Abstract

Introduction

Eukaryotic genomes encode thousands of genes that are linearly arranged along contiguous DNA molecules. Each gene must be transcribed under specific conditions, a process that ultimately requires RNA polymerases to recognize and access the gene’s core promoter. Activation of the RNA polymerase at the promoter is stimulated by an enhancer, a noncoding DNA element with binding sites for transcription factors, which may be encoded nearby or at a distance along the chromosome (Small and Arnosti 2020; Panigrahi and O’Malley 2021; Ray-Jones and Spivakov 2021). The predominant model for interaction between enhancers and promoters involves bringing the 2 elements into close physical proximity via “looping” of the intervening DNA between them, although alternative mechanisms may exist (Alexander et al. 2019; Benabdallah et al. 2019; Karr et al. 2022).

Many studies have aimed to understand how specific enhancer–promoter pairs are chosen such that genome-wide patterns of transcription are faithfully maintained. Recent evidence supports that configurations of local chromosome folding commonly known as topologically associated domains (TADs), characterized by preferred intradomain contacts relative to interdomain contacts, may limit the potential for enhancers and promoters that are encoded in separate domains to communicate with one another. Indeed, examples exist where disruption of TAD structure alters local gene expression in a manner consistent with changes in enhancer–promoter specificity, particularly in mammalian genomes (Lupiáñez et al. 2015; Hnisz et al. 2016; Hanssen et al. 2017; Kragesteen et al. 2018; Despang et al. 2019). However, global depletion of factors that establish TAD boundaries has minimal effect on genome-wide transcription in mammalian cell culture (Nora et al. 2017; Rao et al. 2017), and studies from other model organisms further suggest that changes to TAD structure may only affect expression for a minority of genes in the genome (Meadows et al. 2010; Lee et al. 2016; Shao et al. 2018; Ghavi-Helm et al. 2019; Kaushal et al. 2021).

Another potential mechanism to account for global patterns of gene expression suggests that enhancer–promoter specificity is largely determined by physical proximity of elements along the linear chromosome, with an enhancer generally favoring a promoter that is encoded nearby relative to a promoter that is more distal. This mechanism appears to be nearly universal across the genome of budding yeast, where most upstream activating sequences are encoded within a few hundred base pairs of their target promoters (Dobi and Winston 2007). In Drosophila, where enhancers may be up to several kilobases from their target promoters, an estimated 79–88% of developmental enhancers appear to target their nearest neighboring gene (Kvon et al. 2014), whereas in mammals, where enhancer–promoter distances tend to be even larger, an estimated 27–60% of enhancers act on their closest promoter (Li et al. 2012; Sanyal et al. 2012; van Arensbergen et al. 2014). Linear proximity may therefore account for enhancer–promoter specificity for many genes, but exceptions to this rule are found in larger genomes. Other potential mechanisms for specificity have been suggested to account for these exceptions. For example, analyses in Drosophila have shown that certain enhancers prefer to act on distinct core promoter elements (CPEs), ~6- to 12-bp sequence motifs that surround the transcriptional start site of eukaryotic promoters and facilitate binding of basal transcription factors (Li and Noll 1994; Merli et al. 1996; Ohtsuki et al. 1998; Butler and Kadonaga 2001; Juven-Gershon and Kadonaga 2010; Zabidi et al. 2015; Arnold et al. 2017). Other analyses have shown that some enhancer–promoter pairs rely on a specific sequence proximal to the promoter, often referred to as a “tethering” element, in order to faithfully interact (Qian et al. 1992; Lehman et al. 1999; Calhoun et al. 2002; Akbari et al. 2008; Kwon et al. 2009; Batut et al. 2022).

The above mechanisms suggest that what we observe as “enhancer–promoter specificity” results from an inherent competition among local promoters for the activity of a given enhancer, with factors including chromosome topology, linear proximity, core promoter identity, and tethering elements combining to determine the “winning” promoters that are the primary targets for a given enhancer. Consistent with this, studies in both flies and mammals have shown that an enhancer may switch its activity to a new target promoter when a previously favored target is deleted (Tian et al. 2019; Oh et al. 2021) or a parameter such as linear distance or core promoter identity is altered (Dillon et al. 1997; Ohtsuki et al. 1998; Kmita et al. 2002; Fukaya et al. 2016). Furthermore, biochemical and genetic studies provide evidence that multiple promoters can be activated by a common enhancer, and that neighborhoods of nearby genes frequently share common expression patterns (Li et al. 2012; Sanyal et al. 2012; Fukaya et al. 2016; Quintero-Cadena and Sternberg 2016; Oudelaar et al. 2019; Tian et al. 2019). Thus, promoter competition need not be an all or nothing affair; rather, some competing promoters may receive more of a specific enhancer’s activity, and others may receive less.

In Drosophila, an additional layer of complexity is added to the promoter competition landscape due to extensive somatic homolog pairing, which brings homologous chromosomes into close proximity in virtually all cells of the organism (Joyce et al. 2016; Peterson et al. 2021). The paired configuration of chromosomes allows for enhancers encoded on one chromosome to activate promoters encoded on another chromosome in trans, a phenomenon known as transvection (Lewis 1954). Analyses based on mutant alleles or transgenic constructs have demonstrated that promoters in cis and in trans to an enhancer will compete for the enhancer’s activity, with preference generally being given to the promoter in cis (Martínez-Laborda et al. 1992; Casares et al. 1997; Morris et al. 1999; Gohl et al. 2008; Bateman et al. 2012; Mellert and Truman 2012). However, it is as yet unclear how parameters that impact enhancer–promoter specificity in cis, including linear distance and core promoter identity, can impact competition between promoters in cis and in trans to an enhancer.

Here we use a transgenic approach to assess how factors such as linear spacing, promoter identity, and somatic homolog pairing/transvection can influence enhancer–promoter competition in Drosophila. Our data demonstrate that enhancer–promoter pairs with short intervening linear distances show a high degree of specificity, whereas longer linear distances result in increased promiscuity of the enhancer for other nearby promoters in cis and in trans. We further show that a mutant background with a higher degree of somatic homolog pairing skews an enhancer’s activity toward a promoter in trans at the expense of a promoter in cis. Finally, we show that mutations to CPEs result in decreased activation of the mutant promoter and elevated activation of a competing wild-type promoter, suggesting that enhancer activity in Drosophila is effectively balanced across local promoter competitors in cis and in trans.

Materials and methods

Results

Transgene studies of enhancer function typically position candidate enhancer sequences immediately adjacent to a heterologous promoter and reporter gene. To assess how altering the spacing between an enhancer and promoter can impact gene expression and promoter choice, we created a series of transgenic constructs where the synthetic enhancer GMR (Moses and Rubin 1991) was positioned at varying distances from the hsp70 promoter upstream of an mCherry coding region (Fig. 1a). We employed spacers of 3 sizes: small spacers of 100 or 200bp, typical of the distance between a proximal enhancer and a promoter in Drosophila (Zabidi et al. 2015), and a larger spacer of 3kb, which can support GMR activation of the hsp70 promoter in the context of a transgene (Bateman et al. 2012) and is typical of the distance between a developmental enhancer and promoter in Drosophila (Kvon et al. 2014). Each transgene was inserted into the genome at a common position via RMCE (Bateman et al. 2006), and gene expression was assessed via fluorescence microscopy and quantitative RT–PCR.

Open in a separate window

Fig. 1.

Enhancer–promoter spacing influences strength of transcriptional activation. a) Schematic of transgenes with spacers of varying sizes inserted between GMR and the hsp70 promoter (bent arrow) driving mCherry expression. b) Third instar eye disc from flies carrying a GMR-mCherry transgene lacking a spacer. A and P, anterior and posterior, MF, the morphogenetic furrow that separates differentiated posterior cells from undifferentiated anterior cells. Dashed box indicates general field of view for higher-resolution images in (c) and (d). c, d) mCherry fluorescence in eye discs from flies carrying GMR-mCherry with no spacer (c) and GMR-3kb-mCherry (d). Images represent max-projected confocal z-stacks. e) Relative quantification of mCherry mRNA levels in third instar eye-antennal discs from flies carrying transgenes with spacers relative to those with a transgene lacking a spacer. One-way ANOVA, P<0.0001; Dunnett’s multiple comparisons test relative to no spacer control: 100 bp, P = 0.003; 200 bp, P = 0.0004; 3 kb, P < 0.0001. All transgenes were analyzed as hemizygous insertions.

In larval eye discs carrying a transgene lacking a spacer, we observed robust mCherry fluorescence in retinal cells posterior to the morphogenetic furrow as previously observed for the enhancer GMR (Fig. 1, b and c) (Moses and Rubin 1991). Transgenes carrying small spacers were qualitatively indistinguishable in expression pattern and intensity relative to the transgene lacking a spacer (data not shown), whereas the larger 3-kb spacer resulted in noticeably weaker fluorescence, in agreement with previous observations (Fig. 1d) (Bateman et al. 2012). To quantify relative expression levels, we carried out qRT–PCR using RNA from whole eye-antennal discs carrying each transgene (Fig. 1e). Consistent with the qualitative assessment of fluorescence, the 3-kb spacer between GMR and the hsp70 promoter resulted in a significant decrease in mCherry mRNA levels relative to the construct lacking a spacer (6.9%), whereas small spacers resulted in a modest increase in mCherry mRNA levels relative to the construct with no spacer (120% and 130% expression for the 100- and 200-bp spacers, respectively). Thus, altered spacing between an enhancer–promoter pair can change the strength of transcriptional activation.

Increased enhancer–promoter spacing in cis results in augmented transvection

In Drosophila, prior studies have shown that promoters in cis and in trans to an enhancer will compete for the enhancer’s activity (Bateman et al. 2012; Mellert and Truman 2012). We therefore asked whether altering the spacing between an enhancer and promoter in cis could impact the competition between promoters in cis and in trans to an enhancer. To address this, we employed an enhancerless transgene with a GFP coding region driven by the hsp70 promoter that was integrated in the same genomic location that was analyzed above (Bateman et al. 2012). We established crosses to create larvae carrying the enhancerless hsp70-GFP construct on one homolog and the GMR-hsp70-mCherry construct (with or without spacer elements) on the other homolog such that the GMR enhancer would have a choice between transgenic promoters in cis, resulting in mCherry transcription, and in trans, resulting in GFP transcription (Fig. 2a).

Open in a separate window

Fig. 2.

Enhancer–promoter spacing influences strength of transvection. a) Schematic of genotypes to assess cis-/trans-promoter competition. Crosses were established to create progeny with an enhancerless hsp70-GFP on one homolog and GMR-mCherry constructs with or without spacers on the other homolog. b) Eye discs showing fluorescence of mCherry (representing activation in cis) and GFP (representing activation in trans) in flies carrying GMR-mCherry with no spacer (left column) or GMR-3kb-mCherry (right column). c, d) Relative quantification of mCherry (c) or GFP (d) mRNA levels in eye-antennal discs from flies carrying the experimental setup depicted in (a). Faded bars with dotted outlines in (c) represent relative mCherry mRNA levels from flies lacking the hsp70-GFP transgene on the homologous chromosome as observed in Fig. 1e. “No GMR” in (d) was generated from flies carrying hsp70-GFP on one homolog and no transgene on the other homolog, and was used as a measure of baseline GFP mRNA levels in the absence of enhancer action in trans. One-way ANOVA, P<0.001 for both (c) and (d); Dunnett’s multiple comparisons test relative to no spacer control in cis (c): 100 bp, P=0.002; 200 bp, P=0.16 (not significant); 3 kb, P<0.0001; Dunnett’s multiple comparisons test relative to no spacer control in trans (d): 100 bp, P<0.0001; 200 bp, P<0.0001; 3 kb, P=0.99 (not significant); no GMR, P<0.0001.

We first assessed discs carrying each genotype via fluorescence microscopy. In all cases, we observed GFP fluorescence in a subset of mCherry-positive cells, consistent with previous observations of variegated enhancer action in trans by GMR (Fig. 2b and data not shown) (Bateman et al. 2012). We then used quantitative RT–PCR to compare levels of mCherry and GFP mRNA in discs from the different genotypes (Fig. 2, c and d). Notably, all genotypes carrying GMR resulted in higher levels of GFP mRNA when compared to control discs carrying hsp70-GFP only, providing further evidence that transvection is supported by each construct (Fig. 2d). However, small spacer segments of 100 and 200bp resulted in significantly increased GFP expression relative to the GMR-hsp70-mCherry construct that lacks a spacer, suggesting that the change in cis-spacing between GMR and the hsp70 promoter in these constructs provides a competitive advantage to the promoter in trans (Fig. 2d). Furthermore, analysis of mCherry mRNA levels in these discs shows that the increase in enhancer action in trans comes at the expense of the transgenic promoter in cis, since spacers of 100 and 200bp now lead to decreased mCherry mRNA levels relative to the construct lacking a spacer, whereas the same constructs showed the opposite effect on expression levels in the absence of the promoter in trans (Figs. 1e and ​and2c).2c). In contrast to the small spacers, the presence of the 3-kb spacer results in roughly equivalent levels of GFP trans-activation relative to the GMR-hsp70-mCherry construct lacking a spacer, despite the relatively low levels of mCherry cis-activation caused by the spacer (Fig. 2, c and d). In sum, small changes to the linear spacing of an enhancer and promoter pair can augment the capacity of the enhancer to act in trans.

Increased enhancer–promoter spacing results in ectopic expression of flanking genes

We were surprised that the presence of the 3-kb spacer between GMR and the hsp70 promoter resulted in a dramatic attenuation of mCherry transcription in cis, but no significant change in activation of a promoter in trans when compared to a construct lacking a spacer. We reasoned that the presence of the larger 3-kb spacer could result in a reduction in the specificity of the GMR enhancer for its target promoter, freeing it to act on other local promoters near the transgenic insertion site. We therefore asked whether GMR activity could be detected in the expression patterns of genes flanking the insertion site of the transgene in lines carrying the 3-kb spacer. Our analysis centered on a transgenic insertion site immediately upstream of GstS1 at polytene position 53F8 (Fig. 3a, Supplementary Fig. 1). Indeed, in situ hybridization of third instar larval eye discs using a probe complementary to the GstS1 coding region showed only weak staining across the tissue in control w¹¹¹⁸ discs (Fig. 3b), but robust staining in cells posterior to the morphogenetic furrow from discs carrying a GMR-3kb-cherry transgene with the 3-kb spacer (Fig. 3c), indicating that GstS1 transcription was enhanced by GMR in the transgenic line.

Open in a separate window

Fig. 3.

Increased enhancer–promoter spacing results in ectopic expression of flanking genes. a) Simplified view of the genomic locus surrounding the RMCE insertion site at polytene position 53F. Five genes and their major promoters are depicted, with the table to the right indicating the linear distances in kilobases between GMR and the nearest promoter of each gene for GMR-mCherry and GMR-3kb-mCherry insertions. Two additional small protein-coding genes were not analyzed (CG6984, encoded between GstS1 and Sply, and CG46492, encoded within an intron of CG46491), and an additional weak GstS1 promoter has been annotated to the right of the transgenic insertion site in some annotations (Wangler et al. 2015). Vertical dashed line indicates a TAD boundary; an additional TAD boundary is located roughly 5 kb to the left of CG30460 (Cubeñas-Potts and Corces 2015; Li et al. 2015; Ramírez et al. 2018). All transgenes were inserted such that transcription proceeds from right to left. A more detailed view of the locus is presented in Supplementary Fig. 1, including TAD structure, ATAC-seq/FAIRE-seq peaks, and alternative TSS/splicing of each gene. b, c) RNA in situ hybridization targeting GstS1 in eye discs from control flies lacking a transgene (w¹¹¹⁸, b) or with an insertion of GMR-3kb-mCherry (c). d, e) Relative quantification of mRNA levels from the 5 genes depicted in (a) in larval eye-antennal discs from flies carrying GMR-3kb-mCherry relative to GMR-mCherry (d), or from flies carrying GMR-hsp70-lacZ or a promoterless GMR-(P-)lacZ relative to an enhancerless hsp70-lacZ (e). Unpaired t-tests with Holm–Sidak correction for multiple comparisons in GMR-cherry vs GMR-3kb-cherry discs (d), adjusted P<0.05 for all 5 genes; in GMR-hsp70-lacZ vs GMR-(P-)-lacZ (e), adjusted P<0.05 for GstS1, CG46491, and CG15611 only. f) Allele-specific qRT–PCR of CG46491. Top, schematic showing experimental setup based on TaqMan probes that differentiate homologs based on a polymorphic indel that differs between the transgenic chromosome (allele 2) and DGRP line 443 (allele 1). Bottom, relative quantification of mRNA levels for GMR-3kb-GFP compared to 3kb-GFP lacking an enhancer. Unpaired t-tests with Holm–Sidak correction for multiple comparisons, adjusted P<0.05 for both cis and trans expression. All transgenes were analyzed as hemizygous insertions.

To further explore potential GMR activity on genes flanking our transgenic insertion site, we used quantitative RT–PCR to compare expression levels of 5 nearby genes in discs carrying transgenes with or without the 3-kb spacer (Fig. 3, a and d). Relative to discs carrying GMR-mCherry lacking a spacer, we observe increased expression of 3 nearby genes, GstS1, Sply1, and CG46491, in discs carrying GMR-3kb-mCherry (Fig. 3d). The nearest promoters of these genes were less than 12kb from the GMR enhancer in GMR-3kb-mCherry lines (Fig. 3a), whereas 2 genes with more distal promoters, CG30460 and CG15611, showed reduced expression with the addition of the 3-kb spacer. Note that the change in GMR target specificity in the presence of the 3-kb spacer is unlikely to be due to an ectopic insulator activity of the spacer element: first, roughly 80% of the spacer overlaps a lambda DNA fragment that shows no evidence of insulator activity in a mini-white assay in Drosophila (Chung et al. 1993). Furthermore, an insulator would be expected to have a directional effect, blocking enhancer–promoter interactions in one direction but not the other (Holdridge and Dorsett 1991; Geyer and Corces 1992; Kellum and Schedl 1992; Chung et al. 1993; Roseman et al. 1993; Cai and Levine 1995), whereas our data demonstrate elevated expression of flanking genes on both sides of a transgene carrying the 3-kb spacer (Fig. 2, a and d). In sum, our data show that an increased spacing between GMR and the hsp70 promoter results in increased expression of flanking genes at the expense of robust transgene expression, consistent with a loss of enhancer specificity for the intended target promoter of the transgenic construct.

The increased expression of genes flanking the transgene in the presence of the 3-kb spacer could be due to an inherent competition between promoters for activation by GMR. According to this model, close spacing between GMR and the transgenic hsp70 promoter could constrain the activity of the enhancer, whereas reduced interactions between GMR and the hsp70 promoter in the presence of the 3-kb spacer could free the enhancer to interact with other promoters. To further test this possibility, we asked whether a similar augmentation of flanking gene expression would be seen in the absence of a transgenic promoter. We had previously integrated transgene variants based on GMR, the hsp70 promoter, and a lacZ coding region in the identical genomic location upstream of GstS1 as analyzed above (Bateman et al. 2012). Among these variants were a “complete” transgene, GMR-hsp70-lacZ, an “enhancerless” transgene lacking an enhancer but carrying a minimal promoter and coding region, hsp70-lacZ, and a “promoterless” transgene lacking a promoter but carrying an enhancer and coding region, GMR-(P-)-lacZ. Using the enhancerless construct hsp70-lacZ as a baseline, we use qRT–PCR to assess mRNA levels of genes flanking the transgenic insertion site in larval eye-antennal discs (Fig. 3e). In discs carrying the complete transgene GMR-hsp70-lacZ, we observed elevated levels of all 5 flanking genes relative to the enhancerless construct, demonstrating that the presence of the GMR enhancer results in a modest overall increase of gene expression across the locus in eye-antennal discs. Notably, loss of the hsp70 promoter in the GMR-(P-)-lacZ discs resulted in a further increase in mRNA levels of the 2 genes closest to the insertion site, GstS1 and CG46491, a pattern remarkably similar to that observed for the 3-kb spacer placed between GMR and the hsp70 promoter. Our observations are therefore consistent with a model wherein increasing the spacing between an enhancer and promoter effectively “releases” the enhancer to act on other potential promoter targets.

The increased expression of genes flanking the GMR-3kb-mCherry insertion could result from elevated transcription from flanking promoters in cis and/or in trans to the transgene. To further explore cis- vs trans-promoter usage, we used allele-specific qRT–PCR targeting the flanking gene CG46491 to separately measure transcripts generated from each homologous chromosome (Fig. 3f) (see Materials and Methods). In designing this experiment, we considered that the hemizygous insertion of GMR-3kb-mCherry could have a local disruptive effect on somatic homolog pairing with the potential to negatively impact transvection. We therefore used 2 alternative existing transgenic insertions, GMR-3kb-GFP (analogous to GMR-3kb-mCherry, but instead carrying a GFP coding region) and 3kb-GFP, a construct lacking an enhancer but otherwise identical to GMR-3kb-GFP, as a negative control (Bateman et al. 2012). Since these 2 lines differ from one another only by the 190-bp sequence of the GMR enhancer, we reasoned that they would have a similar impact on local pairing and therefore represent a fair comparison. After normalizing expression of CG46491 in cis and in trans to the transgenic insertion based on the discs lacking GMR, our analysis showed that the presence of GMR-3kb-GFP significantly increased expression from the CG46491 promoter in cis, whereas expression from the CG46491 promoter in trans was instead significantly decreased (Fig. 3f). We observed nearly identical changes in gene expression from the 2 homologous chromosomes in parallel experiments using a w¹¹¹⁸ line that lacks a transgenic insertion as a negative control (data not shown), indicating that the presence of the hemizygous insertion likely has little disruptive effect on local pairing. In sum, our data suggest that the presence of a 3-kb spacer between the GMR enhancer and the hsp70 promoter results in elevated expression of nearby genes primarily from the chromosome in cis, with a potentially disruptive effect on expression from a paired chromosome in trans.

Changing enhancer identity alters the impact of enhancer–promoter spacing

Thus far, our analysis has focused on the synthetic enhancer GMR. To address whether other enhancers may be similarly impacted by altered enhancer–promoter spacing, we created 2 constructs using the minimal sevenless (sev) enhancer, which drives expression in a subset of mature retinal and cone cells in third instar larval eye discs (Bowtell et al. 1991) (Fig. 4a). Surprisingly, fluorescence microscopy of discs carrying a 100-bp spacer between the sev enhancer and the hsp70 promoter showed a sharp decrease in transgene expression relative to discs lacking this spacer (Fig. 4b). This decrease was confirmed by quantitative RT–PCR, which showed expression levels of only 31% for the transgene carrying the spacer relative to the transgene lacking it (Fig. 4c). Since the identical 100-bp spacer did not lead to decreased transgene activation by the enhancer GMR, our data suggest that different enhancer–promoter pairs can be differently impacted by the same change in linear spacing.

Open in a separate window

Fig. 4.

The sev minimal enhancer shows reduced specificity for the transgenic hsp70 promoter in the presence of a 100-bp spacer. a) Schematic of transgenes with the sev minimal enhancer and the hsp70 promoter driving mCherry expression. b) Representative third instar eye discs from flies carrying sev-mCherry transgenes with or without a 100-bp spacer. c, d) Relative quantification of mCherry (c) or GstS1 (d) mRNA levels in third instar eye-antennal discs from flies carrying a transgene with a 100-bp spacer relative to those with a transgene lacking a spacer. Unpaired t-test, P<0.001 in both (c) and (d). Transgenes in panels (a–d) were analyzed as hemizygous insertions. e) Schematic of genotype to assess cis-/trans-promoter competition (left) and relative quantification of mCherry and GFP mRNA levels in eye-antennal discs with or without a 100-bp spacer between the sev minimal enhancer and the hsp70 promoter. Unpaired t-test, P<0.0001 for mCherry in cis, P=0.09 (not significant) for GFP in trans. f) Eye discs showing fluorescence of mCherry (enhancer acting in cis) and GFP (enhancer acting in trans) in flies carrying sev-mCherry with no spacer (top row) or with a 100-bp spacer (bottom row).

To address whether the decreased transgene expression by the sev enhancer in the presence of the 100-bp spacer may be associated with increased enhancer activity on flanking genes, we compared expression levels of the neighboring GstS1 gene in eye-antennal discs carrying the 100-bp spacer to those lacking it. As was the case in the GMR constructs that carried a 3-kb spacer, we observed an increase in GstS1 expression in discs carrying the sev-mCherry construct with a 100-bp spacer relative to those that lacked a spacer (Fig. 4d). We also compared the capacity of the sev enhancer to support transvection with and without a 100-bp spacer (Fig. 4, e and f). In this case, we did not observe a statistically significant increase in trans-activation of GFP by the sev enhancer in the presence of the spacer relative to discs carrying transgenes lacking a spacer, a similar result to that observed for GMR in the presence of a 3-kb spacer (Fig. 3c). In sum, the behavior of the sev enhancer when separated from the hsp70 promoter by a 100-bp spacer is remarkably different from the behavior of GMR with the same spacer fragment; instead, its behavior is more similar to that of GMR in the presence of a 3-kb spacer, supporting that the specificity of interactions between different enhancer–promoter pairs can be differently impacted by changes in their spacing.

Loss of CPEs impacts enhancer–promoter specificity

We found it interesting that the augmented transvection by GMR that is supported by spacers of 100 and 200bp appears to come at the cost of the cis-activity of the enhancer (Fig. 3, b and c), suggesting a balance in the distribution of enhancer activity between promoters in cis and trans. To further explore this balance, we created transgenes with altered CPEs in order to assess how they compete with a wild-type hsp70 promoter in a cis-/trans-promoter competition assay (Bateman et al. 2012). Specifically, we mutated 2 CPEs, either the TATA box or the MTE, of the hsp70 promoter to create “TATAless” and “MTEless” variants of our transgenic reporters (Fig. 5a). In these experiments, none of the reporters carry an added spacer between the enhancer and promoter, allowing us to specifically query the impact of promoter identity on the balance of enhancer activity between competing promoters in cis and in trans.

Open in a separate window

Fig. 5.

Mutations in CPEs impact cis-/trans-promoter specificity. a) Schematic of CPEs (TATA box, Inr, and MTE) in the wt hsp70 promoter and mutations designed to generate TATAless and MTEless variants. Sequence context of the wt promoter is provided to the right, with TATA box in blue, Inr in orange, and MTE in green. The transcription start site (TSS) is boxed. b–e) Relative quantification of GFP (b, e), mCherry (c), or lacZ (d) mRNA levels from eye-antennal discs carrying transgenes with promoter variants: b) hsp70-GFP carrying promoter variants and lacking an enhancer. c) GMR-mCherry with a wild-type promoter in trans to hsp70-GFP carrying promoter variants. Variation in expression reflects the ability of the promoter in trans to compete for GMR activity. d) GMR-lacZ carrying promoter variants. e) hsp70-GFP with a wild-type promoter in trans to GMR-lacZ carrying promoter variants. Variation in expression reflects the ability of the promoter in cis to compete for GMR activity. One-way ANOVA, P<0.0002 for each of (b) through (e); Dunnett’s multiple comparisons test, P<0.001 for each comparison of mutant promoter to wild type in (b) through (e). Transgenes in panels (b) and (d) were analyzed as hemizygous insertions. f) Eye discs showing antibody staining of beta-galactosidase (enhancer acting in cis, top row) and GFP (enhancer acting in trans, middle row) in flies carrying wild type (first column), MTEless (second column), or TATAless (third column) hsp70 promoters upstream of lacZ and a wild-type hsp70 promoter upstream of GFP. g) cis-/trans-promoter competition between GMR-hsp70-mCherry and hsp70-GFP transgenes in wild type and Cap-H2⁰⁰¹⁹/Df(3L)Exel6159 backgrounds. Unpaired t-test comparing wild type to mutant, P=0.0007 for mCherry (cis), P=0.003 for GFP (trans). Note that all transgenes analyzed in Fig. 5 were constructed without spacer elements between the enhancer and promoter.

We first examined variants of the hsp70-GFP construct that lacks an enhancer in order to assess baseline expression levels (Fig. 5b). In the absence of an enhancer, we observe decreases in baseline expression to 59.1% and 32.2% of wild-type levels for the MTEless and TATAless constructs, respectively, demonstrating that loss of either CPE is detrimental to the basal transcriptional output of the hsp70 promoter. We then placed each construct in trans to a GMR-hsp70-mCherry construct and assessed how each variant competes with the wild-type hsp70 promoter in cis to the GMR enhancer by quantifying mCherry mRNA levels (Fig. 5c). If activation of promoters in cis and trans by the GMR enhancer is a balanced competition as suggested by our previous experiments using spacers, we would expect any decrease in trans-activation associated with CPE mutations to result in a concomitant increase in activation of a promoter in cis. Indeed, relative to a control hsp70-GFP with intact CPEs, MTEless-GFP and TATAless-GFP constructs in trans to GMR-hsp70-mCherry result in higher levels of mCherry transcripts, consistent with a balanced competition between the promoters in cis and in trans to the enhancer.

We then explored the converse experiment by creating MTEless or TATAless variants of a promoter in cis to the GMR enhancer. In this case, variant constructs based on an mCherry coding region showed evidence of a cryptic promoter activity, and we therefore used a lacZ coding region instead. Similar to our assessment of baseline expression of promoter variants in the absence of an enhancer (Fig. 5b), activation of each variant by GMR was decreased relative to the activation of an intact hsp70 promoter, further demonstrating the importance of each CPE for efficient promoter function (Fig. 5d). We then assessed how each variant competes with a wild-type hsp70 promoter in trans to the GMR enhancer by placing an hsp70-GFP insert in trans and quantifying GFP mRNA levels (Fig. 5, e and f). Relative to trans-activation of hsp70-GFP by a control GMR-hsp70-lacZ construct with unaltered CPEs, we observed elevated levels of GFP mRNA via trans-activation of hsp70-GFP by GMR-TATAless-lacZ and GMR-MTEless-lacZ, demonstrating that the reduced activation associated with promoter variants in cis is balanced by an increased activation of a competitive promoter in trans.

In a final comparison, we quantitatively assessed enhancer action in cis and in trans in discs lacking Cap-H2, a component of the Condensin II complex (Hartl et al. 2008). Previous analyses have shown that flies lacking Cap-H2 display augmented transvection at several loci, which is thought to be due to increased levels of somatic homolog pairing observed in this mutant background (Hartl et al. 2008; Joyce et al. 2012). Relative to a wild-type Cap-H2 background, we indeed observe increased trans-activation of hsp70-GFP by GMR-mCherry in discs trans-heterozygous for loss-of-function alleles of Cap-H2 (Fig. 5g). Consistent with our observations with small spacer elements, the augmented transvection in a Cap-H2 mutant background is concomitant with a decrease in cis-activation of mCherry by GMR, further suggesting a balanced competition between promoters in cis and in trans to an enhancer. Overall, our data support a natural competition between local promoters for the activity of a nearby enhancer, where spacing/proximity and promoter strength/identity are important components that can influence specificity.

Discussion

Studies from diverse metazoan systems suggest that neighborhoods of promoters can compete for the activity of local enhancers, with both cis- and trans-interactions potentially being supported. Here we explore how changes to the linear distance between an enhancer and a promoter can skew the competition and thereby influence enhancer–promoter specificity in Drosophila, a system where trans-interactions are abundant due to genome-wide somatic homolog pairing. The use of heterologous enhancer–promoter pairs in a transgenic approach allows us to isolate the parameter of linear distance in the absence of sequences that may influence competition at endogenous loci, such as tethering elements. Our data show that close linear spacing between an enhancer and promoter supports target specificity, whereas increasing linear spacing leads to increased promiscuity by the enhancer for other nearby promoter targets in cis and in trans (Fig. 6).

Open in a separate window

Fig. 6.

Model for influence of spacing on enhancer–promoter specificity in Drosophila. Close spacing between an enhancer and promoter results in high enhancer activity on that promoter in cis (broad arrow), and less activity on nearby potential competing promoters in cis and in trans. With greater spacing, enhancer activity increases on competing promoters (dashed arrows), potentially including nearby promoters in trans, at the expense of the target promoter.

Our data show that the synthetic enhancer GMR shows robust activation of the hsp70 promoter when it is encoded within a few hundred base pairs, but reduced activation when placed at a distance of 3kb. Consistent with this, Shlyueva et al. (2014) showed that ecydosone-responsive genes in Drosophila were more highly expressed when an enhancer was within 1kb of its target promoter, and Fukaya et al. (2016) showed that increasing the spacing between an enhancer and promoter from 1 to >7kb results in dramatically reduced frequency of transcriptional bursts. Curiously, analyses similar to ours demonstrate that the transcriptional impact of linear distance between enhancers and promoters varies greatly in different genomes; in Saccharomyces cerevisiae, spacing of just a few hundred base pairs abolishes detectable transcription (Dobi and Winston 2007), whereas in mouse embryonic stem cells, transcriptional output and enhancer–promoter contact frequency decreases steadily over linear spacing of hundreds of kilobases (Zuin et al. 2022). It is likely that species-specific parameters such as chromatin topology, gene density, and species-specific factors that promote or antagonize long-distance enhancer–promoter communication contribute to these differences.

Prior biochemical analyses of Drosophila genome topology have shown that cis-contacts are favored over trans-contacts over short linear distances, but the two become more alike as linear distance increases (AlHaj Abed et al. 2019; Erceg et al. 2019). We therefore hypothesized that, in general, increasing the linear distance between GMR and the hsp70 promoter would be associated with increased activity of GMR on a promoter in trans relative to a promoter in cis. In the case of smaller spacers of 100–200bp, GMR indeed increased its activity on a paired hsp70 promoter in trans at the expense of the promoter in cis, but in the case of a larger 3-kb spacer, there was little difference in the strength of transvection relative to a transgene lacking a spacer. Since the 3-kb spacer resulted in ectopic activation of several endogenous genes surrounding the transgenic insertion site, it appears that other neighboring promoters were favored over the hsp70 promoter in trans at this distance. The elevated expression of nearby neighboring genes was also observed in lines carrying transgenes that lacked a promoter altogether, implying that the 3-kb spacer effectively “releases” the GMR enhancer to act on other targets. These observations are consistent with prior studies of transvection in Drosophila where deletion of a promoter in cis to an enhancer results in increased activation of a promoter in trans (Geyer et al. 1990; Martínez-Laborda et al. 1992; Casares et al. 1997; Morris et al. 1999; Gohl et al. 2008; Bateman et al. 2012; Mellert and Truman 2012). Similarly, deletion of a promoter in human cells also results in enhancer retargeting and ectopic activation of neighboring genes (Oh et al. 2021), suggesting that the “capture” of an enhancer’s activity by a nearby promoter may be a common mechanism to prevent activation of alternative target genes. However, we note that a study in Caenorhabditiselegans using transgenes with spacers to increase linear distances between enhancers and promoters also showed reduced activation of the transgenic promoter with increasing distance, but in this case it was associated with a concurrent reduction in transcription of neighboring genes (Quintero-Cadena and Sternberg 2016). Thus, it may be that cooperative interactions among promoters are also supported in some species and/or genomic contexts.

Our data also provide information on how promiscuous enhancer activity may be distributed among nearby promoters. In discs carrying GMR-3kb-mCherry, 3 flanking genes were found to have elevated expression relative to discs lacking the 3-kb spacer, with the highest activation seen for the 2 genes with promoters closest to the GMR enhancer, CG46491 and GstS1. Notably, GstS1 showed the higher increase in expression despite having a greater linear distance from GMR to its promoter (6.3kb from GMR to the nearest encoded promoter for GstS1 vs 3.4kb for CG46491 in the GMR-3kb-mCherry line), indicating that shared GMR activity is not distributed among promoter targets as a simple function of distance. However, Hi-C analysis of Drosophila cell lines predicts that a TAD boundary exists between the transgenic insertion and the CG46491 promoter, which could account for the unexpected difference in expression (Cubeñas-Potts and Corces 2015; Li et al. 2015; Ramírez et al. 2018). Similarly, Sply1, with a promoter 11.8kb from GMR and no intervening TAD boundary, showed modestly increased expression in GMR-3kb-mCherry discs, whereas CG15611, with 12.0kb from GMR to its promoter but crossing a TAD boundary, did not. Thus, our data support that a combination of linear distance and chromosome topology likely play a role in determining the winners and losers among local promoters competing for promiscuous GMR activity in the presence of a 3-kb spacer. However, some of the changes in gene expression are less easily explained, such as the unexpected decrease in expression of the 2 most distal of the genes analyzed, CG15611 and CG30460, or the allele-specific increase in CG46491 expression exclusively from the promoter in cis to the transgene and concomitant decrease in expression from the promoter in trans. A more thorough analysis of changes to chromosome topology and interactions with endogenous enhancers may help to elucidate these differences.

We were surprised that 2 different enhancers, GMR and the sev minimal enhancer, behaved differently in the presence of a small spacer in our assay. In the case of GMR, activation of the transgenic hsp70 promoter in cis was even more efficient in the presence of a spacer of 100 or 200bp relative to constructs lacking a spacer. Although these spacers correspond to no more than a single nucleosome of DNA, they could potentially promote more efficient looping between the enhancer and promoter, as suggested by in vitro assays that show an optimal looping distance of 500bp for naked DNA and 200bp for chromatin (Ringrose 1999). However, in the presence of the exact same 100-bp spacer sequence and reporter construct, the minimal sev enhancer showed reduced expression of the hsp70 promoter fused to mCherry relative to a transgene lacking the spacer. A similar strong decrease in mCherry expression was observed in an independent line carrying a 200-bp spacer between the sev enhancer and the hsp70 promoter (data not shown), demonstrating that optimal enhancer–promoter spacing varies among different enhancer–promoter pairs. Prior qRT–PCR analysis using lacZ-based constructs without spacers showed that transcriptional activation by the sev minimal enhancer is approximately six times greater than that of GMR (Blick et al. 2016); it may therefore be that a stronger enhancer is more readily impacted by small changes in linear distance to its promoter target relative to a weaker enhancer. In addition, the 2 enhancers differ in the transcription factors that are responsible for their activities, where the 190-bp synthetic GMR consists only of 5 binding sites for the transcriptional activator Glass (Moses et al. 1989; Moses and Rubin 1991), whereas ChIP-seq data for the 476-bp sev minimal enhancer shows binding of Sine Oculis in third instar eye-antennal discs (Jusiak et al. 2014) in addition to binding of Spineless, Pipsqueak, Cyclin G, Vsx2, Glass, and Tailless in different tissues at various stages of development (Celniker et al. 2009; Luo et al. 2020). While it is possible that the reduction in hsp70 promoter activation by the sev enhancer in the presence of the 100-bp spacer could be due to a sequence-specific effect, perhaps interfering with binding of specific transcription factors to the enhancer, our observation that activation of a neighboring gene is increased and that of a promoter in trans is not negatively impacted argues that the function of the enhancer itself is not compromised by the spacer. Furthermore, Camino et al. (2020) previously showed that several enhancers that act in the pupal epidermis are differently impacted by varying enhancer–promoter spacing in a transgenic model. A requirement for close enhancer–promoter proximity is not observed at the endogenous sev locus, where the sev enhancer activates its target promoter from an intronic position approximately 6kb downstream (Bowtell et al. 1991; Rivera et al. 2019). The endogenous sev locus may encode other sequences that support distal activation of the promoter, such as a promoter–proximal tethering element (Qian et al. 1992; Lehman et al. 1999; Calhoun et al. 2002; Akbari et al. 2008; Kwon et al. 2009) or enhancer-associated “remote control” element (Swanson et al. 2010). Alternatively, because the sev promoter is the nearest target to the sev enhancer at its endogenous locus due to the position of the enhancer within the gene body, perhaps competition with other potential target promoters is minimized in this context.

Prior analyses have shown that the activity of Drosophila enhancers is balanced between target promoters in cis and in trans, with the promoter in cis generally being favored. One model to account for the preferred activation of a promoter in cis is that it is simply a “closer” target since it is encoded on the same chromosome, and therefore productive enhancer–promoter interactions are more likely. Our data using spacers show that a small increase in the distance between an enhancer and promoter in cis can skew the activity of the enhancer toward a promoter in trans, as does a CapH2 mutant background with higher levels of pairing, both of which are in general agreement with a “the closer, the better” model. However, we also show that the synthetic enhancer GMR produces reduced transcriptional output from variant hsp70 promoters lacking either a TATA box or an MTE element, and mutant promoters have a reduced capacity to compete for enhancer activity when placed in trans to a wild-type hsp70 promoter. These quantitative data augment prior targeted analyses showing that weak or mutated promoters are poor competitors when placed in trans to a relatively stronger promoter (Morris et al. 2004; Lee and Wu 2006; Mellert and Truman 2012), and demonstrate that cis-/trans-promoter competition is likely a function of both proximity and promoter identity and/or competency, where a stronger or more attractive promoter will sequester enhancer activity away from a weaker or less attractive one. Although structurally wild-type Drosophila chromosomes will likely have near-identical enhancer and promoter sequences in homologous positions, certain developmental events such as stochastic selection of rhodopsin gene expression in photoreceptors rely on trans-interactions between alleles (Johnston and Desplan 2014; Anderson et al. 2017). Thus, skewing of cis- vs trans-interactions by allele-specific differences in regulatory sequences could have meaningful developmental impacts.

Finally, our data suggest that enhancer–promoter linear proximity can be an important component influencing local promoter competition for enhancer activity and may therefore contribute globally to enhancer–promoter specificity in Drosophila. Although other mechanisms clearly exist to account for cases where an enhancer skips nearby genes for a more distal target, this may only be relevant for a minority of genes given that an estimated 12% of developmental enhancers behave in this way (Kvon et al. 2014), and the majority of housekeeping enhancers overlap their presumed target promoters (Zabidi et al. 2015). It is therefore likely that a combination of factors including enhancer identity, linear proximity, chromosome topology, CPEs, and tethering elements combine to govern global cis- and trans-interactions between enhancers and promoters and thereby maintain wild-type patterns of gene expression.

Data availability

Stocks and plasmids are available upon request. The authors affirm that all data necessary for confirming the conclusions of the article are present within the article, figures, and tables.

Supplemental material is available at GENETICS online.

Supplementary Material

Acknowledgments

Contributor Information

Jack R Bateman, Biology Department, Bowdoin College, Brunswick, ME 04011, USA.

Justine E Johnson, Biology Department, Bowdoin College, Brunswick, ME 04011, USA.

Literature cited