Diet composition and selection of Père David's deer in Hubei Shishou Milu National Nature Reserve, China

Abstract

Hubei Shishou Milu National Nature Reserve is an ideal place to restore the wild population of Père David's deer (Elaphurus davidianus). Understanding foraging ecology and diet composition is essential for assessing population development or establishing long-term effective conservation measures for endangered species. However, little is known about the diet composition of Père David's deer and its diet selection mechanism. In this study, we used stable isotope technology to investigate the diet composition of Père David's deer according to various tissues (i.e., fur, muscle, liver, heart, and feces) and seasons, and evaluated the correlation between the nutrient composition of plants and diet composition. Bayesian isotope analysis showed that the autumn and winter diet estimated by fur and fecal samples indicated a diet dominated by C₃ grasses (42.7%–57.2%, mean), while the summer diet estimated by muscle and liver samples was dominated by C₃ forbs (30.9%–41.6%, mean). The Pearson correlation test indicated that the contribution of winter diet composition reflected by fur and fecal samples was associated with correlations with crude protein (r = .666, p < .01) and soluble sugars (r = .695, p < .01). The results indicated that crude protein and soluble sugars were important factors influencing the winter diet selection of Père David's deer. In the context of the current reintroduction facing many challenges, such as habitat fragmentation, wetland degradation, and human disturbance, comprehensively evaluating the diet selection mechanism of Père David's deer under different resource specificities and temporal changes should be considered in the future.

INTRODUCTION

Père David's deer (Elaphurus davidianus) is one of the national key protected animals belonging to Mammalia, Artiodactyla, Cervidae, and Elaphurus (Cao, 1985). Père David's deer became extinct in the wild in 1900 due to war, poaching, and habitat deterioration. In the 1980s, China began to implement the reintroduction project of Père David's deer and first released captive Père David's deer to nature in 1998 (Ding, 2017). With its reintroduction, ex situ conservation and population restoration have gained a great deal of attention in China (Wang et al., 2020; Xu & Yu, 2019; Xue et al., 2022).

Understanding diet composition is essential for assessing population development or establishing long-term effective conservation measures for endangered species. However, little is known about the mechanisms of diet selection, food assimilation, and the interrelationships between Père David's deer and their habitat due to the limitations of research methods and the particularity of research objects (Ding, 2009; Hua et al., 2020; Wang & Wang, 2011). Père David's deer mainly lives in swampy and muddy areas and feeds on the young branches and leaves of grasses and some legumes (Ding, 2004). To study the diet composition of Père David's deer more systematically, Ding et al. (1989) initiated investigations of the vegetation in the territory of Père David's deer, dividing the plants into avoided, foraged, and preferred according to the frequency of feeding and the degree of foraging. Hua et al. (2020) used both fecal microanalysis and direct observation to compare and analyze the feeding habits of Père David's deer in winter in Yancheng Wetland, Jiangsu Province. Phragmites australis and Spartina alterniflora accounted for 96.73% of the diet composition of Père David's deer, while the branches and leaves of woody plants were rarely consumed, accounting for only 3.27% of the diet composition (Hua et al., 2020). To further assess the reintroduction of Père David's deer efforts, researchers also need effective information on changes in diet across seasons, including long-term diet composition and factors that mediate dietary variation.

Hubei Shishou Milu National Nature Reserve is the habitat of the largest wild Père David's deer population in China. It is a typical lake wetland located in the middle reaches of the Yangtze River. However, due to the construction of levees after floods in 1998 and pluvial flooding in the upper reaches of the Yangtze River, the floating aquatic plants were unable to reach the reserve (Li et al., 2012; Zhang et al., 2018). In addition, following the completion of the Three Gorges Dam, the original beach embankment was isolated from natural water exchange between the Tian'E Zhou Oxbow and the Yangtze River (Ding, 2017; Zhang et al., 2018). Human activities, including planting Italian poplar and burning weeds to open up wasteland, accelerated wetland droughts. The affected hygrophytes were gradually replaced by xerophytes and mesophytes, which further accelerated wetland degradation (Li et al., 2016; Zhang et al., 2013; Zhao et al., 2010), soil erosion (Li et al., 2017), and water pollution (O'Hare et al., 2018; Wolka et al., 2018). Wetland degradation may lead to the destruction of wetland ecosystem structure (Davidson et al., 2018), biodiversity reduction (Janne et al., 2021), productivity and wetland function attenuation (Bouma et al., 2014), and other ecological environment deterioration (Fan et al., 2021). Moreover, changes to the vegetation community structure could have a subsequent impact on the largest wild Père David's deer population in China and on the food source of Père David's deer. This further affects the stability of the wetland ecosystem (Hummel et al., 2018; Motta et al., 2020).

Decreases in plant species abundance, diversity, or coverage can affect the foraging choices or habitat use of Père David's deer by changing plant–cervid interactions (Li et al., 2015, 2016). When Père David's deer population abundance increases, it results in a decrease in the diversity and abundance of plants, as seen in other deer populations in Jiangsu, China, Central Japan, and northeastern Illinois (Anderson et al., 2005; Ding, 2017; Iijima et al., 2013). With the increase in trampling frequency of Père David's deer, the soil bulk density increases, while the soil moisture content decreases (Zhou et al., 2010). Meanwhile, available phosphorus, potassium, and other soil components in the surface layers are also continuously reduced (Ding, 2017; Qian et al., 2008; Zhou et al., 2010). This results in lower plant nutrients and reduced biomass, thus preventing the growth of Père David's population within the national reserves. These current and future problems threatening the protection and management of Père David's deer in China could be addressed through the artificial planting of pasture, the implementation of habitat protection and restoration projects, human intervention in population changes (migration in and out) of the population, and other methods. Understanding the dietary composition and quantifying the nutrient composition of Père David's deer is the basis for this work. This is closely related to the type of forage to be planted, the assessment of emigration sites, and the direction of habitat restoration (Xue et al., 2022; Zhang et al., 2021).

Stable isotope analysis can provide quantitative information on the dietary contribution of various foods consumed by Père David's deer (De Smet et al., 2004). Compared with traditional methods such as behavioral observations and stomach content analyses, stable isotope analysis can provide longer-term measurements (McCue et al., 2020). This can help us to understand variations in Père David's deer diet and provide essential information for conservation. Depending on which tissues are analyzed, the stable isotope signature would reflect food assimilation over several time scales, from a few days to the lifetime of the animal (Tieszen et al., 1983). Cell turnover in the tissue is an important factor affecting the turnover rate of different tissues in animals. Taking mammals as an example, feces would reflect food consumption of the last days, the liver for approximately the last month, the heart for the last 1–2 months, and the muscle for the last 3–4 months (Bahar et al., 2014; Caut et al., 2009; Phillips et al., 2014; Roth & Hobson, 2000; Sponheimer et al., 2006). The fur reflects food consumption during fur growth at different time scales (Ayliffe et al., 2004; Cerling et al., 2006; Schwertl et al., 2005), depending on the segment analyzed (Guo et al., 2008; O'Regan et al., 2008; Rogers et al., 2020; Schwertl et al., 2003). We used multiple deer tissues to evaluate the diet of Père David's deer according to different time scales. We also analyzed the nutrient composition of diverse plant species and combined stable isotopes and nutrient compositions to explore the relationship between the diet composition of Père David's deer and the food's nutrient content. Specifically, we aim to address the following scientific aspects of forecasting: (1) Do the stable isotope signatures of Père David's deer vary among tissue types? (2) Does the diet composition of Père David's deer change with the season? (3) Is the diet composition of Père David's deer influenced by the food's nutrient content?

MATERIALS AND METHODS Study sites and sample collection

Hubei Shishou Milu National Nature Reserve (E112°31′-112°36′, N29°46′-29°51′) is located in Hubei Province in China (Figure 1). With a total area of 1567 hm², the reserve is adjacent to the Yangtze River to the south with the Tian'E Zhou Oxbow to the east, located at the southern end of the Jianghan Plain (Zou et al., 2013). There are approximately 72 families, 216 genera, and 321 species of wild higher plants in the reserve. The vegetation types in this area can be divided into three groups: meadow type, swamp type, and aquatic plant type. Among them, herbs are the main plant community, and there are few tree species (Zhang, Li, et al., 2019). Père David's deer is the only large mammal in the reserve.

FIGURE 1. Map of the geographical location and regional distribution of the Shishou Milu National Nature Reserve in Hubei Province, China. Areas a and b are the sampling sites for plant and fecal samples.

Plant samples were collected in January 2021. The activity traces of Père David's deer (resting place traces and foraging plant traces) were investigated during our survey and combined with the observation records of the reserve staff. The main activity ranges of Père David's deer were identified as the Italian poplar forest near the oxbow and the S. matsudana forest near the bank of the Yangtze River (Figure 1). A 4 m² sample plot was established every 150 m in areas a and b. A total of six sample plots (3 m × 2 m) were collected. In each sample plot, aboveground plant parts with signs of foraging were collected with stainless-steel scissors in sealed bags, which were kept in ice boxes to avoid wilting of plant leaves due to temperature changes. Plants with high cover (more than one-fifth of the sample plot) but no obvious signs of feeding were also collected.

Fresh fecal samples from the areas a and b were collected in January 2021. Père David's deer were spotted and followed by professional staff at a distance, fresh droppings were picked up along the way. One dung mound was regarded as being created by one deer. Five dung mounds were mixed into one fecal sample. A total of nine fresh fecal samples were collected. Soil and plant tissue on the surface of the feces were cleaned with brushes and tweezers and stored in sealed bags. Fecal samples were considered to show the diet in January. Fur samples were collected in April 2019 and 2020. The shedding period of Père David's deer is from March to April and August to September each year (Ding, 2017). The difference in color and size between adult and immature shedding fur masses of Père David's deer is obvious. Fur was collected through professional staff, relying on observation at a distance to ensure that one fur mass was regarded as being created by one deer. To avoid the effects of age, only the fallen fur masses from the adult Père David's deer were collected and kept in sealed bags for further processing. Approximately three to four fur masses were mixed into one fur sample, and a total of four fur samples were collected in areas a and b each year, for a total of eight fur samples collected in 2019 and 2020. The fur divided into three sections (fur tip, fur middle, and fur root) can be approximately considered to show the diet composition from October to November, December to January, and February to March (Ding, 2017; Guo et al., 2008; Schwertl et al., 2003). Muscle, liver, and heart samples were taken from one individual adult male Père David's deer that died accidentally (intraspecific struggle) on August 5, 2019, and were sliced with a sterile scalpel and stored at −80°C after being collected in sealed bags. A total of five muscle samples, five liver samples, and three heart samples were collected from this Père David's deer. Muscle samples are considered to show the diet between May and June (Caut et al., 2009; Phillips et al., 2014; Sponheimer et al., 2006). Heart samples are considered to show the diet between June and July (Bahar et al., 2014; Caut et al., 2009; Roth & Hobson, 2000). Liver samples are considered to show the diet in July (Bahar et al., 2014; Caut et al., 2009; Roth & Hobson, 2000; Sponheimer et al., 2006). Research on live animals was performed following the guidelines of the China Wildlife Protection Law, and all research protocols were approved by Hubei Shishou Milu National Nature Reserve.

Sample pretreatments and lipid removal

Plant samples were processed immediately upon return to the laboratory. Fresh, mature, insect-free plant leaves were cut off with scissors and gently rinsed with distilled water. The cleaned leaves were immediately sent to a freeze dryer for drying (−40°C in a vacuum freezing dryer and 48 h of dehydration; all subsequent freeze-drying settings were the same). The dried leaves were ground into a fine powder using a ball mill and packaged. The fecal samples were washed with distilled water and sent to a freeze dryer for drying. After being removed, the samples were ground into a fine powder using a ball mill and packaged.

The fur of Père David's deer can be roughly divided into guard fur, underfur, and intermediate fur, with guard fur being the densest and distributed throughout the body. Guard fur is approximately 40–60 mm long, and its growth rate is approximately 6–10 mm/month (Ding, 2017; Perrin & Campbell, 1980). Eight fur samples (guard fur) were picked out with tweezers and straightened, measured, and divided equally into three sections using scissors, for a total of 24 fur subsamples. After they were washed in an ultrasonic bath using a 2:1 chloroform–methanol solution to remove surface contamination and external lipids (Azorit et al., 2012), the fur subsamples were washed with distilled water and then put into an oven for drying (Hobson et al., 2000).

To further reduce the effect of lipids on stable isotopes, initially processed fur samples and tissue samples from the previous step were subjected to a 2:1 chloroform–methanol solution for further lipid extraction (Ehrich et al., 2011; Rioux et al., 2019). The mixture was shaken and stored at 4°C overnight (18 h; Folch et al., 1957). The supernatant was then removed by centrifugation for 10 min. This procedure was repeated two times (Folch et al., 1957). After three extractions, the samples were dried in an oven. After being dried for 12 h, they were washed with distilled water and then dried in a freeze dryer. After being removed, they were ground into a fine powder using a ball mill and then packaged. All test tissue samples were degreased samples.

Determination of plant nutrients

To further explore the effect of plant nutrients on the dietary composition of Père David's deer, freeze-dried and ground plants were used for subsequent plant nutrient analysis. Three replicates were set up for each plant species. The mean value (±SD) of three replicates per species was used as the measurement of nutrient content.

Crude fat determination was determined according to the direct extraction method (Zou et al., 1999). Freeze-dried plant samples (500 mg) were dissolved in 10 ml of HCl for 50 min. Then, 10 ml of ethanol was added, and the fat was extracted with 25 ml of petroleum ether, mixed, and shaken, and the supernatant was collected. The supernatant was dried (105°C for 2 h) and weighed to calculate the crude fat content (Zou et al., 1999).

The C/N and N contents were determined using an elemental analyzer (Flash EA 1112HT; Thermo Fisher Scientific) in the laboratory of the Food Inspection and Quarantine Center, Shenzhen Custom, China.

The soluble sugar content was determined according to the anthrone–H₂SO₄ method (Nakamura, 1968). A standard curve was established with glucose standards. Freeze-dried plant samples (250 mg) were extracted by adding 10 ml of distilled water in boiling water for 30 min, and the supernatant was mixed with ethyl anthranilate-concentrated H₂SO₄, shaken, and held in a boiling water bath for 1 min. The absorbance was measured at 630 nm by UV spectrophotometry (Nakamura, 1968).

The crude protein content was determined by the Coomassie brilliant blue method (Hayes, 2020). A standard curve was established with bovine serum protein standards. Freeze-dried plant samples (100 mg) were weighed and extracted with distilled water for 2 h at room temperature, the supernatant was mixed with Kaumas Brilliant Blue G250, and the absorbance was measured at 595 nm with UV spectrophotometry (Hayes, 2020).

Condensed tannin (proanthocyanidins) was determined by the acid butanol method (Wei et al., 2014). Freeze-dried plant samples (50 mg) were dissolved in 1 ml of methanol and mixed with 6 ml of 95% butan-1-ol and 5% concentrated HCl in 10 ml test tubes. The tubes were sealed and placed in a water bath at 95°C for 1 h. The color was observed after cooling to room temperature. The alcoholysis under acidic conditions converts the extended units of condensed tannins into colored anthocyanins. The darker the color, the higher the tannin concentration (Wei et al., 2014).

Stable isotope analyses

After the pretreatments, the carbon and nitrogen stable isotope ratios of the samples were analyzed using an elemental analyzer (Flash EA 1112HT; Thermo Fisher Scientific) coupled with an isotope ratio mass spectrometer (Delta V Advantage; Thermo Fisher Scientific) in the laboratory of the Food Inspection and Quarantine Center, Shenzhen Custom, China. Approximately 0.200 mg (±0.001 mg) of sample plant and tissues was weighed and inserted into 4 mm × 6 mm tin cups. The carbon and nitrogen stable isotope ratios of the sample were expressed in delta notation (difference between sample and standard) as parts per thousand (‰):[Image Omitted. See PDF]where R is the abundance ratio of heavy isotopes to light isotopes in the sample, ¹³C/¹²C and ¹⁵ N/¹⁴ N. R_sample is the measured isotope ratio; R_standard is the isotope ratio of reference materials. All results are reported relative to atmospheric nitrogen as the standard for δ¹⁵N and to Cretaceous belemnite (Belemnitella Americana) from the Peedee Formation of South Carolina for δ¹³C. Laboratory standards (glycine and urea) were run every 12 samples to correct any instances of instrument drift. The analytical precision was ±0.1‰ for ¹³C and ±0.2‰ for ¹⁵N.

Mixing models and statistical analyses

For animals with two or more food sources, the proportion of food in animal diets can be determined according to the isotopic mass balance equation:[Image Omitted. See PDF][Image Omitted. See PDF][Image Omitted. See PDF]where δ¹³C_i and δ¹⁵N_i represent the carbon and nitrogen isotopic compositions of consumers, respectively; δ¹³C_j and δ¹⁵N_j represent the carbon and nitrogen isotopic compositions of possible foods, respectively; ΔC and ∆N represent the carbon and nitrogen isotope discrimination values of nutrient grade, respectively; and $$$ {f}_{ij}^{\prime } $$$ is the proportion of food in the consumers' food sources (Parnell et al., 2013; Saito et al., 2001).

To estimate the relative contribution of multiple food sources to Père David's deer diet, the Bayesian isotope mixing model SIMMr (stable isotope mixing models) in R (Jackson et al., 2008; Parnell et al., 2010) was used, as it provides advantages over standard, mass balance multisource mixing models (Phillips & Gregg, 2003). SIMMr can integrate sources of variability associated with multiple sources, trophic enrichment factors, and isotopic values, and its outputs represent true probability density functions rather than a range of feasible solutions (Moore & Semmens, 2008; Parnell et al., 2013; Phillips et al., 2014).

As different tissue types metabolize isotopes at different speeds, before applying a mixing model, these systematic differences must be corrected (Phillips et al., 2014). To more accurately reflect the contribution of each food source to Père David's deer, trophic enrichment factors (TEFs) were used to correct for the enrichment of stable isotope signatures between the tissue and diet of Père David's deer. Since TEF values for various tissues of Père David's deer are not available, the TEFs from the currently published literature were used instead. We are aware that incorrect use of TEF can produce erroneous determinations of food sources (Bond & Diamond, 2011). Estimated TEFs reached 2.31 ± 0.20‰ for ∆¹³C (Caut et al., 2009) and 4.86 ± 0.94‰ for ∆¹⁵ N for fur (Rioux et al., 2020); 1.48 ± 0.14‰ for ∆¹³C (Caut et al., 2009) and 3.02 ± 0.12‰ for ∆¹⁵ N (Caut et al., 2009) for muscle; 0.36 ± 1.86‰ for ∆¹³C (Stephens et al., 2022) and 3.55 ± 0.05‰ for ∆¹⁵ N (Caut et al., 2009) for liver; 2.92 ± 0.01‰ for ∆¹³C (Caut et al., 2009) and 2.95 ± 0.01‰ (Caut et al., 2009) for ∆¹⁵ N for heart; and −0.9 ± 0.2‰ (Sponheimer et al., 2006) for ∆¹³C and 1.9 ± 0.3‰ for ∆¹⁵ N (Steele & Daniel, 1978) for feces. Due to sampling time constraints, the plant samples collected in January 2021 were used as a potential food source for all mixing model analyses. Sixteen species of plants were collected and grouped into four groups according to the stable isotope signature similarity, photosynthetic pathway (Phillips et al., 2014), ANOVA testing, and feeding preferences of Père David's deer (Ding, 2017; Zhang, Fu, et al., 2019; Zhang & Yang, 2017). The four groups were C₃ forbs (C₃-F, n = 10), C₃ grasses (C₃-G, n = 2), C₄ grasses (C₄-G, n = 2), and Other (avoided, n = 2), respectively. The negatively correlated source proportions were combined to gain precision in the calculated proportions (Parnell et al., 2013; Phillips et al., 2014). Therefore, C₃-F, C₃-G, and C₄-G were included in the Bayesian isotope mixing model. Estimates are reported with their 95% credible intervals, and the results are displayed as median, 5% and 95% percentile values (Parnell et al., 2013).

One-way analysis of variance (ANOVA) followed by Tukey's post hoc test was used to test differences in stable isotope values among the potential food sources (C₃-F, C₃-G, and C₄-G) and fur subsamples. Then, a two-way ANOVA was conducted to evaluate the effects of fur subsamples (fur tip, fur middle, and fur root) and sampling time (2019 and 2020) on the stable isotope values of fur. The coefficient of variation (COV = SD/mean) was used to describe the degree of variation in the δ¹³C and δ¹⁵N values of different tissue samples. Pearson correlation analysis was performed to determine the relationship between the nutrient composition of each food source and its proportional contribution to the diet of Père David's deer. The values of nutrients were determined individually for each plant in the four groups (C₃-F, C₃-G, C₄-G, and Other). The values of the proportional contribution to the diet were determined by the mean of the Bayesian mixture model of the three potential food sources (C₃-F, C₃-G, and C₄-G) in the selected samples, and the proportional contribution to the diet for plants (Other) not consumed by Père David's deer was 0. To reduce the error caused by the difference in season and year, three samples (Collected Fur Middle in 2019 and 2020, Feces) were selected for analysis in which the response dietary time window overlapped with the time of plant collection. The Pearson correlation coefficient (r) and p-values were determined using a two-tailed Pearson correlation analysis. All tests were two-tailed, and the acceptable significance level was α = 0.05. The results are reported as the mean ± 1 SD unless otherwise stated. All statistical analyses were performed using SPSS 19.0 (SPSS for Windows; SPSS, Inc.).

RESULTS Stable isotope values of potential food sources and samples of Père David's deer

A total of 16 species of plants were collected in this survey, 14 of which were potential food sources for Père David's deer (Table 1). Arundo donax and Polygonum perfoliatum (Other) have been studied to confirm that they are not consumed by Père David's deer (Ding, 2017; Zhang, Fu, et al., 2019; Zhang & Yang, 2017). One-way ANOVA showed that the δ¹³C and δ¹⁵N values of the three types (C₃-F, C₃-G, and C₄-G) of plants were significantly different (δ¹³C, df = 2/45, F = 1189.73, p < .001; δ¹⁵N, df = 2/45, F = 4.86, p = .012). Tukey's post hoc test showed that the difference in δ¹⁵N values between C₃-G and C₄-G was not significant (df = 1/41, F = 0.06, p = .807). Other group of Arundo donax had the lowest δ¹³C value of −27.93 ± 0.47‰. Based on the one-way ANOVA of data from different sections in the same year, there was a significant difference in the δ¹⁵N value of fur subsamples in different sections in 2019 (df = 2/9, F = 0.43, p = .043), and a significant difference in the δ¹³C value of fur subsamples in different sections in 2020 (df = 2/9, F = 258.77, p < .001). Two-way ANOVA revealed that the fur of different sections had no significant differences in δ¹³C and δ¹⁵N values (Table 2). Sampling time (2019 and 2020) only significantly influenced the δ¹⁵N values of the fur. The interactions between sampling time and section significantly affected the δ¹⁵N values of fur (Table 2).

TABLE 1 Classification, carbon, and nitrogen stable isotope ratios of plants in Shishou Milu National Nature Reserve

Type	Family	Species	Number	δ13C (‰)	δ15N (‰)	Degree of appetite
C3-F	Umbelliferae	Daucus carota	3	−31.84 ± 0.64	−0.70 ± 0.84	++
	Umbelliferae	Apium graveolens	3	−31.98 ± 0.02	0.96 ± 0.04	+
	Polygonaceae	Polygonum lapathifolium	3	−29.56 ± 0.20	−2.99 ± 1.07	++
	Labiatae	Leonurus artemisia	3	−32.65 ± 0.32	−1.31 ± 0.35	+
	Asteraceae	Hemistepta lyrata	6	−31.88 ± 0.62	−0.55 ± 0.92	+
	Geraniaceae	Geranium carolinianum	3	−30.98 ± 0.23	−1.15 ± 0.04	+
	Plantaginaceae	Plantago asiatica	3	−31.92 ± 0.27	−3.09 ± 0.24	++
	Leguminosae	Trifolium repens	3	−31.85 ± 0.61	−1.96 ± 0.69	++
	Brassicaceae	Capsella bursa-pastoris	3	−30.17 ± 0.79	3.03 ± 4.38	+
	Cyperaceae	Carex spp.	3	−31.67 ± 0.33	1.40 ± 3.03	+
		33	−33.08 ± 1.48a	2.58 ± 2.69b
C3-G	Gramineae	Lolium perenne	3	−32.29 ± 0.13	3.64 ± 2.70	+
	Gramineae	Roegneria kamoji	3	−34.26 ± 2.02	0.98 ± 2.41	+
		6	−31.32 ± 0.97b	−0.67 ± 2.54a
C4-G	Gramineae	Cynodon dactylon	6	−14.47 ± 0.37	0.01 ± 0.47	++
	Gramineae	Phragmites australis	3	−13.12 ± 0.26	−1.70 ± 0.52	++
		9	−14.24 ± 0.68c	−0.99 ± 1.49a
Other	Gramineae	Arundo donax	3	−27.93 ± 0.47	−3.06 ± 0.81	−
	Polygonaceae	Polygonum perfoliatum	3	−31.63 ± 0.35	−3.72 ± 0.67	−

Note: Type: C₃-F (C₃ forbs), C₃-G (C₃ grasses), C₄-G (C₄ grasses), and Other (avoided). The last line of each type is the data after the combination of that type and different letters denote significant differences at p < .05. “+” and “−” indicate the food preference of Père David's deer (Ding, 2017; Zhang, Fu, et al., 2019; Zhang & Yang, 2017).

TABLE 2 A summary of two-way ANOVA for δ¹³C and δ¹⁵N values of fur divided into three sections (fur tip, fur middle, and fur root) and sampling time (2019 and 2020) were selected as treatments

Variable	df	δ13C		δ15N
		F	p	F	p
Section	2	2.713	.093	1.350	.284
Sampling time	1	0.080	.780	213.938	.000
Section × Sampling time	2	1.160	.336	6.189	.000

Figure 2 shows the relationship between different tissues and potential food sources. The muscle, liver, and heart samples were collected from the same Père David's deer. The δ¹³C values of fur, feces, muscle, liver, and heart from Père David's deer were close to those of C₃ plants (Figure 2). We analyzed the stable isotope signatures of different tissues and found that the stable isotope signatures of Père David's deer vary among tissue types. We observed the lowest enrichment in the feces (δ¹³C: −29.87 ± 0.64‰, δ¹⁵N: 1.70 ± 1.7‰), followed by the liver (δ¹³C: −26.48 ± 0.31‰, δ¹⁵N: 4.97 ± 0.30‰). This is the same as our predictions for the fecal reflect the diet of the last days, the liver for approximately the last month, and other tissues reflect a longer time window (Bahar et al., 2014; Caut et al., 2009; Phillips et al., 2014; Roth & Hobson, 2000, Sponheimer et al., 2006). The deviations of δ¹³C values of each tissue were small (all COV < 0.05). The deviations of δ¹⁵N values of muscle, liver, and heart were similar (all COV < 0.1). Although feces and fur were collected from many different Père David's deer, the deviations in the δ¹⁵N values of feces (COV = 0.98) were larger than those of fur (all COV < 0.1).

FIGURE 2. Stable isotope signatures of carbon and nitrogen from potential food sources (solid points and error bars, corrected for TEF) and raw data from consumers (fur, muscle, liver, heart, and feces of Père David's deer) in scatter plots. The upper right corner indicates the diet window for the sample response.

Diet composition estimates of Père David's deer

The results of the Bayesian mixture model indicated that the C₃ plants (including C₃-G and C₃-F) were the main diet component for Père David's deer in Hubei Shishou Milu National Nature Reserve. However, there were differences in the contribution of food sources reflected by different tissues of Père David's deer.

The segmented fur samples indicated the composition of the diet at different times. The segmented fur samples (fur tip, fur middle, and fur root) for 2019 indicated that Père David's deer consumed mainly C₃-G (42.7% [11.9–65.3], 45.2% [12.2–68.0], and 42.8% [9.5–68.0]), followed by C₄-G (32.6% [25.6–.6], 31.9% [24.8–40.2], and 32.0% [22.3–43.4]), and C₃-F (24.7% [4.7–.4], 22.9% [3.5–52.9], and 25.2% [4.5–55.9]), respectively (Figure 3). The segmented fur samples for 2020 showed the same trend as those in 2019. Père David's deer consumed mainly C₃-G (36.5% [8.9–62.1], 37.3% [9.9–64.0], and 38.0% [8.7–65.5]), as well as C₄-G (34.8% [28.2–41.8], 31.6% [25.6–38.1], and 30.9% [24.1–38.1]) and C₃-F (28.7% [6.2–53.6], 31.1% [7.1–56.4], and 31.1% [6.7–57.7]), respectively (Figure 3).

FIGURE 3. Proportional contribution of potential food sources (50%, 75%, and 95% confidence intervals) to the diet of Père David's deer in different years (2019 and 2020) using segmented fur (tip, middle, and root) and estimated using stable isotopic mixing models.

The muscle, liver, and heart samples also showed the composition of the summer diet of this unexpectedly dead Père David's deer. Among them, muscle samples showed that the Père David's deer consumed mainly C₃-F (39.0% [10.3–62.0]) but also C₃-G (30.6% [6.0–62.1]) and C₄-G (30.4% [24.9–36.1]) between May and June (Figure 4). Heart samples showed that Père David's deer consumed mainly C₃-F (41.5% [7.4–73]), followed by C₃-G (39.2% [6.0–78.1]) and C₄-G (19.3% [9.3–32.8]) between June and July (Figure 4). Liver samples showed that the Père David's deer consumed mainly C₃-F (41.6% [14.1–63.2]), followed by C₄-G (30.3% [21.0–38.7]) and C₃-G (28.1% [5.5–59.6]) in July (Figure 4).

FIGURE 4. Proportional contribution of potential food sources (50%, 75%, and 95% confidence intervals) to the diet of Père David's deer using different tissue (muscle, liver, heart, and feces) and estimated using stable isotopic mixing models.

Fecal samples represent dietary composition over the previous days and present more individual variability (Phillips et al., 2014). We found that Père David's deer consumed mainly C₃-G (57.2% [26.2–80.4]) in January, and the contribution of C₃-G in fecal samples was much higher than that in other samples, followed by C₃-F (26.3% [5.3–54.3]) and C₄-G (16.5% [12.4–21.0]) (Figure 4).

Nutrient content of the potential food source

The contents of various nutrients from these 16 representative plants are listed in Table 3. The moisture content of different plants ranged from 70% to 80%. The difference between C. dactylon with the highest crude fat content and P. australis with the lowest crude fat content of grasses was 0.8%. The range of crude fat content of forbs was larger than that of grasses, from 2.5% to 6.6%. The C/N ratios of C. dactylon and P. asiatica were 20.09 and 18.54, respectively, which were higher than those of other plants in the range of 9%–13%. The nitrogen content of C₃ plants was higher than that of C₄ plants, but the C/N ratio of T. repens with the highest nitrogen content was the lowest. There were differences in soluble sugar content and crude protein content among different plant types (soluble sugar: df = 3/44, F = 6.18, p = .010; crude protein: df = 3/44, F = 11.93, p < .001). Other (avoided) group of plants (Arundo donax and Polygonum perfoliatum) had the lowest soluble sugar content. Only P. lapathifolium and Apium graveolens contained condensed tannins, and the condensed tannins of P. lapathifolium were higher than those of A. graveolens.

TABLE 3 Nutrient content of plant in Shishou Milu National Nature Reserve

Type	Plant	Moisture content (%)	Crude fat content (%)	C/N	N content (%)	Soluble sugar (%)	Crude protein (%)	Condensed tannins
C3-G	Lolium perenne	74.58 ± 0.50	4.60 ± 0.05	10.92 ± 2.42	4.12 ± 1.15	3.47 ± 0.20	11.87 ± 0.25	−
	Roegneria kamoji	73.94 ± 0.29	5.10 ± 0.69	14.01 ± 1.45	3.53 ± 1.37	2.19 ± 1.23	9.47 ± 0.69	−
C3-F	Daucus carota	80.60 ± 0.11	6.61 ± 0.21	11.88 ± 2.23	3.55 ± 0.71	1.53 ± 0.76	6.92 ± 0.94	−
	Apium graveolens	80.67 ± 0.10	5.17 ± 0.87	11.72 ± 0.25	2.95 ± 0.37	2.16 ± 0.20	4.59 ± 0.23	+
	Polygonum lapathifolium	80.34 ± 0.54	4.83 ± 0.38	13.07 ± 0.12	3.48 ± 0.08	2.41 ± 0.21	9.16 ± 0.87	++
	Leonurus artemisia	77.39 ± 0.73	5.21 ± 0.76	12.42 ± 0.63	2.98 ± 0.54	1.52 ± 0.29	5.36 ± 0.31	−
	Hemistepta lyrata	78.75 ± 0.14	3.76 ± 0.97	12.44 ± 0.82	3.43 ± 0.57	2.49 ± 0.48	11.65 ± 0.55	−
	Geranium carolinianum	73.59 ± 0.02	2.68 ± 0.54	13.14 ± 0.11	3.38 ± 0.03	1.76 ± 0.47	4.92 ± 0.44	−
	Plantago asiatica	81.38 ± 0.97	3.92 ± 0.59	18.54 ± 0.08	2.43 ± 0.06	1.39 ± 0.36	10.41 ± 0.76	−
	Trifolium repens	79.78 ± 0.08	3.46 ± 0.23	9.20 ± 1.13	5.49 ± 1.46	1.43 ± 1.13	8.33 ± 0.02	−
	Capsella bursa-pastoris	76.00 ± 0.36	3.79 ± 0.12	11.29 ± 1.00	3.57 ± 0.19	2.13 ± 1.78	8.84 ± 0.61	−
	Carex spp.	70.35 ± 0.90	2.56 ± 0.29	11.04 ± 0.69	3.90 ± 0.21	2.35 ± 0.28	8.85 ± 0.16	−
C4-G	Cynodon dactylon	71.25 ± 0.79	5.21 ± 0.59	20.09 ± 0.32	2.28 ± 0.21	3.28 ± 0.85	7.69 ± 0.56	−
	Phragmites australis	72.57 ± 0.54	4.41 ± 0.74	13.87 ± 0.16	2.88 ± 0.07	2.24 ± 0.62	8.97 ± 0.69	−
Other	Arundo donax	73.76 ± 0.27	5.04 ± 0.34	11.20 ± 0.12	3.92 ± 0.05	1.09 ± 0.25	3.57 ± 0.09	−
	Polygonum perfoliatum	77.81 ± 0.24	4.57 ± 0.09	12.82 ± 0.18	3.21 ± 0.29	1.37 ± 0.33	4.26 ± 0.30	−

Note: Type: C₃-G (C₃ grasses), C₃-F (C₃ forbs), C₄-G (C₄ grasses), and Other (nonedible). “+” and “−” indicate the presence or absence of condensed tannins.

The Pearson correlation test was used to analyze the correlation between the nutrient composition of each food source and its proportional contribution to the diet of Père David's deer (Table 4). In the three samples selected, soluble sugar and crude protein were correlated (soluble sugar: df = 14, r = .695, t = 3.62, p = .003; df = 14, r = .543, t = 2.42, p = .030; df = 14, r = .501, t = 2.17, p = .048 in order; crude protein: df = 14, r = .666, t = 3.34, p = .005; df = 14, r = .636, t = 3.09, p = .008; df = 14, r = .622, t = 2.97, p = .010 in order) with the proportional contribution of the four source groups (C₃-G, C₃-F, C₄-G, and Other) to the diet.

TABLE 4 Correlations between nutrient composition of each food source and its proportional contribution (C₃-G [mean], C₃-F [mean], C₄-G [mean], and Other [0]) to the diet of Père David's deer by Pearson correlation analysis

	Selected sample	Moisture content/%	Crude fat content/%	C/N	N content/%	Soluble sugar/%	Crude protein/%
Proportional contributions of each dietary source	Fur Middle 2019	−0.226	0.044	0.186	−0.030	0.695*	0.666*
	Fur Middle 2020	0.013	−0.102	0.125	−0.030	0.543*	0.636*
	Feces	−0.014	−0.005	−0.094	0.200	0.501*	0.622*

^*Indicate a significant correlation at the 0.05 levels (two-sided).

DISCUSSION Diet composition of Père David's deer

Research on the feeding habits of ungulates is essential to understanding the interaction between wild animals and the environment. It is also the basis for evaluating population ecology issues such as habitat quality, environmental tolerance, and feeding strategies of wild animals (Zhang et al., 2019). We measured diet composition in multiple tissues, and observed differences in diet composition among feces, fur, and other tissues, which will be useful for determining variations in the diet composition of Père David's deer. We observed that the autumn and winter diets estimated by fur samples indicated a diet dominated by C₃-G (C₃ grasses, 42.7%–57.2%, mean), and the lowest proportional contribution of C₃-F (C₃ forbs, 22.9%–31.1%, mean) to the diet. The summer diet estimated by muscle and liver samples showed that the C₃-F (30.9%–41.6%, mean) was the main plant type consumed by Père David's deer. In the one-way ANOVA and two-way ANOVA of fur subsamples, we found that the results of the effect of the factor sectioning on the stable isotope signature of the fur subsamples were not the same. However, combined with the analysis of the Bayesian mixture model, we believe there is little variation in the composition of the winter diet of Père David's deer. We speculated that the reason for the lack of significant change is that the diet of the Père David's deer was similar during this time. Alternatively, for Père David's deer, fur may have been integrated into the average diet during the entire fur growth (Rogers et al., 2020).

Comparing tissues that respond to dietary information on different time scales could provide more complex response information on food selection (Rogers et al., 2020). We observed that C₃-G was consumed much more from November to January (compared with other months in SIMMr with a probability of approximately 0.41). This is consistent with the results from the fecal samples (collected in January), which demonstrated greater C₃-G consumption. Careful analysis of the food consumed by Père David's deer revealed a larger range of proportional contributions from G₃-G and C₃-F. We speculate that this is related to the fact that the Père David's deer consumed more C₃ plants than C₄ plants. The deviations of δ¹⁵N values of feces (COV = 0.98) were larger than those of fur (COV < 0.1). Fecal samples represent the dietary composition over the previous days and present more individual variability. This indicated that the day-to-day diet is variable, but relatively consistent when considered over several months (Meng et al., 2010; Wang & Wang, 2011). Muscle, heart, and liver samples were all from the same Père David's deer, which affects our prediction of the summer diet composition. However, considering the little variation in habitats (Zhang, Fu, et al., 2019; Zou et al., 2013) and that Père David's deer are gregarious (Ding, 2017), our results can still provide information on the summer diet composition of Père David's deer. Smaller sample sizes may also affect the analysis of diets (Phillips et al., 2014), but we can still reveal dietary trends throughout the seasons in endangered species by using several body tissues that have different turnover time.

Similar to Wang and Wang (2011), who found that Père David's deer consumed more C₃-F plants in spring and summer, we also found that C₃-F became the main food consumed from May to July. This change is the opposite of the autumn and winter months (October to March). We speculate that these changes may be caused by differences in species richness due to seasonal variations in resources (Seto et al., 2015; Taillon et al., 2006). C₃-F biomass is higher in spring and summer and lower in autumn and winter, while graminoids (C₃-G and C₄-G) remain higher in autumn and winter (Zhang, Li, et al., 2019). This change in diet, brought about by changes in biomass, is more evident in the reintroduction of Père David's deer in the Jiangsu wetland in China. In the coastal wetland of Jiangsu Dafeng Milu National Nature Reserve, exotic S. alterniflora has invaded and become the dominant species (Bao & Shi, 2007; Wang et al., 2006; Zhang et al., 2008). Consequently, S. alterniflora gradually became the main food source of Père David's deer, which resulted in a relatively narrow selection of diet composition (Ding, 2009; Ji et al., 2011; Zhang, 2015; Zhao et al., 2010). When the preferred high-quality food resources are limited, ungulates are forced to use low-quality food resources (Gebert & Verheyden-Tixier, 2001; Miranda et al., 2012). This change in feeding behavior is also consistent with other Cervidae (Dumont et al., 2005; Johnson et al., 2001; Zhang et al., 2020).

Our results were different from the diet composition of Père David's deer in northern China (mainly in Jiangsu Dafeng Milu National Nature Reserve and Beijing Milu Ecological Research Center), where C₄ plants were the main food source (70.36%) (Wang & Wang, 2011; Zhang, 2015). Although the change in habitat shows a change in the diet composition of Père David's deer, gramineous plants remain the main food source of Père David's deer (Ding, 2009; Wang & Wang, 2011; Zhang, 2015). We believe this difference in diet is due to differences in habitat types. The geographical location, climate, and other ecological factors of coastal wetlands (Jiangsu Dafeng Milu National Nature Reserve) and lake wetlands (Hubei Shishou Milu National Nature Reserve) influence the diet. The changes in the diet brought about by habitat change prove that the temporal and spatial patterns of plant species and their abundance may be a key factor in determining the diet composition. A study on the correlation between feeding habits of Cervus nippon and geographical location also supported this inference (Takatsuki, 2009). C. nippon mainly feed on Sasa nipponica and other graminoids in southern Hokkaido (northern Japan), while they mainly feed on browse and fruits such as Aucuba japonica, Eurya japonica, and acorns of Lithocarpus edulis in southern Japan (Campos-Arceiz & Takatsuki, 2005; Endo et al., 2017; Takatsuki, 2009). We also observed that Père David's deer were mixed feeders, similar to most Cervidae (Meng et al., 2010), despite differences in the diet composition of Père David's deer and other Cervidae (Zhong et al., 2019). For example, Cervus nippon and Cervus elaphus are also typically mixed feeders (Zhong et al., 2019), with trees and shrubs comprising their main food sources (Cui et al., 2007; Gebert & Verheyden-Tixier, 2001; Krojerová-Prokešová et al., 2010). Thus, we suggest that the availability of food items is an important factor influencing diet composition of Père David's deer. In turn, habitat variation affects the availability of food items (Zhong et al., 2019). This provides direction for the reintroduction of Père David's deer as a targeted habitat restoration and artificial intervention in their diet (Beest et al., 2010).

Effects of nutrient content on diet selection of Père David's deer

The knowledge accumulated in the field of nutritional ecology shows with increasing clarity that animal metabolism and diet selection are associated with synergistic and antagonistic assimilation strategies (Felton et al., 2016). We observed that the crude protein and soluble sugars of plants showed a correlation with their proportional contribution to the diet (Table 4). The results indicated that Père David's deer selectively fed on plants with high protein and soluble sugar contents in autumn and winter. The energy and nutrients of food are absorbed and distributed to various physiologic functions (Boggs, 2009). Protein is one of the most important nutrients for ungulates (Smith, 1978). For most ungulate herbivores, obtaining adequate protein is an important factor affecting diet selection in winter (Berteaux et al., 1998; Demment & Soest, 1985; Illius & Gordon, 1990; Jarman, 1974; Workman & Schmitt, 2012). A study showed that to maintain the metabolic requirements of Odocoileus virginianus in winter, the crude protein content in the diet should be within 13%–16% of the dietary intake (Soest, 2018). In winter, the mean crude protein content of the 14 plant species mainly consumed by Père David's deer in Shishou was 8.36% (Table 3). The crude protein content in the diet of C. elaphus in Europe reaches 5.7% and 5.0% to ensure their protein requirements (Maloiy et al., 1970; Verheyden-Tixier et al., 2008; Yousef Elahi & Rouzbehan, 2008). This suggests that Père David's deer in Shishou can obtain protein to ensure basic needs in winter. Soluble sugars are also an important energy source for ungulates (Zhang, 2015). Soluble sugars increase rumen fluid volume and dilution rate (Schingoethe et al., 1980; Windschitl & Stern, 1988). It also promotes the absorption and utilization of protein in ruminant (Beever et al., 1978). A study by Verheyden-Tixier et al. (2008) on selection of nutrients in red deer indicated that soluble sugar content was an important factor influencing diet selection and was more important than protein content. This is similar to the results of Hobbs et al. (1983), who compared the nutritional ecology of montane ungulates during winter, that ungulates have a strong preference for plants rich in digestible soluble sugars. Our result also is consistent with the diet selection tendency of Père David's deer in other reserves (Jiangsu Dafeng Milu National Nature Reserve and Beijing Milu Ecological Research Center; Wang & Wang, 2011; Zhang, 2015).

Our studies indicated the presence of plants with high condensed tannin levels in the diet of Père David's deer, such as P. lapathifolium and Apium graveolens, and some studies have found that Père David's deer are fond of eating P. lapathifolium (Ding, 2017; Zhang, Fu, et al., 2019; Zhang & Yang, 2017). Tannin, as a typical plant secondary metabolite, has attracted extensive attention. Tannins have the ability to bind and precipitate proteins, which affects the protein and nitrogen retention rates of ungulates (Estell, 2010; Qiu, 2016). Tannins can combine with carbohydrates to a certain extent and affect the nutrient absorption of food by ungulates (Mueller-Harvey, 2006). The intake of tannins by Père David's deer is similar to that by other Cervidae. Bergvall and Balogh (2009) and Bergvall and Leimar (2005) showed Dama dama consumed high-tannin food even in the presence of a low-tannin option. The behavior confirmed that the Cervidae could keep the plant's secondary metabolites at safe levels by adjusting feeding patterns and their bioinvertase system (Chapman et al., 2010; Champagne et al., 2020; Sorensen et al., 2005; Verheyden-Tixier & Duncan, 2000). These results indicate that tannins affect diet selection, but have little effect on Cervidae diets in natural environments (Champagne et al., 2020).

CONCLUSIONS

Our results showed that the autumn and winter diets estimated by fur and fecal samples indicated a diet dominated by C₃ grasses (42.7%–57.2%, mean), and the summer diet estimated by muscle and liver samples was dominated by C₃ forbs (30.9%–41.6%, mean). The Pearson correlation test indicated that crude protein and soluble sugars were important factors influencing the winter diet selection of Père David's deer to some extent. The results obtained therefore suggest the existence of seasonal dietary variation and possible targeted selection for nutrients in Père David's deer. The use of different samples combined with overlapping time periods responds to the dietary variation, which to some extent solves the problem of analytical precision caused by the difficulty of sampling endangered animals. The rich plant resources of the Tian'E Zhou Lake wetland provide a good foundation for the development of the Père David's deer population, but the increase in potential food sources also provides a challenge for the accurate analysis of stable isotopes. We suggest improving the precision of stable isotope analysis by combining different dietary analysis methods to further differentiate and a priori potential food sources. Additionally, to determine the nutritional strategy, a long-term follow-up survey of individual Pere David deer is recommended. The observation of diet structure and selection for nutrients is the basis for assessing population development or establishing long-term effective conservation measures for endangered species. Comprehensively evaluating the diet selection mechanisms for Père David's deer and carefully maintaining the reserve should be highly supported, ongoing goals.

AUTHOR CONTRIBUTIONS

Hao-Lin Wang: Conceptualization (supporting); data curation (equal); formal analysis (lead); investigation (equal); methodology (equal); software (lead); visualization (equal); writing – original draft (lead); writing – review and editing (supporting). Yue Zhao: Data curation (supporting); investigation (equal); methodology (supporting); writing – original draft (supporting). Fei-Jie Wang: Conceptualization (supporting); formal analysis (supporting); investigation (equal); methodology (supporting); software (supporting). Xin-Jia Sun: Conceptualization (supporting); formal analysis (supporting); investigation (equal); methodology (supporting); software (supporting). Jian-Qiang Zhu: Conceptualization (supporting); funding acquisition (supporting); methodology (supporting); resources (equal); writing – review and editing (supporting). Yu-Ming Zhang: Funding acquisition (supporting); investigation (supporting); resources (equal). Shu-Dong Wei: Conceptualization (equal); investigation (equal); project administration (equal); resources (supporting); supervision (supporting); writing – review and editing (supporting). Hui Chen: Conceptualization (equal); data curation (equal); funding acquisition (lead); investigation (equal); methodology (equal); project administration (equal); resources (equal); supervision (equal); visualization (equal); writing – review and editing (lead).

ACKNOWLEDGMENTS

We thank Professor Fenxiao Huang for her assistance in plant sorting and identification. We thank Shishou David's Deer National Nature Reserve for supporting our study in the field. We also thank Pengfei Li, Xuemeng Tang, and Jingyi Dai for their assistance in the field and/or laboratory.

FUNDING INFORMATION

This research was financially supported by the Science and Technology Research Project from the Hubei Provincial Department of Education (D20221301), the Engineering Research Center of Ecology and Agricultural Use of Wetland, Ministry of Education (Yangtze University) (KFT202101), the Opening Project of Henan Province Key Laboratory of Water Pollution Control and Rehabilitation Technology (CJSP2022003), and the Innovation and Entrepreneurship Training Plan for Undergraduate in Hubei Province (S202010489074, Yz2022227).

CONFLICT OF INTEREST

The author declares no conflict of interest.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are openly available in Dryad at https://doi.org/10.5061/dryad.sqv9s4n7c.