Insulator and Ovo Proteins Determine the Frequency and Specificity of Insertion of the gypsy Retrotransposon in Drosophila melanogaster

Abstract

The gypsy retrovirus of Drosophila is quite unique among retroviruses in that it shows a strong preference for integration into specific sites in the genome. In particular, gypsy integrates with a frequency of >10% into the regulatory region of the ovo gene. We have used in vivo transgenic assays to dissect the role of Ovo proteins and the gypsy insulator during the process of gypsy site-specific integration. Here we show that DNA containing binding sites for the Ovo protein is required to promote site-specific gypsy integration into the regulatory region of the ovo gene. Using a synthetic sequence, we find that Ovo binding sites alone are also sufficient to promote gypsy site-specific integration into transgenes. These results indicate that Ovo proteins can determine the specificity of gypsy insertion. In addition, we find that interactions between a gypsy provirus and the gypsy preintegration complex may also participate in the process leading to the selection of gypsy integration sites. Finally, the results suggest that the relative orientation of two integrated gypsy sequences has an important role in the enhancer-blocking activity of the gypsy insulator.

RETROVIRAL DNA integration into the host genome is an essential step for production and replication of viral RNA. It has been traditionally difficult to study the factors controlling selection of integration sites, since most retroviruses integrate throughout the genome with no apparent DNA sequence specificity (Bushman et al. 2005). However, detailed analysis of multiple genomic integration sites in vivo has revealed that retroviruses have a strong preference for certain genomic regions. In particular, retroviruses integrate preferentially into actively transcribed DNA, which will thereafter facilitate transcription of the provirus (Scherdin et al. 1990). The distribution of retroviral integration sites along chromosomes suggests that open chromatin favors retroviral insertion, since integration events are favored in transcriptionally active chromatin and are rare in DNA sequences associated with heterochromatin (Lewinski et al. 2005; Yant et al. 2005). However, chromatin state or DNA accessibility could not be the only factor influencing integration, since different retroviruses manifest preferences for integration that are unlikely to be only the result of chromatin organization. For example, both HIV and murine leukemia virus (MLV) integrate in actively transcribing DNA, but HIV integrates with equal frequency throughout all transcribed DNA, whereas MLV integrates preferentially into transcription start sites (Schroder et al. 2002; Wu et al. 2003).

The mechanism of integration of retrotransposons is fundamentally identical to that of retroviruses (Bushman 2003). However, constraints imposed by small genome sizes have led some retrotransposons to the acquisition of mechanisms for site-specific integration. The best examples of site-specific integration are found in non-LTR retrotransposons such as the Drosophila telomeric elements HeT-A, TART, and TAHRE or rDNA elements such as the Drosophila R1 and R2 non-LTR retrotransposons (Yang et al. 1999; Christensen and Eickbush 2005; George et al. 2006). A number of examples involving integrases from yeast LTR retrotransposons have also shown that retroviral-like integrases have evolved to acquire strong site-specific integration properties (Sandmeyer 2003; Zhu et al. 2003; Brady et al. 2008; Gao et al. 2008). For example, the integration of Saccharomyces cerevisiae retrotransposons Ty1 and Ty3 is associated with RNA polymerase III transcription (Yieh et al. 2000; Bachman et al. 2005; Mou et al. 2006) and the Tf1 retrotransposon from Schizosaccharomyces pombe integrates near RNA polymerase II promoters (Singleton and Levin 2002; Bowen et al. 2003). In other examples, transposable elements are targeted to heterochromatic sites by tethering mechanisms involving interactions between the integrase and DNA binding proteins. Targeting of the Ty5 retrotransposon from S. cerevisiae to heterochromatin, for example, requires a six-amino-acid motif at the C terminus of the Ty5 integrase that interacts with the heterochromatin protein Sir4 (Gai and Voytas 1998; Xie et al. 2001). It has also been proposed that the chromodomain (CHD domain) from certain transposable element chromointegrases targets the retrotransposon for insertion into sites bearing the specific epigenetic marks recognized by the CHD domain (Gao et al. 2008).

In contrast to mammalian retroviruses, Drosophila retroviruses such as gypsy, ZAM, or Idefix, display a high rate of site-specific integration into certain regions of the genome (Mevel-Ninio et al. 1989; Desset et al. 1999; Leblanc et al. 1999; Conte et al. 2000). The mechanisms governing this specificity however are poorly understood, but the genetic tools available in Drosophila provide a unique opportunity to analyze retroviral site integration specificity in higher eukaryotes. In particular, gypsy insertions into the ovo locus occur in the germ line of ∼10% of the female offspring from mothers carrying permissive mutations in the flamenco (flam) locus (Prud'homme et al. 1995; Dej et al. 1998). Gypsy integrations take place specifically into a sequence of ∼1.3 kb spanning the 5′ regulatory region of the ovo gene (Dej et al. 1998). The flam locus is located in the heterochromatin of the X chromosome (Prud'homme et al. 1995) and produces a long noncoding RNA that controls transcription of the gypsy retrovirus through the piwiRNA pathway (Brennecke et al. 2007).

The process of gypsy transposition is maternally regulated, involving maternally inherited gypsy particles that originate in the developing oocyte of flam mutant females. These females fail to produce the flam RNA, allowing the transcription of euchromatic gypsy elements in the follicle cells surrounding the oocyte during oogenesis. Transcription of gypsy in follicle cells leads to the formation of virus particles that infect the oocyte and subsequently participate in the integration of gypsy in the germ line of the resulting embryo after fertilization (Song et al. 1994, 1997). These integration events take place preferentially in the ovo gene, whose product is necessary for the development of the female germ line and the normal progression of oogenesis (Lu et al. 1998). The ovo gene encodes two isoform proteins, Ovo-A and Ovo-B, which have a common DNA-binding domain but different N-terminal domains. Ovo-B positively regulates the ovo promoter, whereas Ovo-A functions as a negative regulator of the ovo promoter (Andrews et al. 2000). Adult females homozygous for a null mutation of the ovo gene do not develop germ line cells. The ovo^D1 allele is caused by a point mutation that creates a new in-frame methionine codon in the 5′ region of ovo, adding an extra amino terminus domain to Ovo-B that is normally only present in the wild-type Ovo-A protein (Mevel-Ninio et al. 1996). The ovo^D1 allele is dominant negative and causes female sterility even when heterozygous. The sterility is due to the expression of Ovo^D1B protein, which is made at the same time of development as Ovo-B but has the repressor activity of Ovo-A; the presence of Ovo^D1B is sufficient to arrest oogenesis at stage 4 (Lu et al. 1998).

Insertion of gypsy into the ovo^D1 allele in a heterozygous female reverts to fertility by preventing the expression of the Ovo^D1B protein, although the reversion occurs only in those germ cells in which gypsy is inserted into the ovo^D1 sequence. The ability of gypsy to integrate specifically into ovo sequences was analyzed by Dej et al. (1998). These studies concluded that gypsy integrates in at least seven different target sites localized within a 200-bp sequence present in the promoter region of the ovo gene. Close analysis of these sites reveals a very relaxed consensus sequence consisting of six alternating pyrimidines and purines. The weak conservation of the observed target sequence suggests that gypsy site-specific integration is not due to a direct interaction of the gypsy integrase with these sequences. Instead, it has been proposed that Ovo proteins may mediate gypsy insertion specificity by promoting protein–protein interactions between Ovo or an associated protein and the gypsy preintegration complex (Labrador and Corces 2001).

The gypsy retrovirus of Drosophila also exhibits the interesting property of blocking enhancers from activating promoters when gypsy is inserted between them. This property is referred to as insulator activity and resides in the Suppressor of Hairy wing [Su(Hw)] binding sites present in a 350-bp sequence located in the 5′-UTR of the gypsy retroviral genome (Geyer et al. 1986; Spana et al. 1988; Gerasimova and Corces 1998, 2001; Gerasimova et al. 2000). In addition, gypsy insulators are also able to buffer transgenes from position effects by preventing heterochromatin from spreading through the chromatin fiber (Roseman et al. 1993). There is mounting evidence suggesting that gypsy insulators function by creating chromatin domains most probably defined by the interaction between adjacent insulator sites in chromosomes. Molecular evidence for such interactions has been obtained by measuring the distance between adjacent gypsy insertions in wild-type and in su(Hw) mutant cells (Gerasimova et al. 2000). These experiments revealed that the two gypsy sequences were significantly closer during interphase when the Su(Hw) protein was present. It has been proposed that such interactions might create chromatin domains by looping out the DNA contained between two interacting insulators. Additional evidence in support of this model has been provided by showing the presence of DNA loops attached at their base to the nuclear matrix by the gypsy insulator in the nucleus of Drosophila imaginal disc cells (Byrd and Corces 2003). Interaction between gypsy insulators is also supported by data showing that two adjacent insulators were able to cancel each other, no longer exerting their enhancer blocking effect when located between the enhancer and the promoter of a reporter gene (Cai and Shen 2001; Muravyova et al. 2001; Kuhn et al. 2003).

The molecular basis for interactions between individual insulators is not well understood but it has been suggested that Modifier of mdg4 [Mod(mdg4)] and CP190, both components of the insulator complex, might facilitate such interactions by mediating protein–protein contacts between the BTB domains present in the two proteins (Gause et al. 2001; Ghosh et al. 2001; Golovnin et al. 2007). These two properties of gypsy, site-specific integration and insulator activity, have been the subject of intense but unrelated studies during the past two decades (Mevel-Ninio et al. 1989; Bucheton 1995; Prud'homme et al. 1995; Dej et al. 1998; Gerasimova and Corces 2001; Labrador and Corces 2001; Capelson and Corces 2004). Here, we attempt to analyze these two properties simultaneously in an effort to understand how insulators might mediate genome organization and how this organization may influence retroviral selection of integration sites through the genome. We have previously developed an assay to show that the 5′ regulatory region of ovo is able to recruit gypsy insertions independently of its position in the genome (Labrador and Corces 2001). We have also provided genetic evidence suggesting that the Ovo protein is directly implicated in such recruitment. We now take advantage of the ability of inducing two consecutive gypsy insertions into a yellow reporter gene to analyze the role that interactions between a gypsy provirus and the gypsy preintegration complex may play in the selection of retrovirus integration sites and the effect of the relative orientation of interacting proviruses on the enhancer-blocking activity of the gypsy retrovirus.

MATERIALS AND METHODS

Cloning DNA sequences containing Ovo binding sites:

The three constructs ovo^WT, ovo^EN, and ovo^DIM were obtained using polymerase chain reaction (PCR) following standard procedures (94° for 1.0 min, 60° for 1.5 min, and 72° for 2.0 min). Each sequence was cloned into the NotI site of the Casper-yellow vector as described (Labrador and Corces 2001). ovo^WT was constructed by PCR amplification using genomic DNA from the Oregon R wild-type strain as template. The primers OVO-F (5′-GATGGGTCGCGGCCGCTTGAAAACAGCAGAAAAATA-3′) and OVO-R (3′-ATACTAATTAAGCATAAATTCGCCGGCGTATTGTAC-5′) were used to amplify ovo^WT. ovo^EN and ovo^DIM were each built from a set of 50–60 overlapping oligonucleotides as follows.

• OVO-dim1: 3′-CTACCCAGCGCCGGCGAACTTTTGTCGTCTTTTTATAT-5′

• OVO-dim2: 5′-CAGCAGAAAAATATGCCAATTTGTTTTGAATTTAACAGATTTT-AATAAATAGTTTTAACTTAATGG-3′

• OVO-dim3: 3′-CAAAATTGAATTACCAAGTCGTGTAAAATTTGAAGGGAT-TCATATACGTATCGGGA-5′

• OVO-dim4: 5′-ATATGCATAGCCCTGTTTTAATTTTTAATTTCAAGTTAATAA-CTTTTATTCACATCATACT-3′

• OVO-dim5: 3′-ATAAGTGTAGTATGAGTCTATCATATAAAATCACAT-GAGAAAACTCTATTAAAGTAGTG-5′

• OVO-dim6: 5′-AGATAATTTCATCACTTGTTGATTCGTCGTTGGCAACTCTGC-ACCCACGATTCCAAATAGGATTA-3′

• OVO-dim7: 3′-AAGGTTTATCCTAATCTTACTGTGTGTGGGTACCTTAACTTC-TTACATTCCTTCTTTCC-5′

• OVO-dim8: 5′-GTAAGGAAGAAAGGGAGTGTGATCGAAAGTCCGTTCC-TTTGCTCAAATAGATAGCAATCGTC-3′

• OVO-dim9: 3′-CTATCGTTAGCAGGCTCGCTTGCCTGTCTGTTTAAAGACTCTT-AGCGTGAAGAAACGAAGAGAG-5′

• OVO-dim10: 5′-TTCTTTGCTTCTCTCATTTTCGGTGATTTTAGAATGCTTGC-TTATTGTGTGTGCACTCGAAAGTTCTATT-3′

• OVO-dim11: 3′-GAGCTTTCAAGATAATCCAAGGTGTCCCAAAAATATGT-ATACTAATTAAGCATAAATTCGCCGGCGTCTTGTAC-5′

• OVO-dim12: 5′-TTTATACATATGATTAATTCGTATTTAAGCGGCCGCAGCAGA-TGCTACTTAACGT-3′

• OVO-en1: 5′-AAATGACGATGGGTCGCGGCCGCTTGAAAACAGCAGAAAAA-TAAAGCCGTTAAAATT-3′

• OVO-en2: 3′-ATTTCGGCAATTTTAACTTTTTCACGTCAAATTTACATTGAC-AATTATACTCGTCATATAGT-5′

• OVO-en3: 5′-TGAGCAGTATATCACGACTACAGTTAGAATTAGCTCTACGG-ATCCT-3′

• OVO-en4: 3′-GAGATGCCTAGGAAAAATGTCAATGTATCGTCTCAGGAACC-TAAAAGGCAACGAAAAAA-5′

• OVO-en5: 5′-TTCCGTTGCTTTTTTATTGAGGCTGTGTGCAAAATAAAGCCG-TTAAAATTGAAGGTTCCAC-3′

• OVO-en6: 3′-AACTTCCAAGGTGCCAATTTACATTGACAATTATACCGA-TAAATATGTCGTCCGAGGA-5′

• OVO-en7: 5′-CAGCAGGCTCCTTTTTACAGTTACATATGATTAATTCGTAT-TTAAGCGGCCGCAGCAGATGCTACTTAACGT-3′

• OVO-en8: 3′-TTGTACAATAAATGATA-5′

We used a two-step PCR amplification as previously described (Dillon and Rosen 1990). In the first PCR step, the ovo^EN template was assembled using 1.0 μl (100 ng/μl) of each of the OVO-en1–OVO-en8 overlapping oligonucleotides plus 0.5 μl Ex Taq polymerase (TaKaRa Biomedicals). This PCR step consisted of seven cycles at 94° for 1.0 min, 50° for 1.5 min, and 72° for 2.0 min. The ovo^DIM template was assembled in identical fashion using OVO-dim1–OVO-dim12. Once the ovo^EN and ovo^DIM templates were assembled, they were subjected to the second round of PCR amplification. The second-step PCR amplification was identical to that used to amplify ovo^WT from genomic DNA, using the same end primers containing NotI restriction sites. After the second-step PCR amplification, cloning of ovo^EN and ovo^DIM into the Casper-yellow vector was identical to the procedure used to clone ovo^WT into the NotI sites of Casper yellow. The Casper-yellow plasmids containing ovo^WT, ovo^EN, and ovo^DIM were sequenced to confirm that all three correspond to the expected sequences shown in Figure 1. The orientation of ovo^WT, ovo^EN, and ovo^DIM in the Casper-yellow vector is opposite to the direction of transcription of the yellow gene. Nucleotide numbers indicating the location of insertions throughout the manuscript are as described by Geyer and Corces (1987). To designate transgenes carrying the Casper-yellow plasmids containing ovo we use the terminology P[y^ovo; w⁺]. The super index ovo^WT, ovo^EN, and ovo^DIM was added to denote the particular ovo sequences cloned into yellow.

Figure 1.—

Ovo DNA sequences cloned into the yellow regulatory region of the different P[y^ovo; w⁺] transgenes. (A) Schematic drawing of the yellow gene with the location of ovo sequences. The black box indicates the upstream regulatory region and the white boxes indicate exons 1 and 2 of the yellow gene. Tissue-specific wing (W), body (B), and bristle (Br) enhancers are indicated as ovals. The open triangle indicates the Ovo sequences inserted between wing and body enhancers and the yellow promoter at nucleotide 1980 of the yellow gene. (B) DNA fragment 508 bp long from the wild-type regulatory region of the ovo gene containing five Ovo binding sites. (C) Synthetic sequence identical to the one shown in B but with the Ovo binding sites mutated to prevent binding of Ovo proteins. (D) Synthetic 314-bp sequence containing eight Ovo binding sites separated by random DNA sequences.

Induction of gypsy mobilization using flam females:

Stocks were maintained on standard cornmeal medium at 25° in a humidity-controlled incubator. The flam flies were maintained as a y v f mal flam/FM3 stock. This stock segregates homozygous flam flies and heterozygous flam/FM3 females. Only heterozygous females were used to maintain the stock each generation. Virgin homozygous y v f mal flam were collected from the stock for crosses requiring homozygous flam flies. ovo^D1 mutants were maintained in a stock by crossing ovo^D1 males to females carrying attached X chromosomes. Flies carrying P[y^ovo; w⁺] transgenes display a yellow wild-type phenotype and were maintained in a y w^67c background (Labrador and Corces 2001). To induce gypsy integrations into the P[y^ovo; w⁺] transgenes, virgin females homozygous for y v f mal flam were crossed to P[y^ovo; w⁺] transgenic males. Since a single insertion in the germ line can give rise to several mutant flies, single female crosses were performed instead of mass crosses. Virgin F₁ females with genotype y v f mal flam/y w^67c; P[y^ovo; w⁺] were individually crossed to y w^67c males. To detect gypsy insertions into the P[y^ovo; w⁺] transgene, the offspring of this cross (F₂ generation) was systematically screened for individuals with a mutant yellow phenotype as follows. Individual y v f mal flam/y w^67c; P[y^ovo; w⁺] females crossed to several y w^67c males were transferred to new vials containing fresh food every 3–4 days. A minimum of three vials was screened per cross and crosses with low numbers of offspring were discarded. The yellow phenotype of the F₂ progeny containing the P[y^ovo; w⁺] transgene was screened under a dissecting microscope and 90–105 flies were counted and screened per cross. This process was repeated two to three times per vial to allow new flies to eclose and the number of flies counted each time was annotated in the vial. Only independent mutant flies (from different mothers) were recorded and crossed in new vials for further analysis. As a positive control, to test whether flam permissive females produced active virus particles, virgin y v f mal flam females were crossed to ovo^D1 males and the offspring were screened for fertile females as described previously (Labrador and Corces 2001). We used the ovo^D1 reversion assay (Prud'homme et al. 1995) to determine the gypsy mobilization activity of the y v f mal flam stock. Five to ten y v f mal flam virgin females were crossed with ovo^D1 males and kept in the same vial for 3–5 days. After this time, parental flies were removed from the vials and the offspring larvae were allowed to develop. Groups of 5–10 ovo^D1/ovo⁺ females from the offspring were transferred to new vials and crossed to wild-type males. ovo^D1/ovo⁺ females are sterile unless the ovo^D1 allele carries a new insertion of the gypsy element integrated into the ovo^D1 regulatory region. After several days, vials were examined for the presence of eggs or larvae. All the mothers in vials containing eggs were dissected under the microscope and females displaying functional ovaries were considered fertile and counted as a gypsy integration event into the ovo^D1 allele.

Analysis of gypsy insertions by PCR and DNA sequencing:

Genomic DNA was extracted from 10–20 flies for the detection of gypsy integrations into P[y^ovo; w⁺] transgenes. DNA extraction was carried out using the potassium acetate quick prep as described (Dej et al. 1998). The primers used to amplify gypsy insertions were gypsy-P3 CTTTGCCGAAAATATGCAATG and gypsy-P1 CAACATGACCG-AGGAGCGGTCATAAAC located outside the LTRs in the 5′ and 3′ ends of gypsy, respectively, and yP0030 GCCCGATTACCACATTGAG, yP1400 GTTGCACAAAA-TTACCGGC, yP1450 CTGTGGGTGCAATGATTAG, yP1120 TCATTGCCGCAAGC-TCTG, and yP2900 CGCCACGGTCCACAGAAGAG, which are located at different points in the regulatory sequences of the yellow gene and therefore could be present adjacent to new gypsy insertions. A combination of P1 or P3 with a yellow primer will detect the insertion of gypsy in either orientation. Alternative combinations of P3 or P1 with alternative primers in the yellow gene can detect the same insertion at the opposite end of gypsy and was used to confirm the results obtained with the first combination of primers. Approximately 50 ng of genomic DNA were used per PCR amplification. Amplified DNA fragments containing gypsy insertion sites were directly sequenced using gypsy-LTR3 AAATCGCTATCGCCACAAGGC or gypsy-LTR5 GCAGCGTGAAG-CAACACTCCC as sequencing primers.

RESULTS

Ovo protein binding sites promote gypsy insertion into a yellow transgene:

We have previously suggested that binding of the Ovo protein to the regulatory region of the ovo gene could account for the specificity of integration into this region of the gypsy retrotransposon (Labrador and Corces 2001). In previous experiments, a DNA fragment containing ∼1.3 kb from the regulatory region of the ovo gene, including the promoter, was inserted into the Casper-yellow transformation vector. Transgenes containing these ovo gene sequences turned into highly specific targets for gypsy insertions, which occurred at a rate similar to that observed at the endogenous ovo gene (Dej et al. 1998; Labrador and Corces 2001).

To provide molecular evidence supporting the hypothesis that Ovo proteins are capable of targeting gypsy insertions into Ovo binding sites, we first asked whether a subset of the ovo sequences used in previous experiments, containing binding sites for the Ovo protein, would also induce a high rate of gypsy insertions. To this end we used PCR to specifically amplify a 508-bp DNA fragment (ovo^WT) containing nucleotides 512–1020 from the ovo gene regulatory sequence (Mevel-Ninio et al. 1996). This fragment lacks the promoter region of the Ovo-A isoform but still contains the promoter region of the Ovo-B isoform and includes five Ovo protein binding sites (Figure 1B). We then cloned the ovo^WT fragment into the NotI site at nucleotide 1980 of the Casper-yellow vector (Figure 1A), which is located between the wing blade and body enhancers and the promoter of the yellow gene (Labrador and Corces 2001). To determine whether the binding of Ovo proteins was necessary to specifically recruit gypsy insertions into Casper-yellow transgenes, we generated a synthetic DNA sequence identical to ovo^WT but containing specific point mutations into the five existing Ovo binding sites. The Ovo binding sites in Figure 1 were identified following the descriptions by Lu et al. (1998).

The induced point mutations result in sequences that strongly deviate from Ovo protein consensus binding sites, whereas the flanking sequences remain identical to those in ovo^WT (Figure 1C). We refer to this sequence as ovo ^DIM. The ovo^DIM fragment was also cloned into the Casper-yellow vector and plasmids containing ovo^WT and ovo^DIM were independently microinjected into y w; Δ2-3 Sb/TM6 embryos. Transgenic lines obtained from both constructs show a wild-type yellow phenotype in a y¹ mutant background, indicating that the yellow ovo^WT and ovo^DIM transgenes (P[y^ovoWT; w⁺] and P[y^ovoDIM; w⁺], respectively) are able to express the Yellow protein in wing blades, body cuticle, and bristle tissues (see Figure 2B).

Figure 2.—

Gypsy-induced yellow phenotypes depend on the integration sites as well as the relative orientation of gypsy. (A) The yellow reporter gene in the Casper-yellow^ovo transformation vector showing the synthetic DNA sequence containing eight Ovo binding sites. (B) y w male containing the P[y^ovoEN; w⁺] 2.4 transgene shows black wing blade, black body, and black bristles. (C and E) Integration of gypsy between the wing and body enhancers results in yellow wings and black body phenotype. (D) Integration of gypsy between the body enhancer and the promoter results in yellow wings and body phenotype. (F) Integration of a second gypsy between the body enhancer and promoter, in opposite orientation to the preexisting one (C and E), results in black wings and yellow body, indicating that the wing enhancer is capable of bypassing the enhancer-blocking activity of the two insulators. (G) Integration of a second gypsy between the body enhancer and promoter, in the same orientation as the preexisting gypsy insertion (C and E), results in yellow wings and yellow body, suggesting that both enhancers, wing and body, are prevented from activating the yellow promoter by the enhancer-blocking activity of the two insulators. (H) Two insertions in opposite orientations with integration sites between the wing and body enhancers result in wild-type yellow phenotype.

To test the role of Ovo binding sites in the specification of gypsy insertion, two independent transgenic lines from each construct were selected and transgenic males homozygous for the insertion were crossed to flam females. Gypsy elements are highly transcribed in the follicle cells surrounding the developing oocytes of these females, producing virus particles that will infect the oocyte and subsequently will integrate into the germ line DNA of the offspring (Pelisson et al. 1994; Song et al. 1994, 1997). Insertions of gypsy into P[y^ovo; w⁺] transgenes can be detected in the F₂ generation by the changes in the yellow phenotype induced by the gypsy insulator after integration of gypsy into the regulatory region of the P[y^ovo; w⁺] transgene (Labrador and Corces 2001). For each experimental cross, controls using the ovo^D1 reversion assay (Mevel-Ninio et al. 1989; Prud'homme et al. 1995) were performed to ensure that gypsy was actively transposing in the flam stock (Table 1). We analyzed an average of 100 F₂ flies per each y v f mal flam/y w; P[y^ovo; w⁺] F₁ parent. The actual number of F₂ descendants counted for each cross was 95–105 and this variability was randomized by using a large number of F₁ progenitors and discarding low yield crosses equally among all samples. Statistical significance was calculated using the χ² contingency tables, assuming that the probability of gypsy insertions into transgenes was uniform across vials within each sample.

TABLE 1.

Frequency of gypsy insertions into P[y^ovo; w⁺] transgenes

Lines	Females tested	New insertionsa	Frequency (%)
ovoD1 controlsb	2134	141	6.61
P[yovoWT; w+] 2.1	214	24	11.21
P[yovoWT; w+] 1.1	181	19	10.50
P[yovoDIM; w+] 2.1	132	0	0
P[yovoDIM; w+] 1.2	121	0	0
P[yovoEN; w+] 2.4	250	14	5.60
P[yovoEN; w+] 4.2	238	10	4.20
P[yovoEN; w+] 2.4 ywing	128	17	13.28

P[y^ovoEN; w⁺] 2.4 vs. P[y^ovoEN; w⁺] 2.4 y^wing χ² = 6.63; P = 0.01. P[y^ovoEN; w⁺] 4.2 vs. P[y^ovoEN; w⁺] 2.4 y^wing χ² = 8.76; P = 0.0031.

^aNumber of females that produced at least one progeny with a new gypsy insertion into a yellow transgene.

^bControls using the ovo^D1 assay were performed in parallel with experiments using transgenic lines to determine the activity of gypsy in flam flies.

Results from these experiments indicate that flies containing P[y^ovoWT; w⁺] transgenes show a high frequency of gyspy insertions into yellow, whereas flies containing P[y^ovoDIM; w⁺] transgenes did not yield a single gypsy insertion. Table 1 indicates the number of flies analyzed and the frequencies of gypsy insertions in the P[y^ovo; w⁺] transgenes from different transgenic lines. From 395 F₁ females analyzed, a total of 43 new insertions occurred into P[y^ovoWT; w⁺] trangenes (10.85%), whereas analysis of 253 F₁ P[y^ovoDIM; w⁺] females did not reveal any new gypsy insertion (Table 1). Therefore, these results suggest that removal of all Ovo binding sites from the ovo^WT sequence eliminates the ability of gypsy to specifically integrate into P[y^ovo; w⁺] transgenes, suggesting that the frequency and specificity of gypsy insertions into the regulatory region of the ovo gene depends on the presence of the Ovo protein.

A synthetic DNA sequence containing an array of eight Ovo binding sites flanked by random intervening sequences actively promotes gypsy insertion into a yellow transgene:

Although the experiments described above strongly suggest a requirement of Ovo proteins to specifically recruit gypsy insertions, we could not discard a role of the flanking sequences in this process. It is possible that these sequences could recruit additional proteins with an active role in targeting gypsy for integration. Therefore, we asked whether the presence of Ovo binding sites flanked by random DNA sequences in a synthetic sequence will also specifically induce a high rate of gypsy insertion. We recreated the Ovo binding sites previously identified in the ovo regulatory region (Lu et al. 1998) to generate a 314-bp synthetic DNA sequence, named ovo^EN (Figure 1D). The ovo^EN sequence has eight Ovo binding sites identical to those found in wild-type ovo sequences (Figure 1A) except that they are flanked by random DNA sequences (Figure 1D). The ovo^EN sequence was inserted into the NotI site of the same Casper-yellow transformation vector used in the previous experiments (Figure 2A) and the resulting Casper-yellow^ovoEN DNA was microinjected into y w ; Δ2-3 St/TM6 Drosophila embryos.

Two transgenic stocks carrying yellow^ovoEN transgenes (P[y^ovoEN; w⁺]) were subsequently analyzed to test the ability of the ovo^EN sequence to specifically promote new gypsy integrations (Table 1). The phenotype of transgenic flies carrying P[y^ovoEN; w⁺] transgenes was indistinguishable from that of wild type, indicating that the synthetic ovo^EN sequence has no effect on the expression of the yellow transgene (Figure 2B). Gypsy insertions into the P[y^ovoEN; w⁺] transgene were induced after activation of gypsy transposition in crosses with flam mutant females. Insertions were identified by their effect on the yellow phenotype in adult flies. Gypsy insertions into the P[y^ovoEN; w⁺] transgene were identified in 24 of 488 (4.92%) analyzed females, further supporting the hypothesis that the presence of Ovo proteins is responsible for recruiting new gypsy insertions into the transgene. The lower frequency of gypsy insertion into ovo^EN vs. ovo^WT sequences may perhaps reflect a role of the sequences flanking the Ovo binding sites in the binding affinity of the Ovo proteins.

Gypsy insertions also occur into sequences flanking Ovo binding sites:

Analysis of the phenotypes emerging from new gypsy insertions into P[y^ovo; w⁺] transgenes unexpectedly revealed that some insertions did not take place into ovo sequences. For example, flies from the strain P[y^ovoEN; w⁺] 2.4 y^wing have wild-type black pigmentation in the body but yellow wing blades. This phenotype could be explained if a gypsy insertion occurred between the wing and body enhancers, with the gypsy insulator preventing the wing enhancer from activating transcription in wing tissues, but allowing the body enhancer to activate transcription in the body cuticle (Figure 2E). The insertion of gypsy between the wing and body enhancers would be unexpected, since the ovo^EN sequences are placed between the body enhancer and the promoter of yellow. The location of the gypsy insertion site in the P[y^ovoEN; w⁺] 2.4 y^wing transgene was determined using PCR. Gypsy-P1 and -P3 primers (see materials and methods) were selected from 5′ and 3′ non-LTR gypsy sequences, respectively, and primers yellow-Pn were selected from sequences present in the 5′ regulatory region of the yellow gene. PCR products were obtained using sets of two primers such that one would match yellow gene sequences and the second would match gypsy sequences. Using this procedure, each combination of primers successfully amplifying a specific DNA fragment provides information on the orientation and the approximate insertion site of the gypsy insertion in the P[y^ovo; w⁺] transgene. The precise location of gypsy integration sites was then obtained by sequencing of the amplified DNA using gypsy LTR-P3 or gypsy LTR-P1 primers. The gypsy insertion in the P[y^ovoEN; w⁺] 2.4 y^wing transgene occurred between the wing and body enhancers, as predicted by the phenotype (Figure 2E), at nucleotide 1111 in the upstream regulatory sequences of the yellow gene and outside of the ovo^EN sequences (Geyer and Corces 1987).

Results from previous work had suggested that gypsy insertions occur only within the DNA fragment containing the binding sites for the Ovo protein, indicating that somehow targeting of the gypy preintegration complex and integration into DNA were part of the same process (Dej et al. 1998; Labrador and Corces 2001). Interestingly, integration of gypsy in the P[y^ovoEN; w⁺] 2.4 y^wing transgene in this case took place 869 bp away from the insertion site of the ovo^EN sequences, suggesting that targeting of gypsy preintegration complexes and gypsy integration itself might take place in two independent events during the process of gypsy integration into genomic DNA. To obtain additional insights into the mechanisms determining the specificity of gypsy integration, we sequenced the insertion sites of a total of 85 insertion events (Figure 3 and Table 2). Results confirmed the previous observation and showed that insertions into P[y^ovo; w⁺] transgenes occurred up to 869 bp upstream and up to 441 bp downstream of the actual ovo sequence (Figure 3). Interestingly, whereas insertions in y^ovoWT occur both into the ovo sequences or into the flanking yellow sequences, insertions induced by the ovo^EN sequence fall mostly into flanking yellow DNA.

Figure 3.—

gypsy integration sites and orientation in the upstream regulatory region of P[y^ovo; w⁺] transgenes. A total of 85 integration sites were characterized at the sequence level (see Table 2). Filled circles correspond to integration events. Arrowheads on top of circles indicate orientation of the insertions with the arrowhead pointing toward the 3′ end of gypsy. Ovo sequences were cloned at nucleotide 1980 in the upstream regulatory sequence of yellow. Integration sites within the ovo sequences are colored. Boldface segments below indicate distances in nucleotides between the first and the last insertion before and after ovo DNA sequences. Numbers in brackets indicate distance between adjacent integration sites. White boxes illustrate the position of wing (W) and body (B) enhancers. The yellow promoter is indicated by a broken arrow. Drawing is not to scale.

TABLE 2.

Target site sequences and integration sites into P[y^ovo; w⁺] transgenes

Target site duplication	Integration site	No. of integrations
TATATGa	1835	21
TGTATGa	2035	15
TATATGa	2031	14
TATAAA	491 in ovoWT	6
TGTATAa	489 in ovoWT	4
TATACAa	2170	3
TATATC	1838	2
TGTATAa	2029	2
TATACT	1842	2
TGTTTG	1430	2
CTGTAT	252 in ovoEN	2
TATAAA	1111	1
TATTAA	2374	1
TATAAC	2220	1
CGTTAT	2210	1
TGTAGA	1676	1
TGTATT	2256	1
TATACAa	2180	1
ACGTTA	2209	1
TATGCT	2036	1
GTGTTT	1429	1
GCACGA	2421	1
TGCACAa	454 in ovoWT	1

^aTarget site duplications conform to the preferred consensus YRYRYR for gypsy integration.

In addition to the Ovo binding sites, there may be other sequences present in ovo^WT but not in ovo^EN that might be responsible for the observed difference in the distribution of gypsy integration sites when transgenes carrying these two sequences are compared. For example, the Ovo-B promoter is present in the ovo^WT sequence and is missing in the ovo^EN sequence. Although there is no experimental evidence to support this explanation, it is possible that the ability to activate transcription by the Ovo-B promoter in the ovo^WT sequence could involve local changes in chromatin structure in a manner that favors gypsy integration. These data suggest that, without other regulatory or promoter regions, Ovo proteins may recruit gypsy preintegration complexes to the vicinity of Ovo binding sites. The specific integration site may be selected later by the gypsy preintegration complexes within a few hundred nucleotides from the Ovo binding sites by a still unknown mechanism.

It has been shown previously that gypsy has a preference for integration into the sequences TATATA and TACATA, later defined as YRYRYR, in which Y is a pyrimidine and R is a purine (Dej et al. 1998). Given that integration sites into this sequence are significantly more abundant than in any other sequence, our results confirm the preference of gypsy for integration into YRYRYR sequences, with 61 of the 85 sequences analyzed conforming to the consensus (Table 2). However, a number of other sequences are also capable of functioning as target sites for gypsy integration, suggesting that, although important, DNA sequence is not the only factor determining integration site selection. Interestingly, the pattern of insertion sites is significantly different at each side of the ovo sequences (Figure 3). Clusters of insertions into sequences distal to the promoter occur at intervals of 145, 159, 246, and 319 nucleotides, suggesting that these intervals may correspond to nucleosomal DNA that is protected and not accessible to the gypsy integrase. On the other hand, insertions into sequences proximal to the yellow promoter occur in a more arbitrary fashion. A large cluster of multiple integrations is observed at 55 bp from the Ovo binding sequences and small clusters or single insertion sites are found at intervals of only 145, 30, 46, 118, and 47 nucleotides. Except for the 145-bp interval between insertions at nucleotides 2036 and 2170, which may suggest the presence of a nucleosome, the remaining insertion sites are randomly scattered, suggesting that positioned nucleosomes may be absent from this region at the time of retroviral integration. Therefore, our results suggest that in addition to the recruitment of gypsy preintegration complexes by Ovo proteins other factors such as the specific DNA sequence of the insertion site, nucleosomal arrangement, or other DNA binding proteins may also play an important role in the final determination of the integration site.

The presence of a gypsy provirus results in high frequency of secondary gypsy insertions:

Interactions between Ovo proteins and components of the gypsy preintegration complex are likely responsible for targeting gypsy insertions to the vicinity of Ovo binding sites. Since gypsy insulator proteins mediate interactions between individual insulators to create chromatin loops (Gerasimova et al. 2000; Cai and Shen 2001; Muravyova et al. 2001; Byrd and Corces 2003; Kuhn et al. 2003), it is possible that the presence of a gypsy provirus in a P[y^ovo; w⁺] transgene will increase the frequency of secondary gypsy insertions. To test this possibility, we compared the insertion rates into yellow-ovo sequences present in transgenes containing or lacking a copy of a gypsy provirus. We used the transgenic line P[y ^ovoEN; w⁺] 2.4 y^wing, which contains a gypsy insertion in the yellow transgene regulatory region between the wing and body enhancers and displays wild-type black pigmentation in the body but yellow wing blades. Starting with this transgenic line we then induced gypsy mobilization using the same flam assay described above. We set up 128 F₁ y w f ml flam/P[y^ovoEN; w⁺] 2.4 y^wing single female crosses and, as in previous assays, we analyzed an average of 100 offspring from each single female cross (range 95–105), screening for changes in the yellow phenotype (Table 1).

Transgenes with new insertions resulted in three new phenotypes when starting from flies with an initial y^wing phenotype (Figure 2): flies with a y² phenotype (yellow wing blades, yellow body, and black bristles), flies with wild-type phenotype (black wing blades, black body, and black bristles), and y^body flies (black wing blades, yellow body, and black bristles). We obtained 17 such events with a final frequency of 13.28%. A Fisher exact test indicates that the number of new gypsy insertions in P[y^ovoEN; w⁺] 2.4 y^wing is significantly higher than the same number in the parental transgenic stock P[y^ovoEN; w⁺] 2.4 (Table 1). Since the initial P[y^ovoEN; w⁺] transgene is inserted in the same genomic location in both stocks, the observed differences are not due to chromosomal position effects. The results suggest that the higher frequency of second gypsy integrations may be due to interactions between the preexisting gypsy element in the genomic copy of the transgene and proteins present in the gypsy preintegration complex.

Insulator function depends on the orientation of gypsy provirus sequences:

To confirm that the new phenotypes described in the previous section are due to new gypsy integration events, we mapped the integration sites and orientation of the putative new insertions. All new phenotypes contain a single additional new gypsy insertion. Interestingly, each phenotype correlates either with the orientation or with a specific integration site of the second gypsy insertion (Figure 2). We obtained a total of 10 independent new insertions resulting in a y^body phenotype and all of them originated from a second insertion, in an orientation opposite to the first one, between the body cuticle enhancer and the yellow promoter (Figure 2F). In these flies the body cuticle enhancer is flanked by two gypsy insertions in opposite orientations and they display wild-type coloration of all tissues except for the body cuticle. We obtained only one event of a y⁺ phenotype, which originated from a second gypsy insertion between the body cuticle enhancer and the wing enhancer but integrated in an orientation opposite to the first one (Figure 2H). Finally, we obtained 6 independent new insertions resulting in y² phenotypes (Figure 2G). Interestingly, all 6 independent new lines carried a second gypsy insertion between the body enhancer and the promoter, the same arrangement observed in y^body strains.

The only difference between flies with y^body phenotypes and y² phenotypes is the relative orientation of the two gypsy insertions. In flies displaying a y^body phenotype the two copies of gypsy are in opposite orientations (Figure 2F) whereas in flies with a y² phenotype the two copies of gypsy are in the same orientation (Figure 2G). These results show that two copies of the gypsy provirus inserted in the regulatory region of the yellow gene have the same properties as two paired insulators in experimental transgenes (Cai and Shen 2001; Muravyova et al. 2001; Kuhn et al. 2003). The presence of a single gypsy provirus prevents the wing and body cuticle enhancers from activating the yellow promoter due to the presence of the gypsy insulator (Figure 2D). However when a second insertion occurs in the opposite orientation, distal enhancers can bypass both gypsy insertions and their respective insulators (Figure 2, F and H).

Although the specific mechanism explaining this property is still unknown, current models suggest that this phenomenon is the result of molecular interactions between the two adjacent insulators and that these interactions loop out the intervening DNA. Enhancers outside the resulting chromatin loop are capable of activating transcription of distal promoters, but enhancers trapped within the chromatin loop are blocked by the interacting insulators. Our results confirm this model when the two copies of gypsy are in opposite orientation and suggest that such interactions can occur at distances much longer than the distances so far reported (usually a few hundred nucleotides). When the wing enhancer bypasses two gypsy insertions in opposite orientation as in Figure 2F and once the distance between the two insertion sites and the size of the two gypsy sequences (7.4 kb) is taken into account, the distance between the two gypsy insulators is >14 kb. Interestingly, these interactions between insulators seem to be abrogated when the two copies of the gypsy provirus are in the same orientation. Since the only genetic difference between gypsy insertion arrangements originating the y^body and y² phenotypes is the relative orientation of the two gypsy elements, the results suggest that the relative orientation of gypsy sequences may affect the ability of the two insulators to pair and thus influence the enhancer blocking properties of gypsy insulators.

DISCUSSION

The gypsy retrovirus of Drosophila may offer valuable clues as to how retroviruses develop strategies to specifically select integration sites into the genome. Results shown here suggest that interactions between the gypsy preintegration complex and, most likely, Ovo proteins are sufficient to promote site-specific integration of gypsy into the ovo locus of Drosophila. Alteration of Ovo binding sites from a wild-type ovo gene fragment abolishes the ability of gypsy to specifically integrate into adjacent sequences. In addition, a synthetic DNA sequence carrying eight Ovo binding sites flanked by random DNA sequences is sufficient to function as a highly specific target for integration of the gypsy retrovirus. Although direct interactions between Ovo proteins and the gypsy preintegration complex have not been substantiated, the data point to a mechanism by which Ovo proteins may tether the gypsy preintegration complexes to their binding sites.

It is tempting to speculate that the gypsy integrase may actually interact with the Ovo proteins and that such interaction may target integration to genome sites enriched in Ovo. Evidence demonstrating tethering of integrases as a mechanism capable of targeting retroviruses to specific DNA binding sites has been shown in experiments using fusion proteins in which the DNA binding domain of phage λ-repressor was fused to the integrase of the HIV retrovirus and successfully showed preferential integration into target DNA near λ-repressor-binding sites (Bushman 1994). Similarly, experiments with yeast retrotransposons have shown that such interactions may occur between the retrotransposon integrase and proteins that target the integration to their cognate chromosomal DNA binding sites (Gai and Voytas 1998; Xie et al. 2001; Sandmeyer 2003; Zhu et al. 2003; Gao et al. 2008).

It has been reasoned that retrotransposons in small genomes such as that of yeast may develop tethering mechanisms of site-specific integration by stimulating interactions between the integrase encoded by the retrotransposon and endogenous proteins, thus minimizing the chances of deleterious mutations induced by retrotransposon integration events. In larger genomes such mechanisms appear infrequently, probably due to the lack of selective pressure from the host genome. In humans for example, sequences related to interspersed retroviruses occupy >50% of the genome and only relatively low frequencies of integration events in specific target spots have been reported (Farnet and Bushman 1997; Suzuki et al. 2004; Bushman et al. 2005; Shun et al. 2007). In Drosophila only a number of specialized non-LTR retrotransposons have acquired specificity of integration associated with specialized chromosomal regions such as telomeres or ribosomal DNA (Jakubczak et al. 1992; Casacuberta and Pardue 2005). Nonvertebrate retroviruses such as ZAM, Idefix, and gypsy appear to be an exception when compared with their vertebrate counterparts, since some degree of sequence specificity and targeted site integration has been described in all three (Mevel-Ninio et al. 1989; Desset et al. 1999; Leblanc et al. 1999; Conte et al. 2000). The high rate of insertion of gypsy into the ovo gene and the role apparently played by the Ovo proteins provide an excellent tool to study the integration mechanism and how retroviruses may acquire integration site specificity in vivo.

Interestingly, even though Ovo proteins appear to have a role in targeting gypsy to their binding sites, gypsy insertion sites do not necessarily occur into the Ovo binding sequences themselves. From a total of 85 sequenced insertion sites, only 13 (15%) occurred into the DNA fragment containing Ovo binding sites; the remaining integration sites fall within an interval of >1300 bp flanking the DNA containing the Ovo binding sites. The analysis of insertion sites suggests that the targeting and the integration mechanisms are uncoupled, with the precise integration sites distributed in a nonrandom manner. Results shown here confirm previous observations suggesting that gypsy has a preference for integration into YRYRYR sequences. However, a variety of other sequences appear to be able to function as integration sites. This disparity makes it difficult to draw a clear conclusion as to what is the mechanism ultimately involved in selecting target sites; however, it is tempting to speculate a role for nucleosome positioning as one of the factors determining the selection of insertion sites by the gypsy retrovirus.

Several indirect lines of evidence suggest such a role. For example, integration frequencies are significantly higher between the promoter of the yellow gene and the Ovo binding sites, indicating a preference that probably reflects a difference in chromatin structure. Forty-two independent integration events occurred into a fragment of 441 bp located between Ovo binding sequences and the promoter of the yellow gene, whereas only 30 integrations occurred in a 869-bp DNA fragment located distal to the promoter and upstream of the Ovo binding sites. This asymmetry does not appear to reflect sequence differences or viability effects, suggesting an epigenetic basis for integration site selection. In addition, insertion sites found distally to the yellow promoter and upstream of the Ovo binding DNA sequences appear to be spatially distributed in four intervals >140 bp, whereas insertions proximal to the promoter are distributed in a random manner, with only one large gap of 145 bp and apparently lacking meaningful spacing intervals. A possible interpretation of these results is that gypsy integration preferentially occurs at specific points of either the nucleosome or the linker DNA. In the distal interval, one spacing >140 bp could reflect nucleosome positioning, whereas nucleosomes may be absent or not positioned in the promoter proximal region of the transgene.

Results presented here also suggest that a preexisting gypsy insertion significantly increases the chances of new gypsy insertions into adjacent sequences by more than twofold. We speculate that such enhancement of insertion frequency might be the result of a tethering mechanism mediated by protein–protein interactions between the gypsy element located in the chromosomal DNA and the gypsy preintegration complex during the normal process by which a new copy of the retrovirus is inserted into the chromosome. Since two copies of the gypsy insulator have been shown to be able to interact with each other (Gerasimova et al. 2000; Byrd and Corces 2003), it is tempting to speculate that interactions between gypsy insulators are responsible for the increased frequency of gypsy insertions. Nevertheless, we cannot rule out the possibility that gypsy sequences other than the insulator or proteins associated with the gypsy element itself are responsible for the observed interactions. If the high frequency of secondary insertions is due to interactions between gypsy insulator proteins present in the provirus and in the preintegration complex, the results would lend support to proposed models suggesting that individual insulators located in different regions of a chromosome can interact to form chromatin loops.

The analysis of phenotypes resulting from double insertions allows further elaboration of this model and offers additional insights into the mechanisms by which insulators affect enhancer–promoter interactions. For example, we have shown that a wing enhancer distal to two adjacent gypsy insertions is capable of bypassing the activity of the two insulators when the two copies of the gypsy provirus are inserted in opposite orientation. These results demonstrate that interactions such as the ones determined genetically in transgenes, involving pairs of 400-bp gypsy insulators (Cai and Shen 2001; Muravyova et al. 2001; Kuhn et al. 2003), also occur between pairs of gypsy insulators embedded in the gypsy provirus and suggest that establishing such interactions is part of the normal life cycle strategy used by the retrovirus. Interestingly, when two gypsy insertions occur in the same orientation, distal enhancers are unable to bypass the two insulators and are blocked from activating the promoter, contrary to what it has been observed with direct repeats of insulator sequences (Cai and Shen 2001; Muravyova et al. 2001; Kuhn et al. 2003). The main difference between the two sets of experiments is the presence of additional DNA sequences in the gypsy provirus. These sequences may be able to form a stem-loop structure when the two copies of gypsy are arranged in opposite, but not when they are in the same, orientation. A similar role has been suggested for the relative orientation of insulator sequences between interacting Mcp insulators (Kyrchanova et al. 2007). The stem-loop structure would allow interactions between insulator proteins present in the two copies of gypsy with opposite orientations but a direct tandem arrangement of the two copies of the provirus would preclude such interactions. These observations support the hypothesis that interactions between paired insulators are required to bypass insulator function and allow enhancer–promoter communication.