To perform our phylogenetic analysis of the MLO family, we used the protein sequences of 12 MLO genes identified in a previous global analysis of the rice MLO family (Liu and Zhu, 2008). We also developed a tree with rice, barley, and Arabidopsis MLO proteins, based on previous global analyses of this gene family (Jorgensen, 1992; Devoto et al., 2003). The protein sequences for our phylogenomic analysis were downloaded from TAIR, GreenPhyl and the Rice Genome Annotation Project Website (Rhee et al., 2003; Ohyanagi et al., 2006; Rouard et al., 2011). After multiple-alignment of those sequences with ClustalX (Higgins et al., 1996), we generated a phylogenetic tree using the Neighbor-Joining method, as incorporated in the MEGA5 tool kit for phylogenetic analysis (Tamura et al., 2011).

Microarray data including Affymetric and Agilent array data were downloaded from the NCBI Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) and Genevestigator (https://genevestigator.com/gv/) (Zimmermann et al., 2004; Cao et al., 2012). We then uploaded the normalized data to the Multi Experiment Viewer (http://www.tm4.org/mev.html) and visualized the data via heatmaps. We used Genevestigator to compare levels of gene expression in several organs and to estimate the functional similarity among rice, barley, and Arabidopsis members.

Circadian rhythms and functional associations with rice gigantea (OsGI) were examined using wild-type (WT) plants and osgi mutants. Ten-day-old rice seedlings were grown on a Murashige Skoog medium under controlled conditions (28°C/25°C day/night, continuous light, and 78% relative humidity). They were then transferred to individual pots and placed in a growth chamber. Beginning at 30 d after germination, their leaves were sampled at 4-h intervals for 48 h. Light- and dark-dependent expression was analyzed using a heterozygous osdxr (rice 1-deoxy-D-xylulose 5-phosphate reductoisomerase) mutant line for which seeds were germinated as described above. Homozygous and WT plants were identified through genotyping, and leaves were collected from 10-day-old plants. Primers used for genotyping analysis are presented in Table S1.

Cold treatment consisted of transferring 10-day-old plants originally grown at 28°C to a 6°C ± 1°C lighted refrigerator. The control plants remained at 28°C. After 48 h, the chilled plants were returned to the 28°C chamber for recovery. Samples were collected at Hours 0, 24, and 48 of low-temperature treatment and after 24 and 48 h of recovery. To simulate heat stress, we exposed 10-day-old plants to 42°C ± 1°C for 0, 3, 6, and 12 h. The control plants remained at 28°C. Leaves of three plants were pooled for one biological replicate and each treatment had three repeats. To assess their genetic response to pathogen infection, we sprayed Magnaporthe oryzae conidia suspension with 0.005% Tween on to 3-week-old plants of Dongjin rice and then placed them in containers for 24 h to maintain 80% humidity under darkness before transferring them to a chamber under a 14-h photoperiod. Control and mock samples were collected by spraying water without and with 0.005% Tween, respectively.

Leaf samples were frozen in liquid nitrogen and homogenized with a TissueLyser II (Qiagen, Hilden, Germany). Total RNA was extracted using RNAiso Plus according to the manufacturer's protocol (Takara Bio, Kyoto, Japan). The qPCR was performed by Qiagen Rotor-Gene Q real-time PCR cycler using follow thermal cycling procedure: 95°C for 10 s, 60°C for 30 s, and 72°C for 1 min. For evaluating tissue-specific expression patterns by real-time PCR, we used rice ubiquitin 5 (OsUbi5, LOC_Os01g22490) (Jain et al., 2006) as an endogenous control to normalize variance in the quality of RNA and the amount of cDNA. LATE ELONGATED HYPOCOTYL (OsLHY) served as the positive control when examining diurnal rhythms. The effects of cold and heat stress were monitored using OsNAC6 (LOC_Os01g66120) (Ohnishi et al., 2005) and OsHSP1 (LOC_Os04g01740) (Moon et al., 2014), respectively, as positive controls. Successful interaction between rice and M.oryzae was convinced by positive control OsPR10a (LOC_Os12g36880) (Lee et al., 2009). All of the primers for these real-time analyses are shown in Table S1.

Twenty-five co-expressed genes of 6 individual MLO genes (OsMLO1, 2, 3, 4, 8, and 9) with top-score Pearson Correlation Coefficient (PCC) value were collected, based on the co-expression tool installed in Genevestigator (Zimmermann et al., 2004). Co-expressed genes might provide useful functional clues for associated MLO genes. Especially, functionally characterized genes collected from The Overview of functionally characterized Genes in Rice online (OGRO) database (Yamamoto et al., 2012) are more effective for this purpose.

Using protein sequences for 12 MLO genes in rice, we generated a phylogenetic tree that incorporated subgroup information and anatomical meta-expression profiles based on 983 Affymetrix array data (Cao et al., 2012) (Figure 1). Starting at the top of the tree, OsMLO3 displayed a medium level of expression in almost all tissues, including roots, shoots, leaves, and flag leaves at the vegetative stage, and anthers and seeds at the reproductive stage. It was most closely linked with OsMLO6. However, very low expression by the latter in whole tissues suggested that OsMLO3 had a more dominant role. Uniquely, OsMLO12 showed strong anther-preferred expression in trinucleate and mature pollen. This profile was consistent with that reported by Yi et al. (2014). Expression of OsMLO1 was high in anthers at the uninucleate stage, especially in indica rice, although not at a level similar to that found in uninucleate pollen (Figure S1). Thus, OsMLO1 might be more active in the anthers than in the pollen grains themselves. The Affymetrix data revealed that OsMLO4 was preferentially expressed in the root tips but at only a low level in other tissues. In contrast, OsMLO2 was ubiquitously expressed. One of its paralogs, OsMLO9, showed dominant expression patterns in reproductive organs, and transcripts were particularly abundant in anthers at the uninucleate pollen stage while OsMLO2 was not detected there (Figure S1). Although OsMLO5 and OsMLO10, as well as OsMLO7 and OsMLO8, are thought to have resulted from segmental duplication, each pairing appears to have retained differential expression patterns from an anatomical perspective. While expression of OsMLO5 was extremely low in the roots, that of OsMLO10 was much higher than for OsMLO5. Moreover, when compared with OsMLO7, OsMLO8 was more predominantly expressed in the shoots, leaves, and anthers at the trinucleate pollen stage (Figure 1). All of these findings suggested that functional specificity and dominance is much greater for OsMLO8 and OsMLO10 than for their homologs OsMLO7 and OsMLO5, respectively. Therefore, with the exception of OsMLO11, for which no probe was printed on the Affymetrix chip, we were able to use meta-expression data to estimate the anatomical functions of all MLO genes examined here.

To confirm the results from our meta-analysis of anatomical expression profiles, we performed real-time PCR with 12 rice MLO genes that are expressed in the roots, shoots, leaves, young panicles, mature flowers, and seeds (at 6 d post-pollination), as well as in anthers sampled at the uni-, bi-, and trinucleate stages (Figure 2). Our findings here closely matched those obtained from meta-analyses of tissue-specific expression profiles. However, we failed to detect any expression of OsMLO7, probably because its transcript level was extremely low. We determined that OsMLO11 was highly expressed in mature flowers and in uni-, bi-, and trinucleate anthers. These results demonstrated the high reliability of meta-expression profiles based on a large collection of transcriptome data.

Although the roles of rice MLO family members have long been studied, only one of those genes has known biological functions. It is known that OsMLO12 regulates pollen hydration (Yi et al., 2014). However, we believed that the expression patterns of other genes in that family might be good indicators for predicting their biological functions. We also expected that a comparative transcriptome analysis of MLO family members from rice and Arabidopsis would provide more accurate evidence for functional conservancy among orthologous gene groups in those two species. Therefore, we combined phylogenetic and meta-expression data from Genevestigator, which uses a similar platform to present expression patterns within an anatomical context.

Our phylogenetic tree covered 29 protein sequences that were aligned from rice (12 MLOs), Arabidopsis (15), and barley (2). This established four clades, i.e., I through IV (Figure 3). Dividing the number of clade for MLO family has yet to be unanimous (Feechan et al., 2008; Pessina et al., 2014; Kusch et al., 2016), we separated 29 MLOs into four clades base on primary evolution and relative known function gene with small numbers of MLO genes. This tree implied that the evolutionary processes for MLO genes from monocots and dicots were rather independent. For example, the 29 members included only two orthologous pairings—AtMLO4/OsMLO11 in Clade III and OsMLO10/HvMLO3 in Clade IV.

For Clade I, the anatomy tool installed in Genevestigator indicated similar expression between AtMLO5 and OsMLO12 (Figure 2). Both AtMLO8 and OsMLO1 were ubiquitously expressed except in the pollen (Figure 2). AtMLO10 was predicted to function in the roots, hypocotyls, radicles, and endosperm.

In Clade II, expression of OsMLO3 in almost all organs/tissues was the most similar to the patterns found for AtMLO2 than AtMLO6, and AtMLO12 expression. Elliott et al. (2002) have demonstrated that introducing OsMLO3 into the barley MLO null mutant genotype mlo-5 leads to 50% susceptibility restoration. Therefore, it appears that OsMLO3 in rice might have a biological function in plant defenses. However, it is unclear whether the presence of a single copy of that gene is sufficient to confer pathogen resistance. Because OsMLO6 was expressed at low levels in most tissues/organs while AtMLO6 and AtMLO12 were highly expressed in roots and radicles, we suspect that the latter two genes have roles in root development that are conserved with AtMLO2.

For Clade III, the expression patterns of AtMLO4, AtMLO11, and OsMLO4 indicated that they have similar functions in the development of root tips and stigma, suggesting that the roles of MLO genes within that clade are conserved. Although data for OsMLO11 were not available using the Affymetrix array platform, we were able to determine its spatio-temporal expression patterns using the Agilent array platform. In doing so, we found that this gene was highly expressed in the roots, inflorescences, pistils, palea/lemma, and ovaries (Figure S2). Therefore, expression was similar between OsMLO11 and OsMLO4 at the reproductive stage. While AtMLO4 displayed similar expression to AtMLO11 in every tissues/organ, AtMLO14 expression is disparity. However, whereas knockouts of AtMLO4 and AtMLO11 caused an abnormal root phenotype, the paralog AtMLO14 was not involved in root development. This might explain their different patterns of expression even though AtMLO11 and AtMLO14 were the closest members in the evolutionary tree (Figure 3). Nevertheless, we expect that the osmlo4 mutant would display the same behavior upon tactile stimulation as that shown by mutants of AtMLO4 or AtMLO11.

In Clade IV, OsMLO8 was preferentially expressed in the leaves and shoots. However, we were unable to find similar expression patterns for Arabidopsis genes in this clade. Instead, AtMLO1, OsMLO2, and OsMLO9 were ubiquitously expressed in most of the anatomical samples, suggesting that these MLO family members have housekeeping functions.

Diurnal rhythm is synchronized with the day/night cycle and is regulated by two mechanisms: light and the circadian clock. To elucidate the diurnal rhythm of our MLO genes, we analyzed their expression patterns using publicly available Agilent 44k array data obtained for diurnal and circadian gene expression in the leaf over the entire life span of field-grown rice plants (Sato et al., 2013). Among the 12 MLO genes, leaf preferentially expressed OsMLO8 showed high expression that was obviously diurnal, while OsMLO1, OsMLO3, and OsMLO9 were also diurnally expressed, albeit at a more moderate level, in mature leaves (Figure 4A). Real-time PCR analyses of genes from 4-week-old leaves indicated that OsMLO1, OsMLO3, and OsMLO8 exhibited cyclic expression every 24 h (Figure 4B).

To obtain insight into the mechanism that regulates the diurnal rhythm of MLO genes, we evaluated a rice mutant. In osgi, diurnal expression of marker gene OsLHY was dramatically down-regulated across all time points whereas the rhythm of expression for OsMLO1, OsMLO3, and OsMLO8 was similar to that detected in WT plants (Figure 4B). These findings indicated that the genes are not related to OsGI or the circadian mechanism.

To determine whether the MLO genes are involved in light-/dark-response mechanisms, we evaluated their expression using Agilent 60K microarray data for a mutant of the light-responsive osdxr, which belongs to the methylerythritol 4-phosphate pathway (Jung et al., 2008) (Figure 5A). Real-time PCR analysis confirmed that OsMLO1 expression was decreased while that of OsMLO3 was much lower in the mutant than in the WT (Figure 5B). Leaf-specific expression of OsMLO8 was suppressed completely in osdxr, suggesting a close connection between OsMLO8 and the light-response pathway. Our results implied that the MLO genes are diurnally expressed and that OsMLO3 is dark-inducible while OsMLO1 and OsMLO8 are light-inducible (Figure 5).

The MLO proteins interact with calmodulin as the second messenger to transfer a signal forward to downstream pathways (Kim et al., 2002a,b; Stein and Somerville, 2002). To investigate whether the diurnal response-dependent OsMLOs function with calmodulin proteins, we examined the differential expression of calmodulin (CaM) and CaM-like (CML), searching for genes with log₂-fold changes >1.8 and p-values <0.05 in the osdxr mutant (Table S2). Subsequent analysis of the microarray data presented four CML genes—OsCML1, OsCML16, OsCML24, and OsCML28—with decreased expression in that mutant (Figure 6A). Real-time PCR analysis also confirmed that all four were significantly down-regulated in osdxr (Figure 6B). Therefore, we propose that these OsCML genes would be primary targets for studying functional relationships among diurnal response-dependent OsMLOs.

Although the barley MLO genes are affected by abiotic stresses such as leaf-wounding and herbicides (Piffanelli et al., 2002), little is known about how genes in that family are influenced by other sources of stress. We used the log₂ fold-change data in response to abiotic stresses (Figure S3) to investigate the meta-expression patterns of 12 rice MLO genes. Differential expression was monitored via real-time PCR with samples under either heat- or cold-stress conditions. As expected, OsMLO2, OsMLO3, OsMLO4, and OsMLO9 were rapidly up-regulated during the first 3 h of heat treatment (Figure 7). Their expression began to decline after 12 h of treatment but was still higher than that measured from the WT control. Among those four genes, OsMLO4 was the most responsive to high temperature, with expression being >30-fold higher than that in control plants after 6 h and remaining at that level for 12 h of treatment. Because OsMLO5, OsMLO6, and OsMLO7 were expressed only at low levels in the leaves, they were eliminated from further analysis of temperature sensitivity.

In response to cold stress, OsMLO1 and OsMLO3 were up-regulated at 48 h after treatment was applied. However, their level of expression declined to that measured from the control when the chilled plants were allowed to recover at 28°C (Figure 8). Both OsMLO4 and OsMLO11 were quickly induced by low temperatures, with expression increasing after just 24 h of treatment and transcripts being retained at higher levels until Hour 48. By contrast, OsMLO9 was down-regulated by 24 h of chilling. This response was the opposite of its upregulation by heat stress, implying that the gene has separate functions in determining the plant response to temperature extremes. All of these findings provided evidence that the expression of these rice MLO genes is influenced by cold and/or heat stress.

“Dongjin” rice plants were inoculated with M. oryzae PO6-6 to monitor the response of MLO genes to pathogen infection. We had expected that OsMLO3 would show differential expression similar to that demonstrated by its homologous members in Arabidopsis and barley. In fact, OsMLO3 was significantly induced at 72 h after infection (Figure 9). Expression of two light-inducible genes—OsMLO1 and OsMLO8—was relatively decreased after 72 h and 24 h, respectively, two closely linked members—OsMLO2 and OsMLO9—were slightly down-regulated at 48 h, and OsMLO11 were significantly reduced expression at 48 h after treatment. These data indicated that rice MLO family members have possible roles in the pathogen response.

Twenty five co-expressed genes of 6 individual MLO genes (OsMLO1, 2, 3, 4, 8, and 9) with the highest PCC value were selected by co-expression analysis tool installed in Genvestigator with 2532 anatomical samples (Table S3). To identify the possible function in which each MLO gene is involved, we searched rice genes with known functions from the OGRO database out of the co-expressed genes. Unfortunately, progress of functional genomics in rice is still restricted and explains functions for 2 or 3% of genes in rice genome (Chandran et al., 2016). Thus, we only found two known genes from co-expressed genes to OsMLO3 or OsMLO8. A known gene, OsHsfA4d/Spl7, one of co-expressed genes to OsMLO3, encodes heat shock transcription factor (Yamanouchi et al., 2002), which activates heat shock genes under high temperature stress. spl7 mutant causes scattered brown spot on leaf and dead cells increase since exposing under heat stress or UV solar radiation. spl7 also accumulated higher amounts of H₂O₂ in response to rice blast fungal elicitor (Kojo et al., 2006). Although the co-expression score (PCC value) of OsMLO3 and Spl7 is 0.6847, this linkage suggests one of downstream responsive genes of OsMLO3 associated with heat stress, cell death and H₂O₂ accumulation. In case of fructose-1,6-bisphosphatase (OscFBP1/MOC2), the co-expression score with OsMLO8 is 0.9314. OscFBP1 converts triose-phosphates to sucrose in the day (Daie, 1993). Loss function of OscFBP1 exhibits monoculm phenotype and reduced photosynthetic rate (Lee et al., 2008; Koumoto et al., 2013). Since leaf- preferred OsMLO8 gene showed the peak of expression level under daylight, its co-expression with OscFBP1 suggests the important role in photosynthesis.

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