Dissection of open chromatin domain formation by site-specific recombination in Drosophila

Looped domains containing single genes or groups of coregulated genes are firmly established as important units of gene regulation and organization of interphase chromatin (Blanton et al., 2003, Murrell et al., 2004, Palstra et al., 2003; Sexton et al., 2009; Sexton et al., 2012; Tolhuis et al., 2002, Van Bortle and Corces, 2012). Although such domains are subject to dynamic changes by tissue-specific gene expression, there is evidence for similarity in the global domain organization of the genome in different cell types of a given species. Such a similarity in the domain organization was first noted by cytology (Beermann, 1953) and was more recently established by molecular methods (Peric-Hupkes et al., 2010; Vatolina et al., 2011; Demakov et al., 2011). However, details on domain structure and the rules regulating the formation and maintenance of looped domains are still not well understood.

Polytene chromosomes of Drosophila are a powerful tool for the study of interphase chromosomes because they allow experimenters to merge cytology with data from molecular biology. The pattern of the majority of their chromomeres, condensed chromatin bands and less condensed interbands, is remarkably constant from cell to cell, between different animals and different organs. Exceptions are loci that differ because of conditional or tissue-specific gene activity. Moreover, the molecular organization of chromomeres in terms of DNA condensation, protein binding and histone modifications appears to be conserved between diploid and polytene cells (Vatolina et al., 2011; Demakov et al., 2011) suggesting that the cytology of polytene chromosomes faithfully reflects the general domain organization of the interphase chromosome. Bands and interband domains differ in their state of condensation with interbands condensed at the level of the 10-nm nucleosome fiber and bands at a level of ≥30-nm fiber (Beermann, 1972, Rykowski et al., 1988). Bands are supposed to reflect domains in a transcriptional inactive state, interbands, by contrast, might be transcribed or have a more open chromatin structure because they contain control elements, promoters and enhancers for genes poised for transcriptional activation (Rykowski et al., 1988; Demakov et al., 2004; Demakov et al., 2011). The finding of paused RNA polymerase in many interbands favors such models.

It is still not clear how condensed (band) and less-condensed interband chromatin is configured and maintained. It simply could be the consequence of the state of transcriptional activity that differs between these domains. However, at least some interbands appear not to be transcribed (Rykowski et al., 1988; Demakov et al., 2004). Alternatively, but not mutually exclusive, there could be a local difference in the epigenetic state caused by local protein binding and histone modifications. A set of proteins mainly recruited to interband chromatin might play a role in open chromatin formation. Besides paused forms of RNA polymerase II, these include nucleosome-remodeling and histone-modifying proteins like Brahma, TRX (Demakov et al., 2011) and the H3S10 kinase Jil-1 (Wang et al., 2001), insulator proteins like BEAF32AB, CP190 and the Drosophila homolog of CTCF (dCTCF) (Van Bortle and Corces, 2012) and proteins of as yet unknown function, like Chriz (also known as Chro) (Gortchakov et al., 2005; Rath et al., 2006) and Z4 (also known as Pzg) (Eggert et al., 2004). Furthermore, interband chromatin is enriched for ‘activating’ chromatin modifications, like acetylation (H4K16ac and H3K9ac), methylation (H3K4me3) and phosphorylation (phospho-H3S10) (Vatolina et al., 2011; Demakov et al., 2011). So far these are mostly correlations and it would be desirable to directly test the contribution of these factors to the formation of local chromatin structure.

Condensed chromatin can be established at ectopic sites by transposition of transcriptionally inactive DNA (Semeshin et al., 1986; Semeshin et al., 1989), and decondensed chromatin also autonomously maintains an open state in a new chromosomal environment (Semeshin et al., 2008). These observations offer the possibility to assay for DNA sequences required for formation of decondensed chromatin. Such elements would establish open chromatin in ectopic positions and recruit the factors and histone modifications necessary for its formation and maintenance. To avoid position effects, the sequences to be tested should be inserted into the same chromosomal environment. Therefore, we designed a condensed domain cassette with an internal attB sequence for site-specific recombination at attP-tagged DNA elements (Groth et al., 2004). Applying this system, we screened for DNA elements that have the capacity to form open chromatin autonomously within a condensed region. By cytogenetic analysis of the ∼17-kb 61C7-8 interband we found that this chromatin can maintain the decondensed state at an ectopic position. By site-specific recombination we mapped the sequences of 61C7-8 essential for open chromatin formation to a 490-bp fragment that overlaps with a promoter element and contains binding sites for interband-specific proteins and boundary element factors.

Our strategy was to create random insertions of attB containing P-element vectors that would form an ectopic condensed domain detectable by microscopy. Later, recombination into such a condensed chromatin cassette should allow us to test DNA fragments for their intrinsic capacity to form open chromatin at an ectopic position (Fig. 1A). Therefore, a P-element vector (Fig. 1A, pUAST-NattBw) with an attB site flanked on its 5′ end by 9 kb of DNA from the Notch-encoding gene (Notch DNA) and on the 3′ end by two 4.2-kb fragments of DNA from the White-encoding gene (white DNA), which were separated by 4 kb of Notch DNA, was inserted into the genome. In salivary glands, the chosen Notch and white sequences are not transcribed and form condensed chromatin (bands) at their endogenous loci. Following germline transformation, 43 lines with independent insertions were obtained and mapped by inverse PCR. The lines had single insertions on all major chromosome arms. A total of 16 of 17 lines studied by in situ hybridization on polytene chromosomes had insertions into decondensed interband regions and formed detectable new bands at their site of insertion. One such ectopic band (strain 42) is shown in Fig. 2A–C′. This suggests that the transposed DNA contains all the information necessary to maintain its condensed chromatin state in salivary glands at an ectopic position. Strain 42 was chosen for further experiments. The insertion is located on chromosome 2L in the interband 21F3-22A1-2. Molecularly, the site of insertion was mapped by inverse PCR to the position 2L 1177706, 117 bp downstream of the gene CG4887 and within the 5′UTR of the gene CG4896 with the 9-kb Notch fragment oriented distally.

Next, we asked whether DNA that forms open chromatin at its endogenous site would maintain its state of condensation within a condensed chromatin environment. For this experiment we chose DNA from the 61C7-8 interband region. DNA from the distal boundary of the 61C7-8 interband was previously isolated by Demakov et al. (Demakov et al., 1993) and the extent of this interband was mapped by chromosomal walking and in situ hybridization by our group (H. S., unpublished observations). In situ hybridization of a 4-kb probe corresponding to the proximal boundary of this interband (Fig. 5A, 61C7/8prox in) is shown in Fig. 4C–C″. Because this 4-kb fragment is 8.3 kb downstream of the 4.7-kb fragment containing the distal boundary (Semeshin et al., 2008) the 61C7-8 interband includes at least 17 kb DNA (Fig. 5A), more than previously estimated (Demakov et al., 1993). According to data from experiments in S2 cells (FlyBase, modENCODE), the whole region is enriched in open active chromatin marks, like H3K4me3 and acetylated histones, and depleted for H3K9me3 and H3K27me3, which are marks for silent condensed chromatin. 61C7-8 contains five genes, four of them encoding proteins and one, bantam, with a short noncoding RNA transcript (Fig. 5A), suggesting that this interband corresponds to an open chromatin domain. The three Chriz-binding sites in 61C7-8 overlap four CP190-binding sites, each one of them in the distal and central portions, and two CP190-binding sites in the proximal part of 61C7-8 domain (Fig. 5A). Given that the fragments at the 61C7-8 boundaries carry active chromatin marks and include strong binding sites for Chriz and the boundary element CP190, we reasoned that they might support autonomous folding into decondensed chromatin. Therefore, 4 kb and 5 kb of DNA from the proximal and distal boundary of the 61C7-8 interband (Fig. 5A) were jointly cloned into a vector containing an attP site. The resulting plasmid with a 9-kb insert from 61C (Fig. 1B, pattP61C) was recombined into the attB site of the condensed chromatin domain cassette of strain 42 (Fig. 1A). From several independent insertions, one was further analyzed by in situ hybridization and PCR. We determined that the complete plasmid pattP61C had recombined in the expected orientation with the distal part of the 61C7-8 sequence facing the centromere. Cytologically, the ectopic chromatin band of strain 42 was split by the insertion and in situ hybridization with pattP61C DNA showed that the newly inserted DNA localized to the middle of the split band (Fig. 2 D–E′). This suggests that the recombined DNA maintains its decondensed chromatin state at an ectopic position. Similarly, the band-forming pUAST-NattBw DNA also maintains its condensed state but split in two parts, one distal and one proximal. We obtained a similar result when we used a vector containing 3 kb DNA from the 3C6-7 interband (Fig. 1B, pattP3C) that forms decondensed chromatin at its endogenous site (Rykowski et al., 1988). Following recombination of this vector, the ectopic band was split by decondensed chromatin that hybridized with the 3C6-7 probe (Fig. 2F–G′). In contrast, recombination of the empty vector (pattPempty, see Fig. 1B; Fig. 2H–I′) or a control vector (Fig. 1B, pattPN) containing 10 kb from the condensed Notch region into this site did not result in a split band (Fig. 2J–K′).

The vectors used for recombination into strain 42 contain a GAL4-inducible UAS-GFP reporter as a selection marker. Gland-specific induction of the GFP reporter gene in the pattP61C recombinant by GAL4 resulted in GFP expression and local puff formation at the site of the insertion that is dependent on strength of induction (Fig. 3A–B′). A similar decondensation was seen after GAL4-driven GFP reporter induction in all pattP recombinants independent of their insertions, indicating that neither of the constructs is inhibitory for gene activity (Fig. 3C–D′). However, this finding might raise the concern that the observed local decondensation results from reporter activation. However, in the absence of the GAL4 induction only the 61C7-8 and 3C6-7 constructs are decondensed (Fig. 2) suggesting that the local decondensation is a property of the inserted DNA and not of transcription of the reporter gene. This was confirmed by analysis of reporter gene activity by qRT-PCR of RNA isolated from salivary glands (Fig. 3E). The GAL4-induced line 42 containing the pattP61C insertion showed a robust GFP expression, but without GAL4 induction GFP expression was very low or absent in this strain as the signal was not significantly different from the w¹¹¹⁸ strain used as a negative control.

The local state of DNA condensation is influenced by protein binding, chromatin remodeling and histone modification. Phosphorylation of H3S10 by the histone kinase Jil-1 has been shown to contribute to local decondensation (Deng et al., 2008), and the Chriz-containing complex is involved in targeting Jil-1 to chromatin (Gan et al., 2011). Given that both Chriz and Jil-1 are found at the endogenous 61C7-8 interband (Fig. 4A–A″,B–B″), we tested whether these proteins were also bound at the ectopic site following recombination. Chriz staining is shown in Fig. 4D–F″. At the 21F3-22A1 interband of the w¹¹¹⁸ control without insertion, we observed a broad domain of Chriz staining with a fairly high intensity (Fig. 4D–D″). Following pUAST-NattBw transposition, a broader distal and a more narrow proximal signal (Fig. 4E–E″, red wedges) flank the inserted condensed chromatin domain cassette that does not bind Chriz. Following recombination of 61C7-8 interband DNA (pattP61C) into the domain cassette, an additional Chriz signal appeared, now within the split band (Fig. 4F–F″, yellow wedge). This signal corresponded to the ectopic 61C7-8 chromatin. The results obtained for Jil-1 binding were identical. A broad staining for Jil-1 in 21F3-22A1 (Fig. 4G–G′), was broken into two unequal parts by the inserted cassette domain (Fig. 4H–H″). On recombination of 61C7-8 DNA (pattP61C), an additional signal appeared within the split band (Fig. 4I–I″). Similarly, we also detected BEAF32 and CP190 binding at the endogenous locus and at the recombined 61C7-8 region (data not shown).

To uncover sequences within the 61C7-8 domain essential for the decondensed state, we tested partial deletions within pattP61C. First, we asked whether both boundaries of 61C7-8 equally contributed to the decondensed state. Therefore, partial deletions of the proximal and distal boundary fragments (pattP61CΔd, pattP61CΔp; see Fig. 1A; Fig. 5A) were recombined into the cassette domain of strain 42. Deletion of ∼2.7 kb from the 5-kb distal fragment of pattP61C, which included a strong Chriz- and CP190-binding region (pattP61CΔd), did not affect the formation of decondensed chromatin. Such a truncated fragment still formed a split band at ectopic sites (Fig. 5B,B′), and bound Chriz (Fig. 5D–D″) and Jil-1 (Fig. 5E–E″). In contrast, a deletion of ∼2.3 kb from the 4-kb proximal region, which removes strong binding sites for Chriz and CP190 (pattP61CΔp) but still contains the intact 5-kb distal fragment, abrogated the capacity of the remaining DNA to form decondensed chromatin at ectopic sites (Fig. 5C,C′), and neither Chriz nor Jil-1 were bound by this element (Fig. 5F–F″,G–G″).

Although both deleted fragments contribute to the decondensed 61C7-8 chromatin, only the proximal fragment is essential for its maintenance. Thus, if Chriz, Jil-1 and CP190 are required for decondensed chromatin formation, their binding is essential at the proximal 61C7-8 fragment only. Additionally, two predicted promoter elements for genes that are moderately expressed in salivary glands (CG3402, Med30 and Rev1) are located within the DNA deleted in the proximal 61C7-8 fragment in pattP61CΔp (Fig. 6A), and these might be required for open chromatin formation. Therefore, we asked whether deletion of either one (pattP61CΔP1 or pattP61CΔP2; see Fig. 1B) or both of these promoter elements (pattP61CΔP1+2; see Fig. 1B) would abrogate the capacity of the proximal fragment to induce decondensed chromatin. Deletion of the more-distal promoter P2 (pattP61CΔP2) did not affect the capacity of the fragment to form decondensed chromatin (Fig. 6D–E′) nor did it affect Chriz and Jil-1 binding (data not shown). However, deletion of the more-proximal promoter P1 alone (pattP61CPΔ1) or in combination with P2 (pattP61CPΔ1+2) abrogated formation of open chromatin (Fig. 6B–C′,F–G′) and abolished Chriz and Jil-1 binding (data not shown). Next, we asked whether open chromatin formation is correlated to the transcription mediated by this promoter. First, we stained with antibodies against the initiating form of RNA polymerase II (RNA-Pol-II₀ phosphorylated at S5). We found that this enzyme bound at the endogenous 61C7-8 locus (Fig. 7E–E″) and within the split chromatin domain cassette formed by the pattP61C-positive control (Fig. 7A–A″) or the P2 promoter deletion (Fig. 7C–C″). In strains with insertions containing P1 promoter deletions (pattP61CΔP1, pattP61CΔP1+2), the chromatin domain cassette was neither split nor stained by RNA-Pol-II₀ (Fig. 7B–B″,D–D″). Next, we used qRT-PCR to measure transcription of truncated Rev1 started from P1 at the ectopic position in salivary glands (Fig. 7F). Owing to its fusion with vector sequences, this transcript can be discriminated from the endogenous Rev1 transcript. The Rev1 transcription rate is almost identical at the endogenous and the transgenic locus. In case of the P1 promoter deletion, Rev1 transcription was reduced at least 5-fold compared to the positive control (pattP61C). Surprisingly, however, the P2 promoter deletion reduces transcription of Rev1 to the same extent. A contribution of P2 to Rev1 transcription is also suggested by the even stronger, 10-fold reduction of Rev1 transcription upon P1+P2 promoter deletion (Fig. 7F). Taken together, these results suggest that the presence of the 490-bp promoter fragment P1, but not the transcriptional activity of Rev1 induced by this promoter, is required for the formation of 61C7-8 open chromatin. We were not able to assess whether the same is true for Med30 because in that case we could not discriminate Med30 transcription at the cassette domain from that at the endogenous locus.

Here, we describe a system to test the relationship between DNA sequence, chromatin structure and gene expression in Drosophila. By site-specific recombination into a condensed chromatin domain cassette, effects of different DNA molecules on the chromatin state can be compared at a molecular and cytogenetic level. Owing to the efficiency of the ΦC31 recombinase (Groth et al., 2004), the new insertions can be obtained at rather high rates to systematically screen a series of modified DNA constructs.

P-elements preferentially insert into open chromatin and remain inconspicuous cytologically in polytene chromosomes unless they carry a DNA fragment of ≥5 kb that is not transcribed. In the latter case, new bands are generated that are visible in electron micrographs (Semeshin et al., 1986; Semeshin et al., 1989). We used this approach to establish a cassette domain in the Drosophila genome that forms an ectopic condensed band on polytene chromosomes that is suitable for cytogenetic analysis. P-elements which contained 21 kb of DNA that is not transcribed in salivary glands, including an internal ΦC31 attB recombination site, were randomly integrated and screened for insertion. A total of 16 of 17 lines that were mapped by in situ hybridization on polytene chromosomes formed new condensed bands within interbands. In line 42, a homozygous viable insertion located to a decondensed interband region (21F3-22A1-2) that has a high gene density. The insertion is in the 5′UTR of the annotated 4.7-kb CG4896 gene and 117 bp downstream of the 4.6-kb CG4887 gene. According to data from FlyBase, transcription of CG4887 is at low and that of CG4896 is at very low levels in salivary glands of third-instar wandering larvae, and the immediately adjacent genes CG5118 and CG5126 also show very low expression. This suggests that this interband is formed by a largely transcription-independent process.

The recombination of attP DNA from the 3C6-7 interband into line 42 resulted in the splitting of the condensed chromatin domain cassette by 3C6-7 chromatin. The extent of the decondensed 3C6-7 interband had previously been mapped by genetics and high-resolution in situ hybridization (Keppy and Welshons, 1977; Rykowski et al., 1988) and it has been suggested that it forms independently of transcription (Rykowski et al., 1988; Semeshin et al., 2008).

Data from the Zhimulev laboratory also suggest a transcription-<~?show=[to]?>independent mechanism for the formation of open chromatin by the 61C7-8 interband (Semeshin et al., 1989; Demakov et al., 1993; Demakov et al., 2004; Semeshin et al., 2008). DNA from the 61C7-8 interband was first cloned starting from a P-element insertion that formed a new band within this interband (Demakov et al., 1993). However, using this approach, it was not possible to map the extent and the boundaries of the 61C7-8 interband. Therefore, based on the knowledge on interbands available at that time, the size of the 61C7-8 interband DNA was estimated to extend to 2–5 kb surrounding the P-element insertion. Later, a corresponding 4.7-kb DNA fragment from 61C7-8 was demonstrated to form decondensed chromatin autonomously (Semeshin et al., 2008). However, walking across the 61C7-8 region by high resolution in situ hybridization (H. S. unpublished), we determined that the 61C7-8 interband is much larger and covers a ∼17-kb DNA segment starting 3.5 kb distal to the P-element insertion that was used to initiate the cloning of this interband (Fig. 5A; Demakov et al., 1993). Consistently, our probe containing the proximal 4 kb of this ∼17-kb segment hybridized within the 61C7-8 interband (Fig. 4C–C″). This interband contains one noncoding and four coding transcripts and, as a whole, is enriched in histone modifications and protein-binding sites typically found in active chromatin. We suggest that 61C7-8 forms an open transcriptionally active chromatin domain both in polytene cells of salivary glands and diploid S2 cells. To investigate the mechanism for open domain formation, we tested a plasmid containing a fusion of the distal 5-kb and proximal 4-kb fragments of the 61C7-8 region, on the assumption that these fragments contain all the elements required for decondensed chromatin formation. As expected, the recombination of this plasmid into the ectopic band of strain 42 induced a splitting of the ectopic band, and in situ hybridization showed that the recombined DNA coincides with the decondensed chromatin of the split. The insertion of a control vector that lacked the 61C7-8 sequences was unable to form open chromatin. Therefore, we conclude that sequences required for formation of the ectopic decondensed chromatin are within the 9-kb 61C7-8 fragment used.

Chromatin proteins that bind at the endogenous 61C7-8 locus are also bound at the new position. We observed prominent binding of Chriz, Z4 (data not shown) and Jil-1, proteins that are typically found in decondensed chromatin as reported previously (Eggert et al., 2004; Gortchakov et al., 2005; Wang et al., 2001). Chriz and Jil-1 physically interact (Rath et al., 2006), and on Chriz knockdown both Jil-1 chromosomal binding and H3S10 phosphorylation are dramatically reduced, indicating that Chriz is required for chromosomal targeting of Jil-1 (Gan et al., 2011). Forced local recruitment of active Jil-1 kinase results in local chromatin decondensation (Deng et al., 2008). Hypomorphic mutations in Jil-1 and Chriz, by contrast, result in chromosome condensation and loss of interphase chromatin structure (Wang et al., 2001; Rath et al., 2006, Gan et al., 2011), similar to what is observed on loss of Z4 protein (Eggert et al., 2004). Besides these factors, we observed binding of the insulator proteins CP190 (Whitfield et al., 1988; Pai et al., 2004) and BEAF 32 (Zhao et al., 1995; data not shown) to the ectopic 61C7-8 position, resembling the situation at the endogenous site. In conclusion, the epigenetic state of 61C7-8 chromatin is conserved at the ectopic position and the reported properties of the bound proteins suggest that they contribute to the decondensed state.

By further dissecting the 9-kb 61C7-8 element, we found sequences from the proximal boundary that are required for the formation of open chromatin. Surprisingly, the distal 5 kb of DNA alone did not support autonomous open chromatin formation in our assay although it binds Chriz and CP190 in several Drosophila cell lines (FlyBase, modENCODE) and forms part of decondensed chromatin of the 61C7-8 interband. Our results are in conflict with data by Semeshin et al. (Semeshin et al., 2008), who used a 4.7-kb 61C7-8 fragment that is included in the distal-most 5 kb of our constructs (Fig. 5A). This 4.7-kb fragment, framed by band-forming chromatin of their pICon(dV)-61C vector, formed decondensed chromatin when inserted into the genome. On removing the 61C7-8 sequence from the construct by FLP/FRT recombination, the remaining vector sequences lost the capacity to form open chromatin, suggesting that this was a property of the 4.7-kb element. Currently, the discrepancy with our observations is not clear. One explanation might be the presence of an hsp70 promoter element in their constructs that was immediately adjacent of their 61C7-8 insertion and might have supported open chromatin formation in its close vicinity. In our constructs, no promoter sequences are present next to the distal fragment and in the absence of the essential element of the proximal fragment condensed chromatin might spread on 61C7-8 interband DNA.

There are two predicted promoter regions in the proximal 61C7-8 fragment. Promoter P1 is immediately upstream of both the Rev1 and Med30 gene that are transcribed in opposite directions. Promoter region P2, which is downstream of Med30, is immediately upstream of CG3402, which is transcribed on the same strand as Med30 (Fig. 6A). Deletion experiments showed that the P1 but not the P2 element is essential for open chromatin formation. All three genes are moderately expressed in salivary glands (FlyBase, modENCODE), suggesting that active transcription from P1 mediates the open chromatin formation. Consistently, we observed that RNA polymerase II phosphorylated at S5 bound at the endogenous and ectopic 61C7-8 chromatin. The P1 deletion, which represses open chromatin formation strongly reduced Rev1 transcription. Surprisingly, the P2 deletion, which shows no repression of open chromatin formation, reduces Rev1 transcription to a similar extent. This argues that it is the presence of the P1 element rather than the transcriptional activity of Rev1 that contributes to open chromatin formation. We do not know whether Med30 transcription is similarly dependent on the P2 promoter as is Rev1. If this is the case, then open chromatin formation and transcriptional activity are uncoupled. If not, then open chromatin formation of 61C7-8 might be caused by transcription from the P1 promoter.

Alternatively, but not mutually exclusive, open chromatin formation might be induced at P1 by an epigenetic mechanism. P1 overlaps with strong binding sites for Chriz and CP190 and other factors, like BEAF32. The chromatin-decondensing function of H3S10 phosphorylation, driven by Jil-1, is one possible mechanism for open chromatin formation, as described above. The Z4/Ptz subunit of the Chriz complex has also been suggested to cooperate with the nucleosome-remodeling factor complex NURF (Kugler et al., 2011) to support open chromatin formation. The whole region is depleted for inactive chromatin marks, like H3K9 and H3K27 trimethylation, and enriched for histone modifications typically found in active chromatin, like histone acetylation and H3K4 trimethylation. This indicates a role for local histone modifiers in the recruitment of activators and chromatin remodeling activities. As a result, a patch of open chromatin for loading paused RNA polymerase II would be formed and the spreading of adjacent more-compact chromatin would be suppressed. P1, which is located between two oppositely transcribed genes, might form a particularly strong site for local open chromatin formation. In conclusion, in such a scenario, open chromatin formation would be induced by factor binding. This would be required for, but uncoupled from, the process of active transcription. If it fails, both processes would be strongly repressed.

For the assembly of the Notch–white band construct (pUAST-NattBw), Notch sequences were amplified from genomic DNA by PCR and cloned into pBluescript. The 3.9-kb Sac2/BamH1 Notch fragment was amplified using primers 5′-ACGCCGCGGATTATATATCTGTTGGAT-3′ and 5′-ACGGGATCCGTGGGTTATTTTACTTTG-3′, the 5.1-kb BamH1/Sma1 Notch-fragment by using primers 5′-ACGGGATCCCGAGATCCAAGAGTCTTA-3′ and 5′-ACGCCCGGGAATGAGAAAATTACACGC-3′, and the 4.0 kb EcoR1/Xho1 Notch-fragment was amplified using primers 5′-ACGGAATTCACATTTAACTTGCGCCAG-3′ and 5′-ACTGGTACCACATACCGAACAAGCAGC-3′. The 4.5-kb EcoR1 white-attB fragment was amplified from pUASTB (Groth et al., 2003) using primers 5′-ACGGAATTCGTCGACGATGTAGGTCA-3′ and 5′-ACGGAATTCAGTACGAAATGCGTCGTT-3′. The mini-white gene was used as a reporter. The whole 17.3 kb sequence consisting of the three Notch-fragments and the white-attB fragment was cut out from pBluescript and cloned into the P-element vector pUAST using Kpn1 and Sph1 restriction sites.

For the assembly of the 61C7-8 interband construct (pattP61C) the proximal and distal part of the 61C7-8 interband was amplified from genomic DNA by PCR and cloned into pTAattP (Groth et al., 2000). The proximal part of 61C7-8 (∼4 kb) was amplified as a HindIII/Kpn1 fragment using primers 5′-ACGGGTACCATTCGCTCTTTCAGCG-3′ and 5′-ACGAAGCTTAGTAGCCGCCCTGAAAAG-3′ and the distal part of 61C7-8 (∼5 kb) was cloned as a Sac1/Kpn1 fragment using primers 5′-ACGGAGCTCCTTTTGCCATCGAATCGG-3′ and 5′-ACGGGTACCATGTGGTCAGCATCGGC-3′. The UAS-GFP reporter gene (∼1 kb) was amplified from pUAS-GFP as a Not1 fragment using primers 5′-CATGCGGCCGCCTCAAGCTTCGAATTCTGCA-3′ and 5′-AATGCGGCCGCATCTAGAGTCGCGGCCGCTTTA-3′. The plasmid pattP61Cprox was obtained as an in situ hybridization probe specific for the proximal part of the 61C7-8 interband by cloning only the proximal HindIII/Kpn1 PCR fragment into pTAattB as described above. The 61C7-8 interband deletion construct pattP61CΔd was obtained by digestion of pattP61C using Hpa1/BstZ17I followed by religation of the vector after removal of the ∼2.7 kb fragment. Similarly, pattP61CΔp was obtained by SnaB1/Swa1 digestion and religation following removal of the ∼2.3 kb fragment.

Potential promoter regions were predicted by using the McPromoter Prediction algorithm and the Promoter 2.0 Prediction algorithm as well as by using modENCODE database to search for characteristic molecular features. For the deletion of the predicted promoter elements within the proximal part of pattP61C (pattP61CΔP1, pattP61CΔP2 and pattP61CΔP1+2), 5′-phosphorylated primers were used that amplified the whole sequence of the pattP61C except the promoter element, followed by blunt-end ligation of the resulting fragment. For the ∼490-bp deletion of promoter-1, primers Ph5′-CAAAATGTGGAAATACGGTCAA-3′ and Ph5′-ATAATGGTTTTAGTGAATGGGTA-3′ were used. For the ∼160-bp deletion of promoter-2, primers Ph5′-ATCGATAAATGATTGCGAGGAAG-3′ and Ph5′-TGTTTGGCAATGTTTCTTAGTTT-3′ were used. For deletion of both promoter elements, the construct carrying the promoter-1 deletion was the template for promoter-2 deletion.

To generate the empty vector control construct (pattPempty) the UAS-GFP reporter gene was amplified from pUAS-GFP using primers 5′-CATGCGGCCGCCTCAAGCTTCGAATTCTGCA-3′ and 5′-AATGCGGCCGCATCTAGAGTCGCGGCCGCTTTA-3′ and the resulting DNA was cloned into pTAattP using Not1. To assemble the 3C6-7 interband construct (pattP3C), 2.6 kb of the 3C6-7 region was amplified from genomic DNA using primers 5′-ATAGGGCCCTTCTGATTCAAGCGGTGTG-3′ and 5′-ACGCTCGAGACTCGGCTTTCGTCTCACTC-3′ and the fragment obtained was inserted into pTAattP using Apa1/Xho1 as restriction sites. The UAS-GFP reporter gene was amplified from pUAS-GFP plasmid as a Not1 fragment using primers 5′-CATGCGGCCGCCTCAAGCTTCGAATTCTGCA-3′ and 5′-AATGCGGCCGCATCTAGAGTCGCGGCCGCTTTA-3′ and inserted into the pattP3C vector. For the condensed band forming control vector (pattPN) a 5.6 kb Sac1 Notch fragment of the pUAST-NattBw construct was cloned into the Sac1 site of the pattPempty control vector. The BstZ17I site of the resulting vector was used for the insertion of a 4.7-kb BstZ17I Notch fragment obtained from pUAST-NattBw.

For the generation of transgenic recipient fly lines, pUAST-NattBw was coinjected with transposase containing helper plasmid pπ25.7wc (Karess and Rubin, 1984) into w¹¹¹⁸ recipient lines. The progeny was screened for insertions based on eye color.

For the targeted recombination of the interband and control constructs, the transgenic recipient line 42 at 21F3-22A1-2 containing an attB-landing site within the band domain was crossed with animals transgenic for ΦC31 integrase on the fourth chromosome (number 23649, M{vas-int.B}ZH-102D; Bloomington). Heterozygous embryos were injected with the desired interband or control construct, respectively. The resulting adult flies were crossed with the salivary-gland-specific diver line G61-GAL4 and screened for recombinants based on GFP expression in salivary glands.

For the genomic localization of the P-element insertions by inverse PCR genomic DNA was digested with Sau3A1 and re-ligated. The fragment including a part of the P-element termini and the adjacent genomic DNA, was amplified by PCR using primers 5′-CACCCAAGGCTCTGTCCCACAAT-3′ and 5′-GTAACGCTAATCACTCCGAACAGGTCA-3′ [for the 5′ primer extension preamplification (PEP) end] or 5′-CCTTAGCATGTCCGTGGGGTTTGAAT-3′ and 5′-CAATCATATCGCTGTCTCACTCA-3′ (for the 3′ PEP end). The PCR product was gel-extracted, sequenced using sequencing primers 5′-ACACAACCTTTCCTCTCAACAA-3′ (for the 5′ PEP end) or 5′-GACACTCAGAATACTATTC-3′ (for 3′ the PEP end) and analyzed by BLAST.

The cytogenetic localization of the insertions was determined by fluorescence in situ hybridization (FISH). Biotin-labeled DNA probes were synthesized by treatment of the corresponding DNA construct with BNT-Mix (Roche) according to the manufacturer's protocol. FISH was performed as described by Langer-Safer et al. (Langer-Safer et al., 1982). For microscopy, a Delta Vision Spectris Optical Sectioning microscope (OSM) was used equipped with 60× and 100× lenses. Data are single sections from image stacks processed by deconvolution using Deltavision SoftWorx.

The transgenic lines were checked for correct insertion by diagnostic PCRs as well as by sequencing. Accordingly, fragments spanning the boundaries and middle parts of the integrated vectors were amplified. The sequence of the primers used for these tests is available from the corresponding author on request.

Polytene chromosomes were prepared from third-instar larvae. Immunostaining was performed as described by Eggert et al. (Eggert et al., 2004). All polyclonal antisera were obtained from Biogenes (Berlin) following immunization with affinity purified proteins expressed in E. coli by our laboratory. Primary antibodies against the following proteins were used: Chriz (rabbit, 1∶1000), CP190 (mouse monoclonal Bx63, 10 µg/ml), Jil1 (rabbit, 1∶1000), Z4 (rabbit, 1∶1000), and RNA PolII phosphorylated on S5 (mouse, 1∶750, Covance USA). As secondary antibodies we used: Alexa-Fluor-488- or -555-conjugated goat anti-mouse-IgG or anti-rabbit-IgG antibodies (Invitrogen) at 1∶1000 dilution. Microscopy was as described for the in situ hybridization.

Salivary glands of 30 third-instar larvae of the desired genotype were dissected and the total RNA was isolated according to the two-step TRIZOL RNA isolation protocol (W.M. Keck Foundation Biotechnology Microarray Resource Laboratory). cDNA was synthesized using RevertAid Premium Reverse Transcriptase (Thermo Scientific) following the manufacturer's protocol and subsequently used as a template for quantitative real-time PCR. The qRT-PCR was performed in a StepOnePlus Real-Time PCR Systems (Applied Biosystems) machine using the Two-Step Power SYBR Green PCR Master Mix (Applied Biosystems). The GFP expression analysis was done using primers 5′-AGGAGCGCACCATCTTCTTC-3′ and 5′-GTCCTCCTTGAAGTCGATGC-3′. The Rev1 expression analysis was done using primers 5′-TGTGAGCGGATAACAATTTCA-3′ and 5′-CATTGTACATAAACCAAACCTGC-3′, for the transgenic region, and primers 5′-TCCGACTTGCGAAATGGATC-3′ and 5′-GTCCTCCTTGAAGTCGATGC-3′, for the endogenous region. Both Rev1 primer pairs were tested and revealed an almost identical efficiency. Actin42a was used as the endogenous control using primers 5′-AGCGGATAACTAGAACTACTCC-3′ and 5′-CTAAAGCTGCAACCTCTTCGT-3′. The cycling parameters were as follows: 10 min at 95°C, 40 cycles of 15 s at 95°C followed by 1 min at 60°C. Finally a melting curve was recorded.

The modENCODE database (www.modencode.org) was used to analyze the interband regions 61C7-8 and 3C6-7 and the P-element insertion sites for their specific molecular features.

We thank Ana Pombo (Max Delbrück Center for Molecular Medicine, Germany) for providing RNA Pol II₀ antisera, Ansgar Klebes for critical reading our manuscript and Alexander Glotov (both Cytogenetics Group, Institute of Biology, Humboldt-University Berlin Germany) for his help in assembling the qRT-PCR. The support by all members of the Cytogenetics group is also gratefully acknowledged.