Reconfiguration of DNA methylation at growth resumption after winter dormancy in Kiwifruit

Abstract

DNA methylation is an epigenetic modification that plays a role in developmental transitions by regulating gene expression. In kiwifruit (Actinidia spp.), the transition from winter dormancy to growth in spring is a critical phase, which sets the stage for flowering and fruit development in the upcoming season. It is unclear if this process is associated with changes in DNA methylation. We found that applying a hypo-methylating agent, 5-azacytidine, to dormant buds at specific stages significantly accelerated shoot emergence in kiwifruit. Genome-wide identification and expression pattern analysis of kiwifruit DNA methyltransferase and demethylase gene families identified that some of these transcripts preferentially accumulated in axillary buds during dormancy in winter or growth resumption in spring. We generated a single-base resolution map of CG, CHG and CHH methylation by whole-genome bisulfite sequencing (BS-seq) and performed a transcriptome analysis to compare dormant and growth-resuming buds. We found genome-wide CHH hyper-methylation in dormant buds and dramatically reduced CHH methylation at the stages when vegetative growth and reproductive development were re-established. Many genes involved in growth and flowering were correlated with the amount of DNA methylation, with specific changes detected in promoter regions. Surprisingly, CHH hypomethylation during the transition from winter dormancy to growth resumption was accompanied by downregulation of transposable element expression. Therefore, alternative mechanisms exist to maintain genome stability and prevent harmful TE activity during growth resumption in kiwifruit buds.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-27058-x.

Introduction

DNA methylation is a universal mechanism of epigenetic regulation in all kingdoms of life¹. Methylation involves dynamic addition or removal of a methyl group to, or from, a cytosine (C) residue in DNA: this affects multiple processes, including genome stability, gene expression, and gene imprinting^2–4. In plants, DNA methylation can occur in three sequence contexts, CG, CHG, and CHH, where H represents a nucleotide other than Guanine (G)^5,6.

Changes in DNA methylation are often classed into two categories, reprogramming and reconfiguration. Reprogramming involves demethylation of DNA, followed by establishment of new DNA methylation patterns, as shown during embryonic development in animals where methylation determines the state of pluripotency and alteration in methylation is involved in transition from stem cells into differentiated cell types⁷. In plants, genome-wide DNA methylation reprogramming involving small RNAs has been described in Arabidopsis thaliana (Arabidopsis) pollen^8,9 and global reprograming during germ cell differentiation establishes hyper-methylation in gametes and hypo-methylation in their companion cells, preserving genome integrity from parent to offspring^10–12. Reconfiguration of DNA methylation, such as increasing methylation during seed development and decreasing methylation during germination, respectively, establish new methylation patterns and levels without initial global demethylation^13–15.

Differential DNA methylation is implicated in regulation of plant developmental transitions. Increases in methylation have been detected during embryogenesis¹³ and in the shoot apical meristem (SAM), which contains a pluripotent stem cell population responsible for growth and floral transition¹⁶. A clear link between DNA methylation and flowering has been established in Arabidopsis^17,18 and sugar beet^19,20. Similarly, hyper-methylation of DNA is associated with vegetative-to-reproductive transition in the SAM in rice²¹. Dynamic differential DNA methylation regulates nodule development in Medicago, showing that both demethylation and hypermethylation are key factors in regulation of plant development and interactions with symbiotic bacteria²², while a conserved feedback circuit involving DNA methylation and specific transcription factors regulates fruit ripening^23,24. Furthermore, DNA methylation has been implicated in plant adaptation and stress responses²⁵.

DNA methylation is dynamic with ongoing alterations in establishment, maintenance and removal of methylation²⁶, and the average methylation patterns in specific cell types result from the balance of these processes²⁷. The establishment of DNA methylation is regulated by DNA methyltransferases, which transfer a methyl group to unmethylated cytosine residues. Three classes of plant DNA methyltransferases recognize specific methylation contexts^28,29. CG methylation is maintained preferentially by the METHYLTRANSFERASE 1 (MET1) family²⁸, whilst CHG methylation is controlled coordinately with histone methylation by CHROMOMETHYLASE (CMT) enzymes^30,31. CHH methylation is mainly mediated by the RNA-directed DNA methylation (RdDM) pathway^32–34, in which 24-nt small RNAs target transposable elements (TEs) and the activity of DOMAINS REARRANGED METHYLTRANSFERASE (DRM) is required to establish methylation. DRMs are responsible for CHH methylation of small TEs or the edges of longer TEs³³, whilst CHH methylation within the internal regions of longer TEs may be dependent on CMT2 chromomethylase^33,34. While considerable redundancy may exist, mutant screens revealed the importance of DNA methylation maintenance in regulation of plant reproductive development, as well as vegetative development and genome integrity in model annual plants such as Arabidopsis^35,36 and rice^37–39.

Demethylation can occur passively, when unmethylated DNA strands are synthesized during DNA replication, or actively, when methylated bases are removed by DNA demethylases. The Arabidopsis genome contains four DNA demethylase genes, DEMETER (DME), REPRESSOR OF SILENCING 1 (ROS1) also known as DEMETER-LIKE 1 (DML1), DML2, and DML3⁴⁰. DML homologs have been identified in other plant species⁴¹ and demethylation associated with aspects of plant development^22,42,43 and responses to abiotic and biotic stress^44–47 in diverse plants.

The interactions between plant genome, transcriptome and epigenome have been studied predominantly in annual plants. However, effects of DNA methylation and demethylation during developmental transitions raises the possibility that DNA methylation plays a crucial role in phenology of perennial lants. In temperate climates, woody perennials cease growth, set buds, and establish winter dormancy to survive unfavorable conditions. Re-establishment of growth in spring involves progression from complete metabolic dormancy to a highly active state, giving rise to rapid shoot development and, in reproductively mature plants, development of flowers. The release of bud dormancy is orchestrated by genetic, biochemical and hormonal changes^48,49 and has been shown to involve epigenetic changes⁵⁰. Histone modifications and non-coding RNAs have been associated with transcription of MADS box genes that act as key regulators of dormancy or growth resumption^51–53. Furthermore, changes in DNA methylation were detected during dormancy and its release in trees including apple, almond, tree peony and chestnut^54–57, and a chilling-responsive DML was associated with poplar budbreak⁴¹.

A common strategy to investigate the role of DNA methylation in plant development involves chemically induced demethylation using agents like 5-azacytidine (5-azaC). 5-azaC is a non-methylable cytosine analogue which incorporates into DNA during replication, thus leading to demethylation across the genome in all sequence contexts^58,59. Furthermore, it irreversibly binds methyltransferases to induce DNA hypomethylation^60,61. A reduction in DNA methylation after the application of 5-azaC affected plant development, including early somatic embryogenesis and flowering in different plant species^62,63. In the context of bud break in perennial plants, 5-azaC has been shown to promote bud break by altering gene expression patterns associated with dormancy release, as observed in grapevine and peony tree^64,65. Further strategies to investigate the role of DNA methylation in plant development include expression and functional studies of methyltransferase and demethylase genes, and analysis of global DNA methylation patterns across developmental stages.

We study Actinidia chinensis (kiwifruit), a woody perennial vine in which growth and floral development in axillary buds are re-established after a winter-dormant period⁶⁶. Our previous studies identified that epigenetic mechanisms, including histone modifications and noncoding RNAs, contribute to regulation of kiwifruit dormancy-related MADS-box genes^52,53. In this study, we investigated the role of DNA methylation in regulation of seasonal growth cycles of kiwifruit. To that end, we used chemical treatments to alter DNA methylation, examined the expression of methylation-related genes, and carried out methylome, transcriptome and TE analysis to compare dormant and non-dormant buds. Treatment with 5-azaC, differential expression of methyltransferase and demethylase genes in buds and marked reduction of CHH methylation at the stage when vegetative growth and reproductive development were re-established all suggested that methylation reprogramming occurred during the resumption of growth after winter dormancy in kiwifruit. Concomitant downregulation of TE expression and upregulation of the RdDM pathway suggested a mechanism for silencing of these potentially harmful TE activities. This study has implications for the management and breeding of perennials with altered phenology that are better suited for changing climate conditions.

Results

Application of 5-azacytidine accelerates shoot outgrowth in Kiwifruit

To study how DNA methylation influences bud dormancy, we monitored the effect of the DNA methylation inhibitor 5-azaC on budbreak at different times during the nongrowing season. Single-node cuttings were excised from kiwifruit vines grown in the field between June and August in 2020, to represent winter dormancy (June–July), and dormancy release in early spring (August) (Fig. 1). Cuttings were treated with 250 μm and 500 μm 5-azaC or water as the control, then maintained at permissive conditions to force budbreak. The average time to budbreak in control cuttings steadily declined with accumulation of chilling (Fig. 1A), consistent with previous findings for ‘Zes008’ and other kiwifruit cultivars⁶⁷. Significantly faster shoot emergence was observed in samples treated with 5-azaC compared with the water-treated controls, but only on cuttings collected in June (Figs. 1B, C). The experiment was repeated in the following year (2021), with sampling performed in autumn and winter. As before, a reduction in days to budbreak in cuttings from May to June was observed (Fig. S1) and earlier shoot outgrowth in samples treated with 5-azaC was noted when the treatment preceded the 100% leaf drop stage (Fig. S1). In summary, application of 5-azaC before and during early stages of dormancy correlated with earlier shoot emergence, suggesting that DNA methylation plays a role in preventing outgrowth of new kiwifruit shoots in autumn and early winter.

Fig. 1.

Treatment with 5-azacytidine (5-azaC) affects budbreak in kiwifruit. (A). Forced budbreak of single-node cuttings was performed at room temperature under long-day conditions after application of 5-azaC. The box plots represent days to budbreak after treatment with 5-azaC or water control. The whiskers represent the minimum and maximum, the bottom and top of the box are the 25th and 75th percentiles, the line inside the box is the 50th percentile (median) and the cross represents the mean. A minimum of 15 cuttings per line was used in a forced budbreak assay. The samples are described in the table below. Student’s t-test was used to determine statistical significance, with P < 0.01 considered significant. (B). Representative cuttings collected on 17 June 2020, photographed 24 days after 5-azaC treatment. (C). Representative cuttings collected on 25 August 2020, photographed 10 days after 5’-azaC treatment.

Identification and differential expression of Kiwifruit methyltransferase and demethylase genes

The relationship between kiwifruit seasonal growth cycles and expression of genes regulating DNA methylation was examined. Interrogation of the kiwifruit genome identified multiple homologs of MET1 and CMT, predicted to be preferentially involved in methylation maintenance, and DRM, predicted to control de novo methylation (Fig. 2A). Similarly, nine gene models were identified by homology to Arabidopsis demethylases (Fig. 2B). Interrogation of expression of these in kiwifruit tissues^53,68 revealed that most were differentially expressed and confined to actively growing terminal buds, with little or no expression in axillary buds (Fig. 2C, Fig. S2). Methyltransferase gene homologs Acc15408 (DNMT2-like), Acc16423, Acc28838 (DRM-like), and Acc29050 (CMT-like) were highly expressed in axillary buds, with peak expression during winter dormancy (June–July). Among the demethylase homologs with prevalent expression in axillary buds, Acc03168, and Acc00310 showed peak expression in August, coinciding with dormancy release (Fig. 2C). These genes showed elevated expression in response to cold, and reduced expression at budbreak in August (Fig. S2), suggesting a role for DNA methylation in growth prevention.

Fig. 2.

Actinidia chinensis (kiwifruit) methyltransferase and demethylase genes. (A). Phylogenetic tree with kiwifruit homologs of methyltransferase genes implicated in maintenance and de novo methylation. (B). Phylogenetic tree of kiwifruit demethylase genes. Alignments and tree construction were performed using Geneious MUSCLE and Geneious Tree builder (Neighbor Joining, Bootstrap 1000 replicates). (C). Gene expression in kiwifruit buds at phenological stages indicated underneath, as described {Voogd, 2022 #47}. The color scheme shows relative expression of each gene ranging from minimum (blue) to maximum (red). Asterisks (*) denote genes predominantly expressed in axillary buds, with peak expression either during winter dormancy or at the time of dormancy release.

Differential DNA methylation in buds during dormancy and growth resumption

The observation that hypomethylation induced by 5-azaC could advance shoot emergence, combined with expression of DNA methylation genes in axillary buds, supported the hypothesis that differential DNA methylation regulates growth and dormancy in kiwifruit. To investigate this, a whole-genome bisulfite sequencing (WGBS) was performed on kiwifruit dormant buds (June) and on buds resuming growth (August) (Fig. 1A, Table S1). WGBS analysis revealed a genome-wide reprogramming of DNA methylation during dormancy and resumption of growth, primarily through global changes in CHH (Fig. 3A). The analysis of differential methylated regions (DMRs) detected 505,738 CHH DMRs, almost 100% of which were hypomethylated in August (during growth resumption) compared with June (dormant buds) (Fig. 3B). Although global CG and CHG methylations were not significantly changed between August and June (Fig. 3A), we detected 228 CG and 10,001 CHG DMRs (Fig. 3B). Over three quarters of CG DMRs (76.8%, 175 DMRs) were hypermethylated in August compared with June (Fig. 3B). In contrast, 99.6% of CHG DMRs showed significant methylation reduction in August (Fig. 3B). To investigate the location distribution of DMRs, we annotated and grouped DMRs in association with the genomic space defined by gene annotation, in which the promoter region was defined as 2 kb region upstream of transcription start site (TSS). This showed 33.6% to 41.1% DMRs were associated with genes and most of them were in the promoter region, irrespective of the cytosine contents or the direction of methylation changes (Figs. 3C, D). The analysis of individual differential methylated cytosine (DMC) showed a similar trend to that detected in DMR analysis (Fig. S3).

Fig. 3.

Differential DNA methylation in kiwifruit dormant axillary buds (June; JUN) and at growth resumption (August; AUG). (A). Genome wide methylation in CG, CHG and CHH contexts. Error bars of the three biological replicates and p-values of t-tests between June and August are as indicated. (B). Volcano plots of differentially methylated regions (DMRs) in CG, CHG and CHH contexts. DMRs showing at least 25% methylation difference (vertical dashed lines) in August compared with June and the adjusted p-value ≤ 0.01 (horizontal dashed line) are shown in black. C-D. Location distribution (promoter, exon, intron or intergenic) of hypomethylated (C) and hypermethylated (D) DMRs (August compared with June). E. Average methylation profiles of all annotated genes and their flanking regions: 2 kb upstream of transcription start site (TSS) and 2 kb downstream of transcription termination site (TTS).

Average methylation profiles of genes

At the gene level, average methylation profiles typical of angiosperms were observed (Fig. 3E). Gene body methylation with a decrease at the transcription start and termination sites were seen in a CG context, while in the CHG and CHH context, hypomethylation of the gene body and a peak of methylation upstream from the transcription start site were detected. Methylation in a CHH, but not a CG or CHG context, was dramatically reduced in the 2-kb regions flanking genes in August compared with June, while no difference in gene body methylation was detected between June and August (Fig. 3E). The methylation percentage was largely correlated with gene expression at both sampling times. An increase in gene body methylation in the CG context correlated with higher expression (Fig. S4). In contrast, an inverse correlation between gene body methylation and expression level was established for CHG and CHH methylation. Overall, these findings support a positive association between the level of gene expression and the levels of upstream CHG and CHH methylation (Fig. S4).

Genome-wide differential methylation in association with gene expression changes

To determine if methylation plays a role in differential gene expression associated with growth and dormancy, a transcriptomic analysis was performed with the bud samples collected in June and August and compared with the WGBS analysis. The RNA-seq analysis (DESeq, padj < 0.05) identified a total of 10,823 differentially expressed genes (DEGs), of which 6573 were up- and 4250 downregulated with a minimum 2-fold higher or lower transcript accumulation in August compared with June, respectively (Table S2).

Analysing DEGs containing DMRs in promoter, intron or exons showed that 6,110 of 6,573 upregulated DEGs and 3,800 of 4,250 downregulated DEGs were solely associated with hypomethylated DMRs, while only a small number (16 upregulated and 27 downregulated genes) contained both hypo- and hypermethylated DMRs (Fig. 4A, Table S3).

Fig. 4.

Distribution of differentially expressed genes (DEGs) and differential DNA methylation (August compared with June). (A). Venn diagram of DEGs and genes with differentially methylated regions (DMRs). (B). Circos plot of kiwifruit chromosomes showing density of annotated genes, DEGs and DMRs in CG, CHG, and CHH contexts across the entire genome. C. Statistical analysis (t-test) of methylation difference of DEGs grouped by DMR location. Diamond symbol indicates mean value of the methylation difference. Dashed lines indicate the cutoff value of differential methylation (± 25%).

The genome-wide density distribution of DEGs and DMRs analysed with 1 Mb window size showed that most of the genomic regions with dense DEGs were also enriched with hypomethylated DMRs (CHG and CHH DMRs in particular) (Fig. 4B, Fig. S5), suggesting that differential methylation occurs in chromosomal regions occupied with transcriptionally regulated genes. Regions with hypermethylation in all contexts have been also noted (Fig. 4B), however DEGs with associated hypermethylation also contain hypomethylated regions (Table S4). DEG-related differential methylation was mostly confined to promoter regions, which were predominantly hypomethylated in the CHH context (Fig. 4C, Fig. S6).

Only a subset of genes showed differential methylation in both the promoter and gene body regions (Table S4), while hypomethylation in all contexts (Fig. 5A) was demonstrated for a single upregulated gene, encoding a hypothetical protein (Acc25030), and three downregulated genes, encoding a RING/U-box superfamily ABA INSENSITIVE RING PROTEIN 2 (Acc01766) critical for abscisic acid (ABA) and high salinity responses during germination⁶⁹, a 4’-phosphopantetheinyl transferase superfamily (Acc16941), and a mitogen-activated protein kinase phosphatase 1 homolog (Acc27684). In addition to a large decrease in CHH methylation in August, differential methylation at specific CG and multiple CHH sites was identified (Fig. 5B). Notably, a significant reduction in CG-context methylation at a specific promoter region approximately 800 bp upstream from the translation start site was associated with expression of Acc25030 in August (Fig. 5C, D). However, both hyper- and hypomethylation of specific C residues in CG context in the promoter region were associated with reduced expression of Acc01766, Acc16941 and Acc27684 in August (Fig. 5C, D), suggesting a more complex regulation and highlighting promoter regions that may play a role in differential expression.

Fig. 5.

Distribution of differentially methylated cytosine (DMC) residues in candidate genes (August compared with June). (A). Venn diagrams indicating the contexts of differentially methylated regions of candidate hypomethylated DEGs. (B). Average methylation site scores from three replicates across methylation contexts and time points in the specified genomic region. The top six panels represent scores (ranging from 0 to 100) for the three methylation contexts (CG, CHG, and CHH) in June (Jun) and August (Aug). The bottom panel illustrates the genomic features in the region, with exons depicted as rectangles and arrows depicting the direction of transcription. (C). DMC residues in CG context in a 5 kb region upstream from the translation start site, plotted as differential between average Aug and Jun methylation percentages. The distance from the translation start site in base pairs is given below. D. Expression (FPKMs ± SE of three biological replicates) of candidate genes.

Bud dormancy and growth-related genes are differentially methylated

Previously, we have identified genes that were differentially expressed at bud dormancy release⁵³, including transcription factors, cell-cycle and flowering genes. Expression of many of these genes was elevated in August compared to July (Table S3) and a substantial CHH hypomethylation was detected in their promoter regions (Table S4), suggesting that demethylation at the promoter regions may be required for increased expression of these genes in spring. For example, CHH hypomethylation affected 37 cyclin genes, all of which were upregulated, as well as 28 upregulated MADS-box genes, homologs of Arabidopsis floral meristem and floral organ identity genes (Table 1). Gene ontology analysis of all upregulated DEGs associated with hypomethylated DMRs showed that they were enriched for development-related categories such as catalytic activity, including oxidoreductase and hydrolase activities, structural molecule activity and cytoskeletal protein binding (Fig. S7A). These enriched molecular function networks coincided with the enrichment of upregulated DEGs involved in photosystems and intracellular organelle components in cellular component networks (Fig. S7B), as well as photosynthesis, cellular component movement and microtubule-based movement enriched in biological process networks (Fig. S7C).

Table 1.

Differential expression and methylation of CYCLIN (CYC) and MADS-box genes in Kiwifruit bud samples between August and June. CYC genes are indicated in blue font and MADS box genes in red font. Kiwifruit FLC-like associated with epigenetic regulation by histone modifications and long non-coding RNAs is indicated by an asterisk.

Annotation		DESeq analysis
Gene_id	TAIR top hit	e value	log2fold change	padj	meth.diff. (August)	Category (DEG_DMR)
Acc00557	AT5G67260 | CYCD3;2	9.5E-80	1.50	2.32E-03	−34.45	UP_HYPO
Acc01102	AT1G20610 | CYCB2;3	2.6E-152	9.31	7.13E-37	−38.80	UP_HYPO
Acc01195	AT5G65420 | CYCD4;1	2E-90	2.60	7.61E-09	−40.91	UP_HYPO
Acc04491	AT1G76310 | CYCB2;4	0	7.51	7.07E-46	−38.12	UP_HYPO
Acc04754	AT4G34160 | CYCD3;1	1.8E-101	3.20	5.10E-61	−36.11	UP_HYPO
Acc04763	AT1G47210 | CYCA3;2	3.1E-150	9.28	1.71E-14	−41.36	UP_HYPO
Acc04782	AT4G37630 | CYCD5;1	1.5E-45	2.32	2.20E-33	−39.08	UP_HYPO
Acc06013	AT4G34160 | CYCD3;1	6.3E-95	7.05	5.09E-11	−36.76	UP_HYPO
Acc06877	AT1G70210 | CYCD1;1	4.5E-63	2.01	7.82E-22	−38.73	UP_HYPO
Acc10304	AT5G67260 | CYCD3;2	4.1E-93	3.36	2.25E-40	−28.12	UP_HYPO
Acc10528	AT5G65420 | CYCD4;1	1.2E-92	2.12	1.81E-25	−35.62	UP_HYPO
Acc10862	AT1G76310 | CYCB2;4	1.5E-155	7.28	3.72E-43	−41.52	UP_HYPO
Acc11766	AT4G34160 | CYCD3;1	1.2E-101	1.53	2.80E-82	−36.59	UP_HYPO
Acc11958	AT1G20610 | CYCB2;3	0	2.66	1.20E-19	−36.98	UP_HYPO
Acc12063	AT4G34160 | CYCD3;1	6.5E-100	1.51	3.44E-44	−30.66	UP_HYPO
Acc12069	AT1G47230 | CYCA3;4	7.1E-156	5.53	1.02E-57	−40.81	UP_HYPO
Acc12153	AT5G65420 | CYCD4;1	1.4E-83	1.96	4.79E-32	−36.77	UP_HYPO
Acc12296	AT1G44110 | CYCA1;1	0	8.17	5.52E-35	−34.95	UP_HYPO
Acc12588	AT1G80370 | CYCA2;4	1.5E-172	1.56	9.44E-12	−41.56	UP_HYPO
Acc13238	AT5G06150 |CYCB1;2	1.6E-161	10.15	3.54E-33	−34.79	UP_HYPO
Acc13514	AT4G03270 | CYCD6;1	7.09E-28	1.33	3.90E-08	−36.89	UP_HYPO
Acc14887	AT1G16330 | CYCB3;1	4.7E-150	1.33	3.44E-38	−33.07	UP_HYPO
Acc17841	AT1G80370 | CYCA2;4	4.4E-172	2.07	9.10E-29	−37.22	UP_HYPO
Acc20604	AT4G03270 | CYCD6;1	2.7E-65	1.71	6.60E-04	−38.88	UP_HYPO
Acc21080	AT1G47230 | CYCA3;4	9.5E-133	3.01	6.82E-14	−39.14	UP_HYPO
Acc21097	AT5G67260 | CYCD3;2	1.2E-80	1.74	2.39E-53	−40.50	UP_HYPO
Acc22522	AT5G11300 |CYCA2;2	1.3E-156	5.19	6.6E-121	−33.58	UP_HYPO
Acc23082	AT3G11520 | CYCB1;3	1.4E-161	8.56	7.52E-61	−33.07	UP_HYPO
Acc23489	AT1G76310 | CYCB2;4	0	10.25	1.58E-17	−39.60	UP_HYPO
Acc23775	AT4G34160 | CYCD3;1	2.9E-102	5.36	2.68E-96	−26.64	UP_HYPO
Acc23788	AT1G47210 | CYCA3;2	1.7E-149	8.83	3.58E-13	−37.15	UP_HYPO
Acc23807	AT4G37630 | CYCD5;1	3.1E-46	3.89	3.43E-49	−34.52	UP_HYPO
Acc23897	AT5G65420 | CYCD4;1	2.6E-84	1.32	1.44E-08	−36.68	UP_HYPO
Acc25062	AT4G03270 | CYCD6;1	3.4E-62	1.67	9.63E-13	−36.04	UP_HYPO
Acc30317	AT5G06150 |CYCB1;2	3.9E-149	9.94	4.89E-42	−31.10	UP_HYPO
Acc30987	AT5G67260 | CYCD3;2	1.4E-104	5.04	2.62E-59	−42.87	UP_HYPO
Acc32381	AT2G26760 | CYCB1;4	6.5E-144	10.21	8.31E-23	−34.36	UP_HYPO
Acc01045	AT3G54340 | AP3	2.3E-73	2.64	1.67E-95	−35.86	UP_HYPO
Acc02284	AT1G69120 | AP1	2.09E-88	8.03	2.96E-14	−32.63	UP_HYPO
Acc02285	AT3G02310 | SEP2	2.3E-106	8.05	2.44E-14	−36.34	UP_HYPO
Acc04040	AT5G60910 |FUL	1.9E-90	8.15	1.06E-14	−32.92	UP_HYPO
Acc04041	AT3G02310 | SEP2	6.3E-109	3.01	5.46E-38	−32.80	UP_HYPO
Acc05042	AT5G20240 | PI	1.04E-89	8.28	2.83E-11	−29.28	UP_HYPO
Acc06158	AT1G24260 | SEP3	6E-122	7.96	2.56E-10	−41.45	UP_HYPO
Acc07652	AT2G45660 |SOC1	3E-86	1.71	3.96E-77	−35.09	UP_HYPO
Acc08270	AT1G24260 | SEP3	9.5E-121	10.32	1.42E-17	−32.24	UP_HYPO
Acc09079	AT2G45650 | AGL6	5.7E-100	6.62	6.29E-14	−38.61	UP_HYPO
Acc10416	AT3G54340 | AP3	1.01E-23	3.50	3.33E-04	−35.32	UP_HYPO
Acc10522	AT2G22540 | SVP	6.5E-113	1.81	4.63E-47	−36.88	UP_HYPO
Acc14105	AT1G69120 | AP1	1.5E-95	10.34	1.13E-17	−39.60	UP_HYPO
Acc14300	AT1G24260 | SEP3	9.9E-123	8.71	1.75E-12	−39.32	UP_HYPO
Acc14601	AT2G45650 | AGL6	3.4E-106	9.14	7.03E-14	−35.05	UP_HYPO
Acc16642	AT2G45660 | SOC1	5.1E-42	2.24	3.56E-39	−33.75	UP_HYPO
Acc18368	AT5G60910 | FUL	1.1E-106	3.28	5.3E-252	−39.91	UP_HYPO
Acc20728	AT4G18960 | AG	9.2E-123	7.89	3.23E-10	−35.76	UP_HYPO
Acc21883	AT5G15800 | SEP1	4.9E-97	10.56	1.99E-18	−36.90	UP_HYPO
Acc22088	AT3G58780 | SHP1	7.3E-46	8.12	7.53E-11	−36.09	UP_HYPO
Acc23601	AT3G54340 | AP3	2.7E-93	5.87	5.81E-20	−34.75	UP_HYPO
Acc24088	AT5G20240 | PI	1.06E-86	6.64	6.58E-07	−36.47	UP_HYPO
Acc26639	AT5G60910 |FUL	1.7E-94	9.42	1.86E-19	−34.02	UP_HYPO
Acc26640	AT3G02310 | SEP2	2.1E-103	8.92	3.52E-13	−34.32	UP_HYPO
Acc27997	AT2G45650 | AGL6	2.5E-43	7.80	1.49E-19	−32.69	UP_HYPO
Acc29300	AT3G02310 | SEP2	1.03E-98	9.43	9.46E-15	−39.66	UP_HYPO
Acc32725	AT1G24260 | SEP3	7.3E-124	10.26	1.94E-17	−38.30	UP_HYPO
Acc32983	AT1G26310 | CAL	3.6E-55	5.10	1.40E-23	−35.78	UP_HYPO
Acc01231	AT1G22130 | AGL104	2.2E-30	−1.26	1.79E-22	−33.92	DOWN_HYPO
Acc05562*	AT5G65070 | MAF4	8E-39	−3.53	0.00E + 00	−31.98	DOWN_HYPO
Acc06844	AT3G57230 | AGL16	6.5E-93	−1.99	3.24E-19	−39.00	DOWN_HYPO
Acc07718	AT5G48670 |AGL80	3.4E-39	−1.55	9.78E-05	−35.51	DOWN_HYPO
Acc08919	AT1G18750 | AGL65	1.9E-50	−1.56	9.99E-29	−35.20	DOWN_HYPO
Acc12070	AT1G77980 | AGL66	4.2E-50	−2.13	1.96E-23	−38.08	DOWN_HYPO
Acc15737	AT2G45660 |SOC1	2.3E-76	−1.85	4.40E-91	−37.25	DOWN_HYPO
Acc24142	AT1G71692 | XAL1	3.06E-78	−1.75	1.62E-32	−28.96	DOWN_HYPO

However, the predominance of CHH hypomethylation in promoter regions was also observed in downregulated DEGs (Fig. 4A, Fig. S6, Table S4) enriched in phosphoprotein phosphatase activity and regulation of transcription and RNA metabolic process (Fig. S7D, E). CHH hypomethylation was detected in downregulated MADS-box genes (Table 1), including an FLC-like shown to be regulated by other epigenetic mechanisms⁵³, although differential methylation of specific C residues in CG context was also detected (Fig. S8). CHH hypomethylation was also prevalent in a set of 518 genes associated with dormancy (Table S5), regardless of their expression (Fig. S9). Such broad-scale hypomethylation from dormancy to growth resumption is therefore unlikely to directly determine the transcriptional activation or suppression of dormancy- and growth-related genes. Instead, it may contribute to genome-wide chromatin reconfiguration, to enable the reprogramming of gene expression for growth.

Reduced TE expression is associated with hypomethylation and accompanied by elevated expression of RdDM pathway genes

CHH methylation serves as a suppressive epigenetic mark for silenced TEs and heterochromatin⁷⁰. The widespread CHH hypomethylation observed during kiwifruit growth resumption indicates a fluctuation in chromatin dynamics and we hypothesized that it may preferentially regulate TE expression during transition to growth and reproduction. To investigate the methylation profiles of TEs, the methylation rates of the 2-kb flanking regions and the body of TEs were averaged for all TE loci (Fig. 6A) and individual TE superfamilies (Fig. S10). The results showed TE methylation in all contexts, with the highest methylation levels in CG, followed by CHG contexts, but methylation rates of both contexts were indistinguishable between June and August samples. However, CHH methylation of all TE superfamilies was reduced by half in August compared with June (Fig. 6A, Fig. S10), concurrent with the genome-wide hypomethylation (Fig. 3A). Surprisingly, the CHH hypomethylation in TEs was not accompanied by TE transcriptional activation; instead, the expression level of most of the TE superfamilies decreased in August, with the expression changes in Copia LTR-retrotransposons, Mutator DNA transposons, MITE elements derived from hAT, and Helitron rolling circle transposons reaching two-fold reductions (Fig. 6B).

Fig. 6.

TE activity in association with genes involving in TE silencing. (A). Average methylation level of the 2-kb flanking regions and TE annotation regions (from 5’ to 3’ annotation boundaries) of all annotated TE loci in kiwifruit dormant buds (June; JUN) and at growth resumption (August; AUG). (B). Heatmap of the log₂fold-change (log₂FC) and scaled expression level (JUN1-3 and AUG1-3) of TE superfamilies. TEs that were significantly downregulated in August compared with June are labelled with stars (*). (C). Heatmap of the log₂fold-change (log₂FC) and scaled expression level (JUN1-3 and AUG1-3) of genes participating in chromatin remodeling, PTGS, and RdDM pathways. Genes that were significantly upregulated in August compared with June are labelled with stars (*).

To examine this unexpected finding, genes that may be involved in silencing of TEs were interrogated. Our results showed that several genes, particularly those involved in post-transcriptional gene silencing (PTGS) and canonical and non-canonical RdDM pathways were significantly upregulated in August bud samples (Fig. 6C), including DDM1 homologue (Acc00983), DICER genes DCL2/3 (Acc20822, Acc12355), the AGO2/4/10 (Acc21787, Acc10653, Acc04231) genes and RDR6 (Acc15429). In addition, kiwifruit homologs of PINHEAD1 (PNH1; Acc23200, Acc13118), as well as MEIOSIS ARRESTED AT LEPTOTENE1 (MEL1; Acc30098), were also upregulated (Fig. 6C). These findings suggest that chromatin remodeling, PTGS and RdDM pathways might work synergistically to prevent TE activation under CHH hypomethylation circumstances at resumption of shoot growth in kiwifruit, while the upregulation of PNH1 and MEL1 implicates their roles in reconfiguration of gene expression by interaction with small RNAs.

Discussion

In this study, we provide multiple lines of evidence demonstrating the role of differential DNA methylation in kiwifruit bud dormancy and shoot growth. Advanced shoot emergence with chemically induced hypomethylation, expression of key methylase and demethylase genes and analysis of genome-wide DNA methylation all support the proposition that DNA methylation correlates with dormancy, and reduced methylation correlates with growth. Differential methylation in kiwifruit buds at different phenological stages has some parallels with the dynamic reconfiguration during seed development and germination^14,15. Although very different in appearance and tissue types, the roles of buds and seeds are aligned. They both protect precious embryos during unfavorable conditions (i.e. cold, drought and osmotic stress) and some of the mechanisms that control bud dormancy and budbreak are comparable to regulation of seed dormancy and germination⁷¹. Just as kiwifruit buds accumulate chilling to be able to break buds, seeds often need cold exposure for germination (cold stratification), and many genes and pathways are shared between bud and seed dormancy onset and release⁷². However, while our study emphasizes the internal epigenetic regulation of bud dormancy and reactivation of growth, the role of prolonged exposure to cold in regulation of DNA methylation in kiwifruit is yet to be explored.

Reduced DNA methylation at growth resumption may be associated with dormancy release, or facilitation of shoot outgrowth. Treatment with 5-azaC was effective throughout autumn, before plants reached 100% leaf drop, showing that demethylation could partially substitute for winter chilling requirement, although only under non-chilling conditions. This is comparable to findings in vernalizing annuals⁷³, demonstrating that reconfiguration of DNA methylation serves as a universal regulatory mechanism for growth and flowering. The inability of 5-azaC to induce budbreak in fully dormant buds (at 100% leaf drop) would suggest that hypo-methylation facilitates growth rather than dormancy release. This is consistent with increasing expression of kiwifruit methyltransferases during winter and reduced transcript accumulation coinciding with budbreak in the field conditions. No such reduction was seen in canes exposed to chilling for up to 4 weeks⁷³, indicating that demethylation correlates with initiation of growth rather than chilling-mediated dormancy release. The steady upregulation of methyltransferase expression during chilling suggested that DNA methylation maintains dormancy and prevents budbreak at unsuitable temperatures. Indeed, chilling accumulation in low- and high-chill varieties of sweet cherry (Prunus avium) was accompanied by increased DNA methylation, predominantly in the CHH context, suggesting a role in response to cold stress, increase in cold tolerance and longer duration of endodormancy, rather than its release⁷⁴. DNA methylation in poplar meristems was also associated with stress response through activation of hormone pathways^47,75. In contrast, a kiwifruit DML demethylase showed increased expression throughout chilling and was further upregulated at budbreak, suggesting that there is active demethylation during accumulation of chilling, consistent with a poplar chilling-dependent DML role in the shift from winter dormancy to a condition that precedes vegetative growth⁴¹.

In kiwifruit, bud dormancy precedes flower development, but the floral fate is believed to be acquired before dormancy, during the active growing season, when undifferentiated meristems are established in the axils of leaf primordia of embryonic shoots in latent buds⁷⁶. Inflorescence development commences with re-establishment of growth in spring⁷⁷, at which time expression of floral genes is detected⁷⁸. The reconfiguration of methylation may underlie or reflect this change in the meristem fate, initiated in the first growing season. This would be comparable to the CHH hyper-methylation associated with vegetative-to-reproductive transition in the SAM in rice²¹. Similarly, the timing and amplitude of DNA methylation determines bolting induction in different sugar beet genotypes²⁰.

DNA methylation represents a major epigenetic mechanism that modulates gene expression⁷⁹. We showed that methylation profiles in all three contexts in kiwifruit genes resemble those described for other flowering plants. As expected, gene expression was positively correlated with gene body methylation in the CG context and negatively correlated with gene body methylation in the CHG and CHH contexts. Demethylation of multiple loci between June and at the time of budbreak (resumption of growth) in August therefore strongly suggested that many of the dormancy and budbreak genes might be regulated through differential methylation. Differential methylation was most prominent in promoter regions, consistent with the concept that methylation acts as a transcriptional repressive mark with the capacity to silence promoter activity⁸⁰. Therefore, hypomethylation of promoters of many genes may be involved in their upregulation at budbreak. However, the inverse correlation between promoter methylation and gene activity was not linear, as evidenced by the observation that the promoters of a significant portion of downregulated genes were also found to be hypomethylated. This is consistent with recent reports that show a much more complex response of promoters to DNA methylation⁸¹, which can explain the largely different findings about methylation and plant developmental processes such as fruit ripening^82,83. Comprehensive analyses of mutant phenotypes³⁶ and targeted manipulation of DNA methylation^22,84 will be instrumental in deciphering the roles and the underlying mechanisms of DNA methylation in gene expression and plant development. Furthermore, this study identifies DMCs in promoter regions of candidate genes, suitable for further study.

TE activity is constantly monitored and suppressed by epigenetic silencing systems, via DNA methylation, histone modification, and small RNA targeting, to facilitate genome stability⁸⁵ and ensure protection of the plant germline²¹. While TEs are often found within the compact heterochromatin regions with dense suppressive epigenetic marks, they can also be located near genes, where they can act as enhancers or repressors, either increasing gene expression in response to environmental or developmental cues, or disrupting gene expression by spreading suppressive epigenetic marks into genes^86–88. Our results showed that TE expression in buds collected during growth resumption was lower than that in dormant kiwifruit buds, concurring with silencing of TEs for the protection of the germline⁸⁹. It also suggested that the prevalent CHH hypomethylation was not largely involved in the activation of TEs acting as gene regulatory elements. Instead, PTGS and RdDM pathway genes⁹⁰ were significantly upregulated in August, compared with June. This observation suggests that the PTGS pathway functions to sequester TE or unwanted transcripts accompanied by CHH hypomethylation, and the RdDM pathway was enhanced to re-establish or maintain transcriptional silencing on TEs during growth resumption from dormancy in kiwifruit buds. Furthermore, upregulation of two ARGONAUTE (AGO) genes, PNH1 and MEL1, important for gene silencing and maintenance of shoot apical meristem^91–93 and acting via interaction with 21-nt phased siRNAs (phasiRNA)^94–96, suggest they may have a role in epigenetic regulation in kiwifruit bud development, providing insight into gene expression reconfiguration during bud dormancy and growth resumption.

In summary, this study demonstrates that DNA methylation is reprogrammed during the transition from winter dormancy to spring growth in kiwifruit (Fig. 7). CHH demethylation precedes the initiation of new differentiation programs, thereby enabling transcriptional reprogramming that underpins phenological changes (Fig. 7). These findings highlight a potential epigenetic regulatory layer controlling transition between dormancy stages, bud break, growth resumption, and subsequent flowering. From an applied perspective, understanding the methylation dynamics underlying dormancy release may provide new strategies to manipulate bud activity through targeted epigenetic interventions. Such approaches could help synchronize bud break, improve flowering consistency, and ultimately enhance yield reliability in commercial kiwifruit orchards. Moreover, the identified DNA methyltransferase and demethylase genes represent promising candidates for breeding programs aiming to develop cultivars better adapted to changing climates and variable winter chilling conditions.

Fig. 7.

A model representing the epigenetic and transcriptional regulation of growth and dormancy in kiwifruit. Epigenetic reprogramming, including a decrease in genome-wide CHH methylation, controls the progression from deep dormancy (endodormacy) to the state when buds regain the ability to grow (ecodormancy) and initiate reproduction. This is followed by transcription factor-mediated activation and repression of genes required for growth and flowering, while differential methylation at specific CG and CHG sites in promoter regions may contribute to differential expression of specific genes. Methylated cytosine residues are presented as red (CHH) or grey (CG, CHG) dots and white dots indicate demethylation. Up- and down-regulation of gene expression is indicated with yellow arrows.

Materials and methods

Plant material and treatments

The study used kiwifruit Actinidia chinensis Planch. var. chinensis ‘Zes008’ (Patent number PP32076), a red-fleshed cultivar typically breaking buds in early spring and flowering in late October in New Zealand. Vines were grown in the field under standard orchard management at the Plant & Food Research orchard near Te Puke, New Zealand (37^o 48’ S, 176° 19’ E). The temperature and chilling accumulation data⁹⁷ were downloaded from MetWatch online (http://www.metwatch.co.nz) and the leaf drop percentage on sampling dates was evaluated in the field. For budbreak assays and 5-azacytidine (5-azaC, Sigma-Aldrich, A2385) treatment, single node cuttings were used as described before⁹⁸. Briefly, 3–4 canes containing ~ 50 axillary buds were collected from three individual plants (biological replicates) on 17th June, 8th July, 29th July and 25th August 2020 and 25th May, 9th June, 16th June and 29th June 2021 and excised into single-node cuttings. The cuttings were immersed in either 250 µmol or 500 µmol of 5-azaC solution, or water (control treatment), respectively, for 48 h in the dark. Then cuttings were placed into a tub containing water and maintained at room temperature under long day conditions (16 h:8 h, light: dark) to force budbreak. The number of days required for visible budbreak was recorded. A minimum of 5 cuttings per biological replicate were used. The remaining buds from the canes collected on 17th June and 25th August 2020 (referred to as June and August, respectively) were sampled into liquid nitrogen and used for RNA and DNA analyses.

Phylogeny

Kiwifruit methyltransferase and demethylase genes were identified using BLASTX with Arabidopsis genes to search the annotated protein sequences in the A. chinensis Red5 genome⁹⁹. Sequence alignment of kiwifruit protein sequences was performed using Geneious (Biomatters, Auckland, New Zealand) (https://www.geneious.com) MUSCLE alignment and the phylogenetic tree were built with Geneious tree builder, using the Neighbor-Joining method, with 1000 bootstrap replicates.

RNA extraction, sequencing, and analysis

Total RNA was extracted using the Spectrum Plant Total RNA kit (Sigma-Aldrich, St. Louis, MO, USA). Library construction, sequencing and initial bioinformatic analyses (Illumina NovaSeq 6000, with paired-end 150-bp reads) were performed at Novogene (www.en.novogene.com). Over 40 million unique reads per sample mapped to the A. chinensis Red5 genome⁹⁹ were interrogated for expression, which were presented as average fragments per kilobase of transcript per million reads (FPKM) ± SE of three biological replicates. Differentially expressed genes (DEGs) were identified using DESeq2¹⁰⁰ and an adjusted p-value < 0.05 by Novogene. Additional filtering was performed to identify genes with FPKM ≥ 1 and log₂foldchange ≥ 1. The heatmaps of DEGs for Fig. 2 and Fig. S2 were created using TBtools v.2. 030¹⁰¹ and Morpheus (https://software.broadinstitute.org/morpheus), whereas the heatmaps for all DEGs associated with DMR (Fig. S6c) and genes involving in chromatin remodeling, PTGS and RdDM pathways (Fig. 5c) were created with the scaled FPKM value and the logarithmically transformed fold changes acquired from DESeq2 using the R package ComplexHeatmap¹⁰². The gene ontology (GO) analysis was conducted using the software BiNGO¹⁰³. Significantly enriched/over-represented GO terms (p < 0.05) were indicated as yellow/orange circles in network graphs.

To conduct TE analysis, TE annotation was acquired using EDTA v1.9.3¹⁰⁴. To analyze TE expression, raw read count mapping to TEs was analyzed using TEtranscripts¹⁰⁵ before transforming to reads per million mapped reads (RPM). The heatmap presented in Fig. 5b was prepared using ComplexHeatmap¹⁰².

Genomic DNA extraction and BS-seq

The bud samples collected on 17th June and 25th August 2020 were used for genomic DNA extraction using the DNeasy plant Mini Kit (Qiagen, Hilden, Germany) following the manufacturer’s instruction. Sodium bisulfite treatment, library construction and sequencing (Illumina NovaSeq 6000, 150-bp paired-end reads) were performed at Novogene (www.en.novogene.com). A total of 1328 million 150-bp paired-end reads were generated from sodium bisulfite-treated DNA with over 99% bisulfite conversion rate for each library. On average, about 209 million sequencing reads were aligned to the A. chinensis kiwifruit genome for each sample.

Methylation analysis

Adapters of all raw reads and the first 10 bases of the reverse reads were trimmed using the software fastp¹⁰⁶. Poly-G tails were also removed Chen, et al¹⁰⁶.. The parameters used with fastp were as follows: --detect_adapter_for_pe --trim_front1 = 0 --trim_front2 = 10 --trim_poly_g. Afterwards, reads were aligned against the reference genome of A. chinensis Red5 genome⁹⁹ using Bismark (v0.23.0)¹⁰⁷ with settings -N 1 -p 8 -X 1000 --score_min L,0,−0.6 -un --ambig_bam. Bismark’s deduplicate_bismark tool was used for removal of PCR duplicates. Cytosine methylation status of all cytosine contexts (CG, CHG, and CHH) was calculated by bismark_methylation_extractor and cytosine reports were prepared using Bismark’s coverage2cytosine.

Identification of differentially methylated cytosine (DMC) sites and regions (DMRs) was performed to compare the June and August samples using MethylKit¹⁰⁸. All cytosine contexts were included for DMC and DMR analyses, with minimum read coverage set to 10 reads for DMC and four reads for DMR analysis. The tile for DMR identification was set to 100 bp, with 100 steps and at least four covered bases. Reads covering different strands were kept separate. A minimum of 25% methylation difference and an adjusted p-value ≤ 0.01 were required for DMC and DMR discovery. The location of DMRs was annotated using the R packages MethylKit and Genomation¹⁰⁹ with Red5 gene annotation⁹⁹. DMRs were binned into four location categories: promoter (within 2 kb upstream of the gene annotation), exon, intron, and intergenic. The circos plot showing DEG and DMR densities across the kiwifruit genome was generated by BioCircos¹¹⁰ in R (https://github.com/lvulliard/BioCircos.R), and the Manhattan plots presented in Fig. S5 were created using ggplot2^111,112. Genes ranked by expression in association with methylation level were analyzed and illustrated as Fig. S4 using MethGET¹¹³. Analysis of the average methylation of the 2-kb flanking regions and gene body, as well as the methylation profile of the 2-kb flanking regions of TEs and the body of TEs, was conducted and illustrated using the scripts published by Bird and colleagues¹¹⁴. Bismark’s cytosine reports were queried to extract gene specific regions. Average methylation site scores from three replicates across methylation contexts and time points were calculated and displayed using ggplot2^111,112 alongside genomic features in the region⁹⁹.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

RW and EVG conceived and designed the experiments. RW and THC performed the experiments. RW, THC, CV and EVG performed the data analysis. RW, THC, CV and EVG wrote the manuscript. ACA approved the final version of the manuscript.

Funding

This work was funded by the New Zealand Ministry of Business, Innovation and Employment (grant C11X2101) and Plant & Food Research internal funding.

Data availability

The datasets generated and analysed during the current study are available in the NCBI BioProject repository ID PRJNA1348359.

Declarations

Consent for publication

All authors read and approved the final manuscript.

Competing interests

The authors declare no competing interests.

Ethics approval

This study did not involve human participants, human data, or human tissue, therefore ethics approval and consent to participate were not required.