Plant materials and growth conditions

Arabidopsis thaliana Columbia ecotype (Col-0) was used as the wild-type. Mutants atx1-2 (SALK_149002C), atx1-4 (SAIL_409_A10) atx2-2 (SALK_117262), atx3-2 (SAIL_582_H12), atx4-1 (SALK_060156), sdg2-4 (SALK_120450), sdg25-1 (SALK_149692C), snd1 (SALK_015495C), nst1 (SALK_120377C) were ordered from the Nottingham Arabidopsis Stock Centre. Mutant atx5-2 (SAIL_705_H05) was obtained from Dr Yan He (China Agricultural University) and Dr Zhengyi Xu (Northeast Normal University). The nst1 snd1 double mutant was produced by genetic hybridization of nst1 with snd1. Arabidopsis seeds were sterilized in 75% (v/v) ethanol for 30 s, 10% (v/v) sodium hypochlorite solution for 10 min, and then washed by sterilized water for several times, finally dispersed on Murashige and Skoog (MS) medium. After exposure to white light (80 μm⁻² s⁻¹) for 10 days and long-day (LD, 16 h of light, 8 h of dark, 18°C), 75% relative humidity condition in a chamber, the seedlings were transferred into soil and grew in the greenhouse in a long day condition (16 h of light, 8 h of dark, 22°C).

Gene cloning, plasmid construction and generation of transgenic plants

To construct the vectors of ATX1pro:Myc-ATX1, ATX1pro:NST1 and ATX1pro:SND1, 2400 bp upstream sequence of ATX1 was amplified from the genomic DNA of A. thaliana by PCR. The coding sequences of ATX1, NST1 and SND1 were amplified from cDNA obtained by reverse transcription of RNA of A. thaliana, respectively. The ATX1 promoter was fused with the coding sequence of ATX1 with Myc label, NST1 and SND1, respectively, and cloned into a modified pCAMBIA-1305.1 vector (GenBank: AF354045.1), in which hygromycin resistance gene has been replaced with kanamycin resistance genes, and the GFP reporter gene was removed.

To generate the promoter-GUS reporter constructs, the promoters of ATX1, ATX2 (−2091 to +1), ATX3 (−2280 to +1), ATX4 (−2165 to +1), ATX5 (−2150 to +1), SDG2 (−2973 to +1) and SDG25 (−2854 to +1) were amplified by PCR and fused with the GUS reporter into pCAMBIA-1305.1, respectively. To construct 35S:Flag-ATX1, the coding sequence of ATX1 was amplified and cloned into a pCXSN-Flag vector (34).

All plant binary constructs were transformed into Agrobacterium tumefaciens cells (strain GV3101). The ATX1pro:ATX1, ATX1pro:Myc-ATX1, ATX1pro:NST1, and ATX1pro:SND1 constructs were stably introduced into the atx1-2 mutant using the floral dip method (35), respectively. Similarly, the 35S:Flag-ATX1 construct was transformed into the wild-type and snd1 nst1, and the ATX1pro:GUS, ATX2pro:GUS, ATX3pro:GUS, ATX4pro:GUS, ATX5pro:GUS, SDG2pro:GUS and SDG25pro:GUS were transformed into wild-type plants described above.

RNA extraction and RT-qPCR

Total RNA was extracted from stems using Biospin Plant Total RNA Extraction Kit (Bioflux, China). Single-strand complementary DNA (cDNA) was synthesized using Primescript RT reagent kit with gDNA Eraser (Takara, Dalian, China). Quantitative real time-PCR (RT-qPCR) reactions were performed with gene-specific primers and the SYBR Premix Ex Taq (Takara, Dalian, China) on a TP800 Real-Time PCR System (Takara, Japan). Primer sequences for RT-qPCR analysis are listed in Supplementary Data Set 8.

Microscopy

For histochemical staining, the basal internode of Arabidopsis inflorescence stems grown to 20 cm was harvested. The stems were fixed with formalin-acetic acid-alcohol (FAA) and embedded in paraffin after dehydration with gradient ethanol. Semi-thin sections (15 μm) were cut using Ultra-Thin Semiautomatic Microtome (FINESSE 325, Thermo). These sections were dewaxed with xylene and rehydrated with gradient ethanol and then observed with OLYMPUS DP73 microscope after staining with 0.1% toluidine blue.

For transmission electron microscopy, the basal stem of Arabidopsis grown to 20 cm was fixed with glutaraldehyde and embedded in spon-812 resin (36). Embedded plant materials were sectioned into 90 nm-thick sections and stained by uranium acetate and lead citrate, then observed with Talos F200X transmission electron microscope. The images were analyzed using ImageJ software (National Institutes of Health, USA; https://imagej.nih.gov/ij/) for quantifying morphological parameters of cells.

Chromatin immunoprecipitation and library preparation

The stem samples were collected at different developing stages from Arabidopsis Col-0, atx1-2 and ATX1pro:Myc-ATX1/atx1-2 plants, respectively. The harvested stems were cross-linked with 1% formaldehyde in a vacuum and ground into powder in liquid nitrogen before Chromatin immunoprecipitation (ChIP). ChIP was performed as described previously (37,38) with three biological replicates. Chromatin was fragmented to ∼500 bp by ultrasound and immunoprecipitated using H3K4me3 specific antibody (Millipore, 04-745), Flag (Sigma, A8592) and Myc (Sigma, M4439), respectively. The immunoprecipitated DNA (5 ng) was used for the ChIP-seq DNA library preparation or RT-qPCR analysis. The immunoprecipitated chromatin was quantified by RT-qPCR using the specific primers given in Supplementary Data Set 8 and calculated to the relative enrichment compared with input. A ChIP-seq library was constructed by Novogene Corporation (Beijing, China).

Transcriptome sequencing

Transcriptome library was constructed by Beijing Genomics institute (Shenzhen, China). DNase I was used to digest double-stranded and single-stranded DNA in total RNA. Magnetic beads were used to purify and recover the reaction products. RNase H was used to remove the rRNA. Purified mRNA was fragmented into small pieces with fragment buffer. cDNA was synthesized using random hexamer primers and M‐MuLV Reverse Transcriptase (RNase H), and then ligated with adaptors. Then PCR was performed with Phusion High-Fidelity DNA polymerase. The libraries were sequenced on a BGISEQ500 platform (BGI-Shenzhen, China), and paired-end reads were generated.

The sequencing reads were filtered using trimmomatic and aligned to Arabidopsis genome (TAIR10) using HISAT (Hierarchical Indexing for Spliced Alignment of Transcripts) (39) and Bowtie2 (40) with default parameters. Gene expression was defined by RSEM (41) and estimated by FPKM. Differential expression analysis between two conditions/groups was performed using the DESeq R package (1.10.1) (42). The resulting P values were adjusted using the Benjamini and Hochberg's approach for controlling the false discovery rate (42). Genes with expression difference more than 1.5 times, P value < 0.01, and FDR < 0.01 were considered as differentially expressed genes. To analyze gene expression patterns during the stem development, their mRNA levels (FPKM) in stage I, II and III were used as the row data. Then, log₂ was taken for the ratio on expression levels of genes between stages II or III and I. Then the calculated data were analyzed using Short Time-series Expression Miner (STEM) with default parameters.

ChIP-seq

ChIP-seq reads were aligned to the A. thaliana genome (TAIR 10) using BWA (Burrows Wheeler Aligner) (43) with default parameters and only uniquely mapped reads were kept. MACS2 (version 2.1.0) peak calling software was used to identify regions of enriched intervals over the background. A q-value threshold of 0.05 was used for all data sets. Different peak analysis was based on the fold enrichment of peaks in different experiments. When the odds ratio between two groups was more than 1.2-fold, this peak was determined as a different peak. To determine genes with co-regulation of transcription and H3K4me3 modification during the stem development, the row data of fold changes in their mRNA (FPKMFC) and H3K4me3 levels (FoldEnrichFC) were normalized by quantiles to make two distributions statistically comparable. Hierarchical clustering figure was generated by Python data visualization library seaborn (http://seaborn.pydata.org/index.html).

GO enrichment analysis

GO enrichment analysis was performed using agriGO (44) with Fisher's exact test. P values, R factors and gene numbers were visualized as Bubble Chart using the OmicShare tools (www.omicshare.com/tools).

In situ hybridization

The basal stem of 6-week-old wild-type plants were harvested and fixed with FAA and embedded in paraffin after dehydration with gradient ethanol. Cross-sections of 15 μm thickness using Ultra-Thin Semiautomatic Microtome (FINESSE 325, Thermo) were used for hybridization with both the antisense and sense probes for ATX1 transcript. A 369-bp cDNA fragment of ATX1 transcripts was amplified using primers listed in Supplementary Data Set 8, and then cloned into pGEM-T Easy vector (Promega). The probes were synthesized and labelled using a DIG RNA Labeling Kit (Roche) following the vendor's recommendations. Pretreatment of stem sections, hybridization and immunological detection were performed as described previously (45). Hybridization was carried out overnight at 50°C.

Western blot

To determine the Myc-ATX1 protein concentrations, 0.1 g of stem samples at different developmental stages for wild-type, atx1-2 and ATX1pro:Myc-ATX1 plants were analyzed by Western blot using a Myc-specific antibody (Sigma, M4439). The Tublin protein detected by anti-alpha-Tublin (Sigma, T5168) was used as an internal reference. ImageJ software was used to quantify the inimmunolabeled signals.

Statistical analysis

Data were analyzed by one-way ANOVA with Tukey's honestly significant difference test or Student's t-test.

Accession Numbers

The accession numbers of genes used in this study are as follows: ATX1 (AT2G31650), ATX2 (AT1G05830), ATX3 (AT3G61740), ATX4 (AT4G27910), ATX5 (AT5G53430), SDG2 (AT4G15180), SDG25 (AT5G42400), NST1 (At2g46770), SND1 (At1g32770), WRKY70 (AT3G56400), LTP (AT2G15050) and ACTIN7 (AT5G09810).

H3K4me3 methylation dynamically regulates secondary wall biosynthesis

In order to observe the secondary cell wall formation in inflorescence stem of Arabidopsis thaliana, we collected sequentially developing stem samples. Microscopic observations of stem sections from the apex to the base of the inflorescence stem showed that secondary cell wall in interfascicular fibers are fully developed in stage IV (Figure 1A). To unravel the transcriptional level of genes involved in secondary wall biosynthesis during inflorescence stem development, we selected the stem samples from stages I to III for transcriptome analysis. All differentially expressed genes by RNA-seq were clustered into eight categories according to their expression profiles at the different stages by using Short Time-series Expression Miner software (STEM) (Figure 1B, Supplementary Data Set 2). As shown in Figure 1B, 2475 genes were classified into category 7, which represents genes up-regulated in the early stages (II versus I) during the inflorescence stem development, whereas only 754 genes (category 5) were activated in the later stages (III versus II), and 963 genes (category 8) were continuously increased from stages I to III.

Figure 1.

Global analysis of dynamic changes of H3K4me3 and mRNA levels during the inflorescence stem development of Arabidopsis. (A) The inflorescence stem of 6-week-old Arabidopsis thaliana were dived into four fragments from the apex to bottom, i.e. 0–3 cm (stage I, no interfascicular fibers), 5–7 cm (stage II, fiber initiation), 10–12 cm (stage III, developing interfascicular fibers), 18–20 cm (stage IV, mature interfascicular fibers with thickened cell walls). if: interfascicular fiber; co: cortex; xf: xylary fiber; ve: vessel; ph: phloem. The scale bar of the whole plant is 1 cm and the scale bars of the sections are 40 μm. (B) Heatmaps of gene expression profiles in the wild-type stem (stage I, II and III) was performed through STEM (83) on the OmicShare tools platform (http://www.omicshare.com/tools) based the RNA-seq data. These data from stage II to III were normalized by taking log₂ in comparison to stage I. The minimum variation of gene expression between two developmental stages was 2-fold (P< 0.05). (C) Hierarchical clustering of RNA-seq and H3K4me3 ChIP-seq of the wild-type stem (stage II versus I). The column vectors were colored by the rankings of normalized fold change of genes. The row vectors which represent genes were divided into 30 clusters (1–30, from top to bottom) by centroid method. Clusters are highlighted by an additional row color column on the left of the heatmap. The red rectangle box represents cluster 7 that conducted to GO analysis. (D) Gene ontology (GO) enrichment analysis for genes of cluster 7. The x-axis showed the enrichment degree of each GO term. The color gradation indicates the P-value of each GO term. The top three enriched GO pathways were GO:0009834 (R factor = 0.34) for secondary cell wall biogenesis, GO:0045491 (R factor = 0.29) for xlyan metabolic process and GO:0009809 (R factor = 0.26) for lignin biosynthetic process.

We further determined the expression patterns of 74 genes associated with secondary wall formation (1,46,47) during the inflorescence stem development in the RNA-seq data. These genes are mainly enriched in category 7, followed by in category 8, and activated at the early stage of the stem development (Supplementary Figure S1, Supplementary Table S2). Among them, 57 members, including NST1, SND1, MYB46 and MYB83, are clustered into category 7 (Supplementary Figure S2). In contrast, some genes involved in cambium cell proliferation, including KANADI3 (KAN3), CLAVATA3/ESR-RELATED44 (CLE44) and REDUCED IN LATERAL GROWTH1 (RUL1) (48–50), are activated at the later stem development stage and clustered into category 5 (Supplementary Figure S2). These results suggest that these genes controlling secondary cell wall biosynthesis in interfascicular fibers are transcriptionally activated at the early stage of the stem development.

Increasing evidence has shown that transcriptional activation directed by H3K4 tri-methylation is very important for multiple developmental processes and environmental responses in plants (51,52). To reveal the possible regulatory role of H3K4me3 methylation in the inflorescence stem development of Arabidopsis, ChIP-seq was performed by using the stem fragments of stages I and II (Supplementary Data Set 3). Compared with the RNA-seq data obtained above, the Venn diagram revealed a high correlation between the up-regulated genes and the set of H3K4me3 hypermethylated genes at stage II versus I (Supplementary Figure S3, Supplementary Data Set 4). To identify genes with the same transcriptional patterns and H3K4me3 level changes during the stem development, we ranked them in an ascending order by their normalized fold changes in mRNA or H3K4me3, respectively. Next, we used the gene rankings in the hierarchical clustering to reveal the global co-regulation patterns of mRNA- and H3K4me3-fold changes. Out of 30 clusters, cluster 7 represents the genes with a most closely correlation between up-regulated transcription and hypermethylated at H3K4me3 (Figure 1C).

To further investigate the functions of these H3K4 hypermethylated genes, which are up-regulated from stages I to II, we performed a Gene Ontology (GO) enrichment analysis for cluster 7 genes and found that the top three enriched GO pathways were involved in secondary wall biogenesis, xlyan metabolic and lignin biosynthetic process (Figure 1D). In these GO pathways, we identified that both the transcription and H3K4me3 levels of two NAC master switch genes, NST1 and SND1, involved in fiber secondary wall thickening (7,8) and other transcriptional regulators, including SND2, MYB20, MYB63 and KNAT7 (1,6,53), were significantly increased in stage II versus I (Supplementary Figure S4A). By contrast, VND6 and 7, which are key regulators for vessel secondary wall formation (11,54,55), were not marked by H3K4me3 in either stage I or II (Supplementary Figure S4B). These results suggest that the dynamic increase of tri-methylation on H3K4 is involved in the specific activation of secondary wall biosynthetic genes during the fiber differentiation stage of the inflorescence stem.

Spatial and temporal expression pattern of ATX1 in interfascicular fibers

It has been reported that seven histone lysine methyltransferases, including ATX1, 2, 3, 4, 5, SDG2 and SDG25, are responsible for the establishment of methyl-H3K4 in Arabidopsis (56,57). To determine which member of these histone methyltransferases is involved in the regulation of secondary wall biosynthesis, spatial and temporal expression patterns of these H3K4 methyltransferase genes were analyzed during the inflorescence stem development. We firstly examined the dynamic expression levels of these genes from stage I to IV by RT-qPCR analysis. All histone lysine methyltransferase genes except for SDG2 were up-regulated in stage II, compared with that in stage I (Figure 2A and B, Supplementary Figure S5A). We next generated transgenic plants harboring all seven H3K4 methyltransferase gene promoters fused to the β-glucuronidase (GUS) reporter gene, respectively. GUS staining showed that ATX2, ATX3 and ATX4 were mainly expressed in the cortical and phloem cells at the developing inflorescence stem, while the expression of SDG2 and SDG25 was restricted in the cortical cells (Supplementary Figure S5B). By contrast, ATX1pro:GUS showed strong expression in the cortical cells at the apex of the stem (stage I), and specific expression in interfascicular fibers (from stage II to IV) rather than in vessel tissues (Figure 2C). To further investigate the expression profiles of ATX1, in situ mRNA hybridization also verified that, in the developing inflorescence stem, ATX1 mRNAs were specifically accumulated in interfascicular fiber cells (Figure 2D). These results indicate that the expression of ATX1 is developmentally associated with secondary wall biosynthesis in the interfascicular fibers.

Figure 2.

The spatial and temporal expression patterns of ATX1 in the developing stem of Arabidopsis. (A) Genome tracks of RNA-seq for ATX1 in the developing stem (from stage I to III) of wild-type plants. (B) RT-qPCR analysis of ATX1 expression in the inflorescence stem from stage I to IV. Relative expression was normalized to that of ACTIN7 (ACT7). Different letters indicate significant differences among different developmental stages of the stem based on one-way ANOVA (P < 0.05). (C) GUS staining of ATX1pro:GUS transgenic plants were used to analyze the spatial and temporal expression pattern of ATX1 in inflorescence stems from stage I to IV. Three independent lines of ATX1pro:GUS were analyzed by staining here and all of them showed consistent pattern of GUS signal. (D) In situ localization of ATX1 in inflorescence stems of wild-type plants. The stem of stage III was cross-sectioned for hybridization with antisense and sense probes of ATX1. Scale bars = 40 μm.

ATX1 positively regulates the secondary wall formation in interfascicular fibers

To determine the regulatory roles of these H3K4-specific methyltransferases in the secondary wall formation during Arabidopsis inflorescence stem development, the mutants of atx1, 2, 3, 4, 5, sdg2 and sdg25 were grown and their phenotypes were analyzed. Compared with the wild-type, two independent loss-of-function mutant alleles of ATX1 exhibited lodging phenotypes (Figure 3A). Cross sections of the basal stems revealed that the wall thickness of interfascicular fibers in the atx1-2 and atx1-4 mutants were reduced to 64.93% and 67.20% of that in the wild-type, respectively, whereas the cell wall thickness of the xylem fibers and vessels were not influenced by ATX1 mutation (Figure 3B and C). To further confirm the effect of ATX1 on secondary wall deposition of interfascicular fibers, the ATX1 genomic DNA driven by its native promoter (−2400 to +6931 bp) was introduced into the background of the atx1-2 mutant (COM-1) (Supplementary Figure S6). The lodging phenotype and thinner cell walls were rescued by complementation with the ATX1 gene (Figure 3A−C, Supplementary Figure S7). Overexpression of ATX1 under the control of the CaMV 35S promoter resulted in 59.75% and 30.59% increase in secondary wall thickness in interfascicular fibers and xylem fibers, compared with the wild-type control, respectively, but had no obvious effects on vessel cells (Figure 3B and C, Supplementary Figure S7). These results indicate a positive correlation between ATX1 expression and the cell wall thickness in interfascicular fibers within the inflorescence stem of transgenic plants. On the contrary, the mutants of atx2, 3, 4, 5, sdg2 and sdg25 did not display lodging stem or deficient in secondary wall deposition in interfascicular fibers (Supplementary Figure S8). These results indicate that ATX1 is required for the secondary cell wall formation in the interfascicular fiber, consistent with its restricted expression patterns in this tissue.

Figure 3.

ATX1 promotes the secondary cell wall synthesis in interfascicular fiber cells. (A) Phenotypes of 6-week-old wild-type, atx1-2 and atx1-4 mutants, COM-1 (complementation) and 35S:ATX1 transgenic plants. (B) Cross sections of the basal stem of wild-type, atx1-2, atx1-4, COM-1 and 35S:ATX1-1 transgenic plants. The interfascicular regions (left) and vascular bundles (right) were observed under the bright field (BF) after stained with toluidine blue or transmission electron microscopy (TEM), respectively. (C) Statistics of secondary cell wall (CW) thickness of fibers and vessels in the inflorescence stem of wild-type, atx1-2, atx1-4, COM-1 and 35S:ATX1-1 transgenic plants. Error bars represent ± SD (n = 15). (D−F) Relative expression of lignin (D), cellulose (E) and xylan (F) biosynthetic genes in the stem of wild-type, atx1-2, atx1-4, COM-1 and 35S:ATX1-1 transgenic plants. Relative expression was normalized to that of ACT7 and set to 1 in the wild-type. Asterisks indicate a significant difference using the Student's t-test (*P < 0.05; **P < 0.01). co: cortex; if: interfascicular fiber; ph: phloem; ve: vessel; xf: xylem fiber. Scale bars = 3 cm in (A), 40 μm in photo taken under BF, 2 μm in photo taken under TEM.

Further phloroglucinol-HCl and Calcofluor White staining and xylan antibody immunolocalization revealed that the deposition of lignin, cellulose and xylan was reduced in the interfascicular fiber cell walls of the atx1-2 and atx1-4 mutants, but increased in 35S:ATX1 transgenic plants compared with the wild-type control (Supplementary Figure S9). Quantitative analysis of secondary wall components also showed that the contents of lignin, cellulose and hemicellulose in interfascicular fibers were decreased by 7.5%, 6.8% and 16.5% in atx1-2, respectively, compared with the wild-type, while increased by 31.4%, 3.5% and 6.6% in transgenic 35S:ATX1 plants, respectively (Supplementary Table S1). Furthermore, the expression of lignin biosynthetic genes (PAL1, CCoAOMT1 and (CAD6) (58), xylan biosynthetic genes (IRX8, IRX9 and IRX14) (59–62), and cellulose biosynthetic genes (CesA4, CesA7 and CesA8) (63–65), were down-regulated in atx1-2 and atx1-4 mutants, but up-regulated in 35S:ATX1 transgenic plants in comparison to the wild-type (Figure 3D−F). These results indicate that ATX1 could promote the secondary cell wall formation in interfascicular fibers by positively regulating the biosynthesis of the main secondary cell wall components.

ATX1 directly associates with the chromatin of NST1 and SND1 to modulate their H3K4me3 and transcript levels

In order to determine ATX1-targeted genes that are involved in the regulation of the secondary cell wall thickening in interfascicular fiber cells at the genome-wide scale, we performed a transcriptome analysis on the developing inflorescence stem (stage II) from 6-week-old wild-type and atx1-2 plants. A total of 3221 genes were down-regulated (|fold change| ≥ 1.5, P < 0.01, FDR<0.01) in atx1-2, compared with the wild-type (Figure 4A, Supplementary Data Set 5). To investigate whether these down-regulated genes in atx1-2 are directly associated targets of ATX1, we carried out an H3K4me3 chromatin immunoprecipitation followed by sequencing (ChIP-seq). In the atx1-2 mutant, H3K4me3 levels were reduced for 554 genes compared with wild-type plants (Figure 4A, Supplementary Data Set 6). The down-regulation and H3K4me3 hypomethylation were overlapped on 107 genes (Figure 4A, Supplementary Data Set 7), including the target genes of ATX1 reported previously, such as WRKY70 and LTP (66), suggesting that ATX1 is directly associated with these 107 genes and positively regulate H3K4me3 levels at these loci in wild-type plants. Among these ATX1-assoicated genes, 26 were also transcriptionally up-regulated and hypermethylated at the early stage (II versus I) of the inflorescence stem development in wild-type plants (Figure 4B, Supplementary Table S3). Interestingly, among these 26 genes, we found two genes encoding known NAC master switches for secondary wall biosynthesis, NST1 and SND1 (7,8) (Figure 4C, Supplementary Figure S10A). Meanwhile, 57 of ATX1-associated genes, including WRKY70, were not up-regulated during the stem development (Supplementary Figure S10C). In addition, some secondary wall regulators, such as SND2, MYB20, MYB63 and KNAT7 (6,53,67,68), were shown to be up-regulated and hyper-H3K4 tri-methylated during the stem development (stage II versus I) in the wild-type, but the elevations in transcription and methylation levels of these genes were not abolished in atx1-2 (Supplementary Figure S10B), suggesting that these genes were not directly regulated by ATX1.

Figure 4.

The transcription and H3k4me3 methylation levels of NST1 and SND1 are up-regulated by ATX1. (A) Venn diagram showing the overlap among genes that are down-regulated in atx1-2 by RNA-seq (|foldchange| > 1.5, P < 0.01) and H3K4me3 hypomethylated genes in atx1-2 compared with the wild-type by ChIP-seq (|foldchange| > 1.2, q-value < 0.05). (B) Venn diagram showing the overlaps among the overlapped genes in (A), the up-regulated genes in wild-type plants (stage II versus I, |foldchange| > 1.5, P < 0.01), and the H3K4me3 hypermethylated genes in wild-type plants (stage II versus I). (C) Genome tracks of RNA-seq and anti-H3K4me3 ChIP-seq data for the loci of NST1 and SND1 in the different stem developmental stages (I and II) of the wild-type and atx1 mutant. (D and E) Diagrams for their gene structures of NST1 and SND1, and the regions of NST1 (R1: −529 ∼ −426 bp; R2: +12 ∼ +95 bp; R3: +514 ∼ +671 bp; R4: +1792 ∼ +1920 bp) and SND1 (R1: -564 ∼ −343 bp; R2: +33 ∼ +156 bp; R3: +264 ∼ +400 bp; R4: +1324 ∼ +1459 bp) were used for ChIP-qPCR analysis, respectively. (F and G) RT-qPCR analysis of NST1 (F) and SND1 (G) expression levels in wild-type, atx1-2, COM-1 and 35S:ATX1-1. Relative expression was normalized to that of ACT7 and set to 1 in wild-type plants. (H and I) ChIP-qPCR analysis of H3K4me3 methylation status at the loci of NST1 (H) and SND1 (I) in wild-type, atx1-2 and 35S:ATX1-1 plants. (J and K) ChIP-qPCR analysis of the association of ATX1 with the loci of NST1 (J) and SND1 (K) using Myc-ATX1 (ATX1pro:Myc-ATX1/atx1-2) and 35S:ATX1-1 (35S:Flag-ATX1) hybridized with Myc or Flag antibody, respectively. Wild-type plants were used as a control. Different letters in H, I, J and K indicate significant differences among genotypes and regions of chromatin based on one-way ANOVA (P < 0.05). Asterisks indicate a significant difference with respect to wild-type for mutant or transgenic plants using Student's t-test (*P < 0.05; **P < 0.01).

Gene expression analysis revealed that the transcript levels of NST1 and SND1 were remarkably down-regulated in the developing stem of atx1-2 and restored to near wild-type levels in the genetic complementation line (Figure 4F and G). Whereas the expression of NST1 and SND1 were increased by 11- and 4-fold in transgenic 35S:ATX1-1 plants, respectively (Figure 4F and G). Several pairs of primers covering approximately 200 base-pair (bp) regions of the NST1 and SND1 loci were designed for ChIP-qPCR (Figure 4D and E). We found that H3K4me3 levels at the R2 region nearby the transcription start site and the R3 region in the second exon of NST1 were distinctly decreased in atx1-2, but increased in transgenic 35S:ATX1-1 plants, compared with the wild-type, while these at the promoter region (R1) and 3′ end (R4) of the gene body were not changed (Figure 4H). Similar results were obtained in the locus of SND1 (Figure 4I). To further verify the direct binding of ATX1 to the loci of NST1 and SND1, we generated transgenic plants harboring ATX1pro:Myc-ATX1 under the background of atx1-2 or 35S:Flag-ATX1 under the background of the wild-type Columbia (Col), respectively. ChIP-qPCR analysis revealed the specific association of the Myc- and Flag-ATX1 proteins with the NST1 and SND1 loci at the R2 and R3 regions (Figure 4J and K). These data indicate that ATX1 could directly bind at the loci of NST1 and SND1, thus promoting their transcription through the regulation of H3K4me3 levels on their chromatin.

ATX1 activates the expression of NST1 and SND1 by positively regulates their H3K4me3 levels during the stem development

Although NST1 and SND1 serves as master regulators for secondary wall thickening in fibers (10), how their expression is regulated during Arabidopsis inflorescence stem development remains largely unknown. To elucidate the role of ATX1 in regulating chromatin modifications of NST1 and SND1 during the stem development, we investigated the transcript accumulation and H3K4me3 levels of NST1 and SND1 during various developmental stages. Consistent with the expression pattern of ATX1 obtained above (Figure 2A), expression levels of NST1 and SND1, were barely detectable at stage I, and significantly increased in wild-type plants from stage I to III, but not distinctly changed in atx1-2 (Figure 5A and B). The expression levels of NST1 and SND1 in the late developmental stem (stage III) of atx1-2 were only approximately 0.21- and 0.31-fold of these in wild-type plants, respectively (Figure 5A and B). As a control, WRKY70, which acts as an activator of salicylic acid (SA)-induced genes and has been identified to be directly bound by ATX1 (66), were not up-regulated during the stem development in either the wild-type or atx1-2 mutant (Figure 5C). Next, we monitored the H3K4me3 levels of the R3 regions in the NST1 and SND1 loci at three developmental stages. In the apex (stage I) of the inflorescence stem of 6-week-old wild-type and atx1-2 plants, basal levels of H3K4me3 were detected at these loci (Figure 5A and B), indicating that ATX1 does not influence H3K4me3 modification of NST1 and SND1 at this stage. Compared with these at stage I, H3K4me3 level in the R3 regions of the NST1 and SND1 loci was significantly increased at stages II and III during the inflorescence stem development in wild-type plants, but not altered in atx1-2 (Figure 5A and B). Our results show that the increase in H3K4me3 and transcript levels of NST1 and SND1 during the inflorescence stem development largely depends on the presence of ATX1.

Figure 5.

ATX1 dynamically regulates NST1 and SND1 expression during the stem development. (A-C) Expression and H3K4me3 levels of NST1 (A), SND1 (B), and WRKY70 (C) in wild-type and atx1-2 at the developmental stem (from stage I to III). The red lines and left Y-axis represent transcription levels. The blue lines and right Y-axis represent H3K4me3 modification. Relative expression was normalized to ACT7 and set to 1 in stage I of wild-type plants. (D) RT-qPCR analysis of transcription levels of ATX1 in wild-type and two of the ATX1pro:Myc-ATX1/atx1-2 transgenic lines at the developmental stem (stage I to III). Relative expression was normalized to ACT7 and set to 1 in wild-type plants. The different letters indicate significant differences among different genotypes based on one-way ANOVA (P < 0.05). (E) Myc-ATX1 protein levels in two of the ATX1pro:Myc-ATX1/atx1-2 transgenic lines (Myc-ATX1-1/2) at the indicated developmental stages of stems. Total protein extracts isolated from wild-type or atx1-2 plants were used as negative controls. Anti- alpha -tublin (Sigma, T5168) was used as a loading control. Anti-Myc antibody (Sigma, M4439) was used to detect Myc-ATX1. (F−H) ChIP-qPCR analysis using anti-Myc antibody to detect the direct association of ATX1 on the loci of NST1 (F), SND1 (G) and WRKY70 (H) at the indicated developmental stages of stems. The primers of the R3 regions of NST1 and SND1, and primer pair flanking the site of translation start site of WRKY70, were used for ChIP-qPCR analysis. Asterisks indicate a significant difference with respect to wild-type plants for each mutant or transgenic plants using Student's t-test (*P < 0.05; **P < 0.01; n.d., no significant different).

In order to further elucidate how ATX1 dynamically regulates their chromatin modifications during interfascicular fiber development, we determined the transcript level, protein abundance and chromatin association of ATX1 with NST1 and SND1 at different developmental stages of the inflorescence stem. ATX1 transcript levels were low in the stage-I stem of ATX1pro:Myc-ATX1 transgenic plants, in which the fused Myc-ATX1 gene driven by the ATX1 native promoter was introduced into the atx1-2 background, but substantially increased at stages II and III (Figure 5D), consistent with the developmental increase of endogenous ATX1 transcripts in the wild-type stem (Figure 2A). We also examined ATX1 protein levels in ATX1pro:Myc-ATX1 transgenic plants using anti-Myc antibody. Immunoblot analysis showed that Myc-ATX1 protein levels were markedly higher in the stage-II and -III stem fragments compared with the stage-I one (Figure 5E). With the increase of ATX1 protein level, the Myc-ATX1 protein levels associated with the NST1 and SND1 loci were also significantly higher at stages II and III than at stage I (Figure 5F and G). In contrast to NST1 and SND1, WRKY70 was not dynamically regulated by ATX1 during the stem developmental process (Figure 5C and H). These results indicate that ATX1 plays an important role in the establishment of H3K4me3 marks and transcriptional reprogramming of NST1 and SND1 for secondary wall biosynthesis of interfascicular fibers during the stem development through the dynamic association and methylation of their chromatins.

ATX1 promotes secondary cell wall biosynthesis in interfascicular fibers through activating the expression of NST1 and SND1

To investigate whether ATX1 regulates secondary cell wall biosynthesis of interfascicular fibers through NST1 and SND1, we introduced the ATX1 gene under the control of the CaMV 35S promoter into the snd1 nst1 double mutant. As shown in Figure 6A−D, overexpression of ATX1 failed to restore the lodging phenotype and the defect in secondary wall thickening in fibers of snd1 nst1. The transcript levels of ATX1 were significantly increased in transgenic plants carrying 35S:ATX1 under the background of the wild-type or atx1-2, compared with the corresponding nontransformed controls (Figure 6E, Supplementary Figure S11). Knockout of NST1 and SND1 also blocked the up-regulation of marker genes for secondary wall biosynthesis triggered by 35S:ATX1 (Supplementary Figure S13A−C).

Figure 6.

ATX1 regulates secondary wall biosynthesis in interfascicular fiber cells through promoting the expression of NST1 and SND1. (A) Phenotypes of wild-type, 35S:ATX1, 35S:ATX1/snd1 nst1, snd1 nst1. (B) Cross sections of interfascicular regions of wild-type, 35S:ATX1-1, 35S:ATX1/snd nst1 and snd1 nst1. (C) Transmission electron micrographs of interfascicular fiber cells of wild-type, 35S:ATX1-1, 35S:ATX1/snd1 nst1 and snd1 nst1. (D) Statistical analysis on secondary cell wall thickness of interfascicular fiber cells of wild-type, 35S:ATX1-1, 35S:ATX1/snd1 nst1 and snd1 nst1. (E) Expression levels of ATX1 in wild-type, 35S:ATX1-1, 35S:ATX1/snd1 nst1 and snd1 nst1. (F) Phenotypes of wild-type, atx1-2, ATX1pro:NST1/atx1-2 and ATX1pro:SND1/atx1-2. (G) Cross sections of interfascicular regions of wild-type, atx1-2, ATX1pro:NST1/atx1-2 and ATX1pro:SND1/atx1-2. (H) Transmission electron micrographs of interfascicular fiber cells of wild-type, atx1-2, ATX1pro:NST1/atx1-2 and ATX1pro:SND1/atx1-2. (I) Statistical analysis on secondary cell wall thickness of wild-type, atx1-2, ATX1pro:NST1/atx1-2 and ATX1pro:SND1/atx1-2. (J and K) Expression levels of NST1 (J) and SND1 (K) in wild-type, atx1-2, ATX1pro:NST1/atx1-2 and ATX1pro:SND1/atx1-2, respectively. Relative expression was normalized to ACT7 and set to 1 in wild-type plants. Different letters in D, E, I, J and K indicate significantly differences among different genotypes based on one-way ANOVA (P < 0.05).

To further determine whether NST1 and SND1 were downstream of ATX1 in controlling secondary wall biosynthesis of the fibers, we transformed the NST1 and SND1 genes driven by the ATX1 promoter into the atx1-2 mutant, respectively (Supplementary Figure S12). We found that the introduction of the ATX1pro:NST1 or ATX1pro:SND1 gene could restore the lodging phenotype and the defect in secondary wall thickening in interfascicular fibers caused by ATX1 mutation (Figure 6F-I, Supplementary Figure S12). In addition, the down-regulation of secondary cell wall biosynthetic genes in atx1-2 was also restored by the ectopic expression of NST1 or SND1 (Figure 6J and K, Supplementary Figure S13D−F). These results demonstrate that NST1 and SND1 were located in the downstream of ATX1 in the regulatory network controlling secondary cell wall biosynthesis in interfascicular fibers.