The formation of new species involves the evolution of barriers to gene exchange. One such barrier is sexual isolation, where divergent mate preferences prevent copulation between taxa. Sexual isolation can evolve via a number of processes, including natural selection, sexual selection, genetic drift, and reinforcing selection to avoid maladaptive hybridization. Conversely, gene flow between populations generally erodes the evolution of sexual isolation. In Timema cristinae walking stick insects, some sexual isolation between populations evolved through ecological divergence in host-plant use, implicating a role for natural selection. However, reinforcement and gene flow also contribute, such that sexual isolation is strongest when migration between hosts is high enough to result in selection to avoid hybridization, but low enough to prevent gene flow from eroding adaptive divergence in mate choice. Both parallel and contrasting patterns can be seen in other Orthopteroids. This variation among groups might reflect 1) the degree to which different groups are intimately associated with their food resources, 2) the types of traits used in mate choice, and 3) how the geographic arrangement of populations affects the opportunity for reinforcement.
Speciation involves the evolution of reproductive isolation between diverging populations. Understanding speciation thus requires determining which reproductive barriers initially reduced gene flow between populations and the evolutionary forces producing them (Mayr 1947, 1963; Coyne & Orr 2004). The evolution of premating isolation caused by divergent mating signals and preferences (sexual isolation hereafter) appears to be an important component of speciation in many taxa (Coyne & Orr 2004). However, questions remain about which processes tend to promote, versus constrain, the evolution of sexual isolation. Here I review the role of 3 processes in the evolution of sexual isolation between populations of Timema walking stick insects: ecological adaptation, reinforcement, and gene flow. I conclude by comparing the results from walking sticks to those from studies of other orthopteroids.
Speciation occurs as an indirect by-product of ecological divergence when barriers to gene flow evolve between populations as a result of ecologically based divergent selection (Mayr 1947, 1963; Schluter 2000; Jiggins et al. 2001). Natural selection causes population divergence in ecologically important traits and these traits (or traits correlated with them) also incidentally cause reproductive isolation. When speciation occurs as a by-product of adaptive divergence, ecologically divergent populations exhibit greater reproductive isolation than ecologically similar populations of similar age (‘ecological speciation’—Funk 1998, Rundle et al. 2000, Schluter 2000, Funk et al. 2002, Rundle & Nosil 2005). The evolution of sexual isolation due to habitat-specific divergent selection on preferences themselves (rather than ecological traits per se) generates the same prediction, and also involves ecological divergence (Endler 1992, Boughman 2002).
Speciation can also involve selection for sexual isolation in geographical regions where hybridization is maladaptive (i.e., reinforcement) (Dobzhansky 1937; see Butlin 1995, Howard 1993, Noor 1999, Servedio & Noor 2003 for reviews). The key prediction of the reinforcement hypothesis is that nonallopatric (geographically contiguous or overlapping) populations will exhibit greater mating discrimination than allopatric (geographically separated) populations, because selection against hybridization can only occur in the former. This pattern of greater divergence in sympatry/parapatry is often called ‘reproductive character displacement’, and previous empirical studies of reinforcement have provided evidence for such a pattern (Noor 1995, Saetre et al. 1997, Rundle & Schluter 1998, Higgie et al. 2000). However, it is unclear how ecological adaptation and reinforcement interact during the speciation process (Schluter 2000, Rundle & Nosil 2005).
Gene flow between populations tends to erode divergence and thus prevent speciation (Mayr 1947, Slatkin 1987, Sanderson 1989, Servedio & Kirkpatrick 1997, Cain et al. 1999). Thus high hetero-specific encounter rates might be expected to prevent the evolution of reproductive barriers. However, heterospecific encounters also provide the opportunity for the selection against hybridization that reinforcement requires (Howard 1993, Noor 1995). Thus the likelihood of reinforcement might be greatest when population sizes are similar and migration rates are intermediate (Sanderson 1989, Servedio & Kirkpatrick 1997, Kirkpatrick 2000).
There have been studies of the evolution of sexual isolation between populations of walking sticks, populations which differ in ecology (host plant use), geography (allopatry vs parapatry) and population size. This variation allows partitioning of the effects of ecology and geography on reproductive isolation and provides the opportunity to explicitly test the predictions outlined above. The results show that the greatest levels of sexual isolation occur under the combined effects of ecological adaptation and reinforcement, whereas high levels of gene flow erode divergence.
Timema walking sticks are wingless, phytophagous insects inhabiting the chaparral of southwestern North America (Crespi & Sandoval 2000). Individuals feed and mate exclusively on the hosts upon which they rest. Most work has focused on T. cristinae, a species feeding upon 2 different host plant species (Ceanothus spinosus and Adenostoma fasciculatum). A ‘population’ of walking sticks is defined as all the individuals collected within a homogenous patch of a single host species (as in Nosil et al. 2002, 2003; Nosil 2004). Thus ‘parapatric’ insect populations are in contact with a population of insects adapted to the alternative host (i.e., they have an adjacent population using the alternative host), whereas ‘allopatric’ populations are geographically separated from all other populations adapted to the alternative host by distances > 50 × the 12-m pergeneration gene-flow distance (Sandoval 1993). Sample sites with both hosts were chosen such that there was only one population on each host species (i.e., each population had only one adjacent population on the alternate host). For simplicity, I use the term ‘hybridization’ to refer to interbreeding between populations on different hosts, but do not imply that the host forms have achieved full species status.
Ecological divergence and the evolution of sexual isolation
Pairs of populations using the same host species can be considered ‘ecologically similar’, whereas those using different host species can be considered ‘ecologically divergent’. Whether sexual isolation was greater between pairs of populations that use the same, vs different, host plant species, was tested by Nosil et al. (2002). A total of 1024 no-choice mating trials were conducted between 28 pairs of populations, with 13 pairs using the same host species and 15 pairs using different host plant species. Consistent with the ecological speciation hypothesis, walking sticks were more likely to copulate if paired with an opposite-sex member from the same host-plant species, than if paired with an opposite-sex member from a different host plant species. Thus the magnitude of sexual isolation detected between pairs of populations using different host plants was significantly greater than the magnitude of isolation detected between pairs of populations using the same host plant (Fig. 1; mean IPSI isolation index = 0.24 vs 0.08 respectively; Mantel test t = 2.24; p < 0.05). Notably, a reciprocal rearing experiment showed that sexual isolation between populations is independent from the host plant species that individuals are reared upon, indicating that sexual isolation likely has a strong genetic basis (Nosil et al. 2003).
Time since divergence is unlikely to confound the results reported above because the degree of sexual isolation observed between populations was not correlated with the genetic distance between pairs of populations (Fig. 2). Moreover, pairs of populations using different host plants were not more genetically divergent from one another than pairs of populations using the same host plant (Fig. 1). Thus ecological divergence in host plant use, rather than neutral differentiation, predicts the magnitude of sexual isolation that evolves.
Reinforcement of sexual isolation
The main prediction of reinforcement was supported; female mating discrimination against foreign males (i.e., males from other populations) is significantly stronger when females are from populations where the 2 host-adapted forms are in geographic contact (i.e., parapatry) than when females are from geographically isolated populations (i.e., allopatry) (Fig. 3; mean between-population copulation frequencies in mating trials involving females from parapatric populations = 28%, sx̄ = 0.45; in trials using females from allopatric populations = 35%, sx̄ = 0.47; p < 0.001 in a logistic regression analysis) (Nosil et al. 2003).
To be relatively certain that reinforcement is the process which caused the pattern of reproductive character displacement documented above, three requirements must be met. First, a history of hybridization is required. Both morphological and molecular data indicate that gene flow (i.e., hybridization) does occur between adjacent pairs of populations on different hosts. Adjacent pairs of populations using different host plants are weakly or not differentiated at mtDNA (mean Fst = 0.07, range = 0.00 to 0.25, n = 7 pairs), while geographically-separated populations are more strongly differentiated (mean Fst = 0.31, range = 0.00 to 0.79, n = 129 pairs; Mantel's t = 2.33, p < 0.01). This result is not restricted to mtDNA, because the magnitude of divergence in color pattern, body size and shape, and host preference is also weaker between parapatric versus allopatric populations (Sandoval 1994a,b; Nosil & Crespi 2004; Nosil et al. 2006). These data suggest that gene flow between neighboring populations occurs in the wild (Coyne & Orr 2004).
Second, reinforcement traditionally requires reduced hybrid fitness. This precondition is likely fulfilled in T. cristinae because hybrid broods contain a higher proportion of color-pattern morphs that are locally noncryptic, than do broods derived from within-population mating (Sandoval 1993). Field experiments and predation trials have shown that natural selection against noncryptic morphs occurs, and is caused by differential visual predation (Sandoval 1994a,b; Nosil et al. 2003; Nosil 2004). Thus hybrids likely suffer higher rates of visual predation than do the parental forms, imposing selection to avoid hybridization. Additionally, females from some parapatric populations exhibit reduced fecundity and oviposition rate following between-population mating, relative to within-population mating (Nosil & Crespi 2006). This reduction in female fitness represents another cost to hybridization, and could also select for mating discrimination against males from the other host.
Third, alternative explanations for the pattern of reproductive character displacement need to be unlikely. Numerous processes other than reinforcement could cause a pattern of reproductive character displacement (Howard 1993, Butlin 1995, Noor 1999 for review). Each of these hypotheses can be viewed as an alternative to reinforcement and examples include ecological character displacement, population ancestry, biased extinction, and male preference for allopatric females. Each of these alternatives was examined, but was unsupported (Nosil et al. 2003 for details). Thus it is likely that reinforcement contributes to the observed pattern of reproductive character displacement.
Relative population size and reinforcement
The 8 parapatric populations studied by Nosil et al. (2003) exhibited a range of variation in the effects of reinforcement (i.e., in the magnitude of sexual isolation). Relative population sizes can affect the likelihood and strength of reinforcement, and may explain some of this variation. Numerous theoretical models have demonstrated that high levels of gene flow between diverging populations can erode the effects of reinforcing selection, and thus prevent reinforcement (Sanderson 1989, Servedio & Kirkpatrick 1997, Cain et al. 1999, Servedio & Noor 2003 for review). However, although migration between divergent populations can result in gene flow between them, it also generates the opportunity for selection against hybridization to occur. Thus migration can exert a dual effect during reinforcement. On the one hand, increasing rarity of a population (i.e., smaller size relative to the population with which it co-occurs) elevates the migration rate into the population. This process raises the probability of encountering mates from the other populations, and thus the opportunity for reinforcing selection (Howard 1993, Noor 1995). On the other hand, such increased opportunity for between-population mating also increases the potential for gene flow between populations, which retards reinforcement. The actual magnitude of mating discrimination that evolves might be expected to reflect a balance between the opposing forces of reinforcing selection and gene flow, with the effects of reinforcement being greatest when population sizes are similar and when migration rates are intermediate.
The dual effects of migration were given explicit theoretical consideration by Kirkpatrick (2000), who examined population divergence in mating preferences between a continent and an island receiving continental migrants. The per generation change in the mean value of a trait used as a cue for assortative mating in the island population (T), is given by the equation:
Two migration terms are represented: m(T'– T) represents the homogenizing effects of gene flow, causing mean trait values between populations to become similar; m (T'– T)I demonstrates that migration can also cause population divergence in mean trait values, by providing the opportunity for selection against migrants and hybrids, which acts as the force driving divergence.
Few empirical studies have examined the effects of migration on the outcome of reinforcement (see Servedio & Noor 2003 for review). The walking stick study of Nosil et al. (2003) explicitly tests for and demonstrates, the dual effects of migration: the magnitude of female mating discrimination against males from other populations is greatest when migration rates between populations adapted to alternate host plants are high enough to allow the evolution of reinforcement, but low enough to prevent gene flow from eroding adaptive divergence in mate choice (i.e., when the sizes of coexisting populations are similar; Fig 4). More specifically, female mating discrimination against males from the alternative host plant was quantified in multiple populations which differed in the relative size of a neighboring, adjacent population. The relative size of the adjacent population is estimated from host-plant patch size, and is positively correlated with mtDNA estimates of gene flow (Nosil et al. 2003). Consistent with the balancing effects of reinforcing selection and gene flow, mating discrimination is low when the study population is allopatric or large relative to its neighbor; it increases rapidly until the sizes of the study and neighboring populations are similar, and it then decreases when the study population is relatively rare (Fig. 4). The generality of the walking stick study is unknown, although at least 3 other studies have documented stronger effects of reinforcement when a species is relatively rarer (i.e., relatively less abundant: Waage 1979, Noor 1995, Peterson et al. 2005). Further studies are required to generate insight into whether migration will tend to facilitate divergence via increased opportunity for reinforcing selection or to act as a homogenizing force.
Collectively, the findings from T. cristinae indicate that both selection and gene flow affect the magnitude of sexual isolation observed in nature. The greatest levels of sexual isolation are observed under the combined effects of ecological adaptation and reinforcement (Fig. 3). There are few other empirical or theoretical studies of how ecology and reinforcement interact during speciation (but see Rundle et al. 2000, Jiggins et al. 2001 for data and Kirkpatrick 2001 for theory).
A number of outstanding questions remain. For the host ecotypes of T. cristinae, the foremost question is which traits have diverged between populations to cause sexual isolation? For example, between-population mating probability is independent of differences between the sexes in color pattern, body size, and body shape (Nosil et al. 2002, Nosil & Crespi 2004). Preliminary data from gas chromatography and behavioral experiments, suggest that host-specific divergence in pheromones is involved (Nosil, Gries & Gries, unpub.), and further studies of olfactory communication are underway. Additionally, the potential role of courtship behavior is yet to be examined. Successful copulation in T. cristinae involves at least 2 distinct stages. First, the male must approach the female and attempt to mount her. Second, the female must allow the male to mount and copulate with her, as males cannot force copulation. A number of behaviors, such as antennal touching, kicking, and leg tapping by the males, are performed at both stages.
Other questions involve the forms of hybrid dysfunction that were most important for reinforcement. Theory predicts that the importance of different types of hybrid dysfunction depends on both the magnitude of hybrid dysfunction and on the genetic covariance between traits reducing hybrid fitness and mating preferences (Barton & Turrelli 1991, Servedio 2001). Finally, the host-associated forms of T. cristinae are unlikely to have achieved species status by any criterion, as indicated by only a 60% barrier to gene flow at the premating level (Nosil et al. 2006) and a general lack of neutral mtDNA differentiation between adjacent populations on different hosts due to ongoing gene flow (Nosil et al. 2003). Thus these host forms represent either an ongoing speciation event or population divergence that has reached equilibrium. A major question is why has complete speciation not occurred? Further studies of more divergent species within this genus may shed light onto the ecological, genetic or geographic factors driving the transition from a host race or ecotype to a species (Sandoval & Nosil 2005).
Parallels and contrasts with other orthopteroids
Orthopteroids have a long history as model systems in speciation research, and different groups exhibit patterns that both parallel and contrast those observed in Timema walking sticks. Here I aim to outline some well-studied groups, rather than to provide an exhaustive review. I first consider the relative roles of natural versus sexual selection, and then examine the evidence for reinforcement.
In T. cristinae, ecological divergence in host plant use promoted the evolution of sexual isolation. Divergence in host plant use has been shown to promote the evolution of sexual isolation in a few other nonorthopteran, herbivorous insect taxa (e.g., Neochlamisus leaf beetles: Funk 1998, Funk et al. 2002 for a review).
In contrast, ecological divergence does not seem critical for the evolution of reproductive isolation (and for sexual isolation in particular) in some Orthoptera. For example, closely related species of Laupala crickets on the Hawaiian Islands are ecologically similar and appear to differ primarily in calling song, a secondary sexual trait used in mate attraction (Mendelson & Shaw 2002, 2005). Differences between species in sexually-selected calling songs form the basis of sexual isolation between them, such that rapid speciation by sexual selection may have occurred. Another example of the evolution of sexual isolation by sexual selection comes from field crickets, where the key feature distinguishing the cryptic sister species Gryllus texensis and G. rubens is pulse rate, and the other conditions for speciation by sexual selection are met (Gray & Cade 2000).
An example where calling song may not be involved comes from montane grasshoppers in the genus Melanopus from the sky islands (Heald 1967) of western North America, where once again there is evidence for relatively rapid speciation (Knowles 2000). Melanopus species are quite morphologically similar, and differ primarily in male genitalia, a trait posited to be under sexual selection (a direct link between sexual isolation and genital divergence is yet to be established). Finally, divergence in mating signals among populations of the grasshopper Chorthippus parallelus is associated with some aspect of the colonization process (i.e., founder events), such that ecological divergence alone does not appear to drive the evolution of sexual isolation (Tregenza et al. 2000).
Perhaps even more striking than the examples above are cases of mosaic hybrid zones in grasshoppers and crickets. In these zones, closely related species do exhibit some habitat (i.e., ecological) specificity, yet reproductive isolation is not always strongly associated with ecological divergence.
Mosaic hybrid zones can be formed by long-distance dispersal into the previously unoccupied region between 2 advancing populations (Nichols & Hewitt 1994, Ibrahim et al. 1996). However, when a close correspondence between particular genotypes and discernible environmental patches is observed, this explanation is implausible (Barton & Hewitt 1985). In such situations, the patchiness of these zones is more readily attributed to some combination of active habitat preference on the one hand and divergent natural selection on the other.
Several such habitat-associated mosaic zones have been documented between species of grasshoppers or crickets (e.g., Gryllus, Rand & Harrison 1989; reviewed in Nosil et al. 2005a), but reproductive isolation is not always clearly ecologically driven. For example, in the mosaic hybrid zone between Allonemobius fasciatus and A. socius, genetic structure is associated with patches of different temperature (Howard 1986). However, the main reproductive barrier between these species appears to be conspecific sperm precedence, rather than sexual isolation (Howard et al. 1998). In some cases there is a partial role for ecology, but other factors also play an apparent role. In the mosaic hybrid zone between Chorthippus brunneus and C. jacobsi, genotypic composition is significantly, but weakly, associated with patches of different vegetation type and local habitat makes only a small contribution in explaining deviations in calling song and stridulatory peg number from clinal expectation (Bridle et al. 2001, Bridle & Butlin 2002). These examples suggest that ecological divergence is not critical for the evolution of sexual isolation between closely-related grasshopper and cricket taxa, at least for the taxa studied.
Why do some groups exhibit strong resource-associated (i.e., ecologically-based) divergence, whereas other groups diversify with less ecological differentiation? Two potential explanations concern the feeding lifestyle of different groups and the types of traits used in mate choice. An explanation for differences among groups is offered by the ‘grazer-parasite hypothesis’; difference among groups could be a function of how closely associated the insects are with their food resources (reviewed by Thompson 1988). In ‘grazer’ species, individuals must move between two or more plant individuals to complete their development. ‘Parasites’ are species where individuals complete development on a single plant individual. It might be expected that ‘parasites’ (such as Timema) are more prone to speciation via resource-associated divergence, due to their closer association with particular food items. Thus perhaps many Orthoptera are not tied closely enough to a specific food resource to have the evolution of reproductive isolation closely associated with resource divergence. Notably, recent molecular studies have shown genetic lineages associated with host plants or ecological features (i.e., host-associated divergence) even in polyphagous, ‘generalist’ grasshoppers (Dopman et al. 2002, VanDyke et al. 2004, Sword et al. 2005). Further studies examining the strength of association with resources in relation to the causes of speciation will likely prove insightful. The prediction is that groups more closely associated with their food resources will be more likely to evolve sexual isolation via resource-associated natural selection.
Another explanation for differences among groups considers the types of traits involved. Once again a clear prediction can be made: the evolution of sexual isolation by natural selection is only expected when traits conferring sexual isolation are under natural selection (i.e., are inherently ‘ecological’). Calling song is inherently sexual (being used to attract mates), such that speciation in singing insects need not involve ecological divergence (a similar argument could be made for other sexual characters such as genitalia). Interestingly though, some models of speciation via sexual selection include a role for natural selection (Lande & Kirkpatrick 1988, Schluter 2000). For example, the ‘sensory drive’ hypothesis predicts that mating signals will evolve in correlation with aspects of the environment (Endler 1992, Boughman 2002). Calling song in birds and frogs has been shown to be affected by the ecological environment (Morton 1975, Ryan et al. 1990, Slabbekoorn & Smith 2002, Seddon 2005) and it would be interesting to consider the role of ecology in affecting the evolution of song in singing Orthoptera. For example, singing Orthoptera need to avoid predation such that predation regimes could affect the evolution of song. Moreover, different types of habitats may vary in their sound transmission properties, imposing habitat-specific selection on signaling systems. Most of the examples discussed exhibit contrasting results to those observed in T. cristinae, and some potential explanations have been outlined. However, there are some parallels as evidenced by the examples of host-associated genetic divergence and habitat-specific differentiation in hybrid zones.
I now turn my attention to reproductive character displacement and reinforcement. As noted by Howard (1993), reproductive character displacement among the singing Orthoptera has played a large role in arguments over the likelihood and importance of reinforcement. Walker (1974) noted that there were few solid cases of reproductive character displacement in male calling song in Orthoptera, but also pointed out that few studies were designed to test for such a pattern. Moreover, displacement could occur in female preferences even if male traits are not displaced, i.e., reinforcement occurs but is detectable only if female preferences are examined (e.g., Hobel & Gerhardt 2003). Otte (1989) reported several cases of male calling song displacement in sword-tail crickets of the Hawaiian islands. Yang and Gerhardt (2006) report similar patterns of song displacement in the southern wood cricket (Gryllus fultoni). In contrast, Veech et al. (1996) found no evidence for character displacement in a detailed study of male song in ground crickets (see also Gregory et al. 1998). Likewise, studies of field crickets (Gryllus texenis and Gryllus rubens) found no evidence for reproductive character displacement (Gray & Cade 2000, Izzo & Gray 2004). Thus the evidence for reinforcement in singing Orthoptera appears equivocal.
The results from T. cristinae suggest that the likelihood and effects of reinforcement are highly dependent on relative population size and migration rates. Variation in the occurrence of reinforcement might be explained by differences among groups in habitat structure (i.e., geographic arrangement of populations), demography, and dispersal characteristics.
In T. cristinae, ecological divergence and reinforcement interact to drive the evolution of sexual isolation. In contrast, gene flow between populations can erode divergence such that both selection and gene flow are important determinants of the evolution of sexual isolation. As discussed, similar and contrasting patterns can be seen in other orthopteroids. This variation among groups might reflect the degree to which different groups are intimately associated with their food resources, the types of traits used in mate choice, and how the geographic arrangement of populations affects the opportunity for reinforcement.
B.J. Crespi and C.P. Sandoval have been integral to the ongoing research on speciation in Timema walking sticks. H. Song provided comments on an earlier version of the manuscript. I thank T.E. Reimchen, A. Mooers, D. Schluter, J. Endler, R. Vos, J. Joy, C. Parent, M. Forister, M. Kirkpatrick, F. Breden, D. Ortiz-Barrientos, H. Rundle, D. Funk and S. Springer for discussion. D. Mclaren, B. Mickelson, M. Vankoeveringe, T. Luchin, M. Javid, T. Leversage, H. Brackett, D. Wick, J. Smith, C. Paul, M. Zheng, and E. Lopez provided field and technical assistance. J. Endler (UCSB) kindly provided lab space. Financial support was provided by the Natural Science and Engineering Research Council of Canada and the Entomological Society of Canada.