Diet of Andean Leaf‐Eared Mice (Phyllotis) Living at Extreme Elevations on Atacama Volcanoes: Insights From Metagenomics, DNA Metabarcoding, and Stable Isotopes Authors: Claudio Quezada‐Romegialli, Marcial Quiroga‐Carmona, Guillermo D'Elía, Chris Harrod, Jay F Storz Published: 2024 ABSTRACT ABSTRACTOn the flanks of > 6000 m Andean volcanoes that tower over the Atacama Desert, leaf‐eared mice (Phyllotis vaccarum) live at extreme elevations that surpass known vegetation limits. The diet of these mice in these barren, hyperarid environments has been the subject of much speculation. According to the arthropod fallout hypothesis, sustenance is provided by windblown insects that accumulate in snowdrifts (“aolian deposits”). Mice may also feed on saxicolous lichen or forms of cryptic vegetation that have yet to be discovered at such high elevations. We tested hypotheses about the diet of mice living at extreme elevations on Atacama volcanoes by combining metagenomic and DNA metabarcoding analyses of gut contents with stable isotope analyses of mouse tissues. Genomic analyses of contents of the gastrointestinal tract of a live‐captured mouse from the summit of Volcán Llullaillaco (6739 m) revealed an opportunistic but purely herbivorous diet, including lichens. Although we found no evidence of animal DNA in gut contents of the summit mouse, stable isotope data indicate that mice from elevations at or near vegetation limits (~5100 m) include a larger fraction of animal prey in their diet than mice from lower elevations. Some plant species detected in the gut contents of the summit mouse are known to exist at lower elevations at the base of the volcano and in the surrounding Altiplano, suggesting that they may occur at higher elevations beneath the snowpack or in other cryptic microhabitats.Keywords: Andes, aolian zone, high altitude, lichenivory, Llullaillaco, Puna de Atacama Introduction Extreme high‐elevation surveys of small mammals in the Central Andes have yielded live captures of numerous specimens of the Andean leaf‐eared mouse Phyllotis vaccarum at elevations at or above the elevational limits of vegetation (Storz et al. 2024). One specimen was captured at 6739 m (22,100 ft) above sea level on the summit of Volcán Llullaillaco, a stratovolcano in the Central Andes that straddles the Argentina–Chile border (Storz et al. 2020). This summit specimen far surpasses previous elevational records for wild mammals in the Andes and Himalayas. Documentation of active burrows of P. vaccarum at > 6100 m on the flanks of Llullaillaco and the discovery of desiccated cadavers (“mummies”) of Phyllotis on the summits of Llullaillaco and several neighboring > 6000 m volcanoes confirm that these mice inhabit extreme elevations well above the apparent limits of vascular plants (Halloy 1991; Steppan et al. 2022; Storz et al. 2023, 2024). Evidence that high‐elevation mice are living in an apparently barren world of rock, ice, and snow prompts numerous questions, perhaps none more basic than: What are they eating? In the perennial winter conditions that prevail at elevations > 6000 m, the scarcity of food poses a special physiological challenge for small endotherms like mice because of the energetic demands of thermoregulation. Moreover, leaf‐eared mice in the genus Phyllotis do not hibernate, so the energetic challenge of sustaining endothermy in cold, hypoxic conditions is especially acute (Storz, Scott, and Cheviron 2010; Storz and Scott 2019). It has been suggested that windblown arthropods and/or vegetation could provide a source of food for animals living at elevations that exceed the limits at which green plants grow (the “aolian zone”) (Swan 1961, 1992). According to this hypothesis, the transport of airborne nutrients from lower elevations sustains life on the upper reaches of a volcano like Llullaillaco similar to the way that the fallout of organic detritus from upper layers of the water column sustains life in the aphotic zone of the deep ocean. Aolian deposits of windblown arthropods along the lee edge of mountain summits and ridgelines can attract birds and other insectivorous animals that normally forage at much lower elevations (Antor 1995; Spalding 1979). If aolian deposits of windblown arthropods (“arthropod fallout”) help sustain populations of P. vaccarum at extreme elevations on Atacama volcanoes, we would expect arthropods to constitute a much larger part of their diet than at lower elevations where the species is mainly granivorous and frugivorous (Bozinovic and Rosenmann 1988; López‐Cortés et al. 2007; Sassi et al. 2017). On the Quinghai–Tibetan Plateau, high‐elevation plateau pikas ( Ochotona curzoniae ) exploit the feces of yak ( Bos grunniens ) as a food source (Speakman et al. 2021). If Phyllotis mice practice a similar form of interspecific coprophagy, feces from Andean camelids such as vicuña (Lama vicugna) and guanaco ( Lama guanicoe ) would provide the most readily available source. Although neither vicuña nor guanaco typically spend much time above the elevational limit of vegetation, both species are known to traverse mountain passes at elevations > 5500 m in the Central Andes (J. Storz, personal observation). It is also possible that the mice feed on lichen that grows on rock substrates (saxicolous lichen) or some cryptic form of vegetation that is not currently known to occur at such extreme elevations. Here, we test the above‐mentioned hypotheses by conducting metagenomic and metabarcoding analyses of gut contents from the world‐record specimen of P. vaccarum that was live captured on the summit of Llullaillaco (6739 m). Since the metagenomic approach involves high‐throughput sequencing of all DNA extracted from a sample without PCR enrichment of specific markers, it is not biased by a priori expectations about which taxonomic groups to expect and is therefore well suited to dietary assessments of omnivorous species (Chua et al. 2021). DNA metabarcoding complements the metagenomic approach and can be used to estimate the diversity and relative abundance of different items in the diet (Deagle et al. 2019; Stapleton et al. 2022). To complement the metagenomic and metabarcoding analysis of the summit specimen, we conducted a stable isotope analysis of liver samples from a larger sample of wild‐caught mice from a broad range of elevations in the Chilean Altiplano and Puna de Atacama (2370–6739 m). We used stable isotope values of three key elements (carbon, nitrogen, and sulfur) from liver tissue to characterize the diet of P. vaccarum over a time span of weeks to months. In the livers of small mammals, isotopic half‐lives are < 1 week for both carbon (δ13C) and nitrogen (δ15N) and we expect a similar half‐life for sulfur (δ34S). Examination of stable isotope values permits inferences about several key components of trophic ecology. Carbon stable isotopes (δ13C) reflect the relative consumption of food derived from different sources of primary production (Cerling et al. 1997; DeNiro and Epstein 1978). The stable isotope of nitrogen is typically used as an indicator of consumer trophic position (TP) (Vanderklift and Ponsard 2003; Quezada‐Romegialli et al. 2018) but can also be used to discriminate between consumption of food from distinct habitats (Harrod et al. 2005). For example, it should be possible to assess the extent to which mice rely on lichen at high elevations. Lichens are typically very 15N‐depleted relative to terrestrial plants (Fogel et al. 2008; Lee, Lim, and Yoon 2009; Pinho et al. 2017), and this holds true for lichens from high elevations (Biazrov 2012; Marris, Hawke, and Glenny 2019; Szpak et al. 2013) and volcanic fumeroles (Tozer et al. 2005). If lichen forms an important part of the diet of P. vaccarum at high elevations, we would expect to observe negative δ15N values. Sulfur stable isotopes (δ34S) are also useful indicators of consumer habitat use as values measured from plants and their consumers often exhibit high levels of spatial variation across biogeochemical gradients (Krouse, Stewart, and Grinenko 1991; Nielsen et al. 1991), as might be expected along the flanks of a historically active volcano like Llullaillaco. By combining metagenomics, metabarcoding, and stable isotope analyses, we tested several hypotheses about the diet of mice living at extreme elevations. The arthropod fallout hypothesis would be supported by the presence of DNA from insects or other arthropods that could be blown upslope, and isotopic estimates of TP would be higher for mice living at especially high elevations on the flanks or summit of the volcano in comparison with those from lower elevations in the surrounding Altiplano. Interspecific coprophagy (or scavenging) would be supported by the presence of DNA from vicuña, guanaco, or other codistributed mammals. Lichenivory would be supported by the presence of DNA from lichen‐associated fungi or green algae and especially low 15N values in mice from high elevations. The particular plants detected in the gut contents of the summit mouse may suggest that some plant species actually occur at much higher elevations than currently assumed, but they may be sparsely distributed in cryptic microhabitats (e.g., in rock crevices or under the snowpack). If that is the case, then the diet of mice living at > 6000 m may include a subset of the same plants that mice feed on at lower elevations. Materials and Methods Sampling We live captured all mice using Sherman live traps and other methods described in Storz et al. (2020, 2024). We collected mice from a broad range of elevations in the Altiplano/Puna de Atacama ecoregions, from 2370 m in the Atacama Desert to the 6739 m summit of Volcán Llullaillaco (Figure 1). We sacrificed mice in the field, prepared them as museum specimens, and preserved liver tissue in 95‰ ethanol as source material for the stable isotope analysis. For the P. vaccarum specimen captured on the summit of Volcán Llullaillaco, we preserved the entire gastrointestinal tract in ethanol as a source of DNA for metagenomic and metabarcoding analyses. All mouse specimens are housed in the Colección de Mamíferos of the Universidad Austral de Chile, Valdivia, Chile. Sampling of Phyllotis vaccarum across an elevational gradient in the Altiplano and Puna de Atacama of northern Chile, Regiόn de Antofagasta. (a) Map of seven collection localities on the flanks of Volcán Lullaillaco and the surrounding Altiplano. (b) Elevational profile of sampling transect, with sampling Localities 1 to 7 shown in ascending order of elevation, from 2370 m (site 1) to 6739 m (the summit of Volcán Lullaillaco, Site 7). (c) Northwest face of Volcán Llullaillaco (24°43.21′ S, 68°32.22′ W). Photo was taken from a point several km northwest of Site 3 on the transect. Photo: J.F. Storz. We collected all mice in accordance with permissions to JFS and GD from the following Chilean government agencies: Servicio Agrícola y Ganadero (SAG, Resolución Exenta # 6633/2020), Corporación Nacional Forestal (CONAF, Autorización # 171219), and Dirección Nacional de Fronteras y Límites del Estado (DIFROL, Autorización de Expedición Científica #68/2020). We handled all mice in accordance with protocols approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Nebraska (Project ID: 1919). Dissection of Gastrointestinal Tract For the mouse captured on the summit of Volcán Llullaillaco (UACH8291), we extracted DNA from contents of the stomach for metagenomic sequencing and DNA metabarcoding. We also dissected the lower gastrointestinal tract into 13 adjoining sections, the cecum and 12 consecutive segments of the colon, ordered from the outlet of the cecum to the rectum (Figure 2), and we extracted DNA from the contents of each section for additional DNA metabarcoding analysis. This approach allowed us to examine temporal changes in the mouse's diet, as determined by gut passage times: the stomach and cecum contain food items ingested within a few hours of its capture, whereas the colon sections and rectum potentially contain food ingested within the previous 2 or 3 days. The metagenomic sequencing represents an unbiased approach to characterize the stomach contents of the mouse, whereas the metabarcoding analysis is designed to test specific hypotheses about the animal's diet (arthropod fallout, interspecific coprophagy, lichenivory, or herbivory at elevations that surpass assumed vegetation limits). Schematic figure of the dissected portions of the gastrointestinal tract of the mouse from the summit of Volcán Llullaillaco. Contents of the upper digestive tract (stomach) were analyzed via metagenomics and DNA metabarcoding. Contents of the lower digestive tract (separated into the cecum and 12 consecutive segments from the outlet of the cecum to the rectum) were analyzed via DNA metabarcoding. Metagenomic Sequencing We sent contents of the dissected sections of the gastrointestinal tract to Azenta Life Sciences (South Plainfield, NJ, USA) for metagenomic analysis. Genomic DNA was isolated using the NucleoMag DNA Microbiome Kit (Takara Bio, Shiga, Japan) and was quantified using a Qubit 2.0 Fluorometer (ThermoFisher Scientific, Waltham, MA, US). NEBNext Ultra II DNA Library Prep Kit for Illumina (New England Biolabs, Ipswich, MA, USA) was used for library preparation following manufacturer's recommendations. Briefly, genomic DNA was fragmented by acoustic shearing with a Covaris S220 instrument, followed by end repair. Adapters were ligated after adenylation of the 3′ ends followed by enrichment by a limited cycle PCR. DNA libraries were quantified using Qubit 2.0 Fluorometer and by real‐time PCR (Applied Biosystems, Carlsbad, CA, USA). Sequencing libraries were sequenced on an Illumina HiSeq instrument using a 2 × 150 bp paired‐end (PE) configuration. Image analysis and base calling were conducted using the HiSeq Control Software (HCS). Raw BCL files were converted to FastQ files and demultiplexed using bcl2fastq v.2.1.9 (Illumina), keeping only > Q30 reads with 150 bp in length. A de novo approach was followed for assembling reads using Spades v3.10 (Bankevich et al. 2012), with a minimum contig length of 1000 bp and using the newly assembled genome of Phyllotis vaccarum (Storz et al. 2023) as a reference genome to discard reads aligned to the host. QUAST (Gurevich et al. 2013) was used to generate statistics, and EMBOSS tools getorf was used to find the open reading frames within the de novo assembled genome. As the basic local alignment search tool (BLAST) might give spurious identification, we used BLAST+ (v.2.6.0) (Altschul et al. 1990) to query assembled contigs in the nucleotide database of GenBank (analyzed on July 22, 2022), considering the first 25 hits of at least 90% similarity. Then, the basic sequence taxonomy annotator (BASTA) was used to assign the most recent common ancestor of all hits per contig following the widely used last common ancestor (LCA) approach. Briefly, accession numbers of BLAST hits are mapped against the taxonomies of genomes published by the National Center for Biotechnology Information (NCBI), which are then resolved to a seven‐level taxonomy. Then, BASTA creates a tree structure of taxonomies of the 25 hits and returns the taxonomic level that is shared by all of the hit sequences. This approach allows us to confirm each taxonomic assignment, especially below the family level. DNA Metabarcoding Analysis and Primer Selection The stomach, cecum, and 12 consecutive segments of the colon were sent to MrDNA (www.mrdnalab.com) for DNA extraction and metabarcoding analysis. We identified and discarded false positives using extraction blanks for all primer combinations and we used four PCR replicates to avoid noise from spurious amplification (Taberlet et al. 2018, Table 1). PCR replicates for each sample of the dissected sections were merged separately before library preparation, allowing a better representation of DNA in each section of the lower gastrointestinal tract. We used the following primer pairs for specific taxonomic groups (Table 1): for plants, we amplified (i) the P6 loop of the chloroplast trnL (UAA) intron using primers P6‐trnLF: 5′‐GGG CAA TCC TGA GCC AA‐3′ and p6‐trnLR: 5′‐CCA TTG AGT CTC TGC ACC TAT C‐3′ (Taberlet et al. 2007), and (ii) the internal transcribed spacer 2 (ITS2) of nuclear ribosomal DNA using primers S2F: 5′‐ATGCGATACTTGGTGTGAAT‐3′ and S2R: 5′‐GACGCTTCTCCAGACTACAAT‐3′ (Chen et al. 2010); for eukaryotic algae and cyanobacteria, we amplified Domain V of the 23S plastid rRNA gene using primers p23SrV_f1 5′‐GGA CAG AAA GAC CCT ATG AA‐3′ and p23SrV_r1 5′‐TCA GCC TGT TAT CCC TAG AG‐3′ (Sherwood and Presting 2007); for fungi (Ascomycota and Basidiomycota), we amplified the internal transcribed spacer (ITS) of nuclear ribosomal DNA using primers: ITS1‐F 5′‐CTT GGT CAT TTA GAG GAA GTA A‐3′ (Gardes and Bruns 1993) and 5′‐GCT GCG TTC TTC ATC GAT GC‐3′ (White et al. 1990); for metazoans, we amplified the cytochrome c oxidase subunit I using primers: mlCOIintF: 5′‐GGW ACW GGW TGA ACW GTW TAY CCY CC‐3′ (Leray et al. 2013) and jgHCO2198: 5′‐TAI ACY TCI GGR TGI CCR AAR AAY CA‐3′ (Geller et al. 2013); and for invertebrates, we amplified the cytochrome c oxidase subunit I using primers: fwhF2: 5′‐GGD ACW GGW TGA ACW GTW TAY CCH CC‐3′ (Vamos, Elbrecht, and Leese 2017) and EPTDr2n: 5′‐CAA ACA AAT ARD GGT ATT CGD TY‐3′ (Leese et al. 2021). TABLE 1Summary statistics for metabarcoding sequence reads for each of the dissected portions of the gastrointestinal tract of the Phyllotis vaccarum specimen captured on the summit of Volcán Llullaillaco (see Figure 2).SampleInput readsFilteredDenoised forwardDenoised reverseMergedNonchimericExtraction blanks2.0091.1731.1501.1421313Stomach755.758350.866349.765349.840242.553237.567Cecum742.138321.525320.278319.642229.464226.427C1548.838283.847283.023282.946213.947204.749C2745.252370.547369.633369.303271.283261.741C3634.524324.615323.596323.317226.881224.844C4551.463254.049253.346253.147206.205203.786C5559.460272.415271.609271.402159.567158.241C6624.732301.630300.660300.059193.624188.859C7670.502313.308311.935311.397214.056205.772C8760.648397.214395.953395.543287.349282.642C9731.646367.289365.790365.438275.122269.431C10658.965318.646317.186316.856215.254212.705C11702.488340.687339.058338.399224.913219.703C12798.662390.642388.897388.447304.740300.431Total9487.0854608.4534591.8794586.8783264.9713196.911 Summary statistics for metabarcoding sequence reads for each of the dissected portions of the gastrointestinal tract of the Phyllotis vaccarum specimen captured on the summit of Volcán Llullaillaco (see Figure 2). Bioinformatic Processing—Metabarcoding Reads were analyzed separately for each primer in R v 4.3.1 (R Core Team 2023) using RStudio 2023.12.1 (RStudio Team 2023) in dada2 (Callahan, McMurdie, et al. 2016; Callahan, Sankaran, et al. 2016) after filtering (maxEE = 2, Q scores > 30; Edgar and Flyvbjerg 2015) and discarding reads with Ns. Error model calculation, noise reduction, read correction, read merging, and removal of chimeric sequences were performed using default settings (Callahan, McMurdie, et al. 2016). Briefly, first low‐quality sequencing reads were eliminated and reads were trimmed to a consistent length as most Illumina sequencing data show a decrease in quality at the ends of sequencing reads. Then, amplicon sequence variants (ASVs) were inferred instead of clustering sequences into operational taxonomic units (OTUs), as ASVs are exact genetic variants that may differ by a single nucleotide without imposing any arbitrary threshold (Callahan, McMurdie, et al. 2016; Callahan, Sankaran, et al. 2016). The DADA2 procedure was then used to dereplicate sequencing reads to remove redundancy. This approach applies an unsupervised learning procedure that distinguishes sequencing errors from true allelic variation (Callahan, Sankaran, et al. 2016), merges inferred forward and reverse sequences, and removes paired sequences that do not perfectly overlap. Finally, chimeras were removed and ASVs identified by DADA2 were assigned to taxa of origin using BLAST+ (v.2.6.0) (Altschul et al. 1990) and BASTA (Kahlke and Ralph 2018) following the same approach described for the metagenomic analysis. For Fungi identification, taxonomic assignments were checked also against UNITE database (Abarenkov et al. 2024) following standard mothur pipeline of taxonomic assignment (Schloss et al. 2009). Stable Isotope Analysis Ethanol‐preserved liver tissues were rinsed in distilled water and freeze dried for 48 h. Once freeze dried, samples were ground to a fine powder using a laboratory bead beater. Ground samples were weighed in 8 × 5 mm pressed standard‐weight tin capsules using a high‐precision microbalance (repeatability = 0.0008 mg). Elemental percentages of carbon, nitrogen, sulfur, and stable isotope ratios (δ13C, δ15N, and δ34S) were measured using a Pyrocube elemental analyzer (Elementar, Langenselbold, Germany) linked to a visION continuous‐flow isotope ratio mass spectrometer (Elementar, Langenselbold, Germany) at the Universidad de Antofagasta Stable Isotope Facility (UASIF), Chile. Stable isotope ratios are expressed using δ notation and are reported in units per mil (‰) relative to the following standards: Vienna Pee Dee Belemnite for carbon, air for nitrogen, and Vienna Canyon Diablo Troilite for sulfur. International standards were used in each batch to provide a multipoint calibration using the ionOS software package v4.1.005 (Elementar, Langenselbold, Germany). Certified reference materials USGS40 and USGS41a were used for carbon and nitrogen and IAEA‐SO‐5, IAEA‐SO‐6, and IAEA‐S2 for sulfur. Repeated analysis of standards showed analytical errors (±1 SD) of ±0.04‰ for δ13C, ±0.06‰ for δ15N, and ±0.6‰ for δ34S. We used two calibration standards, (a) sulfonamide (Elementar, Germany) and (b) an in‐house standard (rainbow trout dorsal muscle), to correct for instrument drift. Statistical Analyses—Stable Isotopes Liver is commonly used as a lipid storage organ in vertebrates and liver lipid content can vary significantly among individuals according to variations in nutritional state and physiological condition. Lipids formed through de novo biosynthesis are isotopically lighter in δ13C values compared to proteins and the dietary sources from which they were formed (DeNiro and Epstein 1977). If the isotopic effect of these lipids is not accounted for, they can affect assessment of consumer δ13C values. Furthermore, as the livers analyzed here were preserved in ethanol, they may have undergone some partial uncontrolled lipid extraction prior to analysis. Variation in individual lipid content can affect comparisons of δ13C values and it is therefore common to use chemical treatments to remove lipids prior to stable isotope analyses. However, the chemical treatment can affect estimated values of other stable isotopes from the same sample. Another possible solution is to use an arithmetic correction that relies on a predictable relationship between lipid content and the elemental ratio between carbon and nitrogen (C:N) in the sample (Kiljunen et al. 2006; Logan et al. 2008). Javornik et al. (2019) found small effects of ethanol storage on δ13C values in mammalian liver but reported no preservation effects on δ15N or δ34S values. Javornik et al. (2019) reported that C:N values decreased after ethanol storage but suggested that lipid‐free 13C values could be reliably estimated mathematically from the C:N ratio. In our samples, liver C:N ratios varied considerably (range = 3.2–5.7, mean ± SD = 3.8 ± 0.5, n = 41). Liver C:N ratios were lower in mice captured at higher elevations (r = −0.36, n = 40, p = 0.024). As there was also a negative relationship between C:N and δ13C within samples from the same collection locality, we estimated lipid‐corrected δ13C values using equation (1a) from Logan et al. (2008), resulting in a mean (±SD) isotopic shift of 1.1‰ ± 0.6‰. All δ13C data that we report are lipid corrected, but liver δ15N and δ34S data are shown without correction. Collection sites spanned ~4400 m of elevation and therefore exhibited considerable variation in vegetation cover and plant species composition. We therefore examined how stable isotopes varied within the dataset by plotting each of the stable isotopes against elevation. We then used PERMANOVA (nonparametric permutation‐based equivalent of ANOVA) to examine whether stable isotope values varied among capture sites. Since we collected a single individual from Site 7 (the summit of Llullaillaco), it was not included in these comparisons. Although PERMANOVA is typically used for multivariate comparisons (MANOVA), it can also be used to make robust univariate comparisons. Finally, to assess the ability of the stable isotope analysis to assign mice to capture location, and to identify mismatches that may be indicative of recent dispersal, we used canonical analysis of principal coordinates (CAP), a distance‐based equivalent of discriminant function analysis (Anderson and Willis 2003). This approach uses multivariate data (e.g., δ13C, δ15N, and δ34S values) to discriminate between groups defined by elevation of capture sites. This approach also allowed us to infer the possible origin of the summit mouse from Volcán Llullaillaco (Site 7). We grouped mice in bins based on their capture elevation (2000–3000, 3000–4000, 4000–5000, and > 5000 m) and we used δ13C, δ15N, and δ34S as dependent variables. We used a leave‐one‐out classification approach to examine relative classification success, and we then used the model to identify the elevational range that provided the best match to values from the Llullaillaco summit mouse. The ability of the CAP model to statistically discriminate between groups was estimated via permutation (n = 9999). PERMANOVA and CAP were both run in the PERMANOVA+1 add‐on to PRIMER 7 (Anderson, Gorley, and Clarke 2008; Clarke and Gorley 2015). We estimated the TP of P. vaccarum at each site using liver δ15N values with those of primary producers collected across a similar (but truncated) elevational range (Díaz et al. 2016). This approach (Cabana and Rasmussen 1996) allows the indirect calculation of consumer TP: TP = λ + (δ15NConsumer − δ15NBaseline)/TDF, where λ is the TP of the baseline taxon, δ15NConsumer is the nitrogen isotopic value of mice at a given site, δ15NBaseline is the nitrogen isotopic value of the baseline at that site, and TDF is the mean ± SD nitrogen trophic discrimination factor (TDF) for mouse liver (here, we use 4.3‰ ± 0.2‰ from Arneson and MacAvoy (2005)). We used plants as our baseline (λ = 1) based on data from Díaz et al. (2016), which were collected in the same region as our study, over an elevational range of 2670–4480 m. Plant δ13C and δ15N values exhibited considerable variation across elevations (Figure 6), and we placed plants into broad elevational intervals (2000–3000, 3000–4000, 4000–5000, and > 5000 m). We then used values from the closest elevational interval to estimate mouse TP at each capture site using tRophicPosition 0.8.0 (Quezada‐Romegialli et al. 2018) in R 4.2.3 (López‐Cortés et al. 2007; R Core Team 2023). Briefly, tRophicPosition uses a Bayesian approach to estimate TP for a population of consumers while accounting for variation in consumer and baseline isotope values. For most sites, we use the onebaseline model (assuming a single baseline), but we used the twoBaselines model for mice from Site 2. This is because the plants from the 3000–4000 m interval showed a bimodal distribution of δ13C values, which indicates the presence of plants using different photosynthetic pathways (e.g., C3 and C4/CAM). Since these groups also showed evidence for a nonnormal distribution of δ15N values, we used the twoBaselines full model, which also uses baseline δ13C. For all model runs, we used the following parameters: chains = 3, number of adaptive iterations = 1000, iterations = 20,000, burn‐in = 1000, and thinning = 10. In case of the summit mouse from Volcán Llullaillaco, we developed an individual model to calculate TP, as tRophicPosition v 0.8.0 currently provides only population‐level estimates of TP. This new model with a one baseline approach was implemented in greta (Golding 2019), which allows the calculation of TP at the individual level. We modeled the baseline for the summit mouse as having a mean and standard deviation of δ15N values of plants > 5000 m with a normal distribution for the mean and a Cauchy distribution for the SD, with a location of plants δ15N SD on a scale of 3 and truncated from 0 to infinite. In this analysis, λ is 1, and the TDF was modeled as having a normal distribution with a mean of 4.3 and SD of 0.2 (Arneson and MacAvoy 2005). We calculated 10,000 samples, with a thinning of 1000 samples as warm‐up and 16 chains. Variation in plant δ13C (A) and δ15N (B) across different elevational zones. Values were taken from Díaz et al. (2016). Note the elevational shift in δ13C values showing dominance of C4 plants at lower elevations, a mix of C3, C4, and CAM plants at mid‐elevations and a shift to C3 plants at higher elevations. Plant δ15N values were similar at lower and mid‐elevations but were relatively 15N depleted at higher elevations. These data were used to define isotopic baselines for estimates of trophic position of P. vaccarum. Due to the selective retention of heavier isotopes during the assimilation of food, consumers are typically isotopically “heavier” than their food (DeNiro and Epstein 1978, 1981). These diet‐tissue shifts are referred to as trophic discriminations and are typically estimated in experimental settings. Arneson and MacAvoy (2005) provided empirical estimates for TDFs in liver from groups of laboratory mice fed diets that differed in the origin of their protein and carbohydrate components. In their control diet, where carbohydrates and proteins originated from the same source, mean ± SD TDFs were 0.7‰ ± 0.3‰ for carbon (Δ13C), 4.3‰ ± 0.2‰ for nitrogen (Δ15N), and −2.1‰ ± 0.1‰ for sulfur (Δ34S). As such, we expect mouse livers to have δ13C and δ15N values that are ~1% and ~4% higher, respectively, than their long‐term average, in combination with δ34S values ~2% lower than the long‐term average. Results and Discussion Metagenomics For the stomach DNA sample of the Llullaillaco summit mouse, we sequenced a total of 423,477,275 reads, yielding 127,043 Mbases, with 92.48% of reads ≥ q30 and a mean quality score of 35.69. We assembled a total of 9138 contigs ≥ 1000 bp in length (21,188,348 bp), with a maximum length of 103,345 bp. Of 991 contigs that were identified at the order level and above, the vast majority were assigned to super kingdom bacteria (699 contigs), with Proteobacteria (563 contigs) and Firmicutes (99) as the dominant phyla. Only 1.3% of contigs (n = 13) were assigned to plants (clade Streptophyta, class Magnoliopsida [= dicotyledons]), and all were assigned to a single representative of the coca family, Erythroxylaceae (Erythroxylum novagranatense). This shrub species is widely cultivated in South America because its leaves are a rich source of the psychoactive alkaloid, cocaine. In the stomach contents of the summit mouse, we detected no traces of DNA from arthropods or from vicuña, guanaco, or other potentially codistributed Andean mammals. Metabarcoding We sequenced a total of 9,487,085 reads as part of the DNA metabarcoding analysis, maintaining 3,196,911 nonchimeric reads after filtering, denoising, and merging reads for all primer combinations and all samples (Table 1, Tables S1–S6). The P6 loop of the chloroplast trnL (UAA) intron marker (Figure 3A,B) yielded an average of 56,507 ± 22,886 (SD) nonchimeric processed reads for each of the dissected sections of the gastrointestinal tract. Consistent with results of the metagenomic analysis, sequences assigned to the family Erythroxylaceae (Erythroxylum novagranatense) predominated in samples from each section, from the stomach to the C12 portion of the colon (Figure 3A). Sequences of E. novagranatense represent 97.3% of the 886,078 sequences derived from all surveyed sections of the gastrointestinal tract. In addition to representatives of Erythroxylaceae, we detected representatives of Amaryllidaceae (0.4% of the total), Poaceae (0.3%), Pinaceae (0.3%), and Malvaceae (0.1%), and Fabaceae and Juglandaceae, which together accounted for 0.05% of the total (Figure 3B). Taxonomic composition of reads identified through metabarcoding for the stomach, cecum, and 12 consecutive portions of the lower GI from the outlet of the cecum to the rectum. (A) Proportion of sequence reads per taxon for the marker trn and (B) abundance of total reads for the same marker with Erythroxylaceae excluded. (C) Proportion of sequence reads per taxon for the marker ITS2 and (D) abundance of total reads for the same marker with Erythroxylaceae excluded. (E) Proportion of sequence reads per taxon for marker 23S and (F) abundance of total reads for the same marker with Erythroxylaceae excluded. The ITS2 marker yielded an average of 19,952 ± 7762 nonchimeric reads for each section of the gastrointestinal tract, and Erythroxylaceae accounted for 87.2% of the 279,336 total reads. Sequences from Pinaceae (6.1% of the total), Amaryllidaceae (5.9%), Urticaceae (0.64%), and Poaceae (0.15%) were also detected, along with traces of Laureaceae, Fabaceae, Ranunculaceae, and Hypericaceae, which together accounted for 0.04% of total reads (Figure 3D). The third marker that provided information for Streptophyta was Domain V of the 23S plastid rRNA gene which yielded 1596 ± 671 reads on average for all sections of the gastrointestinal tract. Again, Erythroxylaceae predominated, accounting for 98.2% of 22,352 total reads (Figure 3E). In addition to Erythroxylaceae and Amaryllidaceae (1.5% of total reads), the next two most abundant taxa were Cyatheaceae and Brassicaceae, which together accounted for < 0.6% of total reads (Figure 3F). For Fungi, the ITS of the nuclear ribosomal DNA marker detected a wide variety of Ascomycota and Basidiomycota orders and families (Figure 4) in the 414,990 read total, with an average of 29,642 ± 12,846 reads for each section of the gastrointestinal tract. For Ascomycota, the most abundant families were Cladosporiacea (28.4% of the total reads for this Phylum), Pleosporaceae (20.2% of the total), Sacharomycetaceae (16.6% of the total), and Nectriaceae (14.1% of the total), whereas the remaining 24 families represent 20.6% of the total, with Phaeococcomycetaceae and Parmeliaceae (lichen‐associated families) accounting for 1.6% of the total. For the Phylum Basidiomycota, the orders Agaricales (40.7% of total reads for this phylum) and Polyporales (36.5% of the total) represent the most abundant groups, whereas Psathyrellaceae (18.8% of the total), Polyporaceae (13.6% of the total), Agaricaceae (12.0% of the total), and the families Meripilaceae, Fomitopsidaceae, and Hyphodermataceae (together accounting for 16.7% of the total) were the most abundant groups across all sections of the gastrointestinal tract. Taxonomic composition of Ascomycota (A) and Basidiomycota (B) at the order and family levels. For Metazoans, the marker cytochrome c oxidase subunit I did not identify ASVs derived from vicuña, guanaco, or other mammals. The marker cytochrome c oxidase subunit I specifically developed for arthropods did not detect any ASV for the group. Diet of the Summit Mouse One of the most puzzling results of the metagenomic and metabarcoding analyses is the predominance of Erythroxylaceae (the coca family) and Amaryllidaceae (the garlic family) in the stomach, cecum, and all 12 of the independently analyzed sections of the colon (Figure 3). Two representatives of Erythroxylaceae, Erythroxylum argentinum and E. cuneifolia , exist at elevations below ~2000 m far to the east of Llullaillaco, but they do not occur in Andean Desert or dry puna habitats. Coca was widely used throughout the Incan Empire and is still used in indigenous Quechua and Aymara communities and, occasionally, by mountain climbers. Upon summiting a particular peak, there is a custom (especially among Argentine climbers) of leaving offerings to Pachamama, an Andean “Earth mother”, Gaia‐type deity. Such offerings are left at the base of rock piles called “apacheta” that serve as summit markers. A typical offering to Pachamama is a sprinkling of coca leaves or a small bag of such leaves at the base of the apacheta. This custom provides a ready explanation for the predominance of Erythroxylaceae in the gut contents of the summit mouse, which had presumably encountered just such an offering on the summit of Llullaillaco. The presence of garlic in the stomach contents of the summit mouse has a similar explanation. In the Argentine Province of Salta (where the western portion of Llullaillaco is located), garlic is a traditional folk remedy for altitude sickness. Argentine climbers are known to chew cloves of garlic during their ascent. As is the case with any unchewed coca that climbers possess upon reaching the summit, it is also customary to leave leftover cloves of garlic at the base of the Summit Apacheta. The predominance of both coca and garlic in the gut contents of the Llullaillaco summit mouse suggests that climbers' offerings to Pachamama on the summits of high Andean summits may sometimes serve as unintentional offerings to opportunistic Phyllotis mice living in an extremely food‐scarce environment. Aside from Erthroxylaceae and Amaryllidaceae, we also detected DNA representative of several plant families such as Fabaceae, Malvaceae, and Poaceae that occur at high elevations at the base of Volcán Llullaillaco and in the surrounding Altiplano (Arroyo et al. 1988, 1998; Luebert and Gajardo 1999; Marticorena et al. 2004). Within Fabaceae, the herb Astragalus pusillus was documented at elevations up to 4300 m on the flanks of Llullaillaco (Marticorena et al. 2004). Within Malvaceae, several perennial herbs such as Cristaria andicola, Nototriche auricoma, and N. clandestina occur at elevations between 4000 and 4500 m (Arroyo et al. 1998) and Cristaria andicola was documented as the most abundant plant species in the diet of Phyllotis at another altiplano study site in northern Chile (López‐Cortés et al. 2007). Within Poaceae (= Gramineae), bunch grasses in the genus Calamagrostis (recognized as Deyeuxia in Arroyo et al. 1998 and Luebert and Gajardo 1999) occur above 4000 m (Marticorena et al. 2004) and are also known to be included in the diet of Phyllotis from the Chilean Altiplano (López‐Cortés et al. 2007). Although all of these plants seem plausible as potential sources of food for the summit mouse, there are no records of vascular plants or other vegetation above ~5000 m on the flanks of Llullaillaco (Arroyo et al. 1988, 1998; Luebert and Gajardo 1999; Marticorena et al. 2004; Storz et al. 2024; Vimercati et al. 2019). However, it is also true that botanical surveys typically do not venture above such elevations, so we should be cautious about interpreting absence of evidence as evidence of absence. Given that sequences representative of Fabaceae, Malvaceae, and Poaceae were detected in the gut contents of the mouse captured at 6739 m, far above the apparent elevational limits of those plant taxa, there are three possible explanations to consider: (1) plant material is carried upslope by the wind and accumulates in sufficient quantities on the lee edge of ridge lines and snowdrifts to provide a source of sustenance for high‐elevation mice (“Aolian deposits”; Antor 1995; Spalding 1979; Swan 1961, 1992); (2) the plants in question actually occur at much higher elevations than previously thought (although they may be scarce and cryptic); or (3) the mouse was not a full‐time summit resident, but rather a transient sojourner that had consumed the plant material at or near the base of the mountain some days prior to its capture. The former two hypotheses cannot be rejected, since few systematic plant surveys have been performed on Llullaillaco or other > 6000 m volcanoes (Storz et al. 2024). In assessing the plausibility of the third hypothesis, it is important to note that the ~1.6 km elevational distance between the summit of Llullaillaco (6739 m) and the apparent vegetational limit (~5000–5100 m; Storz et al. 2024) translates into a linear distance of ~5 km from any side of the volcano. Summiting the volcano from the vegetation limit is roughly equivalent to a direct‐from‐basecamp ascent, a feat that only the most elite mountain climbers could accomplish in a single day. We cannot rule out the possibility that mice undergo upslope/downslope dispersal on a seasonal basis, but such movements could certainly not occur on a daily basis. Moreover, in addition to the live capture of the P. vaccarum specimen on the Llullaillaco summit at 6739 m, video records and identification of active burrows of P. vaccarum between 6145 and 6205 m on the same volcano, and the discovery of mummies and skeletal remains of numerous P. vaccarum on the summits of four different > 6000 m volcanoes in the same mountain chain (Halloy 1991; Steppan et al. 2022; Storz et al. 2020, 2023, 2024) provides a consilience of evidence suggesting that these extreme high‐elevation mice are representative of resident populations. We think it is more likely that potential food plants exist at higher elevations, although they must be scarce and patchily distributed. The plausibility of this hypothesis is supported by the surprising discovery of bryophytes growing in association with active volcanic fumaroles near the summit of Volcán Socompa (Halloy 1991), a 6051 m volcano located 47 km northeast of Llullaillaco along the Argentina–Chile border. Primer set ITS detected sequences from families of two lichen‐associated fungi, Phaeococcomycetaceae and Parmeliaceae (phylum Ascomycota), indicating that Phyllotis feeds on saxicolous lichen, as has been documented in other arctic and alpine mammals during periods of food scarcity (Conner 1983; Seaward 2008; Richardson and Young 1977). During the Arctic winter, cricetid rodents such as snow voles ( Chionomys nivalis ) and northern bog lemmings ( Synaptomys borealis ) feed on tundra lichen (Richardson and Young 1977). Similarly, during winter months in the high alpine, North American pikas ( Ochotona princeps ) feed on lichens under the snowpack (Conner 1983). Terricolous, arboreal, and saxicolous lichens are an important component of the winter diet of Caribou ( Rangifer tarandus ) in the northern Holarctic (Seaward 2008) and arboreal lichens are an important component of the winter diet of Yunnan snub‐nosed monkeys ( Rhinopithecus bieti ) in montane coniferous forests at elevations > 4000 m (Kirkpatrick et al. 1998). Although lichens may serve as a seasonal or short‐term supplement to the normal diet of many mammals living in arctic and alpine environments, their low nutritive value suggest that they are unlikely to represent a year‐round dietary staple for small mammals like Phyllotis that have high metabolic demands. We found no strong support for the arthropod fallout hypothesis, as we did not detect arthropod DNA in the gut contents of our summit mouse. In contrast to high‐elevation pikas on the Quinghai–Tibetan Plateau that feed on yak feces, we found no evidence for interspecific coprophagy in our high‐elevation Phyllotis, as indicated by the absence of metagenomic sequence reads and COI barcodes matching vicuña, guanaco, or any other potentially codistributed mammals. The absence of such sequences also constitutes absence of evidence for scavenging. Stable Isotope Analysis Mice showed considerable variation in values of all three of the stable isotopes examined (Figure 5A–C, Table 2). Patterns included isotopic variation both among sites (indicating an elevational effect) and within sites (indicating individual variation in foraging habits). In the total dataset, δ13C values of P. vaccarum (Figure 5A, Table 2) varied between −22.8‰ and −12.7‰, with more 13C‐enriched values being recorded at lower elevations (< 4000 m). However, mice with relatively 13C‐depleted values were captured at both Sites 1 and 2. There was strong statistical support for intersite differences in δ13C (PERMANOVA: PseudoF 5,34 = 36.6, P9999 perms = 0.0001). Post hoc comparisons indicated significant differences (p < 0.05) between δ13C values for all sites apart from Sites 3 and 4 (t = 1.42, p = 0.13), Sites 3 and 6 (t = 1.92, p = 0.05), and Sites 4 and 5 (t = 1.4, p = 0.14). Phyllotis vaccarum liver δ15N ranged between 6.2‰ and 22.8‰ (Figure 5B, Table 2), with considerable variation among sites (PseudoF 5,34 = 39.3, p 9999 perms = 0.0001). Values were notably 15N‐enriched at Site 1, but included one individual with relatively low δ15N. Post hoc comparisons showed overlap in δ15N values at Sites 2 and 6 (t = 0.10, p = 0.92), Sites 3 and 4 (t = 0.05, p = 0.97), Sites 3 and 6 (t = 2.27, p = 0.05), and Sites 4 and 5 (t = 0.97, p = 0.34). In the cases of both δ13C and δ15N, P. vaccarum showed a similar pattern of 13C‐ and 15N‐enriched values at sites from lower elevations (< 4000 m) relative to individuals captured between 4000 and 5000 m (Figure 6, Table 2). This contrasted with the pattern in δ34S (Figure 5C, Table 2) where P. vaccarum showed less variation in general, with values ranging between −2.5‰ and 2.4‰, but exhibited a positive shift between 34S‐depleted values at sites < 4000 m and 34S‐enriched values > 4000 m. Values of δ34S for mouse livers varied among sites (PseudoF 5,34 = 25.3, p 9999 perms = 0.0001), although variation was lower than that observed for C and N. Post hoc comparisons indicated that δ34S values were similar for mice captured from Sites 1 and 2 (t = 0.93, p = 0.70), Sites 3 and 5 (t = 1.39, p = 0.18), Sites 3 and 6 (t = 0.67, p = 0.52), Sites 4 and 5 (t = 1.35, p = 0.21), Sites 4 and 6 (t = 1.8, p = 0.09), and Sites 5 and 6 (t = 0.84, p = 0.45). Variation in stable isotopes in livers of Phyllotis vaccarum sampled from different elevational zones: (A) δ13C, (B) δ15N, and (C) δ34S. (D) Results of multivariate CAP ordination based on Euclidean distances calculated from combined δ13C, δ15N, and δ34S values. TABLE 2Summary statistics (n, mean ± SD) for Phyllotis vaccarum liver stable isotope analyses of carbon (values shown for δ13C and lipid‐corrected δ13C), nitrogen (δ15N), and sulfur (δ34S) and the elemental C:N ratio.SiteElevation (m) n δ13C (‰)Lipid‐corrected δ13C (‰)δ15N (‰)δ34S (‰)C:NSite 123708−18.3 (±2.2)−16.9 (±2.0)19.2 (±3.6)−1.6 (±0.8)4.2 (±0.8)Site 232408−16.2 (±1.8)−14.9 (±1.8)10.1 (±0.9)−1.5 (±0.7)4.0 (±0.6)Site 341504−22.6 (±0.4)−22 (±0.2)7.6 (±0.8)0.5 (±1.3)3.4 (±0.1)Site 443608−22.8 (±0.5)−21.8 (±0.2)7.6 (±1.0)1.6 (±0.5)3.8 (±0.3)Site 546205−23.3 (±0.2)−22.2 (±0.4)7.1 (±0.1)1.3 (±0.1)3.7 (±0.2)Site 650707−22.0 (±1.1)−20.9 (±0.9)10.2 (±2.1)1.0 (±0.9)3.8 (±0.4)Site 767391−22.0 (—)−21.5 (—)7.0 (—)2.0 (—)3.3 (—) Summary statistics (n, mean ± SD) for Phyllotis vaccarum liver stable isotope analyses of carbon (values shown for δ13C and lipid‐corrected δ13C), nitrogen (δ15N), and sulfur (δ34S) and the elemental C:N ratio. Mice from the highest elevations did not exhibit negative δ15N values, suggesting that lichenivory is not especially common. Assignment to Capture Elevations As abiotic and biotic conditions (e.g., temperature, aridity, UV concentrations, plant nutrient availability, and soil organic content) change with elevation, so too do the biomass and community composition of the primary producers (Díaz et al. 2019), with consequent changes in the availability of food for consumers and isotopic shifts at the base of the food web (Díaz et al. 2016). Given the known elevational gradient in stable isotope values, we can expect that stable isotope values will provide a means of identifying variation in habitat use among mice captured at different elevations. Analysis of the combined stable isotope dataset using CAP showed that mice could be reliably assigned to broad elevational zones using individual δ13C, δ15N, and δ34S values (Figure 5D) (CAP: Trace = 1.71, p = 0.0001). The leave‐one‐out classification cross validation (Table 3) indicated that by using the three isotope ratios, we could assign an individual mouse to a 1000 m interval with ~85‰ success. The CAP model predicted that the 6739 m summit mouse from Volcán Llullaillaco was isotopically most similar to mice captured from the 4000–5000 m interval (Figure 5D). The CAP model also suggested that four other individuals had stable isotope values characteristic of elevational zones distinct from where they were captured (Table 3). Including the summit mouse, this indicates that ca. 15‰ of the P. vaccarum in the study area had stable isotope values suggestive of upslope or downslope dispersal. TABLE 3Results of the canonical analysis of principal coordinates (CAP) leave‐one‐out cross validation to assess the ability of the model to assign individual mice to the 1000 m elevational zone in which they were captured.Known capture altitude (m)Predicted capture altitude (m)‰ Correctly classified2000–30003000–40004000–50005000–60002000–30007001863000–40000701864000–500000161945000–6000002571 Results of the canonical analysis of principal coordinates (CAP) leave‐one‐out cross validation to assess the ability of the model to assign individual mice to the 1000 m elevational zone in which they were captured. Trophic Position Modal estimates of P. vaccarum TP varied between capture sites (Table 4) and ranged from 1.9 at Site 2 to 4.3 at Site 1. The latter estimate is extremely high and reflects the very high δ15N values from mice collected at Site 1 (mean δ15N = 19.2‰). Discounting the results from Site 1, mouse TP estimates were generally similar across sites and were indicative of omnivory with modal values between 1.9 and 2.3 at Sites 2–5. The modal estimate for mice at Site 6 was slightly higher (TP = 3.4), but the credibility limits overlapped with those from all sites apart from Site 2. TP values between 2 and 3 are indicative of an herbivorous diet that includes some animal prey. TP values > 3, as seen at Site 6, indicate a diet dominated by animal prey. The TP of the summit mouse from Volcán Llullaillaco was estimated as 2.2, quite close to that of mice from Sites 3, 4, and 5. TABLE 4Estimates of trophic position for mice captured in different elevational zones. Summary statistics are provided as modal TP (95‰ credibility intervals) and as mean TP ± SD. tRophicPosition modelBaseline data (Díaz et al. 2016)Trophic position (mode, 95‰ credibility intervals)Trophic position (mean ± SD)Site 01 oneBaseline C3 & CAM data combined from 2000–3000 m4.3 (3.4–5.1) a 4.3 ± 0.2Site 02 twoBaselines C3 & CAM/C4 data 3000–4000 m1.9 (1.7–2.2)1.9 ± 0.1Site 03 oneBaseline Plants > 4000 m (all C3)2.3 (1.8–2.9)2.3 ± 0.3Site 04 oneBaseline Plants > 4000 m (all C3)2.3 (1.9–2.7)2.3 ± 0.2Site 05 oneBaseline Plants > 4000 m (all C3)2.2 (1.8–2.5)2.2 ± 0.2Site 06 oneBaseline Plants > 4000 m (all C3)2.9 (2.3–3.5)2.9 ± 0.3Summit mouse oneBaseline‐greta Plants > 4000 m (all C3)2.2 (1.9–2.4)2.2 ± 0.1 a δ15N values of mice captured at Site 1 were unusually high and were markedly 15N enriched relative to the putative baseline (see Figure 5B) and may be artifactual. Estimates of trophic position for mice captured in different elevational zones. Summary statistics are provided as modal TP (95‰ credibility intervals) and as mean TP ± SD. δ15N values of mice captured at Site 1 were unusually high and were markedly 15N enriched relative to the putative baseline (see Figure 5B) and may be artifactual. Conclusions A combination of metagenomic, metabarcoding, and stable isotope data provided new insights into the diet of Phyllotis mice living at extreme elevations that far surpass known vegetation limits. Stable isotope data revealed that Phyllotis vaccarum maintains a mainly omnivorous diet in all elevational zones, and elevational variation in diet reflects variation in vegetation composition and the extent to which the mice rely on animal prey. Estimates of TP based on isotopic data indicated that mice collected near apparent vegetation limits (~5100 m) on the flanks of Llullaillaco rely more heavily on animal prey than mice from lower elevations. Metagenomic and metabarcoding analyses of gut contents from the mouse from the summit of Llullaillaco (6739 m) revealed a strictly herbivorous diet. The absence of animal DNA suggests that mice at extreme elevations do not subsist on wind‐blown arthropods or other animal material. The detection of DNA from lichen‐associated fungi indicates that Phyllotis mice living above known vegetation limits may supplement their diet with saxicolous lichens, as observed for other arctic and alpine mammals during periods of food scarcity in the winter. However, measured δ15N levels indicate that lichen is not an important dietary staple in mice native to any of the surveyed elevational zones. The metagenomic and metabarcoding data also produce a scientific conundrum: the gut contents of a mouse captured at 6739 m elevation contained DNA from several families of native plants that are not known to occur above ~5000 m elevation. Clearly, we have more to learn about the elevational distributions of both the plants and the mice. It is possible that some of the plants identified in the diet of the summit mouse exist at higher elevations than previously supposed, but they occur beneath the snowpack or in other cryptic microhabitats. Author Contributions Claudio Quezada‐Romegialli: conceptualization (equal), formal analysis (equal), investigation (equal), visualization (equal), writing – review and editing (equal). Marcial Quiroga‐Carmona: investigation (equal), writing – review and editing (equal). Guillermo D'Elía: investigation (equal), writing – review and editing (equal). Chris Harrod: conceptualization (equal), funding acquisition (equal), investigation (equal), writing – review and editing (equal). Jay F. Storz: conceptualization (equal), funding acquisition (equal), investigation (equal), project administration (equal), supervision (equal), writing – review and editing (equal). Ethics Statement All animals were handled in accordance with protocols approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Nebraska (Project ID: 1919). All animals were collected in the field with permission from the following Chilean government agencies: Servicio Agrícola y Ganadero (SAG, Resolución Exenta # 6633/2020), Corporación Nacional Forestal (CONAF, Autorización # 171219), and Dirección Nacional de Fronteras y Límites del Estado (DIFROL, Autorización de Expedición Científica #68/2020). Conflicts of Interest The authors declare no conflicts of interest. Supporting information Table S1. Acknowledgments We thank Mario Pérez‐Mamani and Juan Carlos Briceño for assistance and companionship in the field, and we thank Cristina Dorador for logistical assistance. Funding: This work was funded by grants to JFS from the National Institutes of Health (R01 HL159061), National Science Foundation (OIA‐1736249 and IOS‐2114465), and National Geographic Society (NGS‐68495R‐20), and a grant to GD from the Fondo Nacional de Desarrollo Científico y Tecnológico (Fondecyt 1221115). Data Availability Statement The data associated with this study are openly available in the NCBI Sequence Read Archive (SRA), under the BioSample accession number SAMN42772345. References Associated Data Supplementary Materials Table S1. Data Availability Statement The data associated with this study are openly available in the NCBI Sequence Read Archive (SRA), under the BioSample accession number SAMN42772345.