Single‐cell molecular profiling provides a high‐resolution map of basophil and mast cell development

Abstract Background Basophils and mast cells contribute to the development of allergic reactions. Whereas these mature effector cells are extensively studied, the differentiation trajectories from hematopoietic progenitors to basophils and mast cells are largely uncharted at the single‐cell level. Methods We performed multicolor flow cytometry, high‐coverage single‐cell RNA sequencing analyses, and cell fate assays to chart basophil and mast cell differentiation at single‐cell resolution in mouse. Results Analysis of flow cytometry data reconstructed a detailed map of basophil and mast cell differentiation, including a bifurcation of progenitors into two specific trajectories. Molecular profiling and pseudotime ordering of the single cells revealed gene expression changes during differentiation. Cell fate assays showed that multicolor flow cytometry and transcriptional profiling successfully predict the bipotent phenotype of a previously uncharacterized population of peritoneal basophil‐mast cell progenitors. Conclusions A combination of molecular and functional profiling of bone marrow and peritoneal cells provided a detailed road map of basophil and mast cell development. An interactive web resource was created to enable the wider research community to explore the expression dynamics for any gene of interest.


| INTRODUC TI ON
Mast cells are sentinel cells that are strategically positioned throughout the body and allow rapid triggering of the immune system upon infections. 1 Mast cell activation also follows IgE-allergen-mediated crosslinking of the FcεRI receptors in atopic individuals, which causes an allergic reaction. Along with basophils, activation of mast cells results in prompt release of proteases and histamine from the cytoplasmic granules as well as synthesis of cytokines and chemokines. These mediators in turn cause inflammation, vasodilation, and leukocyte recruitment to the site of triggering. 1 Thus, the functions of mature basophils and mast cells have been studied in great detail.
However, less is known about these cells' development.
A hierarchical model with distinct megakaryocyte-erythroid, granulocyte-monocyte, and lymphoid branches, was until recently the dominating representation of hematopoiesis. 2 Single-cell RNA sequencing (scRNA-seq) coupled with cell fate assays now reveals that hematopoietic differentiation more likely represents a landscape of cell states with continuous progression from multi-and bipotent progenitors into each respective cell lineage. [3][4][5][6][7] In particular, single-cell transcriptomics of Lin − c-Kit + mouse bone marrow progenitors uncovers a continuous differentiation from hematopoietic stem cells to bipotent basophil-mast cell progenitors (BMCPs). 4 Microarray analysis of bulk-sorted cells shows distinct gene expression profiles of mature basophils and mast cells. 8 However, investigation of temporal gene expression dynamics during basophil and mast cell specification and maturation is yet to be delineated and requires single-cell resolution.
Here, we combine multicolor flow cytometry-based index sorting with high-coverage scRNA-seq to investigate the basophil-mast cell bifurcation and the differentiation into each respective lineage.
We demonstrate that molecular profiling and pseudotime ordering of single cells highlights genes that are critical for cell differentiation and maturation. The analysis is accompanied with the generation of a user-friendly web resource that allows gene expression to be explored across the single-cell landscape. Finally, we use cell fate assays to show that single-cell transcriptomics and protein epitope data analysis successfully predict the fate potential of the previously uncharacterized BMCP population in the peritoneal cavity. Taken together, the current resource provides a detailed road map of the developmentally related basophils and mast cells, whose activation contributes to allergic diseases.

| Cell isolation and flow cytometry
Experiments involving mice were performed according to the United Kingdom Home Office regulations. PBS with 2% fetal calf serum (Sigma-Aldrich, St Louis, MO) and 1 mmol/L EDTA was injected into the peritoneal cavity of euthanized C57BL/6 mice. The fluid was aspirated following vigorous massage, and the cells were prepared for FACS. Peritoneal lavage samples with excessive blood contamination were discarded before data acquisition. Bone marrow cells were extracted by flushing or crushing the femurs, tibias, and/or ilia. Red blood cells were lysed, and the remaining cells were

G R A P H I C A L A B S T R A C T
Flow cytometry and single-cell gene expression data reconstruct a road map of mouse basophil and mast cell differentiation. Cell fate assays show that previously uncharacterized peritoneal progenitors can differentiate into both basophils and mast cells. An interactive web resource enables the wider research community to explore the gene expression dynamics of differentiating cells. Abbreviations: Ba, basophil; FACS, fluorescence-activated cell sorting; MC, mast cell; Prog, progenitor; scRNA-seq, single-cell RNA sequencing.
prepared for FACS. The cells were sorted with a BD Influx cell sorter (BD Biosciences, San Jose, CA). Cell doublets were excluded with the width parameters. P1 cells and mast cells were sorted two consecutive times for cell culture experiments. The cells were sorted into Terasaki plates (Greiner Bio-One, Kremsmünster, Austria) or 96well plate wells. Visual inspection determined colony sizes following culture, and the size was set to 1 if no live cells were observed in a particular well. Flow cytometry was typically performed on colonies constituting at least 20 cells, and potential to form a particular cell lineage was based on at least 5 events in a given gate, as described previously. 4 Cultured cells were analyzed with the BD Fortessa flow cytometers (BD Biosciences).

| Antibodies and cell staining
Primary cells were incubated with the antibodies integrin β7 DAPI (BD Biosciences) or 7-AAD (Thermo Fisher Scientific) were used to exclude dead cells.

| Flow cytometry analysis
FlowJo v10 (Treestar, Ashland, OR) produced the flow cytometry plots. Diffusion map and principal component analysis (PCA) plots of flow cytometry data were generated using the R programming environment. The flow cytometry events were down-sampled according to the population with the least number of events. Duplicate entries were removed, and the parameters representing fluorescent markers log-transformed. Variables were z-scored and diffusion map plots generated using the destiny and ggplot2 packages. PCA was calculated using the prcomp function. Data projection was performed using the predict function.

| scRNA-seq data analysis
Primary single cells were FACS index sorted into lysis buffer, and scRNA-seq was performed based on the Smart-Seq2 protocol. 9 For details of scRNA-seq data processing, see Supplementary methods. Analysis was performed using the scanpy v1.4 python module 10

| Multicolor flow cytometry analysis reveals the basophil and mast cell differentiation trajectories
Basophil and mast cell differentiation are closely linked, and the cells share a common bipotent progenitor ( Figure 1A). Here, we used mul-   (Table S1). Enrichment analysis of these gene F I G U R E 1 Flow cytometry analysis reveals differentiation trajectories from bipotent basophil-mast cell progenitors to basophils and mast cells. A, Illustration outlining the basophil and mast cell differentiation trajectories. B, Flow cytometry-based gating strategies of (Bi) bipotent basophil-mast cell progenitors (BMCPs) from bone marrow, (Bii) basophil progenitors (BaP) and basophils (Ba) from bone marrow, and (Biii) P1 cells and mast cells from peritoneal cavity. Lineage markers include 7-4, CD5, CD11b, CD19, CD45R/B220, Ly6G/C (Gr-1), and TER119. C, Diffusion map visualization of the flow cytometry data colored by cell type. D, Diffusion map visualization of the flow cytometry data colored by protein expression or light scatter parameters. lists revealed that upregulated genes were enriched for granulocyte immune response terms (Table S2, Figure S1B). Downregulated genes were enriched for cell cycle related terms (Table S2, Figure 2B), suggesting a difference in cell cycle behavior throughout the differentiation process. This observation is in line with other hematopoietic differentiation pathways, where progenitors commonly loose proliferative capacity as they mature into the fully differentiated cell types.

| Single-cell profiling captures progression of basophil differentiation in the bone marrow
To further explore this, we then performed analysis to computationally assign cell cycle state to the single-cell profiles. 15 Consistent with the gene list enrichment analysis, the majority of cells in the BaP gate were assigned to S and G2M states (69%), whereas 87% of cells in the Ba gate were assigned to G1 state ( Figure 2C, D).
The effect of cell cycle status was clear in the diffusion map dimensionality reduction ( Figure S1C), confounding attempts to order cells using pseudotime algorithms. Instead, downregulation of progenitor marker genes such as Cd34 and Kit indicated that ordering cells along PC1 could be used to arrange cells in pseudotime ( Figure S1D). Visualization of index sorting data also showed clear dynamics of the different surface markers along PC1 ( Figure 2E).
As expected, CD34 and c-Kit protein expression showed a negative correlation with pseudotime (compare Figures 1D and 2E), which indicates their downregulation during basophil differentiation. In addition, the basophil marker CD49b (DX5) showed a positive correlation with pseudotime ordering ( Figure 2E).
Using the PC1 pseudotime ordering, we then identified genes that dynamically changed during differentiation ( Figure 2F).
Clustering sorted these dynamic genes into two groups: one increasing and one decreasing with differentiation (Table S3). Basophil differentiation was associated with upregulation of Hdc, which is associated with histamine synthesis, and increased expression of the basophil gene E-cadherin (Cdh1). We further observed downregulation of the proteases Mcpt8, Prss34, and Ctsg and upregulation of the transcription factors Cebpa, Stat5b, and Spi1 ( Figure 2G). To validate the full lists of dynamically regulated genes, we compared these to mast cell and basophil signature genes identified using bulk microarray analysis. 8 Genes upregulated during basophil differentiation exhibited a significant overlap with the previously described basophil signature genes (P = 4.0 × 10 −29 , hypergeometric test, Figure S1Ei), whereas genes that were downregulated during differentiation had significant overlap with the previously described mast cell signature Figure S1Eii).
Together, this analysis offers a description of the dynamics of gene expression during basophil differentiation and highlights changes in cell cycle activity as one of the major occurrences during this maturation process.

| Single-cell gene expression analysis suggests a continuum of mast cell differentiation in the peritoneal cavity
After exploring the basophil progenitors, we next decided to focus on mast cell differentiation in the peritoneal cavity. The flow cytometry data suggested the existence of both peritoneal BMCPs and mast cells (Figure 1), so we performed single-cell RNA sequencing on these primary cell populations to characterize them based on gene expression. A subset of the P1 cells clustered separately from the mast cells in the diffusion map plot, demonstrating a difference between the transcriptome of these cells and the peritoneal mast cells ( Figure 3A). In previous work, we characterized bone marrow BMCPs at the single-cell gene expression level. 4 To examine the similarity of these bone marrow progenitors to the peritoneal mast cell differentiation, single-cell bone marrow BMCP profiles from Dahlin et al 4 were projected onto the peritoneal dataset ( Figure 3B). This demonstrated that the P1 peritoneal cells furthest from the peritoneal MCs were most similar to the bone marrow BMCPs, supporting that these were the most immature cells in the dataset.
To understand expression changes during mast cell maturation, we then performed pseudotime ordering of the peritoneal cells ( Figure 3C). As expected, interrogation of cell surface markers along pseudotime showed a strong downregulation of integrin β7 and strong upregulation of markers such as Sca1 and ST2 (compare Figures 1D and 3D). Genes exhibiting dynamic expression patterns were identified and clustered as for the basophil trajectory (Table S4, Figure 3E). Annotation from the Panther da-  Figure S2C).
To validate the full lists of dynamically regulated genes in the peritoneal mast cell dataset, we compared these to mast cell and basophil signatures identified in Dwyer et al. 8 The upregulated genes significantly overlapped with the mast cell signature genes (P = 3.7 × 10 −65 , hypergeometric test, Figure S2Di), including Ndst2 and Meis2 ( Figure 3H). Some genes showed expression enrichment mainly in the mast cells (Meis2), whereas others were expressed more evenly across the trajectory save for lower expression at the beginning of pseudotime (Ndst2). Similar to basophil differentiation, mast cell differentiation was associated with Hdc upregulation ( Figure 3H). There was also a small overlap between the downregulated genes and basophil signature genes (P = 2.5 × 10 −5 , hypergeometric test, Figure S2Dii). To investigate the link between gene and protein expression, we also interrogated the expression of Itga4 and Itgb7, which encode subunits of integrin β7. Itga4 was significantly downregulated with a similar expression pattern to integrin β7 in the flow cytometry data whereas Itgb7 was not significantly changing in pseudotime ( Figure 3D, H).

| P1 cells in the peritoneal cavity exhibit basophil and mast cell-forming potential
The  Figure 1C, henceforth referred to as the reference dataset ( Figure 4E). We then projected the FACS index sort data onto the principal component space of the reference dataset, and plotted colony size and colony type data ( Figure 4F). Analysis of colony sizes showed that colonies derived from P1 cells were large, whereas cells along the mast cell trajectory exhibited reduced proliferation rate ( Figure 4F, Figure S3). Notably, the cell fate assays revealed that primary P1 cells formed pure basophil colonies, pure mast cell colonies or mixed basophil-mast cell colonies ( Figure 4F, Figure S3A).
Colonies derived from single mast cells were too small to analyze with flow cytometry. However, mast cells cultured in bulk remained mast cells as expected ( Figure 4C-D, Figure S3B). Further analysis of the FACS index sort data revealed that primary cells that formed large colonies comprising basophils and/or mast cells were mainly integrin β7 +/hi P1 progenitors ( Figure S4). This observation agrees with the pseudotime ordering of the single-cell transcriptomics data, which showed that loss of integrin β7 is associated with differentiation. We also cultured the P1 peritoneal cells in erythroid-promoting conditions, as the early basophil-mast cell differentiation is closely linked to the erythrocyte trajectory. 3 However, no erythroid output was observed ( Figure S5), indicating that the P1 cells indeed con- Taken together, the cell culture assays revealed that the protein and gene expression analyses successfully predicted the differentiation state of the P1 cell population in the peritoneal cavity.

| D ISCUSS I ON
Single-cell transcriptomics coupled with index sorting of thousands of bone marrow HSPCs has previously been used to chart erythrocyte and granulocyte-monocyte differentiation. 18,19 BMCPs represent a minor fraction of the bone marrow HSPCs, and capturing the early basophil-mast cell axis therefore requires analysis of tens of thousands of HSPCs. 4 The early differentiation of progenitors with F I G U R E 2 Bone marrow basophil progenitors downregulate cell cycle genes during differentiation. A, PCA of scRNA-seq profiles colored by cell surface marker phenotype. PC, principal component. B, Top 5 GO Biological Process terms associated with the genes significantly upregulated in BaP cells compared to Ba cells, ranked by adjusted P-value. Benjamini-Hochberg correction for multiple hypotheses testing. Genes upregulated in Ba compared to BaP are presented in Figure S1B. We reveal the existence of a progenitor with dual basophil-mast cell-forming potential in the peritoneal cavity. BMCPs have previously been described in the mouse spleen and bone marrow, 4,12,25 and the presence of a bipotent progenitor population indicates that there is a close association between the basophil and mast cell differentiation trajectories. One study has questioned the bipotent nature of splenic BMCPs, 26 as only mast cell colonies were observed following culture. The failure to detect basophils in that study is yet to be explained.
Recent data suggest that the erythroid axis is coupled with the basophil and/or mast cell fates. 3,5,[27][28][29] However, we did not observe erythrocyte-forming potential among P1 cells in the peritoneum.
In agreement with this, BMCPs in the spleen and bone marrow are unable to generate erythrocytes, 4,12 altogether suggesting that loss of erythrocyte-forming potential is an early event along the differ-     ferentiation is yet to be delineated. During mast cell differentiation, we describe the increase of the transcription factor Meis2.
Primary mast cells from human skin express this transcription factor, 39 but the potential function during mast cell differentiation is yet to be described.
Microarray and RNA sequencing analyses reported previously provide detailed gene expression patterns of mature hematopoietic cell populations, including bulk-sorted mature basophils and mast cells. 8,40 We observed that differentiation into basophils and mast cells involves activation of mutually exclusive lineage programs. However, a small subset of the previously reported signature genes is not unique to mature cells, but can also be observed in bipotent progenitors. For example, we show that Mcpt8 expression is not restricted to basophils but is also expressed by BMCPs. Indirect evidence also supports the validity of this observation. 41,42 Transient Mcpt8 expression at the BMCP stage in fact provides an explanation to a major conundrum in the field.
Basophils, identified as Mcpt8-expressing cells, have been reported to exhibit potential to transdifferentiate into mast cells. 43 Our results show that a more likely scenario is that a subset of the previously reported Mcpt8-expressing cells constitutes bipotent BMCPs that can give rise to mast cells.
In summary, here we have reported the generation of a high-resolution single-cell map of the BMCP bifurcation and mast cell and basophil differentiation. A user-friendly interactive website has been created for the wider community to enable further exploration of the data.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no competing interests.