大型食草动物在生态系统中具有关键的功能，向上支撑顶级食肉动物种群，向下控制植物群落的组成和结构，在生态系统结构和功能维持中发挥着至关重要的作用。东北虎豹国家公园旨在保护以东北虎(Panthera tigris altaica)、东北豹(P. pardus orientalis)为旗舰物种的生态系统，是全球生物多样性关注和研究的关键区域。梅花鹿(Cervus nippon)是该区域的主要大型食草动物之一，是维持虎豹国家公园健康生态系统功能的关键。动物的食性选择是整个生态系统相互作用的基础，因此获取高精度的野外食性格局及其时间、空间变异，分析这种格局背后的驱动机制，是理解大型食草动物生态系统功能的根本。本研究利用非损伤采样法、高通量测序技术及DNA宏条形码技术，获取虎豹国家公园梅花鹿的野外食性格局，并结合可食用植物资源调查、植物营养组成分析、次生代谢物分析以及自助餐控制实验，探究食性格局背后的驱动机制。

本研究进行的梅花鹿的定量饲喂实验发现粪便中各物种的测序序列数比例与饲喂比例呈显著正相关，这表明分子食性数据可以基本用于定量反映食物组成。利用野外粪便样品进行分子食性分析结果显示梅花鹿共取食34个科的植物，以双子叶植物为主，包括乔木、灌木、草本多种生活型，以乔木为主。梅花鹿的食性格局存在季节差异，秋冬两季的多样性显著高于春夏两季，食物组成在季节上呈现连续变化，相邻季节的食物组成相似性最高。不同生境的食物组成具有显著差异，蒙古栎林的食物物种丰富度和多样性显著高于杂木林。

总体而言，梅花鹿的野外食性格局符合广食性植食动物的预期，是偏游食者类型，存在时间和空间的变异。根据最优觅食理论，食性选择应符合最小的能量输出，获取最大的能量和营养收益。对于大型食草动物的食性选择机制，目前有三种主流的假说：(1) 频率依赖选择机制：根据食物的可获得性进行选择；(2) 营养驱动的选择：倾向于选择营养和能量更高的食物；(3) 次生代谢物驱动的避毒选择：次生代谢物可能影响营养的吸收效率，所以倾向于选择次生代谢物含量低的食物。为了验证这三种机制在梅花鹿食性格局中的作用，本研究调查了梅花鹿的可食用资源的空间分布、测定了植物营养组成和次生代谢物含量，开展了自助餐实验，控制投喂食物频率、营养和次生代谢物，以此分析梅花鹿的食性选择行为。

一、频率依赖在食性选择中的作用

空间异质性可能导致空间相近的个体面对更加相似的食物资源，如果存在频率依赖，则期望具有更相似的食性选择。我们的研究结果与期望一致，在四个季节以及两种生境中，都存在采样点间距离更近，其食物组成更相似的显著正相关关系。

我们利用样方法调查了可食用资源的空间分布，并分析可食用资源相对丰度和进食比例的相关性。结果显示可食用资源相对丰度和进食比例具有相关性，但是存在季节、生境和不同生活型的变异，主要发现如下：(1) 对于不同生活型植物，乔木的可食用资源相对丰度和进食比例显著正相关，灌木不显著，草本也存在显著相关性，这表现在夏季草本的可食用资源相对丰度和进食比例显著正相关，此外冬季的草本总进食量也显著低于其他季节，符合冬季草本生物量显著降低的事实；(2) 在不同季节，春季和冬季可食用资源相对丰度和进食比例具有显著的相关性，而夏季和秋季不显著；(3) 在不同生境下，蒙古栎林中可食用资源相对丰度和进食比例具有显著正相关，但在杂木林不显著。以上结果表明频率依赖选择机制在梅花鹿食性格局中扮演着重要的角色，并且存在季节、生境以及不同生活型变异。

频率依赖选择一般认为是基于可获得性资源的被动选择，被动选择不一定代表最优选择。然而大型食草动物为了最大化能量收益，这种频率依赖的选择也可能代表其最偏好的主动选择。为了验证这一假说，我们设计了自助餐实验，在投喂食物量一致的情况下，分析梅花鹿的食物选择行为。结果表明，当投喂食物量一致，梅花鹿仍存在食物选择和偏好，且自助餐实验中的偏好和野外食性格局具有显著的正相关性，说明梅花鹿对生境内群落优势种存在主动选择，并且与现有生境具有较好的匹配关系。

二、营养组成在食性选择中的作用

我们测定了18种食物春夏秋三个季节10个营养指标，这些食物分属14个科，在梅花鹿食性中具有不同的进食比例。在物种水平上，绝大部分营养指标在所有季节和生境中都和进食比例没有显著相关。根据食物在种群中和个体中的进食比例，食物被分为主要食物、次要食物、偶吃食物，并比较这三类食物营养组成的差异。主要发现如下：(1) 食物的营养组成具有季节变异，春季粗蛋白最高，夏季次之，秋季最低；总能在季节间没有显著差异；(2) 不同类别食物的营养组成存在差异，即营养选择格局，这种格局具有季节性。夏季具有最明显的营养选择格局，其主要食物的粗蛋白含量和总能都显著高于偶吃食物，春季主要食物的总能显著较高，秋季无显著差异；(3) 不同生境的营养选择格局也存在差异。在蒙古栎林中春季存在对总能的选择，夏季存在对粗蛋白和总能的选择，秋季不存在明显的营养选择格局；而在杂木林中春夏秋三个季节都存在明显的营养选择格局。以上结果表明营养驱动选择和频率依赖选择作用形成互补，不同季节、不同生境可能以不同的选择机制为主。频率依赖机制在春季、冬季、蒙古栎林中更显著，而营养驱动选择在夏季和杂木林更显著。

三、次生代谢物在食性选择中的作用

我们同样测定了春夏秋三个季节上述18种食物的总单宁、总酚、类黄酮和总生物碱，并比较了其与进食比例的关系。主要发现如下：(1) 食物的次生代谢物含量具有季节差异。总单宁在春季和夏季显著高于秋季；类黄酮春季最高，夏天次之，秋季最低；总生物碱在各个季节无显著差异；(2) 食物的次生代谢物含量和进食比例的关系存在季节差异。物种水平上，仅夏季类黄酮含量和进食比例表现出与避毒期望一致的负相关（仅边缘显著），而在春季，类黄酮含量和进食比例呈显著正相关。(3) 次生代谢物含量与进食比例的关系在不同生境中的格局相似。此外，在不同类别食物中不存在主要食物的次生代谢物含量显著低于偶吃食物。以上结果表明次生代谢物在梅花鹿食性选择中影响较弱，仅可能出现在食物资源极大丰富的夏季食性选择中。

为了进一步验证营养和次生代谢物在食性选择中的相对作用，我们设计了自助餐实验，包含高蔗糖低单宁、高蔗糖高单宁、低蔗糖低单宁和低蔗糖高单宁四种食物。研究结果表明梅花鹿显著偏好高蔗糖食物，但是对低单宁的选择仅边缘显著。该结果与野外研究结果一致，证实营养驱动的选择优先于次生代谢物的避毒选择。

本研究是迄今为止第一次系统地探究动物的食性格局及其成因。我们获取了高精度的野生梅花鹿的食性格局，发现了其在时间和空间的变异特征。通过野外调查和自助餐实验，本研究证实了频率依赖选择在梅花鹿食性格局中扮演着重要的角色，而营养驱动选择和频率依赖选择作用形成互补，在不同季节、不同生境可能以不同的选择机制为主。相较而言，次生代谢物的避毒选择在梅花鹿食性选择中影响较弱。尽管频率依赖选择是一种被动选择，但是梅花鹿表现出对优势物种的食性偏好可能是其与现有生境具有较好的匹配关系。

Large herbivores have key functions in the ecosystem by supporting top carnivore populations upwards and controlling the composition and structure of plant communities downwards, playing a crucial role in the maintenance of ecosystem structure and function. The Northeast Tiger and Leopard National Park aims to protect the ecosystem with the Amur Tiger (Panthera tigris altaica) and Amur Leopard (P. pardus orientalis) as flagship species, and is a key area of global biodiversity concern and research. The sika deer (Cervus nippon) is one of the major large herbivores in this region and is essential for maintaining healthy ecosystem function in the park. The feeding habits of animals are fundamental to the interactions of the entire ecosystem. Obtaining high-accuracy wild feeding habits and their temporal and spatial variation, and analyzing the driving mechanisms underlying such feeding habits, are basic to understanding the function of large herbivore in ecosystem. In this study, we used the non-damage sampling method, high-throughput sequencing technology and DNA metabarcoding technology to obtain the wild feeding habits of sika deer in the Northeast Tiger and Leopard National Park, and combined with available food resource survey, plant nutrient composition analysis, secondary metabolite analysis and cafeteria test to investigate the driving mechanism of feeding habits.

The quantitative feeding experiments conducted in this study on sika deer show a significant positive correlation between the sequence count proportion in fecal samples and the feeding proportion of each species, indicating that molecular feeding data can be used to quantitatively reflect food composition. The molecular dietary results of wild sika deer showed a total of 34 families of plants, mainly dicotyledons, including trees, shrubs, and herbs with a variety of life-forms, mainly trees. There were seasonal differences in the feeding habits of sika deer, with significantly higher diversity in autumn and winter than in spring and summer. Food composition showed continuous seasonal changes, with the highest similarity in food composition in adjacent seasons. Food composition in different habitats varied significantly. Species richness and diversity in the Quercus forest had significantly higher than that in the Mixed forests.

Overall, sika deer is a generalist herbivore and a browser, and its feeding pattern varies in time and space. According to the optimal foraging theory, feeding habits should be consistent with minimal energy output for maximum energy and nutritional gain. For the feeding selection mechanism of large herbivores, there are three prevailing hypotheses: (1) frequency-dependent selection mechanism, which selects food based on its availability; (2) nutrition-driven selection, which favors food with higher nutritional and energy benefits; (3) secondary metabolite-driven avoidance selection, which favors food with lower secondary metabolite content that may affect the absorption efficiency of nutrients. To investigate the roles of these three mechanisms in feeding habits of sika deer, this study surveyed the spatial distribution of food resources, determined plant nutrient composition and secondary metabolite content, and conducted the cafeteria test to control the frequency of food administration, nutrition and secondary metabolites as a means to analyze the feeding choice behavior.

I. The role of frequency dependence in feeding choice

Spatial heterogeneity may lead individuals in spatial proximity to face more similar food resources. If frequency dependence exists, more similar feeding choices are expected. Our results were consistent with expectations, with a significant positive relationship between distance between sampling sites and differences in food composition in all four seasons as well as in both habitats.

We investigated the spatial distribution of food resources using a sampling method and analyzed the correlation between the relative abundance of food resources and feeding proportions. The results showed that the relative abundance of food resources and the feeding proportions were correlated, but there were seasonal, habitat and life-form variation. The main findings were as follows: (1) for different life-forms of vegetation, the relative abundance of food resources and food resources were significantly positively correlated for trees, but not for shrubs, and also for herbs, as shown by the significant positive correlation between the relative abundance of food resources and feeding proportions for herbs in summer, in addition to the fact that the total feeding amount of herbs was significantly lower in winter than in other seasons, which is consistent with the fact that herb biomass was significantly lower in winter; (2) In different seasons, between the relative abundance of food resources and feeding proportions were significantly positive correlated in spring and winter, but not in summer and autumn; (3) Under different habitats, between the relative abundance of food resources and feeding proportions were significantly positive correlated in Quercus forests, but not in Mixed forests. The above results suggest that frequency-dependent selection plays an important role in the feeding habits of sika deer, and that there is seasonal, habitat and life-form variation.

Frequency-dependent selection is generally considered to be passive selection based on availability of food resources, and passive selection does not always represent optimal selection. However, frequency-dependent selection may also represent their most preferred active choice for large herbivore in order to maximize energy gains. To verify this hypothesis, we designed a cafeteria test to analyze the food selection behavior of sika deer under a consistent amount of fed food. The results showed that when the amount of food fed was consistent, food choice and preference still existed, and there was a significant positive correlation between food preference in the cafeteria test and wild feeding habit. This result indicated that the sika deer actively selected the dominant species in the community, and had a good match with the existing habitat.

II. The role of nutritional composition in feeding choice

We measured 10 nutritional indices of 18 foods in three seasons, spring, summer and autumn. These foods belonged to 14 families and had different feeding proportions in the feeding habits of deer. At the species level, the vast majority of nutritional indicators were not significantly correlated with feeding proportions in all seasons and habitats. Foods were divided into main, secondary, and rare foods based on the feeding proportions in the population and in individuals, and differences in the nutritional composition of these three food types were compared. The main findings were as follows: (1) the nutritional composition of foods in general had seasonal variation, with crude protein being highest in spring, followed by summer and lowest in autumn; total energy did not differ significantly between seasons; (2) there were differences in the nutritional composition of different types of foods, i.e., nutritional selection patterns, and such patterns were seasonal in character. In summer, there is the most obvious nutritional selection pattern, the crude protein and total energy of main food are significantly higher than those of the rare food. In spring the total energy of main food is the highest, and in autumn there is no significant difference; (3) there are also differences in the nutritional selection pattern of different habitats. In the Quercus forest, there was a selection pattern for total energy in spring, crude protein and total energy in summer, and no obvious pattern of nutritional selection in autumn; while in the Mixed forest, there were obvious patterns of nutritional selection in spring, summer and autumn. The above results suggest that the roles of nutrient-driven and frequency-dependent selection form complementary, and different seasons and habitats may be dominated by different selection mechanisms. Frequency-dependent mechanism was more significant in spring, winter and Quercus forests, while nutrient-driven selection was more significant in summer and Mixed forests.

III. The role of secondary metabolites in feeding choice

We also measured total tannins, total phenols, flavonoids and total alkaloids in the above 18 foods in spring, summer and autumn, and compared the relationship between secondary metabolite content and feeding proportion. The main findings were as follows: (1) There were seasonal differences in the secondary metabolite content of the foods. Total tannins were significantly higher in spring and summer than in autumn; flavonoids were highest in spring, followed by summer and lowest in autumn; total alkaloids did not differ significantly among seasons; (2) there were seasonal differences in the relationship between secondary metabolite content and feeding proportion of foods. At the species level, flavonoid content and feeding proportion showed a negative correlation (only marginally significant) consistent with toxin avoidance expectations in summer, while in spring, flavonoid content and feeding proportion showed a significant positive correlation. (3) The relationship between secondary metabolite content and feeding proportions showed a similar pattern across habitats. In addition, there was no significant lower secondary metabolite content of the main food than the occasional food among different food types. The above results suggest that secondary metabolites have a weak effect on feeding habits of sika deer and may only occur in summer when food resources are greatly abundant.

To further verify the comparative role of nutrition and secondary metabolites in feeding choice, we designed a cafeteria test containing four food items: high sucrose low tannin, high sucrose high tannin, low sucrose low tannin and low sucrose high tannin. The results of the study showed that sika deer preferred the high sucrose food, but the selection of low tannin was only marginally significant. This result is consistent with the results of the field study and confirms the priority of nutrient-driven selection over secondary metabolite selection for toxicity avoidance.

This study obtained high-accuracy feeding habits of wild sika deer and found its spatial and temporal variability characteristics. Through field surveys and cafeteria tests, this study confirmed that frequency-dependent selection plays an important role in the feeding habit of sika deer, while nutrient-driven and frequency-dependent selection complement each other. Different selection mechanisms may be dominated by in different seasons and habitats. In contrast, toxic avoidance selection on secondary metabolites has a weaker influence in the feeding selection of sika deer. Although frequency-dependent selection is a kind of passive selection, sika deer may show a feeding preference for dominant species because it has a better match with the existing habitat.