Physiological and morphological factors affecting leaf sheath reinforcement and their contribution to lodging resistance in rice

ABSTRACT In rice (Oryza sativa L.), leaf sheaths enhance stem strength and lodging resistance. However, little is known about the factors that affect varietal differences and changes over time regarding leaf sheath reinforcement. In the present study, the morphological and physiological traits that are presumed to be related to leaf sheath reinforcement were examined in ‘Chugoku 117’ and ‘Koshihikari’, rice varieties with high and low degrees of leaf sheath reinforcement, respectively. The results showed that Chugoku 117 had thicker leaf sheaths and delayed leaf sheath senescence than Koshihikari, indicating that there were clear varietal differences in these traits. The bending moment at breaking with leaf sheath was correlated with senescence-associated traits, especially strongly correlated with the number of living leaf sheaths in both varieties. Among the components examined in the present study, only starch content was significantly positively correlated with both varieties. On the other hand, the starch in the leaf sheath disappeared in the latter stage of ripening due to translocation to sinks, suggesting that it contributes to stem strength only up to the early stages of ripening. The findings suggest that rice stem strength can be increased by thickening the leaf sheaths and delaying leaf sheath senescence. Thus, in addition to the physical properties of the culm, such as shortening (semi-dwarfing) and strengthening, the findings of the present study provide a new breeding strategy for improving breaking-type lodging resistance in rice. Graphical abstract


Introduction
In the 1960s, the introduction of the semi-dwarf gene semidwarf 1 (sd1) resulted in the development of shortculmed plants in rice (Oryza sativa L.). These plants had a lower center of gravity and significantly increased lodging resistance, which meant that they could be grown with large amounts of nitrogen fertilizer without lodging, resulting in higher yields. This sequence of events is referred as the 'Green Revolution' in rice farming (Hedden, 2003;Khush, 2001). However, in recent years, it has been shown that the yield potential of semi-dwarf varieties using sd1 is reduced by their limited biomass production capacity (Okuno et al., 2014). Furthermore, typhoons are becoming more powerful as a result of climate change (Mei & Xie, 2016), causing lodging issues even in semi-dwarf varieties (Ookawa et al., 2016). Therefore, further improvements in lodging resistance need to focus on increasing the strength of the stem, rather than relying only on semi-dwarfism.
In rice culms, culm thickness and composition have been demonstrated to be associated with breaking strength (Ookawa & Ishihara, 1992, 1993. For example, genes such as ABERRANT PANICLE ORGANIZATION 1 (APO1) and FINE CULM 1 (FC1) increase the bending moment at breaking by thickening the culm (Ookawa et al., 2010;Yano et al., 2015). It has been reported that the increase in culm strength in rice plants is associated with high levels of cellulose, hemicellulose (Mulsanti et al., 2018;Samadi et al., 2019), and lignin (Ke et al., 2019;Liu et al., 2018;Ookawa & Ishihara, 1993), which are cell wall constituents. Furthermore, plants exhibiting high levels of starch accumulation in the culm also showed an increase in culm strength (Ishimaru et al., 2008;Zhao et al., 2019). The delayed senescence of rice plants with the prl5 locus leads to starch accumulation, and consequently, an increase in culm strength Kashiwagi et al., 2006).
On the other hand, leaf sheaths covering the culm have been shown to be important for lodging resistance; most studies have focused on the culm, and the contribution of leaf sheaths on lodging resistance has not been clarified. A previous study that examined the mechanical properties of bending leaf sheaths of the rattan in Calamoideae subfamily showed that thick and persistent leaf sheaths play an important role in the mechanical properties of the stem, and that senescence and shedding of leaf sheaths drastically reduced the rigidity of the old stem base (Isnard & Rowe, 2008). However, the mechanical properties of such climbing species may differ from those of cereals due to differences in the anatomy and mechanical structure of the leaf sheath (Wu & Ma, 2020).
Previous studies on the grass family Poaceae (formerly known as Gramineae) showed that flexural rigidity did not affect stem lodging in oats (Avena sativa L.); however, leaf sheaths contributed to stem flexural rigidity (Niklas, 1990). In Switch Cane (Arundinaria tecta [Walter] Muhl.), leaf sheaths accounted for an average of 33% of total stem flexural rigidity (Niklas, 1998). In the Giant reed (Arundo donax L.), the stem responded to dynamic deflection with damped harmonic bending oscillations, whereas the leaf sheaths surrounding the nodes did not affect damping (Speck & Spatz, 2004). A study on the interspecific differences in the contribution of leaf sheaths to lodging resistance reported that the leaf sheaths in oat contribute relatively more to mechanical reinforcement than do the leaf sheaths in wheat (Wu & Ma, 2020), and other studies on the intervarietal differences among rice varieties reported large differences in leaf sheath reinforcement (Furuhata & Arima, 2007;Miyasaka & Takaya, 1982;Ookawa & Ishihara, 1992). However, even though rice is a major cereal crop in the grass family, few studies have focused on the leaf sheath reinforcement in rice, and the morphological and physiological factors responsible for inter-varietal differences in leaf sheath reinforcement have remained unclear.
In the present study, we focused on the morphological traits of leaf sheath cross-section and physiological traits of leaf sheath senescence and components in two varieties with different stem strength to gain new insights into the factors affecting leaf sheath reinforcement during the ripening period.

Plant materials and cultivation
The rice (Oryza sativa L.) varieties used in this study, Koshihikari and Chugoku 117, were cultivated in 2020. These varieties were selected as plant material in the present study because they have been reported to differ in leaf sheath reinforcement (Ookawa & Ishihara, 1992). The main stems with leaf sheaths were sampled on the day of heading, and then at 2, 4 and 6 weeks after heading. The date of heading was defined as the date when approximately 50% of the main culm of the entire canopy was heading for each variety (August 8 for Koshihikari and August 13 for Chugoku 117). The main stems with the fifth internode from the top as the lowest elongation internode (≥4.0 cm) were used for measurements. For RNA-seq analysis, leaf sheaths were sampled every 5 days before heading (except at heading) because mRNAs related to senescence were expected to dynamically change expression levels prior to the senescence phenotype (Bresson et al., 2018). Field experiments were conducted in the paddy fields of the Field Museum Honmachi, Field Science Center, Faculty of Agriculture, Tokyo University of Agriculture and Technology. The size of one plot was 1.5 m × 1.5 m. Plots were designed so that Koshihikari and Chugoku 117 were arranged in a checkerboard pattern. Plants were collected from three plots per test and the arithmetic mean of the three plots was used as the trait value. Seeds were sown in nursery boxes on 7 May 2020. Seedlings at approximately the four-leaf stage were transplanted into the paddy field on 20 May 2020, with one seedling per hill. The planting density was 22.2 hills m −2 with a spacing of 15 cm × 30 cm. The residues of the previous year's rice were tilled in winter as compost. N, P 2 O 5 and K 2 O were applied at 50 kg ha −1 , 60 kg ha −1 and 60 kg ha −1 , respectively. During the experiment, weed control and pest management were conducted when necessary, and fields were cultivated under irrigation.
Datasets for temperature, precipitation, and sunlight duration during cultivation related to meteorological conditions were obtained from the Automated Meteorological Data Acquisition System (AMeDAS) at Fuchu Observatory (Supplemental Table S1) (Japan Meteorological Agency, 2022).

Evaluation of breaking strength
The bending moment at breaking and section modulus were measured using the method described by Ookawa and Ishihara (1992). A mean of 12 stems (at least 6) were sampled from each plot to measure breaking strength, and the leaf sheaths of about half of the stems were removed. Regardless of the presence or absence of leaf sheaths, all mechanical measurements were performed between the fifth internode. Load was measured by three-point bending test using a Tensilon universal testing machine (RTG-1210; A&D, Tokyo, Japan), placed on a pedestal with a fulcrum distance of 4.0 cm. The bending moment of the internode was calculated using the following equation (1): where M is the bending moment, P is the load, and L is the distance between fulcrums. The section modulus of the culm was calculated by measuring each parameter using digital calipers and the following equation (2) considering the culm as a hollow ellipse: where Z is the section modulus, a 1 is the outer diameter of the minor axis in an oval cross-section, b 1 is the outer diameter of the major axis in an oval cross-section, a 2 is the inner diameter of the minor axis in an oval crosssection, and b 2 is the inner diameter of the major axis in an oval cross-section. The degree of leaf sheath reinforcement was calculated using the following equation (3): where α is the degree of leaf sheath reinforcement, M with is the bending moment at breaking with leaf sheaths, and M without is the bending moment at breaking without leaf sheaths.

Morphological observations and microscopic analysis
The fifth internode with leaf sheaths at each period was fixed in FAA solution (formaldehyde: acetic acid: 50% ethanol at 5:5:90). Thin transverse sections that were prepared from leaf-sheathed internodes with a razor were stained in 0.05% (w/v) toluidine blue solution and observed under a stereo microscope (SZX12; Olympus, Tokyo, Japan). Four individuals as many as possible per replicate were photographed using a digital camera (DP21; Olympus) fitted to the microscope. Image analysis was performed using the Fiji distribution (Schindelin et al., 2012) of ImageJ (Ver. 2.1.0/1.53c). Each trait value shown in the figure was measured (Supplemental Figure  S1). Representative photomicrographs of razor-thin transverse sections were also taken using a scanning electron microscope (TM3000; Hitachi High-Technologies, Tokyo, Japan).

Determination of the number of living leaf sheaths
The number of living leaf sheaths was determined visually by a mean of 12 stems per plot (at least 6), referring to the leaf color scale (Shin et al., 2020), with green or green-yellow leaf sheath as 1, yellow one as 0.5, and brown or absent one as 0.

Determination of chlorophyll contents
Ten 5 mm diameter disks per leaf sheath were prepared from the center of the fifth leaf sheath of three individuals per plot and extracted with acetone until the green color was completely removed. Then, 2.5 mM phosphate buffer (pH 7.8) was added to bring the final acetone concentration to 80% and the absorbance was measured using a microplate reader (Viento XS; DS Pharma Biomedical, Osaka, Japan) and Gen 5 1.10 software (Biotek, Vermont, USA). Based on the absorbance values, the chlorophyll contents were calculated using the method of Porra et al. (1989). For the chlorophyll a/ b ratio, samples with chlorophyll a or b content less than 0.1 were not used as data due to the highly susceptible to measurement error.

Measurement of membrane ion leakage
The measurement of membrane ion leakage was carried out in leaf sheath cells, with some modifications to a previous study (Fan et al., 1997). For each plot, three leaf sheaths on the fifth internode were collected and washed briefly with deionized water. Ten 5 mm diameter disks per leaf sheath were prepared from the center of the leaf sheath; 30 leaf disks from three individuals were placed in tubes containing 25 mL of deionized water, the tubes were stirred gently and the conductivity (C 0 ) was measured immediately using a conductivity meter (D-74; Horiba, Kyoto, Japan) and a conductivity electrode (9383-10D; Horiba). The tube was shaken in a shaker for 3 h at room temperature and the conductivity (C 3 ) was measured. After boiling the samples, the conductivity (C T ) was measured. The percentage of membrane ion leakage (I) was then calculated using the following equation (4):

Total RNA extraction
Three leaf sheaths were collected per plot. The collected samples were placed in tubes and ground in liquid nitrogen. Then, 1 mL of RNAiso Plus (Takara Bio, Shiga, Japan) was added to the ground samples and the samples were homogenized. After incubating at room temperature for 5 min, the sample was then centrifuged (~17,000 ×g for 5 min at 4°C) and the supernatant was transferred to a new tube. Then, 200 µL of chloroform was added to the sample and shaken until a milky white. After standing at room temperature for 5 min, the sample was centrifuged (~17,000 ×g for 15 min at 4°C) and only the top layer of the solution separated into three layers was transferred to a new tube. Then, 750 µL of isopropanol was added to the sample and mixed. After 10 min at room temperature, the sample was centrifuged (~17,000 ×g for 10 min at 4°C) to precipitate the RNA. After decanting the supernatant, 1 mL of 75% ethanol was added and stirred to wash the precipitate. The sample was centrifuged (~10,000 ×g for 5 min at 4°C) and the supernatant was removed. Finally, after drying, it was dissolved in DEPC water.

Library preparation and sequencing
Library preparation was performed with 250 ng of total RNA per sample using Lasy-seq method (Ver. 1.1) (Kamitani et al., 2019; https://sites.google.com/view/ lasy-seq/lasy-seq-protocol). Sequencing was performed using one lane of the HiSeq X Ten Sequencing System (Illumina, California, USA). Fastq files were deposited to DRA (DDBJ Sequence Read Archive) as DRA013775.

Determination of cell wall components and starch
The determination of cellulose, hemicellulose, and holocellulose in leaf sheaths was performed as described in previous studies (Samadi et al., 2019;Schädel et al., 2010), with some modifications. The samples were obtained from the lower 6 cm of the fifth and sixth leaf sheaths dried at 80°C for 72 h. A mean of five dried samples per test were bulked, ground using a shaker mill, extracted twice in 80% ethanol at 80°C for 20 min, and then extracted with 50% ethanol at 80°C for 20 min.
To remove starch from the pellet after ethanol extraction, a 100-µL solution of α-amylase from Bacillus licheniformis (A4551; Sigma-Aldrich, Missouri, USA) dissolved in 200 mM phosphate buffer (~1.2 × 10 4 U mL −1 ) was added and the sample was incubated at 85°C for 30 min. After centrifugation (21,500 ×g for 15 min), the supernatant was discarded and the samples were washed with deionized water and dried completely at 80°C. Then, 1.6 mL of aqueous sodium chlorite solution (6.7 mg mL −1 ) and 0.2 mL of acetic acid were added to the starch-free residue (to remove lignin), and incubated at 80°C for 1 h with vortexing performed every 20 min. Samples were centrifuged (~24,000 ×g for 12 min) and the supernatants were discarded. This operation was repeated until the contents turned white. The pellet was washed twice with deionized water and once with acetone and dried to completion at 80°C. The pellets were weighed after cooling in a desiccator and the constant weight was taken as the holocellulose content. Then, 600 µL of acidic detergent (1 N sulfuric acid and 55 mmol hexadecyltrimethylammonium bromide dissolved in deionized water) was added to the pellets and incubated at 95°C for 1 h with vortexing performed every 20 min. After centrifugation (~24,000 ×g for 12 min), the supernatant was discarded. The pellets were washed twice with deionized water and once with acetone and dried to completion at 80°C. The pellets were weighed after cooling in a desiccator and the constant weight was taken as the cellulose content. The difference between the holocellulose content and the cellulose content was utilized to calculate the hemicellulose content.
The determination of starch in leaf sheaths was performed as described in previous studies (Hendriks et al., 2003;Thormählen et al., 2013), with several modifications. Using the same procedure as described above for the determination of cell wall components, 10 mg of ground dried sample was extracted with ethanol. After extraction, 400 µL of 0.1 M sodium hydroxide was added to the dried precipitate and mixed, followed by incubation at 95°C for 30 min and cooling at room temperature. Then, 80 µL of a mixture (0.5 M hydrochloric acid and 0.1 M sodium acetate buffer [pH 4.9]) was added. If the starch concentration was too high, the following reactions could not be performed sufficiently, so the samples were diluted with 50 mM sodium acetate buffer, if necessary. Then, 80 µL of suspension from each specimen was placed in the well of a microplate (in triplicate). Then, 120 µL of a starch-degradation mixture (~5.4 U mL −1 αamylase from pig pancreas [10102814001; Roche Diagnostics, Mannheim, Germany] and ~2.8 U mL −1 amyloglucosidase from Aspergillus niger [10102857001; Roche Diagnostics] dissolved in 50 mM sodium acetate buffer) was added to each well of the microplate and incubated at 37°C for 16 h. A total of 50 µL of the digested supernatant and the glucose standard solution (0-0.8 mM) dissolved in sodium acetate buffer were transferred to another microplate and 160 µL of a mixture (0.1 M HEPES buffer [0.1 M HEPES and 3 mM MgCl 2 , pH 7.0 adjusted with KOH], 3 mM ATP, 1.4 mM NADP and ~1.7 U mL −1 G6P-DH [10127655001; Roche Diagnostics]) was added and stirred. The absorbance of the specimen was measured at 340 nm using a microplate reader (Viento XS; DS Pharma Biomedical) and analyzed using Gen 5 1.10 software (Biotek). Then, 2 µL of hexokinase (11426362001; Roche Diagnostics) solution dissolved in 0.1 M HEPES buffer (450 U mL −1 ) was added and the absorbance was measured again after the value had stabilized. Absorbance measurements were taken in triplicate for each sample and the mean difference between the readings obtained before and after the addition of hexokinase was used as the value for the sample. From the calibration curve, the glucose concentration was calculated and the glucose content was determined. The glucose content multiplied by 0.9 was used as the starch content.
The determination of lignin was performed by the method of Suzuki et al. (2009) using ground samples as well as the determination of cell wall components described above.

Statistical analysis
Statistical analysis was performed using the R software package (Ver. 4.0.5 & 4.2.1) (R Core Team, 2021-2022). The P-values among the multiple groups were calculated using the Tukey-Kramer test or Dunnett's test, and between the two groups using Welch's t-test. Dunnett's test was performed using the package 'multcomp' (Ver. 1.4.20) (Hothorn et al., 2008).

Contribution of leaf sheaths to culm strength
We investigated the effect of leaf sheaths on lodging resistance-associated traits in rice varieties, Koshihikari and Chugoku 117. Figure 1a shows the bending moment at breaking for each of the fifth internodes with and without leaf sheath in both varieties. Compared to Koshihikari, the fifth internodes of Chugoku 117 with leaf sheaths showed the highest bending moment at breaking of all growth stages. In both varieties, the bending moment at breaking was higher in the fifth internodes with leaf sheaths than in those without leaf sheaths, but these differences decreased over time and almost disappeared at 6 weeks after heading. No significant difference was observed between the fifth internodes of Koshihikari with and without leaf sheaths at any period, but there were significant differences between the fifth internodes of Chugoku 117 with and without leaf sheaths until 4 weeks after heading. Figure 1b shows the degree of leaf sheath reinforcement (DLSR) for bending moment at breaking. Although there were no significant differences, Chugoku 117 showed higher DLSR values compared to Koshihikari at all stages, but the DLSR of both varieties decreased linearly, and the values were marginal at 6 weeks after heading.

Morphological characteristics of leaf sheaths and their contribution to stem reinforcement
To determine the structural factors responsible for the inter-varietal differences observed in leaf sheath reinforcement of a stem, a microscopic examination of transverse sections of the leaf sheaths was performed. Figure 2a-d shows a representative micrograph of the fifth internode and the leaf sheath wrapped around it at heading taken using a stereomicroscope. The leaf sheaths of Chugoku 117 appear to be thicker than those of Koshihikari. Further comparison of the magnified images showed that the width of the web-like parenchyma between aerenchyma in the middle cell layers (the web refers to the part of the H-section beam that connects the flanges; see Supplemental Figure S1) was thicker in Chugoku 117 than in Koshihikari. Figure 2e-h shows representative micrographs of the fifth internode and the fifth leaf sheath wrapped around it with a scanning electron microscope. The leaf sheaths of Chugoku 117 appeared thicker than those of Koshihikari because of the longer web-like parenchyma between aerenchyma. The width of the web-like parenchyma also appeared to be thicker. Starch grains were observed in both varieties and appeared to be particularly dense around the vascular bundles. When both varieties were compared, the abaxial cell layers of the leaf sheath in Chugoku 117 appeared to be more densely packed with starch than Koshihikari. Reticulated membranes across aerenchyma were also observed in both varieties (Supplemental Figure S2).
Furthermore, image analysis of the leaf sheath morphology was performed in cross-section. Table 1 shows the values of each trait relative to the structure of the leaf sheath. The fifth (P = 0.006) and sixth leaf sheaths (P = 0.018) of Chugoku 117 were significantly thicker than the leaf sheaths of Koshihikari. No significant differences in the thickness of the abaxial cell layers (fifth leaf sheath, P = 0.96; sixth leaf sheath, P = 0.67) and adaxial cell layers (fifth leaf sheath, P = 0.78; sixth leaf sheath, P = 0.47) were observed between the two varieties. When the abaxial and adaxial cell layers were compared, the abaxial cell layers of the fifth and sixth leaf sheaths were significantly thicker in both varieties (Table 1; Supplemental Table S2). Except for the abaxial cell layers of Koshihikari, the fifth leaf sheath had a significantly thicker cell layers than the sixth leaf sheath. (Table 1; Supplemental Table S3). There were no significant differences in the thickness of the weblike parenchyma between Koshihikari and Chugoku 117 (fifth leaf sheath, P = 0.052; sixth leaf sheath, P = 0.074). The width of aerenchyma (i.e. between the web-like parenchyma) was not significantly different in the sixth leaf sheath (P = 0.13) between the two varieties, but the fifth leaf sheath of Chugoku 117 was significantly wider than that of Koshihikari (P = 0.001). The outer diameter of the minor axis and section modulus of the culm (the fifth elongation internode counted from the top) were significantly larger in Chugoku 117 than in Koshihikari (Supplemental Table S4).
These results indicated that, compared to Koshihikari, the leaf sheathes themselves were thicker in Chugoku 117 and leaf sheathes were also located more outward from the center of the culm due to this variety having a thicker culm.

Senescence of the leaf sheath
To clarify the changes in leaf sheath senescence during the ripening period, the degree of senescence of the fifth leaf sheath was examined in both varieties. Figure 3a shows the number of living leaf sheaths that were covered around the fifth internode. The number of leaf sheaths tended to be higher in Chugoku 117 than in Koshihikari up to 4 weeks after heading, although there were no leaf sheaths left in either variety at 6 weeks after heading. Although there was a significant difference within each variety between the heading day and 2 weeks after heading, the number of living leaf sheaths in Chugoku 117 was 2.0 at heading and 1.5 at 2 weeks after heading, whereas that in Koshihikari was 1.7 at heading and 0.3 at 2 weeks after heading. The findings indicated that the number of living leaf sheaths decreased more rapidly in Koshihikari than in Chugoku 117 at 2 weeks after heading. Figure 3b shows a representative photograph of the fifth leaf sheath from the top. At 2 weeks after heading, the leaf sheaths of Koshihikari began to turn brown, whereas few changes were observed in Chugoku 117. At 4 weeks after heading, the leaf sheaths of Koshihikari had turned completely brown, whereas those in Chugoku 117 were intermediate in color, being between green and brown. Figure 3c shows the total chlorophyll content of the fifth leaf sheath in both varieties. In Chugoku 117, the total chlorophyll content was 39 mg m −2 on the day of heading and 35 mg m −2 at 2 weeks after heading, with no significant difference in chlorophyll content observed between these periods; these findings indicate that chlorophyll was relatively high and maintained.
However, the chlorophyll content declined significantly to 7.3 mg m −2 at 4 weeks after heading, after which they were almost undetectable. Koshihikari maintained relatively high chlorophyll levels of 42 mg m −2 at heading, but these decreased to 11 mg m −2 after 2 weeks, becoming almost undetectable thereafter. A similar trend was   observed for chlorophyll a and chlorophyll b, respectively (Supplemental Figure S3). The chlorophyll a/ b ratio in Chugoku 117 remained relatively high at 2.9 to 2.3 from heading to 4 weeks after heading, but rapidly decreased at 6 weeks after heading. The chlorophyll a/ b ratio in Koshihikari remained relatively high at 3.3 to 2.7 from heading to 2 weeks after heading, but rapidly decreased at 4 weeks after heading (Supplemental Figure S3). Figure 3d shows the percentages of membrane ion leakage of the cells in the fifth leaf sheath. Membrane ion leakage is a well-established indicator of the damage that occurs to the cell membrane under senescence (Bresson et al., 2018). At heading, both Koshihikari and Chugoku 117 showed relatively low values of 16% and 10%, respectively, but at 2 weeks after heading, membrane ion leakage in Koshihikari increased significantly to 45% while Chugoku 117 showed no significant change at 15%. However, at 4 weeks after heading, the difference between the two varieties narrowed, with membrane ion leakage in Koshihikari and Chugoku 117 increasing significantly to 71% and 66%, respectively, compared to 2 weeks after heading. At 6 weeks after heading, membrane ion leakage in Chugoku 117 decreased, but this may be due to the effect of measurement errors caused by cell damage at the beginning of the experiment, which left few ions in the cell (Supplemental Figure S4). We also inferred the degree of senescence in Koshihikari and Chugoku 117 based on the expression levels of chlorophyll degradation-related genes: SGR (Jiang et al., 2007;Park et al., 2007), NYC1, NOL (Kusaba et al., 2007;Sato et al., 2009), RCCR1, and PAO (Tang et al., 2011); senescence-related transcription factors: OsWRKY23 (Jing et al., 2009), OsNAP (Guo & Gan, 2006;Liang et al., 2014), OsNAC2 (Mao et al., 2017), and OsNAC5 (Sperotto et al., 2009); and senescence-associated genes: Osl2, Osl30, Osl85, Osh36, and Osh69 (Lee et al., 2001). Figure 4 shows the log 2 (FPKM +1) of these genes in the fifth leaf sheath; expression levels of NYC1, OsWRKY23, OsNAP, Osl2, Osl30, and Osl85 significantly increased in Koshihikari at 10 days after heading, and in addition to these genes, SGR, NOL, and Osh69 also significantly increased at 15 days. On the other hand, in Chugoku 117, only SGR and OsNAP were significantly increased at 10 days after heading, and RCCR1 and Osh69 were increased in addition to OsNAP at 15 days. PAO, OsNAC2, OsNAC5, and Osl36 showed no significant change with time compared to 5 days before heading into both varieties. The results suggest that senescence of the leaf sheaths of Koshihikari occurs relatively rapidly compared to Chugoku 117.

Cell wall components and starch in leaf sheaths
We quantified the cell wall components and starch content of the leaf sheaths to clarify the changes over time in the two varieties. Figure 5 shows the changes in the dry matter weight and the components per centimeter of the fifth leaf sheath. Except for lignin, all components per unit length tended to be higher in Chugoku 117 than in Koshihikari throughout the entire investigation period (Figure 5a-f). The rate of the decrease in dry matter weight was higher in Chugoku 117, and the relative difference between both varieties decreased over time (Figure 5a). Holocellulose tended to decrease with time, but there were no significant changes in Koshihikari and Chugoku 117 until 4 and 6 weeks after heading, respectively (Figure 5b). The holocellulose was further divided into cellulose and hemicellulose. Cellulose accounted for a larger proportion of holocellulose than hemicellulose and, similar to holocellulose, there were no significant changes in Koshihikari and Chugoku 117 until 4 and 6 weeks after heading, respectively (Figure 5bd). Hemicellulose showed a relatively large decrease at 2 weeks after heading in Chugoku 117, but did not change significantly until 6 weeks after heading in either variety (Figure 5d). In Chugoku 117, the starch content decreased significantly and relatively rapidly at 2 weeks after heading and was almost absent at 4 weeks after heading (Figure 5e). In Koshihikari, the starch content decreased significantly at 2 weeks after heading and was almost absent at 2 weeks after heading (Figure 5e). Lignin increased over time in both varieties (Figure 5f). The sixth leaf sheath was more senescent than the fifth, and it was not possible to collect sufficient samples from 4 weeks after heading in Koshihikari and at 6 weeks after heading in Chugoku 117; however, up to 2 weeks after heading, Chugoku 117 tended to be higher dry matter weight and each components except for lignin than Koshihikari (Supplemental Figure S5).

Correlation between leaf sheath reinforcement and each trait
We examined the extent to which traits that changed over time in leaf sheaths were related to the bending moment at breaking with leaf sheaths. Tables 2 and 3 show the Pearson product-moment correlation coefficient for each trait in Koshihikari and Chugoku 117. For the traits except the bending moment at breaking with leaf sheaths and number of living leaf sheaths, the values for the fifth leaf sheath were used. Lignin showed an increasing trend over time in this study (Figure 5f), suggesting that it contributed little to the reinforcement of leaf sheaths to internodes. Therefore, correlation analysis was performed without lignin as a factor. In Koshihikari, dry matter weight was most strongly correlated with the bending moment at breaking with leaf sheaths, followed by the number of living leaf sheaths, membrane ion leakage and total chlorophyll content. The number of living leaf sheaths was most strongly correlated with the bending moment at breaking in Chugoku 117, followed by total chlorophyll content, dry matter weight, and starch content. In Koshihikari, relatively strong correlations (|r| ≥ 0.7; P < 0.05) were observed for all traits examined with respect to bending moment at breaking with leaf sheaths, but in Chugoku 117, holocellulose (r = 0.48; P = 0.11) and hemicellulose (r = 0.55; P = 0.083) contents, especially cellulose contents (r = 0.18; P = 0.59), were more weakly correlated than the other traits examined and not significantly correlated. Error bars indicate the standard error (n = 3). ** and * indicate significant differences between 5 days before heading and other days after heading at P < 0.01 and < 0.05, respectively (Dunnett's test).

Predictive model for the bending moment at breaking with leaf sheaths
Finally, to estimate the bending moment at breaking with leaf sheath, a single regression model was developed and fitted using the number of living leaf sheaths as well as the dry matter weight of the fifth leaf sheath, which were commonly correlated with the bending moment at breaking with leaf sheaths in the two varieties. Supplemental Figure S6a and b show a single regression model with the number of living leaf sheaths as the explanatory variable, and a single regression model with dry matter weight per centimeter of the fifth leaf sheath as the explanatory variable. When the number of living leaf sheaths was used as an explanatory variable, the coefficient of determination was 0.73 for Koshihikari and 0.80 for Chugoku 117, indicating that the model was highly accurate for both varieties limited to the present study. When dry matter weight was used as an explanatory variable, the coefficient of determination for Koshihikari was 0.75, indicating that the model was highly accurate, while that for Chugoku 117 was 0.59, which is rather weak accurate.

Morphological factors affecting leaf sheath reinforcement
The bending moment at breaking, which is an indicator of breaking-type lodging resistance, is expressed as the product of the section modulus and the bending stress (Ookawa & Ishihara, 1992). In the present study, the bending moment at the breaking of a culm with leaf sheaths was higher than that of a culm without leaf sheaths, but the difference decreased with time after the heading stage (Figure 1). Previous studies have shown that a higher number of leaf sheaths may increase the strength of a culm (Ookawa & Ishihara, 1992;Takaya & Miyasaka, 1983), and in the present study, the bending moment at breaking with leaf sheaths showed the highest positive correlation with the number of living leaf sheaths (Tables 2 and 3).
Although it should be noted that the number of living leaf sheaths is only one factor because the graphs did not match the behavior of the decrease in the number of living leaf sheaths and the bending moment at breaking with leaf sheaths, these studies and findings suggest that the presence of a high number of living leaf sheaths, which retain their structure, increases the section modulus of a stem and, consequently, the bending moment at breaking. In the present and previous studies, Chugoku 117 had a higher degree of leaf sheath reinforcement (Figure 1; Ookawa & Ishihara, 1992), but the morphological factors involved in this reinforcement have not been clarified. The present study showed that the leaf sheath of Chugoku 117 was 21% thicker in the fifth leaf sheath and 42% thicker in the sixth leaf sheath than that of Koshihikari due to the longer web-like parenchyma between aerenchyma (Figure 2, Table 1). These results suggest that there were differences in leaf sheath morphology among varieties, and that one of the morphological factors responsible for the greater degree of leaf sheath reinforcement in Chugoku 117 was the thickness of the sheath itself. It was also observed that the weblike parenchyma of Chugoku 117 were thicker than they were in Koshihikari, although the difference was not significant in the present study ( Figure 2, Table 1). It has been known that aerenchyma plays an important role in gas diffusion (Yin et al., 2021), but the thicker the aerenchyma causes the thinner the web-like parenchyma, which may have a negative effect on lodging resistance. Since the genes controlling leaf sheath morphology are largely unknown (Toriba et al., 2019), future studies should identify the genes that affect the morphology of different varieties and analyze their functions in relation to lodging resistance. The greater the area of the section modulus away from the neutral axis, the higher the value (the neutral axis is the line along which no stress or strain occurs when a bending moment occurs in a member). Therefore, it can be assumed that the thicker the culm, the farther away from the neutral axis the leaf sheath is located, even if the leaf sheath has the same thickness. In the present study, the leaf sheath of Chugoku 117 was located farther from the neutral axis because its culm was thicker than that of Koshihikari (Figure 2, Supplemental Table S4). This might be also one of the reasons for the high degree of leaf sheath reinforcement observed in Chugoku 117. Therefore, in order to increase the degree of leaf sheath reinforcement, not only should the number and thickness of living leaf sheaths be considered but so should the thickness of the culm itself.

Contribution of cell wall components and starch to leaf sheath reinforcement
In rice culms, cell wall components, such as cellulose, hemicellulose lignin, and non-structural carbohydrates, such as starch, have been reported to be associated with lodging resistance (Ishimaru et al., 2008;Kashiwagi et al., 2006;Li et al., 2015;Liu et al., 2018;Tanaka et al., 2003;Zhao et al., 2019). In rice leaf sheaths, on the other hand, little is known about the relationship of cell wall components and starch to lodging resistance.
In the present study, we evaluated the component content per cm of the fifth leaf sheath because dry matter weight per cm of leaf sheath has been related to stem strength in a previous study (Zhang et al., 2014) and because aerenchyma is sparsely present in the leaf sheath, making it difficult to investigate content per volume as in the internode. The trait most highly correlated with the bending moment at breaking with leaf sheaths was the dry matter weight per length in both varieties, and the starch content when considered as a component only (Tables 2 and 3). The starch content per unit length was 8% of its dry weight in Koshihikari, but it was 29% in Chugoku 117 at heading ( Figure 5). However, while starch was almost undetectable from 2 weeks after heading in Koshihikari and from 4 weeks after heading in Chugoku 117 (Figure 5), the bending moment at breaking with leaf sheaths and the degree of leaf sheath reinforcement continued to decrease after starch disappeared (Figure 1). These results suggest that starch in both varieties contributes relatively more than the other components investigated, especially in the first half of the ripening period.
In the present study, lignin was present in lower amounts in the leaf sheath compared to other components and showed an increasing tendency over time ( Figure 5). On the other hand, in rice, lignin has been reported to be localized in the cortical fiber tissue located on the outer side of a culm and to increase culm strength (Matsuda et al., 1983;Ookawa et al., 2014). These results and studies suggest that in the case of a stem divided into a culm and leaf sheaths, the effect of lignin on leaf sheath strength is minimal, and lignin may contribute strongly only to culm strength. However, a previous study in rice reported that leaf sheaths had the same level of lignin content as culms, although there were differences among rice varieties (Weng et al., 2017). Therefore, further investigation on the contribution of lignin to leaf sheath reinforcement is needed.
In terms of other cell wall components of the fifth leaf sheath, cellulose, hemicellulose, and holocellulose (the sum of these components) showed decreasing trends from the time of heading, but the rate of decrease was less than that of starch and they were relatively more stable in the leaf sheath than starch (Figure 5b-e). In Koshihikari, cellulose (r = 0.70; P = 0.012) and hemicellulose (r = 0.70; P = 0.011) were significantly correlated with bending moment at breaking with leaf sheaths, although they were not significantly correlated with it in Chugoku 117 (Tables 2 and 3). The main components of hemicellulose in the primary cell wall of Commelinid monocots such as rice are glucuronoarabinoxylan and mixed linkage β-1,3/1,4-glucans, which cross-link cellulose microfibrils to form the cell wall (Yokoyama & Nishitani, 2004). It has also been reported that cellulose crystallinity negatively influences lodging resistance (Li et al., 2015). In the present study, it was not possible to observe transverse sections of the leaf sheaths over time, primarily because the leaf sheaths became sludgy after heading. It should also be noted that the sixth leaf sheath became detached over time, which meant that only the fifth leaf sheath was available for physiological examination at all stages. However, the obtained findings suggest that the amount of cellulose and hemicellulose per length of leaf sheath may contribute to the reinforcement of leaf sheaths, whereas the strength may decrease when the hemicellulose cross-linking is broken or when cellulose crystallinity is high, even if the hemicellulose and/or cellulose content is high. Therefore, future research should concentrate not only on the content of cell wall components but also on the structure of the cell wall.

Relationship between leaf sheath senescence and reinforcement
It has been suggested that varieties with late senescence of leaf sheaths maintain high lodging resistance (Furuhata & Arima, 2007;Ookawa & Ishihara, 1992). In the present study, the degree of leaf sheath senescence was examined by investigating physiological factors and gene expression. The results showed that the number of living leaf sheaths was maintained in Chugoku 117 compared to Koshihikari because the expression of genes related to senescence tended to increase rapidly at 10 days after heading in Koshihikari and because the investigation of physiological traits showed that leaf sheath senescence occurred after Koshihikari in Chugoku 117 (Figures 3 and 4). Correlation with the bending moment at breaking with leaf sheaths showed a particularly strong positive correlation with the number of living leaf sheaths in both Koshihikari (r = 0.85; P < 0.001) and Chugoku 117 (r = 0.89; P < 0.001) among senescence associated traits (Tables 2 and 3). These strong correlations may be because the measurement of chlorophyll content and the membrane ion leakage rate were only performed on the fifth internode, whereas the number of living leaf sheaths was a trait that included multiple leaf sheaths covering the culm. The single regression prediction showed a higher coefficient of determination for the number of leaf sheaths than for dry matter weight in both varieties (Supplemental Figure S6). In addition, the dry matter weight used as an explanatory variable in the present study was only used to analyze the fifth leaf sheath, whereas the number of living leaf sheaths included all of the leaf sheaths covering the lowest fifth internode; consequently, the number of living leaf sheaths may more accurately reflects actual growth conditions (Supplemental Figure S6). On the other hand, previous studies have reported that stem strength and characteristics, including leaf sheaths, is variable among varieties and environments (Weng et al., 2017;Zhang et al., 2014). Therefore, in order to establish a more robust and generalized model, it is necessary to verify the model in multiple varieties and environments in the future.
In the present study, the degree of leaf sheath reinforcement converged to zero at harvest time, 6 weeks after heading, even in the late senescing variety Chugoku 117. The risk of lodging appears to be highest at harvest time because the center of gravity is raised by the mature panicle. It has also been reported that the loss of the lower leaf sheaths results in a higher center of gravity of the stem itself, which is independent of the weight of the panicle (Takaya & Miyasaka, 1983). These findings suggest the need to consider the breeding of new varieties and cultivation methods that delay further senescence and thus strengthen the lowest internode until harvest time. However, it has been suggested that leaf sheaths play a role as a sink for accumulating carbohydrates obtained by photosynthesis and nitrogen uptake from roots, which are then translocated to the panicle as the leaves senesce, contributing to seeds ripening (Cock & Yoshida, 1972;Havé et al., 2017;Ishimaru et al., 2004). Thus, there is a combination of source-sink allocation in practice, and the optimal balance of this trade-off must be carefully considered in order to achieve not only lodging resistance but also high yield.
Although the leaf sheath senescence of Chugoku 117 was relatively late in the present study (Figures 3 and 4), a previous study has reported that the percentage of ripened grains and 1000-grain weight were little different from those of Nipponbare, and the yield was higher than that of Nipponbare (Saitoh et al., 2002). These results suggest that the degree of delayed leaf sheath senescence in Chugoku 117 may not have much effect on grain filling. In fact, because typhoons might strike during the ripening period before harvest, delaying senescence may be a key component in lowering culm damage and increasing lodging resistance.

Conclusion
In rice, varietal differences in leaf sheaths that increase stem breaking strength have been reported, but the details of these factors and the changes in leaf sheath reinforcement over time are largely unknown. In the present study, we studied the morphological and physiological characteristics of the lower leaf sheaths that contribute to culm strength in two genetically different rice varieties during the ripening period. The results showed that the leaf sheath of Chugoku 117, which has higher leaf sheath reinforcement, was 21% thicker in the fifth leaf sheath and 42% thicker in the sixth leaf sheath than that of Koshihikari. These findings suggested that the thickness of the leaf sheath itself was one of the factors that contributed to the difference in the degree of leaf sheath reinforcement between varieties. The bending moment at breaking with leaf sheaths was stably and strongly correlated with the number of living leaf sheaths in both varieties (r = 0.85 for Koshihikari and r = 0.89 for Chugoku 117). Although starch in the leaf sheath was present at 8% of dry weight in Koshihikari and 29% of dry weight in Chugoku 117 at heading, it decreased with plant maturity suggesting that it contributed to stem strength only in the early stage of ripening. Thus, thicker leaf sheaths and delayed senescence of the leaf sheaths are expected to improve breaking-type lodging resistance.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
This work was supported by the WISE Program: Doctoral Program for World-leading Innovative & Smart Education of TUAT granted by MEXT, and by JSPS Grant-in-Aid for JSPS Fellows Grant Number 20J13277.