Adaptive Potential of Syzygium maire, a Critically Threatened Habitat Specialist Tree Species in Aotearoa New Zealand

Abstract

The restoration of swampland is vital for the recovery of both biodiversity and cultural values in Aotearoa New Zealand. Syzygium maire, an endemic wetland tree species, is a focus of many wetland restoration efforts. Formerly widespread, extant populations are small, fragmented, and under pressure from myrtle rust. Restoration initiatives may be unknowingly compounding these threats to the species by failing to represent the complete genetic diversity of populations. What genetic diversity remains in remnants and how it is distributed is not known. We therefore aimed to assess the national scale population structure, genetic diversity, and adaptive potential of S. maire to inform species conservation. We identified over 760,000 high-quality single nucleotide variants in 269 reproductive age trees from across the species' range, using low coverage whole genome resequencing. At a national scale, we found five distinct regional-scale genetic clusters, which in turn exhibit local structure and admixture. In the North Island: Northland, Bay of Plenty in the central east, Taranaki in the central west, and Greater Wellington/Manawatū in the south. A single cluster was identified in the South Island, Marlborough. Within-cluster substructure was particularly evident for Greater Wellington/Manawatū. Genetic diversity and fixation indices (F _ST) were relatively uniform across all clusters, and there was some evidence of north to south increase in kinship and shorter time since radiation. These patterns are likely to reflect glaciation cycles that resulted in complex contractions into local microrefugia and subsequent re-radiations of the species over time. Genotype by environment analysis detected genetic variants potentially contributing to environmental adaptation, notably precipitation seasonality. Restoration and conservation goals would best be served by capturing diversity within regional clusters. Information on the geographic and environmentally structured distribution of this tree's genetic diversity supports conservation and restoration strategies through ensuring the complete extant diversity is captured, identifying regions at most risk of genetic degradation, and facilitating planning regarding the movement of adaptive diversity in a changing environment.

FIGURE 1

Locations of sampling sites used for this study. Each circle represents multiple smaller sampling sites (see Table S3 for a breakdown of sampling sites per region). Circle size reflects sample size, while its color corresponds to the assigned genetic cluster according to k‐means analysis. Dark lines represent political borders, while light gray relief represents the remaining native forest. Landcover information was derived from the New Zealand Landcover Database version 5 (LRIS), geographical borders from NZ Coastlines and Islands Polygons (LINZ), and political boundaries from the generalized regional councils 2023 layer (StatsNZ).

FIGURE 2

Clustering analysis of 269 S. maire trees sampled across Aotearoa. Individuals are coloured according to the region from which they were sampled, as follows: Northland (NOR), Bay of Plenty (BOP), Taranaki (TAR), Manawatū (MAN), Greater Wellington (GWE) and Marlborough (MAR). (A) K‐means clustering analysis. Calculations were performed for 20 values of K, but only the first 10 are shown here. Optimal values for K are the first minimum Bayesian Information Criterion (BIC) that occurs before a subsequent increase in BIC. Hence, optimal number of clusters for this analysis was 5. (B) Three‐dimensional representation of discriminant analysis of principal components (DAPC) for a dataset of 188,131 SNPs. Each point depicts a single individual. LD1 explains 66.0% of the variation in the dataset, and largely separates northern and southern populations. LD2 explains 17.8% of the variation, and appears to segregate MAR and TAR from the other populations.

FIGURE 3

Admixture analysis for 203 individual trees sampled across Aotearoa. To generate the plot, ADMIXTURE was run for a dataset of 188,131 SNPs filtered for a minor allele frequency of 0.05, linkage disequilibrium, and related individuals. The outcome is an estimation of the proportion of ancestry for each individual derived from a number of hypothetical ancestral populations (as defined by K). Individuals are organized according to region from north at the left of the plot to southern populations at the right. We assessed the shared ancestor for two to seven common ancestors; the size and colour of each bar reflect the proportional contribution of an ancestral population to that individual. Abbreviations for regions are: Northland (NOR), Bay of Plenty (BOP), Taranaki (TAR), Manawatū (MAN), Greater Wellington (GWE), and Marlborough (MAR).

FIGURE 4

Gene diversity (heterozygosity) for each sampling region. Boxplots represent the median and interquartile range of values, with whiskers depicting the 95% confidence intervals of heterozygosity and outliers depicted as black dots. Mean values of heterozygosity are depicted by +. Statistics were calculated on 188,131 SNPs filtered for linkage disequilibrium and minor allele frequency of 0.05. BET, between population diversity; BOP, Bay of Plenty; GWE, Greater Wellington; MAN, Manawatū; MAR, Marlborough; NOR, Northland; TAR, Taranaki.

FIGURE 5

Kinship of all individuals and regions. (A) Heatmap representing scaled pairwise individual kinship. Calculations were performed in hierfstat and based on 957,330 SNPs filtered for linkage disequilibrium but not minor allele frequency. Inbreeding coefficients are displayed on the diagonal. All values were scaled to the minimum in the dataset. Individuals are grouped into regions from which they were sampled, and subregions are arranged from north to south. (B) Boxplots showing the distribution (median, interquartile range, 95% confidence intervals, with outliers depicted as single unfilled dots) of scaled beta‐kinship values per region. BOP, Bay of Plenty; GWE, Greater Wellington; MAN, Manawatū; MAR, Marlborough; NOR, Northland; TAR, Taranaki.

FIGURE 6

Genotype by environment analysis for precipitation seasonality. (A) RDA axes one and three for SNPs; circles denote individual trees, and environmental variables are overlain as vectors. The length and direction of vectors is proportional to the strength of the relationship of each environmental predictor and the SNPS. Individuals are coloured according to sampling region: Northland (NOR), Bay of Plenty (BOP), Taranaki (TAR), Manawatū (MAN), Greater Wellington (GWE) and Marlborough (MAR). (B) Violin plot depicting data quantity along the gradient of assessed precipitation seasonality. Precipitation seasonality is measured as the ratio of the standard deviation of the monthly total precipitation to the mean monthly precipitation. Each point represents the seasonality at an individual sample location. (C) Manhattan plot of LFMM analysis. The significance of each SNP's association (−log10 of the q‐value) with precipitation seasonality is shown on the y‐axis. Chromosome number and the physical position of SNPs are shown on the x‐axis. A q‐value cutoff of 0.05 (red dashed line) was chosen. SNPs depicted in red were also identified by the RDA analysis and present in the outlier set detected by pcadapt.