Analysis of winter diet in Guizhou golden monkey (Rhinopithecus brelichi) using DNA metabarcoding data

Rhinopithecus brelichi (Guizhou or grey snub-nosed monkey) is among the world’s most critically endangered primates, endemic to China’s Fanjingshan National Nature Reserve (FNNR). Its ranging patterns are influenced by altitude, temperature, vegetation, and food availability. Diet is central to understanding primate survival, behaviour, and habitat relationships. Traditional diet analyses (behavioural observations, stomach contents, faecal microscopy, stable isotopes) have limitations for elusive, endangered species. DNA metabarcoding, combining DNA barcoding with high-throughput sequencing, enables sensitive, fine-scale taxonomic resolution from faecal samples and has been applied to diverse taxa and several primates, including colobines, but information on R. brelichi, especially animal prey in winter, remains scarce. Winter in subtropical montane habitats presents limited high-quality foods, making fallback resources important. The study aimed to answer: (1) What are the main winter dietary components of R. brelichi? (2) What are the growth habits (herb, shrub, tree, liana) of the plants consumed? (3) Are there previously unreported dietary items? The authors hypothesized that: (1) winter diets include shrubs and herbs; (2) plant diet diversity exceeds animal diet diversity; and (3) new dietary species would be detected. Using faecal DNA metabarcoding, they investigated winter diet composition and diversity of plants and animals to infer foraging preferences and identify potential fallback foods.

Permissions were obtained from the FNNR Administration and work complied with China’s Wildlife Protection Law. Non-invasive sampling avoided direct contact with animals. Study site and sampling: FNNR (27°46′50″–28°1′30″ N, 108°45′55″–108°48′30″ E) spans 419 km², elevation 500–2572 m, with vegetation from evergreen broad-leaved to subalpine scrub. R. brelichi mainly inhabits mixed deciduous–evergreen broad-leaved forests. From Dec 2021 to Jan 2022, 31 fresh faecal samples (bead-like, moist, intact) were collected from forest trails and high-activity areas, placed into 50 mL tubes with absolute ethanol, transported, and stored at −80°C. GPS and vegetation type at each site were recorded. DNA extraction: Total DNA was extracted using an OMEGA kit. DNA/RNA quality was assessed by 0.8% agarose gel electrophoresis and NanoDrop A260/A280; DNA was stored at −20°C. PCR and sequencing: Plant diet marker rbcL amplified with primers ZlaF (5′-ATGTCACCACCAACAGAGACTAAAGC-3′) and hp2R (5′-CGTCCTTTGTAACGATCAAG-3′). Animal diet marker COI amplified with COI inTF (5′-GGWACWGGWTGAACWGTWTAYCCYCC-3′) and COI jgHC02198 (5′-TANACYTCNGGRTGNCCRAARAAYCA-3′). PCR (25 µL): 5× Reaction Buffer and 5× High GC Buffer, dNTPs (10 mM), primers (10 µM), 1 µL template, Q5 high-fidelity polymerase; cycling: 98°C 5 min; 30 cycles of 98°C 30 s, 50°C 30 s, 72°C 30 s; final 72°C 5 min; hold 4°C. Amplicons were checked on 2% agarose gels, purified (AMPure), quantified (PicoGreen), and each sample was sequenced on Illumina NovaSeq PE250 (2×250 bp). Bioinformatics and taxonomic assignment: Quality control by FastQC; trimming with Trimmomatic (retain reads ≥150 bp). Merging and chimera removal with FLASH and QIIME2 (2019.4). OTUs were clustered at 97% similarity in QIIME2. Representative sequences were annotated via BLASTN against NCBI taxonomy. Annotations were cross-checked with species distribution records for FNNR to derive taxonomic composition and relative abundance at family, genus, and species levels. Diet metrics and statistics: Percentage of occurrence (%RO) calculated as Ni/ΣNi×100%, where Ni is the number of samples containing a given taxon. Rarefaction/extrapolation (iNEXT) computed Hill numbers (q=0,1,2). The top 20 plant genera by abundance and occurrence were visualized, assigned to orders/families, and grouped by growth habit (herb, liana, shrub, tree, mixed). Figures were produced in Origin 2022. Sequencing output: rbcL yielded 2,418,274 target fragments (mean length ~202 bp), per-sample min 49,561, max 98,503, mean 78,167 ± 10,393. COI yielded 3,105,612 fragments (mean length ~313 bp), per-sample min 72,571, max 161,901, mean 100,155 ± 17,203.

- Plant diet composition: The top plant families (mean relative abundance >1%) included Magnoliaceae (1.16%–75.41%), Rubiaceae (0.1%–80.08%), Lauraceae (0.01%–69.16%), Adoxaceae (0.01%–23.54%), Rutaceae (0–24.07%), Pentaphylacaceae (0.016%–24.30%), Celastraceae (0–12.87%), Smilacaceae (0–9.75%), and Caprifoliaceae (0–8.23%). - Top plant genera: Magnolia (1.16%–75.41%), Morinda (0.072%–79.25%), Viburnum (0.01%–23.54%), Tetradium (0–24.06%), Eurya (0.016%–24.29%), Euonymus (0–12.43%), Smilax (0–9.75%), Lonicera (0–8.23%). - Ubiquity across samples: Families present in all 31 samples: Magnoliaceae, Rubiaceae, Adoxaceae, Pentaphylacaceae, Cornaceae, Lauraceae, Theaceae, Nyssaceae, Apocynaceae. Present in 30/31: Gentianaceae, Aquifoliaceae, Symplocaceae, Araliaceae, Rosaceae. Genera in all 31: Magnolia, Morinda, Viburnum, Eurya, Cornus, Nyssa, Coccocypselum. In 30/31: Gentiana, Ilex, Symplocos. Stewartia and Camellia occurred in 29 samples. - Animal and macrofungal diet: Dominant animal families included Psychodidae (0–71.51%), Trichinellidae (0.6%–27.97%), Staphylinidae (0–4.20%), Scarabaeidae (0–84.13%), Trichoceridae (0–74.81%). Dominant genera included Psychoda (0–71.51%), Oscheius (0–1.11%), Dermestes (0–0.39%). At higher ranks: Diptera (0–75.24%), Trichinellida (0–27.97%), Coleoptera (0–98.81%), Insecta (0.01%–98.84%), Enoplea (0–27.97%). Macrofungal taxa detected included Pleurotus (0–0.49%), Cordycipitaceae (0–14.75%), Agaricales (0–3.13%), Agaricomycetes (0–3.13%). - Diversity metrics (iNEXT): Plant diet showed higher diversity than animal+macrofungal diet. Reported values: species richness 47 (plants) vs 20 (animals+macrofungi), Shannon diversity 40.12 vs 18.09, Simpson diversity 37.44 vs 16.92. - Dietary habits of plants: Among the top 20 plant genera, by relative abundance the habits were dominated by shrubs/trees (45%), with herbs (15%) and lianas (15%). By percentage of occurrence, shrubs/trees dominated (55%), followed by trees (10%) and mixed liana/shrub/tree (10%). - New or noteworthy species-level plant detections (from chloroplast DNA): Holboellia latifolia, Persicaria maculosa, Cinnamomum camphora, Glyptostrobus pensilis, Smilax trinervula, Aucuba japonica, Camptotheca acuminata, Menyanthes trifoliata. - Dominant orders/families in sunburst analyses: By relative abundance and %RO, prominent plant orders/families included Magnoliales, Gentianales, Dipsacales, and families Magnoliaceae, Rubiaceae, Adoxaceae, Rutaceae. - Overall, winter diet was largely composed of shrubs, herbs, and shrub/tree taxa, with insects (Diptera, Coleoptera, Trichoptera, Orthoptera, Hemiptera) and macrofungi as supplementary items.

DNA metabarcoding enabled a fine-scale characterization of the winter diet of R. brelichi in FNNR, revealing strong reliance on certain plant families (Magnoliaceae, Rubiaceae, Adoxaceae, Theaceae, Lauraceae) and genera (Magnolia, Morinda, Viburnum, Tetradium, Eurya). Habit analysis indicated a winter shift towards shrubs and shrub/tree taxa, consistent with optimal foraging when preferred, high-quality foods are scarce and the use of fallback resources. The detection of plant species such as Holboellia latifolia, Persicaria maculosa, and Cinnamomum camphora, some previously unreported for this species, supports the hypothesis that metabarcoding can uncover new dietary items. The plant diet displayed greater richness and diversity than the animal/macrofungal components, aligning with the species’ colobine folivorous adaptations. Animal DNA indicated consumption of multiple insect orders and nematodes; given the likely variability in prey size and foraging behaviour, the study posits that many small insects (e.g., Diptera, Trichoptera, Hemiptera) and nematodes may be incidentally ingested via plant material, water, or soil, whereas larger Coleoptera and Orthoptera could be intentionally preyed upon. Macrofungi (Agaricales, Pleurotaceae, Psathyrellaceae) were also detected and may represent fallback foods during resource-poor periods. These findings refine understanding of winter foraging strategies and nutrient acquisition, including potential contributions of insects to protein and vitamin B12 intake. Conservation implications include prioritizing protection and restoration of key winter plant resources—especially shrubs and shrub/tree species within Magnoliaceae, Rubiaceae, Adoxaceae, Theaceae, Lauraceae—and maintaining habitat heterogeneity that supports these taxa. The results also underscore the utility of DNA metabarcoding for monitoring diets of elusive, endangered primates and guiding habitat management.

DNA metabarcoding revealed that winter diets of Rhinopithecus brelichi in FNNR are dominated by plants, particularly shrubs and shrub/tree taxa from families such as Magnoliaceae, Rubiaceae, Adoxaceae, Theaceae, and Lauraceae, with supplementary consumption of insects (e.g., Diptera, Coleoptera, Trichoptera) and macrofungi (Agaricales). Plant diet diversity exceeded that of animals and macrofungi, and several plant species not previously reported were identified at the species level. These insights improve understanding of winter feeding habits and inform conservation actions focused on protecting key shrub and herbaceous resources. Future work should enhance species-level resolution via metagenomics or ASV approaches and develop local reference libraries to improve taxonomic assignment, along with evaluating primer sets tailored for primate diets.

- Species-level resolution limitations: Chloroplast and COI metabarcoding can suffer from pseudogenes, heterozygosity, and incomplete or biased reference libraries, potentially reducing accuracy and splitting single species into multiple taxa. - Inference of prey acquisition: DNA data cannot reliably distinguish intentional predation from incidental ingestion (e.g., small insects/nematodes on plant material or water) or second-hand predation, complicating interpretation of animal diet components. - Marker and primer bias: Single-locus amplicons may under-detect certain taxa; alternative or additional markers (e.g., ND5, COII) may broaden coverage. - Geographic reference constraints: Reliance on general databases, even when cross-checked with regional floras/faunas, may misassign taxa absent from local reference libraries. - Seasonal and temporal scope: Sampling was limited to winter (Dec–Jan) and one season; diets may vary intra- and inter-annually, limiting generalizability across seasons.