Activity of leaf and spike carbohydrate-metabolic and antioxidant enzymes linked with yield performance in three spring wheat genotypes grown under well-watered and drought conditions


 Background: To improve our understanding about the physiological mechanism of grain yield reduction at anthesis, three spring wheat genotypes [L1 (advanced line), L2 (Vorobey) and L3 (Punjab-11)] having contrasting yield potential under drought in field were investigated under controlled greenhouse conditions., drought stress was imposed at anthesis stage by withholding irrigation until all plant available water was depleted, while well-watered control plants were kept at 95% pot water holding capacity. Results: Compared to genotype L1 and L2, pronounced decrease in grain number (NGS), grain yield (GY) and harvest index (HI) were found in genotype L3, mainly due to its greater kernel abortion (KA) under drought. A significant positive correlation of leaf monodehydroascorbate reductase (MDHAR) with both NGS and HI was observed. In contrast, significant negative correlations of glutathione S-transferase (GST) and vacuolar invertase (vacInv) both within source and sink with NGS and HI were found. Likewise, a significant negative correlation of leaf abscisic acid (ABA) with NGS was noticed. Moreover, leaf aldolase and cell wall peroxidase (cwPOX) activities were significantly and positively associated with thousand kernel weight (TKW). Conclusion: Distinct physiological markers correlating with yield traits and higher activity of leaf aldolase and cwPOX may be chosen as predictive biomarkers for higher TKW. Also, higher activity of MDHAR within the leaf can be selected as a predictive biomarker for higher NGS in wheat under drought. Whereas, lower activity of vacInv and GST both within leaf and spike can be selected as biomarkers for higher NGS and HI. The results highlighted the role of antioxidant and carbohydrate-metabolic enzymes in the modulation of source-sink balance in wheat crops, which could be used as bio-signatures for breeding and selection of drought-resilient wheat genotypes for a future drier climate.

are important to maintain the redox homeostasis under abiotic stress [31] [36] . Studies reported an increase in the activity of MDHAR in rice under drought stress and; enhanced activity of DHAR and GR under drought stress in wheat [33] [35] [36]. Moreover, glutathione-S-transferase (GST) also plays an important role to reduce the oxidative damage within plants [37] [38] to improve the tolerance to different stresses [38]. Plant genotypes having higher activity of these antioxidants are expected to produce more yield under stress conditions. Studies have been conducted to understand the role of these antioxidants during drought stress in wheat [33] [35] however, their role in relation to carbohydrate metabolic enzymes could explain the mechanism of drought in depth.
In this study, the response of three wheat genotypes (L 1 (advanced line), L 2 (Vorobey) and L 3 (Punjab-11)) having contrasting yield potential under drought in eld were investigated in pot experiment under controlled greenhouse condition. Drought stress was imposed at anthesis stage as it is the most sensitive stage of wheat crops to water de cits [1][2] [3]. Our aim was to explore the changes in yield parameters and variation in leaf and spike carbohydrate metabolic and antioxidant enzyme activity signatures and their associations in the three wheat genotypes under both well-watered and drought stress conditions. The results will help to nd discriminating biomarkers in order to devise future strategies for breeding drought resilient wheat cultivars.

Leaf gas exchange and water relations
No signi cant differences in stomatal conductance (Gs), photosynthetic rate (An), relative water content (RWC) and osmotic potential (Y p ) were observed between the three genotypes. Compared to the well-watered controls, drought stress signi cantly decreased Gs, An, RWC and Y p in all genotypes ( Table 2).
Genotypes were signi cantly different for osmotic adjustment (OA) and highest value of osmotic adjustment (OA) was recorded in genotype L 2 and lowest in L 1 (Table 3).

Activity of carbohydrate metabolic enzymes in leaf
Activity of leaf vacuolar invertase (vacInv) was signi cantly different between the three genotypes under control conditions, and the highest activity was recorded in L 3 and lowest in L 1 . Compared to well-watered controls, no signi cant differences were recorded for the activity of vacInv under drought conditions. All genotypes exhibited similar cytoplasmic invertase (cytInv) activity under control conditions, while drought caused a non-signi cant increase of cytInv activity. The activity of cell wall invertase (cwInv) was statistically similar among the genotypes, though L 3 showed a lower activity than L 1 and L 2 under control conditions. Drought signi cantly enhanced the activity of this enzyme in comparison to the well-watered controls. (Table 4).
The activities of AGPase and UGPase were signi cantly different among the three genotypes where, the lowest activities of both enzymes were noticed in L 2 in comparison to the other two genotypes. Compared to the well-watered controls, signi cant reduction of leaf AGPase activity by drought was observed.
Drought did not affect the activity of UGPase. Also, the activity of fructokinase (FK) was signi cantly different among genotypes where, higher activity was recorded in genotype L 3 in relation to the other two genotypes. Drought signi cantly reduced the activity of FK in comparison to well-watered controls. The activity of hexokinase (HXK) was neither affected by genotype nor by drought; whereas, interaction between water*genotype which was 0.2 to 0.07 nkat g -1 Fw was signi cant in genotype L 2 (Table 4).
Phosphoglucomutase (PGM) activity was statistically similar among the three genotypes Activities of phosphoglucoisomerase (PGI) phosphofructokinase (PFK) varied signi cantly among three genotypes and highest were recorded in L 3 in comparison to other two genotypes. Drought did not affect the activity of PGM, PGI and PFK. Non-signi cant differences were recorded among genotypes for the activity of aldolase. Compared to well-watered controls, activity of aldolase was reduced signi cantly under drought (Table 4).

Activity of carbohydrate metabolic enzymes in spike
The activity of vacInv was signi cantly different among genotypes and the highest activity was recorded in L 3 in relation to L 1 and L 2 . Compared to the wellwatered controls, the activity of vacInv was not signi cantly affected by drought. The activity of cytInv enzymes was signi cantly different among three genotypes where, higher activity was recorded in L 1 in comparison to other two genotypes. Compared to well-watered controls the activity of cytInv was signi cantly increased under drought. The activity of cwInv was identical among the three genotypes and it was unaffected by drought (Table 4).
Signi cant differences in the activity of AGPase were found between genotypes and it was highest for L 2 in comparison to other two genotypes. No signi cant effect of drought on the activity of AGPase was noticed. Neither genotype nor drought affected the activity of UGPase signi cantly. The activities of FK and HXK was signi cantly varied between the genotypes, where higher activities of these enzyme were found in L 2 compared to other two genotypes, and FK activity was not signi cantly affected by drought (Table 4). In contrast, the activity of HXK was signi cantly increased under drought as compared to wellwatered controls (Table 4).
Differences were signi cant among genotypes for the activities of PGM and PFK and higher activities were recorded in L 1 in comparison to other two genotypes (Table 4). Signi cant differences for the activity of PGI were noticed among the genotypes and higher activity was recorded in L 2 . However, activities of PGM and PGI were not signi cantly affected by drought. Compared to well-watered controls, drought signi cantly enhanced the activity PFK.
Aldolase activity was neither affected by genotypes nor by drought (Table 4).

Abscisic acid concentration and antioxidants activity in leaf
Leaf ABA concentrations differed signi cantly among the three genotypes where highest ABA concentration was recorded in L 2 compared to other two genotypes. Compared to the well-watered controls, leaf ABA concentration was signi cantly higher under drought conditions. A signi cant interaction between water*genotype was also notice for leaf ABA concentration where pronounced effect was recorded in genotype L 3 (Table 5).
Neither genotypes nor drought changed the activities of DHAR, MDHAR and GR statistically. Difference were signi cant among genotypes for GST where, highest activity was recorded in genotype L 1 as compared to other two genotypes. Compared to the well-watered controls, drought signi cantly increased the activity of GST. Likewise, the interaction of water*genotype was also signi cant and pronounced increase which was 1.49 to 10.27 nkat g -1 FW recorded in genotypes L 3 . Differences were signi cant for the activity of POX among genotypes, where greater activity was observed in genotype L 1 compared to other two. However, non-signi cant differences for the activity of POX were recorded between the well-watered and drought-stressed plants. Similarly, differences were also signi cant among genotypes where, highest activity for cwPOX was observed in genotype L 3 compared to other two genotypes, Moreover, compared to well-watered controls, cwPOX was signi cantly affected by drought. (Table 5).

Abscisic acid and antioxidants activity within spike
The ABA concentration was signi cant among genotypes where, highest ABA was recorded in L 1 compared to other two genotypes. As expected, ABA concentration was signi cantly increased by drought in comparison to well-watered controls. There was also a signi cant interactive effect of water*genotype on spike ABA concentration where pronounced increase of ABA by drought was recorded in genotype L 3 in relation to L 1 and L 3 ( Table 5).
Differences were signi cant among genotypes for GST activity and the highest value was recorded in genotype L 1 and lowest in L 3 . Drought signi cantly increased the activity of GST in comparison to well-watered controls. Likewise, differences were signi cant among genotypes for the activity of DHAR where, highest activity was recorded in L 2 in comparison to the other two genotypes. The activity of GR was signi cantly different between the genotypes and the highest value was recorded in L 1 . No signi cant differences were observed for the activities of DHAR and GR between the well-watered and the drought stressed plants. Signi cant interaction of water*genotype was observed for GR where, more pronounced decrease in the activity of GR by drought was observed in L 1 . Differences were signi cant among genotypes for the activity of POX and the highest value was observed in L 3 . Again, compared to wellwatered controls, non-signi cant effect of drought was recorded for the activity of POX (Table 5). Moreover, neither genotype nor drought signi cantly affected the activity of cwPOX. Differences were signi cant among genotypes for the activity of MDHAR where the lowest value was observed for L 2 as compared to other genotypes. In relation to the well-watered controls, drought did not affect activity of MDHAR (Table 5).

Agronomic parameters
Shoot biomass was identical among the three genotypes, while it was signi cantly reduced by drought in relation to the well-watered controls. There was signi cant interaction between water*genotype on shoot biomass, where more pronounced reduction in plant biomass by drought was recorded in L 2 in comparison to the other two genotypes (Fig. 1a). Grain yield pot -1 (GY) and harvest index (HI) were signi cantly different between the three genotypes, L 3 had the lowest GY and HI in comparison to the other two genotypes. In comparison to the well-watered controls, GY and HI were signi cantly reduced under drought (Fig. 2a &b).
Differences were also signi cant among genotypes for TKW with the highest value recorded for L 3. Drought signi cantly reduced the TKW in comparison to well-watered controls (Fig. 1d). The number of grains spike -1 (NGS) was signi cantly different among three genotypes with the highest NGS recorded for L 2 and lowest for L 3 . Moreover, in comparison to well-watered controls, drought signi cantly reduced NGS. Additionally, signi cant interaction of water*genotype was recorded for NGS, where pronounced grain reduction due to drought was found in genotype L 1 as compared to L 2 and L 3 (Fig. 1e). Kernel abortion (KA) was signi cantly different among all genotypes. Highest KA was recorded in genotype L 3 and lowest in L 1 (Fig. 1f). As compared to well-watered controls, drought signi cantly increased KA. Moreover, interaction between water*genotype was also signi cant and pronounced reduction was noticed in L 1 .

Principal component analysis and combined correlations between yield traits and enzymatic activities
Separated PCA analyses for plants grown under well-watered and drought-stressed conditions were performed visualizing the associations between the yield traits and the enzymatic activities. Principal component 1 (Dim1) and principal component 2 (Dim2) described 26.8% and 18.4% variability among the variables for the well-watered treatment, respectively. Biplot analysis of Dim1 and Dim2 showed that cluster of NGS, GY and HI was closer to An, activity of L.cwInv and L.MDHAR, and these variables were in opposite direction of the cluster for L.vacInv, S.vacInv, and L.cwPOX. The activities of S and L.aldolase clustered closer to BM and in opposite direction to S-cytInv (Fig. 2a). Under drought, 25.3 % and 21.8% of variability was described by PC 1 and PC 2 , respectively (Fig. 2b). Biplot of these PC's showed that NGS, RWC, HI and GY were clustered closer to An, Gs, S.aldolase and L.MDHAR and were in opposite direction of L.vacInv and S.vacInv, KA, TKW and L.cwPOX. 2.7.1. Correlation of leaf parameters with yield-related traits RWC correlated signi cantly and positively with aldolase, cwPOX, OA, An, Gs, E, BM, GY, NGS, TKW and HI. However, this correlation was strong (***) with aldolase, An, Gs, E, BM, GY, NGS and HI, moderate (**) with TKW and weak (*) with cwPOX and OA. RWC chowed a strong signi cant and negative correlation with ABA, cwInv, GST, negative Y p and KA. It was negative but weak with vacInv. (Table 6). A strong signi cant and positive correlation of ABA was recorded with GST and Y p moderate with cwInv and cytInv and weak with KA. ABA showed a strong negative correlation with An, Gs and E moderately negative with NGS and weakly negative with aldolase, cwPOX, BM, GY and NGS. A moderate positive correlation of cwInv was recorded with GST and Y p . It was strong and negative with TKW moderate and negative with An, Gs and E while weak and negative with aldolase, BM and GY. CytInv showed a weak positive correlation with GST and Y p while it was weak and negative with An. VacInv has a weak positive correlation with Y p , moderately negative with HI and weakly negative with GY and NGS. A strong positive correlation of aldolase was estimated with cwPOX, An, Gs, E, BM and TKW and it was moderate and positive with GY while the correlation of aldolase was strong and negative with GST and Y p .
Moderate positive correlation of cwPOX was measured with Gs, E and TKW and it was weakly positive with An and BM while correlation of cwPOX was moderate and negative with GST, Y p and KA. MDHAR showed moderate and positive correlation with OA, NGS, GY and HI and it was moderately negative with KA. A strong positive correlation of GST was estimated with Y p while it was weak and negative with KA. This correlation was strong and negative with An, Gs, E, BM and GY, moderate and negative with NGS, TKW and HI. Negative Y p showed negative correlation with most of the yield related traits except KA. However, this correlation was strong with An, Gs, E, BM, GY and HI while it was moderate and weak with NGS and TKW respectively. A strong and positive correlation of OA was estimated with An while it was weak and positive with NGS moreover, moderate and negative correlation was recorded with KA.
A strong positive correlation of An was noticed with most of yield related traits except TKW where it was moderately positive. A moderate but negative correlation of An was recorded with KA. Correlation of Gs was similar to An except for NGS, TKW and HI where it was moderate and positive. The correlation of Gs weak and negative with KA. E was also showed similar correlation to Gs except it was weak and positive with NGS and HI. BM showed moderate and positive correlation with GY and TKW while it was weak and positive with NGS and HI. A strong and positive correlation of GY recorded with NGS and HI however, correlation was weak and positive with TKW. Like GY, correlation of NGS was strong and positive with HI. In contrast, KA showed strong and negative correlation with GY, NGS and HI (Table 6).

Correlation of leaf parameters with yield related traits
A strong and positive correlation of ABA was recorded with activities of GST, it was moderate and positive with activities of PFK and weak and positive with An. Correlation of ABA was moderate and negative with Gs and weak but negative with E (Table 7). CwInv showed strong and positive correlation with vacInv. A strong positive correlation of vacInv was estimated with UGPase and PFK, moderate and positive with GST and weak and positive with PGM. CytInv showed weak and negative correlation with aldolase, An, BM GY and TKW. The correlation of vacInv was weak and positive with KA and correlation weak and negative with PGM, BM, GY and HI.
UGPase showed moderate and positive with PFK and weak and positive with KA, moderate and negative with TKW and weak and negative with An. A strong and positive correlation of PGM was estimated with PFK while strong and negative with TKW. PFK showed strong and positive correlation with GST. It was strong and negative with TKW moderate and negative with An, Gs and BM and weak and negative with E and GY. Aldolase showed weak and positive correlation with An and weak but negative with GST. A moderately and positive correlation of GST was estimated with KA however, it was strong and negative with An, Gs, E, BM and GY and moderate negative with NGS, TKW and HI.

Discussion
A better understanding of the physiological and biochemical mechanisms attributing to grain yield losses can be tracked by studying the diverse genotypes having varying yield potential under both well-watered and drought-stressed conditions. In this study, three contrasting genotypes (L 1 , L 2 and L 3 ) having different yield potential in eld were selected. Genotype L 1 and L 3 were selected as drought tolerant and drought sensitive, respectively, while L 2 was of intermediate yield potential based upon their performance in the eld.
Drought stress at anthesis causes limited availability of photosynthates which modi es sink capacity [39] and reduces plant biomass, yield and ultimately the harvest index [40]. The reduction of grain yield in wheat caused by "anthesis drought" was attributed to both reduced grain number and individual grain weight [41] [42]. Besides, modi cation of key carbohydrates metabolism enzymes both in leaf and spike could have been associated with the reduced source activity and sink strength which resulted in increased kernel abortion and lowered thousand kernel weight [43]. The limited supply of carbohydrates and alteration in the activity of key carbohydrates metabolic enzymes may induce further modi cations within the plants. Moreover, the production of reactive oxygen species (ROS) and detoxi cation of ROS through antioxidants is one of the most prominent mechanisms of plants response to drought stress [33].

The correlation of An and OA with HI
It has been well established that drought reduces the carbon assimilation and photosynthate supply in crop plants [44]. This limited supply of concurrent photosynthate could also modify dynamics of key carbohydrate metabolism in both source to sink organs [45]. In this study, severe reduction in photosynthesis, stomatal conductance and transpiration rate were recorded under drought condition. The lowered An might have contributed to the decreased grain yield of the wheat genotypes due to source limitation. Osmotic adjustment (OA) is an important mechanism of plant adaptation under drought stress [46]. Moinuddin et al. [47] reported a positive association of OA with grain yield in wheat. In contrast, our results indicate that highest OA in L 2 did not associated a least yield reduction by drought. A high ability of plants to adjust osmotically under drought may help the plants survival during the stress which however, may reduce the yield as OA is often causing a metabolic cost [48]. In agreement with previous ndings of higher metabolic requirement for OA, we found a strong positive correlation of An with OA (Table 6) indicating that higher An could be contributing towards higher OA. HI describes the partitioning of photosynthates into reproductive parts in terms of dry mass. Higher An and HI was recorded in L 2 in comparison to other two genotypes ( Table 2 & Fig. 1c respectively) indicating an important role of photosynthates contribution towards the HI. Consistent with this, Earl and Davis [49] reported a reduced HI due to limited supply of photosynthetic active radiations in maize under drought conditions. The relationship between An and HI was further studied through activity of carbohydrate metabolic enzymes.
3.2. Correlation of HI with leaf ABA concentration and the activity of carbohydrate metabolic enzymes Primarily, decrease in the activity of carbohydrate metabolic enzymes was recorded in the leaf and increased activity was recorded in the spike except aldolase and invertases under drought conditions. Positive correlation of ABA was recorded with cwInv (Table 6) and similar results were reported by Ji et al. [50] in rice peduncle where higher concentration of ABA and higher activity vacInv was recorded under water stress conditions, indicating that higher concentration of ABA may have a role in regulating invertase activity. Several studies reported that subcellular metabolism of carbohydrates within plastids, cytosol and vacuole are involved in stress related responses [51] [52] . A higher activity of vacInv under drought conditions was reported by Yamada et al. [19] and in line with this, we also recorded higher activity of vacInv under drought. On the other hand, the increased activity of invertases in the leaf could result in accumulation of hexoses, which would contribute to more negative Y p as the case for L 3 in the present study (Table 4 & 3 respectively; Table 6). Similar correlation of Y p with cytoplasmic invertase was also recorded in genotype L 1 indicating the role of stored sugars acting as osmolytes under drought conditions (Table 6). Low HI and high vacInv activity in leaf and; high activity of vacInv in spike were recorded in genotype L 3 (Fig. 1c & 4 respectively).
Likewise, highest leaf cytInv and lowest HI was recorded in the same genotype while vice versa for the others. Roitsch and González [16] reported that the activity of vacuolar invertase regulates sugars translocation into reproductive parts under drought stress conditions. It is further indicating that sucrose was being hydrolyzed in the source hereby reducing translocation into the sink causing a lowered HI. Moreover, our experiment showed that correlations of invertase isoenzymes with HI (Table 6) were negative which are in good agreement with earlier ndings [53]. In conclusion, our results show that higher ABA accumulation correlates with the increased activity of invertase, which could be indicative for a function of ABA in regulating invertase activity. However, the increased invertase activities may not facilitate an increased HI in the studied genotypes. Instead, the liberated sugars are being utilized to decrease Y p contributing to OA in the plants.

Correlation between activity of carbohydrate metabolic enzymes and NGS
The association of NGS and TKW with the activity of key carbohydrate metabolic enzymes was further studied to elaborate the HI response. NGS were severely reduced under drought stress and highest KA was recorded in the drought sensitive genotype L 3 and lowest in the intermediate drought responsive genotype L 2 . Reduction in grain number under drought stress has frequently been reported in earlier studies [54] [55] . Cattivelli et al. [56] reported that drought severely affects meiosis at anthesis, which directly impacts grain number and ultimately the grain yield. Simkin et al. [57] reported that grain yield can be improved signi cantly by increasing the photosynthesis. Here, signi cant correlation of An was found with NGS and higher An and NGS were recorded in genotype L 2 in comparison to other genotypes (Table 6). Semenov et al. [58] also reported fewer grains due to decreased photosynthesis. The correlation of NGS with the activity of key carbohydrate metabolic enzymes was studied and negative correlation was recorded with vacInv of sink (Table 7). Higher activity of spike vacInv was recorded in genotype L 3 and this genotype was also having the lowest NGS. Yamada et al. [19] reported abiotic stress-inducible transporter for monosaccharides in Arabidopsis thaliana termed as ESL1 might function coordinately with the activity of vacuolar invertase to regulate osmotic pressure by affecting the accumulation of sugar in plant cells under drought conditions. It is further indicating that limited photosynthetic rates may force plants to utilize stored carbohydrates under severe drought conditions but in our study these carbohydrates were seemingly not utilized to enhance grain number. During glycolysis, sucrose is converted into glucose and fructose by invertases. The hexoses are further phosphorylated with the help of HXK and FK respectively [59]. In the present study, a decreased activity of HXK and FK was recorded in the leaf. No supporting literature is available to explain our ndings. However, Whittaker et al. [60] reported that higher activity of HXK in the leaves of Sporobolus stap anus could be responsible for drought tolerance. Likewise, Fulda et al. [61] reported that SlFRK3, a protein responsible for the activity of FK was upregulated in drought tolerant plants of sun ower under water de cit conditions. Karni and Aloni [62] also reported a decreased activity of FK in anthers under heat stress. These studies although reported in different plant species and tissues yet our studies and previous literature indicate the limited transport of sugars under drought conditions. This limited availability of sugars could induce seed abortion resulting in lower grain numbers. Below, change in HI under drought was further discussed in relation to the role of key carbohydrate metabolic enzymes in grain lling.

Correlation of carbohydrate catalyzing enzymes with TKW
Maintenance of higher TKW is necessary to produce higher grain yield under drought conditions in wheat. In the current experiment highest TKW and lowest HI was recorded in the drought sensitive genotype L 3 under drought conditions. Biplot analysis indicates a close association of TKW with leaf aldolase.; Individually, signi cant and positive correlations of leaf aldolase activity with TKW were noticed (Table 6). Aldolase has been reported to play a key role in physiochemical processes regulating plant development [63] and responses to abiotic stresses [64][65][66] [67] . A successive decline in the speci c activities of aldolase was reported under drought [23]. While, an overexpression of gene encoding leaf aldolase increased photosynthetic rate, enhanced growth and biomass production in tobacco plants [24]. In agreement with previous ndings, here a higher activity of leaf aldolase and higher TKW was observed in L 3 (Table 4 & Fig. 1d respectively). Additionally, Simkin et al. [68] reported that stimulation of sedoheptulose 1,7-bisphosphatase, fructose 1,6-bisphophate aldolase has improved photosynthetic e ciency as well as seed yield in Arabidopsis.. Likewise, role of different intermediate enzymes i.e. UGPase, which is the key enzymes for sucrose synthesis/breakdown [69], PGM, provides intermediate products of glycolysis and PFK, can regulate the glycolysis process through allosteric inhibition [70] was also evaluated. Negative correlations of TKW with spike UGPase, PGM and PFK were also found under drought conditions (Table 7). No supporting literature is available to con rm the results of the present study however, AGPase is reported to have positive correlation with grain llings [27] [29]. Maize and rice transgenes having Shrunken2 gene (Sh2r6hs), which encodes an altered AGPase activity showed increased the biomass and seed weight [71] [72]. Overexpression of the TaLSU I gene has signi cantly increased AGPase activity, which positively correlated with endosperm starch weight, grain number per spike and single grain weight [73], implying that the modi cation of the activities of these enzymes are associated with the grain lling process hereby in uencing the TKW.  (Table 5). Cummins et al. [37] and Roxas et al. [38] also reported an increase in the activity of GST under oxidative stress in transgenic tobacco. Diverging from previous reports, a decrease in the activity of GST was recorded with increasing NGS and signi cant negative correlation of leaf and spike GST with grain yield traits (  Fig. 1e). In contrast, a positive correlation of leaf MDHAR activity with the HI was noticed ( Table 6), implying that plants possessing a higher activity of MDHAR in source tissue would maintain redox homeostasis, which may enhance the resistance of photosynthesis to drought stress thus sustain the HI. Melandri et al. [74] reported higher DHAR activity could reduce drought-induced grain yield losses in rice. In addition, the activity of leaf MDHAR was positively correlated with NGS indicating that higher activity of this antioxidant in the source could have enhanced the drought tolerance of the wheat plants in sustaining the grain number (Fig. 1e), though the underlying mechanisms remain unknown. In line with our results, Sudan et al. [75] reported an increased MDHAR expression and enzyme activity under drought stress. Likewise, Sultana et al. [76] reported that overexpression MDHAR contributes to salt stress tolerance in rice. Eltayeb et al. [77] reported overexpression of MDHAR gene in tobacco is involved in osmotic stress tolerance under drought conditions.
In addition, a positive correlation of leaf cwPOX and aldolase with TKW was noticed in the present study (Table 6). A higher activity of POX under drought was reported by Veljovic-Jovanovic et al. [78] while work of Devi et al. [33] on wheat genotypes suggested a higher POX activity under drought helps plant to sustain grain yield. Our results are in-line with previous ndings supporting that higher activity of leaf cwPOX (Table 6) which may be the reason of less drop in TKW. These nding explains that plant could sustain NGS and TKW through maintaining the higher activities of MDHAR and cwPOX.

Conclusion
Results of this study showed that drought stress at anthesis depressed photosynthesis hereby reduced the source activity and photosynthate supply to the sink. This limited photosynthates supply could have caused reductions in NGS as well as TKW in wheat genotypes. Genotype L 1 maintained higher grain yield both under well-watered and controlled conditions mainly due to maintenance of higher NGS, RWC and Y p while genotype L 3 showed less grain yield mainly due to less RWC, Y p and higher KA under drought conditions. A high activity of aldolase, MDHAR, An, Gs and E in the source leaf might contribute towards sustaining carbohydrates remobilization from source to sink hence sustained NGS as well as HI (Fig. 3) while higher activity of aldolase and cwPOX enabled the plants to maintain a higher TKW. Under drought, a high activity of vacInv and GST in both source and sink may have contributed to osmolytes production which is indicated by less Y p and limited carbohydrate translocation from source and sink and higher utilization of incoming sugars by the sink may have negatively affected NGS and TKW. The ndings of this study provided some insights into the biochemical mechanisms regulating grain yield of wheat in response to drought stress and distinct tolerance to drought was predicted by physiological markers which could be used as important biomarkers for breeding drought tolerant wheat cultivars for a future drier climate.

Plant material and growth conditions
Three genotypes (L 1 , L 2 and L 3 ) of contrasting drought tolerance under eld conditions, developed at the International Maize and Wheat Improvement Centre (CIMMYT), Mexico and the Ayub Agricultural Research Institute (AARI) Pakistan, respectively, were selected. Genotype L 1 and L 3 were drought tolerant and drought sensitive, respectively, while L 2 was of intermediate drought response (Table 1). Four seeds were sown in 4 liters pots ( lled with peat material, Sphagnum, 32% organic matter, pH = 5.6-6.4 and EC = 0.45 mS cm -1 ) and only two seedlings were remained after one week of emergence by thinning.
Twenty-four replications for each genotype were grown under well-watered conditions. After few days of emergence automatic fertigation (irrigation + mixture of essential nutrients) was applied to the plants. Furthermore, weight of each pot was kept at the same water level by manual weighing of the pots. Temperatures for day/night were maintained at 22/16°C while photoperiod was kept at 16/8 hours day/night, respectively. Light conditions were maintained at 0.5 µmol photosynthetic active radiations (PAR) at night and 360 µmol PAR during the day.
Likewise, relative humidity was maintained at 55 to 60 %. At the time when all the plants reached to 50% owering as described using feekes' scale 10.3 [79], 4 replications of each genotype were harvested to study different agro-physiological parameters before applying drought stress. Remaining 20 replications of each genotype were divided into two sets: irrigation was withdrawn during anthesis for one set (10 pots) and water status of the other set (10 pots) was kept at 95% pot water holding capacity. Daily evapotranspiration (ET) of each pot was recorded by weighing. Total transpirable soil water was the change of between the pot weight at 95% water holding capacity (about 3.2 kg pot weight) and when evapotranspiration of the drought plants decreased to 10% of the well-watered plants (when pot weight was ca. 1.6 kg).

Leaf and spikes sampling
Stress was imposed at anthesis until all the plant available water in the pot was consumed. In genotype L 1 and L 2 the drought treatment lasted 9 days while in genotype L 3 the drought treatment lasted 8 days. At the end of the stress period, samples were taken from both well-watered and stressed plants. Two main tillers of each plant were selected for sampling. Flag leaf and attached spike from each of the tillers were taken and snap frozen in liquid nitrogen after tightly wrapping into aluminum foil. These samples were kept at -80 o C until the further use to analyze antioxidant and carbohydrate metabolic enzyme activities and osmotic potential. Then, 5 replications of each treatment were harvested to study their eco-physiology and dry biomass of the plants (Additional le1).

Gaseous exchange and plant water relations
Leaf photosynthetic rate (An, µmol m -2 s -1 ) and stomatal conductance (Gs, mol m -2 s -1 ) were determined from fully expanded ag leaves between 11:00 and 14:00 h with a portable photosynthetic system (LiCor-6400XT, Li-Cor, NE, USA). Measurements were performed at 20°C chamber temperature and 1500 mmol m -2 s -1 photosynthetic active radiation (PAR), and 400 ppm CO 2 concentration in cuvette. Relative water content (RWC) were determined in ag leaves according to the method by Jensen et al. (2000). The RWC was calculated as follow:
To measure osmotic potential (Ψ p ) of the plant tissue, frozen material wrapped in aluminum foil was thawed, squeezed, and a piece of lter paper was dipped into the obtained sap. Ψ p was determined using psychrometers (C-52 sample chambers, Wescor Inc., Logan, UT, USA) connected to a datalogger (Wescor's Dew Point Microvoltmeter, model HR-33T). Likewise, osmotic adjustment (OA) was recorded using following formula; OA = RWC (well-watered) x Ψ p (well-watered) -RWC (drought) x Ψ p (drought)

Extraction of samples for enzymes analysis
Samples extraction was done following the protocol by Jammer et al. [30]. Brie y, leaf and 10 spikelets from the middle of the spike excluding rachis Eleven carbohydrate metabolic enzymes were selected to check their activity within leaf and spike tissue. Dialyzed extract was used for the estimation of vacInv, cytInv, AGPase, UGPase, HXK, FK, PGM, PGI, PFK, Aldolase, and cell wall extract was used to determine the activity of cwInv.

Carbohydrate metabolic enzyme assays
Method described by Jammer et al. [30] was used to determine the activity of invertases. Concisely, 5 µl of the extract were added in at bottom 96-well plates to determine the activity of all invertases. While, 5 μl of 100 mM sucrose and 5 μl of reaction buffer pH 4.5 (454 mM Na 2 HPO 4 /273 mM citric acid) was added into dialyzed and cell-wall extract to determine the activity of vacuolar invertase (vacInv) and cell wall invertase (cwInv) respectively while reaction buffer with pH 6.8 (772 mM Na 2 HPO 4 /114 mM citric acid) was added into dialyzed extract to determine the activity of cytoplasmic invertase (cytInv). Sucrose was not added into control. Likewise, calibration curve was added by glucose standard (0-50 nmol). These plates were incubated at 37 °C for 30 minutes after adding the distilled water to raise the total reaction volume of 50 μl. Plates were put at room temperature for 20 minutes after removing from incubator. 200 μl of GOD-POD reagent (10 U ml -1 GOD, 0.8 U ml -1 POD, and 0.8 mg ml -1 ABTS in 0.1 M potassium phosphate buffer, pH 7.0 was added in each well. The absorbance was measured at 405 nm of plate reader. Principle of Sung et al. [80] was used to determine the activity of all the invertase enzymes. All remaining carbohydrate enzyme activities were determined using higher throughput method described by Jammer et al. [30]. For the activity of HXK and FK was determined following the principle of Petreikov et al. [81]. Moreover, 100 mM fructose, 50 mM NAD, 100 mM ATP, 3500 U ml -1 PGI, 1000 U ml -1 G 6 PDH (from Leuconostoc mesenteroides) and common buffer (composed of 1 M Tris HCl with pH 8.0, 0.25 M EDTA, 0.5 M MgCl 2 ) was used to determine the activity of FK. TPI was not used and 100 mM fructose was replaced with 100mM glucose to assay the activity of HXK. For the activity of UGPase and AGPase, principle of Pelleschi et al. [8] and Appeldoorn et al. [82] was used. Again, glucose and fructose were omitted from the control. For the activity of AGPase and UGPase, common buffer, 10% BSA, 100 mM Na-PPi, 10 mM NADP, 50 mM 3-PG, 1.28 U ml -1 G6PDH from Saccharomyces cerevisiae, 1000 U ml -1 PGM, 50 mM ADP-Glucose (for AGPase) and 100 mM UGP-glucose (for UGPase), was used to determine the activity of AGPase and UGPase respectively. However, for the control samples ADP-glucose and UDP-glucose were omitted. Similarly, principle of Manjunath et al. [83] was used to determine the activity of PGM. In continuation, to assay the activity of PGM 1 M Tris-HCl pH 8.0, 0.5 M MgCl 2 , 500 mM DTT, 10 mM Glc-1,6-bisP, 100 mM Glc-1-P*, 10 mM NADP, 6000U ml -1 G 6 PDH (from S. cerevisiae) was used. Activity of PGI was determined following the principle of Zhou and Cheng [84]. However, to determine the action PGI 10 mM glc-1,6-bisP and 100 mM glc-1-P*, were replaced with fruct-6-P* and; glc-1-P* and fruct-6-P*. Mastermix was prepared using common buffer, 25 mM fruct-1,6-bisP*, 25 mM NADH, GPDH 2100 U ml -1 , TPI 6000 U ml -1 . Additionally, activity of PFK was determined following the principle of Klotz et al. [85]. Similarly, apart from common buffer, 100mM fruct-6-P*, 25 mM NADH, 100 mM ATP, 372 U ml -1 aldolase, GPDH 2100 U ml -1 , TPI 6000 U ml -1 was used for the activity of PFK. Fruct-6-P* was omitted as substrate in the control samples. and activity of aldolase was determined following the principle of Schwab et al. [86]. The absorbance was studied at 340 nm for 30 minutes and deviation of readings/peaks was monitored during this period and calculation of speci c enzyme activity in nkat g FW -1 . Gen5 v3.04.17 software (Biotek Instruments. Inc) was used to measure the absorbance of different enzymes.

Activity of antioxidants enzymes
Methodology described by Fimognari et al. [87] was used to determine the activities of different antioxidant and 96-well plates format was utilized while, the activities were determined photometrically. Brie y, activities for ascorbate peroxidase (APX) was determined based upon the principle of Yoshimura et al. [88].
For the reactions, dialyzed extract was used. Master mix comprised of 50 mM KPO 4 buffer pH 7.6, 0.25 mM ascorbate and 0.5 mM H 2 O 2 was used and absorbance was recorded at 290 nm. Likewise for control H 2 O 2 was omitted [89]. For the activities for catalase (CAT) principle of Aebi [90] was followed.
Master mixed containing 50 mM KPO 4 buffer pH7, 0.001% antifoam agent 204 and 100 mM H 2 O 2 was mixed with dialyzed extract and absorbance was recorded at 240nm. Likewise, for control reactions H 2 O 2 was omitted as mentioned by Fimognari et al. [87]. To determine the activity of peroxidase (POX) or cell wall peroxidase (cwPOX) principle of Polle et al. [91] was used. For determination of POX activity method described by Garcia-Lemos et al. [89] was used.
Again, dialyzed extract was mixed with master-mix containing 100 mM KPO 4 buffer pH 7, 2 mM guaiacol and 0.15 mM H 2 O 2 was used. Absorbance was measured at 450 nm and H 2 O 2 was omitted for control reactions. However, cell wall extract was used for the activity of cwPOX. The activity of superoxide dismutase (SOD) were determined following the principle of McCord and Fridovich [92]. Similarly, dialyzed extract was used to mix with master mix containing 50 mM KPO 4 buffer pH 7.8, 0.1 mM EDTA, 0.05 mM cytochrome c, 10 mM xanthine and 0.0002 U mg -1 xanthine oxidase. Absorbance was recorded at 550nm as described by Fimognari et al. [87]. However, xanthine was omitted in control reactions. Activities of glutathione reductase (GR) principle of Edwards et al. [93] was used. Dialyzed extract was mixed with master mix containing 100mM buffer of Tris HCl with pH 7.8, 25 mM NADPH and 30 mM glutathione oxidized (GSSG). Absorbance was detected at 340nm for 40 minutes and GSSG was omitted for control reactions. For the activity of dehydroascorbate reductase (DHAR) principle of Dalton et al. [94] was followed. Again, dialyzed extract was mixed with master mix comprised of 100 mM KPO 4 with pH 6.5, 50 mM glutathione reduced (GSH) and 50 mM dehydroascorbic acid (DHA). The activity was determined at 290 nm for 40 minutes and DHA was not used in control reactions. To determine the activity of monodehydroascorbate reductase (MDHAR) principle described by Arrigoni et al. [95] was followed. Dialyzed extract was mixed with reaction mixture comprised of 50 mM KPO 4 buffer with pH 7.2, 25 mM NADH, 5U µl -1 ascorbic acid oxidase (OAA) and 50 mM ascorbate.
Activity was measured at 340 nm for 40 minutes and ascorbate was omitted in control reactions. Additionally, the activities of Glutathione S-transferase (GST) were determined following the principle of Li et al. [96]. Again dialyzed extract was mixed with reaction mixture (100 mM KPO 4 buffer with pH 7.4, 50 mM GSH and 2,4-dinitrochlorobenzene (CDNB)). Absorbance was measured at 334 nm for 30 minutes and CDNB was not used for control reactions.

Abscisic acid assay
ABA concentration in leaf and spike samples was determined through an enzyme linked immunosorbent assay (ELISA) using a monoclonal antibody for ABA (AFRC MAC252) (Asch, 2000).

Agronomic traits measurement
At the end of drought treatment pots from each treatment were re-watered until the maturity of the plants. Plant maturity stage was determined as described by Zadoks et al. [79]. Harvesting was done at maturity and following traits were recorded: a)number of grains spike -1 (NGS): Five spikes from each replication were taken and their averages were recorded; b) thousand grain weight (TKW): Thousand grains were counted from each replication and their weight was recorded in grams (g); c) Kernel abortion (KA): KA was recorded using following formula: (number of grains spike -1 /number of orets spike -1 ) x 100 d) Plant biomass pot -1 (BM): Both plants from each pot were harvested from soil level and weight of whole plant was expressed in grams (g); e) Grain yield pot -1 (GY): Spikes from the pot were threshed into grains and their weight was expressed in grams (g); f) harvest index (HI) were recorded using following formula: Grain yield (g)/Biomass (g) x 100

Statistical analysis
Analysis of variance (two-way ANOVA) was done using RStudio 1.0.153.exe to reveal the signi cance of the effect of genotype, water and their interaction on the measured variables at P= 0.05 level. Function PerformanceAnalytics was used to do correlation analysis while devtools, factoextar and fviz_pca_biplot were used to do principal component analysis and to draw biplot between principal component 1 and principal component 2 "L" and "S" are indicating leaf and spike antioxidants or carbohydrate metabolic enzymes or phytohormones in biplot gure like, "L-aldolase" was used for leaf aldolase enzymes and "Saldolase" was used for spike aldolase enzyme.

Additional File
Additional le 1. Schematic diagram of experiment, water consumption of different genotypes during stress and sampling from different treatments and rewatering until maturity Tables   Table 1 Parentage Table 6. Combined correlations of leaf carbohydrate metabolic and antioxidants activity with leaf water relations, gaseous exchange, abscisic acid, and with yield and yield contributing traits   Biplot of PC1 and PC2 derived from PCA analysis under well-watered (A) and drought condition (B) Pre x "L" is indicating leaf antioxidant or carbohydrate metabolic enzymes or phytohormones and pre x "S" is indicating spike antioxidant or carbohydrate metabolic enzymes or phytohormones