Divergent phenotypic response of rice accessions to transient heat stress during early seed development

Abstract Increasing global surface temperatures is posing a major food security challenge. Part of the solution to address this problem is to improve crop heat resilience, especially during grain development, along with agronomic decisions such as shift in planting time and increasing crop diversification. Rice is a major food crop consumed by more than 3 billion people. For rice, thermal sensitivity of reproductive development and grain filling is well‐documented, while knowledge concerning the impact of heat stress (HS) on early seed development is limited. Here, we aim to study the phenotypic variation in a set of diverse rice accessions for elucidating the HS response during early seed development. To explore the variation in HS sensitivity, we investigated aus (1), indica (2), temperate japonica (2), and tropical japonica (4) accessions for their HS (39/35°C) response during early seed development that accounts for transition of endosperm from syncytial to cellularization, which broadly corresponds to 24 and 96 hr after fertilization (HAF), respectively, in rice. The two indica and one of the tropical japonica accessions exhibited severe heat sensitivity with increased seed abortion; three tropical japonicas and an aus accession showed moderate heat tolerance, while temperate japonicas exhibited strong heat tolerance. The accessions exhibiting extreme heat sensitivity maintain seed size at the expense of number of fully developed mature seeds, while the accessions showing relative resilience to the transient HS maintained number of fully developed seeds but compromised on seed size, especially seed length. Further, histochemical analysis revealed that all the tested accessions have delayed endosperm cellularization upon exposure to the transient HS by 96 HAF; however, the rate of cellularization was different among the accessions. These findings were further corroborated by upregulation of cellularization‐associated marker genes in the developing seeds from the heat‐stressed samples.


| INTRODUC TI ON
Present-day agriculture is facing multifaceted challenges. The constant shifts in global food demands associated with exponential growth in world population, shrinking arable land, and increased frequency of extreme events need to be addressed urgently to mitigate future global food security crisis (Godfray et al., 2010;Foley et al., 2011;Röth, Paul, & Fragkostefanakis, 2016). Among the climatic variables, rising temperature is one of the most detrimental for crop yields (Porter & Gawith, 1999;Zhao et al., 2017). The average global temperature is anticipated to increase by 0.2°C per decade over the next few decades (Lobell, Schlenker, & Costa-Roberts, 2011). In this context, a recent study estimates a 3%-8% decline in the yields of rice, wheat, and maize, for each degree-Celsius increase in temperature (Zhao et al., 2017). These three cereal crops are a major food source worldwide; thus, a decline in yields will likely cause increasing malnutrition and socio-economic unrest.
These grains are more prone to breakage during the milling process (Sreenivasulu et al., 2015). Despite our understanding of HS response of rice during vegetative, reproductive, and grain filling stages, information on impact of HS on early seed development in rice is sparse.
Endosperm is the dominant tissue and the primary determinant of seed size in the monocot crops rice, wheat, and maize (Chaudhury et al., 2001;Gao, Xu, Shen, & Wang, 2013). Postfertilization, endosperm undergoes rapid nuclear division without cytokinesis (syncytial phase). The free nuclear divisions are followed by cell wall formation (cellularization), initially around the periphery of a central vacuole and eventually filling the entire inner cavity. Afterward, endosperm development enters the grain filling phase, where it accumulates storage compounds such as starch, proteins, and lipids, thus providing a major food source worldwide (Xing & Zhang, 2010;Zuo & Li, 2014). Historically, the majority of rice-producing regions prefer specific sub-populations, for example, indica in South Asia, temperate japonica in North Asia, and tropical japonica in South America ; however, with the current climate scenario, we need to explore the diverse germplasm to breed climate resilient rice. In this context, we aimed to evaluate the effect of transient HS on early seed development in a diverse set of rice accessions and determine the outcomes of this stress at seed maturity. Our subset of rice accessions represented the four sub-populations (aus, indica, temperate japonica, and tropical japonica). Based on their response, we categorized the accessions into highly sensitive, moderately, and strongly heat tolerant.
Our work provides an important component for the knowledgebase needed to utilize a diverse gene pool to develop heat-resilient rice.

| Plant material and growth conditions
Nine rice accessions from Rice Diversity Panel 1 (RDP1; Zhao et al., 2011) were selected based on their relatively similar flowering time (<120 days) to ensure that the accessions flowered under similar light conditions in the glasshouse. Genetic and geographical diversity of the accessions was also considered (Table 1). Mature seeds were dehusked using a Kett TR-130, sterilized with water and bleach (40%), and germinated in dark on half-strength Murashige and Skoog media. After 5 days, germinated seedlings were transplanted to soil in 4-inch square pots, and plants were grown in control glasshouse conditions: 16-hr light and 8-hr dark at 28 ± 1°C and 23 ± 1°C, respectively and relative humidity of 55%-60% was consistently maintained throughout the plants' life cycle .

| Heat stress treatments
Plants were grown in control conditions until flowering ( Figure 1a). To track developing seeds, florets were marked at the time of fertilization.
Twenty-four hours after fertilization (HAF), plants were kept in either control conditions (16-hr light and 8-hr dark at 28 ± 1°C and 23 ± 1°C) or moved to a reach-in growth chamber for heat stress (HS) treatment (16-hr light and 8-hr dark at 39 ± 1°C and 35 ± 1°C). The HS treatment was applied for either 1 (HS1), 2 (HS2), or 3 (HS3) days corresponding to 48, 72, and 96 HAF, respectively ( Figure 2a). Afterward, plants were moved back to the control condition until maturity for evaluating the effect of HS on mature seed traits. Three independent HS experiments (in 2016, 2017, and 2018) were conducted with 10-20 plants per treatment per accession in each experiment.

| Analysis of developing and mature seeds
To accurately assess the effect of HS on early seed development, only the florets marked at the time of fertilization were considered for downstream analysis. For mature seed phenotyping, seeds harvested at maturity were examined. Total number of fully developed and unfilled or completely sterile seeds were scored to calculate percentage of fully developed seeds. For morphometric analysis, marked mature seeds were dehusked and scanned using Epson Expression 12,000 XL scanner. The images were analyzed using SmartGrain (Tanabata, Shibaya, Hori, Ebana, & Yano, 2012). For mature seed analysis, 400-1,000 marked seeds from 20 to 40 plants were evaluated. For developing seed analysis, marked seeds from control or HS-treated plants were harvested and imaged at 24, 48, 72, and 96 HAF. The images were processed using ImageJ (Abramoff, Magalhães, & Sunanda, 2004) to extract the length of developing seeds. For developing seed analysis, 15-20 marked seeds from 3 to 4 plants were evaluated.

| Histochemical analysis
The control and HS developing seeds (72 and 96 HAF) were harvested and fixed in 1 ml of formaldehyde (2%), acetic acid (5%), and ethanol (60%; FAE solution) for 16 hr. Tissue was then washed with 70% ethanol and stored overnight at 4°C, followed by dehydration with an ethanol series (85%, 95%, and 100%) for 1 hr each at room temperature.
Samples were transferred to xylene (100%) for 2 hr, followed by transfer to 500 µl of a 1:1 mixture of xylene and paraplast tablets and stored at 60°C. Finally, samples were transferred to and embedded in paraplast (Paul et al., 2020). Cross sections (10 µm) were obtained using a rotary microtome (Leica RM2125 RTS). Images from the cross sections were observed using bright-field microscope (Leica DM-2500). For the experiment, six to eight developing seeds for each developmental stage from two independent experiments were considered for cross sections.

| RNA Extraction and RT-qPCR
Total RNA was extracted from developing seeds (48 and 96 HAF) using the RNeasy Plant mini kit (Qiagen), followed by DNase treatment.
One microgram of total RNA was used for cDNA synthesis using the SuperScript VILO cDNA synthesis kit (Invitrogen). The RT-qPCR (10 µl) comprising gene-specific primers, SYBR Green Master Mix (Bio-Rad) and template, was conducted using LightCycler 480 Real-Time PCR System (Roche). A ubiquitin (UBQ5) gene was used as an endogenous control (Jain, Nijhawan, Tyagi, & Khurana, 2006). Data were analyzed using standard methods (Livak & Schmittgen, 2001) and represented as log 2 fold changes (Fragkostefanakis et al., 2015). For all RT-qPCR assays, a minimum of two independent biological and three technical replicates was evaluated. The primers used to analyze the syncytial and cellularization-associated genes are listed in Table S1.

| Statistical analysis
Descriptive statistics and analysis of variance (ANOVA) for mature seed traits were performed in statistical software R version 3.5.3 (R Core Team 2018). Principle component analysis for traits derived from mature seeds was performed using R package FactoMineR and factoextra (Lê, Josse, Rennes, & Husson, 2008). Trait correlations between mature seed traits in control and HS treatments were performed using functions cor.matrix() and corrplot() from the corrplot R package (Wei et al., 2017). Statistical analysis used for analyzing individual experiments is explained in the figure legends.

| Genotypic level heat sensitivity of fully developed mature seeds is conversely related to single-grain weight
To understand the impact of HS on early seed development, selected rice accessions (Table 1) were subjected to transient HS for three days, that is, 24-96 hr after fertilization (HAF; Figure 1a). Only the florets marked at time of fertilization were considered for downstream TA B L E 1 Genetic and geographical information of the rice accessions used in the current study analysis, which ensured accurate assessment of HS impact on early seed development. We detected genotypic differences in response to transient HS during early seed development as measured by the percentage of fully developed marked seeds ( Figure 1 and Table 2).
In this context, 3/9 tested accessions corresponding to indica (IND-1 and 2) and a tropical japonica (TRJ-3) showed drastic reduction in number of fully developed seeds ( Figure 1b). One indica accession (IND-2) showed complete sterility under HS treatment, as we did not recover any developed mature seed, so the respective accession was not part of the downstream mature seed measurements (Figure 1b and c). On F I G U R E 1 Response to post-zygotic transient heat stress exhibits high level of phenotypic variation in seed viability among a diverse set of rice accessions. (a) Florets were marked at the time of fertilization. Twenty-four hours after fertilization (HAF), plants were subjected to HS (39/35°C; day/night) or kept in control conditions (28/23°C; day/night). The plants were heat stressed for 3 days (i.e., until 96 HAF) and returned to control conditions. At maturity, only the marked florets (now seeds) were scored as sterile or fully developed seeds. (b) Percentage of fully developed seeds and (c) single-grain weight from control and heat-stressed plants. Box plots represent range and median for the same. For statistical analysis, paired t test was used to compare the two treatments; n = 700-1,000 marked seeds from 25 to 40 plants, *** indicates p < .001 and ** p < .01. (d) Representative mature seed images from control and transient heatstressed accessions. Images were digitally extracted and scaled for comparison (scale 1 cm). NA: not available as no mature seeds were recovered for the respective accessions the other hand, aus (AUS-1), one of the temperate japonicas (TEJ-2), and three tropical japonicas (TRJ-1, −2, and −4) also showed significant reductions; however, reductions were relatively less severe ( Figure 1b).
One of the temperate japonicas (TEJ-1) exhibited strong resilience to transient HS, as number of fully developed seeds recovered in the heat-stressed plants was comparable to the respective control plants ( Figure 1b). We also measured seed parameters for the seeds that did fully develop for each accession. The indica accession that showed a steep decline for fully developed seeds (IND-1) was not affected in single-grain weight for the fully developed seeds at maturity. Contrarily, the four accessions maintaining the relatively higher mature seed numbers under stress (TEJ-1, TEJ-2, TRJ-1, and TRJ-4) showed a significant reduction with respect to single-grain weight (Figure 1c).
Given the phenotypic variability for the fully developed seeds

| Heat stress penalizes seed length of the relatively resilient accessions
To elucidate the impact of the three-day HS during early seed development on seed size, morphometric measurements on fully developed control and heat-stressed seeds were performed. In this context, area, perimeter, length, width, and length-to-width ratio of the individual seeds were evaluated. A significant difference between control and HS treatments was observed for all measured parameters for certain accessions except for length-to-width ratio, indicating the prevalent treatment effect (Table 3). Interestingly, both temperate japonica accessions (TEJ-1 and TEJ-2) showed a significant reduction in seed length under HS. Further, seed area and width were also reduced for TEJ-2, while these traits were significantly heat-tolerant for TEJ-1 (Table 3). Among the tropical japonica accessions, only TRJ-2 showed a significant reduction in seed area, length, width, and perimeter. Another tropical japonica accession, TRJ-3, showed marginal reduction in seed width under HS. It is notable that the indica accession IND-1 maintained seed size parameters, while the aus accession AUS-1 showed reduction only in seed area (Table 3) indicating that the respective trait is independent. A significant correlation was detected for single-grain weight with seed width under control as well as HS treatments (Figure 3). Single-grain weight was weakly correlated with seed length under control, while no correlation existed between the respective traits under HS (Figure 3).

| Principle component analysis
To

| Resilient accessions sense heat stress prior to the sensitive accessions
Because we observed differences in seed length at maturity for four of the relatively heat-tolerant accessions (TEJ-1, TEJ-2, TRJ-1,

TA B L E 3
Morphometric analysis of mature seeds. and TRJ-2), we next aimed to capture the genotypic differences in growth of the developing seeds exposed to control and HS treat-  (Table 3) and measuring the grain length increase daily provides further insight into genotypic differences to HSR ( Figure 5).

| Histochemical analysis reveals differential rate of endosperm cellularization
The observed differences in seed morphology at maturity and length of developing seed prompted us to explore potential genotypic variation in the cellular biology of early seed development under heat stress. Previous reports have shown that the developing endosperm is at the syncytial stage at or around 48 HAF and completes cellularization around 96 HAF in a model rice variety, Kitaake (Chen et al., 2016;Folsom et al., 2014). We observed that the cellularization timing was affected not only by HS, but also varied among accessions even under control conditions ( Figure 6 and Figure Figure 6 and Figure S2).
On the other hand, genes that are upregulated with the progression of endosperm cellularization relative to the syncytial phase, for example, OSTF1 (Yang, Chung, Tu, & Leu, 2002), SSIIa (Folsom et al., 2014), ZmEBE-1 (French, Abu-Zaitoon, Uddin, Bennett, & Nonhebel, 2014), and others reported by Chen et al (Chen et al., 2016), are referred to as cellularization-associated (Figure 7). For these syncytial and cellularization marker genes, RT-qPCR analy- HAF under control conditions ( Figure 6). As expected, we observed F I G U R E 5 Effect of transient heat stress on length of developing young seeds. Plants with marked florets at the time of fertilization were subjected to control or HS treatment. The HS treatment, initiated 24 HAF, was for either 1 day (24-48 HAF), 2 days (24-72 HAF), or 3 days (24-96 HAF). The marked developing seeds from control and heatstressed plants were harvested at 48, 72, and 96 HAF. Seed length was measured using ImageJ. Representative developing seeds from the respective time-point, treatment, and accessions are shown. Images were digitally extracted and scaled for comparison (scale 1 cm). For statistical analysis, paired t test was used to compare developing seed lengths from control and HS treatments for a given time-point; n = 15-20 marked developing seeds, *** indicates p < .001 and **p < .01. HAF, hours after fertilization down-and upregulation of syncytial-and cellularization-associated genes, respectively, by 96 HAF relative to 48 HAF under control conditions ( Figure 7a). However, the aus accession (AUS-1) showed exceptions, as DRM3 corresponding to syncytial-associated genes showed upregulation, while two cellularization-associated genes (OSTF1 and dirigent) were downregulated by 96 HAF (Figure 7a).
Next, we examined the expression of the selected genes at 48 HAF under HS relative to 48 HAF control (Figure 7b). Three syncytialassociated genes (CMT3, KRP3, and MADS89) were downregulated in all the tested accession for the heat-stressed samples from 48 HAF, wherein DRM3 and MADS79 were significantly upregulated in at least one of the accessions (Figure 7b). Two of the cellularizationassociated genes (expressed protein: LOC_Os05g34510 and OSTF1) were upregulated in all the tested accession, while other cellularization-associated genes (SSIIa, TPP8, ZmEBE-1, CSLA9, and dirigent) were downregulated in at least one of the tested accessions for the heat-stressed samples from 48 HAF (Figure 7b). We also examined the expression of these genes at 96 HAF under HS relative to 96 HAF control (Figure 7c). In this context, 5/7 syncytial-associated (except CMT3 and KRP3) and 6/7 cellularization-associated (ex-

| D ISCUSS I ON
The global mean temperature rise over the last century has negatively influenced crop yields (Battisti & Science, 2009;Challinor et al., 2014;Gourdji, Sibley, & Lobell, 2013;Lobell et al., 2011;Zhao et al., 2017). The situation is expected to worsen given the predicted increase in global temperatures (Gourdji et al., 2013;Zhao et al., 2017). Extensive work has been carried out regarding the impact of HS on different phases of plant development (Bokszczanin & Fragkostefanakis, 2013;Cheng, Sakai, Yagi, & Hasegawa, 2009;Hasanuzzaman, Nahar, Alam, Roychowdhury, & Fujita, 2013;Jagadish et al., 2015;Prasad, Bheemanahalli, & Jagadish, 2017). Both reproductive development and grain filling phases in rice have been documented as sensitive to environmental perturbations (Prasad et al., 2006;Jagadish et al., 2007;Prasad et al., 2008;Jagadish et al., 2010;Coast, Ellis, Murdoch, Quiñones, & Jagadish, 2015;Arshad et al., 2017;Ali et al., 2019;Dhatt et al., 2019). However, genotypic variation in rice germplasm for heat stress responses during early seed development remains largely unexplored. In the present study, we systematically investigated the impact of HS on the early seed development that corresponds to the transition phase of endosperm from syncytial to cellularized. The rate and timing of this transition is one of the main drivers for determining final seed size (Brown et F I G U R E 6 Transient heat stress during early seed development delays endosperm cellularization. (a) Pictorial representation depicting different stages of rice endosperm from syncytial to cellularization. (b) Based on the stages defined in (a), progression of endosperm development (at 72 and 96 HAF) under control and heat stress in the nine rice accessions. This assessment is based on representative cross-section images of developing seed ( Figure S2). rms, radial microtubule system; HAF, hours after fertilization al., 1996;Li et al., 2018;Olsen, 2004 Figures 1 and 2). Given the HS sensitivity of pollination and/or fertilization (Zinn, Tunc-Ozdemir, & Harper, 2010), HS was initiated 24 HAF so that it did not interfere with either event. Furthermore, we only considered seeds that were marked at the time of fertilization. This approach helped us to (i) avoid overestimation of final yield penalty caused by HS and (ii) precisely track the early seed developmental stages with respect to syncytial and endosperm cellularization and their response to HS. Thus, our analysis does not necessarily reflect whole plant yield-level sensitivity or tolerance but rather provides a comprehensive and explicit view of the HS impact on early seed developmental stages. This approach is useful for future work that aims to identify the genetic basis of variation with respect to heat tolerance in the rice germplasm. Having developmental and temporal context will enhance prioritization of genes that increase resilience to transient heat stress during early seed development.
Heat stress response of the nine rice accessions varied widely.
The two indica accessions, IND-1 shs and IND-2 shs ("shs" in the superscript refers to severely heat sensitive), and one of the tropical japonica accessions (TRJ-3 shs ) showed similar heat stress response with respect to proportion of fully developed seeds (Figures 1 and   2). Also, these accessions showed clear separation of the mature seed traits derived from the two treatments, control (Q-1) and HS (Q-2), depicting an overall sensitivity of these accessions (Figure 4).
On the other hand, aus (AUS-1), two temperate japonicas (TEJ-1 and TEJ-2), and three tropical japonicas (TRJ-1, TRJ-2, and TRJ-4) showed relatively stronger association between the control and HS treatments ( Figure 4) (Figure 2b). The four accessions, AUS-1 mht , TRJ-1 mht , TRJ-2 mht , and TRJ-4 mht ("mht" in the superscript refers to moderately heat tolerant), showed drastic reduction in fully developed seeds by either two or three days of HS (HS2 or HS3, respectively) treatment, whereas TEJ-1 sht and TEJ-2 sht ("sht" in the superscript refers to the strongly heat tolerant) showed mild reductions in fully developed seeds in response to HS2 or HS3 ( Figure 2b). Moreover, 4/6 tolerant accessions showed a significant reduction in the mature seed lengths on exposure to three days of HS (Table 3). These reductions are possible repercussions of the significant decline in the developing seed lengths as early as 72 HAF for all the relatively tolerant accessions ( Figure 5). Contrarily, the sensitive accessions did not show any reduction with respect to mature seed length; however, single-grain weight and seed width were significantly compromised for TRJ-3 shs (Table 3) HAF under control conditions ( Figure 6). The RMS defines the nuclear-cytoplasmic domains, thus facilitating the process of endosperm cellularization (Olsen, 2004;Sabelli & Larkins, 2009). In contrast, under HS conditions, the sensitive accessions did not possess RMS by 72 HAF, suggesting substantial delay in the cellularization process ( Figure 6 and Figure S2). On the other hand, 4/6 relatively tolerant accessions (AUS-1 mht , TRJ-1 mht , TEJ-1 sht , and TEJ-2 sht ) either possessed RMS or commenced the cell wall formation procedure under HS by 96 HAF (Figure 6 and Figure S2). The difference in the rate of cellularization in the different accessions might be correlated with their adaptation strategies to HS. Gene expression analysis further validated the findings from the histochemical analysis. The syncytial-specific marker genes were downregulated, while the cellularization-specific marker genes were upregulated at 96 HAF in the control samples ( Figure 7a). This observation is in accordance with the developing seed cross sections as a majority (8/9; except TRJ-2 mht ) of the tested accessions underwent complete endosperm cellularization by 96 HAF (Figure 6 and Figure S2). Similarly, under HS conditions at 96 HAF, 5/7 syncytial-associated and 6/7 cellularization-associated genes were up-and downregulated, respectively (Figure 7c). These results align well with the histochemical findings where all the tested accessions show delay in endosperm cellularization upon exposure to HS ( Figure 6). However, further investigation of gene expression in context of the rate of cellularization will require testing of marker genes corresponding to RMS or cell wall initiation.
In summary, this study provides detailed insight into the divergence of heat stress response mechanisms of a subset of rice accessions on exposure to the transient HS during early seed development. This work also emphasizes the challenges in developing a simple sampling approach that would account for these differences in developmental transitions and impact of HS in a duration-dependent manner across accessions. Further validation of the sensitive and tolerant accessions through -omic analysis will be helpful in molecular dissection. This work will also be useful in developing future phenotyping strategies for large-scale screening and mining allelic variants for heat resilience rice.

ACK N OWLED G M ENTS
We would like to thank Martha Rowe for help with the histochemical analysis.

CO N FLI C T O F I NTE R E S T
The authors declare no conflicts of interest.