Control of Telomere Length in Drosophila

The problem of incomplete end replication of DNA was originally raised by the Russian scientist Aleksey Olovnikov (Olovnikov, 1971, 1973). The main function of telomeric DNA is the compensation of end degradation. In most eukaryotes, telomeric DNA is maintained by the action of telomerase, which is responsible for the synthesis of short 6-9 nucleotide repeats using an RNA component as a template (Greider and Blackburn, 1985). In contrast, telomeres of Drosophila are maintained as a result of retrotransposition of specialized telomeric non-long terminal repeat retrotransposons (Biessmann et al., 1990a; Biessmann et al., 1992a; Levis et al., 1993; Abad et al., 2004b).


Introduction
The problem of incomplete end replication of DNA was originally raised by the Russian scientist Aleksey Olovnikov (Olovnikov, 1971(Olovnikov, , 1973. The main function of telomeric DNA is the compensation of end degradation. In most eukaryotes, telomeric DNA is maintained by the action of telomerase, which is responsible for the synthesis of short 6-9 nucleotide repeats using an RNA component as a template (Greider and Blackburn, 1985). In contrast, telomeres of Drosophila are maintained as a result of retrotransposition of specialized telomeric non-long terminal repeat retrotransposons (Biessmann et al., 1990a;Biessmann et al., 1992a;Levis et al., 1993;Abad et al., 2004b).
Retrotransposons are also found in telomeric regions of such organisms as the silkworm Bombix mori (Okazaki et al., 1995;Takahashi et al., 1997), the green alga Chlorella vulgaris (Higashiyama et al., 1997) and a flagellated protozoan parasite Giardia lamblia (Arkhipova and Morrison, 2001). In Bombyx mori and Chlorella, there is a mixed type of telomere elongation: telomeric retrotransposons are inserted into telomerase-generated sequences (Fig.1). In Drosophila genomes, no telomerase orthologs have been found (Osanai et al., 2006). Elongation of Drosophila telomeres is mediated by specialized telomeric retroelement transpositions onto chromosome ends (Biessmann et al. 1992a; Levis et al. 1993). Recombination represents a bypass mechanism for chromosome length maintenance (Mikhailovsky et al. 1999;Kahn et al. 2000).
This review is focused on the mechanism of telomeric transposition control, which is a crucial step in Drosophila telomere elongation. Telomeric retroelements are arranged in Drosophila telomeres in mixed tandem head-to-tail arrays, with their 3' ends orientated toward the centromere. Telomeric element transcripts serve as a template for transposition according to target-primed reverse transcription. In this case, as well as in telomerase encoding organisms, telomere elongation utilizes reverse transcription, i.e., synthesis of DNA from an RNA template. Drosophila represents a unique model system with an alternative mechanism of telomere maintenance. A characteristic feature of Drosophila telomeres is that the RNA template for telomere elongation is encoded by the telomeric sequences themselves, in contrast to a telomerase RNA component encoded by a separate cellular gene. Regulation of the activity of genes encoding telomerase and the RNA template as well as changes in concentration of the telomere repeat binding proteins play a crucial role in telomere length control in organisms that use telomerase. To understand the mechanism of length control of Drosophila telomeres, it is important to know how the addition of retrotransposon elements onto chromosome ends is regulated. This process is directly associated with the transcriptional regulation of telomeric retrotransposons and chromatin structure in the telomeric region. Components of the telomere capping protein complex are involved in Drosophila telomere length control. Recent studies of Drosophila telomeres have demonstrated the importance of the RNA interference (RNAi) pathway in Drosophila telomere homeostasis. In the Drosophila germline, retrotransposons are silenced by the PIWI-interacting RNA (piRNA) pathway ). The telomeric retroelements HeT-A, TART and TAHRE, which are involved in telomere maintenance in Drosophila, are also the targets of the piRNA-mediated silencing. The abundance of telomeric retroelement transcripts, both sense and antisense, as well as the frequency of their transpositions onto chromosome ends, are controlled by a piRNA-mediated mechanism (Savitsky et al., 2006;Shpiz et al., 2007). piRNAs induce transcriptional silencing of the telomeric retrotransposons (Shpiz et al., 2011), suggesting a putative role for the piRNA pathway in the formation of the telomeric chromatin that protects chromosome ends from fusion and is involved in meiotic and mitotic telomere behavior. Fig. 1. Different modes of telomere elongation. (A) Telomeric DNA is maintained by telomerase, which synthesizes short 6-9 nucleotide arrays using an RNA component as a template. The human-specific telomeric repeat is shown. (B) A mixed type of telomere elongation: the telomeric retrotransposons are inserted into the short repeats formed by telomerase (Bombyx mori, Chlorella). (C) Drosophila telomeres are formed as a result of retrotransposition of specialized telomeric retrotransposons. Telomere-associated sequences (TASs) are indicated.

Structure of Drosophila telomeres
Telomeres are nucleoprotein complexes localized at the ends of linear chromosomes. Based on this, Drosophila telomeres can be subdivided into three domains Biessmann et al., 2005;Frydrychova et al., 2008) (Fig.2). The very end of the chromosome is protected by a special protein complex, the so-called telomeric cap. This structure prevents chromosome ends from end-to-end fusions and degradation by DNA repair mechanisms. The second domain accommodates the telomeric retrotransposon arrays that replace telomerase-generated repeats and supports chromosome end elongation. Lastly, proximally located repetitive complex sequences form subtelomeric or telomere associated sequences (TASs). These domains are characterized by specific chromatin structures that ensure proper telomere functioning.

Drosophila telomeric retrotransposons
The telomeres of Drosophila melanogaster consist of the specialized telomeric retrotransposons HeT-A, TART and TAHRE (Biessmann et al., 1992a;Levis et al., 1993;Abad et al., 2004b). They are LINE (long interspersed nucleic elements) or non-LTR (long terminal repeat)-type retroelements. Spontaneous transpositions of HeT-A, TART and TAHRE to telomeres have been observed, indicating that all three retrotransposon families participate in Drosophila telomere maintenance (Biessmann et al., 1992b;Sheen and Levis, 1994;Kahn et al., 2000;Golubovsky et al., 2001;Shpiz et al., 2007). HeT-A, TART and TAHRE are present at Drosophila telomeres in mixed tandem head-to-tail arrays; their oligo(A) tails always face proximally, towards the centromere. LINE elements use a target-primed reverse transcription mechanism for their transposition (Luan et al., 1993). It has been suggested that Drosophila telomeric retrotransposons can use the 3' protruding end of the chromosome as a primer, but the mechanism of site-specific transpositions of telomeric retroelements to the chromosome end remains unclear. Telomere targeting of retrotransposon mRNA is the most crucial stage in telomere elongation. It has been proposed that Gag proteins target the telomeric retroelement mRNA to chromosome termini in cultured Drosophila cells (Rashkova et al., 2002;Rashkova et al., 2003;Fuller et al., 2010).
HeT-A, the most abundant Drosophila telomeric element, contains a single open reading frame (ORF) encoding a Gag-like RNA-binding protein but lacks reverse transcriptase (RT). TART has two ORFs, encoding the Gag and Pol proteins. The ORF2 has both endonuclease and RT domains. Both elements have unusually long 3' and 5' untranslated regions (UTR).
TAHRE as well as TART has two ORFs ; ORF2 of TAHRE is similar to that of TART. The 5' UTR, ORF1 and 3' UTR of TAHRE are similar to the corresponding regions of HeT-A, which prompted the designation of a newly discovered telomeric element TAHRE (Telomere-Associated and HeT-A-Related Element) (Abad et al., 2004b). It was proposed that a putative ancestral element evolved to provide telomere maintenance in Drosophila ( Fig.2A). TART and TAHRE diverged from a common ancestor. HeT-A lacks ORF2 and may have derived from a processed copy of TAHRE (Abad et al., 2004b). All of the analyzed D. melanogaster stocks have both HeT-A and TART elements, and the copy numbers are approximately 30 HeT-A and 10 TART per genome (Abad et al., 2004a;George et al., 2006). A single complete and three truncated TAHRE were identified in the genome of the Drosophila stock sequenced by the Genome Project (Abad et al., 2004b). Obviously, the structural role is not a primary function for TART and TAHRE, as some telomeres contain neither TART nor TAHRE elements (Levis et al., 1993;Abad et al., 2004a;Shpiz et al., 2007). It has been proposed that the RT activity necessary for HeT-A transposition might be provided by TART or TAHRE (Levis et al., 1993;Rashkova et al., 2002;Abad et al., 2004b). It is noteworthy that autonomous and nonautonomous telomere-specific retrotransposons were described in the genomes of evolutionary distant Drosophila species (Villasante et al., 2007). One of the reasons for the cooperation of several telomeric elements throughout evolution may be the distribution of different roles among elements.
Most of the non-LTR retrotransposons utilize an internal 5'UTR promoter to transcribe a full-length RNA that serves as a template for transposition. An unusual feature of the telomeric retroelements is that the promoters of HeT-A and TAHRE are localized in the 3' UTR and drive transcription of a downstream element (Danilevskaya et al., 1997;Shpiz et al., 2007). An antisense promoter was detected in close proximity to the HeT-A promoter, which drives sense expression (Shpiz et al., 2009). It appears as if the common promoter drives bidirectional expression of HeT-A. The TART element was shown to also be transcribed bidirectionally from internal sense and antisense promoters that are localized within non-terminal direct repeats in the TART 5' and 3' regions (Danilevskaya et al., 1999;Maxwell et al., 2006). An unusual feature of non-coding HeT-A and TART antisense transcripts is that they are spliced (Maxwell et al., 2006;Shpiz et al., 2009). The role of antisense transcripts in the RNA silencing of the telomeric retrotransposons will be discussed below.
According to cytological and genetic data, chromatin in the region of telomeric retrotransposon arrays exhibits euchromatic characteristics. There are several examples of white transgene integration into coding and promoter regions of HeT-A, TART and TAHRE. In most cases, when the insertion is far from a TAS, the normal activity of a reporter gene is observed . The longer the HeT-A array is upstream of the telomeric white transgene located between the TAS and telomeric retroelements, the higher the expression of the reporter gene (Golubovsky et al., 2001). This indicates that telomeric retrotransposon arrays may activate the expression of nearby genes; however, the mechanism of this trans-activation remains unknown. Cytological studies have also characterized the region of telomeric retroelements as a zone of decondensed chromatin similar to euchromatin Biessmann et al., 2005). It is noteworthy that in organisms encoding telomerase, proteins binding to the telomerase-generated repeats form a heterochromatic silencing complex (Schoeftner and Blasco, 2009).

Structure of the Drosophila capping complex
The protein complex at the chromosome end forms a cap that protects DNA ends against the repair system and prevents telomere fusions. The formation of the Drosophila cap does not require specific telomeric sequences at the chromosome ends. Terminally deleted chromosomes in the absence of telomeric and subtelomeric sequences may form a cap as well as natural telomeres (Fanti et al., 1998;Perrini et al., 2004). Several proteins that protect Drosophila telomeres from end-to-end fusion events have been identified . Among them are HOAP (Heterochromatin Protein 1/origin recognition complex-associated protein), HipHop, Moi (Modigliani) and Ver (Verrocchio), which are the founding components of a Drosophila capping complex (Cenci et al., 2003;Raffa et al., 2009;Gao et al., 2010;Raffa et al., 2010). These proteins are highly enriched at telomeres. The HOAP/Ver/Moi complex is a functional analog of shelterin, a protein complex that protects human chromosome ends (Palm and de Lange, 2008). This complex was named "terminin" (Raffa et al., 2009). Terminin accumulation at chromosome ends prevents telomere fusion and helps in recruiting nonterminin components of the Drosophila capping complex. It was proposed that Drosophila lost the shelterin that binds telomeric DNA in a sequence-specific fashion and evolved terminin to bind chromosome ends independent of the terminal DNA sequence. HP1 (heterochromatic protein 1) is another important structural component of the Drosophila chromosome cap (Fanti et al., 1998). In addition to the chromosome cap, this protein is also associated with centromeric regions and many euchromatic sites. Mutations in the HP1 coding gene cause aberrant chromosome associations and telomeric fusions in neuroblast cells, imaginal disks, early embryos and male meiotic cells, providing evidence that HP1 mediates normal telomere behavior in different Drosophila cells and tissues (Fanti et al., 1998). Mutations in genes encoding ATM and ATR kinases, which are the main enzymes of the cell response to DNA damage, and components of the MRN repair complex also cause telomere fusions (Bi et al., 2004;Ciapponi et al., 2004;Oikemus et al., 2004;Silva et al., 2004;Song et al., 2004;Bi et al., 2005). However, these proteins are not stable cap components and, most likely, they mediate HP1/HOAP/Moi recruitment at chromosome ends ( Fig.3). It has been suggested that transient interactions between DNA repair factors and terminal DNA sequences facilitate association of HOAP-HP1 with chromosome termini, which in turn would recruit Moi. A specific feature of the cap assembly is the redundancy of the pathways that affect capping protein recruitment to the telomere. For example, telomeric localization of both HP1 and HOAP requires functional Mre11/Rad50 components of the MRN complex. However, even in the absence of this complex, mitotic chromosomes retain the ability to recruit low levels of HOAP (Ciapponi et al., 2004). ATM and ATR kinases act redundantly in telomere protection (Bi et al., 2005). ATM may fully compensate ATR's absence in telomere capping while ATR may partially compensate for the loss of ATM (Bi et al., 2005). The mutations in the effete and woc genes, which encode an ubiquitin conjugating enzyme and a putative transcriptional factor, respectively, result in Drosophila telomere fusions; however, their role in telomere capping is still unclear (Cenci et al., 1997;Raffa et al., 2005). Heterochromatic proteins such as HP2 (a partner of HP1), Su(var)3-7 (a protein of pericentromere chromatin containing zinc fingers) and SUUR (a protein localized in late replicated regions, which is typical for heterochromatin) bind to polythene chromosome ends in the salivary glands , suggesting that cap represents condensed chromatin structure. The repressive effect of the cap on the expression of terminally located genes extends as far as 4 kb (Mikhailovsky et al., 1999;Melnikova et al., 2004). Thus, in natural telomeres, only very distal telomeric retroelement may be affected by the capping complex.
Capping complexes might control telomere length by affecting the accessibility of the chromosome termini for telomeric element transposition. This process is of great importance in germinal cells. However, the majority of available data regarding cap structure was obtained by inspecting somatic cells, such as interphase salivary gland cells and mitotic larval brain cells. It should be noted that cap structure and the mechanisms of its assembly appear to differ in distinct cells and/or during the cell cycle. For example, a mutation in the gene tefu encoding ATM did not change the localization of HOAP at the telomeres of mitotic chromosomes, but it did decrease the amount of HOAP and HP1 at the telomeres of salivary gland polytene chromosomes (Bi et al., 2004;Oikemus et al., 2004). No detailed information about the cap structure of the germ cell chromosomes has been obtained. HP1 has been found at the ends of normal chromosomes as well as of terminally deleted chromosomes in the nuclei of nurse cells in Drosophila ovaries by immunohistochemistry (Shpiz et al., 2011) . Chromatin immunoprecipitation (ChIP) analysis revealed binding of both HP1 and HOAP with HeT-A in ovaries (Khurana et al., 2010). These data indicate that at least HP1 and HOAP provide chromosome capping in ovarian cells. Recent studies revealed a role of the piRNA pathway components in the assembly of the telomere capping complex in early embryogenesis (Khurana et al., 2010). piRNAs are essential for the retrotransposon silencing in the germline (Vagin et al., 2006;Brennecke et al., 2007;Aravin et al. 2007). armi and aub piRNA pathway components, which encode a putative RNA helicase and a piRNA binding Argonaute protein, respectively, are needed for telomere resolution during mitotic divisions in early embryos (Khurana et al., 2010). Maternally deposited piRNA pathway proteins likely function at this developmental stage (Brennecke et al., 2008). ChIP analysis on mutant armi and aub embryos revealed reduction (but not elimination) of the HOAP and HP1 binding to HeT-A, indicating the role of these components in telomere cap assembly. It is noteworthy that neither ago3 nor rhi piRNA components affect HOAP or HP1 binding to HeT-A. A subpopulation of HeT-A-specific piRNAs was proposed to direct assembly of the telomere cap (Khurana et al., 2010). However, in Drosophila, chromosome capping is a sequence-independent process. aub and armi likely provide a redundant pathway for the recruitment of telomere capping proteins in early embryogenesis.

Subtelomeric region
The TAS is an extended heterochromatic chromosome region that is proximally adjacent to telomeric repeats in different organisms, the function of which remains unclear. In Drosophila, TASs consist of 15-25 kb of satellite-like repeats or fragments of mobile elements. TASs differ in length and sequence among chromosome arms, although some repeats share similarity. 2L TAS has homology with 3L TAS, while X TAS shares homology with 2R and 3R TAS. For example, TASs of the 2L and 3L chromosome arms are similar 460 bp repeats. X TASs are more complex repeats that are 0.9 and 1.8 kb in size (Karpen and Spradling, 1992;Wallrath and Elgin, 1995). The small fourth chromosome of Drosophila has a special TAS structure that is made of fragments of various mobile elements (Cryderman et al., 1999). Drosophila salivary gland polytene chromosome subtelomeric regions look cytologically more condensed when compared with a region resembling euchromatin and containing telomeric retrotransposons Biessmann et al., 2005). The eye color reporter gene white becomes silenced after insertion into a TAS (Karpen and Spradling, 1992;Cryderman et al., 1999;Golubovsky et al., 2001). Inactivation of a reporter gene inserted into a subtelomeric region is referred to as telomeric position effect (TPE). Transgenes in a TAS or close to a TAS demonstrate repressed and variegated expression. The silencing effect spreads to the limited distance from the TAS into the telomeric retrotransposon array (Frydrychova et al., 2007). As mentioned above, the reporter genes in the terminal HeT-A, TAHRE, or TART retroelements do not exhibit repressed expression .
In contrast to the classic position effect related to the suppression of a gene integrated into pericentric heterochromatin, Drosophila TPE does not depend on HP1 and known modifiers of the position effect. An exception is the fourth chromosome, where TPE depends on HP1 (Cryderman et al., 1999). This is not surprising because the fourth chromosome is cytologically located in the chromocenter, a region enriched with heterochromatic proteins. The TAS regions of most chromosome arms contain motifs recognized by the repressors of the Polycomb (PcG) protein family (Cryderman et al., 1999;Boivin et al., 2003;. The chromatin of the TAS region is enriched with methylated Lys27 of histone H3 (H3Me3K27) due to the activity of the histone methyltransferase E(z) (Czermin et al., 2002;Andreyeva et al., 2005). E(z) is a member of the Polycomb group proteins and exhibits chromomethylase activity toward Lys9 and 27 of histone H3 (Czermin et al., 2002). Methylation at these positions is a label that attracts PcG repressor proteins, which results in gene silencing. Interestingly, translocations of the second chromosome TAS to the fourth chromosome and vice versa retained the structure of the chromatin and TPE features in the translocated regions (Cryderman et al., 1999). The translocation of the fourth chromosome TAS to the second chromosome caused a sharp decrease in TPE-induced transgene silencing while dependence of TPE on HP1 within the translocated fragment was retained (Cryderman et al., 1999). This phenomenon may be attributed to changes in nuclear localization of the translocated fragment. Indeed, the fourth chromosome is usually positioned in the chromocenter that is enriched with heterochromatic proteins, whereas translocation transfers it to the periphery of the nucleus. These results suggest that TASs provide sequence-specific assembly of protein complexes involved in larger protein nuclear compartments. In addition, flies with a TAS deficiency on chromosome 2L show an increase in the expression level of telomeric white transgenes located both at homologous and nonhomologous chromosomes, suggesting long-range telomere communication in the nucleus (Golubovsky et al., 2001;Mason et al., 2003;Frydrychova et al., 2007). Based on these data, it is tempting to propose that TASs and associated proteins play a key role in processes related to telomere positioning inside the nucleus.

Mechanisms of Drosophila telomere length control
Telomeres formed by telomerase consist of short repeats like TTAGGG in vertebrates. These repeats form a double-stranded DNA sequence of several thousand base pairs and a single strand 3' overhang of several hundred nucleotides in length. Embedding of the single strand end into the telomeric DNA duplex forms a telomeric t-loop (de Lange, 2004). The telomeric repeats as well as the t-loop are recognized by specific DNA-binding telomeric proteins, representing a platform for assembly of the telomere-protein complex. This complex stabilizes the structure of the t-loop, protects the chromosome end against degradation and controls telomere length by regulating the accessibility of the chromosome end for telomerase (Smogorzewska and de Lange, 2004). In vertebrates, conservative DNA-binding telomeric proteins TRF1 and TRF2 directly bind the duplex of telomeric repeats and recruit other proteins to the telomere (such as the tankyrases and RAP1). TRF1 and TRF2 also associate with proteins of the DNA repair system including the heterodimer Ku70/Ku80, the MRN complex (MRE11, RAD50, NBS1) and ATM-kinase (Chan and Blackburn, 2002;Goytisolo and Blasco, 2002;Smogorzewska and de Lange, 2004). Several dozen proteins are involved in the formation of the telomere complex in mammals. The mode of the complex formation is similar for different organisms and includes sequence-specific recognition of telomeric repeats.
The heterogeneity of terminal sequences distinguishes Drosophila telomeres from the telomeres formed by short repeats. The transposition of three various retroelements to the chromosome end and the process of end degradation results in varied terminal nucleotide sequences, even between chromosomes of the same individual. It remains unclear whether Drosophila telomeres terminate with the long 3' overhang and form the t-loop like structure. Indirect evidence for this possibility is that Drosophila telomeres can be elongated as a result of the recombination of tandem repeats located on the same chromosome (Kahn et al., 2000;Savitsky et al., 2002;Melnikova et al., 2005). It is possible that recombination is facilitated by integration of the chromosome end into the internal region of terminal DNA. Nevertheless, no evidence for the existence of a stable t-loop in Drosophila has been obtained by molecular or cytological methods. Thus, Drosophila telomeres contain neither specific terminal sequences nor t-loop configuration. Although regulation of telomere length in telomerase-expressing organisms is a complex process that depends on numerous factors, it may be described as protein titration on telomeric DNA: a decrease in telomeric proteins on a shortened telomere increases its susceptibility to telomerase in germinal cells or represents a signal for arrest of somatic cell division (Chan and Blackburn, 2002;Smogorzewska and de Lange, 2004).
The activity of genes encoding telomerase and the RNA template are also important factors regulating telomere length (Nugent and Lundblad, 1998). In Drosophila, no specific signaling pathway responsible for the retrotransposon addition to the shortened telomere has been discovered so far. Taking into account the peculiarities of Drosophila telomere structure, one may suggest that two steps are crucial for its telomere elongation: accessibility of the chromosome end for transpositions and control of the transposition frequency through the regulation of the telomeric retrotransposon expression. Table 1 represents a list of genes or genomic loci that affect Drosophila telomere length that confirm this hypothesis and demonstrate a striking crosstalk between the mechanisms of telomere protection and regulation of telomeric retrotransposon expression.

Role of the capping complex in Drosophila telomere length control
The frequency of spontaneous HeT-A transpositions to the chromosome end has been estimated to be in the range of 10 -5 to 10 -3 (Biessmann et al., 1992a;Kahn et al., 2000). Mutations in the Su(var)2-5 gene encoding the major heterochromatic protein and cap component HP1 increase the addition of HeT-A/TART to the ends of terminally truncated chromosomes by more than a hundred times (Savitsky et al., 2002). HP1 mutations cause significant increases in HeT-A and TART expression. Drosophila lines harboring Su(var)2-5 mutations maintain extremely long HeT-A/TART arrays at the natural chromosome termini (Savitsky et al., 2002;Perrini et al., 2004). Telomere elongation and HeT-A/TART derepression are observed in all studied Su(var)2-5 mutants, but one of these mutations that disrupts the chromodomain does not influence HP1 capping capacity (Fanti et al., 1998;Perrini et al., 2004). It appears that the HP1 silencing effect on telomeric retrotransposon expression is more crucial for the negative control of telomere length than its role in the telomere capping. It should be noted here that the expression of HeT-A/TART and cap formation was studied in the somatic tissues of Su(var)2-5 mutants, while terminal attachments causing telomere elongation that are detectable in the progeny take place in the germ cells. Thus, to elucidate the specific role of HP1 in the control of telomere length, its impact on the telomere biology in the germline should be investigated.
The Ku70/Ku80 heterodimer, a component of the DNA repair system, is an essential component of the human telomeric protein complex (Song et al., 2000;d'Adda di Fagagna et al., 2004;Jaco et al., 2004;Myung et al., 2004). A decrease in Ku70 or Ku80 gene dosage in Drosophila causes a sharp increase in the frequency of HeT-A and TART attachments to broken chromosome ends and in terminal DNA elongation by gene conversion (Melnikova et al., 2005). Ku70 mutant flies have elongated telomeres that contain an increased number of HeT-A and TART elements . At the same time, a reduced concentration of Ku70 or Ku80 does not affect HeT-A transcript abundance in flies, and Ku70 null mutation does not cause telomeric fusions Melnikova et al., 2005). A role of the Ku complex in the accessibility of Drosophila chromosome termini for transpositions has been suggested (Melnikova et al., 2005).
In the HOAP mutants, as well as in the double ATM/ATR mutants, HeT-A expression is increased (Bi et al., 2005). However, it is currently unknown whether mutations of other than HP1 cap components lead to excessive telomere elongation.
The dominant factor Telomere elongation (Tel) was genetically identified in the natural Drosophila Gaiano stock that has unusually long telomeres (Siriaco et al., 2002). E(tc) locus, which affects terminal gene conversion, was also mapped to the same chromosome region . These factors might be different alleles of the same, but as yet unidentified, gene involved in Drosophila telomere length control.

Mechanisms of regulation of telomeric retrotransposon expression in somatic tissues
Telomeric retrotransposon transcripts serve as templates for the synthesis of proteins necessary for transposition as well as for reverse transcription primed by the 3' end of the telomeric DNA. As a result of end underreplication, Drosophila telomeres shorten by 75 bp per generation (Biessmann and Mason, 1988;Levis, 1989;Biessmann et al., 1990b), whereas transposition of full-size retrotransposon results in chromosome elongation by 6-12 kb. Evidently, maintenance of normal length of telomeres requires strict control of transposition frequency and transcriptional activity of telomeric retroelements. In spite that the telomeric array exhibits euchromatic features telomeric retrotransposon expression is repressed in somatic tissues. Several negative regulators of telomeric element expression have been shown, such as HP1, ATM and ATR kinases, and PROD protein (Savitsky et al., 2002;Bi et al., 2005;Torok et al., 2007). It was proposed that a special chromatin structure forms along telomeric retrotransposon array providing retrotransposon silencing (Frydrychova et al., 2008). The mechanism of the telomeric element silencing in the somatic tissues is unknown. The role of RNA interference system in this process is still not clear. Twenty-one nucleotide long endogenous siRNAs (endo-siRNAs) have been identified in the somatic cells of D. melanogaster (Czech et al. 2008;Ghildiyal et al. 2008;Kawamura et al. 2008). This class of short RNAs is produced in a Dicer-2-dependent manner and can direct AGO2 to cleave target RNAs. A subset of endo-siRNAs that are homologous to retrotransposons including HeT-A was identified. However, HeT-A expression was not significantly affected by the RNAi pathway disruption in cell culture or imago tissues (Ghildiyal et al. 2008). It should be noted that the HeT-A promoter is active in replicating diploid larval tissues (George and Pardue, 2003). Evidently, endo-siRNAs control the steady-state abundance of HeT-A RNA only in those somatic tissues where the HeT-A promoter is active. Detailed histological analysis is needed to learn more about the contribution of the endo-siRNA pathway to telomeric retrotransposon silencing in the fly soma.
Importantly, we found that the expression of telomeric retrotransposons and the frequency of their transpositions onto chromosome ends are specifically regulated by an RNAi-based mechanism in the germline where heritable transpositions occur (Savitsky et al., 2006). These data are significant for the understanding of the control of Drosophila telomere length.

Role of the piRNA pathway in Drosophila telomeric retrotransposon expression in the germline
The piRNA pathway is directed by a distinct class of 24-30-nucleotide-long RNAs called PIWI-interacting RNAs (piRNAs), which are produced by a Dicer-independent mechanism and associated with Argonaute proteins from the PIWI subfamily Aravin et al., 2008) (Fig.4). The piRNA pathway protects the genome in germ cells from transposable element activity. Piwi, Aubergine (Aub) and Argonaute 3 (AGO3) bind piRNAs and serve as core components of the piRNA machinery in Drosophila ovaries (Saito et al. 2006;Brennecke et al. 2007;Gunawardane et al. 2007;Li et al. 2009;Malone et al. 2009). These proteins are engaged in an amplification loop to mediate the generation of sense and antisense piRNAs from the transposon transcripts . Other proteins such as RNA helicases Spindle-E (Spn-E) and Armitage (Armi), nucleases Zucchini (Zuc) and Squash (Squ), the germline-specific homologue of HP1 Rhino and the product of the vasa locus are involved in transposon silencing in the germline and are required for piRNA production/stabilization (Cook et al. 2004;Vagin et al. 2004;Vagin et al. 2006;Pane et al. 2007;Klattenhoff and Theurkauf 2008;Klattenhoff et al. 2009;Malone et al. 2009). Transposon derepression and transpositions are observed in piRNA pathway mutants, pointing to the primary role of this system in the silencing of parasitic elements (Aravin et al. 2001;Sarot et al. 2004;Vagin et al. 2004;Kalmykova et al. 2005). In spite of their vital genomic function, Drosophila telomeric retroelements as well as other parasitic transposons were shown to be targets of the piRNA-mediated silencing pathway (Vagin et al., 2004;Savitsky et al., 2006;Shpiz et al., 2007;Shpiz et al., 2009). Mutations in the spn-E, aub, piwi, squ and zuc genes and vasa locus result in accumulation of telomeric element transcripts in ovaries (Vagin et al. 2004;Savitsky et al. 2006;Pane et al. 2007;Shpiz et al. 2007;Shpiz et al. 2009), but the same mutations do not affect telomeric element expression in Drosophila testes (A.K., unpublished data). In situ RNA hybridization analysis has revealed different patterns of accumulation of HeT-A, TART and TAHRE transcripts in the ovaries of piRNA mutants (Savitsky et al., 2006;Shpiz et al., 2007). TART transcripts accumulate in supporting nurse cells predominantly at the late stages of oogenesis, whereas HeT-A and TAHRE transcripts are detected in a growing oocyte from the earlier stages of oogenesis. This finding suggests that TAHRE rather than TART is a source of reverse transcriptase for the transpositions of non-autonomous HeT-A elements. Sense and antisense piRNAs specific to telomeric retrotransposons have been revealed in libraries of short RNAs and by Northern analysis (Saito et al. 2006;Savitsky et al. 2006;Brennecke et al. 2007;Shpiz et al. 2007;Malone et al. 2009;Shpiz et al. 2009). Their levels are dramatically lower in the ovaries of piRNA mutants, which correlate with increased expression.
Antisense transcripts of transposable elements are important intermediates in the piRNA pathway because they serve as templates for piRNA generation Gunawardane et al., 2007). Interestingly, both HeT-A and TART produce long non-coding processed antisense transcripts from their internal promoters (Maxwell et al., 2006;Shpiz et al., 2009). The HeT-A antisense transcription start site was mapped to the 3' UTR of this element 150 bp upstream of the sense transcription start site. HeT-A and TART antisense transcripts are targets of the piRNA pathway and accumulate in the germ cell nuclei of the piRNA pathway mutants. Thus, steady-state expression of HeT-A and TART retrotransposons could be a result of an intricate piRNA-mediated interplay of their sense and antisense transcripts (Shpiz et al., 2009). Thus, the piRNA system suppresses excessive retrotransposon activity and maintains transcripts at low levels because this system requires the presence of sense and antisense transcripts to act as triggers of this mechanism.

Role of the piRNA pathway in Drosophila telomere length control and telomeric chromatin assembly
We have shown that increased expression of HeT-A/TART in the piRNA pathway mutants results in an increased rate of their transpositions onto chromosome ends, i.e., telomere elongation (Savitsky et al., 2006). To screen for new transpositions on chromosome ends in the piRNA pathway mutants, we used truncated X chromosomes (designated y TD ) with a break in the yellow locus that controls body and bristle pigmentation. The break is located in the upstream regulatory region and results in the y 2 -like fly phenotype with yellow aristae (a bristle-like part of the antenna). The addition of HeT-A or TART retroelements can be monitored by a yellow-to-black change in aristae pigmentation (Savitsky et al., 2002;Savitsky et al., 2006) (Fig.5). Fig. 5. Genetic assay to measure the frequency of telomeric element attachments to the broken chromosome end. Schematic representation of the telomeric retrotransposon attachments to the broken yellow gene located at the end of the terminally deleted X chromosome. As a result of terminal attachment, yellow is activated, causing yellow-to-black change in aristae pigmentation.
Both spn-E and aub mutations have a strong dominant effect on the frequency of attachments, resulting in a 20-to 100-fold increase in the appearance of flies with blackcolored aristae. In flies carrying a single copy of the spn-E or aub mutant allele, the majority of the new transpositions were TART, whereas in the ovaries of sterile homozygous spn-E females, HeT-A transpositions were more frequent. This difference might be explained by peculiarities of the dosage effect of the piRNA mutations on HeT-A/TART expression. Clusters of flies with identical TART attachments were isolated in the progeny of spn-E and aub mutants, indicating that piRNA-mediated control of terminal transpositions occurs in premeiotic cells. It is noteworthy that despite the greatly increased frequency of HeT-A and TART attachments to the broken ends observed in this assay, spn-E or aub mutant lines do not have detectably longer telomeres on their native chromosomes (Savitsky et al., 2006;Khurana et al., 2010). This indicates that the truncated chromosome end is more sensitive to telomeric element attachments than the native telomere. Despite the fact that the protein cap can be formed on the broken chromosome ends, they lack both telomeric retrotransposons and subtelomeric repeats, which results in considerably different chromatin structure. The attachment of retrotransposons to native telomeres is likely impeded compared to the truncated chromosome. A recent study detected an increase in HeT-A copy number in rhi, armi and ago3 piRNA pathway mutant stocks (Khurana et al., 2010). Among them, only the armi mutation affects HOAP and HP1 capping protein recruitment in early embryogenesis, whereas all of them cause telomeric element derepression in the germline. It seems that transcript accumulation is a main reason for telomere elongation in piRNA pathway mutants. However, piRNAs may affect not only transcript abundance but also chromatin structure.
Distinct short RNA-mediated silencing mechanisms have been described. Short RNAs have been shown to target the associated protein complex to degrade complementary mRNAs that mediate post-transcriptional silencing (Elbashir et al., 2001). Short-RNA-mediated heterochromatin assembly was described in fission yeast, plants and ciliates (Hamilton et al., 2002;Volpe et al., 2002;Liu et al., 2004). In this case, heterochromatinization diminishes the transcriptional capacity of the target locus, resulting in transcriptional silencing. In mice, transposon-specific piRNAs drive transposon promoter methylation in the male germline Kuramochi-Miyagawa et al., 2008). In the Drosophila model, the mechanism of the piRNA-mediated gene silencing has remained obscure. The three Drosophila PIWI proteins PIWI, Aub and Ago3 cleave complementary RNA in vitro, suggesting their involvement in the post-transcriptional degradation of mRNA (Saito et al., 2006;Brennecke et al., 2007;Gunawardane et al., 2007). piRNAs were shown to mediate post-transcriptional retrotransposon mRNA degradation into cytoplasmic bodies in the Drosophila germline (Lim et al., 2009). There is also evidence for the influence of piRNAs on the chromatin state (Josse et al., 2007;Klenov et al., 2007). In our recent study, we addressed the mechanism of piRNA-mediated silencing of telomeric retrotransposons (Shpiz et al., 2011). This problem is of great interest because, in the case of transcriptional silencing, it might be related to the formation of the telomeric chromatin. Using different approaches, we have shown that transcriptional activity of the telomeric retroelements substantially increased in the piRNA pathway mutants. Nuclear run-on assay (Jackson et al., 1998;Core et al., 2008) on ovarian tissues has been used to estimate the density of transcriptionally active RNA-polymerase complexes at telomeric loci. An increase in the nascent transcripts emerging from telomeric loci as well as from some other retrotransposons has been shown. This observation was confirmed by the observation of enrichment of retrotransposon sequences in piRNA pathway mutants with two histone H3 modifications known to be linked to the RNA polymerase II activity (dimethylation of lysines 4 and 79). These data provided strong evidence for piRNA-mediated transcriptional silencing of the telomeric retrotransposon loci in the Drosophila germline. Most likely, transposon defense in the germline is a combination of the piRNA-mediated post-transcriptional and transcriptional silencing. This suggests that piRNAs are important participants in telomeric chromatin assembly.
Telomeric retrotransposon arrays as mentioned above display the features of open chromatin, however, no actively elongating RNA polymerase isoforms have been detected in this region Biessmann et al., 2005). We suggest that piRNAs mediate sequencespecific binding of the inhibition protein complex locally at the HeT-A promoter in the germ cells rather than heterochromatinization along telomeric arrays. Components of the transcription initiation complex of the telomeric retrotransposons may be considered as a putative link between piRNAs and inhibition of the transcription (Fig.6). However, little is known about the transcription factors that regulate telomeric element expression. The PROD protein, which is involved in heterochromatin formation, represses HeT-A expression (Torok et al., 2007). JIL-1 and Z4, which are associated with decompacted chromatin regions, are recruited to the telomeric retrotransposon array and colocalize with euchromatin-specific histone H3 trimethylated at lysine 4 in somatic tissues . The piRNA/PIWI complex is proposed to mediate binding of the transcriptional inhibitors at the HeT-A promoter (yellow filled circle). The piRNA pathway components aub and armi are involved in telomere cap protein recruitment. TAS-specific piRNAs mediate chromatin assembly in this region.
HeT-A/TART derepression was recently reported to be a result of loss of histone H3 lysine 9 trimethylation by the methyltransferase dSETDB1 in the germline (Rangan et al., 2011). dSETDB1 was proposed to be required for piRNA production. This result underlined the importance of chromatin structure for piRNA-mediated expression of telomeric retrotransposons. Thus, piRNAs affect not only transcript abundance but changes in the chromatin state of telomeric retroelements as well. We believe that both factors cause the excessive telomere length detected in piRNA pathway mutants (Table 1). These data underscore the importance of the piRNA pathway in Drosophila telomere homeostasis.
However, there are still more open questions than clear answers in this field. Telomere fusions in early embryos are observed only in aub and armi piRNA mutants, which affect cap formation (Khurana et al., 2010), whereas an increase in the HeT-A copy number was detected in most investigated piRNA pathway mutants. Moreover, TAS regions also produce piRNAs Yin and Lin, 2007;Todeschini et al., 2010). The study of transgenes inserted in TASs led to the discovery of a phenomenon called a telomeric trans-silencing effect (TSE) . A transgene inserted in a TAS can in trans repress the expression of a homologous transgene in the germline. In recent studies, TSE was found to depend on the piRNA silencing pathway and heterochromatin components (Josse et al., 2007;Todeschini et al., 2010), suggesting that TAS may be considered as a platform for piRNA-mediated chromatin assembly.
As mentioned above, the telomeric region of D. melanogaster is subdivided into three distinct subdomains based on DNA composition and chromatin structure: the cap, the retrotransposon array and the TAS region. Interestingly, in spite of the distinct features of these domains, the chromatin structure of each is under the control of the piRNA silencing pathway. This may suggest that there are several levels of Drosophila telomere length regulation and that the piRNA pathway is one of the important participants in this complex process (Fig. 6).

Role of RNAi in the telomere function in different organisms
In the fission yeast Schizosaccharomyces pombe, RNAi is required for heterochromatin assembly Volpe et al., 2002). Short RNAs guide histone methyltransferase to the target locus to methylate lysine 9 of histone H3 (H3K9) with subsequent binding of the heterochromatic protein Swi6. Specific repeated elements (dg and dh) that are present at all major heterochromatic regions, including pericentromeric regions, subtelomeres and the mating-type locus, have been shown to act as targets of RNAi-mediated silencing Volpe et al., 2002). Disruption of the RNAi system leads to defects in telomere clustering during mitosis and meiosis, although the silencing of transgenes inserted within subtelomeric loci, telomere length and telomeric Swi6 localization are not affected (Hall et al., 2003;Sugiyama et al., 2005). It was shown that fission yeast employs two independent mechanisms to maintain gene silencing at telomeres (Fig. 7). A chromatin-remodeling complex is recruited to yeast telomeres via interaction with the telomeric repeat binding proteins Ccq1 and Taz1 or the RNAi machinery that acts through dg and dh repeats embedded within subtelomeric regions (Kanoh et al., 2005;Hansen et al., 2006;Sugiyama et al., 2007). Thus, the maintenance of telomeric chromatin depends on redundant RNAidependent and RNAi-independent mechanisms. The removal of genes encoding RNAi components has little impact on the maintenance of the silencing of reporters or on telomere length; however, it affects telomere dynamics during mitosis and meiosis. The RNAi machinery was proposed to be essential for higher-order chromatin organization at telomeres (Sugiyama et al., 2005). Fig. 7. Telomeric chromatin assembly in S. pombe. A chromatin-remodeling complex is recruited to yeast telomeres via redundant RNAi-dependent and RNAi-independent mechanisms. Subtelomeric repeats generate siRNAs that guide histone methyltransferase to methylate lysine 9 of histone H3 (H3K9me) with subsequent binding of the heterochromatic protein Swi6. Alternatively, telomere binding proteins Ccq1 and Taz1 induce the methylation of histone H3-K9 to recruit Swi6, which results in establishment of telomere heterochromatin.
In Tetrahymena, mutation of the Dcl1p gene that encodes the germline-specific ortholog of the RNase III enzyme Dicer causes serious impairments in meiotic chromosome behavior (Mochizuki and Gorovsky, 2005). The chromosome "bouquet" (i.e., telomere clustering in one region of the nuclear periphery) formation preceding homologous pairing did not occur in meiotic prophase, which resulted in impairments of chromosome segregation and higher lethality of progeny. This observation points to an important role of the RNAi component Dcl1 in chromosome segregation and telomere dynamics during conjugation in Tetrahymena (Mochizuki and Gorovsky, 2005). Taken together, these data suggest that a requirement for RNAi machinery in telomere function is probably conserved among eukaryotes. Telomere and/or subtelomeric arrays in different organisms irrespective of the mode of telomere elongation contain repetitive sequences, which are a potential source of short RNAs and the putative targets for RNAi.

Conclusion
Despite the different mode of chromosome end elongation (telomerase or transpositions), eukaryotic telomeres are functionally equal. They restore receding chromosome ends, protect them from the cell repair system, and are involved in the processes of chromosome positioning in the nucleus and chromosome condensation in meiosis and mitosis. The role of piRNAs in the expression of Drosophila telomeric retrotransposons, control of telomere length and assembly of the telomeric chromatin clearly indicates the importance of the piRNA pathway in Drosophila telomere homeostasis. Mutations in the RNAi machinery have been shown to disrupt telomere function in both S. pombe and Tetrahymena. RNAi likely mediates telomeric chromatin assembly, which plays a crucial role in telomere dynamics. It is tempting to speculate that RNA silencing plays a universal role in telomere function in various organisms irrespective of the mode of telomere elongation.

Acknowledgements
This work was supported by grants to A.K from the Russian Academy of Sciences program for Molecular and Cell Biology and the Russian Foundation for Basic Researches (09-04-00305).