Omega-3 long-chain polyunsaturated fatty acids support aerial insectivore performance more than food quantity

Significance

Insect abundance is an important predictor of survival and performance for many taxa of aerial insectivores, which forage on a mixture of aquatic and terrestrial insects that differ in fatty acid composition, particularly omega-3 long-chain polyunsaturated fatty acid (LCPUFA) content. We raised Tree Swallow (Tachycineta bicolor) chicks on either a high or low quantity of feed with either high amounts of the LCPUFA found in aquatic insects or an equivalent amount of the precursor omega-3 PUFA, alpha-linolenic acid, but low LCPUFA. LCPUFA content was more important for Tree Swallow chick performance than food quantity. Tree Swallows may be timing breeding to coincide with the peak abundance of high-LCPUFA aquatic insects.

Abstract

Once-abundant aerial insectivores, such as the Tree Swallow (Tachycineta bicolor), have declined steadily in the past several decades, making it imperative to understand all aspects of their ecology. Aerial insectivores forage on a mixture of aquatic and terrestrial insects that differ in fatty acid composition, specifically long-chain omega-3 polyunsaturated fatty acid (LCPUFA) content. Aquatic insects contain high levels of both LCPUFA and their precursor omega-3 PUFA, alpha-linolenic acid (ALA), whereas terrestrial insects contain much lower levels of both. We manipulated both the quantity and quality of food for Tree Swallow chicks in a full factorial design. Diets were either high-LCPUFA or low in LCPUFA but high in ALA, allowing us to separate the effects of direct LCPUFA in diet from the ability of Tree Swallows to convert their precursor, ALA, into LCPUFA. We found that fatty acid composition was more important for Tree Swallow chick performance than food quantity. On high-LCPUFA diets, chicks grew faster, were in better condition, and had greater immunocompetence and lower basal metabolic rates compared with chicks on both low LCPUFA diets. Increasing the quantity of high-LCPUFA diets resulted in improvements to all metrics of performance while increasing the quantity of low-LCPUFA diets only resulted in greater immunocompetence and lower metabolic rates. Chicks preferentially retained LCPUFA in brain and muscle when both food quantity and LCPUFA were limited. Our work suggests that fatty acid composition is an important dimension of aerial insectivore nutritional ecology and reinforces the importance of high-quality aquatic habitat for these declining birds.

Aerial insectivores, a paraphyletic group that includes the swallows, swifts, nightjars, and at least five different families of flycatchers, were once abundant throughout both temperate and tropical regions. However, in the last half century, a number of North American aerial insectivores across a diversity of families and species, ranging from Common Nighthawks (Chordeiles minor) and Chimney Swifts (Chaetura pelagica) to Olive-sided Flycatchers (Contopus cooperi) and Tree Swallows (Tachycineta bicolor), have undergone major declines (1, 2). For example, Tree Swallows, one of the best-studied model aerial insectivore taxa in North America, have declined by 36% over the past 2 to 3 decades (2). Experts have proposed several hypotheses, including (i) declines in aerial insects (3), (ii) habitat loss and degradation (3–5), (iii) environmental contaminants (6, 7), and (iv) climate change and phenological mismatch (8, 9). Evidence exists to support all of these hypotheses, yet, at present, the exact causes of aerial insectivore declines remain unresolved, pointing to the need for a more thorough understanding of all aspects of aerial insectivore ecology.

Past studies have documented the importance of food resources for aerial insectivores, and numerous studies have suggested that aerial insectivore declines are linked to decreasing overall insect abundance (e.g., ref. 1). Research on Tree Swallows shows that food availability is linked with chick growth rates and fledging success as well as egg size and composition (10, 11). Winkler et al. (11) found that environmental temperature had a strong effect on patterns of Tree Swallow chick mortality, most likely through its effect on insect activity levels. However, the sheer quantity of food resources may not be the only important factor. Food quality and the potential for mismatch between insect composition and the nutritional needs of aerial insectivores may also be important drivers of reproductive output and overall fitness for these birds (e.g., ref. 12).

Food quality can be defined in many ways, including caloric density, nutrient composition, and digestibility (13). Here, we focus on differences in composition of macronutrients. Aerial insectivores, like all animals, require organic compounds [e.g., vitamins, amino acids, and fatty acids (FAs)] in addition to elemental nutrients (e.g., nitrogen, phosphorus, and calcium) to grow, develop, and complete their life cycles. Omega-3 long-chain polyunsaturated FAs (LCPUFA), in particular the FAs docosahexaenoic acid (22:6n-3, DHA) and eicosapentaenoic acid (20:5n-3, EPA), are especially important organic compounds for most animals, affecting a range of important physiological processes, from immune function to vision and brain development (14). Birds and all other vertebrates must either consume EPA and DHA directly from diet or indirectly by consuming their molecular precursor, the short-chain omega-3 PUFA, alpha linolenic acid (18:3n-3, ALA), and then converting ALA into EPA and DHA. The capability of any particular animal species to convert ALA to the bioactive EPA and DHA depends on whether its diet contains EPA and DHA (14). Mammalian herbivores typically synthesize all EPA and DHA endogenously from ALA, whereas carnivores such as cats must obtain all of their DHA from diet (15). The ability of wild birds to synthesize DHA is not well characterized, but DHA concentrations are inversely related to mass (16). For example, DHA constitutes 12% of FAs in the muscles of House Sparrows (Passer domesticus) (16), which are similar in size to Tree Swallows, but can reach over 20% in Ruby-throated Hummingbird (Archilochus colubris) muscle (17).

In the wild, aerial insectivores consume a combination of terrestrial and aquatic insects (18), which differ in their FA composition (19). Aquatic insects contain much higher levels of LCPUFA than do terrestrial insects, a difference driven by differences in the FA composition of aquatic and terrestrial primary producers (19, 20). Aquatic primary producers, such as diatoms and dinoflagellates, are rich in EPA and DHA (21), which can be incorporated into aquatic insect tissue (22). In contrast, vascular terrestrial plants contain little to no LCPUFA but do contain their molecular precursor ALA (14), which can be either incorporated into tissue or converted to LCPUFA to a minor degree by terrestrial insects (19). As a consequence, from the perspective of LCPUFA content, aquatic insects may constitute a higher-quality food for aerial insectivores than do terrestrial insects.

However, because both aquatic and terrestrial insects contain ALA, the relative value of aquatic insects depends on the capacity of aerial insectivores to convert ALA into LCPUFA (19). The ability to elongate ALA into LCPUFA varies greatly across taxa: Strict carnivores, such as cats (23, 24), and animals from environments rich in LCPUFA, including most marine fish (25), have lost the ability to elongate ALA into LCPUFA and must obtain them directly from diet. In contrast, terrestrial herbivores appear to be relatively efficient at converting ALA to LCPUFA (26). The capacity of aerial insectivores to convert ALA to LCPUFA remains untested, but, as predators living around riparian areas with emergent aquatic insects rich in LCPUFA, they appear likely to be limited by LCPUFA content in diet.

The majority of past studies on avian FA requirements have focused on domesticated herbivorous taxa, especially chickens (e.g., refs. 27 and 28). These studies found domestic hens to be relatively efficient at elongating ALA, EPA, and DHA (e.g., refs. 27 and 28). Far fewer studies have experimentally manipulated dietary FA composition for wild birds (but see refs. 29–32). These studies have found that both dietary composition and elongation capacity of individual species affect avian FA composition (32). However, with the exception of work by Pierce et al. (31) on Red-Eyed Vireos (Vireo olivaceus), these studies have either looked at seed- and fruit-eating passerines or fish-eating seabirds. To our knowledge, no studies have explicitly examined the omega-3 FA requirements of any aerial insectivores. Therefore, we sought to understand the importance of food FA composition for aerial insectivores by varying both food quality and quantity in a balanced factorial experimental design.

In nature, the effects of food quality and quantity may be confounded because parents may provide chicks with an increased quantity of food to make up for low-quality food. To address this, we experimentally manipulated both the quantity and FA composition of food for wild-hatched nestling Tree Swallow chicks. Chicks were fed one of four diets: (i) a high-LCPUFA, high-quantity diet containing EPA and DHA (Hh); (ii) a low-LCPUFA, high-quantity diet containing high ALA and low omega-3 LCPUFA (Lh); (iii) a high-LCPUFA, low-quantity diet (Hl); and (iv) a low-LCPUFA, low-quantity diet (Ll). We assessed size-specific growth rates, body condition, immunocompetence, and basal metabolic rates (BMR) as metrics of performance. We also determined the FA composition of brain and breast muscle tissue from a subset of chicks from each treatment group.

Results

Chicks on high-LCPUFA diets (Table S1) grew significantly more rapidly than those on low-LCPUFA diets (Table S1) regardless of food quantity (ANOVA: treatment F_3,128 = 59.889, P < 0.0001; Figs. 1 A and B and 2). Diet quality was more important than quantity: Even Hl chicks grew significantly faster than did Lh chicks (Table S2 and Figs. 1B and 2). Among the high-LCPUFA groups, Hh chicks grew significantly faster than did Hl chicks (Table S2 and Figs. 1B and 2). Among the low-LCPUFA groups, there were no significant differences between Ll and Lh chicks (Table S2). There were no significant treatment differences in head−bill or tarsus growth rates between treatments (ANOVA for head−bill: treatment F_3,90 = 0.099, P = 0.96; ANOVA for tarsus: treatment F_3,90 = 2.091, P = 0.107; Fig. 1C). Thus, the differences observed were in growth rates for mass, not structural size.

Table S1.

Diet Composition

Characteristic	High-quality diet	Low-quality diet
Ingredients	Soybean protein, dried egg, chicken meal, menhaden meal, corn starch, menhaden fish oil, soybean oil, DL-methionine, dried brewer's yeast, microcrystalline cellulose, lecithin, Aztec marigold extract, wheat flour, xanthan gum, calcium carbonate, choline chloride, taurine, Lactobacillus acidophilus, Lactobacillus casei, rosemary extract, bifidobacterium thermophilum, citric acid, Enterococcus faecium, and vitamin mineral premix	Soybean protein, dried egg, chicken meal, menhaden meal, corn starch, soybean oil, flaxseed oil, DL-methionine, dried brewer's yeast, microcrystalline cellulose, lecithin, Aztec marigold extract, wheat flour, xanthan gum, calcium carbonate, choline chloride, taurine, Lactobacillus acidophilus, Lactobacillus casei, rosemary extract, Bifidobacterium thermophilum, citric acid, Enterococcus faecium, and vitamin mineral premix
Calories (± std. error)*†	5.98 ± 0.088	6.07 ± 0.086
Crude fat (± std. error),*‡ %	24.5 ± 0.433	23.7 ± 0.033
ALA,§ %	1.82	6.25
EPA,§ %	3.74	1.47
DHA,§ %	3.44	1.42
Crude fiber,‡ %	2.2	2
Neutral detergent fiber,‡ %	17.2	11.1
Acid detergent fiber,‡ %	10.1	8.5
Moisture (± std. error),‡ %	7.13 ± 0.033	7.13 ± 0.033
Crude protein (± std. error),*‡ %	52.93 ± 0.145	53.07 ± 0.463
Arginine,¶ %	3.4	3.4
Cystine,¶ %	1.1	1.1
Glycine,¶ %	2.0	2.0
Histidine,¶ %	1.2	1.2
Isoleucine,¶ %	2.5	2.5
Leucine,¶ %	4.2	4.2
Lysine,¶ %	3.6	3.6
Methionine,¶ %	2.1	2.1
Phenylalanine,¶ %	2.6	2.6
Tyrosine,¶ %	1.8	1.8
Threonine,¶ %	2.3	2.3
Tryptophan,¶ %	0.65	0.65
Valine,¶ %	2.8	2.8
Taurine,¶ %	0.25	0.25
Ash,¶ %	<9	<9
Calcium,‡ %	1.05	1.04
Phosphorus,‡ %	0.75	0.75
Potassium,‡ %	0.61	0.65
Magnesium,‡ %	0.1	0.1
Sodium,‡ %	0.218	0.226
Chloride,¶ %	0.40	0.40
Iron,‡ ppm	149	194
Zinc,‡ ppm	28	30
Manganese,‡ ppm	57	42
Copper,‡ ppm	9	9
Iodine,¶ ppm	1.4	1.4
Selenium,‡ ppm	1.04	1.05
Thiamin,¶ ppm	9	9
Riboflavin,¶ ppm	12	12
Niacin,¶ ppm	68	68
Pantothenic acid,¶ ppm	26	26
Choline chloride,¶ ppm	1,710	1,710
Folic acid,¶ ppm	4.2	4.2
Pyridoxine,¶ ppm	10	10
Biotin,¶ ppm	1.7	1.7
Ascorbic acid,¶ ppm	230	230
Vitamin B12,¶ µg/kg	48	48
Vitamin A,¶ IU/kg	11,260	11,260
Vitamin D3,¶ IU/kg	1,845	1,845
Vitamin E,¶ IU/kg	130	130
Vitamin K,¶ ppm	6	6
Beta-carotene,¶ ppm	0.41	0.41

^*Not significantly different based on Kruskal−Wallis rank sum test (calories: χ² value = 2, degrees of freedom = 2, P value = 0.368; crude fat: χ² value = 0.5, degrees of freedom = 2, P value = 0.479; crude protein: χ² value = 2, degrees of freedom = 2, P value = 0.368).

^†Measured caloric content: averages measured directly through bomb calorimetry. Units are in kilocalories per gram of dry feed.

^‡Measured content.

^§Modified fatty acid composition: averages measured directly using fatty acid extraction and chromatography methods described in Methods.

^¶Unmodified dietary components: averages based on standard Mazuri nestling feed.

Fig. 1.

Reaction norms for (A) mass, (B) size-specific mass growth rate, (C) size-specific skeletal growth rate, (D) body condition, (E) immunocompetence, and (F) BMR. Treatment means and SE bars are shown. Black represents high-LCPUFA treatments, and gray represents low-LCPUFA treatments.

Fig. 2.

Chick mass over time. Treatment means and SE bars are shown. Black circles represent our Hh treatment, gray circles represent our Hl treatment, black triangles represent our Lh treatment, and gray triangles represent our Ll treatment.

Table S2.

Size-specific growth rates analysis of variance and Tukey post hoc tests

Variable	Degrees of freedom	F value	P value	Tukey post hoc test for treatment group
Mass growth rate
Date	3	2.565	<0.1	Ll < Hh (<0.001)
Treatment group	3	59.889	<0.0001	Ll < Hl (<0.001)
Nest	8	10.917	<0.0001	Ll = Lh
Individual	32	6.954	<0.0001	Lh < Hh (<0.001)
Residuals	128			Lh < Hl (<0.001), Hl < Hh (<0.001)
Tarsus growth rate
Treatment group	3	2.091	NS	NS
Residuals	90			NS
Head−bill growth rate
Treatment group	3	0.099	NS	NS
Residuals	90			NS

NS, not significant. P > 0.05

Chicks on high-LCPUFA diets were also in significantly better condition (reflected by the ratio of mass to head−bill length and mass to tarsus length) than those on low-LCPUFA diets regardless of quantity (ANOVA for mass to tarsus: treatment F_3,115 = 225.673, P < 0.0001; ANOVA for mass to head−bill: treatment F_3,115 = 276.462, P < 0.0001; Fig. 1D). Chicks on Hh were in significantly better condition than were Hl chicks (Table S3), and Hl chicks were in significantly better condition than were Lh chicks (Table S3 and Fig. 1D). Among the low-LCPUFA groups, chicks on Lh were in significantly better condition than were Ll chicks (Table S3).

Table S3.

Body condition analysis of variance and Tukey post hoc tests

Variable	Degrees of freedom	F value	P value	Tukey post hoc test for treatment group
Mass to tarsus ratio
Date	4	11.632	<0.0001	Ll < Hh (<0.001)
Treatment group	3	225.673	<0.0001	Ll < Hl (<0.001)
Nest	8	101.571	<0.0001	Ll < Lh (<0.001)
Individual	7	3.369	<0.01	Lh < Hh (<0.001)
Residuals	115			Lh < Hl (<0.001), Hl < Hh (<0.001)
Mass to head−bill ratio
Date	4	14.397	<0.0001	Ll < Hh (<0.001)
Treatment group	3	276.462	<0.0001	Ll < Hl (<0.001)
Nest	8	131.713	<0.0001	Ll < Lh (<0.001)
Individual	7	2.755	<0.05	Lh < Hh (<0.001)
Residuals	115			Lh < Hl (<0.001), Hl < Hh (<0.001)

Chicks on high-LCPUFA diets had increased immunocompetence compared with those on low-LCPUFA diets regardless of food quantity (ANOVA: treatment F_3,80 = 38.187, P < 0.0001; Fig. 1E). Even Lh chicks had significantly higher phytohemagglutanin (PHA) immune response ratios than did Ll chicks (Table S4 and Fig. 1E). Among the high-LCPUFA groups, Hh chicks had significantly higher immune response ratios than Hl chicks (Table S4). Among the low-LCPUFA groups, there were no significant differences between Hl and Lh chicks (Table S4).

Table S4.

Immunocompetence analysis of variance and Tukey post hoc tests

Variable	Degrees of freedom	F value	P value	Tukey post hoc test for treatment group
Treatment group	3	38.187	<0.0001	Ll < Hh (<0.001)
Nest	8	1.891	<0.1	Ll < Hl (<0.001)
Residuals	30			Ll < Lh (<0.001), Lh < Hh (<0.01), Hl < Hh (<0.001), Hl = Lh

Patterns in BMR were the reverse of those in immunocompetence (Kruskal−Wallis: χ² = 7.941, df = 3, P < 0.047; Fig. 1 E and F). Hh chicks had the lowest BMR whereas Ll chicks had the highest BMR, and these differences were significant (Table S5 and Fig. 1F). Hl chicks and Lh chicks had similar BMRs, which were not significantly different (Table S5 and Fig. 1F). We also found that Ll chicks had significantly higher BMR than Hl chicks, and Lh chicks had significantly higher BMR than Hh chicks (Table S5 and Fig. 1F).

Table S5.

BMR Kruskal−Wallis and Dunn tests

Treatment group	χ2 value	P value	Dunn post hoc test
Mass-specific BMR	7.941	<0.05	Ll < Hh (<0.01), Ll = Hl, Ll < Lh (<0.05), Lh = Hh, Lh = Hl, Hl < Hh (<0.05)
Whole-organism BMR	6.0362	0.1099	Hh < Hl (<0.05), Hh = Lh, Hh < Ll (<0.05), Lh = Hl, Ll = Ll, Ll = Hl

Degrees of freedom = 3.

We found both treatment and tissue-based differences in FA composition (Tables S6 and S7 and Fig. 3). Hh chicks had significantly higher percentages of EPA in brain, whereas chicks on both high-LCPUFA diets had significantly higher percentages of EPA in muscle (Kruskal−Wallis for brain EPA: χ² = 7.567, df = 3, P = 0.056; Kruskal−Wallis for muscle EPA: χ² = 9.088, df = 3, P = 0.028; Table S6 and Fig. 3). The percentage of DHA in brain was significantly higher in Lh chicks compared with chicks on either high-LCPUFA diet (Kruskal−Wallis: χ² = 8.316, df = 3, P = 0.040; Table S6 and Fig. 3B). In contrast, the percentage of DHA in muscle was highest in Hl chicks, but it was only significantly higher than the percentage of DHA in muscle of Ll chicks (Kruskal−Wallis for muscle DHA: χ² = 3.882, df = 3, P = 0.275; Table S6 and Fig. 3D). The percentage of total omega-3 FAs in brain was also significantly higher in chicks on Lh diets compared with chicks on high-LCPUFA diets (Table S6). We found no significant differences in either the proportion of total omega-6 FAs in either brain or muscle or the percentage of total omega-3 FAs in muscle (Table S6). Muscle had significantly higher percentages of EPA and total omega-6 FAs compared with brain, but it had similar percentages of total omega-3 FAs and significantly less DHA (Table S6).

Table S6.

Fatty acid composition Kruskal−Wallis and Dunn tests

Treatment group	χ2 value	P value	Dunn post hoc test
Brain EPA	7.5662	<0.10	Ll = Hl, Ll = Lh, Hl = Hh, Ll < Hh (<0.01), Lh < Hh (<0.05), Hl < Hh (<0.05)
Brain DHA	8.3162	<0.05	Ll < Hh (<0.10), Ll = Hl, Ll = Lh, Lh < Hh (<0.01), Hl = Hh, Lh < Hl (<0.05)
Muscle EPA	8.3162	<0.05	Ll < Hh (<0.1), Ll < Hl (<0.05), Ll = Lh, Lh < Hh (<0.05), Lh < Hl (<0.01), Hl = Hh
Muscle DHA	3.8824	NS	Ll = Hh, Ll < Hl (<0.05), Ll = Lh, Lh = Hh, Lh < Hl (<0.1), Hh < Hl (<0.1)
Brain omega-3	6.6397	<0.10	Ll = Hh, Ll = Hl, Ll < Lh (<0.1), Hl = Hh, Hl < Lh (<0.05), Hh < Lh (<0.05)
Brain omega-6	4.4559	NS	Ll < Hh (<0.1), Lh = Hh, Lh < Hl (<0.1), Lh < Ll (<0.1), Hl = Hh, Hl < Lh (<0.1)
Muscle omega-3	4.1691	NS	Ll = Hh, Ll = Hl, Ll = Lh, Lh = Hh, Lh < Hl (<0.05), Hl = Hh
Muscle omega-6	5.4485	NS	Lh = Hh Lh = Hl, Lh < Ll (<0.1), Hl < Ll (<0.05), Hl = Hh, Hl < Ll (<0.05)

Degrees of freedom = 3. NS, not significant. P > 0.05

Table S7.

Brain and muscle fatty acid composition two-sample t tests

Tissue type	Degrees of freedom	t value	P value
EPA	16.117	−14.9074	<0.001
DHA	29.437	11.0112	<0.001
Omega-3	28.832	0.9059	NS
Omega-6	17.092	−18.952	<0.001

NS, not significant. P > 0.05

Fig. 3.

Fatty acid composition results for (A) brain EPA, (B) brain DHA, (C) muscle EPA, and (D) muscle DHA. Treatment means and SE bars are shown. Black circles represent our Hh treatment, gray circles represent our Hl treatment, black triangles represent our Lh treatment, and gray triangles represent our Ll treatment.

Discussion

We asked if food quality, in terms of LCPUFA, was as important as food quantity for a model aerial insectivore species, the Tree Swallow. We manipulated food quantity and FA composition in a fully factorial design and assessed performance in Tree Swallow chicks by measuring changes in size (mass, head−bill length, and tarsus length), body condition, and differences in immunocompetence and BMR at the conclusion of the experiment. Overall, we found strong evidence that LCPUFA content is as important, if not more important, than food quantity for aerial insectivores (Figs. 1 and 2). We also found significant differences in the FA composition of chicks on different diets, which suggested that the chicks preferentially retained EPA and DHA (Fig. 3).

Hh chicks grew the fastest and were in the best condition, whereas Ll chicks grew the slowest and were in the poorest overall condition (Figs. 1 and 2). Interestingly, Hl chicks performed better than did Lh chicks (Figs. 1 and 2). Increasing the quantity of low-LCPUFA food had no effect on growth rates or condition. Body mass and condition are two of the most important predictors of Tree Swallow fledgling success and survival in natural systems (33). Our results suggest that wild chicks with access to high-quality food resources, such as aquatic insects, are likely to be in better condition than those with access to only lower-quality terrestrial food resources. In addition, more high-quality food appears to further increase body mass and condition, whereas more low-quality food does not.

We found no significant differences in head−bill or tarsus growth rates across treatments (Fig. 1C). This finding provides evidence that there is strong pressure to develop at a specific rate, even at the cost of overall condition. Tree Swallows nestlings, like other passerines, suffer high mortality from predation (34), and, although body mass and condition are strong predictors of survival after fledging, there appears to be greater pressure to quickly reach fledgling size to be ready to fledge if threatened by predation (35, 36).

We found that food quantity and quality had significant interacting effects on immunocompetence, measured as PHA ratio, across treatments (Fig. 1E). In birds, PHA ratio is an indicator of acquired T-cell proliferation and the ability to produce lymphocytes in response to pathogens (37). We found that Hh chicks had the highest PHA ratios whereas Ll chicks had the lowest ratios and Hl and Lh chicks had equivalent, intermediate PHA swelling responses. Our results suggest that wild Tree Swallow chicks with access to more food, especially high-quality aquatic insects containing EPA and DHA, may be more likely to mount an effective immune response (3, 33). In addition to predators and food deprivation, pathogens are a significant source of early mortality in nestling Tree Swallows (38), and greater immunocompetence from higher food quantity and quality likely increases Tree Swallow chick survival.

Food quantity and quality also had significant interacting effects on BMR across treatments (Fig. 1F). Hh chicks had the lowest metabolic rates and Ll chicks had the highest BMR, either mass-corrected or whole-organism (Table S5). Our low- and high-LCPUFA feeds had equal caloric content (Table S1), so differences in BMR are likely due to effects of feed FA composition and total LCPUFA content, not total energy. The negative relationship between total LCPUFA content of feed and BMR could have resulted from costs of ALA elongation and desaturation to LCPUFA. For example, although feed for Hl and Hh chicks had the same FA composition, Hl chicks consumed less total LCPUFA compared with Hh chicks, and thus may have required additional energy to convert ALA into LCPUFA. Our findings agree with those of previous studies; for example, Pierce et al. (31) found that increasing unsaturated FAs in diet decreased peak metabolic rate for Red-Eyed Vireos. Across all treatments, our findings support the inverse relationship observed by Hulbert et al. (16) between avian body mass and both BMR and breast muscle DHA across bird species.

The FA composition of chicks provided evidence for both dietary accumulation and preferential retention of the long-chain omega-3 FAs EPA and DHA (Fig. 3). In brain tissue, the percentage of EPA was highest in Hh chicks, whose diet contained both the highest percentage and the greatest total amount of EPA (Fig. 3A). Chicks on Hh and Hl diets had the highest percentages of EPA in muscle tissue (Table S5 and Fig. 3C). The Lh and Ll diets may not have contained sufficient amounts of EPA to accumulate dietary EPA, or Tree Swallows may be inefficient at converting ALA to EP;. this suggests that EPA accumulation in the Tree Swallow tissues may be based on dietary availability of EPA, which is consistent with findings in other taxa (28).

In brain tissue, the percentage of DHA was highest in Lh chicks (Fig. 3B). This finding could have stemmed from either increased elongation of ALA or preferential LCPUFA retention in Lh chicks. We suggest a combination of elongation and preferential retention may have been at work: Lh chicks would have had more energy to devote to elongation than did low-quantity chicks and would have had more non-LCPUFA FAs in diet to preferentially oxidize for fuel than did high-LCPUFA chicks. Tree Swallow muscle tissue had significantly less DHA than did brain (Table S6 and Fig. 3), potentially because phospholipid DHA is a key component of neural tissue (39). In muscle tissue, the proportion of DHA in muscle was highest in Hl chicks (Fig. 3D), which we suggest was due to preferential retention because Hl chicks were limited in energy but not in LCPUFA.

Chicks on low-LCPUFA and/or low-quantity diets may have either converted ALA to LCPUFA or preferentially retained LCPUFA already present in tissue. Studies suggest that chicken embryos preferentially remove LCPUFA from yolk (27). However, this does not appear to be the case with altricial chicks, such as Barn Swallows (Hirundo rustica), which contain much less DHA at hatch than do precocial birds (40). We were unable to control maternal FA investment in eggs or parental feeding during the chicks’ first few days of life, and all chicks originated from nest boxes on or near water. Thus, Hl and Lh chicks likely preferentially retained DHA from eggs and early life while oxidizing other dietary fats for energy. In contrast, Hh chicks may paradoxically have had lower tissue DHA concentrations precisely because DHA was abundant in diet, obviating the need to preferentially retain DHA.

Past work on chickens found that higher levels of LCPUFA in diet translated into increased proportions of LCPUFA in breast muscle (e.g., ref. 28). We found that increasing the concentrations of LCPUFA in diet did not necessarily result in increased LCPUFA content in Tree Swallow tissue. Instead, our findings on aerial insectivore chicks are closer to those of past studies on freshwater zooplankton, which have found preferential retention and biomagnification of LCPUFA compared with other FAs regardless of food quality (41). This suggests that there is strong pressure for aerial insectivores to obtain and retain LCPUFA in the face of poor conditions. Our performance data suggest that, when food quantity or quality are low, saving LCPUFA for future use instead of burning them as fuel may result in lower body mass and condition. Further studies using compound-specific stable isotope tracers (e.g., enriched δ¹³C) will be necessary to determine if Tree Swallow chicks are able to convert ALA into LCPUFA and thus whether LCPUFA are beneficial or absolutely essential components of diet.

Previous work on Tree Swallows has attempted to link Tree Swallow breeding season and nestling success with food availability (9, 11). Our findings suggest that the abundance of high-quality aquatic insects relative to Tree Swallow phenology may be a better predictor of breeding success than overall insect abundance. Aquatic insect abundance peaks earlier than terrestrial insect abundance (42), and aquatic insects are often the only available food early in the breeding season (43). Total insect abundance peaks later in the breeding season, yet Tree Swallows complete laying long before peak insect abundance, and their breeding success decreases with lay date (9). Ecologists have generally interpreted these findings to indicate that laying, although earlier than peak insect abundance, places chick rearing, thought to be the most energy-demanding phase of the breeding cycle, at a time of peak food availability (44). Our findings suggest an alternative interpretation, that Tree Swallows, and other birds, may be under selection to time their breeding seasons when insects high in LCPUFA are most available.

Our findings have significant implications for aerial insectivore conservation. Most North American aerial insectivores, including the Tree Swallow, are associated with aquatic or riparian habitats (43). We found evidence that feed containing LCPUFA representative of aquatic insects improves multiple metrics of Tree Swallow performance and that they preferentially retain these high-quality fats. Our study suggests that large quantities of terrestrial insects low in LCPUFA are, at best, no better than even small amounts of aquatic insects, even if they have high amounts of the LCPUFA precursor ALA. Land conservation is not enough for aerial insectivores to survive and thrive: Managers must conserve aquatic habitats that provide aerial insectivores with the highest-quality LCPUFA-rich aquatic insects.

Methods

We collected 44 wild Tree Swallow chicks from nest boxes around Ithaca, New York, from 29 May 2015 to 7 June 2015. To prevent parental abandonment of chicks, we removed all chicks from each nest box. All animal work was approved under Cornell Institutional Animal Care and Use Committee protocol 2001-0051, New York State Department of Environmental Conservation Scientific Collection Permit 1477, and United States Fish and Wildlife Service Migratory Bird Scientific Collection Permit 757670.

Upon return to the laboratory, we weighed and sorted chicks into groups of three to four birds to receive one of four feeding treatments: (i) Hh; (ii) Lh; (iii) Hl; and (iv) Ll. The high- and low-LCPUFA diets were not significantly different in calories, moisture, crude protein, or crude fat (Table S1) and differed only in FA composition (Table S1). All diets were based upon standard commercial Mazuri nestling feed (www.mazuri.com/mazurihandfeedingdiets-1.aspx). Standard nestling diets contained soybean oil as their principal fat source. Our high-LCPUFA diets included a substitution of stabilized menhaden oil for soybean oil in a ratio of 7:3, and low-LCPUFA diets included a substitution of flax oil for soybean oil in a ratio of 1:3. The resulting high-LCPUFA diets contained ∼1.82% ALA, 3.74% EPA, and 3.44% DHA, and low-LCPUFA diets contained ∼6.25% ALA, 1.47% EPA, and 1.42% DHA (Table S1). Low-quantity chicks were fed 4.5% of body mass per feeding session (the point at which begging still occurred at the end of the session), and high-quantity chicks were fed 6% of body mass per session (produced chick satiation at the end of the feeding session) of body mass per feeding. We grouped chicks of similar initial sizes together in nest groups so as to avoid underfeeding or overfeeding individual chicks, and we randomly split up chicks from the same clutch to avoid genetic effects. There were multiple replicates of each food quality and quantity treatment that covered the full range of initial chick masses. Additional details on diets and care are described in SI Methods.

Each chick was weighed four times daily with an Ohaus Scout Pro balance, and the average of that mass was used for calculations. We also measured the head−bill and tarsus length of each chick to the nearest 0.01 mm a minimum of two times over the course of the experiment, with Mitutoyo Digimatic 500 calipers. Growth rates were calculated as [ln(mass or length on day x) – ln(mass or length on day 0)]/(day x – day 0), and body condition was calculated as both the ratio of mass to head−bill length and the ratio of mass to tarsus length. To measure immunocompetence, we used the simplified protocol described by Smits et al. (45) and detailed in SI Methods. To determine BMR, we used an open-flow pull-mode FoxBox respirometry setup coupled with a climate-controlled chamber at a flow rate of ∼490 mL/min following the methods of Lighton (46) and detailed in SI Methods.

We determined the whole-tissue FA composition of brain and pectoral muscle for a subset of chicks from each treatment (n = 4). After euthanasia, we dissected and weighed out brain and pectoral muscle samples from four chicks per treatment. FA methyl esters (FAMEs) were extracted from whole tissues using a modified one-step method (47, 48) and quantified using a BPX-70 (SGE Inc.) column and an HP5890 series II gas chromatography flame ionization detector. Chromatogram data were processed using PeakSimple. Response factors were calculated using the reference standard 462a (Nucheck prep). FAMEs were identified using a Varian Saturn 2000 ion trap with a Varian Star 3400 gas chromatography mass spectrometer run in chemical ionization mass spectrometry mode using Acteonitrile as reagent gas. FA composition data are expressed as percent of total FA. We also calculated total omega-3 PUFAs and total omega-6 PUFAs.

We analyzed mass, size-specific growth rates for mass, tarsus length, size-specific growth rates for tarsus length, head−bill length, size-specific growth rates for head−bill length, the ratio of mass to tarsus length, the ratio of mass to head−bill length, and PHA ratio through ANOVA, using treatment group (the interaction of LCPUFA content and food quantity: Hh, Hl, Lh, and Ll), nest, and individual as predictor variables. For all performance metrics except PHA ratio, which was only measured at the end of the experiment, we also ran analysis of covariance with experiment date as a covariate. We used post hoc Tukey tests to interpret the direction and significance of differences between treatment groups for variables that were significant as main effects and assessed relative support between models using Aikaike's Information Criterion. To detect differences in our smaller datasets on BMRs and brain and muscle FA composition, we used nonparametric Kruskal−Wallis tests and performed Dunn tests to perform pairwise comparisons between treatment groups (Hh, Hl, Lh, and Ll). We also compared differences between brain and muscle fatty acid composition using Welch’s two-sample t tests. All statistical analyses were performed in R (version 3.2.2).

SI Methods

Chick Care.

We labeled each chick with nail polish on its head to distinguish between individuals within nest groups. Nests consisted of folded paper towel layers in plastic bowls. Bedding was changed approximately hourly or when soiled with feces from 0700 to 2100 hours. Two nests bowls were placed in a plastic box covered with a towel, and we placed the box over a heating mat equipped with a thermostat placed within the nest. Nest temperatures were kept at ∼30 °C throughout the experiment, and chicks experienced 12 h of light and 12 h of dark. We cleaned chicks with baby wipes to remove food and fecal residue twice per day.

Chicks were fed for ∼12 h a day when at least half of chicks on the high-quantity diets were begging. All chicks were fed via 1-mL sterile syringes that were washed between feedings and replaced daily. Feeds were made up daily by blending together a 2:1 ratio of feed to water and then refrigerating the mixture until needed. A subset of feed was reblended and warmed to ∼40 °C in a water bath every 2 h.

Diets.

All feeds were based on modifications of Mazuri nestling diets and differed significantly only in fatty acid composition (Table S1). High-LCPUFA diets included a substitution of stabilized menhaden oil for soybean oil in a ratio of 7:3, and low-quality diets included a substitution of flax oil for soybean oil in a ratio of 1:3. The resulting high-LCPUFA diets contained ∼1.82% ALA, 3.74% EPA, and 3.44% DHA, and low-LCPUFA diets contained ∼6.25% ALA, 1.47% EPA, and 1.42% DHA (Table S1). Feed fatty acid analyses were conducted following the same methods described in Methods. Low-quality diets contained minor amounts of EPA and DHA from fishmeal included in nestling diets as a protein source. The two diets were not significantly different in calories, crude protein, or crude fat. Caloric content was measured through bomb calorimetry at the Cornell University Human Nutritional Chemistry Service Laboratory. Additional feed composition analyses were conducted by the Dairy One Forage Lab (Dairy One Cooperative). Diet specifications are provided in Table S1. Chicks on high-quantity diets were fed 6% of body mass each feeding, which was the maximum crop capacity, and those on low-quantity diets were fed 4.5% of body mass each feeding, at which additional begging still occurred.

Immunocompetence.

Chicks were injected with a treatment dosage of 100 μg of PHA-P dissolved in 20 μL of PBS (Sigma-Aldrich; see ref. 49). We measured the initial thickness of the patagium and injected PHA solution into the middle of the patagium, making a mark on the injection site, and then measured the magnitude of the swelling reaction again after 6 ± 0.5 h [for the use of a 6-h period, see, e.g., ref. 50; recently Bonato et al. (51) have shown that there is no statistical difference in the PHA response between 6 h and 24 h after injection]. Tissue thickness at the injection site was measured three times (accuracy 0.01 mm) with a digital micrometer (Mitutoyo) by applying pressure to the point where the skin is lightly moved from pressure of the micrometer, and using the average of these three measures for further analysis. The PHA-induced swelling response index was calculated as the average tissue thickness 6h after the treatment divided by the average thickness before the PHA injection.

Metabolic Rate Calculations.

We acclimated chicks to a T_a of ∼33 °C for 2 h and took respirometry measurements in a multiplexed setup for 90 min, where respired gas was sampled from each individual’s chambers for a period of 10 min. To account for sensor drift, we took baseline measurements lasting 10 min every 30 min and after the completion of the trial. A 3-m copper coil constructed of 6.35-mm-i.d. tubing was placed in-line upstream inside the respiration chamber to equilibrate the temperature between the chamber and the incurrent airstream.

Respired gas was analyzed using a Sable Systems FoxBox field oxygen analyzer in a pull setup following the recommendations in Lighton (46), where samples were scrubbed of water vapor before CO₂ measurements and of CO₂ and water vapor before O₂ measurements using a combination of Drierite (W. A. Hammond Drierite Co.), Soda Lime, and Ascarite (P/N 223913; Sigma-Aldrich). All Drierite was exposed to ambient CO₂ conditions for a minimum of 2 min to equilibrate with ambient atmospheric conditions (52). Gas samples were corrected for dilution effects through the measurements of water vapor pressure and atmospheric pressure throughout the experiment (53). At the beginning and end of each experiment, flow rate calibration was performed by measuring the time to displace an inverted graduated cylinder of water of a known volume with excurrent air (53). All airstream connection tubing was 6.35-mm-i.d. Bev-a-line IV.

Uncompensated flow rate (FR_u) was corrected into standard power-corrected flow rate (FR_i) measured from the FoxBox using the following formula: (i) FR_i = FR_u * (kPa – WVP)/kPa, where kPa is the atmospheric pressure and WVP is the water vapor pressure (both in kilopascals). The rate of oxygen consumption was calculated using equation 11.1 in Lighton (46) from the incurrent (subscript i) and the excurrent (subscript e) airstream measurements. (ii) VO₂ = FR_i (F_iO₂ − F_eO₂)/(1 – F_iO₂), where FRi is the rate for unscrubbed incurrent air, FiO₂ is the measurement from the baseline incurrent sample, and FeO₂ is the rate from the sample of exhaled carbon dioxide free of water vapor (Drierite) and carbon dioxide (Soda Lime and Ascarite). Rate of carbon dioxide production was calculated using equation 11.6 in Lighton (46) from the incurrent (subscript i) and the excurrent (subscript e) airstream measurements. (iii) VCO₂ = [FR_e (F_eCO₂ – F_iCO₂) − F_eCO₂(VO₂)]/(1 – F_eCO₂). With these measures, the rate of carbon dioxide production and oxygen consumption during our respirometry trials were then converted into a mass-specific rate of energy use per unit of time (hour) using the following formula: (iv) MR (milliliters O₂ per hour per gram) = VO₂ (milliliters) * (60)/(individual mass), where the metabolic rate (MR) is the volume of oxygen consumed multiplied by the number of minutes in an hour divided by the mass of the organism.