Introduction

The role of (deep) roots

Roots play a crucial role in a plant’s functions by taking up water and nutrients. Deeper roots enable a plant to reach deeper soil layers, which can be important particularly at later developmental stages, when there is less available water in the surface layers, and water is crucial for yield formation (Thorup-Kristensen and Kirkegaard 2016). During dry conditions, this can help to create resilience to climate change by ensuring access to water. In Australia, deep roots are most important during grain filling as they can access stored deep water, which is more reliable than precipitation (Wasson et al. 2012). While droughts are not so severe in northern Europe as elsewhere, water availability is becoming more erratic; spring and early summer droughts are increasingly common. This affects crops during the reproductive stages and causes yield losses. To develop crop varieties resilient to more variable precipitation patterns, deep rooting should be included in phenotyping and breeding efforts.

In northern Europe, excessive winter precipitation can cause leaching of nitrate deep into the soil, where it is lost from the crop system, polluting water sources and increasing fertiliser costs for farmers. Deep roots can help by increasing N uptake from deep soil layers, enhancing N efficiency and reducing water pollution (Thorup-Kristensen et al. 2009). Therefore, deep rooting is listed as one of the ‘priority traits’ for improving N uptake efficiency (NUP) and thereby reducing the need for excessive fertiliser inputs (Foulkes et al. 2009).

The study of roots so far

In order to breed for deep roots, we have to be able to identify genotypic differences in rooting systems. Roots are difficult to study, as they are hidden by soil, and therefore any attempt to see them requires destruction of the plant or developing methods which allow us to see them in their natural environment. For this reason, root research is still at an early stage, and we know far less about roots than about the plant we see above ground. New methods and technologies are needed for uncovering roots and their functions.

Genetic differences in deep rooting

Genotypic differences in root depth development in wheat have been found at early vegetative stages (Chen et al. 2020; Liao et al. 2006; Rich et al. 2020) and at maturity in controlled root chambers (Manschadi et al. 2006) and tube experiments (Brunel-Saldias et al. 2020; Ytting et al. 2014).

At the field level, however, finding genotypic differences is more complex, amongst the numerous factors that influence root growth in real conditions. A number of field studies have shown genotypic differences in rooting depth, some limited to a few genotypes (Postic et al. 2019; Rasmussen et al. 2015). There are also some field studies of a large number of genotypes, using different methods: a previous study in the RadiMax facility suggested that there were genotypic differences in deep rooting which affected deep N uptake in winter wheat, however more repetitions were needed to verify the genotypic differences (Wacker et al. 2022). In Australia, genotypic differences of 17-39 cm root depth from soil coring were found in 24 wheat genotypes, with a significant interaction in root depth between different genotypes and environments (Botwright Acuña and Wade 2012). In a large panel of Indian and Australian wheat genotypes, rooting depth varied by 27 cm among genotypes, and were consistently deeper in Indian genotypes (Rich et al. 2016).

Genotypic differences in roots and water stress tolerance

The next step is to investigate how deep rooting affects root function and resource acquisition. Genotypes with deeper root growth may have increased water and N uptake. For example, the ‘deep, steep & cheap’ ideotype is expected to increase water and N acquisition (Lynch 2013). Wacker et al. (2022) demonstrate the higher uptake of 15N labelled deep nitrogen due to deeper rooting in winter wheat. However, there is little evidence as yet to show that genotypes with deeper roots are more resilient to drought stress.

Drought tolerance has been linked to deep rooting in controlled experiments, with a small number of genotypes (Figueroa-Bustos et al. 2020; Manschadi et al. 2006), and several studies have shown that deep rooting is related to increased water uptake in the field. In a UK trial of 21 spring wheats, genotypic differences were found in soil water uptake, particularly at soil depths around 0.8 m, which correlated positively with drought tolerance, measured by the difference in grain yield between droughted and well-watered plots (Ober et al. 2014). In Mexico, genotypic differences in deep root mass between 8 wheat genotypes under drought stress, as measured by soil cores down to 120 cm, were associated with drought tolerance, indicated by carbon isotope discrimination (δ13C) in the grain, and cooler canopy temperatures (Lopes and Reynolds 2010). They suggest that better use of deep water resources meant delayed senescence and therefore increased transpiration and yield. Simulated models show that genotypes with faster and more efficient root systems could increase water uptake from dry soils significantly (by up to 40 mm in Australian soils), benefitting yields (Lilley and Kirkegaard 2011).

Quantifying deep roots

There is the further problem of quantifying ‘deep’ roots, as root system development can vary greatly depending on the climate or soil they grow in. Additionally, methods for measuring roots are less well established and able to give precise measurements than typical methods for above-ground plant features are, making it difficult to establish precise quantitative measures of root traits. In the past, studies of rooting depth have used estimates of maximum root depth or total root length as a measure of root depth (Botwright Acuña and Wade 2012; Rasmussen et al. 2015). More recently, new ‘traits’ have been evaluated which aim to address specific aspects of root function, including measures such as root length or root surface density (Postic et al. 2019), and various traits focussing on deep root density have been linked to deep nitrogen uptake in winter wheat (Wacker et al. 2022).

Research into N effects on root growth

While the effects of N on above ground growth are quite well known, for roots it is less clear. Based on controlled studies, we know that root growth will increase in soil with higher N availability (Robinson 1994), while a mild N deficiency can stimulate a ‘foraging response’, encouraging lateral root growth, and severe deficiency will trigger a ‘survival response’, limiting total root length (Giehl and von Wirén 2014; López-Bucio et al. 2003). Griffiths et al. (2022) used X-ray microcomputed tomography (CT) to scan seedlings in a soil column within PVC tubes, finding that under a low N treatment there were both more seminal and lateral roots and an increased root count.

However, in the field these responses are much more complex, and localised effects cannot easily be applied to a whole crop root system in the field. Conditions vary depending on many factors; in particular water availability can have an important effect on the interactions of N and roots. When N supply is not limited, then increased N availability may not affect root growth (Robinson 1994).

Interactions of water and N

Water and nitrogen effects on crops show complex interactions as N is dependent on water for uptake via diffusion and mass flow. However, N availability may also affect uptake of water in ways that are not fully understood; and even less so in the more complex conditions of the field. Under combined N and water stress under rainfed field conditions in France, roots were deeper and yields were higher under drought with a low N compared to a high N treatment (Postic et al. 2019). Under irrigated field conditions in China, root growth was strongest under moderate N (180 kg N/ha) application, while under water stress, reduced N increased root growth (Wang et al. 2014). However, under severe and prolonged drought, deep root growth increased in winter wheat under high N treatment (Barraclough et al. 1989), and drought decreased the root/shoot ratio while reduced N increased it (Barraclough et al. 1991). Limited N may improve a plant’s drought tolerance by reducing overall plant growth and extent of stomata opening, thereby reducing its water requirement and leaving it less susceptible to drought.

This study

The purpose of this study is to investigate differences in deep rooting among a selection of winter wheat genotypes of different backgrounds, and its effect on their resilience. We will study whether deep-rooting genotypes do take up more resources from the deep soil layers, and whether it makes them better equipped to deal with water and nitrogen stress in northern European conditions. The RadiMax facility (Svane et al. 2019) allows for large scale phenotyping of deep roots, with 600 minirhizotron tubes, and Rootpainter software (Smith et al. 2022) for efficient analysis of thousands of images, allowing for genotypic comparisons and imaging over multiple time points. Furthermore, the semi-field nature of the facility allows for some control of conditions, i.e. imposed drought, while also allowing for deep isotope tracer injection and water sensors to more closely follow dynamics of nutrients and water.

This study has the following hypotheses:

Hypotheses

  1. 1.

    There are genotypic differences in deep root growth, and in response to N treatment.

  2. 2.

    Genotypes differ in their resilience to water stress, N limitation and their interactions.

  3. 3.

    Genotypes with deeper roots will be those less affected by water stress.

Materials and methods

The facility

The RadiMax root phenotyping facility at the University of Copenhagen, Denmark, is situated in Taastrup (55.66815°N, 12.30848°E). It is a semi-field root phenotyping facility, consisting of four 40 m long and 9.7 m wide raised concrete beds filled with soil. Each bed is separated from the bulk soil below by a membrane and a gravel layer. The bottom of the facility is V-shaped, sloping from the concrete sides to the middle, reaching 3 m depth at the centre in the two beds. Crop lines were grown 25 cm apart width-wise across the bed. Acrylic minirhizotron tubes 5.5 m long with a 70 mm external diameter are permanently installed in the facility, directly below the crop lines. The tubes are placed at a 24° angle down the sloping sides, allowing for root imaging of the root systems from progressively deeper in the soil, from 0.6 m at the concrete side to a maximum of 2.7 m soil depth in the centre. Irrigation tubes run lengthwise along the sloping base of the beds at 50 cm intervals, allowing for subsoil injection of water and isotope tracers at different depths. Water sensors (TDT volumetric water content (VWC) (Acclima, Inc., Boise, ID, USA)) are located centrally at two positions in each bed, at 50 cm depth intervals. A further set of water sensors are located along the slope of each bed at 20 cm soil depth intervals at two positions per bed. The facility was filled with sandy-loam soils from local fields in 2015, with a nutrient rich topsoil in the top 40 cm and a nutrient-poor subsoil in the remainder of the bed (Svane et al. 2019). Further details of the facility can be in found in Wacker et al. (2022) and Svane et al. (2019).

Genetic material

We tested 14 genotypes of winter bread wheat grown over 2 years, which included breeder lines and commercial cultivars with release dates from 1981 to 2017. Most of the lines were selected from a very diverse panel of ~400 used in the SolACE project (www.solace-eu.net), originating from across Europe and North America. We chose the most modern and geographically relevant cultivars from the panel. Further, two modern Danish commercial wheat genotypes were added to the experiment (see Table 1 for list of genotypes). Each genotype was replicated in 4 crop rows each year, except Creator which had 8. The seeds in 2019 were treated with fungicide and insecticide before sowing.

Table 1 Wheat genotypes used in the experiment
Full size table

Design and management

The wheat was sown in individual rows in early October, in rows 9.7 m long, with 52 seeds sown per metre, and 25 cm between rows. The experiment was divided into 2 blocks (one in each bed of the facility), with each genotype replicated twice per block. The genotypes were randomized within the block, and arranged in pairs so that each alternating pair of rows had a high or low N treatment (see Table 2, Fig. 1B and explanation below). Rainout shelters on rails were placed over the facility from anthesis (14/6/2019 and 6/6/2020) to induce drought effects during grain filling.

Table 2 Management, N treatments, isotope application and imaging dates of semi-field experiments, 2019-2020
Full size table
Fig. 1
figure 1

The RadiMax facility in Taastrup, Denmark. A) Deep and shallow soil profile treatments. The differences in water treatment are due to the differences in the soil profile depth in the deep (1.8-2.8 m depth) and shallow (0.6-1.6 m) parts of the bed, rather than different irrigation levels. Also shown are tracer injection lines (blue) running lengthways along the bed, and minirhizotron tubes (dashed line) which lay along the slope of the facility. Crop rows run across the bed above the minirhizotron tubes. B) Crop rows were arranged in pairs of each genotype (G1, G2 etc), ordered randomly. The dotted line shows the additional N treatment, which was applied in between pairs of genotypes to give alternating high (dark green) and low (light green) N treatments

Full size image

Treatments

Nitrogen fertilizer was applied at two levels: a high treatment of 190 kg N ha−1 (N190) and a low treatment of 60 kg N ha−1 (N60), in three applications (Table 2). At each application, firstly the low N dose was applied to the whole bed. After this the high N treatment based on single row plots was established by applying extra fertilizer in every fourth interrow area, whereby half the lines were grown next to an interrow area with extra N application, and the other half of the lines were not (Fig. 1B).

A drought gradient occurs in the facility due to the ‘V’ shape of the beds. When sampling, the row was divided into 4 sampling areas: two shallow samples on the outside; two deep samples in the middle. The shallow samples at the outer edge of each bed have a soil depth ranging from 0.6-1.6 m. The central area of each bed has a soil depth ranging from 1.8-2.8 m. This is referred to as the deep sample, as crop roots can access deeper soil and thereby more water. Therefore a ‘wet’ and ‘dry’ treatment occurs at each side of the bed within each of the rows. A further labelled sample was taken from the area in between the shallow and deep samples, above the isotope injection, with soil depths from 1.6-1.8 m.

Isotope application and measurement

Isotopic tracers of 15N (Ca15NO3) and 2H labelled water (2H2O) were applied into the subsoil, 15N in both years and 2H in 2020 only. They were injected with pressure into a subsoil irrigation line, running along the length of each side of both beds, at 180 cm depth in 2019; in 2020 the method was adjusted, injecting isotopes into two lines at 160-180 cm soil depth (for dates see Table 2). The isotope mixture of 54.6 L water, 1.51725 g 15N (=8.925 g CaNO3), 1 L 2H labelled water were injected in each side of each bed, giving the equivalent of 0.00759 g 15N and 5 ml 2H labelled water to each side of every crop row. The method was similar to that used by Wacker et al. (2022).

Sampling and sample preparation

The full plant was harvested by hand at ground level at maturity. All plants were harvested for both of the deep and shallow soil profile treatments within each row, and from the labelled area directly above the injection line. These were dried for 24 hours at 80 °C and then threshed and weighed. Measurements were taken of grain, straw and thousand kernel weight (TKW), and grain and straw yield were calculated based on the area of each sample harvested. Straw and grain of both labelled and non-labelled samples were then milled to a fine powder (Cyclone Mill Twister, Retsch, Germany) and packed into aluminium or silver capsules for analysis.

Isotope analysis and calculations

Samples were analysed for δ13C, δ15N, total N and total C at the Stable Isotope Facility, UC Davis, using an isotope ratio mass spectrometer (IRMS) and for δ2H at the Centre for Stable Isotope Research and Analysis, University of Göttingen. Total N and 15N uptake were calculated based on the total sample weight of straw or grain. 15N uptake is corrected for natural abundance of atmospheric 15N of 0.3663 (Junk and Svec 1958). δ13C is measured in grain and straw as an indication of a physiological response to drought, where greater δ13C values indicates more drought stress (Farquhar et al. 1989).

Root imaging and evaluation

Images of the roots were taken in each minirhizotron tube down to 5500 mm tube depth (2.7 m soil depth) at 4 timepoints during each growing season (Table 2), using a multispectral camera to take images at 35 mm intervals (Svane, Dam, et al., 2019). Nearly 30,000 images were taken per imaging campaign. The images were analysed using the Rootpainter software, which is trained to segment root vs non-root in each image, and extracts the root length per image (Smith et al. 2022).

Root traits

Root data for each tube were combined in different ways to calculate a number of ‘traits’ which describe the root profile and deep rooting, based on Wacker et al. (2022). DeepRoot40 (Fig. 2) is selected as the most relevant trait as it has shown strong correlation with δ15N and δ13C values, and has been shown to be an indicator of deep root activity in a previous study of winter wheat in the RadiMax facility, where it is termed TRD_40 (Wacker et al. 2022). It is a measure of the soil depth above the cumulative deepest 40 cm of roots (Fig. 2).

Fig. 2
figure 2

DeepRoot40 – this is a trait for quantifying deep root growth. Starting from the deepest roots, the cumulative root length is added until 40 cm root is measured. This soil depth is the DeepRoot40

Full size image

Data analysis and statistics

Statistical analysis was done using R version 4.2 and R studio (R Core Team 2022), and analysed using gls models (maximum likelihood), with correlation to account for spatial variation across the beds in the RadiMax facility. Statistical significances of variables were tested using analysis of variance (ANOVA) and post hoc pairwise comparisons were tested using the emmeans package (Lenth 2022). R packages including plyr, multcomp, nlme, emmeans, ggplot packages were also used.

Results

Weather and water sensors

Total precipitation during the two growing seasons was similar, with 625 mm in 2019 and 651 mm in 2020. In the first year of the experiment, there was a very wet March followed by a dry April (13 mm precipitation). In the second year of the experiment the winter was milder and had higher autumn/winter precipitation (373 mm October 2019 - February 2020; 120 mm higher than the same period the previous year). In both years, it is clear that the soil profile is filled several times during the winter, with high water levels in all soil layers, followed by rapid drainage. Therefore by early April in both years, there was a water-filled soil profile. In 2019 there was 117 mm total precipitation from 1 April until rain out shelters covered the experiment on 16 June. In 2020 there was a relatively dry spring, with 81 mm total precipitation from 1 April until rain-out tents were used on 6 June (Fig. 3). Water sensors show volumetric water content (VWC) at 50 cm depth intervals throughout the growing season (Fig. 3). These indicate that there was more water use from the soil profile in 2019, and from deeper soil layers, while in 2020 the VWC remained relatively high despite little rainfall in the spring. The VWC at the isotope tracer application depth (180 cm) was similar in both years at the time of the application, 26.1% in 2019 and 25.8% in 2020, falling to 13.6% and 20.9% at maturity (end of July), respectively.

Fig. 3
figure 3

Soil volumetric water content (top) and temperature and precipitation (below) in Taastrup, Denmark (daily mean, min, max temp and monthly precipitation sum). Arrows indicate beginning of drought, when rainout tents were placed over the facility

Full size image

Deep root growth

In both years root growth was observed down to 2 m soil depth and the greatest quantity of deep roots were observed in June (Fig. 4). This is when the roots tend to be clearer and more easily detectable on images, while in July, old roots are starting to fade. The average root profile visibly grew deeper throughout the growing season, although less roots were detectable in the upper soil layers by July. The quantity of roots was less in all soil layers and at each time point in 2020 than in 2019.

Fig. 4
figure 4

Average root profiles across 14 genotypes of winter wheat grown in the RadiMax facility, 2019 & 2020. Values show the mean root length (cm) within each 3.5 cm image, shown in 15 cm intervals

Full size image

Deep rooting varied among genotypes. Deep roots were measured by calculating a number of different traits based on root images; the measure of the rooting depth threshold of 40 cm deep roots in June (DeepRoot40) was selected as the most promising one of these (Fig. 2). Roots grew deeper in 2019 (mean DeepRoot40 137 cm) than in 2020 (mean DeepRoot40 92 cm) (p < 0.05), and in both years the roots grew deeper under the N60 treatment than N190, although only significantly in 2020 (Table 3). In 2019 there were no significant differences between genotypes; in 2020 Mustang had the deepest roots (see Fig. 5). Mean DeepRoot40 by genotype varied between 114 and 151 cm in 2019 and 68-127 cm in 2020.

Table 3 DeepRoot40 under N treatments and years
Full size table
Fig. 5
figure 5

Mean DeepRoot40 by genotype in each year. The values shown indicate the soil depth threshold below which there are 40 cm of root length; the maximum rooting depths that we found were deeper than this. Different letters indicate significant differences between genotypes within year (p < 0.05)

Full size image

Yield: Interactions between water availability and N

Average grain yield was higher in 2019 (607 g/m2) than in 2020 (229 g/m2). Across all treatments, Torril had the highest mean grain yield in 2019 (705 g/m2), and Kvarn in 2020 (758 g/m2). Within the N190 deep soil treatment, Creator had the highest yield in 2019 (820 g/m2), and Kvarn in 2020 (540 g/m2).

While in 2019 the effect of N and soil depth on yield were similar, in 2020 the effect of drought on yield was smaller than the effect of N. TKW was higher in 2019 (p < 0.0001) and there was no effect of N on TKW in either year, but TKW was higher for the deep soil (p < 0.0001) (Table 4).

Table 4 Yield measurements and δ13C under 2 N treatments and soil profile depths, for 2 years
Full size table

Some genotypes were more yield stable across treatments, e.g. CH73641, while Kvarn and Creator (both modern Danish varieties) were more responsive to N treatments (Fig. 6). In 2019 there was also a smaller effect of N level in many cultivars. Grain N% was higher under the high N treatment (Table 4), and there was a significant effect of both genotype and genotype*N (Table 5).

Fig. 6
figure 6

Grain yield (g/m2) under two N and two soil profile depth treatments, mean values and standard errors, for 14 winter wheat genotypes. Different letters show significant differences within genotype and years (p < 0.05)

Full size image
Table 5 Statistical significance of variables
Full size table

In both years there was a significant effect of both N, soil profile depth and genotype on grain yield (p < 0.0001) (Table 5). Deep rooting (DeepRoot40) also had an effect on yield (p = 0.0018) and there was an interaction between year and soil profile depth (p = 0.0121). There were no significant interaction effects on yield between genotype and N or soil profile depth (except in N uptake), but there were differences in how genotypes responded to the different treatments, and in between the years (Fig. 6).

Drought stress, different years, genotypes and treatments

Values of δ13C were lower in 2019 than in 2020 in both grain and straw (Table 4). Grain δ13C was unaffected by soil profile depth in 2019, while in 2020 δ13C was lower in the deep soil; the straw δ13C was lower in the deep soil in both years (Table 4). Comparing the deep soil samples only, in 2019 δ13C was lower in the grain for the N60 treatment than the N190 treatment (p < 0.05) (Table 6). There were significant effects of genotype, year, N and soil profile depth on δ13C values in both grain and straw, and of DeepRoot40 for straw only (Table 5). Genotype differences in δ13C were strongly significant and there was a significant interaction of genotype and year.

Table 6 Isotopes under N treatments and years
Full size table

Yield effects and roots

Deep roots (DeepRoot40) were positively correlated with grain yield in the SoilDEEP section of the facility, i.e. where root growth was not physically limited, in both years and N treatments (Fig. 7). The correlation between yield and deep roots was stronger under the low N treatment in both years.

Fig. 7
figure 7

Correlation of grain yield (deep soil) and DeepRoot40 in June under two N treatments in 2019 and 2020. Values are grouped by genotype means

Full size image

Isotope uptake from deep soil

Uptake of 15N applied at 1.8 m depth, an indicator of deep root activity, was observed in grain in both years. δ15N was higher under N60 than N190 in both years, and higher in 2020 than in 2019 (Table 6). Calculated across genotype means, 15N uptake was positively correlated with deep roots (DeepRoot40) for N60 in 2020 only. For individual row observations, the correlation was positive and significant in both N60 (R = 0.55) and N190 (R = 0.42) treatments in 2019, and in N60 (R = 0.53) in 2020. Genotype effect was approaching significance for δ15N (see Table 5 for p values).

Deuterium (2H) uptake was significantly higher at N190 than N60 (Table 6), but did not vary among genotypes. However there was a positive correlation between DeepRoot40 and δ2H at N60 (Fig. 8c), and DeepRoot40 was approaching significance as a factor for δ2H (p = 0.0598, see Table 5).

Fig. 8
figure 8

Correlations of deep soil deuterium (δ2H) and A) δ13C in grain; B deep applied δ15N in grain; C DeepRoot40 in June. Values are for 2020 and are grouped by genotype means

Full size image

The uptake of the two isotope tracers (δ15N and δ2H) were strongly positively correlated (Fig. 8b), with relatively higher 15N uptake under N60 and relatively lower 2H uptake. DeepRoot40 was positively correlated with both isotopes of 15N and 2H in the grain in 2020, though the correlation to 2H was only significant at N60 (Figs. 8c and 9a).

Fig. 9
figure 9

DeepRoot40 correlations to A) deep soil δ15N in grain; B δ13C grain; C δ13C straw, in 2019 and 2020. Correlations of mean values per genotype. δ13C values are for the deep soil profile

Full size image

Deep roots and drought effects (DeepRoot40, δ15N & δ2H vs δ13C & yield difference)

Grain δ13C was negatively correlated with deep roots (DeepRoot40), indicating less water stress during grain filling. In the straw, δ13C was negatively correlated with DeepRoot40 in N60 in 2020 only (Fig. 9). When looking at individual tubes (rather than genotype means), there is a strongly negative correlation which is significant in both years and N treatments, for both straw and grain (not shown). Both isotopic tracers are correlated negatively to δ13C in 2020, which shows drought stress in the plant (Figs. 8a and 10).

Fig. 10
figure 10

Correlations of deep soil applied δ15N and δ13C values in the grain at maturity in 2019 and 2020. Values are genotype means

Full size image

There was also a positive correlation of individual row observations (rather than genotype means) between the percentage difference in grain yield between the treatments (GrainYieldDEEP – GrainYieldSHALLOW)/GrainYieldDEEP*100 and the root depth, in both years and under both N treatments (not shown). This means that in rows where deeper roots were observed, there was a greater difference between the deep and shallow soil grain yield. When grouped by genotype, there was a positive correlation in 2020 between DeepRoot40 and yield difference, although this was significant for N60 only, and negative in 2019 for N190 (Fig. 11a). The difference in grain δ13C between deep and shallow soil samples was also positively correlated with deep roots for N60 in 2020, meaning a greater difference in δ13C between the deep and shallow soil sample when there were deeper roots (Fig. 11b).

Fig. 11
figure 11

Drought effects. A Difference in grain yield between deep and shallow treatments (%) correlated with DeepRoot40 under two N treatments. Grain yield difference = (GrainYieldDEEP) - GrainYieldSHALLOW/ GrainYieldDEEP *100. B δ13C difference in grain between deep and shallow soil profile (%) correlated with DeepRoot40 in June, under two N treatments. δ13C difference = (δ13C DEEP– δ13CSHALLOW)/ δ13CDEEP*100 . Values are grouped by genotype

Full size image

Discussion

The RadiMax facility

Minirhizotrons in combination with deep application of isotopic tracers allow us to corroborate the root observations with water and nitrogen uptake. While root activity measured via soil water dynamics is an important method (Ober et al. 2014), it can be highly laborious. In the past, a major limitation of analysing roots has been image processing (Postic et al. 2019). This is reduced by using AI software such as Rootpainter (Smith et al. 2022), which allows for rapid analysis of large image sets (we analysed ~90,000 images per year). This is promising for the future of large-scale root phenotyping.

The facility is not entirely representative of field conditions; in order to test a large number of phenotypes and repetitions, crops are grown in single rows, which means that each row may have some border effects. An earlier experiment in the RadiMax facility has shown with 15N tracer applied along the minirhizotron tubes that the roots grow mainly directly above the tube, with very minimal spread of roots to neighbouring rows (Chen et al. 2019). Replication under real field conditions would be helpful to verify our results. The deep soil isotope injection is unique to the RadiMax facility and could not be easily replicated in the field, but the δ13C values could be used in a similar way in a field experiment, to assess drought stress and relate it to deep rooting. The DeepRoot40 measure of deep rooting could be used in other conditions, with minirhizotrons in a field experiment, or in smaller-scale controlled experiments. However, adjustments to the 40 cm threshold may be needed, to reflect different soil and climate conditions which affect where the deep roots become ‘valuable’.

Yields, water stress and isotopes

Yields were particularly low in 2020 due possibly to lack of seed treatment, which likely limited overall plant growth, and therefore yields and root growth were both lower than in 2019. However, the higher yield under the deep soil treatment, and under the high N treatment, shows that the facility succeeded in creating an effective drought in both years. There was more drought stress as shown by δ13C in 2020, and very little spring rainfall, but soil water content (VWC) at anthesis was lower similar in the two years at 0.5 m depth, and in deeper soil layers was higher in 2020, probably as there was less water uptake from the smaller root systems.

The δ13C values in straw and grain show that there was more water stress at both vegetative and grain filling stages in 2020. We can see that there was a water stress during grain filling in the shallow soil treatments, as the TKW was lower in both years. Genotypes CH73641 and Cougar seemed most resilient, with a small difference in yield between drought treatments, particularly in 2020, but were relatively low yielding; Ormil had higher yields and a bigger effect of N but the drought effect on yield was limited. Kvarn was among the highest yielding genotypes but had the greatest yield reduction under drought stress.

Soil profile depth

The water treatment in this experiment was based on the soil depth gradients of the facility, where the deep soil part allows access to water from deeper soil layers not available to plants in the shallow area. The whole experiment was subjected to a drought treatment post-anthesis when the rainout shelters were moved over the facility, to allow significant water stress to develop in the shallow part of the facility. For interpretation of the interactions between root growth, water and N treatments, it should be kept in mind that differences in water availability were found in the subsoil, while differences in N availability were created by N fertilization in the topsoil.

Effect of N fertilization level on deep roots

Previous studies have shown that nitrogen fertilization affects root growth in the field, although the effects vary and are inconclusive. While N has been shown to reduce root growth in water-stressed wheat before (although this measured roots only to 1 m depth) (Wang et al. 2014), the opposite effect has also been shown where high N increased root growth at all depths (Barraclough et al. 1991). Moreover, higher N increased deep root growth and the proportion of roots in the deep soil (80-140 cm) (Wang et al. 2018). In the latter study, there was a relatively high water availability which may have influenced the effect of N. Rasmussen et al. (2015) found contrasting effects of N between years; in one genotype root growth increased across the soil profile when increasing N fertilisation up to 150 kg N/ha, but the highest N levels also reduced root growth in one year; this indicates an optimum N level for root growth, which would vary with water availability amongst other factors. This is also demonstrated in another Danish field trial, where more root growth was found in winter wheat in the soil below 1 m in years where high precipitation led to more N leaching to the deeper soil layers (Thorup-Kristensen et al. 2009). The effects of N on root growth remain uncertain, but are clearly affected by water availability and the placement of the N in the soil. In this study we focussed on deep roots, and found that high N reduced the root depth (DeepRoot40) in one year.

Genotype differences in deep roots

Previous studies have found genotypic differences in deep rooting under field or near-field conditions: Wacker et al. (2022) found that 15N uptake and deep rooting varied among genotypes and Rasmussen et al. (2015) found one cultivar, Hereford, had deeper roots and more deep N uptake (1-2.3 m) than other cultivars. The genotypes used in the present study were quite diverse, with release dates from 1984 to 2017, and from across Europe and north America. While the genotypes studied by Wacker et al. (2022) were less diverse as they were all modern Danish genotypes, they still showed significant differences in deep rooting. Significant genotype x environment interactions have been found in wheat roots (Botwright Acuña and Wade 2012) therefore more studies in different soils and climates are needed, as the existing studies on roots are still very limited compared to above-ground plants.

In addition to root measurements, we also used isotopic tracers, which give more direct results on the affinity for uptake of deep water and N than the root data. The isotope method was improved in 2020 by splitting the application between two irrigation tubes rather than one, allowing a larger sampling area, and this can be seen in the improved correlation between isotopes and roots. Uptake of 2H, although not significantly affected by genotype, was positively correlated with both δ15N and deep rooting suggesting that these three measures of deep roots are reliable indicators of deep root activity. Genotypes with deeper roots took up significantly more 15N, in accordance with Wacker et al. (2022). They found that deep root traits had a predictive effect on uptake of deep-placed N tracer, and that by correcting for genotype this effect increased; this ‘mediation’ shows that some of the genotype differences in tracer uptake are due to genotypic differences in deep rooting.

Effects of N on yields and drought tolerance

The yields of modern Danish varieties Kvarn and Creator were more responsive to N, perhaps as they were bred under higher N fertilisation levels. There were some genotypes which remained relatively stable across treatments in both years; however the small sample area for yields and small number of repetitions means that interactions between genotype, soil profile depth and nitrogen are difficult to show.

Although the effects of N on δ13C were small, high N increased δ2H showing more uptake of deep water and higher water use by bigger plants, making the high N treatment more reliant on deep water during grain filling. Rooting depths were not very different between the N treatments, but increased water use from deep soil layers under high N has been shown in maize (Xie et al. 2022). Higher N can increase the plant biomass, generating more transpiration and water loss from the canopy, and therefore increasing the water demand (Wang et al. 2018).

Deep roots and drought tolerance

The negative correlation of 2H and δ13C shows that drought tolerance was related directly to deep water uptake. This indicates that deep water uptake, rather than any other aspect of plant physiology, reduces drought stress, as measured by δ13C. This finding agrees with Condon et al. (1992) who showed that genotypes that depleted less soil water in the field had lower carbon isotope discrimination (CID) values (i.e. higher δ13C) meaning more water stress. Similarly, a field trial of spring wheat in the UK found that genotypes with deep water uptake (around 1 m) had more drought tolerance (Ober et al. 2014). This finding is supported in our study by the grain yield difference between the wet and dry areas of the bed; the positive correlation in 2020 shows that water from deep soil layers was more valuable to genotypes with deeper roots (Fig. 11a), while for shallower-rooted genotypes, the deep soil is less important. The soil depth effect was similar for the δ13C difference, where genotypes with deep rooting showed the strongest response to soil profile depth on δ13C.

The correlation between uptake of tracers applied at anthesis and δ13C in the grain, indicates that this late root activity enables continued uptake of water during grain filling. That the relationship between δ13C and deep roots was stronger in the grain than the straw also suggests that deep roots played more of a role during grain filling than during vegetative development and straw growth. In 2019, the drought stress was more limited than in 2020, and the deep roots did not affect δ13C. This shows that, while we may screen for deep rooting and, in the RadiMax facility, for deep tracer uptake, under many conditions, the direct value of deep roots for drought tolerance can only be shown when a water stress is present (see Wasson et al. 2012). However the water availability at anthesis was similar in the two years, so the higher δ13C levels may reflect the effect of lack of seed treatment on root growth, rather than drought stress due to soil water limitation.

CID has been used as a breeding tool for drought tolerance in Australian wheat, and could be used to select efficiently for drought resistance due to its high heritability and correlation between CID and grain yields of wheat under drought (Kunz et al. 2022; Richards et al. 2002). However, to make improvements in drought tolerance through breeding, it is necessary to understand the mechanisms that contribute to water stress tolerance. While Condon et al. (1990) attributed some of the genotypic variation in δ13C to stomatal conductance and photosynthetic capacity, we show that deep rooting can contribute some of the mechanisms for drought tolerance, as measured by δ13C. This has been shown for common bean where root length density was positively correlated with CID (White et al. 1990), but until now it had not been shown how this related to roots in wheat.

DeepRoot40 as a measure of deep rooting

The correlation of DeepRoot40 with yield, deep placed isotopic tracers and δ13C confirm that it is a good measure of deep roots, and that deep roots are important for deep N and water uptake. Recently there has been increased focus on deep rooting, density of roots and branching, which allow for ‘capture’ of water and nutrients (Wasson et al. 2012). Among the many root traits which can be measured, we need to find out which root traits are important for the uptake of essential nutrients and water, i.e. to develop root ideotypes for breeding (Foulkes et al. 2009; Lynch 2013; Palta et al. 2011; Thorup-Kristensen and Kirkegaard 2016). In particular, deep roots which can take up leached N and stored water from deeper soil layers, which are not available to less deep rooted genotypes, are found to be important.

While estimates of rooting depth using imaging or soil cores have been used to evaluate deep roots in the field (Botwright Acuña and Wade 2012), this measurement relies heavily on one single point from a small number of images to indicate the deepest root, which can lead to inaccuracy. Measures of root weight density (RWD) at different soil depths, as in Brunel-Saldias et al. (2020) in tube-grown wheat, have indicated that root systems with thinner roots may be more efficient for water uptake. However in order to compare large numbers of genotypes, traits are needed which allow us to evaluate the most critical aspects of the root system; in the case of deep water and N uptake, we need to evaluate deep roots. Wacker et al. (2022) attempted to develop various traits which could quantify deep root density. They considered DeepRoot40 an ‘improved’ measure of rooting depth, as it includes a larger image area in the analysis, increasing reliability. Other studies investigate root distribution at different depths, using root length density (RLD) (Bai et al. 2019; Postic et al. 2019; Rich et al. 2016; Zhang et al. 2020). However, such RLD profiles do not give a single critical value, and make it difficult to identify genotypes with ‘better’ root traits for deep resource uptake. Zhang et al. (2020) demonstrate that beyond a critical density of 1 cm root per cm3 of soil there is little increase in water uptake. In particular in the deep soil where there were less roots, they found that root length density was related to water uptake, while in shallow soil RLD did not relate to water uptake. DeepRoot40 accounts for density and rooting length, and focusses only on the deepest 40 cm of roots; we show that this relates to deep water and N uptake.

Why phenotype roots rather than yield?

We have established a positive correlation between grain yield and deep roots; this alone could suggest that deep roots are beneficial to yield, under water stress. While it might be expected therefore that selecting for yield would indirectly result in selecting for deep roots, in northern Europe water limitation is not always an issue, which means that the value of deep rooting is inconsistent across seasons and sites. This tends to mask the relationship between deep rooting and yield and makes breeding for yield inefficient in breeding for deep roots.

Other factors also in the shoot part of the plant will affect the crop yield response to water deficiency, explained by Wasson et al. (2012), but we can target improved function of roots specifically, by phenotyping them. We have demonstrated that deeper roots corresponded to lower δ13C values and that δ13C and uptake of δ2H tracer from the subsoil were related, and thereby shown that deep roots allow more water uptake and reduce water stress. This demonstrates that deep rooting itself is a beneficial trait for drought tolerance, and that plant breeders should include root phenotyping in their process. Deep rooting is also likely to be important for N use every year, but the yield effects will be marginal, so the main value of this deep N uptake will often be environmental effects more than crop yield.

In future years the expected increasing variability in precipitation means that deep root traits may become more important in selecting for yield stability, and breeders will increasingly need to consider yield stability and resilience as well as yield.

Conclusion

We have found that genotypes do vary in their deep rooting, and that deep roots are a necessary mechanism for deep water and N uptake, contributing to resilience and improved resource use. As climate change progresses, variations in deep roots that increase resilience may become more important. This highlights the importance of including deep roots in future breeding programmes as a desirable ‘trait’ to target in future-proofing crops.

Root research has been neglected in the past, mainly due to the lack of efficient methods for identifying and quantifying deep roots, but we have found that using minirhizotrons and isotopic tracers are effective for evaluating deep root function; we hope that this can contribute to further deep root research.

Further studies are needed to see how these results compare in the field with the more closely related genotypes available to farmers, under different climate and soil conditions, and to continue to develop improved methods of quantifying root systems that clearly relate to root function.