Genotype‐dependent and heat‐induced grain chalkiness in rice correlates with the expression patterns of starch biosynthesis genes

Abstract Starch biosynthesis is a complex process underlying grain chalkiness in rice in a genotype‐dependent manner. Coordinated expression of starch biosynthesis genes is important for producing translucent rice grains, while disruption in this process leads to opaque or chalky grains. To better understand the dynamics of starch biosynthesis genes in grain chalkiness, six rice genotypes showing variable chalk levels were subjected to gene expression analysis during reproductive stages. In the chalky genotypes, peak expression of the large subunit genes of ADP‐glucose pyrophosphorylase (AGPase), encoding the first key step in starch biosynthesis, occurred in the stages before grain filling commenced, creating a gap with the upregulation of starch synthase genes, granule bound starch synthase I (GBSSI) and starch synthase IIA (SSIIA). Whereas, in low‐chalk genotypes, AGPase large subunit genes expressed at later stages, generally following the expression patterns of GBSSI and SSIIA. However, heat treatment altered the expression in a genotype‐dependent manner that was accompanied by transformed grain morphology and increased chalkiness. The suppression of AGPase subunit genes during early grain filling stages was observed in the chalky genotypes or upon heat treatment, which could result in a limited pool of ADP‐Glucose for synthesizing amylose and amylopectin, the major components of the starch. This suboptimal starch biosynthesis process could subsequently lead to inefficient grain filling and air pockets that contribute to chalkiness. In summary, this study suggests a mechanism of grain chalkiness based on the expression patterns of the starch biosynthesis genes in rice.


| INTRODUC TI ON
Physical, granular, and chemical properties are the measures of grain quality, which are dependent on the starch biosynthesis process. Grain chalkiness is a highly undesirable trait in rice, which is genotype-dependent and can also be induced by the high nighttime temperature (HNT) among other factors (Feng et al., 2017;Jagdish et al., 2015;Lanning et al., 2011;Xu et al., 2020). Coordination of the enzymes involved in this process is important to prevent grain chalk that affects the market value, cooking, and eating quality of the rice (Fitzgerald & Resurreccion, 2009;Lisle et al., 2000).
Starch biosynthesis in the developing endosperms of cereal grains is a complex process recently reviewed by Tetlow and Emes (2017). Briefly, starch biosynthesis starts after fertilization, when the endosperm cells multiply, form cell walls, and elongate.
Formation and elongation of cell walls utilizes imported sucrose that is converted to glucose and fructose by cell wall invertase (Wang et al., 2008). These hexose sugars are then transported into endosperm cells, and subsequently, converted to Glucose-1-Phosphate through the action of several enzymes in the cytosol, and ultimately converted to ADP-glucose by ADP-glucose pyrophosphorylase (AGPase), the first key enzyme in the starch biosynthesis pathway. ADP-glucose is transported into amyloplast to serve as the substrate for starch synthases filling the grain with storage starch, an important compound in grain physical quality.
AGPase catalyzed reaction is the rate-limiting step in the process (Stark et al., 1992), and with the reversibility of this reaction, the follow-up expression of granule bound starch synthases (GBSSs) and starch synthases (SS) is important to utilize ADPglucose and prevent the futile cycle of converting ADP-glucose to glucose-1-phosphate.
Granule bound starch synthase I (GBSSI) and starch synthase IIA (SSIIA) are the highly expressed isoforms in the rice endosperm during grain filling stages (Hirose & Terao, 2004;Ohdan et al., 2005;Umemoto & Terashima, 2002;Xing et al., 2016). Mutations in starch synthase genes reportedly affect grain chalkiness by altering the granule morphology from compound polyhedral type to simple spherical type (Kusano et al., 2012;Toyosawa et al., 2016). The simple, spherical granules constitute the chalk portion as they pack loosely and include airspaces (Kaneko et al., 2016;Kim et al., 2004;Lu et al., 2015;Mitsui et al., 2016). Gene expression profiling of chalky and translucent grains in Japanese rice showed genotypic and heatinduced changes in the starch genes. Specifically, starch synthesis genes were found to be upregulated in a near isogenic line showing high chalkiness in comparison to its normal parent .
However, transcriptomics of Nipponbare caryopses ripened in heat or normal temperature showed suppression of starch synthesis genes in heat, even though, heat treatment-induced chalkiness in the grains (Yamakawa et al., 2007). Thus, starch synthesis genes play a major role in determining grain quality; however, their coordination with one another and expression patterns related to chalky or translucent grains has not been fully understood.
Other mechanisms that control grain chalkiness are related to starch accumulation and degradation processes. For example, disruption of the amyloplast's outer envelope membrane during seed maturation leads to the abundance of simple and spherical granules (Toyosawa et al., 2016), and early degradation of starch through amylase activity contributes to grain chalkiness. Micropores on the surfaces of rough amyloplast in the chalky grains indicate starch degradation by amylase activities (Lin et al., 2016). Finally, protein bodies in the endosperm are also implicated in chalkiness. Several studies have shown that chalky rice contains abnormal protein bodies in the endosperm that are large in size and accommodate more air spaces (Fukuda et al., 2011;Nagamine et al., 2011;Ren et al., 2014).
In this study, rice genotypes consisting of well-known chalky varieties and low-chalk cultivars were subjected to gene expression analysis as well as the analysis of grain physical characteristics, starch components, and the granule morphology. The expression patterns of AGPL1, 2, 4, GBSSI, and SSIIA was found to be genotype dependent and heat sensitive, which highlights the importance of coordinated starch biosynthesis during the critical stages of grain filling to produce properly filled, translucent (non-chalky) rice grains.
The disruption in the expression pattern of these starch biosynthesis genes by heat appears to be a part of the mechanism associated with the environment-induced chalkiness.

| Plant materials
Six genotypes, ZHE 733, Nagina 22, Nipponbare, Taggart, Diamond, and LaGrue, representing indica, aus, or japonica subspecies were used in this study. These genotypes included three cultivars (Taggart, Diamond, and LaGrue) developed at Arkansas Rice Research Center. Three replications of each genotype were planted in July 2019 in the greenhouse. When plants were at R0 or R1 stage (Moldenhauer et al., 2018), they were transferred to growth chambers set at 30°C day/22°C night (normal) or at 30°C day/28°C night (HNT) with nighttime starting at 8 p.m. and ending at 6 a.m. Relative humidity and lighting conditions were uniform for the two set-ups. The rice plant culms entering the reproductive stage were tagged and used as the source of samples for different stages, namely before panicle emergence (BP, also called R2 according to Moldenhauer et al., 2018), early flowering/after panicle emergence (AP), 5 days after flowering (DAF), 10, 15, and 20 DAF.
For the granular, physical, and chemical properties, grains from the second panicle were collected at 25 DAF and dried at room temperature for 2 weeks to a moisture content of about ~12%. For gene expression analysis, spikelets from three biological replicates for each genotype/treatment were collected and immediately frozen in liquid nitrogen and stored at −80°C.

| Grain physical property
Grains collected at 25 DAF were dried under room temperature for 2 weeks after harvesting, prior to the observation for chalkiness.
Chalkiness was measured using WinSEEDLE TM with 150 grains for each genotype. The percentage of chalky grains and the average chalk size per grain were taken as measurements of chalkiness.

| Gene expression from databases
Heatmaps were generated to select the appropriate genes of the starch biosynthesis pathway. The Rice Expression Profile Database (Sato et al., 2011) was used under the category datasets and gene expression profile at different ripening stages (7,10,14,21,28,and 42 DAF) relevant to the stages selected for gene expression analysis by quantitative PCR.

| Transcript levels
Total RNA was isolated using Trizol (Invitrogen Inc.) and quantified using Nano-drop 2000 (Thermo-Fisher Inc). Two micrograms of total RNA were treated with RQ1-RNAse free DNase (Thermofisher Inc.), and one microgram of the DNase-treated RNA was used for cDNA synthesis using PrimeScript RT reagent kit (Takara Bio). The expression analysis was performed using TB green Premix Ex Taq II (Takara Bio) on Bio-Rad CFX 96 C1000 with following conditions: 95°C for 30 s. and 40 cycles of 95°C for 5 s +60°C for 30 s. The product specificity was verified by the melt curve analysis. The Ct values of genesof-interest were normalized against 7Ubiquitin fused protein (7UBIQ) as the reference gene. Primers used in the study are given in Table S1.

| Starch granule morphology
Grains were split in two using a microtome for the cross-section per-

| Component analysis
Grains harvested at 25 DAF were dried at room temperature for 2 weeks (~12% moisture content), ground to fine powder in liquid N 2 for the determination of soluble protein content or in a cyclone milling machine for amylose and amylopectin content. The amylose and amylopectin content of grains were determined using the Megazyme amylose/amylopectin assay (K-AMYL) following the manufacturer's method.
Quantification of the soluble fractions of protein was done using 50 mg of grain powder in 1 ml of TE buffer pH 8.0 using a Bradford assay against a standard curve of BSA (0, 2.5, 5.0, 7.5, and 10 µg/ml). Absorbance was read at 595 nm in a Bio-Rad SmartSpec 3000 spectrophotometer.

| Data analysis
The experiment for soluble protein, amylose, and amylopectin contents were conducted in a completely randomized design with three independent replications under six cultivars having a total sample size of 18.
Data were subjected to arcsine transformation and one-way ANOVA.
To determine the significant differences in the amylose and amylopectin content and protein concentration, Tukey's multiple comparison test was used to compare the genotypes under normal condition, and Student's t-test for pairwise comparison in the normal and heat conditions. All statistical analyses were performed in SAS statistical software (version 9.4, SAS Institute Inc.) and results are presented in Tables S2-S6.

| Expression patterns of starch biosynthesis genes
The heatmap based on RiceXPro database showed genes encoding the subunits of rice amyloplastic AGPase are expressed differentially. AGPS1 is somewhat consistent, while AGPL1 and AGPL4, are upregulated early at 7-14 DAF followed by gradual decline in the subsequent stages with AGPL4 expressed at relatively lower levels ( Figure 1). AGPL2 that encodes unique subunit of the cytosolic AGPase is expressed throughout the grain filling stages but shows upregulation in the mid grain filling stages (21-28 DAF) ( Figure 1). The resulting cytosolic ADP-glucose passes through the adenylate transporter, BRITTLE1 (BT1), to enter the amyloplast for starch biosynthesis (Cakir et al., 2016). AGPS2b subunit of the cytosolic AGPase, on the other hand, was expressed at much lower levels throughout the grain filling stages (Figure 1). AGPS2b forms heterotetramer with AGPL2 to form the cytosolic AGPase. For gene expression analysis of the cytosolic AGPase, AGPL2 subunit gene was selected. Similarly, heatmap of GBSS showed that GBSSII is expressed at much lower levels and downregulated during advancing stages of grain filling (14 DAF onwards), while GBSSI is consistently expressed. Next, SSI is downregulated between 7 and 21 DAF and SSIIB is downregulated throughout (7-42 DAF), while SSIIA is consistently expressed ( Figure 1). Therefore, AGPL1, AGPL2, AGPL4, GBSSI, SSIIA, and BT1 were selected for gene expression analysis.

| Grain chalkiness in different genotypes
The six genotypes used in this study were found to have different levels of chalkiness based on which they were classified as high or low chalky. High chalky lines, ZHE 733, Nipponbare, and Nagina 22, contain large opaque areas, while the three low chalky cultivars, Taggart, Diamond, and LaGrue contain no chalk or small chalky areas ( Figure 2a). Furthermore, in high-chalky lines, chalk was observed in all grains, with the majority (average of 82%) showing large chalk (>20% of grain size), in addition to small (<10% of grain size) and medium (11%-20% of grain size) chalk. On the other hand, in low-chalky lines, small chalk was found in the majority of the grains (average of 84%) with a small percentage (2%-4%) showing no chalk (Figure 2b).
Among these, LaGrue was found to contain more chalk (medium sizes) than Taggart or Diamond.

| Expression patterns of AGPL2, AGPL4, GBSSI, and SSIIA
In high chalky lines, AGPase subunit genes are expressed at higher levels during early reproductive stages (BP, AP, or 5 DAF) followed by gradual decline in the subsequent stages (5-20 DAF) when and SSIIA were co-expressed and markedly upregulated during early grain filling stages (5-20 DAF), while AGLP4 was expressed early and remained consistent through early grain filling stages (Figure 3f).

| Component analysis
There is a significant difference in the amylose content among high and low-chalky cultivars (ANOVA, α = 0.05, df = 5, and p < 0.001), where ZHE 733 was the highest (Tukey's test, α = 0.05, df = 12 and p < 0.001). Within the high chalky group, the indica rice ZHE 733, that shows simple and spherical granules (Figure 2c), contained significantly higher amylose fraction. Low chalky cultivars, on the other hand, were found to contain significantly higher amylopectin content (Table 1). Regardless of the chalkiness, amylopectin is higher than amylose in all lines. Variation in the soluble protein content of the endosperm was also observed in the present study (ANOVA, α = 0.05, df = 5, and p = 0.0079). Nipponbare and Nagina 22, the two chalky lines, were found to have highest soluble proteins compared to other genotypes (Tukey's test, α = 0.05, df = 12, p < 0.001).
The high chalky line, ZHE 733, on the other hand, showed a similar level of the soluble protein content as the three low chalky cultivars (Table 1).   Tables S2-S4.

Analyses of variances (ANOVA) summaries are in
Data were transformed using arcsine transformation. **Significant for cultivar/line as source of variation at p = 0.01.
HNT-LaGrue, on the other hand, maintained compound polyhedral granules but showed abundant shreds and micropores on the granule surfaces (Figure 5c).
Finally, amylose and soluble protein contents were determined under normal and HNT conditions. In Diamond, no significant difference in the amylose content (t-test, α = 0.05, df = 5 and p = 0.0757) was observed between the two conditions, but in LaGrue an increase in the amylose content (t-test, α = 0.05, df = 5 and p = 0.0008) was observed in the HNT grains ( Figure 5d). However, the soluble protein content in Diamond (t-test, α = 0.05, df = 5 and p = 0.0005) and LaGrue (t-test, α = 0.05, df = 5 and p = 0.0004) was significantly higher under HNT compared to normal condition (Figure 5e).

| Dynamics of starch biosynthesis genes
ADP-glucose pyrophosphorylase, GBSS, and SS participate in the key steps of starch biosynthesis process. To analyze the cytosolic AGPase, that contributes to the bulk of AGPase activity in rice (Sikka et al., 2001), AGLP2, the large subunit gene unique to cytosolic AGPase was selected. We also analyzed the amyloplastic AGPase through expression analysis of its large subunit genes, AGPL1 and/ or AGPL4, as amyloplastic AGPase is considered critical for the normal levels of storage starch in the grains (Kawagoe et al., 2005;Lee et al., 2007;Sun et al., 2015). Of the GBSS and SS isoforms, GBSS1 and SSIIA were selected based on their steady expression through grain filling stages (Figure 1; Hirose & Terao, 2004), and their roles in controlling amylose and amylopectin content, respectively, in rice (Dobo et al., 2010;Liu et al., 2014;Miura et al., 2018;Nakamura et al., 2005). The interpretation of starch biosynthesis process in this study was based solely on the gene expression analysis. Although, posttranscriptional and posttranslational controls of starch biosynthesis cannot be ignored (Smith et al., 2004;Tetlow et al., 2004;Wang et al., 1995), direct correlation of mRNA abundance and enzyme activity or protein abundance for AGPase, GBSS, and SS in rice and maize (Devi et al., 2010;Ponnala et al., 2014)

| Correlation of granule morphology with gene expression patterns
High chalky lines showed simple granules of small sizes or compound granules of variable sizes (Figure 2c), suggesting inefficient or irregular starch biosynthesis leading to restricted enlargement of the granules (Kawagoe et al., 2005;Lisle et al., 2000). This hypothesis is supported by the expression analysis that showed temporal gap be- In the coordinated expression pattern, AGPL1 and 2, the regulatory subunits of the tetrameric AGPase, and monomeric GBSS1 and SSIIA are upregulated in quick succession through early grain filling stages. As a result, the conversion of glucose-1-phosphate (1P) to ADP-glucose (ADP) occurs in a timely manner to efficiently synthesize amylose and amylopectin. As a result, large polyhedral granules are produced that pack tightly in the grains. In the uncoordinated expression pattern, early upregulation and/or subsequent suppression of AGPase subunit genes creates a temporal gap between the activation of AGPase and that of GBSSI and SSIIA, leading to non-utilization of ADP and its reversal to 1P. As a result, suboptimal starch biosynthesis occurs leading to the formation of smaller granules of spherical (1) or polyhedral shapes (2, 3) accommodating airspaces and large protein bodies (3)  develop pits and micropores that contribute to chalkiness (Kaneko et al., 2016;Lin et al., 2016;Mitsui et al., 2016;Tsutsui et al., 2013).
These micropores are considered the evidence of the enzymatic degradation of starch Zakaria et al., 2002), a hypothesis supported by the genetic analysis that showed improvement in rice grain quality upon suppression of the α-amylase genes (Hakata et al., 2012).

| Starch components and their correlation with grain chalkiness
High amylose content has been correlated with heterogeneous granule formation and grain chalkiness (Man et al., 2014;Zheng et al., 2012). Accordingly, in this study, higher amylose content was observed in high chalky lines (Table 1). Shorter chains (α1-4) and reduced branching (α1-6) in the amylopectin could lead to smaller granules. This analogy is reinforced by observations that high amylose rice contains shorter α-glucan chain lengths, and grains containing short chain amylopectin show higher chalkiness (Park et al., 2007;Patindol & Wang, 2003), possibly from loose packing of small polyhedral granules . On the other hand, higher amylopectin was found in the low chalky lines, which corroborates with other studies that found the association of higher amylopectin with lower chalkiness (Inukai, 2017;Lin et al., 2016). The coordinated expression of AGPase subunit genes and GBSSI and SSIIA in the lowchalky cultivars (Figure 3d- (Abe et al., 2014;Sun et al., 2017). Finally, although genetic relation of protein content with the quality traits in rice has been described (Lin et al., 2016;Liu et al., 2011;Zheng et al., 2012), its correlation with grain chalkiness is not very clear. In this study, variation in the soluble protein content was observed (Table 1), but no clear correlation with grain chalkiness was observed.

| Mechanisms of heat-induced chalkiness
High nighttime temperature induces chalk formation along with the changes in the expression of starch biosynthesis genes (Dhatt et al., 2019;Mitsui et al., 2016;Nevame et al., 2018). In the present study,  (Iwasawa et al., 2009;Tsutsui et al., 2013;Yamakawa et al., 2007;Zakaria et al., 2002). Finally, heat-induced chalk formation in rice is also accompanied with the increase in all classes of storage protein fractions, including albumins and globulins in the early phases of grain development (Lin et al., 2010). However, association of protein content and chalkiness is more complex and likely involves formation of protein bodies with reduced storage proteins accommodating more air pockets (Wada et al., 2018).

| CON CLUS IONS
Efficient synthesis of starch through the coordinated expression of AGPase, GBSS, and SS is arguably the most important mechanism controlling granule morphology. In the coordinated expression pattern, AGPase is upregulated early in the reproductive phase, during the grain filling stages, to produce abundant pool of ADP-glucose, which is quickly utilized by GBSS and SS to produce amylose and amylopectin in the amyloplast. This streamlined process leads to uniform polyhedral granules that pack tightly and produce non-chalky grains ( Figure 5f). However, when AGPase is upregulated before endosperm development or suppressed by heat treatment, only a limited pool of ADP-glucose is presumably available for starch biosynthesis. This uncoordinated process could lead to inefficient starch biosynthesis, producing smaller granules of heterogeneous shapes (Figure 5f).
These simple spherical or heterogeneous granules pack more loosely and accommodate air spaces observed as chalk in the mature grains.

ACK N OWLED G M ENTS
This work was supported by NSF-EPSCoR RII Track-2 program (grant# 1826836). We thank Dr. Flavia Botelho for the guidance on spikelet sampling, Dr. Betty Martin for the scanning electron microscopy, Dr. Bhuvan Pathak for help with protein assays, and Dr.
Anuj Kumar for the help with the grain chalk analysis.

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

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly available in the Box folder hosted by University of Arkansas at https://uark.