L. A. 1990 Hybridsand phylogenetic systematics I. Patterns of character expression inhybrids and their implications for cladistic analysis. Evolution 44: 1685–1700.