A key metabolic gene for recurrent freshwater colonization and radiation in fishes

See allHide authors and affiliations

Science  31 May 2019:
Vol. 364, Issue 6443, pp. 886-889
DOI: 10.1126/science.aau5656

Well prepared

It is well known that species radiate into new niches by adapting to novel environments. But why do some species radiate in this way, while other, related, species do not. Ishikawa et al. looked across sticklebacks to determine why some, originally marine, lineages were able to colonize postglacial freshwater environments (see the Perspective by Weber and Tong). They found that a gene involved in fatty acid desaturation was duplicated in freshwater lineages. Transgenic manipulation of this gene allowed marine lineages to synthesize fatty acids and thus survive on fatty acid–deficient freshwater diets.

Science, this issue p. 886; see also p. 831


Colonization of new ecological niches has triggered large adaptive radiations. Although some lineages have made use of such opportunities, not all do so. The factors causing this variation among lineages are largely unknown. Here, we show that deficiency in docosahexaenoic acid (DHA), an essential ω-3 fatty acid, can constrain freshwater colonization by marine fishes. Our genomic analyses revealed multiple independent duplications of the fatty acid desaturase gene Fads2 in stickleback lineages that subsequently colonized and radiated in freshwater habitats, but not in close relatives that failed to colonize. Transgenic manipulation of Fads2 in marine stickleback increased their ability to synthesize DHA and survive on DHA-deficient diets. Multiple freshwater ray-finned fishes also show a convergent increase in Fads2 copies, indicating its key role in freshwater colonization.

Empty niches can provide organisms with ecological opportunities to diversify (13). Many of the known large adaptive radiations followed invasion of underutilized habitats (1, 2). However, not all lineages appear to take advantage of such opportunities. For example, habitat shifts from marine to freshwater environments have repeatedly triggered radiations, but only in a limited number of fish lineages (46). The physiological and genetic factors causing this variation are unknown.

One of the underappreciated constraints for freshwater colonization by marine animals is the poor nutritional quality of food in freshwater ecosystems. Generally, the food chain in marine environments is rich in ω-3 long-chain polyunsaturated fatty acids, especially docosahexaenoic acid (DHA) (fig. S1A) (7), which is essential for animal development and health (8, 9). However, freshwater ecosystems contain very little DHA (fig. S1A) (7). Here, we tested and confirmed that DHA deficiency can constrain freshwater colonization by marine fishes and identified genetic changes that appear to have enabled some lineages to overcome this constraint.

Three-spined stickleback (Gasterosteus aculeatus species complex) are primarily marine or anadromous (hereafter, we call both marine) fishes, but when new freshwater habitats emerged after glacial retreat, they successfully colonized freshwater habitats and radiated into diverse ecotypes on multiple continents (10) (Fig. 1, fig. S2, and tables S1 and S2). By contrast, the closely related Japan Sea stickleback (G. nipponicus), which diverged from G. aculeatus ~0.68 to 1.5 million years (Ma) ago (11), failed to colonize freshwater environments and remains phenotypically homogeneous (Fig. 1). Although G. nipponicus co-occur with Pacific Ocean populations of G. aculeatus (Pacific Ocean stickleback) in some localities in Japan (1214), have geographical access to many freshwater habitats, and can use freshwater environments for spawning (14), our phylogenomic analysis showed that all known Japanese freshwater populations belong to G. aculeatus (Fig. 1C and fig. S3) (1517).

Fig. 1 Freshwater colonization and diversification in G. aculeatus but not in G. nipponicus.

(A) Sampling sites in Japan: pink squares, G. nipponicus; blue triangles, Pacific Ocean populations of G. aculeatus; green circles, freshwater populations of G. aculeatus. (B) Diversification of key foraging (gill raker number) and armor traits (lateral plate number) in freshwater populations. (C) Double-digest restriction site–associated DNA sequencing (ddRAD-seq) phylogeny of Japanese Gasterosteus indicating that all freshwater populations (green circles) belonged to G. aculeatus rather than G. nipponicus (pink squares). Blue triangles indicate Pacific Ocean marine populations of G. aculeatus. Bar shows the substitution rate. For bootstrap values, see fig. S3.

Because stickleback prey differs in DHA levels between marine and freshwater habitats (fig. S1B), we tested the hypothesis that Japan Sea stickleback may have a lower physiological ability than Pacific Ocean stickleback to survive on DHA-free diets (18). Our rearing experiments showed that, irrespective of salinity, Japan Sea stickleback had higher mortality than Pacific Ocean stickleback starting ~40 days after fertilization when fed DHA-free Artemia (P < 0.01) (Fig. 2A, fig. S4, and table S3); this age is close to the timing of seaward migration in nature (19). Marine-derived diets or Artemia enriched with several fatty acids, including DHA, significantly improved survival of Japan Sea stickleback (P < 0.01) (Fig. 2, A and B). Further, Japan Sea stickleback had a lower DHA content than Pacific Ocean stickleback when fed DHA-free Artemia (P < 0.01 for both the brain and eye) (Fig. 2C and fig. S5), suggesting that they have either lower DHA biosynthetic capabilities or higher rates of DHA degradation or secretion.

Fig. 2 Contribution of higher Fads2 copy numbers to survival with DHA-free diets.

(A) Survival curves of Japan Sea (red) and Pacific Ocean (blue) stickleback fed only DHA-free Artemia and Japan Sea (pink) and Pacific Ocean (light blue) stickleback fed marine-derived diets. (B) Survival curves of Japan Sea stickleback fed DHA-free Artemia (dark red) or Artemia enriched with several fatty acids, including DHA (pink). The two panels indicate independent replicate crosses. (C) DHA contents in the brain and eye of two species fed only DHA-free Artemia. **P < 0.01. (D) Relative copy numbers of Fads2 in males and females of Japan Sea and Pacific Ocean sticklebacks estimated from whole-genome resequencing (WGS) data. Different letters above the boxes indicate significantly different pairs (P < 0.05). (E) Expression levels of Fads2 at 40 to 60 days after fertilization. *P < 0.05. (F) Survival curves of Fads2-transgenic (yellow-green) and GFP-transgenic (gray) Japan Sea stickleback fed only DHA-free Artemia. The two panels indicate independent replicate crosses. (G) Whole-body DHA content of Fads2-transgenics (yellow-green) and GFP-transgenics (gray) at 40 days after fertilization, when Japan Sea stickleback start to die on DHA-free diets. *P < 0.05. RPKM, reads per kilobase per million reads.

Our whole-genome resequencing revealed that Fatty acid desaturase 2 (Fads2), a gene encoding a key enzyme catalyzing desaturation in DHA biosynthesis (fig. S6 and table S4) (2022), has a higher copy number in Pacific Ocean stickleback than in Japan Sea stickleback (Fig. 2D and fig. S7) (F1,12 = 79.8, P < 0.01). Higher Fads2 copy numbers in females compared with males (F1,12 = 11.7, P < 0.01) (Fig. 2D and fig. S7) are due to the X linkage of Fads2 (see below). RNA sequencing further revealed that Pacific Ocean stickleback express Fads2 at higher levels than Japan Sea stickleback (F1,12 = 5.3, P < 0.05 for brain; F1,12 = 7.0, P < 0.05 for eyes) when fed only DHA-free Artemia (Fig. 2E and fig. S8).

To directly demonstrate the effects of Fads2 copy number on survival, we made transgenic Japan Sea stickleback overexpressing Fads2. When fed only DHA-free Artemia, the Fads2-trangenics showed higher survival rates (Fig. 2F) and higher DHA content at 40 days after fertilization than the control GFP-transgenics (P < 0.01) (Fig. 2G and fig. S9). Analysis of an F2 intercross between Pacific Ocean and Japan Sea sticklebacks further showed that hybrids with higher Fads2 copy number had higher survival rates at 40 to 60 days after fertilization (P < 0.01: 10.0 to 12.1% of variance explained) and longer overall life span (P < 0.05) (fig. S10). The higher survival rate of females compared with males in Japan Sea stickleback is consistent with the higher Fads2 copy number in females (P < 0.01) (fig. S11). These data suggest that the lower Fads2 copy number may be a constraint to colonization of DHA-deficient freshwater niches by Japan Sea stickleback.

Fluorescence in situ hybridization (FISH) revealed that Fads2 was located only on the X chromosome [linkage group (LG) 19] in Japan Sea stickleback, but on LG12 and LG19 in Pacific Ocean stickleback (Fig. 3A). This result was confirmed by linkage analysis of Fads2 copy number using an F2 intercross (fig. S12). Genes flanking Fads2 on LG19, but not on LG12, showed conserved synteny with other teleosts (fig. S13). Furthermore, an outgroup, G. wheatlandi, another marine stickleback with no known freshwater populations (10), has Fads2 on LG19 but not on LG12 (table S5) and copy numbers similar to those of Japan Sea stickleback (fig. S14). Thus, LG19 is the ancestral location of Fads2, and copy-and-paste transposition of Fads2 from LG19 to LG12 increased the ability to synthesize DHA in G. aculeatus, but not in G. nipponicus or G. wheatlandi. At the LG12 locus, where a 12-kb insertion containing Fads2 and several types of transposons exist in G. aculeatus (Fig. 3B and fig. S15), G. wheatlandi and G. nipponicus possess transposons without Fads2 (figs. S16 to S18). This suggests that transposons might have mediated the Fads2 transposition and/or that this locus is a hot spot of insertion–deletion mutations (23).

Fig. 3 Extra copy of Fads2 on LG12 in G. aculeatus.

(A) FISH results with 4ʹ,6-diamidino-2-phenylindole (DAPI) nuclear staining. The left panel indicates a Japan Sea stickleback male with one Fads2 copy (green) on the X chromosome (circled). In a Pacific Ocean stickleback male, Fads2 (red) was located on the X chromosome (circle in the middle panel) and both copies of autosomal LG12 (circles in the right panel). The colors for Fads2 and LG19 or LG12 are flipped between the Japan Sea and Pacific Ocean males. Note that LG19 is either the X or Y chromosome, and the LG19 probe detects the region retained on the Y chromosome. (B) Genome structure around Fads2 on LG19 and LG12 of G. aculeatus and the corresponding region on LG12 of G. nipponicus and G. wheatlandi. Arrows and arrowheads indicate genes and repetitive sequences, respectively. White numbers indicate insertion size.

The estimated timing of Fads2 duplication within G. aculeatus is 0.80 Ma ago (95% highest posterior density: 0.47 to 1.16 Ma ago) (fig. S19), which is much earlier than the end of the last glacial period (0.012 Ma ago), when most stickleback freshwater colonization occurred (1, 10). Marine sticklebacks from western North America and Europe also repeatedly colonized freshwater habitats and radiated into diverse ecotypes (10). Our results show that they also have the extra copy of Fads2 on LG12, with copy numbers similar to those of the Pacific Ocean stickleback in Japan (Fig. 4A and figs. S16 and S22C to S22F). These data confirm that transposition onto LG12 occurred before the split between the Pacific and the Atlantic Ocean lineages (0.3 to 0.5 Ma ago) (24). Thus, the preexisting duplication of Fads2 has likely given G. aculeatus an advantage over other Gasterosteus species in colonizing freshwater habitats. However, our estimate suggests that Fads2 on LG12 is younger than the oldest known freshwater Gasterosteus fossil (10). Ancient extinct freshwater species may therefore have possessed additional Fads2 copies somewhere in the genome or adapted to DHA-deprived diets through other mutations.

Fig. 4 Parallel increase in Fads2 copies in freshwater fishes.

(A) Relative copy numbers (CNs) of Fads2 in males and females of Gasterosteus populations: red, Japan Sea; blue, Japanese Pacific Ocean, North American and European marine ecotypes; green, freshwater ecotypes. *P < 0.05, **P < 0.01. g-qPCR, genomic quantitative real-time polymerase chain reaction. (B) Relative CN of Fads2 in Pungitius. A single dot indicates the median of a single population in (A) and (B). (C) Comparison of Fads2 CNs among ray-finned fishes living in marine (–) or freshwater (+) niches.

To investigate whether there are any other loci involved in survivorship on DHA-deficient diets, we conducted quantitative trait locus (QTL) mapping of survival rates using an F2 intercross between the Pacific Ocean and Japan Sea sticklebacks. In addition to a suggestive QTL overlapping the Fads2 gene on LG12 (3.3% of variance in survival explained), one significant and two additional suggestive QTLs were found on different autosomes (fig. S20). The QTL on LG12, but not other QTLs, explained the Fads2 copy-number variation. Two other QTLs, including a significant one, showed overdominance rather than additive effects on survival, which may reflect an epistatic interaction between interspecies alleles (25) (fig. S21). Although survival rate is a polygenic and complex trait, our unbiased QTL analysis confirmed that the additional Fads2 copy of the Pacific Ocean stickleback on LG12 contributes to survivorship on DHA-deficient diets.

Because the Japanese Pacific Ocean stickleback also have increased survivorship with marine-derived diets compared with freshwater diets (Fig. 2A), additional Fads2 duplications beyond the LG12 copy may further increase DHA biosynthetic ability and be beneficial for permanent freshwater residency. Indeed, in Japan, freshwater stickleback populations had even higher Fads2 copy numbers than Pacific Ocean populations (χ22 = 17.1, P < 0.01) (Fig. 4A and figs. S22 and S23). Even within freshwater populations, those that had a longer evolutionary history in freshwater habitats had higher Fads2 copy numbers (χ23 = 35.3, P < 0.01) (Fig. 1C and fig. S22A to S22B). Additional copy number increases also occurred in North American (χ21 = 4.4, P = 0.035) and European freshwater populations (χ22 = 7.2, P = 0.028) (Fig. 4A and fig. S22C to S22F). We confirmed that freshwater populations with additional copies of Fads2 had more DHA than the Pacific Ocean population or a freshwater population with fewer copies when fed only DHA-free diets (F2,8 = 12.6, P < 0.01) (fig. S24). Both linkage analysis and long-read genome sequencing showed that tandem duplications on the X chromosome are responsible for additional copy number increase in both Japanese and Canadian freshwater populations (figs. S25 and S26). Transposons near Fads2 might have facilitated these tandem duplications (fig. S26) (26).

To test the generality of the mechanism, we first investigated nine-spined sticklebacks (genus Pungitius). The freshwater species, P. tymensis and P. kaibarae, had higher Fads2 copy numbers than P. pungitius (P < 0.05), which inhabits only brackish environments in Japan. P. sinensis, which inhabits both freshwater and brackish environments (27, 28), had copy numbers intermediate between those of P. pungitus and freshwater nine-spined sticklebacks (Fig. 4B and fig. S27). Finally, we investigated Fads2 copy numbers in the ray-finned fishes whose whole-genome sequences have been determined (fig. S28). Fish species that form freshwater populations had significantly higher Fads2 copy numbers than entirely marine species (Fig. 4C and fig. S29; MCMCglmm accounting for phylogeny, pMCMC < 0.01), suggesting convergent increases of Fads2 copies in diverse taxa that successfully colonized freshwater habitats.

Gene duplications not only enhance overall gene expression levels, but also allow duplicated copies to acquire new functions (29). Our yeast functional assay suggested that Fads2 genes in the Pacific Ocean stickleback acquired an additional enzymatic function in the DHA synthetic pathway (fig. S6 and table S4). Some of the Pacific Ocean–specific amino acid changes were shared by other freshwater ray-finned fishes (table S6), suggesting that they may be responsible for the acquisition of new enzymatic function. In addition to amino acid changes, both cis- and trans-regulatory changes cause expression differences between Fads2 haplotypes (fig. S30). Given that overexpression of Fads2 rescued the lethality in Japan Sea sticklebacks (Fig. 2, F and G), differences in the copy number itself likely contribute to differences in survival on DHA-deficient diets, although the possible involvement of changes in Fads2 protein sequence and regulation cannot be excluded.

Our data demonstrate that Fads2 is a key metabolic gene important for overcoming the nutritional constraints associated with freshwater colonization in fishes. Intriguingly, Fads shows strong signatures of selection in human populations that colonized polar regions, suggesting the importance of Fads in even more diverse taxa, including humans (30).

Supplementary Materials

Materials and Methods

Figs. S1 to S30

Tables S1 to S7

References (32128)

References and Notes

  1. Further details are available in the supplementary materials.
Acknowledgments: We thank T. Howes, M. Ravinet, K. Yoshida, K. Kawakami, Kitano laboratory members, many field assistants, the Aquatoto Gifu staff, S. Kondo, and L. Kawaguchi for technical assistance; B. Borg, H. Kanazawa, and M. Ozaki for providing fish; and M. Bell for discussion. Funding: This project was supported by JSPS KAKENHI grants 15H02418 and 23113007 (to J.K.), grant 16H06279 (to A.T.), and grants 26870824 and 16K07469 (to A.I.); the Asahi Glass Foundation (to J.K.); the Sumitomo Foundation (to A.I.); and JSPS postdoctoral fellowships 11J04816 to A.I. and 16J06812 to N.K. The work by N.L. and C.J.S. was funded by internal Eawag funds. B.M. was supported by SNF grant 31003A 175614. O.S. was supported by SNSF grant 31003A_163338. D.M. was supported by SNSF grant PDAMP3_123135 (to O.S). Author contributions: A.I. and J.K. conceived and designed the study; all authors collected and analyzed data; A.I. and J.K. wrote the first draft with input from the other authors. Competing interests: The authors declare no competing interests. Data and materials availability: All generated sequence data are available from the DNA Data Bank of Japan (DRA006706, DRA007515 to DRA007519, DRA007568, DRA008129, LC431555, and LC413556) and fatty acid meta-analysis data are available from Dryad ( (31).

Stay Connected to Science

Navigate This Article