The insulator BEAF32 controls the spatial-temporal expression profile of the telomeric retrotransposon TART in the Drosophila germline Free

Insulators are architectural elements of the genome implicated in the organization of higher-order chromatin structures and transcriptional regulation (Chetverina et al., 2017a). Suppressor of Hairy wing [Su(Hw)] (Geyer and Corces, 1992), boundary element-associated factor 32 (BEAF32) (Zhao et al., 1995), GAGA factor (GAF; Trl) (Chetverina et al., 2021; Schweinsberg et al., 2004) and CCCTC Binding Factor (CTCF) (Kellum and Schedl, 1992) are among the DNA-binding proteins associated with insulators in Drosophila. Several co-factors are required for the assembly of insulator protein complexes. Centrosomal protein 190 (CP190) (Pai et al., 2004) and Mod(mdg4) (Ghosh et al., 2001) are the common partners of BEAF32, CTCF and Su(Hw) (Gerland et al., 2017), suggesting their role in the formation of hybrid long-range interactions between heterologous insulators (Liang et al., 2014; Vogelmann et al., 2014). The assembly of insulator complexes is crucial for gene regulation. Transcriptional activity of a genomic locus depends on the architecture of insulator proteins associated with promoters and enhancers. BEAF32 and CP190 occupy promoters of active genes (Cubenas-Potts et al., 2017). Transcription of many BEAF32-associated genes decreases in the absence of BEAF32, suggesting that BEAF32 is important for maintaining active transcription (Jiang et al., 2009). However, Su(Hw) mediates transcriptional repression, which is essential for normal oogenesis (Geyer and Corces, 1992; Melnikova et al., 2019; Soshnev et al., 2013).

The contribution of insulator complexes to telomere integrity is not well understood, but research suggests they have an essential role as telomere architectural elements. In human cells, the boundary element CTCF contributes to telomere replication and transcription (Beishline et al., 2017; Deng et al., 2009). Depletion of CTCF-binding sites at the subtelomeric region containing a promoter of telomere repeat-encoding RNA (TERRA) leads to chromosome abnormalities and anaphase bridges (Beishline et al., 2017). The expression of the telomerase reverse transcriptase (TERT) catalytic subunit in human cancer cell lines was reported to be regulated by architectural protein CTCF (Eldholm et al., 2014; Li et al., 2017; Renaud et al., 2005). In Drosophila, immunofluorescence on polytene chromosomes demonstrated that the architectural protein Chriz (chromo domain protein interacting with Z4; also known as Chromator) (Eggert et al., 2004; Rath et al., 2004), a partner of BEAF32 (Vogelmann et al., 2014), is localized at telomeric regions (Takacs et al., 2012). The insulator protein Mod(mdg4) regulates enhancer activity of subtelomeric repeats and indirectly regulates the expression of the telomeric retrotransposon HeT-A in Drosophila somatic cells (Takeuchi et al., 2022). In this study, we focus on the role of insulators in telomere regulation in the Drosophila germline.

Telomere elongation and telomerase activity are normally restricted to certain cell types, including stem and germ cells (Bekaert et al., 2004), and the shortening of telomeres during development and differentiation serves as a tumor-suppression mechanism. The telomeres of Drosophila melanogaster are maintained by transpositions of specialized telomeric retrotransposons, HeT-A, TART and TAHRE, which are related to the non-LTR retrotransposons of the jockey clade (Abad et al., 2004b; Casacuberta, 2017). Open reading frames (ORFs) of telomeric transposons are closely related, which suggests their origin from a common ancestor (Abad et al., 2004b). TAHRE is considered a relict element represented by a few copies. Telomeric retroelements are randomly interspersed in head-to-tail arrays, although TART copies have a tendency to be clustered (Abad et al., 2004a; George et al., 2006; McGurk et al., 2021). HeT-A, the most abundant telomeric element, is a non-autonomous retroelement with reverse transcriptase (RT) activity likely provided by TART or TAHRE telomeric retrotransposons. Telomere biology in the Drosophila female germline is characterized by the complex regulation of expression of telomeric retrotransposons (Kordyukova et al., 2019; Morgunova et al., 2015). Telomeric retrotransposons provide both a template and enzymatic activity for telomere elongation. The proper balance between expression of enzymatic and structural telomere components is believed to be crucial for the maintenance of telomere length. Heterochromatin protein 1 (HP1) and HipHop, a component of the Drosophila telomere protective complex, repress the transcription of all telomeric retroelements in the germline (Cui et al., 2021; Teo et al., 2018). Piwi-interacting RNAs (piRNAs) mediate epigenetic regulation of telomeric chromatin and the expression levels of telomeric transcripts (Radion et al., 2018; Rozhkov et al., 2013; Savitsky et al., 2006). All telomeric retroelements and subtelomeric regions give rise to the abundant telomere-specific piRNAs, which play an essential role in the expression of telomeric repeats, telomere length control, assembly of telomeric chromatin, and nuclear telomere positioning in the germline (Radion et al., 2018; Savitsky et al., 2006). However, despite their close proximity to each other, HeT-A and TART elements differ in transcriptional regulation and piRNA production (Danilevskaya et al., 1999; Radion et al., 2018; Savitsky et al., 2006). HeT-A and related TAHRE elements are very sensitive to piRNA pathway disruption, whereas TART is much less affected by piRNAs. piRNA loss leads to a dramatic decrease of trimethylated histone H3 lysine 9 (H3K9me3), HP1 and Rhino, a germline-specific ortholog of HP1 and a marker of piRNA source loci (Klattenhoff et al., 2009; Mohn et al., 2014), at telomeres (Radion et al., 2018; Rozhkov et al., 2013). Interestingly, the chromatin states of HeT-A, TAHRE, and transgenes inserted into arrays of these elements change upon loss of piRNA. In contrast, piRNA depletion has little effect on the expression and chromatin state of endogenous TART as well as transgenes inserted into TART arrays, implying a different mechanism for their regulation (Radion et al., 2018). Recently, we found that mutations of the insulator Su(Hw) cause enhanced production of TART mRNAs and piRNAs (Radion et al., 2019). These results encouraged us to seek an alternative mechanism for TART regulation in the germline that could explain its resistance to the piRNA processing machinery. Here, we show that recruitment of the insulator protein BEAF32 to the promoter of TART is required for its transcriptional repression during oogenesis. The knockout of BEAF32 leads to TART overexpression and chromatin changes in the germline, but does not alter the transcription or chromatin state of HeT-A. We further show that BEAF32 controls TART copy number in telomeres. Our data suggest that the BEAF32 insulator complex, being positioned at the TART regulatory region, blocks enhancer–promoter interactions. TART expression is elevated at the earliest stages of oogenesis, in the germ cysts, when BEAF32 levels are normally low. We speculate that insulator-mediated regulation of TART expression ensures well-timed RT activity for telomere elongation in the germline.

The two Drosophila telomeric retroelements HeT-A and TART are regulated differently in the germline. Unlike HeT-A, endogenous TART, as well as transgenes inserted in TART arrays, are not substantially affected by piRNA-mediated chromatin modifications (Radion et al., 2018). We proposed that TART could be isolated from the HeT-A arrays by the chromatin insulator complex. Insulators exhibit enhancer-blocking activity when inserted between an enhancer and promoter. To test the TART regulatory region for enhancer-blocking activity, we used a previously described transgenic enhancer-blocking assay based on the yellow reporter gene (Schwartz et al., 2012). TART elements contain non-terminal direct repeats in the 5′ and 3′ untranslated regions (UTRs) harboring sense and antisense transcription start sites (TSSs) (Maxwell et al., 2006). A 1 kb fragment of the TART promoter region was positioned between the upstream wing and body enhancers and the yellow promoter; the bristle enhancer is located downstream of the yellow promoter (Fig. 1A). Flies carrying a construct with a randomly chosen 1 kb genomic fragment showed no insulator-binding activity and have black pigmentation of the body and wings, whereas flies carrying the gypsy insulator element have a yellow body and wings (Schwartz et al., 2012). In both cases, the flies had black bristles. Insertion of the TART DNA fragment between the yellow wing/body enhancer and promoter prevented the activation of the reporter yellow gene by upstream enhancers yielding transgenic flies with a light body and wings (Fig. 1A). At the same time, expression of the yellow gene in bristles was not suppressed, indicating that the TART DNA fragment does not act as a silencer relative to the bristle enhancer. In vivo excision of the TART fragment by activating the expression of Flp recombinase restored yellow gene expression in the fly body and wings (Fig. 1B), suggesting insulator/silencer activity of the TART fragment in these tissues. The TART-associated protein complex could be a silencer affecting yellow body and wings enhancers. To test this possibility, the TART fragment was positioned upstream of body and wing yellow enhancers. Flies carrying this construct had a black body, wings and bristles, indicating that the TART fragment does not affect the downstream-located yellow regulatory elements and, therefore, lacks silencer activity (Fig. 1C). These genetic assays suggest that the TART regulatory region may act as an insulator.

To identify potential insulator proteins associated with the TART regulatory region, we first analyzed published ChIP-seq (chromatin immunoprecipitation followed by sequencing) data (Wood et al., 2011) and found evidence that BEAF32 binds to the TART promoter region in Kc167 Drosophila cultured cells (Fig. 1D). In contrast, CP190, Su(Hw), BEAF32 and DREF do not bind to HeT-A and TAHRE, with their ChIP-seq signal being comparable to input. Drosophila DNA replication-related element binding factor (DREF) is a transcription factor that binds a similar sequence motif to that of BEAF32 (Matsukage et al., 2008). DREF and BEAF32 have multiple overlapping binding sites and are thought to compete for DNA binding (Gurudatta et al., 2013; Hart et al., 1999). Indeed, two ATCGAT consensus sequences predicted for DREF and BEAF32 overlapping sites (Gurudatta et al., 2013) are identified upstream of the TART TSS (Fig. 1E).

To address the role of BEAF32 in regulation of chromatin structure of the TART promoter in the germline, we performed ChIP-qPCR experiments on ovaries using antibodies against insulator proteins and their partners. Primers specific to the TART promoter are shown in Fig. 1E. As expected, we revealed that BEAF32 and DREF occupy TART but not the HeT-A promoter (Fig. 2). The insulator scs′ from the heat-shock gene locus, known for BEAF32 binding, was used as a positive control (Hart et al., 1997). Among known BEAF32 partner proteins, Chriz and CP190 occupy endogenous TART promoters (Fig. 2). The HeT-A promoter demonstrates low levels of Chriz and CP190 binding. The other Drosophila insulator proteins examined, Mod(mdg4), Su(Hw) and GAF, did not bind to either TART or HeT-A promoters but, as expected, occupied positive-control regions (Fig. S1). We conclude that the TART promoter is distinguished from HeT-A by its specific interactions with insulator proteins, particularly BEAF32 and its partners.

TART is a multicopy retrotransposon family represented by 17 copies per genome on average (McGurk et al., 2021). To place ChIP analysis in a genomic context, we performed ChIP-qPCR on P{EPgy2} transgenes inserted in endogenous TART elements (Biessmann et al., 2005). Using transgenes inserted in telomeric regions, we have previously found that TART retrotransposons are more resistant to the Rhino binding and piRNA production than HeT-A/TAHRE (Radion et al., 2018). Here, we evaluated the BEAF32 occupancy of P{EPgy2} transgenes inserted in the TART regulatory region between BEAF32-binding sites in three independent transgenic strains. EY00802 and EY00453 strains contain P{EPgy2} insertions within chromosome 3R telomeric TART and EY09966 contains a transgene insertion in TART on the 4th chromosome (Fig. 2A). ChIP-qPCR analysis was performed using transgene-specific primers located at ∼300 bp from BEAF32-binding sites in endogenous TART. We found that BEAF32, Chriz, DREF and CP190 bind to all transgenes (Fig. 2C).

To determine the chromatin properties of the TART promoter in euchromatic context, we used transgenic strains containing TART-lacZ constructs. In this construct, a 1 kb promoter region was fused to a lacZ reporter in the CaSpeR-AUG-β-gal vector followed by P-element-mediated transgenesis. Two independent transgenic strains, TD5 and TD11, containing insertions in euchromatin of 2R and 3R chromosome arms, respectively, were established. Localization of transgenes was determined by inverse-PCR (see Materials and Methods) and confirmed by DNA fluorescence in situ hybridization (FISH) on polytene chromosomes of salivary glands (Fig. S2). We demonstrated by ChIP-qPCR using transgene-specific primers located ∼200 bp from the BEAF32-binding site in the TART promoter that, in both transgenic strains, euchromatic transgenic TART fragments associate with BEAF32, Chriz and DREF (Fig. 2C). The lower enrichment of these proteins at telomeric and euchromatic transgenes relative to endogenous TART can be explained by the fact that transgene-specific PCR primers are located at some distance (200-300 bp) from insulator-binding sites. Our data indicate that the recruitment of BEAF32 and its partners to the TART promoter occurs independently of genomic context and is determined by the TART promoter sequence (Fig. 2D).

To address the role of BEAF32 in the expression and chromatin state of telomeric retrotransposons, we used the BEAF32 null allele referred to here as BEAF32_KO (Roy et al., 2007). BEAF32_KO flies are viable but produce underdeveloped ovaries lacking late stages of oogenesis. In wild-type ovaries, BEAF32 is localized to the nuclei of the follicular and nurse cells but no BEAF32 was detected in the BEAF32_KO mutants (Fig. S3). RT-qPCR analysis revealed significant TART derepression in ovaries of BEAF32_KO flies (Fig. 3A). This result is comparable to the derepression of TART observed with mutations in the essential piRNA pathway gene spnE (Malone et al., 2009) (Fig. 3A). RNAi knockdown of BEAF32 in the germline also resulted in TART derepression, confirming the role of BEAF32 in TART expression (Fig. 3A; Fig. S4). In contrast, RNA levels of the telomeric element HeT-A were unaffected by BEAF32 depletions, which underscores that these elements are regulated by different mechanisms.

We then investigated whether BEAF32 mutation could affect genomic TART copy number. To this end, we compared genomic DNA HeT-A and TART relative copy numbers in BEAF32 mutants and laboratory and natural strains (Mackay et al., 2012), including the known long-telomere strains Gaiano (Siriaco et al., 2002) and RAL882 (Wei et al., 2017) using qPCR. Analysis of population genomic data did not previously reveal a difference in the proportion of HeT-A, TAHRE and TART telomeric elements in natural D. melanogaster strains with various telomere lengths (McGurk et al., 2021). Accordingly, both HeT-A and TART showed an increase in copy numbers in Gaiano and RAL-882 strains. In contrast, the BEAF32_KO strain showed an increase in only TART copy number (Fig. 3B). Thus, the BEAF32 null mutation is strongly associated with elevated TART copy number.

To study the chromatin changes accompanying TART derepression in the ovaries of BEAF32 null mutants, ChIP-qPCR was performed. As expected, BEAF32 was not detected at the TART promoter and other loci in BEAF32 null mutants. At the same time, Chriz and DREF occupied the TART promoter even in the absence of BEAF32, suggesting that their binding is independent of BEAF32 (Fig. 3C). Because RNAi of DREF resulted in severe ovary degeneration, we failed to test whether Chriz binding to the TART promoter is mediated by DREF. Immunostaining of wild-type ovaries showed overlapping distribution of BEAF32 and Chriz and their colocalization with the telomere-specific HOAP protein (encoded by the cav gene) in the nurse cell nuclei (Fig. 3D). In the absence of BEAF32, telomeric HOAP colocalizes to Chriz, which is supported by ChIP-qPCR (Fig. 3C,D). Thus, in the absence of BEAF32 TART expression is likely regulated by DREF and Chriz (Fig. 3F).

Next, we analyzed changes in histone modifications associated with active (H3K4me2) and compact (H3K9me3) chromatin at TART in BEAF32 mutants. A significant increase in the active chromatin histone modification H3K4me2 was found at the TART promoter in BEAF32_KO ovaries along with a decrease in the compacted chromatin histone modification H3K9me3 (Fig. 3E). Analysis of previously published RNA polymerase II (PolII) ovarian ChIP-seq data (Andersen et al., 2017) showed that PolII levels peak at the TART promoter (Fig. 4A). However, PolII binding was not significantly changed at the TART promoter according to ChIP-pPCR analysis of BEAF32 null mutants (Fig. 3E).

We found that BEAF32 localizes upstream of the TART TSS and, therefore, can act as an enhancer blocker at the TART promoter, which is typical for insulators. To identify the transcriptional enhancers in the TART regulatory region, we first searched the Drosophila STARR-seq (self-transcribing active regulatory region sequencing) database (Zabidi et al., 2015). STARR-seq libraries contain large collections of functional enhancers identified by quantitative measuring of enhancer activity of genomic DNA fragments in a reporter assay. Enhancer activity was detected 200-1000 bp upstream of the BEAF32-binding site (Fig. 4A,B; Fig. S5). To assess TART enhancer activity directly, we performed luciferase reporter assays. To this end, we cloned 791, 396 and 198 bp fragments of the TART regulatory region upstream of BEAF32-binding site into the hsp70min-Fluc reporter construct and measured the reporter gene activity in transfection experiments using the Schneider 2 (S2) Drosophila cell line. S2 cells were co-transfected with control or TART constructs expressing the lucF reporter as well as the pAc-Rluc plasmid for normalization. Luciferase activity assays revealed similar activation of LucF expression for all studied TART constructs (Fig. 4C), suggesting an enhancer activity of these regions. A similar result was observed in RT-qPCR analysis of the lucF mRNA levels (Fig. S5). Because adding another 600 bp does not increase enhancer activity of the construct, the minimal 200 bp fragment comprises an enhancer itself.

Thus, an enhancer is located upstream of the BEAF32-binding sites in the TART regulatory region, whereas the TSS is localized downstream. Collectively, these results strongly support a role for BEAF32 as an enhancer blocker at TART promoter.

Ovary immunostaining showed BEAF32 in the nuclei of the follicular and nurse cells, but its level was reproducibly lower at early stages of oogenesis in the germarium region 2a containing germ cysts (Fig. 5A,B). To determine the ovarian cell type with reduced BEAF32 expression, we used a morphological marker, Bam protein, which is expressed in the germline until the 8-cell cyst stage (Lie-Jensen and Haglund, 2016). We found that BEAF32 reduction occurs in the 16-cell germ cysts of the germarium region 2a (Fig. S6). Next, we studied the localization of TART transcripts in wild-type ovaries and BEAF32 mutants. TART RNA FISH combined with HOAP immunostaining revealed the statistically significant accumulation of TART sense RNA at telomeres in the nuclei of germline cysts within a germarium region 2a (Fig. 5C). TART RNA foci were observed in all cell nuclei of the germarium 2a stage, but rarely detected at later stages of oogenesis in wild-type ovaries. In BEAF32 null mutants, TART transcripts strongly accumulated in the nurse cell nuclei in egg chambers at 3-5 stages (Fig. 5D). These data are in accordance with increased TART expression in BEAF32 mutants as shown by RT-qPCR.

Next, we compared the effects of BEAF32 knockout and piRNA pathway disruption on TART expression. TART-specific piRNAs have been generated in the Drosophila germline (Ryazansky et al., 2017). Accordingly, piRNA pathway gene mutations cause upregulation of TART expression in ovaries (Fig. 3A; Fig. S7) (Lopez-Panades et al., 2015; Savitsky et al., 2006). Therefore, endogenous TART expression is enhanced in the ovaries of both piRNA pathway mutants and BEAF32 mutants (Fig. 3A), suggesting a possible link between these pathways. We looked at whether BEAF32 knockout affects expression of transposable elements (TEs) and Piwi localization. In the list of TEs, we included blood retrotransposon because its expression is affected by all known piRNA pathway factors (Czech et al., 2013). RT-qPCR analysis revealed largely unchanged levels of HeT-A and non-telomeric retrotransposon RNAs in BEAF32 null mutants relative to heterozygous control (Fig. 6A), indicating that the piRNA pathway is intact in BEAF32_KO mutants. It was previously reported that the germarium region containing germ cysts is characterized by reduced Piwi levels, which causes a burst of transposon expression and piRNA production (Dufourt et al., 2014; Theron et al., 2018). This region, termed ‘Piwiless pocket’, was clearly visible in both in the control and BEAF32_KO germarium (Fig. 6B). It is noteworthy that the area with reduced levels of BEAF32 in the germarium was partially overlapping with the ‘Piwiless pocket’ in wild-type ovaries (Fig. 6B). Thus, repression of TART is normally relaxed at the early stages of oogenesis when both BEAF32 and Piwi proteins are depleted allowing TART expression in the dividing cystoblasts (Fig. 5C). However, some TART transcripts escaped piRNA-mediated silencing and accumulated in the nurse cell nuclei of BEAF32 null mutants at later stages of oogenesis when piRNA-mediated silencing is re-established (Fig. 5C).

We wanted to investigate further the possible cooperation between the piRNA pathway and BEAF32 in the control of TART-lacZ expression. BEAF32 binds to TART-lacZ transgenes (Fig. 2), but, surprisingly, β-galactosidase expression was not distinctly increased in BEAF32 null mutants (Fig. 6C). lacZ expression was also not significantly changed in spnE-depleted transgenic ovaries with strongly derepressed endogenous HeT-A and TART (Fig. S7). It is noteworthy that β-galactosidase activity was observed in the small germarium region, indicating that TART-lacZ is active at the same developmental stage as endogenous TART. Given the dual regulation of TART expression by both piRNA pathway and BEAF32, we thought it was possible that TART-lacZ is actually activated in the absence of BEAF32, but TART-specific piRNAs guide degradation of TART-lacZ RNAs thereby erasing the effect of BEAF32 mutation. Double mutants are commonly used to find a functional link between different factors; however, a compound knockout of BEAF32 and Piwi could not be used owing to severe ovary degeneration caused by both BEAF32 and Piwi depletions. To overcome this experimental problem, we applied epigenetic depletion of the maternal piRNA pool in BEAF32 mutants. It has been reported that maternal age affects TE silencing and expression of piRNAs and piRNA pathway genes in Drosophila (Dramard et al., 2007; Erwin and Blumenstiel, 2019; Theron et al., 2018). Previously, it was shown that the maintenance of ‘long generations’ (offspring is always from aged females during several generations) contributes to the enhancement of anti-transposon protection, which is associated with transposon-specific piRNA accumulation in the Drosophila germline (Brennecke et al., 2008; Grentzinger et al., 2012). We applied the opposite approach – the maintenance of ‘short generations’ (offspring is always from 3-day-old females) – which could lead to depletion of the maternal pool of small RNAs. We established a ‘short-generation’ TART-lacZ line bearing the BEAF32 mutation. As expected, β-galactosidase staining was significantly enhanced in the F7 of ‘short-generation’ transgenic flies bearing the BEAF32 null allele (Fig. 6C). Activation of TART-lacZ in the ovaries of these flies was confirmed by β-galactosidase activity assay (Fig. 6D). This result suggests that TART-lacZ expression is regulated by both BEAF32 and the piRNA pathways. Probably, the maternal pool of TART piRNAs is depleted in the ovaries after several successive ‘short generations’, resulting in reduced degradation of TART-lacZ transcripts and their accumulation in BEAF32 mutants. However, the mechanisms of age-dependent variations of piRNA production are not completely clear and require further elucidation.

In contrast to telomeric TART, TART-lacZ expression was not affected by BEAF32 knockout in the ‘long generation’ stock. A possible explanation could be that TART-lacZ construct contains a single TART repeat, whereas TART full-length copies are bordered by long repeats containing regulatory and insulator-binding regions. We suggest that interactions of architectural proteins occupying TART direct repeats ensure transcriptional integrity of functional TART and provide stability of its transcripts, whereas TART-lacZ transcripts lacking such a protection are more unstable and undergo piRNA-mediated degradation.

Our data suggest a dual control of TART expression by both piRNA pathway and insulator proteins that ensures short-term activation of TART expression at the earliest stages of oogenesis (Fig. 6E). Such intricate regulation may be necessary for strict developmental control of TART-mediated RT activity.

Telomere homeostasis is accomplished by numerous players the activity of which is orchestrated to ensure telomere protection and elongation. Our data shed light on the nature of different mechanisms of regulation of HeT-A and TART, two telomere elements sharing a common genomic niche. We describe a distinct mechanism of regulation of the telomeric retrotransposon TART involved in Drosophila telomere maintenance. This element encodes an RT and may be considered as a functional analog of the telomerase enzyme component. Our study is focused on the transcriptional and chromatin regulation of TART in the germline, where telomere length control is crucial for normal development.

A protein complex consisting of the insulator protein BEAF32, the transcription factor DREF, the chromodomain protein Chriz and the architectural protein CP190 interacts with TART, whereas only low levels of CP190 and Chriz but not BEAF32 and DREF are found at the HeT-A retrotransposon. These findings raise questions about the specific telomeric functions of insulators. A complex composition and various context-dependent activities are typical for insulator protein complexes. Depletion of BEAF32 results in the accumulation of TART transcripts, suggesting a repressive activity of BEAF32 towards TART regulation. An enhancer was identified upstream of BEAF32-binding sites, suggesting that BEAF32 separates the promoter and enhancer that drive TART transcription. A model of competition between BEAF32 and DREF implies that BEAF32 insulates neighboring genes from nearby regulatory elements, whereas, under appropriate conditions, DREF would displace BEAF32 leading to transcriptional activation (Hart et al., 1999). DREF occupies the TART promoter independently of BEAF32 binding, suggesting its direct role in the prompt activation of TART transcription when BEAF32 levels decrease.

The chromatin state of TART elements differs from that of HeT-A and is characterized by an open state and an enrichment by RNA PolII. Telomeric regions, especially HeT-A repeats, belong to the germline piRNA clusters encoding transcripts that serve as piRNA precursors during oogenesis (Radion et al., 2018). However, endogenous TART elements and transgene insertions in TART are resistant to the Rhino binding and piRNA production (Radion et al., 2018). The BEAF32/Chriz complex is a likely candidate to protect TART from the HeT-A-associated piRNA clusters. Moreover, endogenous TART transcripts accumulate in BEAF32_KO ovaries despite the activity of the piRNA pathway. BEAF32 knockout does not affect binding of the chromatin protein Chriz to the TART promoter. Chriz has been shown to have a chromatin-opening role upon tethering to compact chromatin regions (Pokholkova et al., 2018). It is likely that Chriz is involved in the formation of a local region of open chromatin at TART elements in the heterochromatic telomeric regions, whereas BEAF32, acting as an enhancer blocker, mediates the fine-tuned regulation of TART expression during oogenesis. Expansion of TART in the genome of BEAF32 null mutants suggests a TART-specific role of BEAF32 in telomere elongation. Our study demonstrates that architectural proteins play an essential telomeric role in the germline in regulating transcription, chromatin state, developmental profile and copy number of the telomeric retrotransposon TART.

Expression of TART and the TART-lacZ transgene occurs in the germline cysts, which are characterized by a physiological reduction of both BEAF32 and Piwi levels. TART RT protein was also revealed in the germarium in wild-type ovaries by another group (Lopez-Panades et al., 2015), corroborating our data. Restoration of high levels of BEAF32 and Piwi results in TART silencing during later stages of oogenesis. Intriguingly, repression of HeT-A and other transposons regulated by piRNAs, but not by BEAF32, is also loosened in the germarium, where the Piwi level is sharply decreased (Dufourt et al., 2014; Theron et al., 2018). It is proposed that a reduction of Piwi expression allows piRNA amplification to ensure transposon silencing at later stages of oogenesis (Theron et al., 2018). At the same developmental stage, in the germ cysts, we previously observed nuclear sphere-like ribonucleoprotein particles consisting of HeT-A RNA and HeT-A Gag protein, which are thought to be the intermediates of telomere elongation (Kordyukova et al., 2018). Intriguingly, expression of structural and putative enzymatic telomere components, encoded by HeT-A and TART, respectively, is coordinated by distinct mechanisms to provide telomere elongation in the germline. Surprising similarities can be observed with the regulation of structural (TERC) and enzymatic (TERT) telomerase components: TERC is expressed at high levels in most tissues, whereas TERT is repressed in most normal somatic cells and serves as the limiting factor for telomerase activity (Jie et al., 2019; Yashima et al., 1998).

Our results suggest a dual nature of TART regulation by both an insulator complex and the piRNA pathway, which could be determined by the specific organization of the telomeric domain. Functional TART copies have a strong tendency to be clustered in telomeres (George et al., 2006; McGurk et al., 2021). Full-length TART has a pair of non-terminal repeats in the 5′ and 3′ UTRs, which contain regulatory sequences including insulator-binding sites. Most likely, only clustered TART copies or the full-length TART are able to form an isolated domain bordered by the BEAF32/Chriz complex, which ensures TART transcriptional specificity and integrity. It is believed that boundary pairing can form a loop that facilitates and/or protects regulatory interactions (Chetverina et al., 2017a). It was reported that the CTCF-mediated chromatin loop between the enhancer and promoter of TERT plays a role in maintaining TERT expression in human cancer cells (Eldholm et al., 2014). In Drosophila, the chromatin remodeling proteins BEAF32, Chriz, DREF and CP190 are also enriched at sites of enhancers and promoters of Drosophila active genes and are presumably required for their transcriptional integrity (Cubenas-Potts et al., 2017; Jiang et al., 2009). Our data suggest that BEAF32/Chriz-mediated isolation of telomeric domains containing TART provides limited and fine-regulated TART transcription in the germline when piRNA silencing is also relaxed. Given that Gag protein, encoded by HeT-A, provides telomeric localization of TART Gag, it has been suggested that HeT-A ensures the telomere targeting function, whereas TART supplies the reverse transcriptase for the attachments of both elements to the chromosome ends (Rashkova et al., 2003, 2002). We suggest that the simultaneous short-term derepression of both telomeric elements in the ovarian germ cysts is required for telomere elongation control.

We used a null BEAF^AB-KO allele that eliminates production of both BEAF-32A and BEAF-32B isoforms (Roy et al., 2007). Germline knockdown flies were the F1 progeny from the cross of two strains bearing constructs with short hairpin (sh) RNA (spnE_sh, #103913; BEAF32_sh1, #46519; Vienna Drosophila Resource Center) and strain P{UAS-Dcr-2.D}1, w¹¹¹⁸, P{GAL4-nos.NGT}40 (#25751, Bloomington Drosophila Stock Center), providing GAL4 expression under the control of the germline-specific promoter of the nanos gene. spindle-E alleles spn-E¹ and spn-E^hls3987 in a trans-heterozygous state were used. RAL strains were obtained from the Drosophila Genetic Reference Panel collection (Mackay et al., 2012).

The TART-lacZ construct was made in the CaSpeR-AUG-β-gal vector. The TART-B regulatory region corresponding to nucleotides 9008-10079 of the GenBank sequence DMU14101 was amplified on genomic DNA of the iso-1 strain used for D. melanogaster genome assembly and inserted into the EcoRI and BamHI sites of CaSpeR-AUG-β-gal yielding the CaSpeR-TART-AUG-β-gal construct. P-element-mediated germline transformation was performed on Df(1)w^67c23(2), y embryos. Two transgenic strains were established, and genomic coordinates of insertions were determined by inverse PCR. The TD5 strain contains the insertion at chromosome 2R:6501842 (plus genomic strand) and TD11 contains the insertion at chromosome 3R:24218898 (minus genomic strand). X-gal staining of β-galactosidase and quantitative ONPG (o-nitrophenyl-β-D-galactopyranoside) test for β-galactosidase activity in the ovaries were performed according to previously published protocols (Shpiz et al., 2011).

Enhancer-blocking assays were performed using the yellow-attB-mobile vector (Schwartz et al., 2012). The TART-B fragment (the nucleotides 9008-10079 of the GenBank sequence DMU14101) was cloned between wing/body enhancers and promoter of yellow into the XbaI and NheI sites. The yellow-attB-TART construct was integrated in the 51D landing site (#24483, Bloomington Drosophila Stock Center) by φC31 attB recombination. Strains bearing y+; attB mobile-SuHw680 and y+; attB mobile-random constructs were used as positive and negative controls, respectively, in the enhancer-blocking assays (Schwartz et al., 2012). FLP-mediated excision of the TART fragment was performed by crossing transgenic flies with a strain expressing FLP recombinase (w¹¹¹⁸; S2CyO, hsFLP, ISA/Sco; +). To test the silencer activity of TART, the regulatory region of TART-B was inserted upstream of the yellow wing enhancer of yellow-attB-mobile vector. To this end, the yellow wing enhancer was excised from the yellow-attB-mobile construct by digesting with XhoI restriction enzyme, blunted using T4 DNA polymerase, and ligated into blunted CaSpeR-TART-AUG-β-gal/BamH1. The EcoRI-KpnI fragment containing the TART-B fragment and yellow wing enhancer was blunted using T4 DNA polymerase and ligated into blunted yellow-attB-mobile/XhoI yielding the TART-yellow-attB construct. To estimate the levels of yellow, we visually determined the degree of pigmentation in the abdominal cuticle, bristles and wing blades of 3- to 5-day-old females developing at 25°C, with reference to standard color scales. The pigmentation scores were independently determined by two investigators.

Chromatin immunoprecipitation was performed on Drosophila ovaries according to a published procedure (Akulenko et al., 2018). Chromatin was immunoprecipitated with the following antibodies: anti-trimethyl-histone H3 Lys9 (07-523, Millipore); anti-dimethyl-histone H3 Lys4 (07-030, Millipore); anti-RNA polymerase II (05-623, Millipore); anti-GAF (Erokhin et al., 2015); anti-CP190, Su(Hw) and Mod(mdg4)-67.2 (Golovnin et al., 2012); anti-BEAF32 (Melnikova et al., 2021); anti-DREF (kindly provided by Pavel Georgiev, Institute of Gene Biology Russian Academy of Sciences, Moscow, Russia); and anti-Chriz/Chromator (Gortchakov et al., 2005). The primers used in this study are listed in Table S1. Quantitative PCR was conducted with a LightCycler 96 (Roche). The obtained values were normalized to input. The standard error of the mean (s.e.m.) of duplicate PCR measurements for three to five biological replicates was calculated.

Three fragments of 791, 396 and 198 bp of the TART promoter region were PCR amplified on a template of the TART-lacZ construct using primers indicated in Table S1. The resulting amplicons were digested with restriction enzymes, XhoI and KpnI, and ligated into a XhoI/KpnI-digested vector hsp70min-Fluc based on the pGL3 basic vector (Promega) (Chetverina et al., 2017b). Act1-4hsp70min-Fluc construct containing Act5С enhancer region (chromosome X:5899361-5899689, r6.44) upstream of hsp70 promoter was used as a positive control. D. melanogaster embryonic S2 cells were transfected using the FuGENE^® HD Transfection Reagent (Promega) according to the manufacturer's instructions. The cells were harvested 48 h post-transfection for the luciferase activity assay (Dual-Luciferase Reporter Assay System, Promega) or total RNA extraction.

RNA was isolated from the ovaries of 3-day-old females. cDNA was synthesized using random hexamers and M-MuLV reverse transcriptase (Biolabmix). The cDNA samples were analyzed by real-time quantitative PCR on a LightCycler 96 (Roche). The values were averaged and normalized to the expression levels of the ribosomal protein gene rp49 (RpL32) and/or the housekeeping gene Pgd. The s.e.m. for three to five biological replicates was calculated. RNA fluorescence in situ hybridization was performed according to a previously described procedure (Kordyukova et al., 2018) using a digoxigenin (DIG)-labeled, strand-specific riboprobe and anti-DIG antibodies conjugated to FITC (11207741910, Roche). An antisense TART riboprobe containing a fragment of the TART ORF2 corresponding to nucleotides 2377-2888 in GenBank sequence DMU02279 was used.

FISH with polytene chromosomes of salivary glands was performed as previously described (Lavrov et al., 2004). A PCR fragment amplified using white-specific primers (5′-CATGATCAAGACATCTAAAGGC-3′ and 5′-GCACCGAGCCCGAGTTCAAG-3′) was labeled with a DIG DNA labeling kit (11175033910, Roche). Immunostaining was carried out according to a previously described procedure (Kordyukova et al., 2018). The following primary antibodies were used for immunostaining: guinea pig anti-HOAP (1:500; Morgunova et al., 2021), rabbit anti-Chriz (1:200; kindly provided by A. Gorchakov; Gortchakov et al., 2005), mouse anti-BEAF32 (1:250; Developmental Studies Hybridoma Bank), mouse anti-Bam (1:10; Developmental Studies Hybridoma Bank), rabbit anti-Piwi (1:50; ab5207, Abcam). Fluorophore-conjugated secondary antibodies with minimal cross-reactivity to IgG from non-target species (706-606-148, 711-546-152, 715-165-151, 715-545-151, Jackson ImmunoResearch) were used (1:500). To stain the DNA, the ovaries were incubated in PBS containing 0.5 µg/ml DAPI. Two biological replicates were obtained for each experiment. Images were captured using a Zeiss LSM 900 confocal microscope and z-stacks were recorded with a 0.5-µm step. The intensities of BEAF32 or TART RNA fluorescence were calculated using ImageJ tools (http://imagej.nih.gov/ij/). The mean fluorescence intensity of BEAF32 in different germarium regions was measured in ten germariums from two independent experiments. TART RNA fluorescence intensity was measured in 30-40 cell nuclei from each germarium region or from nurse cells (egg chambers at stages 3-5) for each sample; two independent experiments were analyzed. GraphPad Prism 8.0 was used for statistical analysis and graph generation.

For the analysis of ChIP-seq data, we used published datasets of insulator proteins (Wood et al., 2011) and RNA PolII (Andersen et al., 2017). After adapter removal using Cutadapt (Martin, 2011), we mapped the reads passing quality control (allowing zero mismatches) to canonical sequences of telomeric transposable elements using bowtie2 (Langmead and Salzberg, 2012). For STARR-seq analysis, we used published data (Zabidi et al., 2015) and processed them as described. For the analysis, we considered only the libraries made using the housekeeping core promoter (RpS12 core promoter) on S2 cells. STARR-seq reads were mapped to the canonical TART-B sequence or to the D. melanogaster genome (April 2006, BDGP assembly R5/dm3) allowing zero mismatches. For the visualization of both ChIP-seq and STARR-seq data, we plotted normalized read counts using the Gviz package (Hahne and Ivanek, 2016).

We are grateful to M. Savitsky for the yellow-attP vector, Y. Schwartz for the control fly strains for enhancer blocking assay, A. Golovnin for anti-CP190, Su(Hw) and Mod(mdg4)-67.2 antibodies, P. Georgiev for anti-DREF antibodies, A. Gorchakov for anti-Chriz antibodies, C. Hart for BEAF_KO strain, V. Alatortsev for advice on bioinformatic analysis of STARR-seq data, M. Sukhova for participation in molecular cloning, L. Malaev for assistance with statistical analysis, D. Chetverina for critical comments on the manuscript. We thank the Bloomington Stock Center and Vienna Drosophila RNAi Center for fly strains and the Developmental Studies Hybridoma Bank for antibodies. We thank for the provided equipment the User Centre of the Institute of Developmental Biology RAS supported by the IDB RAS Government basic research program 0088-2021-0007.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.201678.reviewer-comments.pdf