The hybrid history of zebrafish

Abstract

Since the description of zebrafish (Danio rerio) in 1822, the identity of its closest living relative has been unclear. To address this problem, we sequenced the exomes of 10 species in genus Danio, using the closely related Devario aequipinnatus as outgroup, to infer relationships across the 25 chromosomes of the zebrafish genome. The majority of relationships within Danio were remarkably consistent across all chromosomes. Relationships of chromosome segments, however, depended systematically upon their genomic location within zebrafish chromosomes. Regions near chromosome centers identified Danio kyathit and/or Danio aesculapii as the closest relative of zebrafish, while segments near chromosome ends supported only D. aesculapii as the zebrafish sister species. Genome-wide comparisons of derived character states revealed that danio relationships are inconsistent with a simple bifurcating species history but support an ancient hybrid origin of the D. rerio lineage by homoploid hybrid speciation. We also found evidence of more recent gene flow limited to the high recombination ends of chromosomes and several megabases of chromosome 20 with a history distinct from the rest of the genome. Additional insights gained from incorporating genome structure into a phylogenomic study demonstrate the utility of such an approach for future studies in other taxa. The multiple genomic histories of species in the genus Danio have important implications for comparative studies in these morphologically varied and beautiful species and for our understanding of the hybrid evolutionary history of zebrafish.

Keywords: danios; genome structure; hybrid species; introgression; phylogenomics; zebrafish.

Graphical Abstract

Fig. 1.

Demographic events and distribution of SDCs in related lineages. Ancestral alleles are denoted “A” and derived alleles “B,” with alleles segregating in multiple individuals in populations over time. Derived alleles present in the ancestral population of the 3 species are shown as colored lines. Alleles arising in ancestral populations after the split from the outgroup and shared by only 2 species (dashed lines) are shown with shading. a) Mutations can occur and reflect the history of the population tree. b and c) Alternative ways of inheriting ancestrally segregating polymorphisms by ILS of alleles. d) Expectations with ILS as the only source of multiple genomic histories (compared to panels b) and c)). e) Expectations with introgression from 1 species. Introgression can also occur from 2 different species, not shown. f) Expectations with HHS. HHS is a special type of HAS.

Fig. 2.

Genealogical discordance and the effects of chromosome structure. a) The unpartitioned genome phylogeny inferred from the concatenated sequence for each species with concordance factors based on the phylogenies inferred for the 250 jackknife partitions. Dashed lines denote relationships not found in the unpartitioned genome topology, but supported by at least 10% of jackknife partitions. Splits supported by <10% of trees are not shown. A gray box demarcates the D. rerio species group. b) Concordance factors of pairwise splits grouping 2 species as sister taxa within the D. rerio species group for the 250 jackknife partitions, with major concordances shown on the left and minor concordances listed on the right. c) Five genealogical histories in the D. rerio species group explain relationships inferred for 93.2% of windows across the genome and a minority of other topologies explain the rest (6.8%, gray). The frequency of the most common jackknife topologies binned according to their position across chromosomes, indicated as a folded chromosome with the 2 telomeres at the right and the chromosome middle on the left. Each bin contains 50 partitions corresponding, for example, to alignments near the ends of each of the 25 Danio chromosomes. The inferred sister species of D. rerio depends on the position of the partition along the chromosome, with the D. rerio—D. kyathit relationship (re, ky; green, see panel c)) supported across the middle of most chromosomes, while the D. rerio—D. aesculapii relationship (ae, re; rust) supported near the telomeres of most chromosomes. Note also that few (ky, ni; purple) trees occurred at chromosome centers and many (ky, ni; purple) trees occupied partitions near the telomeres. d) The relationship of D. choprae relative to other members of the genus phylogeny varied according to chromosome position. The centers of chromosomes supported the placement of D. choprae inferred by the unpartitioned genome phylogeny (green), while chromosome ends showed more variation in the relationships supported.

Fig. 3.

D-statistics show genomic regions affected by introgression. a) D-statistics testing for gene flow between D. nigrofasciatus and D. kyathit (purple) or between D. nigrofasciatus and D. rerio (blue). The assumed topology (left), distribution of D-statistic values across the genome (second from left), and D values across chromosomes for windows of 200 ABBA/BABA sites (right, note the spike near telomeres). The horizontal gray bar shows the 95% confidence intervals under the null expectation of equal ABBA/BABA frequencies. Color scheme as in Fig. 2. b) D-statistics testing for gene flow between D. nigrofasciatus and D. kyathit (purple) or between D. nigrofasciatus and D. aesculapii (pink). c) D-statistics testing for gene flow between D. kyathit and D. rerio (green) or between D. kyathit and D. aesculapii (yellow), demonstrating an anomaly on Chr20, but no consistent effect of chromosome position relative to telomeres. d) D-statistics testing for gene flow between D. aesculapii and D. rerio (rust) or between D. aesculapii and D. kyathit (yellow), again showing an anomaly on Chr20.

Fig. 4.

Distinct genomic histories in the danio genome shown by patterns of SDCs. a) Proportions of the 6 pairwise splits (derived characters present in only 2 species) in the D. rerio species group in sliding windows of 500 pairwise splits across chromosome 2, which displays a representative pattern, and all chromosomes appear in Supplementary Fig. 2. The gray horizontal bar shows the 95% confidence interval under the null expectation of equal proportions of the 6 split patterns. The colored bar at the top shows which splits are enriched at each position along the chromosome. Color scheme as in Fig. 2. b) Distribution of pairwise split proportions from the entire genome (windows of 500 pairwise splits). c) Percentage of genomic windows showing enrichment for pairwise splits in at least 10 genomic windows. The bottom left graph shows the percent of the genome enriched for each combination of pairwise splits. d) Co-occurrence of pairwise split enrichment across all 25 zebrafish chromosomes. Chromosomes are shown unfolded and colored according to which splits are enriched at each position along the chromosome. Color scheme as in panel a). Genomic regions with no significant enrichment are dark gray. e) Effects of chromosome location on pairwise split proportions. Points are plotted according to relative chromosome location of a window along a folded chromosome (centers of chromosomes on the left and ends of chromosomes on the right) and the proportion of each pairwise split in that window. Splines are fitted to the data for each pairwise split across all 25 chromosomes in nonoverlapping windows of 500 pairwise splits. f) Proportion of D. rerio ancestry attributable to the D. aesculapii lineage (orange) or the D. kyathit lineage (green). Analysis was performed on the sequences used previously for jackknife trees binned according to chromosome position.

Fig. 5.

The hybrid history of the zebrafish genome. a) A model for the recent population history of the D. rerio species group showing the hybrid origin of D. rerio from the D. aesculapii and D. kyathit lineages and the introgression of sequences between D. kyathit and D. nigrofasciatus retained mostly at chromosome ends. b) Graphic depicting, across a chromosome, the expected distribution of alleles originating in populations ancestral to only 2 species at the time of HAS. c) Approximate experimentally discovered distribution of alleles from populations shared by only 2 species in modern-day danios.