Abstract
A long-standing question in metazoan gene regulation is how remote enhancers communicate with their target promoters over long distances. Combining genome editing and quantitative live imaging we simultaneously visualize physical enhancer–promoter communication and transcription in Drosophila embryos. Enhancers regulating pair rule stripes of even-skipped expression activate transcription of a reporter gene over a distance of 150 kb. We show in individual cells that activation only occurs after the enhancer comes into close proximity with its regulatory target and that upon dissociation transcription ceases almost immediately. We further observe distinct topological conformations of the eve locus, depending on the spatial identity of the activating stripe enhancer. In addition, long-range activation results in transcriptional competition at the endogenous eve locus, causing corresponding developmental defects. Overall, we demonstrate that sustained physical proximity and enhancer–promoter engagement are required for enhancer action, and we provide a path to probe the implications of long-range regulation on cellular fates.
I. INTRODUCTION
Enhancers play a key role in the control of gene expression and development (Benoist and Chambon, 1981; Levine, 2010; Long et al., 2016). These 50–1500 base pair cis-regulatory elements stimulate transcription from core promoters in a time and tissue specific manner by recruiting context-dependent transcriptional activators and repressors (Levine, 2010; Buecker and Wysocka, 2012). Functions of enhancers are thought to be independent of their position and orientation. Enhancers can act in cis and in trans (Lee and Wu, 2006; Markenscoff-Papadimitriou et al., 2014), can be located upstream or downstream of the core promoter as well as in introns, and as far as a Mb from the promoter (Long et al., 2016). Whole-genome methods have shown that the human genome is riddled with enhancers, with estimates ranging from 200,000 to over a million (Consortium, 2012). Importantly, a significant fraction of these enhancers are located at large genomic distances from the promoters they regulate (Tolhuis et al., 2002; Uslu et al., 2014; Zhang et al., 2013). Even for a compact genome like Drosophila melanogaster, at least 30% of enhancer–promoter interactions occur over 20 kb, and in many cases over intervening genes (Arnold et al., 2013; Ghavi-Helm et al., 2014; Kvon et al., 2014).
In this paper we focus on the mechanism by which enhancers communicate with their target genes over large distances. Several models have been proposed (Benabdallah and Bickmore, 2015; Blackwood and Kadonaga, 1998), the most popular of which involves looping of the flexible chromatin polymer such that distal enhancers come into direct contact with their promoter targets. Early evidence from genetic experiments has already revealed the presence of functional long-range enhancer–promoter communication in both vertebrates and flies (Lettice et al., 2003; Lewis, 1978; Qian et al., 1993; Sagai et al., 2005). Recently, chromatin conformation capture (3C) based approaches, involving the cross-linking of chromatin, revealed the physical proximity of distal enhancers and promoters in a variety of vertebrate cells and tissues (Kagey et al., 2010; Mifsud et al., 2015; Spitz, 2016). Yet, despite these demonstrations of both functional and physical interactions, fundamental questions about how enhancers communicate with their target promoters over large distances and how they activate transcription still remain (Levine et al., 2014).
A major gap in our understanding pertains to the role of physical enhancer–promoter interactions in transcriptional regulation. On the one hand, 3C-based experiments have revealed extensive enhancer–promoter interactions that are conserved among developmental stages, cell fates or evolution, suggesting a permissive role of the physical enhancer–promoter interactions on transcriptional activity (Ghavi-Helm et al., 2014; Jin et al., 2013; Sanyal et al., 2012). On the other hand, lineage-specific enhancer–promoter interactions are found to be prevalent in many developmental contexts, arguing for the possibility of an instructive role (Dixon et al., 2015; Javierre et al., 2016). Part of the reason for this discrepancy lies in the fact that the static picture gained from currently employed techniques can only provide correlative evidence. Furthermore, most genomic experiments rely on bulk assays, which might cover temporal and spatial heterogeneity within the samples.
Crucially, a direct dynamic link between enhancer–promoter proximity and transcriptional activation is lacking. Is proximity needed for transcriptional activation, or is it a consequence of transcriptional activity? Even if the most popular hypothesis is that proximity is necessary for transcriptional activation, the above discussed studies either measure transcriptional activity or physical proximity, and experiments that measure both transcription and physical proximity simultaneously have not been reported. What degree of physical proximity is needed for activation? Is there a topological distinction at the gene locus when different enhancers are activating the gene? It is further unclear whether physical proximity is needed transiently for a limited time interval to establish an active state—a state that can be remembered once proximity is lost—or whether continued transcription requires that the enhancer and promoter remain in sustained proximity. To address these fundamental questions an approach that simultaneously captures the dynamics of enhancer–promoter distances and the dynamics of transcriptional activity is necessary.
Here we have devised an assay that uses a combination of genome editing, genetics, and live single-cell imaging to visualize the relationship between enhancer activation of transcription and physical proximity in real time. We show that activation of transcription can only occur when the enhancer comes into close proximity with its regulatory target and that transcription ceases almost immediately when they move apart. Moreover, we find that proximity alone is not sufficient for enhancer action; rather enhancer action requires a direct engagement of the enhancer with its target promoter. Our results suggest that establishment of physical interactions between enhancers and promoters can be the key rate-limiting step in gene regulation over long distances, and we can exclude mechanisms of transient enhancer–promoter associations leading to stable transcriptional activity.
II. RESULTS
Genetic Design: Homie-dependent long-distance regulation
To determine whether physical proximity is central to enhancer–promoter communication we have taken advantage of a characteristic property of boundary (insulator) elements in flies, namely their ability to pair with themselves, often over large genomic distances. For this purpose we selected a boundary called, homie, which marks the 3 end of the even-skipped (eve) locus (Fujioka et al., 2013; Fujioka et al., 2009). Genetic studies have shown that homie–homie self-pairing interactions can orchestrate enhancer activation of a reporter at distances of at least 2 Mb (Fujioka et al., 2013).
In our experimental system a transgene consisting of the eve promoter (no enhancers) and the lacZ coding sequence is inserted at an attP site located 142 kb upstream of the eve gene (Fig. 1A). When homie is included in this reporter transgene, physical interactions between it and the homie boundary at the endogenous eve locus can be detected in chromosome conformation capture experiments (Fujioka et al., 2016). In fixed embryos sporadic expression of lacZ mRNA is observed solely within the limits of the endogenous eve stripes (Fig. 1B and 1C). These findings indicate that the activation of the lacZ reporter depends on the enhancers in the eve locus 142 kb away, and that at this stage in development, the promoter of the reporter has no spontaneous activity nor does it respond to enhancers near the site of insertion.
Visualization of transcription and enhancer–promoter dynamics
Key to exploring the connection between enhancer action and physical enhancer–promoter proximity is being able to simultaneously visualize the location of the promoter, the location of the enhancers, and the transcriptional activity in living embryos. For this purpose we introduced tags into the endogenous eve gene and the lacZ reporter transgene at −142 kb (Fig. 1A). First, we used genome editing to insert an MS2 stem loop cassette (Bertrand et al., 1998; Larson et al., 2011; Yunger et al., 2010) into the 1st intron of the eve gene. Maternally expressed MS2 coat protein (MCP) fused to a blue fluorescent protein was used to visualize not only nascent eve transcripts but also mark the nuclear location of the eve gene and the associated eve enhancers. Second, to visualize transcriptional activity of the lacZ reporter we placed a PP7 stem loop cassette (Fukaya et al., 2016; Hocine et al., 2013) near the 5 end of the lacZ coding sequence. Nascent transcripts expressed from the reporter can then be visualized by the binding of maternally expressed PP7 coat proteins (PCP) fused to a red fluorescent protein (Fig. 2A). Finally, to mark the position of the lacZ reporter independent of whether the reporter is active, we used the recently developed parS/parB DNA labeling system (Dubarry et al., 2006; Saad et al., 2014). This system employs Burkholderia parS DNA sequences that nucleate the binding of a ParB-GFP fusion proteins, and it is thought to be less disruptive to the local chromatin structure than more traditional DNA labeling systems (Bystricky, 2015).
Using three-color time-lapse confocal microscopy we captured stacks of optical sections of the surface of two-hours old embryos carrying the tagged eve locus and the parS-homie-lacZ reporter. In these stacks we can clearly identify individual fluorescent foci in 70-100 nuclei simultaneously (Fig. 2B). In the blue channel we observed the endogenous activity and transcriptional dynamics of the eve gene in its characteristic seven-striped pattern. This pattern is quantitatively identical to that observed from the endogenous eve gene (Fig. S1). In the green channel we saw the parB foci in all of the nuclei in the embryo, which trace the movement of the lacZ reporter within the nucleus. Finally, in the red channel we observed lacZ expression in a subset of nuclei in the (blue) eve stripes. Consistent with the results from our fixed embryos, lacZ expression is restricted to nuclei that reside within one of the seven eve stripes.
To test our ability to reliably and accurately measure the localization of the reporter and the eve gene we generated a synthetic construct (localization control) in which all three fluorescent proteins are localized within a genomic distance of 2.0 kb (Fig. S2A). By analyzing embryos carrying this construct, we were able to calibrate chromatic aberrations originating from the microscope and also estimate errors in measuring spot localization (Fig. S2B-H). We also tested whether our method of marking the location of the reporter introduces perturbations in the chromatin structure that would hinder our analysis. For this purpose we placed the parS sequence at different locations relative to the lacZ reporter (Fig. S3A). We also implemented the more traditional lacO /LacI system to mark the location of the reporter in the nucleus (Gasser, 2002; Sinclair et al., 2010). No significant difference in chromatin dynamics and transcription kinetics were observed when the parS tag was placed at different locations or replaced by the lacO tag (Fig. S3B-H). Based on these analyses, we conclude that our genomic labeling approach allows us to measure chromatin dynamics with an error of 180±6 nm (mean±SEM).
Our initial visualization of reporter activity revealed a close connection between transcription and physical proximity. In nuclei in which the reporter is inactive, the reporter is well separated from the eve gene. In this case the green focus, which marks the parS sequence in the reporter, and the blue focus, which tags the transcriptionally active eve gene are far from each other, and since there is no red focus the reporter is silent (Fig. 2C). In nuclei in which the lacZ reporter is ON, the parS-homie-lacZ transgene co-localizes with the eve gene. In these nuclei the blue, green and red fluorescent foci appear to be attached together (Fig. 2C). This difference is recovered in measurements of the instantaneous spatial distance between the blue and green foci. In nuclei where red reporter activity is present, the distance between the blue and green foci is significantly shorter than the blue-green distance in nuclei lacking the red focus throughout the interval of observation (Fig. 2D and Fig. S3B-C).
Spatial proximity is necessary for enhancer action
To provide a more detailed picture of how enhancer action is related to spatial proximity, we measured the physical separation between the eve gene (blue) and the lacZ reporter (green) in embryos carrying the experimental transgene, parS-homie-lacZ. We analyzed live images of 2528 nuclei across 35 individual embryos over a 30 min period in nuclear cycle 14 (Fig. 3A). We observed a bi-modal distribution for the time-averaged physical distance (root-mean-squared (RMS) distance, see Methods) that could be modeled as a mixture of two Gaussians (Fig. 3B) at 743±96 nm and at 361±75 nm, respectively (mean±STD).
To confirm that the linkage of the reporter to eve is dependent on homie we replaced the homie sequence with lambda DNA of the same length. We generated a parS-lambda-lacZ transgene and analyzed movies of 870 nuclei across 12 individual embryos, again for 30 min in nuclear cycle 14 (Fig. 3A, yellow). For this data set, the distribution of the MS2-parS (blue-green) RMS distance is uni-modal (Fig. 3C) with a mean at 743±92 nm (mean±STD). While the distance between this transgene and eve fluctuates during the 30 min interval, there is no instance in which sustained physical proximity is established.
When we further tested the effect of reversing the orientation of homie in the original transgene (Fujioka et al., 2016), such that the lacZ reporter is downstream of homie, we find that pairing still occurs, but the regulation of the reporter by the eve enhancers is disrupted. We analyzed movies of 761 nuclei across 10 embryos for this parS-homieR-lacZ transgene. As expected, the corresponding bi-modal distribution resembles that for the regular homie transgene and has two means at 739±104 nm and 334±63 nm(mean±STD), respectively (Fig. 3D).
We next scored the nuclei with respect to the transcriptional activity of the lacZ reporter. Strikingly, all of the parS-homie-lacZ transgene nuclei showing lacZ transcription (i.e. presence of red signal, N=192) have a physical separation that falls within the distribution corresponding to the bound (i.e. homie-linked) conformation (Fig. 3A and 3B). But there are no nuclei in the unbound conformation that also express lacZ. Taken together, the above results demonstrate that homie pairing creates a local chromatin conformation that is permissive to transcription events by ensuring physical proximity between the eve enhancers and the promoter of lacZ. Consistent with this view, none of the nuclei in control parS-lambda-lacZ embryos express lacZ. From these observations we conclude that the eve enhancers must be in close proximity to the lacZ promoter in order to activate transcription.
Necessity for sustained physical association
To assess the temporal relationship between enhancer–promoter proximity and transcriptional activation, we examine the dynamics of physical proximity and activation along our measurement window, in singe nuclei. All nuclei displaying a switch from OFF-to-ON (N=65) were aligned with respect to the time point when nascent transcripts could first be detected (Fig. 4A, see Methods). The data show a sharp switch in activity state with similar rates to those previously reported for nuclei exiting mitosis (Garcia et al., 2013). For the same data set, we then measured the mean distance between the green parS tag in the parS-Homie-lacZ transgene and the eve gene (blue) as a function of time by averaging across all OFF-to-ON time courses with the same alignment (Fig. 4B). We observed a continuous spatial convergence until the onset of transcription at which point the mean distance corresponds to an average separation of about 340 nm. These findings indicate that there is a close connection between the establishment of enhancer–promoter proximity and enhancer activation of transcription.
We also assessed the temporal relationship between the ON-to-OFF switch in transcription and the physical distance between the enhancer–promoter pair. In 42 nuclei that switch their transcriptional activity OFF during our observation window, a drop in transcriptional activity of the lacZ reporter is accompanied by an increase in the mean distance between the parS-Homie-lacZ transgene and the eve gene (Figs. 4C and 4D). Our experiments indicate that there is about a 4 min gap between the time when the parS-Homie-lacZ transgene first begins to detectably separate from the eve gene and a clear decline in transcriptional activity. Since our reporter gene is 5.5 kb in length and the measured RNA polymerase II (Pol II) elongation rate is 1.50.1 kb/min (Garcia et al., 2013), the major part of this delay must be due to engaged polymerases that continue to elongate after the parS-Homie-lacZ becomes unlinked from the eve enhancers. These results fit with a model in which sustained physical association of eve enhancers and the lacZ promoter is necessary for continuous initiation of transcription.
Physical enhancer–promoter engagement leads to distinct topological conformations
While we have demonstrated that enhancer action depends upon sustained physical proximity to the promoter of the lacZ reporter, the proximity generated by homie pairing alone is not sufficient. Among the 307 parS-homie-lacZ nuclei in which homie-pairing likely occurred (i.e. nuclei occupying the short-distance peak in the histogram in Fig. 3B), only 63% express lacZ-PP7 during the 30 min window of observation in n.c. 14. Notably, in the nuclei that show reporter activity the MS2-parS distance is 27±8 nm (mean±SEM) lower than the distance in the nuclei where the lacZ reporter is inactive, suggesting that an additional compaction is associated with reporter activation.
The insufficiency of close proximity is even clearer when we employ the parS-homieR-lacZ reporter. This reporter still facilitates homie-pairing, yet likely renders the lacZ and the eve enhancers on opposite sides of these paired boundary elements (Fig. 3D, Fujioka et al., 2016). Of the 54 nuclei in which the lacZ transgene is close to the eve locus, there are only three in which the eve enhancers activate reporter transcription, and only for brief periods of 3-5 min.
Is transcriptional activation associated with an additional step that promotes physical enhancer–promoter engagement? We have shown that homie pairing establishes a new topological conformation in which the parS-homie-lacZ transgene and the eve locus are in close proximity (Fig. 3B). However, because the eve enhancers are distributed within 20kb of the eve-MS2 locus (Fig. 5A), homie pairing does not necessarily yield a close proximity between eve enhancers and the lacZ reporter. To further examine this point we take advantage of the property that independent eve enhancers regulate individual stripes of the eve pattern along the embryo (Fig. 5A). Thus by examining nuclei from different stripes separately we are able to explore the topology of the locus under different activating enhancers.
We first considered the distance of the eve-MS2 gene relative to the parS tag (blue-green) in nuclei where homie pairing occurs but lacZ transcription is lacking (Red-OFF, Fig. 5B). We observed different distances in nuclei belonging to different stripes, i.e. a dependence of the spatial arrangement on the identity of the activating enhancers (Fig. 5C, Red-OFF). If an activating enhancer is physically engaged with the endogenous eve promoter, it co-localizes with the blue eve-MS2 focus. Thus the distance to the homie pair (green parS tag) should depend on the distance between the activating enhancer and the endogenous homie. Indeed, the distance between the eve gene and the parS tag of the inactive lacZ reporter in stripe 5 is shorter than that observed for nuclei in stripes 4/6 and 3/7, for which the enhancers are located farther away from the parS tag (Fig. 5C). Moreover, for the two enhancers that drive eve expression in a pair of stripes, 4/6 and 3/7, the distance between eve and the parS tag is closely matched within each stripe pair. Finally, for embryos with the parS-homieR-lacZ transgene the same trend is observed in nuclei with homie pairing (Fig. S4A). Taken together, these results argue strongly that the eve enhancers directly engage the endogenous eve promoter to activate transcription and that in each eve stripe a distinct topological conformation is adopted.
Next, if transcriptional activation entails physical engagement, in each stripe we expect that in nuclei where lacZ transcription is active (Red-ON, Fig. 5B) not only will the eve-MS2 gene be in close proximity to the activating enhancer but also the parS-homie-lacZ reporter. Indeed, in the nuclei with the active reporter a significant shortening of the distance between the eve enhancers and the lacZ promoter is observed in stripes 4/6 and 3/7 when lacZ is activated (Fig. 5C), which further argues for transcription associated compaction of the locus. Critically, there is no lacZ activity dependent shift apparent for stripe 5 nuclei as in these nuclei the eve stripe 5 enhancer is already in close proximity to the lacZ promoter.
Finally, a stripe-specific topology is also favored by observations in measurements of the activation probability of the promoter driving lacZ expression. If transcription of the homie-linked reporters is linked to enhancer engagement, a plausible expectation is that activation frequency would be distance dependent. To assess the effect of distance we determined the fraction of homie-linked nuclei in each stripe that express lacZ. As predicted, the fraction of transcriptionally active reporters decreases with increasing distance between the stripe enhancer and the lacZ promoter (Fig. 5D). Stripe 5 has the highest activation probability ( 80%), while stripes 3/7 have the lowest probability of enhancer engagement with the lacZ reporter. The fact that we observe a linearly decaying relationship is consistent with contact probability measurements (Lieberman-Aiden et al., 2009). Together, these results provide compelling evidence that transcriptional activation entails physical enhancer–promoter engagement and is thereby associated with stripe-specific topological compaction of the eve locus.
Promoter competition has phenotypic consequences
In our experiments, the eve stripe enhancer drives expression from two different eve promoters, one for the endogenous eve gene and the other for the lacZ reporter (Fig. 5B, Red-ON). If the activity of the stripe enhancers is limiting, the lacZ reporter will reduce transcription of the eve gene. To determine whether promoter competition occurs in our genomic setup we compared eve transcription (i.e. the intensity of the blue MS2 signal) in individual nuclei in which lacZ is active and nuclei in which lacZ is silent (Fig. 6A, see Methods). For each eve stripe, we measured a 5%-25% reduction in endogenous eve transcription in nuclei in which lacZ is also transcribed compared to nuclei in which lacZ is not transcribed. The average reduction per nucleus is highest for stripe 5, and lowest for stripes 3 and 7.
Eve is a primary pair-rule gene that is responsible for segment patterning. To determine whether the observed reduction in eve transcription has any phenotypic consequences, we crossed males carrying a tagless homie-lacZ transgene at −142 kb (Fig. 1A) to females heterozygous for a wild type eve gene and an eve deficiency (Df(2R)eve). eve is weakly haploinsufficient and 7% of +/Df(2R)eve flies display patterning defects in even-numbered abdominal segments. Consistent with the reduction in the level of eve nascent transcripts, the presence of the homie-lacZ transgene exacerbates eve haploinsufficiency (Fig. 6B-D). Altogether 29% of the Homie-lacZ/Df(2R)eve flies have abdominal defects which corresponds to a 4-fold increase (Fig. 6E). Taken together, the above results suggest that competition between two promoters at the transcriptional level in the early embryo has phenotypic consequences for patterning in the adult. These findings provide evidence that manipulating topological chromatin structures can interfere with developmental programs.
III. DISCUSSION
Despite extensive studies over more than three decades, many questions still remain regarding how enhancers communicate with their target promoters over large genomic distances. Critically, although recent FISH and 3C-based genomic experiments have provided extensive evidence supporting physical interactions for long-range enhancer–promoter communication (Deng et al., 2012; Fabre et al., 2015; Hnisz et al., 2016; Ji et al., 2016; Phillips-Cremins et al., 2013; Rao et al., 2014; Williamson et al., 2016), we still miss a dynamic picture that could disentangle cause from consequence and that could distinguish transient contact with formation of stable topological structures. Here, we have developed a live imaging approach on a well-studied endogenous locus that allows us to probe the dynamics of topological chromatin structures at the single-cell level and its impact on transcriptional activity. We show directly that sustained physical proximity is necessary for enhancer action, and distinct topologies are observed for different pairing of the promoter with an activating enhancer, even when enhancers are clustered together. Though all our results are based on a specific experimental system it seems likely that they will be more broadly applicable.
To examine long-range transcriptional activation we placed a reporter gene at 150 kb distance of the presumptive eve enhancers. Although such distance is unusually large for enhancer–promoter interactions in flies, in higher eukaryotes enhancers are known to function over comparable or greater distances (Sanyal et al., 2012). At such distance the chromatin fiber can display fast random movements, which creates an entropic difficulty for specific long-range chromatin interactions and thus a kinetic barrier for the establishment of a productive pre-initiation complex. Architectural proteins have been suggested to increase contact frequency between chromatin loci separated by large genomic distances and facilitated the formation of sustained physical proximity permitting functional enhancer–promoter interactions (Erokhin et al., 2011; Shen et al., 2012; Symmons et al., 2014). The subdivision of genomes to topologically associating domains (TADs) was further suggested to facilitate local (intra-domain) chromatin interactions and reduce the search space, as exemplified for VDJ recombination in the mammalian immune system (Lucas et al., 2014).
To facilitate the interaction between the enhancers and the reporter in our experimental setup, we thus employ the homie element in our transgene, thereby inducing a stable loop conformation with the endogenous homie element that demarcates the 3 end of the eve locus. This element was shown to be bound by a wide array of Drosophila insulator proteins in the blastoderm embryo, including Su(Hw), CP190, BEAF-32 and dCTCF (Negre et al., 2010). homie-homie pairing between the transgene and the eve locus provides spatial proximity that we find to be necessary to allow the enhancers to activate transcription of the reporter. When the homie element in the transgene is replaced by lambda DNA, transcriptional activity of the reporter gene is not observed. As this transgene should encounter the eve enhancers with roughly the same frequency as the transgenes carrying homie, it demonstrates that transient encounters are not sufficient to activate PolII transcription, and the various transcription factors associated with the eve enhancers and the eve promoter are unable to establish stable and/or productive enhancer–promoter contacts on their own.
Another example in which an architectural element is deployed de novo to stabilize long distance enhancer–promoter interactions is the variably occupied tissue specific dCTCF binding site in the Ultrabithorax intron just upstream of the abx/bx enhancer (Magbanua et al., 2015). More generally, TAD boundaries in mammalian genomes often harbor CTCF binding sites, which are believed to play a major role in genome organization. Intriguingly, these sites were recently shown to have a specific orientation (Guo et al., 2015; Rao et al., 2014). When we examine a reporter construct in which we inverted homie orientation we find that while pairing is still maintained, the frequency of transcriptional activation drops significantly. The cause is likely because the enhancers and the promoters are now found on opposite sides of the paired insulators (Fujioka et al., 2016). Thus, alteration to these binding sites can bare important functional implications and indeed many cancer types accumulate CTCF binding site mutations that modulate chromatin conformations with subsequent influences on transcription (Flavahan et al., 2016; Katainen et al., 2015). Our experimental setup provides an opportunity for exploring the dynamic relationship between topological alterations and transcriptional regulation resulting from mutations in boundary elements.
Overall, our experiments demonstrate that physical proximity is a necessary condition for the eve enhancers to activate lacZ expression. Importantly, we find that transcription ceases upon dissociation of the enhancers and the reporter, suggesting a requirement for sustained physical proximity, and excluding mechanisms of transient enhancer promoter associations leading to stable activity. While we observe this tight relationship between physical proximity and activation, for other genes, loop formation could be a permissive step, with pre-formed enhancer–promoter interactions preceding transcriptional activation (de Laat and Duboule, 2013; Ghavi-Helm et al., 2014; Jin et al., 2013; Montavon et al., 2011). This discrepancy might reflect regulatory strategies that are gene or developmental context dependent. Nevertheless, consistent with the observations that a large amount of reorganization in enhancer–promoter interactions exist upon tissue differentiation (Dixon et al., 2015; Javierre et al., 2016), our results show that establishment of physical interactions between enhancers and promoters could be a rate-limiting step for transcriptional activation and play a key role in regulating transcription and cell fate.
Applying our single-cell live imaging approach in the early fly embryo offers an unprecedented opportunity to examine topological chromatin structures of a locus in a cell-specific way. We provide compelling evidence that eve expressing cells adopt distinct topological chromatin structures according to the stripe-specific identity of the activating enhancer. Notably the eve enhancers are located within 10 kb of the endogenous eve promoter and the homie-paired promoter of the reporter gene. A possible model could be that the adjacent eve enhancers will be found in close proximity to the promoters within a local environment permitting any of the enhancers to activate either of the promoters without further topological distinctions. Our observation of a stripe-specific distance between the locus of the eve gene and the homie-tethered parS tag, and the further compaction of the locus upon concurrent activation of the reporter, seem to exclude such a model. But rather this seems to suggest the necessity of a specific enhancer–promoter engagement for transcriptional activation, even within a relatively short genomic range.
What is the spatial scale that defines enhancer–promoter engagement? Transcriptional activity in our experimental system is only observed when the reporter moves into close proximity to the eve locus. The transition from inactivate to active occurs when the parS tag comes within 350 nm of the eve gene. When the reporter moves more than 350 nm away from the eve gene, transcription shuts off, and already engaged Pol II bring their elongation process to completion. The range of the observed necessary physical proximity is likely to be dictated by the spatial scale set up by the size of the protein complexes that are believed to physically bridge enhancers with promoters (¿2MDa with diameter ¿20nm (Dill et al., 2011)), for example Mediator complex and the Pol II pre-initiation complex (Kagey et al., 2010; Plaschka et al., 2015; Robinson et al., 2016). This requirement of 350 nm in enhancer–promoter proximity in order to activate transcription is comparable with DNA FISH measurements in mammalian systems that estimate the proximity between the shh promoter and its enhancers in mouse limbs (Amano et al., 2009); (Williamson et al., 2016).
Finally, when the endogenous eve locus is presented with an extra ectopic promoter, its activation results in concomitant reduction in endogenous eve transcription, phenotypically leading to patterning defects in the corresponding segments. Usually, competing promoters are physically linked within short genomic distances (¡20kb) (Foley and Engel, 1992; Fukaya et al., 2016). However, our results show that physically associating the competing promoters could in principle also be achieved through dynamic alterations of chromatin conformations. This observation reinforces the notion that rearrangement of topological chromatin structures is able to re-wire enhancer–promoter interactions, and that the rewiring can cause new phenotypes in many developmental contexts. For example in humans, structural chromosome variants that disrupt boundary elements flanking TADs may lead to diseases (Franke et al., 2016; Lupianez et al., 2015). Thus our experiments provide a potential path to synthetically rewire enhancer–promoter interactions to perturb chromatin topologies and thereby interfere with developmental programs (Deng et al., 2012; Deng et al., 2014),
In conclusion, we propose a methodological framework for investigating the relationship between chromatin dynamics and long-range enhancer-mediated transcriptional regulation. With the ability of tracing endogenous loci in individual cells and simultaneously measuring temporal dynamics and spatial information, new mechanistic insights into enhancer–promoter interactions are likely to be uncovered.
Methods Summary
Plasmid construction
Transgenic fly generation
Fluorescent in-situ hybridization
Phenotypic scoring
Microscopy and imaging conditions
Image processing and data analysis
Acknowledgements
We thank P. Schedl for bringing the eve-homie system to our attention. We thank K. Bystricky for introducing us to the ParB/ParS system, and F. Payre and P. Valenti for sharing a ParB-eGFP plasmid and the parS sequence. We thank P. Ratchasanmuang and S. Ryabichko for assistance with cloning and fly husbandry, and the Bloomington Drosophila Stock Center for fly strains. We are also grateful to L. Barinov, S. Blythe, M. Levine, M. Levo, P. Schedl, E.F. Wieschaus for discussion and help with the manuscript. This study was funded by grants from the National Institutes of Health (U01 EB021239, R01 GM097275, and R01 GM050231). HC was supported by the Charles H. Revson Biomedical Science Fellowship.