We have demonstrated for the first time by allele specific 4C-Seq and by extensive DNA FISH analysis that many loci pair with their homologous allele. Pairing is not limited to regions of mono-allelic expression, involves larger chromosomal regions and brings the two homologous chromosomes into close proximity. While pairing events did not coincide with DNA repair, they took place at sites of ongoing transcription.
Homologous pairing has been implicated in the establishment of mono-allelic silencing of the X chromosome. Indeed, during a period of high chromatin mobility, the two X inactivation centres ‘kiss’ which is followed by transient downregulation of Tsix
on one allele, thereby creating a window of opportunity for mono-allelic expression of Xist
. Pairing of homologous alleles is also observed at immunoglobulin loci. One of the paired alleles undergoes RAG mediated cleavage while the other unrearranged allele becomes associated with pericentromeric heterochromatin 
. These two functionally different examples of mono-allelic expression have in common that one of two equivalent genomic copies is chosen at random for expression. This choice requires some kind of trans
-allelic cross talk to ensure that one but only one allele gets inactivated. However, for imprinted regions the situation is different. Here, each allele comes in pre-marked and there is no immediate requirement for communication between homologous alleles. Nevertheless, short interallelic distances were observed in late S phase for the Prader-Willi/Angelman imprinted region in human 
, although this effect was argued by others to be due to the presence of a nucleolus organising region on the same chromosome 
. Here we report high pairing frequency for the Kcnq1
and adjacent Igf2/H19
clusters in the mouse, but not for a number of other imprinted clusters. Pairing at distal chromosome 7 was not limited to the imprinted region, and in fact loss of imprinting did not change pairing frequency. Thus, we conclude that homologous pairing is not a general feature of mono-allelically expressed regions. However, this does not preclude an involvement of pairing and trans
-allelic effects on the regulation of imprinted regions. In fact, it was shown that introducing a third copy of human chromosome 15 disrupted pairing and affected gene expression at the Prader-Willi/Angelman region 
. Speculatively, at the large Kcnq1
domain which is silenced by a coating RNA, homologous pairing might be involved in the escape of imprinting of several interspersed biallelically expressed genes, especially as we find that pairing is associated with transcription.
Homologous pairing has also been speculated to be linked with DNA repair. The genome is constantly challenged by double strand breaks (DSB) brought about by internal or external chemical insults or the collapse of stalled replication forks (for review see 
). These lesions can either be repaired by non-homologous end joining (NHEJ) or homologous recombination (HR). Which repair pathway is used depends on the organism and what caused the double strand break. While HR predominates in yeast, NHEJ plays a more important role in vertebrates. Still, in mammals HR is a common mechanism to repair replication induced damage after fork collapse which leaves a single double strand end. In this scenario the sister chromatid can be used as a template for strand invasion and restart of replication, a process that is helped by sister chromatid cohesion 
. While it can be envisioned that more severe replication blocks may be repaired by HR involving both homologues, we did not find any evidence that links the homologous pairing described here with DSB or HR repair.
A number of recent genome-wide interaction studies in human cells have demonstrated the presence of topologically distinct active and repressive compartments, with trans
associations happening preferentially between transcriptionally active regions 
. Moreover, a high frequency of trans
contacts correlated well with the region’s distance to the edge of the chromosome territory 
. As these studies were not performed in an allele-specific manner, no information about homologous contacts can be drawn. However, it seems likely that for a given region the criteria for a high potential of trans
interactions, namely transcriptional activity and location close to the edge of the chromosome territory, will also apply to homologous associations. In line with our results, this suggests that not transcription of individual genes but large-scale active features of a region determine a regions propensity to form homologous and non-homologous associations.
Although evidence about regional pairing of homologous chromosomes has increased over recent years, it still remains largely unclear how the two homology partners find each other in the crowded nucleus 
. In one model, transcription is the driving force: Transcribed genes are located in transcription factories, organising the linear sequence into nodes and intervening loops. By existence of specialised transcription factories a chromosomal transcription signature is created. Because homologues share the same signature, contact at one node increases the probability of larger regions coming together 
. Indeed, it was recently reported that pairing of the Prader-Willi/Angelman region was reduced upon inhibition of transcription 
. Our results that pairing frequency is correlated with expression, and that pairing events are located in regions of high RNA PolII activity are in line with this hypothesis. More speculatively, our observation that chromosome territories of paired KvDMR alleles can either be touching at the ends, or be partially or fully aligned, might suggest that once homologous contact has been established in one region, chromosomes have the potential to progressively button up along their whole length.
The data presented here suggest that the frequency by which homologous regions pair is determined by several factors, of which we have identified chromosomal position and transcription, with transcription potentially being the driving force of bringing the two homologous together. This could provide the cell with the potential for another layer of regulation: exchange of information in trans
. Interestingly, homologous trans
effects have been reported for multiple loci including imprinted regions. Several studies report introducing mutations into one of the alleles of either the Igf2
or Prader-Willi/Angelman region and finding an unexpected effect on expression of the second allele 
, suggesting that regulatory elements might be functioning in trans
to enhance or supress transcription. Cross-talk is however not limited to transcriptional regulation but has also been shown to affect allelic methylation. Targeting of the unmethylated paternal Snrpn
gene in ES cells was frequently associated with full or partial loss of methylation on the maternal allele 
. Interestingly, allelic methylation was stable when the targeting construct was integrated at heterologous loci, suggesting that both homologues were required to observe a methylation effect in trans
. Similarly, deletion of the unmethylated maternal H19
gene led to reduced methylation of the paternal Igf2
. Vice versa, a mutant Rasgrf1
allele not only attracted methylation to the affected paternal allele, but also in trans
to the maternal allele 
. This methylation mark was stable through meiosis and therefore resembles paramutation. Notably, all of the above examples involve imprinted loci. However, only imprinted loci are routinely analysed in an allele specific manner and other trans
effects might have been missed. Indeed, plasmid DNA containing the beta-globin gene has been shown to physically pair with the homologous region and to transinduce transcription of nearby sequences 
. In contrast, transactivation was not observed between the beta-globin LCR and its target gene when integrated into the same ectopic site on different chromosomes 
. This suggests that while pairing events do not necessarily lead to a change in transcriptional output, they have the potential to do so in certain situations.
Taken together, we propose a model in which not the expression state of individual genes but rather the transcriptional signature of large chromosomal domains can bring homologous regions together. Since global chromosomal movements are constrained this might only be possible in a subset of cells which feature a permissive subnuclear arrangement of chromosomes after the last mitosis. Transient allelic interactions in paired regions could then be stabilised to become functionally relevant. Such close proximity could open up the possibility of allelic cross-talk and transcriptional regulation in trans
, which may in certain circumstances affect normal development and the manifestation of genetic susceptibility to diseases