Life-history traits, pace of life and dispersal among and within five species of Trichogramma wasps: a comparative analysis

Major traits defining the life history of organisms are often not independent from each other, with most of their variation aligning along key axes such as the pace-of-life axis. We can define a pace-of-life axis structuring reproduction and development time as a continuum from less-fecund, longer-developing ‘slow’ types to more-fecund, shorter-developing ‘fast’ types. Such axes, along with their potential associations or syndromes with other traits such as dispersal, are however not universal; in particular, support for their presence may be taxon and taxonomic scale-dependent. Knowing about such life-history strategies may be especially important for understanding eco-evolutionary dynamics, as these trait syndromes may constrain trait variation or be correlated with other traits. To understand how life-history traits and effective dispersal covary, we measured these traits in controlled conditions for 28 lines from five species of Trichogramma, which are small endoparasitoid wasps frequently used as a biological model in experimental evolution but also in biocontrol against Lepidoptera pests. We found partial evidence of a pace-of-life axis at the interspecific level: species with higher fecundity also had faster development time. However, faster-developing species also were more likely to delay egg-laying, a trait that is usually interpreted as “slow”. There was no support for similar covariation patterns at the within-species line level. There was limited variation in effective dispersal between species and lines, and accordingly, we did not detect any correlation between effective dispersal probability and life-history traits. We discuss how expanding our experimental design by accounting for the density-dependence of both the pace of life and dispersal might improve our understanding of those traits and how they interact with each other. Overall, our results highlight the importance of exploring covariation at the “right” taxonomic scale, or multiple taxonomic scales, to understand the (co)evolution of life-history traits. They also suggest that optimizing both reproductive and development traits to maximize the efficiency of biocontrol may be difficult in programs using only one species.


39
Life history describes the life cycle of an organism, how fast and how much it grows, 40 reproduces, and survives. It is the direct product of a collection of phenotypic traits, called 41 life-history traits (Flatt & Heyland, 2011). Those traits include growth and mortality rates, 42 survival, reproductive investment or even the lifespan, and can be age-or stage-specific. 43 When all life-history traits and the values they can take are combined, many pathways can 44 lead to evolutionary success, resulting in the high diversity of what are called life-history 45 strategies, the covariation through time and space of different traits, found across the tree of 46 life. This high diversity can be observed at multiple taxonomic levels, from the phylum level 47 to within species (Gaillard et  histories can be summarised on a small number of key axes, which often reflect trade-offs 51 between life-history components. It is generally accepted that those life-history-trait 52 correlations arise from trade-offs between allocating a certain amount of acquired resources 53 into one trait or another, with limitations arising from a limited pool of resource to draw 54 from, physiological constraints, and from the influence of the environment, resulting in a 55 variety of strategies maximizing fitness (Laskowski et al., 2021;Stearns, 2000). 56 One specific axis has been termed the pace of life and corresponds to a correlation between 57 life-history traits sorting organisms along a fast-slow continuum (Braendle et al., 2011; 58 Stearns, 1983). Many trait combinations can be used to characterize a pace-of-life axis 59 (Gaillard et  Dispersal can be described as any movement potentially leading to a flux of genes or 76 individuals across space (Ronce, 2007), and is a key component influencing both ecological 77 and evolutionary dynamics, so much that it is sometimes described as a life-history trait in 78 its own right (Saastamoinen et al., 2018). Dispersal often covaries with other traits, including 79 other life-history traits (Clobert et al., 2012), in so-called dispersal syndromes (Ronce, 2012). 80 Dispersal syndromes have been observed and compared at multiple taxonomic levels, both 81 across (Stevens et al., 2012(Stevens et al., , 2014 and within species (Jacob et al., 2019

Biological material 130
Trichogramma are endoparasitoids, which means that females lay their eggs inside their 131 hosts, where the larvae will develop by feeding on the host and ultimately killing it, as 132 opposed to ectoparasitoids, who lay their eggs and develop outside their host. As some of 133 Trichogramma hosts are Lepidopteran pest species, several Trichogramma species are used 134 as biological control agents, and have shown to work well (Smith, 1996). For instance, 135 T. brassicae is used on a large scale against Ostrinia nubilalis, the European corn borer (Mertz 136 et al., 1995), and T. evanescens, T. cacoeciae, or a mix of the two species can be used against 137 Cydia pomonella, an apple pest (Sigsgaard et al., 2017

Experimental design 170
We used both single-and two-vial systems to measure life-history traits (Fig. 1). In single-171 vial systems (12 replicates per line), we placed one randomly selected mated Trichogramma 172 female between 24 to 48 hours old into a plastic vial (5 cm diameter, 10 cm height). We also 173 added a non-limiting quantity of irradiated Ephestia kuehniella eggs on a paper strip 174 (hundreds of host eggs in approximatively 1.4 × 1 cm, see Supplemental Figure S2-1). This 175 system was used to measure development time and fecundity traits. In two-vial systems (20 176 replicates per line), the setup was similar to the previous one, with the exception that a see-177 through 40 cm long plastic pipe (5 mm of internal diameter, large enough for species of less 178 than a millimetre in size) connected the first vial (where the wasp was deposited) to another 179 one with the same dimensions, also containing a non-limiting quantity of irradiated eggs. The 180 ends passed through the centre of the foam plugs without protruding from them. For endoparasitoids, the body size is highly dependent on the host size. In our case, all species 201 were maintained and experimented using E. kuehniella as host eggs, which are small enough 202 to allow only one viable descendent (Corrigan et al., 1995) and were provided in high enough 203 quantity to avoid superparasitism (as multiple eggs within one host might affect the viable 204 descendent size). Therefore, we assumed that size variance was probably highly limited, with 205 little to no correlations between hind tibia length (one proxy of individual size) and other 206 traits (Pavlík, 1993) and did not measure size. 207

Fecundity and dispersal 208
A week after isolation, parasitoid larvae were developed enough to blacken the host egg, 209 allowing the visual identification of successfully parasitized eggs (picture in Figure 1). Egg 210 strips (one for single vial, two for two-vial systems) were then photographed (resolution: 211 6016 × 4016 pixels, for a real field of view size of around 12 × 8 cm) using a Nikon D750 212 camera (lens: AF-S Micro NIKKOR 60 mm f/2.8 G ED) fixed above the strips. was at best around a hundred, and each of our host egg strips counted several hundreds of 219 eggs, we can assume that our study was indeed done in a non-limiting context. Furthermore, 220 in general, only one adult emerges from E. kuehniella eggs in the end (Corrigan et al., 1995; 221 Klomp & Teerink, 1966). 222 Egg retention by refusing to oviposit was previously observed in T. principium and 223 T. brassicae (Fleury & Boulétreau, 1993;Reznik et al., 2001Reznik et al., , 1998. Therefore, egg retention 224 may be present in all of the studied species and may affect fecundity measures in the 225 timeframe of our experiment; see below for how this possibility was accounted for in the 226 context of Data analyses. 227 In two-vial systems, effective dispersal (i.e. movement between patches leading to actual 228 gene flow) was measured as a binary response, where one female is considered to have 229 successfully dispersed if at least one parasitized egg was found on the strip present in the 230 second plastic vial. 231

Development time 232
After taking the pictures for fecundity, each isolated host egg strip was checked every day at 233 around 9:00 a.m., 12:00 p.m., and 4:00 p.m. for the presence of emerged individuals. The 234 development time of one replicate was considered to be the number of days between the 235 female in the plastic vial starting to lay eggs and the emergence of the first offspring. Note 236 that the true time is only known to a precision of two days, because of uncertainty in when 237 precisely eggs were laid during the 48 h window after introduction in the system (see Data  238 analyses for how this is accounted for). 239 interpreted as the fecundity of individuals that did not perform egg retention. 264

Data analyses
From now on, we will use "fecundity without retention" to refer to this 265 fecundity component (i.e. the effectively egg-laying individuals only), and 266 "overall fecundity" will refer to the mean number of eggs laid by all individuals, 267 including those potentially doing retention. 268 We used the model architecture described above for two multivariate models. The two 269 multivariate models were fitted to observe how variance in traits and the covariance between 270 traits are partitioned at the inter-and intra-specific levels. The first model incorporated both 271 line and species-level effects, structuring the variance into intra-and inter-specific levels. The 272 second model only had line effects as predictors, and therefore assumed that individuals from 273 two conspecific lines do not resemble each other more than individuals from two randomly 274 selected lines. In both cases, the same predictors were used for all four responses. 275 The first model included species-level effects as a fixed effect, mostly due to the low number 276 of species studied, and line identity was coded as a random effect, while the second model 277 only included line-level random effects. To account for line-level correlations between the 278 response variables, line-level random effects for the two models were modelled as drawn 279 from a shared variance-covariance matrix (Bürkner, 2017a). 280 While phylogenetic comparative methods could be used in this context, as some of the 281 variations could be explained by shared ancestry (Felsenstein, 1985), there is no 282 phylogenetic tree available for all lines used we could include (Hadfield & Nakagawa, 2010). 283 Our first model, splitting variation into species and line components is nonetheless similar to 284 the "taxonomic model" suggested in these cases where tree data are absent (Hadfield & 285 Nakagawa, 2010).

295
Trichogramma bourarachae had lower fecundity without retention and higher development 296 time than Trichogramma brassicae, while T. semblidis only had a lower development time 297 than T. bourarachae but no clear difference in fecundity without retention (Table 2, Figure  298 2B, C). There were no other clear species differences (based on 95 % intervals of pairwise 299 differences) in fecundity or development time. We did not find any evidence for between-300 species differences in effective dispersal probabilities (  Figure 2D). 303 Correlations between traits at the line level were analysed through the random effect 309 correlation/covariance matrix. In the first model, differences across species were modelled 310 with a fixed effect, so they were not included in random effect correlations, while the second 311 model included both species-and line-level random effects. Therefore, any qualitative 312 difference between the two models can be interpreted as an effect at the species level. 313 The only detectable correlations among traits were between fecundity without retention and 314 development time (Table 3, Figure 3). There was a negative correlation between these two 315 traits at the line level in the model where species effects were not partitioned out (-0.62 [-316 0.92; -0.28], see Table 3 bottom, see also the overall pattern Figure 3). However, when 317 looking at the model where species differences are partitioned out into fixed effects (Table  318 3 top), this random effect negative correlation mostly vanishes (-0.22 [-0.76; 0.38]). This 319 reflects the fact that the overall correlation highlighted in Table 3 top is mostly driven by 320 between-species differences in both fecundity and development time (see Figure 2 and 321 species averages in Figure 3). 322 no index letters in common are considered to have "significant" pairwise comparison 331 differences (i.e. the 95 % highest density interval of the difference does not include 0). White 332 dots represent observed means per species, presented for illustrative purposes only (as they 333 are calculated assuming all observed zeroes in egg numbers were attributable to retention, 334 and using the midpoint of the 48 h interval for development time). 335

Identification of one interspecific Pace-of-Life axis in Trichogramma 350
We found a negative between-line correlation between development time and fecundity in 351 this subset of five Trichogramma species, with high fecundity without retention, fast 352 development time on one side, and low fecundity, slow development on the other (Figures  353  2, 3, Table3). This correlation, which matches the classical pace-of-life axis (Healy et al.,354 2019) is mainly or only due to species-level differences: species with higher fecundity also 355 had faster development times (Figures 2, 3), and the line-level correlation vanishes when 356 species differences are partitioned out ( Table 3). We note that even if there is no statistically 357 significant correlation when the variance is structured within species and among lines, the 358 sign of this correlation remains negative (Table 3 top), following a similar tendency to the 359 interspecific negative correlation observed in Table 3  individuals, traits may be more responsive to direct environment variation through 374 phenotypic plasticity, and a shorter evolutionary timescale may lead to lower variation range 375 compared to a higher level (Siefert et al., 2015). 376 However, this pace-of-life finding is based on splitting fecundity into what we interpret as 377 egg retention and fecundity without retention components. While a significant negative 378 correlation with development was found on the latter component of overall fecundity, results 379 are more complex for retention probabilities. Indeed, there is no evidence for the line-level 380 correlation between egg retention and other life-history traits (Table 3). Furthermore, at the 381 species level, faster species (lower development time and higher fecundity in the absence of 382 retention) were also the species with the highest retention probabilities (Figure 2, Table 2). 383 If we interpret retention rates as a trade-off between present reproduction and future 384 opportunities, then high retention can be seen as a "slow" trait; its association with "faster" 385 life history traits may then appear paradoxical. It might be that fecundity in the absence of 386 retention and retention probabilities are not actually separate traits, and that the trait 387 correlations described above derive from their "artificial" separation by the statistical model. 388 However, previous studies indicate that in T. principium, except for prolonged periods of egg-389 retention, individuals manifesting egg retention had similar fecundities in their first days of 390 actual egg-laying and similar lifetime fecundities than individuals that did not (Reznik et al., 391 2001(Reznik et al., 391 , 1998 life-history strategies, like "slow" adult reproduction alongside "fast" offspring survival (that 401 the authors likened to an oak tree life history) or the opposite (represented by mayflies). 402 However, because egg-laying was restricted to a 48 h window in our experiment, we cannot 403 yet confirm this interpretation. Further studies measuring lifetime reproductive success, 404 longevity or the way reproductive effort is spread throughout the lifetime may shed more 405 light on the way life history is structured in Trichogramma wasps. 406

No evidence for a syndrome linking effective dispersal probability and the pace of life 407
Effective dispersal probability varied the least among the four traits measured, with no 408 evidence of between-species or even between-lines differences (Figure 2), and values were 409 rather consistent with previous studies (Dahirel, Bertin, Calcagno, et al., 2021). There was 410 also no correlation between effective dispersal and any of the other traits ( patches completes previous studies on the activity of Trichogramma species. In Wajnberg & 414 Colazza (1998), the authors showed a significant difference in the average area searched 415 within one patch by T. brassicae isofemale lines while our results showed no differences in 416 effective dispersal (Figure 2). In Reznik & Klyueva (2006), T. principium females manifesting 417 egg retention had higher dispersal activity in a continuous environment than females that 418 laid eggs beforehand. This discrepancy may be the result of a focus on different taxonomic 419 levels: Reznik and Klyueva (2006)'s results deal with within-species and within-line 420 covariation, versus between-lines and between-species in the present study. It may also 421 result from differences in experimental designs and metrics used: the dispersal metrics used 422 in Reznik and Klyueva (2006) are based on short-term (less than one day) and short-distance 423 (up to 5 cm) movement on a continuous arena, compared to our experiment (two days and 424 40 cm between discrete patches). In that case, there may still exist in Trichogramma a pace-425 of-life syndrome linking life history to short-term activity and behaviour, but not effective 426 dispersal. Indeed, correlations between short-term movement activity and life-history traits 427 were also found in T. evanescens at the between-line level ( fecundity and competitive abilities are to be favoured for efficiency (Smith, 1996