Evaluation of drought resistance and index screening of foxtail millet cultivars

Drought is one of the main restrictors of food production. With the intensification of global warming, drought problems have become more serious and posed a significant challenge to global food security (Davidson 2016). Arid and semiarid regions account for 52.5% of the total area of cultivated land in China. North China, as the main grain production area in the country, faces a serious water shortage problem. Foxtail millet (Setaria italic L.) is a crop that is mainly planted on dry and barren land in North China, because of its high water use efficiency, drought resistance, and barren tolerance (Sachdev 2021; Sun et al. 2021). However, not all foxtail millet varieties are highly drought tolerant. Therefore, to further explore drought-resistant varieties through the drought-resistance evaluation of germplasm resources is of great significance.

The main difficulty in the study of crop drought resistance is accurately identifying drought resistance and screening drought resistance indicators, which requires analysis of different indicators during different periods (Kamoshita et al. 2008). Researchers in recent years have proposed a variety of identification methods for drought resistance in different crops and studied these indicators in terms of morphology, physiology, and biochemistry (Albacete et al. 2014; Zhang et al. 2018). Identifying drought resistance at the germination, seedling, and mature stages is the main method for identifying drought resistance in foxtail millet germplasm resources (Cattivelli et al. 2008). By studying the drought resistance during different periods, researchers believed that the results of drought resistance identification at the mature stage of foxtail millet were relatively reliable (Zhang et al. 2010). In recent years, research on drought-resistant genes and proteins has accurately analyzed and identified the drought-resistance evaluation (Ghatak et al. 2016; Parvathi et al. 2019); however, it is difficult to apply on a large scale because of the high cost of detection. To avoid the one-sidedness and instability of a single index in the process of drought-resistance evaluation, researchers have used a combination of principal component, membership function, cluster, and stepwise regression analyses to carry out the evaluation of multiple indicators (Wang et al. 2017; Xiao et al. 2021, 2022).

However, the drought resistance of crops should ultimately be reflected in the yield. The previous evaluation method could not reflect the absolute yield of foxtail millet under drought conditions (Wang et al. 2017; Xiao et al. 2021, 2022); furthermore, previous evaluation methods usually lacked the research on physiological indicators. Therefore, further evaluating the yield of foxtail millet under dry land conditions in combination with the drought resistance index (DRI) method and analyzing the relationship between physiological traits and drought-resistant characteristics using a comprehensive evaluation method was necessary.

In addition, the present evaluation of foxtail millet drought resistance at the mature period is mostly conducted under natural conditions in the field (Zhang et al. 2012; Shah et al. 2020; Aberkane et al. 2021), which are greatly affected by different rainfall years; therefore, the research results may not be representative. Most studies on the drought resistance of foxtail millet focused on the germination and seedling stages (Dai et al. 2016), and the research at the mature stage was not systematic.

In this study, normal irrigation and drought stress treatments were set under artificial water control conditions by the rainproof shed. The agronomic and photosynthetic indices of 21 foxtail millet cultivars were utilized, and agronomic traits, yield, and photosynthetic traits were analyzed. The aims of the present study were to (1) investigate the effects of drought agronomic traits and photosynthesis traits, (2) to establish a comprehensive drought-resistance evaluation method, and (3) evaluate drought resistance indices and identify strong drought-resistant cultivars.

Twenty-one foxtail millet cultivars were tested, all of which were the main cultivars in China (Table 1).

The experiment was conducted during the 2019 and 2020 seasons using the field simulation drought method with rainproof shed at Dongyang Experimental Station in China (112°45′E, 37°40′N). The area had a semi-humid continental monsoon climate with an annual average temperature of 9.7 °C, annual average precipitation of 445.5 mm, annual sunshine of 2,662.2 h, and annual frost-free period of 158.9 days. The soil in the field was classified as Calcisol with pH of 8.32.

The experiment used a random block design with three replicates at two irrigation levels. The treatments were as follows: (1) normal irrigation treatment (WT): soil water content of 65–75% (soil field capacity) during the growth period; and (2) drought stress treatment (DT): soil water content of 40–50% (soil field capacity) during the growth period. The soil water content was measured by neutron probe. Each plot was separated by concrete and had an area of 6.0 m² (3 m × 2 m).

At the mature stage, agronomic traits, including plant height (PH), grain weight per ear (GWE), ear weight (EW), ear length (EL), ear diameter (ED), kernel ratio (KR, EW/GWE), biomass (BM), and yield (Y) were recorded.

At the flowering stage, five representative plants were chosen from each plot to measure the photosynthetic parameters of the flag leaf at 9:00–11:00 pm under sunny conditions. Six gas exchange parameters, including transpiration rate (T_r), net photosynthesis rate (P_n), intercellular CO₂ concentration (C_i), stomatal conductance (G_s), leaf temperature (LT), and vapor pressure deficit (VPD), were measured by photosynthetic instrument (LI-6800, USA); relative chlorophyll content (SPAD) was measured by SPAD-502 Chlorophyll Meter.

The experimental data were analyzed using Excel 2013 software (Microsoft Corp., USA). Statistical analysis of the data was conducted using SPSS 18.0 software (SPSS Institute Inc., USA). The average values of all indices in 2019 and 2020 were used as basic data. A paired treatment t-test was used to determine the significance of the average difference between the measured values of each index. Drought resistance coefficient (DC) and DRI were calculated using Equations (1) and (2), respectively (Lan 1998; Liu et al. 2015).

Correlation, frequency distribution, and principal component analyses were performed for the DC value of each indicator. According to Equations (3)–(5), the factor weight coefficient (ω_i), membership function value of each index of each genotype [μ(x_i)] and drought resistance comprehensive evaluation value (D value) were calculated and analyzed (Xangsayasane et al. 2014; Bo et al. 2017).

Based on the gray relational analysis (Wang et al. 2019), the correlation between the DC value of each index and the D and DRI values was analyzed. Finally, according to the D value and DRI value, the weighted pair group method average (WPGMA) was used for cluster analysis to classify the drought resistance level.

Drought stress had a significant effect on the agronomic indices of foxtail millet cultivars (Table 2). The coefficient of variation (CV) of the agronomic indices among the different cultivars was between 4.4 and 25.4%, indicating that the tested foxtail millet cultivars were representative. The DT had significant effects for all cultivars, and the agronomic indices were sensitive to drought stress. In addition, the correlation coefficients of the measured values of the foxtail millet cultivars under the WT and DT treatment ranged from 0.429 to 0.767, which further indicate that the sensitivity to drought stress differed for each index. Therefore, identifying the drought resistance of foxtail millet cultivars is difficult when using a single index.

Compared with the normal irrigation treatment, the agronomic indices of different foxtail millet cultivars changed significantly under DT treatment (Table 3). The DC values of cultivars for different indicators were clearly different, with CV ranging from 5.587 to 17.279%. However, the DC values reflected different drought resistances among different cultivars, and the DC values of each index for the same cultivar also differed, indicating that the sensitivity of each index to drought stress is different.

In addition, the distribution times and frequencies of the DC values of the different indices in the same range varied greatly (Table 4). The distribution frequencies of PH, EL, ED, EW, GWE, KR, BM, and Y in the range of DC > 0.8 were 33.3, 14.3, 28.6, 52.4, 42.9, 100.0, 33.3, and 33.3%, respectively, indicating that EL and ED were more sensitive to drought stress. However, the superposition effect of the indices makes evaluating the drought resistance of cultivars difficult when using a single index.

Correlation analysis revealed various significant correlations were found among the indices (Table 5). Y was significantly and positively correlated with EW (P < 0.01), GWE (P < 0.01), BM (P < 0.01), PH (P < 0.05), EL (P < 0.05), and KR (P < 0.05).

Principal component analysis (PCA) was conducted on the DC values of the agronomic indices (Table 6). The first four principal components accounted for 87.73% of data variance. Therefore, PC1, PC2, PC3, and PC4 can accurately reflect differences in the drought resistance of foxtail millet cultivars. PC1 was highly correlated with GWE and Y, PC2 was highly correlated with ED, PC3 was highly correlated with BM and EW, and PC4 was highly correlated with the KR.

Judging drought resistance using a single index is difficult because of the inconsistent order of the drought resistance coefficients for different indicators and cultivars. Therefore, evaluating the drought resistance of cultivars by analyzing the comprehensive evaluation value (D value) according to the weight coefficient of each index is necessary. The D value of the tested cultivars ranged from 0.177 to 0.886, with an average value of 0.562 and a CV of 37.3%, respectively. The drought resistance of the 21 tested foxtail millet cultivars was ranked according to the D value (Table 7). The cultivars with strong drought resistance were Zhangza 13, Yugu 1, Zhonggu 2, Honggu, and Jiugu 23, and those with weak drought resistance were An-04, Changnong 47, and Jigu 22.

However, the D and DC values did not reflect the high-yield character of cultivars under drought stress. Therefore, the DRI based on yield character should also be used to evaluate cultivars to compensate for the limitations of the D value method. The DRI values of the 21 tested cultivars ranged from 0.649 to 1.569, with an average value of 1.005 and a CV of 27.1%. The tested cultivars were ranked according to their DRI value; the cultivars with strong drought resistance were Gonggu 88, Jiugu 23, Zhonggu 2, Jinmiao K2, and Yugu 1, whereas those cultivars with weak drought resistance were Changnong 47, An-04, and Jigu 22. This result was different from that of drought resistance identification based on the D value, but the overall trend was similar.

Gray relational analysis showed that the degree of correlation between the DC value of each index and the D value in turn was ranked from highest to lowest as Y, GWE, EW, EL, ED, BM, PH, and KR. The correlation degree between the DC value of each index and the DRI value was, ranked from highest to lowest, Y, GWE, EW, EL, BM, KR, PH, and ED (Table 8).

The drought-resistance evaluation indicators of the tested cultivars were statistically analyzed according to the drought resistance level classification of the foxtail millet cultivars (Table 9). The membership function values of most indices, DRI value, and D value decreased with an increase in the drought resistance level. The difference in both the D and DRI values at different drought resistance levels was large, which provides a basis for the classification of drought resistance levels among foxtail millet cultivar.

A comparison of the average drought resistance coefficients of the photosynthetic indices for the tested cultivars (Table 10) showed that the drought resistance coefficients of T_r, P_n, C_i, and G_s were all lower than 0.9 under drought stress, indicating that they were greatly affected by drought stress. However, the T, VPD, and SPAD of most cultivars were approximately 1.000, indicating that drought stress had a relatively low impact on these three indices. In comparing the different drought resistance types (according to the cluster analysis), the drought resistance coefficients of C_i, T, VPD, and SPAD showed no significant differences. The P_n, T_r, and G_s of the different drought-resistant cultivars decreased with the weakening of drought resistance. The DC of the strong drought-resistant cultivars had a P_n that was significantly higher than those of the medium and weakly drought-resistant cultivars (P < 0.05), and the drought resistance coefficients of T_r and G_s of the strong and medium drought resistance cultivars were significantly higher than those of the weakly drought resistance cultivars (P < 0.05).

The correlation analysis between the DRI value and the photosynthetic indices showed a positive correlation with all indices except for VPD (Table 11). DRI was positively correlated with P_n (P < 0.01) and T_r (P < 0.05), while the correlations between G_s and SPAD and DRI values were not significantly but relatively high (0.231 and 0.212, respectively). Correlation analysis between the D value and photosynthetic indices showed a positive correlation for all indices. The D value was significantly positively correlated with P_n (P < 0.05), and the correlation coefficients of T_r and VPD with the D values were 0.322 and 0.229, respectively. The analysis of variance and correlation showed that P_n and T_r could be used as reference indices to evaluate the drought resistance of foxtail millet cultivars.

Crop drought resistance is a reflection of multiple drought resistance characteristics, which cannot be reflected in the results obtained using a single index for identification. The membership function method (D value) was used to evaluate drought resistance based on the comprehensive crop indices (Zou et al. 2020). This method emphasizes the absolute drought resistance of materials, which has significance in basic research on drought resistance breeding but does not reflect the high-yield character of cultivars under drought stress. The DRI method measures crop drought resistance based on the crop yield performance under drought stress and normal water condition (Blum 2005), in order to meet production needs and compensate for the limitations of the DC method.

Zhang et al. (2010) utilized the DRI method to identify the drought resistance of foxtail millet and analyzed the relationship between certain indices and DRI values. However, this study did not consider the importance of each index in the evaluation determined by the index variation coefficient. Some researchers believed that multiple agronomic traits combined with the D value can be used as an evaluation parameter to identify the drought tolerance of foxtail millet effectively and accurately (Lapuimakuni et al. 2018; Xiao et al. 2021, 2022), but it cannot precisely evaluate the yield under drought stress conditions.

In this study, the order of drought-resistant foxtail millet cultivars based on D and DRI values differed to a certain extent, and the DRI value can be used as a supplement to the D value to optimize the drought-resistance evaluation. The correlation analysis showed that the degree of correlation between the D value and yield was lower than that of the DRI value. Some cultivars have higher D values and lower DRI values, which indicates that the drought resistance based on several indices is good, but the yield under drought stress is not necessarily high. Combining the D and DRI values for system cluster analysis of drought resistance on the tested cultivars not only considers the importance of each index but also considers the production demand; therefore, the evaluation results are objective and reliable.

Drought resistance in crops is a complex quantitative trait that involves multiple mechanisms and factors. Therefore, screening for suitable indicators is key for determining the drought resistance. Drought resistance in foxtail millet is a concentrated expression of plant morphology and yield after a series of adaptive changes in morphological structure, physiology, and biochemistry of cells under drought stress (Kamoshita et al. 2008; Song et al. 2017; Gupta et al. 2020). Generally, growth and yield-related indices are reliable indicators of drought resistance. According to the study on drought resistance of foxtail millet, yield character is most sensitive to drought stress (Xiao et al. 2021). Previous studies also revealed that the effect of drought stress on the physiological and biochemical indicators of wheat was greater than that on agronomic indicators; however, the physiological and biochemical indices were greatly affected by the test period (Chernyad'ev & Monakhova 2003; Subrahmanyam et al. 2006).

In this study, drought stress affected agronomic and photosynthetic indices to different degrees, with more significant effects on agronomic indices. The range of drought resistance coefficients of agronomic indices was 0.705–0.796 (except for KR), whereas those of photosynthetic indices were all above 0.800 (except for Ci). However, foxtail millet indices were affected by drought stress to different degrees, with certain correlations between each index. Therefore, evaluating the drought resistance of cultivars objectively and accurately using these indicators is difficult. Gray correlation analysis showed that the four traits with a higher degree of correlation with the D value were GWE, Y, EW, and EL. Unlike wheat and other crops (Subrahmanyam et al. 2006; Cattivelli et al. 2008), the PH and ED of foxtail millet showed a weak correlation with drought resistance and were not suitable to be used as the key indices for screening of drought-resistant cultivars. In addition, variance analysis and correlation analysis showed that the P_n and T_r could also be used as reference indicators to evaluate the drought resistance of foxtail millet. Therefore, in addition to using agronomic traits as screening indicators of drought resistance, further research must focus on analyzing physiological and biochemical indicators. However, these are unstable and difficult to control during the measurement period and are easily affected by factors such as growth environment and period (Reddy et al. 2004). Research on physiological, biochemical, and molecular mechanisms (metabolomics and transcriptomics) must be strengthened in the further drought-resistance evaluation (Mishra et al. 2012; Shah et al. 2020).

Drought stress had a significant effect on most agronomic indices and some photosynthetic indices of foxtail millet cultivars at a mature stage. The combination of D and DRI values was determined to be an appropriate drought resistance identification index. Foxtail millet cultivars with strong drought resistance at a mature stage were selected. The GWE, Y, EW, and EL of foxtail millet can be used as simple and intuitive drought-resistance evaluation indicators. P_n and T_r could also be used as reference indicators to further determine the drought resistance of foxtail millet.

This study was supported by Shanxi Major Research and Development project (201703D211002-2) and Shanxi Agricultural Academy Science Foundation (YCX2020YQ36). We thank the editor and reviewers for their valuable comments and suggestions.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.