Research ArticleAsthma

Distinct immune phenotypes in infants developing asthma during childhood

See allHide authors and affiliations

Science Translational Medicine  05 Feb 2020:
Vol. 12, Issue 529, eaaw0258
DOI: 10.1126/scitranslmed.aaw0258

Infant immunity and asthma

Asthma is traditionally thought of as a type 2 allergic disease, but its presentation and pathogenesis are quite heterogenous. To better understand early immune responses that are associated with asthma, Thysen et al. leveraged samples from a longitudinal cohort of hundreds of children. Blood cells collected at 18 months of age were stimulated in vitro with various ligands, and different immune phenotypes were related to clinical asthma outcomes. The investigators revealed sex differences, an influence of microbial exposure, and immune traits associated with transient or persistent asthma. This systems approach provides valuable information on early life immune profiles and how they relate to asthma development.


Early exposure to environmental triggers may elicit trajectories to chronic inflammatory disease through deregulated immune responses. To address relations between early immune competence and development of childhood asthma, we performed functional immune profiling of 186 parameters in blood of 541 18-month-old infants and examined links between their response phenotype and development of transient or persistent disease at 6 years of age. An abnormal neutrophil-linked antiviral response was associated with increased risk of transient asthma. Children who exhibited persistent asthma at year 6 showed enhanced interleukin-5 (IL-5) and IL-13 production in stimulated T cells at 18 months of age, which was associated with early life bacterial colonization of the airways. These findings highlight the early appearance of distinct immune characteristics in infants developing different asthma endotypes during childhood.


Appropriately regulated immune function in infancy is not only pivotal for successful combat of the first microbial exposures but also essential in preventing later disease development, as inadequately regulated immune responses may lead to poor microbial clearance, resulting in chronic inflammatory programs that may eventually provoke specific disease trajectories. It is generally accepted that newborns are more vulnerable to viral and bacterial infections and show attenuated responses to most vaccines compared to adults (1, 2). From birth, the immune system is characterized by enhanced type 2 immunity at the expense of an incomplete type 1 and type 17 responsiveness (38). Type 1 and type 17 responses emerge after birth and throughout early life, as the infant is exposed to environmental microbes (9, 10). An adequately regulated innate and adaptive immune defense depends on an appropriate response to the encountered microbe, where intracellular microbes (viruses and bacteria) require a type 1 response and extracellular microbes require a type 17 (bacteria and fungi) or type 2 (parasites) response (11). Because certain viral and bacterial infections associate with increased risk of specific diseases during childhood (12), it is important to understand the variation among infant immune responses upon viral and bacterial stimulation. This may further explain the nature of the deregulated immune responses involved in development of common diseases, including transient and persistent phenotypes of childhood asthma (13, 14), which affect up to 30% of preschool children of the industrialized world (15).

To address whether a deregulated immune response to generic microbial exposures in infancy is associated with later disease development, we here performed a comprehensive characterization of the functional immune response to viral- and bacterial-derived ligands, as well as selective T cell stimulators, in 18-month-old infants. We then examined associations between distinct immune response types and asthma development, including transient and persistent asthma, until 6 years of age. Eighteen months of age was selected as the examination point in this study, as it represents a time point where the T cell compartment has undergone extensive in vivo maturation and expansion dependent on the individual environmental exposures in early life (1). We here show that the immune response status at 18 months is useful as a sensor for previous exposures encoded in the T cell compartment of relevance for development of persistent asthma, as well as for monitoring immune phenotypic responses in the innate immune compartment involved in development of transient asthma during childhood.


Circulating immune cell frequencies and response profiles in whole blood of infants

The study is part of the ongoing longitudinal mother-child cohort, COpenhagen Prospective Studies of Asthma in Childhood 2010 (COPSAC2010) (16), where mother-child pairs were enrolled at pregnancy week 24. Demographic data of participating infants are detailed in table S1. The children were clinically monitored from birth and onward, and at 18 months of age, peripheral blood was sampled from 644 of 700 infants (Fig. 1A and fig. S1). Flow cytometry analyses of freshly drawn whole blood were performed in 552 infants and ex vivo stimulation of blood in 567 infants, resulting in a total of 541 infants with combined data. Relative frequencies of 18 immune cell subsets were detected by flow cytometry (Fig. 1B; gating strategy in fig. S2). To study the immune reactivity and diversity of innate immune cells and T cells, whole blood was ex vivo stimulated for 24 hours with seven generic microbial-derived ligands that target different immune receptors and cell types [Toll-like receptor (TLR) 3/4/7/8, NOD2, NLRP3, αβ T cells, or γδ T cells], followed by quantification of 21 secreted cytokines, each related to different immune response types; type 1, type 2, type 17, type 22, and regulatory/regulatory T cell (Treg) responses (colored boxes in Fig. 1C). Type 1–related responses are involved in clearance of intracellular microbes, such as viruses and intracellular bacteria, which activate type I interferon (IFN) from plasmacytoid dendritic cells (pDCs) along with interleukin-12p70 (IL-12p70) and IFN-γ production from myeloid DCs (mDCs) and monocytes that, in turn, induce T helper cell 1 (TH1), cytotoxic CD8 T cell responses, and/or natural killer (NK) cell activation, together resulting in killing of infected cells (17). An effective response to extracellular bacteria and fungi requires phagocytosis mediated via a neutrophil-assisted type 17 response (18), including release of IL-1β, IL-6, IL-23, IL-17A, CXCL8, and CCL20, TH17 activation, neutrophil migration and activation, and, to some extent, type 22–associated release of IL-22 and CCL27. Parasites are cleared by a type 2–based immune response, including production of IL-4, IL-5, IL-13, and chemokine mediators such as CCL17 and CCL24. The type 2 response is based on adaptive TH2 cells and innate cells that mediate eosinophilic activation and immunoglobulin E (IgE)–dependent mast cell activity (19). A tolerogenic and resolving immune response type is primarily propagated by paracrine actions of transforming growth factor–β (TGF-β) and IL-10 and suppressive cells such as Tregs (20). IL-18 is here assigned to the type 1 response due to its effect on TH1 and NK cell activation when in concert with other type 1 cytokines, but when produced in the absence of IL-12p70, IL-18 is reported to mediate TH2 progression of naïve T cells (21) as well as IL-4 and IL-13 production by basophils, which favors TH2 polarization (22). Collectively, we measured a total of 186 immune parameters (8 × 21 cytokines and 18 cell types) from each infant, followed by a comprehensive systems-level data analysis of innate and T cell immune responses to identify common denominators of the functional immune reactivity in 18-month-old infants.

Fig. 1 Comprehensive analysis of circulating innate cell and T cell function in 18-month-old infants.

(A) Study outline. Blood for immune analysis was drawn at 18 months of age, and disease development was evaluated throughout the first 6 years of life. (B) Analytical overview. The 18 listed immune cell subsets were detected by multiparametric flow cytometry in freshly collected whole blood (n = 552). (C) In parallel, diluted whole blood was incubated 24 hours with media only (spontaneous release) or the indicated innate and T cell activating ligands, followed by measurement of cytokines and chemokines within harvested cell-free supernatants using electrochemoluminescence-based immunoassays (n = 567). Concentrations of the pretitrated ligands are available in Materials and Methods. The detected cytokines and chemokines belong to five immune response types, here assigned a specific color-code used in subsequent figures of the manuscript for ease of explanation.

Among the immune cells in circulating blood, neutrophils were the most frequent cell type accounting for 31% of leukocytes, followed by CD4 T cells (26%), B cells (16%), CD8 T cells (12%), classical monocytes (4.2%), CD56dim NK cells (2.0%), eosinophils (1.9%), Tregs (1.7%), and γδ T cells (1.7%) (Fig. 2A and table S2). Circulating DC frequencies were below 1% [blood dendritic cell antigen-1 (BDCA-1 DCs) (0.18%), pDCs (0.12%), and BDCA-3 DCs (0.0061%)]. Recently activated T cells were generally low in abundance with activated CD4 T cells accounting for 0.16% and activated CD8 T cells for 0.012% of leukocytes. Nonclassical and intermediate monocytes as well as CD56bright NK cells accounted for about 0.4%, whereas invariant NKT cells were present at 0.019%. Total immune cell counts were determined in 375 of the infants (table S3) and correlated strongly to the relative cell frequencies within each cell subset (fig. S3).

Fig. 2 Circulating immune cell frequencies and response profiles in whole blood of 18-month-old infants.

(A) Relative frequencies of the 18 immune cell subsets in whole blood (n = 552). (B) Concentrations of cytokines and chemokines from 24 hours unstimulated (spontaneous release) whole blood. (C) Polar charts illustrating the mean fold change of cytokine and chemokine release in innate ligand stimulated versus unstimulated whole blood. For stimulation of innate immune cells, blood was stimulated to activate TLR3 [viral double-stranded RNA (dsRNA)], TLR7/8 (viral ssRNA), TLR4 (bacterial LPS), NOD2 (bacterial peptidoglycan), and NLRP3 (alum with low-dose LPS). For NLRP3, the mean fold change was calculated on the basis of addition or not of alum to a low-dose LPS control culture (5 ng/ml). For T cell stimulation, blood was stimulated with SEB and HDMAPP to activate αβ T cells and γδ T cells, respectively. (B and C) Color codes correspond to the type-response designations from Fig. 1C. N = 567.

Polar charts were used to visualize the type of similarities and dissimilarities in cytokine response patterns in unstimulated blood (spontaneous release, Fig. 2B) and within the individual microbial-derived ligands stimulating different receptors and thus different cell types (Fig. 2C and table S4). The viral-mimicking ligands recognized by TLR3 and TLR7/8 stimulated release of the type 1–associated cytokines IL-12p70 and IFN-γ, resulting in concentrations 10- to 100-fold higher than with the bacterial-mimicking components to TLR4 and NOD2. TLR7/8 activation led to induction of the type I interferon IFN-β that was also slightly induced by TLR3 stimulation, while staying low with the remaining ligand stimulations. Activation of TLR3 and TLR4 led to similar patterns of released type 17–related cytokines, whereas NOD2 activation stimulated sevenfold higher CXCL8 and twofold higher CCL20 production compared to TLR3, TLR7/8, and TLR4 activation. Activation of TLR3, TLR7/8, TLR4, and NOD2 led to similar concentrations of type 2, type 22, and regulatory cytokine release. Alum, activating NLRP3, induced the expected release of IL-1β and IL-18 after stimulation in concert with low amounts of lipopolysaccharide (LPS), a known activator of the NLRP3 inflammasome (23), as compared to the response from blood immune cells stimulated with the viral- and bacterial-derived ligands. Likewise, a principal components analysis (PCA) of the innate ligand responses illustrated greater similarity between cytokine profiles from the microbial stimulations TLR3, TLR7/8, TLR4, and NOD2 as compared to NLRP3 in principal component 1 (PC1) (fig. S4A). PC2 separated the viral-mimicking responses (TLR7/8 and TLR3) from the bacterial-mimicking ligands (TLR4 and NOD2). However, TLR3 and TLR4 profiles, which share the TIR-domain-containing adapter-inducing interferon-β (TRIF)–signaling pathway (24), also partly overlapped. The PCA loading plot signified the type 1–skewed cytokine profile induced by the viral-mimicking TLR3- and TLR7/8-activating ligands and the more type 17–skewed response induced by the bacterial-mimicking TLR4- and NOD2-activating ligands (fig. S4B).

The response of αβ T and γδ T cells after stimulation with distinct polyclonal T cell activators for 24 hours were examined with the aim of identifying the phenotype of circulating T cells previously stimulated in vivo (i.e., activated effector and memory T cells), as de novo activation of naïve T cells would require a longer stimulation period (25). The polyclonal T cell activator SEB (staphylococcal enterotoxin B) was used to stimulate αβ T cells, whereas γδ T cells were activated with the phosphoantigen HDMAPP (1-hydroxy-2-methyl-2-buten-4-yl 4-diphosphate) that specifically stimulates the most abundant γδ T cell subset in peripheral blood, Vγ9Vδ2 (26). We found the major T cell–related cytokines to be released after polyclonal stimulation of αβ T cells, including IFN-γ, IL-4, IL-5, IL-13, IL-31, IL-17A, IL-22, TGF-β, and IL-10 (Fig. 2C and table S4), whereas HDMAPP stimulation of γδ T cells resulted in release of the T cell–related cytokines IFN-γ, IL-5, and IL-17A only (Fig. 2C and table S4).

Sex-dependent differences in innate immune responses at 18 months of age

When next examining sex-dependent differences in the cell frequencies and innate immune responsiveness in the 18-month-old infants, we observed slightly enhanced frequencies of specific cell subsets in girls, including neutrophils, intermediate and nonclassical monocytes, CD56bright NK cells, CD4 T cells, and Tregs, whereas boys had higher BDCA-3 mDC numbers (Fig. 3A). On the other hand, blood-derived immune cells from boys displayed a profoundly enhanced functional response to all innate ligands, with higher type 17–associated cytokine responses, as well as IL-18 and IL-10, as compared to girls (Fig. 3B). Identification of enhanced IFN-β to TLR7/8 stimulation in immune cells from girls is comparable to previous results from women (27, 28).

Fig. 3 Sex-related variations in circulating immune cell numbers and the functional cytokine response after innate ligand stimulation of blood immune cells in 18-month-old infants.

Linear models were applied to compare concentrations of circulating immune cells (A) and cytokine release upon innate ligand stimulation (B) with female sex as the outcome. Positive estimate size (red) corresponds to higher concentrations within girls, and negative estimate size (blue) corresponds to higher concentrations within boys. The statistical significance is reflected in the size of the dot and is reported as false discovery rate (FDR)–adjusted P values. N = 541.

Ligand-dependent associations between immune cell subsets and the functional response pattern

To determine cell-to-cytokine profiles for blood immune cells stimulated with the generic microbial-derived ligands, we next performed pairwise correlation analyses between frequencies of the enumerated 18 cell subsets and the 21 differentially released cytokines across all infants and visualized the associations in covariation heatmaps (Fig. 4). The Spearman rank correlation coefficients (SCCs) of the resulting correlation matrices ranged between −0.5 and 0.5 and revealed several notable cell-to-cytokine relationships when focusing on the most significant (P < 0.01) of these correlations. Monocytes were observed to respond to all five innate stimuli; however, the three monocyte subsets (classical, intermediate, and nonclassical, which are typically reported as one cell type; i.e., monocytes) showed different response types across ligands. Moreover, monocytes were the only cell subsets that correlated with cytokine release after TLR4 and NLRP3 stimulation in all individuals. After viral- and bacterial-mimicking ligand stimulation, the classical monocytes were found to correlate with specific type 1– and type 17–related cytokines, where TLR3 and TLR4 revealed more similar response profiles than TLR7/8 and NOD2. Only classical and intermediate monocytes correlated with IL-18 and IL-1β production after NLRP3 activation and with anti-inflammatory IL-10 upon TLR7/8 and NOD2 stimulation, whereas nonclassical monocytes were associated with IL-18 and IL-1β after TLR3 and TLR4 triggering only. Contrary to monocytes, mDCs and pDCs correlated solely to cytokines induced by the viral-mimicking ligands stimulating TLR3 and TLR7/8, but showed different response profiles. After TLR3 activation, we found strong positive correlations between BDCA-1 DC and the type 1 cytokines IL-12p70 and IFN-γ, indicating that TLR3-induced innate drivers for TH1 differentiation may be initiated by BDCA-1 DC. For TLR7/8, positive correlations were seen between pDC and IFN-β [relying on pDC expression of TLR7 (29)]. TLR7/8 activation also induced strong correlations between the two mDC subsets BDCA-1 and BDCA-3 DC, and IL-23, not identified for other stimuli. Notably, TGF-β and IL-10 were found not to correlate to the mDC subsets upon any stimulation. Neutrophils, like classical monocytes, showed strong correlation to the CCR6-attracting chemokine CCL20 after activation with both the viral- and bacterial-mimicking ligands. For TLR7/8, neutrophils also correlated to CXCL8 and the inflammasome-associated cytokines IL-18 and IL-1β, whereas NOD2 activation induced CXCL8, along with IL-6, from neutrophils. In addition, TGF-β and IL-10 associated with neutrophils upon TLR3, TLR7/8, and NOD2 activation. As expected, we found no associations between any of the five innate ligands and the type 2 mediators, except for a slight positive correlation between IL-5 and eosinophils, and for DCs, after TLR3, TLR7/8, and NOD2 stimulation. Some of the lymphocytes (CD56bright NK cells, B cells, CD4 T cells, and CD8 T cells) were found to correlate to IFN-γ production after innate ligand activation, which may be due to expression of the ligand receptors in lymphocytes (NK cells: TLR3/7/8 and NOD2; B cells: TLR7 and NOD2; T cells: TLR3 and NOD2, as based on Inverse correlations were observed upon innate ligand stimulation between T cell subsets versus type 17 mediators and IL-10.

Fig. 4 Ligand-dependent associations between immune cell subsets and the functional response pattern in blood of 18-month-old infants.

The cell-to-cytokine covariation matrices per activating stimuli from Fig. 2, visualized as SCCs between relative cell frequencies and delta concentrations of released cytokine from stimulated versus unstimulated whole blood. Whole blood was stimulated to activate TLR3 (viral dsRNA), TLR7/8 (viral ssRNA), TLR4 (bacterial LPS), NOD2 (bacterial peptidoglycan), and NLRP3 (alum in concert with low-dose LPS). Unstimulated controls were added media alone. NLRP3 data were based on alum + low LPS stimulation minus the low LPS stimulation control. SCC ranges from −0.5 to 0.5 (legend at the bottom) and are only plotted if P < 0.01. Cytokines are color-coded at the top to match the designations of response types as in Fig. 1C. N = 541.

After short-term (24 hours) polyclonal activation of αβ T cells, the covariation matrix showed CD8 αβ T cells and recently activated CD4 αβ T cells to positively correlate with cytokine release (Fig. 4). Hence, except for IFN-γ, the T cell–related cytokine profile reflected the profile of recently in vivo activated CD4 αβ T cells. These recently activated CD4 T cells correlated with the TH2 cytokines IL-4, IL-5, and IL-13, but not with the TH1 cytokine IFN-γ and the TH17 cytokine IL-17A, suggesting a yet predominant TH2 activation profile in 18-month-old infants. Conversely, the covariation matrix for HDMAPP stimulation showed a correlation between many cytokines and γδ T cells only, justifying that HDMAPP is a pure γδ T cell stimulator (26). Collectively, these cell-to-cytokine covariation matrices illustrated the types of responding blood immune cell subsets beneath the observed overall cytokine profiles and further pinpoint the cellular basis for the different versus overlapping response patterns induced by the microbial ligands.

Large heterogeneity in innate cell response profiles to microbial ligands in 18-month-old infants

While general cell-to-cytokine profiles are relevant to understand overall immune responsiveness in infants, it is also vital to examine underlying response differences between infants, as these might be of relevance for disease trajectories. To examine this, we first focused on identifying whether distinct innate immune response profiles were induced in the 18-month-old infants upon antimicrobial and inflammasome activation of immune cells. We used hierarchical clustering on selected innate ligand-related cytokines (cell-to-cytokine SCC > 0.25) and grouped infants with similar cytokine responses into immune phenotypic clusters [Fig. 5 and fig. S5, clusters 1 to 7 (C1 to C7)]. This subgroup clustering resulted in identification of five to seven response phenotypes, depending on the stimulating innate ligand (TLR3: six clusters; TLR7/8: seven clusters; TLR4, NOD2, and NLRP3: five clusters; Fig. 5). One large proportion of infants (20 to 40%) produced low amounts of all of the innate ligand-related cytokines in response to stimulation and was designated as low responders (C1 in TLR3/4/NOD2/NLRP3; C2 in TLR7/8). Overall, the subgroup clustering illustrated large heterogeneity of the functional response in 18-month-old infants to the generic ligands from viruses and bacteria, as well as from the vaccine adjuvant alum. We found an overall low intra-individual overlap to different microbial ligands, illustrating that an adverse reaction to single-stranded viral RNA does not necessarily result in an adverse immune reaction to LPS or peptidoglycan (PGN) from bacteria or to double-stranded viral RNA (data file S1). This demonstrates that the individual antimicrobial response profile is selective and is based on the type of microbial stimulation. When addressing the influence of various perinatal determinants known to affect asthma development in childhood in the group of low responders, we identified a common denominator of more females among low responders for all microbial-derived ligands, except for NLRP3 stimulation (tables S5 to S9). This finding is in congruence with the enhanced innate-based immune response identified for the male infants.

Fig. 5 Great heterogeneity within innate ligand stimulated immune responses in blood of 18-month-old infants.

For each of the five innate ligands, we selected cytokines for which the cell-to-cytokine SCC was above 0.25 and subgrouped the response profiles based on hierarchical clustering. The color code for the chosen cytokines corresponds to the type-response designations from Fig. 1C. Data are z-score normalized per cytokine and plotted as the average score within each cluster. The identified clusters for each ligand are named by C and a number. The overlap across innate stimuli and response profile (cluster) for each individual is provided in data file S1. The percentage of infants in a given cluster is given by the width of the cluster and printed below each cluster. N = 541.

Abnormal neutrophil-linked response to viral ligands in 18-month-old infants developing transient childhood asthma

We then looked into the relation between these innate immune-defined clusters from stimulated blood of 18-month-old infants and development of childhood asthma. In our cohort of longitudinally clinically assessed children with immune analyses, we observed an overall asthma prevalence of 23% (127 of 541) until 6 years of age. At 6 years of age, 16% had outgrown their asthma diagnosis, which was termed transient asthma, whereas asthma persisted in 7% of the children, here characterized as persistent childhood asthma. To address relations between immune competence at 18 months of age and development of childhood asthma, we examined whether infants within the identified innate response clusters exhibited increased or decreased risk of asthma until 6 years of age.

For overall asthma prevalence, infants with enhanced IL-18, CXCL8, IL-1β, IL-6, and CCL20 production in response to single-stranded viral RNA (the TLR7/8-C7 cluster, Fig. 5) tended to exhibit enhanced risk of asthma until 6 years of age, as compared to infants within all remaining clusters [hazard ratio (HR), 1.74 [1.1 to 2.76]; P = 0.018; Padj = 0.091; fig. S6, A and B). Looking at the previous cell-to-cytokine correlation plot for TLR7/8, these cytokines all associated to neutrophil numbers, and infants in the TLR7/8-C7 cluster displayed enhanced production of IL-18 per neutrophil as compared to the remaining infants (fig. S6C). A similar tendency of increased overall asthma risk was seen for the antibacterial PGN-based response (NOD2) C3 cluster (HR, 1.66 [1.12 to 2.46]; P = 0.012; Padj = 0.057; fig. S6D), which represents individuals displaying enhanced CXCL8, IL-6, CCL20, TGF-β, and IL-10 release to PGN stimulation. All of these cytokines correlated to the number of neutrophils in the cell-to-cytokine plot for NOD2, and infants in the NOD2-C3 cluster displayed increased CCL20 production per neutrophil as compared to the remaining infants (fig. S6E).

When subdividing overall asthma cases until 6 years of age into children with transient asthma and children with persistent asthma at 6 years, we found a significantly enhanced risk of transient asthma in children with the neutrophil-linked response to viral single-stranded RNA (ssRNA; TLR7/8-C7 cluster; HR, 2.17 [1.27 to 3.70]; P = 0.0046; Padj = 0.035; Fig. 6, A and B) as compared to the remaining children. Consistent with this functional response pattern, children who developed transient asthma until 6 years of age also showed enhanced frequencies of circulating neutrophils at 18 months of age (Fig. 6C), suggesting that the transient asthma phenotype may be neutrophil-associated. Infants with an IL-18– and IL-1β–based alum response (NLRP3-C2), characteristic of an intermediate monocytes/neutrophil-linked response profile in response to alum, trended toward reduced risk for development of transient asthma until 6 years of age (NLRP3-C2; HR, 0.32 [0.13 to 0.79]; P = 0.014; Padj = 0.068; fig. S7).

Fig. 6 A distinct antiviral innate response profile in 18-month-old infants enhances the risk of developing transient childhood asthma.

Asthma development was followed longitudinally from birth to 6 years of age in the research clinic and defined as either transient or persistent asthma at 6 years of age. (A) The dot plot displays the relative risk of transient asthma development in infants within the given cluster versus the risk of transient asthma in the remaining infants. An encircled dot indicates the statistically significant association given in the text with Padj < 0.05. (B) Cox proportional hazards regression analysis of transient asthma development until 6 years of age in infants within the TLR7/8-C7 cluster. Percentage of infants in each cluster is given in Fig. 5. The Padj is determined by Benjamini-Hochberg FDR correction. (C) Relative prevalence of blood immune cells at 18 months of age in infants developing transient childhood asthma versus non-asthmatic children at 6 years of age. Cells were identified on the basis of flow cytometry of freshly collected blood and gated as illustrated in fig. S2. N = 541.

The TLR7/8-C7 children with increased risk for transient asthma showed no enhanced risk of persistent asthma at 6 years of age (fig. S8A), underlining that the neutrophil-linked antiviral IL-18–based response is a selective phenotype in 18-month-old infants developing transient asthma during childhood. Likewise, no statistically significant associations were identified between any of the other innate immune stimulated response clusters at 18 months of age and development of persistent asthma (fig. S8A), suggesting that the persistent childhood asthma phenotype may be defined by other immune pathways in 18-month-old infants.

An IL-5– and IL-13–enhanced T cell response in 18-month-old infants precedes development of persistent childhood asthma

We also examined the response profile of stimulated αβ T cells in relation to childhood asthma, as the relative frequencies of in vivo–activated CD4 and CD8 T cells were enhanced in 18-month-old infants developing persistent childhood asthma (fig. S8B). Because frequencies of recently activated CD4 T cells mainly correlated with IL-5 and IL-13 based on the cell-to-cytokine relations, we sought to establish whether this was an overall feature in all 18-month-old infants by use of the clustering approach. We used hierarchical clustering based on the selected T cell–related cytokines IFN-γ, IL-5, IL-13, IL-31, IL-17A, IL-22, TGF-β, and IL-10 to subgroup infants into their predominant cytokine response pattern. The phenotypic clustering resulted in identification of six distinct αβ T cell response phenotypes, termed on the basis of their predominant cytokine production as TH1 (IFN-γ), TH2 (IL-5 and IL-13), TH17 (IL-17A and IL-22), Tregs (TGF-β) (30), and mixed (Fig. 7A and fig. S9). The identified distinct subgrouping of predominant T cell memory responses at 18 months of age highlighted a large interindividual variation within αβ T cell responses at 18 months of age. Among the identified subgroups, around 20% of infants were found to have a predominant IL-5– and IL-13-based TH2 profile in activated αβ T cells, whereas the group of infants with a mixed response profile (10.5%; with high IL-5 and IL-13, concurrent with IL-31, IL-17A, TGF-β, and IL-10) displayed the highest IL-5 and IL-13 concentrations among all 18-month-old infants. Together, the infants with enhanced IL-5 and IL-13 in stimulated αβ T cells (TH2 + mixed) made up 30% at 18 months of age, which illustrates that the earlier reported TH2 dominance in early life (38) may be more balanced at 18 months of age, where 14.7% were TH1, 7% were TH17, and 11.4% were Tregs dominated based on the cytokine profiles. The remaining infants (35.7%) showed low responsiveness of T cells and also displayed significantly lower cell frequencies of recently in vivo activated CD4 T cells in blood circulation than the other children [low responders, 0.011 [0.0068 to 0.017] (median [interquartile range]); remaining infants, 0.013 [0.0087 to 0.019]; P = 0.02], whereas recently activated CD8 T cells did not differ. We found no particular risk determinants to associate with the low responders to SEB stimulation of T cells (table S10).

Fig. 7 IL-5– and IL-13–based T cell profile at 18 months of age associates with development of persistent childhood asthma.

(A) Stimulated αβ T cell response cytokines were selected and used for subgrouping of response profiles based on hierarchical clustering (z score–normalized per cytokine; fig. S9), resulting in six T cell subgroups. The percentage of infants in a given cluster is given by the dimension of the cluster and printed below each cluster. (B) The dot plot displays the prevalence of the indicated asthma phenotype (overall asthma until 6 years of age, persistent asthma at 6 years of age, transient asthma until 6 years of age) of 18-month-old infants within the indicated T cell cluster (x axis) as compared to remaining infants. Data are shown as ratios calculated as the prevalence of infants with disease within the indicated cluster versus the disease prevalence within remaining infants. An encircled dot indicates the statistically significant association given in the text with Padj < 0.05. (C) Cox proportional hazards regression analysis of persistent asthma development (0 to 6 years) within the two IL-5– and IL-13–enriched clusters at 18 months of age compared to remaining infants. The Padj is determined by Benjamini-Hochberg FDR correction. N = 541.

When associating the αβ T cell response subgroups to the risk of overall asthma until 6 years of age, as well as the transient and persistent childhood asthma phenotypes, we found the infants with enhanced IL-5 and IL-13 at 18 months of age (TH2 and mixed clusters) to have significantly higher risk of developing persistent asthma than non–IL-5– and IL-13–based groups (Fig. 7, B and C; HR, 2.31 [1.26 to 4.23]; P = 0.007; Padj = 0.028). Enhanced IL-5 and IL-13 release from stimulated αβ T cells significantly increased the risk of persistent asthma as compared to transient asthma: odds ratios (OR), 3.25 [1.44, 7.48]; P = 0.0049 (table S11). This increased risk is comparable to the risk of developing persistent versus transient asthma for children born by cesarean section (OR, 3.37 [1.25, 9.33], P = 0.016) and is much larger than observed for any of the other asthma-associated perinatal risk factors (table S11). No significant associations were found for any of the T cell subgroups in relation to development of overall childhood asthma and transient asthma until 6 years of age. It is moreover notable that the activated T cell phenotypes in infants with later persistent asthma did not overlap with those of children who developed transient asthma, highlighting that different immune deregulation cues underlie the transient and persistent childhood asthma phenotypes. The same conclusion is apparent when addressing the major risk factors for transient and persistent asthma (table S11).

Early life bacterial colonization of airways in infants is associated with the persistent asthma–associated immune risk cluster at 18 months of age

We next examined whether any of the major perinatal risk factors of asthmatic disease (15) may be linked to the identified immune risk clusters. We found colonization with pathogenic bacteria in the hypopharynx at 1 month of age to be a major risk factor for enhanced IL-5 and IL-13 secretion from stimulated αβ T cells at 18 months of age (TH2 + mixed αβ T cell clusters; OR, 1.87 [1.26, 2.78], P = 0.0020) (Table 1); the risk clusters also associated with persistent asthma. We found none of the major disease risk factors to associate with the TLR7/8-C7 cluster infants with increased risk of transient asthma (table S12).

Table 1 Genetic and environmental determinants of the at-risk immune cluster for persistent asthma.

Logistic regression analyses of genetic and environmental risk determinants on children in the at-risk immune clusters, αβTCR TH2 + mixed, versus remaining children of the cohort. CI, confidence interval.

View this table:


This study represents a large-scale, systems-level analysis of the functional immune response in 18-month-old children followed longitudinally for extensive clinical monitoring of asthma during childhood. When examining the functional immune competence in peripheral blood–derived immune cells, we identified the existence of five to seven different immune response phenotypes per microbial or T cell stimuli at 18 months of age. This revealed a large interindividual variation in functional response phenotypes to generic immune stimulants of bacterial and viral origin. Infants at risk of developing transient asthma until 6 years of age displayed a deregulated antiviral (TLR7/8) response phenotype that associated to type 17 cytokine enhancement without concomitant type 1 immune activation. Such immune phenotype may lead to recruitment of neutrophils upon viral encounter that, without the commonly coupled antiviral type 1 DC activation, might result in ineffective elimination of intracellular viral agents, hence provoking propagation of a neutrophil-associated transient asthma phenotype. Contrarily, children developing persistent childhood asthma showed enhanced IL-5 and IL-13 release from activated αβTCR T cells at 18 months of age, reflective of the classical TH2-based response profile, and this further associated to early airway colonization with pathogenic bacteria. Combined, these findings are indicative for different underlying immune pathologies in infants preceding development of the transient and persistent asthma phenotypes.

The immunological differences associated to sex that we identified to exist in 18-month-old infants, both with respect to immune cell frequencies and at their functional responsiveness, are congruent with previous reports from adults (31, 32). This difference in immune competence in early life suggests that boys and girls may react differently to exposures, such as vaccines, and thus could play a part in the observed sex-dependent variations in vaccine responsiveness in infants (33, 34). In this regard, the presented cell-to-cytokine relationship for each of the microbial-derived ligands may be useful in selection of new vaccine regimens targeting specific cellular subsets.

The approach with stratification of infants into subgroups based on the ex vivo functional response of blood immune cells to generic stimulants enabled the identification of a panel of mediators selectively enhanced in stimulated whole blood cultures of 18-month-old infants at risk of developing transient versus persistent asthma. Because the presently characterized disease-associated immune pathways were identified before disease development, the current strategy adds to previous seminal endotyping approaches founded in active disease settings in atopic dermatitis (35) and childhood asthma (14), the latter stratifying into allergic and nonallergic asthma in 4- to 15-year-old children. Both asthma subgroups, which resemble the herein identified persistent asthma subgroup with asthma at 6 years of age, were in Raedler et al. (14) found to hold a TH2 phenotype.

Intriguingly, from the innate immune clustering profiles, we found that infants differ widely in their immune response toward generic bacterial and viral ligands, with only 25 to 50% of infants responding with an adequate innate-based type 1 and type 17 immune response combined with anti-inflammatory IL-10 and little type 2 cytokine release to bacterial and viral ligands [for TLR3, 43% (C4/C5); TLR7/8, 28% (C5/C6); TLR4, 46% (C2/C3/C4); and NOD2, 33% (C4/C5)]. Likewise, NLRP3 activation with the common vaccine adjuvant alum in concert with low amounts of LPS promoted an inflammasome-based IL-18 and IL-1β response (C2) in about 17% of infants, a response mainly associated with intermediate monocytes and neutrophils, whereas classical monocytes linked to the response of the NLRP3-C3 cluster. The remaining 18-month-old infants produced an inappropriately regulated innate immune response type toward the given ligands or were generally poor responders. This leaves room for several adverse immune-to-disease links to be identified on the basis of longitudinal clinical monitoring, which would be in line with the notion that different, nonoverlapping immune programs may be involved in the trajectory to divergent inflammation-associated diseases (3638). Subgrouping of infants according to their principal αβ T cell–adaptive immune response profile revealed notable preferences for being either a responder with a unique αβ T cell response type (TH1, TH2, TH17, or Treg) or a low responder, as only few infants (11%) exhibited a mixed T cell cytokine profile yet with a predominant IL-5– and IL-13–based activity pattern. IL-5 and IL-13 release from stimulated blood T cells at 18 months of age more than doubled the risk of developing persistent childhood asthma, indicating that a mixed T cell response profile at 18 months of age, if encompassing enhanced IL-5 and IL-13 activity, would not rescue one from the increased risk of developing persistent asthma.

Identification of the link between the TLR7/8-C7 cluster and transient asthma risk made it apparent that the response to ssRNA viruses may be dysregulated in infants at risk of transient asthma. While it is widely accepted that single-stranded RNA viruses such as respiratory syncytial virus (RSV) and rhinoviruses, which stimulate immune responses via TLR7/8 activation, can cause asthma exacerbations (39), it has, to our knowledge, not previously been reported that a deregulated neutrophil-linked antiviral IL-18–based response seems to be selectively linked to transient asthma development, and not to persistent asthma. Moreover, it is also notable that such response pattern is detectable in stimulated blood-derived immune cells already at 18 months of age. It is also less well established that early inadequate antibacterial immune responses may associate with persistent asthma development. We have previously reported an association between early hypopharyngeal colonization with certain airway-associated bacteria and later asthma development in high-risk infants of asthmatic mothers in the COPSAC2000 cohort (40), which was then later identified to relate to an early adverse antibacterial TH2 response (enhanced IL-5 and IL-13) in afflicted children (41). In the current unselected COPSAC2010 cohort, we did not look for antigen-specific antibacterial TH2 responses, but our antigen-independent identification of heightened IL-5 and IL-13 release from polyclonally stimulated T cells from infants with enhanced early bacterial colonization of airways and increased risk of development of persistent asthma points to the same mechanisms of action in this unselected cohort. The TH2 connection to the persistent asthma phenotype might be a classical hallmark of the persistent disease trail, as memory T cells may stay in the body for life and become reactivated at each exposure. If the T cell–triggering antigens derive from colonizing airway bacteria (40, 41), as our data of risk determinants indicate, then such TH2-activated immune phenotype may result in continuous recruitment of eosinophils into airway tissue (IL-5) and enhanced mucus production (IL-13), resulting in triggering of tissue pathologies consistent with a persistent asthma phenotype.

Although the present data exemplify that the functional response in the immune system at 18 months of age may be useful as a sensor of later disease trajectories, it is a study limitation that samples were collected at the predefined 18-month time point as compared to longitudinal profiling in early life, which may have elucidated individual temporal trajectories.

Collectively, this systems immunological approach in 18-month-old infants demonstrated the existence of distinct antiviral, antibacterial, and T cell response phenotypes, which selectively increased the risk of developing transient or persistent asthma during the first 6 years of life. Our findings point to an inadequate innate immune handling of single-stranded RNA viruses as a risk for transient asthma development, while enhanced IL-5 and IL-13 activity in stimulated blood T cells at 18 months of age enhanced the risk for development of persistent asthma at 6 years of age. This latter was associated to early airway colonization with pathogenic bacteria. Further development of functional blood-based assays to detect these selective disease-related immune signatures in at-risk infants may assist early disease endotyping and improve prevention and treatment of childhood asthma.


Study design

Children were enrolled in the COPSAC2010 cohort (16), an ongoing unselected clinical prospective birth cohort of 700 children. The study was conducted in accordance with the guiding principles of the Declaration of Helsinki and approved by the Ethics committee for Copenhagen (H-B-2008-093) and the Danish Data Protection Agency ( 2008-41-2599). Both parents gave their informed consent before enrollment of the children. Blood was sampled between November 2010 and February 2013 in sodium heparin glasses from children at 18 months of age. Blood was analyzed the same day as sampled (mean = 3.5 hours, SD = 1.4 hours).

Statistical analysis

The fold induction of cytokine release (stimulated/unstimulated) was log10-transformed and plotted as polar charts. In addition, a PCA was performed on scaled (mean = 0, SD = 1) concentrations of induced cytokine release [delta cytokines (stimulated − unstimulated)]. Linear models (unpaired Student’s t tests) were used to study the inference of concentrations of individual immune cells and concentrations of cytokines produced (stimulated − unstimulated) in relation to sex. Statistical inference was corrected by the false discovery rate method by Benjamini-Hochberg, and data were visualized using the ggplot2 package for R. Associations between frequencies of cells before stimulation and concentrations of secreted cytokines [delta cytokines (stimulated − unstimulated)] were tested using Spearman correlation and visualized using the corrplot R package, resulting in cell-to-cytokine covariation plots. Following stimulation of TLR3, TLR4, TLR7/8, and NOD2, respectively, release of the innate cytokines IL-12p70, IFN-γ, IFN-β, CXCL10, IL-18, IL-23, CXCL8, IL-1β, IL-6, CCL20, TGF-β1, and IL-10 (delta stimulated − unstimulated) was scaled (mean = 0, SD = 1) and hierarchical clustered (using “hlust” from base R) using the complete linkage method. The same method was applied on delta concentration of IL-18, IL-1β, and CXCL8 after NLRP3 stimulation and on the T cell cytokines IFN-γ, IL-5, IL-13, IL-31, IL-17A, IL-22, IL-10, and TGF-β1 after SEB stimulation of αβTCR cells. Dependent on the given stimuli, five to seven clusters were selected. Selection of the number of clusters was based on á priori knowledge into common ligand-induced immune response types and continued until a cluster appeared, which represented an effective immune response type, and each cluster comprised at least 45 infants. The resulting clusters were visualized per stimuli in a heatmap (pheatmap package for R), where infants were ordered according to the cluster they belonged to. For each stimulus, the mean abundance of each cytokine in each cluster was visualized in additional plots. The risk of asthma development was assessed by Kaplan-Meier curves, where children of each immune response cluster or each immune cell frequency were tested against the remaining infants by Cox proportional hazards regression. Statistical inference to each disease phenotype was corrected by the false discovery rate method by Benjamini-Hochberg. To visualize the differences in disease prevalence between clusters (dot plots), we calculated the ratio (log10-transformed) based on the asthma prevalence in one immune response cluster divided by the asthma prevalence within the remaining children, and visualized using the ggplot2 package for R. Logistic regression analysis was used to test the association between perinatal risk factors and development of persistent versus transient asthma in childhood. The associations between perinatal risk factors and children within a given immune response cluster versus remaining children were analyzed by logistic regression. All presented OR and 95% CIs are univariate. All data were analyzed in R v.3.2.0 (Vienna, Austria). Additional methods are described in the Supplementary Materials.


Materials and Methods

Fig. S1. Overview of blood sampling at 18 months of age.

Fig. S2. Gating strategy for enumeration of immune cells by flow cytometry.

Fig. S3. Correlations between relative cell frequencies and absolute cell counts.

Fig. S4. Multivariate analysis of cytokine profiles from innate ligand-stimulated blood collected at 18 months of age.

Fig. S5. Hierarchical clustering of the functional response profile to innate ligands in blood immune cells from 18-month-old infants.

Fig. S6. Distinct antimicrobial innate response profiles in infants associate to overall risk of childhood asthma.

Fig. S7. Reduced risk of transient childhood asthma in infants within the NLRP3-C2 at 18 months of age.

Fig. S8. Early immune phenotypes in infants that develop persistent childhood asthma.

Fig. S9. Hierarchical clustering of the functional response profile in blood αβ T cells from 18-month-old infants.

Table S1. Demographic data of study cohort.

Table S2. Relative frequencies of immune cell subsets in whole blood at 18 months of age.

Table S3. Whole-blood concentration of immune cell subsets in 18-month-old infants.

Table S4. Concentrations of cytokines and chemokines after 24 hours incubation with media only (unstimulated, i.e., spontaneous release) or ligands targeting the indicated receptors.

Table S5. Genetic and environmental determinants of TLR3 low responders.

Table S6. Genetic and environmental determinants of TLR4 low responders.

Table S7. Genetic and environmental determinants of TLR7/8 low responders.

Table S8. Genetic and environmental determinants of NOD2 low responders.

Table S9. Genetic and environmental determinants of NLRP3 low responders.

Table S10. Genetic and environmental determinants of αβTCR low responders.

Table S11. Genetic and environmental determinants of transient and persistent asthma.

Table S12. Genetic and environmental determinants of at-risk TLR7/8-C7 immune cluster for transient asthma.

Data file S1. Intra-individual matrix comparing the response type to the antiviral and antibacterial ligands and the T cell response type.


Acknowledgments: We thank children and parents participating in the COPSAC2010 cohort, COPSAC nurses, and doctors for blood sampling; K. Vegener, L. B. Rosholm, and L. H. Boje for technical assistance; and P. Lovato and LEO Pharma for scientific input to cytokine selection. Funding: COPSAC is supported by private and public research funds all listed on The Danish Council for Strategic Research funded handling of blood samples and flow cytometry analyses, and LEO Pharma A/S funded MSD plates for cytokine and chemokine detection. A.H.T. was supported by a joint grant by Innovation Fund Denmark, LEO Pharma A/S, and COPSAC. The Lundbeck Foundation, the Ministry of Health, the Danish Council for Independent Research, and the Capital Region Research Foundation have provided core support for COPSAC. The funding organizations did not have any role in design and conduct of the study; collection, management, and interpretation of the data; or preparation, review, or approval of the manuscript. Author contributions: The guarantor of the cohort study is H.B., from conception and design to conduct of the study and acquisition of data, data analysis, and interpretation of data. A.H.T. designed and performed laboratory work, performed all statistical analyses, prepared figures in R, wrote the Materials and Methods section, and assisted in writing the manuscript. J.M.L. assisted with design of experimental work, J.W. with statistical analyses and figure layout, and M.A.R. with data quality control and statistical analyses. J.S., B.C., N.R.F., T.M.P., H.W., S.T., K.B., and H.B. were responsible for clinical assessments and for collection of perinatal exposure data. S.B. was responsible for experimental work, supervision of data analysis, and manuscript and figure layout and writing. All other authors have provided important intellectual input. All authors have agreed that the accuracy and integrity of any part of the work has been appropriately investigated and resolved and all have approved the final version of the manuscript. The corresponding authors had full access to the data and had final responsibility for the decision to submit for publication. No honorarium, grant, or other form of payment was given to anyone to produce the manuscript. Competing interests: Authors declare no potential, perceived, or real conflict of interest regarding the content of this manuscript. Data and materials availability: All data associated with this study are present in the paper or Supplementary Materials. The original raw data from whole blood–stimulated cytokine and chemokine release and flow cytometry files of circulating leukocytes are available via Dryad (doi: 10.5061/dryad.15dv41nsz). Summary- and feature-level data underlying Figs. 5 and 6, figs. S5, S7, S9, and S10, and tables S7 to S10 are provided in Table 1 and tables S1, S11, and S12. A data matrix combining the response type across stimuli per individual is available in data file S1. All other data that support the findings in this study, including clinical data, are available from the corresponding authors through a data transfer agreement. Participant-level personally identifiable data are protected under the Danish Data Protection Act and European Regulation 2016/679 of the European Parliament and of the Council (GDPR) that prohibit distribution even in pseudo-anonymized form, but can be made available under a data transfer agreement as a collaboration effort.

Stay Connected to Science Translational Medicine

Navigate This Article