The microbial biomass measured here was generated in a prior study targeting lipid δ²H analysis (Wijker et al., 2019); all relevant culturing details are recapitulated here. Five wildtype aerobic heterotrophic microbes (Escherichia coli MG1655, Bacillus subtilis PY79, Ensifer meliloti Young 2003, Pseudomonas fluorescens 2-79, and Rhizobium radiobacter C58) and five mutant E. coli organisms carrying specific deletions of dehydrogenase or transhydrogenase genes (glucose 6-phosphate dehydrogenase deleted in JW1841, phosphoglucose isomerase deleted in JW3985, membrane-bound transhydrogenase deleted in PntAB, soluble transhydrogenase deleted in UdhA, and both transhydrogenases deleted in UdhA-PntAB) were cultured on unlabeled glucose for hydrogen isotope analysis, and on ¹³C-labeled glucose (100% 1-¹³C-glucose and a mixture of 20% (wt/wt) U-¹³C₆-glucose + 80% (wt/wt) unlabeled glucose) for metabolic flux analysis. The relative metabolic fluxes were calculated based on the ¹³C-labeling pattern of proteinogenic amino acids—see Wijker et al. (2019) for more details. Wildtype organisms were additionally cultured on acetate, citrate, fructose, pyruvate, and/or succinate in an isotopically constant growth water; as well as on glucose in growth waters with different isotopic compositions, which were manipulated by adding specific volumes of 99.9% purity D₂O to distilled, deionized water. Each organism was grown with 4 g/L of carbon source in M9 minimal medium (prepared as described in Fuhrer et al., 2005) in batch culture on a rotary shaker at 200 rpm, thereby ensuring aerobic conditions were maintained and fermentation was avoided throughout the course of the experiments. The carbon source served as the limiting nutrient in each culture, causing cells to transition to stationary growth phase upon depletion. B. subtilis and R. radiobacter cultures were supplemented with a vitamin mixture, while strains JW1841 and JW3985 were given 50 μg/mL of kanamycin. B. subtilis and E. coli cultures were incubated at 37°C; all other organisms were incubated at 30°C. All wildtype cultures were prepared in duplicate except for organisms grown in D₂O-spiked media (for growth water experiments) and for E. coli, which was grown on pyruvate and acetate in single cultures, and on glucose in two non-replicate cultures: culture #1 was grown along with the rest of the wildtype organisms, E. coli mutants, and growth water experiments for non-E. coli organisms; culture #2 was grown at a later date as one of four cultures in E. coli growth water experiments. Although culturing conditions were identical between E. coli cultures #1 and #2 grown on glucose, the different timing of culturing, and different methods used to process the biomass (see Section 2.2), renders culture #2 a repeat experiment, but not true biological replicate, to culture #1. Culture growth was monitored by measuring optical density at 600 nm (OD₆₀₀), and cells were harvested in late-exponential phase, lyophilized, and stored at -80°C until further processing for lipid and amino acid δ²H analysis (Wijker et al., 2019 and this study, respectively).

A 10–20 mg of dry biomass from each sample was hydrolyzed anoxically in 6N HCl at 110°C for 24 h in tightly capped VOA vials. Following hydrolysis, samples were uncapped and left on the hot plates until the 6N HCl was completely evaporated, then samples were resuspended in 0.5 ml of 0.1N HCl. Amino acids in all samples except E. coli culture #2 were derivatized with 7:6:3 (v/v/v) anhydrous methanol (MeOH), pyridine, and methyl chloroformate (MCF); reagents were added at room temperature, then samples were immediately capped and sonicated for ~5 min (procedure adapted from Husek, 1991a,b and Zampolli et al., 2007). E. coli culture #2 was derivatized with the same reagent bottles and reaction procedure as the other samples, but was placed on dry ice while derivative reagents were added to slow the derivatization reaction. Note that in contrast to most published δ²H_AA studies (Fogel et al., 2016; Newsome et al., 2020; Smith et al., 2022; Mancuso et al., 2023), we avoid using fluorinated derivative reagents, as hydrofluoric acid can form during pyrolysis in the gas chromatograph/isotope ratio mass spectrometer (GC/IRMS), leading to potential hydrogen isotope fractionation (Sauer et al., 2001; Renpenning et al., 2017; Silverman et al., 2022).

The resulting methoxycarbonyl (MOC) esters (Figure 2A) were extracted twice with methyl tert-butyl ether (MTBE) and filtered through a sodium sulfate column to adsorb any water present. Samples were concentrated to ~0.25–0.5 ml under N₂. MOC ester peaks were identified via gas chromatography/mass spectrometry (GC/MS) on a Thermo-Scientific Trace ISQ equipped with a Zebron ZB-5 ms column (30-m × 0.25-mm i.d., 0.25 μm film thickness) and programmable temperature vaporizing (PTV) injector operated in splitless mode, using He as a carrier gas (flow rate = 1.4 ml/min). The GC oven was held at 80°C for 1 min, ramped at 5°C/min to 280°C with no hold, then ramped at 20°C/min to 310°C with a final 5 min temperature hold. Peaks were identified by comparing the relative retention times and mass spectra to those of known MOC ester standards, as well as to mass spectra in the NIST MS Library database.

The δ²H values of MOC esters were measured by a gas chromatograph coupled to an isotope ratio mass spectrometer (Thermo Finnigan Delta⁺XP) using a pyrolysis interface (i.e., GC/P/IRMS). Chromatographic separation was achieved on a thick-film Zebron ZB-5ms column (30-m × 0.25-mm i.d., 1.00 μm film thickness) with a nearly identical chromatographic method as used in GC/MS analysis [exceptions included a higher carrier gas flow rate (1.7 ml/min) and slight modifications to the temperature program to optimize MOC ester separation] so peaks could be identified by retention order and relative height. Measured isotope ratios were calibrated using hydrogen gas of known isotopic composition and are reported in δ notation (in units of ‰, or parts per thousand; Urey, 1948; McKinney et al., 1950) relative to the Vienna Standard Mean Ocean Water (VSMOW) international standard (δ²H = R_AA/R_VSMOW − 1), where R = ²H/¹H. Additionally, an eight-compound fatty acid methyl ester standard mixture was analyzed between every 5–6 samples to verify instrument accuracy and precision. Samples were analyzed in triplicate, and the MOC ester δ²H values were corrected for the addition of methyl hydrogen from the derivative reagents, as well as for the remaining exchangeable amine hydrogen (see Section 2.5). The standard deviation of triplicate analyses for individual amino acids was typically ≤6‰. The average root-mean-square error of the external FAME standard was 3.2‰ across all analyses. δ²H values of the culture media (δ²H_w) were measured previously using a Los Gatos Research DLT-100 liquid water isotope analyzer and calibrated against up to four working standards with δ²H values ranging from –73 to +458‰ (Wijker et al., 2019). Data are reported as apparent fractionations between amino acids (AA) and culture medium water (w) according to the equation  2εAA/w = (δ²H_AA + 1)/(δ²H_w + 1) − 1, with uncertainty propagated as Equation 1

Potential changes in amino acid isotopic compositions during acid hydrolysis were investigated by varying the hydrolysis conditions used (temperature, duration, and O₂ presence). For a control treatment, standard bovine serum albumin (BSA) was hydrolyzed in 6N HCl for 24 h at 110°C under anoxic conditions (achieved by sparging samples with N₂ for 2 min with vigorous shaking). Variations on these conditions were achieved by either hydrolyzing BSA (1) without sparging with N₂ (oxic hydrolysis), (2) at 105°C, or (3) for 20 or 48 h. All conditions were prepared in duplicate. Amino acids were derivatized to MOC esters, extracted, and analyzed via GC/P/IRMS using methods described in Sections 2.2, 2.3.

Additionally, to test for hydrogen isotope exchange with aqueous medium during hydrolysis and derivatization (Hill and Leach, 1964; Fogel et al., 2016; Silverman et al., 2022), BSA and a mixture of pure amino acid standards were separately hydrolyzed (6N HCl, 24 h, 110°C, oxic) and derivatized (0.1N HCl, 7:2:3 v/v/v anhydrous MeOH, pyridine, MCF) in aqueous solvent with different hydrogen isotope compositions.

To determine the δ²H value of MeOH, 100 μg of disodium phthalate with known isotopic composition (Sessions et al., 2002) was derivatized in 2 ml of 20:1 (v/v) anhydrous MeOH:acetyl chloride (70°C, 30 min; Figure 2B). The derivatized product was extracted with 4 ml of 1:1 water:hexane, then measured by GC/P/IRMS, and contribution of disodium phthalate hydrogen was subtracted by mass balance.

The combined δ²H value of MCF and the remaining exchangeable amine hydrogen was characterized via derivatization of p-phenylenediamine (PPD)—a compound with two primary amine groups—with MCF and separate oxidation of PPD to dinitrobenzene (DNB), which has no nitrogen-bound hydrogen. One hundred mM of PPD was dissolved in 5 ml of anoxic dichloromethane with 3 eq. triethylamine (pre-distilled with CaH₂ to remove any HCl generated in the reaction) and 0.1 eq. 4-dimethylamino pyridine, and was subsequently derivatized via an overnight reaction with 2.5 eq. MCF to yield dimethyl 1,4-phenylenedicarbamate (hereafter, “dicarbamate product”, or DCP in Equation 2; Figure 2C). The DCP was purified via flash column chromatography with silica gel. Five mg of DCP was dissolved in 2 ml of anhydrous MeOH, then 2 ml of distilled, deionized water was slowly added. The solution was mixed on a shaker for 2 h to ensure complete equilibration of the two amine hydrogen atoms with water, then the DCP was extracted once with 4 ml MTBE, filtered through a sodium sulfate column, and analyzed by GC/P/IRMS. In a separate reaction, PPD was oxidized to DNB using 8 eq. of m-chloroperbenzoic acid added under refluxing 1,2-dichloroethane in a procedure adapted from Liu et al. (2014) (Figure 2C). DNB was purified via flash column chromatography with silica gel, then 3 mg of DNB was dissolved in MTBE and analyzed via GC/P/IRMS. The δ²H value of MCF and the exchangeable nitrogen-bound hydrogen (MCF+NH) was obtained by solving for F_MCF+NH in the mass balance equation

where F is the fractional abundance (i.e., mole fraction) of ²H in each compound. PPD was chosen for these reactions (1) because of the favorable 8/4 ratio of (MCF + amine)/aromatic hydrogen in the dicarbamate product, and (2) because the exchangeable amine hydrogen atoms on the dicarbamate product and on MOC esters should have similar hydrogen isotope compositions when equilibrated in the same water supply, as the amine hydrogen in the dicarbamate product and in MOC esters share similar intramolecular bonding environments so should be controlled by similar equilibrium ²H/¹H fractionation factors at a constant temperature (Wang et al., 2009).

Hydrogen atoms on amine and carboxyl groups rapidly exchange with water so do not contribute information about native δ²H_AA values. Derivatization of the carboxyl group removes the exchangeable hydrogen, but derivatization of the amine group removes only one of the two exchangeable hydrogen atoms (Figure 2A). In order to determine the isotope compositions of the native, non-exchangeable (i.e., carbon-bound) hydrogen on the amino acids, it is necessary to correct δ²H_AA values not only for the added derivative (MeOH and MCF) hydrogen, but also for the exchangeable amine hydrogen atom remaining after derivatization. A suitable method for this latter correction has eluded prior studies of δ²H_AA thus far, yet is imperative, as errors in reported δ²H_AA values can be on the order of 10 to 100‰ when the amine-bound hydrogen is improperly accounted for (Supplementary Figure S7; Supplementary Section 4). Here we developed a method to characterize the combined δ²H values of MCF and the exchangeable amine hydrogen by derivatizing PPD with MCF, separately oxidizing PPD to DNB (which removes all four of the nitrogen-bound hydrogens; Figure 2C), and analyzing both resulting products using GC/P/IRMS. Importantly, this approach requires derivatizing samples with the same reagents and water as those used to measure PPD.

The δ²H value of the MCF + amine hydrogen was –111.43 ± 1.96‰ when equilibrated with water of δ²H = –86.50 ± 0.36‰. The δ²H value of MeOH was –67.27 ± 1.78‰. Correction for these non-biological contributions generally shifted δ²H_AA values lower (Figure 3). For the relatively ²H-depleted amino acids (leucine, valine, and isoleucine), changes in δ²H values were substantial, reaching up to 123‰ for wildtype organisms grown on glucose. For the relatively ²H-enriched amino acids (proline and phenylalanine), this correction resulted in small or negligible changes, but in some cases resulted in higher δ²H values. However, note that correction of proline δ²H values using this approach may introduce small (<20‰) errors, as proline does not contain amine-bound hydrogen after derivatization, but the isotopic composition of MCF cannot be isolated from our measured δ²H value for MCF + amine hydrogen (see Supplementary Section 4). For other researchers to adopt our correction method, they will need to re-analyze PPD of known δ²H (available by request) with their own reagents and water.

Certain preparatory steps can alter the isotopic compositions of amino acids (reviewed in Silverman et al., 2022). In particular, degradation or non-quantitative recovery of amino acids during acid hydrolysis can lead to isotopic fractionation (e.g., Bada et al., 1989; Phillips et al., 2021). To assess the isotopic consequences of different hydrolysis conditions on δ²H_AA values, standard BSA protein was hydrolyzed at different temperatures (105 or 110°C), for different durations (20, 24, or 48 h), anoxically or with O₂ present. Compared to conventional hydrolysis conditions (6N HCl, 110°C, 20-24 h, anoxic; Silverman et al., 2022), no treatment significantly altered the hydrogen isotope composition of amino acids (Supplementary Table S4).

To investigate whether carbon-bound hydrogen in amino acids exchanges with aqueous medium during hydrolysis or derivatization, BSA and a mixture of amino acid standards were separately hydrolyzed (6N HCl, 110°C, 24 h) and derivatized to MOC esters in solvents with different isotopic compositions. Slopes of regressions of amino acid vs. water δ²H values (Supplementary Figures S9, S10) represent the equilibrium fractionation factor (α_eq) between organic hydrogen and water, multiplied by the fraction of hydrogen exchanged in the amino acid. To estimate the maximum percent of carbon-bound hydrogen exchanged, each slope was divided by α_eq = 0.9, an estimate based on fractionation factors for a variety of hydrogen positions in linear and cyclic organic molecules (Wang et al., 2009, 2013). These calculations indicate that ten amino acids experienced negligible (<2%) hydrogen exchange with aqueous medium during hydrolysis, while tryptophan experienced significant exchange (~27%; Supplementary Figure S9). Asparagine + aspartic acid (Asx), glutamine + glutamic acid (Glx), and tyrosine experienced moderate exchange (4%–10%); this effect has been previously demonstrated through deuterated and tritiated hydrolysis experiments (Hill and Leach, 1964; Fogel et al., 2016) and is likely due to the increased lability of hydrogen adjacent to the polar—R groups. Hydrogen exchange in tryptophan may have occurred through a reversible reaction with sulfur-containing amino acids in the presence of oxygen (common tryptophan degradation mechanisms summarized in Silverman et al., 2022). All amino acids experienced low (<2%) hydrogen exchange during derivatization (Supplementary Figure S10).

The substantial variations in  2εAA/w values within wildtype organisms grown on glucose are summarized in Figure 4A and Supplementary Table S2 for the five amino acids analyzed, and in Supplementary Figure S6 and Supplementary Table S3 for the other amino acids measured in this study. All organisms produced similar  2εAA/w patterns, where phenylalanine and proline were the most ²H-enriched, while isoleucine and valine were the most ²H-depleted. This pattern mirrors that previously observed for E. coli (Fogel et al., 2016), with the exception of phenylalanine, which in our study was significantly more ²H-enriched relative to the average. Within a single organism, the five  2εAA/w values spanned large ranges (220‰–352‰). Valine exhibited the largest variation across organisms (190‰) while proline and leucine  2εAA/w values varied the least (86‰–93‰).  2εAA/w values from biological replicates of wildtype organisms grown on glucose were generally reproducible (within 30‰) except for phenylalanine and isoleucine from P. fluorescens cultures (Supplementary Figure S5; Supplementary Table S2), which differed between replicates for unknown reasons.  2εAA/w values in E. coli cultures grown on glucose also differed by >30‰ for four of the amino acids, but these cultures are not considered biological replicates due to differences in sample preparation (see Sections 2.1, 2.2).

Metabolic flux analysis carried out in a previous study (Wijker et al., 2019) revealed substantial differences in pathways used for glucose breakdown and in carbon fluxes through central metabolic enzymes (Supplementary Figure S3; Supplementary Table S1; Wijker et al., 2019). E. coli and B. subtilis primarily used the EMP pathway for glucose catabolism and excreted high fluxes of acetate. In contrast, E. meliloti and R. radiobacter mainly relied on the ED pathway to metabolize glucose and exhibited moderate fluxes through the TCA cycle. P. fluorescens exhibited high fluxes through the ED pathway and TCA cycle, as well as periplasmic conversion of glucose to gluconate and 2-ketogluconate, and cyclic flux through the EDEMP pathway (Nikel et al., 2015; Wijker et al., 2019).  2εAA/w values for leucine and valine correlated with carbon fluxes through enzymes related to pyruvate synthesis: KDPG aldolase (ED pathway), PEP carboxykinase (anaplerotic pathway), phosphoglucose isomerase (EMP pathway), pyruvate kinase (EMP + ED pathways), and transketolase (PP pathway; Figure 5). Directions of correlations between  2εAA/w values and carbon flux through EMP and ED pathways opposed those observed for lipid/water fractionations ( 2εL/w, presented in Wijker et al., 2019), although these relationships may not be directly comparable, as  2εL/w values are primarily controlled by NADPH metabolism (Zhang et al., 2009; Wijker et al., 2019) while leucine and valine do not inherit hydrogen from NADPH.  2εAA/w values for other amino acids did not correlate with carbon flux through any central metabolic enzyme in wildtype organisms.

Specific dehydrogenase and transhydrogenase genes were deleted in E. coli organisms in a previous study (Wijker et al., 2019) to interrogate the influence of NADPH on lipid δ²H values. Deletion of these genes forced carbon flux through alternative central metabolic enzymes to accomplish glucose catabolism and NADPH balance (Supplementary Figure S4; Table S1). Despite drastic differences in the magnitudes of carbon fluxes, E. coli mutant organisms produced similar  2εAA/w values compared to wildtype E. coli culture #1, differing by <40‰ for any given amino acid (Figure 4B; Supplementary Table S2; see note about E. coli culture #2 in Section 3.3). Lipid δ²H values showed a similar response in these organisms (Wijker et al., 2019), as did δ²H_AA values in E. coli mutants with inhibited glycolysis or oxidative pentose phosphate pathways in a previous study (Smith et al., 2022). Nevertheless, variations in isotopic compositions hint at some control by NADPH, as δ²H_AA values correlated with carbon fluxes through all NADPH-related enzymes (Supplementary Figure S15), with proline exhibiting the strongest correlations (R² = 0.70–0.86), followed by phenylalanine (R² = 0.58–0.71), then isoleucine (R² = 0.36–0.49). The PGI knockout mutant, JW3985, had a severely perturbed metabolism and fell off the regressions in most cases so was excluded from the regression analyses.

In addition to glucose, wildtype organisms were cultured on acetate, citrate, fructose, pyruvate, and/or succinate, which enter central metabolism at different nodes and activate different catabolic pathways for substrate breakdown.  2εAA/w values from biological replicates of wildtype organisms grown on each substrate were generally reproducible (within 30‰) except for proline and phenylalanine in B. subtilis grown on succinate, and proline in R. radiobacter grown on succinate, which differed between replicates for unclear reasons (Figure 6, Supplementary Figure S5, Supplementary Table S2). Fructose led to similar  2εAA/w values as glucose-based growth, while pyruvate and TCA cycle substrates (acetate, citrate, and succinate) led to ²H-enrichment of all amino acids (Figure 6), mirroring phenomena observed for lipids (Zhang et al., 2009; Osburn et al., 2016; Wijker et al., 2019). The amount of ²H-enrichment varied widely for each amino acid. Proline in B. subtilis grown on succinate, and isoleucine and phenylalanine in P. fluorescens grown on acetate, exhibited the largest singular ²H-enrichments (286‰–360‰ higher  2εAA/w values relative to those during glucose-based growth). However, phenylalanine was generally ²H-enriched by the least amount (<100‰ difference between  2εPhe/w values upon growth on TCA cycle substrates relative to on glucose for all organisms except P. fluorescens), while valine was generally ²H-enriched by the greatest amount (142–245‰). These differences were significantly greater than those between growth water (<15‰) or substrate δ²H values (<85‰; Supplementary Table S2).

 2εAA/w patterns were strikingly similar across wildtype organisms and E. coli mutants grown on all carbon substrates (Figures 4, 6), indicating that biosynthetic pathways (as opposed to catabolic pathways) serve as first-order controls on δ²H_AA values. Our interpretation is consistent with Fogel et al. (2016), who observed similar δ²H_AA patterns in E. coli cultured on glucose or tryptone in different growth waters. In this section, we investigate the biochemical factors that set the general  2εAA/w pattern in glucose-grown cultures, focusing our interpretation on hydrogen sources and mechanisms of hydrogen exchange, as well as relevant isotope effects associated with enzymes in central metabolic and biosynthetic pathways. Based on these interpretations, we speculate on the most prominent types of biological information that can be obtained from δ²H_AA measurements. Amino acids are discussed in order from most to least ²H-enriched. As these microbes share the same biosynthetic pathways, variations in ²ε values of a given amino acid across organisms (e.g., Figure 4A) hint at the importance of additional, second-order controls, which are examined in Section 4.2.

While the similar patterns of  2εAA/w values across glucose-grown organisms reveals remarkable consistencies in net isotopic fractionations of central metabolic and biosynthetic pathways, the variations in  2εAA/w values hint at subtle physiological differences in the organisms. Culturing experiments in which the δ²H value of an organism's growth water is manipulated can provide some constraints on the biochemical causes of these differences (presented in Supplementary Section 6.1). The substantial systematic shifts in  2εAA/w values upon growth on different carbon substrates provide further insight into biochemical controls on these isotopic signals, including the differential activation of enzymes across catabolic pathways (Section 4.2.1). Finally, potential differences in  2εAA/w values due to enzymatic variations in biosynthetic pathways are considered (Section 4.2.2).

Here we have explored several hypotheses regarding biochemical controls on δ²H_AA values in aerobic heterotrophic microbes. Our results demonstrate that the overall pattern of amino acid/water fractionations is highly correlated with the individual biosynthetic pathways in organisms, while magnitudes of fractionations are likely controlled by the organic precursor and NADPH isotope compositions. In turn, the δ²H values of organic precursors and the NADPH pool are driven by the relative fluxes through different central metabolic pathways, which vary depending on the catabolic pathways activated for substrate degradation. Together, these results suggest that δ²H_AA values may be useful tracers for carbon processing within organisms and the environment. As the 20 biological amino acids are ubiquitous across the tree of life, and organisms that share similar biosynthetic pathways should produce similar patterns of amino acid/water fractionations, we may expect the δ²H values of amino acids synthesized de novo across microbial and metazoan taxa to be governed by the same controls. Quantitative interrogations of the hypotheses presented in this study will likely require modeling work (e.g., Mueller et al., 2022), but are needed in order to fully understand the information encoded in these signals.

While controls on δ²H_AA values are clearly nuanced and it may not be possible to uniquely relate all δ²H_AA values to simple biological or environmental properties, the different combinations of hydrogen sources in amino acids leads to rich variability in δ²H_AA signals, thus numerous potential applications of δ²H_AA analysis. For example, δ²H_AA values may provide insight into the metabolic strategies that microbes employ for carbon and energy acquisition. Microbes are the major drivers of nutrient and energy cycling in the environment, thereby playing substantial roles in shaping the geochemistry of our planet. While we observed large systematic variations in amino acid/water fractionations across heterotrophic microbes grown on different carbon substrates, we predict that even larger variations should exist between organisms of different metabolic classes (e.g., heterotrophs vs. autotrophs), as is the case for lipids (Zhang et al., 2009; Osburn et al., 2016). If true, δ²H_AA values and patterns may be used to decipher the metabolisms of unculturable organisms, to distinguish contributions by metabolically distinct organisms to geochemical processes in nature (e.g., in largely inaccessible environments such as deep subsurface biospheres), and/or to quantify contributions by different metabolisms to bulk organic matter in the environment (including by mixotrophic organisms that operate on a continuum of carbon and energy acquisition strategies, or across diverse taxa such as plants, algae, bacteria, and fungi). While lipid δ²H values have been suggested as a potential tool for this latter application (Cormier et al., 2022), amino acids may offer distinct advantages, as their δ²H values can be linked to host organisms through the isolation and sequencing of proteins (Gharibi et al., 2022b), and their different combinations of hydrogen sources capture a more diverse suite of information than is encoded in lipids (which inherit hydrogen from essentially the same three sources: acetyl-CoA, NADPH, and water). However, we view amino acid and lipid δ²H analyses as complementary, as each can help inform interpretations from the other. Leucine and valine δ²H may provide the most direct information about the metabolic ‘state' or lifestyle of an organism, as the positioning of pyruvate as a central node in metabolism makes it relatively sensitive to the molecular wiring of central metabolic pathways in cells. In turn, accounting for these metabolic signals may help isolate the NADPH-driven signals in proline and lipid δ²H values, thereby increasing their sensitivity as redox indicators in cells. Moreover, if the majority of phenylalanine hydrogen is indeed equilibrated with water at the level of metabolic intermediates, the δ²H value of phenylalanine may provide insight into water δ²H in places where water sources are unclear (e.g., in environments with intermittent wet/dry cycles), and if such equilibrium is temperature-dependent, phenylalanine could additionally serve as a potential bio-thermometer.

In addition to geomicrobiology-based applications, δ²H_AA analyses may provide useful information about eukaryotic organisms, including their stressors, diets, and migration patterns. The application of δ²H_AA values to human and wildlife forensics is in the early stages of exploration, with links between mammal diet, drinking water, and δ²H_AA values beginning to emerge (Newsome et al., 2020; Mancuso et al., 2023). The biochemical controls discussed here may only be relevant for interpreting δ²H values of non-essential amino acids (e.g., proline) as well as those with relatively high contributions from the gut microbiome (e.g., phenylalanine; Newsome et al., 2020). However, the differentiated tissues in animals, variable residence times of proteins in cells, and integration of numerous dietary hydrogen sources significantly increase the complexity of information encoded in mammalian δ²H_AA signals, which will require detailed investigations to disentangle. δ²H_AA signals in plants likely carry important physiological information as well. Like microbes, plants can synthesize all 20 proteinogenic amino acids, yet their hydrogen metabolism may be simpler to interpret, as plants derive their organic hydrogen exclusively from water. Amino acids are involved in numerous mechanisms of stress alleviation in plants (e.g., Batista-Silva et al., 2019), so their δ²H_AA values may encode information about their physiological status. For example, synthesis and degradation of proline helps maintain redox balance (i.e., the NADP/NADPH ratio) in plants and appears to facilitate drought tolerance (Sharma et al., 2011; Bhaskara et al., 2015; Batista-Silva et al., 2019). In the first step of proline catabolism, proline dehydrogenase removes a carbon-bound hydrogen atom, which should lead to ²H-enrichment of the residual proline pool. Consequently, proline δ²H values may serve as a sensitive indicator of oxidative and drought stress in plants, which are highly important aspects of crop health. This mechanism may additionally play a role in the extreme ²H-enrichment of proline observed in gray seals (Gharibi et al., 2022a), which endure significant periods of oxidative stress while diving. Overall, δ²H_AA analysis has potential to become a highly useful isotopic tool for a variety of diverse applications, which will undoubtedly emerge as we continue to unravel the biochemical mechanisms underpinning δ²H_AA signals in organisms.