Improving the performance of supervised deep learning for regulatory genomics using phylogenetic augmentation.

MOTIVATION: Supervised deep learning is used to model the complex relationship between genomic sequence and regulatory function. Understanding how these models make predictions can provide biological insight into regulatory functions. Given the complexity of the sequence to regulatory function mapping (the cis-regulatory code), it has been suggested that the genome contains insufficient sequence variation to train models with suitable complexity. Data augmentation is a widely used approach to increase the data variation available for model training, however current data augmentation methods for genomic sequence data are limited. RESULTS: Inspired by the success of comparative genomics, we show that augmenting genomic sequences with evolutionarily related sequences from other species, which we term phylogenetic augmentation, improves the performance of deep learning models trained on regulatory genomic sequences to predict high-throughput functional assay measurements. Additionally, we show that phylogenetic augmentation can rescue model performance when the training set is down-sampled and permits deep learning on a real-world small dataset, demonstrating that this approach improves data efficiency. Overall, this data augmentation method represents a solution for improving model performance that is applicable to many supervised deep-learning problems in genomics. AVAILABILITY AND IMPLEMENTATION: The open-source GitHub repository agduncan94/phylogenetic_augmentation_paper includes the code for rerunning the analyses here and recreating the figures.

SOX4 exerts contrasting regulatory effects on labor-associated gene promoters in myometrial cells.

The uterine muscular layer, or myometrium, undergoes profound changes in global gene expression during its progression from a quiescent state during pregnancy to a contractile state at the onset of labor. In this study, we investigate the role of SOX family transcription factors in myometrial cells and provide evidence for the role of SOX4 in regulating labor-associated genes. We show that Sox4 has elevated expression in the murine myometrium during a term laboring process and in two mouse models of preterm labor. Additionally, SOX4 differentially affects labor-associated gene promoter activity in cooperation with activator protein 1 (AP-1) dimers. SOX4 exerted no effect on the Gja1 promoter; a JUND-specific activation effect at the Fos promoter; a positive activation effect on the Mmp11 promoter with the AP-1 dimers; and surprisingly, we noted that the reporter expression of the Ptgs2 promoter in the presence of JUND and FOSL2 was repressed by the addition of SOX4. Our data indicate SOX4 may play a diverse role in regulating gene expression in the laboring myometrium in cooperation with AP-1 factors. This study enhances our current understanding of the regulatory network that governs the transcriptional changes associated with the onset of labor and highlights a new molecular player that may contribute to the labor transcriptional program.

Evolution of a novel regulatory mechanism of hypoxia inducible factor in hypoxia-tolerant electric fishes.

Hypoxia is a significant source of metabolic stress that activates many cellular pathways involved in cellular differentiation, proliferation, and cell death. Hypoxia is also a major component in many human diseases and a known driver of many cancers. Despite the challenges posed by hypoxia, there are animals that display impressive capacity to withstand lethal levels of hypoxia for prolonged periods of time and thus offer a gateway to a more comprehensive understanding of the hypoxic response in vertebrates. The weakly electric fish genus Brachyhypopomus inhabits some of the most challenging aquatic ecosystems in the world, with some species experiencing seasonal anoxia, thus providing a unique system to study the cellular and molecular mechanisms of hypoxia tolerance. In this study, we use closely related species of Brachyhypopomus that display a range of hypoxia tolerances to probe for the underlying molecular mechanisms via hypoxia inducible factors (HIFs)-transcription factors known to coordinate the cellular response to hypoxia in vertebrates. We find that HIF1⍺ from hypoxia tolerant Brachyhypopomus species displays higher transactivation in response to hypoxia than that of intolerant species, when overexpressed in live cells. Moreover, we identified two SUMO-interacting motifs near the oxygen-dependent degradation and transactivation domains of the HIF1⍺ protein that appear to boost transactivation of HIF1, regardless of the genetic background. Together with computational analyses of selection, this shows that evolution of HIF1⍺ are likely to underlie adaptations to hypoxia tolerance in Brachyhypopomus electric fishes, with changes in two SUMO-interacting motifs facilitating the mechanism of this tolerance.

Epigenetic reprogramming of a distal developmental enhancer cluster drives SOX2 overexpression in breast and lung adenocarcinoma.

Enhancer reprogramming has been proposed as a key source of transcriptional dysregulation during tumorigenesis, but the molecular mechanisms underlying this process remain unclear. Here, we identify an enhancer cluster required for normal development that is aberrantly activated in breast and lung adenocarcinoma. Deletion of the SRR124-134 cluster disrupts expression of the SOX2 oncogene, dysregulates genome-wide transcription and chromatin accessibility and reduces the ability of cancer cells to form colonies in vitro. Analysis of primary tumors reveals a correlation between chromatin accessibility at this cluster and SOX2 overexpression in breast and lung cancer patients. We demonstrate that FOXA1 is an activator and NFIB is a repressor of SRR124-134 activity and SOX2 transcription in cancer cells, revealing a co-opting of the regulatory mechanisms involved in early development. Notably, we show that the conserved SRR124 and SRR134 regions are essential during mouse development, where homozygous deletion results in the lethal failure of esophageal-tracheal separation. These findings provide insights into how developmental enhancers can be reprogrammed during tumorigenesis and underscore the importance of understanding enhancer dynamics during development and disease.

The manuscript by Abatti et al. shows that epigenetic reactivation of a pair of distal enhancers that drive Sox2 expression during development (to permit separation of the esophagus and trachea) is responsible for the tumor-promoting re-expression of SOX2 in breast and lung tumors. Intriguingly, the same transcription factors that act on the enhancers during development to either activate or repress them (i.e. FOXA1 and NFIB, respectively) are also required for altering chromatin accessibility of the enhancers and SOX2 transcription in breast and lung cancer cells. With their work, the authors unravel the exact mechanism of how developmentally active enhancers become repurposed in a tumor context and show the relevance of this repurposing event for cancer.

Competition between transcription and loop extrusion modulates promoter and enhancer dynamics.

The spatiotemporal configuration of genes with distal regulatory elements, and the impact of chromatin mobility on transcription, remain unclear. Loop extrusion is an attractive model for bringing genetic elements together, but how this functionally interacts with transcription is also largely unknown. We combine live tracking of genomic loci and nascent transcripts with molecular dynamics simulations to assess the 4D arrangement of the Sox2 gene and its enhancer, in response to a battery of perturbations. We find that alterations in chromatin mobility, not promoter-enhancer distance, is more informative about transcriptional status. Active elements display more constrained mobility, consistent with confinement within specialized nuclear sites, and alterations in enhancer mobility distinguish poised from transcribing alleles. Strikingly, we find that whereas loop extrusion and transcription factor-mediated clustering contribute to promoter-enhancer proximity, they have antagonistic effects on chromatin dynamics. This provides an experimental framework for the underappreciated role of chromatin dynamics in genome regulation.

MYB and ELF3 differentially modulate labor-inducing gene expression in myometrial cells.

Spontaneous uterine contractions are initiated when smooth muscle cells (SMCs) within the uterine muscle, or myometrium, transition from a functionally dormant to an actively contractile phenotype at the end of the pregnancy period. We know that this process is accompanied by gestational time point-specific differences in the SMC transcriptome, which can be modulated by the activator protein 1 (AP-1), nuclear factor kappa beta (NF-κß), estrogen receptor (ER), and progesterone receptor (PR) transcription factors. Less is known, however, about the additional proteins that might assist these factors in conferring the transcriptional changes observed at labor onset. Here, we present functional evidence for the roles of two proteins previously understudied in the SMC context-MYB and ELF3-which can contribute to the regulation of labor-driving gene transcription. We show that the MYB and ELF3 genes exhibit elevated transcript expression levels in mouse and human myometrial tissues during spontaneous term labor. The expression of both genes was also significantly increased in mouse myometrium during preterm labor induced by the progesterone antagonist mifepristone (RU486), but not during infection-simulating preterm labor induced by intrauterine infusion of lipopolysaccharide (LPS). Furthermore, both MYB and ELF3 proteins affect labor-driving gene promoter activity, although in surprisingly opposing ways: Gja1 and Fos promoter activation increases in the presence of MYB and decreases in the presence of ELF3. Collectively, our study adds to the current understanding of the transcription factor network that defines the transcriptomes of SMCs during late gestation and implicates two new players in the control of labor timing.

Enhancer-promoter interactions can bypass CTCF-mediated boundaries and contribute to phenotypic robustness.

How enhancers activate their distal target promoters remains incompletely understood. Here we dissect how CTCF-mediated loops facilitate and restrict such regulatory interactions. Using an allelic series of mouse mutants, we show that CTCF is neither required for the interaction of the Sox2 gene with distal enhancers, nor for its expression. Insertion of various combinations of CTCF motifs, between Sox2 and its distal enhancers, generated boundaries with varying degrees of insulation that directly correlated with reduced transcriptional output. However, in both epiblast and neural tissues, enhancer contacts and transcriptional induction could not be fully abolished, and insertions failed to disrupt implantation and neurogenesis. In contrast, Sox2 expression was undetectable in the anterior foregut of mutants carrying the strongest boundaries, and these animals fully phenocopied loss of SOX2 in this tissue. We propose that enhancer clusters with a high density of regulatory activity can better overcome physical barriers to maintain faithful gene expression and phenotypic robustness.

IFI16 Partners with KAP1 to Maintain Epstein-Barr Virus Latency.

Herpesviruses establish latency to ensure permanent residence in their hosts. Upon entry into a cell, these viruses are rapidly silenced by the host, thereby limiting the destructive viral lytic phase while allowing the virus to hide from the immune system. Notably, although the establishment of latency by the oncogenic herpesvirus Epstein-Barr virus (EBV) requires the expression of viral latency genes, latency can be maintained with a negligible expression of viral genes. Indeed, in several herpesviruses, the host DNA sensor IFI16 facilitated latency via H3K9me3 heterochromatinization. This silencing mark is typically imposed by the constitutive heterochromatin machinery (HCM). The HCM, in an antiviral role, also silences the lytic phase of EBV and other herpes viruses. We investigated if IFI16 restricted EBV lytic activation by partnering with the HCM and found that IFI16 interacted with core components of the HCM, including the KRAB-associated protein 1 (KAP1) and the site-specific DNA binding KRAB-ZFP SZF1. This partnership silenced the EBV lytic switch protein ZEBRA, encoded by the BZLF1 gene, thereby favoring viral latency. Indeed, IFI16 contributed to H3K9 trimethylation at lytic genes of all kinetic classes. In defining topology, we found that IFI16 coenriched with KAP1 at the BZLF1 promoter, and while IFI16 and SZF1 were each adjacent to KAP1 in latent cells, IFI16 and SZF1 were not. Importantly, we also found that disruption of latency involved rapid downregulation of IFI16 transcription. These findings revealed a previously unknown partnership between IFI16 and the core HCM that supports EBV latency via antiviral heterochromatic silencing. IMPORTANCE The interferon-gamma inducible protein 16 (IFI16) is a nuclear DNA sensor that mediates antiviral responses by activating the inflammasome, triggering an interferon response, and silencing lytic genes of herpesviruses. The last, which helps maintain latency of the oncoherpesvirus Epstein-Barr virus (EBV), is accomplished via H3K9me3 heterochromatinization through unknown mechanisms. Here, we report that IFI16 physically partners with the core constitutive heterochromatin machinery to silence the key EBV lytic switch protein, thereby ensuring continued viral latency in B lymphocytes. We also find that disruption of latency involves rapid transcriptional downregulation of IFI16. These findings point to hitherto unknown physical and functional partnerships between a well-known antiviral mechanism and the core components of the constitutive heterochromatin machinery.

Transcriptional regulation and chromatin architecture maintenance are decoupled functions at the Sox2 locus.

How distal regulatory elements control gene transcription and chromatin topology is not clearly defined, yet these processes are closely linked in lineage specification during development. Through allele-specific genome editing and chromatin interaction analyses of the Sox2 locus in mouse embryonic stem cells, we found a striking disconnection between transcriptional control and chromatin architecture. We traced nearly all Sox2 transcriptional activation to a small number of key transcription factor binding sites, whose deletions have no effect on promoter-enhancer interaction frequencies or topological domain organization. Local chromatin architecture maintenance, including at the topologically associating domain (TAD) boundary downstream from the Sox2 enhancer, is widely distributed over multiple transcription factor-bound regions and maintained in a CTCF-independent manner. Furthermore, partial disruption of promoter-enhancer interactions by ectopic chromatin loop formation has no effect on Sox2 transcription. These findings indicate that many transcription factors are involved in modulating chromatin architecture independently of CTCF.

Histone macroH2A1 is a stronger regulator of hippocampal transcription and memory than macroH2A2 in mice.

Histone variants H2A.Z and H3.3 are epigenetic regulators of memory, but roles of other variants are not well characterized. macroH2A (mH2A) is a structurally unique histone that contains a globular macrodomain connected to the histone region by an unstructured linker. Here we assessed if mH2A regulates memory and if this role varies for the two mH2A-encoding genes, H2afy (mH2A1) and H2afy2 (mH2A2). We show that fear memory is impaired in mH2A1, but not in mH2A2-deficient mice, whereas both groups were impaired in a non-aversive spatial memory task. However, impairment was larger for mH2A1- deficient mice, indicating a preferential role for mH2A1 over mH2A2 in memory. Accordingly, mH2A1 depletion in the mouse hippocampus resulted in more extensive transcriptional de-repression compared to mH2A2 depletion. mH2A1-depleted mice failed to induce a normal transcriptional response to fear conditioning, suggesting that mH2A1 depletion impairs memory by altering transcription. Using chromatin immunoprecipitation (ChIP) sequencing, we found that both mH2A proteins are enriched on transcriptionally repressed genes, but only mH2A1 occupancy was dynamically modified during learning, displaying reduced occupancy on upregulated genes after training. These data identify mH2A as a regulator of memory and suggest that mH2A1 supports memory by repressing spurious transcription and promoting learning-induced transcriptional activation.

Nuclear RNA Isolation and Sequencing.

Most transcriptome studies involve sequencing and quantification of steady-state mRNA by isolating and sequencing poly (A) RNA. Although this type of sequencing data is informative to determine steady-state mRNA levels, it does not provide information on transcriptional output and thus may not always reflect changes in transcriptional regulation of gene expression . Furthermore, sequencing poly (A) RNA may miss transcribed regions of the genome not usually modified by polyadenylation which includes many long non-coding RNAs including enhancer RNA (eRNA). Here, we describe nuclear RNA sequencing (nucRNA-seq) which investigates the transcriptional landscape through sequencing and quantification of nuclear RNAs which are both unspliced and spliced transcripts for protein-coding genes and nuclear-retained long non-coding RNAs.

Transcriptional control of parturition: insights from gene regulation studies in the myometrium.

The onset of labour is a culmination of a series of highly coordinated and preparatory physiological events that take place throughout the gestational period. In order to produce the associated contractions needed for foetal delivery, smooth muscle cells in the muscular layer of the uterus (i.e. myometrium) undergo a transition from quiescent to contractile phenotypes. Here, we present the current understanding of the roles transcription factors play in critical labour-associated gene expression changes as part of the molecular mechanistic basis for this transition. Consideration is given to both transcription factors that have been well-studied in a myometrial context, i.e. activator protein 1, progesterone receptors, oestrogen receptors, and nuclear factor kappa B, as well as additional transcription factors whose gestational event-driving contributions have been demonstrated more recently. These transcription factors may form pregnancy- and labour-associated transcriptional regulatory networks in the myometrium to modulate the timing of labour onset. A more thorough understanding of the transcription factor-mediated, labour-promoting regulatory pathways holds promise for the development of new therapeutic treatments that can be used for the prevention of preterm labour in at-risk women.

A flexible repertoire of transcription factor binding sites and a diversity threshold determines enhancer activity in embryonic stem cells.

Transcriptional enhancers are critical for development and phenotype evolution and are often mutated in disease contexts; however, even in well-studied cell types, the sequence code conferring enhancer activity remains unknown. To examine the enhancer regulatory code for pluripotent stem cells, we identified genomic regions with conserved binding of multiple transcription factors in mouse and human embryonic stem cells (ESCs). Examination of these regions revealed that they contain on average 12.6 conserved transcription factor binding site (TFBS) sequences. Enriched TFBSs are a diverse repertoire of 70 different sequences representing the binding sequences of both known and novel ESC regulators. Using a diverse set of TFBSs from this repertoire was sufficient to construct short synthetic enhancers with activity comparable to native enhancers. Site-directed mutagenesis of conserved TFBSs in endogenous enhancers or TFBS deletion from synthetic sequences revealed a requirement for 10 or more different TFBSs. Furthermore, specific TFBSs, including the POU5F1:SOX2 comotif, are dispensable, despite cobinding the POU5F1 (also known as OCT4), SOX2, and NANOG master regulators of pluripotency. These findings reveal that a TFBS sequence diversity threshold overrides the need for optimized regulatory grammar and individual TFBSs that recruit specific master regulators.

The recycling endosome protein Rab25 coordinates collective cell movements in the zebrafish surface epithelium.

In emerging epithelial tissues, cells undergo dramatic rearrangements to promote tissue shape changes. Dividing cells remain interconnected via transient cytokinetic bridges. Bridges are cleaved during abscission and currently, the consequences of disrupting abscission in developing epithelia are not well understood. We show that the Rab GTPase Rab25 localizes near cytokinetic midbodies and likely coordinates abscission through endomembrane trafficking in the epithelium of the zebrafish gastrula during epiboly. In maternal-zygotic Rab25a and Rab25b mutant embryos, morphogenic activity tears open persistent apical cytokinetic bridges that failed to undergo timely abscission. Cytokinesis defects result in anisotropic cell morphologies that are associated with a reduction of contractile actomyosin networks. This slows cell rearrangements and alters the viscoelastic responses of the tissue, all of which likely contribute to delayed epiboly. We present a model in which Rab25 trafficking coordinates cytokinetic bridge abscission and cortical actin density, impacting local cell shape changes and tissue-scale forces.

Transcriptional enhancers: from prediction to functional assessment on a genome-wide scale.

Enhancers are cis-regulatory sequences located distally to target genes. These sequences consolidate developmental and environmental cues to coordinate gene expression in a tissue-specific manner. Enhancer function and tissue specificity depend on the expressed set of transcription factors, which recognize binding sites and recruit cofactors that regulate local chromatin organization and gene transcription. Unlike other genomic elements, enhancers are challenging to identify because they function independently of orientation, are often distant from their promoters, have poorly defined boundaries, and display no reading frame. In addition, there are no defined genetic or epigenetic features that are unambiguously associated with enhancer activity. Over recent years there have been developments in both empirical assays and computational methods for enhancer prediction. We review genome-wide tools, CRISPR advancements, and high-throughput screening approaches that have improved our ability to both observe and manipulate enhancers in vitro at the level of primary genetic sequences, chromatin states, and spatial interactions. We also highlight contemporary animal models and their importance to enhancer validation. Together, these experimental systems and techniques complement one another and broaden our understanding of enhancer function in development, evolution, and disease.

Enhancer-gene rewiring in the pathogenesis of Quebec platelet disorder.

Quebec platelet disorder (QPD) is an autosomal dominant bleeding disorder with a unique, platelet-dependent, gain-of-function defect in fibrinolysis, without systemic fibrinolysis. The hallmark feature of QPD is a >100-fold overexpression of PLAU, specifically in megakaryocytes. This overexpression leads to a >100-fold increase in platelet stores of urokinase plasminogen activator (PLAU/uPA); subsequent plasmin-mediated degradation of diverse α-granule proteins; and platelet-dependent, accelerated fibrinolysis. The causative mutation is a 78-kb tandem duplication of PLAU. How this duplication causes megakaryocyte-specific PLAU overexpression is unknown. To investigate the mechanism that causes QPD, we used epigenomic profiling, comparative genomics, and chromatin conformation capture approaches to study PLAU regulation in cultured megakaryocytes from participants with QPD and unaffected controls. QPD duplication led to ectopic interactions between PLAU and a conserved megakaryocyte enhancer found within the same topologically associating domain (TAD). Our results support a unique disease mechanism whereby the reorganization of sub-TAD genome architecture results in a dramatic, cell-type-specific blood disorder phenotype.

The pregnant myometrium is epigenetically activated at contractility-driving gene loci prior to the onset of labor in mice.

During gestation, uterine smooth muscle cells transition from a state of quiescence to one of contractility, but the molecular mechanisms underlying this transition at a genomic level are not well-known. To better understand these events, we evaluated the epigenetic landscape of the mouse myometrium during the pregnant, laboring, and postpartum stages. We generated gestational time point-specific enrichment profiles for histone H3 acetylation on lysine residue 27 (H3K27ac), histone H3 trimethylation of lysine residue 4 (H3K4me3), and RNA polymerase II (RNAPII) occupancy by chromatin immunoprecipitation with massively parallel sequencing (ChIP-seq), as well as gene expression profiles by total RNA-sequencing (RNA-seq). Our findings reveal that 533 genes, including known contractility-driving genes (Gap junction alpha 1 [Gja1], FBJ osteosarcoma oncogene [Fos], Fos-like antigen 2 [Fosl2], Oxytocin receptor [Oxtr], and Prostaglandin G/H synthase 2 (Ptgs2), for example), are up-regulated at day 19 during active labor because of an increase in transcription at gene bodies. Labor-associated promoters and putative intergenic enhancers, however, are epigenetically activated as early as day 15, by which point the majority of genome-wide H3K27ac or H3K4me3 peaks present in term laboring tissue is already established. Despite this early exhibited histone signature, increased noncoding enhancer RNA (eRNA) production at putative intergenic enhancers and recruitment of RNAPII to the gene bodies of labor-associated loci were detected only during labor. Our findings indicate that epigenetic activation of the myometrial genome precedes active labor by at least 4 days in the mouse model, suggesting that the myometrium is poised for rapid activation of contraction-associated genes in order to exit the state of quiescence.