See(quence) and ye shall find: higher-order genome folding in intact single cells

Abstract

Payne et al. (2020) combine in situ imaging and ex situ sequencing via spatially resolved unique molecular barcodes to query higher-order genome folding patterns in intact single nuclei from mouse embryos and human fibroblasts.

It has long been known that DNA is folded non-randomly in the mammalian nucleus. Over the last two decades, chromosome-conformation-capture (3C) experiments based on molecular proximity ligation have been combined with high-throughput sequencing to enable the inquiry of physical contact frequency between distal genomic loci genome-wide. Early applications of Hi-C confirmed the presence of chromosome territories and revealed the features of “A” and “B” compartments, while later improvements to 3C-based methods led to the genome-wide detection of topologically associating domains (TADs) (Dixon et al., 2012), subTADs (Phillips-Cremins et al., 2013), and long-range looping interactions (Rao et al., 2014) (Figure 1A).

Figure 1. In situ genome sequencing for the inquiry of higher-order genome folding patterns in intact single nuclei.

(A) Resolution of methods used to map the features of higher-order genome folding.

(B) In situ genome sequencing (IGS) overview. (1) Adapters are inserted by Tn5 transposase and DNA is tagged with unique molecular identifiers (UMI) through ligation of DNA hairpins to adaptor sites. (2) Tagged DNA is amplified through rolling circle amplification and (3) UMI coordinates are mapped through sequencing by ligation in situ. (4) Amplicons are recovered and sequenced to generate spatially-resolved sequences.

(C) Future IGS applications may include targeting transposon insertion by fusion to transcription factors or CRISPR-Cas proteins

Hi-C and other 3C-based methods have achieved high (<1 kb genomic bin) resolution by averaging contact frequency across a population of cells. However, how each unique folding pattern varies over time and the extent of heterogeneity in individual cells remains an open question. Single-cell variants of 3C-based methods are limited by low sensitivity for subMb (subMegabase)-scale features and often involve the destruction of the biological sample. More recently, high genome-coverage imaging techniques based on multiplexed fluorescence in situ hybridization (FISH) have detected individual chromosome territories, A/B compartment domains, and some TADs and their boundaries in intact single nuclei. Multiplexed DNA FISH-based approaches visualize chromatin folding with two approaches: either by tiling probes in sequential bins (2-kb up to 1-Mb sized) across genomic regions ranging from 1–30+ Mb in size (Bintu et al., 2018; Mateo et al., 2019; Nir et al., 2018; Su et al., 2020; Takei et al., 2021), or by simultaneously imaging barcoded probes to spatially register 1–3-Mb-sized bins distributed across the genome (Nguyen et al., 2020; Su et al., 2020; Takei et al., 2021). Current multiplexed probe-based imaging methods require the design of FISH probes to pre-selected regions and, therefore, depend on prior knowledge of the genome sequence. In new work, Payne et al. (2020) report the development of in situ genome sequencing (IGS), a sequence-agnostic method to image and sequence hundreds to thousands of genomic fragments in the same nucleus in intact single cells.

IGS introduces unique molecular identifiers (UMIs) at thousands of loci in situ to bridge imaging and sequencing within the same nucleus. Several molecular innovations contribute to this technical advance. First, IGS employs a Tn5 transposase to fragment and insert sequencing adapters into the genomic DNA in situ (Figure 1B). By contrast to ATAC-seq, nuclear DNA is chemically denatured to remove histones, and the authors provide evidence that adapters integrate randomly without bias to transcription start sites. Second, DNA hairpin ligation and rolling circle amplification were used to introduce (1) a UMI barcode, (2) primer sites for rolling circle amplification of the UMI, adaptor, and unique genome sequence in situ, (3) primer sites for UMI sequencing in situ, and (4) primer sites for sequencing of both the UMI and genomic sequence upon extraction, or ex situ (Figure 1B). Importantly, by introducing the UMI in the DNA hairpin, and not by transposon-directed deposition, investigators can titrate the hairpin concentration to allow for precise calibration of UMI density to avoid overcrowding in the nucleus during the in situ sequencing step. Third, UMIs are sequenced by ligation in situ using fluorescence imaging to register their spatial subnuclear location. Finally, after DNA extraction, the amplified genetic region and UMIs are sequenced ex situ by conventional paired-end sequencing. Together, both in situ and ex situ reads are computationally registered by their common UMIs to reveal the nuclear positioning of each genomic bin.

In this first instantiation of IGS, hundreds to thousands of UMIs were spatially resolved in single nuclei from intact human fibroblasts and mouse embryos. Despite the limited number of reads, IGS data binned at Mb resolution confirmed the presence of chromosome territories and recapitulated some aspects of A/B compartment patterns (Figure 1A). Moreover, imaging and sequencing the same genome facilitated the assignment of haplotypes to the allele-specific chromosome territories. By distinguishing haplotypes in 4-cell stage mouse embryos, the authors report that chromosome positions in sister nuclei showed greater concordance than those of cousin nuclei, thus suggesting that memory of nuclear positioning might be transmitted through mitosis in early development. Overall, these results demonstrate that IGS yields insights into global nuclear organization by localizing specific A/B compartments and allele-specific chromosome territories, defining haplotypes, and relating specific genomic bins to immunostained subnuclear structures.

Within the fixed volume of the nucleus, IGS can distinguish up to 4,000 UMIs per 3 billion base pair genome. Thus, UMI density limits the capacity of IGS to detect 3D genome features at subMb scales, i.e., TADs, subTADs, and loops. UMIs were detectable at an order-of-magnitude higher frequency in mouse embryo cells due to the larger nuclear volume compared to fibroblasts. The authors raise the compelling possibility of combining IGS with expansion microscopy, a method of physically enlarging specimens that preserves nanoscale features, as a means of enlarging nuclear volume. Therefore, there is great potential for future IGS variants that increase the number of spatially resolved UMIs per nucleus to enhance subMb-scale feature detection. Future efforts to increase the throughput of IGS will catalyze the quantification of the full range of single-cell variation for each distinct folding pattern and integrate such variation with genome-wide transcription and epigenetic profiles in the same cells.

Future IGS variants (Figure 1C) that couple the transposase to a CRISPR-Cas system would enable guide RNA-directed adaptor integration to specific genomic locations. Fusion of the transposase to transcription factors or protein components of subnuclear bodies could enable IGS to simultaneously probe where proteins bind on the linear genome and where those binding sites spatially localize with respect to folding patterns. Hi-C and imaging data to date together support a model in which distal genetic and epigenetic features spatially co-localize in pairwise or multifactorial interactions at the nuclear interior, subnuclear bodies, or the periphery to participate in complex spatiotemporally regulated gene expression patterns. Yet, the cause and effect relationship between each folding pattern and gene expression regulation across developmental and disease contexts remains hotly debated. We envision the power of IGS and other imaging-based methods will be realized by coupling single-cell architecture maps with perturbations. It will be critical to pair the hypothesized perturbative effect with consideration of the architectural length scale assayable by a given imaging technique. Unifying Hi-C, single-cell imaging (Bintu et al., 2018; Mateo et al., 2019; Nir et al., 2018; Su et al., 2020; Takei et al., 2021) and in situ sequencing (Nguyen et al., 2020; Payne et al., 2020) approaches will catalyze the endeavor to unravel and understand the mammalian genome’s structure-function relationship.