Suboptimization of developmental enhancers

A gene regulatory network for specification and morphogenesis of a Mauthner Cell homolog in non-vertebrate chordates

Transcriptional regulation of gene expression is an indispensable process in multicellular development, yet we still do not fully understand how the complex networks of transcription factors operating in neuronal precursors coordinately control the expression of effector genes that shape morphogenesis and terminal differentiation. Here we break down in greater detail a provisional regulatory circuit downstream of the transcription factor Pax3/7 operating in the descending decussating neurons (ddNs) of the tunicate Ciona robusta. The ddNs are a pair of hindbrain neurons proposed to be homologous to the Mauthner cells of anamniotes, and Pax3/7 is sufficient and necessary for their specification. We show that different transcription factors downstream of Pax3/7, namely Pou4, Lhx1/5, and Dmbx, regulate distinct "branches" of this ddN network that appear to be dedicated to different developmental tasks. Some of these network branches are shared with other neurons throughout the larva, reinforcing the idea that modularity is likely a key feature of such networks. We discuss these ideas and their evolutionary implications here, including the observation that homologs of all four transcription factors (Pax3/7, Lhx5, Pou4f3, and Dmbx1) are key for the specification of cranial neural crest in vertebrates.

Mechanisms for DNA Interplay in Eukaryotic Transcription Factors

Like their prokaryotic counterparts, eukaryotic transcription factors must recognize specific DNA sites, search for them efficiently, and bind to them to help recruit or block the transcription machinery. For eukaryotic factors, however, the genetic signals are extremely complex and scattered over vast, multichromosome genomes, while the DNA interplay occurs in a varying landscape defined by chromatin remodeling events and epigenetic modifications. Eukaryotic factors are rich in intrinsically disordered regions and are also distinct in their recognition of short DNA motifs and utilization of open DNA interaction interfaces as ways to gain access to DNA on nucleosomes. Recent findings are revealing the profound, unforeseen implications of such characteristics for the mechanisms of DNA interplay. In this review we discuss these implications and how they are shaping the eukaryotic transcription control paradigm into one of promiscuous signal recognition, highly dynamic interactions, heterogeneous DNA scanning, and multiprong conformational control.

Multimeric transcription factor BCL11A utilizes two zinc-finger tandem arrays to bind clustered short sequence motifs

BCL11A, a transcription factor, is vital for hematopoiesis, including B and T cell maturation and the fetal-to-adult hemoglobin switch. Mutations in BCL11A are linked to neurodevelopmental disorders. BCL11A contains two DNA-binding zinc-finger arrays, low-affinity ZF2-3 and high-affinity ZF4-6, separated by a 300-amino-acid linker. ZF2-3 and ZF4-5 share 73% identity, including five out of six DNA base-interacting residues. These arrays bind similar short sequence motifs in clusters, with the linker enabling a broader binding span. Crystallographic structures of ZF4-6, in complex with oligonucleotides from the β-globin locus region, reveal DNA sequence recognition by residues Asn756 (ZF4), Lys784 and Arg787 (ZF5). A Lys784-to-Thr mutation, linked to a neurodevelopmental disorder with persistent fetal globin expression, reduces DNA binding over 10-fold but gains interaction with a variable base pair. BCL11A isoforms may form oligomers, enhancing chromatin occupancy and repressor functions by allowing multiple copies of both low- and high-affinity ZF arrays to bind DNA. These distinctive properties, apparently conserved among vertebrates, provide essential functional flexibility to this crucial regulator.

Rewriting regulatory DNA to dissect and reprogram gene expression

Regulatory DNA provides a platform for transcription factor binding to encode cell-type-specific patterns of gene expression. However, the effects and programmability of regulatory DNA sequences remain difficult to map or predict. Here, we develop variant effects from flow-sorting experiments with CRISPR targeting screens (Variant-EFFECTS) to introduce hundreds of designed edits to endogenous regulatory DNA and quantify their effects on gene expression. We systematically dissect and reprogram 3 regulatory elements for 2 genes in 2 cell types. These data reveal endogenous binding sites with effects specific to genomic context, transcription factor motifs with cell-type-specific activities, and limitations of computational models for predicting the effect sizes of variants. We identify small edits that can tune gene expression over a large dynamic range, suggesting new possibilities for prime-editing-based therapeutics targeting regulatory DNA. Variant-EFFECTS provides a generalizable tool to dissect regulatory DNA and to identify genome editing reagents that tune gene expression in an endogenous context.

Interpreting regulatory mechanisms of Hippo signaling through a deep learning sequence model

Signaling pathway components are well studied, but how they mediate cell-type-specific transcription responses is an unresolved problem. Using the Hippo pathway in mouse trophoblast stem cells as a model, we show that the DNA binding of signaling effectors is driven by cell-type-specific sequence rules that can be learned genome wide by deep learning models. Through model interpretation and experimental validation, we show that motifs for the cell-type-specific transcription factor TFAP2C enhance TEAD4/YAP1 binding in a nucleosome-range and distance-dependent manner, driving synergistic enhancer activation. We also discovered that Tead double motifs are widespread, highly active canonical response elements. Molecular dynamics simulations suggest that TEAD4 binds them cooperatively through surprisingly labile protein-protein interactions that depend on the DNA template. These results show that the response to signaling pathways is encoded in the cis-regulatory sequences and that interpreting the rules reveals insights into the mechanisms by which signaling effectors influence cell-type-specific enhancer activity.

Predicting the DNA binding specificity of transcription factor mutants using family-level biophysically interpretable machine learning

Sequence-specific interactions of transcription factors (TFs) with genomic DNA underlie many cellular processes. High-throughput in vitro binding assays coupled with machine learning have made it possible to accurately define such molecular recognition in a biophysically interpretable way for hundreds of TFs across many structural families, providing new avenues for predicting how the sequence preference of a TF is impacted by disease-associated mutations in its DNA binding domain. We developed a method based on a reference-free tetrahedral representation of variation in base preference within a given structural family that can be used to accurately predict the effect of mutations in the protein sequence of the TF. Using the basic helix-loop-helix (bHLH) and homeodomain families as test cases, our results demonstrate the feasibility of accurately predicting the shifts (ΔΔΔG/RT) in binding free energy associated with TF mutants by leveraging high-quality DNA binding models for sets of homologous wild-type TFs.

Transfer learning reveals sequence determinants of the quantitative response to transcription factor dosage

Deep learning models have advanced our ability to predict cell-type-specific chromatin patterns from transcription factor (TF) binding motifs, but their application to perturbed contexts remains limited. We applied transfer learning to predict how concentrations of the dosage-sensitive TFs TWIST1 and SOX9 affect regulatory element (RE) chromatin accessibility in facial progenitor cells, achieving near-experimental accuracy. High-affinity motifs that allow for heterotypic TF co-binding and are concentrated at the center of REs buffer against quantitative changes in TF dosage and predict unperturbed accessibility. Conversely, low-affinity or homotypic binding motifs distributed throughout REs drive sensitive responses with minimal impact on unperturbed accessibility. Both buffering and sensitizing features display purifying selection signatures. We validated these sequence features through reporter assays and demonstrated that TF-nucleosome competition can explain low-affinity motifs' sensitizing effects. This combination of transfer learning and quantitative chromatin response measurements provides a novel approach for uncovering additional layers of the cis-regulatory code.

Modeling and designing enhancers by introducing and harnessing transcription factor binding units

Enhancers serve as pivotal regulators of gene expression throughout various biological processes by interacting with transcription factors (TFs). While transcription factor binding sites (TFBSs) are widely acknowledged as key determinants of TF binding and enhancer activity, the significant role of their surrounding context sequences remains to be quantitatively characterized. Here we propose the concept of transcription factor binding unit (TFBU) to modularly model enhancers by quantifying the impact of context sequences surrounding TFBSs using deep learning models. Based on this concept, we develop DeepTFBU, a comprehensive toolkit for enhancer design. We demonstrate that designing TFBS context sequences can significantly modulate enhancer activities and produce cell type-specific responses. DeepTFBU is also highly efficient in the de novo design of enhancers containing multiple TFBSs. Furthermore, DeepTFBU enables flexible decoupling and optimization of generalized enhancers. We prove that TFBU is a crucial concept, and DeepTFBU is highly effective for rational enhancer design.

Enhancer cooperativity can compensate for loss of activity over large genomic distances

Enhancers are short DNA sequences that activate their target promoter from a distance; however, increasing the genomic distance between the enhancer and the promoter decreases expression levels. Many genes are controlled by combinations of multiple enhancers, yet the interaction and cooperation of individual enhancer elements are not well understood. Here, we developed a synthetic platform in mouse embryonic stem cells that allows building complex regulatory landscapes from the bottom up. We tested the system by integrating individual enhancers at different distances and confirmed that the strength of an enhancer contributes to how strongly it is affected by increased genomic distance. Furthermore, synergy between two enhancer elements depends on the distance at which the two elements are integrated: introducing a weak enhancer between a strong enhancer and the promoter strongly increases reporter gene expression, allowing enhancers to activate from increased genomic distances.

Positive autoregulation of Sox17 is necessary for gallbladder and extrahepatic bile duct formation

Expression of SRY-box transcription factor 17 (Sox17) in the endodermal region caudal to the hepatic diverticulum during late gastrulation is necessary for hepato-pancreato-biliary system formation. Analysis of an allelic series of promoter-proximal mutations near the transcription start site (TSS) 2 of Sox17 in mouse has revealed that gallbladder (GB) and extrahepatic bile duct (EHBD) development is exquisitely sensitive to Sox17 expression levels. Deletion of a SOX17-binding cis-regulatory element in the TSS2 promoter impairs GB and EHBD development by reducing outgrowth of the nascent biliary bud. These findings reveal the existence of a SOX17-dependent autoregulatory loop that drives Sox17 expression above a critical threshold concentration necessary for GB and EHBD development to occur, and that minor impairments in Sox17 gene expression are sufficient to impair the expression of SOX17-regulated genes in the nascent GB and EHBD system, impairing or preventing development.

ChromBPNet: bias factorized, base-resolution deep learning models of chromatin accessibility reveal cis-regulatory sequence syntax, transcription factor footprints and regulatory variants

Despite extensive mapping of cis-regulatory elements (cREs) across cellular contexts with chromatin accessibility assays, the sequence syntax and genetic variants that regulate transcription factor (TF) binding and chromatin accessibility at context-specific cREs remain elusive. We introduce ChromBPNet, a deep learning DNA sequence model of base-resolution accessibility profiles that detects, learns and deconvolves assay-specific enzyme biases from regulatory sequence determinants of accessibility, enabling robust discovery of compact TF motif lexicons, cooperative motif syntax and precision footprints across assays and sequencing depths. Extensive benchmarks show that ChromBPNet, despite its lightweight design, is competitive with much larger contemporary models at predicting variant effects on chromatin accessibility, pioneer TF binding and reporter activity across assays, cell contexts and ancestry, while providing interpretation of disrupted regulatory syntax. ChromBPNet also helps prioritize and interpret regulatory variants that influence complex traits and rare diseases, thereby providing a powerful lens to decode regulatory DNA and genetic variation.

Real-time visualization of reconstituted transcription reveals RNA polymerase II activation mechanisms at single promoters

RNA polymerase II (RNAPII) is regulated by sequence-specific transcription factors (TFs) and the pre-initiation complex (PIC): TFIIA, TFIIB, TFIID, TFIIE, TFIIF, TFIIH, Mediator. TFs and Mediator contain intrinsically-disordered regions (IDRs) and form phase-separated condensates, but how IDRs control RNAPII function remains poorly understood. Using purified PIC factors, we developed a Real-time In-vitro Fluorescence Transcription assay (RIFT) for second-by-second visualization of RNAPII transcription at hundreds of promoters simultaneously. We show rapid RNAPII activation is IDR-dependent, without condensate formation. For example, the MED1-IDR can functionally replace a native TF, activating RNAPII with similar (not identical) kinetics; however, MED1-IDR squelches transcription as a condensate, but activates as a single-protein. TFs and Mediator cooperatively activate RNAPII bursting and re-initiation and surprisingly, Mediator can drive TF-promoter recruitment, without TF-DNA binding. Collectively, RIFT addressed questions largely intractable with cell-based methods, yielding mechanistic insights about IDRs, condensates, enhancer-promoter communication, and RNAPII bursting that complement live-cell imaging data.

Functional variant rs9344 at 11q13.3 regulates CCND1 expression in multiple myeloma with t(11;14)

Multiple myeloma (MM) is a plasma cell (PC) malignancy characterized by cytogenetic abnormalities, such as t(11;14)(q13;q32), resulting in CCND1 overexpression. The rs9344 G allele within CCND1 is the most significant susceptibility allele for t(11;14). Sequencing data from 2 independent cohorts, CoMMpass (n = 698) and Mayo Clinic (n = 661), confirm the positive association between the G allele and t(11;14). Among 80% of individuals heterozygous for rs9344 with t(11;14), the t(11;14) event occurs on the G allele, demonstrating a biological preference for the G allele in t(11;14). Within t(11;14), the G allele is associated with higher CCND1 expression and elevated H3K27ac and H3K4me3. CRISPR/Cas9 mediated A to G conversion resulted in increased H3K27ac over CCND1 and elevated CCND1 expression. ENCODE ChIP-seq data supported a PAX5 binding site within the enhancer region covering rs9344, showing preferential binding to the G allele. Overexpression of PAX5 resulted in increased CCND1 expression. These results support the importance of rs9344 G enhancer in increasing CCND1 expression in MM.

Targeted mutagenesis of specific genomic DNA sequences in animals for the in vivo generation of variant libraries

Understanding how the number, placement and affinity of transcription factor binding sites dictates gene regulatory programs remains a major unsolved challenge in biology, particularly in the context of multicellular organisms. To uncover these rules, it is first necessary to find the binding sites within a regulatory region with high precision, and then to systematically modulate this binding site arrangement while simultaneously measuring the effect of this modulation on output gene expression. Massively parallel reporter assays (MPRAs), where the gene expression stemming from 10,000s of in vitro-generated regulatory sequences is measured, have made this feat possible in high-throughput in single cells in culture. However, because of lack of technologies to incorporate DNA libraries, MPRAs are limited in whole organisms. To enable MPRAs in multicellular organisms, we generated tools to create a high degree of mutagenesis in specific genomic loci in vivo using base editing. Targeting GFP integrated in the genome of Drosophila cell culture and whole animals as a case study, we show that the base editor AIDevoCDA1 stemming from sea lamprey fused to nCas9 is highly mutagenic. Surprisingly, longer gRNAs increase mutation efficiency and expand the mutating window, which can allow the introduction of mutations in previously untargetable sequences. Finally, we demonstrate arrays of >20 gRNAs that can efficiently introduce mutations along a 200bp sequence, making it a promising tool to test enhancer function in vivo in a high throughput manner.

The emergence and evolution of gene expression in genome regions replete with regulatory motifs

Gene regulation is essential for life and controlled by regulatory DNA. Mutations can modify the activity of regulatory DNA, and also create new regulatory DNA, a process called regulatory emergence. Non-regulatory and regulatory DNA contain motifs to which transcription factors may bind. In prokaryotes, gene expression requires a stretch of DNA called a promoter, which contains two motifs called -10 and -35 boxes. However, these motifs may occur in both promoters and non-promoter DNA in multiple copies. They have been implicated in some studies to improve promoter activity, and in others to repress it. Here, we ask whether the presence of such motifs in different genetic sequences influences promoter evolution and emergence. To understand whether and how promoter motifs influence promoter emergence and evolution, we start from 50 'promoter islands', DNA sequences enriched with -10 and -35 boxes. We mutagenize these starting 'parent' sequences, and measure gene expression driven by 240,000 of the resulting mutants. We find that the probability that mutations create an active promoter varies more than 200-fold, and is not correlated with the number of promoter motifs. For parent sequences without promoter activity, mutations created over 1500 new -10 and -35 boxes at unique positions in the library, but only ~0.3% of these resulted in de-novo promoter activity. Only ~13% of all -10 and -35 boxes contribute to de-novo promoter activity. For parent sequences with promoter activity, mutations created new -10 and -35 boxes in 11 specific positions that partially overlap with preexisting ones to modulate expression. We also find that -10 and -35 boxes do not repress promoter activity. Overall, our work demonstrates how promoter motifs influence promoter emergence and evolution. It has implications for predicting and understanding regulatory evolution, de novo genes, and phenotypic evolution.

Bicoid-nucleosome competition sets a concentration threshold for transcription constrained by genome replication

Transcription factors (TFs) regulate gene expression despite constraints from chromatin structure and the cell cycle. Here we examine the concentration-dependent regulation of hunchback by the Bicoid morphogen through a combination of quantitative imaging, mathematical modeling and epigenomics in Drosophila embryos. By live imaging of MS2 reporters, we find that, following mitosis, the timing of transcriptional activation driven by the hunchback P2 (hb P2) enhancer directly reflects Bicoid concentration. We build a stochastic model that can explain in vivo onset time distributions by accounting for both the competition between Bicoid and nucleosomes at hb P2 and a negative influence of DNA replication on transcriptional elongation. Experimental modulation of nucleosome stability alters onset time distributions and the posterior boundary of hunchback expression. We conclude that TF-nucleosome competition is the molecular mechanism whereby the Bicoid morphogen gradient specifies the posterior boundary of hunchback expression.

Crosstalk between paralogs and isoforms influences p63-dependent regulatory element activity

The p53 family of transcription factors (p53, p63 and p73) regulate diverse organismal processes including tumor suppression, maintenance of genome integrity and the development of skin and limbs. Crosstalk between transcription factors with highly similar DNA binding profiles, like those in the p53 family, can dramatically alter gene regulation. While p53 is primarily associated with transcriptional activation, p63 mediates both activation and repression. The specific mechanisms controlling p63-dependent gene regulatory activity are not well understood. Here, we use massively parallel reporter assays (MPRA) to investigate how local DNA sequence context influences p63-dependent transcriptional activity. Most regulatory elements with a p63 response element motif (p63RE) activate transcription, although binding of the p63 paralog, p53, drives a substantial proportion of that activity. p63RE sequence content and co-enrichment with other known activating and repressing transcription factors, including lineage-specific factors, correlates with differential p63RE-mediated activities. p63 isoforms dramatically alter transcriptional behavior, primarily shifting inactive regulatory elements towards high p63-dependent activity. Our analysis provides novel insight into how local sequence and cellular context influences p63-dependent behaviors and highlights the key, yet still understudied, role of transcription factor paralogs and isoforms in controlling gene regulatory element activity.

Distinct H3K9me3 heterochromatin maintenance dynamics govern different gene programmes and repeats in pluripotent cells

H3K9me3 heterochromatin, established by lysine methyltransferases (KMTs) and compacted by heterochromatin protein 1 (HP1) isoforms, represses alternative lineage genes and DNA repeats. Our understanding of H3K9me3 heterochromatin stability is presently limited to individual domains and DNA repeats. Here we engineered Suv39h2-knockout mouse embryonic stem cells to degrade remaining two H3K9me3 KMTs within 1 hour and found that both passive dilution and active removal contribute to H3K9me3 decay within 12-24 hours. We discovered four different H3K9me3 decay rates across the genome and chromatin features and transcription factor binding patterns that predict the stability classes. A 'binary switch' governs heterochromatin compaction, with HP1 rapidly dissociating from heterochromatin upon KMT depletion and a particular threshold level of HP1 limiting pioneer factor binding, chromatin opening and exit from pluripotency within 12 h. Unexpectedly, receding H3K9me3 domains unearth residual HP1β peaks enriched with heterochromatin-inducing proteins. Our findings reveal distinct H3K9me3 heterochromatin maintenance dynamics governing gene networks and repeats that together safeguard pluripotency.

Optimization of the Irf8 +32-kb enhancer disrupts dendritic cell lineage segregation

Autoactivation of lineage-determining transcription factors mediates bistable expression, generating distinct cell phenotypes essential for complex body plans. Classical type 1 dendritic cell (cDC1) and type 2 dendritic cell (cDC2) subsets provide nonredundant functions for defense against distinct immune challenges. Interferon regulatory factor 8 (IRF8), the cDC1 lineage-determining transcription factor, undergoes autoactivation in cDC1 progenitors to establish cDC1 identity, yet its expression is downregulated during cDC2 differentiation by an unknown mechanism. This study reveals that the Irf8 +32-kb enhancer, responsible for IRF8 autoactivation, is naturally suboptimized with low-affinity IRF8 binding sites. Introducing multiple high-affinity IRF8 sites into the Irf8 +32-kb enhancer causes a gain-of-function effect, leading to erroneous IRF8 autoactivation in specified cDC2 progenitors, redirecting them toward cDC1 and a novel hybrid DC subset with mixed-lineage phenotypes. Further, this also causes a loss-of-function effect, reducing Irf8 expression in cDC1s. These developmental alterations critically impair both cDC1-dependent and cDC2-dependent arms of immunity. Collectively, our findings underscore the significance of enhancer suboptimization in the developmental segregation of cDCs required for normal immune function.

A DNA base-specific sequence interposed between CRX and NRL contributes to RHODOPSIN expression

Gene expression emerges from DNA sequences through the interaction of transcription factors (TFs) with DNA cis-regulatory sequences. In eukaryotes, TFs bind to transcription factor binding sites (TFBSs) with differential affinities, enabling cell-specific gene expression. In this view, DNA enables TF binding along a continuum ranging from low to high affinity depending on its sequence composition; however, it is not known whether evolution has entailed a further level of entanglement between DNA-protein interaction. Here we found that the composition and length (22 bp) of the DNA sequence interposed between the CRX and NRL retinal TFs in the proximal promoter of RHODOPSIN (RHO) largely controls the expression levels of RHO. Mutagenesis of CRX-NRL DNA linking sequences (here termed "DNA-linker") results in uncorrelated gene expression variation. In contrast, mutual exchange of naturally occurring divergent human and mouse Rho cis-regulatory elements conferred similar yet species-specific Rho expression levels. Two orthogonal DNA-binding proteins targeted to the DNA-linker either activate or repress the expression of Rho depending on the DNA-linker orientation relative to the CRX and NRL binding sites. These results argue that, in this instance, DNA itself contributes to CRX and NRL activities through a code based on specific base sequences of a defined length, ultimately determining optimal RHO expression levels.

A change in cis-regulatory logic underlying obligate versus facultative muscle multinucleation in chordates

Vertebrates and tunicates are sister groups that share a common fusogenic factor, Myomaker (Mymk), that drives myoblast fusion and muscle multinucleation. Yet they are divergent in when and where they express Mymk. In vertebrates, all developing skeletal muscles express Mymk and are obligately multinucleated. In tunicates, Mymk is expressed only in post-metamorphic multinucleated muscles, but is absent from mononucleated larval muscles. In this study, we demonstrate that cis-regulatory sequence differences in the promoter region of Mymk underlie the different spatiotemporal patterns of its transcriptional activation in tunicates and vertebrates. Although in vertebrates myogenic regulatory factors (MRFs) such as MyoD1 alone are required and sufficient for Mymk transcription in all skeletal muscles, we show that transcription of Mymk in post-metamorphic muscles of the tunicate Ciona requires the combinatorial activity of MRF, MyoD and Early B-cell Factor (Ebf). This macroevolutionary difference appears to be encoded in cis, likely due to the presence of a putative Ebf-binding site adjacent to predicted MRF binding sites in the Ciona Mymk promoter. We further discuss how Mymk and myoblast fusion might have been regulated in the last common ancestor of tunicates and vertebrates, for which we propose two models.

How to scan naked DNA using promiscuous recognition and no clamping: a model for pioneer transcription factors

Most DNA scanning proteins uniquely recognize their cognate sequence motif and slide on DNA assisted by some sort of clamping interface. The pioneer transcription factors that control cell fate in eukaryotes must forgo both elements to gain access to DNA in naked and chromatin forms; thus, whether or how these factors scan naked DNA is unknown. Here, we use single-molecule techniques to investigate naked DNA scanning by the Engrailed homeodomain (enHD) as paradigm of highly promiscuous recognition and open DNA binding interface. We find that enHD scans naked DNA quite effectively, and about 200000-fold faster than expected for a continuous promiscuous slide. To do so, enHD scans about 675 bp of DNA in 100 ms and then redeploys stochastically to another location 530 bp afar in just 10 ms. During the scanning phase enHD alternates between slow- and medium-paced modes every 3 and 40 ms, respectively. We also find that enHD binds nucleosomes and does so with enhanced affinity relative to naked DNA. Our results demonstrate that pioneer-like transcription factors can in principle do both, target nucleosomes and scan active DNA efficiently. The hybrid scanning mechanism used by enHD appears particularly well suited for the highly complex genomic signals of eukaryotic cells.

CRISPR tiling deletion screens reveal functional enhancers of neuropsychiatric risk genes and allelic compensation effects (ACE) on transcription

Precise transcriptional regulation is critical for cellular function and development, yet the mechanism of this process remains poorly understood for many genes. To gain a deeper understanding of the regulation of neuropsychiatric disease risk genes, we identified a total of 39 functional enhancers for four dosage-sensitive genes, APP, FMR1, MECP2, and SIN3A, using CRISPR tiling deletion screening in human induced pluripotent stem cell (iPSC)-induced excitatory neurons. We found that enhancer annotation provides potential pathological insights into disease-associated copy number variants. More importantly, we discovered that allelic enhancer deletions at SIN3A could be compensated by increased transcriptional activities from the other intact allele. Such allelic compensation effects (ACE) on transcription is stably maintained during differentiation and, once established, cannot be reversed by ectopic SIN3A expression. Further, ACE at SIN3A occurs through dosage sensing by the promoter. Together, our findings unravel a regulatory compensation mechanism that ensures stable and precise transcriptional output for SIN3A, and potentially other dosage-sensitive genes.

A suboptimal OCT4-SOX2 binding site facilitates the naïve-state specific function of a Klf4 enhancer

Enhancers have critical functions in the precise, spatiotemporal control of transcription during development. It is thought that enhancer grammar, or the characteristics and arrangements of transcription factor binding sites, underlie the specific functions of developmental enhancers. In this study, we sought to identify grammatical constraints that direct enhancer activity in the naïve state of pluripotency, focusing on the enhancers for the naïve-state specific gene, Klf4. Using a combination of biochemical tests, reporter assays, and endogenous mutations in mouse embryonic stem cells, we have studied the binding sites for the transcription factors OCT4 and SOX2. We have found that the three Klf4 enhancers contain suboptimal OCT4-SOX2 composite binding sites. Substitution with a high-affinity OCT4-SOX2 binding site in Klf4 enhancer E2 rescued enhancer function and Klf4 expression upon loss of the ESRRB and STAT3 binding sites. We also observed that the low-affinity of the OCT4-SOX2 binding site is crucial to drive the naïve-state specific activities of Klf4 enhancer E2. Altogether, our work suggests that the affinity of OCT4-SOX2 binding sites could facilitate enhancer functions in specific states of pluripotency.

Protocol to electroporate DNA plasmids into Ciona robusta embryos at the 1-cell stage

Electroporation is a technique to introduce DNA constructs into cells using electric current. Here, we present a protocol to electroporate DNA plasmids into Ciona robusta embryos at the 1-cell stage. We describe steps for setting up and conducting electroporation. We then detail procedures for collecting, fixing, and mounting embryos and counting expression. This protocol can be used to study the expression of enhancers via reporter assays, manipulating cells using genes or modified genes such as dominant negatives, and genome editing. For complete details on the use and execution of this protocol, please refer to Song, et al.1.

Dissection of core promoter syntax through single nucleotide resolution modeling of transcription initiation

How the DNA sequence of cis-regulatory elements encode transcription initiation patterns remains poorly understood. Here we introduce CLIPNET, a deep learning model trained on population-scale PRO-cap data that predicts the position and quantity of transcription initiation with single nucleotide resolution from DNA sequence more accurately than existing approaches. Interpretation of CLIPNET revealed a complex regulatory syntax consisting of DNA-protein interactions in five major positions between -200 and +50 bp relative to the transcription start site, as well as more subtle positional preferences among transcriptional activators. Transcriptional activator and core promoter motifs work non-additively to encode distinct aspects of initiation, with the former driving initiation quantity and the latter initiation position. We identified core promoter motifs that explain initiation patterns in the majority of promoters and enhancers, including DPR motifs and AT-rich TBP binding sequences in TATA-less promoters. Our results provide insights into the sequence architecture governing transcription initiation.

Distinct H3K9me3 heterochromatin maintenance dynamics govern different gene programs and repeats in pluripotent cells

H3K9me3-heterochromatin, established by lysine methyltransferases (KMTs) and compacted by HP1 isoforms, represses alternative lineage genes and DNA repeats. Our understanding of H3K9me3-heterochromatin stability is presently limited to individual domains and DNA repeats. We engineered Suv39h2 KO mouse embryonic stem cells to degrade remaining two H3K9me3-KMTs within one hour and found that both passive dilution and active removal contribute to H3K9me3 decay within 12-24 hours. We discovered four different H3K9me3 decay rates across the genome and chromatin features and transcription factor binding patterns that predict the stability classes. A "binary switch" governs heterochromatin compaction, with HP1 rapidly dissociating from heterochromatin upon KMTs' depletion and a particular threshold level of HP1 limiting pioneer factor binding, chromatin opening, and exit from pluripotency within 12 hr. Unexpectedly, receding H3K9me3 domains unearth residual HP1β peaks enriched with heterochromatin-inducing proteins. Our findings reveal distinct H3K9me3-heterochromatin maintenance dynamics governing gene networks and repeats that together safeguard pluripotency.

Mutagenesis Sensitivity Mapping of Human Enhancers In Vivo

Distant-acting enhancers are central to human development. However, our limited understanding of their functional sequence features prevents the interpretation of enhancer mutations in disease. Here, we determined the functional sensitivity to mutagenesis of human developmental enhancers in vivo. Focusing on seven enhancers active in the developing brain, heart, limb and face, we created over 1700 transgenic mice for over 260 mutagenized enhancer alleles. Systematic mutation of 12-basepair blocks collectively altered each sequence feature in each enhancer at least once. We show that 69% of all blocks are required for normal in vivo activity, with mutations more commonly resulting in loss (60%) than in gain (9%) of function. Using predictive modeling, we annotated critical nucleotides at base-pair resolution. The vast majority of motifs predicted by these machine learning models (88%) coincided with changes to in vivo function, and the models showed considerable sensitivity, identifying 59% of all functional blocks. Taken together, our results reveal that human enhancers contain a high density of sequence features required for their normal in vivo function and provide a rich resource for further exploration of human enhancer logic.

Anatomy of a superenhancer

Interferon regulatory factor-8 (IRF8) is the lineage determining transcription factor for the type one classical dendritic cell (cDC1) subset, a terminal selector for plasmacytoid dendritic cells and important for the function of monocytes. Studies of Irf8 gene regulation have identified several enhancers controlling its activity during development of progenitors in the bone marrow that precisely regulate expression at distinct developmental stages. Each enhancer responds to distinct transcription factors that are expressed at each stage. IRF8 is first expressed in early progenitors that form the monocyte dendritic cell progenitor (MDP) in response to induction of the transcription factor CCAAT/enhancer-binding protein alpha (C/EBPα) acting at the Irf8 +56 kb enhancer. IRF8 levels increase further as the MDP transits into the common dendritic cell progenitor (CDP) in response to E protein activity at the Irf8 +41 kb enhancer. Upon Nfil3-induction in CDPs leading to specification of the cDC1 progenitor, abrupt induction of BATF3 forms the JUN/BATF3/IRF8 heterotrimer that activates the Irf8 +32 kb enhancer that sustains Irf8 autoactivation throughout the cDC1 lifetime. Deletions of each of these enhancers has revealed their stage dependent activation. Surprisingly, studies of compound heterozygotes for each combination of enhancer deletions revealed that activation of each subsequent enhancer requires the successful activation of the previous enhancer in strictly cis-dependent mechanism. Successful progression of enhancer activation is finely tuned to alter the functional accessibility of subsequent enhancers to factors active in the next stage of development. The molecular basis for these phenomenon is still obscure but could have implications for genomic regulation in a broader developmental context.

Systematic dissection of sequence features affecting binding specificity of a pioneer factor reveals binding synergy between FOXA1 and AP-1

Despite the unique ability of pioneer factors (PFs) to target nucleosomal sites in closed chromatin, they only bind a small fraction of their genomic motifs. The underlying mechanism of this selectivity is not well understood. Here, we design a high-throughput assay called chromatin immunoprecipitation with integrated synthetic oligonucleotides (ChIP-ISO) to systematically dissect sequence features affecting the binding specificity of a classic PF, FOXA1, in human A549 cells. Combining ChIP-ISO with in vitro and neural network analyses, we find that (1) FOXA1 binding is strongly affected by co-binding transcription factors (TFs) AP-1 and CEBPB; (2) FOXA1 and AP-1 show binding cooperativity in vitro; (3) FOXA1's binding is determined more by local sequences than chromatin context, including eu-/heterochromatin; and (4) AP-1 is partially responsible for differential binding of FOXA1 in different cell types. Our study presents a framework for elucidating genetic rules underlying PF binding specificity and reveals a mechanism for context-specific regulation of its binding.

An activity-specificity trade-off encoded in human transcription factors

Transcription factors (TFs) control specificity and activity of gene transcription, but whether a relationship between these two features exists is unclear. Here we provide evidence for an evolutionary trade-off between the activity and specificity in human TFs encoded as submaximal dispersion of aromatic residues in their intrinsically disordered protein regions. We identified approximately 500 human TFs that encode short periodic blocks of aromatic residues in their intrinsically disordered regions, resembling imperfect prion-like sequences. Mutation of periodic aromatic residues reduced transcriptional activity, whereas increasing the aromatic dispersion of multiple human TFs enhanced transcriptional activity and reprogramming efficiency, promoted liquid-liquid phase separation in vitro and more promiscuous DNA binding in cells. Together with recent work on enhancer elements, these results suggest an important evolutionary role of suboptimal features in transcriptional control. We propose that rational engineering of amino acid features that alter phase separation may be a strategy to optimize TF-dependent processes, including cellular reprogramming.

Specification and survival of post-metamorphic branchiomeric neurons in a non-vertebrate chordate

Tunicates are the sister group to the vertebrates, yet most species have a life cycle split between swimming larva and sedentary adult phases. During metamorphosis, larval neurons are replaced by adult-specific ones. The regulatory mechanisms underlying this replacement remain largely unknown. Using tissue-specific CRISPR/Cas9-mediated mutagenesis in the tunicate Ciona, we show that orthologs of conserved hindbrain and branchiomeric neuron regulatory factors Pax2/5/8 and Phox2 are required to specify the 'neck', a cellular compartment set aside in the larva to give rise to cranial motor neuron-like neurons post-metamorphosis. Using bulk and single-cell RNA-sequencing analyses, we characterize the transcriptome of the neck downstream of Pax2/5/8. We present evidence that neck-derived adult ciliomotor neurons begin to differentiate in the larva and persist through metamorphosis, contrary to the assumption that the adult nervous system is formed after settlement and the death of larval neurons during metamorphosis. Finally, we show that FGF signaling during the larval phase alters the patterning of the neck and its derivatives. Suppression of FGF converts neck cells into larval neurons that fail to survive metamorphosis, whereas prolonged FGF signaling promotes an adult neural stem cell-like fate.

Plant enhancers exhibit both cooperative and additive interactions among their functional elements

Enhancers are cis-regulatory elements that shape gene expression in response to numerous developmental and environmental cues. In animals, several models have been proposed to explain how enhancers integrate the activity of multiple transcription factors. However, it remains largely unclear how plant enhancers integrate transcription factor activity. Here, we use Plant STARR-seq to characterize 3 light-responsive plant enhancers-AB80, Cab-1, and rbcS-E9-derived from genes associated with photosynthesis. Saturation mutagenesis revealed mutations, many of which clustered in short regions, that strongly reduced enhancer activity in the light, in the dark, or in both conditions. When tested in the light, these mutation-sensitive regions did not function on their own; rather, cooperative interactions with other such regions were required for full activity. Epistatic interactions occurred between mutations in adjacent mutation-sensitive regions, and the spacing and order of mutation-sensitive regions in synthetic enhancers affected enhancer activity. In contrast, when tested in the dark, mutation-sensitive regions acted independently and additively in conferring enhancer activity. Taken together, this work demonstrates that plant enhancers show evidence for both cooperative and additive interactions among their functional elements. This knowledge can be harnessed to design strong, condition-specific synthetic enhancers.

Two redundant transcription factor binding sites in a single enhancer are essential for mammalian sex determination

Male development in mammals depends on the activity of the two SOX gene: Sry and Sox9, in the embryonic testis. As deletion of Enhancer 13 (Enh13) of the Sox9 gene results in XY male-to-female sex reversal, we explored the critical elements necessary for its function and hence, for testis and male development. Here, we demonstrate that while microdeletions of individual transcription factor binding sites (TFBS) in Enh13 lead to normal testicular development, combined microdeletions of just two SRY/SOX binding motifs can alone fully abolish Enh13 activity leading to XY male-to-female sex reversal. This suggests that for proper male development to occur, these few nucleotides of non-coding DNA must be intact. Interestingly, we show that depending on the nature of these TFBS mutations, dramatically different phenotypic outcomes can occur, providing a molecular explanation for the distinct clinical outcomes observed in patients harboring different variants in the same enhancer.

Transient interactions modulate the affinity of NF-κB transcription factors for DNA

The dimeric nuclear factor kappa B (NF-κB) transcription factors (TFs) regulate gene expression by binding to a variety of κB DNA elements with conserved G:C-rich flanking sequences enclosing a degenerate central region. Toward defining mechanistic principles of affinity regulated by degeneracy, we observed an unusual dependence of the affinity of RelA on the identity of the central base pair, which appears to be noncontacted in the complex crystal structures. The affinity of κB sites with A or T at the central position is ~10-fold higher than with G or C. The crystal structures of neither the complexes nor the free κB DNAs could explain the differences in affinity. Interestingly, differential dynamics of several residues were revealed in molecular dynamics simulation studies, where simulation replicates totaling 148 μs were performed on NF-κB:DNA complexes and free κB DNAs. Notably, Arg187 and Arg124 exhibited selectivity in transient interactions that orchestrated a complex interplay among several DNA-interacting residues in the central region. Binding and simulation studies with mutants supported these observations of transient interactions dictating specificity. In combination with published reports, this work provides insights into the nuanced mechanisms governing the discriminatory binding of NF-κB family TFs to κB DNA elements and sheds light on cancer pathogenesis of cRel, a close homolog of RelA.

Multiplex profiling of developmental cis-regulatory elements with quantitative single-cell expression reporters

The inability to scalably and precisely measure the activity of developmental cis-regulatory elements (CREs) in multicellular systems is a bottleneck in genomics. Here we develop a dual RNA cassette that decouples the detection and quantification tasks inherent to multiplex single-cell reporter assays. The resulting measurement of reporter expression is accurate over multiple orders of magnitude, with a precision approaching the limit set by Poisson counting noise. Together with RNA barcode stabilization via circularization, these scalable single-cell quantitative expression reporters provide high-contrast readouts, analogous to classic in situ assays but entirely from sequencing. Screening >200 regions of accessible chromatin in a multicellular in vitro model of early mammalian development, we identify 13 (8 previously uncharacterized) autonomous and cell-type-specific developmental CREs. We further demonstrate that chimeric CRE pairs generate cognate two-cell-type activity profiles and assess gain- and loss-of-function multicellular expression phenotypes from CRE variants with perturbed transcription factor binding sites. Single-cell quantitative expression reporters can be applied in developmental and multicellular systems to quantitatively characterize native, perturbed and synthetic CREs at scale, with high sensitivity and at single-cell resolution.

Interpreting cis-regulatory mechanisms from genomic deep neural networks using surrogate models

Deep neural networks (DNNs) have greatly advanced the ability to predict genome function from sequence. However, elucidating underlying biological mechanisms from genomic DNNs remains challenging. Existing interpretability methods, such as attribution maps, have their origins in non-biological machine learning applications and therefore have the potential to be improved by incorporating domain-specific interpretation strategies. Here we introduce SQUID, a genomic DNN interpretability framework based on domain-specific surrogate modeling. SQUID approximates genomic DNNs in user-specified regions of sequence space using surrogate models-simpler quantitative models that have inherently interpretable mathematical forms. SQUID leverages domain knowledge to model cis-regulatory mechanisms in genomic DNNs, in particular by removing the confounding effects that nonlinearities and heteroscedastic noise in functional genomics data can have on model interpretation. Benchmarking analysis on multiple genomic DNNs shows that SQUID, when compared to established interpretability methods, identifies motifs that are more consistent across genomic loci and yields improved single-nucleotide variant-effect predictions. SQUID also supports surrogate models that quantify epistatic interactions within and between cis-regulatory elements, as well as global explanations of cis-regulatory mechanisms across sequence contexts. SQUID thus advances the ability to mechanistically interpret genomic DNNs.

Transfer learning reveals sequence determinants of the quantitative response to transcription factor dosage

Deep learning approaches have made significant advances in predicting cell type-specific chromatin patterns from the identity and arrangement of transcription factor (TF) binding motifs. However, most models have been applied in unperturbed contexts, precluding a predictive understanding of how chromatin state responds to TF perturbation. Here, we used transfer learning to train and interpret deep learning models that use DNA sequence to predict, with accuracy approaching experimental reproducibility, how the concentration of two dosage-sensitive TFs (TWIST1, SOX9) affects regulatory element (RE) chromatin accessibility in facial progenitor cells. High-affinity motifs that allow for heterotypic TF co-binding and are concentrated at the center of REs buffer against quantitative changes in TF dosage and strongly predict unperturbed accessibility. In contrast, motifs with low-affinity or homotypic binding distributed throughout REs lead to sensitive responses with minimal contributions to unperturbed accessibility. Both buffering and sensitizing features show signatures of purifying selection. We validated these predictive sequence features using reporter assays and showed that a biophysical model of TF-nucleosome competition can explain the sensitizing effect of low-affinity motifs. Our approach of combining transfer learning and quantitative measurements of the chromatin response to TF dosage therefore represents a powerful method to reveal additional layers of the cis-regulatory code.

Recruitment of CTCF to the SIRT1 promoter after Oxidative Stress mediates Cardioprotective Transcription

Because most DNA-binding transcription factors (dbTFs), including the architectural regulator CTCF, bind RNA and exhibit di-/multimerization, a central conundrum is whether these distinct properties are regulated post-transcriptionally to modulate transcriptional programs. Here, investigating stress-dependent activation of SIRT1, encoding an evolutionarily-conserved protein deacetylase, we show that induced phosphorylation of CTCF acts as a rheostat to permit CTCF occupancy of low-affinity promoter DNA sites to precisely the levels necessary. This CTCF recruitment to the SIRT1 promoter is eliciting a cardioprotective cardiomyocyte transcriptional activation program and provides resilience against the stress of the beating heart in vivo . Mice harboring a mutation in the conserved low-affinity CTCF promoter binding site exhibit an altered, cardiomyocyte-specific transcriptional program and a systolic heart failure phenotype. This transcriptional role for CTCF reveals that a covalent dbTF modification regulating signal-dependent transcription serves as a previously unsuspected component of the oxidative stress response.

Context dependent activity of p63-bound gene regulatory elements

The p53 family of transcription factors regulate numerous organismal processes including the development of skin and limbs, ciliogenesis, and preservation of genetic integrity and tumor suppression. p53 family members control these processes and gene expression networks through engagement with DNA sequences within gene regulatory elements. Whereas p53 binding to its cognate recognition sequence is strongly associated with transcriptional activation, p63 can mediate both activation and repression. How the DNA sequence of p63-bound gene regulatory elements is linked to these varied activities is not yet understood. Here, we use massively parallel reporter assays (MPRA) in a range of cellular and genetic contexts to investigate the influence of DNA sequence on p63-mediated transcription. Most regulatory elements with a p63 response element motif (p63RE) activate transcription, with those sites bound by p63 more frequently or adhering closer to canonical p53 family response element sequences driving higher transcriptional output. The most active regulatory elements are those also capable of binding p53. Elements uniquely bound by p63 have varied activity, with p63RE-mediated repression associated with lower overall GC content in flanking sequences. Comparison of activity across cell lines suggests differential activity of elements may be regulated by a combination of p63 abundance or context-specific cofactors. Finally, changes in p63 isoform expression dramatically alters regulatory element activity, primarily shifting inactive elements towards a strong p63-dependent activity. Our analysis of p63-bound gene regulatory elements provides new insight into how sequence, cellular context, and other transcription factors influence p63-dependent transcription. These studies provide a framework for understanding how p63 genomic binding locally regulates transcription. Additionally, these results can be extended to investigate the influence of sequence content, genomic context, chromatin structure on the interplay between p63 isoforms and p53 family paralogs.

Exhaustive identification of genome-wide binding events of transcriptional regulators

Genome-wide binding assays aspire to map the complete binding pattern of gene regulators. Common practice relies on replication-duplicates or triplicates-and high stringency statistics to favor false negatives over false positives. Here we show that duplicates and triplicates of CUT&RUN are not sufficient to discover the entire activity of transcriptional regulators. We introduce ICEBERG (Increased Capture of Enrichment By Exhaustive Replicate aGgregation), a pipeline that harnesses large numbers of CUT&RUN replicates to discover the full set of binding events and chart the line between false positives and false negatives. We employed ICEBERG to map the full set of H3K4me3-marked regions, the targets of the co-factor β-catenin, and those of the transcription factor TBX3, in human colorectal cancer cells. The ICEBERG datasets allow benchmarking of individual replicates, comparing the performance of peak calling and replication approaches, and expose the arbitrary nature of strategies to identify reproducible peaks. Instead of a static view of genomic targets, ICEBERG establishes a spectrum of detection probabilities across the genome for a given factor, underlying the intrinsic dynamicity of its mechanism of action, and permitting to distinguish frequent from rare regulation events. Finally, ICEBERG discovered instances, undetectable with other approaches, that underlie novel mechanisms of colorectal cancer progression.

Cis-regulatory interfaces reveal the molecular mechanisms underlying the notochord gene regulatory network of Ciona

Tissue-specific gene expression is fundamental in development and evolution, and is mediated by transcription factors and by the cis-regulatory regions (enhancers) that they control. Transcription factors and their respective tissue-specific enhancers are essential components of gene regulatory networks responsible for the development of tissues and organs. Although numerous transcription factors have been characterized from different organisms, the knowledge of the enhancers responsible for their tissue-specific expression remains fragmentary. Here we use Ciona to study the enhancers associated with ten transcription factors expressed in the notochord, an evolutionary hallmark of the chordate phylum. Our results illustrate how two evolutionarily conserved transcription factors, Brachyury and Foxa2, coordinate the deployment of other notochord transcription factors. The results of these detailed cis-regulatory analyses delineate a high-resolution view of the essential notochord gene regulatory network of Ciona, and provide a reference for studies of transcription factors, enhancers, and their roles in development, disease, and evolution.

A change in cis-regulatory logic underlying obligate versus facultative muscle multinucleation in chordates

Vertebrates and tunicates are sister groups that share a common fusogenic factor, Myomaker (Mymk), that drives myoblast fusion and muscle multinucleation. Yet they are divergent in when and where they express Mymk. In vertebrates, all developing skeletal muscles express Mymk and are obligately multinucleated. In tunicates, Mymk is only expressed in post-metamorphic multinucleated muscles, but is absent from mononucleated larval muscles. In this study, we demonstrate that cis-regulatory sequence differences in the promoter region of Mymk underlie the different spatiotemporal patterns of its transcriptional activation in tunicates and vertebrates. While in vertebrates Myogenic Regulatory Factors (MRFs) like MyoD1 alone are required and sufficient for Mymk transcription in all skeletal muscles, we show that transcription of Mymk in post-metamorphic muscles of the tunicate Ciona requires the combinatorial activity of MRF/MyoD and Early B-Cell Factor (Ebf). This macroevolutionary difference appears to be encoded in cis, likely due to the presence of a putative Ebf binding site adjacent to predicted MRF binding sites in the Ciona Mymk promoter. We further discuss how Mymk and myoblast fusion might have been regulated in the last common ancestor of tunicates and vertebrates, for which we propose two models.

Overlapping binding sites underlie TF genomic occupancy

Sequence-specific DNA binding by transcription factors (TFs) is a crucial step in gene regulation. However, current high-throughput in vitro approaches cannot reliably detect lower affinity TF-DNA interactions, which play key roles in gene regulation. Here, we developed PADIT-seq ( p rotein a ffinity to D NA by in vitro transcription and RNA seq uencing) to assay TF binding preferences to all 10-bp DNA sequences at far greater sensitivity than prior approaches. The expanded catalogs of low affinity DNA binding sites for the human TFs HOXD13 and EGR1 revealed that nucleotides flanking high affinity DNA binding sites create overlapping lower affinity sites that together modulate TF genomic occupancy in vivo . Formation of such extended recognition sequences stems from an inherent property of TF binding sites to interweave each other and expands the genomic sequence space for identifying noncoding variants that directly alter TF binding.

One-Sentence Summary

Overlapping DNA binding sites underlie TF genomic occupancy through their inherent propensity to interweave each other.

Interpreting cis-regulatory mechanisms from genomic deep neural networks using surrogate models

Deep neural networks (DNNs) have greatly advanced the ability to predict genome function from sequence. Interpreting genomic DNNs in terms of biological mechanisms, however, remains difficult. Here we introduce SQUID, a genomic DNN interpretability framework based on surrogate modeling. SQUID approximates genomic DNNs in user-specified regions of sequence space using surrogate models, i.e., simpler models that are mechanistically interpretable. Importantly, SQUID removes the confounding effects that nonlinearities and heteroscedastic noise in functional genomics data can have on model interpretation. Benchmarking analysis on multiple genomic DNNs shows that SQUID, when compared to established interpretability methods, identifies motifs that are more consistent across genomic loci and yields improved single-nucleotide variant-effect predictions. SQUID also supports surrogate models that quantify epistatic interactions within and between cis-regulatory elements. SQUID thus advances the ability to mechanistically interpret genomic DNNs.

Proformer: a hybrid macaron transformer model predicts expression values from promoter sequences

The breakthrough high-throughput measurement of the cis-regulatory activity of millions of randomly generated promoters provides an unprecedented opportunity to systematically decode the cis-regulatory logic that determines the expression values. We developed an end-to-end transformer encoder architecture named Proformer to predict the expression values from DNA sequences. Proformer used a Macaron-like Transformer encoder architecture, where two half-step feed forward (FFN) layers were placed at the beginning and the end of each encoder block, and a separable 1D convolution layer was inserted after the first FFN layer and in front of the multi-head attention layer. The sliding k-mers from one-hot encoded sequences were mapped onto a continuous embedding, combined with the learned positional embedding and strand embedding (forward strand vs. reverse complemented strand) as the sequence input. Moreover, Proformer introduced multiple expression heads with mask filling to prevent the transformer models from collapsing when training on relatively small amount of data. We empirically determined that this design had significantly better performance than the conventional design such as using the global pooling layer as the output layer for the regression task. These analyses support the notion that Proformer provides a novel method of learning and enhances our understanding of how cis-regulatory sequences determine the expression values.

The contributions of DNA accessibility and transcription factor occupancy to enhancer activity during cellular differentiation

During gene regulation, DNA accessibility is thought to limit the availability of transcription factor (TF) binding sites, while TFs can increase DNA accessibility to recruit additional factors that upregulate gene expression. Given this interplay, the causative regulatory events in the modulation of gene expression remain unknown for the vast majority of genes. We utilized deeply sequenced ATAC-Seq data and site-specific knock-in reporter genes to investigate the relationship between the binding-site resolution dynamics of DNA accessibility and the expression dynamics of the enhancers of Cebpa during macrophage-neutrophil differentiation. While the enhancers upregulate reporter expression during the earliest stages of differentiation, there is little corresponding increase in their total accessibility. Conversely, total accessibility peaks during the last stages of differentiation without any increase in enhancer activity. The accessibility of positions neighboring C/EBP-family TF binding sites, which indicates TF occupancy, does increase significantly during early differentiation, showing that the early upregulation of enhancer activity is driven by TF binding. These results imply that a generalized increase in DNA accessibility is not sufficient, and binding by enhancer-specific TFs is necessary, for the upregulation of gene expression. Additionally, high-coverage ATAC-Seq combined with time-series expression data can infer the sequence of regulatory events at binding-site resolution.

DNA-guided transcription factor cooperativity shapes face and limb mesenchyme

Transcription factors (TFs) can define distinct cellular identities despite nearly identical DNA-binding specificities. One mechanism for achieving regulatory specificity is DNA-guided TF cooperativity. Although in vitro studies suggest that it may be common, examples of such cooperativity remain scarce in cellular contexts. Here, we demonstrate how "Coordinator," a long DNA motif composed of common motifs bound by many basic helix-loop-helix (bHLH) and homeodomain (HD) TFs, uniquely defines the regulatory regions of embryonic face and limb mesenchyme. Coordinator guides cooperative and selective binding between the bHLH family mesenchymal regulator TWIST1 and a collective of HD factors associated with regional identities in the face and limb. TWIST1 is required for HD binding and open chromatin at Coordinator sites, whereas HD factors stabilize TWIST1 occupancy at Coordinator and titrate it away from HD-independent sites. This cooperativity results in the shared regulation of genes involved in cell-type and positional identities and ultimately shapes facial morphology and evolution.

Affinity-optimizing enhancer variants disrupt development

Enhancers control the location and timing of gene expression and contain the majority of variants associated with disease1-3. The ZRS is arguably the most well-studied vertebrate enhancer and mediates the expression of Shh in the developing limb4. Thirty-one human single-nucleotide variants (SNVs) within the ZRS are associated with polydactyly4-6. However, how this enhancer encodes tissue-specific activity, and the mechanisms by which SNVs alter the number of digits, are poorly understood. Here we show that the ETS sites within the ZRS are low affinity, and identify a functional ETS site, ETS-A, with extremely low affinity. Two human SNVs and a synthetic variant optimize the binding affinity of ETS-A subtly from 15% to around 25% relative to the strongest ETS binding sequence, and cause polydactyly with the same penetrance and severity. A greater increase in affinity results in phenotypes that are more penetrant and more severe. Affinity-optimizing SNVs in other ETS sites in the ZRS, as well as in ETS, interferon regulatory factor (IRF), HOX and activator protein 1 (AP-1) sites within a wide variety of enhancers, cause gain-of-function gene expression. The prevalence of binding sites with suboptimal affinity in enhancers creates a vulnerability in genomes whereby SNVs that optimize affinity, even slightly, can be pathogenic. Searching for affinity-optimizing SNVs in genomes could provide a mechanistic approach to identify causal variants that underlie enhanceropathies.