Genomewide association study of a homogeneous multiple sclerosis cohort: tumefactive demyelination

Abstract

Background:

Tumefactive demyelination (TD) is a rare variant of multiple sclerosis (MS) characterized by tumor-like lesions that often require aggressive management. Genomewide association studies (GWAS) identified variants associated with MS; similar analyses in TD are lacking.

Objective:

A GWAS was performed to identify variants associated with TD.

Methods:

The case-control study included 142 TD cases and 293 controls. TD patients were required to have a demyelinating event and MRI showing one or more lesions. Controls were patients without a neurologic or systemic inflammatory disease or cancer. Logistic regression was used to compare cases versus controls for each variant; age, sex, and principal components were included as covariates. A p-value threshold of 5×10⁻⁸ was GWAS significant and 5×10⁻⁶ nominally significant. A polygenic risk score (PRS) was compared across TD and controls.

Results:

Variants on chromosome 14 (rs117797734, p=2.06x10⁻¹¹, OR=13.14) and chromosome 6 (most significant rs6936540, p=5.5x10⁻⁷, OR=2.61) near DCBLD1 were significant. Seven non-MHC and two MHC variants associated with MS were associated with TD. The PRS was significantly higher in TD versus controls.

Conclusions:

We identified novel regions associated with TD, demonstrating the importance of performing GWAS in homogenous subtypes of MS. Further validation and functional experiments are necessary.

INTRODUCTION

Genomewide association studies (GWAS) have identified germline variants associated with MS susceptibility^1-6, and led to implication of CNS proteins in MS severity⁷. The largest published GWAS of MS validated 233 variants: 32 variants in the major histocompatibility complex (MHC), 200 autosomal non-MHC variants, and 1 variant on chromosome X ³. These GWAS to date have treated MS as a single disease, with one exception. A GWAS published in 2009 evaluated age of onset, multiple sclerosis severity scale, brain parenchymal volume, and T2 lesion load amongst MS patients diagnosed as relapsing remitting or secondary progressive⁴. Tumefactive demyelination (TD) is a rare subtype of demyelinating disease characterized by a large, tumor-like lesion in the CNS that often requires aggressive management⁸, occurs in 1.9% of individuals with MS⁹, and has an age-adjusted annual incidence rate of 0.56/100,000. It has been demonstrated in cancer that performing GWAS within homogeneous subtypes can identify variants with clinically relevant odds ratios (OR)¹⁰. With improved diagnostics of MS subtypes, GWAS should also be performed within homogeneous cohorts of MS¹¹. We present the first GWAS of TD, with the goal of understanding the molecular mechanisms associated with the development of TD.

METHODS

Participants

This study was approved by the local Institutional Review Board. All patients consented to use their medical records for research. Patients were identified through subspeciality MS Clinics; the year of disease onset ranged between 1988-2020. All patients underwent a clinical assessment by a fellowship trained MS neurologist who confirmed the diagnosis of TD. Inclusion required a demyelinating event and a brain MRI showing one or more demyelinating lesions (minimum transverse diameter ≥10 mm). Non-MS controls were identified from the Mayo Clinic Biobank, which is a prospective biobank open to all Mayo Clinic patients. Controls were defined by excluding individuals with any neurologic or systemic inflammatory disease or cancer and were matched by age and sex to a larger set of patients in the parent study and were not matched to the TD cohort specifically.

Methods

Patients were genotyped on the Affymetrix Axiom^™ Precision Medicine Diversity Array Plus. Quality control included testing for Hardy–Weinberg equilibrium on the non-MS controls (HWE; p-value < 10⁻⁶), duplicate and relatedness checks, sex checks, variant call rates (>95%), and subject call rates (>95%). Imputation was performed using the TOPMed Imputation Server, utilizing the HG38 reference. STRUCTURE and 1000 Genome data¹² were used to determine racial groups; analyses were limited to subjects that were at least 75% European ancestry. Population stratification was evaluated using Eigenstrat and principal components significantly associated with case-control status (p-value < 0.05) were included as covariates in the logistic model, which allows for adjustment of population stratification to limit false positive results. Logistic regression was used, and age, sex, and the first two principal components included as covariates. A p-value threshold of 5×10⁻⁸ was considered GWAS significant and 5×10⁻⁶ was considered nominal significant. Analyses stratified by sex were also performed; z-statistic was used to test for a difference in OR between males and females.

A MS-based polygenic risk score (PRS) was estimated using an additive logistic regression model and the published coefficients of the non-MHC variants from a large MS GWAS (Supplementary Table S2¹³). The PRS was calculated on 142 TD patients and 293 non-MS controls. A TD-based PRS was estimated using the same set of variants; however, using the estimated coefficients from the TD GWAS. Both the MS- and TD-based PRS were standardized to zero mean and unit variance using the 293 non-MS controls. Logistic regression was subsequently used to test for a significant difference in PRS between TD cases and normal controls.

RESULTS

A total of 142 TD patients (140 with tumefactive MS and 2 with MOGAD) and 293 controls with at least 75% European ancestry were included in the analyses. TD patients were 56% female and the median age at diagnosis was 41 years. Controls were 46% female, and the median age was 53 years. Additional clinical and demographic characteristics for the TD cohort are available in Table 1.

Table 1:

Description of non-MS normal controls and tumefactive demyelination (TD) cases.

	Controls (N=293)	TD (N=142)
Sex, n (%)
Female	135 (46.1%)	79 (55.6%)
Male	158 (53.9%)	63 (44.4%)
Age (year)
Median	53	41
Range	21, 90	18, 83
Duration of index attack (Days), n (%)
< 2 Weeks		7 (4.9%)
2-6 Weeks		30 (21.1%)
6-12 Weeks		25 (17.6%)
> 12 Weeks		44 (31.0%)
Unknown		36 (25.4%)
Index MRI lesion types, n (%)
Solitary lesion		85 (59.9%)
Any Balo Lesion		17 (12.0%)
Multiple tumefactive lesions		40 (28.2%)
Butterfly lesion, n (%)
No		126 (88.7%)
Yes		16 (11.3%)
Unique CSF oligogoclonal bands
Median (IQR)		1.0 (0.0, 3.0)
Range		0.0, 14.0
Missing data		62
2+ Unique CSF oligoclonal bands, n (%)
No		45 (56.3%)
Yes		35 (43.8%)
Missing data		62

Abbreviations: CSF: cerebrospinal fluid; IQR: interquartile range; TD: tumefactive demyelination;

A GWAS was performed on the142 TD patients and 293 controls (Supplementary Table S1). One variant on chromosome 14, rs117797734, reached GWAS significance (p=2.06x10⁻¹¹, OR=13.14) (Figure 1; Table 2). The variant had a high allele frequency in TD cases (19.2%) and low frequency in the controls (2.6%); the control frequency aligns with the European 1000 Genome frequency (1.2%)¹². The variant is in a gene desert region with high recombination (Figure 2a). No significant eQTLs in normal brain cortex, whole blood or EBV-transformed lymphocytes were observed within 5 Mb of rs117797734 (GTEx portal, data not shown). Eighteen variants had a nominal association with TD. Seventeen were in or near the discoidin, CUB and LCCL domain containing 1 (DCBLD1) gene on chromosome 6. The most significant variant was rs6936540 (p=5.5x10⁻⁷, OR=2.61) (Table 2; Figure 2b). The DCBLD1 region was GWAS significant in the discovery stage of a previously published MS GWAS; however, it was not validated³. The variant observed in our TD GWAS (rs6936540) is in high linkage disequilibrium (D’=0.95) with the variant identified in the previous MS GWAS (rs62433107). The TD variant had a significant positive correlation (false discovery rate (FDR)<5%) with DCBLD1 expression in CD14⁺ monocytes and in peripheral blood mononuclear cells (PBMCs) from subjects with relapsing remitting MS³. A GWAS was performed stratified by sex and no variants reached GWAS or nominal significance when comparing the OR between males and females (Supplementary Table S1).

Figure 1:

Manhattan plot demonstrating GWAS results comparing TD cases versus controls. The red line denotes genomewide significance of 5x10⁻⁸ and the blue line denotes a nominal p-value of 5x10⁻⁶.

Table 2:

Variants associated with risk of tumefactive demyelination (TD).

Chrom	Position (GRCh38/hg38)	Variant	Gene	A2	A1	A1 allele TD freq	A1 allele Control freq	OR (95% CI)	P-value	Genotyped	R2
Novel variants identified in the TD GWAS
14	97817989	rs117797734		C	T	0.192	0.026	13.14 (6.19,27.90)	2.06x10−11	Imputed	0.85
6	117463884	rs6936540	DCBLD1	A	G	0.331	0.184	2.61 (1.79,3.79)	5.54x10−7	Imputed	0.99
Validation of non-MHC GWAS results that were identified and validated by the International Multiple Sclerosis Genetics Consortium (IMSGC; Science 365, eaav7188 (2019)).
17	73335776	rs9900529	GRB2	G	C	0.278	0.194	1.76 (1.23,2.51)	0.0021	Imputed	0.98
5	35877505	rs10063294	IL7R	G	A	0.470	0.567	0.66 (0.48,0.89)	0.0069	Imputed	0.99
4	122119449	rs17051321	TNIP3	T	C	0.706	0.777	0.65 (0.46,0.92)	0.0140	Imputed	0.98
17	38252660	rs883871	NR1D1	A	G	0.819	0.875	0.61 (0.40,0.93)	0.0210	Imputed	0.98
8	128814091	rs6990534	PVT1	G	A	0.259	0.317	0.67 (0.48,0.94)	0.0220	Genotyped	1.00
1	120267505	rs483180	PHGDH	C	G	0.315	0.367	0.72 (0.52,0.99)	0.0430	Genotyped	1.00
1	65429319	rs72922276	JAK1	A	G	0.930	0.896	1.76 (1.02,3.04)	0.0440	Genotyped	1.00
Validation of MHC results that were identified and validated by the International Multiple Sclerosis Genetics Consortium (IMSGC; Science 365, eaav7188 (2019)).
6	32619077	rs9271366	DRB1/DQA1	A	G	0.250	0.137	2.23 (1.48,3.35)	0.00012	Genotyped	1.00
6	31572980	rs2229092	LTA	A	C	0.088	0.043	1.99 (1.09,3.61)	0.024	Genotyped	1.00

Abbreviations: Chrom=chromosome; CI=confidence interval; freq=frequency; OR= odds ratio

Figure 2:

Locus zoom plots for the genomic region surrounding (a) rs117797734 on chromosome 14 and (b) rs6936540 on chromosome 6.

We also evaluated the 200 autosomal non-MHC and 32 MHC variants that were previously shown to be associated with risk of MS³ (Supplementary Table S2 and S3, respectively). Seven non-MHC variants and two MHC variants had p<0.05 and an OR that was in the same direction (Table 2).

Using published coefficients from a large MS GWAS¹³, a MS-based PRS was calculated on 142 TD patients and 293 non-MS controls, demonstrating that the TD cohort had a significantly larger PRS (Figure 3a). The OR per standard deviation of the PRS in TD versus controls was 1.55 (95% CI: 1.26-1.92, p<0.001). A TD-based PRS was calculated using the same variants; however, using the estimated coefficients from the TD GWAS (Figure 3b). The OR per standard deviation of the PRS in TD versus controls was 5.39 (95% CI: 3.97-7.60, p<0.001).

Figure 3:

Standardized polygenic risk score (PRS) for non-MS normal control versus tumefactive demyelination patients, calculated using the coefficients from (a) the large, published MS GWAS, or (b) using the same variants, however, using coefficients estimated from the case-control analysis of tumefactive demyelination versus controls.

DISCUSSION

TD is commonly the first symptomatic attack of MS or MOGAD and can be disabling or fatal. Molecular mechanisms leading to the severe initial attack may give insights into the heterogeneity in disease activity. The TD GWAS identified one variant that reached genomewide significance (rs117797734) and is ~900 kb from the nearest protein coding gene and ~29 kb from the nearest non-protein coding gene. This suggests that rs117797734 may be participating in long-range interactions as an enhancer or super-enhancer for a distant gene. The variant is in a region with a high rate of recombinations; thus, imputation will not perform well in the region and could explain why only a single variant was observed. The imputation quality for rs117797734 appears adequate; the imputation R2 was 0.84 and the observed allele frequency for in the 293 non-MS controls was similar to what is reported by 1000 Genomes¹². In silico analyses identified candidate transcription factors: HDX, RARB, RARG, GCM1, GTF2B, and TEAD3. The region was not identified in large published MS GWAS^{3, 13}; thus, the association may be specific to TD and requires validation in an independent cohort of TD patients.

We also identified eighteen nominally significant variants in or near DCBLD1. While this region was identified in the discovery stage of a large MS GWAS (14,802 MS and 26,703 controls), it was not subsequently independently validated using 20,360 MS and 19,047 controls³. The variant (rs62433107) identified previously had an OR=1.04, modeling risk of the G-allele³. This previous study combined relapse-remitting and progressive MS and did not specify if MOGAD patients were included. The MOG-IgG testing has only been widely available since 2017 and the MOGAD diagnostic criteria was published in 2023¹⁴. We identified eighteen DCBLD1 variants as nominally significant in the TD GWAS. The DCBLD1 region was identified when analyzing only 142 TD cases. This sample size is one hundred times smaller than the previous study on MS³, demonstrating the importance of performing GWAS in more homogenous subtypes. The TD variant (rs6936540) had an OR=2.61, signifying that variants with potentially clinically relevant effect sizes can be identified when analyzing homogeneous cohorts. DCBLD1 is a scaffolding transmembrane protein, and biochemical pathways are largely undefined. Overexpression of DCBLD1 leads to proliferation and metastasis in cervical cancer¹⁵, and is associated with poor outcomes in many cancers^16-19. Potential mechanisms include upregulation of the pentose phosphate pathway and integrin signaling pathway. DCBLD1 maps to chromosomal region 6q22.1, immediately between and adjacent to ROS1 and GOPC. ROS1 is frequently amplified or rearranged in many cancers²⁰ and ROS1 is often activated by rearrangements with GOPC, especially in brain tumors²¹. It is possible that the DCBLD1 locus may be involved in the activation of ROS1 and/or GOPC in TD. For instance, a common feature of TD and glioblastoma is the presence of multinucleated Creutzfeldt astrocytes^22-24. Further studies are needed to understand the role of DCBLD1 in TD.

We also evaluated the 200 autosomal non-MHC and 32 MHC variants that were previously shown to be associated with risk of MS³. Seven non-MHC variants were associated with risk of TD with a p-value<0.05. Of the 32 MHC variants, 13 were available in the TD GWAS, and two were significantly associated with TD with a p-value<0.05. Previous GWAS studies on MS identified variants in the MHC region as GWAS significant. Even though imputation was performed using TOPMed, which included the MHC region, the TD GWAS did not identify variants in the MHC region at either a GWAS significance or nominal significance level. However, variants in this region were observed if the p-value threshold was reduced further (Supplementary Table S1). Particularly, rs3135388 is a tag single nucleotide polymorphism (SNP) for the primary MS risk variant HLA-DRB1*15:01^{5, 25, 26}. In the TD GWAS, rs3135388 had an OR=2.13 (p=0.00041), which agrees with the OR observed in MS patients²⁵. Future work should include evaluating this region more carefully in a larger cohort of TD patients.

Using coefficients estimated from a large MS GWAS¹³, we calculated an MS-based PRS on our cohort of TD and non-MS controls. The PRS in the TD cohort had a statistically significantly larger PRS than the non-MS controls. The OR in TD versus controls (OR=1.55) is in alignment with published results for MS versus controls (OR=1.70)¹³. These ORs are also in alignment with cancer studies, such as breast cancer, where the OR per standard deviation was 1.71 across all breast cancers, 1.81 in ER-positive, and 1.48 in ER-negative breast cancer^{27, 28}. We also calculated a TD-based PRS using the same variants; however, using the estimated coefficients from the TD GWAS. This resulted in a larger OR (OR=5.39), demonstrating that some of the MS variants have larger effect sizes in the more homogeneous subset of TD patients.

Our study has limitations. Due to sample size restriction, we were only able to analyze individuals who were primarily European ancestry. Additionally, we did not have a validation set and thus it is important to acknowledge that all results require future validation, which will likely require a consortium effort due to the rareness of TD. Future experiments are necessary to finemap and identify functional mechanisms of the candidate variants. While sex-stratified analyses were performed, they had minimal power due to the small sample size.

We performed the first GWAS on TD. By evaluating a more homogeneous subtype of demyelination, two novel candidate variants were identified. Furthermore, we observed that seven non-MHC and two MHC variants originally identified in MS were also associated with risk of TD. Such observations can help to understand the mechanisms leading to the severe initial attack of TD.

Data availability:

The GWAS data are available in dbGaP (phs003806.v1.p1).