The histone deacetylase Rpd3 regulates the heterochromatin structure of Drosophila telomeres

Telomeres are specialized structures at the end of eukaryotic chromosomes that are required to preserve genome integrity, chromosome stability and nuclear architecture. Telomere maintenance and function are established epigenetically in several eukaryotes. However, the exact chromatin enzymatic modifications regulating telomere homeostasis are poorly understood. In Drosophila melanogaster, telomere length and stability are maintained through the retrotransposition of specialized telomeric sequences and by the specific loading of protecting capping proteins, respectively. Here, we show that the loss of the essential and evolutionarily conserved histone deacetylase Rpd3, the homolog of mammalian HDAC1, causes aberrant telomeric fusions on polytene chromosome ends. Remarkably, these telomere fusion defects are associated with a marked decrease of histone H4 acetylation, as well as an accumulation of heterochromatic epigenetic marks at telomeres, including histone H3K9 trimethylation and the heterochromatic protein HP2. Our work suggests that Drosophila telomere structure is epigenetically regulated by the histone deacetylase Rpd3.


Introduction
Eukaryotic linear chromosomes have evolved special nucleoprotein structures at their ends, the telomeres, which counterbalance the incomplete replication of terminal DNA and protect chromosome termini from fusion events (De Lange, 2005;Ferreira et al., 2004). Telomere structure is conserved in most eukaryotes and, with few exceptions, consists of short repetitive double-stranded G-rich sequences that end in a single-stranded overhang. Telomeric DNA is elongated by telomerase, a specialized reverse transcriptase that adds short DNA repeats by reverse-transcribing the template region of its RNA component (Nugent and Lundblad, 1998). Telomere repeats bind sequence-specific factors, which in turn recruit additional telomeric proteins, forming multiprotein complexes that are crucial for chromosome-end homeostasis. A well-documented example of such terminal complexes is the six-protein complex shelterin that coats human chromosome ends allowing cells to distinguish telomeres from sites of DNA damage (Palm and de Lange, 2008). Telomeric and subtelomeric DNA from different organisms contain histone modifications that are typical of heterochromatin (Blasco, 2007). Interestingly, recent studies have highlighted some of the roles played by chromatin remodelers and covalent modifiers in the establishment of epigenetic marks essential for telomeres structure and function (Altaf et al., 2007).
The model organism Drosophila melanogaster lacks telomerase activity and fly chromosome ends do not consist of telomerasegenerated simple repeats. Drosophila telomeres are elongated by transposition of three different, but related, non-long terminal repeat retrotransposons called HeT-A, TART and TAHRE (Mason et al., 2008). Like their yeast and vertebrate counterparts, fly telomeres are also protected by multiprotein complexes Ciapponi and Cenci, 2008). Interestingly, the elongation of Drosophila telomeres is uncoupled from their capping function, as chromosomes ends devoid of retrotransposons can assemble regular telomeres and can be transmitted for many generations. There is also evidence that terminally deleted Drosophila chromosomes can be properly capped regardless the sequence of terminal DNA Mason et al., 2008). Thus, Drosophila telomere protection relies on sequence-independent epigenetically determined structures.
The remarkable functional similarity between mammalian and fly telomere-protecting mechanisms makes Drosophila an excellent model organism to study telomere biology . In order to identify factors responsible for the deposition of epigenetics marks required for the establishment and maintenance of telomere structure and function, we conducted a small biased pilot genetic screen scoring for telomere morphology defects on polytene chromosomes. We found that loss of the essential and evolutionary conserved histone deacetylase Rpd3, the homolog of mammalian HDAC1, causes aberrant telomeric fusions on polytene chromosome ends. Surprisingly, these Rpd3-knockdown-induced telomere fusion defects are associated with a marked decrease of histone H4 acetylation, as well as an accumulation of heterochromatic epigenetic marks at telomeres, including histone H3K9 trimethylation and the overloading of the heterochromatic protein HP2. Our work suggests that telomere chromatin structure is regulated by the histone deacetylase Rpd3, providing one of the first insights into the epigenetic determination of telomere organization in higher eukaryotes.

Loss of the histone deacetylase Rpd3 causes polytene chromosome telomeric fusions
Polytene chromosomes represent a special structural organization of Drosophila salivary glands, consisting of polyploid interphase nuclei, which originate by repeated rounds of DNA replication without cell division. Drosophila polytene chromosomes have proven to be an invaluable cytogenetic tool to examine chromosome structure and, at the same time, to rapidly assess the genome-wide distribution of chromatin-binding proteins. Indeed, several studies have revealed that the genome-wide distribution of countless structural and regulatory proteins on polytene chromosomes reflects cellular and interphase chromosome physiology occurring in other developmental stages or tissues (Johansen et al., 1999;Stephens et al., 2004). Therefore, we conducted a biased genetic screen on a small group of genes encoding known chromatin remodelers and members of the histone deacetylase family, by searching for telomere morphology defects on polytene chromosomes. As canonical mutations in these genes are embryonic or early-larval lethal, thereby preventing their role in polytene chromatin structure being addressed, we decided to use salivary-gland-specific RNA interference (RNAi), with a classic GAL4::UAS driving system (Steitz et al., 1998), to knockdown gene expression in the corresponding VDRC (Vienna Drosophila RNAi Center) lines. To drive UAS-controlled induction of the RNAi in salivary glands we used eyeless (ey) promoter-mediated GAL4 expression (eyGAL4); in addition to the eye imaginal disks, the eyeless promoter has been shown to also efficiently drive the expression of GAL4 in salivary glands (Corona et al., 2007).
RNAi-mediated knockdown for some of the genes analyzed led to modest alterations in polytene chromosome structure (data not shown). However, loss of Sin3A or Rpd3 histone deacetylase function on polytene chromosomes (supplementary material Fig. S1A-F), using multiple VDRC lines with no predicted off-target sites, resulted in the formation of apparently elongated chromosome ends that also underwent frequent telomeric fusions . Interestingly, eye-specific RNAi-mediated knockdown of Sin3A or Rpd3 caused developmental and differentiation defects in adult eyes (supplementary material Fig. S2A-C) confirming a role for the Sin3A-Rpd3 histone deacetylase complex in cell division and proliferation (Ahringer, 2000). However, silencing of Sin3A or Rpd3 in larval neuroblasts, using a daGAL4 driver (Corona et al., 2007), did not cause telomeric fusions on mitotic chromosomes (supplementary material Fig. S2D-F). Indeed, previous experiments have suggested that the establishment of telomere function in polytene and mitotic chromosomes are similar, although they probably rely on different mechanisms of maintenance and protection Rashkova et al., 2002), thereby explaining why chromosome-end fusions specifically occur only after RNAi-mediated silencing of Sin3A or Rpd3 on interphase polytene chromosomes.
In order to quantify the chromosome-end fusions in Sin3A and Rpd3 RNAi lines, we developed a scoring system to measure the penetrance of telomeric attachments (supplementary material Fig.  S3). On the basis of this scoring system, the observed telomeric fusions upon Sin3A and Rpd3 RNAi were not due to independent mutations present in the driving (eyGAL4) or UAS-RNAi (UAS-Sin3A-RNAi; UAS-Rpd3-RNAi) lines (Fig. 1F), nor are they indirectly caused by the activation of the RNAi machinery in salivary glands because the overexpression of Dicer2 (eyGAL4; UAS-Dicer2) does not cause an increase in telomere fusions when compared with wild-type Df(1)yw chromosomes (Fig. 1F) (Dietzl et al., 2007). Our scoring system also revealed that knockdown of the histone deacetylase catalytic subunit encoded by Rpd3 yielded a higher frequency of telomere fusions when compared with knockdown of Sin3A (Fig. 1F). Therefore, to address further the role of the Sin3A-Rpd3 deacetylase complex in telomere 2042 Journal of Cell Science 124 (12) RNAi larvae is based on the number of fused interphase polytene chromosome telomeres scored. Class 1 corresponds to zero fusions, 2 to two telomeres fused, 3 to three telomeres fused, 4 to four telomeres fused and 5 to five telomeres fused (supplementary material Fig. S3). For each genotype tested the statistical significance (P value) of the telomere fusion defect was calculated and compared with the telomere fusion distribution observed in the Df(1)yw wild type line. The presence of class 2 and 3 fusions in the Df(1)yw wild-type line is mostly due to the variability in chromosome spreading preparations, which causes some telomeres to overlap or fall very close. homeostasis, we decided to focus our telomere fusion characterization only on the Rpd3-encoded catalytic subunit.
Abnormal telomere elongation and chromosome-end fusions have been observed in Drosophila polytene chromosomes as consequence of mutations in the HP1-encoding Su(var)205 gene and the Telomere elongation (Tel) gene (Perrini et al., 2004;Siriaco et al., 2002). Telomeres of polytene chromosomes are characterized by the presence of three distinct chromatin domains: the telomereassociated sequences (TASs), the terminal tract of the HTT retrotransposons (HeT-A, TART and TAHRE) and the cap, the mostterminal region that is associated with capping proteins that bind independently of the DNA sequence (Andreyeva et al., 2005;Cenci et al., 2005). In both Su(var)205 and Tel mutants, elongated telomeres result from an increase in HeT-A and TART elements in the HTT domain. It has been hypothesized that the presence of many tandemly arranged copies of telomeric retrotransposons on long polytene telomeres can increase the likelihood of recombination between telomeric sequences leading to telomeretelomere attachments Perrini et al., 2004;Siriaco et al., 2002).
To ascertain whether the apparent telomere elongation and fusions upon Rpd3 RNAi were caused by an increase in the amount of telomeric retrotransposable elements at chromosome ends, we conducted classic fluorescent in situ hybridization (FISH) analysis using HeT-A, TART and TAHRE sequences as probes. This analysis revealed that the levels of HeT-A (supplementary material Fig. S4A-C), TART (supplementary material Fig. S4D-F) and TAHRE (supplementary material Fig.  S4G-I) sequences were very similar in wild-type telomeres and Rpd3 RNAi fused telomeres. These results were also confirmed by dot-blot analysis conducted on genomic DNA from Rpd3 RNAi and Tel mutant larval salivary glands. As expected, the dot-blot analysis revealed a significant increase of telomeric genomic DNA HeT-A sequences in Tel mutants but not upon Rpd3 RNAi, when compared with wild type ( Fig. 2A). Furthermore, because Rpd3 plays a key role in transcription repression (De Rubertis et al., 1996;Pile and Wassarman, 2000;Rundlett et al., 1996), we checked whether Rpd3 knockdown would affect transcription of telomeric retrotransposons. RT-PCR on RNA samples extracted from salivary glands of wild-type and Rpd3 RNAi larvae did not show any significant difference in transcription of HeT-A sequences (Fig. 2B). Collectively, our data indicate that, unlike for Tel and Su(var)205, upon Rpd3 RNAi telomeres are not longer than wild-type, suggesting that the 'apparently elongated telomeres' Rpd3 RNAi phenotype is due to changes in chromatin condensation.

Loss of Rpd3 affects histone acetylation at telomeres
To assess whether the apparent elongation and fusion phenotypes of the telomere tip upon Rpd3 RNAi were linked to chromatin condensation changes underlying telomeric histone acetylation defects, we performed immunofluorescence analysis on wildtype and Rpd3 RNAi chromosomes using antibodies directed against acetylated histones H3 and H4 (AcH3 and AcH4; specifically recognizing K9 and K14 histone H3 acetylations and K5, K8, K12 and K16 histone H4 acetylations), which represent the main histone targets of Rpd3 (Kadosh and Struhl, 1998). Interestingly, loss of Rpd3 caused a substantial increase in the global levels of both acetylated H3 and H4 histones on polytene chromosomes ( Fig. 3A-D), without affecting chromatin loading or expression of non-acetylated histones (supplementary material Fig. S5). Furthermore, western blot analysis on chromatin extracts derived from salivary gland cells, confirmed that chromosomes are loaded with higher levels of both acetylated histones H3 and H4 upon Rpd3 knockdown when compared with that on wildtype chromosomes (Fig. 3E,F).
Remarkably, upon Rpd3 RNAi high-magnification images of telomeres immunostained for AcH3 and AcH4, revealed that, whereas the levels of histone H3 acetylation staining slightly increased, the levels and extension of histone H4 acetylation were greatly reduced (Fig. 3BЈ,DЈ, white arrowhead and double-arrowed bar). To confirm these findings we measured the levels of histone acetylation specifically associated with telomeric HeT-A. To this aim, we optimized a chromatin immunoprecipitation (ChIP) technique on chromatin isolated from salivary glands using the telomere-associated protein HP2 (Fig. 4A). As expected, using antibodies directed against AcH3 and AcH4 we found a substantial increase in HeT-A-associated histone H3 acetylation upon Rpd3 RNAi with respect to that in wild type (Fig. 3G,H, lanes 2 and 6). By contrast, the histone H4 acetylation associated with HeT-A sequences was dramatically reduced (Fig. 3G,H, lanes 3 and 7) when compared with that in wild type. These ChIP data are consistent with our immunostaining results and collectively indicate that Rpd3 RNAi has different effects on histone acetylation at telomeres.
In male polytene chromosomes histone H4 lysine 16 acetylation (AcH4K16) specifically and exclusively localizes to the X chromosome (Gelbart et al., 2009) and is excluded from the male X telomere tip (Fig. 5A, white arrowhead). In Rpd3-knockdown polytene chromosomes AcH4K16 is still excluded from X chromosome ends but this loss of localization seems much wider than that for wild-type X chromosome tips (Fig. 5B,C, doublearrowed bar). Interestingly, ectopic AcH4K16 localization starts to appear on some autosomal chromatin loci (Fig. 5B,C, green arrowheads), probably as a consequence of loss of global Rpd3 histone deacetylase activity. ChIP analysis also confirmed a marked reduction of AcH4K16 on HeT-A telomeric sequences upon Rpd3 RNAi (Fig. 5D,E). Taken together, our analyses on histone acetylation suggest that Rpd3 knockdown causes a significant increase of global H3 and H4 histone acetylation. However, although histone H3 acetylation levels are higher also within telomeric sequences upon Rpd3 RNAi, acetylation of histone H4 (including that at K16) appeared substantially and specifically decreased at the level of HTT sequences.

2043
Rpd3 telomere fusions HeT-A sequences was measured by dot blot using a radiolabeled HeT-A probe hybridized to genomic DNA extracted from wild-type (wt; 1), Rpd3 RNAi (2) and Tel mutant (3) salivary glands. A radiolabeled RpL32 probe was used as a loading control. (B)HeT-A transcription was evaluated by RT-PCR on total RNA derived from wild-type (1) and Rpd3 RNAi (2) salivary glands. Primers amplifying the ribosomal protein RpL32 were used to normalize the RT-PCR data.

Rpd3 regulates heterochromatin marks on polytene telomeres
To investigate whether the marked decrease in histone H4 acetylation levels associated with telomeric HeT-A upon Rpd3 RNAi affected other telomere-associated epigenetic modifications, we analyzed the distribution of three common heterochromatin histone modifications: trimethlyated histone H4K20 (trimeH4K20), trimethlyated histone H3K9 (trimeH3K9) and trimethlyated histone H3K27 (trimeH3K27). Normally, trimeH4K20 is found in the heterochromatic chromocenter and in numerous euchromatic bands (Fig. 5H) (Ebert et al., 2006), trimeH3K27 is associated with the chromocenter, telomeric TASs and a subset of about ~100 bands (Fig. 5J) (Andreyeva et al., 2005), whereas trimeH3K9 localizes at pericentric heterochromatin, at telomeres and at a number of sites along chromosome arms (Fig. 5F). We found that Rpd3 knockdown decreased the global localization of all three modified histones on polytene chromosome arms (Fig. 5G,I,K) when compared with wild type. However, although the levels of trimeH4K20 and trimeH3K27 were reduced also at telomeres upon Rpd3 RNAi (Fig. 5IЈ,KЈ, white arrowheads), the amount of H3K9 trimethylation increased in Rpd3-knockdown telomeres (Fig. 5GЈ, white arrowhead). As it has been previously shown trimeH3K9 localizes at the telomeric HTT domain (Andreyeva et al., 2005) and HTT telomeric sequences are not amplified in telomeres upon Rpd3 RNAi (Fig. 2), the observed trimeH3K9 accumulation combined with the reduced trimeH3K27 levels could reflect a change in heterochromatin structure in telomeres upon Rpd3 RNAi.
As we were unable to find ChIP primers specific for TAS sequences (data not shown), we conducted FISH analysis in order to obtain some low-resolution information about global chromatin organization of TAS in both wild-type and telomeres and upon Rpd3 RNAi. The FISH data indicate that, whereas in wild-type strains the TASs showed a wide punctuated localization at telomeres (Fig. 6A, double arrowed bars in 6AЈ and 6AЉ), in Rpd3-knockdown TAS FISH signals on telomeres were slightly more compacted (Fig. 6B, double arrowed bar on the 3L-X chromosome fusion and white bar on the 2L unfused chromosome in BЈ). This apparent slightly compacted TAS localization pattern in Rpd3-knockdown polytene chromosomes was not due to a reduction of TAS copies, as their number did not appear different from that observed in the w 1118 control strain (data not shown). Thus, it is probably a (E,F)Western blot analysis of histone H3 and H4 acetylation on chromatin protein extracts derived from wild-type and Rpd3 RNAi salivary glands. Chromatin protein loading was normalized with an antibody recognizing the linker histone H1. (G)ChIP analysis was performed on chromatin extracted from wild-type and Rpd3 RNAi salivary glands and was immunoprecipitated using antibodies against acetylated H3 and H4. Chromatin immunoprecipitation was performed in triplicates and analyzed by PCR using specific primers for the telomeric HeT-A sequences and GAPDH as a control for non-telomeric targets. (H)Quantification of the anti-AcH3 and anti-AcH4 antibody ChIP experiments, relative to 'input' chromatin.

Fig. 4. ChIP on salivary gland polytene chromatin specifically detects
HeT-A sequences. ChIP analysis was performed on chromatin extracted from wild-type and Rpd3 RNAi salivary glands and immunoprecipitated using an antibody against HP2. Chromatin immunoprecipitation was analyzed by PCR using specific primers for the telomeric HeT-A sequences and ACT5C as a control for non-telomeric targets.
consequence of a change in chromatin compaction that affects the organization of the sequences and/or probe accessibility to its specific complementary sequence.

Loss of Rpd3 causes an accumulation of telomeric HP2
It is widely accepted that trimeH3K9 at chromosome ends is a key conserved mark for a proper localization of telomeric capping proteins (Schoeftner and Blasco, 2009). Thus, we wanted to verify whether accumulation of trimeH3K9 in telomeres upon Rpd3 RNAi influenced the localization of proteins known to have essential functions in Drosophila telomere maintenance. Immunofluorescence using antibodies directed against specific telomere capping proteins showed that Rpd3 knockdown did not affect the localization or abundance of HOAP (encoded by the caravaggio gene) (Fig. 7A-C,J) (Cenci et al., 2003;Shareef et al., 2001), HP1 (Fig. 7D-F,J) (Fanti et al., 1998;Perrini et al., 2004) or Woc (Fig. 7G-I,J) (Raffa et al., 2005) in either fused or nonfused telomeres, indicating that capping proteins localize normally upon Rpd3 RNAi. The fact the HP1 localization and function is not affected as consequence of changes in telomeric levels of trimeH3K9 strengthens previous observations that the interaction between the HP1 chromodomain and trimeH3K9 is not required for the capping function of HP1 Perrini et al., 2004).
Because specific histone modifications are crucial for the localization of heterochromatin proteins, we sought to determine whether alterations in both AcH4 and trimeH3K9 caused by loss of Rpd3 influenced the distributions of heterochromatin proteins that, like HP1, associate with chromosome ends in Drosophila. We therefore analyzed the telomere localization pattern of HP2, an HP1-interacting factor with a role in the structural organization of chromosomes and in heterochromatin-induced gene silencing (Shaffer et al., 2006). In polytene chromosomes HP2 localizes at the chromocenter, at distinct chromosomal bands and at the cap of telomeres (Andreyeva et al., 2005;Shaffer et al., 2006). However, unlike HP1, HP2 is not a capping protein and is not required to prevent chromosome fusions (Shaffer et al., 2006). Surprisingly, immunofluorescence ( Fig. 5L-MЈ; Fig. 7J) and ChIP (Fig. 4B) analyses revealed a striking accumulation of HP2 at fused and non-fused telomeres upon Rpd3 RNAi, although the localization of HP2 at the chromocenter and other chromosomal regions remained indistinguishable from wild type (data not shown).

Discussion
We have identified the Rpd3-Sin3A histone deacetylase complex as a factor establishing proper chromatin structure of Drosophila telomeres. Strikingly, loss of Rpd3 dramatically affected the organization of polytene chromosome ends resulting in telomere prone to fusions. Interestingly, the observations that upon Rpd3 RNAi telomeric fusions are not seen in mitotic cells and that 2045 Rpd3 telomere fusions

Fig. 5. Loss of Rpd3 alters site-specific histone H4K16 acetylation and causes an accumulation of trimeH3K9 and HP2 at telomeres.
Immunofluorescence analysis with an antibody recognizing H4K16 acetylation (AcH4K16) on wild-type (wt) (A) and not-fused (B) or fused (C) Rpd3 RNAi male polytene chromosomes. The inserts show a magnified image of the male X telomere tips. The white arrowhead and the double-arrowed bar indicate changes in the levels of AcH4K16 in telomere tips upon Rpd3 RNAi. The green arrowheads indicate the appearance of non-specific autosomal H4K16 acetylated chromatin domains on chromosomes upon Rpd3 RNAi. (D)ChIP analysis was performed using antibodies against acetylated H4K16 (AcH4K16) to immunoprecipitate chromatin extracted from wild-type and Rpd3 RNAi salivary glands. Chromatin immunoprecipitation was performed in triplicates and analyzed by PCR using specific primers for the telomeric HeT-A sequences and GAPDH as a control for non-telomeric targets. (E)Quantification of the anti-AcH4K16 antibody ChIP experiment, relative to input chromatin. Immunofluorescence on (F,H,J) wild-type and (G,I,K) Rpd3 RNAi polytene chromosomes using antibodies recognizing trimethylated histone H3K9 (trimeH3K9), trimethylatated H4K20 (trimeH4K20) and trimethylated H3K27 (trimeH3K27). (GЈ,IЈ,KЈ) A magnified image of the white-boxed areas in G, I and K, respectively, shows in the detail the changes in trimeH3K9, trimeH4K20 and trimeH3K27 histone modifications occurring at the telomere fusions sites. The asterisk indicates the chromocenter and the white arrowheads point to telomere fusions. Immunofluorescence staining with the anti-HP2 antibody on (L) wild-type and Rpd3 RNAi mutant polytene not-fused (M) and fused (MЈ) chromosome ends shows that the HP2 protein accumulates in telomeres upon Rpd3-knockdown.
telomere capping proteins localize normally further suggest that the activity of Rpd3 is not required for chromosome end protection.
Fused polytene telomeres have been always associated with an increase in HTT copies. Surprisingly, our FISH analysis, along with dot-blot results, clearly indicated that in Rpd3-knockdown telomeres the amount of the HeT-A, TART and TAHRE elements remained as high as in wild type. The apparent telomere elongation we observed upon Rpd3 RNAi probably reflects a change in telomere chromatin condensation in Rpd3 mutants, underlying changes in histone acetylation. As a result of this change in telomere chromatin condensation, Rpd3 mutants possess telomere tips that appear 'longer' than usual by DAPI staining.
A combination of immunostaining and ChIP data indicated that loss of Rpd3 by RNAi leads to profound global, as well as telomeric, chromatin epigenetic changes. Upon specific Rpd3 RNAi telomere epigenetic changes included a decrease in H4 acetylation, including H4K16Ac, and an increase of trimeH3K9 at the level of telomere tips. Moreover, the HP2 protein, which normally binds to the telomere cap, was highly enriched upon Rpd3 RNAi. Because we were unable to perform ChIP on TAS sequences, we could not directly assess whether the epigenetic changes induced by Rpd3 RNAi also affected TAS sequences. Nevertheless, the apparent slight compaction of TAS organization detected by FISH in polytene chromosomes upon Rpd3 RNAi suggests that loss of Rpd3 affects the structural chromatin organization of this region.
The question of how the accumulation of HP2 on telomeric chromatin with reduced H4Ac and increased trimeH3K9, is linked to the fusion of telomeres upon Rpd3 RNAi still remains open. One intriguing possibility is that an overload of HP2 drives excessive heterochromatization at chromosome ends upon Rpd3 RNAi, inducing telomeric stickiness reminiscent of chromocenter heterochromatin coalescence. Alternatively, an excess of 2046 Journal of Cell Science 124 (12) Fig. 6. Chromatin organization of TASs in Rpd3 RNAi polytene chromosome telomeres. FISH analysis of TASs in wild-type (wt) (A) and Rpd3 RNAi (B) unfused and fused telomeres. Arrowhead point to chromosome fusion. Wild-type TASs show a wide and punctuated localization at telomeres, whereas upon Rpd3 knockdown TASs present a slight but highly penetrant (100%; n50) reduction in the extension of the FISH signal along both chromosome axes, probably reflecting a change in chromatin compaction following loss of Rpd3 or probe accessibility. (AЈ,AЉ,BЈ) Magnified images of the TAS FISH signals shown in A and B, respectively. Double arrowed bars show the extension of FISH signals along the telomere-centromere axis. Note that extension of TAS FISH signals in Rpd3 RNAi telomeres is slightly reduced when compared with that in wild type. The white bar indicates an area with a substantial reduction of FISH signal along the width of an unfused Rpd3-knockdown 2L telomere. Note that only the HP2 percentage of signal intensity is significantly higher (*P<0.001,  2 test) on mutant telomeres with respect to that on wild-type chromosome ends (asterisk). heterochromatization might also impair telomere replication leading to stalled replication forks, which are known to cause fusigenic telomeres (Ferreira et al., 2004).
In conclusion, our work provides an important insight into the epigenetic determination of telomere chromatin organization in higher eukaryotes. As polytene chromosomes are known to be useful cytogenetic tools to address the question of structural and functional organization of telomeres in Drosophila, our results might represent a foothold for further studies to understand the interplay existing between histone acetylation and deacetylation events and telomere homeostasis.

Drosophila stocks and genetic crosses
Flies were raised on cornmeal-yeast-agar medium containing Tegosept. Drosophila RNAi strains and w 1118 ;P w + ] were obtained from the Vienna Drosophila RNAi Center (http://stockcenter.vdrc.at/control/main).

Protein extracts and western blotting analysis
Salivary glands total protein extracts from the Df[yw], eyGAL4/UAS-Sin3A-RNAi and eyGAL4;UAS-Rpd3-RNAi mutant third-instar larvae were obtained dissecting the glands in 0.7% NaCl, following vortexing in of RIPA buffer (2 l for each pair of glands; 1% NP40, 0.5% sodium deoxicolate, 0.1% SDS in PBS). The homogenate was centrifuged at 20,000 g for 30 min and the supernatant was flash frozen in liquid nitrogen and stored at -80°C (Burgio et al., 2008). Salivary glands chromatin extracts were prepared as described previously (Corona et al., 2007). Proteins were fractionated by SDS-PAGE and analyzed by protein blotting using rabbit antibodies against Sin3A and Rpd3 (Pile and Wassarman, 2000), acetylated histone H3 and H4 (Upstate) and against histone H1 (Corona et al., 2007). The chemiluminescent HRPconjugated secondary antibodies were developed using the Super Signal West Femto substrate (Pierce) and acquired with the ChemiDoc XRS imager (Bio-Rad).

RT-PCR and dot-blotting of HeT-A sequences
For RT-PCR, total RNA was isolated from Df[yw] and eyGAL4;UAS-Rpd3-RNAi larval salivary glands by using TRIzol (Invitrogen). The total RNA was reverse transcribed by using the oligo d(T) GeneAmp RNA PCR kit (Applied Biosystems). The first-strand cDNA was used as a template for PCR reactions with primers specific for HeT-A and RpL32 (HeT-A forward, 5Ј-GGCAGCATATTAGCGC -GTACA-3Ј and HeT-A reverse, 5Ј-TTTGCCGCCAGCTTTTGT-3Ј; RpL32 forward, 5Ј-CAAGAAGTTCCTGGTGCACAAC-3Ј and RpL32 reverse, 5Ј-AAACGCGGTTCTGCATGAG-3Ј) (Perrini et al., 2004). To exclude amplification of genomic DNA contamination present in our RNA samples, we always included a mock RT-PCR reaction lacking the reverse transcriptase as a negative control. Agarose gels were stained with ethidium bromide and images were acquired with the ChemiDoc XRS Imager (Bio-Rad). For dot blots, genomic DNA from Df[yw], eyGAL4;UAS-Rpd3-RNAi and Tel mutant larval salivary glands was obtained by using classical phenol-chloroform-isoamyl alcohol extraction and then cross-linked on Hybond-N+ membrane (Amersham). A PCR-amplified HeT-A probe was obtained using the pSC-23Zn plasmid as a template and the primers specific for HeT-A used for the RT-PCR described above (Perrini et al., 2004). The RpL32 probe was obtained from cDNA retro-transcribed from the total RNA of salivary glands and was PCRamplified with the primers used for RT-PCR described above (Perrini et al., 2004). Membranes were hybridized with HeT-A and RpL32 probes labeled with the Random Primed DNA Labeling kit (Roche). Radioactive signals were acquired with the Personal Molecular Imager (BioRad).