Allergic phenotypes and asthma in multiple sclerosis: Epidemiologic and genetic relationships

Abstract

Background:

Multiple sclerosis (MS) and allergic disorders (ADs) are classically seen as immunologically distinct, and prior studies suggest potential inverse associations and overlapping genetic contributors.

Objectives:

Compare the prevalence of asthma, and dietary and environmental allergic phenotypes in persons with MS (PwMS) versus unaffected controls, compare AD prevalence by MS clinical course, and evaluate associations between MS genetic risk, HLA-DRB1*15:01 and a non-MHC genetic risk score (GRS), and ADs in PwMS.

Methods:

We analyzed survey and genotype data from 1542 PwMS and 700 controls from the Accelerated Cure Project. Logistic regression models were adjusted for key confounders. Genetic analyses included 1252 unrelated non-Hispanic White PwMS, assessing the main effects of HLA-DRB1*15:01 and the GRS. Benjamini–Hochberg procedure controlled the false discovery rate.

Results:

PwMS had lower odds of seasonal allergies compared to controls (OR = 0.79; 95% CI: 0.65, 0.96), with a non-significant reduction in primary progressive versus relapsing at onset MS (OR = 0.58; 95% CI: 0.37, 0.92). Among PwMS, HLA-DRB1*15:01 was associated with a higher odds of gluten allergies/intolerance, while the GRS was inversely associated with childhood-onset seasonal allergies.

Conclusions:

MS is associated with reduced seasonal allergies. MS genetic risk factors appear to shape allergy susceptibility in PwMS.

Introduction

Multiple sclerosis (MS) and asthma, atopic, and allergic disorders (ADs) are immune-mediated conditions with distinct immunological profiles. MS is prototypically viewed as a Th1/Th17-driven autoimmune disease while ADs are Th2-skewed and involve IgE-mediated hypersensitivities.^1,2 Early European case-control studies, ranging up to 423 cases and 643 controls, reported inverse relationships between MS and asthma and allergic rhinitis (seasonal allergies).^3–5 A U.S. study of 271 pediatric MS cases and 418 controls observed non-significant reductions in environmental (p = 0.08) and food (p = 0.16) allergies in MS cases prior onset.⁶ However, it remains unclear whether inverse relationships extend to US adults with MS, particularly in the context of allergic endophenotypes.

Findings from a case-only US study of 1349 persons with MS (PwMS) revealed those with food allergies had higher relapse rates and were more likely to exhibit gadolinium-enhancing lesions (p < 0.05).⁷ There was suggestive evidence that AD prevalence varied by clinical course (p = 0.052), with a trend toward lower prevalence among progressive-onset MS cases.⁷ These associations are clinically relevant given that two-thirds of PwMS in that study reported at least one allergy, spanning environmental, dietary, and medication-related triggers.⁷ In addition, the prevalence of asthma in PwMS has doubled over the past two decades, from 7% in the early aughts to over 15% in 2018.^8,9 The drivers of these trends remain unclear but may involve changes in environmental exposures, improved diagnostic recognition, or interactions between novel exposures and host genetic susceptibility.

While MS and ADs are immunologically distinct,¹⁰ overlapping genetic risk loci exists. Variation within the major histocompatibility complex (MHC) on chromosome 6p21.33, specifically the HLA-DRB1*15:01 allele that encodes a class II molecule, is the strongest risk locus for MS, with 200 non-MHC autosomal variants contributing additional modest risk.¹¹ Similarly, multiethnic genome-wide association studies (GWAS) implicate the MHC and class II molecules in asthma susceptibility and IgE regulation, though fine-mapping of HLA alleles remains limited.^12–15 Several non-MHC MS risk loci, including IL7R, IL2RA, TNFAIP3, and CLEC16A, have been implicated in AD etiology.^12,13 Despite these overlaps, no compelling immunogenetic evidence supports presence of genetic correlations between MS and ADs,^16–18 and Mendelian randomization studies have not supported a causal relationship in either direction.^18–20 Importantly, many immune loci exhibit pleiotropic and sometimes opposing effects across autoimmune conditions (e.g. HLA-DRB1*15:01 is protective in type 1 diabetes), suggesting that genetic associations between MS risk loci and ADs could plausibly operate in either direction.

Given these gaps, our study pursued four objectives. First, to confirm the previously reported inverse associations between MS and asthma or seasonal allergies in a large, US case-control of adult-onset MS (hypothesis: PwMS will have a lower prevalence of asthma and seasonal allergies). Second, to explore whether other less common allergic phenotypes, including specific food and environmental allergies, differ between PwMS and unaffected controls (hypothesis: PwMS may have a different prevalence of less common ADs). Third, to examine whether allergic prevalence varies by MS clinical course at onset (relapsing vs progressive), building on preliminary findings that suggest a lower AD prevalence in progressive at onset MS (hypothesis: AD prevalence will differ by presenting disease course). Fourth, to evaluate whether MS genetic risk (HLA-DRB1*15:01 and the non-MHC genetic risk score (GRS)) associate with allergic phenotypes among unrelated PwMS (hypothesis: MS genetic risk will confer differential risk for allergic phenotypes given the known pleiotropy of immune-related loci).

Materials and methods

Study population and study design

The Accelerated Cure Project for MS recruited PwMS and unaffected controls from communities of 10 US MS specialty centers between May 2006 and December 2014 (www.acceleratedcure.org; eMethods). Participant inclusion/exclusion criteria have been described.²¹ Neurologists confirmed MS cases met diagnostic criteria,^22,23 participants were ⩾18 years at onset, provided informed consent at enrollment (including future multi-omic use of biological samples), and completed detailed epidemiologic surveys administered by trained staff. Cases were invited to refer unrelated and related cases and controls (i.e. siblings and spouses), with controls defined as individuals without a prior demyelinating event characteristic of MS, transverse myelitis, acute disseminating encephalomyelitis, neuromyelitis optica spectrum disorder, or optic neuritis and who did not have a diagnosis of any other demyelinating disease.²¹ Based on these data, we constructed a case-control study. Genetic data were available for a subset of unrelated non-Hispanic White (NHW) MS cases, and has been previously described (see eMethods).

Dependent variables

Dependent variables included multiple AD phenotypes. Participants were asked if they had seasonal allergies or asthma and to report age at onset. For both traits, we created two definitions: any history and childhood-onset (before age 18)—the latter for sensitivity analyses. In addition, six open-ended prompts invited participants to report food, drug, or other allergies. Responses ranged from general categories (e.g. seafood) to specific items. These were reviewed and coded into indicator variables for seafood (including shellfish), gluten, dairy, eggs, nuts, latex, pet, or soy allergies by multiple authors blinded to MS status (F.B., S.Z., L.J.L.), with all coding reviewed until consensus was reached (F.B.). AD phenotypes were based on self-report.

Statistical analyses

Descriptive summaries were generated and compared using Wilcoxon rank-sum tests or Fisher’s exact tests for continuous or categorical variables, respectively. For Objective 1, multivariable logistic regression models examined the relationship between MS status (independent variable) and presence seasonal allergies (dependent variable), adjusting for age, sex, birth year (to account for cohort effects), race (Non-Hispanic White (NHW), Non-Hispanic Black, Other), smoking status (ever/never), years of education (as a proxy for socioeconomic status and diagnostic awareness), and recruitment site (to account for geography). Cluster-robust standard errors accounted for dependency due to household or familial relatedness, correcting variance estimates without altering coefficients. This model was repeated for asthma, and the Benjamini–Hochberg procedure was used to control the false discovery rate (FDR-BH). For sensitivity analyses, we also examined associations for seasonal allergies or asthma before age 18.

For Objective 2, similar models were conducted for each of the other 8 ADs (dependent variable) and MS status (independent variable), with an FDR-BH correction. Objective 3 used similar models for each AD (dependent variable) but MS disease course was the independent variable: relapsing remitting MS (RRMS) versus primary progressive MS (PPMS) at onset; FDR-BH correction was applied.

For Objective 4, the analyses were restricted to unrelated NHW PwMS for whom genetic data was available (eMethods).^24,25 The model included main effects for HLA-DRB1*15:01 (0, 1, or 2 alleles) and the unweighted GRS for 198 of the 200 non-MHC risk variants, adjusting for age, sex, smoking, education, and ancestry (first five multidimensional scaling components)²⁵ with the standard errors accounting for possible dependency by recruitment site. The model assessing gluten allergy/intolerance additionally adjusted for self-reported celiac disease. Analyses were conducted using RStudio (glm function) and STATA v17.0 (logit for regression models; multiproc for FDR-BH correction) (StataCorp, TX).

Ethics statement

This secondary analysis of de-identified epidemiologic and genetic data from the Accelerated Cure Project for MS was deemed non-human subjects by the Case Western Reserve University Institutional Review Board (Protocol Number: IRB-2016–1583).

Results

The study included 2242 participants, comprising 1542 PwMS and 700 unaffected controls with ~90% identifying as NHW (Table 1). The average age at interview was similar between groups (46.3 years in PwMS vs 47.6 years in controls), with cases more likely to be female (MS = 77.7%, n = 1198 vs controls = 66.6%, n = 466) and slightly more educated on average (15.8 vs 15.2 years). Smoking history was more prevalent among PwMS (47.2%; n = 728) than controls (41.8%; n = 292). Among PwMS, the average age at MS onset was 33.5 years. At onset, 92.7% (n = 1429) had RRMS and 7.3% (n = 113) had PPMS.

Table 1.

Study population characteristics comparing all MS cases to controls, and MS cases by disease course at onset.

Characteristic		All	MS	Controls	MS versus HC	Disease course at onset		RR versus PP
					p-valuea	RRMS	PPMS	p-valuea
N		2242	1542 (68.8%)	700 (31.2%)		1429 (92.7%)	113 (7.3%)
Age at interview (mean, SD)		46.7 (13.1)	46.3 (11.1)	47.6 (16.7)	0.0016	45.6 (10.9)	54.7 (8.8)	< 0.0001
Birth year (mean, SD)		1961 (13.3)	1962 (11.2)	1961 (16.9)	0.019	1962 (11.1)	1953 (8.8)	< 0.0001
Female (%; n)		74.2% (n = 1664)	77.7% (n = 1198)	66.6% (n = 466)	< 0.0001	78.7% (n = 1125)	64.6% (n = 73)	0.0009
Years of education (mean, SD)		15.6 (3.0)	15.8 (2.9)	15.2 (3.22)	<0.0001	15.8 (2.9)	15.7 (2.9)	0.721
Ever smoker (%; n)		45.5%	47.2%	41.7%	0.011	46.3%	58.4%	0.014
		(n = 1020)	(n = 728)	(n = 292)		(n = 662)	(n = 66)
Age of onset (mean, SD)		–	33.5 (9.9)	–	–	32.9 (9.5)	41.6 (10.3)	< 0.0001
Race (%; n)	Non-Hispanic	90.3%	90.1%	90.3%	0.719	90.3%	88.5%	0.589
	White	(n = 2022)	(n = 1390)	(n = 632)		(n = 1290)	(n = 100)
	Non-Hispanic	7.2%	7.5%	6.9% (n = 48)		7.3% (n = 104)	9.7% (n = 11)
	Black	(n = 163)	(n = 115)	2.9% (n = 20)		2.4% (n = 35)	1.8% (n = 2)
	Other	2.5% (n = 57)	2.4% (n = 37)

Abbreviations: HC = unaffected controls; PP/PPMS = primary progressive multiple sclerosis; RR/RRMS = relapsing remitting multiple sclerosis; SD = standard deviation.

^ap-values for Wilcoxon rank-sum tests or Fisher’s exact test for continuous and categorical variables, respectively.

Objective 1: seasonal allergies and asthma in MS cases versus controls

Seasonal allergies were the most reported AD condition in both PwMS (36.8%) and controls (41.9%) (Table 2). Adjusted logistic regression showed that PwMS had a significantly lower odds (21% lower or 0.79 times the odds) of reporting seasonal allergies (OR = 0.79; 95% CI: 0.65, 0.96). This significant relationship persisted when restricting to childhood-onset seasonal allergies (OR = 0.75; 95% CI: 0.59, 0.96). Asthma and childhood-onset asthma were slightly less prevalent in PwMS versus controls, but these associations did not reach statistical significance.

Table 2.

Prevalence of seasonal allergies and asthma and adjusted associations for MS versus controls (objective 1).

Allergy phenotype	All MS (n = 1542)	Controls (n = 700)	MS versus HC
			OR (95% CI)
Seasonal allergies	36.8% (n = 567)	41.9% (n = 293)	0.79 (0.65, 0.96)a
Seasonal allergies before 18	18.6% (n = 287)	22.0% (n = 154)	0.75 (.59, 0.96)a
Asthma	11.5% (n = 177)	13.9% (n = 97)	0.78 (0.58, 1.03)
Asthma before 18	6.8% (n = 105)	8.9% (n = 62)	0.75 (0.53, 1.03)

Abbreviations: HC = unaffected controls; OR = odds ratio; CI = confidence interval.

Associations from logistic regression models adjusted for age, gender, birth year, years of education, ever smoker status, race, and recruitment site, and clustering of the standard errors to account for household/family relationships.

^aTwo-sided uncorrected p < 0.025 and significant after FDR-BH correction.

Objective 2: other less common allergies in MS cases versus controls

No statistically significant differences were observed for diet or other environmental ADs (prevalence < 4% for each AD phenotype in MS cases and controls) (Table 3).

Table 3.

Prevalence of less common diet/environmental allergic phenotypes and adjusted associations for MS versus controls (objective 2).

Allergy phenotype	All MS (n = 1542)	Controls (n = 700)	MS versus HC
			OR (95% CI)
Seafood	2.9% (n = 44)	1.7% (n = 12)	1.75 (0.89, 3.42)
Gluten	1.2% (n = 19)	0.6% (n = 4)	1.83 (0.61, 5.51)
Dairy	2.5% (n = 39)	3.6% (n = 25)	0.74 (0.44, 1.26)
Eggs	0.8% (n = 12)	0.4% (n = 4)	1.60 (0.44, 5.76)
Nuts	1.7% (n = 26)	1.1% (n = 8)	1.40 (0.67, 2.95)
Latex	1.0% (n = 15)	0.4% (n = 4)	1.89 (0.49, 7.23)
Pets	1.0% (n = 16)	1.3% (n = 9)	0.73 (0.30, 1.79)
Soy	0.3 (n = 5)	0	-

Abbreviations: HC = unaffected controls; OR = odds ratio; CI = confidence interval.

Associations from logistic regression models adjusted for age, gender, birth year, years of education, ever smoker status, race, and recruitment site, and clustering of the standard errors to account for household/family relationships.

Objective 3: allergy prevalence by MS disease course

PPMS cases had a lower prevalence of seasonal allergies compared to those with RRMS at onset (24.1% vs 37.8%) (Table 4). In adjusted models, PPMS cases had a lower odds (42% lower or 0.58 times the odds) of seasonal allergies than RRMS at onset cases (OR = 0.58; 95% CI: 0.37, 0.92). A similar inverse relationship was also observed for childhood-onset seasonal allergies (OR = 0.73; 95% CI: 0.54, 0.97). However, when controlling the FDR for multiple testing, these associations were no longer significant. For other allergic phenotypes, there were no meaningful differences by MS disease.

Table 4.

Prevalence of allergic phenotypes and adjusted associations for PPMS versus RRMS (objective 3).

Allergy phenotype	RRMS (n = 1429)	PPMS (n = 113)	PP versus RR
			OR (95% CI)
Seasonal allergies	37.8% (n = 540)	23.9% (n = 27)	0.58 (0.37, 0.92)a
Seasonal allergies before 18	19.4% (n = 277)	8.8% (n = 10)	0.73 (0.54, 0.97)a
Asthma	11.8 (n = 169)	7.1% (n = 8)	0.84 (0.41, 1.73)
Asthma before 18	6.9% (n = 99)	5.4% (n = 6)	1.22 (0.50, 2.99)
Seafood	2.8% (n = 40)	3.6% (n = 4)	1.12 (0.40, 3.12)
Gluten	1.1% (n = 16)	1.8% (n = 2)	2.61 (0.54, 12.61)
Dairy	2.4% (n = 34)	4.4% (n = 5)	2.61 (0.97, 7.03)
Eggs	0.8% (n = 11)	0.9% (n = 1)	0.95 (0.19, 4.84)
Nuts	1.8% (n = 26)	0	–
Latex	1.0% (n = 14)	0.9% (n = 1)	1.16 (0.16, 8.26)
Pets	1.1% (n = 16)	0	–
Soy	0.4% (n = 6)	0	–

Abbreviations: OR = odds ratio; CI = confidence interval; PP/PPMS = primary progressive multiple sclerosis; RR/RRMS = relapsing remitting multiple sclerosis.

Associations from logistic regression models adjusted for age, gender, birth year, years of education, ever smoker status, race, and recruitment site, and clustering of the standard errors to account for household/family relationships.

^aUncorrected two-sided p < 0.05, but not significant after FDR-BH correction.

Allergic phenotypes and MS genetic susceptibility in PwMS

Among 1252 unrelated non-Hispanic White PwMS, genetic models evaluated associations between HLA-DRB1*15:01 copy number, MS GRS, and ADs (Table 5). After FDR-BH correction, there were two significant associations: (1) a one-allele increase in HLA-DRB1*15:01 was associated with a 3.47-fold (95% CI: 1.98, 6.08) increase in the odds of gluten allergies/intolerance among PwMS, and (2) a one-allele increase in the GRS was associated with decrease in the odds (0.98 times the odds (95% CI: 0.97, 0.99)) of childhood-onset seasonal allergies in PwMS.

Table 5.

Adjusted associations for MS genetic susceptibility and allergy phenotypes in 1252 unrelated non-Hispanic White PwMS (objective 4).

Allergy phenotype	HLA-DRB1:15:01 OR (95% CI)	GRS OR (95% CI)
Seasonal allergies	1.02 (0.85, 1.22)	0.99 (0.98, 1.00)
Seasonal allergies before 18	1.10 (0.88, 1.38)	0.98 (0.97, 0.99)b
Asthma	0.99 (0.83, 1.19)	1.00 (0.99, 1.02)
Asthma before 18	1.10 (0.77, 1.56)	1.02 (1.00, 1.04)
Seafood	0.92 (0.53, 1.59)	0.97 (0.94, 1.01)
Glutena	3.58 (1.98, 6.47)c	1.07 (1.01, 1.13)d
Dairy	1.06 (0.74, 1.52)	0.98 (0.95, 1.00)
Eggs	1.64 (0.87, 3.08)	1.06 (0.98, 1.16)
Nuts	1.39 (1.04, 1.86)d	1.03 (0.99, 1.06)
Latex	1.31 (0.69, 2.48)	0.98 (0.94, 1.02)
Pets	1.16 (0.60, 2.24)	0.98 (0.95, 1.03)
Soy	0.55 (0.14, 2.05)	0.94 (0.97, 1.00)

Abbreviations: GRS = genetic risk score; OR = odds ratio; CI = confidence interval.

The main effects for HLA-DRB1*15:01 and GRS adjusting for age, sex, education, smoking status, and genetic ancestry, with clustering of the standard errors for account for possible dependency by recruitment site.

^aAlso adjusted for self-reported diagnosis of celiac disease.

^bUncorrected two-sided p = 0.0015 and significant after FDR-BH correction.

^cUncorrected two-sided p = 0.000025 and significant after FDR-BH correction.

^dUncorrected two-sided p < 0.05, but not significant after FDR-BH correction.

Discussion

This study provides new insight into the immunoepidemiology of ADs in MS. We observed a lower prevalence of seasonal allergies among US PwMS relative to unaffected controls, with the more pronounced differences for childhood-onset seasonal allergies. Among PwMS, those with PPMS trended to have a lower prevalence of seasonal allergies than those who presented with RRMS, though it was not significant after adjusting for multiple testing. Genetic analyses revealed that MS genetic risk, HLA-DRB1*15:01 and a non-MHC GRS, were associated with specific ADs among PwMS.

Our findings align with prior studies reporting an inverse association between MS and specific ADs. The key significant associations include inverse relationships between MS and asthma in a 1995–1999 Welsh study, and atopic allergy, allergic asthma, and allergic rhinitis in two early aughts Italian studies.^3–5 In the US pediatric MS study from the early 2010s, 74 cases had a low prevalence of environmental allergies compared to 132 controls (13% vs 20%), while the difference was not statistically significant (p < 0.1), the trend is consistent with our findings.⁶ And while our inverse association between MS and asthma was not significant, the direction of effect aligns with the prior Welsh study. There is, however, a Danish study that failed to observe an association between IgE positivity against inhalant allergens and MS incidence.²⁶ Although difficult to interpret, those results did not account for study heterogeneity, which was a likely source of bias as data was pooled from five studies that measured IgE positivity for different antigens, using different assays, and in samples collected in non-overlapping time intervals.²⁶ It is not clear whether our reduced prevalence of seasonal allergies in PwMS compared to controls is primarily due to differences in immunopathologies, such as Th1/Th17-skewness in MS versus Th2-skewed IgE-mediated hypersensitivities.^1,2 Another plausible mechanism might be differences in B cell isotype class switching mechanisms (such as from IgM or IgD to IgE) and subsequent IgE antibody response,²⁷ between PwMS and controls. For example, direct class switching can lead in the production of lower affinity IgE antibodies, while sequential class switching results in higher affinity IgE antibodies.²⁸ However, there is limited research on isotype class switching in MS and merits further inquiry. In addition, the hygiene hypothesis would not explain this inverse relationship as decreased early exposure to infections is hypothesized to be a shared mechanism that may explain the increasing incidence in autoimmune diseases and ADs.²⁹

We also observed a non-significant trend for a lower seasonal allergy prevalence in those presenting with PPMS compared to RRMS (uncorrected p < 0.05). In post hoc analyses of the current data, comparing PPMS and secondary progressive (SP) MS to RRMS separately, the trend for PPMS versus RRMS remained unchanged from those reported in Table 4, while SPMS did not differ from RRMS (OR = 0.8 (95% CI: 0.6, 1.2); uncorrected p > 0.25). This suggests that if a relationship was underpowered to detect, it may reflect differences in initial disease course. This hypothesis is tentatively supported by post hoc analyses of a US study of PwMS that reported findings for: 28 progressive at onset cases with environmental allergies (pollen, grass, tree, dust, mite, pets, or other), 32 progressive at onset cases with no known allergies, 558 relapsing at onset with environmental allergies, and 395 relapsing at onset cases with no known allergies.⁷ These numbers result in an unadjusted prevalence ratio of 0.64 (95% CI: 0.39, 1.04; p = 0.07). Together, these two sets of findings suggest that there may be a lower prevalence of seasonal allergies in progressive at onset cases—however, confirmation is required.

Genetically, the primary MS risk allele, HLA-DRB1*15:01, was associated with a higher prevalence of gluten allergy/intolerance among PwMS. There has been little research into the genetic underpinning of non-celiac gluten sensitivity;³⁰ thus, we could not find evidence to support or refute this finding in PwMS. A systematic review of gluten-related antibodies in PwMS found inconsistent results across several modestly sized studies that varied in methodology—and none considered genetic variants.³¹ Our association was independent of celiac disease (which had a prevalence 1% in all PwMS; 28% of PwMS reporting a gluten allergy/intolerance in this study), which is driven by the HLA-DQ2.5 haplotype (comprised of HLA-DQA1*05:01 and HLA-DQB1*02:01 and present in 90% of celiac cases) or HLA-DQ8.³⁰ We examined linkage disequilibrium between 28 MHC risk variants for celiac disease reported in the GWAS Catalog (https://www.ebi.ac.uk/gwas/ ) and rs3135388 (tagging SNP for HLA-DRB1*15:01) in 1000 Genomes European samples (eTable). Several of these celiac disease risk SNPs had non-random associations with rs3135388, which is expected given the complex linkage disequilibrium structure of the MHC. However, while a few SNPs showed high D’, r2 was extremely low (< 5%), suggesting our observed association between HLA-DRB1*15:01 and gluten allergy/intolerance in PwMS is not due to correlations with known celiac disease MHC risk alleles and haplotypes, but instead may reflect an independent immunogenetic signal specific to MS-related mechanisms.

The significant inverse relationship between the GRS and childhood-onset seasonal allergies suggest pleiotropic effects of non-MHC MS risk variants in PwMS. A similar non-significant trend was also present for seasonal allergies irrespective of onset age (uncorrected p = 0.1). This finding, in part, supports our key finding for Objective 1—the presence of an inverse relationship between seasonal allergies and MS. Notably, a recent genetic study of allergic rhinitis did not observe a genetic correlation with MS, but did find a significant enrichment of MS-related genes and MS-relevant biological processes, including Th17 cell differentiation, JAK-STAT signaling pathway, Epstein-Barr virus infection, and others, among genes associated with allergic rhinitis.³¹ These results, therefore, also suggest an independent immunogenetic signal specific to PwMS. Several of these overlapping loci have diverse immunological function, such as regulating IgE class switching, Th2 differentiation, and mast cell homeostasis.^32,33 Therefore, it is yet plausible to hypothesize that alleles conferring risk for MS may also have pleotropic functions modifying risk for seasonal allergies in PwMS.

There are strengths and limitations of the current study. First, this study leverages a relatively large U.S. case-control study (in contrast to prior studies) included genetic data and detailed questionnaire data, that explicitly included questions on seasonal allergies and asthma, with six open-ended prompts to report dietary and other allergies, as well as the age at onset of these ADs. Our findings add to those reported in earlier European studies. Multiple likely confounders were adjusted for, including birth year, education, smoking status, and geography. This is the first study evaluating allergic endophenotypes by MS clinical course, that adjusts for key confounders, and investigates genetic relationships for MS genetic susceptibility and specific allergic endophenotypes in PwMS. The primary limitation is that many allergic endophenotypes were not common and therefore absence of a relationship may have been a function of limited statistical power. Second, a few allergic definitions may have been broad (i.e. seafood allergies), which may have led to misclassification. Third, these data are based on self-report which raises the concern for recall bias, however there is little evidence as we would have hypothesized PwMS to have over-reported allergic phenotypes. Fourth, the representativeness of cases and controls is uncertain. The CDC estimates that 25% of U.S. adults have allergic rhinitis,³⁴ lower than among controls in this study, though allergy prevalence varies by region and climate. Cases were recruited from tertiary centers and may not reflect the general MS population.

Participation rates were unavailable, and diagnostic updates beyond neurologist confirmation at enrollment were not possible, raising the possibility that a few cases may reflect myelin oligodendrocyte glycoprotein antibody-associated disease. Fifth, we did not account for disease modifying therapies which would have been correlated with MS status; however, we did investigate childhood-onset seasonal allergies which would be in the time interval before the onset of MS and observed as consistent association. Finally, the genetic analyses were restricted to NHW PwMS, it is unknown if similar relationships might exist in other ethnoracial populations with MS.

Conclusion

This study provides novel evidence of an inverse association between MS and seasonal allergy prevalence in the United States; it suggests there may be lower prevalence in PPMS compared to RRMS, and the genetic analyses revealed that both HLA-DRB1*15:01 and non-MHC MS risk alleles modulate AD endophenotype susceptibility. Our findings demonstrate that detailed phenotypic and genetic data should be essential components of future studies. Considering the pivotal role of B cells in both MS and ADs, a compelling avenue for future studies would be to examine whether differential class switching underlies the observed differences between PwMS and controls, and among PwMS with varying combinations of genetic risk, as it may illuminate overlapping yet divergent pathways between MS autoimmunity and atopy.

Supplementary Material

Supplemental material for this article is available online.

Data Availability

Data are available from the Accelerated Cure Project for Multiple Sclerosis (www.acceleratedcure.org).