Whole Genome Sequencing Identifies CRISPLD2 as a Lung Function Gene in Children With Asthma

Abstract

Background

Asthma is a common respiratory disorder with a highly heterogeneous nature that remains poorly understood. The objective was to use whole genome sequencing (WGS) data to identify regions of common genetic variation contributing to lung function in individuals with a diagnosis of asthma.

Methods

WGS data were generated for 1,053 individuals from trios and extended pedigrees participating in the family-based Genetic Epidemiology of Asthma in Costa Rica study. Asthma affection status was defined through a physician’s diagnosis of asthma, and most participants with asthma also had airway hyperresponsiveness (AHR) to methacholine. Family-based association tests for single variants were performed to assess the associations with lung function phenotypes.

Results

A genome-wide significant association was identified between baseline FEV₁/FVC ratio and a single-nucleotide polymorphism in the top hit cysteine-rich secretory protein LCCL domain-containing 2 (CRISPLD2) (rs12051168; P = 3.6 × 10⁻⁸ in the unadjusted model) that retained suggestive significance in the covariate-adjusted model (P = 5.6 × 10⁻⁶). Rs12051168 was also nominally associated with other related phenotypes: baseline FEV₁ (P = 3.3 × 10⁻³), postbronchodilator (PB) FEV₁ (7.3 × 10⁻³), and PB FEV₁/FVC ratio (P = 2.7 × 10⁻³). The identified baseline FEV₁/FVC ratio and rs12051168 association was meta-analyzed and replicated in three independent cohorts in which most participants with asthma also had confirmed AHR (combined weighted z-score P = .015) but not in cohorts without information about AHR.

Conclusions

These findings suggest that using specific asthma characteristics, such as AHR, can help identify more genetically homogeneous asthma subgroups with genotype-phenotype associations that may not be observed in all children with asthma. CRISPLD2 also may be important for baseline lung function in individuals with asthma who also may have AHR.

Asthma is a disease with a strong genetic basis and substantial heterogeneity^{1, 2, 3} combined with early life and environmental factors⁴ that is characterized in three general domains based on most traditional approaches: (1) reversible airway obstruction, (2) airway hyperresponsiveness (AHR), and (3) airway inflammation.^3,5 Despite these features, patients differ across a wide range of clinical manifestations, inflammatory characteristics, severity, and outcomes,³ stemming from differing underlying endotypes and causes.^2,6 Asthma outcome has benefited from personalized care programs; however, despite the current national asthma guidelines,^7,8 its prevalence and economic burden continues to increase worldwide.^3,5 Common clinical symptoms shared across different asthma endotypes, along with variables including age and time of onset, sex, and lung function, make it extremely complicated to define appropriate treatment regimens and strategies for disease control. Although targeted therapies based on disease stratification and severity are available,^9,10 precise identification and classification of phenotypes and endotypes has become imperative and will benefit from conducting data-driven large multicenter genome-wide studies.

Albeit genome-wide association studies of asthma have identified and replicated associations of several single-nucleotide polymorphisms (SNPs) with asthma,^{11, 12, 13, 14} these variants explain only a fraction of the total of disease variants.¹ The use of broad categorizations tailored toward clinical classifications further complicate gene discovery through misclassification of case and control subjects that leads to reduced statistical power from misclassification. This explanation has become an impetus for creating more genetically homogeneous disease subgroups.

Whole genome sequencing (WGS) enables a more focused assessment of low-frequency and rare genetic variants; however, this assessment is not feasible until WGS data are available in a large enough number of individuals for sufficient statistical power to evaluate these low-frequency variants. To date, to our knowledge, only two WGS studies of asthma have been published.^15,16 A recently published WGS study identified several viral and bacterial species in children with asthma and pneumonia¹⁷; however, to our knowledge, none have focused on the genetic basis of lung function among individuals with asthma.^15,16 In this study, we present, to our knowledge, the first WGS analysis of lung function phenotypes by using family-based association tests (FBATs)^18,19 for extended pedigrees of 1,053 individuals who participated in the Genetic Epidemiology of Asthma in Costa Rica study (GACRS) cohort.²⁰

Materials and Methods

Study Population

The study included children with asthma who were aged 6 to 14 years from the GACRS family-based trios and probands from extended pedigrees who were recruited from a genetically homogeneous Hispanic population isolate in the Central Valley of Costa Rica with one of the highest prevalences of asthma worldwide (24% in children).²¹ Enrollment and baseline characteristics of the Costa Rican trios have been described previously in detail.²⁰ In brief, unrelated children were eligible if they had a high probability of having at least six great-grandparents born in the Central Valley of Costa Rica and they had asthma, defined as physician-diagnosed asthma plus a history of at least two episodes of troublesome lung symptoms or asthma attacks in the prior year.²² Children with an asthma diagnosis in this cohort also were evaluated for AHR by using a methacholine challenge. Additional probands were included from the extended pedigrees in which asthma was diagnosed by using the same diagnostic criteria. Further details can be found in e-Appendix 1.

Ethics

Oral and written parental consent and participating child’s assent were obtained. The study was approved by the Partners Human Research Committee at Brigham and Women’s Hospital (Boston, MA; Protocol No. 2000-P-001130/55).

Whole Genome Sequencing

WGS was performed as a part of the TOPMed program offered through the NHLBI. Details about the TOPMed sequencing and variant calling are available in e-Appendix 1.

Statistical Analysis

Single-Variant Family-Based Association Scans

Single-variant association analyses across four quantitative lung function phenotypes were performed using FBATs^18,19 that use the within-family information and are therefore robust to population stratification.²³ The FBAT approach tests for association between the offspring phenotype and the Mendelian residual. The Mendelian residual is described by the offspring genotype minus the expectation under Mendel’s law, computed using the parental genotypes. If parental genotypes are missing, FBAT uses the sufficient statistic approach to compute the expectation under the null hypothesis.²⁴ Treating the parental genetic data and the offspring phenotype in this manner means that this approach is robust against population stratification and phenotype misspecification. For the analysis, we used the publicly available software implementation FBAT Toolkit (https://sites.google.com/view/fbat-web-page).

The association analyses for the lung function phenotypes (prebronchodilator FEV₁, prebronchodilator FEV₁/FVC ratio, postbronchodilator [PB] FEV₁ , PB FEV₁/FVC ratio) were performed using two primary statistical models: (1) an unadjusted model in which the phenotype was the mean-centered lung function variable and (2) an adjusted model in which the phenotype was the lung function variable residual of a linear regression model that adjusted for sex, height, and dichotomized age (< or ≥18 years). We performed these analyses by using (1) all probands with available phenotypic information (531 probands of 1,053 individuals from the extended pedigrees and trios), including individuals with and those without an asthma diagnosis, and (2) only probands with an asthma diagnosis (n = 302). A minimum of five informative families was required for all of these analyses, which corresponds approximately to a minor allele frequency (MAF) cutoff of 0.3%.

Multivariate Family-Based Association Scan

Given the correlation between lung function phenotypes, we also used a multivariate generalized estimating equation (GEE) FBAT-GEE model.²⁵ FBAT-GEE accounts for the correlation between phenotypes and therefore minimizes multiple comparisons by analyzing all phenotypes simultaneously, resulting in one joint P value to evaluate whether the genetic variant is associated with any of the lung function phenotypes.

Rare Variant Analysis

Because of the limited sample size with WGS in GACRS, rare variant analyses were underpowered. Therefore, we did not present them here.

Replication Cohorts

We identified six validation cohorts of children with an asthma diagnosis including four from the National Heart, Lung, and Blood Institute (NHLBI) Trans-Omics for Precision Medicine (TOPMed) consortium (CAMP²⁶: Childhood Asthma Management Program, GALAII²⁷: Gene-Environments and Admixture in Latino Americans, SAGE²⁷: Study of African Americans, Asthma, Genes and the Environment and SAPPHIRE²⁸: Study of Asthma Phenotypes and Pharmacogenomic Interactions by Race-ethnicity) and two independent cohorts (HPR²⁹: Hartford-Puerto Rico (HPR) and COPSAC2000: Copenhagen Prospective Study on Asthma in Childhood^{30, 31}). Among these, three cohorts also identified that most of these children with asthma also had confirmed AHR (CAMP,²⁶ HPR,²⁹ COPSAC³⁰). Rest of the three cohorts did not have information about AHR (GALAII, SAGE,²⁷ SAPPHIRE²⁸). Details about the replication study populations can be found in e-Appendix 1.

Replication Analyses

Children with a diagnosis of asthma were included from each cohort for the replication. A total of 5,451 subjects with both genotype and phenotype data were included from CAMP (n = 769), HPR (n = 490), SAPPHIRE (n = 802), GALA II (n = 2,203), SAGE (n = 1,124), and COPSAC2000 (n = 63) as described in Table 1. Linear models with lung function variables, prebronchodilator and PB FEV₁ and FEV₁/FVC ratios, as outcome were used in both adjusted and unadjusted models in which the adjusted model included age, sex, and height. For CAMP, the models were additionally adjusted for the first two principal components, race and the interaction between genotype and race because this was the only replication population with multiple ethnicities. In SAPPHIRE, the models were adjusted additionally for genome-wide percentage of African ancestry. Combined P values were calculated across the replication cohorts by means of ascertainment schemas. P values were combined using the weighted z-score method, a widely used and robust method for combining P values in meta-analysis,³² from the Bioconductor survcomp³³ package in R because of variable sample sizes across cohorts.

Table 1.

Baseline Characteristics and Lung Function Measures of the Whole Genome Sequencing Population

Characteristic	Cohort
	GACRS	CAMP	COPSAC2000	HPR	SAPPHIRE	GALA II	SAGE
No. of probands	Total: 531 With Asthma: 302	769	63	490	802	2,203	1,124
Age, mean (SD), y	19.23 (16.49)	8.92 (2.14)	7.09 (0.32)	10.1 (2.67)	31.72 (14.72)	12.7 (3.3)	14.9 (5.3)
Male, No. (%)	271 (51)	469 (61)	33 (52)	267 (54)	304 (38)	1,199 (54)	551 (49)
FEV1, mean (SD), L	2.25 (0.93)	1.66 (0.48)	1.39 (0.23)	1.95 (0.70)	2.58 (0.77)	2.46 (0.83)	2.56 (0.76)
FEV1/FVC ratio, mean (SD)	0.84 (0.08)	0.79 (0.08)	0.91 (0.06)	0.83 (0.09)	0.76 (9.56)	0.84 (0.08)	0.83 (0.09)
PB FEV1, mean (SD), L	2.35 (0.93)	1.82 (0.51)	1.47 (0.23)	2.06 (0.73)	2.81 (0.76)	2.71 (0.87)	2.79 (0.81)
PB FEV1/FVC ratio, mean (SD)	0.87 (0.07)	0.85 (0.07)	0.96 (0.04)	0.85 (0.09)	0.80 (8.96)	0.89 (0.06)	0.87 (0.07)

No. of children indicates those with asthma diagnosed. GACRS corresponds to the discovery cohort; CAMP, COPSAC2000, and HPR correspond to the replication cohorts in which individuals had a physician’s diagnosis of asthma that was confirmed with airway hyperresponsiveness in most participants; SAPPHIRE, GALA II, and SAGE correspond to the replication cohorts that used a physician’s diagnosis of asthma with no information about airway hyperresponsiveness.

CAMP = Childhood Asthma Management Program; COPSAC2000 = Copenhagen Prospective Study on Asthma in Childhood 2000; GACRS = Genetic Epidemiology of Asthma in Costa Rica study; GALA II = Genes-Environments and Admixture in Latino Americans; HPR = Hartford-Puerto Rico; PB = postbronchodilator; SAGE = Study of African Americans, Asthma, Genes and Environments; SAPPHIRE = Study of Asthma Phenotypes and Pharmacogenomic Interactions by Race-Ethnicity.

Results

Baseline Characteristics

On the basis of 1,053 individuals from extended pedigrees, FBAT identified 317 nonsingleton nuclear families and 248 trios for the analysis. Baseline characteristics for the 531 GACRS samples that were used for phenotype construction and the replication cohorts are listed in Table 1.

Single-Variant FBAT

After the quality control and data cleaning procedure, 25.8 million autosomal SNPs remained for single-variant analysis. Additional details are available in e-Appendix 1.

Lung Function Phenotypes

Single-variant association analyses for the four lung function phenotypes led to association P values for 12.2 million SNPs, with λ inflation factors that ranged, across variants with more than 10 informative families, between 1.023 and 1.067. e-Tables 1 through 4 provide lists of all SNPs with association P values lower than 10⁻⁵ for each lung function phenotype analyzed in all children in the GACRS. In the evaluation of the covariate-unadjusted analysis, rs12051168 (Fig 1) achieved genome-wide significance for unadjusted baseline FEV₁/FVC ratio (P = 3.6 × 10⁻⁸) (e-Table 1). SNP rs12051168 is on chromosome 16 (16: 84879324) in the intronic region of the cysteine-rich secretory protein LCCL domain-containing 2 (CRISPLD2) gene with an MAF of 0.39 in Hispanics (Table 2). Another SNP, rs12919905, also in the intronic region of CRISPLD2 gene (chromosome 16: 84880424) suggestively achieved the genome-wide significance threshold (P = 5.34 × 10⁻⁸) (e-Table 1). The corresponding P value for rs12051168 in the covariate-adjusted model for FEV₁/FVC ratio was 5.61 × 10⁻⁶ (e-Table 1, Table 3). The adjusted models also showed similar associations for the other lung function phenotypes: baseline FEV₁ (P = 3.3 × 10⁻³), PB FEV₁ (P = 7.3 × 10⁻³), and PB FEV₁/FVC ratio (P = 2.7 × 10⁻³)(Table 3). The associations with lung function phenotypes were slightly attenuated in the covariate-adjusted model including those with asthma only (Table 3).

Figure 1.

LocusZoom plot highlighting the top hit (rs12051168) of the single-variant whole genome sequencing family-based association test analysis. The plot shows the covariate-unadjusted results with −log¹⁰ P values (y-axis 1), recombination rates (y-axis 2), and genome-wide significance threshold (dashed line) for SNPs (rs12051168 as highlighted by the red dashed line, rs12919905 adjacent to rs12051168 as highlighted by the gray dashed line) in the CRISPLD2 gene on chromosome 16. cM = centimorgan; Mb = Megabase; SNP = single-nucleotide polymorphism.

Table 2.

Replication SNPs, Their MAFs, and Ethnicity Across Cohorts

Cohort
Discovery	SNP	MAF	Ethnicity
GACRS	rs12051168	0.392	Costa Rican
	rs12051468	0.375
Replication
CAMP	rs12051168	0.37	White
	rs12051468	0.38
CAMP	rs12051168	0.14	Black
	rs12051468	0.38
HPR	rs12051168	0.32	Puerto Rican
	rs12051468	0.39
COPSAC2000	rs12051168	0.46	European, Danish
	rs12051468	0.43
SAPPHIRE	rs12051168	0.15	Black
	rs12051468	0.40
GALA II	rs12051168	0.32	Hispanic
	rs12051468	0.34
SAGE	rs12051168	0.15	Black
	rs12051468	0.39

MAF = minor allele frequency; SNP = single-nucleotide polymorphism. See Table 1 legend for expansion of other abbreviations.

Table 3.

Phenotypes Across Discovery and All Replication Cohorts for SNP rs12051168

Discovery Cohort
Phenotype	Cohort	z Score	P Valuea (All Samples)	P Valuea (With Asthma)	FBAT-GEE (Joint Phenotypes)
FEV1	GACRS	−2.940	3.3 × 10−3	1.3 × 10−2	4.5 × 10−5
FEV1/FVC ratio		−2.685	5.6 × 10−6	9.1 × 10−6	…
PB FEV1		−4.541	7.3 × 10−3	1.6 × 10−2	…
PB FEV1/FVC ratio		0.003	2.7 × 10−3	3.5 × 10−3	…

Replication Cohorts With Asthma Diagnosis and Confirmed AHR in Most Participants
Phenotype	Cohort	P Valuea			Meta-P Weighted z Score
Baseline FEV1/FVC ratio	CAMP	.055			.015
	HPR	.076			…
	COPSAC2000	.11			…

Replication in All Cohorts
Phenotype	Cohort	P Valuea			Meta-P Weighted z Score
Baseline FEV1/FVC ratio	CAMP	.055			.35
	HPR	.076			…
	COPSAC2000	.11			…
	SAPPHIRE	.71			…
	GALA II	.59			…
	SAGE	.58			…

Boldface indicates significant or suggestive P values. The CAMP cohort was adjusted additionally for race and interaction between genotype and race, which was either significant or suggestive in the models for baseline FEV₁/FVC ratio (P = .051) and PB FEV₁ (P = .054), and PB FEV₁/FVC ratio (P = .023); however, race was not significant in any model.

AHR = airway hyperresponsiveness; FBAT = family-based association test; GEE = generalized estimating equation. See Tables 1 and 2 legends for expansion of other abbreviations.

^aAdjusted for sex, age, and height.

The closest functional variant to the rs12051168 SNP in the CRISPLD2 gene was a splice donor at another SNP, rs185968132, which is illustrated in Figure 2. Furthermore, the SNPs within the 17q21 region, the most replicated childhood asthma locus, which encompasses the ORMDL3 (ORMDL Sphingolipid Biosynthesis Regulator 3) gene^{12, 13, 34}, had nominally significant (P <.05) associations with lung function. In the unadjusted analysis, the functional SNP, rs12936231,³⁵ yielded nominal associations with the baseline FEV₁/FVC ratio (P = .044), FEV₁ (P = 7.6 × 10⁻³) and PB FEV₁ (P = 8.9 × 10⁻³).

Figure 2.

The figure provides a haplotype overview of the SNPs in the CRISPLD2 gene on chromosome 16 showing linkage disequilibrium (LD) between the SNPs and corresponding haplotype blocks. TSS = transcription start site.

Multivariate FBAT and Haplotype Analyses

Using FBAT-GEE, we observed an overall significant association between the rs12051168 SNP and the lung function phenotypes (P = 4.5 × 10⁻⁵) (Table 3). To investigate further, we used the FBAT haplotype test¹⁸ for rs12051168, and the closest coding variant in CRISPLD2, rs12051468, was in linkage disequilibrium with rs12051168 (Table 4). This finding indicated that these two associated SNPs do not show an independent signal and that minor alleles are associated with decreased lung function. Haplotype TA has a frequency of 0.58 and is associated with increased lung function, whereas haplotype CG has a frequency of 0.35 and is associated with decreased lung function.

Table 4.

Haplotypes Showing Significant Associations for rs12051168 and for rs12051468

Haplotype	Allele Frequency	z Score	P Value
2-2 (TA)	0.58	4.38	1.2 × 10−5
1-1 (CG)	0.35	−4.30	1.8 × 10−5
1-2	0.04	−1.14	.25
2-1	0.02	0.29	.77

Replication in Independent Asthma Cohorts

Table 2 describes the MAFs of the two lung function SNPs of interest, rs12051168 and its coding variant rs12051468, in discovery and in replication cohorts. Notably, although white and Hispanic cohorts had a similar MAF for rs12051168, the black cohorts had a much lower MAF (∼0.15). Table 3 summarizes the replication findings in all cohorts. Although rs12051168 was not significant overall in all of the cohorts, this finding was replicated across the three independent cohorts (HPR, CAMP, and COPSAC2000) in which most children with asthma also had confirmed AHR (combined P value for baseline FEV₁/FVC ratio = .015) (Table 3).

Discussion

In this WGS family-based analysis of lung function phenotypes among children with asthma, we identified a plausible suggestive association between baseline FEV₁/FVC ratio and the rs12051168 SNP in the CRISPLD2 gene, a gene that has been implicated in pharmacogenetics mechanisms of asthma previously.³⁶ Nominally significant associations also were detected between rs12051168 and the baseline FEV₁ and PB FEV₁/FVC ratio. A multivariate GEE model detected a significant joint association between rs12051168 and the lung function phenotypes. This finding was replicated and showed a trend for all lung function phenotypes when meta-analyzed in three cohorts with a parental report of physician-diagnosed asthma with AHR confirmed by means of a methacholine challenge in most subjects, suggesting that this lung function association at rs12051168 may be specific to individuals with AHR. We were not able to replicate this finding in other cohorts because AHR was not measured in the other cohorts, so we could not evaluate the subjects on the basis of this criterion. Therefore, our results suggest that we can achieve better accuracy with a more stringent asthma definition and ascertainment. Together, these findings further help confirm those of previous studies^{36, 37, 38} that suggest CRISPLD2 is a lung function candidate gene for children with asthma that may be specific to those who also have confirmed AHR.

There is significant prior evidence that CRISPLD2, a glucocorticoid-responsive gene and regulator of immune response, could be a potential candidate for pharmacologic targeting to treat asthma. We previously reported that the CRISPLD2 gene on chromosome 16 may play a role in modulating two important asthma pharmacogenetic phenotypes in response to inhaled corticosteroid use.³⁶ The study highlighted several SNPs within CRISPLD2, or spanning 50 kilobase pairs on either side, that were nominally associated with inhaled corticosteroid resistance and bronchodilator response.³⁶ This prior work also identified nominally significant associations between SNPs in the CRISPLD2 gene and changes in lung function after investigating bronchodilator response to an inhaled short-acting β₂-agonist³⁶ in a discovery and replication study in more than 2,000 people with asthma.³⁷

Previous functional RNA-sequencing studies of human airway smooth muscle cell lines found that treatment with dexamethasone, a glucocorticoid,³⁶ was associated with increased CRISPLD2 messenger RNA and protein expression and that CRISPLD2 expression was induced by the proinflammatory cytokine IL-1β, suggesting that CRISPLD2 regulates antiinflammatory effects of glucocorticoids in the airways.³⁶ These findings were replicated in other studies by analyzing publicly available expression data from human airway smooth muscle cells treated with dexamethasone³⁸ and fluticasone.³⁹ An immune regulatory role of the CRISPLD2 protein is further supported by studies showing its potential for regulating endotoxin function.⁴⁰ However, it has been reported that even mutation or overexpression of late gestation lung 1 (LGL1) can lead to cell proliferation and metastasis via binding and upregulation of miR-652-3p in patients with non-small cell lung cancer.⁴¹

Although the regulatory role of CRISPLD2 in maintaining proper lung function is accumulating and being realized in humans, its role in crucial developmental processes leading to lung formation are widely reported in mouse studies. CRISPLD2, also known as LGL1, has been shown to regulate fetal lung development in experimental studies of rats.⁴² Several studies of rat lung models have shown that the glycoprotein-inducible protein encoded by CRISPLD2 is implicated in both branching morphogenesis and alveologenesis and that the absence of LGL1 is lethal to embryos.^{42, 43, 44} One of the studies highlights that postnatal LGL1 deficiency and knockout leads to production of inflammatory cytokines, altered pulmonary lung function, lung injury, and possible predisposition to an early onset of adult lung disease.⁴³ A 2015 human fetal airway fibroblast study⁴⁵ confirmed the importance of CRISPLD2 for fetal lung development, which may affect lung function phenotypes in childhood.

The association between CRISPLD2 and baseline FEV₁/FVC ratio in this study suggests that CRISPLD2 not only regulates treatment response to inhaled β₂-agonists and glucocorticoids but also may play a larger role in asthma pathogenesis by altering baseline lung function. These findings expand on our previous work and suggest that variants within CRISPLD2 may (1) affect altered baseline lung function, a key characteristic of asthma, and (2) be specific to children with confirmed AHR.

The top hit for the rs12051168 SNP within the CRISPLD2 gene was for baseline FEV₁/FVC ratio but not FEV₁. Children with asthma can have an abnormal FEV₁/FVC ratio despite a normal FEV₁ or FVC, a condition termed dysanapsis, which is believed to occur when early life lung volume and airway length growth outpace the increase in airway caliber.⁴⁶ We also previously have shown that airway dysanapsis was associated with worse lung function outcomes⁴⁷ in children with asthma who were obese,⁴⁶ highlighting further the importance of reevaluating asthma phenotypes and definitions. Therefore, to avoid potential observation bias, we chose to focus on lung function, one of the common attributes and features of asthma phenotypes in general.

The major strength of our WGS data is the broad coverage of the genome, providing in-depth information about copy number variations; noncoding and intergenic regions; and ability to access rare frequencies, particularly rare variants with MAF lower than 1%. However, we were not able to evaluate rare variants because of limited statistical power. Notably, we still identified a compelling common variant in our WGS analysis: the rs12051168 SNP with an MAF of 0.39 in the population in our study and 0.30 in the 1000 Genomes Project population.⁴⁸ The family-based single-variant FBAT approach,⁴⁹ with analysis of trios of children, parents, and extended pedigrees used in our study, is a significant advantage compared with population-based designs because it is robust against population admixture and stratification and allows both linkage and association tests,¹⁹ which could be used in future studies.

Despite these strengths, our study was subject to several limitations. Most notably, the small sample size makes the analysis underpowered and prohibits the evaluation of rare genetic variants. Another limitation may be the lack of consistency of phenotypes between the discovery and replication populations; most of the children had mild to moderate asthma with AHR in the discovery population, while the asthma phenotypes in the replication populations were heterogeneous.

Conclusions

In conclusion, we present the first, to our knowledge, WGS analysis of lung function phenotypes among children with asthma, identifying a variant in CRISPLD2 as associated with the baseline FEV₁/FVC ratio. Although this finding did not replicate in all six cohorts, there was nominal replication in three cohorts of children with a diagnosis of asthma among whom most also had confirmed AHR. Our findings suggest that more refined evaluation of asthma phenotypes focused on the physiologic characteristics of asthma is useful for identifying genetic variants for specific asthma domains. With CRISPLD2 known to regulate the antiinflammatory effects of glucocorticoids in airway smooth muscle cells, our WGS data extend those findings and suggest that CRISPLD2 may play a larger role in asthma pathogenesis, which is biologically plausible because mechanistic studies have proposed a role of CRISPLD2 in lung organogenesis.