Supplementary Figure 1. TAS-Seq outperformed the BD Rhapsody WTA kit in terms of
gene-detection sensitivity.
a. Diagram of the experimental workflow. Using 1600 frozen mouse adult spleen cells, cDNA
synthesis was performed by the same BD Rhapsody cartridge, and resultant beads were separated
into two partsthem which were processed by TAS-Seq and BD commercial WTA kit, respectively. b.
Scatter plot representation of the read number/detected gene number of each cell of TAS-Seq
(salmon, 736 cells) and BD WTA (blue, 715 cells) datasets of mouse spleen cells. A Violin plot of
the detected gene number of each dataset is also shown on the right side of the scatter plot. p = 2.2 ×
10-163, W = 45757.5 by Wilcoxon rank-sum test. c. Hexagonal pseudocolor plot of library quality
metrics of the TAS-Seq and BD WTA datasets. Blue, green, and red indicate more cells located
within the same hexagonal area. Note that mitochondrial gene proportion, ribosomal protein gene
proportion, and ribosomal RNA proportion were similar between the two datasets, suggesting
comparable performance on the library metrics between TAS-Seq and BD WTA kit.
Supplementary Figure 2. Gating scheme for identification of murine lung cell subsets by flow
cytometry. Single-cell suspension of 8-week-old C57BL/6J female murine lung, subjected to
TAS-Seq analysis, was analyzed by flow cytometry. a. Gating scheme of murine lung endothelial
cells, epithelial cells, smooth muscle cells (SMC)/pericytes, and fibroblasts. b. Gating scheme of
murine lung myeloid cell subsets. c. Gating scheme of murine lung lymphoid cell subsets.
Supplementary Figure 3. Comparison of detected gene number of each cell subset of murine
lung between TAS-Seq and Smart-seq2 datasets. Violin plot representing the distribution of
detected gene number of each dataset among commonly-detected cell subsets in TAS-Seq (salmon)
and Smart-seq2 (green). Boxplot shows mean, upper and lower quantile of detected genes. **p <
0.01, ***p < 0.001, ****p < 0.0001 by Wilcoxon rank-sum test. Exact p-values and W statistics are
shown in Supplementary Table 2.
Supplementary Figure 4. TAS-Seq data of the lungs of a human RA-ILD patient. Visualization
of cell clustering results of TAS-Seq data of human RA-ILD lungs by Seurat v2.3.4 package in 2D
FIt-SNE space. The stacking plot showed the composition of each annotated cell. Each annotated
cell was colored commonly between FIt-SNE and stacking plot. Detail of the cell annotation and
associated marker genes are represented in Supplementary Table 3.
Supplementary Figure 5. Gating scheme for identification of human RA-ILD lung cell subsets
by flow cytometry. Single-cell suspension of non-fibrotic and fibrotic lung samples from a human
RA-ILD patient, subjected to TAS-Seq analysis, was analyzed by flow cytometry. a. Gating scheme
of human lung endothelial cells, epithelial cells, smooth muscle cells (SMC)/pericytes, and
fibroblasts. b. Gating scheme of human lung leukocytes.
Supplementary Figure 6. Difference of CellChat-inferred Cell-cell interaction network of
murine lungs between TAS-Seq, Smart-seq2, and 10X Chromium v2/v3 datasets. a. Heatmap
representation of detected pathways within the network at several thresholds of minimum expression
of genes in each cell subset (from 0.05 to 0.5). Detected pathways are colored by magenta, and
undetected pathways are colored by grey. Commonly-detected pathways are separately shown in
Supplementary Table 4 to show the difference between datasets. Fibroblast growth factor (FGF),
bone morphologic protein (BMP), sonic hedgehog (HH), NOTCH, and WNT signaling are
highlighted by red arrows. b. Scatter plot of incoming (target) and outgoing (source) signaling
strength within cell-cell interaction network of each cell subset (minimum expression of genes in
each cell subset ≥ 0.15, a minimum number of expressed cells ≥ 10, a threshold of the significance
of the interaction ≤ 0.05). Dot size represents the sum of the number of incoming and outgoing
signaling of each cell subset. Vascular endothelial cells, alveolar type 2 cells, and Inmthi alveolar
fibroblasts were strongly connected with the TAS-Seq and Smart-seq2 dataset network, and cell
subsets were more strongly connected in TAS-Seq dataset than the Smart-seq2 dataset. The
contribution of alveolar type 2 cells was depleted in 10X v3 and v3 datasets. c. Circle plot
visualizations of all of the cell-cell interaction networks of TAS-Seq, Smart-seq2, 10X v2/v3 datasets.
Circle sizes are normalized to the cell number of each subset. Edge width represents communication
strength (wider edge means stronger communication between source and target cell subsets),
normalized among all datasets. Edge colors are the same as the color of their source cell subsets.
Abbreviations of cell subsets were shown in Supplementary Table 8.
Supplementary Figure 7. Relative contribution of each cell subset to growth factor genes and
interleukin genes in steady-state murine lungs. a-d. The relative contribution of each cell subset
to each gene was calculated based on the TAS-Seq or 10X Chromium v3 datasets of murine lungs.
Cumulative raw read counts (TAS-Seq data) or raw UMI counts (10X v3 data) of each gene were
calculated for each cell subset and z-scaled. Each row represents each gene, and each column
represents each cell subset.
Supplementary Table 1. Cell annotations, associated marker genes, and associated references
for mouse datasets.
Supplementary Table 2. Exact p-value of Wilcoxon rank-sum test of Supplementary Figure 3.
Supplementary Table 3. Cell annotations, associated marker genes, and associated references
for human RA-ILD lung datasets.
Supplementary Table 4. List of the commonly-detected pathways in CellChat analysis.
Supplementary Table 5. List of the antibodies used for this study.
Supplementary Table 6. List of the primer sequences used for this study.
Supplementary Table 7. All of the identified marker genes of each dataset by Seurat analysis.
Supplementary Table 8. Abbreviations of the cell subset names.