Spatial epigenomics

Epigenomic assays ask which parts of the genome are open, which carry particular histone modifications, and where DNA is methylated. Performed on dissociated nuclei, they lose the tissue architecture that determines much of what regulates what. Spatial epigenomics performs the same chemistry on an intact section, so chromatin state keeps its coordinates.

Two chemistries dominate. Transposition-based accessibility profiling uses the Tn5 transposase, which inserts sequencing adapters preferentially into nucleosome-free DNA — the enzyme both cuts and tags in one step, so accessible regions become sequenceable fragments directly. For histone marks, an antibody against the modification of interest is used to tether a protein A-fused transposase or nuclease to its target, so cutting happens only near bound antibody. Both work on fixed tissue.

Location is then encoded by deterministic barcoding rather than by droplets. A microfluidic device lays parallel channels across the section and flows a first set of barcodes; the device is rotated ninety degrees and a second set is flowed, so every intersection receives a unique pair. The section is thereby divided into a grid of pixels of defined size — tens of micrometres in early implementations, single-digit micrometres in later ones — whose addresses are known by construction rather than inferred.

Sparsity is the defining constraint

Accessibility is measured at two copies of the genome per diploid cell, not at the tens or hundreds of copies a transcript can have. There is no amplification of the underlying template before the measurement, so per pixel the data amount to a modest number of fragments distributed across a genome of billions of bases — a matrix that is near-binary and overwhelmingly zero. Analysis compensates by aggregating: pooling reads over peak regions, over gene bodies plus flanking sequence, or over neighbouring pixels and clusters. Consequently statements about a single pixel are weak, and confident results are statements about domains.

What it measures that expression does not

Accessibility and histone marks report regulatory potential and cell state rather than current output, and they change on different timescales from transcription — poised or primed regions can be open without corresponding transcript. Transcription-factor motifs enriched in accessible regions give an inferred regulatory programme, which is a hypothesis about causation that expression data alone cannot supply. Where the two layers are collected on the same section, the useful analysis is the constraint one places on the other.

What this page does not cover

Spatially resolved measurement of transcripts and proteins is a separate technology with different trade-offs, described under spatial transcriptomics. Bisulfite-based DNA methylation mapping does not translate straightforwardly to this format, because the chemistry that converts unmethylated cytosine also degrades DNA; spatial methylation methods exist but are less mature than accessibility and histone-mark profiling. Three-dimensional genome organisation in situ is likewise a distinct and less developed area.

The honest summary of the field’s maturity: the barcoding chemistry is established and reproducible, the biology it can resolve is limited principally by how few molecules exist per cell to measure, and validation against dissociated-nucleus reference data remains a routine part of interpreting a result.