Enhancing Asthma Pharmacogenetics Through Subtype‐Specific Associations

Asthma is a heterogeneous disease comprised of multiple subtypes/endotypes. One common clinical manifestation of type 2 (T2) inflammation is the presence or absence of substantial eosinophilia. Inhaled corticosteroids (ICS) are the most prescribed asthma controller medications, but ~30% of patients fail to respond to ICS [1]. Pharmacogenetic variants in CRHR1 [1], GLCCI1 [2] and T-gene [3] are associated with ICS response, but to date have been insufficient to form prognostic biomarkers [4]. We hypothesised that endotypic-specific pharmacogenetics analyses would enhance the predictive ability of these variants. We tested this hypothesis by evaluating the effects of eosinophilia on genetic response to ICS. For our methods, we utilised prior genotyping [1-3] from the CAMP cohort, a multicentre, randomised, double-blinded clinical trial testing the safety and efficacy of inhaled budesonide versus nedocromil versus placebo over a mean of 4.3 years [5, 6]. CAMP enrolled 1041 children aged 5–12 years with mild to moderate asthma. Entry and ICS response criteria were as previously described [1-3], with the primary outcome of 2-month post-ICS FEV1 change following randomisation. Baseline eosinophils were dichotomised at 300/ul. We used an additive model in PLINK2 [7] to analyse three variants rs3127412 (T gene), rs37972 (GLCCI1) and rs242941 (CRHR1) and stratified by high versus low eosinophilia. R was used to calculate the odds ratio (OR) using logistic regression between upper and lower tertiles of FEV1 change for high and low eosinophils. Significance was set at p < 0.05. Area under the ROC curve metrics was calculated using auc(), roc () and roc.test() from the ‘pROC’ package [8]. A Welch t-test was performed for continuous clinical and demographic features and a chi-squared test for sex. We analysed 168 Caucasian subjects with genotypes and ICS FEV1 data. No significant differences (all p > 0.05) were found between low and high eosinophil groups in age (mean 8.8 vs. 8.7 years), age of onset (mean 2.7 vs. 3.2 years), sex (43% vs. 44% female), atopy (73.1% vs. 76.6%) or baseline FEV1 per cent predicted (95.3% vs. 92.4%). We noticed significant differences in PC20 (mean logPC20 0.372 vs. −0.233, p < 0.001) and lung function change after ICS (4.9% for low eosinophils vs. 11.3% for high eosinophils, p = 0.0007) between low and high eosinophil groups. Consistent with previous analyses [1-3], we noted a significant association between each of the three variants and overall lung function response to ICS in the entire cohort. In the eosinophil stratified analyses, rs3127412 and rs37972 were associated with 6.3% (p = 0.009) and 6.2% (p = 0.009) increase in FEV1 per allele for high eosinophils, compared to non-significant increases of 1.2% (p = 0.36) and 2.4% (p = 0.07) for low eosinophils. Figure 1a,b shows boxplots revealing associations for high eosinophils, but not for low eosinophils. For logistic regression, we noticed a similar trend to our additive models. We observed OR [95% CI] of 3.04 [1.4–6.6] and 2.7 [1.3–5.6] for SNPs rs3127412 and rs37972 with p-values 0.005 and 0.009, respectively, in high eosinophils. For low eosinophils, we did not observe a significant association for either SNPs with p-values 0.72, 0.25 and OR 1.2 [0.5–2.7], 1.6 [0.7–3.7], respectively. Our analyses revealed that for both variants, subjects in the high eosinophil group are approximately three times more likely to exhibit an elevated lung function response to ICS compared to the low eosinophil group. While the CRHR1 variant was associated with the whole cohort, there were no strata-specific effects. Additionally, we used the area under the ROC curve (AUC) to assess the predictive value of the SNP across the groups. Figure 1c illustrates the AUC that includes the T and GLCCI1 SNPs. SNPs within the low eosinophil group were poorly predictive of ICS response (AUC 0.54), whereas the SNPs demonstrated good prognostic ability in the high eosinophil group (AUC 0.72). SNPs within the combined cohort demonstrated modest predictive ability (AUC 0.61). Overall, our findings suggest that the prognostic value of pharmacogenetic SNPs is significantly enhanced in the context of a specific asthma endotype (here high eosinophil counts) and was reduced in patients with low eosinophil counts. Two of the three tested variants were associated with a > 12% improvement in FEV1 (comparing homozygous wild type vs. homozygous mutant) following ICS therapy within the high eosinophil group but showed no substantive change in the low eosinophil group. These variants also markedly increased the potential clinical pharmacogenetic prediction (AUC 0.72 vs. 0.54 in high vs. low eosinophil groups). These findings suggest that the prognostic value of these SNPs was significantly enhanced when limited to patients with high eosinophil counts and was reduced when limited to patients with low eosinophil counts. We acknowledge that unmeasured confounders, such as allergen exposures or other environmental triggers between visits, could have affected FEV1 and were not accounted for in this study. We also recognise that the lack of additional T2 biomarkers, such as FeNO, is a limitation of our study. In conclusion, we observed a pharmacogenetic link between two variants where the change in lung function was enhanced following ICS treatment for asthmatic individuals with higher levels of eosinophil compared to lower levels of eosinophil. This highlights the significance of personalised medicine methods in asthma management and suggests that larger studies involving more detailed looks at the pharmacogenetic basis for endotype/subtype response are warranted. Shraddha Piparia: data curation, formal analysis, methodology, visualization, writing – original draft. Julian Hecker: data curation, writing – review and editing. Upasna Srivastava: writing – review and editing. Rinku Sharma: writing – review and editing. Manaswitha Khare: writing – review and editing. Alvin Kho: data curation, writing – review and editing. Scott T. Weiss: project administration, writing – review and editing. Michael McGeachie: supervision, writing – review and editing. Kelan Tantisira: conceptualization, data curation, funding acquisition, statistical support, methodology, project administration, resources, supervision, writing – review and editing. This work was funded by the NIH R01 HL162570, R01 HL127332 and R01 HL161362. The author Scott T. Weiss receives royalties from UpToDate and is on the Board of Histolix a digital pathology company. The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.