Suppressing GhCKX optimizes the trade‐off between photosynthesis and water use efficiency in cotton mature leaves under drought

SUMMARY

Studies on drought stress physiology have primarily focused on the newest fully expanded leaves, in which drought reduces cytokinin (CTK) content, thereby disrupting water balance and photosynthesis. Mature (old) leaves usually exhibit greater sensitivity to drought than the newest fully expanded ones. However, the impact of higher endogenous CTK levels on photosynthetic rate (A _N) and water use efficiency (WUE) in drought‐stressed mature leaves remains underexplored. To address this, a drought experiment was performed using wild‐type (WT) cotton and two CYTOKININ DEHYDROGENASE (GhCKX) suppression lines with elevated CTK levels. Here, suppressing GhCKX optimized the trade‐off between A _N and intrinsic WUE (iWUE) in mature leaves during drought, since not only higher A _N but also higher iWUE were observed in transgenic lines than the WT under drought. This optimization was driven by shifts in the relationships of mesophyll conductance (g _m) and maximum carboxylation rate (V _cmax) to stomatal conductance (g _s), specifically higher g _m/g _s and V _cmax/g _s ratios, resulting from larger enhancements in g _m and V _cmax than in g _s, caused by up‐regulated CTK. Moreover, higher g _m in drought‐stressed transgenic lines than drought‐stressed WT resulted from thinner cell wall, shorter distance between adjacent chloroplasts, and larger ratio of chloroplast area to mesophyll area facing the intercellular airspace; and their superior V _cmax was attributed to increased Rubisco content and initial Rubisco activity. This study reveals that promoting endogenous CTK simultaneously enhances A _N and iWUE in mature leaves under drought via the coordination of photosynthetic constraints, offering a sustainable strategy to improve the drought resistance of crops.

Significance Statement

GhCKX‐suppressed cotton lines exhibited higher photosynthesis and water use efficiency in mature leaves than wild‐type plants under drought, which was driven by shifts in the relationships of mesophyll conductance (g _m) and maximum carboxylation rate (V _cmax) to stomatal conductance (g _s), specifically higher g _m/g _s and V _cmax/g _s ratios.

This study reveals that elevating endogenous cytokinin levels by suppressing GhCKX simultaneously enhances photosynthesis and water use efficiency in mature leaves of drought‐stressed cotton, and this is achieved by promoting mesophyll conductance and carboxylation rate to a greater extent than stomatal conductance. This study provides a novel strategy for future breeding of drought‐tolerant crops via the coordination of photosynthetic constraints.

INTRODUCTION

As a vital global fiber crop, cotton (Gossypium hirsutum) is typically cultivated in hot and arid regions. During its growth, particularly in the flowering and boll‐forming stages, drought events frequently occur (Yu et al., 2023; Zhao et al., 2019), seriously impacting cotton growth and development, which in turn reduces yield and quality (Hu et al., 2018). Prior studies have confirmed that the decline in cotton yield and quality is largely attributed to reductions in leaf photosynthesis under drought stress, which consequently result in inadequate supply of photo‐assimilates for boll loading (Chastain et al., 2014; Zhao et al., 2019). Therefore, enhancing photosynthesis is essential for sustaining cotton production in water‐scarce environments. However, in many instances, improvements in leaf photosynthesis often result in decreased water use efficiency (WUE), since the trends of net photosynthetic rate (A _N) and intrinsic water use efficiency (iWUE) are frequently inversely related (Flexas et al., 2016). Consequently, the simultaneous enhancement of A _N and iWUE has become an urgent requirement for achieving drought‐resistant crop production.

Research indicates that 60–87% of the total photo‐assimilate requirements for each developing reproductive unit of cotton are derived from the leaf subtending the cotton boll (LSCB) (Wullschleger & Oosterhuis, 1990). Thus, maintaining a high photosynthetic capacity in LSCBs is crucial for the development of cotton bolls. Due to cotton's indeterminate growth habit, maturity of its reproductive organs and their LSCBs differs at different positions on the cotton plant. However, the majority of LSCBs are fully expanded mature (old) leaves during cotton's flowering and boll‐forming stages. Nevertheless, previous research has mainly concentrated on the newest expanded leaves and the effects of drought stress on their ontogeny (Chastain et al., 2016). It should be noted here that several studies across multiple crops, including cotton, have revealed that mature leaves exhibit greater sensitivity to drought stress than the newest expanded ones (Chastain et al., 2016; Jung, 2004; Meng et al., 2015; Taulavuori et al., 2010). Therefore, investigating the photosynthetic performance and water status of mature cotton LSCBs will be of greater significance than that of the newest expanded ones. Limited reports have documented an increase in iWUE for these mature LSCBs under drought stress; however, a decrease in their photosynthetic capacity has been observed, inhibiting the development of their subtending bolls (Gao et al., 2021; Hu et al., 2024). Simultaneously enhancing the photosynthesis and iWUE of these mature LSCBs under drought conditions presents a considerable challenge.

Photosynthesis is governed by the combination of CO₂ diffusion and CO₂ fixation, with the former being primarily influenced by stomatal (g _s) and mesophyll (g _m) conductance, and the latter being closely linked to photosynthetic biochemistry, particularly with the maximum carboxylation rate (V _cmax) in the Calvin cycle (Flexas et al., 2012; Sakoda et al., 2021). iWUE at leaf level is defined as the ratio of A _N to g _s, and reducing g _s is often viewed as a direct way to enhance iWUE. However, any reduction in g _s inevitably leads to a decrease in A _N as it restricts the entry of substrate, CO₂, into the leaves (Flexas et al., 2016). Therefore, the improvement in iWUE usually involves a trade‐off with A _N. It is not feasible to improve both iWUE and A _N simultaneously by solely regulating g _s (Hu et al., 2025). Flexas et al. (2013, 2016) emphasized that to achieve concurrent improvements in A _N and iWUE, it is crucial to promote g _m and/or V _cmax at a given g _s level. Alternatively, adjustments should be made to the parallel changes in g _m and/or V _cmax with respect to g _s, specifically by increasing the ratio of g _m/g _s and/or V _cmax/g _s (Gago et al., 2014).

As a physical process, the CO₂ diffusion from the atmosphere to carboxylation sites is strongly modulated by leaf structural properties. For instance, g _s primarily depends on stomatal morphological characteristics, for example, stomatal pore area (a _max), size (SS), and density (SD) (Lawson, 2009). g _m is limited by mesophyll anatomical structures such as specific leaf weight (LMA), the thickness of leaf (T _leaf), and mesophyll (T _mes) (Flexas et al., 2012). Furthermore, g _m comprises two conductances: gas phase (g _ias) and liquid phase (g _liq). g _ias is primarily determined by the volume fraction of intercellular airspace (f _ias) (Han et al., 2018; Lu et al., 2016), while g _liq is largely dependent upon the morphology of liquid phase components, such as the thickness of cell wall (T _cw) and chloroplast stroma (T _chl), the spacing between neighboring chloroplasts (D _chl‐chl), and the surface area of chloroplast (S _c/S) and mesophyll (S _m/S) exposed to intercellular airspace per unit leaf area (Gago et al., 2020). The extent to which any of these traits influence g _s and g _m depends critically on the species and environmental context (Hu et al., 2024, 2025; Lei et al., 2022; Xiong & Flexas, 2021; Zou et al., 2022). Regarding V _cmax, it is closely linked to Rubisco properties like Rubisco content, activation and performance (Flexas et al., 2016; Sakoda et al., 2021). Prior research has indicated that water shortage can lead to reduced SD, SS, and/or a _max, resulting in decreases in g _s (Hu et al., 2025; Ouyang et al., 2017). Accordingly, drought has been shown to increase LMA, T _cw, etc. and/or to decrease S _m/S, f _ias, S _c/S, etc. in cotton, ultimately leading to reductions in g _m (Han et al., 2019; Hu et al., 2024; Zou et al., 2022). Bota et al. (2004) concluded that the effects of altered Rubisco activity and content on A _N are strongly species‐dependent. Past studies have shown that drought restricts initial Rubisco activity and activation state in cotton leaves, thereby reducing V _cmax and consequently affecting A _N (Carmo‐Silva et al., 2012; Hu et al., 2023).

Cytokinin (CTK) is crucial in governing plant tolerance to abiotic stress (Merewitz et al., 2011a). In more detail, promoting endogenous CTK levels can promote chlorophyll content, leaf A _N, root and shoot biomass, as well as yield of drought‐stressed crops (Kuppu et al., 2013; Zaheer et al., 2019). Nevertheless, the impacts of endogenous CTK levels on g _s are inconsistent. Some studies reported that enhancing endogenous CTK, by overexpressing the isopentenyl‐transferases (IPT) gene that encodes isopentenyl‐transferase involved in CTK synthesis, significantly promoted the g _s of mature leaves in drought‐stressed corn (Zea mays) (Oneto et al., 2016) and rice (Oryza sativa) (Reguera et al., 2013). In contrast, the same treatment had no effect on the g _s in the newest expanded leaves of tomato (Solanum lycopersicum) (Glanz‐Idan et al., 2022) and in either mature or the newest expanded leaves of Agrostis stolonifera (Merewitz et al., 2011a) under drought. Based on these findings, it has been proposed that the influence of CTK on g _s under water deficit is both age‐dependent and species‐specific (Farber et al., 2016). Regarding g _m‐related parameters, a few studies have observed that enhancing endogenous CTK levels, by overexpressing IPT, not only altered thylakoid arrangement within chloroplasts of tobacco (Nicotiana tabacum) under drought (Rivero et al., 2009), but also up‐regulated beta‐glucosidase activity to remodel cell wall in drought‐exposed leaves of A. stolonifera (Merewitz et al., 2011b). Additionally, down‐regulating endogenous CTK by overexpressing cytokinin dehydrogenase (CKX) gene, encoding CTK dehydrogenase, which can degrade CTK, reshaped the structure of cell wall in barley (Hordeum vulgare) leaves under water deficit (Vojta et al., 2016). These studies suggested that altering endogenous CTK levels may influence g _m by changing certain structural characteristics of mesophyll tissue. Notably, to date, only one study has reported that high endogenous CTK levels, achieved by suppressing CKX, promoted g _m in the newest expanded leaves of drought‐stressed cotton (Hu et al., 2025), while no relevant information regarding the effects of altered CTK levels on g _m in mature crop leaves under drought conditions exists. Regarding V _cmax, Kuppu et al. (2013) found that enhancing endogenous CTK, by overexpressing IPT, did not affect V _cmax in the newest expanded cotton leaves either under full or reduced irrigation regimes. Similar to g _m, the impact of altered endogenous CTK levels on V _cmax of mature crop leaves under drought remains unclear.

Drought is known to reduce endogenous CTK levels and worsen water balance and photosynthesis in plants (Hu et al., 2025; Merewitz et al., 2011b). Previous studies indicated that boosting endogenous CTK via GhCKX suppression could delay the aging and shedding of mature cotton leaves (Xu et al., 2019; Zhao et al., 2015). Whether this elevation improves mature leaves' water use and carbon assimilation under drought is unconfirmed. We hypothesized that increasing endogenous CTK levels can optimize the trade‐off between photosynthesis and WUE, and boost both A _N and iWUE in cotton mature LSCBs under water deficit. To verify this, GhCKX‐suppressed transgenic cotton with elevated CTK levels, along with their wild‐type (WT) controls, were used to (i) investigate the effects of enhanced endogenous CTK on A _N, iWUE, and their trade‐off in drought‐stressed cotton mature LSCBs, and (ii) explore how changes in g _s, g _m, and V _cmax influence the trade‐off between A _N and iWUE, as well as identify key factors that regulate g _s, g _m, and V _cmax in drought‐stressed cotton mature LSCBs.

RESULTS

Leaf water status

Drought conditions (soil relative water content [SRWC] [45 ± 5]%) resulted in a lower pre‐dawn leaf water potential (Ψ _PD) than the well‐watered conditions (SRWC [75 ± 5]%) for the WT and GhCKX‐suppressed lines (CR‐3 and CR‐13) (Figure 1A,B). Nonetheless, transgenic lines exhibited higher Ψ _PD values than the WT under water deficit conditions.

Figure 1.

Plant water status and biomass for cotton lines under well‐watered and drought conditions.

(A) Dynamic changes in soil relative water content (SRWC).

(B) Dynamic changes in pre‐dawn leaf water potential (Ψ _PD).

(C) Images showing the phenotypic performance of different cotton lines under well‐watered and drought conditions, scale bar = 20 cm.

(D) The biomass of the whole plant for cotton lines under well‐watered and drought conditions. WT, wild‐type control plants. CR‐3 and CR‐13, GhCKX suppression lines. Values are means of three (A, B) or four (D) replicates ± standard error (SE). Different capital letters indicate significant differences at the probability level of 0.05 between different water treatments in the same plant line (Student's t‐test). Different lowercase letters indicate significant differences at the probability level of 0.05 between different plant lines within the same water treatment (LSD test).

Plant biomass and boll weight

Under well‐watered conditions, no significant differences were observed in phenotypic performance between WT and transgenic lines (Figure 1C). Drought inhibited the cotton growth, whereas transgenic lines exhibited better growth performance than the WT under water deficit conditions, as supported by their higher biomass (Figure 1D). Moreover, drought significantly decreased the boll weight for all plant lines as compared to well‐watered conditions (Figure S1). However, the boll weight was higher in GhCKX‐suppressed lines than in WT plants under drought conditions.

Leaf photosynthetic properties, iWUE and their relationships

Relative to well‐watered conditions, drought resulted in significant decreases in net photosynthetic rate (A _N), stomatal conductance (g _s), mesophyll conductance [g _m(Harley) and g _m(anatomy), estimated using the variable J method and using anatomical characteristics, respectively], and maximum carboxylation rate (V _cmax) (except for V _cmax in CR‐13) (Figure 2A–D; Figure S2A), but increased leaf WUE including iWUE_(GE) derived from gas exchange, as well as iWUE_(Isotope) derived from isotopic discrimination (Figure 2E,F). Notably, transgenic lines showed larger values of A _N, g _s, g _m(Harley), g _m(anatomy), V _cmax, iWUE_(Isotope), and iWUE_(GE) than WT plants under water deficit conditions.

Figure 2.

Photosynthetic properties and intrinsic water use efficiency in mature LSCBs of cotton lines under well‐watered and drought conditions.

(A) Net photosynthetic rate (A _N).

(B) Stomatal conductance (g _s).

(C) Mesophyll conductance estimated using the variable J method (g _m(Harley)).

(D) Maximum carboxylation rate (V _cmax).

(E) Intrinsic water use efficiency estimated from the gas exchange measurements (iWUE_(GE)).

(F) Intrinsic water use efficiency estimated from the carbon isotope (iWUE_(Isotope)). WT, wild‐type control plants. CR‐3 and CR‐13, GhCKX suppression lines. Values are means ± SE (n = 6). Different capital letters indicate significant differences at the probability level of 0.05 between different water treatments in the same plant line (Student's t‐test). Different lowercase letters indicate significant differences at the probability level of 0.05 between different plant lines within the same water treatment (LSD test).

Tight linear relationships were found between g _m(Harley) and g _m(anatomy) (R ² = 0.55, P < 0.001) (Figure S2B), as well as between iWUE_(Isotope) and iWUE_(GE) (R ² = 0.80, P < 0.001) (Figure S3), indicating that the estimated g _m and iWUE values are reliable. Consequently, g _m(Harley) and iWUE_(Isotope) were used for further analysis.

The A _N correlated positively with g _s (R ² = 0.66, P < 0.001), g _m(Harley) (R ² = 0.75, P < 0.001), and V _cmax (R ² = 0.92, P < 0.001) under drought conditions, but only with V _cmax (R ² = 0.29, P < 0.05) under well‐watered conditions (Figure S4A). To assess the difference in the coefficients of these regression lines, representing the relationship between A _N and V _cmax under drought or well‐watered conditions, a covariance analysis was employed (Barrett, 2011), and results showed that the slopes of these regression lines were similar (Table S3). The iWUE_(Isotope) also exhibited positive correlations with g _m(Harley) (R ² = 0.30, P < 0.05) and V _cmax (R ² = 0.22, P < 0.05) under water deficit conditions (Figure S4B).

Under drought conditions, A _N exhibited positive correlations with g _m(Harley)/ g _s (R ² = 0.24, P < 0.05), as well as V _cmax/ g _s (R ² = 0.22, P < 0.05) (Figure S4A). Meanwhile, iWUE_(Isotope) also showed a stronger positive correlation with g _m(Harley)/ g _s (R ² = 0.44, P < 0.01) and V _cmax/ g _s (R ² = 0.55, P < 0.001) (Figure S4B). However, under well‐watered conditions, no such correlations were observed.

Significant linear correlations were detected between g _s and g _m(Harley), as well as between g _s and V _cmax under drought or well‐watered conditions (Figure 3A,B). However, the covariance analysis indicated that the slopes of the regression lines, representing the relationship between g _m(Harley) (Table S4), V _cmax (Table S5) with g _s, differed between drought and well‐watered conditions.

Figure 3.

Correlations between CO₂ diffusion conductance and carboxylation rate.

(A) Stomatal conductance (g _s) versus mesophyll conductance estimated using the variable J method (g _m(Harley)).

(B) g _s versus maximum carboxylation rate (V _cmax). The significance for the linear regression model was checked using F‐test at the probability level of 0.05. The black dotted and solid lines correspond to the regression lines under well‐watered and drought conditions, respectively.

Relative limitations of g _s, g _m, and biochemistry on A _N

Compared with the well‐watered treatment, drought significantly increased the relative limitation of g _s on A _N (l _s), while the relative limitation of biochemistry on A _N (l _b) was significantly decreased (Figure 4). A comparison among the plant lines revealed that, under water deficit, compared with WT plants, the transgenic lines displayed substantially lower relative limitations on A _N from both mesophyll conductance (l _m) and l _b, while l _s showed an opposite pattern.

Figure 4.

The relative limitations of photosynthesis for cotton lines under well‐watered and drought conditions.

The total relative limitation (1.0) of photosynthesis is the sum of stomatal conductance (l _s), mesophyll conductance (l _m), and biochemical (l _b) limitations. WT, wild‐type control plants. CR‐3 and CR‐13, GhCKX suppression lines. Values are means ± SE (n = 6). Different capital letters indicate significant differences at the probability level of 0.05 between different water treatments in the same plant line (Student's t‐test). Different lowercase letters indicate significant differences at the probability level of 0.05 between different plant lines within the same water treatment (LSD test).

Stomatal traits and their linkages with g _s

The representative stomatal images were presented in Figure S5, and Table S1 showed the quantified stomatal anatomical traits. Drought did not affect stomatal density (SD) on either leaf surface for all lines. Nevertheless, regardless of water conditions, both CR‐3 and CR‐13 exhibited higher adaxial and abaxial SD in relation to the WT. Drought triggered a reduction in adaxial and abaxial stomatal size (SS) for all lines (except for adaxial SS in WT) and lowered the maximum pore area (a _max) on both leaf surfaces across all lines. Under drought conditions, adaxial and abaxial SS were comparable among lines, but CR‐3 and CR‐13 possessed a larger abaxial a _max than the WT.

Regression analysis further revealed significant positive correlations between g _s and adaxial SD under drought (R ² = 0.32, P < 0.01) and well‐watered conditions (R ² = 0.36, P < 0.01) (Figure 5A), and covariance analysis indicated no difference in the slopes of these correlations between water regimes (Table S6). No significant correlations were found between g _s and adaxial a _max across plant lines under both well‐watered and drought conditions (Figure 5C). Additionally, g _s correlated positively with abaxial SD (R ² = 0.29, P < 0.05) and a _max (R ² = 0.19, P < 0.05) under drought conditions (Figure 5B,D).

Figure 5.

Relationships between stomatal conductance (g _s) and stomatal anatomical traits.

(A) Adaxial stomatal density (SD).

(B) Abaxial SD.

(C) Adaxial stomatal pore area (a _max).

(D) Abaxial a _max. The significance for the linear regression model was checked using F‐test at the probability level of 0.05. The black dotted and solid lines correspond to the regression lines under well‐watered and drought conditions, respectively.

Anatomical features and their linkages with g _m

Representative leaf anatomical images are shown in Figure S6, with corresponding quantitative data in Table S2. Anatomical traits including specific leaf weight (LMA), leaf density (LD), and the volume fraction of intercellular airspace (f _ias) showed no significant differences among lines under either water regime. Drought had no influence on the thickness of leaf (T _leaf) and mesophyll (T _mes), chloroplast length (L _chl) and thickness (T _chl), and the distance between cell wall and chloroplast (D _cw‐chl) for all lines (except for T _leaf and T _mes in CR‐13). Drought universally promoted cell wall thickness (T _cw), while CR‐3 and CR‐13 maintained significantly lower T _cw than the WT under water deficit conditions. Relative to well‐watered conditions, drought reduced the chloroplast number (N _chl) in WT plants and increased the distance between adjacent chloroplasts (D _chl‐chl) for all plant lines. However, transgenic lines possessed higher N _chl but lower D _chl‐chl than the WT under drought. Drought also reduced the ratio of exposed chloroplast area to mesophyll area (S _c/S _m) in the WT and reduced the surface areas of both mesophyll (S _m/S) and chloroplasts (S _c/S) facing the intercellular airspace across all lines. Under drought, the transgenic lines exhibited higher S _c/S _m and S _c/S than the WT, while S _m/S was similar among all lines.

Moreover, g _m(Harley) showed positive correlations with N _chl (R ² = 0.44, P < 0.01), S _c/S _m (R ² = 0.29, P < 0.05), and S _c/S (R ² = 0.35, P < 0.01), but negative correlations with T _cw (R ² = 0.44, P < 0.01) and D _chl‐chl (R ² = 0.26, P < 0.05) under drought conditions (Figure 6A–F). Under well‐watered conditions, g _m(Harley) correlated positively only with S _m/S (R ² = 0.20, P < 0.05) and S _c/S (R ² = 0.29, P < 0.05) (Figure 6E,F). Covariance analysis revealed that the slope of the g _m(Harley)‐S _c/S regression did not differ between the two water conditions (Table S7).

Figure 6.

Relationships between mesophyll conductance (g _m) and leaf anatomical features. In the present study, g _m(Harley) values were used to perform regression analysis.

(A) Cell wall thickness (T _cw).

(B) Chloroplast number (N _chl).

(C) Distance between adjacent chloroplasts (D _chl‐chl).

(D) The ratio of chloroplast area to mesophyll cell area facing the intercellular airspace (S _c/S _m).

(E) The surface area of mesophyll cells exposed to leaf intercellular airspace per unit (S _m/S).

(F) The surface area of chloroplasts exposed to leaf intercellular airspace per unit (S _c/S). The significance for the linear regression model was checked using F‐test at the probability level of 0.05. The black dotted and solid lines correspond to the regression lines under well‐watered and drought conditions, respectively.

Rubisco properties and their relationships with V _cmax

Rubisco content, initial and total Rubisco activities were decreased by drought for all lines as compared to those under well‐watered conditions (Table 1). Under drought conditions, Rubisco content, initial Rubisco activity along with the activation state were higher in the transgenic lines than the WT.

Table 1.

The characteristics of Rubisco enzyme in mature LSCBs of wild‐type and GhCKX suppression lines under different water treatments

Treatment	Plant	Rubisco content (g m−2)	Initial Rubisco activity (μmol m−2 sec−1)	Total Rubisco activity (μmol m−2 sec−1)	Rubisco activation state (%)
Well‐watered	WT	3.65 ± 0.18Aa	54.25 ± 3.42Ab	78.86 ± 4.15Aa	68.76 ± 2.01Ab
	CR‐3	3.68 ± 0.19Aa	63.07 ± 1.11Aa	82.22 ± 3.19Aa	77.37 ± 3.73Aa
	CR‐13	3.50 ± 0.18Aa	59.56 ± 2.00Aab	87.51 ± 1.74Aa	68.05 ± 1.72Bb
Drought	WT	1.13 ± 0.08Bb	38.18 ± 2.04Bb	50.40 ± 2.92Ba	75.86 ± 1.07Ab
	CR‐3	1.71 ± 0.11Ba	47.27 ± 2.01Ba	57.69 ± 2.78Ba	82.21 ± 2.12Aa
	CR‐13	1.90 ± 0.11Ba	46.93 ± 1.63Ba	56.37 ± 2.48Ba	83.60 ± 2.68Aa

WT, wild‐type; CR‐3 and CR‐13, GhCKX suppression lines. Values are means ± SE (n = 6). Different capital letters indicate significant differences at the probability level of 0.05 between different water treatments in the same plant line (Student's t‐test). Different lowercase letters indicate significant differences at the probability level of 0.05 between different plant lines within the same water treatment (LSD test).

The V _cmax showed positive correlations with Rubisco content (R ² = 0.52, P < 0.001) (Figure 7A), initial Rubisco activity (R ² = 0.23, P < 0.05) (Figure 7B), and Rubisco activation state (R ² = 0.34, P < 0.01) (Figure 7D) under drought conditions. Total Rubisco activity was not observed to be correlated with V _cmax under either water regime (Figure 7C). Moreover, V _cmax correlated positively only with Rubisco activation state (R ² = 0.23, P < 0.05) (Figure 7D) under well‐watered conditions. Covariance analysis indicated that the slope of the V _cmax‐Rubisco activation state regression line differed significantly between water conditions (Table S8).

Figure 7.

Relationships between maximum carboxylation rate (V _cmax) and Rubisco characteristics.

(A) Rubisco content.

(B) Initial Rubisco activity.

(C) Total Rubisco activity.

(D) Rubisco activation state. The significance for the linear regression model was checked using F‐test at the probability level of 0.05. The black dotted and solid lines correspond to the regression lines under well‐watered and drought conditions, respectively.

Leaf GhCKX expression and CTK contents

Drought up‐regulated GhCKX expression and reduced CTK contents (Figure S7A,B) in all lines relative to well‐watered conditions. However, under drought, transgenic lines exhibited reduced GhCKX expression but higher leaf CTK levels relative to the WT.

Leaf GhRCA expression

The expression of Rubisco activase (GhRCA) in leaf was down‐regulated by drought in all lines, with the transgenic lines exhibiting higher expression than the WT under drought conditions (Figure S8).

Leaf reactive oxygen species levels

Relative to well‐watered conditions, drought led to an accumulation of superoxide anion (O₂ ^·−) and hydrogen peroxide (H₂O₂) for all lines (Figure S9A,B). However, the transgenic lines exhibited reduced O₂ ^·− and H₂O₂ contents compared to the WT under drought.

Antioxidant enzymes and non‐enzymatic substances

Compared with well‐watered conditions, drought stress enhanced the activities of superoxide dismutase (SOD), peroxidase (POD), and catalase (CAT), as well as increased the contents of ascorbic acid (AsA) and glutathione (GSH) for all plant lines (Figure S10A–E). Notably, these antioxidant enzymes (except for CAT in CR‐3) activities along with AsA and GSH contents were higher in transgenic lines than the WT under drought conditions.

DISCUSSION

Past research has indicated that drought resistance at the leaf level is dependent on leaf age (Taulavuori et al., 2010). In particular, since young leaves are not fully mature prior to the onset of drought, drought stress might significantly affect their ontogeny (Chastain et al., 2016), while the impacts of drought on mature leaves' ontogeny are relatively smaller. Nevertheless, mature leaves exhibit greater sensitivity to drought stress in many crops, including cotton, compared to the newest expanded leaves (Chastain et al., 2016; Jung, 2004; Meng et al., 2015; Taulavuori et al., 2010), as more pronounced decreases in A _N and leaf water potential were found in mature leaves than in the newest expanded leaves under drought conditions (Chastain et al., 2016; Zhao et al., 2019). Previous investigations have demonstrated that promoting endogenous CTK content, by overexpressing IPT or suppressing CKX, can mitigate the drought‐caused adverse effects on A _N in the newest expanded cotton leaves (Hu et al., 2025; Kuppu et al., 2013) without affecting leaf water potential (Hu et al., 2025). In this study, promoting endogenous CTK content (Figure S7B), by suppressing GhCKX (Figure S7A), not only enhanced A _N (Figure 2A) but also improved Ψ _PD (Figure 1B) in mature cotton leaves under drought, indicating that the alleviating effect of CTK on the adverse effects of drought depends on the age of leaves. In contrast, promoting endogenous CTK content did not alter the A _N in mature leaves of drought‐stressed creeping bentgrass, although it did increase their relative water content and osmotic potential (Merewitz et al., 2011b). Observed differences between our results and those of Merewitz et al. (2011b) might be because the effects of CTK on A _N in drought‐stressed plants are species‐specific (Farber et al., 2016). However, in agreement with Merewitz et al. (2011a), promoting endogenous CTK levels enhanced iWUE in mature leaves under drought conditions (Figure 2E,F). Consequently, we are led to accept our initial hypothesis, which stated that increasing endogenous CTK levels can simultaneously enhance A _N and iWUE in mature LSCBs of cotton under water deficit conditions. Rivero et al. (2007) thought that this was because promoting endogenous CTK, by overexpressing IPT, delayed the senescence process of leaves under drought conditions, which was translated into sustained photosynthesis. Furthermore, in the current study, sustained higher A _N in drought‐stressed transgenic lines than in the drought‐stressed WT was beneficial for the growth of the plants (Figure 1C,D), and ultimately increased boll weight (Figure S1).

The superior performance of GhCKX suppression lines in both A _N and iWUE relative to the WT under drought was driven by facilitating g _m and V _cmax rather than decreasing g _s

Previous studies have confirmed that drought stress can improve iWUE, however, this improvement usually involves a trade‐off with A _N (Aranda et al., 2007; Han et al., 2016). In support of this perspective, under drought, all plant lines exhibited a trade‐off between increased iWUE and reduced A _N compared to well‐watered conditions (Figure 2A,F). Several studies have reported that any drought‐related reduction in A _N can be attributed to decreases in g _s, g _m, and/or V _cmax (Hu et al., 2024; Zeng et al., 2024; Zou et al., 2022). Accordingly, in the current study, A _N, g _s, g _m and V _cmax of all plant lines, were decreased markedly by drought (except for V _cmax in CR‐13) (Figure 2A–D). Moreover, although A _N was positively correlated with V _cmax under either water condition, it was positively correlated with g _s and g _m(Harley) only under drought conditions (Figure S4A), implying that the impact of promoting endogenous CTK on the relationships between A _N and g _s, g _m and V _cmax varied under different moisture conditions. Ultimately, this was manifested as a positive correlation between A _N and g _m(Harley)/ g _s as well as V _cmax/ g _s under drought conditions, but such relationships were not observed under well‐watered conditions (Figure S4A). Similarly, iWUE was positively correlated with g _m(Harley) and V _cmax, as well as with g _m(Harley)/ g _s and V _cmax/ g _s under drought conditions, while no such correlation existed under well‐watered conditions (Figure S4B). These findings suggest that the increased A _N and iWUE in transgenic lines relative to the WT during drought could be due to the relatively larger increases in g _m and V _cmax than in g _s. Consistent with this interpretation, l _m and l _b were markedly lower, while l _s was significantly higher in the GhCKX suppression lines in relation to the WT, under drought conditions (Figure 4). In addition, the relatively larger increases in g _m and V _cmax than g _s could be attributed to the alteration in the coordinated trends of g _s, g _m, and V _cmax in response to drought when endogenous CTK increased, since the slopes of the regression lines, representing the relationships between g _s and g _m(Harley) (Figure 3A; Table S4) as well as between g _s and V _cmax (Figure 3B; Table S5) differed between drought and well‐watered conditions based on the covariance analysis. These results confirm the hypothesis proposed by Gago et al. (2014), which posits that to achieve simultaneous gains in A _N and iWUE, the typical parallel changes of g _m and/or V _cmax with respect to g _s must be altered, specifically by elevating their ratios relative to g _s.

Higher g _s in GhCKX ‐suppressed lines than the WT under drought was primarily associated with increased abaxial SD and a _max

In general, g _s is governed by SS, SD, and aperture (Ouyang et al., 2017). Aligning with Li et al. (2021) and Zhou et al. (2017), our results showed that drought significantly reduced SS and a _max on both leaf surfaces across all plant lines (except for adaxial SS in WT plants) (Table S1). However, SD in mature leaves was unaffected by drought, which is in contrast to some reports where drought altered SD in the newest expanded leaves of grass (Leymus chinensis) (Xu & Zhou, 2008), basil (Ocimum basilicum) (Driesen et al., 2023), and cotton (Dubey et al., 2023; Hu et al., 2025). This might be because the stomatal differentiation of mature leaves had been completed before their exposure to drought, so the number of stomata had been determined (Hu et al., 2024). A comparison among the cotton lines revealed that under drought conditions, the transgenic lines exhibited similar SS to the WT (Table S1), ruling out SS as a cause for g _s differences between plant lines. Instead, the transgenic lines had higher SD on both leaf surfaces and abaxial a _max than the WT under drought. Additionally, tight relationships were observed between g _s and SD on both leaf surfaces as well as abaxial a _max (Figure 5A–D), implying that these traits might explain the higher g _s of transgenic plants relative to the WT during drought. However, further covariance analysis indicated that the slopes of the regression lines representing the relationship between g _s and adaxial SD did not differ between water regimes (Table S6), meaning that the effects of GhCKX suppression on adaxial SD are moisture‐independent. Furthermore, g _s is primarily governed by stomatal pore area, and SD's impact on g _s often requires concurrent changes in a _max (Lawson, 2009; Zhang et al., 2019). Given that adaxial a _max did not differ among plant lines under either moisture condition (Table S1), we can conclude that the superior g _s in GhCKX‐suppressed lines under water shortage is mainly attributable to increased abaxial SD and a _max.

Higher g _m in GhCKX ‐suppressed lines than the WT under drought was primarily attributed to decreased T _cw and D _chl‐chl, and increased S _c/S _m

Mesophyll conductance (g _m) is often driven by the structural characteristics of mesophyll tissue (Flexas et al., 2012). LMA, LD, and f _ias showed no significant variations among the plant lines under either water condition (Table S2). Furthermore, GhCKX suppression lines had similar T _leaf, T _mes (except for CR‐13), D _cw‐chl, L _chl (except for CR‐3) and T _chl to WT under drought (Table S2), suggesting that these structural characteristics of mature cotton leaves, under drought, were independent of the enhanced CTK levels and thus did not underlie the differences in g _m between the transgenic and WT plants. Contrastingly, in line with prior findings of Zou et al. (2022), T _cw was significantly increased by drought (Table S2). However, GhCKX suppression plants exhibited a significantly reduced T _cw relative to the WT under drought, indicating a shorter CO₂ diffusion path through the cell wall in the drought‐stressed GhCKX suppression lines. Higher levels of CTK in mutants can influence the remodeling process of the cell wall by regulating cell wall remodeling enzymes (e.g., beta‐glucosidase, actin, and beta‐5 tubulin) in response to drought (Merewitz et al., 2011b), which might be the reason for the lower T _cw in drought‐stressed GhCKX suppression plants in this study. Moreover, further analysis showed a positive impact of smaller T _cw on g _m under drought conditions (Figure 6A), indicating that the lower T _cw observed in the transgenic plants was a key factor in their superior g _m compared to the WT under drought. Furthermore, chloroplast arrangement is one of the key determinants influencing g _m (Liu et al., 2021). Consistent with research in mature cotton leaves under drought (Zou et al., 2022), drought increased D _chl‐chl in all plant lines (Table S2), while the GhCKX suppression lines maintained a lower D _chl‐chl than the WT under drought, which was conducive to the diffusion of CO₂ within the cytoplasm, contributing to their higher g _m. A negative correlation between D _chl‐chl and g _m across plant lines under drought further supported this finding (Figure 6C). As chloroplasts usually align parallel to the cell membrane, an increase in the number of chloroplasts will inevitably lead to a reduction in the distance between them (Takagi, 2003). This was also the case in this study, as N _chl was higher in transgenic plants than the WT and was positively correlated with g _m under drought (Figure 6B; Table S2). In addition, the higher N _chl in the transgenic lines relative to WT plants under drought would be due to their low O₂ ^·− and H₂O₂ contents within leaves (Figure S9), since low ROS reduced chloroplast degradation (Yang et al., 2024). Xu et al. (2016) found that high CTK levels can up‐regulate antioxidant defenses via promoting antioxidant enzymes (SOD, POD, and CAT, etc.) and non‐enzymatic systems (AsA and GSH, etc.). Supporting their findings, GhCKX suppression lines exhibited higher activities of SOD, POD, and CAT (except for CAT in CR‐3), as well as higher contents of AsA and GSH than the WT under drought conditions (Figure S10), which resulted in their low O₂ ^·− and H₂O₂ contents.

Consistent with previous reports of drought reducing S _c/S in mature cotton leaves (Hu et al., 2024), a decrease in this parameter was observed in all lines under drought (Table S2). Nevertheless, the GhCKX suppression lines maintained a higher S _c/S than the WT under drought, meaning that more chloroplasts were arranged along diffusion paths in the drought‐stressed GhCKX suppression lines. Significant positive g _m(Harley)‐S _c/S correlations were detected under either water condition (Figure 6F), highlighting the impact of the distribution of chloroplasts toward the intercellular airspace on g _m, while further covariance analysis indicated that the effects of GhCKX suppression on S _c/S were not different under different moisture conditions (Table S7). According to Lu et al. (2016), variations in S _c/S might be attributed to alterations in S _m/S and the S _c/S _m ratio. Similar to the S _c/S, under drought conditions, g _m(Harley) correlated positively with S _c/S _m (Figure 6D), but not with S _m/S (Figure 6E). Based on that, we conclude that the higher S _c/S observed in transgenic lines, in relation to the WT, under water shortage was due to an increased S _c/S _m.

Superior V _cmax in GhCKX ‐suppressed lines than the WT under drought was primarily driven by increased Rubisco content and initial Rubisco activity

Previous studies reported that drought lowered the V _cmax of mature leaves in chestnut oak (Quercus prinus), beech (Fagus sylvatica), black gum (Nyssa sylvatica), and red maple (Acer rubrum) (Cano et al., 2013; Wilson et al., 2000). In accordance with those reports, a lower V _cmax was found in mature cotton leaves for all lines under drought (Figure 2D). However, compared with the WT, V _cmax was higher in transgenic lines under drought, indicating that enhanced CTK can promote CO₂ fixation in drought‐stressed mature cotton leaves. This is in contrast to a prior study on the newest expanded cotton leaves (Kuppu et al., 2013), where no differences in V _cmax were observed between WT plants and overexpressed IPT mutants with increased CTK levels under reduced irrigation. Jordi et al. (2000) found that tobacco mutants with increased CTK levels allocated more nitrogen to mature rather than the newest expanded leaves when exposed to stress, thereby delaying the decomposition of Rubisco to maintain photosynthesis and resist leaf senescence. Hence, the observed divergence between this study and Kuppu et al. (2013) might be attributed to alterations in Rubisco degradation, which arise from variations in nitrogen distribution between the newest expanded and mature leaves of CTK mutants under drought conditions. Supporting this perspective, a higher Rubisco content in mature leaves was found in GhCKX‐suppressed lines relative to the WT under drought (Table 1), and V _cmax positively correlated with Rubisco content under drought conditions (Figure 7A). Apart from Rubisco content, V _cmax is also determined by Rubisco activity (Carmo‐Silva et al., 2012; Hu et al., 2023; Zhang et al., 2009). In the present study, GhCKX suppression lines exhibited markedly higher initial Rubisco activity than the WT under drought (Table 1), which in conjunction with the significant positive correlation between V _cmax and initial Rubisco activity under drought conditions (Figure 7B), explained the increased V _cmax in drought‐stressed GhCKX suppression lines. Furthermore, Rubisco activity is regulated by Rubisco activase (RCA) (Carmo‐Silva & Salvucci, 2013). A higher GhRCA expression was observed in transgenic lines relative to the WT under drought (Figure S8), which was the reason for their higher initial Rubisco activity. Supporting our suggestion, Hudeček et al. (2023) have reported that CTK exerts a significant regulatory effect on the activity of RCA. In addition to that, a significant positive correlation between V _cmax and Rubisco activation state was detected under either water treatment (Figure 7D). However, the covariance analysis indicated the slopes of the regression lines, representing the relationship between V _cmax and Rubisco activation state (Table S8), differed between drought and well‐watered conditions, indicating that the effects of GhCKX suppression on Rubisco activation state were different under different moisture conditions. Taking into consideration that the effects of GhCKX suppression on initial Rubisco activity also varied under different moisture conditions, we speculate that this might be the reason for the variation in the effects of GhCKX suppression on Rubisco activation state under different water treatments.

CONCLUSION

In summary, elevating endogenous CTK, by suppressing GhCKX, optimized the trade‐off between A _N and iWUE, with simultaneous improvements in A _N and iWUE in mature cotton leaves under drought. Moreover, these improvements were attributed to the increased g _m/g _s and V _cmax/g _s ratios, which were achieved by larger enhancements in g _m and V _cmax than in g _s caused by elevated CTK in drought‐stressed mature leaves. The higher g _s in GhCKX‐suppressed lines than the WT under drought was mainly ascribed to increased abaxial SD and a _max. Furthermore, the increased g _m in GhCKX‐suppressed lines than the WT under drought was due to decreased T _cw and D _chl‐chl and increased S _c/S _m, and their superior V _cmax was mainly associated with increased Rubisco content and initial Rubisco activity. In short, our findings demonstrate that promoting endogenous CTK levels offers a sustainable strategy for improving photosynthesis and water utilization of crops by coordinating photosynthetic limitations rather than through a simple trade‐off.

MATERIALS AND METHODS

Experimental design

This study used cotton cultivar Jimian 14 (WT) and transgenic lines with suppressed GhCKX expression (CR‐3 and CR‐13) (Xu et al., 2019). Seeds were sown on May 10, 2024, in pots (10 L) filled with clay soil within a semi‐open greenhouse shielded from rain by retractable plastic covers. Pots with two mutant lines were placed adjacent to the WT controls, and the distance between pots was not such that the plants would block each other. Meanwhile, pots were rearranged weekly to minimize edge effects. After germination, uniform and vigorous seedlings were retained, with one seedling per pot. Other agricultural management was consistent and followed local high‐yield protocols. Prior to treatment, to ensure optimal cotton growth, the SRWC was adjusted to (75 ± 5)% (Wang et al., 2016). Upon flowering at the first node of the second fruit branch, the fully expanded mature leaves subtending the flowers (LSCBs) were labeled. Simultaneously, two distinct watering conditions were initiated: (i) a well‐watered control (SRWC [75 ± 5]%) and (ii) drought stress (SRWC [45 ± 5]%) (Wang et al., 2016). There were 50 replications for each line under each water condition. SRWC monitoring and control were conducted every 2–3 days (Figure 1A) following the methods of Wang et al. (2016). Tagged leaves were sampled for analysis 15 days after treatment imposition. Subsequently, drought‐stressed plants were rehydrated to the control SRWC until boll maturation.

Determination of leaf water status

The pre‐dawn water potential (Ψ _PD) of adjacent leaves to the labeled LSCBs was measured every 2–3 days using a pressure chamber (3005, Soil‐moisture Equipment Corp., Goleta, CA, USA). Three replicates were performed for each treatment (n = 3).

Determination of plant biomass and boll weight

Images showing phenotypic performance of plants were taken using a smart camera (D810, Nikon, Tokyo, Japan). The biomass of the whole plant was recorded after drying the fresh plant tissues to stable weight at 80°C. During the boll maturation stage, the final weight of the bolls subtending to the tagged leaves was also recorded. Four replicates were performed for each treatment (n = 4).

Gas exchange and chlorophyll fluorescence

One day before the measurements were performed, pots were moved into a growth chamber. The conditions inside the growth chamber were set at 65% relative humidity, 25°C/20°C day/night air temperature, 1500 μmol m⁻² sec⁻¹ photosynthetic photon flux density (PPFD). Leaf gas exchange and fluorescence measurements on tagged mature LSCBs were performed using a LI‐6400XT gas exchange system equipped with a fluorometer (Li‐Cor, Lincoln, NE, USA). Inside the gas exchange chamber, the leaf temperature was set at 25°C, the PPFD was set at 1500 μmol m⁻² sec⁻¹ (red:blue = 90%:10%), the air flow rate was set at 500 μmol sec⁻¹, and the reference CO₂ concentration was set at 400 μmol mol⁻¹ with a CO₂ mixture (Hu et al., 2024). Once a steady‐state was achieved, leaf gas exchange parameters, steady‐state (F _s) and maximum (F'_m) fluorescence were logged. Six replicates were performed for each treatment (n = 6). Measurements were performed between 09:00 and 15:00 h.

The actual quantum efficiency of PSII (Φ _PSII) was calculated as follows: Φ _PSII = (F'_m − F _s)/F'_m. The electron transport rate (J _f) was calculated as follows: J _f = Φ _PSII × PPFD × αβ; α and β denote the leaf absorptance and partitioning of photons between photosystem I and II, respectively. The αβ was obtained from the slope of the relationship between net photosynthetic rate (A _N) and 0.25 × PPFD × Φ _PSII, which was determined via the measurements of light response curves at an oxygen level of less than 2% (Valentini et al., 1995; Yin & Struik, 2009).

The variable J method was applied to estimate mesophyll conductance (g _m(Harley)) (Harley et al., 1992):

$$g_{\mathrm{m}\left(\text{Harley}\right)}=\frac{A_{\mathrm{N}}}{C_{\mathrm{i}}-\frac{\left[J_{\mathrm{f}}+8\times \left(A_{\mathrm{N}}+R_{\mathrm{d}}\right)\right]\times {\mathrm{\Gamma }}^{*}}{J_{\mathrm{f}}-4\times \left(A_{\mathrm{N}}+R_{\mathrm{d}}\right)}}$$ (1)

where C _i denotes the intercellular CO₂ level, Γ* denotes the CO₂ compensation point in the absence of respiration, Γ* is conservative across C₃ species under the same temperature (Xiong, 2023), and hence, a value of 37.43 μmol mol⁻¹ was adopted in this study (Bernacchi et al., 2002). R _d denotes daytime respiration, which was computed to be half of dark respiration (R _n) (Han et al., 2018). R _n was measured using a LI‐6400XT gas exchange system in the dark. After gas exchange measurements, the J _f/A _N ratios for all well‐watered plant lines were calculated and they were at 7.75–8.91, within the range of 7.5–10.5 for non‐stressed C₃ plants (Perera‐Castro & Flexas, 2023), indicating that the control plants were under optimal conditions.

Estimation of maximum carboxylation rate (V _cmax)

The CO₂ response curves on tagged mature LSCBs were measured by adopting the automatic program of the gas exchange analyzer (reference CO₂ concentrations were set as: 400, 200, 100, 50, 400, 600, 800, 1000, 1200, 1500, 1800 μmol mol⁻¹). The leakage of the leaf chamber was corrected according to Flexas et al. (2007). Six replicates were performed for each treatment (n = 6), and measurements were performed between 8:30 and 16:30 h each day. The chloroplastic CO₂ level (C _c) was derived from the formula of C _c = C _i − A _N/g _m(Harley) (Harley et al., 1992). Furthermore, V _cmax was calculated using whole data points rather than parts of the curve based on the A _N‐C _c curves (Miao et al., 2009):

$$\begin{array}{c}{A}_{\mathrm{N}}=\frac{\left(C_{\mathrm{c}}-{\mathrm{\Gamma }}^{*}\right)\times V_{\text{cmax}}}{K_{\mathrm{c}}\times \left(1+\frac{O_{\mathrm{i}}}{K_{\mathrm{o}}}\right)+C_{\mathrm{c}}}-R_d\end{array}$$ (2)

where O _i refers to the intracellular O₂ level, and K _c and K _o denote the Michaelis–Menten values for carboxylation and oxygenation by Rubisco, respectively. The values of O _i with 210 mmol mol⁻¹, K _c with 272.38 μmol mol⁻¹, and K _o with 165.82 mmol mol⁻¹ at 25°C were used in this study (Bernacchi et al., 2002).

Intrinsic water use efficiency (iWUE)

Two distinct approaches were employed to estimate iWUE: one derived from gas exchange (iWUE_GE, the ratio of A _N to stomatal conductance [g _s]), and the other based on carbon isotope analysis (iWUE_Isotope).

Tagged mature LSCBs were dried and ground. About 0.5 mg of powder was placed in tin cups and combusted to determine carbon isotope ratio using isotope ratio mass spectrometer (DELTA V Advantage, Thermo Fisher Scientific, Waltham, MA, USA) coupled to an elemental analyzer (multi EA S 5000, Analytik Jena, Jena, Germany). Six replicates were performed for each treatment (n = 6). The carbon isotope composition (δ¹³C) was then calculated as follows:

$$\begin{array}{c}{\mathrm{\delta }}^{13}{\mathrm{C}}_{\text{sample}}=1000‰\times \left(\frac{R_{\text{sample}}}{R_{\text{standard}}}-1\right)\end{array}$$ (3)

where R denotes the ¹³C/¹²C ratio, and the standard reference is PeeDee Belemnite (PDB).

Isotopic discrimination (∆¹³C) was further determined via the following equation:

$$\begin{array}{c}{\Delta }^{13}\mathrm{C}=\frac{{\mathrm{\delta }}^{13}{C}_{\mathrm{a}}-{\mathrm{\delta }}^{13}{C}_{\text{sample}}}{\frac{{\mathrm{\delta }}^{13}{C}_{\text{sample}}}{1000}+1}\end{array}$$ (4)

where δ¹³C_a denotes the atmospheric carbon isotope ratio, which can be estimated as −0.029 × t + 49.032 (Driscoll et al., 2020), with t being the sampling year. In this study, t = 2024, and hence, a value of −9.66‰ was used.

The calculation of iWUE_(Isotope) employed the following expression:

$$\begin{array}{c}{\text{iWUE}}_{\text{Isotope}}=\frac{C_{\mathrm{a}}}{1.6}\times \frac{\left(b-{\Delta }^{13}\mathrm{C}\right)}{b-a}\end{array}$$ (5)

where C _a denotes the CO₂ concentration in the air, which can be estimated as 2.09 × t − 3811 (Driscoll et al., 2020). Similarly, the sampling year is 2024, and hence a value of 419.16 μmol mol⁻¹ was used in this study. a and b denote the fractionations associated with the stomatal pores (4.4‰) and carboxylation (29‰), respectively (Xiong, 2023).

Analysis of quantitative limitation on A _N

The relative limitations imposed by g _s (l _s), g _m (l _m), and photosynthetic biochemistry (l _b) on A _N were assessed using the following equations (Grassi & Magnani, 2005):

$$\begin{array}{c}{l}_{\mathrm{s}}=\frac{\frac{g_{\mathrm{tot}}}{g_{\mathrm{s}}}\times \frac{{\partial A}_{\mathrm{N}}}{{\partial C}_{\mathrm{c}}}}{g_{\mathrm{tot}}+\frac{{\partial A}_{\mathrm{N}}}{{\partial C}_{\mathrm{c}}}}\end{array}$$ (6)

$$\begin{array}{c}{l}_{\mathrm{m}}=\frac{\frac{g_{\mathrm{tot}}}{g_{\mathrm{m}}}\times \frac{{\partial A}_{\mathrm{N}}}{{\partial C}_{\mathrm{c}}}}{g_{\mathrm{tot}}+\frac{{\partial A}_{\mathrm{N}}}{{\partial C}_{\mathrm{c}}}}\end{array}$$ (7)

$$\begin{array}{c}{l}_{\mathrm{b}}=\frac{g_{\mathrm{tot}}}{g_{\mathrm{tot}}+\frac{{\partial A}_{\mathrm{N}}}{{\partial C}_{\mathrm{c}}}}\end{array}$$ (8)

where g _tot is the conductance to CO₂ from the leaf surface to chloroplast:

$$\begin{array}{c}{g}_{\mathrm{tot}}=\frac{1}{\frac{1}{g_{\mathrm{s}}}+\frac{1}{g_{\mathrm{m}}}}\end{array}$$ (9)

In this study, values of g _m(Harley) were adopted to conduct relative limitation analysis. The ∂A _N /∂C _c values were estimated from the slope of A _N‐C _c curves under C _c concentration <100 μmol mol⁻¹ (Lu et al., 2016).

Stomatal trait determinations

Following the measurements of leaf gas exchange, six leaf segments (5 × 5 mm) were sampled from the central area of the same leaves. Leaf segments were fixed in 10% (w/v) NaOH solution for 24 h and then bleached with 50% ethanol solution (Xiong et al., 2018). After staining with 5% safranin O solution (in ethanol), images of both adaxial and abaxial leaf sides were photographed using a light microscope equipped with a smart camera (BX41, Olympus, Tokyo, Japan). The following stomatal traits: guard cell length (GL), stomatal density (SD) and width (SW), as well as stomatal pore width (PW) and length (PL) were quantified using ImageJ software (NIH, Bethesda, MD, USA). The stomatal size (SS) and maximum pore area (a _max) were calculated using 0.25 × π × GL × SW and 0.25 × π × PW × PL, respectively, as previously suggested (Zhang et al., 2019). Stomatal traits were accounted from six independent fields of view per treatment (n = 6).

Leaf anatomical measurements

Six leaf segments (2 × 2 mm) were collected from the central region of tagged LSCBs that were used for gas exchange measurements (avoiding major and secondary veins) and were immersed immediately in 2.5% glutaraldehyde solution (pH = 7.4) for 24 h. The procedures to obtain semi‐thin (1 μm) and ultrathin (80 nm) sections were described in previous studies (Hu et al., 2024; Zou et al., 2022). The light and electron sections were observed and photographed using a light microscope (BX41, Olympus, Tokyo, Japan) and a transmission electron microscope (HT7800, Hitachi, Tokyo, Japan), respectively. Anatomical traits in cross‐sectional mesophyll and ultrastructure were measured with six replicates per treatment (n = 6) using ImageJ software, including the thickness of leaf (T _leaf) and mesophyll (T _mes), cell wall thickness (T _cw), distance between two neighboring chloroplasts (D _chl‐chl), distance from cell wall to chloroplast (D _cw‐chl), chloroplast number (N _chl), chloroplast length (L _chl) and thickness (T _chl), total exposed lengths of mesophyll cells (L _m) and chloroplasts (L _c) toward the intercellular airspace. Furthermore, the volume fraction of intercellular airspace (f _ias), along with the mesophyll (S _m/S) and chloroplast (S _c/S) surface areas exposed to intercellular airspace per unit leaf area, were estimated according to Syvertsen et al. (1995):

$$\begin{array}{c}\begin{array}{c}{f}_{\mathrm{ias}}=\frac{S_{\mathrm{ias}}}{S_{\mathrm{mes}}}\end{array}\end{array}$$ (10)

$$\begin{array}{c}\begin{array}{c}\frac{S_{\mathrm{m}}}{S}=F\times \frac{L_{\mathrm{m}}}{W}\end{array}\end{array}$$ (11)

$$\begin{array}{c}\begin{array}{c}\frac{S_{\mathrm{c}}}{S}=\frac{S_{\mathrm{m}}}{S}\times \frac{L_{\mathrm{c}}}{L_{\mathrm{m}}}\end{array}\end{array}$$ (12)

where S _ias and S _mes denote the total cross‐sectional area of intercellular airspace and mesophyll tissues, respectively. F denotes curvature corrective factor, and a value of 1.42 was used according to Thain (1983). W denotes the measured section's width. The surface area ratio of chloroplasts to mesophyll cells exposed to intercellular airspace (S _c/S _m) was calculated using Formulas (11) and (12).

In addition, specific leaf weight (LMA) and leaf density (LD) were measured as suggested by Carriquí et al. (2020) with six replicates per treatment (n = 6).

Estimation of mesophyll conductance based on leaf anatomy

A gas diffusion model as modified by Tosens et al. (2012) and Tomás et al. (2013), was used to estimate anatomy‐based mesophyll conductance (g _m(anatomy)). Initially, g _m(anatomy) was divided into two parts: (i) a gas phase conductance from the sub‐stomatal cavities to the outer surface of the cell wall (g _ias), and (ii) a liquid phase conductance from the outer surface of the cell wall to the carboxylation sites in the chloroplast stroma (g _liq):

$$\begin{array}{c}\begin{array}{c}{g}_{\mathrm{m}\left(\text{anatomy}\right)}=\frac{1}{\frac{1}{g_{\mathrm{ias}}}+\frac{R\times T_{\mathrm{k}}}{H\times g_{\mathrm{liq}}}}\end{array}\end{array}$$ (13)

where T _k denotes the absolute temperature, H denotes the gas constant, and R denotes the Henry's constant. H/(R × T _k) allows for the conversion of g _liq to its corresponding equivalent g _ias. The g _ias was estimated using the following equation:

$$\begin{array}{c}\begin{array}{c}{g}_{\mathrm{ias}}=\frac{D_{\mathrm{a}}\times f_{\mathrm{ias}}}{\frac{1}{2}\times T_{\mathrm{mes}}\times \mathrm{\varsigma }}\end{array}\end{array}$$ (14)

where D _a and ς denote the diffusion coefficient and the diffusion path tortuosity for CO₂ within the gas phase, respectively, and estimates of D _a (1.51 × 10⁻⁵ m⁻² sec⁻¹ at 25°C) and ς (1.57 m m⁻¹) are provided by Xiong (2023).

The g _liq is calculated by the sum of the reciprocal of partial conductances:

$$\begin{array}{c}\begin{array}{c}\frac{1}{g_{\mathrm{liq}}}=\left(\frac{1}{g_{\mathrm{cw}}}+\frac{1}{g_{\mathrm{pl}}}+\frac{1}{g_{\mathrm{ct}}}+\frac{1}{g_{\mathrm{en}}}+\frac{1}{g_{\mathrm{st}}}\right)\end{array}\times \frac{S_{\mathrm{c}}}{S}\end{array}$$ (15)

where serial conductances correspond to the diffusive components of cell wall (g _cw), plasmalemma (g _pl), cytosol (g _ct), chloroplast envelope (g _en), and chloroplast stroma (g _st). Furthermore, the conductance in a given component during the diffusion process of CO₂ was estimated as:

$$\begin{array}{c}\begin{array}{c}{g}_i=\frac{p_{\mathrm{i}}\times D_{\mathrm{w}}\times r_{\mathrm{i},\mathrm{f}}}{\Delta L_{\mathrm{i}}}\end{array}\end{array}$$ (16)

where D _w denotes the diffusion coefficient of CO₂ within the liquid phase (1.79 × 10⁻⁹ m⁻² sec⁻¹ at 25°C), ΔL _i (m) denotes the diffusion path length, which is usually assumed to equal the thickness of a component, the dimensionless factor (r _i,f) was set to 1 for cell wall, and 0.3 for both cytosol and stroma (Tomás et al., 2013), p _i denotes the effective porosity, which was set to 1 for cytosol and stroma (Han et al., 2018), whereas the p _i of cell wall was determined using −0.3733 × T _cw + 0.3378 (Tosens et al., 2016), and an estimate of 0.0035 m sec⁻¹ was assigned for both g _pl and g _en (Lei et al., 2022), since the structures of plasmalemma and chloroplast envelope could not be quantified from light or electron sections.

Assay of Rubisco content and activity

Rubisco content was assessed using the previously established procedures of Ye et al. (2024) with six replicates per treatment (n = 6). Briefly, leaf discs (10 cm²) from tagged LSCBs were ground with 5 ml of 50 mM Tris–HCl buffer (pH 7.8) containing 5 mM beta‐mercaptoethanol and 12.5% (v/v) glycerol. Following centrifugation (12 000  g ) for 15 min at 4°C, the supernatant was collected for Rubisco content determination. An identical volume of a solution containing 2% (w/v) sodium dodecyl sulfate, 4% (v/v) beta‐mercaptoethanol, and 10% (v/v) glycerol was added to the supernatant, after which the mixture was boiled in a water bath for 5 min. The samples were loaded into a polyacrylamide gel and separated by SDS‐PAGE. After that, the gel was dyed with Coomassie Brilliant Blue R‐250 solution for 3 h. Formamide was used to extract the subunits of Rubisco, and its absorbance at 595 nm was measured. Rubisco content was then calculated by matching a standard curve of bovine serum albumin (A8013, Solarbio, Beijing, China).

Initial and total Rubisco activities were assessed following the protocols of Ye et al. (2024) with six replicates per treatment (n = 6). The supernatant of leaf samples was prepared as described above. Initial Rubisco activity was determined by adding 50 μl of supernatant, 0.65 ml of working buffer (containing 50 mM Tris–HCl buffer, pH 7.8; 10 mM MgCl₂, 1 mM EDTA, 0.2 mM NADH, 20 mM NaHCO₃, 5 mM dithiothreitol, 5 mM ATP, 10 U ml⁻¹ creatine phosphokinase, 10 U ml⁻¹ 3‐phosphoglyceric phosphokinase, 10 U ml⁻¹ glyceraldehyde‐3‐phosphate dehydrogenase and 5 mM phosphokinase), and 50 μl of 10 mM RuBP into a cuvette. Changes in absorbance at 250 nm were recorded for 60 sec. Total Rubisco activity was determined by activating 50 μl of supernatant with 0.65 ml of working buffer for 10 min and then adding 50 μl of 10 mM RuBP into a cuvette to start the determination. The initial activity normalized to total activity was defined as Rubisco activation state.

Analysis of endogenous CTK content

Total endogenous CTK levels in tagged LSCBs were assayed via HPLC‐MS/MS (4000 Q TRAP system) following the protocols of Zeng et al. (2012), and three replicates were performed for each treatment (n = 3).

Determination of O₂ ^·− and H₂O₂ contents

The contents of superoxide anion (O₂ ^·−) and hydrogen peroxide (H₂O₂) in tagged LSCBs were determined using assay kits (BC3595, BC1290, Solarbio, Beijing, China) with three replicates per treatment (n = 3).

Determination of antioxidant enzymes and non‐enzymatic substances

The activities of SOD, POD, and CAT, as well as the contents of AsA and GSH in tagged LSCBs were determined following the procedures of Xu et al. (2016). Three replicates were performed for each treatment (n = 3).

Gene expression

Reverse transcriptase‐polymerase chain reaction (RT‐qPCR) was conducted following the established procedures of Hu et al. (2025), with Ubiquitin and actin serving as internal reference genes. All primers for targeted and reference genes were listed in Table S9. Gene expression levels were determined via the 2^−ΔΔCt method, and three replicates were performed for each treatment (n = 3).

Data analysis

Statistical tests were performed using SPSS 26.0 (SPSS Inc., Chicago, IL, USA). Significant differences in physiological and morphological traits among plant lines within the same water conditions or the same plant line under the different water conditions were detected by one‐way analysis of variance followed by least significant difference test or Student's t‐test at P < 0.05 level, respectively. Relationships between parameters were fitted in Origin 2025 (Origin Lab, Northampton, MA, USA) following the Akaike Information Criterion (Hu et al., 2024). Differences in the coefficients of the fitted regression lines were evaluated by covariance analysis (Barrett, 2011). All figures were prepared using Origin 2025.

AUTHOR CONTRIBUTIONS

WH and ZZ designed the study; YY, YL, HY, SW, and WZ conducted the experiments; YY, DAL, YL, and WH performed the analysis; YY, DAL, ZZ, and WH wrote and revised the manuscript.

CONFLICT OF INTEREST

All authors declare that they have no competing interests.

Supporting information

Figure S1. Boll weight for all plant lines under well‐watered and drought conditions.

Figure S2. The relationship between g _m(Harley) and g _m(anatomy).

Figure S3. The relationship between iWUE_(GE) and iWUE_(Isotope).

Figure S4. Correlations of A _N and iWUE_(Isotope) with g _s, g _m(Harley) and V _cmax.

Figure S5. Representative images of stomatal traits in mature LSCBs.

Figure S6. Representative images of mesophyll anatomical traits in mature LSCBs.

Figure S7. The relative expression of GhCKX and CTK content in mature LSCBs.

Figure S8. The relative expression of GhRCA in mature LSCBs.

Figure S9. O₂ ^·− and H₂O₂ contents in mature LSCBs.

Figure S10. Antioxidant enzymes and non‐enzymatic substances in mature LSCBs.

Table S1. Stomatal anatomical characteristics in mature LSCBs.

Table S2. Leaf anatomical characteristics of mature LSCBs.

Table S3. The results of ancova to compare the slopes of regression lines depicting the relationship between A _N and V _cmax.

Table S4. The results of ancova to compare the slopes of regression lines depicting the relationship between g _m(Harley) and g _s.

Table S5. The results of ancova to compare the slopes of regression lines depicting the relationship between V _cmax and g _s.

Table S6. The results of ancova to compare the slopes of regression lines depicting the relationship between g _s and adaxial SD.

Table S7. The results of ancova to compare the slopes of regression lines depicting the relationship between g _m(Harley) and S _c/S.

Table S8. The results of ancova to compare the slopes of regression lines depicting the relationship between V _cmax and Rubisco activation state.

Table S9. Primer sequences of genes used for RT‐qPCR.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available on request from the corresponding author.