Meta‐analysis of primary producer amino acid δ15N values and their influence on trophic position estimation

Compound‐specific stable isotope analysis of individual amino acids (CSIA‐AA) has emerged as a transformative approach to estimate consumer trophic positions (TPCSIA) that are internally indexed to primary producer nitrogen isotope baselines. Central to accurate TPCSIA estimation is an understanding of beta (β) values—the differences between trophic and source AA δ15N values in the primary producers at the base of a consumers’ food web. Growing evidence suggests higher taxonomic and tissue‐specific β value variability than typically appreciated. This meta‐analysis fulfils a pressing need to comprehensively evaluate relevant sources of β value variability and its contribution to TPCSIA uncertainty. We first synthesized all published primary producer AA δ15N data to investigate ecologically relevant sources of variability (e.g. taxonomy, tissue type, habitat type, mode of photosynthesis). We then reviewed the biogeochemical mechanisms underpinning AA δ15N and β value variability. Lastly, we evaluated the sensitivity of TPCSIA estimates to uncertainty in mean βGlx‐Phe values and Glx‐Phe trophic discrimination factors (TDFGlx‐Phe). We show that variation in βGlx‐Phe values is two times greater than previously considered, with degree of vascularization, not habitat type (terrestrial vs. aquatic), providing the greatest source of variability (vascular autotroph = −6.6 ± 3.4‰; non‐vascular autotroph = +3.3 ± 1.8‰). Within vascular plants, tissue type secondarily contributed to βGlx‐Phe value variability, but we found no clear distinction among C3, C4 and CAM plant βGlx‐Phe values. Notably, we found that vascular plant βGlx‐Lys values (+2.5 ± 1.6‰) are considerably less variable than βGlx‐Phe values, making Lys a useful AA tracer of primary production sources in terrestrial systems. Our multi‐trophic level sensitivity analyses demonstrate that TPCSIA estimates are highly sensitive to changes in both βGlx‐Phe and TDFGlx‐Phe values but that the relative influence of β values dissipates at higher trophic levels. Our results highlight that primary producer β values are integral to accurate trophic position estimation. We outline four key recommendations for identifying, constraining and accounting for β value variability to improve TPCSIA estimation accuracy and precision moving forward. We must ultimately expand libraries of primary producer AA δ15N values to better understand the mechanistic drivers of β value variation.


| BACKG ROU N D
Knowledge of an organism's position in a food web is foundational to understanding the structure and function of ecological communities (Leibold et al., 1997;Persson, 1999;Post et al., 2000). Characterizing the nitrogen isotope variation at the base of the food web (δ 15 Nbaseline ) remains one of the biggest challenges to accurately estimating consumer trophic position using stable isotope analysis (SIA).
Over the past decade, compound-specific isotope analysis of amino acids (CSIA-AA) has transformed our ability to study food web dynamics as it allows for the simultaneous reconstruction of δ 15 N baseline and consumer trophic ecology without needing to independently identify and isotopically characterize the relative contribution of δ 15 N baseline . Trophic dynamic studies using CSIA-AA are premised on differential isotopic fractionation of individual AAs during trophic transfer (McMahon & McCarthy, 2016), with AAs broadly categorized as those that do ('trophic' AA) and do not ('source' AA) undergo significant trophic fractionation (Popp et al., 2007). Importantly, because the δ 15 N values of source AAs (phenylalanine: Phe; methionine: Met; lysine: Lys; tyrosine: Tyr) remain relatively unaltered moving through the food web, they provide an internally indexed estimate of δ 15 Nbaseline . These data, when used in combination with the δ 15 N values of trophic AAs [glutamic acid (Glu) + glutamine (Gln): Glx; aspartic acid (Asp) + asparagine (Asn): Asx; alanine: Ala; isoleucine: Ile; leucine: Leu; proline: Pro; valine: Val], provide a method for estimating consumer trophic positions (TP CSIA ) calibrated to each consumer's integrated δ 15 N baseline . This ability to simultaneously partition isotopic variability resulting from changes in δ 15 N baseline versus changes in the number of trophic transfers has been used to reveal foraging patterns of cryptic species and systems (Gerringer et al., 2017;Saccò et al., 2019), reconstruct ancient diets and food webs (Jarman et al., 2017;McMahon et al., 2019), and resolve complex trophic dynamics and patterns in animal movement (Dale et al., 2011;Matsubayashi et al., 2020).
Understanding of AA-specific trophic discrimination factors (TDFs) and primary producer δ 15 N values is fundamental to accurate trophic position estimation using CSIA-AA ( Figure 1). TP CSIA is most commonly estimated using the equation: where δ 15 N Trophic AA and δ 15 N Source AA represent the δ 15 N values of consumer trophic and source AAs (typically Glx and Phe, respectively); TDF Trophic AA-Source AA is the trophic discrimination factor reflecting changes in trophic and source AA δ 15 N values between diet and consumer (TDF Trophic AA-Source AA = Δ 15 N Trophic AA -Δ 15 N Sourcec AA ); and β (beta) is the difference between trophic and source AA δ 15 N values in the primary producer(s) at the base of the food web.
A confluence of evidence suggests variability in AA-specific TDFs within and among taxa is mechanistically linked to variation in animal physiology and biochemistry (Bradley et al., 2015;McMahon & McCarthy, 2016;Nielsen et al., 2015), with diet quality and mode of nitrogen excretion emerging as major drivers of variability (Germain et al., 2013;. Similarly, seminal research identified characteristic differences in β values among different groups of primary producers-algae, C 3 plants and C 4 plants-that are now routinely applied in trophic ecology studies . Although research into and consideration of AA-specific TDF variability continues to grow (Figure 2), study of β values, including sources of heterogeneity and influence on TP CSIA , has lagged far behind.
Through integration of data from cultivated and wild primary producers,  first identified what have become conventionally applied β values for aquatic algae (mean ± SD: +3.4 ± 0.9‰) and vascular plants (C 3 : −8.4 ± 1.6‰, C 4 : −0.4 ± 1.7‰). However, multiple recent studies have demonstrated there can be substantial AA δ 15 N value variation within these groups that can influence β values, particularly among terrestrial plants. For example,  identified an ~3.5‰ difference in β values between woody and herbaceous C 3 plants, and differential isotopic fractionation of Phe appears to also contribute to differences in β values among mangrove species  and among some grass tissues .  and  demonstrated that the phenology of plant tissue synthesis can strongly influence tissue-specific primary producer AA δ 15 N values, with early growth leaves and flowers having different isotopic compositions and β values than those generated later in the season.  observed differences in AA δ 15 N values within and among cultured eukaryotic microalgae and cyanobacteria, which may yield divergent β values for these taxa. Such currently underappreciated isotopic and β value variation within and among primary producers holds the potential to strongly impact TP CSIA estimates.
Given the rise in application of CSIA-AA to trophic ecology studies, there is a pressing need to re-visit β to identify ecologically (1) TP CSIA = 1 + δ 15 N Trophic AA − δ 15 N Source AA − TDF Trophic AA -Source AA , accuracy and precision moving forward. We must ultimately expand libraries of primary producer AA δ 15 N values to better understand the mechanistic drivers of β value variation.

| VARIAB ILIT Y IN PRIMARY PRODUCER AMINO ACID NITROG EN ISOTOPE R ATI OS
Our discussion focuses primarily on β values derived from Glx and Phe (β Glx-Phe ) because they are the most commonly measured and applied trophic and source AAs for estimating TP CSIA . However, we also present β values for all trophic (Asx, Ala, Ile, Leu, Pro and Val) and 'metabolic' (Thr) AAs relative to all source AAs (Phe, Lys, Met and Tyr) with discussion of potential alternative useful pairings wherever pertinent ( Table 2; Table S1; Figure 3; Figure S1). These data come from a mix of natural and human manipulated primary producers (e.g. laboratory or farm settings; Table 1), which may contribute to β value variation in some taxa dependent on N source and N assimilation pathway (see Insights into β Variability: Nitrogen Assimilation and Amino Acid Biosynthesis; Figure S2). Marine phytoplankton data were primarily from laboratory cultures, whereas macroalgal, seagrass and freshwater primary producer data were primarily from natural environments. Approximately two-thirds of the terrestrial plant data were from cultivated plants in suburban/urban or farm settings.

| Vascular primary producers
Further study of non-vascular terrestrial plants will shed important light on relationships among lignin biosynthesis, AA nitrogen isotope fractionation and β value variability.
In contrast, limited data suggest β Glx-Phe values were generally similar among grass tissues ( Figure 3; Figure S1), although β values that include the trophic AA Ile (β Ile-Phe , β Ile-Lys , β Ile-Tyr ) appear higher for grass structural tissues relative to both reproductive and leaf tissues ( Figure S1a,b,d). These forb, grass and woody plant tissue-specific β value patterns extended to most β X-Phe combinations that could be evaluated ( Figure S1); there were generally insufficient data to evaluate tissue-specific β X-Lys , β X-Tyr and β -Met variation. While tempting to speculate about potential underlying mechanisms, a necessary limitation of this meta-analysis is that it required pooling of limited Note: Natural includes non-directly human manipulated/cultivated samples (e.g. wild plants). Culture includes directly human manipulated/cultivated samples, either in laboratory or farm settings. POM = particulate organic matter. n genera is the number of autotroph genera analysed; specieslevel identification was not always reported. n samples is the total number of autotrophs samples.
TA B L E 1 Summary of published research devoted to characterizing amino acid δ 15 N values in photosynthetic organisms (plants/macrophytes, eukaryotic macro-and microalgae, cyanobacteria, chemoautotrophic bacteria). See Figure 3 for characterization of tissue types analysed. Dashes denote where genus-level identification was not reported  TA B L E 2 β Glx-Phe and β Glx-Lys values (samples size, mean and standard deviation) by broad taxonomic groupings for data presented in Figure 3. See Table S1 for the full suite of summary β X-Phe and β X-Lys data and complementary β X-Tyr and β X-Met data data from a variety of taxa across space and time and is thus not adequately designed to robustly evaluate all potential sources of variability. Our understanding of tissue-specific β value variation would thus be enhanced through more comprehensive sampling of multiple tissue types from the same individuals or species. Similarly, more targeted experimental studies would allow for quantitative evaluation of how suites of factors (e.g. vascularity, lignin production, N assimilation pathway, AA concentrations, AA turnover rates) interact to influence β values.
Marine non-vascular primary producers were the second most analysed group of autotrophs in our meta-analysis (30.3% of analysed taxa; Table 1). We found that β X-Phe values were generally similar among green (n = 3), brown (n = 7) and red (n = 4) macroalgae and among cyanobacteria orders Nostocales (n = 7), Oscillatoriales as silica-lined cell walls, a central vacuole that stores nutrients and functional urea uptake and utilization pathways (Bromke, 2013;Falkowski et al., 2004;Tozzi et al., 2004), which may contribute to differences in AA δ 15 N patterns relative to other phytoplankton. Of note, almost all β values for eukaryotic microalgae and cyanobacteria to date were derived from laboratory cultures. Although methodologically challenging, further study of wild-collected phytoplankton taxa is needed to better assess natural β value variability.
Freshwater non-vascular primary producers are woefully understudied within the CSIA-AA literature (3.8% of analysed taxa; Table 1), with only four studies reporting AA δ 15 N data . Freshwater eukaryotic microalgae β Glx-Phe values (+4.2 ± 0.7‰, n = 5) were similar to those of their marine counterparts (+3.9 ± 1.6‰, n = 22, p > 0.05; Figure 3). Lack of taxonomic identifications prevented further analysis of the limited freshwater non-vascular autotroph data but this is an area of important future research given the rapid growth in TP CSIA applications in freshwater systems.
Chemoautotrophs, which derive energy from the oxidation of inorganic compounds, are among the least characterized groups within the AA δ 15 N literature. Data for chemoautotrophic bacteria in this meta-analysis were derived from just two studies:  and . Notably, chemoautotroph β Glx-Phe values were similar to those of the much more studied eukaryotic microalgae and cyanobacteria groups, which may be because the analysed microbes all use the phenylpyruvate pathway for Phe synthesis . Nevertheless, chemoautotrophs exhibit considerable physiological and phylogenetic diversity, which may ultimately influence AA biosynthesis and β values within this diverse group of organisms (Nakagawa & Takai, 2008). Given their importance to food webs in extreme environments such as hydrothermal vents, cold seeps, natural gas and methane seeps, and anoxic sediment waters (Nakagawa & Takai, 2008), expanded investigation of nitrogen isotope dynamics associated with chemoautotroph AA biosynthesis will improve food web studies within these systems.

| IN S I G HTS INTO β VARIAB ILIT Y: NITROG EN A SS IMIL ATI ON AND AMINO ACID B I OSYNTHE S IS
In this section, we discuss nitrogen assimilation and AA biosynthesis and degradation pathways to highlight the potential mechanisms leading to variation in β values within and among primary producer groups. We focus primarily on terrestrial plants because they are the most analysed group of autotrophs and exhibit the largest range in β values. However, also make note that many of the processes discussed are directly applicable to non-vascular autotroph nitrogen metabolism. Generally, variation in the most often employed β val- Autotrophs acquire and assimilate nitrogen using multiple complex pathways (Figure 4). The extent to which autotrophs utilize each of the various nitrogen sources (e.g. NO − 3 , NH + 4 , N 2 , AA) depends on a number of factors, including taxonomy, growth state, microbial symbiont status and environmental conditions (Jackson et al., 2008;Nacry et al., 2013;Szpak, 2014). These processes can contribute to variation in δ 15 N baseline (i.e. the bulk tissue δ 15 N values of primary producers at the base of a food web) and individual AA δ 15 N values because there are differential kinetic isotope effects associated with each inorganic nitrogen acquisition pathway (Lachmann et al., 2019;Werner & Schmidt, 2002) such that bulk tis- grown using various N sources (e.g. animal manure, synthetic fertilizers) exhibit more variable Glx and Phe values (tomato: 10.5 and 10.7‰, wheat: 9.6 and 11.0‰) than β Glx-Phe values (tomato: 2.4‰, wheat: 5.1‰; . However, there is some evidence that fertilizer application rates can slightly impact β Asp-Phe values , but the effects of nitrogen supply rates on primary producer AA concentrations, AA δ 15 N values and β values have not been systematically tested, including in the above examples, and warrant further study.
There are two N assimilation pathways-direct uptake of AAs from the environment and N 2 fixation by microbial symbiontsthat may contribute to variation in β values within and among taxa.
Many autotrophs, including terrestrial plants, all bacteria and some eukaryotic microalgae, have the capability to directly uptake AAs from their environment (i.e. soil or water; Kielland, 1994;Palenik & Morel, 1990;Zehr & Kudela, 2011). Given that δ 15 N values of soil AAs are highly variable because they originate from a variety of organismal sources and undergo varying degrees of degradation before assimilation (Philben et al., 2018), differential rates of direct AA uptake could increase variation in β values among primary producers. The relative importance of this incorporation mechanism varies greatly and is more likely to occur in regions with high soil organic matter content and low decomposition rates (e.g. the Arctic tundra; Kielland, 1994) than in ecosystems with high decomposition and mineralization rates (but see Gioseffi et al., 2012). In addition, autotrophic symbioses with microbes that can access atmospheric nitrogen sources, particularly N 2 (e.g. plants and N 2 -fixing Rhizobium, diatoms and N 2 -fixing cyanobacteria), can uniquely influence Ala, Asp, Asn, Glu and Gln production rates (Lambers et al., 2008;Liu et al., 2018;Werner & Schmidt, 2002) (Galili et al., 2001;. The extent to which Phe and Lys are catabolized for secondary compound synthesis varies due to a variety of factors, such as growth form and environmental conditions (Galili et al., 2001;Sharma et al., 2019). Phe may display greater isotopic fractionation than Lys and thus a wider range in δ 15 N values as it is typically catabolized more frequently; approximately a third of plant organic matter is synthesized from Phe (Maeda & Dudareva, 2012;Pascual et al., 2016). Phe is a precursor to many secondary compounds, including lignin, flavonoids and tannins (Maeda & Dudareva, 2012;Vogt, 2010).  reported a positive correlation between the concentrations of lignin-a structural compound-and Phe δ 15 N values. This relationship occurs because the first step in Phe catabolism for the synthesis of secondary metabolites is deamination, in which Phe loses an amine group to become cinnamate (Deng & Lu, 2017;. This initial deamination step exhibits isotopic discrimination (Hermes et al., 1985) such that Phe molecules containing 14 N are preferentially deaminated, leaving behind a 15 Nenriched residual pool of Phe δ 15 N values . Thus, plants composed of more lignin, like trees, typically have relatively higher Phe δ 15 N values (and thereby lower β Glx-Phe values) than herbaceous plants containing less lignin . Lys, on the other hand, can act as an alternative respiratory substrate, increasing in concentration during periods of abiotic stress, and is also a precursor to metabolites involved in plant immunity (Galili et al., 2001;Zeier, 2013). Recent work on a freshwater green alga and terrestrial C 3 , C 4 and CAM plants found that Lys δ 15 N values were lower and less variable than those of Phe across producer groups and were strongly correlated to bulk tissue δ 15 N values (A. C. Besser, unpublished data), a pattern that likely explains the low variability in β X-Lys values relative to β X-Phe values observed herein (Figure 3; Figure S1).
Phe δ 15 N values did not correlate with bulk tissue δ 15 N values in these producer groups, highlighting the highly variable nature of Phe δ 15 N values.
In contrast to Phe and Lys, Glu and Gln are unlikely to contribute significantly to β value variability. These AAs are central to plant nitrogen metabolism and have high turnover rates, particularly in leaves (Kruse et al., 2003). During assimilation, plants add NH + 4 , reduced from NO − 3 and NO − 2 via nitrate reductase and nitrite reductase, respectively (Werner & Schmidt, 2002), to an α-ketoglutarate molecule using glutamate dehydrogenase to form Glu or an existing Glu molecule using glutamine synthase to form Gln (Lambers et al., 2008;Werner & Schmidt, 2002). Once assimilated, nitrogen is primarily shuttled among compounds within plant tissues via Glu and Gln (Figure 4), which donate amine groups during AA biosynthesis via transamination reactions (Werner & Schmidt, 2002). As a result, Glu and Gln, which make up the highest proportion of AAs in plant proteomes (Hildebrandt et al., 2015), are likely isotopically well-mixed within the internal nitrogen pool, potentially leading to lower variability in their δ 15 N values as compared to other AAs.
Lastly, tissue-specific differences in plant trophic or source AA δ 15 N values, and by extension β values, likely occur due to differing AA concentrations and turnover rates among plant tissue types (Camargos et al., 2004;Kruse et al., 2003;Mapelli et al., 2001).
Differences in AA concentrations are largely determined by free, or soluble, AA concentrations, which vary greatly with changing nitrogen supply conditions (Caputo & Barneix, 1997), rather than changes in protein-bound AA concentrations. For example, Glu and Gln are among the most abundant free and protein-bound AAs in Canavalia ensiformes seeds and leaf tissues, but among the least abundant AAs in seedling plant stem tissues (Camargos et al., 2004). Phe displays the opposite pattern, being most abundant in seedling plant stem tissues but least abundant in leaf and seed tissues (Camargos et al., 2004). AAs, and thereby AA δ 15 N values, typically turnover more quickly in leaves than in other tissues and in younger tissues than in older tissues (Kruse et al., 2003). Different AA turnover rates among tissue types may lead to significant differences in AA δ 15 N values and, in turn, more variable β values among tissue types.
However, more experimental studies measuring plant AA concentrations, turnover rates and δ 15 N values among tissue types are needed to constrain the relative magnitudes of variation. primary producer assemblages support focal food webs. Systemand consumer-specific β Glx-Phe value ranges were derived from our meta-analysis (Figure 3), whereas TDF Glx-Phe ranges were derived from variation related to consumer and system type, including reductions due to shifts in diet quality and/or mode of nitrogen excretion (see meta-analysis by McMahon & McCarthy, 2016).

| S EN S ITIVIT Y OF TROPHI C P OS ITI ON E S TIMATE S TO UN CERTAINT Y IN B E TA VALUE S
In the terrestrial case study, we modelled TP CSIA for four consumers within an insect food web on a mature apple orchard (TP 2: apple aphid, Aphis pomi; TP 3: hoverfly, Eupeodes sp.; TP 4: parasitoid wasp, Bothriothorax sp.; TP 5: hyperparasitoid wasp, Pachyneuron albutius; . Only vascular primary producers contribute nitrogen to this food web. We used the  single-TDF equation in combination with the published consumer δ 15 N Glx and δ 15 N Phe data to estimate consumer TP CSIA for all combinations of ecologically relevant β Glx-Phe values between −12.0 and 0.0‰ from this study and TDF Glx-Phe values between 6.5 and 8.5‰ (mean = 7.5‰) as observed for terrestrial insects (e.g. Takizawa et al., 2020).
Our analyses revealed that TP CSIA estimates were highly sensitive to both univariate and bivariate changes in β Glx-Phe and TDF Glx-Phe values ( Figure 5; Table 3), and that sensitivity varied as a function of consumer and system type. Across all three case studies, there was a general pattern of increasing TP CSIA sensitivity to both β Glx-Phe and TDF Glx-Phe values with increasing trophic position. This is evidenced by the narrowing of the space between each isocline (i.e. 0.25 TP bin) with each trophic transfer, which indicates that changes in parameter values will lead to bigger changes in TP CSIA estimates for higher-order consumers relative to lower-order consumers. For example, across the three case studies TP CSIA estimates spanned 1.1-3.0 for primary consumers but 2.1-5.1 for the top predators (Table 3), illustrating that unconstrained variance in parameter values can yield 2+ unit variability in TP CSIA estimates. Moreover, the magnitude of variability in TP CSIA increased in relation to the range of possible β Glx-Phe values (Table 3) Lastly, our sensitivity analysis revealed that assumptions of TDF Glx-Phe variability within food webs can have complex effects on the relative influence of β and TDF values on TP CSIA estimates. This is illustrated by the abrupt shift in pattern of isocline slopes between TL 3 and 4 in the freshwater case study (Figure 5b), where we used an ecologically relevant lower TDF Glx-Phe value range (3.5-5.5‰) for osprey relative to the aquatic consumers (6.5-8.5‰) to reflect a coupled shift in diet quality and mode of nitrogen excretion. After this transition point, there are large areas in bivariate β-TDF value space where TP CSIA estimates are very insensitive to changes in TDF Glx-Phe but highly sensitive to changes in β Glx-Phe value (e.g. β between −4.0 and +4.0). These results contrast sharply with those derived from a sensitivity analysis where we assumed TDF Glx-Phe was constant for all freshwater consumers (range: 5.5-7.5‰; Figure S4), which yields more predictable and consistent isocline patterns across consumers but at the expense of higher sensitivity of TP CSIA estimates to changes in β and TDF values (narrowed isocline bins).
Collectively, these findings show that in some cases β Glx-Phe values are more influential to TP CSIA estimation than TDF Glx-Phe values. However, CSIA-AA practitioners will likely rarely know a priori which parameter is more important, highlighting the need for more critical evaluation of both β and TDF values in addition to TP CSIA equations (e.g. single-vs. multi-TDF) in all CSIA-AA trophic ecology studies. TA B L E 3 Consumer δ 15 N Glx and δ 15 N Phe data, baseline scenario parameters (β and trophic discrimination factors, TDF Glx-Phe ), and baseline scenario trophic position (TP CSIA ) estimates used to evaluate how variation in mean β values and TDFs influence mean TP CSIA estimates ( Figure 5). Max-Min TP CSIA represents the difference between the maximum and minimum TP CSIA estimated for each consumer within the sensitivity analysis

Consumer data
Baseline scenario  Table 3 for list of data sources and justification for parameter means and ranges   Figure 5). Such efforts will also help fill data gaps for the myriad of underanalysed primary producer taxonomic groups (e.g. bryophytes, pteridophytes and gymnosperms) and expand our understanding of ecologically relevant sources of β value variability (e.g.

Max-Min TP
taxonomic, tissue type, mode of photosynthesis, N assimilation pathway). Subsequent recommendations pertain to scenarios where in-situ sampling is infeasible (e.g. historical ecology, archaeology).
2. Within systems dominated by vascular or non-vascular primary producers, use the most up to date β values and variance estimates, currently resulting from this meta-analysis, as starting points for TP CSIA estimation (vascular = −6.6 ± 3.4‰; non-vascular = +3.3 ± 1.8‰; Table 2; Table S1). The β values in this meta-analysis integrate all published AA δ 15 N data to date and thus most accurately reflect known β value variation among primary producers. Importantly, this recommendation includes propagation of the higher β value error estimates presented here using Monte Carlo simulation or simple Taylor expansion of the canonical TP CSIA equation (Equation 1) to estimate TP CSIA uncertainty more accurately (e.g. through the propagate package in R; Gelwicks & Hayes, 1990). However, we encourage all CSIA-AA practitioners to carefully consider what β values would be most appropriate for their study system, including calculation of different β values using the raw data collated in this meta-analysis in combination with newly collected or published data in the future.
Nevertheless, our non-vascular β values may be appropriate to use in most marine applications where seagrasses are absent and terrestrial inputs are negligible (i.e. no vascular autotroph contribution) because known variability is relatively well-constrained.
Within terrestrial systems, we encourage practitioners to further constrain β values through purposeful identification of likely primary producer taxonomic groups and producer tissues that are important to focal food webs (e.g. grasses only, cacti only, estimated mix of woody and herbaceous plants).
3. For systems with both vascular and non-vascular primary producers, estimate β mix values using pertinent indices as endmembers in mixing models. When consumers use food webs with both vascular and non-vascular autotrophs, β values must reflect the realized admixture of primary producers (i.e. β mix ) for each consumer . Multiple molecular tools are available for use in tandem with mixing models to estimate the relative contribution of vascular and non-vascular primary producers to consumer food webs, such as bulk tissue δ 13 C and δ 15 N values (e.g. , AA δ 13 C 'fingerprinting' (e.g. Bowes et al., 2020; Chua et al., 2020), fatty acid analysis (Hebert et al., 2016) and Δ Met-Phe (δ 15 N Metδ 15 N Phe ; . Notably, phylogenetically distinct AA δ 13 C 'fingerprints' can be used to differentiate resource use among a diverse array of baseline energy sources (e.g. terrestrial plants, algae, seagrasses, bacteria, fungi; Larsen et al., 2013).
When combined with consumer AA δ 15 N analyses, this collection of approaches can more accurately characterize consumer trophic dynamics in modern and historical food webs (e.g. Jarman et al., 2017).
4. Within systems dominated by vascular autotrophs, such as terrestrial food webs, consider using Lys instead of Phe as the source AA for TP CSIA estimation. Our meta-analysis revealed that vascular plant β values calculated using the source AA Lys are considerably less variable across primary producer taxonomic groups and tissue types than those that use the source AA Phe (Figure 3), which may ultimately yield more precise TP CSIA estimates for food webs primarily supported by vascular autotrophs. The use of vascular β Glx-Lys is also advantageous because the data are approximately normally distributed and thus avoid distributional biases present in the vascular β Glx-Phe dataset ( Figure 3). As a result, we advance the call for optimization of analytical protocols to collect the full suite of measurable AAs rather than just focusing on Glx and Phe (Bradley et al., 2015;McMahon & McCarthy, 2016;Nielsen et al., 2015;O'Connell, 2017). This recommendation comes with the caveat that Lys is sometimes more analytically challenging to measure on standard gas chromatography-combustion-isotope ratio mass spectrometers given its late retention time and potential to co-elute with tyrosine (Tyr). With careful attention to late run temperature ramp structure, high-quality Lys δ 15 N data are achievable and likely worth the investment, particularly for terrestrial food web studies. 2049307. We thank two anonymous reviewers for comments on this manuscript.

CO N FLI C T S O F I NTE R E S T
The authors declare that there is no conflict of interest.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/ publo n/10.1111/2041-210X.13678.