Allee effect in larval resource exploitation in Drosophila: an interaction among density of adults, larvae, and micro-organisms

Authors


Bregje Wertheim, Department of Biology, University College London, Darwin Building, Gower Street, London WC1E 6BT, U.K. E-mail: B.Wertheim@ucl.ac.uk

Abstract

Abstract 1. Aggregation pheromones can evolve when individuals benefit from clustering. Such a situation can arise with an Allee effect, i.e. a positive relationship between individual fitness and density of conspecifics. Aggregation pheromone in Drosophila induces aggregated oviposition. The aim of the work reported here was to identify an Allee effect in the larval resource exploitation by Drosophila melanogaster, which could explain the evolution of aggregation pheromone in this species.

2. It is hypothesised that an Allee effect in D. melanogaster larvae arises from an increased efficiency of a group of larvae to temper fungal growth on their feeding substrate. To test this hypothesis, standard apple substrates were infested with specified numbers of larvae, and their survival and development were monitored. A potential beneficial effect of the presence of adult flies was also investigated by incubating a varying number of adults on the substrate before introducing the larvae. Adults inoculate substrates with yeast, on which the larvae feed.

3. Fungal growth was related negatively to larval survival and the size of the emerging flies. Although the fungal growth on the substrate was largely reduced at increased larval densities, the measurements of fitness components indicated no Allee effect between larval densities and larval fitness, but rather indicated larval competition.

4. In contrast, increased adult densities on the substrates prior to larval development yielded higher survival of the larvae, larger emerging flies, and also reduced fungal growth on the substrates. Hence, adults enhanced the quality of the larval substrate and significant benefits of aggregated oviposition in fruit flies were shown. Experiments with synthetic pheromone indicated that the aggregation pheromone itself did not contribute directly to the quality of the larval resource.

5. The interaction among adults, micro-organisms, and larval growth is discussed in relation to the consequences for total fitness.

Introduction

An aggregation pheromone is a substance, emitted by an individual, that induces aggregative behaviour in conspecifics (Shorey, 1973). These pheromones have evolved several times in insects (Wertheim, 2001), which implies that under certain circumstances selection can favour individuals that seek out conspecifics. Although a number of advantages is apparent for individuals in large groups (see for example Pulliam & Caraco, 1984; Parrish & Edelstein-Keshet, 1999), the evolution of aggregative behaviour requires that even at low densities, advantages outweigh all associated costs. Such a situation could arise when survival or reproduction is hampered at low population densities, i.e. with an Allee effect.

The Allee effect was named after W. C. Allee, for his prominent role in construing adaptive value to grouping behaviour in animals (Allee et al., 1949; Courchamp et al., 1999). Recently, Stephens et al. (1999) provided a clear definition for the Allee effect: ‘A positive relationship between any component of individual fitness and either numbers or density of conspecifics’. This definition emphasises that the Allee effect is based on an advantage to the individual, and it accepts a wide range of mechanisms to achieve such benefits (for examples, see Courchamp et al., 1999; Stephens & Sutherland, 1999). Additionally, a distinction was made between component Allee effects, manifesting positive density dependence on a component of fitness, and demographic Allee effects, concerning total fitness. The latter mostly depict the hump-shaped relationship that is commonly associated with Allee effects, where at some densities the positive density dependence dominates, but at other densities costs outweigh the benefits. The component Allee effect reflects only the isolated mechanism that yields a benefit to aggregation, which might or might not be sufficient to compensate for the connected costs.

If it is expected for a specific species, that the evolution of an aggregation pheromone was driven by positive density dependence at low densities, a component Allee effect can be sought. In controlled experiments, the fitness consequences of isolated mechanisms can be measured, and may provide insight into the selective force that has promoted the use of aggregation pheromone.

The fruit fly Drosophila melanogaster Meigen uses an aggregation pheromone that is emitted by males and mated females (Bartelt et al., 1985; Wertheim, 2001). In a previous field study, potential costs and benefits to the use of an aggregation pheromone in this fly species were identified (Wertheim, 2001). The adult flies aggregate, feed, and breed on decaying fruits (Spieth, 1974). The pheromone induces aggregated oviposition in females (Wertheim, 2001). Consequently, the larvae also have an aggregated distribution across resource substrates. The larvae of D. melanogaster feed on yeasts that develop on a substrate (Brito Da Cunha et al., 1951; Cooper, 1959; Begon, 1986), but at low larval densities, fungi and moulds can overtake a substrate instead (Ashburner, 1989). In laboratory cultures on artificial substrates, larvae suffer high mortality at low larval density (Sang, 1956; Ashburner, 1989).

The aim of the work presented here was to identify a component Allee effect in resource exploitation by D. melanogaster larvae, which might explain the evolution of the use of aggregation pheromone in this species. It is hypothesised that an Allee effect in D. melanogaster larvae can arise from an increased efficiency of a group of larvae to temper fungal growth; however a recent study with larvae of D. subobscura feeding at a natural resource (rowan berries) showed no indication of any facilitation when feeding in groups (Hoffmeister & Rohlfs, 2001). It is possible that the type of substrate on which the larvae develop determines the need for aggregated distributions. In any case, larval density by itself might be insufficient to yield a benefit regarding resource exploitation. In fact, it is likely that adults contribute to the quality of the larval substrates, because a number of Drosophila species is known to inoculate substrates with different species of yeast (Gilbert, 1980; Fogleman & Foster, 1989; Morais et al., 1995).

To test for an Allee effect by larval densities, and a potential added role for adult flies, standard substrates were infested manually with specified numbers of larvae, or standard substrates were infested through oviposition by adult flies. The standard substrates consisted of mashed apple, which constitutes a fairly benign resource for larval development, or apple chunks, which are both more natural and more demanding for larvae to exploit (B. Wertheim, pers. obs.). The observations during the first experiment inspired a more thorough procedure for a second experiment. Before infesting the standard substrates manually with specified numbers of larvae, each standard substrate received a preliminary treatment, consisting of incubation with a specified number of adult flies of specified sex, or the addition of the synthetic aggregation pheromone or its solvent. The fungal growth on the substrates was recorded throughout the larval development. The survival, developmental time, and thorax length of the emerging flies were compared across larval densities and preliminary treatments.

Materials and methods

Insects

The culture of D. melanogaster originated from wild strain individuals reared from apples, collected in an orchard in Wageningen, The Netherlands in 1995. The insects were subsequently reared on an agar gel medium of 40 g yeast, 135 g sugar, and 23 g agar in 1 litre of water with 8 ml nipagin solution added, at 20 ± 2 °C and LD 16:8 h photoperiod. Adult populations were kept in cages (40 × 30 × 30 cm) under the same climatic conditions.

For the experiments, larvae and adults were isolated. To obtain isolated larvae, adults were allowed to oviposit on water agar (20 g agar in 1 litre of water) with a thin layer of living baker's yeast (Fermipan, Gist Brocades, Delft, The Netherlands; 250 mg ml−1 water). The eggs were rinsed from the yeast with water, and after one additional thorough rinse, transferred to Petri dishes with water agar and incubated at 25 ± 2 °C. Within 3 h after hatching, the larvae were introduced to the standard substrates for the experiments (see below). To obtain isolated adult flies, pupae were rinsed from the rearing medium, transferred to vials with water agar, and separated by sex 1 or 2 days prior to emergence (the male sex combes on the front tarsi are visible through the pupal case). The emerged adults were kept for 3–5 days in vials with water agar (8 ml nipagin solution added per litre during preparation) before experimentation, and were provided with honey on a strip of filter paper.

Standard substrates for experiments

Two types of substrate were used in the experiments: mashed apples and apple chunks. Mashed apple was prepared from Golden Delicious apples, ground in a blender with 1.5 dl water added for each litre of pulp. The pulp was frozen on preparation, and defrosted at room temperature before use. Each standard substrate consisted of 8 ml of the pulp in a Petri dish with ridges at the lid to allow some ventilation (5 cm diameter; expt 1) or in a 175-ml container with a ventilation hole, plugged with cotton wool (Greiner, 5 cm diameter; expt 2). The apple chunks [11 ± 2 g (expt 1) or 10 ± 0.5 g (expt 2)] were cut from Golden Delicious apples, such that one side was covered with skin. In one of the unskinned sides, a 1-cm indentation was pressed using a 1-cm diameter glass rod to create a bruised spot where the first-instar larvae could initiate feeding. Each chunk was placed in a glass vial (8 cm height, 4 cm diameter) with a layer of moist vermiculite, and the vials were plugged with cotton wool.

Experiment 1

To create a series of substrates with different larval densities, a specified number of isolated larvae (n = 1, 2, 4, 8, 16, 32) was introduced to a standard substrate (mashed apple or apple chunk) using a small brush. Alternatively, mashed apple substrates were offered to an adult population for oviposition for different periods of time. In the latter case, the resulting eggs were counted to represent the larval density, assuming no egg mortality. The standard substrates were incubated in a climatised room at 20 ± 2 °C and LD 16:8 h photoperiod. The development of fungal growth was categorised (0–10, 25, 50, 75, and 100% of the substrate surface) at 2–4 day intervals during larval development, and emerged adults were counted to calculate the percentage survival. Ten to 20 replicate series were prepared for each of the three treatments. The replicate series that were prepared on the same day were grouped in blocks.

Experiment 2

Standard substrates were treated prior to the introduction of a specified number of isolated larvae (n = 0, 1, 4, 8, 24). The preliminary treatments were according to an incomplete factorial design (Table 1), and consisted of incubation with adult fruit flies in specified numbers of specified sex, or the addition of synthetic aggregation pheromone (cis-vaccenyl acetate, 99% pure, Ipo Pherobank, Wageningen, The Netherlands, 4.5 μg dissolved in 15 μl hexane) or the solvent (15 μl hexane). For the incubation of substrates with adults, adult flies were released in a container (12 × 25 × 10 cm) with five mashed apple substrates that would later constitute a series of larval densities. The number of flies that was released per container was 0, 5, 20, or 40 to obtain an average of zero, one, four, or eight flies per substrate, or zero or four adults were released in each vial with an apple chunk. After incubation for 16–18 h in a climatised room at 21 ± 1 °C, the adults were removed from the substrates. Before introducing the isolated larvae, the eggs that were deposited by the virgin females on the mashed apple substrates were removed using a pair of tweezers; the other mashed apple substrates were also disrupted with pairs of tweezers. After introduction of the larvae, the standard substrates were incubated in a glasshouse at 20 ± 3 °C at ambient spring light conditions. Several series of larval densities for each treatment were prepared on four different days (block factor), resulting in 10–17 replicate series per treatment in total.

Table 1.  The preliminary treatments for the standard substrates in expt 2.

Abbreviation

Standard substrate

Preliminary treatment
Factorial design
Replications
AdultsSexPheromone
MA/0FMashed appleNo adults0No17
MA/1FMashed apple1 virgin female1FemaleNo17
MA/4FMashed apple4 virgin females4FemaleNo17
MA/8FMashed apple8 virgin females8FemaleNo17
MA/4MMashed apple4 virgin males4MaleYes16
MA/cvaMashed appleSynthetic pheromone0Yes16
MA/hexMashed appleHexane (solvent)0No16
CH/0FApple chunksNo adults0 No10
CH/4FApple chunks4 virgin females4 No10

The percentage of substrate area that was covered with fungal growth was recorded at days 1, 3, 5, 7, 10, and 12 after introducing the larvae. The emerged adults were counted and collected daily from day 9 to day 13 and at 16 days after introducing the larvae, and stored in ethanol. The thorax length of the emerged flies was measured from the base of the most anterior humeral bristle to the posterior tip of the scutellum (French et al., 1998) under a stereo-microscope with an eyepiece micrometer. To analyse the differences in developmental time and in thorax length of the emerged flies (expt 2), the average values per standard substrate were calculated (separately for male and female thorax length).

Statistics

The data were analysed using generalised linear models (GLM, SAS, v. 6.1, Proc Genmod, with dscale option to correct for overdispersion when the scaled deviance exceeded a value of 1). For the percentage of substrate covered with fungal growth and the percentage survival, a binomial distribution with logit link function was specified. For the averaged developmental time and averaged male and female thorax lengths per standard substrate, a normal distribution with identity link was specified. The data were analysed separately for the different types of standard substrate (mashed apples, apple chunks), the isolated larvae and oviposited eggs (expt 1), and the data for expts 1 and 2. The block factors were included in all analyses. The factors that were tested were: number of larvae, number of adults, sex of the adults (nested), quadratic and third-order terms of numbers of larvae and adults, and pheromone presence (see also Table 1). The number of larvae was log-transformed [Ln(larvae + 1)] to avoid strong leverage for the high host density. Second- and third-order terms were included to capture non-linear relationships between density and a fitness component, such as asymptotic or hump-shaped relationships. The results for the minimal adequate models are presented. The minimal adequate model was obtained by backward elimination of all those factors from the full model, for which removal caused an insignificant increase in deviance (F-test), starting at higher-order terms; the F-values of the significant terms describe the increase in variance compared with the previous (stripped-down) model from which the term was removed (Crawley, 1993). The degree of fungal cover was subsequently added to the minimal adequate model to test for an added explanatory value of the fungal growth, above and beyond the other factors.

Results

Experiment 1

In mashed apple, the larvae experienced competition rather than an Allee effect. The survival of larvae was negatively density dependent for both the eggs, oviposited by flies (first-order term: F1,99 = 16.13, P < 0.001; second-order term: F1,99 = 16.19, P < 0.001) and the isolated larvae, introduced manually (first-order term: F1,58 = 10.74, P < 0.01). At first sight, the density-dependent pattern between the treatment with eggs oviposited by flies (Fig. 1a) and the treatment with manually infested larvae (Fig. 1b) appeared to be different, with a hump shape in the former (Fig. 1a), suggesting an Allee effect. The statistical models, however, described monotonically declining survival rates for both cases, indicating competition. Nonetheless, the observation prompted a more close examination of the effect of adults on substrates in relation to larval development. In the apple chunks, survival was not density dependent (first-order term: F1,97 = 0.24, P > 0.05; Fig. 1c). The development of fungi (mainly Mucor spp., Penicillium spp., and Aspergillus spp.) was reduced strongly at high larval density in all the treatments (mashed apple, oviposited eggs, second-order term: F1,99 = 5.23, P < 0.05; mashed apple, isolated larvae, first-order term: F1,58 = 246.55, P < 0.001; apple chunks, first-order term: F1,96 = 19.96, P < 0.001).

Figure 1.

The survival of Drosophila melanogaster larvae at different larval densities in standard substrates (expt 1). The different densities were obtained by (a) offering mashed apple substrate patches to adult fruit flies for varying amounts of time or (b) by introducing specified numbers of isolated larvae manually on mashed apple substrates or (c) on apple chunks. See methods for explanation. The average percentage survival and 95% CI (error bars) were calculated on angular-transformed data and back-transformed.

Experiment 2

In mashed apple, again, no Allee effect was found for larval density on any fitness component. For adult density, however, a significant Allee effect was found on larval survival. The survival of the larvae was related positively to the numbers of adults that were incubated on the substrates prior to the introduction of larvae (first-order term: F1,449 = 5.13, P < 0.05; Fig. 2a). For larval density, only negative relationships with fitness components were found, indicating strong competition. Increased larval density reduced survival (first-order term: F1,449 = 18.06, P < 0.001; Fig. 2a), prolonged developmental time (first-order term: F1,435 = 4.51, P < 0.05; second-order term: F1,432 = 17.43, P < 0.001; Fig. 2b), and reduced the thorax length of emerging females (second-order term: F1,369 = 12.20, P < 0.001; Fig. 2c). The density of adults (first-order term: F1,371 = 7.97, P < 0.01) and the interaction term between larval and adult density (F1,370 = 5.59, P < 0.05) were significant in the statistical model for female thorax length, indicating a negative relationship between adult density and size at low larval density, and a positive relationship at high larval density. The thorax length of males described a slight hump-shaped relationship to larval density (first-order term: F1,367 = 5.93, P < 0.05; second-order term: F1,363 = 5.49, P < 0.05; third-order term: F1,362 = 4.09, P < 0.05; Fig. 2d). Adult density did not contribute to the explanatory power of this model. At higher adult densities, developmental period was increased slightly (first-order term: F1,435 = 6.36, P < 0.05).

Figure 2.

Average fitness components for larvae reared at different densities in mashed apple standard substrates, which had been incubated with varying densities of adults, prior to the introduction of the larvae (see methods). (a) The survival (+ 95% CI) and (b) developmental time (± SEM) of Drosophila melanogaster larvae and the thorax length (± SEM) of emerging (c) females and (d) males. The open symbols depict the averages from the data; the closed symbols depict the minimal adequate model, as calculated using generalised linear models (GLM). For clarity, the different series are shifted slightly around the value of x (larval density).

The development of fungi on the mashed apple was influenced by the density of larvae, and by the density and sex of adults during the preliminary treatments. Increased larval and adult densities strongly reduced the percentage of substrate that was covered with fungi (for larvae: significant first-, second-, and third-order terms from the third day onwards; for adults: significant first- and second-order terms during the first 5 days; see Table 2 and Fig. 3). Incubation with virgin females reduced fungal growth more strongly than incubation with virgin males (significant on day 1 and day 7, statistical trend on day 5; Table 2). Survival of the larvae was, above and beyond the larval and adult effects, related negatively to the degree of fungal cover on the substrate during the early development of larvae (F1,339 = 11.14, P < 0.001, fungal cover at day 3). After the pupation of larvae (from approximately day 6), fungal growth increased and short developmental times correlated with a higher degree of fungal cover during late larval development (significant from day 7 onwards; day 7: F1,435 = 13.31, P < 0.001; day 10: F1,435 = 6.67, P < 0.05; day 12: F1,388 = 9.85, P < 0.01). The degree of fungal cover during very early development had a slight positive relationship with female thorax length (at day 1: F1,231 = 4.71, P < 0.05). The presence of pheromone did not contribute to explanatory power in any of the described statistical models (P > 0.05).

Table 2.  The development of fungi on the two standard substrate types in expt 2. For each census day, the significant factors in the minimal adequate models for fungal cover are presented. Adult denotes the number of adults during the preliminary treatment, with the sex of the adults nested (see also Table 1 ), larva is the log-transformed density of larvae, NA = not applicable; – indicates that the factor was not significant in the particular substrate type.
Significant factors in minimal adequate modelMashed appleApple chunks
F statistic P -value F statistic P -value
Day 1
AdultF1,358 =13.37 P <0.001 Residuals indicate unreliability of statistical results
Adult×adultF1,354 =11.07 P <0.001
Adult(sex)F1,357 =6.18 P <0.05
Day 3
LarvaF1,428 =112.28 P <0.001
Larva×larva×larvaF1,424 =5.03 P <0.05
AdultF1,428 =19.01 P <0.001 F1,76 =4.83 P <0.05
Adult×adultF1,424 =4.77 P <0.05 NA
Day 5
LarvaF1,565 =81.47 P <0.001 F1,94 =45.48 P <0.001
Larva×larvaF1,563 =11.22 P <0.001
Larva×larva×larvaF1,562 =4.19 P <0.05
AdultF1,565 =7.96 P <0.01 F1,94 =7.75 P <0.01
Adult(sex)F1,564 =3.70 P <0.055 NA
Day 7
LarvaF1,565 =70.92 P <0.001 F1,94 =54.46 P <0.001
Larva×larvaF1,563 =19.05 P <0.001
Larva×larva×larvaF1,562 =16.42 P <0.001
Adult F1,94 =8.09 P <0.01
Adult(sex)F1,563 =5.06 P <0.05 NA
Day 10
LarvaF1,567 =59.22 P <0.001 F1,94 =54.91 P <0.001
Larva×larvaF1,563 =14.42 P <0.001
Larva×larva×larvaF1,562 =11.33 P <0.001
Adult F1,94 =6.84 P <0.05
Day 12
LarvaF1,501 =37.97 P <0.001 F1,86 =30.39 P <0.001
Larva×larvaF1,499 =11.99 P <0.001
Larva×larva×larvaF1,498 =11.15 P <0.001
AdultF1,501 =7.22 P <0.01
Adult×larvaF1,499 =13.39 P <0.001
Figure 3.

Fungal growth on mashed apple standard substrates with different preliminary treatments and different numbers of larvae. The average percentages of substrate that were covered with fungi were calculated on angular-transformed data and back-transformed.

In the apple chunks, again no Allee effect was found for larval density, but significant Allee effects were found for adult density on larval fitness components. In substrates that had been incubated with adult flies, both the survival (F1,74 = 9.55, P < 0.01; Fig. 4a) and thorax length of emerging adults (females: F1,56 = 16.91, P < 0.001, Fig. 4c; males: F1,58 = 7.53, P < 0.01, Fig. 4d) were increased considerably. The larvae experienced competition at increasing larval densities, expressed by a reduced survival (statistical trend, F1,74 = 3.72, P = 0.055; Fig. 4a), longer developmental times of the flies (F1,71 = 6.82, P < 0.05; Fig. 4b) and reduced thorax length in emerging females (F1,56 = 6.02, P < 0.05; Fig. 4c). Male thorax length was not affected significantly by larval density (P > 0.05; Fig. 4d).

Figure 4.

Average fitness components for larvae reared at different densities in apple chunks, which had been incubated with zero or four adult flies, prior to the introduction of the larvae (see methods). (a) The survival (+ 95% CI) and (b) developmental time (± SEM) of Drosophila melanogaster larvae and the thorax length (± SEM) of emerging (c) females and (d) males at different larval densities. The open symbols depict the averages from the data; the closed symbols depict the minimal adequate model, as calculated using generalised linear models (GLM). For clarity, the different series are shifted slightly around the value of x (larval density).

The development of fungal growth on apple chunks (Fig. 5) was reduced after incubation with adult flies (significant from the third day onwards; Table 2) and with increasing larval densities (first-order terms significant from the fifth day onwards; Table 2). On top of the effects of larval and adult density, a high degree of fungal cover correlated with longer developmental times (day 5: F1,70 = 5.14, P < 0.05; day 7: F1,70 = 7.22, P < 0.01) and smaller thorax length for both males (day 5: F1,57 = 9.65, P < 0.01; day 7: F1,57 = 5.82, P < 0.05) and females (day 5: F1,55 = 17.63, P < 0.001; day 7: F1,55 = 19.49, P < 0.001).

Figure 5.

Fungal growth on apple chunks with different preliminary treatments and different numbers of larvae. The average percentage of substrate area that was covered with fungi was calculated on angular-transformed data and back-transformed.

Discussion

Resource exploitation by increasing densities of D. melanogaster larvae in itself did not indicate a component Allee effect, but rather competition, just as was found for D. subobscura (Hoffmeister & Rohlfs, 2001). Increased larval densities reduced survival, increased developmental time, and yielded smaller emerging flies. For adult density, however, a distinct component Allee effect was demonstrated. Increased density of adult flies on substrates, prior to the introduction of larvae, enhanced larval survival and, in apple chunks, yielded larger flies. For Drosophila, larger body size confers a higher fitness (Partridge et al., 1986; McCabe & Partridge, 1997; Reeve et al., 2000). Thus, through the effects of adult density on larval fitness components, significant benefits for the aggregated oviposition in adult female fruit flies were revealed. Whether these benefits are sufficient to attribute the evolution of aggregation pheromone in D. melanogaster to this component Allee effect will be discussed in relation to the interactions with fungi (see below). A high degree of fungal cover was related statistically to reduced larval survival in mashed apple and prolonged developmental times and reduced sizes of the emerging flies in apple chunks. Both the prior presence of adults on a substrate and high larval densities reduced fungal growth.

Interactions among density of adults, larvae, and micro-organisms

A complex of direct and indirect interactions among densities of adults, larvae, and micro-organisms may have led to the Allee effect that was shown. The hypothesis that an Allee effect in larval resource exploitation can arise from an adverse interaction with fungi growing on the resource was supported by the negative relationship between larval fitness components and fungal cover. Yet, this relationship is statistical, leaving the cause undetermined: was the reduction in larval survival caused by the fungi or did the fungi prosper because the larvae had died? The literature indicates that fungi can have deleterious effects on Drosophila larvae (Atkinson, 1979; Ashburner, 1989; Hodge et al., 1999), and the diminishing size and increased developmental time of the survivors for increasing degrees of fungal cover in apple chunks suggest the same. From this, it is inferred that the fungi were harmful to the larvae, and a reduction in fungal cover is considered as an enhancement of the larval resource.

It is unclear whether the fungi exerted a direct detrimental influence on the larvae through, for example, toxins, or indirectly through an interaction with the growth of yeast on which the larvae feed. Similarly, the reduction in fungal growth by prior adult presence on a substrate might have been caused directly by the inoculation of an anti-fungal agent, or indirectly through inoculation of yeasts. Fungi and yeast compete for resources in fruit, and any process that favours one of the competitors could result in a shift in the competitive interaction. The tunnelling activities of larvae may destroy fungal mycelium, while the inoculation of substrates with yeast by adults might give the yeast a head start. There is some evidence to suggest a mutualistic interaction between Drosophila and yeasts, which might eventually suppress the development of fungal growth: (1) Yeast development is faster within insect-infested resources (Fogleman & Foster, 1989) and the succession in yeast communities is driven through inoculations by adult Drosophila (Morais et al., 1995). (2) When baker's yeast was added to standard substrates of mashed apple, no fungal growth was visible (B. Wertheim, unpublished). (3) Female fruit flies were reported to vector a higher diversity of yeasts than males (Morais et al., 1995), which corresponds with the observations of a stronger effect on fungal growth of prior incubation with females than with males. It is therefore likely that the positive effect of adult density on larval fitness arose partly from the inoculation of yeasts, which also reduced fungal growth. It should be noted that the experimental procedure was conservative with respect to yeast inoculation by adults: the adults were not provided with living yeasts and any yeast they inoculated was either obtained as larva or derived from the (water rinsed) pupal cases and honey. Adults under natural conditions can obtain yeasts from other fruits and alternative sources, potentially enlarging their beneficial effects on a larval resource. The effects of fungal cover and adult density were not fully exchangeable in the statistical models, indicating that other processes also operated.

The timing of yeast and larval introductions in a substrate could influence the outcome of the multiple interactions. It is conceivable that the enhancement of a substrate by adults through inoculation with yeasts can only succeed up until a certain degree of fungal infestation. With the experimental procedure that was used, the natural temporal sequence in the species interactions was mimicked: Drosophila melanogaster oviposits on fruit in the early stages of decay when no major fungal growth is yet apparent. Several studies have shown that the oviposition time window in Drosophila is rather narrow and species specific (Nunney, 1990; Shorrocks & Bingley, 1994), which may be related to the succession and development of fungi and yeasts. Moreover, the interaction between D. melanogaster larvae and fungi facilitated subsequent resource exploitation for D. hydei larvae (Hodge et al., 1999), potentially promoting association between the fruit fly species. This illustrates the intricate nature of several density-dependent interactions that are affected by the use of aggregation pheromones in fruit flies.

The Allee effect and the evolution of aggregation pheromone

When it is accepted that fungi are indeed detrimental for larval development, directly or indirectly, the results support the hypothesis that a benefit of aggregated oviposition can arise through suppressing fungal growth. This component Allee effect arises at increased adult densities. Whether the aggregation of ovipositing females enhances the resource quality sufficiently to outweigh the negative effects of increased larval competition will in general depend on the characteristics of the resource (e.g. carrying capacity, constitution, resilience), the residing and invading micro-biota (density, succession, and species composition), and other insect species that utilise the resource (e.g. competitors or facilitators, their clutch sizes, timing of invasion, and the degree to which competition is contest or scramble type). It is noteworthy that the beneficial effects of prior presence of adults were clearly more pronounced in the apple chunks (a demanding substrate for the larvae) than in the mashed apple substrates (a fairly benign substrate). Furthermore, the marked decrease in fungal development at high larval density might result in additional benefits, when the fungal species are more noxious than those that infested the substrates in these laboratory experiments. The balance between costs and benefits is thus largely dependent on a variety of ecological conditions and will differ for each resource in each environment.

To translate the results on component Allee effects to the scale of total fitness, i.e. to a demographic Allee effect, would require many additional experiments, for example on other fitness components [an inadequate yeast diet can cause reduced and delayed fertility and fecundity (Dudgeon, 1954; Cooper, 1959)] and in other substrate types (Courtney et al., 1990). Furthermore, stochasticity in mortality risk should be incorporated explicitly in an analysis for a demographic Allee effect because aggregation might reduce mortality only under adverse conditions, while the advantage is lost under ideal conditions (Lockwood & Story, 1986). Moreover, all ecological interactions that are affected by the use of aggregation pheromone must be incorporated in a cost–benefit analysis, to judge when the use of aggregation pheromone would be promoted by natural selection. As such, it is meaningless to calculate some fitness index that describes the combined positive and negative density dependence for this laboratory experiment, because these costs and benefits vary largely for different ecological conditions.

This leaves the answer incomplete to the question of whether the use of aggregation pheromone in D. melanogaster can be explained by an Allee effect for larval resource exploitation after aggregated oviposition; however a number of prerequisites appears to be met in this system. As stated in the introduction, aggregative behaviour can evolve provided that benefits outweigh all costs even at low densities. The mechanism that was investigated in this study could produce such a scenario. When on average the fitness of an ovipositing female, comprising the quantity and quality of its offspring, is enhanced when another female shares the same oviposition substrate, both females benefit when seeking out each other. The aggregation pheromone in Drosophila is produced by males and transferred to females during copulation, together with many other male accessory gland products (Bartelt et al., 1985; Partridge et al., 1986; Chapman et al., 1995). Therefore, there are no physiological costs of producing the pheromone for the female. Recently mated females have both a high emission rate of pheromone and a high oviposition rate (Bartelt et al., 1985; Partridge et al., 1986), making the pheromone a reliable indicator of the presence of other gravid females. Finally, aggregating insects can be more successful in transmitting micro-organisms and causing rot than single individuals (Fleischer et al., 1999). Because rot and fermentation are exactly what renders a substrate suitable for Drosophila larval development (Atkinson & Shorrocks, 1977), this is in strong agreement with the suggestion that aggregation pheromone has evolved in response to a benefit of aggregated oviposition. Clearly, the topic deserves much more study to address all questions. Such an effort may also shed light on the numerous other insect species that both possess aggregation pheromone and have a mutualistic interaction with micro-organisms (Wertheim, 2001). The evolution of aggregation pheromone may somehow be facilitated or even directed by the interaction with micro-organisms.

Acknowledgements

We gratefully acknowledge Dine Volker for identifying the fungi, Frodo Kindt for lending the eyepiece micrometer, and Leo Koopman for culturing the flies. Thomas Hoffmeister and an anonymous referee gave valuable comments on an earlier version of the manuscript.

Ancillary