Polygenic Prediction of Cellular Immune Responses to Mumps Vaccine
==================================================================

* Brandon J. Coombes
* Inna G. Ovsyannikova
* Daniel J. Schaid
* Nathaniel D. Warner
* Gregory A. Poland
* Richard B. Kennedy

## Abstract

In this report, we provide a follow-up analysis of a previously published genome-wide association study of host genetic variants associated with inter-individual variations in cellular immune responses to mumps vaccine. Here we report the results of a polygenic score (PGS) analysis showing how common variants can predict mumps vaccine response. We found higher PGS for IFNγ, IL-2, and TNFα were predictive of higher post-vaccine IFNγ (p-value = 2e-6), IL-2 (p = 2e-7), and TNFα (p = 0.004) levels, respectively. Control of immune responses after vaccination is complex and polygenic in nature. Our results suggest that the PGS-based approach enables better capture of the combined genetic effects that contribute to mumps vaccine-induced immunity, potentially offering a more comprehensive understanding than traditional single-variant GWAS. This approach will likely have broad utility in studying genetic control of immune responses to other vaccines and to infectious diseases.

Keywords
*   polygenic scores
*   mumps
*   mumps vaccine
*   measles-mumps-rubella vaccine
*   cellular immunity
*   cytokines
*   chemokines

## Introduction

Understanding mumps immunity is of paramount importance, especially considering the recent increase in mumps outbreaks despite widespread vaccination efforts1. The resurgence of mumps in the United States, even among highly immunized populations, underscores the need for comprehensive research to understand the driving factors behind vaccine-induced immunity such that better control this disease results. Our previous work, a genome-wide association study (GWAS) conducted by Ovsyannikova et al. in 2023, shed light on the host genetic variants associated with inter-individual variation in mumps vaccine-induced cellular immune responses2. This GWAS provided crucial insights into the genetic factors influencing mumps vaccine-induced immunity. Work done by our lab and others with mumps and other viral pathogens has demonstrated that vaccine-induced immune responses are multigenic3,4. Polygenic scores (PGS) summarize the estimated effects of many genetic variants on an individual’s phenotype, thereby offering a powerful approach to comprehensively assess the genetic basis of complex traits. PGS have been used to examine COVID-19 disease severity5,6 and risk of infection-induced sequelae7. The approach has also been proposed as an enrichment strategy to lower costs and increase success in clinical trials8. Surprisingly, PGS have never been applied to immunogenetic studies of vaccine responsiveness. We employed PGS to evaluate the genetic determinants underlying mumps vaccine-induced cellular immune responses in the same cohort previously studied by us2. This novel approach enables capture of combined genetic effects that contribute to mumps immunity, potentially offering a more comprehensive understanding than traditional single-variant GWAS. This approach holds the potential to significantly enhance our understanding of vaccine-induced immune responsiveness and could be used to reverse engineer new vaccine candidates based on a fuller understanding of immune response mechanisms9.

## Methods

The immune assays described herein are identical to those described in our previous vaccine studies10-12 and as used in our recent published GWAS2.

### Study Subjects

Our study cohort was a sample of 1,406 healthy children, older adolescents, and healthy adults selected from two cohorts with existing genotyping data from prior immunogenetics studies10,11,13-15. The demographic and clinical characteristics of these cohorts have been previously published10,11,13,16. Briefly, the cohort consisted of 1,406 healthy subjects (35.8 % females, 94.4 % European) with a mean time from last measles-mumps-rubella (MMR) vaccination to blood draw of 5.3 years (25 % and 75 % interquartile range/IQR 0.3, 15.5), and age at enrollment of 19.5 years (IQR 15.0, 23.0). The Institutional Review Boards of the Mayo Clinic (Rochester, MN) and the NHRC (Naval Health Research Center, San Diego, CA) approved the study, and written informed consent was obtained from each subject.

### Cytokine and Chemokine Response Measurements

We measured cytokine (IL-2, IL-6, IL-10, IFNα2a, IFNγ, IL-1β, and TNFα) and chemokine (IP-10, MCP-1, MIP1α, and MIP1β) responses from cultured PBMCs stimulated *in vitro* with mumps virus, as described and previously reported12. Here we focused on the five immune response outcomes for which we had previously found significant genome-wide SNP associations: IL-2, IFNγ, IL-1β, TNFα, and MCP-1. The coefficient of variation (CV) for these assays varied from 10% to 31% depending on the cytokine/chemokine.

### Genotyping and Imputation

The cohorts were genotyped using the Illumina Omni 1M array for school-age subjects recruited from the Rochester, MN area (n=964), and using the Illumina 550 and 650 SNP arrays (for the European and African-American subjects, respectively) for young adults recruited from the San Diego, CA area (n=807) 10,11,14. Each genotyping batch was filtered using standard quality control filters as described in Ovsyannikova et al. 2023 and then imputed using the TOPMed imputation reference panel17 retaining only well-imputed SNPs (dosage-R2 > 0.8) with MAF > 0.01.

Our initial GWAS of mumps-specific cytokines/chemokines was focused on the n=1,406 (903 males and 503 females) subjects with European ancestry (i.e., Rochester cohort, n=748 subjects, San Diego cohort, n=658 subjects). The covariates that were significantly associated with cytokine/chemokine levels were the following: age at last MMR vaccination; cohort (San Diego vs. Rochester); sex; age at blood draw; year of blood draw and the first 10 principal components (PCs) that were constructed from the genome wide SNP arrays. These associated covariates were used as independent variables in linear models with quantile-normalized cytokine/chemokine trait as dependent. The “residualized” cytokines from these models were then used as the dependent in each PGS association test.

### Polygenic Score (PGS) Analyses

We used a “leave-one-out” (LOO) approach to test whether the genetic prediction of cytokine response can predict observed cytokine response. In this LOO approach, GWASs were repeated as described in Ovsyannikova et al. (2023)2 except they were performed separately in the independent Rochester and San Diego cohorts using the same models. We then generated PGS within each cohort using the GWAS summary statistics from the other cohort. These LOO PGSs were constructed for each cytokine using PRSice218. To account for linkage disequilibrium (LD) among SNPs, clumping was performed (clump-r2 0.1 clump-kb 250). We evaluated a series of p-value thresholds (*p**T* = 5e-7, 1e-7, 1e-6, 1e-5, 0.0001, 0.001, 0.01). Each PGS was then standardized. The PGS for each cytokine represents a genetically predicted cytokine. In each cohort, the genetically predicted cytokine (PGS) constructed with varying p-value thresholds was tested for association with the observed residualized cytokine using linear regression while adjusting for the first two principal components (PCs) of ancestry. The results from each LOO PGS analysis were then meta-analyzed across cohorts using the meta R package using a fixed-effect meta-analysis. In some cases, the PGS calculated under a stricter threshold (e.g., genome-wide significant variants only; p<5e-8) for one cohort was not possible to be estimated in the other cohort as no variants fit that threshold. In these cases, the results from only one cohort was reported for that p-value threshold.

## Results

### Study Sample Characteristics

The analyzed cohort (n=1,406) consisted of subjects with European (EUR, 94.4%) or Asian (5.6%) ancestries. The sex distribution was 903 males (64.2%) and 503 females (35.8%). The average age at enrollment was 19.5 years. Additional demographic details can be found in Ovsyannikova et. al.2 The current study focused on five immune outcomes: IL-2, IFNγ, IL-1β, TNFα, and MCP-1 (See Table 1) as these were the outcomes for which we had previously identified significant genome-wide SNP associations (IL-2, IFNγ, IL-1β, TNFα) or came close to the 5 × 10−8 threshold (MCP-1).

View this table:
[Table 1.](http://medrxiv.org/content/early/2024/02/27/2024.02.23.24303277/T1)

Table 1. Cytokine and Chemokine Responses in the Study Cohort (n=1,406)

### Polygenic Score Analysis Results with Mumps Cellular/Inflammatory Immune Responses

We first performed GWASs of vaccine outcomes (IFNγ, IL1β, IL-2, MCP1, and TNFα) separately in the cohort from Rochester (N = 728) and the cohort from San Diego (N = 599). In a LOO approach, we tested for association of LOO-PGS constructed within each of those cohorts using the oppositive cohort’s GWAS summary statistics and meta-analyzed the results from each cohort. Figure 1 shows the average variance explained by each LOO PGS for each cytokine at various p-value thresholds. In the meta-analysis, IFNγ (best *p**T* = 5e-8; *β*=0.16 (0.04); R2=3.1%; p-value = 9e-6), IL-2 (best *p**T* = 1e-5; *β*=0.13 (0.03); R2=2.1%; p-value = 2e-7), and TNFα (best *p**T* = 1e-6; *β*=0.11 (0.02); R2=1.2%; p-value = 0.004) were each significantly predicted by their respective PGS.

![Figure 1.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2024/02/27/2024.02.23.24303277/F1.medium.gif)

[Figure 1.](http://medrxiv.org/content/early/2024/02/27/2024.02.23.24303277/F1)

Figure 1. 
Bar plots depicting the average variance explained in each residualized cytokine outcome across cohorts when using a LOO PGS for each cytokine with different p-value thresholds of inclusion of variants. The meta-analysis p-value for association is annotated above each bar.

We noted weak signals for IL1β and MCP-1, suggesting that the genetic contributions of the secretion of these factors is weaker than what is observed for the other factors. For IL-2 and TNFα, lowering the p-value threshold to include more SNPs in the analysis resulted in both an increase in variance explained and a lower meta-analysis p-value. This increased signal tapered off as the p-value threshold for including SNPs in the PRS increased to 0.0001. This suggests that there are a limited number of additional genetic factors influencing secretion of these factors (i.e., control of these cytokines is likely multi-genic but not poly-genic). For IFNγ, the genome-wide significance p-value threshold explained the most variation. However, the PGS was not calculated at this threshold (nor the p<1e-7 and 1e-6 thresholds) in the Rochester cohort because in the San Diego cohort no variant met these thresholds. Therefore, the strictest p-value threshold (p<1e-5) with available variants in both cohorts actually achieved the smallest meta-analysis p-value.

Our initial results suggested there are some SNPs below the genome-wide significance level that appear to improve prediction, therefore we examined the SNPs reaching the 1 × 10−6 threshold and looked up the gene most connected to each SNP according to the variant-to-gene (V2G) score (Table 2)19,20. For IFNγ, 6 additional variants (mostly in the HLA region) were identified as potentially contributing to the variance explained. For IL-2, 7 additional variants in the HLA region were identified. For TNFa, we identified an additional SIGLEC 5/14 variant as well as a variant in NCOR2 and a variant in ATXN7L3B.

View this table:
[Table 2.](http://medrxiv.org/content/early/2024/02/27/2024.02.23.24303277/T2)

Table 2. SNPs Associated with p < 1e-6 with Increased Outcome Variance Explained by PRS

## Discussion

In our cohort of 1,406 mumps vaccine recipients, we found that higher polygenic scores for IFNγ, IL-2, and TNFα were all predictive of higher post-vaccination cytokine responses to *in vitro* stimulation with inactivated mumps virus. The *in vitro* stimulation elicited a recall response from virus-specific T cells without the complications of viral infection, acting as a probe allowing us to evaluate the magnitude and quality of the T cell response that would be expected upon exposure to mumps virus. Furthermore, the PRS clearly indicates that including more variants than those that were genome-wide significant improved the strength of association. While widely used in some fields (e.g., psychiatry), our study demonstrates that PGS scores can also be calculated for vaccine immune response and that such scores can be associated with the same immune outcome in independent cohorts. The increased variation in outcome explained by the PRS indicate that there is additional power to including information from variants not reaching the standard genome-wide threshold. Thus, PRS approaches can supplement GWAS by providing additional information about genetic control of viral vaccine responses.

We have already reported genetic associations with HLA loci and mumps-specific IFNγ and IL-2 secretion2. Our PRS found additional polymorphisms in the same region (HLA) also contributed to the variation in IL-2 secretion; however, for IFNγ we identified not only additional HLA loci, but also SNPs near *FAM117B* and *RAPH1* genes that may potentially influence cytokine secretion. *FAM117B* (also known as *ALS2CR13)*, is a gene associated with the neurodegenerative disorders juvenile ALS (amyotrophic lateral sclerosis) and mast syndrome.21 RAPH1 encodes for an adapter protein that functions in cell migration. Interestingly, this gene is also associated with juvenile ALS. With respect to TNFα, our PRS identified one additional SNP in SIGLEC5/14 as well as a SNP in *ATXN7L3B* and a SNP in *NCOR2. ATXN7L3B* encodes for the Ataxin 7 like protein 3B, which regulates histone deubiquitination, while *NCOR* encodes for a nuclear receptor co-repressor that interacts with retinoic acid receptor. Immune-related functions for these genes have not been identified and further investigation will be necessary to determine what role they might play in the cytokine response to mumps vaccination.

To the best of our knowledge, this study represents the first use of use of polygenic scores to study genetic influences on cellular immune responses after vaccination. Our results indicate that this approach has promise, as it identified additional genetic variants potentially contributing to inter-individual variation in immune responses to mumps virus stimulation. These additional variants/regions may provide information regarding the genes/pathways involved in regulating the immune response and represent targets for functional studies to elucidate the biological mechanisms behind the statistical associations. It is likely that calculation of polygenic risk scores will complement existing GWAS approaches when investigating multigenic control of immune responses.

## Data Availability

All data produced in the present work are contained in the manuscript.

## Financial support

Research reported in this publication was supported by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health under award number R01AI-127365, R37AI-48793, and R01AI-33144. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

## Potential conflicts of interest

Dr. Poland is the chair of a Safety Evaluation Committee for novel investigational vaccine trials being conducted by Merck Research Laboratories. Dr. Poland provides consultative advice to AiZtech; Emergent, GlaxoSmithKline, Invivyd, Merck & Co. Inc., Moderna, Novavax, Syneos Health, and Valneva.

These activities have been reviewed by the Mayo Clinic Conflict of Interest Review Board and are conducted in compliance with Mayo Clinic Conflict of Interest policies.

Drs. Poland, Ovsyannikova and Kennedy hold patents related to vaccinia, influenza, and measles peptide vaccines. Drs. Poland, Ovsyannikova and Kennedy have received grant funding from ICW Ventures for preclinical studies on a peptide-based COVID-19 vaccine. This research has been reviewed by the Mayo Clinic Conflict of Interest Review Board and was conducted in compliance with Mayo Clinic Conflict of Interest policies.

Dr. Poland is an adviser to the White House and World Health Organization on COVID-19 vaccines and monkeypox, respectively. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

## Acknowledgments

We thank the Mayo Clinic Vaccine Research Group staff and the study participants.

*   Received February 23, 2024.
*   Revision received February 23, 2024.
*   Accepted February 27, 2024.


*   © 2024, Posted by Cold Spring Harbor Laboratory

The copyright holder for this pre-print is the author. All rights reserved. The material may not be redistributed, re-used or adapted without the author's permission.

## References

1.  1.Centers for Disease Control and Prevention. Mumps cases and outbreaks. December 8, 2022. [https://www.cdc.gov/mumps/outbreaks.html](https://www.cdc.gov/mumps/outbreaks.html).
    
    

2.  2.Ovsyannikova IG, Haralambieva IH, Schaid DJ, Warner ND, Poland GA, Kennedy RB. Genome-wide determinants of cellular immune responses to mumps vaccine. Vaccine 2023; 41(44): 6579–88.
    
    

3.  3.Kennedy RB, Ovsyannikova IG, Haralambieva IH, et al. Multigenic control of measles vaccine immunity mediated by polymorphisms in measles receptor, innate pathway, and cytokine genes. Vaccine 2012; 30(12): 2159–67.
    
    

4.  4.Pankratz VS, Vierkant RA, O’Byrne MM, Ovsyannikova IG, Poland GA. Associations between SNPs in candidate immune-relevant genes and rubella antibody levels: a multigenic assessment. BMC Immunol 2010; 11(1): 48.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1186/1471-2172-11-48&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=20923569&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F02%2F27%2F2024.02.23.24303277.atom) 

5.  5.Fritsche LG, Nam K, Du J, et al. Uncovering associations between pre-existing conditions and COVID-19 Severity: A polygenic risk score approach across three large biobanks. PLoS Genet 2023; 19(12): e1010907.
    
    

6.  6.Bountress KE, Bustamante D, Ahangari M, et al. The impact of the COVID-19 pandemic on alcohol use disorder symptoms: Testing interactions with polygenic risk. J Am Coll Health 2024: 1–6.
    
    

7.  7.Llucià-Carol L, Muiño E, Cullell N, et al. Genetic Architecture of Ischaemic Strokes after COVID-19 Shows Similarities with Large Vessel Strokes. Int J Mol Sci 2023; 24(17).
    
    

8.  8.Fahed AC, Philippakis AA, Khera AV. The potential of polygenic scores to improve cost and efficiency of clinical trials. Nat Commun 2022; 13(1): 2922.
    
    

9.  9.Poland GA, Ovsyannikova IG, Kennedy RB. Personalized vaccinology: A review. Vaccine 2017; 36(36): 5350–7.
    
    

10. 10.Ovsyannikova IG, Kennedy RB, O’Byrne M, Jacobson RM, Pankratz VS, Poland GA. Genome-wide association study of antibody response to smallpox vaccine. Vaccine 2012; 30(28): 4182–9.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.vaccine.2012.04.055&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=22542470&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F02%2F27%2F2024.02.23.24303277.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000305780400009&link_type=ISI) 

11. 11.Haralambieva IH, Ovsyannikova IG, Kennedy RB, et al. Genome-Wide Associations of CD46 and IFI44L Genetic Variants with Neutralizing Antibody Response to Measles Vaccine. Human Genetics 2017; 136(4): 421–35.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1007/s00439-017-1768-9&link_type=DOI) 

12. 12.Riggenbach MM, Haralambieva IH, Ovsyannikova IG, Schaid DJ, Poland GA, Kennedy RB. Mumps virus-specific immune response outcomes and sex-based differences in a cohort of healthy adolescents. Clin Immunol 2022; 234: 108912.
    
    

13. 13.Ovsyannikova IG, Haralambieva IH, Kennedy RB, et al. Impact of cytokine and cytokine receptor gene polymorphisms on cellular immunity after smallpox vaccination. Gene 2012; 510: 59–65.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.gene.2012.08.021&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=23009887&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F02%2F27%2F2024.02.23.24303277.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000313611700008&link_type=ISI) 

14. 14.Kennedy RB, Ovsyannikova IG, Pankratz VS, et al. Genome-wide genetic associations with IFN gamma response to smallpox vaccine. Human Genetics 2012; 131(9): 1433–51.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1007/s00439-012-1179-x&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=22661280&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F02%2F27%2F2024.02.23.24303277.atom) 

15. 15.Voigt EA, Haralambieva IH, Larrabee BL, Kennedy RB, Ovsyannikova IG, et al. Polymorphisms in the Wilms Tumor gene are associated with interindividual variations in rubella virus-specific cellular immunity after measles-mumps-rubella II vaccination. The Journal of Infectious Diseases 2018; 217(4): 560–6.
    
    

16. 16.Kennedy RB, Ovsyannikova IG, Pankratz VS, Haralambieva IH, Vierkant RA, Poland GA. Genome-wide analysis of polymorphisms associated with cytokine responses in smallpox vaccine recipients. Human genetics 2012; 131(9): 1403–21.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1007/s00439-012-1174-2&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=22610502&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F02%2F27%2F2024.02.23.24303277.atom) 

17. 17.Das S, Forer L, Schönherr S, et al. Next-generation genotype imputation service and methods. Nat Genet 2016; 48(10): 1284–7.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1038/ng.3656&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=27571263&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F02%2F27%2F2024.02.23.24303277.atom) 

18. 18.Choi SW, O’Reilly PF. PRSice-2: Polygenic Risk Score software for biobank-scale data. Gigascience 2019; 8(7).
    
    

19. 19.Ghoussaini M, Mountjoy E, Carmona M, et al. Open Targets Genetics: systematic identification of trait-associated genes using large-scale genetics and functional genomics. Nucleic Acids Res 2021; 49(D1): D1311–d20.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1093/NAR/GKAA840&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F02%2F27%2F2024.02.23.24303277.atom) 

20. 20.Mountjoy E, Schmidt EM, Carmona M, et al. An open approach to systematically prioritize causal variants and genes at all published human GWAS trait-associated loci. Nat Genet 2021; 53(11): 1527–33.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1038/S41588-021-00945-5&link_type=DOI) 

21. 21.Hadano S, Hand CK, Osuga H, et al. A gene encoding a putative GTPase regulator is mutated in familial amyotrophic lateral sclerosis 2. Nat Genet 2001; 29(2): 166–73.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1038/ng1001-166&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=11586298&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F02%2F27%2F2024.02.23.24303277.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000171374100022&link_type=ISI)