Zein proteins make up the majority of maize (Zea mays) endosperm proteins, and the most abundant of those are the α-zeins, which are encoded by highly duplicated gene families (Holding and Messing, 2013). Through coordinated gene expression and specific interactions during the middle phase of endosperm development (Woo et al., 2001;Kim et al., 2002;Wu and Messing, 2010;Guo et al., 2013), zeins are organized into ER-localized and discretely organized proteins bodies (Lending and Larkins, 1989). Mutations that reduce zein abundance, either direct mutations or mutations of regulatory factors, result in changes to protein body shape, size, and number and partly or completely disrupt the formation of vitreous endosperm resulting in opaque kernels (Holding, 2014). Such opaque mutants include floury-1 (fl1) and opaque-1 (o1) (Holding et al., 2007;Wang et al., 2012). The most well-known of the maize opaque mutants is opaque-2 (o2), which has been widely studied because of its increased lysine and tryptophan content (Mertz et al., 1964). The amino acid improvements in o2 result from substantially reduced accumulation of lysine-devoid zeins and the compensatory increase in lysinecontaining non-zein proteins. Although the non-zein changes largely affect the proteome globally, discrete qualitative changes in especially lysine-rich proteins are also important (Morton et al., 2016). Cloning of the O2 gene revealed that it encodes a bZIP transcription factor that regulates α-zeins (Schmidt et al., 1990). O2 is now known to regulate most classes of zein genes (Li et al., 2015) and to act in concert with other transcription factors, OHP and PBF (Zhang et al., 2015). Furthermore, O2 plays a wider role in orchestrating storage acquisition in endosperm since it both directly and indirectly regulates several starch synthesis genes (Zhang et al., 2016).
The Arabidopsis RNA-DIRECTED DNA METHYLATION 4 (RDM4) mutant shows a dwarf plant phenotype and vulnerability to cold stress (He et al., 2009;Chan et al., 2016). It also exhibits pleiotropic developmental abnormalities under normal growth conditions and extensive modification of the RNA polymerase II (Pol II)-mediated transcriptome, and the RDM4 protein was shown to associate with Pol II, Pol IV, and Pol V (Matzke et al., 2015). The complexity of the factors that interact with Pol II may provide differential transcription in response to specific biotic, abiotic and developmental scenarios. For example, the Arabidopsis rdm4 mutant showed diminished expression of several cold responsive genes under normal development, providing a clue that the gene is involved in adaptation to at least one abiotic stress (He et al., 2009). Subsequently, it was shown that cold hypersensitivity is one aspect of the pleiotropic rdm4 mutant phenotype and, most significantly, that the RDM4 protein directs Pol II transcription of these cold-responsive genes in an RNA-directed DNA methylation (RdDM)-independent manner (Chan et al., 2016). This raises the question of the possible involvement of RDM4 in cold responsiveness in other plant species including monocots. It also invites the question of whether RDM4 and other factors are involved in other transcriptional scenarios involving stress-related or tissue-specific gene expression such as in maize endosperm.
We previously used γ-radiation to create deletion mutants in the B73 background including opaque kernel mutants, and implemented a bulked segregant RNA-seq (BSR-seq) and exome-seq based mapping pipeline (Jia et al., 2016). We subsequently developed bulked segregant exome sequencing (BSEx-seq) for mapping and identifying a short deletion on chromosome 10 in mutant 1486 (Jia et al., 2018a, b). The causal deletion had an estimated size of 6248 bp with genomic coordinates of two break points at 1 540 105 and 1 546 352, and covered three predicted genes (GRMZM2G098603, GRMZM2G098596, and GRMZM2G176546) based on the B73 genome assembly v3. The deletion was confirmed by our exome-seq analysis and genomic PCR (Jia et al., 2018b). In the present study, we established the causality of the loss of RDM4 for opaque kernel and plant phenotype in this mutant, henceforth named Zm-rdm4. Through characterization of its phenotype and changes of transcriptome and proteome, we highlight the differences between rdm4 and o2 in the nature of zein reduction and non-zein proteome rebalancing in the opaque kernel endosperm. We present leaf transcriptome data that may suggest an overarching regulatory role for RDM4 in contributing to tissue-specific and stress-related gene expression networks.
The rdm4 mutant was created as previously described (Jia et al., 2016(Jia et al., , 2018b)), and grown in the greenhouse and field for collection of leaf and developing ears. For endosperm RNA-seq, 20 d after pollination (DAP) embryo DNA was extracted for genotyping according to a previously published method (Jia et al., 2018b) while endosperms were flash frozen in liquid nitrogen and stored at -80 °C. Wild type (WT), heterozygous and opaque mutant kernels were identified based on two genomic PCRs by using two primer pairs, RDM4_E3F and RDM4_E2R, for a 526-bp fragment inside the gene RDM4, and RDM4_E5F and AMP_E3R for fragments of only ~1 kb in rdm4 and ~7 kb in WT (see Supplementary Table S1 at JXB online). Endosperm RNAs genotyped for both homozygous WT and mutant types were extracted, processed by RNase-free DNase I, and purified using the Plant RNeasy Kit (Qiagen) for RNA-seq.
Seedling cold stress treatments were performed by first germinating the seeds in standard potting compost in 3-inch peat pots, growing in the greenhouse and subsequently in a Conviron growth chamber. The seed-ling28 treatment consisted of 14 d growth at 28 °C in the greenhouse (16 h day, 8 h night). For mild cold stress of seedling10, plants were grown at 28 °C for 11 d followed by 10 °C for 3 d. For severe cold stress of seedling2, seedlings were grown at 28 °C for 11 d, 10 °C for 3 d followed by 2 °C for 1 d. Wild type and mutant seedlings were determined based on both the seed opaque phenotype and leaf genomic DNA genotyping. DNA was extracted from leaves of ~14-day-old plants and genotyped for WT, heterozygous and mutant plants, before seedling RNA was extracted and purified for RNA-seq.
cDNA was synthesized for constructing Illumina sequencing libraries by using NEBNext® Ultra™ II DNA Library Prep (NEB, Ipswich, MA, USA). Groups of six endosperm and 18 seedling samples were sequenced for RNA-seq on separate HiSeq 2500 lanes, with a 125-bp paired-end run and v4 chemistry. The exome capture and sequencing were performed on pooled genomic DNA of 14-day-old normal and mutant seedlings by using SeqCap EZ Developer Maize Exome kit (Roche NimbleGen, Madison, WI, USA) (Jia et al., 2018b). All the sequencing was performed in the University of Minnesota Genomics Center. The data are deposited in the Short Reads Archive (SRA) database (http://www.ncbi.nlm.nih. gov/sra) of NCBI with the accession number of PRJNA306879.
We designed primers using Primer3 and our locally developed perl script for batch processing to do the RT-PCR and qRT-PCR (see Supplementary Table S1). Total RNA from endosperm and seedling was extracted by using the Plant RNeasy Kit (Qiagen) and DNase I treatment. First-strand complementary DNA was synthesized using the Bio-Rad iScript™ cDNA Synthesis Kit (Bio-Rad, Hercules, CA, USA). The qRT-PCR analyses were performed in a My iQ icycler (Bio-Rad) using IQ SYBR Green super mix (Bio-Rad), with three biological replicates using maize gene Zm00001d018145 as an internal control since it was found to have invariant C t values between all samples. The relative transcript levels were determined using 2 -∆∆Ct method.
Total protein as well as alcohol-soluble zein and aqueous non-zein fractions were extracted from mature endosperm according to a previous method (Wallace et al., 1990). Maize endosperm was ground, and then suspended in extraction buffer (12.5 mM Na-borate pH 10, 0.1% SDS, 1% β-mercaptoethanol). After pelleting insoluble material (such as starch), the supernatant was diluted in ethanol to 70% ethanol to precipitate the non-zein proteins. The non-zein protein pellet was washed three times with 70% ethanol, air dried, and stored at -80 °C. SDS-PAGE was used to compare single kernels of normal (i.e. WT and heterozygous) and mutant types. All the comparisons between the mature normal and mutant kernels were based on the same amount of kernel flour (Jia et al., 2016;Morton et al., 2016).
For LC-MS/MS analysis, kernels of 20 DAP were genotyped for homozygous mutant and WT alleles, with four replicates. Non-zein pellets were thawed and redissolved in 7 M urea, 2 M thiourea, and 5 mM DTT, and 100 µg of each sample was reduced at 37 °C for 2 h. Proteins were alkylated with 15 mM iodoacetamide, and then quenched with an equimolar amount of DTT. Samples were diluted 9-fold and subjected to trypsin digestion, before being redissolved in 0.1 M triethylammonium bicarbonate. Eighty micrograms of each sample was labeled with 0.8 mg of TMT10-plex reagent using labels 127N-130C only (Thermo Fisher Scientific, Waltham, MA, USA), combining all eight samples into one single 8-plex sample. Three hundred micrograms of this combined sample was sub-fractionated offline into 96 fractions using high pH reverse phase C18 chromatography (ACQUITY UPLC® CSH™ C18, 1.7 µm, 2.1×150mm, Waters Corp.) at pH 10 and then recombined to give a total of 12 fractions according to the previous strategy (Yang et al., 2012). Each of the 12 fractions from the high pH reverse phase run was analysed by LC-MS/MS on an RSLCnano system (Thermo Fisher Scientific) coupled to a Q-Exactive HF mass spectrometer (Thermo Fisher Scientific). The samples were first injected onto a trap column (Acclaim PepMap™ 100, 75 µm×2 cm, Thermo Fisher Scientific) for 3.3 min at a flow rate of 5 µl min -1 , 2% acetonitrile, 0.1% formic acid before switching inline with the main column. Separation was performed on a C18 nano column (Acquity UPLC® M-class, Peptide CSH™ 130A, 1.7 µm, 75 µm×250 mm, Waters Corp.) at 260 nl min -1 with a linear gradient from 5% to 35% over 96 min. The LC aqueous mobile phase contained 0.1% (v/v) formic acid in water and the organic mobile phase contained 0.1% (v/v) formic acid in 80% (v/v) acetonitrile. Mass spectra for the eluted peptides were acquired on a Q Exactive HF mass spectrometer in data-dependent mode using a mass range of m/z 375-1500, resolution 120 000, automatic gain control (AGC) target 3×10 6 , maximum injection time 60 ms for the MS1 peptide measurements. Data-dependent MS2 spectra were acquired by higher-energy collisional dissociation as a top15 experiment with a normalized collision energy set at 33%, AGC target set to 1×10 5 , 60 000 resolution, intensity threshold 5×10 4 and a maximum injection time of 100 ms. Dynamic exclusion was set at 30 s and the isolation window set to 1.2 m/z to reduce co-isolation.
Data were analysed in Proteome Discoverer 2.2 software (Thermo Fisher Scientific) connected to Mascot 2.6.1, which searched the common contaminants database cRAP (116 entries, www.theGPM.org) and the Uniprot reference proteome database for Z. mays (ID: UP000007305 retrieved on 20171001, 99 369 entries). Peptides were validated by Percolator with a 0.01 posterior error probability (PEP) threshold. The data were searched using a decoy database to set the false discovery rate to 1%. The reporter abundance was normalized using total peptide amount. The protein ratios were calculated using summed abundance for each replicate separately and the geometric median of the resulting ratios was used as the protein ratios. The mass spectrometry proteomic data were deposited into the ProteomeXchange database via ProteomeXchange submission tool with the dataset identifier PXD016445.
Protein-bound and free amino acids were extracted and quantified according to a previously published method (Li et al., 2018a;Yobi and Angelovici, 2018). The in silico calculation of amino acid content was conducted according to a previous publication (Morton et al., 2016).
Structural variations were called by Lumpy (Layer et al., 2014) to help determine the deletion positions. RNA-seq data were analysed with the standard protocol with trimming, mapping by TopHat2 (Kim et al., 2013) on the maize genome assembly v4 (downloaded from ftp://ftp. ensemblgenomes.org/pub/plants/release-44/fasta/zea_mays/dna/), and determining mRNA expression fragments per kilobase per million mapped reads (fpkm) values by Cufflinks (Trapnell et al., 2012). For functional enrichment analysis, a bin-wise Wilcoxon test in MapMan v3.6 was conducted (Usadel et al., 2006(Usadel et al., , 2009)). R functions prcomp() and ggbiplot() were employed for principal component analysis (PCA) and visualization. Significantly differentially expressed genes (DEGs) and differential proteins (DPs) were identified by the R package DESeq2 (Love et al., 2014).
The maize deletion mutant rdm4 had apparently normal-sized opaque kernels for the mutant alleles (-/-), through which light cannot penetrate (Fig. 1A). The mutant was crossed with Mo17 to make an F 2 mapping population and confirm the robust heritability of the mutant opaque kernel phenotype since segregation of opaque and vitreous kernels occurred in F 2 ears (Fig. 1B). Self-pollinating the heterozygous F 1 plants containing rdm4 allele gave F 2 kernels segregating normal sized opaque kernels for the mutant alleles (-/-), and normal sized vitreous kernels for alleles (-/+ and +/+), with ratios of about 1:2:1, indicating the recessive nature of the causal mutation. The homozygous opaque seeds of rdm4 germinated normally and gave rise to morphologically normal seedlings. However, vegetative growth was slightly slower and adult plants were shorter, as shown at 2-week, 1-month, and adult stages (Fig. 1C-F). Male fertility was negatively affected such that only a few seeds were produced when plants were self-pollinated (Fig. 1G). Female fertility appeared to be normal as moderately wellfilled ears resulted from using other pollen sources.
The causal deletion that was initially annotated to contain three genes (GRMZM2G098603, GRMZM2G098596, and GRMZM2G176546) was mapped and identified previously by using BSEx-seq (Jia et al., 2018b). Here we re-estimated this deletion to be 5731 bp with genomic coordinates of two break points at 1 100 088 and 1 105 820 based on the B73 genome assembly v4, rather than 6248 bp previously estimated from the B73 v3 genome (Fig. 2). Two paired-end and 12 single-end reads in exome-seq, and the Sanger sequences of cloned PCR product with the primer pair of RDM4_E5F and AMP_E3R both covered this junction site and two break points (Fig. 2). The deletion actually contains only two predicted genes, an RDM4like gene (Zm00001d023237) and an AMP-binding-like protein (Zm00001d023238). The opaque mutant phenotype results from loss of the RDM4-like gene (see confirmation below).
To confirm the causal mutant gene in the deletion corresponding to opaque kernel and vegetative phenotypes, we obtained 10 UniformMU insertion lines in the two genes (see Supplementary Table S2). For all the UniformMU progeny, it was shown that mutations in the AMP-binding like gene had normal kernels and plants (Supplementary Fig. S1A) indicating that loss of this gene is not responsible for either phenotype. In contrast, segregation for opaque kernels was observed in the heterozygous UniformMU insertion lines for RDM4 gene. In the four UniformMU lines for RDM4, including UFMU000942 and UFMU06648, the received seeds and progeny seeds segregated for an opaque kernel phenotype similar to mutant rdm4 (Supplementary Fig. S1B,C) and opaque seeds gave rise to plants with reduced height like mutant rmd4 (Fig. 3). In the genotyped homozygous mutant plants for the lines of UFMU000942 and UFMU06648, the same vegetative phenotype was observed. For example, plant heights were significantly different between WT and mutant (Fig. 3B,C,E), and similar mutant plant heights were seen for UFMU000942 and rdm4 (Fig. 3D). The homozygous UFMU000942 and UFMU06648 mutant alleles produced almost sterile tassels (Fig. 3A), which was consistent with this phenotype in mutant rdm4. Finally, allelism test crosses between heterozygous rdm4 plants and the independent UniformMu alleles in UFMU00942 and UFMU06648 showed that these mutants fail to complement each other (Fig. 3F,G), and mutations in RDM4 cause the observed seed and vegetative phenotypes. RDM4 comprises 10 exons and nine introns, and encodes a protein containing 318 amino acid residues.
The opaque kernels in rdm4 likely result from the striking reduction in zein storage proteins in an analogous way to opaque-2 (Fig. 4A). However, zein protein and mRNA expression analysis showed some key differences from o2. While the most striking effect in o2 was reduction of the 22 kDa α-zeins, in rdm4 the 19 kDa α-zeins gene families (Z1A, Z1B, and Z1D) were clearly more reduced than the 22 kDa α-zeins (Z1C) at both the protein and the transcript level (Fig. 4A). This suggests that RDM4 may play a role in high-level mRNA expression of the extensively duplicated 19 kDa α-zein genes. Intriguingly, the O2 gene itself showed substantial reduction in expression (see Supplementary Fig. S2A,B), raising the possibility that RDM4 acts upstream of O2 in the endosperm mRNA expression hierarchy and may even play an overarching role in endosperm development. Unlike the opaque-2 mutant, which shows a substantial global increase in non-zein proteins, SDS-PAGE showed that such proteome rebalancing in rdm4 is not as pronounced as in o2 (Fig. 4B). Instead, rdm4 shows discrete increases and decreases of a few nonzein proteins (Jia et al., 2018b). Similar profiles of zein and non-zein proteins were observed in rdm4 deletion allele and UniformMU alleles (Fig. 4C,D). Heterozygous kernels had similar zein mRNA expression values as homozygous WT kernels in both RT-PCR and qRT-PCR, and this also applied to other related genes such as O1, O2, O10, OHP1, and OHP2 (Supplementary Fig. S2A,B). This slight and selective proteome rebalancing was reflected in an increase in protein-bound lysine (Lys, 1.4-fold, P<0.001) which is less than observed in the o2 mutant (2 fold, P<0.001) (Table 1). The accumulation of most free amino acids was increased in Fig. 2. Causal deletion covering RDM4 gene in mutant rdm4. A deletion was identified exactly underneath the linkage peak by BSEx-seq, and covered the promoters and partial exons of two genes, Zm00001d023237 (RDM4) and Zm00001d023238. The deletion was determined to be of 5731 bp, with two joint sites' sequences shown, based on the cloning and sequencing of PCR product by primers (RDM4_E5F and AMP_E3R mature o2 seeds (average 3.3-fold increase across all amino acids; Table 1), which is likely a result of reduced zein synthesis and reduced incorporation into non-zein proteins. Mutant rdm4 seeds showed an even more profound increase in free amino acids in rdm4 (average 7.8-fold increase across all amino acids; Table 1), which likely reflects the general reduction in protein synthesis.
A shotgun proteome analysis involving 10-plex Tandem Mass Tag (TMT 10 ) labeling of the non-zein proteins extracted from developing endosperm at 20 d after pollination (DAP) was WT and the mutants could be separated spatially by PC1 (Fig. 5A). We compared our proteomic data with that of the o2 mutant in our previously published work (Morton et al., 2016), to identify commonalities and differences in rebalancing of the non-zein proteome. The proteomic analysis showed that the calculated lysine (abbreviated as K) content of the most abundant non-zein proteins (top 10 and top 20) in rdm4 and o2 was consistent with the measured lysine increase in Table 1, compared with WT (Fig. 6A). Lysine increase in the up-regulated non-zein proteins was globally significant in rdm4 (see Supplementary Table S3), and less than that in o2. The amino acid content changes in rdm4 were very distinct from other opaque mutants (Morton et al., 2016). We compared the mutant and WT based on the abundance data, and produced 4075 and 303 DPs, respectively, in rdm4 versus WT and o2 versus WT (Table 2). It is not meaningful to draw inferences from the number of DPs in rdm4 compared with that of o2. This is because the proteomic data in the two mutants were collected using two different methodologies. However, we can compare the relative number of increased and decreased proteins in each mutant. A higher proportion of DPs were decreased in o2 (61.39%) than in rdm4 (51.04%). Given that the SDS-PAGE (Fig 4B ) clearly showed that proteome rebalancing was different in o2 compared with rdm4, it is likely that the highly abundant proteins visible on the SDS-PAGE and contributing to the general increase in the non-zein proteome in o2 are among the 38.61%. The fold changes of both increased and decreased DPs appeared to be greater, on average, in o2 than in rdm4, as indicated by more green and red dots further away from the zero axis in Fig. 6B. We also calculated the size of the DPs (P<0.05) whose peptide sequences were downloaded from the Uniprot database in the two mutants and showed a difference of fold change of the increased (green dots in Fig. 6B) and decreased (red dots in Fig. 6B) proteins between o2 and rdm4 over their length distribution. It is likely that some of the commonly increased or decreased proteins in rdm4 and o2, as well as common proteins with opposite trends, were visible as differential bands in the SDS-PAGE gels of non-zein proteins in rdm4 and o2 (Fig. 4B). Of the DPs (P<0.05 and |log2(fold Relative free and protein-bound (PB) amino acid fold increases (more than 1) or decreases (less than 1) in mature kernel flour from B73 opaque-2 and B73 rdm4 mutants compared with normal sibling kernels. Ratios were calculated from average % (w/w) amino acid against kernel flour from four biological replicates (data not shown). In PB amino acids, N/D represents asparagine/aspartic acid, Q/E represents glutamine/glutamic acid, and W (tryptophan) and C (cysteine) were not detected with the acid hydrolysis used. 'All aa' is the average change across all detected amino acids.
Downloaded from https://academic.oup.com/jxb/article/71/19/5880/5872007 by University of Nebraska-Lincoln Libraries user on 06 July 2022 change)|>0.5) found in rdm4 from the proteomics, 37 of these were also among the DPs found in o2 although some of them were opposite in their increase or decrease between the two mutants. The DPs accordingly separated into four groups. One protein was increased in both rdm4 and o2 (green colored line in Fig. 6B), five proteins were increased in rdm4 and decreased in o2 (yellow colored line in Fig. 6B), 17 proteins were decreased in rdm4 and increased in o2 (red colored line in Fig. 6B), and 14 proteins were decreased in both rdm4 and o2 (blue colored line in Fig. 6B). For example, elongation factor 1-α (B6TWN7) as an indicator of protein synthesis was increased and decreased in rdm4 and o2, respectively (Table 3). These differences suggest that proteome rebalancing of non-zein proteins in the two mutants is quite distinct. Interestingly, we found proteins with different regulation patterns in subcellular locations. Of the 98 DPs (P<0.05, fold change>2), 51 and 47 DPs were significantly increased and decreased, respectively (Supplementary Table S3). It is notable that for subcellular location, 18 out of 51 increased DPs predicted were predicted to be nuclear localized, while all the 47 down-regulated DPs were predicted to be non-nuclear localized, based on the search in the UniProt website. We identified 50 kDa and 27 kDa γ-zeins decreased in the list of most significant DPs (Supplementary Table S3). γ-Zeins are known to partially fractionate with non-zein proteins, unlike the hydrophobic α-zeins. The enrichment analysis of DPs on the STRING website showed that the increased DPs were enriched in the KEGG pathway of 'ribosome biogenesis', while the decreased DPs were enriched in 'plant hormone signal transduction'. Endosperm RNA-seq analysis in WT versus rdm4 was used to test whether they were consistent with the protein abundances in the mutant versus WT. Of the 51 increased DPs (see Supplementary Table S3), 46 of the corresponding transcripts were with fold change >1, and 39 transcripts were significant (P<0.05). Of the 47 decreased DPs, 41 of the corresponding transcripts were with fold change <1, 18 of which were significant. For example, elongation factor 1-α (Uniprot ID: B6TWN7; Gene ID: Zm00001d021788) and rRNA methyltransferase (Uniprot ID: A0A1D6FKG8; Gene ID: Zm00001d009597) as indicators of protein synthesis were increased with a consistency in protein and transcript abundance, as was histone 2B (Uniprot ID: B4FZQ6; Gene ID: Zm00001d005789). Down-regulated corresponding genes were also significantly differentially abundant between WT and mutant in RNA-seq. For example, the transcript abundance of granule-bound starch synthase (Uniprot ID: Q5NKP6; Gene ID: Zm00001d033937) also corresponded to a protein decrease. The 20-DAP endosperm RNA-seq analysis was used to glean clues as to the function of RDM4 in endosperm. Genotyping using embryo DNA was used to identify homozygous mutant and WT alleles. Based on genotyping results, endosperm tissue was used for RNA extraction for RNAseq. The three replicates of WT and mutant were grouped well by PC1 in the PCA (Fig. 5B). As expected, the RDM4 (Zm00001d023237) transcript was not detected in the mutant, while it was in abundance in WT. In total, 3581 DEGs, including 2022 up-regulated and 1559 down-regulated transcripts, were found significant (P<0.05) between WT and mutant (see Supplementary Table S4). More than 80 down-regulated DEGs were annotated as transcription factors, including O2 (Zm00001d018971), and zein genes were comprehensively down-regulated, including 50 kDa γ-zein (Zm00001d020591), 19 kDa zein (Zm00001d048848, Zm00001d030855, Zm00001d048847, Zm00001d048849, Zm00001d048850, Zm00001d048851, Zm00001d048852), 22 kDa α-zein (Zm00001d048806 and Zm00001d048807), and 16 kDa zein (Zm00001d005793). These were consistent with RT-PCR results (Supplementary Fig. S2). In addition, >109 transcription factors were up-regulated. These results suggest that developing endosperm zein and non-zein mRNA expression profiles were significantly influenced by the loss of RDM4. To identify the key pathways affected, the fold-change values of 1500 DEGs (P<0.001) for endosperm were subjected to a Wilcoxon bin-wise statistical test and plotted in MapMan (Supplementary Fig. S3). This showed a pathway enrichment, including an increase of protein biosynthesis (especially ribosome biogenesis) and histones in endosperm of rdm4, and a decrease of components of RdDM, vesicle trafficking, solute transport, redox homeostasis, ethylene biosynthesis, lipid metabolism, starch metabolism, and protein modification.
Given the evidence that RDM4 plays a role in mediating cold stress in Arabidopsis (Chan et al., 2016), we addressed whether transcript profiles suggest it might play a similar role in maize stress responses. While RDM4 is equally likely to play a broader role in stress responsiveness, we focused on cold for the above reason. We investigated the transcriptomic changes resulting from the loss of RDM4 in maize vegetative development in normal (28 °C, 16 h day, 8 h night for 14 d; seed-ling28), mild cold stress (28 °C for 11 d followed by 10 °C for 3 d; seedling10) and near freezing severe cold stress (28 °C for 11 d, 10 °C for 3 d followed by 2 °C for 1 d; seedling2) conditions. Rdm4 mutant plants were visibly less healthy than WT after 3 d at 10 °C (seedling10), exhibiting significant leaf necrosis. Similarly, WT plants were obviously more resistant to severe cold stress (seedling2) since their leaves were less necrotic than those of mutant plants (see Supplementary Fig. S4A,B). Transcriptomic analysis in leaf tissue confirmed that the expression of RDM4 was absent in the mutant as expected. A different transcript pattern was observed in seedling28 of rdm4 mutant, with 848 up-and 1180 down-regulated DEGs between WT and mutant (Supplementary Table S4). The pathways of photosynthesis, nutrient uptake, cell cycle organization, and redox homeostasis were impaired in seedling28 of rdm4 mutant, while lipid metabolism, protein biosynthesis and modification, cell wall organization, and chromatin organization were enhanced (Supplementary Fig. S3).
The RNA-seq analysis identified extensive changes in transcriptome for response to mild and severe cold stresses that appeared to be independent of RDM4. Cold stress (seedling10 and seedling2) separated the samples from seedling28 based on the transcripts in RNA-seq (Fig. 5B) and resulted in up-/ down-regulated DEGs, with 1329/849 and 345/739 for seed-ling10 and seedling2, respectively (see Supplementary Table S4). Based on the transcriptome data, we identified tissuespecific expression in endosperm and leaf (Fig. 5C), and found only 18 transcripts were all significantly changed in the four comparisons of mutant versus WT in endosperm, seedling28, seedling10 and seedling2 (Table 4). A spearman correlation between RDM4 and all the other genes was calculated across all the 18 seedling samples, and the significance was assessed by using Student's t-test. In total, 1306 genes with significant P values (P<0.05) for correlation coefficient were observed. In this way, we identified five genes significantly positively correlated to RDM4 (Table 4). In the same way, 29 DEGs (P<0.01) in the three pair-wise comparisons for seedling leaf were shown with significant correlation values (P<0.05), including 13 genes negatively correlated and 16 genes positively correlated (Supplementary Table S5).
To dissect the possible contribution of RDM4 to the response to cold stress in the seedlings of rdm4 mutants, we conducted four pairwise comparisons of transcriptomic expression, i.e. in WT leaves between seedling28 and seedling10 (for the DEGs by both RDM4 and cold stress) and between seedling28 and seedling2 (for the DEGs by both RDM4 and cold stress), in rdm4 mutant leaves between seedling28 and seedling10 (for the DEGs by only cold stress) and between seedling28 and seedling2 (for the DEGs by only cold stress). In the Venn diagram, 4624 DEGs overlapped in the four comparisons (Fig. 5D, Supplementary Table S4), and they contributed to response to cold stress, but their regulation was not directly related to RDM4. We observed some DEGs apparently related to RDM4. There were 8410 and 10 559 DEGs in a comparison of seedling28 versus seedling10 and seedling28 versus seedling2, respectively in WT. Of these, 2177 (25.9%) genes were not differentially expressed in rdm4 mutant (WT seed-ling28-seedling10 versus mutant (MU) seedling28-seedling10, i.e. 1003 + 844 + 68 + 262=2177 in Fig. 5D). Near freezing treatment for seedling2 induced more genes with a potential regulation of RDM4, as 4310 (40.8%) genes were not found in WT seedling28-seedling2 versus MU seedling28-seedling2, i.e. 1904 + 1003 + 459 + 944=4310 (Fig. 5D). This suggested that the loss of RDM4 caused a subset of cold-stress-responsive genes to lose responsiveness to cold. This subset was larger in the more severe cold stress.
Of the large group of DEGs with a potential regulation of RDM4, we selected a subset of representative genes whose cold-induced transcript expression was not observed or was much less pronounced in rdm4 leaves to conduct a real time qRT-PCR analysis. The results supported that RDM4 may play an overarching role in coordinating transcriptional responses to cold (Fig. 7; Supplementary Fig. S5). RDM4 was induced by cold stress in WT (Fig. 7A). Based on the expression pattern shaped by a loss of RDM4 and cold stress, we observed four expression patterns. First, some genes' expression was increased in rdm4 mutant at 28 °C but not increased in rdm4 in response to cold (see the left panel in Fig. 7B-E; Supplementary Fig. S5A-L), which indicated a loss or alleviated repression in WT in cold stress and was consistent with RNA-seq results. We observed strong induction in response to cold (10 °C and/or 2 °C) in WT, but a similar cold induction is either substantially reduced or absent in rdm4 mutant (see the right panel in Fig. 7B-E; Supplementary Fig. S5A-L). Second, transcript expression of DNA-directed RNA polymerase II subunit 1 (NUCLEAR RNA POLYMERASE B1, i.e. NRPB1, Zm00001d024583) and histone 3 was unchanged or slightly increased in rdm4 mutant at all temperatures, and cold-induced expression changes in WT were similar in rdm4 mutant (Fig. 7F,G), as well as methyl transferase (Zm00001d027329) and auxin transcription factor (Zm00001d018414) (see Supplementary Fig. S5M,N). Third, expression of jasmonate methyltransferase (Zm00001d052827) was impaired in rdm4 mutants, and cold stress also contributed to its expression changes (Fig. 7H), which is consistent with the correlation results (see Supplementary Table S5). Fourth, conversely, the expression of cytokinin riboside (Zm00001d038862) and 60S ribosomal protein (Zm00001d007287) was significantly increased in rdm4 mutant at all temperatures, and we found strong induction in response to cold in WT, but a lesser response to cold in rdm4 mutant (Fig. 7I,J).
When the expression of zein storage proteins is interrupted by loss of regulatory factors such as opaque-2, compensatory mechanisms result in the increase of non-zein proteins and relatively constant levels of total proteins in maize seeds. This compensatory shift has been referred to as 'proteome rebalancing' (Holding, 2014;Wu and Messing, 2014;Li and Song, 2020). Proteome rebalancing has also been observed in sorghum (da Silva et al., 2011), rice (Kawakatsu et al., 2010;Takaiwa et al., 2018) and soybean (Schmidt and Herman, 2008).
Vegetative phenotypes were reported in the Arabidopsis mutant for RDM4/DEFECTIVE IN MERISTEM SILENCING 4 (DMS4) (He et al., 2009;Kanno et al., 2010) and no mention was made of a role in seed storage proteins. In the maize rdm4 mutant, similar vegetative phenotypes were accompanied by an opaque kernel phenotype resulting from reduced zeins and the resultant proteome rebalancing led to altered amino acid profiles (Table 1; Fig. 6A). Loss of RDM4 function substantially changes the endosperm transcriptome and proteome, although it is not yet known the extent to which RDM4 is directly and indirectly involved in regulatory processes. The decrease in zeins and increase in non-zeins (proteome rebalancing) were shown in rdm4 reflected by a proteome that was completely different from WT, and supported by RNA-seq data. The classical mutant opaque-2 has long been known to increase kernel lysine content through reduced accumulation of zeins and a compensatory increase in non-zeins and this results from the loss the multi-function O2 transcription factor (Moro et al., 1996;Holding and Larkins, 2006). Although rdm4 also shows generally reduced zeins, the SDS-PAGE comparison shows that the increase in non-zein proteins is not as pronounced as that of o2. This may reflect a more generalized and overarching role for RDM4 than for O2 in the endosperm proteome. We therefore compared the proteome changes of WT:rdm4 to that of WT:o2. Although the two proteomic datasets were derived from different methodological platforms that yielded very different total numbers of DPs, we were able to identify proteins that were either commonly increased or commonly decreased or showed opposite trends between rdm4 and o2 (Fig. 6B).
Interestingly, comparison of the data sets suggested higher fold changes of DPs in o2 than in rdm4, which might suggest that a relative high abundance of the minority fraction of nonzein proteins in o2 made the most substantial contribution to the visible bands on SDS-PAGE gels (Fig. 4B). The generally higher fold changes in DPs in o2 than in rdm4 is consistent with O2 being required for expression of only a subset of genes expressed in the endosperm while RDM4 likely contributes to the transcriptional in a more general sense and for a greater fraction of genes expressed in the endosperm. Overall, these results suggest that proteome rebalancing is a complicated process likely affected by multiple levels of regulation.
It is known that zein accumulation accounts for almost a half of storage proteins in maize, but zeins lack lysine and tryptophan (Holding, 2014). The proteome analysis of o2 mutant indicated that some lysine-rich proteins, such as sorbitol dehydrogenase and glyceraldehyde-3-phosphate dehydrogenase, had increased abundance in mature kernels of o2 mutant (Jia et al., 2013). Since the rdm4 mutant displays negative and pleiotropic vegetative and reproductive phenotypes that would preclude its use as a source of increased lysine grain, we were nonetheless interested to understand the basis of its increased kernel lysine. We used proteomic data to evaluate both increased and decreased proteins on the amino acid content, and found that the in silico-calculated lysine content was increased in the top 10 and top 20 most abundant proteins (Fig. 6A) and in most of the significant up-regulated proteins (see Supplementary Table S3). The results explained the lysine increase by the DPs, which were not the same proteins as those causing the lysine increase in o2. For example, elongation The left graphs show that expression of the four genes was increased in rdm4 mutant at 28 °C but not increased in rdm4 in response to cold. The right graphs show strong induction of the genes in response to cold (10 °C and/or 2 °C) in wild type, but a similar cold induction is either substantially reduced or absent in rdm4 mutant. (F, G) The left graphs show that expression of the two genes was slightly increased in rdm4 mutant at all temperatures. The right graphs show that cold induced expression changes in wild type (increases or decreases) are similar in rdm4 mutant suggesting regulation of this gene by cold stress is independent of RDM function. (H) The left graph shows that expression of this gene was reduced in rdm4 mutant at all temperatures. The right graph shows strong induction of this gene in response to cold (10 °C and factor 1-α, which showed a correlation with lysine increase (Habben et al., 1995), was significantly increased in rdm4 with lysine content of 10.5% (Supplementary Table S3). The analysis confirmed the substantial albeit distinct qualitative and quantitative increase in key lysine-rich non-zein proteins may contribute to lysine improvement in rdm4, although this lysine increase is less significant than that in o2 (Morton et al., 2016).
Potential role of RDM4 in regulation of known cold-stress-responsive genes RDM4 has been reported to be involved in the cold response in Arabidopsis, and affects the expression of several cold-stressresponsive genes, especially CBFs/DREBs via promoter occupancy (He et al., 2009). The RdDM pathway is not thought to be required for cold stress responses, and the association of RDM4 and Pol II is activated under chilling conditions (Chan et al., 2016). We observed more than 2000 cold-stressresponsive genes with significantly differential transcript expression in response to cold in WT, which were not significantly differentially regulated in rdm4, possibly suggesting an overarching function for RDM4 in abiotic stress tolerance (Figs 5D, 7, Supplementary Fig. S5). For example, inducer of ICE1 (Fig. 7E) is a master regulator of CBFs in cold stress, and mitogen-activated protein kinase 3 (MPK3, Supplementary Fig. S5D) negatively regulates CBF expression in plants (Li et al., 2017). ICE1, a bHLH transcription factor, was identified as an inducer of CBF expression and a key regulator of CBF genes under cold conditions (Li et al., 2016). Higher levels of superoxide dismutase (Fig. 7C) were found after freezing treatment in the overexpression of ICE1 (Zuo et al., 2019). Overexpression of the ABC transporter gene (Fig. 7D) increases abiotic stress tolerance in Arabidopsis (Chen et al., 2018). Glycerol-3-phosphate acyltransferase (Supplementary Fig. S5F) plays a pivotal role in cold resistance in a variety of plant species (Li et al., 2018b).
Jasmonic acid (JA) and its methylated ester, methyl jasmonate (MeJA), have been shown to control aspects of flowering, plant growth and development, and responses to various abiotic stresses (Huang et al., 2017). A lower DNA methylation level and significant up-regulation of the genes involved in JA biosynthesis were found in salt-tolerant line ND98 in sweet potato (Zhang et al., 2017). The JA pathway was involved in plant development and stunted plant growth was accompanied by increased jasmonate as a growth inhibitor (Zhang and Turner, 2008). In the CBF signaling pathway, epigenetic regulation plays an important role in modulating mRNA expression under cold stress (Ding et al., 2019), and RDM4 is important for Pol II occupancy at the promoters of CBF2 and CBF3 genes (Chan et al., 2016). Overexpression of JASMONATE ZIM-DOMAIN 1 (JAZ1, Supplementary Fig. S5I) or JAZ4 repressed freezing stress responses, and jasmonate acted as an upstream signal of the ICE-CBF transcriptional pathway to positively regulate freezing stress responses in Arabidopsis (Hu et al., 2013). JA was involved in leaf senescence and tolerance to cold stress (Hu et al., 2017). In this study, jasmonate O-methyltransferase and other jasmonate-induced genes were found significantly down-regulated in rdm4 mutant leaves (see Supplementary Tables S4,S5). Based on the strong positive correlation and expression pattern, it is hypothesized that jasmonate-related mRNA expression is highly dependent on RDM4, and regulates the downstream cold stress responses and vegetative growth in maize leaves.
Histones had increased abundance in both proteomic and transcriptomic analysis in endosperm and seedling of rdm4 (Supplementary Table S3; Table 4; Fig. 7G) as had the histonerelated proteins such as histone deacetylase 2b (B6SK06). We observed a similar increase for NRPB1 in rdm4 mutant (Fig. 7F). RDM4 ortholog Iwr1 was found in yeast, and is important for the import of Pol II from the cytoplasm into the nucleus and its regulation (He et al., 2009). Interestingly, there is an expression increase for NRPB1, but a decrease for histone, in WT in response to cold stress (Fig. 7F,G). RDM4 interacts with Pol II, Pol IV, and Pol V in plants (He et al., 2009;Movahedi et al., 2015;Xie and Yu, 2015). RDM4 was shown with affinity with the largest Pol-IV subunit (NUCLEAR RNA POLYMERASE D1, NRPD1), RNA-DEPENDENT RNA POLYMERASE 2 (RDR2), CLASSY 1 (CLSY1), and SAWADEE HOMEODOMAIN HOMOLOG 1 (SHH1) (Law et al., 2011), which did not have significantly changed mRNA expression in rdm4 in this study.
In addition, the deletion in rdm4 mutant covers two genes, including AMP-binding-like protein (Zm00001d023238), which is also annotated as long chain acyl-CoA synthetase 6 (LACS6). LACS is a gene family with redundant roles involved in fatty acid metabolism, and multiple family members have been found, such as six in rodent and human (Teodoro et al., 2017), 34 in Brassica napus (Xiao et al., 2019), nine in Arabidopsis (Zhao et al., 2019), and nine in maize based on the annotation v4. The additional alleles and complementation tests showed that RDM4 is responsible for the seed, vegetative growth and male sterility phenotypes, although the loss of AMP-bindinglike protein might also contribute to differentially abundant mRNAs and proteins identified in RNA-seq and LC-MS/MS experiments, respectively. Further studies are needed to address the possible effects of the loss of Zm00001d023238.
The identification and molecular characterization of rdm4 has highlighted a potential overarching role for RDM4 in normal seed, vegetative and reproductive development as well as in stress responses in maize. Zein and non-zein proteins in kernels were significantly influenced by a loss of RDM4, which resulted in proteomic rebalancing analogous to but substantially different from that observed in o2. Global transcriptional changes were shown in endosperm and leaf, including many transcription factors. A radically altered transcriptome in the rdm4 mutant in response to cold may suggest an overarching function for RDM4 in abiotic stress tolerance. Studies to characterize possible roles for RDM4 in regulating the seed proteome and the transcriptional response to cold in terms of protein and nucleic acid interaction with RDM4, and whether or not DNA methylation is involved, are needed.
Downloaded from https://academic.oup.com/jxb/article/71/19/5880/5872007 by University of Nebraska-Lincoln Libraries user on 06 July 2022
°C) in wild type, and lesser induction by cold in rdm4 mutant. (I, J) The left graphs show that expression of this gene is significantly increased in rdm4 mutant at all temperatures. The right graphs show strong induction of this gene in response to cold in wild type, but a lesser response to cold in rdm4 mutant. Downloaded from https://academic.oup.com/jxb/article/71/19/5880/5872007 by University of Nebraska-Lincoln Libraries user on 06 July 2022