The reticulate nature of angiosperm evolution has been acknowledged from numerous studies supporting organelle capture (22). However, of the estimated 300,000 species of angiosperms, there are <10 clearly documented diploid or homoploid hybrid species based on molecular data (3–5, 6, 9, 18–19, 21). To understand the process of evolution, one first needs the tools to detect the patterns. In the case of reticulate evolution, the necessary tools include morphological, cytological, and molecular characters. Many hypotheses of diploid hybrid speciation are based on morphological and cytological characters (8), but support for this mechanism of evolution has been lacking due to the absence of molecular markers that can clearly distinguish closely related parental species.

The PCR was instrumental in the development of new molecular markers for population studies (23). New methods for generating molecular markers based on PCR include random amplified polymorphic DNA (RAPD), simple sequence repeat (SSR), amplified fragment length polymorphism, and inter-simple sequence repeat (ISSR) techniques. Each of these methods has many advantages as well as limitations (23). For example, RAPD markers are relatively easy to generate but may not be variable enough for some applications or may have problems with reproducibility; SSR markers are extremely powerful for estimating genetic diversity but require a lot of effort for primer development; amplified fragment length polymorphism produces many variable bands but usually requires the purchase of patented kits and specialized equipment; ISSR markers are extremely variable and have proven to be sensitive enough to differentiate cultivars (24–34) but haven’t been tested in natural populations.

In this paper, we test the utility of ISSR markers in natural populations by assaying the variability of ISSR banding patterns in a known hybrid complex of Penstemon (35–37) where other molecular data were available for comparison (38–40). We demonstrate that the hypervariability of ISSR markers revealed in studies of cultivar differentiation (24–34; reviewed in ref. 23) also occurs in natural populations, and these markers are sensitive enough to thoroughly test hypotheses of hybrid speciation and introgression in contrast to previous studies (38–40).

Background on Hybrid Complex. Straw (35–37) proposed two classic examples of diploid hybrid speciation in Penstemon (Fig. 1). These purported cases of hybrid speciation were hypothesized to have resulted from ethological isolation (35–37), where different classes of pollinators were adapted to different floral morphologies, or as a result of ecological isolation (Fig. 1). Over a series of three studies based on allozymes and restriction-site variation of nuclear rDNA and chloroplast DNA, Wolfe and Elisens (38–40) found numerous markers that separated bird-pollinated Penstemon centranthifolius (sect. Gentianoides) from insect-pollinated species (sect. Spectabiles). However, within sect. Spectabiles, there were no allozyme markers and few rDNA and chloroplast DNA (cpDNA) markers to distinguish among species. Penstemon spectabilis (purported hybrid derivative I; Fig. 1) lacked additive patterns of rDNA markers of Penstemon grinnellii and P. centranthifolius, but most populations had some individuals with molecular markers from P. centranthifolius. The overall pattern observed did not support a hypothesis of hybrid speciation but was more consistent with introgression of genes from P. centranthifolius into P. spectabilis. In the second hybrid speciation scenario (purported hybrid derivative II; Fig. 1), P. spectabilis and Penstemon clevelandii shared most alleles, had identical rDNA profiles, and had only 2–4 cpDNA restriction-site differences. P. clevelandii had numerous unique alleles, and two of five populations surveyed also shared molecular markers with P. centranthifolius but not in a pattern that supported the hypothesis of hybrid speciation. Instead, the overall pattern supported pollen-mediated gene flow from P. centranthifolius into P. clevelandii and a sister-species relationship between P. clevelandii and P. spectabilis. Because the few nuclear markers inadequately differentiated the insect-pollinated species of sect. Spectabiles, the hypotheses of hybrid speciation could neither be negated nor supported for P. spectabilis or P. clevelandii (38–40).

Figure 1 Schematic of diploid hybrid speciation hypotheses in Penstemon. Gray arrows represent directionality of hybrids as proposed by Straw (35–37). Length of the black arrows represent relative degree of gene flow between species supported by four molecular (more ...)

Background on ISSR Markers. Simple sequence repeats (microsatellites) are 1–6-nucleotide tandem repeats scattered throughout the genomes of most organisms assayed to date (reviewed in ref. 23). ISSR primers are designed from dinucleotide or trinucleotide simple sequence repeats and are anchored on either the 5′- or 3′-end of the microsatellite with 1–3 nucleotides (23, 25, 27). Where primer matches are found on opposite strands of the DNA molecule within an amplifiable distance, a discrete PCR product results, which can be assayed using nonradioactive or radioactive labeling techniques. ISSR markers are generated by single-primer PCR methods similar to RAPD techniques but have the advantage of exhibiting higher degrees of polymorphism compared with most RAPD markers (23, 28, 32, 33). Each amplicon represents a genetic locus with alleles present or absent. The markers are inherited in a dominant or codominant Mendelian fashion but are scored as diallelic dominant markers for data analysis.

DNA Analysis. One-hundred twenty-two DNA accessions representing 39 populations of species in the hybrid complex were assayed for ISSR polymorphisms (Table 1). Each DNA accession represented one to four individuals and a subset of the DNAs previously assayed using markers from nuclear rDNA and cpDNA restriction-site variation (39, 40).

Table 1 Distribution and percent polymorphic ISSR bands in Penstemon hybridcomplex

PCR Amplification and Scoring of Bands. Eight ISSR primers were assayed in single-primer reactions (814.1: (CT)₈-TG; 7: (CT)₈-RG; 17899: (CA)₆-RG; 17898: (CA)₆-RY; AW3: (GT)₆-RG; 17901: (GT)₆-YR; 17902: (GT)₆-AY; M1: CAA-(GA)₅₎. Standard reaction conditions were 0.4 μM primer, 1× Taq polymerase buffer, 0.2 μM dNTPs, 0.25–0.5 unit Taq polymerase (GIBCO/BRL), 3 mM MgCl₂, and 0.5 μl DNA in a 25-μl volume. The thermocycle program was 1.5 min at 94°C; 35 × 40 s at 94°C, 45 s at 44°C or 45°C, 1.5 min at 72°C; 45 s at 94°C, 45 s at 44°C, 5 min at 72°C; 6°C soak. PCR reactions were characterized on 1.5% agarose gels in 1× TAE buffer (5 h at 80 mA or until bromphenol blue marker dye migrated 10 cm from origin). Gels were stained with ethidium bromide and were documented digitally using an Alpha Innotech imaging system (Alpha Innotech, San Leandro, CA). The digital image files were transferred in a TIFF format to a PowerMac 7500 and analyzed using the BioMax 1D image analysis software (Eastman Kodak). Fragment sizes were estimated based on 1-kb ladder size standards (GIBCO/BRL) according to the algorithm provided in the BioMax 1D software. Fragment sizes were used to assign loci for each primer. Bands were scored as diallelic for each assigned locus (1 = band present; 0 = band absent).

Data Analysis. We calculated the total number of bands and distribution of bands across taxa, number of polymorphic bands, number of fixed bands, and bands shared among taxa. A distance matrix was generated from the raw data matrix of 1s and 0s by using an unpublished computer program written by Vera Ford (University of California, Davis). Pairwise average similarity comparisons between groups of DNA accessions representing specific taxa were calculated.

A total of 270 bands was scored for the five taxa in the hybrid complex. Four bands were fixed across all populations of all taxa (percent polymorphic across five taxa = 98.5). Individual DNA accessions could be genotyped with one to three ISSR primers. Private and marker ISSR bands were found for each species of the Penstemon hybrid complex (Table 1) in contrast to previous studies using other molecular data (38–40). The percent polymorphic bands ranged 88–95 for each species and was 72% for the hybrid taxon, Penstemon × parishii (the F₁ hybrid between P. centranthifolius and P. spectabilis). Species-specific marker bands and private bands were present in all species, and P. × parishii had two private bands.

Several marker bands for P. centranthifolius were found in some populations of P. spectabilis, but only one non-introgressive band (found outside of sympatric and hybridizing populations) was shared exclusively between P. centranthifolius and P. spectabilis (Table 2) in agreement with patterns previously observed for allozyme and rDNA restriction-site data (38, 39). P. clevelandii exclusively shared 15 bands with P. centranthifolius, of which 7 were not congruent with a pattern of introgression; 18 bands were exclusive with P. spectabilis, with 7 apparently not due to introgression. Patterns for P. spectabilis and P. grinnellii were nearly identical in regard to bands exclusively shared with P. centranthifolius (Table 2, upper and lower diagonals), and the number of bands exclusively shared between pairs of taxa in sect. Spectabiles, other than P. clevelandii:P. spectabilis, were nearly identical.

Table 2 Patterns of markerbands

Average similarity values (Table 3) based on band sharing among populations reveal that P. spectabilis and P. clevelandii are more similar to one another than P. spectabilis is to P. × parishii, and P. clevelandii is more similar to P. × parishii than P. × parishii is to its other parental species, P. centranthifolius. The average similarity between P. × parishii and P. clevelandii is nearly identical as the average similarity of individuals within P. × parishii.

Table 3 Matrix of average similarity based on bands present that are shared between populations oftaxa

If P. spectabilis were a recent hybrid-derivative of P. centranthifolius and P. grinnellii, one would expect an additive profile of parental bands in most individuals in all populations of P. spectabilis; an ancient hybrid derivative species would likely exhibit additive profiles among individuals in most populations, and a pattern of ancient introgression would show “foreign” markers present among a few individuals in some populations (40). However, populations of P. spectabilis lacked an additive profile of bands of P. centranthifolius and P. grinnellii, exhibited previously observed patterns of introgression of genetic markers from P. centranthifolius, and had evidence of low levels of gene flow between the two insect-pollinated species (Fig. 1, Table 2). The combined evidence from allozyme, nuclear rDNA restriction-site variation, and ISSR markers negates the hypothesis of a diploid hybrid origin of P. spectabilis but is consistent with pollen-mediated gene flow as proposed by Wolfe and Elisens (40).

Penstemon clevelandii had an additive profile of species-specific ISSR markers and had the highest number of non-introgressive, exclusively shared marker bands with both of its purported parental species (Fig. 1, Table 2). In addition, the average similarity (46%) between P. clevelandii and P. × parishii (the natural hybrid between P. centranthifolius and P. spectabilis) was higher than P. × parishii was to either of its parental taxa (35–39%) and was nearly identical to the average similarity within P. × parishii (Table 3). The morphological similarities between P. clevelandii and P. × parishii include numerous floral and leaf characters (35, 39). P. × parishii is also commonly found wherever P. centranthifolius and P. spectabilis occur sympatrically in contrast to the rare occurrence of observed hybrids between the purported progenitors of P. spectabilis. Taken together, the propensity for hybridization between the parental taxa, morphological similarities between P. clevelandii and P. × parishii, observed ecological isolation between P. clevelandii and its purported parental species, and high genetic similarity between P. clevelandii and P. × parishii all support a diploid hybrid origin of P. clevelandii (Fig. 1). However, because P. clevelandii has many unique allozyme (38) and ISSR alleles as well as distinct chloroplast haplotypes (40), we suggest an ancient hybrid origin for this species (40).

This study demonstrates that sensitive molecular markers are needed to identify diploid hybrid plant species. Unlike allopolyploid speciation where parental taxa may be highly differentiated, all of the diploid hybrid plant species characterized with molecular data represent hybridization between closely related species, which may be difficult to resolve with standard markers. Although reticulate evolutionary patterns are common in angiosperms, studies using allozymes, restriction-site variation of nuclear and chloroplast genomes, ITS nucleotide sequencing, and RAPD markers have all suggested that diploid hybrid speciation is a rare event in evolution (5, 8, 15, 19, 22, 40). It is clear that numerous species-specific markers for purported parental species are required for documentation of diploid hybrid species. With closely related species, some molecular markers may not offer sufficient variability or consistency to differentiate the parental taxa. However, our results clearly show that in cases where conventional molecular markers fail to reveal patterns of diploid hybrid speciation, hypervariable nuclear markers such as ISSR bands may provide an attractive alternative for assessing whether diploid hybrid speciation is an evolutionary mechanism of importance in angiosperms. Documentation of the diploid or homoploid nature of hybrid species is the first step in understanding the underlying evolutionary mechanisms of hybrid speciation.

We thank D. Crawford and L. Leege for comments on the manuscript and E. Esselman for helpful discussion. A.D.W. especially thanks R. M. Straw for his contributions to the biosystematics of Penstemon and his support and enthusiasm for the critical assessment of his wonderful hypotheses. The National Science Foundation (DEB 9123080; DEB 9632675) and The Ohio State University provided funding in support of this research.