Vertebral samples were collected from nine blue sharks and six porbeagles caught across the North Atlantic. Three blue sharks were captured in offshore waters south of Canary Islands (eastern blue sharks), three in the mid-Atlantic Ridge area northwest of the Azores (central blue sharks), and three in oceanic waters between Cape Hatteras and the Gulf of Maine (western blue sharks; Figure 1). Blue sharks were bycaught by pelagic longlines targeting swordfish, tunas and sharks, by research and commercial fishing vessels using pelagic longlines or during shark fishing tournaments using rods and reels. Two porbeagles were bycaught on the continental shelf in the Celtic Sea during commercial gillnet fishery targeting gadiform fish and retained under dispensation as part of a fishery bycatch study (eastern porbeagles; Ellis and Bendall, 2015; Ellis et al., 2015). One porbeagle was bycaught by a fisherman near the Faroe Islands and donated to research. Three porbeagles were caught between Massachusetts and Grand Banks (off southern Newfoundland) by commercial vessels using pelagic longlines (western porbeagles; Figure 1; for metadata see Table 1).

Trunk vertebrae were excised from above the branchial chamber in porbeagles and western blue sharks, and vertebrae were removed posterior to the skull in central and eastern blue sharks. All vertebrae were immediately frozen after dissection. Vertebral centra were then defrosted, physically cleaned of excess muscle and connective tissue and air-dried for 3–10 days, depending on dimensions (Kim and Koch, 2012). A 6 mm section was cut from each centrum using a low-speed diamond-bladed Isomet^® saw. This section was divided in two halves: a thinner (5–5.4 mm) section was cut from one half and used for bulk tissue isotope analysis, the other half was used for amino acid analysis. Approximately equidistant samples (mean ± SD interval between adjacent samples among vertebrae: 0.59 ± 0.07 mm; Table 1) were collected along the vertebral radius by manually cutting the thinner half section with a scalpel. Equidistant samples were chosen over sampling growth bands, as growth bands can be difficult to identify, particularly in blue sharks (Magozzi personal observation), and are compressed toward the vertebral edge due to reduced growth/accretion rates with age (Natanson et al., 2002; Skomal and Natanson, 2003), which makes it hard to sample each single growth band and obtain sufficient material for isotopic analysis. For this reason, the temporal resolution of isotope chronologies from equidistant samples was higher (e.g., <1 year) closer to the vertebral centrum and lower toward the edge (e.g., a few years) but sufficient to depict isotopic variation at ecologically relevant time-scales for all specimens and life-history stages.

Vertebral samples were decalcified by exposure to 2 ml of 1 M HCl for 48 h to remove any potential influence of ¹³C-enriched bioapatite on the δ¹³C values of bulk cartilage collagen (Hussey et al., 2012; Kim and Koch, 2012; Christiansen et al., 2014; see also Tuross et al., 1988). During treatment samples were kept at a constant temperature of 4°C to reduce decalcification rates and prevent collagen dissolution and damage. Samples were then washed five times with 2 ml of Milli-Q water, frozen and freeze-dried.

Samples weighing <0.5 mg (i.e., the minimum weight required for dual carbon and nitrogen isotope analysis of bulk protein) were combined to the lighter adjacent sample(s) until the combined weight exceeded 0.5 mg. Due to the typical bowtie shape of vertebral sections, more samples generally required to be combined closer to the centrum or even further along the radius in portions of the vertebra characterized by large water or air content. Thus, the combination of samples often resulted in decreased temporal resolution at the core. Single or combined samples with a weight of >1.5 mg were powdered and ∼1.25 mg of material was randomly selected for analysis. Blue shark samples were analyzed at full resolution (i.e., all adjacent samples), porbeagle samples at half resolution (i.e., one every other sample) for distances along the vertebral radius <7 mm and at full resolution afterward. This choice was made to minimize analytical costs, whilst also achieving a sufficient temporal resolution throughout ontogeny in the two species. Evidence from satellite-tagging studies indicates that, while juvenile blue sharks may undergo extensive latitudinal movements (Queiroz et al., 2010; Vandeperre et al., 2014), porbeagles tend to remain on the shelf until they conduct an ontogenetic habitat shift toward offshore waters with the onset of maturity (Biais et al., 2017) and are therefore unlikely to express large isotopic variation during early life-history stages. The vertebral radius at maturity is estimated to be ∼15 mm in porbeagles from age at maturity equal to 13 years (Natanson et al., 2002); a value of 7 mm was selected here as a more conservative cutoff to account for approximations in the sampled interval.

The procedures for sample preparation for amino acid isotope analysis were similar to those for bulk tissue analysis except that the half section used for this analysis was 6 mm thick and more adjacent samples required to be combined at the core and along the vertebral radius, as the minimum weight for compound-specific analysis is 4 mg (prior to hydrolysis and derivatization; Houghton personal communication). Vertebral samples were analyzed at full resolution for central and eastern blue sharks, at half resolution for all other specimens to minimize analytical costs. Thus, the temporal resolution of amino acid analysis was generally lower than for bulk tissue analysis but sufficient to identify breakpoints in amino acid isotope chronologies.

For compound-specific analysis, samples were initially acid-hydrolyzed (after combining adjacent samples) in 1 ml 6 N HCl at 110°C (temperature range: 105–115°C) for 20 h to liberate single amino acids from protein. Vials were tightly closed during hydrolysis to prevent sample and acid evaporation. After hydrolysis, samples were dried with a gentle stream of N₂ at 60°C (temperature range: 55–65°C) for 30–60 min, depending on sample weight and excess HCl volume. Samples were then dissolved in 100 μl 0.1 N HCl and preserved at 4°C until derivatization.

A total of 50 μl of the sample+HCl solution were derivatized by adding 35 μl of methanol, 30 μl of pyridine, and 15 μl of methyl chloroformate (MCF); after additions, the mixture was vortexed for 30 s. Amino acid-MCF derivates were separated from the reaction mixture by adding a volume of chloroform that depended on sample original weight: 50 μl for 4–10 mg, 75 μl for 11–22 mg, and 100 μl for >23 mg. After chloroform addition, the mixture was vortexed for another 30 s. At this stage the reaction mixture was stratified and the organic layer of amino acid-MFC derivates was separated from the aqueous layer using electrophoresis pipette tips and used for analysis.

Carbon and nitrogen isotopic compositions of bulk cartilage collagen were measured using a PDZ Europa ANCA-GSL elemental analyzer interfaced to a PDZ Europa 20–20 isotope ratio mass spectrometer (Sercon Ltd., Cheshire, United Kingdom) at the UC Davis Stable Isotope Facility. Samples from Northeast Atlantic porbeagles were transported to the United States accompanied by a CITES Export Permit (No. 529713/01). All samples were analyzed in duplicate. Analytical quality-control was assessed through repeated measurement of laboratory-internal reference materials (nylon, glutamic acid, and USGS-41; for details see Magozzi, 2017), which were in turn calibrated to international reference materials. Nylon was used as a drift reference to correct for variation over the course of a run, glutamic acid to calculate the elemental totals and apply a linear correction to the isotope values. The isotope values were then scaled to two reference materials of known isotopic compositions: nylon and USGS-41. Precision was calculated as the standard deviation of isotope values for bovine liver NIST 1577 across runs (i.e., for the time period over which the samples were analyzed), accuracy as the difference between the mean isotope value for bovine liver and its accepted value. Precision was 0.03‰ for δ¹³C values and 0.10‰ for δ¹⁵N values, accuracy 0.01‰ for δ¹³C values and 0.06‰ for δ¹⁵N values. Final δ¹³C and δ¹⁵N values were expressed relative to the international reference materials V-PDB (Vienna PeeDee Belemnite) and Air, respectively. The raw ratio (R) of the heavy (¹³C, ¹⁵N) to the light (¹²C, ¹⁴N) isotope in a sample was converted to a δ-value expressed in ‰ using Eq. (1):

where X is the heavier isotope. R_standard refers to the raw ratio of an internationally accepted reference gas for the analyzed isotope.

Derivatized samples were analyzed with gas chromatography-combustion-isotope ratio monitoring mass spectrometry (GC-C-irm-MS) at the Woods Hole Oceanographic Institution. Samples were dissolved in dichloromethane (DCM) and injected on column in splitless mode at 260°C and separated on an Agilent VF-23ms column (length: 30 m, inner diameter: 0.25 mm, film thickness: 0.25 μm; Agilent Technologies, Wilmington, DE, United States) in an Agilent 689N Gas Chromatograph (GC). Sample concentrations were adjusted to achieve a minimum of 2 V output for all amino acids. Gas chromatography conditions were set to optimize peak separation and shape as follows: initial temperature 80°C held for 1 min; ramped to 260°C at 6°C min^{– 1}; held for 3 min. The separated amino acid peaks were combusted online in a Finnigan GC-C continuous flow interface at 930°C and then measured as CO₂ on a Thermo Finnigan Mat 253 irm-MS (Agilent Technologies 689N GC). Standardization of runs was achieved using intermittent pulses of a CO₂ reference gas of known isotopic composition. All samples were run in duplicate. Carbon isotopic compositions were recovered for the essential amino acids: valine, isoleucine, leucine, threonine, and phenylalanine; and for the non-essential amino acids: alanine, glycine, proline, aspartic acid, and glutamic acid. A typical chromatogram from the analysis of carbon isotope ratios in amino acid-MCF derivates is displayed in Supplementary Figure 2. Values of δ¹³C for threonine showed poor chromatography and inconsistencies among replicates, therefore were omitted from data analysis.

Two internal lab reference materials (AA1 and AA2) were created with single amino acids with known δ¹³C values. Additionally, 20 mg of lyophilized muscle from Atlantic cod (Gadus morhua, Linnaeus 1758) were acid-hydrolyzed and re-dissolved in 100 μl 0.1 N HCl and 50 μl of the solution used as reference material. All reference materials were concurrently derivatized with the samples and dissolved in 100 μl chloroform. Derivatization correction factors were determined for each amino acid based on known δ¹³C values of the amino acids in the reference materials prior to derivatization, and applied to each sample to adjust for the introduction of exogenous carbon and kinetic fractionation from derivatization. As no international reference materials are currently available for the analysis of carbon isotope ratios in individual amino acids and no large inter-laboratory calibration has yet come up with consensus δ¹³C values, mean values across three laboratories (i.e., the Fish Ecology Laboratory, the Marine Biological Laboratory, and the UC Davis Stable Isotope Facility; Supplementary Table 1) were used. Reference materials were interspersed among samples and analyzed 4–6 times. Precision was determined as the standard deviation of the δ¹³C values for each amino acid in the cod standard during the time period of sample analysis. Precision was 0.31‰ for valine, 0.34‰ for isoleucine, 0.40‰ for leucine, 0.30‰ for threonine, 0.26‰ for phenylalanine, 0.13‰ for alanine, 0.30‰ for glycine, 0.32‰ for proline, 0.40‰ for aspartic acid, 0.22‰ for glutamic acid.

To account for differential vertebral growth rates throughout ontogeny, linear distance of samples along the vertebral radius was converted to age using validated (for blue shark and porbeagle trunk vertebrae; Natanson et al., 2002; Skomal and Natanson, 2003) and estimated (for blue shark cervical vertebrae; this study) vertebral width:body size relationships and validated body size:age (Von Bertalanffy) growth curves for blue sharks (Skomal and Natanson, 2003) and porbeagles (Natanson et al., 2002; for details on age estimation see Supplementary Material). Clearly, due to reduced growth/accretion rates with age, the same sampled interval corresponded to “less years” integrated at smaller distances along the vertebral radius, “more years” integrated at larger distances along the radius. Thus, conversion of sample distance to age accounted for time-series expansion closer to the vertebral centrum and time-series compression toward the edge. Differences in growth and longevity between blue sharks and porbeagles were accounted for by parameters in vertebral width:body size relationships and Von Bertalanffy growth curves. Age at birth was 0 years and age at maturity was 5 and 13 years in blue sharks and porbeagles, respectively (Natanson et al., 2002; Skomal and Natanson, 2003); these values were used as cutoffs to distinguish between pre-birth, juvenile and adult life-history stages when examining isotope chronologies. Sample age is not displayed in plots for pre-birth samples, as Von Bertalanffy growth curves are not applicable to estimate age during the pre-birth stage. Age classes were determined for each species by subsetting the maximum age range across individuals by 1 year-intervals, and samples were assigned to their age classes based on their estimated age. Age classes were sufficiently small to allow the identification of relatively continuous ontogenetic patterns across individuals.

Principal component analysis (PCA) was applied to the δ¹³C values of essential and non-essential amino acids to reduce the data to two dimensions (PC1 and PC2) explaining the majority of variance in δ¹³C values of amino acids. For non-essential amino acids, spacing (i.e., difference) in δ¹³C values was calculated between independent pairs of glycolytic (lipid-derived) and Krebs cycle (lipid+protein-derived) amino acids (glycine and alanine compared to glutamic acid and proline, respectively) and explored as a potential indicator for shifts in macronutrient contents of diets and/or changes in metabolic routing of macronutrients to amino acid synthesis (Wolf et al., 2015; Wang et al., 2018, 2019).

In order to test for common isotope patterns across individual blue sharks and porbeagles, the values of each variable (i.e., bulk protein δ¹³C and δ¹⁵N values, PC1 scores for δ¹³C values of essential and non-essential amino acids and spacing in δ¹³C values between glycolytic and Krebs cycle amino acids) in each sample were normalized to the individual mean (life history-normalization) to emphasize within-individual variation in ontogenetic profiles, whilst reducing among-individual differences. Ontogenetic patterns in life history-normalized δ¹³C and δ¹⁵N values of bulk protein and differences between species and among areas and individuals were tested with generalized additive mixed models (GAMMs). Estimated sample age was added as a smoother, species and area as parametric fixed effects and individuals as a random effect (Pinheiro and Bates, 2000; Zuur et al., 2014). Model fits were calculated for full models and optimal models were selected with Akaike Information Criteria (AIC; Zuur et al., 2014). Generalized additive mixed models were fitted using the mgcv R package (Wood, 2006) and all analysis were performed using R v.4.0.2 (R Core Team, 2019).

Bulk protein δ¹³C values ranged between -16.73 and -13.95‰ (mean ± SD: -15.13 ± 0.65‰) in blue sharks and between -15.80 and -13.71‰ (mean ± SD: -14.55 ± 0.38‰) in porbeagles. Values of δ¹⁵N varied between 9.09 and 14.14‰ (mean ± SD: 11.47 ± 1.11‰) in blue sharks and between 9.42 and 16.29‰ (mean ± SD: 12.84 ± 1.47‰) in porbeagles. The δ¹⁵N range in porbeagles was, however, strongly influenced by ¹⁵N-enriched subadult and adult samples from Northwest Atlantic individuals.

Despite high among-individual variability in individual isotopic profiles (Supplementary Figure 4), we identified common, broad ontogenetic patterns in life history-normalized δ¹³C and δ¹⁵N values (δ¹³C_n, δ¹⁵N_n) of bulk protein across individuals of each species (Figure 2). Ontogenetic patterns in δ¹³C_n and δ¹⁵N_n values were non-linear (smoother ‘estimated sample age’: p-value = 2⋅10^–16, F_δ^13C = 26.46; F_δ^15N = 26.29) and differed significantly between species (maximum p-value = 0.006, F = 7.62; Supplementary Tables 2, 3). In blue sharks, δ¹³C_n values increased progressively during juvenile and early adult growth, reaching a relatively steady state in the adult life stage (Figure 2A). Values of δ¹⁵N_n increased during juvenile growth and decreased in adult stages (Figure 2B). Both δ¹³C_n and δ¹⁵N_n values were higher in the pre-birth stage and decreased around birth (δ¹⁵N_n values to a lesser extent; Figures 2A,B). Pre-birth δ¹³C_n values were highest across life-history and comparable to adult levels, whereas pre-birth δ¹⁵N_n values were higher than late adult levels. Among-individual variances in both δ¹³C_n and δ¹⁵N_n values were relatively high but constant across life-history and lower than in porbeagles (Figures 2A,B).

In porbeagles, mean δ¹³C_n values increased in the juvenile stage (until an age of ∼7–8 years), then remained relatively constant until capture (Figure 2C). Pre-birth δ¹³C_n values were higher than post-partum values but generally lower than adult levels. Among-individual variance in δ¹³C_n values was markedly higher than in blue sharks, particularly in the pre-birth stage and early juvenile and subadult stages associated with shifts in mean δ¹⁵N_n values (see below). Mean values of δ¹⁵N_n showed two stepwise increases, a small one just after birth and a larger one in the subadult stage (at ∼11–12 years of age), whereas values in intermediate juvenile and adult periods remained relatively constant (Figure 2D). The stepwise pattern in the subadult stage was driven by a step increase in δ¹⁵N_n values in individuals from the Northwest Atlantic, and the plateau phase in the adult stage was also only seen in these sharks (Supplementary Figures 4M–O). Similar to δ¹³C_n values, pre-birth δ¹⁵N_n values were higher than values after birth but lower than values in the adult stage (Figures 2C,D). Among-individual variance in δ¹⁵N_n values was broadly consistent with that in δ¹³C_n values, peaking in the pre-birth stage and early and late juvenile stages.

Carbon isotope compositions of individual amino acids were analyzed for 128 blue shark vertebral samples (n = 196 bulk collagen samples). For porbeagles, 80 samples were analyzed for carbon isotopes in amino acids (n = 133 bulk protein samples). Mean ± SD δ¹³C values for each amino acid in blue shark and porbeagle samples are presented in Supplementary Table 4. Individual isotope chronologies for δ¹³C values in single amino are presented in Supplementary Figures 5–13.

Values of δ¹³C of non-essential amino acids (δ¹³C_non–EAA) were more positive than values of essential amino acids (Supplementary Figures 5–13), and the bulk protein δ¹³C value corresponded to the weighted average of δ¹³C_EAA and δ¹³C_non–EAA values. Co-variations between bulk protein and single amino acid δ¹³C values are shown in Figures 3, 4 for essential and non-essential amino acids, respectively. Bulk protein δ¹³C values were higher and spanned a smaller range in porbeagles. Individual blue sharks grouped by capture area, implying that spatially determined among-individual variation explained more of the variance than the within-individual term. Within individual sharks, amino acid δ¹³C values co-varied more with bulk protein δ¹³C values in blue sharks compared to porbeagles, implying a common mechanism influencing multiple amino acids in the same direction and therefore contributing to bulk collagen δ¹³C values. This was especially clear for the non-essential glycolytic amino acids alanine and glycine. In porbeagles, there was very weak co-variance between bulk protein δ¹³C values and δ¹³C values for any single amino acids, implying that mechanisms underpinning variation in bulk protein δ¹³C values varied among- and within-individuals, likely complicating efforts to interpret bulk collagen δ¹³C values in an ecological or behavioral context. Aspartate/aspartic acid and, to a lesser extent, glutamate/glutamic acid δ¹³C values separated blue shark and porbeagle individuals into two clusters but these clusters did not correspond to different capture areas.

Bulk protein δ¹³C and δ¹⁵N values in blue sharks, normalized by life-history, showed a common pattern of increasing values during juvenile growth (Figures 2A,B), despite relatively high variation in individual profiles (Supplementary Figures 4A–I). Increases in δ¹³C values in marine animals are commonly interpreted as baseline effects resulting from increased use of warmer waters closer to the equator (McMahon et al., 2013; Magozzi et al., 2017) and/or more coastal foodwebs with higher levels of primary production (O’Learly, 1981; Miller et al., 2008). Increases in δ¹³C values could also reflect increases in trophic position, however the lack of a corresponding trend in δ¹⁵N values in blue sharks implies that trophic level-change alone could not explain the δ¹³C pattern. In the eastern North Atlantic in particular, blue sharks are believed to be primarily oceanic sharks (Vandeperre et al., 2014, 2016; Coelho et al., 2018), therefore ontogenetic trends in bulk protein δ¹³C values in blue sharks are likely to be interpreted as reflecting a life-history migration resulting in increased use of resources from relatively warm southern locations through the juvenile and early adult life stages. In the Northwest Atlantic, juveniles are often observed near the shelf, particularly in summer, meaning that isotopic signals in the juvenile stages could as well be influenced by coastal processed. The increase in δ¹⁵N values during juvenile growth and subsequent decrease in the adult stage (Figure 2B) is also consistent with baseline effects associated with an increasing dependence on resources living in warmer waters (McMahon et al., 2013; Schmittner and Somes, 2016) rather than a strong ontogenetic trophic shift.

Ontogenetic profiles in bulk protein δ¹³C and δ¹⁵N values in porbeagles are considerably more challenging to interpret compared to blue sharks. Life history-normalized bulk protein δ¹³C values displayed a minor ontogenetic increase with approximately half of the range seen in blue sharks, and high among-individual variance in both early and late juvenile life stages (Figure 2C). Bulk protein δ¹⁵N values were relatively flat until the late juvenile stage, when a pronounced step increase was observed (Figure 2D). Similar to blue sharks, the ontogenetic profiles in bulk protein δ¹³C values in porbeagles may imply a gradual increase in reliance on resources with higher δ¹³C values (e.g., closer to the equator and/or more coastal resources) throughout juvenile growth but with a high degree of among-individual variance. The step increase in δ¹⁵N values in late juvenile stages may potentially be linked to a change in diet or trophic level as inferred for other lamnid sharks (Estrada et al., 2006; Kerr et al., 2006), as there is no clear association with δ¹³C values that might imply a habitat shift. Such a step increase in δ¹⁵N values was, however, only observed in individuals from the Northwest Atlantic (Supplementary Figures 4M–O), possibly suggesting a delay in trophic shift in the Northeast Atlantic population, movement across larger nitrogen isoscapes in the eastern part of the basin, and/or differences in age estimation models not accounted for here.

In summary, bulk protein profiles in blue sharks may be interpreted as suggesting common ontogenetic changes in space use, toward greater use of oceanic and warmer waters with increasing age. For porbeagle sharks, a lack of clear, consistent, isotopic patterns, and a relatively small increase in δ¹³C values, would likely imply no strong, common ontogenetic trend in space use, and potentially location- and population-specific variations in life-history movements.

Essential amino acids can only be synthesized by primary producers and bacteria, and must be acquired by consumers directly through the diet. Variations in bulk tissue isotope values reflecting horizontal movements should therefore be replicated consistently in the δ¹³C values of essential amino acids, without confounding influences from trophic discrimination and physiology (Reeds, 2000; McMahon et al., 2010).

In blue sharks, at least a component of the positive ontogenetic trend in bulk protein δ¹³C values was matched by a common ontogenetic pattern of increasing δ¹³C values of essential amino acids (Figure 5A). Inferences drawn from bulk protein δ¹³C values are therefore supported by patterns in δ¹³C values of essential amino acids. That is, the ontogenetic increase in bulk protein δ¹³C values observed in these sharks during juvenile and early adult growth can be explained by progressively increased dependence on resources from waters characterized by more positive δ¹³C baselines (e.g., warmer waters). The interpretation of a life-history migration to southern waters is consistent with results from satellite-tagging studies in the eastern North Atlantic, which indicate that juvenile and subadult blue sharks conduct seasonal movements between temperate and subtropical waters (Queiroz et al., 2010; Vandeperre et al., 2014; see also Campana et al., 2011), whereas adult sharks undertake a directional migration to tropical waters possibly associated with parturition and pupping (Vandeperre et al., 2014, 2016; Coelho et al., 2018).

Maternal δ¹³C values of essential amino acids were higher than post-partum values and comparable to adult levels (Figure 5A) as seen in bulk protein δ¹³C values, indicating that pupping might occur in isotopically distinct areas with more negative δ¹³C baselines than those used by the mothers while provisioning eggs and by adult sharks. Adult δ¹³C values of essential amino acids approached maternal levels by ∼5–6 years of age, which approximately corresponds to the onset of maturity in blue sharks (Pratt, 1979; Skomal and Natanson, 2003).

In stark contrast to observations in blue sharks and bulk collagen values, δ¹³C values of essential amino acids in porbeagles, expressed as life history-normalized PC1_EAA scores, decreased during juvenile growth until an age of ∼6–7 years (Figure 5C). Assuming that essential amino acids reflect baseline conditions, this trend implies a common trend of increasing assimilation of carbon from prey with low δ¹³C values (e.g., resources from offshore and/or colder waters), as the sharks grow. The interpretation of porbeagle isotopic life-histories based on bulk protein and essential amino acid δ¹³C values are therefore difficult to reconcile. Porbeagles are believed to conduct an ontogenetic habitat shift from shelf to shelf-edge and oceanic environments (Bendall et al., 2013; Ellis and Bendall, 2015; Ellis et al., 2015; Biais et al., 2017), and have been shown to conduct extensive latitudinal migrations (Pade et al., 2009; Saunders et al., 2011; Biais et al., 2017). Contrasting patterns in bulk protein and essential amino acid δ¹³C values could be possibly reconciled by an increasing trophic level, if trophic level increased sufficiently to raise δ¹³C values to counteract the decrease in δ¹³C values of essential amino acids. However, this was not supported by the trend in bulk collagen δ¹⁵N values, which would be expected to be more sensitive to changes in trophic level. Satellite-tag data are therefore consistent with the interpretations based on patterns in essential amino acid δ¹³C values (Figure 5C), not with those based on bulk protein δ¹³C values.

Pre-partum δ¹³C values reflected carbon assimilation during maternal provision and represented the highest values that were found across life-history (Figure 5C). This pattern suggests that pregnant female porbeagles may exploit isotopically distinct (more positive) foraging grounds compared to those used by subadult and adult sharks and is consistent with the observation of a directional movement to subtropical waters by tagged mature (presumably pregnant) females in the Northwest Atlantic (Campana et al., 2010; see also Natanson et al., 2019).

Ontogenetic profiles in PC1 scores for δ¹³C values of non-essential amino acids in blue sharks also showed a common increase (Figure 5B), consistent with a baseline effect expressed in both essential and non-essential amino acids. As all non-essential amino acids are influenced by baseline effects, the isotopic spacing in δ¹³C values between glycolytic and Krebs cycle amino acids to some extent controls for the baseline influence (assuming that the rate of incorporation of newly ingested carbon is consistent between the two amino acids). For blue sharks, observed ontogenetic patterns in life history-normalized isotopic spacing varied among individuals resulting in no clear common trend (Figures 6A,B and Supplementary Figures 16A–I). This pattern also suggests that the common signal seen in bulk protein δ¹³C values in blue sharks is largely controlled by baseline spatial influences rather than dietary and/or physiological effects.

Ontogenetic trends in life history-normalized PC1 scores for δ¹³C values of non-essential amino acids in porbeagles are unclear (Figure 6D) but ontogenetic profiles of normalized spacing in δ¹³C values between glycolytic and Krebs cycle amino acids show consistent, systematic trends with a strong increase after birth, followed by U-shaped patterns throughout juvenile growth (Figures 6C,D and Supplementary Figures 16J–O). The biochemical or nutritional mechanisms underpinning differences in isotopic spacing between non-essential amino acids remain uncertain. However, the presence of systematic trends in isotopic spacing between amino acids assumed to be responsive to differences in macronutrient (lipid and protein) assimilation and use (Wolf et al., 2015; Wang et al., 2018, 2019) implies that amino acid isotope data hold information associated with individual nutrition, and that bulk protein δ¹³C values may be very difficult to interpret reliably.

Therefore, we suggest that in porbeagles relatively minor isotopic variability associated with ontogenetic transitions between juvenile foraging in more coastal, shelf waters and greater dependence on cooler, offshore and slope environments in adult stages are masked in bulk protein δ¹³C values by individual variation and by physiological influences potentially associated with differential use of lipid and protein macronutrients. In contrast, common ontogenetic movement transitions across latitudinal gradients in δ¹³C values in blue sharks are large enough to be clearly expressed in bulk protein δ¹³C values and overwhelm any shared ontogenetic trends in macronutrient use.