A DNA adenine demethylase impairs PRC2-mediated repression of genes marked by a specific chromatin signature

Background The Fe (II)- and α-ketoglutarate-dependent AlkB family dioxygenases are implicated in nucleotide demethylation. AlkB homolog1 (ALKBH1) is shown to demethylate DNA adenine methylation (6mA) preferentially from single-stranded or unpaired DNA, while its demethylase activity and function in the chromatin context are unclear. Results Here, we find that loss-of-function of the rice ALKBH1 gene leads to increased 6mA in the R-loop regions of the genome but has a limited effect on the overall 6mA level. However, in the context of mixed tissues, rather than on individual loci, the ALKBH1 mutation or overexpression mainly affects the expression of genes with a specific combination of chromatin modifications in the body region marked with H3K4me3 and H3K27me3 but depleted of DNA CG methylation. In the similar context of mixed tissues, further analysis reveals that the ALKBH1 protein preferentially binds to genes marked by the chromatin signature and has a function to maintain a high H3K4me3/H3K27me3 ratio by impairing the binding of Polycomb repressive complex 2 (PRC2) to the targets, which is required for both the basal and stress-induced expression of the genes. Conclusion Our findings unravel a function of ALKBH1 to control the balance between the antagonistic histone methylations for gene activity and provide insight into the regulatory mechanism of PRC2-mediated H3K27me3 deposition within the gene body region. Supplementary Information The online version contains supplementary material available at 10.1186/s13059-023-03042-4.


Background
N6-adenine methylation (6mA) was first characterized as a DNA modification in bacteria and only identified in a few lower eukaryotes for a long time.More recently, 6mA has been identified in the genomic DNA of a wide range of eukaryotic organisms, from basal fungi to animals and plants, and there exist significant variations in genomic distribution, abundance, and functions of DNA 6mA among eukaryotes [1,2].Depending on the organisms, DNA 6mA may function as an epigenetic mark for either transcriptional activation or repression [3][4][5][6][7].DNA 6mA can be induced during development or under stress.For instance, 6mA is upregulated during the development of mouse trophoblast stem cells, specifically at stress-induced DNA double helix destabilization (SIDD) regions [8].In rice plants, 6mA was found to be induced by abiotic stress [9].Deposition of DNA 6mA prevents chromatin binding of SATB1, a crucial chromatin organizer that interacts with SIDD regions, revealing a role of 6mA at the boundaries between euchromatin and heterochromatin to restrict the spread of euchromatin [8].However, in animals such as Drosophila, C. elegans, and mammals [5,[10][11][12][13][14], 6mA levels are orders of magnitude lower than those in unicellular eukaryotes such as Tetrahymena [15] and Chlamydomonas [3].In higher plants (Arabidopsis and rice), DNA 6mA was detected at intermediate levels [4,7,9].Whether low-level 6mA can be accurately detected or even exists in some organisms is currently debated [16][17][18].
Histone H3 lysine 27 tri-methylation (H3K27me3) is a repressive chromatin mark deposited by the Polycomb repressive complex 2 (PRC2).In plants, about 10 to 20% of genes are marked by H3K27me3 [29][30][31][32], which are mainly involved in developmental processes and stress responses in plants [29,33,34].However, changes in H3K27me3 levels in mutant plants affecting enzymes involved in the H3K27me3 dynamics do not generally correlate with the overall gene expression changes [35][36][37], suggesting that additional chromatin events may be involved in H3K27me3-mediated gene repression.By contrast, H3K4me3 is an active chromatin mark associated with a large number of expressed genes.Genes marked by both H3K4me3 and H3K27me3 are so-called bivalent genes, which are in many cases poised for activation by cellular stimuli.But how H3K4me3 and H3K27me3 control the expression of bivalent genes is generally unclear.
In this work, we explored the ALKBH1 genome-wide targets and its function in 6mA demethylation, chromatin modifications, and gene expression in plants.Our study uncovered that the ALKBH1-binding antagonizes PRC2 function to maintain a high H3K4me3/H3K27me3 ratio in its target genes to upregulate their expression.These genes are highly enriched in stress response.The results extend the function of ALKBH1 to controlling histone methylation of genes marked by a specific chromatin signature and provide insight into the regulatory mechanism of PRC2 function and H3K27me3 spreading in the gene body.

Mutation of ALKBH1 had a limited effect on the genome-wide 6mA in rice
The rice ALKBH family has 7 members (Additional file 1: Fig. S1a) [7].The rice ALKBH1 (LOC_Os03g60190) was shown to localize in the nucleus and have an in vitro 6mA demethylase activity toward ssDNA [7].To study the function of rice ALKBH1 in genomewide 6mA, we analyzed the wild type (Nipponbare, NIP) and the alkbh1 CRISPR mutant lines (Additional file 1: Fig. S1b) [7] by using the 6mA DNA immunoprecipitation sequencing (DIP-seq) method [3].From the DIP-seq, we obtained 32,000 to 37,000 peaks (P < 10 −5 ) in the wild-type and mutant lines relative to the respective inputs (Additional file 2: Table S1).However, metaplots and scatter plots of the peaks revealed no clear difference in the 6mA levels in genes between the wild type and the mutant, except for some increases in TE-related genes (TEG) in the mutant (Fig. 1a, b).The limited effect of the alkbh1 mutation on the genome-wide 6mA may be due to a redundant function between ALKBH1 and other members of the family in rice, as a C. elegans homolog of ALKBH4 (NMAD-1) was shown to demethylate 6mA [11,38].Alternatively, the rice ALKBH1 may target only discrete genomic sites for 6mA demethylation which could not be revealed by the antibody-based DIP-seq method.Previous studies showed that in rice 6mA is enriched in R-loops which are mainly detected in the flanking regions of genes [39].Analysis of the R-loop 6mA revealed higher levels (P < 0.0001) in both Watson DNA strand-related R-loops (wR-loops) and Crick DNA strand-related R-loops (cR-loops) [39,40] in the mutants than wild type (Fig. 1c), consistent with the results that mammalian ALKBH1 preferentially targets unpaired DNA for 6mA demethylation [25].In fact, 29% of the 6mA peaks colocalized with R-loops (Fig. 1d).In alkbh1, about 48% (671/1399) of the upregulated peaks localized to R-loops compared with 184/1329 downregulated peaks in R-loops (Fig. 1d, e).The analysis suggested a potential role of ALKBH1 to reduce 6mA from R-loops in the rice genome.

ALKBH1 is required for the expression of genes with a specific chromatin signature
RNA-seq analysis of seedling tissues uncovered 1047 downregulated and 386 upregulated (> 2 fold, q < 0.05) genes in the alkbh1 mutants compared to the wild-type plants (Fig. 2a), suggesting that ALKBH1 functioned as a positive regulator of gene expression.The downregulated genes were enriched mainly for transcriptional activity and stress response (Fig. 2b), suggesting that ALKBH1 may be a higher hierarchical regulator of basal expression of stress-responsive genes.However, there was no clear association between the gene expression change and the 6mA variation in the mutant (Additional file 1: Fig. S1c).Chromatin modification heatmaps indicated that the downregulated genes had much lower levels of 5′-cytosine methylation (5mC) at CG sequences (mCG) in the body region than the randomly selected 1500 genes (with averaged expression levels) in wild-type background (Fig. 2c).In addition, the downregulated genes exhibited higher levels of H3K27me3 and H3K4me3 in the body region than the control genes in wild-type background.In plants, H3K4me3 mainly peaked at the gene transcriptional start sites (TSS), as observed in the control genes, but the mark was found to be spread all over the body region toward the transcription termination site (TTS) in the downregulated genes (Fig. 2c).By contrast, the upregulated genes in the mutants showed no clear Fig. 1 ALKBH1 demethylates 6mA in R-loops in rice.a Metaplots of 6mA levels of all genes in alkbh1 compared with wild type.b Scatterplot of 6mA levels between WT and alkbh1.DIP-seq peaks merged from all samples (n = 42,387) are shown.The x-axis and y-axis represent the normalized read number at each peak of WT and alkbh1, respectively.Blue dots represent the downregulated peaks; red dots represent the upregulated peaks; differential peaks (fold change > 1.5, P < 0.05) overlapping between two replicates were considered as significantly changed.c Metaplots of 6mA levels at R-loops [39] in alkbh1 compared with wild type.d Venn diagram showing the percentages of 6mA DIP-seq peaks that overlap with DRIP-seq peaks [39] (DNA: RNA hybrid immunoprecipitation and sequencing for R-loops).P-values were calculated by hypergeometric tests.e DRIP-seq and 6mA DIP-seq signals at a representative genomic region.Gray areas display the difference in signals between the mutant and WT difference in chromatin modifications from the control genes in wild-type background, except for higher levels of H3K27me3 (Fig. 2c).
To confirm the function of ALKBH1 in gene expression, we produced and analyzed the ALKBH1 overexpression (OE) plants that showed no visible growth phenotype (Additional file 1: Fig. S1d, e).RNA-seq analysis detected 2388 up-and 1114 downregulated genes in the OEALKBH1 plants (Fig. 2d).As the downregulated genes in the mutants, the upregulated genes in the OE plants were also enriched for Fig. 2 ALKBH1 activates genes marked by both H3K27me3 and H3K4me3 but depleted of mCG in the body region.a RNA-seq analysis of differentially expressed genes (DEGs) (> 2-fold; q-value < 0.05) in alkbh1 compared with the wild type (WT).b GO enrichment of the downregulated DEGs in alkbh1.c Metaplots and heatmaps of chromatin modification profiles in the alkbh1 downregulated (n = 1047 blue) and upregulated (n = 386 red) genes and control genes with similar transcription levels (CK) (n = 1500 gray) in the wild-type background.Expression levels of the DEGs in the mutant and the control genes in wild-type plants are shown by boxplots and heatmaps.d RNA-seq analysis of DEGs (> 2-fold; q-value < 0.05) in the overexpression lines (OEALKBH1) compared with the wild type (WT).e GO enrichment of upregulated DEGs in OEALKBH1.f Metaplots and heatmaps of chromatin modification profiles in the OEALKBH1 downregulated (n = 1114 blue) and upregulated (n = 2388 red) genes and control genes (n = 1500 gray) in the wild-type background transcriptional activity and stress response (Fig. 2e) and displayed much higher levels of H3K27me3 and H3K4me3 and lower levels of gene body mCG than the control genes (Fig. 2f ).Actually, 265 downregulated genes in the mutants were upregulated in the OE plants (Additional file 1: Fig. S1f, Additional file 3: Table S2).The data indicated that ALKBH1 was required for the expression of genes enriched for H3K27me3 and/or H3K4me3 but without mCG in the gene body in wild-type rice seedling tissues.
Further analysis of the mean transcript levels of genes showing only H3K27me3 or H3K4me3 or both modifications in wild-type seedling mixed tissues indicated that the alkbh1 mutation and overexpression had more effect on genes marked with H3K27me3 and H3K4me3 than with H3K27me3 or H3K4me3 alone in the context of mixed tissues (Additional file 1: Fig. S1g).
In plants, gene body mCG is associated with gene activity and is detected only in species that express the CHG methyltransferase CMT3 [41,42].A correlation between mCG and 6mA in gene body regions was observed [7].In the rice genome, there were 23,938 genes with (i.e., > 40% methylation) and 7986 genes without (i.e., < 10% methylation) body mCG (Additional file 1: Fig. S2a).The genes without body mCG showed lower 6mA and higher ATAC (open chromatin) signals than those with body mCG (Additional file 1: Fig. S2b).By contrast, these two categories of genes displayed a striking difference in H3K27me3 and H3K4me3: the genes without body mCG showed much higher levels of H3K27me3 and H3K4me3 than those with the body DNA methylation (Additional file 1: Fig. S2b).In addition, the body CG-methylated genes were marked by H3K4me3 only at the TSS, while the body unmethylated genes had H3K4me3 spread from the TSS toward the TTS (Additional file 1: Fig. S2b), reminiscent of the downregulated genes in the alkbh1 mutants and the upregulated genes in the OE plants (Additional file 1: Fig. S2c).This may reflect the antagonistic relationship between H3K4me3 and CG methylation enzymes [43,44].The association of high levels of H3K27me3 and/or H3K4me3 with low levels of DNA methylation in the gene body likely represents a chromatin signature of genes downregulated in alkbh1 mutant and upregulated in OE plants.

ALKBH1 binds to genes marked by the specific chromatin signature
To further study genomic targets of ALKBH1, we produced an antibody of the protein using a peptide as antigen (Additional file 1: Fig. S3a) and performed ChIP-seq analysis of the wild type and, as a control, the alkbh1 plants (Additional file 1: Fig. S3b, Additional file 2: Table S1).We detected 2394 ALKBH1-binding peaks (or 1744 binding genes) in the wild type (Fig. 3a).The ALKBH1 binding located mainly in the gene body with the highest binding level toward TTS (Fig. 3b).Heatmap analysis indicated that the ALKBH1 binding levels mirrored that of H3K4me3, whereas the binding genes showed a more or less uniformed level of H3K27me3 (Fig. 3b).Scatter plots revealed that 33% (324/1047, P-value = 0) of the downregulated genes in the alkbh1 mutants and 15% (349/2388, P-value = 0) of the upregulated genes in the OE plants were ALKBH1binding genes (Fig. 3c, d).In these DEGs, the ALKBH1-binding and H3K4me3 levels were higher and the body mCG was much lower than the remaining binding genes (Fig. 3e).These DEGs also showed higher H3K27me3 levels than the remaining binding genes (Fig. 3e).The downregulated DEGs in the mutants displayed a higher H3K4me3/ H3K27me3 ratio than the control genes in the wild-type background (Fig. 3f ).The analysis confirmed that ALKBH1 targeted genes with the specific combination of chromatin modifications (i.e., high H3K4me3 and H3K27me3 and low body mCG) to stimulate gene expression.In the context of mixed tissues, either a threshold of the ratio between H3K4me3 and H3K27me3 was required for ALKBH1 binding to stimulate gene expression, or the ALKBH1-binding allowed maintaining an H3K4me3/H3K27me3 ratio favorable for gene expression.

ALKBH1 is required for maintaining a high H3K4me3/H3K27me3 ratio in the target genes
To test the above hypotheses, we first analyzed the histone methylation levels in alkbh1 and wild type plants by immunoblots.The level of H3K27me3, but not H3K4me3 nor H3K9me2, was augmented in the mutants but reduced in the OE plant (Fig. 4a, Additional file 1: Fig. S4a, b).The results suggested that ALKBH1 might modulate H3K27me3 to control the H3K4me3/H3K27me3 ratio on the target genes.Next, we performed H3K27me3 ChIP-seq analysis of the wild type and the alkbh1 plants and detected a total of 11,473 peaks (or 8729 genes), representing about 14% of the rice genes (Fig. 4b, Additional file 2: Table S1), as found in other studied [29,35,37].In the mutant, 1542 peaks (or 1058 genes) showed an increase but only 324 peaks (264 genes) showed a decrease (fold change > 1.5, P-value < 0.05) of H3K27me3 (Fig. 4b), confirming the immunoblot results.Among the 1058 genes, 225 were ALKBH1-binding genes (Additional file 4: Table S3), in which the increases of H3K27me3 in the mutant were more important than remaining 833 genes (Fig. 4c, d) and showed lower expression in the mutants but higher expression in the OE lines than wild type (Fig. 4e), suggesting that ALKBH1-binding reduced H3K27me3 and stimulated gene expression.Interestingly, the 225 ALKBH1-binding genes displayed much higher levels of H3K4me3 than the remaining 833 genes in the wild-type background, while the total 1058 genes with increased H3K27me3 in the mutant had similar H3K27me3 levels in a wild-type background (Fig. 4f ).This supported the hypothesis that ALKBH1 preferentially bound to genes marked by a high H3K4me3/H3K27me3 ratio (Fig. 3f ) and that ALKBH1 functioned to maintain Fig. 4 The alkbh1 mutation increased genome-wide H3K27me3.a Histone H3K4me3, H3K9me2, and H3K27me3 levels in alkbh1 compared to wild types (WT) were detected by immunoblotting.H3 was tested as a loading control.Original images are shown in Additional file 8. b Scatterplots of H3K27me3 ChIP-seq peaks in WT and alkbh1.Total peaks are merged from all samples (n = 11,473).The x-axis and y-axis represent the normalized read number at each peak of WT and alkbh1, respectively.Blue dots represent the downregulated peaks; red dots represent the upregulated peaks; differential H3K27me3 peaks (fold change > 1.5, P < 0.05) overlapping between two replicates were considered as significantly changed.c Heatmaps and metaplots of ALKBH1-binding signals in the genes that showed increased H3K27me3 in alkbh1 (n = 1058), of which 225 showed ALKBH1 binding, and the remaining (833 gray) are unbound genes.d Metaplots of elevated levels of H3K27me3 for ALKBH1-binding and non-binding genes in alkbh1 versus WT. e Boxplots of the expression levels of the two gene categories in c in the wild-type, mutant, and OE plants.The y-axis represents the gene expression level in log2 (TPM + 0.01).The center line stands for the median, box limits stand for the upper and lower quartiles, and whiskers stand for 1.5 × the interquartile range.The dotted line is TPM equal to 1. Statistical significance between comparisons was calculated by Student's t-test.*P < 0.05, **P < 0.01, ***P < 0.001.ns, no significant difference.f Heatmaps and metaplots of ALKBH1 binding and chromatin modifications signals in the genes that showed increased H3K27me3 in alkbh1 (n = 1058).The top 225 genes are ALKBH1-binding genes (See figure on next page.)the ratio by restraining H3K27me3.In addition, the 225 genes showed higher levels of ATAC signals at TSS and TTS and lower levels of 6mA, mCG, and mCHG at gene body and R-loop at TTS in wild-type background (Fig. 4f ).Within the 225 genes, 35 were downregulated in the mutant, and 64 were upregulated in the OE plants (Additional file 5: Table S4).More than half (18/35) of the genes encode transcription factors, particularly WRKY, NAC, MYB, and ZOS family members (Fig. 5a, Additional file 5: Table S4), many of which have been shown to regulate stress response [45][46][47].The expression, H3K27me3, and ALKBH1-binding of 7 representative genes in wildtype and alkbh1 plants were validated by RT-PCR and ChIP-PCR (Fig. 5b-e).All of the tested genes were found to be bound by ALKBH1, and the highest binding activity was detected near the TTS (Fig. 5d), confirming the ChIP-seq data.All tested genes showed decreased expression and increased H3K27me3 in the gene body in the mutant, except for CPK4 in which no H3K27me3 was detected (Fig. 5e).

ALKBH1 inhibits H3K27me3 deposition by preventing PRC2 (EMF2b) binding
The increases of H3K27me3 in the mutants suggested that ALKBH1 might either remove or inhibit H3K27me3 deposition.A H3K27me3 demethylase activity of ALKBH1 was not excluded, as its catalytic domain is similar to the conserved histone demethylase JmjC domains [21].In addition, ALKBH1 could demethylate histone H2A [26].However, tests of ALKBH1 demethylase activity on histones in plant cells turned out to be negative (Additional file 1: Fig. S5).Alternatively, ALKBH1 might prevent PRC2 function to deposit H3K27me3 from its binding targets.To test this hypothesis, we compared the ALKBH1 binding peaks (genes) with those of EMF2b in the rice tissues [37].EMF2b binds a large number of genomic loci, only part of them showed H3K27me3  [37].It is suggested those EMF2b-binding sites without H3K27me3 may be pre-nucleation sites for PRC2-mediated H3K27me3 deposition [48,49].Actually, more than 60% (1455/2394) of the ALKBH1-binding peaks or genes (1068/1744) were also bound by EMF2b (Fig. 6a), indicating that ALKBH1 mainly associated with the PRC2-binding targets in the rice genome.Interestingly, the ALKBH1-binding genes displayed the highest EMF2b-binding levels especially in the gene body and at TTS regions (Fig. 6a), suggesting that ALKBH1 may preferentially associate with the loci broadly bound by EMF2b.In addition, the genes bound by both ALKBH1 and EMF2b showed higher H3K27me3 levels than those targeted only by EMF2b in the wild type, which was likely due to the fact that a large portion of the EMF2b-binding genes had no H3K27me3 [37].The alkbh1 mutation had more effects on H3K27me3 levels of the genes bound by both ALKBH1 and EMF2b than those bound by EMF2b alone (Fig. 6b, c), supporting the hypothesis that ALKBH1 impaired PRC2-mediated H3K27me3 deposition on a subset of PRC2 targets.Among the 255 ALKBH1-binding genes that showed increased H3K27me3 in alkbh1 mutant, 137 were also bound by EMF2b (Fig. 6d), of which 30 were upregulated and 2 were downregulated in the alkbh1 mutants (Additional file 6: Table S5, Fig. 6e).The observations suggested that ALKBH1 may either compete with EMF2b for binding to the target genes or impair the EMF2b (or PRC2) function to deposit H3K27me3 from nucleation sites toward gene body to a threshed level that leads to gene repression.ChIP-PCR analysis indicated that EMF2b bound 6 of the above analyzed 7 ALKBH1 target genes and the binding levels were largely increased in the gene body of WRKY53, WRKY71, NAC3, and ZOS3 in the alkbh1 background (Fig. 6f ), supporting the assumption that ALKBH1 inhibits the binding of EMF2b in genes with the above-described specific chromatin feature.
To investigate whether 6mA was involved in the ALKBH1 function to control H3K27me3, we analyzed the 6mA peaks that colocalized with R-loops (Fig. 1d).These peaks corresponded to 7484 genes, of which 4425 were marked by H3K27me3 (Additional file 1: Fig. S6a).The marked genes showed higher 6mA, R-loop, H3K4me3, ATAC, and ALKBH1-binding signals than the remaining unmarked 3059 genes that were, by contrast, enriched for TE-related genes (1648/3059), H3K9me2 and DNA methylation (Additional file 1: Fig. S6b), suggesting that ALKBH1 targeted 6mA-methylated R-loops in euchromatin regions.In addition, the H3K27me3 signals of the marked genes displayed a reverse trend of H3K4me3 and ALKBH1 binding signals (Additional file 1: Fig. S6b).The alkbh1 mutation reduced the expression levels of the 4425 genes but had no effect on the 3059 genes (Additional file 1: Fig. S6c).In the alkbh1 mutant, 733 genes of the H3K27me3-marked genes showed increased 6mA levels compared with 106 genes with decreases of the modification (Additional file 1: Fig. S6d), and 392 genes showed increased H3K27me3 compared with 78 genes with decreases of the mark (Additional file 1: Fig. S6e).Collectively, the analysis supported the hypothesis that ALKBH1 targeted R-loops within euchromatin to control 6mA and H3K27me3 levels.

ALKBH1 is required for stress-responsive gene expression and plant tolerance
Because the downregulated genes in the mutant and the upregulated genes in the OE plants are enriched for stress-responsive function, and many of the 35 genes directly activated by ALKBH1 have been previously shown to function in plant stress responses, we tested stress-induced gene expression and/or tolerance of the transgenic plants to salt, heat, and biotic (infection by Xanthomonas oryzae pv.oryzae, Xoo) stresses.For gene expression, we treated the plants cultured in a hydroponic nutrient solution supplemented with 180 mM NaCl for 12 h or incubated the plants at 42 ° C for 3 h and sampled them for RT-PCR, and the tests revealed that all of the tested genes were induced by the salts and most by the heat stress.Nearly all of the tested genes showed a lower expression in the mutants before and after the stress treatments (Fig. 7a).For plant tolerance experiments, after treatment with NaCl for 3 days, hydroponic cultured plants were transferred to the normal growth medium for 7 days for recovery.The recovery rates of the mutants were actually higher and the OE plants were lower than the wild-type plants (Fig. 7b).Although no clear phenotype was observed during heat stress, the seed setting rate of the mutant was greatly reduced under high temperature (Fig. 7d).By contrast, compared to the wild type, the mutants appeared to be more susceptible and the OE plants were more resistant to the Xoo infection (Fig. 7c).The data suggest that ALKBH1dependent gene expression may have distinct effects in plant response to biotic and abiotic stresses.

Multiple functions of ALKBH1 in chromatin modification
The present work provides evidence that rice ALKBH1 has a function to restrain H3K27me3 to stimulate the expression of genes with a specific chromatin signature in plants.Although the catalytic domain resembles that of histone demethylase [21], the rice ALKBH1 appears to have no H3K27me3 demethylase activity.The modulation of H3K27me3 by ALKBH1 is likely achieved by inhibiting PRC2 (EMF2b) binding or spreading at the common targets.However, at this stage, it is not excluded that the 6mA demethylase activity of the protein might be also involved, as it was shown in mammalian cells that 6mA deposition preserves genome-wide levels of ubiquitination of H2A lysine 119 (H2Aub) and Polycomb silencing [38].H2Aub is deposited by PRC1 and recognized by PRC2 to deposit H3K27me3.However, immunoblot analysis did not detect (See figure on next page.)Fig. 7 ALKBH1 is required for stress-induced gene expression and is involved in plant tolerance to stress.a RT-PCR analysis of ALKBH1 target genes expression in plants treated with salt (green) and high temperature (red).Three biological replicates were performed.Student's t-test was used to detect significance.Significance was calculated by comparing the mutant with the wild type and marked on the mutant column.b Recovery phenotypes of wild type and the transgenic plants from salt stress.The seedlings of Nipponbare (NIP, the wild type control with alkbh1 and OEALKBH1), alkbh1 and OEALKBH1 grown under 12-h light/12-h dark conditions for 14 days, transferred to 180 mM NaCl for 3 days and recovered for 7 days, respectively.Scale bar, 5 cm.Data are presented as mean ± SD (n > 3 biological replicates, and 35 plants were tested in each of the biological replicates).Significance of differences (***P < 0.001) was determined by Student's t-test.Significance was calculated by comparing the mutant or overexpression with the wild type and marked on the mutant and overexpression column.c Leaf phenotypes of 8-week-old WT, alkbh1 and OEALKBH1 plants inoculated with the Xoo isolate PXO99A.Lesion lengths were measured and the values are means ± SEM (n > 30 biological replicates).Significance was calculated by comparing the mutant or overexpression with the wild type and marked on the mutant and overexpression column.d Phenotype and seed setting rate of the mutant and overexpressed plants under normal and high-temperature conditions.Data are presented as mean ± SD (n > 20 biological replicates).Significance of difference (***P < 0.001 was tested by Student's t-test).Significance was calculated by comparing the mutant or overexpression with the wild type and marked on the mutant and overexpression column any difference in H2Aub between the wild type and the alkbh1 mutants or OE plants (Additional file 1: Fig. S4b, c).In addition, 6mA and R-loops were detected in the tested ALKBH1 target genes, but in the regions where the ALKBH1-binding is absent or low (Fig. 5a).This suggests that ALKBH1 may have a function to restrict 6mA spreading and to reduce H3K27me3 in R-loops, which has been shown to enhance PRC2 function [50,51].This is supported by the data showing that ALKBH1 preferentially bound to genes enriched for 6mA-methylated R-loops in euchromatin regions and that the alkbh1 mutation led to increases of both 6mA and H3K27me3 in this gene category (Additional file 1: Fig. S6).Although no clear variation of H3K4me3 in the alkbh1 mutants was observed, the ALKBH1-dependent high H3K4me3/H3K27me3 ratio to simulate the expression of the target genes may also involve a function of ALKBH1 to stabilize or recruit H3K4me3 methyltransferases.In Drosophila, the DNA 6mA demethylase DMAD, an ortholog of 5mC demethylase ten-eleven translocation (TET) that is distally related to the AlkB dioxygenase, interacts with the Trithorax-related complex protein Wds (for H3K4me3 deposition) by dynamically demethylating intragenic 6mA [12,52].Accumulation of 6mA by depleting DMAD coordinates with Polycomb proteins and contributes to transcriptional repression of these genes [53].No TET homolog of 5mC demethylase is found in plants so far.The distal relationship of ALKBH1 with TET and its multiple demethylase substrates allow us to speculate that ALKBH1 may have an activity to maintain low mCG levels in the body region of the target genes.

ALKBH1 targets a specific chromatin signature poised for stress response
In this work, we showed that ALKBH1 preferentially targeted genes marked by the specific chromatin signature (i.e., high H3K4me3 and H3K27me3 but depleted of mCG) in the body region.H3K4me3 is deposited by Trithorax complexes and positively regulates transcription, whereas H3K27me3 is deposited by PRC2 and promotes gene repression [54].In plants, H3K27me3 is mainly found in the gene body, and its deposition in the Arabidopsis FLC gene triggered by cold temperature maintains the repression state of the gene across multiple rounds of cell division during growth [55,56].By contrast, at the genomic scale, H3K4me3 is enriched at the TSS site of expressed genes in plants.The present and previous data suggest the two antagonstic histone methylation marks may co-exist in the body region of a set of genes in plants [36], although it is not excluded that the co-marking may be due to cell type heterogeneity of the analyzed tissues.This is reminiscent of the so-called bivalent genes of developmental transcription factors that are lowly expressed in early embryonic cells and are poised for lineage-specific expression during differentiation in mammals.However, the bivalency of the mammalian genes occurs mainly at the promoters or enhancers.In addition, the rice genes marked by H3K4me3 and H3K27me3 are depleted of mCG (Additional file 1: Fig. S2).This is consistent with the observations that PRC2 and H3K4 methyltransferases localize to unmethylated CG islands in the promoter regions in mammalian embryonic cells, and conversely, the presence of H3K27me3 and/or H3K4me3 opposes DNMT activity [57,58].In plants, an antagonistic relationship between H3K4me3 and CG methyltransferase MET1 was also established [59].The two antagonistic histone methylations in the gene body likely mark a transient or intermediate state of gene activity, as those genes are enriched for stress responses in rice plants [36].This is consistent with the functional enrichments of downregulated genes detected in the alkbh1 mutants.Our data indicate that ALKBH1 has a function to maintain gene expression levels by modulating the H3K4me3/H3K27me3 ratio.This is supported by the requirement of ALKBH1 for the stress-induced expression of the tested genes and stress tolerance phenotypes of the mutant and OE plants.

Conclusions
In this work, we discovered that the evolutionarily conserved dioxygenase ALKBH1, previously shown to demethylate 6mA from single-stranded or unpairing DNA, is required to remove 6mA from R-loops in euchromatin regions of the rice genome and has a function to inhibit H3K27me3 or to reduce the H3K27me3/H3K4me3 ratio to promote gene expression.The ALKBH1 function in histone methylations and gene expression essentially relies on impairing PRC2 binding to target genes to deposit the repressive H3K27me3 mark.These results uncover a novel function of ALKBH1 and provide new insights into the regulatory mechanism of PRC2 function to control gene expression.

Rice mutant and overexpression lines
All experiments were performed in the rice (Oryza sativa spp.japonica) cultivar Nipponbare (NIP).The alkbh1#1 and alkbh1#3 mutants are two previously characterized lines in the NIP background [7].The alkbh1#2 mutant was produced using a CRISPR-Cas9 system.CRISPR-P (http:// crispr.hzau.edu.cn/ CRISP R2/) was used for gRNA design.The sequence GAG GCG AGC CTT CTG CCG CGGGG in ALKBH1 was selected as the specific target.The Gibson Assembly method (Kpn I digestion) was used to assemble the gRNA oligonucleotides into the CRISPR system.For the overexpression experiment, the ALKBH1 full-length cDNA without a stop codon was amplified from the NIP cDNA using primers OEALKBH1-F and OEALKBH1-R (Additional file 7: Table S6) and then cloned into the overexpression vector pU1301 in frame with 3XFLAG tag at the 3′ end with Kpn I and BamH I sites.alkbh1#1 and OEALKBH1#1 were used for high-throughput sequencing.

Plant growth
For high-throughput sequencing, rice seeds were surface sterilized with 75% (v/v) ethanol (2 min) and 0.1% (w/v) HgCl 2 (12 min) before sowing.Seedlings were grown on half-strength Murashige and Skoog (MS) solid medium (with 0.15% [w/v] phytagel and 2% [w/v] sucrose) in sterilized tubes for 12 days.Plants were grown in a growth chamber under a 16-h light/8-h dark photoperiod with a day temperature between 28 and 32°C and a night temperature between 23 and 26 °C.

Salt and heat stress treatments
For physiological analysis, seeds were immersed in water at 37 °C for 2 days; then, the germinated seeds were placed into the bottomless 96-well plate that grew in hydroponic nutrient solution [35].For seedling stress tests, 2-week-old plants were transferred to the hydroponic nutrient solution containing 180 mM NaCl for growing 3 days.Subsequently, they were moved to a normal hydroponic nutrient solution for recovery.After 7 days, the survival rate was calculated according to the percentage of seedlings alive.
For identifying rice thermotolerance at the reproductive stage, the rice plants at the heading stage were transferred from the experimental field to the greenhouse (38 °C, 12 h light/35 °C, 12 h dark) till to maturity [60].
The 2-week-old seedlings were transferred to a hydroponic nutrient solution containing 180 mM NaCl at the seedling stage to grow for 12 h, and then leaf samples were harvested for RNA extraction and RT-PCR.The 2-week-old seedlings were transferred to 42 °C high-temperature incubators for 3 h at the seedling stage, and then leaf samples were harvested for RNA extraction and RT-PCR.

Pathogen inoculation
Rice flag leaves were inoculated with X. oryzae pv.oryzae (Xoo) strains at the booting (panicle development) stage by the leaf-clipping method as previously described [61].Xoo strains used in this study included Philippine strains PXO99.Disease was scored by measuring the lesion length 14 days after inoculation.The bacterial growth rate in rice leaves was determined by counting colony-forming units as described previously [62].

6mA DIP-seq and analysis
For 6mA IP-seq, rice genomic DNA was extracted from 2-week-old sterile-cultured seedlings and incubated with RNase A overnight.The 6mA IP-seq was performed using published protocols [3].Briefly, 5-μg genomic DNA was sonicated into 200-400 bp and end-repaired, followed by adapter ligation.DNA was denatured at 95 °C for 10min and chilled on ice for 10 min.An aliquot of the denatured DNA was saved as input.The remaining DNA was immunoprecipitated overnight at 4 °C with anti-6mA antibodies purchased from SYSY (202-003).DNA was eluted by 6mA salt competition.After size selection and PCR amplification, the DNA fragments were sequenced with the HiSeq 2500 (Illumina).Low-quality reads were removed from raw data by using the Trimmomatic package (http:// www.usade llab.org/ cms/ uploa ds/ suppl ement ary/ Trimm omatic/ Trimm omatic-Src-0.35.zip), and the clean data were mapped to the MSU7.0 rice reference genome (ftp:// ftp.plant biolo gy.msu.edu/ pub/ data/ Eukar yotic_ Proje cts/o_ sativa/ annot ation_ dbs/ pseud omole cules/ versi on_7.0/ all.dir/) using Bowtie2 with parameters permitting less than two mismatches.The reads were mapped to bacterial, fungal, and archaeal genomes (ftp:// ftp.ncbi.nlm.nih.gov/ genom es/ genba nk/).Then, MACS (model-based analysis of chromatin immunoprecipitation followed by sequencing (ChIP-seq) [63]) was introduced to locate enriched regions to call 6mA peaks by comparing reads from the IP sample with the input sample.The DiffBind (version 3.13) Bioconductor package [64] was used to find differential peaks between alkbh1 and WT samples.The DESeq2 method implemented in DiffBind was used to test the differential peaks.Fold change > 1.5 and P-value < 0.05 were used to define differential binding peaks.

ALKBH1 antibody preparation
Preparation of polyclonal antibodies in rabbits using rice ALKBH1-specific peptide (amino acid [aa] 21-38) as antigen.The peptide (C-RQKARLPRGPVHEKSLEQ) (24mg) was synthesized in vitro.After three cycles of injections, the antisera were obtained, affinity-purified using antigen-coupled magnetic beads, and mixed with 20% glycerol for storage at − 80 °C.

ChIP
Two grams of 2-week-old rice seedling was cross-linked by 1% (v/v) formaldehyde and used for chromatin extraction.Chromatin was fragmented to around 200 bp by sonication and then incubated with antibody-coated beads (Invitrogen/Life Technologies; 10001D) overnight.After extensive washing, immunoprecipitated chromatin was de-cross-linked and retrieved, and unimmunoprecipitated chromatin was used as input.The precipitated and input DNA samples were analyzed by highthroughput sequencing or by RT-PCR with gene-specific primers listed in (Additional file 7: Table S6).Anti-H3K27me3 (Millipore; 07-449), anti-ALKBH1 (this work), anti-EMF2b (Tan et al 2022), and anti-IgG (Abcam; ab37415) antibodies were used.

ChIP-seq data analysis
DNA from chromatin immunoprecipitation was used to construct sequencing libraries following the protocol provided by the Illumina TruSeq ChIP Sample Prep Set A and sequenced on Illumina HiSeq 2500.Trimmomatic (version 0.32) was used to filter out low-quality reads.Clean reads were mapped to the rice genome (MSU7.0;http:// rice.plant biolo gy.msu.edu/) by Bowtie2 (version 2.4.2) using default parameters.SAMtools (version 1.9) was used to remove potential PCR duplicates.The bam files were first converted to Wiggle files, and bigwig files were generated using bam-Coverage with the parameters "-bs 10 -normalizeUsing RPKM" in DeepTools [68], and the data were imported into the Integrated Genome Browser for visualization.The two replicates were merged using MACS2 in peak calling (version 2.1.1)[63], for H3K27me3 and ALKBH1 with the parameters "--broad --broad-cutoff 0.1".The reads from input were used as controls for H3K27me3 and ALKBH1 ChIP-seq, respectively.The DiffBind (version 3.13) Bioconductor package [64] was used to find differential peaks between alkbh1 and WT samples.The DESeq2 method implemented in Diff-Bind was used to test the differential peaks.fold change > 1.5 and P-value < 0.05 were used to define differential binding peaks.

Reverse transcription and RT-PCR
One microgram of total RNA was reverse-transcribed in a reaction volume of 20 μL using DNase and reverse transcriptase (Vazyme; Nanjing, China; R233-01) according to the manufacturer's instructions to obtain cDNA.RT-PCR was performed as follows: 95 °C for 10 s, 45 cycles of 95 °C for 5 s, and 60 °C for 40 s.Disassociation curve analysis was performed at 95 °C for 15 s, 60 °C for 20 s, and 95 °C for 15 s.Three biological replicates were performed (separate experiments), each with three technical repeats (three identical samples within an experiment).Data were collected using the ABI PRISM 7500 sequence detection system.The rice ACTIN1 gene was used as the internal control.The primers used for RT-PCR are listed in (Additional file 7: Table S6).

Phylogenetic analysis
The ALKB sequences of E. coli, mouse, human, and Arabidopsis were downloaded from published data and used for alignment with rice sequences [5,13,69].The ClustalX program was used to generate alignments of ALKB protein sequences.The phylogenetic tree generation was performed by MEGA7.
• fast, convenient online submission • thorough peer review by experienced researchers in your field • rapid publication on acceptance • support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year

•
At BMC, research is always in progress.

Learn more biomedcentral.com/submissions
Ready to submit your research Ready to submit your research ?Choose BMC and benefit from: ? Choose BMC and benefit from:

Fig. 3
Fig. 3 ALKBH1 genome-wide binding characteristics.a Metaplots and heatmaps of ALKBH1-binding signals (centered in the binding peaks) in the wild type and alkbh1#1 mutant.b Metaplots and heatmaps of H3K27me3 and H3K4me3 levels in ALKBH1-binding genes.c Scatterplots of the differentially expressed genes in alkbh1 with the ALKBH1-binding genes.The x-axis and y-axis respectively represent the normalized reads number of the DEGs (log 2 FC > 1, q-value < 0.05) in alkbh1 and in the wild type.The blue dots represent the ALKBH1-binding genes.P-values were calculated by hypergeometric tests.d Scatterplots of the DEGs in OEALKBH1 with the ALKBH1-binding genes.The x-axis and y-axis respectively represent the normalized reads number of the DEGs (log 2 FC > 1, q-value < 0.05) in OEALKBH1 and in the wild type.The red dots represent the ALKBH1-binding genes.P-values were calculated by hypergeometric tests.e Metaplots and heatmaps of chromatin modification profiles in the ALKBH1-binding genes that are downregulated in alkbh1 (n = 324 blue), upregulated in OEALKBH1 (n = 349 red), and unchanged in the mutant or OE plants (n = 1133 gray).f Boxplots of H3K4me3/H3K27me3 ratios of the ALKBH1-binding genes that are downregulated in the mutant (324), upregulated in OE plants (349), and unchanged in the transgenic plants (1133).Statistical significance between wild type and alkbh1 was calculated by Student's t-test.*P < 0.05, **P < 0.01, ***P < 0.001.ns, no significant difference

Fig. 5
Fig. 5 ALKBH1-binding inhibits H3K27me3 and promotes the expression of target genes.a Integrative Genomics Viewer of 7 stress-related transcription factor genes targeted by ALKBH1.b Three regions (red horizontal bars) of the 7 genes tested by ChIP-PCR analysis with H3K27me3 and ALKBH1 antibodies.c RT-PCR analysis of the 7 genes in wild-type and alkbh1 mutant lines.P2 primers were used.d Anti-ALKBH1 ChIP-PCR detection of ALKBH1 association with the 7 genes in wild-type and alkbh1 plants.e ChIP-PCR detection of H3K27me3 levels in the 7 genes in wild-type and alkbh1 plants.Values are the means of three biological replicates, and error bars represent ± SD.Statistical significance between wild type and alkbh1 was calculated by Student's t-test.*P < 0.05, **P < 0.01, ***P < 0.001.n, no significant difference

Fig. 6
Fig. 6 ALKBH1 binds to a subset of the PRC2 targets the rice genome.a Metaplots and heatmaps of ALKBH1-binding EMF2b binding genes (n 6240).b Metaplots of H3K27me3 levels of the genes (1068) commonly bound by EMF2b and ALKBH1 (upper) compared to those bound by EMF2b only (lower) in WT and alkbh1.c The fold change (log 2 ) of H3K27me3 of the two gene populations in alkbh1 relative to wild type.The P-value was calculated based on the t-test.d The Venn diagram between 1068 genes commonly bound by EMF2b and ALKBH1 and 225 ALKBH1-binding genes that shown increased H3K27me3 in the alkbh1 mutants.e Expression levels (in log 2 TPM) of the 137 genes in wild type and alkbh1.f Anti-EMF2b ChIP-PCR detection of EMF2b binding to the 7 ALKBH1-targeted genes in wild type and alkbh1.Data represent the means ± SD (three separate experiments per sample).Statistical significance between wild type and alkbh1 was calculated by Student's t-test.*P < 0.05, **P < 0.01, ***P < 0.001.n, no significant difference