EHT

Injured endocardium obtains characteristics of haemogenic endothelium during adult zebrafish heart regeneration

1. Single cell sequencing data analysis:
   To investigate the endocardial-to-haematopoietic transition (EHT) after heart injury, we integrated 6 published zebrafish single-cell RNA sequencing datasets containing steady-state and injured heart samples at different post-injury timepoints: 1) GSE138181 (Koth); 2) GSE153170 (Bakker); 3) GSE172511 (Sun); 4) GSE159032 and GSE158919 (Hu); 5) GSE188511 (Kapuria); and 6) GSE145980 (Ma).

   These datasets were first preprocessed and analysed separatedly (pre-analysis). Main steps included: 1)Quality control; 2)Normalization (10000 final counts, log-transformed); 3)Cell Cycle Assignment; 4)Select highly variable genes; 5)PCA and UMAP calculation

   Then, these datasets were integrated by normalized counts and used to conduct Combined_analysis:
   (1) Highly variable genes selection (min_mean=0.02, max_mean=3, min_disp=0.3)
   (2) PCA calculation
   (3) Harmony integration (top 50 PCA)
   (4) UMAP calculation
   (5) Cell type definition (Euclidean distance with Hu et al's annotation)

   To investigate the EHT clusters, leiden clustering was performed directly or with the 'restrict_to' parameters (Subcluster_definition.ipynb).Also, to further investigate the inflammatory response in the endocardium after heart injury, the normalized counts of the endocardial cells from Hu et al’s data were extracted and re-analysed (Endo_subcluster.ipynb)(Supplementary Figure 3).

   The rank-gene-groups function in Scanpy was used to identify marker genes for each subcluster with the method set to ’t-test'. The plots were generated using the umap/dotplot plotting functions in Scanpy (EHT_plottings.ipynb):
   (1) Figure 1 (D-G)
   (2) Figure 2 (A,B,D)
   (3) Figure 3
   (4) Figure 6 (C,F)
   (5) Figure 7 (A,B,C,E)
   (6) Supplementary Figure 1 (D)
   (7) Supplementary Figure 4 (A,B)
   (8) Supplementary Figure 6 (A,B)

2.Lineage analysis:
Scar lineage tracing information was obtained from the GEO database GSE159032. Lineage tree construction was performed using the R script Iterative_tree_building.R in the original manuscript, excluding the DMSO/IWR/morphine treatment samples:
· (1) Iterative_tree_building
· (2) Calculate transition probablities
· (3) Visulization (Figure 4B)

3.Trajectory analysis:
To investigate the possible transition between clusters, we conducted PAGA (Partition-based graph abstraction) and FateID analysis on wild-type cells:
· (1) PAGA:
1)PAGA calculation
2)Filtering and positioning (Figure 4A)
· (2) FateID:
1)Cluster definition
2)Fate probability calculation
3)Fate bias visualization (Figure 4E,4F)
· (3) Cellrank:
1)Diffusion map calculation
2)Root cell selection (Supplementary Figure 6C)
3)Diffusion pseudotime calculation (Figure 4D)
4)Transition matrix calculation (pseudotime kernel) (Figure 4C)

4.runx1 functional analysis:
To understand the role of runx1 in EHT, we interrogated the Koth et al 2020 data set clustering in integrated data set:
· (1) Koth data overview (Supplementary Figure 5A,5B)
· (2) (runx1-citrine,kdrl-mCherry) double-positive cells analysis:
1)Re-analysis and subclustering (Figure 6A)
2)Psuedotime analysis (Figure 6B,6C)
· (3) Endocardial cells analysis
1)Re-analysis and subclustering (Figure 6D, Supplementary Figure 5C)
2)Psuedotime analysis (Figure 6E)

5.Background contamination analysis (SoupX) (Supplementary Figure 4C):
To investigate the influence of the soup contamination on the gene expression evaluation, we conducted contamination analysis using SoupX on the representative datasets:
· (1) GSE138181_All (WT/runx1KO, 0/3dpi)
· (2) GSE159032_GSM4817944 (WT_3dpi)
Main steps include:
1)Contamination probability prediction
2)Non-expressing cluster definition
3)Soup fraction evalutaion

6.Bridging clusters analysis (Supplementary Figure 2):
To investigate the clusters connecting between the endocardial cells, leukocytes and thrombocytes, we further subclustered Leiden_v2 clusters and checked their:
· (1) Gene expression
· (2) Proportions over time

The scripts for data type conversion between R and Python could be found in conversions.