Variation and inheritance of cytosine methylation patterns in wheat at the high molecular weight glutenin and ribosomal RNA gene loci

There is now substantial evidence that the activity of chromosome segments is correlated with modifications of specific bases, most commonly cytosine. 5-methylcytosine occurs especially frequently in plant genomes in the symmetrical dinucleotide CG or in the trinucleotide CXG (see Table 1). In plant species, from 12 to 33 % of the cytosine residues are methylated (Wagner and Cespius, 1981). It is likely that the modification of cytosine provides information affecting many kinds of nuclear processes. The best documented is in mammalian genomes where the stable, inherited modification of DNA sequences around a gene can influence the binding of transcription and other factors to the gene and hence influence its expression (Cedar, 1988). Genes which contain a high proportion of methylated cytosines are usually inactive and may be inaccessible to the transcription machinery, while those which are active or potentially active are not methylated at critical cytosine residues. Results for plant species consistent with this general conclusion have been published for Agrobacterium T-DNA genes (Hepburn et al. 1983; Gelvin et al. 1983; van Slogteren et al. 1984 and Peerbolte et al. 1986), for rRNA genes (Blundy et al. 1987; Watson et al. 1987 and Flavell et al. 1988), for maize zein storage protein genes (Bianchi and Viotti, 1988) and for maize transposable elements (Chomet et al. 1987; Chandler and Walbot, 1986; Chandler et al. 1988; Schwartz and Dennis, 1986; Fedoroff, 1989 and Martienssen et al. 1990).

The studies on transposable elements (Fedoroff, 1989; Martienssen et al. 1990) are particularly interesting, because they reveal not only the correlation between activity and the presence of specific unmethy-lated cytosines around the start sites of transcription, but also the activation and demethylation of partially methylated copies in a specific developmental pattern and by the presence of additional active copies of the element in the genome. The altered methylation pattern can be inherited giving the element a different activity potential in the next generation. Whether the loss of methyl groups is due to active transcription interfering with the methylation process or whether it is the consequence of regulatory proteins binding to the DNA before transcription and thus interfering with methylation is often unclear. However, several studies have shown that methylated DNA is not transcribed when introduced into cells (Cedar, 1988).

Specific methyltransferases which use S-adenosyl-methionine as the donor of methyl groups are present to modify the cytosine residues (Kirnos et al. 1981). Few detailed studies of the processes involved and their control have been carried out in plants, but Kirnos and collaborators have provided some data which show that newly replicated DNA fragments are undermethylated (Vanyushin and Kirnos, 1988). Much of the methylation occurs after replication by a process that does not involve DNA repair. The methylases presumably recognise the hemimethylated symmetrical CG or CXG motifs and methylate the cytosine in the new strand, unless something such as a DNA-bound protein interferes with the process. The methylation process must be very efficient to ensure faithful propagation of the cytosine methylation pattern.

Estimates of the error frequency in the methylation process have not been made in plants. However, in a major survey of maize plants regenerated from tissue culture, many changes in cytosine methylation were recorded especially in a particular genotype (Brown, 1989). These changes appear to be stably transmitted from one generation to the next. While phenotypic changes are frequent in these plants propagated through tissue culture, there is no evidence yet to prove that this is due to altered gene expression as a result of failure to methylate a specific cytosine residue.

Recently we have initiated studies to investigate variation in cytosine methylation in hexapioid bread-wheat at specific loci that encode high molecular weight glutenin subunits which are major seed proteins. We have also studied the inheritance of this variation. This seemed a worthwhile study because if methylation of cytosine residues is an important feature of the control of the activity of chromosome segments, then a better understanding of the control and fidelity of the methylation process in plants possessing high levels of cytosine methylation is desirable.

Another way of examining the control of cytosine methylation is to study a large number of copies of the same gene. The ribosomal RNA genes constitute a very large multigene family in wheat, with between 8000 and 15 000 members. Members of the family can be compared within and between genotypes to examine variation in cytosine methylation.

In this paper we summarise results of these studies which demonstrate a strong correlation between the status of cytosine methylation at specific sites and gene expression, although cytosine methylation at other sites appear not to show such correlation. Variation in cytosine methylation at specific sites was uncovered between inbred plants and between members of the rRNA multigene family. Some of the methylation modifications are inherited through meiosis.

The genes for high molecular weight (HMW) glutenin subunits are important because they are major contributors to the visco-elastic properties of dough made from flour (Payne et al. 1981 ; Flavell et al. 1989). There are three pairs of HMW glutenin genes in hexapioid wheat, one pair on each of the closely related group 1 chromosomes, 1A, IB and ID. When nuclear DNA from the variety Chinese Spring is restricted with fiamHI and probed with the cDNA encoding a HMW glutenin, eight fragments ranging from 2.0 to 10.2kbp are detected (Fig. 1A). From studies on DNAs from aneuploid lines lacking chromosomes 1 A, 1B or ID the chromosomal origin of each of the fragments has been determined. When DNAs from different seedlings of Chinese Spring, regenerated from callus tissue initiated from scutellum, were treated with BamHl and Hhal (an enzyme which does not cleave GCGC sites when the internal cytosine is methylated) then several different hybridisation patterns were obtained. Out of a sample of 36 seedlings studied, 7 variant patterns were seen. Some are shown in Fig. 1A. The variation is due to variation in cytosine methylation at one or more of the HhaI sites on the BamHI fragments containing the HMW glutenin genes. For example, in one plant (E2a, Fig. 1A) the 10.2 kbp fragment of the locus on chromosome ID is cleaved to give a 5.1 kbp fragment, while in another (Ela) it is cleaved to give a 2.9 kbp fragment (Fig. IB). Occasionally both kinds of fragment are visible (E3a, Fig. 1A). Such variation was unexpected given that Chinese Spring seeds are highly inbred. Seedlings grown directly from inbred seed were also examined. Similar variation was seen. These results show that variation in cytosine methylation arises during plant life cycles and probably also during tissue culture.

The variation is stably inherited in most cases. This was established by making crosses between individuals whose methylation patterns in leaf DNAs are different. The F| progeny had the additive pattern of hybridising fragments expected if homozygous chromosomal patterns were transmitted through meiosis, the zygote and perpetuated somatically. The methylation patterns of F, plants were unaffected by which plant donated the egg or the pollen. This implies that the single pollen and egg cells had the same methylation patterns as the leaf and stem cells. A few F₂ progeny have also been examined. Parental and heterozygous patterns were detected endorsing once again the stability of methylation patterns during meiosis. In addition some new methylation patterns have been detected in F₂ progeny. The time in development when the variation arose is not known. It could have occurred in the F, plants before meiosis. during meiosis, post meiosis during zygote formation or during development and growth of the F₂ plant. The methylation change did not occur in all copies of the glutenin gene subunits which implies it might have arisen late during seedling development. Alternatively, the change may have occurred in only one of the two alleles. The discovery of new variation is not surprising, given that 3 out of our original sample of 19 Chinese Spring seeds were variants.

The HMW glutenin genes are expressed only in the seed. Therefore it was interesting to study the methylation patterns in and around the loci in DNA isolated from the developing seeds and compare them with those of other organs. After treatment with and Hhal, similar patterns of fragments were observed in the endosperm and leaf samples. However, some additional 1.4 and 1.6 kbp fragments were produced in the endosperm and these must be due to cleavage of Hhal sites close to or within a HMW glutenin gene. These sites are not cleaved in leaf DNA of the same plant. Also, the sites giving rise to the 2.9 kbp fragments (see Fig. IB) were unmethylated in seed of the plant where they were methylated in the leaf cells. These discoveries imply that the undermethylation of these specific sites in the seed, most of which is the endosperm tissue, is associated with gene expression.

In the second system we examined the heterogeneity amongst the members of the large family of ribosomal RNA genes in wheat. Cytological studies at the light and electron microscope level as well as studies on stocks carrying major ribosomal DNA deletions have shown that there is an excess number of genes and not all are required (Flavell et al. 1988). Indeed perhaps fewer than 20% are used. What determines how many and which subset of the genes are used? Are the choices made during every cell generation? Are the active genes marked differentially from the inactive set?

The rRNA genes are organised in long tandem arrays called nucleolus organisers. The plethora of rRNA genes are remarkably uniform in sequence due to recombination and gene conversion events which homogenise the sequences over time within and between the complex loci (Flavell, 1985). However, there is considerable heterogeneity within the gene family with respect to the patterns of cytosine methylation. Many of the genes are methylated at most or all of the CG residues assayed by restriction enzymes. Another subset of the genes contains one or a few unmethylated cytosines but many different sites are involved. The third class carries an unmethylated CCGG site at 165 base pairs upstream from the start of transcription.

The number of genes which carry one or more methylated CCGG sites was studied in related plants that differ greatly in the total number of rRNA genes (Flavell et al. 1988). As the total number of rRNA genes increases, the proportion containing one or more unmethylated CCGG sites decreases. In addition amongst the genes that contain unmethylated CCGG sites, fewer genes contain more than one unmethylated site. This relationship suggested that the total number of unmethylated CCGG sites in rDNA might be relatively constant between genotypes, the sites being distributed (but not at random) among the available rRNA genes. The extent to which the number of unmethylated cytosines is controlled is difficult to define with precision but it is clear that the cytosine methylation patterns in rDNA are highly regulated. Furthermore, these different patterns of methylation correlate with nucleolus activity as determined cytologically (Flavell et al. 1988) and described below.

In hexaploid wheat there are two major nucleolar organiser loci, on chromosomes IB and 6B respectively. Each ribosomal RNA gene of the array is associated with an intergenic region. This region contains amongst other features an array of 135 bp repeats as shown in Fig. 2 (Barker et al. 1988). In some varieties the genes at the IB locus produce a more active nucleolus than those at the 6B locus while in other varieties it is the reverse. The genes at the two loci are often distinguishable because in the intergenic region there are different numbers of 135 bp sub-repeats which are responsible for the production of restriction fragments of different lengths when DNA is treated with an appropriate restriction endonuclease. It has thus been possible by combining cytological assays of nucleolus volume with the methylation status of specific restriction fragments to examine the relationship between nucleolus volume, gene number and methylation status of the genes in specific IB and 6B loci present together in the same cell. (Flavell et al. 1988; Sardana and Flavell, unpublished).

The loci which give the larger nucleolus in a cell invariably have a larger number of rRNA genes without a methylated cytosine at the -165 CCGG site. Conversely, an inactive or weakly active nucleolus organiser has a much higher proportion of its rRNA genes methylated at all the CCGG and GCGC sites assayed. These results imply a correlation between the extent of cytosine methylation and locus activity; an active locus is associated with non-methylation of cytosine residues. The genes at a more active locus also have more 135 bp repeats in the intergenic region. This correlation suggests that 135 bp repeats may act as enhancers of gene activity, as the equivalent repeats have been shown to do in Xenopus (Reeder, 1984). The 135 bp repeats also have an interesting pattern of cytosine methylation. In almost all those genes with the -165 CCGG site unmethylated, one or more of the 135bp repeats has a GCGC site unmethylated. The observation that only one or a few of the 135 bp repeats in such genes are unmethylated is especially interesting because the primary sequences of all the 135 bp repeats are essentially identical. Furthermore, the distribution of unmethylated 135 bp repeats in the tandem array is not random. The probability of a 135 bp repeat containing an unmethylated GCGC site is much greater in the 3’ half of the array. The more active nucleoli are therefore characterised by having more genes with unmethylated 135 bp repeats close to the promoter. How is this achieved? How is this subset of genes distinguished by unmethylated sites selected?

A model and working hypothesis to relate gene methylation and expression has been presented else-where (Flavell et al. 1986, 1989). We have proposed that specific proteins in limiting concentrations can bind co-operatively to sequences in the promoter and the 135 bp repeats when certain cytosines are not methylated. We have recently recognised a protein species in cell extracts that can bind to sequences in these regions (S. Jackson and R. B. Flavell, unpublished observations). When the proteins are bound, the DNA is prevented from being methylated in these regions, which in turn prevents it being wrapped up in heterochromatin and thereby becoming inaccessible to transcription complexes. The model further postulates that these proteins facilitate the binding of transcription complexes and thus play an important role in the control of transcription. The co-operative binding of such proteins and transcription processes in a competitive manner between the rRNA gene variants available would lead to a situation where sufficient genes with a greater capacity to bind transcription complex proteins are distinguished from the total set of rRNA genes by lacking methylated cytosines at specific sites. Because the work on rRNA genes and the studies on the HMW glutenin region presented earlier show that methylation patterns are inherited somatically and even through meiosis, then it is reasonable to propose that the marking of the genes by de-methylation ensures that the same subset, by and large, is used from one cell generation to the next and the competition process between genes for specific protein binding does not involve all the genes in every cell cycle.

However, when new rRNA genes are introduced via a sexual cross, the situation changes because genes of a different competitive ability may be added. This is frequently the case when wide crosses are made. Suppression of the nucleolar organisers of one species in the presence of those of another is common in plants and animals (Flavell et al. 1986). We have described the situation in wheat-rye hybrids (Flavell, 1989) and also in the wheat lines in which the nucleolar organiser bearing chromosome from a wild relative, Aegilops umbellulata, has been added (Martini et al. 1982). The wheat nucleolus organisers are suppressed while the extra one of the Aegilops species is very active. Consistent with the results presented above, many more of the wheat rRNA genes are methylated at all sites assayed in the presence of the Aegilops umbellulata chromosome, while many of the Aegilops umbellulata rRNA genes have unmethylated sites in the intergenic regulatory regions (Flavell et al. 1988).

A similar situation occurs in wide crosses between Hordeum species. Cytological analyses of the fertilised egg cells and cells of subsequent generations (R. Finch, M. D. Bennett and R. B. Flavell, unpublished observations) have shown that while nucleolar organisers of both parents are active in the fertilised egg cell, only those of one parent can be observed after 2 or 3 cell cycles. We presume that these cytological changes are accompanied by the molecular changes in methylation, and therefore the restructuring of the rRNA gene loci to mark the set of genes potentially most useful to the cell occurs as soon as the new genotype is established, i.e. in the fertilised egg. This subset, we predict, is perpetuated by somatic inheritance of the methylation pattern.

The analyses of both sorts of loci described here are consistent with the conclusion that methylation of cytosine at CG dinucleotides is very common in plants and that loss of the methyl group from specific sites is correlated with gene expression or the potential for the gene to be expressed.

As we have described for the rRNA genes, this specific interference with the methylation process is likely to involve the binding of specific regulatory proteins and possibly transcription complexes. Although it can be argued whether the initial event interfering with methylation comes before or after transcription, the inheritance of a methylation pattern raises the important point that once the pattern is established in cells of a tissue or physiological state, it is perpetuated through cell division to predispose the gene in derived cells to be accessible to regulatory proteins and transcription complexes. Thus, the choice of which gene template to use need not be made de novo in each daughter cell after cell division. Rather, it is established for a particular cell lineage and stably inherited through cell division. Chromosome or gene locus marking associated with the modification of bases may therefore play a very significant role in development. This theme is elaborated in the discoveries of genomic imprinting in mammals in which cytosine methylation patterns are determined by whether the gene was inherited from the father or the mother (Sapienza et al. 1987; Reik et al. 1987). It is also relevant to the studies on plant transposable elements whose methylation state varies during development, altering the potential activity of the element (Federoff, 1989; Martienssen et al. 1990). The activity of the element itself or another element in the genome is involved in the methylation change, suggesting that the transposase product may interact with the element’s DNA and interfere with the methylation process. Because in plants the egg and pollen cells are formed from somatic lineages, changes such as these in somatic cells can be inherited meiotically to alter the potential activity of an element in the next generation. The extent to which inherited epigenetic changes are a source of genetic variation in evolution needs to be assessed. Methylated cytosine residues have higher mutation rates than non-methylated residues (Anitiquera and Bird, 1988). Therefore, active genes may be shielded from mutation compared with inactive ones.

While some cytosine methylation changes are associated with developmental changes in gene expression it is obvious that a change in the DNA template is insufficient for gene expression - the appropriate transcription factors are also essential. Thus errors in the methylation process would not lead inherently to errors in transcription, although in certain situations this could occur. In this paper we have described methylation variants at specific sites around the HMW glutenin loci and within the large rRNA multigene family of a specific plant. These may result from ‘errors’ and have no consequences in leaf cells either because the transcription factors are not present (glutenin subunit genes) or because the particular sites necessary for transcription are not unmethylated. If several sites must be unmethylated to facilitate access to transcription factors, then the probability of all these being unmethylated by error is very low, and biological aberrations will be similarly low. Therefore the inherited variation that we have revealed around the HMW glutenin locus may be genomic noise of little biological significance. No variation in the levels of HMW glutenins accumulated in seeds has been found to be associated with this variation (J. Rogers, unpublished observations). However, such variation provides a useful marker to study the inheritance of specific chromosome fragments. This is especially valuable in a highly inbred organism where RFLP variation is not as extensive as in other species. Exploitation of cytosine methylation variation in pedigree analysis has already been described for human families (Silva and White, 1988).