ICES Journal of Marine Science: Journal du Conseil Advance Access originally published online on November 19, 2007
ICES Journal of Marine Science: Journal du Conseil 2008 65(1):17-24; doi:10.1093/icesjms/fsm166
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Experimental study of the dependence of embryonic development of Trachurus trachurus eggs on temperature
1 Estação Piloto de Piscicultura de Olhão, L-IPIMAR/CRIPSul, Av. 5 de Outubro s/n, 8700-305 Olhão, Portugal
2 L-IPIMAR, Av. de Brasília, 1449-006 Lisboa, Portugal
Correspondence to M. E. Cunha: tel: +351 289715346; fax: +351 289715579; e-mail: micunha{at}ipimar.pt
Cunha, M. E., Vendrell, C., and Gonçalves, P. 2008. Experimental study of the dependence of embryonic development of Trachurus trachurus eggs on temperature. – ICES Journal of Marine Science, 65: 17–24.To determine the effect of temperature on the development rates of artificially fertilized eggs of Trachurus trachurus, experiments were carried out at temperatures ranging from 10.5°C to 19°C. Egg development through to hatching only took place at 11.7–19°C. At lower temperature, eggs did not develop beyond the stage where the outline of the embryo was clearly discernible and a defined median line of the embryonic shield (stage 4 in this study) was apparent. Development time took from 46 h at 19°C to 126 h at 12°C. A generalized linear model of the stage-dependent development time (age) as a function of incubation temperature was developed. The data are also compared with those reported in the literature and related to sea temperature on the spawning grounds.
Keywords: egg development, NE Atlantic, temperature, Trachurus trachurus
Received 14 April 2007; accepted 17 October 2007; advance access publication 19 November 2007.
 |
Introduction |
|---|
 Top Introduction Material and methodsResultsDiscussionReferences |
|---|
Horse mackerel (Trachurus trachurus) are heavily exploited throughout the Northeast (NE) Atlantic. Estimates of spawning-stock biomass (SSB), and detailed knowledge of population age structure and spawning characteristics, are used to produce an appropriate total allowable catch (TAC) in Atlantic–Iberian waters (Priede, 1994).
The daily egg production method (DEPM) (Lasker, 1985) is an appropriate method to estimate the SSB of fish with indeterminate fecundity and pelagic eggs, such as horse mackerel. An important requirement of this approach is the need to know the age-dependent abundance of eggs on the spawning grounds, assuming constant mortality. Accurate staging of the eggs is important because each stage reflects a possible age range, based on the estimated time of day they were spawned and local water temperature (Lo, 1985). Previous models of temperature-dependent egg development rates for T. trachurus were developed by King et al. (1977) for South Africa, and by Pipe and Walker (1987) for the southwest coast of Ireland. Although the work of Pipe and Walker (1987) dealt with a population of T. trachurus from the NE Atlantic, the stock in Atlantic–Iberian waters is considered to be a discrete, more southerly population (Murta, 2003), and it may be differently adapted to the local environment. Spawning of the Atlantic–Iberian stock takes place mainly in waters ranging from 15°C to 17°C (mean 15.7°C), some 3°C warmer than the stock studied by Pipe and Walker (1987) in the Celtic Sea. Moreover, the stages described by Pipe and Walker (1987) do not provide sufficient differentiation to provide accurate age estimates at temperatures <17°C.
Different approaches have been used to model the time taken for fish eggs to develop at various temperatures. Lo (1985) fitted the development time for each stage separately and used non-linear fitting algorithms to estimate parameters, assuming a log-normal error distribution. Chambers and Leggett (1989) proposed the use of event analysis to quantify stage transition in early life history studies. Using the latter approach, Pepin et al. (1997) found that a Weibull error distribution was more appropriate. A multinomial approach could have been appropriate to describe stage-classified populations (Ibaibarriaga et al., 2007). Generalized linear models (GLMs) are recommended by Jiao and Chen (2004) to identify error structures in sequential population analysis.
In this study, horse mackerel eggs from Atlantic–Iberian waters were fertilized in vitro and allowed to develop to hatching under controlled temperature conditions. We present a revised description of easily identifiable embryonic development stages that is intended to provide greater resolution for age determination. We developed a GLM of the age-structured sequential egg population (with multiple stages per observation) as a function of water temperature. GLMs offer a maximum-likelihood-based method that provides a systematic framework by which parameters can be estimated when the model error belongs to the exponential family. One of the advantages of GLMs is that they can readily deal with many types of error structures (Jiao et al., 2004). We paid specific attention to identifying the appropriate error structure of the model, and considered log-normal, gamma, and Poisson distributions in modelling and diagnostic analysis.
|
Material and methods |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
During a survey aboard RV "Capricórnio" in February 2004, and aboard RV "Noruega" in March 2005 and February 2006, hydrated oocytes and semen from several mature horse mackerel were stripped by gentle pressure on their abdomen. Two experiments were performed during the 2004 cruise and one during each of the other two cruises (Table 1). The oocytes and the semen were stirred together with a small amount of seawater, then transferred into 1-l glass beakers filled with seawater. After 30 min, the floating (fertilized) eggs were carefully collected and washed with local filtered seawater, and distributed into several 1-l glass beakers filled with filtered local seawater and incubated in baths at different controlled temperature. Two replicate experiments were performed for each temperature.
|
Incubations were performed at six temperatures, four lower and two higher than ambient. The lower temperatures were maintained by four recirculation reservoirs, each containing titanium evaporators and temperature probes connected to a control thermostat. The higher temperatures were maintained with two commercially available aquarium heaters. The overall system (Figure 1) consists of six 60-l tanks, each containing six 1-l beakers that float when supported by Styrofoam plates. Gentle aeration of filtered seawater in the beakers was accomplished with an aquarium pump.
|
Temperatures were set from 10°C to 20°C (see Table 1 for initial settings). The choice of these temperatures was based on Portuguese oceanic and coastal water temperatures during the horse mackerel spawning season, which range from 13°C to 18°C. The temperature in each tank was slightly different from the initial settings, as indicated in Table 1, and varied within ±0.5°C.
Time 0 was defined as the time when the fertilized eggs were placed in each beaker. From then on, in the first experiment, eggs were sampled every hour during the first 18 h, and subsequently every 2 h until hatching. In the second experiment, hourly sampling was only carried out during the first 12 h. During the third experiment, sampling was random, with eggs collected at intervals that varied between 1 and 12 h. During the fourth experiment, sampling was carried out every 2 h during the first 18 h and every 4 h thereafter. At each egg collection, the time and the temperature of each incubation tank were recorded. Eggs were sampled randomly and immediately fixed in 4% formaldehyde for later staging. Hatched larvae were allowed to grow but not to feed. Eggs were examined under a binocular dissecting microscope, and classified into one of the 11 stages shown in Figure 2, each lasting <24 h.
|
Embryonic development stages were adapted from the sardine egg classification of Gamulin and Hure (1955):
- Stage 1: First segmentation, which, under dim reflected light, is easily visible (Figure 2a). The stage lasts until individual cells are easily distinguishable from each other, and counting them is possible (64 cells). Equivalent to stages IA of Pipe and Walker (1987), and 1 of King et al. (1977).
- Stage 2: Cleavage proceeds until a blastodermal cap is formed, making individual cells indiscernible (Figure 2b). Equivalent to stages IA of Pipe and Walker (1987), and 1 of King et al. (1977).
- Stage 3: Development of the segmentation cavity. First appearance of a germinal ring, with a small peak at one pole (the embryonic shield) (Figure 2c). Equivalent to stages IB of Pipe and Walker (1987), and 1 of King et al. (1977).
- Stage 4: First appearance of an embryonic axis. The outline of the embryo is clearly defined in the median line of the embryonic shield. The embryo develops, but the head and tail are not yet discernible (Figure 2d). Equivalent to stages II of Pipe and Walker (1987), and 1 and 2 of King et al. (1977).
- Stage 5: The cephalic region becomes apparent and an outline of the optic vesicles can be discerned. Germinal ring development proceeds around the yolk (Figure 2e). Equivalent to stages II of Pipe and Walker (1987), and 2 of King et al. (1977).
- Stage 6: This is marked by the closure of the blastopore. The tail end develops and is swollen at its tip. Abdominal somites take shape. The angle formed by the tail and yolk is
90° (Figure 2f). Equivalent to stages II of Pipe and Walker (1987), and 2 and 3 of King et al. (1977).
- Stage 7: The embryo tail begins to separate from the yolk mass. The angle formed by the tail and yolk is <90°. Pupils can be discerned in the eyes. Pigment spots appear in two rows along the dorsal body contour and especially around the oil globule (Figure 2g). Equivalent to stages III of Pipe and Walker (1987), and 3 and 4 of King et al. (1977).
- Stage 8: Growth of the tail still short of three-quarters of the egg circumference (Figure 2h). Equivalent to stages III of Pipe and Walker (1987), and 4 of King et al. (1977).
- Stage 9: Embryo length is three-quarters of egg circumference. Pigment spots develop in the caudal region (Figure 2i). Equivalent to stages III of Pipe and Walker (1987), and 4 of King et al. (1977).
- Stage 10: Embryo length exceeds three-quarters of egg circumference until the tail reaches the head (Figure 2j). Equivalent to stages IV of Pipe and Walker (1987), and 5 of King et al. (1977).
- Stage 11: Tail grows past the embryo head (Figure 2k), and hatching takes place. Equivalent to stages IV of Pipe and Walker (1987), and 5 of King et al. (1977).
- Stage 2: Cleavage proceeds until a blastodermal cap is formed, making individual cells indiscernible (Figure 2b). Equivalent to stages IA of Pipe and Walker (1987), and 1 of King et al. (1977).
Data analysis
The data are available from the senior author on request. Because the initial numbers of eggs in the beakers were unknown, it was not possible to determine either total abundance or mortality. Instead, the abundance of each stage in the samples was converted into ratios in relation to the total number of eggs in that sample, and these were used as weighting factors. Ages were calculated as the time elapsed from when the fertilized eggs were placed in each of the beakers (time 0) to sampling. The unexpected presence of egg stages whose probabilities were <0.05 was considered as zero, because they could represent dead eggs, and experiments where no eggs developed past a certain stage were not considered for the model.
The abundance of eggs at age was the outcome of an experiment involving the random sampling of fixed time intervals followed by observation of the number of discrete events per sampling unit, which often have the characteristics of a Poisson process (Dowdy and Wearden, 1990). We assumed that the appropriate probability model for the number of occurrences of egg stages in the specified time interval and temperature is a Poisson distribution, as described by Dowdy and Wearden (1990).
We started by modelling the dependence of development time on temperature for each stage separately. Because the relationship between the mean and the variance of the age of each egg stage was exponential, we loge-transformed the development time. The model was fitted using ordinary least squares, and tested for equality of intercepts and slopes within stages, using analysis of covariance (Dowdy and Wearden, 1990). Because the differences between slopes were not significant, we assumed a common trend for all stages, and fitted a simpler generalized linear model (GLM; McCullagh and Nelder, 1989; Zuur et al., 2007), with a Poisson distribution and log-link function to describe the exponential relationship between eggs-at-age (with multiple stages per observation) with incubation temperature, using the egg stage as a categorical variable. Age (Age_hoursi) was the response variable, and the temperature of the incubation bath (Tempi) and egg stage (Stagei) were explanatory variables. Ratios of the abundance of eggs of each age at each stage were used as weighting factors.
The Poisson regression model can be summarized as
|
|
|
|
The estimated slope (β) is the rate of ageing with temperature and the intercept (
) the mean development time for all stages at 0°C. Factor is a factorial measure for each stage and represents the mean age at constant temperature.
Other than the Poisson error distribution, we explored two other error structures commonly used in ontogenic development of marine fish: log-normal and gamma. Of all the error assumptions used in a GLM, the one that gives the most homogeneous deviance residuals is considered the most appropriate. To diagnose whether the model errors are homogeneous, we used its residuals (Zuur et al., 2007) plotted against the predicted age of the eggs (given the model and estimates of its parameters), and the plots were inspected visually. A distinct pattern indicated lack of homogeneity. We validated our choice of "best" model using quantile–quantile plots (QQ–plots; Zuur et al., 2007). The residuals from each particular error distribution were plotted against quantile points (the difference between a quantile and percentile point is a factor of 100), to determine how well the model fitted the data. The best fit was considered the one where the resulting points lay closer to a straight line.
The Brodgar (v 2.5.2) software package (© Highland Statistics Ltd, 2000), which contains an interface to the statistics package R (version 1.8.1), was used for the analysis.
|
Results |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
The dependence of development time of each egg stage on temperature illustrates the temporal evolution of egg stages, with the distribution of older stages displaced to the upper right corner of each plot (Figure 3). The first two stages (1 and 2) displayed great variability in age at lower temperature (Figure 3) because of the very long development times. In fact, eggs did not hatch at the coldest temperature of the incubation baths (10°C), which corresponded to mean incubating temperatures of 10.6–11°C (Table 1). The embryos must have died in the early stages of cleavage, because the oldest stage attained by these eggs was Stage 4 (Figure 2d). At 12°C (mean 11.7°C), the second coldest temperature, the eggs were able to hatch and the total development time was less than at 13°C. However, this result was derived from a single experiment performed in February 2006 (Table 1).
|
The temperature dependence of development time is strong, all stages exhibiting an exponential relationship between age and temperature (Table 2). All stages also share a common slope, because the rate of development with temperature was the same within the 11 stages, with no significant differences (F10,614 = 0.195, p > 0.999).
|
Intercepts and slopes obtained with a generalized linear model with Poisson error distribution were highly significant (p > 0.001; Table 3) and the deviance residuals of the model were homogeneous (Figure 4). When log-normal and gamma error distributions were used, the residuals were heterogeneous (Figure 4). The standard deviance of the residuals lay closer to a straight line in the QQ–plots (Figure 5), showing that the assumption of a Poisson distribution of model errors is acceptable, but not ideal. The goodness of fit of the model is also demonstrated by the straight line with slope 1 of the plot of predicted against observed egg development time (Figure 6), although with slightly greater spread for the older ages, reflecting the greater variability at older age.
|
|
|
|
The result of fitting the model is shown in Figure 7. With this model, the time-lag to 50% of hatching (Stage 11) at the lowest and highest incubation temperatures (12°C and 19°C, respectively) are 126 and 46 h. The slope obtained with the GLM (Table 3; –0.145) corresponds to an ageing rate 15% faster for each degree rise in temperature. The average change in development rate for a 10°C change in temperature (Q10) was 4.3.
|
|
Discussion |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
The experimental study presented here made it possible to describe for the Atlantic–Iberian horse mackerel stock the exponential increase in age of eggs with decreasing temperature using egg stage as a categorical variable. The results showed that all stages exhibited the same semi-log relationship between age and temperature, with a common temperature factor of exp(kT), where k = –0.145. It is likely that the development of eggs is governed by a single underlying process rather than by a succession of different processes with differing temperature dependences. The controlling processes are probably parts of the respiratory metabolism, and they prevent temperature from disorganizing development. Temperature-dependence is quite strong and a 10°C rise in temperature leads to an ageing rate 4.3 times faster. It is interesting that the magnitude of the temperature effect observed in this study is similar to that observed for species living in very different temperature conditions, such as cod (Gadus morhua) and haddock (Melanogrammus aeglefinus). Data from Pepin et al. (1997) yielded a Q10 of 3.7 for cod, and from Page and Frank (1989) a value of 4.6 for haddock. This suggests that the process that drives the maturation of eggs evolved with a compensation for temperature. The implication is that differences in embryonic development time in fish species may be caused not by temperature but rather by the size of the embryo (Clarke, 1983).
The use of this information, together with the new staging scheme described here, permits better assignment of age, in days, to field-collected eggs than previously possible. Although earlier descriptions of the egg stages of horse mackerel (King et al., 1977; Pipe and Walker, 1987) do exist, some stages lasted considerably longer than 1 day at the range of temperatures found in Atlantic surface waters of the Iberian Peninsula during the horse mackerel spawning season (13–18°C), and therefore may have limited the accuracy of assignment of age. The longest period between stages using the egg staging methodology suggested here (between stages 6 and 7) is 13 h at 13°C, but this still allowed good age resolution even at low temperature. According to Berenbeim et al. (1973), the optimum temperature for development of Trachurus eggs off the Iberian Peninsula is 
16°C. At that temperature, the total development time for horse mackerel eggs in Atlantic waters of the Iberian Peninsula (as determined here) and in the Celtic Sea (Pipe and Walker, 1987) are similar, 70 h. In contrast, the development time off Namibia at 16°C was much shorter, i.e. 51 h (King et al., 1977). Pipe and Walker (1987) attribute this difference, at least in part, to the fact that King et al. (1977) used eggs from the wild and conducted their experiments on eggs already at the blastodisc stage (stage 2 in the present work, and stage IA of Pipe and Walker, 1987).
In general, horse mackerel egg development off the Atlantic coast of the Iberian Peninsula lasts from 46 h at 19°C to 126 h at 12°C. These results are similar to the values of 50% hatching obtained by Pipe and Walker (1987) for the Celtic Sea. The main difference is the temperature at which eggs developed through to hatching. In Atlantic waters of the Iberian Peninsula, T. trachurus did not hatch at a mean temperature <11.7°C, but in the Celtic Sea (Pipe and Walker, 1987) some eggs hatched at 10.4°C, though with high mortality. These results suggest that off the Iberian Peninsula, spawning to be successful must take place at temperatures >12°C, whereas in the Celtic Sea it can be at slightly lower temperature, but still >11°C, supporting the observations ofLetaconnoux (1951), who suggested that spawning would not commence where surface water temperature was <11°C.
|
Acknowledgements |
|---|
We thank editor Pierre Pepin and two anonymous reviewers for constructive comments on the submitted manuscript, and A. J. C. G. Murta for similar input to an earlier version. We also thank the crew of RVs "Capricórnio" and "Noruega" for their support in conducting the experiments. The work forms part of the QCA-3/MARE/FEDER-NeoMAV project. PG was supported through a grant from Plano Nacional de Recolha de Dados-PNAB/EU DCR-Data Collection Regulation.
|
References |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
-
Berenbeim D. Ya., Overko S. M., Sedletskaya V. A. On regularities in the variability of the optimum surface temperature in the course of development of the eggs and larvae of horse mackerel, Trachurus trachurus Linne. ICES Document CM 1973/J (1973) 8:4.
Chambers R. C., Leggett W. C. Event analysis applied to timing in marine fish ontogeny. Canadian Journal of Fisheries and Aquatic Sciences (1989) 46:1633–1641.
Clarke A. Life in cold water: the physiological ecology of polar marine ectotherms. Oceanography and Marine Biology: An Annual Review (1983) 21:341–453.
Dowdy S., Wearden S. Statistics for Research. (1990) 2nd edn. New York: Wiley-Interscience. xvii+629.
Gamulin T., Hure J. Contribution a la connaissance de l’ecology de la ponte de la sardine (Sardina pilchardus Walb.) dans l’Adriatique. Acta Adriatica (1955) 7:1–23.[Medline]
Ibaibarriaga L., Bernal M., Motos L., Uriarte A., Borchers D. L., Lonergan M. E., Wood S. N. Characterization of stage-classified biological processes using multinomial models: a case study of anchovy (Engraulis encrasicolus) eggs in the Bay of Biscay. Canadian Journal of Fisheries and Aquatic Sciences (2007) 64:539–553.
Jiao Y., Chen Y. An application of generalized linear models in production model and sequential population analysis. Fisheries Research (2004) 70:367–376.[CrossRef][Web of Science]
Jiao Y., Chen Y., Schneider D., Wroblewski J. A simulation study of impacts of error structure on modelling stock–recruitment data using generalized linear models. Canadian Journal of Fisheries and Aquatic Sciences (2004) 61:122–133.
King D. P. F., O’Toole M. J., Robertson A. A. Early development of the South African maasbanker Trachurus trachurus at controlled temperatures. Fisheries Bulletin South Africa (1977) 9:16–22.
Lasker R., ed. An egg production method for estimating spawning biomass of pelagic fish: application to the northern anchovy, Engraulis mordax. NOAA Technical Report, NMFS 36 (1985) 99.
Letaconnoux R. Contribution a l’étude des espèces du genre Trachurus et spécialement du Trachurus trachurus (Linne 1978). Mémoirs office Scientifique et Technique des Pêches Maritimes (1951) 15:67.
Lo N. C. H. A model for temperature-dependent northern anchovy egg development and an automated procedure for the assignment of age to staged eggs. An Egg Production Method for Estimating Spawning Biomass of Pelagic Fish: Application to the Northern Anchovy, Engraulis mordax—Lasker R., ed. (1985) 43–50. NOAA Technical Report, NMFS 36.
McCullagh P., Nelder J. A. Generalized Linear Models. (1989) New York: Chapman and Hall. 511.
Murta A. J. C. G. Estrutura populacional do carapau Trachurus trachurus (L. 1758) na costa Portuguesa. Tese de Doutoramento, Faculdade de Ciências, Universidade de Lisboa (2003) 211.
Page F. H., Frank K. T. Spawning time and egg stage duration in Northwest haddock (Melanogrammus aeglefinus) stocks with emphasis on George and Brown Bank. Canadian Journal of Fisheries and Aquatic Sciences (1989) 46(Suppl. 1):68–81.
Pepin P., Orr D. C., Anderson J. T. Time to hatch and larval size in relation to temperature and egg size in Atlantic cod (Gadus morhua). Canadian Journal of Fisheries and Aquatic Sciences (1997) 54(Suppl. 1):2–10.
Pipe R. K., Walker P. The effect of temperature on development and hatching of scad, Trachurus trachurus L. eggs. Journal of Fish Biology (1987) 31:675–682.[CrossRef][Web of Science]
Priede I. G. Spawning biology, distribution and abundance of mackerel, Scomber scombrus, and horse mackerel, Trachurus trachurus, in the north east Atlantic. Final Report to the Directorate General for Fisheries (DG XIV) of the Commission of the European Communities. Project MA2 436 (1994).
Zuur A. F., Ieno E. N., Smith G. M. Analysing Ecological Data. (2007) New York: Springer. 700.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  A. G. Murta and C. Vendrell Using the EM algorithm to age fish eggs ICES J. Mar. Sci., May 1, 2009; 66(4): 607 - 616. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||