Pot experiments were performed out in a greenhouse with a retractable transparent plastic top and walls in 2018 and 2019 at Pailou Experimental Station, Nanjing Agricultural University, Nanjing (118°50′E, 32°02′N), China. The weather data during the experimental period was presented in Supplementary Figure 1 . The daily maximum temperature and daily minimum temperature from July 1 to August 31 were 33.3 °C and 26.0 °C in 2018, 32.8 °C and 24.9 °C in 2019, and the average sunshine duration was 7.6 h and 6.0 h in 2018 and 2019, respectively. The cotton cultivars selected for the experiment were Dexiamian 1 and Yuzaomian 9110, with a similar individual boll growth period of about 38 days, as reported in our previous experiment (Zou et al., 2020). On 7 April 2018 and 12 April 2019, seeds were sown in a nursery bed. Uniform seedlings with three true leaves were transplanted in 13 L plastic pots, which were filled with 12 kg clay soil containing 1.1 and 1.2 g kg^-1 total nitrogen (N), 65.1 and 71.3 mg kg^-1 available N, 127.9 and 138.7 mg kg^-1 available potassium (K),21.5 and 22.5 mg kg^-1 available phosphorus (P), 17.8 and 18.3 g kg^-1 organic matter in 2018 and 2019. A single plant was planted in each pot, and each pot was considered as a replication. Each treatment consisted of 58 pots, and each pot was evenly fertilized with 0.64 g P₂O₅ pot^-1, 1.28 g K₂O pot^-1 and 2.78 g N pot^-1 annually.

All plants were well-watered to keep the soil relative water content (SRWC) at (75 ± 5)% until roughly 50% of the flowers on the first fruiting position on the 4-6th fruiting branches above the cotyledon node opened, which were started on 24 June 2018 and 30 June 2019. Then, three soil water treatments consisting of control (SRWC at 75 ± 5%), moderate drought (SRWC at 60 ± 5%) and severe drought (SRWC at 45 ± 5%) were established according to previous studies (Wang et al., 2016b). Soil water content was determined following the method described by Liu et al. (2008). Soil samples from 0 to 20 cm depth were collected during soil water treatment from 1800 to 1900 h with a punch (2 cm diameter) from different pots of each treatment and then composite samples were collected. Fresh weight of the samples was determined and then these samples were oven-dried at 105°C for 8 h. Soil water content was expressed as g water g^-1 dried soil. Cotton plants would be watered to the upper soil water limit (SRWC of 80%, 65%, 50%) in the next early morning, when SRWC of each treatment was smaller than the lower limit (SRWC of 70%, 55%, 40%). SRWC was measured every 2-3 days, and its trends in this study are shown in Supplementary Figure 2 . The water treatments lasted until boll opening, 38 days, meaning that the number of days of drought corresponded to the days post anthesis (DPA) for the flowers of the first fruiting node on the 4-6th fruiting branches. On the day of establishment of different treatments, the date of flowering for the previously-noted fruiting branches was marked for subsequent sampling.

About eight marked cotton bolls with subtending leaves and bracts for each treatment were sampled on 10, 17, 24, 31 and 38 DPA at 0900-1000 h. By severing the main vein, the subtending leaves were rinsed with distilled water, then separated into two halves. The bracts were also washed with distilled water. The capsule walls were removed from the boll with sterile tweezers on ice. Half of the samples (subtending leaves, bracts and capsule walls) were immediately placed in liquid-nitrogen, then stored in an ultralow temperature freezer (-80 °C) for following analysis, the other half were dried at 105 °C for 30 min, then by 72 h at 70 °C to achieve a constant weight.

Total chlorophyll content in the three subtending leaves, bracts, and capsule walls for each treatment at 10, 17, 24, 31 and 38 DPA was measured using the ethanol and acetone extraction colorimetric method according to Moran (1982). The Pn of above photosynthetic organs from three plants per treatment was determined from 0900 to 1100 h at 10, 17, 24, 31 and 38 DPA using a portable photosynthesis system (Li-6400, Li-COR, Lincoln, NE, USA). A fluorometer leaf chamber (6400-40) was used for measuring Pn of the subtending leaf and bract according to the following settings: 380 ± 5 μmol mol^-1 CO₂ concentration, 1500 μmol m^-2 s^-1 photosynthetic photon flux density (red and blue light sources), (65 ± 5)% air relative humidity. A conifer chamber (6400-05) was used for measuring the Pn of the entire boll following the same settings except that the red and blue light sources were replaced by white light.

Dried samples of the subtending leaf, bract or capsule wall were ground to a fine powder, then 0.15 g were added to 5 ml ethanol that was 80 percent (v/v). The mixtures were exposed in an 80 °C water bath for 40 min before being filtered. The foregoing procedure was performed three times, and the three supernatants were mixed and diluted in 80% ethanol to make 25 ml. The sucrose content was determined with the supernatant according to Hendrix (1993).

Rubisco was obtained from frozen samples in accordance with Hu et al. (2015), then the obtained extract (5 μl) was added to 985 μl of reaction solution as described by Zahoor et al. (2017a). Then, to drive the beginning of the reaction, appended 50 mM ribulose-1,5-bisphosphate for 10 μl. The activity was detected through detecting the oxidation rate of nicotinamide adenine dinucleotide over 1 min at 340 nm at 25°C.

Frozen samples (0.3 g) were used to extract the SPS and SuSy, which were ground in cooled extraction solution (5 ml) containing 10 mM MgCl₂, 1 mM phenylmethylsulfonyl fluoride, 2% polyvinylpyrrolidone, 1 mM ethylene glycol bis-(2-aminoethyl ether)-tetraacetic acid, 1% (v/v) Triton X-100, 1 mM ethylenediaminetetraacetic acid, 0.5% (w/v) bovine serum albumin, and 50 mM N-(2-hydroxyethyl) piperazine-N’-(2-ethane sulfonic acid)-NaOH buffer (pH 7.5). The supernatant was obtained after centrifuging the homogenate for 15 minutes at 15,000 ×g at 4°C. The SPS, SuSy activities were determined by evaluating the sucrose synthesis in accordance with Wang et al. (2017).

The soluble acid and alkaline invertases were extracted using the same method for extraction of SPS and SuSy, and measured with minor modifications as reported by Shu et al. (2009). The activity of acid invertase was assayed using the aforementioned extract (200 μl) with 2.3 ml reaction solution [1 M sucrose, 200 mM acetic acid-NaOH (pH 5.0)], then the alkaline invertase activity was determined using the extract (200 μl) with 2.3 ml reaction solution [200 mM sodium acetateacetic acid (pH 7.5), 1 M sucrose]. Then the remaining steps were consistent with Shu et al. (2009).

At 10 and 31 DPA, the subtending leaf, bract or capsule wall from different three plants of each treatment were exposed separately to a sealed, transparent chamber which contained ¹³CO₂ (Shanghai Engineering Research Center of Stable Isotope, China) for 5 ml, for 4 h (08:30-12:30) with three replications. After 24 h (the next afternoon at 12:30), the isotope-labeled subtending leaves, with their corresponding bolls (including capsule walls, seeds and fibers) and bracts, the isotope-labeled bracts, with their corresponding bolls and subtending leaves, the isotope-labeled capsule walls with their corresponding subtending leaves, bracts, seeds and fibers were collected to estimate carbon isotope composition (δ¹³C), using an EA-1110 elemental analyzer (Carlo Erba Thermoquest, Milan, Italy) at 1020°C coupled to an isotope ratio mass spectrometer (Finnigan MAT, Bremen, Germany). Carbon isotopic value was expressed in δ¹³C(‰), and was calculated as:

Where (¹³C/¹²C)_standard is the standard value of the Pee Dee Belemnite for the isotope ratio, and (¹³C/¹²C)_standard is the sample value for the isotope ratio respectively. To estimate the amount of ¹³C added by labelling, δ notation is expressed in atom% as follows:

Where R_PDB (0.00112372) is the (¹³C/¹²C)_standard of the Pee Dee Belemnite.> To calculate the ¹³C content of the samples, we used the equation referring to Ruehr et al. (2009):

Where ¹³C content (expressed as mg) is the total amount of ¹³C in the organs, atom%_s is the atom% of the labelled sample, atom%_n is the atom% of the unlabeled natural sample, w is the dry weight of organ biomass (expressed as mg dry biomass), and C% is the percentage of total C in the sample. ¹³C allocation proportion of each organ was calculated based on ¹³C content, and mathematically expressed as:

Where ¹³C_i is the ¹³C content of the organ component, and ¹³C_total is the sum ¹³C content of the source-sink system including subtending leaf, bracts, and boll (capsule walls, seeds and fibers).

The contributions to boll weight of different photosynthetic organs were determined using the modified approach published by Araus et al. (1993). From 10 DPA to 38 DPA, aluminum foil was used to cover the subtending leaves, bracts, and capsule walls, respectively, from three different plants per treatment. To allow organs for gas exchange, a needle was used to create small holes in the foil. The biomass of seed and fiber per boll was assessed at 38 DPA of subtending leaf-darkened, bract-darkened or capsule wall-darkened plants. The relative contributions of the subtending leaf, bract and capsule wall to boll weight were computed with the following equation:

At plant maturity, all open bolls on the nine plants for each treatment were counted and harvested by hand to determine boll number per plant. Thereafter, samples were transferred to the lab, where they were weighed to obtain a total seed cotton yield per plant and an average boll weight.

The samples of subtending leaf, bract and capsule wall were used for quantitative real-time PCR (qRT-RCR) with two cultivars, Dexiamian 1 and Yuzaomian 9110; two water treatments, control [SRWC(75 ± 5)%] and 10-day, 31-day drought [SRWC(45 ± 5)%]; and three biological replicates per treatment at 0900-1000 h in 2019. The Plant Total RNA Extraction Kit was used to extract total RNA in the samples (TIANGEN, Beijing, China). RNA purity was determined using NanoDrop ultraviolet spectrophotometer (Nanodrop Technologies, Wilmington, DE, USA), and RNA fragment length or integrity was detected using Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). Primer Premier 5.0 was used to design the primers, which were then produced commercially by Invitrogen Corporation (Beijing, China) ( Supplementary Table 1 ). PrimeScript RT reagent Kit with gDNA Eraser (TaKaRa, Kusatsu, Japan) was utilized to make first-strand cDNA, then qRT-PCR was done using an Applied Biosystem 7500 real-time PCR system. The program was set as follows: initial denaturation at 95 °C, 30 s, followed by denaturation with 40 cycles at 95 °C, 5 s, then collecting fluorescence at 60 °C, 40 s, to establish the melting curve, use 95 °C, 10 s, 60 °C, 60 s, 95 °C, 15 s after the amplification, and slowly heated from 60 °C to 99 °C. Internal control 18S rRNA (L24145) was used to quantify and standardize gene expression (Wang et al., 2017).

All parameters were measured for a minimum of three biological replicates and analyzed with the SPSS Statistics 26.0 (IBM, Armonk, NY, USA). The impacts of water treatment, cultivar, and their interaction for yield components were investigated using a two-way analysis of variance (ANOVA) following Duncan’s multiple range test, different lower-case letters indicate significantly difference at P< 0.05; and other indicators with comparison in two groups were analyzed by using Student’s t-test. * and ** represent P< 0.05 and P< 0.01. Figures were designed by Origin 2019 (OriginLab, Northampton, MA, USA).

The number of bolls per plant, boll weight, seed cotton yield under drought were considerably lower than that of well-watered plants (control; Table 1 ). Noticeable disparities were uncovered in the above indicators between two cultivars. In 2018-2019, compared to control, under SRWC(60 ± 5)% and SRWC(45 ± 5)%, boll weight for Dexiamian 1 decreased by 17.0-17.5% and 21.4-22.1%, respectively. Furthermore, boll weight for Yuzaomian 9110 decreased by 20.0-21.3% and 33.2-33.4%, respectively.

Compared with the control, chlorophyll content of the subtending leaf was significantly decreased under SRWC(60 ± 5)% and SRWC(45 ± 5)% ( Supplementary Figure 3 ). However, the chlorophyll content of the bract did not alter significantly amongst the soil water deficit treatments. Compared with the control, chlorophyll content of the capsule wall was significantly decreased under SRWC(45 ± 5)% at 24, 31 DPA. The average Pn of subtending leaf in two years was reduced by 3.6-19.5% and 19.6-47.5% under SRWC(60 ± 5)% and SRWC(45 ± 5)% for Dexiamian 1; and by 6.8-26.2% and 25.8-50.5% for Yuzaomian 9110, from 10-38 DPA ( Figure 1 ). There were little significant differences in Pn between SRWC(60 ± 5)% and the control for the bract or capsule wall in either cultivar; however, a significant difference in Pn between SRWC(45 ± 5)% and the control was observed for the bract and capsule wall prior to 38 DPA, and the average Pn of bract and capsule wall in two years was reduced by 24.1-32.9% and 19.4-38.8% under SRWC(45 ± 5)% for Dexiamian 1, and by 26.4-40.4% and 26.9-46.2% for Yuzaomian 9110, respectively, from 10-31 DPA. The chlorophyll content and Pn under SRWC(60 ± 5)% decreased significantly only in the subtending leaf, but not in the bract and capsule wall, suggesting that the subtending leaf might be more sensitive to moderate drought than the other two photosynthetic organs. While the Pn under SRWC(45 ± 5)% decreased markedly in all three photosynthetic organs. Therefore, in order to reveal and compare the relative contribution of photosynthesis by different photosynthetic organs to boll weight under drought, we focused our efforts on the SRWC(45 ± 5)% treatment.

The δ¹³C values of the subtending leaf, bract and capsule wall at 10 and 31 DPA of two years were decreased under SRWC(45 ± 5)% ( Figure 2 ), and the decrease in the subtending leaf was greater than the bract and capsule wall. In addition, the reduction of δ¹³C value in the three photosynthetic organs under SRWC(45 ± 5)% of Yuzaomian 9110 was more significant than that of Dexiamian 1.

Under SRWC(45 ± 5)% at 10 DPA, the average sucrose content in two years increased by 35.6% for Dexiamian 1 and 46.2% for Yuzaomian 9110 in the subtending leaf ( Figure 3 ); and decreased by 28.1% for Dexiamian 1 and 39.9% for Yuzaomian 9110 in the capsule wall, relative to well-watered conditions. However, the sucrose content of the bract did not alter significantly between the control and SRWC(45 ± 5)%. Under SRWC(45 ± 5)% at 31 DPA, the sucrose contents of three photosynthetic organs in two years were increased, and the increase in the subtending leaf was greater than the bract and capsule wall. Moreover, Yuzaomian 9110 was more affected than Dexiamian 1, considering the sucrose contents of three photosynthetic organs under SRWC(45 ± 5)%.

Compared with the well-watered control, Rubisco activities of the aforementioned photosynthetic organs were decreased under SRWC(45 ± 5)% for both cultivars ( Figure 4 ). The activities of SPS, SuSy, the acid and alkaline invertase in the three photosynthetic organs were obviously increased at 10 and 31 DPA under SRWC(45 ± 5)% for both cultivars, compared to the control ( Figures 5 , 6 ). The amplitude of variation in the subtending leaf was greater than the bract and capsule wall, and compared to Dexiamian 1, Yuzaomian 9110 was relatively sensitive to SRWC(45 ± 5)% when it comes to the activities of SPS, SuSy, the acid and alkaline invertase.

After feeding the subtending leaf with ¹³CO₂ at 10 and 31 DPA, a notable decrease of ¹³C allocation proportion in the corresponding bract plus boll (including capsule wall, seed and fiber) was detected under severe drought ( Figure 7A ), and the average ¹³C allocation proportion of two years was decreased by 7.4% and 20.5% for Dexiamian 1, and 9.1% and 23.9% for Yuzaomian 9110, respectively, compared to the control; however, there was no significant difference in the ¹³C allocation proportion in the corresponding subtending leaf plus boll between SRWC(45 ± 5)% and the control after feeding the bract with ¹³CO₂. After feeding the capsule wall with ¹³CO₂, the ¹³C allocation proportion in the corresponding subtending leaf, bract, seed plus fiber was markedly decreased at 31 DPA under SRWC(45 ± 5)%, compared with the control.

Additionally, after feeding the subtending leaf with ¹³CO₂ at 10 and 31 DPA, ¹³C allocation proportion in the corresponding seed plus fiber was decreased under SRWC(45 ± 5)% ( Figure 7B ), and the average ¹³C allocation proportion of two years was decreased by 7.8% and 15.0% for Dexiamian 1, and 11.5% and 17.9% for Yuzaomian 9110, respectively, compared to the control. Conversely, the ¹³C allocation proportion in the corresponding seed plus fiber was not decreased after feeding the bract or capsule wall with ¹³CO₂.

The level of GhSUT1 expression in the subtending leaf was reduced at 10 DPA and 31 DPA for both cultivars under SRWC(45 ± 5)%, compared to the well-watered plants ( Figure 8 ), and the reduction in GhSUT1 expression of Yuzaomian 9110 was larger than that of Dexiamian 1. GhSUT1 expression did not alter significantly in the bract and capsule wall at 10 DPA, but a significant reduction was notable at 31 DPA under SRWC(45 ± 5)% compared to the control for both cultivars.

Compared to well-watered plants, the relative contribution of the subtending leaf to boll weight decreased by 24.7% in Dexiamian 1 and 28.3% in Yuzaomian 9110 under severe drought ( Figure 9 ), but the relative contributions of bract and capsule wall were increased in both cultivars.

Previous studies have documented substantial declines in yield that were partially a function of boll weight reductions when drought stress was experienced at cotton boll formation stage (Wang et al., 2016b; Zou et al., 2020). Consistent with previous results, one of cotton yield components, boll weight under moderate and severe drought, were substantially lower than that of the well-watered in the present study. Furthermore, the decrease in boll weight for cotton cultivar (Yuzaomian 9110) was significantly greater than another cultivar (Dexiamian 1) ( Table 1 ), which is in accordance with previous studies (Mehrabadi et al., 2015; Zou et al., 2020) where boll development of drought-sensitive cotton varieties was more susceptible to water stress compared with drought-tolerant cotton varieties. The carbohydrates required for boll filling are derived from leaves and other non-laminar photosynthetic tissues, each differing in their relative contribution to final boll mass (Simkin et al., 2019). Additionally, leaves, bracts and bolls each differ in their susceptibility to environmental stress (Hu et al., 2013). In this study, the relative photosynthetic contribution of the subtending leaf to boll weight was decreased by 24.7-28.3% under severe drought, while the relative contribution of bracts and capsule wall photosynthesis to boll weight was increased by 8.1-14.0% and 4.1-22.9% under severe drought ( Figure 9 ). Despite those increases, boll weight was significantly decreased under drought, implying the lower photosynthetic rates of the subtending leaf and consequently, its reduced contribution might be the main driver of the observed reductions in boll weight under drought stress. Similar to our results, Zhang et al. (2011), in experiments with wheat (Triticum aestivum L.), reported that non-leaf organ photosynthetic contributions to yield not only rose with declining supply of water but were even higher than contributor of the flag leaf.

Drought has been known to have detrimental effects on leaf photosynthesis and metabolism (Wang et al., 2016a), with the extent of damage depending mainly, but not only, on the intensity of drought stress. For example, Zhao et al. (2019) reported that drought reduced subtending leaf photosynthesis, where photosynthetic rates were positively associated with soil water content for a given treatment. Zou et al. (2022a) observed that the mesophyll conductance limitations increased and significantly contributed to photosynthetic rate reductions of subtending leaf under short- or long-term drought stress conditions. Accordingly, in this study, moderate and severe drought decreased Pn of the subtending leaf in both cultivars, irrespective of their drought-tolerance with the decrease being significantly larger under severe compared to moderate drought stress conditions ( Figure 1 ). Increasing drought stress severity resulted in a similar decreasing pattern in the subtending leaf’s chlorophyll content ( Supplementary Figure 3 ), which was rather expected, taking in consideration the pivotal role of chlorophyll in the photosynthetic process (Eggink et al., 2001). In contrast to the subtending leaf, Pn and chlorophyll content of the capsule wall were markedly decreased only under conditions of severe drought stress ( Figure 1 ; Supplementary Figure 3 ). This is in accordance with Gao et al. (2021) who also noted that drought stress has to be severe enough in order to result in reductions in chlorophyll content and Pn of the capsule wall since moderate drought had no effect on either trait in their study. Similar to the subtending leaf and the capsule wall, bract Pn were also decreased under severe drought stress conditions ( Figure 1 ); however, contrastingly, those decreases were not accompanied by reductions in bract chlorophyll content ( Supplementary Figure 3 ). Zhang et al. (2015) reported comparable results and speculated that this might be related to the restriction of CO₂ assimilation in the bract. Evaluating stable carbon isotope abundance provides a way to evaluate source strength, and typically, lower Pn implies lower δ¹³C (Wu et al., 2019; Zhang et al., 2019). In our study, severe drought led to the reduction in δ¹³C values of all three photosynthetic organs in both cultivar of Dexiamian 1 and Yuzaomian 9110 compared to well-watered conditions ( Figure 2 ), confirming the marked restrictions in CO₂ assimilation that were indicated by the decreased Pn. Moreover, the reduction observed in the δ¹³C value, and consequently CO₂ assimilation, of the subtending leaf decrease was much larger, compared to that of the bract and/or the capsule wall, and this could partially explain the decrease in the relative photosynthetic contribution of the subtending leaf to the boll weight under severe drought that was discovered in our study. Furthermore, larger decreases in Pn and δ¹³C values of subtending leaf were found in Yuzaomian 9110 than Dexiamian 1 under severe drought ( Figures 1 , 2 ), indicating the photosynthetic assimilation of Yuzaomian 9110 was more sensitive to drought than that of Dexiamian 1.

Carbon assimilation is greatly dependent on Rubisco activity (Parry et al., 2013). In our study, Rubisco activity of the three photosynthetic organs was significantly decreased under severe drought conditions ( Figure 4 ), confirming that CO₂ fixation and formation of 3-phosphoglycerate was substantially constrained. Because sucrose is the primary photosynthate synthesized for subsequent transport to sink organs like the cotton boll (Zahoor et al., 2017a), one would anticipate a reduction in its content for source leaves exposed to photosynthesis-limiting drought. Nevertheless, despite the reduced CO₂ fixation, sucrose contents of the subtending leaf, at 10 DPA and 31 DPA, as well as the bract and capsule wall at 31 DPA, were significantly increased ( Figure 3 ). Accumulation of carbohydrates, in photosynthetic and non-photosynthetic organs under drought stress conditions, is a common mechanism of plants to maintain cell turgor and prevent cell dehydration (Liu et al., 2004; Muller et al., 2011; Durand et al., 2016). Especially in cotton, several studies have noted increases in sucrose under drought stress conditions and they have been credited with alterations in sucrose metabolizing enzymes activities (Zou et al., 2022b) or reductions in sucrose translocation (Lemoine et al., 2013). SuSy is most often considered to catalyse the reaction of sucrose decomposition in sink organs (Winter et al., 1997), however, SuSy activity of photosynthetic organs was measured in the synthetic sucrose direction of this study. In the current study, the activities of sucrose-synthesizing enzymes, SPS and SuSy, and sucrose-hydrolyzing enzymes, acid and alkaline invertase, were significantly increased in all three photosynthesizing tissues of drought stressed plants, compared to the control ( Figures 5 , 6 ). In partial agreement with our results, Pilon (2015) found that carbohydrate metabolism in cotton leaves was compromised by water-deficit conditions. Zahoor et al. (2017a) observed substantial increases in SPS and SuSy activities of drought-stressed cotton plants, accompanied by concomitant increases in leaf sucrose content. As mentioned above however, sucrose translocation out of photosynthesizing tissues also determines tissue sucrose content (Koch, 2004). In our study, expression of GhSUT1, encoding a sucrose transporter in the subtending leaf, was down-regulated at all sampling dates under severe drought stress ( Figure 8 ), indicating that a likely reduction in sucrose export and consequently resulting in sucrose accumulation in drought-stressed subtending leaf. Expression of GhSUT1 in the bract and capsule wall was not affected at 10 days into the stress; however, it was substantially lower at 31 days into the stress compared to the control. At 10 DPA, sucrose content in the bract was not influenced by drought, but sucrose content in the capsule wall was significantly lower under severe drought ( Figure 3 ), indicating that drought-induced changes in sucrose content of bracts and capsule walls during this period are governed by the balance between sucrose synthesis and degradation, not transport. However, sucrose contents of bract and capsule wall were increased at 31 DPA ( Figure 3 ), indicating that despite the delayed response, compared to the subtending leaf, severe drought stress markedly inhibited sucrose export from the bract and capsule wall resulting in substantial increases in their sucrose content. Taking in consideration that sucrose translocation from the bract and capsule wall was only observed at 31 days into stress, a time point at which boll development has been almost completed, we speculate that this might be the reason why the photosynthetic contribution of the bract and capsule wall to boll weight was not decreased or even increased under severe drought stress.

In support of our speculation, after feeding the bract or the capsule wall at 10 DPA with ¹³CO₂, the ratio of ¹³C allocated to other organs was not altered by severe drought ( Figure 7 ). Contrastingly, after feeding the subtending leaf with ¹³CO₂, the ratio of ¹³C content of corresponding bract plus boll (capsule wall, seed, fiber) to total ¹³C content (subtending leaf, bract plus boll) was markedly decreased under severe drought stress, compared to CK ( Figure 7A ), indicating that, among the three photosynthetic organs, only the export of photosynthate from the subtending leaf to other organs was restricted by severe drought at 10 and 31 DPA. In accordance with our results, Zahoor et al. (2017a) observed that water stress reduced leaf photo-assimilates partitioning towards other organs. Furthermore, only under severe drought was the distribution of ¹³C from the subtending leaf to seed and fiber decreased ( Figure 7B ), meaning that the photosynthate distribution within these non-leaf organs was not readily affected by drought stress. Bearing in mind that both the bracts and the capsule wall are closer to the seeds and fiber, compared to the subtending leaf, we speculate that the shorter distance the photosynthate produced by these non-leaf organs needs to cross (Hu et al., 2012) might be the reason why the relative photosynthetic contribution of the subtending leaf to boll weight was the only one to decrease under drought. This is in agreement with our previous conclusion that the subtending leaf is more sensitive to drought stress than non-leaf tissues.