Natural selection in Ipomopsis hybrid zones: implications for ecological speciation


Author for correspondence: Diane R. Campbell Tel: +1 949 8242242 Fax: +1 949 8242181 Email:


The fitness of hybrids relative to parental species plays an important role in models of speciation. In ecological speciation, reproductive isolation evolves owing to divergent natural selection, which implies reduced fitness of hybrid phenotypes. The source of the divergent selection in flowering plants may be animal pollinators, or environmental features of habitats that lead to physiological adaptations. Reciprocal transplants, combined with examination of pollination and other fitness components, allow exploration of these mechanisms of speciation. Much of this information is available from hybrid zones between Ipomopsis aggregata and Ipomopsis tenuituba. Pollination studies reveal some disruptive selection on corolla width, but also geographical variation in pollinator preference, and hybrids do not generally suffer lower pollination. Survival of hybrids depends on both genotype and environment. One genotypic class of hybrids is as fit or more fit than the parents, while another type suffers reduced fitness in parental environments. The dynamics of this hybrid zone involve a complex mixture of selection mediated by pollinators and other sources, and this combination of selection may have contributed to the original speciation.


Natural hybridization is common in flowering plants, with one recent survey concluding that hybrids comprise 6–22% of all angiosperm species (Ellstrand et al., 1996). The fitness of those hybrids compared with the parental species has been of considerable interest to evolutionary biologists, who have presented several alternative models to explain genetic differentiation across hybrid zones (reviewed in Arnold, 1997). The fitness of hybrids can also inform us about mechanisms of speciation. In ecological speciation, reproductive isolation evolves as a consequence of divergent natural selection, which implies reduced fitness of hybrid or intermediate phenotypes (Schluter, 2000). Here, I develop some relationships between studies of hybrid zones and particular mechanisms of ecological speciation in flowering plants. I then review the empirical evidence, with special emphasis on the Ipomopsis aggregata (Polemoniaceae) group. The genus Ipomopsis figured prominently in Verne Grant's early work on plant speciation (Grant & Grant, 1965) and has continued to be the focus of many ecological and evolutionary studies on topics that include competition for pollination, gene flow and genetic structure, herbivory, hybridization and nectar robbing.

Ecological speciation in flowering plants

Speciation requires both divergence of phenotype and the development of reproductive isolation. In ecological speciation the initial divergence of phenotype results from divergent (if in allopatry) or disruptive (in sympatry) selection, and reproductive isolation then arises directly or indirectly as a consequence of that divergent selection. Many biologists have suggested that pollination is a critical ecological factor in speciation of flowering plants, and interactions with animal pollinators (or abiotic pollination) could certainly be the source of the divergent selection. This viewpoint was well-developed by Grant (1949). In his book with Karen Grant on flower pollination in the Polemoniaceae (Grant & Grant, 1965), he described the process: ‘When the specializations for different classes of pollinators approach or reach a stage of mutual exclusiveness, these differences contribute to the reproductive isolation between the species involved, and permit such species to evolve henceforth along genetically independent lines.’ If two types of flower morphologies specialize on different classes of pollinators, we would expect hybrids with intermediate morphology to receive few pollinator visits. In this scenario of pollinator-mediated divergent selection (sensuWaser & Campbell, 2003), hybrids are expected to have a low pollination success, and this low hybrid fitness itself represents a form of postzygotic reproductive isolation. Furthermore, because pollinators are agents of gene flow as well as selection, the same set of traits (genes) under divergent selection can produce prezygotic reproductive isolation as a byproduct (Waser & Campbell, 2003). If this is so, sympatric speciation would be relatively easy to achieve, as the usual difficulty in generating linkage disequilibrium between ecological traits and traits that control mating preferences would be circumvented.

Ecological speciation could also involve adaptation to components of different environments unrelated to animal pollinators, such as soil type (McNeilly & Antonovics, 1968) or presence of herbivores or pathogens. In this case of environment-mediated divergent selection, hybrids between two new species might have low fitness because of genomic incompatibilities (nuclear or cytonuclear epistatic effects of alleles at two loci or chromosomal rearrangements) involved in adaptation to different environments, or because hybrids are poorly adapted specifically to the parental environments (Schluter, 2000). In the former case hybrids would have low fitness in all environments whereas in the latter case hybrid fitness would be environment dependent. Hybrid fitness can also be environment-dependent if hybrids of particular genetic background have high survival in ‘hybrid’ habitats characterized by disturbance, leading to bounded hybrid superiority (Moore, 1977). In any of these mechanisms, prezygotic reproductive isolation could develop through reinforcement (if the speciation event was in allopatry) or through an association with loci that control mating (in sympatry; Dieckmann & Doebeli, 1999).

These mechanisms of ecological speciation generate predictions about hybrid fitness. These predictions apply in a strict sense to current speciation events. However, we can also gain insight by testing them in recently divergent species that can still form hybrids, with the caveat that current reproductive isolation may not necessarily reflect the initial reproductive isolation during speciation. Pollinator-mediated divergent selection predicts that hybrids will have low pollination success because of divergent or disruptive selection on floral traits. In this case, it will be of interest to test whether those same floral traits also control reproductive isolation. Environment-mediated divergent selection predicts that hybrids will have low survival, or reproduction unrelated to pollination, because of ecological selection. This low fitness could occur in all environments (because of genomic incompatibilities between changes favored in different environments) or only in parental environments. These predictions can be tested by forming hybrids of known genetic background, planting them into different environments, as in a reciprocal transplant experiment, and then examining to what extent fitness differences at each site arise from pollinator-mediated selection vs other forms.

Although the importance of reciprocal transplants in assessing hybrid fitness is now well-recognized (Rieseberg & Carney, 1998), we do not have a complete set of such information for any case of natural hybridization. However, in recent years several investigators have completed large pieces of such a study. In the next sections I emphasize the work on Ipomopsis, and then more briefly make comparisons with other genera.

Ipomopsis hybrid zones

The genus Ipomopsis was described by Grant with the most recent revision in Grant & Wilken (1986). They included three species in the Ipomopsis aggregata group: I. arizonica, I. aggregata and I. tenuituba. Ipomopsis arizonica appears not to hybridize with the other species in nature, due at least in part to a combination of mechanical isolation and poor pollen performance of I. arizonica on I. aggregata stigmas (Wolf et al., 2001). Because hybrid formation is rare, low fitness of hybrids appears to play little role, at present, in maintaining the species integrity of I. arizonica.

Ipomopsis aggregata and I. tenuituba each have several subspecies. They were considered semi-species by Grant & Wilken (1986) and are closely related according to a cpDNA phylogeny that indicates recent speciation, extensive hybridization, or both (Wolf et al., 1993). They differ most conspicuously in floral traits; I. aggregata has relatively short and wide corollas of red coloration in the subspecies aggregata, whereas I. tenuituba has pale pink/violet to white flowers with long narrow corolla tubes (Fig. 1). Grant & Grant (1965) described reproductive isolation between these species of a prezygotic sort that implied pollinator-mediated divergent selection: ‘Floral isolation exists between the related species Ipomopsis aggregata and I. tenuituba in the Rocky Mt. region, which are normally pollinated by hummingbirds and hawkmoths, respectively’. More recently Grant & Wilken (1988) described situations where reproductive isolation is incomplete, and hybrid zones are evident. These hybrid zones may represent areas of secondary contact between previously divergent taxa or sites of independent primary divergence (Wolf et al., 1997).

Figure 1.

Flowers of Ipomopsis: (a) Ipomopsis aggregata aggregata; (b) F1 hybrid; (c) Ipomopsis tenuituba tenuituba.

Since 1992, my colleagues and I have studied contact zones between the two species with a major goal of measuring hybrid fitness and its dependence upon the environment. This work builds upon our earlier studies of pollinator-mediated selection on flower traits in Ipomopsis aggregata. We have focused especially upon a hybrid zone that occurs along an elevational gradient at Poverty Gulch in Gunnison County, Colorado, USA (Campbell et al., 1997).

Plants of I. aggregata aggregata grow in the valley at elevations of 2900 m and below, plants of I. tenuituba tenuituba on steep slopes at elevations above 3100 m, with natural hybrids found on the dry talus slopes at intermediate elevations. Several floral traits, including the width and length of the corolla tube and flower corolla, show clinal variation along the 3 km stretch separating the parental populations (Campbell et al., 1997). The main pollinators of both species of Ipomopsis in these parental sites and the natural hybrid zone are hummingbirds, Selasphorus platycercus and S. rufus (Campbell et al., 1997). Although I. tenuituba has traits frequently associated with hawkmoth pollination, hawkmoths (Hyles lineata) have appeared in this area in only 2 yr since we began extensive observations in 1992. Occasional other visitors to both species have included swallowtail butterflies and solitary bees (Campbell et al., 2002a). At this contact site, flowering times of the two species overlap almost completely, so seasonal reproductive isolation is unimportant (Campbell et al., 2002b).

Using a reciprocal transplant approach, seed progeny in 20 full-sib families resulting from within species crosses and 12 families resulting from between species crosses (F1 hybrids) were planted as seed into sites where the parental species and hybrids naturally grow (Campbell & Waser, 2001). All types of crosses yielded similar numbers of seeds. Ipomopsis is convenient for such a study because there is no seed bank, allowing experimental seedlings to be identified, and because it is monocarpic it is possible to measure lifetime fitness by following plants until their first and only reproductive bout at a median age of 5 yr (Campbell, 1997). Hybrid survival over a 5-yr period depended strongly on the direction of cross (i.e. on which species served as the maternal parent). In both I. aggregata and I. tenuituba habitats, hybrids with I. aggregata as the mother (AT in Fig. 2) survived similarly or better than all other genotypic classes, while the reciprocal F1 hybrid (TA in Fig. 2) performed poorly. However, at a site where natural hybrids were found, both F1 hybrids had high survival. Lifetime fitness measured as number of seeds produced per seed planted for a given family showed similar patterns at the two higher elevation sites; at the I. tenuituba site only AT hybrids and I. tenuituba had nonzero fitness, while at the hybrid site all four types of plants survived and reproduced successfully (D. Campbell, unpubl. data). At the I. aggregata site, lifetime fitness of I. tenuituba was zero even though survival was relatively high and as a result only I. aggregata plants and AT hybrids had high fitness. This difference results in part from the usually poor pollination success of I. tenuituba (see below).

Figure 2.

Proportion surviving from seed to age 5 yr or flowering for parental and F1 hybrid offspring at the Ipomopsis tenuituba and Ipomopsis aggregata parental sites and a natural hybrid site. Error bars for proportion surviving are standard errors across 10–12 full-sib families. AA, aggregata × aggregata; AT, aggregata × tenuituba; TA, tenuituba × aggregata; TT, tenuituba × tenuituba. In each case, the first letter indicates the maternal parent (cross type, site, and cross type × site, all P < 0.05). Adapted from Campbell and Waser (2001).

In classic speciation models, the low fitness of hybrids occurs because substitution of an allele at one locus is detrimental in combination with an allele at a second locus (Dobzhansky–Muller incompatibility). These epistatic effects on fitness will not be seen until the F2 generation (Fenster et al., 1997). To test for such effects, a second experiment is underway, in which F2 and backcrosses of F1 to the parental species were planted as seed into two environments. After 4 yr, F2 hybrids still have survival that is statistically indistinguishable from that of the parents; there is no hybrid breakdown in early survival (D. Campbell, unpubl. data).

To distinguish between pollinator-mediated divergent selection and environment-mediated selection, it is necessary to determine the extent to which any low hybrid fitness is the result of poor pollination rather than poor survival or some aspect of reproduction unrelated to pollination. While data on pollination are unavailable from the transplant experiments, we have examined pollination success and rates of herbivory in separate studies. Campbell et al. (2002a) examined pollen transfer in experimental arrays of potted plants with equal numbers of the two parental species and hybrids. The main pollinators in these arrays were hummingbirds, which visited all three types of plants but with a preference for I. aggregata or hybrids. When natural hybrids were used, they received a number of pollinator visits and pollen grains (estimated with dyes) that was intermediate to that of the two parental species. When F1 and F2 hybrids were used, they had the highest pollen receipt of all types of plant. Simulation models show that these patterns can be explained largely by observed patterns of pollinator visitation (ethological isolation), with mechanical isolation playing a smaller role (Campbell et al., 2002a). Nevertheless, one contribution to the high pollen receipt of hybrids is that hummingbirds are most efficient, on a per visit basis, at depositing pollen onto the slightly inserted stigmas of hybrid flowers (Campbell et al., 1998).

In most years, hummingbirds are the main pollinators throughout this natural hybrid zone. In that situation hybrids do not suffer from low pollination, and selection on floral traits is largely directional, favoring traits that are characteristic of I. aggregata and include wide corollas, high nectar production, and red color (Fig. 3a) (Melendez-Ackerman et al., 1997). All these traits enhance visitation rate by hummingbirds, and wide corollas also allow birds to enter more deeply into the flower and remove more pollen (Grant & Temeles, 1992; Campbell et al., 1996). In rare years, hawkmoths are also present, and then some floral traits may experience disruptive selection. In one such year, the per flower visit rate at a natural hybrid site was lowest to plants with an intermediate corolla width, as hummingbirds preferred to visit wide flowers and hawkmoths preferred to visit narrow flowers (Fig. 3b). Disruptive selection on flower color was not observed; while hummingbirds preferred to visit red flowers (Melendez-Ackerman & Campbell, 1998), hawkmoths showed no preference. Hawkmoth preference for pale flowers has been observed in other studies of Ipomopsis (Elam & Linhart, 1988). The lack of color preference at Poverty Gulch might result from the diurnal behavior of hawkmoths at these high elevation sites that cool rapidly at dusk. Another contact site along the Black Canyon of the Gunnison in Colorado is lower in elevation and warmer, and there the same species of hawkmoth flies at dusk and forages only on I. tenuituba (G. Aldridge, pers. comm.). Interestingly, the two species appear to hybridize little at this site. Thus, under certain ecological circumstances, the presence of both hummingbirds and hawkmoths may generate the disruptive selection required for pollinator mediated speciation. However, at Poverty Gulch, because of the usual predominance of hummingbirds, only rare years could fit this model of ecological speciation.

Figure 3.

Rates of pollinator visitation (visits per flower per hour) as a function of mean corolla width for the plant. (a) Directional selection in experimental arrays of potted Ipomopsis aggregata, Ipomopsis tenuituba and natural hybrid plants visited by hummingbirds only. Multiple regression detected significant selection on corolla width holding corolla length and flower color constant (P < 0.05). (b) Disruptive selection in natural hybrid populations visited by hummingbirds and hawkmoths. The surface is the best fitting quadratic regression of visitation rate (standardized relative to the mean) on corolla width and length, holding flower color (optical density) equal to the mean for natural hybrids (quadratic term P < 0.05). Adapted from Campbell et al. (1997).

It is not yet clear if the disruptive selection on corolla width is directly associated with reproductive isolation owing to ethological or mechanical isolation. We do know that the difference in flower color between the two species is associated with some reproductive isolation despite the absence of disruptive selection on this trait. When flowers in experimental arrays containing both species were painted red to match those of I. aggregata, a lower percentage of seeds were sired by conspecific pollen than in similar experimental arrays with unmanipulated flowers retaining the natural color difference (Fig. 4).

Figure 4.

Number of seeds sired per flower for various combinations of male and female parent within experimental arrays of Ipomopsis aggregata, Ipomopsis tenuituba and natural hybrid plants (male parents: filled columns, I. aggregata; open columns, I. tenuituba; hatched columns, hybrid). (a) Flowers on all plants in this array were painted red. (b) Flowers in this array were unmanipulated. With the natural color difference between species retained, there were many more seeds sired by I. aggregata on I. aggregata, and the percentage of seeds with conspecific parents thereby increased. Adapted from Melendez-Ackerman & Campbell (1998).

In the reciprocal transplant study, at least at the site where I aggregata normally grows, plants of that species and hybrids had higher total seed production than I. tenuituba. This difference in reproductive success is consistent with the patterns in pollination success. However, many other factors can also influence reproduction, including herbivory. In this hybrid zone, seeds of Ipomopsis are often consumed before dispersal by an Anthomyiid fly. Hybrids generally have resistance to the seed predators that falls between the levels experienced by the two parental species (Campbell et al., 2002b); however, the seed predators appear to have relatively small influence on lifetime hybrid fitness. Similarly, I. aggregata formosissima, I. tenuituba latiloba, and their natural hybrids respond indistinguishably to herbivory by caterpillars and ungulates (Anderson & Paige, 2003).

Unraveling the causes of the differences in survival is a daunting challenge. We are in the process of exploring differences in survival that correlate with aspects of physiology. The I. tenuituba site is drier than the I. aggregata site. The natural hybrid site is even drier, and temperatures near the ground fluctuate strongly as the rocks heat up in the intense sunlight, suggesting the hypothesis that natural hybrids, and perhaps TA hybrids, survive particularly well there because of higher water-use efficiency. We are currently testing this idea by measuring photosynthetic rate and water-use efficiency of plants in our reciprocal transplants.

All of the data so far are most consistent with a speciation model in which hybrid fitness is low only in parental environments, and even then only for hybrids of one cytoplasmic background. These differences in fitness result from differences in survival as well as pollination success, suggesting that a mixture of pollinator-mediated selection and environment-mediated divergent selection might have contributed to speciation, or at least to maintaining the species differences. While pollinator-mediated selection is reasonably well understood in this system, we do not yet know the reasons for the patterns in survival. Moreover the story is more complex than the usual scenario of ecological speciation in that the cause of low hybrid fitness in the F1 generation is not only environment-dependent, it appears to be strong only on the tenuituba cytoplasmic background. This implies cytonuclear interactions, which are environment dependent. We are currently testing for these using cytoplasmic markers inherited through the mother and nuclear DNA markers. Our studies to date make the prediction that, if the hybrid zone is stable, natural hybrids growing near parental sites will have I. aggregata cytoplasm whereas those growing near the center of the hybrid zone could have I. tenuituba cytoplasm. Moreover, there should be interactions between particular cpDNA and nuclear markers on survival, and, if physiology is involved, on water-use efficiency as well.

For these two species of Ipomopsis, reproductive isolation arises not only from prezygotic isolation due to preferences of pollinators and postzygotic effects on hybrid fitness, already discussed, but also a more subtle mechanism of postzygotic reproductive isolation. While F1 and F2 hybrids produce pollen in similar amounts and viability to the two parental species, the hybrid pollen is at a competitive disadvantage when it arrives in mixture with conspecific pollen on a stigma (Campbell et al., 2003). This happens even though heterospecific pollen suffers no such competitive disadvantage (Alarcon & Campbell, 2000), as is common in some other genera, including Iris (Carney et al., 1996).

As originally argued by Grant & Grant (1965), the floral traits separating these two species of Ipomopsis are subject to differential selection by pollinators. However, the two species are not neatly adapted to hummingbird and hawkmoth pollination. Hummingbirds do visit I. aggregata preferentially, leading to selection largely favoring traits associated with the classic hummingbird syndrome. However, surprisingly, flowers of that species are less effective at receiving pollen than are hybrid flowers. Moreover, at the contact site that is most intensively studied, flowers of I. tenuituba normally receive more visits from hummingbirds than from hawkmoths. In addition, ecological factors are clearly important to the differential ability of the two species and hybrids to survive at different elevations. Those environmental factors may interact with pollination in a way that helps to maintain the habitat separation of the species. For example, hummingbirds are more effective at transferring pollen from aggregata to tenuituba flowers (generating a TA hybrid) than in the reverse direction (Campbell et al., 1998), which might have helped to prevent the generation of large numbers of AT hybrids in the I. aggregata habitat, where they otherwise survive well. In summary, selection in this hybrid zone involves elements mediated both by pollinators and by environmental features, with the form of selection being more complex than that described by any one of the simple scenarios for ecological speciation.

Fitness of hybrids in other plant genera

In only a few other cases of natural hybridization do we have studies of hybrid fitness under field conditions (Arnold & Hodges, 1995). Several studies compared the fate of seeds collected from parental and natural hybrid populations. Using this approach, Levin and Schmidt (1985) found that natural hybrids between two subspecies of Phlox drummondii, also in the Polemoniaceae, had fitness indistinguishable from that of the two parental species. In a study of two subspecies of sagebrush (Artemisia tridentata) also involving seeds collected from parental and natural hybrid sites, selection was environmentally dependent, with the most fit type in each planting, including the hybrid site, being the type that naturally grows there (Wang et al., 1997). A similar study in a Prunella hybrid zone also found evidence for a homesite advantage, but in this case no evidence for bounded hybrid superiority in the natural hybrid site (Fritsche & Kaltz, 2000). None of these cases involved planting hybrids of known genetic background. However, Emms & Arnold (1997) planted F1 and F2 rhizomes along with rhizomes of their parental species Iris fulva and Iris hexagona. The F1 hybrids survived and grew as well or better than the parental types in all habitats. Fitness of the F2 was highly variable, and under greenhouse conditions was influenced by positive and negative nuclear–nuclear and cytonuclear epistasis (Burke et al., 1998). In summary, these results support the view that hybrids can, in at least some circumstances, have fitness as high as the parents, but there is no clear pattern to when hybrid fitness is environment dependent.

In most of these cases we have little idea of the pollination or physiological mechanisms that might have influenced fitness of the hybrids. Indeed, neither pollination success nor physiological traits of natural hybrids are well studied (but see Schwarzback et al., 2001) The Louisiana irises are one exception in which reciprocal transplants were complemented by that sort of information. In experimental arrays, F1 hybrids between two species of Iris were visited more frequently by pollinators then the two parental species (Wesslingh & Arnold, 2000). In the greenhouse, backcrosses to I. fulva showed lower maximum photosynthetic rate than the parental types, perhaps contributing to lower total biomass across all watering treatments (Johnston et al., 2001). By contrast, backcrosses to I. brevicaulis outperformed I. fulva. So, as in Ipomopsis, selection on hybrids appears to be mediated both by pollinators and by environmental factors.

Future directions

Reproductive isolation is fundamental to speciation. This reproductive isolation can be prezygotic or postzygotic, with the latter expressed as reduced hybrid fitness. Specific patterns to this reduced hybrid fitness may suggest particular models of ecological speciation, as illustrated above. While a number of recent studies have used reciprocal transplants to distinguish genomic incompatibilities from poor adaptation to parental environments as the cause of low hybrid fitness, we are still mostly ignorant about the extent to which the selection on hybrids is pollinator-mediated vs environmentally driven. More studies examining pollinator visitation rates to parental and hybrid plants, and the percentage of flights (and pollen movement) that are between species rather than within species are certainly needed. However, the physiological performance of hybrids under natural conditions is perhaps even more poorly understood. One approach to filling this gap would be to measure physiological traits (e.g. allocation to roots vs shoots, maximum photosynthetic rate and water-use efficiency) on hybrids of known genetic background. Such a study would be most informative if the hybrids were planted in a reciprocal transplant experiment. If pollinator visitation, survival and reproduction were measured on the same plants, it would be possible to see to what extent the fitness differences are explained by physiological adaptation vs pollination. Comparing the physiology of hybrids to parents allows one to test the hypothesis that selection on physiology differs across two parental habitats or between parental and hybrid habitats, as expected if speciation is due to environment-mediated divergent selection.

Another promising approach for the future is to combine the use of experimental gardens in the field with molecular methods to determine the genetic basis for different modes of prezygotic and postzygotic isolation. To date, most studies of genetic architecture have focused on plants growing in the greenhouse. Traits thought to be important to reproductive isolation are genetically mapped on greenhouse plants, as in studies of Helianthus (Kim & Rieseberg, 1999) and Mimulus (Bradshaw et al., 1998), and separate field studies can then be used to explore their role in reproductive isolation. As pointed out by Rieseberg et al. (1999), genetic mapping can be used in natural hybrid zones to test more directly for quantitative trait loci influencing hybrid fitness and the extent of reproductive isolation in the field. Their method requires a large number of species-diagnostic markers. That requirement can be muted by performing controlled crosses between individual plants (which will differ at more loci) and planting the F2 seeds into multiple habitats. One could then examine correlations of markers specific to the original parents with physiology, lifespan, seed production and visitation rates by specific pollinators. Such studies involving the measurement of hybrid fitness components in the field hold much promise for documenting mechanisms of ecological speciation in flowering plants.


The research on Ipomopsis hybrid zones was funded by National Science Foundation (NSF) grants DEB 9407144 and DEB 9806547. G. Aldridge, S. Kimball and C. Wu provided comments on the manuscript.