Spatial and temporal genetic variation in an exploited reef fish: The effects of exploitation on cohort genetic structure

Abstract Many coral reef fishes are fished, often resulting in detrimental genetic effects; however, reef fishes often show unpredictable patterns of genetic variation, which potentially mask the effects of fishing. Our goals were to characterize spatial and temporal genetic variation and determine the effects of fishing on an exploited reef fish, Plectropomus leopardus, Lacepède (the common coral trout). To determine population structure, we genotyped 417 Great Barrier Reef coral trout from four populations sampled in 2 years (1996 and 2004) at nine microsatellite loci. To test for exploitation effects, we additionally genotyped 869 individuals from a single cohort (ages 3–5) across eight different reefs, including fished and control populations. Genetic structure differed substantially in the two sampled years, with only 1 year exhibiting isolation by distance. Thus, genetic drift likely plays a role in shaping population genetic structure in this species. Although we found no loss of genetic diversity associated with exploitation, our relatedness patterns show that pulse fishing likely affects population genetics. Additionally, genetic structure in the cohort samples likely reflected spatial variation in recruitment contributing to genetic structure at the population level. Overall, we show that fishing does impact coral reef fishes, highlighting the importance of repeated widespread sampling to accurately characterize the genetic structure of reef fishes, as well as the power of analysing cohorts to avoid the impacts of recruitment‐related genetic swamping. The high temporal and spatial variability in genetic structure, combined with possible selection effects, will make conservation/management of reef fish species complex.

| 1287 TABOUN eT Al. they typically have a narrow thermal tolerance range (Tewksbury et al., 2008), and slight increases in temperature and CO 2 concentrations can reduce performance Pratchett et al., 2017). Furthermore, it is likely that harvesting can reduce resilience of populations to the stress of climate change (Planque et al., 2010).
For example, physiological states are related to vulnerability to exploitation (discussed by Lennox et al., 2017), and specifically, highperformance aerobic scope phenotypes may be targeted by fishing (Duncan et al., 2019). Given the susceptibility of reef fishes to the combined stressors of climate change and fishing pressure, it is likely that those stressors also impact the genetic composition of individuals and populations.
Dispersal is an important coral reef fish life-history trait resulting in population connectivity, an important factor in coral reef management plans (Almany et al., 2017;Planes et al., 2009;Roberts, 1997). Population genetic studies in marine species show typically high gene flow (Clarke et al., 2010;Unsworth et al., 2008;Waples, 1998), also true for coral reef fishes (Mora & Sale, 2002;Williamson et al., 2016). Although adult coral reef fish tend to be sedentary after settlement (Sale, 1991), early pelagic life stages are highly dispersive (Hepburn et al., 2009). However, genetic differentiation, despite high gene flow, has been reported in reef fishes; for example, Salas et al. (2010) found subtle genetic structure in Stegastes partitus (bicolor damselfish) within Costa Rica-Panama reefs.
Additionally, Hogan et al. (2012) found evidence of local retention in the S. partitus (bicolor damselfish). Similarly, Taylor and Hellberg (2003) reported strong genetic differentiation in populations of Elacatinus evelynae, a cleaner goby, despite a larval pelagic duration of 21 days. Nevertheless, high levels of dispersal pose challenges for studying population differentiation within coral reef fishes, as new recruits cause a mixing of genotypes after each reproductive cycle (Hepburn et al., 2009). Given that many reef fishes remain on reefs post-settlement (Sale, 1991), individual cohorts should display higher genetic structure. Thus, cohort-level analyses allow genetic diversity and divergence of 'populations' to be estimated with reduced interference from gene flow, with the expectation that as the cohorts age, the effects of selection and drift should accumulate and become more pronounced.
Although reef fishes generally show limited genetic structure, they can exhibit unpredictable spatial and temporal patterns, perhaps reflecting stochastic larval dispersal and variable larval retention (Hepburn et al., 2009;Hogan et al., 2012). Some marine species exhibit genetic structure that does not adhere to population genetic models, a phenomenon referred to as chaotic genetic patchiness, first described by Johnson and Black (1982). This phenomenon has been observed in various species, including Acanthaster planci (Crown-ofthorns starfish; Nash et al., 1988), sea urchins (Strongylocentrotus purpuratus, see Edmands et al., 1996;Echinometra mathaei, see Watts et al., 1990), Spisula ovalis (a clam, see David et al., 1997), S. partitus (the bicolor damselfish; Hogan et al., 2010;Lacson & Morizot, 1991), among others. The variable and unpredictable spatial and temporal genetic structure often reported among marine fish populations highlights the importance of characterizing patterns of genetic structure over time to inform conservation and management actions with respect to a population's evolutionary potential and hence their susceptibility to environmental and exploitation stress.
Standing genetic variation is important in evaluating the evolutionary potential of populations (Barrett & Schluter, 2008).
Populations that harbour high genetic diversity are typically better able to respond to selection because there is an increased chance that the favoured genotypes are present within the population.
However, genetic variation can be reduced through exploitation (Allendorf et al., 2008), and this can negatively impact the evolutionary potential of harvested populations. For example, Hauser et al. (2002) used scale-extracted DNA to show a significant decline in neutral genetic diversity in Pagrus auratus (New Zealand snapper) associated with aggressive harvesting. Losses of genetic diversity due to fishing pressure have also been reported for Gadus morhua (North Sea cod; Hutchinson et al., 2003), Diplodus sargus (white seabream; Pérez-Ruzafa et al., 2006), Hoplosthethus atlanticus (orange roughy; Smith et al., 1991) and Rexea solandri (gemfish; Ovenden et al., 2020). Therefore, combined stressors, such as fishing pressure and climate change, may erode the evolutionary potential for populations to adapt, ultimately leading to population decline and possible extirpation.
While marine fish population genetic structure is expected to vary over space and time, individual species also present specific challenges for applications of conservation and management genetics. For example, exploited reef fishes that are experiencing climate change effects but also undergo sequential hermaphroditism may be the most vulnerable to the genetic consequences of harvesting because size-selective fishing removes larger (and thus sex-specific) individuals and potentially those with adaptive phenotypes (Duncan et al., 2019). The common coral trout, Plectropomus leopardus, is an iconic species found on Indo-Pacific reefs and undergoes protogynous sequential hermaphroditism . This species is the most economically important harvested fish on the Great Barrier Reef (GBR; Leigh et al., 2014). Additionally, Plectropomus species are a top predator, and thus, impacts on these species may have cascading effects (Graham et al., 2006). It is thus essential that we characterize the genetic structure of this species on the GBR and integrate these data into management and conservation plans (Palumbi, 2003).

Although there have been published population genetic studies on
Plectropomus, research on this species is limited . Williamson et al. (2016) used microsatellite marker-based parentage analyses to examine larval dispersal within P. leopardus and found dispersal can be quite extensive (median dispersal distances = 190 km).
Such high dispersal potential is consistent with the lack of genetic structure reported in the same study . Van Herwerden et al. (2009) found that some P. leopardus populations have meta-population genetic structure and dynamics. Furthermore, patterns of recruitment can vary through space and time; for example, Russ et al. (1996) found one cohort of P. leopardus dominated the population on two marine reserves on the GBR for at least 3 years.
The economic and ecological importance of this fish, coupled with its particular vulnerability, highlights the need to define patterns of gene flow, genetic structure, and genetic diversity, and how these may be affected by fishing pressure.
The genetic consequences of exploitation in the common coral trout remain unknown, despite its economic and ecological value.
In this study, we first use microsatellite data from four unexploited populations to characterize spatial and temporal genetic variation in genetic diversity and structure. Secondly, we sampled a single cohort of fish over 3 years (ages 3-5) from eight reefs that differ in fishing pressure and tested for the effects of exploitation on genetic diversity and structure. We found chaotic spatial and temporal genetic structure makes repeated and widespread sampling critical when characterizing genetic structure in species such as P. leopardus. However, analysing individual cohorts reduces the effects of variable recruitment, making it possible to assess the effects of fishing pressure on genetic diversity and structure -a critical need for effective management and conservation. Our data also highlight possible changes in adaptive potential of populations that are facing pressure from exploitation on a backdrop of climate change.

| Sample collection
The GBR, located on the eastern coast of Australia, is the world's largest coral reef ecosystem ( Figure 1). As part of a large GBR project designed to test the effect of line fishing on GBR coral reef fishes, coral trout were collected from multiple reefs across much of the GBR (Mapstone et al., 2004). To reduce the introduction of potential sampling biases, times of peak spawning activity were avoided for the collection of the fish, as these periods could lead to unintentional sampling of spawning aggregations unassociated with true population genetic signals (Mapstone et al., 2004). Furthermore, sampling was standardized spatially and temporally on each reef (see Mapstone et al., 2004, for additional details on sampling). Therefore, it is likely that our samples are representative of the local population.
The fish were collected using line fishing from reefs located in four regions: Lizard Island, Townsville, Mackay and Storm Cay ( Figure 1a).
Additionally, fish were sampled at eight reefs within two sub-regions for a targeted fishing effects analysis; the reefs were categorized as Green (closed to fishing), Blue (open to fishing and subjected to pulse fishing in 1999) and Manipulated (closed to fishing but subjected to pulse fishing in 1999; Figure 1b). Pulse fishing included the maximum level of fishing possible without incentive for 12-month periods by commercial and recreational fishers, as described in Mapstone et al. (2004). Reefs classified as 'Green' act as controls that provide measures of natural changes in genetic structure. Both the Blue and Manipulated reefs were treated with the same experimental pulse fishing schedule and allowed us to assess the fishing effects on populations compared to the Green controls. Otoliths were collected from all fish; however, age was determined only for the fish in the cohort genetic analysis (see below). The age of the fish was estimated as described in Mapstone et al. (2004) through otolith analyses. DNA was extracted from the otoliths following the protocol of Heath et al. (2007).
Two types of DNA analyses are included in this study, constituting two related, but separate analyses: (1) spatial/temporal genetic analyses and (2) cohort effects of fishing genetic analyses. The spatial/temporal samples include individuals obtained from four reefs, closed to fishing, sampled in 1996 and 2004 (Table 1 and Figure 1a).
These samples allowed us to characterize spatial and temporal patterns of genetic variation at the population level. DNA samples included in the cohort study include individuals from a single cohort sampled over 3 years (1998-2000, ages 3-5) from eight reefs that differ in fishing pressures (see Table 1, Figure 1b and Table S1). The reefs in the cohort study are located in two regions, each region had four reefs: two Green, one Manipulated and one Blue (Figure 1b).
Since long-range and inter-reef movements are relatively rare in coral trout (Matley et al., 2015;Sumpton et al., 2008), the cohort study allowed us to assess random and selection-based genetic changes occurring in these populations over time. We also tested for fishing pressure effects with considerable power due to decreased interference from new recruits.

| Genotyping
Multiplex PCRs of 16 microsatellite loci were performed in 7.0 µl reactions (See Table S2  and 10 µl of diluted first-round PCR product. Thermocycling conditions used for the ligation PCR were as follows: 94°C for 2 min, six cycles of 94°C for 30 s, 30 s at 60 and 72°C for 60 s, followed by 72°C for 5 min. The barcoded products were pooled, run on an agarose gel, and the bands were excised and cleaned using GenCatch Gel Extraction Kit (Epoch Life Science Inc.). The pooled amplicons were assessed on a 2100 Bioanalyzer (Agilent Technologies), diluted to 60 pM and sequenced using an Ion Torrent NGS personal genome machine (Thermo Fisher Scientific). We also sequenced negative controls from the first PCR.
We used QIIME (Caporaso et al., 2010) to demultiplex sequence data by individual sample code and to separate sequences based on primer sequences. We used NGS-usat software (Roy et al., 2021) | 1289 to genotype individuals. This R-based software counts the number of microsatellite repeats within each sequence read and generates frequency-based plots of read number versus repeat number.
Automatic genotypes were generated, but alleles were visually verified and adjusted as necessary without knowledge of population assignment to reduce the potential for introducing biases. To determine F I G U R E 1 Maps showing the sampling sites for the coral trout used in this study. Panel (a) Map of the Australian Great Barrier Reef showing the sample sites for coral trout reefs included in spatial/temporal study (Table 1). Reef IDs from north to south: 14147, 18071, 20142 and 21132. Panel (b) Map of eight study sites within two spatial regions located on the Great Barrier Reef for coral trout populations included in cohort study (Table 1) the minimum sequencing depth required to reliably genotype individuals, we determined the number of reads at which heterozygosity estimates reached a plateau, which determined the minimum number of reads required for genotyping. We only included marker loci with five or more alleles across all samples and that were genotyped in 15 or more individuals per population for further analyses.

Population genetic diversity
Unless otherwise stated, statistical analyses were conducted using R Core Team (2019). We first characterized genetic diversity within populations by calculating the number of alleles, observed heterozygosity, expected heterozygosity and Hardy-Weinberg Equilibrium (HWE) deviations using R package 'diveRsity' (Keenan et al., 2013).
To verify that each locus was independent, we tested for linkage disequilibrium for locus pairs in each population and year using 'genepop' v1.1.7 R package (Rousset, 2008 (Rousset, 2008).

Population genetic divergence
We estimated global F ST to quantify differentiation among subpopulations for both sample years separately using an analysis of molecular variance (AMOVA) with significance tested over 9999 permutations implemented in GenoDive v.3.04 (Meirmans, 2020).
We also used GenoDive v.3.04 (Meirmans, 2020)  isolation was a function of distance among reefs, we tested for isolation by distance (IBD) using the 'adegenet' package in R (Jombart, 2008), which creates matrices of both pairwise genetic (Reynold's distance; Reynolds et al., 1983) and geographic (Euclidean distance) distance based on genotypes and reef coordinates and performs a Mantel test.

Population genetic diversity
To assess the impacts of fishing pressure on genetic diversity in the cohort study populations, we used the R package 'diveRsity' (Keenan et al., 2013) to first obtain measures of diversity (heterozygosity and allelic richness). To eliminate potential biases caused by missing genotypes, we only used individuals that were successfully genotyped for all loci, (n = 684, with population sample sizes ranging from 14-39) to estimate allelic richness. To test for the effects of spatial region on genetic diversity, we used a linear mixed effects model in the 'lme4' v1.1-21 R package (Bates et al., 2015) with random effects for locus and fixed effects for spatial region, fishing pressure status, and cohort age. We used likelihood ratio tests to compare the full model to a reduced model to test for the significance of terms of interest.
We used a second linear mixed effects model to assess interaction effects; this model included fixed effects for fishing pressure status, cohort age plus the interaction between cohort age and fishing pressure status, and random effects for reef and locus. We then used a likelihood ratio test as described above, and, where effects were significant, we conducted pairwise tests to determine the factor driving significance using the 'lsmeans' v2.30-0 R package (Lenth, 2016).
Furthermore, the effects of genetic drift should increase as the cohort ages and experiences cumulative losses. Since this process will drive changes in allele frequencies over time, we used an exact contingency-table test in the 'genepop' v1.1.7 R package (Rousset, 2008), to test for changes in allele frequency distributions as the cohorts aged.

Population genetic divergence
As a cohort ages, and cohort size declines, genetic drift will increase and should be reflected in greater population genetic divergence.
We predicted that populations that experience fishing pressure would undergo even faster divergence relative to our 'Green' reef controls due to greater cohort losses, leading to elevated genetic change. To detect spatial variation in cohort composition and to determine if patterns of genetic structure varied as the cohort aged, we first estimated Global F ST for all three sampled cohort ages using GenoDive v3.04 (Meirmans, 2020) using an AMOVA with 9999 permutations. We also calculated per locus F ST values for all pairwise comparisons among reefs within each spatial region as well as across spatial regions for each year using GenoDive v3.04 (Meirmans, 2020

| Pairwise relatedness
As individuals age within a cohort, the mean relatedness among individuals is expected to increase if selection acts to favour certain families (Hogan et al., 2010). To determine if population relatedness was changing as the cohort aged, we used the 'Demerelate' v0.9-3 R package (Kraemer & Gerlach, 2017) to estimate pairwise relatedness among individuals within a population for each of the three age classes within the cohort. Demerelate uses a jackknife method to estimate the relatedness across random samples to assess neutral expectations and to aid in the selection of the most suitable estimator choice -we used the relatedness estimator of Queller and Goodnight (1989). Furthermore, Demerelate also provides estimates of the number of siblings and half siblings present in a population sample. Because we expected the cohort to experience selection as it aged, we expected that empirical data will contain more siblings and half siblings than the simulated datasets. To eliminate potential biases caused by missing genotypes, we only used individuals that were successfully genotyped for all loci, (n = 684, with population sample sizes ranging from 14 to 39).
We used two linear mixed effects models implemented in the 'lme4' R package (Bates et al., 2015) to investigate the impacts of fishing pressure and age on relatedness. The first model included reef as a random effect and fixed effects for fishing pressure status, cohort age and spatial region. The second model included fixed effects for spatial region, fishing pressure status, cohort age, and age by fishing pressure status interactions and reef as a random. We then tested the significance of terms of interest using a likelihood ratio test and used the 'lsmeans' v2.30-0 R package (Lenth, 2016) to conduct pairwise tests for effects that were significant based on the results from the global tests.

| RE SULTS
In total, we genotyped 1286 individuals, with 417 from the spatial/ temporal study and 869 from the cohort study. Statistical analyses were carried out using nine microsatellite loci for the spatial/ temporal study and eight loci for the cohort study. Seven loci were removed from both studies due to low number of alleles (PL05, PL07, PL11, PL13, PL14, PL15 and Pma191). PL09 was removed from the cohort study due to low genotyping success in one of the eight populations.

| Population genetic diversity
We found that 14 of 72 tests for HWE deviations were significant (Table S3)

| Population genetic divergence
We found evidence for both spatial and temporal variation in genetic structure; however, the patterns were not consistent (  (Table 2). Furthermore, exact tests of allele frequency distribution differences among populations showed significant differentiation between three pairs of reefs, but not between the two reefs that differed in 1996 (Table 2).
Isolation by distance analyses showed that genetic divergence observed in 1996 was not a function of distance among the reefs 1996 (p = 0.748; r 2 = 0.032); however, the opposite was true in 2004 (p = 0.047; r 2 = 0.77). These patterns of spatial population divergence are again, consistent with the chaotic genetic patchiness previously reported in coral reef fish populations (Hepburn et al., 2009;Hogan et al., 2010). It is likely that the 'chaotic' nature of the spatial and temporal variation we observed is at least partly due to our sampling of mixtures of genetically distinct cohorts, resulting from variable recruitment success.

| Population genetic diversity
To determine the effects of cohort age and fishing pressure on genetic diversity, we measured allelic richness and heterozygosity in the cohorts across the three sampling years (Figure 3 and Table S6).
Because we found no spatial region effects on allelic richness or heterozygosity, we dropped this term from all subsequent analyses (Table S7). Although we found no age effects on cohort heterozygosity, we found significant age (p = 0.0090) effects on allelic richness and age-by-status interaction (p = 0.093) and status (p = 0.083) effects that approached significance. As expected, allelic richness declined with age ( Figure 4a, Table S7). To determine if fished reefs were more likely to undergo genetic change, we performed exact tests of allele frequency distribution change and found that the cohort underwent significant shifts in allele frequency distribution on 7 of the 8 reefs, with no apparent difference in the fished reefs (Table 3). Furthermore, there was variation in the timing of these shifts, depending on the reef examined.

| Population genetic divergence
We estimated Global F ST for all three sampled cohort ages and found significant differentiation in all three age groups (Age 3:  Table S8) to test for effects on cohort genetic structure and found no significant effects for any of the terms tested (Table S9). Although we expected that the populations would show more divergence as the cohort aged, we found that F ST values remained stable as the cohort aged from Age 3 to 5, independent of fishing pressure status (Figure 4b,c).

| Pairwise relatedness
Using pairwise relatedness estimates, we found that there were a greater number of full siblings (χ 2 = 143.8, df = 1, p < 0.00001) in the entire cohort dataset than random expectation. Using two linear mixed models, we found that age effects and status-by-age interaction effects on pairwise relatedness were significant (Table S10).
Overall, the average relatedness of the eight populations showed an increase from age 3 to 4 and remained stable from ages 4 to 5.
( Figure 4d). In general, the Blue and Green reefs showed an increase in relatedness as the cohort aged, while Manipulated reefs showed an increase in relatedness from ages 3 to 4, followed by a decrease in relatedness from ages 4 to 5 ( Figure 5).

| D ISCUSS I ON
We detected both temporal and spatial variation in the genetic structure of P. leopardus on the GBR. However, the patterns of divergence were indicative of chaotic genetic patchiness, with unpredictable variation over both space and time (Hepburn et al., 2009;Hogan et al., 2010;Johnson & Black, 1982). In general, genetic differentiation among reefs was low, indicating high levels of gene flow among populations. The combination of low overall divergence and high spatial and temporal variation is likely due to recurring recruitment waves of settling juveniles that led to changing mixtures of population genetic signatures (Hogan et al., 2012;Johnson & Black, 1982;Selkoe et al., 2006). In such a high gene flow environment, using a cohort approach to study population differentiation is valuable because the lack of adult dispersal decreases the chances of sampling an individual that immigrated into the 'population', but also allows us to assess changes that may be relevant at the population level.

F I G U R E 3
Mean genetic diversity (± 95% confidence intervals) estimates for the three years each cohort was sampled for coral trout from reefs differing in fishing pressure status (cohort study). 'Status' describes the fishing pressure status of the population (see Table 1; fishing pressure status refers to the fishing pressure status of the population: Green reefs are closed to fishing, Manipulated reefs were closed to fishing, but pulse fished in 1999, while Blue reefs were completely open to fishing and also pulse fished in 1999). Likelihood ratio tests showed overall age effects for allelic richness but not heterozygosity

Age
Our cohort study showed that selection acted on the cohort as it aged, but that there were few effects of line fishing on genetic diversity and population differentiation. However, intense pulse fishing pressure did drive changing patterns of pairwise relatedness as the cohort aged, indicating that different mechanisms of selection were likely acting to drive these changes.
Our spatial/temporal study showed that there was weak genetic structure among P. leopardus populations on the GBR and this genetic structure changes over time. Previous studies on common coral trout genetic structure on the GBR found no evidence of divergence ; however, the recruitment process in this species, and other coral reef fishes, likely leads to differences in genetic population structure depending on when and where samples are taken. For example, Russ et al. (1996) observed a particularly strong cohort of P. leopardus that was overrepresented in the population. Patterns of variable recruitment can be extremely important in driving chaotic genetic patchiness, likely due to a relatively small number of individuals contributing disproportionately to the next generation (Larson & Julian, 1999). Variable patterns of genetic structure are common in coral reef fishes; for example, one study showed that there was subtle genetic divergence in S. partitus (the bicolor damselfish) within the Mesoamerican barrier reef system (Hogan et al., 2010) despite other studies reporting no evidence of divergence (Ospina-Guerrero et al., 2008). These inconsistencies in reported genetic structure within populations of coral reef fishes likely reflect high variation in recruitment and local retention (Hogan et al., 2012), which highlights the need for widespread monitoring repeated over time.
Interestingly, we saw elevated population differentiation and significant IBD in 2004, potentially indicating more restricted larval dispersal than in 1996. In general, larval dispersal is the main mechanism of connectivity in reef fish, as adults tend to remain on reefs once they have settled (Sale, 1991). However, climate changerelated stressors may impact patterns of connectivity among reefs.  'Spatial region' refers to the region in which the reef is located, 'Status' refers to the fishing pressure status of the reef, and 'Reef' refers to the identification number assigned to the reef by the GBRMPA (Table 1). The numbers before 'shift' refer to the two age groups that were being compared. Bolded p values are significant.  et al., 2007). Given that P. leopardus larvae can disperse over large distances , it is likely that larval survival plays an important role in the connectivity of these populations, and in turn, their population dynamics and genetic structure.

TA B L E 3
As a cohort ages, genetic divergence among reefs should increase as selection and random genetic change occur, independent of fishing pressure. However, fishing pressure should increase those effects as fishing can be selective and reduced population size enhances random genetic change (Hoffman et al., 2011;Rodríguez-Zárate et al., 2013). We know fishing pressure can contribute to selection (Barot et al., 2005;Haugen, & Vollestad, 2001;Morgan & Colbourne, 1999;Walsh et al., 2006) and would also tend to increase pairwise relatedness within a cohort. Our cohort analysis showed that changes in genetic structure and diversity occurred as the cohort aged. Overall, we found evidence of genetic change occurring both within and among reefs. As expected, overall allelic richness declined as the cohort aged, likely due to random genetic changes occurring as the size of the cohort inevitably decreased. Planes and Romans (2004) reported allozyme heterozygosity declining over time in a single cohort of D. sargus (white seabream). Although we expected a decrease in heterozygosity associated with cohort age, we did not find that effect, perhaps due to the lower sensitivity of heterozygosity as a genetic diversity measure relative to allelic richness. We would expect changes in allele frequency distribution to be the most sensitive measure of genetic change, and we found that allele frequency shifts tended to be common, showing that cohorts were likely to undergo genetic changes as they aged.
As a cohort ages, selection and immigration may occur and mean pairwise relatedness should change to reflect those cumulative changes (Hogan et al., 2010;Planes & Lenfant, 2002). For example, Planes and Lenfant (2002) reported significant fluctuations in mean pairwise relatedness in D. sargus (white seabream) cohorts, with the direction of change depending upon the specific cohort and the age of the individuals. We also found that the direction of change in pairwise relatedness varied with cohort age. Although our observed changes in relatedness may be explained by changing selective pressures (e.g., increased susceptibility of younger individuals to environmental stressors; Pörtner & Peck, 2010), it is also possible it may be due to immigration patterns. While increased relatedness as the cohorts age is consistent with selection reducing the number of families (Hogan et al., 2010), reduced relatedness as the cohort ages can most reasonably be explained by immigration of unrelated individuals (Planes & Lenfant, 2002). Although inter-reef movements are thought to be rare for P. leoaprdus (Davies, 2000;Matley et al., 2015;Sumpton et al., 2008), they have been observed and are typically associated with spawning aggregations (Zeller, 1998), with older individuals more likely to make inter-reef movements. Given that the mean age of maturity is 2.5 years in this species (Ferreira, 1995), we would be more likely to observe inter-reef movements in the older individuals in our study, as they are more likely to be sexually mature.
Furthermore, males are more likely to move to join spawning aggregations (Zeller, 1998), thus straying is more likely in older (male) individuals of this protogynous species. Finally, the reefs included in the cohort study are in close proximity, making inter-reef movements at least possible, if perhaps still unlikely. Hence, spawning aggregationrelated dispersal of older individuals can provide an explanation of our observed pattern of change in relatedness, where increases in relatedness only occurred in the younger cohort ages. Furthermore, the population size of the older aged fish in a cohort is lower, and they are thus more likely to have mean relatedness affected by even rare immigration.
We found no evidence for losses in genetic diversity associated with exploitation among the coral trout at the eight reefs we sampled. There have been reports of harvested populations being resilient to losses in genetic diversity; for example, microsatellite analyses showed that high fishing pressure on a population of Illex argentines (squid) resulted in no significant loss of genetic diversity (Adcock et al., 1999). Additionally, we predicted that cohorts on fished reefs would be more likely to undergo shifts in allele frequencies; however, we found that seven of the eight reefs underwent significant shifts in allele frequency distribution with no apparent bias towards fished (Blue) reefs. This may be due to underlying differences among the populations; for example, habitat differences may lead to variation that masks the effects of fishing pressure (Ludwig et al., 1993;Mapstone et al., 2004;Payet et al., 2020). Additionally, Payet et al. (2020)  of an effect on Blue reefs may be explained by the ongoing fishing pressure resulting in population adaptation to fishing-related selection through social learning, which is thought to play a role in evading capture (discussed by Leigh et al., 2014). However, a different pattern of pairwise relatedness was observed for Manipulated reefs -the much larger increase in relatedness from age 3 to 4 in the manipulated populations may reflect strong selection on naïve populations, buffered at age 5 by possible gene flow, perhaps facilitated by the disturbance associated with the pulse fishing. Furthermore, the individuals in the Manipulated populations may also adapt to fishing pressure through social learning and thus are less affected at older ages. This is consistent with underwater visual survey data that showed pulse fishing did have an impact at the population level for Manipulated reefs (Mapstone et al., 2004). The susceptibility to individuals being captured is likely an interplay between size and naivety, with younger individuals likely being more naïve to fishing pressure, making them more susceptible to being captured and older individuals being larger, and therefore, more likely to be captured.
However, larger fish accumulate toxins that cause human health issues (Lewis & Holmes, 1993;Wong et al., 2005) and are therefore less valuable for fisheries, which may reduce selection pressure on older individuals. Furthermore, the stable relatedness between Ages 4 and 5 on the Blue and Green reefs may reflect not only reduced natural mortality (and hence scope for selection) as the fish age, but perhaps also increased inter-reef movements decreasing relatedness. Because we detected pulse fishing pressure effects at the cohort level, it is likely that fishing pressure plays an important role in shaping genetic structure at the population level.
In general, coral reef fishes tend to be challenging for population genetic applications in conservation and management, given their larval dispersal and diverse life histories. However, reef fishes are experiencing multiple stressors, and it is critical that these iconic species are conserved for both economic and ecological purposes, and genetic analyses can provide valuable insights. In this study, we focused on a particularly problematic reef fish species -P. leopardus are top-predators that are actively exploited both commercially and recreationally on the GBR. Here, we report spatial and temporal chaotic genetic structure, a pattern previously reported for other reef fishes, but not for P. leopardus. Using DNA from known-aged fish, we tracked an individual cohort as it aged, allowing us to examine populations without the confounding effects of new recruits changing the genetic structure of the 'populations'. Furthermore, we have shown that pulse fishing likely affects P. leopardus, likely through selection, and resilience to stressors is likely population-specific, and may be unpredictable, as other studies on P. leopardus populations have shown (Mapstone et al., 2004;Mclean et al., 2011). Variation in genetic patterns and resilience to stressors needs to be incorporated into models allowing for the sustainable harvest of coral reef fishes.

ACK N OWLED G EM ENTS
We thank Dr. J. Smit for her assistance in DNA extractions, S.

CO N FLI C T O F I NTE R E S T
The authors report no conflict of interest associated with this work.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data for this study are available at: https://doi.org/10.5061/dryad.