Lunar cycles of coho salmon, Oncorhynchus kisutch: I. Growth and feeding

Biological rhythmicity, the periodicity of which is correlated with major events in the fluctuating physical (abiotic) environment, is a widespread phenomenon in the living world. The general properties of the observed rhythms in a variety of life forms are remarkably similar, suggesting that they are fundamental and ancient phenomena (see reviews by Brown, 1973; Saunders, 1977) which permit organisms to make continuous and frequent adjustments in advance of changing environmental or social conditions. They also allow for the coordination of behavioural, physiological and reproductive activities within a species and between the species of an integrated ecosystem.

Circadian (Leatherland, McKeown & John, 1974; de Vlaming, Sage & Tiegs, 1975; Leatherland & Nuti, 1982) and annual (Shul’man, 1974; Saunders & Henderson, 1970; de Vlaming, 1972) rhythms have been well documented in teleost fishes. Also, in teleosts there are several documented cases of a correlation between lunar cycles in association with annual rhythms and the timing and synchronization of physiological and metabolic events (Walker, 1949, cited by Grau et al. 1981; Mason, 1975; Grau et al. 1981).

Observations in our laboratory and anecdotal reports from several local hatcheries suggested a rhythmic pattern of food intake in salmonid fish; the fish showed marked cyclic variation in their feeding activity with apparent peaks and troughs in food intake. Since growth patterns are closely correlated with food intake, it is likely that alterations in food intake in the short term are reflected in short-term changes in growth. Such growth patterns would not necessarily be identified in the conventional long-term growth rate studies, particularly those involved in growth modelling (e.g. Corey, Leith & English, 1983), but have been found in brown trout, Salmo trutta (Brown, 1946), and rainbow trout, Salmo gairdneri (Wagner & McKeown, 1985; Farbridge, 1985). This study examines the short-term cyclical growth patterns in coho salmon, Oncorhynchus kisutch, during the parr and smolt stages of development, and evaluates the possible interrelationship between the cycles of growth, the cycles of food intake and lunar cycles.

The coho salmon, Oncorhynchus kisutch, used in the study were hatched and raised in the laboratory. Gametes were obtained in 1982, from salmon at the Platte River Hatchery, Michigan, and in 1983 from salmon taken from the Credit River, Streetsville, Ontario. The eggs of 15-20 females were fertilized with sperm from 2–3 males. After water hardening on site for approximately 2 h, the eggs were returned to the laboratory where they were maintained in horizontal hatchery trays in constantly running well water at 10 ± 1 °C.

In all experiments the fish were maintained in constantly running and aerated well water (the seasonal water temperature range was 9–11 °C) under a 12h:12h L:D artificial photoperiod (L = 08.00 h to 20.00 h) unless otherwise indicated. They were fed with either a commercial salmonid diet (Martin Feed Mill, Elmira, Ontario) or a diet formulated by Dr J. Hilton, Department of Nutritional Science, University of Guelph. The composition of the two diets was as follows: Martin’s diet – herring meal 35 %, wheat middlings 35 %, soybean meal 20%, soybean oil 6%, linseed oil 1 %, vitamin mix 2 %, mineral mix 1 % ; Hilton’s diet – fish meal 40 %, wheat middlings 13%, soybean meal 10%, gelatin 10%, poultry by-product meal 7%, shrimp meal 5 %, brewer’s yeast 5 %, fish oil 7 %, vitamin mix 2 %, mineral mix 1 %.

Fish body masses (± 0·05 g) were measured by placing groups of fish (the number depended on the size of the animals) into a tared water-filled container.

To examine possible relationships between growth in mass (and/or length) and feeding of coho salmon parr, 450·6-month-old fish [means 12·9 ± 1·27 (S.D.) g, 9·78 ± 1·64 cm] were randomly distributed between nine aquaria (801, 50fish/ aquaria) ; water temperature throughout the experiment was 11·5 ± 1·5 °C. Fish were fed to satiety three times a day with Hilton’s diet; the feeding order of tanks was randomized daily and records of food consumption were kept. Food consumption was expressed as a fraction of the mean fish mass at the beginning of each 4-day interval (see below).

The aquaria were divided into three study groups, each consisting of three replicates (three aquaria/group). The study groups were weighed in sequence at 4-day intervals (i.e. day 0 – group 1, day 4 – group 2, day 8 – group 3, day 12 – group 1, etc.), in lieu of the morning feeding, from June 23 to September 19, 1983. Thus, masses were obtained at 4-day intervals but fish in any one group were only weighed at 12-day intervals; this design was chosen to eliminate the possibility of stress-related effects on growth rate. The fish in any group were weighed in groups of 10. The standard lengths (± 0·05 cm) of 30 fish from each group (10 from each replicate) were measured on the appropriate mornings.

To determine whether sampling from a common stock would reveal the same growth pattern as that seen in the experimental design above, approximately 1100 coho salmon parr were distributed between two circular aquaria (5001); water temperature throughout the experiment was 11·5 ± 1·5°C. The fish were fed to satiety three times a day with Hilton’s diet.

Fifty fish from each aquarium were weighed in groups of 10, at 4-day intervals in lieu of the morning feeding from June 23 to July 1, 1983. Thereafter 100 fish from each tank were weighed at 4-day intervals until August 22, 1983.

Two hundred and forty 1-year-old coho salmon smolts (mean mass 80·8 ± 8·76 g) were distributed randomly between 12 compartments of four circular aquaria (5001; 20fish/compartment); water temperature throughout the experiment was 10·0 ± 1·5°C. Differences in total mass between compartments were adjusted to 10·0 g or less at the beginning of the study. The fish were fed to satiety three times a day with the commercial diet.

The compartments were allocated randomly to three study groups, each consisting of three replicates. In addition, three ‘reference’ groups were maintained; these were weighed only at the beginning and the end of the study in order to evaluate whether the weighing procedure itself affected overall growth. The study groups were weighed in sequence at 4-day intervals, in lieu of the morning feeding (see experiment 1), from March8 to May 15, 1984. All groups were weighed on May 19, 1984.

In experiment 3, the total mass of fish in each aquarium at all times of sampling was expressed as a fraction of the total mass of the fish in the aquarium at the end of the study. These were termed ‘corrected’ masses in that they were corrected for any differences in the mass of fish in a given aquarium at the end of the study. These masses represented the fraction of the final mass achieved by the time of sampling and were used to calculate a ‘corrected’ growth rate. ‘Corrected’ growth rates could not be calculated for experiment 1 because fish were removed periodically from aquaria at the end of the experiment to analyse various tissue constituents.

In experiments 1 and 3, replicates were randomly allocated a subgroup designation; aquaria with the same subgroup designation were compared to obtain a growth rate for a given interval. Growth rates for each interval were expressed as a mean ± S.D. In experiment 2, growth rates were calculated for each aquarium and expressed as mean ± S.D.

In order to test for significant differences between apparent peaks and troughs and to test for the presence of a semi-lunar periodicity, growth rates (means ± S.E.M.) for mass were plotted on a common semi-lunar scale (Astronomical Almanac 1983, 1984, 1985) and means were compared using one-way analysis of variance. Where significance is indicated (P< 0·05), differences between peaks and troughs in growth were compared with the least significant means procedure (Steele & Torrie, 1960); the critical level of significance for testing hypotheses was P ⩽ 0·05.

Between June 27 and September 18, 1983, the coho salmon parr showed cyclical patterns of growth in mass (Fig. 1A). When a sine curve with semi-lunar periodicity was superimposed over the data for growth in mass it was observed that the growth rate pattern was in phase with the sine curve from June 27 to July 25, then fell out of phase with the semi-lunar sine curve after July 25 but was in phase from August 26 to September 19.

When the data were plotted on a semi-lunar scale, growth was lowest at the time of new and full moons and increased thereafter (Fig. 2).

The cyclical pattern of growth in relative length was not as pronounced as that for growth in mass (Fig. IB). From June27 to July25 the apparent cycle of growth in length was out of phase with that of growth in mass. Any obvious cycling was lost after July 25 when growth in mass fell out of phase with a semi-lunar sine curve, but appeared to be re-established after August 26, although it was then in phase with that of growth in mass.

The observed peaks in the amount of food consumed/mean fish mass occurred approximately 2–4 days after the peak for growth rate (Fig. 1C). Again, the pattern was less distinct for a period after July 25 but became clearer towards the end of the study.

There was no consistent pattern of cycling of growth rates in coho salmon parr subsampled from a common stock (Fig. 3).

The coho salmon smolts showed a cyclical pattern of growth similar to that observed in the parrs, whether or not masses were corrected for the final mass achieved by each aquarium at the end of the study (Fig. 4). The largest ‘corrected’ growth rate was recorded on March 8, 1984, 12 days before the spring equinox (March 20, 1984). The ‘corrected’ growth rate pattern after the spring equinox, although not in complete synchrony, kept in phase with that of a semi-lunar sine curve.

A cyclical pattern was also evident with ‘uncorrected’ growth rates (Fig. 4B) although it was slightly different from that seen with ‘corrected’ growth rates.

When ‘corrected’ and ‘uncorrected’ growth rates after the spring equinox were plotted on a semi-lunar scale both showed a similar cyclical pattern (Fig. 5). Although the curves of ‘corrected’ and ‘uncorrected’ growth rates appeared to be slightly out of phase, lower growth rates tended to be associated with the times of new and full moons or slightly thereafter.

The mass gain of fish in the ‘reference’ aquarium was not significantly different from that of fish in the study aquaria.

Both parr and smolt stages of coho salmon displayed rhythmical patterns of growth, with the fish alternating between periods (approximately 1 week in length) of rapid and slow growth in mass.

Growth rhythms have also been reported in brown trout (Brown, 1946) and rainbow trout (Wagner & McKeown, 1985; Farbridge, 1985). Other evidence for short-term growth rhythms in teleosts is derived from daily, bimonthly and monthly patterns of otolith growth (Pannella, 1971, 1974; Campana, 1984), which are characterized by a period of increasing daily increment width followed by a period of decreasing daily increment width (Campana, 1984), a pattern identical to that shown in this study.

Growth in length also showed evidence of a similar rhythmical pattern as growth in mass in coho salmon parr, although the pattern was not as distinct. At the beginning of the study the periods of rapid growth in length corresponded with the periods of slow growth in mass; this relationship was lost after July 25. A similar relationship between growth in mass and length was observed in brown trout (Brown, 1946). Swift (1955), while studying annual changes in growth rate in brown trout, noted that monthly averages of growth rates for mass and length always varied in the same way but that the cycle for mass preceded that for length. Wagner & McKeown (1985) suggest that the periods of reduced growth in rainbow trout might be analogous to the anabolic refractory periods seen in rats when tissues are unresponsive to growth hormone. However, if this relationship between growth in length and mass does exist it would suggest that the phenomenon is more complicated than a simple reduction in growth. Instead, there may be a switch in processing resources. One possible explanation is that the fish are partitioning the processes involved with growth in mass from those involved with growth in length. Presumably, the fish enter a period of ‘assimilation’ followed by a period of ‘lengthening’ during which some of the body mass laid down during the ‘assimilation’ period is converted into an increase in length. Indeed, brown trout with a condition factor (ratio of mass:length³) of less than 1·08 grew relatively faster in mass than length, while fish with a condition factor of more than 1·08 grew relatively faster in length (Brown, 1957). The biological value of partitioning the two processes is not clear. It may be that the two events are too metabolically demanding to occur simultaneously or that the controlling factors (endocrine system, etc.) would be in conflict if the two processes were to occur simultaneously. Higgs et al. (1977) observed that the administration of bovine growth hormone and 17-α methyltestosterone resulted in greater growth in length than in mass in coho salmon, while the administration of L-thyroxine (T4) had the opposite effect.

Negative growth rates for both mass and length were recorded on a few occasions, although their significance is unknown since they could be a result of the experimental design.

Food consumption showed a similar rhythmical pattern to growth in mass. Brown (1946) also reported biweekly fluctuations in appetite of brown trout. The periods for increased feeding activity were observed to occur slightly after those for growth in mass; the apparent independence of food intake and growth when food is not restricted suggests that both food intake and body mass gain, or growth, are regulated. Indeed, Brown (1946) observed an increase in the efficiency of food utilization before periods of maximum growth in mass and food consumption.

Fish are stressed by handling and this can result in reduced growth (Schreck, 1982). The stress of the weighing protocol was minimized by rapid handling but could not be completely eliminated, particularly since short intervals between weighings were desired. Consequently, the experimental design attempted to balance these considerations. In studies of growth cycles of rainbow trout, several aquaria were sampled at 4-day intervals (Farbridge, 1985). Because this was likely to be stressful, the studies were kept short. Such an experimental design, whilst appropriate for the rainbow trout, is less useful in a species such as coho salmon which is more susceptible to handling stress. However, studies of a longer time series were necessary, especially if possible Zeitgeber were to be considered.

Consequently, the masses of fish in different groups on different dates were compared to obtain a growth rate for a given interval. Thus, the time between weighings for any particular group was increased without having to increase the growth rate interval. In doing this, the assumption was made that when aquaria are randomly allocated an equal biomass of fish, they also possess the same growth potential over time. In other words, at any given point in the time series, all groups weigh approximately the same.

A possible criticism of this experimental design is that this assumption cannot be made and that the masses of different aquaria cannot be compared to calculate a growth rate for a given interval. Any apparent rhythm may simply reflect differences in masses between groups. However, several points argue against that assertion. First, the rhythmic patterns in experiments 1 and 3 followed over several cycles are not easily explained by the above argument given the random block design of the experiments. Moreover, the experimental design has been used on three separate occasions, two presented here and one presented elsewhere (Farbridge, 1985) with the same results. Furthermore, if certain groups had exhibited consistently higher growth rates throughout the study period, one might expect to find significant differences in mass between the groups at the end of the study. This was not found in any of the studies. Moreover, expression of the masses as a fraction of the final body mass achieved by an individual group supports this conclusion, since, if cycling patterns were a result of different growth rates in different groups, ‘corrected’ growth rates calculated on the basis of final mass should eliminate the cycling patterns; this did not happen. Also, the phase shifting of growth in mass, food consumption and possibly growth in length in experiment 1 argues against a chance observation ; one would expect that the group exhibiting the greatest growth and, hence, occupying the peak position for growth in mass would simultaneously occupy the peak positions for growth in length and food consumption. The pattern observed when growth rates from experiments 1 and 3 were plotted on a semi-lunar scale specifically supports a semi-lunar rhythm. Lower growth rates appear to be associated with the time of new and full moons. If the observed growth rhythm had been a result of a 12-day weighing cycle, then one would expect it gradually to become out of phase with a semi-lunar cycle at an approximate rate of 2 days/cycle. This was not observed even in experiment 1 where a semi-lunar rhythm is more tenuous. Finally, the cyclical pattern of amino acid uptake by scales, liver and muscle RNA:DNA ratios, carcass water content, haematocrit, and plasma thyroxine, triglyceride, glucose and cholesterol levels during the semi-lunar period strongly suggest the existence of a semilunar cycle in physiology and thus, indirectly, growth (Farbridge & Leatherland, 1987). These results were based on measurements taken over several different lunar cycles and thus do not suffer the same problems of experimental design as those reported in this paper.

The fish were started on the experimental feeding regime several weeks before the beginning of the weighing schedule. This should have precluded the possibility of an initial increased feeding rate initiating the growth rate pattern which may have been maintained simply by digestion rates and degree of stomach fullness. Given the combined length of the pre-experimental and experimental periods one would otherwise have expected to see a slow loss of synchronicity between fish.

It is not possible to assign precise dates to the times of peaks and troughs of growth, because there is no information on the growth of the fish during the weighing interval itself. The growth may actually peak some time during the interval. Consequently, the observed peaks only represent the approximate positions of periods of rapid growth. This fact is most pertinent when speculating on possible Zeitgeber.

The lunar cycle is the most obvious environmental factor that might be serving as a Zeitgeber for the synchronization of the observed growth rhythm. A biological rhythm may be defined as having a lunar periodicity if the peaks and troughs of the rhythmical process appear once or twice in every lunar month at the same time, that is at the time of a certain lunar phase (Hauenschild, 1961). Periods of rapid growth were correlated with lunar phases. New and full moons appear to be associated with periods of lower growth while more rapid growth appears to occur some time midway between new and full moons. The precise phase relationship between growth and lunar phenomenon is difficult to determine from these data, primarily because of the relatively long time between weighings (4 days).

Rounds (1983) reported a 10- to 14-day rhythm of acetylcholine receptor escape from atropine blocking in cockroach hearts. This rhythm could not be explained simply by the lunar phase cycle. The 10- to 14-day cycle seemed to be the product of two pulses, 12 h apart, progressing across successive days. The times of pulse progression did not correspond to 50 min day⁻¹ which could indicate upper and lower transits of the moon (tidal variations) but instead to the rate of change of moonset across successive days (30–80 min day⁻¹). Perhaps these fish are also using a similar lunar cue of moonrise and/or moonset. If so, it is important to consider that coho salmon feed primarily by sight and, consequently, depend on illumination for feeding. If their growth is influenced by lunar phenomena, the resulting rhythm is likely to be a result of an interaction between lunar day (24·8 h) and solar day (24·0 h) phenomena. It is important to remember that these fish have been maintained under a constant photoperiod from the time of their emergence from the egg. The pattern of growth may take on a different appearance in the natural environment.

The mechanism which allows these fish to respond to lunar and perhaps solar cues (see below) is not known. In marine organisms, the daily changes in tides may directly determine physiology and behaviour. Yet marine teleosts that depend on tides for feeding exhibit lunar and semi-lunar checks (discontinuities) in otolith growth. Check formation is normally associated with a stressful incident in the life history of the fish (Pannella, 1980); stresses reduce branchial uptake of calcium, resulting in a calcium-poor structure which is usually prominent relative to the surrounding daily increments (Campana, 1983). However, an analogous mechanism does not appear to exist during the formation of lunar checks (Campana, 1984). These observations suggest that it is not simply tidal variation which determines the pattern of growth. Certainly, one can envisage the advantages of anticipating tidally determined feeding behaviour.

The unusually high growth rate followed by a rapid trough in growth just before the spring equinox was observed in rainbow trout in 1984 (Farbridge, 1985) and in coho salmon parr in 1985 (Farbridge & Leatherland, 1987). Observations suggest that the spring equinox is an important event for these fish.

The poor documentation of the growth rhythms in these salmonid species is at first perplexing. However, in most studies on growth, fish are weighed at intervals of 1 week or longer. Obviously, the growth rhythm described in this study would have been difficult to detect if a similar weighing regime had been employed. Also, the growth rhythm was lost when fish were sampled from a common stock. Numerous studies have obtained growth data by subsampling. However, because of the inherent variation in size of these fish, this method may not be suitable for detecting short-term changes in growth.