Spatially clustered type I interferon responses at injury borderzones

Animals

Adult C57BL/6J mice were purchased from the Jackson Laboratory at 10 weeks of age and were housed in a pathogen-free environment at University of California San Diego (UCSD) facilities. Mice were maintained under a 12 h–12 h light–dark cycle with ad libitum access to normal chow and sterile drinking water at 21 °C with 60% humidity. All animal experiments were approved by the subcommittee on Animal Research Care at UCSD (Institutional Animal Care and Use Committee, S17144). Irf3^{−/−} mice were bred in house from existing colonies. Ccr2^{−/−} (strain 004999), Ifnar1^− (028288), Cgas^{−/−} (026554) and Sting^gt/gt (017537) mice were purchased from Jackson Laboratory. Nuclear reporter mice (tdTom-NLS) were purchased from Jackson Laboratory (025106). Cardiomyocyte-specific nuclear rupture reporter mice were generated by crossing floxed tdTom-NLS with Myh6^cre/+ mice. Irf3^fl/fl were a gift donated from T. Taniguchi. Cell-type-specific transgenics were generated by crossing Irf3^fl/fl mice with transgenic mice expressing Cre recombinase under the control of the following promoters purchased from Jackson Laboratory: Myh6^cre/+ (011038), Col1a1^creERT2/+ (016241), Cx3cr1^creERT2/+ (025524), S100a8^cre/+ (0216141), Tie2^cre/+ (008863). Inducible deletion was accomplished by intraperitoneal injection of tamoxifen 20 mg ml^−1 in corn oil or vehicle injection for 5 consecutive days in Col1a1^creERT2/+ and Cx3cr1^creERT2/+ mice. The mice were given 7 days to recover after the last tamoxifen or vehicle injection before any surgical procedures were conducted. All of the experiments were performed with 12–14-week-old mice and were carried out using age-matched groups, and both male and female mice were used in WT control groups without randomization and without statistical predetermination of sample sizes. All of the animal experiments were approved by the Subcommittee on Animal Research Care at UC San Diego. No field collected samples were used in this study.

Permanent ligation (MI) surgery and cardioprotective therapy

For permanent ligation, mice were intubated and ventilated with 2% isoflurane. Thoracotomy was performed at the fourth left intercostal space was performed to expose the heart and visualize the left anterior descending artery (LAD). The LAD was permanently ligated with an 8-0 nylon suture in mice with MI, and the hearts were collected at various timepoints (1 h, 4 h, day 1, day 3, day 7 or day 28 after surgery). Intercostal space and skin were sutured closed using 6-0 prolene sutures. Cardioprotective therapy of anti-IFNAR antibodies was performed as previously described. Mice were treated with two intraperitoneal doses of 500 μg of MAR1-5A3 IFNAR neutralizing antibody at 8 h and 48 h after MI surgery (BioXCell, BE04241). Hearts were collected at day 3 after MI.

Cardiac needle trauma

For in vivo trauma of the myocardium, we used needle pass injury as previously described¹⁶. In brief, mice were anaesthetized under 2% isoflurane and partial thoracotomy was performed above the fourth left intercostal space; a chest retractor was inserted and placed between the third and fourth intercostal spaces. After visualization of the heart and LAD, a 28 G bevelled needle was inserted into the lateral LV free wall directly to the right of the LAD at the position in which we normally perform permanent ligation in MI; the needle did not penetrate through to the endocardium. The 28 G bevelled needle was held in position in the midventricular wall for 3 s before being withdrawn. The chest retractor was then removed, and the intercostal space was closed using 6-0 prolene sutures. To reduce complications due to pneumothorax, a sterile 20 G flexible angiocatheter was placed within the pleural space before removal of chest retractor. The intercostal space and dermis were closed using 6-0 prolene sutures. After the skin was sutured and closed, a syringe was attached to the angiocatheter and negative pressure was manually applied simultaneously as the catheter was withdrawn. Surgical glue was then applied to the remainder of the skin incision.

Culturing and differentiation of human iPS-cell-derived fibroblasts

Human H9 embryonic stem cells were commercially acquired, tested for mycoplasma contamination and cultured on Matrigel-coated six-well plates (WiCell). Cell lines were tested for mycoplasma contamination determined by PCR but were not subjected to additional verifications. H9 embryonic stem cells were maintained in mTeSR1 medium (StemCell Technologies, 85851) until 90% confluency. Once confluent, cells were split with mTeSR1 medium and 5 μM ROCK inhibitor Y27632, counted and plated at a confluency of 0.5 million cells per ml on Matrigel-coated 12-well plates (Tocris, 1524). Cells were fed daily for 3 days and, the next day, cells were fed with RPMI basal medium and 4 μl of 36 mM of GSK3 inhibitor CHIR99021 (Sigma-Aldrich, R7388; Tocris, 4423) to begin differentiation. Then, 1 day after differentiation with GSK3 inhibitor, the medium was removed and cells were fed with only RPMI basal medium. Then, 2 days later, combined medium was prepared as follows: 1 ml RPMI medium, 1 μl of 5 mM IWP2 (2.5 μM final concentration). Cells were fed with RPMI medium after 2 days and, after another 2 days, cells were detached with ACCUTASE and inactivated by 20% fetal bovine serum (FBS) in RPMI medium (RPMI20) (StemCell Technologies, 07922). Cells were resuspended in LaSR basal medium containing 5 μM Y27632 ROCK inhibitor and seeded onto gelatin-coated 12-well plates at a density of 5,000 cells per well. For the next 5 days, cells were fed with 12 ml of LaSR medium (advanced DMEM/F12 medium; 6.5 ml GlutaMax, 500 μl of antioxidant 100 mg ml^−1 ascorbic acid solution) + 3 μM CHIR99021 GSK3 inhibitor.

Epicardial cells were removed from the plates with Accutase and quenched with RPMI20. Cells were resuspended in LaSR medium supplemented with 5 μM Y27632 ROCK inhibitor. After overnight cell attachment, epicardial cells were differentiated into fibroblasts and maintained in LaSR medium supplemented with 10 ng ml^−1 bFGF (RnD Systems, 233-FB). After differentiation, human iPS-cell-derived fibroblasts were then replated and expanded on gelatin-coated six-well plates and maintained on human cardiac fibroblast medium (MSDS 315-500, Cell Application).

In vitro fibroblast treatment with TGFβ and IFNβ

Mouse L929 fibroblasts originally derived from connective tissue and purchased from the American Type Culture Collection were used for in vitro cell culture experiments and tested for mycoplasma contamination but no additional cell authentication methods were performed (ATCC CCL-1). Human iPS-cell-derived fibroblasts or L929 cells were seeded onto 10 mm gelatin-coated, treated culture plates and maintained with 10% FBS/1% penicillin-supplemented DMEM until reaching confluency. Fibroblasts were subsequently subcultured and/or seeded onto six-well gelatin-coated plates for experimental treatments and designated as passage 3 cells. For in vitro treatment of cells, we used recombinant human TGFβ (10 ng ml^−1, Peprotech 100-21) and/or human IFNβ (10 ng ml^−1, Peprotech 300-02) suspended in culture medium for 24 h before collecting for downstream analysis.

Fibroblast functional assessment with collagen gel contraction assay

Human iPS-cell-derived cardiac fibroblasts or L929 fibroblasts were enzymatically dissociated from confluent culture plates, pelleted by centrifugation and quantified using a haemocytometer. Approximately 250,000 cells were suspended into 2 mg ml^−1 collagen hydrogel solution and adjusted to 1 ml with 10% FBS-supplemented clear DMEM without phenol red (Advanced Biomatrix, 5074; Gibco, 31053028). The 1 ml solution is added to 12-well culture plate and placed in 37 °C incubator for 30 min to allow solidification of hydrogel. After solidification, gels were released from the sides of the wells by careful separation using a pipette tip traced along the perimeter of gel. Gels were supplemented with 1 ml of clear culture medium and photographed to determine the gel area pre-contraction. After 24 h, gels were treated with TGFβ (10 ng ml^−1) as the positive control for gel contraction and/or IFNβ (10 ng ml^−1). For relevant experiments, anti-IFNAR antibody was administered immediately after TGFβ/IFNβ cotreatment with 1 μg ml^−1 suspended in full medium and imaged at 24 h and 3 days after treatment. The gel area was determined using scaled images in ImageJ.

RNA isolation and quantitative PCR

RNA was isolated from myocytes using the RNeasy Mini Kit (Qiagen, 74536) and reverse transcribed using high-efficiency enzymes (Applied Biosystems, 438813). Quantitative PCR was then performed using TaqMan primers for the following mouse transcripts Col1a1 (Mm00801666_g1), Bgn (Mm001191753_m1), Postn (Mm01284919_m1), Sparc (Mm05915229_s1), Irf7 (Mm00516793_g1), Oasl1 (Mm00455081_m1), Ifnb1 (Mm00439552_s1), Irf3 (Mm00516784_m1), Cxcl10 (Mm00445235_m1), Ifit1 (Mm07295796_m1), Isg15 (Mm01705338_s1) and Gapdh (Mm99999915_g1). In the studies done on human iPS-cell-derived cells, the following human transcripts were used: POSTN (Hs01566750_m1), SPARC (Hs00234160_m1), COL1A1 (Hs00164004_m1), IRF7 (Hs00164004_m1), IFI27 (Hs01086373_g1), IFIT3 (Hs01922752_s1), and GAPDH (Hs02786624_g1). For genotyping of transgenic animals, earsnip samples were digested in NaOH and neutralized by Tris HCL. DNA was extracted and amplicons were amplified for gel electrophoresis. The primer sequences used for each transgenic line are included in Supplementary Table 4.

Collection of ruptured ventricles

As permanent LAD ligation in mice produces the mechanical defect of ventricular rupture comparable to clinical observations of myocardial rupture, we performed MI surgery in a cohort of 40 WT mice aged 12–15 weeks purchased from Jackson Laboratory (000664). To preserve the integrity of RNA of ruptured ventricles, mice were closely monitored for 12 consecutive hours each day starting from day 3 to day 14 after MI, which is the window of susceptibility for a rupture event to occur in mice. When an acute mortality was observed, the chest cavity was immediately opened, and hearts were immediately perfused with cold PBS and collected. Tissue was quickly observed under a surgical microscope and then flash-frozen by embedding in optimal cutting temperature (OCT) compound. Each acute mortality event was observed for (1) the presence of blood in the chest cavity and (2) visualization of the rupture site under the microscope. These two criteria qualified a collected sample as a ruptured ventricle.

pDC depletion

Mice were pretreated with intraperitoneal doses of 500 μg of anti-mouse CD317 (BST2) antibody or isotype control antibody once a day for 3 days before MI surgery (BioXCell, BE0311). Hearts were collected at day 3 after infarct, flow sorted and processed for scRNA-seq.

Chimeric bone marrow transfer

For transfer of bone-marrow-derived cells from WT mice into Irf3^{−/−} (WT to KO) experiments, recipient mice were irradiated for 12 min using a 10 Gy dose of ionizing radiation. Bone marrow cells were isolated from the femurs of WT or Irf3^{−/−} donor mice, counted and resuspended in 1 ml of 5% bovine serum albumin/PBS solution. Approximately half a million cells were suspended per 100 μl and donated through retroorbital injection into irradiated recipients.

Immunohistochemistry and nuclear rupture imaging

All hearts collected for downstream analysis were perfused with first 10 ml of cold PBS contained within a syringe attached to a 28 G needle to remove contaminating blood. Tissue was collected and embedded in OCT compound and flash-frozen in an isopentane bath cooled by dry ice. OCT-embedded hearts were sectioned into 10-μm-thick, short-axis sections for use with H&E staining, immunofluorescence and spatial transcriptomics assays (Visium and MERFISH). H&E staining was performed according to the manufacturer’s suggested protocol.

The 3D distribution of cytosolic NLS signal was visualized and imaged using the Nikon AXR point-scanning confocal microscope. For each field of view taken at ×60 magnification, 15–20 optical slices were obtained and used for maximal-intensity projections. Ruptured cardiomyocyte nuclei were normalized to the total number of nuclei analysed in each field of view acquired from the RZ or BZ adjacent to infarct.

Chromagenic in situ hybridization

Fixed frozen tissue from infarcted mouse hearts was prepared for chromogenic in situ hybridization by perfusion fixation with 4% paraformaldehyde, collected and embedded in OCT compound. Cardiac short-axis cross-sections were cut into sections of 10 μm in thickness, dehydrated with ethanol, and underwent tissue pretreatment including hydrogen peroxide incubation, transcript target retrieval and protease treatment according to manufacturer’s suggestions for the RNAscope 2.5 HD Duplex Assay (ACDBio). The following probes were conjugated to HRP-based Green and AP-based Fast Red chromogens to detect Nppa and Ifit1 transcripts in mouse cardiac tissue: Mm-Nppa-C1 (418691) and Mm-Ifit1-C2 (50071-C2).

Sequencing-based spatial transcriptomics

OCT-embedded cardiac tissue blocks were cryosectioned in short-axis orientation at approximately 10 μm in thickness with the cryostat temperature set to −22 °C. The sections were stained with H&E, and images were obtained using ×20 magnification on the Nikon Eclipse Ti2-E widefield microscope. The sections were then processed for spatially resolved gene expression using the Visium Spatial Transcriptomics Kit according to the manufacturer’s protocol (10x Genomics). The permeabilization time of infarcted mouse hearts was previously optimized and determined to be 30 min. Quality control for cDNA and libraries were performed on the Agilent TapeStation before sequencing on the Illumina NovaSeq 6000 instrument. The resulting sequencing data were processed, and images were aligned using the SpaceRanger v.1.3.1 pipeline (10x Genomics).

Quality control, normalization and integration for sequencing-based spatial transcriptomics

Sequencing-based spatial transcriptomic assays were initially preprocessed to assess variance in feature counts/spot and the biological differences in cell density across heterogenous tissue morphology of the infarcted heart. Visual assessment of the underlying H&E confirmed lower molecular counts in tissue regions of dead and necrotic tissue of the IZ, whereas high molecular counts were consistently seen in BZ areas. Normalization was performed using Seurat’s SCTransform v.2 method based on negative binomial models that account for technical artifacts such as sequencing depth variations but detect and preserve highly variant biological features. This is performed by placing a lower bound on the s.d. of low-expressed genes when using Pearson residuals to estimate highly variable features. Replicate data of experimental conditions were split into two Seurat objects labelled as the control condition and the stimulated condition; our control condition consisted of all samples that served as our negative control, that is, infarcted hearts that lacked a type I interferon response (Irf3^{−/−}, Ifnar^{−/−}, Cgas^{−/−}, Sting^gt/gt). All WT day-3 post-MI hearts, all tissue-specific Irf3^{−/−} mice and Ccr2^{−/−} mice were assigned to the stimulated Seurat object. The control dataset was normalized using SCTransform and dimensional reduction was performed using principal component analysis. The stimulated group was similarly normalized using SCTransform() and principal component analysis. To perform integration of the two categorized Seurat objects using Pearson residuals, FindIntegrationAnchors() was performed followed by PrepSCTIntegration() on a merged list of the two objects to anchor and integrate the datasets together. Graph-based clustering using k-nearest neighbours function FindNeighbors() was performed and shared nearest neighbours were identified using FindClusters() at a resolution of 0.40 on the entire integrated dataset.

Moran’s I test statistic, Sepal score and IFNIC colony-size quantification

The spatial distribution of genes was examined using the Moran’s I test statistic, which is a spatial autocorrelation coefficient used to quantify and measure spatial enrichment and distribution of individual genes. Moran’s I was chosen as it is independent from differential gene expression and unrelated to clustering information. This metric scale ranges from 1 (significant spatial enrichment) to 0 (homogenously distributed throughout biological sample). Moran’s I revealed that ISGs are spatially enriched and formed colonies with a high autocorrelation test statistic, indicating numerous focal IFN responses. As orthogonal validation of identifying spatially variable genes, we used the Sepal method, which quantifies spatially clustered features using a diffusion-based simulation. Highly expressed, clustered gene expression are assigned a high Sepal score as it would take a longer diffusion time (d^t) to reach homogenous distribution across space.

To define and quantify ISG-expressing colonies, we considered first-order, contiguous neighbours as a requisite to be designated as an IFNIC colony (Supplementary Fig. 3). We first compared log-normalized expression of Ifit1, Rsad2 and scored ISG transcripts in WT versus Irf3^{−/−} day 3 infarcted tissue to determine that Irf3^{−/−} mice are appropriate to use as a negative control for ISG expression. We then compared the relative frequency of Ifit1 and Rsad2 in all of our biological replicate samples between WT and Irf3^{−/−} day 3 MI spatial transcriptomic samples to determine a threshold or cut-off limit for the expression levels of each ISG. Once the threshold was determined, we created a binomial ‘neighbourhood matrix’ consisting of ISG⁺ pixels and neighbouring pixels also with ISG⁺ expression above the designated threshold expression value of 0.70 for Ifit1, 0.45 for Rsad2, 3.0 for ISG score spatial data and 10.0 for ISG score counts. When quantifying counts of the ISG score per spot, the threshold value was determined as 10 (Extended Data Fig. 1). This was performed by assigning each pixel with a value of 0–6 corresponding to the number of ISG⁺ pixels (N) and quantified by taking the intersect of a binary neighbourhood matrix (positive ISG neighbours) with a binary ISG classification matrix and summing by each column in the generated matrix. k-means nearest-neighbour clustering of the chosen ISG transcript (for example, Ifit1) was then performed, and this approach yielded 4–8 colonies that were significantly absent in Irf3^{−/−} mice. The IFNIC colony area was calculated based on the 55 μm diameter of an individual Visium spot, and the distance between the centres of each adjacent spot measures 100 μm.

Activated fibroblasts, IZ and BZ mapping strategy

We mapped BZ or IZ labels to spatial transcriptomic clusters using a list of differentially expressed genes compiled from sn/scRNA-seq data of cardiomyocytes, innate immune cells and fibroblasts based on our previously described method and publication (Supplementary Table 2). To classify CM-rich spatial clusters, we evaluated the gene-set scores found uniquely elevated in post-MI samples specific to the BZ. Mapping of gene-set scores from CM snRNA-seq data to space was performed using area under the receiver operating characteristic (AUROC) analysis. Clusters with an AUROC > 0.7 were positively classified. For IZ clusters, we performed subclustering to determine immune cell niches to map to space. To map activated fibroblasts to space, we performed subclustering of our integrated snRNA-seq dataset to determine subclusters of fibroblasts designated as activated (Postn⁺) or non-activated (Postn^−). Activated Postn⁺ fibroblasts were used in downstream analysis and the average expression of genes within this subset was determined comparing day 3 post-MI snRNA-seq data from WT samples versus Irf3^{−/−} or Ifnar^{−/−} samples. Zones were quantified and normalized to total UMI and scaled by SCTransform. This analysis was performed with all spots from a representative day 3 post-MI sample and with IZ pixels (defined by clustering) to further explore heterogeneity at a timepoint predetermined to yield the highest expression of ISGs by snRNA-seq and spatial transcriptomic data. Correlation tests were also performed with gene set scores to confirm colocalization patterns inferred from clustering analyses.

We next curated subset-specific gene lists using Seurat’s FindMarkers() function (logfc.threshold = 0.50, min. pct = 0.25, assay = “SCT”) in comparing respective clusters to relevant transcriptional neighbours. Gene lists were filtered to remove genes with adjusted P > 0.0001 and sorted by log-transformed fold change, and gene scores were generated from the top ten genes of each cluster, which were then summed in each spatial assay.

Determination of IFNIC colony localization

To assess whether observed IFNIC colonies localized to the infarct BZ at a rate greater than chance occurrence, the Monte Carlo simulation method was used to approximate random sampling of our dataset. The coordinates of Visium spots with overlying tissue were extracted in which a random coordinate or location was chosen for each simulation. During a simulation, the randomly chosen coordinate was assigned a neighbouring spot (N + 1) and assessed for any overlap or adjacency with BZ^high pixels. Each simulation was performed 500 times until N + 1 = 12 for each sample. The probabilities generated from the simulations were transformed using χ² contingency analysis and P values were assessed using Fisher’s Exact tests to compare probability versus outcome, that is, the percentage of clusters that overlapped with any BZ^high pixels.

Inverse spatial patterning of activated fibroblast and IFNIC signatures

Centroids were selected from each IFNIC colony in analysed samples and primary, secondary and tertiary neighbours were then assigned. Differential gene expression analysis was performed using Wilcoxon rank-sum tests and each neighbour and centroids of ISG colonies were compared. Line scans were performed by measuring gene scores along a vector drawn from the IZ to the BZ. We previously determined gene scores by spatial clustering of transcripts that characterized the BZ and IZ described in detail above. Here we used an ISG gene score determined from our previously generated datasets and spatially clustered genes. Gene scores were reported as a function of distance from a reference vector line (orthogonal to image analysis line scans).

For IZ and BZ neighbour analysis, we quantified the fraction of BZ and ISG pixels in second-order neighbours (defined by clustering analyses). Results were binned on the basis of the reference pixel classification; IZ contained primarily transcripts designated as innate immune process or cells whereas BZ contained primarily transcripts designated from BZ myocytes and activated fibroblasts. From this, we quantified both scores as a function of distance (0–400 μm) relative to the reference pixels.

Analysis of IFNIC colonies using infarcted human tissue

Spatial transcriptomic analysis of infarcted human samples was determined using deposited datasets from a recently published study using the same genome-wide Visium platform and from our previously published samples. Patient samples were integrated using SCTransform described above, and metrics of spatial autocorrelation and size of IFNIC colonies were performed using the methods described above and in the same manner as for mouse spatial transcriptomic analysis. BZ and ischaemic zone genes were transcriptionally determined using snRNA-seq data of ‘CM2’ from the referenced multiomic study, and the tables are included in Supplementary Table 3.

RNA MERFISH imaging-based spatial transcriptomics

To perform RNA MERFISH and sequential imaging in the infarcted murine heart, a 33-gene probe library targeting Tnnt2, Ttn, Ankrd1, Nppa, Shroom3, Nppb, Xirp2, Flnc, Col1a1, Col6a3, Postn, Cxcl5, Adgre1, Cd68, Ccr2, Chil3, S100a4, Ly6c2, Timd4, Lyve1, Cxcr2, Csf3r, Ly6g, Retnlg, S100a8, Pecam1, Flt1, Ifna2, Ifnb1, Ifit1, Ifit2, Ifit3 and Cxcl10 was designed and constructed to include cell-type-specific marker genes and genes covering the type I IFN signalling pathway. Approximately 20–60 barcoded encoding probes were designed to target specifically 40 nucleotide subregions of selected transcripts^49,50. Ventricular short-axis tissue sections were cut on a cryostat at 16 μm in thickness onto silanized coverslips and were fixed in 4% PFA at room temperature for 10 min. The samples were then permeabilized with 5% SDS in PBS for 10 minutes followed by 80% ethanol in water for a few hours. Tissue hybridization of encoding probes was performed for approximately 16 h in a humidified oven at 47 °C. After overnight hybridization and washing for 30 min with 40% formamide in 2× SSC with 0.1% Tween-20, the samples were cast in Bis/Acrylamide to cross-link and stabilize acrydite-modified encoding probes. Tissue clearing with proteinase K digestion or photobleaching of samples was performed to quench background autofluorescence in the tissue samples⁵¹. Sequential hybridization and stripping of fluorescently labelled readout probes was performed using an automated custom-built fluidics system. Imaging was performed on a custom-built system with a ×60 objective lens as previously described⁵². After data collection, raw images underwent fitting analysis and image registration to correct for drift that occurred during image acquisition. Single mRNA molecules were computationally decoded, and the total transcript signal and DAPI nuclear stain were used to perform cell segmentation with machine learning algorithm Cellpose⁵².

Quality control, normalization and integration for RNA MERFISH

Data analysis of RNA MERFISH data was performed with single-cell sequencing analysis tools such as Scanpy and Squidpy^53,54. Quality-control metrics were determined by visualizing distribution plots of the raw data for total transcript counts and number of genes with more than one counts per cell. These metrics were used to determine covariates that may affect the quality of the dataset and filtered the data accordingly. Quality-control metrics are displayed in Supplementary Fig. 2. The data were then normalized using Scanpy and log-normalization of the total number of transcripts. We performed RNA MERFISH experiments with multiple cardiac tissue sections and experimental conditions to control for batch effects. Data integration across sections from different animals was performed using Scanpy’s principal component-based method ingest. We used the WT day 3 MI mouse cardiac section as our reference dataset and performed Leiden clustering with subsets of marker genes from our encoding probe gene panel as follows: BZ cardiomyocytes (Nppa, Flnc, Ankrd1), fibroblasts (Col1a1, Col6a3), macrophages (Cd68, Adgre1), neutrophils (Cxcr2, Csf3r) and endothelial cells (Pecam1, Flt1). Cells from infarcted Irf3^{−/−} mice were ingested or embedded into the UMAP space of the annotated cells in WT mice we used as reference.

DNA MERFISH: genome-scale in situ chromatin imaging

DNA MERFISH was performed as previously described^55,56,57. In brief, the mouse genome was partitioned into 260 distinct DNA loci that spanned all 21 mouse chromosomes. Each genomic loci spanned around 30 kb and was targeted by 200–300 specific DNA-encoding probes. These primary probes were synthesized from an oligonucleotide pool from Twist Biosciences and each contained a 40-nucleotide target sequence specific to a region of DNA loci. Sample preparation was performed as previously described and hybridization of probes contained pools of both acrydite-modified RNA and DNA encoding libraries to facilitate downstream analysis of cell identity and DNA localization. Fluorescently labelled readout probes were combinatorially labelled. Hybridization and stripping were performed using a custom-built fluidics and microscopy system (see the ‘RNA MERFISH imaging-based spatial transcriptomics’ section above). After data collection, raw images underwent fitting analysis and image registration to correct for drift. DNA loci were computationally decoded using signal intensity between fields of view in multiple fluorescence channels to filter noise from signal. Nuclear-localized DNA probes were determined using neighbourhood-based clustering with minimally ten decoded spots per 5 μm radius. Sparse, decoded DNA probes that met the criteria for positive signal but were outliers to the neighbourhood distance were designated as extranuclear DNA. Nuclear shearing artifacts from cryosectioning were controlled and excluded from analysis by removing data in the first and last 5 μm regions of the tissue sections. Cell segmentation was performed with Cellpose on DAPI staining and RNA molecules. Cells with extranuclear DNA underwent quality-control analysis as described above, normalized and ingested with cells from our WT day 3 MI cells from the RNA MERFISH experiment.

Density-based clustering of applications with noise analysis

Analysis of RNA MERFISH IFNIC colonies was performed using the density-based clustering algorithm (DBSCAN). This method considers whether a set of features are densely grouped or homogenously distributed with a certain radius ε (distance-based eps = 70). ε was unbiasedly selected using k-nearest neighbour analysis. This algorithm identified eight groups of densely clustered ISG transcripts (Ifit1, Ifit2, Ifit3, Cxcl10) as ISG expression in low-density regions that were designated as outliers or ‘scattered’ ISG expression.

Nuclear morphology

In cardiac sections that had undergone RNA MERFISH analysis, cardiomyocyte nuclei were identified by the colocalization of hybridized Tnnt2 probes and DAPI fluorescence present in both the BZ and RZ. Approximately 200 nuclei were quantified in each region by performing nuclear segmentation and creating a binary segmentation mask. These masks were analysed using ImageJ Particle Analysis to measure cardiomyocyte-specific nuclear solidity. The solidity of a nucleus is calculated as the area of the DAPI-stained nuclei divided by the area of the convex hull; large deviations below solidity ratio of 1.0 indicate irregularity of the nuclear contour.

Single-cell isolation, flow cytometry and cell sorting

Whole-cell suspensions were isolated from freshly collected hearts as previously described. In brief, hearts were enzymatically digested for 45 min in continuous agitation at 37 °C in 450 U ml^−1 collagenase, 125 U ml^−1 collagenase XI, 60 U ml^−1 DNase and 60 U ml^−1 hyaluronidase. Cell suspensions were then filtered through 40 μm nylon mesh cell strainer containing flow cytometry staining buffer (FACS) and stained at 4 °C with DAPI to exclude permeabilized cells and anti-mouse cocktail directed against major haematopoietic lineage markers Terr119 (1:100, TER119), B220 PE (1:100, RA3-6B), CD49b (1:200, DX5) and CD90.2 (1:500, 53-2.1). Secondary staining of myeloid and stromal cell subsets was performed using an anti-mouse antibody cocktail against CD11b (1:100, M1/70), and CD45.2 (1:100, 104), Ly6G (1:100, A1A8) and F4/80 (1:100, BM8). Primary and secondary master mixes were suspended in FACS. Flow cytometry was performed on the SONY MA900 multi-application cell sorter.

Single-nucleus isolation

Single-nucleus suspensions were isolated from frozen ventricular tissue as previously described. In brief, mouse ventricles were collected, weighed and minced before flash-freezing by immersion in liquid nitrogen. For isolation of nuclei, minced ventricles were suspended in 0.6 ml nucleus lysis buffer supplemented with 0.2 U μl^−1 RNase inhibitor (Sigma-Aldrich, NUC101; Enzymatics, Y9240L). Minced tissue was further homogenized with 2 ml dounce grinder for approximately 10 strokes with an A-sized pestle and 20 strokes with a B-sized pestle (Sigma-Aldrich, D8938). The lysates were treated with an additional 1 ml of lysis buffer and incubated for an additional 2 min; the lysates were filtered through consecutive 100, 50 and 20 μm strainers (CellTrics, 04-004-2318, 04-004-2317 and 04-0042-2315). Nuclei were pelleted by centrifugation at 1,000g for 5 min at 4 °C. Subsequent washes were performed until the final suspension of nuclei was made using 2% BSA in PBS supplemented with RNase inhibitor. The nucleus suspension was treated with 10 μg ml^−1 4′,6-diamidino-2-phenylindole (DAPI), counted on a haemocytometer and the volume was adjusted to produce a final suspension of 1,000 nuclei per μl.

sn/scRNA-seq analysis

Microfluidic droplet-based separation of single cells or individual nuclei was performed in which each was encapsulated with reagents for reverse transcription of mRNA, barcodes and unique molecular identifiers (UMIs) (10x Genomics Chromium). Paired-end sequencing was performed using Illumina dye sequencing on the NovaSeqX Plus instrument. Demultiplexing of pooled samples and low-level analysis were performed using the Cell Ranger v.6.1.1 pipeline from 10x Genomics in which the sequenced samples were mapped to a mouse reference transcriptome (refdata-gex-mm10-2020-A, which includes introns), and redundant UMIs were eliminated.

Quality control, normalization and integration for sn/scRNA-seq

Normalization was performed to account for variability in depth of sequencing reads per cell or nuclei as previously described. The total transcript count for each nucleus was scaled to 10,000 molecules, and raw counts for each gene were normalized to the total count of captured transcripts associated with the barcoded cells or nucleus and natural log transformed. Nuclei that contained at least 200 uniquely expressed genes represented in at least three nuclei were retained for further analysis. Ribosomal and haemoglobin transcripts were excluded to avoid incorporation of transcriptional artifacts or technical variables accrued during nuclei isolation. Moreover, quality control was performed to assess the quantity of low-quality/dying cells by assessing the percentage of mitochondrial transcripts present using the PercentageFeatureSet function and excluded nuclei containing more than 5% mitochondrial content. Highly variable genes across individual datasets were identified with the FindVariableFeatures function using Seurat R package v.4.3, which performs variance-stabilizing transformation with subsequent selection of 4,000 genes with the highest feature variance. Doublets and aggregated nuclei were determined by assessing non-endogenous gene markers (for example, the presence of CM genes such as Myh6 in the fibroblast subset) and ambient RNA were removed using SoupX, which displayed subsets after filtering and removing doublets/multiples. Integration of multiple scRNA-seq and snRNA-seq datasets was performed in Seurat using canonical correlation analysis (CCA) to identify anchors between datasets and anchored with the mutual nearest-neighbour method using the Seurat FindIntegrationAnchors function.

Statistics

Statistical analyses were performed using Prism 9 (GraphPad Software). Group size was based on previous experience and no statistical methods were used in selecting a predetermined sample size. Comparisons between two groups were analysed using two-tailed unpaired Student’s t-tests for parametric analyses or two-tailed Mann-Whitney nonparametric tests. For comparison of multiple groups, one-way or two-way ANOVA was used to test main effects with Dunnett’s, Tukey’s or Bonferroni’s post hoc analysis performed for multiple-comparison testing. Kruskal–Wallis testing was performed for multiple comparisons in one-way ANOVA analysis with Dunn’s correction. Two-way ANOVA was used to test main effect with Tukey’s post hoc analysis to test interaction between groups. Differential expression analyses were performed using nonparametric Wilcoxon rank-sum tests with Benjamini–Hochberg FDR adjustment. Unless otherwise stated, every experiment was repeated two or more times. Data represent mean ± s.e.m. as indicated in the figure legends.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.