1. Introduction

Functional traits of plants are key attributes related to plant colonization, growth, and survival (Violle et al., 2007); they reflect the adaptation of plants to the external environment and can determine ecosystem functions (Díaz et al., 2004; Gross et al., 2017). Consequently, functional trait-based approaches have been widely used to answer a variety of ecological questions, including plant life history strategies (Westoby et al., 2002; Kraft et al., 2015), community assembly (Lavorel and Garnier, 2002; McGill et al., 2006; Kraft et al., 2008; ), and the relationship between functional diversity and ecosystem functioning (Gross et al., 2017; Brun et al., 2022). It is generally assumed that trait constellations, rather than individual traits, are more effective in addressing plant adaptation and responses to environmental change (Díaz et al., 2016) because of the interdependency and coordination of multiple traits during plant growth and survival. The past two decades have seen persistent exploration of essential trait combinations, such as trait syndromes (Tjoelker et al., 2005), multidimensional trait space (Díaz et al., 2016; Mouillot et al., 2021), the economic spectrum of trait relationships in different plant organs (Wright et al., 2004; Osnas et al., 2013; Kong et al., 2019; Li et al., 2022a), and plant trait networks (He et al., 2020; Li et al., 2022b). Thus, the selection of representative traits and simplification of essential trait combinations reflecting plant strategies and environmental responses is a stepping stone to the development of trait-based plant ecology.

The variation in global plant traits has been simplified into two general dimensions, corresponding to strategies of resource economics and organ size. The first trait dimension, the leaf economics spectrum (LES), showed a systematic shift from conservative to acquisitive resource acquisition strategies as a result of a trade-off between leaf construction costs and growth potential (Wright et al., 2004, 2005). Subsequently, with the improvement in trait data availability, the second trait dimension of plant organ size was observed, linking whole-plant form and function (Díaz et al., 2016; Joswig et al., 2022). However, the current trait relationships do not include traits specifically related to cold resistance, which plays an important role in the life history strategies of alpine plants. Whether alpine plants have unique trait relationships because of their adaptation to alpine environments remains to be clarified.

Can the trait relationships that affect alpine plant growth, survival and stress resistance be captured by these two general trait dimensions? Alpine plants have evolved a set of traits to adapt to the harsh alpine environments of high altitude, low temperature, intensive solar radiation and high wind speed (Zhang et al., 1988), such as small leaf size, thick leaves (Hultine and Marshall, 2000), slow growth, and long lifespan (Geng et al., 2014). Beyond economics and size traits, non-structured carbohydrates (NSC) play critical roles in alpine plants survival and stress resistance as a result of the allocation trade-off of NSC to growth, metabolism and storage (Fig. 1). In the carbon (C) reserves of plants, part of the carbohydrates fixed by photosynthesis are allocated to plant organ construction, known as stable reserve of structural carbohydrates (SC), while other part serves as a mobile C pool in the form of NSC (Fig. 1). NSC mainly consist of starch and soluble sugars. The starch can be hydrolyzed to soluble sugars and translocated to other organs to support metabolic activities or new tissue growth (Herms and Mattson, 1992; Fig. 1). When carbon assimilation exceeds carbon demand, soluble sugars can be conducted to starch as a carbon storage for plant future use. However, NSC storage is also an active process, which could be conduct not only after all other carbon demands being met (Wiley and Helliker, 2012; Carbone et al., 2013). Under stressful conditions, plants tend to enhance construction cost to produce thicker leaves. Meanwhile, in mobile C pool, starch can be converted to soluble sugars to improve plant cold and drought resistance by increasing cell osmotic pressure (Sala et al., 2012; Du et al., 2020; Song et al., 2022), in which NSC storage is allocated at the expense of growth and metabolism (Carbone et al., 2013; Zhou et al., 2021; Fig. 1). Thus, the storage and transformation of NSC under stressful conditions would reflect a strategy of stress resistance beyond plant construction costs and growth potential.

Combining the above insights, we propose the following hypothesis: beyond the trait dimensions of leaf size and resource economics, NSC would reflect an important dimension of cold resistance in alpine plants. To test this hypothesis, we measured 12 leaf traits critical to leaf construction, growth, and stress resistance in 143 species, ranging from alpine steppes to alpine meadows along an environmental gradient on the Tibetan Plateau. Meanwhile, we estimated a direct indictor of cold-tolerance, that is, the lethal temperature causing 50% frost damage (LT₅₀) in 11 species at one of these sites. We firstly investigated the variation and potential leaf trait dimensions of alpine plants and then verified the third leaf trait dimension is related to cold-tolerance by analyzing the relationships of NSC and LT₅₀.

2. Materials and methods 2.1. Study area

The study area ranges from east to west along the Tibetan Plateau (87.20°E–94.74°E, 29.73°N–31.78°N) at an altitude of 3475–4807 m (Fig. 2). The average annual temperature ranges from −4 to 12 ℃, and the average annual precipitation rainfall is 150–800 mm. There are two primary types of grasslands in the region: alpine steppe and alpine meadow. The alpine steppe is dominated by grasses, such as Stipa purpurea, Stipa aliena, and Poa araratica, and alpine meadows are dominated by sedges, such as Carex aridula, Carex tibetikobresia, and Carex alatauensis. Based on the FAO90 soil classification system (https://www.fao.org/soils-portal/soil-survey/soil-maps-and-databases/harmonized-world-soil-database-v12/en/) (Fischer et al., 2008), the main soil types were Gelic Leptosols, Haplic Phaeozems, Gelic Cambisols, Mollic Gleysols, and Haplic Luvisols. Because of the topography and monsoons, the spatiotemporal distribution of precipitation differs greatly within the study area (Gao et al., 2019).

2.2. Field survey and laboratory analysis

In August 2021, seven sites along an environmental gradient from east to west on the Tibetan Plateau were investigated. Four plots (200 m × 200 m) were established in representative communities at each site for leaf and soil sampling. Each plot was set at least 1 km apart. In each plot, eight quadrats (0.5 m × 0.5 m) were surveyed to determine the community composition and structure. Leaf samples from dominant and common species were collected from each plot according to a unified protocol. For each plant species, sun-exposed and mature leaves (leaf blades for grasses) from five to ten individuals or clusters, which provided enough leaf samples for traits measurement and chemical analysis were collected. Soil samples of 0–10 cm layers were collected around the quadrats at 30–50 points using a soil sampler (diameter 6 cm) in each plot. Then soil was mixed together as the soil sample of this plot. In total, we collected 28 soil samples and 143 species belonging to 38 families from 7 sites (41 species occurred more than one site). The species list and sampling coordinates were showed in Appendix A.

For each leaf sample selected for leaf trait analysis at each site, we randomly selected a certain number of fully developed leaves based on leaf size with little signs of damage. Sample storage, processing, and trait measurements were performed according to standardized protocols (Pérez-Harguindeguy et al., 2013). We measured 12 leaf traits critical to leaf construction, growth, and stress resistance: leaf area (LA), leaf water-saturated fresh mass (leaf sf mass), leaf-saturated water content (LSWC), leaf dry matter content (LDMC), specific leaf area (SLA), leaf nitrogen content (LNC), leaf phosphorus content (LPC), leaf nitrogen to phosphorus ratio (Leaf N: P), chlorophyll content (CC), leaf carbon content (LCC), leaf soluble sugar content (LSS), and leaf starch content (LSC) (see Appendix A for detailed methods).

The collected soil samples were packed in aluminum boxes, weighted using an analytical balance, then, oven-dried at 105 ℃ for 24 h to determine soil water content (SWC). Each soil sample, sieved through 2-mm mesh screens, was air-dried for approximately 100 g in a vent room and ground into a fine powder using a ball mill (Retsch MM400, Germany) for soil chemical and physical property analyses. We measured the soil pH (soil pH), soil available nitrogen (SAN), soil available phosphorus (SAP), soil available potassium (SAK), and soil organic carbon (SOC) (see Appendix A for detailed methods).

A cold-tolerance experiment was conducted at Nagqu, one of the seven sites, and LT₅₀ of 11 species (Table S1) were estimated using electrolyte leakage method (see Appendix A for detailed methods).

2.3. Environmental data

At each sample site, the longitude, latitude, and altitude data were recorded using a portable GPS. The climate data include mean annual temperature and precipitation (MAT and MAP) from the Climate Data Centre of China (http://data.cma.cn/), mean annual wind speed and solar radiation (wind speed and solar radiation) from WorldClim (https://worldclim.org/data/worldclim21.html) (Fick and Hijmans, 2017).

2.4. Statistical analyses 2.4.1. Correlation analysis and trait clustering

To define groups of correlated traits and examine the synergistic and trade-off relationships among traits, we conducted a Pearson correlation analysis of the 12 traits after log-transformation. We calculated the correlation coefficients. Then, to determine the groups of correlated traits, we used a hierarchical clustering algorithm, based on absolute Pearson correlation coefficient, which employs Euclidean distance to calculate the distance between data matrices. Hierarchical clustering grouped species traits with the least distance and highest similarity into clusters (Joswig et al., 2022). We used the 'rcorr' and 'hclust' functions in R package 'Hmisc', respectively.

2.4.2. Principal component analysis

To investigate the spectrum of alpine plant trait variation, we performed principal component analysis for 12 traits. Prior to analysis, the values for each trait were centralized and standardized by Z-score normalization [(trait value - mean of trait value)/standard deviation of trait value]. We used the 'PCA' function in the R package 'FactoMineR'.

2.4.3. Variation partitioning

We selected environmental factors that represent the alpine environment: strong radiation, high winds, cold temperatures and some regional aridity. Climate variables include MAT, MAP, solar radiation, and wind speed. Soil variables include soil pH, SWC, SOC, and available nutrient (SAP, SAN and SAK).

To determine trait-environment relationships, we used linear mixed-effect models ('lmer' function in R package 'lme4') to quantify the relative effect of environmental factors on trait axes. We treated environmental variables as fixed effects and site as a random effect. Given that the environmental variables were strongly coupled with each other, we used stepwise regression ('stepAIC' function in R package 'MASS') to select climate and soil variables to avoid multiple collinearity. Only the environmental factors with significant effects (P < 0.05) on leaf traits were included in the linear mixed-effect models. Overall model significance and goodness-of-fit were judged by using the likelihood ratio statistic and assessing change in Akaike's information criterion (AIC) scores. Models with lower AIC values were chosen as the final best-fit models (Table S2).

To investigate how leaf traits on the third dimension vary with LT₅₀, we performed reduced major axis regression using the 'lmodel2' package.

All analyses were performed using the R v.4.2.2. Depending on the method of analysis, the data were transformed into the most suitable form.

3. Results 3.1. Three-dimensional spectrum of alpine plant leaf traits

Three main dimensions were identified for 12 alpine plant leaf traits by principal component analysis and together explained 53.5% of the trait variations (Fig. 3). The first leaf trait dimension (principal component 1; PC1) was related to leaf size and structure. Species with high scores on PC1 tend to have high LA, high leaf sf mass, LSWC, and low LDMC, while species with negative scores on PC1 tend to have low LA, low leaf sf mass, and high LDMC (Fig. 3). The second leaf trait dimension (principal component 2; PC2) was associated with resource economics, including nutrient concentrations (LNC, LPC, and leaf N: P), photosynthetic capacity-related traits (CC), and leaf area per unit of carbon investment (Fig. 3). High SLA associated with high LNC, LPC, and CC (Fig. 3) implies a tradeoff between leaf construction cost and plant growth potential. Furthermore, leaf carbon compounds (LCC, LSC, and LSS) were clustered in the third leaf trait dimension (principal component 3; PC3), which was related to stress resistance.

3.2. The response of leaf trait dimensions to environmental factors

Climate and soil factors jointly regulated the trait groups, accounting for about 66.8%–85.1% of the variation in the three trait dimensions (Fig. 4 and Table S3). Climatic and soil factors had different effects on these three trait axes. PC1 was mainly influenced by water related factors, in which SWC accounted for the largest contribution (Fig. 4 and Table S3). The variation on PC2 was mainly partitioned by soil fertility and water related factors. Soil pH (soil pH was associated with soil fertility as a result of significant negative correlations with SAN, SAP and SOC; Table S4) and SAK together explained 62.6% of the trait variation on PC2 (Fig. 4 and Table S3). PC3, the axis clustered by leaf C and NSC, showed strong temperature partitioning of trait variation (Fig. 4 and Table S3).

The influence of temperature on each individual trait of PC3 was not the same. The variation of LCC was largely explained by soil fertility (soil pH and SAK), while MAT did not show significant effect (Fig. 4 and Table S3). Meanwhile, MAT was the largest contributor to the explained variance in LSC and LSS, accounting for 36.2% and 56.3%, respectively (Fig. 4 and Table S3). LSC and LSS all increased with the decrease of MAT, while LSS also increased with the decrease of SWC (Table S5). In general, the significant negative correlations between leaf traits on PC3 and MAT suggested that the third leaf trait dimension was related to cold-tolerance of alpine plants.

3.3. LSS is the most critical trait associated with cold-tolerance

To verify whether the third leaf trait dimension was indeed associated with cold-tolerance of alpine plants, we tested the correlations between traits on PC3 and LT₅₀. Although LCC, LSC and LSS all showed negative correlations with MAT, a significant negative correlation was only found between LSS and LT₅₀ (R² = 0.26, P = 0.05; Fig. 5). LCC and LSC did not represent remarkable relationships with LT₅₀ (P > 0.05). This result indicated that PC3 was able to reflect the cold-tolerance of alpine plants to some extent, in which LSS was the most critical trait.

4. Discussion 4.1. The third dimension of alpine plant leaf traits is related to alpine plants cold resistance

It is a key problem to understand what environmental factors determine plant trait spectrum. In our study, different climate and soil variables have different importance in trait-environmental relationships. The effects of climate and soil variables on the first two dimensions of leaf trait variation in alpine plants were consistent with those on the global scale (Joswig et al., 2022). That is, climate variables mainly explain size traits, while soil variables mainly explain economics traits. However, climate regulations on size traits were different in alpine plants compared with global plants. For global plants, sufficient water, sunlight and warm temperatures allowed plants to grow fast and to have bigger size (Díaz et al., 2016; Joswig et al., 2022). While the negative correlation between solar radiation and PC1 (Table S5) indicated higher solar radiation corresponds to smaller leaves for alpine plants. Meanwhile, it was worth noting that wind speed, rather than temperature, had more significant effect on the first axis of alpine plant size. In brief, sufficient water, lower intensity of solar and slower wind speed allows bigger leaf size of alpine plants.

Our result shows that, beyond the two general dimensions (leaf size and structure and resource economics) (Díaz et al., 2016; Joswig et al., 2022), cold resistance represented by NSC was also an important dimension in alpine plants leaf spectrum. The third dimension of alpine plant leaf traits reflected sensitivity to temperature changes (Fig. 4; Tables S3 and S5). The direct evidence from LT₅₀ indicates that the trait in the third dimension that the most important contributor to cold-tolerance was LSS (Fig. 5). Under cold stress, LSS increases as osmotic protection, either as the result of a conservative, hedging strategy as a form of cryoprotection from upregulation (Sala et al., 2012), or the result of sugar accumulation due to physiological limitations of growth at low temperatures (Körner, 2003, 2015), or a combination of the two (Blumstein et al., 2023). Temperature-sensitive starch-degrading enzymes promote more conversion of starch to soluble sugars (Carbone et al., 2013).

4.2. NSC should not be ignored in leaf economic spectrum of alpine plants

LCC and NSC, which clustered on the third leaf trait axis, were all related to plant carbon reserves (Fig. 1). After being fixed by photosynthesis, parts of carbohydrates are allocated to plant organ construction, known as stable reserve of SC, while other parts serve as a mobile C pool in the form of NSC. Previous studies of plant trait spectrum almost exclusively focused on traits highly related with stable C reserves, such as LCC, which was more strongly associated with resource economics (Joswig et al., 2022). Within trait spectrum absence of LCC, leaf dry mass content, significantly related to LCC, was associated with plant size and structure for global plant species (Díaz et al., 2016) or with resource economics for tundra species (Thomas et al., 2020). Nevertheless, when NSC was considered in the trait spectrum, LCC and NSC were clustered into a new dimension separating from plant size and structure and resource economics. It suggests that the variations of NSC make a linkage of different plant strategies to cope with environmental changes and C allocation among different C reserves (Würth et al., 2005; Thalmann and Santelia, 2017; Fig. 1).

Plant trait spectrum and strategies associated with structural C has been well known. That is, species with higher C input per unit leaf area, generally have a longer leaf life span, and slower growth rate, being less sensitive to environmental changes. In our study, species with expensive leaf construction have better cold resistance, which revealed by a significant positive correlation between SLA (1/LMA) and LT₅₀ (R² = 0.104, P = 0.045; Table S6). However, the relationships between NSC and the classical trait spectrum and the underlying plant strategies are still unclear. NSC has been proved playing a positive role in plant resistance of drought (Würth et al., 2005; Du et al., 2020), low temperature (Hoch et al., 2002; Hoch and Körner, 2012), freeze–thaw events (Sala et al., 2012), and pest and disease (Thalmann and Santelia, 2017). The level of NSC is related to plant growth and metabolism rate. Previous study on Inner Mongolian semi-arid grassland community exhibited that higher plant growth rate decreased the NSC level in leaves as a result of increased carbohydrate consumption. In contrast, increased photosynthesis of plant tended to increase the NSC level in leaves (Wang et al., 2017). Alpine plants developed a strategy to accumulate more NSC at the expense of growth in response to low temperatures (Hoch et al., 2002; Hoch and Körner, 2012; Zhou et al., 2021). The enhanced construction cost and lager NSC storage limited C input to new tissue growth and metabolism (Fig. 1). This might provide a new perspective to the classical 'fast-slow' spectrum in understanding the growth strategy of alpine plants.

5. Conclusions

Our results suggested that NSC should not be ignored in leaf economic spectrum for alpine plants. The third dimension of leaf traits related to LCC, LSS and LSC clustered together reflects the trade-offs in the allocation of carbon assimilation products in growth, metabolism and storage, which is an extension of the resource-utilization strategy beyond plant construction and growth. LSS had a significant negative correlation with LT₅₀, confirming that the third dimension of leaf traits has an important function in cold-tolerance. NSC, serving as a mobile C pool, should be emphasized in future studies, especially to explore plant life strategies under stressful environments.

Acknowledgements

We are grateful for the Nagqu Integrated Observation and Research Station of Ecology and Environment, Nam Co Station for Multisphere Observation and Research, Chinese Academy of Sciences, South-East Tibetan Plateau Station for Integrated Observation and Research of Alpine Environment, and Xainza Alpine Steppe and Wetland Ecosystem Observation Station. This work was supported by the National Natural Science Foundation of China (32192461, 32271619 and 32160285) and the Natural Science Foundation of Science & Technology Department of Qinghai (2020-ZJ-952Q).

CRediT authorship contribution statement

Yuan Wang: Writing – review & editing, Writing – original draft, Visualization, Software, Methodology, Investigation, Formal analysis, Data curation. Ji Suonan: Investigation, Data curation, Conceptualization. Kun Liu: Supervision, Investigation. Yanni Gao: Supervision, Methodology. Sihao Zhu: Methodology, Investigation, Formal analysis. Qian Liu: Investigation. Ning Zhao: Writing – review & editing, Writing – original draft, Investigation, Funding acquisition, Data curation, Conceptualization.

Declaration of competing interest

All co-authors have been and agree with the contents of the manuscript and there is no potential conflict of interest to report. We certify that the submission is original work and is not under review at any other publication.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.pld.2024.10.001.