Detrimental effects of heat stress on grain weight and quality in rice (Oryza sativa L.) are aggravated by decreased relative humidity

Plant materials

Six rice cultivars with different heat tolerances at the grain filling stage were sourced from our previous studies. They include three rice hybrids: Liangyou27 (LY27), Liangyou6 (LY6), and Zhuliangyou47 (ZLY47), and three conventional varieties: 16343, R168, and IR64. R168 had a smaller grain weight and quality difference between temperature conditions. IR64 and the other four varieties showed sensitivity to high temperature (Table S1).

Growth conditions

A pot experiment was conducted in 2017 on Yangtze University experimental farm (Jingzhou City, 112°09′E, 30°21′N, 32 masl) in the western part of the Jianghan Basin in China.

Seeds were sown on April 8, 2017. Twenty-day-old seedlings were transplanted to each plastic pot (inner diameter 30 cm, height 30 cm) containing 12.5 kg soil and 8 g N:P:K compound fertilizer (26:10:15). After transplantation, the soil surface in the pots was kept submerged until maturity in natural conditions. Tillers were cut off after they emerged, leaving only the main stem in each plant throughout the entire growth period.

Treatment

Six controlled environments with a combined air temperature and relative humidity (RH) level were simulated in a growth chamber (AGC-MR, Zhejiang Qiushi Environment Co., Ltd., Zhejiang, China). Experimental treatments included two RH levels (i.e., RH75% and RH85%), three temperature levels (daily maximum temperatures of 33 °C, 35 °C, and 37 °C), and two durations (8 and 15 days after anthesis) (Fig. 1). Hourly temperature changes were recorded daily, and the results are shown in Fig. 1. The RH was kept constant in each treatment environment.

Panicles heading on the same day were marked for treatments after anthesis. Six pots of each cultivar were subjected to each combination of the controlled environment (temperature × RH combinations) for eight and 15 days, respectively. The plants grown under natural conditions (six pots per cultivar) were used as controls. Temperature and RH are shown in Table S2.

Measurements of grain weight, milling quality, and chalkiness

Filled grains were harvested and sun-dried to a moisture content of 13%. We measured thousand-grain weights with three replications. Samples were stored for 3 months (15–20 °C, RH: 10–20%) and then the milling quality and chalkiness were measured. Rice grain samples (30 g for each cultivar per treatment) were dehulled and polished for 30 s (JDMZ100, Beijing Dongfu Jiuheng Instrument Technology Co., Ltd., Beijing, China) to obtain milled rice. The head rice was then separated and weighed. The weight of the head rice to the sample weight (30 g) was calculated as the head rice rate. The chalkiness of the head rice was evaluated with a rice appearance quality tester (JMWT12, Beijing Dongfu Jiuheng Instrument Technology Co., Ltd., Beijing, China).

Stress tolerance estimation

Stress tolerance was evaluated by the membership function value (MFV) based on the theory of fuzzy mathematics (Zadeh, 1965). We used a method modified from Chen et al. (2012) and Liu, Yang & Hu (2015). The heat-tolerant coefficient (HC) was calculated as the ratio of the value in the combinative treatment of temperature × RH × duration to that in the control of the same cultivars for individual traits, using the following equation:

$$\mathrm{H}{\mathrm{C}}_{ijk}={\displaystyle \frac{{\mathrm{X}}_{ijk}}{\mathrm{C}{\mathrm{K}}_{ij}}}$$where HC_ijk is the heat-tolerant coefficient of the trait (i) for cultivar (j) in treatment (k), X_ijk is the value of the trait (i) for the cultivar (j) in the treatment (k), CK_ij is the value of the trait (i) for the cultivar (j) under the control condition.

As with HC, MFV for grain weight and head rice rate were calculated, using the following equation:

$${\mathrm{M}\mathrm{F}\mathrm{V}}_{ijk}={\displaystyle \frac{{\mathrm{H}\mathrm{C}}_{ijk}-\mathrm{M}\mathrm{i}\mathrm{n}\left(\mathrm{H}\mathrm{C}{}_i\right)}{\mathrm{M}\mathrm{a}\mathrm{x}\left(H{C}_i\right)-\mathrm{M}\mathrm{i}\mathrm{n}\left(H{C}_i\right)}}$$where MFV_ijk is the membership function value of heat tolerance of the trait (i) for the cultivar (j) in treatment (k), HC_ijk is the same as earlier defined, HC_i is the heat-tolerant coefficient value of the trait (i) over all cultivars and all treatments.

The chalkiness value tends to increase after treatment, which is contrary to the changes in grain weight and milled rice rate, and the absolute value of chalkiness under control conditions is usually very small. The MFV for chalkiness was calculated using the following equation:

$${\mathrm{M}\mathrm{F}\mathrm{V}}_{ijk}=1-{\displaystyle \frac{X_{ijk}-\mathrm{M}\mathrm{i}\mathrm{n}\left(X_i\right)}{\mathrm{M}\mathrm{a}\mathrm{x}\left(X_i\right)-\mathrm{M}\mathrm{i}\mathrm{n}\left(X_i\right)}}$$where X_i is the value of the trait (i) over all cultivars and all treatments.

For these three traits, MFVs are dimensionless, real number interval [0,1], standing for individual cultivar′s heat tolerance under treatments.

Data analysis

We used Microsoft Excel 2019 for data entry and collating and conducted the analysis of variance (ANOVA) using the R package “agricolae” in R 3.6.0 to determine the effect of treatment factors on grain weight and quality. We performed comparisons between treatment means using the least significant difference test (LSD) at P ≤ 0.05.

A full model was established via function “lm” in R package “stats” to evaluate the treatment factors and their interactions on grain weight, head rice, and chalkiness. Then, a stepwise backward selection based on the Akaike’s information criterion (AIC) which performed with function “stepAIC” in R package “MASS” in R 3.6.0 was used to find out the factors that affected the grain weight, head rice rate, and chalkiness most (Venables, Ripley & Venables, 2002). The significance of each predictor was tested via a Student’s t-test, all treatment factors and their interactions left in the model were significant at the 0.01 level.

The full model of multiple linear regression as follow:

$$Y={\beta }_1{X}_1+{\beta }_2{X}_2+{\beta }_3{X}_3+{\beta }_4{X}_4+{\beta }_{12}{X}_1{X}_2+{\beta }_{13}{X}_1{X}_3+{\beta }_{23}{X}_2{X}_3+{\beta }_{123}{X}_1{X}_2{X}_3+e$$where Y represents grain weight, head rice rate, and chalkiness in each treatment, respectively; X₁ represents the max daily temperature in each treatment; X₂ represents the relative humidity in each treatment; X₃ represents the treatment duration days in each treatment; and X₄ represents the grain weight, head rice rate, and chalkiness of each cultivar under controlled conditions to eliminate the differences in the variety; β₁, β₂, β₃, and β₄, are linear coefficients, β₁₂, β₁₃, β₂₃, and β₁₂₃ represent interaction coefficients, with e representing the intercept.

Grain weight

The effects of temperature, RH, and temperature × RH on grain weight were significant among six cultivars (Table 1). The duration of treatment also had a significant effect on grain weight, except for R168. Grain weight was significantly reduced in cultivars at RH 75% coupled temperature treatments, and the reduction was higher than that in RH 85% over the same three temperature treatments (Fig. 2). Under the same temperature and RH conditions, grain weight significantly decreased from 8 d to 15 d of the duration treatment (Fig. 2). The impact of different RH regime treatments (RH75% vs. RH85%) on grain weight was smaller at 37 °C than at 33 °C and 35 °C (Fig. 3).

Multiple regression analysis showed that grain weight decreased when temperature and durations increased, which was the opposite effect when RH increased (Table 2).

MFVs of each trait displayed in Table 3 showed the heat tolerance of each cultivar under treatments, with the mean value as a comprehensive index for the evaluation of heat stress tolerance of each cultivar. R168 was found to have the highest mean MFV (0.70), followed by IR64 (0.55), then LY27, LY6, ZLY47, and 16343 (0.54, 0.50, 0.45, and 0.44, respectively). R168 also showed the smallest difference of MFVs between RH conditions. The treatment of 35°C by RH75% at 15 days showed the largest mean MFV difference among the cultivars and could be used to evaluate high-temperature tolerance.

R168 and IR64 were also the two cultivars with the highest mean MFVs of grain weight and the smallest difference between RH conditions (Table 3). Their grain weight decreased >2 g at 37 °C (Fig. 2). ZLY47 and 16343 showed high MFVs under temperatures of 33 °C and their grain weight decreased >2 g at 35 °C or 37 °C. LY27 and LY6 showed small MFVs at RH75% over temperature treatments, and their grain weight decreased >2 g at all three temperature treatments.

Head rice rate

The head rice rate of all cultivars was significantly affected by temperature and RH (Table 1). Durations also showed a significant detrimental effect on the head rice rate except in cultivar ZLY47. Interactive effects of temperature × durations and RH × durations were significant for the head rice rate among cultivars, except LY6. The head rice rate significantly decreased in cultivars at RH75% over temperature treatments, and the reduction was higher than that at RH85% over the same temperature treatments when compared to the control (Fig. 4). The difference in head rice rate between the two RH treatments (RH75% vs. RH85%) was greatest at 35 °C and lowest at 37 °C (Fig. 3), suggesting that RH has a pronounced effect on temperature-induced head rice loss at 35 °C. A pronounced difference was found at 33 °C only in cultivars whose head rice rate was dramatically decreased (LY6 and ZLY47, Fig. 4).

R168 was the most heat-tolerant in terms of head rice rate, except for the treatments of 37 °C × RH 75%, with an MFV of >0.70 (Table 3). In addition, R168 showed a stable and higher head rice rate than other cultivars, even at 37 °C (Fig. 4). The MFVs of head rice rate for 16343 was ≥0.85 at 33 °C but dropped sharply at 35 °C in both humidity treatments (Table 3). For other cultivars, the temperature caused a sharp drop in response to RH. At RH 75% over temperature treatments, the temperature was 33 °C for LY6 and ZLY47, or 35 °C for LY27; under RH85%, the temperature was 35 °C for LY6 and ZLY47, or 37 °C for LY27.

Chalkiness

For all cultivars, the temperatures, RH, and the interaction of temperature × RH had a significant effect on chalkiness (Table 1). The treatment durations had a significant effect on chalkiness for cultivars except 16,343. The majority of cultivars had higher levels of chalkiness at RH75% than at RH85%, even with the same temperature and duration treatment (Fig. 5). The difference in chalkiness between RH treatments (RH75% vs. RH85%) was most pronounced at 35 °C, followed by 33 °C (Fig. 3). However, the difference was the smallest at 37 °C. Multiple regression analysis showed that chalkiness was higher when the temperature or durations increased (Table 2).

R168 and IR64 showed the highest MFVs of chalkiness in all treatments (Table 3). R168 also showed the smallest chalkiness difference between RHs (Fig 5). Compared to the control, the chalkiness of R168 sharply increased (more than 10%) only at 37 °C by the RH treatments; this was the case for IR64 at 35 °C and 37 °C, or for the remaining cultivars at all three temperature treatments (Fig. 5).

The interaction effect of temperature and RH

During the flowering stage, panicle temperature rather than air temperature was curvilinearly related to spikelet fertility (Weerakoon, Maruyama & Ohba, 2008). Rice could homeostatically adjust panicle temperature via transpiration cooling for optimal growth. The air-panicle temperature difference was altered under different RH conditions (Fukuoka, Yoshimoto & Hasegawa, 2012; Yoshimoto et al., 2011). Yan et al. (2008) showed that at temperatures of 31.5–33.5 °C, the temperature difference between the air and panicle was about 2 °C under humid atmospheric conditions (~86% RH) or about 5 °C under a dry atmospheric condition (~48% RH). Rice panicles benefit from their transpiration cooling under heat stress at flowering (Matsui et al., 2014). However, our results showed decreased RH, coupled with a high temperature at the grain filling stage reduced grain weight and lowered grain quality. The benefit from the decreased RH by transpiration cooling may not be the only scenario of RH and temperature interaction.

Theoretically, there may be a balance between transpiration cooling and a water deficit in rice panicles exposed to heat. Zhao & Fitzgerald (2013) reported that a daily maximum temperature ranging from 30 °C to 33 °C and a lower RH led to higher head rice yield and lower chalkiness. Meanwhile, Wada et al. (2011) showed dry winds at day/night temperatures of 34/26 °C caused water deficiencies in panicles and restricted starch accumulation, which led to a decline in the quality of the rice. We found that decreased humidity coupled with temperature treatments caused grain weight reduction and quality loss. This effect was most pronounced at 35 °C, and to a lesser extent at 33 °C, and insignificant at 37 °C. This indicates that transpiration cooling may not be enough to compensate for the adverse effects of water deficit when rice is exposed to high-temperature stress (35 °C and 33 °C).

Physiological mechanism of temperature and RH interaction

Water fluctuations for growth and transpiration are linearly superimposed (Nonami & Hossain, 2010), and the impaired ability of stomatal regulation of rice spikelets has a greater evaporative demand under high temperature (Garrity, Vidal & O’Toole, 1986), leaving the spikelets and grains at risk of water deficit (Tanaka & Matsushima, 1971). Water deficiency in the panicle caused by dry, hot winds was detrimental to rice grain weight and quality formation (Hiroshi et al., 2012; Kang et al., 2003; Wada et al., 2011). Low relative humidity may reduce the assimilation capacity of leaf photosynthesis (Tanaka & Matsushima, 1971). The normal maturing process of grains is dependent on a sufficient water supply (Cochrane, Paterson & Gould, 2000; Ferrise, Bindi & Martre, 2015). Thus, a higher temperature with lower humidity may induce high temperature-forced maturity.

Panicle water potential is temporarily reduced during periods of dry, hot wind stress (Wada et al., 2011). The osmotic adjustment of endosperm cells with increased transport was activated to maintain kernel growth, but starch biosynthesis was slowed (Wada et al., 2014). The vacuolar structures in the cytosol were preserved during maturity because of osmotic adjustments, resulting in ring-shaped chalkiness (Hatakeyama et al., 2018). High-temperature stress led to substantial solutes to be accumulated in endosperm cells by osmotic adjustment. This was accompanied by the partial inhibition of amyloplast development and the formation of protein bodies, which caused air spaces to remain in endosperm cells during grain dehydration, resulting in a chalky appearance (Wada et al., 2019). Similar mechanisms of starch synthesis restriction were found under high temperature and dry wind conditions, meaning that decreased RH may aggravate the high-temperature effect on grain weight and quality loss.

The osmotic regulation ability of plants is closely related to the heat tolerance during the vegetative growth period (An, Zhou & Liang, 2014; Jiang & Huang, 2001; Zhou et al., 2018; Zou et al., 2016), and the study of Wada et al. (2014) suggests that this relationship is also applicable at the filling stage of rice. We speculate that the differences in variety performance in this study may be related to osmotic regulation, but further study is needed.

Varietal differences under RH in combination with temperature treatments

A daily mean air temperature over 25 °C during the grain filling stage can cause a loss of quality in the rice grains (Morita, Wada & Matsue, 2016; Wu, Chang & Lur, 2016). However, the magnitude of heat stress-induced damage varied by genotype. Cooper, Siebenmorgen & Counce (2008) reported a reduction in grain quality in susceptible cultivars at nighttime temperatures of 20 °C, but this effect was not reflected in heat-tolerant cultivars at the nighttime temperature of 30 °C. We found that grain weight and quality traits did not change linearly with increasing temperatures and only changed dramatically at a certain temperature, with varied responses among different cultivars (Figs. 2, 4, and 5). This was similar to the response of spikelet fertility to high temperatures during flowering (Jagadish, Craufurd & Wheeler, 2007).

The identification of heat-tolerant rice germplasms (e.g., N22) may present opportunities to breed heat-tolerant rice at the flowering stage (González-Schain et al., 2015; Tetsuo & Shouichi, 1978). The mechanism leading to grain quality losses is more complex than the mechanism of heat-induced yield loss (Jagadish, Murty & Quick, 2015), which brings more challenges for screening and identifying tolerant varieties. Previous studies paid closer attention to temperatures in controlled environments when investigating rice heat tolerance at the grain filling stage (Chen et al., 2017; Shiraya et al., 2015; Tanamachi et al., 2016). We showed that changes in grain weight and quality are affected by interactions between temperature and humidity and found that humidity is important in evaluating varietal heat tolerance. We suggest that the effect of humidity should be considered in multi-variety tolerance screening and identification. The optimum combination of 35 °C by RH75% by 15 days should be recommended to screen for heat tolerance of rice. Our results bring attention to the detrimental interactive effects of high temperature and humidity on rice yield and quality and are of interest to breeders and agronomists to adjust breeding targets. R168, the most heat-tolerant cultivar used in this study, showed smaller differences in grain weight and quality between two RH regimes. This variety may be selected as a heat-tolerance variety to improve rice yield and quality under climate change.