Spatial Sequencing Maps Native DNA Modifications and Transcriptomes in Mammalian Tissues

A research team led by scientists reporting in Cell Research has established a spatial multi-omics sequencing platform called Spatial-DiSMHT-seq, which jointly profiles native DNA cytosine modifications alongside full transcriptomes directly within intact mammalian tissues. Cytosine modifications such as 5-methylcytosine, commonly designated 5mC, and 5-hydroxymethylcytosine, or 5hmC, serve as foundational epigenetic marks that regulate cell-type-specific gene expression programs during organismal development and tissue homeostasis. While recent advances in single-cell epigenomics have begun unraveling cellular heterogeneity, existing spatial methylome profiling strategies have consistently lacked the resolution required to distinguish and capture both 5mC and 5hmC modifications concurrently within complex spatial microenvironments. The newly developed technology bypasses traditional bisulfite and enzymatic conversion protocols, using direct single-molecule long-read sequencing to preserve both native modification states and spatial tissue organization.

To overcome chemical degradation and amplification biases inherent to standard bisulfite-based epigenomic assays, the developers incorporated third-generation Oxford Nanopore sequencing into a spatially resolved microfluidic barcoding workflow. Conventional chemical treatments often fragment genomic DNA and fail to differentiate between 5mC and 5hmC without pairing complex chemical steps, whereas direct nanopore electrical signal recording recognizes both epigenetic marks on intact single molecules. In the Spatial-DiSMHT-seq workflow, frozen tissue sections undergo gentle fixation followed by tagmentation, where genomic DNA is tagged through Tn5 transposition. Simultaneously, polyadenylated messenger RNAs are captured in situ for reverse transcription using a specialized biotinylated DNA adaptor containing a universal ligation linker and a poly-T sequence, setting up both nucleic acid species for parallel spatial registration.

Spatial indexing within the intact tissue architecture is achieved through deterministic barcoding in tissue, an approach adapted from microfluidic DBiT-seq technology. Within microfluidic flow channels placed directly over the tissue slice, both transposed genomic DNA fragments and synthesized complementary DNA molecules are sequentially ligated to two distinct sets of spatial barcodes, designated barcode A and barcode B. Following spatial barcode addition, specific library preparation adapters for Oxford Nanopore sequencing are ligated to the molecules before the cellular crosslinks are reversed. This reversal releases the barcoded genomic DNA and complementary DNA into solution. The biotin-labeled cDNA is subsequently isolated and enriched using streptavidin-coated magnetic beads, underwent template switching, and processed for high-throughput next-generation sequencing, while the genomic DNA supernatant is harvested separately for PCR-free long-read nanopore sequencing.

The researchers evaluated the sequencing performance and spatial resolution of Spatial-DiSMHT-seq across multiple mammalian biological specimens, including Carnegie stage 15 to 16 human embryos, embryonic day 9.5 mouse embryos, and adult mouse testis sections. The assays were conducted at spatial resolutions of either 10 or 20 micrometers per pixel. Across these diverse tissue datasets, the nanopore sequencing reads demonstrated high sequence quality, yielding mean Phred quality scores ranging from 17 to 22. Median unique-read counts fell between 113 and 452 reads per pixel, while median read lengths ranged from 533 base pairs to 972 base pairs. For reads exceeding 500 base pairs in length, alignment rates to the respective human and mouse reference genomes were exceptionally high, landing between 98.9 percent and 99.3 percent.

Direct base-modification calling from the unamplified nanopore electrical current signals allowed the simultaneous detection of both 5mC and 5hmC across single molecules without requiring separate conversion libraries. On average, the captured sequencing reads covered approximately 496 to 2,272 CpG sites per spatial pixel, with 88.4 percent to 93.7 percent of all mapped reads carrying detectable DNA modification signals. Across the sequenced reads, 86.8 percent to 92.7 percent contained 5mC modifications, whereas 15.7 percent to 35.9 percent contained 5hmC modifications. Genome-wide analysis indicated that global CpG 5mC levels remained stable between 69 percent and 74 percent across all analyzed tissue sections, while global CpG 5hmC levels ranged from 2.6 percent to 3.9 percent and displayed distinct tissue-associated spatial distribution patterns.

To validate the biological fidelity and quantitative precision of Spatial-DiSMHT-seq measurements, the investigators compared the spatial methylome data against established orthogonal genomic profiling methods. Measurements of 5mC obtained by Spatial-DiSMHT-seq demonstrated strong statistical concordance with conventional whole-genome bisulfite sequencing, exhibiting a Pearson correlation coefficient of r = 0.866. Because whole-genome bisulfite sequencing cannot differentiate 5mC from 5hmC, the authors benchmarked 5hmC calls against independent enzymatic 5hmC sequencing datasets in embryonic day 9.5 mouse embryos, finding moderate agreement with a Pearson correlation coefficient of r = 0.54. Similarly, regional 5hmC levels detected across human embryonic brain sections correlated moderately with published fetal brain datasets derived from paired bisulfite and oxidative bisulfite profiling, showing a Pearson correlation coefficient of r = 0.53.

Beyond benchmarking against independent technological platforms, Spatial-DiSMHT-seq successfully reproduced known structural distribution properties of 5hmC originally identified in fetal brain Tet-assisted bisulfite sequencing datasets. The spatial nanopore calls showed pronounced 5hmC enrichment within actively transcribed gene bodies and putative active enhancer sequences, alongside marked depletion of 5hmC at promoter CpG islands belonging to active genes. Analytical sensitivity tests confirmed that the identified patterns of 5mC differentially methylated regions and 5hmC differentially hydroxymethylated regions remained consistent and stable across different analytical settings. Concurrently, the platform captured robust transcriptomic profiles from the exact same spatial coordinates, recovering median gene counts per pixel of 4,400 with 7,730 unique molecular identifiers in mouse testis at 10-micrometer resolution, and up to 5,623 genes with 12,922 unique molecular identifiers in embryonic tissues at 20-micrometer resolution.

When applied to adult mouse testis sections, Spatial-DiSMHT-seq captured extensive epigenetic remodeling during germline maturation and meiotic division. RNA-based clustering identified 10 spatial clusters that aligned closely with known histological testicular compartments. The spatial modification data recapitulated the genome-wide wave of DNA demethylation that occurs at meiotic entry, revealing corresponding single-molecule shifts in 5mC and 5hmC composition. The platform also identified expected DNA modification features at meiotic recombination hotspots and transposable elements during active meiosis. In embryonic day 9.5 mouse embryos, the platform resolved 14 RNA-defined anatomical clusters, identifying reduced 5mC and elevated 5hmC levels at active tissue-specific enhancers annotated in EnhancerAtlas 2.0. Notably, spots enriched for the dorsal aorta displayed a distinct low-5mC and low-5hmC profile characterized by the expression of Dnmt1 alongside low expression of de novo methyltransferases Dnmt3a and Dnmt3b, paired with high expression of the proliferation marker Mki67, indicating passive replication-coupled epigenetic dilution.

Extending the platform to Carnegie stage 15 to 16 human embryo sections revealed 18 distinct spatial clusters, linking tissue-specific 5mC differentially methylated regions and high-5hmC clusters to specialized cell lineages such as cardiomyocytes and hepatoblasts. Similar to the mouse dorsal aorta, human hepatoblast-enriched areas and limb vascular plexus spots displayed reduced 5mC and 5hmC states coupled with high proliferation signatures. Furthermore, in the human embryonic hindbrain, the authors reconstructed dynamic spatiotemporal trajectories of neurogenesis from radial glia to immature neurons. Along this dorsal-to-ventral differentiation axis, global 5mC levels gradually decreased while 5hmC levels increased, a pattern mirrored across the gene bodies of neuronal identity-associated genes. Radial glia exhibited higher expression of DNA methylation maintenance machinery, whereas differentiating immature neurons up-regulated de novo methyltransferases and TET enzymes.

The study authors emphasize several important technical and analytical constraints inherent to the current methodology. Because Spatial-DiSMHT-seq uses a PCR-free preparation to preserve native modifications on single unamplified molecules, read counts and CpG coverage per individual pixel remain relatively sparse compared to amplified single-cell platforms. This coverage sparsity reduces statistical power when detecting localized modifications at small regulatory sequences and rare cell types, while physical cellular admixture within 10- or 20-micrometer pixels can blend distinct epigenetic signals. Furthermore, technical replicates showed perfect reproducibility for transcriptomic profiles with r = 1.00 and high concordance for 5mC levels with r = 0.93, but lower quantitative reproducibility for 5hmC with r = 0.60, requiring caution when interpreting subtle localized 5hmC changes until deeper sequencing coverage and expanded biological replication are achieved.

Subject of Research: Spatial single-molecule co-profiling of DNA methylation and transcription in mammalian tissues

Article Title: Spatial DNA-directed single-molecule methylome and transcriptome co-profiling in mammalian tissues

Article References: Yu, H., Yuan, S., Tang, X., Lu, Y., He, A., Wu, H., Zhou, J., Liang, G., Ruan, K., Li, N., Huang, X., Jiang, X., Chen, H., Liu, C., & Li, W. (2026). Spatial DNA-directed single-molecule methylome and transcriptome co-profiling in mammalian tissues. Cell Research. https://doi.org/10.1038/s41422-026-01320-x

Image Credits: AI Generated

DOI: 10.1038/s41422-026-01320-x

Keywords: Spatial-DiSMHT-seq, DNA methylation, 5-methylcytosine, 5-hydroxymethylcytosine, spatial transcriptomics, Oxford Nanopore sequencing, single-molecule sequencing, epigenomics, mammalian embryogenesis, DBiT-seq, mouse embryo development, human embryonic development, hindbrain neurogenesis, PCR-free sequencing, epigenetic remodeling, microfluidic barcoding