Nutritional adaptation in the marine to freshwater establishment process – how do diet and genotype shape phenotype?

TL;DR

This study explores how diet and genetic factors influence the adaptation of marine fish to freshwater environments.

Contribution

The study reveals how diet and genetic capacity for DHA synthesis affect phenotypic responses in sticklebacks during freshwater adaptation.

Findings

Diet and population history strongly influence stickleback phenotype and performance in freshwater.

Sticklebacks with marine-derived fatty acids showed better growth and condition.

Freshwater populations accumulated more DHA than marine populations despite fads2 copy number variation.

Abstract

Nutrients, including vital organic compounds, vary in availability across ecosystems, with the potential to act as a strong source of selection for traits that increase nutrient acquisition and biosynthesis. Compared with freshwater ecosystems, marine ecosystems are much richer in the omega-3 long-chain polyunsaturated fatty acid docosahexaenoic acid (DHA) and thus marine animals establishing new freshwater populations are faced with the challenge of acquiring DHA. However, the relative roles of DHA synthesis capacity and diet in the freshwater establishment process remain unresolved. We used common garden experiments to explore phenotypic responses to dietary nutrient content in threespine sticklebacks (Gasterosteus aculeatus) that varied in their genetic capacity for DHA synthesis. We found that diet as well as presumed metabolic adaptation to freshwater nutritional environments (population identity) had strong effects on stickleback phenotype and performance. Sticklebacks enriched with marine-derived fatty acids including DHA grew more and were in better condition compared with controls. Those fed diets with more DHA also accumulated more DHA in muscle tissue. Freshwater sticklebacks accumulated more DHA compared with those from a marine population. However, populations with greater fads2 copy number did not consistently have higher performance or DHA accumulation. Together, these results suggest that DHA availability during development can strongly influence phenotype and performance, with the potential to act as a source of selection. Our findings also suggest that mechanisms beyond the accumulation of fads2 copies, such as plasticity in expression or variation in other DHA synthesis pathway genes, could be important adaptations to the nutritional constraints of freshwater.

Summary: Together, the LC-PUFA content of diet and freshwater establishment history of threespine sticklebacks (Gasterosteus aculeatus) influence fish performance, mortality and LC-PUFA content under common garden conditions.

Figures

Stickleback source populations and experimental design including copy numbers of populations and fatty acid content of diets. (A) Mean, standard deviation and individual replicates (smaller points) of relative fads2 copy number from field-sampled sticklebacks from populations used in the experiments (Chriesbach: n=24 female, n=19 male; Réveillon: n=2 female, n=2 male; Binnenkanal: n=21 female, n=17 male; Clitunno: n=18 female, n=12 male; Chessel: n=21 female, n=25 male). (B) Non-metric multidimensional scaling (NMDS) visualization of chironomid fatty acid content sampled monthly throughout the experiment (n=7 per treatment). (C) Stickleback source populations (CH, Switzerland; FR, France; IT, Italy) and number of replicate tanks (n=5 fish per tank) for each of the two chironomid dietary treatments used in common garden experiments. (D) Location of sampled European freshwater populations. Populations are ordered based on mean female parent copy number. Chriesbach is estimated to be the youngest population from the Baltic Sea lineage, which is estimated to have established freshwater populations within the last 10–5 kya. Binnenkanal, Chessel and Réveillon are assumed to be of mixed European freshwater lineages established during the Pleistocene (up to 40 kya) in southern Europe as well as after the Pleistocene (i.e. ∼11.7–5 kya) in Central and Northern Europe. Clitunno is estimated to be from an older (≥25 kya) pre-glacial southern Mediterranean lineage. The fads2 copy number data for marine sticklebacks from Puget Sound (USA) are taken from Ishikawa et al. (2019).

Effect of diet and population on mortality. Mean and standard deviation of mortality (i.e. proportion of fish dying per tank) for each population by dietary treatment combination; smaller points represent mortality for individual tanks (n=5 tanks per treatment for Binnenkanal, Chessel, Chriesbach and marine populations and n=3 tanks per treatment for Clitunno). The experiment was run a single time in the laboratory. Data were analyzed with bias-reduced logistic regression followed by post hoc Tukey-adjusted pairwise population comparisons (see Materials and Methods). Post hoc comparisons indicated that marine populations had higher mortality at the P=0.018, 0.018, 0.018 and 0.06 level versus Binnenkanal, Chessel, Chriesbach and Clitunno, respectively.

Effect of diet and population on body mass, length and condition. Mean and standard deviation of stickleback (A) mass, (B) length and (C) condition (ratio of mass to length) at all three experimental measurement time points, with onset of Vibrio-induced mortality for each population indicated by the gray shaded areas. Mean mass and length were measured at the individual level and were replicated as follows: Binnenakanal: n=25/25, 20/24, 19/24; Chriesbach: n=24/24, 21/21, 16/15; Chessel: n=25/25, 25/25, 25/25; Clitunno: n=14/15, 14/15, 12/15; marine: n=15/17, 15/17, 11/12 for control/treated fish at 10, 16 and 22 weeks, respectively. Condition was calculated at the tank level and was replicated as follows: Binnenkanal, Chriesbach, Chessel: n=5/5 control/treated tanks throughout the experiment; Clitunno: n=3/3 control/treated tanks throughout the experiment; marine: n=5/5 control/treated tanks at 10 weeks, n=3/4 control/treated tanks at 16 weeks and n=2/4 control/treated tanks at 22 weeks. The experiment was run a single time in the laboratory. Data were analyzed with either a general linear mixed model (GLMMs; mass and length) or a general linear model (GLM; condition) followed by post hoc Tukey-adjusted pairwise comparisons (see Materials and Methods). Throughout the experiment, mass varied with treatment at the P<0.001 level and condition at P<0.02. Condition and length both varied with population at the P<0.05 level throughout the experiment while mass only varied with population (P<0.05) at 10 weeks. See Results and Tables 3–5 for all interactions and pairwise comparisons.

Effect of diet, population and sex on fatty acid content. Mean and standard deviation of (A) ALA, (B) EPA and (C) DHA content (μg g−1) from fish muscle tissue; smaller points represent individual fish. (D) The DHA synthesis pathway, with ALA, EPA and DHA highlighted. Sample sizes for A–C were as follows: Binnenkanal n=9/11, Chessel n=10/10, Chriesbach n=9/9, Clitunno n=7/7, marine n=7/5 and Réveillon n=3/3 control/treated fish. The experiment was run a single time in the laboratory. Data were analyzed with GLMM (mass and length) followed by post hoc Tukey-adjusted pairwise comparisons (see Materials and Methods). ALA varied based on population (P<0.05), sex (P<0.05) and their interaction (P=0.05). EPA varied based on population (P<0.05). DHA content varied with population (P=0.001) and the interaction of population and treatment (P<0.05). See Results and Table 6 for all interactions and pairwise comparisons.

Tables

Table 1.. fads2 copy number by sex and population of parents of experimental fish

Predictor	Estimate	CI	P
(Intercept)	2.65	2.38–2.92	<0.001
Population [Chessel]	0.42	0.03–0.80	0.034
Population [Chriesbach]	−1.11	−1.49–0.73	<0.001
Population [Clitunno]	0.09	-0.32–0.49	0.673
Sex [m]	−1.42	−1.83–1.01	<0.001
Population [Chessel]×Sex [m]	−0.23	−0.78–0.33	0.432
Population [Chriesbach]×Sex [m]	0.63	0.06–1.20	0.031
Population [Clitunno]×Sex [m]	−0.57	−1.19–0.06	0.076

Table 2.. Mortality of fish in the experiment

	Odds ratio	CI	s.e.	d.f.	t-ratio	P
Predictor
(Intercept)	0.02	0.00–0.32				<0.001
Treatment [treated]	1.00	0.02–52.36				1.000
Population [Chessel]	1.00	0.02–52.36				1.000
Population [Chriesbach]	1.00	0.02–52.36				1.000
Population [Clitunno]	1.65	0.02–87.21				0.807
Population [Marine]	75.29	4.12–1377.06				<0.001
Treatment [treated]×Population [Chessel]	1.00	0.00–269.80				1.000
Treatment [treated]×Population [Chriesbach]	1.00	0.00–269.80				1.000
Treatment [treated]×Population [Clitunno]	1.00	0.00–274.52				1.000
Treatment [treated]×Population [Marine]	0.18	0.00–11.45				0.438
Contrast
Binnenkanal/Chessel	1.000		1.428	36	0.000	1.000
Binnenkanal/Chriesbach	1.000		1.428	36	0.000	1.000
Binnenkanal/Clitunno	0.608		0.871	36	−0.348	0.997
Binnenkanal/Marine	0.031		0.033	36	−3.282	0.018
Chessel/Chriesbach	1.000		1.428	36	0.000	1.000
Chessel/Clitunno	0.608		0.871	36	−0.348	0.997
Chessel/Marine	0.031		0.033	36	−3.282	0.018
Chriesbach/Clitunno	0.608		0.871	36	−0.348	0.997
Chriesbach/Marine	0.031		0.033	36	−3.282	0.018
Clitunno/Marine	0.051		0.054	36	−2.795	0.060

Table 3.. General linear mixed models (GLMMs) of fish mass at each time point

	Estimate	Odds ratio	CI	s.e.	d.f.	t-ratio	P
Predictor
Time point 1
(Intercept)	0.37		0.32–0.43				<0.001
Treatment [treated]	0.15		0.08–0.23				<0.001
Population [Chessel]	0.09		0.01–0.16				0.021
Population [Chriesbach]	0.16		0.09–0.24				<0.001
Population [Clitunno]	0.20		0.12–0.29				<0.001
Population [Marine]	0.05		−0.03–0.13				0.204
Treatment [treated]×Population [Chessel]	−0.06		−0.16–0.05				0.281
Treatment [treated]×Population [Chriesbach]	−0.03		−0.14–0.07				0.536
Treatment [treated]×Population [Clitunno]	−0.10		−0.22–0.02				0.104
Treatment [treated]×Population [Marine]	0.05		−0.06–0.16				0.371
Time point 2
(Intercept)	0.51		0.43–0.58				<0.001
Treatment [treated]	0.16		0.06–0.25				0.001
Population [Chessel]	0.10		0.01–0.20				0.037
Population [Chriesbach]	0.11		0.01–0.21				0.028
Population [Clitunno]	0.20		0.09–0.31				0.001
Population [Marine]	0.13		0.00–0.25				0.046
Treatment [treated]×Population [Chessel]	−0.06		−0.19–0.07				0.380
Treatment [treated]×Population [Chriesbach]	−0.01		−0.14–0.13				0.938
Treatment [treated]×Population [Clitunno]	−0.13		−0.29–0.02				0.085
Treatment [treated]×Population [Marine]	0.08		−0.08–0.24				0.326
Time point 3
(Intercept)	0.61		0.50–0.73				<0.001
Treatment [treated]	0.26		0.10-0.41				0.001
Population [Chessel]	0.14		−0.01–0.30				0.070
Population [Chriesbach]	0.05		−0.11–0.22				0.530
Population [Clitunno]	0.18		0.00–0.36				0.051
Population [Marine]	0.04		−0.25–0.34				0.761
Treatment [treated]×Population [Chessel]	−0.07		−0.28–0.14				0.529
Treatment [treated]×Population [Chriesbach]	0.10		−0.13–0.33				0.406
Treatment [treated]×Population [Clitunno]	−0.02		−0.27–0.23				0.875
Treatment [treated]×Population [Marine]	0.05		−0.29–0.38				0.787
Contrast
Time point 1
Binnenkanal/Chessel		−0.059		0.027	27.181	−2.225	0.201
Binnenkanal/Chriesbach		−0.147		0.026	30.399	−5.558	0.000
Binnenkanal/Clitunno		−0.153		0.030	31.227	−5.043	0.000
Binnenkanal/Marine		−0.076		0.028	31.297	−2.676	0.081
Chessel/Chriesbach		−0.088		0.026	30.665	−3.324	0.018
Chessel/Clitunno		−0.094		0.030	31.436	−3.103	0.031
Chessel/Marine		−0.017		0.028	31.538	−0.592	0.975
Chriesbach/Clitunno		−0.007		0.030	34.410	−0.217	0.999
Chriesbach/Marine		0.071		0.028	35.025	2.517	0.110
Clitunno/Marine		0.077		0.032	34.786	2.425	0.133
Time point 2
Binnenkanal/Chessel		−0.072		0.035	22.091	−2.091	0.259
Binnenkanal/Chriesbach		−0.109		0.035	34.743	−3.118	0.028
Binnenkanal/Clitunno		−0.130		0.039	31.285	−3.356	0.017
Binnenkanal/Marine		−0.166		0.041	31.155	−4.087	0.002
Chessel/Chriesbach		−0.037		0.035	23.027	−1.050	0.830
Chessel/Clitunno		−0.058		0.039	22.612	−1.496	0.575
Chessel/Marine		−0.093		0.040	23.120	−2.308	0.178
Chriesbach/Clitunno		−0.021		0.039	32.305	−0.548	0.981
Chriesbach/Marine		−0.057		0.041	32.086	−1.394	0.636
Clitunno/Marine		−0.035		0.044	29.991	−0.803	0.928

Table 4.. General linear models (GLMs) of fish condition at each time point

	Estimate	Odds ratio	CI	s.e.	d.f.	t-ratio	P
Predictor
Time point 1
(Intercept)	0.12		0.11–0.14				<0.001
Treatment [treated]	0.04		0.02–0.06				<0.001
Population [Chessel]	0.01		−0.01–0.03				0.234
Population [Chriesbach]	0.03		0.01–0.05				0.005
Population [Clitunno]	0.04		0.02–0.06				0.001
Population [Marine]	0.01		−0.01–0.04				0.221
Treatment [treated]×Population [Chessel]	−0.01		−0.04–0.02				0.568
Treatment [treated]×Population [Chriesbach]	−0.02		−0.05–0.01				0.236
Treatment [treated]×Population [Clitunno]	−0.03		−0.06–0.00				0.078
Treatment [treated]×Population [Marine]	0.00		−0.03–0.03				0.814
Time point 2
(Intercept)	0.15		0.13–0.16				<0.001
Treatment [treated]	0.03		0.01–0.05				0.002
Population [Chessel]	0.02		0.00–0.04				0.098
Population [Chriesbach]	0.02		0.00–0.04				0.061
Population [Clitunno]	0.06		0.03–0.08				<0.001
Population [Marine]	0.02		0.00–0.05				0.064
Treatment [treated]×Population [Chessel]	−0.01		−0.04–0.02				0.381
Treatment [treated]×Population [Chriesbach]	−0.01		−0.04–0.02				0.619
Treatment [treated]×Population [Clitunno]	−0.04		−0.07–0.00				0.025
Treatment [treated]×Population [Marine]	0.00		−0.04–0.03				0.821
Time point 3
(Intercept)	0.16		0.14–0.18				<0.001
Treatment [treated]	0.04		0.02–0.07				0.002
Population [Chessel]	0.02		−0.01–0.05				0.115
Population [Chriesbach]	0.01		−0.02–0.04				0.552
Population [Clitunno]	0.05		0.02–0.08				0.002
Population [Marine]	0.02		−0.02–0.05				0.350
Treatment [treated]×Population [Chessel]	−0.01		−0.05–0.03				0.513
Treatment [treated]×Population [Chriesbach]	0.02		−0.02–0.05				0.433
Treatment [treated]×Population [Clitunno]	0.00		−0.05–0.04				0.887
Treatment [treated]×Population [Marine]	−0.01		−0.06–0.04				0.662
Contrast
Time point 1
Binnenkanal/Chessel		−0.008		0.007	33	−1.162	0.772
Binnenkanal/Chriesbach		−0.021		0.007	33	−2.928	0.045
Binnenkanal/Clitunno		−0.026		0.008	33	−3.125	0.028
Binnenkanal/Marine		−0.012		0.008	33	−1.539	0.546
Chessel/Chriesbach		−0.013		0.007	33	−1.820	0.380
Chessel/Clitunno		−0.018		0.008	33	−2.160	0.220
Chessel/Marine		−0.003		0.008	33	−0.452	0.991
Chriesbach/Clitunno		−0.005		0.008	33	−0.584	0.997
Chriesbach/Marine		0.010		0.008	33	1.264	0.714
Clitunno/Marine		0.014		0.009	33	1.669	0.466
Time point 2
Binnenkanal/Chessel (control)		−0.017		0.010	33	−1.656	0.474
Binnenkanal/Chriesbach (control)		−0.019		0.010	33	−1.871	0.353
Binnenkanal/Clitunno (control)		−0.058		0.012	33	−4.877	0.000
Binnenkanal/Marine (control)		−0.022		0.012	33	−1.850	0.363
Chessel/Chriesbach (control)		−0.002		0.010	33	−0.214	1.000
Chessel/Clitunno (control)		−0.041		0.012	33	−3.442	0.013
Chessel/Marine (control)		−0.005		0.012	33	−0.416	0.993
Chriesbach/Clitunno (control)		−0.039		0.012	33	−3.257	0.021
Chriesbach/Marine (control)		−0.003		0.012	33	−0.230	0.999
Clitunno/Marine (control)		0.036		0.013	33	2.707	0.074
Binnenkanal/Chessel (treated)		−0.004		0.010	33	−0.416	0.993
Binnenkanal/Chriesbach (treated)		−0.012		0.010	33	−1.168	0.769
Binnenkanal/Clitunno (treated)		−0.020		0.012	33	−1.707	0.443
Binnenkanal/Marine (treated)		−0.018		0.011	33	−1.681	0.459
Chessel/Chriesbach (treated)		−0.008		0.010	33	−0.752	0.942
Chessel/Clitunno (treated)		−0.016		0.012	33	−1.347	0.664
Chessel/Marine (treated)		−0.014		0.011	33	−1.288	0.700
Chriesbach/Clitunno (treated)		−0.008		0.012	33	−0.696	0.956
Chriesbach/Marine (treated)		−0.006		0.011	33	−0.579	0.977
Clitunno/Marine (treated)		0.002		0.012	33	0.157	1.000
Time point 3
Binnenkanal/Chessel		−0.015		0.010	32	−1.574	0.524
Binnenkanal/Chriesbach		−0.016		0.010	32	−1.625	0.493
Binnenkanal/Clitunno		−0.048		0.011	32	−4.252	0.002
Binnenkanal/Marine		−0.012		0.012	32	−1.022	0.843
Chessel/Chriesbach		0.000		0.010	32	−0.051	1.000
Chessel/Clitunno		−0.032		0.011	32	−2.889	0.050
Chessel/Marine		0.003		0.012	32	0.290	0.998
Chriesbach/Clitunno		−0.032		0.011	32	−2.845	0.055
Chriesbach/Marine		0.004		0.012	32	0.333	0.997
Clitunno/Marine		0.036		0.013	32	2.768	0.066

Table 5.. GLMMs of fish length at each time point

	Estimate	Odds ratio	CI	s.e.	d.f.	t-ratio	P
Predictor
Time point 1
(Intercept)	3.25		3.11–3.39				<0.001
Treatment [treated]	−0.13		−0.33–0.08				0.230
Population [Chessel]	0.17		−0.03–0.36				0.089
Population [Chriesbach]	0.30		0.10–0.49				0.003
Population [Clitunno]	0.28		0.05–0.51				0.018
Population [Marine]	0.21		0.00–0.41				0.050
Treatment [treated]×Population [Chessel]	0.04		−0.24–0.33				0.763
Treatment [treated]×Population [Chriesbach]	0.32		0.04–0.61				0.026
Treatment [treated]×Population [Clitunno]	0.21		−0.12–0.53				0.218
Treatment [treated]×Population [Marine]	0.39		0.09–0.69				0.012
Time point 2
(Intercept)	3.37		3.24–3.50				<0.001
Treatment [treated]	0.35		0.18–0.52				<0.001
Population [Chessel]	0.32		0.15–0.49				<0.001
Population [Chriesbach]	0.32		0.14–0.49				0.001
Population [Clitunno]	0.10		−0.10–0.30				0.319
Population [Marine]	0.55		0.33–0.77				<0.001
Treatment [treated]×Population [Chessel]	−0.14		−0.37–0.10				0.245
Treatment [treated]×Population [Chriesbach]	0.00		−0.24–0.25				0.993
Treatment [treated]×Population [Clitunno]	−0.16		−0.44–0.11				0.248
Treatment [treated]×Population [Marine]	0.12		−0.16–0.40				0.405
Time point 3
(Intercept)	3.70		3.52–3.88				<0.001
Treatment [treated]	0.47		0.23–0.71				<0.001
Population [Chessel]	0.33		0.09–0.57				0.008
Population [Chriesbach]	0.11		−0.15–0.36				0.414
Population [Clitunno]	0.03		−0.27–0.32				0.867
Population [Marine]	0.46		0.00–0.91				0.049
Treatment [treated]×Population [Chessel]	−0.14		−0.48–0.19				0.398
Treatment [treated]×Population [Chriesbach]	0.01		−0.35–0.37				0.959
Treatment [treated]×Population [Clitunno]	−0.14		−0.53–0.26				0.504
Treatment [treated]×Population [Marine]	−0.13		−0.66–0.39				0.617
Contrast
Time point 1
Binnenkanal/Chessel (control)		−0.168		0.098	31.712	−1.709	0.443
Binnenkanal/Chriesbach (control)		−0.297		0.099	32.414	−3.004	0.038
Binnenkanal/Clitunno (control)		−0.279		0.117	34.840	−2.389	0.142
Binnenkanal/Marine (control)		−0.205		0.104	31.712	−1.969	0.304
Chessel/Chriesbach (control)		−0.129		0.099	32.414	−1.305	0.690
Chessel/Clitunno (control)		−0.111		0.117	34.840	−0.953	0.874
Chessel/Marine (control)		−0.037		0.104	31.712	−0.358	0.996
Chriesbach/Clitunno (control)		0.018		0.117	35.376	0.150	1.000
Chriesbach/Marine (control)		0.092		0.105	32.336	0.875	0.904
Clitunno/Marine (control)		0.074		0.122	34.570	0.608	0.973
Binnenkanal/Chessel (treated)		−0.211		0.105	32.336	−2.016	0.281
Binnenkanal/Chriesbach (treated)		−0.620		0.105	32.336	−5.913	0.000
Binnenkanal/Clitunno (treated)		−0.485		0.119	31.712	−4.090	0.002
Binnenkanal/Marine (treated)		−0.594		0.112	33.494	−5.309	0.000
Chessel/Chriesbach (treated)		−0.408		0.099	33.121	−4.105	0.002
Chessel/Clitunno (treated)		−0.274		0.114	32.238	−2.404	0.140
Chessel/Marine (treated)		−0.382		0.107	34.324	−3.587	0.009
Chriesbach/Clitunno (treated)		0.134		0.114	32.238	1.178	0.764
Chriesbach/Marine (treated)		0.026		0.107	34.324	0.243	0.999
Clitunno/Marine (treated)		−0.108		0.121	33.239	−0.899	0.895
Time point 2
Binnenkanal/Chessel		−0.250		0.060	29.417	−4.200	0.002
Binnenkanal/Chriesbach		−0.317		0.062	33.250	−5.105	0.000
Binnenkanal/Clitunno		−0.019		0.070	30.807	−0.275	0.999
Binnenkanal/Marine		−0.612		0.072	29.632	−8.506	0.000
Chessel/Chriesbach		−0.067		0.060	30.820	−1.108	0.801
Chessel/Clitunno		0.231		0.068	28.965	3.405	0.015
Chessel/Marine		−0.362		0.070	27.953	−5.150	0.000
Chriesbach/Clitunno		0.298		0.070	31.880	4.250	0.002
Chriesbach/Marine		−0.296		0.072	30.597	−4.079	0.003
Clitunno/Marine		0.593		0.079	29.258	−7.511	0.000
Time point 3
Binnenkanal/Chessel		−0.256		0.085	25.644	−3.014	0.042
Binnenkanal/Chriesbach		−0.111		0.091	32.472	−1.217	0.742
Binnenkanal/Clitunno		0.043		0.101	29.458	0.412	0.993
Binnenkanal/Marine		−0.389		0.134	37.286	−2.912	0.045
Chessel/Chriesbach		0.144		0.089	30.821	1.620	0.497
Chessel/Clitunno		0.298		0.099	28.175	3.010	0.040
Chessel/Marine		−0.133		0.132	36.481	−1.009	0.850
Chriesbach/Clitunno		0.154		0.105	33.507	1.468	0.590
Chriesbach/Marine		−0.277		0.136	40.069	−2.036	0.268
Clitunno/Marine		−0.431		0.143	37.259	−3.016	0.035

Table 6.. GLMM of DHA content of fish

Predictor	Estimate	CI	P
(Intercept)	3.201	2.87–3.18	<0.001
Treatment [treated]	0.05	−0.18–0.28	0.648
Population [Chessel]	−0.16	−0.37–0.04	0.122
Population [Chriesbach]	0.06	−0.16–0.27	0.599
Population [Clitunno]	0.34	0.14–0.54	0.001
Population [Marine]	−0.66	−0.91–−0.42	<0.001
Population [Reveillon]	0.10	−0.15–0.36	0.415
Sex [m]	0.09	−0.13–0.32	0.411
Treatment [treated]×Population [Chessel]	0.13	−0.18–0.44	0.415
Treatment [treated]×Population [Chriesbach]	−0.13	−0.50–0.23	0.466
Treatment [treated]×Population [Clitunno]	0.13	−0.18–0.44	0.407
Treatment [treated]×Population [Marine]	0.46	0.07–0.85	0.020
Treatment [treated]×Population [Reveillon]	−0.01	−0.40–0.38	0.953
Treatment [treated]×Sex [m]	−0.05	−0.36–0.27	0.756
Population [Chessel]×Sex [m]	−0.03	−0.35–0.29	0.848
Population [Chriesbach]×Sex [m]	−0.26	−0.58–0.06	0.108
Population [Clitunno]×Sex [m]	−0.07	−0.43–0.29	0.700
Population [Marine]×Sex [m]	0.32	−0.02–0.67	0.063
Population [Reveillon]×Sex [m]	0.24	−0.23–0.71	0.310
Treatment [treated]×Population [Chessel]×Sex [m]	0.00	−0.43–0.44	0.983
Treatment [treated]×Population [Chriesbach]×Sex [m]	0.10	−0.37–0.58	0.674
Treatment [treated]×Population [Marine]×Sex [m]	−0.29	−0.80–0.22	0.254

Full text

INTRODUCTION

Fish have established new freshwater populations and lineages from marine ancestral populations numerous times (e.g. Seehausen and Wagner, 2014). Such establishments have required them to adapt to novel abiotic conditions, notably lower salinity (Lee and Bell, 1999; Velotta et al., 2022), as well as novel biotic conditions, including resources that vary in nutritional composition compared with those in marine ecosystems. In particular, the physiologically important omega-3 long chain polyunsaturated fatty acids (n-3 LC-PUFA) eicosapentaenoic acid (EPA; 20:5n-3) and docosahexaenoic acid (DHA; 22:6n-3) are both abundant throughout marine food webs and relatively scarcer, especially in the case of DHA, in freshwater food webs (Twining et al., 2021). EPA and DHA are critically important for a host of physiological functions in animals, including neural, immune and cardiac function, and reproduction (Bell and Tocher, 2009; Brenna et al., 2009; Brenna and Carlson, 2014; Das, 2006) and serve as limiting nutrients for many fish (Sargent et al., 1995; Tocher, 2003). As limiting nutrients, n-3 LC-PUFA availability may serve as a source of selection for fish and other marine taxa seeking to exploit the novel nutritional conditions of freshwater environments, and thus promote the evolution of nutritional adaptations. Consequently, as marine fish establish new freshwater populations, they can provide a unique window into the role of nutrients in the process of adaptation to freshwater environments.

One way that consumers can escape the n-3 LC-PUFA limitation is through increased synthesis from dietary precursors such as alpha linolenic acid (ALA; 18:3n-3; Twining et al., 2021). Some animals can source all of the DHA that they require by adding double bonds and carbon atoms to these precursors via the Sprecher or Δ4 pathway and then using a combination of fatty acid desaturase (FADS) and elongase (ELOVL) enzymes (Oboh et al., 2017), encoded by fads and elovl genes, respectively (Castro et al., 2012, 2016; Monroig and Kabeya, 2018). Freshwater fishes appear to have adapted to freshwater nutritional constraints through changes in the function, efficiency and expression of these genes (Zhang et al., 2023). For instance, freshwater flatfish species have increased DHA synthesis capacity through changes in the efficiency and function of the fads2 and elovl5 genes (Matsushita et al., 2020), while freshwater threespine sticklebacks have increased their copy number of fads2 (Ishikawa et al., 2019).

In this study, we focus on threespine stickleback (Gasterosteus aculeatus), which have repeatedly established freshwater populations from ancestral marine populations, across a range of different freshwater nutritional environments (i.e. in lakes as well as streams) across the Northern Hemisphere following deglaciation (Fang et al., 2018, 2020; McKinnon and Rundle, 2002; Hudson et al., 2024; Ishikawa et al., 2021). Ishikawa et al. (2019) found that multiple freshwater lineages of threespine stickleback had more copies of the fads2 gene than contemporary marine stickleback populations, presumably allowing freshwater lineages to more readily synthesize DHA. In addition, stream fish in the Pacific Northwest with more DHA-poor benthic diets had a higher fads2 copy number than lake fish that consumed more DHA-rich pelagic prey (Ishikawa et al., 2021). Across lakes in Greenland, stickleback in lakes with a higher proportion of DHA-rich copepods had fewer fads2 copies compared with those from nearby lakes with fewer copepods, suggesting that local dietary DHA availability can select for nutritional adaptation (Hudson et al., 2024).

Previous data suggest that nutritional habitat-based differences can lead to increased fads2 copy number and may consequently improve stickleback performance in DHA-poor environments as part of their adaptation to freshwater. However, this has not been tested experimentally along a gradient of marine to freshwater populations varying in their fads2 copy number and history of freshwater establishment. Importantly, while previous experimental studies have compared the phenotype and performance either marine (i.e. Ishikawa et al., 2019) or freshwater (i.e. Hudson et al., 2022) population pairs, we lack comparative data on marine and freshwater stickleback performance and phenotypes under common garden conditions. Moreover, previous studies comparing the fatty acid content of two marine or freshwater populations did not manipulate diet, leaving the importance of fads2 copy number versus diet quality in shaping freshwater stickleback phenotype unclear.

Here, we explored adaptation to the nutritional constraints of freshwater environments in threespine sticklebacks, through common garden experiments with both marine and freshwater fish. We included freshwater populations that had contrasting fads2 copy numbers from lineages of variable estimated ages since freshwater establishment. We also manipulated dietary n-3 LC-PUFA availability to understand how n-3 LC-PUFA synthesis and n-3 LC-PUFA retention from diet affect fatty acid phenotype and performance, including survival. This design allowed us to examine both the general freshwater nutritional adaptation process (i.e. marine–freshwater differences) and the specific importance of fads2 (i.e. variation across freshwater populations) in this process.

MATERIALS AND METHODS

Study design

To understand the effects of both population and diet on fatty acid phenotype as well as performance, we conducted a common garden experiment, where we manipulated nutritional quality directly for multiple populations of sticklebacks that varied in their fads2 genotype (Table 1, Fig. 1; Tabel S1). Five European freshwater populations from across Baltic, northern/central European and southern European Mediterranean lineages, plus one North American marine population (Puget Sound, USA) were raised under common conditions on chironomid-based diets that varied in DHA content, mimicking either marine or freshwater prey. We examined the effects of diet and population identity (representing genetically divergent lineages) on survival, growth, body condition and fatty acid phenotype of these populations.

Stickleback source populations and experimental design including copy numbers of populations and fatty acid content of diets. (A) Mean, standard deviation and individual replicates (smaller points) of relative fads2 copy number from field-sampled sticklebacks from populations used in the experiments (Chriesbach: n=24 female, n=19 male; Réveillon: n=2 female, n=2 male; Binnenkanal: n=21 female, n=17 male; Clitunno: n=18 female, n=12 male; Chessel: n=21 female, n=25 male). (B) Non-metric multidimensional scaling (NMDS) visualization of chironomid fatty acid content sampled monthly throughout the experiment (n=7 per treatment). (C) Stickleback source populations (CH, Switzerland; FR, France; IT, Italy) and number of replicate tanks (n=5 fish per tank) for each of the two chironomid dietary treatments used in common garden experiments. (D) Location of sampled European freshwater populations. Populations are ordered based on mean female parent copy number. Chriesbach is estimated to be the youngest population from the Baltic Sea lineage, which is estimated to have established freshwater populations within the last 10–5 kya. Binnenkanal, Chessel and Réveillon are assumed to be of mixed European freshwater lineages established during the Pleistocene (up to 40 kya) in southern Europe as well as after the Pleistocene (i.e. ∼11.7–5 kya) in Central and Northern Europe. Clitunno is estimated to be from an older (≥25 kya) pre-glacial southern Mediterranean lineage. The fads2 copy number data for marine sticklebacks from Puget Sound (USA) are taken from Ishikawa et al. (2019).

We chose populations from European lineages that varied in their time of freshwater establishment (Fang et al., 2018; Hudson et al., 2021) (Table 1; Table S1). Stickleback populations from the Baltic Sea lineage, which have been introduced around the Bodensee (i.e. Lake Constance/Chriesbach, Switzerland), were estimated to have established freshwater populations within the last 10–5 kya during the post-Pleistocene (Fang et al., 2020; Hudson et al., 2021; Liu et al., 2016; Mäkinen and Merilä, 2008; Marques et al., 2019). Populations from the Swiss midlands (i.e. Binnenkanal), western Switzerland (i.e. Chessel) and the Seine drainage around Paris, northern France (i.e. Réveillon) were assumed to be of mixed European lineage. These admixed populations originate from lineages that established during the Pleistocene in southern Europe in the lower Rhône and other Mediterranean drainages up to 40 kya as well as those that established after the Pleistocene (i.e. ∼11.7–5 kya) in northern and central Europe in the Seine, Rhein and other Atlantic or North Sea drainages (Fang et al., 2020; Hudson et al., 2023; Lucek et al., 2010). We also included a central Italian population (i.e. Clitunno River) estimated to be from an older (≥25 kya) pre-glacial southern Mediterranean lineage (Coll-Costa et al., 2024; Fang et al., 2018). All freshwater populations came from streams because macroinvertebrate prey from streams typically have lower DHA availability compared with those in lakes, making them the most challenging freshwater nutritional environment for establishment by animals of marine origin (Twining et al., 2021).

Fish collection and breeding in Switzerland was permitted at the Swiss federal level under permits 33583 and LU02/2021. In Italy, sampling was authorized under permit no. 003169_2022 from the Umbria region. Fish from France were moved across the French-Swiss border under Nagoya protocol agreement TREL2103218S/461.

Breeding and husbandry

In the spring of 2022, we first visually assessed the egg ripeness of females and only sampled those that looked most ready to breed. We also sampled adult males that were in breeding color. Animals were killed by immersion in tricaine methanesulfonate (i.e. MS-222) and rinsed with clean stream water. We gently released clutches of eggs from females into Petri dishes. We then dissected testes from males, mashed these between two glass slides, and washed this mixture over the eggs with stream water. After letting the egg clutches and testes sit for ∼5–10 min to allow fertilization to occur, we washed fertilized clutches into clean plastic test tubes, leaving head space for gas exchange, and brought tubes back to the lab in coolers of cool stream water.

In the Réveillon River at Villecresnes (France), we caught two males, which we mated with one ripe female, resulting in one split clutch (i.e. half of the clutch had one male parent and the other half had the other male parent). In the Clitunno River (Italy), we produced two successful clutches from two males and two females. In Grand Canal Chessel, Binnenkanal and Chriesbach (Switzerland), we produced eight, five and five successful clutches, respectively, from male and female pairs. After we had bred individuals in the field, we took fin clips from each and stored these in absolute ethanol for genetic analysis (see below). In addition, we acquired nine clutches of Puget Sound (USA) marine sticklebacks that were already in culture at the University of Bern.

All clutches were initially placed in plastic cups with window screen mesh bottoms that were suspended in aquaria to allow water circulation and gas exchange. Air bubblers were placed on the bottom of the aquaria directly under the cups in order to ensure oxygenation and frequent water movement. We visually inspected all clutches daily and manually removed eggs with fungi with the aid of a dissecting microscope. Eggs took between 8 and 14 days to hatch, with entire clutches hatching within a few hours of each other. When eggs had hatched, we removed the air bubblers from the bottom of the tanks and reduced air flow until fish began swimming freely and feeding. We initially fed juveniles with Salt Lake Artemia spp. nauplii that were hatched in salt water, rinsed and filtered into fresh water. When fish were approximately 28–35 days old, we began introducing them to finely chopped frozen chironomid larvae, gradually acclimating them to a full, non-chopped chironomid diet. We used aquarium heaters to keep all aquaria at approximately 18°C throughout the experiment and we added NaCl (100 mg l^−1^) to all aquaria to help prevent fungal infections while maintaining salinity below that of typical marine conditions. To minimize tank-specific effects, we moved fish to new tanks halfway through experiments. Tank water was changed weekly or more frequently if fish showed signs of disease or stress.

In late summer to autumn of 2022, when fish were approximately 75–90 days old (variation due to differences in clutch hatching dates within the same population), we assigned them to experimental treatments. Clutch number and egg survival to initial larval stage varied across populations, resulting in different numbers of clutches per population, so we mixed clutches within populations and did not include clutch as a random effect in our models (see below). Fish from multiple mixed clutches within populations were placed in each 100 l aquarium and then randomly assigned to diet treatments with either ad libitum thawed chironomids or ad libitum thawed chironomids soaked in a marine oil enriching mixture (Selco Easy DHA artemia enricher, INVE Aquaculture; Fig. 1B,C; Fig. S1). Because of variation in breeding success and clutch survival among populations, we created five replicated tanks of each diet treatment for the marine population as well as the three Swiss populations (i.e. Chessel, Binnenkanal and Chriesbach), three replicate tanks per diet treatment for the Italian population, and one tank per treatment for the French population. We took samples of both diet treatments monthly for analysis of fatty acid content.

We examined the effects of diet on growth and body condition by weighing and photographing each fish three times over the course of the experiment at approximately 158±5 days (10 weeks in experiment), 201±6.7 days (16 weeks in experiment) and 242±8 days (22 weeks in experiment) old (means±s.d.). We used ImageJ (https://imagej.net/ij/docs/guide/) to calculate fish standard length from photos taken of fish next to a ruler and calculated average tank body condition as tank average fish mass divided by tank average fish length. We concluded experiments when fish were approximately 246–270 days old and close to reproductive age. At this point, we photographed each individual in a cuvette with color blocks and length bars, then weighed it, and killed it via MS-222 overdose. We then dissected each fish to confirm its sex based on its gonads and took skinned tail muscle samples for fatty acid analysis.

Marine populations experienced high mortality within the first 2 months of our experiments. One individual from a marine oil-enriched diet treatment died within the first 8 days of the experiment. Two additional marine tanks on the control chironomid diet experienced 80% mortality (i.e. four fish from each tank of five died) within the first 2 months of the experiment. We replaced all of these individuals with fresh individuals from stock tanks of the same age and mix of parents (from breeding described above) that had been fed control chironomids. All individuals in one of these tanks subsequently died before the next measurement. We included fish from the one surviving replacement tank in fatty acid content analysis, but not in subsequent analysis of mortality or performance.

A widespread Vibrio infection occurred throughout our aquaria facility midway through our rearing study, causing 56%, 30%, 22%, 16%, 7% and 4% mortality in marine, French, Binnenkanal, Chriesbach, Clitunno and Chessel populations, respectively. Fish in stock tanks began showing symptoms in late November and Vibrio infection was confirmed by mid-December, by which time fish in experiments (∼200 days old) began getting sick as well. Diagnosis of sick fish and assessment of potential treatments was conducted by the Institute for Fish and Wildlife Health at the University of Bern. In mid-December, we began treating fish showing signs of disease (fast breathing, lying at the bottom of the tank, not eating) with Enrox^®^ 10% (100 mg ml^−1^ enrofloxacin). Tanks with fish showing signs of disease were given an initial 1 ml dose of Enrox, followed by a 50% water change and then subsequent dosing 3 and 5 days later.

Genetic analysis

We used genomic quantitative PCR (g-qPCR) to investigate fads2 copy number variation across our five freshwater populations (Fig. 1A; Table S1). In addition to the parents from all freshwater sites, we also included an additional juvenile female caught from the Reveillon River (France); an additional 16 females and 10 males from Clittuno River (Italy); an additional 16 females and 12 males from Binnenkanal (Switzerland); an additional 13 females and 17 males from Chessel (Switzerland); and an additional 17 females and 14 males from Chriesbach (Switzerland).

First, we extracted DNA from fin clips preserved in absolute ethanol using Qiagen^®^ DNeasy Blood & Tissue Kits and used the resulting extract for g-qPCR. After extraction, we quantified DNA concentration with a Nanodrop 1 (Thermo). As in Hudson et al. (2024) we modified our g-qPCR method from Ishikawa et al. (2019), using the same forward and reverse primers, and a TaqMan probe on a region conserved among fads2 haplotypes. These primers amplify both fads2 on stickleback chromosome 12 as well as the transposed copy (and any duplications of this) on chromosome 19 (i.e. the X chromosome). Following Ishikawa et al. (2019), we used thyroid stimulating hormone β2 (tshβ2), to calculate the relative copy number of fads2 (henceforth copy number) for each individual. tshβ2 is a single copy gene without known duplications in threespine stickleback, meaning that the two copies present in a diploid stickleback are normalized to a copy number of one. g-qPCR reactions were performed on a LightCycler^®^ 480 System (Roche) with Real-Time PCR Master Mix (Roche); we ran each reaction in duplicate (i.e. two reactions with fads2 primers and two with tshβ2 primers). We then calculated relative fads2 copy number using the delta-delta Ct method (Livak and Schmittgen, 2001) to compare the mean Ct values for fads2 and tshβ2 across replicates. We excluded samples from further analysis and/or reanalyzed them when fads2 or tshβ2 copy number differences between replicates were greater than 0.5. As we normalized copy number relative to the single copy gene tshβ2, a relative copy number of one for fads2 reflects a total genomic copy number of two.

Fatty acid content analysis

We quantified fatty acid content of 90 sticklebacks and their chironomid diets from our experiments (Table S1). We extracted lipids three times by immersing tissues in a 1:2 methanol:dichloromethanol (DCM) mixture in ashed glass vials, placing these vials in an ultrasonic ice water bath for 10 min, transferring the extract into new ashed and pre-weighed test tubes with glass Pasteur pipettes, and drying down the extract with nitrogen gas (N_2_). We then added 50 µl of C23:0 at a 483.6 µg ml^−1^ concentration to each extracted lipid sample via Hamilton syringe as a recovery standard and then added 1 ml of boron trifluoride (BF_3_) to each tube with a glass pipette, followed by vortexing and 5 min of water bath sonication. We then placed tubes in an oven at 100°C for 2 h. After removing methylated samples from the oven and allowing them to cool, we added 2 ml MilliQ purified water to each, followed by approximately 1 ml hexane. We then vortexed samples, and used glass pipettes to transfer the hexane fraction to an HPLC vial three times. After transferring hexane to vials, we again dried them down with N_2_, and resuspended fatty acid methyl esters (FAME) in 500 µl ethyl acetate for gas chromatography flame ionization detection (GC-FID) analysis. We analyzed samples on a Shimadzu GC-2010 Plus with a 30 m InertCap 5MS/NP GC column using a 47 min run starting at 70°C and ramping up to 150°C and then 320°C running in splitless injection mode with a column flow of 2.07 ml min^−1^. We identified compounds based on retention times relative to a Supelco FAME mix (G4-C24, 18919-1AMP) using the software GC Solution (Shimadzu). We then calculated the content of each compound in µg g^−1^ by dividing peak area by the peak area of our C23:0 recovery standard, multiplying this by C23:0 total mass added to the sample, and dividing by the mass of dried tissue.

Statistical analysis

How does the fatty acid content of chironomid diets vary across treatments?

We visually compared control versus marine oil-treated chironomids (Fig. S1) fed to sticklebacks using non-metric multidimensional scaling using the metaMDS function (vegan, version 2.6-4) in R. All data were log transformed. We included the following fatty acids: 13:0, 14:0, 15:0, 15:1, 16:0, 16:1, 17:0, 18:0, 18:1n-9t, 18:2n-6c (linoleic acid, LIN), 18:2n6t or 18:1n-9c, 18:3n-3 (ALA) or 18:3n-6, 20:0, 20:1, 20:2, 20:4n-6 (arachidonic acid, ARA), 20:5n-3 (EPA), 21:0, 22:0, 22:2, 22:5n-3 (docosapentaenoic acid, DPA), 22:5n-6, 22:6n-3 (DHA), 23:3n-3 or 20:3n-6, and 24:0. No marine oil-treated chironomids contained the fatty acid 22:1n-9 so we did not include it in our NMDS visualization even though it was detected in control chironomids.

How does fads2 copy number vary across freshwater populations?

To understand whether there was significant variation in fads2 relative copy number between our four of five freshwater European stickleback populations (the Reveillon population was excluded because of its small sample size), we used a general linear model (GLM, base R) with a Gaussian error distribution and an identity link function to test for differences among populations, between sexes and their interaction. We performed post hoc pairwise comparisons between individual populations as well as between sex by population pairs using the R package emmeans (version 1.10.5; https://CRAN.R-project.org/package=emmeans) with Tukey correction for multiple comparisons.

Were there initial differences in length, mass and/or condition between populations?

We also used GLM with a Gaussian error distribution and an identity link function to examine differences in length, mass and condition at the start of experiments from the Clittuno, Chessel, Chriesbach, Binnenkanal and marine populations. Ten fish were randomly measured from each population of mixed clutches.

How did population and/or treatment influence mortality, size and body condition?

We used logistic regression to examine variation in mortality by population, treatment and their interaction in five of our six experimental populations (excluding the Reveillon population, which lacked replicate tanks). Because some populations had little to no mortality (i.e. Clitunno and Chessel), we used a Firth bias-reduced logistic regression model (logistf, version 1.26.1) with a binomial error distribution and logit link function. We modeled mortality as mortality of all fish that originally started in treatments, not including replacement control fish from stock tanks. In cases where population and/or an interaction of population and treatment were significant, we performed post hoc Tukey tests with emmeans.

We sought to understand how experimental dietary treatment and population influenced performance in five of our six experimental populations (i.e. excluding the Reveillon population, which lacked replicate tanks). We tested for differences in mass and length of sticklebacks using general linear mixed models (GLMM, lme4 version 1.1-34) with population, treatment and their interaction as fixed effects and tank as a random effect using models with Gamma error distribution and log link functions. Because body condition (i.e. mean mass of fish in tank/mean length of fish in tank) and changes in mean mass, length and body condition were measured at the tank level, we used GLM with a Gamma error distribution and log link function to examine differences in these variables based on population, treatment and their interaction. Sex was not included in any of these models because fish were only sexed at the end of the experiment. We created separate models of mass and length at each of the three time measurement points (i.e. measurement 1, approximately 10 weeks into experiments; measurement 2, approximately 16 weeks into experiments; and measurement 3, approximately 22 weeks into experiments). In cases where population and/or an interaction of population and treatment were significant, we performed post hoc Tukey tests with emmeans.

How did fatty acid content vary by population, treatment and/or sex?

We used GLMMs with a Gaussian family and identity link function to examine variation in the content of palmitic acid (16:0), oleic acid (18:1n-9), LIN, ALA, ARA, EPA, DPA and DHA in fish from all six experimental populations, with sex, population, treatment and the three-way interaction of population, treatment and sex as fixed effects and tank identity as a random effect. In cases where population and/or interactions of population and treatment were significant, we performed post hoc Tukey tests with emmeans. Because sex was unbalanced across populations and treatments, we subset emmeans models by sex.

All statistical analyses were conducted in R (http://www.R-project.org/) and graphics were created with the package ggplot2 (version 3.4.4; Wickham, 2011). Model fits were assessed with marginal R^2^ values and appropriateness of model fits was assessed with residual plots (ggResidpanel version 0.3.0).

RESULTS

fads2 copy number varies across European stickleback populations

As expected given the location of the fads2 gene on the X chromosome, XY males had a significantly lower fads2 copy number than XX females across all sites (Fig. 1A, Table 1). Fish collected from Chriesbach had a significantly lower fads2 copy number compared with fish collected from Binnenkanal (Binnenkanal–Chriesbach; estimate=0.798, s.e.=0.1459, d.f.=149, t-ratio=5.492, P<0.0001), Chessel (Chessel–Chriesbach; estimate=1.103, s.e.=0.139, d.f.=149, t-ratio=7.955, P<0.0001) and Clittuno (Chriesbach–Clittuno; estimate=−0.602, s.e.=0.157, d.f.=149, t-ratio=−3.830, P=0.001), while fish collected from Chessel had a significantly higher fads2 copy number than that of fish collected from Clittuno (Chessel–Clittuno; estimate=0.501, s.e.=0.153, d.f.=149, t-ratio=3.267, P=0.007) as well as Chriesbach. The four fish collected from the Reveillon River had values that were intermediate between those of Chriesbach and Binnenkanal (Fig. 1A).

Initial variation in condition among populations

Prior to the onset of the experimental dietary treatment, fish from Grand Canal Chessel were significantly lighter and shorter and of lower condition (i.e. ratio of mass to length) compared with those from other populations (Tables S2–S4).

Effect of diet and population on mortality

Survival in experiments varied strongly by population. Mortality occurred in two waves: there was an initial period of low mortality from a variety of causes during the first 2 months of the experiment and a later period of high mortality due to a widespread Vibrio infection across our entire facility. Throughout our experiments, significantly more marine sticklebacks died compared with other populations (Fig. 2, Table 2).

Effect of diet and population on mortality. Mean and standard deviation of mortality (i.e. proportion of fish dying per tank) for each population by dietary treatment combination; smaller points represent mortality for individual tanks (n=5 tanks per treatment for Binnenkanal, Chessel, Chriesbach and marine populations and n=3 tanks per treatment for Clitunno). The experiment was run a single time in the laboratory. Data were analyzed with bias-reduced logistic regression followed by post hoc Tukey-adjusted pairwise population comparisons (see Materials and Methods). Post hoc comparisons indicated that marine populations had higher mortality at the P=0.018, 0.018, 0.018 and 0.06 level versus Binnenkanal, Chessel, Chriesbach and Clitunno, respectively.

Effect of diet and population on length, mass and condition

Compared with sticklebacks on standard chironomid diets, fish fed a diet supplemented with marine fats grew significantly heavier and increased their condition significantly more (i.e. mass per length) throughout the experiment and were longer by the later parts of the experiment (Fig. 3, Tables 3–5). By 10 weeks into the experiment, prior to the Vibrio outbreak in our facility, fish treated with marine fats were longer, heavier and in better condition compared with those on the control chironomid diet (Fig. 3, Tables 3–5). Fish from Binnenkanal and Chessel were also significantly lighter and of lower condition than those from Chriesbach or Clitunno. Binnenkanal fish on control diets were also significantly shorter than control fish from Chriesbach, while treated Binnenkanal fish were shorter than treated fish from Chriesbach, Clitunno and marine populations, and treated fish from Chessel were shorter than treated fish from Chriesbach and marine populations.

Effect of diet and population on body mass, length and condition. Mean and standard deviation of stickleback (A) mass, (B) length and (C) condition (ratio of mass to length) at all three experimental measurement time points, with onset of Vibrio-induced mortality for each population indicated by the gray shaded areas. Mean mass and length were measured at the individual level and were replicated as follows: Binnenakanal: n=25/25, 20/24, 19/24; Chriesbach: n=24/24, 21/21, 16/15; Chessel: n=25/25, 25/25, 25/25; Clitunno: n=14/15, 14/15, 12/15; marine: n=15/17, 15/17, 11/12 for control/treated fish at 10, 16 and 22 weeks, respectively. Condition was calculated at the tank level and was replicated as follows: Binnenkanal, Chriesbach, Chessel: n=5/5 control/treated tanks throughout the experiment; Clitunno: n=3/3 control/treated tanks throughout the experiment; marine: n=5/5 control/treated tanks at 10 weeks, n=3/4 control/treated tanks at 16 weeks and n=2/4 control/treated tanks at 22 weeks. The experiment was run a single time in the laboratory. Data were analyzed with either a general linear mixed model (GLMMs; mass and length) or a general linear model (GLM; condition) followed by post hoc Tukey-adjusted pairwise comparisons (see Materials and Methods). Throughout the experiment, mass varied with treatment at the P<0.001 level and condition at P<0.02. Condition and length both varied with population at the P<0.05 level throughout the experiment while mass only varied with population (P<0.05) at 10 weeks. See Results and Tables 3–5 for all interactions and pairwise comparisons.

By 16 weeks into experiments, while fish from several populations had become infected with Vibrio (at different ages and time in experiment due to differences in hatch dates), overall patterns remained relatively unchanged (Fig. 3, Tables 3–5). Supplemented fish continued to be heavier, longer and in better condition compared with those on control diets. Fish from Binnenkanal were still significantly lighter than those from Chriesbach and Clitunno as well as those from the marine population. Clitunno fish were in significantly better condition than those from Binnenkanal, Chessel and Chriesbach and marginally better condition than marine fish by the second measurement. By this time point, marine fish were longer than those from all other populations. Fish from Binnenkanal were also significantly shorter than those from Chessel and Chriesbach, while fish from Clitunno were significantly shorter than those from Chessel and Chriesbach, and but these effects no longer interacted with treatment. Marine fish were still longer at this point compared with fish from Binnenkanal and Clitunno.

At 22 weeks, after all populations had been exposed to Vibrio, fish supplemented with marine oils remained heavier, longer and in better condition compared with those on control diets (Fig. 3, Tables 3–5). Although many marine fish died (see above), those that did survive until the third measurement were significantly longer compared with those from other populations. There were no differences in mass by population by the third measurement. Meanwhile, fish from Clitunno remained in significantly better condition compared with fish from Binnenkanal and in marginally better condition compared with fish from Chriesbach and marine populations.

Fatty acid phenotype varies with treatment, population and sex

Population, treatment, sex and their interaction all influenced the content of n-3 and other PUFA (Fig. 4) as well as the energy storage fatty acids palmitic and oleic acid. In the case of DHA, marine fish had significantly lower DHA content than other populations, but marine fish supplemented with marine oils rich in DHA (Fig. 1B) were able to achieve higher muscle DHA content than those on control diets (Fig. 4C, Table 6; Table S5). Fish from Clitunno (our oldest estimated population) had significantly higher DHA than other populations regardless of dietary treatment (Fig. 4C, Table 6; Table S5).

Effect of diet, population and sex on fatty acid content. Mean and standard deviation of (A) ALA, (B) EPA and (C) DHA content (μg g−1) from fish muscle tissue; smaller points represent individual fish. (D) The DHA synthesis pathway, with ALA, EPA and DHA highlighted. Sample sizes for A–C were as follows: Binnenkanal n=9/11, Chessel n=10/10, Chriesbach n=9/9, Clitunno n=7/7, marine n=7/5 and Réveillon n=3/3 control/treated fish. The experiment was run a single time in the laboratory. Data were analyzed with GLMM (mass and length) followed by post hoc Tukey-adjusted pairwise comparisons (see Materials and Methods). ALA varied based on population (P<0.05), sex (P<0.05) and their interaction (P=0.05). EPA varied based on population (P<0.05). DHA content varied with population (P=0.001) and the interaction of population and treatment (P<0.05). See Results and Table 6 for all interactions and pairwise comparisons.

Although chironomids treated with marine oil contained more of the omega-3 PUFA DHA precursors (Fig. 4D), namely ALA, EPA (Fig. 1B) and DPA, fish did not differ in their content of these fatty acids based on dietary treatment (Fig. 4A,B; Tables S6–S8). EPA content was significantly lower in Chessel and marine fish than in those from Binnenkanal, but did not vary based on sex or its interactions (Table S7). ALA and DPA content was also significantly lower in fish from Chessel and marine populations, but these effects varied based on interactions between population, sex and/or treatment (Tables S6 and S8). Specifically, females did not vary in their ALA content (Table S6), but males from Chessel and marine populations supplemented with DHA had lower ALA content than those from Binnenkanal (Table S6), while males from Chessel and Chriesbach had higher DPA than males from Binnenkanal (Table S8).

Fish did not differ in their content of the omega-6 PUFA LIN based on diet, sex, population or their interaction (Table S9). The n-6 LC-PUFA ARA was more complex: fish from Chessel had significantly lower ARA content than those from Binnenkanal or marine populations and male marine fish supplemented with DHA had higher ARA than those from Chessel (Table S10).

Content of the energy storage fat palmitic acid was significantly lower in DHA-treated marine male fish, but did not otherwise vary based on dietary treatment, population, sex or their interaction (Table S11). Oleic acid was lower in Chessel fish and well as in those from our marine population, but this effect in marine fish varied based on interactions: marine male fish had higher oleic acid content than females, but treated marine males had lower oleic acid content than others (Table S12).

DISCUSSION

We sought to reveal the roles of diet and metabolic capacity in determining nutritional phenotype and the consequences of this for growth and survival during the freshwater establishment process. To understand capacity for nutrient synthesis and accumulation, we conducted common garden experiments with threespine stickleback populations that varied in their time of freshwater establishment and their fads2 copy number, suggesting among-population differentiation in their putative genetic capacity for n-3 LC-PUFA synthesis. To understand the effects of dietary DHA intake, we also manipulated the DHA content of diets in our experimental treatments. We found that the DHA content of diet had a strong effect on DHA phenotype in marine fish (i.e. those with the lowest fads2 copy number). We also observed higher rates of growth and body condition across individuals with diets that were rich in DHA and other LC-PUFA. Both muscle fatty acid content and performance also varied among populations. Importantly, across dietary treatments, mortality was highest and DHA content was lowest in marine populations. Together, these results suggest that dietary n-3 LC-PUFA availability during an individual's development, together with freshwater establishment history, can strongly influence stickleback fatty acid phenotype and survival, with the potential to act as a source of selection as populations become established in freshwater.

Common garden conditions reveal effects of diet and metabolism on nutritional phenotype

Manipulating nutrient availability for stickleback populations with variable fads2 genotypes in common garden experiments helped elucidate the processes of n-3 LC-PUFA synthesis and n-3 LC-PUFA retention from diet across populations. Supplementation with marine-derived fatty acids, including DHA, resulted in marine fish with higher muscle DHA, demonstrating that marine sticklebacks will preferentially retain dietary DHA in spite of their more limited ability to synthesize it. Interestingly, while the marine oil-supplemented diets contained more of the n-6 LC-PUFA ARA and the n-3 LC-PUFA DPA, muscular ARA and DPA content did not vary with dietary treatment, suggesting that DHA availability was more limiting than ARA or DPA availability. This metabolic plasticity in response to variable nutritional environments might be an adaptive solution in environments with high nutritional heterogeneity (Raubenheimer et al., 2012), such as the streams used as sources for our freshwater populations.

Echoing the findings of previous studies, we found that lineage-based phenotypic differences in nutritional metabolism became apparent while rearing fish under common garden laboratory conditions. For instance, the oldest experimental lineage from the Clitunno River in Italy, which had some of the highest fads2 copy numbers, had a higher DHA content compared with other populations. Previously, Hudson et al. (2022) found that sticklebacks from an older Lac Léman population (Grand Canal Chessel) accumulated more DHA compared with a more recently established Bodensee population. Here, we found that fish from Chessel and Chriesbach (i.e. Bodensee) populations, despite differing in growth and condition (Fig. 3), had DHA phenotypes that were similar to each other and intermediate compared with those of other populations (Fig. 4C). In addition to these differences within freshwater lineages, we found that marine sticklebacks in experiments had significantly lower muscle DHA content compared with freshwater lineages, highlighting both their more limited ability to synthesize DHA (control treatments) and their more limited ability to accumulate DHA from diet (marine oil-supplemented diets). Together, this provides further evidence that lineages with a longer history of freshwater establishment are able to accumulate DHA more efficiently than those of marine or more recent freshwater establishment. However, in contrast to our expectations, freshwater populations with a higher fads2 copy number did not consistently have a correspondingly higher DHA content, suggesting that other genes (e.g. elovl) may underlie the nutritional adaptation process in freshwater.

Intriguingly, sticklebacks from Clitunno and Chessel populations, which had varying degrees of ancestry from pre-glacial Mediterranean lineages, both performed better and, in the case of Clitunno, accumulated more DHA, than other freshwater lineages despite differing from each other in fads2 copy number. Chessel fish also experienced the lowest mortality of all populations, echoing findings from previous mesocosm studies (Best et al., 2017; Hudson et al., 2023). Fish from Clitunno, the oldest freshwater population in our experiments, outperformed those from Chessel in terms of DHA accumulation even though fish from Chessel had a significantly higher fads2 copy number. Clitunno sticklebacks accumulated more DHA and were also in better condition than fish from other populations regardless of diet. These populations may have additional metabolic adaptations for freshwater nutritional environments beyond increased fads2 copy number. Adaptations beyond increases in fads2 copy number could include changes in elovl elongase genes involved in DHA synthesis as seen in flatfish (Matsushita et al., 2020), as well as an increased ability to take advantage of DHA when it is available via changes in nutrient retention.

As in previous studies (e.g. Ishikawa et al., 2019, 2021; Hudson et al., 2024), freshwater female sticklebacks in our experimental populations had a higher fads2 copy number than their male counterparts (Fig. 1A). As females invest large amounts of lipids, including LC-PUFA, in their eggs, it may be especially advantageous for females to have a higher fads2 copy number. We found that stickleback muscle tissue varied in ALA and DPA content based on sex and as well as interactions between sex, population and treatment, such that females had higher ALA and DPA content than males (Fig. 4A). ALA is a precursor to the n-3 LC-PUFA EPA, DPA and DHA, while DPA is an intermediate between EPA and DHA (Fig. 4D). This suggests that within populations, males and females, despite varying in their fads2 copy number, were able achieve similar EPA and DHA levels in muscle tissue, albeit with potential costs in terms of their reserves of other precursor and/or intermediate fatty acids. Future studies could examine whether females preferentially allocate their EPA and DHA to their gonads compared with males and thus have higher LC-PUFA content at the whole-body level.

Diet determines growth, but population origin determines survival

Across all populations, sticklebacks supplemented with marine-derived fatty acids, including DHA, grew larger and were in better condition than those on the standard DHA-poor chironomid diet (Fig. 3). While diet had strong effects on growth, condition and fatty acid phenotype of sticklebacks in our common garden experiments, population origin had the strongest effects on survival. Fish throughout our aquarium facility were unintentionally exposed to a Vibrio infection midway through experiments, allowing us to opportunistically examine effects of diet and population on performance under the challenge of infection. Regardless of dietary treatment, fish from marine populations had significantly higher mortality than those from other populations (Fig. 2, Table 2). In contrast, populations from lineages that established the earliest in freshwater (e.g. Clitunno and Chessel) had zero mortality on marine supplemented diets and the lowest mortality of all populations on standard treatments (treatment differences not significant). Fish from more recently established Binnenkanal and Chriesbach populations had intermediate levels of mortality across both dietary treatments. Population-based differences in mortality were not explained by fads2 copy number, suggesting that additional genetic mechanisms, including those unrelated to nutrition, could influence survival in the face of infection.

Conclusions

Overall, we found clear evidence of metabolic adaptation to freshwater nutritional environments. Multiple European freshwater stickleback lineages accumulated increased genetic capacity for nutrient synthesis (i.e. greater fads2 copy number). These populations, especially those with the longest history of freshwater establishment, were able to accumulate more DHA from both synthesis and preferential retention compared with marine sticklebacks. In contrast, DHA content in marine sticklebacks varied based on dietary DHA availability, suggesting little use of synthesis. Freshwater populations also suffered lower mortality compared with the marine population. Together, these results suggest that DHA availability in diet may act as a source of selection as fish and other taxa of marine origin establish new populations in freshwater nutritional environments. Threespine stickleback are a classic example of a species whose ancestral form displays developmental plasticity that has ultimately given rise to multiple ecotypes through parallel selection (e.g. Wund et al., 2008). To further evaluate the effect of nutrients on selection, future studies should extend common garden rearing conditions to evaluate the effects of developmental plasticity throughout the stickleback life cycle, including evaluating the effects of diet and fads2 during early post-larval development (i.e. during artemia feeding, as in Ishikawa et al., 2019), during the pre-reproductive stage (current study), as well on reproductive investment and success later in life.

In general, our findings on the freshwater nutritional adaptation process echo studies on a diversity of other metabolic adaptations that have allowed species and populations to take advantage of novel nutritional resources. For instance, in humans, the repeated evolution of lactose tolerance probably increased survival on milk products during periods of nutritional stress and disease (Evershed et al., 2022). Likewise, the ability to survive on freshwater resources and to accumulate DHA via both synthesis and selective retention appears to have repeatedly evolved across threespine sticklebacks, including across the multiple European lineages included in this study, as well as more broadly (Ishikawa et al., 2019). Future studies should also examine whether fads2 duplication and phenotypic differences in fatty acid accumulation and performance exist across other closely related taxa, such as fourspine sticklebacks (Apeltes quadracus), brook sticklebacks (Culaea inconstans) and ninespine sticklebacks (Pungitius pungitius; e.g. Liu et al., 2022), that have encountered similar nutritional challenges while establishing freshwater populations.

Supplementary Material

10.1242/jexbio.251462_sup1Supplementary information