**4. Discussion**

We investigated the e ffect of a drought-stress QTL introgressed into the elite rice line KDML105. Using leaf water potential under drought stress, we found that introgression line CSSL104 manifested drought tolerance similar to that of the parent line DH103. Both tended to have a better ability to maintain water status in the first fully expanded leaves. After six days of drought stress, both lines had about a 22% higher leaf water potential than the KDML105 parent. Drought stress limits water uptake from the rice root and reduces water availability in the cells, which is critical for survival under drought stress. Water depletion can compromise photosynthesis and cell growth [34], and three main maintenance mechanisms are used by plants to o ffset water loss: leaf rolling, stomatal closure, and osmoregulation [35,36]. Evidence for the drought tolerance of CSSL104 was also provided by the lower leaf-death score and higher Fv/Fm displayed by CSSL104 compared to KDML105 (Figure 1).

Drought stress resulted in decreased stomatal conductance in all lines (KDML105, DH103, and CSSL104; Table 1). This was a water-preservation mechanism that also resulted in a decline in the net photosynthesis rate. After three days of drought stress, the photosynthesis parameters, net photosynthesis rate, stomatal conductance, transpiration rate, ΦPSII, electron transport rate, intercellular CO2 concentration, and Fv'/Fm', were similar among all lines. After six days under drought stress, the net photosynthesis rate of DH103 and CSSL104 were about twofold higher than the net photosynthesis rate of KDML105. In comparison with normal plants, KDML105 rice had a nearly 70% reduction in photosynthesis rate, while DH103 and CSSL104 had only a 46 and 44% reduction, respectively. Interestingly, the ΦPSII and electron transport rate of all lines were similar, while the stomatal conductance of KDML105 was 58% lower than DH103. These findings sugges<sup>t</sup> that stomatal closure could be one of the major factors contributing to the decline in the net photosynthesis rate of KDML105. Although the stomatal conductance of CSSL104 was lower than DH103 by 42%, CSSL104 could maintain a net photosynthesis rate (Table 1). These indicated that this CSSL is better adapted than its drought-tolerant parental line. It is possible that a KDML105 locus contributed to maintenance of the photosynthesis process through an epistatic interaction with the introgressed DH103 region.

Using whole genome sequence comparison and co-expression network analysis, we characterized the molecular fingerprint of the introgression. The first detected DNA segments introgressed, while the second identified genes connected with the drought response. During this stress, 18 genes were highly co-expressed with other genes (Figure 3A and Table 2). Nine of them (*CPFTSY*, *NDH-O*, *SOQ1*, *LHCB3*, *RRF*, *PGRL1*, *HCF244*, *NAD(P)-linked oxidoreductase,* and *MRL1*) are involved in the photosynthesis process, and CPRF1 is essential for chloroplast development. These findings indicate that the drought-adaptation QTL a ffects photosynthetic genes whose modulation maintains the net photosynthesis rate of CSSL104.

The function of the identified genes is as follows. CPFTSY (chloroplast FtsY, i.e., chloroplast signal-recognition particle) is required for light-harvesting chlorophyll *a*/*b*-binding protein (LHCP) integration into thylakoids [25], and NDH-O is the subunit required for the NADH dehydrogenase-like (NDH) complex assembly that functions in cyclic electron flow [25,37]. SOQ1 is required to maintain light harvesting e fficiency especially during nonphotochemical quenching (NPQ) recovery [26]. Light-harvesting chlorophyll (LHC) functions as a light receptor to capture light energy and deliver it to photosystems. The *Lhcb3* gene product regulates the rate of state transition by changing the excitation energy transfer and charge separation [38]. RRF is a ribosome recycling factor in chloroplast [39]. *RRF* is required to maintain photosystem II e fficiency (Fv/Fm) and proper stacking of the internal membranes of chloroplast. Loss of these functions led to a lower growth rate for the *rrf* mutant compared to WT [28], which is consistent with the phenotype documented in this study (Figure 4A). *PGRL1A* (*AT4G22890*) and *PGRL1B* (*AT4G11960*) are paralogous genes whose products switch linear electron flow to cyclic electron flow. PGRL1 is the elusive ferredoxin-plastoquinone reductase (FQR) [29].

During drought stress, plants shift the electron transfer route from linear to cyclic to balance the energy flow from light reaction to Calvin cycle and photorespiration [37,40,41]; this change is due to CO2 limitations caused by stomatal closure. It was recently shown that tomato with co-silencing of the *PGR5*/*PGRL1A* gene was more susceptible to cold stress [42]. This is consistent with our finding that the *pgrl1* mutant had a significantly lower growth rate than WT in both intermediate and severe drought stress (Figure 5B,C). HCF244 is required for the biogenesis of photosystem II (PSII), and specifically for the synthesis of the reaction center proteins [30–32]. The NAD(P)-linked oxidoreductase gene is involved in redox reactions, but the details are still unclear. *MRL1* is the only gene in this study whose product is involved in the Calvin cycle. It is involved either in the processing or in the stabilization of the large subunit (LS) of *RuBisCO* transcripts [43].

Not surprisingly, impairment of these photosynthesis genes significantly reduced growth under normal conditions (Figure 4A). Moreover, we could not obtain homozygous lines of the *soq1*, *hcf244*, and *cprf1* mutation, probably because the homozygous mutants are lethal. However, under drought-stress conditions, some of the mutants in the genes involved in the light reaction process (*ndhO* (*at1g74880*), *lhcb3* (*at5g54270*), *rrf* (*at3g63190*), and *at2g27680* mutants) displayed significantly higher growth than WT. These responses suggested that the decrease in light energy harvest during drought stress could prevent damage to chloroplasts and prolong survival of photosynthetic tissues. The *mrl1* mutant was expected to lack of the ability to stabilize the *rbcL* mRNA, which could result in a decline in Calvin cycle activity. The higher stability index of the *mrl1* mutant line under intermediate and severe drought stress indicated that a slower rate of the Calvin cycle may help plants cope with drought stress.

Based on the growth phenotype under drought stress, the *ndhO*, *lhcb3*,*rrf*, and *mrl1* mutants showed a higher growth rate than WT (Figure 4B,C). The homologs in rice of these genes are *LOC\_Os01g72800*, *LOC\_Os07g37550*, *LOC\_Os07g38300*, and *LOC\_Os10g10170*, respectively. Therefore, we would like to propose that these genes contribute to drought-tolerance regulation in rice by mediating adaptation in the photosynthesis process. *LOC\_Os01g72800* is located in the previously reported drought-tolerance QTL between RZ14 and R117 [21], while the other three genes are not. Collectively, our results indicate that the combination of SNP analysis with co-expression network analysis is a suitable method for drought-tolerance gene prediction. This approach will help future exploration to identify the candidate genes for abiotic stress tolerance.
