Enhancer-Promoter Communication Is Regulated by Insulator Pairing in a Drosophila Model Bigenic Locus

Abstract

The complexity of regulatory systems in higher eukaryotes, featuring many distantly located enhancers that nonetheless properly activate the target promoters, has prompted the hypothesis that the action of enhancers should be restricted by insulators. Continuing our research on the functional role of insulators and the consequences of their interaction in Drosophila, we studied the interplay of different Su(Hw)-dependent Drosophila insulators. The set of transgenic constructs comprised two consecutive genes (yellow and white) with their enhancers and insulator elements differently arranged in between and/or around the gene(s). All insulators were found to interact in twin or mixed tandems, demonstrating the bypass phenomenon. However, insulator pairing around a gene did not always improve its isolation from an outside enhancer. On the other hand, merely two insulator elements (identical or different) in appropriate positions can permit the expression of one gene but not the gene next to it or, conversely, largely block the transcription of the first gene, while allowing full enhancement of the second, or make them behave similarly. Thus, the results of this study support the model that loop formation by insulators is an essential component of insulator action on a positive and negative regulation of an enhancer-promoter communication.

The complexity of regulatory systems in higher eukaryotes, featuring many distantly located enhancers that nonetheless properly activate the target promoters, has prompted the hypothesis that the action of enhancers should be restricted by elements called insulators. Generally, insulators are defined by two properties: (i) enhancer-blocking activity, preventing communication between an enhancer and a promoter separated by the insulator, and (ii) boundary function (barrier activity), preventing repressive chromatin spreading (4, 6, 11, 12, 16, 18, 19, 32, 52, 53, 54, 55, 56). Their function as barriers to repressive chromatin is relatively easy to understand (19, 52, 53, 54) and will not be considered here. As to the other function, implied to restrict the promiscuity of transcription enhancers and known as enhancer blocking (6, 11, 19, 32), the actual progress is not very impressive.

The Drosophila genome contains many sequences with insulator properties (see references 1, 6, 28, and 32). The best-studied insulator in Drosophila is found in the gypsy retrotransposon (mdg4) (36, 50). It contains 12 degenerate repeats of the binding motif for the zinc finger protein Su(Hw), which is indispensable for its function (22, 27). Among numerous potential Su(Hw) binding sites dispersed throughout the wild-type genome, rarely three or more motifs occur in reasonable proximity to each other (1, 39, 42). However, the 1A2 insulator downstream of the yellow gene, with only two Su(Hw) binding sites, functions as an effective insulator (24, 38).

Upon critical analysis, there are very few things we know more or less certainly. (i) The stimulation of reporter gene (e.g., yellow) expression by a corresponding enhancer can be prevented if a single insulator is interposed in the DNA sequence between the enhancer and the gene promoter, whereas in any other place the insulator is ineffective (the definitive property known as position dependence) (8, 17, 20, 25, 26, 30, 31, 44, 47, 57-59). Insulators may vary in apparent blocking strength (10, 47, 26, 46, 57). (ii) At least some insulators can interact with each other (9, 13, 14, 15, 25, 37, 43). For instance, when two gypsy insulators in tandem are interposed in the same yellow enhancer-reporter system, they both appear to lose their enhancer-blocking ability (9, 37). This phenomenon is known as insulator bypass, mutual neutralization, or cancellation. Some other insulators in twin or mixed tandems, however, are reported not to allow bypass (33, 34). (iii) Insulator pairing can give rise to a chromatin loop (3, 7).

Again, even these issues should be considered with caution. Although we know that an insulator impedes enhancer-promoter communication, we do not know how this occurs. Much of the research on insulators in Drosophila has been done with the white gene as reporter; these results and ensuing conclusions now need reconsideration, because all such systems contained the white-abutting resident insulator (Wari) (13). This element can interact equally well with another Wari and with unrelated insulators (gypsy or 1A2), thereby significantly altering the results of insulator assays. The problem with “loop domains” is not their existence but rather their functional significance; e.g., two or three copies of the gypsy insulator can interact over considerable distances, but partitioning of the expression construct by closed DNA loops supposed to arise from such interactions does not itself always ensure enhancer blocking (45).

In the present study, we (i) estimated the expression of two Drosophila genes in a large set of constructs with various arrangements of the same or different Su(Hw)-dependent insulators and (ii) analyzed the aggregate data to discern general patterns. The diverse changes in the enhancer-promoter communication that can be caused by insulator interactions are systematized, and some additional features and fallacies of simplistic “loop domain” models are revealed. We also show how versatile expression control in a model bigenic locus can be achieved by properly placing the insulator elements. In this respect, we corroborate the basic idea that gave rise to the insulator research.

MATERIALS AND METHODS

Plasmid construction.

The 3-kb SalI-BamHI fragment containing the yellow regulatory region (yr) with the body and wing enhancers (fragment −2873 to −1266 bp relative to the transcription start site) (21, 35) was subcloned into the pGEM7 plasmid digested with BamHI and XhoI. The white eye enhancer (Ee, fragment −1465 to −1084 bp relative to the white transcription start site) (40) was then inserted at −1868 from the yellow transcription start site (yr-Ee).

The 5-kb BamHI-BglII fragment containing the yellow coding region (yc) was subcloned into CaSpeR2 (yc-C2). The CaSpeR2 vector contains the mini-white gene and defective inverted repeats of P element (41). The pCaSpeRΔ700 vector without insulator located at the 3′ side of the mini-white gene (Wari insulator) was constructed from pCaSpew15(+RI) plasmid as described in reference 13.

The 340-bp fragment containing the Su(Hw)-binding region (Gy) was PCR amplified from the gypsy retrotransposon. The 454-bp sequence of the 1A2 insulator (1A2) was PCR amplified with pr-1 (5′-GGAGTACTACTACCAGGC-3′) and pr-2 (5′-CAAGAACATTTCCGATATG-3′) primers. The plasmid containing four reiterated Su(Hw)-binding sites (S⁴) was provided by E. Savitskaya. These fragments were inserted between two lox or two frt sites.

En1A2_-343YW and EnY(1A2)W.

The 1A2 insulator was inserted in yr-Ee cleaved with KpnI at position −343 from the yellow transcription start site (yr-Ee-1A2). The lox(1A2) fragment was inserted in CΔ700 digested with EcoRI at position −400 relative to the white transcription start site [CΔ700-lox(1A2)]. The yr-Ee-1A2 piece was inserted into the yc-CΔ700 plasmid. The yr fragment was inserted into the CΔ700-lox(1A2) plasmid.

En(S⁴)YW, En(S⁴_-343YW, and EnY(S⁴)W.

The lox(S⁴) was inserted in either yr-Ee cleaved with Eco47III at −893 or KpnI at −343 from the yellow transcription start site [yr-Ee-lox(S⁴) or yr-Ee-lox(S⁴_-343)] or in CΔ700 digested with EcoRI at −400 relative to the white transcription start site [CΔ700-lox(S⁴)]. The yr-Ee-lox(S⁴) or yr-Ee-lox(S⁴_-343) piece was inserted into the yc-CΔ700 plasmid. The yr fragment was inserted into the CΔ700-lox(S⁴) plasmid.

En(1A2)YW(1A2), En(1A2)YW(S⁴), and En(1A2)YW(Gy).

The frt(1A2) piece was inserted in yr-Ee cleaved with Eco47III at position −893 from the yellow transcription start site [yr-Ee-frt(1A2)]. The lox(1A2), lox(S⁴), and lox(Gy) pieces were inserted in CΔ700 digested with EcoRI at position −400 relative to the white transcription start site [CΔ700-lox(1A2) and CΔ700-lox(S⁴)]. The yc was subcloned into CΔ700-lox(1A2) and CΔ700-lox(S⁴), giving yc-CΔ700-lox(1A2) and yc-CΔ700-lox(S⁴). The yr-Ee-frt(1A2) set was subcloned in yc-CΔ700-lox(1A2) and yc-CΔ700-lox(S⁴) digested with XbaI and BamHI.

En1A2(1A2)Y(1A2)W, EnS⁴(S⁴)Y(S⁴)W, EnGy(Gy)YGyW, En1A2(Gy)Y(Gy)W, and En1A2(S⁴)Y(S⁴)W.

The 1A2, Gy, and S⁴ sequences were inserted in yr-Ee digested with Eco47III at position −893 from the yellow transcription start site (yr-Ee-1A2, yr-Ee-Gy, and yr-Ee-S⁴). The frt(1A2), frt(S⁴), and frt(Gy) pieces were inserted in the yr-Ee-1A2, yr-Ee-Gy, or yr-Ee-S⁴ digested with KpnI at position −343 from the yellow transcription start site [yr-Ee-1A2-frt(1A2), yr-Ee-Gy-frt(Gy), yr-Ee-S⁴-frt(S⁴), yr-Ee-1A2-frt(Gy), and yr-Ee-1A2-frt(S⁴)]. The BamHI-Eco47III fragment from yc-C2 was subcloned into CΔ700 (yc-CΔ700). The lox(1A2), lox(S⁴), and lox(Gy) pieces were inserted into yc-CΔ700 digested with BglII at position +4964 relative to the yellow transcription start site [yc-lox(1A2)-CΔ700, yc-lox(S⁴)-CΔ700, and yc-lox(Gy)-CΔ700]. The yr-Ee-1A2-frt(1A2), yr-Ee-S⁴-frt(S⁴), yr-Ee-1A2-frt(Gy), and yr-Ee-1A2-frt(S⁴) sets were inserted into yc-lox(1A2)-CΔ700, yc-lox(S⁴)-CΔ700, or yc-lox(Gy)-CΔ700 digested with XbaI and BamHI.

En1A2(1A2)WY.

The fragment with yellow regulatory and coding regions (yr-yc) was inserted into CΔ700 digested with EcoRI (CΔ700-yr-yc). The yr-Ee-1A2-frt(1A2) was inserted into CΔ700-yr-yc digested with XbaI.

En1A2(1A2) ΔYW.

The yr-Ee-1A2 was inserted into yc-CΔ700 digested with XbaI and BamHI. The frt(1A2) piece was inserted in the yr-Ee-1A2-yc-CΔ700 digested with KpnI at positions −343 and at +792 from the yellow transcription start site [yr-Ee-1A2-frt(1A2)-yc-CΔ700].

Generation of transgenic lines and genetic crosses.

All flies were maintained at 25°C on the standard yeast medium. The construct together with P25.7wc, a P element having defective inverted repeats used as a transposase source, was injected into y ac w¹¹¹⁸ preblastoderm embryos as described previously (29). The resulting flies were crossed with y ac w¹¹¹⁸ flies, and the transgenic progeny were identified by the color of their eyes and cuticle structures. The chromosome localization of various transgene inserts was determined by crossing the transformants with the y ac w¹¹¹⁸ balancer stock carrying dominant markers, In(2RL),CyO for chromosome 2 and In(3LR)TM3,Sb for chromosome 3. The transformed lines were tested for transposon integrity and copy number by Southern blot hybridization. Only single-copy transformants were taken into the study.

The lines with excisions were obtained by crossing the transposon-bearing flies with the Flp (w¹¹¹⁸; S² CyO, hsFLP, ISA/Sco; +) or Cre (y¹, wⁱ; CyO, P[w+,cre]/Sco; +) recombinase-expressing lines. The Cre recombinase induces 100% excisions in the next generation. A high level of FLP recombinase was produced by heat shock of late embryos and second- or third-instar larvae at 37°C for 2 h. The excisions generated by Flp or Cre recombinases were confirmed by PCR analysis with pairs of primers flanking the following insertion sites: at position −893 relative to yellow (5′-ATCCAGTTGATTTTCAGGGACCA-3′ and 5′-TTGGCAGGTGATTTTGAGCATAC3′), at position −343 relative to yellow (5′-TAGATCGTCAAATAAAGTCCCTA-3′ and 5′-GTTTGGTATGATTTTTGGCCTTC-3′), between yellow and white (5′-TTTTCTTGAGCGGAAAAAGCGGA-3′ and 5′-ATCTACATTCTCCAAAAAAGGGT-3′), and at the 3′ end of white (5′-CTAATATCCTGCGCCAGCTCCT-3′ and 5′-ACGTGACTGTGCGTTAGGTCCTGT-3′).

To determine the levels of yellow and white expressions, we visually estimated the degree of pigmentation in the abdominal cuticle and wing blades (yellow) and in the eyes (white) of 3- to 5-day-old males developing at 25°C. For yellow, a five-grade scale was used, with grade 1 corresponding to the total loss of yellow expression and grade 5 corresponding to wild-type pigmentation. Identical data were obtained for the wing and body pigmentation in all experiments. On the nine-grade scale for white, bright red (R) and white (W) eyes corresponded to the wild type and the total loss of white expression, respectively. Intermediate levels of eye pigmentation in the order of decreasing gene expression were brownish red (BrR), brown (Br), dark orange (dOr), orange (Or), dark yellow (dY), yellow (Y), and pale yellow (pY). The pigmentation scores were independently determined by two investigators, who examined 30 to 50 male flies from each of the two independent crosses for every transgenic line. These scores (every unit representing one line) were entered into the corresponding table and used to assess changes in gene expression.

RESULTS AND DISCUSSION

Model system.

Since insulators (enhancer blockers) are supposed to take part in differential control of gene expression in vivo, our purpose in the present study was not just to find out whether different insulator elements can interact with each other but also to check whether any manifestations of such differential control could be observed in a simple model system. Our model comprised two consecutive genes transcribed in the same direction, the cognate enhancers grouped upstream, and insulator elements in different arrangements in between and/or around the gene(s).

As the test genes, we chose yellow and white, which have been extensively used in insulator studies (13, 14, 22, 24, 25, 33, 37, 38, 39, 42, 43, 44, 45, 46). Note, however, that we initially removed the Wari-containing sequence 3′ adjacent to mini-white in the standard plasmids (13). yellow is responsible for dark pigmentation of the larval and adult cuticle and its derivatives. Two upstream enhancers stimulate its expression in the body cuticle and wing blades (21, 35). The white gene, required for eye pigmentation, is also controlled by an upstream eye enhancer (40). In all constructs, the eye enhancer was inserted between the wing and body enhancers (W-E-B, collectively designated En; Fig. 1).

FIG. 1.

Schemes of transgenic constructs used in the present study. Schemes of the gypsy (Gy), S⁴, and 1A2 insulators are presented in the top of figure. Small boxes represent Su(Hw)-binding sites. The third site (filled box) in Gy was used to make S⁴. In construct schemes (not to scale) the yellow and white genes are shown as gray and white rectangles, respectively; the arrows indicate the direction of transcription (empty box in scheme G means deleted promoter). The wing, eye, and body enhancers are represented by contiguous W-E-B squares shaded according to target genes. The 1A2, Gy, and S⁴ insulators are shown as black rectangles, ovals, and circles, respectively. Downward arrows indicate cleavage sites for Cre or Flp recombinase; in construct names, the corresponding excisable elements are in parentheses.

The natural gypsy insulator (Gy) consists of 12 potential binding sites for the Su(Hw) protein, appearing as degenerate 12-bp cores connected by variable AT-rich sequences (48, 51). The S⁴ element was generated by tetramerization of a 30-bp oligonucleotide corresponding to the third Su(Hw) binding site and its flanks in the gypsy insulator. The four reiterated Su(Hw) binding sites were reported to function as a strong Su(Hw)-dependent insulator (48). The 1A2 insulator has two near-gypsy-consensus Su(Hw)-binding sites essential for its activity (24, 38). These elements are schematically shown in the top of Fig. 1.

The constructs tested in the present study contained one to three copies of 1A2, alone or combined with S⁴ or Gy, or only S⁴ or Gy in three copies (Fig. 1). Parentheses in construct designations and short downward arrows in the schemes indicate the elements flanked by lox or frt sites for in vivo excision by crossing (23, 49), as outlined in Materials and Methods; such excisions are denoted by “(Δ)” in the primary (expression) data (Fig. 2 to 6).

FIG. 2.

Comparison of the enhancer blocking activities of 1A2 and S4 in different positions relative to the enhancers and promoters of the yellow and white genes. Schemes of transgenic constructs are shown on top of each part of the table. The unified ranks for the levels of yellow and white activation by the enhancers are indicated by “++” (strong activation), “+” (weak activation), and “−” (no activation) in rectangles corresponding to the genes. The “yellow” column shows the numbers of transgenic lines with the yellow pigmentation levels in the abdominal cuticle (reflecting the activity of the body enhancer); in most of the lines, the pigmentation levels in wing blades (reflecting the activity of the wing enhancer) closely correlated with these scores. The “white” column shows the numbers of transgenic lines with the white pigmentation levels in eyes (reflecting the activity of the eye enhancer). “N” is the number of lines in which flies acquired a new y of w phenotype relative to the control “[0]” lines generated by deleting the corresponding insulator (1A2 or S⁴). “T” is the total number of lines examined for each particular construct. For other designations, see Fig. 1.

FIG. 6.

Effects of interaction between two insulators enclosing both genes. “N” is the number of lines in which flies acquired a new y of w phenotype relative to the control “[1]” lines carrying constructs with one insulator (1A2 or S⁴) at position −893 relative to the yellow transcription start site. For other designations, see Fig. 1 and 2.

In one series of constructs (Fig. 1A, J, and M), the frt-flanked 1A2 was inserted at position −893 relative to the yellow transcription start site (i.e., after the grouped enhancers), while a lox-flanked 1A2, S⁴, or Gy was inserted at the 3′ side of the white gene. Thus, we had two excisable insulators surrounding the two genes and thereby separated by about 10 kb.

In another series (Fig. 1H, K, and N), “invariant” 1A2 was at position −893, while two excisable 1A2, S⁴, or Gy were inserted at position −343 (frt-flanked) and position +4964 (lox-flanked) relative to the yellow transcription start site. With these constructs, we could compare the enhancer-blocking activity of the 1A2 insulator alone and in combination with assorted insulators placed between the enhancers and the yellow promoter and/or after the yellow gene.

Finally, three copies of S⁴ (second and third excisable) or Gy (second excisable) were placed at position −893, position −343, and position +4964 (Fig. 1L and O) to make a full analog of the 1A2 construct shown in Fig. 1H.

Four more 1A2-contaning constructs served as controls for insulator or gene position: a single insulator was placed close to the promoter of either gene, yellow (−343, Fig. 1B) or white (−400, i.e., +4964 relative to yellow; Fig. 1E); a 1A2(1A2) tandem was followed by yellow from which the promoter was deleted (−Y), whereby the whole sequence to the white promoter became a ∼5-kb spacer (Fig. 1G); and the 1A2(1A2) tandem was followed by white (first gene) and next by yellow (second gene) (Fig. 1I).

Finally, S⁴-containing constructs served as controls for this insulator: a single insulator was placed close to the promoter of yellow (−893 and −343, Fig. 1C and D) or white (−400, Fig. 1F).

The effects of insulator elements and their combinations and rearrangements on gene expression were deduced by comparing the phenotypic distributions of fly lines carrying the basic constructs and their derivatives produced by in vivo excision of a particular element.

It was shown that the gypsy insulator inserted at position −893 almost completely blocked the interactions of yellow and white enhancers with their promoters (22, 37, 33, 45). In the present study, the 1A2 or the S⁴ insulator at the same position caused strong but not complete blocking of both enhancers (Fig. 2). For semiquantitative evaluation, we considered that gene activation through the single Gy insulator was absent and gene activation through the 1A2 insulator or S⁴ was weak (Fig. 2). The effect of in vivo excision of a particular insulator from the construct was interpreted as opposite to the effect of its addition therein. If additional insulator(s) significantly changed the level of gene expression (for example, compared to the invariant 1A2) in more than half of transgenic lines, we assumed that the level of gene activation either increased from weak to strong or decreased to zero. The results of this evaluation are summarized in the figures. For every kind of construct (initial or derivative) present in the genome, the body and eye pigmentation grades in a sufficient number of independent transgenic lines were scored by using standard scales (as shown in reference 13). The unified ranks obtained in this way (strong [++], weak [+], or none [−]) were included in the construct schemes to indicate the levels of gene activation by the enhancers.

A single 1A2 at position −893, just as at position −343, allowed only weak stimulation; i.e., the insulator was effective at any place between the enhancer and promoter (Fig. 2A and B). When placed after yellow, 1A2 did not affect its expression, but gene white was activated only weakly (Fig. 2C). Finally, 1A2 placed on the 3′ side of white had no effect on the expression of the genes (Fig. 2D), in full conformity with the definitive position dependence of the enhancer-blocking function. The results obtained with the S⁴ insulator were similar (Fig. 2E to H).

Pairing between tandem insulators mainly affects the nearest gene.

First, we analyzed constructs with two insulators placed after the enhancers, i.e., relatively close to the yellow promoter but more than 5 kb away from the white promoter (Fig. 3).

FIG. 3.

Effects of interaction between tandemly placed insulators (selective influence on the proximal gene). (A) A diagram explaining the model (drawn roughly to scale), with the hexagon representing an insulator and the box at the beginning of the gene showing its promoter; arrows indicate the direction of transcription (here and in Fig. 4 to 6). (B to G) The constructs inserted into the fly genome correspond to those shown in Fig. 1 or derived from them by in vivo excision. “N” is the number of lines in which flies acquired a new y of w phenotype relative to the control “[1]” lines carrying constructs with one insulator (1A2 or S⁴) at position −893 relative to the yellow transcription start site. For other designations, see Fig. 1 and 2.

Let us first consider the nearest gene, yellow, since these results can be compared to available data. We observed appreciable stimulation of yellow across any tandem of Su(Hw)-related insulators tested, 1A2-(1A2), 1A2-(Gy), 1A2-(S⁴), or S⁴-(S⁴). This is the “bypass” phenomenon attributed to mutual neutralization of the enhancer-blocking properties of insulators upon their pairing. Thus, we offer experimental evidence that this effect of pairing is equally strong in twin or mixed tandems. However, it is also clear that yellow stimulation in the absence of insulators (Fig. 2) was stronger than in the presence of any tandem (Fig. 3). Thus, pairing of insulators leads to only partial neutralization of their enhancer blocking activity.

As to the white gene, its stimulation by the enhancer in constructs with a single 1A2 or S⁴ (Fig. 2A and E) was weak, as in the case of yellow. This confirms that insulators themselves are not selective with respect to enhancers (8, 22, 26, 33, 44, 47, 57). Indeed, the communication signal must be essentially the same for the genes thus regulated. Nevertheless—and this is where surprises spring up—the enhancer effect on white remained weak when yellow was activated across the tandems (Fig. 3B to E). Thus, it appears that enhancer-promoter communication through the same region is alleviated for one gene but not for the other.

This differential behavior cannot be convincingly explained in the conventional terms of insulator properties. It is possible that the yellow promoter interferes with proper communication between the enhancer and the white promoter. However, this cannot account for the observed difference, since deletion of the yellow promoter from a 1A2-(1A2) tandem construct (Fig. 3F) expectedly abolishes all yellow function but does not change anything for white.

These results may indicate that tandem pairing of insulators facilitates the enhancer action on the nearest gene mainly by drawing it still closer (reducing the distance to the promoter approximately by half, as suggested in Fig. 3A). The only condition that does change differently for the two consecutive genes upon tandem pairing is the relative enhancer-promoter distance. For yellow, this distance is reduced approximately by half, sincce the loop closure takes out the tandem spacer (>0.5 kb) and at least some of the insulator length. Combined with partial neutralization of the enhancer-blocking ability, this contributes to considerable facilitation of the enhancer-promoter communication. For white, such a contraction has little effect (if any) for the simple reason that it is an order of magnitude smaller than the distance to the promoter and, hence, the relative gain is negligible.

This idea was tested by interchanging the genes as shown in Fig. 3G: in this construct, white is proximal and yellow is distal. As expected, the responses of the particular genes were also interchanged, whereas the “block diagrams” remained exactly the same: tandem pairing improved the enhancer-promoter communication only for the proximal gene.

Thus, we can reasonably assume that the overall attenuation of the enhancer-promoter communication in general case depends both on the enhancer blocking activity of the insulator(s) and on the total distance between the enhancer and the target promoter. It seems likely that insulator tandems equally block both promoters. However, only the proximal promoter is moved in close proximity to the enhancer and “wins” the competition with insulator tandems for the enhancer.

Insulator pairing around the first gene creates a bypass to the second gene.

Next, we examined a panel of constructs where, in addition to a tandem of insulators after the enhancers (as in Fig. 3), there was also an insulator at position −400 relative to the white transcription start site, i.e., between the two genes (Fig. 4).

FIG. 4.

Effects of interaction between three insulators enclosing the yellow gene. “N” is the number of lines in which flies acquired a new y or w phenotype relative to the control “[1]” lines carrying constructs with one insulator (1A2 or S⁴) at position −893 relative to the yellow transcription start site. For other designations, see Fig. 1 and 2.

On the whole, the situation proved to be opposite to that shown in Fig. 3. white was markedly stimulated by its enhancer in all cases, although the gene and the enhancer were separated by three insulators surrounding the yellow gene. Conversely, stimulation of the enhancer-proximal yellow was absent or weak in constructs with 1A2-1A2 or 1A2-Gy “bypass tandems” (Fig. 4B and C) but could be quite distinct in other cases (Fig. 4D to F). Thus, Fig. 4 shows how an additional intergene insulator radically alters the behavior of the model bigenic locus, with its effect being the same as that of gene interchange (see Fig. 3G).

This paradox is explainable in view of the spatial arrangement of such constructs due to the interaction of three insulators they contain (Fig. 4A). With physical events being generally the same as in the “tandem bypass” [this time, the intergene insulator comes into contact with the insulator(s) after the enhancers to close a loop], we obtain the communication scheme that is no longer sequential. Instead, there is a “node” of three interacting insulators, and the distances to the yellow and to the white promoters are approximately the same. As a result, the enhancer is capable of effectively stimulating the white promoter. At the same time, the yellow gene is within a chromatin loop, which may lead to conformational and/or steric hindrances to the interaction of the enhancers and the yellow promoter. This mechanism of insulation is corroborated by the results of two studies showing that an enhancer can be blocked by its isolation from a target promoter in a quite small DNA loop (2, 5).

Interestingly, yellow placement between two S⁴ after 1A2 (Fig. 4D) or amid three copies of gypsy (Fig. 4E) or S4 (Fig. 4F) in a similar arrangement did not at all reinforce its blocking, in contrast to similar constructs with three 1A2 (Fig. 4B) or two Gy after 1A2 (Fig. 4C). Thus, the nature of interacting insulators appears to be critical for conformational isolation of the gene from enhancers in a chromatin loop.

This conclusion becomes more obvious in the light of data on the panel of constructs with only two insulators flanking the yellow gene (Fig. 5). Strong white stimulation by its enhancer in all transgenic lines carrying these constructs (Fig. 5B to F) indicated that it did not matter much whether two or three identical or different insulators were interposed between the enhancer and the white promoter. At the same time, yellow activation changed in opposite directions, depending on the nature of insulators combined. Placing yellow between two 1A2 or between 1A2 and Gy considerably reduced its activation by the enhancers, compared to the situation with the single 1A2 insulator (Fig. 5B and C). In contrast, S⁴ obviously interacted with 1A2 in the “insulator bypass assay” for yellow (Fig. 5D) and white (Fig. 4D and Fig. 5D) but had no effect on yellow expression when placed downstream of the yellow gene in addition to the invariant 1A2 (Fig. 5D).

FIG. 5.

Effects of interaction between two insulators enclosing the yellow gene. “N” is the number of lines in which flies acquired a new y or w phenotype relative to the control “[1]” lines carrying constructs with one insulator (1A2 or S⁴) at position −893 relative to the yellow transcription start site. For other designations, see Fig. 1 and 2.

Thus, S⁴ behaved quite comparably to 1A2 in the sequential schemes (Fig. 2 and 3), which was in line with their relative similarity in strength and pairing-neutralization ability. However, when two S⁴ insulators flanked the yellow gene, its activation was much stronger than in the case of single S⁴ insulator inserted between the enhancers and promoter (Fig. 5F). A similar picture was observed in the construct with three “strong” Gy insulators (Fig. 4E).

The factors underlying this situation are obscure. It may well be, for instance, that the artificial S⁴ element can acquire an unusual spatial structure due to DNA bending over mechanically reiterated AT-rich tracts (51), which, upon forming a chromatin loop by pairing, bring the enhancers and the promoter of the yellow gene in close proximity to each other. A similar mechanism might explain why a chromatin loop formed by three Gy insulators has an effect on the enhancer-promoter communication for the yellow gene. It seems likely that the topology for yellow and white in these cases is equal, and the effective enhancer-promoter interaction in these constructs is accomplished for both genes. These experiments show once again that enclosing a transcription-related element (gene) in a loop does not necessarily result in its functional isolation.

Insulator pairing around two genes equalizes activities of their enhancers.

Finally, we considered data on constructs with insulators surrounding both genes (Fig. 6). With any second (downstream) insulator except S⁴ (Fig. 6D), no stimulation of either gene was observed. This uniform attenuation of the enhancer-promoter communication is evidence for similarity of interactions in the 1A2-1A2 and 1A2-Gy pairs of insulators separated by more than 10 kb, which is in perfect agreement with previous observations (13, 45). As shown in the Fig. 6A, the topology of these constructs is similar to that in Fig. 5A with two insulators, except that the large loop now contains both genes.

In these constructs, the linear distances between enhancers and promoters remain the same as without insulator pairing. Once again, S⁴ did not change the enhancer-blocking activity of the 1A2 insulator, in contrast to the second 1A2 and gypsy insulators. Hence, the responses ranged from no apparent effect to complete blocking but, in every case, they were similar for the two “enclosed” genes. It appears that the interacting insulators in these constructs facilitate communication of the enhancers with promoters located outside the construct, downstream from the white gene (Fig. 6A).

Conclusions.

Our data show that merely two copies of the same or different insulators, provided they are appropriately arranged within a relatively small and simple construct inserted into the chromosome, can facilitate the expression of one gene (yellow) but not the next one (white) (Fig. 3) or, conversely, largely block the stimulation of the first gene while allowing full enhancement of the other (Fig. 4 and 5) or make them behave similarly (Fig. 6).

Taking into account the results of our previous studies (13, 44), we may conclude that the outcome depends on the nature of interacting insulators, as well as on the distances between all of the elements involved (enhancers, insulators, and promoters) and their relative “strengths.” Thus, formation of the chromatin loops by interacting insulators do not form independent chromatin domains but rather can be involved in regulation of proper enhancer-promoter communication.