Learning the histone codes of gene regulation with large genomic windows and three-dimensional chromatin interactions using transformer

The quantitative characterization of the transcriptional control by histone modifications (HMs) has been challenged by many computational studies, but still most of them exploit only partial aspects of intricate mechanisms involved in gene regulation, leaving a room for improvement. We present Chromoformer, a new transformer-based deep learning architecture that achieves the state-of-the-art performance in the quantitative deciphering of the histone codes of gene regulation. The core essence of Chromoformer architecture lies in the three variants of attention operation, each specialized to model individual hierarchy of three-dimensional (3D) transcriptional regulation including (1) histone codes at core promoters, (2) pairwise interaction between a core promoter and a distal cis-regulatory element mediated by 3D chromatin interactions, and (3) the collective effect of the pairwise cis-regulations. In-depth interpretation of the trained model behavior based on attention scores suggests that Chromoformer adaptively exploits the distant dependencies between HMs associated with transcription initiation and elongation. We also demonstrate that the quantitative kinetics of transcription factories and polycomb group bodies, in which the coordinated gene regulation occurs through spatial sequestration of genes with regulatory elements, can be captured by Chromoformer. Together, our study shows the great power of attention-based deep learning as a versatile modeling approach for the complex epigenetic landscape of gene regulation and highlights its potential as an effective toolkit that facilitates scientific discoveries in computational epigenetics.

The Histone Modifications of Neuronal Plasticity

Nucleosomes composed of histone octamer and DNA are the basic structural unit in the eukaryote chromosome. Under the stimulation of various factors, histones will undergo posttranslational modifications such as methylation, phosphorylation, acetylation, and ubiquitination, which change the three-dimensional structure of chromosomes and affect gene expression. Therefore, the combination of different states of histone modifications modulates gene expression is called histone code. The formation of learning and memory is one of the most important mechanisms for animals to adapt to environmental changes. A large number of studies have shown that histone codes are involved in the formation and consolidation of learning and memory. Here, we review the most recent literature of histone modification in regulating neurogenesis, dendritic spine dynamic, synapse formation, and synaptic plasticity.

PDTM-39. CRACKING THE HISTONE CODE: REVEALING THERAPEUTIC EPIGENETIC TARGETS THROUGH HISTONE H3 TAIL ANALYSIS OF DIFFUSE INTRINSIC PONTINE GLIOMA

INTRODUCTION Diffuse intrinsic pontine glioma (DIPG) is a highly morbid pediatric cancer. Up to 80% harbor a Histone H3K27M mutation, which alters wild type Histone H3 protein post-translational modifications (PTMs) and genomic enrichment patterns to impact chromatin structure and transcription regulation. We previously identified tumorigenic patterns of H3K27Ac/bromodomain co-enrichment in DIPG, and demonstrated pre-clinical efficacy of bromodomain inhibition (JQ1). Here, we employ a novel proteomics platform, developed at our institution, to further elucidate the impact of H3K7M mutation on glioma histone codes and response to bromodomain inhibition. METHODS Epiproteomic analysis was performed on pediatric glioma cell lines (H3K27 WT n=2, H3K27M n=2) to characterize 95 distinct Histone H3.3 and H3.1 N-terminal tail modification states. Cells were treated with JQ1 or DMSO, and collected at 0h, 24h, 48h. Histones were extracted from isolated nuclei, immunopurified, and analyzed by LC-MS/MS. Results were integrated with RNA-Seq and ChIP Seq results (H3K27M, H3K27Ac, H3K27me3, H3K4me1, H3K4me3) from the same DIPG cell lines. Pediatric glioma tissues (H3K27M WT n=3, H3K27M n= 9) were similarly analyzed to validate cell line results. RESULTS Cell PTM profiles cluster by H3 mutation status on unsupervised analysis. Relative H3 PTM abundance were compared across cell lines by tumor location, H3 mutation status, and in response to treatment. Significant differential genomic enrichment H3K27M and H3.3 WT proteins, H3K27Me3 and H3K27Ac were observed between mutant and wild type cell lines with epigenetic-targeted therapy, correlating with cell transcriptomes. CONCLUSIONS Histone H3 tail epiproteomic analysis reveals DIPG histone codes in situ, revealing the effects of bromodomain inhibition on the tumor epigenetic landscape and providing new insight to the mechanism of tumor formation and therapy response. Further investigation of the utility of these signatures as biomarkers for diagnosis and longitudinal monitoring of treatment response are therefore currently underway.

A mass spectrometry-based assay using metabolic labeling to rapidly monitor chromatin accessibility of modified histone proteins

Histone post-translational modifications (PTMs) contribute to chromatin accessibility due to their chemical properties and their ability to recruit enzymes responsible for DNA readout and chromatin remodeling. To date, more than 400 different histone PTMs and thousands of combinations of PTMs have been identified, the vast majority with still unknown biological function. Identification and quantification of histone PTMs has become routine in mass spectrometry (MS) but, since raising antibodies for each PTM in a study can be prohibitive, lots of potential is lost from MS datasets when uncharacterized PTMs are found to be significantly regulated. We developed an assay that uses metabolic labeling and MS to associate chromatin accessibility with histone PTMs and their combinations. The labeling is achieved by spiking in the cell media a 5x concentration of stable isotope labeled arginine and allow cells to grow for at least one cell cycle. We quantified the labeling incorporation of about 200 histone peptides with a proteomics workflow, and we confirmed that peptides carrying PTMs with extensively characterized roles in active transcription or gene silencing were in highly or poorly labeled forms, respectively. Data were further validated using next-generation sequencing to assess the transcription rate of chromatin regions modified with five selected PTMs. Furthermore, we quantified the labeling rate of peptides carrying co-existing PTMs, proving that this method is suitable for combinatorial PTMs. We focus on the abundant bivalent mark H3K27me3K36me2, showing that H3K27me3 dominantly represses histone swapping rate even in the presence of the more permissive PTM H3K36me2. Together, we envision this method will help to generate hypotheses regarding histone PTM functions and, potentially, elucidate the role of combinatorial histone codes.

Bivalent Histone Codes on WNT5A during Odontogenic Differentiation

Lineage-committed differentiation is an essential biological program during odontogenesis, which is tightly regulated by lineage-specific genes. Some of these genes are modified by colocalization of H3K4me3 and H3K27me3 marks at promoter regions in progenitors. These modifications, named “bivalent domains,” maintain genes in a poised state and then resolve for later activation or repression during differentiation. Wnt5a has been reported to promote odontogenic differentiation in dental mesenchyme. However, relatively little is known about the epigenetic modulations on Wnt5a activation during tooth development. Here, we investigated the spatiotemporal patterns of H3K4me3 and H3K27me3 marks in developing mouse molars. Associated H3K4me3 methylases (mixed-lineage leukemia [MLL] complex) and H3K27me3 demethylases (JMJD3 and UTX) were dynamically expressed between early and late bell stage of human tooth germs and in cultured human dental papilla cells (hDPCs) during odontogenic induction. Poised WNT5A gene was marked by bivalent domains containing repressive marks (H3K27me3) and active marks (H3K4me3) on promoters. The bivalent domains tended to resolve during inducted differentiation, with removal of the H3K27me3 mark in a JMJD3-dependent manner. When JMJD3 was knocked down in cultured hDPCs, odontogenic differentiation was suppressed. The depletion of JMJD3 epigenetically repressed WNT5A activation by increased H3K27me3 marks. In addition, JMJD3 could physically interact with ASH2L, a component of the MLL complex, to form a coactivator complex, cooperatively modulating H3K4me3 marks on WNT5A promoters. Overall, our study reveals that transcription activities of WNT5A were epigenetically regulated by the negotiated balance between H3K27me3 and H3K4me3 marks and tightly mediated by JMJD3 and MLL coactivator complex, ultimately modulating odontogenic commitment during dental mesenchymal cell differentiation.

The histone codes for meiosis

Meiosis is a specialized process that produces haploid gametes from diploid cells by a single round of DNA replication followed by two successive cell divisions. It contains many special events, such as programmed DNA double-strand break (DSB) formation, homologous recombination, crossover formation and resolution. These events are associated with dynamically regulated chromosomal structures, the dynamic transcriptional regulation and chromatin remodeling are mainly modulated by histone modifications, termed ‘histone codes’. The purpose of this review is to summarize the histone codes that are required for meiosis during spermatogenesis and oogenesis, involving meiosis resumption, meiotic asymmetric division and other cellular processes. We not only systematically review the functional roles of histone codes in meiosis but also discuss future trends and perspectives in this field.

Biochemical Analysis of Histone Succinylation

Posttranslational modification (PTM) of proteins is used to regulate protein activity and stability. Histone PTMs are regarded as some of the most important, as they can directly regulate gene expression through chromatin reorganization. Recently, histone proteins were found to undergo succinylation, adding to other well-known PTMs such as acetylation, methylation, and phosphorylation. However, there is little information regarding the enzyme which catalyzes histone lysine succinylation. In fact, it is unclear whether this reaction is enzymatic. In this study, we tested histone succinylation activity in vitro using cell nuclear extracts of HepG2 cells. Although whole nuclear extracts did not show histone succinylation activity, we found that an SP 1.0 M KCl fraction of nuclear extracts indeed had such activity. These data offer the first direct evidence that histone succinylation is an enzymatic PTM as are other histone codes in the nucleus.