Introduction

Thermokarst lakes, emerging from the thawing of permafrost in high altitude and latitude regions1,2, are significant aquatic ecosystems prevalent across the Arctic and sub-Arctic areas3, and the Qinghai-Tibet Plateau4. They are emblematic of the broad challenges and opportunities associated with climate-induced environmental change. Attributing to their extensive distribution and substantial reserves of water, carbon, and nutrients5,6, thermokarst lakes are increasingly recognized as dynamic hotspots of hydrological, ecological, and biogeochemical processes7,8. Accelerating climate change is causing rapid permafrost thaw, resulting in substantial changes in size and abundance of thermokarst lakes, as well as their biogeochemical processes9,10. These changes further highlight the role of high latitude and elevation lakes as sensitive indicators of environmental shifts11,12. A critical aspect of these lakes would be their biogeochemical stoichiometry, particularly the ratios of carbon (C), nitrogen (N), and phosphorus (P). These nutrient stoichiometric ratios are fundamental because they influence primary productivity, determine the availability and limitation of essential nutrients for various organisms, and shape the overall structure and function of aquatic ecosystems13,14,15. Understanding these stoichiometric relationships helps unravel the complex web of ecological interactions, including nutrient cycling, energy flow, and the dynamics of plankton communities, which are essential for maintaining ecosystem health and stability16,17,18. Despite the importance, however, nutrient stoichiometry in thermokarst lakes remains underexplored.

The elemental composition and balance of C, N, and P (C:N:P) play a foundational role in shaping the structure and function of aquatic ecosystems16,18. The nutrient composition of lake water and seston (the suspended particulate matter) can serve as a crucial starting point for investigation19. The balance and dynamic of C:N:P in these matrices provide insights into the availability and limitation of nutrients for biota. Phytoplankton, as the primary producers responsible for fixing carbon and providing the foundational energy source for the food web, are highly responsive to changes in nutrient availability and stoichiometry20,21. The composition and abundance of phytoplankton are closely tied to nutrient availability and ratio, with different species exhibiting varied nutrient requirements and stoichiometric flexibility16,22,23. The availability of N and P influences the nutrient composition of phytoplankton communities, which in turn shape the ecological and biogeochemical dynamics of the lake18,24. Thermokarst lakes represent dynamic zones where various environmental factors such as nutrient inputs, temperature changes, and hydrological processes converge, thereby influencing the composition and interactions of biological communities within these ecosystems8,25. With permafrost thaw and increased terrestrial inputs, thermokarst lakes are experiencing increases in nutrient content, driven by the release of previously trapped nutrients from thawing permafrost and increased runoff from surrounding landscapes, which can have far-reaching consequences on lake productivity and the composition of aquatic communities26,27,28. Ecological stoichiometry is a powerful tool for understanding how the availability and balance of elements in organisms and their environments influences the structure and function of ecosystems. In thermokarst lakes, nutrient availability and ratio are critical for determining nutrient limitations, and their dynamics can directly affect primary producers as well as higher trophic levels like zooplankton. Evaluating nutrient stoichiometry in these lakes is fundamental for unraveling the ecological processes driving phytoplankton dynamics, which subsequently have bottom-up cascading influences on other biotic component of thermokarst lake ecosystems. Understanding these processes is crucial for predicting how these ecosystems will respond to ongoing environmental changes in the era of global warming.

The stoichiometry of N and P elements in aquatic environments can influence the growth and productivity of primary producers, and may cascade through the food web, impacting the composition and dynamics of zooplankton communities and beyond16. Owing to cold temperature and low productivity, the majority of thermokarst lakes and ponds support low biodiversity and relatively simple food webs, with few or no intermediate and top-level predators29,30. Changes in nutrient stoichiometry within lakes can influence the quality and quantity of zooplankton food sources, thereby shaping the fitness and composition of these communities31,32,33,34,35. Previous studies suggest that various environmental factors can be important determinants of zooplankton communities in thermokarst lakes, such as temperature, pH, conductivity, nutrient availability, primary productivity, and predation36,37,38. Zooplankton community dynamics and interactions with phytoplankton are influenced by nutrient availability, and, in turn, affect the nutrient stoichiometry of these lakes16,39. Understanding zooplankton community dynamics is integral to comprehending nutrient biogeochemical processes and tracking ecological changes40,41. However, due to accessibility challenges, very little is known about zooplankton communities in thermokarst lakes in permafrost regions36,37,38.

Despite the growing body of research on thermokarst lakes, significant gaps remain in our understanding of nutrient dynamics and community interactions within these lakes. This study aims to address these gaps by posing the following research questions: (1) How do variations in nutrient availability and stoichiometry influence phytoplankton in thermokarst lakes? (2) How do these changes, in turn, affect zooplankton? Our investigation seeks to uncover the intricate web of interactions linking nutrient availability, phytoplankton, and zooplankton communities, and the broader consequences of these interactions on the thermokarst lake ecosystem as a whole. We hypothesize that shifts in C:N:P ratios, influenced by permafrost thaw and other biogeochemical processes, significantly alter phytoplankton community structure and productivity, which subsequently impacts zooplankton communities and overall ecosystem dynamics. Ultimately, our findings will contribute to a deeper comprehension of the consequences of environmental change for thermokarst lakes, offering valuable insights into the broader challenges presented by climate-induced permafrost thaw.

Results

Stoichiometry pattern of thermokarst lakes

For the studied thermokarst lakes across the Qinghai-Tibet Plateau, nutrient concentrations and stoichiometric ratios varied by orders of magnitude in both water and seston (Fig. 1, Supplementary Fig. 1 and Table 2). The average concentration of DOC, TN, and TP in water was 19.18 ± 17.48 mg/L (mean ± SE), 1.37 ± 0.96 mg/L, and 0.06 ± 0.05 mg/L, respectively (Supplementary Table 2). DOC, TN, and TP exhibited high variability with coefficients of variation (CV) of 0.91, 0.70, and 0.83, respectively). The average stoichiometric ratios of DOC:TN, DOC:TP, and TN:TP were 20:1 (CV = 0.68), 950:1 (CV = 0.59), and 65:1 (CV = 0.87), respectively (Supplementary Table 2). The standard major axis regression analysis indicated that the relationships between DOC, TN, and TP were isometric, with slopes not significantly different from 1 (Fig. 1 and Supplementary Table 3).

Fig. 1: Relationships between C, N, and P in water and seston.
figure 1

All data were lg-transformed. The blue lines represent the linear regression, and the shadow areas are the 95% confidential interval.

Full size image

The average concentration of seston carbon, nitrogen, and phosphorus (PC, PN, and PP) was 1.34 ± 2.37 mg/L (CV = 1.78), 0.18 ± 0.21 mg/L (CV = 1.15), and 0.04 ± 0.04 mg/L (CV = 1.19), respectively. The average stoichiometric ratios of PC:PN, PC:PP, and PN:PP were 31:1 (CV = 1.27), 224:1 (CV = 0.88), and 19:1 (CV = 1.07), respectively. The variation of nutrient concentrations and stoichiometric ratios in seston was even higher than that of the corresponding components in water column. All three elements in seston (PC, PN, and PP) were closely correlated with TN, and TP in the water (Supplementary Fig. 2a).

Despite the large variations in PC, PN, and PP concentrations, they were closely correlated with each other. The standard major axis regression analysis indicated significantly positive associations between PC, PN, and PP (Fig. 1 and Supplementary Table 3). The relationships between PC and PN, and PP were allometric (slopes significantly different from 1 with p < 0.05), suggesting disproportionate changes of PC, PN, and PP. Specifically, PN and PP increased faster relative to PC, while PN increased faster relative to PP.

Phytoplankton community

Metagenomic sequencing showed that phytoplankton communities were predominantly composed by cyanobacteria in terms of species richness and relative abundance (Fig. 2a, Supplementary Fig. 3 and Table 4). The average species richness of phytoplankton was 497 ± 45 (Fig. 2a) with cyanobacteria contributing 357 ± 26 and eukaryotic algae 140 ± 35. The relative abundance of cyanobacteria and eukaryotic algae was 90.6 ± 13.6% and 9.4 ± 13.6%, respectively (Fig. 2a). Phytoplankton richness, particularly cyanobacteria richness, was positively correlated with mean annual temperature, while negatively correlated with mean annual precipitation (Supplementary Fig. 2b). Additionally, lake area was an important factor, being negatively correlated with the relative abundance of cyanobacteria, while positively correlated with both species richness and relative abundance of eukaryotic algae (Supplementary Fig. 2b). TP was negatively correlated with cyanobacteria richness and the Shannon diversity of eukaryotic algae (Fig. 2b). However, the TN:TP ratio was not associated with phytoplankton diversity or relative abundance (Fig. 2b). Distance-based redundancy analysis showed that the distribution of phytoplankton communities and cyanobacteria sub-communities were significantly associated with TP and TN:TP ratio (Fig. 2b).

Fig. 2: Phytoplankton communities in thermokarst lakes.
figure 2

a Alpha diversity between different taxonomic groups. The different lowercase letters represent significant differences assessed using the Wilcoxon rank-sum test. b Relationships between environmental variables and phytoplankton alpha diversity. The numbers represent correlation coefficient with p < 0.05. c Distance-based redundancy analysis shown the relationships between environmental variables and phytoplankton communities. d networks shown the relationship between individual phytoplankton and water nutrients (left) and seston nutrients (right).

Full size image

At the species level, the dominant cyanobacteria were Aphanizomenon_flos-aquae (relative abundance of 14.3%), Synechococcaceae_bacterium (8.7%), and Synechococcus_sp. (6.9%), Cyanobium_sp. (6.0%), and Cyanobium_usitatum (5.2%). The top three abundant eukaryotic algae were Diacronema_lutheri (1.9%), Scenedesmus_sp. (1.0%), and Raphidocelis_subcapitata (0.8%). Considering nutrient concentration and stoichiometric ratios, 67 cyanobacteria and 9 eukaryotic algae species were significantly associated with TN, 14 cyanobacteria associated with TP, while 27 cyanobacteria and 10 eukaryotic algae associated with TN:TP (Fig. 2d). Regarding seston nutrient stoichiometry, 15, 63, 24, 10, 28, and 39 phytoplankton species were significantly correlated with PC, PN, PP, PC:PN, PC:PP, and PN:PP, respectively (Fig. 2d).

Relationships of seston and phytoplankton with zooplankton

Zooplankton communities in our studied thermokarst lakes were generally simple (Supplementary Table 5), characterized by low diversity (ranging from 2 to 15 species) and relatively low population density (ranging from 0.7 to 166.8 ind./L). In general, Copepod exhibited significantly higher diversity and density than Cladocera (Fig. 3a and Supplementary Fig. 4). Zooplankton communities were influenced by various environmental factors, including geological, physicochemical, as well as stoichiometric variables (Fig. 3 and Supplementary Fig. 2c). PN:PP was negatively associated with zooplankton density and richness, particularly with copepod diversity and richness. PC:PP was negatively associated with zooplankton richness, particularly with Cladocera richness. Moreover, copepod density was positively correlated with PP and PC:PN.

Fig. 3: Zooplankton communities in thermokarst lakes.
figure 3

a Alpha diversity between different taxonomic groups. The different lowercase letters represent significant differences assessed using the Wilcoxon rank-sum test. b Relationships between seston nutrient and zooplankton alpha diversity. The numbers represent correlation coefficient with p < 0.05. c Distance-based redundancy analysis shown the relationships between environmental variables and zooplankton communities. d networks shown the relationship between individual zooplankton and seston nutrients (left) and individual phytoplankton (right).

Full size image

Distance-based redundancy analysis showed that the distribution of zooplankton communities and Copepod subcommunities was significantly associated with longitude, latitude, elevation, conductivity, pH, PP, and PN:PP (Fig. 3c). The distribution of Cladocera subcommunities was significantly associated with lake area, TN, and PN:PP. For individual zooplankton, one Cladocera was positively correlated with PC, and three copepoda were associated with PN, PP, and/or PC:PN:PP (Fig. 3d).

In terms of relationships between phytoplankton and zooplankton, Alona guttata, Alona rectangular, Ceriodaphnia quadrangular, and Daphnia sp. had strong interactions with eukaryotic algae (Fig. 3d). However, Chydorus sphaericus, Daphnia magna, Harpacticoida sp., and Nauplius had strong interactions with cyanobacteria. In addition, Acanthodiaptomus tibetanus, Calanidae Copepodite, Cyclopoidea Copepodite, Cyclops vicinus vicinus, and Eucyclops serrulatus had interactions with both cyanobacteria and eukaryotic algae (Fig. 3d).

Cascading influences from nutrient to phytoplankton, and to zooplankton

The structural equation modeling (SEM) results elucidate the intricate relationships among nutrient stoichiometry, phytoplankton, and zooplankton communities within the studied thermokarst lakes across the Qinghai-Tibet Plateau (Fig. 4). Nutrient stoichiometry in water showed a negative effect on the nutrient stoichiometry in seston (β = -0.146, p < 0.05, Fig. 4). The phytoplankton community, divided into eukaryotic algae and cyanobacteria subcommunities, exhibited distinct responses to nutrient concentration and stoichiometry in water. Eukaryotic algae subcommunities were positively influenced by TP (β = 1.137, p < 0.001,) and the stoichiometry in water (β = 0.708, p < 0.01) while negatively by TN (β = -0.448, p < 0.01) (Fig. 4). However, cyanobacteria subcommunities showed no significant direct relationship with the stoichiometry in water but a positive relationship with TN (β = 0.493, p < 0.01). The zooplankton community, represented by copepods and cladocerans, also displayed differential responses to nutrient stoichiometry in seston. Copepods were negatively influenced by the stoichiometry in seston (β = −0.383, p < 0.1), while Cladocera did not show directly influences by seston stoichiometry. Moreover, a robust negative effect between cladoceran and copepod was confirmed (β = −0.636, p < 0.01), suggesting competitive interactions between these two zooplankton groups (Fig. 4). However, no significant relationships were found between phytoplankton and zooplankton.

Fig. 4: Structural equation modeling shown potential bottom-up cascading influences from nutrient stoichiometry to phytoplankton to zooplankton.
figure 4

The red and green lines represent negative and positive relationships with “*”, “**” and “***” represent p < 0.1, p < 0.05, and p < 0.01. The gray lines represent the pathways considering in the model but with p > 0.1. The symbols “↑” and “↓” aside the stoichiometric ratios indicate a positive or negative relationship between the ratio and the first PCA component of the stoichiometry variable, respectively.

Full size image

Discussion

The results from this study on thermokarst lakes across the Qinghai-Tibet Plateau provided significant insights into the complex interplay between nutrient stoichiometry, plankton community structure, and ecosystem dynamics within these unique aquatic systems. These dynamics include changes in species composition, interactions between phytoplankton and zooplankton, nutrient cycling processes, and the stability of the lake ecosystems in response to variations in nutrient availability and environmental conditions. The observed high variability in nutrient concentrations and stoichiometric ratios in both water and seston underscores the heterogeneity of thermokarst lake ecosystems, likely reflecting differences in underlying geological, hydrological, and biological processes42,43,44. The high variability in DOC, TN, and TP concentrations, along with their stoichiometric ratios, indicates a dynamic balance between inputs (e.g., from thawing permafrost, atmospheric deposition, and catchment runoff) and internal processing (e.g., microbial decomposition, nutrient uptake by biota)26,45. Variations in DOC:TN:TP ratios are essential for regulating both heterotrophic and autotrophic metabolism, which in turn affects nutrient cycling46. The isometric relationships between DOC, TN, and TP suggest that their biogeochemical cycles might be coupled in thermokarst lakes.

In contrast, the disproportionate changes among PC, PN, and PP in seston, with PN and PP increasing faster relative to PC, suggest differential utilization by phytoplankton and other microorganisms, and their sensitivity to variations in nutrient availability and limitation, particularly nitrogen and phosphorus inputs from permafrost thaw and other sources20,47. Notable deviations of seston N:P were commonly found in aquatic ecosystems18,48. The observed patterns that PC:PN had a higher variability than PN:PP and PC:PP can be explained by that PC:PN might be largely driven by common physiological adjustment strategies across all phytoplankton, while PN:PP can be driven by ecological selection for taxonomic groups with different phosphorus storage capacities16,49. Nutrient concentrations may affect the elemental composition of seston by controlling phytoplankton community structure48,50. Additionally, factors such as temperature, water nutrient levels, latitude, and seasonality have been shown to control seston C:N:P ratios51,52. Alternative explanations for our findings could involve considering the impact of permafrost thaw, a critical aspect of thermokarst lake formation and evolution, which significantly influences nutrient release and availability. For example, C-enriched detritus from terrestrial sources might be abundant in thermokarst lakes, contributing significantly to the PC pool and leading to elevated and hypervariable PC:PN ratios53. This suggests that phytoplankton biomass alone may not fully account for the seston stoichiometry, and the presence of detritus could alter the nutrient dynamics and energy flow in these systems. Although we did not have direct data on the proportion of detrital carbon in the PC pool, it is reasonable to assume that the presence of detritus not only inflated the PC:PN ratio but also influenced the quality of food available to zooplankton, as detrital material is less nutritious compared to living phytoplankton54. Identifying the mechanisms that underpin C:N:P variability is essential to understanding the factors that limit primary productivity, the effect of resource competition on community structure, and long-term regulation of aquatic biogeochemistry.

The dominance of cyanobacteria in the phytoplankton communities, along with observed environmental correlations, suggested that these primary producers may highly adapted to the specific conditions of thermokarst lakes. Cyanobacteria are known for their ability to fix atmospheric nitrogen in nitrogen-limited environments, a process that is particularly advantageous when inorganic nitrogen is scarce. N-fixation allows certain cyanobacteria to thrive in environments where phosphorus becomes the primary limiting nutrient55,56. In our study, we observed a negative relationship between cyanobacteria richness and TP, which may indicate their ability to outcompete eukaryotic algae under lower phosphorus availability and overall nutrient limitation57,58,59. Moreover, the presence of dominant cyanobacterial species known to fix nitrogen (e.g., Aphanizomenon flos-aquae and Synechococcaceae_bacterium) further supported that N-fixation may play a role in structuring the phytoplankton communities observed in our study. The species-level associations with TN, TP, and their ratios, provided insights into the nuanced preferences and tolerances of different phytoplankton taxa to nutrient availability and stoichiometry. Distance-based redundancy analysis further emphasized the significant role of nutrient stoichiometry in influencing phytoplankton community assemblages, with notable association of phytoplankton distribution with TP and TN:TP60,61. Shifts in lake N:P stoichiometry are known to alter ecological nutrient limitation of phytoplankton growth62,63. Thus, enhanced N and/or P inputs from thawing permafrost and the resulting nutrient imbalance might shift phytoplankton biomass and diversity, and finally may alter the structure and functioning of the entire lake ecosystems.

The wide range of PC:PN:PP alongside the highly constrained C:N:P ratios in Cladocera and Copepoda, suggests that these zooplankton may experience varied elemental limitation. Nutrient limitation in zooplankton production is a consequence of the tight elemental composition in zooplankton and the variable elemental composition in their food sources64,65,66. The differential responses of copepods and cladocerans to nutrient stoichiometry and phytoplankton communities likely reflect their distinct nutritional requirements and competitive abilities. Generally, Cladocera have lower N:P than Copepoda and fed non-selectively, making copepoda more sensitive to nutrient stoichiometry and nutrient limitation67,68. In addition, zooplankton diversity and abundance in our studied thermokarst lakes were negatively associated with PC:PP and PN:PP, indicated the pivotal role of seston nutrient stoichiometry in structuring zooplankton communities18,69. Distance-based redundancy analysis suggested that zooplankton distribution associated with both abiotic factors (like longitude, latitude, elevation, conductivity, and pH) and nutrient-related variables, further illustrated the complex interplay between environmental conditions and the ecological niches occupied by these organisms70,71. The interactions between specific zooplankton taxa and phytoplankton groups reveal a complex food web dynamic within these lakes, where certain zooplankton species exhibit preferences for or stronger interactions with either cyanobacteria or eukaryotic algae72,73. This specificity in feeding relationships can significantly impact energy flow and nutrient cycling within thermokarst lake ecosystems, potentially influencing the stability and resilience of these communities to environmental changes. Projected thawing permafrost and thermokarst lake succession are likely to substantially influence lake food webs by altering lake environments, particular by shifting nutrient stoichiometry. The discrepancy in the N:P ratio between zooplankton and their food has implications for nutrient recycling39,74,75.

Synthetically, structural equation modeling analysis provided a comprehensive overview of the bottom-up influences from nutrient stoichiometry to phytoplankton/seston, and to zooplankton, highlighting the interconnectedness of biogeochemical and biological processes in thermokarst lakes. In SEM, PC and DOC were not included as direct drivers due to that PC only has a minute impact on DOC (PC/DOC = 0.079), while PN and PP had a much larger impact on TN and TP, respectively (PN/TN = 0.154, PP/TP = 0.763). The phytoplankton community, divided into eukaryotic algae and cyanobacteria subcommunities, exhibited distinct responses to nutrient concentrations and stoichiometry in water. The results suggested that nutrient availability and stoichiometry can differentially influence phytoplankton subcommunities, potentially through varying nutrient uptake and utilization strategies21,60,62,63. For example, cyanobacteria can fix atmospheric nitrogen, use high-affinity phosphate uptake systems, and exhibit flexible stoichiometry, allowing them to thrive in nutrient-limited environments76,77,78. In contrast, eukaryotic algae rely on rapid nutrient uptake during pulses, efficient nutrient storage, and selective uptake of nitrogen forms59,79. The zooplankton community, represented by copepods and cladocerans, also displayed differential responses to nutrient stoichiometry in seston. Copepods were more sensitive to changes in nutrient quality of their food sources68,80. The significant negative correlation between copepods and cladocerans indicates competitive interactions, likely driven by resource or habitat partitioning strategies that allow these groups to coexist in the dynamic environment of thermokarst lakes81,82. For example, copepods typically prefer deeper, colder waters where they feed selectively on higher-quality food sources. Cladocerans, in contrast, are often found in the warmer, more turbid surface layers, and are generalist filter feeders, consuming a broad range of seston83,84. As discussed above, we acknowledge that the presence of detritus adds complexity to these relationships and likely moderates the strength of the bottom-up effects from nutrient stoichiometry to zooplankton. However, the SEM still provides valuable insights into the broader nutrient dynamics, as it captures the overall relationships between nutrient stoichiometry and plankton communities, including potential effects of detrital inputs.

While our study focused primarily on phytoplankton stoichiometry and its role in nutrient dynamics, it is important to acknowledge the contributions of bacteria and detritus in thermokarst lake ecosystems85,86, although we did not measure those in our study. Heterotrophic bacteria, particularly those utilizing DOC, play a significant role in carbon and nutrient cycling, often competing with phytoplankton for limiting nutrients87. In lakes with high DOC levels, bacterial activity may be particularly important, influencing nutrient availability and organic matter decomposition. Similarly, detritus, in the form of particulate organic matter, contributes to the total seston pool, impacting the carbon, nitrogen, and phosphorus stoichiometry available to higher trophic levels88. The balances between these components—bacteria, detritus, and phytoplankton—likely shape the overall nutrient dynamics in thermokarst lakes. Although our current analysis emphasized phytoplankton, future research should quantify the relative contributions of bacterial biomass and detrital matter to more fully understand nutrient stoichiometry and food web interactions in these complex systems.

Conclusions

Our research on the Qinghai-Tibet Plateau’s thermokarst lakes elucidated the intricate connections between nutrient stoichiometry and plankton communities. Notably, we observed significant fluctuations in nutrient concentrations and their stoichiometric balances, highlighting the dynamic and responsive nature of these ecosystems to permafrost melt and biogeochemical activities. The results highlighted the pivotal role of nutrient availability in shaping phytoplankton communities and seston quality, which in turn influenced zooplankton dynamics and overall lake ecosystem. These insights into interconnected nature of nutrients and planktons in thermokarst lakes were critical for understanding the impacts of climate change on these sensitive environments. Our work contributed to the broader knowledge of aquatic ecosystems under changing climatic conditions, emphasizing the importance of ecological stoichiometry and integrated approaches to limnological research in high-latitude and elevation regions.

Methods

Study area, field sampling, and chemical analysis

This study was conducted in July and August in 2021 and 2022 across the Qinghai-Tibet Plateau (Fig. 5). A total of 68 thermokarst lakes were sampled, covering a wide range of longitudes (90.59E to 98.59E), latitudes (29.65 N to 35.52 N), and elevations (3569 m to 4959 m above the sea level). The mean elevation of the studied lakes was 4426 ± 41.4 m (mean ± SE). All these lakes are closed water bodies with no inflow or outflow.

Fig. 5: Study area and sampling sites.
figure 5

The distribution of the permafrost on the Qinghai-Tibet Plateau97,98 were downloaded from the National Tibetan Plateau Data Center (https://data.tpdc.ac.cn/en/).

Full size image

Due to the shallow depth of the lakes, water samples were collected at a depth of 0.3-0.5 m. In each lake, water samples were collected in triplicate using acid clean bottles, which were transported to the laboratory in a cooler for further processing. Conductivity and pH of the lake water were measured in situ using a multiparameter instrument (YSI ProPlus, Yellow Springs, Ohio). In the laboratory, dissolved organic carbon (DOC) was analyzed using a Shimadzu TOC Analyzer (TOC-VCPH, Shimadzu Scientific Instruments, Columbia, Maryland) after the filtration through glass fiber filters (GF/F, Whatman, UK). Unfiltered water samples were used to measure total nitrogen (TN) and total phosphorus (TP). TN was quantified by ion chromatography after persulfate oxidation (EPA 300.0). TP was measured using the ascorbate acid colorimetric method after oxidation (EPA 365.3). Given the important role of seston particles—which include phytoplankton cells, detritus, and other particulate organic matter—as the major food source for zooplankton89,90, seston particle carbon (PC), seston particle nitrogen (PN), and seston particle phosphorus (PP) were measured. From each bottle, two 200–500 ml subsamples were taken and filtered through two 25-mm glass fiber filters with a pore size of 0.7 µm (Whatman, UK). One filter sample was used to measure PC and PN, and the other one was for measuring PP. PC and PN were analyzed using an elemental analyzer (LECO 628, USA). PP was analyzed using the ascorbate acid colorimetric method after digestion with potassium persulfate (EPA 365.3). In order to reflect the elemental balance in these lakes and food quality for zooplankton, the stoichiometry ratios of DOC:TN:TP and PC:PN:PP were calculated as molar ratios. The processing, retention, and transport of C, N, and P in lakes are largely determined by their balance91. The raw data of the sample location and element’s concentration was presented in Supplementary Table 1.

Phytoplankton and Zooplankton sampling and analyzing

Phytoplankton samples were collected in 22 lakes and identified using metagenomic sequencing. Phytoplankton samples were collected in triplicate by filtering 200 mL water onto a 0.2-μm polycarbonate membrane filter (Whatman, UK). For each lake, the triplicates were combined into one composite sample for DNA extraction. DNeasy PowerWater kit (Qiagen, Hilden, Germany) was used to extract DNA from filter samples following the manufacturer’s protocol. The DNA extractions were quantified using the Qubit dsDNA BR assay kit in a Qubit 2.0 fluorometer. Paired-end sequencing libraries were created with the Illumina TruSeq Nano DNA LT Sample Preparation Kit, and sequencing was carried out on an Illumina PE150 platform. Raw reads in FASTQ format were processed with Trimmomatic (v0.36) to trim and filter the data. The metagenome was assembled using MEGAHIT (v1.1.2) with default parameters. The scaffold was divided into new contigs (Scaftig) using gaps within the scaffold as breakpoints and only those Scaftigs with a length of 500 bp or more were used for open reading frame (ORF) prediction. The ORFs were predicted and translated into amino acid sequences using prodigal (v2.6.3). A non-redundant gene set was built for all predicted genes with CDHIT (v4.6.7) and the longest gene was selected as the representative for each set. The clean reads of each sample were aligned against the non-redundant gene set (with 95% identity and 90% coverage) using bowtie2 (v2.2.9) and the gene abundance information was counted for each corresponding sample. The abundance of genes was calculated as \({G}_{k}=({r}_{k}/{L}_{k})\times (1/{\sum }_{i=1}^{n}\left({r}_{i}/{L}_{i}\right.)\). In this equation, k symbolizes a specific gene, r stands for the number of reads for gene k, L represents the length of the gene, and n denotes the total number of genes. The taxonomy of the species was obtained as a result of the corresponding taxonomy database of the NR Library, and the abundance of the species was calculated using the corresponding abundance of the genes. Eukaryotic algae and cyanobacteria were extracted from the metagenomic data as the phytoplankton community.

Zooplankton samples were collected with a plankton net from 68 lakes. For each lake, 15–30 L water samples were collected from the open water region using a water column sampler (hanging on a 5-meter-long telescopic pole) and then filtered through the plankton net91. Zooplankton samples were immediately preserved in 5% formalin. Samples were identified to species in the Freshwater Ecology Research Center, Institute of Hydrobiology, Chinese Academy of Sciences. Zooplankton density was presented in the unit of individuals per liter (ind./L).

Statistical analyses

To assess the isometric/allometric relationships between C, N, and P concentrations in water and seston, the log-log function log10(y) = a × log10(x) + b was used in standard major axis regression, and the slope was tested in comparison to 1. R package SMART 3.4-892 was used for standard major axis regression. Differences in alpha diversity between different groups were assessed using the Wilcoxon rank-sum test. Distance-based redundancy analysis (dbRDA), a constrained ordination method using non-Euclidean distance measures, was conducted using “vegan 2.5-7” package93 to assess the relationships between environmental variables and phytoplankton and zooplankton communities. Additionally, network analyses were used to show the relationships between individual phytoplankton and zooplankton species and nutrients in water and seston. In networks, only significant (p < 0.05) correlations were used. The networks were visualized using Gephi 0.1094. Structural equation modeling was conducted to assess the potential bottom-up influences from nutrient stoichiometry to phytoplankton to zooplankton using the “lavaan” package95. In the structural equation modeling, nutrient stoichiometry and plankton communities were reduced in dimension by principal component analysis (PCA), respectively, using the “prcomp” function of the “vegan 2.5-7” package, with the first axis (PCA1) used in the structural equation modeling. All the statistical analyses were carried out in R 4.3.196.