Natural variation identifies new effectors of water-use efficiency in Arabidopsis

We leveraged the measurements of integrated WUE as δ¹³C of above-ground biomass of well-watered plants or plants subjected to terminal drought from Kenney et al. (21). We reanalyzed the core subset of 185 accessions (Dataset S1). Most accessions had a higher WUE in the drought treatment (higher δ¹³C, F = 42.89, P < 0.0001; Fig. 1A). Only a few accessions had decreased or stable WUE in the drought treatment compared to the unstressed (wet) control (Fig. 1A). GWAS was performed using the plasticity of WUE (Fig. 1B; plasticity = drought − well-watered control (wet); mean of −0.585 and range −3.35 to 1.75) as well as the WUE of the well-watered control and drought treatments (SI Appendix, Fig. S1). While all three sets of GWAS results identified potentially interesting associations, we focused on the WUE plasticity GWAS for further study, since the ability to change WUE in response to changing environmental conditions may be an important adaptive trait whose regulation is poorly understood. Moreover, we hypothesized that selecting genes from WUE plasticity GWAS would discover genes affecting both sides (transpiration or biomass accumulation) of the WUE ratio. The SNPs most significantly associated with WUE plasticity were distributed across several genomic regions (Fig. 1C). As GWAS generates many candidates, selecting the most promising genes for the follow-up functional study is challenging. Thus, we used several layers of data to identify and prioritize a large set of candidate genes for high-throughput testing, rather than studying the molecular mechanisms of any particular gene (SI Appendix, Fig. S2A). We first generated a list of candidate genes for which any part of the gene body (untranslated regions [UTRs], introns, and exons) was within 5 kb of one or more of the top 500 lowest P-value SNPs (nominal P = 4.37 × 10⁻⁵ to 4.2 × 10⁻³). The 1,058 genes so identified (Dataset S2) were then categorized based on their stress-induced gene expression to further determine which genes were most likely to be involved in WUE plasticity. Stress-responsive changes in gene expression were identified by referring to transcriptomic studies (29, 30) (Datasets S3 and S4) that examined gene expression during acclimation to an extended duration of nonlethal drought stress conditions similar to those used to measure WUE plasticity in the current study. This analysis identified 198 genes that were associated with the top 500 WUE plasticity SNPs and that were up-regulated during drought stress (SI Appendix, Fig. S2B and Dataset S5). These genes were further subjected to Gene Ontology (GO) enrichment analysis, which identified 10 significant functional annotation terms (SI Appendix, Figs. S2C and S3 and Dataset S6). To incorporate a range of gene functions into our validation experiments, we selected the top 10 genes from each enriched GO term based on the lowest P values of their associated SNPs. We also included all 19 genes from the transcriptional regulation GO cluster. This filtering identified 72 candidate genes (Dataset S7). One limitation of this approach is that it may discard poorly annotated genes. We found 29 such genes annotated as proteins of unknown functions associated with the top 500 SNPs and were transcriptionally responsive to stress (Dataset S5, genes highlighted in red). Thus, we also included those 29 genes in our functional study.

We isolated 88 homozygous T-DNA insertion lines covering 70 candidate genes (Dataset S9). We were unable to obtain homozygous T-DNA lines for the other 31 prioritized genes. (T-DNA alleles for all prioritized genes are given in Dataset S8). These homozygous lines were tested for their effect on WUE by gravimetric analysis using a high-throughput closed container system with well moistened soil (Fig. 1D and SI Appendix, Fig. S4A and B). This first round of screening identified mutants for 27 genes that significantly differed from wild type for WUE (SI Appendix, Fig. S4A and Datasets S10 and S11). Among these 27 genes, only 9 genes had previously known roles in abiotic stress (Dataset S12).

These mutants were further analyzed for their effect on WUE by two independent gravimetric analyses using larger containers (237 mL) to allow the plants a larger rooting volume and space for rosette biomass (SI Appendix, Fig. S4C). We also measured δ¹³C of above-ground biomass, leaf surface temperatures, SD, and stomatal index (SI) (Dataset S13). This set of measurements included mitochondrial editing factor11 (mef11-5) and open stomata kinase (ost1-3; OST1 is also referred to as SnRK2.6) mutants as controls known to affect water loss. mef11-5 is hypersensitive to ABA-mediated stomatal closure and exhibits elevated leaf temperatures and reduced water loss under water stress (31, 32). In contrast, ost1-3 has reduced ABA sensitivity, leading to reduced stomatal closure, greater leaf water loss, and reduced leaf temperature (33, 34). Note that mutants for two genes, Cytochrome P 450 family 76 (CYP76C2) and AT3G58660, did not germinate for unknown reasons, and so we were unable to include them in our further study. We found a significant positive correlation (R² = 0.646, P < 0.001) for WUE between smaller containers (high-throughput system) and larger containers for these mutants (SI Appendix, Fig. S5A).

These experiments identified genes that altered either side (biomass accumulation or transpiration) of WUE (Fig. 2A). For example, plant cysteine oxidase 5 (pco5-1 and pco5-2) had an increase in biomass (2.8-fold), which gave it a higher WUE despite the fact that it consumed substantially more water (1.9-fold) than the Col-0 wild type (Fig. 2A, purple circles). A few other mutants had no change in biomass but showed moderately reduced water consumption (significant based on nominal P value), leading to increased WUE (Fig. 2A, blue circles). Several other mutants had moderately decreased WUE mainly due to decreased biomass accumulation with no significant change in water consumption (Fig. 2, green circles). We also found mutants with strong reduction in biomass as well as water consumption. However, changes in biomass had a relatively larger impact on WUE in these mutants (Fig. 2A, orange circles). The two control mutants mef11-5 and ost1-3 showed high and low WUE, respectively (Fig. 2A), consistent with their effects on ABA sensitivity. Interestingly, mef11-5 was the only mutant in our analysis where decreased biomass was associated with increased WUE. Across these mutants, there was a strong correlation between above-ground biomass gain and water consumption (Fig. 2B; R² = 0.89, P < 0.001). We also found that WUE was moderately correlated with biomass (Fig. 2C; R² = 0.628, P < 0.001) and weakly correlated with water consumption (Fig. 2D; R² = 0.364, P < 0.001), suggesting that the majority of mutants analyzed primarily altered the biomass side of the WUE ratio. In addition, we found a relatively weak correlation between WUE and SD (Fig. 2E; R² = 0.142, P < 0.017), and no significant relationship between WUE and stomatal index (SI Appendix, Fig. S5B; R² = 0.083, P < 0.057), suggesting that changes in SD or SI were not the main driver of altered WUE in our set of mutants. The correlation between gravimetrically determined WUE and δ¹³C was significant and positive (Fig. 2F; R² = 0.546, P < 0.001). We discuss sets of discovered genes based on their impact on WUE, biomass, and water consumption below.

We found that mutants of Plant cysteine oxidase 5 (pco5-1 and 2), Defective UGE in root (dur-1), Calcium-dependent protein kinase 23 (cpk23), and Nuclear speckle localized RNAa (nsra-1 and 2) had increased WUE, indicating that the proteins encoded by these genes have a negative effect on WUE (Fig. 3). None of these mutants affected SD (SI Appendix, Fig. S6A). Two mutant alleles of PCO5 had a strongly significant increase in gravimetric WUE and δ¹³C while also producing more biomass and consuming more water (Fig. 2A, purple circles). Thus, mutants of PCO5 were more water productive despite having significantly reduced leaf temperatures (Fig. 3 and SI Appendix, Fig. S7) indicative of a higher transpiration rate. Since pco5 mutants had no effect on SD, the reduced leaf temperature suggests that stomatal size regulation of stomatal aperture may instead be altered. PCO5 is one of five cysteine oxidases involved in N-end rule protein degradation and hypoxia response (35).

Mutants of three other genes in this category, cpk23, dur-1, and both nsra alleles, had significantly increased WUE. These mutants had no effect on biomass but showed moderately reduced water consumption, suggesting that they altered the transpiration side of WUE (Figs. 2A, blue circles and Fig. 3). Among these, cpk23 had the strongest effect on WUE. dur-1 had a marginally nonsignificant (P = 0.08) increase in δ¹³C, consistent with the gravimetric analysis. (For marginally nonsignificant effect, nominal P values are listed in the text and in figures throughout.) cpk23 and dur-1 had elevated leaf temperatures, whereas nsra-1 and nsra-2 had decreased leaf temperatures compared to wild type (Fig. 3 and SI Appendix, Fig. S7). The control mutant mef11-5 had a moderate decrease in biomass and strongly reduced water consumption leading to increased WUE. It had no effect on SD (SI Appendix, Fig. S6A) but did have elevated leaf temperature (Fig. 3 and SI Appendix, Fig. S7), consistent with previous reports (31, 32). The elevated leaf temperatures of cpk23, dur-1, and mef11-5 are consistent with their reduced water consumption. However, the decreased leaf temperatures for the NSRa mutants contrast with their reduced water consumption. It is possible that although NSRa mutants did not have significantly reduced biomass, they may have had reduced leaf area that could explain the apparent mismatch between water consumption and leaf temperature. DUR encodes a uridine diphosphate (UDP)-glucose epimerase (UGE) involved in UDP-arabinose biosynthesis, possibly affecting cell wall properties (36). NSRa is an RNA binding protein with no previous information to connect it to stress responses or stomatal development. NSRa, and related NSRs, affect alternative splicing and thus could influence activity of downstream genes leading to increased WUE.

Mutants of 13 genes had a strong decrease in WUE mainly due to a strong reduction in biomass even though their water consumption was also reduced (Fig. 2A, orange circles and Fig. 4). None of these mutants affected SD (SI Appendix, Fig. S6B). Mutants of the first five genes in this category mterf defective in Arabidopsis (mda1-3), cytochrome P450, family 7070a3 (cyp707a3-3), dehydration responsive element-binding protein 2a (dreb2a-1 and 2), general control nonrepressible 20 (gcn20-2), and a hypothetical protein with unknown function (at1g49170) had a strong decrease in biomass as well as strongly reduced water consumption (Fig. 4). However, biomass was more strongly reduced, leading to a substantial decrease in WUE in these mutants. Decreased δ¹³C in these mutants was also consistent with the gravimetric WUE. MDA1, CYP707A3, and DREB2a, were previously reported to affect ABA and water stress signaling, whereas GCN20 has roles in hormone signaling (37). MDA1 encodes for mitochondrial transcription termination factor (mTERF). mda1 was previously shown to enhance salt and osmotic stress tolerance probably due to reduced sensitivity to ABA (38). CYP707A3 acts as an ABA 8′-hydroxylase involved in ABA catabolism (39). cyp707a3-3 had elevated leaf temperatures (Fig. 4 and SI Appendix, Fig. S8) consistent with its reduced water consumption and with previous observations of lower stomatal conductance and higher basal ABA levels in unstressed cyp707a3 plants (39). DREB2a is a transcription factor that functions in both water and heat-stress responses (40). Of the two dreb2a alleles we examined, dreb2a-3 had a stronger effect than dreb2a-4, consistent with dreb2a-4 being a knockdown, rather than knockout (41). Both mutant alleles of DREB2a had decreased leaf temperatures (Fig. 4 and SI Appendix, Fig. S8), indicating that they may affect WUE by controlling stomatal aperture. GCN20 encodes an ABC transporter family protein. Mutants of GCN20 were defective in MAMP/bacterium-triggered stomatal closure but respond normally to salicylic acid (SA) and ABA (37). gcn20-2 had a decrease in leaf temperatures and no change in SD, consistent with a previous report (37). The AT1G49170 mutant also had decreased leaf temperature (Fig. 4 and SI Appendix, Fig. S8).

Mutants of the next six genes in this category include osmosensitive calcium-permeable cation channel 3.1/early responsive to dehydration 4 (osca3.1/erd4), jasmonic acid (JA), transporter 4 (jat4/atjat4), histone acetyl transferase (hac5), suppressor of max2 1–like 5 (smxl5), AT1G03687, and grana deficient chloroplast 1 (gdc1) had strong decreases in biomass with moderately reduced water consumption, indicating that biomass was the driver for the decreased WUE of these mutants (Fig. 2A, orange circles). OSCA3.1 and JAT4 have known or highly probable roles in stress or hormone signaling (42, 43) and the mutants of OSCA3.1 and JAT4 showed strong decreases in WUE. Similarly, hac5-6, smxl5-2, mutants of AT1G03687, and gdc1-2 all had strongly decreased WUE. HAC5 functions in the transcriptional repression of genes related to flowering time and floral development (44). hac5-6 had a marginally nonsignificant decrease in leaf temperatures (P = 0.09) (Fig. 4 and SI Appendix, Fig. S9). SMXL5 promotes secondary phloem formation (45) and smxl5-2 had decreased δ¹³C consistent with gravimetric WUE. The altered WUE in smxl5-2 might be due to up-regulation of several stress-related pathways in this mutant (45). The function for AT1G03687 is unknown. Both mutants of AT1G03687 had strong decreases in δ¹³C consistent with gravimetric WUE. GDC1 is a ankyrin domain containing chloroplast protein essential for grana formation (46). gdc1-2 had a marginally nonsignificant decrease in water consumption (P = 0.08) and δ¹³C (P = 0.057) and showed increased leaf temperature. Consistent with previous reports, gdc1-2 exhibited pale green leaves and ceased growth at the vegetative stage (46), indicating that this mutant had decreased WUE because of impaired photosynthesis.

The next gene in this category was IQ-domain 11 (IQD11), another Ca²⁺ responsive protein involved in regulation of microtubule orientation (47). Among two T-DNA alleles, only iqd11-1 showed a strong decrease in WUE, biomass, and δ¹³C, while iqd11-2 had a more moderate effect on these phenotypes and was not significant for δ¹³C. iqd11-2 has a T-DNA insertion in the 5′ UTR and thus is likely a knockdown rather than knockout mutant (Dataset S11). Note that iqd11-1 had a marginally nonsignificant decrease in water consumption (P = 0.058). Both mutants of IQD11 had decreased leaf temperatures (Fig. 4 and SI Appendix, Fig. S9). Parallel 1/Nucleoline-1 (PARL1/NUC1) is involved in many aspects of ribosomal biogenesis and affects leaf venation (48). parl1-2 is the only mutant in this category that had a moderate decrease in WUE. It had decreased δ¹³C consistent with gravimetric WUE and showed elevated leaf temperature (Fig. 4 and SI Appendix, Fig. S9).

The changes in leaf temperatures for mutants in this category suggest their role in stomatal response (Fig. 4 and SI Appendix, Figs. S8 and S9). The elevated leaf temperatures found for cyp707a3-3, gdc1-2, and parl1-2 was consistent with their reduced water consumption; however, decreased leaf temperatures found for mutants of DREB2a, GCN20, HAC5, and IQD11 contrasts with their reduced water consumption. Here again, change in leaf area or rosette architecture may explain the seeming mismatch between leaf temperature and water consumption.

Mutants for six genes, With No Lysine (WNK11), Plethora 1 (PLT1), AT2G45460, AT3G62220, Basic Penta Cysteine (BPC2), and Carbon Catabolite Repressor 4b (CCR4b), had decreased WUE and reduced biomass but no significant change in water consumption (Fig. 5 and SI Appendix, Fig. 2A, green circles) or SD. WNK11 belongs to the WNK protein kinase family, which is involved in various processes, such as ABA sensitivity, proline accumulation, regulation of flowering time, and intracellular signaling through G protein (49); however, the function of WNK11 itself is unknown. wnk11-1 had strongly decreased WUE and biomass while others had moderate effects in this category. The decreased δ¹³C in wnk11-1 was consistent with gravimetric analysis. PLT1 is a negative regulator in ABA inhibition of root growth (50). plt1 had a marginally nonsignificant (P = 0.054) decrease in δ¹³C, consistent with gravimetric WUE. BPC2 functions in negative regulation of osmotic stress tolerance and it also determines β-1,4-galactan accumulation in response to salt stress (51, 52). bpc2 also had decreased δ¹³C consistent with gravimetric WUE. It is possible that WNK11, PLT1, and BPC2 affect WUE through mechanisms related to ABA or osmotic stress signaling.

AT2G45460 and AT3G62220 encode proteins of unknown function. For AT3G62220, we observed disparate results for two independent T-DNA lines; only at3g62220-1 (promoter insertion) had a significant effect on biomass, while at3g62220-2 (exonic insertion) had a marginally nonsignificant (P = 0.08) effect. Conversely, at3g62220-2 had a significant effect on leaf temperature and SD while at3g62220-1 had no significant effect (Fig. 5 and SI Appendix, Figs. S6C and S10). ccr4b-1 had a marginally nonsignificant decrease in WUE (P = 0.08) and it had decreased leaf temperatures. CCR4b determines the poly(A) length of transcripts related to starch metabolism and may also affect degradation of stress responsive RNAs via its interaction with Pumilio RNA-binding protein 5 (53). The control mutant ost1-3 had a moderate effect on WUE and biomass, a strong decrease in δ¹³C, and an expected decrease in leaf temperatures (Fig. 5 and SI Appendix, Fig. S10). Surprisingly the altered leaf temperatures did not affect the overall water consumption in these mutants and none of these mutants influenced SD except for at3g62220-2 that had increased SD (SI Appendix, Fig. S6C). There is no prior information for genes in this category to link them to WUE or traits related to plant water relations.

Finally, we found mutants of natural resistance associated macrophage protein 4 (NRAMP4) and carbamoylphosphate synthetase subunit B/Venosa3 (CarB/VEN3) had marginally nonsignificant decreases in WUE without significantly affecting final biomass or water consumption (SI Appendix, Fig. S11 and Fig. 2A: gray circles). In addition, these mutants did not affect δ¹³C or SD (SI Appendix, Figs. S11 and S6D) and carb did not affect δ¹³C or leaf temperatures (SI Appendix, Fig. S11). NRAMP4 is functionally redundant with NRAMP3 in maintaining photosynthesis under cadmium and oxidative stress (54). CarB is involved in the conversion of ornithine into citrulline in arginine biosynthesis; however, the carB mutant used in this study is a knockdown mutant that expresses a reduced level of carB (55), suggesting that a carB knockout mutant may have greater effect on WUE.