Unexpectedly, we have found that more than 87% of MOM1 protein is dispensable for the gene silencing function, according to the functional analysis of a series of deletion derivatives of the MOM1 gene. We have also demonstrated that a “miniMOM1”, comprising 22 N-terminal amino acids, an NLS and 197 amino acids including CMM2, retains silencing activity, as reflected by drastically reduced levels of TSI expression and almost complete transcriptional suppression at a transgenic GUS locus. Therefore, minor contribution of the N-terminal part of MOM1 to its silencing activity seems to be apparent. In addition, a drastic reduction in protein size leading to alterations in physical properties (e.g. a predicted isoelectric point of 5.2 for MOM1 and 8.8 for “miniMOM1” and a change in net charge from −62 to +4.3) can also contribute to the incomplete silencing mediated by “miniMOM”. Nevertheless, the results of MOM1 deletion analysis and the successful replacement of Arabidopsis CMM2 by the CMM2 of poplar provide strong evidence that CMM2 is the most critical element of the MOM1 protein for its silencing function, not only in Arabidopsis but also in other plants. Obviously other domains, also these clearly dispensable for TSI and transgene silencing, may still be required for epigenetic regulation at other, as yet unidentified, target loci.
Although, MOM1 proteins are CHD3 derivatives, the domains shared with CHD3 chromatin remodeling factors apparently became obsolete after the acquisition of CMM2. This is also evident for Pt
MOM1, which has a structure largely similar to CHD3 proteins. For example, the SNF2 domain of Pt
MOM1, shown to be critical for the function of CHD3 proteins, acquired mutations of conserved amino acids essential for CHD3 activity 
). Several indispensable amino acids are replaced in MOM1 homologs from different plant species and, remarkably, these replacements are identical in MOM1 proteins from different plant species. It is difficult to provide a simple explanation for this unusual sequence drift since the loss of remodeling functions of SNF2 should not be under a direct, strong selection pressure for particular types of mutations. In any case, the pattern of these mutations provides specific signatures to MOM1 SNF2 domains (Figure S5
) and suggests that acquisition of CMM2 and degeneration of SNF2 occurred in species ancestral to vascular plants. The SNF2 domain of Arabidopsis
MOM1 underwent the most drastic alterations due to an internal deletion. This relatively recent event seems to be accompanied by the formation of the RS1 sequence duplication. Alignment of Arabidopsis
and poplar sequences flanking RS1 suggests that extensive deletion and the formation of RS1 removed not only part of the SNF2 domain but also a PHD finger and chromodomains. Clearly, this event provides the best illustration of the dispensability also of the PHD finger, and chromodomain for the MOM1 silencing function.
CHD3 proteins of human and Drosophila
, known as Mi-2, act as components of a multi-subunit chromatin repression complex NuRD (Nu
emodelling and D
eacetylating), which combines nucleosome remodelling and histone deacetylation activities 
. The Arabidopsis
genome encodes two CHD3-like proteins: PKL with a potentially functional SNF2 domain and the still uncharacterized At5g44800 with an SNF2 domain containing mutations in amino acids essential for chromatin remodeling activity (Figures S5
and data not shown). PKL is involved in transcriptional repression of genes that are active only at a particular time and place during sporophyte development 
. However, there is little evidence at present for the involvement of chromatin remodeling and histone deacetylation in PKL-mediated gene repression.
The exact mechanism of MOM1-mediated silencing is not known, but MOM1 and PKL both seem to contribute to transcriptional suppression or restriction of levels of ectopic reactivation of TSI transcription. The multilayer nature of epigenetic regulation and the necessity for backup mechanisms have been documented recently for Arabidopsis
gene silencing associated with DNA methylation changes 
. However, in this case, interaction between the major and evolutionary highly conserved gene silencing mechanisms, such as DNA and histone methylation, was investigated and the backup deficiencies were found to have very drastic developmental consequences indicative of the destabilization of central epigenetic functions. The effects of pkl
and the combination of these mutations have much more subtle effects. This can be explained by the characteristics of MOM1 targets and their association with bivalent epigenetic marks. The number of such loci is low 
and their reactivation is likely controlled at multiple levels, as illustrated here by the cooperative activities of MOM1 and PKL. It is remarkable that despite the clearly divergent evolution of MOM1 in terms of protein properties, it has retained its functional relationship to the CHD3 proteins. The CHD3 origin of MOM1 and the silencing in cooperation with PKL suggest that MOM1 function is also linked to histone acetylation changes. Although global changes in histone acetylation properties were not observed in mom1
, more subtle target-specific acetylation changes cannot be ruled out.
Whatever the precise molecular mechanism(s) of heritable transcriptional repression mediated by MOM1 might be, it is remarkable that increasing complexity of epigenetic gene regulation has resulted from the emergence of supplementary and cooperating levels and/or mechanisms of epigenetic control. Genomic or cDNA sequences encoding CMM2 are present in many species of vascular plants, even as distant as club-moss Selaginella moellendorffii
, pine Pinus taeda
and various monocotyledonous and dicotyledonous species. Remarkably, we failed to detect CMM2 in the moss Physcomitrella patens
or in green algae Chlamydomonas reinhardtii
, although the sequences of both genomes are complete (http://genome.jgi-psf.org/
). Therefore, it appears that the emergence of CMM2, and thus MOM1 proteins, coincided with the appearance of vascular plants. Since MOM1-related proteins are not present outside of the plant kingdom, it can be envisaged that novel, highly specialized epigenetic factors and functions can appear only in a narrow subset of organisms through diversification of the general, evolutionary conserved epigenetic regulators. So far the biological role of the CHD3/MOM1 sub-diversification remains unclear but it is intriguing that it seems to have assisted the major evolutionary step in the emergence of land plants.