Rhizomes are modified subterranean stems that grow horizontally and can produce roots and shoots on their nodes, enabling vigorous asexual proliferation in various perennial species (Gizmawy et al., 1985; Li et al., 2022). From a physiological point of view, the rhizome is also the main energy storage organ of many perennial plants and plays a determinate role in their survival in harsh environment (Paterson et al., 1995). In agriculture, the rhizome trait is of significance due to both its positive and negative effects. On the one hand, strong rhizomes largely contribute to the competitiveness and invasiveness of weeds (Jang et al., 2006). On the other hand, rhizomes are also dispersal in many forage crops and may be utilized for developing perennial grain crops (Hu et al., 2011). Notably, it has been speculated that genes controlling rhizome development could be used to prolong the nutritional life cycle of certain plant species, potentially converting major annual grain crops into perennial ones (Paterson et al., 1995; Hu et al., 2003; Cox et al., 2006; Glover et al., 2010). As such, understanding the genetic mechanisms underlying rhizome development is not only instrumental in devising control or productivity-enhancing strategies for rhizomatous plants, but also beneficial to sustainable food production and ecosystem maintenance given the multiple ecological threats posed by our current annual crop dominated agricultural system (Hu et al., 2003; Cox et al., 2006; Glover et al., 2010).

Previous efforts to dissect the genetic basis of the rhizome trait were largely hindered by the sparsity of genomic resources. Recently, Oryza longistaminata is emerging as a useful model system for exploring rhizome development (Hu et al., 2003; Li et al., 2022). Originated from Africa, O. longistaminata is the only rhizomatous wild Oryza species that has the same AA genome type as the cultivated rice species O. sativa (Morishima, 1967; Tao and Sripichitt, 2000; Hu et al., 2003; Sacks et al., 2003; Ghesquiere, 2008). Hence, it can be crossed with well-studied rice cultivars for which high-quality genomic information is available, a strategy that is widely used for mapping genes or quantitative trait loci (QTLs) for agronomic traits. It is noteworthy that through such strategy, known as wide hybridization, a number of perennial rice lines have been successfully developed and are currently on trial in many Chinese provinces, representing an environmentally and economically sound rice production system (Zhang et al., 2017, 2019; Huang et al., 2018; Zhang et al., 2021).

An early study (Meakawa et al., 1998) that took advantage of the hybridization between O. sativa and O. longistaminata has suggested that the rhizomatous growth habit of O. longistaminata was segregated as a single dominant trait. The single dominant allele responsible for the rhizome trait was termed Rhz, and it was loosely linked to the liguleless (lg) locus on chromosome 4 with a recombination value of 37 ± 3.6% (Meakawa et al., 1998). However, the rhizomatous phenotypes displayed pronounced variations in the F₂ population, indicating the presence of other modifying genes. Based on a complete simple sequence-repeat map, Hu et al. (2003) identified two dominant-complementary loci, termed Rhz2 and Rhz3, that predominantly control rhizomatousness in O. longistaminata, and the authors also revealed many QTLs affecting rhizome abundance (Hu et al., 2003). Recently, through entire population genotyping mapping and selective genotyping mapping using three F₂ populations, over 10 major- or minor-effect rhizome-regulating QTLs were identified; however, none of these QTLs could be able to function alone, indicating that interactions among multiple QTLs are required for proper rhizome development (Fan et al., 2020). Nevertheless, to our knowledge, no reports have detailed how the rhizome-related QTLs interact with each other.

Understanding the relationships between complex genotypes and their underlying phenotypes is still one of the main challenges in modern genetics (Mackay et al., 2009). To better characterize the genetic networks underlying complex traits, a theoretical framework founded upon knowledge of signal transduction pathways has been proposed (Zhang et al., 2011). In this framework, the principle of hierarchy is defined as one-way functional dependency of downstream genes on upstream regulators, and functional genetic units (FGUs) refer to a group of functionally dependent genes acting at each level of a signaling pathway (Zhang et al., 2011). This framework was later employed to explore the pleiotropic effects of SD1, whose mutant alleles greatly contribute to the Green Revolution, and three genetic systems (SD1-mediated, -repressed, and -independent) were revealed, comprising 43, 38, and 64 FGUs, respectively, and jointly controlling growth, development and productivity of rice (Zhang et al., 2013). The functionality of this framework was further demonstrated in an investigation of the genetic basis of submergence tolerance (Wang et al., 2015). The putative networks consisted of 296 loci that were grouped into 167 FGUs, and the directional links between and among the nodes (the detected loci) suggested that submergence tolerance in rice is genetically controlled by a number of positively regulated signaling pathways (Wang et al., 2015).

Here, to gain insights into the genetic basis of rhizomatousness in O. longistaminata, a large F₂ population containing 818 individuals was used to map loci that control various rhizome-related traits. These identified loci were further interwoven into putative networks to elucidate their interconnections. Our results highlight the intricate genetic regulation on rhizome development and may provide information for pinpointing key rhizome-related genes that may be utilized in future perennial rice breeding programs.

In the present study, we substantially expanded the genetic repertoire for rhizome development. A recent manuscript (Fan et al., 2020) identified over 10 loci related to rhizome growth, including five major-effect loci—qRED1.2, qRED3.1, qRED3.3, qRED4.1, and qRED4.2, some of which are partially overlapped with two previously mapped loci determining the presence of rhizome, Rhz2 and Rhz3 (Hu et al., 2003). Here, we identified 13 Rhz-regulating loci and as many as 51 QTLs controlling rhizome abundance. Not surprisingly, some of them are overlapped with those previously reported QTLs. Comparing with Fan et al. (2020), eight of our identified loci (qRN-1-2, qRLSUM-2, qRHZ-3-1, qRHZ-3-1, qRHI-3, qRB2-4-2, qRHI-11, and qRHZ-11) are overlapped with their QTL intervals. Among these, two Rhz-regulating loci, qRHZ-3-1 and qRHZ-11, are located in the intervals of qRED3.1 and qRED-11, respectively, highlighting the potential prerequisite role of these loci, and qRHI-3 and qRHI-11, two intervals controlling multiple rhizome traits, were found to coincide with qRED3.3 and qRED11 (Fan et al., 2020). Meanwhile, many rhizome abundance QTLs we identified are closely linked with their major-effect loci, such as qRB2-2 and qRED2.2, qRN-3 and qRED3.1, and qRLSUM-4-2 and qRED4.1 (Fan et al., 2020). In general, these two studies are mutually confirmative; however, it is important to note that, instead of using the average rhizome extension distance (measured in soil surface, Fan et al., 2020), we treated the presence of rhizomes as a qualitative trait and carefully evaluated specific rhizome-related traits by digging the plants up. This may explain why our results do not completely include the previously identified QTLs, and such comprehensive phenotypic evaluation also evidently gave rise a larger number of identified QTLs in our study. Furthermore, the results presented here are also largely consistent with our previous study (Hu et al., 2003). A total of six pairs of loci are overlapped, and many of these loci pairs were identified to control the same or similar trait, such as qRN-1-2 and QRl1, and qRBD3-2 and QRbd2. The disparity between our current and previous results may stem from the distinct populations. Moreover, compared with previously used simple-sequence repeat (Hu et al., 2003) and insertion/deletion markers (Fan et al., 2020), the SNP markers used in this study have relatively wider coverage and higher density in the whole genome; thus, more QTLs have emerged.

Multiple genetic networks were established on the basis of classical QTL mapping (Zhang et al., 2011; Figures 5, 6). We observed a clear hierarchical structure in some networks for individual traits (Figure 5). This indicates that some loci, such as qRHZ-4 in the Rhz network (Figure 5A) and qRLsum-2 in the RLsum network (Figure 5C), may play a more significant regulatory role than the other loci in the same network since upstream loci are likely prerequisite for downstream loci to execute their functions. Given that the other traits measured in this study all depend on the presence of rhizomes, the two top-located nodes in the RHZ network are presumably of primary importance for proper rhizome development. Furthermore, in the comprehensive network containing all the identified loci, qRHI-2 and qRHI-5 were the hub loci connecting network-RHZ with network-RB and network-RN/L. In network analyses, hub nodes are generally considered to be critical (Zhang et al., 2011). In our case, considering that no other networks would exist if network-RHZ collapses, qRHI-2 and qRHI-5 are therefore considered as secondarily important loci for rhizome development. It is attempting to speculate that these four loci might be the minimal set of loci required for rhizome development, and this is worth of further investigation. Of note, the concept of FGU in this study is defined as a pair of mutual functional dependent genes, whose encoding proteins may not necessarily physically interact with each other, and FGUs in our networks were generated by comparing the genes’ contributions to the phenotype. Compared with RIL and doubled haploid populations, F₂ populations usually have a larger number of distinct genotypes, and the efficacy of detecting genetic networks using F₂ populations is weakened due to the possible errors occurred in multiple comparisons (Zhang et al., 2011). Such weaking effect would be amplified if there are a large number of loci controlling the trait (Zhang et al., 2011). Therefore, even though our preliminary functional verification suggested that the two hub loci, qRHI-2 and qRHI-5, are likely both needed for the rhizome phenotype, we could not exclude the possibility that other loci are also required here. Further verification using rhizomatous accessions and functional characterization of these loci would be needed to unravel the complex genetic basis of rhizome development.

In this study, we detected 62 loci that are putative regulators for rhizome development. By grouping these loci into FGUs, we provided evidence that rhizome growth is controlled by a multi-locus, multi-layered genetic network. Based upon the network structure and the interactions among the nodes, we predicted four loci, two major-effect loci for rhizome initiation and two hub loci connecting the individual genetic networks, as key loci. Functional verification using rhizome-free RILs confirmed that none of these key loci could solely initiate rhizome morphogenesis and suggested that qRHI-2 and qRHI-5 are likely both needed for rhizome initiation.

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding authors.

FH and SZ designed the experiments. WL, SZ, LH, GH, and JZ performed the experiments. FH, ZL, and WL wrote the manuscript. WL, SZ, and ZL analyzed the data. All authors have read and approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.