The following rice germplasms are paddy rice landraces and all obtained from the Yuanyang Hani terraced fields, Yunnan Province, China, including Acuce (Oryza sativa cv. indica, Acuce), Qi Xian Gu (Qxg), Da Leng Shui (Dls), Lao Jing Hong Jiao (Ljhj), Ma Wei Gu (Mwg), Ye Bai Gu (Ybg), Man Che Hong Nuo (Mchn), Hei Gu (Heig), Duo Dian (Dd), Nuo Gu (Ng), Meng La Gu (Mlag), Che Bu (Cb), Hua Gu (Huag), Jiu Yue Nuo (Jyn), Ma Zha Nuo (Mzn), Ai Zhe Gu (Azg), Chang Wei Nuo (Cwn), Yun Xiang (Yx), Mao Lai Gu (Mlig), Che Zuo (Cz), Ga Niang Hong Gu (Gnhg), Xiao Gu (Xg), Chuan Bai Gu (Cbg), Shi Yue Bai Gu (Sybg), Gan Di Gu (Gdg), Ban Jiu Gu (Bjg), Meng La Nuo (Mln), Xi Bai Gu (Xbg), Xiao Pi Gu (Xpg), Xiao Hua Gu (Xhg), Si Ma Che (Smc), Hong Jiao Lao Jing (Hjlj), Ma Xian Gu (Mxg), Hua Ke Nuo (Hkn), Gan Tian Nuo (Gtn), Le Che Che Ma (Lccm), Da Pi Gu (Dpg), Xiao Hua Nuo (Xhn), and Ai Jiao Gu (Ajg).

To observe the rice agronomic traits, including the root and tiller phenotypes, rice seeds were germinated in the field soil and grown for 3 weeks. The rice seedlings were then transplanted into plastic basins and grown in experimental fields located in Xiao Guo Xi Village, Caoba Town, Mengzi City, Yunnan Province (103°37′67′′ E, 23°49′39′′ N) with irrigation condition and Yunnan Agricultural University, Kunming, Yunnan Province (102°45′29′′ E, 25°8′26′′ N) with drought condition for 5 months, respectively. We repeated the experiment three times in April 2019, 2020, and 2021.

To detect the expression levels of DRO1 in roots and tillers of the 38 rice varieties from the Yuanyang Hani terraced fields, rice seeds were germinated in soil and grown for 3 weeks, and then the rice seedlings were transplanted into plastic basins and grown in experimental fields located in in Yunnan Agricultural University with drought condition for 3 months in 2022.

Rice seeds were surface-sterilized with 70% alcohol for 90 s, followed with treatment with 2% sodium hypochlorite for 14 min, and then washed five times with sterilized water. Seeds were cultured on Murashige-Skoog (MS) medium supplemented without sucrose and phytohormone for 3 days under dark condition, and then grown for 4 days under light condition. The 7-day old rice seedlings were used to observe the phenotypes of primary roots and lateral roots, and detect gene expression levels. The lateral roots were counted from primary root emerging lateral root growth point, and calculated the number of lateral roots per unit primary root length.

To construct the expression vectors 35S::DRO1Δexon5, the genomic DNA of DRO1 with deletion of the fourth intron and the fifth exon was amplified from the pBWA(V)HII-ProDRO1A::DRO1A plasmid containing complete DRO1 sequence (GenBank no. MH939159.1). Gene-specific primers DRO1-FP1 and DRO1-RP1 (Supplementary Table 1) were synthesized for PCR amplification. The PCR reaction mixtures were prepared with 4 μL primer pair DRO1-FP1 and DRO1-RP1, 5 μL 10 × PCR Buffer, 5 μL dNTPs (2 mM), 3 μL Mg²⁺ (25 mM), 1 μL Neo enzymes mix (1 U/μL), and 50 ng of plasmid template, then added ddH₂O to 50 μL. PCR reactions were performed under the following conditions: denaturation at 94°C for 2 min, followed by 29 cycles of 98°C for 10 s, 58°C for 30 s, and 68°C for 1 min. The PCR products were cloned into BGV002 vector and transcribed under the 35S promoter, and the recombinant expression vector was transferred into the Agrobacterium tumefaciens strain EHA105. Rice Acuce transformation was performed as previously described by the Agrobacterium-mediated method (Nishimura et al., 2006) at Biogle Company (Hangzhou Biogle Co., Ltd., Hangzhou, China).

Rice roots derived from tissue culture seedlings and rice seedlings grown under experimental fields with irrigation and drought stresses were pooled together. The total RNA isolation and complementary DNA (cDNA) synthesis followed methods described previously (Zhao et al., 2021). The relative quantitative expression levels of the DRO1, LAZY1, TAC1, qSOR1, PIN1b, LEA19, WAXY, ARF15, Os03g0624000, CRL1, Os06g0311000, WOX11, SHR1, PIN1A, PIN1C, and PIN2 genes were determined using an ABI QuantStudio 7 Flex Real-Time PCR system (Applied Biosystems, Waltham, MA, United States). The 10 μL reaction mixture was prepared with 5 μL PowerUp SYBR Green Master Mix (Thermo Fisher Scientific, Waltham, MA, United States) containing 0.5 μL cDNA template and 0.8 μL primer pairs DRO1-rFP and DRO1-rRP1, LAZY1-rFP and LAZY1-rRP, TAC1-rFP and TAC1-rRP, qSOR1-rFP and qSOR1-rRP, PIN1b-rFP and PIN1b-rRP, LEA19-rFP and LEA19-rRP, WAXY-rFP and WAXY-rRP, ARF15-rFP and ARF15-rRP, Os03g0624000-rFP and Os03g0624000-rRP, CRL1-rFP and CRL1-rRP, Os06g0311000-rFP and Os06g0311000-rRP, WOX11-rFP and WOX11-rRP, SHR1-rFP and SHR1-rRP, PIN1A-rFP and PIN1A-rRP, PIN1C-rFP and PIN1C-rRP, PIN2-rFP and PIN2-rRP (Supplementary Table 1) for DRO1, LAZY1, TAC1, qSOR1 and, PIN1b, LEA19, WAXY, ARF15, Os03g0624000, CRL1, Os06g0311000, WOX11, SHR1, PIN1A, PIN1C, and PIN2, respectively. The gene-specific primer pairs DRO1-rFP and DRO1-rRP2, DRO1-rFP and DRO1-rRP3 (Supplementary Table 1) were used to detect the expression levels of endogenous DRO1 and transgene DRO1 in transgenic rice lines #1, #4, and #8. To detect gene haplotype of DRO1 in the 38 rice varieties, the gene-specific primer pair DRO1-FP2 and DRO1-RP2 (Supplementary Table 1) were used to amplify the DRO1 gene. The Actin gene (GenBank no. AK060893) acted as the internal control, and was amplified using the primer pair Actin-FP and Actin-RP (Supplementary Table 1). Real-time PCR was performed under the following conditions: denaturation at 95°C for 2 min, followed by 40 cycles of 95°C for 45 s, 54–58°C for 30 s, and 72°C for 1 min. Three biological replicates were made. The relative expression level was calculated by using the 2^–ΔΔCt method. The PCR products of DRO1 gene were checked by using electrophoresis on 2% agarose gels and sequenced (Tsingke Biogle Co., Ltd, Beijing, China) to analysis DRO1 haplotypes.

The Search Tool for Retrieval of Interacting Genes/Proteins (STRING) database was used to build the protein–protein interaction (PPI) network between the IGT family proteins and proteins related to root development and auxin response, which provides directly physical connections as well as indirectly functional relationships (Szklarczyk et al., 2021). The protein sequences of qSOR1, DRO1, TAC1, and LAZY1 were used as query sequences, and we used STRING to find proteins that interact with them. The co-expression data were downloaded from Rice RNA-seq Database and Rice FREND database (Sato et al., 2013; Yu et al., 2022), which provide rice expression profiles and RNA-seq data. The rice microRNA (miRNA) family mature sequences were retrieved from Plant miRNA Encyclopedia (Guo et al., 2020), which is a database of miRNA loci (MIR) in many species including O. sativa. The IGT and PIN family genes were used as target genes for miRNA target prediction study with the psRNATarget program (Dai et al., 2018). Plant miRNA Encyclopedia and psRNATarget are two common tools which are wildly used in plant microRNA target analysis. Next, the PPI network, microRNA regulation network and co-expression network were visualized by Cytoscape (version 3.7.1) (Shannon et al., 2003). Then we combined the PPI network, microRNA regulation network and co-expression network with purple lines, green lines and orange dot lines and highlighted the proteins involved in auxin response and root development with red nodes and green nodes, respectively.

Image J 1.41 was used to measure root length, width, growth angle, and surface area. Measurement of the root growth angle followed methods described in previous report (Zhao et al., 2021). SPSS version 19.0 (IBM Inc., Armonk, NY, United States) was used to analyze differences in gene expression and normality distributions of length-width ratios in the 38 rice varieties. P < 0.05 indicated a statistical difference and P < 0.01 indicated a statistically significant difference. Pearson’s correlation coefficient was calculated by using the SPSS software and was used to analyze correlations between the tiller number and root area, root width, root length, and root length and growth angle. All images were processed in Photoshop.

Rice varieties that obtained from Yuanyang Hani’s terraced fields grown under irrigation condition showed different plant height and heading date (Supplementary Figure 1). We further observed the root phenotypes of 38 rice varieties (Supplementary Figure 2). The 38 varieties showed differences in tiller number and root area (Supplementary Table 2), with positive correlations between tiller number and root area, and between tiller number and root width (Supplementary Figure 3). The RSA of these 38 rice varieties could be classified into six groups according to the normality distributions of the ratio of rice root length and width (Supplementary Figure 4). The ratio distributions were 1.3–1.4, 1.4–1.5, 1.5–1.6, 1.6–1.7, 1.7–1.8, and 1.8–1.9, and the architectures of these groups were typified by the races Da Leng Shui (Dls), Hei Gu (Heig), Ai Zhe Gu (Azg), Ban Jiu Gu (Bjg), Ma Xian Gu (Mxg), and Xiao Hua Nuo (Xhn), respectively (Supplementary Figure 2; Supplementary Table 2).

To further examined the rice agronomic traits, the six typical rice varieties Dls, Heig, Azg, Bjg, Mxg, and Xhn were grown in experimental fields with irrigation or drought conditions (Figures 1A–L). Compared with the rice varieties grown under irrigation condition (Figures 1A,C,E,G,I,K), rice seedlings showed short root phenotypes when rice grown under drought condition (Figures 1B,D,F,H,J,L). We also found significant differences in root length (Figure 1M), root growth angle (Figure 1N), and tiller number (Figure 1O) between the six rice varieties. In the six rice varieties, compared with rice seedling grown under irrigation condition, the rice varieties Heig and Azg also had small root growth angles when rice seedlings grown under drought stress (Figure 1N), it showed that rice varieties Heig and Azg had deep-rooting phenotype to adapt to drought stress.

We found that there were positive correlations between tiller number and root length in the varieties Dls, Heig, Azg, Bjg, and Mxg, but there was no obvious positive correlation in the variety Xhn grown under irrigation condition (Figure 2). However, there were positive correlations between tiller number and root length in the varieties Dls, Heig, Bjg, Mxg, and Xhn, but there was no obvious positive correlation in the variety Azg grown under the drought condition (Figure 2). We analyzed the correlation between root growth angle and root length when rice seedlings grown on irrigation or drought conditions (Figure 3), it showed a significant negative correlation between these two traits in the varieties Dls, Heig, Azg, Mxg, and Xhn, but no significant negative correlation was observed in rice Bjg (Figure 3).

To investigate the role of the IGT family genes in rice root and tiller development, we analyzed the expression levels of the DRO1, LAZY1, TAC1, and qSOR1 genes in these organs. Under irrigation condition, in all varieties studied, compared with the level of DRO1, expression levels of LAZY1 were increased in roots. Furthermore, with the exception of Dls, the expression level of qSOR1 gene in roots was the lowest of all the genes studied. The expression of these four genes showed a similar pattern (LAZY1 > TAC1 > DRO1 > qSOR1) in the varieties Azg, Bjg, and Mxg. The pattern of gene expression (LAZY1 > DRO1 > TAC1 > /≈qSOR1) was also similar in the landraces Xhn and Dls, with expression of TAC1 and qSOR1 in Dls being almost the same. The variety Heig showed a different gene expression pattern (TAC1 > LAZY1 > DRO1 > qSOR1) to the other landraces (Figure 4A). Under drought condition, in all varieties studied, compared with the level of DRO1, expression levels of LAZY1 and qSOR1 were significantly increased in roots. Furthermore, with the exception of Xhn, the level of TAC1 expression also was increased in roots. The expression of these four genes showed a similar pattern (qSOR1 > LAZY1 > TAC1 > DRO1) in the varieties Dls, Heig, Azg, Bjg, and Mxg. The variety Xhn showed a different gene expression pattern (qSOR1 > LAZY1 > DRO1 > TAC1) to the other landraces (Figure 4C).

When we examined the expression levels of the IGT family genes in rice tillers, all rice varieties examined showed similar patterns of gene expression (TAC1 > DRO1 > LAZY1 > qSOR1) grown under irrigation condition, although in the case of Dls, the expression levels of DRO1 and LAZY1 were very similar. In all varieties, the expression of TAC1 was very much higher than that of the other examined genes, and that of qSOR1 was very low (Figure 4B). Furthermore, under drought condition, rice varieties Dls, Heig, Azg, and Xhn showed similar patterns of gene expression (qSOR1 > TAC1 > LAZY1 > DRO1), the variety Bjg and Mxg showed a different gene expression pattern (TAC1 > qSOR1 > LAZY1 > DRO1) to the other landraces (Figure 4D), although in the rice varieties Heig, Bjg, and Xhn, the expression levels of DRO1 and LAZY1 were very similar. Taken together, these results suggest that the DRO1, LAZY1, TAC1, and qSOR1 genes have different roles in the development of the rice roots and that of the tillers under different growth conditions.

Regulation of auxin distribution by PIN transporters is important in formation of the RSA (Lavenus et al., 2016; Zhao et al., 2021). Therefore, we analyzed the expression levels of PIN1b in the rice roots and tillers. When rice varieties were grown on irrigation or drought conditions, we found that the expression levels of PIN1b in the roots were significantly higher than that in the tillers (Figures 4E,F), and the increased levels of PIN1b in the roots of different rice varieties showed different expression patterns when rice seedlings grown under irrigation condition (Mxg > Dls > Azg > Heig > Bjg > Xhn) or grown under drought stress (Dls > Azg > Heig > Bjg > Xhn > Mxg) (Figures 4E,F).

To further detect the role of DRO1 in regulating rice root growth, we analyzed the DRO1 expression patterns of rice roots and tillers in the 38 rice varieties grown under drought condition. Compared with levels of DRO1 in tillers, 24 rice varieties had higher DRO1 levels, 12 rice varieties showed lower levels, and two rice varieties had no obvious differences in roots (Supplementary Figure 5). When we analyzed the DRO1 gene haplotype, it showed two single nucleotide polymorphisms (SNPs) (C-T, A-C) located in the exon 3 in the 38 rice varieties (Supplementary Figure 6). Next, we constructed transgenic Acuce rice lines overexpressing DRO1 gene with a deletion of the fifth exon (35S::DRO1Δexon5) that contains an EAR-like motif (Figures 5A,B). We wanted to test whether the EAR-like motif could be responsible for the role of DRO1 in regulating rice RSA. It showed that the T2 homozygous transgenic rice lines showed shorter root length (Figures 6B–E; Supplementary Figures 7B–E), less lateral roots (Figure 6F), and had no obvious differences but with decreased tendency in the tiller number (Supplementary Figure 7F) when compared with the wild type (Figure 6A; Supplementary Figure 7A). We found that there were higher levels of DRO1 in the primary roots in the transgenic rice lines (#1, #4, and #8) than the wild type (Figure 6G). When we detected the transgene DRO1 levels of primary root in the transgenic rice lines (#1, #4, and #8), the levels of transgene DRO1 were higher than the endogenous DRO1 gene (Supplementary Figure 8). This result suggests that the EAR-like motif located in the fifth exon is required for the DRO1 in regulating rice RSA.

To explore the effect of DRO1 on other the IGT family members, we tested the expression levels of DRO1, LAZY1, TAC1, and qSOR1 in the primary root of wild type and transgenic rice lines. It showed that there was a similar expression pattern (qSOR1 > LAZY1 > DRO1 > TAC1) in the meristem region, the elongation and differentiation zones in wild-type Acuce seedlings, and there was a different expression pattern (qSOR1 > TAC1 > LAZY1 > DRO1) in the culm of wild-type Acuce seedlings (Figure 6H). However, we found that the expression levels of TAC1, LAZY1, and qSOR1 were decreased in the meristem region, the elongation and differentiation zones of primary root and culm in the transgenic rice lines (Figure 6H). These data indicate that the malfunction of DRO1Δexon5 affects the expression of TAC1, LAZY1, and qSOR1 genes.

To find the relationship between the IGT family genes and other genes involved in auxin response and root development, a PPI network combined with co-expression network and microRNA targeting network was constructed. A total of 17 proteins and 14 microRNAs were showed in the network (Figure 7). Figure 7 showed that five proteins directly interacted with DRO1 protein, while three proteins directly interacted with LAZY1 protein. It also showed that genes qSOR1, TAC1, LAZY1, and DRO1 were regulated by different number of microRNAs.

It was reported that protein WOX11 may interacted with protein CRL1 and ARF15 (Zhang et al., 2018), and ARF family genes may regulate DRO1 gene (Mai et al., 2014). These relationships also showed in our predicted network according to Figure 7. Protein PROG1, which has function in controlling tiller angle and number (Jin et al., 2008), interacted with protein LAZY1. This showed that LAZY1 might be involved in controlling roots and tillers, which is consistent with previously reported function of this gene (Taniguchi et al., 2017). Protein WOX11 and auxin response factor 15 (ARF15), which were involved in auxin response, interacted with DRO1, WOX11 was thought to be an integrator of auxin and cytokinin signaling, and it was involved in regulating the cell proliferation during crown root growth (Zhao et al., 2009). SHR1 and CRL1, which were involved in root development, also interacted with DRO1. SHR1 plays an important role in cell division and tracheary element development of roots (Xing et al., 2021), while the CRL1, which is the target of auxin response factor 1, plays a significant role in crown root formation (Inukai et al., 2005). It is suggested that DRO1 might be involved in the auxin related pathway in addition to the root development. Our network also showed that CRL1 had an indirect interaction with ARF15. What’s more, proteins of LAZY1 and DRO1 may indirectly interact with each other through protein LEA19 and WAXY, and DRO1 had an indirect interaction with PIN family proteins according to the Figure 7. To further detect the regulation of DRO1 and its interaction partners, we tested gene expression levels that directly or indirectly interact with DRO1 showed in Figure 7 in the transgenic rice line (#8). Compared with the wild type, the levels of LEA19, SHR1, and PIN1A were decreased, the levels of WAXY, ARF15, Os03g0624000, CRL1, and WOX11 were significantly increased in the primary roots of transgenic rice line (#8) (Figure 8). Based on this result, we speculated that DRO1 may regulate the levels of DRO1 interacting partners to affect the expression of other IGT family genes (Figure 6H) in root and tiller development, which associated with changes in auxin transport.

According to Figure 7, TAC1, Os03g0624000, and LAZY1 displayed co-expression patterns, suggesting that TAC1 and LAZY1 may be involved in the same biological process. What’s more, qSOR1, PIN family genes and CRL1 exhibited co-expression patterns with Os06g0311000. We randomly selected rice varieties Azg, Bjg, and Nipponbare (NPB) to detect gene expression patterns, and it showed that the levels of TAC1, Os03g0624000, and LAZY1 (Figures 9A–C) and Os06g0311000, CRL1, PIN1A, PIN1C, PIN2, and qSOR1 (Figures 9D–F) had similar expression in roots, tillers and leaves, respectively. We found that PIN1A, PIN1C, PIN2 had lower expression levels in the roots of rice varieties NPB, Azg, and Bjg (Figure 9D), and qSOR1 had higher levels in the roots of rice NPB and Bjg, but lower level in rice variety Azg (Figure 9D). PIN family proteins have a critical function in auxin transportation and root development (Li et al., 2019). We hypothesized that qSOR1 has a role in responding to auxin transport of rice roots based on the network of IGT family genes.

The regulatory relationship between the IGT family genes and mircroRNAs were also displayed in Figure 7. We found that the IGT family genes were regulated by different microRNAs (Supplementary Table 3). It showed that qSOR1 was only targeted by miR5491, while DRO1 was targeted by seven microRNAs (Supplementary Table 3). The DRO1 gene was targeted by miRNA156 family (Supplementary Table 3). DRO1 and PIN2 were both target genes of miR156a, miR156b, and miR156c, while LAZY1 and AUX4 were targeted by miRN2268a and miRN2268b (Supplementary Table 3). These microRNAs regulation network suggested that function of the IGT family genes are tightly connected with microRNAs in regulating root and tiller development associated with auxin transport.