COMPONENTS OF REPRODUCTIVE ISOLATION BETWEEN NORTH AMERICAN PHEROMONE STRAINS OF THE EUROPEAN CORN BORER

Authors


Abstract

Of 12 potential reproductive isolating barriers between closely related Z- and E-pheromone strains of the European corn borer moth (Ostrinia nubilalis), seven significantly reduced gene flow but none were complete, suggesting that speciation in this lineage is a gradual process in which multiple barriers of intermediate strength accumulate. Estimation of the cumulative effect of all barriers resulted in nearly complete isolation (>99%), but geographic variation in seasonal isolation allowed as much as ∼10% gene flow. With the strongest barriers arising from mate-selection behavior or ecologically relevant traits, sexual and natural selection are the most likely evolutionary processes driving population divergence. A recent multilocus genealogical study corroborates the roles of selection and gene flow (Dopman et al. 2005), because introgression is supported at all loci besides Tpi, a sex-linked gene. Tpi reveals strains as exclusive groups, possesses signatures of selection, and is tightly linked to a QTL that contributes to seasonal isolation. With more than 98% of total cumulative isolation consisting of prezygotic barriers, Z and E strains of ECB join a growing list of taxa in which species boundaries are primarily maintained by the prevention of hybridization, possibly because premating barriers evolve during early stages of population divergence.

The biological species concept (BSC) posits that speciation is caused by reproductive isolating barriers that prevent gene flow between diverging populations (Dobzhansky 1937; Mayr 1942; Coyne and Orr 2004). Two methods are commonly used to investigate the causes of speciation. The comparative approach seeks to understand how isolating barriers evolve. In practice, this is achieved by measuring the strength of reproductive isolation for several barriers across a large group of taxa that vary in divergence time. This information can then be used to estimate the sequential order of isolating-barrier evolution (i.e., which evolved first, second, etc.) (e.g., Coyne and Orr 1989, 1997; Presgraves 2002; Mendelson 2003; Christianson et al. 2005), or to identify biological traits associated with enhanced diversification (e.g., Mitter et al. 1988; Arnqvist et al. 2000). The comparative approach, although ideal for issues related to evolutionary rate, suffers from at least one major limitation. Namely, speciation may often require the evolution of numerous and diverse barriers before being complete (Coyne and Orr 2004), but measuring isolation across a phylogeny for even a single barrier poses exceptional technical demands. Over 150 different Drosophila species were used in what is perhaps the best-known comparative study on speciation, but general conclusions about the initial causes of speciation in this group remain out of reach because only three barriers were considered (Coyne and Orr 1989, 1997).

An alternative approach for understanding the causes of speciation emphasizes the question that comparative studies commonly neglect. What are all of the habitat, morphological, behavioral, physiological, and other barriers that restrict gene flow between diverging populations? Rather than quantifying a small number of barriers across a large phylogeny, “case studies” of speciation take a comprehensive view of individual speciation events by evaluating numerous and diverse forms of reproductive isolation that could prevent a small number of closely related populations from fusing (McMillan et al. 1997; Ramsey et al. 2003; Husband and Sabara 2003; Hurt et al. 2005; Nosil et al. 2005; Martin and Willis 2007; Mendelson et al. 2007; Nosil 2007; Lowry et al. 2008; Schwander et al. 2008; Schluter 2009). Besides providing the only means for understanding the genetic and evolutionary mechanisms contributing to individual speciation events, case studies have the potential to confront many outstanding issues in speciation research.

One open question that case studies can address is how the strength of isolation varies among all of the individual components. There is a current debate about the role of one or a few strong barriers to gene flow versus many weaker ones (Nosil et al. 2005; Lowry et al. 2008; Schluter 2009). This distinction matters for at least two reasons. First, it relates to the ease in which speciation evolves. All else being equal, speciation between population pairs that are isolated by a few strong barriers will have a less complicated genetic architecture (e.g., in terms of number of loci) and evolutionary basis (e.g., in terms of genetic drift, natural or sexual selection, and sexual conflict), compared to speciation in which dozens of weaker barriers are involved. Second, if barrier strength typically increases with time, a barrier cannot have played a major historical role during speciation unless its contemporary strength is large (Coyne and Orr 2004). Testing the order of barrier evolution commonly requires a comparative approach because case studies lack phylogenetically independent contrasts (e.g., Coyne and Orr 1997; Mendelson 2003; but see Christianson et al. 2005). Nevertheless, under this assumption explicit hypotheses about the major historical causes of speciation can be formulated for case studies in which strong individual components are identified. The geographic mode of divergence bears on the debate about the role of a few strong or many weak barriers because gene flow can constrain divergence, although opportunities for reinforcement can only exist among nonallopatric populations.

A related problem to the strength of individual components of reproductive isolation is concerned with the relative contribution of those components to total cumulative reproductive isolation. Isolating barriers operate sequentially over organisms’ lifetimes and contribute to total reproductive isolation by preventing gene flow that has escaped earlier-acting barriers (Coyne and Orr 1989, 1997; Ramsey et al. 2003). Therefore, a barrier will disproportionately restrict gene flow if it operates at early life stages, before other barriers have had their effect. A barrier's relative or “sequential” strength is meaningful for understanding speciation because it informs a disagreement about the major contemporary causes for the maintenance of diverging populations, which is typically couched in terms of the relative importance of premating barriers (e.g., those that derive from behavioral, habitat, or temporal differences) versus postzygotic barriers (e.g., those that derive from hybrid inviability or infertility) (Ramsey et al. 2003; Husband and Sabara 2003; Nosil et al. 2005; Martin and Willis 2007; Mendelson et al. 2007; Nosil 2007; Lowry et al. 2008; Schwander et al. 2008; Schluter 2009). Because premating barriers operate first, they may often be the strongest impediments to gene flow even if postzygotic barriers are stronger when acting alone. One implication for the sequential nature of isolation and a speciation process in which diverse barriers accumulate is that the barriers that are most important for maintaining species boundaries may ultimately be early-acting forms (i.e., premating- or postmating-prezygotic [e.g., those stemming from differences in gametic compatibility or reproductive structures]), regardless of whether strong postzygotic barriers evolved first (Coyne and Orr 1989, 1997). However, the generality of this prediction hinges on the relative rates in which early and late-acting barriers evolve, an issue of substantial controversy (Coyne and Orr 1997; Kirkpatrick and Ravigne 2002; Mendelson 2003; Coyne and Orr 2004).

A third outstanding issue that case studies can address is the relationship between patterns of hybridization and gene flow in the field, and the biological traits that create these patterns. Quantifying the maintenance (or erosion) of genetic differences between natural populations has become relatively straightforward following the emergence of modern genetics tools and methods of analysis. However, a biological context for understanding these patterns is conspicuously absent for a majority of population and species pairs. In addition to establishing this linkage, estimates of genetic exchange can be calculated from cumulative reproductive isolation (e.g., Husband and Sabara 2003; Schwander et al. 2008), which are independent of those derived from genetic markers (e.g., Dopman et al. 2005; Malausa et al. 2007a). Finally, measures of total cumulative isolation can be brought to bear on taxonomic issues because the BSC defines species in terms of reproductive barriers. Between some taxa, biological species status may be warranted because the combined action of all isolating barriers reduces gene flow to nearly zero (e.g., Ramsey et al. 2003; Husband and Sabara 2003).

Here, we apply a case study approach to evaluate multiple forms of premating barriers, postmating-prezygotic barriers, and postzygotic barriers to gene exchange between North American Z- and E-pheromone strains of Ostrinia nubilalis, the European corn borer (ECB) moth. Pheromone strains of ECB are an exceptional model system to investigate speciation because multiple trait differences have been reported that could restrict gene flow, and because many of these differences have a clear genetic component. In sequential order of when they appear over the ECB life history, we quantified the strength of reproductive isolation for seasonal temporal isolation, circadian temporal isolation, male behavioral isolation, female behavioral isolation, mechanical isolation, gametic isolation (oviposition and fertilization), F1 hybrid inviability, behavioral hybrid (male) infertility, physiological hybrid infertility (oviposition and fertilization), and F2 and BC inviability. Our study has three main goals: (1) to confirm past (often anecdotal) data on reproductive barriers and formally quantify isolation strength; (2) to evaluate new potential reproductive barriers; and (3) to address the three outstanding issues in speciation research that case studies are well suited to confront.

Materials and Methods

THE STUDY SYSTEM

The ECB is native to Europe, North Africa, and Western Asia, but it is clear that three, and possibly more, introductions from Europe occurred into multiple eastern North American locations during the period 1909–1914 (Caffrey and Worthley 1927; Mutuura and Munroe 1970). With their recent introduction, most phenotypic and genetic variation in North American ECB will have originated in Europe, with subsequent evolution associated with a relatively “new” environment after introduction and expansion from Massachusetts, New York (NY), and Ontario, to most of the United States east of the Rockies. Population surveys reveal that at many sites in Europe and North America only Z strain males and females exist, that Z- and E-strain moths exist sympatrically at a number of sites, and that the E strain rarely exists alone (Klun and Cooperators 1975; Kochansky et al. 1975; Cardé et al. 1978; Anglade and Stockel 1984). ECBs often are found in or near agricultural crops, including maize (Zea mays) (Hudon and LeRoux 1986), but the species is polyphagous (>200 host species are known, review in Ponsard et al. 2004). Nevertheless, local host–plant specificity has been found in Europe (Bethenod et al. 2005; Malausa et al. 2008).

Z and E strains of ECB are an established model for behavioral isolation, in which two major genes determine trait differences in components of the pheromone communication system (female pheromone production and long-distance male orientation) that are responsible for behavioral isolation (Roelofs et al. 1987; Klun and Huettel 1988; Dopman et al. 2004). Populations in North America also show evidence for at least two forms of temporal isolation. In NY, differences in the number of generations per year (bivoltine or univoltine) can contribute to seasonal temporal isolation between sympatric populations. In this region, strains often differ in breeding cycle (i.e., univoltine Z [UZ] and bivoltine E [BE]) but the Z strain shows geographic variation such that BE or UZ populations can be found with bivoltine Z (BZ) moths (Roelofs et al. 1985; Glover et al. 1991). Although differences in life cycle occur in NY, across most of its range voltinism is broadly overlapping between strains, and the number of generations increases to—three to four in southern latitudes (Sorenson et al. 1992; Showers 1993). Breeding cycle differences can be a consequence of environmental variation, but inheritance studies suggest a strong sex-linked genetic component for a trait difference (postdiapause development time) that determines breeding cycle variation in NY (McLeod 1978; Reed et al. 1981; Glover et al. 1991, 1992; Dopman et al. 2005). A second potential form of temporal isolation was reported by Liebherr and Roelofs (1975), in which differences in the 24-h rhythm of mating is responsible for “circadian” temporal isolation between a UZ and a BE population. Common garden experiments were used by Liebherr and Roelofs, arguing for a strong genetic component for this difference.

Forms of postmating-prezygotic and postzygotic isolation have also been studied for North American populations of ECB. In addition to mating rhythm, Liebherr and Roelofs (1975) investigated gametic barriers and found that only one of three E × Z matings produced fertile eggs. Although this loss of fecundity was not significant (P > 0.5), their result suggests that a larger sample size might confirm the presence of genetic incompatibilities operating between copulation and zygote formation. As early as the 1940s hybrid inviability was assessed in the ECB (Arbuthnot 1944), but heterosis rather than inviability was discovered between a multigeneration and a single-generation population. Likewise, Liebherr and Roelofs (1975) found that of the few F1 crosses that were successful, F1-hybrid larvae were more than 1.5 times more likely to survive to adults compared to Z and E parentals. Finally, reduced viability of F2 and backcross offspring were reported by both Arbuthnot (1944) and Liebherr and Roelofs (1975). Although these studies support F1 hybrid vigor and second-generation “hybrid breakdown” as general phenomena, questions regarding strain identity or low cross replication leave uncertainty about this interpretation.

Although numerous reproductive barriers exist, debate about the taxonomic status of ECB strains has been fueled by their morphological and genetic similarities (e.g., Cardé et al. 1978; Dopman et al. 2005; Frolov et al. 2007; Malausa et al. 2007a, b), and by potential variation in reproductive barrier strength. Beyond trait differences that are associated with reproductive isolation, Z- and E-pheromone strains are difficult to distinguish (Liebherr 1974; Cardé et al. 1978), and genetic surveys suggest that gene flow is ongoing or recent. Of the many loci that have been screened, only a single locus reveals appreciable genetic differentiation (between BE and UZ/BZ moths in North America), the gene encoding Triose phosphate isomerase (Tpi) (Dopman et al. 2005 and references therein). In addition to life-cycle variation, populations show geographic differences in host–plant use. In North America, where our study focuses, strains have overlapping diets (e.g., Roelofs et al. 1985; Klun and Huettel 1988; Glover et al. 1991), but there is potential for host-associated reproductive isolation in France where strains use different hosts (Bethenod et al. 2005; Malausa et al. 2008). Differences in host use have prompted some to suggest that populations in France are in fact distinct species (Frolov et al. 2007). Adding to the taxonomic debate is the discovery of a similar pheromone communication system in the ECB's sister species, O. scapulalis (Kim et al. 1999; Huang et al. 2002) that casts doubt on the biological species status of this more geographically restricted relative (Frolov et al. 2007).

As our most comprehensive knowledge about reproductive isolation comes from North American populations of ECB, we focused our study using pheromone strains from this region. With the exception of seasonal and circadian temporal isolation, barriers were measured using a laboratory colony initiated from a BE population in Geneva, NY (hereafter BE colony), and a colony made from a UZ population in Bouckville, NY (hereafter UZ colony). Both colonies were initiated using ∼500 male and ∼500 female ECB collected as larvae, pupae, and adults, and were mass reared using ∼100 males and ∼100 females each generation (see Dopman et al. 2004 for further details). Seasonal isolation was studied using field populations in NY, whereas circadian isolation was measured using the results from Liebherr and Roelofs (1975). Because seasonal isolation was quantified under field conditions, environmental effects cannot be excluded. However, for all other potential barriers a common garden rearing scheme was used such that detected differences likely have a strong genetic component, and for some barriers major QTL have been previously detected.

PREMATING BARRIERS

Seasonal temporal isolation

In NY, ECB larvae diapause over the winter, then pupate and emerge as adults in the spring and summer. Differences in postdiapause development time, the time to pupation under temperature and photoperiod conditions conducive to breaking diapause, in large part determine number of generations per year (Glover et al. 1991). Glover et al. (1992) used the sex-linked marker Tpi to show that postdiapause development (hereafter PDD) time is affected primarily by a major gene (Pdd) or genes on the sex chromosome. Dopman et al. (2005) confirmed these findings and showed that Pdd is tightly linked with Tpi on the Z chromosome. Because the periods of flight of UZ and BZ or BE moths are displaced in time (Eckenrode et al. 1983), and because moths with different developmental programs may be incompatible, genetic differences affecting voltinism can have important consequences for the extent of gene flow in natural populations.

We estimated the consequences of genetic differences in PDD for seasonal temporal isolation by monitoring annual changes in abundance of Z- and E-strain moths in NY. Populations associated with sweet corn (Z. mays var. rugosa), a common host–plant species, were monitored over a 9-year period (1999–2007) at 41 different localities (Fig. 1), but not all sites were monitored across all years. The monitoring schedule was designed to sample both bivoltine populations (in June and August) and univoltine populations (in July) (Eckenrode et al. 1983; Roelofs et al. 1985), and typically began on the last week of May and ended on the last week of September (19 weeks). A passive sampling method was used in which adult male moths were captured in Scentry (Billings, MO) Heliothis traps that were baited with synthetic Z- or E-strain sex-pheromone lures (Trécé; Adair, OK), which were replaced every two weeks. At each location, one Z and one E trap were placed 40 m apart in habitat where adult ECB commonly aggregate (wild herbaceous vegetation, Showers et al. 1976). To maximize catches, traps were positioned near cornfields at “preferred” developmental stages (Spangler and Calvin 2000), but when corn reached the less attractive, dry brown-silk stage, traps were moved to neighboring cornfields that were at an earlier, more attractive developmental stage (i.e., late whorl or green silk). Secondary fields were within 1 km of the primary field, but most were less than 0.5 km away. Traps were checked once per week by a network of cooperators (http://www.nysipm.cornell.edu/scouting/scnetwork/). To estimate the effect of breeding cycle on temporal isolation between strains, we subsampled the complete dataset to identify those sites with E and Z populations in sympatry (defined as >100 Z-strain and >100 E-strain moths). Additionally, only years with at least 10 consecutive weekly measurements were included. Kolmogorov–Smirnov tests were used to determine whether strains showed significant differences in breeding cycle.

Figure 1.

New York localities where local abundance of Z- and E-strain moths was monitored. White place marks indicate sympatric localities where >100 E and >100 Z ECB males were captured in at least one year (see Table 1). Image from Google Earth (http://earth.google.com/).

For our purposes, we assume that male ECB collected using Z-pheromone lures were from the Z-pheromone strain and those collected using the E lure were from the E strain. As a result, all hybrids and those pure-strain moths that flew to the opposite pheromone blend were misidentified. However, in the laboratory, only a small fraction of E-strain males complete upwind long-distance flight to the Z-pheromone blend, and no Z-strain males fly to the E-pheromone source (E male flights to Z pheromone: ∼1%, n= 110 males; Z male flights to E pheromone: 0%, n= 112 males) (Roelofs et al. 1987; Glover et al. 1990; Linn et al. 1997). Considering these results, few pure-strain moths will be misidentified. Hybrids do show partial response to the Z-pheromone blend but the number of hybrid moths in NY is relatively low (mean = 5.9%, n= 2 mixed localities) (Roelofs et al. 1985). Because F1 hybrids show intermediate PDD times (Glover et al. 1992), any hybrids that do fly to Z- or E-blend pheromone traps would make estimates of seasonal temporal isolation conservative.

Circadian temporal isolation

We turned to investigate how differences in the 24-h rhythm of mating affects gene flow (via circadian temporal isolation) by revisiting the data from Liebherr and Roelofs (1975), in which mating periodicity was investigated for a UZ strain from London, Ontario, Canada and a BE strain from Aurora, NY. Moths were reared under a common environment (24°C and 65–70% r.h.), and patterns of mating were observed under a 16:8 hour L:D (light:dark) photoperiod to simulate day length for months in which active reproduction occurs (May and June). (Day lengths for May and June are roughly equivalent at Aurora, NY and London, Ontario.) We used these data to calculate an index of isolation.

Male and female behavioral isolation

Using a specialized pheromone gland, female ECB emit a specific pheromone blend that is used by males with the “appropriate” behavioral response for long-distance orientation and navigation toward the signaling female. In the Z strain, females produce and males preferentially orient to a 3:97 mixture of (E) and (Z)-11-tetradecenyl acetates (Δ11–14:OAc), whereas in the E strain, females produce and males orient to a 99:1 (E)/(Z) blend (e.g., Roelofs et al. 1987; Glover et al. 1990; Linn et al. 1997; Dopman et al. 2004). Hybrid females produce an isomeric mixture that is intermediate in composition (∼65:35 E:Z) and F1 males orient to all sources except the E source. Pheromone production is controlled by a single autosomal locus with two main alleles (Klun and Maini 1979; Roelofs et al. 1987; Zhu et al. 1996; Dopman et al. 2004), whereas upwind orientation to sex pheromone stimuli in a 1–2 m flight tunnel is controlled by a single sex-linked locus (Roelofs et al. 1987; Glover et al. 1990; Dopman et al. 2004). Although there are clearly two sexual communication systems with limited cross attraction in the laboratory, a reduced frequency of hybrids can be found in nature (e.g., Roelofs et al. 1985; Klun and Huettel 1988; Glover et al. 1991). As differences in sexual communication can lead to assortative mating or in reduced ability of hybrids to obtain mates (via behavioral “dysfunction”), male orientation to female pheromones has important implications for patterns of gene flow in the field. We estimated the effect of long-distance (∼2 m) male behavioral isolation on gene flow by using the data from Roelofs et al. (1987), Glover et al. (1990), and the 30 ug experiment in Linn et al. (1997). These studies used the same UZ and BE colonies that we use here to measure long-distance orientation response for Z, E, and hybrid males to a range of sex pheromone sources.

As ECBs are known to aggregate in the field (Showers et al. 1976; Malausa et al. 2005), measures of long-distance behavioral isolation may not accurately reflect natural conditions. To assess how mating behavior is affected by spatial scale, we made observations under shorter distances (<20 cm) using the UZ colony and the BE colony. Initial no-choice observations were made of E female × E male (E × E) and Z female × Z male (Z × Z) pairings to characterize major elements of the mating behavioral sequence. Following these observations, we quantified the success rate of each behavioral element in no-choice experiments using the four possible types of Z and E mating pairs (E × E, E × Z, Z × E, and Z × Z). To simulate mating patterns that result when ECB aggregate, each trial consisted of a male released into a one-pint ventilated, disposable, and clear plastic tub that contained a resident female and a water source. One-day-old virgin males and females were used in all experiments. Moths were used only once. Humidity (65 ± 10%) and temperature (22–24°C) were controlled.

Moth behavior was monitored following Liebherr and Roelofs (1975), in which a light source (Petzl headlamp) was filtered with 2 thicknesses of number 92 Kodak Wratten filter and 2 thickness of typing paper. Continuous observations were made from 5.5–6.5 h of the dark cycle or until mating, which we defined as a state of genital clasping that lasted more than 20 min. Although both strains were reported to mate during the observation period (Liebherr and Roelofs 1975), we tested each moth for mating motivation. Female motivation was assessed by observing the exposure of the sex-pheromone gland, which is consistent with release of sex pheromone (Webster and Cardé 1982). Males that were actively flying before the experiment were assumed to be adequately motivated. Indices of the strength of reproductive isolation were calculated for behavioral elements that significantly differed between experimental groups, as defined using contingency-table tests. For these tests, we distinguish between the reciprocal interstrain crosses (i.e., Z × E and E × Z) by the sex of the moth expressing the observed behavior.

POSTMATING-PREZYGOTIC BARRIERS

Gametic isolation

To confirm the results of Liebherr and Roelofs (1975), in which one of three BE × UZ matings was successful, we estimated gametic isolation by measuring egg production and fertilization over the lifetime of female ECB. UZ and BE colonies were used for experiments in which single one-day-old females were paired for life with single one-day-old males under controlled laboratory conditions (see Behavioral Isolation). Unlike behavioral trials, mating was not observed. Females were allowed to lay eggs on wax paper until death. Spermatophores persist in females after death; therefore, we dissected females to infer the number of times that each female had mated. Fertilization was assessed by a change in egg color. Eggs that are unfertilized remain cream colored, whereas in fertilized eggs the developing larval head capsule visibly darkens. From replicates crosses, Mann–Whitney U-tests were used to detect significant differences in fitness components that could result in reproductive isolation. Interstrain crosses were compared to the within-strain cross that had the same maternal strain (i.e., Z × Z vs. Z × E and E × E vs. E × Z).

POSTZYGOTIC BARRIERS

F1 inviability

In response to uncertainty about the generality of heterosis in ECB hybrids (i.e., Arbuthnot 1944; Liebherr and Roelofs 1975), we evaluated F1 survival. We tested for heterosis (or inviability) in F1-hybrids by comparing survival of F1 and pure-strain offspring produced from the crosses that were used to characterize postmating-prezygotic isolation. Viability was quantified at the early zygote stage as the number of fertilized eggs that hatched. Data analysis followed that used for postmating-prezygotic isolation.

Behavioral and physiological infertility

In addition to generating behavioral isolation, mating behavior differences between populations can cause postzygotic isolation if behaviors in hybrids are either intermediate to parentals or if they are disrupted (review in Servedio and Noor 2003). By taking advantage of published data on long-distance male orientation, we were able to assess the importance of this potential reproductive barrier. In their laboratory experiment of long-distance flight orientation to synthetic pheromone blends, Roelofs et al. (1987) demonstrated that F1 males have the highest level of orientation to the intermediate, hybrid blend, but at an intensity that is reduced compared to that of pure males to their blends (Fig. 2A). Hybrid males may experience infertility due to failures or delays in mating compared to Z- or E-strain males because of these “dysfunctional” behaviors. However, an appreciable level of hybrid orientation to the Z pheromone (46%) suggests that relative fitness will depend on the frequency of females in the population (i.e., hybrid males would be expected to suffer most when hybrid and Z females are rare). We explored the consequences of behavioral hybrid dysfunction by first applying contingency-table tests to confirm the significance of differences among ECB males (Roelofs et al. 1987). We then calculated fitness for F1 and parental males as the total probability of male orientation to a female. Fitness was calculated using the estimated frequency of females from mixed sites in the field. We used values from Roelofs et al. (1985) for Geneva, NY, which has BE, UZ, and hybrid moths with the E strain predominating. We also used values from Klun and Huettel (1988) for Beltsville, MD, which has BE, BZ, and hybrid moths, but with the Z strain predominating. In both studies, diagnostic differences in female sex-pheromone blend were used to identify strains.

Figure 2.

(A) Proportion of parental and F1 males attracted to parental and hybrid sex-pheromone blends (n≥ 40 for each strain/pheromone blend combination, modified from Roelofs et al. (1987)). Reciprocal F1 males showed similar orientation and were combined (G-test, P > 0.5). (B) Estimated frequency of parental and hybrid female ECB at Geneva, NY (n= 46 moths, Roelofs et al. 1985) and Beltsville, MD (n= 78 moths, Klun and Huettel 1988). (C) Estimated relative male fitness at Geneva, NY and Beltsville, MD, suggesting that behavioral hybrid dysfunction occurs at some localities. Fitness was calculated as the total probability of male orientation to a female, given estimated female frequencies at each location.

A second form of hybrid infertility that we were interested in has a physiological basis. We measured lifetime oviposition and percent fertility in F1 hybrids. F1 moths (E/Z and Z/E) from UZ and BE colonies were paired with F1 or parental adults to generate six cross types: two maternal backcrosses (BC) (Z/Z × Z/E and E/E × E/Z), two paternal backcrosses (E/E × E/Z and E/Z × Z/Z), and two F2 intercrosses (E/Z × E/Z and Z/E × Z/E). Fitness differences between F1 and parental moths were assessed by comparing each second-generation cross to fecundity and fertility values from parental crosses. Data collection and analysis methods followed those described for postmating-prezygotic isolation.

F2 and BC inviability

We sought to confirm evidence for “hybrid breakdown” in second-generation ECB offspring (Arbuthnot 1944; Liebherr and Roelofs 1975), by measuring zygote viability. We used the approach described for measuring fitness of F1-hybrid offspring.

STRENGTH OF REPRODUCTIVE ISOLATION

We took the quantitative approach of Ramsey et al. (2003) and Coyne and Orr (1989, 1997) in calculating an index of the absolute strength (AS) of reproductive isolation for each individual barrier when it operates alone (“reproductive isolation” in Ramsey et al. 2003). AS for the nth component of reproductive isolation was generically defined from mean fitness components as:

image(1)

Values of AS typically ranged from 0, which indicated free gene flow, and 1, which indicated complete isolation. AS could also be negative, in which case hybridization and gene flow was favored. We restricted our analysis to isolating barriers that were statistically significant in at least one comparison.

Next, we calculated the sequential strength (SS) of each barrier (“absolute contribution” in Ramsey et al. 2003), total (T) or stage-specific cumulative reproductive isolation, and the relative contribution of each barrier to total isolation. Isolating barriers were ordered according to the life stage when they had their effect. For the nth barrier, SSn depends on ASn and the amount of gene flow that has escaped earlier-acting barriers. Hence, SSn was defined as:

image(2)

From these values, T (or stage-specific cumulative reproductive isolation) was calculated by summing SS across barriers (e.g., T =∑SSi for i= 1 to n). Finally, we measured the relative contribution of each barrier to T (i.e., SSn÷ T).

Of the barriers that were measured, seasonal temporal isolation and behavioral hybrid male dysfunction ought to show predictable differences in AS between Z and E population pairs. In NY, the strength of seasonal isolation will differ between Z and E populations because of genetic differences (in PDD) within the Z strain contribute to BZ or UZ breeding cycles (Glover et al. 1991). In contrast, male dysfunction will vary because of environmental dependence (stemming from differences in female frequency). We used several estimated values of AS for these barriers to explore the extent to which different Z and E populations in nature might be reproductively isolated.

Results

PREMATING BARRIERS

Seasonal temporal isolation

Across years, a total of 16,949 and 38,674 male ECB were trapped using E- and Z-pheromone lures, respectively. We found evidence for sympatry (>100 E and >100 Z moths) at 12 localities (Fig. 1). Nine localities showed evidence for sympatry across multiple years, resulting in a total sample size of 30. E and Z moths captured using E and Z lures differed significantly in time of seasonal occurrence across all years and locations (Kolmogorov–Smirnov test, P < 0.03).

We estimated the strength of seasonal temporal isolation based on the number of hybrid and pure-strain offspring expected after one generation of random mating. By assuming that male and female frequencies were the same, we generated this expectation using Hardy–Weinberg proportions. For each week i, the expected frequency of hybrid (2piqi) and pure-strain (p2i+q2i) offspring was calculated from the fraction of the total number moths (ni) that were of the Z (pi) strain and the E (qi) strain. The expected number of hybrid and pure-strain moths each week was then calculated by multiplying the number of moths (ni) by their expected frequencies. Summing across weeks (w) provided an estimate of the total number of hybrid and parental offspring produced in the season. In this manner, we estimated absolute-seasonal isolation (ASseasonal) as:

image(3)

To adjust for the effect of abundance asymmetry on the index (e.g., if one strain was rare the index indicated high isolation even if strains shared breeding cycles), we forced the total number of Z and E moths across the season to be equal. With this formulation, the index reflected differences in proportional abundance between strains across the season.

When breeding cycles were completely overlapping ASseasonal= 0 and when breeding cycles were completely separated ASseasonal= 1. High seasonal isolation (ASseasonal > 0.7) commonly reflected the presence of sympatric UZ and BE populations (e.g., Farmington in 2000) (Fig. 3A). Similarly, low isolation (ASseasonal < 0.4) often revealed sympatric BZ and BE populations or BZ, UZ, and BE moths in sympatry (e.g., Dresden in 2003) (Fig. 3B). The relationship between breeding-cycle and isolation, however, was not always clear and patterns were sometimes variable across years at the same locality (Table 1).

Figure 3.

Premating, temporal isolation. Examples of strong (A) and weak (B) seasonal temporal isolation. (A) In 2000, strains at Farmington showed strong seasonal isolation, with Z moths displaying a predominant univoltine breeding cycle and E moths displaying a predominant bivoltine breeding cycle. (B) Seasonal differences in male flight activity at Dresden, NY, in 2003, in which males of both strains showed overlapping (bivoltine) breeding cycle patterns and weak seasonal isolation. (C) Circadian differences in time of reproduction between Z- and E-strain ECB during scotophase (from Liebherr and Roelofs (1975)).

Table 1.  Strength of seasonal temporal isolation at sympatric New York sites.
LocationYearASseasonalWeeksTotal ETotal Z
Baldwinsville20000.44518268232
 20060.60116109144
Batavia20020.45114106443
Dresden20030.29814280179
Farmington20000.75119338344
 20020.62618215243
 20030.38817904275
 20040.85519326230
Hall20010.52019182189
 20020.55018337282
 20030.42017765530
 20040.43818830831
 20060.62519101111
LeRoy20000.53919158614
Lockport19990.50619182275
 20000.75418144797
Medina20000.46118261468
 20010.61118109284
Penn Yan20000.73119209343
 20040.70219130490
 20060.62419175824
 20070.66718225198
Rush19990.60319126112
 20000.68118135270
 20020.65818476541
Spencerport20000.71618109399
 20020.65815116195
Williamson20000.63619173345
 20010.28118171186
 20020.55116110250

Circadian temporal isolation

With respect to temporal isolation from differences in circadian time of mating (AScircadian), Liebherr and Roelofs (1975) showed that Z-strain moths mated, on average, 1.7 h earlier than E-strain moths (mean E = 6.8, n= 18; mean Z = 5.1, n= 24; t-test, P < 0.01; Fig. 3C). Using the data from Liebherr and Roelofs (1975), we estimated AScircadian with equation (3), except that the time interval over which isolation was based was hours rather than weeks. Without other barriers, the absolute strength of AScircadian was estimated to be 0.525.

Male and female behavioral isolation

Observations from a total of 119 no-choice mating experiments suggested that after the release of sex-pheromone by female ECB, successful mating under short distances consists of three additional elements: male orientation and courtship, female acceptance, and copulation. If a male is sexually receptive to the female, orientation and courtship is initiated. This stage consists of the male flying and then walking to the signaling female with modified abdominal scales (“hair pencils”) everted and wings fanning (e.g., Nakano et al. 2006). The male then positions himself perpendicular to the female and attempts copulation. Female acceptance of a courtship attempt may then lead to clasping of genitalia and mating.

Within-strain pairings indicated that mating failures could occur at any stage of the behavior sequence. The release of sex pheromone appeared to be required for initiation of the sequence as otherwise, males did court (n= 20). However, pheromone release did not always result in orientation and courtship (attempts = 90%, n= 41) and not all courtship attempts were accepted by females (acceptance = 86%, n= 37). Female rejection consisted of walking or flying away from a courting male or of abdominal bending such that her genitalia were inaccessible (Royer and McNeil 1992; Nakano et al. 2006). Finally, mating was suspended when male copulation attempts failed to result in clasping of genitalia (clasping = 78%, n= 32). Failures at the genital clasping stage might result from incompatibilities in reproductive structure rather than behavior (i.e., mechanical isolation). Therefore, we report on genital clasping as a potential form of postmating-prezygotic isolation below.

Compared to that seen within-strain, mating behaviors in interstrain pairings commonly failed. Males were likely to court females of the same strain, but not females of the other strain (Fisher's test; E × E [n= 21] vs. Z × E [n= 46], P < 0.01; Z × Z [n= 20] vs. E × Z [n= 32], P < 10−5; Fig. 4A). Similarly, females were likely to accept males of the same strain, but reject males of the other strain (Fisher's test; E × E [n= 19] vs. E × Z [n= 8], P < 0.02; Z × Z [n= 18] vs. Z × E [n= 23], P < 0.03; Fig. 4B).

Figure 4.

Premating, behavioral isolation. (A) Male orientation to live females over short distances (<20 cm). (B) Female acceptance of male courtship in short-distance trials. (C) Male orientation to synthetic pheromones over long distances (∼2 m) (pooled from Roelofs et al. (1987), Glover et al. (1990), and Linn et al. (1997)). Cross type designations list females first. Numbers in parentheses are sample sizes.

Behavioral isolation was calculated using equation (1). For male behavioral isolation (AS♂  orient.), we considered short-distance attraction (<20 cm, this study; Fig. 4A) and long-distance orientation (∼2 m, [pooled from Roelofs et al. (1987), Glover et al. (1990), and the 30 ug experiment in Linn et al. (1997)]; Fig. 4C). With short-distance attraction, AS♂  orient. was 0.448 for Z × E pairings and 0.722 for E × Z pairings (Fig. 4A). With long-distance attraction, AS♂  orient. was 0.988 for Z × E pairings and 1.0 for E × Z pairings (Fig. 4C). For female discrimination behavior, AS♀  discrim.= 0.712 in E × Z pairings and 0.298 in Z × E pairings (Fig. 4B).

POSTMATING-PREZYGOTIC BARRIERS

Mechanical isolation

In contrast to overt male and female behaviors, there were no significant differences in the proportion of successful clasping attempts among the four types of mating pairs (Fisher's test; E × E [n= 13], Z × E [n= 13], Z × Z [n= 12], E × Z [n= 2], P > 0.25; Fig. 5A). The limited sample size for E × Z pairings at this last stage of the mating sequence, however, prohibits a fully informative test of this possible mechanical form of postmating-prezygotic isolation.

Figure 5.

Postmating-prezygotic isolation. (A) Proportion of clasping successes in within and between strain pairings. (B) Mean oviposition and (C) mean percent fertility in within and between strain crosses. Cross-type designations list females first. Numbers in parentheses are sample sizes.

Gametic isolation

Gametic barriers were assessed by recording the fate of 8754 eggs from 40 unobserved crosses. Pairings that failed to produce offspring were excluded; thus, our analysis reflects a minimum strength of gametic isolation. Interstrain pairings failed to produce offspring more often (P < 0.001, proportion successful, Z × E = 0.34, Z × E = 0.27 vs. 0.71 for within-strain pairings), but this probably reflects behavioral incompatibilities rather than severe postmating-prezygotic problems (or complete F1 inviability, below). Females that laid live progeny mated only once, as indicated by the presence of only a single dissected spermatophore. Over their lifetime, Z females mated to Z males laid about 280 eggs and E females mated to E males laid about 248 eggs (U-test, P > 0.4). Our results confirm gametic isolation in the ECB, in that oviposition was reduced in an interstrain cross (Fig. 5B), but it is in the opposite direction to that seen by Liebherr and Roelofs (1975). Z females mated with E males lay on average 194 eggs, a significant reduction in lifetime productivity compared to when they mated with Z males (U-test, P < 0.03). E females paired with Z males were only slightly less fecund than when mated to E males (mean = 227.1; U-test, P > 0.4). Isolation was computed for lifetime oviposition differences using equation (1). The absolute strength of isolation (ASovipos.) was 0.309 for Z × E crosses and 0.084 for E × Z crosses. Unlike oviposition, percent fertilization did not significantly differ between within-strain crosses (Z strain = 90.2%, E strain = 98.5%, P > 0.05), or between these and interstrain crosses (Z × Z vs. Z × E, U-test, P > 0.2; E × E vs. E × Z, U-test, P > 0.15; Fig. 5C).

POSTZYGOTIC BARRIERS

F1 inviability

In contrast to previous studies (Arbuthnot 1944; Liebherr and Roelofs 1975), we did not find evidence for heterosis in hybrid ECB, nor did hybrids experience significantly reduced survival. Mean viability for E-strain and Z-strain zygotes was 96.8% and 94.2%, respectively (U-test, P > 0.27; Table 2). For the Z × E cross, mean viability was 95.5% (vs. Z × Z, U-test, P > 0.65), whereas mean viability was 92% in the Z × E cross (vs. E × E, U-test, P > 0.07).

Table 2.  Physiological forms of postzygotic isolation. Viability of F1 zygotes, fertility of F1 adults, and viability of backcross and F2 zygotes.
Cross type♀×♂nMean ovipositionPMean fertility (%)PMean viability (%)P
  1. 1Mean viability of Z×E versus Z×Z (mean=94.3%) and mean viability of E×Z versus E×E (mean=96.8%).

  2. 2Versus pooled parental values.

  3. 3Significant after Bonferroni correction (α=0.05÷6 second-generation crosses).

Parental (pooled)E×E and Z×Z20267.3N/A93.6N/A95.3N/A
F11Z×E10(Fig. 5) (Fig. 5) 95.50.677
 E×Z10(Fig. 5)(Fig. 5)92.00.070
Maternal backcross2Z×Z/E10173.20.008395.00.19897.80.082
 E×E/Z 5260.40.94092.31.00095.61.000
Paternal backcross2Z/E×E10190.00.09387.40.49798.10.010
 E/Z×Z10202.60.01297.20.35094.80.900
F22E/Z×E/Z10289.90.45196.40.32793.30.940
 Z/E×Z/E10231.90.24897.50.49898.30.152

Behavioral and physiological infertility

To test for behavioral dysfunction, we revisited the data from Roelofs et al. (1987) and confirmed that F1 males show inhibited orientation towards all pheromone blends compared to Z- or E-strain males to their sex-pheromone blends (G-test, P < 1 × 10−5, n≥ 40 for each strain/pheromone blend combination; Fig. 2A). Similarly, as was expected (e.g., Glover et al. 1991; Linn et al. 1997), differences in hybrid-male orientation towards the intermediate (hybrid) pheromone blend (56%) and the Z-pheromone blend (46%) were not significantly different (G-test, P > 0.2), but hybrids showed limited response to the E-pheromone blend (4%) (G-test, P < 1 × 10−8). Using orientation values from Roelofs et al. (1987) (Fig. 2A) and the estimated frequency of females from Roelofs et al. (1985) (Geneva, NY; n= 46 females) and Klun and Huettel (1988) (Beltsville, MD; n= 78 females) (Fig. 2B), we calculated relative fitness for F1 and parental males. As a result of their promiscuous, although less vigorous behavior, hybrid male fitness exceeded that of E males when hybrid and Z females were in the majority (e.g., Beltsville, MD, in Fig. 2C). In contrast, when E females were in the majority (e.g., Geneva, NY in Fig. 2C), F1 behavior conferred reduced mating success when compared to both parental males. Using these fitness values, we defined an index of isolation for behavioral hybrid dysfunction (ASF1♂  dys.) as:

image(4)

At the Geneva site, ASF1♂  dys. was 0.184 for the Z strain and 0.753 for the E strain. At Beltsville, isolation was positive for the Z strain (0.345), but negative for the E strain (−1.422), reflecting the superior fitness of F1 males at this location.

We tested for a physiological form of F1 infertility by measuring oviposition and fertility for 55 different second-generation crosses (six cross types), which yielded a total of 12,178 eggs. As in gametic isolation experiments, pairings that failed to produce any progeny were excluded and female dissections recovered a single spermatophore in each female from every successful cross. The exclusion of unsuccessful pairings could not have influenced measures of F1 infertility (or F2/BC inviability, below) because all the six second-generation crosses were as successful as within-strain crosses (P > 0.06). Mean oviposition of each second-generation cross was compared to pooled parental crosses (pooled parental cross mean = 267.3). Oviposition was significantly reduced in both backcrosses to the Z strain (Z × Z/E = 173.2; E/Z × Z = 202.6) (Table 2), but only the maternal backcross remained significant after Bonferroni correction (P < 0.008). When divided by cross class, only BC pairings were less productive than parentals (mean = 198.8, U-test, P < 0.01). Both parental and F2 intercrosses (mean = 260.9, U-test, P > 0.8) showed similar levels of productivity. Reproductive isolation was estimated using equation (4), with F1 fitness conservatively estimated using mean oviposition of all second-generation crosses (224.7). Thus, isolation stemming from reduced oviposition in the second-generation (ASF1  ovipos) was 0.159. In contrast to patterns seen for F2 or BC oviposition, none of the second-generation crosses significantly differed in mean percent fertility compared to parentals (U-test, P≥ 0.198; Table 2). With 87.4% of eggs fertilized, the Z/E × E paternal backcross had the lowest fertility and an F2 intercross (Z/E × Z/E) had the highest percent fertility (97.5%).

F2 and BC inviability

We were not able to confirm “hybrid breakdown” in BC or F2 offspring (Arbuthnot 1944; Liebherr and Roelofs 1975). Mean viability for second-generation zygotes ranged from 93.6% to 98.1% (Table 2). The only cross that significantly differed from parental crosses showed higher survivorship (Z/E × E = 98.1%, P < 0.01), but this value was not significant after correcting for multiple tests.

CUMULATIVE REPRODUCTIVE ISOLATION

We explored how differences in the AS of seasonal isolation and behavioral hybrid male dysfunction could affect cumulative isolation by investigating two scenarios. “Strong” reproductive isolation was modeled after locations in which strains differ in voltinism (BE and UZ) and the E strain is at high population density. Under this scenario, seasonal isolation was calculated as the mean index across years for Farmington, NY (Table 1), and behavioral dysfunction used the value calculated for Geneva, NY (Fig. 2B,C). “Weak” reproductive isolation was modeled after sites in which strains are both bivoltine (BZ and BE) and the Z strain predominates. Under this scenario, seasonal isolation was calculated as the value for Dresden, NY (Table 1), and dysfunction was equal to the estimate for Beltsville, MD (Fig. 2B,C).

Violation of several assumptions regarding trait variation and environmental effects may result in inaccurate estimates of the strength of reproductive isolation between Z- and E-pheromone strains. With the exception of seasonal isolation, barriers were evaluated between UZ and BE moths. Therefore, an assumption under the weak model is that barrier strength is identical for UZ versus BE population pairs and BZ versus BE pairs. Trait covariation has not been extensively studied in the ECB. However, the observation that genetic variation across many loci is shared between BZ and UZ populations (e.g., Coates et al. 2004; Dopman et al. 2005), coupled with the finding that both populations are strongly differentiated with respect to BE moths (at Tpi) (Dopman et al. 2005), suggests that gene flow has been relatively recent between BZ and UZ populations. A second assumption that applies to both models is that circadian temporal isolation, which was measured by Liebherr and Roelofs (1975) between an Ontario (UZ) and a NY (BE) strain, is equivalent between UZ and BE populations in NY (which were studied here). Finally, we estimate total isolation assuming that all barriers have a genetic basis. However, a combination of genetic (PDD) and environmental effects account for seasonal isolation because breeding cycle can only be measured in the field. In comparison, phenotypic differences contributing to other barriers will have a larger genetic component because they were measured in the laboratory using a common garden regime.

Table 3 shows estimates of AS, SS, and T under the two isolation models. In both models, we adjusted for uncertainty in the spatial scale in which male orientation occurs (i.e., aggregated vs. dispersed females) by using the mean of AS for short-distance (Fig. 4A) and long-distance orientation (Fig. 4C). Under the “strong” model, T from the measured traits was 99.4% (using mean SS values of E × Z and Z × E pairings, Table 3). Under the “weak” model, this value was 96.4%. Although this difference may not be statistically significant because our estimates of reproductive isolation are based on means with unknown confidence intervals, the biological significance of increased gene flow could be dramatic. Given that barriers operate sequentially over the ECB life history, earlier-acting barriers contributed more to T in both scenarios (Table 3; Fig. 6). As a result, differences between models were largely attributed to different values of ASseasonal rather than ASF1♂  dys. The relative contribution of seasonal isolation to total isolation was 66% (SSseasonal= 0.655) and 31% (SSseasonal= 0.298) under the “strong” and “weak” models (Fig. 6). In contrast, behavioral dysfunction contributed <1% in both (mean strong SSF1♂  dys.= 0.006, mean weak SSF1♂  dys.=−0.015). Because seasonal isolation operates before all other barriers (ASseasonal=SSseasonal), weak values resulted in increased sequential strength for all later-acting barriers. Weak ASseasonal also contributed to more asymmetry in T between the Z- and E-pheromone strains. This effect stemmed from an increase in sequential strength of asymmetric later-acting barriers, specifically male behavioral isolation (Fig. 4A). Indeed, T showed strong asymmetry only under the weak model (Z × E = 90.7%; E × Z = 99.3%) (Table 3).

Table 3.  Reproductive isolating barriers between Z-, and E-pheromone strains of Ostrinia nubilalis.1
ClassIsolating barrier Absolute strength  (strong2) Absolute strength  (weak2) Sequential strength  (strong2) Sequential strength  (Weak2)
 E×Z Z×E Mean E×Z Z×E Mean E×Z Z×E Mean E×Z Z×E Mean
  1. 1Absolute strength of reproductive isolation (AS) typically describes the intensity of gene-flow reduction between strains relative to that occurring within strains and in the absence of other barriers (see text for details). Sequential strength of isolation (SS) describes the intensity that the barrier prevents gene flow that has escaped earlier acting barriers. 0, free gene flow; 1, no gene flow. Cross-type designations list females first.

  2. 2In a scenario of strong isolation, seasonal isolation was calculated from the mean index across years for Farmington (Table 1) and behavioral dysfunction used the values calculated for Geneva (Fig. 2). In a scenario of weak isolation, seasonal isolation equaled that estimated for Dresden, NY (Table 1), and dysfunction equaled that estimated for Beltsville, MD (Fig. 2).

  3. 3Values of absolute isolation were not statistically significant (P>0.05).

  4. 4Calculated using mean oviposition of parental and second-generation crosses. Only one of six second-generation crosses was significant after Bonferonni correction (P<0.008, Table 2).

PrematingSeasonal temporal isolation 0.655 0.655 0.655 0.298 0.298 0.298 0.655 0.655 0.655 0.298 0.298 0.298
 Circadian temporal isolation 0.525 0.525 0.525 0.525 0.525 0.525 0.181 0.181 0.181 0.369 0.369 0.369
 Male orientation 0.862 0.717 0.790 0.862 0.717 0.790 0.141 0.117 0.129 0.287 0.239 0.263
 Female discrimination 0.712 0.298 0.505 0.712 0.298 0.505 0.016 0.014 0.017 0.033 0.028 0.035
Postmating-prezygoticClasping 0 0 0 0 0 0 0 0 0 0 0 0
 Oviposition 0.0843 0.309 0.197 0.0843 0.309 0.197 0.001 0.010 0.003 0.001 0.020 0.007
 Fertilization 0 0 0 0 0 0 0 0 0 0 0 0
PostzygoticF1 Inviability 0 0 0 0 0 0 0 0 0 0 0 0
 F1-male behavioral dysfunction 0.184 0.753 0.469 0.345−1.422−0.539 0.001 0.017 0.006 0.004−0.065−0.015
 F1 Oviposition 0.1594 0.1594 0.1594 0.1594 0.1594 0.1594 0.001 0.001 0.001 0.001 0.018 0.007
 F1 Fertilization 0 0 0 0 0 0 0 0 0 0 0 0
 F2 and BC Inviability 0 0 0 0 0 0 0 0 0 0 0 0
 Total Isolation       0.996 0.995 0.994 0.993 0.907 0.964
Figure 6.

Relative contribution of isolating barriers to total reproductive isolation (T) under “strong” and “weak” isolation scenarios in which seasonal isolation and behavioral dysfunction varied in absolute strength. Values for each barrier were obtained using mean sequential strength (see Table 3).

Discussion

Reproductive isolating barriers are among the primary forces generating and maintaining biodiversity in nature (Dobzhansky 1937; Mayr 1942; Coyne and Orr 2004). Here, we report on a case study of North American Z- and E-pheromone strains of ECB, in which three major classes of reproductive barriers–premating, postmating-prezygotic, and postzygotic–are considered. Our evaluation of 12 potential isolating barriers demonstrates that reproductive isolation arises from manifold sources whose effects are far from uniform. Consequently, the genetic and evolutionary basis for speciation will be complex. Although the cumulative action of all barriers produces nearly complete reproductive isolation, polymorphism in strength of reproductive isolation allows for substantial opportunity for gene flow. Indeed, the ECB likely exists as a mosaic of incompletely isolated populations that vary both in the strength and form of reproductive isolation. Nevertheless, of the three classes of barriers that were measured, premating forms appear to be the most potent in maintaining the integrity of sympatric Z and E populations in nature, possibly because they evolve rapidly between diverging populations.

PREMATING BARRIERS

Seasonal temporal isolation

Although accounts of temporal isolation may be numerically rare, the phylogenetic breadth of taxa that experience temporal isolation suggests that this paucity reflects limited study rather than low evolutionary frequency (e.g., fish [Quinn et al. 2000], insects [Cooley et al. 2003], flowering plants [Martin and Willis 2007], corals [Knowlton et al. 1997], algae [Clifton 1997]). Indeed, observing mating activity across breeding seasons represents a major technical challenge. In spite of these difficulties, and in support of its prevalence, nearly 30 years ago Roelofs and colleagues uncovered ecological differences contributing to seasonal temporal isolation in the ECB and proceeded to characterize sympatric UZ, BZ, and BE voltinism “races” in NY (Eckenrode et al. 1983; Roelofs et al. 1985). Since this pioneering work, however, the consequences of ecological differences in breeding cycle on patterns of gene flow have remained unclear.

Our survey of 41 localities (Fig. 1) over a 9-year-period (1999–2007) records the persistence of sympatric bivoltine and univoltine populations of ECB in NY. The degree of asynchrony in breeding cycle across many sites indicates that voltinism can severely restrict gene flow between pheromone strains (e.g., Fig. 3A). Indeed, our results show that when BE and UZ moths are sympatric, as much as 80% of gene flow is prevented by seasonal isolation alone (Table 1). However, strains at several sites in NY (Fig. 3B; Table 1) and in other geographic areas (e.g., North Carolina, Sorenson et al. 1992) have overlapping breeding cycles, or patterns of PDD times that are consistent with overlapping breeding cycles (NY, Glover et al. 1991; France, Thomas et al. 2003). These findings indicate that seasonal temporal isolation is only a locally important barrier to gene flow between Z and E strains, and that, overall, breeding cycle differences constitute a rather ineffective isolating barrier. Indeed, at the same study site, we found that the strength of temporal isolation can vary dramatically between years. At Farmington, ASseasonal was above 0.6 in three of the four sampled years, but in 2003, it fell to 0.388 (Table 1). The cause of these fluctuations is unknown, but environmental effects or sampling artifacts are two possible explanations. Changes in weather are known to affect the number and shape of ECB breeding cycles (e.g, via temperature-dependent development; Beck 1983). Alternatively, changes in farming practices, including the use of BT corn or insecticides, can affect the size of annual flights through differential mortality (e.g., Burkness et al. 2001). Such changes could increase the breeding cycle overlap for populations that are genetically bivoltine and univoltine (e.g., at Pdd). Finally, as traps were positioned on 41 active farms across 9 years, the uniformity of sampling in some years may have been compromised.

Although breeding cycle differences can be a consequence of environmental variation, voltinism in NY can be at least partially explained as a byproduct of genetic differences at the major, Z-linked QTL affecting PDD time (McLeod 1978; Reed et al. 1981; Glover et al. 1991, 1992). Therefore, the origin of seasonal isolation can be understood as the evolution of developmental rate. It has been suggested that trait divergence resulted from selection, most likely from an ecological source, because Tpi shows evidence for selection and is tightly linked to the Pdd locus (Dopman et al. 2005). PDD time may have evolved because it influences fitness through its effect on generation time, or because it helps coordinate the performance of life-cycle activities (e.g., reproduction, growth, diapause) at appropriate times in the season (Bradshaw et al. 2004; Bradshaw and Holzapfel 2007).

Circadian temporal isolation

Besides seasonal temporal isolation, at least some ECB strains are temporally isolated through differences in 24-h, circadian time of mating (Liebherr and Roelofs 1975). Under laboratory conditions, Liebherr and Roelofs reported that BE-strain moths showed delayed mating compared to UZ-strain moths, which mate soon after darkness (Fig. 3C). Other interpretations for this pattern have been proposed (Webster and Cardé 1982), and pheromone strains in Europe display different circadian rhythms (Karpati et al. 2007). Nevertheless, a ∼50% reduction in gene flow would be expected if the asynchrony in mating time reported by Liebherr and Roelofs persists in the field (Table 3). The basis for circadian mating rhythms in the ECB is unclear, but a genetic component seems likely because differences were detected after more than eight generations in a common laboratory environment. At least in Drosophila, circadian mating periodicity is genetically controlled by an endogenous clock (e.g., Sakai and Ishida 2001).

Male and female behavioral isolation

Considering the prevalence of hybridization between animal taxa that lack strong differences in mate selection preferences or signals, Mayr (1963) argued that behavioral isolation is the most important form of reproductive isolation. In ECB, the importance of behavioral isolation is unequivocal. When Z and E moths are dispersed (1–50 m), numerous authors have reported that the use of different preferences and signals (i.e., male orientation to volatile sex pheromones released by females) results in limited cross attraction (e.g., Roelofs et al. 1987; Bontemps et al. 2004), and nearly complete reproductive isolation (Fig. 4C). What is not clear from these studies, however, is the extent that these preferences and cues prevent hybridization when moths are not dispersed. In general, behavioral isolation is usually examined in one context, for example, a single spatial scale or a single ecological or laboratory setting (Coyne et al. 2005; but see Craig et al. 1993). In the ECB, the spatial context for mating is important because moths of both strains can be found to aggregate in the field where mate location or attraction could conceivably occur without the use of long-distance olfactory cues (Showers et al. 1976; Malausa et al. 2005).

By observing mating behavior, our study confirms that over very short distances (<20 cm), males of both strains continue to show orientation specificity towards females of their own strain (Fig. 4A). Combined with a recent investigation demonstrating that adding sex pheromone fails to increase interstrain mating (Pelozuelo et al. 2007), this result can be taken as prima facie evidence that male orientation to female cues, most likely to female sex pheromones, causes behavioral isolation over a broad range of spatial scales. However, isolation strength is reduced when males and females are in close proximity (AS♂  orient., UZ × BE: 0.448 [<20 cm] vs. 0.988 [∼2 m]; E × Z: 0.722 [<20 cm] vs. 1.0 [∼2 m]), suggesting that moth dispersion could shape the extent that differences in male behavior reduce gene flow. Further studies are needed to determine the reasons for increased rates of male orientation, but two possible explanations are environmental dependence (e.g., flying 1–2 m upwind to sex pheromone stimuli vs. a combination of walking and short flights to signaling females), or the presence of an unidentified short-distance female cue (other than known sex pheromones) that elicits male orientation and which is similar in both strains.

Previous research has attributed patterns of assortative mating between Z and E moths to male orientation specificity alone (e.g., Cardé et al. 1978; Klun and Huettel 1988; Glover et al. 1991). However, like many other animal species (Andersson 1994), we find that ECB females display choosy mate-selection behavior that manifests as active discrimination when confronted by courting moths of the opposite strain (Fig. 4B). When rejecting a male's courtship attempt, many females contort their abdomen so that copulation is impossible, others simply walk or fly away. As a result, female behavioral isolation in some pairings is as strong as male behavioral or seasonal temporal isolation (E × Z: AS♀  discrim.= 0.712) (Table 3). Recently, it has been demonstrated that a male pheromone, produced by the hairpencil, is critical for female acceptance of courting males (Lassance and Löfstedt 2009). Combined with our results, this finding suggests a role for this male pheromone in female behavioral isolation. In addition to helping explain the deficit of hybrids in nature, male and female mate preferences shed light on biases in the type of hybrids that are produced. Both male and female behaviors are asymmetric, with a deficit of successes when E females pair with Z males (Fig. 4A,B). Although behavioral asymmetries such as these are common between animal species (Kaneshiro 1980; Arnold et al. 1996), and it was known that Z males were more selective (Roelofs et al. 1987; Glover et al. 1991; Linn et al. 1997), it is unusual that both sexes of the Z strain are more discerning. Regardless of the evolutionary explanation for this pattern, the proximate consequence is strong asymmetry in hybridization potential. This potential seems to be realized in the field, as an extensive survey of ECB in NY failed to uncover E female × Z male hybrids (Glover et al. 1991).

Male behavioral isolation stems from changes at an uncharacterized, autosomal fatty-acid reductase that affects pheromone blend ratios (Zhu et al. 1996; Roelofs et al. 1987; Dopman et al. 2004), along with changes at a sex-linked, major QTL for orientation (Glover et al. 1990; Dopman et al. 2004). A candidate gene(s) affecting male orientation has yet to be identified; however, a gene(s) that functions after peripheral sensory processing of pheromone cues is likely (Willett and Harrison 1999; Linn et al. 1999; but see Karpati et al. 2008). As noted by Coyne and Orr (2004), monogenic inheritance of male behavioral isolation in ECB contrasts with polygenic inheritance in other species. With this simple genetic basis, it is difficult to understand how new preferences or signals could have evolved because most mutations are expected to have large and deleterious phenotypic effects (review in Löfstedt 1993). The existence of rare ECB males that orient towards females from both the ECB and the Asian corn borer (ACB, O. furnacalis) provides a partial solution (Reolofs et al. 2002)–some ECB males may have been preadapted to orient towards the first mutant Z- or E-strain female, mitigating initial fitness costs associated with pheromone blend evolution. Regarding female behavioral isolation, both proximate and ultimate causes will remain unknown until the sensory modalities involved have been elucidated. Acoustic communication occurs in the ACB, in which males produce an ultrasonic courtship “song” (Nakano et al. 2006), but an olfactory system is also possible (Royer and McNeil 1992; Lassance and Löfstedt 2009).

POSTMATING-PREZYGOTIC BARRIERS

Mechanical and gametic isolation

“Cryptic” reproductive barriers, which include mechanical and gametic forms of postmating-prezygotic isolation, although difficult to observe have been documented across diverse organisms (e.g., in broadcast spawners [Swanson and Vacquier 1997], in insects [Gregory and Howard 1993; Herndon and Wolfner 1995; Price et al. 2001; Nosil and Crespi 2006], in fish [Mendelson et al. 2007], review in Coyne and Orr 2004). We find that mechanical isolation is either limited or absent between Z and E strains (Fig. 5A), but our ability to test for clasping failures was restricted by a small sample size in E × Z pairings. Limited mechanical isolation would corroborate the findings of Cardé et al. (1978), in which genitalic differences between strains that might contribute to mechanical isolation were not detected. In contrast to mechanical isolation, we find that ECB strains exhibit partial gametic isolation, confirming preliminary results obtained over 30 years ago (Liebherr and Roelofs 1975). Gametic incompatibilities are moderate in intensity and are significant in only one cross direction. When Z females mate with E males, lifetime egg production drops by ∼30% (Fig. 5B; ASovipos= 0.309, Table 3), but fertilization rate remains high (Fig. 5C).

Postmating-prezygotic problems are often highly variable and asymmetric between diverged populations or species, and cross failures can have multiple causes (Price et al. 2001; Nosil and Crespi 2006; Mendelson et al. 2007). In the ECB, explanations involving reduced remating frequency can be rejected based on the presence of a single transferred spermatophore in all females, but many other potential mechanisms remain. It is possible that fertilization in Z × E pairings proceeds normally, but foreign seminal proteins or unrecognized copulatory behavior fails to stimulate egg laying (e.g., Herndon and Wolfner 1995). Alternatively, in some interspecific crosses (e.g., Gregory and Howard 1993), reduced oviposition reflects poor fertilization and the retention of unfertilized eggs. Even with good demonstrations of cryptic isolation (e.g., Swanson and Vacquier 1997; Price et al. 2001; Nosil and Crespi 2006; Mendelson et al. 2007), the processes driving evolution are poorly understood, and thus represent a clear avenue for further work.

POSTZYGOTIC BARRIERS

Behavioral and physiological infertility

Behavioral hybrid infertility or dysfunction has two primary causes that are often difficult to distinguish (Stratton and Uetz 1986; Frey and Bush 1996; Noor 1997; Servedio and Noor 2003; Linn et al. 2004; Coyne and Orr 2004). Failures or delays in mating can result when hybrid behaviors are intermediate compared to parental individuals (extrinsic behavioral dysfunction) (e.g., Stratton and Uetz 1986). Alternatively, failures or delays can result when otherwise “normal” hybrid behaviors are inhibited or disrupted by physiological or neurological abnormalities (intrinsic behavioral dysfunction) (e.g., Noor 1997). Our reanalysis of Roelofs et al. (1987) suggests that both extrinsic and intrinsic dysfunction contribute to reduced F1 male mating success in the ECB (Fig. 2A). Although F1 males show moderate response to the Z pheromone and therefore do not exhibit exclusively intermediate behavior, few hybrids respond to the E-strain blend. Similarly, although hybrids show appreciable orientation to the Z and hybrid blends, they do not respond as readily as Z and E males to their own pheromones.

Due to orientation patterns among Z, E, and hybrid males, behavioral hybrid dysfunction becomes apparent only under restricted circumstances. Specifically, dysfunction and intermediate levels of isolation occur when E females predominate (e.g., Geneva, NY in Fig. 2B,C; mean ASF1♂  dys.= 0.469, Table 3). When Z and hybrid females are at high frequency (e.g., Beltsville, MD, in Fig. 2B,C), hybrids experience intermediate relative fitness and introgression is promoted (mean ASF1♂  dys.=−0.539, Table 3). Two relevant points suggest that behavioral dysfunction could be weak in nature. First, our measure is based on long-distance attraction, but males may orient to a wider spectrum of female genotypes when aggregated (Fig. 4A vs. C). Second, conditions favorable to hybrid male orientation are probably common in the field because Z females are typically the most numerous genotype (Table 1; Klun and Cooperators 1975; Anglade and Stockel 1984). Whatever its impact on gene flow, the evolution of behavioral dysfunction is linked to male behavioral isolation because both barriers have a common genetic basis.

In contrast to behavioral dysfunction, the fertility effects of hybrid physiology or development is one of the most commonly studied forms of reproductive isolation (review in Coyne and Orr 2004). In the ECB, Liebherr and Roelofs (1975) reported that F2 intercrosses yielded fertile eggs, whereas only ∼70% of backcrosses were successful. Our measure of fertility (percent fertile eggs) indicates that both F2 and BC pairings proceed normally (∼90% fertile, Table 2), but BCs do suffer disproportionately in terms of fecundity (∼25% fewer eggs than F2 or parental crosses). However, among all second-generation crosses, oviposition drops by only ∼15% compared to crosses occurring within-strain. Thus, our results indicate that physiological hybrid infertility is a rather weak isolating barrier (ASF1  ovipos.= 0.159, Table 3). Inheritance patterns of oviposition in first and second-generation crosses (Fig. 5B; Table 3) might be consistent with an incompatibility between an autosomal recessive allele in the Z strain and a dominant E-strain allele, but more work is needed to explore the genetic and evolutionary basis of this phenotype.

THE PHENOTYPIC ARCHITECTURE OF SPECIATION

The number and strength of isolating barriers that have evolved before speciation is complete, which we refer to here as the “phenotypic architecture” of speciation, provides insight into several outstanding problems in speciation research. One important issue is whether this phenotypic architecture typically consists of one or a few strong barriers to gene flow or whether many weaker barriers are involved. The Z and E strains of ECB most clearly fall into the latter category, given that seven of 12 potential forms of isolation restrict gene exchange, but all are less than 80% complete (Table 3). Comparable studies have revealed a similar pattern (e.g., Chamerioin angustifolium populations, Husband and Sabara 2003; sister species of Mimulus, Ramsey et al. 2003; lineages of Pogonomyrmex ants, Schwander et al. 2008; Etheostoma darters, Mendelson et al. 2007; review in Coyne and Orr 2004), suggesting that speciation is commonly an extended and complex process that requires an accumulation of barriers of intermediate strength. Nevertheless, the role of one or a few strong barriers to gene flow has been documented between some diverging populations (review in Funk et al. 2002; Nosil et al. 2005; Lowry et al. 2008).

An evolutionary explanation for these two contrasting phenotypic architectures of speciation may be related to differences in the geographic mode of divergence. A history that includes long intervals of sympatry might result in a small number of barriers because gene flow will constrain the evolution of new phenotypic differences (possibly to a subset driven by strong selection [Rice and Hosert 1993]), and impede some types of evolutionary processes from promoting divergence (e.g., by genetic drift or “mutation-order” speciation [Schluter 2009]). Those few barriers that do evolve in sympatry must arise rapidly and be great in strength (e.g., behavioral isolation in Drosophila, Coyne and Orr 2004), or else population fusion will occur. In contrast, geographical separation facilitates the origin of trait differences and reproductive barriers by almost any evolutionary process (besides reinforcement), and restricted migration may facilitate the persistence and accumulation of trait differences and barriers of varying strength (Rice and Hosert 1993). Under this reasoning, the complex phenotypic architecture of speciation that is observed between ECB strains may have been promoted by an extended historical period(s) of restricted gene flow, such as that imposed by physical separation.

Uncovering manifold causes of speciation suggests a high level of complexity in the underlying genetic architecture. That is, the genetics of speciation will typically consist of more loci with the potential for more elaborate control (e.g., epistasis, additive), phenotypic effects (e.g., minor versus major), and chromosomal distributions, as the number of barriers preventing gene flow increases. Besides underscoring the obvious truth that disentangling the genetics of speciation in many (or most) species will require tremendous effort, a more profound implication is how this genetic complexity will shape patterns of gene flow when populations are sympatric.

The genetic architecture of speciation has consequences for genome-wide patterns of gene flow because chromosomal regions linked to reproductive barrier loci will introgress across species boundaries at lower rates than unlinked regions (Barton and Hewitt 1981; Harrison 1990; Wu 2001; Ting et al. 2000). As a result of this effect, species or racial boundaries are often thought of as “semipermeable” or “porous,” with permeability depending on the genetic marker (Barton and Hewitt 1981; Harrison 1990; Wu 2001; Via 2009). For a semipermeable genome, in regions that are unlinked to barrier loci, gene flow will constrain local adaptation but facilitate the spread of new, globally beneficial alleles. Conversely, restricted gene flow near barrier loci will promote further population divergence. Thus, the phenotypic architecture of speciation will shape the genomic context in which diverging populations change in response to spatial or temporal heterogeneity. Speciation in which a large number of reproductive barriers are involved is predicted to increase the size of the semipermeable fraction of diverging genomes, whereas the strength of each barrier (and possibly when it operates over an organism's life history) may affect the degree of permeability at underlying loci.

Z and E strains of ECB support the semipermeable genome model of speciation. In upstate NY, even where they exist together in the same fields, E and Z borers remain differentiated for Tpi suggesting that gene flow is limited or absent for this locus. Patterns of variation at Tpi may be a consequence of close physical linkage to Pdd (Dopman et al. 2004, 2005). In contrast, other nuclear genes (two of which, kettin and Ldh, are also on the sex chromosome) reveal no evidence of differentiation, because of recent or ongoing hybridization and introgression (Dopman et al. 2005). With a relatively large number of reproductive barriers between Z and E strains (Table 3), the ECB provides an important study system in which to investigate the consequences of the phenotypic and genetic architectures of speciation for introgression and local adaptation.

A complex phenotypic architecture of speciation presents challenges for understanding the evolutionary forces driving speciation. However, only a subset of barriers and evolutionary processes will have been important in initiating the speciation process. Testing hypotheses about these early-evolving barriers typically requires a comparative approach (e.g., Coyne and Orr 1997; Mendelson 2003; but see Christianson et al. 2005), but Coyne and Orr (2004) noted that it is unlikely for a barrier to have played a major historical role unless its contemporary strength is large. Therefore, premating barriers may be among the most historically important between strains of the ECB (Table 3).

There are at least two reasons why premating barriers might be predisposed to rapid evolution–they may be common targets of selection, or they could be more mutable. Comparative studies suggest that behavioral isolation can rapidly evolve by reinforcement in sympatry or as a byproduct of sexual selection in allopatry (Coyne and Orr 1989, 1997; Mendelson 2003). For reinforcement to explain patterns in the ECB, hybrids must experience reduced fitness. Hybrid ECB are relatively fit in the laboratory (Fig. 2C; Table 2; Calcagno et al. 2007), but reinforcement cannot be ruled out until ecologically dependent costs (e.g., Rundle and Whitlock 2001) have been excluded. In contrast, conditions favoring sexual conflict are known. Male ECB fitness increases with more mating (Royer and McNeil 1993), but the act of mating reduces female survival (Fadamiro and Baker 1999). In contrast to behavior, breeding cycle differences are unlikely to have been historically important because Z and E strains breed at similar times at most sympatric sites (Glover et al. 1991; Sorenson et al. 1992; Thomas et al. 2003). Nevertheless, circumstantial evidence suggests that PDD time might be prone to rapid adaptive evolution in the ECB (Dopman et al. 2005; Malausa et al. 2007a). Empirical studies have documented rapid evolution of circannual and circadian rhythms in both animals and in plants (e.g., Weber and Schmid 1998; Tauber et al. 2007), with diapause traits being among those phenotypes showing adaptive responses in insects (e.g, Tauber et al. 2007).

Premating barriers might also rapidly evolve because their genetic determinants are predisposed to mutation. In yeast, the spontaneous gene duplication/deletion rate is orders of magnitude larger than the nucleotide substitution rate (Lynch et al. 2008), and the genes determining pheromone differences between Ostrinia are in a multigene family (Roelofs et al. 2002; Roelofs and Rooney 2003). Together, these results suggest that the number of pheromone genes that are polymorphic in copy number might be large and that the (in)activation of different paralogs could provide a major source of “raw material” for rapid sex-pheromone evolution. Although pheromone blend differences between ECB strains seems to be from allelic changes rather than from alternative paralog usage, mutational mechanisms could be responsible for rapid behavioral isolation in the Ostrinia genus, and perhaps, in other species.

THE IMPORTANCE OF PRE- VERSUS POSTZYGOTIC BARRIERS TO SPECIATION

There has been long-standing debate about the relative contributions of prezygotic versus postzygotic barriers to speciation (e.g., Coyne and Orr 1989, 1997; Kirkpatrick and Ravigne 2002). It is clear that the timing of the evolution of barriers during the often-extended process of speciation is most clearly related to the historical or evolutionary causes of speciation. In this sense, those barriers that evolved first (e.g., as identified by comparative studies) are the most “important.” However, a barrier's relative or “sequential” strength over an organism's life history informs a related issue about the contemporary or ecological causes of speciation in which the main question is how diverging populations remain distinct. Accumulating evidence supports the interpretation that premating barriers may be most important forms of isolation between in both contemporary and historical population pairs, but the debate is far from resolved (Ramsey et al. 2003; Husband and Sabara 2003; Nosil et al. 2005; Martin and Willis 2007; Mendelson et al. 2007; Nosil 2007; Lowry et al. 2008; Schwander et al. 2008; Schluter 2009).

For the ECB, with more than 98% of total cumulative reproductive isolation consisting of premating barriers, our findings broadly confirm early studies (e.g., Arbuthnot 1944; Liebherr and Roelofs 1975), which implicate prezygotic barriers as the most important for the maintenance of Z and E strains in nature (Table 3; Fig. 6). In the weak isolation scenario, which probably represents a close approximation to that experienced by many ECB strains in the field, temporal isolation (seasonal + circadian) contributes nearly two-thirds to the total, whereas another ∼third is contributed by behavioral isolation (male + female) (Fig. 6). All other isolating barriers combined (postmating-prezygotic + postzygotic) contribute a little more than 1% at most to total isolation. Granted, not all isolating barriers have been measured in the ECB. Patterns consistent with “hybrid breakdown” have been documented in life stages other than those investigated here (i.e., in larvae, Arbuthnot 1944; Liebherr and Roelofs 1975), and strong extrinsic (postzygotic and prezygotic) barriers may have gone undetected in our benign laboratory setting (e.g., environmentally dependent postzygotic incompatibilities at Pdd). Nevertheless, some Z and E populations in Europe differ in traits associated with specialization to different host plants (Thomas et al. 2003; Calcagno et al. 2007; Malausa et al. 2008). If these differences prevent gene flow then Z and E strains of ECB would almost certainly join the growing list of incipient species that support premating barriers as the strongest impediments to gene flow (e.g., Ramsey et al. 2003; Husband and Sabara 2003; Nosil 2007; Martin and Willis 2007; Mendelson et al. 2007; Lowry et al. 2008).

TOTAL REPRODUCTIVE ISOLATION AND PATTERNS OF GENE FLOW IN NATURE

For most species the biological context for the maintenance or erosion of genetic differences between populations in nature is obscure, but we can begin to characterize this relationship for Z and E strains of the ECB. One expected pattern is a negative association between estimates of cumulative isolation and estimates of genetic exchange in the field. Based on measures of cumulative isolation, sympatric populations of UZ and BE moths should produce fewer hybrid offspring and experience less gene flow than sympatric BZ and BE moths (Table 3). In agreement with this prediction, the proportion of hybrid females is smaller at the UZ/BE Geneva site than at the BZ/BE Beltsville site (Fig. 2B; Roelofs et al. 1985; Klun and Huettel 1988), and genetic differences have been maintained for ∼20 years between UZ and BE populations at Geneva (Glover et al. 1991; Dopman et al. 2005). Presumably, more recent evidence for introgression would be found between strains at Beltsville and at other sites where seasonal isolation is weak.

Although it is unlikely that measures of cumulative isolation translate easily into quantitative assessments of hybridization or gene flow, cumulative isolation can inform taxonomic issues because the BSC defines species by the presence of reproductive barriers (Ramsey et al. 2003; Husband and Sabara 2003; Martin and Willis 2007). This is particularly useful for Z and E strains of ECB because their morphological and genetic similarity (e.g., Liebherr 1974; Dopman et al. 2005) has led to regular taxonomic debate and the emergence of two opposing views. Thirty years ago, Cardé and colleagues (1978) argued on the basis of allozyme, pheromone, and hybridization data that pheromone strains are distinct species. More recently, Frolov et al. (2007) asserted that Z and E strains using maize, such as those investigated here, belong to one species, whereas those using non-maize hosts (e.g., mugwort) belong to another. With as much as ∼10% gene flow permitted (in Z × E crosses, Table 3) and more than 15% hybrids at some sites (Fig. 2B; Klun and Huettel 1988), claims of distinct species seems unjustified. On the other hand, Frolov et al.'s (2007) treatment of maize-feeding strains as conspecific appears to be at odds with our finding that total isolation approaches completion if seasonal isolation is strong (Table 3). These contrary interpretations can be reconciled by interpreting Z- and E- pheromone strains as a mosaic of incompletely isolated populations that differ in levels of total reproductive isolation (e.g., due to voltinism [here] or host–plant use [Thomas et al. 2003; Calcagno et al. 2007; Malausa et al. 2008]). A more dynamic application of the BSC that acknowledges this diversity is necessary for the ECB. When combined with a comprehensive genetic survey, such a perspective can lead to a better understanding of the process in which alleles at reproductive barrier loci spread across populations before ultimately giving rise to completely isolated and distinct species.

Conclusion

Our study determined that speciation between the closely related Z and E strains of ECB has diverse phenotypic causes, which are incomplete when acting alone but also when operating together. The presence of numerous incomplete barriers between these partially isolated populations suggests that speciation in this lineage is a slow process relative to the rate in which isolating barriers evolve. Along with recent investigations, our observations indicate that premating barriers represent the main impediments to gene flow between species, potentially because premating barriers evolve at early stages of the speciation process. Testing for rapid evolution of premating barriers requires a comparative approach, with numerous barriers measured across a large group of taxa that vary in levels of evolutionary divergence. One such study has been conducted in Lepidoptera (Presgraves 2002), but with only postzygotic forms of isolation characterized, the general principles governing the origin of new Lepidopteran species have yet to be discovered. Estimating the strength of behavioral, temporal, and other forms of reproductive isolation across the Ostrinia genus and among related species holds promise for revealing these general principles.


Associate Editor: C. Jiggins

ACKNOWLEDGMENTS

We are grateful to R. Harrison, L. Hatch, D. Hartl, C. Jiggins, and three anonymous reviewers for helpful comments on this manuscript and to C. Linn, W. Roelofs, and K. Poole for advice, suggestions, and materials. This research was partially supported by a National Institutes of Health Kirschstein-NRSA Postdoctoral Fellowship (GM080090-01 to E.B.D.).

Ancillary