Multiple Lines of Evidence Indicate Limited Natural Recruitment of Golden Perch ( Macquaria ambigua ) in the Highly Regulated Lachlan River

: Freshwater ecosystems and their associated biota have been negatively impacted by the human development of water resources. Fundamental to restoration activities for target species is an understanding of the factors a ﬀ ecting population decline or recovery. Within Australia’s Murray–Darling Basin, recovery e ﬀ orts to address the population decline of native freshwater ﬁsh include stock enhancement, habitat restoration, and the delivery of environmental water. Essential to guiding future management actions is information to assess the e ﬃ cacy of these e ﬀ orts. We undertook a study to investigate whether natural spawning and recruitment, stock enhancement, or a combination of the two is contributing to sustaining populations of golden perch ( Macquaria ambigua ) in the highly regulated Lachlan River, Australia. Otolith microchemistry and genetic analyses were used as complementary tools to determine the source (hatchery origin or wild-spawned) of existing populations in the catchment. We identiﬁed that natural spawning and recruitment was contributing to riverine populations in some years but that populations were heavily reliant on stocking. It was not possible to distinguish hatchery and wild-born ﬁsh using genetic tools, highlighting the value of using multiple lines of evidence to establish causal mechanisms contributing to population recovery. R.T., R.P.D., J.D.T. T.G.E.; validation, T.E., F.D., R.T., R.P.D. and J.D.T.; formal analysis, F.S.; resources, T.E., F.D. and J.D.T.; data curation, F.S. and T.E.; writing—original draft preparation, F.S.; writing—review and editing, T.E., F.D., R.T., R.P.D., J.D.T. and T.G.E.; visualization, F.S.; supervision, T.E., and J.D.T.; project administration,


Introduction
Recruitment is a fundamental requirement for sustaining fish populations [1,2]. In freshwaters, fish and other biota have undergone global population decline, primarily as a result of anthropogenic disturbances such as climate change, river regulation, over-exploitation, pollution, and pests and diseases [3][4][5][6]. Often, these disturbances directly impact a species' ability to complete life-history requirements such as spawning, migration, and recruitment into adult populations [7,8]. Consequently, stock enhancement is often used as a remediation tool to subsidize a vulnerable fish population with no or limited natural recruitment, with varying degrees of success [9][10][11]. In many cases, stocked fish have been shown to have poor post-stocking survival and reproductive success [12][13][14]. In catchments Under this long-term isolation scenario and the hatchery practice regime, we hypothesized that the source of M. ambigua recruitment in the Lachlan River in the last twenty years will be either natural recruitment or stocking from the hatcheries. The otolith core will have developed at the site of the natal origin of a fish, while the otolith edge will have developed in the habitat from which the fish was sampled. Consequently, we assumed a fish born in the Lachlan River would have the same chemical signature in both the core and the edge of the otolith, while a stocked fish would have a different chemical signature in the otolith core, reflecting the hatchery conditions, relative to the edge, reflecting conditions in the Lachlan River.

Otolith Sampling
We collected otoliths from a total of 44 M. ambigua sourced from two sites on the Lachlan River (upper Lachlan and lower Lachlan) ( Figure 1). An additional 19 young of year (YOY) juvenile M. ambigua were collected from a commercial hatchery (Murray Darling Fisheries) as a control group to provide a hatchery reference collection. (Table 1). Fish were humanely euthanized following the animal ethics protocol approved by the University of Canberra Animal Ethics Committee, AEC 17-18 and Fisheries NSW Animal Care and Ethics permit 14/10. The otoliths were separated from the head using sterile forceps (washed with 100% EtOH followed by a wash with distilled water), and otoliths were washed in Milli Q (Millipore, Merck) water. Air-dried otoliths were stored in sterile 5 mL tubes in dry Under this long-term isolation scenario and the hatchery practice regime, we hypothesized that the source of M. ambigua recruitment in the Lachlan River in the last twenty years will be either natural recruitment or stocking from the hatcheries. The otolith core will have developed at the site of the natal origin of a fish, while the otolith edge will have developed in the habitat from which the fish was sampled. Consequently, we assumed a fish born in the Lachlan River would have the same chemical signature in both the core and the edge of the otolith, while a stocked fish would have a different chemical signature in the otolith core, reflecting the hatchery conditions, relative to the edge, reflecting conditions in the Lachlan River.

Otolith Sampling
We collected otoliths from a total of 44 M. ambigua sourced from two sites on the Lachlan River (upper Lachlan and lower Lachlan) ( Figure 1). An additional 19 young of year (YOY) juvenile M. ambigua were collected from a commercial hatchery (Murray Darling Fisheries) as a control group to provide a hatchery reference collection. (Table 1). Fish were humanely euthanized following the animal ethics protocol approved by the University of Canberra Animal Ethics Committee, AEC 17-18 and Fisheries NSW Animal Care and Ethics permit 14/10. The otoliths were separated from the head using sterile forceps (washed with 100% EtOH followed by a wash with distilled water), and otoliths

Otolith Processing and Laser Ablation ICP-MS Analysis
The core of each otolith was externally marked with a permanent marker after examining the surface and placed on a mounting tray with the marked side upward. The otoliths were then embedded in a 2-part epoxy resin containing 5:1 resin (ChemAlart) to hardener (methyl ethyl ketone peroxide catalyst (Nuplex)) and hardened at 45 • C for 4 h. Resin blocks were then sectioned using a low-speed diamond saw (TechCut 4™, ALLiED). To expose the core of the otoliths, each block was sectioned into 3 slices (200-300 µm each) with the marked position of the otolith in the middle slice ( Figure S1). The sections were washed with Milli-Q water and examined for the shape of the sulcus groove in each slice. The slice containing the core (based on the shape of the sulcus groove) was selected for further analysis. The selected otolith sections were polished on a 3M lapping film ALLiED) to expose the core as well as to remove saw marks. The polished otolith sections were mounted on a glass slide using a mounting adhesive (Crystal Bond 509, ProSciTech).
Laser ablation-inductively coupled plasma-mass spectrometry (LA-ICP-MS) was performed on the samples at the Research School of Earth Science, Australian National University using a 193 nm ArF beam (LPF 200, Lambda Physik) and an iCap RQ (ThermoFisher) quadrupole-ICP-MS. Samples were pre-ablated to remove surface contaminants; the otoliths were then analyzed for a suite of trace element concentrations based on previous work [22,47]. The analyzed elements included 24 Mg, 86 Sr, 137 Ba, 55 Mn, 57 Fe, 59 Co, 60 Ni, 63 Cu, and 66 Zn. The analyses were bracketed by analyses of NIST 612 (National Institute of Standards and Technology) and the in-house Davies Reef Coral standard [48] before and after each sample batch. Samples were normalized to Ca and then evaluated against the coral standard for Mg, Sr, and Ba and against the NIST 612 for all other elements.
Spot analyses were carried out by ablating square 50 µm spots with a repetition rate of 10 Hz and an ablation time of 60 s per spot at the core of the otoliths, as well as at the outer edge. These two locations on the otolith were chosen because it would be expected that hatchery-sourced fish would have a different signature at the edge compared to the core of the otolith, while wild-bred fish should have the same signature at the edge and core. This assumption is based on the geographic distance and differences in source water between the commercial hatchery (source of the hatchery group) and the Lachlan River.

Genetic Data
In this study, we used a subset of samples used in [49] (Shams et al. submitted manuscript 2020) for genetic analysis. We followed the tissue sample collection procedure as described in Shams et al. (submitted manuscript 2020). Briefly, fin clips and/or muscle tissue from the same individuals used for the otolith analysis (44 M. ambigua from Lachlan river and 19 hatchery YOY) ( Table 1) were collected using sterile scissors and washed with 100% EtOH to avoid genetic contamination. Tissue samples were stored at 4 • C in 95% EtOH until further use.
Genomic DNA extraction was performed from fin clip and muscle tissue using protocols developed by Diversity Array Technology Pty Ltd (DArTseq™) [50]. The quality of genomic DNA was evaluated by running agarose gel electrophoresis (1.2% agarose). Genotyping by sequencing was performed by DArTseq™ using a combination of DArT complexity reduction methods and next-generation sequencing following protocols described in [50][51][52][53][54][55]. Markers with a high call ratio (threshold, 0.25) were filtered based on their polymorphic information content (>0.025) using a software package developed by DArT PL (DArTsoft14). Here, a high call ratio represents the least proportion of missing values (to call both alleles of a marker) caused by sequencing error or low-quality genomic DNA.

Otolith Microchemistry
For each element, we tested for a difference between the otolith core and edge of wild-caught fish using Student's t-test with the statistical computing software R (version 3.3.1) [56].
To determine the structure of the M. ambigua population in the Lachlan River and to identify hatchery and wild-born fish, we performed non-metric multidimensional scaling (NMDS) [57] and a PERMANOVA test [58] based on the Bray-Curtis distance using otolith trace element composition (Table S1). The analysis was carried out using the function metaMDS in the R package 'vegan' (version 2.5-3) [59]. The settings included a Wisconsin double standardization, a maximum number of 100 runs to find the best solution, and 1000 permutations for calculating the coefficient of determination (r 2 ) and significance (p-value) [57]. The automatic rotation of the ordination space displays the value's main gradient in the first axis (NMDS1). In addition to the NMDS, we performed a hierarchical cluster analysis using individual dissimilarities based on Euclidean distance to identify groupings of fish based on otolith trace element composition. The cluster analysis was performed using the Ward.D2 implementation of the clustering method (minimum variance method) described by Ward [60,61]. All analyses were carried out with the statistical computing software R (version 3.3.1) [56].

Single Nucleotide Polymorphism (SNP) and SilicoDArT Data Analysis
Data processing, quality assessment, and filtration to reduce potential errors from low-quality genomic DNA and sequencing errors were performed using dartR (version 1.1.11) [62], a package developed for the statistical computing software R. To differentiate hatchery-bred stocked fish from wild-born fish, Principal Component Analysis (PCA) using the packages "factoextra" [63] and "factoMineR" [64] in R was performed on the Single Nucleotide Polymorphism (SNP) and SilicoDArT data.

Otolith Micro-Chemical Composition
Four of the nine elements analyzed showed a significant difference in concentration (ug·g −1 ) between the otolith core and edge for wild-caught fish from the Lachlan River ( 24 Mg, 86 Sr, 55 Mn, and 66 Zn; Student's t-test p-values all < 0.01) ( Figure 2, panel LL_edge vs. LL_core and UL_edge vs. UL_core).
The NMDS clearly separated the chemical signatures from the edges of otoliths from fish from the Lachlan River and those of the hatchery (YOY) samples (stress: 0.07) (Figure 3a). A PERMANOVA test with 999 bootstrapping using the Bray-Curtis dissimilarity index found a significant difference between the clusters (p-value = 0.001, F model = 76.35, ANOVA for homogeneity of multivariate dispersion p = 0.39), indicating a unique chemical signature for each locality. The analyses were then repeated on the lower Lachlan and upper Lachlan edge and core samples separately (Figure 3b,c). For the lower Lachlan, this revealed two distinct groups comprising core and edge samples (Figure 3b). The signature from the hatchery reference samples was more similar to the lower Lachlan core chemical signature than the edge signature of the lower Lachlan River individuals. In addition, a small group of individuals displayed similar core and edge signatures (Figure 3b), suggesting wild-born individuals from the lower Lachlan. Cluster analysis using elements significantly contributing to the core and edge variation ( 24 Mg, 86 Sr, 55 Mn, and 66 Zn) identified three primary groups (clusters) representing predominantly edge, core, and hatchery signatures. This analysis clearly identified six M. ambigua individuals for which core signatures were clustered with the edge signature indicating natural recruitment in the lower Lachlan catchment. Within the core and hatchery (YOY) groups, four individuals displayed core signatures that clustered with the hatchery reference group while the remaining individuals displayed core signatures that grouped separately (Figure 3d). The NMDS clearly separated the chemical signatures from the edges of otoliths from fish from the Lachlan River and those of the hatchery (YOY) samples (stress: 0.07) (Figure 3a). A PERMANOVA test with 999 bootstrapping using the Bray-Curtis dissimilarity index found a significant difference between the clusters (p-value = 0.001, F model = 76.35, ANOVA for homogeneity of multivariate dispersion p = 0.39), indicating a unique chemical signature for each locality. The analyses were then repeated on the lower Lachlan and upper Lachlan edge and core samples separately (Figure 3b,c). For the lower Lachlan, this revealed two distinct groups comprising core and edge samples (Figure 3b). The signature from the hatchery reference samples was more similar to the lower Lachlan core chemical signature than the edge signature of the lower Lachlan River individuals. In addition, a small group of individuals displayed similar core and edge signatures (Figure 3b), suggesting wild-born individuals from the lower Lachlan. Cluster analysis using elements significantly contributing to the core and edge variation ( 24 Mg, 86 Sr, 55 Mn, and 66 Zn) identified three primary groups (clusters) representing predominantly edge, core, and hatchery signatures. This analysis clearly identified six M. ambigua individuals for which core signatures were clustered with the edge signature indicating natural recruitment in the lower Lachlan catchment. Within the core and hatchery (YOY) groups, four individuals displayed core signatures that clustered with the hatchery reference group while the remaining individuals displayed core signatures that grouped separately (Figure 3d).
In the upper Lachlan River samples, the NMDS revealed three distinct groups comprising core, edge, and hatchery reference samples, but the cluster analysis using only those elements significantly In the upper Lachlan River samples, the NMDS revealed three distinct groups comprising core, edge, and hatchery reference samples, but the cluster analysis using only those elements significantly contributing to the core and edge variation suggests more complex groupings within the data set ( Figure 3e). The cluster analysis identified two primary clusters separating the core and edge measurements of the Lachlan River samples from the hatchery reference collections. Five tentative sub-clusters belonging to two major groups of Lachlan River samples and hatchery samples were evident; however, the groupings were less distinct within the clusters (Figure 3e). The results suggest none of the upper Lachlan samples were wild-born, while five individuals displayed a core signature that clustered with the hatchery reference group.
Our analysis revealed the presence of both hatchery-bred stocked and wild-born M. ambigua from the Lachlan River. Among the 44 fish collected from Lachlan River, a total of six M. ambigua were identified as wild-born fish (14%), while the rest of them were either hatchery-bred stocked fish or from unknown sources (Figure 4). The observation of annual rings in the otolith revealed that all six wild-born M. ambigua, sampled from the lower Lachlan, were spawned in 2002,2007,2009,2014 (one fish each year), and 2010 (two fish). A comparison between the daily discharge data from the lower Lachlan (Figure 4) revealed no clear association between high flow events and the spawning year of the six wild-born fish.
Water 2020, 12, x FOR PEER REVIEW 7 of 16 however, the groupings were less distinct within the clusters (Figure 3e). The results suggest none of the upper Lachlan samples were wild-born, while five individuals displayed a core signature that clustered with the hatchery reference group.

Genetic Data Analysis
The Diversity Arrays Technology (DArT) pipeline scored 5207 single nucleotide polymorphisms (SNPs) and 6956 SilicoDArT (presence-absence) markers after filtering based on polymorphic information content (>0.025). For the present study, we filtered out 2124 SNP markers based on the call ratio (0.90) and reproducibility average (0.90), resulting in a total of 3083 SNPs (null allele = 2.1%). Further filtering based on the Hardy-Weinberg equilibrium (HWE) was performed to remove markers under selection. Finally, 1481 SNPs were used for the present study. The same criteria (call ratio and polymorphic information content) were also used to filter out low-quality SilicoDArT markers. After filtering, 2902 (nNull allele = 5.7%) SilicoDArT markers were used for further analysis. Filtering for SilicoDArT loci under selection was not performed, as the allelic form of the loci is unknown (presence and absence of restriction fragment).   Despite a small cluster of hatchery reference individuals (hatchery YOY) (Figure 5a), the principal component analysis (PCA) suggests no specific clustering using SNP markers. High similarity among the Lachlan River samples and hatchery YOY and a lack of genetic structure means that the SNP markers could not conclusively distinguish the wild-born fish from the hatchery-bred stocked or unknown-source fish. We also performed a principal component analysis using only Lachlan River samples testing for differences between wild-born fish identified a priori and the rest from the otolith data ( Figure 5b) and could find no genetic differences. The principal component analysis using SilicoDArT markers found a similar result, although it did indicate differences between the lower Lachlan fish (widely spread across PC1/Dim1 in the PCA plot) and upper Lachlan fish (a single cluster, which included the hatchery-bred YOY fish) (Figure 5c,d).
information content (> 0.025). For the present study, we filtered out 2124 SNP markers based on the call ratio (0.90) and reproducibility average (0.90), resulting in a total of 3083 SNPs (null allele = 2.1%). Further filtering based on the Hardy-Weinberg equilibrium (HWE) was performed to remove markers under selection. Finally, 1481 SNPs were used for the present study. The same criteria (call ratio and polymorphic information content) were also used to filter out low-quality SilicoDArT markers. After filtering, 2902 (nNull allele = 5.7%) SilicoDArT markers were used for further analysis. Filtering for SilicoDArT loci under selection was not performed, as the allelic form of the loci is unknown (presence and absence of restriction fragment).
Despite a small cluster of hatchery reference individuals (hatchery YOY) (Figure 5a), the principal component analysis (PCA) suggests no specific clustering using SNP markers. High similarity among the Lachlan River samples and hatchery YOY and a lack of genetic structure means that the SNP markers could not conclusively distinguish the wild-born fish from the hatchery-bred stocked or unknown-source fish. We also performed a principal component analysis using only Lachlan River samples testing for differences between wild-born fish identified a priori and the rest from the otolith data ( Figure 5b) and could find no genetic differences. The principal component analysis using SilicoDArT markers found a similar result, although it did indicate differences between the lower Lachlan fish (widely spread across PC1/Dim1 in the PCA plot) and upper Lachlan fish (a single cluster, which included the hatchery-bred YOY fish) (Figure 5c,d).

Discussion
The aim of this study was to investigate whether there was evidence of the natural recruitment of the existing M. ambigua population at the Lachlan River, or whether all the fish were derived from hatchery stocking. In doing so, we tested the effectiveness of genetic and otolith data as comparative tools. The analysis of otolith microchemistry identified limited natural recruitment along with the co-existence of wild-born and hatchery-bred stocked fish in the lower reach of the Lachlan River. While the data set was less clear about the origin of the upper Lachlan individuals, it appears that no natural recruitment was evident in our samples. Genetic data were not able to clearly distinguish hatchery-bred stocked and wild-born fish. Understanding the effects of stocking on population composition is essential to evaluate the effectiveness of stocking to inform effective management and population recovery [31]. To date, the mark-recapture of chemically or physically tagged fish is the predominant method used to identify hatchery-bred stocked fish in Australia [17,[65][66][67]. Otolith trace element composition has been widely applied in identifying the natal origin of fish [68][69][70][71][72][73]. This method has also been used to identify hatchery-bred stocked fish in wild salmonid populations [24,74,75].
We performed an otolith spot measurement to compare the chemical composition of the core (indicating natal origin) and edge (indicating current habitat). We observed that 24 Mg, 86 Sr, 137 Ba, and 55 Mn were the elements significantly contributing to the variation between core and edge composition. This is consistent with previous observations of otolith trace element composition in carp (Cyprinus carpio) from the Lachlan River [22].
In the upper Lachlan samples, no natural recruitment was evident using otolith microchemistry data. However, the clustering of the otolith edge, core, and hatchery reference groups was more complex than for the samples from the lower Lachlan. A different chemical signature was displayed by the core of otoliths compared to the edge as well as the YOY hatchery samples, indicating that the natal origin of these fish was neither the Lachlan River nor the Murray Darling Fisheries. Although our data could not conclusively identify the natal origin of these fish, plausible explanations for such results are (i) that they are fish stocked from hatcheries other than the Murray-Darling Fisheries such as DPI Fisheries ( Figure 1) and (ii) that they are caused by interannual variability in the multi-element composition in the otolith, or a combination of both.
The fact that no association between high-flow events in the lower Lachlan and a priori recruitment events was evident in the present study suggests a flow-independent recruitment pattern in this species, which has also been reported in previous studies [40,76]. However, this could be an artifact of there being very few individuals from each spawning year. Larger sample sizes from each cohort will be required to provide conclusive evidence.
M. ambigua is a highly migratory, long-lived species and widely distributed in the MDB [34]. Within this region, M. ambigua is considered as a single panmictic population with high gene flow across connected catchments, resulting in high genetic similarity among individuals from different drainages of the MDB [77,78]. Our approach to identify natal origin using SNP and SilicoDArT markers was unsuccessful. Such an outcome would be expected if (i) either one or both parents of the river-born fish were previously stocked hatchery-bred fish, which would indicate the natural recruitment of stocked golden perch in the Lachlan River; or (ii) the broodstock being used in the hatchery were previously stocked fish from a different cohort. The longevity of the species, the intensity of M. ambigua stocking in Lachlan River, and the hatchery quality assurance program (the frequent replacement of old broodstock with wild-caught fish from the Lachlan River), coupled with abundant stocked fish in the river (the results from this study), suggests the second assumption is more likely to be correct. However, the natural recruitment in stocked fish is not infeasible considering a similar result in other species [16].
One possible limitation of this study that resulted in the inability of the genetic data to identify natal origin could be the low genomic representation of the hatchery samples. We sampled only a single cohort for the hatchery reference sample for the present study. Given the high genetic similarity of the species within the MDB as well as in Lachlan River, we recommend further analysis using multiple hatchery cohorts to demonstrate the effectiveness of genetic markers to identify hatchery-bred stocked and wild-born fish. The necessity of multiple-cohort sampling was also reflected in the otolith microchemistry analysis. We recommend further otolith study using multiple cohorts of hatchery reference as well as covering all the hatcheries contributing to the Lachlan River stocking. This will also reduce the probable error caused by physiological influences on otolith microchemistry, which was not considered in the present study [79,80].
There are increasing numbers of examples of studies that combine otolith microchemistry and genetic analyses [24,[81][82][83][84]. As a combined approach, most studies have used otolith data to perform the fine-scale identification of spawning grounds and genetic data to investigate gene flow and connectivity among populations. Focusing on stock enhancement, we identified only one study where a similar research question to ours was addressed using a combination of genetic and otolith data. Perrier, Daverat, Evanno, Pécheyran, Bagliniere, and Roussel [24] used this multi-marker approach to infer information about the breeding activity of stocked Salmo salar (Atlantic Salmon) in wild conditions. However, the utility of a multi-marker approach is dependent on the presence of fine-scale genetic differentiation between hatchery and wild fish. For genetically homogenized populations, otolith trace elements are the more useful approach to identify sub-populations from different localities (e.g., Svedäng, André, Jonsson, Elfman, and Limburg [84]).
In conclusion, the present study has found evidence of limited natural recruitment in the lower Lachlan River, while no natural recruitment was evident in the upper Lachlan. The study successfully demonstrates the utility of otolith microchemistry analysis to identify stocked and wild-born M. ambigua, particularly where a lack of genetic structure rules out other approaches. Along with the use of otoliths, we propose that parent-offspring assignment approaches with genetic data from the hatchery broodstock could be effective in terms of the genetic differentiation of stocked and wild-born fish.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4441/12/6/1636/s1, Figure S1: A schematic diagram of otolith processing. The best section containing core of the otolith was selected to mount on glass slide. 1 and 3: 200-300 µm sections from each side of the core. 2: the otolith section with the core; Table S1