Stable isotope fingerprinting traces essential amino acid assimilation and multichannel feeding in a vertebrate consumer

Abstract

Animals often consume resources from multiple energy channels, thereby connecting food webs and driving trophic structure. Such ‘multichannel feeding’ can dictate ecosystem function and stability, but tools to quantify this process are lacking. Stable isotope ‘fingerprints’ are unique patterns in essential amino acid (EAA) δ13C values that vary consistently between energy channels like primary production and detritus, and they have emerged as a tool to trace energy flow in wild systems. Because animals cannot synthesize EAAs de novo and must acquire them from dietary proteins, ecologists often assume δ13C fingerprints travel through food webs unaltered. Numerous studies have used this approach to quantify energy flow and multichannel feeding in animals, but δ13C fingerprinting has never been experimentally tested in a vertebrate consumer. We tested the efficacy of δ13C fingerprinting using captive deer mice Peromyscus maniculatus raised on diets containing bacterial, fungal and plant protein, as well as a combination of all three sources. We measured the transfer of δ13C fingerprints from diet to consumer liver, muscle and bone collagen, and we used linear discriminant analysis (LDA) and isotopic mixing models to estimate dietary proportions compared to known contributions. Lastly, we tested the use of published δ13C source fingerprints previously used to estimate energy flow and multichannel feeding by consumers. We found that EAA δ13C values exhibit significant isotopic (i.e. trophic) fractionation between consumer tissues and diets. Nevertheless, LDA revealed that δ13C fingerprints are consistently routed and assimilated into consumer tissues, regardless of isotopic incorporation rate. Isotopic mixing models accurately estimated the proportional diets of consumers, but all models overestimated plant‐based protein contributions, likely due to the digestive efficiencies of protein sources. Lastly, we found that δ13C source fingerprints from published literature can lead to erroneous diet reconstruction. We show that δ13C fingerprints accurately measure energy flow to vertebrate consumers across tissues with different isotopic incorporation rates, thereby enabling the estimation of multichannel feeding at various temporal scales. Our findings illustrate the power of δ13C fingerprinting for quantifying food web dynamics, but also reveal that careful selection of dietary source data is critical to the accuracy of this emerging technique.

Intercontinental analysis of temperate steppe stream food webs reveals consistent autochthonous support of fishes

“…Next, we ran a Bayesian isotope mixing model using the MixSIAR package in R (Stock et al, 2018) to estimate proportional resource use by consumers, using the same inputs as we did for the LDA. Bayesian mixing models can estimate proportional dietary contributions even in cases of complex, multichannel feeding (Manlick & Newsome, 2022). The MixSIAR model, which explicitly included site and species as nested factors, was set to run for 300,000 iterations (burn‐in = 200,000) on three parallel Monte Carlo Markov chains with a thinning interval of 100 using non‐informative (generalist) priors (Equ.…”

The coupling of green and brown food webs regulates trophic position in a montane mammal guild

“…Given the broad co‐occurrence of woody and herbaceous vegetation, the observed error in TP estimates may be unavoidable in C 3 ‐based forest ecosystems until linkages between energy channels are more proximately identified. Despite these intricacies, amino acid isotope analysis remains the best tool ecologists have to map and quantify multichannel linkages in wild systems (Manlick & Newsome, 2022; Steffan et al, 2017), a key step in disentangling food web complexity and informing models of ecosystem stability (Rooney et al, 2006; Wolkovich et al, 2014).…”

“…To quantify multichannel feeding, we used δ 13 C values from six EAAs: isoleucine (Ile), leucine (Leu), lysine (Lys), phenylalanine (Phe), threonine (Thr), and valine (Val). We combined measured values with published EAA δ 13 C data (Larsen et al, 2009; Appendix S1) and characterized δ 13 C fingerprints for sources using linear discriminant analysis (LDA), which maximizes separation of groups in multivariate space and can distinguish green from brown energy channels (Larsen et al, 2009; Manlick & Newsome, 2022). We trained the LDA using EAA δ 13 C values of the three source groups, used leave‐one‐out cross‐validation to assess classification accuracy, and then used the model to predict group membership for animal consumers.…”

“…To quantify the proportional assimilation of energy channels by consumers, we used Bayesian mixing models in the R package simmr version 0.4.5 (Parnell & Inger, 2016). Following Manlick and Newsome (2022), we parameterized source and consumer mixtures using LDA coordinates from δ 13 C fingerprinting and ran two separate models: (1) animals grouped by species to estimate proportional use of energy channels at the species level and (2) animals grouped by individual to estimate proportions for each consumer (Appendix S1: Section S1). To corroborate estimates from mixing models, we also quantified energy channeling with an independent LDA bootstrap using 10,000 iterations of our LDA model with random sampling of energy sources (Fox et al, 2019; Manlick & Newsome, 2022).…”

“…Using a montane small-mammal community as a model, we first measured the δ 13 C fingerprints of consumers and three potential energy channels (plants, fungi, bacteria). We then used Bayesian isotopic mixing models to estimate the proportional assimilation of green and brown energy by consumers from different functional groups (e.g., omnivore vs. insectivore; Manlick & Newsome, 2022). Lastly, we calculated TP for individual consumers via amino acid ECOLOGY δ 15 N analysis, and we used multiple linear regression to test the relationship between brown energy assimilation and TP.…”