The rice cultivar Nipponbare (Oryza sativa L. ssp. japonica) was utilized in this study. Plants were grown under natural conditions in a rice paddy field at Shandong Academy of Agricultural Science, and the Arabidopsis zou-4 mutant (Col-0 background) that corresponds to GABI_584D09 was obtained from the Gabi-Kat collection of T-DNA inserts (Rosso et al., 2003). The Arabidopsis growth method used in this study follows that outlined in previous work (Zhang et al., 2017).

A total of 211 bHLH protein sequences from rice were mined from Plant Transcription Factor Database V4.0, and a phylogenetic reconstruction of the AtZOU gene encompassing rice bHLH proteins was performed by applying the neighbor-joining method (Saitou and Nei, 1987) using the software MEGA7 (Tamura et al., 2007). A bootstrap consensus tree inferred from 1,000 replicates was considered sufficient to represent the evolutionary history of included taxa (Felsenstein, 1985). Branches corresponding to partitions reproduced in less than 50% of bootstrap replicates were collapsed, and the percentage of trees in which associated taxa clustered together (1,000 replicates) were highlighted adjacent to corresponding branches (Felsenstein, 1985). The tree resulting from this analysis was drawn to scale, with branch lengths given using the same units as the evolutionary distances used to infer the phylogeny. Evolutionary distances were computed using the Poisson correction method (Zuckerkandl and Pauling, 1965) and are presented in terms of the number of amino acid substitutions per site. All positions containing alignment gaps and missing data were eliminated via pairwise sequence comparisons (i.e., the pairwise deletion option), an overall total of 671 positions in the final dataset.

Total RNA was isolated from developing rice seeds and leaves under normal growth conditions using the Plant Total RNA Extraction Kit (Bioteke Corporation, Beijing) and first strand complementary DNA (cDNA) was synthesized using the AMV Reverse Transcription System (Takara, Beijing). A 3′ A overhang from EasyTaq (TransGenBiotech) was added to all PCR fragments and these were then cloned into the pMD18-T vector (Takara, Beijing, China). The primers OL0069 and OL0070 were used for OsZOU-1 clone, while OL0071 and OL0072 were used for OsZOU-2 clone (Supplementary Table S1).

Rice total RNA extraction, reverse transcription, and qRT-PCR were all performed following previously described protocols (Dou et al., 2017), as was Arabidopsis RNA extraction (Zhang et al., 2017). The reference genes Actin1 (rice) and EIF4 (Arabidopsis) were used for expression level calculations (Xing et al., 2013; Dou et al., 2017) in all cases and all primers are listed in Supplementary Table S1.

In order to analyze OsZOU expression patterns in rice, total RNA was isolated from a range of different tissues, including root, stem, leaf, unopened flower, and kernels at between two and nine DAP. Similarly, to analyze OsZOU-1 expression levels in different rice seed segments, the endosperm and embryo were obtained separately from rice kernels at five DAP following the protocol outlined by Zhang et al. (2017). Error bars denote the standard deviation (SD) calculated using three technical replicates on pooled material from three WT plants.

In order to analyze OsZOU expression levels in Atzou-4 mutants carrying the pAtZOU::OsZOU-1 or pAtZOU::OsZOU-2 constructs, five independent transgenic lines were examined in each case. Three different plants from each line were used for biological replicates and the Atzou-4 mutant was selected as a negative control. Three consecutive siliques from the mid to late heart stage of each plant were picked and flash frozen in liquid nitrogen for RNA extraction (Xing et al., 2013). OsZOU expression was then quantified relative to the AtZOU level in Col-0 seeds; the error bars in these cases represent SDs calculated from three biological replicates (Student’s t-test value: ^∗∗P < 0.01).

To analyze the expression levels of OsZOU-1 in the OsZOU-1 knock-down transgenic lines, the seeds at five DAP of WT line and knock-down lines KDZ-2, KDZ-4 were examined by qRT-PCR. Three homozygous plants from each line were examined, and the expression values are averages for three plants. Thus, OsZOU-1 expression level in WT seeds was set as 1 and the error bars in this case correspond to the SD calculated from three biological replicates (Student’s t-test value: ^∗∗P < 0.01).

To analyze OsZOU-1 expression levels in OsZOU-1 over-expressed transgenic lines, unopened flowers of OXZ-4, OXZ-5, and WT plants were picked, flash frozen in liquid nitrogen, and used for RNA extraction and qRT-PCR. The OsZOU-1 expression level in WT plants was again set as 1, and error bars in this case correspond to the SD calculated from four technical replicates on pools of eight unopened flowers from WT, OXZ-4, and OXZ-5 individuals, respectively (Student’s t-test value: ^∗∗P < 0.01).

In order to construct pAtZOU::OsZOU-1 and pAtZOU::OsZOU-2 expression vectors, cloned OsZOU-1 and OsZOU-2 cDNA from pMD18-T vectors were digested with XbaI and HincII and cloned into the pRT101 vector. This enabled construction of the vectors ZYH040 and ZYH041 which were then digested with SalI and HindIII before gene fragments were separately cloned into a binary vector pCAMBIA1301 to construct ZYH057 and ZYH058. A 1.6-kb region upstream from the AtZOU start codon in pSC-B (Zhang et al., 2017) was then digested with SalI, and the promoter fragment was inserted into the binary vectors ZYH057 and ZYH058 to construct the pAtZOU::OsZOU-1 and pAtZOU::OsZOU-2 expression vectors ZYH065 and ZYH067. All recombinant plasmids were verified via sequencing, and the correct ones were introduced into Agrobacterium tumefaciens GV3101 independently for plant transformation. All Arabidopsis transformations were performed following previously described protocols (Zhang et al., 2017).

Homozygous transgenic Arabidopsis plants were screened using previously developed protocols (Zhang et al., 2017), and transgenic lines were further verified for the proper T-DNA insertion containing the exogenous OsZOU-1 and OsZOU-2 derived by the AtZOU promoter. The Atzou-4 mutant background in obtained OsZOU-1 and OsZOU-2 transgenic lines was identified using PCR based on the presence of a relevant T-DNA insertion site.

A series of TB staining tests were performed on Arabidopsis seedlings following protocols developed in previous work (Tanaka et al., 2004; Zhang et al., 2017); tests were therefore performed independently on OsZOU-1#1, #4, and #5 and OsZOU-2#1, #3, and #5 Arabidopsis transgenic plants. To do this, a total of 50 Arabidopsis seeds from each plant were grown uniformly on 15 cm plates; these were then dyed and photographed and 20 seedlings from each plant line were harvested in order to quantify their TB absorption values. Roots and seed coats (both of which stain dark when treated with TB) were completely removed before the cotyledons and hypocotyl were incubated with 80% ethanol and subjected to spectrophotometer analysis. Mean values for OsZOU-1#1, #4, and #5 represent results for OsZOU-1 transgenic lines, while those for OsZOU-2#1, #3, and #5 represent results for OsZOU-2 transgenic lines. Error bars in these cases denote SD values for three plants (Student’s t-test value: ^∗∗P < 0.01, ^∗P < 0.05).

Three homozygous plants from the knock-down rice lines KDZ-2 and KDZ-4 were used for TB staining analyses in each case. Six seeds from each plant were placed in plates containing tap water in a greenhouse (subject to cycles of 14 h light and 10 h dark at 28°C) for germination. This water was changed once a day to keep it clean, and rice seedlings were placed in plates containing staining solution [i.e., 0.4% (v/v) Tween-20 and 0.05% (w/v) TB] for 2 min on the third day after germination. Seedlings were then washed under running tap water until no visible blue color remained before they were photographed.

The OsZOU-1 cDNA from the pMD18-T vector was digested with EcoRI and HincII and connected into a pRT101 vector that had been digested using EcoRI and SmaI. This enabled the ZYH070 construction which was then digested using HindIII and EcoRI before a DNA fragment was connected into the binary vector pCAMBIA1300 to construct ZYH071. A 1.9 kb region upstream from the OsZOU-1 start codon was then amplified using PCR with the primers ZOUECORIF and ZOUECORIR and subcloned into vector pMD18-T. The vector containing the OsZOU-1 promoter was then digested with EcoRI, and the promoter fragment was connected into the binary vector ZYH071 to construct the pOsZOU-1::OsZOU-1^AS expression vector ZYH074. Similarly, to construct a pOsZOU-1::GUS expression vector, ZYH076 containing the OsZOU-1 promoter was digested with PstI and BamHI and connected into the binary vector pCXGUS-P (Chen et al., 2009). This process enabled construction of the GUS report vector ZYH081, while to construct a OsZOU-1 over-expression vector, OsZOU-1 was amplified using PrimeSTAR HS DNA Polymerase (Takara, Beijing, China) and the primers OL0069 and OL0070. A series of 3′ A overhangs were then connected to PCR fragments using EasyTaq (TransGenBiotech, Beijing, China) and the products were directly connected into the binary vector pCXUN (Chen et al., 2009) by T-A cloning enabling construction of the OsZOU-1 over-expression vector ZYH048. All of these vectors were then transformed into the Nipponbare (Oryza sativa L. ssp. japonica) rice cultivar using the Agrobacterium-mediated co-cultivation method (Hiei et al., 1994). Hygromycin-resistant plants were then screened for the presence of the transgene via PCR using gene specific primers, and seeds from different transgenic individuals were harvested separately. Primary transformants were referred to as T1 transgenic plants, and homozygous individuals were screened using an MS medium containing hygromycin B; T3 homozygous knock-down transgenic individuals were used for subsequent phenotypic and molecular analyses.

GUS report rice kernels at five DAP were cut into longitudinal sections with a razor blade and placed in pre-cooled 90% acetone for 20 min. These kernels were then rinsed twice with a X-Gluc free GUS staining solution, following a previously developed procedure (Maharjan et al., 2011).

Cell numbers and size measurements for OsZOU-1 over-expressed lines and WT lines were recorded based on ten seeds of each plant via photographs taken with a Nikon Eclipse E800 microscope. Cell numbers and lengths at the widest point of glumes in each case were measured using the software Image J. Error bars in this case correspond to SDs calculated from all measurements (Student’s t-test value: ^∗∗P < 0.01).

In order to identify ZOU orthologous genes, a AtZOU amino acid sequence was used as a bait to search for orthologs within the rice genome (Phytozome). Results show that two ZOU homologs (LOC_Os04g35010 and LOC_Os02g34320) were highly conserved with AtZOU and the corresponding genes were termed OsZOU-1 and OsZOU-2, respectively. Amino acid sequences alignment of OsZOUs with AtZOU revealed conservation within the bHLH DNA-binding domain (pfam00010) and C-terminal region (Figure 1). Phylogenetic analysis of all the 211 bHLH family proteins in Oryza sativa subsp. japonica with AtZOU revealed that OsZOU-1 and OsZOU-2 were located within the same clade with AtZOU, while the other bHLH family proteins were not (Supplementary Figure S1).

A series of qPCR experiments were performed using RNA extracted from vegetative and reproductive tissues, including roots, stems, leaves, unopened flowers, and developing kernels at different stages to analyze tissue specific expression of OsZOUs (Figure 2A). These results reveal that OsZOU-1 is mainly expressed in kernels between five and nine DAP and that there is a peak at five DAP when endosperm cellularization has completed and no visible expression of OsZOU-1 can be detected in roots, leaves, and stems. In contrast, data show that OsZOU-2 is mainly expressed in leaves and stems, at much lower levels in roots, and is almost completely absent from kernels (Figure 2A).

To further dissect the promoter activity of OsZOU-1 in developing seeds, twenty-three independent pOsZOU-1::GUS transgenic plant were developed in which 1.9 kb of OsZOU-1 promoter drived GUS coding sequence. Seeds at seven DAP and nine DAP were stained with a GUS solution and visualized under microscope (Figure 2B); these results reveal a GUS signal was detected in entire endosperm of kernels at seven DAP and nine DAP but not in embryo. Thus, to further determine the expression position of OsZOU-1 in developing rice kernels, the relative expression levels of OsZOU-1 in dissected embryos and endosperms at five DAP were examined. The qPCR results reveal that OsZOU-1 is highly expressed in the endosperm, but not in the embryo (Figure 2C). These results show that OsZOU-1 is specifically expressed in developing kernel endosperm but not in the embryo.

In order to investigate whether, or not, OsZOUs are involved in endosperm breakdown, Atzou-4 mutants were transformed with OsZOU-1 and OsZOU-2 controlled by the AtZOU promoter (pAtZOU::OsZOUs), respectively (Figure 3). Five independent transgenic lines of OsZOU-1 and OsZOU-2 were tested separately in this experiment and OsZOU-1 and OsZOU-2 were successfully expressed in seeds in all cases (Figure 3E). Phenotype analyses were therefore performed on three transgenic lines in which OsZOU-1 or OsZOU-2 were moderately expressed. Observation revealed that OsZOU-1 almost completely rescued the shriveled seed phenotype and that the mature dry seeds from OsZOU-1 transformation lines remained plump albeit irregularly shaped in some cases (Figures 3A,C). Mature seeds from OsZOU-2 transgenic lines were shriveled, similar to their counterparts from Atzou-4 mutant lines (Figures 3B,D).

Developing seeds from transgenic lines were examined in order to verify the function of OsZOUs during embryonic and endosperm growth (Figures 4A–D). Results showed that the Col-0 embryo was almost completely bent with little endosperm retention (Figure 4A), while the embryo of the Atzou-4 mutant was hindered at the heart/torpedo stage and a large volume of endosperm persisted, which resulted from PCD failure (Figure 4B). The embryos from the OsZOU-1 transgenic line were bent and occupied most of the embryonic sac, similar to Col-0, meanwhile the endosperms from the OsZOU-1 transgenic lines exhibited degradation with less persistent compared to Atzou-4 mutants (Figure 4C). In contrast, the seeds from the transgenic lines with over-expressed OsZOU-2 were similar in their development to the Atzou-4 mutant as embryo morphology was arrested in the heart/torpedo stage and a large volume of endosperm was preserved (Figure 4D). The results of this study indicate that OsZOU-1 rescues the defect in endosperm breakdown and embryo expansion partly, although OsZOU-2 is unable to recover seed development in Atzou-4 mutants.

The integrity of embryonic cuticle in transgenic lines was also examined in this study because of breaks observed in Atzou mutants. As discussed, hydrophobic cuticle defects were detected using the TB hydrophilic dye (Tanaka et al., 2004); thus, on the seventh day after germination, seedlings of zou-4 mutants, Col-0 WT plants, and OsZOU transgenic lines were dyed with a TB aqueous solution and imaged using a microscope. Observations reveal that the cotyledons of zou-4 lines were strongly stained, while the dye was hardly able to penetrate this region in Col-0 WT plants. Staining results for the cotyledons of OsZOU-2 transgenic plants were similar to those of zou-4 mutants, while OsZOU-1 transgenic individuals were only weakly dyed (Figures 4E–H). Finally, the TB absorption capacity of seedlings was also quantified using spectrophotometry, yielding results consistent with microscope observations (Figure 4I). These findings indicated heterologous expression of OsZOU-1, but not OsZOU-2 also partially rescued cuticle formation in Atzou-4 mutant.

In order to further investigate the function of OsZOU-1, a series of knock-down plants were also generated as part of this study. Thus, OsZOU-1 knock-down lines were generated via expression of full length antisense OsZOU-1 cDNA sequences under the control of a 1.9 kb OsZOU-1 endogenous promoter. A total of 63 independent T1 transformation plants were generated and T-DNA insertions were verified by PCR. Observations show that these transgenic plants exhibited no visible defects with the exception of lower seed rates than their WT counterparts (Figure 5A and Supplementary Figures S3, S5); meanwhile, most mature transgenic kernels were not completely filled (Figure 5B) with the full seed percentages less than 20% in the ten primary transgenic plants caculated (Supplementary Figure S4). Data show that OsZOU-1 expression levels in the T1 knock-down plants KDZ-2 and KDZ-4 were about 33 and 16% of the WT kernel level at five DAP, respectively (Figure 5C); histological analysis also revealed that embryos remained adhering to the surrounding endosperm in knock-down kernels at 12 DAP, whereas there was endosperm cavity between endosperm and embryo in wild type kernel (Figure 5D).

The integrity of embryonic cuticle in knock-down rice lines was examined using a TB staining test. Thus, seedlings of WT, KDZ-2, and KDZ-4 individuals were stained with a TB solution and observed on the third day after germination. Results show that just coleoptile margins were weakly stained in WT seedlings, while the whole of this structure in KDZ-2 and KDZ-4 plants were strongly stained (Figure 5E and Supplementary Figure S6). This result suggests that the rice seed cuticle becomes defective when OsZOU-1 expression is knocked down.

A series of over-expressed transgenic lines, in which OsZOU-1 cDNA was driven by a maize ubiquitin promoter for efficient expression in monocots (Chen et al., 2009),were generated to explore OsZOU-1 function in rice. Although seven independent transgenic plants were generated, their phenotypes were similar and so just two, OXZ-4 and OXZ-5, were selected for further analysis. These plants exhibited dwarfism and leaf curling, had narrowed leaves, glumes, and kernels, and poor fertility (Figures 6A–E and Supplementary Figures S7, S8). The expression of OsZOU-1 in unopened OXZ-4 flowers was 206 times the level in WT plants, and in unopened OXZ-5 flowers was 152 times the level in WT plants (Figure 6E). At the same time, OXZ-4 plant was infertile and more dwarfed than their OXZ-5 counterpart (Figure 6A and Supplementary Figure S7), which exhibited the consistency of the phenotypic severity with the expression levels. Although the OXZ-5 plant produced fertile seeds, germination was difficult and inconsistent (Figure 6F and Supplementary Figure S9).

The results of this study also show that the narrow kernel phenotype is inherited maternally. All the glumes and kernels of over-expressed plants were narrow in this study; these narrow kernels developed into both over-expressed and WT plants that had normal glumes and kernels (Figure 6G). This means that kernel width is controlled by the glume, a component of the female parent, and not by the genotype. Cell number and length along the widest part of the glume was therefore measured in order to explore the direct cause of this narrow phenotype; data show that the glume cell numbers in over-expressed plants exhibited no obvious differences when compared to the WT although the cell length of glumes in OXZ-4 (61.89 μm) and OXZ-5 (64.78 μm) were both significantly smaller than in the WT (73.85 μm) (Figure 6H). These results show that the correct expression of OsZOU-1 is important for both rice seed and plant development.

Although ZOU has previously been shown to be important in endosperm breakdown and cuticle formation during Arabidopsis seed development (Wei et al., 2002; Yang et al., 2008; Xing et al., 2013; Denay et al., 2014), its function in rice seed development remains poorly understood. The results of this study show that although there are two most closely related homolog of AtZOU in rice, just OsZOU-1 complements the defects of cuticle integrity, endosperm breakdown, and embryo expansion in the Atzou-4 mutant, whereas OsZOU-2 failed to rescue Atzou-4. In OsZOU-1 knock-down lines, the endosperm cavity did not form at 12 DAP with embryo adhering to the endosperm, which might be caused by the PCD failure in the endosperm adjacent to the embryo since endosperm cavity appears as soon as the endosperm adjacent to the embryo has degraded in maize and wheat (Young and Gallie, 2000; Cosségal et al., 2007). This result implies that OsZOU-1 shows functional conservation with AtZOU.

The ZOU gene is found ubiquitously from mosses to seed plants, its presumed ancestral function is promoting the separation of the embryo from surrounding gametophytic tissues, as well as their subsequent breakdown (Yang et al., 2008). To date, ZOU orthologs have been identified in Arabidopsis, soybean, maize, and rice; interestingly, just only one copy of ZOU in each species can complement the Arabidopsis zou mutant phenotype, even though in soybean with duplicated genome. This may imply that although the ZOU gene performs important and conserved roles during endosperm development, a number of duplicated genes have diverged and lost this original function because of its highly specific expression pattern.

The phylogenetic analysis presented in this paper also suggests that ZOU genes in monocotyledonous plant species can be divided into three subgroups (Supplementary Figure S2). In this context, it is interesting to note that ZOU orthologous genes that complement the zou mutant phenotype, such as OsZOU-1 and ZmZOU, occur in clade1 while their corresponding homologs generated by genomic duplication or gene duplication all occur in clade3. Similarly, the ZOU genes that are duplicated in Brachypodium distachyon (Bradi) and B. stacei (Brast) occur within clade2; whether, or not, these perform a similar role to their counterparts in clade1 currently remains unclear. Since OsZOU-2 was highly expressed in vegetative tissues, and cannot complement the zou mutant phenotype, it can be hypothesized that the ZOU orthologousgenes in clade3 may have lost the original function of this gene via changes of their expression patterns or protein sequence conservation.

Differentiation in endosperm development between dicots and monocots confirms that research on ZOU in species of the latter is valuable for exploring its functional conservation or variation associated with endosperm evolution. As a result, further experiments are needed to explore the role of ZOU in regulating endosperm development in rice, and comparisons of the role of ZOU in monocotyledon species versus certain dicotyledonous ones that possess non-degraded endosperm are also useful in investigating the evolution of endosperm development process.