Rice arbuscular mycorrhiza as a tool to study the molecular mechanisms of fungal symbiosis and a potential target to increase productivity

Rice (Oryza sativa L.) is a monocot model crop for cereal molecular biology. Following the emergence of molecular genetics of arbuscular mycorrhizal (AM) symbiosis in model legumes in the 1990s, studies on rice genetic resources have considerably contributed to our understanding of the molecular mechanisms and evolution of root intracellular symbioses. In this review, we trace the history of these studies and suggest the potential utility of AM symbiosis for improvement in rice productivity.


Introduction
Arbuscular mycorrhizal (AM) symbiosis is an ancient endosymbiosis that originated more than 400 million years ago (Remy et al. 1994). Most land plants engage in a cooperative relationship with obligate biotrophic fungi of the phylum Glomeromycota. Strigolactones (SLs) secreted by plant roots into the rhizosphere act as branching factors for AM fungal hyphae (Akiyama et al. 2005). Fungal exudates induce symbiotic Ca 2+ spiking in the host cells, which is crucial for AM fungal infection. AM fungi penetrate the root epidermis through hyphopodia. The pre-penetration apparatus (PPA), an intracellular pre-infection structure, is formed in the host cells to determine the penetration path of fungal hypha from the epidermis to the cortex (Genre et al. 2005, Oldroyd 2013, Recorbet et al. 2013. A highly branched hyphal structure, an arbuscule, develops in the cortical cells and serves for nutrient exchange between the host plant (which provides photosynthates) and the AM fungus (which provides mainly phosphate along with other nutrients). External hyphae growing from the mycorrhizal roots allow host plants to assimilate phosphate from outside the root zone ( Fig. 1) and therefore to survive in phosphorus-deficient soils.

Identification of AM-specific phosphate transporters
Identification of AM-specific phosphate transporters was the starting point of molecular genetic research on AM symbiosis in rice. Using genome-wide data for rice (Oryza sativa L.), Paszkowski and colleagues have identified 13 phosphate transporter genes (PT) and determined that OsPT11 is specific to AM symbiosis (Paszkowski et al. 2002). The OsPT11-GFP fusion protein is specifically localized in the periarbuscular membrane surrounding young and mature arbuscules, where active phosphate transfer seems to occur (Kobae and Hata 2010). Using OsPT11 and OsPT13 knockout and knockdown mutants, Yang et al. (2012) have shown that not only OsPT11 but also OsPT13 is involved in AM symbiosis development. Although OsPT13 is essential for AM symbiosis, symbiotic phosphate uptake is independent of OsPT13. These results indicate functional specialization: OsPT11 may be responsible for both AM development and symbiotic phosphate uptake, whereas OsPT13 may act as a sensor to detect the phosphate level appropriate for arbuscule development (Yang et al., 2012).

Rice as a model monocot plant to study the evolution of root nodule and AM symbioses
Since the emergence of two legume models during 1990's, Lotus japonicus and Medicago truncatula, the door for analysis of molecular mechanisms of root symbioses has been opened. Legumes have two mutualistic symbiosis systems, root nodule (RN) and AM symbiosis.
Before the molecular genetics era in legume symbiosis, the presence of host plant genes governing both RN and AM symbioses has been suggested (Resenders et al. 2001). In line with this, over half of non-nodulation mutants of L. japonicus and M. truncatula also showed non-mycorrhization phenotypes , Oldroyd 2013. A set of genes shared by both symbioses was assumed to encode key components of the "common symbiosis pathway" (CSP) that regulates intracellular infection with rhizobial bacteria and AM fungi. Extensive studies since the 2000s have revealed that these CSP components function as activators of Ca 2+ spiking (LjSYMRK/MtDMI2, LjCASTOR, LjPOLLUX/ MtDMI1, LjNUP85, LjNUP133, and LjNENA), a decoder of symbiotic Ca 2+ signals (LjCCaMK/MtDMI3), and activators of downstream symbiotic signaling pathways (LjCYCLOPS/MtIPD3, MtVapyrin, LjNSP1, MtNSP2, and LjCerberus, reviewed by Kouchi et al. 2010 andOldroyd 2013).
AM symbiosis is formed between the majorities of plant species, while RN symbiosis is limited to a clade within Eurosid I. Thus, it has been proposed that RN symbiosis has originated from AM symbiosis by using basic CSP and a newly acquired set of genes for the RN specific pathway (Kistner and Parniske 2002). To examine this hypothesis, comparative analysis of orthologous genes between legumes and non-legumes became essential. Among the non-leguminous plants, rice is the most suitable monocot model plant, because its genome databases and mutant resources have already been established (Miyao et al. 2003, Hirochika et al. 2004). Zhu et al. (2006) reported the presence of putative rice orthologs of the legume genes NFR1, NFR5, SYMRK/ DMI2, CASTOR, POLLUX/DMI1, DMI3, MtNSP1, and MtNSP2. Studies on knockout rice mutant lines showed that the OsCSP genes, OsCASTOR, OsPOLLUX, OsC-CaMK, and OsCYCLOPS, are directly involved in AM symbiosis (see references in Table 1).
Heterologous complementation of legume symbiotic mutants with rice orthologs revealed that almost all rice genes examined rescued mycorrhization-defective phenotypes of corresponding mutants of L. japonicus or M. truncatula (Table 1). These results show functional conservation of these genes between rice and legumes in the regulation of AM symbiosis.
OsSYMRK, which encodes a symbiotic leucine-rich repeat (LRR) receptor kinase, failed to rescue bacterial endosymbiosis of the nodulation-defective Ljsymrk mutant . In contrast to the other   OsCSP genes, which have similar domain structures as those of legumes , Yano et al., 2008, a comparison of structure among SYMRK orthologs showed the presence of additional domains at the N-terminus of LjSYMRK. Thus, 'full-version' LjSYMRK regulates both RN and AM symbioses, whereas 'short-version' OsSYMRK regulates AM symbiosis only, suggesting that stepwise domain acquisition by SYMRK has contributed to the evolution of RN symbiosis ).
In L. japonicus, CASTOR and POLLUX are twin cation channels that function non-redundantly (Imaizumi-Anraku et al. 2005, Venkateshwaran et al. 2012. Neither rhizobial nor mycorrhizal symbioses are complemented in the roots of Ljpollux mutant expressing OsPOLLUX (OsPOLLUX/Ljpollux), while both rhizobial and mycorrhizal symbioses occur in the roots of OsCASTOR/Ljcastor . In this view, incomplete rescue of Ljpollux by OsPOLLUX may be caused by incompatibility between LjCASTOR and OsPOLLUX in Ljpollux mutant, conversely, heterologous combination of OsCASTOR and LjPOLLUX may be sufficient to rescue Ljcastor mutant. In M. truncatula, DMI1, the ortholog of POLLUX, acts in the absence of CASTOR, because a serine-to-alanine substitution in its filter region, a highly conserved ADS/AGNHA amino acid residues in the 4 th transemembrane domain, confers the integrated function of CASTOR and POLLUX (Venkateshwaran et al., 2012). In the Mtdmi1 mutant, OsPOLLUX cannot rescue rhizobial endosymbiosis, probably because OsPOLLUX retains the Ser residue in the filter region (Imaizumi-Anraku et al. 2005, Chen et al. 2009).
OsCCaMK and OsCYCLOPS rescue nodulationdefective phenotypes of the corresponding mutants of M. truncatula and/or L. japonicus (Godfroy et al. 2006, Chen et al. 2007, Yano et al. 2008, Chen et al. 2008. CCaMK/DMI3 encodes a Ca 2+ , CaMdependent protein kinase, which consists of kinase, CaM-binding (CaMBD), and EF-hand domains. CCaMK is a key player in RN and AM symbioses and acts as a decoder of symbiotic Ca 2+ signals. Functional analysis of CCaMK revealed the dispensability of its CaMBD and EF-hand domains for AM symbiosis, because the mutants with deleted CaMBD and EF hand domains are able to accommodate AM fungi (Shimoda et al., 2012, Takeda et al. 2012). On the other hand, full-length CCaMK is indispensable for the infection of rhizobial bacteria (Shimoda et al. 2012). These results raise the possibilities that CCaMK plays a role in sorting the signals derived from AM or RN symbioses and that RN symbiosis requires more complex regulation of CCaMK through CaMBD/EF hand domains (Takeda et al. 2012). The domain structure of CCaMK orthologs from nonleguminous mycorrhizal plants is similar to that of legumes . In abscisic acid signaling, OsCCaMK induces antioxidant defense (Shi et al. 2012(Shi et al. , 2014, which implies its yet unidentified roles in nonleguminous plants. CYCLOPS, a phosphorylation target of CCaMK, plays a role in transactivation of the NIN gene in the RN-specific pathway (Singh et al. 2014), whereas the downstream target of CYCLOPS in AM symbiosis remains unknown.
These complementation studies discussed above suggest that the CSP genes, except SYMRK, have remained unchanged during the evolution of RN symbiosis and constituted the molecular basis for the symbiotic signaling.

Is CSP also involved in responses to infecting microorganisms in rice?
There is a myriad of microorganisms in the rhizosphere. Do they interact with plant roots via CSP? The rice blast fungus Magnaporthe oryzae infects the roots of Oscastor-1 and Osccamk-2 mutants similarly to those of wild type, indicating that the OsCSP genes are not required for this infection (Marcel et al. 2010). Endophytic colonization of the rice mutants Oscastor-1, Osccamk-2, and Oscyclops (NC2794) by rhizobia also appears not to be affected under laboratory conditions (Chen and Zhu 2013). However, the OsCCaMK genotype affects the diversity of the bacterial community in rice roots in paddy and upland fields (Ikeda et al. 2011). Minamisawa and colleagues showed that the abundance of Alphaproteobacteria (Sphingomonadales and Rhizobiales), which are ubiquitous in the environment, was drastically decreased on the Osccamk-1 roots in comparison with wild-type roots (Ikeda et al. 2011). In the paddy fields under low nitrogen conditions, the methanotrophic community (mainly Methylomonas and Methylomicrobium in type I methanotrophs) was less diverse in the Osccamk-1 roots than in the wild-type roots; the enhanced CH 4 flux of the Osccamk-1 mutant suggests a positive influence of OsCCaMK on CH 4 oxidation (Bao et al. 2014).

Choice between symbiosis and defense response: two faces of OsCERK1
In RN symbiosis, CSP is activated upon recognition of rhizobial symbiotic signal molecules, Nod factors, by host lysine motif (LysM) receptor-like kinases, NFR1/ LYK3 and NFR5/NFP. The Nod factors are diversely modified derivatives of lipochitooligosaccharides (LCOs) and are also called Nod-LCOs. Similarly to rhizobia, AM fungi secrete diffusible signal molecules (Kosuta et al. 2003). One type of the mycorrhizal (Myc) factors, Myc-LCOs, are structurally similar to Nod-LCOs, but have a simpler structure (Maillet et al. 2011), implying that the receptors for AM fungal signals are also similar to those for Nod-LCO receptors. This suggestion is supported by the fact that a single copy of an NFR5/NFP homolog participates in both RN and AM symbioses in Parasponia, the only known non-legume plant that establishes rhizobial symbiosis .
Myc-LCOs improve mycorrhization (Maillet et al. 2011). In M. truncatula, they induce the expression of a number of genes, which partially overlap with but are largely distinct from the genes induced by Nod-LCOs (Czaja et al. 2012). Curiously, most of these responses in M. truncatula require NFP (Maillet et al. 2011, Czaja et al. 2012, although the mycorrhization phenotype of Mtnfp appears not to be affected. These perplexing results might indicate the different dependence of AM symbiosis on the Myc-LCO/NFP system among plant species. It is noteworthy that NFR5/NFP homologs are found in a wide range of mycorrhizal plants (Zhu et al. 2006, Zhang et al. 2007), suggesting their importance.
Homologs of NFR1/LYK3 are also conserved among plants (Zhu et al. 2006, Zhang et al. 2007, De Mita et al. 2014. OsCERK1, the closest rice homolog of NFR1/ LYK3, recognizes chitin or peptidoglycan and triggers plant immunity responses (Miya et al. 2007, Shimizu et al. 2010, Willmann et al. 2011. Although their physiological roles-acceptance or rejection of infecting microbes-are seemingly opposite, these genes may be descendants of a common ancestral gene (Nakagawa et al. 2011, Oldroyd 2013, Miyata et al. 2014). This view is supported by transient induction of a number of defense genes by Nod-LCOs in an NFR1-dependent manner in L. japonicus (Nakagawa et al. 2011) and by the findings that co-expression of either LjNFR1 with LjNFR5 or MtLYK3 with MtNFP causes cell death in Nicotiana benthamiana (Madsen et al. 2011, Pietraszewska-Bogiel et al. 2013. Therefore, NFR1/LYK3 also bears some traits related to defense responses, a feature reminiscent of CERK1. The ectodomains of NFR1 and CERK1 distinguish between Nod-LCOs and chitin oligomers (Broghammer et al. 2012, Wang et al. 2014, and their intracellular kinase domains play crucial roles in triggering nodulation or immunity responses, respectively (Petutschnig et al. 2010, Madsen et al. 2011. Despite the contrasting functions of the two kinases, chimeric constructs consisting of the ectodomain of LjNFR1 and the kinase domain of OsCERK1 rescue the nodulation defect of the Ljnfr1 mutant (Nakagawa et al. 2011, Miyata et al. 2014, suggesting the ability of OsCERK1 to regulate symbiotic interactions. Indeed, the Oscerk1 null mutants (Kouzai et al. 2014a) are impaired in both chitin-triggered immunity and AM symbiosis (Miyata et al. 2014). Yet, these two functions are not necessarily coupled because the Oscebip mutant (Kouzai et al. 2014b) has defective chitintriggered immunity but normal AM symbiosis (Miyata et al. 2014). In the oscerk1 null mutants, AM fungal infection is initially blocked at the root surface but prolonged cultivation eventually allows penetration of fungal hyphae into the root cortex and arbuscular formation (Miyata et al. 2014). In contrast, OsCERK1knockdown lines clearly show an impaired mycorrhization phenotype (Zhang et al. 2015) even at the time point corresponding to "prolonged cultivation" in oscerk1 null mutants in the study by Miyata et al. (2014). Although the mycorrhizal conditions differed in the two studies, different phenotypes of the oscerk1 null mutants and OsCERK1-knockdown lines might imply the existence of a functionally redundant gene. The bifunctionality of OsCERK1 raises a question of how the same receptor selectively triggers the opposite physiological responses. In response to microorganisms, OsCERK1 appears to recruit appropriate receptor partners to form symbiotic or defensive receptor complexes by which the initial interactions are determined.

Reverse genetic approaches of rice provide new insights into AM symbiosis
Reverse genetic approaches have identified novel AM symbiosis genes in rice. Genome-wide microarray analysis of mycorrhized rice roots revealed 224 genes the expression of which was altered in response to mycorrhization (Güimil et al. 2005). These OsAM genes were used for molecular phenotyping of the Oscsp mutants, which resulted in identification of an alternative signaling pathway independent of CSP . Furthermore, 76 of OsAM genes showed similar expression patterns of mono-and dicotyledonous plants, shedding light on a conserved molecular mechanism governing AM symbiosis (Güimil et al. 2005).
The expression of two half-size ABC transporters, STR1 and STR2, is enhanced during AM symbiosis, and these transporters are directly involved in arbuscule development (Gutjahr et al. 2012).
Two markers, OsPT11-GFP, localized in the periarbuscular membrane (Kobae and Hata 2010) and GFP-AM42, localized in the arbuscule, enable observation of arbuscule development (Kobae and Fujiwara 2014). Long-term live imaging of transgenic OsPT11-GFP and GFP-AM42 rice roots revealed a short life-span of arbuscules and repetitive de novo colonization of rice roots by AM fungi.
Plant hormones, e.g., abscisic acid, gibberellin, auxin, and SLs, are involved in AM symbiosis (Foo et al. 2013). In rice, the overexpression of osmiR393, a microRNA that targets several auxin receptors, results in downregulation of auxin receptor genes and hampers arbuscule development (Etemadi et al. 2014).

Relationship between strigolactone biosynthesis and mycorrhization
Identification of SLs as branching factors (Akiyama et al. 2005) has led to a breakthrough in understanding of the molecular mechanisms of the pre-contact stage of AM symbiosis. SLs also act as plant hormones to regulate shoot branching (Domagalska and Leyser 2011). In rice, the causative genes of the increased-tiller-outgrowth mutants, d17, d10, d27, d14 and d3 have been identified as SL-biosynthesis (D17, D10 and D27) or SL-signaling (D14 and D3) genes. In d10 and d17 mutants, the rates of colonization by AM fungi are decreased, whereas arbuscule structures are not affected (Gutjahr et al. 2012), indicating that SLs are dispensable for arbuscule development. Because SLs are undetectable in root exudates of these mutants (Umehara et al. 2008), their mycorrhization phenotypes are likely to be associated with incomplete induction of hyphal branching, which may determine the lower frequency of AM fungal infection.
The crosstalk between the SL synthesis pathway and the GRAS transcription factors NSP1 and NSP2 has been shown in rice. A decreased expression of D27 was detected in the Osnsp1/Osnsp2 double knockdown lines ) and in the nsp1, nsp2, and nsp1/ nsp2 background of M. truncatula and L. japonicus , Takeda et al. 2013, Nagae et al. 2014). However, the colonization rate of the Ljnsp1 mutant was not restored by application of SL analog, GR24, suggesting that the attenuation of the SL biosynthesis pathway is not a major determinant of the mycorrhization defect of this mutant (Takeda et al. 2013).
The d3 and d14 mutants are SL-insensitive and synthesize more SLs and induce highly branched hyphae around their roots in comparison with wild type (Umehara et al. 2008, Yoshida et al. 2012. As expected, d14 shows a high-mycorrhization phenotype. However, the development of hyphopodia on d3 roots is arrested (Yoshida et al. 2012) and the mode of action of D3 at the initial penetration stage of AM fungi remains elusive.

Conclusions and Perspective
As reviewed here, rice has made a major contribution to the elucidation of molecular mechamisms of AM symbiosis, as a model monocot plant. On the other hand, rice is one of three major cereals and has a direct impact on food supplies. Conventional agriculture largely depends on the input of phosphorus fertilizer; the output of mineral phosphorus is expected to reach a peak around 2030, and then to take a downward turn (Cordell et al. 2009). Under these circumstances, rice cultivars that need less phosphorus are an important breeding target (Kochian 2012). As a successful example of such approach, the protein kinase Pstol1, initially detected in cultivar Kasalath as a QTL for phosphorus-deficiency tolerance (Pup1), was used to confer tolerance to phosphorus deficiency in phosphorus-starvation-intolerant modern cultivars. Pstol1 acts as an enhancer of root growth, thereby allowing host plants to absorb more phosphorus (Gamuyao et al. 2012).
In general, rice is grown in flooded fields, where mycorrhization is inhibited under anaerobic conditions, in part because of reduced spore abundance of AM fungi in comparison with dry conditions (Ilag et al. 1987). Regardless of rice varieties, colonization by AM fungi is substantially reduced under flooded conditions (Vallino et al. 2009). AM symbiosis can be used for rice cultivation under flooded conditions (Secilia and Bagyaraj 1992, Solaiman and Hirata 1997, 1998; however, rice breeding to use AM symbiosis for phosphate uptake remains out of reach. Even though modern rice cultivars are grown under anaerobic conditions, the amount of inorganic phosphate (Pi) uptake through AM symbiosis accounts for 70 % of total Pi absorbed by rice cultivar Nipponbare under experimental condition. In addition, cultivar IR66 maintains elevated expression levels of AM marker genes under both aerobic and anaerobic field conditions, suggesting high dependency of rice Pi uptake on the mycorrhizal pathway (Yang et al. 2012). Although rice flooding has a negative impact on AM fungal colonization, AM fungi that had entered into rice roots before flooding remain viable in the roots under anaerobic conditions. Furthermore, basic functionality of AM symbiosis is not affected by flooding (Maiti et al. 2008 andVallino et al. 2014). Therefore, rice seems to be a potential target for breeding to increase productivity through AM symbiosis.