Introduction

Rice (Oryza sativa L.) is a staple food crop consumed by more than 3.5 billion people worldwide, and its stable production is essential to meet the demands of a growing global population1. Plant height is a crucial agronomic trait affecting rice yield2,3,4. While overly short plants may limit biomass accumulation, excessively tall plants are prone to lodging. Thus, it is essential to optimize plant height for specific cultivation practices for maximizing rice productivity3,4,5. In rice, lodging occurs mainly during the maturation stage, when the structural support of the above and under-ground parts weakens with the increased grain weight disrupting the balance6,7,8. In Korea, frequent typhoons during this period have caused significant lodging damage, leading to a reduction in both rice yield and quality, as well as a decrease in harvesting efficiency6,8.

Numerous efforts have been made to overcome the yield losses caused by lodging. The development of semi-dwarf rice varieties by introgressing the semi-dwarf1 (sd1) gene reduced lodging under high-fertilizer conditions and increased global rice yields by approximately 20–30%, achieving the widely known ‘Green Revolution’3,9,10. However, as relying on a single gene reduces genetic diversity, research efforts have been made to identify alternative dwarfing genes11. To date, more than 60 recessive dwarf and semi-dwarf mutants, in addition to sd1, have been identified in rice. However, most of these mutants exhibit excessive dwarfism or negatively affect yield components, limiting their use in breeding programs10,12. Therefore, it is essential to explore new breeding materials that can reduce plant height in rice without negatively affecting yield or grain quality.

Recent advances in gene-editing technologies have made it possible to precisely edit target genes13,14. In particular, CRISPR (clustered regularly interspaced short palindromic repeats)/Cas9 (CRISPR-associated protein 9) is a simpler and more efficient technology for generating targeted mutations compared to the previously available technologies such as ZFN (zinc finger nuclease) and TALEN (transcription activator-like effector nuclease)15,16. Using CRISPR/Cas9-mediated mutagenesis of the SD1 gene, semi-dwarf rice lines in an elite variety background have been developed17,18,19. In addition, many studies have focused on regulating plant height by editing different target genes using CRISPR/Cas9. For example, rice lines edited for OsFBK4 exhibited a semi-dwarf phenotype due to reduced internode cell length20. Similarly, miR5504-edited lines exhibited reduced cell expansion and division, resulting in short plant height and panicle length, while exhibiting increased tillering21. Also, the SD8 gene-edited lines showed optimized rice architecture by reducing flag-leaf angle and plant height without yield penalty, suggesting the potential for increased yield under high-density planting conditions22.

MORE AXILLARY GROWTH2 (MAX2), an F-box leucine-rich protein, is known as DWARF3 (D3) in rice and serves as a key component of the SKP1-CUL1-F-box protein (SCF)-type ubiquitin ligase (SCFD3/MAX2) complex, which is involved in the signaling of strigolactones (SLs) and karrikins (KARs)23,24,25,26. SUPPRESSOR OF MAX2 1 (SMAX1)/SMAX1-LIKE (SMXLs) proteins are degraded by the receptor-SCFD3/MAX2 complex during the signaling of SLs and KARs, triggering various biological processes27,28,29. In Arabidopsis, eight SMXL family genes, including SMAX1, have been identified, and nine SMXL family genes have been identified in rice26. However, SMXL4/SMXL5 are not involved in SL/KAR signaling but function as key regulators of primary phloem formation30. In Arabidopsis, HEAT SHOCK PROTEIN-RELATED (AtHSPR) encodes a nuclear-localized protein with ATPase activity and is also known as SMXL431,32. AtHSPR/SMXL4 plays a positive role in the GA- and light intensity-dependent regulation of seed set and flowering and enhances salt tolerance31,33,34. However, the function of SMXL genes is largely unknown in rice.

Samkwang is a Korean japonica rice variety with high yield and good grain quality35,36. However, with an average culm length of 87 cm under standard cultivation in regional adaptability trials, Samkwang is relatively tall and highly susceptible to lodging during the ripening stage, particularly due to the increasing frequency of typhoons36. Although efforts have been made to overcome these limitations, conventional breeding approaches encounter challenges in preserving the genetic background of Samkwang. Therefore, it is necessary to obtain breeding materials that reduce lodging while maintaining the elite genetic background of Samkwang. In this study, we screened a CRISPR/Cas9-mediated rice mutant population in the genetic background of Samkwang and identified an SMXL4-edited line (designated as smxl4) with reduced plant height. This line was phenotypically analyzed for major agronomic traits and yield to assess its potential as a breeding material. Furthermore, transcriptome analysis was conducted to investigate the molecular function of the SMXL4 gene during internode elongation.

Results

Screening of smxl4 from a CRISPR/Cas9-mediated rice mutant population

From a large-scale CRISPR/Cas9-mediated rice mutant population developed at NIAS, RDA37, we obtained a subset composed of 104 T1 lines targeting 45 different genes (mainly transcription factors) in the background of Samkwang. This population was grown in the field for visual screening and analyzed for major agronomic traits including days to heading (DTH), culm length (CL), panicle length (PL) and panicle number (PN) (Fig. S1). As a potential breeding material for improving lodging resistance of Samkwang, we selected SMXL4-edited lines (hereafter referred to as smxl4) exhibiting short stature (CL of 70.2 cm, 17.7 cm shorter than Samkwang) with stable growth (Fig. S2). SMXL4 (Os04g0298700), located on chromosome 4, consists of three exons and two introns, with a coding sequence (CDS) length of 3,195 bp encoding 1,064 amino acids (Fig. 1A). Polymerase chain reaction (PCR) and sequencing of the target gRNA region from smxl4 and Samkwang verified that smxl4 carries a 1 bp insertion at the 861st nucleotide of the first exon of SMXL4, resulting in a frameshift mutation (Fig. 1B). Cas9-free screening was conducted among the T1 sister lines of smxl4, and we selected one Cas9-free line (smxl4-2-2) used for further experiments (Fig. 1C, Fig. S3). In addition, Sanger sequencing was carried out for the top eight candidate sites selected from 315 potential off-target sites predicted by Cas-OFFinder38, and no off-target mutations were detected at any of the examined sites (Table S1, Fig. S4).

Fig. 1
Fig. 1The alternative text for this image may have been generated using AI.
Full size image

Verification of SMXL4 gene editing. (A) Schematic diagram of the gene structure and CRISPR-Cas9 meditated gene editing target sequence in SMXL4. (B) DNA sequence analysis of SMXL4 from Samkwang (WT) and smxl4. (C) Cas9-free screening of smxl4 plants (Unlabeled lanes are samples irrelevant to this study; the full-length agarose gel image is provided in Fig. S3).

Phenotypic evaluation of smxl4

In the field experiment arranged as a randomized complete block design with three blocks (see Methods section for details), the plant height of Samkwang and smxl4 T2 plants were measured every two weeks from 39 days after transplanting (DAT) to 79 DAT. The plant height of smxl4 was significantly shorter than that of Samkwang throughout the growth period, except at the panicle initiation stage (Fig. 2A, B, Table S2). Compared with Samkwang, the first, second, and third internodes of smxl4 were significantly reduced by 15%, 8%, and 19%, respectively, while the fourth and fifth internodes showed no significant differences (Fig. 2C, D, Table S3). Panicle exsertion of smxl4 was also reduced by 85% (Fig. 2E).

Fig. 2
Fig. 2The alternative text for this image may have been generated using AI.
Full size image

Phenotypic comparison of Samkwang (WT) and smxl4 plants. (A) Phenotype of wild-type (WT) and smxl4 plants at heading stage (79 DAT). Bar= 10 cm. (B) Average plant height of WT and smxl4 plants during growth. DAT: days after transplanting. (C), (D) Internode morphology and average internode length in WT and smxl4. Bar= 10 cm. (E) Panicle exsertion phenotype of WT and smxl4. Bar= 5 cm (*P < 0.05, nsnot significant).

To evaluate the potential of smxl4 as breeding material, we analyzed its agronomic and yield-related traits in comparison with Samkwang (Fig. 3, Table S3). The mean DTH of Samkwang and smxl4 were 75.3 and 76.0 days, respectively, with no significant difference (Fig. 3A). The CL and PL of smxl4 (69.6 cm, 15.5 cm) were 16% and 20% shorter than those of Samkwang (83.3 cm, 19.4 cm), respectively (Fig. 3B, C). However, PN of smxl4 (14.3) was 22% greater than that of Samkwang (11.7) (Fig. 3D). The grain number per panicle (GNP) and 1000-grain weight (TGW) of smxl4 were 112.8 and 21.1 g, respectively, representing a 17% and 14% reduction compared to Samkwang (136.7 and 24.5 g, respectively) (Fig. 3E, F). The assessment of grain yield per plot (YPP, hulled rice harvested from 40 plants per plot) showed that the YPP of smxl4 (1,228 g) decreased by 22% compared to Samkwang (1,582 g) (Fig. 3H). Taken together, these results showed that the semi-dwarf phenotype of smxl4 is likely accompanied by a significant yield reduction under standard field conditions. Grain size evaluation indicated no significant differences in grain area, length, and width between Samkwang and smxl4 (Fig. 4A–D). We also measured the seed germination rate and seedling growth in both Samkwang and smxl4. There was no significant difference in the initial germination rates between Samkwang and smxl4 on day 1 and day 2 (Fig. 4E). However, on day 3 and day 4, the seed germination rates of smxl4 were 78% and 84%, respectively, which were significantly lower than those of Samkwang at 95.3% and 99.0%, respectively. Seedlings grown for two weeks showed that the root length of smxl4 was significantly shorter than that of Samkwang, without significant difference in shoot length (Fig. 4F, G).

Fig. 3
Fig. 3The alternative text for this image may have been generated using AI.
Full size image

Analysis of major agronomic and yield-related traits in Samkwang (WT) and smxl4. Comparison of days to heading (A), culm length (B), panicle length (C), panicle no. per plant (D), grain no. per plant (E), 1000-grain weight (F), spikelet fertility (G), yield per plot (rough rice weight harvested from 40 plants) (H). Data are means ± SE (**P < 0.01, *P < 0.05, nsnot significant).

Fig. 4
Fig. 4The alternative text for this image may have been generated using AI.
Full size image

Seed and seedling traits of Samkwang (WT) and smxl4. (A)-(D) Comparison of grain phenotypes between WT and smxl4. Bar=5 mm (E) Comparison of seed germination rate in WT and smxl4. (F), (G) Root and shoot lengths of 2-week-old WT and smxl4 seedlings. Bar= 1 cm. Data are means ± SE (***P < 0.001, **P < 0.01, *P < 0.05, nsnot significant).

To validate the genetic effect of SMXL4 editing, agronomic traits were further analyzed in an F2 population derived from a cross between smxl4 and the non-mutagenized cultivar Samkwang. Consistent with observations in smxl4 T2 plants relative to Samkwang (Fig. 4), CL, PL, GNP, and TGW were significantly reduced in plants carrying the homozygous smxl4 mutation compared with those carrying the homozygous wild-type allele (Table S4). However, because the F2 population was grown under late-planting conditions (transplanting on June 27, 2025) with a shorter growth duration, the magnitude of reduction in these traits associated with the smxl4 mutation was smaller than that observed in T2 plants grown under regular planting conditions (transplanting on June 5, 2023). Taken together, these results indicate that the smxl4 mutation contributes to reductions in plant height and yield components.

RNA-seq analysis of smxl4

To elucidate the molecular basis controlling stem elongation in smxl4 relative to Samkwang, RNA-seq was performed on the second internode at the booting stage and the uppermost internode at the heading stage. We constructed and sequenced 12 RNA-seq libraries from three biological replicates at each growth stage for two genotypes. One sample was excluded due to poor sequencing quality, resulting in 11 libraries for downstream analysis. PE-150 sequencing generated an average of 53.8 million reads per sample (Table S5). After adaptor trimming and quality filtering, an average of 53.7 million clean reads per library was obtained. The Q20 value was above 98.3% for all libraries, while the Q30 value was above 94.1%. On average, 59.1% of the reads were uniquely mapped, and 91.4% were pseudoaligned. These results indicated that the data were suitable for downstream analysis. Principal component analysis (PCA) showed that PC1 accounted for 77% of the variance, clearly separating samples by growth stage (booting and heading) (Fig. 5A). However, Samkwang and smxl4 were not distinctly separated, suggesting that transcriptional differences between the genotypes were relatively minor compared to those driven by growth stage variation. Sample-to-sample clustering analysis further supported the PCA results (Fig. 5B).

Fig. 5
Fig. 5The alternative text for this image may have been generated using AI.
Full size image

Transcriptome analysis of smxl4 relative to Samkwang (WT). (A) Principal component analysis (PCA); (B) Distance heatmap of RNA-seq samples from WT and smxl4. (C-D) Volcano plots for differentially expressed genes (DEGs) in smxl4 relative to WT at booting stage (BT) and heading stage (HD); (E) Venn diagrams of the DEGs in smxl4 relative to WT at BT and HD.

For DEG analysis, the clean reads were mapped and quantified using Kallisto39, followed by analysis with DESeq240. Significant DEGs were identified based on padj < 0.05 and |log2foldchange| ≥2. A total of 84 DEGs were identified at the booting stage, of which 77 were upregulated and 7 were downregulated in smxl4 relative to Samkwang (Fig. 5C). At the heading stage, 120 DEGs were identified, including 94 upregulated and 26 downregulated genes in smxl4 relative to Samkwang (Fig. 5D). Among these, 51 genes were commonly upregulated at both stages, whereas only one gene was consistently downregulated (Fig. 5E; Table 1). Detailed DEG lists for each stage are summarized in Tables S6 and S7. GO enrichment analysis of the commonly upregulated DEGs at both stages revealed enrichment for terms related to cell wall synthesis, methylation, protein modification, and degradation (Fig. 6A, Table S8). Specifically, three genes involved in cell wall synthesis and degradation were commonly upregulated in smxl4 at both stages. These included genes encoding the beta subunit of polygalacturonase 1 (BURP16; Os10g0409400), which degrades pectin41; beta-1,3-glucanase 11 (GNS11; Os07g0539100), an enzyme that hydrolyzes cell wall polysaccharides42; and cellulose synthase-like protein D3 (CSLD3; Os08g0345500).

Table 1 Selected differentially expressed genes in smxl4 relative to Samkwang at the booting stage and heading stage.
Full size table
Fig. 6
Fig. 6The alternative text for this image may have been generated using AI.
Full size image

Gene ontology (GO) enrichment analysis of up-regulated and down-regulated genes in smxl4 relative to Samkwang at booting stage (BT) and heading stage (HD). (A) Commonly up-regulated genes at both BT and HD. (B) Down-regulated genes at BT. (C) Up-regulated genes and (D) down-regulated genes at HD.

Additionally, genes encoding defense-related proteins were also upregulated in smxl4, including genes encoding PID3 (Os06g0330100), a blast resistance protein43; and JA upregulated protein 1 (Os07g0442800), which regulates jasmonic acid biosynthesis and signaling44. Furthermore, genes encoding a major facilitator superfamily (MFS) protein (Os08g0409900), a UDP-glucuronosyl/UDP-glucosyltransferase family protein (Os05g0527700), which is involved in plant hormone glycosylation45,46, and S-adenosyl-L-methionine synthetase (OsSAMS3; Os01g0293000), which is involved in regulating plant development via histone and DNA methylation47, were upregulated. In contrast, only one gene, OsSPX2 (Os02g0202200), involved in phosphate (Pi) sensing and Pi starvation responses48, was commonly downregulated at both stages.

Booting stage

Among the 32 DEGs uniquely identified at booting stage, 26 were upregulated, while six were downregulated in smxl4 (Table 2). Upregulated DEGs associated with cell wall metabolism included genes encoding PECTIN METHYLESTERASE 17 (OsPME17; Os05g0361500) for pectin degradation, and dirigent protein (DIR; Os07g0643800), which is involved in lignin biosynthesis pathway49. Three genes associated with defense mechanisms were also upregulated, including those encoding pathogenesis-related protein 5 (PR5; Os05g0538500)50, chitinase 11 (CT11; Os03g0132900)51 and lectin receptor-type kinase (Os07g0575600)52. Genes involved in transport included those encoding PHOSPHATE TRANSPORTER 11 (PT11; Os01g0657100) and peptide transporter (PTR2; Os04g0441800). Downregulated genes were significantly enriched in GO terms related to lipid metabolism such as lipid binding, lipid transport, and lipid metabolic process. Among these, DEGs included a gene encoding nonspecific lipid-transfer protein (LTP) (Fig. 6B, Table S8). Additionally, a gene encoding arabinogalactan protein (AGP; Os03g0245300), which play roles in cell expansion53,54,55 and stem elongation56, was also downregulated.

Table 2 Selected differentially expressed genes in smxl4 relative to Samkwang at the booting stage.
Full size table

Heading stage

Among the 78 DEGs uniquely identified at the heading stage, 43 were upregulated, whereas 25 were downregulated in smxl4 plants. The GO enrichment analysis of upregulated genes showed significant enrichment in terpenoid biosynthetic process and response to fungus within the biological process, and terpene synthase activity and magnesium ion binding in the molecular function category (Fig. 6C, Table S8). In addition to OsTPS4, which was commonly upregulated, genes encoding OsTPS20, OsTPS41, OsTPS29, and a cytochrome P450 family protein involved in terpene biosynthesis57 were also upregulated at the heading stage. Two genes encoding caffeic acid O-methyltransferase (COMT), a key enzyme in the phenylpropanoid metabolism pathway of lignin biosynthesis58, were upregulated. Genes encoding germin-like protein (GLP; Os08g0189200), which confer broad-spectrum disease resistance59, and OsLOX8 (Os08g0509100), a member of the lipoxygenase family involved in plant immunity60, were also upregulated (Table 3).

Table 3 Selected differentially expressed genes in smxl4 relative to Samkwang at the heading stage.
Full size table

Enriched GO terms in downregulated DEGs at the heading stage included two auxin-related terms within the biological process (Fig. 6D, Table S8). The downregulated DEGs included genes encoding SHORT INTERNODES1 (OsSHI1), which suppresses the transcriptional activity of IPA1 to regulate plant architecture and integrate multiple hormone signaling pathways61,62. Additionally, a gene encoding SHORT INTERNODES-related sequences63 (SRSs; Os06g0712600), which play crucial roles in plant growth, development, and stress regulation, was also downregulated (Table 3).

Discussion

Plant height is a crucial agronomic trait that directly influences lodging tolerance and crop yield4. While the widespread cultivation of semi-dwarf varieties has contributed to improving yield, limited genetic diversity may increase susceptibility to biotic and abiotic stresses11,64. In this study, we leveraged a previously developed large-scale CRISPR/Cas9-edited rice population in the Samkwang background37 to screen semi-dwarf breeding materials. The SMXL4-edited line exhibited reduced internode length, resulting in decreased plant height, and was accompanied by pronounced changes in other agronomic traits and yield components (Figs. 2 and 3). Genetic analyses in the F2 population further confirmed that modification of SMXL4 was associated with reduced plant height and yield related traits, supporting a role of SMXL4 in rice growth and development. However, lodging tolerance is determined not only by plant height but also by morphological, anatomical, and biochemical properties of culms, including the number and area of vascular bundles, culm thickness and diameter, and the levels of lignin, cellulose, and hemicellulose65,66. As these traits were not evaluated in the present study, the potential of smxl4 as a lodging-tolerant breeding material remains to be determined. Moreover, the significant yield penalties observed in smxl4 suggest that the direct use of smxl4 in breeding programs should be approached with caution.

In Arabidopsis, the SMXL gene family is classified into four subclades based on phylogeny and function27. Although these genes are found throughout seed plants, their physiological functions remain unclear in species other than Arabidopsis28,30,67. SMXL4 and SMXL5 have been demonstrated to play crucial roles in phloem formation in Arabidopsis30,68. The RNA-binding protein JULGI (JUL) binds to the G-quadruplex in the 5′ UTR of SMXL4/5 mRNAs, directly suppressing their translation and consequently inhibiting phloem formation. JUL deficiency leads to an increase in phloem cell number and enhanced seed size and weight69,70. In tomato, knockdown of SlJUL, a translational repressor of SlSMXL5, enhanced phloem development and transport efficiency without causing growth retardation, thereby increasing fruit set and yield71. Similarly, Cho et al. (2018) reported that the silencing of JUL1 homologues in rice also increased the number of phloem cells, supporting an evolutionarily conserved role of the JULGI–SMXL4/5 module in rice69. This implies that SMXL4 is likely to play a regulatory role in phloem development in rice. Interestingly, while grain size including area, length, and width did not differ between Samkwang and smxl4, the TGW was reduced in smxl4 (Figs. 3 and 4, Table S4). Given the potential function of SMXL4 in phloem development, this reduction is likely attributable to decreased assimilate transport to the grain, possibly leading to changes in grain density and internal structure. However, previous studies by Cho et al. (2018) and Nam et al. (2022) demonstrated that optimizing the expression of genes involved in phloem development enhances assimilate transport capacity, thereby increasing crop yield69,71. Therefore, further investigation into the impact of SMXL4 expression levels on rice productivity could provide insights into its potential as a breeding material for yield improvement.

In the Arabidopsis C24 background, the T-DNA insertion mutant athspr/smxl4 showed reduced rosette leaf size and inflorescence stem height, delayed flowering and development, and deficient seed germination31,33,34, consistent with phenotypes observed in rice smxl4, including developmental delays, decreased plant height and panicle length, and reduced seed germination. However, in contrast to previous findings in Arabidopsis, there were no significant differences in heading date or seed appearance traits (including area, length, and width) between Samkwang and smxl4, suggesting that the GA-mediated control of flowering time and seed set by SMXL4 may differ between Arabidopsis and rice. It has been reported that single mutants of SMXL3, SMXL4, or SMXL5 exhibited no obvious phenotypic differences compared to wild-type; however, double and triple mutants showed reduced primary root length, indicating a dose-dependent role of these genes in growth regulation30. Yang et al. (2015) also reported no significant differences in root length between wild-type and athspr31; however, Yuan et al. (2023) demonstrated a reduction in primary root length in athspr, indicating that the AtHSPR-KNAT5-OFP1 module regulates root development through modulation of GA20ox1 expression72. Similarly, in this study, two-week-old smxl4 seedlings exhibited reduced root length but no significant difference in shoot length compared to Samkwang, suggesting that SMXL4 may be involved in root development in rice. However, given that previous studies have shown variability depending on ecotype or experimental method, further investigation into the role of SMXL4 in rice root development is necessary. Collectively, these results suggest that SMXL4 knockout leads to growth defects in both Arabidopsis and rice, supporting its critical role in plant growth; however, they also suggest that species-specific differences may exist in SMXL4-mediated growth regulation and hormonal responses.

To evaluate the transcriptomic effects of SMXL4 editing, RNA-seq analysis was conducted on Samkwang and smxl4, identifying 84 and 120 DEGs from elongating internodes at the booting and heading stages, respectively. Among the common DEGs identified at the two stages, 51 genes were commonly upregulated, while only one gene was commonly downregulated in smxl4 relative to Samkwang. In rice, the cell wall primarily consists of cellulose, hemicellulose, and lignin73, and cell expansion involves loosening the existing cell wall structure and synthesizing and depositing new cell wall components74. The commonly upregulated genes included three genes encoding cell wall-related enzymes (BURP16, GNS11, CSLD3; Table 1), suggesting cell wall remodeling in association with shortened internodes in smxl4. However, this is inconsistent with previous reports indicating that cell wall hydrolases typically promote cell elongation75. Given that smxl4 plants exhibited reduced plant height compared to Samkwang from early developmental stages, the upregulation of these enzymes may represent a compensatory response, enhancing cell elongation and cell-wall synthesis. GNS11 encodes a β-glucanase (OsBGL) protein reported to contribute to sheath blight resistance in rice through regulation of callose accumulation76. In addition, the upregulation of genes encoding PID3, OsTPS4, and JA Upregulated Protein 1 suggests a potential role of SMXL4 in defense responses.

It has been reported that interaction between pectin and cellulose is critical for regulating internode elongation77. Pectin methylesterase (PME) catalyzes the hydrolysis of ester bonds in pectin, facilitating its degradation by pectinases such as polygalacturonases (PG). This process contributes to cell wall loosening and ultimately supports cell expansion7. At the booting stage, genes encoding OsPME17 and DIR were uniquely upregulated, while those encoding AGP and LTP were downregulated. These results suggest that cell elongation at the booting stage may have been promoted by cell wall-loosening enzymes, whereas reduced AGP expression might have limited cell expansion. At the heading stage, OsSHI1 and SRS were downregulated. Given that overexpression of OsSHI1 has been reported to significantly reduce plant height in rice61, this result may indicate a compensatory response to the growth reduction observed in smxl4. Taken together, these results suggest that modification of SMXL4 could affect the expression of cell wall remodeling-related genes during the period of internode elongation.

Unexpectedly, the reduced plant height observed in smxl4 was not accompanied by downregulation of genes involved in cell elongation. Therefore, the reduced plant height in smxl4 may be explained by different mechanisms: (1) a regulatory mechanism related to plant architecture may differ from that of previously reported dwarf or semi-dwarf mutants, which typically exhibit reduced internode cell length or number5,20,78; (2) a compensatory response may be involved, characterized by increased expression of cell-wall-related enzymes, possibly resulting from altered developmental processes caused by the knockout of SMXL4, a potential regulator of phloem formation; (3) the upregulation of defense-related genes in smxl4 suggests enhanced defense responses, which may be associated with growth reduction due to the growth–defense trade-off79. However, as histological analysis of internode tissues was not performed in this study, it remains unclear whether the differential expression of cell wall-related genes influenced cell elongation and expansion. Considering the potential role of SMXL4 as a regulator of phloem development in rice and the consistently reduced plant height observed in smxl4 throughout development, transcriptome analyses of various tissues across multiple growth stages may provide further insights into its genetic functions.

In summary, we identified an SMXL4-edited rice line exhibiting reduced plant height. The knockout of SMXL4 led to decreased CL and PL, as well as reductions in yield components such as GNP and TGW, lower seed germination rate, and shorter seedling root length. These results indicate that SMXL4 plays a broad role in rice growth and development. Transcriptomic analysis further revealed that knockout of SMXL4 may induce cell wall remodeling and alter the growth–defense balance. As the functional role of SMXL4 in rice remains largely uncharacterized, this study provides a foundation for future research into its physiological roles and potential utility in breeding programs.

Methods

Plant materials and growth conditions

We obtained a T1 rice population composed of 104 lines developed by editing various transcription factors using CRISPR/Cas9 in the background of Samkwang, from the National Institute of Agricultural Sciences (NIAS)37. The seeds were sown on April 25, 2022 and transplanted to the experimental field of NIAS on May 23, 2022, with one row per line and 20 plants per row, and the spacing of 20 cm between plants and 30 cm between rows. From this population, the SMXL4 (Os04g0298700) edited lines (smxl4) with good growth and a short-statured phenotype were selected. For yield assessment, the T2 seeds of smxl4 were sown on May 8, 2023 and transplanted on June 5, 2023 with Samkwang in the same field at NIAS in a randomized complete block design (RCBD), with three replicates composed of eight rows per replicate and 20 plants per row, with the same spacing as described above. T3 plants of smxl4 were crossed with the wild-type Samkwang and self-pollinated to generate the F2 segregating population (n = 84), which was sown on June 8, 2025 and transplanted on June 27, 2025, in the same field at NIAS. Other cultivation practices followed the standard rice cultivation guidelines of the Rural Development Administration80.

Phenotypic evaluations

In 2022, days to heading (DTH), culm length (CL), panicle length (PL) and panicle number (PN) were measured in the T1 population composed of 104 lines. DTH was recorded as days from transplanting to heading when ~ 40% of the plants within a row headed, and CL, PL, and PN were measured from five random plants in a row upon maturity. In 2023, the plant height of Samkwang and smxl4 was measured every two weeks throughout the entire growth period. For each replicate, at least three plants were measured from the ground to the tip of the flag leaf on the main stem. For the comparison of internode length of Samkwang and smxl4 at maturity, the length of five internodes of the main stem was also measured from three plants per replicate. The heading date was defined as the date when approximately 40% of the plants reached heading in each plot, and DTH was calculated as the number of days from transplanting to heading. Approximately 45 days after heading, CL, PL, and PN were measured from 10 randomly selected plants per replicate. CL was measured as length from the ground to the panicle neck on the main stem, and PL was measured from the panicle neck to the tip of the panicle. Panicle exsertion was defined as the distance from the lamina joint to the panicle neck at maturity stage, and if the panicle neck was enclosed within the leaf sheath, it was recorded as 0 cm. At approximately 55 days after heading, three plants with uniform growth per replicate were harvested to measure grain number per panicle (GNP), 1000-grain weight (TGW) and spikelet fertility (SF). Additionally, 40 plants per replicate were harvested to measure grain moisture content and grain weight, and then yield per plot (YPP) was calculated by adjusting the grain weight to 15% moisture content. In 2025, CL, PL, PN, GNP, TGW, and SF were evaluated in the F2 population described above.

To evaluate grain size, images of 100 seeds per replicate were captured using a digital camera α7RⅢ (Sony Co., Tokyo, Japan), and grain area, length, and width were measured using ImageJ software. For testing germination, the harvested seeds of Samkwang and smxl4 were subjected to dormancy-breaking treatment in an incubator at 40℃ for 14 days. The seeds were then sterilized with 1% sodium hypochlorite for 20 min, and then thoroughly rinsed with distilled water. Germination assays were conducted by evenly distributing 100 seeds of each genotype (Samkwang and smxl4) in petri dishes (three replicates) lined with two layers of filter paper (90 mm diameter), and 10 mL of distilled water was added to each dish. The dishes were then transferred to an incubator at 28℃±2℃ and kept in the dark for 4 days. Germination was monitored every 24 h, and the seeds were considered germinated if the emerged radicles measured at least 1 mm.

For testing seedling growth, the sterilized seeds (as described above) were sown on petri dishes with distilled water and kept at 28℃±2℃ for 2 days to ensure uniform germination. Afterward, 4 seeds of each genotype (Samkwang and smxl4) were placed in germination pouches (12 cm × 18 cm, PhytoAB Inc., USA) moistened with distilled water. The experiment was arranged in an RCBD with five replications. The seeds and germination pouches were placed in a plastic zipper bag with 10 mL of distilled water, arranged between Plexiglas plates (20 cm × 30 cm; PhytoAB Inc., USA), and maintained at 28 °C ± 2 °C (10 h day/14 h night) for 2 weeks. Shoot and root lengths were measured 2 weeks after sowing.

Statistical analysis and visualization were performed using the R program (v 4.2.1) and Microsoft Excel 2019, respectively. For field experiments, the average from each plot was used to represent an individual plot. ANOVA was conducted to determine the significant differences in plant height, internode lengths, panicle exsertion, agronomic traits and plot yield between Samkwang and smxl4, with significance levels indicated as P < 0.001 ***, P < 0.01 **, P < 0.05 *, and ns (not significant). Germination and seedling traits were analyzed following the same criteria.

Cas9-free screening and off-target assessment

Polymerase chain reaction (PCR) was performed to verify the removal of Cas9 in the gene-edited lines. As a positive control, primers for the Heading date 2 (Hd2) gene were included81, consistently amplifying a 1,062 bp fragment in all samples to ensure reliability (Table S9). The PCR amplicons were run on a 1% agarose gel at 135 V for 30 min, and the absence of Cas9 was verified using a UV gel documentation system (Analytik Jena, Germany). To verify the CRISPR/Cas9-induced mutation in smxl4, the SMXL4 genomic region covering the gRNA target region was amplified by PCR using the primers listed in Table S9. The PCR amplicons were sequenced using the CES (Capillary Electrophoresis Sequencing) service of Macrogen, Korea.

Potential off-target mutation sites were predicted using Cas-OFFinder (http://www.rgenome.net/cas-offinder/)38, allowing up to three nucleotide mismatches and one DNA and/or RNA bulge in the target region. A total of 315 potential off-target sites were identified, and their corresponding genomic positions were determined using BLAST searches against the Rice Annotation Project Database (RAP-DB, https://rapdb.dna.affrc.go.jp) based on the Nipponbare reference genome. The eight top-ranked off-target sites were amplified by PCR using the primers listed in Table S9, and the amplicons were sequenced as described above.

RNA-seq library construction and sequencing

To conduct transcriptome analysis of Samkwang and smxl4, internode samples of main stems were collected at booting and heading stages from the field experiments in 2023 described above. At booting stage (65 days after transplanting, DAT), the 2nd internodes were collected from plants with a panicle length of approximately 8 cm, while at the heading stage (72 DAT), the top internodes were collected from plants with approximately 10% panicle emergence. The samples were collected in the field between 10:00 am and 12:00 pm, immediately frozen in liquid nitrogen, and stored at -80℃. Three biological replicates from the plots arranged in RCBD were used per genotype at each time point. The samples were ground to fine powder in liquid nitrogen using a mortar and a pestle, and total RNA was extracted using the RNeasy Plant Mini Kit (74904; QIAGEN) by following the manufacturer’s instructions. Total RNA was treated with DNase I (M0303S; NEB) to remove gDNA and purified by using Agencourt RNAClean XP (A63987; Beckman Coulter, USA). The total RNA was analyzed using Agilent Bioanalyzer 2100 to determine the RNA integrity value (RIN), and quantified with the Qubit™ 3.0 Fluorometer (Q33216; Invitrogen, USA). The mRNA was enriched from one microgram of total RNA with an RIN value greater than 7.0 using Dynabeads® mRNA Purification Kit (61006; Invitrogen, USA). RNA-seq libraries were constructed using MGIEasy RNA Directional Library Prep Set (1000006386; MGI, China) following the manufacturer’s instructions. The 12 libraries with an insert size of approximately 250 bp were sequenced on the MGISEQ-2000 platform (MGI, China) with paired-end 150 bp reads.

Analysis of differentially expressed genes (DEGs)

After evaluating read quality with FastQC (v0.11.5), the raw sequencing data were trimmed using fastp (v0.23.2, https://github.com/OpenGene/fastp) to remove low-quality reads, short reads, and overrepresented sequences (e.g., adapter contamination, poly-G and poly-X tails). Using Kallisto (v0.44.0), the clean reads were aligned to the Oryza sativa japonica transcriptome (Oryza_sativa.IRGSP-1.0.cdna.all.fa, Ensembl Plants) and quantified39. Differentially expressed genes (DEGs) were identified between Samkwang and smxl4 plants at each growth stage (three biological replicates per genotypes) using DEseq2 package (v1.42.0) of the R program (v4.3.2). We filtered out the genes with less than 10 counts. Genes with adjusted p-value (padj) < 0.05 and |log2(foldchange)| ≥ 2 were considered as significant DEGs. Volcano plots for the DEseq2 output were generated using the Enhanced Volcano package in R82. Information on gene description was supported by ShinyGO 0.80. Gene ontology (GO) enrichment analysis of the DEGs was performed using Rice Gene Index database (RGI; https://riceome.hzau.edu.cn/) with default parameters83.