Differentially methylated genomic regions of lettuce seeds relate to divergence across morphologically distinct horticultural types

Abstract

Heritable cytosine methylation plays a role in shaping plant phenotypes; however, no information is available about DNA methylation in cultivated lettuce (Lactuca sativa), one of the most important leafy vegetables. Whole-genome bisulfite sequencing (WGBS) performed on seeds of 95 accessions from eight morphologically distinct horticultural types (Batavia, butterhead, iceberg, Latin, leaf, oilseed, romaine and stem) revealed a high level of methylation in lettuce genome with an average methylation of 90.6 % in the CG context, 72.9 % in the CHG context and 7.5 % in the CHH context. Although WGBS did not show substantial differences in overall methylation levels across eight horticultural types, 350 differentially methylated regions (DMR) were identified. Majority of the 41 pivotal DMR overlapped with genomic features predicted or confirmed to be involved in plant growth and development. These results provide the first insight into lettuce DNA methylation and indicate a potential role for heritable variation in cytosine methylation in lettuce morphology. The results reveal that differences in methylation profiles of morphologically distinct horticultural types are already detectable in seeds. Identified DMR can be a focus of the future functional studies.

Lettuce is the most economically important leafy vegetable cultivated in moderate climates worldwide. Whole-genome bisulfite sequencing (WGBS) performed on seeds of 95 accessions from eight morphologically distinct horticultural types (Batavia, butterhead, iceberg, Latin, leaf, oilseed, romaine and stem) revealed a high level of methylation in lettuce genome. These results provide the first insight into lettuce DNA methylation and indicate a potential role of heritable variation in cytosine methylation in lettuce morphology. The results reveal that differences in methylation profiles of morphologically distinct horticultural types are detectable already in seeds.

Introduction

Cytosine methylation is an epigenetic modification that influences plant development through silencing of transposable elements (TE), the formation of heterochromatin and regulation of gene expression (Kawakatsu et al. 2016). In plants, cytosine methylation can occur at CG, CHG and CHH sites, where H represents A, C or T nucleotide. It has been hypothesized that the short-time methylome evolution is caused mainly by spontaneous epimutations due short-term environmental stresses, while the long-time methylome evolution is predominantly driven by repeated historical stress and that this type of methylation is associated with genomic changes (Vidalis et al. 2016; De Kort et al. 2020). Analyses of association between morphological variation and epigenetic markers (Róis et al. 2013; Roy et al. 2015; Ma et al. 2018) together with the Mendelian inheritance pattern of these markers (Yi et al. 2010) indicate the involvement of heritable epigenetic changes in shaping variation of phenotypes in natural populations (Cubas et al. 1999; Róis et al. 2013; Wilschut et al. 2016; Browne et al. 2021) and cultivated species (Roy et al. 2015; Ma et al. 2018; Avramidou et al. 2021). Traits controlled by heritable epigenetic changes include but are not limited to plant, leaf and flower morphology (Cubas et al. 1999; Róis et al. 2013; Roy et al. 2015; Ma et al. 2018; Avramidou et al. 2021; Browne et al. 2021), composition (Yi et al. 2010), and disease resistance (Browne et al. 2021). Comparison of methylation patterns in cultivated maize (Zea mays) and its wild ancestor teosinte (Z. mays ssp. parviglumis) revealed that variation in DNA methylation has changed and that these changes may be related to domestication (Xu et al. 2020). Epigenomes of plants are highly diverse; in some species (e.g. Prunus mume, Jatropha curcas) epigenetic diversity is larger than genetic diversity (Yi et al. 2010; Ma et al. 2018), while in others (e.g. sweet cherry, Prunus avium) genetic diversity is greater (Avramidou et al. 2021). It needs to be pointed out, however, that several of these studies used only low-resolution techniques (e.g. methylation-sensitive amplified polymorphism—MSAP) and may not have covered completely the genetic variation of the genome.

Analyses of the whole-genome bisulfite sequencing (WGBS) data revealed an extensive DNA methylation variation throughout angiosperms (Niederhuth et al. 2016); however, information is lacking about methylation pattern in lettuce (Lactuca sativa), the most popular leafy vegetable produced mostly in moderate climates worldwide. L. sativa is the only cultivated species from the genus Lactuca that includes around 100 species, most of which are indigenous to Asia and Africa. Results of RNA sequencing indicate that the single domestication event of lettuce occurred about 10 800 years ago in the Fertile Crescent, and that L. serriola (prickly lettuce) is the progenitor of the cultivated lettuce (Zhang et al. 2017). Domestication was accompanied by the loss of seed shattering (Wei et al. 2021) and an absence of spines along the leaf midvein.

Cultivated lettuce displays an extensive variation in colour, shape and texture (Křístková et al. 2008) that makes this vegetable an ideal model to study the effect of differences in genome methylation on morphological divergence. Based on the morphological characteristics, lettuce cultivars are commonly divided into several horticultural types and subtypes (Simko et al. 2014): Batavia, butterhead, iceberg, Latin (grassé), leaf (looseleaf, cutting), oilseed, romaine (cos) and stem (stalk). Stem type is cultivated mainly for edible stems, while the oilseed type is used to produce cooking oil from its relatively large seeds. Both stem and oilseed-type plants develop long stems without forming rosettes characteristic for other types of lettuce. Rosette-forming types of lettuce are mainly grown for edible leaves that are consumed raw or minimally processed. Butterhead cultivars produce small spherical heads with pliable leaves and oily texture. Romaine cultivars exhibit leaves that are much longer than wide and are typically upright to form an elongated head. Latin types produce quite small heads with upright leaves that are somewhat longer than wider; however, their leaf texture is softer than that of romaine. Leaf-type lettuce is the most variable, with a large diversity in leaf colour, shape, size, texture, margin and blistering combinations. Iceberg and Batavia cultivars are frequently grouped together into crisp (crisphead) type that is known for its thick, crisp leaves that cup and form a spherical head. Heads of iceberg type are large, dense and tight, while those of Batavia are smaller and softer. All rosette-forming types of lettuce produce ca. 50–150 cm long stems (bolting) when the plant life cycle transition from vegetative to reproductive stage. The separation of stem type from its ancestral cultivated lettuce occurred about 900 years ago and those of butterhead, Batavia and romaine about 500 years ago (Zhang et al. 2017). The modern group of late-bolting iceberg cultivars was developed only in 1940s.

The objectives of the current study were to (i) determine the frequency of cytosine methylation in lettuce genome, (ii) compare methylation levels in seeds of eight morphologically distinct types of lettuce, (iii) examine the existence of differentially methylated regions (DMR) at morphologically distinct types, and (iv) if DMR were detected, identify predicted functions of genomic features at DMR.

Results and Discussion

To investigate the effect of genome-wide differences in methylation pattern on lettuce morphological divergence, WGBS was performed on a set of 95 accessions [see Supporting Information—Fig. S1] from eight morphologically distinct types (Batavia = 11, butterhead = 10, iceberg = 11, Latin = 7, leaf = 16, oilseed = 3, romaine = 27 and stem = 10). Horticultural types of lettuce show a very large variation in the plant and leaf morphology and the rate of their development (Sthapit Kandel et al. 2020). It was evidenced that such differences in leaf physiological age and the rate of development can affect methylation profiles (Bartels et al. 2018) as these change during plant growth (Zhang et al. 2018). To determine whether differences in methylation profiles of horticultural types can be detected already in seeds, the current study focused on analysing seeds stored under identical environmental conditions. Plants of all accessions were cultivated for at least three growing cycles in identical environmental conditions and seeds were stored at the same facility to minimize differences in methylation due to dissimilar environmental conditions. Across all types, the average genome-wide cytosine methylation levels were 90.6, 72.9 and 7.5 % in CG, CHG and CHH sites [see Supporting Information Figs S2–S4], respectively, which is similar to previous estimates in maize (Z. mays) (84.8 % CG, 73.2 % CHG and 5.8 % CHH), lower than in sugar beet (Beta vulgaris) (92.5 % CG, 81.2 % CHG and 18.9 % CHH), but substantially higher than in Arabidopsis thaliana (30.4 % CG, 9.9 % CHG and 3.9 % CHH) (Niederhuth et al. 2016) [see Supporting Information—Fig. S5]. The high methylation level in L. sativa was expected because of its relatively large size genome (2.5 Gb) and a high repeat content (Reyes-Chin-Wo et al. 2017), both of which are positively associated with DNA methylation level in angiosperms (Niederhuth et al. 2016). The fractions of CG, CHG and CHH contexts in the total number of methylated cytosine sites were 44.0, 34.2 and 21.8 % (Fig. 1), respectively, the values almost identical to that found in switchgrass (Panicum virgatum) (44, 35 and 21 %) (Yan et al. 2018) (Fig. 2).

Figure 1.

Fraction of methylated cytosines according to three sequence contexts. Genomes of 95 lettuce accessions from 8 horticultural types.

Figure 2.

Fractions of methylated cytosines for plant species. Ternary graph shows proportion of the genome methylated according to three sequence contexts. Data for 10 species were compiled by Tong et al. (2021). Lettuce data are the mean values of all 95 accessions tested in the current study (differences across lettuce accessions were minimal).

No significant difference was found in the number of methylated sites across different horticultural types except for a higher (P = 0.037) number of methylated sites in iceberg when compared to stem lettuce. Twenty-eight pairwise comparisons between eight horticultural types revealed 2361 differentially methylated cytosine sites (differentially methylated position—DMP, Fig. 3, Supporting Information—Data S8), 1281 of them detected in more than one paired comparison. The number of DMP between two types ranged from 49 DMP between iceberg and Batavia to 1025 DMP detected between iceberg and oilseed lettuce [see Supporting Information—Fig. S6]. Across all horticultural types of lettuce, 34.4 % of DMP were located in promoter regions, 0.9 % in exons, 3.0 % in introns and 61.7 % in intergenic regions. When the frequency of hypermethylated versus hypomethylated DMP was compared across eight types, the highest proportions of hypermethylated sites were observed in butterhead, Latin and stem types [see Supporting Information—Fig. S7, bottom panel].

Figure 3.

Differentially methylated position (DMP) and regions (DMR) in lettuce genome. Concentric circles show (from outer to inner) the size and the name of each chromosome, positions of DMP (lines), positions of DMR (points) and the absolute difference in methylation levels between pairs of horticultural types at each DMR (difference is shown if ≥25 %, the grey line indicates 50 % difference), and positions of pivotal DMR (bars). Detailed data for all pairwise comparisons between horticultural types are in Supporting Information—Data S8–S10.

Although the genome-wide methylation level was similar in all horticultural types, 350 DMR were identified (Figs 3 and 4, Supporting Information—Data S9); 169 of them detected in multiple paired comparisons between two horticultural types. Differentially methylated regions were located in promoter regions (43.2 %), exons (4.4 %), introns (2.3 %) and intergenic regions (50.9 %). When methylation levels at DMR were compared across types, significantly higher number of hypermethylated DMR were identified in butterhead and oilseed types, while the highest frequency of hypomethylated DMR were found in iceberg lettuce (Fig. 5). More stringent specifications of methylation differences identified 41 pivotal DMR (PDMR, Fig. 3, Supporting Information—Data S10) that had at least one methylation difference ≥50 %, P-value < 0.001 and were detected in five or more pairwise comparisons. Dendrogram based on methylation levels at PDMR (Fig. 6) matched well with morphological differences and known historical divergence of horticultural types in lettuce (Zhang et al. 2017; Wei et al. 2021) as was observed in other species such as maize (Roy et al. 2015).

Figure 4.

Differentially methylated regions (DMR) detected at pairwise comparisons. Columns on the left show the pair of compared horticultural types (Rom—romaine, Ice—iceberg, But—butterhead, Leaf—leaf, Bat—Batavia, Lat—Latin, Stem—stem and Oil—oilseed) and the number of identified DMR. Proportion of DMR overlapping with genomic features are shown in the left panel. Proportion of hyper- and hypomethylated DMR are shown in the right panel. If the ratio of hyper- and hypomethylated DMR is different from 1:1 (at experiment-wise α level of 0.05), the horticultural type with significantly more hypermethylated regions is named.

Figure 5.

Differentially methylated regions (DMR) found in eight horticultural types. The number of DMR detected for each horticultural type (top panel). Note that a DMR can be detected at several pairwise comparisons; therefore, the sum of DMR from all types is larger than the total number of unique DMR (350). Proportion of DMR overlapping with genomic features (middle panel). Proportion of hyper- and hypomethylated DMR (bottom panel). If the ratio of hyper- and hypomethylated DMR is different from 1:1 (at experiment-wise α level of 0.05), arrows indicate whether there is excess of hypermethylated (white arrow) or hypomethylated (black arrow) regions at the type.

Figure 6.

Dendrograms based on pivotal differentially methylated regions (PDMR) and SNP linked to PDMR. The neighbour joining algorithm was applied to produce dendrogram from methylation levels at each PDMR (left panel) and from SNP linked to PDMR (right panel). Bootstrap values from 1000 cycles of resampling are given at branching points.

Functions were predicted for 28 out of 41 PDMR [Supporting Information—Data S10]. Most of the predicted proteins are involved in plant growth and developmental processes, such as strengthening the primary cell wall (extensin-2), flowering and lignin synthesis (lipid transfer protein EARLI 1), shoot apical meristem formation during embryogenesis (homeobox protein SHOOT MERISTEMLESS [STM]), early floral meristem identity (MADS-box), embryogenesis, oogenesis and flowering (WUSCHEL), rapid growth (kiwellin-1), cell wall restructuring during ageing (xyloglucan endotransglucosylase/hydrolase 2), auxin-activated signalling (auxin-induced protein 15A), abscisic acid-induced signalling (ricin B-like lectin EULS3), biosynthesis of brassinolide (delta (24)-sterol reductase), and synthetizing secondary metabolites that function as signals for plant growth and development (cytochrome P450 genes).

To determine a potential cofounding effect of genetic variants on morphological differences among horticultural types, single nucleotide polymorphism (SNP) variations were identified in the two 1 kb genomic areas flanking PDMR (Xu et al. 2020). A significant SNP—horticultural type association was detected in the genomic areas of 18 out of 34 mapped PDMR [see Supporting Information—Data S10]. However, additional SNP with a significant effect on morphological differences may be potentially found if the search were extended over a larger upstream and downstream areas. Though the current study found that 47 % of PDMR were not located in the proximity of genetic variants significantly associated with morphological differences observed among horticultural types, it certainly does not suggest that methylation is a driving force in morphological divergence as there are a large number of loci across lettuce genome that cause morphological variation (Zhang et al. 2017; Wei et al. 2021). Detailed functional studies are needed to unequivocally determine morphological variations caused by cytosine methylation rather than genetic variants (Yu et al. 2020; Wang et al. 2022). Meantime, analyses of SNP and PDMR can provide a statistical basis for their possible functional involvement.

Two dendrograms produced from PDMR and from SNP in the proximity of PDMR indicate similarity in clustering of stem, oilseed and romaine types, and iceberg with Batavia type (Fig. 6). Butterhead type showed in both dendrograms proximity to Latin type, while leaf type that includes genotypically and morphologically the most diverse set of accessions, did not show a stable relationship with any other horticultural type. These results are highly similar to those obtained with target region amplified polymorphism markers (Simko and Hu 2008) that clearly separated romaine, iceberg and butterhead type accessions, while leaf type accessions were not placed into a distinct cluster. Leaf-type accessions were, however, well separated from both iceberg and romaine accessions similarly as in the present study.

Recent functional analyses of the genomic regions associated with morphological variation in lettuce provided evidence that at least one of the PDMR discovered through statistical analyses in the current study has a biological function in controlling the development of lettuce head and leaf shape. PDMR on Chr. 7 (starts at 18 850 001, Supporting Information—Data S10) overlaps XM_023887058.2, the predicted homeobox protein STM that is needed for establishing and maintaining shoot apical meristem formation during A. thaliana embryogenesis and organ separation (Spinelli et al. 2011). Hypermethylation of the XM_023887058.2 region was observed in both heading (crisp) types of lettuce (iceberg and Batavia). LsKN1, a lettuce homologue of STM located in this PDMR, is necessary for lettuce head formation (Yu et al. 2020). In the heading type lettuce, there is a CACTA-like TE inserted into the first exon of LsKN1 (LsKN1▽). The TE sequences act as a promoter of high expression of LsKN1▽ (Yu et al. 2020). Besides iceberg and Batavia types, the hypermethylation in the XM_023887058.2 region was also observed in non-heading leaf lettuce. This is likely due to the fact that the gene also plays a role in the leaf shape formation. The LsKN1 allele that is upregulated by the insertion of a TE transforms pinnately leaves to palmately lobbed (Wang et al. 2022), the shape observed only in leaf-type lettuce. It is conceivable that other detected PDMR (or DMR) are involved in lettuce morphological diversity such as the one on Chr. 5 (starts at 7 308 001, Supporting Information—Data S10) that encompasses fatty acid desaturase 4 (XM_023893343.2). This enzyme is involved in biosynthesis of polyunsaturated fatty acids, which are a part of lipid metabolism but also alter plant growth. XM_023893343.2 was significantly more methylated in butterhead lettuce than in other types. Leaves of butterhead lettuce are characteristically softer, more pliable, have more oily texture and are less crunchy than leaves of other lettuces.

Previous studies performed on other plant species found that DNA methylation may be associated with variation in plant morphology (Cubas et al. 1999; Róis et al. 2013), including leaf morphology (Ma et al. 2018; Avramidou et al. 2021; Browne et al. 2021). Sequencing of genetic and epigenetic markers linked to leaf phenotypic parameters in P. mume revealed that some of the predicted candidate genes likely play a role in plant development and metabolism (Ma et al. 2018). Additional evidence of DNA methylation being involved in vegetative growth (Bräutigam and Cronk 2018; Zhang et al. 2018; Kumar and Mohapatra 2021) comes, besides others, from maize (Candaele et al. 2014) and Arabidopsis (Yamamuro et al. 2014) studies that identified several developmental genes whose expressions are regulated by methylation.

The current study describes cytosine methylation in seeds of cultivated lettuce. A high level of methylation was observed in all tested accessions with no substantial difference in overall methylation levels across eight horticultural types. Additional analyses identified 350 genomic regions (41 of them pivotal) that were differentially methylated in some horticultural types. Majority of PDMR overlapped with genomic features predicted or confirmed to be involved in plant growth and development. These results suggest that heritable variation in cytosine methylation plays a role in morphological divergence of lettuce horticultural types and that the differences in methylation profiles of horticultural types can be detected already in seeds. Though the current study does not provide biological proof that the detected differences in lettuce genome methylation contribute to functionally relevant morphological variation, it identified the genomic regions that can be a focus of the future functional studies.

Materials and Methods

Plant material and DNA sequencing

The L. sativa diversity panel used in this study comprised 95 lettuce accessions from eight horticultural types (Batavia = 11, butterhead = 10, iceberg = 11, Latin = 7, leaf = 16, oilseed = 3, romaine = 27 and stem = 10). Seeds of each accession were harvested from a single plant grown in a greenhouse, stored for 6 months (dark, low humidity, −30 °C) and used for DNA extraction. The genomic DNA extracted with Qiagen DNeasy Plant Mini Kit was sent out to Genewiz (currently Azenta Life Sciences) for whole-genome sequencing and WGBS using Illumina HiSeq platform. Sequencing quality of raw data was assessed with FASTQC 0.11.9. Only high-quality samples with consistently high Q scores (>30) across the length of the read were retained for further analyses. Sequencing adapters and low-quality bases in raw reads were trimmed using Trim Galore! 0.6.6 with cutadapt 3.1. Cleaned reads were then aligned to the Lsat_Salinas_v7 reference genome (Reyes-Chin-Wo et al. 2017) using Bismark 0.23.1 with bowtie2 2.4.2. Methylation information was extracted from the deduplicated bam file using Bismark 0.23.1. When statistical analyses were performed using all paired comparisons (N = 28) between eight horticultural types, the calculated P-values were adjusted for multiple comparisons, maintaining the experiment-wise α level of 0.05.

Identification of individual, DMP sites

Initial analyses were performed to discover differentially methylated CpG sites between all pairs of horticultural types. All accessions from a pair of horticultural types were used in the analysis using CpG sites with a minimum coverage of eight reads across all samples. Differentially methylated position analysis was performed with methylKit 1.4.1. The logistic regression was employed to calculate DMP P-values (Akalin et al. 2012). The P-values were subsequently corrected for multiple testing with the Sliding Linear Model (SLIM) method (Wang et al. 2011) implemented in methylKit package. Differentially methylated position sites with the adjusted experiment-wise P-value ≤ 0.05 and an absolute difference in methylation level between two horticultural types ≥25 % were annotated with Transcription Start Sites (TSS) information from Refseq (https://www.ncbi.nlm.nih.gov/refseq/) (O’Leary et al. 2016). Results of the annotation analyses were then used to calculate the percentage of DMP overlapping with genomic features.

Identification of DMR

To identify DMR that are conceivably more relevant than DMP in a functional biological context (Robinson et al. 2014; Hüther et al. 2022), the genome was tiled with a window size of 1 kb and a step size of 1 kb. The number of methylated and unmethylated CpG sites at the given region was obtained for each accession. The fraction of methylated CpG sites for the given region was pooled across all accessions within each horticultural type. To compare the fraction of methylated CpG sites between two horticultural types, the logistic regression model implemented in methylKit 1.4.1 was applied (Akalin et al. 2012), followed by the SLIM method (Wang et al. 2011) to correct P-values for the multiple hypothesis testing. DMR with the adjusted experiment-wise P-value ≤ 0.05 and an absolute difference in methylation level between two horticultural types ≥25 % were annotated with TSS information from Refseq. Genes were classified to be associated with DMR when the gene body or its 2-kb flanking region overlapped with DMR (Tong et al. 2021). Differentially methylated regions detected in at least 5 out of 28 comparisons between pairs of horticultural types, with at least one adjusted experiment-wise P-value < 0.001 and differential methylation level ≥50 % were assigned to the group of PDMR and used for more detailed analyses. Methylation differences at PDMR were applied to calculate dendrogram of horticultural types using the neighbor joining algorithm of Past v.4.01 (Hammer et al. 2001) with 1000 bootstrap cycles. Biological function of predicted proteins at PDMR was obtained from and UniProt (https://www.uniprot.org/) (The_UniProt_Consortium 2023).

Hypermethylated and hypomethylated DMP and DMR

When two horticultural types showed significant difference at a DMP or DMR they were classified as being relatively hypermethylated or hypomethylated at the DMP or DMR. The frequencies of hypermethylated and hypomethylated DMP and DMR found at each horticultural type were compared to 1:1 ratio using the Chi-square goodness of fit test in JMP pro v.15.0.0 (Cary, NC, USA).

Identification of genetic variants in proximity of PDMR

To test the relationship between genetic variants that were positioned in the proximity of PDMR and horticultural types, SNP located 1 kb (Xu et al. 2020) upstream and downstream of each PDMR were identified. At each SNP, the association between alleles and horticultural types was tested using the Fisher’s exact probability test (experiment-wise α level of 0.05). The alleles at these SNP were also used to produce a genetic dendrogram with Past v.4.01 (Hammer et al. 2001) as described for PDMR.

Supporting Information

The following additional information is available in the online version of this article –

Figure S1 . Cytosine fraction in genomes of 95 lettuce accessions from eight horticultural types.

Figure S2. Methylation level at CG sites of 95 lettuce accession from eight horticultural types.

Figure S3. Methylation level at CHG sites of 95 lettuce accession from eight horticultural types.

Figure S4. Methylation level at CHH sites of 95 lettuce accession from eight horticultural types.

Figure S5. Non-metrical multidimensional scaling (NMDS) plot calculated from methylation levels of 39 plant species. Calculation is based on methylation levels at CG, CHG, and CHH sites detected in the current study and those compiled by Niederhuth et al. (2016) and Tong et al. (2021). AL: Arabidopsis lyrata, AT: Arabidopsis thaliana, ATr: Amborella trichopoda, BD: Brachypodium distachyon, BO: Brassica oleracea, BR: Brassica rapa, BV: Beta vulgaris, CC: Citrus clementina, CR: Capsella rubella, CS: Cannabis sativa, CSa: Camellia sinensis var. assamica, CSs: Camellia sinensis var. sinensis, CuS: Cucumis sativus, EG: Eucalyptus grandis, ES: Eutrema salsugineum, FV: Fragaria vesca, GM: Glycine max, GR: Gossypium raimondii, LJ: Lotus japonicus, LS: Lactuca sativa, MD: Malus domestica, ME: Manihot esculenta, MG: Mimulus guttatus, MT: Medicago truncatula, OS: Oryza sativa, PA: Picea abies, PH: Panicum hallii, PP: Prunus persica, PT: Populus trichocarpa, PV: Panicum virgatum, PVu: Phaseolus vulgaris, RC: Ricinus communis, SB: Sorghum bicolor, SL: Solanum lycopersicum, ST: Solanum tuberosum, SV: Setaria viridis, TC: Theobroma cacao, VV: Vitis vinifera, ZM: Zea mays.

Figure S6. Differentially methylated position (DMP) detected at pairwise comparisons. Columns on the left show the pair of compared horticultural types (Rom – romaine, Ice – iceberg, But – butterhead, Leaf – leaf, Bat – Batavia, Lat – Latin, Stem – stem, and Oil – oilseed) and the number of identified DMP. Proportion of DMP overlapping with genomic features are shown in the left panel. Proportion of hyper- and hypomethylated DMP are shown in the right panel. If the ratio of hyper- and hypomethylated DMP is different from 1:1 (at experiment-wise α level of 0.05), the horticultural type with significantly more hypermethylated CpG sites is named.

Figure S7. Differentially methylated position (DMP) sites found at eight horticultural types. The number of DMP detected for each horticultural type (top panel). Note that a DMP can be detected at several pairwise comparisons therefore the sum of DMP from all types is larger than the total number of unique DMP (2,361). Proportion of DMP overlapping with genomic features (middle panel). Proportion of hyper- and hypomethylated DMP (bottom panel). If the ratio of hyper- and hypomethylated DMP is different from 1:1 (at experiment-wise α level of 0.05), arrows indicate whether there is excess of hypermethylated (white arrow) or hypomethylated (black arrow) CpG sites.

Data S8. Differentially methylated positions (DMP).

Data S9. Differentially methylated regions (DMR).

Data S10. Pivotal differentially methylated regions (PDMR).

Phenome, Genome and Environment. Chief Editor: Colleen Doherty

Contributions by the Author

I.S.: conceptualization, methodology, software, validation, formal analysis, investigation, resources, data curation, writing–original draft, writing–review & editing, visualization, supervision, project administration, funding acquisition.

Conflict of Interest Statement

The author declares no conflict of interest.

Data Availability

The data underlying this article are available in the article and in its online Supporting Information.