Large Root Cortical Cell Size Improves Drought Tolerance in Maize^,,,

Large cortical cells substantially reduce root respiration, permitting greater root growth and exploration of deep soil, thereby improving water acquisition, plant growth, and yield under drought.

Abstract

The objective of this study was to test the hypothesis that large cortical cell size (CCS) would improve drought tolerance by reducing root metabolic costs. Maize (Zea mays) lines contrasting in root CCS measured as cross-sectional area were grown under well-watered and water-stressed conditions in greenhouse mesocosms and in the field in the United States and Malawi. CCS varied among genotypes, ranging from 101 to 533 µm². In mesocosms, large CCS reduced respiration per unit of root length by 59%. Under water stress in mesocosms, lines with large CCS had between 21% and 27% deeper rooting (depth above which 95% of total root length is located in the soil profile), 50% greater stomatal conductance, 59% greater leaf CO₂ assimilation, and between 34% and 44% greater shoot biomass than lines with small CCS. Under water stress in the field, lines with large CCS had between 32% and 41% deeper rooting (depth above which 95% of total root length is located in the soil profile), 32% lighter stem water isotopic ratio of ¹⁸O to ¹⁶O signature, signifying deeper water capture, between 22% and 30% greater leaf relative water content, between 51% and 100% greater shoot biomass at flowering, and between 99% and 145% greater yield than lines with small cells. Our results are consistent with the hypothesis that large CCS improves drought tolerance by reducing the metabolic cost of soil exploration, enabling deeper soil exploration, greater water acquisition, and improved growth and yield under water stress. These results, coupled with the substantial genetic variation for CCS in diverse maize germplasm, suggest that CCS merits attention as a potential breeding target to improve the drought tolerance of maize and possibly other cereal crops.

Suboptimal water availability is a primary constraint for terrestrial plants and a primary limitation to crop production. In developing countries, the problem of yield loss due to drought is most severe (Edmeades, 2008, 2013), and the problem will be further exacerbated in the future due to climate change (Burke et al., 2009; Schlenker and Lobell, 2010; Lobell et al., 2011a; IPCC, 2014; St. Clair and Lynch, 2010). The development of drought-tolerant crops is therefore an important goal for global agriculture. Breeding for drought adaptation using yield as a selection criterion is generally not efficient, since yield is an integration of complex mechanisms at different levels of organization affected by many elements of the phenotype and the environment interacting in complex and often unknown ways. Trait-based selection or ideotype breeding is generally a more efficient selection strategy, permitting the identification of useful sources of variation among lines that have poor agronomic adaptation, elucidation of genotype-environment interactions, and informed trait stacking (Lynch, 2007; Araus et al., 2008; Richards et al., 2010; Wasson et al., 2012; York et al., 2013; Lynch, 2014).

Under drought stress, plants allocate more resources to root growth relative to shoot growth, which can enhance water acquisition (Sharp and Davies, 1979; Palta and Gregory, 1997; Lynch and Ho, 2005). The metabolic costs of soil exploration by root systems are significant and can exceed 50% of daily photosynthesis (Lambers et al., 2002). With a large root system, each unit of leaf area has more nonphotosynthetic tissue to sustain, which may reduce productivity by diverting resources from shoot and reproductive growth (Smucker, 1993; Nielsen et al., 2001; Boyer and Westgate, 2004). Genotypes with less costly root tissue could develop the extensive, deep root systems required to fully utilize soil water resources in drying soil without as much yield penalty. Therefore, root phenes that reduce the metabolic costs of soil exploration, thereby improving water acquisition, are likely to be valuable for improving drought tolerance (Lynch and Ho, 2005; Zhu et al., 2010; Lynch, 2011; Richardson et al., 2011; Jaramillo et al., 2013; Lynch 2014).

Maize (Zea mays) is the principal global cereal. Maize production is facing major challenges as a result of the increasing frequency and intensity of drought (Tuberosa and Salvi, 2006), and this problem will likely be exacerbated by climate change (Lobell et al., 2011b). The Steep, Cheap, and Deep ideotype has been proposed for improving water and nitrogen acquisition by maize when these resources are limited (Lynch, 2013). This ideotype consists of root architectural, anatomical, and physiological traits that may increase rooting depth and thereby improve water acquisition from drying soils. Anatomical phenes could influence the metabolic cost of soil exploration by changing the proportion of respiring and nonrespiring root tissue and affecting the metabolic cost of tissue construction and maintenance, which is an important limitation to root growth and plant development under edaphic stress. Specific anatomical phenes that may contribute to rooting depth by reducing root metabolic costs include components of living cortical area (LCA; Jaramillo et al., 2013), including root cortical aerenchyma (RCA), cortical cell size (CCS), and cortical cell file number (Lynch, 2013).

RCA consists of large air-filled lacunae that replace living cortical cells as a result of programmed cell death (Evans, 2004). Previous studies have demonstrated that RCA improves crop adaptation to edaphic stress by reducing the metabolic cost of soil exploration and exploitation (Fan et al., 2003; Zhu et al., 2010; Postma and Lynch, 2011a, Saengwilai et al., 2014a). RCA is associated with a disproportionate reduction of root respiration, thereby permitting greater root growth and acquisition of soil resources (Fan et al., 2003; Zhu et al., 2010). SimRoot modeling indicated that RCA can substantially increase the acquisition of nitrogen, phosphorus, and potassium in maize by reducing respiration and the nutrient content of root tissue (Postma and Lynch, 2011b). Under water stress in the field, maize genotypes with more RCA had deeper roots, better leaf water status, and 800% greater yield than genotypes with less RCA (Zhu et al., 2010). Under nitrogen stress in the field and in greenhouse mesocosms, maize genotypes with more RCA had greater rooting depth, greater nitrogen capture from deep soil strata, greater nitrogen content, greater leaf photosynthesis, greater biomass, and greater yield (Saengwilai et al., 2014a).

LCA refers to the living portion of the cortex that remains after the formation of aerenchyma (Jaramillo et al., 2013). Recently, we reported that LCA is an important determinant of root metabolic cost and a better predictor of root respiration than RCA (Jaramillo et al., 2013). In that study, maize lines contrasting in LCA were grown under well-watered or water-stressed conditions in soil mesocosms, and LCA was associated with a reduction of specific root respiration. These results provided the impetus to investigate the relative contribution of each component of LCA to metabolic cost. Our focus here is on root CCS.

Plant cell size varies substantially both among and within species (Sugimoto-Shirasu and Roberts, 2003). Cell size in a given species and tissue is under genetic control and results from the coordinated control of cell growth and cell division (Sablowski and Carnier Dornelas, 2014). The increased volume of individual cells is attributable to cytoplasmic growth and cell expansion (Marshall et al., 2012; Chevalier et al., 2014). Cytoplasmic growth is the net accumulation of macromolecules and cellular organelles, while cell expansion refers to increased cell volume caused by enlargement of the vacuole (Taiz, 1992; Sablowski and Carnier Dornelas, 2014). Lynch (2013) proposed that large CCS would decrease the metabolic costs of root growth and maintenance, both in terms of the carbon cost of root respiration and the nutrient content of living tissue, by increasing the ratio of vacuolar to cytoplasmic volume.

The objective of this study was to test the hypothesis that large CCS would reduce specific root respiration (i.e. respiration per unit of root length), which under water stress would result in greater root growth, greater acquisition of subsoil water, better plant water status, and improved plant growth and yield. Diverse sets of genotypes (including landraces and recombinant inbred lines [RILs]) contrasting for CCS were evaluated under water stress and well-watered conditions in soil mesocosms in controlled environments, in the field in the United States using automated rainout shelters, and in the field in Malawi. Our results demonstrate that substantial variation for CCS exists in maize and that this variation has substantial effects on the metabolic cost of soil exploration and thereby water acquisition under drought.

RESULTS

Substantial Variation for CCS Exists in Maize

We observed substantial variation for CCS in maize (Fig. 1; Supplemental Table S1). Cell sizes vary across the root cortex, being greatest in the center of the cortex (midcortical band) and decreasing toward the epidermis (outer band) and endodermis (inner band; Table I). Cell size in the midcortical band was correlated with the size of cells in the outer band (r = 0.75, P < 0.05) and the inner band (r = 0.45, P < 0.05). In this study, the median cell size for the midcortical region was chosen as a representative value for the root CCS. There was over 3-fold variation for CCS among Malawian landraces in MW2012-1 (one of the field experiments in Malawi in 2012), with the largest cells having a cross-sectional area of 533 μm² and the smallest cells having a cross-sectional area of 151 μm² (Supplemental Table S1). Among RILs in greenhouse experiment I (GH1), variation for CCS was over 5-fold (101–514 μm²; Supplemental Table S1). Cortical cell diameter was weakly correlated with cell length (Supplemental Fig. S1).

Figure 1.

Cross-sectional images of second nodal crown roots of maize showing genotypic differences in root CCS. The roots were ablated 10 to 20 cm from the base. The images were obtained using laser ablation tomography. Bars = 100 μm. [See online article for color version of this figure.]

Table I. Summary of ANOVA (F ratio) for the effects of soil moisture regime (treatment) and genotype on shoot biomass, D₉₅, and stomatal conductance in greenhouse mesocosm experiments (GH2 and GH3).

^¶P from 0.1 to 0.05, *P from 0.05 to 0.01, **P from 0.01 to 0.001. Treatment is the moisture regimes imposed, and phenotype is the phenotype class (i.e. large CCS and small CCS), and conductance is the stomatal conductance (mol m⁻² s⁻¹).

Effect	GH2		GH3
	Biomass	D95	Biomass	D95	Conductance
Treatment	51.04**	46.89**	32.47**	29.80**	16.02**
Phenotype	9.77**	6.91**	15.39**	15.29*	5.09*
Phenotype × treatment	2.70¶	6.75**	28.86**	11.86*	0.72

To describe the pattern of cell size variation among various tissues, we measured the size of parenchyma cells in the leaf midrib, mesocotyl cortex, and primary root cortex. Generally, the cells were larger in the leaf midrib and mesocotyl compared with the root cortex (Supplemental Fig. S2). Cell size variation in the root cortex was correlated with size variation in the mesocotyl cortex (r = 0.54, P < 0.01) and leaf midrib (r = 0.40, P < 0.01). The relationship between mesocotyl and leaf midrib cell size was weak and not significant (r = 0.21, P not significant).

The correlation between the CCS of second nodal crown roots in well-watered and water-stressed conditions was positive and significant in PA2011 (the 2011 field study in Pennsylvania; see “Materials and Methods” for description of studies) (r = 0.83, P < 0.05), PA2012 (the 2012 field experiment in PA) (r = 0.94, P < 0.05), and MW2012-1 (r = 0.79, P < 0.05). However, the cells were relatively larger in well-watered compared with water stressed conditions. Trait stability across environments was estimated as the correlation coefficient between CCS measured on young plants in soil mesocosms (30 d after planting; GH2) and mature plants in the field (70 d after planting; PA2011). A positive correlation was found between CCS in mesocosms (GH2) and in the field (PA2011; r = 0.59, P < 0.05).

The selected RILs in experiments GH2, GH3, PA2011, PA2012, and MW2012-2 did not show consistent variation in other root phenes, such as root cortical cell file number, total cortical area, RCA, and root diameter (Supplemental Tables S2 and S3). No significant correlations were observed between CCS with either LCA (r = 0.12, P > 0.353) or RCA (r = 0.004, P > 0.945) in GH1.

Large CCS Is Associated with Reduced Root Respiration, Greater Root Growth, and Better Plant Water Status under Drought in Mesocosms and in the Field

The respiration rate of root segments was measured in plants grown in soil mesocosms in a greenhouse in three experiments (GH1, GH2, and GH3). Large CCS was associated with substantial reduction of specific root respiration; increasing CCS from 100 to 500 μm² approximately halved root respiration (Fig. 2).

Figure 2.

Relationship between root respiration per unit of length and CCS for GH1 (r² = 0.46, P = 0.009), GH2 (r² = 0.59, P = 0.001), and GH3 (r² = 0.52, P = 0.018) in greenhouse mesocosms 30 d after planting. Each point is the mean of at least three measurements of respiration.

In the mesocosms, water stress significantly reduced rooting depth (depth above which 95% of total root length is located in the soil profile [D₉₅]) 30 d after planting in GH2 and GH3 (Table I; Fig. 3). Under water stress, lines with large CCS correlated with 21% (GH2) and 27% (GH3) deeper D₉₅ than lines with small CCS, and there was no relationship in well-watered conditions (Fig. 3; Supplemental Fig. S3). Water stress in GH3 reduced stomatal conductance by 68% and leaf CO₂ assimilation by 43% at 30 d after planting (Table I; Fig. 4). Under water stress, lines with large CCS had 50% greater stomatal conductance and 59% greater leaf CO₂ assimilation than lines with small CCS (Fig. 4).

Figure 3.

Relationship between D₉₅ and CCS for GH2 (r² = 0.48, P = 0.001) and GH3 (r² = 0.45, P = 0.01) in the greenhouse mesocosms 30 d after planting. The regression line is shown only for the significant relationships. Data include both water-stressed (WS) and well-watered (WW) conditions.

Figure 4.

Leaf carbon dioxide exchange rate (A) and stomatal conductance (B) for genotypes with large and small cortical cells in the greenhouse mesocosms (GH3) 30 d after planting under both water-stressed (WS) and well-watered (WW) conditions. Data shown are means ± se for three lines per group (n = 4). Means with the same letters are not significantly different (P < 0.05).

In the field at Rock Springs, Pennsylvania, under water stress, lines with large CCS had 41% (PA2011) and 32% (PA2012) greater D₉₅ and 22% (PA2011) and 30% (PA2012) greater leaf relative water content (RWC) than lines with small CCS (Figs. 5 and 6, A and B; Table II; Supplemental Fig. S4). In the field in Malawi, lines with large CCS had 20% greater leaf RWC than lines with small CCS (MW2012-2; Fig. 6C). In addition, genotypes with deeper D₉₅ had greater leaf water status than genotypes with shallow D₉₅, while there was no relationship in well-watered conditions (PA2011).

Figure 5.

Relationship between D₉₅ and CCS for PA2011 (r² = 0.41, P = 0.01) and PA2012 (r² = 0.42, P = 0.05) in the field 80 d after planting under both water-stressed (WS) and well-watered (WW) conditions. The regression line is only shown for the significant relationships.

Figure 6.

Leaf RWC for genotypes with large and small cortical cells at 60 d after planting in the field in PA2011 (A), PA2012 (B), and MW2012-2 (C) in both well-watered (WW) and water-stressed (WS) conditions. Data shown are means ± se for three lines per group (n = 4). Means with the same letters are not significantly different (P < 0.05).

Table II. Summary of ANOVA (F ratio) for the effects of soil moisture regime (treatment) and genotype on yield, shoot biomass, D₉₅, and leaf RWC in three field experiments (PA2011, PA2012, and MW2012-2).

*P from 0.05 to 0.01, **P from 0.01 to 0.001, ***P < 0.001; values without asterisks are not significant. Treatment is the moisture regimes imposed, genotype is the phenotype class (large CCS and small CCS), and RWC is the leaf RWC (%).

Effect	PA2011			PA2012				MW2012-2
	Biomass	D95	RWC	Biomass	D95	RWC	Yield	Biomass	RWC	Yield
Treatment	49.5***	6.5*	21.0***	41.3***	4.9*	35.7***	46.2***	166.0***	58.8***	81.49***
Genotype	6.7**	7.2***	9.7***	3.9**	0.6	9.8**	10.1***	23.6***	2.0*	20.0**
Genotype × treatment	4.3**	1.6*	5.7**	5.5***	0.3	4.2**	10.7***	13.4***	8.0***	16.6***

Lines with Large CCS Had Lighter Stem Water δ¹⁸O and Greater Reliance on Deep Soil Water under Water Stress in the Field

Soil water δ¹⁸O (the ratio of ¹⁸O and ¹⁶O isotopes) was significantly more enriched in the upper 20 cm of the soil profile and a progressively lighter isotopic signature was observed with increasing depth (Fig. 7). However, the majority of change in this signature was in the topsoil: 0 to 10 cm and 10 to 20 cm (approximately 2.09‰). The soil water δ¹⁸O signatures below 30 cm showed no significant difference with depth (Fig. 7) and were aggregated as deep water. The stem water δ¹⁸O signatures ranged from −8.9‰ to −6.2‰ (Table III). Lines with large CCS collectively had a 32% lighter δ¹⁸O signature than lines with small CCS (Table III). The isosource isotopic mixing model was used to determine the proportional contribution of different soil layers (i.e. 10 cm, 20 cm, and deep water) to plant water uptake. Lines with large CCS had greater average reliance on deep water and were less reliant on shallow water from the top two soil layers than lines with small CCS (Table III).

Figure 7.

Soil water oxygen isotope composition in six soil layers in the rainout shelters (PA2011). Sampling was done 65 d after planting. Values are means ± se of three observation points in the rainout shelters. Means with the same letters are not significantly different (P < 0.05).

Table III. δ¹⁸O of xylem water measured for six genotypes contrasting in CCS under water stress 65 d after planting.

Proportional water use by depth (%) from different soil layers where deep is the aggregate of three deep soil layers (Fig. 7), calculated using multisource mixing model analysis (Phillips et al., 2005). δ¹⁸O values are means ± se.

Classification Based on CCS	RIL	δ18O of Xylem Water	Proportional Water Use by Depth
			10 cm	20 cm	Deep
Large CCS	26	−7.9 ± 0.306	0	35	65
	59	−8.0 ± 0.105	0	24	76
	178	−8.9 ± 0.19	0	2	98
Small CCS	131	−6.4 ± 0.253	0	81	19
	132	−6.2 ± 0.166	0	93	7

Large CCS Was Associated with Greater Plant Growth and Yield under Water Stress

In the mesocosms, water stress reduced shoot biomass by 42% in GH2 and 46% in GH3 (Table I; Fig. 8). Under water stress, lines with large CCS had 80% (GH2) and 83% (GH3) greater shoot biomass than lines with small CCS at 30 d after planting (Table I; Fig. 8). In the field, water stress reduced shoot biomass 46% (PA2011), 38% (PA2012), and 53% (MW2012-2) at 70 d after planting (Table III; Fig. 9). Under water stress in the field, lines with large CCS had greater shoot biomass than lines with small CCS by 51% (PA2011), 81% (PA2012), and 100% (MW2012-2). However, CCS was not associated with biomass in well-watered conditions (Fig. 9). Water stress reduced grain yield by 47% (PA2012) and 46% (MW2012; Table II; Fig. 10). Under water stress, lines with large CCS had greater grain yield than lines with small CCS by 145% (PA2012) and by 99% (MW2012-2; Fig. 10).

Figure 8.

Shoot biomass for genotypes with large and small cortical cells in the mesocosms 30 d after planting in GH2 (A) and GH3 (B) in the mesocosms in both well-watered (WW) and water-stressed (WS) conditions. Data shown are means ± se for three lines per group (n = 4). Means with the same letters are not significantly different (P < 0.05).

Figure 9.

Shoot biomass for genotypes with large and small cortical cells in the field 70 d after planting in PA2011 (A), PA2012 (B), and MW2012-2 (C) in both water-stressed (WS) and well-watered (WW) conditions. Data shown are means ± se for three lines per group (n = 4). Means with the same letters are not significantly different (P < 0.05).

Figure 10.

Grain yield for genotypes with large and small cortical cells in the field in PA2012 (A) and MW2012-2 (B) in both water-stressed (WS) and well-watered (WW) conditions. Data shown are means ± se (n = 4). Means with the same letters are not significantly different (P < 0.05).

DISCUSSION

Our results support the hypothesis that large CCS increases drought tolerance by reducing root metabolic costs, permitting greater root growth and water acquisition. In the soil mesocosms, large CCS was correlated with substantial reduction of root respiration per unit of root length. Under water stress, lines with large CCS had deeper rooting, greater stomatal conductance and leaf CO₂ assimilation, and greater shoot biomass than lines with small CCS. Under water stress conditions in rainout shelters in the United States and in the field in Malawi, lines with large CCS had greater root depth, greater exploitation of water from deep soil strata, greater leaf RWC, and substantially greater grain yield than lines with small CCS.

We observed substantial variation for CCS among diverse maize genotypes, including landraces and RILs. Among RILs, variation was approximately 5-fold. In contrast, 3-fold variation was observed within selected landraces collected across Malawi (Supplemental Table S1). The greatest variation for CCS occurs within the midcortical band (Supplemental Table S1). In contrast, outer and inner bands had smaller levels of variation within a population (Supplemental Table S1). The reduced variation and smaller cells in outer and inner bands could be related to their functions: the inner band represents highly specialized cells with important functions for the regulation of radial transport in the roots, and the outer band provides protection against pathogen entry. In addition, the multiseriate epidermal band cells have been associated with root mechanical strength (Striker et al., 2007). Root CCS was correlated with the size of cells of the mesocotyl cortex and leaf midrib parenchyma. We also found excellent correlation between CCS measured in water-stressed plants and well-watered plants in the field (Supplemental Fig. S5), although the cells were relatively larger in well-watered compared with water-stressed conditions. We also found good correlation of CCS variation between greenhouse mesocosms and field-grown plants. We conclude that phenotypic screening of large numbers of maize genotypes can be conducted effectively in the greenhouse with young plants, which is considerably faster and cheaper than in the field.

Larger CCS was correlated with substantial reductions in specific root respiration (Fig. 2). We propose that as cell size increases, the thickness of the cytoplasm between the plasma membrane and tonoplast is maintained, so that cell enlargement is mostly due to vacuolar enlargement. Increased vacuolar volume relative to cytoplasm may reduce respiration because metabolic activity is greater in the cytoplasm than in the vacuole.

Our results demonstrate that metabolically efficient roots reduce the effects of drought by permitting access to deep soil water (Figs. 3 and 5; Supplemental Figs. S3 and S4). Rooting depth was positively correlated to CCS under water stress, while there was no relationship in well-watered conditions (Figs. 3 and 5). These results suggest that the benefit of reduced metabolic demand for root growth is particularly important under water stress. We interpret the large magnitude of this effect as evidence of an autocatalytic effect, whereby incrementally greater root growth leads to better water acquisition, which positively reinforces root growth via greater shoot carbon gain (Fig. 4A). Indeed, lines with large CCS had greater stomatal conductance and leaf CO₂ assimilation in mesocosms (Fig. 4B) and greater RWC in the field than lines with small CCS under water stress, which was directly related to rooting depth (Fig. 3).

An additional benefit of reduced root costs could be reduced competition for photosynthates among competing sinks, including developing seeds. In maize, yield losses due to drought are related to carbohydrate availability during the reproductive phase (Boyer and Westgate, 2004). It is difficult to distinguish this indirect benefit of large CCS for yield from the more direct benefit of CCS for root growth and soil exploration, because of the tightly coupled integration of water stress effects on photosynthesis, reproduction, and source-sink relationships. Structural-functional plant modeling may provide useful insights in this context by allowing the quantification and independent manipulation of resource allocation among competing plant sinks. Functional-structural plant model SimRoot has provided such insights in the context of the effects of RCA on internal resource allocation in maize (Postma and Lynch, 2011a, 2011b). However, SimRoot is not currently parameterized for reproductive growth.

The δ¹⁸O signature in soil water is used as a natural tracer for water sources captured from the soil, because no isotopic fractionation occurs during water uptake and transport (Ehleringer and Dawson, 1992; Dawson and Pate, 1996; Ehleringer et al., 2000). We used natural variation in the δ¹⁸O signature of soil water in the profile to provide insight into the potential link between root depth and water acquisition (Figs. 4–7). The net effect of evaporation is an enrichment of heavy isotopes in the topsoil. In the subsoil, the isotope signature is attributed to the combination of the evaporation effect and the isotopic signatures of irrigation water and rainfall, resulting in a gradient of δ¹⁸O with soil depth (Fig. 6). In this study, we hypothesized that large CCS improves drought tolerance in maize by reducing the metabolic cost of soil exploration, enabling deeper soil exploration and greater water acquisition. Stem water δ¹⁸O signatures showed that lines with large CCS had 32% lighter isotope signatures and greater dependency on deep soil water than lines with small CCS (Table III). The difference in stem water δ¹⁸O between lines with large CCS and small CCS could be attributed to their differences in rooting depth (Figs. 3 and 5).

We found that large root CCS was associated with large cell size of parenchyma in the leaf midrib, and we anticipate that this relationship also holds with other parenchyma cells in the leaf. Large leaf cells might influence the metabolic efficiency for light capture per unit of volume, by analogy with our hypothesis that large root cortical cells reduce the metabolic cost of capturing soil resources. However, the relationship between cell size and photosynthesis remains unresolved, as illustrated by the contradictory hypotheses put forward by several authors. A negative correlation between photosynthesis and mesophyll cell size in several species was reported (El-Sharkawy and Hesketh, 1965; Wilson and Cooper, 1970; El-Sharkawy, 2009), attributed to the increase in cell surface area per volume with reduced cell size, but Dornhoff and Shibles (1976) observed no correlation between photosynthesis and cell size in soybean (Glycine max). In contrast, others (Warner et al., 1987; Warner and Edwards, 1988, 1989) argued that large cells have greater photosynthetic capacity than smaller cells. Thus, our understanding of the relationship between cell size and photosynthesis is still rudimentary and merits further research. It is noteworthy that, in this study, we found that lines with large root CCS had greater photosynthetic rates than lines with small CCS under water stress, while there were no differences in well-watered conditions (Fig. 4A).

We attempted to employ near-isophenic RILs with common root phenotypes other than CCS (Supplemental Tables S2 and S3) to evaluate the utility of CCS under water-limited conditions. Near-isophenic RILs permit the analysis of the physiological effects of variation in CCS while holding other aspects of the plant phenotype as constant as possible to minimize the confounding effects of other root phenes. Therefore, differences in growth between large CCS and small CCS are most readily explained by variation of CCS rather than by other root anatomical differences. In addition, we found that there was no correlation between CCS and RCA. These results are consistent with other studies. Burton et al. (2013), working with a large population of Zea spp., reported that RCA was not correlated to any anatomical phene. Genotypic variation for CCS was not associated with genotypic variation for other obvious features of the plant phenotypes observed in nonstressed plots other than those reported here.

In many low-input agricultural systems, drought and low soil fertility are primary constraints to crop production. We anticipate that large CCS may have special utility in low-input systems for increasing the acquisition of deep soil resources like nitrate and water, particularly in leaching environments. The utility of CCS for increasing the acquisition of these deep soil resources may interact with other root phenes such as steep root angles (Lynch, 2013; York et al., 2013).

Optimum plant density is a key consideration for maximizing maize grain yield (Cox and Cherney, 2012; Reeves and Cox, 2013). In this study, all field trial plants were planted at the same standard density of 53,000 plants ha⁻¹, while in the greenhouse, one plant was planted per mesocosm. The fact that our results in mesocosms and two field environments agree with each other is evidence that plant density did not affect the utility of CCS in the conditions used in this study. However, in low-input systems, many farmers plant maize at lower densities than used in this study and also intercrop maize with other crops. In contrast, in high-input systems, economically optimum plant densities for maize are between 74,000 and 89,000 plants ha^–1. The influence of CCS on plant performance under low or high density is unknown. This could possibly be addressed using modeling approaches considering the large number of parameters that may affect this relationship, including other anatomical and architectural features of the root phenotype and varying soil conditions, nitrogen regimes, and precipitation patterns.

Drought and mechanical impedance are two stressors that commonly cooccur in agroecosystems and have a potential to overlap in their impacts on plant growth, consequently affecting the utility of the trait in a particular environment. Small cells have a greater density of cell walls per volume, providing rigidity and strength, and thus are more resistant to buckling and deflection than large cells (Weijschedé et al., 2008). This is particularly important for root penetration in hard soil, which is common under drought. On the other hand, as demonstrated in this study, large cells are important in reducing the metabolic cost for soil exploration, permitting root growth into deeper soil domains and thereby enhancing water acquisition under drought. In addition, large CCS may also interact with other phenes that enhance root penetration in hard soils, such as root diameter and root hairs, and may synergistically enhance root penetration under combined stress of drought and mechanical impedance. In this context, it is noteworthy that we observed benefits to large CCS whether plants were grown in soil mesocosms, a silt loam soil in the United States, or a sandy clay loam soil in Malawi. These potential tradeoffs and synergisms should be understood better to guide crop breeding programs.

These results add to a growing body of evidence that phenes and phene states that reduce the metabolic costs of soil exploration improve the capture of limiting soil resources (Lynch and Ho, 2005; Lynch, 2014; Lynch et al., 2014). Such phenes include the production of an optimal number of root axes, biomass allocation to metabolically efficient root classes, and reduced tissue respiration (Miller et al., 2003; Jaramillo et al., 2013; Lynch, 2014; Saengwilai et al., 2014b). CCS is an example of an anatomical phene that affects the metabolic costs of soil exploration by affecting tissue respiration. RCA also affects tissue respiration by converting living cortical cells to air spaces through programmed cell death. Maize genotypes with high RCA formation have reduced root respiration, greater rooting depth, greater water acquisition under drought (Zhu et al., 2010), and greater nitrogen acquisition under nitrogen limitation (Saengwilai et al., 2014a). Similarly, maize genotypes with reduced cortical cell file number have less root respiration, greater rooting depth, and greater water acquisition under drought (Chimungu et al., 2014). The deployment of root phenotypes with greater metabolic efficiency of soil exploration represents a novel, unexploited paradigm to develop crops with greater resource efficiency and resilience (Lynch, 2014).

CONCLUSION

Our results demonstrate that large CCS improves drought tolerance in maize by reducing the metabolic costs of soil exploration. Large CCS substantially reduces root respiration. Under water stress, lines with large CCS had deeper roots, better exploitation of deep soil water, greater plant water status, greater leaf photosynthesis, and greater shoot biomass and grain yield than lines with small CCS. Our results are entirely supportive of the hypothesis that large CCS reduces the metabolic costs of soil exploration, leading to greater water acquisition in drying soil (Lynch, 2013). There is substantial variation for CCS in maize that can be exploited for crop improvement. We suggest that large CCS may have broad relevance in graminaceous crop species lacking secondary root growth, including rice (Oryza sativa), wheat (Triticum aestivum), barley (Hordeum vulgare), oat (Avena sativa), sorghum (Sorghum bicolor), and millet (Pennisetum glaucum). Large CCS may also be useful for nitrogen capture in leaching environments. Although potential fitness tradeoffs for large CCS are not known, it is noteworthy that we observed substantial benefits to large CCS whether plants were grown in soil mesocosms, a silt loam soil in the United States, or a sandy clay loam soil in Malawi.

MATERIALS AND METHODS

Plant Materials

Based on preliminary experiments conducted under optimal conditions in the field and greenhouse, a set of 16 intermated B73 by Mo17 (IBM) lines (Supplemental Table S2) of maize (Zea mays) was used to assess the impact of phenotypic variation of CCS on root respiration (GH1). A set of six IBM lines contrasting in CCS was selected for the GH2 and PA2011 experiments, and another set of six IBM lines also contrasting in CCS was selected for the GH3 and PA2012 experiments (Supplemental Table S2). The IBM lines are from the intermated population of B73 × Mo17 and were obtained from Shawn Kaeppler (University of Wisconsin; Senior et al., 1996; Kaeppler et al., 2000) and designated as Mo (Supplemental Table S2). In Malawi, a set of 43 maize landraces was screened for CCS variation in the field (MW2012-1; Supplemental Table S1). The landrace entries were obtained from the Malawi Plant Genetic Resource Center. Based on MW2012-1, a set of six landraces contrasting in CCS was selected for MW2012-2 (Supplemental Table S1).

Greenhouse Mesocosm Experiments

A total of three experiments were carried out under the same conditions in 2 consecutive years (Supplemental Table S2). The experiments were conducted in a greenhouse at University Park, Pennsylvania (40°4′N, 77°49′W), using 14-h day/10-h night, 23°C day/20°C night, and 40% to 70% relative humidity. The experiments were carried out with natural light between 500 and 1,200 μmol photons m⁻² s⁻¹ of photosynthetically active radiation, and supplemental light between 500 and 600 μmol photons m⁻² s⁻¹ of photosynthetically active radiation was provided with 400-W metal-halide bulbs (Energy Technics) for 14 h per day. Plants were grown in soil mesocosms consisting of polyvinyl chloride cylinders 1.5 m in height by 0.154 m in diameter, lined with transparent high-density polyethylene film, which was used to facilitate root sampling. The growth medium consisted of 50% (v/v) commercial-grade sand (Quikrete), 35% (v/v) vermiculite (Whittemore), 5% (v/v) Perlite (Whittemore), and 10% (v/v) topsoil (Hagerstown silt loam top soil: fine, mixed, mesic Typic Hapludalf). Mineral nutrients were provided by mixing the medium with 70 g per mesocosm of Osmocote Plus fertilizer (Scotts-Sierra Horticultural Products) consisting of (w/w each) 15% nitrogen, 9% phosphorus, 12% potassium, 2.3% sulfur, 0.02% boron, 0.05% copper, 0.68% iron, 0.06% manganese, 0.02% molybdenum, and 0.05% zinc. Seeds were germinated in darkness at 28°C ± 1°C for 2 d prior to transplanting two seedlings per mesocosm and then thinned to one per mesocosm 5 d after planting.

At harvest (i.e. 30 d after planting), the shoot was removed, the plastic liner was extracted from the mesocosm and cut open, and the roots were washed carefully by rinsing the medium away with water. This allowed us to recover the entire plant root system. Samples for root respiration measurement were collected 10 to 20 cm from the base of three representative second whorl crown roots per plant. Root respiration (CO₂ production) was measured 15 to 20 min after cutting the shoot using a portable infra-red gas analyzer (Li-Cor 6400, Li-Cor Biosciences) in closed-system mode equipped with a 56-mL chamber. The change in CO₂ concentration in the chamber was monitored for 3 min. During the time of measurement, the chamber was placed in a temperature-controlled water bath at 27°C ± 1°C. Following respiration measurements, root segments were preserved in 75% ethanol for anatomical analysis as described below.

Root length distribution was measured by cutting the root system into seven segments in 20-cm depth increments. Roots from each increment were spread in a 5-mm layer of water in transparent plexiglass trays and imaged with a flatbed scanner equipped with top lighting (Perfection V700 Photo; Epson America) at a resolution of 23.6 pixels mm⁻¹ (600 dots per inch). Total root length for each segment was quantified using WinRhizo Pro (Regent Instruments). Following scanning, the roots were dried at 70°C for 72 h and weighed. To summarize the vertical distribution of the root length density, we used the D₉₅ (Schenk and Jackson, 2002).

Root segments and leaves were ablated using laser ablation tomography (B. Hall and J.P. Lynch., unpublished data) to obtain images for anatomical analysis. In brief, laser ablation tomography is a semiautomated system that uses a pulsed laser beam (Avia 7000; 355-nm pulsed laser) to ablate root tissue at the camera focal plane ahead of an imaging stage. The cross-section images were taken using a Canon T3i camera with 5× micro lens (MP-E 65 mm) on the laser-illuminated surface. Root images were analyzed using RootScan, an image-analysis tool developed for analyzing root anatomy (Burton et al., 2012). CCS was determined from three different images per root segment. CCS was calculated as median cell size. Based on preliminary experiments, the cortex was divided into three radial bands: outer (0–0.25 of the cortex from the epidermis), midcortical (0.25–0.75), and inner (0.75–1). In this study, the median cell size for the midcortical band was chosen as a representative value for the root CCS.

Experiment I (GH1)

A randomized complete block design was used in this experiment, with time of planting as a blocking factor replicated three times. A set of 16 IBM lines (Supplemental Tables S2 and S3) was planted, and water stress was imposed by withholding water starting 14 d after planting. Plants were harvested for root respiration measurements and anatomical analysis 35 d after planting. An experiment was conducted to characterize the pattern cell size variation in different parts of the day maize plant. A set of 12 IBM lines was planted in nutrient solution until the two-leaf stage. In this experiment, the leaf midrib, mesocotyl, and primary root of each plant were sampled for sectioning. To assess the relationship between cortical cell cross-sectional area and cell length, cortical cell lengths and diameter were measured on laser-ablated root longitudinal sections. These measurements were made on the midcortical region on each of the three sections for each genotype. The cell length or diameter on each section was taken as the median of measurements on 20 to 30 cells.

Experiments II (GH2) and III (GH3)

Two experiments were conducted, in fall 2011 (GH2) and summer 2012 (GH3). A set of six genotypes was planted in each experiment (Supplemental Tables S2 and S3). A randomized complete block design with four replications was used in both experiments with time of planting as a blocking factor. Planting was staggered by 7 d. In both experiments, the irrigated mesocosms (control) each received 200 mL of water every other day, to replenish water lost by evapotranspiration, and in stressed mesocosms, water application was withheld starting 5 d after planting to allow the plants to exploit residual moisture to simulate terminal drought. An SC-1 leaf porometer (Decagon) was used for stomatal conductance measurements from the abaxial side of third fully expanded leaves at 28 d after planting in GH3. All of the stomatal conductance measurements were made between 9 and 11 am. Plants were harvested 30 d after planting for root respiration measurements, root length distribution, and shoot biomass. The dry matter of the shoot and root was measured after drying at 70°C for 72 h, and root length distribution was determined as described above.

Field Experiments, Rock Springs, Pennsylvania

Field Sites and Experimental Setup

Two experiments were conducted in rainout shelters located at the Russell E. Larson Agricultural Research Center in Rock Springs, Pennsylvania (77°57′W, 40°42′N) during the summers of 2011 (PA2011) and 2012 (PA2012). The soil is a Hagerstown silt loam (fine, mixed, mesic Typic Hapludalf). Both experiments were arranged as split plots in a randomized complete block design with four replications. The main plots were composed of two moisture regimes, and the subplots contained six lines contrasting in CCS. The plants were hand planted on June 15, 2011, and June 25, 2012. Plants in both trials were planted in three-row plots with 0.75-m interrow spacing and 0.25-m in-row spacing to give a plant population of 53,000 plants ha⁻¹. The drought treatment was initiated 35 to 40 d after planting using two automated rainout shelters. The shelters (10 × 30 m) were covered with a clear greenhouse plastic film (0.184 mm) and were automatically triggered by rainfall to cover the plots, excluding natural precipitation throughout the entire growing season. The shelters automatically opened after rainfall, exposing experimental plots to natural ambient conditions whenever it was not raining. Adjacent nonsheltered control plots were rainfed and drip irrigated when necessary to maintain the soil moisture close to field capacity throughout the growing season. Soil moisture content at different soil depths (20, 35, and 50 cm) was monitored at regular intervals (Supplemental Fig. S5) using a TRIME FM system (IMKO).

Plant Measurements

Leaf RWC was used as a physiological indicator of plant water status. To measure leaf RWC, fresh leaf discs (3 cm in diameter) were collected from the third fully expanded leaf for three representative plants per plot 60 d after planting and weighed immediately to determine fresh weight. The discs were floated in distilled water for 12 h at 4°C with minimal light. Discs were then blotted dry and again weighed to determine turgid weight. After being dried in an oven at 70°C for 72 h, discs were weighed again for dry weight. Leaf RWC was calculated according to Barrs and Weatherley (1962).

Root growth and distribution were evaluated by collecting soil cores 80 d after planting. A soil coring tube (Giddings Machine) 5.1 cm in diameter and 60 cm long was used for sampling. The core was taken midway between plants within a row. The cores were sectioned into six segments of 10-cm depth increments and washed. Subsequently, the washed roots were scanned using a flatbed scanner (Perfection V700 Photo; Epson America) at a resolution of 23.6 pixels mm⁻¹ (600 dots per inch) and analyzed using the image-processing software WinRhizo Pro (Regent Instruments).

Shoots and roots were evaluated 75 d after planting. To accomplish this, shoots from three representative plants in each plot were cut at soil level. The collected shoot material was dried at 70°C for 72 h and weighed. Root crowns were excavated by the shovelomics method (Trachsel et al., 2011). Three 8-cm root segments were collected 10 to 20 cm from the base of a second whorl crown root of each plant and used to assess CCS. The segments were preserved in 75% ethanol before being processed as described above. At physiological maturity, grain yield was collected from 10 plants per plot.

Field Experiments, Lilongwe, Malawi

Assessing Phenotypic Variation of CCS in Malawi Germplasm (MW2012-1)

The experiment was conducted at Bunda College research farm, Lilongwe, Malawi (33°48′E, 14°10′S), in 2012 under optimum conditions (i.e. the plots were rainfed but only rarely were they severely moisture stressed). The soil is a Lilongwe series sandy clay loam (Oxic Rhodustalf). The experiment was arranged as a randomized complete block design with three replications. Each plot consisted of a single 6-m-long row and 25 cm between plants. Root crowns were excavated by shovelomics (Trachsel et al., 2011). Three 8-cm root segments were collected 10 to 20 cm from the base of a representative second whorl crown root of each plant and used to assess CCS. The segments were preserved in 75% ethanol before being processed as described above.

Utility of CCS under Water-Limited Conditions (MW2012-2)

The field experiment was conducted at Bunda College research farm, Lilongwe, Malawi (33°48′E, 14°10′S), during summer 2012 (i.e. August to November). The area was selected because it is the main maize production area, accounting for more than 20% of the area planted to maize in Malawi per annum, and regarded as having representative soil types and agronomy for maize-growing areas in Malawi. A set of six maize genotypes contrasting in CCS was planted (Supplemental Tables S2 and S4). The experiment was arranged as a split plot in a randomized complete block design with four replications. The main plots were composed of two moisture regimes, and the subplots contained six genotypes contrasting in CCS. The trial plants were planted in single-row plots with 0.75-m interrow spacing and 0.25-m in-row spacing to give a plant population of 53,000 plants ha⁻¹. At planting, both the control and stressed plots received the recommended amounts of irrigation. Drought stress was managed by withholding irrigation starting 6 weeks after planting, so that moisture stress was severe enough to reduce yield and shoot biomass by 30% to 70%. Control plots, which received supplementary irrigation, were planted alongside the stressed plots separated by a 5-m-wide alley. At each location, the recommended fertilizer rate was applied during planting and by top dressing 3 weeks after planting. Leaf RWC was determined 60 d after planting as described above. Shoots and roots were evaluated 75 d after planting. The collected shoot material was dried at 70°C for 72 h and weighed. Root crowns were excavated by shovelomics (Trachsel et al., 2011). Three 8-cm root segments were collected 10 to 20 cm from the base of a representative second whorl crown root of each plant and used to assess CCS. The segments were preserved in 75% ethanol before being processed as described above. At physiological maturity, grain yield was collected from each plot.

Soil and Plant Sampling for δ¹⁸O Analysis

In PA2011, soil samples were collected adjacent to plants in the rainout shelter 65 d after planting using a 5-cm-diameter soil core. Soil cores were taken to the maximum achievable depth of 60 cm. The cores were immediately separated into 10-cm increments: 10, 20, 30, 40, 50, and 60 cm. The corresponding maize stems were collected at the same time that soil was sampled, approximately 8 to 10 cm of the stem was collected just above ground level, and the epidermis was immediately removed. Soil and maize stem samples were put in snap vials, sealed with Parafilm to prevent evaporation, and refrigerated immediately. Cryogenic vacuum distillation (West et al., 2006; Koeniger et al., 2010) was used to extract soil water and crop stem water. In cryogenic vacuum distillation, two glass tubes were attached to a vacuum pump. The sample was placed in one tube and frozen by submerging the tube in liquid nitrogen, and then both tubes were evacuated by vacuum pump to create a closed U-shape configuration. After that, the tube containing sample was heated, while the collection tube was still immersed in liquid nitrogen to catch the vapor. Samples were weighed and oven dried after extraction to ensure that the extraction time was sufficient to vaporize all the water in the samples. The water samples were analyzed at the Penn State Institutes of Energy and the Environment. Stable isotopic analyses were performed using an L2130-i δD/δ18O Ultra High Precision Isotopic Water Analyzer (Picarro). Results were expressed as parts per thousand deviations from the Vienna Standard Mean Ocean Water. To determine the percentage contribution of soil water at depth to the signature of water within the plant’s xylem, an isotopic mixing model was used (Phillips et al., 2005). IsoSource version 1.3.1 (Phillips and Gregg, 2003) was used to evaluate the relative contribution of each soil layer to plant xylem water signature. The fractional increment was set at 1% and tolerance at 0.1.

Data Analysis

Data from each year were analyzed separately, since different sets of genotypes were used. For mesocosm data, for comparisons of genotypes, irrigation levels, and their interaction effects, a two-way ANOVA was used. Field data were analyzed as a randomized complete block split-plot design to determine the presence of significant effects due to soil moisture regime, genotype (or phenotype group), and interaction effects on the measured and calculated parameters. Mean separation of genotypes for the different parameters was performed by Tukey’s honestly significant difference test. Unless noted otherwise, honestly significant difference values were only reported when the F test was significant at P ≤ 0.05. Linear regression analysis was used to establish relationships between CCS and measured or calculated parameters. Data were analyzed using R version 3.0.0 (R Development Core Team, 2014).

Supplemental Data

The following materials are available in the online version of this article.

• Supplemental Figure S1. Correlation between cortical cell diameter and cell length.

• Supplemental Figure S2. Transverse section showing parenchyma cells in maize.

• Supplemental Figure S3. Root length density at different soil depths for genotypes with large CCS and small CCS under water-stressed and well-watered conditions in the greenhouse with corresponding D₉₅.

• Supplemental Figure S4. Root length density at different soil depths for genotypes with large CCS and small CCS under water-stressed and well-watered conditions in the field (PA2011) with corresponding D₉₅.

• Supplemental Figure S5. Change in soil moisture content at different depths in well-watered and water-stressed plots for PA2012.

• Supplemental Table S1. Descriptive statistics of root cortical cell size variation for maize in the greenhouse and in the field (MW2012-1).

• Supplemental Table S2. Summary of the experiments.

• Supplemental Table S3. List of genotypes selected from IBM population and root anatomical phenes.

• Supplemental Table S4. List of genotypes selected from Malawi maize breeding program populations and root anatomical phenes.

Glossary

LCA

living cortical area

RCA

root cortical aerenchyma

CCS

cortical cell size

RIL

recombinant inbred line

D₉₅

depth above which 95% of total root length is located in the soil profile

RWC

relative water content