Innovation and appropriation in mycorrhizal and rhizobial Symbioses

Abstract

Most land plants benefit from endosymbiotic interactions with mycorrhizal fungi, including legumes and some nonlegumes that also interact with endosymbiotic nitrogen (N)-fixing bacteria to form nodules. In addition to these helpful interactions, plants are continuously exposed to would-be pathogenic microbes: discriminating between friends and foes is a major determinant of plant survival. Recent breakthroughs have revealed how some key signals from pathogens and symbionts are distinguished. Once this checkpoint has been passed and a compatible symbiont is recognized, the plant coordinates the sequential development of two types of specialized structures in the host. The first serves to mediate infection, and the second, which appears later, serves as sophisticated intracellular nutrient exchange interfaces. The overlap in both the signaling pathways and downstream infection components of these symbioses reflects their evolutionary relatedness and the common requirements of these two interactions. However, the different outputs of the symbioses, phosphate uptake versus N fixation, require fundamentally different components and physical environments and necessitated the recruitment of different master regulators, NODULE INCEPTION-LIKE PROTEINS, and PHOSPHATE STARVATION RESPONSES, for nodulation and mycorrhization, respectively.

A review of progress in plant symbiosis, emphasizing more recent advances in understanding where mycorrhizal and root nodule symbioses overlap and diverge and how these systems evolved.

Introduction

Across angiosperms, the two most important beneficial interactions of plants are arbuscular mycorrhizal (AM) symbiosis, in which most species of land plants participate, and nodulation, which is carried out by most (approximately 16,000) legume species and a handful of nodulating nonlegumes. The importance of these interactions to terrestrial ecology and to agriculture is immense, and notwithstanding 20 years of accelerated research progress buoyed by the rise of model legume systems and genomics, our understanding of these interactions still has a long way to go (Roy et al., 2020). Nonetheless, cumulative advances have been made in the areas of microbial signal perception (which includes crosstalk with immune pathways) and systemic signaling (which integrates symbiotic infection with plant nutrient homeostasis). More recently, progress has been made in understanding the processes and mechanisms underlying symbiotic infection and the regulation of the transition from the infection to nutrient acquisition phases of symbiosis. Here we review progress in these areas, with emphasis on more recent advances, focusing on where mycorrhizal and root nodule symbioses overlap, where they diverge, and how these systems evolved.

Perception of symbiotic and pathogenic factors from microbes

Microbe/Pathogen-associated molecular patterns (MAMPs/PAMPs) are small, conserved molecular patterns that are recognized by the plant innate immune system. MAMPs include carbohydrate-based molecules such as chitin, peptidoglycan (PGN), lipopolysaccharides, and exopolysaccharides (EPS; Boller and Felix, 2009; Liu et al., 2012a). Symbiotic signaling molecules are structurally similar to carbohydrate-based MAMPs, and their recognition allows host plants to quickly differentiate compatible symbionts from the greater population of microbes.

Lipo-chitooligosaccharides (LCOs) secreted from rhizobia or AM fungi (AMF) have been identified as key signals used to communicate with host plants, allowing the entry of the microbial partners into plant cells (Roche et al., 1991; Schultze et al., 1995; Chabaud et al., 2002; Genre et al., 2013; Oldroyd and Downie, 2004; Kosuta et al., 2003; Maillet et al., 2011; Miller and Oldroy, 2012). Nod-LCOs generally consist of four or five β-1, 4-linked N-acetyl glucosamine units and a nonreductive terminal sugar acylated by fatty acid (FA) chains (C16:0, C18:1). Different substitutions are added to the terminal sugar residues of the glucosamine backbone depending on the rhizobia species, which plays a role in host range specificity (Roche et al., 1991; Schultze et al., 1995; D’haeze and Holster, 2002). AM signals are acylated, chitin-like molecules called Myc-LCOs (Maillet et al., 2011; Liang et al., 2014; Limpens et al., 2015; Sun et al., 2015). In addition to Myc-LCOs, short-chain chitooligosaccharides consisting of four or five N-acetyl glucosamine (GlcNAc) residues can induce symbiotic calcium (Ca²⁺) spiking in different plant species, indicating that these chitooligosaccharides, CO4 and CO5, also play important roles in plant–mycorrhiza symbiosis (Genre et al., 2013; Sun et al., 2015;  He et al., 2019). Actinobacteria of the genus Frankia also induce nodulation in some species of Fagales, Rosales, and Cucurbitales, and genes for LCO biosynthesis have been identified in Frankia genomes (Nguyen et al., 2016, 2019). The common use of LCOs as signaling molecules across mycorrhizal, actinorhizal, and rhizobial interactions reflects the stepwise evolution of these symbioses, which is discussed below. Interestingly, LCOs are widely distributed across fungal species, including pathogenic and saphrophytic fungi, and contribute to fungal growth and development (Bonhomme et al., 2021; Rush et al., 2020). Collectively, LCOs appear to be common regulatory signals of fungi, which have been recruited and specialized as symbiotic signals in different plant–microbial interactions.

Although carbohydrate-based MAMPs such as chitin and PGN are structurally related to LCOs, plants interpret MAMPs as danger signals and respond by triggering innate immunity, which hampers pathogen infection (Boller and Felix, 2009; Macho and Zipfel, 2014; Wu et al., 2014; Zipfel, 2014). Chitin, the major component of the fungal cell wall, is a polymer of GlcNAc units that repeat to form long chains with β-1,4 linkages. Chitin heptamers and octamers can effectively trigger plant immune responses (Liu et al., 2012b). PGN, which is present in almost all bacterial cell walls, is a polymer consisting of alternating residues of β-1,4-linked GlcNAc, and N-acetyl muramic acid residues that are cross-linked by short peptide bridges (Willmanna et al., 2011; Dworkin, 2014; Gust, 2015).

Nod-LCOs and Myc-LCOs inhibit plant immunity and facilitate infection of rhizobia and mycorrhizal fungi (Shaw and Long, 2003; Maillet et al., 2011; Lopez-Gomez et al., 2012; Liang et al., 2013; Feng et al., 2019;  He et al., 2019). Feng and associates showed that chitin (CO4–CO8, especially CO8) and PGN also induce symbiosis signaling in Medicago truncatula and that a combination of chitin/PGN and Nod (NF)/Myc factor (MF) recognition enhances symbiotic responses (Feng et al., 2019), indicating that symbiotic receptors can also respond to immunity-associated MAMPs.

Recognition systems of symbiotic factors and carbohydrate-based MAMPs

A growing body of evidence indicates that the perception of chitin-like MAMPs/PAMPs or symbiotic signals (LCOs) involves lysine-motif (LysM)-containing receptor-like kinases (LYKs) or LysM-containing proteins (LYPs)/LysM domain proteins (LYMs), the latter of which lack an intracellular kinase domain. The paralogous LysM proteins rice (Oryza sativa) CHITIN ELICITOR RECEPTOR KINASE 1(CERK1; OsCERK1), banana (Musa acuminata) LYK1 (MaLYK1), M. truncatula LYK9 (MtLYK9), and MtNFP play roles in both symbiotic and pathogenic interactions with microbes (Rey et al., 2013, 2015; Miyata et al., 2014; Zhang et al., 2015a, 2019b; Feng et al., 2019; Gibelin-Viala et al., 2019), indicating that the perception of symbiotic signals is closely intertwined with innate immune signaling. These findings highlight the similarities and distinctions between plant symbiosis and immunity.

Perception of symbiotic factors

In legumes, two receptor-like kinases (LjNFR1 and LjNFR5 in Lotus japonicus, MtLYK3 and MtNFP in M. truncatula, and NOD FACTOR RECEPTOR 1a (GmNFR1) and GmNFR5 in soybean [Glycine max]) have been identified that mediate Nod (nodulation) factor responses (Amor et al., 2003; Limpens et al., 2003; Madsen et al., 2003; Radutoiu et al., 2003; Indrasumunar et al., 2010, 2011; Broghammer et al., 2012). LjNFR5/MtNFP/GmNFR5 consists of three extracellular LysM domains and an intracellular atypical kinase domain that lacks an activation loop and does not autophosphorylate in vitro, and so is believed to be a pseudo-kinase (Madsen et al., 2003; Arrighi et al., 2006; Indrasumunar et al., 2010;). Nonetheless, deletion of the atypical kinase domain of MtNFP abolished its function, indicating that this domain serves some essential purpose, such as mediating interactions with other proteins (Lefebvre et al., 2012; Pietraszewska-Bogiel et al., 2013). LjNFR1/MtLYK3/GmNFR1 contains three extracellular LysM domains and an intracellular active kinase domain that shows autophosporylation activity in vitro and is required for NF signaling (Indrasumunar et al., 2011; Madsen et al., 2011; Pietraszewska-Bogiel et al., 2013). LjNFR1/MtLYK3 and LjNFR5/MtNFP form heterodimers and possibly homodimers at the cell surface (Madsen et al., 2011). LjNFR1 can phosphorylate the cytoplasmic domain of LjNFR5 in vitro, suggesting that phosphorylation might be important for initiation of downstream signaling (Madsen et al., 2011).

LjNFR1 and LjNFR5 directly bind to NFs with high affinity (nanomolar range), resembling the concentrations required for the activation of symbiotic signaling in legume plants (Broghammer et al., 2012). The transfer of LjNFR1 and LjNFR5 to M. truncatula extended its host range to include Mesorhizobium loti strains; these symbionts are normally restricted to L. japonicus and its close relatives (Radutoiu et al., 2007). The specificity of NF perception by these LysM receptor-like kinases has been linked to specific amino acid residues in their LysM domains (Radutoiu et al., 2007; Bensmihen et al., 2011; Bozsoki et al., 2020). These findings suggest that LjNFR1/MtLYK3/GmNFR1 and LjNFR5/MtNFP/GmNFR5 both contribute to Nod-LCOs perception. However, the structures of their ligands and receptors are still unknown.

Myc-LCOs are structurally similar to NFs, suggesting that similar systems are employed for their perception. Myc-LCO treatment stimulated lateral root formation in M. truncatula in an NFP-dependent manner, indicating that Myc-LCOs could be perceived by MtNFP in M. truncatula (Maillet et al., 2011). However, no apparent role of MtNFP in mycorrhizal symbiosis has been observed in M. truncatula (Zhang et al., 2015a). MYC FACTOR RECEPTOR 1 (OsMYR1) directly binds to Myc-CO4 and is required for AM symbiosis (He et al., 2019). OsCERK1, which is required for chitin-triggered immunity in rice, is also required for mycorrhizal infection (Miyata et al., 2014; Zhang et al., 2015a). Interestingly, Myc-CO4 promotes OsMYR1–OsCERK1 protein complex formation, suggesting that a similar mechanism underlies the perception of NFs and MFs.

Parasponia andersonii is a nonlegume plant that can undergo both rhizobial and mycorrhizal symbioses. PanLYK1 and PanLYK3 are required for AM and rhizobia symbioses, whereas the Pannfp1 Pannfp2 double mutant is defective in nodulation but has wild-type levels of AM colonization (Rutten et al., 2020). However, the suppression of PannNFP by RNA interference reduced both nodulation and mycorrhization (Op den Camp et al., 2011), indicating that additional NFP homologs might function in mycorrhizal symbiosis in this species. In the nonleguminous plants petunia (Petunia hybrida) and tomato (Solanum lycopersicum), PhLYK10, and SlLYK10 (a homologous protein of NFP) are required for AM symbiosis (Girardin et al., 2019). Interestingly, the expression of PhLYK10 and SlLYK10 partially restored nodulation in Mtnfp mutant plants (Girardin et al., 2019), suggesting PhLYK10 and SlLYK10 have the potential to participate in root nodule symbiosis signaling.

NFs and MFs activate a common symbiotic signaling pathway. Despite the similarities between the signaling molecules and convergence of the signaling pathways, the signaling outputs are distinct. Therefore, a high degree of specificity in the perception of these symbiotic factors must be required to activate processes that promote either rhizobial infection or fungal colonization. This specificity seems to involve the differential recognition of NF and MF signals by LysM receptor-like kinase complexes and the activation of appropriate downstream gene expression (Mulder et al., 2006; Radutoiu et al., 2007). The next challenge will be to elucidate how LysM domains can discriminate between Nod and Myc signals.

Perception of chitin, PGN, and EPS

Chitin Elicitor Binding Protein (CEBiP) is a glycosylphosphatidylinositol-anchored protein containing three LysM domains. Chitin-triggered immune responses in rice are dependent on OsCEBiP (Kaku et al., 2006). OsCEBiP directly binds to the acetyl moieties of the chitin fragment (Hayafune et al., 2014). OsLYP4 and OsLYP6 are dual-specificity receptors that can bind both chitin and PGN (Liu et al., 2012a). AtLYK4/5/LjLys13/14/MtLYR4 are required for chitin-induced immunity. AtLYK5 shows higher affinity to chitin than AtCERK1, indicating that AtLYK5 is a chitin receptor in Arabidopsis (Lohmann et al., 2010; Cao et al., 2014; Bozsoki et al., 2017).

AtCERK1/LjLYS6 (LjCERK6)/MtLYK9/PanLYK3 are essential for chitin recognition in Arabidopsis thaliana, L. japonicus, M. truncatula, and P. andersonii, respectively (Miya et al., 2007; Wan et al., 2008; Liu et al., 2012b; Bozsoki et al., 2017, 2020; Rutten et al., 2020). Due to their lack of cytoplasmic domains or intracellular typical kinase domains, the LYPs and LYK5 must interact with CERK1 to form a complex in order to initiate downstream responses (Shimizu et al., 2010; Ao et al., 2014). AtCERK1 also forms a complex with AtLYM1 and AtLYM3 to directly recognize PGN (Willmanna et al., 2011). AtCERK1/OsCERK1 may act as co-receptors that form different complexes with specific partners for perception of PGN and chitin. Two motifs in the LysM1 domain of chitin receptor LjLYS6 (LjCERK6) or LjNFR1 have been identified that determine the specificity of ligand recognition (Bozsoki et al., 2020).

LjLYS11 can bind to the chitin oligomer CO5, and replacing its ligand-binding site with six amino acids from the corresponding region of LjNFR5 allowed it to specifically bind to LCO (Gysel et al., 2021). Therefore, specific motifs or regions of the LysM receptors are responsible for the recognition of specific ligands. Exactly how these receptors coordinate in the perception of chitin, PGN, and symbiotic factors remains to be further elucidated.

EPS are high-molecular weight polymers composed of a single sugar (homopolysaccharides) or complex mixtures of sugars (heteropolysaccharides) that suppress host defense and facilitate bacterial infection (Sutherland, 1994; Denny, 1995; Aslam et al., 2008; Jones et al., 2008). The LysM domain-containing receptor-like kinase EPR3 functions as an EPS receptor (Kawaharada et al., 2015, 2017) by forming a compact structure consisting of three putative carbohydrate-binding modules that enable it to directly perceive EPS from different bacterial species (Wong et al., 2020).

The evolution of LYKs for MAMP and symbiotic signal recognition

AtLYK5–AtCERK1 and OsMYR1–OsCERK1 (or MtNFP/LjNFR5-MtLYK3/LjNFR1) protein complexes function in an analogous fashion to perceive pathogenic and symbiotic signals, respectively (Madsen et al., 2011; Pietraszewska-Bogiel et al., 2013; Cao et al., 2014; Moling et al., 2014). OsCERK1/MaLYK1/MtLYK9/PanLYK3 are required for both plant immunity and AM symbiosis (Shimizu et al., 2010; Miyata et al., 2014; Feng et al., 2019; Gibelin-Viala et al., 2019; Zhang et al., , 2015a, 2019b). Similarly, Medicago NFP and PanLYK3 have dual functions in defense and rhizobial symbiosis (Rey et al., 2013). Co-expression of the symbiotic receptors LjNFR5–LjNFR1 or MtNFP–MtLYK3 in Nicotiana benthamiana leaves triggered a spontaneous cell-death phenotype and induced the expression of defense genes (Madsen et al., 2011; Pietraszewska-Bogiel et al., 2013). Thus, the following questions arise: When did these perception systems emerge? How did they evolve? What is the relationship between the perception systems for chitin/PGN and symbiotic factors in early land plants?

Phylogenetic analysis revealed a clear separation of LjNFR1/MtLYK3/CERK1 (clade 1) and LjNFR5/MtNFP/AtLYK5 (clade 2); this separation is not based on their involvement in perceiving chitin or symbiotic factors (Figure 1; Supplemental File S1). This clustering appears to have a functional basis, given that the perception of chitin and symbiotic factors requires heterodimerization of receptors from clades 1 and 2. In both the liverwort (which undergoes AM symbiosis) and the moss Physcomitrella patens (which does not), clades 1 and 2 are represented by two LYKs. It is therefore plausible that in liverworts, these two types of LYKs form a heterocomplex to recognize both symbiotic and immune signals, while the same heterocomplex in mosses, which are unable to form mycorrhiza, serves only in immunity. This model is supported by the finding that two clade 1 LYKs play functionally redundant roles in chitin-induced immune responses in moss (Bressendorff et al., 2016). Intriguingly, in Selaginella moellendorffii, clades 1 and 2 are each represented by a single LYK gene, further supporting the hypothesis that a shared perception system for chitin and symbiotic factors existed in early land plants (Figure 1). Although the possibility of alternative receptors for chitin or symbiotic factors cannot be ruled out, current phylogenetic evidence is more consistent with a common perception system in liverwort and S. moellendorffii (Figure 2).

Figure 1.

Phylogenetic topology of plant LYK proteins. Molecular phylogeny of LysM-containing receptor-like kinases in different plant lineages. Protein sequences were aligned by Clustal X2 (http://www.clustal.org/), and the phylogenetic tree was generated by MEGA version 10 (http://www.megasoftware.net) using the maximum-likelihood method. The numbers at the branches are bootstrap values calculated from 1,000 bootstrap replicates. The protein sequences used in the phylogenetic tree are shown in Supplemental File S1. Color ranges represent clades 1 and 2. Different font colors represent subfamilies a–g. Subfamily a comprises EPS receptors. Subfamily c comprises COs and MF receptors. Subfamily d comprises Nod factor receptors. Subfamily e comprises LCO receptors. Subfamily g comprises CO receptors.

Figure 2.

The perception of symbiotic factors and related MAMPs from an evolutionary perspective. For liverworts and pterophytes, the homologs of CERK1 and LYK4/5 form a complex which can perceive either symbiotic factors or MAMPs to mediate the appropriate signaling, with an LYP also potentially contributing to plant immunity. Since there are no direct data to suggest that mosses can form symbiotic relationships with microbes, it is expected that this system is mainly used by them mosses to perceive immunity-related MAMPs. In angiosperms, LYKs underwent functional differentiation and evolved separate recognition systems for the perception of symbiotic factors and MAMPs, with some LYKs maintaining their ancient functions in both symbiosis and immunity. The question marks indicate currently unknown components or regulation.

In angiosperms, the systems mediating the recognition of carbohydrate-based MAMPs and symbiotic factors, while allowing for high specificity, partially overlap, and these two systems interact with each other through bi-functional co-receptors, such as OsCERK1/MaLYK1/MtLYK9/MtNFP/PanLYK3, and so on (Figure 2). LCOs consist of a chitin backbone made of GlcNAc units with an N-acylated nonreducing terminal glucosamine unit. The perception of specific LCOs and COs is determined by specific motifs in the LysM domains of the receptors. Recent data suggest that the process of functional differentiation and specialization of receptors occurs through substitutions within key motifs in the ligand-binding region. However, so far, our knowledge of the factors driving the evolution of LYKs remains limited and could be related to nutritional stress, interspecies competition, or other factors. In addition to LYKs, LYPs also serve as chitin or PGN receptors (Figure 2). In Arabidopsis, AtLYM1 and AtLYM3 specifically function in PGN-induced immune responses, and AtLYM2 is a chitin receptor (Willmanna et al., 2011; Faulkner et al., 2013). In rice, the homologs of AtLYM1 and AtLYM3, OsLYP4 and OsLYP6, are dual function receptors involved in the perception of chitin and PGN (Liu et al., 2012a). The origin and evolution of LYPs as it relates to their function are still unclear.

Why is specificity so important? NF-independent nodulation

Host–symbiont specificity, a hallmark of legume nodulation, serves to match legume hosts with their optimal partners, with the goal being to engage the rhizobia with the highest N-fixation efficiency. Considering a large number of LysM family members in the legume lineage, it is likely that the expansion of this receptor family was a key event facilitating the evolution of host specificity. However, some legumes, such as the semiaquatic legume Aeschynomene indica, can be nodulated by rhizobia via a NF-independent pathway (Giraud et al., 2007; Zhang et al., 2019c). In some cases, the nodulation of A. indica is enabled through use of type III secreted effectors (Teulet et al., 2019), while in others it appears that neither effectors nor NFs is required (Giraud et al., 2007). The use of effectors by rhizobia to bypass LCO signaling was also reported for soybean (Okazaki et al., 2013). Interestingly, both LYK3 and NFP are present in the A. indica genome (Quilbé et al., 2021). NFP, as discussed above, may be retained to support mycorrhization, while the presence of the LYK3 homolog could suggest its importance for effector-triggered nodulation; it may instead be a pseudogene, as it is expressed at very low levels. Regardless of the mechanisms involved, by circumventing the plant’s ability to choose its symbiont, rhizobia capable of NF-independent nodulation are ostensibly “cheaters” that exist alongside more beneficial symbionts in nature (Fujita et al., 2014; Moyano et al., 2017; Gano-Cohen et al., 2020). This idea is supported by the evolution of R-genes in soybean that target component of the Bradyrhizobium japonicum type III secretion system (Tsukui et al., 2013). The successful legume shopkeeper is one that can distinguish these shoplifters from their preferred clientele: rhizobia that are highly efficient at N fixation. The lock-and-key system of NFs and flavonoids is the legume’s most important security measure.

Signal transduction in mycorrhizal and rhizobial symbioses

Our knowledge of the pathway between the perception of NFs/MFs and the induction of nucleus-associated Ca²⁺ oscillations, which are required for the establishment of nodule and AM symbioses, still contains many gaps. An LRR receptor-like kinase known as SYMBIOSIS RECEPTOR KINASE (SYMRK) in L. japonicus, GmSYMRK in G. max and its ortholog DOES NOT MAKE INFECTIONS2 (DMI2) in M. truncatula are essential for NF and MF signaling (Endre et al., 2002; Stracke et al., 2002; Indrasumunar et al., 2015). The cytoplasmic domain of SYMRK shows autophosphorylation activity in vitro (Chen et al., 2012; Samaddar et al., 2013), which has led to the proposal that SYMRK functions as a co-receptor in a complex with the LysM containing receptor-like kinases to mediate NF and MF signal perception (Figure 3), although no trans-phosphorylation between MtDMI2 and MtLYK3 was reported.

Figure 3.

The common symbiotic signaling pathway. In M. truncatula and L. japonicus, LYK3/NFR1 and NFP/NFR5 form a receptor complex to perceive NFs. In rice, CERK1 (homolog of LYK3/NFR1) and MYR1 (homolog of NFP/NFR5) recognize MFs in a similar manner. Following the recognition of the symbiotic factor, nodule and AM symbioses share common signaling components. DMI2/SYMRK and RLCKs interact with the symbiotic receptor complex and contribute to nodule and AM symbioses. HMGR associates with DMI2/SYMRK to promote the production of the potential second messenger MVA. MVA somehow activates the Ca²⁺ channels DMI1/POLLUX and CASTOR, which may interact with CNGC15 to regulate the influx of Ca²⁺ into the cytoplasm from the ER lumen-perinuclear space. The Ca²⁺ ATPase MCA8 then recaptures Ca²⁺ from the nucleoplasm and nucleus-associated cytoplasm, thereby resetting the system. DMI1, CNGC15, and MCA8 thus form a regulatory loop to maintain continuous Ca²⁺ spiking. The nuclear pore complex is also required for Ca²⁺ oscillations, but its role is unclear. CCaMK decodes Ca²⁺ spikes and phosphorylates the transcriptional activator IPD3/CYCLOPS. DELLA acts as a bridge to connect IPD3 and NSP2/DIP1. NSP2/DIP1 and NSP1/RAM2 form different heterodimers to output nodule or AM signaling through NIN and PHR2, respectively. MAC8, a sarco/ER Ca²⁺ ATPase.

The receptor-like cytoplasm kinase (RLCK) LjNICK4 interacts with LjNFR5 and shuttles to the nucleus, where it regulates nodule number (Wong et al., 2019). Three other RLCKs also contribute to AM symbiosis, but their roles are unclear (Bravo et al., 2016). The enzyme needed to produce mevalonate (MVA), 3-hydroxy-3-methylglutaryl-CoA reductase (HMGR), interacts with MtDMI2 and is required for Ca²⁺ spiking (Kevei et al., 2007; Venkateshwarana et al., 2015). Exogenous application MVA, a precursor of potential secondary messengers, is sufficient to induce Ca²⁺ spiking, which is dependent on MtNFP and MtDMI1 (Venkateshwarana et al., 2015). MtDMI1 and its orthologs LjCASTOR and LjPOLLUX, which are located in the nuclear envelope and are required for symbiotic Ca²⁺ spiking, are Ca²⁺-regulated Ca²⁺ channels (Ane et al., 2004; Capoen et al., 2011; Charpentier et al., 2008; Granqvist et al., 2012; Imaizumi-Anraku et al., 2005; Peiter et al., 2007; Riely et al., 2007). In addition, CYCLIC NUCLEOTIDE-GATED CHANNELS (CNGC15s) in M. truncatula function as Ca²⁺ channels to modulate nuclear Ca²⁺ release from the endoplasmic reticulum (ER) lumen-perinuclear space and form a complex with MtDMI1 in the nuclear envelope (Charpentier et al., 2016). The ATPase MCA8, which localizes to the nuclear envelope and ER membrane, also contributes to symbiotic Ca²⁺ spiking (Capoen et al., 2011). Three components of the nuclear pore complex, NENA, NUP85, and NUP133, are also required for symbiotic Ca²⁺ spiking and regulate AM and rhizobial colonization (Groth et al., 2010; Kanamori et al., 2006; Saito et al., 2007; Figure 3).

The decoding of Ca²⁺ oscillations involves the Ca²⁺ and calmodulin (CaM)-dependent serine/threonine protein kinase LjCCaMK/MtDMI3 (Levy et al., 2004; Mitra et al., 2004). The activation of Ca²⁺- and CaM-dependent serine/threonine protein kinase (CCaMK) is sufficient to initiate downstream symbiotic events; constitutively activated CCaMK induces the spontaneous formation of nodules and the AM-associated prepenetration apparatus (Gleason et al., 2006; Takeda et al., 2012; Tirichine et al., 2006). LjCCaMK/MtDMI3 interacts with and phosphorylates the transcriptional activator LjCYCLOPS/INTERACTING PROTEIN OF DMI3 (MtIPD3; Messinese et al., 2007; Yano et al., 2008; Horvath et al., 2011; Ovchinnikova et al., 2011). The GRAS protein DELLA, a central regulator of gibberellic acid (GA) signaling, connects MtIPD3 and the downstream GRAS transcription factors NODULATION SIGNALING PATHWAY1/2 (NSP1/2) to regulate nodulation (Jin et al., 2016). NSP2 forms a complex with NSP1, which binds to the promoters of nodulation-specific genes such as the transcription factor genes NODULE INCEPTION (NIN) and ERF REQUIRED FOR NODULATION (ERN1) and regulates their expression (Cerri et al., 2012; Hirsch et al., 2009; Marsh et al., 2007; Middleton et al., 2007; Murakami et al., 2006). Intriguingly, ERN1/ERN2 appear to operate in parallel with NIN to control infection: NIN controls the expression of nodulation-specific genes such as RHIZOBIUM-DIRECTED POLAR GROWTH (RPG), CBS1, NODULATION PECTATE LYASE (NPL), and NF-YA1, and ERN1/2 are required for the expression of ENOD11 and ENOD12, which are also induced during mycorrhization (Garg and Geetanjali, 2007; Journet et al., 2001; Figures 3 and 4).

Figure 4.

The formation and regulation of symbiotic infection structures. A, Steps in infection thread initiation and elongation. Following the molecular dialog between the host plant and rhizobia, the root hair curls and deforms to form an infection chamber. The infection thread (a tubular invagination of the cell wall and PM) originates from the infection chamber and extends anticlinally across the root hair cell. Rhizobia colonize the infection threads as they grow to form a continuous conduit through the root hair cell and underlying cortical cells where the infection threads branch. The process is driven by NF signaling, which activates a set of transcription factors (NSP1/2, NIN, SHR/SCR, DELLAs, and Nuclear Factor Y Subunit A/B1[NF-YA1/B1]) that in turn drive the expression of key structural genes including genes encoding RPG, DREPP, SYFO1, and infectosome components LIN, VAPYRIN, and EXO70H1 and trigger cytoskeletal changes mediated through ROPs-GEFs and components of the SCAR/WAVE complex. The infectosome (LIN/VPY/EXO70H4) directs the polar growth of infection threads. B, The initiation and development of arbuscular mycorrhiza. Following hyphal contact, the epidermal host plant cell forms a prepenetration apparatus (PPA). The hypha, guided by the PPA, extends anticlinally across the cell into the apoplast. The fungal hypha then grows laterally along the root axis where PPA-like structures Figure 4: (continued) direct hyphal entry into cells of the inner cortex. Within these cells, the hypha forms a thick trunk that ramifies into highly branched arbuscules. These events are cued by MF signaling, which activates the transcriptional regulators PHR2, RAM1, and DELLAs. In turn, these transcription factors activate the structural genes required for intracellular accommodation (encoding infectosome components VAPYRIN, EXO70I, and other components involved in endo/exocytosis, including VAMP72, SYP13II, and so on). Arbuscule establishment is dependent on successful host nutrient provision (FatM, RAM1/2, KASI/II/III, STR/STR2, and SWEETS) and acquisition (PT4). The infectosome directs the growth of infection structures, including the formation of arbuscule branches. C, The infectosome complex. The growth of membranes and the reinforcing cell wall-like matrix required for symbiotic infection rely heavily on endo/exocytosis. The exocyst mediates interactions between transport vesicles and the target membrane prior to SNARE-mediated fusion. The fusion of these vesicles with their target sites drives membrane expansion and delivers transmembrane proteins in vesicle membranes and soluble cargo in the vesicle lumen. The functions of the symbiosis-specific proteins VAPYRIN and LIN remain unclear: they might function in vesicle targeting, or they may themselves be cargoes, since they localized to the growing tips of infection threads/arbuscules.

Profiling of root hairs of rhizobia-inoculated L. japonicus ern1 plants showed that ERN1 controls the expression of SPI1, CYM1, and SIR1; these genes are induced in response to both mycorrhizal fungi and rhizobia (Liu et al., 2019c). This suggests that NIN and ERN1/2 operate in parallel to regulate nodulation-specific and common symbiotic branches, respectively, of gene expression to promote rhizobial infection. Both ERN1 and NIN are directly transcriptionally activated by the common symbiotic transcription factor IPD3/CYCLOPS (Yano et al., 2008; Singh et al., 2014; Cerri et al., 2017). In M. truncatula, ERN1 and NIN are expressed independently, while in L. japonicus, when induced by rhizobia, the increased expression of ERN1 does not require NIN, but the induction of NIN appears to require ERN1, although this would ideally be tested using multiple ern1 alleles (Yano et al., 2017; Liu et al., 2019c).

Another GRAS family member, REQUIRED FOR ARBUSCULAR MYCORRHIZATION 1 (RAM1) is AM-specific and is required for AM colonization. RAM1 regulates the expression of RAM2, which encodes a glycerol-3-phosphate acyltransferase that is essential for the AM symbiosis (Gobbato et al., 2012; Park et al., 2015). DELLA plays a similar role in AM interactions, bridging interactions between MtIPD3 and the downstream GRAS protein DELLA INTERACTING PROTEIN1 (DIP1) to promote AM symbiosis (Yu et al., 2014). DIP1 also interacts with RAM1, suggesting that analogous IPD3/GRAS complexes operate in AM (DELLA/RAM1/DIP1) and nodule symbioses (DELLA/NSP1/NSP2), perhaps by integrating negative GA-related inputs via DELLA (Pimprikar et al., 2016; Yu et al., 2014). Further evidence for this comes from MIG1, a GRAS protein that interacts with DELLA1 in the cortex and is required for cortical radial cell expansion during mycorrhization, a process that is sensitive to GA (Heck et al., 2016). REQUIRED FOR ARBUSCULE DEVELOPMENT 1 (RAD1), like DELLA1, is required for AM symbiosis, and its transcript levels increase during AM colonization (Xue et al., 2015; Rey et al., 2017; Figures 3 and 4). NSP1/2 also contribute (to a lesser degree) to mycorrhizal colonization (Catoira et al., 2000; Kaló et al., 2005; Smit et al., 2005; Maillet et al., 2011; Nagae et al., 2014). PHOSPHATE STARVATION RESPONSES (PHRs) are master regulators of AM symbiosis, potentially regulating the expression of 87% of AM-regulated genes. OsPHRs directly interact with the promoters of genes encoding the key transcription factors OsRAM1 and OsWRI5A, and the transporters OsPT11 and OsAMT3;1 via P1BS elements (Shi et al., 2021).

Symbiotic infection and accommodation of mycorrhiza and rhizobia

Legume plants use the common symbiosis pathway to respond to similar symbiotic signals such as NF and MFs from microbes (Oldroyd, 2013). This pathway results in two different forms of symbiosis involving the formation of distinct anatomical structures: AM fungi are accommodated by legumes in differentiated cortical cells, whereas rhizobia initiate the de novo formation of nodules (Parniske, 2008; Oldroyd et al., 2011). The evolution of nodulation, which serves to promote the acquisition of fixed N, from the AM symbiosis, which is centered on the procurement of P, required a switch from phosphorus (P)-driven regulation to control by the plant’s nitrogen (N) systems.

Regulation of rhizobial accommodation

Nodule development

The legume nodule is a functionally specialized organ that originates from root cortical and vascular cells (Timmers et al., 1999; Gustavo Gualtieri, 2000; Xiao et al., 2014). Developmentally, the legume cortex is distinct compared to nonlegumes: it can de-differentiate in response to symbiotic signals from rhizobia or from phytohormones, thereby enabling de novo organogenesis of nodules to accommodate the N-fixing rhizobia. In general, nodule formation in legumes occurs predominantly opposite xylem poles, a phenomenon controlled by ethylene (Penmetsa et al., 2003, 2008). Legume nodules, which are classified as determinate (L. japonicus) or indeterminate (M. truncatula), are postembryonic organs that consist of cortex-derived cells and multiple vascular strands located at the periphery of the nodule that originate from the root vascular bundle (Xiao et al., 2014). Determinate nodules initiate from the middle (L. japonicus) or outer (common bean [Phaseolus vulgaris]) root cortex (van Spronsen et al., 2001). In contrast, both inner and middle cortical cells become mitotically activated during indeterminate nodule initiation (Timmers et al., 1999; Xiao et al., 2014). Here we highlight progress on our understanding of how nodule organogenesis is coordinated through the integration of symbiotic and phytohormone signaling.

The transcription factors NSP1, NSP2, and NIN, which are essential for both rhizobial infection in the epidermis and cell division in the cortex, are also required for spontaneous nodulation induced by autoactive forms of CCaMK/DMI3 and LHK1/CRE1 (Schauser et al., 1999; Catoira et al., 2000; Oldroyd and Long, 2003; Smit et al., 2005; Kaló et al., 2005; Gleason et al., 2006; Tirichine et al., 2007; Marsh et al., 2007; Hayashi et al., 2010; Madsen et al., 2010). NIN mutations abolish nodule organogenesis and infection thread formation, and its overexpression is sufficient to induce local cell divisions in the absence of rhizobia (Marsh et al., 2007; Schauser et al., 1999; Soyano et al., 2013; Vernie et al., 2015). The transcriptional induction of NIN by CYCLOPS (which binds to its cis-element CYC-RE) is also crucial for nodule organogenesis (Yano et al., 2008; Singh et al., 2014). Intriguingly, the regulation of NIN expression during organogenesis requires the actions of distal elements. The L. japonicus daphne mutant (and its M. truncatula counterpart daphne-like) contains different mutations in a distal enhancer that prevent proper cytokinin-mediated cortical activation of NIN, leading to normal epidermal infection but the lack of nodules (Liu et al., 2019b; Yoro et al., 2020, 2014). Both the CCAAT-box complex subunits NF-YA1/NF-YB1 and LBD16 are direct regulatory targets of NIN in L. japonicus (Soyano et al., 2013). Ubiquitous co-expression of NF-YA1, NF-YB1, and LBD16 led to the formation of infected nodules on L. japonicus daphne roots and the production of pseudonodules in the absence of rhizobia in wild-type roots (Soyano et al., 2019).

Two parallel signaling pathways, auxin/PLETHORA (PLT) and SHORT ROOT (SHR)/SCARECROW (SCR), are important for nodule organogenesis. PLTs are expressed in the nodule primordia. RNAi-mediated knockdown of PLTs in M. truncatula significantly decreased nodule number and led to the development of aberrant nodules with inactive meristems (Franssen et al., 2015). The SHR/SCR module is required for cortical cell divisions during nodule organogenesis (Dong et al., 2021). The cortical expression of SHR/SCR is crucial for the activation of cortical cell division induced by overexpression of NIN or by the exogenous application of cytokinin, suggesting that SHR/SCR acts downstream of cytokinin-NIN signaling (Dong et al., 2021). Several cytokinin response regulators were also induced by NF treatment, and depletion of their function significantly decreased nodule number (Gonzalez-Rizzo et al., 2006; Op den Camp et al., 2011; Ariel et al., 2012; van Zeijl et al., 2015; Tan et al., 2020). The current data suggest that the cortical SHR/SCR module predisposed legume cortical cells for cytokinin-induced divisions, after which LBD16 was recruited from lateral root development for nodule organogenesis (Dong et al., 2021;  Soyano et al., 2019). Intriguingly, the M. truncatula double mutant scr scarecrow-like23 (scl23), like the compact root architecture 2 (cra2) mutant, produces fewer nodules but more lateral roots, suggesting that fundamentally different mechanisms control how these lateral organs form (Dong et al., 2021; Huault et al., 2014). This difference may lie in the fact that lateral roots initiate through the action of auxin, whereas nodule initiation is a cytokinin-mediated process. However, the relationship between nodule organogenesis and lateral root formation remains to be elucidated.

Symbiotic infections

Downstream of the symbiotic signaling events and the resulting transcriptional cascades are the structural genes that mediate the intracellular accommodation process, that is, the formation of tubular intracellular infection structures composed of a cell wall like matrix that allows transcellular colonization of AM hyphae or rhizobia (Genre et al., 2008; Tsyganova et al., 2021). The formation and growth of these structures require the ongoing activation of the common symbiotic pathway, which is accompanied by nucleus-associated Ca²⁺ oscillations (Marie et al., 1992; Den Herder et al., 2007). During both symbioses, anticlinal cytoplasmic bridges form. These bridges, resembling phragmosomes that form in premitotic cells, predict the future paths of infection threads or hyphal passage sites (Murray, 2016). During nodulation, the formation of these preinfection structures in root hairs cells is associated with increased expression of cell cycle-related genes, including genes encoding cyclins, CDPKs, and the mitotic kinase AURORA1 (Breakspear et al., 2014). While none of these cell cycle components have yet been linked to mycorrhiza, the accumulation of the cell plate anchoring factor T-PLATE at AM fungal infection sites and ectopic cell divisions in AM roots have been reported (Russo et al., 2019). The first structural infection gene shown to be required for both nodulation and mycorrhization was VAPYRIN (Murray et al., 2011). VAPYRIN was later shown to form a complex with LUMPY INFECTION (LIN) and exocyst EXO70 subunits (EXO70s) and localize to the tips of growing infection threads (Liu et al., 2019a, 2021b; Figure 4). The role of another essential component, RPG, which is specific to nodulation, remains to be resolved (Arrighi et al., 2008; Griesmann et al., 2018). The importance of interactions within this “infectosome” complex, for instance whether VAPYRIN serves mainly as a scaffold protein, and which (if any) members of the complex are targeted by the LIN’s E3-ligase activity, remain unresolved.

A role for cytoskeleton

Secretory vesicle transport to the site of exocytosis is mediated by microtubules and actin filaments. Given its housekeeping functions, actin is expected to play many roles in the symbiosis processes, including nuclear movement and cell growth and division. Other roles for actin may be more specific to the symbiosis, such as in the redirection of the polar growth of root hairs in response to NFs (Fournier et al., 2015; Fournier et al., 2008), which is associated with a rapid increase in the number of F-actin plus ends and their re-localization to infection thread initiation sites (Zepeda et al., 2014). F-actin assembly in plants can be regulated by formins and/or the ARP2/3 complex under the control of plant-specific Rho-GTPases (ROPs; Craddock et al., 2012;; Henty-Ridilla et al., 2013; Feiguelman et al., 2018), both of which have been implicated in rhizobial infection. Several members of the SCAR/WAVE complex, which promotes actin nucleation, were identified in early forward genetic screens to identify infection mutants (Gavrin et al., 2015; Hossain et al., 2012; Qiu et al., 2015; Yokota et al., 2009), while reverse genetics has been used to implicate several the Rac/ROP small GTPases, a class of regulators essential for F-actin-mediated processes including tip growth of pollen tubes (Fu et al., 2001).

In M. truncatula, G. max, and L. japonicus, the deregulation of ROP-Guanine-nucleotide Exchange Factor (GEF) signaling is known to interfere with rhizobial infection (Ke et al., 2012; Kiirika et al., 2012; Wang et al., 2014b, 2021a; Lei et al., 2015; Ke and Peng, 2019; Gao et al., 2021). In contrast, the role of ROP-GEF signaling in mycorrhiza formation has not been extensively investigated. Interestingly, silencing of MtROP9 promoted colonization by AM and the fungal pathogen A. euteiches but negatively affected rhizobial infection (Kiirika et al., 2012). Some insights have been gained into how ROP-GEF signaling modules interface with symbiotic signaling; LjNFR5 interacts with LjROP6 (Ke et al., 2012), and GmNFR1a and GmNFR5a were recently shown to interact with GmROP9 (Gao et al., 2021). Moreover, GmNFR1a phosphorylates GmGEF2a, which promotes its activation of GmROP9, shedding light on how NF signaling directs cytoskeleton remodeling during the re-direction of root hair growth and infection thread formation. The L. japonicus ortholog of the Arabidopsis GEF protein SPIKE1, which regulates the SCAR/WAVE complex to control actin nucleation (Basu et al., 2008), activates LjROP6 (Liu et al., 2020a). Ljrop6 mutants form defective infection threads and have root hairs featuring disordered actin filaments. This finding suggests that ROP-GEFs may act through the WAVE/SCAR complex to modulate changes in actin during infection, but direct links remain to be demonstrated. These or other ROP-GEF modules may mediate other infection processes, such as NF or hormonal signaling.

Relative to actin, progress in our understanding of the role of microtubules in infection has lagged. However, the recent characterization of the microtubule-binding protein DEVELOPMENTALLY REGULATED PLASMA MEMBRANE POLYPEPTIDE (DREPP) provides a starting point. DREPP localizes to symbiosis-specific membrane nanodomains, and both mutation of DREPP and its overexpression decreased rhizobial infection (Su et al., 2020). Upstream of these events, a role for SYMBIOTIC FORMIN1 (SYFO1) has recently been identified (Liang et al., 2021). Class I formins are evolutionarily conserved integral membrane proteins with roles in actin and microtubule organization that mediate interactions between the plasma membrane and cell wall. Loss of SYFO1 resulted in fewer rhizobia-induced deformations and infection structures, the latter of which had normal morphology, suggesting a role in the initiation of infections, rather than their growth per se. Based on their findings, the authors proposed a role for SYFO1 in the coordination of the cell wall and plasma membrane during rhizobium-induced focal membrane deformations.

In addition to the importance of these studies in understanding rhizobial infection mechanisms, this research vein offers a special opportunity to plant biologists to study the function of the cytoskeleton and its interactions with the membrane and cell wall in a system where the re-direction of cell growth can be reliably and rapidly cued and is easily observed. In comparison, few actin/microtubule-related genes have been implicated in mycorrhiza formation, but a recent study showed a role for the microtubule-associated protein Tomato Similar to SB401 in the formation and function of arbuscules (Ho-Plagaro et al., 2021; Figure 4).

Arbuscule development

After penetrating into roots, the mycorrhizal fungal hyphae spread intercellularly in the inner cortex and enter cortical cells to repeatedly dichotomously branch, resulting in the formation of arbuscules, which are surrounded by the host-derived periarbuscular membrane (PAM). In arbuscule-containing cells, the ER surrounds the arbuscule branches, the Golgi bodies and peroxisomes relocate alongside the branched hyphae, and the vacuoles become highly invaginated to accommodate arbuscule development (Pumplin and Harrison, 2009). The microfilaments form a dense network that coats the hyphal branches, and the microtubules form bridges between adjacent fungal branches (Genre and Bonfante, 1998). The plastids, the source of lipid biosynthesis, increase in number and form a complex network surrounding the arbuscules (Fester et al., 2001). The number of mitochondria also increases, and they aggregate in the vicinity of arbuscules (Lohse et al., 2005). Besides arbuscule formation, mycorrhizal colonization also activates ectopic cortical cell divisions (Russo et al., 2019). Live cell imaging revealed that the PAM is composed of two distinct parts: the “trunk domain” labeled by MtBCP1 and AtPIP2a, which resembles the plasma membrane; and the “branch domain” labeled by the symbiotic-specific phosphate transporter inorganic phosphate [P(i)] Transport protein 4 (MtPT4), which is clearly distinct from the plasma membrane (Pumplin and Harrison, 2009).

The peri-arbuscular membrane in well-developed arbuscules provides the interface for bidirectional nutrient exchange, where carbon provided by the host plant is exchanged for mineral nutrients, particularly phosphate from the AM fungi. FAs synthesized in host plants are transported to AM fungi as a major organic carbon source (Jiang et al., 2017; Luginbuehl et al., 2017; Figure 5). Genes involved in FA biosynthesis and transport in the host plant are induced by mycorrhizal infection and are required for mycorrhizal symbiosis, including genes encoding Fatty acyl-ACP thioesterase M (FatM; Bravo et al., 2017; Jiang et al., 2017), RAM2 (Wang et al., 2012), β-Keto-Acyl ACP Synthase I (KASI; Keymer et al., 2017), KASII (Jiang et al., 2017), KASIII (Keymer et al., 2017), and STR/STR2 (Zhang et al., 2010; Gutjahr et al., 2012). In the plastids of arbusculated cells, during FA biosynthesis, KASIII catalyzes FA elongation from C2:0-ACP to C4:0-ACP, the C4:0-ACP is then elongated by KASI to C16:0-ACP (Keymer et al., 2017). ACP is then cleaved from 16:0-ACP to release 16:0 FAs via FatM. The 16:0 FAs are converted to 16:0 β-monoacylglycerol in the ER, a process catalyzed by RAM2 (Bravo et al., 2017;  Wang et al., 2012). Finally, 16:0 β-monoacylglycerol is thought to be exported to the peri-arbuscular interface by the half-size ATP-binding cassette (ABC) transporters STR/STR2 (Jiang et al., 2017). Mutants of FatM, RAM2, KASI, and KASII display defective arbuscules due to the lack of lipid transfer to the mycorrhizal fungi (Zhang et al., 2010; Gutjahr et al., 2012; Wang et al., 2012; Groth et al., 2013; Bravo et al., 2016, 2017; Jiang et al., 2017; Keymer et al., 2017; Luginbuehl et al., 2017; Figures 4 and 5).

Figure 5.

Mature arbuscules and nodules. A, Mature arbuscules show thin and higher-order branching structures surrounding by a plant-derived PAM. The PAS, the apoplastic region between the fungal plasma membrane and PAM, is the site of nutrient exchange between the host plant and AM fungi. Three transport channels are located in the PAM: STRs, which are thought to be responsible for transporting the FAs synthesized by plants to the PAS; PT4, which transports the phosphorus released by the AM fungi to the host plant, a process that requires a proton gradient generated by the HA1 for phosphorus uptake; and MST2, a host MST that contributes to arbuscule development. FAS, FA synthase; 2-MAG, 2-Monoacylglycerol; B, The gene regulatory gradient in M. truncatula indeterminate nodules. SHR1/2, SCL23, ERN1, NSP1, and NSP2, along with CCAAT-box components NF-YA1, NF-YB16/18 drive cell division and support infection in the meristem and infection zone. NIN, supported by its transcriptional activator IPD3, acts as a master regulator of infection, but also plays an important role in the N-fixation zone, an activity that requires DNF1. Precise activation of DME expression in the interzone reshapes the methylome to allow the expression of genes essential for N fixation, including NLP2 which, along with NIN, promotes the expression of leghemoglobins and NCRs; leghemoglobins create the microaerobic environment required for biological nitrogen fixation.

Intriguingly, FAs induce hyphal growth and secondary spore formation by mycorrhizal fungi under asymbiotic conditions (Kameoka et al., 2019; Sugiura et al., 2020). Additionally, a monosaccharide transporter (MST2) from the AM Glomus sp. takes up sugars in arbuscules and is required for arbuscule development (Helber et al., 2011). Several sugar transporters (SWEETs) in various plant species are also induced during mycorrhizal symbiosis (An et al., 2019; Manck-Gotzenberger and Requena, 2016; Zhao et al., 2019), but the mutation of SWEET1b in M. truncatula resulted in no significant symbiotic defects (An et al., 2019). Altogether, this data suggest that the carbon transferred from plants to mycorrhizal fungi is mainly in lipid form.

The most recognized benefit of AM symbiosis for host plants is improving P(i) nutrition; for example, mycorrhizal-colonized rice receives >70% of its P(i) through the AM symbiosis (Yang et al., 2012). Phosphate is acquired by a mycorrhizal phosphate transporter expressed in extraradical hyphae (Harrison and van Buuren, 1995; Xie et al., 2018; Maldonado-Mendoza et al., 2001). The phosphate is then translocated from extraradical hyphae to intraradical hyphae in the form of polyphosphate, where it is hydrolyzed into phosphate and released into the periarbuscular space (PAS; Rasmussen et al., 2000; Viereck et al., 2004; Kuga et al., 2008; Tani et al., 2009; Hijikata et al., 2010; Figure 5). P is then taken up by plant cells through phosphate transporter (PHT) proteins, which consist of four subfamilies (PHT1–4) (Rausch and Bucher, 2002). Members of the PHT1 subfamily, including MtPT4 and OsPT11, are transcriptionally induced in AM roots in many plant species such as M. truncatula and O. sativa (Harrison, 2002; Paszkowski et al., 2002). MtPT4 is localized to the PAM and mediates P(i) uptake from the PAS in M. truncatula (Javot et al., 2007). The PT4 family comprises H⁺/P(i) symporter proteins, whose function requires the plasma membrane H⁺-ATPase (HA) to create a proton gradient across the PAM (Krajinski et al., 2014; Wang et al., 2014a).

Bacteroids and arbuscules: analogous structures with overlapping and distinct mechanisms

The purpose of symbiotic infection is the eventual establishment of intracellular nutrient exchange structures. Symbiosomes and arbuscules fulfill corresponding roles, each serving as the membrane interface for nutrient exchange between the host and symbionts (Parniske, 2008). The key difference between the two symbioses at the trophic level is that nutrients obtained via arbuscules can be channeled outside of the plant via intraradical hyphae, whereas bacteroids are strictly enclosed within host cells. Symbiosome function does not require NF signaling (Den Herder et al., 2007), which is reflected in the lack of expression of NF receptors in the N₂-fixation zone of mature Medicago nodules (Roux et al., 2014).

In terms of evolution, nodulation in legumes is thought to have evolved directly from an actinorhizal lineage, which in turn evolved from a mycorrhizal ancestor (van Velzen et al., 2018, 2019). In this series of symbiotic adaptations, there was a stepwise progression of the symbiotic membrane interface, starting with the highly branched mycorrhizal arbuscules that evolved to maximize the nutrient exchange area, which is surrounded by a matrix material (Bonfante et al., 1990; Bonfante and Perotto, 1995). This feature was then modified to accommodate the filamentous N-fixing actinorhiza, whose symbiotic vesicles are also enclosed in a matrix (Newcomb and Wood, 1987; Berg, 1999), and then was finally converted to independent organelle-like bacteroids in the legume progenitor (Coba de la Pena et al., 2017). This evolutionary legacy is evidenced in the genomes of nodulating nonlegumes by the presence of homologs of NIN and RPG in the actinorhizal lineages; these genes are required for infection by rhizobia but not by mycorrhizal fungi (Griesmann et al., 2018;  Wu et al., 2022). Similarly, NODULES WITH ACTIVATED DEFENSE1 (NAD1), which is required for N fixation in M. truncatula, is present in actinorhizal but not mycorrhizal plants (Wang et al., 2016a). Interestingly, legume NIN and RPG homologs contain deletions that are not present in nonlegumes, suggesting protein-level neofunctionalization (Zhao et al., 2021).

Densely packed symbiosomes likely provide increased surface area for nutrient exchange compared to the branching architecture of the arbuscule; compared to the plasma membrane, the arbuscule gives an ∼10-fold enhancement in surface area, while the symbiosome membrane surface in nodules is estimated to be 100 times that of the plasma membrane (Alexander et al., 1989; Brewin, 1990). This advantage, as well as the absence of a surrounding matrix on symbiosomes, which could offer a further advantage in terms of nutrient exchange, may have in part driven the symbiotic switch from actinorhiza to rhizobial nodulation. The symbiosomes apparently form through endocytosis, whereas the primary cellular process required for growing symbiotic infection structures is exocytosis, which is mediated by the cooperative action of the exocyst and SNAP receptor (SNARE) complexes (Figure 4). Endo/exocytosis are tightly coupled processes (Zhang et al., 2019a), as revealed by the overlap of genes required for infection thread formation and symbiosome biogenesis.

Endo/exocytosis

The processes that mediate the membrane invagination required for arbuscular growth and symbiosome formation are, at least in part, shared. Knockdown of VESICLE-ASSOCIATED MEMBRANE PROTEIN 72 (VAMP72) homologs in M. truncatula and L. japonicus blocked formation of both arbuscules and symbiosomes (Ivanov et al., 2012; Sogawa et al., 2018). Similarly, M. truncatula vapyrin mutants, which form only truncated arbuscules, also form defective rhizobial infection threads, eventually allowing the formation of a limited number of normal-looking symbiosomes (Murray et al., 2011). However, this effect on symbiosome formation could be direct (i.e. altered endocytosis) or may solely reflect decreased capacity of the mutant to make infection threads. Some overlap exists in the exocyst components induced by rhizobia and those induced in mycorrhizal roots, such as EXO70I, which decorates the tips of both growing infection threads and arbuscule branches, and VAMPs (Gavrin et al., 2017; Liu et al., 2019a). EXO70I plays a positive role in arbuscule formation, but has no obvious role in nodulation, perhaps reflecting genetic redundancy (Zhang et al., 2015b). Mutation of an EXO70 family member in soybean resulted in reduced nodulation and nonsymbiotic phenotypes (Wang et al., 2016b). Silencing of synaptotagmins, which are thought to mediate Ca²⁺-triggered vesicle fusion of exocyst vesicles, delayed rhizobia release from infection threads and blocked symbiosome maturation (Gavrin et al., 2017). Silencing of the t-SNARE gene SYNTAXIN OF PLANTS 13II (SYP13II) also resulted in defects in the symbiotic interface (Huisman et al., 2016), blocking both arbuscule and bacteroid formation. Mutation of the nonsymbiosis-specific syntaxin gene SYP71 in L. japonicus resulted in plants that formed enlarged symbiosomes (Hakoyama et al., 2012; Figure 4).

No specificity was observed in SNARE interactions of M. truncatula, with all VAMP72s tested being able to interact with SYP123A. However, only two of three nonsymbiotic VAMP72s tested could complement the loss of arbuscule formation in VAMP721d/e RNAi plants, while all SYPs tested restored arbuscule formation in SYP132α knockdown roots (Huisman et al., 2020). In addition, the nonsymbiotic isoform SYP132B was sufficient for normal mycorrhizal interactions (Huisman et al., 2020). Therefore, while the current data do not exclude the possibility of differential channeling of symbiotic cargo through specialized isoforms and specific v-SNARE/t-SNARE interactions, it is possible that the main neofunctionalization in these components has been in their expression patterns rather than in their functions/interactions. Nonetheless, the noninterchangeability of certain components, which has also been documented for EXO70 components in Arabidopsis (Markovic et al., 2021), suggests specialization of the exocyst in plants. Some of the components adopted for symbiotic infection appear to have retained their original roles in plant development, such as VAMP72s, which locate to the apices of growing roots hairs and are required for root hair growth (Fournier et al., 2015; Sogawa et al., 2018).

Maintenance of the symbiotic interface

It is clear that to some extent the endo/exocytotic processes in the two symbioses overlap. Nonetheless, the highly specialized microaerobic environment required for N fixation has driven extensive neofunctionalization of both the host and symbiont (Schulte et al., 2021; Wheatley et al., 2020; Yurgel et al., 2021). This includes the recruitment of different types of transporters to support nutrient exchange, nutrient supply, nutrient export, as well to supply crucial co-factors such as iron, which is needed for N fixation (Brear et al., 2020; Escudero et al., 2020; Kryvoruchko et al., 2018; Liu et al., 2020b; Walton et al., 2020); symbiotic transport has recently been reviewed (Banasiak et al., 2021). Two Nodule Cysteine Rich peptides (NCRs) play important roles in N fixation (Horvath et al., 2015; Kim et al., 2015), but they are limited to legumes with indeterminate nodules with persistent infections (Downie and Kondorosi, 2021) and act directly on the rhizobia (Barriere et al., 2017). Other NCRs have demonstrated roles in blocking infection by specific strains, leading to bacterial cell death and early nodule senescence (Wang et al., 2017; Yang et al., 2017).

Several proteins of unknown function have been implicated in bacteroid formation or maintenance, such as NODULE-SPECIFIC PLAT DOMAIN1, NAD1, DEFECTIVE IN NITROGEN FIXATION2 (DNF2), and Symbiotic CYSTEINE-RICH RECEPTOR KINASE (SymCRK) (Bourcy et al., 2013; Berrabah et al., 2014; Wang et al., 2016a; Domonkos et al., 2017; Pislariu et al., 2019). While the biochemical functions of these proteins remain uncertain, the loss-of-function results in defense-like responses, with strong accumulation of reactive oxygen species and flavonoids and the development of necrosis. Whereas Medicago nad1 mutants show excessively activated innate immunity, leading to cell death in nodules, mutants with knocked out expression of MtCDPK5 and its phosphorylation targets RBOHs showed partial or near-full recovery of the nodule phenotypes, providing insights into the mechanism of NAD1 (Yu et al., 2018). A candidate for regulating the suppression of defense-like responses is NIN, which was recently shown to play a role in symbiosome development (Liu et al., 2021a). The weak nin-16 allele, which supports normal rhizobial infection, forms nodules that accumulate phenolic compounds and have underdeveloped bacteroids, along with decreased expression of SymCRK, NAD1, and REGULATOR OF SYMBIOSOME DIFFERENTIATION, providing evidence that NIN’s role in nodulation is not limited to infection.

Regulating the switch from infection to N fixation

The indeterminate nodules of M. truncatula and other species feature a gradient of development across which cells divide (meristem), become infected by rhizobia (infection zone), switch to endoreduplication with cell enlargement and form symbiosomes (interzone), and fix N (N-fixation zone). The meristem and infection zone both feature dividing cells, functions that are reflected in the expression of the transcriptional regulators SHR1/2, SCL23, NSP1, NSP2, and ERN1. NIN is expressed everywhere except the meristem, peaking in the proximal infection zone, which is mirrored by DNF1.

The interzone, which represents a critical developmental phase in which cells stop expressing genes required for infection and begin to express genes required for N fixation, is marked by expression of the DNA demethylase gene DEMETER (DME), while the genes encoding the NIN-regulator IPD3/CYCLOPS and the NIN-like protein 2 (NLP2) are expressed in the interzone and N-fixation zone (Figure 5). DME is required for the expression of numerous genes expressed in the interzone/N-fixation zone, including leghemoglobins and NCR proteins, suggesting it plays a major role in the transcriptional switch from infection to N fixation (Satge et al., 2016). One of the downregulated genes in the DME-knockdown plants was NLP2. A recent study of nlp2 mutants revealed decreased expression of leghemoglobin and NCR genes. Both NIN and NLP2 bind to the promoters of leghemoglobin genes and activate their expression (Jiang et al., 2021). Another recent study found that NIN is proteolytically cleaved in nodules and that this requires the signal peptidase component DNF1. A NIN mutant that prevented this processing was able to complement the infection phenotype of the nin null mutant but formed small white nodules phenocopying weak nin alleles, suggesting that the cleavage of NIN is required for its switch from activating infection-related genes to controlling genes required for N fixation (Feng et al., 2021). These recent findings suggest that in addition to its role in infection, NIN acts cooperatively with NLP2 and DME to activate the expression of genes required for N fixation.

Autoregulation of nodulation and mycorrhization

Symbiotic interactions with both mycorrhizal fungi and rhizobia are dynamically coordinated with external nutrient availability. N and phosphate-deficient soils promote the establishment of symbiosis with rhizobia and mycorrhiza, respectively. To optimize nodule number, legumes have a systematic regulatory mechanism termed autoregulation of nodulation (AON; Kosslak and Bohlool, 1984; Delves et al., 1986; Reid et al., 2011b; Nishida et al., 2018a, 2018b), and recent discoveries suggest this system also operates in arbuscular mycorrhization.

In the past two decades, the molecular basis of AON has become increasingly clear. The hallmark of AON initiation is the production of the small peptides in roots belonging to the CLAVATA (CLV) 3/EMBRYO SURROUNDING REGION (CLE) family. These CLE peptides are widely distributed in the genomes of legumes (Hastwell et al., 2017), including GmRIC1 and GmRIC2 in soybean (Reid et al., 2011a, 2011b; Ferguson et al., 2014), LjCLE-RS1 and LjCLE-RS2 in L. japonicus, MtCLE12 and MtCLE13 in M. truncatula, and PvRIC1 and PvRIC2 in P. vulgaris (Okamoto et al., 2009; Mortier et al., 2010; Reid et al., 2011a, 2011b; Ferguson et al., 2014; Figure 6). The production of CLE peptides follows the early infection events of rhizobia, nodule development, and even NF treatment (Caetano-Anolles and Gresshoff, 1990; van Brussel et al., 2002; Li et al., 2009; Okamoto et al., 2009; Reid et al., 2011a). CLEs are also produced in response to high nitrate levels to help balance shoot and root growth (Araya et al., 2016; Mens et al., 2020). In soybean and L. japonicus, nitrate application triggers CLE peptide accumulation in roots (GmNICs, LjCLE-RS2/RS3, and LjCLE40) (Okamoto et al., 2009; Reid et al., 2011a, 2011b; Lim et al., 2014; Nishida et al., 2016). Although these nitrate-induced CLEs (NICs), like those induced by rhizobia, are perceived by GmNARK (Lim et al., 2014; Reid et al., 2011a, 2011b), this effect appears to act locally rather than systemically, as nodulation was inhibited in roots overexpressing GmNIC1 but was normal in wild-type roots of the same plants (Reid et al., 2011a, 2011b).

Figure 6.

The autoregulation of nodulation. In nitrogen-deficient soil, legume plants secrete flavonoids that attract compatible rhizobia to attach to their root surfaces and induce the rhizobia produce NFs. In soybean, NFR1a and NFR5a recognize NFs and activate downstream components, such as the transcription factor NIN and key symbiotic gene ENOD40, to ensure the establishment of root nodule symbiosis. Rhizobia-Induced CLE (RIC) peptide production occurs following the early rhizobial infection events. RICs are transported by the xylem to the shoot where they are perceived by a receptor complex consisting of NARK/HAR1/SUNN and/or CLV2, KLV, and CRN. KAPP1 and KAPP2 can be phosphorylated by NARK and in turn, reduce its phosphorylation. After RIC perception, shoot-derived regulators (cytokinin and miR2111) are synthesized in leaves and transported by the phloem to the root to suppress nodule formation. TML, a suppressor of nodule emergence, is directly regulated by miR2111. NNC1, rhizobia inoculation, and NF treatment can induce miR172c upregulation, which represses the expression of NNC1 to release the inhibition of ENOD40 expression, leading to nodule formation. NARK negatively regulates miR172c expression via the activity of shoot-derived cytokinin. Low N conditions trigger the production of CEPs, which are perceived by CRA2 in shoots, resulting in higher miR2111 levels to permit nodule formation. High soil nitrate promotes the production of NIC peptides and RICs. Unlike RICs, NICs are recognized locally by NARK in roots. Light is also a primary regulator of nodulation; light induces the synthesis of FTs and Soybean TGACG-motif binding Factors (STFs) in shoots, which move to roots and activate the expression of symbiosis-associated genes, such as NIN, NF-YA1, and NF-YB1. STF, light-induced TGACG-MOTIF BINDING FACTOR3/4; CRY, Cryptochrome; NARK, CLV1-like leucine-rich repeat receptor kinase; NRSYM1, Nitrate unresponsive symbiosis 1.

CLE peptides are transported through the xylem from roots to shoots where they are recognized by a receptor complex that consists of a leucine-rich repeat receptor-like kinase termed GmNARK/LjHAR1/MtSUNN/PvSym29 and by other components, including CLV2, CORYNE (CRN), and KLAVIER (KLV; Nishimura et al., 2002; Searle et al., 2003; Schnabel et al., 2005; Krusell et al., 2011; Ferguson et al., 2014). Mutation in these receptors results in hyper/supernodulation. The posttranslational modification of CLE peptides, such as triarabinosylation, is required for their binding efficiency and biological activity (Shinohara and Matsubayashi, 2013; Hastwell et al., 2015, 2019; Corcilius et al., 2017). Perception of these root-derived peptides induces GmNARK phosphorylation, which can be reduced by two phosphatases: KINASE-ASSOCIATED PROTEIN PHOSPHATASE1 (KAPP1) and KAPP2 (Miyahara et al., 2008). GmNARK/LjHAR1 activation promotes the synthesis shoot-derived inhibitors (SDIs) that move to the roots through the phloem. The kelch repeat-containing F-box protein TOO MUCH LOVE (TML) is specifically expressed in roots and functions as a downstream signaling component; tml mutants show a hypernodulation phenotype (Magori et al., 2009; Takahara et al., 2013; Figure 6).

The AON and nodulation signaling pathway are tightly interlinked and reciprocally regulated. NIN plays an essential role during the early steps of nodule initiation by activating LjCLE-RS1/2 expression (Marsh et al., 2007; Schauser et al., 1999; Soyano et al., 2014). In turn, AON negatively regulates NIN expression and NIN-activated cortical cell division, and thus they form a regulatory loop (Soyano et al., 2014). miR172s function as pivotal regulators that fine tunes the nodulation signaling pathway and AON, and the expression of soybean miR172c is induced by rhizobial inoculation (Wang et al., 2009, 2014c; Yan et al., 2013). The effect of miR172 might be exerted through it cleavage of the mRNA of the AP2 transcription factor NODULE NUMBER CONTROL1 (NNC1; Mathieu et al., 2009; Sakuma et al., 2002; Saleh and Pages, 2003), preventing the direct activation of GmENOD40, GmRIC1, and GmRIC2 expression by NNC1 (Wang et al., 2014c, 2019). GmNINa upregulates the expression of miR172c (Wang et al., 2014c, 2019). Therefore, together, the GmNINa–miR172c/NNC1 module functions as a junction between the nodulation and AON signaling pathways.

Mounting evidence points to miR2111, which is transported from shoots to roots where it targets TML, as one of the key SDIs of AON (Xu et al., 2013; Zhang et al., 2014, 2021; Tsikou et al., 2018). Conversely, C-terminally Encoded Peptide (CEP) peptides promote nodulation via their action on the CRA2 receptor (Imin et al., 2013). The CRA2 receptor is required to maintain a high level of miR2111 expression in shoots to promote the ability of roots to nodulate under low N conditions (Gautrat et al., 2020). Certain CEPs and CLEs are regulated by NIN (Laffont et al., 2020). In addition to symbiotic regulation, these peptides are also influenced by N availability. Nitrate regulation of nodulation in legumes occurs through NLP1 and NLP4 (Lin et al., 2018; Nishida et al., 2021), and NLP1 directly activates nitrate-dependent CLE expression (Luo et al., 2021). The emerging picture is that NLP transcription factors integrate symbiotic and nutrient signals to control nodule number through small peptide signaling (Figure 6).

Cytokinin also acts as an SDI that contributes to AON suppression. In L. japonicus, ISOPENTENYLTRANSFERASE3 contributes to the shoot-derived cytokinin biosynthesis, which inhibits nodule formation, and its activation is dependent on LjCLE-RS1/2-LjHAR1 signaling (Sasaki et al., 2014). Other plant hormones can also act as SDIs. The application of methyl jasmonate to shoots inhibits infection by rhizobia, but this suppressive effect is independent of LjHAR1 (Nakagawa and Kawaguchi, 2006).

Phosphate is a major limiting factor for nodule development and symbiotic N fixation (Cabeza et al., 2014; Chen et al., 2011; Hernandez et al., 2009; Nasr Esfahani et al., 2017; Sulieman et al., 2013). Intriguingly, P deficiency conditions activate the AON signaling pathway to inhibit nodule development (Isidra-Arellano et al., 2020). The light signal also helps regulate nodulation. In G. max, light-induced TGACG-motif binding factor 3/4 (GmSTF3/4, homologous proteins) and FLOWERING LOCUS T (GmFTs) move from shoots to roots. Rhizobia activate CCaMK, which phosphorylates GmSTF3 to promote GmSTF3–GmFT2a complex formation, which in turn regulates the expression of NIN, NF-YA1, and NF-YB1 (Wang et al., 2021b). Interestingly, the HY5 homolog ASTRAY/Ljsym77 plays a negative role in determining root nodule number in L. japonicus (Nishimura et al., 2002). The mechanisms underlying these differences in nodulation phenotypes need to be further explored.

In M. truncatula, AM colonization and high phosphate conditions also induce CLE expression, suggesting that CLE peptides could be regulators of AM symbiosis (Le Marquer et al., 2019). Furthermore, overexpression of either MtCLE33 or MtCLE55 inhibited mycorrhization and reduced strigolactone levels in a SUNN-dependent manner (Muller et al., 2019). Shi and coworkers recently found that PHR2 acts as a global regulator that orchestrates AM symbiosis, P(i) nutrient capture, and plant development (Shi et al., 2021), indicating that PHR2 is a key component of the autoregulation of mycorrhization. Altogether, it appears that the CLE perception module integrates symbiotic and nutrient signals to control the level of endosymbiotic engagement. A future challenge will be to dissect the mutual regulatory mechanisms between PHR2 and CLE peptides.

Future prospects

As rhizobia and AMF are microbes, they might be initially recognized by host plants as potential pathogens. Their MAMPs could be recognized by host plants and cause immune responses that, if left unchecked, would hamper the establishment of symbiosis (Kouchi et al., 2004; Lohar et al., 2006; Stacey et al., 2006; Libault et al., 2010; Lopez-Gomez et al., 2012). In compatible interactions, host immune responses are transient and suppressed during infection (Cardenas et al., 2008; Nakagawa et al., 2011). Rhizobia and AMF must employ various strategies to evade or suppress host plant immunity during the early stage of symbiosis (Hubber et al., 2004; Scheidle et al., 2005; Aslam et al., 2008; Okazaki et al., 2010, 2013; Kloppholz et al., 2011; Lopez-Gomez et al., 2012; Liang et al., 2013; Margaret et al., 2013; Gourion et al., 2015). NFs and MFs inhibit MAMP-triggered immunity in both legume and nonlegume plants (Liang et al., 2013; Feng et al., 2019; Rey et al., 2019; Zhang et al., 2021), but the detailed mechanism needs to be further studied. How do plants engage with beneficial microorganisms while at the same time restricting pathogens? Studies performed to date have tended to uncouple plant symbiosis and immune signaling pathways and dissect just one of these processes. In future studies, these two signaling pathways need to be jointly considered in order to obtain a complete picture of the signaling pathways governing symbiotic interactions.

AM fungi are important components of the soil ecosystem; the extra-radical mycelium extends into the soil and acquires nutrients for the plant. AM symbiosis promotes the accumulation of rhizobia in the rhizosphere (Wang et al., 2021c), suggesting that the large surface area of the extraradical mycelium might provide a nutrient-rich niche to support the colonization and growth of rhizobial communities. The study of microbes associated with mycorrhizal fungi and their role as the third component of symbiosis is thus a promising field.

Downstream of Ca²⁺ signaling and its interpreters, the common signaling cascade sharply diverges, with NIN mediating the expression of a large proportion of nodulation-specific structural genes, including genes encoding the CCAAT-box heterocomplex, RPG1, and NPL, which play decisive roles in infection. PHR2 appears to be NIN’s counterpart in mycorrhization, controlling the expression of genes encoding the key transcription factors RAM1 and WRI5, which control both epidermal penetration and arbuscule establishment and function (Shi et al., 2021). The recruitment of PHR2, a central regulator of P homeostasis, reflects the profound importance of the symbiosis for P acquisition. In legumes, the recruitment of NIN, a member of the NLP family that controls nitrate homeostasis, may be related to the central role of nitrate in energy generation under hypoxia, a central requirement for efficient N fixation (Jiang et al., 2021; Feng et al., 2021). These two transcription factors together represent the evolutionary impetus behind the symbioses: the acquisition of P and N, the two key growth-limiting nutrients in plants. Therefore, fully understanding how these key factors are differentially regulated is essential to address how two systems can be controlled through a common signaling pathway. In terms of evolution, studies are needed to elucidate the details of whether and how nodulation in legumes evolved from an actinorhizal lineage, and how these actinorhizal plants in turn evolved from a mycorrhizal ancestor, in order to fill the evolutionary gap between mycorrhization and nodulation.

Supplemental data

The following material is available in the online version of this article.

Supplemental File S1. The protein sequences used to construct the phylogenetic tree.

 

These authors contributed equally (D.W. and W.D.).

All authors contributed to writing the article.

The authors responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (https://academic.oup.com/plcell) are: Ertao Wang (etwang@cemps.ac.cn).