Caroline Hill: Projects

The dynamics of TGF-β signalling

Understanding how cells respond to ligands in complex physiological and pathological contexts in vivo requires a knowledge of the dynamics of signalling in addition to an understanding of the molecular wiring of the pathway. We have addressed this by using a combined experimental and computational modelling approach to study TGF-β signalling dynamics. In particular, we have dissected the mechanisms that determine cellular responsiveness to acute and chronic signals, as well as to repeated ligand stimulations. We have shown that stimulation of cells with TGF-β produces a transient response that attenuates over time, resulting in desensitised cells that are refractory to further acute stimulation.

This loss of signalling competence depends on ligand binding, but not on receptor activity, and is only reverted after the ligand has been depleted. Using surface receptor biotinylation experiments we have shown that TGF-β binding triggers the rapid depletion from the cell surface of signalling-competent receptors, with the type I and type II receptors exhibiting different degradation and trafficking kinetics.

In collaboration with Bernhard Schmierer (Karolinska Institute, Stockholm) we have gone on to generate a computational model of TGF-β signal transduction from the membrane to the nucleus that incorporates this mechanism. We have used the model to predict which parameters are critical determinants of the dynamics of TGF-β signalling, and find that the rate of receptor turnover and the ratio between ligand-induced and constitutive receptor degradation are crucial, but not the initial levels of surface receptors. The model also predicts that autocrine signalling, as demonstrated in many tumours, would severely compromise TGF-β responses, and we have been able to confirm this experimentally using a panel of tumour cell lines (Vizán et al., 2013; Science Signalling. 6(305):ra106).

We are now extending this work to other TGF-β superfamily members and have performed several whole genome high throughput screens to find new regulators of TGF-β signalling dynamics.

A new branch of TGF-β signalling through Smad1/5

Several years ago we made the unexpected discovery that TGF-β induces phosphorylation of Smad1/5 in the majority of epithelial cells, cancer cell lines and fibroblasts that we have tested, in addition to inducing Smad2/3 phosphorylation.

We want to establish the function of this branch of TGF-β/Smad signalling and determine its importance in tumour development and dissemination.

We have shown that this pathway results in the formation of phosphorylated Smad1/5-Smad4 complexes as well as complexes between phosphorylated Smad1/5 and phosphorylated Smad2/3.

Using ChIP-seq we have identifed the genomic binding sites of the Smad1/5-Smad4 complexes in response to TGF-β which has led us to identify the target genes activated.

We are now using CRISPR-Cas9 technology to knockout this pathway in cancer cells, which we will then use in in vivo tumourigenic assays to elucidate the importance of this branch of TGF-β signalling in cancer.

Regulation of transcription by activated Smad complexes

The main function of the TGF-β superfamily/Smad pathways is to induce new programmes of gene expression, but apart from previous work in my lab showing that Smads activate transcription by remodelling the chromatin template, little is understood about how exactly they achieve this.

Our work in this area is focused on Nodal signalling, which is not only critical for embryonic development, and for maintaining pluripotency of human ES cells, but is exploited by cancer cells to promote tumour progression and metastasis.

We are using the embryonic carcinoma cell line, P19 which respond to Nodal and Activin both acutely and chronically as a model system. As well as being a model for cancer, these cells express a combination of pluripotency factors and mesendoderm markers in response to Activin/Nodal.

To characterise the transcriptional responses to Activin we performed RNA-seq on P19 in the non-signalling state, in cells treated with Activin for short (1 h) or extended (8 h) times, and in the untreated state (autocrine signalling). This has allowed us to define four main classes of response genes (transiently induced; induced sustained; delayed induced; repressed), which is enabling us to define different classes of enhancers that are co-regulated.

In the same conditions we have also performed ChIP-seq analysis for Smad2, for total histone H3, for two different histone modifications characteristic of active chromatin (H3K27ac and H3K9ac), and for two different forms of RNA Polymerase II (initiating state phosphorylated at Ser 5, and elongating state phosphorylated at Ser 2).

On-going work integrating the RNA-seq data with all of these ChIP-seq datasets is revealing exciting and novel insights into the mechanism whereby activated Smad2-containing complexes find their targets in chromatin and activate transcription.

Motif enrichment analysis is being used to define what other cofactors bind with Smad2 at enhancers of Activin-responsive genes, distinguishing between those that bind with Smad2 1 h after Activin stimulation and those that are synthesised in response to Activin and bind with Smad2 at later time points. This is yielding a number of interesting candidates which are being experimentally verified.

Spatial regulation of BMP signalling

To understand how TGF-β superfamily ligands function in vivo we use early zebrafish embryos as a model system. We are particularly interested in how these ligands are regulated, how they function in a dose-dependent manner and how they contribute to tissue specification. This year we have been very successful in understanding how BMP signalling is regulated and how it contributes to tissue patterning at the neural plate border.

A key tool for this work has been a transgenic zebrafish line that expresses an mRFP reporter driven by BMP responsive elements. We have determined the mechanism whereby the ventral to dorsal BMP gradient is set up in early zebrafish embryos and show that it occurs as a result of graded transcription of the ligands, with little or no diffusion of the ligands (Ramel and Hill, 2013; Dev Biol. 378(2): 170-82). This is in contrast to well known mechanisms of BMP gradient formation determined in Drosophila which involve either extensive diffusion, or shuttling of the BMP proteins from their site of production.

We have gone on to determine how BMP signalling is remodelled post-gastrulation and have shown that the ventral to dorsal gradient evolves into two distinct stripes at the neural plate border: one coinciding with the neural crest, and the other abutting the epidermis. In between is a region devoid of BMP activity, which is specified as preplacodal ectoderm (PPE). We have identified the BMP ligands involved and determined the underlying mechanism (Figure 1).

We have demonstrated a key role for the secreted BMP regulator, Crossveinless 2, as a factor essential to concentrate BMP activity in the stripes, and show that the cell autonomous BMP inhibitor, BAMBI-b is unregulated in the PPE under the control of the Distalless transcription factors Dlx3b/4b, and is an excellent candidate to inhibit BMP signalling in this domain (Reichert et al., 2013; Development. 140(21): 4435-44).

Figure 1

Figure 1: Tissue patterning at the neural plate border in zebrafish embryos. Upper panels: Double fluorescent in situs (DFISH) showing the expression pattern in four somite stage zebrafish embryos of cvl2, dlx3b and n-cadherin (n-cad). cvl2 marks the BMP activity domains, the inner of which is the neural crest; dlx3b marks the PPE; n-cad marks the neural plate. ncad and dlx3b have been labelled with the same colour to allow visualisation of three different mRNAs. The view is dorsal, with the anterior to the top right. Lower panels: DFISH for cvl2 and dlx3b showing how the expression domains evolve over time from late gastrulation (95% epiboly) to the one somite stage. Lateral views with dorsal to the right. V, ventral; D, dorsal; A, anterior; P, posterior; PPE, preplacodal ectoderm; NP, neural plate (Click to view larger image)

Spatial regulation of Nodal signalling

To understand how TGF-β superfamily ligands function in vivo we are using early zebrafish embryos as a model system. We want to determine how ligand activity is regulated, how the ligands function in a dose-dependent manner and how they contribute to tissue specification.

This year we have been focusing on spatial and temporal control of Nodal signalling. To visualise Nodal signalling in vivo, we generated a transgenic zebrafish line in which an EGFP reporter is controlled by 3 copies of the Activin-responsive enhancer (ARE), which binds a complex of activated Smad2-Smad4 with the transcription factor FoxH1.

Using EGFP mRNA as a readout we can track active Nodal signalling in a developing embryo, and have made some surprising discoveries. Shortly after Nodal signalling is initiated, it is visible in the embryo margin in a shallow dorsal-ventral gradient, but contrary to long standing assumptions in the field, it extends only four or five cell diameters towards the animal pole, coincident with the expression domains of cyclops and squint (the fish Nodal ligands), and lefty1/2 (the antagonists), and is not graded in this direction.

The predominant model to explain the generation of domains of Nodal signalling is the reaction-diffusion model, which is based on the assumption that a highly diffusible inhibitor (in this case Lefty1/2) and a less diffusible activator (Nodal ligands) can create a network as a result of short-range activation and long-range inhibition. Our findings do not support this model. Instead, our results are leading us to an alternative model, whereby a temporal delay in the translation of the ligand antagonists Lefty1/2 allows Nodal signalling to become established in four to five cell tiers at the margin, at which point Lefty protein levels reach a sufficiently high threshold to prevent further spread of Nodal signalling. We are determining the mechanism underlying this behaviour. Our data is suggesting that rather than a spatial gradient of Nodal signalling in the early embryo, the gradient may be temporal.

A new class of small RNA involved in gene regulation

Work in the lab on spatial regulation of Nodal signalling in Xenopus embryos led us to a large-scale next generation sequencing project to define all the small RNAs in early Xenopus tropicalis embryos. We identified a number of new miRNAs, which are currently informing our work on spatial and temporal regulation of TGF-β superfamily ligands in vertebrate development. However, most strikingly we found that the vast majority of small RNAs sequenced were previously unannotated.

By combining the small RNA sequencing with mRNA profiling, we identified a new class of small non-coding RNAs, which align in clusters to introns of specific protein-coding genes. We show that these small non-coding RNAs are derived from remnants of transposable elements that are present in introns, and have named them siteRNAs for small intronic transposable element RNAs to reflect this.

We find that genes containing clusters of siteRNA are transcriptionally repressed as compared with all genes. Furthermore, we show that this is true for individual genes containing siteRNA clusters, and moreover, these genes are enriched in specific repressive histone modifications.

These data suggest a new mechanism of siteRNA-mediated gene silencing in vertebrates, and provide an example of how mobile elements can affect gene regulation (Harding et al., 2014; Genome Res. 24(1):96-106).

We are now extending this work to the zebrafish system.

Caroline Hill
+44 (0)20 379 61251

  • Qualifications and history
  • 1989 PhD, Cambridge University, UK
  • 1991 Postdoctoral Research Fellow, Imperial Cancer Research Fund, UK
  • 1995 Established lab at Ludwig Institute for Cancer Research, UCL Branch, UK
  • 1998 Moved lab to the Imperial Cancer Research Fund, UK (in 2002 the Imperial Cancer Research Fund became Cancer Research UK)
  • 2015 Group Leader, the Francis Crick Institute, London, UK