Natural Variation in the Flag Leaf Morphology of Rice Due to a Mutation of the NARROW LEAF 1 Gene in Oryza sativa L.

Abstract

We investigated the natural variations in the flag leaf morphology of rice. We conducted a principal component analysis based on nine flag leaf morphology traits using 103 accessions from the National Institute of Agrobiological Sciences Core Collection. The first component explained 39% of total variance, and the variable with highest loading was the width of the flag leaf (WFL). A genome-wide association analysis of 102 diverse Japanese accessions revealed that marker RM6992 on chromosome 4 was highly associated with WFL. In analyses of progenies derived from a cross between Takanari and Akenohoshi, the most significant quantitative trait locus (QTL) for WFL was in a 10.3-kb region containing the NARROW LEAF 1 (NAL1) gene, located 0.4 Mb downstream of RM6992. Analyses of chromosomal segment substitution lines indicated that a mutation (G1509A single-nucleotide mutation, causing an R233H amino acid substitution in NAL1) was present at the QTL. This explained 13 and 20% of total variability in WFL and the distance between small vascular bundles, respectively. The mutation apparently occurred during rice domestication and spread into japonica, tropical japonica, and indica subgroups. Notably, one accession, Phulba, had a NAL1 allele encoding only the N-terminal, or one-fourth, of the wild-type peptide. Given that the Phulba allele and the histidine-type allele showed essentially the same phenotype, the histidine-type allele was regarded as malfunctional. The phenotypes of transgenic plants varied depending on the ratio of histidine-type alleles to arginine-type alleles, raising the possibility that H²³³-type products function differently from and compete with R²³³-type products.

THE leaf of grasses typically consists of a relatively narrow blade and sheath enclosing the stem, and venation is parallel in the blade and the sheath (Esau 1977). Because large leaves intercept more light, the leaf area of the blade strongly affects final yield in cereal crops (Watson 1952). To produce plants that intercept light efficiently, leaf angle has been a target in breeding programs because erect leaves can capture more sunlight (Sinclair and Sheehy 1999). It was demonstrated that a brassinosteroid-deficient mutant with erect leaves showed increased grain yield under dense planting conditions (Sakamoto et al. 2005). It is also essential to understand the mechanism of development and the natural variations in morphology of the flag leaf since photosynthesis in the top three leaf blades of the plant, especially flag leaf, makes the largest contribution to the grain yield of rice (Tanaka 1958; Yoshida 1972).

The developmental processes of the flag leaf are the same as those of other leaves. In rice, the longitudinal strands in the leaf comprise the midrib, large vascular bundles, and small vascular bundles (Hoshikawa 1989). According to Inosaka (1962) and Itoh et al. (2005), the midrib and large vascular bundles initiate at the base of the leaf primordium and develop acropetally in the leaf and basipetally in the culm (stage P2 in leaf development). It takes about one plastochron to initiate a small vascular bundle after the initiation of the midrib. When the leaf sheath and blade start to differentiate (the beginning of stage P3), small vascular bundles become visible between the large vascular bundles at the base of leaf primordia. Small vascular bundles form acropetally in the leaf blade and basipetally in the stem. Later (stage P3), a small vascular bundle develops between the midrib and a large vascular bundle near the leaf tip and extends basipetally through the leaf blade. Then, more small vascular bundles form between large vascular bundles sequentially from the midrib toward the leaf margin. After the rapid elongation of the leaf blade (stage P4) and the leaf sheath (stage P5), the leaf becomes mature and growth is complete (stage P6).

Natural variations in flag leaf size have been reported for 491 rice accessions from Japan and 666 accessions from other countries (Matsuo 1952). Flag leaves were wider in accessions from Java, western China, and Latin America and narrower in those from north China, central China, and Russia. Flag leaves were longer in accessions from Java and India, but shorter in those from Taiwan, central China, and south China. In genome-wide association studies (GWAS) of 413 diverse accessions, significant loci accounted for ∼24% of variance in the width of the flag leaf (WFL), and three loci on chromosomes 1, 4, and 7 contributed 5.7, 5.0, and 6.1% of phenotypic variance, respectively (Zhao et al. 2011).

The NARROW LEAF 1 (NAL1) gene was located at ∼31.2 Mb on chromosome 4 (hereafter, all genomic positions are based on Os-Nipponbare-Reference-IRGSP-1.0), close to one of the single-nucleotide polymorphisms contributing to the variation in WFL reported by Zhao et al. (2011). NAL1 was originally isolated as a gene affecting vascular patterns in a study on a classic dwarf mutant with a narrow leaf (Qi et al. 2008) and was later shown to affect WFL, total spikelet number per panicle, photosynthetic rate, and chlorophyll content (Chen et al. 2012; Fujita et al. 2013; Takai et al. 2013; Zhang et al. 2014). NAL1 encodes a plant-specific protein, and the nal1 mutant with an in-frame deletion of 10 amino acids in exon 4 showed reduced basipetal polar auxin transport and fewer longitudinal veins, compared with wild type (Qi et al. 2008). A recent analysis of mutant rice with a NAL1 null allele implied that NAL1 is also involved in control of the cell cycle and cell division from the initial stage of leaf development onward (Jiang et al. 2015).

The NAL1 gene exhibits natural variations that are associated with plant morphology in rice. The NAL1 allele in Koshihikari, a temperate japonica cultivar, has three amino acid differences (R233H, A475V, and V484I) compared with NAL1 in the indica cultivar Takanari. A study on near-isogenic lines revealed that the Koshihikari NAL1 allele decreased the thickness of the flag leaf and increased the ratio of leaf area to dry mass, i.e., specific leaf area (Takai et al. 2013). The NAL1 allele in Daringan, a tropical japonica landrace, has the same predicted amino acid sequence as that of NAL1 in Koshihikari. When the Daringan allele was introduced into an indica cultivar with the same NAL1 protein as Takanari (in an IR64 background), the progeny showed increased flag leaf width, more vascular bundles, greater root biomass, more spikelets, and increased grain yield per square meter (Fujita et al. 2013), suggesting that NAL1 affects various yield-related traits.

In this study, we extensively examined world-wide collections of rice accessions using a variety of analyses, including principal component analysis (PCA) of morphological characters, GWAS, quantitative trait loci (QTL) analysis, and positional cloning, to explore the cause of the natural variations in the flag leaf morphology of rice. We found that an amino acid-altering mutation in NAL1, which occurred during rice domestication, significantly contributes to the natural variations in rice flag leaf morphology.

Materials and Methods

Plant materials

The rice accessions used in this study are summarized in Supporting Information, Table S1, and details are provided in File S1.

A total of 103 accessions from the National Institute of Agrobiological Sciences (NIAS) Core Collection represented natural variations in Oryza sativa L.; 56 accessions were from the World Rice Core Collection (WRC) (Kojima et al. 2005), and 47 accessions were from the Core Collection of Japanese Landraces (Ebana et al. 2008). Seeds of all of these accessions were planted in 2013 to collect data for the PCA (Table S1 and Table S2).

A total of 102 accessions (Table S1 and Table S3) were selected to represent the breeding history of rice in Japan in the GWAS. These accessions were selected because their population structure was simpler than that of the accessions in the NIAS Core Collection (Yamasaki and Ideta 2013). Seeds of these accessions were planted in 2009 and 2010.

Seeds of world-wide collections of rice accessions, including those described above, hybrid populations derived from a Takanari × Akenohoshi cross, and most of the chromosomal segment substitution lines (CSSLs) were sown in April. The seedlings were transplanted in May in 2009–2013 into an experimental field at NIAS, Tsukuba, Japan. These plants were used to evaluate flag leaf morphology traits such as the length and width of the flag leaf, number of vascular bundles, distance between vascular bundles, and thickness of the flag leaf. Seeds of the CSSLs that were used to evaluate the Phulba allele of NAL1 were sown in May 2013, and seedlings were transplanted in June.

Seeds of recombinant fixed lines containing either Akenohoshi or Takanari fragments were sown on April 20 and May 20, and seedlings were transplanted on May 20 and June 18 in 2010 and 2011, respectively, in an experimental field at the Tokyo University of Agriculture and Technology, Fuchu, Japan. These plants were used to evaluate yield and yield-related traits such as flag leaf size, dry weight of leaf and culm, and number of spikelets per plant.

Evaluation of flag leaf morphology

Plants with a heading date between July 20 and September 15 were used to evaluate flag leaf morphology. Flag leaves were sampled in the field, and their width and/or length were measured. In 2012, the widest part of the flag leaf was measured as WFL in the field using a ruler. To perform the measurement tasks efficiently, we developed an application program to record measured data using an iPod touch (Apple Inc. Cupertino, CA). Additional details about evaluating flag leaf morphology are provided in File S1.

Principal component analysis

A flag leaf was sampled from the main or equivalent culm of each plant, and nine traits were scored to identify the primary factors affecting flag leaf morphology by PCA (Table 1). Scores from five plants were averaged to represent each accession. In addition to length, width, and thickness of the flag leaf, the number of small or large vascular bundles and the distance between vascular bundles were recorded. The thickness of the flag leaf was represented by the thickness at three points around the second vascular bundle next to the midrib, i.e., the point of the large vascular bundle, the point of the small vascular bundle, and the point of the motor cell. All nine traits were used for the correlation analysis and PCA, which was conducted using JMP version 10.0 (SAS Institute, Cary, NC).

Figure 1.

Regions associated with flag leaf width and/or length in rice. (A) Genome-wide association analysis of flag leaf width (top) and length (bottom) using 102 rice accessions and 1596 DNA markers in 2009 and 2010. (B) Regions associated with flag leaf width and/or length were detected by evaluating four sets of CSSLs in which the whole genome was covered by at least one donor fragment. Each CSSL was compared with the background accession Koshihikari. When the CSSL is significantly different from the background, regions covered by donor segment are shown in green or light green boxes.

Genome-wide association analysis

We conducted the GWAS of flag leaf morphology using 102 rice accessions (Table S3) and 1596 DNA markers (Nagasaki et al. 2010; Yamasaki and Ideta 2013) tagged by Tagger (De Bakker et al. 2005). The GWAS model included the effects of multiple QTL and population structure and was based on the Bayesian method (Iwata et al. 2007; Yamasaki and Ideta 2013). A marker was regarded as significant if its mean gamma value was higher than the specific threshold of gamma = 0.1.

Quantitative trait loci analysis

Simple sequence repeat (SSR) markers spread throughout the genome were used for the QTL analysis. The SSR markers were selected from those described in McCouch et al. (2002) and the International Rice Genome Sequencing Project (2005). Polymorphisms were detected as described by Ebana et al. (2011). Genomic DNA was extracted from leaves of each plant by the cetyl trimethylammonium bromide method. A total of 96 and 149 markers were used to analyze 93 F₂ (Akenohoshi/Takanari) plants and 95 backcrossed inbred lines (BILs) (Jarjan/Koshihikari//Koshihikari), respectively. Linkage maps were constructed with MAPMAKER/EXP 3.0 software (Lander et al. 1987) using the Kosambi map function.

The length and width of the flag leaf were the averages of the three largest flag leaves per plant in the F₂ population (Akenohoshi/Takanari) or averages of the largest flag leaf of five plants per line in the BIL population (Jarjan/Koshihikari//Koshihikari).

The QTL analyses were performed by composite interval mapping as implemented by QTL Cartographer 2.5 software (http://statgen.ncsu.edu/qtlcart/WQTLCart.htm). Genome-wide threshold values (α = 0.05) were used to detect QTL based on the results of 1000 permutations (Churchill and Doerge 1994).

One-way analysis of variance

The 103 accessions from the NIAS Core Collection were divided into two groups according to the 1509th nucleotide (233rd amino acid) of their NAL1 gene; 62 accessions had a guanine (arginine) residue, and 41 accessions had an adenine (histidine) residue. To test the equality of the two groups, one-way analysis of variance for length and width of the flag leaf, number of vascular bundles, distance between small vascular bundles, and thickness of the flag leaf was performed using Microsoft Excel 2007. The variance between two groups was divided by the total variance to obtain the coefficient of determination (R²).

Production and evaluation of transgenic plants

Details of the production of transgenic plants are provided in File S1.

The copy number of transgenes in T₀ and T₁ plants was determined based on the amount of NAL1 genes relative to ubiquitin 2 (RUBQ2, AF184280) genes, as determined by quantitative real-time PCR. The primers used are shown in Table S4. The T₁ plants were classified into groups depending on the copy number of transgenes, and each group was compared to the vector control group using the F-test and t-test in Microsoft Excel 2007. To identify the T₁ plant groups that differed significantly from the vector control group, we used Tukey’s “Honest Significant Difference” method in the “multcomp” package of R v. 3.2.0 software (R foundation, Vienna).

Molecular phylogenetic analysis

The transcribed region of NAL1 was amplified from the 56 accessions from the WRC (Kojima et al. 2005) by PCR with the primers shown in Table S4. The PCR products were sequenced, and phylogenetic analyses were conducted with MEGA6 (Tamura et al. 2013). The complete deletion option was selected to use only sites shared among aligned sequences, and the T92 + G model was selected as the best-fit model with the lowest Bayesian information criterion score. The maximum-likelihood method was used to infer the phylogenetic tree and ancestral nucleotide sequences.

Homology searches in databases of other species

Homology searches were conducted using TBLASTN with the predicted NAL1 protein sequence of an accession from the NIAS Core Collection, Bei Khe (WRC 03), as the query. The searches were conducted against the genome databases of plants shown in Table S9B, and the sequence with the lowest E-value was selected for each species.

Real-time PCR analysis

Details on determining the transcript levels of NAL1 at various developmental stages and the transgene copy number in transgenic plants are provided in File S1. The primers used for real-time PCR are shown in Table S4. TaqMan real-time PCR was performed as described previously (Kojima et al. 2002). All TaqMan probes (Operon) were 3′-labeled with Black Hole Quencher-1a dye. The RUBQ2 probe was 5′-labeled with VIC, and HPT and NAL1 probes were 5′-labeled with the reporter dye FAM.

Data availability

Accessions of NIAS Rice Core Collection (Table S2 and Table S8) are available at https://www.gene.affrc.go.jp/databases-core_collections_wr_en.php for WRC, and https://www.gene.affrc.go.jp/databases-core_collections_jr_en.php for JRC.

Accessions having NIAS Genebank accession number (JP No) (Table S3) is accessible by ‘Plant Search’ in NIAS Genebank Databases (https://www.gene.affrc.go.jp/databases_en.php). As for wild rice (Table S9), accessions of Wild Core Collection Rank 1 provided by National Institute of Genetics (NIG) are available at http://www.shigen.nig.ac.jp/rice/oryzabase/strain/wildCore/about.

Results

First component of PCA explained 39% of total variance and WFL had highest loading

To identify the morphological characters of the flag leaf that primarily contribute to the natural variations in rice, we evaluated the following nine traits: length of the flag leaf (LFL), WFL, number of large vascular bundles (NLVB), number of small vascular bundles (NSVB), number of small vascular bundles between large vascular bundles (NSVB bet LVB), distance between small vascular bundles (DSVB), and thickness of flag leaf at the point of the large vascular bundle (TFL_LVB) at the point of the small vascular bundle (TFL_SVB) and at the point of the motor cell (TFL_MC) (Table S2, right). There were positive correlations (1) among the four traits WFL, NLVB, NSVB, and NSVB bet LVB; (2) between DSVB and WFL, TFL_SVB, or TFL_MC (Table S5); and (3) among the three thickness of flag leaf traits. The highest estimated correlation coefficient was 0.87.

The above nine traits were used in the PCA. This analysis covered 103 accessions from the NIAS Core Collection that represented the natural variations in O. sativa (Table 1). In the PCA, the first three components accounted for >80% of the total variance. PC1 and PC2 accounted for 39.2 and 27.5% of total variance, respectively, and the following variables showed substantial loadings: WFL, NLVB, NSVB, and NSVB bet LVB for PC1; TFL_LVB, TFL_SVB, and TFL_MC for PC2; and LFL and DSVB for PC3. WFL and its correlated traits were, therefore, the most significant characters to distinguish the natural variations in the flag leaf morphology of rice.

Table 1. Eigenvectors for nine flag leaf morphology traits in principal component analysis of 103 rice accessions from NIAS Core Collection.

Principal component		1st	2nd	3rd
Eigenvalue		3.52	2.47	1.21
Contribution ratio		39.2	27.5	13.5
Cumulative contribution		39.2	66.6	80.1
Trait	Abbreviation
Length of flag leaf	LFL	−0.13	0.08	0.78
Width of flag leaf	WFL	0.50	−0.05	0.21
No. of large vascular bundles	NLVB	0.39	−0.27	0.03
No. of small vascular bundles	NSVB	0.48	−0.24	−0.07
No. of small vascular bundles between large vascular bundles	NSVB bet LVB	0.43	−0.16	−0.10
Distance between small vascular bundles	DSVB	0.28	0.26	0.50
Thickness of flag leaf
At point of large vascular bundlea	TFL_LVB	0.15	0.42	−0.25
At point of small vascular bundleb	TFL_SVB	0.13	0.57	−0.03
At point of motor cellb	TFL_MC	0.22	0.51	−0.17

^aSecond large vascular bundle from midrib.

^bNear second large vascular bundle from midrib.

Scatter plots of PC1/PC2 or PC2/PC3 exhibited overlapping but distinct distributions of the indica, temperate japonica, and tropical japonica subgroups (Figure S1; see also Table S6) and reflected the morphological features of the three cultivars. Flag leaves tended to be thin and short with a moderate width in indica accessions; narrow, thick, and long in temperate japonica accessions; and wide, thick, and short in tropical japonica accessions.

QTL on chromosomes 4 and 8 contributed to natural variations in WFL

The GWAS for WFL (Figure 1A) identified a marker RM6992 at 30.8 Mb on chromosome 4 with a gamma value >0.1 over 2 years. Another marker, Rik6718, at 25.3 Mb on chromosome 8, also had a gamma value >0.1 in 2009, while no marker with gamma values >0.1 were detected for LFL.

We conducted whole-genome analyses of four CSSLs: two with indica accessions of O. sativa (Kasalath and IR64) as the donor, one with a temperate japonica accession (LAC23) as the donor, and one with Oryza rufipogon (IRGC-ACC104814) as the donor, in the Koshihikari background. These analyses, except for one using the LAC23 as the donor, identified a QTL for WFL on the long arm of chromosome 4 but no QTL on chromosome 8 (Figure 1B).

The variation in WFL was further studied by a QTL analysis followed by a mapping study using recombinant fixed lines. The analyses of the F₂ population derived from a Takanari × Akenohoshi cross revealed two QTL, one of which was located in the region near 32.2 Mb on chromosome 4 (Table S7 and Figure S2A). Three QTL, including one located in the region near 18.8 Mb on chromosome 8, were identified in the analysis of BILs derived from a BC₁F₁ (Jarjan/Koshihikari//Koshihikari) plant (Table S7). In both analyses, QTL for WFL were not linked to markers for heading date (HD), whereas QTL for LFL and for HD on chromosome 6 were linked to the same marker.

QTL for WFL on chromosome 4 was located within a 10.3-kb region

We analyzed the effect of the QTL on chromosome 4 on WFL using a pair of lines with the same fixed Takanari/Akenohoshi background except at two regions: one at 31.2–35.0 Mb on chromosome 4 and the other at 2.3–4.5 Mb on chromosome 7, which has no QTL (Figure S2, A and B). Compared with the line with the Takanari-derived allele, the line with the Akenohoshi-derived allele at the 31.2- to 35.0-Mb region on chromosome 4 showed greater WFL, spikelet number, hulled grain weight per plant, and panicle weight per growing area (Figure S2B). This result indicated that the 31.2- to 35.0-Mb region on chromosome 4 likely contained a QTL for WFL and/or yield.

We further examined an additional 77 recombinant fixed lines with the same background as that of F₄ line_10-7-58-16-1 or F₄ line_10-7-58-16-5 (Figure S2B). These analyses further restricted the QTL to a 10.3-kb region within the 31.2- to 35.00-Mb region on chromosome 4. This region contained only three ORFs: Os04t0615000 (NARROW LEAF 1), Os04t0615100, and Os04t0615200 (Figure S2C). Among the three genes, NAL1 had three nucleotide differences that would alter the encoded amino acid (one in exon 3 and the other two in exon 5). Another difference between Takanari- and Akenohoshi-derived alleles was the length of a poly-glutamine region, that is, a difference in the number of CTG repeats, in exon 2 of Os04t0615200 (Table 2 and Figure S2C).

Table 2. Four amino acid-altering differences in 10.3-kb region harboring QTL for flag leaf width: amino acid-altering nucleotide differences present in exon regions in 10.3-kb region.

	NARROW LEAF 1 (NAL1)			Os04t0615200
	3rd exon	5th exon		2nd exon
Nucleotide site	1509	2727	2753	No. of CTG repeats starting from 115th nucleotide
Amino acid residue	233	475	484	No. of glutamine residues starting from 39th amino acid
Parents used in F2 analysis
Takanari	G (R: arginine)	C(A: alanine)	G(V: valine)	3
Akenohoshi	A (H: histidine)	T(V: valine)	A (I: isoleucine)	4

Given that NAL1 was the best candidate for the QTL for WFL among the three ORFs, we examined its promoter region, since a previous report suggested that some NAL1 alleles were expressed at different levels (Zhang et al. 2014). A pair of recombinant fixed lines, F₅ line_10-7-52-84-3-1 and -2 (Figure S2D), which had either the Takanari or Akenohoshi fragment in the upstream promoter region but shared the Akenohoshi coding region, exhibited the same WFL and yield. The promoter region accounted for none or little of the variations in WFL and/or yield (Figure S2D), suggesting that mutations in the coding region, not the promoter region, caused the variation in WFL in this case.

R233H amino acid substitution in NAL1 caused variation in WFL and its phenotypic effect depended on the ratio of copy number of R²³³ type to H²³³ type

Eleven CSSLs with different donor fragments at the 10.3-kb region in the Koshihikari background (Figure 2) were used to identify the causative mutation of the QTL for WFL. Of the 11 CSSLs shown in Table 3, only 2 had the same WFL as Koshihikari (Table 3, column P). The donors of these CSSLs were Hayamasari or Tupa121-3. Among the four differences mentioned above, only the R233H amino acid difference in NAL1 coincided with the observed differences in WFL (Table 3).

Figure 2.

Chromosome 4 of 14 CSSLs in which the 10.3-kb region harboring the QTL for flag leaf width was replaced by a fragment from the donor. Vertical lines on chromosomes indicate positions of SSR or INDEL markers. White and colored regions indicate Koshihikari background and donor fragments, respectively. Triangle indicates 10.3-kb region. In three CSSLs with Naba, IR 64, or Deng Pao Zhai as the donor, heterozygous regions are shown in light color. Background of CSSLs in the remaining 11 chromosomes was largely from Koshihikari, as shown at right. Of 12 CSSLs with O. sativa as donor, 11 CSSLs (except for 1 with a Phulba fragment) were used to determine cause of variation in flag leaf width harbored in the 10.3-kb region.

Table 3. Four amino acid-altering differences in 10.3-kb region harboring QTL for flag leaf width: flag leaf width of CSSLs in the Koshihikari background.

		NARROW LEAF 1 (NAL1)			Os04t0615200
		3rd exon	5th exon		2nd exon
	Nucleotide site	1509	2727	2753	No. of CTG repeats starting from 115th nucleotide
	Amino acid residue	233	475	484	No. of glutamine residues starting from 39th amino acid
						Flag leaf width (mm; n = 5)
	Accession	Nucleotide (amino acid)				Year	CSSL	Control (Koshihikari)	P
Background	Koshihikari	A (H)	T(V)	A(I)	4
Donor of CSSL	Hayamasari	A (H)	T(V)	A(I)	4	2012	11.2 ± 0.45	11.6 ± 0.55	NS
	Tupa 121-3	A (H)	T(V)	A(I)	3	2012	12.0 ± 0.00	12.4 ± 0.55	NS
	Khao Nam Jen	G (R)	T(V)	A(I)	3	2012	9.0 ± 0.00	12.4 ± 0.89	6.2E-05
	Naba	G (R)	T(V)	A(I)	3	2013	9.4 ± 0.55	12.4 ± 0.55	2.5E-05
	Shuusoushu	G (R)	T(V)	A(I)	3	2013	9.6 ± 0.55	13.2 ± 1.10	1.7E-04
	Basilanon	G (R)	C(A)	A(I)	3	2013	9.8 ± 0.45	11.8 ± 0.84	1.5E-03
	Kasalath	G (R)	C(A)	G(V)	3	2013	9.0 ± 0.71	11.4 ± 0.55	9.4E-05
	IR64	G (R)	C(A)	G(V)	3	2012	10.2 ± 0.45	12.6 ± 0.55	3.2E-05
	Bei Khe	G (R)	C(A)	G(V)	3	2012	8.8 ± 0.45	12.2 ± 0.45	1.1E-06
	Deng Pao Zhai	G (R)	C(A)	G(V)	3	2013	10.8 ± 0.45	12.2 ± 0.45	1.1E-03
	Bleiyo	G (R)	C(A)	G(V)	3	2013	8.6 ± 0.55	13.2 ± 0.45	4.9E-07

To gain insight into the R233H amino acid mutation, we compared eight traits between Koshihikari and CSSLs whose donor was one of four O. sativa accessions (Hayamasari, Tupa 121-3, Khao Nam Jen, Bei Khe) or one of two O. rufipogon accessions (IRGC-ACC104814, IRGC-ACC101941) (Figure S3 and Figure 2). The seven traits in addition to WFL were LFL, NLVB, NSVB, TFL_SVB, DSVB, number of cells between motor and epidermal cells (NCME), and number of cells between large vascular bundle and epidermal cells (NCLVBE). We also compared these traits between a pair of recombinant fixed lines that had a Takanari fragment (F₄_10-7-58-16-5) or an Akenohoshi fragment (F₄_10-7-58-16-1) at the 31.2- to 35.00-Mb region (Figure S2B). These analyses confirmed that the 233rd amino acid substitution in NAL1 affected NSVL, NCME, and NCLVBE, as well as WFL (Figure S3).

To further explore the effect of the R233H amino acid mutation on WFL, we transformed two lines (SL2013 with R-type NAL1 and Akenohoshi with H-type NAL1) with either Takanari NAL1 or the Takanari NAL1_R233H vector (see File S1). We observed the phenotypes of the T₀ and T₁ generations (Figure 3) and found that the WFL increased when the H construct was introduced into the R-type background, whereas the WFL decreased when the R construct was introduced into the H-type background. There was no apparent effect when the R construct was introduced into the R-type background or when the H construct was introduced into the H-type background. In the R-type background (Figure 3B), WFL was correlated with NSVB and DSVB but not with NLVB. Interestingly, the WFL, NSVB, and DSVB of transgenic plants increased as the copy number of H-type alleles increased from zero to two, but their values did not exceed those of Koshihikari (H-type) even if the copy number of H-type alleles was more than two. In the H-type background (Figure 3C), the copy number of the H-type allele had little or no effect on traits related to WFL. Therefore, the proportion rather than the actual copy number of introduced vectors (H-type alleles) was important for WFL. Interestingly, the wider flag leaves with increased NSVB and DSVB were thinner at the point of the small vascular bundles (Figure 3D).

Figure 3.

Flag leaf morphology of plants transformed by the NAL1 gene with the endogenous promoter. Vectors or background with arginine or histidine at the 233rd amino acid residue of NAL1 are R-type and H-type, respectively. SL2013 and Takanari are R-type; Koshihikari and Akenohoshi are H-type. Vector with Takanari allele is R vector; R vector in which the 233rd arginine is replaced by histidine is H vector. (A) T₀ transgenic plants harboring R vector, H vector, and empty vector as control in SL2013 or Akenohoshi background. (B) T₁ transgenic plants harboring R vector, H vector, and vector control in SL2013 background. Copy number of introduced vectors, total copy number of NAL1, and proportion of H-type are shown below. Each experimental group was compared to vector control (t-test), and P-values are shown above plots. Different letters below plots indicate significant differences (P < 0.05, Tukey’s HSD test). (C) T₁ transgenic plants harboring H vector in Akenohoshi background. (D) T₁ transgenic plants harboring H vector or R vector in SL2013 background.

R233H amino acid mutation occurred during O. sativa domestication and accounted for 13% of total variation in WFL among rice accessions from the global collection

We sequenced the transcribed region of the NAL1 gene from 69 accessions from the NIAS Global Core Collection, the four parents of the CSSLs, and the two parents of the hybrid population used in the QTL analysis (Table S8A). The obtained sequences were classified into seven types according to their predicted amino acid sequences (Figure 4). Notably, 23 accessions including Nipponbare or Koshihikari had a retrotransposon insertion in exon 1 with duplicated regions at the 3′ end of exon 1 and the 5′ end of exon 2 (type IV′), although the nucleotide sequences in the coding region were intact and essentially the same as that in Akenohoshi (type IV). The type IV′ allele was also present in LAC23 and Jarjan, which explained why no QTL on chromosome 4 was detected in the analysis of a CSSL with LAC23 as donor and from BILs derived from BC₁F₁ (Jarjan/Koshihikari//Koshihikari). One accession, Phulba, had an allele in which the deletion of a large part of the inserted retrotransposon resulted in a stop codon near the end of exon 1 (type V).

Figure 4.

Seven types of NAL1 genes observed in nature. Type IV′ with retrotransposon insertion presumably encodes the same amino acid sequence as that encoded by type IV. In total, 75 accessions (69 accessions from NIAS Global Core Collection, 4 donors of CSSLs not included in Core Collection, and 2 parents used to produce hybrid population for mapping QTL for flag leaf width) were classified into seven types. Boxes indicate transcribed regions; regions translated into polypeptide are shown in orange. Gray bar shows inserted retrotransposon.

Using the nucleotide sequences of only the shared region, and excluding the retrotransposon region, the maximum-likelihood tree was estimated by MEGA6 by inferring ancestral amino acid sequences under the best-fit evolutionary model of T92 + gamma (Figure 5). As for the 233rd amino acid, accessions with R²³³ or H²³³ were distributed in all three subgroups: indica, temperate japonica, and tropical japonica. However, most of the accessions with R²³³ were in the indica subgroup.

Figure 5.

Maximum-likelihood tree of rice NAL1 genes from 75 accessions. Types of NAL1 protein (see Figure 4) and numbers of accessions for each subgroup (indica, tropical japonica, and temperate japonica) are shown. Note that the nucleotide sequences used in the phylogenetic analysis are same for the 15 accessions included by “Nipponbare*³”, and that Nipponbare*³ includes type IV, type IV′ and type V at the encoded protein level due to indels of retrotransposon. (Right) The 233th amino acid of NAL1, 1509th nucleotide of the NAL1 gene, and the retrotransposon insertion are shown. *¹Jarjan and Kalo Dhan. *²Muha, Jhona 2, Napal 8, Anjana Dhan, Tupa 121-3, Surjamukhi, and ARC7291. *³Includes one type IV (Akenohoshi, temperate japonica,), 13 type IV′ (two indica, i.e., Shwe Nang Gyi and Kaluheenati, five tropical japonica, i.e., Khao Nok, Jaguary, Padi Perak, Rexmont, and Urasan 1, five template japonica, i.e., Nipponbare, Dianyu 1, Koshihikari, Hayamasari, and LAC23, and one unknown, Calotoc) and one type V (Phulba, temperate japonica). *⁴Ryou Suisan Koumai, Tadukan, Pinulupot 1, and Khau Mac Kho. *⁵Naba, Shuusoushu, Keiboba, Khau Tan Chiem, and Khao Nam Jen. *⁶Vary Futsi and Deejiaohualuo. *⁷Puluik Arang and Neang Menh. *⁸Bei Khe, Davao 1, Co 13, Qingyu (Seiyu), Lebed, Milyang 23, Neang Phtong, Pokkari, Chin Galay, Vandaran, IR64, and Takanari. *⁹Neang Menh, Radin Goi Sesat, Kemashin, and Rambhog. *¹⁰Ratul and Local Basmati. *¹¹Kasalath, Jena 035, ARC 7047, and Badari Dhan.

Exon 3 of NAL1 was sequenced for 23 accessions from 10 wild rice species, and all had G at the 1509th nucleotide and encoded the amino acid R at the 233rd residue (Table S9A). Interestingly, the survey of NAL1 homologs in the genome databases of plants of various taxa (Table S9B) revealed that species in the Streptophyta, including land plants (Embryophyta) had NAL1 homologs with R at the corresponding amino acid position, while species in the Rhodophyta or Chlorophyta had no detectable NAL1 homolog. This finding suggested that the original amino acid residue at this position in the rice NAL1 was R, which later mutated into H.

We conducted a one-way analysis of variance of 103 accessions from the NIAS Core Collection to examine the functional significance of the G1509A nucleotide substitution (R233H amino acid substitution). The results showed that the substitution accounted for 13, 20, and 6% of the total variation in WFL, DSVB, and NSVB, respectively (Table 4).

Table 4. One-way analysis of variance based on the 1509th nucleotide (233rd amino acid) in 3rd exon of NAL1 using 103 accessions from NIAS Core Collection.

	1509th nucleotide (233rd amino acid) in 3rd exon
	G (arginine), n = 62	A (histidine), n = 41	One-way analysis of variance
Trait	Average	Average	P	R2
Length of flag leaf (mm)	330 ± 5165	349 ± 3984	NS	—
Width of flag leaf (mm)	13.2 ± 5.0	15.3 ± 9.4	1.39E-04	0.13
No. of large vascular bundles	12.4 ± 1.8	12.9 ± 3.1	NS	—
No. of small vascular bundles	44.9 ± 78.1	49.8 ± 109.8	1.32E-02	0.06
No. of small vascular bundles between two large vascular bundles	4.63 ± 0.24	4.90 ± 0.35	1.52E-02	0.06
Distance between small vascular bundles (μm)	196 ± 368	215 ± 250	2.13E-06	0.20
Thickness of flag leaf (μm)
At point of large vascular bundlea	231 ± 860	232 ± 508	NS	—
At point of small vascular bundleb	115 ± 191	115 ± 142	NS	—
At point of motor cellb	98.2 ± 126.5	98.5 ± 92.6	NS	—

^aSecond large vascular bundle from midrib.

^bNear second large vascular bundle from midrib.

H²³³-type NAL1 gene is likely to be a malfunctional allele

One accession in the NIAS Core Collection, Phulba, had an allele encoding a truncated NAL1 protein of only 148 amino acids, in contrast to the full NAL1 with 582 amino acids (type V in Figure 4). The phenotypes of Koshihikari with the H-type allele (type IV′ in Figure 4) and a CSSL containing a Phulba-derived fragment of NAL1 in the Koshihikari background were compared (Figure 2). There were no significant differences in any of the traits (Figure 6), indicating that the phenotypes of the Phulba allele and H-type allele were essentially the same and that both were likely to be malfunctional.

Figure 6.

Comparison of flag leaf morphology and culm length among three genotypes: Phulba allele (type V), Koshihikari allele (type IV′), and Phulba/Koshihikari allele. No significant differences were observed among the three types.

NAL1 expression peaked at stage P3 when small vascular bundles form

Immature flag leaves were sampled to survey the transcript levels of NAL1 throughout flag leaf development. Leaves at stage P2 and at the beginning of stage P3 were too small to excise intact, and so the whole shoot apex was used instead. In the period from stage P2 to P6, NAL1 transcript levels peaked at stage P3. At this stage, there was no difference in NAL1 transcript levels among IR64 (R-type; type I in Figure 4), Koshihikari (H-type; type IV′), and Phulba (truncated protein; type V) (Figure 7).

Figure 7.

Gene expression of NAL1 relative to that of ubiquitin (UBQ). (A) Changes in gene expression throughout early developmental stages of flag leaf in Koshihikari (type IV′) (mean ± SD, n = 3). Note that shoot apex includes flag leaf primordia. Stages of leaf: P2, hood-shaped primordium; P3, formation of blade-sheath boundary; P4, rapid elongation of leaf blade, P6, growth is completed. (B) Comparison of IR64 allele (type I), Koshihikari allele (type IV′), and Phulba allele (type V) in Koshihikari background at stage P3 (mean ± SD, n = 3).

Discussion

H²³³ in NAL1 was likely selected during rice domestication

Seven different NAL1 types have been detected among the rice accessions examined so far. The progenitor of cultivated rice had R²³³ in its NAL1 protein, given that all 23 accessions from 10 wild rice species including O. rufipogon, from which O. sativa originated, had R²³³ (Table S9A). The presence of the R-type allele of NAL1 in land plants also suggests that the R-type allele is the ancestral form and that the H-type allele is derived from it (Figure 5). The R233H amino acid mutation, the insertion of a retrotransposon, and the subsequent deletion of the retrotransposon distinguished type IV from type III, type IV′ from type IV, and type V from type IV′, respectively (Figure 4). Our analyses raise the possibility that the three major subgroups of cultivated rice differentiated after the insertion of the retrotransposon into the H-type allele as both R²³³ (type I, II, III, VI, and VII) and H²³³ (type IV and IV′) were present among the indica, temperate japonica, and tropical japonica subgroups (Figure 4 and Table S8A).

According to Huang et al. (2012), ancient japonica was first domesticated from a specific population of O. rufipogon in southern China. Subsequently, indica originated from crosses between ancient japonica and local O. rufipogon as the initial cultivar of temperate japonica spread into Southeast and South Asia. Considering this pattern of distribution, we propose an evolutionary scenario in which the R233H amino acid substitution and the following retrotransposon insertion occurred at the differentiation of ancient japonica and that the temperate japonica, indica, and tropical japonica accessions with H-type alleles originated from crosses between O. rufipogon and ancient japonica harboring the mutation.

Although further population genetics studies are required to confirm this scenario, the current data suggest that the H-type allele has been selected for during rice domestication. The evidence for this hypothesis can be summarized as follows: the angiosperms examined so far have retained R-type NAL1 homologs for a long time during evolution, implying a selective advantage of R-type alleles over others in the natural environment. However, both H-type and R-type alleles have persisted among the three rice subgroups during the history of rice cultivation. The H-type allele directly or indirectly leads to increased WFL, increased number of spikelets per panicle, and often increased yields, although its effects vary among accessions (see below). Therefore, the H-type allele seems to have provided some desirable characters, e.g., increased grain yields, with little, if any, adverse effects under artificially controlled environments. We propose that NAL1 has been one of the important genes in the domestication of cultivated rice.

Amino acid change in NAL1 is important for increased WFL

Previous reports have suggested that there are differences among rice accessions in terms of the quantity and quality of NAL1 gene expression. It was reported that the H-type allele in HB277 but not the R-type allele in D50 was subject to extensive alternative splicing, while both were expressed at similar levels (Chen et al. 2012). In Koshihikari (type VI′), only 20% of NAL1 transcripts contained no retrotransposon, and a NAL1 protein the same size as that encoded by the R-type allele was synthesized (Takai et al. 2013). Furthermore, it was shown that NAL1 expression was significantly higher in Nipponbare (type IV′) than in 93-11 (type I), although the expression level was shown to vary among developmental stages (Zhang et al. 2014). Overexpression of the Nipponbare allele in Nipponbare increased FWL and LWL through increasing the distance between vascular bundles (Zhang et al. 2014), indicating that the NAL1 transcript level is an important factor in leaf morphology.

Nonetheless, most of the current evidence indicates that the different effects of NAL1 types on flag leaf phenotypes are mainly due to the amino acid at the 233rd position, rather than to changes in transcript levels. First, there was no phenotypic difference between a pair of recombinant fixed lines with the same H²³³-type NAL1 driven by either the Takanari or Akenohoshi promoter (Figure S2D). Second, there was no difference in gene expression among the IR64 allele (type I), the Koshihikari allele (type IV′), and the Phulba allele (type V) at stage P3 (Figure 4 and Figure 7B). Third, transformation by two vectors driven by the same Takanari promoter with different amino acids at the 233rd residue resulted in marked differences in phenotype (Figure 3).

Products of the malfunctional H²³³-type allele still regulate WFL and compete with R²³³-type products

Although the H-type allele is thought to be malfunctional, it must retain some function because the phenotype of plants homozygous for H-type alleles differed from that of plants homozygous for the null allele. The leaves of plants with the null allele were 50% narrower and shorter than those of the H-type plants (Nipponbare, type IV′) (Jiang et al. 2015). The H-type allele was neither recessive nor dominant, as transgenic plants with both the R-type allele and the H-type allele exhibited intermediate phenotypes (Figure 3B). The effect on phenotype did not depend on the dosage of R-type or H-type genes, but on the ratio of the copy numbers of R-type to H-type (Figure 3). Therefore, the H²³³-type NAL1 products likely compete with R²³³-type products to regulate WFL.

Notably, the deletion of two amino acids in exon 1 resulted in a phenotype similar to that of the null mutant, with a 50% reduction in the width and length of the leaf (Jiang et al. 2015). Compared to a deletion in exon 1, a 10-amino-acid deletion in exon 4 had a milder effect: it reduced the blade width by 37% and the blade length by 13% (Qi et al. 2008). The Phulba allele, which produces a truncated protein consisting of only exon 1 (type V), resulted in a phenotype similar to that of the H-type allele (type IV′), which appeared to have no obvious defects (Figure 6). Therefore, the protein domain encoded by exon 1 seems to be essential for the regulation of WFL and LFL, as well as the competition of R²³³-type and H²³³-type products.

NAL1 regulates flag leaf width and exhibits various pleiotropic effects

NAL1 derived from Nipponbare (type IV′) was shown to promote cell division in the anticlinal direction and suppress it in the periclinal direction with no or little effect on cell size in the leaf (Jiang et al. 2015). This is consistent with our observations, where WFL, NSVB, DSVB were larger and TFL_SVB was smaller in the line transformed with the H-type allele (Figure 3). Takai et al. (2013) reported that the R-type allele derived from Takanari (type I) increased leaf thickness and the H-type allele derived from Koshihikari (type IV′) decreased leaf thickness. These results can be explained if the R-type allele is regarded as an allele with a weaker effect than the H-type allele on cell division in the anticlinal and periclinal directions. This would also explain the increase in WFL when the Daringan allele (H-type allele) was introduced into IR64 (R-type allele) (Fujita et al. 2013) and when the Nipponbare allele (H-type allele) was introduced into 93-11 (R-type allele) (Zhang et al. 2014). However, it could not explain the results of Chen et al. (2012), where the R-type allele increased WFL in mapping of QTL for WFL using hybrid populations originated from a cross between D50 (R-type allele) and HB277 (H-type allele). This discrepancy might be because of the presence of a large number of alternatively spliced forms of the H-type allele and a low level of the NAL1 protein (Chen et al. 2012). However, further analyses are required to identify the exact reason for this discrepancy.

Gene expression of NAL1 was observed as early as the formation of leaf primordia (Jiang et al. 2015). NAL1 was highly expressed in the vascular bundle, especially in the phloem (Qi et al. 2008), and its expression peaked at the P3 stage when the small vascular bundle formed. Therefore, its main effect appears to be on the formation of vascular bundles, especially small vascular bundles, which subsequently affect leaf width and morphology. Nonetheless, the R233H amino acid mutation of NAL1 exhibits a pleiotropic effect. In addition to the increase in WFL and the number of vascular bundles, Nal1 increased the total number of spikelets per panicle, root dry weight, and the rate of filled grains (Fujita et al. 2013). Improvements to the panicle, such as increased panicle length, more spikelets per panicle, and more secondary branches per panicle, were also observed after introducing the H-type allele (Zhang et al. 2014). Since NAL1 is expressed in the culm, coleoptile, crown root, lateral root, and panicles (Qi et al. 2008; Fujita et al. 2013; Jiang et al. 2015), its regulation of the cell cycle and cell division in these tissues likely explains the pleiotropic effects of the Nal1 mutation.

With respect to the grain yields, the effects of R-type and H-type alleles seem to vary. Higher photosynthetic rate per area was attained when the Koshihikari allele (H-type) was replaced by the Takanari allele (R-type) due to the increased number of mesophyll cells between vascular bundles and the larger total mesophyll area between the vascular bundles (Takai et al. 2013). Meanwhile, when IR64 (R-type allele) was replaced by the Daringan allele (H-type allele), grain yield per area increased despite the decreased number of panicles per plant and lower-1000-grain weight (Fujita et al. 2013). When the 93-11 allele (R-type) was replaced by the Nipponbare allele (H-type), the yield and chlorophyll content in the leaf increased, and panicle morphology was affected (Zhang et al. 2014). Grain yield is also affected by other factors including the primary structure and expression level of NAL1, the background genome, and the growth conditions/environmental factors. Therefore, the outcomes of different types of NAL1 alleles in various rice accessions are not easily predicted.

NAL1 may be involved in auxin transport and response

It has been speculated that NAL1 is involved in polar auxin transport, since a nal1 mutant lacking 10 amino acids from exon 4 produced fewer vascular bundles and contained lower levels of OsPIN1 protein, compared with wild type (Qi et al. 2008). The fact that the expression of the AUXIN RESPONSE FACTOR gene family (ARF1, ARF2, and ARF3) and PIN1 was reduced in the NAL1 null mutant (Jiang et al. 2015) supports the involvement of NAL1 in the auxin response. In this regard, it is notable that a NAL1 homolog was found in Klebsormidium flaccidum, a terrestrial alga with a primitive body plan in the phylum Streptophyta (Table S9B). This species produces multicellular and nonbranching filaments without differentiated or specialized cells, but its genome contains most of the genes required for auxin biosynthesis, auxin receptors, auxin sensing, and auxin transport. Also, auxin (indole-3-acetic acid) was detected in tissues of this terrestrial alga by mass spectrometry (Hori et al. 2014). Future research, including studies on the phylum Streptophyta, will shed light on the function of NAL1 and its involvement in the auxin response.