Interactive effects between CDHR3 genotype and rhinovirus species for diagnosis and severity of respiratory tract infections in hospitalized children

ABSTRACT Rhinovirus (RV) is the leading pathogen causing childhood wheezing, with rhinovirus C (RV-C) species reported to cause asthma exacerbation. Allele A of single-nucleotide polymorphism (SNP) CDHR3_rs6967330 upregulates epithelial expression of RV-C receptors which results in more severe asthma exacerbations in children. Nevertheless, there are limited data on interactions between CDHR3 variants and their impact on severity of RV-related pediatric respiratory tract infections (RTIs). Medical records of RV-related RTIs in children aged below 18 years who were hospitalized in two public hospitals in 2015–2016 were independently reviewed by two paediatricians. Archived nasopharyngeal aspirates were retrieved for RV detection and sequencing as well as CDHR3 genotyping. HaploView v.5.0 and generalized multifactor dimensionality reduction (GMDR) analysis were employed for haplotypic assignment and gene-environment interaction analyses. Among 1019 studied cases, our results confirmed the relationship between RV-C species and more severe RTIs. Besides the top risk variant rs6967330-A, we identified rs140154310-T to be associated with RV-C susceptibility under the additive model [odds ratio (OR) 2.53, 95% CI 1.15–5.56; P = 0.021]. Rs140154310 was associated with wheezing illness (OR 2.38, 95% CI 1.12–5.04; P = 0.024), with such association being stronger in subjects who wheezed due to RV-C infections (OR 2.71, 95% CI 1.32–5.58; P = 0.007). Haplotype GAG constructed from rs4730125, rs6967330, and rs73195665 was associated with increased risk of RV-C infection (OR 1.71, 95% CI 1.11–2.65; P = 0.016) and oxygen supplementation (OR 1.93, 95% CI 1.13–3.30; P = 0.016). GMDR analyses revealed epistatic interaction between rs140154310 and rs6967330 of CDHR3 for RV-C infection (P = 0.001), RV-C-associated lower RTI (P = 0.004), and RV-C-associated wheeze (P = 0.007). There was synergistic gene-environmental interaction between rs3887998 and RV-C for more severe clinical outcomes (P < 0.001). To conclude, rs140154310-T is another risk variant for RV-C susceptibility and more severe RTIs. Synergistic epistatic interaction is found between CDHR3 SNPs and RV-C for RTI severity, which is likely mediated by susceptibility to RV-C. Haplotypic analysis and GMDR should be included in identifying prediction models of CDHR3 for childhood asthma and RTIs. IMPORTANCE This case-control study investigated the interaction between CDHR3 genotypes and rhinovirus (RV) species on disease severity in Hong Kong children hospitalized for respiratory tract infection (RTI). There were synergistic effects between RV-C and CDHR3 SNPs for RTI severity, which was mainly driven by RV-C. Specifically, rs6967330 and rs140154310 alone and their epistatic interaction were associated with RV-C-related and severe RTIs in our subjects. Therefore, genotyping of CDHR3 SNPs may help physicians formulate prediction models for severity of RV-associated RTIs.

KEYWORDS CDHR3, children, gene-environment interaction, respiratory tract infection, rhinovirus, wheezing W heezing is a major cause for pediatric hospitalization, and early-life wheezing is commonly the first presentation of childhood asthma.Viral infection is the leading cause of acute wheeze, whereas human rhinovirus (RV) is one of the most commonly detected pathogens in children hospitalized for wheezing illness as well as the main trigger for asthma exacerbation (AE) (1,2).Infants suffering from RV-induced wheeze are at risk of recurrent wheezing and later asthma development (3).Among RVs, the rhinovirus C (RV-C) species is associated with more severe wheezing pheno types compared to the A and B species (4,5).The asthma risk allele rs6967330-A on CDHR3 was related to the susceptibility to RV-C infection as the encoded variant Tyr529 was associated with higher surface expression of RV-C-specific receptor (6)(7)(8).The COPSAC2010 and COAST birth cohorts also supported this risk allele to be associ ated with more frequent respiratory tract infections (RTIs) within the first three years of life and particularly due to RV-C (9).Comparatively, there were limited data on other single-nucleotide polymorphisms (SNPs) of CDHR3 and their roles in RV-C-asso ciated RTIs.We hypothesized that other SNPs of CDHR3 also played significant roles in determining susceptibility and severity of RV-C infection.This cross-sectional study investigated the association between CDHR3 SNPs and severity of RV-C infection among Hong Kong children hospitalized for RTIs.
To minimize biased assessment of RTI severity, subjects with clinically significant co-morbidities or respiratory co-infections were excluded (Fig. S1).Table S1 describes the demographics and CDHR3 genotype distribution in analyzed and excluded subjects.Interestingly, RV-B cases had higher rate of respiratory co-infections than RV-A and RV-C (44% vs 21.5% vs 17.3%; Table S2).

CDHR3 genotypic association for RV-C susceptibility and clinical outcomes
Table S3 shows tested MAFs of 10 SNPs selected for genotyping.Rs448025 and rs543085868 were previously found to be monoallelic for the major allele in our local children, and they were not analyzed.The genotyping efficiency of eight tagging SNPs were over 95%, and MAFs in our subjects were comparable to those reported for Southern Han Chinese (CHS) in the 1000 Genomes database.
Table 3 summarized the genotypic associations with outcomes of RV infection.Under the additive model, the minor allele of rs140154310 was associated with higher risk of wheezing illness (OR 2.38, 95% CI 1.12-5.04;P = 0.024) after adjusting for age and gender.There was stronger association between this SNP and RV-C-associated wheeze (OR 2.71, 95% CI 1.32-5.58;P = 0.007), implying synergistic effect between rs140154310 and RV-C for wheezing.This SNP was also associated with RV-C-associated lower RTI (OR 2.55, 95% CI 1.24-5.24;P = 0.011), while the minor allele of rs73195665 was associated with systemic corticosteroid treatment (OR 1.69, 95% CI 1.04-2.75;P = 0.035).In addition, the minor allele of rs408223 was inversely associated with RV-C-associated lower RTI (OR 0.67, 95% CI 0.47-0.96;P = 0.025).Nonetheless, the top SNP rs6967330 was not associated with any clinical diagnoses and measures of RTI severity.

Haplotypic association for RV-C susceptibility, diagnosis, and RTI severity
The linkage disequilibrium (LD) plot for the eight SNPs spanning ~10 kbp on CDHR3 locus revealed two haplotype blocks (Fig. S2).Rs3887998 and rs73195657 defined HapA, while rs4730125, rs6967330, and rs73195665 defined HapB.Table 4 summarizes the results for haplotypic associations.HapB (GAG frequency 0.046) was associated with an increased risk of RV-C infection compared to other species (OR 1.71, 95% CI 1.11-2.65;P = 0.016).This risk haplotype was also associated with the need for oxygen supplement (OR 1.93, 95% CI 1.13-3.30;P = 0.016).In the meanwhile, HapA (AC frequency 0.050) was associated with lower RTI (OR 1.55, 95% CI 1.02-2.37;P = 0.041) but not other clinical parameters.None of the haplotypes was associated with RV-C-associated lower RTI, wheezing illness, and AE.

Interactions among CDHR3 SNPs on RV-C susceptibility and clinical outcomes
After adjustment for age and gender, GMDR analyses revealed the combination of tagging SNPs rs140154310 and rs6967330 to be the significant model predicting risk for RV-C infections, RV-C-associated lower RTI, and wheezing illness (P < 0.05, Table 5; Tables S4 and S5).Having a minor allele at either of these two loci was associated with increased risk for RV-C (OR 1.62, 95% CI 1.17-2.23;P = 0.003; Fig. 1A).Similarly, high-risk genotypes were associated with increased risk of RV-C-associated lower RTI (OR 1.57, 95% CI 1.13-2.19;P = 0.008; Fig. 1B) and wheezing illness (OR 1.56, 95% CI 1.11-2.18;P = 0.010; Fig. 1C) when compared with the low-risk genotype.GMDR analyses also identified a significant four-locus model from rs140154310, rs146004234, rs4730125, and rs6967330 to be associated with increased susceptibility for RV-C (TA > 55%, P < 0.001; Table 5).There was no epistatic interaction among the tested SNPs for clinical diagnoses and RTI severity (Tables S6 and S7).

Gene-environmental interaction between CDHR3 SNPs and RV-C infection
Gene-environmental interaction analysis from GMDR found a synergistic relationship between RV-C and CDHR3_rs3887998 for more severe RTI infections in terms of oxygen supplement (TA 68.9%, P < 0.001; Table 6) and systemic corticosteroid treatment (TA 72.7%, P < 0.001; Table 7).Subjects with major allele for rs3887998 and RV-C were at risk of oxygen supplement during hospitalization (OR 4.19, 95% CI 2.67-6.58;P < 0.001).

DISCUSSION
This case-control study extended the knowledge of CDHR3 by reporting a new risk SNP rs140154310 and its epistatic interaction with the known risk variant rs6967330 on increased susceptibility and severity of RV-C-associated RTIs.We identified genotypic and haplotypic associations between rs140154310 and rs6967330 for RV-C susceptibil ity.These factors also interacted to confer increased risk of wheezing illness from RV infection and requirements for oxygen supplement and systemic corticosteroid treatment.CDHR3 is the RV-C-specific surface receptor on airway epithelial cells, while the ubiquitous ICAM-1 and LDLR are the receptor for RV-A and RV-B (10).Rs6967330-A was consistently reported to be a risk allele for early-onset severe asthma and more frequent wheezing and RTIs, particularly those caused by RV-C infections (9,(11)(12)(13).Compared with the dominant rs6967330-G (Cys529), the asthma risk allele rs6967330-A (Tyr529) led to higher surface expression of transmembrane protein CDHR3, whereby increasing RV-C binding and replication (6).In contrast to the intuitive assumption that rs6967330-A was the late mutation that predisposed asthmatics to RV-C infections, evolutionary studies suggested that this allele was the ancient ancestral allele (14).Rs6967330-G was the more modern missense mutant that selected out under the pressure of evolving RV-C species to protect the general population from this viral infection by encoding the defective CDHR3 protein and downregulating its surface expression (15,16).
Apart from replicating the known association between RV-C and rs6967330, our data suggested that CDHR3_rs140154310, an intron SNP adjacent to the top SNP rs6967330, synergistically contributed to RV-C susceptibility and severity of wheezing.The main genetic variants of CDHR3 differed across populations (17).Among our Chinese subjects, rs140154310 was associated with increased risk of RV-C-associated RTI.This SNP was also involved in the two-locus GMDR model for RV-C susceptibility and RV-C-associated lower RTI and wheezing (Table 5; Tables S4 and S5).The previous study from our team also reported rs140154310 to be associated with more frequent wheezing and diminished lung function indices in Chinese preschool children (13).However, in silico prediction indicated low probability of this mutant for regulating protein transcript splicing (13).
In our current cohort with RV-associated RTIs, we identified genotypic and haplo typic associations between CDHR3 risk genotypes and RV-C infection and extended this genetic association to the requirement for oxygen supplement.Subgroup analyses showed that children with risk genotype infected with RV-C had even higher risks of wheezing illness (OR increased from 2.38 to 2.71) and oxygen supplement (OR from 4.04 to 4.19) by gene-environmental interaction analysis.The missense variant rs6967330 of CDHR3 increased RV-C susceptibility by upregulating surface expression of RV receptor CDHR3 in ciliated airway epithelium (7)(8)(9)18).We hypothesize that the synergistic effect between CDHR3 SNPs on RV-C susceptibility and more severe RTIs is mediated through regulation of surface expression of trans-membrane CDHR3 protein.
Experimental plasmid transfection of CDHR3 with rs140154310 and rs6967330 mutants into Hela cell lines could be the first step to understand the effect of rs140154310-T on transcription (mRNA) and surface expression (protein immunofluorescence) of the RV-C receptor CHDR3.Alternatively, we could measure CDHR3 mRNA level in peripheral blood or in vitro culture of epithelial cells from wheezing children who carried the two CDHR3 mutants so that we could analyze their correlations with the severity of wheezing illnesses.
Our study suggested that CDHR3 haplotype block with rs4730125, rs6967330, and rs73195665 (GAG) was associated with RV-C susceptibility and requirement of oxygen supplement.Moreover, GDMR analysis revealed epistatic interaction between CDHR3  SNPs and RV-C infection on the severity of RTIs.More recent evidence suggested genetic and environmental risk factors combined to determine the presenting RTI phenotype.In asthma development, host and environmental microbiota modulated the host immunity along with early-life allergen exposures to affect hosts' anti-viral immune responses, airway remodeling and inflammation, and asthma development (19).Host risk geno type also played a pivotal role in this intricate interaction.However, it was unclear how the genetic expression was regulated by environmental stimuli.Hammar et al. described decreased expression of CDHR3 mRNA in the peripheral blood leukocytes of preschool children with RV-associated wheezing, particularly among those with the risk A allele at CDHR3_rs6967330 (20).Forno and coworkers found an epigenetic prediction model based on DNA methylation of CDHR3 and other asthma genes to distinguish  6 and 7).Within each cell, the left bar represents cases with the corresponding RV-C outcomes, and the right bar shows the "control group." The numbers above bars denote sums of the GMDR scores.Subjects with high-risk genotypes formed by these two SNPs had OR of 1.62 (95% CI 1.17  between atopic and non-atopic phenotypes in school-age asthmatic children (21).Further investigations that integrate CDHR3 risk haplotype, epigenetic profile, resulting mRNA expression, RV-C infection, and clinical severity indices might help to identify genotype prediction model for asthma and RTI severity in paediatric patients.Furthermore, the human body functions as a sophisticated entirety with complicated interplay networks, so a single SNP or gene cannot be the sole determinant for severity of RV-associated RTIs.On the other hand, different genes would interact with each other to regulate anti-viral immunity.The locus 17q21 was widely recognized as an asthma gene cluster that regulated airway remodeling and chemokine responses (22).RV stimuli increased the expression of ORMDL3 and GSDMB on 17q21 locus in peripheral blood mononuclear cells (23).More recently, Eliasen et al. identified interaction between CDHR3 and GSDMB in children with early-onset severe asthma and homozygous G allele at GSDMB_rs2305480 (24).They further reported that this interaction might be mediated through upregulation of anti-viral cytokine IL-17A in peripheral blood.However, they did not detect any interaction between CDHR3 and GSDMB for RV-C susceptibility.
This study had several limitations.Firstly, RV titer could not be measured in the archived NPA samples.It remains unknown whether higher RV-C viral load mediated more severe RTIs.Secondly, the genotypes of other asthma genes were not determined in our cohort, so we were unable to examine any gene-gene interaction outside the CDHR3 locus.Future prospective studies with larger samples size and freshly collected respiratory samples are required to validate our findings of gene-gene and gene-envi ronment interactions for the severity of RV-associated RTIs.

Study population and design
This cross-sectional study retrieved nasopharyngeal aspirates (NPAs) archived in microbiology laboratories of two public hospitals under the New Territory East cluster in Hong Kong.These NPA samples were subjected for RV and host CDHR3 genotyping.The relationship between severity of RTIs and genetic and environmental factors were examined.The medical records of children aged below 18 years hospitalized for RTIs in these two hospitals between January 2015 and December 2016 were screened.On each calendar day, the first two enterovirus/rhinovirus-positive cases were selected to avoid seasonality bias.This study excluded subjects with co-morbidities that affected the assessment of clinical severity of RV infections such as compromised immunity, aspiration pneumonia, and unremitted bronchopulmonary dysplasia and co-infection with other respiratory pathogens.

Rhinovirus genotyping
As a standard practice, NPA was collected within 24 hr from patients hospitalized for RTIs.Following multiplex respiratory virus detection, the remaining NPA samples were stored at −80°C until analysis (2).Details on RV genotyping were described in the online supplement.

SNP selection and genotyping of CDHR3
Tagging SNPs of CDHR3 were selected using tagging algorithm of HaploView v.5.0 (Broad Institute, Cambridge, MA), following our published criteria (13).Briefly, the studied SNPs had MAF of ≥0.01 and r 2 of ≥0.8 for LD in CHS as cited in the 1000 Genomes database.
The selection region covered 5 kbp both upstream and downstream from rs6967330, with this top SNP being forced to be included in the tagging process.Host genomic DNAs extracted from archived NPA samples were subjected to CDHR3 genotyping by TaqMan SNP Genotyping Assays (Applied Biosystems) using QuantStudio 12K Flex Real-Time PCR System (Applied Biosystems).

Assessment of clinical outcomes
Two paediatricians independently reviewed the medical records for patients' discharge diagnoses and RTI severity.The exclusion criteria for clinical severity analyses inclu ded (i) medical records with insufficient details for the severity of RTIs; (ii) co-infec tion with other viral or bacterial pathogens; (iii) co-morbidities predisposing to more severe outcomes, such as unremitted bronchopulmonary dysplasia, recurrent aspira tion pneumonia, and immunocompromised conditions (congenital immunodeficiency, long-term treatment with systemic corticosteroids, and chemotherapy) within 3 months.Respiratory diagnoses were categorized into (i) upper RTI, (ii) wheezing illness, and (iii) lower RTI.The severity of RTIs was assessed based on duration of hospitalization, transcutaneous oxygen saturation level, requirement for oxygen supplement, treatment with systemic corticosteroid, admission to pediatric intensive care unit, and requirement for assisted breathing (see online supplement).

Statistical analysis
Demographic and clinical characteristics were compared by t-test or analysis of variance for continuous variables and Pearson's χ 2 test or Fisher exact test for categorical variables respectively.Hardy-Weinberg equilibrium for selected SNPs was evaluated with χ 2 exact test.Multivariable logistic and linear regression was used to analyze the association between CDHR3 SNPs and dichotomous and continuous clinical outcomes, adjusting for age and gender as covariates.Bonferroni correction was used to adjust for multiple statistical comparisons.Haploview v.5.0 (Daly Lab, Cambridge, MA) (25) was applied to identify haplotypes and calculate pairwise LD coefficients.Associations between haplotypes and clinical outcomes were analyzed by multivariate regression using R package haplo.stats(https://www.r-project.org/)(26).
GMDR was used to examine gene-gene interactions among SNPs and gene-environ mental interaction between SNPs and RV-C infections (27).Empirical P values from GMDR prediction models were estimated by comparing the observed average prediction error to the distribution of average prediction error under the non-association null hypothe sis from 5000 permutations.Models of maximized prediction accuracy were selected for ≥8/10 cross-validation consistency (CVC) and ≥0.55 testing accuracy (TA).Similar to multivariable regression, haplotype association and GMDR analyses were also adjusted for age and gender as covariates.Statistical comparisons were made two tailed using SPSS v.28 (IBM Corporation, Chicago, IL), with nominal significance level set at 0.05.

FIG 1
FIG 1 The best two-locus model derived from the GMDR analysis for RV-C-related outcomes.The best two-locus model was composed of SNPs rs6967330 and rs140154310 for (A) RV-C infection, (B) RV-C-associated lower RTI, and (C) RV-C-associated wheezing.Dark shading cells indicate high-risk genotype; light gray cell indicates low-risk genotype; and empty cells are non-shading.No homozygous minor allele exists for rs140154310 in our cohort (Tables6 and 7).Within each cell, -2.23, P = 0.003) for RV-C infection; OR of 1.57 (95% CI 1.13-2.19,P = 0.008) for RV-C-associated lower RTI; and OR of 1.56 (95% CI 1.11-2.18,P = 0.010) for RV-C-associated wheezing compared with those with low-risk genotypes.GMDR, generalized multifactor dimensionality reduction.

TABLE 1
Relationship between RV-C infection and risks for wheezing illness and RTI severity c

TABLE 2
Associations between RV-C susceptibility and tagging SNPs of CDHR3

TABLE 3
Associations between tagging SNPs of CDHR3 and clinical outcomes of our children under the additive model a Statistically significant results are highlighted in bold.September/October 2023 Volume 11 Issue 5 10.1128/spectrum.01181-235

TABLE 4
Haplotypic associations between CDHR3 SNPs and clinical outcomes from RV infections (Continued on next page)

TABLE 4
Haplotypic associations between CDHR3 SNPs and clinical outcomes from RV infections (Continued) a Adjusted for age and sex as covariates.b Statistically significant results are highlighted in bold.

TABLE 5
GMDR associations between tagging SNPs of CDHR3 and RV-C infection b a Adjusted for age and sex as covariates.b CVC, cross-validation consistency; SNP, single-nucleotide polymorphism; TA, testing accuracy.c Statistically significant results are highlighted in bold.

TABLE 6
Gene-environmental interactions between RV-C and CDHR3 SNPs on requirement for oxygen supplement b a Adjusted for age and sex as covariates.b CVC, cross-validation consistency; SNP, single-nucleotide polymorphism; TA, testing accuracy.c Statistically significant results are highlighted in bold.

TABLE 7
Gene-environmental interactions between RV-C and CDHR3 SNPs on systemic corticosteroid treatment b Adjusted for age and sex as covariates.b CVC, cross-validation consistency; SNP, single-nucleotide polymorphism; TA, testing accuracy.
a c Statistically significant results are highlighted in bold.