Maize brace roots provide stalk anchorage

Abstract Mechanical failure, known as lodging, negatively impacts yield and grain quality in crops. Limiting crop loss from lodging requires an understanding of the plant traits that contribute to lodging‐resistance. In maize, specialized aerial brace roots are reported to reduce root lodging. However, their direct contribution to plant biomechanics has not been measured. In this manuscript, we use a non‐destructive field‐based mechanical test on plants before and after the removal of brace roots. This precisely determines the contribution of brace roots to establish a rigid base (i.e. stalk anchorage) that limits plant deflection in maize. These measurements demonstrate that the more brace root whorls that contact the soil, the greater their overall contribution to anchorage, but that the contributions of each whorl to anchorage were not equal. Previous studies demonstrated that the number of nodes that produce brace roots is correlated with flowering time in maize. To determine if flowering time selection alters the brace root contribution to anchorage, a subset of the Hallauer's Tusón tropical population was analyzed. Despite significant variation in flowering time and anchorage, selection neither altered the number of brace root whorls in the soil nor the overall contribution of brace roots to anchorage. These results demonstrate that brace roots provide a rigid base in maize and that the contribution of brace roots to anchorage was not linearly related to flowering time.


| Plant biomechanics and lodging-resistance
The failure of crop plants to maintain a vertical position is referred to as lodging and can affect both crop yield and grain quality (Berry et al., 2004;Fedenko et al., 2015;Rajkumara, 2008). Plant lodging occurs due to weak stalk and/or failed root anchorage, and multiple factors contribute to lodging susceptibility. Susceptibility factors include field management practices, meteorological factors, biotic stresses, and plant characteristics (Berry et al., 2004;Rajkumara, 2008;Stamp & Kiel, 1992). Natural lodging is difficult to study due to the multiple and unpredictable factors that influence the susceptibility. To overcome this challenge, proxy measures have been developed to determine the plant characteristics that optimize for lodging-resistance (Erndwein et al., 2020). These proxy measures rely on principles of biomechanics to link plant mechanical properties with lodging-susceptibility.
In maize, lodging has been shown to reduce yield by 3%-25% (Carter & Hudelson, 1988;Flint-Garcia et al., 2003;Tirado et al., 2020). For stalk lodging, biomechanical measures of stalk breaking strength and stalk flexural stiffness have shown high association with lodging-resistance (Robertson et al., 2016;Sekhon et al., 2020). These biomechanical properties have been further associated with variation in plant characteristics (e.g. stem morphology), and the underlying stem material properties (e.g. bending strength) (Robertson et al., 2016;Stubbs et al., 2020). In contrast, there is a limited understanding of how root systems contribute to lodging-resistance.

| The role of brace roots in lodging-resistance
Maize has specialized brace roots that develop in a whorl from above-ground stem nodes (Blizard & Sparks, 2020). The uppermost whorls of brace roots may remain aerial with the lower whorls penetrating into the soil. As their name suggests, the brace roots that penetrate the soil have been proposed to provide anchorage and limit root lodging Sharma & Carena, 2016;Shi et al., 2019). Previous studies linking brace roots to root lodging-resistance have identified the brace root traits that are correlated with higher lodging-resistance as (a) a higher number of roots in a whorl Sharma & Carena, 2016), (b) more overall brace root whorls entering the soil (Sharma & Carena, 2016;Shi et al., 2019), and (c) a higher brace root spread width (Sharma & Carena, 2016).
While these studies support the importance of brace roots for lodging-resistance, a direct assessment of the contribution of brace roots to plant biomechanics has not been reported.

| Genetic regulation of brace roots
Although optimal brace root traits for root lodging-resistance have been identified, there are limited reports to define the genetic basis of these traits. The studies that have been reported primarily focus on the total number of brace root whorls (Gu et al., 2017;Ku et al., 2012;Zhang, et al., 2018a;Zhang, et al., 2018b), with only one study differentiating between the number of brace root whorls that enter the soil and the total number of brace root whorls (Ku et al., 2012). In this study, an analysis of recombinant inbred line (RIL) and immortalized F2 (IF2) populations identified both shared and independent quantitative trait loci (QTL) for brace root whorls in the soil and the total number of brace root whorls (Ku et al., 2012).
One of the shared QTL on chromosome 10 explained 16.36% (RIL) and 17.88% (IF2) variation of total brace root whorls and 3.50% (RIL) and 6.37% (IF2) of the variation in the number of brace root whorls in the soil (Ku et al., 2012). While the gene or genes underlying this major QTL have not been determined, the location overlaps with CCT1 (CONSTANS, CONSTANS-LIKE, TOC1}; B73 v4: Zm00001d024909), a major determinant of photoperiod responses in maize (Hung et al., 2012). A separate study of a maize-teosinte BC 2 S 3 population showed significant overlap for brace root whorl number QTL and flowering time QTL (Zhang, et al., 2018b), suggesting a link between the control of flowering time and brace root whorl numbers. It is unknown if modulating flowering time is sufficient to alter the number of brace root whorls in the soil, and thus affect lodging-resistance.

| Existing methods to quantify lodgingresistance
Prior studies of root lodging-resistance have employed destructive measures of pulling/pushing forces or root failure moment (Erndwein et al., 2020). However, it is more advantageous to use non-destructive testing strategies that enable the direct measurement of brace root contribution. The Device for Assessing Resistance to Lodging IN Grains (DARLING) is ideal for non-destructive field-based mechanical testing (Cook et al., 2019). DARLING is a portable device equipped with a variable height load cell (force sensor) on a vertical post and a hinged footplate equipped with an inertial measurement unit (IMU). This device acquires continuous force-rotation data, which can be used to assess plant biomechanics. Previously, this device has been used to measure stalk flexural stiffness, which has been associated with stalk bending strength and stalk lodging-resistance (Robertson et al., 2016;Sekhon et al., 2020). Measurements of stalk properties with DARLING assume that the plant acts as a cantilever beam with a rigid boundary condition at the base (soil).
The rigidity at the base is provided by the anchorage of the root system, with both below-ground and above-ground (brace root) contributions. However, the relative contribution of these different root types has not been assessed.
In this study, we show that brace roots significantly contribute to the anchorage of maize and that more brace root whorls in the soil lead to a greater contribution. However, flowering time selection is not sufficient to alter the number of brace root whorls in the soil nor the relative contribution to anchorage. Together these results establish the methods for a direct assessment of the contribution of brace roots to anchorage, and highlight limitations for future studies aimed at modulating the brace root contribution to anchorage.

| Plant stiffness is reduced throughout the day
The biomechanics of a plant before senescence is hypothesized to vary during the course of a day due to the variable water content (high in the morning and decreasing throughout the day), which alters turgor pressure and is thus also hypothesized to alter plant biomechanics. To quantify how biomechanics is impacted by plant physiology and time of day, two replicates of CML258 inbred plants were measured with DARLING at the VT stage (103 days after planting) at 9:00 a.m., 12:00 p.m., and 4:00 p.m. (Figure 1).
As expected, the Force-Deflection slope is highest, indicating a stiffer plant, at 9:00 a.m. and decreases throughout the day ( Figure 1; p = 2.36 -06 ). Thus, any biomechanical testing applied to green plants must be time-matched to circumvent these time of day effects.

| Neither the number of tests nor plot replicate affects plant stiffness
Non-destructive measures of plant biomechanics are acquired under the assumption that the plant is elastically deformed and does not experience permanent deformation. To test this assumption, the same plants from replicate A and B were tested three times in a series before returning to the beginning. For each plant, there was variation in the Force-Deflection slope for the three tests, but this variation was unrelated to the test order ( Figure 1; p = .716). Further, there was no replication effect between the two plots assayed (p = .972), showing that the time of day was the dominant factor for influencing plant biomechanics. The variation observed between repeat tests is likely due to the re-positioning of the DARLING footplate relative to the plants that occur with each subsequent test and/or the noise of data collection that impacts the extraction of the slope from the Force-Deflection data. This variation represents the technical variation that is introduced by manual field-based measurements. Despite this variation, a high repeatability is found for the Force-Deflection slope measurements (Repeatability, R = .876, SE = 0.0245, p = 2.31E −49 ; Figure S2). These results indicate that neither repeat testing nor plot replication affects the Force-Deflection slope in the CML258 inbred line, and there is high repeatability with the non-destructive measurement of plant biomechanics with the DARLING.

| Plant stiffness is higher during senescence than the tasseling stage
To remove the impact of time of day on the Force-Deflection slope, all subsequent analyses were conducted during plant senescence and just prior to or immediately after harvest. The same CML258 plants that were tested at VT were tested again at R5 prior to harvest. There was a consistent increase in the Force-Deflection slope at R5 (Figure 1) compared to plants measured at VT. There was again no effect of plot replication on the Force-Deflection slope at R5.
These data demonstrate that plant biomechanics change across the plant lifespan, with increasing stiffness during reproductive development relative to late vegetative development.

F I G U R E 1 Plant Stiffness Changes with Time of Day and Developmental
Stage. CML258 plants were tested at tasseling VT (103 days) and dent R5 (128 days) stages from two replicates (Plot A and B). At VT, within replicate A, each plant was tested three times as indicated by the colored dots (blue-first, red-second, and black-third tests). The Force-Deformation slope is reduced throughout the day (p = 2.36E −06 ). The effect of replication and repeat testing was not significant (p = .716 and .972, respectively). The same plants were tested at R5 and show higher Force-Deflection slope compared to any time point at VT. Plant 11 and 12 in replicate A at R5 were not used due to weed control damage 2.4 | Brace roots significantly contribute to plant anchorage Brace roots are suggested to contribute to plant anchorage and root lodging-resistance, but their contribution has not been directly tested. CML258 plants have two to three whorls of brace roots that enter the soil (Figure 3a). Although there was no effect of plot replication on the Force-Deflection slope in Figure  measures. There was a moderate correlation between the difference measure and the initial Force-Deflection slope ( Figure S2a, r = .59, p = .004). This is not surprising, as the higher the initial Force-Deflection slope the more potential there is to reduce the slope upon brace root removal. However, when normalizing for the relative contribution by use of a ratio, the correlation disappears ( Figure S2b, r = −.17, p = .461). The initial Force-Deflection slope is predominantly determined by the stalk properties, and thus we would not expect any correlation with the relative brace root contribution to anchorage. To separate from the initial stalk mechanics, we utilize the relative ratio measurements for the remainder of this F I G U R E 2 Methods for Testing the Contribution of Brace Roots to Anchorage. The measure of plant biomechanics with all brace roots intact is labeled "A." For brace root removal experiments, the excision of sequential whorls starting at the top are labeled alphabetically.
Here is an example of a plant with three whorls of brace roots, with two that enter the soil. The uppermost whorl of brace roots that enter the soil is excised and the measurement is labeled "B." The next whorl of brace roots that enter the soil is then excised and the measurement is labeled "C" study. These ratios are a quantification of the relative contribution of brace roots to anchorage, and a lower value indicates a higher contribution from the brace roots.

| Brace root whorls have differential contribution to anchorage
Despite a consistent reduction in Force-Deformation slope upon removal of brace root whorls, there is variation in the relative contribution within the CML258 inbred line (Figure 3d). Previous studies have shown that the number of brace root whorls that enter the soil is correlated with root lodging-resistance Sharma & Carena, 2016). Thus, we examined the role of different numbers of brace root whorls in the soil.
There is a significant correlation between the brace root contribution ratio and the number of brace root whorls in the soil

| Selection for early flowering does not affect the number of brace root whorls in the soil or the brace root contribution to anchorage
Previous work has described a relationship between the number of nodes producing brace roots and flowering time in maize (Zhang, et al., 2018b). However, the described relationship does not take into account the number of brace root whorls that enter the soil, nor their relative contribution to anchorage. We set out to determine if selection for flowering time can alter the number of brace root whorls that enter the soil, and thus alter the contribution of brace roots to anchorage. For this analysis, a subset of lines from the Tusón population was analyzed (Teixeira et al., 2015). The opposing effects of these variables are remarkable, considering that the DTS and PH are positively correlated (r = .8; Table S2) with each other across the selection experiment (Teixeira et al., 2015). Similar regressions were performed using plant per se data, instead of line means. The plant per se data uncovered the same relationships of DTS and PH to the Force-Deflection slope.
Closer inspection of the changes in the Force-Deflection slope across the generations of selection shows that significant changes did not occur until the eighth generation (g8; Figure 5c), where there was a significant reduction from g6 (Tukey HSD p = .009) followed by an increase in the Force-Deflection slope at g10 (Tukey HSD p = .015). At these same generations, a change in flowering time was not significant (Tukey HSD p = .83; Figure S4). Mean ear height also increased between g8 (119 cm) and g10 (126 cm), and mean PH decreased marginally between g8 (248 cm) and g10 (243 cm) for the lines used in this analysis ( Figure S4). Although these differences were not statistically significant, the greater selection efficiency for lodging resistance and co-selection of PH in later generations of the Tusón population may explain the bounce observed in the Force-Deflection slope in g10 (Figure 5c). Indeed, greater selection efficiency for standability in later generations may select for specific plant architectures and explain the increase in the Force-Deflection slope in g10. There are three hypotheses that may explain these differences. However, the displacement of the stalk during testing is no more or less than we would expect from wind, people, or animals moving through the field. Thus, we do not expect the testing itself to induce additional changes in the biomechanics. While additional work is required to fully define the underlying biology of these dynamic changes in the Force-Deflection slope, these results highlight the importance of field-based non-destructive tests to capture biomechanical dynamics.

| Brace roots provide anchorage
The measured Force-Deflection slope has a significant contribution from brace roots, which provide a rigid attachment base for the F I G U R E 3 Brace Roots Contribute to Anchorage. (A) The number of brace root whorls in the soil was evaluated for CML258 plants in two replicates and there was a significant effect of replicate (p = .0308). (B)The Force-Deformation slope was measured from CML258 plants upon sequential removal of brace root whorls. Test was performed with all whorls intact A, excision of the highest whorl B, the next highest whorl C, and so on. Plants with two whorls are represented by black and with three whorls by blue dots. There is an overall reduction in the slope upon removal of brace root whorls (p = 5.88E −11 ). (C) This observation is extended to the individual plant, where the subsequent removal of brace root whorls results in a reduction of the Force-Deflection slope. (D) Using the paired data measurements, the contribution of brace roots to anchorage is represented in two ways: 1) a ratio of the Force-Deflection slope with no brace roots to the Force-Deflection slope with all brace roots and 2) the difference in the Force-Deflection slope with all brace roots and the Force-Deflection slope with no brace roots. FDwo = Force-Deflection slope without brace roots. FDw = Force-Deflection slope with brace roots maize stalk (Figure 3). This contribution is likely due to brace roots acting as guy-wires to provide stability to the free-standing stem.
In this way, the removal of brace roots will effectively lengthen the beam-in other words, move the anchorage point on the stem from the point of brace root attachment to the soil surface. One hypothesis is that the role of brace roots can be explained by this change in effective beam length. We manually measured the attachment height of brace roots for the CML258 plants in Plot A, which ranges between 0.015 m and 0.032 m. When accounting for this change in length, and calculating the ratio contribution of brace roots, the effect of brace roots increases rather than decreases ( Figure S5). Therefore, the contribution of brace roots cannot be explained by changing the length of the beam and instead demonstrates that brace roots contribute to stalk anchorage by providing a stable base.

| More brace root whorls have a greater contribution to anchorage
Our results demonstrate that the more brace root whorls that enter the soil, the greater the overall reduction in the Force-Deflection slope ( Figure 4). While this is again consistent with the idea of brace roots functioning as guy-wires, there are some notable differences with this analogy. With guy-wires, the higher set of wires contributes more than the lower set. With brace roots, the lowest whorl contributes the most and the higher whorls successively less. We hypothesize that this is due to the depth of the roots in the soil; with the lowest whorl having more time to grow subterranean and provide anchorage. However, determining the depth of brace roots in the soil is a difficult challenge due to the complexity of the maize root system and limited field-based root phenotyping approaches (Clark et al., 2020).

F I G U R E 4 Brace Root Whorls
Have Differential Contribution to Anchorage. (a) The reduction of Force-Deflection slope after removal of all brace root whorls is significantly correlated with the number of whorls in the soil (r = −.50, p = .02). (b) Each whorl of brace roots in the soil has a differential contribution to the Force-Deflection slope. The effect of whorl is significant by ANOVA, p = 2.4431E −06 . The brace root whorl closest to the soil contributes the most and the additional whorls above contribute successively less. (c) The contribution of the lowest whorl is highly correlated with the overall contribution (r = .85, p = 1.302E −06 ). Plants with two whorls are represented by black and with three whorls by blue dots. Lines represent a generalized linear model (glm) fit and shading indicates a 95% confidence interval. A lower ratio indicates a higher contribution of brace root whorls. WR = whorl

| Selection for flowering time does not affect the relative contribution of brace roots to anchorage
Previous work has linked the number of brace root whorls with flowering time (Ku et al., 2012;Zhang, et al., 2018b). However, it was unknown if modulating flowering time affects the number of brace root whorls in the soil or their contribution to anchorage. Although there was a significant change in flowering time across selection in these populations ( Figure S4), there was no change in the number of brace root whorls in the soil nor their relative contribution to anchorage (Figure 5a,b).
The lack of a relationship between flowering time and brace root traits may result from the source of flowering time variation in this particular subset of the Tusón population. In the complete Tusón population, twelve loci were identified at an FDR of 10% that that contributed to flowering time adaptation across the 10 cycles of selection (Wisser et al., 2019). Of these, the non-photoperiod sensitive allele of a major determinant of photoperiod responses in maize, cct1 was highly relevant for the flowering time selection (Wisser et al., 2019). Due to the selection of "average" lines for our analysis, four of these loci (including cct1) were monomorphic in all generations including g0 and g2 and another three loci had a single line with an alternative allele (Table S3). Of the remaining five markers, only two have more than five lines with alternative alleles and none of these alleles were significantly associated with time to maturity, or the Force-Deflection slope as determined by regression analyses (Marker data are in Table S3; Slopes are in Table S4). Despite the absence of variation in these major photoperiod genes, the subset of lines evaluated in this study still shows a significant decrease in days to anthesis and silking in Delaware ( Figure S4a

| There is a separable and opposing effect of plant height and flowering time on plant stiffness
Interpretations of the relationship between the initial Force-Deflection slope (stiffness) and changes in flowering time over the selection cycles in the Tusón population is complicated by a few factors. The first is the small sample size of S1 lines evaluated in this study. For instance, the lines sampled as representatives for selection cycle g10 have higher mean ear height than g8. This is not the case for the Tusón population which flowers earlier and shows a gradual and significant decrease (Tukey's HSD p < .05) in ear height from g8 to g10 (Teixeira et al., 2015). The fortuitous selection of lines with unusual height in this subset may confuse the predicted slope for calculations of genetic effects of selection across the experiment, but may have helped to expose opposing effects of PH and DTS regulators on plant stiffness.
Additionally, poor sampling of alternate alleles at the significant flowering time loci in this population may limit the detection of associations between DTS and the Force-Deflection slope. Flowering time in maize is a complex trait (Buckler et al., 1993)  Overall, these results are the first report of repeat measures to uncover the dynamics of plant biomechanics in the field. These results also provide the first direct measurement of the brace root contribution to anchorage. The number of brace root whorls in the soil is important for the overall contribution to anchorage. However, selection for flowering time alone is not sufficient to alter the number of brace roots in the soil or their contribution to anchorage. These results provide the foundation for future studies aimed at uncovering the variation in brace root contribution to anchorage, defining the genetic basis of this trait, and uncovering the brace root phenotypes (in addition to whorl number) that alter the contribution to anchorage. and the contribution of brace roots was determined as described below. Two plants from the first replicate were unable to be tested at R5 due to damage incurred during weed control measures.

| Plant material
For investigating the link between flowering time and the brace root contribution to anchorage, the Hallauer's Tusón (Teixeira et al., 2015) population was analyzed.
The Tusón base population (g0) was produced by one generation of intermating between five maize accessions: PI449556, PI583912, NSL283507, PI487940, and PI498583. This base population was then subjected to open pollination in isolation and selection was made for early anthesis/silking and standability (reduced root and/or stalk lodging) for 10 generations (Teixeira et al., 2015).
Based on results from (Teixeira et al., 2015), a subpopulation from even-numbered generations was selected based on flowering times in Delaware closest to the mean for each generation. These lines were planted in a randomized block design in Newark, DE in the summer of 2018. Data were collected from two to three plants derived from three to five accessions in each even-numbered generation, with up to two replicated plots (Table S1). To overcome the moisture variability in green plants, samples were tested for the brace root contribution to anchorage after dry-down (154-158 DAP).

| Data collection
Field-based measurements of plant biomechanics were collected with DARLING devices (Cook et al., 2019). Here the devices were used to non-destructively flex the plant and extract the slope of the For testing plants during senescence or dry-down, the test with all brace roots intact was labeled "A" (Figure 2). The top whorl of brace roots that entered the soil was then manually excised with shears, and the test was repeated and labeled "B." The process was repeated for all subsequent whorls of brace roots that entered the soil and labeled alphabetically. Here, brace roots were defined as any nodal roots visible above the soil surface. Cook et al. (2019) explain how DARLING can be used to measure the flexural stiffness of maize stalks by bending while measuring both the force and deflection. The calculation of flexural stiffness assumes that the plant acts as a cantilever beam with a rigid boundary condition at the base (soil). Under these conditions, the flexural stiffness is purely a property of the stalk itself. Here we assume that the stalk properties are unchanging and thus any change in Force-Deflection slope from brace root removal is due changing the boundary condition of anchorage. So as not to imply that these changes are due to altering the stalk itself, here we report and analyze the Force-Deflection slope.

| Force-Deflection calculations
To convert the measured rotation to deflection, IMU data were first converted from degrees to radians with the following equation: where pi equals 3.1415.
The rotation in radians was then converted into deflection ( ) with the following equation: where h is the height of the applied load (m) and is the rotational angle (radians).
Three different approaches were compared to extract the slope from the Force-Deflection data measured from the DARLING devices ( Figure S1a). In the first approach, a line was fit to the complete dataset, which includes loading and unloading data for all three cycles. For the second and third approaches, the data were automatically parsed into loading and unloading data points based on a user-defined number of monotonically increasing data. Here, ten consecutively increasing deflection data points were required to define a single load cycle, and each load must span at least 0.02 m to be considered a cycle. For the second approach, a line was fit to all three loading cycles together. In the third approach, a random sample consensus (RANSAC) was used to fit a line to each of the cycles and the slope of the line that fits the curve with the longest continuous data points was kept. Comparison of the different methods to extract slopes revealed that all three slopes have Pearson correlation r ≥ .76 and p < 2.2 E −16 (Figure S1b-d). While it is ideal to extract the slope from only the loading data, the variation in the manual acquisition of data in the field environment often results in the unreliable automated identification of the loading portion of the curve (e.g. wind will introduce noise that prevents the identification of monotonically increasing data points). To avoid introducing potential bias by the manual definition of loading data, the slopes of lines that were fit to the complete dataset (loading and unloading data) were used in this manuscript. Fitting a line to the complete dataset has a high correlation with fitting data to just the loading data (r = .86, p < 2.2 E −16 ), thus this selection is unlikely to influence the results.

| Brace root contribution calculations
Since the removal of brace roots does not change the properties of the stalk, any differences in the Force-Deflection slope can be = deg × pi∕180.
attributed to a change in the boundary condition and interpreted as a change in anchorage at the base. The overall brace root contribution to anchorage was calculated through relative and absolute approaches. In the relative approach, the ratio was calculated as Ratio = FDwithout/FDwith, where FD is the Force-Deflection slope. For plants with three brace root whorls in the soil the calculation is Ratio = D/A ( Figure 2) and for plants with two whorls in the soil, Ratio = C/A. In the absolute approach, the difference was

| Genotypic analysis of selection
Marker genotypes for the subset of the Tusón population used in this study were obtained from (Wisser et al., 2019). To test the ef-

| Statistics
To determine the sample sizes necessary to identify differences in the Force-Deflection slope, G*Power was used to compute sample size

ACK N OWLED G M ENTS
We would like to acknowledge members of the Sparks lab for reading and providing feedback on this work, and Dr. Randall Wisser for providing Tusón seed stocks.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest associated with the work described in this manuscript.

AUTH O R CO NTR I B UTI O N S
JWR, LE, and EES collected data. RSK, AS, BPD, and EES analyzed data. TW provided germplasm and managed field sites. DDS provided DARLING devices. All authors contributed to the writing and editing of the manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
All raw data, the code used to process data, and the analyzed data are available at: https://github.com/EESpa rksL/ab/Reneau_et_al_2020.