Boris Krasnov, Ramon Science Center, PO, Box 194, Mizpe Ramon 80600, Israel (fax: +972 8 6586369; e-mail: email@example.com).
1Fleas Xenopsylla conformis and Xenopsylla ramesis parasitize the rodents Meriones crassus and Gerbillus dasyurus. Previously reported isodar analysis of host selection suggested that at low density, X. conformis parasitizes M. crassus only, but with an increase in flea density, G. dasyurus is also parasitized. Xenopsylla ramesis at low density parasitizes both hosts equally, whereas at high density prefers M. crassus. The ultimate cause of underlying these patterns was suggested to be differential fitness reward of parasitism on a particular host species, while the mechanism can be adaptive host selection by fleas.
2To justify the suggested fitness–density relations, we studied reproductive success in X. conformis and X. ramesis feeding on either M. crassus or G. dasyurus. We hypothesized that reproductive success would differ between two host species for each flea species.
3Xenopsylla conformis parasitizing M. crassus produced more eggs than when parasitizing G. dasyurus, whereas time of oviposition and larval hatching did not depend on host species. In contrast, egg production in X. ramesis did not differ between host species. However, fleas fed on M. crassus needed fewer bloodmeals, oviposited earlier and hatching of their larvae took less time than those of fleas fed on G. dasyurus.
4Patterns of egg production and development time in both fleas were consistent with the hypothesis that their between-host distribution arose from adaptive host selection strategy.
The main assumption of the theory of habitat selection is that maximization of the reproductive success of an individual can be achieved by selection of those habitats that guarantee the greatest fitness output (Rosenzweig 1981, 1991). When fitness declines with density (Fretwell & Lucas 1970), individuals will successively reside in habitats of ever declining quality as their population size grows. Consequently, habitats that differ either quantitatively or qualitatively, or both, will support different numbers of individuals.
By knowing the form of density-dependent relationships, we can infer the habitat-selection response simply by examining patterns of density in different habitats using isodar analysis (Morris 1987a,b, 1988, 1990). Isodar analysis allows testing of whether consumers tend to arrange themselves across habitats as in an Ideal Free Distribution (Fretwell & Lucas 1970). This analysis produces isodars, which are plotted in density space as lines at every point of which the fitness of individuals in one habitat is assumed to be equal to that of individuals in another. Thus, isodars can be used to infer the nature of differences between habitats perceived by the consumers and the mode of habitat-selection strategy (Morris 1987a,b; Morris 1988). Each mode of habitat-selection strategy produces its own characteristic isodar, whereas intercepts and slopes of isodars indicate how a consumer perceives the between-habitat difference (Morris 1988, 2003). Isodar analysis has been successfully applied to infer habitat preferences in different animal taxa (see Morris 2003). However, experimental testing of isodar predictions as well as that of the main assumption of habitat-selection theory (see above) have rarely been conducted (Morris 1989; Morris & Davidson 2000). Moreover, the testing of habitat-selection patterns was often based on some substitute of fitness (e.g. Kotler & Brown 1999) rather than on direct fitness measurements.
Fleas (Siphonaptera) are characteristic parasites of small mammals. Traits of flea life history suggest an important role of habitat selection in their evolution and host preference. They usually alternate between periods when they occur on the body of their host and when they occur in their burrows. In most cases, pre-imaginal development is entirely off-host. The larvae usually feed on organic debris in the burrow of the host. The majority of rodent fleas parasitize more than one host species and the degree of association between a particular flea species and a particular host species varies (Marshall 1981). In addition, an individual flea is able to relocate from one host individual to another of the same or different species via free dispersal (e.g. Marshall 1981; Kuznetzov et al. 1999) or during direct between-host contacts (Krasnov & Khokhlova 2001).
We previously applied isodar analysis to study host (= habitat) selection of some flea species, each infesting two rodent species (Krasnov, Khokhlova & Shenbrot 2003). We considered the host population as a habitat of a parasite population and determined the relationship between densities on two alternative hosts for each parasite. This relationship appeared to be non-random and the resulted isodars indicated divergent fitness-density curves (= divergent population regulation, sensuMorris 1988) for fleas Xenopsylla conformis mycerini (Rothschild) and Xenopsylla ramesis (Rothschild) both parasitic on gerbillids Meriones crassus Sundevall and Gerbillus dasyurus (Wagner). Isodars showed that both fleas perceived M. crassus as a superior host compared with G. dasyrus. The divergent fitness–density curves suggest a divergence in habitat suitability with density and that, at any given density, consumers in one habitat are more efficient at converting resources into offspring than they are in another habitat (Morris 1987a,b, 1988). In addition, if the habitat that allows more efficient foraging is also more productive, it produces greater fitness reward at all density combinations (Morris 1988). Isodar slopes were > 1·0 for both fleas. Therefore, the fitness rewards for individuals of both fleas parasitizing M. crassus are expected to be higher than for those parasitizing G. dasyurus. Furthermore, the isodar intercept for X. conformis was greater than 0. This means that at low density, X. conformis only parasitizes the ‘best’ host (M. crassus), whereas with an increase in flea population size, the less preferable host (G. dasyurus) is also parasitized. In contrast, the isodar for X. ramesis passed through the origin indicating that at low density this flea parasitizes both hosts equally. However, with an increase in flea population size, their pressure on the ‘high-quality’ host increases at a higher rate than that on the ‘low-quality’ host and, thus, a preference for M. crassus over G. dasyurus occurs.
Isodar analysis of host preferences in fleas indicated that the ultimate cause underlying these patterns can be differential fitness reward of parasitism on a particular host species, while the mechanism can be adaptive host selection by fleas. Therefore, this analysis leads to clear testable hypotheses regarding the difference in reproductive success of a flea species parasitizing different hosts and, thus, justification of supposed fitness–density relations. Furthermore, the direct experimental testing of reproductive success in consumers exploiting different habitats (= hosts for parasites) can validate the isodar analysis itself (Morris & Davidson 2000). We hypothesized that reproductive patterns in both X. conformis and X. ramesis differ on different host species and that this difference in reproductive patterns is manifested differently in the two fleas. To test these hypotheses we studied egg production and development in X. conformis and X. ramesis feeding on either M. crassus or G. dasyurus.
Materials and methods
fleas and rodents
Fleas were obtained from our laboratory colonies started from field-collected specimens on M. crassus and G. dasyurus. Rodents were maintained individually in glass cages (60 × 50 × 40 cm) that contained a steel nest box with a screen floor and a pan containing a mixture of sand and dried bovine blood (larval nutrient medium). We infested a host with 12–15 (M. crassus) or 5–6 (G. dasyurus) newly emerged fleas. Gravid female fleas deposited eggs in the substrate and bedding material. We collected all substrate and bedding material bi-weekly and transferred them into an incubator (FOC225E, Velp Scientifica srl, Milano, Italy) where flea development and emergence took place at 25 °C temperature and 75% relative humidity. Details of the flea-rearing procedure can be found elsewhere (Krasnov et al. 2002a,b). Colonies of fleas were maintained at 25 °C and 75% RH with a photoperiod of 12 : 12 (L : D) h. In this study we used only newly emerged 2-day old fleas that did not feed from emergence until the implementation of experimental treatments.
We used six adult male M. crassus and six adult male G. dasyurus from our laboratory colonies. Progenitors of the colony were captured at the Ramon erosion cirque, Negev Highlands, Israel (30°35′ N, 34°45′ E). The rodents were maintained as described above. They were offered millet seed and fresh leaves of alfalfa (Medicago sp.) ad libitum. No water was available as the alfalfa supplied enough moisture for the rodents’ needs.
The study was conducted under permits from Israel Nature and National Parks Protection Authority and Ben-Gurion University Committee for the Ethical Care and Use Animals in Experiments.
We selected randomly 80 females and 20 males of each flea species and assigned them randomly to two experimental treatments that differed in host species. Rodents were placed in wire mesh (5 × 5 mm) tubes (15 cm length and 5 cm diameter for M. crassus and 10 cm length and 2 cm diameter for G. dasyurus) that limited movement and did not allow self-grooming. The tubes with rodents were placed in individual plastic pans and five X. conformis or X. ramesis (four females and one male) were placed on each rodent for 2 h. We then collected the fleas by brushing the hair of the rodent with a tooth-brush until all fleas were recovered. Each female flea was examined under a light microscope (without dissection, between two glass slides, 40× magnification) and the proportion of the midgut engorgement with the blood and the degree of egg development (early, middle or late) were determined visually (see Krasnov et al. 2002d for details). Stages of egg development were determined based on the oocyte size, degree of yolk deposition and changes in follicular epithelium (see Lindsay & Galloway 1997 and Krasnov et al. 2002d for details). After feeding, fleas were placed in plastic cups (200 cm2) with their bottom covered by a thin layer of sand, transferred into an incubator and maintained at 23 °C air temperature and 92–95% RH (four females and one male per cup). The second and subsequent feedings were conducted daily using the procedure described above. Each consecutive feeding of a flea was done on a different randomly selected individual host of the same species. After every feeding and microscope examination, fleas of the respective treatment were randomly distributed among plastic cups (four females and one male per cup). Females with eggs in the late stage of development were fed a final time, assigned an individual number and placed individually in 20 mL glass vials that contained a thin layer of sand and small pieces of filter paper. Each vial was covered with a nylon screen held by a rubber band. Vials were returned to incubators and checked twice a day. After oviposition, we counted eggs, checked each flea under the light microscope to guarantee the absence of underdeveloped eggs at the late development stage and returned it to the colony. Vials with eggs were checked every 8 h until all larvae hatched. An egg was considered dead if no larva hatched after 30 days.
The distributions of the majority of dependent variables (proportion of midgut engorgement after feeding, time to oviposition, number of eggs produced per female and time to larva hatching) did not conform to the assumptions of parametric tests and transformation of data did not lead to normality (Shapiro–Wilk's tests, W = 0·7–0·9, P < 0·01). Thus, nonparametric statistics were used. We compared these parameters between host species using Mann–Whitney U-test separately for X. conformis and X. ramesis. To test for the possible confounding effect of host individual on reproductive parameters of each flea within each host treatment, we applied Kruskal–Wallis anova by ranks separately for each dependent variable for each flea species on each host species. No effect of host individual on any studied parameter was found (Kruskal–Wallis H = 1·02 = 2·47, P > 0·6).
To evaluate the effect of host species on the number of feedings necessary for egg development, we estimated the number of feedings necessary for 50% of females to develop their eggs (NF50) using the logistic model analogous to the half-maximal response model of pharmacological research (Neter, Wasserman & Kutner 1990)
where P is the percentage of females with fully developed eggs, NF is number of feedings, and b is slope of the function. We applied least squares estimation procedures via the quasi-Newton algorithm. Estimated parameters were compared within flea species between host species using t-test.
The influence of host species on percentage egg survival was analysed using Yate's corrected χ2 test for 2 × 2 contingency tables separately for X. conformis and X. ramesis. We used Bonferroni correction to correct the alpha level to compensate for carrying out multiple tests. Therefore, the alpha level was set as 0·01. Data are presented as means ± SE.
The pattern of midgut engorgement of the two flea species after 2 h of feeding on M. crassus or G. dasyurus is presented in Fig. 1. In X. conformis, the proportion of midgut engorged with blood was significantly higher when fleas were fed on G. dasyurus than when fed on M. crassus (Mann–Whitney test, Z = −2·7, P < 0·01). Xenopsylla ramesis engorged their midgut similarly when fed on either host species (Mann–Whitney test, Z = 1·8, P = 0·1).
The relationships between number of feedings and percentage of females with fully developed eggs fitted well to the model (equation 1) (Table 1). The number of feedings necessary for 50% of females to achieve full development of eggs did not differ between host species in X. conformis (t = 0·6, P = 0·5), but differed significantly in X. ramesis, being higher when a flea fed on G. dasyurus than on M. crassus (t = 2·9, P < 0·01) (Table 1). Time from the first feeding to oviposition did not depend on host species in X. conformis (Mann–Whitney test, Z = 0·71, P = 0·4), whereas in X. ramesis it was significantly shorter when fleas fed on M. crassus than on G. dasyurus (Mann–Whitney test, Z = 3·97, P < 0·01) (Fig. 2a).
Table 1. Number of feedings necessary for 50% flea females to achieve full egg development (NF50 ± SE) when feeding on different hosts. V, variance explained by model (equation 1); R, coefficient of correlation; b1, slope of the function (all significant, P < 0·05)
5·6 ± 0·3
4·2 ± 0·9
5·4 ± 0·1
7·8 ± 1·3
5·5 ± 0·4
3·5 ± 0·8
4·2 ± 0·2
5·3 ± 1·1
Egg production by females flea when fed on the different rodent species is presented in Fig. 2(b). Female X. conformis fed on M. crassus produced significantly more eggs than those fed on G. dasyurus (Mann–Whitney test, Z = 2·29, P < 0·01). In contrast, the mean number of eggs produced by female X. ramesis did not differ between host species (Mann–Whitney test, Z = 0·62, P = 0·4). In addition, survival of X. conformis eggs differed significantly between host species, being higher when females fed on M. crassus than on G. dasyurus (74% vs. 93%, respectively, Yate's corrected χ2 = 5·38, P < 0·01). Eggs of X. ramesis survived similarly on both rodent species (93% and 91% of eggs survived when fleas were fed on M. crassus and G. dasyurus, respectively, Yate's corrected χ2 = 0·07, P = 0·79).
Time to X. conformis larva hatching was similar for females fed on both host species (Mann–Whitney test, Z = −0·21, P = 0·8; Fig. 2c). Larvae of X. ramesis hatched significantly faster for females fed on M. crassus than for females fed on G. dasyurus (Mann–Whitney test, Z = 5·46, P < 0·01; Fig. 2c).
Our hypotheses were supported by the data. Xenopsylla conformis parasitizing M. crassus demonstrated higher reproductive success compared with conspecifics parasitizing G. dasyurus. Females of this species that fed on M. crassus produced more eggs than females fed on G. dasyurus. Moreover, eggs produced by fleas when feeding on M. crassus survived better than those on G. dasyurus. In addition, food requirements necessary for successful egg development appeared to be lower if a flea exploited M. crassus rather than G. dasyurus. This suggests that the fitness reward of individuals exploiting M. crassus is greater than that of individuals exploiting G. dasyurus, all else being equal, and, thus, confirms the isodar for X. conformis that demonstrated sharp host selectivity even at low density. However, the hatching time of X. conformis larvae did not depend on host species.
In contrast, X. ramesis did not have any direct reproductive advantage when feeding on a particular host. This confirms the isodar for X. ramesis that indicated random host selection at low density. Nevertheless, exploitation of M. crassus compared with that of G. dasyurus provided an indirect reproductive benefit in terms of its temporal pattern. Fleas that fed on M. crassus needed fewer bloodmeals, eggs in their oviducts developed faster and larval hatching took less time compared with fleas that fed on G. dasyurus. Xenopsylla ramesis's isodar indicated a preference for M. crassus at high density. Indeed, at high density, the food resource for larvae (the amount of organic matter in the burrow) can be a limiting factor and, thus, shorter development time can be advantageous for an individual flea.
The reproductive patterns of X. conformis and X. ramesis proved to be consistent with the hypothesis that they are using adaptive host selection strategies. Indeed, in X. conformis exploitation of M. crassus provided greater fitness reward than exploitation of G. dasyurus. By behaving in an IFD-like way, individual fleas can maximize their reproductive success by only parasitizing M. crassus, as occurs when flea population size is low. However, when flea population size grows, the amount of organic matter in the host burrow presumably becomes a limiting factor for larval growth and development. Consequently, between-larva competition for food resources intensifies (Marshall 1981) and their survival decreases, thus, decreasing the per capita fitness of X. conformis. Some fleas will select G. dasyurus when increased density reduces fitness rewards for those fleas parasitic on M. crassus to levels comparable with G. dasyurus, owing to intensification of between-larva competition. Survival of their larvae is likely to be higher because of their lower density in burrows of G. dasyurus. Therefore, the lower egg production of fleas parasitic on G. dasyurus will be compensated for by higher survival of the larvae (with consequent fitness rewards), and there will thus be an equalization of fitness rewards between hosts at different within-host densities. From this point on, further increase in flea population size will be divided equally between both hosts (Morris 1988). Another factor that presumably decreases flea fitness with increasing density can be an increase in host defensiveness. The more defensive the host, the more likely a biting ectoparasite will be interrupted before completion of a bloodmeal (e.g. Kelly, Mustafa & Dye 1996). The amount of blood taken by a haematophagous insect is, in turn, closely linked to its fecundity (e.g. Takken, Klowden & Chambers 1998).
Xenopsylla ramesis at low population size achieves similar maximum fitness when exploiting either host and, thus, select hosts randomly. As in X. conformis, amount of larval food resources would become a limiting factor with an increase in flea population size. Under these conditions, larvae that hatch earlier (i.e. from females exploiting M. crassus) will most likely have an advantage over larvae that hatch later (i.e. from females exploiting G. dasyurus) simply because the former would consume rapidly most of the available food. The cumulative time to larval hatching consists of a period of egg development inside the female and a period of egg development after oviposition. Both these periods are shorter when a flea feeds on M. crassus than on G. dasyurus. Therefore, parasitism on M. crassus compared to that on G. dasyurus produces a ‘delayed’ fitness advantage which is manifested only at high flea density. In addition, shorter development time allows an increase in the number of flea generations per breeding season.
The feasibility of the proposed scenarios can be indirectly supported by our long-term observations on the distribution of M. crassus and G. dasyurus (1992–2003, twice a year trapping sessions in 14–24 sample plots). In areas inhabited by X. conformis (but not by X. ramesis) rodent densities are relatively low (Krasnov et al. 1996). Burrows of M. crassus and G. dasyurus are almost completely separated, although spatial overlapping between home ranges of individual M. crassus and G. dasyurus is relatively high (G. Shenbrot & B. Krasnov, unpublished). Consequently, processes of between-larva competition for food presumably occur independently and are spatially separate for fleas parasitic on M. crassus and for fleas parasitic on G. dasyurus. In contrast, in areas inhabited by X. ramesis (but not X. conformis), densities of both hosts are relatively high (Krasnov et al. 1996) and their distributions are spatially overlapped, so that G. dasyurus was repeatedly recorded in burrows of M. crassus (G. Shenbrot & B. Krasnov, unpublished data). This suggests the spatial co-occurrence of larvae of fleas fed on M. crassus and larvae of fleas fed on G. dasyurus and, thus, competition for food between these groups. Consequently, host-dependent differences in the time of larval hatching supposedly affect the competitive outcome in X. ramesis, but not in X. conformis.
Results of the present study support our earlier suggestion that qualitative between-host difference was represented by some parameters of host-blood biochemistry (Krasnov et al. 2003). Furthermore, this difference seemed to be perceived differently by the two fleas. Xenopsylla conformis consumed more blood per bloodmeal when fed on G. dasyurus than on M. crassus, whereas X. ramesis required a higher number of bloodmeals on G. dasyurus than on M. crassus for successful egg development. This suggests that the efficiency of digestion of M. crassus blood was higher than that of G. dasyurus blood in both fleas. Host preference based on blood biochemistry has been demonstrated for other ectoparasites. It is largely unclear what these blood parameters are, although, for example, a preference for human blood over mouse blood in the female mosquito Aedes aegypti has been shown to be based on its amino acid composition, specifically the isoleucine content (Harrington, Edman & Scott 2001).
Xenopsylla conformis and X. ramesis are not unique in that their reproductive patterns differ on different host species. A decrease in reproduction in fleas fed on one but not another host has been reported for other flea species. The rat fleas Xenopsylla cheopis and Xenopsylla astia did not reproduce when fed on man (Seal & Bhattacharji 1961). Fecundity and egg hatchability in X. cheopis were higher when the fleas were fed on Rattus rattus than on Bandicota bengalensis (Prasad 1969). Egg production in Parapulex chephrenis and Xenopsylla dipodilli was higher when fleas fed on their specific hosts (Acomys cahirinus and G. dasyurus, respectively) than when fed on other rodent species (Krasnov et al. 2002c). However, these studies either dealt with highly host-specific fleas (e.g. Krasnov et al. 2002c) or compared between characteristic and highly non-characteristic host species (e.g. Seal & Bhattacharji 1961). Consequently, it is not surprising that the resulted effects were sharp and clear. Host-generalist fleas, such as X. conformis and X. ramesis, represent a more challenging pattern. Narrative description or calculation of standard parasitological indices (e.g. prevalence and intensity of infestation for a particular host species) of their between-host distribution did not provide an indication of their host preferences, nor the mechanisms underlying their host distribution (Theodor & Costa 1967; Krasnov et al. 1997). In contrast, isodar analysis allowed detection of the mode with which a flea perceives differences between host species, and an explanation of between-host flea distribution through subtle differences in fitness as a consequence of host selection.
This study was supported by the Israel Science Foundation (Grant no. 663/01–17·3 to for BRK, ISK and GIS). We thank two anonymous referees for their most helpful comments. The experiments comply with the laws of State of Israel. This is publication no. 157 of the Ramon Science Center and no. 406 of the Mitrani Department of Desert Ecology.