Actinopteryiian fishes exhibit the most diverse range of reproductive behaviors found in any vertebrate group (Breder and Rosen 1966; Sadovy de Mitcheson and Liu 2008). Individual species may be unisexual, gonochoristic (separate sexes), or hermaphroditic (sequential or simultaneous), and mating systems range from broadcast group spawning to self and internal fertilization, including monogamy through various degrees of polygamy by either or both genders (Mank and Avise 2006). Courtship and spawning behaviors in fishes, including male alternative reproductive tactics (MARTs), have formed the basis for theoretical and empirical studies examining sexual selection. A significant component of the observed variety in reproductive modes is the level of parental care of offspring. Broadcast spawners typically provide no parental care, whereas nest builders, brooders, and live bearers often expend considerable energy in defending eggs and developing young. Approximately 21% of the 422 recognized taxonomic families of bony fishes contain species that provide some degree of parental care of offspring (Avise et al. 2002), including 15.5% to 22.5% of families in California and Pacific Canada (DeMartini and Sikkel 2006). Unlike in other vertebrates, about 70% of the parental care is provided by males in fish species that exhibit uniparental care (Blumer 1982; Avise et al. 2002).

In recent years, parentage analysis of offspring using highly polymorphic microsatellite DNA markers has provided unprecedented resolution of fish mating systems (DeWoody and Avise 2001; see review by Avise et al. 2002; Mackiewicz et al. 2005). These behaviors include monogamous or multiple matings, male cuckoldry, and a variety of MARTs to attain surrogate parentage, such as egg mimicry, nest piracy and egg-thievery, multiple paternity in female-pregnant species, and sex role reversal in male-pregnant species (e.g., pipefish, seahorses). Parentage analysis has also revealed the presence of MARTs that would probably go unobserved or unrecognized in many field studies and provides a means to quantify fitness consequences, in terms of reproductive output, in natural populations. Other behaviors in complex mating systems, such as patterns of egg or offspring cannibalism, can be readily assessed using parentage analysis.

The hexagrammid Atka mackerel Pleurogrammus monopterygius is a semidemersal, schooling species distributed along continental shelves from Asia to North America. Atka mackerel are obligate demersal spawners that attach egg clutches to hard substrates (often in crevices) that are protected from cannibals and other egg predators by territorial males. Males aggregate and acquire nest sites before spawning at depths from 32 to 144 m (Lauth et al. 2007). Nesting typically occurs on rocky substrates in areas exposed to moderate currents, and there is considerable variation among locations with respect to relief, slope, and continuity (Lauth et al. 2006). Nest sites average about 2 m² (Fritz and Lowe 1998) and tend to be adjacent to one another (Lauth et al., 2006). Atka mackerel develop conspicuous seasonal sexual dichromatism, and both sexes display ephemeral color changes while engaged in seasonal courtship and territorial defense behaviors (Lauth et al., in press). Females can potentially spawn up to six adhesive egg clutches during the spawning season, each containing approximately 6,700 eggs (McDermott et al. 2007), and males may brood multiple clutches within their territories. Spawning typically begins in late June, and females in spawning condition have been documented from July through October (McDermott and Lowe 1997). Nesting males have been observed in late October, suggesting that the combined mating and brooding phases may extend from summer well into the fall (Lauth et al. 2007).

Specific details of the mating system in Atka mackerel are unknown, but probably include some of the polygamous strategies found in other greenlings. Sequential polyandry (either within or among spawning seasons) by female hexagrammids has been documented by parentage analysis in several species (Crow et al. 1997; King and Withler 2005). Crow et al. (1997) reported that about 40% of the nests of kelp greenling Hexagrammos decagrammus contained clutches from multiple females, and DeMartini (1987) reported similar proportions for painted greenling Oxylebius pictus. Less well understood are MARTs, which may include adoption of a “sneaker” strategy by adjacent nest-tending males or nonterritory holding males to engage in parasitic fertilizations. Sneaking has been reported in hexagrammids species (personal communication cited in Crow et al. 1997; Munehara et al. 2000), and the reproductive consequences of this behavior have been quantified using molecular genetic techniques in the fat greenling H. otakii (Munehara and Takenaka 2000) and lingcod Ophiodon elongatus (Withler et al. 2004; King and Withler 2005).

Nest raiding of eggs by both sexes and male filial cannibalism have been reported in other hexagrammids (DeMartini 1987; Munehara and Miura 1995), and adult Atka mackerel are egg cannibals (Yang 1999; Rand 2007). In species with paternal care, male filial cannibalism parasitizes the fecundity of females as a strategy to maximize lifetime reproductive success at the expense of current broods (Rohwer 1978; Sargent 1992) and may benefit females if it increases the probability of successfully rearing the remaining offspring (Lindström 2000). Male energy reserves have often been considered to be the primary factor driving filial cannibalism (DeMartini 1987; Lindström 2000), and male condition may vary seasonally, depending upon the duration of parental care. The extended breeding and brooding period in Atka mackerel probably imposes significant energy demands upon territorial males, but patterns of cannibalistic behaviors in Atka mackerel have not been investigated.

In this study, we used genetic parentage analyses of embryos produced in both captive and natural populations of Atka mackerel to assess the mating system and patterns of egg cannibalism by adults. Genotype data from polymorphic DNA microsatellite loci were obtained for embryos in egg clutches (defined as an event of egg deposition by females) to document mating behaviors in captive populations through parentage analyses. Additional inferences on aspects of the mating system were derived from analyses of clutches collected in the field or retrieved from the stomachs of adult cannibals, where parental data were unavailable. Results from natural populations were contrasted with those observed in captivity to examine the full range of reproductive behaviors in Atka mackerel and the predominance of these behaviors relative to other members of the family. A second objective was to assess the extent of heterocannibalism and filial cannibalism by both genders. The relatedness of male and female cannibals with the embryos they had consumed was determined using microsatellite DNA markers, providing the first quantitative assessment of gender-specific cannibalism in this species.

Methods

Samples from captive populations

Parentage was assessed for embryos in clutches produced in captivity from two separate aquaria: a small tank (about 10,250 L, 1.5 m deep) and a large tank (about 400,000 L, 6.4 m deep) at the Alaska SeaLife Center, where adult Atka mackerel had successfully spawned for several years (Table 1). The small tank contained two males and 4 or 5 females in 2004–2006; the tank had 4.0 m² of spawning substrate available in 2004 and 7.2 m² in 2005 and 2006. The large tank contained 13 females and 8 males in 2005 and 11 females and 6 males in 2006; this tank had more than 50 m² of spawning substrate available in both years (Table 1). Fin clips from all adults were collected for parentage analysis, and the identities of territorial males were recorded based upon natural markings.

Eggs were removed from the small tank soon after first discovery to deter cannibalism, and the time, nest location, and identity of the guardian male were recorded. During 2004 and 2005 eggs were incubated at controlled temperatures ranging from 3.9°C to 9.9°C. Samples for genetic analysis were taken in middle to late stage development when eyes were readily visible: approximately 4–10 weeks following fertilization, depending upon temperature. We subsequently found that we could determine parentage of embryos at much earlier stages of development. Thus, clutches were removed from nesting sites after first discovery in 2006 and incubated at 4–6°C for approximately 2 weeks before preservation in 95% nondenatured ethanol. In 2005 and 2006 only three to six embryos were genotyped per clutch to establish maternity of clutches used in study of fecundity and egg energy content (S. McDermott and colleagues, National Marine Fisheries Service, unpublished).

During the 2005 season, Atka mackerel were allowed to spawn undisturbed in the large tank from the onset of spawning in July until the end of August. Four of the eight resident males in the tank successfully brooded clutches that season. A first group of clutches from two territories, each guarded by a single male throughout the season, were removed on the same date. They were again allowed to spawn undisturbed for 12 d, when a second group of clutches was removed. Spawning continued for an additional 38 d before the final group of clutches was removed. Once removed, all eggs were incubated to the eyed stage as described above. In 2006 we allowed adult Atka mackerel to spawn undisturbed for nearly the entire spawning season to evaluate the effects, if any, of immediate egg removal on reproductive behaviors. Clutches were collected from a single territory held by one of the males sampled during the previous year.

Samples from wild populations

Adult Atka mackerel were collected by trawling between Unimak Island and Amchitka Island, Alaska, during three cruises during 2004–2007 (Table 1). Egg cannibalism was deduced from the distended appearance of the stomach during routine dissections for sex determination. Fish were immediately frozen whole on the vessel and maintained at −20°C until thawed for analysis in the laboratory. Stomach contents were dissected from putative cannibals, and partial clutches were counted, briefly washed in seawater, and preserved in nondenatured 95% ethanol for parentage analysis. Six intact clutches were collected by bottom trawl aboard a chartered commercial bottom trawler in the Aleutian Islands (52.046°N, 172.025°W) on October 13, 2006 (McDermott, unpublished), immediately frozen at −20°C, and subsequently preserved in ethanol.

Extraction, amplification, and analysis of DNA

Genomic DNA from samples of eggs and fin clips (from captive adults and field-collected cannibals), was extracted using DNeasy tissue kits (Qiagen Inc., Valencia, California) according to the manufacturer's instructions with one modification for egg samples: owing to the small size of developing embryos, an elution buffer volume of 60 μL was used to provide a more concentrated DNA template. Four polymorphic microsatellite loci (Pmo70, Pmo152, Pmo367, and Pmo399) isolated from Atka mackerel (Spies et al. 2005) and two loci isolated from lingcod (Oel42 and Oel32; S. Young, Washington Department of Fish and Wildlife, Olympia, Washington, personal communication) were amplified via the polymerase chain reaction (PCR), as described in Spies et al. (2005). Bovine serum albumin (New England Biolabs, Ipswich, Massachusetts) was added to a final concentration of 0.5 mg/mL in PCR reactions involving embryos. Loci were amplified and scored two at a time (following order: Pmo152, Pmo70, Pmo399, Oel32, Pmo367, and Oel42) on eggs randomly selected from clutches of wild and captive populations until sibling relationships could be accurately assessed or, in the case of cannibalism, the cannibal was excluded as a parent. Thus, all loci were not amplified in all samples if fewer loci were sufficient to meet criteria for unequivocal parental identification (captive samples) or determination of sibship relationships. Genotyping was conducted using a LI-COR 4200 Infrared Automatic DNA Sequencer (LI-COR Biotechnology, Lincoln, Nebraska) and analyzed with LI-COR Saga genotyping software.

Parentage analysis

Binomial sampling theory predicts that a sample of n individuals will detect the proportion (p) of offspring contributed by an individual parent with 95% certainty if

(DeWoody and Avise 2001). Thus, genotyping 96 individuals had the power to detect embryos produced by other parents if their contributions to a clutch exceeded 3% and to detect a contribution of 12% or more by sampling 24 embryos. Based upon these calculations, we initially performed exhaustive sampling of 96 embryos from clutches produced in the large tank and subsequently altered the sampling regime based upon those results. Six of the 17 clutches produced in 2005 were exhaustively sampled (≥90 eggs) and the remaining sample sizes exceeded 20 in all but four cases. Genotypes of all possible parents were known for all loci in captive fish, allowing for unambiguous parental assignments by simple exclusion. Parentage in the small tank in 2004 was determined by genotyping 24–96 embryos per egg mass. In subsequent years (2005 and 2006), only three to six embryos were sampled from each clutch to identify the female parent.

Incompatibility of genotypes from ingested embryos provided conclusive proof of nonparentage for egg cannibals. Microsatellite data from 553 adult Atka mackerel collected during 2004–2006 in the Aleutian Islands (Canino et al. 2010) were used to estimate allele frequencies at the six loci used for parentage analysis. Exclusion probabilities (the expected proportion of offspring for which the cannibal would be excluded as a parent) were determined using a frequency-based method (Jamieson and Taylor 1997). Because multilocus genotpyes of cannibals could be determined, we calculated exclusion probabilities using the general formula for one unknown parent. We also calculated the probability of identity, P_ID (i.e., the probability that two different individuals have identical genotypes), for each locus using the formula provided by Waits et al. (2001),

where p_i is the frequency of the ith allele at a locus. The P_ID value over six loci was calculated as the product of probabilities for individual loci.

In general, we attempted to genotype a minimum of 24 individuals at four loci from each clutch to determine sibling relationships and conducted more exhaustive sampling when preliminary results indicated the presence of half-sibs or unrelated individuals. Construction of full-sib and half-sib families detected in cannibalized and field-collected clutches (parental genotypes unknown) was conducted using PEDIGREE 2.2 (available at  http://herbinger.biology.dal.ca:5080/Pedigree/) and subsequently verified by visual inspection.

Two regions of mitochondrial DNA were sequenced to screen for variation. A 450 base pair (bp) segment of the mitochondrial D-loop region was amplified via PCR in 20 individuals from across the species range (Gulf of Alaska to northern Japan) via Pro-L and H16498 primers previously used for successful detection of maternal haplotypes in kelp greenling (Crow et al. 1997) and lingcod (Withler et al. 2004). Next, a 790-bp segment of the cytochrome b gene was amplified using Cytb-F and Cytb-R primers following (Kimura et al. 2007) in nine individuals. The PCR amplicons were purified and sequenced at the High-Throughput Genomics Unit at the University of Washington ( http://www.htseq.org).

Results

No sequence variation was found among 20 individuals in the mitogenomic D-loop region or for cytochrome b in nine individuals (data not shown). We concluded that haplotype variation was likely to be at frequencies too low to be informative for maternal identification in clutches, a conclusion that was subsequently verified by additional D-loop sequencing (see Canino et al. 2010). Parentage analysis was thus conducted using only microsatellite DNA loci.

The six microsatellite loci had 12–92 alleles each, resulting in single locus exclusion probabilities ranging from 0.424 to 0.873 and a combined multilocus exclusion probability exceeding 0.999 (Table 2). Even when only two loci were used (Pmo70 and Pmo152), the exclusion probability was greater than 0.97. The P_ID values for the six microsatellite loci ranged from 0.258 to 0 0.310, providing an overall value of 0.0005 across all loci. These highly polymorphic loci provided sufficient power to discriminate among closely related individuals, enabling us to determine genetic identities for parental assignment or exclusion.

Discussion

Molecular assessment of the mating system in Atka mackerel enabled documentation of behaviors previously inferred from field studies and confirmed reproductive tactics characteristic of other hexagrammid species. The vivid dichromatism of males and ephemeral color and behavior displays by both sexes indicate a highly dynamic mating system probably shaped by strong intrasexual and intersexual selection. Conspicuous nuptial coloration is selected by females as a conditional expression of the genotypic and phenotypic quality of males, as well as their nutritional condition, social rank, and mating motivation and success (Kodric-Brown and Brown 1984; Kodric-Brown 1998). Within a given brood year, approximately 15% of adult males display nuptial coloration, whereas the remaining nonbreeding males make no apparent energetic investment. These seasonal patterns appear to be unrelated to male length and thus are unlikely to represent age-dependent alternative reproductive tactics (e.g., Alonzo et al. 2000), but they may reflect a large energetic cost imposed by sexual selection. Carotenoid-based pigmentation in many fishes must be obtained from the diet and is energetically costly for males to produce and further work should examine patterns of annual and seasonal reproductive status. Ephemeral color displays during courtship and agonistic encounters described for the permanently dichromatic painted greenling (DeMartini 1985) and seasonally dichromatic Atka mackerel (Lauth et al., in press) indicate a complex social environment during spawning determined by strong competition for access to territories and mates in males and by readiness to spawn in females.

Atka mackerel exhibited a polygynandrous mating system, resulting from serial polyandry and polygamy within the breeding season. Male polygyny appeared to be achieved primarily through territorial defense in captive populations, although some cuckoldry of guardian males did occur in the small exhibit tank and probably to a greater degree in natural populations. Polyandry in Atka mackerel results from batch spawning of eggs throughout the breeding season (McDermott et al. 2007) coupled with low nest site fidelity by females. Females in captivity sometimes exhibited male preference for mating, although sample sizes were inadequate to assess this behavior. Most of the clutches produced in natural populations (69%) and nearly all of those produced in captivity (98%) indicated monogamous pairings. Nest-guarding males in the large exhibit tank apparently monopolized all fertilization events within their territories. Two documented instances of parasitized fertilizations occurred in the small tank (Table 3). The average number of embryos genotyped per clutch in partial clutches consumed by cannibals and whole clutches taken in trawls (32.5 individuals) would have only detected contributions from extra-pair matings exceeding 10%, but more exhaustive sampling of several clutches indicated that strictly monogamous spawning events are common in natural populations.

Parentage analysis using microsatellite DNA markers unequivocally documented parasitic fertilizations of clutches by males in captivity, but the responsible behaviors were not directly observed. Male cuckoldry is common among fishes exhibiting paternal care and several alternative reproductive tactics have evolved in response to competition for access to mates or their gametes (Taborsky 1994, 2001; Gross 1996). The most common MARTs documented in nest-tending species by parentage assessment are fertilizations by sneaking or satellite males, resulting in the nest-tending male being the sire of most, but not all of his custodial offspring (DeWoody and Avise 2001: Avise et al. 2002; Mackiewicz et al. 2002; Mackiewicz et al. 2005). Sneaking constitutes potential costs to nest-tending males in terms of lowered reproductive output (Rico et al. 1992; Largiader et al. 2001) and the energetic expense of rearing unrelated young. This strategy is associated with strong sexual dimorphism and active courtship and nest defense behaviors by males (Gross 1996; Taborsky 1998), characteristics common to hexagrammids in general. Sneaking has been observed (Munehara et al. 2000) in three species of Hexagrammos and verified using molecular genetic analyses in the fat greenling (Munehara and Takenaka 2000). Sneaking is a likely, but unobserved, source of male cuckoldry in Atka mackerel. The frequency of clutches produced in captivity by monogamous pairings of females with territorial males was lower in the small tank containing limited spawning substrate than in the large tank where larger amounts of spawning substrate were available. Density-dependent sneaking has been reported in the bitterling Rhodeus sericeus (Reichard et al. 2004). Our experimental design may have favored territoriality when densities were low in the captive tanks and prompted sneaking or other fertilization strategies when they were higher. Another potential MART that may explain parasitized fertilizations is sperm competition from satellite males. Munehara and Takenaka (2000) observed two patterns of sperm emission in male fat greenling and different rates of fertilization success. When the satellite intruder male emitted a large amount of sperm after ejaculation by the territorial male, the sneaker fertilized more eggs. If both males emitted sperm alternately, the territorial male sired a larger number of offspring in the clutch.

Nest-tending males may tolerate fertilizations by nonterritorial males if they help to attract females or aid in nest defense or maintenance (Taborsky 2001). The mixed paternity observed in two single clutches of eggs from the small exhibit tank in 2004 and 2006 resulted from parasitized fertilizations, although sneaking by males was not directly observed. Discrepancies between the apparent guardian and the genetic sire of at least some embryos occurred in all 3 years, suggesting the possibility of additional MARTs. One tactic may be temporary nest takeovers by the nonterritorial male to acquire matings. This may increase reproductive fitness if females prefer males already brooding clutches in the nest (Marconato and Bisazza 1986; DeMartini 1987; Kraak and Groothuis 1994; Goulet 1998) or if there is mate-choice copying by females (Jamieson 1995). Cuckolded males may later return to the nest to care for unrelated offspring if alloparental care improves the possibility of additional spawnings. This potential MART has been documented genetically in other fishes (DeWoody et al. 2001), including lingcod (Withler et al. 2004). Nest takeovers are consistent with the observation that nest site fidelity for male Atka mackerel may not extend for the entire spawning season. Lauth et al. (2007) inferred duration for male nesting and brooding of 33–141 d, based upon archival tag data for four adults. It is unknown if the cessation of nesting represents expulsions by other males or voluntary abandonment of a site.

The presence of unrelated half-sib and full-sib embryos found within single putative clutches from natural populations (Table 5) was a surprising result that may, in part, be explained by egg deposition patterns by females and MARTs inferred from results from captive populations. Previous studies have considered a clutch to be the reproductive output from an individual female—an assumption (DeMartini 1986) inferred from the physical discreteness, yolk coloration, and stage of embryonic development in kelp greenling and whitespotted greenling Hexagrammos stelleri and documented using genetic studies in kelp greenling (Crow et al. 1997) and lingcod (Withler et al. 2004). Although this also appeared to be the general case for Atka mackerel in captivity, some clutches produced in wild populations indicated that females may deposit their eggs on existing clutches, fusing them to each other (Figure 1). In the congeneric Okhotsk Atka mackerel Pleurogrammus azonus, females used their pectoral fins or side of their belly to wedge fertilized clutches into gaps between rocks from various directions (Munehara and Markevich 2003), and female captive Atka mackerel have been observed to push freshly fertilized eggs into crevices using their mouths (J. Guthridge, unpublished data). Koya et al. (1995) reported that the adhesive surrounding the eggs of the fat greenling required several hours to harden, and these behaviors may help to retain clutches after deposition until they fully adhere to the spawning substrate. The extreme mitogenomic homogeneity exhibited by Atka mackerel (Canino et al. 2010) precluded using mitochondrial DNA haplotypes to determine individual female contributions to clutches retrieved from cannibals or captured in trawls, but the probable sources of complex assortments of related and unrelated offspring within a single clutch are multiple egg depositions and manipulations by females, accompanied by MARTs, such as nest takeover or parasitized fertilizations.

Egg cannibalism constitutes a significant dietary component of Atka mackerel during the spawning and brooding seasons (Yang 1999; Rand 2007). Both genders engage in cannibalism in roughly equal frequencies, but males consume more than twice as much by weight (K. Rand, National Marine Fisheries Service, personal communication). Unlike territorial males, females can forage during the mating and brooding phases in seasonal reproduction and are generally not in proximity of defended nests unless engaged in spawning. Female fat greenling have been observed pecking at egg clutches already present in nests, causing attendant males to repel them without attempting to mate (Munehara et al. 2000). Nest raiding by female Atka mackerel has been observed (R. Lauth, National Marine Fisheries Service, personal communication), suggesting it is the most likely source for opportunistic heterocannibalism by this gender.

Filial cannibalism represents a range of compromise strategies between current and future reproductive success (Rohwer 1978) and is common for males in species with uniparental care. DeMartini (1987) reported apparent heterocannibalistic raiding of nests and some paternal filial cannibalism by guardian male painted greenling. Filial cannibalism has also been observed in fat greenling, when clutches detached from seaweeds to which they had been fastened and were no longer under direct male control (Munehara and Miura 1995). Males may cannibalize some fraction of their offspring as an investment to maximize future reproductive success by enhancing their own survival and that of the remaining brood, whereas females can only benefit from partial, but not total, consumption of their young by males. This results in a conflict between the sexes over the male's investment in the current brood (Lindström 2000). Empirical studies have shown that small and young broods are more often fully cannibalized than larger and older ones that have greater paternal investment by the male (DeMartini 1987; Lindström and Sargent 1997). That was not the case in the single instance of filial cannibalism documented in this study; all of the embryos in the partial clutch were in a later eyed stage. However, most were unrelated to the cannibal, and his relationship to the clutch (e.g., guardian or opportunistic cannibal) was unknown.

Although the parental energy reserve of the male has often been considered as the primary factor in filial cannibalism, mate availability may also be important (Kondoh and Okuda 2002). Males may receive an energetic surplus of eggs when females are abundant, resulting in an energy reserve that can be reallocated to future reproduction. Conditions favoring mate availability include a female-biased sex ratio, an extended period of male care, and a short refractory period between spawnings by females (Kondoh and Okuda 2002).These conditions may be met seasonally, to some degree, in Atka mackerel. The occurrence of adult nonbreeding males in the population (Lauth et al., in press) biases the operational sex ratio (ratio of ready-to-mate males to ready-to-mate females; Emlen and Oring 1977) towards females. Average realized batch fecundity of females is 4.6 clutches per season (McDermott et al. 2007), resulting in clutch production perhaps every 2–3 weeks. The spawning and brooding period for males can last from late July to January (Lauth et al. 2007), and the sole instance of filial cannibalism detected in our study occurred when the spawning phase of seasonal reproduction was largely over (October) and some of the brooding phase remained. Although more extensive sampling would be required to characterize the extent of filial cannibalism by males, the prevalence of heterocannibalism we observed suggests that densities in communal nesting areas may be sufficiently high to make conspecific nest raiding a more common strategy.

Our results underscore both the power and limitations of genetic parentage assessment alone for making inferences about reproductive behaviors and their consequences in natural populations. The study confirmed some general aspects of the mating system in Atka mackerel (e.g., serial polygamy and polyandry) inferred from existing studies on the family but did not resolve more complex behaviors (e.g., MARTs and egg deposition patterns by females) contributing to the clutch genetic diversity observed in wild populations. Future efforts should include pairing observational and parentage approaches in studies of captive and wild populations, that would more fully elucidate the range of reproductive behaviors in both genders.