Background
Food allergy has emerged as a considerable public health concern, affecting up to 0.1–5.7% of children and adolescents under 18, and 0.1–3.2% of adults in westernized countries [
1]. Food allergic reactions are either mediated by an immunological mechanism [
2], involving allergen-specific Immunoglobulin E (sIgE), cell-mediated mechanisms in the absence of sIgE in serum or may show etiologies of both [
3,
4]. The underlying cellular mechanisms causing adverse reactions to food in subjects with a suspected food allergy but where sIgE is not detected are largely unknown [
5]. In Norway, about 50% of cases reported to The Norwegian Register of Adverse Reactions to Food did not have detectable sIgE to a standard panel of 12 food allergens in serum. Most of these cases (95%) were also negative for sIgE to inhalant allergens [
6].
Diagnosing food allergy is highly challenging and complex. A careful assessment of the clinical history is currently the most important tool for the diagnosis of food allergy [
7]. Although oral food challenges (especially the double-blind placebo-controlled food challenge) are the golden standard for an objective diagnosis of food allergy, food challenges are infrequently conducted outside the academic context as the procedure is resource-intensive, requires highly equipped specialists, and carries the risk of inducing a severe anaphylactic reaction [
8,
9]. If an IgE-mediated food allergy is suspected, assessment of sIgE either in blood or via skin prick tests is recommended to identify the offending food. However, the relation between sIgE serum levels and adversity of the allergic responses varies, thus sIgE can occur in subjects without clinical food allergy symptoms and vice versa [
10‐
12]. Also, other serum markers, such as IgG4 or cytokines, have been evaluated but were not validated as reliable diagnostic markers [
13,
14].
Whereas a few cellular in vitro procedures have been considered to support diagnosis [
4], there are still large knowledge gaps regarding cellular mechanisms.
Recent advances in cell cytometry, combining high-dimensional assessment of cellular phenotype and function with data-driven statistical algorithms [
15,
16] allow for capturing the complexity of cellular immune mechanisms in a new scope. In this regard, our objective in this explorative study was to identify phenotypical and/or functional immune cell signatures characteristic for patients reporting adverse reactions to food. The overarching goal was to obtain new insight into the cellular mechanisms of food allergy, contributing to a more accurate clinical diagnosis.
By the use of mass cytometry/CyTOF (cytometry by time of flight), we performed comprehensive profiling of peripheral blood mononuclear cells (PBMC) from adult patients reporting adverse reactions to food and healthy controls. The patients were grouped according to sIgE-positive or sIgE-negative serology to common food and inhalant allergens. Using a combination of manual gating strategies and data-driven approaches [
15], immune cell profiles and functional cell subpopulations differing between the groups of participants were identified.
Discussion
By the use of broad antibody panels and data-driven analyses, we identified several cell populations where the combination of marker expression was significantly different between individuals with adverse reactions to food (both sIgE-positive and sIgE-negative to common food and inhalant allergens), and healthy controls. While no differences were seen for more traditionally targeted serum markers (antibodies and proteins/cytokines), our results illustrate the great potential of high-dimensional analyses on single-cell level, such as mass cytometry, for the identification of new diagnostic biomarkers and/or mechanistic knowledge.
Whereas cell population phenotypic markers alone did not differ between the groups, several functional markers, such as the expression of activation markers in unstimulated cells, or (co) expression of cytokines in PMA/ionomycin-stimulated cells did differ. Interestingly, most of the cell features were similar in the patients reporting adverse reactions to food regardless of the presence of sIgE in serum.
Most IgEpos patients, except for P11 and P9, had sIgE to both food and inhalant allergens. In spite of the suspicion of cross-allergies, the results suggest that the cellular signature and several immune cell mechanisms may be common for these two groups of patients. Furthermore, it indicates that the group of sIgE-neg patients indeed were food allergic patients according to the nomenclature in Johansson et al. (2001), defined by responses mediated by an immunological mechanism. Although we cannot exclude that the food allergy patients in the IgE-neg group have sIgE in serum not detected by our ImmunoCap panel and food extract dot blot matrix, our results indicate that cellular mechanisms may be important and that cellular features may serve as helpful diagnostic markers of food allergy. Our results also underline the large knowledge gap of the underlying cellular mechanisms of hypersensitivity reactions in general.
The most striking observation was an apparent impairment of polyfunctionality of both Th (CD4+) and Tc (CD8+) cells in the food allergic individuals compared to healthy controls, although only reaching statistical significance in the IgEneg group. This result was consistent, with both lower abundance of TNF-α and IFN-γ producing Th and Tc cell subpopulations as well as lower expression of these cytokines per cell. Some of the subpopulations of these polyfunctional Th and Tc cells also co-expressed IL-2.
Likewise, TNF-α and IFN-γ co-expression was lower in NK cells. The effects on polyfunctional cells were supported by a similar pattern in several groups of parent-child clusters. In agreement, reduction in TNF-α and IFN-γ individually was observed even after manual gating on NK cells, CD4+ and CD8+ T cells. Further, we can exclude the possibility that the reduction in the two cytokines was not due to methodological errors such as lack of antibody staining or other systematic errors since other cell clusters with high expression of TNF-α and/or IFN-γ were not reduced in the allergy groups.
Polyfunctional cells are cells simultaneously expressing two or more immune mediators (cytokines/chemokines) [
18], and have been described for CD4
+ and CD8
+ T cells, NK cells and monocytes [
19,
20]. Polyfunctional cells have been shown to provide a more effective immune response to various pathogens such as human immunodeficiency virus (HIV) [
21],
Leishmania major [
22], and
Mycobacterium tuberculosis infections [
23], than cells that produce only single cytokines, and reflect functional efficiency in vaccination [
24]. Polyfunctional T cells have also been shown to play a role in certain autoimmune diseases [
25]. Functional consequences of lower levels of polyfunctional T cells in food allergy may, therefore, be hypothesized. On the other hand, the lower abundance and TNF-α/IFN-γ cytokine response to PMA/ionomycin could also be a result of cell exhaustion in the observed Th, Tc, and NK cell populations [
26‐
28] and/or Th2-skewing of T cell responses in the two allergy groups, as would be expected in particular for the IgEpos group [
29]. The observation can depend on the choice of PMA/ionomycin as the stimulant since the stimulus strongly influences the immune signature [
30]. Nevertheless, our results indicate that certain cell populations from the two allergy groups respond with altered ability for combined cytokine production compared to the control group in the present setup. This points to polyfunctional cells as a potential diagnostic biomarker for food allergy and deserves focus in future studies
.
Both TNF-α and IFN-γ have previously been reported to be relevant for food allergic responses [
23]. In agreement with our current findings, Osterlund et al. have reported decreased frequencies of IFN-γ expressing CD4
+ T cells [
31] and decreased production of TNF-α in culture supernatants of PBMC from children with cow’s milk allergy [
32].
CITRUS did not detect expression of the Th2 cell cytokines IL-5, IL-10, or IL-13, cytokines that are strongly associated with food allergy [
33]. The reason for this could be the type of stimuli, as described above, or the low frequencies of allergen-specific cells taking the limited amount of acquired cells into consideration [
34].
In unstimulated cells, the cell count within each subpopulation did not differ significantly between the groups, neither the percentage of the conventional cell populations identified by manual gating nor in populations clustered based on all markers or only the phenotypic markers (by unsupervised clustering in CITRUS). However, for food allergic individuals, the expression of the activation/functional markers CD371, CD69, CD28, HLA-DR, and CD25 per cell was higher and CD23 was lower in several parent-child groups of cell subpopulations. The effects were observed in both allergy groups, and several of the activation markers were altered in the same cells. For instance, monocyte subpopulations showed higher levels of CD371 (inhibitory receptor [
35]) and CD69 (early activation marker [
36]) in both allergy groups and with simultaneously decreased CD23 (low affinity IgE receptor [
17]) expression only in the IgEneg allergy group. Taken together with recent literature suggesting that monocytes might have a pivotal role in some non-IgE-mediated disorders [
37], our observations suggest that the activation status of monocytes could be of interest for further studies of diagnostic markers and mechanisms of food allergy. Furthermore, also the higher expression of HLA-DR and CD28 are biologically relevant since they are involved in the crosstalk between antigen-presenting cells (e.g. B cells/DCs) and T cells during activation and crucial for the maintenance of immune homeostasis [
38,
39]. Moreover, increased CD25 expression (high-affinity heterotrimeric IL-2 receptor [
40]) might counterbalance the decreased expression of IL-2. Our data indicate that the expression of markers indicative of activation state and/or function are more potent in reflecting disease-dependent characteristics/features than cell population frequencies.
We exploited the power of unsupervised cluster analyses to identify subpopulations of cells with combinations of phenotypic and functional markers that were overlooked in a manual gating approach. While the choice of clustering algorithms and statistical analyses will to a certain degree influence the outcomes, our main findings were confirmed in different ways, strengthening the conclusions: i) by running all clustering analyses at least twice with similar results ii) in most cases, the results obtained by the SAM method for identifying cell clusters differing between the groups (with corrections for multiple comparisons) were confirmed by the PAMR results (data not shown), iii) the application of the relatively conservative Kruskal-Wallis test for pairwise comparisons between the three groups led to a reduction in significant cell populations reported.
The antibody panels were designed to cover a broad spectrum of different immune cells, but still only included a selection of markers to study activation, maturation, and proliferation status of the immune cells. In future studies, the use of whole blood should be considered, to assess neutrophils, basophils, and especially eosinophils, which are strongly associated with several disorders not associated with sIgE [
5]. As suggested by Goswami et al. (2017), food-antigen specific cells with a pathogenic phenotype in allergic patients with an sIgE-negative serology might not be found in circulation, but be mainly localized to the gastrointestinal tract [
37]. However, in the search for diagnostic markers, blood is the preferred matrix.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.