top of page


How Enhancers Work

During Development

An Enhancer is a piece of DNA that tells the gene when and where in the embryo to turn on or off. While we know a lot about genes, we know very little about enhancers. Without a sufficient understanding of how enhancers work, we don't know how genes are controlled or how whole genetic systems work. The goal of our research is to understand How Enhancers Work, the lingering mystery of Genetics.

Our Working Model for Enhancers

Enhancer big model.png

We view an enhancer as harbouring a microprocessor (See Figure above): Based on inputs it receives from other genes, the microprocessor activates or deactivates the gene it controls. An enhancer also harbours a knob that tunes the intensity of gene activity. In addition, based on our previous work (Research Direction 1 below),  we posit that an enhancer (or a system of enhancers) harbours a second knob that tunes the speed of gene regulation (i.e. how fast the gene responds to changes to its inputs). In our lab, we devise novel working models for how each of these enhancer components (the microprocessor and the two tuning knobs) works, and present novel experimental frameworks to test each model. These models have the potential of changing our understanding of how enhancers work, how to decipher the function of specific enhancers, and how to synthetically design new ones.

Research Direction 1

The Speed Knob: How do enhancers control the speed of gene regulation?


How do enhancers regulate wavelike gene expression patterns?

A common problem in development is how to partition a tissue into domains of different identities. The most popular mechanism for how to do so is the ‘French Flag’ model, where thresholds of a morphogen gradient sets the boundaries between different embryonic fates within a tissue. The French Flag model explains well the patterning of the anterior-posterior (AP) axis of the fruit fly Drosophila, hence its popularity. However, studying various organisms showed that pattern formation is oftentimes mediated by dynamic waves of gene expression, a fact that cannot be explained by the French Flag model. Based on our experimental studies of AP patterning in another insect, the beetle Tribolium castaneum, we developed a new model of pattern formation: the ‘Speed Regulation’ model. In this model, a morphogen gradient modulates the ‘speed’ of gene regulation within gene regulatory networks (i.e. how fast genes respond to changes in their inputs), which results in the induction of gene expression waves.  This mechanism seems to be general and explains experimental observations in a wide range of tissues and species; however, its molecular underpinnings is currently unknown. Our goal in Research Direction 1 is to understand how the speed of gene regulation is modulated at the molecular/enhancers level. To do so, we start with a working model and develop strategies to test/modify it.

Working Model

Much like a car is controlled by two pedals (gas and brake), our model posits that a gene is regulated by two enhancers: one ‘dynamic’ that induces rapid changes in the gene output (equivalent to a car’s gas pedal), and another ‘static’ that stabilizes gene activity (equivalent to a car’s brake pedal). By modulating the balance between the potency of dynamic vs static enhancers, the tuning of speed of gene regulation is achieved. The model was inspired by our study of how multiple enhancers regulate the same gene expression domain in the early Drosophila embryo. Using computational modelling, the suggested scheme was shown to be logically sound; however, it awaits experimental verification.

gene expression waves.png
enhancer switching model.png
mathmematical model.png

​​Experimental Approaches

(1) Enhancer discovery and activity analysis using open chromatin genomic assays and enhancer reporter assays in Tribolium

triboium atac seq experiments.png

​(2) Live Imaging using the MS2-MCP system in Tribolium

run6 kymo.png

Research Direction 2

The Intensity Knob: How do enhancers control the intensity of gene activity?

Especially focusing on:

The role of enhancer RNAs (eRNAs) in Transcription

In higher eukaryotes, an enhancer does not directly mediate gene activity, but rather switches on the gene’s promoter, which acts like a valve that allows for pumping Pol II molecules along the gene, producing mRNAs (Figure below). There is currently no satisfactory model for how this happens, and consequently, how enhancers control the intensity of gene activity.

traditional transcription model.png

However, four recent discoveries, albeit puzzling, suggest a path forward:

(1) Transcription happens in bursts: the gene’s valve repeatedly opens and closes. It is currently unknown why and how these bursts are generated.

(2) Pol II molecules form clusters that bind the gene very briefly, before they quickly disperse. Nonetheless, transcription persists long after the Pol II cluster has already vanished, and it is currently unknown where the rest of Pol II molecules comes from.

(3) Transcription does not only happen at the promoter, but also at specific sequences within (and sometimes outside) the enhancer that have promoter-like motifs (hereafter called ‘e-promoters’). Whereas transcription at the promoter produces long stable mRNAs that code for proteins, transcription at e-promoters produces short unstable and non-coding ‘enhancer RNAs’ (eRNAs). Transcription of eRNA has been shown to influence gene activity, although the mechanism by which this happens remains unknown.

(4) Gene promoters do not only loop and bind enhancers, but also scans and binds other DNA sequences around and beyond the enhancer.

It is currently unclear how these observations could fit into a model of transcription. In Research Direction 2, we present a hypothesis that brings these facts together in a coherent model, and propose a novel experimental approach to test/modify it.

Working model

In our model, the arrival of Pol II clusters to the gene locus is episodic and scarce (Figure below). As many Pol II molecules as possible should then be stored upon the arrival of a Pol II cluster to support a steady flow of Pol II molecules along the gene. We posit this happens by utilizing not only the main valve of the gene (the gene’s main promoter), but using several other smaller valves (e-promoters) that store Pol II by keeping them busy transcribing eRNAs. In this way, e-promoters keep a stock of Pol II molecules so that when a Pol II cluster vanishes, Pol II molecules at e-promoters are recycled at the gene’s main promoter and used to support a steady flow of Pol II molecules along the gene. To achieve this, the promoter scans the whole gene locus and dwells at e-promoters to load Pol II molecules. Eventually, the promoter runs out of Pol II molecules before the next Pol II cluster is recruited, and mRNA production ends;  hence, transcription appears to happen in bursts. According to this model, transcription intensity is modulated by varying the frequency of Pol II cluster recruitment, and the efficiency of the gene’s promoter and e-promoters in anchoring Pol II molecules to DNA.

my transcription model.png

​​Experimental Approaches

(1) Studying eRNA transcription landscapes using imaging in the early Drosophila embryo

endogeneous eRNA.jpg
kr locus erna sketch.png

​(2) Assaying the effect of sequestering e-promoters (by deleting them or blocking them using insulators) in live embryos using the MS2-MCP system in the early Drosophila embryo

RD 3

Research Direction 3

The Microprocessor: How do enhancers make decision?


An enhancer can be seen as a microprocessor with certain computational logic that takes the activities of other genes (usually TFs) as inputs and delivers a decision as to turn on or off the gene. The traditional view is that this logic is set by arranging binding sites within the enhancer to recruit specific TFs, where each TF has a specific influence on transcription (either activating or repressing). However, taking this view as a design strategy for synthetic enhancers often fails: simple concatenation of individual TF binding sites does not automatically result in a functional enhancer. Recent studies are beginning to reveal as why this is the case: TFs seem to interact with each other and bind enhancers together in ‘collectives’, where the influence of one TF depends on other TFs in the collective. This means that there is no way that we can understand enhancer logic based on the isolated influence of each TF, and that enhancer logic should be studied case by case using time-consuming experiments. This is a serious problem, since there are several thousands of enhancers in the genome of a typical animal. Here we present a hypothesis that suggests a scalable strategy to decipher enhancer logic and design new ones.


Here we posit that TF collectives are evolutionary conserved, difficult to modify, and mediate specific transcriptional logic. Furthermore, since the evolution of multiple protein-protein and protein-DNA interactions are required to form a new TF collective with a new logic, we argue that TF collectives are actually limited in number, and new transcriptional logics are created by recruiting one or more of these fixed set of TF collectives. This means that the enhancer’s microprocessor is built by combining a set of standard computational modules that have been optimized over long evolutionary times. This suggests that one can decipher the enhancer logic of large number of enhancers by identifying just a limited number of TF collectives and the regulatory logic they mediate. Creating a synthetic enhancer will be then reduced to combining different ‘enhancer submodules’ each recruiting different TF collectives, where their additive activities mediate the required transcriptional logic.



Experimental approach


Since one enhancer might recruit several TF collectives, ChIP-seq analysis is unsuitable to disentangle which TF belongs to which TF collective. Furthermore, performing ChIP-seq for all TFs that regulate a single process (which are oftentimes plenty) is prohibitively laborious and expensive. Excitingly, recent data suggest that it is possible to detect TF collectives using imaging, as TFs were found to bind DNA in large clusters, detectable using imaging. These clusters are very dynamic and disperse within seconds after their formation. Hence, the co-localization of two different TF clusters strongly indicates that they bind together to enhancers. Indeed, using super-resolution STED microscopy, we found that many TFs involved in the early patterning of the Drosophila embryo co-localize at frequencies much higher than chance (Figure above), as has been previously shown for other TFs. TF collectives, however, are expected to be composed of more than two TFs, and hence, here we use the multiplexed visualization system to catalogue the set of TF collectives involved in patterning the AP axis of the early Drosophila embryo, and document how often the same TF collective is used at different enhancers. By observing which TF collectives bind which enhancers in which cell in both WT and mutant embryos, we then decipher the regulatory logic of each TF collective using computational modelling. This will be followed by an enhancer dissection step to find minimal enhancer submodules that recruit single TF collectives. Finally, we use synthetic biology approaches create new transcriptional logic by combining various enhancer submodules.

bottom of page