Drosophila has a unique mechanism of telomere maintenance. Instead of using the telomerase holoenzyme as most eukaryotes, Drosophila replenishes the telomeres by specific transpositions onto the end of the chromosomes of three retrotransposons, HeT-A, TART and TAHRE [1, 2]. The telomeric retrotransposons are completely excluded from euchromatin and share unique characteristics, possibly linked to their telomeric role, that separate them from their non-LTR counterparts. Orthologues of HeT-A and TART have been cloned and studied from species more than 60 MY distant (D.melanogaster - D.virilis), demonstrating that the telomeric retrotransposons predate the separation of the extant species as well as the robustness and reliability of this mechanism of telomere maintenance [3, 4]. Surprisingly, HeT-A and TART orthologues, although committed to the essential function of telomere replication, are far from being static, and while maintaining their basic structures allow their sequence to change rapidly, evolving faster than euchromatic genes and other retrotransposons . This trend of fast sequence change also results in differences within the same Drosophila species and for the D. melanogaster
HeT-A element two previous studies have suggested the presence of a small number of subfamilies coexisting in the same stock [6, 7].
Previous studies have attempted to classify the genomic copies of the HeT-A element in several subfamilies according to their variability in the 3'UTR  and also in the ORF . These studies found four subfamilies considering ORF variability and two considering 3'UTR variability. Taking into account that those studies were based in a limited number of genomic copies, our first objective was to perform an exhaustive survey at genomic level in order to obtain a more accurate picture of the real variability of the HeT-A element.
Other retroelements also form subfamilies in a given genome, as for example Tnt1 in tobacco and L1 in mammalian genomes [8, 9]. In the case of Tnt1, the different subfamilies have acquired different sequences at their regulatory regions that ensure the expression of a particular subfamily in response to different external factors, widening and diversifying in that way the number of opportunities for transposition . In the case of L1, although remnants of several subfamilies exist in a given genome, only one subfamily seems to be active at a time . Whether the existence of different HeT-A subfamilies has a putative role related to its own survival as a retrotransposon or to its telomeric function is still unknown. Studies comparing the number and dynamics of the different subfamilies between wild type and telomeric mutant stocks are needed to answer this question.
With the completion of the heterochromatic genome project  and the assembly of some telomeres for the particular Drosophila strain used in the sequencing project (isogenic strain 2057 yellow (y
); cinnabar (cn
) brown (bw
) speck (sp
) [7, 13]) it was possible to obtain the first detailed view of the telomere structure in Drosophila melanogaster. Because the telomeric retrotransposons suffer from terminal erosion while being at the end of the chromosome, 5' truncated copies were expected. These two studies actually revealed that more HeT-A copies in the telomeric arrays have maintained ORFs and other regions needed for function than had originally been expected. The existence of functional copies in proximal regions of these long telomere arrays suggests that these interior sequences may be renewed more frequently than previously thought. In this case, the turnover in these arrays does not simply replace terminal sequence lost in DNA replication but is also necessary for rebuilding a large fraction of the telomere when needed . If this were the case, it would be even more important to keep a fair number of HeT-A copies capable of active transposition to replenish or make up new telomeres whenever needed. Alternatively, full length elements in the middle of the array could also be explained by the simultaneous transposition of more than one HeT-A element, or access of HeT-A transposition intermediates to the end of the chromosome when terminal erosion has not yet taken place, for example when the capping complex is disassembled. Finding which HeT-A copies are actively being transcribed and whether transcriptional differences exist in response to subfamily affiliation or specific position in the telomeric array would help to better understand telomere biology in Drosophila.
Besides specialized structures composed of repeated sequences, telomeres are also composed of specific proteins and RNA . Because Drosophila telomere elongation is dependent on elements located within the telomere itself, the presence of RNA at telomeres is not a surprise. Some years ago a non-coding antisense RNA containing telomeric repeats was reported in different mammals and named TERRA (Telomeric-Repeat containing RNA) . TERRA RNA seems to be an important component of the telomeric heterochromatin with different regulatory functions in chromatin as well as direct regulation of the telomerase activity. Similarly to TERRA RNA, the telomeric retrotransposons HeT-A and TART are also transcribed from the antisense strand [17, 18] and a recent report strongly suggests that short RNAs coming from the telomeres might have a role in telomere function, protection and development in Drosophila . Therefore, with the new discoveries of non-coding RNAs being involved in different regulatory functions, as mentioned above, further studies on the potential role for the antisense RNAs of HeT-A and TART at Drosophila telomeres become crucial.
The work presented here aims at better characterizing the extent of HeT-A variability and determining if different HeT-A subfamilies could become more successful in a given stock. For this, we have chosen two different stocks, a wild type stock, Oregon-R, and a mutant stock with longer telomeres, the GIII stock bearing an Oregon-R background where the third chromosome from the Tel-1 mutant was introduced almost 10 years ago . Although the cytological position of the Tel-1 mutation has been determined (3L (69)), it is still unknown the molecular cause responsible for the extremely long telomeres of the GIII stock (approximately ten times longer than in any wild type stock). We have amplified and sequenced from DNA and RNA sources in these two stocks some of the most variable regions inside the HeT-A sequence, and have been able to identify all previously defined HeT-A subfamilies. Moreover we describe five previously unreported HeT-A families, demonstrating that HeT-A variability is even greater than expected. Our results show that most HeT-A subfamilies are actively transcribed and some of the found variations allow us to draw the recent history for a number of HeT-A copies. Finally, we have done a wide study on HeT-A antisense transcription and found that most antisense transcripts suffer different alternative splicing with remarkable conservation. Interestingly, we find that the GIII and Oregon-R stocks mainly differ in which subfamilies contribute the most to antisense transcription. This leads us to suggest that due to the importance of non-coding RNAs in gene and heterochromatin regulation, the reported differences could explain in part the greater expression of HeT-A and ultimately the longer telomeres of the GIII stock.