GWAS Identifies DPP6 as Risk Gene of Cognitive Decline in Parkinson’s Disease

Abstract

Background

Cognitive decline is among the most common non-motor symptoms in Parkinson’s disease (PD), while its physiological mechanisms remain poorly understood. Genetic factors constituted a fundamental determinant in the heterogeneity of cognitive decline among PD patients. However, the underlying genetic background was still less studied.

Methods

To explore the genetic determinants contributing to cognitive decline in PD, we performed genome-wide survival analysis using a Cox proportional hazards model in a longitudinal cohort of 450 Chinese patients with PD, and further explored the functional effect of the target variant. Additionally, we built a clinical-genetic model by incorporating clinical characteristics and polygenic risk score (PRS) to predict cognitive decline in PD.

Results

The cohort was followed up for an average of 5.25 (SE = 2.46) years, with 95 incidents of cognitive impairment. We identified significant association between locus rs75819919 (DPP6) and accelerated cognitive decline (p = 8.63E-09, beta = 1.74, SE = 0.30). Dual-luciferase reporter assay suggested this locus might be involved in the regulation of DPP6 expression. Using data set from the UK Biobank, we identified rs75819919 was associated with cognitive performance in the general population. Incorporation of PRS increased the model’s predictability, achieving an average AUC of 75.6% through fivefold cross-validation in 1 000 iterations.

Conclusions

These findings improve the current understanding of the genetic etiology of cognitive impairment in PD, and provide a novel target DPP6 to explore therapeutic options. Our results also demonstrate the potential to develop clinical-genetic model to identify patients susceptible to cognitive impairment and thus provide personalized clinical guidance.

Cognitive dysfunction is one of the most detrimental symptoms of Parkinson’s disease (PD), which significantly affects the patients’ quality of life and places a substantial burden on the caregivers (1). Mild cognitive impairment affects approximately 20%–33% of newly diagnosed PD patients (2), while dementia develops in up to 80% of patients as the disease progresses (3,4). Consequently, it is crucial to identify risk factors that contribute to the progression of cognitive decline in order to facilitate early diagnosis, prevention, and treatment of PD dementia.

In recent years, emerging evidence has highlighted the significant role of genetic factors in the cognitive performance of individuals with PD (5,6). For instance, mutations in the β-glucocerebrosidase gene (GBA), the most common genetic risk factor of PD accounting for about 3%–4% of patients (7), were established to be associated with accelerated cognitive decline in PD (8). Moreover, other genes like MAPT, APOE, and COMT were suggested to be associated with PD dementia (9), though they have not been specifically identified as risk genes for PD. Additionally, the cumulative polygenic risk score (PRS) generated based on variants associated with risk of PD has been found to be linked to faster cognitive decline in patients with PD (10). With the advent of next-generation sequencing, extensive efforts have been made in the past decade to unravel the genetic factors associated with PD risk through case-control genome-wide association studies (GWAS). However, the focus on understanding the genetic basis of cognitive impairment in PD has been limited to a small number of studies. GWAS involving European-ancestry individuals have been conducted to investigate the genetic variants modulating clinical outcomes including cognitive impairment in PD, while few risk variants were identified (11–13). Therefore, genetic analysis in longitudinal cohorts with detailed clinical observation of cognitive progression is urgently needed.

In this context, we explored the genetic modifiers for cognitive decline in a Chinese PD cohort. We found a novel locus rs75819919 (DPP6) associated with an accelerated rate of cognitive decline. Further exploration in the UK Biobank data set suggested this variant was associated with cognitive performance in the general population. These findings shed light on the genetic mechanisms underlying cognitive impairment in PD, and highlight the potential implications of DPP6 in cognitive function across diverse populations.

Materials and Methods

Study Design and Participants

An overview of the workflow is described in Figure 1. The patients with PD were recruited from the Department of Neurology at West China Hospital of Sichuan University. The patients were diagnosed by 2 neurologists specializing in movement disorders according to either the United Kingdom PD Society Brain Bank Clinical Diagnostic Criteria (14), or the Movement Disorder Society Clinical Diagnostic Criteria (15). We excluded patients meeting the following criteria from the analysis: (1) those with a disease duration of more than 3 years since symptom onset. For these patients, the cognitive function might have already declined to some extent as the disease progressed; (2) those who exhibited cognitive impairment at baseline, as assessed by the Beijing version of the Montreal Cognitive Assessment (MOCA) with a cutoff score below 22, considering the overall lower education level of the older people (16); (3) those who have received deep brain stimulation (DBS) during follow-up because DBS might influence disease progression (17); (4) those who had other diseases, which might influence the cognitive function such as vascular lesions, progressive supranuclear palsy, vitamin B12 deficiency, or hypothyroidism. All participants have signed informed consent and the study was approved by the West China Hospital of Sichuan University.

Figure 1.

Schematic analysis workflow.

Clinical Evaluation

The enrolled patients underwent a repeated series of neurological evaluations at baseline and during follow-up by 2 trained interviewers. Cognitive performance was evaluated by MOCA (range: 0–30) (16), and cognitive performance during disease progression was assessed on a regular basis of ~1 year. Cognitive impairment was defined as a decrease of at least 3 points from the baseline MOCA score, as previous studies have suggested that a reliable change in cognitive score during long intervals was 3 to 4 points (10,18). The time to cognitive impairment was determined as the duration from the baseline examination to follow-up examinations when a 3-point decrease was first observed. For those patients without a 3-point decrease, time to event was defined as the difference between the date of baseline examination and the censor date taken as the date of the last valid follow-up. Other demographic and clinical features such as the Unified PD Rating Scale Part III (UPDRS3) and Hoehn-Yahr stage were measured as described previously (19).

Genotyping and Quality Control

Genomic DNA was extracted from peripheral blood leukocytes using standard phenol-chloroform procedures and then genotyped on the Illumina Infinium Asian Screening Array-MD v1.0 for a total of ~0.66 million single nucleotide polymorphisms (SNPs) via standard protocol. Imputation was performed using Minimac4 with reference panel Genome Asia Pilot (20) after phasing with Eagle 2.4 (21).

Extensive quality control procedures were performed. Imputed SNPs with r² < 0.8 and genotype calls with genotype quality < 0.8 were removed. Individuals with a call rate below 95%, or mismatched sex inferred from X-chromosome analysis using Plink (22) were excluded. Principal component analysis was conducted based on 1 000 Genomes Project haplotypes, and identified outliers were removed. Additionally, SNPs with missingness > 0.05 or out of Hardy-Weinberg Equilibrium (p < 1E-06) or with minor allele frequency (MAF) < 0.01 were removed, as well as non-autosomal SNPs.

Genome-Wide Association Analysis

To identify risk loci associated with the cognitive function of PD, we performed a multivariate linear regression analysis on the MOCA score at baseline assuming an additive model using Plink, adjusting for sex, age at onset (AAO), disease duration, levodopa equivalent daily dose (LEDD), education level, comorbidities, which might influence cognitive function including hypertension, stroke, and diabetes, as well as the 5 five principal components to account for population stratification. The significant level was set as p = 5E-08 at genome-wide.

To investigate the genetic factors associated with cognitive decline in PD, we further performed genome-wide survival analysis using a Cox proportional hazards model, adjusting for sex, AAO, MOCA at baseline, disease duration, LEDD, education level, comorbidities including hypertension, stroke, and diabetes, and the first 5 PCs. Survival analysis was conducted using the R package “gwasurvivr” (The R Foundation, Vienna, Austria).

In addition, we calculated PRS of PD risk for each individual, and evaluated the correlation between PRS and cognitive decline using a Cox proportional hazards model. PRS was calculated based on full summary statistics from the largest GWAS on PD risk in the European population (23) with PRSice (24). Then PRS was converted to Z scores to make the scale of analyses more easily interpretable.

Exploration of Significant Variants in the UK Biobank Data Set

We further expanded our investigation by exploring the association between the significant variants identified in our analysis and general cognitive performance using data from the UK Biobank. The UK Biobank is a large-scale prospective health study that recruited over 500 000 individuals from various regions across the United Kingdom between 2006 and 2010 (25). Because the MOCA score was not collected in the UK Biobank study, we analyzed 3 cognitive tests that assess general intelligence and processing speed, namely prospective memory, reasoning, and reaction time. All 3 tests were administered via computerized touchscreen interface at the initial assessment visit. Details of the cognitive test could be found in Supplementary Table 1 and on the UK Biobank website, https://biobank.ctsu.ox.ac.uk/crystal/label.cgi?id=100026. Data were obtained from the UK Biobank under an approved application (ID: 98992).

In our analysis, we excluded participants who reported chronic neurological diseases, which could directly affect cognitive function (Supplementary Table 2). The International Classification of Disorders (ICD)-10 system (field codes: 41202, 41204, and 41270) was used to identify those diagnosed with the target neurological diseases. Individuals with missing values in cognitive measures were removed for each test. We conducted phenotype correlation analysis using the Pearson correlation test to assess the relationships between different cognitive tests. Linear regression analysis was performed adjusting for sex and age on the cognitive ability score assuming an additive model in the general population, as well as in the PD population from the UK Biobank. PD patients were identified as described in previous study (26). Imputed autosomal genotype data of the UK Biobank was provided as part of the data release (field ID: 22828).

Clinical-PRS Prediction Model

To explore the potential utility of PRS of cognitive decline in clinical trials, we built a clinical-PRS model to predict cognitive impairment within 5 years since symptom onset. Predictability was estimated with receiver operating characteristic curves. Clinical factors associated with cognitive decline of PD were first filtered in a backward stepwise fashion based on the Akaike information criterion, considering that only a few clinical factors were available to choose from. Single factors associated with cognitive decline of PD were assessed using a logistic regression model. To address the issue of potential overfitting and biased results when testing the predictive performance in the same cohort used for model training, we built and tested the prediction model in 1 000 training and test sets randomly generated from the entire study population. In each iteration, we calculated the beta and p values of each SNP from the beginning in the training set (randomly 80% individuals selected), and used these values to predict cognitive decline in the corresponding test set (remaining 20% individuals). Statistical analyses were performed in R v3.5.3.

Cell Culture

The SH-SY5Y cells were obtained from the China Center for Type Culture Collection, Wuhan, China, and validated using an STR profiling method. SH-SY5Y is a thrice-subcloned cell line derived from the SK-N-SH neuroblastoma cell line, and serves as a model for neurodegenerative disorders. The SH-SY5Y cells were cultured by DMEM supplemented with 10% fetal bovine serum, 100 U/mL penicillin, 100 mg/mL streptomycin, and 1% minimum Eagle’s essential medium nonessential amino acids (all from Thermo Fisher Scientific, Waltham, MA) at 37°C, 5% CO₂ with a constant humidity environment.

Dual-Luciferase Reporter Assay

The pGL3-promoter vector was from Promega (Promega Corporation, Madison, WI). About 1 000 bp region around the lead SNP with wild-type or mutant site were cloned into the pGL3-promoter vector, which contained the SV40 promoter upstream of the firefly luciferase reporter gene. The SH-SY5Y cells were plated in 24-well plates and cotransfected with 1 μg of pGL3-promoter-WT or pGL3-promoter-MT and 0.2 μg of pRL-TK vector (Promega) as an internal control by using Lipofectamine 2000. Moreover, the pGL3-promoter vector was transferred as a systemic control. Luciferase activity was measured 48 hours after transfection and the assay was conducted according to the standard protocol of Promega Dual-Luciferase Reporter Assay System. Minimum of 5 independent experiments were performed, each including 3 technical replicates.

Results

Genome-Wide Association Analysis at Baseline

We first explored the genetic determinants contributing to the baseline cognitive function of patients with PD. A total of 1 529 patients remained after quality control for the association analysis. The average disease duration since symptom onset was 1.48 (0.82) years. The average AAO (SD) was 58.53 (11.51) years old with a sex ratio of 1.16 (male/female: 821/708). The genomic inflation factor λ was 1.03, suggesting minimal bias from population stratification. The MOCA score distribution was shown in Supplementary Figure 1. However, no variant was identified to be significantly associated with the baseline cognitive function (Supplementary Figure 2).

Genome-Wide Cox Regression Analysis

Then we explored the risk loci associated with the cognitive decline of PD. A total of 450 patients with regular follow-up were involved in the survival analysis, among whom 95 demonstrated cognitive impairment. The average AAO (SD) was 57.12 (SE = 11.16) with a sex ratio of 1.30 (male/female: 254/196). The average disease duration since symptom onset was 1.44 (SE = 0.78) years, and the cohort was followed up for an average of 5.25 (SE = 2.46) years. The MOCA score distribution was shown in Supplementary Figure 3. We observed a mean change in the MOCA score of −0.29 (SE = 1.64) points per year. Among the 95 patients with cognitive decline, the mean change was −1.99 (SE = 2.18) points per year. The genomic inflation factor λ was 1.03, suggesting minimal bias from population stratification (Figure 2A). We identified one significant locus rs75819919 (DPP6) (p = 8.63E-09, beta = 1.74, SE = 0.30; Figure 2A and B, Supplementary Table 3), which resulted in an average decrease of ~1.75 MOCA score per year, compared with a decrease of ~0.16 MOCA score per year for patients without this variant (Figure 2C).

Figure 2.

Genome-wide survival analysis for cognitive decline of Parkinson’s disease. (A) Manhattan plot and QQ plot showing GWAS results for cognitive decline of PD. The locus rs75819919 (DPP6) reached genome-wide significance level (p < 5E-08). (B) Regional plot of the association signal at rs75819919 by LocusZoom. (C) Kaplan-Meier survival curves of cognitive decline of PD for rs75819919. Given that no individuals carried the homozygous genotype for the target variant rs75819919, a dominant model was applied. PD = Parkinson's disease; GWAS = genome-wide association studies.

Functional Exploration of the Significant Variants

Because most loci identified by GWAS are thought to regulate gene expression, we screened the significant SNPs for cis-expression quantitative trait loci (eQTL) signals in GTEx (27), eQTLGen (28), and PsychENCODE (29). However, the identified locus was not a cis-eQTL in these data sets. Given the low MAF of rs75819919 in the population and these eQTL data sets were based on individuals of European descent, we conducted a dual-luciferase reporter assay in SH-SY5Y cells to examine the impact of the rs75819919 G>T variant on DPP6 expression (Figure 3A and B). We observed that the G>T variant decreased the ratio of firefly/Renilla luciferase reporter gene expression compared to the wild-type, indicating that G>T suppressed the expression of DPP6 (Figure 3C). Therefore, the identified significant locus might be related to the cognitive decline of PD by modulating the expression of DPP6.

Figure 3.

Luciferase reporter assay of the role of rs75819919 on gene expression. (A, B) The schematic diagram shows the structure of firefly luciferase reporter plasmid. (C) The plasmid with the mutant variant G>T significantly decreased the ratio of firefly/Renilla luciferase reporter gene expression compared to the wild-type.

Exploration of Significant Variants in the UK Biobank Data Set

Of the 502 376 participants from the UK Biobank data set, 7 438 with target neurological diseases were excluded. We identified significant phenotypic correlation between the 3 cognitive tests, suggesting these test scores were associated (Supplementary Table 4). Our analysis revealed an inverse association between rs75819919 and cognitive performance in the general population from UK Biobank (Supplementary Table 5), which aligned with our findings that patients with this variant exhibited faster cognitive decline. These results suggested that rs75819919 was associated with overall cognitive performance, not only in individuals with PD but also in the general population. Notably, the coefficients of the association between rs75819919 and fluid intelligence and prospective memory were small, suggesting the effect of this variant on these 2 cognitive functions was limited. However, this variant exhibited no association with the 3 cognitive tests in the PD population from UK Biobank, potentially owing to the limited sample size (Supplementary Table 6). Therefore, additional replication was still necessary.

Polygenic Risk Score and Cognitive Function

Polygenic risk score based on risk loci identified from GWAS on disease risk has been recognized as a genetic predictor for specific clinical characteristics in PD (30). In the calculation of PRS based on summary statistics from the previous GWAS, the best-fit p value threshold was .007, and the variance explained by the chosen SNPs was 0.016. We found that higher PRS led to faster cognitive decline (p = .048, beta = 0.22, SE = 0.11), consistent with previous results based on individuals of European ancestry (10). Therefore, the risk loci for PD status also infer the overall risk of cognitive impairment in PD.

Clinical-PRS Model Predicts Cognitive Impairment in PD

To evaluate the potential utility of PRS in clinical trials, we tested whether PRS of cognitive decline could provide valuable information in predicting cognitive impairment within 5 years compared with other clinical variables. The selected clinical variables included AAO, sex, LEDD, Hoehn-Yahr stage, score of UPDRS3, Hamilton Depression Rating Scale, Hamilton Anxiety Rating Scale, Frontal Assessment Battery, MOCA, presence of motor fluctuation, dyskinesia, restless legs syndrome, hallucination, constipation, freezing of gait, and falls at baseline. We tested the model in unrelated patients using fivefold cross-validation. PRS of cognitive decline performed the second best (AUC = 0.566, 95% CI: 0.404–0.727) in predicting cognitive impairment within 5 years compared with other clinical factors (Figure 4A). In the 1 000 iterations, PRS of cognitive decline was chosen 480 times, ranking the third (Figure 4B). The average AUC of the clinical-PRS model was 0.756 (95% CI: 0.620–0.891), representing a 10.4% increase compared with model using clinical variables only (AUC: 0.685, 95% CI: 0.523–0.842).

Figure 4.

Stability of the PRS of cognitive decline. (A) The average AUC of each predictor variable. (B) The number of times selected into the model out of 1 000 iterations for each predictor variable. AUC = area under the curve; PRS = polygenic risk score.

Discussion

A number of environmental and genetic factors contributed to the cognitive impairment (5,31–33). However, few studies have investigated the genetic background for cognitive function, and previous studies were mostly based on individuals of European ancestry. To fill this research gap, we performed genome-wide survival analysis in a Chinese cohort. Our analysis led to the identification of a novel locus, rs75819919 (DPP6) as a genetic modifier for cognitive decline in PD. Further exploration in the UK Biobank data set suggested this variant was associated with cognitive performance in the general population. Moreover, we integrated PRS into a prediction model and observed a substantial improvement in predicting accuracy for cognitive impairment. Overall, our study contributes to a deeper understanding of the genetic basis of cognitive impairment in PD, specifically within the Chinese population.

DPP6 is a transmembrane protein mainly expressed in the brain, where it binds to the potassium channel K_v4.2 to enhance its expression and controls the dendritic excitability in hippocampal CA1 pyramidal neurons (34). Studies conducted on DPP6-knockout mice have demonstrated that loss of DPP6 function impairs hippocampal synaptic development and induces behavioral impairments in recognition, learning, and memory (35), highlighting the importance of DPP6 in cognitive function. Additionally, genetic analysis has revealed that disrupting of the coding sequence or deleterious variants in DPP6 was related to early-onset Alzheimer’s disease (AD) and frontotemporal dementia (FTD), suggesting DPP6 may have a broader role in the regulation of cognitive function in different disease contexts (34). In our study, we established the link between DPP6 and cognitive impairment in PD from the genetic perspective. Although the clinical manifestations vary among PD, AD, and FTD, these results supported the notion that these diseases may share common underlying molecular mechanisms in cognitive impairment. Notably, we found that executive function showed the most significant decrease in the MOCA score in the patients. Executive function is a set of mental skills including working memory, flexible thinking, and self-control, and is affected by multiple brain regions such as the hippocampus (36). Previous study has shown that the hippocampal neurons lacking DPP6 showed a sparser dendritic branching pattern along with fewer spines throughout development and into adulthood, and these deficits led to fewer functional synapses and occur independently of the potassium channel subunit Kv4.2 (37). Meanwhile, it has been established that mutations in DPP6 resulted in a decrease of DPP6 protein levels and supported a haploinsufficiency mode of the pathogenesis (34). The dual-luciferase reporter assay in the current study revealed that the identified locus rs75819919 could regulate the expression of DPP6. Together, these results suggested that dysregulation of DPP6 expression might contribute to the shared underlying mechanism of cognitive impairment in various neurodegenerative disorders.

A previous large-scale genome-wide survival study was conducted in the European population to pinpoint genetic variations influencing the progression of cognitive decline. This study identified 3 target variants, namely rs182987047 (RIMS2), rs138073281 (TMEM108), and rs8050111 (WWOX) (13). The first 2 variants were rare in the Asian population and were not available in our data set. The third variant, rs8050111, did not demonstrate significance in our data set (p = .95, beta = 0.01, SE = 0.19). These findings imply ancestral disparities in the genetic background influencing the progression of cognitive decline. Previous study conducted on patients of European ancestry has demonstrated that common genetic variants associated with PD risk have a small but consistent effect on disease etiology, combining to expedite faster motor and cognitive decline (10). Our study expanded these findings by showing that a higher PRS was associated with faster cognitive decline in the Asian population. Due to the polygenic architecture of complex traits like PD, single variant might have limited effect on the pathogenesis or the progression of the disease. Previous studies have demonstrated that clinical characteristics together with genetic information from multiple trait-related genes could improve prediction accuracy for traits like impulse control disorder in PD (38). However, the contribution of individual genes to cognitive decline may be small, and the performance of model was susceptible to the chosen target genes. In this context, we built a predictive model by combining clinical traits and PRS of cognitive decline, which substantially improved the prediction accuracy. Compared with each clinical factor, PRS of cognitive decline still demonstrated good predictability for cognitive impairment in unrelated samples. These findings suggested the potential utility of the clinical-PRS model as a predictive tool for identifying individuals at higher risk of cognitive decline in PD, who would benefit from interventions aimed at slowing down the progression of cognitive impairment. However, it is important to note that our model was validated only in the Chinese population. Further studies are needed to replicate these findings in diverse populations and to better understand the complex relationships between clinical and genetic risk factors for cognitive impairment in PD.

The present study investigated the genetic factors contributing to cognitive decline of PD in a Chinese cohort. There were a few limitations worth mentioning. Firstly, the sample size in the survival analysis was relatively small, potentially leading to the oversight of variants with less pronounced effects and raising the likelihood of chance influencing the observed results. Secondly, there exists a lack of replication for the identified significant locus. Due to the limited number of samples with regular follow-up, we adopted a one-stage design combining all available samples to maximize statistical power. Consequently, further studies are imperative to validate these findings in other populations within East Asian or Chinese PD cohorts, thereby refuting the possibility that the results are specific to a particular subpopulation. Thirdly, it should be noted that MOCA serves as a rudimentary screening tool for cognitive decline, particularly in patients in the early stages. Subsequent investigations incorporating more comprehensive neuropsychological tests may offer additional insights. Fourth, we utilized summary statistics of GWAS in the European population when calculating PRS of disease risk, which might bring some bias to the results. Further replication in the Asian population was still necessary.

Conclusion

We performed a genome-wide survival analysis on cognitive impairment of PD in a Chinese PD cohort and identified rs75819919 as a risk locus. Functional exploration suggested this locus might influence cognitive function by regulating the expression of DPP6. These findings expanded current knowledge on the genetic architecture for the cognitive function of PD, and underscored the potential of the clinical-PRS prediction model in facilitating personalized treatment options and clinical trial designs for patients at higher risk of cognitive decline.

Funding

This research was supported by the funding of the National Key Research and Development Program of China (grant no. 2021YFC2501200), and the Sichuan Science and Technology Program (grant nos. 2022ZDZX0023 and 2021YJ0415).

Conflict of Interest

None.

Data Availability

Summary statistics of risk of PD based on European individuals used in the PRS calculation were downloaded from iPDGC (https://pdgenetics.org/resources).

Author Contributions

C.L.: investigation, methodology, execution, and writing; Y.H.: investigation and execution; R.O.: investigation and execution; Q.W.: blood samples collection; L.Z.: clinical data collection; K.L.: clinical data collection; J.L.: patients enrollment; X.C.: patients enrollment; W.S.: patients enrollment; B.Z.: patients enrollment; Y.W.: patients enrollment; H.S.: design, conception, supervision, and funding acquisition. All authors read and approved the final manuscript.