Population Genomics of New Zealand Pouched Lamprey (kanakana; piharau; Geotria australis)

Abstract

Pouched lamprey (Geotria australis) or kanakana/piharau is a culturally and ecologically significant jawless fish that is distributed throughout Aotearoa New Zealand. Despite its importance, much remains unknown about historical relationships and gene flow between populations of this enigmatic species within New Zealand. To help inform management, we assembled a draft G. australis genome and completed the first comprehensive population genomics analysis of pouched lamprey within New Zealand using targeted gene sequencing (Cyt-b and COI) and restriction site-associated DNA sequencing (RADSeq) methods. Employing 16 000 genome-wide single nucleotide polymorphisms (SNPs) derived from RADSeq (n = 186) and sequence data from Cyt-b (766 bp, n = 94) and COI (589 bp, n = 20), we reveal low levels of structure across 10 sampling locations spanning the species range within New Zealand. F-statistics, outlier analyses, and STRUCTURE suggest a single panmictic population, and Mantel and EEMS tests reveal no significant isolation by distance. This implies either ongoing gene flow among populations or recent shared ancestry among New Zealand pouched lamprey. We can now use the information gained from these genetic tools to assist managers with monitoring effective population size, managing potential diseases, and conservation measures such as artificial propagation programs. We further demonstrate the general utility of these genetic tools for acquiring information about elusive species.

Graphical Abstract

Diadromous fish species, fishes which spend portions of their life cycles in both fresh and marine environments, comprise ~1.5% of all described fishes yet they amount to ~3% of all endangered species (McDowall 1992). These species are likely at enhanced risk since they must find passages through freshwater barriers and combat the threats of both freshwater and marine systems. These threats include habitat degradation, overexploitation, invasive species, poor forestry practices, pollution, climate change, disease, marine overexploitation, and bycatch (McDowall 1992; Dudgeon et al. 2006; He et al. 2017; WWF 2018).

Geotria australis Gray, 1851 (pouched lamprey) is known as kanakana by southern iwi (Indigenous New Zealand tribes) and piharau by northern iwi in Aotearoa New Zealand. It is a diadromous, jawless vertebrate in Geotriidae (Petromyzontiformes) (Figure 1) and is one of the most widely distributed lampreys; G. australis is found in Chile, Australia, and throughout Aotearoa New Zealand, including Te Ika-a-Māui (the North Island), Te Waipounamu (the South Island), Rakiura (Stewart Island), and Rēkohu/Wharekauri (the Chatham Islands) (McDowall 1990, 2002; Tedd and Kelso 1993; James 2008). Similar to salmon, Geotria are anadromous; eggs and larvae develop in freshwater for approximately 3–4 years, juveniles migrate downstream to the ocean where they feed and mature, and mature adults migrate back into freshwater where they spawn and die (Baker et al. 2017; Miller et al. 2021). At all life stages Geotria are prey for multiple species, including endangered species, and likely act as ecosystem engineers during their larval stages (Tickell 1964; McDowall 1990; Moore and Wakelin 1997; Docker et al. 2015).

Figure 1.

Geotria australis (kanakana/piharau; pouched lamprey). (A) Migrating adult, (B) oral disc of migrating adult, (C) juvenile, (D) adult. Photo credit: A = AKM, B = Matthew Dale, C = Jonah Yick, D = Robert Holdaway. Photo D was reprinted from Population structure in anadromous lampreys: Patterns and processes, In Press, Copyright (2021), with permission from Elsevier.

Despite their cultural importance, biological importance, and moderate size (migrating mature adults ~500 mm length), pouched lamprey are often physically overlooked in both freshwater and marine environments. As a nocturnally moving species that hides during the day, pouched lamprey are often missed or visually misidentified as eels (Kelso and Glova 1993; Jellyman and Glova 2002). This has led them to be referred to as an “elusive” or “secretive” species. Adults and juveniles are not large enough to fit satellite tags, hampering attempts to both monitor their marine life stage and determine if adults are philopatric, that is, if they return to natal or “home” streams.

Against this dearth of knowledge around population connectivity, pouched lamprey abundance appears to be declining in Aotearoa New Zealand (Best 1941; McDowall 1990; McDowall 2011; Kitson et al. 2012; Dunn et al. 2018). Reasons for these declines are unknown, although, anthropogenic stressors and disease are possible causes. Lamprey reddening syndrome (LRS), a syndrome associated with external hemorrhaging around the gill-pores and along the body, has been recorded on upstream migrating pouched lamprey from the South Island since 2011 (Brosnahan et al. 2018). The lack of rudimentary knowledge concerning the geographical relationships (population structure), degree of philopatry, and migration patterns of pouched lamprey in New Zealand has greatly hindered the ability of managers to implement robust conservation policy for this species. The potential risk of LRS transmission between the North and South Islands is a particular concern. In addition, understanding why certain pouched lamprey population patterns exist (e.g., recent origin/limited gene flow, and [or] more ancient origin/elevated gene flow) will allow managers to better understand the vulnerability and stability of the current pouched lamprey population(s).

Here, we investigate population connectivity throughout Aotearoa New Zealand based on 186 pouched lamprey sampled from 5 North Island and 5 South Island locations. We use a newly assembled draft genome, restriction site associated DNA sequencing (RADSeq; Elshire et al. 2011; Dodds et al. 2015), and targeted sequencing of mitochondrial DNA (mtDNA) to investigate genetic structure, gene flow, and historical demography of pouched lamprey in Aotearoa New Zealand. Using these inferences, we then demonstrate how genomic methods, such as RADSeq, can be used to enhance information about the current (and historical) population structure, size, and migration of pouched lamprey and other elusive species. We further illustrate how this information can be used to assist managers with translocation, propagation, disease management, and population monitoring through time.

Materials and Methods

Sample Collections

New Zealand National Institute of Water and Atmospheric Research (NIWA) scientists and iwi (Māori tribal) representatives collected pouched lamprey adults and larvae from 2016 to 2019 using electrofishing, hand-catching, and trapping methods. Specifically, 258 larvae and 37 adults were collected (Supplementary Table S1). Engagement with local iwi was undertaken before all collection events. Following recommendations by Hale et al. (2012), approximately 30 individuals were targeted during each sampling event; however, tissue quality and relatedness scores following initial data analysis determined the final number of individuals incorporated in our analyses. Tissues were collected from the 295 total individuals during 12 sampling events at 10 sites (North Island: Mokau, Piha, Kaniwhaniwha, Pauatahanui, and Mangakino; South Island: Styx, Waiau, Waikawa, Okuti, and Pourakino) (Figure 2). The Kaniwhaniwha and Waikawa sites were sampled twice. Sample identification (ID), location, year sampled, life stage, library number, raw read counts, number of loci, Aotearoa Genomic Data Repository accession links, targeted markers used, and NCBI GenBank accession numbers are provided in Supplementary Table S1. For adult pouched lamprey, a fin clip from the caudal fin or from the second dorsal fin was placed into RNAlater® [Thermo Fisher Scientific], and for larvae, the entire individual was placed in 15 mL vials of RNAlater®. Additional muscle and gonadal tissue were collected from one mature male sampled at the Okuti site, preserved in RNAlater®, and used for the draft genome assembly. All tissues were stored at −20 °C until they were used for DNA extractions.

Figure 2.

Aotearoa New Zealand pouched lamprey sampling locations and marine currents (inset). Blue points represent Te Ika-a-Māui/North Island sites. Green points represent Te Waipounamu/South Island sites. Striped blue and striped green points, along with asterisks (*), denote a site sampled on more than one occasion. Numbers below the site names denote the total number of larvae and adults collected from the site.

DNA Extraction and Sequencing

DNA was extracted from subsampled tissues following Gemmell and Akiyama (1996), quantified via Qubit Fluorometer (Thermo Fisher Scientific), and DNA integrity assessed using gel electrophoresis.

Targeted Mitochondrial DNA Markers

Mitochondrial DNA cytochrome b (Cyt-b) sequences were amplified from 94 New Zealand pouched lamprey using Geotria496L and Phe1612H primers (Lang et al. 2009). The 20 μL polymerase chain reaction (PCR) was carried out using Bioline Biotaq DNA polymerase and comprised of the following: 14.1 μL ddH₂O, 2 μL of 10× reaction buffer, 1.2 μL of MgCl₂ (50 mM), 0.5 μL of each dNTPs (10 mM), 0.2 μL Taq polymerase (5 μ/μL), 0.5 μL of forward primer (10 μM), 0.5 μL reverse primer (10 μM), and 1 μL template DNA. The thermoprofile consisted of an initial denaturation of 94 °C for 3 min, followed by 30 cycles of 94 °C for 1 min, annealing at 60 °C for 1 min, and extension at 72 °C for 2 min, with a final 10 min extension. Cytochrome c oxidase I (COI) sequences were amplified using the primers (FishF1 and FishR1) following Ward et al. (2005). PCR products were purified using either PALL AcroPrep96 filter plates (http://www.pall.com) or exonuclease 1 and shrimp alkaline phosphatase (https://www.nucleics.com/DNA_sequencing_support/exonucleaseI-SAP-PCR-protocol.html), and sequenced (reverse and forward directions) using BigDye v3.1 dye terminator chemistry, on an Applied Biosystems 3730xl DNA Analyzer (Applied Biosystems), following dye-terminator removal using columns prepared with Sephadex G-50 Fine grade (Cytiva Life Sciences). All available Mordacia (lamprey outgroup taxa, Mordaciidae) and G. australis Cyt-b and COI sequences were downloaded from NCBI GenBank and included with the newly sequenced Cyt-b and COI regions in downstream analyses (sequences and accession numbers in Supplementary Table S1). All new sequences from Aotearoa New Zealand were deposited into the Aotearoa Genomic Data Repository (https://repo.data.nesi.org.nz/TAONGA-LAMPREY/collection/TLCMC00005-00002).

Geotria australis Genome Sequencing and Assembly

Since reference-based RADSeq analyses were found to outperform de novo approaches (Rochette et al. 2019) and position-relative-to-scaffold information was needed for downstream outlier analyses, a genome was sequenced and assembled from a male G. australis. DNA was extracted from RNAlater preserved gonadal and muscle tissues from one male adult pouched lamprey (Supplementary Table S1, NZKIN19_M1_T1) using a Qiagen DNeasy Blood and Tissue Kit. The extracted DNA was submitted to the Otago Genomics Facility where it was quality checked, prepared into 2 separate libraries (Rubicon Thruplex DNA-seq), and HiSeq 2500 V4 sequenced (paired end, 2 × 125 bp) on one lane. Initial assessment of genome size and read sampling was performed by examining the distribution of k-mers for k = 21 that was generated using Jellyfish v.2.2.4 (Marçais and Kingsford 2011) and analyzed using GenomeScope2.0 (Ranallo-Benavidez et al. 2020). Evaluation of the sequenced reads with FastQC v0.11.8 showed a high level of duplication primarily due to the presence of telomere-like sequences. Duplicates were removed with reference-free duplicate remover hts_SuperDeduper from HTStream_1.0.0 with default parameters. Retained reads were assembled using Abyss-pe v.2.2.3 (Simpson et al. 2009) based on the construction of a de Bruijn graph for k-mers with an experimentally chosen k = 100. To assess the completeness/quality of the assembly, the assembly was searched for expected low copy orthologs using BUSCO v. 5.1.3 (Seppey et al. 2019) and k-mer based metrics were generated using Merqury v1.1 (Rhie et al. 2020). BUSCO was run with parameters: --limit = 20 (number of candidate contigs considered per ortholog) and the parameter -tcov of MetaEuk (Levy Karin et al. 2020) subroutine was set to 0.2 to accommodate the fragmentary nature of the assembly. Searches for conserved single copy orthologs were performed using databases for metazoan conserved orthologs (metazoa_odb10: 23 February 2021) and a previously reported set of 233 Core Vertebrate Genes (Hara et al. 2015) that was reformatted to be compatible with BUSCO v.5. To further resolve the presence/absence of missing conserved orthologs, missing vertebrate BUSCOs were aligned to the genome using tblastn from blastall suite of the Basic Local Alignment Search Tool (BLAST; blast-2.2.17) with disabled dust filter (parameter-F) to preserve search at low complexity regions (e.g., GC rich) that are known to be enriched in lamprey genomes (Smith et al. 2013). To investigate the presence of muscle-specific k-mers, we extracted the corresponding subregions (lengths 21 to 3088 bases) from 9104 contigs and aligned them to NCBI NR with blastn (Altschul et al. 1990) with word size 11. Contigs with subregions with more than 200 bs were also used in targeted alignments to flatworms (taxid:6157) from the NCBI nt database.

Restriction Site-Associated DNA Sequencing

A specific RADSeq method, termed genotyping-by-sequencing (GBS; Elshire et al. 2011), was used following modifications proposed by Dodds et al. (2015). The PstI restriction enzyme was chosen for its ability to interrogate vertebrate genomes at a high depth of coverage (De Donato et al. 2013). A total of 5 libraries of a 193–318 bp size range (selected via a Labgene Scientific Pippin Prep) were sequenced on an Illumina HiSeq 2500 platform (single-end 1 × 101 bp). Of these 5 libraries, one was created in 2016–2017 with 39 samples and sequenced over 2 lanes; 3 were created in 2018 with 252 samples and sequenced over 3 lanes; and one was created in 2019 with 4 samples and sequenced in 1 lane (Supplementary Table S1). Negative controls and repeated individuals for each sequencing run, that is, the same individual was included in the 2017 and 2018 sequencing rounds (not included in the totals above), were included to detect contamination and ensure genotyping consistency. The raw RADSeq reads were deposited into the Aotearoa Genomic Data Repository (https://repo.data.nesi.org.nz/TAONGA-LAMPREY/collection/TLCMC00005-00002).

Bioinformatics and Analyses

Targeted Mitochondrial DNA Markers (Cyt-b and COI)

Sanger-sequenced Cyt-b and COI sequences were manually inspected and edited (low-quality bases and primer sites removed) using Geneious Prime v2020.1.1. The resulting forward and reverse consensus sequences for each individual sequenced were aligned using MAFFT (Katoh and Standley 2016) with the E-INS-i strategy. Each alignment was then manually inspected with Geneious Prime. Sequence similarity and uncorrected distances were evaluated using BLAST and Geneious Prime, respectively. Haplotype networks (median joining network; Bandelt et al. 1999) were generated with PopART (Leigh and Bryant 2015).

Restriction Site-Associated DNA Sequencing

Raw sequencing reads were quality assessed with FastQC (0.11.9) and assembled to reference using Stacks2 v2.52 (Rochette et al. 2019). First, raw reads were demultiplexed, filtered for adapters, and quality filtered (--clean, --quality, --rescue) with Stacks2 process_radtags following default parameters for each of the 6 lanes individually. The resulting reads (~67% retained) were then mapped (~61% retained) to the indexed G. australis genome scaffolds with BWA-MEM v0.7.17 (Li 2013). Samtools v1.10 (Li et al. 2009) was used to convert output BWA-MEM files to sorted BAM files. The Stacks2 pipeline was then run manually using gstacks with default parameters to identify and genotype SNPs. Using these preliminary data, duplicate and closely related individuals were removed using the custom R script SNP_mismatch (https://github.com/laninsky/SNP_comparisons/blob/master/SNP_mismatch.R, created by AA), and the VCFtools v0.1.16 (Danecek et al. 2011) relatedness statistical analyser --relatedness2 was used to confirm that first degree relatives were identified and removed. VCFtools was used to remove individuals with more than 50% missing data, and Stacks2 populations was used to phase haplotypes at each locus for the remaining individuals, retaining only loci genotyped in a minimum of 80% of individuals in a population (-r 80). The resulting Stacks2-assembled dataset was exported and subjected to further filtering. VCFtools was used to retain only biallelic sites with minor allele count >2, a minimum read depth of 10, and loci found in 80% or more individuals. The final filtered dataset was exported for downstream analyses, excluding the CubSFS (R package, Waltoft and Hobolth 2018) downstream analyses of which no minor allele count filter was applied.

To ensure our results were robust to the choice of assembly program, we also generated a SNP dataset with Ipyrad v0.7.28 (Eaton and Overcast 2020) de novo assembling sequence reads using default parameters with a few customizations (“gbs” datatype, “2” stricter filter adapters, “5” max uncalled bases in consensus, “8” max heterozygotes in consensus, and “20” max number of SNPs per locus). The resulting Ipyrad- de novo-assembled dataset was further pruned for duplicate individuals/close-kin and filtered (minor allele count > 2, min depth 10, loci in 80% or more individuals, biallelic sites only).

Population Structure

STRUCTURE v2.3.4 (Pritchard et al. 2000) was used with an admixture model (Pritchard et al. 2000) and Bayesian iterative algorithm to determine the underlying number of genetic clusters (K) for pouched lamprey. The final filtered dataset (see above) followed the STRUCTURE-specific filtering recommendations of Linck and Battey (2019). Lambda was first inferred for a K of one and was fixed at the obtained value for all subsequent runs. Five replicates were run for each K between 1 and 7, each replicate run starting from a different random seed. Replicates were run for 100 000 MCMC replicates after a burn-in of 50 000. Trials with longer chains did not produce different results and likelihood values were consistent across replicates for each given K, suggesting burn-in was long enough to reach stationarity within runs, and convergence across runs. To determine the optimal K values, the Evanno method (Evanno et al. 2005) was then attempted with both Clumpak v1.1.2 (Jakobsson and Rosenberg 2007) “best K” and Structure Harvester v0.6.94 (Earl 2012). Individual assignments to the best-fitting K genetic clusters and mean likelihood probabilities were then determined from the STRUCTURE Q matrices output using Clumpak and plotted using the Clumpak “main pipeline.”

Model-free principal component analysis (PCA) and discriminant analysis of principal components (DAPC) were performed using the R package adegenet v2.1.3 (Jombart 2008; Jombart and Ahmed 2011). DAPC analyses were performed de novo using the find.clusters to determine the number of groups, optim.a.score to determine the optimal number of PCs, and the lowest BIC value to select the optimal K. PCA and DAPC methods were additionally used to evaluate whether genetic structure was concordant with geographical gradients (latitude and longitude), islands, life stages, and collection years (2015–2016 adults, 2016–2018 larvae).

Patterns of Diversity

Expected heterozygosity, observed heterozygosity (H_O), Wright’s inbreeding coefficient (F_IS), standardized multilocus heterozygosity (sMLH), and genetic diversity (locus and population) were estimated for the 10 Aotearoa New Zealand sampling sites (Figure 2) with HIERFSTAT (R package, Goudet 2005) and inbreedR (sMLH analyses follow Coltman et al. 1999; Stoffel et al. 2016). Nei (1987)F_ST was estimated with HEIRFSTAT, while Weir and Cockerham (1984)F_ST was estimated using StAMPP (R package; Pembleton et al. 2013) and HEIRFSTAT. Pairwise F_ST comparisons and global F-statistics were tested for significance using 1000 bootstrap replicates (95% confidence interval). Effective population size (N_e) was estimated with NeEstimator v2.1 (Do et al. 2014) for the entire Aotearoa New Zealand region (both with the 10 sampling locations considered as separate populations and as one metapopulation) and the North and South Islands separately utilizing a random mating model and the linkage disequilibrium method. A <0.05 allele frequency cut off value (P_crit = 0.05) and a 0 cutoff value (P_crit = 0+, all alleles used) were used, and jackknife 95% confidence intervals constructed.

Isolation by distance (IBD) analyses were performed using a pairwise per individual genetic dissimilarity matrix calculated from the Stacks2 final filtered dataset using 2 R packages: Adegenet v2.1.3 to create a Genlight-class object, and dartR v1.8.3 to convert this to a similarity matrix (Jombart 2008; Jombart and Ahmed 2011; Gruber et al. 2018). Distance between sampling locations was measured from the corresponding river mouth (coastal entrance) as individuals exit the river mouth upon maturity (e.g., they do not traverse land). The 10 sampling locations (datum WGS84) were then linked to individuals and a distance matrix between the locations was calculated using a custom R script (https://github.com/wpearman1996/PopGenScripts/blob/main/LC_distance_calculation.R; van Etten and Hijmans 2010; Pearman et al. 2020; van Etten and van Etten 2020). The resulting geographic pairwise distance matrix and genetic pairwise distance matrix were then vectorized in R and the ecodist v2.0.7 R package was used to estimate a Mantel coefficient and corresponding confidence limits (P-values; Goslee and Urban 2007). In addition, a Mantel test was performed using a genetic distance matrix of mean F_ST per population (Weir and Cockerham 1984; F_ST values calculated with StAMPP).

Gene flow and genetic diversity among sampling regions were estimated with estimated effective migration surfaces (EEMS; Petkova et al. 2016). EEMS, like PCA, can summarize population structure visually; however, unlike PCA, EEMS is not influenced by uneven collection of data among sampling sites (see Petkova et al. 2016, Figure 2). EEMS is useful for datasets where population structure is not entirely consistent with isolation by distance as it identifies areas where genetic similarity deviates from the genetic similarity expected under an isolation by distance model. We used the pairwise genetic dissimilarity matrix and sample coordinates previously calculated for the Mantel test and created a habitat file using the Google Maps API v3 Tool (http:www.birdtheme.org/useful/v3tool.html), including only marine habitat as the environment that pouched lamprey migrates within. A custom R script (created by D. Petkova) was used to run EEMS (Supplementary File S2) as the habitat map included holes (the land extent of the 2 main New Zealand Islands) and setup was complicated by New Zealand’s close proximity to the international date line (longitude 180°). Three independent runs for multiple deme sizes (291, 561, 749, 996; EEMS manual recommended; Petkova et al. 2016) were completed using 6 million Markov chain Monte Carlo (MCMC) iterations and 2 million burn-in iterations for a total of 12 runs. The rEEMSplots R package was then used to produce plots (run conversions, observed and fitted dissimilarities, posterior probabilities) and surfaces (migration [m] and diversity [q]). Migration was considered statistically significant if the posterior probability [Pr] of effective migration rates [m] was greater than the overall mean migration rate [0], given the average squared genetic difference [D], by over 95% for all runs except one run [deme 561].

Outlier Analysis

Outlier analyses were conducted to identify loci showing higher than background differentiation (F_ST) to detect loci that were potentially under selection. The R package OutFLANK (Whitlock and Lotterhos 2015) was used following an adjusted OutFLANK script (T. Oosting, https://github.com/tomoosting/). A binary biallelic genotype table (.bim file) and sample information file (.fam file) were created using PLINK v1.9 (S. Purcell and C. Chang, www.cog-genomics.org/plink/1.9/; Chang et al. 2015) and used with chromosome information created from scaffolds of the assembled G. australis genome to detect outlier SNPs. The analysis was performed using a false discovery threshold of 0.05%, minimum heterozygosity of 0.1, and left and right trim fractions of 0.1. Outliers with the highest F_ST estimate were further thinned to every 500 000 base pairs (“sliding_window”) to ensure putative outliers were unlikely to be in physical linkage. A Discontiguous Megablast search (max E-value 0.05, max target sequences 20) of the NCBI nucleotide collection (nr/nt) database was performed with Geneious Prime v2022.0.1 on the sequences potentially under divergent selection to attempt to identify function.

Historical Demography

Demographic history was estimated with CubSFS v1.0 considering all samples as representing a single population. Methods followed those listed in (Cole et al. 2019). The final filtered dataset, a VCF file containing only variant sites, was first modified to include monomorphic sites following the GitHub fastsimcoal2_inputs protocol (Alexander 2021; https://github.com/laninsky/bats_and_rats/tree/master/fastsimcoal2_inputs). This VCF file (monomorphic sites included) was then used to create a SFS file using easySFS (Overcast 2021; https://github.com/isaacovercast/easySFS), and CubSFS (used with R version 4.0.2) was used to estimate demographic history using 29 knots and a t_m of 0.35 coalescent units.

Results

Sequencing, Assembly, and Filtering

Targeted mitochondrial DNA markers

The 766 bp Cyt-b sequence region was identical in all of the 63 sequenced New Zealand pouched lamprey individuals and showed pairwise similarity of 98.04% (15 fixed substitutions) and 99.86% (1 fixed substitution) to 2 previously sequenced (> 390 bp) G. australis sequences from GQ206164 (Chile) and GQ206165 (Western Australia), respectively (Figure 3). A single transition was observed between the GenBank mitogenome-trimmed Cyt-b sequence from New Zealand (NC_029404; Ren et al. 2016) and our samples, which may be the result of a sequencing error. The New Zealand pouched lamprey haplotype shared ~85.025% sequence identity with Geotria macrostoma haplotypes (GenBank accessions MT478645–53). The 589 bp trimmed targeted COI sequence region was identical in all 20 New Zealand individuals sequenced during this study and with the 14 Australian (Victoria, South Australia, and Tasmania) sequences from GenBank. The New Zealand pouched lamprey COI region additionally showed pairwise similarity of 98.30% (10 fixed substitutions) and 99.32–99.83% (3–4 fixed substitutions) to the 2 Chile and 8 Western Australia sequences listed in GenBank.

Figure 3.

Median joining haplotype network (created with PopART) and pouched lamprey sampling locations. Haplotype network: colors correspond to locations, sizes of circles correspond to the frequency of the sequences belonging to a specific haplotype, numbers correspond to the number of fixed substitutions (mutations) from the large Aotearoa New Zealand haplotype, and the Aotearoa New Zealand haplotype acquired from mitochondrial data (NC_029404) is colored in grey-blue. All of the Aotearoa New Zealand samples represented in the haplotype networks, except for the mitogenome haplotype, were collected during the current study. All of the Australian and Chilean samples were acquired from GenBank.

Geotria australis Genome Sequencing and Assembly

Nearly 200 million paired end reads were sequenced from DNA of gonadal tissue and a similar number from muscle DNA of a male adult pouched lamprey (Supplementary Table S3). Based on the distribution of k-mers within these reads the haploid genome length was estimated to be 1.07 Gb with 38.5% of the genome existing as single-copy sequence with an estimated heterozygosity of 0.67% (on average one polymorphism per 149 bp). Inclusion of both gonad and muscle sequences in the same assembly yielded a total assembly length of 0.8 Gb that was distributed across 1 899 277 scaffolds with N50 scaffold of 966 bases and a maximal scaffold length of 46 391 nucleotides. Searches for single-copy orthologs detected 75% of the expected core vertebrate orthologs (of 233 total) and 79% of the expected metazoan orthologs (of 954 total; Hara et al. 2015). For comparison, the chromosome-scale assembly of the sea lamprey genome contains 90% of the expected core vertebrate orthologs and 75% of metazoan orthologs (Smith et al. 2018). It is also worth noting that we were able to recover partial alignments of 58/59 missing core vertebrate orthologs by aligning them directly to the assembly with tblastn. Partial alignments with at least 60% identity covered from 5.5% to 89% of the orthologs’ length.

The high rate of BUSCO detection is consistent with k-mer based estimates of genome completeness and quality, which indicate a consensus error rate of <1/million bases and 91% assembly completeness when testes and muscle reads are considered together, though it is important to recognize that the assembly was generated from these same reads. Of the total assembly k-mers, 0.17 million k-mers were found to be muscle-specific, that is, were missing from the testes read set, compared to 1.66 million k-mers that were found to be testes-specific, that is, were missing from the muscle read set. This is consistent embryonically programmed DNA elimination events that result in the removal of specific sequences from somatic cell lineages, as has been observed in other lamprey species (Smith et al. 2009, 2018, 2020). However, a more complete assessment of programmed DNA eliminations in pouched lamprey will await assembly improvement and the coordinated collection of additional samples from reproductively mature individuals. Investigation of muscle-specific k-mers, or those undersampled in testes, revealed that 42 corresponding contigs aligned to high copy rDNA sequences from flatworms (several from Stegodexamene anguillae), which might reflect the presence of this parasite within the muscle tissue of the sequenced animal. These sequences were pruned from the assembly prior to submission to Aotearoa Genomic Data Repository. Taken together our analyses indicate that while the assembly is fragmented due to the presence of repeats and reliance on Illumina short fragment libraries, the vast majority of sequence content is present in the assembly and the assembly can act as a valuable reference for reduced representation sequencing datasets.

RADSeq Sequencing

RADSeq sequencing yielded ~900 million single-end 101 bp Illumina sequences (178 524–4 830 235 reads per individual). Since reference-based RADSeq analyses were found to outperform de novo approaches (Rochette et al. 2019), we used the Stacks2 reference genome pipeline that resulted in 372 081 genotyped loci across the 295 samples (258 larvae and 37 adults), with 31.42% of variants scored as missing. The Stacks2 dataset used for downstream analyses consisted of 16 672 SNPs characterized across 186 samples with 17 samples removed due to missing data and 92 removed through relatedness filters. Each of the 92 identified sibling-pairs were collected from the same sampling site, of which 6.28% proportion of variants were scored as missing, and ranged from 4 to 45 individuals per location (Supplementary Table S1).

The Ipyrad de novo assembly pipeline resulted in an initial dataset consisting of 290 pouched lamprey (258 larvae and 32 adults), with 5 individuals removed due to missing data, 115 407 variants, and 62.1% of variants scored as missing. Following additional filtering and pruning of sibling-pairs, the final Ipyrad dataset consisted of 8884 SNPs (with 6.65% of variants scored as missing) across 202 individuals. When datasets were subject to the same filtering methods (e.g., retained reads with minor allele count >2; data not shown), downstream analyses were concordant between the Stacks2 and Ipyrad datasets. Given this concordance, we present the results based on our Stacks2 dataset for the remainder of this article.

Population Structure

The data used for Structure consisted of 186 individuals and 16 629 SNPs. Preliminary Structure runs identified 0.3544 as the optimal lambda value, which was then fixed for subsequent runs. As K = 1 had the highest likelihood, the Evanno method could not be implemented, and unsurprisingly, structure plots for higher levels of K revealed admixture over all sampling locations (Figure 4).

Figure 4.

STRUCTURE v2.3.4 (Pritchard et al. 2000) plots of the final filtered Stacks2-assembled RADSeq dataset (see text) for pouched lamprey from 10 sites in Aotearoa New Zealand. Plots for each of the 7 genetic clusters (K) are displayed as rows and the 10 sampling sites are displayed as columns. Colors represent different genetic clusters. “N” and “S” denote North and South Island sites respectively. K = 1 (not shown) had the highest likelihood and therefore the Evanno method could not be implemented.

A PCA also demonstrated little separation of sampling site locations when plotted against PC1 and PC2, explaining only 0.304% and 0.302% of the genomic variance, respectively (Figure 5). However, PCA plots of individuals colored by gradients in latitude and longitude suggested slight separation along PC1 (Figure 5). This trend appears to be driven by subtle population structure between the islands, and nonoverlapping locations, as no association was found with latitudinal and longitudinal coordinates when the dataset was broken into each island and analyzed separately (Supplementary Figure S4). No structure was seen by collection years and life stages, however because collection years and life stages were confounded with sampling location, we cannot state unequivocally that such structure does not exist (Supplementary Figure S5).

Figure 5.

Principal component analysis (PCA) plots of pouched lamprey from 10 sites in Aotearoa New Zealand colored by (A) sampling location, (B) latitudinal position, and (C) longitudinal position. Symbols “(N)” and “(S)” represent South Island and North Island, respectively. Ellipses = 95% confidence.

De novo DAPC analyses of population structure using an optimal a-score (55 principal components retained, 9 discriminant functions) identified the lowest Bayesian Information Criterion (BIC) value (970.0136) at k = 1, thus suggesting the data fits best in a single cluster. Plots of discriminant function one also demonstrated a single contiguous cluster (no gaps between clusters of individuals). However, subtle structure was observed in DAPC scatter plots when individuals were assigned to sample sites; Piha Stream—the most northerly of our sample locations—showed slight separation except from the North Island locations of Mangakino Stream and Mokau River (Figure 6). The 12 most admixed individuals (defined as those having no more than 0.5 probability of membership to any group) were from Waiau, Pauatahanui, Kaniwhaniwha, Waikawa, Pourakino, and Okuti (Supplementary Figure S6). Of all the sampled locations, the least amount of admixture appeared in Piha Stream and the Styx River (Supplementary Figure S7).

Figure 6.

Discriminant analysis of principle components (DAPC) plot of pouched lamprey from 10 sites in Aotearoa New Zealand colored by sampling location. A plot of the cumulated variance explained by the eigenvalues is positioned in the top left corner. A plot of the eigenvalues that were retained is positioned in the bottom right corner. Sampling sites are listed in a latitudinal manner with the most northern site at the top left and the most southern site at the bottom right.

We did not detect isolation by distance (IBD) using a Mantel test between coastal river entrances corresponding to the 10 sampling sites (allelic data per individual: Mantel r = 0.0005 and 2-tailed P-value = 0.9730, and pairwise F_ST between populations: Mantel r=0.1606 and 2-tailed P-value = 0.2220). In agreement with the Mantel test results, estimated effective migration surfaces (EEMS) plots of distances between demes and observed dissimilarity between demes suggested a poor linear relationship (R² ~ 0.20–0.48), that is, that spatial data patterns were poorly explained by IBD.

Results from the EEMS analyses demonstrated higher than average historical gene flow between some sampling sites. It is ideal for the MCMC chains of the EEMS individual runs to converge upon one posterior value, and all of our runs converged around a log posterior value of ~88 580. Higher than average migration rates were consistently seen between the southernmost sites (Waiau, Pourakino, and Waikawa; Figure 7 with mean migration rates log[m] > 1), and were statistically significant. Higher than average migration was also seen for the 4 most northerly sites (Piha, Kaniwhaniwha, Mangakino, and Mokau; Figure 7 with mean migration rates log[m] > 0.5), albeit with less support: posterior probability Pr{m > 0 | D} exceeded 90% for 7 out of the 12 runs. The “central” sites (Pauatahanui Stream, Styx River, and Okuti River) did not have consistent migratory corridors or migration barriers across all runs. Scatter plots of between-deme and within-deme genetic dissimilarity components, however, suggested a poor fit of the EEMS model to our data. A strong linear relationship was expected between the observed and fitted values yet was only occasionally produced (R² ~ 0.86–0.94 within demes, R² ~ 0.08–0.73 between demes).

Figure 7.

EEMS posterior mean migration rates m (on the log10 scale) for pouched lamprey in Aotearoa New Zealand. Green indicates higher than average migration and brown indicates lower than average migration. Black circles represent deems consisting of one or more sampling sites and are proportional in size to the number of isolates in the deme. Major marine currents are drawn in blue.

Patterns of Diversity

Locus-specific F_ST estimates ranged from −0.0480 to 0.51940 (mean = 0.0010) among the 10 sampling locations, with only a small number of SNPs yielding higher values (i.e., >0.05) (Supplementary Figure S8). F_ST values averaged over all loci equaled 0.0022 and 0.0018 for Nei (1987) and Weir and Cockerham (1984), respectively. Observed heterozygosity (H_o), observed gene diversity, and inbreeding coefficient (F_IS), averaged over all loci equaled 0.1106, 0.1156, and 0.0434, respectively. Mean pairwise F_ST values among populations ranged between −0.0001 and 0.0067 according to Nei (1987) and −0.00008 and 0.00660 according to Weir and Cockerham (1984) (Table 1). Although we detected statistically significant genetic differentiation among our sampling locations (at alpha = 0.05), the values of F_ST for these comparisons were low indicating very little restriction in gene flow between locations and/or extremely recent shared ancestry between locations. Similarly, no statistically significant genomic differentiation was found among collection years (2015–2017; pairwise F_ST = −0.0117 to 0.0287) or life stages (migrating adults and larvae; pairwise F_ST = 0.0081 adults, −0.0066 larvae).

Table 1.

Pairwise F_ST estimates [Wright (1949), updated by Weir and Cockerham (1984)] of Aotearoa New Zealand pouched lamprey from the 10 collection sites

	Piha (N)	Kaniw (N)	Manga (N)	Mokau (N)	Paua (N)	Styx (S)	Okuti (S)	Waika (S)	Poura (S)	Waiau (S)
Piha Stream (N)
Kaniwhaniwha Stream (N)	0.00153**
Mangakino Stream (N)	0.00000	0.00044
Mokau River (N)	0.00341**	0.00147**	0.00174**
Pauatahanui Stream (N)	0.00323**	0.00075	0.00153**	0.00331**
Styx River (Drain) (S)	0.00505**	0.00320**	0.00314**	0.00598**	0.00336**
Okuti River (S)	0.00660**	0.00422**	0.00247*	0.00721**	0.00124	0.00333*
Waikawa River (S)	0.00346**	0.00217**	0.00291**	0.00338**	0.00153**	0.00152**	0.00269**
Pourakino River (S)	0.00310**	0.00187**	0.00185**	0.00300**	0.00198**	0.00241**	0.00345**	0.00001
Waiau River (S)	0.00187**	0.00081**	0.00028	0.00222**	0.00005	0.00231**	0.00231*	0.00132**	0.00141**

Note: Asterisks denote pairwise F_STP-values (1000 bootstraps): “*” = significant at <0.05, “**” = significant at <0.005. Symbols “(N)” and “(S)” represent South Island and North Island, respectively.

Observed heterozygosity, expected heterozygosity (H_S), inbreeding coefficient, and standardized multilocus heterozygosity (sMLH; takes into consideration the variation of the analysed loci) ranged from 0.1038 to 0.1156, 0.1078 to 0.1181, 0.0218 to 0.0619, 0.96108 to 1.03938, respectively, across populations (Table 2). All metrics were very similar among populations, consistent with the recent shared ancestry/high gene flow suggested by the population structure analyses.

Table 2.

Estimates of genetic diversity for Aotearoa New Zealand pouched lamprey sampling sites

Sampling location	n	H  O	H  S	F  IS	sMLH
Piha Stream (N)	12	0.1111	0.1164	0.0455	1.0016
Kaniwhaniwha Stream (N	25	0.1095	0.1153	0.0508	0.9875
Mangakino Stream (N)	19	0.1106	0.1162	0.0481	0.9963
Mokau River (N)	9	0.1048	0.1078	0.0277	0.9792
Pauatahanui Stream (N)	11	0.1119	0.1162	0.0374	1.0092
Styx River (Drain) (S)	10	0.1087	0.1148	0.0533	0.9795
Okuti River (S)	4	0.1038	0.1106	0.0619	0.9611
Waikawa River (S)	45	0.1118	0.1163	0.039	1.0110
Pourakino River (S)	30	0.1102	0.1153	0.0449	0.9923
Waiau Stream (S)	21	0.1156	0.1181	0.0218	1.0394

Note: H  _O, average observed heterozygosity; H_S, average gene diversities within populations heterozygosity; average Wright’s inbreeding coefficient (F_IS, following Nei [1987]), and standardized multilocus heterozygosity (sMLH) over all loci. Symbols “(N)” and “(S)” represent South Island and North Island, respectively.

When individuals were grouped by sampling location (10 sites) N_e was estimated for all populations as negative or “infinite.” NeEstimator sampling error is calculated from unbiased estimators using a known sample size, and a larger than expected sampling error can occur by chance resulting in a negative estimate of N_e. When a negative estimate of N_e occurs, it is often interpreted as “infinite,” or that “there is no evidence for variation in the genetic characteristic caused by a finite number of parents—it can all be explained by sampling error” (see Do et al. 2014). When considering all sampled pouched lamprey as a single population, N_e was estimated at 3326.7 for the <0.05 allele frequency (95% confidence interval: 3186.2–3480.0) and 8066.6 when all alleles were used (95% confidence interval: 7791.1–8362.1). When each island was assessed as a single population (North and South), N_e was estimated at 2847.2 for the <0.05 allele frequency (95% confidence interval: 2615.8–3123.1) and 8917.6 when all alleles were used (95% confidence interval: 8101.9–9915.7) for the North Island, while N_e was estimated at 4111.4 for the <0.05 allele frequency (95% confidence interval: 3758.2–4537.3) and 11535.3 when all alleles were used (95% confidence interval: 10569.9–12694.4) for the South Island.

Outlier Loci

OutFLANK identified 12 potential adaptive SNPs from all sampling locations (Supplementary Figure S9). Principal component (PC) analyses of the SNPs revealed a lack of structure as was observed for neutral loci. The NCBI blast analysis of the 12 outlier sequences recovered 5 unique hits (Supplementary Table S10). These included a beta-1,3-galactosyl-O-glycosyl-glycoprotein beta-1,6-N-acetylglucosaminyltransferase (GCNT1)-like miscRNA that is possibly involved in protein modification, a sorting nexin-6 (SNX6)-like mRNA possibly involved in intracellular trafficking, a zinc finger MYM-type protein 3 (ZMYM3)-like miscRNA possibly involved in the regulation of cell morphology, a solute carrier organic anion transporter family member 1C1 (SLCO1C1)-like mRNA possibly involved in the transport of organic anions, and a ruthenus protein promyelocytic leukemia (PML)-like mRNA possibly involved in apoptosis/senescence/DNA damage response/viral defence (UniProt Consortium 2021).

Historical Demography

Demographic reconstructions of the single pouched lamprey population using CubSFS suggested that there was a recent decline in the effective population size (N_e) since the last glacial maximum (LGM, ~20 000 years ago). CubSFS runs testing variations in the knots and coalescent units all produced plots with peak N_e around the LGM. The 0–30 000 year plot created using 29 knot with a t_m of 0.35 coalescent units had knots at 0.00, 20605.49, and 41433.01 years before present and thus appeared to provide adequate resolution for the assessment of the N_e around the LGM (Supplementary Figure S11).

Discussion

Determinants for the Lack of Differentiation: Gene Flow and/or Recent Colonization

We assessed population connectivity, historical migration, and population size of pouched lamprey from Aotearoa New Zealand using single-gene and RADSeq data in order to guide management of pouched lamprey and similar “elusive” species. Both our single-gene and RADSeq datasets demonstrated a lack of strong population structure within Aotearoa New Zealand, which is not surprising given that several anadromous Northern Hemisphere lamprey species also show limited population structure (e.g., Bryan et al. 2005; Goodman et al. 2008; Yamazaki et al. 2014; Bracken et al. 2015). Similar to Northern Hemisphere lampreys, the lack of structure in New Zealand pouched lamprey is likely a consequence of gene flow and/or recent colonization.

Northern Hemisphere anadromous lampreys are thought to be non-philopatric (do not home to natal streams; Bjerselius et al. 2000; Waldman et al. 2008; Spice et al. 2012; Moser et al. 2015; Mateus et al. 2021), which means they generally show increased gene flow and weak population structure relative to philopatric species (e.g., salmonids; Bryan et al. 2005; Goodman et al. 2008; Yamazaki et al. 2014; Bracken et al. 2015). Given the lack of population structure detected in our dataset, natal philopatry is also unlikely for pouched lamprey in Aotearoa New Zealand. Reduced population structure amongst pouched lamprey and other lampreys may also arise from behavioral, morphological, and reproductive traits that increase the chances of gene flow across a wider geographic range. For instance, reduced population structure may occur through nonspecific highly mobile host/prey choice, or through continuous selection for individuals with larger body sizes (possibly associated with longer marine migrations).

Colonization of pouched lamprey from a recent ancestral population may also have led to the lack of overall variation seen across Aotearoa New Zealand. At the Last Glacial Maximum (LGM; 18–26 kyr; Shulmeister 2017), if pouched lamprey were excluded from most of New Zealand due to environmental barriers such as glaciers (permanent ice) or steepened sea surface temperature gradients (Lorrey and Bostock 2017), and then recolonized New Zealand when conditions became more favorable, we might expect to see the high population genetic homogeneity we observe. This scenario, however, seems unlikely for 2 reasons. First, pollen, macrofossil, insect, and geomorphic evidence suggests that permanent ice did not cover all of New Zealand during the LGM; a large majority of the vegetation was shrubland-grassland, especially on the coastlines which extended further than the present day (Wood et al. 2017). Our historical demographic reconstructions suggest that effective population size of pouched lamprey peaked around the LGM, thus it may be that pouched lamprey in New Zealand were benefited by the expanded coastlines, widespread bogs, and more prevalent shrublands. Second, population genetic studies of lamprey have demonstrated that genetic differentiation can occur in less than 200 years. For example, studies using microsatellite loci demonstrated that significant genetic differentiation occurred between sea lamprey populations in the lower and upper Great Lakes after the opening of the Welland Canal in 1892 (Bryan et al. 2005). If pouched lamprey had roughly similar migration rates between populations, it would be expected that at least weak population structure would have developed over the 20 000 years since the LGM. Unfortunately, pouched lamprey fossil data are lacking from New Zealand, and many other areas, since lampreys lack ossified anatomical structures and do not fossilize well. Future studies that evaluate the population genomics and divergence times of Geotria across the Southern Hemisphere may help determine if pouched lamprey are a recently introduced population to Aotearoa New Zealand and whether recent colonization has contributed to the observed lack of differentiation in pouched lamprey.

Migration and Dispersal in the Marine Environment

The lack of genetic structure between the North and South Islands of Aotearoa New Zealand is indicative of ongoing gene flow or recent colonization (see previous section). This suggests pouched lamprey are able to migrate between islands and overcome hurdles in the marine environment, such as strong currents, steep temperature gradients, and upwelling (Vincent et al. 1991; Stevens et al. 2012; Chiswell et al. 2017) either through prey attachment or free swimming behaviors. This dispersal capability is further supported by observer and commercial at-sea data, as well as museum vouchers that demonstrated that New Zealand pouched lamprey may travel at least 400 km from the New Zealand shoreline (Fisheries New Zealand Official Information Act request dated 4 March 2019 and A. Miller 2019, Personal Communication with Auckland War Memorial Museum curators, 27 May 2020). Suspected prey species of lampreys (e.g., cetaceans; Pike 1951; McDowall 1990; Miočić-Stošić et al. 2020) are thought to make long-range migrations across New Zealand and pouched lamprey may make long-range migrations with them by attaching to them as “hitchhikers.”

However, despite the overall suggestion of a largely panmictic metapopulation, we detected some subtle neutral-loci population structure. Marine currents around New Zealand (Figure 2 inset) may also influence the dispersal ability of pouched lamprey—either when free-swimming, or due to the impacts on prey that they attach to, or “hitchhike,” with. Piha Stream and the Styx River appear to have lower admixture which may be a result of converging currents, thus possibly reducing the gene flow to and from these locations. The number of migrants, not the proportion of migrants (demographic connectivity), determines the genetic connectivity of a population, regardless of population size (Lowe and Allendorf 2010). Thus, populations can sometimes be demographically uncoupled even if they have small F_ST values. Consequently, although philopatry has not been shown to occur in other lampreys it cannot be ruled out entirely as a factor for the subtle structure seen in pouched lamprey, potentially resulting from environmental preferences, or limited breeding habitats in the streams available to them. For example, possible anthropogenic stressors on streams located on the North Island may explain the slight structure observed between islands. However, a limited amount of suitable breeding habitat does not appear to be supported in the Auckland region (North Island) as there are streams containing suitable habitat in the region, yet semi-extensive surveys have determined that pouched lamprey are only found in select streams within these regions. Furthermore, philopatric anadromous species such as sturgeons, American shad, and salmonids generally appear to have higher pairwise F_ST values across neutral loci (e.g., Hasselman et al. 2013; Whitaker et al. 2020; White et al. 2021), and our neutral loci values between pouched lamprey sampling sites were low (−0.0001 to 0.0067) even for anadromous lampreys with comparable geographic ranges (Table 2 in Mateus et al. 2021). Thus, we believe philopatry is unlikely in pouched lamprey.

Adaptive variation could also be an explanation for the subtle neutral-loci population structure through selection for phenotypes (e.g., total length and weight) that are more fit for environments with certain temperature, flow, and lunar conditions, as distributional differences of Northern Hemisphere species and G. australis have been observed associated with variation in these environmental parameters (Macey and Potter 1978; Sloane 1984; Moser et al. 2015; Potter et al. 2015; Golovanov et al. 2019; Arakawa and Yanai 2021; Miller et al. 2021). Our outlier analyses recovered 12 potential adaptive SNPs and suggested that 5 genes, involved in multiple cell functions, may be contributing to the subtle structure of the neutral loci. Previous RADSeq studies have found SNPs that localized to genes of multiple functions in Lampetra and Entosphenus species (Hess et al. 2013, 2014; Mateus et al. 2013; Rougemont et al. 2017). Genes predicted to be involved in development and body morphology in particular, appear important to the genomic diversity of lampreys (Hess et al. 2020). Here, we find possible commonality for several genes and gene functions. The genes BAX/CASP7 and SLC35F5 (solute carrier) that were associated with apoptosis and transmembrane functions, respectively, were discovered as outliers in Entosphenus tridentatus RNASeq data (Hess et al. 2013), and may hold similar functions to the PML and SLCO1C1 (solute carrier) genes also associated with apoptosis and membrane transport from our putative adaptive loci. Additional genes associated with adaptive loci in E. tridentatus, such as ZNF385D and PCDH15, were identified later in E. tridentatus (Hess et al. 2014) and may be similar to our zinc finger protein (ZMYM3) and cell adhesion (GCNT1) genes. In addition, immune genes were associated with RADSeq sequence differences between Lampetra fluviatilis and L. planeri (Mateus et al. 2013; Rougemont et al. 2017) and our PML gene may be playing a similar immunological role. Immunity genes such as PML, GCNT1, as well as others, may prove important for future studies that aim to better understand lamprey syndromes and diseases such as LRS. Overall, since only 12 SNPs were recovered it difficult to determine how informative the SNPs are to the structure of the population in Aotearoa New Zealand. Future studies that include joint genetic sampling and morphological traits, such as those seen in association analyses by Hess et al. (2020) on Pacific lamprey (E. tridentatus), will help determine if selection could be impacting population structure in pouched lamprey.

More broadly, colonization history and rates of intercontinental dispersal of Geotria in the Southern Hemisphere are uncertain. Presently our understanding of the systematics of the Southern Hemisphere lampreys is based on morphological characters and mitochondrial DNA markers.

Mitochondrial DNA markers and morphological characters suggest a deep split between Atlantic G. macrostoma sampled from Argentina and Pacific G. australis (Nardi et al. 2020; Riva-Rossi et al. 2020). Within G. australis, Chilean populations show mitochondrial sequence distinctions from Australasian populations, and within Australasia, Western Australian populations appear genetically differentiated from Aotearoa New Zealand, Tasmania, and South Australian populations (this study [Figure 3] and published works; Nardi et al. 2020; Riva-Rossi et al. 2020). Although limited to 3 Chilean samples, the genetic differences of the Chilean samples from the Australian and New Zealand samples suggest, at least from mtDNA, that little to no migration occurs between the Pacific and Atlantic coasts of South America, that trans-Pacific migration is rare, and that migration between Western Australia and other Australasian locations may also be restricted.

In this study, we aimed to better understand the population structure of pouched lamprey within Aotearoa New Zealand and to better understand how they are related to other Southern Hemisphere pouched lamprey. RADSeq data and mtDNA data were used, respectively. RADSeq data have been shown to be more suitable for examining historical relationships and contemporary fine-scale gene flow (e.g., Longo et al. 2020), while mtDNA is suitable for examining larger-scale phylogenetic relationships. An investigation into the finer-scale population structure of Southern Hemisphere is warranted and will require nuclear gene regions to better understand these relationships and resolve if dispersal and/or vicariance explains the current distribution of Geotria in New Zealand and in the Southern Hemisphere.

Management and Conservation of Pouched Lamprey and Other Elusive Species

Estimating population size, habitat preferences, behaviors, population structure, and migration corridors is challenging for elusive species. This study, however, illustrates the benefits of genetics for acquiring conservation-relevant information for these species.

Elusive species are those that have a low probability of detection and include both abundant and rare species. The reasons for their low probability of detection vary by species, however, the impacts this has on the management and conservation of rare species is often universally detrimental. Since elusive species are under-detected by surveying and monitoring equipment they often get counted incorrectly, or not counted at all (undefined/“NA”) in population censuses. This lack of detection also creates large knowledge gaps about their preferred environments, inter- and intra-species behaviors, and migrations/movements, among others. For example, many threatened tropical mammals are poorly researched due to the difficulties detecting them (Jambari et al. 2019). This even includes large species such as the tiger (Panthera tigris) in India where monitoring and surveying difficulties have led to decades of misleading information and poor conservation practices (Karanth et al. 2003). Large elusive fishes, such as eels are also considered difficult to sample quantitatively since species such as the European eel (Anguilla anguilla) are thought to avoid detection during electrofishing and trapping surveys, leading to inadequate abundance estimates (Degerman et al. 2019). As one of these elusive species, pouched lamprey are routinely under-detected in general stream surveys using standard methods (e.g., electrofishing, spotlighting, and trapping) due to their secretive habitats and tendency to migrate on receding flood waters. For instance, standard electrofishing pulse settings are often inadequate for pulsing large adults or sedentary larvae (Joy et al. 2013). This has made it difficult for pouched lamprey managers, stakeholders, and kaitiaki to accurately infer important population demographics and implement robust conservation policies. Here, we show that molecular methods such as RADSeq can increase our understanding of the elusive pouched lamprey and can be used on other elusive species, such as eels, to detect similar important population structure and migration information. Molecular methods such as RADSeq often require only a limited amount of physical sampling—sometimes only a single sampling event is needed. Thus, unlike most standard monitoring methods that require the routine collection of individuals, these molecular methods are much more feasible for elusive species. In addition, the samples/tissues collected for molecular analyses can often be stored and used for multiple studies across multiple disciplines, such as genetic and isotope analyses. The information gained from these molecular methods can then be used to improve conservation initiatives and assist future management decisions (see below).

The lack of genetic structure in our data suggests the presence of ongoing gene flow between regions within Aotearoa New Zealand. This has ramifications for potential future translocation efforts between neighboring streams as well as sourcing individuals for potential propagation programs. Multiple case studies, led by tribal and government agencies, of Northern Hemisphere lampreys have demonstrated the success of artificial propagation programs, and have shown that successful programs can be created by reusing existing hatchery facilities with some adjustments (Lampman et al. 2021). These methods are transferrable to pouched lamprey; however, both translocations and propagation programs would warrant considerable research, careful consultation, and be investigated in collaboration with iwi partners, to ensure cultural values are adhered to and that any genetic variation, even if subtle, is maintained (such as those protocols created for Pacific lamprey, see Hess et al. 2014; Lampman et al. 2016).

Geotria australis is known to have LRS and hemorrhagic septicemia in Aotearoa New Zealand and Australia, respectively. In New Zealand, pouched lamprey were first seen with LRS, a syndrome causing external hemorrhaging and body reddening, in Southland rivers in 2011 and it has since been annually documented in rivers of the South Island. The cause of LRS is unknown and most documented cases were lethal or thought to be associated with mass mortalities (Brosnahan et al. 2018). A lack of organized reporting has made monitoring LRS difficult, and there is concern that LRS may spread across cohorts (Kitson et al. 2012) and to the North Island. Our results suggest that if LRS is not associated with particular environmental conditions of Southland, that is, if it is transmittable to other individuals and or has a genetic basis for susceptibility, that it will likely be seen in North Island individuals since gene flow appears to occur between the North and South Islands. Thus, LRS should be the subject of ongoing monitoring and North Island managers should be especially observant for it.

In addition to genetic structure, we also investigated the estimated effective population size (N_e) of New Zealand pouched lamprey, finding our N_e estimate of a single population to be ~3327–8067. This was lower than the lowest N_e estimates, 50 000, for European sea lamprey (Almada et al. 2008) that were sampled on an approximately similar geographic scale. However, the European sea lamprey N_e estimates were inferred using a different method than ours: the control region (d-loop) with the estimate of diversity index (=2N_eμ). N_e estimates of other lamprey species have been inferred using similar methods to ours (e.g., RADSeq data and NeEstimator), but these estimates were focused on much smaller geographic scales such as creeks and rivers (e.g., Rougemont et al. 2017; Sard et al. 2020). N_e between islands in Aotearoa New Zealand, however, did meet expectations. Greater N_e estimates were found to occur in the South Island than in the North Island, which is similar to anecdotal observations that pouched lamprey are less abundant in the North Island compared to the South Island. New Zealand pouched lamprey are considered “Threatened—Nationally Vulnerable” since they were predicted to occupy ≤100 ha (1 km²) and to decline by 10–50% over 3 generations (qualifiers: “data poor” and “secure overseas”; Goodman et al. 2014; Dunn et al. 2018). Direct comparisons of N_e to adult census size (N_c) are problematic because most natural populations have some degree of population structure, lack discrete generations, and do not maintain a constant population size. For example, effective population size estimates of Chinook salmon were commonly underrepresented due to their fluctuating population size between years and since only a few individuals will produce many offspring; Shrimpton and Heath 2003). However, N_e analyses, such as the one performed in this study, will be useful tools for monitoring pouched lamprey N_e through time. This monitoring may act as a rough proxy for monitoring the expected declines in N_c and the successes of specific conservation measures such as translocation and propagation programs.

Conclusions

Here, we used RADSeq and targeted markers to improve our knowledge of the population structure and genetic variation of pouched lamprey (G. australis) in Aotearoa New Zealand: the first genome-wide population study of any Southern Hemisphere lamprey species. Our mitochondrial DNA (Cyt-b and COI) analyses and RADSeq data revealed that gene flow occurs throughout Aotearoa New Zealand and that this lack of population structure may be influenced by recolonization following the last glacial maximum. However, our historical demographic reconstructions differ from those expected under a recolonization scenario and thus support the hypothesis that long-distance migration explains the lack of structure in New Zealand pouched lamprey. We recovered fine scale structure between sites and along geographical gradients (latitudinal and longitudinal); however, this was not reflected in our putative outlier analyses. Our isolation by distance tests were insignificant, suggesting other potential factors such as philopatry and ocean currents may be subtly influencing pouched lamprey gene flow. The techniques used in this study can be applied broadly to other elusive species (e.g., eels) to better understand their distribution, effective population size, and population structure. This structure and abundance information will be highly useful for ongoing pouched lamprey monitoring, conservation, and disease-management strategies.

Funding

This work was supported by a Ministry of Business, Innovation, and Employment New Zealand contract C01X1615 to C.B. and administered by NIWA; the University of Otago; a University of Otago Doctoral Scholarship to A.K.M.; a National Institutes of Health Minority Health International Research Training Programme run through the University of California Santa Cruz to N.J.G. and A.K.M.; and a Ministry of Business, Innovation & Employment’s Research Infrastructure programme grant that supported the New Zealand eScience Infrastructure (NeSI) high performance computing facilities utilized in this research.

Data Availability

We have deposited the primary data underlying these analyses as follows: The RADSeq DNA sequences and targeted DNA sequences are available in Aotearoa Genomic Data Repository at https://repo.data.nesi.org.nz/TAONGA-LAMPREY/collection/TLCMC00005-00002. The draft genome assembly is available in Aotearoa Genomic Data Repository at https://repo.data.nesi.org.nz/TAONGA-LAMPREY/collection/TLCMC00005-00003. All data will be shared upon reasonable request.