There are LINE-1 (L1) retrotransposons in most mammalian genes, but these are mainly found in the introns, so it was assumed that they rarely influenced gene expression. However, new data show that these elements of selfish DNA can in fact affect transcription of endogenous genes. So, rather than just being quiet non-contributors, these new findings paint a picture of these elements as the disruptive naughty schoolchildren of the genomic classroom.

Despite L1's abundance in mammalian genomes, L1 trancripts have been difficult to detect. Given the succesful expression of L1 in non-mammalian systems, this difficulty indicated that there might be a mammal-specific mechanism for suppressing LI transcription. To investigate this possibility, Jeff Han and colleagues fused one of the two human L1 open reading frames (ORF2) to a GFP ORF and then measured levels of RNA expression relative to a control lacZ/GFP fusion. They found that the ORF2 sequences caused a large decrease in gene expression, whether they were in sense (GFPORF2) or antisense (GFPORF2AS) orientation.

Interestingly, the authors did find a difference between sense and antisense constructs in the way that they achieved this transcriptional downregulation. Cloning and sequencing of the abundant lower molecular mass species that is expressed when ORF2 is in the antisense orientation indicated that premature polyadenylation accounted for the shortfall in full-length transcripts. By contrast, premature polyadenylation only accounted for 15% of the decrease in full-length transcripts from GFPORF2.

Measurement of the RNA half life and real-time PCR showed that an increase in RNA degradation could not account for the reduction in GFPORF2 full-length transcripts. Neither did ORF2 inhibit transcriptional initiation, because a nuclear run-on assay indicated that the polymerase density in the early region of GFPORF2 transcripts was equivalent to that of the lacZ control. However, the same assay revealed that polymerase density decreased as it was assayed further along the ORF2 sequence: so, ORF2 seems to suppress gene expression by inhibiting transcriptional elongation.

Han and colleagues followed up their experiments with a neat bit of bioinformatics. They compared the amount of L1 sequence in the genes with the highest and lowest levels of expression in humans. The results were striking: highly expressed genes contained small amounts of L1 sequence, whereas genes with low levels of expression contained large amounts of L1 sequence. These patterns hold even when differences in total intron content or isochore location are accounted for.

The dovetailing of the authors' experimental and bioinformatic data provides extremely strong circumstantial evidence that L1 insertions decrease gene expression in vivo. Moreover, this new work tallies with previous findings that gene expression is suppressed when full-length L1 sequences are inserted into introns.

The authors also raise the intriguing possibility that the mammalian genome might have co-opted transcriptional suppression of L1 (which probably first evolved to prevent excessive mutagenic retrotransposition) as a mechanism for fine-tuning the relative levels of expression of different genes. If this model is correct, perhaps L1 is better viewed as more of a genomic classroom assistant than a problematic pupil?