Let's now consider the second model: noncoding RNA

This is exemplified by the maternally imprinted Igf2R locus, in the Figure. 

In this case, the locus contains: a gene for a long noncoding RNA, called Airn, which is partly overlapped, on the opposite strand, to the Igf2R gene. At the level of the Igf2R gene promoter, there is a secondary DMR. The ICR is exactly placed at the start of Airn and, indeed, it represents the Airn promoter.  On the right side, there are other three genes, encoding for transporters, two of which are expressed only maternally. 

How does this locus function? When the ICR is methylated (Black circle, in the maternal allele), this ICR will not function as a promoter (indeed, it is a CG-rich hypermetylated inactive promoter) and the Airn long noncoding RNA will not be produced. In this condition, the promoters of Igf2R, Slc22a2 and Slc22a3 are activated, and direct transcription of these genes.

Conversely, when the ICR is unmethylated (paternal allele), it works as a very active promoter, directing transcription of Airn. In this condition, promoters of Igf2R, Slc22a2 and Slc22a3 are heterochromatic, inactive.

How can transcription of a long noncoding RNA have such an effect on these genes ?

Researchers made different experiments to understand this. One possibility was that active transcription in the antisense direction (i.e. transcription of the Airn gene) was incompatible with transcription in the other sense (since Airn and Igf2R are in part overlapping). To explore this possibility, they deleted the leftmost part of Airn gene and found that this abrogates silencing of Igf2R and other genes.

Is this result compatible with the "transcriptional interference" hypothesis ?