Uncovering tissue-specific binding features from differential deep learning.

Transcription factors (TFs) can bind DNA in a cooperative manner, enabling a mutual increase in occupancy. Through this type of interaction, alternative binding sites can be preferentially bound in different tissues to regulate tissue-specific expression programmes. Recently, deep learning models have become state-of-the-art in various pattern analysis tasks, including applications in the field of genomics. We therefore investigate the application of convolutional neural network (CNN) models to the discovery of sequence features determining cooperative and differential TF binding across tissues. We analyse ChIP-seq data from MEIS, TFs which are broadly expressed across mouse branchial arches, and HOXA2, which is expressed in the second and more posterior branchial arches. By developing models predictive of MEIS differential binding in all three tissues, we are able to accurately predict HOXA2 co-binding sites. We evaluate transfer-like and multitask approaches to regularizing the high-dimensional classification task with a larger regression dataset, allowing for the creation of deeper and more accurate models. We test the performance of perturbation and gradient-based attribution methods in identifying the HOXA2 sites from differential MEIS data. Our results show that deep regularized models significantly outperform shallow CNNs as well as k-mer methods in the discovery of tissue-specific sites bound in vivo.

Cdx is crucial for the timing mechanism driving colinear Hox activation and defines a trunk segment in the Hox cluster topology.

Cdx and Hox transcription factors are important regulators of axial patterning and are required for tissue generation along the vertebrate body axis. Cdx genes have been demonstrated to act upstream of Hox genes in midgestation embryos. Here, we investigate the role of Cdx transcription factors in the gradual colinear activation of the Hox clusters. We found that Hox temporally colinear expression is severely affected in epiblast stem cells derived from Cdx null embryos. We demonstrate that after initiation of 3' Hox gene transcription, Cdx activity is crucial for H3K27ac deposition and for accessibility of cis-regulatory elements around the central - or 'trunk' - Hox genes. We thereby identify a Cdx-responsive segment of HoxA, immediately 5' to the recently defined regulatory domain orchestrating initial transcription of the first Hox gene. We propose that this partition of HoxA into a Wnt-driven 3' part and the newly found Cdx-dependent middle segment of the cluster, forms a structural fundament of Hox colinearity of expression. Subsequently to initial Wnt-induced activation of 3' Hox genes, Cdx transcription factors would act as crucial effectors for activating central Hox genes, until the last gene of the cluster arrests the process.

Cdx and T Brachyury Co-activate Growth Signaling in the Embryonic Axial Progenitor Niche.

In vertebrate embryos, anterior tissues are generated early, followed by the other axial structures that emerge sequentially from a posterior growth zone. The genetic network driving posterior axial elongation in mice, and its disturbance in mutants with posterior truncation, is not yet fully understood. Here, we show that the combined expression of Cdx2 and T Brachyury is essential to establish the core signature of posterior axial progenitors. Cdx2 and T Brachyury are required for extension of a similar trunk portion of the axis. Simultaneous loss of function of these two genes disrupts axial elongation to a much greater extent than each single mutation alone. We identify and validate common targets for Cdx2 and T Brachyury in vivo, including Wnt and Fgf pathway components active in the axial progenitor niche. Our data demonstrate that integration of the Cdx/Hox and T Brachyury transcriptional networks controls differential axial growth during vertebrate trunk elongation.

Polarized regulatory landscape and Wnt responsiveness underlie Hox activation in embryos.

Sequential 3'-to-5' activation of the Hox gene clusters in early embryos is a most fascinating issue in developmental biology. Neither the trigger nor the regulatory elements involved in the transcriptional initiation of the 3'-most Hox genes have been unraveled in any organism. We demonstrate that a series of enhancers, some of which are Wnt-dependent, is located within a HoxA 3' subtopologically associated domain (subTAD). This subTAD forms the structural basis for multiple layers of 3'-polarized features, including DNA accessibility and enhancer activation. Deletion of the cassette of Wnt-dependent enhancers proves its crucial role in initial transcription of HoxA at the 3' side of the cluster.

Diabetes Inhibits Gr-1+ Myeloid Cell Maturation via Cebpa Deregulation.

Recruitment of innate immune cells from the bone marrow (BM) to an injury site is required for effective repair. In diabetes, this process is altered, leading to excessive recruitment and retention of dysfunctional myeloid cells that fail to promote angiogenesis, prolong inflammation, and block healing. The aberrant myeloid phenotype is partially mediated by stable intrinsic changes to developing cells in the BM that are induced by the diabetic (db) environment, but the exact mechanisms remain largely unknown. Here, we show that the db-derived Gr-1(+)CD11b(+) immature myeloid population has widespread misexpression of chromatin-remodeling enzymes and myeloid differentiation factors. Crucially, diabetes represses transcription of the key myeloid transcription factor CEBPA via diminished H3 Lys 27 promoter acetylation, leading to a failure in monocyte and granulocyte maturation. Restoring Cebpa expression by granulocyte colony-stimulating factor reverses the db phenotype and rescues myeloid maturation. Importantly, our data demonstrate a possible link between myeloid cell maturation and chronic inflammation.

Hoxa2 selectively enhances Meis binding to change a branchial arch ground state.

Hox transcription factors (TFs) are essential for vertebrate development, but how these evolutionary conserved proteins function in vivo remains unclear. Because Hox proteins have notoriously low binding specificity, they are believed to bind with cofactors, mainly homeodomain TFs Pbx and Meis, to select their specific targets. We mapped binding of Meis, Pbx, and Hoxa2 in the branchial arches, a series of segments in the developing vertebrate head. Meis occupancy is largely similar in Hox-positive and -negative arches. Hoxa2, which specifies second arch (IIBA) identity, recognizes a subset of Meis prebound sites that contain Hox motifs. Importantly, at these sites Meis binding is strongly increased. This enhanced Meis binding coincides with active enhancers, which are linked to genes highly expressed in the IIBA and regulated by Hoxa2. These findings show that Hoxa2 operates as a tissue-specific cofactor, enhancing Meis binding to specific sites that provide the IIBA with its anatomical identity.

Chromatin immunoprecipitation and chromatin immunoprecipitation with massively parallel sequencing on mouse embryonic tissue.

Regulation of gene expression must be tightly controlled during embryonic development. A central mechanism to control gene expression is the binding of sequence-specific transcription factors to cis-regulatory elements in the genome. Chromatin immunoprecipitation (ChIP) is a widely used technique to analyze binding of transcription factors and histone modifications on chromatin; however, it is limited to looking at a small number of genes. ChIP with massively parallel sequencing (ChIP-seq) is a recently developed powerful tool to analyze genome-wide binding of transcription factors and histone modifications and provides a vast amount of information into the regulation of gene expression. This chapter describes how ChIP and ChIP-seq are performed on mouse embryonic tissue.

Mouse Hoxa2 mutations provide a model for microtia and auricle duplication.

External ear abnormalities are frequent in newborns ranging from microtia to partial auricle duplication. Little is known about the molecular mechanisms orchestrating external ear morphogenesis. In humans, HOXA2 partial loss of function induces a bilateral microtia associated with an abnormal shape of the auricle. In mice, Hoxa2 inactivation at early gestational stages results in external auditory canal (EAC) duplication and absence of the auricle, whereas its late inactivation results in a hypomorphic auricle, mimicking the human HOXA2 mutant condition. By genetic fate mapping we found that the mouse auricle (or pinna) derives from the Hoxa2-expressing neural crest-derived mesenchyme of the second pharyngeal arch, and not from a composite of first and second arch mesenchyme as previously proposed based on morphological observation of human embryos. Moreover, the mouse EAC is entirely lined by Hoxa2-negative first arch mesenchyme and does not develop at the first pharyngeal cleft, as previously assumed. Conditional ectopic Hoxa2 expression in first arch neural crest is sufficient to induce a complete duplication of the pinna and a loss of the EAC, suggesting transformation of the first arch neural crest-derived mesenchyme lining the EAC into an ectopic pinna. Hoxa2 partly controls the morphogenesis of the pinna through the BMP signalling pathway and expression of Eya1, which in humans is involved in branchio-oto-renal syndrome. Thus, Hoxa2 loss- and gain-of-function approaches in mice provide a suitable model to investigate the molecular aetiology of microtia and auricle duplication.

Differential distribution of the Ca (2+) regulator Pcp4 in the branchial arches is regulated by Hoxa2.

Branchial arches are externally visible tissue bands in the head region of all vertebrate embryos. Although initially formed from similar components, each arch will give rise to different head and neck structures. In a screen designed to characterize the molecular control of branchial arch identity in mouse, we identified Pcp4 as a second branchial arch-specific molecular signature. We further show that the transcription factor Hoxa2 binds to Pcp4 chromatin and regulates Pcp4 expression in the second arch. Hoxa2 is also sufficient to induce Pcp4 expression in anterior first arch cells, which are Pcp4-negative.

Histone methyltransferase ASH1 orchestrates fibrogenic gene transcription during myofibroblast transdifferentiation.

UNLABELLED: Transdifferentiation of hepatic stellate cells (HSCs) to a myofibroblast-like phenotype is the pivotal event in liver fibrosis. The dramatic change in phenotype associated with transdifferentiation is underpinned by a global change in gene expression. Orchestrated changes in gene expression take place at the level of chromatin packaging which is regulated by enzymatic activity of epigenetic regulators that in turn affect histone modifications. Using expression profiling of epigenetic regulators in quiescent and activated primary HSCs we found a number of histone methyltransferases including MLL1, MLL5, Set1 and ASH1 to be highly up-regulated during transdifferentiation of HSCs. All of these histone methyltransferases regulate methylation of lysine 4 of histone H3, which is a signature of actively transcribed genes. We therefore postulated that one or more of these enzymes may be involved in positively influencing expression of profibrogenic genes. CONCLUSION: We find that ASH1 directly binds to the regulatory regions of alpha smooth muscle actin (αSMA), collagen I, tissue inhibitor of metalloproteinase-1 (TIMP1) and transforming growth factor beta1 (TGFß1) in activated HSCs while depletion of ASH1 caused broad suppression of fibrogenic gene expression. We also discovered that MeCP2 positively regulates ASH1 expression and therefore identify ASH1 as a key transcriptional activator component of the MeCP2 epigenetic relay pathway that orchestrates coordinated induction of multiple profibrogenic genes.

Genome-wide occupancy links Hoxa2 to Wnt-ß-catenin signaling in mouse embryonic development.

The regulation of gene expression is central to developmental programs and largely depends on the binding of sequence-specific transcription factors with cis-regulatory elements in the genome. Hox transcription factors specify the spatial coordinates of the body axis in all animals with bilateral symmetry, but a detailed knowledge of their molecular function in instructing cell fates is lacking. Here, we used chromatin immunoprecipitation with massively parallel sequencing (ChIP-seq) to identify Hoxa2 genomic locations in a time and space when it is actively instructing embryonic development in mouse. Our data reveals that Hoxa2 has large genome coverage and potentially regulates thousands of genes. Sequence analysis of Hoxa2-bound regions identifies high occurrence of two main classes of motifs, corresponding to Hox and Pbx-Hox recognition sequences. Examination of the binding targets of Hoxa2 faithfully captures the processes regulated by Hoxa2 during embryonic development; in addition, it uncovers a large cluster of potential targets involved in the Wnt-signaling pathway. In vivo examination of canonical Wnt-ß-catenin signaling reveals activity specifically in Hoxa2 domain of expression, and this is undetectable in Hoxa2 mutant embryos. The comprehensive mapping of Hoxa2-binding sites provides a framework to study Hox regulatory networks in vertebrate developmental processes.

Cross-talk between DNA methylation and active histone modifications regulates aberrant expression of ZAP70 in CLL.

Zeta-associated protein of 70 kD (ZAP70) is a recognized adverse prognostic marker in chronic lymphocytic leukaemia (CLL) associated with enhanced B-cell receptor signalling, significantly more aggressive disease course and poor overall survival. Zeta-associated protein of 70 kD is ordinarily expressed in T cells where it has a crucial role in T-cell receptor signalling, whereas its aberrant expression in CLL leads to enhanced B-cell receptor signalling and significantly more aggressive disease course. Although much is known about the activation of ZAP70 following engagement of T-cell receptor, there are little data on the regulation of ZAP70 gene expression in normal T cells or CLL. To understand the molecular events underpinning epigenetic regulation of ZAP70 gene in CLL, we have defined ZAP70 promoter region and outlined the regions crucial in regulating the gene activity. Following a direct comparison of ZAP70+ and ZAP70- primary CLLs, we show ZAP70 promoter in expressing CLLs to be associated with a spectrum of active histone modifications, some of which are tightly linked to aberrant DNA methylation in CLL. Cross-talk between histone modifications and reduced DNA methylation culminates in transcriptional de-repression of ZAP70. Moreover, treatment with histone deacetylase (HDAC) and DNA methylation inhibitors results in recovery of ZAP70 expression, which provides a possible explanation for the failure of HDAC inhibitors in CLL treatment and might serve as a cautionary warning for their future use in treatment of this leukaemia.

ZAP70 in chronic lymphocytic leukaemia.

The protein tyrosine kinase zeta-chain associated protein kinase (ZAP70), normally expressed in T cells and a subset of B cells, is solely expressed in poor prognosis chronic lymphocytic leukaemia and implicated in enhanced B cell receptor signalling. As a result, the expression of this protein provides an ideal prognostic marker for the disease. A previous study has shown differential CpG methylation of a 5' region of ZAP70 in leukaemic lymphoid cells, although no further epigenetic studies have been reported. Further investigation into the expression of ZAP70 may therefore provide targets for therapies.