Embryonic Temperature Modulates Muscle Growth Characteristics in Larval and Juvenile Herring

Muscle growth in fish differs from that in mammals in that new fibre production continues throughout much of adult life (Greer-Walker, 1970; Weatherley and Gill, 1985). The first muscle fibres are formed prior to the end of somitogenesis, starting in the most rostral myotomes and proceeding caudally (Van Raamsdonk et al. 1978; Hanneman, 1992; Johnston et al. 1995; Weinberg et al. 1996). Myoblasts at the horizontal septum have been shown to elongate to span the entire somite width to form a small number of mononuclear muscle pioneer fibres per myotome in zebrafish (Brachydanio rerio) (Hanneman, 1992) and Atlantic herring (Clupea harengus) (Johnston et al. 1995). In all fish species studied, the majority of embryonic muscle fibres arise from the fusion of several myoblasts to form multinucleated muscle fibres (Waterman, 1969; Nag and Nursall, 1972; Van Raamsdonk et al. 1978). Embryonic red and white muscle fibre types can be distinguished on the basis of their morphological (Waterman, 1969; Vieira and Johnston, 1992), histochemical (Van Raamsdonk et al. 1978; Batty, 1984) and immunocytochemical characteristics (Mascarello et al. 1995). Before hatching in herring, mitotic activity was observed in myoblasts between the embryonic muscle fibres, which are thought to correspond to presumptive satellite cells (Johnston et al. 1995). On the basis of ultrastructural observations in rainbow trout (Oncorhynchus mykiss) and several cyprinid fish, it was suggested that these myoblasts originate from and proliferate within the adjacent mesenchymal tissue lining and migrate into the muscle via the myosepta (Stoiber and Sänger, 1996). During the juvenile phase, most if not all of the myogenic precursor cells become enclosed within the basal lamina of muscle fibres to form the satellite cell population (Veggetti et al. 1990; Koumans et al. 1991). The satellite cells are thought to be the source of the additional nuclei that are required for the enlargement of existing muscle fibres (hypertrophy) and new fibre production (hyperplasia) (Koumans and Akster, 1995). Post-larval growth is quite plastic, with the relative contribution of hyperplasia and hypertrophy varying during ontogeny and between species and stocks according to their ultimate body size (Weatherley and Gill, 1985). In the European sea bass [Dicentrarchus labrax (L.)], hyperplastic growth was found to be more important at ambient temperatures (16–20 °C) than at 13 °C (Nathanailides et al. 1996), indicating that environmental conditions can also influence the mechanism of muscle growth.

The embryos of Clyde herring develop on the sea bed during March at a mean temperature ranging from 4.8 °C to 10 °C, depending on natural climatic variation (Jones and Jeffs, 1991). The transparent larvae spend an extended period in the plankton before completing metamorphosis to the juvenile stage some time during late summer when the sea temperature has risen to 12–16 °C. In contrast, autumn-spawning herring in the Irish Sea (Manx herring) undergo embryonic development at higher temperatures (10–15 °C), but experience declining temperatures during ontogeny, causing them to over-winter as larvae and complete metamorphosis in the spring. Temperature has a marked effect on the relative timing of embryonic myogenesis in spring-spawning Clyde herring (Johnston, 1993; Johnston et al. 1997). Although the timing of myotube formation was similar at 5 °C to that at 15 °C, the assembly of myofibrils and the innervation of the somites was progressively delayed with respect to somite stage as rearing temperature was reduced (Johnston et al. 1997). These effects persisted through the early part of larval life, such that the length at which the red muscle fibres became multiply innervated was inversely related to temperature, occurring at 12–14 mm at 12 °C, but not until 16–19 mm at 5 °C. Similarly, the expression of embryonic isoforms of myosin light chain 2 and troponin T in white muscle was switched off at progressively shorter body lengths as rearing temperature was increased (Johnston et al. 1997). Temperature has also been shown to influence the number and size distributions of the embryonic muscle fibres in Atlantic salmon (Salmo salar L.) (Stickland et al. 1988; Johnston and McLay, 1997), Atlantic herring (Vieira and Johnston, 1992) and plaice (Pleuronectes platessa L.) (Brooks and Johnston, 1993). In Clyde herring, the number of undifferentiated myoblasts present at hatching was found to be significantly higher at 8 °C than at either 5 °C or 12 °C (Johnston, 1993), and preliminary experiments indicated that this might influence future muscle growth characteristics (Johnston et al. 1996).

The aim of the present study with Clyde and Manx herring (Clupea harengus L.) was to test the hypothesis that temperature during the early life history fixes some component of growth phenotype by modulating the rates of muscle fibre recruitment. In addition, we have extended our earlier observations on the effects of temperature on muscle development in Clyde herring to determine the relative importance of hypertrophic and hyperplastic growth pathways between hatching and the juvenile stage. The results show that thermal environment modulates muscle cellularity with respect to body length in both the larval and juvenile stages after metamorphosis.

Mature Atlantic herring (Clupea harengus L.), 30–35 cm in total length, were caught by trawling during the spawning season. A spring-spawning (Clyde) stock was sampled during March from the Ballantrae Bank in the Firth of Clyde in 1994, 1995 and 1996. An autumn-spawning (Manx) stock caught in the Irish Sea off the Isle of Man during September 1995 was also used. In all cases, the gonads were dissected at sea and transported on ice to the Dunstaffnage Marine Laboratory, Oban, Scotland.

Four different larval rearing experiments were performed. In all cases, the eggs from 6–10 females were fertilised by pooled sperm from five or six males. The eggs were first dissected free from the ovaries and scattered on glass plates (250 mm×600 mm), keeping the eggs from each female on separate plates. Unfertilized eggs become adhesive on contact with sea water. The plates were then immersed in tanks of sea water containing the sperm for a few minutes at either 5–8 °C (spring) or 11–12 °C (autumn). Fertilization success rate was 90–95 %. Unfertilized eggs were removed with a razor blade. The plates were incubated in 1000 l tanks of recirculating sea water at controlled temperatures. A few days after hatching, algal-fed rotifers were introduced into the tanks. After a further 2–3 days, larvae were fed twice daily on live food, brine shrimps (Artemia sp.) supplemented with marine copepods (Acartia sp.). At 25 mm total length (TL), larvae were weaned onto a proprietary dried food (Ewos, Bathgate, UK).

Clyde herring were reared under three thermal regimes in the spring/summer of 1994 and 1996 (Fig. 1). Eggs were incubated at nominal constant temperatures of 5 °C, 8 °C or 12 °C (±0.35 °C range). The larvae were reared in three 1000 l tanks per temperature under natural photoperiod. Following hatching, the water temperature of each group was allowed to rise slowly until it reached the ambient seawater temperature in mid-June. The main growth experiment was carried out in 1994 (Fig. 1A). The use of gradually rising temperatures simulated the seasonal warming that occurs in the natural environment and increased the percentage of larvae surviving to metamorphosis (estimated to be 5–15 %). To avoid confusion, these complex temperature regimes are referred to by their starting temperatures throughout this paper. The temperature in the ‘12 °C regime’ peaked at 17 °C some 120 days after hatching and thereafter declined such that the temperature of all three groups was approximately 12.5 °C after 175 days (Fig. 1A). The experiment was repeated in 1996 (Fig. 1B), although on this occasion the 5 °C group of larvae did not survive beyond 22 mm TL, and therefore only the data for the 8 °C and 12 °C groups are presented.

In 1995, eggs of Clyde herring were reared at constant temperatures of 5 °C and 8 °C (±0.35 °C range), and the larvae transferred to the ambient seawater temperature at first feeding.

The temperature closely tracked the surface water temperature in Dunstaffnage Bay, fluctuating around (mean) values that increased from 10 °C in late April to 13.5 °C in July. Similar experiments on autumn-spawning (Manx) herring were carried out in 1995, using a temperature regime appropriate for that time of the year. Eggs of Manx herring were reared at nominal constant temperatures of 10 °C and 13.5 °C, and the larvae were transferred to the ambient seawater temperature at first feeding. In these experiments, the ambient seawater temperature decreased from 13.5 °C at transfer to approximately 10 °C at the end of the experiment. In both experiments, fish from each group were reared in two tanks per temperature, fed ad libitum and sampled after similar periods at the common temperature in order to provide a similar growth opportunity.

The larvae were anaesthetized in a 1:5000 (v/v) solution of benzocaine in sea water or in a 1:10 000 (m/v) solution of bicarbonate-buffered MS 222 (Sigma Chemicals, Poole, Dorset) in sea water. The total length (TL) of each larva was measured using a binocular microscope. Specimens were fixed for 24 h in Bouin’s fluid, transferred to 70 % ethanol and processed for wax histology. Serial 7 μm sections from 0.4 TL (measured from the tip of the snout to the tail and corresponding to the pelvic fin insertions in fish greater than 25 mm TL) larvae were cut transversely to the longitudinal body axis and mounted on gelatin-coated slides. Sections were dewaxed with xylene, rehydrated and either stained with haematoxylin–eosin or processed for immunocytochemistry. Sections for antibody staining were treated with 3 % (v/v) hydrogen peroxide to block endogenous peroxidase activity, washed with phosphate-buffered saline (PBS), pH 7.4 (at 4 °C), and blocked with normal sheep serum. The sections were incubated overnight at 4 °C with a monoclonal antibody against herring fast muscle myosin light chains and processed for immunoperoxidase staining as previously described (Johnston and Horne, 1994).

The proliferation of myogenic cells was investigated using 5-bromo-2′-deoxyuridine (BrdU), which was incorporated into replicating DNA and subsequently localised with a mouse anti-BrdU monoclonal antibody. The labelling solution (Amersham International, UK) was added to the sea water to give a final concentration of 30 μg ml⁻¹. The larvae were transferred to beakers floating in the holding tanks to maintain the temperature during the labelling period and subsequently killed by an overdose of anaesthetic (MS 222) after 1, 3, 5, 8, 10 or 24 h. BrdU-labelled fish and controls were fixed in Bouin’s fluid and processed for wax histology. The sections were dewaxed, and BrdU incorporation was detected using a commercially available monoclonal antibody (Amersham International, UK) as previously described (Johnston and Horne, 1994). The procedure results in black staining at sites of BrdU incorporation. Sections were counter-stained with 0.5 % (m/v) Nuclear Fast Red to visualise the unlabelled nuclei.

Sections cut transversely to the longitudinal body axis and stained with haematoxylin–eosin were used to determine the number and cross-sectional areas of all the muscle fibres in myotomal cross sections. The outline of each muscle fibre in half the myotomal cross section was traced using a microscope drawing arm (magnification ×40). The resulting drawing was enlarged three times on a Xerox machine, and the cross-sectional areas of the muscle fibres were determined by digital planimetry using a Video-Plan image-analysis system (Kontron Electroniks, Basel). For the transfer experiments with Clyde herring, the nuclei were marked on the drawings, enabling the density of nuclei per unit muscle cross-sectional area to be quantified. The numbers of fibres and nuclei from half-sections were multiplied by 2 on the assumption that the fish were bilaterally symmetrical.

Myosin heavy chains were isolated from larvae reared under the different temperature regimes in 1994. Herring were sampled at hatching and at approximately 15, 20, 25 and 40 mm TL (up to 30 fish were pooled for each sample). Larvae were killed by an overdose of anaesthetic in a 1:2000 (m/v) solution of bicarbonate-buffered MS 222 in sea water. White muscle samples were isolated on a cooled microscope dissection stage. For the smallest larvae, the trunk was skinned as this has been shown to remove the single superficial layer of red muscle effectively (Crockford and Johnston, 1993). In addition, 1–2 g of white (fast) muscle and red (slow) muscle was dissected from the dorsal epaxial myotomes just behind the dorsal fin from four adult Clyde herring, 30–35 cm TL.

Using a chilled hand-held glass homogenizer, tissue samples were homogenized in 20 vols of ice-cold preparation buffer (containing in mmol l⁻¹): Tris/HCl, 10; NaCl, 50; EDTA, 1; pH 7.4; and the following proteolytic enzyme inhibitors, 50 μg ml⁻¹ phenylmethylsulphonyl fluoride, 0.5 μg ml⁻¹ leupeptin, 1 μg ml⁻¹ pepstatin A and 0.2 trypsin inhibitor units ml⁻¹ aprotinin (units as defined by the supplier; Sigma Chemicals, Poole, UK). Washed myofibrils were prepared as previously described (Crockford and Johnston, 1993).

Sodium dodecylsulphate–polyacrylamide gel electro-phoresis (SDS–PAGE) was carried out as described by Laemmli (1970) with the inclusion of 10 mmol l⁻¹ DL-dithiothreitol (DTT) in the sample buffer. Myofibrils were dissolved in 60 mmol l⁻¹ Tris/HCl, pH 6.75, 2 % (m/v) SDS, 10 % (v/v) glycerol, 10 mmol l⁻¹ DTT and 0.002 % (m/v) Bromophenol Blue to give a final protein concentration of 2 mg ml⁻¹. The samples were heated to 80 °C for 3 min and centrifuged at 5000 g for 5 min prior to use.

Myosin heavy chains were purified on a 0.75 mm thick, 8 % (m/v) acrylamide gel and identified by their relative molecular mass using standard proteins of known molecular mass (Sigma Chemicals, Poole, Dorset). Gels were rapidly stained in 0.001 % (m/v) Coomassie Blue G-250, and the myosin heavy chain bands were cut out with a razor blade. The gel segments containing the myosin heavy chain were placed in the sample wells of a 1 mm thick 15 % (m/v) SDS–PAGE gel and digested with 10 μl of protease solution per well, as described previously (Crockford and Johnston, 1993). Digests were carried out using papain (12.5 μg ml⁻¹) from Papaya latex, endoproteinase Glu-C (25 μg ml⁻¹) from Staphylococcus aureus strain V8 and α-chymotrypsin (2.5 μg ml⁻¹) from bovine pancreas (Sigma Chemicals, Poole, Dorset). Gels were fixed for 2 h in 12 % (m/v) trichloroacetic acid, 3 % (m/v) sulphosalicylic acid, then thoroughly rinsed prior to silver staining (Bloom et al. 1987).

The following method was used to compare the peptide maps of myosin heavy chain(s). Photographs of the gels were photographically enlarged so that the closest two bands were 1 mm apart. The gel was then divided into 1 mm elements and scored according to whether a band was absent or present in that element. In order to compare two lanes, the total number of bands in both lanes (b) and the number of bands unique to each lane (u) were measured. An estimate of difference (D) is given by twice the number of unique bands divided by the total number of bands plus the number of unique bands, i.e. D=2u/(b+u). Values of D are zero if the banding patterns of the two lanes are identical and 1 if the banding patterns of the two lanes are completely different from each other.

Data on the growth of herring larvae and juveniles were fitted using a second-order polynomial, and the resulting curve was differentiated (Mathematica Software, Wolfram Research Inc., USA). The hypothesis that there was a single population underlying the fitted curves was tested by multiple regression analyses (Zar, 1984). The transfer experiments were analysed using a two-way analysis of variance (ANOVA) design with embryonic temperature and age as between-subject factors (SPSS, SPSS Inc). Further post-hoc tests analysis were carried out using Tukey’s honest significant difference tests. Profile analysis was used to compare the slopes of the frequency distribution of muscle fibre diameters. Other statistical tests used are detailed in the text.

In the 1994 experiments, the larvae hatched at 9.13±0.56 mm at 5 °C (N=39), 9.48±0.64 mm at 8 °C (N=39) and at 9.03±0.50 mm total length (N=15) (TL) at 12 °C (mean ± S.D.). Feeding began prior to the complete absorption of the yolk-sac, which occurred after approximately 9–13 days at 5 °C and 4–6 days at 12 °C. The relationship between age (days post-hatch) and total length was fitted using a quadratic equation (Fig. 2A), and growth rate was estimated by differentiating the fitted line (Fig. 2B). The initial growth rates of larvae up to 75 days were relatively similar at different temperatures, as previously reported (Johnston et al. 1997).

However, for the whole data set, the hypothesis that the TL/age relationships calculated for each of the three thermal regimes were estimating the same population regression function was rejected (F_(6,359)=26.8; P<0.0001). Growth rate varied with age (post-hatch) and was significantly lower at 5 °C than at either 8 °C or 12 °C (Fig. 2B).

Consistent differences were observed in the morphological development of the larvae with respect to body length at the different thermal regimes. Herring hatch as transparent eel-shaped larvae with a pronounced dorsal and ventral primordial fin-fold (Fig. 3A). The dorsal fin began to differentiate from the primordial fin at 12–13 mm TL (Fig. 3B). Dorsal fin rays were evident with a binocular microscope at 13.5 mm in all the 12 °C larvae (six examined), but not until 16 mm TL at 5 °C (10 larvae at 13.5–15.5 mm TL examined had no dorsal fin rays). Flexion of the notochord and development of the first caudal fin rays were observed at 16.5–18.5 mm TL (Fig. 3C,D), both characters developing somewhat earlier at 12 °C than at 5 °C (Fig. 3G,H). The anal fin started to develop from the ventral primordial fin at 16 mm TL (arrowhead in Fig. 3C) and was fully formed by 22 mm TL (Fig. 3E). For a given body length within this size range, the anal fin was relatively more advanced at 12 °C than at 5 °C (compare Fig. 3I and 3J). Early stages of the pelvic fins were visible in all eight of the 12 °C larvae examined at 20–22 mm TL (Fig. 3J,K), but pelvic fins were only observed in two out of seven fish examined in this size range at 5 °C (Fig. 3I). Metamorphosis from the larval to the juvenile stage involves inter alia the development of gill filaments, the production of red muscle fibres at the horizontal septum, the formation of scales, the appearance of haemoglobin in the blood and the silvering of the body (DeSilva, 1974). Metamorphosis started at approximately 23 mm TL and was completed by 37 mm TL in all the temperature regimes. From the regression equations in Fig. 2A, it was estimated that the fish reached 37 mm TL after 177 days at 5 °C, 117 days at 8 °C and 101 days at 12 °C, at which point they had growth rates of 0.24 mm day⁻¹ at 5 °C, compared with 0.31 mm day⁻¹ at 8 °C and 0.33 mm day⁻¹ at 12 °C (Fig. 2B).

At hatching, the myotomes (at approximately 0.4 mm TL) contained approximately 120 red muscle fibres and 320 white muscle fibres per myotome (Fig. 4A). There were no significant differences in the number of muscle fibres per myotomal cross section at hatching in larvae reared at 5, 8 and 12 °C, contrasting with results from similar experiments on Clyde herring carried out in 1991 (Vieira and Johnston, 1992) and 1993 (Johnston et al. 1995), in which incubation temperature influenced muscle cellularity in embryos, although somewhat differently in each year. The red muscle fibres in yolk-sac larvae constituted a single superficial layer around the lateral surfaces of the trunk (Fig. 4A). At the hatch stage, both red and white muscle fibres stained intensely with a monoclonal antibody to adult herring fast muscle myosin light chain 3 (LC3_f) (Fig. 5A,B), as reported previously (Johnston and Horne, 1994). By approximately 18 mm TL, the red muscle fibres at the horizontal septum no longer stained for anti-myosin LC3_f (Fig. 5C). Expression of fast muscle light chains in red muscle fibres was switched off completely by 22 mm TL at 12 °C, but not until 28 mm TL at 5 °C (not illustrated). Red muscle fibres with the adult pattern of myosin expression (unstained for LC3_f) were then added at the horizontal septum, externally to the layer of embryonic/larval red muscle fibres, starting at the same body lengths, i.e. at shorter larval lengths at 12 °C and 8 °C than at 5 °C (Fig. 4C). There had been a substantial thickening of the red muscle layer at the horizontal septum by 35 mm TL (Figs 4D, 5D).

Myosin heavy chains (MHCs) were electrophoretically purified from adult red and white muscles. Peptide maps of MHCs produced by digestion with papain (Fig. 6A), endoproteinase Glu-C (Fig. 6B) and α-chymotrypsin (Fig. 6C) were significantly different for adult white and red muscles (compare lanes 6 and 7) (estimate of difference, D=0.57), consistent with differences in primary amino acid structure. The embryonic white muscle fibres in 1-day-old larvae contained a distinct isoform of myosin heavy chain(s) from that found in adult white muscle (compare lanes 1 and 6, Fig. 6A–C) (D=0.28). The transition from the embryonic to the adult pattern of myosin heavy chain expression in white muscle occurred between 20 and 25 mm TL. Peptide maps of MHCs prepared from the white muscle of larvae 15 mm TL (lane 2) and 20 mm TL (lane 3) were identical to that of the embryonic MHC isoform(s) (lane 1) (D=0), whereas peptide maps of MHCs from larvae 25 mm TL (lane 4) and 40 mm TL juveniles (lane 5) were identical to the adult pattern (lane 6) (D=0) (Fig. 6A–C). Similar results were obtained in larvae reared at 5, 8 and 12 °C (Fig. 7A–C). The sampling was not sufficiently frequent to determine whether the body length at which the adult MHC(s) was first detected varied between the temperatures, as has been reported for the expression of myosin light chain 2 isoforms (Johnston et al. 1997).

The relationship between the total cross-sectional area of red and white muscles and body length is illustrated in Fig. 8A,B. The cross-sectional area of red muscle increased by approximately 90-fold, whereas the cross-sectional area of white muscle increased some 160-fold between hatching and metamorphosis (at 37 mm TL), with no significant differences between the three thermal regimes (Fig. 8A,B).

The increase in muscle cross-sectional area in larvae from hatching to 12 mm TL was due entirely to the hypertrophy of the embryonic muscle fibres present at hatching (Fig. 9A,B). The recruitment of new white muscle fibres probably started at 14–15 mm TL, and by 15.6 mm TL there was an increase in both the total number of fibres and the number of fibres in the 10–15 μm diameter size class (Fig. 9C). Discrete germinal zones of small-diameter white fibres were present at the dorsal and ventral apices of the myotome in larvae at 15.6–18.4 mm TL (arrowheads in Fig. 4B). Very small diameter white muscle fibres were observed scattered throughout the myotome from 22 to 24 mm TL, consistent with the general activation of satellite cells (Fig. 4C). In order to quantify these early stages of muscle growth further, five larvae per size class were analysed at hatching and at 14, 37 and 72 days post-hatch in the 5 °C group, when TL was 8.5 mm, 11.5 mm, 15.6 mm and 18.4 mm respectively. The total cross-sectional area of white muscle fibres was 0.010±0.00043 mm² at 8.5 mm TL, 0.029±0.0026 mm² at 15.6 mm TL and 0.056±0.015 mm² at 18.4 mm TL (mean ± S.E.M.). A one-way ANOVA revealed highly significant increases in the total cross-sectional area (P<0.01), the number of fibres per myotome (P<0.001) and the mean cross-sectional area of the largest 200 muscle fibres (P<0.001). A Tukey–Kramer multiple-comparisons test showed a significant increase in the number of white muscle fibres between fish at 11.5 mm TL (261±8) and 18.4 mm TL (546±85) (P<0.001) (means ± S.E.M.). In five larvae of 18.4 mm TL sampled 55 days after hatching at 12 °C, the total cross-sectional area of white muscle fibres per myotome was 0.076±0.0082 mm² and the number of white muscle fibres per myotome was 584±27 (means ± S.E.M.), values that are not significantly different from those obtained for these parameters in 5 °C larvae (P>0.05; Mann–Whitney test). At metamorphosis, the largest fibres were 50–55 μm in diameter, whereas the most numerous size class was still the 5–10 μm class, indicating active recruitment of new fibres (Fig. 9E).

In contrast, the recruitment of new red muscle fibres did not start until 22 mm TL (Fig. 10A–E). At 5 °C, the total cross-sectional area of red muscle fibres increased from 0.0013±0.00014 mm² at 8.5 mm TL to 0.0031±0.00060 mm² at 18.4 mm TL (Tukey’s test) (mean ± S.E.M.), whereas the numbers of fibres were identical (120–140 per myotomal cross section) at these stages (Fig. 10A,D). Towards the end of metamorphosis, at 35 mm TL, the most numerous size class of fibres was the 15–20 μm class and the largest fibres were 25–30 μm in diameter (Fig. 10E).

Muscle recruitment was quantified by counting the number of red and white fibres per myotomal cross section at different stages (Fig. 11A,C). The relationship between fibre number and age post-hatch was fitted with a series of quadratic equations. There was significant variation in the body length and the number of myotomal muscle fibres in larvae sampled 50 days post-hatch (Fig. 11A,C). However, for both muscle fibre types, the hypothesis that there was a single population underlying the fitted regressions at the different temperatures was rejected (red muscle: F_(6,69)=9.54; P<0.0001; white muscle: F_(6,69)=12.27; P<0.001). The rate of muscle fibre recruitment (Fig. 11B,D) was estimated by differentiating the relationships in Fig. 11A,C. By the completion of metamorphosis at 37 mm TL there were 460 red muscle fibres per myotome at 5 °C (177 days post-hatch), 523 at 8 °C (117 days post-hatch) and 562 at 12 °C (101 days post-hatch). At this body length, the rate of red muscle recruitment was approximately 5 day⁻¹ at 5 °C and 10 day⁻¹ at both 8 °C and 12 °C (Fig. 11B). In juvenile herring of 50 mm TL, there were 30.4 % more red muscle fibres per myotome at 12 °C (1209) than at 5 °C (927) (Fig. 12A). At 37 mm TL, the number of white muscle fibres per myotome was estimated to be 4708 at 5 °C, 5600 at 8 °C and 5604 at 12 °C (Fig. 12B); with recruitment rates of 66 day⁻¹ at 5 °C and 103 day⁻¹ at 8 °C and 12 °C (Fig. 11D). It was calculated that in 50 mm TL juveniles there were 23.4 % more white muscle fibres per myotomal cross section at 12 °C (12 065) than at 5 °C (9775).

Muscle hypertrophy has often been assessed by measuring the increase in the mean diameter of the fibres (Kryvi and Eide, 1977; Romanello et al. 1987; Weatherley et al. 1980; Kiessling et al. 1991). However, mean fibre diameter is a complex parameter which is increased by hypertrophic growth and decreased by the addition of new fibres, both of which occur simultaneously. In the present study, the increase in diameter of the largest 200 white fibres was measured to obtain an estimate of the rate of hypertrophy on the assumption that this population of fibres corresponded to the embryonic muscle fibres laid down prior to hatching. The mean cross-sectional area of the largest 200 fibres for all the fish was plotted against age and a second-order polynomial was fitted (Fig. 13A). The hypothesis that there was a single population underlying the fitted second-order regressions was rejected (F_(6,66)=7.16; P<0.0001). Differentiating the fitted relationships provided a measure of the rate of hypertrophy, μm² day⁻¹, of the oldest muscle fibres (Fig. 13B). Rates of fibre hypertrophy at 37 mm TL were estimated to be 15.8 μm² day⁻¹ at 5 °C, 21.4 μm² day⁻¹ at 8 °C and 18.8 μm² day⁻¹ at 12 °C (Fig. 13B). At the intermediate temperature regime, the mean diameter of these fibres increased from 8.9±0.5 μm at hatch (N=5 fish) to 20.7±1.4 μm at 23 mm TL (N=4 fish) and to approximately 60 μm in juveniles 50 mm TL (calculated from Fig. 11A), equivalent to a rate of hypertrophy of approximately 40 μm² day⁻¹ (Fig. 13B).

The experiment on the effect of different thermal regimes on growth was repeated in 1996 (Fig. 1B). Since the larvae reared at 5 °C did not survive past 22 mm TL, only the data for the 8 °C and 12 °C regimes are presented (Fig. 14A,B). The relationships between the number of red and white muscle fibres per myotome and total body length were remarkably similar in 1994 and 1996 (Fig. 14A,B).

In sagittal sections, a wide variation in the morphology of muscle nuclei was evident (Fig. 15A–D), ranging from elongated cylinders within the muscle fibres (arrows in Fig. 15C) to more spherical shapes in isolated cells and myotubes (Fig. 15A,B). Multi-nucleated myotubes were common in the white muscle of fish at 16–50 mm TL, indicating the active recruitment of new muscle fibres (Fig. 15A illustrates a fish of 18 mm TL and Fig. 15D illustrates fish of 24 mm TL). Bromo-deoxyuridine labelling experiments revealed significant mitotic activity (Fig. 15E), and daughter nuclei were observed both randomly distributed on the surface of muscle fibres (Fig. 15A) and within myotubes (Fig. 15B,D). The elongated nuclei occur within the sarcolemma of muscle fibres and were probably post-mitotic myofibre nuclei, whereas the majority of the remainder were myoblasts either proliferating or at various stages on the pathway to terminal differentiation. Since there was a continuous spectrum of nuclear morphologies, it was not possible to classify reliably the various proportions at the light microscope level in this species. Commonly, the majority of nuclei within a myotube were found to have divided within a 6–8 h labelling period (Fig. 15A,D). Our results indicate that both hypertrophy and hyperplasia involve a rapidly proliferating population of myogenic precursor cells. Following a 24 h labelling period, 44.9±4.6 % (mean ± S.E.M., N=5 fish) of the total muscle nuclei were labelled with BrdU. Dividing nuclei associated with hypertrophy were observed in the superficial fibres of larvae at 18 mm TL, prior to the onset of red muscle recruitment at 22 mm TL (Fig. 15D).

In order to test the hypothesis that early thermal experience influenced subsequent muscle growth characteristics, herring were reared at different temperatures until first feeding and then transferred to a common temperature regime. The experimental design for the transfer experiment with Clyde herring is illustrated in Fig. 16. Fish were randomly sampled on the day of transfer and after 60 days and 80 days at the common temperature. At the time of transfer to the ambient seawater temperature, the total lengths of the two groups of larvae analysed for muscle cellularity parameters were not significantly different: 12.9±0.2 mm for fish reared at 5 °C (N=8) and 11.9±0.2 mm (N=8) for fish reared at 8 °C (mean ± S.E.M.). Fish total length and muscle cellularity parameters (number of fibres, mean and total cross-sectional area of fibres) were analysed using a two-way ANOVA design with embryonic temperature and age as between-subject factors. The total length of larvae following transfer showed a significant variation with embryonic temperature (F_(1,37)=6.62; P<0.02) and age (F_(2,37)=303.6; P<0.0005), and there was also a significant interaction between embryonic temperature and age (F_(2,37)=32.2; P<0.005). After 80 days at ambient temperature, the total length of the 5 °C group was 29.7±0.7 mm (N=7) and that of the 8 °C group was 32.1±1.2 mm (N=7) (mean ± S.E.M.).

The total cross-sectional area of red muscle fibres varied with embryonic temperature (F_(1,37)=32.16; P<0.0005) and was 89 % higher in the 8 °C than in the 5 °C group 80 days after transfer (Fig. 17A). A two-way ANOVA revealed significant effects of age (F_(2,37)=179.2; P<0.0005) and embryonic temperature (F_(1,37)=32.2; P<0.005) on the mean cross-sectional area of red muscle fibres (Fig. 17B). Although the main effect of temperature on the total number of red muscle fibres was not significant (Fig. 17C), fish exposed to 8 °C as embryos had 63 % more ‘adult red’ muscle fibres at the horizontal septum than the 5 °C group (F_(1,37)=11.51; P<0.002) (Figs 17D, 18), and there was a significant interaction between age and temperature (F_(2,37)=21.5; P<0.0005).

In the 5 °C group, the total cross-sectional area of white muscle fibres per myotome increased 34-fold during the experiment from 0.020±0.0012 mm² at transfer (N=8) to 0.26±0.050 mm² (N=8) after 60 days and 0.68±0.076 mm² (N=7) after 80 days at the ambient temperature (mean ± S.E.M.) (Fig. 19A). After 60 days of common growth opportunity, the total cross-sectional area of white muscle was 52.5 % greater in the 8 °C than in the 5 °C group, and after 80 days the discrepancy had increased to 87.7 %. The two-way ANOVA revealed highly significant effects of age (F_(2,37)=116.6; P<0.0005) and embryonic temperature (F_(1,37)=22.69; P<0.0005) on the total cross-sectional area of white muscle as well as a significant interaction between age and embryonic temperature (F_(2,37)=11.0; P<0.005).

The number of white muscle fibres per myotome at the level of the dorsal fin was approximately 300 at first feeding in both temperature groups (Fig. 19B). White muscle fibre number had increased 8.7-fold in the 5 °C fish and 13.8-fold in the 8 °C fish after 80 days at a common temperature (Fig. 19B). The two-way ANOVA revealed highly significant main effects of embryonic temperature (F_(1,37)=6.08; P<0.02) and age (F_(2,37)=95.0; P<0.0005) on the number of white muscle fibres per myotome, and a significant interaction between temperature and age (F_(2,37)=5.50; P<0.05). A Tukey’s honest significant different test revealed a significant increase in fibre number between the transfer and 60 day samples (P<0.05) and between the 60 day and 80 day samples (P<0.05). The rates of muscle fibre recruitment between 60 and 80 days post-transfer were approximately 49 fibres day⁻¹ in the 5 °C group and 112 fibres day⁻¹ in the 8 °C group (inset in Fig. 19B). There were main effects of embryonic temperature (F_(1,37)=25.3; P<0.0005) and age (F_(2,37)=192.3; P<0.0005) on the mean cross-sectional area of white muscle fibres (Fig. 19C) and a significant interaction between embryonic temperature and age (F_(2,37)=6.68; P<0.005). The mean cross-sectional area of white fibres was approximately 68 μm² at transfer in both groups, increasing after 80 days to 257±8.9 μm² (N=7) in the 5 °C fish and 326.8±15.0 μm² (N=6) in the 8 °C fish (mean ± S.E.M.) (not illustrated). To obtain a measure of the rate of fibre hypertrophy, the mean cross-sectional area of the 200 largest white muscle fibres was measured in each sample (Fig. 19C). Between transfer and 60 days post-transfer, the calculated rate of hypertrophy was approximately threefold higher in 8 °C than in 5 °C embryos (inset in Fig. 19C). The rate of hypertrophy of the embryonic fibres continued to increase with the rising temperature between 60 and 80 days post-transfer, although the difference between the two groups was less pronounced (inset Fig. 19C). A two-way ANOVA revealed significant main effects of age (F_(2,37)=233.9; P<0.0005) and embryonic temperature (F_(1,37)=43.38; P<0.0005) as well as a significant interaction between embryonic temperature and age (F_(2,37)=13.58; P<0.005). The significant interactions between embryonic temperature and age for the number of fibres and the mean cross-sectional area of the largest 200 fibres were interpreted in terms of a general acceleration of muscle growth in the fish reared at 8 °C until first feeding, since both the rates of new fibre recruitment (Fig. 11D) and hypertrophy (Fig. 13B) increased during the course of larval life. The frequency distributions of fibre diameters between fish reared at 5 °C and 8 °C until first feeding were very similar at transfer (Fig. 20A,B) and after 80 days at the common temperature (Fig. 20E,F). The percentage of fibres in the smallest size class (0–5 μm) was 11.0 % for the 5 °C group and 11.8 % for the 8 °C group. White muscle fibres with diameters greater than 35 μm made up 9.3 % of the total in the 5 °C fish and 8.6 % in the 8 °C fish (Fig. 20E,F). In contrast, profile analysis revealed a significant difference in the slopes of fibre diameter distributions between the groups after 60 days at the common temperature (P<0.05) (Fig. 20C,D). Interestingly, there was a significant difference in the regression equations relating the total cross-sectional area of white muscle fibres and the number of fibres per myotome in the two groups (P<0.05), with more fibres per unit cross-sectional area in the larvae reared at 5 °C until first feeding (Fig. 21). This provided evidence for subtle effects of embryonic temperature on muscle cellularity and the relative contributions of fibre recruitment and hypertrophy to the increase in muscle girth.

The density of muscle nuclei was determined at transfer and after 80 days at the common temperature (Table 1). The density of nuclei per square millimetre of muscle fibre cross-sectional area decreased from 10 281 mm⁻² at first feeding to 2746 mm⁻² after 80 days in the fish initially reared at 5 °C. The density of total muscle nuclei was not significantly different in the two groups at first feeding, but was 31.5 % higher in the 8 °C than in the 5 °C group after 80 days (P<0.05; Mann–Whitney U-test (Table 1).

The temperature regimes used for the transfer experiments with Manx herring are illustrated in Fig. 22. Larvae were transferred to the common ambient temperature at first feeding, 7 days after hatching at 13.5 °C (15 days post-fertilization) and 10 days after hatching at 10 °C (25 days post-fertilization). In contrast to experiments with spring-spawning fish (Fig. 15), there was a decrease in temperature during ontogeny for the autumn spawners (Fig. 22). Manx herring were smaller at first feeding (11.72±0.13 mm TL at 10 °C and 10.27±0.14 mm TL at 13.5 °C; N=40 per temperature) (mean ± S.E.M.) than Clyde herring. At 86 days post-transfer, the total cross-sectional area of white muscle was four- to fivefold smaller (Fig. 23A) than in the Clyde herring 80 days post-transfer (Fig. 17A). Rates of muscle recruitment (inset Fig. 23B) and hypertrophy (inset Fig. 23C) in Manx herring were both substantially slower than reported in the earlier experiments with Clyde herring (Fig. 19B,C). A two-way ANOVA revealed significant effects of age (F_(2,35)=151.8; P<0.0005) and embryonic temperature (F_(1,35)=15.0; P<0.005) on the number of white muscle fibres per myotomal cross section in Manx herring, with a significant interaction between age and embryonic temperature (F_(2,35)=3.33; P<0.05). There were also significant effects of age (F_(2,35)=185.6; P<0.0005) and embryonic temperature (F_(1,35)=10.0; P<0.005) on the mean cross-sectional area of the 200 largest white muscle fibres and a significant interaction between age and temperature (F_(2,35)=4.54: P<0.05). Tukey’s post-hoc tests revealed significant age-related effects between first feeding and 40 days post-transfer and between 40 days and 86 days at the common temperature (P<0.05). Rates of muscle fibre recruitment (Fig. 23B) and hypertrophy (Fig. 23C) were generally higher in embryos incubated at 10 °C than at 13.5 °C following transfer to ambient temperature.

Organogenesis in clupeoids begins prior to somite production in the embryo and continues during the larval stage (O’Connell, 1981; Blaxter, 1988; Johnston, 1993). In the present study, we have shown an uncoupling between growth and the body length at which several characteristics critical for swimming develop. For example, the formation of anal fin rays and pelvic fins occurred at shorter body lengths in larvae reared at 12 °C than at 5 °C (Fig. 3). Previously we reported that the adult pattern of myosin LC2 expression was established at 11 mm TL at 12–15 °C, but not until 15 mm TL at 5 °C (Johnston et al. 1997). There were also complex changes in the expression of the thin filament Ca²⁺-regulatory proteins, troponin I and troponin T, during ontogeny. Again the isoforms characteristic of adult muscle were acquired at shorter body lengths as rearing temperature was raised (Johnston et al. 1997). In addition, the adult multi-terminal pattern of innervation to the red muscle fibres developed at 12–14 mm TL at 12 °C, but was delayed until 16–19 mm TL at 5 °C (Johnston et al. 1997).

At hatching, viscous forces are important for swimming in fish larvae at all but the highest speeds, but as body length increases and the fins are formed between 12 mm and 24 mm TL the larvae move into a hydrodynamic regime where reactive forces predominate (Weihs, 1980). This change in hydrodynamic regime is accompanied by a transition from an anguilliform to a subcarangiform style of swimming (Batty, 1984) and an increasingly complex repertoire of behaviour as the sense organs develop (Blaxter, 1988). During ontogeny, there is a gradual change in the phenotype of red and white myotomal muscle fibre types associated with a reduction in mass-specific aerobic metabolism and a decrease in the contraction speed of the muscle fibres due to scaling effects (Batty, 1984; El-Fiky et al. 1987; Scapolo et al. 1988; Goolish, 1991; Johnston et al. 1996). The composition of myosin heavy chains and light chains is a major determinant of muscle contraction speed (Schiaffino and Reggiani, 1996). In the present study, we have shown a change from the expression of embryonic to adult isoform(s) of myosin heavy chain in white muscle at 20–25 mm TL (Fig. 6). The embryonic red muscle fibres present in yolk-sac larvae express fast myosin light chain isoforms characteristic of white muscle in adult Atlantic herring (Fig. 5), as was previously reported for the barbel (Barbus barbus) (Focant et al. 1992). In the present study, we found that fibres towards the dorsal and ventral margins of the superficial red muscle layer continued to stain for LC3_f until 22 mm TL at 12 °C, whereas expression of the fast light chain isoform was not switched off until 24–28 mm TL at 5 °C.

Thus, in this and previous studies, we have shown that different rearing temperatures produce phenotypic variation in a large number of characteristics critical to swimming over the length range 12–24 mm. It is highly likely that such variations in phenotype will affect swimming style and/or performance and quite possibly the survival of larvae under natural conditions, an idea that now needs to be tested by experiment. Variations in the development of the digestive tract and other characteristics in different environments have also been reported in other species of marine fish larvae, suggesting this may be a general phenomenon (Sieg, 1992; Hunt von Herbing et al. 1996).

Three distinct phases of muscle formation can be distinguished in herring. The first phase leads to the formation of the embryonic muscle fibres. Experiments in different years indicate considerable variation in the effects of temperature during development on the number and diameters of embryonic muscle fibres in Clyde herring (present study, Vieira and Johnston, 1992; Johnston, 1993; Johnston et al. 1995), which may be related to variations in egg quality including the amount of yolk, the amino acid content and the concentration of maternal growth factors. Post-embryonic growth involves a population(s) of undifferentiated myoblasts which are thought to arise prior to hatching (Johnston et al. 1995; Stoiber and Sänger, 1996). The initial two- to threefold increase in the cross-sectional area of muscle largely involves the hypertrophy of the embryonic muscle fibres formed prior to hatching (Figs 4, 9A, 10B). The second phase of myogenesis involves distinct germinal zones of myoblasts at the dorsal and ventral extremities of the myotome (Fig. 4B), as described in sea bass (Veggetti et al. 1990), plaice (Brooks and Johnston, 1993) and sea bream (Rowlerson et al. 1995). In herring, these proliferative zones became exhausted at 22–25 mm TL.

The third phase of myogenesis involves the general activation of myoblasts scattered throughout the myotome (Fig. 4C). Muscle fibre number began to increase at greater body lengths in red (22–28 mm TL) than in white (15 mm TL) muscle fibres (Fig. 10A,B), suggesting that at least some components of the signalling pathways involved are local to the muscle. Rates of red and white muscle fibre recruitment increased progressively throughout the second half of larval life and in juveniles, reaching approximately 15 fibres day⁻¹ at 50 mm TL and 150 fibres day⁻¹ at the intermediate temperature regime (Fig. 9B,D). In contrast, Rowlerson et al. (1995) found that there were distinct phases of hyperplastic growth during ontogeny in the sea bream (Sparus auratus). White muscle fibres with diameters less than 5 μm accounted for 2 % of the total at hatch, 33 % after 20 days, dropping to less than 1 % at 60 days, before increasing again by 90–150 days after hatching in juvenile fish.

Muscle is a post-mitotic tissue, and therefore fibre recruitment and hypertrophy require a source of additional nuclei. In mammals, in which post-embryonic muscle growth is entirely by fibre hypertrophy, the source of these additional nuclei is the satellite cells. Satellite cells were first described by Mauro (1961) as small spindle-shaped cells with a heterochromatic nucleus enclosed within the basal lamina of muscle fibres. Satellite cells have also been observed in the juvenile and adult stages of various fish species (Nag and Nursall, 1972; Kryvi and Eide, 1977; Egginton and Johnston, 1982; Veggetti et al. 1990; Koumans et al. 1991). In contrast, the larval stages of most species examined including sea bass (Veggetti et al. 1990), Atlantic herring (Johnston, 1993) and various cyprinid species (Stoiber and Sänger, 1996) possess free undifferentiated myoblasts dispersed between the myotomal muscle fibres. In sea bass, true satellite cells enclosed within the basal lamina of muscle fibres were first observed at 40 mm TL (Veggetti et al. 1990). In the present study, we found a wide range of nuclear morphology in the muscle tissue of larval and juvenile herring. Myofibre nuclei were generally spindle-shaped (arrows in Fig. 15C) whereas myoblast nuclei were more spherical (arrowheads in Fig. 15C). Myotubes were commonly observed on the surface of muscle fibres in fish greater than 18 mm TL, often containing daughter nuclei that had recently divided (Fig. 15A,B,D).

Studies on the in vivo behaviour of myoblasts in growing rats suggest that the satellite cells can be subdivided into two categories. Approximately 80 % of the satellite cells divided with a cell cycle time of 32 h, whilst the remaining 20 % divided more slowly (termed reserve cells) and were not fully labelled after 9 days of infusion with BrdU (Schultz, 1996). It has been suggested that these reserve cells correspond to the stem cell population that generates the myo-nuclei-producing cells (producer cells) by an asymmetric division (Schultz, 1996). A series of in vitro cloning and subcloning experiments combined with computer modelling also indicated that myogenic precursors in chicken could undergo either asymmetric or symmetric divisions (Quinn et al. 1988). Asymmetric divisions resulted in one stem cell and one cell committed to terminal differentiation. Each committed cell underwent further symmetrical divisions to produce up to 32 terminally differentiated muscle cells (Quinn et al. 1988). In vivo experiments in the rat, involving the tandem continuous infusion of BrdU and [³H]thymidine, showed that the producer cell population underwent a limited number of divisions prior to fusion and terminal differentiation. There is also evidence from tissue culture experiments in mammals (Schultz and Lipton, 1982) and fish (Koumans et al. 1993) that is consistent with the idea that satellite cells are not a homogeneous population. Satellite cells expressing the myogenic regulatory factor myogenin apparently follow a programme of limited proliferation, and it has been suggested that these cells represent a subset nearing terminal differentiation (Yablonka-Reuveni and Rivera, 1994).

Koumans and co-workers have investigated the behaviour of fish satellite cells in tissue culture (Koumans et al. 1993; Koumans and Akster, 1995). Only 10 % of myogenic cells isolated from juvenile carp 4–5 cm in standard length (SL) were in a proliferating phase, with the remainder staining for desmin, one of the first expressed muscle-specific proteins. Koumans et al. (1993) suggested that hyperplastic muscle growth is largely dependent on a population of post-mitotic cells formed at an earlier stage of differentiation. Our in vivo experiments of myogenic cell proliferation do not support this hypothesis. In many cases, all the nuclei in newly formed myotubes were found to be BrdU-positive within 6–8 h of labelling, indicating they had undergone recent divisions prior to fusion (Fig. 15A,B,D). Other BrdU-positive cells were associated with the surface of muscle fibres and were presumably involved in hypertrophic growth (Fig. 15D,E).

Although the total cross-sectional area of muscle was similar for a given body length at the different temperature regimes, there were significant differences in the numbers of red and white fibres per myotome (Fig. 12A,B). This indicates that the relative contribution of hypertrophy and hyperplasia to growth varies with temperature, as reported for the sea bass (Nathanailides et al. 1996). Either there are separate populations of myogenic precursors responsible for hyperplastic and hypertrophic growth in larvae and juveniles or the cellular environment regulates the fate of cells committed to terminal differentiation. The former is quite possible since, although embryonic red and white muscles in mammals are derived from distinct myogenic lineages (Stockdale, 1992), the progeny of retrovirus-labelled myoblasts can fuse with any fibre type, indicating that extrinsic signals override the intrinsic programme of gene expression (Hughes and Blau, 1992).

Early thermal experience is known to produce irreversible changes in phenotype in fish and other ectotherms (Johnston et al. 1996). For example, in some fish species including the Atlantic silverside (Menidia menidia) (Lagomarsino and Conover, 1993) and the pejerrey (Odontesthes bonariensis) (Strussmann et al. 1996), environmental temperature has a pronounced effect on sex ratios. The temperature regime during development has also been shown to influence pigmentation patterns (Schmidt, 1919), meristic characteristics (Hempel and Blaxter, 1961) and ultimate body size (Atkinson, 1996).

In embryos and yolk-sac larvae of Clyde herring reared at 8 °C, the total cross-sectional area of myotomal muscle was 88 % higher after 80 days at ambient temperature than in fish initially reared at 5 °C (Fig. 14A), and both fibre number and mean diameter were greater. This probably reflects a faster rate of overall growth since the body lengths of larvae were greater in the 8 °C (34.0 mm TL) than in the 5 °C (29.7 mm TL) groups (P<0.05). Koumans et al. (1991) found that the ratio of myofibre to myosatellite cell nuclei determined by electron microscopy remained relatively constant in common carp (Cyprinus carpio) between 5 cm and 20 cm SL. In a previous study, the number of myoblasts (presumptive satellite cells) per mm² was found to be almost twice as high at hatching in embryos incubated at 5 °C as in those incubated at 8 °C (Johnston, 1993). Presumably, myoblasts identified using ultrastructural criteria include both muscle stem cells and proliferating (producer) cells committed to terminal differentiation. Although presumptive satellite cells were not quantified in the present study, the total numbers of muscle nuclei in fibres and myoblasts were significantly higher in the 8 °C than in the 5 °C fish 80 days after transfer to the common temperature (Table 1).

Embryonic temperature was also found to influence muscle growth in autumn-spawning Manx herring, although in this case absolute growth rates were slower and the relative growth performance of fish was greater in individuals initially reared at 13.5 °C than at 10 °C. Indirect evidence that embryonic temperature modulates muscle growth characteristics up to 3 weeks after first feeding was obtained in the Atlantic salmon (Nathanailides et al. 1995). They found that salmon reared at 11 °C had fewer larger-diameter fibres than fish incubated at cooler ambient temperatures but that, because of a faster rate of hypertrophy in the cooler temperature group after hatching, these differences in fibre size had disappeared 3 weeks after first feeding. It was suggested that the faster rate of hypertrophy at the cooler temperatures was related to a higher density of muscle nuclei, although the interpretation of these experiments is complicated by the continuing differences in the temperature regime experienced after hatching.

In conclusion, we have shown that early thermal experience modulates rates of muscle fibre recruitment and hypertrophy throughout the larval stages and that such effects may be distinct from the effects of temperature on the embryonic muscle fibres. The mechanism is unknown, but our working hypothesis is that it involves alterations in the numbers of muscle stem cells and/or in the numbers and properties of the numerous cell types involved in growth regulation. Proliferation and differentiation of the myogenic precursor cells are mutually exclusive events regulated by a balance of opposing cellular signals and multiple and highly redundant control systems. The MyoD family of basic helix–loop–helix (HLH) DNA-binding proteins or myogenic regulatory factors (MRFs) plays a pivotal role in inhibiting proliferation and promoting terminal differentiation (reviewed in Olson, 1992; Rudnicki and Jaenisch, 1995). MRFs form heterodimers with other widely expressed HLH proteins and bind to the E-box sequence (CANNTG) found in many muscle-specific promoters and enhancers (Mezzogiorno et al. 1993; Olson, 1992; Rudnicki and Jaenisch, 1995). The activity of MRFs is negatively regulated by peptide growth factors, including fibroblast growth factor (Clegg et al. 1987; Hannon et al. 1996), transforming growth factor type-β (Olson et al. 1986), as well as activated oncogenes, and the HLH protein Id, which lacks a basic region and forms inactive heterodimers with other HLH proteins (Olson, 1992; Rudnicki and Jaenisch, 1995). A new member of the transforming growth factor-β (TGF-β) family, recently described in mouse (GDF-8 or myostatin), is a negative modulator of muscle growth (McPherron et al. 1997). In animals in which the GDF-8 gene was disrupted by gene targetting, both body size and the mass of individual muscles were several-fold greater than in wild-type animals. Insulin-like growth factor 1 (IGF-1), in contrast, is thought to have a role in promoting differentiation (Allen and Boxhorn, 1989). Early thermal experience could therefore potentially modulate one or more of the numerous genes and regulatory circuits determining muscle fibre number and muscle mass.

It is worth emphasising that such effects of early thermal experience are relatively subtle compared with the large variations in growth rate observed between different genetic strains and with different stocking densities, ration levels, diet compositions, temperatures and photoperiodic regimes (Weatherley and Gill, 1987). Greer-Walker et al. (1972) found that the number of white muscle fibres per myotome in North Sea herring showed consistent variation between different stocks. Although this may be due to genetic variation, the results of the present study indicate that temperature variation during early development could provide an alternative or additional explanation.

Recruitment is a highly variable and poorly understood process involving abiotic and biotic factors, including variations in the abundance and patchiness of prey and predator species (Gotceitas et al. 1996; Hinckley et al. 1996). Using fuzzy-logic mathematical techniques and time-series data, inter-annual variations in year class strength in sandfish (Arctoscopus japonicus) have been correlated with temperature variation (Sakuramoto et al. 1997). Other studies have found correlations between temperature and egg and larval abundance (Laprise and Pepin, 1995). The survival of planktonic fish larvae is thought to vary spatially with mesoscale and larger-scale ocean circulation patterns, which can be modelled with a combination of hydrodynamic and biological individual-based models of early life stages (Hinckley et al. 1996). Evidence that early thermal experience affects larval phenotype extends beyond the locomotory system and has been obtained in phylogenetically diverse species. For example, rearing temperature has been shown to influence the relative timing at which the jaws become functional in relation to yolk depletion in various tropical species (Fukuhara, 1990), the appearance of cephalic spines in turbot (Scophthalmus maximus) (Gibson and Johnston, 1995) and variation in the morphological complexity of the gastrointestinal system in cod (Gadus morhua) larvae (Hunt von Herbing et al. 1996). Taken together, these studies indicate the importance of considering the direct effect of development temperature on phenotype for incorporation in any individual-based models of larval mortality.

This work was supported by a thematic grant from the Natural Environment Research Council awarded under the Inter-Universities Marine Research Initiative 1990–1997. We are grateful to Dr Robert Batty and Mr Simon Morely for assistance with the larval rearing and to the hospitality of the Director and staff of the Dunstaffnage Marine Laboratory. We thank James Wakeling for writing the curve-fitting programme, Dr M. McKracken from the Institute of Mathematics for statistical advice and Professor John Blaxter, who kindly commented on a draft of the manuscript. The project benefited greatly from the histological expertise of Mr Ron Stuart.