Consistent and Heritable Alterations of DNA Methylation Are Induced by Tissue Culture in Maize

Abstract

Plants regenerated from tissue culture and their progenies are expected to be identical clones, but often display heritable molecular and phenotypic variation. We characterized DNA methylation patterns in callus, primary regenerants, and regenerant-derived progenies of maize using immunoprecipitation of methylated DNA (meDIP) to assess the genome-wide frequency, pattern, and heritability of DNA methylation changes. Although genome-wide DNA methylation levels remained similar following tissue culture, numerous regions exhibited altered DNA methylation levels. Hypomethylation events were observed more frequently than hypermethylation following tissue culture. Many of the hypomethylation events occur at the same genomic sites across independent regenerants and cell lines. The DNA methylation changes were often heritable in progenies produced from self-pollination of primary regenerants. Methylation changes were enriched in regions upstream of genes and loss of DNA methylation at promoters was associated with altered expression at a subset of loci. Differentially methylated regions (DMRs) found in tissue culture regenerants overlap with the position of naturally occurring DMRs more often than expected by chance with 8% of tissue culture hypomethylated DMRs overlapping with DMRs identified by profiling natural variation, consistent with the hypotheses that genomic stresses similar to those causing somaclonal variation may also occur in nature, and that certain loci are particularly susceptible to epigenetic change in response to these stresses. The consistency of methylation changes across regenerants from independent cultures suggests a mechanistic response to the culture environment as opposed to an overall loss of fidelity in the maintenance of epigenetic states.

CLONAL propagation of plants and animals is expected to produce individuals identical to the donor. However, this is often not the case. Plant tissue culture involves dedifferentiation, or return to a “stem-cell-like” state, which involves dynamic reprogramming at the chromatin level to induce the formation of callus. Subsequently, proliferating cells start to redifferentiate when specific changes in the balance of growth regulators are introduced in the culture medium, ultimately leading to organogenesis or regeneration into whole plants (Grafi et al. 2011; Miguel and Marum 2011). This process represents a traumatic stress to plant cells and often provokes an array of genetic and epigenetic instabilities that are somatically and meiotically heritable (Phillips et al. 1994; Kaeppler et al. 2000). This suite of molecular and phenotypic phenomena is collectively termed somaclonal variation (Larkin and Scowcroft 1981).

At the cellular and molecular levels, somaclonal variation is composed of chromosome rearrangements, polyploidy and aneuploidy, DNA sequence changes, activation of quiescent transposable elements (TEs), and epigenetic variation reflected by altered DNA methylation patterns (Karp and Maddock 1984; Brettell et al. 1986; Dennis et al. 1987; Peschke et al. 1987; Armstrong and Phillips 1988; Brown et al. 1991; Kaeppler et al. 2000). Our current study focuses on genome-wide changes in cytosine methylation.

Direct evidence for heritable epigenetic changes in tissue culture regenerants and their progenies has come from studies on cytosine methylation patterns. RFLP and AFLP technologies have been used to scan methylation patterns following tissue culture in numerous plant species (Kaeppler and Phillips 1993; Jaligot et al. 2002; Bednarek et al. 2007; Schellenbaum et al. 2008; Rodriguez Lopez et al. 2010). Additionally, in Arabidopsis, methylation levels of chromosome 4 were profiled using McrBC-digested DNA on a tiling array, and it was reported that cell suspension culture has a different epigenomic profile compared to wild-type plants, such that certain TEs become hypomethylated and certain genes become hypermethylated (Tanurdzic et al. 2008). In contrast, changes in DNA methylation of tissue culture regenerated rice plants are frequently hypomethylation events that are heritable for several generations and are sometimes correlated with changes in gene expression of nearby genes (Stroud et al. 2013).

Heritable epialleles have been characterized in progeny of regenerated plants. For instance, epialleles of the P locus, which controls cob color in maize, were associated with reversible hypermethylation of the second intron (Rhee et al. 2010). Interestingly, a naturally occurring epiallele at P has also been reported (Sekhon et al. 2007), indicating that similar epigenetic responses to stress that occur in tissue cultured cells may also occur in plants grown in the field.

To investigate the effect that tissue culture processes have on DNA methylation patterns in regenerated maize plants and their progenies, we generated genome-wide maps of DNA methylation patterns, of callus, regenerated plants, and progeny of regenerated plants using meDIP. We determined that methylation changes in tissue culture are frequent, and a portion of these DNA methylation changes is heritable through sexual generations.

Materials and Methods

Tissue culture and regeneration

Immature zygotic embryos were harvested from donor plants of maize genotype A188 ears, 10–12 days after pollination (DAP). The culture process was as described by Frame et al. (2011) with minor modifications. Briefly, ears were surface sterilized for 20 min in 50% (v/v) commercial bleach and sterilized deionized water plus one drop of Tween 20 and rinsed in sterilized deionized water three times, 5 minutes per rinse. Immature embryos were harvested with a flame-sterilized spatula in a sterilized, ethanol-washed fume hood, and plated, embryo axis-side down, onto Chu’s N6 initiation/maintenance media (Chu et al. 1975) with modifications to initiate calli formation (Supporting Information). Cultures were grown at 28 ± 1° in darkness for a total of 6 months from the date of embryo plating with regular subculture approximately every 2 weeks; embryogenic callus was selected at each subculture to maintain the cultures. After 6 months in culture, embryogenic calli were selected and subcultured on modified R1 media for 2 weeks at 28 ± 1° in darkness, and were subsequently transferred to modified R2 media for 2 weeks at 28 ± 1° under a 16-hr photoperiod for plant regeneration (Supporting Information). Individual plantlets were transplanted into magenta boxes filled with R2 media for ∼10 days until viable roots were established. Plants were then removed from the media, rinsed with tap water, and transplanted into degradable peat pots in the greenhouse and grown under standard conditions.

Plant materials and DNA isolation

Maize plants (A188 genotype) were grown at the Walnut Street Greenhouse (University of Wisconsin, Madison, WI). The growing conditions were 27°-day and 24°-night temperatures with 15 hr light and 9 hr dark (6:00 am to 9:00 pm). Maize plants were watered daily as needed. Callus used for DNA extraction was collected from each cell line prior to plating onto R₁ media after 6 months of culturing. The uppermost flag leaf of R₀ plants and the 3rd leaf of R₁ plants were harvested for DNA extraction to conduct meDIP-ChIP profiling. All tissues were immediately flash frozen in liquid N₂. DNA was isolated using a modified CTAB method (Saghai-Maroof et al. 1984).

meDIP-ChIP epigenomic profiling and qPCR assays

Whole-genome profiling of DNA methylation was performed to document changes in DNA methylation following tissue culture. DNA was isolated from three different sibling plants of genotype A188 that had not been subjected to tissue culture to serve as controls. Four independent cultures were initiated from immature embryo tissue (10–12 DAP) and multiple primary regenerants were derived from these two cultures and grown in greenhouse conditions to isolate leaf tissue from the same stage as the control plants. DNA methylation levels were profiled by immunoprecipitation of methylated DNA followed by hybridization to an oligonucleotide microarray as described previously (Eichten et al. 2011). Methyl-sensitive qPCR assays were performed as described according to a previous study (Li et al. 2014). Primers used for qPCR are listed in Table S1.

RNA isolation and calculation of RNA-seq expression values

The uppermost ear leaf of R₁ plants was harvested for RNA extraction. RNA was extracted via TRIzol according to the manufacturer’s instructions (Life Technologies, Gaithersburg, MD) and purified with a RNeasy kit (Qiagen, Valencia, CA). mRNA was isolated from ∼5 μg of total RNA, fragmented, converted to cDNA, and PCR amplified according to the Illumina RNA-seq protocol (Illumina, San Diego, catalog no. RS-100-0801). Sequencing was done at the University of Wisconsin Biotechnology Center using the Illumina HiSequation 2000 at the University of Wisconsin Biotechnology Center. Illumina barcodes were used to multiplex the samples. Generated sequence reads were 101-bp paired-end reads. Total number of reads for each library varied from 18 to 58 million reads; information for each sample can be found in Table S2. Sequences are available in the Sequence Read Archive at the National Center for Biotechnology Information (accession no. SRP040690). Sequence reads for each sample were mapped to V2 of the B73 reference pseudomolecules (http://ftp.maizesequence.org/) (Schnable et al. 2009) using Bowtie version 0.12.7 (Langmead et al. 2009) and the splice site aware aligner TopHat version 1.2.0 (Trapnell et al. 2009). The minimum and maximum intron length was set to 5 bp and 60,000 bp, respectively; all other parameters were set to the default values. Gene model annotation was not provided during the read mapping. Normalized gene expression values expressed as fragments per kilobase pair of exon model per million fragments mapped (FPKM) were determined using Cufflinks version 0.9.3 (Trapnell et al. 2010) with the −G and −r option. The maximum intron length was set to 60,000 bp and the quartile normalization option was used. The default settings were used for all other parameters.

Results

Discovery of tissue-culture-induced differentially methylated regions

DNA methylation profiles of maize-embryo-derived callus, regenerated plants, and noncultured control plants were evaluated to assess the frequency and nature of methylation variation induced by the stress of tissue culture. Four independent cell cultures, each originating from an independent immature embryo, were initiated from the maize inbred A188. The independent cultures were named cell lines (CLs) -3, -4, -5, and -8, and multiple R₀ plants were regenerated from each of these cell lines. The ear leaf was collected from six independent regenerated plants from CL-3, one plant from CL-4, six plants from CL-5, and two plants from CL-8. Primary regenerants (R₀ plants) were named with their CL number listed first, followed by hyphens, and then R₀ plant number (e.g., R₀ plant 3-4 is the fourth progeny derived from CL-3). In addition, the corresponding ear leaf was collected from three different control A188 plants, which were sibling plants originating from the same seed source and were not subjected to tissue culture. Whole-genome profiling of DNA methylation was performed using meDIP followed by hybridization to an oligonucleotide microarray (Eichten et al. 2011). A comparison of the DNA methylation profiles for the three control plants did not identify any differentially methylated regions (DMRs) between the sibling plants. However, a comparison of the DNA methylation profiles from plants that had been subjected to tissue culture to the average of the nontissue culture controls identified a number of DMRs.

DMRs were identified by comparing the average DNA methylation level for all regenerants from the same CL with the control samples or by comparing the average methylation level of all regenerants with the average methylation level of the controls as determined by meDIP (Table 1 and Table S3). The scan for DMRs was run separately for chromosome 9 and genome-wide as chromosome 9 had a higher probe density than the rest of the genome. In total, 479 DMRs were discovered in the genome-wide scan, and zero DMRs were detected between the three noncultured control samples. Many of the DMRs exhibit consistent changes in all samples subjected to tissue culture relative to controls (Figure 1, A and B). A greater proportion of DMRs had hypomethylation in the regenerated plants relative to the controls (67% of genome-wide DMRs, 61% of chromosome 9 DMRs) as opposed to hypermethylation events. Plants regenerated from CL-8 and CL-4 exhibited a higher number of DMRs than CL-3 and CL-5. This higher rate may reflect biological differences or may be a technical artifact resulting from lower sampling of these two cell lines and less robust average values for calling DMRs. A separate DMR scan was conducted using the full set of probes for chromosome 9, which provides about four times higher probe density for this chromosome (Figure S1). This analysis with higher probe density identified the same DMRs as the lower resolution genome-wide scan and also identified additional DMRs. These additional DMRs had similar lengths but were identified due to increased statistical power provided by the additional probe density. This suggests that there are likely additional DMRs that are present on other maize chromosomes but not captured in our scan due to probe density.

Table 1. DMRs in primary regenerants.

	Chromosome 9 hypermethylated DMRs	Chromosome 9 hypomethylated DMRs	Genome-wide hypermethylated DMRs	Genome-wide hypomethylated DMRs
Cell line 3 (n = 6)	1	14	30	86
Cell line 5 (n = 6)	1	14	23	75
Cell line 8 (n = 2)	40	58	113	220
Cell Line 4 (n = 1)	11	22	25	120
All R0 plants (n = 15)	0	10	12	80
Nonredundant DMRs	52	81	158	321

Figure 1.

DMRs in plants regenerated from tissue culture. (A and B) Two examples of DMRs are shown. The relative level of DNA methylation is shown for each of the controls (A188 C1-3) and R₀ plants. Black indicates low methylation; red indicates higher levels of DNA methylation. These regions show high levels of DNA methylation in all control plants but show reduced levels of DNA methylation in many of the regenerated plants. The scale for each of the images and the location of the DMR is shown near the top. The tracks near the bottom show the level of methylation in B73 and Mo17 plants and the genes located near the DMRs. (C) Hierarchical clustering of DNA methylation levels for genome-wide DMRs in all plants (n = 479). The level of DNA methylation for each regenerant and control plant relative to the average of the three control plants was calculated and used to perform hierarchical clustering (Ward’s method). The heatmap coloring indicates hypermethylation (red) or hypomethylation (yellow) relative to the control. (D) The proportion of DMRs that were supported in one, two, three, or all four CLs is shown for the hypomethylated and hypermethylated DMRs.

Hierarchical clustering analysis based on DNA methylation levels in each regenerated plant and in the control plants revealed that many of the DMRs exhibit consistent changes in multiple plants and in multiple cell lines (Figure 1C and Figure S1). Overall, there were 92 robust genome-wide DMRs that were identified in all CLs relative to the controls (Table 1), indicating that certain regions of the genome are consistently exhibiting DNA methylation alterations in tissue culture and these are heavily biased toward hypomethylation events. The number of CLs that support each of the DMRs was assessed for all DMRs found in at least one CL (Figure 1D). The DMRs that are hypomethylated in regenerated plants compared to the controls are more frequently observed in multiple CLs (∼75% of DMRs), and over one-third of them are observed in all four CLs (Figure 1D). In contrast, the majority (53%) of the hypermethylated DMRs are only observed in one of the various CLs, indicating that hypermethylated DMRs are comparatively more stochastic events in tissue culture.

A subset of the tissue culture DMRs was selected for methyl-sensitive qPCR validation (Figure S2). Digestion of genomic DNA with the methylation dependent enzyme FspEI followed by qPCR allows for semiquantitative detection of DNA methylation levels. The majority (8/10) of the hypomethylated DMRs could be validated (Figure S2A) while only one of the five hypermethylated DMRs was supported by this analysis (Figure S2B). The lower validation rate for the hypermethylated DMRs could reflect the fact that many of these regions start with low levels of DNA methylation and gain additional levels following tissue culture. However, the methyl-sensitive qPCR approach provides relatively little discrimination between partial and full DNA methylation.

The use of meDIP-array profiling does not allow the discrimination of different types of DNA methylation. Several recent studies have used whole-genome bisulfite sequencing to document genome-wide patterns of DNA methylation in CG, CHG, and CHH sequence contexts in the maize inbred B73 (Eichten et al. 2013; Gent et al. 2013; Regulski et al. 2013). To understand what types of DNA methylation were being affected by tissue culture, we assessed the level of CG, CHG, and CHH methylation levels for the genomic regions that are hypomethylated following tissue culture using the B73 data from Eichten et al. (2013). Approximately one-third (46/144) of the tissue culture hypomethylated DMRs contain high levels of CG and/or CHG methylation in B73 seedling tissue. None of these DMRs had levels of CHH methylation >2%. This suggests that many of the tissue culture hypomethylated DMRs are not active targets of de novo methylation in leaf tissue of B73.

Many tissue culture DMRs are heritable

DMRs could occur in tissue culture and be somatically heritable through regeneration, or alternatively, they could arise during the regeneration process. To test these possibilities, DNA methylation levels in a sample of callus tissue from CL-3 were profiled and included in the clustering (Figure 1C and Figure S3). The callus sample clustered together with samples from regenerated plants of CL-3, suggesting that most of the methylation changes observed in tissue from regenerated plants are already present in callus tissue prior to regeneration.

We next investigated the heritability of the DNA methylation changes observed in the R₀ plants by examining methylation profiles of several R₁ plants, which were generated by self-pollination of the primary regenerants. Whole-genome profiling of DNA methylation levels was performed for two R₁ offspring resulting from the self-pollination of three R₀ plants; plant 3-4 (offspring termed 3-4.1 and 3-4.3), plant 3-7 (termed 3-7.3 and 3-7.7), and plant 5-13 (termed 5-13.11 and 5-13.12). In addition, methyl-sensitive qPCR was used to assess DNA methylation changes at seven of the DMRs in a larger number of R₀ and R₁ plants (Figure 3 and Figure S4). The relative levels of DNA methylation were profiled in 7 control plants, 5 cell lines (callus tissue), 11 additional R₀ plants and 24 R₁ plants that include multiple sibling offspring of 4 different R₀ plants. Clustering of DNA methylation levels at DMRs for the R₁ plants, the parental R₀ plants, and the controls reveals that many of the hypomethylated DMRs observed in R₀ plants are stably inherited in the R₁ progeny (Figure 2A and Figure S3). The majority (∼60% averaged across all 3 R₀ plants tested) of hypomethylated DMRs present in the R₀ plants were detected in both offspring, while a much smaller portion (7% across all 3 R₀’s) of hypermethylated DMRs were stably inherited in both R₁ offspring (Figure 2B). The analysis of DNA methylation levels in the offspring of different R₀ plants provide examples of relatively consistent DMRs that are heritable, as well as examples that show incomplete penetrance and segregation (Figure 2, C and D and Figure 3). Some DMRs, such as DMR622 and DMR609, show highly consistent effects in all R₀ and R₁ plants relative to controls (Figure 2C and Figure 3, A and B). Other DMRs, exemplified by DMR485 (Figure 2B and Figure 3C) exhibit loss of DNA methylation in some samples but not others. In general, the R₀ plants that have reduced DNA methylation levels for this region generate offspring that show similar reduced levels of DNA methylation. However, there are examples of R₀ plants without reduced DNA methylation and some families of R₁ plants appear to exhibit segregation for DNA methylation levels (Figure 3). Combined, these results provide evidence for stable inheritance of many hypomethylation DMRs induced by tissue culture and provide evidence that some of these events may show variable behavior in R₁ siblings that could be explained by heterozygosity for DNA methylation state in some R₀ plants.

Figure 3.

Validation of the heritability of tissue culture-induced DMRs (Panels are example DMRs; A is DMR622, B is DMR609, and C is DMR485). Three of the tissue-culture-induced hypomethylation DMRs that were validated by methyl-sensitive qPCR were further profiled in additional samples (four additional validated DMRs are pictured in Figure S4). For each sample, the value shows the difference in C_t for amplification with and without the methylation dependent enzyme FspEI. Higher values reflect high levels of DNA methylation; low values reflect low methylation. The samples include 11 additional R₀ samples, 24 R₁ samples, five callus samples and seven control plants not subjected to tissue culture. The 24 R₁ samples include four plants siblings each from three families (parent plants R₀ 3-10, R₀ 3-4, and R₀ 3-7) and eight siblings from R₀ 5-13.

Figure 2.

The majority of hypomethylation observed in R₀ plants is heritable to R₁ progeny. DNA methylation patterns are shown in R₁ plants. (A) Hierarchical clustering of the level of DNA methylation for each of the genome-wide DMRs in three R₀ plants and for two R₁ progeny derived by self-pollination for each of these three plants. The values were normalized by comparing them to the average of the three control (nontissue culture) plants. The DNA methylation level for these DMRs in the three controls is also shown. (B) The DMRs that show altered methylation level in each of the R₀ plants relative to the controls was determined. The proportion of these DMRs that were supported by only one of the R₁ plants (gray) or both of the R₁ plants (black) was determined for each of the three families. The hypomethylation DMRs were generally heritable and observed in both progeny. In contrast, the hypermethylation DMRs were not often inherited. (C and D) Several examples of DMRs are shown for the control plants, the three R₀ parents, and two R₁ offspring for each of the R₀ plants. The location of the DMR and scale are indicated near the top of each image and the blue symbols near the bottom indicate genes. The relative level of DNA methylation (black, low; red, high) is shown for each of the samples.

Characterization of tissue culture DMR location and effects on gene expression

The DMRs were characterized to determine whether they were preferentially located near genes or transposable elements (Figure 4). The location of hypo- or hypermethylated DMRs was compared to annotated maize genes (Figure 4A). In addition, we generated a control set of 10,000 randomly generated pseudo-DMRs by selecting sets of adjacent probes that have similar sizes relative to the actual DMRs. Permutation analysis of the random probe sets (100 sets of 500 random probe sets) was used to assess significance for the genomic context of hypo- or hypermethylated DMRs. Both the hypo- and hypermethylated DMRs are depleted for overlapping the 5′ or 3′ end of maize genes (Figure 4A), and only the hypermethylated DMRs are enriched for being located >2000 bp upstream of genes or 3′ of genes (Figure 4A). It is possible that the tissue-culture-induced DMRs reflect processes targeted to TEs. For each DMR, the nearest TE was identified using the MTED annotation of TEs (Schnable et al. 2009). The hypomethylated DMRs are enriched for not having any TEs within 500 bp relative to randomly selected regions (Figure 4B). In contrast, the hypermethylated DMRs are enriched for being located near nonspreading LTR elements (Figure 4B). Additionally, we assessed the biological function and tissue-specific expression patterns for genes located near DMRs. There was no evidence for significant enrichment of specific biological processes or Gene Ontology (GO) terms among the genes located near hypomethylated or hypermethylated DMRs. Similarly, the genes located near DMRs do not show enrichment for particular tissue-specific expression patterns.

Figure 4.

Characterization of genomic context of tissue culture DMRs and misregulated expression of DMR genes. (A) The location of each of the tissue culture DMRs (hypomethylated or hypermethylated) was assessed relative to maize genes. In addition, a set of 10,000 randomly created DMRs was generated using our probe spacing to assess enrichment for localization of DMRs relative to genes. The standard deviation for the random values was determined by calculating the proportion of the random probe sets in each category for 100 subsamples of 500 random sets. The “*” above bars indicates values that are significantly (P < 0.05) different from the values observed for random probes sets. (B) The location of each DMR was compared to the nearest annotated transposable element. The nearest elements were classified as terminal inverted repeat (TIR) elements, LTR_spreading, or LTR_nonspreading. The spreading or nonspreading is based upon the assignments from a previous study in which some families of LTR elements were found to result in spread of DNA methylation to low-copy flanking sequences (Eichten et al. 2012). The error bars and significance (“*”) of the random sets of probes were determined in the same method used in A. (C) The proportion of hypo-DMR genes ≤1 kb from TSSs that have significantly altered expression levels in R₁ progeny relative to the average FPKM of two noncultured control A188 plants (fourfold up/down-regulation). Genes with zero mRNA-seq reads in all samples were removed from analyses.

It is possible that loci epigenetically sensitive to tissue-culture stress also exhibit variability in natural populations. To examine this possibility, tissue culture DMRs were compared to naturally variable DMRs (n = 3720) in 51 maize genotypes that had not been subjected to tissue culture (Eichten et al. 2013). The overlap between the tissue culture DMRs identified in this study and the 3720 DMRs identified as examples of natural variation was significant (P < 0.01, based on a permutation test of random sets of DMRs being compared to the tissue culture DMRs) but limited, as 8.7% of the hypomethylated tissue culture DMRs and 5.7% of the hypermethylated tissue culture DMRs overlap with the natural variation DMRs among different maize genotypes. The DNA methylation levels in other maize inbreds are often more similar to the control A188 plants than the R₀ plants (Figure 1, A and B). The average DNA methylation level in the diverse noncultured genotypes was compared to the average level for the A188 controls and for the R₀ plants for all hypo- or hypermethylated DMRs. Diverse nonregenerated maize lines exhibit average DNA methylation levels that are similar to the control A188 plants for 63% of the hypomethylated DMRs and levels that are similar to the R₀ A188 plants for 21% of the hypomethylated DMRs, while the remaining 16% of hypomethylated DMRs have intermediate DNA methylation levels in other maize lines (Figure S5). This suggests that for the majority of hypomethylated DMRs, the DNA methylation state that results from tissue culture is unusual. However, a significant portion of these hypomethylated DMRs (21%) from tissue culture exist in this hypomethylated state in natural maize populations, potentially indicating that these loci are more prone to epigenetic variation naturally (Figure S5). For the hypermethylated DMRs, the control plants are similar to diverse maize lines at 25% of the regions, and the R₀ plants are more similar to diverse maize lines at 53% of the DMRs.

RNA-seq was performed on leaf tissue from the same R₁ plants used to assess heritability of DNA methylation differences. We found that loss of methylation at promoters (defined at the 1 kb 5′ of the transcription start site (TSS), 15 genes with hypo-DMRs) was associated with aberrant expression levels of 23% of genes, averaged across all R₁ plants (Figure 4C and Table S4). This proportion was higher than the percentage of all genes genome-wide that fit these criteria (15%). Hypomethylation was more frequently associated with up-regulation of gene expression (71.4% of genes) as opposed to down-regulation of gene expression (28.6% of genes), which was substantially higher than the average genome-wide rate of up-regulation of differentially expressed genes (52%). Unlike the Stroud et al. (2013) study in rice, we did not find evidence that the DMRs located particularly close to the TSS were more likely to be associated with altered expression. The majority (81%) of the misregulated hypo-DMR genes had confirmed heritable DMRs in the R₁ generation. The analysis of the R₁ transcriptome suggests that tissue culture can have heritable effects upon gene expression level for some genes located near hypomethylated DMRs.

Discussion

In this study we have investigated the effect that the tissue culture process has on genome-wide DNA methylation patterns of regenerated plants by generating profiles of DNA methylation in multiple R₀ plants generated from an inbred line. Somaclonal variation has been well characterized in a number of plants species, and generally results in the occurrence of unexpected phenotypes in progeny that arise from tissue culture (Kaeppler and Phillips 1993; Kaeppler et al. 2000; Thorpe 2006; Rhee et al. 2010; Miguel and Marum 2011; Neelakandan and Wang 2012). The rate of occurrence of aberrant phenotypes is higher than observed for normally propagated plants and appears to reflect a higher “mutagenic” rate during tissue culture. Detailed characterization of particular somaclonal variants has revealed evidence for novel TE insertions as well as epigenetic changes (Rhee et al. 2009, 2010). It is worth noting that the consistent DNA methylation changes observed in this study and in a prior study in rice (Stroud et al. 2013) would predict a tissue culture “syndrome” that would be observed in many of the plants regenerated from culture. However, this is generally not observed. Instead, most phenotypic abnormalities are only observed in a subset of plants and often segregate as recessive alleles. This may suggest that the hypomethylation events, which are consistent among regenerants and may result in gain-of-function alleles, are unlikely to be related to phenotypic somaclonal variation. Instead, the more stochastic hypermethylation events may explain some of the somaclonal variation observed following tissue culture.

Our results are consistent with previous studies that assessed specific genomic regions (Kaeppler and Phillips 1993; Zhang et al. 2009; Linacero et al. 2011; Gonzalez et al. 2013) and with a recent genome-wide analysis of rice plants regenerated from tissue culture (Stroud et al. 2013). The bulk of DNA methylation is not affected by tissue culture. However, a subset of genomic regions exhibit altered DNA methylation levels. Similar to a recent study in rice (Stroud et al. 2013), we found that losses of DNA methylation following tissue culture are more common than gains of DNA methylation. In general, we hypothesize that similar phenomena are affecting somaclonal variation within both maize and rice. In both species, embryogenic callus was induced from scutellum tissue of immature embryos. It could be possible that different methods of tissue culture affect methylation patterns differently, as suspension culture of Arabidopsis caused a prevalence of DNA methylation increases in genic regions as opposed to losses (Bednarek et al. 2007; Tanurdzic et al. 2008). Interestingly, numerous DMRs are consistently found in many independent lines and these appear to reflect homozygous changes that occur to both alleles during tissue culture. Contrarily, there are also a number of DMRs that exhibit stochastic behavior or incomplete penetrance. The analysis of R₁ descendants reveals that the majority of hypomethylation events are heritable. BNL5.09, one of the genomic regions identified as exhibiting altered DNA methylation via RFLP analysis (Kaeppler and Phillips 1993) was also found in our study as a consistently hypomethylated DMR (DMR666).

The molecular mechanism that results in altered DNA methylation is not clear. It is possible that the loss of DNA methylation reflects the absence of cellular machinery that maintains DNA methylation patterns in some cell types, as suggested by previous studies (Zhang et al. 2009; Wang et al. 2013). However, the fact that the majority of genomic regions maintain their DNA methylation levels in culture would argue against these broad effects. An alternative explanation is that DNA demethylation machinery could be targeted to specific genomic regions during tissue culture consistent with previous reports (Zhang et al. 2013). The fact that the tissue-culture-induced DMRs are observed in callus samples as well as R₀ plants suggests that many of the DNA methylation changes occur early in the tissue culture process. However, it is possible that some of the changes with incomplete penetrance may be the result of on-going processes during culture.

The functional consequences of altered DNA methylation during tissue culture are not clear. Tissue culture DMRs do not exhibit strong enrichments for location relative to genes or transposons. Only a subset (∼23%) of genes located near tissue culture DMRs in maize exhibit altered expression levels. It is possible that this process of altered DNA methylation plays an important role at a subset of the DMRs and that the other regions with altered methylation have limited consequences. One interpretation is that the tissue culture DMRs do not reflect a process that is critical for somaclonal variation but instead reflect an initial process that is required for culturing plant cells. Remodeling of chromatin at specific sites may play a critical role in passaging cells into culture. These changes may be heritable but only play an important role at early stages of tissue culture. An intriguing corollary to this scenario is that preexisting difference in chromatin may influence the culturability of different genotypes. Many plant species exhibit intraspecific variation for the ability to grow in tissue culture and regenerate fertile plants. Variation in chromatin state of critical genes among different individuals could contribute to differences in the culturability. A significant, albeit small (∼8.6%), proportion of DMRs overlap with DMRs identified in a previous study (Eichten et al. 2013) that are due to natural variation in 51 maize genotypes, indicating natural variation for DNA methylation state at a number of tissue-culture DMRs. These may be loci that are epigenetically sensitive in natural populations under field growth conditions, potentially as a mechanism to dynamically and heritably respond to stress.

In summary, our results indicate that production of plants using tissue culture produces an epigenetic footprint in regenerated plants that is stable over multiple generations. Specific chromosomal regions are specifically sensitive to the stress, and similar variation occurs across independent cultures. Methylation variation in tissue culture is frequent compared to sexually derived progenies, but does not result in a tissue culture syndrome, which would be reflected by similar phenotypic alterations due to consistently altered genomic sites. Nevertheless, altered methylation patterns are often heritable through sexual generations, and some of these changes may contribute to heritable phenotypic changes in tissue-culture-derived progenies.