Seedlings acquired from Minqin Desert Experimental Station were cultivated in greenhouse at Yuzhong Experimental Station, Lanzhou University, China (35°56′N, 104°09′E, 1750 m a.s.l.), starting from December 2017 (except Ammopiptanthus mongolicus cultivated from May 2018). Three PFTs (6 woody species, N-fixing legumes: Ammopiptanthus mongolicus, Caragana korshinskii; N-fixing non-legumes: Elaeagnus angustifolia (Bertrand and Lalonde, 1985), Hippophae rhamnoides (Korekar et al., 2012); non-N-fixing plants: Nitraria tangutorum, Haloxylon ammodendron) comprising of first-year seedlings were selected as target plant materials. Each seedling was grown in a plastic pot filled with sand separately without competition.

To clarify the relationship between plant drought stress and varying levels of soil water content, we have referenced the work of Li et al. (2019a, 2019b, 2024) across several studies to provide context for our experimental design. Specifically, Li et al. (2024) analyzed soil samples from 30 locations along a 4500 km east–west transect in the drylands of northern China, varying in aridity index ranged from 0.04 to 0.59, and conduct a laboratory experiment under relative soil water contents of 20%, 40%, and 60%. Complementing this, Li et al. (2019a, 2019b) documented soil water content measurements from the Hexi Corridor and Xinjiang regions within China’s drylands, noting that soil water content varied from 0 to 0.497. Considering these empirical realities and practical constraints of our experimental setup, we opted to conduct our study using the soil water content levels representative of China’s dryland conditions in this study, which are consistent with those described by Li et al. in their researches (Li et al., 2019a, 2019b, 2024). All pots were assigned to one of the four water gradients: 75% soil water holding capacity (SWHC, which absolute water content is 16.7%, marked as control treatment), 50% SWHC (moderate treatment), 25% SWHC (severe treatment) and 5% SWHC (extreme treatment) with 9–15 replicates (variations of replicates were because the death of the specific seedlings) of each water gradient. The drought treatment lasted for at least 3 months before initiating ¹⁵N pulse labelling.

¹⁵N labeling procedure took place 3–4 hours before separating plants into aboveground and belowground parts (labeling hours were recorded for every plant; Zhang et al., 2020). A stick was used to prick four holes about 1cm away from the stem of a plant and 15cm deep evenly before 1.5mg ¹⁵N dissolved in 8ml solution via a syringe injected into the soil. Pots from the same water gradient, one quarter of randomly selected plants were labeled with ¹⁵N-KNO₃, one quarter with ¹⁵N-(NH₄)₂SO₄, one quarter with ¹³C-¹⁵N-Glycine (all three were 99.99 atom% ¹⁵N enrichment purchased from Shanghai Chem-Industry Institute), the remaining plants were served as the control treatment and were added with a mixture of ¹⁴N-KNO₃, ¹⁴N-(NH₄)₂SO₄ and ¹²C-¹⁴N-Glycine. For example, when 1.5mg ¹⁵N of KNO₃ was injected, 1.5mg ¹⁴N of (NH₄)₂SO₄ and 1.5mg ¹⁴N of Glycine would be injected at the same time for a plant.

¹⁵N recovery rate of each N form (NO₃ ^-, NH₄ ⁺, Glycine) from the labeled ¹⁵N pool (nmol ¹⁵N uptake per gram plant per hour) is calculated as follows:

Where NUR (nmol g^-1 h^-1) is the plant ¹⁵N uptake rate, APE (= atom_labelled% − atom_unlabelled%) is the atom% excess (¹⁵N atom% difference in plants between the labelled and the unlabeled), Plant [N] (mg g^-1) is the whole-plant N content, t (h) is labeling time for plant. N uptake capacity (NUC) is plant total ¹⁵N uptake rate, i.e. the sum of plant NUR of NO₃ ^-, NH₄ ⁺, Glycine (Hong et al., 2018).

Whole-plant net photosynthesis and dark respiration were measured using a Li-8100 automated CO₂ flux system (LiCor, Nebraska, USA) in a high-lucid organic glass box (0.4×0.4×0.4m; Wang et al., 2015), with triplicates for each pot. The glass box, equipped with two circulating fans, facilitated CO₂ concentration variation recorded every 3 minutes during each measurement.

Plants were exposed to sodium lamp (Philips SON-TARGO, Amsterdam, Holland), beyond the glass box, providing a quantum flux density (PAR) of 800–1000 μmol m^-2 s^-1 for at least 2h to initiate photophosphorylation. Subsequently, plants were placed inside the glass box for the measurement of photosynthetic rate (P). Whole-plant dark respiration (R_dark) was measured approximately after 30 minutes of dark adaption in the dark chamber. Following these measurements, cut the plants in pots at the base of stems to separate the plants into two parts, i.e., aboveground parts (including leaves and stem, termed shoots hereafter) and belowground parts (i.e., roots). Carefully remove soil from the roots to avoid damage to fine roots and root underwent thorough washing with distilled water three times before conducting respiratory measurements on shoots (R_shoot) and roots (R_root). Afterwards, place the plants in a self-fabricated closed-circuit system which was a cylindrical dark chamber (317.8 cm² in base area and 4823.9 cm³ in volume) connected to a LI-8100 automated CO₂ flux system. Triplicates were recorded for every plant every 3 minutes during the measurement.

Whole-plant ICAE is calculated as follows:

Where ICAE is instantaneous carbon assimilation efficiency, P_net (P_net = P_gross + R, nmol s^-1, where R = R_shoot + R_root) is whole-plant net photosynthetic rate per plant, P_gross (P_gross = R_dark + P_light, nmol s^-1) is whole-plant gross photosynthetic rate per plant. The unit of gas exchange rate is converted by the formula of gas exchange rate (nmol s^-1) = observed gas exchange rate (μmol m^-2 s^-1)×cross sectional area of the chamber (m², i.e. 0.16 for glass box and 0.03178 for cylindrical dark chamber in this study)×10³ (Wang et al., 2015).g_s,A and E were measured at leaf level using a portable open gas-exchange system (LI-6400, LiCor, Lincoln, NE, USA). The gas exchange chamber was supplied with a photosynthetic photon flux density of 1200 μmol m^-2 s^-1 and equipped with an LED light source.

For iWUE at leaf level, it is calculated as

Where iWUE (μmol mol^-1) is intrinsic water use efficiency at leaf level, A (μmol m^-2 s^-1) is photosynthesis rate at leaf level, g_s (mol m^-2 s^-1) is stomatal conductance.

Plant and soil sampling were collected approximately four hours after ¹⁵N labeling. After shoot and root respiration measurement, each plant was segmented into components (leaves, stem and root) and measured the fresh mass, and then dried in an oven at 65°C for 48h (to a constant mass), measured the dry mass of components and ground with a Mixer Mill MM400 (Retsch, Haan, Germany). Plant C, N content and ¹⁵N atom% were measured using an isotope ratio mass spectrometer coupled with an elemental analyzer (Thermo Fisher Scientific, DELTA VIRMS).

All analyses were carried out using R statistical software v. 4.0.5 (R Core Team, 2021). One-way ANOVA with Tukey’s HSD test (HSD.test function in the agricolae package) was performed to detect the differences in plant traits among N-fixing legumes, N-fixing non-legumes and non-N-fixing plants across SWC gradients by evaluating the pairwise differences at p< 0.05 level. Two-way ANOVA was performed (aov function in the stats package) to test the interaction of SWC and PFTs on plant functional traits at p< 0.05 level.

To further illustrate the different responses of plant traits to drought among N-fixing legumes, N-fixing non-legumes and non-N-fixing plants, correlations between pairwise plant trait and SWC were used separately for each PFT. Relationships between selected plant trait and SWC were tested with linear regression (Couso and Fernandez, 2012). Data of plant traits were log₁₀-transformed to normalize statistical distribution. The figures were generated using the ggplot2 package in R.

Drought significantly altered leaf dry biomass (LDM), leaf fresh biomass (LFM), A, g_s and E, with the effects varying among different plant functional types (PFTs) for all leaf traits except g_s, while the interaction of PFTs and drought exerted significant influence on leaf dry-fresh mass ratio (LDFR), leaf nitrogen concentration (Leaf [N]) and leaf carbon to nitrogen ratio (Leaf C:N) ( Table 1 ). Drought significantly altered LDM, LFM, Leaf [N], Leaf C:N and E in two N-fixing plants, but not in non-N-fixing plants, while leaf mass fraction (LMF), A and g_s only changed with drought in N-fixing non-legumes but no significant changes were observed in N-fixing legumes or non-N-fixing plants. LDFR, specific leaf area (SLA) and iWUE showed insignificant changes along SWC gradients among the three PFTs ( Figure 1 ).

Drought significantly altered stem dry biomass (SDM), stem fresh biomass (SFM), stem dry-fresh mass ratio (SDFR) and stem nitrogen concentration (Stem [N]), with specific effects being varied to each PFTs for all stem traits, while the combined influence of PFTs and drought was deemed insignificant on all stem traits ( Table 2 ). Drought induced significant alterations in both SDM and SFM of all three PFTs, while Stem [N] showed changes in two N-fixing plants but not in non-N-fixing plants. SDFR only experienced changes in N-fixing legumes while stem mass fraction (SMF) and stem carbon to nitrogen ratio (Stem C:N) were changed in N-fixing non-legumes responding to drought ( Figure 2 ).

Drought triggered significant alterations in root dry biomass (RDM), root fresh biomass (RFM), root dry-fresh mass ratio (RDFR) and root nitrogen concentration (Root [N]). These effects were unique to each PFT for all root traits, while the combination of PFTs and drought exerted a significant influence on RDM, RFM, Root [N] and root carbon to nitrogen ratio (Root C:N) ( Table 3 ). The repercussion of drought manifested in significant alterations in both RDM and RFM across all three PFTs. RDFR and Root [N] underwent changes in two N-fixing plants but not in non-N-fixing plants, whereas root mass fraction (RMF) and Root C:N were affected solely in N-fixing non-legumes ( Figure 3 ).

Drought induced significant modifications in whole-plant dry biomass (TDM), whole-plant fresh biomass (TFM), whole-plant dry-fresh mass ratio (TDFR), Plant [N], whole-plant mass-specific respiration rate (R_mass), ICAE andNUC, with distinct effects to each PFTs for all whole-plant traits. However, the combined influence of PFTs and drought exerted significant influence on TFM, TDFR, Plant [N], whole-plant carbon to nitrogen ratio (Plant C:N), R_mass and NUC ( Table 4 ). Drought induced alterations were also observed in TDM, TFM, Plant [N], Plant C:N and NUC for all three PFTs. Only TDFR changed in two N-fixing plants but not in non-N-fixing plants, while ICAE only displayed changes in N-fixing non-legumes. R_mass, on the other hand, only changed in non-N-fixing plants. Root to shoot ratio (R/S) showed insignificant changes along SWC gradients for all three PFTs ( Figure 4 ).

In context of leaf traits i.e., LDM and E ( Supplementary Figures 1A, J ), stem traits i.e., SFM and Stem [N] ( Supplementary Figures 2B, E ), root traits i.e., RDM, RFM and RDFR ( Supplementary Figures 3A–C ), and whole-plant traits i.e., TDM, TFM and TDFR ( Supplementary Figures 4A–C ), the represented absolute slopes of trait-SWC linear regression models in N-fixing non-legumes demonstrated higher values than that in N-fixing legumes, with the lowest values observed in non-N-fixing plants. Similarly, R ² between traits and SWC decreased from N-fixing non-legumes, N-fixing legumes to non-N-fixing plants in LDM, LFM, LMF, A, E, SMF, Stem [N], RDM, RFM, RDFR, Root [N], TFM, TDFR, Plant [N] and Plant C:N ( Supplementary Figures 1A, B, D, H, J, 2D, E, 3A–C, E, 4B, C, E, F ).

Absolute slopes of trait-SWC linear regression models in N-fixing plants displayed higher values than non-N-fixing plants in LFM, LMF, Root [N], Plant [N], ICAE and NUC ( Supplementary Figures 1B, D, 3, 4E, H, I ). SLA, g_s, SMF and Stem C:N showed greater values in N-fixing non-legumes than non-N-fixing plants, followed by N-fixing legumes in absolute slopes of trait-SWC linear regression models ( Supplementary Figures 1E, I, 2D, F ). However, LDFR, Leaf [N], SDM, SDFR, RMF and R/S indicated higher values in N-fixing legumes than non-N-fixing plants, followed by N-fixing non-legumes ( Supplementary Figures 1C, F, 2A, C, 3, 4D ). Notably, for Leaf C:N, A, iWUE, Root C:N, Plant C:N, R_mass, the absolute slopes values in non-N-fixing plants were much greater than N-fixing plants ( Supplementary Figures 1G, H, K, 3F, 4F, G ).

Plants dynamically adjust their growth and biomass distribution in response to drought conditions. This investigation revealed consistent decreases in the biomass of leaves, stems, roots and whole plants across N-fixing legumes, N-fixing non-legumes and non-N-fixing plants (except LDM and LFM of non-N-fixing plants; Figures 1A, B , 2A, B , 3A, B , 4A, B ), in agreement with previous studies (Rodiyati et al., 2005; Gebauer et al., 2020). On the one hand, drought limits biomass accumulation with reduction of SWC which leads to lower nutrients availability (Matías et al., 2011; Usmani et al., 2020). Simultaneously, plants adjust biomass distribution to improve their adaptations to drought (Puglielli et al., 2021). Leaf, stem, root and plant biomass in N-fixing plants were higher than non-N-fixing plants indicating that N supply is the main factor contributing the differences under higher soil water content, however steeply decreasing in biomass of N-fixing plants under lower soil water content in this study. The differences in plant functional types responses to drought suggest a shift from N limitation under sufficient water availability to water limitation under extreme drought (Hofer et al., 2017). In this study, drought stress had a notable impact on the LMF, SMF and RMF of N-fixing non-legumes, but had insignificant effect on R/S of three PFTs and biomass distribution (LMF, SMF, RMF) of N-fixing legumes and non-N-fixing plants ( Figures 1 – 4D ). The decline of LMF and the concurrent increase of RMF (particularly evident under extreme treatment compared to control treatment) with drought in N-fixing legumes suggested adaptive adjustments in plant functions and improved ability to obtain limited resources in harsh environment (Freschet et al., 2018; Du et al., 2019). The findings of Tobita et al. (2011) showed R/S of Alnus hirsute (N-fixing non-legumes) responded insignificantly to drought, aligning with the outcomes of this study.

C and N are important nutrients in maintaining plant growth and functionality, with C:N stoichiometry serving as key indicators of plant responses to environmental shifts (Agren, 2008). Results showed that leaf, stem, root and whole-plant nitrogen concentration ([N]) and C:N in N-fixing plants significantly varied (except Stem C:N and Root C:N of N-fixing legumes) with increasing drought severity, while non-N-fixing plants did not changed (except Plant [N] and Plant C:N, Figures 1F, G , 2 – 4E, F ), which illustrated that N concentration and C:N tends to be more stable in non-N-fixing plants compared to their N-fixing counterparts. Leaf [N] and Leaf C:N partly determines plant growth, as changes in protein, chlorophyll and rRNA production influence plant N content because of the cost in plant growth (Agren, 2004; Yan et al., 2019). Drought-induced shifts in plant metabolism may further contribute to changes in N concentration and C:N across different organs and whole plants (Sardans et al., 2012).

Water is a crucial factor in plant growth and closely correlated with photosynthesis and respiration (Huang et al., 2020a, 2020b). The initial response to water deficit is stomatal closure to prevent water loss, leading to a subsequent reduction in transpiration and improvement in water use efficiency. However, this inevitably results in a decrease in photosynthesis due to limited availability of water, CO₂ and water deficiency in leaves (Limousin et al., 2015; Yan et al., 2016). In this study, drought significantly decreased A, g_s and E of N-fixing non-legumes, only E significantly declined in N-fixing legume, while these parameters remained relatively unchanged in non-N-fixing plants ( Figures 1H–J ). It indicated non-N-fixing plants are more tolerant to drought compared to N-fixing plants, ensuring maintenance of leaf photosynthesis during drought (Zhao et al., 2020). However, iWUE varied insignificantly among all the three PFTs ( Figure 1K ), with distinction to Li et al. finding that iWUE was very sensitive to soil moisture variation, suggesting desert plants are most adaptive to arid conditions despite the different PFTs (Li et al., 2022a). This study revealed that drought had no significant effect on the R_mass of N-fixing plants, but led to a rapid increase in non-N-fixing plants ( Figure 4G ). Plant respiration plays a crucial role in producing energy and C-rich precursors for biomass in physiological processes (Wendering and Nikoloski, 2023). Increasing plant respiration induced by drought may reflect needs of ATP, associated with various physiological aspects like plant growth, cell maintenance, and ion transport, which improved plant tolerance to drought consequently (Atkin and Macherel, 2009; Birami et al., 2020). ICAE allows for the standardized comparison of plant C gains between different PFTs under drought conditions (Limousin et al., 2015). This study found a significant alteration in ICAE for N-fixing non-legumes under drought, but insignificant changes in N-fixing legumes and non-N-fixing plants ( Figure 4H ).

Sensitivity of plant traits to drought indicated distinctive responses in leaf, stem, root and whole-plant characteristics among different PFTs (Geng et al., 2012). Drought changes plant morphological and physiological traits, representing a complex of plant adaptation process of different PFTs to drought conditions, often involving a trade-off between trait and fitness (Li et al., 2022c). Some previous studies have shown that increasing drought was more harmful to non-N-fixing plants compared to N-fixing plants (Hofer et al., 2016, 2017). However, our research did not align with this conclusion. Specific significant effects on each PFT for all plant traits (except g_s) were observed, demonstrating that drought led to distinct changes in plant traits among three PFTs in this study. In general, the relationship between trait and SWC progressively weakened from N-fixing non-legumes, N-fixing legumes to non-N-fixing plants ( Supplementary Figures 1–4 ). This implied a more sensitive response to decreasing soil water supply in N-fixing non-legumes and N-fixing legumes than non-N-fixing plants. The possible discrepancy between our findings and previous studies may be attributed to the way in which drought treatments were applied, including sustained exposure to low water availability (e.g. decreased precipitation) that may not lead to the climax of biomass loss and persistent severe drought (our study), and the differences between species (Sui et al., 2011; Hofer et al., 2016, 2017). In this study, sharp decline in SWC may rendered N-fixing plants incapable of nodulating and fixing N, potentially leading to plant mortality (Pawlowski and William, 2008; Cooper and Scherer, 2012). This resulted in a substantial reduction in N availability for N-fixing plants, while N availability of non-N-fixing plants remained unaffected by N-fixing microsymbionts, thus N-fixing plants and non-N-fixing plants responded differently to drought stress.