Periodic DNA Methylation in Maize Nucleosomes and Demethylation by Environmental Stress*

When maize seedlings were exposed to cold stress, a genome-wide demethylation occurred in root tissues. Screening of genomic DNA identified one particular fragment that was demethylated during chilling. This 1.8-kb fragment, designated ZmMI1, contained part of the coding region of a putative protein and part of a retrotransposon-like sequence. ZmMI1 was transcribed only under cold stress. Direct methylation mapping revealed that hypomethylated regions spanning 150 bases alternated with hypermethylated regions spanning 50 bases. Analysis of nuclear DNA digested with micrococcal nuclease indicated that these regions corresponded to nucleosome cores and linkers, respectively. Cold stress induced severe demethylation in core regions but left linker regions relatively intact. Thus, methylation and demethylation were periodic in nucleosomes. The following biological significance is conceivable. First, because DNA methylation in nucleosomes induces alteration of gene expression by changing chromatin structures, vast demethylation may serve as a common switch for many genes that are simultaneously controlled upon environmental cues. Second, because artificial demethylation induces heritable changes in plant phenotype (Sano, H., Kamada, I., Youssefian, S., Katsumi, M., and Wabilko, H. (1990)Mol. Gen. Genet. 220, 441–447), altered DNA methylation may result in epigenetic inheritance, in which gene expression is modified without changing the nucleotide sequence.

DNA of higher eukaryotes is characterized by the presence of 5-methylcytosine (m 5 C) 1 nucleotides, comprising up to 30% of the total cytosines. In vertebrates, m 5 C is located almost exclusively in CpG, whereas in plants it occurs in both CpG and CpNpG (1). The distribution within the genome is non-random and varies depending on the tissue and the developmental stages. The physiological function of m 5 C is essentially to silence gene expression, which is important for host DNA de-fenses against incorporation of "parasitic" DNA (2,3). Two systems are involved: one directly blocks transcriptional machinery attached to promoter regions of genes by altering DNA structure and the other indirectly interferes with transcription by influencing nucleosome conformation and stability (4). Recent studies indicated the latter to occur frequently, as shown by various disorders due to abnormal chromosome structures that are caused by defective DNA methylation (5). One hallmark of cancer cells is local hypermethylation and global hypomethylation of chromosomal DNA (6). Abnormal methylation in the promoter regions of regulatory genes may indeed result in cancer development (7). Using antisense inhibition of DNMT1, a maintenance type DNA methyltransferase, about 10% of all genes in cultured mouse cells were found to be activated (8). These observations confirmed that DNA methylation functions as a global repressor of gene expression (9). The reverse case, i.e. global demethylation has also been inferred to be critical during embryogenesis in mammals (9,10). What controls the on-off switch for DNA methylation, however, is still largely unclear. While a set of DNA methyltransferases has been identified in various organisms, including plants, the presence of DNA demethylases is controversial (7). Among several candidate proteins, 5-methylcytosine DNA glycosylase was shown to induce genome-wide demethylation upon transfection into mouse myoblasts (11). Whatever the mechanism may be, reprogramming of DNA methylation appears to be fundamental in normal development (10).
Epigenetic inheritance is defined as change in gene expression without base sequence alteration (12). This typically occurs during somatic cell differentiation, in which the clonal expansion of a single cell leads to a diversity of cell types (13). Such a cellular inheritance is common during ontogeny but is usually erased before gametes are produced (14). In plants, however, it has long been known that epigenetically acquired traits can be sexually transmitted, as exemplified by flax (Linum) (15). In this case, epigenetic changes were induced by external factors such as nutrients and temperature (16,17). Consequently, the idea has been proposed that an environmental stimulus can induce heritable chromatin modifications as an adaptive response (18). It is established that some clonal epigenetic changes are mediated through DNA methylation (6,13), but evidence is limited for involvement of the latter in inheritance of acquired characteristics (19). We have previously shown that a single exposure of germinated rice seeds to the DNA demethylating agent, 5-azadeoxycytidine, induced dwarfism at maturity (20). Genomic DNA isolated from dwarf plants showed a 16% reduction in the m 5 C content in comparison with DNA from untreated plants. Both hypomethylation and dwarfism were transmitted to progeny for at least three generations (20). Thus, the acquired phenotype due to acquired changes in DNA methylation was heritable. However, whether or not such an epigenetic inheritance * This work was supported by grants for the Core Research for Evolutional Science and Technology (CREST) from the Japan Science and Technology Corporation (JST), and Grant JSPS-RFTF 00L01604 for the Research for the Future Program from the Japan Society for the Promotion of Science. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
In this report we provide support for the conclusion that levels of DNA methylation change in the cores of nucleosomes in maize root tissues in response to cold stress or environmental cues. This finding allowed us to speculate that DNA methylation may function as a common switch of gene expression and that naturally induced changes in DNA methylation may result in heritable epigenetic modification of gene expression.

EXPERIMENTAL PROCEDURES
Plant Materials and Cold Treatments-Maize seeds (Zea mays L. cv. Golden Arrow) were germinated and hydroponically grown in a onefifth strength Murashige and Skoog medium (Nihon Seiyaku, Tokyo, Japan) under continuous light for 13 days at 23°C and 70% relative humidity in a growth cabinet. Cold-pulse experiments were performed by transferring 13-day-old seedlings to an incubator at 4°C for 6 days and then returning samples to 23°C for a further 7 days. Light and humidity conditions were kept constant throughout the experimental periods.
High Pressure Liquid Chromatography Analysis-Root tissues were collected from 10 seedlings at the end of each treatment step, and DNA was extracted using a Nucleon phytopure kit (Amersham Biosciences). A 100-g aliquot of genomic DNA was incubated with 110 ng of RNase A (Nacalai Tesque, Kyoto, Japan) at 37°C for 2 h to remove RNA contamination. After ethanol precipitation and denaturation at 100°C for 5 min, DNA was digested with 2 units of nuclease P1 (Sigma) at 37°C for 20 h, followed by dephosphorylation with 20 units of calf intestine alkaline phosphatase (Takara, Otsu, Japan) at 37°C for 2 h. The sample was fractionated by Ultrafree-MC PL-10 microcentrifuge tubes (Millipore, Bedford, MA), and the permeate was injected into a Supelcosil LC-18-S column (Supelco, Bellefonte, PA). Separation was performed with a 2.5-20% methanol gradient in the presence of 50 mM KH 2 PO 4 (pH 4.3).
Methylated CpG Island Screening-DNA samples were extracted from roots of control untreated seedlings and 6-day-old cold-treated seedlings as described above. Methylated CpG island amplification was performed by successive digestion of 5 g of genomic DNA using SmaI and XmaI (24). SmaI cleaves unmethylated 5Ј-CCCGGG-3Ј, giving blunt-end products, whereas XmaI cleaves the same sequence regardless of the methylation status, generating cohesive-end products. To prepare a cohesive-end adaptor, 50 M each oligonucleotides 5Ј-CCGG-TAGCTAATGAACCAT-3Ј and 5Ј-ATCGATTACTTGGTA-3Ј were mixed in TE buffer and annealed at 65°C for 5 min. A 500-ng aliquot of digested DNA was then ligated to 1 pmol of adaptor with 400 units of T4 ligase (Takara) at 37°C for 16 h. This procedure yields only XmaIdigested DNA ligated to the adaptor. The resultant fragments were amplified by PCR with 30 cycles of 96°C for 20 s, 58°C for 25 s, and 72°C for 2 min using the single adaptor-specific primer 5Ј-ATGGT-TCATTAGCTACCGGG-3Ј, ExTaq (Takara) and 250 M dNTPs. Differential display screening was performed with DNA fragments amplified by second PCR using the initial PCR product as a template and one of 10-base oligonucleotide primers (Operon Technologies, Alameda, CA). Several hundred PCR products were separated on an agarose gel and stained with ethidium bromide. Migration patterns were compared between DNA samples originating from control and cold-treated seedlings, and fragments amplified exclusively from control DNA were selected as putative demethylated CpG islands.

Southern Hybridization, Genomic Library Screening, and Reverse
Transcriptase-PCR-Genomic DNA samples from roots were digested with either MspI or HapII (Takara), separated on agarose gels, and transferred to Hybond-N ϩ membranes (Amersham Biosciences). For library screening genomic DNA was digested with BamHI and fractionated on an agarose gel. A fraction containing ϳ1-1.6 kb of DNA was extracted and ligated to a ZAP Express BamHI/alkaline phosphatase-treated vector (Stratagene). Membranes were probed with a radioactively labeled 599-bp ZmMI1 fragment (GenBank/EBI accession no. AF453523). For reverse transcriptase-PCR analysis total RNA was prepared by the aurintricarboxylic acid method (25) and treated with DNaseI (Takara) and cDNA were synthesized using an oligo-dT 17 as the primer and Super Script II (Invitrogen). cDNA was used as a template for ZmMI1 transcript-specific amplification (forward, 5Ј-GTCCGGGA-GCTTCCTTAAGC-3Ј and reverse, 5Ј-CCTTCAATGAGCTCCTGCTC-3Ј) or actin transcript-specific amplification (forward, 5Ј-CGAACAAC-TGGTATTGTGATGG-3Ј and reverse, 5Ј-TGCTGAAAAGTGCTGAGA-GAAG-3Ј). Samples were separated by electrophoresis on agarose gels and visualized with ethidium bromide staining.
Cytosine Methylation Mapping-Total genomic DNA was digested with XbaI (Takara) and subjected to bisulfite modification (26). The modified DNA was subjected to PCR using ZmMI1-specific primers (forward, 5Ј-GAGGAAGAGAAAGGGAGAG-3Ј and reverse, 5Ј-AAATC-CATTTTCTATTCATTTATTC-3Ј), ExTaq enzyme (Takara), 250 M each dATP and dTTP, and 125 M each dGTP and dCTP. To remove partially and non-modified sequences, PCR products were digested with AluI, for which sites were found to be rarely or never methylated by preliminary methylation mapping. A 1015-bp AluI-resistant fraction was extracted after an agarose gel electrophoresis, ligated to the pGEM-T Easy Vector (Promega, Madison, WI), and cloned in JM109 (Stratagene, La Jolla, CA). The sequence was determined with an ABI PRISM BigDye Terminator DNA sequencing kit and a 3100 Genetic Analyzer automated sequencer (Applied Biosystems, Foster City, CA).

Genome-wide Demethylation upon Cold
Stress-Analyses of maize DNA methyltransferase ZmMET1 during chilling (28) and of vernalization in Arabidopsis thaliana (29) revealed that cold stress might induce modification of the DNA methylation status. To obtain confirmatory evidence, we examined the level of total m 5 C in maize seedlings grown at 23°C for 13 days and then maintained at 4°C for up to 6 days. At the end of this treatment, growth had halted and tissue necrosis was apparent on leaf edges. Plants were then returned to normal culture conditions at 23°C for an additional 7 days to allow recovery of frail growth, although the plants ultimately died after 2 weeks. Genomic DNA was extracted from leaf blades, stem mesocotyls, and root tissues at each step and subjected to nuclease P1 and phosphatase digestions to yield nucleosides, which were analyzed by high pressure liquid chromatography. The methylation level estimated from the ratio of m 5 C to cytosine was little affected in leaf blades and stem mesocotyls, m 5 C accounting for ϳ20% of the total cytosines throughout the treatment period (Fig. 1, A and B). In contrast, the methylation level clearly decreased in roots upon cold treatment (Fig. 1C). A notable feature was that, even after seedlings were returned to normal growth conditions, the decreased methylation level did not recover but rather continued to decline, reaching one third of the original level (Fig. 1C). The global status of methylation and its alteration upon cold treatment was histochemically examined in 8-day-old seedlings using antibodies against m 5 C. The distribution of m 5 C was not random in and among chromosomes (Fig. 2). In a chromosome, heavily methylated regions alternated with undermethylated regions. Some chromosomes were less methylated than others (Fig. 2). A hypomethylated region in chromosome 4 was assigned to the knob, which consists of TR-1 repeats forming fold-back DNA segments (30). However, no distinct difference was observed in methylation patterns among chromosomes from untreated and cold-treated seedlings (data not shown), probably due to the inability of antibodies to provide quantitative analysis under the experimental condition.
Isolation and Characterization of Demethylated Genomic Sequences-To identify specific sequences demethylated in root tissues upon cold stress, methylated CpG island amplification (24) and PCR differential display screening were performed. Initial screening identified nine genomic fragments that were possibly demethylated on chilling (data not shown), one of which (registered as GenBank/EBI accession no. AF453523) was further used as the probe and analyzed by Southern blot hybridization using MspI and HapII m 5 C-sensitive restriction enzymes. An ϳ0.6-kb fragment was distinct in DNA from coldtreated seedlings, whereas it was absent in DNA from un-treated seedlings (Fig. 3A). This indicated two cytosines in CCGG to be methylated in controls but demethylated in chilled samples, this being confirmed later by direct methylation mapping sequencing (data not shown). The isolated fragment was used as the probe to screen a genomic DNA library, and a 1756-bp fragment was finally isolated (registered as Gen-Bank/EBI accession no. AF453522) (Fig. 3B). This genomic fragment, designated as ZmMI1, showed a similarity at the 3Ј-region from position ϳ1400 with the PREM-1 long terminal repeat (registered as GenBank/EBI accession no. U03681), a putative retroelement (Fig. 3B). A part of common sequences between 480 and 800 matched well to a maize 348-bp expressed sequence tag clone, which was induced upon salt stress to seedlings (registered as GenBank/EBI accession no. AI967092) (Fig. 3B). The copy number of ZmMI1 was estimated to be ϳ40 -50 by Southern blot signal intensity analysis (data not shown). To identify possible transcription of this sequence, DNA samples were extracted from 13-day-old seedlings grown under 23°C (day 0), from seedlings that were further cold-treated at 4°C for 8 days (day 8), and from seedlings returned to 23°C for 3 days (day 11) and 7 days (day 18) after cold treatment. Reference cultivation without cold treatment was performed in parallel. The ratio of m 5 C to total cytosine is expressed in percentage (%) by shaded and closed bars for untreated and cold-treated samples, respectively.

FIG. 2. DNA methylation patterns in chromosomes.
Methylated sites were revealed by indirect immunofluorescence labeling with monoclonal anti-m 5 C antibodies. Samples were from 8-day old seedlings grown at 23°C. DAPI staining images (A) and fluorescein isothiocyanate signal images by staining with monoclonal anti-m 5 C antibodies followed by the second antibodies (B) are merged (C). Bars indicate 5 m. Arrowheads indicate chromosome 4, which is enlarged in insets. A distinctly weak signal site was found on cytogenetically visible knob of chromosome 4 (indicated by arrows). a cDNA library was screened and six clones were obtained. All of the sequences, however, slightly differed each other, probably reflecting transcription from different loci (registered as GenBank/EBI accession nos. AF468668 through AF468672). The position of poly(A) addition was around 1100 (Fig. 3B). The level of transcription was examined by reverse transcriptase-PCR and found to increase during cold treatments (Fig. 3C). These results indicate that ZmMI1 consists of a retroposon-like sequence and a part of a gene, which is demethylated and transcriptionally activated upon cold stress.
Periodic Distribution of m 5 C in the ZmMI1 Region-The methylation status in ZmMI1 was directly examined by the bisulfite modification method. Genomic DNA was extracted from roots of three batches each containing ϳ10 seedlings grown, handled as described above, and subjected to bisulfite treatment. An 886-bp region of ZmMI1 (positions between 440 and 1326, Fig. 3B) was amplified by PCR and cloned, the sequence determined, and the positions and frequencies of m 5 C mapped ( Fig. 4 and see also the supporting information). The results revealed two distinct features of DNA methylation: periodic oscillation of the pattern and non-random demethylation on cold treatment. m 5 C does not occur uniformly but rather periodically within ϳ200 bases. The heavily methylated region, in which up to 100% of the methylatable sites are saturated with m 5 C, comprises ϳ45 bases. The relatively undermethylated region, in which only ϳ60% of methylatable sites are m 5 C, comprises ϳ145 bases. The two alternate, forming a periodic oscillation pattern in terms of DNA methylation level (Fig. 4A).
Non-random Demethylation by Cold Stress-An average of 38.4% cytosine residues in ZmMI1 were methylated when seedlings were cultivated at 23°C (Fig. 4, A and D). This methylation level declined to 24.7% after 5 days of chilling (Fig. 4, B  and D). When cold-treated samples were returned to 23°C and cultivated for additional 7 days (cold-chased), demethylation FIG. 3. Properties of ZmMI1. A, Southern blots analyses. Genomic DNA extracted from root tissues (10 g per lane) from seedlings untreated (23°C) or cold-treated (4°C) was digested with either MspI or HapII, separated on an agarose gel, blotted, and probed with a 599-bp radioactively labeled ZmMI1 fragment (registered as GenBank/EBI accession no. AF453523). The size of the hybridizing signals is indicated along with the position of DNA molecular markers. B, schematic representation of ZmMI1 genome DNA sequence (registered as Gen-Bank/EBI accession no. AF453522) isolated from a genomic DNA library, showing partial homology with the cDNA fragments (registered as GenBank/EBI accession nos. AF468668 through AF468672) isolated from a cDNA library screening with an expressed sequence tag clone (registered as GenBank/EBI accession no. AI967092) and a retroelement PREM-1 partial sequence (registered as GenBank/EBI accession no. U03681). Dotted lines indicate sequences not determined. The region subjected to methylation mapping is indicated by double arrows. C, induction of ZmMI1 transcripts upon cold treatments. Total RNA was isolated from seedlings treated or untreated with cold stress as described above, and reverse transcriptase-PCR was performed to quantify ZmMI1 transcripts. actin transcripts were simultaneously estimated for the RNA loading control. Note that ZmMI1 transcripts increased concomitantly with, but decreased independently, on demethylation. persisted and the level declined further to 22.5% (Fig. 4, C and  D). However, the change did not occur uniformly but rather preferentially in the ϳ145-base regions, with depression of some methylation sites from 75% to 5%. The ϳ45-base region with high methylation remained generally unaffected (Fig. 4, B  and C). Another aspect of the demethylation was cytosine motif specificity. Demethylation of CpG was more distinct than that of CpNpG (Fig. 4D). Some isolated cytosine residues (orphan C) in these samples appeared to be remethylated (Fig. 4D). Because CpG and CpNpG are uniformly distributed along the ZmMI1 sequence, some factor(s) independent of sequence may be responsible for periodic methylation and preferential demethylation.
Association of the Methylation Pattern with the Nucleosomal Structure-The periodic oscillation pattern in every 200 bases suggests an association with the nucleosome structure. To assess this possibility, root nuclei were extracted from seedlings and subjected to digestion with micrococcal nuclease, which preferentially cleaves chromatin DNA in the spacer regions. Resultant ϳ200-base fragments were isolated (Fig. 5A) and subjected to PCR in parallel with control undigested genomic DNA. Primers were designed to independently amplify five ϳ150-base fragments, each containing either hypermethylated or hypomethylated regions (Fig. 5B). When genomic DNA was used as the template for PCR with five sets of primers, corresponding fragments were equally well amplified (Fig. 5C). When nuclease-digested 200-base DNA fragments were employed, fragments containing hypomethylated regions were more efficiently amplified than those containing hypermethylated regions (Fig. 5C). This suggested that micrococcal nuclease preferentially cleaved DNA around the hypermethylated regions. In other words, ϳ145-base regions enriched with hypomethylated cytosines were resistant to nuclease and hence corresponded to the core, whereas ϳ45-base regions enriched with hypermethylated cytosines were sensitive to nuclease and hence corresponded to the linker regions. We concluded therefore that DNA in the core and linker regions of nucleosomes was hypo-and hypermethylated, respectively. However, whether or not such differential methylation patterns correlated with the mode of nuclease digestion was not clear from the present finding.

DISCUSSION
Nucleosomes are organized into precise positions on DNA by signals in the sequence (31). One such signal has been proposed to be m 5 C, but the experimental evidence is controversial. For example, opposite results as to preferential binding of histone H1 to methylated DNA have been reported (32,33), and currently the interaction between histones and methylated DNA is in general not considered to be specific. However, under certain circumstances, nucleosome assembly appears to depend on CpG methylation (31,34). This suggests a dynamic interaction between methylated DNA and proteins that constitute chromatin. Methylation of histones was shown to be essential for triggering DNA methylation in filamentous fungi, and it was suggested that propagation of DNA methylation patterns is dependent on feed-back loops between modification of chromatin proteins and DNA (35). Our result showing periodic oscillation of DNA methylation within the nucleosomal structure is consistent with this view and further suggests a positive role for methylation in determining the chromatin infrastructure. Our findings also substantiate occurrence of demethylation upon environmental stress.
During differentiation of mammalian cells, genome-wide loss of DNA methylation occurs, for which active or passive mechanisms have been proposed (11,36). In active demethylation, m 5 C is enzymatically removed from DNA, and in passive demethylation, newly replicated DNA is not methylated (11). Our results with maize indicate that genome-wide demethylation occurs, possibly through active demethylation, because chilled tissues immediately ceased DNA replication (28). A rapid and active decrease in global DNA methylation was also observed during seed germination of Silene latifolia (37). Active demethylation is catalyzed by 5-methylcytosine DNA glycosidase (11,36), although such activity has not yet been found in plants. A specific feature of our findings was that demethylation predominantly occurred at the nucleosome core regions as revealed by direct methylation mapping. A question arises, then, as to how such a differential demethylation pattern is formed. Currently we have no explanation for this, but some clues for speculation are available. For example, nucleosome cores are proposed to be located outside the 30-nm fiber, which is composed of a unit of six nucleosomes. Linker regions are connected by histone H1 inside the fiber. This structure makes 5-methylcytosine DNA glycosylase more accessible to the core than to the linker. Alternatively, 5-methylcytosine DNA glycosylase may be selectively recruited to the core by specific protein(s) such as, for example, 5-methylcytosine binding proteins. Another distinct finding of the direct methylation mapping is ubiquitous demethylation at CpG, CpNpG and other sites, sug-

FIG. 5. Methylation pattern in nucleosomes.
A, isolation of mononucleosomal DNA fraction. Intact chromatin extracted from isolated maize seedlings nuclei was subjected to micrococcal nuclease digestion, and resulting DNA fragments were subjected to agarose gel electrophoresis. A 200-bp fragment indicated by the arrow was collected and used for further analysis. B, schematic illustration of methylation frequency and PCR positions. Methylation frequency on the ZmMI1 locus in untreated seedlings is illustrated after Fig. 4A. Fragments 1, 3, and 5 were expected to be amplified from hypermethylated regions and Fragments 2 and 4 from hypomethylated regions. C, template efficiency. PCR was performed with 200-bp mononucleosomal DNA fragments (upper gel) or with total genomic DNA (lower gel) as the template. Products were subjected to agarose gel electrophoresis and visualized with ethidium bromide staining. Numbers correspond to those illustrated in B. D, ratio of amplification from a 200-bp DNA template to those from total genomic DNA. Amount of each fragment was densitometrically estimated from ethidium bromide staining, and relative ratio of obtained values for a 200-bp DNA template to those for genomic DNA (taken as 100%) was determined. Numbers correspond to those illustrated in B.
gesting no restriction to only one strand of DNA. This means that, for remethylation, de novo methyltransferase activity is necessary. If this enzyme is absent in quiescent cells, for example, in root tissues, the methylated status can not be restored. This could account for the apparently progressive demethylation in cold-chased seedlings.
While the average methylation level of total genomic DNA in untreated root tissues was ϳ18%, that of ZmMI1 was 38%, 2-fold higher. This indicates a heterogeneous distribution of m 5 C in the genome, consistent with the histochemical observations, and suggests ZmMI1 to be located in heterochromatin. However, because demethylation occurred throughout the genome, including the ZmMI1 sequence, the periodic methylation pattern found in the latter may be common to other parts of genome. If so, the occurrence of demethylation mainly in the core regions of nucleosomes could induce an alteration of chromatin structure and thus influence gene expression. Increase in ZmMI1 transcripts substantiates this hypothesis. The advantage of this feature is that expression of many genes could be simultaneously controlled. Indeed, it was estimated that plants may express as many as several hundred genes upon cold stress (38). It is attractive to speculate that plants have developed the methylation/demethylation system to simultaneously regulate a vast number of genes at once, rather than to individually regulate each gene. Alternatively, because ϳ70% of the maize genome is composed of retroposon-like sequences (39) and because demethylation has been shown to reactivate transposons (40), the global demethylation possibly transcriptionally activates such sequences to translocate, resulting in secondary activation/inactivation of other genes (41). However, the reason why such demethylation occurs only in root tissues is not clear. One explanation might be that above-ground parts are equipped with a more complete self-defense system than those underground, as exemplified by the presence and absence of maintenance-type DNA methyltransferase in the respective tissues (28).
Heredity of epigenetic modification, or acquired traits, in response to environmental conditions has long been speculated, and DNA methylation has been proposed as one of its promoters (13). In animal cells, however, DNA methylation patterns are strictly regulated in somatic cells so that upon gametogenesis those that are acquired are usually completely erased and not transmitted to the progeny (14). In plants, the distinction between somatic and germ cells is less obvious in comparison with animals, making possible transmission of acquired DNA methylation patterns. We have experimentally shown this by treatment of germinated rice seeds with 5-azadeoxycytidine, a chemical that powerfully induces demethylation of DNA in vivo (20). At maturity, plants exhibited a global demethylation and altered phenotypes, including dwarfism. The acquired traits and demethylation patterns were inherited for up to at least six generations. 2 Thus alteration of DNA methylation can induce changed expression of some genes, resulting in a new phenotype, both of which are heritable (20).
Our above-mentioned experiments suggested that the acquired trait due to an acquired methylation pattern is sexually transmittable, but evidence of methylation patterns changing depending on the environment was missing. The present study showed that this can indeed occur. However, maize roots are differentiated tissues consisting of somatic cells and they do not form germ cells. In this context, the demethylation pattern would not be transmittable to progeny. The fact that no obvious demethylation was observed in stem mesocotyl tissues, which contain cells with the potentiality to develop into germ cells, suggests that these latter are well protected against environmental stresses in terms of DNA methylation. However, if a DNA methylation pattern did change, it could result in heritable epimutations. If such mutations were advantageous for survival, they might persist for generations. An example is the flower morphology change in Linaria vulgaris, considered to be the result of hypermethylation of the Lcyc gene which occurred 250 years ago (42). Also, in plants, vegetative reproduction is not rare as seen with tuber propagation, making it possible to directly transmit altered methylation patterns to the next generation. We therefore speculate that Lamarckian inheritance does exist, being mediated through DNA methylation.