Parkinson disease (PD) is the second most common neurodegenerative disorder in the elderly, which is characterized by motor disabilities such as tremor, bradykinesia, rigidity as well as non-motor symptoms including autonomic dysfunction, hyposmia, rapid eye movement sleep behavior disorder (RBD), cognitive impairment and other clinical features [1, 2]. Both motor and non-motor disabilities are significant to patients’ quality of life and increase financial burden to the families and the society [3]. In addition to various combination of motor and non-motor presentations, the disease progression of each patient varies quite differently [4]. The rate of motor progression is the major factor related to prognosis and quality of life [5], although almost every patient with PD has non-motor symptoms ranging from 8 to 12 different symptoms, which is now considered as another crucial factor for driving quality of life of patients in PD [6]. Identifying factors influencing the progression of motor deterioration of PD is particularly important for the predicting the prognosis of patients with PD and discovery of novel therapies for PD. Genetic contributions to PD have been well acknowledged [2, 7, 8, 9]. Meta-analysis of genome-wide association studies (GWAS) in 2017 and 2019 have successfully identified more than 40 independent single nucleotide polymorphisms (SNPs) and 90 SNPs associated with PD, respectively [10, 11]. These 90 variants explained 16–36% of the heritable risk of PD depending on prevalence [11]. Recent studies showed that patients carrying a certain gene mutation have distinct clinical presentations and disease progression from sporadic PD [12, 13]. PD susceptibility genes have been associated with PD clinical features, such as onset age, subtype, and progression of PD [14, 15, 16, 17, 18, 19, 20, 21, 22, 23]. However, these studies were either based on cross-sectional PD cohorts or with limited genetic variants involved. Although recent genome-wide association study (GWAS) revealed genetic contributions to PD motor and cognitive progression in longitudinal PD cohorts [24], the input of multiple clinical quantitative measures and the algorithm to reflect the clinical progression of PD are also challenges to truly reveal genetic factors association with the progression of PD. This study aims based on recent GWAS identified PD risk SNPs to investigate genetic and clinical associations with rapid motor progression in a longitudinal PD cohort with five-year follow up [10, 11].

In this study, we found there were more male than female affected with PD, but not related to other clinical features (Table 1). In contrast to other studies [35, 36], age at onset did not associate with PD motor progression via statistical analysis of either ANOVA taking onset age as continuous variables and binomial the motor progression as fast and slow, or using Chi Square after binomial onset age outcomes to early and late onset, which is possibly due to similar factors contributing to the fast motor progression in either early or late onset of PD.

We found that among the non-motor symptoms, early and severe autonomic dysfunction was associated with rapid motor progression, which is consistent with previous research [37]. AUC did not reach a higher level when considering all NMS in this study, which indicates that autonomic dysfunction is the major predictive factor of motor progression among non-motor symptoms in this study. Fast motor deterioration may reflect rapid neuronal loss within the substantia nigra of PD, while autonomic dysfunction is considered due to enervation of the cranial vagus nerve, which is composed of the largest part of the parasympathetic nervous system. The dorsal motor nucleus of the vagus nerve located in the medulla oblongata is the earliest affected site by $\alpha$-synuclein related pathologies within the brain in idiopathic PD, and then the $\alpha$-synuclein related pathologies take upwards advancement pattern affecting pontine tegmentum and the substantia nigra of midbrain [38]. The association between early autonomic dysfunction and rapid motor progression of PD is coherent with the anatomical and pathological connections of PD.

Our results showed that multigenetic network with autonomic dysfunction together provided the optimal predictive model for fast motor progression in PD (AUC = 0.821, Figs. 1,2, Table 4). We identified seven genetic variants (LINC00693 rs6808178, CHMP2B rs115185635, IP6K2 rs12497850, TMEM175 rs34311866, OGFOD2 rs11060180, CAB39L rs9568188, and DLG2 rs3793947 were associated with rapid motor deterioration in PD (Table 3), and only LINC00693 rs6808178 passed multiple testing correction. Long non-coding RNA (lncRNA) has been shown function in regulating gene expression and cellular homeostasis [39]. Alterations in lncRNAs expression levels have been shown in brain tissues affected by PD compared to controls [40, 41]. Unfortunately, previous studies did not examine LINC00693. A recent study using machine learning model found that rs6808178 was one of two key variants among 90 SNPs for developing PD using the same PPMI cohort [42], which suggest LINC00693 is important in the onset and progression of PD.

For the other six probable fast motor progression associated genes, allele-specific expression of TMEM175 is identified in human brain [43], and TMEM175 deficiency could lead to lysosomal and mitochondrial dysfunction, as well as $\alpha$-synuclein deposition [44]. In the PD risk protein coding genetic interaction network (Fig. 2), TMEM175 is a hub protein with connection of SNCA, GBA, LRRK2, ACMSD, GAK, MCCC1 and FAM47E, while SNCA, GBA and LRRK2 have been shown associated with rapid motor progression [14, 18, 45, 46]. CHMP2B is located in the Lewy bodies of human brainstem [47, 48], and mutations of CHMP2B may contribute to endosomal and autophagic pathway in $\alpha$-synuclein deposition of PD [48]. IP6K2 is related to cell apoptosis, and its protein is highly expressed in granule cells of the cerebellum and related to impairment of locomotor function [49]. Mutations in DLG2 lead to reductions in excitatory synaptic input of dorsal striatum in mice [50]. The molecular mechanisms contributing to PD of DLG2 need further investigation, so are the same for OGFOD2 and CAB39L.

We took the rapidity of motor progression of PD, which is the main concern of PD patients, as our study question, rather than the disease progression profile, which incorporates other non-motor symptoms, such as pain, anxiety, fatigue and so on, to ensure the stability of the predictive model. A variety of research approaches have been adapted in investigating genetic associations with motor progression of PD in the literature, such as dichotomizing the cohort into fast motor progressors or not (Table 5), or differences in actual motor UPDRS score over time, etc. (Supplementary Table 3). SNCA, MAPT, LRRK2, GBA, PARK16, ATP8B2, SLC44A1and polygenic risk score have been shown associations with the progression of PD [14, 17, 18, 22, 23, 24, 31, 46, 51]. The definition of fast motor progression in different studies results in different proportion of fast motor progression ranging from 15% to 30% (Table 5) [17, 18, 31]. The rapid motor progression defined in this study has relatively smaller UPDRS-III score change, which is possibly because that the patients in PPMI are at the early stage of PD, and UPDRS-III is evaluated at post dose. Huang et al. [17] and Wang et al. [18] conducted cross-sectional studies and reveled motor progression associated with SNCA & MAPT and SNCA respectively. Besides, Paul et al. [31] elucidated that PRS is significantly associated with the progression of PD, and thus it is reasonable of taking all 44 SNPs as a whole to predict motor progression in this study. Oosterveld et al., Deng et al. and Davis et al. [46, 51] carried out linear mixed model to compare motor progression between carriers and non-carriers (Supplementary Table 3). Tan et al. [22] based on mixed-effects regression model using PCA and testified the relevance of PCA scores with 90 genetic factors. Iwaki et al. [22, 23] carried out two meta-analyses with large samples. However, they aimed at investigating genetic associations with motor severity rather than motor progression. Our study showed that genetic and clinical features together demonstrated more predictivity towards fast motor progression, while others only took genetic factors into account (Table 5 and Supplementary Table 3).

This study also has its limitations: First, the NeuroX genetic data did not include all 90 SNPs recently summarized [11]. Consequently, we could not assess the associations of missing SNPs, leading to not ideal AUC value for prediction of rapid motor progression. Second, patients enrolled in this study were mostly at early stage of PD, and some genetic variants may affect the disease progression at later stages, which were unable to be discovered in this study.

Our study indicated the rapidity of motor progression of PD could be attributed to clinical and genetical factors. Additional genetic information is required to enhance the value of genetic predicative model. Furthermore, the prediction of genetic associated molecular pathways & network for rapid motor progression in PD identified in this study warrant validation in other PD cohorts.

LC did data extraction from the PPMI database, statistics analysis and drafted the manuscript. YJ provided instructions of statistical and bioinformatic analysis of the study. YP interpreted clinical data and revised the manuscript. YH designed the project and critically revised the manuscript.

Application to data usage was approved by the scientific committee of PPMI, and this study involved database usage was approved by the Ethics Board of the Beijing Tiantan Hospital, Capital Medical University of China (KY 2018-031-02).

Data used in the preparation of this article were obtained from the Parkinson’s Progression Markers Initiative (PPMI) database (www.ppmi-info.org/data). For up-to-date information on the study, visit www.ppmi-info.org. PPMI — a public-private partnership — is funded by the Michael J. Fox Foundation for Parkinson’s Research and funding partners, including Abbvie, Allergan, Amathus Therapeutics, Avid Radiopharmaceuticals, Biogen, Biolegend, Bristol-Myers Squibb, Celgene, Covance, GE Healthcare, Genentech, Golub Capital, Handl Therapeutics, Insitro, Janssen Neuroscience, Lilly, Lundbeck, Merck, Meso Scale Discovery, Neurocrine, Pfizer, Piramal, Prevail Therapeutics, Roche, Sanofi Genzyme, Servier, Takeda, Teva, Ucb, Verily and Voyager Therapeutics.

The study was funded by National Natural Science Foundation of China (NSFC, 82071417, YH).

The authors declare no conflict of interest.