Type 1 Immune Responses Related to Viral Infection Influence Corticosteroid Response in Asthma

Abstract

Rationale: Corticosteroid-responsive type 2 (T2) inflammation underlies the T2-high asthma endotype. However, we hypothesized that type 1 (T1) inflammation, possibly related to viral infection, may also influence corticosteroid response. Objectives: To determine the frequency and within-patient variability of T1-high, T2-high, and T1/T2-high asthma endotypes and whether virally influenced T1-high disease influences corticosteroid response in asthma. Methods: Patients in SARP-3 (Severe Asthma Research Program-3) had sputum collected at baseline, after intramuscular (triamcinolone acetonide) corticosteroid treatment, and at 1- and 3-year follow-ups. Sputum cell RNA was used for whole-transcriptome gene network and viral metagenomic analyses. We then profiled patients as highly expressing T1 and/or T2 gene networks and established the influence of these endotypes on corticosteroid responsiveness and the likelihood of viral transcript detection in the airways. Measurements and Main Results: We found that 22% and 35% of patients with asthma highly expressed T1 and T2 network genes, respectively, and that 8.5% highly expressed both networks. Asthma severity outcomes were worse in T2-high compared with T1-high asthma and most severe in the T1-high/T2-high subgroup. Corticosteroid treatment strongly suppressed T2 but poorly suppressed T1 gene expression, and corticosteroid-associated improvements in FEV₁ occurred only in patients with T1-low/T2-high disease and not in patients with T1-high/T2-high disease. Viral metagenomic analyses uncovered that 24% of asthma sputum samples tested positive for a respiratory virus, and high viral carriage was associated with 14-fold increased risk of T1-high disease. Conclusions: Airway T1 immune responses are relatively common in asthma, are largely corticosteroid resistant, and are associated with subclinical viral infection.

Keywords: sputum; transcriptomics; type 1 asthma; type 2 asthma; viral metagenomics.

Figure 1.

Sampling design for this study. Adults with and without severe asthma yield blood, sputum, and deep phenotyping samples at baseline and follow-up visits, such that inflammatory endotype expression and its associations with disease traits and steroid response can be investigated. RNA-seq = RNA-sequencing; T1 = type 1; T2 = type2; TA = triamcinolone acetonide.

Figure 2.

The frequency, intersection, and clinical characteristics of patients with airway type 1 (T1)-high and T2-high asthma endotypes. (A) A heat map of 36 T1 network genes as they are expressed among 782 SARP (Severe Asthma Research Program) samples hierarchically clustered into T1-low (T1-L) and T1-high (T1-H) groups. A key cytokine of T1 inflammation, IFNG, is highlighted. (B) Similar to A, but based on 35 T2 network genes. Key cytokines of T2 inflammation, IL-4, IL-5, and IL-13, are highlighted. (C) An UpSet plot visualizing the intersection between baseline T1 and T2 status among 260 participants, where the number of samples in each of the four dual endotype groups is on the y-axis and percentage of samples in each group is given over each bar. (D) Box plots depicting elevated immune cell counts (log₁₀) in the sputum of individuals sampled at baseline (n = 259) who also exhibit T1 and/or T2 inflammatory endotypes (from left to right within a plot, n = 132, 35, 70, and 22). *Log₁₀ immune cell counts significantly differ from the dual T1- and T2-low group (P < 0.05) based on an ANOVA that accounts for sex, age, and ethnicity (P values, from left to right: 9.70 × 10⁻¹², 1.33 × 10⁻⁰⁸, 1.18 × 10⁻⁰⁴, 6.02 × 10⁻⁰⁵, 2.00 × 10⁻⁰⁵, 0.0403, 0.0380). Data beyond the end of whiskers are not shown. Box centers = median; upper and lower box bounds = first and third quartiles; whiskers extend from these bounds up to 1.5× interquartile range.

Figure 3.

Coincident type 1 (T1) inflammation in T2-high (T2-H) patients blunts the FEV₁ response to triamcinolone acetonide. (A) Box plots show patient changes (Δ) in T1 network eigengene expression from the baseline visit, where steroid was administered, to the post-steroid treatment time point, among T1-high and T1-low (T1-L) groups (n = 128 for T1-low; n = 30 for T1-high). *Significant (P value = 0.0130) reduction in T1 network expression based on a mixed model that contrasts T1 expression before and after steroid treatment in T1-high individuals, with participant as a random effect. (B) Box plots show patient changes (Δ) in T2 network eigengene expression from baseline to the post-steroid treatment time point, among T2-high and T2-low groups (n = 102 for T2-low; n = 56 for T2-high). *Significant (P value = 1.73 × 10⁻¹⁷) reduction in T2 network expression based on a mixed model that contrasts T2 expression before and after steroid treatment in T2-high individuals, with participant as a random effect. (C) Box plots depicting changes in T2 network eigengene expression in response to steroid (Δ) among the four dual T1/T2 endotype groups (from left to right, n = 85, 17, 43, and 13). *Significant reduction in T2 network expression based on a mixed model that contrasts T2 expression before and after steroid treatment within a given endotype group (P values, from left to right: 1.29 × 10⁻¹⁹, 4.79 × 10⁻⁷). (D) Plot of the relationship between Δ T2 network eigengene expression with steroid against baseline (presteroid) T2 network eigengene expression for T2-high individuals, stratified by their T1 status, where red = T1-L/T2-H (n = 43) and purple = T1-H/T2-H (n = 13). Loess curves are overlain on the data points. Spearman correlation coefficients and P values for the two endotype groups are given. (E) Box plots depicting patient percentage FEV₁ responses to steroid (Δ) among the dual T1/T2 endotype groups (from left to right, n = 133, 35, 70, and 22). *Significant elevation in percentage FEV₁ based on a mixed model that contrasts percentage FEV₁ before and after steroid treatment within a given endotype group (P value = 4.51 × 10⁻⁸). ^†FEV₁ response to steroid is significantly different between groups based on the same mixed model (P values, left to right: 5.31 × 10⁻⁴, 0.0238). (F) Plot of the relationship between Δ FEV₁ with steroid against Δ T2 network eigengene expression with steroid for T2-high individuals, stratified by their T1 status, where red = T1-L/T2-H (n = 43) and purple = T1-H/T2-H (n = 13). Loess curves are overlain on the data points. Spearman correlation coefficients and P values for the two endotype groups are given. For all box plots, data beyond the end of whiskers are not shown. Box centers = median; upper and lower box bounds = first and third quartiles; whiskers extend from these bounds up to 1.5× interquartile range.

Figure 4.

Airway endotype is a patient characteristic with variable expression over time. (A) An alluvial diagram depicting patient persistence/variability of type 2 (T2) status across three time points (n = 72). (B) An alluvial diagram depicting patient persistence/variability of T1 status in the same individuals as in A across three time points (n = 72). T1-H = T1-high; T1-L = T1-low; T2-H = T2-high; T2-L = T2-low.

Figure 5.

Lower airway respiratory virus carriage is associated with expression of the type 1 (T1) endotype. (A) Pie charts describing the proportion of samples carrying virus as identified using our metagenomics pipeline (left; n = 605) or the proportion of different virus types observed among samples carrying virus (right; n = 146). (B) Density distribution for the log₁₀ number of virus reads detected among virus-positive samples (n = 146). The dashed line at 1,000 reads marks our threshold to discriminate between low and high virus detection. (C) Box plots visualizing increased T1 network expression among high virus samples (n = 39) compared with samples with no (n = 459) or low (n = 107) virus. Overlaid points are colored by number of log₁₀ virus reads in a sample. *Group differences are significant based on a mixed model predicting T1 network expression as a function of virus status, with participant as a random effect (P values, left to right: 1.35 × 10⁻³⁴, 1.30 × 10⁻³¹). (D) Bar plot depicting the proportion of samples that carry no, low, or high virus within T1-low (n = 478) and T1-high (n = 127) samples (proportions are listed on each bar). OR and P values, describing the excess chance of carrying low virus (compared with no virus; top) or of carrying high virus (compared with no/low virus; bottom) when T1-high compared with T1-low, were calculated using GEE logistic regression. (E) Box plots showing lower percentage FEV₁ among T1-high individuals with high virus loads (n = 29) compared with individuals who are T1-low (n = 443) or who are T1-high while carrying low/no virus (n = 96). *Significant differences in FEV₁ between groups based on a mixed model with a random effect for participant (P values, from left to right, are 0.00937 and 0.0195). GEE = generalized estimating equations; HRV = human rhinoviruses; OR = odds ratio; RSV = respiratory syncytial virus.