Development of a model estimating root length density from root impacts on a soil profile in pearl millet (Pennisetum glaucum (L.) R. Br). Application to measure root system response to water stress in field conditions

Pearl millet is able to withstand dry and hot conditions and plays an important role for food security in arid and semi-arid areas of Africa and India. However, low soil fertility and drought constrain pearl millet yield. One target to address these constraints through agricultural practices or breeding is root system architecture. In this study, in order to easily phenotype the root system in field conditions, we developed a model to predict root length density (RLD) of pearl millet plants from root intersection densities (RID) counted on a trench profile in field conditions. We identified root orientation as an important parameter to improve the relationship between RID and RLD. Root orientation was notably found to depend on soil depth and to differ between thick roots (more anisotropic with depth) and fine roots (isotropic at all depths). We used our model to study pearl millet root system response to drought and showed that pearl millet reorients its root growth toward deeper soil layers that retain more water in these conditions. Overall, this model opens ways for the characterization of the impact of environmental factors and management practices on pearl millet root system development.


Introduction
Pearl millet (Pennisetum glaucum (L.) R. Br., syn. Cenchrus americanus (L.) Morrone) is a 44 cereal crop domesticated in the Western part of Sahel about 5,000 years ago [1]. It is well 45 adapted to dry tropical climate and low-fertility soils and therefore plays an important role for 46 food security in arid and semi-arid regions of sub-Saharan Africa and India. In these areas, 47 pearl millet is one of the most important sources of nutritious food [2, 3] and is the staple crop 48 for nearly 100 million people [4,1]. Its grain is rich in protein, essential micronutrients and 49 calories. It is also gluten-free and has hypoallergenic properties [4]. In a context of climate 50 change leading to unpredictable weather patterns and rising temperatures in West Africa [5,  However, this will depend on the availability of phenotyping methods to characterize and 57 exploit the available genetic diversity and identify interesting target traits. 58 Drought and low soil fertility are among the most important factors limiting pearl millet 59 yield. The root system is responsible for water and nutrient uptake, and root system 60 architecture is therefore a potential target in pearl millet breeding program to address these 61 constraints. It is also an important trait to consider when analyzing the impact of agricultural 62 practices. However, despite tremendous progress in the genetic characterization of root 63 development, root system architecture phenotyping remains challenging particularly in 64 agronomically-relevant field conditions. The root length density (total length of roots per unit 65 of soil volume; RLD) is a key factor to estimate the soil volume explored by a root system 66 and consequently the amount of water and nutrients available to the plant [7][8][9][10][11][12]. Therefore, 67 RLD could be used to screen drought-resistant varieties. 68 The aim of this study was to develop a technique to map RLD in pearl millet plants from 69 simple measurements in field conditions. For this we analyzed the relationship between RLD      Roots sampling for model development 120 We adapted a method previously described to estimate the RLD from intersections between 121 roots and the face of a soil trench profile (root intersection density or RID; [10-12, 16 and 122 17]). Trench profiles were dug perpendicularly to the sowing rows and at two distances (30 123 then 10 cm) from the plant stalk base (Fig 1A). Three-sided incomplete steel cubes with 124 sharpened edges facilitating penetration into the soil were used to sample soil cubes (Fig   125   1BC). The sampling device was pressed into open soil profile (trench profile) until its rear 126 plane was aligned with the soil profile ( Fig 1D) and then cut out of the soil to obtain a cube of 127 soil ( Fig 1E). A second sample was taken at the same depth and distance from the plant but

136
Measurements were repeated four times per variety (384 cubes in total) and repeated 137 measurements were averaged (i.e., same variety, same seeding rate, same date and same 138 position).

139
The same protocol was used in the validation test, except that measurements were conditions. We first analyzed the diversity of these four pearl millet varieties for root and 191 shoot characters. There were significant differences in root length densities (RLD m ) and root 192 biomass densities (Fig 2A,B). Some varieties had deeper root systems than others. By contrast 193 no significant differences were observed between varieties for shoot traits such as biomass 194 ( Fig 2C). Hence, these four varieties had contrasted root systems and were deemed suitable 195 for model development. 196 We then analyzed the relationships between measured RLD and root intersection densities to the plant (S1 Table). The Pv index only depended on depth. As a consequence, the results

249
The results of our statistical tests on the different models are summarized in Table2.

255
Considering all roots (fine and thick), statistical tests showed that the measured and 256 calculated RLD values were significantly closer with the geometrical model than with the 257 empirical model ( Table 2). The MB induced by both models was an underestimation of RLDs 258 for low root intersection densities, generally at depth (Tables 2 & 3, Fig 5A,B). The results obtained using models estimating only the fine roots (diameter < 1mm), were 263 close to those obtained using model estimating all roots together. There were good 264 relationships between measured and calculated values of fine roots for each variety except GB 265 8735 (Fig 5C, Table2).

266
RLD for thick roots (diameter > 1 mm) ranged from 0 to 2000 m.m -3 , about ten times 267 lower ( Fig 5D) than those for fine roots (Fig 5C). The model construction showed that when  However, the latter was only usable for RLDs > 500 m.m -3 and its use is therefore limited.

291
Response of pearl millet root system to water stress 292 We next used our model that takes into account all roots to study the effect of water stress   Altogether, these results indicate efficient field dry-down and imposition of water limited 309 conditions from topsoil to a depth of around 90 cm in the drought stress treatment.

310
Agromorphological characteristics were then measured at the end of the cycle (99 DAP).

311
SL28 is a dual-purpose pearl millet variety selected for both fodder and grain production.

312
Accordingly, it shows a very large biomass and grain production compared to the inbred line 313 LCICMB1 in well-watered conditions (Fig 6A,B). Moreover, these two lines showed 314 contrasted responses to drought stress conditions. SL28 showed a very strong and significant 315 reduction in both biomass and grain production in response to water stress while these traits 316 were not significantly affected in LCICMB1 (Fig 6A,B). 317 We used the geometrical model for all roots to estimate RLD from RID along soil profiles.

318
Measurements were performed in both well-watered and drought stress conditions for both 319 lines at 43 and 71 DAP, i.e. at the beginning and at the end of the water stress period. Three 320 days after stress application (43 DAP), the RLD profiles were not significantly different for 321 well-watered and drought stress conditions for both lines (Fig 7A,B) indicating that the 322 change in water availability had not significantly impacted root architecture at this stage.

323
However, 31 days after stress application (71 DAP), we observed strong and significant 324 changes in RLD profiles between well-watered and drought stressed plants (Fig 7C,D). For drought as compared to well-watered conditions (Fig 8A,B). Hence, our data demonstrate that 331 upon drought conditions, both pearl millet lines reduced root growth in the dry topsoil layers 332 and increased their root growth in deeper soil horizons. 333 We used our RLD data to estimate the total length of the root system of SL28 and 334 LCICMB1 per plot surface (m.m 2 ) between the soil surface and the root front. Drought stress 335 had contrasted impact on total root length per m 2 in both lines. We observed a strong and 336 significant increase in total root system length in LCICMB1 and a non-significant reduction 337 in total root length in SL28 (Fig 9A). In the water stress treatment, the ratio between total root 338 length (m.m -2 ) and shoot biomass (g.m²) increased in both lines indicating a stronger resource 339 allocation to root growth ( Fig 9B). However, this increase was limited and non significant in 340 SL28 while it was large (> 4 times) and significant in LCICMB1 (Fig 9B). Hence, upon 341 drought stress both pearl millet lines seem to reallocate resources to root growth but this 342 reallocation was stronger in LCICMB1.

344
Here, in order to study pearl millet root system in field conditions, we developed a model 345 to evaluate root length density (total length of roots per unit of soil volume; RLD) from root 346 intersection densities (i.e., the number of root impacts on a vertical soil surface; RID We therefore used our phenotyping method to analyze the response of pearl millet root 389 system to water stress during the vegetative phase in field conditions. Our experiments were 390 performed during the dry season on two germplasms with contrasted characteristics: a dual-391 purpose variety that develops a large aerial biomass and is sensitive to drought and an inbred 392 line with a more limited biomass and that is less sensitive to drought. Our results clearly show 393 that water stress leads to a reallocation of carbon for root growth combined to a reduction of 394 RLD in topsoil layers and to an increase in root system depth. It demonstrates that upon 395 drought stress, pearl millet increases its root growth in deeper soil layer that retain some 396 water. While we cannot conclude from such a small sample, we can hypothesize that this 397 response is adaptive, i.e., that it contributes, with other strategies such as reduction in water 398 loss and temporal regulation of water uptake, to pearl millet tolerance to drought stress.

399
Further work will be needed to test this hypothesis.

400
In conclusion, we developed a simple way to evaluate and map pearl millet RLD