Introduction

In recent years, activation of the nucleotide-binding oligomerization domain-, leucine-rich repeat, and pyrin domain-containing 3 (NLRP3) inflammasome has been implicated in Parkinson’s disease (PD) by numerous functional studies using different models1. Inflammasomes are protein complexes that serve as signaling platforms for the activation of the immune response. The NLRP3 inflammasome comprises three main components: NLRP3 (encoded by the NLRP3 gene), apoptosis-associated speck-like protein containing a caspase activating recruitment domain (encoded by PYCARD), and caspase-1 (CASP1). NLRP3 is expressed in microglia and when activated, it leads to secretion of the cytokines IL-1β and IL-18, which leads to neuroinflammatory response and pyroptosis2.

The evidence for the involvement of the NLRP3 inflammasome in PD is mainly derived from in vitro and in vivo cell and animal models, by interacting with α-synuclein, mitochondria, and other mechanisms. For example, early research suggested that in human monocytes, α-synuclein may directly trigger the NLRP3 inflammasome3. Similar results have been reported in other cell and animal models4,5. Other studies in cell and animal models have suggested that the NLRP3 inflammasome may be involved in toxin-mediated PD and that there could be an interplay between mitochondria and the NLRP3 inflammasome in PD pathogenesis6,7. In humans, one study reported that a genetic variant in NLRP3 may affect its expression and the risk of PD8. Several studies in cells and postmortem brain tissues from PD patients and controls reported alterations in the NLRP3 inflammasome in PD8,9,10. However, there are no thorough human genetic studies of the NLRP3 inflammasome in PD, although such studies can help with inferring causality. Nevertheless, there is a suggestion that the NLRP3 inflammasome may be a good target for therapeutic development in PD, and several compounds targeting the NLRP3 inflammasome are in different stages of development11. Considering that clinical trial success rates increase significantly when supported by genetic evidence12, it becomes crucial to conduct a thorough genetic analysis of the proposed target.

In this study, we aimed to examine whether human genetics data supports NLRP3's involvement in PD and the development of therapeutics targeting NLRP3 for PD. We analyzed common and rare variants in the NLRP3 inflammasome components in large PD cohorts and further performed pathway-specific analyses of polygenic risk scores (PRS) and Mendelian randomization (MR) analyses. Our results do not support an important role for the NLRP3 inflammasome in PD nor its being a good target for therapeutic development in sporadic PD.

Results

No association between NLRP3 inflammasome genes and PD

We examined common variants from the largest available PD risk GWAS (N cases/proxy-cases = 49,053; N controls = 1,411,006)13. We did not observe any associations between PD and variants in genes composing the NLRP3 complex (NLRP3, PYCARD, and CASP1) and the genes encoding the cytokines released by its activation, IL-1β and IL-18 in neither GWAS on participants from European ancestry nor the diverse non-European Global Parkinson’s Genetics Program (GP2) cohorts (Fig. 1; Supplementary Fig. 1). While the PYCARD gene is located near one of the GWAS loci (rs11150601) within SETD1A, PYCARD is just outside of the linkage disequilibrium (LD) block, i.e. there are no variants within or in regulatory regions of PYCARD that are in LD (r2 < 0.2) with the variants that surpassed the GWAS level of significance. We then performed PRS analyses for the three NLRP3 inflammasome genes from 14,828 PD cases and 13,283 controls across 7 cohorts (detailed in Supplementary Table 1). Overall, the PRS explains a very small portion of the variance in PD (1.39E-06–0.001) and was not associated with PD (Fig. 2, Supplementary Table 2).

Fig. 1: Locus zoom plot of NLRP3, CASP1, PYCARD, IL-1β, and IL-18 genes (±500 kb) in Parkinson’s disease GWAS.
figure 1

The SNP with the lowest P-value in the studied gene or locus is highlighted by red square. The punctuated line represents GWAS level of significance p < 5 × 10−8. SNPs shown in gray are below this significance level, while those in blue exceed it. X axis: Chromosomal position (in Mb). Y axis: Negative log10 of the P-values from GWAS.

Full size image
Fig. 2: Pathway-specific PRS analysis of NLRP3 inflammasome genes.
figure 2

OR odds ratio, CI confidence interval, PPMI Parkinson’s Progression Markers Initiative, APDGC Autopsy-confirmed Parkinson Disease GWAS Consortium, IPDGC International Parkinson Disease Genomics Consortium, NINDS National Institute of Neurological Disorders and Stroke Repository Parkinson’s Disease Collection, NGRC NeuroGenetics Research Consortium, UKBB UK Biobank.

Full size image

We also analyzed rare variants in two independent cohorts, including 2943 patients and 18,486 controls, followed by a meta-analysis (Supplementary Table 3). We did not observe any associations between any subsets of variants in any of the genes comprising the NLRP3 inflammasome and PD (Supplementary Table 4). We then performed an analysis including all the variants in all three genes combined, and in this analysis too, we did not observe any associations between rare variants and PD (Supplementary Table 4).

MR does not support NRLP3 as a druggable target for PD

Using summary-data-based Mendelian Randomization (SMR), we investigated whether the modulation of the NLRP3 inflammasome could be a target for therapy. Initially, we established that NLRP3, CASP1, IL-1β, IL-18 are recognized within the database of druggable genes, highlighting their potential relevance for drug development14. PYCARD, however, did not meet these criteria. We then performed SMR, where exposure was quantitative trait loci (QTL) of NLRP3 genes from tissues that are relevant to PD pathogenesis. In the present study, we used the Genotype-Tissue Expression (GTEx) project v8 release (All brain tissues, blood, and liver), PsychENCODE, and BrainMeta/brain-eMeta15,16,17. As an outcome for SMR, we used the most recent PD risk13, PD age-at-onset18 GWASs, and the largest publicly available PD motor and cognitive progression GWASs19,20. Our analysis did not reveal any potential causal associations between the QTL data tested in this study and PD in tissues relevant for PD after correction for multiple comparisons (Supplementary Table 5).

Discussion

Our results, using large-scale human genetic, transcriptomic, and methylomic data, do not support the NLRP3 inflammasome as important in PD pathogenesis or as a good target for drug development. There were no associations of common or rare variants nor of PRS for the NLRP3 inflammasome, with risk of PD. When we considered the three NLRP3 genes as druggable targets, there was no evidence that altering their expression at the RNA level may have an effect on risk, onset, or PD progression.

While using MR to infer efficient druggability is not a definitive test, it can still provide valuable information. For example, a recent MR study was able to replicate the beneficial effects of tumor necrosis factor (TNF) inhibition in Crohn’s disease and ulcerative colitis, and its deleterious effect in multiple sclerosis21. The same study also suggested that TNF inhibition might not be beneficial for PD.

Understanding the role of a drug compound is essential when planning clinical trials. Studies that are not guided by genetic evidence are more likely to fail12. Currently, several phase 1 clinical trials targeting neuroinflammation and particularly NLRP3-inflamassome are being conducted22. The discordance between the hypothesis underlying these clinical trials targeting NLRP3 pathway in PD and our findings suggests that efforts to target the NLRP3 inflammasome in PD should be critically evaluated. It is important to select therapeutic strategies based on robust human genetic and biomarker evidence to reduce the chances of trial failure. Perhaps targeting the NLRP3 inflammasome could work specifically in individuals in which this pathway is pathologically activated, but this approach is not being taken, to the best of our knowledge. Subpopulations of patients with distinct genetic or environmental risk factors where the NLRP3 pathway plays a role in disease pathogenesis may exist. However, additional effort on defining the subpopulation of PD patients with neuroinflammation particularly with induction of the NLRP3 pathway, should be considered in clinical research targeting the NLRP3 inflammasome. Future studies exploring potential gene-environment interactions may further explain the role of the NLRP3 inflammasome in specific subsets of PD patients.

Our study has several limitations that need to be acknowledged. The GWASs on PD progression that were used could be underpowered. Further analysis using larger datasets should be performed when they become available to confirm our findings. The activation of NLRP3 inflammation involves a complex interplay of various proteins, such as NEK711. Further studies examining other genes encoding these activators in the context of PD are important. Another important limitation of our study is the reliance solely on genetic and transcriptomic data to infer the role and druggability of the NLRP3 inflammasome in PD. This approach does not account for post-translational modifications and the complex regulation at the protein level, which are critical for the functional activity of the NLRP3 inflammasome. Finally, the SMR analysis is dependent on the quality of the expression data used for exposure, and variations in quality across datasets might influence the results.

In conclusion, our analyses do not provide human genetic evidence for the involvement of the NLRP3 inflammasome in PD, suggesting potentially limited druggability from a genetic perspective.

Online methods

Study populations

To examine whether common variants in the NLRP3 inflammasome components may be associated with PD, we used summary statistics from the largest European PD GWAS13 and also analyzed genes of interest using the data from GP2 (release six) in several ancestry populations (detailed in Supplementary Table 6). Quality control analyses for both samples and variants have been previously described (https://github.com/GP2code/GenoTools). To study the association of common variants (minor allele frequency >1%) with PD in GP2 cohorts, we performed logistic regression, adjusting for age, sex, and the top five principal components in each ancestral population. For each of the cohorts, we created locus zoom plots using locuszoomr R library23 for the NLRP3, CASP1, PYCARD, IL-1β, and IL-18, loci with ±500 kb around each gene. We then created pathway-specific PRS for the NLRP3 inflammasome using available individual-level data from cohorts of European ancestry (detailed in Supplementary Table 1).

In our MR analysis, we utilized the following summary statistics datasets: PD risk GWAS13, PD age-at-onset GWAS with 17,415 cases18, and PD progression data from GWAS studies conducted by Iwaki et al.19 and Tan et al.20 The PD progression traits in the study by Iwaki et al.19 were measured using observational study meta-analysis of clinical scales data, we specifically used UPDRS Part III (N cases = 1398), MMSE (N cases = 1329), and MoCA (N cases = 1000) scores. In the study by Tan et al.20, PD progression was assessed using scores for motor, cognitive, and composite progression in 3364 PD patients with an average follow-up of 4.2 years.

To analyze rare variants, we performed an analysis in two cohorts with available whole-exome and whole-genome sequencing data with a total of 2943 PD patients and 18,486 controls (Supplementary Table 3). Whole-genome sequencing was available from the Accelerating Medicines Partnership—PD (AMP PD) initiative cohorts (https://amp-pd.org/; detailed in the Acknowledgment). Whole-exome data was available from the UK biobank (UKBB) cohort, which was accessed using Neurohub (https://www.mcgill.ca/hbhl/neurohub). All participants signed written informed consent. The ethics committee of McGill University gave ethical approval for this work.

PRS pathway analysis

In order to examine the potential genetic association of the NLRP3 complex as a whole in PD (as opposed to analysis of specific SNPs), we calculated pathway PRS using PRSet for the three genes encoding the components of the NLRP3 complex (NLRP3, PYCARD and CASP1)24. In this analysis, we only included participants of European origin and removed first- or second-degree relatives. Sex discrepancy analysis was conducted to compare the recorded biological sex of individuals in the dataset with their genetically inferred sex, determined by rates of heterozygosity and homozygosity on the X chromosome. This analysis used the --check-sex function in PLINK 1.9, where males were expected to exhibit an X chromosome homozygosity estimate greater than 0.85, and females less than 0.25. This method enabled the identification and exclusion of samples with potential sex mismatches (where the reported sex at recruitment does not match the genetic sex), thereby enhancing the accuracy of subsequent genetic analyses. Only common SNPs with minor allele frequency >0.01 and p value < 0.05 were included in the analysis. We conducted LD clumping, removing variants with r2 > 0.1 and within a 250 kb distance. We performed a permutation test with 10000 repetitions to generate an empirical p value for our gene set of interest. We used age-at-onset for cases, age-at-enrollment for controls, sex, and the top 10 principal components as covariates.

Whole-exome and whole-genome sequencing data analysis

To determine whether rare variants in the genes encoding the components of the NLRP3 inflammasome (NLRP3, PYCARD, and CASP1), we extracted genetic data from whole-exome and whole-genome sequencing datasets. Our analysis included only participants of European ancestry, and we excluded any first or second-degree relatives from the study. For whole-genome sequencing data, we performed quality control as previously described25. In brief, we included samples with a mean coverage of 25x and a rate of missing genotypes per sample less than 5%. For the UK Biobank’s whole-exome sequencing data, we used the Genome Analysis Toolkit (GATK, v3.8) to perform quality control. We applied the recommended filtration parameters for whole-exome sequencing data, which included a minimum depth of coverage of 10x and a minimum genotype quality score of 2026. The human reference genome hg38 was used for alignment.

We analyzed the association of rare variants with minor allele frequency <0.01 using the optimized sequence kernel association test (SKAT-O)27. The variants were grouped into different categories: all rare variants, all non-synonymous variants, loss-of-function variants (stop, frame-shift, and canonical splice-site variants), and variants with a combined annotation-dependent depletion score ≥20 (representing 1% of the top deleterious variants). To meta-analyze the two cohorts, we used the metaSKAT R package28.

Mendelian randomization

If modulation of the NLRP3 inflammasome is a target for therapy, then genetically driven differences in its expression, or that of the cytokines released following its activation, IL-1β, and IL-18, should be causally linked to PD risk or progression. To examine this possibility, we used SMR. SMR utilizes summary-level data to determine whether a causal relationship exists between an exposure and an outcome. In our specific case, we examined if differences in expression levels of the NLRP3 genes (using QTL) are associated with risk, age at onset, and progression of PD. As exposure, we used different QTL data from various studies and tissues, including methylation, gene expression, and chromatin QTLs. All the QTLs we used were collected from the same resource, and we conducted analyses using SMR software developed by Yang Lab with standard settings (https://yanglab.westlake.edu.cn)29,30. Specifically GTEx project v8 release (All brain tissues, blood, and liver), PsychENCODE, and BrainMeta/brain-eMeta15,16,17. As an outcome for SMR, we used the most recent PD risk13, PD age-at-onset18 GWASs, and the largest publicly available PD progression GWASs19,20. The Bonferroni-corrected significance threshold was set at p < 0.05/185 = 0.00027.