Personality and gonadal development as sources of individual variation in response to GnRH challenge in female great tits

Abstract

Seasonal timing of reproduction is a key life-history trait, but we know little about the mechanisms underlying individual variation in female endocrine profiles associated with reproduction. In birds, 17β-oestradiol is a key reproductive hormone that links brain neuroendocrine mechanisms, involved in information processing and decision-making, to downstream mechanisms in the liver, where egg-yolk is produced. Here, we test, using a simulated induction of the reproductive system through a Gonadotropin-Releasing Hormone (GnRH) challenge, whether the ovary of pre-breeding female great tits responds to brain stimulation by increasing oestradiol. We also assess how this response is modified by individual-specific traits like age, ovarian follicle size, and personality, using females from lines artificially selected for divergent levels of exploratory behaviour. We show that a GnRH injection leads to a rapid increase in circulating concentrations of oestradiol, but responses varied among individuals. Females with more developed ovarian follicles showed stronger responses and females from lines selected for fast exploratory behaviour showed stronger increases compared to females from the slow line, indicating a heritable component. This study shows that the response of the ovary to reproductive stimulation from the brain greatly varies among individuals and that this variation can be attributed to several commonly measured individual traits, which sheds light on the mechanisms shaping heritable endocrine phenotypes.

1. Introduction

Reproduction is one of the most demanding stages of the life cycle in many organisms. To balance the costs inherent to breeding with the benefits in terms of fitness, rearing of offspring must be timed appropriately to match the period when energy sources are most abundant in the environment [1–3]. To time their breeding, many animals use various endogenous and environmental cues that herald the approach of suitable times. Those cues stimulate the neuroendocrine system to orchestrate a cascade of physiological reactions that are necessary to make the reproductive system fully operational [4]. Our current understanding of the mechanisms responsible for the pre-breeding activation of the reproductive system mostly comes from the study of the signalling properties of photoperiod [5,6], which was mostly studied in males [7–9]. Timing of breeding has, however, been shown to be under female control [10–13]. This highlights the need for a better understanding of the mechanisms that activate and regulate the reproductive physiology in females specifically [9,14]. Besides focusing on males, the majority of studies on reproductive physiology have also focused on central tendencies (i.e. averages between groups) instead of analysing variation within and among individuals [15]. Hormone concentrations are notoriously variable, even between individuals that are at a same life-history stage [16,17]. Such variation, when heritable, is the main target of natural selection, and by extension the engine of microevolution and adaptation [18–20]. Ignoring individual variation and phenotypic plasticity in physiological traits like hormones limits our capacity to link physiology with more evolutionary aspects of seasonal timing, which is required for predicting how individuals can cope with environmental changes [21].

In most vertebrates, the perception of photoperiod (and possibly other cues) by dedicated receptors activates or inhibits the gonadotropin-releasing (GnRH) and -inhibitory (GnIH) hormone systems at the level of the hypothalamus. These neuro-hormones are released in the portal veins to regulate the release of the gonadotropins, luteinizing hormone (LH) and follicle-stimulating hormone (FSH), by the anterior pituitary into the general circulation. LH and FSH then bind to receptors in the ovary and testis, stimulate their development, the gametogenesis and the production of steroid hormones, which are involved in other physiological and morphological changes (e.g. secondary sexual characters) and increase the probability of sexual behaviours to occur. In female birds, one particular steroid is 17β-oestradiol that stimulates the production by the liver of the yolk precursors vitellogenin (VTG) and very low-density lipoproteins (VLDL) [16]. VTG and VLDL constitute about 90% of yolk dry weight. In particular, 17β-oestradiol induces a change in the liver production of VLDL from large, generic VLDL, to small VLDL particles that can pass through the pores of the granulosa cell layer of the ovarian follicles [14]. This change in VLDL production coincides with the transition from the slow to the rapid follicle growth stage during which the majority of proteins and lipids are added to the yolk to produce large yellow follicles, and which precedes ovulation and egg-laying by only a few days [14]. Oestradiol therefore lies at the crossroads between the brain and the downstream effector mechanisms of reproduction.

Historically, the hypothalamo–pituitary–gonadal (HPG) axis described above has been viewed as a system where the brain plays the prominent role, meaning a top-down model of regulation where the GnRH/GnIH systems are in control, and the levels below (pituitary and gonadal secretions) are simply responding to the stimulations by the upper levels (hypothalamus). While this model seems to explain male reproductive development well, in females, it is more complex. For example, exposing birds to long photoperiods in captivity leads to a complete sexual activation with fully developed gonads and active sperm in males, but females rarely reach the final ovarian development and egg-laying stages, despite elevated concentrations of gonadotropins [22–24]. Furthermore, injecting GnRH fails to increase oestradiol concentrations in female song sparrows (Melospiza melodia) [25] and zebra finches (Taenopygia guttata) [26], and GnRH-induced levels of LH and testosterone do not correlate in female dark-eyed juncos (Junco hyemalis) [27]. This suggests that long photoperiods and high levels of GnRH and LH/FSH are not sufficient to trigger the final stages of gonadal development in females, and that steroid production by the ovary does not simply mirror what happens at the brain level. The fine control of rapid follicular development and ovulation by supplemental cues could thus lie at a more downstream tier of the HPG-axis, at the level of the ovary or the liver [11,14,28]. The relative independency of the downstream part of the HPG-axis in females is further supported by the discovery that the largest follicle in the ovary of the chicken controls the LH surge necessary for its ovulation, through circadian clock gene proteins that it produces [29]; and that in starling sex steroid secretion by the gonads can be regulated by alteration of GnIH present locally in the gonads [30]. In such a model of the downstream regulation of reproduction, the critical switch for triggering egg-laying might be the response of the ovary to gonadotropins, or of the liver to oestrogens.

In this study, we tested whether the ovary of female great tits maintained under a naturally long photoperiod in spring, responds to an artificial and rapid elevation of GnRH by increasing the release of 17β-oestradiol (E₂), and we explored how several individual factors, namely ovarian follicle size, personality, and age of the females, modulate this individual response. We thus aimed to test whether and how individual characteristics of the phenotype modulate ovarian oestradiol and its responses to brain stimulation. Using a repeated-measure design and GnRH challenges in females from lines artificially selected for divergent levels of exploratory behaviour [31], an operational measure of avian personality [32], we were able to assess the change in circulating E₂ concentration relative to each individual's initial value, thus with each individual acting as its own control.

2. Methods

(a). Subjects

We used 16 captive, non-breeding female great tits for this experiment. Birds had been hand-reared at the Netherlands Institute of Ecology (NIOO-KNAW) following standardized procedures [31]. The birds originated from the fourth generation of two lines bidirectionally selected for a combination of two personality traits. One line was selected for fast exploration and boldness, the other for slow exploration and shyness [33–35]. Birds were housed individually in cages (0.9 × 0.4 × 0.5 m) with solid bottom, top, sides, and rear walls, a wire-mesh front and three perches [36], in a room equipped with large open windows for natural lightning and airflow, so that birds were exposed to natural spring photoperiod and temperature. The experiment was conducted in May, at a time when day length is 16 h 20 min. Birds had visual and auditory contact with each other. Diet consisted of a mixture of commercial egg mix, cow heart, dry bird food, vitamins, and minerals. Birds also had access to mealworms and sunflower seeds. Food and water were provided ad libitum.

(b). Experimental procedures

Each bird was taken out of its cage and an initial blood sample (approx. 150 µl) was collected from the jugular vein, using a syringe (Easy Touch Insulin 0.3 ml, with 31 g, x5/16 inch needle, Houston, TX, USA). Blood was collected within 3 min of capture (average: 1 min 48 s, range: 1 min 20 s–2 min 10 s). Birds were then injected with 1.25 µg chicken GnRH-I (H-3106; Bachem, Bubendorf, Switzerland) dissolved in 50 µl of phosphate-buffered saline (PBS) solution [27,37,38]; 25 µl in each pectoral muscle. After this, birds were returned to their home cage. After 30 min, birds were anaesthetized with isoflurane (Abbott Laboratories, Lake Bluff, IL, USA), a second blood sample was collected from the jugular vein (maximal volume possible) and birds were euthanized by increasing the amount of isoflurane, as part of another study (van Oers et al., unpublished results). Blood samples were stored on ice, centrifuged and plasma was stored at −80°C until hormone assays.

(c). Hormone assays

Steroid hormones were extracted from plasma using solid phase extraction with C18 columns (100 mg C18 material, Sep-Pak Vac 1cc, Waters, Milford, MA, USA) to remove substances present in the matrix, mainly lipids, that could interfere with the assay. We used solid phase extraction columns instead of traditional organic solvents because organic solvents can lead to variable recoveries, or incomplete removal of interfering substances, when extracting steroids from samples with high lipid content, like those of females that are close to laying [39,40]. C18 columns were first primed with absolute ethanol, and equilibrated with deionized water. Deionized water was added to the plasma samples (25 µl), before loading onto the C18 columns. Columns were then washed, and steroids were eluted with 90% high-performance liquid chromatography (HPLC)-grade methanol. The eluates were dried at 40°C under a nitrogen flow. Dried extracts were resuspended in PBSg (PBS with 0.1% gelatin) containing 0.7% ethanol [41]. Average recovery after extraction, estimated from samples spiked with a small amount of ³H-E₂, was 46.1%. Final hormone concentrations were corrected accordingly. During validation procedures, we performed multiple estimates of recovery with the C18 columns, and the average recovery rate was 49.8%, with a standard deviation of only 6.7%. This suggests that our recovery values are quite constant, and by extension that the E₂ concentrations reported in the results are accurate.

Resuspended samples were then assayed in singleton using a commercially available double-antibody ¹²⁵I-E₂ radioimmunoassay (DSL-4800, Ultra-sensitive Estradiol RIA, Beckman Coulter, Brea, CA, USA) that was modified by dilution of the tracer and the antibody to increase the sensitivity of the assay, following [41]. Assay tubes were counted on a gamma counter (Automatic Gamma Counter, Perkin Elmer, Waltham, MA, USA). Concentrations of E₂ were obtained using a linear regression with the log-transformed concentrations of the standards provided in the assay kit. All samples were run in a single assay, and the intra-assay coefficient of variation, as estimated by assaying one high- and one low-concentration E₂ standard in triplicate, was 3.6%. Assay sensitivity was 0.6 pg ml⁻¹. It was defined as the highest point on the standard curve whose standard deviation did not overlap that of the blank standard [40]. No sample was found to be below the detection limit. The antibody used in the immunoassay is highly specific to oestradiol, with low cross-reactivity with related molecules (electronic supplementary material, table S1).

We validated the assay procedure for great tits and blue tits by serially diluting several plasma samples from these two species. The percentages of tracer bound (B/B0) from serially diluted samples were parallel to the standard curves (see electronic supplementary material, figure S1), indicating that the assay works correctly for those species, and that substances potentially interfering with the RIA were appropriately removed during solid phase extraction [39].

(d). Statistical analyses

We modelled plasma concentrations of E₂ with linear mixed effect models, using the function lmer of the package lme4 [42] in R (v. 3.2.1) [43]. The time period of blood collection (before or after GnRH injection), the selection line for personality (fast or slow explorer, n = 6 and 10, respectively), and the age class (old or young, where young = 3–4 years old, old = 8–13 years old) were used as categorical explanatory variables. We treated age as a categorical variable because there was little variation within age classes. Because small passerines of more than 8 years are rare, the old age class only had three birds. Even if the fits of the models with age were found to be good, the corresponding results will still be interpreted cautiously because of the low sample size in the old age-class. The size of the largest ovarian follicle (n = 16) was log-transformed to achieve normality, and used as a continuous explanatory variable. Individual identity was included as a random intercept to account for repeated measures on the same individuals. To test the effects of the above variables on plasma E₂ concentration, we ran several biologically relevant models that included each variable and its interaction with time period (table 1). We did not run a single model with all variables and interactions because it would have led to over-parametrization of the model [44]. Inferences were obtained using a Bayesian framework [45]. Candidate models in table 1 were then ranked using the widely applicable information criterion (WAIC), which is the Bayesian version of the Akaike information criterion (AIC), and which is based on the joint posterior distribution of the model parameters rather than the point estimates [45]. Within WAIC, the estimate of the effective number of parameters can be calculated in two ways, and we here used WAIC2 (package blmeco [45]) that most strongly resembles the cross-validation [45]. Like for AIC, the smaller the WAIC2 value, the better the model fits the data. We retained the models with the lowest WAIC2 scores and inferences were then drawn using improper, flat prior distributions. To obtain the posterior distributions, we directly simulated 1000 values from the joint posterior distribution of the model parameters using the function sim of the package arm [46]. The means of the simulated values from the joint posterior distribution of the model parameters were used as estimates, and the 2.5 and 97.5% quantiles as lower and upper limits of the 95% credible intervals. Posterior probabilities for effect sizes of particular interest (the ‘degree of belief’, [45]) were calculated by the proportion of simulated values from the posterior distribution being smaller/larger than a biologically relevant threshold (e.g. smaller/larger than another group level).

Table 1.

Model selection for 17β-oestradiol concentration in female great tits as a function of time (before or after GnRH injection, ‘t’), personality selection line (fast or slow, ‘p’), age (old or young, ‘a’), or ovarian follicle size (f). In all models, individual identity was included as a random intercept. Models are sorted by ascending order of WAIC2.

models	WAIC2	ΔWAIC2
t + f + (t × f)	301.7	0
t + p + (t × p)	302.9	1.2
t + a + f + p	317.9	16.2
t + a + (t × a)	320.3	18.6
t	323.8	22.1

3. Results

The mean plasma concentrations of E₂ were 60.5 pg ml⁻¹ (s.d. 23.0) before the GnRH injection and 98.9 pg ml⁻¹ (s.d. 52.7) 30 min after the injection. Female great tits (n = 16) thus had plasma concentrations of E₂ that on average increased by 38.4 pg ml⁻¹ (95% CrI 17.1–60.9 pg ml⁻¹, from a simple univariate model including only time period as a predictor and individual identity as a random intercept) following the injection of GnRH. According to the model selection procedure using WAIC (table 1), plasma E₂ concentrations of female great tits were best explained when, in addition to the time period relative to the GnRH injection, personality line (fast or slow exploring), or the size of the largest ovarian follicle were included in the models, together with their respective interactions with time period.

Females that have large ovarian follicles increased E₂ more during the 30 min post-injection period than females that have small follicles (figure 1a), with a posterior probability of having a higher increase in female with large follicles greater than 0.999. If baseline concentrations of E₂ do not reliably vary with ovarian follicle size (i.e. before the GnRH injection; posterior probability of 0.88), after the GnRH injection, E₂ levels are markedly higher in females with large follicles (posterior probability greater than 0.999) (figure 1a). Only four females had at least one yolky follicle (follicle size of 2.9 mm or higher).

Figure 1.

Effect of ovarian follicle size (a), selection line for exploratory behaviour (i.e. personality) (b), and age (young = 3–4 years old, old = 8–13 years old) (c) on variation of plasma 17β-oestradiol concentrations following an injection of GnRH to 16 female great tits. Filled circles and vertical bars, estimated means and 95% credible intervals. Open circles and dotted lines, individual data, with variation in thickness (a) or colour (b and c) of the lines depicting the different groups of birds. Although ovarian follicle size (a) was treated as a continuous variable in the statistical models, follicles were subdivided into three groups (small, medium, and large follicles) for clarity of graphical representation. (Online version in colour.)

Similarly, fast exploring female great tits increased E₂ more during the 30 min post-injection period than slow exploring females (figure 1b), with a posterior probability of having a higher increase in fast explorers greater than 0.999. While E₂ concentrations do not differ between fast and slow females before the injection of GnRH (posterior probability of 0.58), they do clearly differ 30 min later (posterior probability greater than 0.999).

If young females tend to increase E₂ more than old females, leading to young females having higher E₂ concentrations than old females 30 min after the GnRH injection (posterior probability of 0.997; figure 1c), the model with age has a much higher WAIC than models including follicle size or personality (table 1). Age can thus not be considered as exerting an important influence on plasma E₂ in the present study.

4. Discussion

Here, we show that female great tits exposed to natural photoperiod in spring respond to a GnRH injection by increasing their circulating concentrations of E₂. Absolute concentrations and responses to GnRH were, however, highly variable between individuals, with a fourfold variation in baseline concentrations, and with some females elevating their E₂ concentrations by up to 58 pg ml⁻¹, while others did not respond or even lowered their levels after the GnRH injection (figure 1). We found that personality and size of the largest ovarian follicles contributed to explain this variation between individual females. This shows that we identified important individual characteristics shaping reproductive physiology of female great tits.

The fact that female great tits increased their E₂ concentrations indicates that the ovaries of those captive females responded to GnRH. This contradicts earlier results in song sparrows [25] and zebra finches [26], but is in agreement with earlier studies in dark-eyed juncos (J. hyemalis), where females had more E₂ after GnRH or LH challenges than before [27,47]. The junco data were, however, not repeated measures from the same individuals, but resulted from pools of plasma from different females [27], or from samples of different females before and after the challenges [47]. Our results thus suggest that female ovarian follicles were able to prompt the liver to produce VTG, the small VLDL particles defining the onset of the rapid follicular growth stage and, by extension, egg-laying. Such capacity to rapidly increase E₂ concentration in the blood circulation suggests that the ovary may not be the main downstream tier that blocks egg-laying as long as all the necessary conditions (cues) are not met. One hypothesis that remains is that this main downstream regulator is the liver, which might only respond to the elevated concentrations of E₂ in some circumstances. This would be in agreement with data in female starlings where injection of E₂ leads to an increase in VTG, but not of VLDL [48], even though in zebra finches, both yolk precursors were shown to increase in response to E₂ injections [49,50]. We did not measure plasma VLDL here because our post-GnRH injection protocol was designed to measure E₂ only. Measuring the liver response to GnRH would have required a different sampling scheme. If in vitro, the apoVLDL-II mRNA increases very rapidly after oestrogen stimulation in chicken [51], how quickly this increase in transcript availability translates into secretion of VLDL in the bloodstream is unknown, let alone in wild species. In the aforementioned studies on zebra finches, yolk precursors only started to be measured 1–5 days after the first E₂ injection [49,50]. Alternatively, one could argue that even though E₂ increases, it might not increase sufficiently to stimulate the liver. Although to our knowledge circulating concentrations of E₂ in female great tits have never been published before, and even if comparison between assay protocols is debatable, we have measured plasma E₂ concentration in some female great tits that were breeding in captivity. During egg-laying, average E₂ levels were around 150 pg ml⁻¹, thus higher than what we describe here, while levels in those same females were at 80 pg ml⁻¹ during the nest-building period (Caro et al., unpublished data). The latter is intermediate between our mean baseline (approx. 60 pg ml⁻¹) and GnRH-induced (100 pg ml⁻¹) concentrations, showing that the concentrations observed here are comparable to those that occur in females that are close to egg-laying. Our results thus show that E₂ is a critical component linking the brain neuroendocrine mechanisms resulting from information processing and decision-making, to more downstream effective mechanisms in the liver, where egg-yolk is produced just before laying. We should now determine whether and how the liver could modulate these upstream stimulations transiting by the gonads, in order to adjust the seasonal timing of reproduction to prevailing environmental conditions.

The size of the largest ovarian follicle was one of the factors that explained inter-individual differences in E₂ concentrations. If baseline levels of E₂ were not affected by follicle size, the larger the follicles, the more E₂ was released in response to GnRH (t × f interaction, figure 1a). This suggests that the enzyme aromatase, which synthetizes E₂ from testosterone [52], present in the cells of the theca externa of the ovarian follicles has responded to the GnRH-induced elevation of LH and/or FSH by increasing the production of E₂ [53,54]. Since large follicles contain more theca cells than smaller follicles, the steroidogenesis capacity of those large follicles is higher, resulting in a higher response to GnRH in females with large ovarian follicles. Alternatively, GnRH could have induced the release of gonadal stores of E₂. Although steroids are assumed to be released in the bloodstream directly after their production by the gonads, some studies have suggested that they could be stored locally in cells of various organs. E₂ has indeed been suggested to accumulate in the lipid bilayer of cell membranes where it could stay until picked up by a carrier protein [55,56]. Sex hormone binding globulins (SHBG), which bind E₂ and T with high affinity, have also been reported to accumulate in various reproductive tissues, such as the ovary, uterus, or spermatozoids [57,58]. Female great tits could thus potentially have stored some E₂, stores that would have been released following the injection of GnRH. We did not measure LH and FSH, which could have shed light on the multi-level interactions between the brain, the pituitary, and the ovary. Large follicles could possess more gonadotropin (but also GnRH) receptors, and variation in gonadotropin responses to GnRH could also contribute to the observed variation in E₂. Besides the fact that FSH antibodies are, to our knowledge, not currently available for birds, GnRH-induced LH concentrations have been shown to peak (within 5 min) and return to basal levels (within 30 min) very rapidly in different bird species [47,59,60]. As for VLDL, the optimal time point for measuring GnRH-induced LH is therefore very different than for E₂, and further studies are warranted to elucidate how the various tiers of the HPG-axis interact in the minutes/hours following brain stimulation.

The selection line for exploratory behaviour (used as a proxy for personality [31,61]) from which the females originated was another important individual characteristic that explained the variation in E₂ between females. Personality was found to be associated with variations in (neuro-) endocrine profiles in various taxa, including fish, mammals, and birds [62–64]. Glucocorticoids, for example, tend to fluctuate less in response to a stressor in bold than in shy individuals [65–67], but see [35]. While another steroid, testosterone, was also shown to differ between lines selected for diverging levels of personality in both males and females [68,69]; to our knowledge, the relationship between behavioural profiles and oestrogens levels has never investigated in birds. We show here that fast explorers (also described as bold or proactive) responded more strongly to the GnRH challenge than slow explorers did (often described as shy or reactive). E₂ is the main sex steroid in females and, besides its role in the final stages of egg production, it is also generally associated with aggressiveness in both males and females [52,70–73], and fast explorers tend to be more aggressive [74–76]. Thus, although fast explorers do not necessarily present higher baseline concentrations of sex steroids ([68,77], this study), our results suggest that they have a higher capacity to mount a rapid E₂ response when challenged, which could explain why fast explorers tend to be more aggressive and dominant in agonistic interactions [76]. Importantly, the fact that E₂ differs between selection lines implies some level of genetic variation in the physiological mechanisms characterizing those lines [78]. Along those lines, we also found that the size of the largest follicle was influenced by the line of selection for personality, with fast and bold line females having larger ovarian follicles (p = 0.007, data not shown). To our knowledge, such a relationship has not been described before in great tits or other species. How exactly the two traits interact is currently unclear, but this suggests on the one hand that the personality effects on E₂ are likely mediated by gonadal size, and on the other hand that there has been a correlated selection for the size of the gonads during the establishment of the two selection lines for personality traits (exploration and boldness). Disentangling the personality and gonad size effects on E₂ will require further exploration, possibly on a sample of birds that vary largely in their personality scores without having been artificially selected. In any case, such kind of genetic variation, combined with the inherently large variations in hormone titres among individuals, indicates that in the wild, natural selection can act on the mechanisms underlying reproduction and shape the phenotypes via microevolution [21]. In changing environments, like under global warming, micro-evolution might become necessary for organisms to mitigate the ecological impact of those changes [79,80]. Describing genetic variation in the mechanisms underlying the phenotype is a first necessary step, but future studies will need to assess at what rate those mechanisms can respond to selection to determine whether microevolution will proceed sufficiently fast to cope with the rate of environmental change.

The age of great tit females did not convincingly affect their E₂ concentrations. Even though a difference in E₂ levels between age classes was visible following the GnRH injection, with lower concentrations in old females (figure 1c), the WAIC score of the model including age was quite high and the interaction between age and time period was not particularly reliable. This might result from a lack of power due to the small number (3) of very old females available (depicted by the large credible intervals around the estimated means, figure 1c). Senescence seems, however, widespread among vertebrate taxa, including in physiological traits like immune function, muscular mass, or fecundity (reviewed in [81]). While there have been many studies looking at how ageing affects reproductive performances in terms of offspring produced (e.g. [82–84]), relatively fewer have tried to quantify how ageing changes the underlying physiological mechanisms of reproduction [85], especially in birds and in wild species [86]. In Florida scrub-jays (Aphelocoma coerulescens), for example, baseline levels of LH and oestradiol did not differ between females of different ages, but both baseline and GnRH-induced levels of LH and testosterone decreased with age in males, suggesting that senescence can affect some reproductive functions in this species [87]. Additional work is thus warranted to describe ageing mechanisms in birds in general, and great tits in particular.

In conclusion, while hormone titres are notoriously variable between individuals, our study identifies several important sources of variation in E₂ among female great tits. These results are important both on a physiological perspective, for understanding the factors that shape endocrine profiles around reproduction, but also in an evolutionary context, since those individual variations underlie the evolution of phenotypes by providing the raw material on which natural selection can act.

Ethics

Procedures were approved by the Institutional Animal Care and Use Committee: the Royal Academy of Sciences—Animal Experiment Committee (KNAW-DEC NIOO-CTE07.05 to K.v.O. and NIOO-11.09 to S.P.C.).

Data accessibility

Data available from the Dryad Digital Repository at: https://doi.org/10.5061/dryad.t6r3mq1 [88].

Authors' contributions

S.P.C. conceived the study, designed the study, carried out the experiment, carried out the laboratory work, analysed the data, and wrote the manuscript; C.A.C. carried out the laboratory work, analysed the data, and wrote the manuscript; K.v.O. and M.E.V. conceived and designed the study, and wrote the manuscript.

Competing interests

We declare we have no competing interests.

Funding

S.P.C. was supported by an NWO-VENI (863.09.011), and a WBI-World grants.