Methods

Copepods were collected from high inter-tidal pools at Aguilar Point, British Columbia (48° 51´ 28” N, 125° 09´ 38” W). Samples from five pools were collected in September 2009 and transported directly to the laboratory at nearby Bamfield Marine Sciences Centre, Bamfield, B.C., Canada. From this initial group, two selection lines were established using truncation selection: male-biased and female-biased. For the male-biased line only females that produced the most male-biased brood sex ratios (BSRs) were allowed to breed in the next generation and for female-biased lines only females that produced the most female-biased BSRs were allowed to breed in the next generation (see Alexander et al. 2014 for complete details). At the end of six generations this resulted in a male-biased line with a mean (± SD) proportion male BSR of 0.64 ± 0.029 and female-biased line with a mean (± SD) proportion male BSR of 0.35 ± 0.031 (Alexander et al. 2014). While in the lab, Tigriopuswere housed in filtered sea water (sea water drawn from 15 m deep in the ocean and filtered through a 1-micron filter), held at room temperature (20 °C) and fed ad libitum a mixture of ground TetraMin® and Spirulina flakes (50:50 by weight) suspended in filtered sea water at a ratio of 0.01 g/mL.
Line crosses were carried out between the two selection lines using offspring produced by the eight females (of >55 sampled) with the most biased brood sex ratio in each of the two selection lines (omitting any families with <6 individuals of either sex). Offspring from these 16 families were mated in an incomplete diallel cross design among four families (within family crosses were not done) from each line, replicated twice (Table 1).

F1 Crosses

Crosses were done by taking individual males and females from the appropriate families and placing them together in the well of a 12-well culture plate filled with 5 mL of filtered sea water (FSW). Males were removed after 7 days, and each female was subsequently checked twice daily for the presence of a mature (red) egg sac. Females with red egg sacs were placed on moist filter paper and the egg sac gently removed using a fine insect pin (Burton 1985). The egg mass was placed into an empty well of a 6-well culture plate filled with 10 mL of FSW. Females were returned to their home well and continued to be checked for an egg sac; for 50 of the 73 crosses a second ripe egg sac was removed, and sex ratio estimates calculated for the second brood. Females typically have anywhere from 3-6 broods; all eggs are fertilized from sperm stored by the female during a single mating (Burton 1985).
When the brood of egg sacs placed in 6-well plates reached the copepodid stage (10-12 days post-hatching), copepodids were isolated from one another by transferring individuals to an individual well in 24-well plate filled with 2.5 mL FSW. At maturity, the number of males and females in each brood was counted to estimate BSR. A total of 71 (of a possible 112) F1 crosses that produced a minimum of 12 offspring for BSR estimate were analyzed.

Backcrosses

F1 cross offspring from 19 of the 34 families with parents from two different selection lines (FM1, MF1, FM2, MF2; Table 1) were backcrossed to male- and female-biased sex ratio parent lines (using the next generation of parental selection lines, i.e., generation 8). Offspring from all but three of these F1 families were used as both sires and dams, generating reciprocal backcrosses. A total of 223 of 280 possible reciprocal backcrosses for the 19 F1 families were done (Appendix Figure 1).
The first mature egg sac of each female in a backcross was plated into a 6-well plate, as above. For some crosses, we also plated a second egg sac. Offspring of these crosses were allowed to mature in 6-well plates and sex ratio determined when the brood matured by anesthetizing all individuals in the brood using 10% MgCl2 (in FSW) prior to counting males and females.

Statistical Analysis

Phenotypic means and variances of brood sex ratio for each cross were calculated. We then tested whether the observed variance in sex ratio differed significantly (under- or over-dispersion) from that of the expected binomial distribution by generating random numbers of males and females in each observed brood (keeping observed brood size), as recommended by Krackow et al. (2002) and modifying code from Postma et al. (2011).
A pedigree analysis (‘animal model’; Wilsonet al. 2010; Postma et al.2011) was used to estimate variance components of sex inTigriopus . We created a pedigree that included all 19,571 individuals sexed and used the R package ‘MCMCglmm’ (Hadfield 2010) to fit a generalized linear mixed model and generate Bayesian posterior distributions for estimating additive genetic variance, variance due to maternal effects, and heritability for sex. Each individual can only have one measure (it is either male or female) and therefore we cannot estimate residual variance and must instead fix it at some arbitrary value in the model (Postma et al. 2011); we used 1, finding this gave us the most well-behaved model (other values led to increased autocorrelation). We started with a model with no fixed effects and animal and dam as random effects, with the link function set as family = threshold. Six chains were run in parallel, each with a burn-in of 100,000, thinning of 5,000 and one million iterations, for a total sample size of 1080. We estimated effective sample size in all Bayesian models run to confirm validity of models. After ensuring chains were well-mixed, data from all chains were combined and the R Package QGglmm (de Villemereuil et al. 2016; de Villemereuil 2018) used to obtain heritability on the observed scale.