Pollen host choice in bees is in many cases highly conserved, which might partly be due to physiological limitations of bee larvae to digest non-host pollen. These limitations need to be overcome in order to incorporate new pollen hosts; however, the mechanisms underlying such host expansion are poorly understood.
In this study, we examined intra- and interpopulational variation in the ability of larvae of the solitary bee species Osmia cornuta (Megachilidae) to develop on a non-host pollen diet of Ranunculus acris (Ranunculaceae) by comparing larval performance within and between five geographically distant European populations.
The majority of bee larvae from all tested populations died when reared on the Ranunculus pollen diet. Between 10% and 43·5% of all larvae per population reached the cocoon stage, and 48% of these emerged as viable adults from the cocoons, indicating that the physiological ability to cope with the unfavourable properties of Ranunculus pollen exists in each population.
The bee larvae of one population exhibited significantly reduced survival on the Ranunculus pollen diet compared with three of the four other populations.
Although bees that successfully developed on the Ranunculus pollen diet showed a distinctly prolonged development time, exhibited higher mortality during diapause and reached a considerably lower adult weight compared with individuals fed the control pollen diet, several of the Ranunculus fed individuals were able to reproduce and to sire viable offspring.
This study provides the first evidence for both intra- and interpopulational variation in the physiological ability of solitary bees to digest non-host pollen. This variation might enable host expansion and subsequent host shifts in response to natural selection.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Bees collect large quantities of pollen and nectar for their own nourishment and as food for their offspring (Michener 2007). The plants exploited by bees are remarkably diverse, and host plant spectra differ widely between bee taxa (Westrich 1989). According to the breadth of their host plant spectrum, bees are classified as pollen specialists, which restrict pollen collection to one plant family, subfamily or genus (‘oligolecty’), or as pollen generalists, which harvest pollen from plants of two or more different plant families (‘polylecty’) (Robertson 1925; Cane & Sipes 2006; Müller & Kuhlmann 2008).
Recent studies suggest that many polylectic bee lineages are derived from oligolectic ancestors (Müller 1996; Sipes & Tepedino 2005; Danforth et al. 2006; Larkin, Neff & Simpson 2008; Michez et al. 2008), that host shifts among oligoleges are probably always preceded by a shorter or longer period of polylecty (Sedivy et al. 2008) and that closely related generalist bee species may exploit different hosts (Sedivy, Müller & Dorn 2011). In all these cases, bees have incorporated new hosts into their diet range. However, the mechanisms underlying such host expansion in bees are poorly understood. It has been shown that bees usually broaden their diet range while continuing to utilize the original host and that host shifts frequently occur to plants that are already exploited by close relatives (Müller 1996; Sipes & Tepedino 2005; Sedivy et al. 2008). These findings suggest that certain plant traits impose constraints on the bees, which render host expansion difficult. In fact, host plant choice seems to be highly conserved in bees, and incorporation of new hosts appears to occur much less often than previously thought (Sedivy et al. 2008). Constraints that impede the easy and rapid acquisition of new pollen hosts may be due to neurological limitations related to the recognition or handling of flowers (Praz, Müller & Dorn 2008a) or to physiological limitations related to pollen digestion, including lack of essential nutrients (Levin & Haydak 1957; Suárez-Cervera et al. 1994; Williams 2003; Praz, Müller & Dorn 2008b). Thus, bees have to overcome these constraints in order to broaden their pollen diet, which then allows for subsequent respecialization, that is, host plant shift.
A possible mechanism underlying host expansion and subsequent host shift is genetic variation in the ability of individuals of the same population to successfully overcome constraints imposed by non-hosts. Such intrapopulational variation has recently been reported for herbivorous insects. Larval survival of the oriental fruit moth (Grapholita molesta) that typically feeds on a wide range of rosacean fruit trees varied considerably between individuals reared on a diet supplemented with juglone, the main defence compound of the non-host plant walnut, which belongs to the phylogenetically distant family of Juglandaceae (Piskorski, Ineichen & Dorn 2011). Similarly, larval survival of the highly generalist gypsy moth (Lymantria dispar) varied considerably between individuals fed leaves of Eucalyptus, a plant genus that does not naturally occur in the distribution range of this Palearctic moth species and that contains secondary metabolites, which are unfamiliar to L. dispar (Matsuki et al. 2011). In both herbivorous species, the ability to develop on a chemically and phylogenetically unfamiliar host is present in some individuals even though the moths do not exploit or never have encountered the plant species. Anecdotal evidence suggests that similar intrapopulational variation in the ability to successfully use certain plant hosts might also exist in bees. When reared on a diet of buttercup (Ranunculus) pollen, all larvae of the pollen generalist mason bee species Osmia cornuta of a south German population died except for two females, which developed into dwarf-sized adults (Sedivy, Müller & Dorn 2011). Interestingly, the same bee species has been reported to collect Ranunculus pollen in Spain and in northern Italy (Márquez, Bosch & Vicens 1994; Nepi, Pacini & Pinzauti 1997), and Osmia bicornis, a close relative of O. cornuta, is known to frequently collect pollen of Ranunculus and to successfully develop on it (Westrich 1989; Vicens, Bosch & Blas 1993; Sedivy, Müller & Dorn 2011).
Genetic variation among individuals in their ability to successfully exploit a certain plant is a condition for natural selection to act towards the incorporation of that plant into the normal diet of a population. Correspondingly, the proportion of individuals that are able to overcome constraints imposed by potential hosts is expected to differ among populations that are under differing selection regimes. In fact, interpopulational variation in host plant use or performance is well known among herbivorous insects (Jaenike 1990 and references therein), but to our knowledge, such interpopulational variation has not been reported for bees so far.
In this study, we analysed intra- und interpopulational variation in the ability of larvae of O. cornuta (Latreille 1805) to develop on a pollen diet of Ranunculus acris L. by comparing larval performance within and between five geographically distant European populations. We hypothesized that (i) some individuals in every population are able to overcome the unfavourable properties of a Ranunculus pollen diet and that (ii) differences in the performance of larvae reared on a Ranunculus pollen diet exist between the five populations.
Materials and methods
Osmia cornuta (Apoidea: Megachilidae) is a common solitary bee species in Central Europe and is widespread throughout the Palearctic (Peters 1978; Westrich 1989; Fig. 1). Its distribution ranges from northern Africa in the south to northern Germany in the north and from Portugal in the west to Turkmenistan in the east (Peters 1978). The species is subdivided into three subspecies and the nominate form, O. c. cornuta, occurs in Central Europe (Peters 1978). It mainly inhabits warmer areas below 500 m and has an early seasonal flight period that usually lasts from the beginning of March until the beginning of May (Westrich 1989). Osmia cornuta is broadly polylectic and collects pollen from plants belonging to many different families (Westrich 1989; Márquez, Bosch & Vicens 1994). Although small quantities of Ranunculus pollen have been found in brood cell provisions of O. cornuta in southern Europe (Márquez, Bosch & Vicens 1994; Nepi, Pacini & Pinzauti 1997), Ranunculus is not a regular pollen host in Central Europe as suggested by field observations and the almost complete absence of Ranunculus pollen in the scopal pollen loads of collected females (M. Haider, S. Dorn, A. Müller unpublished data). Osmia cornuta nests in a great variety of pre-existing cavities and willingly accepts hollow bamboo sticks as nesting site. Female bees build several brood cells per nest, which are provisioned with a mixture of pollen and nectar before a single egg is laid on top of each provision. The hatched larva feeds on the pollen–nectar provision and completes its development by autumn, after which the fully developed adult overwinters in its cocoon within the brood cell.
To test for intra- and interpopulational differences in the capability of O. cornuta to develop on a pollen diet of Ranunculus, we selected five populations of the same subspecies, O. c. cornuta, originating from Belgrade (Serbia), Bologna (Italy), Erfurt (Germany), Konstanz (Germany) and Troyes (France) (Fig. 2). The distance between these five populations ranges from 380 to 1300 km. Regular gene flow between the five populations is unlikely as all populations are separated from each other by the Alps or other mountain ranges, which are unsuitable habitats for the thermophilous O. cornuta.
Origin of the Pollen Diet
Larvae of O. cornuta were reared on a pollen diet of R. acris (Ranunculaceae) collected by females of the Ranunculus specialist Chelostoma florisomne (Apoidea: Megachilidae) at Gletterens in western Switzerland. Reed stems containing brood cells of C. florisomne were collected and taken to the laboratory, where they were carefully opened longitudinally with a knife and the pollen/nectar provisions withdrawn. As a control, we used a pollen diet of Sinapis arvensis (Brassicaceae) collected by the females of O. cornuta in a large walk-in cage (see below). Species of Brassicaceae are known to be regular pollen hosts of O. cornuta (Westrich 1989; Márquez, Bosch & Vicens 1994; M. Haider, S. Dorn, A. Müller unpublished data). Prior to use in the experiments, all freshly collected provisions were stored at −20 °C for at least 24 h. In addition, we also used provisions that were collected during the previous season and stored at −20 °C for up to 1 year.
Experiments with the populations of Erfurt and Konstanz were conducted during spring 2010. Experiments with the populations of Belgrade, Bologna, Troyes and a second part of the Konstanz population were conducted during spring 2011. About 100 female and 100 male cocoons from each population (one male and one female cocoon per nest) were overwintered in a climate chamber (E7⁄2; Conviron, Winnipeg, MB, Canada) at −3 °C from December until end of April, resulting in a delay of emergence of about 2 months. This emergence delay was necessary as the normal flight period of O. cornuta usually overlaps only marginally with that of C. florisomne, which is active from May to June. After overwintering, the cocoons were placed in artificial nesting stands inside walk-in cages (10 × 8 × 3·5 m) covered with gauze. For each population, a separate walk-in cage was used. Each cage was provided with about 800 S. arvensis plants in 400 pots as pollen and nectar sources and with hollow bamboo stalks and moist soil as nesting resources. The cocoons of the Konstanz population were placed in an artificial nesting stand outside the cages at the Experimental Research Station of the ETH Zürich at Eschikon, and the emerged bees were allowed to collect pollen and nectar freely. The delayed emergence prevented these bees from coming into contact with the local population of O. cornuta. Of each population, we transferred between 60 and 80 eggs onto each of the two pollen diets, except for the Troyes population, of which fewer eggs were transferred due to an unexpectedly short flight period of this population. Two eggs per female were removed from the nests, one of which was later placed on a Ranunculus pollen diet and one on a control pollen diet. As the females often build more than one nest during their flight period, females of all populations were individually marked with numbered opalite plates to avoid the use of more than two eggs per individual bee in our experiments. The eggs were carefully transferred with a thin pair of tweezers onto the pollen diets previously placed in artificial brood cells. These artificial cells were made of small blocks of beech wood (4 × 2 × 2 cm) provided with a drilled burrow (length 2 cm, width 0·8 cm), open both at the top and at the front side. These openings were covered with coverslips attached to the block with transparent adhesive tape to permit free viewing into the burrow. Based on the average weight of the provisions from ten female brood cells, we provided 600 ± 5 mg of pollen diet per artificial brood cell. As eggs could not be sexed, each cell was provided with the same quantity of pollen diet despite the fact that under natural conditions male cells receive less food than female cells, which is explained by the distinctly smaller size of adult males (Westrich 1989). This experimental artefact did not appear to bias our results as the male larvae did never consume the entire cell provision in contrast to the female larvae that usually consumed the entire provision before they started to spin the cocoon.
After egg transfer, the artificial brood cells were incubated in constant darkness within a climate chamber (E7⁄2; Conviron) under the following conditions: 25 °C for 16 h followed by a gradual reduction in temperature to 10 °C within four hours followed by a gradual increase back to 25 °C within another four hours. Relative humidity was constantly held at 60%. Larval development was checked every second day, and the following developmental stages were recorded: (i) egg hatching; (ii) feeding without defecating; (iii) feeding and defecating; (iv) start of spinning silk; (v) completion of cocoon. After the cocoons had been completed, cells were kept in the climate chamber until autumn to allow for the development of the bees to the adult stage. In autumn, all cocoons were weighed with the adult bees inside (AB204; Mettler Toledo, Switzerland) and afterwards stored at 4 °C in constant darkness for overwintering. In early spring, the sex of the bees was determined through a small hole cut at the anterior end of the cocoon with a pair of small scissors. Males are easily distinguished from females by their white facial pilosity (Fig. 1). Sexing was performed in early spring only, as opening the cocoons already in autumn would have caused high mortality of the overwintering bees.
Reproductive Capability of the Ranunculus Fed Individuals
The ability to reproduce was tested for all individuals of the Bologna, Belgrade, Konstanz and Troyes populations that successfully developed on the Ranunculus pollen diet in 2011. In contrast to 2010, when six females developed into viable adults, no females, but only males emerged from the cocoons in 2012, which might have been due to the production of predominantly male eggs by the nesting O. cornuta females triggered by a temporary pollen-shortage phase in the walk-in cages in 2011 (see Strohm & Linsenmair 1997; Kim 1999; Bosch 2008). Therefore, the reproductive ability could only be tested for 18 males. These males were allowed to fly within a large walk-in cage together with 18 females that had developed on a S. arvensis pollen diet. Completed nest were collected to determine the ploidy of the larvae. As in the Hymenoptera unfertilized eggs develop into haploid males, whereas fertilized eggs develop into diploid individuals that are generally females (Elias, Mazzi & Dorn 2009), production of diploid progeny by the Sinapis fed females would unambiguously indicate that their eggs had been fertilized by the sperm of the Ranunculus fed males. For the ploidy analyses, two post-defecating larvae were sampled per nest, including the innermost larva, which is usually female, and the outermost larva, which is usually male. Ploidy was determined using flow cytometry following the protocol provided by Ruf, Dorn & Mazzi (in press), using the head of the larvae for the extraction of cell material. For extraction and staining of cell material, we used CyStain® PI absolute T (Partec GmbH, Münster, Germany). Flow cytometric analyses were performed using a BD FACSCalibur™ multicolour flow cytometer (Becton, Dickinson and Company, Franklin Lakes, NJ, USA) equipped with a blue argon laser (excitation: 488 nm) and a band pass filter of 585 nm to detect PI fluorescence. For each sample, 15 000 nuclei were measured in an FL2-W/FL2-A gated region containing haploid, diploid and tetraploid cells, using the software CellQuest™ Pro (Becton, Dickinson and Company). The ploidy of the larvae was determined using flow cytometric DNA histograms of the head of adult O. cornuta males and females as a reference.
Eggs that did not hatch and larvae that undoubtedly died from external factors, such as mite attack or mechanical damage during handling, were excluded from all analyses. For analysis of survival, all larvae that completed their cocoons were counted as survivors, irrespective of whether they later successfully completed metamorphosis. Cocoons were regarded as completed when they became intransparent. Survival of larvae on the different pollen diets was analysed using Kaplan–Meier survival statistics (Lee & Wang 2003). The number of days between hatching and completion of the cocoon was considered to be ‘censored data’: individuals that died before the completion of the cocoon represented the exact observations for which the event (death) occurred. Those that completed the cocoon, that is the survivors, were considered the censored observations and thus withdrawn from survival calculations. Differences between survival distributions were analysed with log-rank tests, using the option ‘pairwise for each stratum’ implemented in spss Statistics 19.0 (SPSS Inc., Chicago, IL, USA) and controlled with false discovery rate correction (Benjamini & Hochberg 1995). Development time (time between egg hatching and completion of the cocoon) and adult weight (including cocoon) were analysed for those bees that successfully reached the adult stage because sex determination was not possible in the larval stage. Differences in the development time were analysed by applying a generalized linear model using pollen diet, population and sex as fixed factors. Differences in the adult weight were analysed for males only by applying a generalized linear model using pollen diet and population as fixed factors. The number of Ranunculus reared females that completed metamorphosis was too low to apply statistical tests. Significant differences between the populations were identified using Bonferroni post hoc tests. Statistical analyses were conducted with spss Statistics 19.0 for Macintosh OS X.
A total of 504 transferred eggs hatched (Table 1). Three larvae that died from mechanical damage and one larva that died from an infection with mites were excluded from all analyses. Larval survival on the control (Sinapis) pollen diet was >93% for each of the five populations (Table 1), and 95% of the survivors on the control pollen diet developed into viable adults, indicating that our rearing method had at most a marginal effect on larval performance.
Table 1. Larval survival, development time and number of viable adults of individuals derived from five different populations of Osmia cornuta reared on a control pollen diet of Sinapis arvensis and on a pollen diet of Ranunculus acris
Number of eggs hatched
Development time [days]
No. viable adults
Development time represents the number of days between hatching and completion of the cocoon of all bees that developed into viable adults. Differences in the survival of the populations were tested using Kaplan–Meier analysis. Group heterogeneity: populations sharing the same letter did not differ significantly at P < 0·05 (post hoc test: pairwise log-rank tests with false discovery rate control). m = males, f = females.
Larval survival on the Ranunculus pollen diet was significantly reduced in all five populations (Kaplan–Meier analysis, log-rank tests; Belgrade: χ2 = 32·331, P < 0·001; Bologna: χ2 = 70·097, P < 0·001; Erfurt: χ2 = 112·365, P < 0·001; Konstanz: χ2 = 54·597, P < 0·001; Troyes: χ2 = 11·407, P = 0·001; Fig. 3). The majority of all larvae died on the Ranunculus pollen diet; however, some individuals from every population reached the cocoon stage (Table 1). The percentage of these survivors ranged from 10·0% among larvae of the Erfurt population to 43·5% among larvae of the Troyes population. 47·5% of all survivors successfully developed into viable adults (Table 1).
Larval survival on the Ranunculus pollen diet was significantly different between the populations (Kaplan–Meier analysis, log-rank test, χ2 = 15·272, P < 0·004). This difference was due to a significantly reduced survival of larvae of the Erfurt population compared with larvae of all other populations except the Bologna population (Kaplan–Meier analysis, pairwise log-rank tests; Erfurt-Troyes: χ2 = 8·598, P = 0·003; Erfurt-Konstanz: χ2 = 9·639, P = 0·002; Erfurt-Belgrade: χ2 = 8·971, P = 0·003; Erfurt-Bologna: χ2 = 0·717, P = 0·397; Table 1; Fig. 3). Larval survival did not differ significantly between the other four populations (results of statistical tests not shown).
Larval development time did not differ between males and females reared on the same pollen diet (Wald χ2 = 0·949, d.f. = 1, P = 0·330). Development time on the Ranunculus pollen diet was significantly prolonged in larvae of all five populations (32–54 days, median: 42) compared with development time on the control pollen diet (26–44 days, median: 34) (Wald χ2 = 131·582, d.f. = 1, P < 0·001; Fig. 4).
Development time on the Ranunculus pollen diet differed significantly between larvae derived from the different populations (Wald χ2 = 15·922, d.f. = 4, P = 0·003): development time of larvae of the Belgrade population was significantly prolonged compared with that of larvae of the Erfurt (Bonferroni post hoc test: P = 0·046) and the Konstanz (Bonferroni post hoc test: P = 0·015) populations. Development time did not differ significantly between all the other populations (results of statistical tests not shown). There was no significant interaction between population and pollen diet (Wald χ2 = 4·179, d.f. = 4, P = 0·382) nor between pollen diet and sex (Wald χ2 = 0·949, d.f. = 1, P = 0·330).
Adult weight of males reared on the Ranunculus pollen diet was significantly lower in all populations (46·5–98·1 mg, median: 67·8 mg) compared with those reared on the control pollen diet (48·4–178·5 mg, median: 110·4 mg) (Wald χ2 = 73·034, d.f. = 1, P < 0·001; Fig. 5).
Adult male weight did not differ significantly between populations (Wald χ2 = 1·901, d.f. = 4, P = 0·754); however, there was a significant interaction between population and pollen diet (Wald χ2 = 11·383, d.f. = 4, P = 0·023). This interaction was due to the presence of significant interpopulational differences in the weight of individuals reared on the control pollen diet and the absence of such interpopulational differences between individuals reared on the Ranunculus pollen diet. Sinapis fed males of the Erfurt population were significantly heavier than those of the Belgrade and Bologna populations (Bonferroni post hoc tests: Erfurt-Belgrade: P < 0·001; Erfurt-Bologna: P = 0·019) and Sinapis fed males of the Konstanz population were significantly heavier than those of the Belgrade population (Bonferroni post hoc test: P = 0·016).
Reproductive Capability of Ranunculus Fed Individuals
To test their reproductive capability, males of O. cornuta that successfully developed on a Ranunculus pollen diet were allowed to mate with females reared on a Sinapis pollen diet, and 52 offspring larvae from 26 nests were analysed for their ploidy. Diploid individuals were found in 19 nests, clearly indicating that at least some of the Ranunculus fed males produced fertile sperm.
The results of this study clearly demonstrate that the ability to develop on a pollen diet of R. acris differs within and between populations of the solitary bee species O. cornuta. In each of the five tested populations, a small proportion of the Ranunculus fed individuals reached the adult stage, indicating that the physiological ability to cope with the unfavourable properties of Ranunculus pollen exists in each population. Such intrapopulational variation in larval performance on non-host diets has not been previously recorded in bees. It was recently observed, however, in herbivorous insects. In two generalist moth species and a generalist beetle species, the physiological ability to develop on an unfamiliar host was found to vary considerably between larvae of the same population (Fox, Nilsson & Mousseau 1997; Matsuki et al. 2011; Piskorski, Ineichen & Dorn 2011). In contrast, specialized leaf beetles of the genus Ophraella exhibited only little genetic variation influencing larval survival, oviposition and feeding responses on unfamiliar hosts (Futuyma, Keese & Funk 1995). Differences in the ability of generalist vs. specialist herbivores to cope with unfamiliar hosts might be due to a broader range of non-specific detoxifying tools used by generalists compared with specific tools required by the specialists to cope with the secondary chemistry of their exclusive hosts (Krieger, Feeny & Wilkinson 1971; Li, Schuler & Berenbaum 2003; Ramsey et al. 2010). Correspondingly, the experimental finding that all larvae of two pollen specialist bee species quickly died on a Ranunculus pollen diet (Praz, Müller & Dorn 2008b) suggests that the larvae of the generalist O. cornuta might possess non-specific physiological tools that enable at least some individuals to successfully develop on a Ranunculus pollen diet.
Individuals of O. cornuta that developed on the Ranunculus pollen diet showed a distinctly prolonged development time, exhibited higher mortality during diapause and reached a considerably lower adult weight compared with individuals fed the control pollen diet. This finding adds further evidence in support of the unfavourable properties of R. acris pollen for bee larval development (Praz, Müller & Dorn 2008b; Sedivy, Müller & Dorn 2011). The mechanism underlying the adverse effect of Ranunculus pollen on larval development of unspecialized bee species is still unknown (Sedivy et al. 2012). Its unfavourable properties might result from deficiencies or imbalances in nutrient content, from structural properties that render extraction of pollen nutrients difficult, from interference of the pollenkit with nutrient assimilation or from toxic secondary pollen compounds (see Praz, Müller & Dorn 2008b and Roulston & Cane 2000 for a discussion of possible unfavourable pollen traits). Pollen of Ranunculus has an intermediate protein content as has the pollen of Sinapis (Wille et al. 1985; Roulston, Cane & Buchmann 2000), and it contains all the amino acids known to be essential for the honeybee, albeit in rather low concentrations (McLellan 1977; Szczęsna 2006). Therefore, the content of both proteins and essential amino acids in the pollen of Ranunculus appears to be sufficient for a successful development. We cannot exclude, however, that other essential nutrients are lacking in Ranunculus pollen, such as sterols, which insects need for the synthesis of certain hormones, such as ecdysone (Svoboda et al. 1978; Rasmont et al. 2005). The fact that 80% of the larvae of O. cornuta already died within 14 days after onset of feeding, that the larvae turned conspicuously green shortly before they died and that the surviving larvae, irrespective of whether female or male, did not consume the entire cell provisions before they started to spin a cocoon might possibly indicate the presence of an unknown toxic compound in the pollen of Ranunculus.
While it is well known that small adult body size considerably reduces bee reproductive fitness (Kim 1997; Seidelmann, Ulbrich & Mielenz 2010; but see Bosch & Vicens 2006), the Ranunculus fed adult males of O. cornuta were able to reproduce and to sire diploid offspring. Provided that Ranunculus fed adult females of O. cornuta are fertile as well, intrapopulational variation in the ability of O. cornuta to digest R. acris pollen might enable the incorporation of R. acris into the normal pollen host spectrum of O. cornuta, if the exploitation of R. acris flowers provides a fitness benefit within a particular population. The exploitation of R. acris might be favoured by a lack of sufficient pollen from other sources due to natural fluctuations in local flower abundances or due to anthropogenic habitat alteration, which may alter insect–host associations (Tabashnik 1983; Singer, Thomas & Parmesan 1993; Singer et al. 2008). Furthermore, it might be selected for by a large increase in the abundance of R. acris flowers in the bees' habitats due to fertilization, which favours the nitrophilous R. acris (Dorioz, Fleury & Jeannin 1987; Rudmann-Maurer et al. 2008), or by a closer phenological matching between the flowering period of R. acris and the flight period of O. cornuta due to climate change, which might cause advances in the phenology of both bees and their host plants (Hegland et al. 2009; Bartomeus et al. 2011; Forrest & Thomson 2011). Interestingly, climate change was recently found to lead to locally asynchronous advances in the phenologies of bees and their host flowers, such that the bees' flight period advanced less quickly than the hosts' flowering time (Forrest & Thomson 2011; but see Bartomeus et al. 2011). An earlier advance of the flowering period of R. acris relative to the flight period of O. cornuta would extend the temporal overlap in phenology between O. cornuta and R. acris, which is currently very brief in many regions of Central Europe, and thus might enable natural selection to act towards the incorporation of R. acris into the diet of O. cornuta.
Beside intrapopulational variation, our study also revealed variation between populations in the performance of O. cornuta larvae on Ranunculus pollen. While interpopulational variation in larval performance on an unfamiliar host has been shown in a generalist beetle herbivore (Fox, Nilsson & Mousseau 1997), our study provides the first documentation of such variation in the ability of a solitary bee to digest pollen of a particular non-host. Larvae of the Erfurt population exhibited significantly reduced survival on the Ranunculus pollen diet compared with larvae of all other populations except the Bologna population. We hypothesize that the absence of R. acris in the surroundings of Erfurt, which is due to a very dry climate in this region (M. Möhler, personal communication), might possibly account for this low larval survival rate. Similarly, in herbivorous insects, interpopulational variation in both larval performance and female host plant preference are usually explained by differences in the local host plant spectrum (Jaenike 1990; Keeler & Chew 2008; Downey & Nice 2011).
Our study focused on the performance of O. cornuta larvae on an unfamiliar pollen host. Incorporation of a new pollen host, however, does not only require physiological adaptations of the bee larvae to cope with the chemistry of the new pollen, but also neurological adaptations of the mother bee to recognize and exploit the host flowers. In fact, female host plant preference and larval performance are not necessarily correlated in herbivorous insects (Mayhew 1997, 2001) or in bees (Praz, Müller & Dorn 2008a) as is suggested by the ‘preference–performance hypothesis’ or ‘optimal oviposition theory’ (Jaenike 1978). Females of the pollen specialist solitary bee species Heriades truncorum refrained from collecting pollen on the flowers of two non-hosts although pollen of both non-host species were experimentally found to support larval development (Praz, Müller & Dorn 2008a). In that study, however, floral morphology of the non-hosts differed considerably from that of the exclusive host, suggesting that the Heriades females were not able to detect or collect the non-host pollen. In contrast, the flowers of Ranunculus have freely accessible anthers and strongly resemble the flowers of Rosaceae, which is the preferred host plant taxon of O. cornuta (Tasei 1973; Márquez, Bosch & Vicens 1994). Moreover, O. cornuta, a highly polylectic species, collects pollen from host species that differ widely in flower morphology (Westrich 1989) and, therefore, likely possesses a broader repertoire of pollen collection behaviours than the oligolectic H. truncorum. Harvesting pollen of Ranunculus flowers should, therefore, not pose a major challenge for O. cornuta and increased co-occurrence of R. acris with O. cornuta is expected to result in quick acceptance of the Ranunculus flowers by pollen collecting females of O. cornuta if their exploitation is favoured by natural selection.
In conclusion, this study provides first evidence for both intra- and interpopulational variation in the physiological ability of solitary bees to digest non-host pollen. This variation might provide a basis for natural selection to act in favour of host expansion and subsequent host shifts.
We thank M. Herrmann, M. Möhler, M. Perrigault, F. Sgolastra and L. Stanisavljevic for providing O. cornuta cocoons and information on local bee populations; the staff of the neolithic village of Gletterens for providing bee nests; C. Sedivy for expertise in bee rearing; A. Imboden for help with plant breeding; M. Eckhardt for help with experiments, in particular with flow cytometry; M. Kisielow (Flow Cytometry Laboratory ETH) for support in flow cytometry; A. Ferrer for statistical support; M. Waldburger for providing walk-in cages; and H. Kirk, J. Bosch, I. Alves-dos-Santos and an anonymous reviewer for helpful comments on the manuscript. All experiments comply with the current laws of Switzerland, where they were performed.