Untargeted metabolomics reveals anion and organ‐specific metabolic responses of salinity tolerance in willow

SUMMARY

Willows can alleviate soil salinisation while generating sustainable feedstock for biorefinery, yet the metabolomic adaptations underlying their tolerance remain poorly understood. Salix miyabeana was treated with two environmentally abundant salts, NaCl and Na₂SO₄, in a 12‐week pot trial. Willows tolerated salts across all treatments (up to 9.1 dS m⁻¹ soil EC_e), maintaining biomass while selectively partitioning ions, confining Na⁺ to roots and accumulating Cl⁻ and ${\mathrm{SO}}_4^{2-}$ in the canopy and adapting to osmotic stress via reduced stomatal conductance. Untargeted metabolomics captured >5000 putative compounds, including 278 core willow metabolome compounds constitutively produced across organs. Across all treatments, salinity drove widespread metabolic reprogramming, altering 28% of the overall metabolome, with organ‐tailored strategies. Comparing salt forms at equimolar sodium, shared differentially abundant metabolites were limited to 3% of the metabolome, representing the generalised salinity response, predominantly in roots. Anion‐specific metabolomic responses were extensive. NaCl reduced carbohydrates and tricarboxylic acid cycle intermediates, suggesting potential carbon and energy resource pressure, and accumulated root structuring compounds, antioxidant flavonoids, and fatty acids. Na₂SO₄ salinity triggered accumulation of sulphur‐containing larger peptides, suggesting excess sulphate incorporation leverages ion toxicity to produce specialised salt‐tolerance‐associated metabolites. This high‐depth picture of the willow metabolome underscores the importance of capturing plant adaptations to salt stress at organ scale and considering ion‐specific contributions to soil salinity.

Significance Statement

As a biorefinery and phytoremediation crop, willow could help address the global challenge of soil salinisation; yet, despite broad abiotic stress tolerance, the metabolic strategy underlying salinity tolerance in this non‐model crop remains poorly understood. This study uses untargeted metabolomics to capture responses during salinity tolerance, revealing extensive organ‐ and anion‐specific changes, including evidence of sulphur incorporation in ion toxicity mitigation and emphasising the need to consider salt type as a driving factor of soil salinity.

INTRODUCTION

Soil salinisation represents a major threat to agricultural productivity and environmental health on a global scale. It is estimated that over 20% of irrigated cropland is impacted by salt accumulation (FAO, ²⁰¹⁵), reducing yields and critical ecosystem biodiversity (Zörb et al., 2019). Climate change and human activities are predicted to intensify both natural and anthropogenic soil salinisation processes, further threatening food security (Singh, 2021). Soil salinity is commonly assessed using electroconductivity of water‐saturated soil extract (EC_e), typically categorised as non‐saline (<2 dS m⁻¹), slightly saline (2–4 dS m⁻¹), moderately saline (4–8 dS m⁻¹) and strongly saline (>8 dS m⁻¹), with few plants, primarily halophytes, capable of tolerating EC levels above 8 dS m⁻¹ (FAO, ¹⁹⁸⁸). While salt stress research typically investigates NaCl, Na₂SO₄ represents another widespread salt derived from anthropogenic sources like industrial sulphur emissions and agricultural practices, as well as natural volcanic and marine environments (Dick et al., 2008). The use of salt‐tolerant biomass crops able to sequester salt and improve soil health could help reclaim and repurpose salt‐impacted areas (Abreu et al., 2022; Quinn et al., ²⁰¹⁵).

Soil salinisation results from soluble ions accumulation such as Na⁺, K⁺, Mg²⁺, Ca²⁺, Cl⁻, ${\mathrm{SO}}_4^{2-}$ and ${\mathrm{CO}}_3^{2-}$. When in contact with roots, these ions can induce both osmotic stress and ion toxicity (Munns & Tester, 2008). The osmotic component, primarily driven by any dissolved salt concentration, alters water potential gradients, disrupting cell turgor pressure and elongation. In response, stomatal closure limits water loss via transpiration, yet compromises photosynthate supply and light energy balance, ultimately impairing growth through water deficit, reduced carbon availability and photo‐oxidative damages (Munns, Passioura, et al., 2020). Ion toxicity, however, is dependent on the nature of the ion, with different elements inducing unique patterns of signalling, transport, oxidative damage, energy allocation and hormone‐regulated growth processes (Llanes et al., 2013; Reginato et al., ²⁰²¹; Richter et al., ²⁰¹⁹).

Chloride (Cl⁻), an essential micronutrient for plants, contributes to vital physiological processes (Wege et al., 2017). As a major osmotically active solute, chloride fluxes are involved in vacuolar turgor required for cell expansion and are implicated in stomatal aperture regulation, contributing to water balance and gas exchange (Teakle & Tyerman, 2010). Additionally, chloride acts as a counter‐ion, regulating membrane potential, intracellular pH gradients and electrical signalling in the cytoplasm (White & Broadley, 2001) and serves as a cofactor in the oxygen‐evolving complex of photosystem II (Ifuku, 2015). However, in salt‐impacted soils, excess chloride damages plants, leading to impaired photosynthesis (Geilfus, 2018) and reduced ${\mathrm{NO}}_3^{-}$ uptake (Abdelgadir et al., 2005). In contrast, sulphate (${\mathrm{SO}}_4^{2-}$), the major bioavailable form of sulphur, is an essential macronutrient involved in physiological and biochemical processes (Prasad & Shivay, 2018; Watanabe & Hoefgen, ²⁰¹⁹). Sulphur is required for the production of the amino acids cysteine and methionine, peptide derivatives such as glutathione and disulfide bonds contributing to enzyme activity and stability. The production of chlorophyll, vitamins and certain plant defence specialised metabolites, such as glucosinolates, also depend upon sulphur (Chan et al., 2019; Rausch & Wachter, ²⁰⁰⁵). Plants regulate sulphate uptake and assimilation through sulphate‐specific transporters to avoid toxicity (Buchner et al., 2004); however, excess sulphate ions can create nutrient deficiency by competing with the absorption of other ions, especially calcium and phosphorus, as demonstrated in Brassica rapa (Reich et al., 2017).

Plants mediate salt stress through a coordinated series of physiological, structural, biochemical and molecular adjustments (Arif et al., 2020; Deinlein et al., ²⁰¹⁴; Gupta & Huang, ²⁰¹⁴; van Zelm et al., ²⁰²⁰; Yang & Guo, ²⁰¹⁸). These responses vary between species (Sanchez et al., 2008) and depend on the type of ions involved (Reich et al., 2017; Richter et al., ²⁰¹⁹). A primary adaptive response involves osmotic stress management through the accumulation of inorganic ions (Rodriguez et al., 1997; Wege et al., ²⁰¹⁷) or organic osmoprotectants such as amino acids, polyamines, quaternary ammonium compounds, sugars and polyols (D'Amelia et al., ²⁰¹⁸; Slama et al., ²⁰¹⁵), although this response is neither exclusively induced by excess salt nor dependent on salt type (Munns, 2002). While the biosynthesis of these osmoprotectants imposes a metabolic energy cost (Munns, Day, et al., 2020), their functions extend beyond osmotic adjustment to detoxification processes such as antioxidant activity and chaperone‐mediated protein stabilisation (Yancey, 2005). Other key tolerance mechanisms include ion homeostasis through selective transport and compartmentalisation into the vacuole or apoplast (Bose et al., 2017; Manishankar et al., ²⁰¹⁸), upregulation of antioxidant defence (Gill & Tuteja, ^2010a), growth modulation via hormonal signalling (Ryu & Cho, 2015), xylem alteration and loading (Janz et al., 2012), modification of root architecture (Zou et al., 2022) and symbiotic associations with microorganisms such as arbuscular mycorrhizal fungi (Evelin et al., 2009).

Salicaceae, including poplars (Populus sp.) and willows (Salix sp.), are common phytoremediation and biorefinery crops on marginal lands (Baker et al., 2022; Gonzalez et al., ²⁰¹⁸; Rodzkin & Volk, ²⁰¹⁸), with high potential for growth on salinised soils. Some cultivars can maintain transpiration and growth under saline conditions up to 6–8 dS m⁻¹ (Bilek et al., 2020; Chen & Polle, ²⁰¹⁰; Huang et al., ²⁰²⁰; Mirck & Zalesny, ²⁰¹⁵). This tolerance likely engages high metabolic diversity, including reactive oxygen species (ROS) protective phenolics and other antioxidants (Sui & Wang, 2020; Zhou et al., ²⁰²⁰). However, the understanding of the metabolomic toolkit employed beyond these compounds, and their variation across different organs and salts, remains limited.

In this pot‐based greenhouse study, Salix miyabeana was exposed to (moderate) NaCl and Na₂SO₄ treatments (EC_e ~ 6 dS m⁻¹). Responses to these distinct salt treatments, which provided equal Na⁺ concentrations and similar osmotic pressure, are compared to detect anion‐specific components of salinity tolerance. Physiological trait measurements, ionic profiling and untargeted metabolomics across roots, stems and leaves were assessed with the aim of delineating general and ion‐specific metabolomic responses to salinity across multiple organs (Figure 1). Additionally, a high Na₂SO₄ treatment (EC_e ~ 9 dS m⁻¹) was included to further explore organ‐specific metabolic responses under sulphate‐dominated salinity stress and to evaluate the effects of increased ionic and osmotic stress conditions as compared to moderate Na₂SO₄.

Figure 1.

Workflow for untargeted metabolomics approach.

A randomised pot trial treated willows with salt solutions or water controls (7 replicates per treatment) to explore variation in anion‐specific salt tolerance mechanisms across organs. Metabolites were extracted from root, stem and leaf samples and analysed by LC–MS/MS. Resulting chromatographic data were pre‐processed using MZmine for feature mass detection, cross‐sample alignment and quality filters. Statistical analysis used a cross‐sample aligned peak area table. After log transformation and Eigen MS normalisation, univariate analysis was used to factorially compare feature peak areas across willow organs in controls, while pairwise comparisons were used to identify differentially abundant (DA) compounds between each salt treatment compared to controls. Parametric or non‐parametric tests were used based on feature peak area normality and variance homogeneity before a false discovery rate correction was applied to account for multiple tests. Annotation strategy used precursors information and corresponding spectral fingerprints (when available) to query multiple databases and for in silico structure prediction to generate a list of annotation candidates (unsupervised annotation). After standardising compounds identity information and assigning prioritisation criteria, the best candidate was retained. Annotation was further refined for compounds identified as differentially abundant due to salt treatment and used data‐driven (structural similarity networking) and knowledge‐driven (literature) curation.

RESULTS

Physiological effects of salt treatments and mineral elements accumulation

After 84 days of growth, including 28 days of salinity treatments (Figure S1C), soil electrical conductivity (EC_e) was 0.9 (±0.2 standard error [SE]) dS m⁻¹ in controls (Figure 2A) and was significantly higher in both moderate NaCl and Na₂SO₄ treatments, at 6.5 (±0.2 SE) and 5.5 (±0.5 SE) dS m⁻¹ (Tukey HSD, α < 0.05). High Na₂SO₄ treatment significantly increased soil conductivity further to 9.1 (±0.5 SE) dS m⁻¹.

Figure 2.

Physiological assessment of salt stress on Salix miyabeana ‘SX64’.

(A) Following 56 days of root establishment, salinity in planted pots was progressively built up over 28 days to reach the final saturated soil electroconductivity (EC), as inferred from a (1:1) soil‐to‐water extract.

(B) Representative willow photographs, on harvest day, in response to salt treatments (white scale bars: 50 cm).

(C–F) The impact of moderate NaCl, moderate Na₂SO₄ and high Na₂SO₄ was quantified on (C) total biomass (dry weight), (D) photosystem II efficiency evaluated using the LI‐COR 600 parameter phiPSII, (E) stomatal conductance g _sw, (F) stems and leaves moisture contents. Boxes represent median and quartiles while whiskers extend to the largest value within 1.5 times the inter‐quartile range (IQR) from the hinge. Dots represent individual plants (n = 7). Significant differences (P < 0.05) between treatments are indicated by letters above the boxes and were determined by two‐way ANOVA followed by Tukey's post hoc test.

S. miyabeana maintained a healthy phenotype throughout the experiment (Figure 2B) and salt treatments did not induce differences in aboveground growth rates or organ biomass yields compared to controls (Figures S1–S3), with total biomass averaging 131 (±5 SE) g dry matter (DM) for the control and all salt treatments (Figure 2C). Quantum efficiency Φ_PSII ranged from 0.44 to 0.61 and was not significantly different between treatments (Figure 2D). Salt‐treated willows showed significant reductions in stomatal conductance to water vapour (g _sw) ranging from 0.13 to 0.19 mol m⁻² sec⁻¹ (Figure 2E), compared to g _sw of 0.29 (±0.04 SE) mol m⁻² sec⁻¹ in control willows. Leaf moisture content (MC) was 1.97 (±0.07 SE) $g_{{\mathrm{H}}_2\mathrm{O}}$ $g_{\mathrm{DM}}^{-1}$ in controls and significantly increased to 2.28 (±0.10 SE) $g_{{\mathrm{H}}_2\mathrm{O}}$ $g_{\mathrm{DM}}^{-1}$ under moderate NaCl treatment. Moderate Na₂SO₄ resulted in intermediate leaf MC, at 2.09 (±0.04 SE) $g_{{\mathrm{H}}_2\mathrm{O}}$ $g_{\mathrm{DM}}^{-1}$, while high Na₂SO₄ treatment significantly decreased compared to both moderate salt treatments, measuring 1.75 (±0.04 SE) $g_{{\mathrm{H}}_2\mathrm{O}}$ $g_{\mathrm{DM}}^{-1}$ (Figure 2F). None of the treatments induced significant differences in stem moisture content, which was maintained at an average of 1.34 (±0.03 SE) $g_{{\mathrm{H}}_2\mathrm{O}}$ $g_{\mathrm{DM}}^{-1}$.

Salt treatments distinctly altered ion total contents and concentrations across plant organs (Figure 3). Total sodium increased from 2.78 (±0.1 SE) mmol(Na) in controls to 9.4–13.0 mmol(Na) in salt‐treated willows (Figure 3A), with the highest concentration in roots (3‐fold above controls) (Figure 3B). NaCl treatment elevated total chloride content to 26.3 (±2.1 SE) mmol(Cl) as compared to 6.2 (±0.7 SE) mmol(Cl) in controls, particularly in leaves (3.4‐fold above controls). Both Na₂SO₄ treatments resulted in similar total sulphur content of 14.1–15.7 mmol(S), as compared to 7.4 (±0.4 SE) mmol(S) in controls; however, the proportion of inorganic sulphur ${\mathrm{SO}}_4^{2-}$ did not vary significantly across all treatments, averaging 27% of the total sulphur (Figure S5). Total potassium content was 21.3 (±0.6 SE) mmol(K) in controls and significantly higher in both moderate and high Na₂SO₄ treatments, ranging from 28.4 to 28.9 mmol(K). Across all four treatments, there were no significant variations in total phosphorus, calcium and magnesium content (Figure 3A). However, compared to controls, leaf calcium concentration decreased in both Na₂SO₄‐treated plants, while leaf magnesium concentration increased in NaCl‐treated plants only (Figure 3B).

Figure 3.

Elemental analysis of mineral nutrients.

Elemental analyses include sodium (Na), chloride (Cl), sulphur (S), potassium (K), phosphorus (P), calcium (Ca) and magnesium (Mg).

(A) Total element content (mmol) combined from the leaves, stems and roots presented. The bars indicate mean values within each organ ± standard error (n = 3).

(B) Element concentration within a single organ (μmol g_DW ⁻¹) is shown. The dots represent mean values ± standard error (n = 3). Letters indicate significant differences (P < 0.05) between control, moderate NaCl, moderate Na₂SO₄ and high Na₂SO₄ treatments and were determined using two‐way ANOVA followed by Tukey's post hoc test. Mineral elements ratios (K⁺/Na⁺; Ca²⁺/Na⁺; K⁺/Ca²⁺) are presented in Figure S4.

Organ partitioning of the Salix metabolome

After 84 days of growth, untargeted metabolomic profiling of control willows, without salt treatment, revealed a total of 5043 metabolic compounds across all three organs, with 2063 compounds in roots, 2950 compounds in stems, and 3601 compounds in leaves (Figure 4A). Shannon and inverse Simpson indices showed a significant sequential increase in metabolic diversity from roots to stems to leaves (Figure 4B). Principal coordinate analysis (PCoA) separated samples based on organ (Figure 4C), PC1 and PC2 explaining 52 and 26% of variance, respectively, with significant differences between groups (PERMANOVA, P‐value <0.05). Leaves and stems exclusively shared 481 compounds, stems and roots exclusively shared 371 compounds while leaves and roots exclusively shared 83 compounds (Figure 4D).

Figure 4.

Willow metabolic composition.

Organs (leaves, stems and roots) were separately collected from control willows and analysed for metabolomic profiling.

(A) Total number of consistently detected metabolites within each major organ (present in at least 4 replicates out of 7).

(B) Comparison of organ metabolic richness using Shannon and Inverse Simpson diversity indices. Boxes represent median and quartiles while whiskers extend to the largest value within 1.5 times the inter‐quartile range (IQR) from the hinge. Dots represent individual plants (n = 7). Different letters above the boxes indicate significant differences between organs (ANOVA, Tukey HSD, P < 0.05).

(C) Principal coordinate analysis depicting dissimilarity in organ composition based on Bray–Curtis distances; ellipses are for standard deviation.

(D) Venn diagram illustrating the distribution of metabolites across organs as unique to one or shared between multiple.

(E) Cumulative peak areas of metabolites present in each organ are represented as stacked bar charts. Each bar is divided into three categories: unique to the respective organ, shared with one other organ and shared across all three.

(F) Cumulative peak area and number of core metabolites (compounds present across the entire plant, with no significant variation of abundance between organs) classified per metabolic pathways, with composition details of main superclasses.

(G) Count of compounds is significantly higher in one organ compared to the other two (ANOVA, Tukey HSD, P < 0.05), revealing dominant organ‐specialisation of some metabolic superclasses. Superclass abbreviations are as follows: ac., acid; alk., alkaloid; gly., glycoside.

A total of 1318 compounds were present in all three organs (Figure 4D), representing 73–96% of the cumulative peak area in each organ (Figure 4E). Differential abundance analysis between organs showed that, of these shared compounds, 278 did not significantly differ. They were defined as ‘core metabolites’ and included 24 (iso)flavonoids, 20 phenolic acids, 9 phenylpropanoids, 22 small peptides, 14 pseudoalkaloids, 8 tryptophan alkaloids, 10 nucleosides, 5 saccharides, 6 monoterpenoids and 2 fatty acids and conjugates (Figure 4F). High‐intensity compounds included adenosine, glutamate, myrciacitrin III, salicortin, salicin, benzaldehyde, nicotinamide, galactosylglycerol and fructose (Tables S3 and S4).

Conversely, 3259 compounds were organ‐specialised metabolites with significantly increased abundance in one organ, including 1719 compounds exclusively detected in leaves, 780 in stems and 291 in roots. Leaf‐specialised compounds were dominated by the terpenoid pathway (Figure 4G), with a high number of (seco)iridoid monoterpenoids, diterpenoids, sesquiterpenoids, steroids, as well as the alkaloids pathway, particularly tyrosine and tryptophan derived alkaloids. Stem‐specialised compounds were dominated by shikimates and phenylpropanoids pathway, with a larger relative number of lignans, along with fatty esters. Root‐specialised compounds were dominated by highly water‐soluble (polar) compounds (Figure S6), including saccharides and nucleosides. Each organ also had a highly specific set of flavonoids and small peptides.

Salix metabolomic response to moderate and high Na₂SO₄

Principle component analysis (PCA) separated samples by control, moderate and high Na₂SO₄ treatments in roots (PERMANOVA, P‐value <0.05), with PC1 explaining 17% and PC2 explaining 11% of the variance, while no separation was observed between treatments in stems and leaves (Figure 5A). Moderate Na₂SO₄ led to 571 differentially abundant (DA) compounds across the whole plant, including 218 in roots, 105 in stems and 248 in leaves (Figure 5B,D). Comparatively, high Na₂SO₄ led to 976 DA compounds across the whole plant, including 574 in roots, 131 in stems and 271 in leaves (Figure 5C,D).

Figure 5.

Willow metabolic response to sodium sulfate (Na₂SO₄) salt and its variation with dose.

(A) Principal coordinate analysis (PCoA) based on Bray–Curtis distance, illustrating metabolic profile dissimilarity between control, moderate Na₂SO₄ and high Na₂SO₄ at organ scale (n = 7), including standard deviation ellipse (95% confidence interval).

(B) Organ distribution of differentially abundant (DA) compounds between control and moderate concentration Na₂SO₄‐treated willows (P < 0.05).

(C) Organ distribution of differentially abundant compounds between control and high concentration Na₂SO₄‐treated willows (P < 0.05).

(D) Overview of significantly increased and decreased compounds across organs, comparing DA of control against moderate Na₂SO₄ and control against high Na₂SO₄.

(E) Distribution of increased and decreased DA compounds across metabolic pathways for moderate and high Na₂SO₄ treatments compared to controls.

(F) Metabolic response overlap across both Na₂SO₄ concentrations, representing the number of DA compounds common and unique to a single dose DA, at the organ scale.

(G) Dose‐dependent correlation of compounds consistently DA across Na₂SO₄ levels within the same organ (excluding salt‐induced/salt‐suppressed compounds with infinite fold change), representing the fold change (log2) of average peak areas (n = 7) under moderate Na₂SO₄ relative to control against high Na₂SO₄ relative to control.

(H) Sub‐selection of main superclasses altered by moderate or high concentration of Na₂SO₄ compared to controls, illustrating increased and decreased compound distribution at organ scale. DA compounds shared between both treatment levels are highlighted by the darker section of the bars.

(I) Distribution of increased and decreased sulphur‐containing DA compounds at organ scale, comparing both levels of Na₂SO₄. Shared compounds are highlighted by the darker section of the bars.

(J) Molecular network of compounds MS/MS spectra similarities (cosine >0.6) computed in the GNPS environment and visualised in Cytoscape. Dots are colour‐coded based on the presence of sulphur (pink represents S‐containing compounds) and dot size highlights significantly different compounds (larger dots are DA compounds). Major clusters containing S‐compounds and structurally similar DA compounds are circled by an ellipse.

Taken together, the Na₂SO₄‐salt response involved 989 DA compounds distributed across the main pathways as 255 shikimates/phenylpropanoids, 102 alkaloids, 98 carbohydrates, 82 amino acids/peptides, 72 terpenoids, 50 fatty acids and 19 polyketides (Figure 5E). In response to moderate Na₂SO₄ treatment, DA compounds in roots and stems were mostly depleted (58–60%), while DA compounds in leaves were mostly enriched (85%) (Figure 5D,H). Higher Na₂SO₄ induced a larger metabolite enrichment across most pathways, except for further depletion in carbohydrates, while stems maintained a reduced metabolite pattern and leaves maintained increased metabolites across the major pathways.

Over the whole plant, 327 compounds were consistently altered in both moderate and high Na₂SO₄ treatments, representing 21% of DA compounds in roots, 24% in stems and 53% in leaves (Figure 5F). This consistent response was further reflected in an R ² = 0.93 correlation between the fold changes (log₂) moderate and high Na₂SO₄ treatments compared to controls (Figure 5G).

In roots, moderate Na₂SO₄ reduced 30 compounds within the shikimates/phenylpropanoids pathway, including 10 flavonoids, 6 phenolic acids and 4 lignans (Figure 5H). Comparatively, high Na₂SO₄ treatment increased 106 compounds from that pathway, including 39 (iso)flavonoids, 22 phenolic acids, 12 phenylpropanoids, as well as 9 lignans. These included the accumulation of compounds such as 4‐coumaroyl shikimate, syringin, procyanidol, catechin 3‐O‐(1‐hydroxy‐6‐oxo‐2‐cyclohexene‐1‐carboxylate), 8‐hydroxypinoresinol 4'‐glucoside, guaicylglycerol‐coniferyl ether, glucose‐conjugated salicylic acid and salicin. Five sphingolipids, including phytosphingosine, were consistently depleted by both Na₂SO₄ concentrations, while 2 and 9 saccharides were significantly reduced under moderate and high Na₂SO₄, respectively. In contrast, 14 nucleosides and 18 small peptides increased under high Na₂SO₄, as did 10 tryptophan alkaloids, 5 tyrosine alkaloids, 6 monoterpenoids, 11 steroids and 6 triterpenoids. In stems, moderate and high Na₂SO₄ treatments enriched 3 and 7 small peptides, such as proline, while 4 and 6 small peptides were depleted, respectively. Additionally, 4 and 12 saccharides were depleted, including glucose and trehalose (Figure 5H), whereas 11 and 10 saccharides were enriched in leaves. Common, dose‐responsive compounds in leaves were dominated by 44 shikimates/phenylpropanoids, encompassing the accumulation of 13 flavonoids, 9 phenylpropanoids and 9 phenolic acids. Commonly regulated metabolites also included 23 alkaloids and 16 amino acids/peptides (Figure 5H).

A total of 139 DA compounds altered by one or both concentrations of Na₂SO₄ (9% of all metabolomic changes induced by sulphur salt treatments) were annotated as sulphur‐containing (S‐compounds), corresponding to 79 distinct compounds. These S‐compounds similarly accumulated in leaves under both Na₂SO₄ concentrations (26–27 compounds), while their accumulation in roots was markedly higher under high Na₂SO₄ (37 compounds) as compared to moderate Na₂SO₄ (8 compounds) (Figure 5I). Increased S‐compounds included peptides such as (E)‐1‐(l‐cystein‐S‐yl)‐N‐hydroxy‐omega‐(methylsulfanyl)hexan‐1‐imine and N‐[(benzyloxy)carbonyl]‐l‐cysteinylglycine; nucleosides such as S‐adenosylmethionine; and stilbenes such as resveratrol 4‐sulphate. Depleted S‐containing compounds putatively included peptides such as S‐(hydroxymethyl)glutathione; and indoles such as camalexin. While high‐confidence annotation of these S‐compounds was limited, a subgroup of 13 DA S‐compounds was within the same structural cluster (Figure 5J).

Salix metabolomic response to NaCl treatment

Comparing metabolomes from controls with NaCl‐treated willows, PCA revealed a significant separation between control and NaCl treatment in roots (PERMANOVA, P‐value <0.05), partial overlap in stems (PERMANOVA, P‐value <0.05) and no observed separation in leaves (Figure 6A). NaCl treatment led to DA in 1063 DA compounds, with 475 in roots, 481 in stems and 107 in leaves (Figure 6B). Among these, 768 DA compounds were unique to one organ, while 135 were shared among two or all organs (Figure 6C). Notably, 98 DA compounds were exclusively shared between roots and stems, and 24 were significantly depleted in all three organs due to NaCl treatment.

Figure 6.

Willow's metabolic response to sodium chloride (NaCl) salt.

(A) Principal coordinate analysis (PCoA) based on Bray–Curtis distance, illustrating metabolic profile dissimilarity between control and moderate NaCl at organ scale (n = 7), including standard deviation ellipse (95% confidence interval).

(B) Overview of significantly increased and decreased compounds across organs.

(C) Organ distribution of differentially abundant (DA) compounds between control and moderate concentration NaCl‐treated willows (P < 0.05).

(D) Distribution of increased and decreased DA compounds across metabolic pathways.

(E) Metabolic profile at pathway level of DA compounds shared between multiple organs and unique to roots, stems or leaves.

(F–I) Sub‐selection of main pathways or superclasses altered by moderate NaCl compared to controls, illustrating increased and decreased compound distribution at organ scale for (F) carbohydrates, including saccharides and nucleosides, (G) small peptides, (H) and 3 major superclasses of shikimates and phenylpropanoids pathway: salicinoids, lignans, phenylpropanoids, flavonoids, as well as (I) terpenoids, including monoterpenoids and steroids. Dots are colour‐coded by regulation (blue: decreased, red: increased by salt treatment compared to control). Base mean area reflects the average peak area across all compared samples.

Across all organs, NaCl‐responsive metabolites comprised 206 DA compounds within the shikimates/phenylpropanoids pathway, 92 alkaloids, 92 carbohydrates, 77 amino acids/peptides, 46 terpenoids, 42 fatty acids and 14 polyketides (Figure 6D). While shikimates/phenylpropanoid pathway compounds were the most diverse, carbohydrates represented the highest proportional change within a pathway, constituting >27% of total carbohydrates (Figure S7). This was driven by significant depletion in 27 carbohydrates across multiple organs, including glucose, fructose and sucrose, as well as organ‐specific depletion of 36 carbohydrates (Figure 6E,F). Tricarboxylic acid cycle intermediates, such as malate, citrate and 2‐oxoglutarate, were also depleted across the entire plant (Figure S8) as well as phosphorylated compounds, such as phosphocholine, glycerophosphoglycerol and (R)‐3,5‐diphosphomevalonate.

Within organs, leaves had a limited response to NaCl treatment (Figure 6B,C), but depleted compounds putatively included 2‐oxoglutarate, malate, aspartate and choline, while accumulated compounds included (indol‐3‐yl)acetyl‐myo‐inositol l‐arabinoside (IAA derivative) and spermidine (Tables S3 and S7). Stems and roots had a larger metabolic response, markedly within the shikimate/phenylpropanoid pathway, altering 117 and 103 compounds, respectively (Figure 6E). This large group of differentially abundant phenolic compounds was primarily associated with flavonoids, phenylpropanoids, lignans and phenolic acids (including salicinoids). Salicinoids were net increased in both stems and roots (Figure 6H). Lignans, phenylpropanoids and flavonoids showed opposite trends between organs, with a net decrease in stems (more compounds were reduced than accumulated), contrasting with a net increase in roots (more compounds were accumulated than reduced) (Figure 6H). Stems had a significant increase of myricetin 3‐O‐glucoside, kaempferide 3‐O‐glucoside, procyanidol, salicin and grandidentatin, alongside the decrease of eriodictyol 7‐glucuronide, ramontoside and a hydroxycyclohexen‐oyl ester catechin derivative. In roots, populin, tremuloidin and benzoyl‐β‐d‐glucoside were decreased, while (iso)salicin, grandidentatin, syringin, coniferin, catechin, procyandin C and salipurposid were increased.

A balanced pattern of metabolomic change was observed in nitrogen metabolism in response to NaCl treatment, with enrichment of 147 DA compounds and depletion of 140 DA compounds. This was particularly evident as 30 small peptides increased across all organs, including proline, tryptophan and glutathione, while 35 small peptides decreased, including acetylproline, threonine and arogenate, along with glutamate (Figure 6G). Other increasing N‐containing compounds included purine nucleosides like adenosine, phytohormones such as kinetin‐7‐N‐glucoside, as well as sphingolipids, such as phytosphingosine and sphinganine. Concurrently, terpenoids were enriched, especially monoterpenoids, with 11 enriched monoterpenoids in stems and 9 in roots (Figure 6I,J).

Salix anion‐specific metabolomics response to NaCl and Na₂SO₄ (at the same EC)

Comparing the 1063 and 571 DA compounds altered by moderate NaCl and moderate Na₂SO₄ revealed 103 DA compounds commonly regulated in roots, 53 in stems and 7 in leaves (Figure 7A,B). These generalised salinity responsive metabolites included 21 shikimates/phenylpropanoids, 11 carbohydrates, 8 amino acids and peptides, 5 terpenoids, 5 alkaloids and 4 fatty acids (6 other and 93 unknown) (Figure 7B). Commonly enriched compounds included (indol‐3‐yl)acetyl‐myo‐inositol l‐arabinoside and 2‐phenylacetamide in leaves; proline, grandidentatin and fraxiresinol 1‐O‐glucoside (a lignan) in stems; salipurposid and syringin in roots. Commonly depleted compounds included glucose and trehalose in stems; glycerophosphoglycerol, galactoglycerol and benzoyl‐beta‐d‐ in roots.

Figure 7.

General willow salt response and anion‐specific tailored responses.

(A) Venn diagrams showing DA metabolic responses overlap in roots, stems and leaves under moderate NaCl and Na₂SO₄ stress.

(B) Pathway profile of commonly increased and decreased compounds in response to moderate salt stress. The fold change graph illustrates the distribution of these compounds at the organ scale. Each dot represents a compound, with its base mean and fold change averaged across both moderate salt treatments.

(C) Superclass profiles of NaCl‐ and Na₂SO₄‐specific responses. Superclasses are categorised as net enriched, equal or net decreased based on the relative number of increased versus decreased compounds. Differential patterns for nucleosides and saccharides are highlighted in bold.

(D) Metabolic response to moderate NaCl and moderate Na₂SO₄ categorised based on compounds distribution across the whole plant: salt‐induced compounds (not detected in controls), core compounds (present in all organs at stable state in controls) and organ‐specialised compounds: roots, stems or leaves (present in a higher abundance in a specific organ, based on controls). Bar graphs are split into decreased and increased compounds to highlight the patterning of the DA compounds.

Anion‐specific responses included 764 DA compounds unique to moderate NaCl treatment and 391 DA compounds unique to the moderate Na₂SO₄. Both salt treatments induced the accumulation of salt‐specific monoterpenoids, as well as phenolic compounds, including flavonoids, phenylpropanoids and phenolic acids. Moreover, NaCl treatment led to a net decrease of saccharides and nucleosides, whereas Na₂SO₄ treatment resulted in their net accumulation (Figure 7C). An additional contrasting effect was observed with sphingolipids, with an increase in NaCl but a decrease in Na₂SO₄ treated trees.

An anion‐specific effect could also be observed when comparing the change in constitutively produced compounds that were core (present and relatively stable between organs) and organ‐specialised (significantly different between organs), but also in salt‐induced compounds (not detected in controls). In both moderate salt treatments, roots primarily reduced root‐specialised compounds rather than core compounds, while significantly enriched compounds were predominantly newly synthesised (salt‐induced), rather than root‐specialised or core compounds (Figure 7D). Similarly, in stems, reduced compounds were largely stem‐specialised rather than core compounds under both salt types; however, enriched compounds were predominantly stem‐specialised under NaCl, comprising 53 enriched DA compounds as compared to 20 salt‐induced and 23 core enriched DA compounds. Conversely, Na₂SO₄ treatment resulted in 2 stem‐specialised, 5 salt‐induced and 3 core‐enriched DA compounds. In leaves, NaCl treatment reduced 13 leaf‐specialised and 4‐core DA compounds while enriching 26 leaf‐specialised, 20 salt‐induced and 4‐core DA compounds. In contrast, leaves treated with Na₂SO₄ decreased 22 leaf‐specialised and 4 core compounds while increasing 5 leaf‐specialised and 63 core compounds.

DISCUSSION

Salix salinity tolerance response

The comparison of a non‐saline control, moderately saline NaCl and Na₂SO₄ and highly saline Na₂SO₄ (Figure 2A) enabled capture of the anion‐specific effects (Cl⁻ and ${\mathrm{SO}}_4^{2-}$) and severity effects (moderate vs. high Na₂SO₄) on willow (S. miyabeana ‘SX64’).

Photosynthetic capacity and biomass productivity were maintained compared to non‐saline controls, even at high soil salinity levels of 9 dS m⁻¹ (Figure 2C,D), supporting the suitability of willow for phytoremediation. This aligns with research by Hangs et al. (2011), reporting salinity tolerance in 37 willow varieties, where most cultivars could tolerate moderate soil salinity and the most salt‐tolerant cultivars maintained yield under high soil salinity. Leaf moisture content increased under moderate salinity, alongside a >35% reduction of stomatal conductance (Figure 2E). This indicates reduced transpiration accompanied by osmotic adjustment to sustain turgor, potentially mitigating ion toxicity through absorbed salt dilution (Munns, Passioura, et al., 2020; Nguyen et al., ²⁰¹⁷). However, subsequent lower leaf moisture content suggests this water balance might not be maintained as salinity severity increases (Figure 2D).

Salt treatments did not reduce the acquisition of essential nutrients and even increased potassium uptake alongside salt ions, although these ions were selectively partitioned (Figure 3). Sodium accumulation was predominantly restricted to the roots and influx did not increase linearly with external concentration, suggesting a robust sodium exclusion mechanism (Figure 3A). Such exclusion can occur through root cell wall Na⁺ binding (Byrt et al., 2018) or competing uptake of K⁺ (Hauser & Horie, 2010), as proposed in the salt tolerant Salix interior (Major et al., 2017) and Salix eriocephala ‘LAR’ (Huang et al., 2024). In contrast, the high chloride uptake and transport throughout the plant (Figure 3) can be facilitated by passive anion channels (Skerrett & Tyerman, 1994), resulting in leaf chloride concentrations here reaching the upper optimal level for non‐halophytes (~30–560 μmol(Cl) g_DW ⁻¹ [Geilfus, ²⁰¹⁸]). This accumulation, coupled with increased leaf water content, suggests that Cl⁻ may serve as a readily available osmolyte, potentially conferring osmotic benefits similar to those observed in Nicotiana tabacum (Franco‐Navarro et al., 2016). However, given that cytosolic chloride toxicity is thought to occur at concentrations of 5–20 mmol(Cl) L(cytosol)⁻¹ (Teakle & Tyerman, 2010), the accumulated Cl⁻ may also require compartmentalisation into vacuoles, apoplasts or older leaves to mitigate cellular damage and optimise metabolic burden. In comparison to chloride, sulphate uptake was more constrained but still resulted in increased transport throughout the plant, where it was either stored as ${\mathrm{SO}}_4^{2-}$ or metabolised into organic sulphur compounds.

The maintenance of biomass yields despite reduced stomatal conductance, which can lead to decreased carbon assimilation (Munns, Day, et al., 2020), suggests effective metabolic plasticity and ion management strategies, making this experimental system suitable for assessing willow metabolomic tolerance mechanisms without the confounding metabolic effects associated with severe salt‐induced physiological damage.

High organ‐specific diversity in the Salix metabolome

Untargeted metabolomic analysis of control trees revealed substantial diversity across the three main vegetative organs (roots, stems and leaves), with over 5000 compounds detected, approximately 62% of which could be annotated at confidence levels 1–3 (Sumner et al., 2007), providing a detailed overview of the willow metabolome at a whole plant scale (Figure 4). Previous metabolomic studies in single organs of Salix have captured up to 1639 metabolites in bark (Zhou, Guo, et al., 2022), 2735 (Kaling et al., 2015), 600 (Jia et al., 2020) and 2,440 (Aliferis et al., 2015) metabolites in leaves and 800 metabolites in roots (Xia et al., 2021). By integrating analyses of multiple plant organs, a group of common metabolites across organs was identified, representing 26% of metabolomic diversity and over 70% of the putative abundance (estimated from peak area, while acknowledging potential influences of ionisation efficiency differences across compounds, matrix effects and the non‐linearity of response curves) (Figure 4D). Of these, 287 metabolites had stable abundance across organs and represent their potential plant‐wide basal, or ‘core’ metabolism (Figure 4F); a recent metabolomic concept that can also denote evolutionarily conserved metabolites across species (Drapal et al., 2022; Dussarrat et al., ²⁰²²; McLaughlin et al., ²⁰²³). Many of these core metabolites were associated with central pathways like glycolysis, tricarboxylic acid cycle (TCA) cycle derivatives and amino acid metabolism, reflecting primary metabolism across plant tissues. Others included bioactive phenolics found at high levels in Salicaceae, such as characteristic salicinoids (e.g. salicin, salicortin) associated with antioxidant, antimicrobial and anti‐herbivory functions in willow (Boeckler et al., 2011; Brereton et al., ²⁰¹⁷; Meier et al., ¹⁹⁸⁸; Pobłocka‐Olech et al., ²⁰⁰⁷; Tyśkiewicz et al., ²⁰¹⁹; Volf et al., ²⁰¹⁵).

Beyond these common metabolites, willow organs had clear metabolic differences (Figure 4G). Leaves contained the most diverse metabolome, aligning with leaf function as a source of complex photosynthates. Similar trends have been evidenced in other species, including Brachypodium distachyon and Nicotiana attenuata, where untargeted metabolomics studies revealed extensive metabolic complexity in leaves compared to other vegetative organs (Li et al., 2016; Mahood et al., ²⁰²³). Among the generally high diversity of leaf metabolome, the prominence of terpenoids may be linked to their known function in light harvesting, development, defence and signalling (Pichersky & Raguso, 2018). Stems accumulated more lignans compared to the other organs, reflecting woody tissue functions of structure and storage (Köhler et al., 2020). Additionally, the metabolome of stems had a large overlap with both leaves and roots, underscoring their function as conduits for biochemical transport and dialog between aerial and belowground processes (Figure 4D). Comparatively, despite critical functions of water and nutrient absorption, exudation, anchoring structure and storage, roots had the lowest metabolic diversity (Figure 4A,B). Roots were enriched in saccharides and flavonoids, thought to mediate soil interactions like microbial symbiosis (Endo et al., 2021), metal tolerance (Osawa et al., 2011), nutrient uptake (Cesco et al., 2012) and root architecture (Buer et al., 2010). The root metabolome was also distinctly richer in polar metabolites (Figure S6), which can more readily diffuse through the soil environment within the water sector and could therefore play a role as exudates involved in shaping the rhizospheric biotic (Sasse et al., 2018) and abiotic (Badri & Vivanco, 2009) environment. These distinct metabolomes reflect the different physiological roles of each organ, describing different metabolic activities and degrees of functional specialisation (Li et al., 2016) as well as serving as a baseline to enable understanding of key responses to stimuli such as salt stress.

Sulphate salt induces dose‐dependent metabolomic responses in Salix

Sulphate salinity induced extensive organ‐specific metabolic reprogramming (Figure 5A–D,H), affecting 10 and 16% of all metabolites at moderate and high concentrations, respectively. Doubling the Na₂SO₄ concentration in soil resulted in a twofold increase in metabolic changes in roots, but more limited responses in stems and leaves. This localisation of metabolic response to the roots, which serve as the primary defence barrier against excess ions and osmotic stress, is likely decisive in mitigating salinity stress impact aboveground. Notably, consistent levels of salt ion uptake and accumulation despite increased soil salinity indicate effective ion management (Figure 3), allowing leaves to maintain similar adaptive metabolomic strategies under both sulphate salt treatments (Figure 5F,G). Previous studies reported sulphate salt tolerance in willows (Huang et al., 2020) associated with ion imbalance restriction to roots, similarly preventing physiological impact on leaves at moderate salinity levels.

In roots, Na₂SO₄ dose differentially affected shikimates/phenylpropanoids, with a reduction of flavonoids, phenolic acids and lignans under moderate levels, contrasting their net increase under higher sulphate salinity (Figure 5H). This may indicate a developmental slowdown under mild stress, superseded by a required metabolic burst under more severe salinity levels. The lower osmotic potential associated with higher salinity likely drove the accumulation of lignans (Figure 5H) to enhance structural protection (as lignin precursors) and cavitation protection (Reyt et al., 2021), while the accumulation of antioxidative flavonoids can enhance secondary oxidative stress tolerance (Nakabayashi et al., 2014). Under high Na₂SO₄, the sugar donor UDP‐d‐galactose significantly increased alongside phenolics, indicating a disruption in glycosylation and flavonoid bioactivity regulation (Xiao et al., 2014), while increases in iridoid monoterpenoids, such as lamiide, can prevent lipid peroxidation to protect cell membrane integrity (Delaporte et al., 2002). The high diversity of salt‐responsive metabolites captured here implies additional functions beyond ROS scavenging, likely contributing to the observed Na₂SO₄ stress tolerance. Moreover, the reduction in sphingolipids, which regulate ions through early sensing and signalling, could limit salt uptake below toxic thresholds (Jiang et al., 2019; Liu et al., ²⁰²²).

Primary metabolites declined in stems under both sulphur salt concentrations, including reductions in small peptides, saccharides and nucleosides (Figure 5H). A slowdown of developmental processes and primary metabolism would be expected under salt stress, although there was no accompanying (short‐term) impairment to total biomass production (Figure 2C). The comparatively limited disruption of secondary metabolites in stems indicates maintenance of homeostasis and primary resource allocation to support root functions, effectively restricting salt stress propagation.

In leaves, the salt‐responsive metabolites were broadly shared between both Na₂SO₄ levels, characterised by dose‐dependent accumulation, including saccharides, small peptides and precursors of phenolic compounds (Figure 5F,G). Again, this illustrates the buffering role of roots in protecting essential carbon fixation functions and energy‐related compound production, even under severe salt exposure.

An increase of sulphur‐containing compound synthesis under ${\mathrm{SO}}_4^{2-}$ treatment and upregulation of sulphur assimilation and transport pathways is a characterised plant adaptation mechanism to effectively utilise excess ${\mathrm{SO}}_4^{2-}$ (Buchner et al., 2004; Davidian & Kopriva, ²⁰¹⁰). Here, a substantial enrichment of small peptides occurred in roots under sulphur salinity (Figure 5H), with around half incorporating sulphur. This accumulation of sulphur‐containing compounds was consistent throughout the plant at both concentrations (Figure 5I) and was predominantly cysteine‐and glutathione‐based small peptides and some fatty acyl thioesters and alkaloids. The integration of sulphate ions into organic sulphur‐containing compounds has the potential to reduce ion imbalance in planta during salt stress and could represent a tolerance mechanism in willow. Moreover, the increase of structurally related compounds here (Figure 5I), although their functional information is limited, suggests activation of a bespoke, yet‐to‐be‐described, metabolic response to Na₂SO₄. Watanabe & Hoefgen (2019) highlighted a need for integrated sulphur‐omics to overcome this shortfall in functional understanding of sulphur‐containing molecules in plants, of particular relevance to salinity‐tolerant non‐model organisms such as willow.

Salt anions shape Salix metabolomic tolerance responses

The salt response across all treatments involved an extensive proportion of the metabolome (28% of metabolites), reflecting the highly plastic nature of plant metabolism under environmental challenges. While activation of the calcium‐sensing signalling pathway by Na⁺ (Munns & Tester, 2008) and disruption of the Na⁺/K⁺ ratio (Zhang et al., 2018) are commonly considered to trigger salt responses in plants, anion‐driven metabolomic responses have been less well explored. Comparison of metabolomic responses to moderate NaCl and Na₂SO₄ treatments here, at equivalent soil electroconductivity and sodium concentrations, revealed only 3% of metabolites were consistently responsive, highlighting extensive anion‐specific changes (Figure 7).

The common or generalised, salt metabolite changes were asymmetrically distributed across organs, with ~15‐fold more generalised salt responses in roots than in leaves (Figure 7B). These represent common metabolic strategies to tolerate (anion agnostic) osmotic stress in roots at the root–soil interface and restrict sodium translocation, while leaves might be more susceptible to the (anion dependent) ion toxicity of translocated Cl⁻ and ${\mathrm{SO}}_4^{2-}$. The generalised salt response included common accumulation of the phenylpropanoid precursor 2‐phenylacetamide in leaves (a salt biomarker in the euhalophyte Suaeda salsa [Bao et al., ²⁰²³]), osmoprotective molecules in stems such as proline and a highly antioxidative coumaric acid conjugate of salicin, grandidentatin (Si et al., 2009), as well as potential cell wall structural components in roots, such as the monolignol precursors syringin.

NaCl treatment, unlike Na₂SO₄, induced a comparatively small metabolomic response in leaves (Figure 6A–C). Of the limited responses, a significant accumulation of spermidine, a polyamine known for its acid‐ and ROS‐neutralising properties, could help protect cell membranes, reduce chlorophyll loss and leaf senescence (Gill & Tuteja, ^2010b). The auxin conjugate, (indol‐3‐ylacetyl)‐myo‐inositol l‐arabinoside, also increased, suggesting oxidative stress‐specific developmental regulation (Bajguz & Piotrowska, 2009; Ludwig‐Müller, ²⁰¹¹). A larger change under NaCl salt exposure was the reconfiguration of carbon resources through the shikimate/phenylpropanoids pathway (Figure 6D,H). Specifically, lignans and phenylpropanoids had organ‐specific responses: substantial accumulation in roots and significant depletion in stems (Figure 6H). As major lignin precursors, these lignan and phenylpropanoid dynamics could represent the expected structural adaptations within the roots and stems. In particular, lignin deposition can act as a barrier regulating ion uptake and translocation to aerial parts (Byrt et al., 2018; Neves et al., ²⁰¹⁰; Reyt et al., ²⁰²¹).

Flavonoids, whose antioxidative capacity can facilitate localised neutralisation of ROS (Nakabayashi et al., 2014), also had marked organ‐specific responses, with substantial accumulation in roots and variable regulation in the stems (Figure 6H). As flavonoids, phenolic acids, lignans and phenylpropanoids all derive from the same substrates (phenylalanine and the pivotal intermediate cinnamic acid), the organ‐specific shift suggests a coordinated metabolic response promoting defence in the directly exposed roots while simultaneously restricting the availability and energy expense of these compounds in the stems. By comparison, moderate Na₂SO₄ did not induce similar extensive accumulation of phenolic compounds in roots or stems, but this response was triggered under higher Na₂SO₄ concentration.

Accumulation of osmoprotectants under salt stress is one of the most prevalent and well‐described adaptive strategies to reduce cell osmotic potential and stabilise cellular components (D'Amelia et al., ²⁰¹⁸; Yang & Guo, ²⁰¹⁸). Surprisingly, except for stem proline enrichment under both NaCl and Na₂SO₄, the levels of expected organic osmoprotectants, namely hydrophilic and low molecular weight sugars, polyamines or amino acids, were largely reduced in roots and stems under NaCl while accumulating in leaves under Na₂SO₄, more distal to the primary site of osmotic stress (Figures 5H, 6F,G and 7C). This response contrasts with salt‐tolerant poplars which can accumulate soluble sugars as a tolerance mechanism (Jouve et al., 2004; Watanabe et al., ²⁰⁰⁰) and might be a distinct strategy used by S. miyabeana to mitigate osmotic stress through the uptake of inorganic ions, including root‐restricted Na⁺ and anions (Cl⁻ and ${\mathrm{SO}}_4^{2-}$) translocation to leaves. Similarly, Ottow et al. (2005) reported preferential sodium uptake by Populus euphratica, where Na⁺ served as a readily available osmoprotectant stored in the apoplast or bound in the cell wall, alongside reduced calcium and soluble carbohydrate levels.

The anion‐dependent regulation pattern of osmoprotectants also reflected broader differing metabolic adjustments within central metabolism. Although the reduction of stomatal conductance was similar between both salts and photosystem II efficiency (Φ_PSII) preserved (Figure 2), the impact on foliar glucose and energy‐related compounds was distinct. NaCl induced the plant‐wide reduction of glucose and multiple TCA cycle intermediates (Figure S8), representing impaired photosynthetic carbon fixation and decreased flux through glycolysis and the TCA cycle, limiting mitochondrial respiration and overall energy supply. A similar reconfiguration of resources and metabolic processes has been observed under salt stress in barley roots (Wu et al., 2013), where glycolysis was inhibited, and in Arabidopsis thaliana (Hartmann et al., 2015), where a shift to non‐oxidative fermentative energy metabolism can occur. This was coupled with reduced levels of small peptides, nucleosides and other simple carbohydrates like fructose and sucrose in stems and roots (Figures 6F,G and 7C), possibly driven by increased carbon demand for secondary metabolism involved in salt stress tolerance. By comparison, Na₂SO₄ induced an increase in glucose in leaves (Figure S8), alongside an accumulation of saccharides, small peptides and nucleosides, consistent even under high Na₂SO₄. This suggests that willow can sustain chemical energy production in leaves in the presence of sulphate, unlike with chloride and plant‐wide production of salt‐tolerance supporting metabolites, underscoring the importance of understanding anion toxicity in salt tolerance and implicating organic compound sulphate incorporation in the prevention of ionic imbalance. As an example, the accumulation of sulphur‐containing metabolites (Figure 5I,J), including 2,3‐di‐O‐sulfopropylglucose, aligns with the known mechanism of sugar sulphonation that promotes salt tolerance through ion regulation, osmotic tolerance and cell wall integrity in algae (Aquino et al., 2011).

Both chloride and sulphate salts activated the terpenoid pathway (Figure S9), known to mitigate salt stress in mangroves (Basyuni et al., 2009). NaCl induced significant accumulation of terpenoids from the chloroplastic methylerythritol phosphate (MEP) pathway across the whole plant, including diterpenoids as well as iridoid and secoiridoid monoterpenoids, reflecting transcriptomic changes observed in salt‐tolerant poplar (Chen et al., 2018). Na₂SO₄ treatment similarly enriched terpenoids through the MEP pathway but also activated the cytoplasmic mevalonate pathway in roots, also linked to salt tolerance in poplar (Wei et al., 2020). This sulphate‐salt‐associated activation of the mevalonate pathway increased with higher Na₂SO₄ levels, including enrichment of the hormone precursor abscisic aldehyde in roots and leaves. This suggests an adaptive stress response where willows are positioned to quickly synthesise ABA, signalling stomatal closure, root architecture changes and osmoprotectant accumulation (Ryu & Cho, 2015), with ABA‐aldehyde oxidase mediated regulation. In contrast, ABA precursor enrichment was not observed with NaCl, indicating chloride ion interference with signalling that may leave willows more vulnerable to salt stress.

Overall, willow responded to salt stress by modulating both core and organ‐specialised metabolites (Figures 4 and 7D). Within each organ, there was a tendency for greater depletion of organ‐specialised compounds over core metabolites, indicating higher resilience of core metabolism under salt stress (Figure 7D). Contrasting this, Na₂SO₄ treatment resulted in the enrichment of core metabolites in leaves, while NaCl enriched organ‐specialised compounds. This anion‐specific difference further highlights the nuanced metabolic mechanisms deployed to face distinct salt stress conditions.

CONCLUSION

The willow cultivar S. miyabeana ‘SX64’ demonstrated salinity tolerance up to 9 dS m⁻¹, maintaining biomass productivity through effective ion management strategies, including root‐based sodium exclusion, chloride accumulation potentially serving as an osmolyte and sustained potassium uptake, while exhibiting physiological adaptations such as reduced stomatal conductance to reduce osmotic stress. Untargeted metabolomics uncovered an extremely high diversity of metabolites in willow, the majority of which represent organ‐specialised metabolites aligned with distinct physiological functions, but which also include over 250 putative plant‐wide stable, or core, metabolites. The leaves had the highest metabolite diversity, but when subjected to different saline soil conditions, roots, as the frontline of exposure, had the most extensive biochemical response, mitigating ion toxicity and osmotic pressure changes to enable plant tolerance. Accompanying this was a plant‐wide reallocation of energy and metabolic resources, coupled with prioritisation of antioxidant flavonoids, able to protect vulnerable leaf photosynthetic functions, constituting the signature of a general willow salt response. However, the largely specialised nature of this response revealed that in planta, salt anion type becomes an increasingly stronger discriminating force of metabolomic profile when moving from roots to leaves. Indeed, the metabolic adjustment differed extensively in leaves, where Na₂SO₄ induces key components of the core metabolome to mitigate complex osmotic and toxic stress even as soil salinity increases, whereas NaCl induces more organ‐specialised metabolic tools. Salinity levels in agriculture are currently categorised by soil EC, however, given the substantial anion‐dependent differences observed in plant metabolic responses, soil EC measurement might be insufficient to reflect the diversity of soil salinity types, their actual osmotic potential, as well as the anion‐specific plant metabolic toolkit necessary for plant salt tolerance.

EXPERIMENTAL PROCEDURES

Experimental design, plant growth, sample collection and physiological traits

Salix miyabeana ‘SX64’, a willow cultivar selected for its high biomass yields on a variety of sites, including saline soils, and with agronomic traits suitable for large‐scale plantation (Hénault‐Ethier et al., 2017; Mirck & Volk, ²⁰¹⁰), was grown from cuttings in a polyethylene grow tunnel within 15 L pots filled with all‐purpose growing soil (Promix™), sand and vermiculite (1:1:1 volume‐based). After 56 days of growth for root establishment, salt treatments were applied every 3 days during watering over 28 days to progressively build up soil salinity and to avoid salt shock response (Shavrukov, 2013). The four treatments were: tap water control (EC_water = 0.3 mS cm⁻¹), 51 mM NaCl (EC_solution = 6.5 dS m⁻¹) and 26 mM Na₂SO₄ (EC_solution = 6.7 dS m⁻¹), both considered moderately saline, and 51 mM Na₂SO₄ (EC_solution = 13.4 dS m⁻¹), considered highly saline (Mirck & Zalesny, 2015). The treatments were arranged in a randomised block design with seven replicate plants (seven independent pots, n = 7) per treatment (Figure 1). Leaves, stems and roots were sampled after 28 days of treatment (84 days of growth). A representative subsample of each fraction was immediately flash‐frozen in liquid nitrogen and stored at −70°C for metabolomic analysis.

Shoot heights and numbers were measured weekly. Fresh weight (FW) at harvest, oven‐dried biomass weight (DW) and moisture content were measured for each organ. The LICOR‐600 porometer/fluorometer was used to measure photosynthetic parameters, including stomatal conductance and the quantum efficiency in light of photosystem II Φ_PSII (Genty et al., 1989). Soil salinity was inferred at harvest from the electroconductivity of the supernatant solution from a soil: water mixture (1:1) and converted to the electroconductivity of saturated soil extract (EC_e) using the correlation EC_e = 2.23 EC_(1:1) − 0.58 (R ² = 0.98) (Sonmez et al., 2008).

Elemental analysis of roots, stems and leaves was performed on ground dry biomass powder (<500 μm) by Natural Resources Analytical Laboratory at University of Alberta (AB, Canada). Briefly, 0.1 ± 0.02 g of dry powder was digested overnight, with 5 mL of Trace Metal Grade HNO₃ using the Microwave Accelerated Reaction System (method EPA 3051a [modified]), diluted to 25 mL with milliQ water and analysed for mineral nutrients (Al, B, Ca, Cu, Fe, K, Mg, Mn, Mo, Na, Ni, P, S, Zn) using a Thermo iCAP6300 Duo inductively coupled plasma‐optical emission spectrometer (ICP‐OES). Additionally, the anions Cl⁻ and ${\mathrm{SO}}_4^{2-}$ were determined by extracting 0.2 ± 0.01 g of dry powder with 50 mL of 2% acetic acid using a reciprocal shaker for 30 min, followed by a 0.45 μm PTFE syringe filtration. Chloride and sulphate were measured using the ferrithiocyanate colorimetric method (EPA Method 325.2) and the barium chloride turbidimetric method (EPA Method 375.4), respectively.

Extraction of metabolites

Metabolites from willow roots, stems and leaves were extracted according to de Vos et al. (2007) with some modifications. Briefly, plant tissues were freeze‐dried for 72 h and homogenised using a precooled bead mill Tissuelyser II (Qiagen) at 30 Hz for 2–4 cycles of 60 sec each. From the dry tissue powder, 50 ± 0.5 mg was suspended in 1 mL of methanol:water (75:25%v/v), also containing the internal quantitative standard resorcinol at 5 mmol L⁻¹. Samples were vortexed for 10 sec, sonicated in a 135 W/42 Hz ultrasound bath for 10 min at room temperature and centrifuged at 11 000× g for 2 min at 20°C. The supernatant was filtered through a 0.2 μm centrifuge filter, diluted by 4 with methanol 75% and placed in the liquid chromatography–tandem mass spectrometry (LC–MS/MS) autosampler conditioned at 4°C.

Liquid chromatography–tandem mass spectrometry (LC–MS/MS) acquisition

Chromatographic separation was performed using an Agilent 1260 Infinity system, where 10 μL of extract was injected onto a Zorbax Eclipse Plus C18 column (4.6 × 100 mm, 3.5 μm) at 30°C. The mobile phase was solvent A (water, 5% methanol, 0.1% formic acid) and solvent B (methanol, 0.1% formic acid) with a flow rate of 0.4 mL min⁻¹ and an 80‐min elution gradient: 100% A hold for 20 min, then a linear increase from 0 to 100% B over 50 min and 100% B hold for 10 min. This gradient allowed the detection of highly water‐soluble (polar) compounds within the first isocratic phase. Untargeted full‐scan (100–1300 Da) MS acquisition was performed on extracts from all three organs (roots, stems and leaves) of each of the seven replicates (84 samples total), blanks and quality control samples using an Agilent Q‐TOF 6530B mass spectrometer equipped with an Agilent Jet Stream ion source operating in positive ion mode (ESI [+]). Gas temperature was 300°C, drying gas flow was 5 L min⁻¹ and the nebuliser pressure was 45 psig. Sheath gas temperature was 250°C with a gas flow of 11 L min⁻¹. Data‐dependent MS/MS acquisition was performed on three biological organ replicates per treatment and acquired at a collision energy of 20 and 35 eV. Precursor isolation used an intensity threshold set to 500, with a maximum of 4 precursors per cycle and 3 spectra per precursor released after 0.3 min.

Untargeted metabolite data processing

Raw LC–MS data were converted to mzML files by using msConvert (Chambers et al., 2012) and processed using MZmine 3.3.0 (Schmid et al., 2023) (Figure 1). Background noise cut‐off was 1000 for MS spectra and 0 for MS/MS spectra, with a minimum 5000 peak intensity threshold for feature detection. Chromatogram deconvolution used the local minimum search algorithm, isotopes were grouped and feature alignment across samples was performed with 20 ppm mass tolerance (80% weighting) and 1 min retention time tolerance (20% weighting). Feature quality control filtering applied isotope patterns (minimum two), minimal occurrence of four across all samples and a maximal blank: sample average ratio (1:3). Features eluting within <0.2 min retention time and with peak shape correlated by >85% (MZmine metaCorrelate algorithm) were assigned to a feature correlation group and used for curation of adducts (including >10 potential adduct forms), complex formation and in‐source fragments. Despite this extensive quality filtration (Figure S10), some remaining features may still include in‐source fragments or adducts.

Metabolites annotation and molecular networking

Metabolic feature annotation was conducted using a bespoke pipeline that combined multi‐database queries, advanced cheminformatic tools and in‐depth curation of differentially abundant (DA) features (Figure 1). In the first unsupervised annotation step (Figure S11), compound identification was performed using in silico prediction with SIRIUS 5.7.1, combining raw formula and structural elucidation with CSI:FingerID (Dührkop et al., 2019), as well as compound class prediction with CANOPUS (Dührkop et al., 2021). Detailed performance assessment of this tool can be found in previous studies (Dührkop et al., 2021; Mahood et al., ²⁰²³). Here, annotation candidates were retained only when the confidence score or posterior probability was >0.5. In parallel, all features were queried against a set of MS and MS/MS databases, including all databases available in the GNPS environment (e.g., GNPS, MassBank, MoNA and ReSpect) (Wang et al., 2016), experiment‐ and organism‐specific databases from the PlantCyc repository (Hawkins et al., 2021) and from an in‐house library specifically compiled for Salicaceae based on literature and authentic standards. Additional searches were conducted using non‐specific databases accessed via the CEU Mass Mediator platform (Gil‐de‐la‐Fuente et al., 2019). For each annotation tool, only the top‐ranked match per feature was retained, requiring high mass accuracy (mass deviation ≤10 ppm from the theoretical value) and spectral similarity (cosine score ≥0.6) between experimental MS/MS fragmentation patterns and reference library spectra.

For features with multiple annotation candidates, a cross‐tool comparison was performed (Annotation integration step), standardising all outputs to determine the most reliable identification (Figure S12). This cross‐tool reinforcement step prioritised annotation agreements between tools and resolved conflicts by leveraging key parameters, including probability scores (when available), mass deviation, adduct prediction and molecular network agreement. This resulted in a consolidated list of annotated features (Table S3), hereafter tentatively referred to as compounds.

In the second step, the annotation of DA compounds was refined using both data‐driven and knowledge‐based curation. Data‐driven approach leveraged feature‐based ion identity molecular networking (Chen et al., 2021; Zhou, Luo, et al., ²⁰²²), based on Spec2Vec and cosine spectral similarity metrics (Huber et al., 2021) (at 0.6 threshold), to manually resolve annotation ambiguities. Knowledge‐driven curation took advantage of published compound lists and spectra from comparable experiments and plant species not deposited in public databases to improve annotation.

To reflect annotation reliability, each putatively identified compound was assigned a confidence level (1–5) (Schymanski et al., 2014), as reported in Table S3, following best reporting practices (Alseekh et al., 2021). A total of 326 compounds were annotated at level 2, mainly supported by high‐quality MS/MS spectral library matches (similarity ≥0.75); 937 compounds were assigned at level 3, based on MS/MS information (fragmentation patterns, molecular network) and supported by Salicaceae or plant‐specific library evidence; 2554 compounds were at confidence level 4, relying primarily on MS data and Salicaceae or plant‐specific libraries; and 1395 remained at level 5 (unknown) due to insufficient information for formula or chemical class assignment (Figure S12D). Chemical taxonomy was assigned to metabolites based on the Natural Product Classification system (Kim et al., 2021) using the automated NP‐Classifier batch workflow on the Center for Computational Mass Spectrometry platform. All putative compounds were retained for statistical accuracy. All annotations should be considered putative due to potential limitations of database content and cheminformatic predictions, and low confidence (level >3) annotation should be interpreted with caution. Molecular networks were generated in the GNPS analysis environment (Nothias et al., 2020), and visualisation used Cytoscape software (Shannon et al., 2003).

Statistical analysis

Physiological and metabolomic data were processed using R (R Core Team, ²⁰²²) and statistical significance was set at a Benjamin–Hochberg (BH) false discovery rate (FDR) adjusted P‐value threshold of <0.05. Physiological comparisons between treatments used a one‐way ANOVA, followed by a Tukey HSD post hoc test (n = 7, except n = 3 for mineral element analysis). Metabolite peak areas were log‐transformed and normalised using EigenMS (Karpievitch et al., 2014) prior to statistical analysis (Figure 1). To identify differentially abundant (DA) metabolites, pairwise comparisons of peak area for each treatment (n = 7) were performed against controls using either a t‐test or Wilcoxon rank sum test with FDR (BH) control, depending on whether assumptions of normality and equal variance were met, or through detection as structural zeros (He et al., 2014) (metabolites detected ≥4 replicates of one condition, and not detected in any replicates of the compared condition). For metabolite comparison across organs, a Kruskal–Wallis test was first performed before post hoc pairwise comparisons between each organ pair (Wilcoxon rank sum tests with FDR control). The core metabolome was tentatively defined as compounds that did not significantly vary across control organs and met a metabolic stability coefficient of variation of peak area below 30%. Principal coordinate analysis (PCoA) was performed using the MixOmics package (Rohart et al., 2017) and PERMANOVA on Bray–Curtis distances was used to test significant differences between groups.

AUTHOR CONTRIBUTIONS

ES, NJBB, ML and FEP conceived and designed the study. ES and AF designed the metabolomics workflow. ES and MS conducted the mass spectrometry. ES, AF, EG and NJBB designed the statistical analyses. ES analysed the data and drafted the manuscript. All authors reviewed and commented on the results and manuscript.

CONFLICT OF INTEREST

The authors declare that there is no conflict of interest.

Supporting information

Figure S1. Experimental setup for salt stress assessment of Salix miyabeana ‘SX64’.

Figure S2. Willow organs biomass (dry weight) under control and saline conditions.

Figure S3. Growth rate based on cumulative stem height per day.

Figure S4. Mineral elements ratios.

Figure S5. Organic and inorganic sulphur.

Figure S6. Histogram of organ‐specialised metabolites retention time.

Figure S7. Proportional change within each pathway.

Figure S8. Regulation of glycolysis and TCA cycle pathways.

Figure S9. Regulation of terpenoid pathway under salt stress.

Figure S10. Feature count throughout data processing.

Figure S11. Unsupervised annotation details.

Figure S12. Annotation integration details.

Table S1. Soil electroconductivity.

Table S2. Physiological data.

Table S3. Annotation details.

Table S4. Organ partitioning of the Salix metabolome.

Table S5. Salix metabolomic response to moderate Na₂SO₄.

Table S6. Salix metabolomic response to high Na₂SO₄.

Table S7. Salix metabolomic response to moderate NaCl.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available in the supporting information of this article and on the GNPS‐MassIVE online dataset repository: ftp://massive.ucsd.edu/v08/MSV000095768/.