Growth potential can affect timing of maturity in a long-lived semelparous fish

Abstract

Many diadromous fishes such as salmon and eels that move between freshwater and the ocean have evolved semelparous reproductive strategies, but both groups display considerable plasticity in characteristics. Factors such as population density and growth, predation risk or reproduction cost have been found to influence timing of maturation. We investigated the relationship between female size at maturity and individual growth trajectories of the long-lived semelparous European eel, Anguilla anguilla. A Bayesian model was applied to 338 individual growth trajectories of maturing migration-stage female silver eels from France, Ireland, the Netherlands and Hungary. The results clearly showed that when growth rates declined, the onset of maturation was triggered, and the eels left their growth habitats and migrated to the spawning area. Therefore, female eels tended to attain larger body size when the growth conditions were good enough to risk spending extra time in their growth habitats. This flexible maturation strategy is likely related to the ability to use diverse habitats with widely ranging growth and survival potentials in the catadromous life-history across its wide species range.

1. Introduction

A major question in life-history theory has been what determines the reproductive strategies of species and how these are adaptive to diverse biotic and abiotic selection pressures. Many short-lived plants and animals reproduce once and die (semelparity) and the alternative, repeated reproduction is known as iteroparity [1]. Among long-lived organisms semelparity is relatively rare, but diverse taxonomic groups including plants, insects, cephalopods, marsupial mammals and fishes contain long-lived semelparous species [1]. Long-lived semelparity is exceptionally common in the diadromous fishes, with the best-known examples being salmon and anguillid eels. They have an inverse pattern of diadromy though, with spawning occurring in freshwater and growth occurring in the ocean for salmon and vice versa for eels [2].

Both groups show various degrees of plasticity in aspects of their life histories, with eels ranging widely in their entry far into freshwater and in their growth rates and age at maturation [3–5]. As for many species, age and size at maturation are key life-history traits that affect growth rate, survival and fecundity [6], but for eels it is unclear what triggers the maturation process [7]. The relationship between size and the maturation probability has been intensively studied in teleost fishes, but how growth affects size and age at maturation is not yet fully understood [8,9]. The European eel (Anguilla anguilla) lives across a wide range of latitudes from North Africa to Scandinavia, so they have variable ages and sizes at maturation [4,5,10] and these environmental differences may affect the timing of the start of the maturation–migration process. Factors such as population density, feeding success and growth or predation risk may also affect timing of reproduction, as shown in salmonids [11].

The 4000–7000 km, or longer, migration of European eels requires that enough energy be stored in the body to enable completion of the long journey [9], so there must be a threshold-size that eels must attain before starting to mature and become silver eels (adult migratory stage). Female eels may use a size-maximizing strategy [12] because fecundity is entirely determined by body size, but extending the growth phase can increase the risk of mortality. This trade-off between size at maturity and growth rate suggests that females begin the process of silvering when growth is slowing down (see discussion in [13]).

We investigated the link between maturation probability and body size in European eels. We tested the hypothesis that the probability of becoming a maturing silver-stage eel was linked to eel growth, with poor growth leading to a higher probability that silvering maturation may be triggered at smaller sizes.

2. Material and methods

Body size and age data of silver-phase female European eels (N = 338) collected from Europe (Ireland, France, the Netherlands and Hungary) during 2000–2007 were used to provide a wide overview of population-level traits [14]. Data were extracted from the EU-FP5th EELREP database (Loire, Ste. Eulalie, Grevelingen, Balaton, Certes, Nive and Rhine river systems), and additional data from the Burrishoole ([13,15], R. Poole 2008, unpublished data) and Corrib (R. Poole 2005, unpublished data) rivers were obtained.

Individual growth trajectories were acquired using data from otoliths (calcium carbonate structures in fish inner ears). Back-calculation analysis of the relationship between body size and otolith radius followed [15]. Measurements were made of radius (mm) of the ith annulus (R_i), which is the distance from the otolith mark at recruitment to the ith annulus, and of the radius (mm) of the otolith (R). The total length (L_T) of the fish at age i years (L_i, mm) was estimated using the following formula: L_i = L_r + (L – L_r)R_iR⁻¹, where L_r is the mean L_T of glass eels recruited at the coast [5], and L is the L_T at capture (mm). The annual body increment (G_i, mm yr⁻¹) was calculated as G_i = L_i – L_i−1.

The average growth rate during a given period until the year preceding silvering (differential in individual size–age relationship) and growth acceleration/deceleration (second-order differential of individual size–age relationship) were selected as explanatory variables. Logistic regression models describing the probability of silvering were constructed to test whether body size and growth history were significant proximate cue(s) for silvering. The basic form for these logistic models was logit(p) = log_e [p(1 − p)⁻¹] = c₀ + c₁L_t,j + c₂ L'_t,j + c₃L''_t,j + Inds, where p is the probability of silvering (early maturation), c₀ is a constant, c₁ is the coefficient for the size effect (L_t,j) at various ages t of individual j, c₂ is the coefficient for the individual's growth (L'_t,j) from age t–4 to age t of individual j, c₃ is the coefficient for the acceleration/deceleration of growth (L''_t,j) between age t–4 ∼ t and age t–5 ∼ t–1 of individual j, and Inds is the random effect of individuals.

Maturation probability was fitted to a sigmoid curve using the Bernoulli distribution and non-informative priors were deployed as priors and hyperpriors. Three Markov chain Monte Carlo chains were initiated at the maximum-likelihood estimates and were run for 5000 iterations as a burn in, after which every fifth iteration was recorded to remove autocorrelation, until 1000 samples were obtained. To assess the significance of the parameters, we confirmed that the estimated probability distributions of parameters did not include zero within the distribution range. Convergence of the estimated parameters was confirmed by the iteration figures and by noting that R̂ was near 1. The random effect was assessed by the deviance information criterion (DIC) to calculate the significance of the model. R (R2Winbugs.package) and WinBugs [16] were used for analysis.

3. Results

The total length at silvering of female eels from Ireland, France, the Netherlands and Hungary (N = 338, figure 1) varied widely, from 436 to 982 mm, with just one smaller eel (377 mm). The size distribution showed that for the onset of silvering maturation in eels a certain minimum size must be achieved, which appears to be about 430 mm (one outlier) [14]. The age and growth rate of silver eel females ranged widely, from 4 to 44 years and from 12.1 to 148.2 mm yr⁻¹, respectively. The annual body increment (G_i) of females and its variability declined slowly with age until about age 10, then declined more rapidly until age 20, after which they stabilized (figure 1).

Figure 1.

(a) The estimated mean body increment (growth in 1 year) and standard deviation (bars) at each age and (b) size and age at silvering maturation of individual female European eels. (Online version in colour.)

In the model, size (L_t,j), average growth (L'_t,j) and the acceleration/deceleration of growth (L''_t,j) had significant effects on female maturation probability. The significantly improved mixed-model fit was confirmed by its lower DIC (1554) compared with a fixed-effects model (1619). Parameter convergence was confirmed visually, and R̂ values ranged from 1.02 to 1.15 (table 1). The estimated distribution of L_t,j positively affected, whereas those of L'_t,j and L''_t,j negatively affected, the silvering probability (figure 2). Furthermore, the distribution ranges of these parameters did not include zero within the ranges (table 1), indicating that lower growth in recent years and/or a large growth deceleration in growth leads to a higher probability of silvering.

Table 1.

Parameter estimates with 95% confidence intervals from the best probability model for silvering maturation of female European eels with random effects of individuals.

parameter	estimate ± s.d.	95% CI	R̂
body size (L)	0.016 ± 0.003	0.011 to 0.021	1.15
recent 5-year growth (L')	−0.015 ± 0.003	−0.021 to −0.009	1.13
inclination of growth (L'’)	−0.032 ± 0.005	−0.041 to −0.023	1.02

Figure 2.

Probability curves of silvering maturation as a function of body size (total length) in the Bayesian model for (a) recent 5-year growth (L'_t,j) and (b) the deceleration of growth (L''_t,j) with several values of growth or trends. (Online version in colour.)

4. Discussion

This study demonstrated that when the recent growth in their present habitat was poor, European eels tended to start the silvering maturation process and leave their growth habitat to begin their spawning migration. This indicates that constant high growth leads to a larger size at silvering, whereas poor or decreasing growth results in silvering of eels earlier at smaller sizes. Our results also suggested that the silvering size was linked with both deceleration of growth and the growth rate. Similarly, a study using rainbow trout introduced into a range of habitats found that factors such as density and growth affected the characteristics of age and size at reproduction [11].

Our model is in agreement with the dynamics of internal factors during silvering. In silver eels, growth-related physiological traits are weakened whereas maturation-related physiological traits are enhanced [17,18]. From an endocrinological point of view, the silvering corresponds to a conversion of the somatic growth mode into a maturation mode. This type of internal biological process might be genetically mediated and related to the genetic background of individuals [19].

For the European eel, our model is likely applicable to the empirical understandings of eel growth [20]. For example, some steady-growth older female eels remain as the top predators in their ecosystems, like Ireland [13,15], possibly because the further stable moderate growth within present habitats results in postponement of silvering onset. Whereas, in a region with unstable environments like Italian lagoons [4], a trade-off between somatic growth and the risk of mortality before reproduction would lead females to employ the tactic of shifting to maturation at the earliest possible opportunity.

These types of trade-offs have been suggested in various other diadromous fishes, with salmon having alternative reproductive strategies in landlocked forms [3], or the semelparous osmerid fish ayu has some small females that spawn more than once [21]. Eels can adopt several migration scenarios in whether they migrate upstream or remain in estuarine or even coastal marine habitats [22], and their freshwater habitats can vary among streams, rivers and lakes. Subsequent environmental sex determination [23] after a migratory choice of the growth habitat may facilitate adaptation to the environmental conditions of the growth phase [24].

The relationships between diadromy and the reproductive strategies of eels (many small eggs in the offshore ocean; diverse growth habitats) might be different from other diadromous fishes (much larger eggs in freshwater; relatively uniform growth habitats). Migration costs, sexual selection or environmental variability have been hypothesized to explain an empirical correlation between semelparity and large demersal salmonid egg size (reflects size at maturation) [25,26]. In eels, body size determines the number of eggs and, therefore, the fitness (total fecundity) of the females [27], making their decision to migrate earlier or later important, in addition to size thresholds of energy content [10] for the long migration.

Their plasticity in being able to use such a range of growth habitats that can vary greatly in eel density, food availability/growth potential and predation risk may require a flexible strategy to determine their age and size at reproduction, as indicated here. A variety of species show geographical variations in life-history traits, but the semelparous European eel seems to face an interesting extreme trade-off for their fitness at maturation and the start of their 4000–7000 km migrations to the spawning area in the Sargasso Sea.

Ethics

There are no conflicts with the animal ethics and conservation policy of Biology Letters for the fish species of the IUCN Red List in this study. The present work entirely relies on published data of the EU FP5 project EELREP (Q5RS-CT-2001-01836).

Data accessibility

Data are available from the Dryad Digital Repository: Yokouchi K, Daverat F, Miller M, Fukuda N, Sudo R, Tsukamoto K, Elie P, Poole R. Data from: Growth potential can affect timing of maturity in a long-lived semelparous fish. Dryad Digital Repository. (http://dx.doi.org/10.5061/dryad.42c1t4t) [14].

Authors' contributions

K.Y., F.D., K.T., P.E. and W.R.P. conceived the idea of the study. K.Y. and F.D. designed the study and conducted data analyses. K.Y., M.J.M., N.F. and R.S. contributed to interpretation of data and modelling. K.Y., F.D., M.J.M., N.F. and R.S. drafted the manuscript. All authors contributed to revising the manuscript, agreed to be held accountable for the content therein and approved the final version of the manuscript to be published.

Competing interests

We have no competing interests.

Funding

K.Y. was supported by the FRA and a Research Fellowship for Young Scientists (no. 245842) from the JSPS. K.Y., F.D. and P.E. were supported by the CPER programs of Aquitaine, France.