Immune cell profiling reveals natural killer and T cell subpopulations to be associated with atopic dermatitis severity.

Abstract

Atopic dermatitis (AD) is a chronic inflammatory skin disease characterized by a genetically determined impaired epidermal barrier function and predisposition to develop other atopic comorbidities.1 The role of adaptive and innate immunity, in particular IL-4 and IL-13 and T lymphocyte populations in AD pathogenesis are established, but the exact cellular mechanisms are not fully understood. At least partially, this is attributed to technical limitations of flow cytometry, that is, limited channels and signal spillover. Mass cytometry overcomes these limitations and allows advanced bioinformatic methods. An unbiased comprehensive analysis of the total peripheral immune cell pool in AD is not available. Peripheral leucocytes were analysed from 60 well-defined adults (18–55 years) comparable regarding sex and age and recruited at the Charité – Universitätsmedizin Berlin within 5 months. Written informed consent was obtained from all participants. All procedures were approved by the local ethics committee (EA1/080/16) following the Declaration of Helsinki. The AD group was defined by international criteria,2 including age-related eczema course, elevated serum IgE (>100 kU/L) and aeroallergen sensitization, absence of systemic treatment. The clinical severity determined by SCORing Atopic Dermatitis (SCORAD) was applied to stratify into two subgroups: moderate (SCORAD 15–40, ADm, n = 20, mean 28.8 ± 7.0 and severe: >40, ADs, n = 20, mean 56.7 ± 12.5, intergroup T-test p < .001). The reference group with healthy donors (HD) presented without history for atopy (self and family), chronic diseases and the absence of medication intake. The groups were comparable between age (Chi-square Pearson Test p = .153) and sex (ANOVA p = .546). The analysis includes 39 surface markers considering all immune cell lineages in the blood. An unbiased, automated cell clustering algorithm (ImmunoClust) identified frequencies of 70 cell clusters and 103 linked populations, of which 44 differed significantly between the groups to be changed in AD samples when compared to healthy donors (HD). In ADs 49 populations and in ADm 20 populations were found (p < .01). Most of them can be attributed to the CD4+, CD8+, eosinophil or NK lineage. A principal component analysis (PCA) using frequencies of AD-related immune cell clusters and the SCORAD values revealed a clear separation of HD and ADm from ADs samples (Figure 1A), while HD and ADm samples were overlapping. The candidate populations identified by the unsupervised analysis were confirmed by manual gating. The frequencies of eosinophils in both AD groups were higher compared to controls with increased frequencies of CD45RO+ eosinophils in ADs (Figure 1B), which were described as being less apoptosis-prone compared to other CD45 isoforms.3 The role of eosinophils is currently investigated applying anti-IL-5 receptor alpha antibodies (NCT03563066, NCT04605094). No change in frequencies of Th2 cells (CD3 + CD4 + CD45RO + CRTh2+) were observed in the ADs and ADm group respectively (p = .25–.66, Figure 1C). Type 2 innate lymphoid cells (ILC2) were present at very low frequencies in the blood limiting a precise quantification, but were rather reduced in AD (p = .002, Figure 1C). Our data suggest that Th2 and ILC2 cells exert their functions in in the inflamed skin, and are only rarely detectable in the periphery.4 Unexpectedly, reduced frequencies of total CD56+ NK cells were identified in AD samples by the unbiased analysis and confirmed by manual gating. These results support an increased NK cell apoptosis in the blood or an increased extravasation in the inflamed skin.5 The frequency of a CD38 + CD69+ NK cell subset, resembling most probably activated NK cells after tissue egress, was increased in ADm, an effect much more pronounced in AD (Figure 1D) and indicating NK cell activation in the AD skin,6 for example, by driving Th2-cytokines in keratinocytes.4 Besides Th2 also increased frequencies of activated T helper effector memory cells (CD3 + CD4 + CD8–CD45RO + CD45RA–CCR7–CD127–CD27 + HLADR) and activated Treg's (CD3 + CD4 + CD25hi CD127–CD38 + HLADR) were identified in all AD samples, which could be confirmed by manual gating (Figure 1E), with additional subtle differences on other CD4+ and CD8+ T cell subpopulations. These data underline the important role of T cells in AD, which was also highlighted in a recent study using mass cytometry analysis on CD4+ T helper cells in AD.7 Of note, no significant association to ADs or ADm compared to HD were observed in naïve, memory, CD27+/IgD+ B cells or plasmablasts as such or considering activation marker coexpression, for example, CD38 or CD69, myeloid cells, including neutrophils, basophils, CD14+ monocytes stratified into classical and non-classical cells considering CD11c or activation marker coexpression (data not shown). Next, we aimed to correlate the most prominent cell subpopulations identified in this study, including eosinophils, NKT cells, activated CD8+ or CD4+ T memory cells and activated Treg, but not NK cells with severity determined by SCORAD. In order to avoid a bias by the HD group, we determined with predicted disease severity (pEASI, calculated by the serum concentrations of soluble CD25, IL-22 and C-C-chemokine 17 [CCL17, TARC]) from all the sera (Figure 2 and data not shown). These data indicate that the cytokine serum levels as components of the pEASI might be associated with the cell populations identified and support a potential regulatory role of these cell types in AD. In conclusion, mass cytometry-based immune cell profiling revealed the presence of immune cell subsets of different lineages including eosinophils, NK, NKT and activated T cell populations, which may contribute to AD pathophysiology. A better understanding of the inter-cellular interactions of these subsets may improve our understanding of AD pathomechanism and will open new treatment options of AD beyond cytokine blockade or JAK inhibition. M.W. designed the study, V.G. performed all experiments and analysed the data, S.B. established the antibody panel and CyTOF-measurements, T.S. analysed the data, H.M. analysed the data, A.G. supervised the experimental procedures, G.H. designed the study and analysed the data. All coauthors contributed to writing of the manuscript. We thank Heike Hirseland and Dennis Ernst for outstanding technical support of the project. Dr. Worm reports grants from Deutsche Forschungsgemeinschaft (DFG), during the conduct of the study; personal fees from Biotest AG, personal fees from Sanofi-Aventis Deutschland, personal fees from ALK-Abelló Arzneimittel, personal fees from Mylan, personal fees from Leo Pharma, personal fees from Regeneron Pharmaceuticals, personal fees from DBV Technologies, personal fees from Stallergenes, personal fees from Bencard Allergie, personal fees from Aimmune Therapeutics UK limited, personal fees from Novartis AG, personal fees from Allergopharma, personal fees from HAL Allergie, personal fees from Biotest, personal fees from AbbVie Deutschland, personal fees from Eli Lilly Deutschland, outside the submitted work. Dr. Grützkau reports grants from Deutsche Forschungsgemeinschaft (DFG), grants from Leibniz Gesellschaft, grants from IMI-JU, during the conduct of the study. Dr. Heine reports grants from Deutsche Forschungsgemeinschaft (DFG), during the conduct of the study; personal fees from Allergopharma, personal fees from Biotest, personal fees from Eli Lilly, personal fees from Sanofi, personal fees from Abbvie, outside the submitted work. This project was supported by the Deutsche Forschungsgemeinschaft Grant/Award Number: SFB650TRR130 (DFG – SFB650 TP5 to MW, GH and DFG – SFB650 Z6 to AG, and further SFB financial support to TP5 for GH, MW, AG, DFG – SFB-TRR130-P19 to MW and GH, Leibniz Science Campus Chronic Inflammation and IMI JU-funded project RTCure to A.G.).