Introduction
Parasitoid fitness strongly depends on the availability and quality of hosts because hosts provide all resources required for larval development (Caron et al., Reference Caron, Myers and Gillespie2010). Host quality may, for instance, affect parasitoid oviposition rate, offspring survival, longevity, sex ratio, body size and fecundity (King, Reference King2000; Silva-Torres & Matthews, Reference Silva-Torres and Matthews2003; Silva-Torres et al., Reference Silva-Torres, Ramos Filho, Torres and Barros2009). Consequently, a female's fitness returns depend on her ability to identify suitable hosts, viz. on the process of host selection during which host quality is assessed (Hassell, Reference Hassel2000; Caron et al., Reference Caron, Myers and Gillespie2010). Several factors may determine the quality of a host, with a particularly important one being host size, as larger hosts provide more food to developing larval parasitoids (Hardy et al., Reference Hardy, Griffiths and Godfray1992; Zaviezo & Mills, Reference Zaviezo and Mills2000; Silva-Torres et al., Reference Silva-Torres, Ramos Filho, Torres and Barros2009). Another important factor affecting host quality is whether a given host has been parasitized previously (Silva-Torres et al., Reference Silva-Torres, Ramos Filho, Torres and Barros2009). Larvae developing in superparasitised hosts may compete for limited resources, which may result in a fitness reduction for each individual through various mechanisms, including smaller size and lower fecundity (Charnov & Skinner, Reference Charnov and Skinner1984; van Alphen & Visser, Reference van Alphen and Visser1990; Godfray, Reference Godfray1994).
To avoid detrimental effects of superparasitism, many parasitoid species seem to have evolved the ability to discriminate between parasitized and unparasitized hosts (Ueno & Tanaka, Reference Ueno and Tanaka1994; Bell et al., Reference Bell, Marris, Prickett and Edwards2005; Silva-Torres et al., Reference Silva-Torres, Ramos Filho, Torres and Barros2009). In solitary parasitoids, surplus eggs or larvae are often even killed by physical attack or physiological suppression (Hubbard et al., Reference Hubbard, Marris, Reynolds and Rowe1987; Godfray, Reference Godfray1994; Mackauer & Chau, Reference Mackauer and Chau2001). Therefore, superparasitism in solitary species was historically considered as a failure to discriminate between parasitized and unparasitized hosts (van Lenteren, Reference van Lenteren1976). Parasitoids may avoid superparasitism by means of early patch leaving (Rosenheim & Mangel, Reference Rosenheim and Mangel1994), transient paralysis (Desneux et al., Reference Desneux, Barta, Delebecque and Heimpel2009) or by detecting scent marks (Hubbard et al., Reference Hubbard, Harvey and Fletcher1999). However, the ability to discriminate may not necessarily result in an avoidance of superparasitism (Montoya et al., Reference Montoya, Liedo, Benrey, Barrera, Cancino and Aluja2000, Reference Montoya, Benrey, Barrera, Zenil, Ruiz and Liedo2003; Burton-Chellew et al., Reference Burton-Chellew, Koevoets, Grillenberger, Sykes, Underwood, Bijlsma, Gadau, van de Zande, Beukeboom, West and Shuker2008); and, at least in some parasitoids, host discrimination is only poorly developed (Caron et al., Reference Caron, Myers and Gillespie2010). Further, superparasitism may, under certain circumstances, increase offspring number (Mackauer & Chau, Reference Mackauer and Chau2001; Gu et al., Reference Gu, Wang and Dorn2003; Silva-Torres & Matthews, Reference Silva-Torres and Matthews2003; Keasar et al., Reference Keasar, Segoli, Barak, Steinberg, Giron, Strand, Bouskila and Harari2006). Thus, detrimental effects on individual offspring may be counterbalanced by higher offspring numbers, resulting in a net fitness gain for female parasitoids (Vet et al., Reference Vet, Datema, Janssen and Snellen1994; Silva-Torres et al., Reference Silva-Torres, Ramos Filho, Torres and Barros2009). Consequently, superparasitism may actually be advantageous to females, at least when hosts are scarce and when egg load is concomitantly high (Weisser & Houston, Reference Weisser and Houston1993; Yamada & Miyamoto, Reference Yamada and Miyamoto1998; Mackauer & Chau, Reference Mackauer and Chau2001; Silva-Torres et al., Reference Silva-Torres, Ramos Filho, Torres and Barros2009). Superparasitism is predicted to increase with decreasing host abundance (Hubbard et al., Reference Hubbard, Harvey and Fletcher1999) and increasing egg load (Sirot et al., Reference Sirot, Ploye and Bernstein1997). If females are time-constrained and face a high risk that not all eggs can be deposited, discriminating between parasitized and unparasitized hosts may not pay off (Hughes, Reference Hughes1979).
The occurrence of self-superparasitism (i.e. depositing an egg on a host that has been previously parasitized by the same female) is even more difficult to understand, as it will inevitably induce competition among siblings. Self-superparasitism may result from not discriminating between parasitized and unparasitized hosts, or alternatively from females laying multiple-egg clutches (Rosenheim & Hongkham, Reference Rosenheim and Hongkham1996; Mackauer & Chau, Reference Mackauer and Chau2001). Even self-superparasitism (or laying multiple egg clutches) may be beneficial to female parasitoids, as long as the overall fitness gain is higher in hosts receiving more than one compared to those only receiving one egg (Ito & Yamada, Reference Ito and Yamada2005). This may even be the case in solitary parasitoids, if (i) more eggs increase the probability of host rejection by conspecific females, thus protecting a female´s offspring from competition, if (ii) the chance that a female´s offspring succeed in competition with conspecific offspring increases with the number of eggs laid, or if (iii) the risk that all offspring are killed by a conspecific female or the host's immune response decreases with increasing egg number (van Alphen & Visser, Reference van Alphen and Visser1990; Godfray, Reference Godfray1994; Rosenheim & Hongkham, Reference Rosenheim and Hongkham1996; Mackauer & Chau, Reference Mackauer and Chau2001).
Against the above background, we here explore self-superparasitism in relation to host abundance in the solitary parasitic wasp Spalangia cameroni (Hymenoptera: Pteromalidae). This species is used worldwide for biological control of house and stable flies (Skovgard & Nachman, Reference Skovgard and Nachman2004; Birkemoe et al., Reference Birkemoe, Soleng and Aak2009). Specifically, we predicted that self-superparasitism will increase with decreasing host abundance. By scoring lifetime offspring production, we further investigated the fitness consequences of depositing more than one egg per host in this species, thus testing whether self-superparasitism may exert beneficial effects in a solitary parasitoid, as has been predicted by some researchers (van Alphen & Visser, Reference van Alphen and Visser1990; Ito & Yamada, Reference Ito and Yamada2005). As host, we used an economically important pest species of fruits, the Mediterranean fruit fly Ceratitis capitata (Diptera: Tephritidae) (Fimiani, Reference Fimiani, , and Hooper1989; Liquido et al., Reference Liquido, Shinoda and Cunningham1991).
Materials and methods
Study organisms
The parasitoid S. cameroni was originally described from the Hawaiian Islands, but has a worldwide distribution owing to intentional introductions to control filth fly populations (Boucek, Reference Boucek1963). Females are synovigenic (eclosing with 12–40 mature eggs), and destructive host feeding is required to mature additional eggs (Gerling & Legner, Reference Gerling and Legner1968; King, Reference King2002). This species is considered a solitary ectoparasitoid, as the larvae develop on the host inside the puparium of various dipteran species, with only one larva completing development per host (Boucek, Reference Boucek1963; Gerling & Legner, Reference Gerling and Legner1968). Adults probably feed on nectar or pollen because they readily accept honey in laboratory experiments (e.g. Legner & Gerling, Reference Legner and Gerling1967; King, Reference King2000; Tormos et al., Reference Tormos, Beitia, Böckmann and Asis2009).
The experimental host, C. capitata, has a wide distribution, including the Mediterranean area, where it attacks more than 250 commercially-gown fruits (Fimiani, Reference Fimiani, , and Hooper1989; Fletcher, Reference Fletcher, Robinson and Hooper1989; Liquido et al., Reference Liquido, Shinoda and Cunningham1991). It has up to five generations per year in favourable areas of Spain (Muñiz & Gil, Reference Muñiz and Gil1984) and produces usually about 300 (but up to 1000) eggs per female (Weems, Reference Weems1981; Fletcher, Reference Fletcher, Robinson and Hooper1989). We are currently studying S. cameroni as a potential biological control for this economically important pest (Pérez-Hinarejos & Beitia, Reference Pérez-Hinarejos and Beitia2008; Tormos et al., Reference Tormos, Beitia, Böckmann and Asis2009, Reference Tormos, Beitia, Alonso, Asís and Gayubo2010). The parasitoids and hosts used here originated from colonies maintained at the Instituto Valenciano de Investigaciones Agrarias (IVIA) in Moncada (Valencia, Spain). The parasitoid colony was founded in 2003 with individuals that had emerged from pupae of C. capitata that had been collected in the field near Bétera (Valencia) (Falcó et al., Reference Falcó, Verdú and Beitia2004). The C. capitata host colony was established in 2002 by collecting attacked fruits at various locations in the province of Valencia.
Experimental design
Throughout all experiments, parasitoids were reared in a climate cabinet at 24.5±0.5°C, 60±10% relative humidity, and a 16 h:8 h light-dark cycle. They were maintained in translucent plastic boxes (1 litre) covered by muslin for ventilation. Honey and mineral water were provided for adult feeding ad libitum. For experiments, only freshly eclosed parasitoids (<20 h) were used. Host pupae were offered in Petri dishes (Ø 60 mm) located in the centre of each box. All host pupae used were 3–5 days old in order to minimize confounding effects of host age (King, Reference King1998). To further minimize effects of host size on sex ratio (King & King, Reference King and King1994), only pupae of similar size and colour (as judged visually) were used. All material (Petri dishes, boxes, etc.) used in the experiments was either new or washed in de-ionized water prior to use.
To investigate self-superparasitism in relation to host abundance, two different experiments (A and B) were carried out as detailed below. Both experiments used one parasitoid pair (i.e. one male and one female) per box and two treatments, with parasitoid pairs being provided either two (treatment 2P) or ten host pupae (10P) per day. While experiment A focused on the number of S. cameroni offspring produced, experiment B focused on the number of eggs laid per host. Carrying out two experiments was necessary because counting parasitoid eggs is impossible without dissecting the hosts and, thus, harming parasitoid offspring. In experiment A, Petri dishes with host pupae were removed daily, tightly sealed, checked ten days later for the number of eclosed (=surviving) hosts and additionally 40 days later for the number and sex of eclosed parasitoids. The latter time span is sufficient to allow all individuals to finish development (Moon et al., Reference Moon, Berry and Peterson1982; Tormos et al., Reference Tormos, Beitia, Böckmann and Asis2009). Additionally, the number of dead host pupae which had not died from parasitation (but e.g. from probing or host feeding) was recorded. Experiment B used the same setup as above, but here all host pupae were dissected soon after removal from the boxes (typically within 24 h) in order to identify (i) the proportion of pupae parasitized, (ii) the number of eggs per pupa, and (iii) the numbers of intact (being regularly shaped) and dead eggs (being more irregular) per pupa.
Thirty replicate boxes (=parasitoid pairs) were used per treatment and experiment, resulting in a total of 120 boxes (2 experiments×2 treatments×30 replicates). Due to space limitations within the climate cabinet, only 40 boxes, though, could be used at a time. Therefore, experiments were divided into three consecutively analyzed blocks, with each block containing ten replicates per treatment and experiment. Throughout, females were provided with host pupae until day 28 of adult life or death (if females died before day 28). A time span of 28 days was chosen because no later egg-laying was observed in the 20 females used in the first block (experiment B). Therefore, our data reflect lifetime reproductive investment throughout. Males that had died before females were replaced to assure egg fertilisation. Throughout both experiments, baseline mortality of host pupae was scored using 20 pupae not having had any contact to parasitoids per day.
Statistical analyses
From the 120 replicates in total (2 experiments×2 treatments×10 replicates×3 time blocks), five had to be excluded from further analyses, as either no offspring (experiment A, three replicates) or no eggs (experiment B, two replicates) were produced during the entire female lifespan. Differences in the number of offspring, parasitized hosts, hosts killed additionally, total number of eggs per female, eggs per pupa, and the percentage of dead eggs across treatment groups were tested using general linear models, with treatment as a fixed factor and block (nested within treatment) as a random factor. Note that using generalized models resulted in qualitatively identical results. Differences in treatments over time were tested by repeated measures ANOVAs, using the first 20 days of the oviposition period (a longer period would lead to an increasing reduction in sample size through death). Obtained offspring sex ratios were tested against even sex ratios using chi-square tests. Throughout the text, means are given ±1 SE. All statistical tests were calculated using SPSS (version 12.0; Inc., 2004; Chicago, IL, USA) or JMP (version 7.0.1; SAS Institute, 2007; Cary, NC, USA).
Results
Females provided with ten host pupae per day laid significantly more eggs (63.7±3.8 vs 41.2±2.8), parasitized significantly more hosts (49.4±3.0 vs 17.9±0.9), killed significantly more hosts in addition to parasitism (22.6±1.6 vs 7.8±0.7) and produced significantly more offspring (27.3±1.7 vs 13.8±0.9) compared to females provided with two pupae per day (table 1). Treatment effects were strikingly pronounced during the first ten days of the oviposition period, as indicated by significant interactions between treatment and time in repeated measures ANOVAs (number of eggs laid: F19,33=3.2, P=0.0018; pupae parasitized: F19,33=9.5, P<0.0001; additionally killed pupae: F19,36=3.8, P=0.0003; offspring produced: F19,36=8.7, P<0.0001; fig. 1). Offspring sex ratio (M/F) was female-biased in both treatments (1:1.7 each; treatment 2P: Chi2=24.9, P<0.0001; treatment 10P: Chi2=13.4, P=0.0002). The mean number of eggs per pupa (1.3±0.03 vs 2.1±0.08) was significantly lower when provided ten compared to two host pupae, and there was an according tendency for the proportion of all eggs that was dead (9.0±1.2% vs 18.1±1.2%; table 1). Although superparasitism (defined here as the presence of >1 egg per pupa) remained higher throughout the oviposition period in females being provided two host pupae per day (non-significant interaction between treatment and time in repeated measures ANOVA: F19,56=0.26, P=0.99), differences in the proportion of dead eggs were particularly pronounced during oviposition days 4–11 (interaction: F19,51=1.8, P=0.0209; fig. 2). Host mortality in the controls without contact to parasitoids was low and constant over time (3.0%±0.4%). In contrast, host mortality (in addition to successfully parasitized pupae) in experiment A was much higher (23.2%±1.3% in treatment 10P; 54.7%±2.0% in treatment 2P; table 1).
Number of (a) eggs laid, (b) pupae parasitized, (c) additionally killed pupae and (d) offspring produced (means±1 SE) by Spalangia cameroni over time in relation to host abundance (2P, two host pupae per day; 10P, ten host pupae per day) (–▪–, 10P; 
, 2P).
Percentage of self-superparasitism: (a) more than one parasitoid egg per host pupa; and (b) dead parasitoid eggs (means±1 SE) for Spanlangia cameroni over time in relation to host abundance (2P, two host pupae per day; 10P, ten host pupae per day). Data after day 20 were excluded due to low sample size (–▪–, 10P; , 2P).
Results of a general linear models analysis testing for the effects of lifetime exposure of S. cameroni to two vs ten pupae of C. capitata per day (treatment) and experimental block (random effect, nested within treatment). Variables examined were numbers of eggs laid, hosts parasitized, the proportion of dead eggs, eggs per host (all data from experiment B), offspring produced and hosts killed in addition to those successfully parasitized (data from experiment A). Significant P-values are given in bold. Significant block effects indicate trait variation over time.
Note that, based on the above mean values, offspring production per egg laid was slightly higher in the 10P compared to 2P treatment (0.43 vs 0.33 offspring per egg). However, offspring production per pupa parasitized was higher in the 2P (0.77 offspring per pupa) compared to the 10P treatment (0.55). As the data on offspring produced (experiment A) and eggs laid per pupae parasitized (B) stem from different experiments, a formal statistical comparison of the above ratios is difficult. Using chi square tests based on the total numbers of offspring produced, eggs laid and pupae parasitized for illustrative purposes here reveals significant differences in the proportion of offspring per egg laid (χ21=23.2, P<0.0001) and of offspring per pupa (χ21=8.2, P=0.0041).
Discussion
As expected and in accordance with earlier findings on parasitoids, our results show that lifetime fecundity (egg laid), the number of pupae parasitized and offspring numbers were all positively affected by host abundance (Legner, Reference Legner1969; He et al., Reference He, Teulonz and Wang2006; Dannon et al., Reference Dannon, Tamo, van Huis and Dicke2010; Mahmoudi et al., Reference Mahmoudi, Sahragard and Sendi2010). Female-biased offspring sex ratios as found here have also been repeatedly documented for S. cameroni and other parasitoids (e.g. Moon et al., Reference Moon, Berry and Peterson1982; King, Reference King1989; Carleton et al., Reference Carleton, Quiring, Heard, Hebert, Delisle, Berthiaume, Bauce and Royer2010; Peters, Reference Peters2010). Regarding parasitation rates, it should be noted that, in both treatments, eggs were distributed unevenly across hosts, i.e. on some pupae more than one egg was deposited while others were not used at all for oviposition. Thus, an obvious question is why all available hosts were not used by S. cameroni? One explanation is that S. cameroni needs hosts not only for parasitation but also for host feeding (alimentation). Host feeding is essential for triggering egg production but is destructive and, therefore, excludes using the same host for oviposition (non-concurrent destructive host feeders: Gerling & Legner, Reference Gerling and Legner1968; Shi et al., Reference Shi, Zang, Liu, Ruan and Sun2009; Zang & Liu, Reference Zang and Liu2010). Consequently, females are forced into a trade-off between alimentation and reproduction when hosts are scarce (Heimpel & Collier, Reference Heimpel and Collier1996). This notion is supported by the fact that approximately four times more pupae were killed without being parasitized, most likely caused by host feeding, when hosts were abundantly available. At the same time, higher levels of alimentation, facilitated by the higher host abundance, may have contributed to the increased egg numbers in the 10P treatment.
Interestingly, a substantial proportion of parasitoid eggs was found to be dead, especially early in the oviposition period (i.e. when many eggs were laid) and in the 2P treatment. Although there are alternative explanations (e.g. host immune defence or egg resorbtion, with the latter resulting in the deposition of the resorbed (=dead) eggs: Gerling & Legner, Reference Gerling and Legner1968). However, egg resorbtion is expected to decrease rather than increase with lower host abundance (Quezada et al., Reference Quezada, Debach and Rosen1973; Rosenheim et al., Reference Rosenheim, Heimpel and Mangel2000), although the results of Richard & Casas (Reference Richard and Casas2009) indicate that egg resorbtion most likely occurs at intermediate host densities. Consequently, not only overall energy gain but also timing seems to play an important role. Thus, in our study, it seems most likely that the eggs were killed by the females themselves (ovicide of own eggs: Godfray, Reference Godfray1994; Yamada & Kitashiro, Reference Yamada and Kitashiro2002; Collier et al., Reference Collier, Hunter and Kelly2007). Egg death might occur accidentally while superparasitizing, or could be a deliberate ovicide by the female, following from her inability to recognize her own eggs. As female S. cameroni strongly concentrate parasitation on the dorsum of the abdomen of the host pupae and as the host used here is relatively small (Gerling & Legner, Reference Gerling and Legner1968; Tormos et al., Reference Tormos, Beitia, Böckmann and Asis2009), killing of eggs through both above mechanisms is certainly possible.
As expected, the occurrence of more than one egg per pupa increased with decreasing host availability (Sirot et al., Reference Sirot, Ploye and Bernstein1997; Hubbart et al., Reference Hubbard, Harvey and Fletcher1999; He et al., Reference He, Teulonz and Wang2006; Keasar et al., Reference Keasar, Segoli, Barak, Steinberg, Giron, Strand, Bouskila and Harari2006), though it occurred even when hosts were abundant. Note that up to six eggs (deposited by a single female within 24 h) were found in one puparium, while at the same time other hosts remained completely untouched. This suggests that S. cameroni either does not distinguish between parasitized and unparasitized hosts (Hubbard et al., Reference Hubbard, Marris, Reynolds and Rowe1987; Roitberg & Mangel, Reference Roitberg and Mangel1988; Visser, Reference Visser1993; but not Wylie, Reference Wylie1972) or that, at least occasionally, multiple-egg clutches are laid by individual females (Rosenheim & Hongkham, Reference Rosenheim and Hongkham1996; Mackauer & Chau, Reference Mackauer and Chau2001). Unlike an inability to discriminate between parasitized and unparasitized hosts resulting in superparasitism, multiple-egg clutches represent a response of females to environmental conditions, such as host scarcity (Mackauer & Chau, Reference Mackauer and Chau2001). In our case, the lack of a decline in superparasitism with female age, despite a concomitant reduction in egg load thus reducing female time constraints, may favour an inability to recognize parasitized hosts as explanation. The failure to use some of the host pupae provided would then result from host feeding and from variation in host quality, with females repeatedly favouring a given host pupa over others. Note, though, that the ability to discriminate may not necessarily result in an avoidance of superparasitism, especially since self-superparasitism and producing multiple-egg clutches, respectively, may yield fitness gains (Montoya et al., Reference Montoya, Liedo, Benrey, Barrera, Cancino and Aluja2000, Reference Montoya, Benrey, Barrera, Zenil, Ruiz and Liedo2003; Mackauer & Chau, Reference Mackauer and Chau2001; Keasar et al., Reference Keasar, Segoli, Barak, Steinberg, Giron, Strand, Bouskila and Harari2006; Burton-Chellew et al., Reference Burton-Chellew, Koevoets, Grillenberger, Sykes, Underwood, Bijlsma, Gadau, van de Zande, Beukeboom, West and Shuker2008; Silva-Torres et al., Reference Silva-Torres, Ramos Filho, Torres and Barros2009).
Following up on the question whether laying more than one egg per pupa might be adaptive in S. cameroni, it is interesting to note that the ratio between eggs laid and offspring produced was higher in the 10P compared to the 2P treatment. Consequently, the higher offspring number in females with hosts abundantly available is a function of both their higher egg number and an increased survival probability of offspring during development from egg to adult. This is in line with other studies reporting detrimental effects of superparasitism for individual offspring (van Alphen & Visser, Reference van Alphen and Visser1990; Sousa & Spence, Reference Sousa and Spence2000; Ahmad et al., Reference Ahmad, Ahmad, Mishra and Sheel2002; but not e.g. Mackauer & Chau, Reference Mackauer and Chau2001). Offspring production per parasitized pupa, in contrast, was higher when hosts were scarce and levels of self-superparasitism concomitantly high. This finding indicates that self-superparasitism may be beneficial here by increasing offspring numbers per pupa (cf. Mackauer & Chau, Reference Mackauer and Chau2001; Gu et al., Reference Gu, Wang and Dorn2003; Silva-Torres & Matthews, Reference Silva-Torres and Matthews2003; Keasar et al., Reference Keasar, Segoli, Barak, Steinberg, Giron, Strand, Bouskila and Harari2006). As in our experiment, there was no interference with conspecific females or other parasitoids; the positive effect documented here may result from an enhanced chance to overcome the host's immune defence (van Alphen & Visser, Reference van Alphen and Visser1990; Godfray, Reference Godfray1994; Rosenheim & Hongkham, Reference Rosenheim and Hongkham1996; Mackauer & Chau, Reference Mackauer and Chau2001), e.g. by injecting venoms (Rivers, Reference Rivers2004). The relatively low overall survival rates of S. cameroni on the given host (<50%) indicate that the host's immune response is a limiting factor for offspring development in S. cameroni. Consequently, higher egg numbers per host, as found under limited host availability, will evidently result in a higher number of offspring per parasitized pupa.
In summary, our study highlights the importance of host availability on parasitoid egg numbers, offspring production, eggs per pupa, number of hosts attacked and ovicide. The attack and mortality rates suggest that S. cameroni should be further investigated as a biological control agent for C. capitata. The patterns of self-superparasitism and putative ovicide by S. cameroni suggest a lack of discrimination between parasitized and unparasitized hosts combined with an inability of females to recognize their own offspring, or alternatively the deposition of multiple egg clutches as an adaptive strategy. However, we found only weak evidence for negative effects of self-superparasitism. Perhaps depositing more than one egg per pupa is even beneficial when hosts are scarce and survival rates unpredictable, thus forcing females into a trade-off between searching time and oviposition. Future experiments should test whether and how S. cameroni females respond to hosts previously used by themselves or by conspecific females, and whether superparasitism results from depositing single eggs repeatedly on the same hosts or from multiple-egg clutches.
