The authors declare that the transgenic experiments described herein comply with the current biosafety laws of the country in which they were performed.

The isolation and cloning of the AlSAP gene from A. littoralis was described previously (Ben Saad et al., 2011, 2012). The AlSAP gene and protein sequences are deposited in the GenBank (DQ885218) and UniProtKB (A1YAQ3) databases. The rice genome contains 12 SAP genes encoding A20 and AN1 domains. The relationship between AlSAP and OsSAP proteins was established by Ben Saad et al. (2010); the SAP gene encoding the protein sharing the highest amino acid identity (79%) is OsSAP9 (Os07g07350) which is the likely ortholog of AlSAP in rice (Supplementary Figure S1).

We evaluated six AlSAP-lines and their NT azygous controls in our field trials (Supplementary Figures S2, S3). RN4, RN2, and RN5 (T5 homozygous lines) expressed AlSAP under control of the 35S CaMV promoter and accumulated AlSAP transcripts at increasing levels (Ben Saad et al., 2012). RN1 and RN3 (T3 homozygous material) expressed AlSAP under the control of the maize Ubiquitin1 regulatory region (Christensen et al., 1992). RN6, also a T3 homozygous material, expressed AlSAP under the rice ACT1 gene regulatory region (Mc Elroy et al., 1990). T3 homozygous RN3, RN1, and RN6 plantlets accumulated AlSAP transcript at moderate, medium, and high levels in leaves, respectively, at the six leaf stage under greenhouse conditions. Each of the three lines harbored a single copy of the transgene (Supplementary Figure S3).

In Trial 1, six homozygous AlSAP-transformed lines (RN1–6), together with their transformation controls, and non-transformed WT-NIP plants were evaluated in a natural upland rain-fed field from November 2012 to March 2013 at CIAT Santa Rosa station (Villavicencio, Colombia). The transgenic and WT-NIP lines were sown in three replicate 2 m × 1 m plots with a spacing of 25 cm × 10 cm and received regular fertilization. The plants experienced a 42-day rain-free period, which began at panicle initiation and extended for approximately 28 days after attainment of 50% flowering (Supplementary Figure S5).

Trials 2 and 3 were performed in the rainout shelter facilities at CIAT headquarters (Palmira, Colombia); the drought treatment was imposed at panicle initiation stage by suspending irrigation. In Trial 2 (April–July 2013), RN2, RN5, and RN6 were evaluated, together with their NT controls and WT-NIP, while in Trial 3 (June–October 2013) only RN5 and RN6 were assessed, together with their NT controls, and WT-NIP. Lines were sown in a randomized complete block design, with three replicates. Each plot had six rows of 16 plants with a spacing of 20 cm × 10 cm. WT-NIP plants were inserted after every four transgenic line plants for comparison. Drought was imposed for 22 days (Trial 2) and 19-days (Trial 3), after which the plants received full irrigation.

For both Trials 2 and 3, a duplicate set of materials were also sown under flooded conditions in another field, 5 m away from the rainout shelter, to evaluate differences in GY between transgenic and control rice under irrigated conditions.

Harvest was completed for all trials 107–114 days after sowing (DAS).

The progress of drought-stress was monitored by measuring changes in soil moisture at four depths, 20, 40, 60, and 80 cm, using Aquapro sensors (Reno, NV); polycarbonate access tubes were installed at regular intervals within each replicate, and readings were taken regularly during the stress period. Humidity data were transformed to soil matric potential (Ψ_soil) values using soil water retention curves (SWRC) built for each location. Aquapro soil moisture values (%) were converted to volumetric soil water content using the equation:

where 𝜃_vol is the volumetric soil moisture content (%), A is the Aquapro measurement value (Aquapro %), and b is a constant depending on soil type (Tromp-van Meerveld and McDonnell, 2006). SWRC were obtained experimentally using the pressure plate apparatus method (Bittelli and Flury, 2009).

Single plant yield and yield components were measured in each plot for all trials. Yield parameters were scored for 15 plants. Unfilled and filled grains were independently counted and weighed. Grain weight was used to determine final yields at 14% moisture content. The following agronomic traits were scored as established in the Standard Evaluation System for Rice (IRRI, 2013): flowering date, plant height (cm), total aerial biomass (g), panicle length (cm), number of tillers, number of panicles, total grain weight (g), and grain fertility (%).

The rate of light-saturated photosynthesis (A_sat), stomatal conductance (G_s), and intercellular CO₂ concentration (C_i) of expanded flag leaves were measured from 9:00 to 12:00 h using an LI-6400 portable photosynthesis open system (Li-Cor Biosciences, Lincoln, NE, United States). Measuring conditions were as follows: photosynthetic photon flux intensity = 1500 μmol/m²/s, leaf temperature = 27.0°C ± 1°C, ambient CO₂ concentration = 380 ± 10 μmol/mol, and relative humidity = 60%. The rate of CO₂ saturated photosynthesis (A_max) was measured under the conditions described above, except that the ambient CO₂ concentration was 1800 μmol/mol.

Leaf rolling (LR) was characterized at regular intervals in drought-stressed plants using a scoring scale from 0 (unrolled) to 9 (tight rolling) according to the Standard Evaluation System for Rice (IRRI, 2013). Leaf drying (LD) was evaluated at the peak of stress using a scoring system from 0 (no symptoms) to 9 (dead plant). The drought recovery score (DRR) was recorded 10 days after rewatering using a scoring system from 1 (90–100% recovery) to 9 (0–19% recovery) (IRRI, 2013). All evaluations were performed at noon using whole plants.

On three different occasions during Trial 2 (before stress, at stress peak, and 10 days after rewatering), Ψw was evaluated (pre-dawn) in drought-stressed plants. Measurements were performed on fully expanded flag leaves, as previously described (Turner, 1981), using a pressure chamber (Soil Moisture Equipment 3005F01) with a measurement limit of -15 MPa.

The antioxidant capacity was determined using the ABTS^∙+ test system, based on the method of (Du et al., 2009). The level of lipid peroxidation in leaf tissues was measured as MDA content, following a published protocol (Ueda et al., 2013).

Procedures for Southern and northern blot analyses, and RT-PCR were the same as those detailed by Ben Saad et al. (2012). Total DNA and RNA were isolated from leaf tissue of six leaf stage seedlings, and hybridizations performed using P³² labeled Npt2, Hpt, or AlSAP gene probes. RT-PCR analyses were performed using RNA isolated from leaves and primers specific to AlSAP and the housekeeping gene, OsExp (Caldana et al., 2007), used as an internal expression standard (Supplementary Table S1). For samples from field grown plants, inflorescences were collected from the master tiller of three independent plants per line at flowering stage, pooled and stored. RNA isolation and northern blot procedures were performed as indicated above (Ben Saad et al., 2012).

Data for days to flowering, height, total aerial biomass, GY, yield components, leaf MDA, and TEAC-ABTS were subjected to General Linear Model (GLM) analysis, appropriate for a randomized complete block design. Differences between mean values from WT-NIP- and AlSAP-transformed lines were compared using Dunnett’s test. A GLM was also used to test the effect of drought treatment (days after the last irrigation), and repetition on the soil matric potential (Ψ_soil) measured at a depth of 40 cm in Trials 2 and 3. All statistical analyses were performed using SAS v9.2 (SAS Institute Inc., Cary, NC, United States). Data (means ± standard errors of raw data) are presented graphically using GraphPad Prism (GraphPad PRISM for Windows v.7.0, GraphPad Software Inc., La Jolla, CA, United States).

Drought intensity and duration varied among trials with harsher conditions in Trial 1 (Santa Rosa station). In this trial, drought progressed rapidly in the first 20 cm of soil, with Ψ_soil falling under the permanent wilting point (PWP, -1.6 MPa) 10 days after stress initiation (DASI). Changes in the 40 cm layer proceeded more slowly, with Ψ_soil reaching -1.6 MPa by day 22. Thereafter, Ψ_soil decreased steadily until the first rains occurred. A reduction in Ψ_soil was also observed at 60 cm depth; however, minimum values were approximately -1.0 MPa by the end of the rain-free period (Figure 1A). In the rainout shelter in Trial 2 (Palmira station), Ψ_soil in the 20 cm soil layer also fell below the PWP 10 DASI and reached very low values during the drought period (Figure 1B). However, Ψ_soil changes at depths of 40 and 60 cm were less pronounced than those in Trial 1. Drought developed in a similar fashion in Trial 3 (Palmira), although the overall stress intensity was less severe than that in Trial 2 (Figure 1C). No significant differences in Ψ_soil were detected between replicates in any trials, indicating that drought-stress progressed uniformly among replicates (Supplementary Table S2).

The Nipponbare root system normally develops in the first 40 cm of soil (Hanzawa et al., 2013). Thus, we considered changes in Ψ_soil at depths of 0–40 cm good indicators of the level of drought-stress experienced by plants in the field.

In Trial 1, six homozygous AlSAP-transformed lines (RN1, RN2, RN3, RN4, RN5, and RN6) (generations T3–T5) were evaluated together with their azygous NT controls and WT-NIP. RN2, RN4, and RN5 harbor the AlSAP coding sequence under the control of the CaMV35S promoter (Ben Saad et al., 2012), while RN1, RN3, and RN6 contain one copy of a T-DNA containing the AlSAP coding sequence directed by the maize Ubi-1 promoter or the rice ACT1 promoter (Supplementary Figures S2, S3). All materials reached 50% flowering around 45–47 DAS, indicating that AlSAP accumulation did not affect plant phenology. The accumulation of AlSAP transcripts was monitored by northern blot analysis in inflorescence tissues collected from field grown plants at the initiation of drought stress imposition (Supplementary Figure S4). AlSAP transcript accumulation in these tissues was comparable in RN5 and RN6 and lower in RN2. On the other hand, the level of accumulation in RN1, RN3, and RN4 was very low compared to their respective levels in leaves of young plants at the previous generation. This was particularly surprising for the RN1 line which was identified as a medium expresser. Previous literature report describes high activity of the maize Ubi-1 regulatory regions in reproductive organs of rice (Cornejo et al., 1993). This may tentatively be ascribed to a regulation of the maize ubiquitin promoter under field conditions or to a silencing of the transgene.

In this trial, the plants were subjected to 42 days of drought stress from panicle initiation to 28 days after the plants reached 50% flowering, which caused a marked reduction in Ψ_soil. Under these severe stress conditions, RN2, RN5, and RN6 were able to maintain their green biomass (Supplementary Figure S6) and exhibited higher GY than WT-NIP and NT controls, with relative gains of 46, 97, and 115%, respectively, compared with NT controls (Figure 2A). The NT controls tended to be more affected by drought stress than WT-NIP and, although this difference was not significant, it was consistent. However, the RN1 and RN3 NT control lines exhibited significantly lower GY (p < 0.001) than either the WT-NIP or NT controls of the other transformation events following the field drought-stress. This depressive effect could be due to somaclonal variation. We also see previously that these two lines eventually expressed AlSAP at a low level under field conditions. Interestingly, a beneficial effect of AlSAP expression was still observed in these variants; however, this was not sufficient to cause yields significantly higher than those of WT-NIP. The four other lines could be grouped into two classes according to their performance relative to WT-NIP, and their level of accumulation of AlSAP transcripts. The low level of transcript expression in the RN4 line did not result in enhanced yield protection under drought stress at flowering, whereas the medium and high expression levels found in RN2, RN5, and RN6 conferred enhanced tolerance to these lines. Therefore, lines RN1, RN3, and RN4 were not included in subsequent drought trials.

Trials 2 and 3 were conducted using a rainout shelter to impose drought-stress. In Trial 2, RN2, RN5, and RN6 consistently achieved higher GYs than their NT controls, exhibiting GY increases of 57, 92, and 69%, respectively (Figure 2B). RN5 and RN6 also exhibited superior GY in Trial 3 (Figure 2C). The higher GYs were associated with significantly higher numbers of panicles (p < 0.001), higher fertility (p < 0.001), and higher accumulation of biomass in AlSAP-lines (p < 0.001) (Table 1). All lines flowered around the same date, 53–55 DAS; thus, they were at very similar physiological stages when subjected to drought (Table 1).

In Trial 2, lines RN1–6 were also grown under irrigated conditions, along with their NT controls and WT-NIP to evaluate the effect of AlSAP accumulation in the absence of drought-stress. Interestingly, under irrigated conditions, all NT controls behaved in the same manner as WT-NIP, while the transgenic lines tended to outperform both types of controls (Figure 2D). This indicates that the depressed background represented by the RN1 and RN3 NT controls influence performance only under drought conditions and, could, therefore, affect stress-related traits. Irrigated plots were also established in the third field trial. Under irrigation, the accumulation of AlSAP significantly enhanced GY (p < 0.001). As observed under drought-stress, this effect was related to both higher numbers of panicles (p < 0.001) and accumulation of vegetative biomass (p < 0.001) (Table 1).

Together, these results demonstrate that AlSAP expression increases GY under drought-stress imposed at the reproductive stage, through a mechanism that involves the maintenance of vegetative biomass, a higher number of panicles, and enhanced grain fertility. Furthermore, AlSAP-lines also exhibited higher yields, related to higher panicle numbers and increased vegetative biomass, when grown under irrigation.

In Trial 2, we compared the drought sensitivity of AlSAP-lines, their NT controls, and WT-NIP by monitoring the progress of LR, LD, and DRRs (IRRI, 2013). LR appeared earlier and proceeded more rapidly in WT-NIP plants, reaching a score of 5.5 at the peak of stress (22 DASI) (Figure 3A), indicating that leaves were fully cupped. At this time, LR scores for RN5 and RN6 were 3.0 and 3.6 (leaves folding), respectively. At peak stress, the LD score for WT-NIP was 3.2 (drying up to ¼; of the length of the leaf), while RN5 and RN6 scores were 1.6 (slight tip drying) (Figure 3B). DRR was evaluated 10 days after rewatering, RN5 and RN6 showed better recovery (DRR = 1.7) than WT-NIP (DRR = 3.0). These results indicate that the best-performing AlSAP lines were much less sensitive to drought and exhibited improved recovery after rehydration compared with controls (Figure 3C).

In rice, the ability to maintain ψ_w at the flowering and grain filling stages is related to variation in GY under drought-stress (Sibounheuang et al., 2006). Furthermore, there is a negative correlation between ψ_w and the percentage of spikelet sterility in rice plants subjected to drought at flowering (Jongdee et al., 2002). Therefore, in Trial 3 we evaluated whether the GY differences observed among RN5, RN6, and WT-NIP were related to an improved ability to maintain higher ψ_w.

No significant differences in ψ_w were detected before the imposition of drought stress, with ψ_w values oscillating between -0.15 and -0.50 MPa (Figure 4). After 19 days under drought-stress, ψ_w decreased significantly in all lines, reaching values of -0.90 to -2.10 MPa. RN6 plants showed slightly, but not significantly, higher ψ_w than RN5 and WT-NIP. Rewatering of plants resulted in ψ_w recovery to pre-drought values, with no significant differences among lines (Figure 4).

Flag leaves of AlSAP-lines and WT-NIP showed a marked reduction in A_sat under the stress conditions imposed in Trial 2 (Figure 5A). Plants exhibited A_sat rates of approximately 30 μmol/m²/s before the imposition of drought; however, rates were reduced by approximately 70% at the peak of drought stress. The reduction in A_sat was matched by a strong diminution in G_s and C_i (Figures 5B,C), indicating that stomatal closure intensely limited photosynthesis.

To establish whether the reduction in A_sat was exclusively due to diffusive limitations, we evaluated the photosynthetic responses of RN5 and WT-NIP under a saturating CO₂ concentration (A_max) at regular intervals during the drought period and 10 days after rewatering (Figure 6A). Under irrigated conditions, A_max was approximately 46 μmol/m²/s; however, it reduced progressively during drought reaching values of 21 μmol/m²/s, at the peak of stress. After rewatering, A_max did not fully recover to pre-drought levels. These results indicate that the reduction of photosynthesis also involved biochemical damage. The similar responses exhibited by RN5 and WT-NIP suggest that AlSAP expression did not offer protection to the photosynthetic machinery under the high stress levels imposed in Trial 2.

We also evaluated A_max under the milder stress conditions imposed in Trial 3. RN6 was also included in this evaluation. Similar to Trial 2, we observed a progressive diminution of A_max in all lines as drought-stress intensified (Figure 6B). However, A_max was reduced to a lesser extent in AlSAP-lines, and recovered almost entirely after rehydration. At the peak of stress, A_max in AlSAP-lines was reduced by 24% compared to the initial rate, while in its NT control A_max diminished by 52%. After reirrigation, AlSAP-lines showed a 95% recovery of A_max; however, the NT control attained only 59%. Thus, it appears that AlSAP may exert a “protective” effect on photosynthesis, specifically at the biochemical level, and that this effect is modulated by the intensity of drought-stress.

Previous work on AlSAP-transformed rice demonstrated that detached leaves of AlSAP-lines maintained high chlorophyll content than controls after exposure to high concentrations of H₂O₂, indicating tolerance to oxidative stress. Additionally, AlSAP-lines showed enhanced expression of the genes OsSodA and OsAPX1, encoding the enzymes manganese superoxide dismutase and ascorbate peroxidase, respectively, which are involved in oxidative protection (Ben Saad et al., 2012).

Based on these results, in Trial 3 we evaluated whether a superior capacity for removal of reactive-oxygen-species (ROS) was involved in the enhanced drought tolerance observed in AlSAP-lines. We quantified the content of MDA in leaves to monitor the degree of oxidative stress induced by drought (Figure 7). Leaf MDA concentrations increased in response to drought in all lines and continued to rise after rehydration (p < 0.01). RN5 and RN6 exhibited significantly higher MDA content than NT controls before the drought, at the peak of drought-stress, and 10 days after rewatering (p < 0.01). No association was observed between the concentration of MDA and chlorophyll content (data not shown). All plants exhibited decreases in chlorophyll content during the experiment; however, no significant differences were observed among lines (data not shown).

We also evaluated the capacity of plants to remove ROS using the ABTS^•+-based system, which estimates total antioxidant capacity (Tirzitis and Bartosz, 2010). No significant differences were observed among lines at any time point (Figure 7); however, a marked increase in antioxidants was observed in response to drought. The antioxidant capacity decreased after rewatering, although it remained higher than that observed before stress, possibly due to stress priming (Wang et al., 2014).

In this investigation, we characterized the field performance of rice lines (O. sativa L. ‘Nipponbare’) transformed with the AlSAP gene isolated from the halophyte grass A. littoralis. Our results demonstrate that the constitutive expression of AlSAP did not involve growth, development or phenology trade-offs, and confirm a role for AlSAP in enhancing rice tolerance to drought stress.

AlSAP expression improved the tolerance of rice to drought stress imposed at the reproductive stage, and this effect translated into higher GY. In our three independent field trials, AlSAP-lines achieved higher GYs than NT and WT controls when exposed to drought-stress of different intensities, complementing previous findings by Ben Saad et al. (2012) indicating superior drought tolerance at the vegetative stage. The higher GY corresponded with a significant increase in panicle number and spikelet fertility. No differences were observed in the number of spikelets per panicle, nor in individual grain weight. We disregarded the hypothesis that differences in panicle number and spikelet fertility were due to phenology, as all the lines flowered at around the same time and drought stress began at heading, when the number of panicles was already determined. In fact, we observed no difference in panicle number between plants grown under irrigated and drought conditions for either WT-NIP or AlSAP-lines.

For RN2, RN4, and RN5, the same ranking in stress tolerance (RN5 > RN2 > RN4) correlating with levels of AlSAP transcript accumulation was observed under field drought conditions and previous laboratory and greenhouse experiments (Ben Saad et al., 2012). The RN6 line, harboring the AlSAP coding sequence under the control of the rice ACT1 promoter, performed similarly to RN5. In contrast, somaclonal variation in Ubi1 NT controls, led to inferior performance under drought conditions. This depressed background likely prevented expression of the full benefit of a low level of AlSAP accumulation in the transgenic lines. Variation among plants derived from tissue culture is an extensively reported phenomenon, including in transgenic lines tested for drought-tolerance (Xiao et al., 2007, 2009), and has both genetic (Miyao et al., 2012) and epigenetic (Stroud et al., 2013) origins. These reports stress the requirement for evaluation of homozygous transgenic lines alongside their azygous, sibling progenies.

Photosynthesis and yield are positively correlated in rice under both irrigated and drought conditions (Centritto et al., 2009; Gu et al., 2012). However, the GY differences observed between AlSAP-lines and WT-NIP under drought-stress were not clearly associated with A_sat in flag leaves. This result contrasts with those obtained by Ben Saad et al. (2012), which demonstrated enhanced photosynthetic performance in RN5 under mild and severe drought applied at the vegetative stage, a response engendered by the capacity of this transgenic line to maintain elevated G_s. Similar to that report, we found that stomatal closure limited A_sat under drought conditions; however, we did not detect an improved capacity to maintain open stomata in AlSAP-lines. Furthermore, evaluation of A_max during drought and after rehydration indicated that similar non-stomatal limitations operate in AlSAP-lines and WT-NIP under severe stress conditions.

Although we did not observe differences in instantaneous A_sat on a single leaf basis, it is still possible that AlSAP-lines may have achieved improved photosynthetic performance. RN5 and RN6 exhibited moderate LR, which can increase photosynthesis and GY in rice (Zhang et al., 2009; Zou et al., 2011). LD was also lower in AlSAP-plants, and they were able to recover their vegetative biomass almost entirely after drought. Additionally, they showed significantly higher above-ground dry matter accumulation under both irrigated and drought conditions, indicating that increased biomass contributed to carbon assimilation and GY. This combination of traits may have led to higher carbon fixation rates, at either the leaf or the whole plant basis, in AlSAP-lines, particularly during the active grain filling phase, which occurred close to the peak of drought-stress.

Delayed LR indicates an ability to sustain turgor, despite drought-stress; for example, through increased water uptake or osmotic adjustment. We did not observe differences in Ψw that could explain the reduction of LR in AlSAP-lines, although Ψw was slightly higher in RN6 at stress peak. However, it is interesting that AlSAP-lines maintained Ψw values similar to those of WT-NIP, despite their elevated biomass accumulation, which would be expected to reduce this characteristic (Pantuwan et al., 2002). Neither was differential regulation of G_s observed in AlSAP-lines; hence, the reduced LR may be associated with improved access to water in the soil. The characterization of the response of AlSAP-seedlings to different abiotic stresses demonstrated their improved ability to maintain root growth and development (Ben Saad et al., 2012). Thus, it would be interesting to evaluate the development of the root system of these plants in the field.

It is important to highlight that AlSAP expression reduced LR but did not affect leaf morphology or plant development (data not shown; Supplementary Figure S2). A similar effect has been reported in rice plants overexpressing the Arabidopsis TFs CBF3/DREB1A (CBF3), ABF3, and SDIR1 (Oh et al., 2005; Zhang et al., 2009), or OsNAC TFs (Hu et al., 2006; Jeong et al., 2010). The molecular mechanisms underlying this effect remain unknown, but may include the interaction of some of these, or similar TFs, with other components in drought-stress signaling pathways, in which SAPs may also participate.

The imposition of stress at different phenological stages may influence plant responses to drought and the mechanisms by which AlSAP contributes to drought tolerance. Ben Saad et al. (2012) performed physiological investigations on the last expanded leaf of AlSAP rice plants at the 5–6 leaf, early tillering stage. In contrast, we evaluated photosynthesis parameters using the flag leaf, which is physiologically different to other leaves, and suffers distinctly under stress conditions. Hormone balance, oxidative damage, water regulation, and particularly drought-induced changes in photosynthesis, proceed distinctly in the flag leaf (Chaves et al., 2003; Biswal and Kohli, 2013). Drought-stress during anthesis reduces the rate of photosynthesis of the flag leaf and increases the remobilization of assimilates from it to sustain grain demand, accelerating its senescence, which has further impacts on photosynthesis (Peng and Ismail, 2004). In contrast, during vegetative growth and especially in young leaves, drought depresses the photosynthesis rate without inducing senescence (Peng and Ismail, 2004). The mechanisms involved in reducing photosynthesis are different from those that enhance remobilization, and do not necessarily involve AlSAP.

Evaluation at distinct physiological stages may also have involved developmentally associated differences in AlSAP transcription levels or expression patterns. We found that a higher accumulation of transcript correlates positively with drought-tolerance in AlSAP-lines. Similar observations were reported for rice, tobacco, and durum wheat (Ben Saad et al., 2010, 2011, 2012). These results may indicate that complete expression of the effects of AlSAP depends to some degree on gene transcript, or protein, accumulation. Alternatively, expression in certain organs/tissues or at specific growth stages may be required for AlSAP to achieve its full potential.

Co-occurrence of multiple abiotic stresses is another factor to consider. In the field, crops encounter a combination of stresses that could require different or antagonistic responses, and consequently may activate varying molecular mechanisms (Mittler, 2006; Mazzucotelli et al., 2008; Salekdeh et al., 2009). Drought and heat stresses are classic examples of this kind of interaction that may elicit antagonistic physiological responses and activate the transcription machinery in a non-additive fashion (Atkinson and Urwin, 2012). In our field trials, plants coped with both drought and high diurnal temperatures. Although, AlSAP appeared to confer tolerance to both heat and drought stresses applied independently to transformed tobacco (Ben Saad et al., 2010), its effect in response to the simultaneous imposition of these factors has not been tested and cannot be predicted from these individual studies (Salekdeh et al., 2009).

It is also possible that the improved yields observed in AlSAP-lines exposed to drought during the reproductive stage is not the result of better performance at the time of stress, but rather signifies a primed status, acquired during vegetative growth, due to the higher accumulation of carbon reserves resulting from enhanced photosynthetic performance at the vegetative stage.

Evidence gathered to date demonstrates that AlSAP overexpression in rice is effective in conferring drought tolerance at both the vegetative and reproductive stages. Stress protection during distinct phenological stages was also reported for AlSAP-transformed tobacco and durum wheat. Importantly, AlSAP expression exerted similar effects on growth and yield of durum wheat, tobacco, and rice, under both normal and stressed conditions (Ben Saad et al., 2010, 2011, 2012). These similar responses indicate that the beneficial role of AlSAP in regulating growth and development may be universal in monocotyledonous and dicotyledonous plants.

The gains in GY in AlSAP-transformed rice ranged from 26 to 115%, depending on the line and the severity of the stress exerted in the field. These gains are very high compared with reports for other transgenic rice lines expressing genes conferring drought tolerance and challenged with milder stress intensity, as indicated by the GY loss observed in controls under drought vs. standard conditions (see Table 1 for AlSAP-lines). For example, overexpression of AP37 in rice enhanced drought tolerance at the reproductive stage and resulted in yield gains of 16–57% above controls. However, in this study, the stress intensity was limited, since plants remained irrigated to avoid LR, resulting in a GY loss of only 32% in the WT line (Oh et al., 2009). In another study, rice plants overexpressing OsNAC10 showed enhanced drought tolerance at the reproductive stage, and GY increased by 25–42%; however, again milder stress conditions were applied (Jeong et al., 2010). The expression of HYR, which afforded a 14–39% GY increase under stress, was also evaluated under mild drought stress, since GY in the control line reduced by only 30% (Ambavaram et al., 2014).

A limited number of studies applied drought intensity more comparable to our experimental conditions. Plants overexpressing LOS5 and ZAT10 showed gains ranging from 11 to 36% compared with their controls, which suffered an 82% yield reduction (Xiao et al., 2009). Plants overexpressing AtEDT1/HDG11 exhibited a GY gain of 65% over the control, the yield of which was reduced by 83% when subjected to severe drought (Yu et al., 2013). More remarkably, rice lines expressing OsCPI1 showed a 2.5–3-fold superior GY over the control, the yield of which dropped by 90% under severe drought conditions (Huang et al., 2007).

Grain production penalties, due to panicle sterility and abnormalities, under unstressed conditions have frequently been reported when transgenic TFs are constitutively expressed (Dubouzet et al., 2003; Oh et al., 2009; Park et al., 2013). In our study, the use of three distinct promoters, CaMV35S, maize Ubi-1, and rice ACT1, inducing a range of constitutive gene expression levels, imposed no penalties in either rice growth or GY under irrigated conditions. Conversely, we observed modest, but consistently higher GY in AlSAP rice lines under paddy fields. These results contrast with the effects reported previously from several studies of SAP over-expression. To date, ectopic expression of five (OsSAP1, OsSAP8, OsSAP7, OsSAP9, and OsSAP11) of the 18 rice SAP genes has been investigated in heterologous or homologous systems. Two examples of SAPs (OsSAP1 and OsSAP8) over-expressed in rice under the maize ubiquitin promoter resulted in an enhanced tolerance to abiotic stresses, but detrimental effects on GY under standard growth conditions (Kanneganti and Gupta, 2008; Dansana et al., 2014). A third example (OsSAP7) demonstrated a negative effect of SAP overexpression in Arabidopsis abiotic stress tolerance, due to its role as negative regulator of ABA responses (Sharma et al., 2015).

The fact that these detrimental effects are absent in rice plants expressing the AlSAP gene underlines the potential for the application of AlSAP transgenic lines to reduce yield losses in crop species facing environmental stress.

This is the first report of field evaluation of transgenic plants expressing an A20/AN1 stress-associated protein. We consistently observed that AlSAP accumulation enhanced GY under drought, and did not penalize either yield or plant growth under irrigated conditions. Improved GY was associated with higher biomass accumulation, a greater number of productive panicles, and lower panicle sterility, even under severe stress. Milder environments should be tested since it is possible that further GY gains can be attained by AlSAP expression under more favorable conditions. It will also be important to introduce AlSAP in other genetic backgrounds, notably elite upland cultivars with established basal levels of drought-tolerance, to determine whether AlSAP can further enhance their ability to resist drought.

Recent reports of transgenic protection of rice from drought-stress associated beneficial transgene effects with stronger root systems (Jeong et al., 2010, 2013; Redillas et al., 2012; Ambavaram et al., 2014). The influence of AlSAP on root development and hydraulic conductance require further assessment in specifically designed experiments. Also, a combination of transgenes with different modes of action, and that improve distinct traits contributing to yield under drought could be attempted.

We note apparent discrepancies in the effects of AlSAP in protection of photosynthesis observed in our greenhouse and field experiments. Several hypotheses to explain these findings have been presented and require further exploration. To provide additional clues regarding the physiological and molecular mechanisms operating in AlSAP crops, a better understanding of AlSAP gene function is required; in particular, investigations to identify proteins interacting with AlSAP in rice and decipher its regulation network will be beneficial. Such studies will provide opportunities to consolidate the protection afforded by AlSAP through stress-inducible/developmentally regulated expression strategies and combined expression with other genes (e.g., NAC genes) affording abiotic stress tolerance through different modes of action.