Shankar Balasubramanian

Mapping the molecular identities and functions of cells within their spatial context is key to understanding the complex interplay within and between tissue neighborhoods. A wide range of methods have recently enabled spatial profiling of cellular anatomical contexts, some offering single-cell resolution. These use different barcoding schemes to encode either the location or the identity of target molecules. However, all these technologies face a trade-off between spatial resolution, depth of profiling, and scalability. Here, we present B arcoding by A ctivated L inkage of I ndexes (BALI), a method that uses light to write combinatorial spatial molecular barcodes directly onto target molecules in situ, enabling multiomic profiling by next generation sequencing. A unique feature of BALI is that the user can define the number, size, shape, and resolution of the spatial locations to be interrogated, with the potential to profile millions of distinct regions with subcellular precision. As a proof of concept, we used BALI to capture the transcriptome, chromatin accessibility, or both simultaneously, from distinct areas of the mouse brain in single tissue sections, demonstrating strong concordance with publicly available datasets. We also developed an integrated instrument that automates combinatorial barcode writing on tissue sections, enabling high-throughput profiling. BALI therefore combines high spatial resolution, high throughput, compatibility with standard histological pipelines, and workflow accessibility to enable tunable spatial multi-omic profiling.

Targeted protein degradation can intervene with the function of disease-related proteins, but most current approaches rely on direct ligand engagement of a protein target, limiting their applicability to proteins that are difficult to bind selectively. Here we present a conceptually unique approach to degrade proteins associated with DNA G-quadruplex (G4) secondary structures in a chromatin context. G4s are non-canonical nucleic acid structures that form at regulatory regions of transcriptionally active genes in open chromatin, and are abundant in cancer states. Although many proteins recognize or regulate G4 structures, selectively targeting G4-binding proteins in their native chromatin environment is challenging. Our bifunctional molecules are proteolysis-targeting chimeras that bind naturally occurring G4s, recruit E3 ubiquitin ligases and degrade G4-specific transcription factors and chromatin remodellers such as FUS, SMARCA4 and ATRX. These proteins are important therapeutic targets that play crucial roles in transcription regulation and DNA repair. Our approach has the potential to be exploited in a therapeutic strategy to target diseases characterized by aberrant G4 activity, such as cancers.

DNA G-quadruplexes (G4s) are key structural features in chromatin that are important to genome function. G4s have an apparent capacity to recruit a wide variety of proteins, including chromatin remodelers, yet the molecular basis and biophysical principles governing these interactions remain poorly understood. Here, we sought to build insights into the interactions of chromatin remodeler SMARCA4 with G4s using a biophysical approach. We found that SMARCA4 selectively recognizes the G4 structure over duplex and single-stranded DNA. SMARCA4 binds a wide range of G4s with different topologies and loop lengths with similar low nanomolar affinities. SMARCA4 was also observed to have a longer residency time on the G4 structure compared to that of other known protein-DNA interactions. We also found that the D1 (DExx-c) helicase domain of SMARCA4, which is important for tethering SMARCA4 to chromatinized DNA, was the predominant binding domain for G4 recognition. Our findings reveal new insights into how G4s interact with proteins, which may have important implications for understanding G4-mediated genome mechanisms.

Epigenetic modifications govern chromatin dynamics and cell state. However, current methods cannot simultaneously resolve the presence of multiple DNA modifications at co-occurring chromatin-associated features. It is thus not clear how these features are physically coupled and how their combinations regulate genome function. To address this key question, we report 6-base-CUT&Tag, a method for simultaneous 6-base DNA sequencing at target chromatin features. Using 6-base-CUT&Tag to profile 5-methylcytosine (5mC) and 5-hydroxymethylcytosine (5hmC) at co-occurring histone modifications in mouse embryonic stem cells (mESCs), we identify feature-dependent 5mC/5hmC signatures previously unresolvable with untargeted or bisulfite-based workflows. We show that DNA methylation and hydroxymethylation are specifically coupled with the H3K4me1 mark in mESC enhancers and that H3K4me1-derived signatures robustly distinguish different enhancer functional states.

Background Early cancer detection has the potential to significantly improve treatment outcomes and survival rates. This study investigates the roles of 5-methylcytosine (5mC) and 5-hydroxymethylcytosine (5hmC) as biomarkers for early-stage colorectal cancer (CRC) detection in cell-free DNA (cfDNA). Methods We analyzed cfDNA from 37 treatment-naive CRC patients and 32 healthy controls using duet evoC 6-base whole genome sequencing (A, C, T, G, 5mC, 5hmC). We then built CRC classifiers using differentially methylated regions identified in stage IV CRC tissue samples. We compared the performance of models using 5mC-only, 5hmC-only, modified cytosine, or 5mC and 5hmC features. Results Here we show that models combining measurements of 5mC and 5hmC significantly enhance diagnostic accuracy (AUC = 0.95) compared to traditional approaches that conflate these markers (modified C, AUC = 0.66). Notably, of all the DMRs considered, almost half show an increase in 5hmC at stage I and a corresponding decrease in 5mC at stage IV, suggesting that 5hmC can effectively track regions undergoing demethylation during tumor development. Conclusions These results support the hypothesis that distinguishing between 5mC and 5hmC improves the sensitivity of liquid biopsy tests for early-stage cancer detection.

DNA double-strand breaks (DSBs) are highly deleterious lesions that can trigger cell death or carcinogenesis if unrepaired or misrepaired. In mammals, most DSBs are repaired by non-homologous end joining (NHEJ), which begins when Ku70/80 binds DNA ends and recruits DNA-PKcs to form the DNA-PK holoenzyme. Although recent cryo-EM studies have resolved several NHEJ assemblies, how these factors access DSBs within nucleosomes remains unclear. Here, we present cryo-EM structures of human Ku70/80 and DNA-PK bound to nucleosomes. Ku70/80 binds the DNA end and bends it away from the nucleosome core, while the Ku70 C-terminal SAP domain makes an additional, specific DNA contact. Our DNA-PK–nucleosome structure further reveals the opening of the Ku80 vWA domain, and we show that non-hydrolysable ATP promotes synapsis by stabilising the Ku80-mediated DNA-PK dimer. These structures reveal a model for DSB recognition on nucleosomal DNA and provide insights relevant to targeting NHEJ in cancer therapy.

Natural, covalently modified cytosine bases within genomic DNA function as important epigenetic markers. Approaches for single-base-resolution sequencing of cytosine modifications typically deploy chemistry for modification-selective C-to-T code conversion and can require error-prone subtractive analysis of complex data. Here we report the sequencing of an epigenetic base by exploiting an unnatural base pair system. This approach relies on hydrogen-bonding complementarity between a malononitrile adduct of 5-formylcytosine and protonated 3,7-dideazaadenine. The specificity of this unnatural base pair was studied by biophysical DNA thermal melting analysis and by template-directed incorporation by DNA polymerase enzymes. Base pair selectivity was enhanced by controlling the protonation state of 3,7-dideazaadenine. We exemplify use of this unnatural base pair to sequence 5-formylcytosine in a DNA template using a Sanger-type format. There is scope for this base pair and the general concept to be implemented on further sequencing platforms that exploit Watson–Crick base pairing to directly identify epigenetic bases.

In mammalian genomes, cytosine modifications form a layer of regulatory information alongside the genetic code. Decoding this information is crucial to our understanding of biology and disease. Established sequencing methods cannot simultaneously resolve cytosine’s three most common forms—cytosine (C), 5-methylcytosine (mC), and 5-hydroxymethylcytosine (hmC)—across both strands of the DNA double helix. Thus, how epigenetic information is distributed in DNA remains unclear. Here, we present S trand- Co upled T andem C ytosine H ydroxymethylation and methylation sequencing (SCoTCH-seq): an accurate and quantitative, base-resolution approach to sequence genomes, together with mC and hmC, in both strands of the same DNA fragment. We show that different forms of cytosine combine across the double helix at CpG sites to form discrete information states in the mouse epigenome. These CpG states have distinct genomic distributions—including at promoters, enhancers, and gene bodies—and have different relationships with transcription. We show that while all possible forms of hydroxymethylation occur, hmC is predominantly asymmetric and that different forms of asymmetric hmC are not equivalent. Our findings demonstrate that 5-hydroxymethylcytosine combines with different cytosine variants across the DNA double helix to form distinct states of regulatory information.