Two Distinct Maternal Lineages of Threespine Stickleback (Gasterosteus aculeatus) in a Small Norwegian Subarctic Lake

Abstract

A subarctic Norwegian watercourse is known for its population of threespine sticklebacks (Gasterosteus aculeatus Linnaeus 1758) with unusual pelvic spine polymorphism; the upper lake contains a mixture of specimens that are normal-spined, asymmetric-spined, and spineless. In contrast, the downstream lakes and the nearby marine site contain only the normal spined morph. We investigated the maternal-line genetic structure in the watercourse using mitogenomics. A total of 242 sticklebacks representing two lakes and a marine site were assessed based on individual and pooled DNA sequencing. While two distinct mitogenome clades were detected in the upper lake (Lake Storvatnet), only one of these clades was present in the downstream lake. The marine site pooled DNA sample, however, contained several mitogenome haplotypes. We present mitogenome sequence features that include gene-specific single nucleotide polymorphisms, molecular phylogeny, and genetic differentiation assessments based on pairwise comparisons of pooled population samples from each site. The Lake Storvatnet mitogenomes belong to two distinct Euro-North American (ENA) clades; one of the ancestral lineages likely corresponds to the original maternal lineage in the watercourse, and the other was introduced more recently. We hypothesize that the second invader carried nuclear genomic features responsible for the observed present-day pelvic spine polymorphism in Lake Storvatnet.

1. Introduction

The threespine stickleback (Gasterosteus aculeatus), which represents a well-established fish model system in phenotypic evolution and evolutionary genomics, has been reported to repeatedly colonize freshwater lakes across the northern hemisphere since the last ice age [1]. Transitions from marine to freshwater environments have resulted in notable morphological changes in body armor, such as reduction or loss of lateral plates and pelvic spines, which represents a classic example of parallel evolution. The genetics of body armor reduction has recently been investigated in more detail, including the molecular mechanism of pelvic reduction [1]. These studies, based on pelvic-reduced stickleback from Paxton Lake in British Columbia, Canada, revealed interesting mutation features linked to the PelA enhancer region of the Pitx1 transcription factor gene [2,3,4].

Pelvic-reduced stickleback morphologies have been observed in northern Europe, and one of the best-studied examples is from the Vesterålen archipelago in subarctic Norway. Here, Klepaker and co-workers reported complete or partial loss of pelvic spines in 60% of the stickleback population in a small lake (Lake Storvatnet) on the island of Langøya [5,6,7]. Recently, we performed molecular genetic analyses of the Lake Storvatnet population, with a focus on Pitx1 enhancer structures [8]. While interesting mutational features were observed in the PelA enhancer, there was no clear association with pelvic status, suggesting a more complex genetic mechanism compared to that of the Paxton Lake sticklebacks.

To gain more information about the genetic background, and probably also the long-term evolutionary history of Lake Storvatnet sticklebacks, we analyzed complete mtDNAs and reconstructed maternal lineage relationships using the mitogenome as a molecular marker. Different pelvic spine morphs were included in this study, as follows: normal-spined, spineless, and asymmetric-spined. The asymmetric group consists of specimens with >0.2 mm length difference between the two pelvic spines. In addition, we investigated stickleback mitogenomes from a downstream lake in the same watercourse and a nearby marine site. The latter two sample sites contain normally spined sticklebacks only [8].

MtDNA appears as a suitable and applicable genetic marker in vertebrates due to its maternal inheritance and multi-copy nature [9]. The structure, gene content, and function are highly conserved and involve 37 canonical genes [10,11,12,13]. The derived gene products represent a subset of hydrophobic proteins essential to oxidative phosphorylation, two rRNAs of the highly specialized mitochondrial ribosome, and a complete set of 22 tRNAs important for decoding and protein synthesis. The threespine stickleback has previously been investigated using specific mtDNA markers at the population level. Most studies include cytochrome b gene sequences in combination with the non-coding control region [14,15,16,17]. More recently, complete stickleback mitogenome sequences have become available and applied in taxonomic analyses. Among these are US specimen isolates from Alaska and Wisconsin [18] and complete mitogenomes derived from whole-genome sequencing datasets of sticklebacks sampled in Norway, Denmark, and Japan [19,20]. A recent mitogenomic analysis of the threespine stickleback revealed two main clades designated Trans-North Pacific (TNP) and the Euro-North American (ENA) [20]. In the current study, we investigated mitogenome molecular features and mitogenome diversity among sticklebacks from Lake Storvatnet, a downstream lake, and a nearby marine site by single specimen and pooled specimen datasets. The analyses revealed that Lake Storvatnet sticklebacks harbor mitogenome haplotypes that belong to two divergent ENA clades.

2. Materials and Methods

2.1. Fish Sampling and Morphology Scores

Sticklebacks were sampled at three locations in Vesterålen, Northern Norway (Figure 1). A total of 162, 40, and 40 specimens were collected and included in the mitogenome analyses from Lake Storvatnet, Lake Gjerdhaugvatnet, and a nearby marine site, respectively. Specimens were assessed for pelvic spine morphology features. All sticklebacks from Lake Gjerdhaugvatnet and the marine site were normal-spined. Lake Storvatnet sticklebacks (160 specimens) were divided into four population samples (40 specimens in each) based on their pelvic spine morphology, as follows: normal-spined, asymmetric-spined, spineless, and mixed-spined. The latter group (mixed-spined) was included as a representative population sample from Lake Storvatnet, consisting of 11 normal-spined specimens, 19 spineless and spine-reduced specimens, and 10 asymmetric-spined (left and right) specimens. Thus, a total of six population samples, each consisting of 40 specimens, were targeted for pooled sequencing. In addition, mitogenomes from two Lake Storvatnet specimens (one normal spine and one spineless) were sequenced individually.

2.2. DNA Isolation and DNA Pooling

DNA was isolated from a muscle sample anterior to the caudal fin. Tissue was homogenized using bead beating in DNA/RNA Shield solution (Zymo Research, Irvine, CA, USA) using a Dremel 8220 rotary tool (MP Biomedicals, Solon, OH, USA), as previously reported [8]. DNA was extracted from homogenized samples either with the MasterPure^TM DNA Purification kit (Biosearch Technologies, Hoddesdon, UK) or the Monarch genomic DNA purification kit (New England Biolabs, Ipswich, MA, USA) following the standard protocols described in the kit manuals. The quality and concentrations of DNA in samples were further measured as previously described [8]. The final total DNA concentration of each specimen sample was ≥20 ng/µL. Six pooled population samples from the three locations, Lake Storvatnet, Lake Gjerdhaugvatnet, and the marine site, were prepared, each pool consisting of an equal amount of total DNA (100 ng) of each of the 40 specimens.

2.3. DNA Sequencing

Two specimens from Lake Storvatnet, one normal-spined (STV-Ida31) and one spineless (STV-Ida61), were subjected to whole-genome sequencing based on the Ion PGM^TM system and Ion^TM 318 v2 chips. Sequencing libraries were constructed from approximately 400 bp DNA fragments, and Ion Torrent sequencing was performed essentially as previously described [21]. About 47 billion and 102 billion sequence nucleotides were obtained from STV-Ida31 and STV-Ida61 libraries, respectively, with average read lengths of 239 bp and 328 bp. The pooled DNA samples were shipped to the Norwegian Sequencing Center, University of Oslo, for whole-genome sequencing using Illumina technology. Sequencing was performed on an Illumina NovaSeq instrument generating paired-end reads of 150 bp.

2.4. Mitogenome Assembly and Analysis

The mitogenomes were assembled using QIAGEN CLC Genomics Workbench 23 (https://digitalinsights.qiagen.com/, accessed on 8 July 2024), and reads were mapped to a published stickleback reference sequence (MH205729 from Bear Pew Lake, AK, USA). The average mitogenome coverages of STV-Ida31 (16,545 bp) and STV-Ida61 (16,540 bp) correspond to 72 times and 154 times coverages, respectively. All positions were unambiguously determined. Each pooled DNA sample generated 796 to 1146 million paired Illumina reads. In the PoolSeq analysis, mitochondrial reads were mapped to STV-Ida31 as the reference sequence. Approximately 14,000 times coverages of the mitogenome were estimated in each pool sample, which correspond to an average coverage of 350 times for each specimen. The Alta-region mitogenome sequences were assembled using a synthetic pool creation approach, based on downloaded whole genome reads from the Sequence Read Archive under the BioProject accession number PRJNA693136 [19]. Three different locations were assessed, including one marine site (Altafjord) and two lake sites (Lake Klubbvatnet and Lake Jossavannet). Ten samples were downloaded for each location. For each location, the forward and reverse FASTQ files from downloaded samples were concatenated to create synthetic pools. Concatenated FASTQ files were then aligned to the mitochondrial reference genome STV-Ida31 using the Spliced Transcripts Alignment to a Reference (STAR) aligner (https://github.com/alexdobin/STAR, accessed on 8 July 2024). The mapping produced a sorted BAM file for each synthetic pool, enabling subsequent analyses of location-specific mitochondrial variations. Sequence information of the Alta-region mitogenomes is given in Table S4. Two mitogenomes from Limfjord (Denmark) and two from Akkeshi River (Japan) were assembled from whole genome sequencing data [20] available at Sequence Read Achieve (SRA) under accession numbers ERR104980, ERR104981, DRR032282, and DRR032292, respectively.

2.5. Molecular Phylogeny

Molecular phylogeny was inferred by using the tree-building method in QIAGEN CLC Genomics Workbench 23 (https://digitalinsights.qiagen.com/, accessed on 8 July 2024) based on maximum likelihood. The sequence alignment includes 15,650 nucleotide positions representing the complete mitogenome, excluding the control region. The sequence alignment was model-tested prior to tree construction using QIAGEN CLC Genomics Workbench 23. The best-fit model was general time reversible (GTR + G + T), and tree topologies were evaluated through bootstrapping (1000 bootstrap replicates).

2.6. Genetic Differentiation Statistics

Genetic differentiation between mitogenome populations was quantified using F_ST statistics [22]. F_ST values range from 0 to 1, where smaller values indicate lesser differentiation between populations while larger values indicate allele frequencies that are highly differentiated [23]. F_ST assessments were performed on pairwise pooled samples, essentially as described by [24]. A sliding window strategy of 30 nt with 15 nt overlap was used that included the F_ST-sliding.pl Perl script, which is part of the PoPoolation2 software toolkit (version 2_1201) [25]. The F_ST data were imported into R (version 4.2.0, R studio v 1.4.1717) for graphical population comparisons and downstream analysis.

3. Results

3.1. Mitogenome Sequence of the Reference Specimen STV-Ida31

Complete mitogenome sequences of two G. aculeatus specimens (fully spined STV-Ida31 and pelvic spineless STV-Ida61) from Lake Storvatnet, Norway (Figure 1), were determined using the Ion Torrent technology. The sequences of STV-Ida31 and STV-Ida61 were found to be 16,545 bp and 16,540 bp, respectively, and possessed a gene content and organization typical of vertebrates (Figure 2A). STV-Ida31 was selected as a reference sequence in our study and thoroughly annotated according to vertebrate mitogenome convention, conserved mRNA polyadenylation sites, and tRNA and rRNA secondary structure folding (Table S1). All 13 protein genes showed TAA as their predicted stop codons, and 7 of these were generated by posttranscriptional polyadenylation. Secondary structure predictions of the mt-tRNAs (Figure S1), mtSSU rRNA (Figure S2), and mtLSU rRNA (Figure S3) followed the typical patterns of teleost mitochondrial RNAs [11,12,21]. The 892 bp control region contained a heteroplasmic tandem repeat (HTR) array consisting of an 11 bp motif that varies both in primary sequence and in copy number [26]. Variants of four copies are shown in the mitogenome sequences (Figure 2A). The HTR array region (corresponding to positions 15,669–15,729 in STV-Ida31) was excluded in downstream comparisons and analyses.

3.2. Mitogenome Comparison between STV-Ida31 and STV-Ida61

The two mitogenomes from Lake Storvatnet differed at 88 nucleotide positions (excluding the HTR array), which show the co-occurrence of two divergent mitochondrial genotypes within a small, presumably isolated lake population (Table S2, columns 1 and 2). Two of the tRNAs were found to have variable sites in their structures (Figure 2B and Table S1). Furthermore, three and four single nucleotide polymorphisms (SNPs) were noted in mtSSU and mtLSU rRNAs, respectively, all located in regions known as variable in vertebrates (Figure 2C, Tables S2 and S3). We detected 67 SNPs in protein-coding regions, and all 13 mitochondrial genes were represented (Table S3). Here, thirteen SNPs were non-synonymous and resulted in conservative amino acid changes (Figure 2A). The remaining 12 SNPs were located in non-coding regions (Table S2, column 2).

To further assess the relationship between the two mitogenome haplotypes, STV-Ida31 and STV-Ida61, we performed a phylogenetic analysis. In addition to the Lake Storvatnet mitogenomes, we included available G. aculeatus mitogenome sequences in GenBank (two from Lake Superior, Wisconsin, USA, and one from Bear Paw Lake, Alaska, USA) and retrieved several mitogenome sequences from available genomic datasets. The latter included a mitogenome consensus sequence (GHV-maj1) from the nearby Lake Gjerdhaugvatnet located downstream in the same watercourse (Table S2, column 3), three Alta-region (Finnmark County, Norway) mitogenomic sequences, two mitogenomes from Limfjord (Denmark), and two from Akkeshi River (Japan) (Table S4). G. wheatlandi (Blackspotted stickleback) was used as an outgroup in the analysis. The alignment was based on complete mitogenome sequences, excluding the control region (CR). A representative maximum likelihood (ML) phylogenetic tree is presented in Figure 3. The phylogenetic analysis showed that the mitogenomes belong to the following two ancient clades, which is highly supported by the bootstrap analysis (100%): the TNP haplotype group represented by specimens from Japan and the ENA haplotype group consisting of specimens from Norway, Denmark, and the USA. Interestingly, the two mitogenome types from Lake Storvatnet clustered separately, one with each of the two main clades of the ENA (supported by a bootstrap of 100%). While STV-Ida31 clusters together with mitogenomes from the nearby Lake Gjerdhaugvatnet and a Lake Superior specimen (Wisconsin, US), STV-Ida61 appears more related to a marine mitogenome from the Alta-region and additional samples from Norway, Denmark, and the USA.

3.3. Pairwise Comparisons of Populations Based on Pooled DNA

We investigated mitogenome diversity in Lake Storvatnet, Lake Gjerdhaugvatnet, and a nearby marine site using an approach based on pairwise comparisons of datasets based on pooled DNA sequencing (PoolSeq). Pooled DNA samples were obtained using equimolar pooling of total DNA isolated from the muscle tissue of 40 specimens from each location (Figure 4), and the pooled samples were sequenced using Illumina technology. Approximately 14,000 times coverage of the mitogenomes was generated for each pool sample, corresponding to 350 times coverage of individual mitogenomes. Lake Storvatnet was originally represented by four different datasets (4 × 40 specimens) based on pelvic spine morphology (normal-spined, asymmetric-spined, spineless, and mixed-spined). Since the initial analysis concluded that the four datasets from Lake Storvatnet were practically identical in mitogenome SNPs and distribution of variants (see Figure S4), we used the mixed dataset as a representative for Lake Storvatnet in the mitogenome analyses below.

When applying a stringent SNP criterion (minimum variant frequency of ≥5%), we identified 82 SNP loci in the mixed dataset from Lake Storvatnet, mainly with two alleles corresponding to that of the STV-Ida31 and STV-Ida61 haplotypes (Figure 4; Table S2, column 4). Furthermore, the SNP allele frequency was largely stable at most sites at an average of 38% for the STV-Ida31 variant and 62% for the STV-Ida61 variant (Figure 4; Table S2, column 4). This result supports that the STV-Ida31 and STV-Ida61 haplotypes represent two clearly divergent clades, which are represented by approximately 15 and 25 specimens, respectively, in the mixed pooled 40 specimen sample from Lake Storvatnet. A similar SNP frequency trend was noted in the three additional pooled samples from Lake Storvatnet.

The SNP frequencies of stickleback mitogenomes from Lake Gjerdhaugvatnet (Figure 4; Table S2, column 3) suggest the presence of only one major haplotype that is strongly affiliated with the STV-Ida31 haplotype (see Figure 3). Only five SNP loci were identified within the dataset, all with two alleles with variant frequencies of approximately 40% (minor) and 60% (major). The Lake Gjerdhaugvatnet mitogenome haplotype (GHV-maj1) was reconstructed and applied in the phylogenetic analysis presented in Figure 3.

More variations were observed within the marine dataset with 101 variable positions compared to the STV-Ida31 sequence (Figure 4; Table S2, column 5). Here, 71 SNPs had two alleles and one had three alleles, but the allele frequencies were highly variable between loci indicating a mix of haplotypes. The marine dataset also contained 11 private SNP loci not found in Lake Storvatnet or Lake Gjerdhaugvatnet (Table S2, column 5). Finally, we note that most SNP loci allele variants (72%) present at Lake Storvatnet also were present in the marine dataset.

3.4. Mitogenome Genetic Differentiation between Sample Sites

Genetic differentiation between Lake Storvatnet datasets and between sample sites was assessed by pairwise comparisons using a window-based method to calculate the fixation index F_ST across the mitogenomes. We used the same setting as previously described for the Atlantic cod mitogenome [24] of sliding windows of 30 nt, with a 15 nt overlap. First, we assessed possible mitogenomic genetic differentiation between pelvic morph specimens in Lake Storvatnet (normal-spined, asymmetric-spined, spineless, and mixed pooled datasets). As expected, no differentiation was detected giving an F_ST value at or close to 0 (Figure S4), which strongly supports that mitogenome haplotypes are not associated with present-day pelvic morphs.

Next, we analyzed the genetic differentiation between the Lake Storvatnet pooled dataset and that of Lake Gjerdhaugvatnet (Figure 5A). Most positions were identical (F_ST = 0) among the mitogenomes assessed. However, some of the mitochondrial genes (e.g., COI, ND4, and ND5) contained positions that were fixed for different nucleotides in the two lakes (FST = 1.0; Table S2, columns 2, 3, and 4). Interestingly, and in accordance with the SNP analyses (Table S2), a number of sites throughout the mitogenome are divergent between the Lake Gjerdhaugvatnet, which is an observation that corresponds well to the two-mitogenome haplotype situation in Lake Storvatnet. The marine SNP analysis (Table S2, column 5) concluded that several mitogenome haplotypes appear present among the 40 specimens in the pool sample. This result is reflected in the genetic differentiation pairwise analysis comparing the marine population to that of Lake Storvatnet (Figure 5B) and Lake Gjerdhaugvatnet (Figure 5C). Here, most SNPs noted in Lake Storvatnet were also found in the marine sample. Lake Gjerdhaugvatnet harbors some private alleles not seen in Lake Storvatnet or the marine sample, an observation explained by the fact that only one major mitochondrial haplotype (GHV-Maj1) is present.

4. Discussion

We report mitogenome analyses of sticklebacks sampled from two freshwater lakes in a small watercourse and a nearby marine site in subarctic Norway. While the marine sample contained a mixture of several mitogenome haplotypes, the downstream lake (Lake Gjerdhaugvatnet) harbored only one major single lineage among the 40 specimens assessed in the dataset. The upper lake (Lake Storvatnet), however, contained two distinct mitogenome clades. Interestingly, sticklebacks in Lake Storvatnet have been reported to be unusual due to different morph features regarding the pelvic spines [6,8].

The complete primary mitogenome sequence of two individuals, STV-Ida31 and STV-Ida61, was determined, and their mitogenomes were compared in more detail. The haplotype sequences were highly divergent, differing in rRNA and tRNA genes, all protein-coding genes, and non-coding regions. Complex I genes (encoding NADH dehydrogenase subunits) were found to harbor most non-synonymous substitutions, an observation corroborating recent findings [20]. A phylogenetic analysis based on mitogenome sequences revealed interesting and surprising relationships. (1) STV-Ida31 and STV-Ida61 were both members of the ENA mitogenome clades but located in different clusters. (2) STV-Ida31 was closely related to the mitochondrial haplotype (GHV-Maj1) found in downstream Lake Gjerdhaugvatnet. This implies that STV-Ida31 and GHV-Maj1 represent the original maternal lineage of sticklebacks in the watercourse. (3) STV-Ida61, on the other hand, was more closely related to a marine haplotype from Altafjord, Northern Norway, a sampling site that corresponds to an aerial line distance of approximately 400 km compared to Lake Storvatnet. (4) Each of the two ENA mitogenome clusters was interspersed with US haplotypes from Lake Superior (Wisconsin), a situation resembling that of Lake Storvatnet. However, it is important to note the enormous difference in size between Lake Storvatnet (0.2 km²) and Lake Superior (82,103 km²).

What could be the origin of STV-Ida61 in Lake Storvatnet? Based on the observed close relationship between STV-Ida61 and the marine haplotype in Altafjord, it is tempting to speculate that STV-Ida61 is a result of a more recent colonization by marine sticklebacks. Repeated colonization of freshwater habitats is relatively common in sticklebacks [15,17,27,28], and the PoolSeq data from the assessed marine site was found to carry almost all private STV-Ida61 SNPs compared to that of STV-Ida31. To further challenge the idea of a marine invader, there will be a need to sequence characterize individual mitogenome haplotypes from nearby marine sites to assess nucleotide-level relationships to the STV-Ida61 haplotype.

Is there a link between the mitochondrial haplotypes and pelvic spine morphology in Lake Storvatnet? The STV-Ida31 and STV-Ida61 mitogenomes were isolated and analyzed from normal-spined and spineless stickleback specimens, respectively. However, mitogenome PoolSeq analyses representing all morph types (see Figure 4 and Table S4) unambiguously conclude that there is no obvious connection between the mitogenome haplotypes and pelvic morphs in the present-day sticklebacks in Lake Storvatnet. However, these results may not exclude a possible linkage at an earlier stage in the past. There are fundamental differences in inheritance patterns between the uniparental mitogenome and the biparental nuclear genome in which the phenotypic traits are encoded. Thus, we speculate that one plausible explanation could be that Lake Storvatnet was invaded and subsequently experienced a second colonization by pelvic spineless sticklebacks carrying the STV-Ida61 mitogenome haplotype. While the mitogenome haplotype became preserved in the population due to uniparental inheritance and a lack of intermolecular recombination, nuclear genome traits responsible for the spineless morph were soon randomized due to Mendelian inheritance and meiotic recombination. This hypothesis could be further challenged by whole genome sequence analyses of specimens representing different morphs and sample site locations.

5. Conclusions

In the present study, we investigated the mitogenomic sequence features of threespine sticklebacks in a Norwegian watercourse known for its population of sticklebacks with unusual pelvic spine morphs. Two distinct mitogenome haplotypes were detected in the upper lake, and we argue that these haplotypes represent different mitogenome clusters of the Euro-North American sticklebacks. While one of the mitogenome haplotypes represents the original maternal line in the watercourse, the other could have been introduced through a more recent colonization event. We hypothesize that the marine invader was carrying genomic features responsible for the present-day mixture of pelvic spine morphs in Lake Storvatnet.