Impact of rare and common genetic variation in the interleukin-1 pathway on human cytokine responses

Background The interleukin (IL)-1 pathway is primarily associated with innate immunological defense and plays a major role in the induction and regulation of inflammation. Both common and rare genetic variation in this pathway underlies various inflammation-mediated diseases, but the role of rare variants relative to common variants in immune response variability in healthy individuals remains unclear. Methods We performed molecular inversion probe sequencing on 48 IL-1 pathway-related genes in 463 healthy individuals from the Human Functional Genomics Project. We functionally grouped common and rare variants, over gene, subpathway, and inflammatory levels and performed the Sequence Kernel Association Test to test for association with in vitro stimulation-induced cytokine responses; specifically, IL-1β and IL-6 cytokine measurements upon stimulations that represent an array of microbial infections: lipopolysaccharide (LPS), phytohaemagglutinin (PHA), Candida albicans (C. albicans), and Staphylococcus aureus (S. aureus). Results We identified a burden of NCF4 rare variants with PHA-induced IL-6 cytokine and showed that the respective carriers are in the 1% lowest IL-6 producers. Collapsing rare variants in IL-1 subpathway genes produces a bidirectional association with LPS-induced IL-1β cytokine levels, which is reflected by a significant Spearman correlation. On the inflammatory level, we identified a burden of rare variants in genes encoding for proteins with an anti-inflammatory function with S. aureus-induced IL-6 cytokine. In contrast to these rare variant findings which were based on different types of stimuli, common variant associations were exclusively identified with C. albicans-induced cytokine over various levels of grouping, from the gene, to subpathway, to inflammatory level. Conclusions In conclusion, this study shows that functionally grouping common and rare genetic variants enables the elucidation IL-1-mediated biological mechanisms, specifically, for IL-1β and IL-6 cytokine responses induced by various stimuli. The framework used in this study may allow for the analysis of rare and common genetic variants in a wider variety of (non-immune) complex phenotypes and therefore has the potential to contribute to better understanding of unresolved, complex traits and diseases. Supplementary Information The online version contains supplementary material available at 10.1186/s13073-021-00907-w.

For instance, activation of the inflammasome allows for cleavage and activation of caspase-1 0 5 1, with subsequent activation and release of pro-inflammatory cytokines IL-1β and IL-18.

0 6
Conversely, autophagy is an important process involved in cell homeostasis, but is also able 1 0 7 to directly inhibit the inflammatory response by removing inflammasome components and 1 0 8 damaged mitochondria(4). 1 0 9 1 1 0 Defects in IL-1 pathway signaling and its specific members have been linked to 1 1 1 various inflammation-mediated diseases(2, 5). Pro-inflammatory members of the IL-1 family, 1 1 2 e.g. IL-1β and IL-18, play an important role in a variety of (auto-)inflammatory or immune 1 1 3 2 8 0 individuals in two groups that were subjected to binary trait association. Specifically, for each 2 8 1 cytokine-stimulus combination the SKAToR was applied twice: 1) 1% highest cytokine 2 8 2 producers versus all other individuals ( TOP BT, Figure 1D.II), and 2) 1% lowest cytokine 2 8 3 11 producers versus all other individuals ( LOW BT; Figure 1D.III). In two cases, C. albicans-2 8 4 induced IL-1ß production low-producers and LPS-induced IL-6 production high-producers, 2 8 5 no distinctive categories could be created due to equal cytokine measurements at the 1% 2 8 6 cut-off, and as such the groups were extended to 7 and 9 respectively. We followed up on 2 8 7 associations based on >1 variant, that were nominally significant (unadjusted P-values) in 2 8 8 the continuous associations and recurrent in either TOP BT or LOW BT. In order to give meaning to our detected associations, we extracted the residual 2 9 2 (corrected for covariates age and gender) cytokine production from the SKAT null-model and 2 9 3 correlated those to the genotype categories, where applicable. For all plots, correlations and 2 9 4 cytokine levels mentioned from hereon, the cytokine production therefore has been 2 9 5 corrected for age and gender. In the case of set-based common variant associations we 2 9 6 correlated cytokine production to the three separate genotype categories; homozygous 2 9 7 reference, heterozygous, homozygous alternative, whereas in the case of set-based rare 2 9 8 variant associations we correlated cytokine production to only two categories; absence or 2 9 9 presence of one of the rare variants in the set.

0 0
Additionally, considering accumulating evidence for a role of non-coding genetic 3 0 1 variation in health and disease(46, 47), we followed-up on common variant associations by 3 0 2 using the publicly available genotype data from the FG500-cohort, generated with the 3 0 3 commercially available SNP chip Illumina HumanOmniExpressExome-8v1.0 (for further 3 0 4 details we refer to previously published work(18, 48)). We extracted all common variants 3 0 5 (based on cohort AF ≥ 5%) within NCBI RefSeq 'Whole Gene' gene regions and extended 3 0 6 the start position by 50kB(49) for the following sets: IL36A, IL38, IL-30s subpathway, pro-3 0 7 inflammatory phenotype and anti-inflammatory phenotype. Variant sets were pruned for 3 0 8 Linkage Disequilibrium (LD) based within cohort metrics and the commonly used R 2 cut-off 3 0 9 of >0.8, by means of the snpStats package in R. The final pruned set of variants, termed 3 1 0 tagSnps, were subjected to the same SKAT with default weights, to test for association with 3 1 1 12 continuous IL-1ß (N=428) and IL-6 (N=425) cytokine production. Finally, for the purpose of 3 1 2 correlating significant non-coding common variant sets to cytokine levels, we calculated an 3 1 3 allelic score for all variants in the set. An allelic score is a way to collapse multidimensional 3 1 4 genetic data associated with a risk factor into a single variable(50). We slightly adapted the 3 1 5 allelic score calculation to our SKAT-based test results, into a weighted (based on AF-based 3 1 6 Beta.Weights function from SKAT package), directional (based on increasing or decreasing 3 1 7 cytokine production over the genotype categories) allelic score. Specifically, we inferred the prioritizing one SNP for investigating epigenetic effects we computed the same linear 3 2 9 models now using the log-transformed cytokine production as criterion variable and next to 3 3 0 the SNP also age and gender as covariates. The predictive capacity of the linear models, as 3 3 1 reflected by the model p-value, was used as a measure for impact of a specific SNP on 3 3 2 cytokine production and as such prioritized rs1562305 for more in-depth follow-up.

3
Epigenetic effects of rs1562305 were studied using a Hi-C dataset(51) and Pol II ChiA-PET

4 9
killed Candida albicans 10 6 CFU/mL (C. albicans) and 1x10 6 /mL Staphylococcus aureus (S. in our extensive quality filtering due to homology regions, and was therefore excluded from 3 6 6 3 9 6 performed in this study can be found in Additional File 6. 3 9 7 3 9 8 Association landscapes show similarities and differences in IL-1ß and IL-6 response 3 9 9 We created holistic heatmap overviews termed 'association landscapes', for the

2 8
The inflammatory-phenotype level analysis (Figure 3), revealed a strong 4 2 9 association between rare variants in genes with pro-inflammatory effects and LPS-induced 4 3 0 IL-6 cytokine production (SKAToR adj P-value=2.42E -04 and SKATO adj P-value=1.99E -03 ), that    that the alternative allele presented with 1) a higher frequency as compared to the ancestral 4 7 4 allele, and 2) a higher cytokine production (residual, after correcting for co-variates age and 4 7 5 gender). Figure 5A shows that for each of these variants the IL-1ß and IL-6 in vitro cytokine 4 7 6 production in response to C. albicans stimulation decreases over the genotype categories.

8 1
While the rest of our study focused on coding variants, i.e. variants that likely have a 4 8 2 direct effect on protein function, we additionally aimed to get insight into the impact of non- Disequilibrium (LD), using within cohort R 2 metrics. Figure 5B shows that by subjecting 4 9 1 these expanded sets to the same SKAT, we identified associations between C. albicans-4 9 2 induced IL-6 production and non-coding common variants in IL36A (P-value=0.049), IL38 (P-4 9 3 value=0.007) and pro-inflammatory genes (P-value=0.019), and between C. albicans-4 9 4 induced IL-1ß production and non-coding common variants in IL38 (P-value=0.046). We 4 9 5 visualized these associations by calculating an allelic score, with single SNP directions and 4 9 6 MAF-based variant weights incorporated, for all significant sets in 5B. In Figure 5C the 4 9 7 weighted directional allelic score is displayed in correlation with cytokine production, 4 9 8 demonstrating that for each set there is an increasing cytokine production with increasing 4 9 9 score, highest for IL38 set with a correlation (R 2 ) of 0.025. Finally, to evaluate the individual 5 0 0 contribution of non-coding common variants in a set, we organized single SNP effect 5 0 1 estimates based on their direction and significance (Additional File 7). A more in-depth 5 0 2 follow up on IL36A set non-coding SNP rs1562305, revealed its non-coding activity by a In this study we identify and characterize rare and common genetic variation in genes 5 1 3 related to the IL-1 pathway, and determine their impact on the inter-individual variability of 5 1 4 stimulus-specific in vitro cytokine responses, measured in whole blood from healthy 5 1 5 individuals. In addition, we assess the relative contribution of rare as compared to common 5 1 6 variants, as well as the joint effect of rare and common variants, by employing various 5 1 7 grouping strategies.

1 8
Over the past decades, multiple studies have identified a role for common genetic 5 1 9 variation on cytokine level and response, however a significant proportion of inter-individual 5 2 0 variability remains to be determined (17)(18)(19)(20). Common and rare variants have mostly been 5 2 1 studied separately, and the differentiation between them is an arbitrary decision.

2 2
Considering increasing evidence that specific combinations of variants with variable 5 2 3 frequencies accounts for variability in phenotypic presentation (12)(13)(14), in particular for a 5 2 4 combination of phenotypic characteristics that do not fit one specific clinical diagnosis(58),

2 5
we hypothesized this concept might also apply to the inter-individual variability in cytokine 5 2 6 responses. As such, this study aimed to provide a more holistic view, that considers the 5 2 7 interplay of variants of different allele frequencies. For this purpose, we sequenced the 5 2 8 coding regions of 48 genes related to the IL-1 pathway in almost 500 healthy individuals,

2 9
and assessed in vitro IL-1ß and IL-6 production by whole blood in response to LPS, PHA, C.  The strongest rare variant association detected in this study was between CASP1 5 3 6 and LPS-induced IL-6 cytokine production. CASP-1 protein, encoded by CASP1 gene, is 5 3 7 responsible for cleavage of the inactive mediators IL-1ß, IL-18 and IL-33 into their active 5 3 8 form. The association in this study was based on five rare variants: two private variants (one

4 1
The fact that we observed a burden of rare variants not with IL-1ß but with IL-6 production, also reflect an unnoticed effect of these variants on preceding IL-1ß production, that 5 4 7 subsequently influence the induction of IL-6.

4 8
We detected an almost equally strong rare variant burden between NCF4 and both