Genetic analysis of water loss of excised leaves associated with drought tolerance in wheat

Plant material

The mapping population consisted of 94 doubled haploid lines (CSDH) from the cross between hexaploid wheat (Triticum aestivum L.) genotypes Chinese Spring (CS) and SQ1 (a breeding line) according to Quarrie et al. (2005) and available from the John Innes Centre, Norwich (mike.ambrose@bbsrc.ac.uk).

Excised leaf dehydration

Excised-leaf water loss in the CSDH population was determined using the procedure of Clarke & McCaig (1982a). After six weeks vernalization at 4 °C, plants were grown under well-watered conditions in a temperature-controlled glasshouse until leaf 4 ligule emerged. This leaf was then detached, quickly transferred to a nearby walk-in growth chamber maintained at 20 °C, 50% relative humidity and continuous light of 250 µmol m⁻² s⁻¹ (HPS “Agro” lamps, Osram), placed on a v-shaped card support (Fig. S1.) and water loss monitored. Leaf weights were recorded immediately (0 h), after 3 and 6 h and finally after drying at 70 °C for 48 h.

The mapping population was tested three times (Experiment I, II and III) under ostensibly identical conditions in the growth chamber in consecutive years (2007–2009), with three replicate leaves sampled per CSDH line on each occasion.

ELWL after 3 h, 6 h and from 3 to 6 h was calculated as water loss per unit initial water content (ELWLW), according to Clarke & McCaig (1982a) as follows (1)${\displaystyle {\mathrm{ELWLW}}_{0-3\mathrm{h}}=\frac{\left({\mathrm{FW}}_0-{\mathrm{FW}}_3\right)}{\left({\mathrm{FW}}_0-\mathrm{DW}\right)},}$ (2)${\displaystyle {\mathrm{ELWLW}}_{3-6\mathrm{h}}=\frac{\left({\mathrm{FW}}_3-{\mathrm{FW}}_6\right)}{\left({\mathrm{FW}}_3-\mathrm{DW}\right)},}$ (3)${\displaystyle {\mathrm{ELWLW}}_{0-6\mathrm{h}}=\frac{\left({\mathrm{FW}}_0-{\mathrm{FW}}_6\right)}{\left({\mathrm{FW}}_0-\mathrm{DW}\right)}}$ where FW₀, FW₃ and FW₆ are fresh weight after 0, 3 and 6 h, respectively, and DW is dry weight after drying at 70 °C.

Leaf length and width were measured before dehydration to estimate leaf surface area (LA) as length × width × 0.78: a factor previously determined to be appropriate for wheat leaf four (Quarrie & Jones, 1977). From this, rate of water loss/cm² (ELWLA) during the first 3 h, 3–6 h and 6 h of excision was calculated as for Eqs. (1)–(3), substituting LA for water content, as well as initial leaf FW/cm².

As ELWL has also been expressed in the literature on the basis of initial leaf FW or DW, ELWL on the basis of FW (ELWLF) and DW (ELWLD) were calculated (substituting FW-DW in Eqs. (1) to (3) with FW₀ or DW, respectively).

Leaf morphological and anatomical measurements

In experiment (III), prior to dehydration the basal ca. 2 cm of each leaf 4 was cut and placed into 70% ethanol solution. Leaf cross-sections were prepared manually, and analysed on a microscope slide under a bright-light microscope at ×5 magnification. Sections were photographed with a digital camera (LEICA DC 300) and leaf thickness measured at the midrib and along the lamina: two measurements in the “valley” between two secondary vascular bundles and two measurements at the thickest part across vascular bundles on each side of the midrib. Thus, CSDH line mean midrib thickness and lamina thickness were based on three measurements at the midrib and 24 measurements across the lamina (three leaves × (four maximum widths+four minimum widths)).

Leaf segments were also examined directly at ×10 magnification to count stomata per field of view (0.761 mm²). Two fields of view either side of the midrib on the lower leaf surface were selected randomly on each leaf segment.

Grain yield from field and pot trials

Grain yield per plant from field trials with 95 CSDH lines in Norwich, UK in 1997 and 1998 (mean of five random plants per CSDH line) were described in Quarrie et al. (2005), and in Zaječar, Serbia in 2000, 2001 and 2002 (yield and plant number per row) were described in Quarrie et al. (2005), Quarrie et al. (2006). Grain yield per plant from pot trials in Krakow, Poland in 2007, 2008, 2010 and 2011 with irrigated and droughted treatments were described in Czyczyło-Mysza et al. (2013) and Czyczyło-Mysza (2013).

Further yield trials were carried out in the field in Zaječar in 2004–2005 and 2005–2006, with plants grown from autumn sowings (30th October, 2004 and 14th November, 2005, respectively) as described for previous years in Quarrie et al. (2005), with two treatments. Two replicate plots were rainfed and two replicate plots were covered with a rain-out shelter at the beginning of tillering, from 4th (2005) and 5th (2006) April, as described in Dodig et al. (2002). The shelter stayed over droughted plots until maturity and reduced light intensity by around 50%.

Pot trials with 94 CSDH lines were carried out in Krakow in 2006, with seedlings transferred to a glasshouse (three plants per line) with or without vernalisation, as described in Czyczyło-Mysza et al. (2013). Only the main ear was sampled for grain weight. Two other trials, under the same conditions described for 2010 and 2011 by Czyczyło-Mysza et al. (2013), were carried out in 2012 and 2013. Yield per plant was recorded in irrigated and droughted treatments with three replicate plants per CSDH line and treatment.

Grain yield per plant was also measured in the following trials. In 1994, all CSDH lines were multiplied in the field from a spring sowing with one row per line at Morley Experimental Station, Norfolk, UK, and in a soil glasshouse trial at the John Innes Centre, Norwich, UK using 73 CSDH lines. Spring-sown plants, one row per CSDH line in two replicate plots, were watered until 16th May and then again from 6th July, to give a drought stress during flowering and early grain filling. Field trials with 95 CSDH lines were sown in the autumn with two replicate plots of three rows per CSDH line in 2002–3 and 2003–4 in Conselice and Idice, northern Italy, organised by Società Italiana Sementi Spa (Ravaglia, 2005). The ozone fumigation trial of 2003 using open-top chambers at Newcastle University’s Close House Experimental Station, UK (Quarrie et al., 2006) included two additional ozone fumigation treatments not reported in Quarrie et al. (2006), of nominally 25 ppb and 50 ppb, with four chambers per treatment and one pot with three plants per CSDH line in each chamber. The Close House ozone fumigation trial was repeated in 2005, with non-filtered air (NFA) and NFA plus 50 ppb ozone treatments. Plants were grown and treated exactly as described in Quarrie et al. (2006), except that ozone fumigation began 8 d earlier, during the rapid tillering phase.

In total, these 52 year × site × treatment trials for yield per plant included 19 regarded as control (little stress) and 12 where a drought treatment was the major stress, reducing yields significantly, by at least 10%. For these 12 trials a mean drought effect was calculated for each CSDH line as mean drought yield/mean control yield.

A further estimate of yield per plant for each CSDH line under favourable and stressed conditions was calculated using genotype × environment plots for each line as described in Quarrie et al. (2005). Linear regressions of genotype yields from the 52 trials on site mean yields were used to calculate yield per plant for each CSDH line at site mean yields of 2 and 7 g/plant (equivalent to ca. 2.5 and 9 t ha⁻¹, respectively).

Phenotypic analysis

Phenotypic data were analysed using GenStat v. 17, and Microsoft EXCEL™ (trait correlations and broad-sense heritabilities). Normality of distributions of CSDH means for each trait were tested using the Shapiro–Wilk normality test (Shapiro & Wilk, 1965). Two-way analysis of variance (ANOVA) was carried out for each trait to determine effects of experiments, lines and experiments × lines interaction. Relationships amongst traits were calculated based on CSDH line means.

QTL analysis

The CS × SQ1 genetic map of Czyczyło-Mysza et al. (2013), with 702 non-duplicated markers (80 RFLP, 227 SSR, 81 AFLP, 292 DArT, 14 EST, five biochemical and three phenotypic markers) distributed on 21 chromosomes, was slightly adjusted by replacing missing marker scores with scores predicted from flanking markers, when these both had the same allele score, and reanalysed to achieve the best-fit marker orders and to remove occasional order inconsistencies in the Czyczyło-Mysza et al. (2013) map. This adjusted genetic map of 3,640.5 cM (Kosambi mapping function) was used for QTL analysis. QTLs for CSDH line mean data in each experiment were identified using single-marker analysis (SMA) with Windows QTLCartographer version 2.5 (Wang, Basten & Zeng, 2011) or QTL Cartographer v. 1.17j, 28 January 2005 for Macintosh, and Windows QTLCartographer was used for composite interval mapping (CIM) of ELWL and other leaf traits. A QTL from CIM was accepted when the LOD score was greater than 2.

To allow for trait variation from experiment to experiment and to compare QTL coincidences amongst traits on an equal scale, single marker LRmapqtl output was modified by expressing marker additive effects as ratios of the Minimum Significant Additive Effect (MSAE), determined as the minimum absolute one-star [*] P ≤ 0.05 additive effect for a particular trait and experiment. Thus, 1 = marker additive effect equal to MSAE. Using this procedure, a marker additive-effect ratio (MAR) of one is equivalent to P = 0.05 with LRmapqtl, and P = 0.01, 0.001 and 0.0001 with LRmapqtl was determined to be equivalent to ratios of ca. 1.32, 1.65 and 1.92, respectively, with these CSDH lines.

For each trait and marker, a mean MAR was calculated as the mean of MARs from each experiment. Thus, the three-experiment mean of MARs (Method 1) was derived as: trait 3-replicate mean → LRmapqtl → marker additive effect → additive-effect ratio → I-III mean MAR (output for ELWLW_0−3h illustrated in Fig. S2, Method 1). This allowed SMA results for all traits to be compared using the same scale and facilitated graphical display. As peak additive-effect ratios for a particular trait frequently occurred at the same markers in each experiment, though not always reaching one-star (5%) significance with LRmapqtl, the arbitrary threshold of 0.5 × MSAE was used for including a mean MAR for graphical presentation. SMA for traits measured only in Experiment III are also presented as additive-effect ratios. Positive and negative MARs indicate alleles increasing the trait coming from CS (CS alleles increasing) and SQ1 (SQ1 alleles increasing), respectively.

QTLs declared to be present using mean MARs (Method 1) were chosen to be those equivalent to P ≤ 0.05 significance using marker additive effects from LRmapqtl output for the three-experiment mean ratio (Method 2), determined as follows: trait 3-replicate mean_I → trait mean (I+II+III)/3 → LRmapqtl → marker additive effect → MAR (output for ELWLW _0−3h illustrated in Fig. S2, Method 2). These QTLs were characterised in Table S1, and identified with arrowheads in Figs. 1 and 2, Fig. S2.

A coincidence of QTLs was assumed when SMA MAR maxima and/or CIM LOD score maxima were within 10 cM, representing a minimum precision typical for QTL detection (Mangin & Goffinet, 1997).

Phenotypic analyses

Genetic analyses

To reduce the complexity of genetic analysis of ELWL and leaf traits, only 3-experiment mean data for the four ELWL traits ELWLW_0−3h, ELWLA_0−3h, ELWLW_3−6h and ELWLA_3−6h, as well as 3-experiment mean data for leaf length, width, area and Experiment III data for midrib thickness are described in detail here. CIM was used with 3-expt-mean data, and QTL peaks coincident between SMA and CIM are identified in Table S1. About 20% of QTLs classified as significant by SMA were also significant using CIM.

Coincidence of QTLs between ELWL and constitutive leaf traits

As phenotypic correlations between ELWL and constitutive leaf traits were almost invariably negative (Table 4), the coincidence of QTLs for ELWL and constitutive leaf traits is compared (Figs. 3A and 3B) with traces of MARs inverted for ELWLW_0−3h, ELWLA_0−3h, ELWLW_3−6h and ELWLA_3−6h. Although phenotypic correlations between leaf constitutive traits and measures of ELWL were much more significant for ELWLW than for ELWLA, QTLs were coincident between all measures of ELWL and each of the four constitutive leaf traits: QTL coincidences with ELWLW_0−3h were leaf length—4, width—2, area—5, thickness—4; ELWLA_0−3h with length—2, width—4, area—4, thickness—3; ELWLW_3−6h with length—6, width—3, area—6, thickness—3, and ELWLA_3−6h with length—2, width—2, area—4, thickness—1. All four measures of ELWL were coincident with leaf trait QTLs distal on 1AL and near 3B centromere (Fig. 3A). QTLs specific for only ELWLW were coincident with leaf trait QTLs on 4B at the dwarfing gene Rht-B1, 5DL, 7AL and 7DS.

Genetic analysis of yield per plant

Figures 4A and 4B shows MARs from SMA for five measures of yield (control, droughted, droughted/control, yield at 7 and 2 t ha⁻¹), together with mean MARs for all 52 yield trials, ELWLW_0−3h and ELWLA_0−3h. Yield QTLs combined using Method 1 from all trials were consistently present with increasing alleles from CS on chromosomes 1D, 4A, 4B, 4D, 5A, 7A, 7B, and from SQ1 on chromosomes 1D, 2B, 2D, 3D, 4B, 5D, 6B and 7A.

Many peak MARs were consistent across the four measures of yield/plant (arrowheads in Fig. 4) but differed from those for yield drought/control ratio (Fig. 4). Major differences in QTLs between the 19 control and 12 droughted yields were present on 16 chromosomes. Increasing alleles were contributed by CS for control-specific QTLs on chromosomes 4D, 5D and 7B, and by SQ1 on chromosomes 1A, 2B, 3B, 3D, 4B and 7A. Drought-specific QTLs were present on 1A, 1D, 4B and 5A (increasing allele from CS), as well as 1D, 2A, 3D and 6B (SQ1 alleles increasing). QTLs with MARs ≥1 for yield at a site yield of 7 t ha⁻¹, were located on chromosomes 1A (CS alleles increasing) and 2B, 3D, 4D and 7A (SQ1 alleles increasing). For yield at a site yield of 2 t ha⁻¹, QTLs were found on chromosomes 1B, 2B, 4A and 7A (CS alleles increasing), as well as 2D, 3B and 6A (SQ1 alleles increasing).

Yield response to drought (drought/control) showed highly significant (P < 0.001) QTLs on chromosomes 1D and 5D (increasing alleles from CS and SQ1, respectively), with other major QTLs (P < 0.01) on chromosomes 1A, 3A, 4B and 7A (increasing alleles from CS), and 2A and 6B (increasing alleles from SQ1).

The physiological control of ELWL and its genetic variation

We selected leaf 4 from glasshouse-grown plants for our genetic analysis of ELWL, as the mapping population varies considerably in phenology (Quarrie et al., 2005), with days to flag leaf emergence varying over two weeks (Table S5). Others have also studied leaf water loss in young plants (Golestani Araghi & Assad, 1998; Rampino et al., 2006), though Clarke (1983) found a genotype × environment interaction comparing glasshouse and field-grown plants sampled near anthesis.

ELWLW was usually highly negatively correlated with aspects of leaf size: leaf DW, leaf area, leaf length, as well as midrib thickness (Table 4), implying that a longer path length for water to reach the epidermis slowed rate of water loss from the leaf surface. Significant negative associations between rate of water loss and leaf area have been found by others in both sorghum (Ali et al., 2011) and wheat flag leaves (Sayed & Bedawy, 2016). Although variation in ELWLA could also reflect variation in leaf thickness, no significant relationship was found between ELWLA and leaf midrib thickness. Thus, distance from vascular bundles to the epidermis per se was unlikely to be a factor determining rate of water loss. The highly significant negative correlation between midrib thickness and ELWLW_0−3h likely reflected the greater structural requirement of a thicker midrib as leaf length increased.

Stomatal number per unit area was not a factor in determining genotypic variation in ELWL amongst the CSDH lines (Table 4), though Wang & Clarke (1993) found a highly significant positive correlation between rate of water loss up to 2 h from excision and stomatal frequency amongst 12 hexaploid wheats. Furthermore, genotypic variation in the rate of water loss was unlikely to indicate genotypic variation in stomatal aperture as correlations for a given measure of ELWL between 0–3 h and 3–6 h (Table 4) were all significant, and stomata would be expected to have closed within a few minutes of leaf detachment as leaves lost turgor. Nevertheless, genotypic differences in non-stomatal water loss due to variation in cuticular thickness or composition, already reported for wheat (e.g., Clarke & Richards, 1988; Jäger et al., 2014; Bi et al., 2016; Bi et al., 2017a), could have contributed to the variation amongst CSDH lines in ELWL. It was not possible with the hand sections of Experiment III to assess cuticle thickness.

The genetic control of ELWL and candidate genes

As additive effects using SMA varied up to 1.7-fold amongst experiments, additive ratios (Method 1) were used to compare between experiments and traits. Our QTL analyses (Table S1) demonstrated a broad genetic control of ELWL with QTLs distributed across several chromosomes, with increasing alleles from both parents, though few QTLs were stably expressed every year. MAR traces of ELWLW for different time intervals in 2007 (Fig. S6) implied the same genetic control of water loss for each time interval. Figure 1 confirms the extensive coincidence between 0–3 h and 3–6 h QTLs for both ELWLW and ELWLA. Therefore, it is probable that genetic variation in water loss was determined largely by non-stomatal characteristics. Although no cuticular traits were measured in our detached leaf experiments, visual assessment of CSDH line leaf waxiness at the tillering phase in the field in 2004 scored from 1 (very little visible wax) to 3 (thick greyish wax) showed QTLs coincident with those for ELWLW_0−3h on 5BL and 5DL (QELWLW₀₋₃.csdh-5B.2 and QELWLW₀₋₃.csdh-5D.1) (Table S1).

Yang et al. (2009) and Mei (2012) have reported preliminary information on the genetic control of ELWL in wheat. Yang et al. (2009) identified QTLs on chromosomes 1D, 4A, 6B and 6D, with those on 6B (near locus Xgwm193) and 6D (near locus Xbarc173) being coincident with weak ELWL QTLs in our mapping population. Mei (2012) reported an additional QTL for ELWL on chromosome 2A where we also had a weak QTL effect.

Two likely classes of candidate gene for regulating leaf water loss would be those regulating water flow to the epidermis and those regulating its evaporation from the leaf surface through the cuticle. Aquaporins are water channel proteins belonging to the Major Intrinsic Protein superfamily of integral membrane proteins which specifically facilitate the passive flow of water molecules across cellular membranes (Maurel, 1997). Forrest & Bhave (2010) assigned several aquaporin genes to wheat chromosome bins. Plasma membrane aquaporin genes PIP1;1, PIP1;2, PIP2;2 and PIP2;1 were located in bins corresponding to ELWL QTLs on 2BS, 6AL and 7AS, respectively (Table S1). Tonoplast membrane aquaporins TIP1;2 and TIP2;1 were in bins corresponding to ELWL QTLs on 4BS and 6BL, respectively. An aquaporin gene listed in the GrainGenes wEST SQL bin-mapped markers database (http://wheat.pw.usda.gov/cgi-bin/westsql/map_locus.cgi; downloaded June 2006 as an Excel^MS file and searched for ”aquaporin”), BE403397, was located on bins C-2AL1-0.85, 6AL4-0.55-0.90 and 6BL5-0.40-1.00, each coincident with bins for ELWL QTLs (Table S1).

A recent publication by Bi et al. (2017b) reported the location of genes for several transcription factors regulating cuticle biosynthesis genes on the group 5 long arms, 6BL and 6DL. Two more genes with high sequence identity were found on 4A and 4D. QTLs for measures of ELWL were present on 5AL, 5BL, 5DL and 6BL. Only weak effects on ELWL were found on 4A and 4D. The well-characterised leaf waxiness genes W1 and Iw1 (Wu et al., 2013; Hen-Avivi et al., 2016) map distally on 2BS, where a QTL for ELWLA_3−6h was located (Table S1). Thus, some of the genes influencing ELWL may be associated with the regulation of water transport through aquaporins and genes for wax biosynthesis.

ELWL as a trait for improving drought tolerance

Many authors have proposed excised-leaf water loss (or water retention), measured on the basis of either leaf water, fresh weight or dry weight, as a selection criterion to help improve drought tolerance (e.g., Dedio, 1975; Clarke & McCaig, 1982b; Yang, Jana & Clarke, 1991; Dhanda & Sethi, 1998; David, 2010). Indeed, significant positive relationships between excised-leaf water retention and yield have been found in studies on wheat genotypes under drought conditions (Clarke et al., 1989; Petcu, 2005; Geravandi, Farshadfar & Kahrizi, 2011), though not always (Clarke et al., 1989; Dabiry et al., 2015).

Clarke & Townley-Smith (1986) and Clarke (1987) demonstrated that selection for both high and low excised-leaf water retention in durum wheat crosses gave yield advantages for selections with low ELWL, but only under drought conditions. We therefore tested the efficacy of ELWL_0−3h as a selection criterion for yield in the CSDH population and compared this with leaf 4 length, a much simpler trait to measure and one very similar to ELWL in its phenotypic correlations with yield under both control and droughted conditions (Tables 5A, 5B). Although we demonstrated a yield advantage for the 10 CSDH lines with both the lowest ELWL_0−3h and longest leaves compared with the 10 lines at the opposite end of the trait rankings, as a breeding criterion, the advantage of selecting for low ELWL_0−3h was much less apparent (Table 6), with an overall yield advantage of only around 3% compared with the remaining lines in the population, with no clear additional benefit under droughted conditions. Therefore, as a selection criterion, ELWL_0−3h would probably be no more effective than leaf 4 length in improving wheat yields.