Single‐cell <scp>RNA</scp> sequencing of human blood eosinophils reveals plasticity and absence of canonical cell subsets

Abstract

Eosinophils are leukocytes involved in homeostasis and diseases like asthma. The existence of eosinophil subpopulations remain controversial, with eosinophils being classified as type 1, type 2, inflammatory, resident, homeostatic, or developmental depending on the publication,1 with no clear indication of the existence of intrinsic subpopulations or environmental dependence. Unsupervised methods could provide data on eosinophil true subclassifications, which could be of importance for the efficacy of asthma therapy.2 Therefore, the aim of this study is to analyse blood eosinophils using single-cell RNA sequencing (scRNAseq) (BD Rhapsody™) from three asthmatic subjects and three healthy controls (Table S1). A total of 23,031 isolated individual cells were sequenced, which were mainly annotated as eosinophils (File S1; Figure S1A). Unsupervised clustering yielded 6 clusters (C0–5; Figure 1A; Figure S1). Figure 1A represents individually each study subject, depending on being healthy or asthmatic and in the experiment day (each day an asthmatic and a healthy sample were run in parallel), being batch differences considered in the in-silico analysis. Clusters 0, 1, 2, 3, and 5 were relatively homogeneous, whereas C4 is the more diverse cluster, mainly drifting from the third asthmatic individual (Figure 1A; Figure S1). Gene expression variation between each cluster was low, with C0 having 16 differentially expressed genes, C2 with 2, C3 with 7, and C5 with 4; and only C1 with 41 and C4 with 102 having high numbers (Table S2). As seen in Figure 1B; Figure S2, the most representative genes for each cluster (and between conditions) are genes previously related to eosinophils and asthma (CCR3 and CAT for C1, SP100 and SP110 for C3, CLC and S100A8-9 for C4, or CCL4 and CCL4L2 for C5). Expression of eosinophil markers was conserved between clusters. Interestingly, C1 differentially expressed CCR3, ANXA1, SIGLEC10, and ITGB2, whereas C4 showed higher expression of CLC and SELL (CD62L) (Figure 2A). We applied 3 scores composed of averaged gene expression (Table S3), with clusters having similar scores, even between disease status (Figure 2B), except C4 with higher Th2 asthma score (Figure 2C). Notably, C4 presents the highest transcriptomic variation (fivefold) between healthy and asthmatics (Table S4). EnrichR analysis (Figure 1C; Figure S3) evidenced that C0 was enriched in ATPase, ion transmembrane transport GO, and NOD or TNFα pathways (Figure 1C; Figure S3). C1 is characterized by pathways and processes related to chemokines, pathogen immune defence, calcium, and GPCR signalling. C2 has no process or pathway annotated, probably due to low differential mRNA expression associated. C3 expresses nucleoside triphosphate and tuberculosis responses (Figure S3). C4 and C1 share leukocyte transendothelial migration, granule secretion, oxidative stress, and immune response mechanisms like TLR/NOD (Figure 1C; Figure S3). Besides this, C4 also is enriched in processes and pathways related metabolic and cellular activity (Figure 1C; Figure S3). Finally, C5 presents immune-related mechanisms. TTRUST analysis revealed that C1 was enriched in eosinophil transcription factors, C3 in interferon responses, and C5 in NF-κB and STAT-6 (Figure S3). C4 had no specific transcription factor network, whereas C1 presents highest number of associated transcription factors, which could describe eosinophils in different maturation state. This is also supported with the data showing that both C1 and C4 share pathways and functions related to immunity (against pathogens, granule secretion, oxidative stress) but C4 presents higher amount of asthma related gene expression and increase cellular metabolism and activation processes, a sign of cellular activation. According to our results, blood eosinophil transcriptomics is homogeneous, compromising the existence of intrinsic subpopulations. A 2016 study using mouse asthma models identified two different eosinophil lung subpopulations (resident and inflammatory) distinguished by phenotype and function as well as their possible reaction to IL-5, which opened the field to the idea of the existence of eosinophil subpopulations.3 The independency of eosinophils form IL-5 was recently questioned, mice-anti-IL-5 treatment depleted all eosinophil populations, suggesting that eosinophils exist on an activation continuum.4 As previously mentioned, other studies also proposed different eosinophil subtypes,1 most of them described in mouse, such as eosinophils with differential expression of GR-1 and cytokine expression in lung in allergic reactions.5 Many studies on mice intestinal diseases have also shown different eosinophil subpopulations, such as CD11chi intestinal antigen presenting eosinophils that differ from those from the intestine lamina propria and blood.6 A study from Diny and coworkers shed light on the eosinophil differences in intestine describing that the aryl hydrocarbon receptor partially shapes eosinophil transcriptome in response to environmental triggers.7 Furthermore, scRNAseq and in vivo experiments of gastrointestinal disease mice models concluded that eosinophil sub specialization is a continuous differentiation process controlled by the tissue milieu, its signalling molecules and microbiota, a process that is sustained on their lineage plasticity and the sequential ontogeny observed for the eosinophil precursor-blood-intestine subpopulations in the trajectory analysis.8 In accordance, Abdala–Valencia et al also proposed that it is the tissue microenvironment which shapes the eosinophils phenotype and function.9 These data support our findings, with all clusters having immune response pathways and eosinophil activation being prominent in C4 (increased S100A4, S100A8, CLC, SELL, associated with worst lung function or male sex), which reinforce that there might not be intrinsic eosinophil subpopulations in blood, but rather when the cell arrives to the tissue, then microenvironment changes eosinophils by virtue of their plasticity. Recently it has been published that CD62Llow inflammatory blood eosinophils are abundant in asthmatics.10 Conversely, we found that C4 does not fit this paradigm of CD62Llow inflammatory eosinophils, as it was increased in the asthmatic with worst lung function and presents higher Th2 asthma score, in addition to the fact that it is the most distinct cluster between healthy volunteers and asthmatics. A possibility is that being CD62L cytokine-dependent (downregulated by IL-5 and GM-CSF and upregulated by IFNγ),11 its levels on eosinophils may be affected by tissue microenvironment and disease status more than define eosinophil subpopulations. Being this study the first report using scRNAseq of human blood eosinophils we provide with an unsupervised approach into how are eosinophils at a transcriptomic level, proving that there are indeed differences between the clusters, but the overall results highlight certain cellular homogeneity in blood and differences attributable to activation states, differing from results for other immune cells like lymphocytes or monocytes,12 and from eosinophils located at tissue.8 Furthermore, as all clusters homogeneously represented in healthy and asthmatics blood eosinophils points to the hypothesis that eosinophils have high plasticity and acquire differential maturation, phenotype and functional states after tissue localization and stimulation as previously mentioned. It is worth mentioning that this study also presents limitations, first, although the number of sequenced cells is fine (loaded 20,000–40,000 cells per sample, retaining around 23,031 sequenced cells after filtering, a high number given eosinophil fragility) the sample size is small, and accounts for a small representation of the asthmatic population heterogeneity, which should be improved in future studies, where the inclusion of more and different subjects of study could provide even with data on asthma phenotypes and endotypes and their relationship with blood eosinophils. Moreover, all patients included in the study were phenotyped as severe eosinophilic asthmatics (and over time were treated with biologics), but present heterogeneous characteristics (lung function, FeNO, exacerbations…), which can be observed for C4, that mainly derives from one subject with worst lung function, high FeNO and no exacerbations the last year, proving that, as previously mentioned, more studies should be performed in order to better define eosinophils in the context of asthma heterogeneity. In conclusion, this is the first human blood eosinophil scRNAseq from healthy and asthmatic showing that at the transcriptional level, eosinophils are a homogeneous population (even between healthy/asthmatic conditions) with few gene and in silico differences resembling transient cellular activation phenotypes modulated by homeostatic and pathophysiological status. VdP conceived of the manuscript and designed the study. JMR-M, SN-G, SC, AB, CL-S, MG-M, ZG-dC, and JAC performed experiments; JMR-M, SN-G, AD, CR-R, CT, FS-C, and VdP performed data analysis. D-B, JB, MV-M, EJP-R, OS-P, MJR-N, and JS included patients and retrieved clinical data. JMR-M, SN-G, and VdP edited and wrote the manuscript. All authors contributed to the article and approved the submitted version. The authors would like to thank all the patients for their voluntary participation as well as all technical and nursing staff involved in the project (Manuela Garcia del Potro, Erica Aguado Wakui, Esther Gamella Álvarez, María Remedios Marquina Valero, Almudena Batanero Rodríguez and Raquel García Latorre). The authors also recognize Oliver Shaw, for his revision and editing in English. This work was supported by ISCIII—Instituto de Salud Carlos III and co-funded by the European Union, FIS (Fondo de Investigación Sanitaria—Spanish Health Research Fund) grants PI21/00896 and FI19/00067; Miguel Servet Program (CP23/00017); Ciber de Enfermedades Respiratorias (CIBERES); Ayudas para el fomento de la investigación de la Fundación de la SEAIC grants 22A07 and A21_09; Comunidad de Madrid grant PEJ-2021-AI_BMD-22320 and FEDER funds (Fondo Europeo de Desarrollo Regional). JMR-M reports receiving payments for lectures and educational events form Astra Zeneca and GSK. D-B reports having been under contract with the Instituto de Salud Carlos III (Rio Hortega Research Contract). MJR-N reports receiving research grant support from AstraZeneca, and having received payments for lectures from AstraZeneca and SANOFI. MV-M reports receiving payments for lectures and support for attending meetings and/or travel by AstraZeneca, GSK, Gebro, and Organon S.A.; as well as having received grant support for research from Organon S.A. VdP has received honoraria (advisory board, speaker) and/or institutional grant/research support from AstraZeneca and GSK and has held an unpaid leadership or fiduciary role in committees belonging to the EAACI. The rest of authors declare no conflicts of interest. The data that support the findings of this study are available from the corresponding author, [VdP], upon reasonable request, and scRNAseq is deposited in the repository: https://www.ebi.ac.uk/biostudies/arrayexpress/studies/E-MTAB-14010. Figure S1. Figure S2. Figure S3. Data S1. Table S1. Table S2. Table S3. Table S4. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Topics & Concepts