QTL-seq identifies NAL1 and OsOFP19 as additive regulators of tiller number in rice (Oryza sativa L.)

Abstract

Background

Tiller number is a critical component of rice yield, as it directly influences overall productivity. While upland rice varieties are well adapted to lowland environments and prove resilient to fluctuating water availability, their typically low tillering capacity limits their performance in lowland ecosystems where conditions are more conducive to achieving higher yields.

Results

To facilitate the marker-assisted selection (MAS) breeding of upland rice cultivars suitable for lowland conditions, we performed QTL-seq analysis using populations derived from a cross between a high-tillering lowland indica parent (PTT1) and a low-tillering upland tropical japonica line (NDCMP49). Two major QTLs associated with tiller number were identified on chromosomes 4 and 5 and designated as qTN4 and qTN5, respectively. Candidate gene analysis revealed NAL1 and OsOFP19 as putative genes underlying these loci. Functional validation of NAL1 using CRISPR-Cas9 knockout mutants confirmed its role as a negative regulator of tillering, as two independent alleles of nal1 mutant plants exhibited significantly increased tiller numbers compared with the wild type. Marker–trait association analysis further supported the additive effect of qTN4 (NAL1) and qTN5 (OsOFP19), indicating their potential for pyramiding in breeding programs. Functional KASP markers of NAL1 and OsOFP19 were developed and successfully validated in segregating populations, demonstrating their applicability for marker-assisted selection.

Conclusions

Collectively, these findings advance our understanding of the genetic regulation of tillering in rice and provide molecular tools for improving plant architecture and yield in upland rice varieties cultivated under lowland conditions.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12870-025-07239-6.

Background

Global food security is increasingly threatened by the dual challenges of climate change and rapid population growth [1]. As a staple food for more than half of the world’s population, rice (Oryza sativa L.) plays a central role in sustaining global food systems [2]. To meet future food demands under increasingly variable environmental conditions, increasing the productivity and adaptability of rice is a key objective of crop improvement programs. Currently, the majority of the world’s rice is grown in lowland ecosystems, which rely on the constant water supply to maintain flooded or waterlogged soils that are essential for high yields [3]. In contrast, rice is also grown in upland ecosystems, where aerobic soil conditions and limited water availability have driven the evolution of various adaptive traits [4].

In Southeast Asia, the rice germplasm exhibits considerable genetic diversity, with indica being the predominant subspecies, while aus, temperate japonica and tropical japonica varieties are also grown in certain agroecological environments [5]. In Thailand, for example, the rice gene pool is dominated by indica, while tropical japonica varieties are mainly grown in the arid mountainous regions [6]. Tropical japonica is thought to have diverged from temperate japonica by adapting to the highlands and mountainous regions [7], acquiring traits such as tall stature, deep root system and lower tillering ability. These adaptations are beneficial under drought-prone conditions but often result in lower yield potential compared to varieties adapted to lowland conditions [8]. However, when cultivated under favorable lowland conditions, upland rice often does not produce high yields due to its inherently low tillering ability, which directly limits the number of panicles and total grain production. Improving the tillering ability of upland rice is a key strategy to improve its adaptation and performance in lowland systems.

The number of tillers is an important yield component as it directly determines the number of productive panicles per plant [9]. In water-abundant environments, increased tillering can contribute to higher biomass, better resource utilization and greater yield stability [10, 11]. Despite the physiological importance of tillering, its genetic regulation is complex and involves a network of developmental and hormonal pathways. Auxin and strigolactones act as negative regulators of tiller bud outgrowth, while cytokinin and brassinosteroids promote tiller formation [12, 13]. Tiller development is initiated by the formation of axillary meristems (AMs), which then give rise to tiller buds. Several genes have been shown to regulate these processes, including MONOCULM1 (MOC1), MONOCULM3TILLER ABSENT1/SRT1, LAX PANICLE1 (LAX1) and LAX2 [14–20]. NAL1 (Narrow leaf 1) encodes a plant-specific protein implicated in regulating leaf morphology, photosynthetic capacity, and yield-related traits in rice. Previous work by Jiang et al. linked NAL1 to tiller number using a mutant derived from random mutagenesis in the Nipponbare background [21]. Further studies have shown that microRNAs and dwarfing genes are involved in the modification of tillering phenotype [13, 22, 23]. Although numerous genes involved in tillering development have been identified, relatively few have been translated into useful molecular markers for breeding purposes. In addition, little attention has been paid to tillering traits in upland rice, although they are of great value for increasing yield in lowland cultivation. Given the particular architecture and adaptation of the upland germplasm, there is a clear need to investigate the genetic basis of tillering in this group to enable targeted trait improvement.

In the present study, we used QTL-seq, a next-generation sequencing-based bulked segregant analysis (BSA) method [24], to identify genomic regions associated with tiller number in rice. F₂ and F_2:3 populations derived from a cross between an indica lowland parent with high tiller production (PTT1) and a tropical japonica upland line with low tiller production (NDCMP49) were used for QTL-seq analysis. Key QTLs were mapped and candidate genes were identified. Functional markers were developed and validated to facilitate marker-assisted selection (MAS). This work provides new insights into the genetic regulation of tillering in upland rice and facilitates development of molecular markers for breeding programs aimed at increasing yield potential in lowland environments.

Materials and methods

Plant materials and phenotype evaluation

An F₂ population derived from a cross of Niaw Dam Chaw Mai Pai 49 (NDCMP49) x Pathum Thani 1 (PTT1) was used for a QTL-seq analysis to identify QTLs associated with tillering ability under lowland (waterlogged) conditions. NDCMP49, an upland tropical japonica cultivar, was used as a low tiller number (low-TN) parent with an approximate TN of 9 tillers per plant, and PTT1, an elite Thai fragrant lowland indica rice cultivar with an approximate TN of 20 tillers per plant, was used as a high tiller number (high-TN) parent. Three-week-old seedlings from a total of 492 F₂ individuals from this cross were grown together with their parents in the field at a distance of 25 cm x 25 cm between the plants. A total of 2.3 g of chemical fertilizer N-P-K (21-7-14) was applied per plant in three splits at 15, 30 and 50 days after transplanting. The standing water level was maintained at 5 cm above the soil throughout the season. The field experiments were conducted at the rice field of the Rice Science Center, Kasetsart University, Kamphaeng Saen, Nakhon Pathom, Thailand.

Bulk construction, DNA isolation and whole-genome re-sequencing

The tillering ability of each F₂ line and the parental lines was evaluated by counting the tiller number at the maximum tillering stage. A total of 30 F₂ individuals with the highest tiller number and the other 30 individuals with the lowest tiller number were selected and assigned into high-TN and low-TN groups, respectively. The leaf samples of all 60 lines were collected individually and stored at -80 °C until they were used for DNA isolation. To select the F₂ lines with robust phenotype from the preselection, we reconfirmed the phenotype in F_2:3 families corresponding to the selected F₂ lines. Fifteen plants of each F_2:3 family were grown in the field under the same growing conditions as previously described and the tiller number from each plant was counted at the maximum tillering stage. Ten F_2:3 families with the highest average tiller numbers and another ten F_2:3 families with the lowest average tiller number were selected and grouped into high-TN and low-TN bulks. The DNA of the individual F₂ lines corresponding to the 20 selected F_2:3 families was extracted from the stored leaf samples using Dneasy^® Plant Mini Kit (QIAGEN, Hilden, Germany). The DNA of the individual F₂ plants as well as that of the both parents was sequenced the whole genome using an MGI-Seq platform at the China National Gene Bank (CNGB) Shenzhen, China.

Sequencing data analysis and identification of genomic regions associated with tillering ability

Raw sequencing data of each sample were quality-filtered and trimmed using Trimmomatic [25] to remove adapter sequences and low-quality reads, defined as those containing more than 15% bases with a Phred score below 30 or more than 5% ambiguous bases (N). Reads were retained only if ≥ 85% of bases had a Phred score ≥ 30. The resulting high-quality reads were subjected to QTL-seq analysis using the QTL-seq pipeline v2.2.2 (https://github.com/YuSugihara/QTL-seq). The analysis was performed as follows: First, a PTT1-based pseudo-reference genome was generated by aligning clean PTT1 reads to the publicly available Nipponbare reference genome (IRGSP1.0) using BWA-MEM with default parameters [26], followed by conversion, sorting, and indexing with SAMtools [27]. Variants were called with BCFtools mpileup [28] and filtered with thresholds of minimum read depth ≥ 10 and maximum read depth ≤ 250. The Nipponbare genome was then modified by substituting PTT1-specific variants to create the PTT1 pseudo-reference. Equal numbers of clean reads from each F₂ individual were pooled into high-tiller-number (high-TN) and low-tiller-number (low-TN) bulks. Pooled reads were aligned to the PTT1 pseudo-reference genome, and SNPs and indels were called for each bulk. Only variants detected in both bulks were retained for SNP-index calculation, defined as the proportion of alternative allele reads to total read depth. The Δ(SNP-index) was calculated by subtracting the high-TN SNP-index from the low-TN SNP-index [24]. SNP-index and Δ(SNP-index) values were averaged using a sliding window of 2 Mb with a 10 kb step size, and plotted along the 12 rice chromosomes. Statistical significance was determined using the pipeline’s simulation-based approach, which performs 10,000 Monte Carlo simulations based on bulk sizes and population type to generate 95% (p < 0.05) and 99% (p < 0.01) confidence intervals for each window. Genomic regions where the observed Δ(SNP-index) exceeded these thresholds were considered significantly associated with the trait, with peaks above the 99% interval regarded as the most likely QTL candidates. SNPs within QTL regions were annotated using the Variant Effect Predictor (VEP: https://plants.ensembl.org/Oryza_sativa/Tools/VEP), and homologs of previously characterized genes carrying high- to moderate-impact SNPs were identified as potential candidate genes associated with tiller number.

Marker development and QTL validation

The design of the markers was determined by the preliminary marker-trait association of each sequenced F₂ line (n = 20). SNPs within selected genomic regions were tested for association with phenotype using ANOVA for a single marker as described in [29]. Briefly, F₂ lines were grouped by SNP alleles and then one-way ANOVA was tested for significant differences between group means. The F-test was used to determine the trait-associated SNP. SNPs showing both high to moderate variant impact and high association with tiller number were targeted for the development of allele-specific PCR genotyping assays using the Kompetitive Allele-Specific PCR Genotyping (KASP) system (LGC Genomics) or TaqMan™ Real-Time PCR Assays (Thermo Fisher Scientific). The KASP genotyping test was performed according to [30], and the TaqMan assay was performed according to [31]. A further marker-trait analysis of a larger F₂ population was performed using the genotypes and phenotypic data obtained from 500 progeny. Two-way analysis of variance (ANOVA) was performed using jamovi version 2.6 (https://www.jamovi.org) to evaluate the main and interaction effects between the two QTL genotypes on the tiller number.

Generation of nal1 mutant lines and growing condition

Two nal1 mutant lines (T1 generation), nal1-1 (BG110220C10) and nal1-2 (BG110336H11) generated via the CRISPR-Cas9 system in the japonica rice variety ZhongHua 11 (ZH11) were obtained from Biogle Gene Tech Co, Ltd (Jiangsu, China). The sgRNA sequence for nal1-1 was 5′-GGAATCCCACTGCTGTCCCG-3′, and that for nal1-2 was 5′-GACGATAAGGCGCAGCTCTCCGG-3′. Details of edited sites are available at http://biogle.cn/geo/index/geo/val/BG110220C10 for nal1-1 and http://biogle.cn/geo/index/geo/val/BG110336H11 for nal1-2. Both the nal1 mutants and ZH11 wild-type plants were grown in pots under waterlogged and dry conditions, simulating lowland and upland environments, respectively, in the greenhouse at South China Agricultural University, China.

Results

Tiller number assessment in parental lines and F₂ population

The development of tillers in the parental lines NDCMP49 and PTT1 was observed over time. In the early growth phase (i.e. before 50 days after sowing, DAS), no significant difference in the number of tillers was observed between the two parental varieties. However, from 60 DAS, PTT1 began to show increased tiller production, with 6–7 tillers per plant observed, compared to 3–4 tillers in NDCMP49. This trend continued throughout plant development, with maximum tiller production reached at 80 DAS. At this stage, PTT1 formed about 20 tillers, while NDCMP49 formed about 7 tillers (Fig. 1A–B). The number of tillers was analyzed at 80 DAS in 492 F₂ individuals from the NDCMP49 × PTT1 cross. The trait showed a continuous, approximately normal distribution, ranging from 2 to 41 tillers, with a mean of 18.78 (Fig. 1C). Transgressive segregation was evident as 36.78% of F₂ individuals exceeded the tiller count of the PTT1 parent (TN = 20), while 3.45% produced fewer tillers than NDCMP49 (TN = 7). To construct high and low-tiller number (TN) bulks for QTL-seq analysis, 30 F₂ individuals with the highest and 30 with the lowest tiller numbers were preliminarily selected. These individuals were further developed to generate F_2:3 families in which tillering ability was reassessed. From this set, ten families with the highest and ten with the lowest number of tillers were selected. The corresponding F₂ individuals were then used to form high-TN and low-TN bulks, which were subjected to whole-genome resequencing and QTL-seq analysis (Fig. 1D).

Fig. 1.

Development of tiller numbers in PTT1, NDCMP49 and F₂ populations and QTL analysis workflow. A Appearance at the maximum tillering stage (80 DAS) of NDCMP49, PTT1 and representative F₂ lines selected for QTL-seq analysis. B Development of tiller number in PTT1 and NDCMP49 from 50 to 80 DAS, C Distribution of traits based on the number of tillers of the F₂ population at the maximum tillering stage. The vertical lines indicate the average number of tillers of the two parents. The plants chosen for the high-TN and low-TN bulks are indicated in green and red, respectively. D Schematic diagram of the experiment workflow

Whole-genome re-sequencing and analysis of parents and high/low-tiller F₂ bulks

Whole-genome re-sequencing was performed on 20 selected F₂ lines (consisting of the high-TN and low-TN bulks) and the two parental lines using the MGI-seq platform. The total number of high-quality, cleaned reads of NDCMP49 and PTT1 was 65.55 million and 78.54 million, respectively, corresponding to approximately 23.41-fold and 27.53-fold genome coverage (based on an estimated rice genome size of ~ 400 Mb). Sequencing of individual high-TN and low-TN F₂ lines yielded an average of 126.91 million (between 62.56 and 166.42 million) and 141.57 million (between 86.05 and 190.62 million) paired-end reads, respectively. To create the high-TN bulk, 26.49 million clean reads (corresponding to ~ 10-fold coverage) were randomly selected from each of the 10 high-TN samples. A similar approach was used to construct the low-TN bulk. The clean reads from the selected individuals were pooled to generate high-TN and low-TN bulks for subsequent analysis (Table 1). Alignment of the PTT1 and NDCMP49 reads to the Nipponbare reference genome identified 1,921,313 and 585,671 SNPs, respectively. Variant calling from the pooled reads revealed 959,597 SNPs in the high-TN bulk and 925,060 SNPs in the low-TN bulk (Table S1). A total of 436,481 common SNPs supported by more than eight reads were selected for QTL-seq analysis (Table 2).

Table 1.

Summary of BGI-seq data of parental lines and High-TN and Low-TN bulks

Sample	Cleaned read (million)	Cleaned base (Gb)	%Alignment	Average depth coverage (x)
NDCMP49	65.55	9.83	92.72	23.41
PTT1	78.54	11.69	93.79	27.53
Low-TN bulk	264.9	40	92.95	100
High-TN bulk	264.9	40	92.61	100

Table 2.

Chromosome-wise distribution of common SNPs and indels (read depth > 8) identified in high- and low-tiller number bulks

Chr.	Length	SNPs (Depth > 8)	Indels (Depth > 8)
1	43,270,923	61,521	20,308
2	35,937,250	32,559	10,169
3	36,413,819	51,911	16,096
4	35,502,694	31,763	9418
5	29,958,434	37,464	11,014
6	31,248,787	42,264	12,861
7	29,697,621	34,128	10,184
8	28,443,022	29,889	9167
9	23,012,720	41,349	12,256
10	23,207,287	20,735	6115
11	29,021,106	34,434	9441
12	27,531,856	18,464	5537
Total	373,245,519	436,481	132,566

QTL-seq analysis and identification of genomic regions associated with the tiller number in rice

To identify genomic regions associated with tiller number in rice, SNP-index values for each bulk and the Δ(SNP-index) of 436,481 high-confidence SNPs were calculated and plotted across all 12 rice chromosomes (Fig. 2). Significance thresholds were determined based on a null distribution generated assuming no QTL. Two candidate genomic regions were identified where the average Δ(SNP-index) exceeded the 99% confidence interval, indicating a statistically significant association with the trait (Table 3; Fig. 2). On chromosome 4, the QTL labeled qTN4 was mapped in the interval between 30.7 and 34.3 Mb, with a peak at 33.2 Mb. On chromosome 5, qTN5 was detected in the region between 9.5 and 15.7 Mb, with a peak at 12.6 Mb (Fig. 2).

Fig. 2.

SNP-index and Δ(SNP-index) diagrams from the QTL-seq analysis. A Pseudomolecules of the Nipponbare reference genome (IRGSP 1.0), B Upper probability values at 99% confidence (P < 0.01), C Upper probability values at 95% confidence (P < 0.05), D The sliding window plots of ∆(SNP-index), with a window size of 2-Mb and 10-kb increments, E Lower probability values at 95% confidence (P < 0.05), F Lower probability values at 99% confidence (P < 0.01), G Sliding window plots of the average SNP index values in low-bulk, with a 2-Mb window size and 10-kb increments, H The sliding window plots of the average SNP-index values in high-bulk, with a 2-Mb window size and 10-kb increments, I Candidate genomic regions containing QTLs for tiller number

Table 3.

Summary of illumina sequencing data of parental lines and F₂ bulks

QTL	Chr	QTL Region (Mb)	Peak SNP (Mb)	High-TN bulk SNP index	Low-TN bulk SNP index	Confidence interval (95%)	Confident interval (99%)	Delta (SNP index) of peak SNP
qTN4	4	30.7–34.3	33.2	0.17–0.22	0.65–0.70	0.33	0.42–0.43	0.52
qTN5	5	9.5–15.7	12.6	0.16–0.31	0.58–0.79	0.32	0.42	0.49

Candidate gene annotation and validation

Candidate genes within the qTN4 and qTN5 regions were annotated using the RAP-DB database. Genes that had both annotated biological functions and functional SNPs were prioritized, resulting in the identification of 175 genes in the qTN4 region and 30 genes in the qTN5 region (Table S2). Based on their known involvement in the regulation of lateral leaf growth and panicle development, NARROW LEAF1 (NAL1; Os04g0615000) was selected as a candidate gene for qTN4 and and OVATE protein gene, OsOFP19 (Os05g0324600), as a candidate for qTN5 (Fig. 3). Three non-synonymous SNPs were identified within NAL1. A G-to-A substitution at position 4:31,212,801 in exon 3 resulted in an amino acid change from arginine to histidine (R-to-H). Two additional SNPs were detected in exon 5: a C-to-T substitution at 4:31,214,019, resulting in a change from alanine to valine (A-to-V), and a G-to-A substitution at 4:31,214,045, resulting in a change from valine to isoleucine (V-to-I) (Fig. 4A). In OsOFP19, a non-synonymous G-to-A substitution was identified at position 5:15,069,810 in exon 1, resulting in an amino acid change from methionine to valine (M-to-V) (Fig. 4B).

Fig. 3.

SNP-index plots and candidate genes on chromosomes 4 and 5. A and B show the SNP-index plots for low-TN bulk (top, red dots), high-TN bulk (middle, green dots) and Δ(SNP-index) values (bottom, cyan dots) along chromosomes 4 and 5, respectively. Gray shaded areas indicate genomic regions identified as QTLs. Annotated genes within these regions are shown as shaded rectangles, with arrowheads indicating the direction of the open reading frame (ORF). The candidate genes NAL1 and OsOFP19 are highlighted by red shaded rectangles

Fig. 4.

Gene structures of NAL1 and OsOFP19 and polymorphisms in the parental lines and selected F₂ individuals. A The structure of the NAL1 gene shows the exon–intron organization and the positions of three non-synonymous SNPs that distinguish PTT1 and NDCMP49. The amino acid substitutions resulting from each SNP are indicated. B Structure of the OsOFP19 gene showing the position of a non-synonymous SNP identified between the two parents along with the corresponding amino acid change

We also analyzed the genotypes of the two candidate genes in the 20 individual F₂ lines used for the QTL-seq analysis. For NAL1, F₂ lines carrying the homozygous genotypes of NDCMP49 (A/A at position 4:31,212,801; T/T at 4:31,214,019; and A/A at 4:31,214,045) had a significantly lower number of tillers (TN = 11.60 ± 0.72, p < 0.01), while lines with the homozygous PTT1 genotype (G/G, C/C and G/G at the corresponding positions) had a higher number of tillers (TN = 24.15 ± 2.45) (Fig. 5). For OsOFP19, the F₂ lines with the homozygous NDCMP49 genotype (A/A at position 5:15,069,810) also had a significantly lower number of tillers (TN = 11.84 ± 1.01), while the lines with the homozygous PTT1 genotype (G/G) had a significantly higher value (TN = 22.90 ± 0.92, p < 0.01). Of note, the F₂ lines heterozygous for NAL1 had a comparable number of tillers to the NDCMP49 homozygotes (TN = 11.67 ± 1.01, no significant difference), suggesting a recessive effect. In contrast, the heterozygous lines for OsOFP19 had a wide range of tiller numbers (TN = 10.12–27.94), including both high and low values, suggesting possible incomplete dominance or variable expressivity.

Fig. 5.

Association between sequenced SNP genotypes of NAL1 and OsOFP19 and tiller number in selected F₂ individuals used for QTL-seq analysis. A–C Tiller number distribution among NAL1 genotypes based on SNPs at positions 4:31,212,801, 4:31,214,019 and 4:31,214,045. D Tiller number distribution among OsOFP19 genotypes based the SNP at position 5:15,069,810. Violin plots show the distribution of tiller number across genotype classes, overlaid with boxplots and individual data points (jittered). Genotype classes for the SNPs at positions 4:31,212,801, 4:31,214,019, and 5:15,069,810 are labeled as G/G = homozygous PTT1 allele, A/A = homozygous NDCMP49 allele, and A/G = heterozygous, while genotype classes for 4:31,214,019 are labeled as C/C = homozygous PTT1 allele, T/T = homozygous NDCMP49 allele, and T/C = heterozygous

Marker-trait association analysis for NAL1 and OsOFP19

To evaluate the association between SNP genotypes in the candidate genes NAL1 and OsOFP19 and tiller number, 500 F₂ individuals from a cross between NDCMP49 and PTT1 were genotyped using a KASP marker (NAL1_4_31214045) and a TaqMan probe marker (OFP19_5_15069810), developed from G/A SNPs located in exon 3 of NAL1 (4:31,214,045) and exon 1 of OsOFP19 (5:15,069,810), respectively (Table S3). Single marker analysis revealed that NAL1_4_31214045 had a highly significant association with tiller number (p < 2.2 × 10^− 16) and explained 22.07% of the phenotypic variance (PVE), while OFP19_5_15069810 showed a weaker but still significant association (p = 1.99 × 10^− 5), with a PVE of 3.65% (Table 5; Fig. 6A, B).

Table 5.

Analysis of single and multiple markers in the F₂ population

Marker	Tested Sample	Chr.	Position of SNPs	p-Value	PVE (%)
Single marker
NAL1_4_31214045	439	4	31,214,045	< 2.2E-16	22.07
OFP19_5_15069810	498	5	15,069,810	1.99E-05	3.65
Multiple markers
Tested sample	436

	Estimate	Std. Error	t value	Pr(>|t|)	PVE (%)
(Intercept)	12.9235	1.1781	10.97	< 2e-16 ***	24.13
NAL1_4_31214045	4.201	0.3853	10.903	< 2e-16 ***
OFP19_5_15069810	-1.3884	0.384	-3.615	0.000336 ***

Fig. 6.

Marker–trait association analysis of NAL1 and OsOFP19 SNP genotypes with tiller number in the F₂ population derived from the NDCMP49 × PTT1 cross. A Single-marker analysis of the KASP marker NAL1_4_31214045, based on a G/A SNP in exon 3 of NAL1. B Single-marker analysis of OFP19_5_15069810, based on a G/A SNP in exon 1 of OsOFP19. C Two-marker regression analysis indicates an additive effect between the PTT1 alleles at NAL1 and OsOFP19. In A and B, genotypes are shown as G/G: homozygous for the PTT1 allele, A/A: homozygous for the NDCMP49 allele, and G/A: heterozygous. In C, combined genotypes across the two loci are represented in the format [NAL1 genotype]/[OsOFP19 genotype]. For example, GG/GG: homozygous PTT1 at both loci, GG/AA: homozygous PTT1 at NAL1 and homozygous NDCMP49 at OsPFP19, AA/GG: homozygous NDCMP49 at NAL1 and homozygous PTT1 at OsOFP19, AA/AA: homozygous NDCMP49 at both loci, and AG/AG: heterozygous at both loci

To further evaluate the combined effect of the two markers, a multiple marker regression analysis was performed. The model showed that both markers contributed significantly to the variation in tiller number, with NAL1_4_31214045 showing a strong positive effect (estimate = 4.201, p < 2 × 10^− 16) and OFP19_5_15069810 showing a modest but significant negative effect (estimate = − 1.3884, p = 0.000336). The overall model explained 24.13% of the phenotypic variance (Table 5; Fig. 6C). The results confirm that the PTT1 allele in NAL1 contributes significantly to the increase in tiller number, while the PTT1 allele in OsOFP19 exerts an additive effect, albeit smaller. To test for potential epistatic interactions between the two QTLs, a two-way analysis of variance (ANOVA) was conducted. The analysis showed significant main effects of both NAL1 and OsOFP19 on tiller number, but no significant interaction between the loci (F(_{4, 427}) = 0.57, p = 0.68; Table S4), indicating that their effects are additive rather than epistatic. These results suggest that NAL1 plays an important role in regulating tiller number in this population and that the effect of OsOFP19 may be context-dependent or involve interaction with other loci.

Validation of NAL1 function in regulating tillering ability

CRISPR-Cas9-edited lines targeting NAL1 were studied to determine the gene’s role in regulating tiller number. The nal1-1 line contained a 19-bp deletion, and nal1-2 carried a 10-bp deletion within the NAL1 coding sequence (exon 1). Both mutations are predicted to cause frameshift disruptions. At the maximum tillering stage, the edited lines produced a greater number of tillers than the wild type Zhonghua 11 (ZH11; a lowland Japonica variety). The average number of tillers was 6.75 in nal1-1, 5.75 in nal1-2 and 2.75 in ZH11 (Fig. 7). Statistical analysis confirmed significant differences in tiller number among the three genotypes (p < 0.05, Duncan’s test). The increased number of tillers in the mutant lines was evident in the early vegetation stage and remained consistent through 49 days after sowing (Fig. 7B). The increased number of tillers and narrow leaf width were consistently observed in the mutant lines grown under dry condition, simulating upland condition (Figure S1). These phenotypic results observed in the edited lines are consistent with QTL-seq and marker-trait association results implicating NAL1 in the control of tillering in rice. In addition to increased tillering, both mutants exhibited a narrow leaf phenotype, with average leaf widths of 0.38 cm (nal1-1) and 0.39 cm (nal1-2), compared with 0.92 cm in the wild-type background (Fig. 7A). These findings confirm that loss-of-function mutations in NAL1 promote increased tillering and reduced leaf width, supporting its pleiotropic role in rice plant architecture.

Fig. 7.

Whole plant phenotype of ZH11 (WT) and NAL1 lines at the grain-filling stage under anaerobic condition (A). Tiller number of ZH11 (WT), nal1-1 and nal1-2 lines from the 0th day to the 49th day after sowing (B). Comparison of the total tiller number per plant (C). Data in C is presented as means ± SE (n = 4). Different lower-case letters denote significant differences between lines according to Duncan’s test (p < 0.05). Scale bar is 5 cm. in (A)

Discussion

To improve the performance of upland rice in favorable lowland ecosystems, it is essential to decipher the genetic basis of important agronomic traits such as tillering, which directly influence yield potential under high-input conditions [32]. Numerous quantitative trait loci (QTLs) associated with tiller number have been previously identified using traditional and molecular mapping approaches [33–35]. In recent years, next-generation sequencing (NGS)-based bulked segregant analysis (BSA), i.e. QTL-seq, has emerged as an effective tool to rapidly identify genomic regions associated with phenotypic variation, especially in populations descended from parents with contrasting phenotypes [36].

In this study, we used QTL-seq to analyze an F₂ population derived from a cross between the high tillering lowland indica cultivar PTT1 and the low tillering upland tropical japonica line NDCMP49. QTL-seq analysis identified two genomic regions on chromosomes 4 and 5 that were significantly associated with variation in tiller number. Within these regions, two candidate genes, NAL1 (Os04g0615000) and OsOFP19 (Os05g0324600), were identified as likely contributing to the observed phenotypic differences. Both genes have previously been associated with plant architecture traits. NAL1 has been extensively studied for its role in regulating leaf width via auxin-mediated acid growth mechanisms and cell expansion [37]. The pleiotropic role of NAL1 has also been associated with vascular bundle formation, vein patterning and cell division [38], as well as physiological traits such as chlorophyll content and photosynthetic efficiency [39]. Favorable alleles of NAL1 from japonica germplasm have been introgressed into indica backgrounds to improve yield traits, including spikelet number, root architecture, and vascular development [40, 41]. Although NAL1 has traditionally been associated with narrow leaf morphology, a growing body of evidence, including our results, suggests that it also plays a broader role in shoot architecture and yield-related traits [42]. Our results, derived from two independent nal1 alleles (nal1-1 and nal1-2) generated via CRISPR/Cas9 in the ZH11 wild-type background, provide strong genetic evidence consistent with the findings of Jiang et al. [21], who reported similar phenotypes in a mutant generated by random mutation in the Nipponbare background. Consistently, our evaluation under dry (upland) conditions also showed that nal1 mutants produced more tillers than wild-type plants, indicating that the negative regulatory role of NAL1 in tiller production is maintained across contrasting water regimes. Together, these studies confirm that loss of NAL1 function leads to increased tiller number, supporting its role as a negative regulator of tiller production in rice.

Within the qTN5 interval, we identified 30 annotated genes, among which OsOFP19 emerged as the most compelling candidate based on our QTL mapping results and prior genetic reports. While additional fine mapping and functional assays are needed to confirm causality, our discovery of previously unreported, naturally occurring missense mutations in OsOFP19 provides novel evidence for its functional importance. The consistent association of OsOFP19 with tiller number in the segregating population derived from two genetically distant parental lines highlights its robustness across genetic backgrounds, positioning it as a priority target for future functional studies and rice breeding programs. OsOFP19 was functionally characterized as part of a protein complex with OSH1 (Oryza sativa homeobox 1) and DLT (DWARF AND LOW-TILLERING) that integrates the brassinosteroid (BR) signaling and developmental pathways to modulate plant architecture [43]. Considering that tiller formation is associated with axillary meristem initiation and bud outgrowth, the involvement of OsOFP19 in hormonal regulation emphasizes its functional importance.

To validate the association between these loci and tiller number, we developed an allele-specific KASP marker targeting a functional SNP in NAL1 and a TaqMan probe marker targeting a functional SNP in OsOFP19. Marker trait analysis in the F₂ population showed that the NAL1 marker explained a substantial proportion of the phenotypic variance (PVE = 22.07%), while the OsOFP19 marker contributed a smaller but significant effect (PVE = 3.65%). When both markers were analyzed together, an additive effect was observed, with the PTT1 allele in OsOFP19 enhancing the effect of the PTT1 allele in NAL1. This suggests that pyramiding these alleles could be beneficial for breeding high-tillering lines. Additive effects between QTLs have long been recognized as important for the improvement of complex traits, as they provide the opportunity for cumulative genetic gains across loci [44]. The development of robust KASP and TaqMan Probe markers linked to these QTLs provides a practical and scalable tool for marker-assisted selection (MAS). KASP and TaqMan genotyping is known for its specificity, cost-effectiveness and usefulness in large-scale breeding pipelines [45, 46], making it well suited for improving upland rice varieties to enhance their performance in waterlogged lowland environments.

In summary, our study identified NAL1 and OsOFP19 as important genes associated with variation in tiller number in rice. The functional significance of NAL1 was supported by CRISPR-Cas9 knockout lines that showed significantly increased number of tiller and narrow leaf phenotype. These results provide new insights into the genetic control of tillering in rice and offer molecular tools to improve shoot architecture in future breeding programs. Further studies of the molecular mechanisms underlying the function of NAL1 and OsOFP19 are essential to fully elucidate their role in meristem regulation and yield optimization.

Conclusion

In this study, two major QTLs, qTN4 and qTN5, associated with tiller number in rice were successfully identified by QTL-seq analysis. These loci, located on chromosomes 4 and 5, were used to develop functional KASP or TaqMan Probe markers that effectively track genotype segregation in the F₂ population. The validated markers exhibited a strong association with the trait and showed an additive effect when combined. These results highlight the potential application of NAL1- and OsOFP19-based markers in marker-assisted selection (MAS) and provide valuable tools to improve tillering capacity and increase productivity of upland rice varieties in future breeding programs.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

BSA

Bulk segregant analysis

CRISPR-Cas9

Clustered regularly interspaced short palindromic repeats and CRISPR-associated protein 9

DAS

Days after sowing

InDel

Insertion/Deletion

KASP

Kompettitive Allele-Specific PCR

MAS

Marker-assisted selection

NAL1

Narrow leaf 1

NDCMP49

Niaw Dam Chaw Mai Pai 49

OFP19

Ovate family protein 19

PTT1

Pathum Thani 1

QTL-seq

Quantitative trait loci-sequencing

SNP

Single nucleotide polymorphism

TN

Tiller number

Author contributions

S.A., T.Toojinda, J.L.S. and S.W. conceived the study. S.A., M.S., V.R., R.X. and S.W. designed the experiments. T.R., N.P., T.Thianthavon, R.D. and P.M. performed experiments. T.R., N.P. and W.A. analyzed data. T.R., N.P., and M.K.P. wrote the manuscript. S.A., S.W., J.E.G. reviewed the manuscript. All authors read and approved the final manuscript.

Funding

This work was partially financially supported by the Kasetsart University Research and Development Institute, KURDI (Grant No. FF (KU) 55.67), the Office of the Ministry of Higher Education, Science, Research and Innovation, and the Thailand Science Research and Innovation through the Kasetsart University Reinventing University Program 2024, the National Research Council of Thailand (NRCT Grant Number: NRCT-RTA/812/2563), the Thailand Rice Science Research Hub of Knowledge (NRCT Grant Number: N35E680080), and the National Science, Research and Innovation Fund, Thailand Science Research and innovation (TSRI Grant Number: FFB680075/0337) through the National Science and Technology Development Agency (NSTDA).

Data availability

The datasets generated and analyzed in this study are available in the GenBank of NCBI with the project accession of PRJNA1287108. Submitted data will remain private until related manuscript has been accepted. All data generated or analyzed during this study are included in this published article and its supplementary information files.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.