Diet is an integral part of an animal’s ecology, and therefore obtaining accurate diet information is essential, but challenging, for any study species. In the age of next-generation sequencing, DNA barcoding is a powerful tool for diet determination. However, numerous factors such as DNA sequencing technique and animal digestive ability can introduce artifacts when estimating the relative abundance of dietary components. Here, we evaluate the accuracy of plant metabarcoding in quantifying the amount of creosote bush (Larrea tridentata) and juniper (Juniperus osteosperma) in the diet of desert woodrats (Neotoma lepida). To do this, we sequenced the P6 loop of the chloroplast trnL (UAA) intron amplified from feces of captive desert woodrats fed fresh, plant-based diets consisting of variable amounts of creosote bush and alfalfa-based rabbit chow and a constant amount of juniper. Relative read abundances were compared across dietary treatment groups to determine whether barcoding yielded accurate quantifications.
Results/Conclusions
For both plant species, content was both overestimated and underestimated within and among individual animals at each dietary level. For example, despite comprising a constant 15% of the diet, the relative abundances of juniper reads in the feces varied from 10-75% with no discernable pattern. Diet estimation was, however, improved when individuals were averaged across dietary treatment. Using plant metabarcoding techniques, we were able to detect all dietary components, including trace amounts of creosote that constituted only 1% of the diet. In addition, we did not detect any plant species not present in the diet. Differences in food intake, digestion, and amplification of dietary components may influence the accuracy of the approach. Our results indicate that while DNA metabarcoding can provide indicators of dietary composition, the use of relative abundances to estimate diet should be restricted to the population level, and interpreted with caution.