In situ genome sequencing resolves DNA sequence and structure in intact biological samples

Abstract

Understanding genome organization requires integration of DNA sequence and three-dimensional spatial context; however, existing genome-wide methods lack either base pair sequence resolution or direct spatial localization. Here, we describe in situ genome sequencing (IGS), a method for simultaneously sequencing and imaging genomes within intact biological samples. We applied IGS to human fibroblasts and early mouse embryos, spatially localizing thousands of genomic loci in individual nuclei. Using these data, we characterized parent-specific changes in genome structure across embryonic stages, revealed single-cell chromatin domains in zygotes, and uncovered epigenetic memory of global chromosome positioning within individual embryos. These results demonstrate how IGS can directly connect sequence and structure across length scales from single base pairs to whole organisms.

Figure 1:. Method for in situ genome sequencing.

(A) In situ genomic DNA library construction. i) Legend. ii) Adaptor insertion. iii) Insert circularization by hairpin ligation, followed by in situ rolling circle amplification (RCA). iv) Clonal amplicons contain primers for in situ and ex situ sequencing. (B) Workflow for in situ genome sequencing. i) In situ sequencing localizes unique molecular identifiers (UMIs). 4-channel imaging of two representative amplicons over 18 rounds of in situ sequencing. ii) Amplicon dissociation following in situ sequencing. iii) PCR and ex situ sequencing of amplicons associates genomic sequences with UMIs. (C) Top: paired-end sequences are spatially localized by integrating in situ and ex situ sequencing data. Bottom: matched reads, colored by chromosome, are overlaid on their imaged amplicon library (below). (D) In situ sequenced nuclei from cultured fibroblasts and intact embryos at the PN4 zygote, late 2-cell, and early 4-cell stages, with spatially-localized reads colored by chromosome.

Figure 2:. IGS characterizes spatial features of the human genome.

(A) DAPI stain of a PGP1f nucleus after in situ library construction. (B) 601 spatially-localized reads in the same PGP1f nucleus, colored by chromosome. (C) Exploded view reveals conformations of chromosome territories, shown as in situ reads (balls) connected according to sequential genomic position (sticks). (D) Genome-wide population mean pairwise distance matrix of 106 PGP1f cells binned at 10 Mb. (E) Chromosome size vs. normalized mean radial distance from the nuclear center for 106 diploid-resolved PGP1f cells. Error bars denote 95% confidence interval of the mean determined by bootstrapping. (F) The 103 most abundant repetitive elements ordered by radial bias, defined as the variability of binned distances relative to a permuted background from the nuclear center for 106 PGP1f cells. The dashed grey line represents the threshold for elements shown in (G). (G) Radial enrichment/depletion by binned distance from the nuclear center for the repetitive elements with the strongest radial bias from (F). (H) Ball-and-stick models for Chr 1–4 in the same single-cell, demonstrating spatial polarization between the p and q arms of each chromosome. (I) Genomic distance vs. spatial distance for Chr 1, distinguishing intra-arm and inter-arm measurements. Error bars: standard deviation. Dashed: range in which both measurements can be compared at reasonable sampling depth (n > 20 per 1 Mb bin). (J) Intra-arm and inter-arm distance distributions in the dashed range in (I) are distributed differently (n = 819 intra-arm, 766 inter-arm, 144 Chr 1 territories, K-S test, p < 10⁻¹⁶). Violin plot indicates median and range.

Figure 3:. IGS enables high-resolution genomic and spatial profiling of intact early mouse embryos

(A) Workflow. B6C3F1 x B6D2F1 embryos at the zygote, 2-cell, and 4-cell stages are pooled, fixed and immobilized in a polyacrylamide gel. Following in situ sequencing, DAPI and immunofluorescence staining of CENP-A and Lamin-B1 are performed. (B) Representative zygote with 7,374 spatially-localized reads colored by chromosome (left), distance to the nuclear lamina (middle), and distance to nearest nucleolus precursor body (right). (C) Amplicons from (B), with reads colored by parental haplotype assignment for the intact embryo (top), reads colored by genomic position for Chr 3 homologs (middle), and reads colored by parental haplotype assignment for Chr 3 homologs (bottom). Boxes show two haplotype-informative Chr 3 SNPs. (D) An exploded view of chromosome territories from (B) for the maternal (left) and paternal pronuclei (right).

Figure 4:. IGS characterizes developmental transitions in embryonic genome organization.

(A) Exploded view of a single nucleus from a 2-cell embryo colored by chromosome territories, haplotype, centromere-telomere position, and GC content. (B) 2-cell embryo with spatially-localized reads colored by parental haplotype assignment (left) and haplotype separation score (right). (C) Boxplots showing mean haplotype separation score per cell across developmental stages (left; K-S test, p < 10⁻⁴ and p < 10⁻⁸). Grey dots represent mean scores of single cells. Distribution mean (red line), 95% confidence interval, (red box), and 1 standard deviation (blue box) are indicated. Two cells representing extreme scores (> 1 SD) are shown (right). (D) Nucleus from (A) with spatially-localized reads colored by centromere-telomere position, shown from two angles 90 degrees apart (left). Black dots indicate the position of CENP-A as identified from immunostains. Chr 1 and Chr 15 homologs from this cell are shown (right) to illustrate the Rabl-like configuration. (E) Mean centromere-telomere position of spatial neighbors as a function of centromere-telomere position for each stage. (F) Chr 12 homologs from a representative zygote with spatially-localized reads colored by GC content (left) and distance to lamina (right). (G) Plots showing the relationship between GC content and average distance to the nuclear lamina for 1 Mb bins in Chr 12 of the maternal and paternal zygotic pronuclei. Zygotic lamina-associated domains (LADs) defined by DamID are displayed below. (H) Boxplots showing Spearman’s ρ between GC content and distance to lamina for 1 Mb bins, partitioned by haplotype and developmental stage (K-S test, p < 10⁻⁵ and n.s.). Dots represent single chromosomes. Distribution mean (red line), 95% confidence interval, (red box), and 1 standard deviation (blue box) are indicated. n = 24 zygotes, 40 2-cell, 49 4-cell nuclei for all panels. Scale bars: 5 μm in all directions.

Figure 5:. IGS reveals single-cell domains in zygotes.

(A) Global relationship between genomic and spatial distance in zygotes for all chromosomes, distinguishing the parental genomes. (B) Visualization of Chr 11 homologs in two zygotes according to parent-of-origin. (C) Population ensemble mean spatial distance matrix for paternal Chr 11, constructed at 2.5 Mb resolution (24 zygotic pronuclei, 2317 reads). (D) Comparison across measurement modalities for the population of paternal zygotic Chr 11. Top row: Hi-C defined eigenvalues and compartment calls. Middle row: DamID-defined population lamina associated domains. Bottom row: lamin-proximal and lamin-distal regions defined with IGS (24 zygotic pronuclei, 2317 reads). (E) Top left: single-cell mean distance matrix for paternal Chr 11 in a representative zygote, with single-cell domain boundaries (SCDs) marked below (263 reads). Top right: visualization of individual paternal SCDs in the same zygote. To assist visualization, two SCDs are shown in color (purple, gold), while the remaining SCDs are shown in grey. Bottom left: single-cell mean distance matrix for paternal Chr 11 in a second representative zygote, with single-cell domain boundaries marked below (213 reads). Bottom right: visualization of three paternal SCDs in the second zygote. To assist visualization, three SCDs are shown in color (magenta, lime, cyan) while the remaining SCDs are shown in grey. (F) Comparison of single-cell and ensemble domain boundary strengths spanning all detectable boundaries in Chr 1–19+X (74 ensemble boundaries, 1057 single-cell boundaries, K-S test, p < 10⁻¹⁷). (G) Scaled distance from SCD boundary versus observed/expected median distance to nuclear lamina, measured genome-wide (Chr 1–19+X, N = 1262 SCDs). Envelope indicates 95% confidence interval determined by bootstrapping.

Figure 6.. IGS uncovers epigenetic memory of global chromosome positioning within single embryos.

(A) Positioning of Chr 1 and 3 in the cells of 2-cell embryos 37 (top) and 41 (bottom). (B) Pairwise correlations between autosome distance matrices for the cells in (A). Intra-embryo and inter-embryo correlations are shown in blue and orange, respectively. (C) Probability distributions of correlations between autosome distance matrices for intra-embryo and inter-embryo pairs of cells among 2-cell embryos. K-S test, p < 10⁻¹⁵; n = 20 intra-embryo pairs and n = 760 inter-embryo pairs, among 20 2-cell embryos. (D) Positioning of Chr 2 and 4 in the cells of 4-cell embryo 45. Pairs of cells are putatively classified as sister and cousin cells based on correlation of global chromosome positioning, with the most correlated pair classified as sisters. Correlations between sister and cousin cells are shown in blue and red, respectively. (E) Probability distributions of correlations between autosome distance matrices for pairs of putative sister cells, cousin cells, and inter-embryo pairs of cells among 4-cell embryos. K-S test, p < 10⁻¹⁴ for sisters vs. inter-embryo and p < 10⁻³ for cousins vs. inter-embryo; n = 18 sister pairs, n = 36 cousin pairs, and n = 933 inter-embryo pairs, among 13 4-cell embryos. (F) Model of epigenetic memory transmission within clonal lineages.