Genetic canalization of nutrient resorption in the widespread grass Phragmites australis under salt stress

To establish that the salinity treatment imposed a physiologically meaningful stress before evaluating NuRE responses, we first examined multilevel indicators of salt stress.

Principal component analysis (PCA) of the metabolome revealed a clear separation between salt-treated and control plants (Figure 1a), with the first two principal components explaining 16.51% and 7.81% of the total variance, respectively; orthogonal partial least squares discriminant analysis yielded a Q² (cum) of 0.233, indicating that the treatment effect is detectable albeit accompanied by substantial genotypic variation. A total of 484 metabolites showed significant differential accumulation between treatments (|log₂ fold change|>1, false discovery rate [FDR] <0.05, variable importance in projection [VIP]>1) (Figure 1b). Among the metabolites, canonical stress markers were prominently altered: N,N-dimethylglycine (a glycine betaine precursor) and sucrose 6-phosphate were significantly upregulated (log₂FC = +1.04 and +2.93, respectively), while glutathione (log₂FC = –0.27) and N-caffeoylputrescine (log₂FC = –1.34) were significantly downregulated. Collectively, these results demonstrate that the salinity treatment induced widespread metabolic reprogramming consistent with osmotic and oxidative stress responses.

Figure 1 with 1 supplement see all

Download asset Open asset

Salt treatment drastically reduced above-ground biomass by more than 60% (control: 835.7±36.4 g, salt: 310.4±15.0 g; Welch’s t=13.26, p<0.001; Figure 2a). Sodium (Na) accumulation, a direct indicator of salt uptake, increased dramatically across all leaf developmental stages. In green leaves, Na concentrations rose from 475 to 1929 mg kg⁻¹ (p<0.001), in yellowing leaves from 508 to 2027 mg kg⁻¹ (p<0.001), and in litter leaves from 1379–3722 mg kg⁻¹ (p<0.001; Figure 2b).

Figure 2 with 2 supplements see all

Download asset Open asset

The elemental composition of leaves was also significantly altered by salinity. PCA of 11 elements revealed that PC1 (41.9% variance) separated nutrient-related elements (C, N, P, K, Cu, Zn) from stress-related elements (Si, Mn, Na), while PC2 (14.3%) was primarily driven by Mn, Zn, Cu, Fe, and Si (Figure 2—figure supplement 1a). Salinity significantly increased PC2 scores (p<0.001, multivariate analysis of variance [MANOVA]), indicating a shift in elemental balance (Figure 2—figure supplement 1b and c). Notably, the effect of salinity on elemental composition was stage- and lineage-dependent: during the Green stage, all lineages showed significant responses; during the Yellowing stage, only CN_N and EU lineages were affected; and by the Litter stage, no significant salinity effect was detected in any lineage (Figure 2—figure supplement 1d–f).

The PCA of leaf N, P, and K concentrations showed a clear separation between green leaves and senesced litter along the first principal component (PC1, Figure 2—figure supplement 2a). This separation was driven by the strong positive loadings of all three nutrient concentrations on PC1 (Figure 2—figure supplement 2b). This pattern indicates that the dominant trend in the data is the distinction between nutrient-rich green leaves and nutrient-poor senesced litter, a pattern that directly results from substantial nutrient resorption during senescence.

The results of the linear mixed-effects models indicated a significant effect of phylogeographic group on the resorption efficiency of N, P, and K (p<0.001); however, neither salinity nor the interaction between salinity and phylogeographic group had a significant effect (Figure 3, Figure 3—figure supplement 1). For instance, the CN_N group exhibited the lowest N and P resorption efficiencies among all groups, whereas both the CN_N and CN_E groups showed high K resorption efficiency. Meanwhile, the results revealed a significant effect of ecotype on the resorption efficiency of N and P (p<0.001), but not on K (p=0.694) (Figure 4, Figure 4—figure supplement 1). It should be noted that the ecotype analysis here was based only on Chinese populations, as ecotype classification was not available for non-Chinese populations. A significant interaction between salinity and ecotype was observed for N resorption efficiency (p=0.038), indicating that salinity treatment enhanced N resorption in the saltmarsh ecotype relative to the freshwater ecotype. Among the ecotypes, coastal saltmarsh populations displayed the highest N and P resorption efficiencies, while populations from inland plateau saltmarsh had the lowest.

Figure 3 with 1 supplement see all

Download asset Open asset

Figure 4 with 1 supplement see all

Download asset Open asset

A significant positive association was observed between N concentrations in green leaves and litter under control conditions (β>1, p<0.05; Figure 5a), indicating the presence of nutrient concentration control over leaf N resorption in P. australis under non-saline conditions. However, this relationship disappeared under salt treatment (p>0.05; Figure 5a), suggesting that salinity disrupted the N concentration control strategy. For P, a significant positive association was maintained under both control and salt conditions (β>1, p<0.05; Figure 5b), indicating robust nutrient concentration control over leaf P resorption in P. australis regardless of salinity. In contrast, correlations for K were nonsignificant under both conditions (p>0.05; Figure 5c), suggesting the absence of nutrient concentration control for this element.

Figure 5 with 1 supplement see all

Download asset Open asset

Significant positive relationships were detected between N and P resorption efficiencies (Figure 5d), as well as between the resorbed N:P ratio and the green leaf N:P ratio under both control and salt conditions (Figure 5e), supporting the presence of stoichiometric control over P resorption. Similarly, N and K resorption efficiencies were positively correlated (Figure 5g), and the resorbed N:K ratio was significantly related to the green leaf N:K ratio under both conditions (Figure 5h), indicating stoichiometric control also regulates K resorption.

Additionally, significant linear relationships were found between the log₁₀-transformed resorbed ratio and log₁₀-transformed green leaf ratio for N:P and N:K, with slopes greater than 1 (Figure 5f and i), demonstrating an ‘inverted’ nutrient limitation control for P and K relative to N. The OLS slope estimates (Figure 5—figure supplement 1) generally aligned with these resorption control strategies inferred from the SMA regressions.

The linear mixed-effects models revealed that NuREs of N, P, and K were significantly positively correlated with the absolute latitude of population origin (Figure 6). Variation partitioning indicated that latitude had the largest individual contribution among the measured predictors, but its contribution was modest for N, P, and K resorption (individual R²=0.092, 0.057, and 0.094, respectively; Table 1). The very small contributions of green-leaf P concentration (R²=0.006) and stoichiometry (R²<0.001) further indicate that most variation in P resorption was associated with factors not represented in the present models.

Figure 6 with 1 supplement see all

Download asset Open asset

Raw and carbon-standardized NuRE were strongly correlated for N, P, and K (Pearson’s r=0.997–0.999; Figure 6—figure supplement 1), and analyses using raw NuRE did not alter the results for phylogeographic group, ecotype, salinity, their interactions, or latitude.

We first confirmed that the salinity treatment imposed a strong and sustained physiological stress at multiple biological levels. Metabolomic profiling revealed clear separation between control and salt-treated plants, with 484 metabolites differentially accumulated, including canonical markers of osmotic and oxidative stress such as N,N-dimethylglycine (a glycine betaine precursor) and sucrose 6-phosphate (Figure 1). Above-ground biomass declined by more than 60% (Figure 2a), and leaf Na concentrations increased three- to fivefold across all developmental stages (Figure 2b). Elemental composition also shifted significantly, with salinity increasing PC2 scores (mainly driven by Mn, Zn, Cu, Fe, Si) in a stage- and lineage-dependent manner (Figure 2—figure supplement 1). Collectively, these results confirm that the treatment was physiologically effective, providing a robust foundation to evaluate NuRE responses. In this study, metabolomic profiling was used primarily to confirm that the salinity treatment induced broad physiological stress, rather than to resolve genotype-specific metabolic mechanisms underlying NuRE variation. Integrating metabolite profiles with genotype-level NuRE responses would be a valuable direction for future mechanistic research. Against this background of confirmed stress, the intraspecific variation in NuRE we observed in P. australis reflects divergent genetic origins, not phenotypic plasticity. Our results demonstrate that phylogeographic group and ecotype are primary determinants of NuRE, while salt stress elicited no overall plastic response in NuRE. This finding highlights the conservative nature of NuRE under this non-nutrient stress condition.

Nutrient resorption is established as a key plant functional trait governing nutrient conservation strategies and ecosystem processes (Killingbeck, 1996; Kobe et al., 2005). Therefore, deciphering the drivers of its intraspecific variation (genetic adaptation versus plasticity) is central to functional ecology (Liu et al., 2025; Zhang et al., 2022). Our finding that this variation is shaped by long-term adaptive evolution rather than a general short-term acclimation capacity underscores that macroecological patterns can arise primarily from genetic differentiation. Such trait conservatism, also observed in desert plants under saline stress (Luo et al., 2021), suggests evolutionary optimization under consistent selective pressures, which may constrain phenotypic adjustments to sudden environmental shifts. The predominance of genetic over plastic controls highlights the necessity of incorporating intraspecific diversity into future models of nutrient cycling.

The lack of a plastic increase in NuRE under salinity presents a physiological puzzle. Theoretically, if salt stress inhibits root nutrient uptake through ionic competition or root damage, enhancing resorption from senescing leaves could serve as a critical internal compensation mechanism to maintain minimum nutrient levels for growth and metabolism. This aligns with ecological principles of optimal allocation and stress adaptation, where optimizing internal recycling is a predicted strategy under resource limitation (Nicotra et al., 2010). However, our data did not support this expected plastic increase, a phenomenon also noted in some desert plants under saline stress (Luo et al., 2021). This discrepancy suggests the existence of substantive physiological constraints that override this optimal strategy. Salt stress may disrupt the finely tuned cellular senescence program necessary for efficient nutrient remobilization, possibly through the accumulation of incompatible solutes or oxidative damage that affect element stability (Zuo et al., 2024), or through a preferential reallocation of energy towards immediate survival mechanisms (e.g., osmotic adjustment and ion exclusion). Therefore, the canalization of NuRE may reflect an evolutionary ‘optimum’ under stress (Huang et al., 2022), where the metabolic and logistical costs of enhancing resorption outweigh its potential benefit, or where the core resorption machinery is intrinsically vulnerable to ionic interference (Alseekh et al., 2025).

Our findings appear to contrast with studies investigating other abiotic stressors. For example, a 5-year field experiment simulating climate change, specifically warming and reduced rainfall, reported significant declines in resorption efficiency for multiple nutrients across several shrub species (Prieto and Querejeta, 2020). This discrepancy may stem from differences in the type and duration of stress applied. Our inference of canalization is therefore limited to the absence of a plastic NuRE response during one growing season under the imposed salinity treatment. Chronic, more severe, or multigenerational salinity exposure may produce acclimatory, epigenetic, or transgenerational responses that cannot be evaluated here. Moreover, although altered phloem loading, disruption of senescence-associated remobilization, and reallocation towards osmotic adjustment are plausible explanations for the observed response, we did not directly measure these mechanisms. Long-term experiments combined with targeted molecular and transport measurements are needed to distinguish among these possibilities. Because only one moderate salinity level (10 ppt) was tested, our results do not exclude plastic NuRE responses at higher salinity or along a dose-response gradient. Future experiments should determine whether a threshold exists beyond which the apparent stability of NuRE is overcome.

Alternatively, nutrient resorption may exhibit stressor-specific plasticity, being more responsive to soil nutrient availability and climate-related shifts than to ionic stress such as salinity (Drenovsky et al., 2019; Luo et al., 2021; Yang et al., 2025; Yuan and Chen, 2015). Moreover, the contrast underscores the potential importance of experimental duration (long-term gradual stress versus short-term shock) in shaping plant physiological responses. Such comparisons highlight the need for integrative approaches across stress types and time scales to fully understand the plasticity and evolution of nutrient conservation strategies.

Our study revealed distinct regulatory mechanisms governing the resorption of N, P, and K in P. australis. Phosphorus resorption was consistently influenced by initial green tissue P concentration under both control and saline conditions, supporting the nutrient concentration control hypothesis. In contrast, N resorption followed concentration control only under non-saline conditions, with this relationship disappearing under salinity. Potassium resorption showed no concentration dependence. Notably, N and K resorption exhibited tight stoichiometric coordination, as reflected in stable N:P and N:K ratios across treatments. Functionally, slopes greater than 1 indicate that changes in green-leaf N:P or N:K are accompanied by disproportionate changes in the corresponding resorbed ratio, consistent with relatively greater recovery of P or K than of N across the observed nutrient gradient. This pattern may contribute to maintaining internal N:P:K balance during regrowth. It may also reflect stronger conservation of P, the high mobility of K, or constraints on the remobilization of N retained in structural or metabolic compounds. Because the experiment did not include nutrient additions or direct measurements of remobilization costs, these explanations remain hypotheses.

These element-specific strategies align with broader patterns documented across biomes and taxa. For instance, the prevalence of stoichiometric control for multiple nutrients, including N and K, has been demonstrated at a global scale (Sun et al., 2023), suggesting a universal mechanism that complements nutrient limitation effects. Notably, K resorption did not follow a concentration-dependent pattern, despite this element typically exhibiting high resorption efficiencies globally (often >70%; Vergutz et al., 2012). This suggests that high K recovery may be driven by factors other than green leaf K concentration. One possible explanation is the partial functional substitution of K⁺ by Na⁺ during osmotic adjustment. Na⁺ can replace part of the nonspecific vacuolar osmotic function of K⁺ in plants and has been reported to become a major osmoticum in P. australis from higher-salinity habitats (Wakeel et al., 2011; Zhao et al., 1999). Increased Na⁺ accumulation may therefore reduce reliance on K⁺ for osmotic adjustment, potentially contributing to the weak relationship between green-leaf K concentration and K resorption efficiency observed here. Furthermore, the preferential resorption of P under certain conditions echoes patterns found in nutrient-limited systems such as tropical forests and alpine permafrost ecosystems, where P resorption can surpass N resorption due to strong local constraints (Yang et al., 2025). This suggests that the strong concentration-dependent control of P resorption we observed may be a widespread adaptation to P limitation, which can override other biogeographic expectations. The stability of these stoichiometric relationships under salt stress implies that internal nutrient regulation is a conservative trait, potentially shaped by evolutionary history more than immediate abiotic context. Such stoichiometric flexibility may be especially critical in environments subject to periodic nutrient imbalances, enabling plants to maintain functional homeostasis.

It is important to acknowledge that our experimental design did not include a nutrient addition treatment, which represents the most direct approach to test the nutrient limitation control hypothesis (Chen et al., 2021). Future studies incorporating nutrient manipulations would provide more definitive evidence. However, our use of resorbed nutrient ratios (e.g., Resorbed N:P) as a proxy for relative nutrient limitation is a well-established methodology in resorption ecology (Kobe et al., 2005; Sun et al., 2023), and has been effectively applied in other recent studies (e.g., Yang et al., 2025). The significant intraspecific variation across our studied populations inherently created a broad gradient of nutrient status and demand. Within this context, the consistent ‘inverted’ nutrient limitation pattern (i.e., a slope significantly >1 for the relationship between log Resorbed N:P and log Green N:P) provides indirect evidence consistent with nutrient-limitation control at the intraspecific level in P. australis, but it cannot establish causal nutrient limitation. Direct factorial nutrient-addition experiments would be required to determine whether the observed resorption patterns are driven by the relative limitation of N, P, or K.

Latitude consistently predicted NuRE in our study, with populations from higher latitudes exhibiting greater resorption for all nutrients. This macroecological pattern likely reflects adaptive responses to environmental constraints such as shorter growing seasons, lower temperatures, and reduced nutrient availability, which collectively favor more conservative nutrient use strategies. Although latitude explained more variation than the other measured predictors, its individual contribution remained modest. The substantial unexplained variance indicates that additional climatic, edaphic, demographic, or genetic factors also contribute to NuRE variation. We therefore interpret latitude as a significant but limited correlate of NuRE rather than as a dominant determinant.

This gradient aligns with biogeographic patterns observed in other widely distributed species. For instance, increasing NuRE with latitude has been documented in temperate trees such as Quercus variabilis across China, where N, P, K, and S resorption all increased significantly with latitude, paralleled by declining nutrient concentrations in senesced leaves (Sun et al., 2016). Such convergent evolution of nutrient conservation strategies across distinct lineages and biomes suggests a general principle in plant adaptation to high-latitude environments. However, the relative contribution of genetic differentiation versus phenotypic plasticity to these latitudinal clines may vary by species and ecosystem. For example, in the grass Stipa breviflora, phenotypic plasticity to soil nutrient availability was the dominant factor explaining latitudinal variation in resorption, with genetic effects playing a modest role (Zhang et al., 2022). This contrasts with our findings in P. australis, wherein geographic origin (reflecting genetic or deep-time adaptive processes) outweighed plastic responses to salinity. Therefore, while latitudinal clines in resorption traits are common, our study demonstrates that they can arise predominantly from genetic differentiation rather than phenotypic plasticity.

Our findings demonstrate that NuRE in P. australis is a genetically conservative trait with limited phenotypic plasticity to salt stress. This conclusion is reinforced by the clear evidence that the salinity treatment was effective at multiple levels (metabolic reprogramming, biomass reduction, sodium accumulation, and altered elemental balance) yet NuRE remained unchanged. Such a decoupling of a core functional trait from an effective environmental stress challenges the prevailing assumption of broad trait plasticity to environmental change (Nicotra et al., 2010) and underscores the role of evolutionary constraints, such as stabilizing selection, in canalizing this trait (Luo et al., 2021). Crucially, we show that while salinity can disrupt specific regulatory patterns (e.g., concentration control for N), the overall efficiency of nutrient recovery remains fixed and is predominantly governed by geographic origin (phylogeographic group and ecotype). The separate-model approach avoids including the two correlated classification schemes as simultaneous independent predictors, but it does not fully disentangle deep phylogeographic history from recent habitat-associated differentiation. We therefore interpret the ecotype analysis as complementary evidence of habitat-associated differentiation rather than as an effect independent of phylogeographic history.

This conservatism reframes how we predict ecosystem function under global change. Since trait variation is phylogenetically anchored, the pace and direction of ecosystem nutrient cycling changes will be forecast primarily by the pre-existing genetic composition of plant populations. In rapidly salinizing habitats, adaptive potential therefore relies on standing genetic variation rather than short-term acclimation, making the conservation of intraspecific genetic diversity critical for ecosystem resilience (Liu et al., 2021). To accurately project biogeochemical feedbacks, models must integrate intraspecific variation and phylogeographic history alongside environmental drivers (Vahsen et al., 2023).

These principles have direct management and biosecurity relevance. P. australis is widely introduced for wetland restoration and can be invasive (Meyerson et al., 2025). Our study shows that key functional traits like NuRE are strongly conserved within lineages. For instance, the invasive European lineage in North America (EU_NA) and its native progenitor (EU) showed no significant divergence in NuRE (Figure 3), a functional stasis consistent with observations of genetic and epigenetic stability during invasion (Liu et al., 2018). Consequently, the transfer or expansion of distinct genotypes (e.g., high-NuRE types) could establish populations with inherent, non-plastic nutrient cycling signatures in new ecosystems. Therefore, the functional traits of source populations must be a key criterion in restoration projects, and monitoring of invasive genotypes should account for their potential to alter local nutrient dynamics.

By evaluating multiple nutrients across diverse genetic lineages, our study reveals that classical resorption hypotheses operate as conditional and hierarchical regulatory strategies (Alseekh et al., 2025; Chen et al., 2021). Their influence is element-specific: concentration control firmly governs P resorption (Kobe et al., 2005) but is disrupted for N under salinity, while stoichiometric control persists throughout (Sun et al., 2023). This integrative framework moves beyond single hypotheses and offers a more realistic basis for predicting plant responses to complex stress. To advance this framework, future studies should (a) combine nutrient additions with salt stress to test interactive effects on resorption control; (b) apply transcriptomic or metabolomic tools to elucidate the regulatory pathways underlying these canalized traits (Huang et al., 2022); and (c) examine whether conservative resorption is commonly observed across other foundation species. These steps are vital for assessing the adaptive capacity of plant communities in a changing world.

In conclusion, this study demonstrates that intraspecific variation in NuRE in P. australis is primarily determined by genetic origin and geographic context, particularly latitude, rather than by plastic responses to non-nutrient stressors. Despite clear evidence of effective salt stress, including substantial biomass reduction, pronounced sodium accumulation, and widespread metabolic reprogramming, NuRE showed no overall plasticity. Instead, we identified distinct, element-specific control strategies: under control conditions, N resorption followed concentration control, a pattern disrupted by salt stress; P resorption was consistently regulated by initial tissue concentration across treatments; and both N and K resorption were under strong stoichiometric control, with evidence of inverted nutrient limitation for P and K relative to N. These patterns reflect evolutionary adaptations to local nutrient conditions and are likely conserved across widely distributed plant species. The limited phenotypic plasticity of NuRE under salt stress suggests that ecosystem resilience to rapid non-nutrient environmental change may depend more on standing genetic variation than on short-term acclimation. Consequently, integrating phylogenetic and macroecological perspectives, particularly the phylogeographic composition of plant populations, is essential for accurately forecasting future nutrient cycling dynamics in a changing world.

The common garden experiment was conducted in Changle County, Weifang City, Shandong Province, China (36°22′ N, 118°53′ E). On April 5, 2024, we collected 110 individuals (genotypes) of P. australis rhizomes from the common garden for cultivation. No specific permits were required for collecting P. samples, as the species is an unprotected weed. To disentangle the effects of deep evolutionary history from local adaptation, we employed a dual classification of our plant materials. The 110 individuals were assigned to six phylogeographic groups (AU, CN_N, CN_E, EU, EU_NA, NA; Supplementary file 1A) based on genetic evidence from chloroplast haplotypes, nuclear microsatellites, and genomic analyses (Liu et al., 2022; Saltonstall, 2002; Wang et al., 2024). Concurrently, the populations from China were categorized into three ecotypes reflecting their source habitats: freshwater wetland, coastal saltmarsh, and inland plateau saltmarsh (Supplementary file 1A). Ecotype classification was not applied to non-Chinese populations due to insufficient habitat documentation at the time of collection.

The plants were planted in barrels (total volume 25 L; top diameter 32.5 cm, bottom diameter 28.4 cm, height 38.5 cm) containing 20 L of a substrate composed of soil, peat moss, and river sand in a 2:1:1 volume ratio (Figure 1—figure supplement 1). The salinity treatment was applied as a 1% NaCl solution (10 ppt), calculated based on the estimated porewater volume of the 20 L soil system. Each individual was planted in two barrels: one for the salt treatment and one for the control. This concentration falls within the range encountered by P. australis in saline wetlands and represents an ecologically recurrent stress that is moderate for most lineages of P. australis but severe for specific genotypes (e.g., those native to North America) (Eller et al., 2017; Meyerson et al., 2025), thus capturing the natural variation in salinity tolerance across its global distribution. This was achieved by adding 100 g of NaCl dissolved in water to each barrel on June 2 and 9, for a total of 200 g. Green leaves and senesced litter were sampled for elemental analysis at two critical phenological stages: one month after salt addition (July 11, 2024; coinciding with active growth) and six months later (December 4, 2024; coinciding with seasonal senescence), respectively.

At each sampling, we collected fully expanded, healthy green leaves from the top of the shoots, typically the second to fourth leaf from the apex. Senesced litter was identified as leaves that had completely yellowed, which typically corresponded to the same second to fourth positional nodes prior to senescence. This standardized protocol ensured the comparison of equivalent physiological tissues across all individuals and treatments.

The plant leaves and litter were dried at 45℃ to constant weight, cut with scissors, placed in a 2 mL centrifuge tube and placed in two small steel balls, and ground with a grinder (Shanghai Jingxin Industrial Development Co., Ltd., Shanghai, China) at 55 Hz for 5 min to powder. Total carbon (C) and nitrogen (N) of leaves was measured using an element analyzer (UNICUBE, Germany) after weighing 2 mg (accurate to three decimal places) of plant samples in tin paper. A total of 0.1 g of the sample was weighed and placed in a polytetrafluoroethylene digestion tube, to which nitric acid was added, and then digested by microwave to obtain the digestive solution. The concentrations of K and P in the digest were determined by ICP-OES (iCAP Pro, Thermo Scientific, Germany).

Besides C, N, P, and K, other seven elements (Cu, Zn, Fe, Mn, Mg, Si, Na) were quantified in leaf tissues. Concentrations of Na, Si, Fe, Mn, and Mg were determined by inductively coupled plasma optical emission spectrometry (ICP-OES; iCAP Pro, Thermo Scientific), while Cu and Zn were measured by inductively coupled plasma mass spectrometry (ICP-MS; NexION 1000G, PerkinElmer, USA).

For untargeted metabolomic profiling, approximately 50 mg of fresh leaf tissue was extracted with 1000 μL of ice-cold extraction solvent (methanol:acetonitrile:water = 2:2:1, v/v/v) containing 20 mg/L of 2-chloro-L-phenylalanine as an internal standard. The mixture was homogenized with steel beads (45 Hz, 10 min), sonicated in an ice-water bath (10 min), and incubated at –20°C for 1 h. After centrifugation (12,000 rpm, 15 min, 4°C), the supernatant was vacuum-dried, and the residue was reconstituted in 160 μL of acetonitrile:water (1:1, v/v). Chromatographic separation was performed on an Acquity UPLC HSS T3 column (1.8 μm, 2.1×100 mm) using a Waters Acquity I-Class PLUS UHPLC system coupled to a Waters Xevo G2-XS QTOF mass spectrometer. The mobile phases consisted of 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B) at a flow rate of 400 μL/min with the following gradient: 0–0.5 min, 5% B; 0.5–5.5 min, 5–50% B; 5.5–9.0 min, 50–95% B; 9.0–10.5 min, 95% B; 10.5–12.0 min, 95–5% B. The injection volume was 2 μL. Mass spectrometry was operated in both positive and negative ion modes with an electrospray ionization source; capillary voltage was set to 2500 V (positive) or –2000 V (negative), cone voltage to 30 V, ion source temperature to 100°C, desolvation gas temperature to 500°C, cone gas flow to 50 L/h, and desolvation gas flow to 800 L/h. Data were acquired in MSe mode over a mass range of m/z 50–1200, alternating between low (off) and high (10–40 V) collision energies. Raw data were processed using Progenesis QI software for peak extraction, alignment, and annotation against the online METLIN database, public databases, and an in-house library.

For statistical analysis, raw metabolomic data were log₂-transformed, missing values were imputed with half of the minimum positive value, and constant metabolites were removed. Data were autoscaled (mean-centered and scaled to unit variance) prior to PCA. Permutational multivariate analysis of variance (PERMANOVA) with 9999 permutations was performed on Euclidean distances to test the effect of salinity while accounting for genotype as a covariate. Orthogonal partial least-squares discriminant analysis (OPLS-DA) was used to extract VIP scores. Differentially accumulated metabolites between salt and control groups were identified using limma with empirical Bayes moderation; significance was defined as |log₂ fold change|>1, FDR<0.05, and VIP >1.

The continuous growth strategy and substantial leaf mass heterogeneity in P. australis preclude reliable mass loss correction factor determination. Uncontrolled physical losses during asynchronous senescence (notably a high rate of leaf tip abscission) decouple dry mass reduction from physiological nutrient resorption. Therefore, to ensure equitable comparisons, nutrient concentration in senesced leaves were standardized for carbon-basis shifts using the fractional change in the measurement basis protocol. Specifically, litter nutrient concentrations were multiplied by the ratio of green leaf carbon content to litter carbon content (Cgreen/Clitter) (Prieto and Querejeta, 2020). The Cgreen/Clitter ratios used for standardization varied among samples, with a mean of 1.06; a one-sample t-test confirmed this mean was significantly greater than 1 (p<0.001), justifying the correction for a systematic shift in the carbon measurement basis. This approach directly quantifies nutrient withdrawal per structural carbon unit, neutralizing mass heterogeneity through carbon-standardization, bypassing fragmentation artifacts, and remaining valid across continuous growth gradients. Finally, NuRE for N, P, and K was calculated as the percentage of nutrient recovered from senescing leaves: NuRE = (Green nutrient concentration – Litter nutrient concentration) / Green nutrient concentration ×100. Raw NuRE was calculated without carbon standardization and compared with carbon-standardized NuRE using Pearson correlations; the main linear mixed-effects analyses were also repeated using raw NuRE.

A PCA was conducted to visualize the overall structure of multi-element composition (N, P, K) across leaf types and salinity treatments, using the rda function in the vegan package after standardizing the data.

Prior to multivariate analysis, all 11 elemental concentrations were standardized to zero mean and unit variance. A global PCA was performed using the FactoMineR package in R. To evaluate the overall effect of salinity on elemental composition, a MANOVA was conducted on PC1 and PC2 scores using Pillai’s trace test. Pairwise comparisons between control and salt treatments were performed using Welch’s t-test for PC1 and PC2 separately. To assess whether salinity effects differed across genetic lineages and developmental stages, MANOVA was repeated for each lineage ×stage combination.

Phylogeographic group and ecotype were analyzed in separate linear mixed-effects models because ecotype classifications were available only for Chinese populations and were partly associated with phylogeographic structure. Each model included salinity treatment and either phylogeographic group or ecotype as fixed effects, with genotype fitted as a random effect to account for the paired experimental design in which each genotype was exposed to both control and salt conditions. When significant effects were detected (p<0.05), post hoc pairwise comparisons were performed using Tukey’s HSD test via the emmeans package, with additional sensitivity analyses conducted on the within-genotype difference (Δ) between salt and control treatments.

The three fundamental control strategies over leaf nutrient resorption were assessed following established conceptual frameworks. To assess the nutrient concentration control strategy, we analyzed the relationships between log10-transformed nutrient concentrations in green leaves and senesced litter. A slope (β) significantly greater than 1 was interpreted as evidence of nutrient concentration control, indicating that resorption efficiency decreases as green leaf nutrient concentration increases. Conversely, a slope not significantly different from or less than 1 suggested the absence of this control strategy. To evaluate the stoichiometric control strategy, we examined the relationships between N and P resorption efficiencies, as well as between the resorbed N:P ratio and the N:P ratio in green leaves (Yang et al., 2025). Consistent positive correlations in both relationships were taken as support for stoichiometric control, reflecting proportional resorption aligned with initial nutrient ratios in green tissues. Nonsignificant or negative correlations indicated a deviation from this strategy. For the nutrient limitation control strategy, we tested the relationship between log10-transformed ratios of resorbed nutrients (e.g., N:P) and log10-transformed ratios in green leaves (Kobe et al., 2005). A slope significantly less than 1 was considered evidence of nutrient limitation control, implying that the resorption efficiency ratio between elements shifts in response to their relative limitation. A slope ≥1 suggested the ‘inverted’ nutrient limitation control (Sun et al., 2023).

To rigorously estimate the functional (structural) slopes between bivariate nutrient traits (e.g., green leaf vs. senesced litter concentrations), we employed standardized major axis (SMA) regression implemented via the smatr package in R (Warton et al., 2012). SMA regression was selected as the primary analytical method because it is a Model II regression approach that appropriately accounts for measurement and biological variation in both variables—a critical consideration given the observational nature of our data where both nutrient concentrations in green leaves and senesced litter are measured with error. This method provides unbiased estimates of the true structural relationship between variables, unlike ordinary least squares (OLS) regression that assumes the predictor variable is error-free and systematically underestimates slope magnitudes under these conditions (Warton et al., 2012).

For each bivariate relationship, we performed SMA regression analyses separately for each salinity level. The smatr package was used to estimate the SMA slope (β) along with its 95% confidence interval and to calculate the coefficient of determination (R²). Crucially, we specifically tested whether each estimated SMA slope significantly deviated from theoretical values, particularly from 1. This slope comparison against 1 was conducted using the built-in permutation tests in the smatr package, which generated p-values indicating the statistical significance of the deviation. For comparative purposes and to maintain consistency with the methodological approaches employed in foundational studies of nutrient resorption (e.g., Kobe et al., 2005; Sun et al., 2023) that relied on OLS regression, we also calculated and reported OLS slope estimates in supplementary materials. However, all formal hypothesis testing and interpretation regarding slope values, particularly tests of deviation from 1:1 relationships, are based on the more appropriate SMA regression results, which represent robust, unbiased estimates of the functional relationships between nutrient traits.

The influence of geographical origin (absolute latitude) on NuRE was analyzed using linear mixed-effects models (LMMs) fitted with the lmer function in R package lme4 (Bates et al., 2015), where phylogeographic group was included as a random intercept term to account for non-independence among individuals from the same source. Model assumptions of normality and homoscedasticity were verified by inspecting residual plots, and NuRE values were arcsine-square-root transformed when necessary to meet these assumptions. To quantify the individual contributions of the fixed factors (latitude, leaf nutrient concentration, and stoichiometry) to the variance in NuRE, we performed a variation partitioning analysis following the LMMs with phylogeographic group as a random intercept. This variation partitioning was conducted using the glmm.hp package in R (Lai et al., 2022), which implements hierarchical partitioning to calculate the marginal R² for each predictor within the fitted linear mixed-effects model framework. This approach allowed us to estimate the unique proportion of variance in NuRE attributable to each fixed factor, independent of the random effect of phylogeographic group. All figures were generated using ggplot2 and assembled with the patchwork library. All statistical analyses were performed in R version 4.3.0.

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

This work was supported by the National Natural Science Foundation of China (32470388; U22A20558; 42501350) and Shandong Provincial Natural Science Foundation (ZR2024QC197). Lele Liu is supported by the Cyrus Tang Foundation through the Tang Scholar Program.

1. Preprint posted: May 19, 2026
2. Sent for peer review: May 19, 2026
3. Reviewed Preprint version 1: August 13, 2026
4. Reviewed Preprint version 2: October 5, 2026
5. Version of Record published: October 26, 2026

You can cite all versions using the DOI https://doi.org/10.7554/eLife.111792. This DOI represents all versions, and will always resolve to the latest one.

© 2026, Liu, Sheng et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.