Compound-specific stable hydrogen isotope ( δ 2 H) analyses of fatty acids: A new method and perspectives for trophic and movement ecology

Rationale: Compound-specific stable isotope analysis (CSIA) is a powerful tool for a better understanding of trophic transfer of dietary molecules in and across ecosystems. Hydrogen isotope values ( δ 2 H) in consumer tissues have potential to more clearly distinguish dietary sources than 13 C or 15 N values within and among habitats, but have not been used at the fatty acid level for ecological purposes. Methods: Here we demonstrate a new online high-capacity gas chromatography – isotope ratio mass spectrometry technique ( 2 H-CSIA) that offers accurate and reproducible determination of δ 2 H values for a range of fatty acids from organisms of aquatic food webs. Results: We show that lipid extracts obtained from aquatic sources, such as biofilms, leaves, invertebrates, or fish muscle tissue, have distinctive δ 2 H values that can be used to assess sources and trophic interactions, as well as dietary allocation and origin of fatty acids within consumer tissue. Conclusions: The new 2 H-CSIA method can be applied to evaluate sources and trophic dynamics of fatty acids in organisms ranging from food web ecology to migratory connectivity.


| INTRODUCTION
Tracing trophic transfer of dietary energy sources and molecules in and among aquatic and terrestrial food webs is critical for understanding and quantifying trophic interactions of species. Stable isotope analyses (e.g. δ 13 C, δ 15 N, δ 34 S) of primary producers and consumer tissues (e.g. algae, leaves, feathers, muscle tissue, etc.) have been used since the 1980s to construct trophic models for terrestrial, freshwater, and marine ecosystems. [1][2][3][4][5] Bulk tissue δ 2 H or bulk lipid δ 2 H analyses have been applied to track animal migration patterns and provenance of lipid synthesis [6][7][8] and offer low-cost routine approaches, but suffer from several intrinsic challenges: (i) bulk tissue isotopic averaging of all molecular components, such as carbohydrates, proteins, and lipids, all of which undergoing anabolic and catabolic processes at different turnover times, thus distorting the bulk δ value depending on the current physiological state of the biological sample; (ii) isotopic fractionation by biochemical processes; (iii) bulk stable isotopic overlap of available food sources; and (iv) specific to any H isotope exchange with environmental water. [9][10][11] In the past decade, technological advances in compound-specific isotope analysis (CSIA) of specific molecules extracted from tissues (e.g. essential or nonessential amino acids) have offered a deeper view of trophic transfer, and trophic ecology in general, but these discrete molecular (or compound-specific) stable isotope assays require specialized preparation and complex gas or liquid chromatographic interfaces to combustion-based isotope ratio instruments, and, so far, are largely limited to 13 C or 15 N. 12,13 More recently, the use of hydrogen isotopes (δ 2 H) has provided a promising approach owing to large and predictable H isotope patterns between aquatic and terrestrial sources, season, i.e. differences in precipitation amounts and ambient temperature, as well as the significant bulk tissue H isotopic differences commonly observed between terrestrial and aquatic primary producers. 10,14,15 The extension of H isotopes towards compound-specific assays of fatty acids has long been seen as encouraging, as has been shown for amino acids. 16,17 Furthermore, a recent study comparing bulk plasma fatty acid composition with feather bulk δ 2 H could infer relative inputs to swallows from terrestrial and aquatic insects. 18 Still, the use of 2 H-CSIA for lipids in trophic ecology is largely unexplored and mainly focused on plant lipid and wax extract studies, 17,[19][20][21] or paleobiology and geosciences. 22,23 Fatty acids are quantifiable tracers used to assess the trophic transfer of specific dietary energy sources, such as bacterial-, algal-, or terrestrial-derived fatty acids. [24][25][26] However, it is problematic to differentiate amongst these different dietary sources of fatty acids; for example, the essential polyunsaturated fatty acids (PUFA) α-linolenic (ALA) and linoleic (LIN) acids are synthesized in terrestrial and also aquatic primary producers, 27 and hence cannot be used as clear dietary source-specific markers. The application of δ 2 H is exceptionally promising compared to δ 13 C because large bulk H isotopic differences have already been seen between individual diet sources and consumers both spatially and temporally. 8,[28][29][30] In most lipids, the C-H bonds of the alkyl chains are chemically stable and nonexchangeable with ambient water, hence preserving the original H isotope source signal, but may occur over much longer geological timescales, 31 in contrast to the rapid H isotope exchange with water, for example, in carbohydrates. Thus, δ 2 H values of nonessential components will reflect (with biological H isotope fractionation) the deuterium abundance of local environmental water, while δ 2 H values in essential components should resemble those of the consumer's diet, as has been shown for amino acids. 16 34 For example, the enzymatic processes for conversion of precursors to long-chain PUFA involve elongation and desaturation, and possibly to some extent retro-conversion back to short-chain PUFA, which seems likely to be a process that alters the δ 2 H values inessential fatty acids. In addition, source water and the seasonal variations in 2 H content of water sources within the sampling area need to be considered to establish baseline H isotope conditions. A major technological and analytical challenge in achieving δ 2 H assays of individual lipids in environmental samples is the exceedingly low natural abundance of 2 H in combination with low contents of the lipids of interest for most samples. This requires improved detection limits and the need for highly specialized chemical and isotopic analyses that inherently have larger uncertainties than are commonly expected for routine bulk sample deuterium measurements (e.g. less than ±2-4 ‰) due to greater analytical complexity.
Previous attempts to use δ 2 H in an ecological context failed mainly due to the low H 2 signal intensity of the mass spectrometer, which led to unreliable or highly uncertain results that were difficult or ambiguous to interpret. The aim of the study reported here was twofold. First, we present a new analytical method and optimized isotope ratio mass spectrometry (IRMS) workflow that utilizes high-capacity gas chromatography (HC-GC) to isolate essential and nonessential lipids from environmental samples, along with online thermochemical conversion to H 2 gas, and introduce the setup and provide the parameters for reliable and reproducible determinations of

| Environmental samples
A suite of natural lipid samples extracted from leaves, stream periphyton, macroinvertebrates, and fish that were previously sampled, preserved, and stored 35 was used to optimize the analysis and evaluate the potential of compound-specific δ 2 H analysis of fatty acids. Briefly, the environmental samples were collected from the subalpine Ybbs River catchment near Lunz am See, Austria (47 45 0 N, 15 12 0 E), between July and October 2016, were subsequently freeze-dried and stored (Virtis Genesis freeze dryer). All samples were weighed and homogenized using a glass rod or food processor (leaves). Lipids were extracted from the samples as described elsewhere, 36 using a chloroform-methanol mix (2:1 v/v) and sonication. Samples were vortexed and centrifuged three times to remove nonlipid phases. The extracted lipids were evaporated to a final volume (1.5 mL) using a stream of N 2 gas. For fatty acid methyl ester (FAME) formation suitable for GC, the samples were incubated with a sulfuric acid-methanol mix (1:100 v/v) for 16 h at 50 C followed by the addition of an equal normality of KHCO 3 and hexane.
Samples were manually shaken, vortexed, and then centrifuged. The supernatant organic FAME layers were collected, pooled, and again concentrated under N 2 gas to 30 μL to be used for GC characterization and H isotope analysis. Samples were stored under inert N 2 atmosphere at À80 C. Technologies) was used. For both columns, the GC injector port was held at 250 C. In spitless mode, 3 μL of each sample was injected which was previously determined to be the maximal volume without backflash using a split/splitless liner with single taper (4 mm Â 6.3 mm Â 78.5 mm, 453A1355, Thermo Scientific) following activation of the purge flow after 1 min. For the thinner 60 m column, the GC temperature program started at 80 C for 2 min and ramped up by 30 C min À1 to 175 C, then by 5 C min À1 to 200 C, and finally by 2.4 C min À1 to 250 C which was maintained for 30 min for a total GC run time of 62 min. The temperature program for the shorter high-capacity 30 m GC column used for isotope analyses started at 80 C for 2 min, was ramped up by 30 C min À1 to 175 C, and then by 5 C min À1 to the 240 C maximal temperature, and held for 35 min.

| FAME analysis using GC
The total run time for the 30 m HC-GC column used for H isotope analysis was 52 min.

| δ 2 H measurements
For the δ 2 H isotope analyses, the GC-resolved FAME were passed through a high-temperature (1400 C) thermochemical carbonconditioned open ceramic reduction reactor (hydrothermal carbonization; HTC) which quantitatively pyrolyzed the FAME molecules to pure H 2 gas, with the sample carbon reduced and retained as graphite in the reactor. The HTC reactor was preconditioned using 2 μL hexane injections before each GC sample sequence of 40-60 samples, followed by reverse flushing with He for 1 h. Each sample batch session was preceded with triplicates of USGS70 and USGS71 standards for data normalization purposes and performance testing. After every 10 unknown sample runs, these high and low δ 2 H standards were repeated to ensure system stability and to check for instrumental drift.
The He flow rate through the HTC reactor was kept constant at 1.1 mL min À1 to ensure stable and quantitative conversions to H 2 .
Ionization energy of the IRMS source was optimized to obtain the highest total H 2 + ion current yield without excessive formation of He 2+ ions. An H 3 + -factor determination was performed before and after each measurement batch sequence using a dilution series of reference gas and was found to be low and stable (3.0 ± 0.1 ppm/V). The starting temperature of 40 C was held for 1 min followed by an increase to 150 C at a rate of 30 C min À1 . Each vial was measured eight times and δ 2 H values of methanol were found to be consistent at À146.8 ‰ ± 2.6 (VSMOW). Other methods such as injection of different concentrations of methanol in toluene or variation of split ratio provided nearly identical average δ 2 H values, but with significantly higher standard deviations (data not shown). Thus, it can be assumed that no H isotopic fraction for methanol was induced. It was reported that the effect of H isotopic fractionation induced by the process of fatty acid methylation is usually below the detection limit of IRMS systems. 37 For the introduced methyl group, we performed "methanol corrections" to avoid biases due to potentially large isotopic differences between the δ 2 H value of the methanol and those of fatty acids in the sample: where H n is the number of H atoms for each fatty acid, δ 2 H FAME the observed value for the FAME, δ 2 H FA the desired value for the target fatty acid and δ 2 H Me the previously determined value of the methanol used for transmethylation. All values reported here refer to the δ 2 H FA values.

| Data analysis
The δ 2 H values of individual FAME compounds were determined were also tested. 38 In our FAME samples, peaks were only considered acceptable for isotope data analysis if the H 2 amplitude exceeded 450 mV and the peak area was more than 4 V s. All H 2 peaks were validated manually and start-and-end-point corrected, as necessary.
While manual background correction can be effective, it is subject to individual operator preferences and requires experience. For this study, no manual background corrections were performed. The HC-GC setup used in this study moreover did not allow for a clear separation between C18:1n-7 and C18:1n-9, and therefore both peaks were combined as one area thereby providing a general weighted-average δ 2 H value for C18:1 isomers. 38

| Reference materials and normalization
All H isotope analyses were calibrated against the FAME-C20  Figure 1A). We observed that the methyl group significantly shifted F I G U R E 1 Assessment of fatty acid δ 2 H bias from preparative methylation. (A) Theoretical shifts between the derivatized FAME and original fatty acid of interest as a function of number of H atoms of the acyl chain as dependent on the isotope difference between methanol used for derivatization and the fatty acid in the sample. (B) Observed differences between FAME and fatty acid after correction for methyl group for data. Significant biases in δ 2 H, as for LIN and ARA or ALA and EPA/DHA, may lead to data misinterpretation and reduce comparability between different laboratories the observed δ 2 H value of FAME ( Figure 1B). The observed differences between the laboratory-derivatized FAME and original fatty acid of interest were eliminated after a correction for the methyl group. Uncorrected differences in δ 2 H, such as for LIN and arachidonic acid (ARA) or ALA and eicosapentaenoic acid (EPA)/ docosahexaenoic acid (DHA), might lead to misinterpretation and reduce the comparability between different laboratories and even batches of methanol used for analysis; hence we recommend that a methyl bias correction is routinely applied.
To ensure robust H isotope analyses for more complex mixtures of FAME, a fish liver sample was measured four times sequentially  Figure S1). Overall, more 2 H peaks met the acceptable criteria for δ 2 H determinations (>450 mV; >4 V s) using the 60 m GC column (15 versus 14); however, only two peaks were identified in the periphyton sample and the overall reproducibility for δ 2 H was poorer than when using the 30 m column, which had an average standard deviation of 2.8‰ versus 2.0‰ (Table 1).
One disadvantage of the standard thin 60 m column was a higher chance of peaks overlapping for some sample types due to the broad nature of FAME peaks. By contrast, H 2 sample peak intensity increased by a factor of ca 2 by when using the 30 m HC-GC column, thereby improving both the accuracy and precision of the δ 2 H measurements. There was only a slight loss in chromatographic performance with the 30 m GC column, but for most of the FAME peaks of ecological interest a return of H 2 signal to baseline levels between two analytes could be achieved.

| Adjustment of environmental samples for δ 2 H analysis
The abundance of fatty acid components in nature varies depending on organism, tissue or sample type, time, site, and other complex factors. We intended to develop a generalized robust method to conduct lipid δ 2 H assays for a broad range of sample types, needed for studies in (aquatic) ecology, and to determine the required amount of sample for successful use in HC-GC/ 2 H-IRMS analysis. For this process optimization we used a small subset of the archived samples ( Figure S2). T A B L E 1 Differences in analytical performance between standard and highcapacity GC columns. A fish liver and one biofilm sample measured five times on both columns as described in the methods section. All baseline-separated peaks >450 mV and >4 V s were used for analysis.

| Aquatic invertebrates
Aquatic invertebrate samples (n = 19; including Leuctra sp.,    Table 2).   to terrestrial leaves contribute more to overall bulk tissue deuterium depletion. 29 As noted, the latter is caused by multiple steps during lipid synthesis, 33 and thus, when using δ 2 H, detailed knowledge of the physiological processes is required. 34 33 Therefore, additional isotope research on consumer physiology for species of interest is needed, especially in combination with δ 2 H and δ 13 C-CSIA, to significantly increase the isotopic resolution for food web analyses in the future.

| CONCLUSIONS
This study provides a new reproducible protocol for the determination of δ 2 H-CSIA in biological lipid samples, and we show promising ecological applications such as resolving resource utilization in aquatic and terrestrial food webs (endogenous versus exogenous energy sources). As bulk δ 2 H has been used for migration studies in the past, we also suggest δ 2 H of fatty acids as a potential sensitive site-specific marker. Finally, our method is applicable to a wide range of other ecological applications, such as fatty acid metabolism and routing in and within consumer tissues, food authenticity, and overall consumer response to diet change.

ACKNOWLEDGMENTS
The authors would like to thank the two anonymous reviewers for their valuable input, which significantly improved the manuscript. This work has been supported by the Austrian Science Fund (FWF project "AlphaOmega" P-28902-B25).

PEER REVIEW
The peer review history for this article is available at https://publons. com/publon/10.1002/rcm.9135.