The findings presented in this report underscore the value of tissue- and stage-specific studies of imprinted domains and have yielded several surprising insights. Our data and studies of other imprinted genes confirm that loss of imprinting in specific tissues is a normal developmental event. We find that strain-specific effects influence expression levels and timing of imprinting loss. We also find two co-existing mechanisms responsible for imprinting within the Kcnq1 domain in the embryo. Importantly, by mapping the higher order structure of a specific region, we reveal a novel mechanism by which an antisense non-coding RNA can regulate transcription.
Kcnq1ot1 is ubiquitously expressed and it had been assumed that it was imprinted in every tissue. In this study, we find that Kcnq1ot1 is biallelically expressed in the heart, with the transition from mono- to biallelic occurring in parallel with the Kcnq1 switch. This loss of imprinting is not seen in other tissues at that stage, indicating that there are mechanisms that shift the control of the region to meet tissue-specific needs. Perhaps the strong cardiac enhancers acting on Kcnq1 also exert influence on the Kcnq1ot1 promoter, activating the previously silent maternal allele.
The maternal methylation mark encompassing the Kcnq1ot1
promoter is not lost during the activation of transcription. This is in line with abundant data showing that gametic methylation is very stable 
, and suggests that the repressive effect of DNA methylation can be overcome. In fact, our data show that maternal transcription of Kcnq1ot1
ncRNA is initiated at an alternative promoter region downstream of the methylation mark. Several examples of alternative promoter usage bypassing DNA methylation have been reported 
, and we propose that this mechanism may be commonly deployed to fine-tune expression of imprinted genes for tissue-specific needs.
Interestingly, the Kcnq1ot1 molecule produced from the maternal chromosome is half the length of the paternal transcript. This could be because expression of maternal Kcnq1 from early stages blocks or competes with the Kcnq1ot1 in some way and impedes completion of the transcript. Another possibility is that an alternative transcriptional termination signal is used. These two mechanisms are not mutually exclusive. Surprisingly, the reactivated ncRNA does not repress Cdkn1c, even though it accumulates locally in a manner similar to the paternal Kcnq1ot1. It remains to be determined if this is because of the diminished length of the RNA or if cardiac cells do not provide repressive co-factors that are necessary for the spread of silencing.
Our data reveal a genetic background effect on the level of expression of Kcnq1, with the CAST/EiJ allele exhibiting higher abundance than the C57BL/6J allele at all stages in the developing heart. In addition to being expressed less abundantly, a C57BL/6J paternal allele is reactivated at a later time point than a CAST/EiJ one. Whether these two features are related needs to be determined, but our own analyses and publicly available genomic data lead us to hypothesize that strain-specific cis-regulatory polymorphisms may explain both phenomena. For example, sequence differences in transcription factor binding or affinity for enhancers could determine higher levels of Kcnq1 expression from the CAST/EiJ allele, and greater accessibility to reactivation when inherited paternally.
In exploring the cardiac regulation of the Kcnq1 gene, our findings delineate two phases, independently of strain-specific differences: one, in early development, when Kcnq1 exhibits imprinted expression independently of Kcnq1ot1; and two, in the neonate, when Kcnq1 is biallelic and its expression levels are modulated by Kcnq1ot1.
Our analysis of the K-term
mutant mouse shows that Kcnq1
expression is imprinted in early cardiac development even in the absence of the ncRNA. This result excludes a role for Kcnq1ot1
in establishing or maintaining repression. Thus, there is an independent mechanism that silences the paternal Kcnq1
allele during early development. This mechanism is cardiac-specific, because absence of Kcnq1ot1
does release Kcnq1
from repression in tissues other than the heart. Although a secondary methylation mark at Cdkn1c
is reportedly dependent on expression of Kcnq1ot1
, the Kcnq1
CG-rich promoter is never methylated. Thus, monoallelic expression of Kcnq1
is likely due to chromatin conformation or trans-acting factors. Perhaps paternal expression of Kcnq1ot1
opens the chromatin and makes a tissue-specific silencer available to factors that repress Kcnq1
. These factors would be present in early development and would disappear upon full maturation of the heart (). Another possibility is that an inhibitory factor (IF) for Kcnq1
has a cognate bind site within a differentially methylated region. If the factor is methylation-sensitive, it would only be able to bind the unmethylated paternal allele, thus rendering the paternal Kcnq1
allele inactive. In fact, a silencing domain (SD) has been delimited downstream of the Kcnq1ot1
promoter, overlapping one of the differentially methylated CG-islands (). Interestingly, our bisulfite sequencing results for the paternal KvDMR show a trend towards acquisition and spreading of methylation that could be explained by loss of the inhibitory factor binding at the SD (Figure S3
Model for regulation of Kcnq1 in the embryonic heart.
Early paternal repression of Kcnq1
may also be due to the presence of CTCF
on the paternal KvDMR 
binding is methylation-sensitive. CTCF
could negatively impact expression of Kcnq1
in early embryonic stages by repressing it directly or blocking it from access to enhancers required at that stage (, inhibitory factor (IF) would be CTCF
). Enhancer activity has been shown for a sequence immediately upstream of the Kcnq1ot1
promoter, a position that would be blocked by CTCF
Comparison of Kcnq1
abundance and domain conformation between wild-type and mutant mice reveals what may be the key role for Kcnq1ot1
in later cardiac development. Our results show that absence of the ncRNA leads to much increased Kcnq1
levels, accompanied by a wider range of promoter contacts with sequences throughout the region. Although several of the regions contacted exhibit the marks of regulatory sequences, the Kcnq1
promoter is only associated with them if Kcnq1ot1
is prematurely terminated. The ectopic interactions reflect a greater flexibility of the chromatin fiber in the absence of Kcnq1ot1
transcription and may contribute to the increased Kcnq1
levels. It is imperative to further elucidate the molecular basis of the tissue-specificity of this effect. Subnuclear localization of the Kcnq1
domain in the presence or absence of the Kcnq1ot1
RNA will be an interesting avenue to pursue, to determine whether the ncRNA can act to tether the region to specific nuclear compartments 
, . We also need to explore whether these mechanisms are more widely used by other ncRNA, especially those that are antisense to coding genes.
Tissue-specificity of Kcnq1ot1
repression has been previously reported, specifically for the neighboring Cdkn1c
. In fact, Cdkn1c
imprinting is maintained in the kidney, liver and lung even with a truncated Kcnq1ot1
. Several studies have also shown lineage-specific imprinting regulation in the placenta 
. Indeed, these observations collectively underscore the need to identify cis
- and trans
-acting factors and the dynamics of their interactions to test hypotheses relative to the tissue-specific effects of Kcnq1ot1
Overall, our results are in accordance with the idea that the Kcnq1ot1
ncRNA itself is not responsible for directly silencing neighboring genes 
. We posit that distinct cis
-acting elements are made available for regulatory contacts depending on whether the act of transcription occurs or not, as previously suggested 
. These associations have an additional layer of regulation that imposes the tissue-specificity of the patterns.
In conclusion, our results paint picture of the regulation at the Kcnq1
domain that is much more complex than previously assumed. Imprinting centers, which regulate extended domains of genes, are limited and counteracted by transcriptional mechanisms that may have evolved in response to the emergence of imprinting. Our studies on the Kcnq1
domain underscore that the expression pattern of a domain is ultimately determined by cooperative and competitive dialogues between enhancers, silencers and insulators 
and that understanding complex genetic loci will require identifying all the regulatory sequences and determining their linear order, tissue-specificity and three-dimensional structure.