Salinization alters the elemental balance of wetlands and induces variations in plant survival strategies. Sulfur (S) plays vital roles in serving regulatory and catalytic functions in stress resistance of plants. Yet, how plant S and its relationships with nitrogen (N) vary across natural environmental gradients are not well documented. We collected 1366 plant samples and 230 water and sediment samples from 230 wetlands in Tibetan Plateau and adjacent arid regions of western China, to analyze the effects of environmental variables on plant S accumulation and N-S correlations. We found that plant S correlated with N in unimodal patterns. Salinity, rather than temperature or nutrient supply, promoted disproportionate accumulation of S but limited N uptake, inducing decoupling of N-S correlation in plants. Towards high salinity, the faster increasing rates of total S than that of glutathione, the most abundant organic-S compound in plant resistance, provided potential evidences explaining the decoupled plant N-S correlation. A salinity of 3.9‰ was calculated to be a threshold at which substantial changes in plant N-S correlation occurred. We designed a conceptual model to illustrate the mechanisms driving variations of N-S correlation in plants and environments along salinity gradient. In addition, salinity reduced species richness and drove community reassembly by filtering species with high S concentrations at community scale. Our study addressed the critical roles of S in plant resistance under adverse conditions. Studies on biogeochemical cycles of S and N in wetland ecosystems will further enhance our understanding of plant responses to future climate change.

We focused on two plant forms: emergent plants and submerged plants. Emergent plants were defined as hydrophytes that grew in waterlogged soil with their roots submerged in water and their leaves in air. Submerged plants were those whose whole bodies were submerged in water with their roots in sediments. Using a uniform protocol, plant samples were collected across the study area in July and August of 2018 and 2019. For emergent plants, 30-50 fully expanded and intact leaves were sampled randomly for each species. For submerged plants, we collected 30-50 segments of plant shoots (20 cm from the tips) and then removed all the leaves from the segments. In total, we collected 82 species, belonging to 25 families and 37 genera from 230 sites, for a total of 1366 samples (Dataset S1). The total 1366 samples consisted of two groups: 1019 samples from all 82 species in 230 sites for measuring concentrations of leaf N (LN) and leaf S (LS) and 347 samples from 44 common species in 146 sites for measuring concentrations of leaf GSH. The 347 samples were collected simultaneously with the corresponding samples in the 1019 samples, but was stored as fresh samples in liquid nitrogen during field investigation. The 1019 samples were oven-dried at 70℃ for 72 h and then ground into fine powder.

Before sampling, the pH and salinity of the undisturbed water were measured in situ with a handheld multiparameter meter (PROPLUS, YSI, USA). Then, 500ml water samples were collected from midway between the water surface and the soil substrate. The total nitrogen contents of water were analyzed using a multiparameter portable colorimeter (DR900, HACH, USA) with a digestion solution, while the total SO₄^2- contents of water were determined by ion chromatography (ICS-1000, DIONEX, USA). Three samples from the upper 0-20 cm of sediments were collected randomly at each site and then mixed, air dried, the roots and rocks removed, and then ground and sieved through a 0.15mm mesh for further analysis. The pH and conductivity of the sediments were measured by dissolving samples into a mass-based soil-water ratio of 1:5. Total N and S concentrations of both plants and sediments were measured from powdered samples by an elemental analyzer (UNICUBE, Elementar, Germany). According to the protocol of Cang and Zhao (2013), plant GSH concentrations were measured from fresh samples based on the 5,5'-dithio bis-(2-nitrobenzoic acid) colorimetric (DNTB method).