Spatial epitranscriptomics reveals A-to-I editome specific to cancer stem cell microniches

Abstract

Epitranscriptomic features, such as single-base RNA editing, are sources of transcript diversity in cancer, but little is understood in terms of their spatial context in the tumour microenvironment. Here, we introduce spatial-histopathological examination-linked epitranscriptomics converged to transcriptomics with sequencing (Select-seq), which isolates regions of interest from immunofluorescence-stained tissue and obtains transcriptomic and epitranscriptomic data. With Select-seq, we analyse the cancer stem cell-like microniches in relation to the tumour microenvironment of triple-negative breast cancer patients. We identify alternative splice variants, perform complementarity-determining region analysis of infiltrating T cells and B cells, and assess adenosine-to-inosine base editing in tumour tissue sections. Especially, in triple-negative breast cancer microniches, adenosine-to-inosine editome specific to different microniche groups is identified.

Fig. 1. Spatial-histopathological examination-linked epitranscriptomics converged to transcriptomics with sequencing (Select-seq) enables full-length spatial transcriptomics and epitranscriptomics at single-nucleotide resolution.

a Schematic of the Select-seq protocol. Selective isolation of target regions in tumour sections was performed using a near-infrared pulsed laser following immunofluorescence staining. Full-length transcripts extracted from each targeted region are tagged with barcodes used for tracking the target region. Multi-modal analysis of the full-length transcriptome is connected with the spatial and staining information using barcodes included in the sequencing data. b Targets selected from a triple-negative breast cancer (TNBC) patient tumour section. The tissue was stained with Hoechst dye and anti-ALDH1 and CD44 antibodies (scale bar, 100 μm). We obtained the above results from a single tissue section. c Transcriptomics and epitranscriptomics of target region 72 containing 5–30 cells at single-nucleotide resolution and its gene expression profiles, transcript isoforms, B cell receptor sequences, and adenosine-to-inosine (A-to-I) editing events. d Each transcriptomic and epitranscriptomic data point was mapped to the tissue based on the barcodes.

Fig. 2. The Spatially-Resolved Laser-Activated Cell Sorting (SLACS) device produces high-quality spatial-histopathological examination-linked epitranscriptomics converged with transcriptomics with sequencing (Select-seq) data from single cells and ten cells.

Source data are provided as a Source Data file. a Experimental design and an example of single-cell isolation. b Single-cell isolation of the SLACS device. Scale bar, 100 µm. c Fragments per kilobase of transcript per million mapped reads (FPKM) values for the paraformaldehyde (PFA)- and methanol (MeOH)-fixed cells (n = 60 biologically independent cells examined over 2 independent experiments). Interquartile range (IQR) of boxplot is between Q1 and Q3 and centre line indicates median value. Whiskers of boxplot is extended to the maxima and minima. Maxima is Q3 + 1.5*IQR and minima is Q1 − 1.5*IQR. d Correlation between the mRNA sequencing profiles from bulk mRNA-seq and Select-seq with ten types of PFA-fixed cells and Select-seq with ten types of MeOH-fixed cells. e Number of genes detected (FPKM) (left) and exon alignment percentage in three different cell lines fixed with PFA (right) (n = 92 biologically independent cells examined over 3 independent experiments). Interquartile range (IQR) of boxplot is between Q1 and Q3 and centre line indicates median value. Whiskers of boxplot is extended to the maxima and minima. Maxima is Q3 + 1.5*IQR and minima is Q1 − 1.5*IQR. f Representative 3’ end bias of the full-length transcriptomes. g Principal component analysis (PCA) and h unsupervised clustering heatmap of the cells analyzed with Select-seq. i Representative transcript isoform diversity from two samples. j B cell receptor (BCR) analysis of the IM-9 cell line.

Fig. 3. Tumour sections from TNBC patients reveal the spatial transcriptomic landscape of immunofluorescence (IF)-stained tissue sections.

a Number of genes detected in the isolated target regions according to the four staining groups (n = 106 biologically independent samples (ROI)). Interquartile range (IQR) of boxplot is between Q1 and Q3 and centre line indicates median value. Whiskers of boxplot is extended to the maxima and minima. Maxima is Q3 + 1.5*IQR and minima is Q1 − 1.5*IQR. b Haematoxylin and eosin (H&E) stained serial tissue section (left). ERBB2 gene expression data from Select-seq (middle). RNA-fluorescence in situ hybridization (FISH) of serial tissue sections (right) (scale bar, 500 μm). We obtained the above results from three different tissue sections. c Lehmann TNBC subtyping. Red and green boxes indicate upregulated and downregulated gene pathways. BL1 basal-like type 1, BL2 basal-like type 2, IM immunomodulatory, ML mesenchymal-like, MSL mesenchymal stem cell-like, LAR luminal androgen receptor. Source data are provided as a Source Data file. d Gene expression heatmap of the target regions. e Principal component analysis (PCA) of the target regions. f RNA velocity analysis of the target regions. Arrow indicates the A-to-I-edited sample in GPX4. g Mean and standard deviation (SD) of immunosuppressive gene signature and GPX4 gene expression in different spatial groups. h Spatial mapping of signature genes related to immunosuppression. Yellow, green, red, and blue marks indicate (i) CD44⁺/ALDH1⁺, (ii) CD44^low/−/ALDH1^high, (iii) CD44^high/ALDH1^low/−, and (iv) CD44⁻/ALDH1⁻ regions, respectively, as determined by IF.

Fig. 4. Spatial analysis of the immune cell repertoire.

a Spatial mapping of the signature genes related to B cell maturation (n = 106 biologically independent samples (ROI)). Bar plot indicates mean value and error bar indicates standard deviation. b Six representative BCR sequences recovered using Select-seq and matched for gene expression related to B cell maturation. Overall, both heavy and light chain sequences from 42.9% of samples were recovered using Select-seq. Representative sequences including isotype, CDR3, matched V and J genes for each heavy and light chain. c Matched spatial context of recovered BCR sequences. Most of the BCRs were detected in the CD44low/−/ALDH1high stromal region, and very few BCRs were detected in the CD44high/ALDH1low/− duct region. Four of the six HCDR3 amino acid sequences were recovered from the CD44−/low/ALDH1high region, and two were recovered from the CD44−/ALDH1− region. d Single-cell deconvolution of each ROI (left) and ROI group (right). Source data for b–d are provided as a Source Data file.

Fig. 5. Spatial A-to-I editome marks characteristic features for different staining groups.

Source data are provided as a Source Data file. a Percentage of A-to-I-edited samples per staining group (top) (n = 106 biologically independent samples (ROI)). Interquartile range (IQR) of boxplot is between Q1 and Q3 and centre line indicates median value. Whiskers of boxplot is extended to the maxima and minima. Maxima is Q3 + 1.5*IQR and minima is Q1 − 1.5*IQR. Bar graphs of ADAR gene expression according to the spatial groups (bottom). Bar plot indicates mean value and error bar indicates standard deviation. b Portions of A-to-I-edited regions in the genome (top). Profiles of editing in the non-repetitive regions, excluding upstream and downstream regions (middle). Portions of non-synonymous (Non-syn) and synonymous (Syn) editing events among the exonic-edited regions (bottom). c A-to-I editome landscape of different microniche groups.

Fig. 6. Non-synonymous A-to-I editing signatures are preserved in the stiaining microniche groups.

Source data are provided as a Source Data file. a A-to-I editome of distant-clustered ROIs show preserved A-to-I signatures. b Tree with height clustered with physical distance between ROIs (left) and distribution of these clusters in the tissue. c A heatmap of single-base RNA editing event in selected genes related to iron mediation. d Volcano plot of fold change values of genes in samples with A-to-I-edited GPX4 variants compared to those without the variant. Microniches from four TNBC tissues (from patients B, C, D, and E) were analyzed. Blue dots indicate ferroptosis-related genes, red dots indicate upregulated genes in CSCs, and green dots indicate downregulated genes in CSCs.