Spatial detection of fetal marker genes expressed at low level in adult human heart tissue

Abstract

Heart failure is a major health problem linked to poor quality of life and high mortality rates. Hence, novel biomarkers, such as fetal marker genes with low expression levels, could potentially differentiate disease states in order to improve therapy. In many studies on heart failure, cardiac biopsies have been analyzed as uniform pieces of tissue with bulk techniques, but this homogenization approach can mask medically relevant phenotypes occurring only in isolated parts of the tissue. This study examines such spatial variations within and between regions of cardiac biopsies. In contrast to standard RNA sequencing, this approach provides a spatially resolved transcriptome- and tissue-wide perspective of the adult human heart, and enables detection of fetal marker genes expressed by minor subpopulations of cells within the tissue. Analysis of patients with heart failure, with preserved ejection fraction, demonstrated spatially divergent expression of fetal genes in cardiac biopsies.

Conflict of interest statement

P.L.S. and J.L. are founders of a company that holds IP rights to the presented technology.

Figures

Figure 1

Schematic of spatially resolved gene expression analysis on adult heart. (a) In order to establish optimal conditions for retrieval and attachment of mRNA in situ, a quality control assay is first used. Intensity of signals generated from the Cy3-cDNA footprint is then scored to examine optimal permeabilization conditions. (b) A spatially barcoded microarray, printed with 1,007 clusters (i.e features) of reverse-transcription oligo-dT primers, are then used with established permebilization conditions. Each feature on the array contains more than 200 million probes, all sharing a unique DNA sequence (barcode) specific to that feature. The barcode is used in the downstream analysis to link each feature’s position within the tissue to the mRNA captured at that position. Finally, following reverse transcription and tissue removal, barcoded cDNA is enzymatically released from the array and used to generate sequencing libraries.

Figure 2

Gene expression patterns within and between individuals. (a,b) Saturation curves showing numbers of unique transcripts (a) and unique genes (b) per feature, against subsets of raw reads. (c and d) Scatterplots showing Pearson correlation of gene expression between consecutive tissue sections from LV and RAA. (e) Scatterplot showing Pearson correlation of gene expression between different cardiac regions of the same individual. Letters correspond to tissue sections specified in Table 1. (f) Principal Component Analysis score plot showing separation of LV and RAA sections along PC1, and separation of individuals along PC2. Loadings show genes that make a major contribution to the separation. (g) Heatmap of gene expression across all features from one LV and one RAA tissue section from individual 1.

Figure 3

Spatially resolved gene expression patterns within cardiac tissue sections. (a) Upper panel: Images of one hematoxylin- and eosin-stained replicate of each sample. Images have been cropped to match the area of the spatially barcoded microarray. Lower panel: Results of applying t-SNE analysis to data from RAA (4 samples in total; t-SNE 1) and LV (6 samples in total; t-SNE 2). Similar colors within the t-SNE plots reflect similar gene expression patterns. (b–i) Cell composition of right atrial appendage. (b) Selected features within the RAA of individual 1 corresponding to the colored areas in results of the t-SNE 1 analysis shown in (a). (c, d and e) Magnified images of corresponding areas in (b), showing: (c) an adipocyte-tissue-rich region, (d) a fibrous-tissue-rich region, and (e) a cardiomyocyte-rich region. (f) Selected features within the RAA of individual 3 corresponding to the colored areas in results of the t-SNE 2 analysis shown in (a). (g,h and i) Magnified images of corresponding areas in (f), showing: a pericardial fibrous-tissue-rich region (g,h) and a cardiomyocyte-rich region (i). (j) Barplots showing that the blue area (corresponding to the cardiomyocyte-rich region) contains ~2-fold more unique transcripts and ~1.6-fold more unique genes than the red area (corresponding to adipocyte- and fibrous-rich tissue). (k) Barplots showing that pericardial fibrous-tissue contains similar numbers of unique transcripts and genes per feature to cardiomyocyte-rich regions. Error bars indicate standard errors. (l) Spatially resolved expression heatmaps showing four differentially expressed genes between the blue and red area of individual 1, and between the blue and green area of individual 3.

Figure 4

Detection of fetal gene expression within adult heart tissue sections. (a) Barplots showing the average number of relative fetal gene counts per feature. All numbers have been log-transformed after adding a pseudocount of 1. In the ST data, the number of features is based on features displaying read counts from indicated fetal marker gene. In the bulk treated ST data, the number of features is based on the total amount of features within the dataset. (b) Spatial view of individual features within RAA of individual 1 displaying expression of the HOPX gene. (c) Magnified image of corresponding area in (b) illustrating an example of features expressing HOPX.