A NOVEL PREFERENCE FOR AN INVASIVE PLANT AS A MECHANISM FOR ANIMAL HYBRID SPECIATION

Authors


Present address: MIDI Labs, Inc., 125 Sandy Drive, Newark, Delaware 19713.

Present address: University of Florida, Department of Zoology, 223 Bartram Hall, Gainesville, Florida 32611

Abstract

Homoploid hybrid speciation—speciation via hybridization without a change in chromosome number—is rarely documented and poorly understood in animals. In particular, the mechanisms by which animal homoploid hybrid species become ecologically and reproductively isolated from their parents are hypothetical and remain largely untested by experiments. For the many host-specific parasites that mate on their host, choosing the right host is the most important ecological and reproductive barrier between these species. One example of a host-specific parasite is the Lonicera fly, a population of tephritid fruit flies that evolved within the last 250 years likely by hybridization between two native Rhagoletis species following a host shift to invasive honeysuckle. We studied the host preference of the Lonicera fly and its putative parent species in laboratory experiments. The Lonicera fly prefers its new host, introduced honeysuckle, over the hosts of both parental species, demonstrating the rapid acquisition of preference for a new host as a means of behavioral isolation from the parent species. The parent taxa discriminate against each other's native hosts, but both accept honeysuckle fruit, leaving the potential for asymmetric gene flow from the parent species. Importantly, this pattern allows us to formulate hypotheses about the initial formation of the Lonicera fly. As mating partners from the two parent taxa are more likely to meet on invasive honeysuckle than on their respective native hosts, independent acceptance of honeysuckle by both parents likely preceded hybridization. We propose that invasive honeysuckle served as a catalyst for the local breakdown of reproductive isolation between the native parent species, a novel consequence of the introduction of an exotic weed. We describe behavioral mechanisms that explain the initial hybridization and subsequent reproductive isolation of the hybrid Lonicera fly. These results provide experimental support for a combination of host shift and hybridization as a model for hybrid speciation in parasitic animals.

Speciation is the key process underlying biological diversity. In animals, speciation is almost always described as the split of an ancestral species into two new taxa. The alternative—hybrid speciation—during which hybridization between two species gives rise to a third species, is rarely described in animals (Coyne and Orr 2004). Hybrid speciation can occur in two ways. One is polyploid hybrid speciation, which results in a new hybrid taxon that has a different chromosome number than its parents, thereby attaining instantaneous reproductive isolation (Bullini 1994). The other is homoploid hybrid speciation during which no change in chromosome number occurs. Well-documented cases of homoploid hybrid speciation are rare in both animals and plants. It is likely that ecological divergence in general promotes hybrid speciation, but the particular ecological conditions facilitating homoploid hybrid speciation have only been studied in detail in some plant species and remain otherwise poorly understood (Gross and Rieseberg 2005). Understanding the ecological conditions for hybrid speciation is crucial for evaluating its importance in animal evolution. Is hybrid speciation indeed a rare mode of speciation because the necessary ecological circumstances are rarely found in nature? Or, are favorable conditions common, suggesting that homoploid animal hybrid speciation has been overlooked due to methodological difficulties and an inherent bias in zoology (Dowling and Secor 1997)?

In a previous study, we described the Lonicera fly, a recently discovered population of tephritid fruit flies that likely represents a case of homoploid hybrid speciation (Schwarz et al. 2005). The Lonicera fly is a member of the Rhagoletis pomonella species complex that comprises several taxa of fruit parasitic tephritid fruit flies, all of which are highly host-specific (Berlocher 2000). This new population was formed by hybridization between the blueberry maggot, R. mendax, and the snowberry maggot, R. zephyria, which infest blueberries and huckleberries (Vaccinium spp. and Gaylussacia spp.) and snowberries (Symphoricarpos spp.), respectively (Bush 1966; Schwarz et al. 2005). The hybrid Lonicera fly is, however, found only on nonnative species of honeysuckle from the Lonicera tatarica complex (Green 1966). Prior to our report (Schwarz et al. 2005), no infestation of Lonicera spp. has been previously reported in North America (Foote et al. 1993), and samples of native L. dioica berries that were collected in proximity (i.e., within flight range) to infested nonnative honeysuckle were not infested by tephritids (D. Schwarz, unpubl. data). The Lonicera fly therefore most likely formed in conjunction with a recent host shift to nonnative honeysuckle. While the analysis of multilocus genetic data demonstrates a hybrid origin of the Lonicera fly, it also shows that the Lonicera fly is not a “hybrid zone” maintained by continuous immigration from both parents, but rather is a self-sustaining population (Schwarz et al. 2005).

Hybrids are expected to be less competitive than their parent taxa in the respective parental habitats (Anderson 1948), and theoretical models have shown that one critical criterion for homoploid hybrid speciation is the ecological isolation of the new hybrid species from its parent taxa (Buerkle et al. 2000). In accordance with this prediction, plant hybrid species are found in novel and extreme habitats and have become ecologically adapted to their new environments (Rieseberg 1997; Rieseberg et al. 2003). In the case of the Lonicera fly, the new host, honeysuckle, could provide such a novel habitat, which would release it from parental competition.

The other necessary condition for the formation of a new hybrid species is reproductive isolation from its parent taxa (Buerkle et al. 2000). In homoploid plant hybrids, such as sunflower species, reproductive isolation is associated with chromosomal rearrangements, but it is also conceivable that the adaptation to a hybrid-specific habitat alone could result in reproductive isolation (Buerkle et al. 2000). The rapid evolution of reproductive isolation following a shift and adaptation to a newly introduced host has been documented in great detail for the closely related R. pomonella (Berlocher and Feder 2002). A similar mechanism could be at work in the evolution of the Lonicera fly, resulting in hybrid speciation without the need to invoke chromosomal rearrangements. Such a mechanism does not, however, exclude some role of chromosomal rearrangements or an interaction between chromosomal rearrangements and ecological adaptation (Rieseberg 2001; Feder et al. 2003a,b).

This unique combination of hybridization and host shift could therefore provide a robust ecological solution to the two problems of homoploid hybrid speciation: ecological and reproductive isolation (Mayr 1963). We have previously argued that the ecological conditions of the Lonicera fly system and resulting hybrid speciation should be common, because both host-specific parasites and opportunities for hosts shifts (e.g., by geographic range expansions) are common (Schwarz et al. 2005). To provide experimental evidence for this model, we examine in this study whether this ecological scenario has indeed resulted in reproductive and ecological isolation by testing the host choice of the Lonicera fly and its parents.

Like other specialist herbivores, Rhagoletis species have evolved behaviors and sensory adaptations that allow them to find their specific host plants (Prokopy and Papaj 2000; Linn et al. 2003). The ability to discriminate among suitable and unsuitable hosts is critical for specialized insects, because these species often possess physiological or life-history adaptations for their preferred host (Filchak et al. 2000). Choosing the wrong host can result in the loss of fitness (Bierbaum and Bush 1990; Rausher 1993). In Rhagoletis, host choice is also central to reproductive isolation among the taxa because mating takes place on the host plant (Prokopy et al. 1971; Smith and Prokopy 1980; Smith and Prokopy 1982). In addition, host-independent mechanisms of assortative mating are only weak barriers to gene flow in the R. pomonella species complex (Smith 1986; Schwarz 2004). It is, therefore, assumed that host choice equals mate choice in the R. pomonella species complex. This combination results in an interaction between traits for host-specific fitness and traits for assortative mating that greatly facilitates sympatric speciation (Diehl and Bush 1989). The behaviors leading to host choice underlie this interaction. The study of host-choice behavior is therefore essential for understanding speciation in the Lonicera fly.

Host-choice behavior is also crucial for understanding whether a host shift preceded hybridization (“host-shift-first” hypothesis) or vice versa (“hybridization-first” hypothesis). If the parental taxa do not discriminate against honeysuckle, one cannot exclude that hybridization is the consequence of a host shift or at least oviposition mistakes. As stated above, host fidelity acts as the principal mechanism of reproductive isolation in Rhagoletis (Prokopy et al. 1971; Smith and Prokopy 1980). If R. mendax and R. zephyria colonize the same host, that is, honeysuckle, then this mechanism of reproductive isolation no longer exists and host-independent sexual isolation is too weak to prevent matings between R. mendax and R. zephyria on the new host (Smith 1986; Schwarz 2004). This would also be the case if both parents met on one of the two parental hosts due to an error in host choice. Whether hybridization preceded a host shift or vice versa depends on which error is more likely to occur—the choice of honeysuckle or the choice of the other parent's host fruit.

Previous studies have elucidated much of the host-choice behavior in Rhagoletis (reviewed in Prokopy and Papaj 2000). Most studies have concentrated on R. pomonella, but some work has also been conducted on other Rhagoletis taxa (Diehl and Prokopy 1986; Boller et al. 1998). Fruit odors have been shown to act as cues for discrimination between different host-fruit species. Rhagoletis pomonella and R. mendax show differences in their antennal sensory response to hawthorn and blueberry (Frey et al. 1992), and the host races of R. pomonella can distinguish between apple and hawthorn volatiles (Linn et al. 2003). Flies will not only seek out volatile blends of preferred host-fruit species (= preference sensu stricto), but will also actively avoid non-host-fruit volatiles when tested against a blank (Forbes et al. 2005). Due to the methodological difficulties of producing a true blank control in our tests of host choice via oviposition in real fruit, our observed “preference” could result from either preference sensu stricto or avoidance or both. On the fruit, flies explore the fruit surface, and it is thought that they use their tarsal receptors, antennae, and mouthparts to assess fruit quality based on chemical and physical characteristics (Prokopy and Papaj 2000). These behaviors in females are followed by probing the fruit with the ovipositor to obtain information about the chemical and physical properties of the fruit flesh (Prokopy and Papaj 2000) and, finally, oviposition. After oviposition, females mark the fruit surface with oviposition-deterring pheromone that deters oviposition both by that female and conspecific individuals (Prokopy and Papaj 2000).

In our experiments, we tested only female flies. A female's choice of where to place an egg most profoundly influences ecological isolation, because it determines the distribution of larvae, the life stage that exploits the resource. Male host-choice behavior will, however, affect nuclear gene flow just as much as female behavior. Previous studies on Rhagoletis (Prokopy et al. 1988; Linn et al. 2003) have found no differences in male and female host preference. For the sake of this study, we therefore assume that our measures of female host choice are representative for both males and females.

We used laboratory tests to examine the female host choice of the Lonicera fly, its two parent taxa, and the outgroup taxon R. pomonella to test two hypotheses: (1) The Lonicera fly has acquired a preference for invasive honeysuckle. It prefers honeysuckle over the native hosts of other Rhagoletis taxa. (2) Invasive honeysuckle represents a unique resource for the Lonicera fly. Parent taxa will prefer their native hosts over honeysuckle.

Materials and Methods

FLY POPULATIONS AND REARING PROCEDURE

Individuals of all four taxa studied, R. mendax, R. zephyria, R. pomonella, and the Lonicera fly, were collected in the summer and early fall of 2000, 2002, and 2003 by picking infested host fruit and rearing out the larvae, which pupated in the laboratory. Rhagoletis zephyria, R. pomonella, and the Lonicera fly were all collected in locations within 40 kilometers of State College, Pennsylvania. In Pennsylvania, R. mendax infests wild blueberries at low population densities. We could not collect it locally in sufficiently high numbers and had to rely on material from abandoned blueberry orchards in other states (Fennville, MI; East Wareham, MA; and Tinton Falls, NJ). We assume that any observed differences are general differences between taxa and not the result of local adaptation. The validity of this assumption is supported in part by the very weak geographic differentiation of R. mendax from both domestic and wild blueberries (Berlocher 1995). Pupae were put into diapause at 4°C for at least 5 months and then incubated at 22°C to induce eclosion (18:6 L:D photoperiod). Eclosed flies were kept under identical conditions, divided into weekly cohorts and sorted by sex to control age and mating status (Prokopy and Papaj 2000). Individuals were between 14 and 28 days old at the time of the behavioral assay. One week before the experiments, females and males were crossed at a ratio of 2:1 (in most cases 30 females and 15 males), and females remained in these cages until tested. Flies were transferred to a clean cage 1 day before an experiment (P. Gienapp and K. Heubel, pers. comm.).

HOST FRUITS

All host fruits used in the experiments were collected in Centre County, Pennsylvania. Highbush blueberry fruit—Vaccinium corymbosum, one of the hosts of R. mendax—was collected at two locations: Bear Meadows Natural Area (Rothrock State Forest) and from cultivated plants in Port Matilda, PA. Rhagoletis zephyria‘s host, snowberry, Symphoricarpos albus laevigatus, was obtained from a site on the Penn State campus, University Park, PA, and from a newly established planting in Port Matilda, PA. Honeysuckle fruit and the native host of R. pomonella—hawthorn, Crataegus sp.—were also collected from plants on the Penn State campus. Except for the blueberries and the Port Matilda snowberry site, at which there was no or rare infestation, respectively, fruit was covered with mesh bags on the host plant before fly emergence in the field to prevent infestation of experimental fruit. Fruit was picked shortly before the experiments and was continually handled with latex gloves that were frequently changed to avoid fruit contamination with semiochemicals (Prokopy 1972).

Preliminary studies of fly and host-plant phenology were conducted to determine the suitable fruit stage for oviposition by the flies (D. Schwarz, unpubl. ms.). In the case of honeysuckle fruit, flies were only present when the fruit was fully ripened. On hawthorn, the peak in flytrap catches occurred while red fruit was starting to soften. These fruit stages that coincide with maximum fly activity in the field were therefore used in the laboratory experiments. Because snowberries continually fruit, different fruit stages co-occurred during the time when R. zephyria was caught on traps. Field observations of female flies in June demonstrated that females will oviposit in green (diameter >4 mm) fruit, leading us to use this fruit stage in all no-choice experiments. Later in the season, however, we also observed females using white fruit for oviposition in the field leading us to compare R. zephyria's response to green and white snowberry fruit in preliminary choice trials. These trials suggested that R. zephyria prefers white (diameter ca. 10 mm) snowberry fruit. We therefore used white fruit in three-way choice experiments with all taxa and, in addition, tested R. zephyria in a three-way choice set-up with green snowberry fruit. We detected R. zephyria's preference for white snowberry fruit after we had completed almost all no-choice replicates involving snowberry fruit. A shortage of time and material prohibited repeating these replicates with white snowberry fruit. Despite this difference in snowberry ripening stage, the response of R. mendax and the Lonicera fly proved to be consistent between no-choice and choice experiments (see Results and Discussion). Previous studies had shown that R. mendax will oviposit into ripening (turning from green to blue) blueberries in the field (Lathrop and Nickels 1932), but preliminary experiments demonstrated that R. mendax showed a stronger response to freshly ripened, blue and soft fruit in the laboratory. This latter fruit stage was used in all reported experiments.

OBSERVED BEHAVIORS

Female behaviors were recorded with event recording software (The Observer, Noldus, Wageningen, The Netherlands), generating time event tables that specified when and for how long a predefined behavioral state was observed. Only behaviors displayed on fruit were recorded. The behavioral states fall into three categories.

While searching the fly walks over the fruit with its wings tilted 45° and lifted above the abdomen. During this behavior, a fly appears to examine fruit, and searching always precedes oviposition behavior. Oviposition behavior falls into two distinct behaviors. The first is ovipositor probing during which the female inserts her ovipositor into the fruit. This behavior is part of assessing the fruit quality for oviposition and does not necessarily result in oviposition (Prokopy and Papaj 2000). However, if ovipositor probing is followed by the second behavioral type, marking, the female laid an egg in the host fruit (Prokopy 1972). While marking, the female walks swiftly around the fruit with her everted ovipositor touching the surface and applies oviposition-deterring pheromone that serves as a signal that a specific fruit has received an egg. When not searching, probing or marking, females engage in general behaviors (e.g., cleaning, resting, etc.).

NO-CHOICE HOST-ACCEPTANCE TESTS

Experimental procedure

To test the acceptance of different host fruits by females under no-choice conditions, a single naïve (i.e., without previous fruit contact as an adult) female was released on an experimental array of a single host-fruit species. All of the 16 possible fly/fruit combinations between Lonicera fly, R. mendax, R. zephyria, and R. pomonella and their respective host fruits honeysuckle, blueberry, snowberry (green stage, >4 mm), and hawthorn were tested with the exception of R. pomonella on blueberry fruit. Multiple fruits were used to control for differences in individual fruit quality.

The experimental arena consisted of a glass plate (220 × 220 mm) with 14 fruits arranged in a hexagonal fashion with each fruit being equidistant (30 mm) from its six neighbors. Six fluorescent plant lights (P40PL/AQ 40W, General Electric, Louisville, KY) and two greenhouse lights (Lucalox 400W high pressure sodium bulbs, General Electric, Louisville, KY) provided 4000 lux of light at the center of the arena, and the temperature at this point was 31°C. Flies were moved to the experimental area for acclimation before the experiments, which were performed between the seventh and 11th hour of the insects' photoperiod (Katsoyannos et al. 1980).

An observation was started when a fly walked out of the aspirator onto a fruit in the center of the array and “recognized” the fruit. This meant that the fly had to spend enough time on the fruit to allow us to start the observation with the event recording software (ca. 3–5 s). This definition excludes instances in which the fly just used the fruit as a stepping-stone for flying off to the cage wall. If a fly did not respond with fruit recognition for three consecutive trials, it was placed back in the culture cage to be used in a later replicate. It was assumed that such uncooperative flies, even though they spent a brief amount of time in the experimental arena, did not acquire experience with fruit that could influence their behavior in a successful replicate (Prokopy et al. 1982). Each female was observed until it flew or walked off the experimental arena or the maximum observation time of 15 min passed. Individual females were used only in one experimental replicate. Any fruit that experienced probing or marking was immediately replaced with fresh fruit after that replicate. The entire array of fruit was exchanged after every five replicates. The fly taxa were alternated between replicates to correct for possible differences in environmental conditions or host-fruit quality. For the sample sizes at which each fly/fruit combination was tested, see Figure 1.

Figure 1.

Fruit acceptance under no-choice conditions. (A) Lonicera fly, (B) Rhagoletis mendax, (C) R. zephyria, and (D) R. pomonella. The proportion of females that probed host fruit at least once during an observation is shown in grey. The proportion of females showing no oviposition behavior is shown in white. The response proportions in each fly taxon (= row) were compared by pairwise Fisher's exact tests. Bold lower-case letters indicate significant differences after correction for multiple comparisons. Brackets indicate nonsignificant comparisons with P-values <0.15. The R. mendax/hawthorn comparison was excluded from significance testing due to small sample size. Sample sizes are indicated at the bottom right of each graph.

Measure of fruit acceptance and statistical analysis

If a fly probed a fruit, it also marked the fruit in the majority of observations. We therefore limited our measure of fruit acceptance to the proportion of females that probed a specific host-fruit species at least once during an observation. Fruit acceptance by each fly taxon was compared in a pairwise fashion in 2 × 2 contingency tables using Fisher's exact test. For these comparisons (as for all other analyses in this study) an alpha of 0.05 was set as statistically significant and adjusted by serial correction for multiple comparisons (Holm 1979).

THREE-WAY CHOICE TESTS

Experimental procedure

For the three-way choice experiment, a naïve female of the Lonicera fly or its two parental species was released on a hexagonal array consisting of two blueberry, two honeysuckle, and two white snowberry fruits. In addition R. zephyria was also tested on an array in which the choice consisted of green snowberry, honeysuckle, and blueberry fruit. Preliminary three-way choice trials had shown that it was more practical to use a closed arena than the open arena of the no-choice experiment. The closed arena consisted of a polystyrene petri dish (15 mm deep by 100 mm diameter) and was exposed to the same lighting conditions as the no-choice arena (see above). Two fruits each of three different host-fruit species were placed in alternating order on the vertices of a hexagon with 30 mm long edges. To avoid effects of fruit order, the sequence of the fruits was changed every five replicates when the entire set of fruit was exchanged for fresh fruit. If a test fruit experienced oviposition behavior, it was immediately replaced after the end of the replicate. In addition, the dish was turned 45° before each trial to randomize any effects of lighting or temperature in the experimental area. To start a replicate, a mated, naïve female (see above) was introduced through a hole in the lid of the petri dish that represented the center of the hexagon formed by the experimental fruit. Female behavior was recorded for a period of 15 min. Due to limitations on fly material, individuals that did not touch fruit during the experimental period were re-used in a three-way choice experiment after a rest period of at least 1 day. All females that came into contact with fruit were excluded from future replicates. As in the no-choice test, the experiments were conducted during the seventh to 11th hour of the flies' 16-h photoperiod.

Measure of host choice and statistical analysis

In the three-way choice experiments, both frequency of oviposition in the three host-fruit types and the allocation of search time on the three hosts were used as measures of preference. Only behavior that occurred before the first fruit was marked with oviposition-deterring pheromone was included in the analysis. Females will avoid marked fruit (Prokopy 1972), which leads to a skewed ratio of the different host-fruit types after the first fruit has been chosen for oviposition. If no fruit was marked during a replicate, the entire duration of the observation was used for data analysis. Oviposition behavior was measured by the number of females from each fly taxon that marked blueberry, snowberry, or honeysuckle fruit as their first choice. We compared the observed distributions among the three taxa with pairwise log-likelihood ratio contingency tests. In addition, we conducted a separate log-likelihood ratio contingency test to compare R. zephyria's response to arrays containing white and green snowberry along with blueberry and honeysuckle. We adjusted α for the four tests to 0.0125. The second measure of host-fruit choice was the proportion of the total observation time a female searched a particular host-fruit species. This standardization of the absolute search time became necessary because total observation time differed between replicates depending on if and when the first fruit marking occurred. As a female allocated on average only ca. 10% of the observational time to searching during a single replicate, we regarded search times on the different hosts as independent. The search data for each fly taxon were therefore analyzed by analysis of variance (ANOVA) followed by Tukey's test for multiple comparisons where applicable.

Results

NO-CHOICE HOST-ACCEPTANCE TESTS

Under no-choice conditions the Lonicera fly showed a statistically significant preference for honeysuckle fruit (Fig. 1A). It probed honeysuckle fruit in 50% of the replicates, but oviposition behavior was shown by only 20% and 10% of the females that were tested on hawthorn and snowberry fruit, respectively. No Lonicera fly females probed blueberry fruit (Fig. 1A). Unlike the Lonicera fly, the described taxa did not show different levels of fruit acceptance but either accepted or rejected a given fruit species. Both parent taxa probed honeysuckle fruit to the same degree as their native host-fruit species but did not probe each other's host-fruit species (Fig. 1B,C). In contrast, the outgroup taxon R. pomonella did not probe honeysuckle fruit and also discriminated against snowberry fruit (Fig. 1D). The response of the described species is dichotomous, but while R. zephyria and R. pomonella females probed their native fruit in ca. half of the replicates, R. mendax accepted its native host in only 20% of the observations (Fig. 1B). Hence, the difference between R. mendax's response to its native fruit, blueberry, and to R. zephyria's host, snowberry, is not statistically significant (the adjusted α for multiple tests is 0.017) but the probability that the observed difference was due to chance alone was small (P= 0.051). Rhagoletis zephyria also discriminated against R. pomonella's native host hawthorn, whereas R. mendax probed hawthorn fruit (Fig. 1B). Because this latter combination was tested in only eight replicates, it represents rather anecdotal information and is therefore excluded from significance testing. We did not test R. pomonella females on blueberry fruit, but a previous study found that R. pomonella did not oviposit in blueberries under no-choice conditions (Diehl and Prokopy 1986).

THREE-WAY CHOICE TESTS

The Lonicera fly spent, on average, ca. 1.7 times more time examining honeysuckle fruit than either of the two parental taxa in the three-way choice tests (marginally significant with P= 0.053, ANOVA, Fig. 2A). Rhagoletis mendax preferred its native blueberry over both honeysuckle and snowberry fruit with a three-fold higher allocation of its search time (Fig. 2B). Rhagoletis zephyria spent similar proportions of its search time on its native snowberry fruit and on honeysuckle fruit but discriminated against blueberry fruit with a ten-fold difference in the proportion of search time (Fig. 2C). However, when white snowberry fruit was replaced with green snowberry fruit R. zephyria preferred honeysuckle fruit over both the nonripe state of its native host-fruit species and blueberry fruit (Fig. 2D).

Figure 2.

Allocation of search behavior under three-way choice conditions. (A) Lonicera fly, (B) Rhagoletis mendax, (C) R. zephyria, and (D) R. zephyria in tests containing green snowberry fruit. Bar graphs show the mean search time on each host-fruit species as a proportion of the total observation time (±SE). Snow, snowberry; hon, honeysuckle; blue, blueberry. Letters indicate significant differences in Tukey's test for multiple comparisons. P-values were obtained by a one-way analysis of variance (ANOVA). N, total number of females tested.

Only a small proportion of the females that were tested in the three-way choice test (8–24%) marked fruit (our indicator of successful oviposition, see Materials and Methods), resulting in small sample sizes for the comparison of marking preference in the three-way choice (Fig. 3). Nevertheless, the differences in host discrimination were so pronounced that they were statistically significant despite the small sample sizes. As in the no-choice experiment, we also observed the Lonicera fly's preference for honeysuckle under three-way choice conditions. It chose honeysuckle fruit more than three times as often as either of the parental host fruits for marking (Fig. 3A). Four out of five R. mendax individuals picked blueberry fruit as their first choice, whereas one R. mendax female marked snowberry fruit and no R. mendax females chose honeysuckle fruit for marking in the three-way choice test (Fig. 3A). Rhagoletis zephyria never marked blueberry fruit as its first choice. Eight R. zephyria chose white snowberry fruit and three chose honeysuckle fruit for marking (Fig. 3A). A comparison of the marking choices by pairwise contingency tests showed significant differences between R. mendax, the Lonicera fly, and R. zephyria (Fig. 3). When white snowberry was replaced by green snowberry in the three-way choice experiment, R. zephyria did not mark blueberry, but R. zephyria selected honeysuckle fruit three times more frequently than snowberry fruit as first choice for marking (Fig. 3B, P= 0.111 when compared to R. zephyria on white snowberry fruit by a contingency test).

Figure 3.

Proportions of host-fruit species as first choice for fruit marking under three-way choice conditions. (A) Experiments with white snowberry fruit in choice array. (B) Results for Rhagoletis zephyria when white fruit was exchanged for green fruit. Black, blueberry; grey, honeysuckle fruit; white, snowberry fruit. Bold lower-case letters indicate significant differences in log-likelihood ratio contingency tests after correction for multiple comparisons. Sample sizes are indicated at the bottom right of each graph.

Discussion

We will first use the results of the no-choice and choice experiments to discuss our two hypotheses regarding the host preference and isolation of the Lonicera fly before addressing the question of whether a host shift preceded hybridization or vice versa.

HYPOTHESIS 1: THE LONICERA FLY PREFERS INVASIVE HONEYSUCKLE

The Lonicera fly prefers honeysuckle fruit under both no-choice and choice conditions supporting Hypothesis 1. Both parent taxa accept honeysuckle fruit under no-choice conditions in similar proportions as their native host-fruit species (Fig. 1B,C). The acceptance of honeysuckle as such is therefore not new to the Lonicera fly, but the hybrid origin taxon displays a preference for its new host that is absent in its parent species. The magnitude of the new preference for honeysuckle fruit is even more evident in the choice experiments, in which both parents prefer their native host-fruit species for oviposition, whereas the Lonicera fly prefers honeysuckle fruit (Fig. 3). Rhagoletis zephyria appears to regard unripe, green snowberry fruit as an inferior oviposition substrate when also exposed to ripe honeysuckle fruit (Figs. 2D, 3B). Our use of green snowberry fruit in the no-choice experiments might therefore have resulted in a lower response of the Lonicera fly to snowberry than we would have observed had we used white fruit. However, R. zephyria accepts green snowberry fruit ca. four times more often than the Lonicera fly in the no-choice test (P= 0.013, Fisher's exact test, Fig. 1A,C), suggesting a generally lower level of snowberry acceptance by the Lonicera fly as compared to R. zephyria. In addition, the Lonicera fly's preference for honeysuckle fruit under no-choice conditions is consistent with the results of the three-way choice test in which we used white fruit. As the tested flies were reared from field collected larvae that matured in their native host fruit, we cannot exclude larval conditioning as an influence on the host choice of adult females. But previous studies on R. pomonella demonstrated a genetic basis of host-odor discrimination by experimental crosses (Dambroski et al. 2005) and failed to detect a significant influence of the larval environment on adult preference for host volatiles by rearing the offspring of different host races in apples as a common host (Linn et al. 2003; Dambroski et al. 2005).

The observed preference of the Lonicera fly for honeysuckle fruit validates the proposed model that the host shift has resulted in the adaptation of the hybrid origin population to its new host. This preference for honeysuckle is consistent with predictions from models of adaptive speciation following a host shift (Diehl and Bush 1989). It is conceivable that early generation hybrids could show an immediate preference for honeysuckle (see discussion below), but population genetic evidence shows that the Lonicera fly is a self-sustaining population and not an active hybrid zone (Schwarz et al. 2005). The preference of the Lonicera fly for honeysuckle—regardless of its underlying genetic basis—is therefore a phenotype that is maintained by the Lonicera fly population and not the product of recurrent parental hybridization.

In its preference for the new host, the Lonicera fly differs from the results of similar host-acceptance tests in the host races of R. pomonella. These experiments found that the derived apple race of R. pomonella does not prefer R. pomonella's original host hawthorn over introduced apples (Prokopy et al. 1988). The Lonicera fly therefore shows a pattern of host acceptance that is more similar to that of described species than to that of a described host race that arose without hybridization. Still, the Lonicera fly does not completely discriminate against other host fruits as the parent taxa and R. pomonella do in the no-choice experiment (with the exception of R. mendax accepting hawthorn, Fig. 1A–D), and in the three-way choice experiment, the Lonicera fly is the only taxon that accepts all three taxa as a first oviposition choice. In addition, the preference for honeysuckle shown in the allocation of search time on the three host-fruit species is less pronounced than the preference of the parent taxa for their native host fruit. The phenotypic signature of hybridization is therefore still evident in the Lonicera fly, which combines the host acceptance spectra of both parent species. This suggests that hybridization between R. mendax and R. zephyria resulted in increased phenotypic variation. There is both empirical and theoretical evidence (Rausher 1993; Futuyma et al. 1995) that suggests that a lack of genetic variation constrains the ability of specialist herbivores to acquire a new host plant. Although our results indicate that the parents do not lack the necessary phenotypic variation to, at least initially, colonize honeysuckle (Fig. 1B,C, see discussion below), it is an open question whether the mixing of R. mendax and R. zephyria genes could have facilitated and accelerated specialization to honeysuckle. Hybridization should particularly increase the variability of quantitative traits (Barton 2001), and transgressive segregation is a well known consequence of hybridization (Rieseberg et al. 1999). Results of quantitative genetic models have been used to argue that this increase in variation should increase the likelihood of sympatric and parapatric speciation (Seehausen 2004) and the adaptive role of hybridization during speciation events has been experimentally demonstrated in plants (Rieseberg et al. 2003). It is unclear how the new preference for honeysuckle evolved. Is it the product of rapid (Thompson 1998) but gradual adaptation to the new host following the initial host shift and hybridization within the last 250 years (Rehder 1947) or did it form as an immediate consequence of hybridization? If early generation hybrids do indeed prefer honeysuckle, then Lonicera fly would present a system for ad hoc hybrid speciation comparable to allopolyploid hybrid speciation that can instantaneously result in reproductive isolation (Bullini 1994). Both experimental studies using artificial F1 and other early generation hybrids and explicit quantitative genetic models are necessary to validate this effect in the Lonicera fly and other host-specific parasites.

The preference of the hybrid origin Lonicera fly for its new host honeysuckle is consistent with a model of homoploid hybrid speciation solely by prezygotic isolation following the adaptation to a new habitat (Buerkle et al. 2000). Given the ecological scenario we do not have to invoke postzygotic isolation as a consequence of chromosomal incompatibilities to explain the reproductive isolation of the Lonicera fly. This does not, however, preclude a role for chromosomal rearrangements in the evolution of the Lonicera fly. There is indirect evidence for chromosomal differences between R. mendax and R. zephyria (Schwarz et al. 2005) and the role of these rearrangements in the formation of the Lonicera fly should be the subject of future study.

Many studies describing homoploid animal hybrid species are limited to the inference of hybrid origin from morphological and genetic evidence, but some provide experimental support for proposed mechanisms of animal hybrid speciation. A hybrid origin Heliconius butterfly species is isolated from its two progenitors by assortative mating that is based on a unique wing color pattern that could be recreated from the parental taxa by a series of artificial crosses (Mavarez et al. 2006). Similarly, the hypothesized mechanism for the hybrid origin of the swordtail fish Xiphophorus clemenciae via sexual selection has been supported by mate choice experiments with artificial hybrids (Meyer et al. 2006). In swallowtail butterflies temporal isolation of a hybrid swarm occurred by delayed emergence supporting the hypothesis that the recently described species Pterourus appalachiensis could have arisen by this mechanism (Scriber and Ording 2005). Some chromosomal races of house mice in the Alps are thought to have arisen within hybrid zones between parental races and experiments have shown that F1 offspring from these chromosomal races show decreased fertility (Pialek et al. 2001). Two hybrid species in a complex of crater lake cichlids represent distinct ecotypes with unusual physiological and morphological adaptations (Schliewen and Klee 2004). Strong selection against immigrants as an isolating factor has been deduced from a population genetic study of contact zones between a hybrid sculpin lineage and one of its progenitors (Nolte et al. 2006). Compared to the numerous studies on mechanisms of isolation in uniparental animal speciation (Coyne and Orr 2004) our knowledge of homoploid hybrid speciation is limited.

HYPOTHESIS 2: INTRODUCED HONEYSUCKLE REPRESENTS A UNIQUE RESOURCE FOR THE LONICERA FLY

No-choice and choice experiments yield different results for the evaluation of this hypothesis. The results of the no-choice experiment prompt us to reject the hypothesis that honeysuckle is a unique resource for the Lonicera fly. Both parent species, R. mendax and R. zephyria, accepted honeysuckle fruit in similar proportions as their native host-fruit species. We regard the no-choice experiment as the conservative estimate of parental influx in the field because the conditions of the no-choice experiment provide a realistic model of what flies experience in nature. In most cases, females will encounter patches of the same fruit rather than true choice situations, even if the two different host plants grow adjacent to one another. As in nature, flies in our no-choice experiments were not forced to remain in a given patch of fruit. Once a fly flew or walked off the experimental arena we terminated our observation. Results from bioassays with R. pomonella further support our contention that the no-choice tests are biologically relevant and do not represent artificially forced conditions because R. pomonella, the outgroup taxon, did not accept honeysuckle while both parent taxa, R. mendax and R. zephyria, did.

The three-way choice experiments nevertheless show an innate preference of both parents for their native host-fruit species, even though this effect could be masked by environmental conditions in the field. These experiments provide us also with important insights about the relative roles of R. mendax and R. zephyria in the colonization of honeysuckle. Presented with a choice, R. mendax discriminated against honeysuckle fruit, measured both by search time and oviposition choice. In contrast, R. zephyria females examined snowberry and honeysuckle fruit for similar amounts of time and—even though it preferred snowberry—also oviposited in honeysuckle. This asymmetry between R. zephyria and R. mendax could mirror the greater relatedness between honeysuckle and snowberry, both of which are in the Caprifoliaceae, whereas blueberry belongs to the Ericaceae. Host shifts are thought to be more likely between chemically similar host plants (Becerra 1997). A study of host-fruit volatiles that elicit electrophysiological responses in Rhagoletis has indeed found that the volatile profiles of snowberry and honeysuckle fruit are more similar than are those of blueberry and honeysuckle fruit (Olsson 2005). The observed phenotypic asymmetry between R. mendax and R. zephyria and their respective host fruits corresponds to a greater genetic similarity between R. zephyria and the Lonicera fly that could be the result of asymmetric historic or on-going introgression (Schwarz et al. 2005).

If we accept that our no-choice experiments represent a more realistic representation of natural conditions, we are faced with the question of whether the absence of behavioral isolation in the laboratory corresponds to a lack of isolation in the field. The assignment of individual Lonicera fly multilocus genotypes revealed no individuals that could be classified as R. mendax, but showed several Lonicera flies with genotypes that were indistinguishable from R. zephyria. At present, we cannot distinguish whether these individuals are actual immigrants or R. zephyria-like genotypes that were reassembled by random mating (Schwarz et al. 2005). An apparent absence of parental immigrants in the field—despite the laboratory acceptance of honeysuckle by R. mendax and R. zephyria—could be a function of the spatial distribution of host plants in the landscape. A survey of the State College area (Pennsylvania) showed that the three hosts are not equally distributed throughout the landscape but form largely distinct patches. Honeysuckle is very abundant and borders both urban areas with planted snowberry and forest lands with native blueberry and huckleberry populations (D. Schwarz, unpubl. data). Infestation of honeysuckle fruit by the parent species could occur at the edges of the host-plant patches, and parental immigrants would not be detected in Lonicera fly samples, unless samples were taken in the direct vicinity of the parental hosts. The parapatric distribution of the abundant honeysuckle and the parental hosts at the landscape scale could provide spatial refugia from parental competition. Another, though not mutually exclusive, possibility for ecological isolation is fitness trade-offs on the different hosts. Despite the parent species' acceptance of honeysuckle, both R. mendax and R. zephyria larvae could be less fit on honeysuckle fruit than Lonicera fly offspring. Both R. mendax and R. zephyria individuals can complete larval development in honeysuckle fruit (D. Schwarz, unpubl. data), but detailed transplant studies are necessary to detect fitness trade-offs. Such trade-offs have been observed between R. mendax and R. pomonella (Bierbaum and Bush 1990).

DID INTRODUCED HONEYSUCKLE CATALYZE THE HYBRIDIZATION OF R.MENDAX AND R. ZEPHYRIA?

Under no-choice conditions, both R. mendax and R. zephyria accept honeysuckle fruit but not the native host fruit of the other parent taxon. The proportion of R. mendax females probing blueberry or honeysuckle fruit is low, resulting in small, but insignificant, probabilities that the discrimination against snowberry is the effect of chance alone (Fig. 1B). The second caveat is the use of green snowberry fruit for the no-choice tests. Given R. zephyria's preference for white over green snowberry fruit (Figs. 2 and 3) we cannot exclude that R. mendax would have accepted white snowberry fruit. Although R. zephyria accepts a large proportion of green snowberry fruit in no-choice trials, the ripening stage of snowberry remains a potential confounding factor. If the observed pattern holds up to further scrutiny in future experiments, it will allow us to draw an important inference about the initial formation of the Lonicera fly. Given this pattern, the “host-shift-first” hypothesis is a more parsimonious explanation than the “hybridization-first” hypothesis as R. mendax and R. zephyria flies are more likely to meet on honeysuckle than on either parental host. At the same time, there is no evidence for substantial gene flow between the two parents (Schwarz et al. 2005). This observation is consistent with the Lonicera fly's discrimination against its parental hosts that we observed in this study. Both findings make it unlikely that the Lonicera fly serves as a genetic bridge between the two parent taxa. The acceptance of honeysuckle fruit therefore represents a pre-adaptation that was present prior to the introduction of plants from the L. tatarica complex to North America and not a consequence of gene flow from the Lonicera fly. The introduction of invasive honeysuckle from the L. tatarica complex could have facilitated the breakdown of reproductive isolation between two native insect species, because it instantaneously provided a host acceptable to both R. mendax and R. zephyria. Such apparent promotion of hybridization between native insects could represent a novel mechanism by which introduced plants affect native communities (Strauss et al. 2006). Habitat alterations have long been regarded as an important stimulus for the formation of plant hybrids (Anderson 1948; Ellstrand and Schierenbeck 2000). Here we provided experimental evidence of how an anthropogenic habitat disturbance (the invasion by introduced honeysuckle) triggered animal hybridization. Such human induced habitat alterations have also been invoked to explain the original formation hybrids in other examples of animal hybrid speciation (Reich et al. 1999; Nolte et al. 2005).

BEHAVIORAL ISOLATION THROUGHOUT THE RHAGOLETIS POMONELLA SPECIES GROUP

The described species R. mendax, R. zephyria, and R. pomonella each discriminated against the hosts of the other two described species (Fig. 1). To date R. pomonella and R. mendax (Diehl and Prokopy 1986) and the host races of R. pomonella are the only Rhagoletis taxa that have been compared by experiments similar to this study. Here we provide experimental evidence that this previously observed pattern of host specialization also applies to other Rhagoletis species. Our finding confirms universal host specialization among R. pomonella, R. mendax, and R. zephyria that is predicted by taxonomic, genetic, and theoretical studies on the R. pomonella species complex (Bush 1966; Diehl and Bush 1989; Berlocher 2000). The only exception to this specialization in our experiments is the acceptance of hawthorn by R. mendax. This response had been previously observed (Diehl and Prokopy 1986). The same study also found that R. pomonella will not accept blueberry fruit, and they could not replicate R. mendax's acceptance of hawthorn under field conditions (Diehl and Prokopy 1986).

The described species' discrimination against each other's native host fruit in this study provided a baseline against which we could compare responses toward honeysuckle fruit and the behavior of the Lonicera fly. The motivation of Rhagoletis flies to discriminate against unsuitable oviposition substrates appears to be modified by the state of the individual and environmental conditions (Prokopy and Papaj 2000). In the field, different Rhagoletis taxa are known to occasionally “sample” fruit from plant species other than their native hosts and, under laboratory conditions, R. pomonella can be induced to lay eggs into a variety of fruits from different plant families (Bush 1966). The unambiguous responses displayed by the described Rhagoletis taxa toward native and nonnative host fruits therefore serves as a positive control for experiments involving honeysuckle fruit or the Lonicera fly. Given this control we can be confident that acceptance of honeysuckle fruit for oviposition by both parent taxa under no-choice conditions represents a biological effect and not an experimental artifact.

Conclusions

Our results document the rapid emergence of a fruit parasite's preference for a new host following host shift and hybridization within the last 250 years. This finding provides strong support for our model that a combination of hybridization and host shift can lead to homoploid hybrid speciation in animals. We could not provide evidence in the Lonicera fly system that behavioral mechanisms alone would result in decreased gene flow and competition from the parent taxa, as the parent species both accepted honeysuckle fruit under more realistic no-choice conditions. This finding calls for studies that test the role of spatial structure and host-specific fitness trade-offs to explain the—at least partial—isolation of the Lonicera fly demonstrated by population genetic analysis (Schwarz et al. 2005). The lack of parental discrimination against honeysuckle suggests, however, that a specialist parasite could fail to identify an introduced host as “wrong” habitat (Forbes et al. 2005). In animals with host fidelity as the most important mechanism of reproductive isolation, such new hosts could serve as meeting grounds for interspecific matings. Hybridization in native host specialists could be an important consequence of the increasing number of human-mediated introductions of exotic plants. We have no reason to believe that the ecological conditions of the Lonicera fly system are exceptional and rare. Instead, as both specialized animals (Price 1980) and range expansions of specialists or hosts, for example, by human introductions (Mack et al. 2000) or climate change (Parmesan and Yohe 2003), are common, there could be multiple opportunities for Lonicera fly-like hybrid speciation.

Associate Editor: M. Noor

ACKNOWLEDGMENTS

We thank Clinton Cario and Benjamin Matta for technical assistance, Jeffrey Feder, Sridhar Polavarapu (deceased), and Anne Averill for providing samples of R. mendax, and Thomas Baker, Frank Hanson, Mohamed Noor, Wendell Roelofs, and John Tooker for comments on the manuscript. Partial funding for this study came from the National Science Foundation DEB-0343771.

Ancillary