Does female nuptial coloration reflect egg carotenoids and clutch quality in the Two-Spotted Goby (Gobiusculus flavescens, Gobiidae)?

Authors


†Author to whom correspondence should be addressed. E-mail: Andreas.Svensson@bio.ntnu.no

Summary

  • 1Carotenoid-based ornamentation has often been suggested to signal mate quality, and species with such ornaments have frequently been used in studies of sexual selection.
  • 2Female Gobiusculus flavescens (Two-Spotted Goby) develop colourful orange bellies during the breeding season. Belly coloration varies among mature females, and previous work has shown that nest-holding males prefer females with more colourful bellies. Because males invest heavily in offspring during incubation, the evolution of this preference can be explained if colourful females provide males with eggs of higher quality.
  • 3We tested this hypothesis by allowing males to spawn with ‘colourful’ and ‘drab’ females and comparing parameters including egg carotenoid concentration, clutch size, hatchability and larval viability between groups. We also investigated relationships between egg carotenoid concentration and clutch quality parameters.
  • 4Eggs from colourful females had significantly higher concentrations of total carotenoids than drab females, and photographically quantified belly coloration was a good predictor of egg carotenoid concentration.
  • 5Colourful females produced slightly larger clutches, but female belly coloration was not related to any measure of clutch quality. In addition, there were no significant relationships between egg carotenoids and clutch quality. Females with high levels of egg carotenoids spawned slightly earlier, however, possibly because they were more ready to spawn or because of male mate choice.
  • 6Our results call into question the generality of a causal link between egg carotenoids and offspring quality.

Introduction

Evolution of ornament-preference systems relies on the ornaments conferring a benefit to both signaller and receiver. To the signaller, the benefit is to be chosen by members of the other sex, despite the costs of carrying the ornament (Darwin 1871; Zahavi 1975). To the recipient, the benefit may be to gain information about the quality of the signaller. Accurate estimation of the quality of a sexual partner may be of importance to the choosing sex, since selecting a high-quality mate may yield direct (Price, Schluter & Heckman 1993; Siefferman & Hill 2003) or indirect benefits (Reynolds & Gross 1992; Kirkpatrick & Barton 1997). Numerous studies have investigated female preference for male ornaments and the potential benefits females may receive by selecting highly ornamented males. For example, females may benefit by mating with males that have higher fertility (Pitcher & Evans 2001), provide better parental care (Hoelzer 1989) or are bolder (Godin & Dugatkin 1996). Alternatively, the female can benefit by obtaining offspring with genes that make them more attractive (Fisher 1930; Brooks 2000; Gwinner & Schwabl 2005) or result in better body condition (Parker 2003).

In contrast, the presence of ornamentation in females has been relatively little studied but represents a growing area of interest (Amundsen, Forsgren & Hansen 1997; Amundsen 2000; Berglund 2000; Foote, Brown & Hawryshyn 2004; Massironi, Rasotto & Mazzoldi 2005; Amundsen & Pärn 2006). Extravagant traits in females have traditionally been thought to be a non-adaptive side effect, through genetic correlation, from selection on male traits (Lande 1980; Nordeide 2002). Alternatively, female ornaments have been suggested to be used primarily in intrasexual aggression (Beeching et al. 1998; Torres & Velando 2005). In recent years, however, several studies have shown that female ornaments may be important in male mate choice (Amundsen & Forsgren 2001; Griggio et al. 2005; Torres & Velando 2005). In many teleost fishes, males provide sole paternal care of eggs, thus heavily investing in offspring. By doing this, the male may limit his own potential reproductive rate (Clutton-Brock & Parker 1992), and could therefore benefit from discriminating among females of different reproductive potential. To increase the number of offspring resulting from a mating, males may mate with the most fecund females. To select such females, the male can use an easily recognizable physical cue such as body size, which may be directly related to female fecundity (Itzkowitz et al. 1998; Kraak & Bakker 1998). However, males that control nests large enough to accommodate several clutches should pay less attention to the fecundity of individual females. Similarly, if the variation in female fecundity is small, or if the relationship between female size and fecundity is weak, males are not expected to pay much attention to female size (Parker 2003; Pélabon et al. 2003).

Apart from differing in number of eggs, females may also differ in the quality of their eggs (Brooks, Tyler & Sumpter 1997). Although egg size has been repeatedly suggested as an important factor (Bernardo 1996; Hendry, Day & Cooper 2001), maternally derived components in eggs such as hormones (Brooks et al. 1997), lipids (Rennie et al. 2005; Salze et al. 2005) or carotenoids (Blount, Houston & Møller 2000) may also significantly affect offspring fitness.

Carotenoids are lipid-soluble pigments, synthesized by photosynthetic organisms, meaning that animals must obtain them through their diet (Goodwin 1984). Carotenoids are largely responsible for red and yellow pigmentation of flesh, integument and eggs in many bird and fish species (Blount et al. 2000). These pigments have received increasing attention because, in addition to their role in signalling, they are thought to have a variety of beneficial effects on animal physiology, for example as precursors for vitamin A (Moren, Næss & Hamre 2002), in immune defence (Kawakami et al. 1998; Amar et al. 2001), in protection from light (Craik 1985) or as antioxidants (reviewed in Edge, McGarvey & Truscott 1997; Christiansen et al. 1995; McGraw 2005). In species with carotenoid-based ornamentation, the ornaments have been suggested to signal a capacity to obtain carotenoids in the diet and to incorporate them in the body and, in the case of females, into eggs (Bortolotti et al. 2003; McGraw, Adkins-Regan & Parker 2005). Increased carotenoid levels in the maternal diet have also been shown to positively affect egg quality in invertebrates (George et al. 2001) and in birds (Surai & Speake 1998). The presence of carotenoids in eggs may be critical for the development of fish embryos, as suggested by high juvenile mortality in carotenoid deficient eggs of Salmo salar (Atlantic Salmon) afflicted by the M74 syndrome (Pettersson & Lignell 1999). Evidence for a link between the carotenoids used in signals and their positive effects on physiology is still limited and equivocal, and the issue is controversial. Recently, the view that carotenoid displays are directly linked to antioxidant capacity has been criticized by Hartley & Kennedy (2004). There are also several studies that have failed to find beneficial effects of carotenoids. For example, Christiansen & Torrissen (1997) found no relationships between carotenoid concentration of the eggs and fertilization rate, survival to the eyed stage and larval performance in S. salar. Similarly, Lygren et al. (1999) measured several health parameters in juvenile salmon but found none of these to be affected by dietary carotenoids. Further studies investigating the relationship between ornamentation, carotenoids and quality parameters are therefore required.

Although a male preference for female ornamentation has been shown in species with conventional sex roles (Amundsen & Forsgren 2001; Griggio et al. 2005), few studies have attempted to investigate what a male stands to gain by selecting more ornamented females. In Gobiusculus flavescens (Two-Spotted Goby) there is only a weak male preference for female size, explained by a relatively weak relationship between female size and clutch size (Pélabon et al. 2003). In addition this species shows little variation in egg size (C. Pélabon, T. Amundsen & E. Forsgren, unpublished data). On the other hand, there is a clear male preference for females with more colourful orange bellies, and the degree of orange coloration varies greatly among mature females (Amundsen & Forsgren 2001).

In this study we investigated whether more colourful G. flavescens females produce clutches of a higher quality and with higher carotenoid concentration. We also tested whether egg carotenoid concentration was directly related to clutch quality.

Materials and methods

study species

Gobiusculus flavescens is a small, semipelagic marine fish with a distribution from Norway to Spain (Johnsen 1944; Collins 1981), which exhibits paternal care of the eggs (Gordon 1983; Skolbekken & Utne-Palm 2001; Bjelvenmark & Forsgren 2003). G. flavescens are sexually dimorphic with both males and females exhibiting visual ornamentation during the reproductive season. While males have brightly coloured fins and iridescent blue lateral spots, females develop strikingly orange bellies during the breeding season (Amundsen & Forsgren 2001). The nuptial belly coloration of female G. flavescens derives from two sources (Svensson et al. 2005). First and most important, the colourful gonads are directly visible through the semitransparent abdominal skin. Second, the presence of red and yellow chromatophores in the abdominal skin overlying the gonads adds to the coloration of the belly (Svensson et al. 2005). Female courtship involves approaching the male and bending the body (sigmoid display), a behaviour that seems to emphasize the round and colourful belly (Amundsen & Forsgren 2001).

Spawning takes place inside empty mussel shells or in fronds of algae. After courtship the female deposits eggs inside the nest and the male fertilizes and cares for them until hatching (Gordon 1983; Bjelvenmark & Forsgren 2003). Eggs hatch within 10 days at 16 °C (see Results), but hatching time can range from 7 to 21 days, depending on temperature and the degree of mechanical disturbance, e.g. by water movement or male activity (Miller 1986; Falk-Petersen et al. 1999; Skolbekken & Utne-Palm 2001; Folkestad 2002). The larvae have relatively small yolk sacs, are free-swimming and forage immediately after hatching (Folkestad 2002).

general protocol

This study investigated the relationships between female nuptial coloration, egg carotenoids and egg and larval quality in G. flavescens. From daily catches of mature females, we selected the most drab and colourful and allowed these to spawn with nest-holding males. The clutches obtained were incubated without the male (to avoid confounding effects of paternal care) and several parameters related to egg and larvae quality were quantified. The belly coloration of females was quantified using digital image analysis, and egg carotenoid concentration was quantified with high-performance lipid chromatography (HPLC).

fish collection, husbandry and spawning

The study was carried out at the Kristineberg Marine Research Station (58°15′ N, 11°27′ E), in Fiskebäckskil, Sweden in June 2002. Fish were caught with handheld dip nets while snorkelling or with lift nets from land. All fish were placed in single-sex storage tanks with a continuous flow of surface sea water at ambient temperature (16·3 ± 2·2 °C, range 10·5–21·4 °C). Round, i.e. sexually mature, females were put in a separate holding tank. Males were fed Artemia nauplii ad libitum. To avoid effects of food intake on eggs and coloration, females were not fed during the study (maximum 5 days for each female).

Males were placed individually in 20-L tanks (25 × 30 × 30 cm3) with a plastic plant and an 80 mm long, 25 mm Ø PVC tube as nest substrate. To facilitate the removal of laid eggs, a 60 × 80 mm2 clear acetate sheet was placed inside the tube. Males that were observed inside the nest and/or courted a female presented to them in clear bottles, were regarded as nest holding and used in the study.

Each day of collection, drab and colourful individuals were selected from the group of newly caught round females (group size range: 10–110 females). This selection was similar to that of Amundsen & Forsgren's (2001) study of male response to female coloration. A standardized array with two spotlights was directed against the holding tank to aid the selection. The most colourful and drab females were selected and each placed in a beaker with sea water. This process was repeated until the remaining fish were of intermediate colour. Intermediate-coloured females were released. Over the course of the study, a total number of 870 round females were collected and of these 190 were selected and introduced to males. Each day, on average five drab and five colourful females were selected from a group of 45 females. Thus, the selected females roughly comprised the upper and lower 11% of round females with regard to belly coloration.

The selected females were photographed (see below), carefully blotted and weighed to the nearest mg, and measured (total length, TL) to the nearest 0·5 mm on a millimetre grid. The roundness of females was defined as the residual from a length–mass regression performed on all spawning females using log-transformed data (see Pélabon et al. 2003). Females were released individually into tanks containing a nest-holding male. The spawning tanks were observed more than five times per day between 8 am and 6 pm, and the times of spawnings were recorded. If no spawning had occurred within 48 h, the female was transferred to another male. If spawning had not occurred within 48 h with the second male, the female was released.

clutch handling and egg incubation

The clutches were hatched in incubators to remove any effects of paternal care, such as differences in filial cannibalism or fanning behaviour (Bjelvenmark & Forsgren 2003). This also allowed standardization of the mechanical disturbance of the clutches, which can affect time until hatching (Falk-Petersen et al. 1999). Once a spawning was completed, the female was removed. Males were left undisturbed for 2–6 h after spawning so as not to interfere with fertilization. Thereafter, the acetate with the eggs was removed from the nest and photographed (see below). All except 150–250 eggs were scraped off the acetate, transferred to a 1·5 ml test tube and frozen for later carotenoid analysis. A second photograph was taken of the reduced clutch (see below). The acetate with the reduced clutch was placed on a Perspex strip in a (25 × 20 × 30 cm3, 15 L) incubator tank with filtered sea water adjusted to a salinity of 25 psu. The incubator tanks were placed in a dimly lit, temperature controlled room set to 16 °C, and had continuous vigorous aeration. Each day, we recorded the numbers of undeveloped (clear), dead (cloudy) or eye-spotted (visible dark spots) eggs. On the day before hatching was expected, aeration was reduced to a minimum in order not to harm newly hatched larvae. The number of remaining, unhatched eggs was recorded when hatching was completed.

larval viability

A viability study was conducted to test for differences in larval nutritional reserves. On the day of hatching, 50 larvae were removed from each incubator and placed in five plastic cups, each with 0·2 L filtered sea water (10 larvae in each cup). Every day, the number of dead larvae was counted. Larvae that did not move, even when gently touched with a nylon thread, were noted as dead.

photography and image analysis

Females were selected over a period of 26 days, and it is possible that the person performing the selections (P.A.S.) involuntarily varied the criteria for ‘drab’ and ‘colourful’ during this time period. To obtain a measure of belly coloration that was consistent over time and allowed quantification of colour differences, standardized digital photographs were taken of all females before the first introduction to a male. In a dark room, fish were individually placed in a 7 × 4 × 2 cm3 aquarium containing sea water. A Canon D30 digital camera equipped with a Canon 50 mm f/2·5 EF Compact Macro lens (Canon Norge AS, Oslo, Norway) and a Senz stereo macro flash (Photax AB, Nybro, Sweden) was used to take an image of each side of every fish (Fig. 1). Exposure time (1/60), aperture (f8) and flash power settings were kept constant for all photographs. Both sides of each female were photographed. As females were not fed, and all of them spawned within 3 days of photography, we assumed that female belly coloration would not change between photography and spawning.

Figure 1.

An example of a photograph used to quantify the belly coloration of female G. flavescens (Two-Spotted Goby). The dashed line indicates the area selected for the colour quantification.

Photographs of females were analyzed using Adobe Photoshop 4·0 (Adobe Systems Inc.). The digital images were converted to L*a*b*, a colour space recommended by the Commission International de l’Eclairage (CIE). CIE L*a*b* consists of three parameters: the L* value (lightness) representing the relative lightness ranging from total black to total white, the a* value (‘redness’) representing the balance between red and green, and the b* value (‘yellowness’) representing the balance between yellow and blue. In contrast to, for example RGB colour space, CIE L*a*b* is a standardized, perceptually uniform and device-independent colour space (Chen, Hao & Dang 2004). It has been frequently used in fish colour quantification, especially in connection with carotenoid-based coloration (e.g. Skrede & Storebakken 1986; Smith, Hardy & Torrissen 1992; Hatlen, Jobling & Bjerkeng 1998; Craig & Foote 2001). Redness (a*) was chosen for the analysis of belly coloration, as it produces the best correlation between belly and gonad colour (Svensson et al. 2005), as well as between belly colour and gonad carotenoid concentration (P. A. Svensson, T. Amundsen, J. D. Blount & C. Pélabon, unpublished data).

The belly area was selected with the Photoshop lasso tool and was defined as the roughly elliptic area between the anal pore, the pectoral fin base and the blue spots below the lateral line (Fig. 1). The mean redness (a*) of the selected area was measured using the histogram tool. The values from the left and right side of each female were averaged prior to analysis.

Acetates with clutches were photographed in sea water on a light table with the same camera and lens (f4, 1/60 s). This first photograph was used to determine the full clutch size. After scraping off and removing all but 150–250 eggs, a second photograph was taken to determine the number of eggs that were placed in the incubator. The number of eggs was counted using ImageTool (UTHSCSA, San Antonio, TX).

On the day of hatching, 10 larvae from each female were randomly selected from the incubator and placed on a microscope slide with a drop of MS222 solution. When the larvae were immobile, photographs were taken of each larva using the camera above, mounted on a compound light microscope with 40× magnification. The length of the larva (SL) was measured in ImageTool by drawing a line from the front of the upper jaw to the caudal peduncle. Length measurements were calibrated using photographs of micrometer scales.

analysis of carotenoids

It has been hypothesized that the colourful bellies of female G. flavescens originate from carotenoid-rich eggs visible through the abdominal skin (Amundsen & Forsgren 2001). To test this hypothesis, egg carotenoids were quantified using HPLC at the Scottish Agricultural College, Auchincruive, UK. Excess water was removed from test tubes with eggs by gentle centrifugation followed by placing a folded 6 cm filter paper in each tube for 5·0 min. The samples were weighed to the nearest 0·1 mg. All samples were homogenized in acetone, 5% sodium chloride and hexane. Extraction was performed twice, combining the hexane fractions after centrifugation. Hexane was evaporated under N2 gas at 60 °C, and the remaining fraction was redissolved in 50:50 methanol : dichloromethane. Total carotenoids were determined as described by Horak et al. (2004), using a Spherisorb, type S5NH2 5 µm reverse phase HPLC column, 25 cm × 4·6 mm (Phase Separations Ltd, Deeside, UK). Chromatography was performed using a mobile phase of methanol:water (97:3, v/v) at a flow rate of 1·5 ml min−1 and total carotenoids were detected as a single peak at 445 nm, using astaxanthin as a standard.

statistics

Data analyses were performed in R 2·2·1 (R Development Core Team 2005, http://www.r-project.org). Proportions were arcsine square root transformed before analyses. Regressions with binary response variables were performed using generalized linear models with binomial error distribution (Crawley 2002). In the viability study, time of death of 50 larvae from each female was recorded. Preliminary analysis revealed within-female variation in this parameter to be non-significant. Therefore, larval viability was averaged for each female prior to further analyses.

Results

comparison of drab and colourful females

Image analysis confirmed a difference in belly coloration between the two groups, i.e. females selected as colourful had significantly higher redness (a*) values than females selected as drab (Table 1). The drab and colourful females did not differ significantly in length, body mass or roundness (Table 1). Females in the colourful group laid clutches with significantly higher concentrations of total carotenoids compared with drab females (Table 2). Of the 190 females introduced to nest-holding males, 48 spawned. This spawning success is similar to other studies of Two-Spotted Gobies in captivity (T. Amundsen, I. Barber, J. Bjelvenmark, Å. A. Borg, E. Forsgren, C. Pélabon & P. A. Svensson, unpublished data). There was no significant difference between the two groups in the proportion of females that spawned (Table 2). Similarly, there was no difference between the groups in the time it took before spawning commenced (Table 2). Colourful females (as judged by visual inspection) laid significantly larger clutches than drab females (Table 2). Of the 48 clutches, 35 contained eggs that developed, while 13 clutches had no developing eggs. The proportion of eggs developing eyes and the proportion of eggs hatching did not differ between drab and colourful females (Table 2). There was a non-significant tendency for larvae from colourful females to be longer on the day of hatching, compared with larvae from drab females (Table 2). Larval viability was quantified using time of death (days after hatching) of larvae. There was no significant difference in time of death between colourful and drab females (Table 2).

Table 1.  Comparison of belly coloration and size between G. flavescens females visually categorized as either ‘drab’ (n = 22) or ‘colourful’ (n = 26). Belly coloration was measured as CIE a* (‘redness’) in digital photographs (see text for details). Roundness was calculated as the residuals from a regression between length and mass using log-transformed data
Measured parameterDrab femalesColourful femalestP
MeanSDMeanSD
Belly coloration (CIE a*)142·23·71147·12·485·482< 0·0001
Total length (mm)43·952·2844·002·990·058    0·954
Body mass (g)0·7370·100·7220·120·437    0·664
Roundness0·0090·04−0·0070·041·382    0·174
Table 2.  Comparison of egg carotenoid concentration and measures of female, clutch and larval quality between G. flavescens females visually categorized as either ‘drab’ or ‘colourful’ (see text for details)
Measured parameterDrab femalesColourful femalesdfTest statisticP
MeanSDnMeanSDn
Total carotenoid concentration in eggs (µg g−1) 3·45 1·4122 4·68 1·362646t = 3·0800·003
Proportion females that spawned 0·23 100 0·33 90 1χ2 = 2·0270·155
Time until spawning (h)31·9316·22229·0213·62646t = 0·6770·502
Clutch size (number of eggs)12062562213602142545t = 2·2490·029
Proportion of clutches with hatching eggs 0·77 22 0·69 26 1χ2 = 0·0890·765
Incubation time (day)10·150·261710·15 0·321833t = 0·0580·954
Proportion of eggs with eyespots 0·960·0717 0·94 0·201833t = 0·0410·968
Proportion of eggs that hatched 0·920·2217 0·93 0·211833t = 0·2230·825
Larval standard length on day of hatching (mm) 2·600·1016 2·66 0·081832t = 1·9770·057
Larval viability (time of death after hatching) 5·220·6216 5·45 0·651731t = 1·0630·296

The drab/colourful two-group design was chosen because it corresponded to the selection design in Amundsen & Forsgren's (2001) study, showing male mate preference for colourful females. In addition, we quantified female belly coloration through image analysis of digital photographs. This allowed examination of relationships between clutch quality and female coloration on a continuous scale using regression analyses. These results, presented in Table 3, were largely in agreement with the original two-group design (Table 2). However, the effect of female coloration on clutch size was not present in the regression analysis (Table 3). Two non-significant trends also appeared in these regressions; a negative trend between photographic belly coloration and time until spawning, and a positive trend for colourful females to have larger larvae (Table 3).

Table 3.  Linear regression analyses between photographically determined belly coloration (CIE a*, ‘redness’) and reproductive parameters reflecting female, clutch and larval quality in G. flavescens females
y parameterR2SlopeSEdfP
  • *

    Binary regression using a GLM with binomial error distribution.

Egg total carotenoid concentration (µg g−1)0·4270·2360·041 45< 0·0001
Proportion females that spawned*0·0120·043188    0·772
Time until spawning (h)0·079−1·0480·529 46    0·054
Clutch size (number of eggs)0·001−1·8589·20145    0·841
Proportion of clutches with hatching eggs*−0·0410·08546    0·632
Incubation time (day)0·0220·0100·01233    0·399
Proportion of eggs with eyespots0·0050·2080·49633    0·677
Proportion of eggs that hatched0·0280·6810·69433    0·334
Larval standard length on day of hatching (mm)0·1010·0070·00432    0·066
Larval viability (time of death after hatching)0·0160·0200·02831    0·476

relationships between egg carotenoid concentration and quality parameters

Female belly coloration measured in photographs before spawning and the total carotenoid levels in the laid eggs were significantly positively related (Fig. 2). However, there was overlap between the two visually selected groups (colourful/drab) in both carotenoid concentration and belly coloration measured in photographs (Fig. 2). In addition, egg carotenoid concentration explained only about 42% of the variation in belly coloration (Table 4). To test the more proximate hypothesis that egg carotenoid concentration in itself is related to clutch quality, we performed regression analyses between egg carotenoid concentration and the various quality parameters (Table 4). After inspection of Cook's distances, one outlier with particularly high leverage effect was removed (egg carotenoids: 0·79 µg g−1, belly coloration (a*): 147·4). There was a significant negative relationship between egg carotenoid concentration and the time females spent with males before spawning commenced (Table 4). This is consistent with the negative trend found between belly coloration and time until spawning (Table 3). There were no significant relationships between egg carotenoid concentrations and clutch size, incubation time, the proportion of eggs developing eyes, hatching success, larval size or larval viability (Table 4).

Figure 2.

Relationship between total carotenoid concentration in eggs and belly coloration measured as CIE a* (‘redness’) from digital photographs taken of female G. flavescens before spawning. Solid circles indicate females visually judged as ‘colourful’; open circles indicate females judged as ‘drab’.

Table 4.  Linear regression analyses between egg carotenoid concentration (µg g−1) and reproductive parameters reflecting female, clutch and larval quality in G. flavescens females
y parameterR2SlopeSEdfP
  • *

    Binary regression using a GLM with binomial error distribution.

Female belly coloration (CIE a*)0·427 1·81 0·31245< 0·0001
Time until spawning0·110−3·438 1·46045    0·023
Clutch size (number of eggs)0·02129·5424·9744    0·331
Proportion of clutches with hatching eggs*−0·363 0·25045    0·147
Incubation time (day)0·037 0·036 0·03233    0·277
Proportion of eggs with eyespots0·027−2·607 2·54233    0·312
Proportion of eggs that hatched0·013 1·194 1·80633    0·513
Larval standard length on day of hatching (mm)0·008 0·006 0·01230    0·628
Larval viability (time of death after hatching)0·015 0·043 0·06530    0·514

Discussion

Our results suggest that colourful G. flavescens females produce eggs with higher carotenoid concentration, but not of higher quality. Colourful and drab females laid eggs with similar hatching success and time until hatching. Since G. flavescens larvae forage immediately after hatching (Folkestad 2002), it is likely that even small differences in larval reserves will affect the duration of the window of opportunity, i.e. the time until the ‘point of no return’ when exogenous food must be present (e.g. Peña & Dumas 2005). The larval viability study aimed at detecting such differences in nutritional reserves. Despite large variations in time of death between clutches, neither the belly coloration of the mother nor the carotenoid content of the eggs affected larval viability significantly. Likewise, there were no significant relationships between the egg carotenoid concentration and the other offspring quality parameters. There was however, a significant negative relationship between egg carotenoids and the time it took for a female to spawn.

female belly coloration – a signal of fecundity?

The colourful belly of G. flavescens originates mainly from the pigmented gonads being visible through the transparent abdominal skin (Svensson et al. 2005). Our results revealed that females visually selected as colourful had significantly higher egg carotenoid levels compared with females selected as drab (Table 2). In addition, there was a significant positive relationship between egg carotenoid concentration and belly coloration measured from photographs (Table 4, Fig. 2). The nuptial signal in female G. flavescens therefore seems to reliably signal a capacity to obtain carotenoids in the diet and incorporate them into the eggs, as previously shown in birds (Bortolotti et al. 2003; McGraw et al. 2005). Despite this, the only significant quality difference between the two groups was that colourful females laid c. 150 more eggs than drab females. It may therefore seem reasonable to explain female belly coloration simply as a signal of fecundity. The result should be interpreted cautiously, however, as there was no relationship between belly coloration measured in photographs and clutch size (Table 3). Even if the observed pattern is genuine, it is problematic to infer that female belly coloration serves as a signal of fecundity in male mate choice. The colourful group consisted of the most colourful 11% of all the females. A choosy male would therefore have to reject 9 of 10 females to increase his clutch size with, on average, 75 eggs, compared with a male mating randomly with regard to female belly coloration. Male G. flavescens will normally mate with several females in quick succession and may care for two to four clutches simultaneously, i.e. 2000–5000 eggs (P. A. Svensson, J. Bjelvenmark, E. Forsgren & T. Amundsen, unpublished data). Accordingly, to reject several females in order mate with one slightly more fecund seems disadvantageous in comparison with obtaining two or more clutches, even if these should be slightly smaller. A similar female signal exists in Knipowitschia panizzae (Lagoon Goby), where ready-to-spawn females develop a yellow patch on their belly, and the area of this patch indicates female fecundity (Massironi et al. 2005). In K. panizzae, however, the male typically cares for one clutch at a time and compared with G. flavescens, differences in female fecundity will be more directly related to the male's reproductive success. Taken together, it seems unlikely that the colourful belly in female G. flavescens is a mere signal of fecundity. This calls for alternative or additional explanations for the reported strong male preference for colourful females (Amundsen & Forsgren 2001).

egg carotenoids, belly coloration and time-to-spawn

Regression analyses were performed to investigate the relationship between egg carotenoid concentrations and belly coloration, and to test for possible relationship between egg carotenoids and clutch quality. Egg carotenoid concentration explained almost half the variation in belly coloration (Fig. 2 and Table 4). Therefore belly coloration, measured as CIE a* in digital photographs constitute a viable, non-invasive method to estimate the concentration of egg carotenoids in this species. There was a significant negative relationship between egg carotenoid concentration and the time females spent with males until spawning commenced (Table 4). There was also a non-significant tendency for belly coloration measured in photographs to be negatively related to time until spawning (Table 3). As egg carotenoids relate strongly to belly coloration, the difference in time-to-spawn could be interpreted as an effect of male choosiness, i.e. a greater male interest in colourful females with higher egg carotenoid levels (Amundsen & Forsgren 2001). Alternatively, the result might have been generated by a difference in female readiness to spawn. If egg carotenoids accumulate as eggs mature, females with high egg carotenoid levels could have more mature eggs (Bjerkeng, Storebakken & Liaaen-Jensen 1992; Dierenfeld et al. 2002). These females might then be the most ready, and therefore most eager, to spawn (McLennan 2005). Neither of these relationships were very strong, however (Table 4), and the large number of tests call for caution in interpretation. In addition, there was no corresponding difference in time until spawning in the two-group design (Table 2).

egg carotenoids and offspring quality

From the present literature it is hard to predict if and exactly how maternally acquired carotenoids should affect offspring fitness, as such studies are rare and have yielded equivocal results. In general, carotenoids are suggested to offer protection from free-radical-induced cell damage through their function as antioxidants (reviewed in Edge et al. 1997). Carotenoids could therefore be important for proper embryonic development, and be expected to affect hatching rates or the survival of newly hatched larvae (Tsushima et al. 1997; George et al. 2001; McGraw et al. 2005). In Palaemonetes pugio (Grass Shrimp), egg carotenoids decrease during embryo development, whereas other antioxidants such as enzymes increase (Winston, Lemaire & Lee 2004). Thus, it is likely that maternally acquired carotenoids have the greatest role early on in development, before more complex antioxidant systems are assembled. Effects of maternal carotenoids are also expected to dissipate after the onset of exogenous feeding, as shown in Hippoglossus hippoglossus larvae (Rønnestad et al. 1998b). Despite choosing to quantify ‘early parameters’ such as hatching success and larvae nutritional reserves, and despite a seven-fold range in egg carotenoid concentrations (1·1–7·5 µg g−1), we found no significant effects of egg carotenoids on the quality of G. flavescens eggs and larvae. This corroborates a study on S. salar by Christiansen & Torrissen (1997), who also found no effects of egg carotenoids on fertilization rate, hatching rate and larval survival. Such lack of effects could be due to even the poorest clutches containing enough carotenoids to avoid carotenoid deficiency. Adult G. flavescens feed mainly on calanoid copepods, a food source very rich in carotenoids (Berg 1979; Costello, Edwards & Potts 1990; Van Nieuwerburgh et al. 2005). We found G. flavescens to have fairly high concentrations of egg carotenoids compared with other species, as all females laid eggs with more than 1 µg g−1, and 90% of the females laid eggs with more than 2 µg g−1. For salmonids, Craik (1985) suggested that a critical level for hatching might be 1–3 µg g−1, but Pettersson & Lignell (1999) found a mere 0·11 µg g−1 to be enough for healthy development of eggs in S. salar (Baltic Salmon). Eggs of wild Gadus morhua (cod) have as low carotenoid levels as 0·7 µg g−1 (Grung, Svendsen & Liaaen-Jensen 1993), while other teleosts seem to make do without egg carotenoids altogether, as for example Pseudocaranx dentex (Vassallo-Agius et al. 1998) and H. hippoglossus (Rønnestad et al. 1998b). Investigations of four other gobies occurring in the same area indicate that egg carotenoids in these species are less than half of that in G. flavescens (0·2–3 µg g−1; P. A. Svensson, J. D. Blount, P. F. Surai, T. Amundsen & E. Forsgren, unpublished data). Although such interspecific comparisons may be problematic, it is possible that even the drabbest G. flavescens females had sufficient egg carotenoids to ensure proper egg development and hatching.

High carotenoid levels may be important for life stages later than the ones investigated in this study. For example, carotenoids may affect juvenile growth rate (Christiansen, Lie & Torrissen 1994) or immune response (Christiansen et al. 1995; Amar et al. 2001; Biard, Surai & Møller 2005). It is also possible that egg carotenoids may affect general pigmentation of the embryo (Pan, Chien & Cheng 2001), or function as a source of pigments important for photoreception (Rønnestad, Helland & Lie 1998a). Egg carotenoids may also have benefited the larval immune system in a way that was not detected in our study (Kawakami et al. 1998; Amar et al. 2001), as the filtered water in the incubator tanks may have had unusually low levels of natural pathogens. To merely note larvae as dead or alive might also be too crude a measurement to detect differences in larval quality, and future studies should incorporate some estimate of larval vigour.

In conclusion, we found the nuptial coloration of female G. flavescens to be strongly related to the carotenoid content of the eggs. Despite this, our study provided little support for the idea that female coloration reflects egg or larval quality. Experimental manipulations of female coloration and more sensitive measures of larval quality are needed before this hypothesis can be rejected altogether.

Acknowledgements

Thanks to Uwe Berger, Jens Bjelvenmark, Åsa A. Borg and Tanja Viio for field assistance, to Kenny McIsaac for lab assistance and to Elisabet Forsgren for helping develop the framework for this study. We thank the editors and two anonymous referees for helpful comments. The study was supported by the MARE programme of the Research Council of Norway. The fieldwork was supported by Kristineberg Marine Research Station, the Royal Swedish Academy of Science and the Nordic Academy for Advanced Study (NorFA). The procedures were approved by the Swedish Animal Welfare Agency (Göteborgs djurförsöksetiska nämnd), Dnr 125–2002.

Ancillary