Performance of Host-Races of the Fruit Fly, Tephritis conura on a Derived Host Plant, the Cabbage Thistle Cirsium oleraceum: Implications for the Original Host Shift

The thistle-infesting fruit fly Tephritis conura Loew (Diptera: Tephritidae) forms host races on the melancholy thistle, Cirsium hetewphyllum (L.) Hill (Asterales: Asteraceae) and the cabbage thistle, Cirsium olemceum (L.). Scop. Previous research indicates that the host shift occurred from C. hetewphyllum to C. oleraceum. In this paper we address whether the host shift involved physiological adaptations by studying oviposition acceptance and survival of the two host races on the derived host C. oleraceum. Performance differed significantly between host races. T. conura originating from C. oleraceum produced adults in 75% of all egg-laying trials in contrast to only 6.6% in T. conura originating from C. hetewphyllum. Population fitness components measured as a function of life-stage was linear decreasing for T. conura on C. oleraceum but stepwise for T. conura on C. heterophyllum. Low performance of T. conura on C. hetewphyllum was determined by low plant acceptance and high mortality during the larval stage, whereas hatching (at least one larva per batch) and pupae survival were not affected.


Introduction
Host-specific insects are estimated to represent 25-40% of all animal species (Zwölfer 1975;Bush and Butlin 2004). It has been proposed that many of these phytophagous insects have arisen via host shifts to a new plant with subsequent specialization (Bush 1975). The colonization of a new host theoretically requires only a genetic change in host preference (Dres and Mallet 2002). However, adaptations to different host-plant phenologies (e.g. Craig et al. 1993;Feder et al. 1997; Thomas et al. 2003) and host plant physiology may be required. Plants that differ in the amount of secondary metabolites (Roitberg and Isman 1992) and nutritional value (Häggström and Larsson 1995) can result in reduced growth and survival of the feeding larvae (Sharma and Norris 1991). Host-dependent larval performance has been demonstrated in numerous studies (e.g. Via 1991; Craig et al. 1993Craig et al. , 1997Katagura and Hosogai 1994), indicating that physiological adaptations play an important role in the course of a host shift. Distinguishing the causes responsible for divergence is important for reconstructing how host races evolve. If larvae develop on different plants equally well, only a change in host preference might have been the critical requirement to initiate the host shift. However, if insects from the ancestral host population were not physiologically adapted to the novel host, the initial shift would have been maladaptive but after adaptation perhaps physiological adaptations became a prerequisite for diversification.
A promising candidate for host-mediated speciation is the thistle-infesting fruit fly Tephritis conura Loew (Diptera: Tephritidae), which infests flower heads of several Cirsium (Asterales: Asteraceae: Cardueae) species (Zwölfer 1988) including Cirsium heterophyllum (L.) Hill and Cirsium oleraceum (L.) Scop. Genetic analyses have shown that gene flow among T. conura infesting C. heterophyllum or C. oleraceum is severely restricted (Seitz and Komma 1984;Diegisser et al. 2006a,b) and that host-specific morphological adaptations exist (Diegisser et al. 2007). In hostchoice experiments, T. conura significantly preferred to fly to and to stay on the host-plant species from which they had emerged (Komma 1990;Diegisser 2005). As T. conura mate exclusively on their host, host preferences cause assortative mating, which is a prerequisite for further differentiation. Preliminary experiments could show that T. conura have adapted behaviourally to the phenology of host plants; in the lab, T. conura on C. heterophyllum become significantly earlier sexually active than T. conura on C. oleraceum after hibernation, which coincides with plant phenology in the field (Diegisser 2005).
These findings provide evidence for two host races, i.e. partially reproductively isolated populations due to differential host specialization. Phylogeographic mtDNA data strongly suggest that C. heterophyllum is the ancestral host, from which flies shifted to C. oleraceum (Diegisser et al. 2006a). No study has yet examined whether T. conura's host shift from C. heterophyllum to C. oleraceum has involved physiological adaptations. In this paper, the results of an oviposition experiment are presented which estimates acceptance and performance of both T. conura host races on the derived host, C. oleraceum. By doing so, it is concluded that T. conura from the ancestral C. heterophyllum population may have adapted physiologically to C. oleraceum in the course of the host shift.

Biology of T. conura
The univoltine T. conura completes larval development in flower heads of several Cirsium species. As soon as the host plants start developing flower buds in late spring, T. conura imagoes reappear from unknown hibernation sites. After mating, which occurs exclusively on the host plants, females lay egg batches in young buds. Depending on the host plant, eggs are laid from the middle of May (C. heterophyllum) to the end of June (C. oleraceum). Within the developing flower heads, larvae feed on the florets (first instar) and the receptacle area (second and third instar). Larvae pupate about three weeks after oviposition; imagoes emerge three weeks after pupation.

Oviposition experiment
Rosettes were collected for experiments from C. oleraceum in the Bavarian Fichtelgebirge (Eimersmühle, N49°53.092/E011°36.975, 396m asl) at the beginning of May 2003. All plants (N = 100) were potted in 5L flowerpots and kept in a greenhouse. In the beginning of June, the rosettes started developing buds that could be used for oviposition experiments. Buds were formed on new stem growth in the rosette. In general, 5-10 apical buds developed together in a tight cluster. C. oleraceum produces buds continuously throughout the flowering season. Continuous bud development ensured that all flies were offered buds in the same phenological stage and of the same size.
T. conura were collected in the field as soon as they re-colonized the host populations after hibernation. Female T. conura were collected from C. heterophyllum at Bischofsgrün (N50°03.479/E011°46.588, 630m asl) at the end of May 2003. Female T. conura were collected from C. oleracum at Eimersmühle (see above) in the middle of June 2003. Preliminary experiments in 2002 showed that T. conura have already mated when first found in the field and that females kept in a cage would oviposit into the cage's gauze if no host plant was offered. These flies survived for +30 days after being caught.
T. conura infesting C. heterophyllum were collected about 10 days before the first C. oleraceum in the lab were ready for experimental trials while T. conura infesting C. oleraceum were caught two weeks later when plants in the lab were ready for experiments (see above). To standardize the experiments between the host races, all flies were kept in the lab for an average of 12 days before starting the experiments.
Experiments were performed by offering single females apical buds of a C. oleraceum plant. Flies and buds were enclosed in a gauze bag. Buds had a diameter between 6 -11mm diameter, which is the size of buds accepted by both host races in the field. Previous lab experiments showed that T. conura adapted to C. heterophyllum or C. oleraceum accept buds between 5.8-13.7mm and 3.4-11mm diameter, respectively (Eschenbacher 1982;Romstöck 1982). Buds are enveloped by small leaflets, which have to be pierced by the fly's ovipositor in order to deposit the eggs into the flower head. As a consequence, oviposition attempts can be inferred from tiny dark holes in the enveloping leaflets. Each plant was checked daily for such holes. If punctures were detected, females were removed two days later. If no holes had been found after a week, the trial was terminated.
Four weeks after removal of flies, flower heads were cut and placed separately in small jars covered with gauze. As T. conura pupates within three weeks after oviposition (see above) this transfer most likely did not affect the development of flies within a flower head. The jars were checked daily for emerging flies. After four weeks in a jar, each flower head was dissected to check for non-emergence of larvae and pupae. The total time involved from oviposition to flower-head dissection was eight weeks. This time period is sufficient to ensure quantitative emergence because T. conura completes larval development within six weeks.
The experiment estimated population fitness components at four life-stages. (1) Oviposition attempts, defined as the number of flies accepting buds. (2) The number of flies successfully producing larvae. The presence of larvae can easily be inferred from the damaged brownish/darkbrown colour of the receptacle area. Damage is not seen in non-viable oviposition attempts. (3) Survival from larvae to pupae was quantified as the number of flies producing pupae, and the number of pupae. (4) Survival from pupa to adult was estimated as the number of females producing adults and as the difference between emerging adults and the number of pupae per flower head. Individual mortality in the egg and larval stages were not estimated because the number of eggs and larvae could not be determined reliably without destructive or subsampling of the host materials.
(2) and (3) are therefore qualitative estimates of a fly's ability to produce at least one surviving offspring per egg batch. Between-species plant acceptance and survival were tested with independent 2 x 2 contingency table " 2 tests.

Discussion
The oviposition experiment revealed that T. conura adapted to C. heterophyllum exploited the derived host C. oleraceum significantly less well than T. conura adapted to C. oleraceum. This was determined by unwillingness to accept C. oleraceum buds and by high mortality during the larval stage, whereas survival of at least one egg to hatching after oviposition attempts and pupae to imago were not affected. This shows that T. conura adapted to C. oleraceum were significantly better adapted than T. conura adapted to C. heterophyllum flies to the larval environment provided by C. oleraceum flower heads. The two stages not affected by mortality can be viewed as stages that are not directly involved in interactions with the host plant. Different performance may be biased if the host races responded differently to the induced oviposition delay or if the developmental buffer time was not long enough to ensure survival of T. conura adapted to C. heterophyllum on the Figure 1. The percentage of Tephritis conura adapted to Cirsium oleraceum (N = 28, grey bars) and T. conura adapted to C. heterophyllum flies (N = 45, black bars) accepting C. oleraceum inflorescences for oviposition and producing offspring depicted as a function of proceeding life-stage. Note that survival decreases linearly for T. conura adapted to C. oleraceum and decreases step-wise for T. conura adapted to C. heterophyllum. novel host. If these factors did influence the results, they also imply that the host races will respond differently to conditions experienced in the field; conditions that will be of advantageous to the T. conura adapted to C. oleraceum host race.
The inclusion of a novel host may involve "host-range expansion" where an insect population retains the ability to utilise old hosts (Hare 1990). Alternatively, the insect may be unable to adapt simultaneously to both, resulting in a "host shift" with host plant-associated fitness tradeoffs (Via, 1990). To evaluate which of these performance scenarios influence the evolution of T. conura host-races, reciprocal experiments are needed. Regrettably, despite great efforts, we have not been able to bring C. heterophyllum rosettes to develop buds in the greenhouse to perform the reciprocal experiments (for details see Diegisser 2005). Attraction tests (Diegisser 2005) showed that host races are attracted significantly more to their own host plants, implying that acceptance of C. heterophyllum flower heads will be significantly lower for T. conura adapted to C. oleraceum. Data for naïve flies showed that female T. conura adapted to C. oleraceum preferred their own plants, which required C. oleraceum to become sexually active (Diegisser 2005). This suggests that female T. conura adapted to C. oleraceum will not accept C. heterophyllum in the field.
For the life-stage with performance differences within the flower head, larvae to pupae, the field evidence shows that heterophyllum flies are most fit on C. heterophyllum, but, unfortunately, no field data is available for the reciprocal mortality of T. conura larvae adapted to C. oleraceum on C. heterophyllum. Field data show that T. conura adapted to C. heterophyllum do not experience significant intrinsic egg and larval mortality on C. heterophyllum (Romstöck 1987;Romstöck-Völkl 1990). The only factors significantly reducing survival during development seem to be parasitoids and/or predators and bud abortion (Romstöck-Völkl 1990) but even in the presence of predators performance is much better on the original host with mean number of imagoes per infested flower = 5.41 (N flowers = 78) (T. Diegisser and J. Johannesen, unpublished observations) vs. 3.67 in this study. The low number of surviving T. conura adapted to C. heterophyllum on C. oleraceum is likely not a consequence of increased competition in smaller C. oleraceum flower heads, as evidenced by larval capacity in Table 1. Also, Romstöck-Völkl (1990) calculated the receptacle surface needed per larvae in C. heterophyllum buds to be 6-8 mm 2 , which would allow 8-10 adult T. conura adapted to C. heterophyllum to develop in C. oleraceum. Thus, we conclude that the T. conura population adapted to C. heterophyllum is best adapted to C. heterophyllum. Furthermore, evolved avoidance mechanisms of T. conura adapted to C. oleraceum towards C. heterophyllum limits gene flow significantly, but data on performance within flower head are still missing. Despite low performance of T. conura adapted to C. heterophyllum on C. oleraceum, the experiment showed genetic variation for survival on the wrong, low quality host. Significant within-population variability in performance has been found in many herbivore populations (see Via 1990), including tephritid host races (Craig et al. 1997).
As selection on performance is most likely very strong, pre-adapted genotypes would spread rapidly and dominate the gene pool of C. oleraceum infesting flies within a few generations. Adaptation may be especially rapid when performance is determined by a few genes with large effects (Sezer and Butlin 1998). But why should T. conura shift to a low quality host plant? Cirsium heterophyllum and C. oleraceum have different temperature requirements and generally only co-occur in narrow contact zones. If the normal host is locally rare or becomes absent, oviposition into a novel host might be favoured despite poor larval performance. The experiment showed that, if forced, oviposition was attempted at low frequency on the wrong host. Indeed, mtDNA data point towards a peripatric origin of T. conura adapted to C. oleraceum (Diegisser et al. 2006a). Theoretical predictions that oviposition preference will shift towards less suitable hosts if hosts which provide a suitable larval environment become scarce (Levins and MacArthur 1969;Jaenike 1978) has been verified in empirical studies where alternative hosts were accepted if 'high quality' hosts were absent (Singer 1983;Etges and Heed 1987;Courtney and Forsberg 1988).
In conclusion, it has been demonstrated that the T. conura host race on C. heterophyllum have adapted physiologically to C. oleraceum in the course of a host shift. It is likely that low abundance of C. heterophyllum favoured the host shift to C. oleraceum.