Higher social status is expected to result in fitness benefits as it secures access to potential mates. In promiscuous species, male reproductive success is also determined by an individual’s ability to compete for fertilization after mating by producing high quality ejaculates. However, the complex relationship between a male’s investment in social status and ejaculates remains unclear. Here we examine how male social status influences ejaculate quality under a range of social contexts in the pygmy halfbeak Dermogenys cf. collettei, a small, group-living, internally fertilizing freshwater fish. We show that male social status influences ejaculate traits, both in the presence and absence of females. Dominant males produced faster swimming and more viable sperm, two key determinants of ejaculate quality, but only under conditions with frequent male-male behavioral interactions. When male-male interactions were experimentally reduced through the addition of a refuge, differences in ejaculate traits of dominant and subordinate males disappeared. Furthermore, dominant males were in better condition, growing faster and possessing larger livers, highlighting a possible condition-dependence of competitive traits. Contrary to expectations, female presence or absence did not affect sperm swimming speed or testes mass. Together, these results suggest a positive relationship between social status and ejaculate quality in halfbeaks, and highlight that the strength of behavioral interactions between males is a key driver of social-status dependent differences in ejaculate traits.

Study populations and rearing conditions

Two experiments were performed on sexually mature halfbeaks that were F1 or F2 descendants of fish originating from two different sources. Experiment 1 used fish descendant from adults obtained by a commercial supplier (Ruinemans Aquarium B.V., Montfoort, The Netherlands), while Experiment 2 used descendants of a wild caught parental population from the Tebrau River, Malaysia. For both experiments, juvenile fish were sexed at the initial period of sexual development at around two-months of age. Male halfbeaks were then housed in same-sex tanks at a range of densities (approx. 15 – 30 males per tank) until sexual maturity (~4 months). Differences in housing densities during development can influence allocation to reproductive traits (Gage 1995). However, as densities were not biased in one treatment or experiment, any potential effects of rearing density are more likely to add statistical noise than to systematically influence the direction of observed effects. At sexual maturity, males were transferred to individual tanks housing a female to allow males to gain mating experiences before entering the experiment. After a week, the female was removed and males were kept in isolated tanks for another 3 weeks, at which point males entered the experiments. Mating and isolating males prior to the treatment ensured that sperm reserves were replenished. All tanks were oxygenated, contained ~2 cm of gravel and live and artificial plants. Fish were maintained on a 12:12 light:dark cycle at 27°C and fed daily with a mix of ground flake flood, freeze dried Artemia and once per week additional Drosophila melanogaster. Experiments were approved by the Stockholm Animal Research Ethical Board (permit number 3867-2020 and 2393-2018).

Experiment 1: Social status, socio-sexual environments and investment in reproductive and condition traits

The influence of social status on male behavior, physiology and reproductive traits was assessed under a range of socio-sexual environments. Males were photographed one day prior to the start of the experiment using a photo-chamber (30 x 20 x 20 cm) fitted with a scale. Using these images, body length (from tip of the lower beak to the caudal peduncle, see Reuland et al. 2019) was measured and size-matched pairs of males, henceforth called dyads, were created. Males in each dyad were statistically indistinguishable in total body length (average absolute difference within dyad ± SE: 0.44 ± 0.10 mm, n = 32; paired two-sample t-test assuming equal variance: t(31) = 0.30, p = 0.77). Male dyads were placed in experimental tanks (40 x 25 x 30 cm), and allowed to freely interact and establish social dominance hierarchies. Experimental tanks were divided into two chambers, with male dyads being placed in the larger chamber (26 x 25 x 30 cm). A removable opaque divider separated the larger and smaller chamber. Since we were interested in assessing the influence of social status across different socio-sexual contexts, male dyads were allocated to one of three experimental treatments that varied male access to females, including i) a ‘no female’ treatment (n = 9), where the smaller chamber was left empty, ii) a ‘visual access to female’ treatment (n = 12), where the female was placed in the smaller chamber and a transparent divider was added (in addition to the opaque barrier) that separated the female from the male dyad, and iii) a ‘free access to female’ treatment (n = 11), where females could freely interact with the male dyad, allowing both visual and physical contact among all fish.

Male dyads were observed two times per day for 20 min on Days 1, 2, 6 and 10 to determine social status. Observational times were chosen to both capture frequent agonistic interactions early on during the establishment of social roles (Day 1 and 2) as well as to observe the stability of the formed dominance hierarchies (Day 6 and 10). During the initial behavioral observation on the morning of Day 1, the opaque partition was lifted to give males visual access to a female (visual access to female treatment) or physical access to a female (free access to female treatment). To standardize handling, partitions to the empty chamber were also lifted in the no female treatment. Behaviors were recorded by applying a continuous sampling method and recording all the activity that occurred while the animals were observed (all activities within the dyad). Male social status within each dyad was determined by recording displacement behaviors, where one (aggressor) male approaches another or exerts an agonistic behavior and as a consequence the opposing male increases the distance between the two males to more than 2/3 of a body length again. Dominance indexes were calculated for both males in the dyads using the formula: (1 - Displacements by rival maleTotal number of displacements in the dyad ), where values of 1 indicate complete social dominance (males only displaced rival males) and values of 0 indicate complete social subordination (males were only displaced by rivals males). We considered social dominance to be clearly resolved when one male in the dyad displaced the rival male in >70 % of the interactions, which was the case in all dyads (average ± SE dominance index for males classified as dominant: 0.98 ± 0.01, n = 32, range: 0.75 – 1). Furthermore, to assess the frequency of male-male agonistic interactions within dyads, total agonistic behaviors between males (including gill flare, parallel swimming, chasing, beak-locking and biting) were recorded.

On the morning of Day 11, males were transferred to isolated tanks and given new anonymous IDs in preparation for data collection (i.e. males were processed blind to their social status or socio-sexual treatment). Males were photographed in the photo-chamber and then euthanized in a benzocaine solution (final concentration 400 mg benzocaine in 1 L diH₂O, where initial stock solution = 100 mg benzocaine to 1 ml Ethanol). After euthanasia, males were rinsed in distilled water then placed on their side onto a slide covered with saline solution (0.9 % NaCl in diH₂O) and viewed under a dissecting scope (S9 stereo microscope, Leica Microsystems, Wetzlar, Germany). Because male halfbeaks are partially transparent, the posterior part of the testicular duct is visible externally (Downing and Burns 1995). The testicular duct transfers sperm to the andropodium, a modified anal-fin used to transfer sperm to females (Meisner and Burns 1997). Sperm were extracted into a saline solution by applying gentle pressure with a blunt instrument to the posterior part of the testicular duct. Halfbeak sperm cells are arranged in unencapsulated bundles called spermatozeugmata, with sperm tails on the outside and sperm heads facing towards the inside of the aggregate (Downing and Burns 1995). Spermatozeugmata structure remains largely intact when sperm are stripped into a saline solution and cells remain largely inactivated, assuring that sperm bundles can be collected before activation. The saline/sperm solution was then transferred into an equal volume of Hanks’ balanced salt solution (modified, with sodium bicarbonate, without phenol red, H8264, Sigma-Aldrich, St. Louis, United States) and briefly mixed to activate the sperm. After activation, sperm were transferred to a polyvinyl alcohol coated multi-well slide (Multitest slide, MP Biomedicals, Santa Ana, United States) and sperm swimming speed (median curvilinear track velocity VCL, average path velocity VAP, straight line velocity VSL) was analysed using a computer-assisted-sperm-analysis (CASA) system consisting of a computer and camera attached to a microscope (ISAS V1 system, PROiSER R+D, Paterna, Spain). As VCL was strongly positively correlated with both VAP (r = 0.92, n = 63) and VSL (r = 0.97, n = 63), and since cell trajectories were not expected to be linear due to the lack of any egg attraction or ovarian fluid effects in this study, we used VCL in all subsequent analyses (note that results are consistent irrespective of the metric used).

We also recorded change in body area during the experiment, as well as male condition (see below) and liver mass after the experiment. The change in body area (omitting the fins) during the experiment was measured as an approximation for a male’s growth using the photos taken one day before dyad formation and on the 11^th day of the experimental treatment. We analysed body growth using photographs. Photographs were used rather than weighting fish before the experiment to minimise handling of the animals before formation of male dyads. To assess male condition, after males were euthanized at the end of the experiment, they were lightly dried with paper towel, measured for body length (the distance from the eye to the caudal peduncle excluding the beak) and weighed to the nearest 0.01 mg on a fine scale (XS105, Mettler Toledo, Ohio, USA). Condition factor was calculated using the formula: (body mass/body length³) * 100, which is commonly used as an indicator of overall fish health (Ricker 1975; Froese 2006). In teleost fishes, the liver is an important store for energy reserves and relative liver mass is commonly used as an indicator of a fish’s energy status (Tyler and Dunn 1976; Wootton et al. 1978; Culbert and Gilmour 2016). Therefore, after weighting the fish, males were dissected and their liver weighed. Together, these metrics are likely candidate traits to assess energy reserves in halfbeak fishes, as they are commonly used in other fish species.

Males investing more on postcopulatory competitive traits typically have larger testes relative to their body mass (Simmons and Fitzpatrick 2012). Thus, as an indicator for male expenditure on postcopulatory traits, testes of males were dissected and weighted.

Experiment 2: Social status, male-male interactions and ejaculate traits

We investigated how experimental manipulation of male-male interactions influences social status-dependent differences in ejaculate traits. This experiment focused in detail on responses in a range of ejaculate traits, as Experiment 1 suggested that ejaculates are responsive to differences in social status among males (see Results). Males were photographed up to three days prior to the formation of male dyads (see Experiment 1). Size matched male dyads (average absolute difference within dyad ± SE: 0.38 ± 0.09 mm, n = 44; paired two-sample t-test assuming equal variance: t(43) = -0.98, p = 0.33) were placed in experimental tanks that were identical to the ‘visual access to female’ treatment in Experiment 1. We manipulated the extent to which males in a dyad were able to interact. Dyads were placed in tanks where i) males could interact freely (as in Experiment 1), henceforth called ‘- Refuge’ treatment (n = 24), or ii) the opportunity for males to interactions was experimentally reduced by introducing a refuge, henceforth called ‘+ Refuge’ treatment (n = 20).  Experimental reduction in male interactions was achieved through the addition of a small opaque wall (5cm width) placed in the chamber, which created a refuge, reduced visual contact between the males, and thus reduced the frequency of male-male interactions (see Results).

Male dyads were observed for 20 min on three to four occasions (as in Experiment 1) and social status was determined based on behavioral interactions between the males (as in Experiment 1). Dominance index scores for the dominant males in each dyad ranged from 0.54 to 1 (average ± SE: 0.87 ± 0.02, n = 44). Eight dyads (four in the ‘- Refuge’ and four in the ‘+ Refuge’ treatments) were omitted from the final analysis as, in each dyad, one male did not win at least 70% of agonistic encounters and thus the dominance hierarchy was unclear (scores between 0.54 and 0.7). This reduced the final sample size for analysis to n = 20 for the ‘- Refuge’ treatment and n = 16 for the ‘+ Refuge’ treatment. We present a statistical analysis including these dyads in the supplementary. Comparisons of frequencies of agonistic interactions between treatments only incorporated behavioral data collected on days common to all dyads, i.e. Day 1 and Day 8. Frequencies of male-male agonistic interactions within dyads were recorded as described for Experiment 1 (i.e. gill flare, parallel swimming, chasing, beak-locking, biting).

On the morning of Day 13, males were transferred to isolated tanks and given new anonymous IDs (to ensure experimenters were blind to male social status and treatment). Males were lightly sedated (final concentration 67 mg benzocaine in 1 L diH₂O, where initial stock solution = 100 mg benzocaine to 1 ml Ethanol), stripped of sperm, and sperm swimming speed was measured as described in Experiment 1. As for Experiment 1, we present only VCL in all subsequent analyses, but results were consistent irrespective of the metric used for sperm swimming speed. A subset of the sperm was dyed with a solution of Propidium Iodide (ex503-530/em640) and Acridine Orange (ex536/em600-640) in order to assess sperm viability (Vitaltest, NordicCell, Copenhagen, Denmark). Propidium iodine and acridine orange are fluorogenic compounds binding to nucleic acids and thus staining sperm nuclei. The cell-permeable acridine orange stains both alive and dead cells green, while propidium iodine can only enter cells with a loss of membrane integrity, staining them red. Thus, alive cells are stained green, while dead cells are stained red. In instances where cells were dyed both green and red, meaning that both dyes had entered the cell, cells were classified as dead. The light-sensitive dyes were kept in the dark at all times, as were sperm cells after addition of the dye. After adding the dyes, sperm was briefly mixed, then transferred to a microscopic slide. Cells settled on the slide for about 2 min (average ± SE: 130 ± 2.27 s, n = 65) in the dark. After all cells had settled on one focal plane, images of the cells were taken under a microscope (200x magnification; UB 200i Series Microscope and C13-ON camera, PROiSER R+D, Paterna, Spain). Viability was assessed by counting the proportions between red or green-red (dead) and green (alive) cells. Samples where less than 100 cells were recorded were omitted from the final analysis to ensure reliability of the data (n = 9). From non-fixed samples, images of sperm cells were taken (400x magnification; UB 200i Series Microscope and C13-ON camera, PROiSER R+D, Paterna, Spain) and 20 morphologically normal (i.e. with all sperm components clearly distinguishable) sperm per male were measured to assess sperm morphology differences. Sperm morphology was assessed by measuring the length of the (i) sperm head, (ii) midpiece, (iii) flagellum using ImageJ v1.52r (Schneider et al. 2012) and (iv) total sperm length was assessed by summing respective head, midpiece and flagellum measurements. Lastly, sperm samples were fixed by adding 5% formalin and sperm cells were counted using a microscope and a counting chamber (Neubauer Improved). Samples with low final sperm counts (< 200) were removed from the final statistical analysis (n = 2).

Statistical analysis

Experiment 1: The effect of male access to females on male agonistic behavior was analysed in a linear model with the average frequency of agonistic interactions between male dyads per 20 min (per observational unit) as the response variable, and experimental treatment as the independent variable. Agonistic interactions were square-root transformed to normalise the data and ensure a better fit of the model.

To determine the effects of male social status and male access to females on sperm swimming speed, testes mass, male condition factor, and change in body area during the experiment, we constructed linear mixed effects models with the respective trait as the response variable. Male social status and treatment (male access to females), as well as their interaction term, were treated as independent variables. Dyad ID was included in the model as a random-effect. To prevent overfitting, liver mass was analysed irrespective of dyad ID as a linear model with male social status, treatment, and their interaction term as independent variables. When analysing sperm swimming speed, data was weighted by the number of motile sperm measured for each male to account for variation in measurement number among males (number of motile sperm cells average ± SE: 133 ± 9, n = 63, range 25 - 315). For testes and liver mass, body mass was included as a covariate in the model to account for allometric effects.

Additionally, we tested the relationship between sperm swimming speed and male body condition by constructing a linear model using within dyad differences in sperm swimming speeds (dominant – subordinate male) as the dependent variable and differences in body condition, access to females, and their interaction term as covariates.

Experiment 2: The effect of refuge provision on male agonistic behavior was analysed in a linear model with the average frequency of agonistic interactions between male dyads per 20 min (per observational unit) as the response variable and experimental treatment (‘-’ and ‘+’ refuge) as the independent variable. Agonistic interactions were square-root transformed to normalise the data and ensure a better fit of the model.

Experimental treatments (‘-’ and ‘+’ refuge) were performed sequentially (first ‘- Refuge’ dyads, then dyads of ‘+ Refuge’ treatments), not simultaneously as in Experiment 1. To account for the sequential nature of the treatments, statistical analyses were performed using standardized trait values, where absolute sperm trait values were divided by treatment means (individual value / mean value within the respective refuge, i.e. ‘-’ or ‘+’, treatment group). Comparing standardised values accounts for potential block effects that can emerge from sequential sampling, and thus allows for more appropriate comparisons of treatment effects. Linear mixed effects models were constructed with either average sperm swimming speed, sperm count, or sperm head length as the dependent variable; male social status, treatment (‘-’ and ‘+’ refuge ) and their interaction effect as independent variable; and dyad ID as a random effect. To prevent overfitting, sperm viability, sperm midpiece length, sperm tail length and total sperm length were analysed irrespective of dyad ID as linear models with male social status, treatment, and their interaction term as independent variables. However, results were consistent irrespective of the inclusion or exclusion of dyad ID. As in Experiment 1, sperm swimming speed data was weighted by the number of motile cells recorded to account for variation in the number of sperm measured among males (number of motile sperm cells average ± SE: 320 ± 30, n = 63, range 21 - 989). Sperm count was square-root transformed to fit the assumptions of normality. Likewise, midpiece length was transformed using a standardized  Yeo-Johnson transformation (λ = 5, mean = 6.31, sd = 1.33) (Yeo and Johnson 2000) included in the VGAM package (Yee 2010).

All analyses were completed using R version 4.0.0 (R Core Team 2020). Significant effects were obtained using the ‘Anova’ function included in the ‘car’ package (Fox and Weisberg 2011). Non-significant interaction effects were dropped from final models. Model fit was assessed through visual inspection of the residuals, with outliers being dropped from the final model (i.e. one measurement for frequencies of agonistic interactions in Experiment 1 and one measurement for sperm viability in Experiment 2).