Dr Mark Jervis, Cardiff School of Biosciences, Cardiff University, PO Box 915, Cardiff CF10 3TL, UK. Fax: (0)29 20874305. E-mail:Jervis@cf.ac.uk
1 Ecologists concerned with life-history strategies of parasitoid wasps have recently focused on interspecific variation in the fraction of the maximum potential lifetime egg complement that is mature when the female emerges into the environment. Species that have all of this complement mature upon emergence are termed ‘pro-ovigenic’, while those that do not are termed ‘synovigenic’. We document and quantify the diversity of egg maturation patterns among 638 species of parasitoid wasps from 28 families.
2 We test a series of hypotheses concerning variation in ‘ovigeny’ and likely life-history correlates by devising a quantitative index – the proportion of the maximum potential lifetime complement that is mature upon female emergence.
3 Synovigeny, which we define as emerging with at least some immature eggs, was found to be by far the predominant egg maturation pattern (98·12% of species). Even allowing for some taxonomic bias in our sample of species, pro-ovigeny is rare among parasitoid wasps.
4 There is strong evidence for a predicted continuum in ovigeny index among parasitoid wasps, from pro-ovigenic (ovigeny index = 1) to extremely synovigenic species (ovigeny index = 0).
5 As predicted, synovigenic species are longer-lived than pro-ovigenic ones, and ovigeny index and life span are negatively correlated across parasitoid taxa, suggesting a life span cost of concentrating reproductive effort early in adult life.
6 There is equivocal evidence that host feeding (i.e. consumption of host haemolymph and/or tissues by adult wasps) is confined to synovigenic parasitoid wasps. It is also not certain from our analyses whether host feeding is associated with a relatively low ovigeny index.
7 As predicted, egg resorption capability is concentrated among producers of yolk-rich eggs. Also, the hypothesis that it is associated with a tendency towards a low ovigeny index is supported. Parasitoid species that produce yolk-rich eggs also exhibit a lower ovigeny index than species that produce yolk-deficient eggs.
8 Ovigeny index appears to be linked to parasitoid development mode (koinobiosis–idiobiosis).
9 We conclude that ‘ovigeny’ is a concept applicable to insects generally.
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.
Parasitoids are insects whose females lay their eggs in or on other invertebrates (mostly other insects) and whose larvae feed on the host and eventually kill it. The number of eggs that a female parasitoid deposits during her lifetime is determined by interactions among three processes: the number of suitable hosts that are encountered, the number of eggs that are matured over the female’s life span, and behavioural manipulation of the oviposition rate. Environmental stochasticity is expected to prevent exact matching of egg and host availability (Rosenheim 1996), but there is some evidence that natural selection adjusts egg production characteristics to approach the expected rate of host encounter (Price 1975; Rosenheim 1996; Sevenster, Ellers & Driessen 1998). Behavioural plasticity also allows parasitoid females to improve the match between host and egg availability (Charnov & Skinner 1984; Mangel & Heimpel 1998).
Parasitoid egg production traits that are potentially important in improving congruence between egg and host availability include: the total number of stem line oocytes (known as oogonia; their number sets the maximum potential fecundity), the maximum number of mature (i.e. fully chorionated) eggs that can be stored within the ovaries at a given time, the rate of egg maturation, the size of mature eggs and the capacity for, and rate of, egg resorption (Price 1975; Jervis & Kidd 1986). An important outcome of interactions among these characteristics is that, of the total number of oocytes a female carries, the fraction of the maximum potential lifetime egg complement that is mature when she emerges into the environment varies greatly among parasitoid taxa. Flanders (1950) classified parasitoid wasps that have all or nearly all of their eggs mature prior to the start of oviposition as ‘pro-ovigenic’ and those that continue to mature eggs throughout their reproductive life as ‘synovigenic’. This distinction continues to be widely applied in the literature on parasitoids, and recently several models have contrasted the effects of pro-ovigeny and synovigeny both on the evolution of parasitoid life-history strategies (Chan & Godfray 1993; Hunter & Godfray 1995; West, Flanagan & Godfray 1999; Ellers, Sevenster & Driessen 2000; Rosenheim 2000; Rosenheim, Heimpel & Mangel 2000) and on host–parasitoid population dynamics (Kidd & Jervis 1989; Shea et al. 1996).
While Flanders’ concept has been useful, it also glosses over much of the diversity found in parasitoid egg maturation strategies, particularly among synovigenic species. Also, it was formulated before the development of modern life-history theory, particularly that predicting a trade-off between reproduction and life span (Roff 1992; Stearns 1992). Here, we review and explore in more detail the diversity in egg maturation patterns among the parasitoid Hymenoptera, and place it in the context of life history theory in general and of that pertaining specifically to the reproductive strategies of parasitoids. Our main goal is to test a series of hypotheses that relate ovigeny to various life-history traits in parasitoids. We briefly discuss each hypothesis in turn.
1 A continuum of ovigeny is likely to exist, in which species can range from pro-ovigenic (all eggs are mature at emergence), through weakly synovigenic (most eggs are mature at emergence) to extremely synovigenic (no eggs are mature at emergence) (Jervis & Copland 1996; Quicke 1997) (note that our definition of pro-ovigeny is more conservative than that of Flanders). This prediction is based on published data on egg maturation patterns which tentatively indicate that there may be such a continuum.
2 Across parasitoid wasp taxa, life span should be negatively correlated with the proportion of oocytes mature upon emergence, and so pro-ovigenic wasps should be shorter-lived than synovigenic ones (Flanders 1950). Life-history theory predicts that a concentration of ‘reproductive effort’ towards the start of adult life should incur a life-span cost. Recently, Ellers & van Alphen (1997) provided evidence that an inverse relationship exists between life span and the proportion of oocytes mature at emergence among geographical strains of the parasitoid wasp Asobara tabida (Nees).
3 Host feeding (i.e. the consumption of host haemolyph and/or tissues by adult parasitoids) should be confined to synovigenic species, as suggested by Flanders (1950) (see also hypothesis 4).
4 Host feeding should be associated with a relatively low proportion of oocytes mature upon female emergence. The reasoning behind this is as follows: whereas female pro-ovigenic and weakly synovigenic wasps presumably require nutrients more for maintenance purposes, strongly synovigenic wasps need to fuel both maintenance and reproduction. Trehalose and proline, both of which can occur in significant amounts in host blood, can potentially be used for somatic maintenance in the broad sense (Gilbert & Jervis 1998). Host tissues and blood are a rich source of egg production nutrients: proteinaceous materials, fats and sugars and also essential vitamins and salts that are either scarce or absent in other foods (Jervis & Kidd 1986; Heimpel & Collier 1996; Rivero & Casas 1999a). At least some synovigenic, host-feeding species can allocate to reproduction, over a long time period (~2 weeks in Dinarmus basalis Rondani), nutrients originally acquired from a single meal taken early on in life (Rivero & Casas 1999a).
5 The production of yolk-deficient (hydropic) eggs should be associated with a higher proportion of eggs mature upon female emergence than the production of yolk-rich (anhydropic) eggs (see hypotheses 6, 7 and 8).
6 Egg resorption capability should be concentrated among anhydropic egg-producers, as hydropic eggs contain little or no yolk for the female to salvage.
7 Egg resorption capability should be associated with a relatively low proportion of oocytes mature upon female emergence. Resorption usually precludes oviposition and can be a time-consuming process (Jervis & Kidd 1986), so it is likely to be practised by longer-lived species, long life span being a trait predicted to be associated with a low proportion of oocytes mature upon emergence (see hypothesis 2).
8 Egg maturation strategy should be linked to the mode of parasitism. Two alternative, supposedly divergent (Mayhew & Blackburn 1999) life-history strategies are widely recognized as occurring among parasitoids with respect to development mode: ‘koinobiosis’ in which the host continues to develop, grow (if attacked as a larva), reproduce (if attacked as an adult), and remain active after attack, and ‘idiobiosis’ in which the host does not continue to develop, grow or remain active after attack (Askew & Shaw 1986; Shaw & Huddleston 1991). Several authors (Askew 1975; Force 1975; Price 1975; Blackburn 1991a, 1991b) have proposed that divergent suites of traits are associated with koinobiosis and idiobiosis, and that natural selection operates on the life history strategies of the two sorts of parasitoids to magnify the initial differences. This view was summarized by Godfray (1994) as the ‘dichotomous hypothesis’. Space does not permit a detailed discussion of all the life-history parameters and the various selection pressures acting upon them (see the aforementioned references); we confine ourselves here to discussion of the parameters we consider to be most relevant to ovigeny. Mayhew & Blackburn (1999), using phylogeny-based statistical techniques applied to a literature database, established that koinobioints, compared with idiobionts, have a shorter adult life span and produce smaller (= yolk-poor) eggs, and have a higher realized fecundity and a higher maximum rate of oviposition.
We hypothesize that ovigeny index may be linked to development mode for the following reasons: (i) a higher realized fecundity can be achieved by producing smaller (hydropic) eggs – a recognized trade-off (Godfray 1994); (ii) female adult koinobionts are shorter-lived than those of idiobionts (Mayhew & Blackburn 1999), and correlated with this should be a higher proportion of oocytes mature upon emergence (hypothesis 2); (iii) koinobiosis is supposed to select for a higher oviposition rate (including a minimal, if any, preovi-position period) which can be achieved by a tendency towards pro-ovigeny (Blackburn 1991b); (iv) koinobionts spend longer as pupae compared with idiobionts (Blackburn 1991a), and this ought to allow more time for oocytes to develop prior to female emergence.
Materials and methods
We used an ‘ovigeny index’, calculated as the fraction of the maximum potential lifetime egg complement that is mature upon emergence into the environment. In pro-ovigenic species the index is 1, whereas in synovigenic species it is less than 1. We use the term ‘emergence’ broadly, encompassing both emergence into the environment from an enclosure such as a host covering structure (i.e. cocoon, puparium, scale covering, leaf mine or similar ‘cell’) and eclosion, directly into the environment, of the parasitoid female from its pupal integument.
We classified parasitoid wasps as either pro-ovigenic or synovigenic and, where possible, calculated the ovigeny index using information mainly obtained from the literature but supplemented by unpublished research. Such information on the biology of parasitoid egg maturation varied in its usefulness both for assessing the form of ovigeny and for deriving ovigeny indices. We list below the forms in which information was available. Data, whether published or unpublished, were used only if the contents of the female’s reproductive tract had been examined directly in assessing the degree of egg maturity.
Females of a known age
Some sources give information on the state of the ovaries in newly emerged females. This information indicates whether the ovaries contain either only mature or at least some immature eggs – and therefore enable us to determine whether the species is pro-ovigenic or synovigenic, and in many cases gives the relative numbers of immature and mature oocytes in the ovaries. Alternatively, some sources provide information on the state of the ovaries in females of a known age or series of known ages starting beyond emergence. Finally, some sources provide numerical and in a few cases figurative information (line drawings/photographs) on the numbers and/or proportions of immature and mature oocytes in the ovaries of females of specified ages over the life span. Information of this type tends to give the most complete picture of the temporal dynamics of egg development (see Edwards 1954; Djambong & Laugé 1977; van Lenteren et al. 1987; Volkoff & Daumal 1994).
Females of an unspecified age
Some sources provide information on the state of the ovaries of females of an unspecified age. For instance Iwata (1955, 1959, 1960, 1962, 1966) provides information of this kind on a considerable number of species, particularly of Ichneumonoidea. Iwata counted and tabulated the numbers of mature and immature oocytes in one or more dissected, field-collected females of a species (he ignored the smallest developing oocytes, see Iwata 1955). A potential problem with such data sources arises when females are recorded as having only mature eggs in their ovaries. In the absence of other data it is not clear whether such females are synovigenic and close to the end of life (i.e. no more oocytes remain to be matured) or pro-ovigenic and newly emerged. However, this problem arose very infrequently, and where it did the data could usually be ignored owing to the availability of unambiguous data on other females of the same species.
Synovigeny is also indicated if: (a) females that have been exposed to hosts have higher egg loads than females of equivalent age that have been deprived of hosts (Drost & Cardé 1992); (b) egg load varies with type of diet, provided diet-based biases in oviposition rate are accounted for; and (c) the egg load of newly emerged females is significantly less than the measured realized fecundity.
Where available, data on the following variables were included in the analyses: life span, maximum potential lifetime fecundity (see below), egg resorption capability and feeding habit. We used life-span data for females given hosts as well as food wherever possible, and in all cases longevity had been measured in the laboratory. In the very few cases where a significant period of postreproductive life was reported for a species (e.g. see Jervis, Kidd & Almey 1994) female life span was taken to be the period of reproductive life plus any preoviposition period; the post-reproductive period was disregarded because it tends to be a laboratory artefact in parasitoids. Where more than one mean value was available for longevity (longevity varied, for example, with dietary regime), the highest mean was used as the measure of life span.
Data on maximum potential lifetime egg complement were used in calculating the ovigeny index among those synovigenic parasitoid wasps which emerge with > 0 mature eggs. These data referred either to the maximum average potential fecundity or to an approximation of this – the maximum average realized fecundity, depending on which was available. For six species (each marked with an asterisk in Table 2) a highly imprecise ovigeny index was obtained; these species were not included in the comparative analyses involving ovigeny index as a variable. Where more than one index could be calculated for a species (e.g. two indices were available for each of the two geographical strains of Asobara tabida), the mean of the indices was used in our analyses.
Table 2. Ovigeny index, calculated for 67 parasitoid wasp species. Superscripts: a some females emerge with a small proportion of immature eggs; b index varies with female body size; c varies geographically; d varies with photoperiod in wasp cultures; e varies with host instar attacked, probably via effects on female body size. In A. hesperidum the phenotypic variation is most likely to be correlated with variation in resource acquisition during the larval stage, since pro-ovigeny-tending species have minimal need to allocate resources to storage. However, phenotypic variation in moderately to strongly synovigenic species will be also be linked to differential resource allocation, during the pupal stage, to reproduction and storage; this is suggested by data for the Asobara tabida SIN and KOS populations (discussed under Anticipated host availability). Species marked with an asterisk were not included in the comparative analyses involving ovigeny index as a variable
The taxonomic distribution of ovigeny indices suggests that ovigeny index is, to some extent, phylogenetically constrained. For example, all the Mymaridae in Table 2 have an index of 1, while all the Bethylidae have an index of 0. In order to avoid taxonomic pseudoreplication in our analyses, we used statistical techniques that take account of the potential constraints of phylogeny upon variation in trait values (Grafen 1989; Harvey & Pagel 1991; Purvis & Rambaut 1995). When investigating the relationship between continuous variables (life span, ovigeny index), we used Harvey & Pagel’s (1991) method of independent comparisons. The information on phylogenetic relationships necessary to make such comparisons was compiled from Ronquist et al. (1999) (Hymenoptera), Ronquist et al. (1999), Brothers (1999) and Carpenter (1999) (Aculeata), Ronquist (1999) (Cynipoidea), J.S. Noyes in Heraty, Woolley & Darling (1997) (an intuitive tree for Chalcidoidea), J. Heraty (personal communication) (an intuitive tree for Aphelinidae), Gauthier et al. (2000) (Eulophidae), Quicke & Belshaw (Braconidae, unpublished tree, based on 28S D2 rDNA sequence data) and Belshaw et al. (1998) (Ichneumonidae). Where relationships could not be resolved, means for the unresolved taxa were used. Continuous variables were transformed to logarithms, where necessary, to achieve normality, except in the case of ovigeny index, where angular transformation was used. When analysing combinations of continuous and dichotomous variables, we used Burt’s (1989) method, followed by paired t-tests on nodal means (Purvis & Rambaut 1995). One-tailed tests were applied when the a priori hypothesis made them appropriate. Because of the controversy currently surrounding phylogeny-based techniques in comparative studies (Ricklefs & Starck 1996; Björklund 1997; Price 1997), we supplemented these with traditional cross-species analyses that do not take phylogeny into account. Below, we distinguish between results obtained by phylogeny-based statistical techniques (‘PT’) and results obtained by techniques that are not ‘phylogenetically aware’ (‘TT’). A copy of the main database is available from M.A. Jervis.
Data were available for a satisfactory number of species in the testing of some individual hypotheses. For example, in testing the first part of hypothesis 2, i.e. synovigenic parasitoids are longer-lived than pro-ovigenic ones, data from 73 species were available. By contrast, there was limited scope to control for the effects of some variables when testing other hypotheses. However, this did mean that there was little overlap between data used to test different hypotheses, so precluding the need to apply multiple tests corrections.
Based on the various types of information used, we were able to identify the egg maturation type (pro-ovigenic or synovigenic) shown by 638 species belonging to 28 families within the ‘Parasitica’ and the Aculeata of the order Hymenoptera (Table 1). Of these, the vast majority (98·12%, n = 626) are synovigenic, while only 12 species (1·88%) are unambiguously pro-ovigenic. We encountered little difficulty in categorizing the majority of species we obtained information on, and where species had been previously categorized, there was little disagreement.
Table 1. Parasitoid wasp families represented in our sample, their size, and the numbers of pro- and synovigenic species we categorized within each
Number of species in our sample
Number of ‘strictly’ pro-ovigenic species
Number of synovigenic species
We examined the numbers of pro-ovigenic and synovigenic species in each family of parasitoid wasps, in relation to the size of that family (Table 1). Data on family sizes were taken from Gaston (1991), Gauld & Bolton (1988), Johnson (1992), Noyes (1990 and personal communication) and Vlug (1993). Six families from three superfamilies contained pro-ovigenic species (Platygastridae, Eucoilidae, Eucharitidae, Mymaridae, Pteromalidae and Aphelinidae, Table 1). We took the proportions of pro-ovigenic and synovigenic species in our sample of each family to be representative of that family. The grand mean proportion, weighted by the estimated number of described species in each family, indicates that we ought to have found 27 pro-ovigenic and 611 synovigenic species (4·23% and 95·77%, respectively), a significant difference compared with our sample (χ2 = 8·33, P < 0·005). Thus, even accounting for taxonomic (family) bias in our sample, pro-ovigeny is rare among parasitoid wasps. Of course, the assumption that the proportions of the different ovigeny types per family sample are representative is questionable (although accounts of oviposition behaviour suggest it is probably correct for the Eucharitidae, see Clausen 1940, 1941), and several of our family samples (e.g. the Ichneumonidae) are certainly strongly biased towards particular subfamilies. Only a more extensive survey can provide a conclusive answer as to the real representation of the different ovigeny types among parasitoid wasps. Quantitative information on ovigeny was available for only a small proportion (≈ 10%) of the 638 species that we established to be pro-ovigenic or synovigenic. Table 2 presents ovigeny indices for these 67 species. The data strongly support previous suggestions (Jervis & Copland 1996; Quicke 1997) that there is a continuum of ovigeny from an index of 0 (extreme synovigeny) to an index of 1 (pro-ovigeny) (hypothesis 1) Fig. 1). Our subsample of synovigenic species clearly had a non-normal distribution of ovigeny indices (Fig. 1; Shapiro–Wilkes’W-test for normality: P < 0·001), with 35% of species emerging with no mature eggs (index = 0), a median value of 0·05 and positive skewness (g1 = 0·85). The mean index for synovigenic species was 0·22 with a variance of 0·07. The maximum index value for a synovigenic species was 0·75. The gap between 0·75 and 1 in Fig. 1 may be attributed to undersampling and/or evolutionary instability of ovigeny values in that range, but note that there are three species (not plotted) with an index of ‘~1’.
Thirty-four independent comparisons between life span and ovigeny index were available for testing hypothesis 2. These revealed a significant negative relationship between life span and the proportion of oocytes mature upon adult emergence (Fig. 2) (PT: line forced through the origin, slope = − 0·292, t33 = −4·51, P < 0·0001; TT: slope = −0·375, F1,39, P < 0·0001). Ovigeny index accounted for 38% of the variation in life span, correcting for phylogeny. Hierarchical anova revealed that the largest percentage of the variation in life span occurred at the level of the species within the genus (46·4%) and in ovigeny index at the level of the family within the superfamily (51·4%). The smallest percentage of the variation in both life span and ovigeny index was at the level of the genus within the family (9·4% and 4·7% respectively).
We examined whether any life-history variable(s) on which we had gathered data accounted for the residuals in the relationship between life span and ovigeny. We could use only traditional statistical techniques, as statistical power was too low using a phylogeny-based analysis. When the effect of ovigeny index on life span was removed, residual life span was not significantly related to either female feeding habit, egg type, egg resorption capability, larval development mode or female fecundity.
Intraspecific variation was not included in the above analysis, but geographical populations (Asobara tabida SIN and KOS strains) and subsamples of a population (Trichogramma nubilale Ertle & Davis large and small wasps) that had the largest proportion of oocytes mature upon emergence also had the shortest life span (Ellers & van Alphen 1997; Olson & Andow 1998).
Comparing the life span of pro-ovigenic vs. synovigenic parasitoids (i.e. testing Flanders’ 1950 observation that the latter are longer-lived, hypothesis 2), we found that the life span of synovigenic species (mean ± SE: 25·8 ± 2·2 days, n = 63) is indeed significantly greater than that of pro-ovigenic species (9·0 ± 1·5 days, n = 10) (PT: paired t6 = 1·89, P = 0·027; TT: t71 = 6·25, P < 0·0001).
Although host feeding has been unequivocally recorded in many synovigenic species (Jervis & Kidd 1986) it is unclear whether it is actually confined to synovigenic wasps (hypothesis 3). The pro-ovigenic aphelinid Eretmocerus sp. displays behaviour suggestive of host feeding: females use the ovipositor to sting and then probe the whitefly host close to the latter’s anal region, and they then apply their mouthparts to the wound. However, it needs to be established whether they consume mainly the host’s blood or mainly the honeydew contained in the host’s hind gut (if they take mainly honeydew, they are very likely to do so from the gut rather than from elsewhere, because the hosts – atypically for Sternorrhynha – jettison their honeydew well away from the patch; M.J.W. Copland, personal communication).
Host feeding species tend on average to have a lower ovigeny index (hypothesis 4) (host feeders: 0·088 ± 0·066 [n = 15] vs. others: 0·479 ± 0·079 [n = 27]) and this difference is significant according to the traditional analysis (TT: t40 = −3·81, P = 0·0002) but not the phylogeny-based analysis (PT: paired t3 = −1·00, P > 0·10). Note that the data set included Eretmocerus sp. as a host feeder. Because there is some doubt as to whether this species mainly consumes host materials or honeydew, we tested hypothesis 4 again, this time coding Eretmocerus sp. as a parasitoid wasp which does not consume host tissues or blood. Again, the phylogeny-based test failed to reveal an association (PT: paired-t, t4 = 0·47, P > 0·2) whereas the traditional test did (TT: one-tailed, t40 = 3·72, P = 0·0003). However, it should be noted that in both of the above phylogeny-based analyses the number of transitions in the tree was low (i.e. statistical power was low). The results of similar analyses, based on a taxonomically more extensive data set, could well stand in favour of hypothesis 4.
The data unequivocally support hypothesis 5, which argues that producers of hydropic eggs have on average a significantly higher ovigeny index than producers of anhydropic eggs (0·585 ± 0·070 [n = 30] vs. 0·108 ± 0·062 [n = 22]) (PT: paired t4 = 2·52, P = 0·043; TT: t49 = 4·91, P < 0·00001). (Note that whereas, according to the dichotomous hypothesis, hydropic egg production is considered to be a means of maximizing fecundity [see hypothesis 8], we found no statistically significant link between ovigeny index and fecundity.)
As predicted (hypothesis 6), the ability to resorb eggs was recorded mainly (indeed, only) from anhydropic egg-producers, which themselves appear to be confined to synovigenic species (note that anhydropic egg-producers include species that do not host feed). Also as predicted (hypothesis 7), egg resorbers have a significantly lower ovigeny index than non-resorbers (0·022 ± 0·009 [n = 12] vs. 0·575 ± 0·069 [n = 28]) (PT: paired t4 = 2·54, P = 0·032; TT: t38 = 7·92, P < 0·00001).
Koinobionts tend to have on average a higher ovigeny index than idiobionts (hypothesis 8) (0·473 ± 0·081 [n = 25] vs. 0·313 (0·076 [n = 31]), and this difference is significant according to the phylogeny-based analysis but not according to the traditional analysis (PT: one-tailed paired t6 = 2·56, P = 0·011; TT: one-tailed t53 = 1·48, P = 0·072). As Godfray (1994) notes, the dichotomous hypothesis pertains mainly to parasitoids attacking host larval stages. Although Mayhew & Blackburn (1999) found a strong degree of support for the hypothesis in an analysis that encompassed parasitism of a broader array of host stages, we chose to test hypothesis (8) further, using the subset of our database containing only parasitoids of larval stage hosts. Again, koinobionts tend on average to have a higher ovigeny index than idiobionts (0·477 ± 0·093 [n = 19] vs. 0·008 ± 0·005 [n = 11]) but in this case the difference is statistically very highly significant according to the traditional analysis and falls just short of being significant according to the phylogeny-based analysis (PT: one-tailed paired t2 = 2·83, P = 0·0527; TT: one-tailed t28 = 5·08, P = 0·000065). Taking the aforementioned results as a whole, it is reasonable to conclude that hypothesis 8 is upheld, i.e. ovigeny is linked to development mode. Support for this stance comes from the results of ancillary tests we conducted specifically to establish whether the dichotomous hypothesis holds when tested against our data set. Using our data for parasitoids of larval stages, the trends for two major life-history variables covered by the dichotomous hypothesis (egg type, life span) are all in the predicted direction: towards hydropic egg production in koinobionts (one-tailed Fisher’s exact test, PT: n = 11, P = 0·175; TT: n = 54, P = 0·0006) and towards a shorter life span in koinobionts (PT: paired-t, t4 = 2·95, P = 0·021; TT: one-tailed t, t25 = 2·47, P = 0·022). (There were insufficient data for fecundity to be considered.) Ovigeny therefore appears to be the final, major piece in the jigsaw of life-history traits forming the dichotomous hypothesis. However, we lacked sufficient information to be able to control for the effects of major correlates (e.g. life span and egg type) of ovigeny index and development mode, and therefore see whether the observed relationship still holds. Given a better data set, future investigators might explore this using phylogenetic multiple regression techniques (see Mayhew & Blackburn 1999).
In summary, despite the potentially confounding effects of using a data set derived from different studies carried out under different conditions by different researchers, we found support for most of the hypotheses tested.
What enables some parasitoid wasp species to emerge with all their eggs mature, and what constrains other species from doing so? How are resources allocated in females, what external nutrient inputs are required during adult life, and how do these relate to ovigeny index and development mode? What environmental selective forces are acting on females to determine ovigeny index? How do our results relate to insects in general? We address these questions below.
Resources carried-over from the larva
In parasitoid wasps the size and nutritional quality of the host attacked, the period of time available for larval development and the intensity of larval competition all combine to determine the amount of carry-over of resources to the pupal and adult stages (Godfray 1994; Jervis & Copland 1996), and so ought, a priori, to be a major determinant of intraspecific variation in ovigeny index. Within parasitoid wasp species, adult body size is strongly correlated with larval feeding history (Mackauer, Sequiera & Otto 1997; Harvey et al. 1999), and the following variables are correlated strongly with body size: fat and glycogen reserves, ovariole number, egg size, egg load at emergence, lifetime fecundity and longevity in the absence of food (e.g. Opp & Luck 1986; Harvey et al. 1994; Visser 1994; Heimpel, Rosenheim & Mangel 1996; Jervis & Copland 1996; Ellers, van Alphen & Sevenster 1998; Olson & Andow 1998; Olson et al. 2000). There is also evidence that ovigeny index varies with body size (Table 2), although currently precise data exist for only one species, Trichogramma nubilale (Olson & Andow 1998). In T. nubilale large wasps have a higher realized lifetime fecundity, have a higher egg load upon emergence and are longer-lived than small wasps, suggesting a higher degree of resource carry-over. However, ovigeny index in this species decreases with increasing body size, the reason being that the proportionate increase in starting egg load with body size is smaller than the proportionate increase in fecundity. This is likely to be a general phenomenon, reflecting a higher degree of allocation, in larger wasps, both to the fat body (to fuel a longer potential life span) and to lifetime fecundity (to provide greater numbers of ovarioles and oocytes and to provide an increase in egg size).
Interspecifically, body size correlates poorly with life span and fecundity in parasitoid wasps (Blackburn 1991a). Whether it does so with ovigeny index remains to be determined; currently, we lack sufficient data on body size to investigate this.
Considering only functions obviously related to reproduction and survival, the main sites of carried-over resource allocation in parasitoid wasps are the ovaries (as lipids and proteins in the yolk of the oocytes, in the later stages of their development) and the fat body (as lipid, protein and glycogen) (King & Richards 1969; Al-Darkazly 1977; Ellers 1996; Ellers & van Alphen 1997; Rivero & Casas 1999b; Olson et al. 2000). In pro-ovigenic parasitoids carried-over resources are allocated mainly, if not entirely, to egg maturation. Because of their short potential life span females should allocate minimal, if any, resources to the fat body. We reason that pro-ovigeny will result if two conditions are satisfied: (i) the amount of carry-over from the larval stage has to be sufficient to offset not only the metabolic costs of egg development during the pupal stage but also the costs of pupal development; (ii) pupal duration time has to be sufficiently long to allow all oocytes to be matured prior to adult emergence.
It is not known whether synovigenic species differ from pro-ovigenic species in the amount of resource carry-over (but see below). However, even with an equivalent amount of carry-over, synovigeny ought to result because a fraction of the carried-over resources is allocated to storage in the fat body (although it is possible that a short pupal duration period may also contribute to the synovigeny of some species). Given the across-species negative correlation between life span and ovigeny index, the degree of resource allocation to fat body storage is, we hypothesize, likely to be higher the stronger the degree of synovigeny, i.e. the lower the ovigeny index.
Strong synovigeny could be interpreted as reflecting a low amount of resource carry-over compared with weak synovigeny and pro-ovigeny, and the existence of an ovigeny continuum may be partly attributable to a link between carry-over and the dichotomy in parasitoid wasps with respect to development mode. Many koinobionts can manipulate the host’s physiology and feeding behaviour to their own benefit (e.g. Godfray 1994; Harvey et al. 1999), so they might be better than idiobionts at manipulating the amount of carry-over to the pupal stage and be more likely to reach adulthood with a higher proportion of oocytes mature.
We would expect a preoviposition period to be confined to synovigenic wasps, given both their carried-over resource allocation pattern – to storage as well as egg maturation – and their longer potential life span compared with pro-ovigenic wasps. As far as we can ascertain, this is indeed the case. With extreme synovigeny (no mature eggs present, an ovigeny index of zero) there is typically a preoviposition period of several days, although this constitutes a small proportion of the life span −0·15 being the highest recorded, in the host feeding ichneumonid Priopodus nigricollis (Thomson) (Quednau & Guevremont 1975). However, even with moderate synovigeny there can be a preoviposition period. For example, Meteorus trachynotus (Viereck) females emerge with some mature, and presumably layable, eggs but nevertheless have a preoviposition period of at least 1 day (Thireau & Régnière 1995). The functional significance of the reproductive delay in this species is not known; however, given that M. trachynotus is solitary, the delay is unlikely to be attributable to the need to accumulate sufficient mature eggs to produce an optimum egg clutch size (but see Rosenheim & Honghkam 1996). The existence of a preoviposition period may be inadvertently overlooked in many parasitoid wasps: a substantial number of species do not eclose directly into the environment (Quicke 1997), and some fraction of the egg complement may be matured between eclosion and emergence. For instance, Nasonia vitripennis (Walker) females spend 20–30 h inside the host’s puparium prior to emergence during which period of time they mature some eggs (Edwards 1954).
Externally derived nutrient inputs
In parasitoid wasps, external nutrient inputs from adult feeding derive from various sources in the field: floral and extrafloral nectar, ‘homopteran’ (= Sternorrhynchan) honeydew, pollen, leaf trichome exudates, fungal spermatial fluid, living and decaying plant tissue and host blood and tissues (Gilbert & Jervis 1998; Jervis 1998). The nutrients from such foods can be put to various uses by female synovigenic parasitoid wasps: storage, ovigenesis, somatic maintenance and locomotion (e.g. see Clarke & Smith 1967; Al-Darkazly 1977; Olson 1980; Rivero & Casas 1999a,b), although in some species sugar meals apparently cannot be used to synthesize lipids (Asobara tabida: Ellers 1996; Macrocentrus grandii (Goidanich): Olson et al. 2000). The contribution of feeding to survival and fecundity is well documented for a diverse range of parasitoid species held under laboratory conditions. Several of these species are known to be synovigenic according to the criteria used here (see Jervis & Kidd 1986; van Lenteren et al. 1987; Heimpel & Collier 1996). Although hypothesis 4 was not unequivocally supported by the available data, we hold to the view that synovigenic species with a low ovigeny index should be more likely than other wasps to host feed. Feeding on pollen – a potential (but generally suboptimal) alternative to host materials –appears to be rare among parasitoid wasps (for a discussion of the disadvantages of pollen feeding, see Gilbert & Jervis 1998; Jervis 1998). Therefore, strongly synovigenic parasitoid wasps are most likely to fuel their egg production demands by host feeding. This is not to say, however, that all parasitoid wasps with a low ovigeny index will host feed: known exceptions are the three Bathyplectes species listed in Table 2.
Rivero & Casas (1999b) note that the current picture of internally and externally derived nutrient allocation in synovigenic parasitoids is sketchy, and they recommend a methodology for identifying, tracking and quantifying resources within adults (see also Olson et al. 2000). The same methodology could also be applied to nutrient carry-over from the larva, through the pupa, to the adult (see also Spradbery & Sands 1981; Boggs 1997a). The stage is therefore set for quantitatively relating ovigeny to resource capital, income and expenditure.
Anticipated host availability
The degree of habitat heterogeneity, in terms of the reproductive opportunities available to female wasps, is a major candidate selective force in the evolution of ovigeny index. Natural selection is expected to lead to reproductive strategies that approach, but do not necessarily attain, a quantitative match between egg supply and availability of suitable hosts. The main factor preventing precise matching is thought to be variation in host availability (Rosenheim 1996; Sevenster et al. 1998; van Baalen 2000), but variation in factors that affect egg maturation, such as temperature and adult food availability, may also contribute to the mismatch. It is in the face of this variation that reproductive strategies, including physiological and behavioural traits, are under selection to balance host and egg availability. What is the role of variation in ovigeny index in this process? While pro-ovigeny (or, more generally, a high ovigeny index) provides the benefit of maximizing the number of eggs available early in life, this must be balanced against potential losses in reproductive plasticity later in life. In other words, the closer to being pro-ovigenic parasitoids are, the less will they be able to adjust their reproductive output to variation in host encounter (Ellers et al. 2000; van Baalen 2000). This disadvantage is likely to increase with the magnitude of the variance in host encounter rates (Ellers et al. 2000). Since host populations often exhibit aggregated spatial distribution patterns (which can be taken to be equivalent to a high variance in host encounter rates), this reasoning might be considered sufficient to explain the rarity of strict pro-ovigeny. However, most of the pro-ovigenic species listed in Table 2 are associated with hosts that typically have an aggregated distribution, e.g. Trybliographa rapae (Jones 1986), and the Mymaridae (Waloff & Jervis 1987).
A related disadvantage of a tendency towards pro-ovigeny is reduced life span. To the extent, therefore, that a negative correlation exists between female survival and early reproductive effort (as is evident from this study), pro-ovigenic-tending species are sacrificing opportunities for future reproduction for the ability to take advantage of a high host encounter rate early on in life.
The inability to precisely match egg supply and host availability results in individual parasitoids being either egg-limited or time-limited (Driessen & Hemerik 1992; Getz & Mills 1996; Rosenheim 1996; Heimpel & Rosenheim 1998; Heimpel, Mangel & Rosenheim 1998; Sevenster et al. 1998; Heimpel 2000; van Baalen 2000). How might ovigeny index influence the risk of egg- or time-limitation in parasitoids? We have already argued that synovigeny is likely to provide a high degree of flexibility in the timing of reproduction. Synovigeny in general, including a low ovigeny index, may therefore result in a minimum of either time- or egg-limitation (Ellers et al. 2000). Synovigenic species may still experience significant levels of transient egg-limitation, however, depending upon the relationship between the egg maturation rate and the rate of encounter with hosts (Heimpel & Rosenheim 1998; Heimpel et al. 1998; Casas et al. 2000; Rosenheim, Heimpel & Mangel 2000). Females of extremely synovigenic species (an ovigeny index of zero) would be subject to severe egg-limitation around the start of adult life if their females forage before and during maturation of the initial egg complement. Furthermore, egg resorption – a ‘last resort’ survival tactic involving the realloaction of resources from eggs to somatic maintenance (Jervis & Kidd 1986) – can greatly enhance the risk of egg-limitation in synovigenic species (Rosenheim et al. 2000), and this is a form of egg-limitation not experienced by pro-ovigenic species. (Although we found that egg resorption capability is absent from hydropic egg producers [hypothesis 6], it should be borne in mind that such wasps tend to have greater numbers of ovarioles compared to anhydropic egg producers (Price 1975). Therefore, some of them ought, in principle, to be capable of compensating for the low yolk content per egg by resorbing a large number of eggs at one time. The fact that no such species have been recorded to date suggests that resorption, even of large numbers of yolk-poor eggs concurrently, is not a viable option.)
There is good evidence for phenotypic plasticity in resource allocation, post-emergence, in parasitoids: females can, under conditions of high host availability, shift their egg production ‘schedule’ towards an earlier age (synovigenic species: Aphelinus semiflavus Howard, Aphidius smithi Sharma & Subba Rao: Mackauer 1982, 1983; Dicondylus indianus Olmi: Sahragard, Jervis & Kidd 1991; Trichogramma minutum Riley: Bai & Smith 1993; Asobara tabida: Ellers & van Alphen 1997; pro-ovigenic species: Leptopilina boulardi Barbotin, Carton & Kelner-Pillaut: Kopelman & Chabora 1992). In the first four of the aforementioned synovigenic species there was no concomitant shortening of life span. It may be reasoned that this was because females had opportunities to feed (on either supplementary food, host honeydew or host blood – D. indianus was actually observed to feed). However, in honey-provisioned A. tabida KOS-population females, the concentration of egg production early in life (under conditions of high host availability) occurred – in accordance with life-history theory, at the cost of life span (Ellers & van Alphen 1997). A similar negative correlation occurred in the pro-ovigenic species L. boulardi. In the latter species, females continuously provided with hosts and given sugar-rich food were shorter-lived than females given limited access to hosts, but produced the same numbers of progeny, suggesting that life span reduction was the direct consequence of a high egg production rate, the cost of which could not be met by feeding.
Although to date there have been no quantitative genetic studies of the type done on non-parasitoid insects (e.g. Partridge & Farquhar 1981; Service 1989; Zwaan, Bijlsma & Hoekstra 1995), it is reasonable to suppose that natural selection can effect significant shifts in the egg production schedules of parasitoids and so account for both the ovigeny index continuum and the ovigeny index-life span negative correlation revealed here. Presently, there is only circumstantial evidence that the SIN and KOS strains of A. tabida (which form part of a north–south geographical cline in Europe, Kraaijeveld & Van der Wel 1994) are genetically different. However, it is reasonable to surmise that the different ovigeny indices of these strains (which have the same mean realized lifetime fecundity under favourable conditions), coupled with an inversely related life span and an inversely related allocation to fat body storage, are the result of natural selection, adapting the populations to local conditions, in particular the average availability of hosts (see Kraaijeveld & Van der Wel 1994; Ellers & van Alphen 1997).
Except for pro-ovigenic species and those extremely synovigenic species that emerge with no mature eggs, we calculated the ovigeny index using fecundity data, and wherever possible potential fecundity was used. However, under conditions of time-limitation, the potential fecundity is by definition greater than the realized (i.e. actual) fecundity. If realized fecundity, in the field, were to be used to estimate ovigeny instead of the potential fecundity, one could envisage cases of ‘functional pro-ovigeny’ in otherwise synovigenic species, if the number of eggs at emergence equals the realized fecundity by virtue of low host encounter rates and/or high parasitoid mortality rates (see Wharton 1993; also below). More generally, an ovigeny index based on realized fecundity will always be higher than one based on potential fecundity under conditions of time-limitation.
Anticipated host availability cannot, however, be invoked as the major selective force in the evolution and maintenance of ovigeny index in all parasitoids, eucharitid wasps being a case in point. Their pro-ovigeny is undoubtedly linked more to a high fecundity (up to 15 000 eggs in some species), required to offset a low probability of the phoretic planidium larvae attaching themselves to foraging worker ants (see Price 1975).
Extrinsic mortality acting upon the adult female
Mortality such as that inflicted by predators, weather and starvation is another candidate force in the evolution of ovigeny index. Predation certainly has the potential to substantially shorten the life span of adult parasitoids (Rosenheim 1998), as has food scarcity (Jervis & Kidd 1986; van Lenteren et al. 1987). Egg maturation rates in Aphytis species suggest that females are able to lay little more than a single egg load and are therefore ‘functionally pro-ovigenic’ under conditions of high predation or starvation risk (Heimpel, Mangel & Rosenheim 1998).
Heimpel, Rosenheim & Mangel (1997b) estimated the rate of predation by a number of generalist predators on adults of the synovigenic parasitoid Aphytis aonidiae. The expected life span, calculated from these estimated predation rates, was less than 1–2 days during some parts of the year, compared to a median of 20·5 days in the laboratory when supplied with sugar-rich food and hosts. Adult mortality rates of this magnitude can presumably lead to selection for shifts in the egg maturation schedule towards the start of adult life (i.e. a higher numerator in the ovigeny index), provided the mortality is concentrated among older females (Stearns & Hoekstra 2000) (but note that the opposite appears to be the case with Aphytis aonidiae). With regard to food scarcity as an extrinsic selective force, given the known decline with age in the size of the fat body in synovigenic species, older females will presumably be more prone than younger ones to starvation under conditions of food scarcity, and this will select for a shorter life span and thus a higher ovigeny index.
Ovigeny in other insects
Marked interspecific variation in ovigeny index is not unique to parasitoid wasps: it also occurs in parasitoid flies (Jervis & Copland 1996), Lepidoptera (Boggs 1986, 1997a, 1997b) and probably many other kinds of insects. Boggs (1986, 1997a) argued that for Lepidoptera, the higher the potential importance of adult feeding to reproduction (measured in terms of the proportion of oocytes that are mature upon emergence), the longer is the life span and thus the time that age-specific fecundity stays high. Based on her own empirical studies, she concluded that: ‘the relative availability of nutrients from juvenile and adult resource pools most likely drives this cross-species correlation between feeding habits and timing of reproduction’. This accords well with the concept of ovigeny in parasitoid wasps. Boggs (1997a) constructed a graphical model which relates the form of the age-specific fecundity curve to adult dietary requirements for egg production, and takes into account diet quality (which ranges from species feeding only on nectar to species feeding on both nectar and pollen) defined as the degree to which all required nutrient types are present in the food. Her model predicts that age-specific fecundity in species requiring no food during adult life will show a decline from the time of emergence, and a relatively short life span (Boggs’ curve A, equivalent to pro-ovigeny in parasitoid wasps). The model includes three other curves, all for synovigenic species. The curve for nectar-feeding Lepidoptera emerging with some immature eggs (Boggs’ curve B, equivalent to weak synovigeny in parasitoid wasps) shows no decrease or increase in age-specific fecundity early on in life, then subsequently declines. Lepidoptera emerging with no mature eggs fall into two groups: those that require only a carbohydrate diet (Boggs’ curve C) and those that require additional materials that are contained in pollen (Boggs’ curve D). Both types are predicted to show an increase in fecundity early in life, and life span in curve C is intermediate between curves B and D. In contrast to Lepidoptera, parasitoid wasps that feed both on sugar-rich food and on proteinaceous materials (host haemolymph) include species that emerge with some mature eggs (Jervis & Kidd 1986). Nevertheless, we make the following attempt to place the age-specific fecundity curves of parasitoid wasps in the context of Boggs’ model.
The curve for the pro-ovigenic wasp Leptopilina boulardi corresponds roughly to curve A; it is concave as opposed to convex; Kopelman & Chabora 1992). The curves for the synovigenic species Gryon pennsylvanicum (Vogt & Nechols 1993), Venturia canescens (Trudeau & Gordon 1989; Harvey et al. unpublished) and Asobara tabida (Ellers & van Alphen 1997) correspond roughly to Curve B. The curve for Bathyplectes anurus, a non-host-feeding, synovigenic species that emerges with no mature eggs, corresponds closely to curve C (see Yeargan & Latheef 1977). That for the non-host feeder Coccophagus atratus is also similar, overall, to Boggs’ curve C, if one ignores the peaks and troughs during the mid-phase of life (see Donaldson & Walter 1988). The curves for species that host-feed (taken to be equivalent to pollen feeding in Lepidoptera, i.e. Boggs’ curve D) show a range of forms, being either similar to that of B. anurus, i.e. Boggs’ curve C (e.g. Edovum puttleri: Corrigan & Lashomb 1990; Aphelinus albipodus: Bernal, Waggoner & González 1997), or showing an initial increase which is soon followed by a decline (Tamarixia radiata: Chien, Chu & Ku 1995; Cephalonomia stephanoderis: Infante et al. 1992), i.e. not corresponding closely to any of Boggs’ curves.
The relationships that were outlined by Boggs (1997a) were intended to be general in scope, and she used just three species of Lepidoptera to illustrate them (see also Boggs 1986, 1997b). Our analysis provides some support for the generality of her conclusions for insects as a whole, by demonstrating a degree of parallelism between the two major taxa of insects. We conclude that ‘ovigeny’ is a potentially unifying concept for aiding understanding of the evolution of life-history strategies among insects generally.
We thank the following who provided valuable advice and/or information: Jacques van Alphen, Hasan Basibiyuk, Robert Belshaw, Tim Blackburn, Jerome Casas, C.C. Chien, Tim Collier, Mike Copland, Jim Cronin, Jacintha Ellers, Mike Fitton, Alan Grafen, Ian Hardy, John Heraty, Hefin Jones, Arthur Kopelman, Jana Lee, Suzanne Lewis, Peter Mayhew, Nick Mills, John Noyes, Kazuro Ohno, Susan Opp, Gabriela Perez-Lachaud, Craig Phillips, Andrew Polascek, Donald Quicke, Peter Randerson, Jay Rosenheim, Cetin Sengonça, Mark Shaw, Christine Taylor, Sujaya Udayagiri, Felix Wäckers, Gimme Walter and Susan Weller. We dedicate this paper to the memory of Stanley Flanders, in recognition of his seminal contributions to parasitoid wasp biology.
Received 15 August 2000; revision received 9 January 2001