A phenotypically distinct human Th2 cell subpopulation is associated with development of allergic disorders in infancy

Abstract

Background:

Little is known about the ontogeny of T cell immunity during infancy in farming and urban lifestyle due to lack of immunophenotyping in such birth cohorts.

Methods:

Two birth cohorts (farming and urban) at differing risks and rates of allergic diseases were compared. Blood mononuclear cells were collected from infants at birth, and 6 and 12 months of age. Full spectrum flow cytometry followed by traditional gating and the Scalable Weighted Iterative Flow-clustering Technique (SWIFT) high dimensional analysis were used to identify cell populations that differed between farming and urban infants. Additionally, single-cell RNAseq and multiplex cytokine assays were used to assess the function of cell populations of interest.

Results:

Several regulatory T cell (Treg) subpopulations were elevated in farming lifestyle and in non-atopic infants. A unique effector memory CD25⁺CD127⁺CD161⁻CD49d⁺ CCR4⁺CRTH2⁺ Th2 population was elevated at 6 months in urban infants and in those who developed atopic dermatitis and/or food allergy and allergic sensitization. Although this population shared Th2 and IL-9 skewing with Th2A cells, the population uniquely failed to express CD161, produced more IL-2 and TNF-α, and upregulated the differentially expressed genes (DEGs), FOXP3 and the cytokine inducible SH2-containing protein gene (CISH) relative to Th2A cells. This population has been termed Th2B cells.

Conclusion:

We describe a unique effector memory Th2 population elevated in urban high-risk infants, potentially implicated in the development of allergic disease.

Graphical Abstract

• Unbiased clustering identifies a CD25⁺CD127⁺CD161⁻CD49d⁺CCR4⁺CRTH2⁺ Th2 cell population, termed Th2B cells.

• Th2B cells are elevated in urban lifestyle infants and those with atopic outcomes.

• This cell population is phenotypically distinct from both conventional Th2 cells and Th2A cells.

INTRODUCTION

The prevalence of allergic diseases, such as atopic dermatitis (AD), food allergy (FA), and asthma, has increased over the past few decades¹. Allergic diseases are characterized by a T helper 2 (Th2) response^2,3. Recently, a population of pathogenic effector Th2 cells⁴ was described in 4- to 26-year-old individuals with established peanut allergy⁵. These allergen-specific Th2A cells express a memory, terminally differentiated phenotype, and co-express the chemoattractant receptor-homologous molecule that is expressed on Th2 cells (CRTH2), C-type lectin receptor CD161, and CD49d (integrin α4). This cell population has increased production of IL-5 and IL-9 in comparison to their conventional Th2 counterparts⁵. The role of these cells in peanut immunotherapy in childhood was recently described^6,7, but the ontogeny of these cells in early life has not been previously assessed.

Although heredity predisposes to allergic disease, environmental factors, including a Western lifestyle, microbial exposures, and diet likely play a role in the risk of developing atopy⁸. Despite the exact mechanisms behind Th2 skewing in allergic disease remaining unclear, the original hygiene hypothesis proposed that decreased microbial exposure in early childhood contributes to Th2 skewing due to a lack of T helper 1 (Th1) activation through microbial stimulation⁹. Several epidemiological studies have revealed a lower prevalence of asthma and allergic diseases in rural, farming communities, supporting the hygiene hypothesis^10–14. An increase in regulatory T cells (Tregs) is associated with farming lifestyle children, who are at low risk for asthma or allergic disease development¹⁵.

There is limited insight into the mechanisms of the development of allergic diseases within the first year of life¹⁶— a critical window when allergic disease begins to develop. This is due to a paucity of birth cohorts with immune cell phenotyping before the onset of atopic disease. To identify biomarkers of either allergic sensitization or protection in early life, cord and peripheral blood samples from infants in two populations at different risk of allergic disease: urban/suburban Rochester, NY infants from atopic families, and a traditional farming lifestyle Old Order Mennonite (OOM) community of western New York at low risk for allergic diseases (Zooming in to Old Order Mennonites “ZOOM” study)¹⁷ were studied (Figure 1A). Immunophenotyping was performed from samples collected up to one year of age by full spectrum flow cytometry for CD4⁺ T cells and Treg populations. Herein, we report a unique effector memory CD161⁻CCR4⁺CRTH2⁺ Th2 population that co-expresses CD25 and CD127, which we have termed Th2B cells. This population emerges by 6 months of age in urban, high-risk infants and in those who go on to develop AD, FA, and/or sensitization.

Figure 1: Overall study timeline; Total and memory CD4⁺ cells and Treg populations change with age.

A) Overall study timeline. Regulatory T cells (Tregs) were defined as CD25⁺Foxp3⁺CD127⁻. Of those, memory cells were defined as being CD45RO⁺. (B) Percentage of total and memory CD3⁺CD4⁺ Th cells over time. (C) Percentage of total and memory Tregs over time. (B and C) CB n = 40, 6m n = 40, 12m n = 69. (D) Total and memory CD3⁺CD4⁺ cells in OOM versus urban infants. (E) Total and memory Tregs in OOM versus urban infants. (D and E) CB n = 20 OOM + 20 urban, 6m n = 20 OOM + 20 urban, 12m n = 35 OOM + 34 urban. Box indicates median and interquartile range in all panels. Points that fall outside of whiskers are outliers determined by Tukey’s test. One-way ANOVA was used to assess differences between ages in panels A and B. Mann-Whitney U-test was used to assess differences between OOM and urban in panels C and D. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. OOM, Old Order Mennonite. CB, cord blood

METHODS

Study population

Mother-infant dyads were recruited prenatally during the second and third trimesters from the OOM of Western New York and urban/suburban Rochester, NY populations as part of the “Zooming in on Old Order Mennonites” (ZOOM) cohort, described elsewhere in detail¹⁷, see Fig 1A for the study schematic.. All participants were recruited with written informed consent. The study protocol was approved by the University of Rochester Medical Center Research Subjects Review Board (#1220). Cord blood mononuclear cells (CBMCs) were collected from infants at birth and peripheral blood mononuclear cells (PBMCs) were collected at 6 and 12 months of age. Only infants that completed their 12-month visit were included in this study, culminating in 78 OOM and 79 Rochester infants. We subsampled infants for full spectrum flow cytometry analysis, resulting in 20 OOM and 20 Rochester cord blood (CB) and 6-month blood as well as 35 OOM and 34 Rochester 12-month blood from a total of 56 OOM and 51 Rochester individual infant samples utilized in the present study. Due to limited sample volume, CB and 6- and 12-month peripheral blood samples with the most numerous PBMCs available were chosen for full spectrum flow cytometry. Additionally, our chosen samples included all individuals that were diagnosed with allergic disease if samples were available to allow comparisons based on allergic outcomes.

Assessment for infant outcomes of allergic diseases

Details regarding assessment of infant outcomes of allergic diseases is detailed elsewhere¹⁷. In brief, allergy symptoms were assessed via surveys and during in-person study visits performed on all infants at 1–2 weeks, 6 weeks, 6 months, 12 months, 18 months and 24 months by the sub-investigators. In addition, blood draw and specific IgE testing was attempted at 1 and 2 years of age in every subject. Plasma samples were assessed for specific IgE measured by ImmunoCAP (Thermofisher) to cow’s milk, hen’s egg white, peanut, cat, dog, and house dust mite. Sensitization was defined by a specific IgE cut-off value of 0.35 kU/l.

Any children in both cohorts who displayed symptoms of allergic disease, such as chronic rash, hives, food reactions, food intolerance, vomiting, diarrhea, bloody stools, wheeze/persistent cough and chronic rhinitis, were evaluated by the study physician, and allergen-specific IgE tests (ImmunoCAP, Thermofisher), skin prick tests and oral food challenges, as appropriate to refute or confirm food allergy, were performed guided by these symptoms. Food allergy (FA), including non-IgE-mediated FA, was diagnosed based on the NIAID sponsored expert panel guidelines on diagnosis and management of FA¹⁸. The diagnosis of atopic dermatitis (AD) was based on recurring or chronic red, pruritic rash in a distribution age-typical for AD, also queried from medical records. Food or aeroallergen sensitization was defined by a positive skin test or specific IgE. Data on outcomes of allergic disease made by 2 years of age in this study are detailed in Supplemental Table 1. The average age of symptom onset was 6.4 (range, 1–24) months. The proportion of subjects included in this study that developed AD, FA, and/or sensitization to food or aeroallergens was significantly different between OOM and urban lifestyle infants (p=0.0012 by Fisher’s exact test).

Sample Collection

Umbilical CB and peripheral blood were collected in tubes containing anticoagulant (heparin or ACD-A). CB was processed within 2–18 hours¹⁹ and peripheral blood within 1–6 hours. Blood was diluted 1:2 in Hanks’ Buffered Saline Solution (HBSS) (Gibco, Thermo Fisher Scientific) prior to processing. PBMCs were isolated using density centrifugation in Ficoll-Paque (Sigma-Aldrich) using SepMate tubes (STEMCELL Technologies). Interphase cells were collected and centrifuged to obtain a PBMC pellet. Remaining plasma was collected and frozen at −80°C for storage. PBMCs were counted after ACK lysis and 2×10⁶ to 7×10⁶ PBMCs were frozen down in fetal bovine serum (FBS) (Corning) 20% DMSO (Corning) in liquid nitrogen for storage.

Full Spectrum Flow Cytometry

Frozen PBMCs were thawed and stained with fixable green Live/Dead (Thermo Fisher Scientific) to exclude dead cells during analysis. PBMCs were then blocked with Human TruStain FcX Fc Receptor Blocking Solution (BioLegend). They were then stained with a cocktail of fluorophore-conjugated antibodies for surface markers (Supplemental Table 2). Stained PBMCs were fixed and permeabilized using the Foxp3/Transcription Factor Staining Buffer Set (eBioscience). They were then stained with a cocktail of fluorophore-conjugated antibodies for intracellular markers (Supplemental Table 3). UltraComp eBeads (Invitrogen, Thermo Fisher Scientific) were used for single-stained positive controls. Full spectrum flow cytometry was performed on an Aurora Full Spectrum Flow Cytometer (Cytek Biosciences) with 4 lasers (UV, Violet, Blue, and Red). Spectral unmixing was performed using SpectroFlo software (Cytek Biosciences) and data was analyzed using FlowJo (BD Biosciences).

SWIFT Registration and Analysis

After spectral unmixing, sample data were preprocessed by censoring 1) saturated events at or above the limit of detection in any raw channel; and 2) time-defect events that contribute to inconsistent signals over time. To reduce batch effects, swiftReg²⁰ was performed. A batch-concatenate was produced, representative of all samples in batch 3, including both unstimulated and stimulated samples. A SWIFT^21,22 cluster template was produced from the batch 3 concatenated consensus sample, then similar batch-concatenates of batches 1 and 2 were registered to this batch 3 reference cluster template by neighbor-dependent cluster registration (NDCR) in swiftReg. This produced two registration templates that contained the registration vectors that needed to be applied to each cluster, to register batches 1 and 2 to batch 3. These vectors were then applied to all individual samples in batches 1 and 2. This produced a set of preprocessed, registered samples used for subsequent manual (traditional) gating and algorithmic analysis. A large concatenate sample (4 million events) was produced by sampling random events from all registered 12-month samples. This concatenate was clustered in SWIFT, resulting in a cluster template representative of all samples. All individual samples were then assigned to this master cluster template, as previously described^20–23. This procedure allows direct comparisons for each cluster between different biological groups. A UMAP projection of the clusters generated using SWIFT algorithm are shown in Supplemental Figure 5. See Supplemental Methods for full detail.

PBMC Isolation and Culture

Three previously cryopreserved PBMC samples from one- to two-year-old atopic infants were chosen for cell sorting. PBMCs were thawed and then blocked for 10 minutes at 4°C with Human TruStain FcX Fc Receptor Blocking Solution (BioLegend). For bulk RNAseq, cells were immediately stained with antibodies listed in Supplemental Table 4 and sorted into CD4⁺CD25⁻ and CD4⁺CD25⁺CD127⁺ using FACSAria II (BD). Representative gating plots for these populations are provided in Supplemental Figure 11. Information regarding the number of sorted events per sample is provided in Supplemental Table 5. Cryopreserved PBMCs from 2-year-old infants were stained with antibodies listed in Supplemental Table 6 and were FACS sorted to enrich Th1 (CD4⁺CXCR3⁺), Th2 (CD4⁺CDTH2⁺CD25⁻CD127⁻), Th2A (CD4⁺CRTH2⁺CD161⁺), and CD4⁺CRTH2⁺CD161⁻CD25⁺CD127⁺ cells. The number of sorted events per sample for cytokine production is shown in Supplemental Table 7. Cells were then cultured in AIM-V medium (Gibco, Thermo Fisher Scientific) 2.5% human AB serum (GeminiBio) and stimulated with 50ng/mL PMA (InvivoGen) and 1μM/ml ionomycin (Sigma) for 4 hours for one-year-old samples and overnight for two-year-old samples. Culture supernatants were collected from both one- and two-year-old samples for cytokine analysis and cell pellets from one-year-old samples were lysed with RLT Lysis Buffer (Qiagen), and RNA was extracted using Qiagen RNeasy Plus Mini Kit.

For scRNAseq experiments, previously cryopreserved PBMC samples from four food allergic 10 to 18-month-old children (see Supplemental Table 8) were stimulated with 100 μg/ml crude peanut extract, 100 μg/mL β-lactoglobulin, 100 μg/ml casein, 1 mg/ml ovalbumin (OVA), and 1 mg/ml ovomucoid (OVM) overnight prior to flow sorting. PBMCs were collected and then blocked for 10 minutes at 4°C with Human TruStain FcX Fc Receptor Blocking Solution (BioLegend) followed by immediate staining with CD4 and CRTH2 antibodies (Supplemental Table 6) and flow sorting of CD4⁺CRTH2⁺ cells. Information regarding the number of sorted events per sample is provided in Supplemental Table 5.

RNAseq and Analysis

RNA concentration was determined with the NanoDrop 1000 spectrophotometer (NanoDrop, Wilmington, DE) and RNA quality assessed with the Agilent Bioanalyzer 2100 (Agilent, Santa Clara, CA). 1 ng of total RNA was pre-amplified with the SMARTer Ultra Low Input kit v4 (Clontech, Mountain View, CA) per manufacturer’s recommendations. 150 pg of cDNA was used to generate Illumina compatible sequencing libraries with the NexteraXT library preparation kit (Illumina, San Diego, CA) per manufacturer’s protocols. The amplified libraries were hybridized to the Illumina flow cell and sequenced using the NextSeq 550 sequencer (Illumina, San Diego, CA). Single end reads of 75nt were generated for each sample. Sequence quality was adequate in all samples (Supplemental Report 1). Differential expression analysis was performed using DESeq2–1.28.1 with a P-value threshold of 0.05 within R version 4.0.2 (https://www.R-project.org/). Heatmaps were generated using the pheatmap package were given the rLog transformed expression values. Gene ontology analyses were performed using the EnrichR package.

scRNAseq and Analysis

Cellular suspensions were loaded on a Chromium controller (10x Genomics, Pleasanton, CA, USA) to generate single-cell Gel Bead-in-Emulsions (GEMs) and scRNAseq libraries were prepared using Chromium Single-Cell 5′ v2 Library & Gel Bead Kit (PN-1000265, PN-1000264, 10x Genomics). The 5’ gene expression libraries were sequenced following 10X Genomics read length and sequencing depth guidelines on the NovaSeq6000 sequencer (Illumina). Samples were demultiplexed using ‘cellranger mkfastq’ within 10X provided cellranger-7.0.1 using default parameters. For each sample, GEX libraries were aligned to 10X provided reference refdata-gex-hg38–2020-A using ‘cellranger mulit’ within cellranger-7.0.1 using default parameters. Seurat (v4.2.0) was used to aggregate scRNAseq data from all four samples and data was then filtered based on number unique genes and mitochondrial content followed by log normalization and expression scaling. Uniform Manifold Approximation and Projection (UMAP) was used to visualize data.

Culture Supernatant Cytokine Quantification

Culture supernatant levels of GM-CSF, IFNγ, IL-1β, IL-2, IL-4, IL-5, IL-6, IL-7, IL-8, IL-10, IL-12 (p70), IL-13, and TNF-α from stimulation experiments with samples from one-year-old subjects were assessed using the MILLIPLEX MAP Human High Sensitivity T Cell Panel Premixed 13-plex kit (Millipore) according to manufacturer’s protocol. Plate was read using a Luminex 200 instrument (PerkinElmer) and results were analyzed using xPONENT software.

For samples from 2-year-old subjects, culture supernatant IL-2, IL-4, IL-5, IL-6, IL-9, IL-10, IL-13, IL-17A, IL-17F, IL-22, IFN-γ and TNF-α were measured using the LEGENDplex^™ Human Th Cytokine Panel 12-plex kit (BioLegend) according to the manufacturer’s protocol. Samples were read using an Aurora Full Spectrum Flow Cytometer (Cytek Biosciences) with 4 lasers (UV, Violet, Blue, and Red). Quantification of cytokines was done using LEGENDplex^™ data analysis software suite (BioLegend).

Plasma Cytokine Quantification

Plasma supernatant levels of GM-CSF, IFNγ, IL-1β, IL-2, IL-4, IL-5, IL-6, IL-7, IL-8, IL-10, IL-12 (p70), IL-13, and TNF-α from one-year-old infants were assessed using the MILLIPLEX MAP Human High Sensitivity T Cell Panel Premixed 13-plex kit (Millipore) according to manufacturer’s protocol. Plate was read using a Luminex 200 instrument (PerkinElmer) and results were analyzed using xPONENT software.

Statistical Analyses

T cell subtypes were compared between lifestyles and disease outcomes using Mann-Whitney U tests followed by Benjamini-Hochberg multiple comparisons correction and between time points using a one-way analysis of variance (ANOVA). For SWIFT clustering data, cluster sizes were compared between OOM and Rochester using a two-sided unpaired t-test followed by Benjamini–Hochberg correction for multiple comparisons. Cytokine levels were compared using a paired ratio t-test. All statistical analyses other than SWIFT were performed in R using RStudio or GraphPad Prism.

RESULTS

Treg cell subsets associate with OOM lifestyle and non-atopic outcomes

Total CD4⁺ T cells and memory CD4⁺ T cells increased from birth to 6 months of age, thereafter remaining steady (Figure 1B), as did both total and memory Tregs (Figure 1C). When analyzed based on lifestyle groups, there were no differences in total and memory CD4⁺ T cells between lifestyles (Figure 1D). Although there were no differences in total Tregs, interestingly, at 12 months of age, urban infants possessed a higher percentage of memory Tregs in comparison to the OOM (Figure 1E).

Memory Tregs that co-express integrin β7 and CD49d (integrin α4), markers of gut-homing (gating strategy shown in Supplemental Figure 1A), were elevated in OOM compared to urban infants at 12 months of age (Figure 2A). Furthermore, memory Tregs possessing the T cell immunoreceptor with Ig and ITIM domains (TIGIT) were higher in OOM in comparison to urban lifestyle at 6 months of age and trending higher at 12 months of age (Figure 2A). Other subpopulations of Tregs were comparable between OOM and urban lifestyle (Supplemental Figure 2, also see Supplemental Figure 8 and 15 for gating). We also found a similar increase in TIGIT⁺ memory Tregs in 6-month-old infants who did not develop AD and/or FA compared to those who did (Figure 2B), accompanied by an increase at 12 months in other Treg markers such as programmed cell death protein-1 (PD-1) and CD38, which have been considered to indicate a suppressive profile²⁴, as well as C-lectin type receptor CD161. However, the identified Treg subsets did not differ in absolute count (Supplemental Figure 3). Longitudinally, the frequency of TIGIT⁺ and PD-1⁺ memory Tregs decreased with age from birth to 6 months of age (p<0.0001) and from 6 to 12 months of age (p<0.0001), respectively (Fig 2B, statistics not displayed). CD38⁺ memory Tregs, however, increased from birth to 6 months of age (p<0.0001), but decreased from 6 to 12 months of age (p=0.0488).

Figure 2: Treg cell subsets and Th2A populations differ between lifestyles, and according to atopic dermatitis, food allergy and sensitization outcomes.

(A) Percentage of memory Treg subsets over time split between OOM and urban (CB n = 20 OOM + 20 urban, 6m n = 20 OOM + 20 urban 12m n = 35 OOM + 34 urban). (B) Percentage of memory Treg subsets over time split between no disease and atopic dermatitis (AD) and/or food allergy (FA) diagnosis within the first year of life (CB n = 30 no disease + n = 10 AD and/or FA, 6m n = 30 no disease + n = 10 AD and/or FA, 12m n = 56 no disease + n = 13 AD and/or FA). (C) Percentage of Th2A cells based on cohorts, AD and/or FA outcomes, and sensitization outcomes. Th2A cells defined as CD3⁺CD4⁺CRTH2⁺CD161⁺CD49d⁺CD45RO⁺CD27⁻ (CB n = 30 not sensitized + n = 10 sensitized, 6m n = 31 not sensitized + n = 9 sensitized, 12m n = 56 not sensitized + n = 13 sensitized; n for lifestyle and AD and/or FA outcomes same as panels A and B). Box indicates median and interquartile range. Points that fall outside of whiskers are outliers determined by Tukey’s test. Mann-Whitney U-test was used to assess statistical significance between groups. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. OOM, Old Order Mennonite. CB, cord blood. AD, atopic dermatitis. FA, food allergy.

Th2A cells associate with urban lifestyle and sensitization

Th2A cells are a CD4⁺ memory, allergen-specific, pathogenic effector Th2 subset that has increased production of type 2 cytokines in comparison to conventional Th2 cells^5,25. Th2A cells were defined as CD3⁺CD4⁺CRTH2⁺CD161⁺CD49d⁺CD45RO⁺CD27⁻ as previously described⁵ (gating strategy shown in Supplemental Figure 1B). We found a trending increase of Th2A cells in urban lifestyle compared to OOM infants and a significant increase of Th2A cells in IgE-sensitized (a positive skin test or specific IgE to food and/or aeroallergens, in the presence or absence of clinical symptoms) infants compared to non-sensitized, both at 12 months of age (Figure 2C). However, there was no association between AD and/or FA outcomes within the first 2 years of life and Th2A cells (Figure 2C). Additionally, other subpopulations of memory CD4⁺ T cells were comparable (Supplemental Figure 4).

SWIFT clustering analysis reveals CD25⁺CD127⁺CD161⁻CCR4⁺CRTH2⁺ effector memory Th2 cells elevated in urban infants

We performed an unbiased clustering analysis utilizing the SWIFT method²² to further identify cell populations that differ between our cohorts. This method provides a high-resolution cluster map of the samples, resulting in the identification of large numbers of clusters that allows small but significant sub-populations to be identified^21,26. Using an unpaired t-test and a false discovery rate (FDR) of 0.1, we identified five clusters (58, 59, 241, 370, and 535) at 12 months of age that were significantly larger in the urban infants (Figure 3, A and D). Two of these clusters (59 and 241) were also significantly larger in urban infants at 6 months of age. Figure 3B shows a z-score heat map characterizing each of the five clusters based on their surface and intracellular marker expression profile. Interestingly, four of the clusters (59, 241, 370, and 535) were found to co-express CD25 and CD127, indicative of recent cellular activation and maintenance of memory, respectively. Their expression patterns shared several other similarities (Figure 3, B and C). Elevated expression of CRTH2, CCR4, and GATA-3 suggest that this population possesses a Th2-like phenotype. Additionally, high expression of CD45RO and low expression of CCR7 are indicative of an effector memory population. They also express CD49d, i.e. integrin α4 chain which constitutes one of the chains of that form α4β1 and α4β7 integrins that mediate migration of T cells to mucosal sites. One of the CD25⁺CD127⁺ Th2 clusters, cluster 241, identified via SWIFT, had elevated expression levels of CD27, which is a marker of recent T cell receptor (TCR) activation and switch to transitional memory cells^27,28. Analysis of these clusters in each sample between OOM and urban infants found that clusters 59 and 241 predominated the four Th2 clusters and the four Th2 clusters tended to increase with age (Figure 3D). In summary, these four clusters represent a unique CD25⁺CD127⁺CD161⁻CCR4⁺CRTH2⁺ effector memory Th2-like population, henceforth termed Th2B cells, which is elevated in the urban infants at high risk of allergic diseases. As opposed to the Th2A cells described by Wambre et al.⁵, these cells have low expression of CD161.

Figure 3: SWIFT clustering analysis identifies CD25⁺CD127⁺CD161⁻CD49⁺CCR4⁺CRTH2⁺ effector memory Th2 populations elevated in urban infants.

(A) Volcano plot identifying clusters based on false discovery rate (FDR) of 0.1 that differed in size between OOM and Rochester at birth (green), 6 months old (red), and 12 months old (black). (B) z-score heatmap of markers for each cluster that differed between OOM and urban infants. (C) Dot plots and histograms from a combination of ten randomly selected combined 12-month-old samples for each cluster showing expression of markers designated in x and y axes. (D) Histogram showing individual infant samples and their respective cluster sizes for clusters 59, 241, 370, and 535 for each time point between OOM and urban infants. OOM, Old Order Mennonite.

The fifth cluster, cluster 58, which is elevated at 12 months of age in urban infants, is characterized by high expression of CD45RO, CD25, and Foxp3, and low expression of CD127 and CCR7 (Figure 3, B and C). This is indicative of an effector memory Treg population, which is further reinforced by high expression of the co-inhibitory molecules cytotoxic T-lymphocyte associated protein 4 (CTLA-4), TIGIT, and PD-1. This is consistent with our findings above using traditional gating (Figure 1D).

Compared to the five clusters elevated in urban infants, only one cluster, 439, was significantly elevated in OOM at 6 months of age. However, the size of this cluster in most samples was less than 100 cells per one million CD4⁺ cells. Cluster 439 is characterized by increased expression of CD45RO, CXCR5, PD-1, ICOS, Foxp3, CXCR3, T-bet, and ROR-γt, (Figure 3, B and C; Supplemental Figure 6) suggesting that it represents a Th1-biased memory follicular regulatory T cell (Tfr) population, which has not been previously characterized in the literature.

Th2B cells are associated with allergic sensitization

Graphing the trajectory of the Th2B clusters 59, 241, 370, and 535 as well as effector memory Treg cluster 58 shows them to be elevated in urban infants at 12 months of age (Figure 4A). Secondly, clusters 59, 241, and 370 were elevated in urban infants at 6 months of age (Figure 4A). Furthermore, infants who did develop AD and/or FA, compared to those who did not, exhibited an increased size of clusters 59 and 241, but no difference in size of clusters 370 and 535 at 6 months (Figure 4B). Likewise, clusters 59 and 241 were significantly elevated in IgE-sensitized individuals at 6 months of age (Figure 4C).

Figure 4: Th2B cells are an early marker for infants at risk for allergic disease.

Comparisons of cluster size (number of cells per 1×10⁶ CD4⁺ cells) of Th2B clusters and cluster 58 identified by SWIFT, between (A) OOM and urban (CB n = 20 OOM + 20 urban, 6m n = 20 OOM + 20 urban, 12m n = 35 OOM + 34 urban), (B) infants who did and did not develop AD and/or FA during the first year of life (CB n = 30 no disease + n = 10 AD and/or FA, 6m n = 30 no disease + n = 10 AD and/or FA, 12m n = 56 no disease + n = 13 AD and/or FA), and (C) infants who did or did not become sensitized within the first year of life (CB n = 30 not sensitized + n = 10 sensitized, 6m n = 31 not sensitized + n = 9 sensitized, 12m n = 56 not sensitized + n = 13 sensitized). (D) Traditionally gated CD25⁺CD127⁺CD161⁻CRTH2⁺CCR4⁺ cells based off lifestyle, AD and/or FA outcomes, and sensitization outcomes. See above for details on n per group. Box indicates median and interquartile range. Points that fall outside of whiskers are outliers determined by Tukey’s test. Mann-Whitney U-test was used to determine statistical significance between groups. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. OOM, Old Order Mennonite. CB, cord blood. AD, atopic dermatitis. FA, food allergy.

To further confirm SWIFT findings, we performed traditional gating for the Th2B cells (See Supplemental Figure 1C for gating strategy, and for full stain and FMO controls, and Supplemental Table 9 for cell counts of Th2B and other subpopulations). As expected, traditional gating confirmed SWIFT cluster results demonstrating an increase in Th2B cells in urban lifestyle at both 6 months and 12 months of age (Figure 4D). There was also an increase in these cells in those with allergic outcomes at 6 months of age as well as in those with IgE-sensitization at 6 and 12 months of age (Figure 4D). These significant patterns were also observed in the absolute count of Th2B cells (Supplemental Figure 9). To isolate the effects of allergic disease outcomes on Th2B cells, we compared them between allergic and non-allergic infants in the urban population only. There was a significant increase in traditionally gated Th2B cells and trending increase in cluster 59 in 6-month-old infants (Supplemental Figure 10). This suggests that the elevation of Th2B cells and cluster 59 in infants who develop AD and/or FA are not due to confounding lifestyle differences. The lack of significant differences in the other Th2B clusters may be due to low sample size or the notion that lifestyle is the main driver of cluster 241, 370, and 535 separation between groups rather than allergic disease outcomes (Supplemental Figure 7). However, linear regression analysis reveals that the elevation of traditionally gated Th2B cells as well as clusters 59 and 241 in infants who develop AD and/or FA remains significant even when correcting for lifestyle (p<0.05), although this is not true for clusters 370 and 535.

CD4⁺CD25⁺CD127⁺ cells demonstrate a phenotypically distinct Th2-skewed gene expression profile

To further understand the gene expression profile of this CD4⁺CD25⁺CD127⁺ cell population, RNAseq was performed on isolated CD4⁺CD25⁺CD127⁺ cells treated with PMA/ionomycin from 3 atopic infants and compared to stimulated autologous isolated CD4⁺CD25⁻ effector cells. Compared to stimulated, conventional CD4⁺CD25⁻ effector Th cells, a total of 326 genes were differentially expressed between the two populations, with 202 genes being upregulated and 124 genes being downregulated in CD4⁺CD25⁺CD127⁺ cells compared to CD4⁺CD25⁻ cells (Figure 5A, Supplemental Table 10). Because we are mostly focused on CD4⁺ T cell subsets and association with disease in this study, we chose to focus on GO/KEGG database pathways identified via Enrichr analysis (Supplemental Table 11) that fulfill these criteria. More specifically, we focused on pathways associated with general T cell, Th1, Th2, Th17, and Treg responses and allergic and autoimmune diseases (Figure 5B). According to the KEGG database, CD4⁺CD25⁺CD127⁺ cells upregulated several genes in pathways associated with T cell receptor signaling, cytokine-cytokine receptor interaction, Th1/Th2 differentiation, and asthma (Figure 5B). Such genes include IL4, IL5, IL13, CSF2 (encodes GM-CSF), and GATA3 (Figure 5B). At the same time, CD4⁺CD25⁺CD127⁺ cells downregulated genes associated with several autoimmune disorders identified via the KEGG database including Type I diabetes mellitus, autoimmune thyroid disease, inflammatory bowel disease, and rheumatoid arthritis, as well as with Treg differentiation identified via the GO database (Supplemental Table 12). Such genes include human leukocyte antigen genes as well as FOXP3, CTLA4, and LILRB2 (Figure 5B). In addition, compared to CD4⁺CD25⁻ cells, CD4⁺CD25⁺CD127⁺ cells secreted more IL-4, IL-5, IL-13, and GM-CSF, which is consistent with gene expression analysis (Figure 5C).

Figure 5: CD25⁺CD127⁺ T cell population associated with urban lifestyle and development of sensitization possesses a functional Th2 phenotype.

(A) Volcano plot showing differentially expressed genes and (B) heatmap showing differentially expressed genes clustered by associated pathway between isolated and stimulated CD4⁺CD25⁺CD127⁺ or autologous CD4⁺CD25⁻ cells from three one-year-old urban infant PBMC samples. Samples 1 (subject ID 219) and 2 (subject ID 240) were derived from egg-sensitized infants with AD, whereas sample 3 (subject ID 271) was derived from an egg-sensitized infant with no AD (Supplemental Table 1). (C) Culture supernatant concentration of IL-4, IL-5, IL-13, and GM-CSF from stimulated sorted cells normalized to cell count. Each point represents one sample. Paired ratio t-test was used to determine statistical significance between groups for cytokine secretion. (D) Mann-Whitney U-test was used to calculate differences in levels of IL-8 and TNF-α between OOM and urban. Fisher’s exact test was used to calculate differences in the proportion of detectable IL-10 or IFN-γ between OOM and urban. Each point represents one infant. Data are presented as geometric mean and ± 95% confidence interval. *P < 0.05, **P < 0.01. OOM, Old Order Mennonite.

Systemic anti-inflammatory, Th1, and pro-inflammatory cytokines are elevated in OOM infants

At 12 months of age, OOM infants’ plasma had elevated levels of pro-inflammatory cytokines IL-8 and TNF-α. Additionally, OOM samples had an increased proportion of detectable IL-10 and IFN-γ compared to urban infants (Figure 5D). There were no differences in systemic Th2 cytokines (Supplemental Figure 12). Most of the cytokines were undetectable at 6 months of age (Supplemental Table 13).

Differential cytokine profile of Th2A and Th2B cells

To better understand the relationship between Th2B and Th2A cells, we sorted both populations along with conventional Th1 and Th2 cells from cryopreserved PBMC samples from 9 two-year-old infants. Unlike our PBMC samples from one-year-olds, the sample volumes from two-year-olds are larger and allow for isolation to the necessary resolution to specifically sort Th2A and Th2B cells with adequate cell numbers. Th2A cells were sorted as CD4⁺CRTH2⁺CD161⁺ cells, Th2B cells as CD4⁺CRTH2⁺CD161⁻CD25⁺CD127⁺, Th2 as CD4⁺CRTH2⁺CD161⁻CD25^+/−CD127^+/−, and Th1 as CD4⁺CXCR3⁺. Cells were stimulated and cultured overnight with PMA/ionomycin, and secreted cytokines were quantified. Enriched Th2, Th2A and Th2B cells, upon stimulation, had comparable production of Th2 cytokines, which was significantly higher than that of Th1 cells, whereas Th1 cells produced significantly more IFNγ (Figure 6A). Enriched Th2A and Th2B cells produced more IL-9, IFNγ, IL-17, IL-22, IL-6, and IL-10 compared to Th2 cells. Compared to Th2A cells, Th2B cells produced more IL-2 and TNFα (Figure 6A). Supplemental Figure 13 shows the changes upon stimulation from baseline.

Figure 6. Functional and transcriptional characterization of Th2B cells.

(A) Cytokine concentrations in culture supernatants of FACS sorted Th1 (CD4⁺CXCR3⁺), Th2 (CD4⁺CDTH2⁺CD161⁻CD25^+/−CD127^+/−), Th2A (CD4⁺CRTH2⁺CD161⁺), and Th2B (CD4⁺CRTH2⁺CD161⁻CD25⁺CD127⁺) cells from 2-year-old children. From cryopreserved PBMCs, sorted cells were cultured overnight in the presence of PMA/ionomycin. Statistical analysis to compare cytokine concentrations between cell populations were made using Kruskal-Wallis and Dunn’s test with Benjamini-Hochberg correction. Colored dots represent the summary of three experiments. (B-C) Single cell RNA sequencing (scRNAseq) on CD4⁺CRTH2⁺ Th2 cells isolated from four 10- to 18-month-old food allergic subjects. PBMCs were stimulated with food allergens, including crude peanut extract, milk, and egg antigens overnight prior to flow sorting of CD4⁺CRTH2⁺ Th2 cells. scRNAseq was performed followed by sample aggregation and subsequent analysis using Uniform Manifold Approximation and Projection (UMAP). Th2A cells (CD25⁻CD127⁺CD161⁺) and Th2B (CD25⁺CD127⁺CD161⁻) cells were detectable in all four samples and clustered separately indicating that they are distinct populations. (D) Upon differential expression analysis, FOXP3 and the cytokine inducible SH2-containing protein gene (CISH) were upregulated in Th2B cells relative to Th2A cells.

Elevated expression of FOXP3 and CISH in Th2B cells distinguish them from Th2A cells

In order to better differentiate Th2A and Th2B cells, we performed single cell RNA sequencing (scRNAseq) on CD4⁺CRTH2⁺ Th2 cells isolated from four 10- to 18-months-old food allergic subjects (Supplemental Table 8). PBMCs were stimulated with food allergens, including crude peanut extract, milk, and egg antigens overnight prior to flow sorting of CD4⁺CRTH2⁺ Th2 cells. scRNAseq was performed on all four sorted samples followed by sample aggregation and subsequent analysis using Uniform Manifold Approximation and Projection (UMAP). Th2A cells (CD25⁻CD127⁺CD161⁺) and Th2B (CD25⁺CD127⁺CD161⁻) cells were detectable in all four samples and clustered separately (Figure 6, B and C; Supplemental Figure 14), indicating that they are distinct populations. Of the differentially expressed genes (DEGs), FOXP3 and the cytokine inducible SH2-containing protein gene (CISH) were upregulated in Th2B cells relative to Th2A cells (Figure 6D). Other DEGs where the differences were more subtle are listed in Supplementary Table 14.

DISCUSSION

Investigation of early biomarkers for allergic diseases in infancy is limited due to a lack of immunophenotyping in allergy cohorts, especially within the first year of life— a niche that our study fulfills. As part of a larger cohort study examining the effects of farming lifestyle on infant immune development, we utilized a comprehensive T cell panel, full spectrum flow cytometry and SWIFT, an unbiased, high-dimensional clustering analysis able to detect rare cell populations. This allowed identification of an effector memory Th2 population that has a unique CD25⁺CD127⁺CD161⁻ CD49d⁺CCR4⁺CRTH2⁺ phenotype. This population was elevated already at 6 months of age in those infants later developing AD and/or FA, suggesting that this unique cell population is an early marker of allergic disease. Additionally, we identified several Treg subpopulations associated with farming lifestyle as well as with protection against AD and/or FA outcomes and sensitization— with notable populations associated with gut-homing and suppressive phenotypes.

The CD25⁺CD127⁺CD161⁻CCR4⁺CRTH2⁺ cells discovered here co-expresses Th2 transcription factor GATA-3, chemokine receptor CCR4, and prostaglandin receptor CRTH2, indicating that these cells possess a Th2 phenotype. Additionally, high expression of CD45RO and low expression of CCR7 are indicative of an effector memory population. They also express CD49d, i.e. α4 chain, which constitutes one of the chains that form α4β1 and α4β7 integrins that mediate migration of T cells to mucosal sites. In contrast to pathogenic effector Th2A cells which have been demonstrated in older individuals with established peanut allergy⁵, the cell population identified here expresses CD25 and has a low expression of CD161, therefore termed here Th2B cells. CD161 is thought to be a possible co-signaling surface molecule expressed in various memory T pools but its function is still being elucidated in CD4+ T cell populations, with both co-stimulatory and inhibitory functions being described.²⁹ Th2B cells co-expressed CD25 and CD127, which have been associated with recent cellular activation and maintenance of memory, respectively. Few studies^30,31 have described T cells that co-express CD25, a component of the IL-2 receptor, and CD127, a subunit of the IL-7 receptor. One such study examined immune markers of type 1 diabetes (T1D) disease progression and associated elevated CD25⁺CD127^hi cells with slower T1D progression³¹. Lastly, compared to the other CD25⁺CD127⁺ Th2 clusters, cluster 241, identified via SWIFT, had elevated expression levels of CD27, which is a marker of recent T cell receptor (TCR) activation and switch to transitional memory cells^27,28.

Th2B cells were significantly elevated already at 6 months of age, whereas the Th2A population was elevated only at 12 months in urban infants and infants who became IgE-sensitized. This sequence of observations begs the question of whether Th2B cells act as a precursor for Th2A cells. Th2A and Th2B cells, indeed, may possess a similar functional phenotype as they exhibited a largely overlapping cytokine secretion profile upon stimulation, characterized by a high expression and production of Th2 cytokines IL-4, IL-5, IL-13 similar to conventional Th2 cells. Compared to Th2 cells, cell populations enriched for Th2A and Th2B cells showed an elevated production of IL-9 as previously described⁵, but also IL-2, IL-6, TNFα, Il-10, and IL-17 and IL22, which may reflect a younger age group assessed. Compared to Th2A cells, Th2B cells produced more IL-2 and TNFα, indicative of a more activated and proinflammatory phenotype.

scRNAseq analysis revealed that Th2A and Th2B cells form separate clusters and, in comparison to Th2A cells, Th2B cells upregulate FOXP3 and the cytokine inducible SH2-containing protein gene CISH. This suggests that these two cell populations are distinct from another. Although Foxp3 is traditionally thought of as a marker of Tregs, it can also be transiently expressed following recent cellular activation³². Prior research has demonstrated that ex vivo stimulation of CD4⁺ T cells from egg-allergic children with hen’s egg white resulted in upregulation of CISH³³, particularly in CD25⁺ T cells and IL-4⁺ T cells. Together, upregulation of FOXP3 and CISH in the Th2B cells isolated from food allergic infants in response to stimulation with egg, milk, and peanut allergens further suggests that Th2B cells are allergen-activated. Recently, increased methylation of CISH locus on chromosome 3 was associated with decreased gene expression and less numerous allergic sensitizations³⁴. This locus has not been associated with asthma or allergic diseases in genome-wide association studies, although CISH expression is induced by IL-13 and has been implicated in eosinophilic inflammation. The authors concluded that increased expression of CISH is associated with increased sensitization to allergens and that the regulation of CISH expression may be epigenetically mediated.

The suppressive function of Tregs is quintessential to the development of immunological tolerance and prevention of allergic disease³⁵. We identified several memory Treg populations that are positively associated with farming lifestyle as well as protection against allergic disease and/or sensitization outcomes. At 6 months of age, OOM and individuals who did not develop AD and/or FA had increased TIGIT⁺ memory Tregs. Additionally, at 12 months of age, infants who did not develop AD and/or FA had an elevated proportion of memory Tregs bearing suppressive markers PD-1, ICOS, and CD38. TIGIT, PD-1, and ICOS are co-inhibitory molecules on Tregs that promote their suppressor function^36–38. The finding of CD38 expression on farm lifestyle associated Tregs is a novel one. Unlike TIGIT, PD-1, and ICOS, CD38 is transmembrane glycoprotein that is not a co-inhibitory molecule. However, like the co-inhibitory molecules, CD38 is a marker for highly suppressive Tregs^24,39. Interestingly, we also identified a memory Treg population that is elevated in the urban cohort (but not in those with atopic outcomes), and observed in our SWIFT analysis as cluster 58. Our data are in fact consistent with the Swedish FARMFLORA birth cohort report of elevated Tregs in a non-farming population⁴⁰. Further functional studies on isolated memory Tregs from these populations are warranted. Finally, to establish oral tolerance and prevent the development of food allergy, it is necessary that Tregs home to the gut where they undergo expansion⁴¹. On that note, we found that OOM had increased gut homing memory Tregs⁴² at one year of age that may relate to protection against the development of food allergy shown by us before^12,17,43. However, our study does not have enough food allergy outcomes for direct correlation.

SWIFT analysis also identified two other clusters, clusters 58 and 439, that were distinct from the CD25⁺CD127⁺ Th2 clusters. Cluster 58, which was elevated in 12-month-old urban infants, represented an effector memory Treg population as defined by its high expression of CD45RO, CD25, Foxp3, CTLA-4, TIGIT, and PD-1 and low expression of CD127 and CCR7. However, functional assays on isolated memory Tregs from our low and high risk for allergy populations are warranted. Cluster 439 was elevated in 6-month-old OOM infants. Its expression of CD45RO, CXCR5, PD-1, ICOS, and Foxp3 is indicative of memory follicular regulatory T (Tfr) cells. Additionally, elevated expression of CXCR3, T-bet, and ROR-γt in this cluster may suggest Th1-biased Tfr cells, which has not been previously characterized in the literature.

Due to the paucity of infant allergy cohorts assessing T cell subsets, understanding of early events in allergic sensitization and atopic disease is limited, a niche that our study fulfills. Additional strengths of our study include the use of SWIFT which allows unbiased exploration of rare T cell subsets. Our functional studies using RNAseq and protein production analysis further characterizes the Th2 phenotype and function of our Th2B cells. However, studies confirming our observed findings with a larger sample size and studies regarding allergen-specificity, developmental relationship to Th2A cells, and functional studies on a pure Th2B population are warranted. Because our samples are derived from infants, sample volume and cell frequency remain obstacles for these studies. Further limitation is that in our cohort the number of allergic outcomes is small and therefore analyses comparing markers between atopic and non-atopic infants are limited in their power. Similarly, interrogation of specific lifestyle factors, such as diet, vaccinations, illness, living environment, tobacco use, pets just to name a few, responsible for the differences is not possible in the study of this size.

In conclusion, the use of comprehensive immunophenotyping combined with a unique clustering method led us to discover a novel Th2 subset implicated in urban lifestyle and increased risk for development of atopic disease as well as several Treg subsets associated with anti-allergy protection. We believe these data to be foundational findings in the field of early biomarkers for atopic disease, which will allow for the identification of infants at risk of developing atopic disease and provide targets for preventative therapies.

Abbreviations:

CCR4

C-C chemokine receptor type 4

CRTH2

chemoattractant receptor-homologous molecule on T helper type 2 cells

Foxp3

forkhead box P3

scRNAseq

single cell RNA sequencing

Th

T helper cell