Slowing or preventing the progression of Parkinson’s disease (PD) remains a major unmet goal. To this end, therapeutic interventions that target pathogenic mechanisms in genetically-defined subgroups offer an attractive strategy. Along with GBA, pathogenic variants in the gene encoding leucine-rich repeat kinase 2 (LRRK2) represent the most common genetic cause of PD, with G2019S being the most prevalent pathogenic variant. Not surprisingly, a pipeline of LRRK2-targeted therapies is in clinical development1. Although current studies target populations with manifest PD, intervention in the pre-manifest phase of the disease offers an opportunity for disease prevention. Thus, non-manifesting carriers (NMCs) of LRRK2 pathogenic variants are a unique prodromal PD population for disease prevention studies. However, to enable such trials two major knowledge gaps must be addressed. First, establish the trajectory of clinical and biomarker changes during the prodromal stage in LRRK2 NMCs and second, elucidate clinical outcomes and/or biomarkers that can identify the LRRK2 NMC subgroup at the highest risk of developing PD. The latter is especially critical because LRRK2 pathogenic variants have a low, variable (25–40%) and age-modified lifelong penetrance2. Although there is a growing body of literature examining motor, non-motor and biomarker characteristics of LRRK2 NMCs3,4,5, there are limited longitudinal data from large prospective studies. We have previously reported the presence of subtle motor and non-motor signs of PD in LRRK2 NMCs compared to healthy controls (HCs) enrolled in the Parkinson’s Progression Markers Initiative (PPMI) study in cross-sectional analysis but indicated that longitudinal data will be essential to confirm the findings and establish trajectory and baseline predictors of progression in that cohort6.

The objectives of the present analysis were to: (1) systematically evaluate 2-year longitudinal change in clinical and dopamine transporter (123-I Ioflupane (DatScan®)) imaging (DATscan) characteristics of LRRK2 NMCs compared with HCs enrolled in the PPMI study; (2) report for the first time baseline and longitudinal data on an array of biofluid biomarkers; and (3) assess baseline predictors of progression.


Baseline demographic and DAT imaging characteristics

176 LRRK2 G2019S NMCs and 185 HCs were included in the analysis. Five NMCs and no HCs developed PD in the course of the study (Table 1). Baseline demographics, PD family history, presence of hyposmia, presence of a DAT deficit, and percent of participants meeting MDS prodromal research criteria above 80% are presented in Table 1. The demographic characteristics and comparison to HCs are consistent with our previous report6. Considering that hyposmia is a significant prodromal risk for PD in general, University of Pennsylvania Smell Identification test (UPSIT) scores are presented as raw values and by different percentile cut offs. There was no difference in severe hyposmia (≤15th percentile for age and sex) between NMCs and HCs but a higher percent of NMCs (41% vs 27%) had moderately reduced sense of smell. A minority (13%) of NMCs had a DAT deficit, which was not significantly different (p = 0.0546) compared to HCs. Only 7% of NMCs met MDS prodromal criteria.

Table 1 Demographics and baseline characteristics.

Clinical and biological characteristics of the participants who developed PD (“phenoconverters”)

5 NMCs developed clinically defined PD at the time of data analysis. Baseline demographic characteristics of these participants are presented in Table 1 and longitudinal clinical characteristics compared to the rest of the cohort are presented in Suppl. Table 4. Considering the small sample size, no statistical comparisons were performed. These participants were older and all females (Table 1). As expected, they had higher baseline and longitudinal Movement Disorders Society Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) total and subscores. 4 of 5 had baseline DAT data available and all showed DAT deficit (Table 1). Only 2 of 5 had hyposmia and 3 of 5 met MDS prodromal criteria at baseline. The diagnosis of PD was made within 12 months of baseline in 4 participants and at 48 months in 1. None of the participants developed other neurodegenerative conditions.

Longitudinal data

Clinical characteristics

Longitudinal change in a wide range of clinical characteristics in NMCs at the group level were compared to HCs (Table 2). Consistent with our previous report6, at baseline, NMCs scored significantly worse than HCs on numerous clinical scales. However, there was no significant longitudinal change in any of these measures over 2 years. Since DAT deficit signifies the presence of presynaptic striatal dopaminergic deficiency, analyses of the longitudinal change in clinical characteristics in NMCs subdivided by DAT deficit (<65% versus above) (Table 3) was performed. There was no significant group by time effect, indicating that there was no difference between groups in the longitudinal change from baseline to follow-up; though it is noteworthy that few participants had a baseline DAT deficit (13%). We repeated the same analysis dichotomizing the carriers on the baseline degree of hyposmia (≤15th percentile or above) and similarly did not observe a significant group by time effect (Suppl. Table 1).

Table 2 Change in clinical characteristics among LRRK2 G2019S carriers vs healthy controls.
Table 3 Change in clinical characteristics among LRRK2 G2019S carriers with vs without baseline DAT deficit.

DAT imaging

There also was no significant longitudinal decline in caudate (p = 0.1929), putamen (p = 0.0865) or striatum (p = 0.0910) specific binding ratios (SBRs) in NMCs. There was a trend but no significant difference in SBR values at 2-year follow-up in NMCs compared to HCs at baseline after adjusting for multiple comparisons (Suppl. Table 2).

Biofluid biomarker characteristics

We analysed biofluid biomarkers in cerebrospinal fluid (CSF) Abeta, total tau and phospho-tau; and serum urate and neurofilament light chain (NfL) (Table 4). There were no differences in any of these biomarkers between NMCs and HC at baseline and no difference in longitudinal changes between groups. Of note, CSF AD biomarkers are presented as median (range) values as well as above or below cut-offs established in AD literature for the sake of comparison and cross referencing7,8. We also ran analyses of urine bis(monoacylglycerol) phosphate (BMP) as a biomarker of lysosomal dysfunction previously shown to be elevated in LRRK2 cohorts9. Consistent with our published data, there was significant baseline elevation of all tested urinary BMP species in NMCs compared to HCs (Table 4). Relative to HCs, levels for NMCs were higher by 3.8-fold (95% CI 3.1–4.5) for total di-18:1-BMP, 6.1-fold (95% CI 5.3–7.1) for total di-18:1-BMP, and 7.4-fold (6.4–8.6) for 2,2′-di-22:6 BMP. However, there was no longitudinal change in any BMP species in NMCs (Suppl Table 3). CSF alpha-synuclein measures were not available, as they are still under analysis. The polygenic risk score did not differ between NMCs and HCs.

Table 4 Baseline and longitudinal biofluids among LRRK2 G2019S carriers vs healthy controls.

Association of biofluid biomarkers with clinical and DAT imaging characteristics

Utilizing MDS-UPDRS Part III score as a measure of evolving motor parkinsonism, none of the fluid biomarkers included in our array correlated with baseline or longitudinal change in MDS- UPDRS Part III (data not shown). We repeated the analysis looking at correlation between baseline polygenic risk score, biofluid biomarkers, and baseline DAT (mean striatum SBR). The only significant correlations were with the total di-22:6 (Spearman’s rho = −0.28; 95% CI −0.42 to −0.13) and 2,2′-di-22:6 (−0.27; −0.41 to −0.12) BMP species (Table 5). However, there was no correlation of any baseline biofluid biomarker level, including polygenic risk score and BMP, with the 2-year change in mean striatum SBR (Table 5). There was no significant effect of baseline serum urate level on change in any of the clinical characteristics (data not shown).

Table 5 Correlations between baseline biomarkers vs baseline and longitudinal change in DAT specific binding ratios among LRRK2 G2019S carriers.


We present the longitudinal clinical, DAT imaging, and biofluid biomarker data for a large LRRK2 G2019S NMC cohort compared to HCs. To our knowledge, this is the largest comprehensive longitudinal dataset reported to date. These data prompt several novel observations that are highly relevant to future PD prevention trials.

Consistent with slow progression and low penetrance, only 5/176 carriers developed clinically defined PD. These observations highlight the essential importance of identifying clinical and biological markers for the enrichment of the NMC cohort to identify those with the greatest risk of developing PD for future therapeutic interventions. We have systematically explored a number of such predictors here.

We and others have previously reported higher prevalence of PD motor and non-motor features in LRRK2 G2019S NMCs compared to HCs in cross-sectional studies3,4,5,10,11,12. However, our data do not show any significant longitudinal change in these clinical outcomes. The most likely explanation is the relatively slow progression of the phenotypic changes, consistent with incomplete penetrance of this variant and lack of risk-based enrichment in this cohort.

MDS prodromal research criteria were developed to identify participants with higher likelihood of prodromal PD13,14. MDS prodromal criteria have shown a wide range of positive predictive values (19–81%) in prodromal cohorts and perform best in REM sleep behavior disorder (RBD) cohorts, consistent with the fact that they are heavily weighted for the presence of idiopathic RBD15,16,17. Only 7% of our NMCs met MDS prodromal criteria at baseline, which may explain the lack of meaningful progression in our cohort. On the other hand, only 60% of the converters met the MDS prodromal criteria at baseline suggesting that other risk factors may underlie the progression to clinical PD in LRRK2 G2019S carriers. MDS prodromal criteria have been applied in another LRRK2 G2019S cohort and showed 47% positive predictive value over 5 years12. In aggregate, these findings argue that MDS criteria are less sensitive in LRRK2 cohorts and additional variables will be important to identify the population at the highest risk of progression. These findings also highlight the essential need for biomarker enrichment of LRRK2 G2019S NMC cohorts to identify participants at risk for progression.

The presence of DAT deficit is the strongest biomarker predictor of progression to clinically defined PD in RBD and hyposmic prodromal cohorts18,19 and should be an important enrichment biomarker for genetic NMC cohorts as well. Reduction of SBR values in LRRK2 G2019S NMCs compared to HCs was reported by other groups, though in smaller cohorts and cross-sectional observations5,4,11,20,21. Only 13% of the LRRK2 G2019S NMCs had baseline DAT deficit, which is the likely explanation for the lack of significant progression using a range of outcome measures. However, there was no significant progression on any clinical measure even in the sub-group of participants with DAT deficit (Table 3). At least three possibilities may have contributed to this observation: the relatively small number of participants with DAT deficit at baseline, short observation period (2 years), or a slower progression rate in LRRK2 G2019S NMCs compared to other prodromal cohorts. Further clarification of this observation is of essential importance for experimental therapeutics and will require longer follow-up in larger cohorts. Enrichment based on DAT deficit was not applied in the original PPMI NMC cohorts as it was essential to establish the prevalence of such in the overall study population. Current PPMI enrollment criteria include enrichment based on DAT deficit; however, the cut-off value (60–80%) will be established from the prospective longitudinal study. Our analyses here indicate a major challenge for DAT deficit-based enrichment of LRRK2 NMC cohorts. Specifically, >80% of currently enrolled LRRK2 G2019S NMCs will not qualify for the high-risk prodromal group if we apply the cut-off of <65% DAT deficit. Thus, additional enrichment strategies are required for the LRRK2 NMC prodromal group.

Hyposmia is a common prodromal feature in PD and has been used for staged enrichment of prodromal cohorts19. However, it is less common in LRRK2 PD and specifically in G2019S carriers4. Consistent with our and others’ previous reports, the prevalence of severe hyposmia in LRRK2 G2019S NMCs was low and baseline hyposmia did not associate with longitudinal change in any clinical variables. Notably, though, baseline hyposmia was associated with higher odds of baseline DAT deficit (odds ratio=3.52, 95% CI 1.10 to 11.27).

The strongest and most novel attribute of our dataset is the richness of biomarker characterization. We structured analyses to assess if any of the tested biomarkers: (1) separated NMCs from HCs; (2) demonstrated longitudinal change; or (3) were associated with baseline or longitudinal changes in DAT or motor parkinsonian features. None of the CSF or serum biomarkers tested had significant findings in any of these analyses. Confirming our previously reported data, NMCs had significantly elevated BMP levels at baseline compared to HCs9. Baseline BMP levels correlated with baseline mean striatum SBRs but did not show longitudinal change and did not correlate with the 2-year change in DAT values. In a separate analysis (not shown) there was no significant difference in BMP levels between LRRK2 NMCs and LRRK2 with PD. As such BMP is an important biomarker to study LRRK2-driven pathogenic mechanisms or modulation of LRRK2 activity by therapeutics (i.e., target modulation biomarker) but does not appear to be a biomarker of progression or risk to develop PD in prodromal individuals. CSF Abeta, total tau or p-tau levels did not differ between LRRK2 G2019S NMCs versus HCs. These indices are of interest considering that some LRRK2 carriers have tau rather than alpha-synuclein pathology22 but a direct test would require tau imaging studies in LRRK2 pathogenic variant carriers. Polygenic risk score also was not associated with the change in DAT.

We recognize that our study has several limitations. First, we do not have data on CSF monomeric or aggregated alpha-synuclein levels since the biosamples have not been fully analyzed for these markers. However, based on previous reports, we do not expect to see a significant longitudinal change in monomeric or aggregated alpha-synuclein23. Alpha-synuclein seed amplification assays (SAA; also termed PMCA or RTQuIC), while still only qualitative, have a high potential to detect prodromal alpha-synuclein pathology24,25 and thereby represent a promising biomarker for the enrichment of an at-risk population. In the PPMI cohorts, alpha-synuclein SAAs on the CSF and skin are being conducted currently and data will be shared once available. These studies could help identify the stage at which alpha-synuclein pathology develops in prodromal participants. SAA data will be of paramount importance in all prodromal cohorts, and specifically in LRRK2 carriers, to confirm the presence of alpha-synuclein driven pathology to select individuals appropriate for alpha-synuclein targeted interventions.

We and others have reported several LRRK2-associated biomarkers including blood metabolomics, urinary proteome, and inflammatory markers26,27,28,29. In addition, LRRK2-associated pharmacodynamic biomarkers have also been developed as reviewed recently27,30. Some of these distinguish LRRK2 manifest and NMCs from sporadic PD and HC. However, none have been shown to correlate with longitudinal disease progression. Importantly, specific biomarkers/panels from these studies are being tested on PPMI biosamples to assess their utility as prognostic or progression markers. There are few pilot studies showing promising data on novel functional MRI sequence imaging biomarkers that will require further testing31. Digital biomarkers offer tremendous promise for baseline enrichment and longitudinal follow-up of the NMC cohorts32. These are being collected in PPMI and other cohorts and will be reported as the data become available. Lastly, our analysis was restricted to G2019S LRRK2 carriers. While G2019S is the most common LRRK2 pathogenic variant, it is not the only one. While we identified 16 R1441G NMCs, the sample size was too small to run any comparative analysis and considering data on difference in phenotypic characteristics and progression, we decided not to combine the analysis (see Suppl. Table 5). Other studies targeting diverse ethnic populations will have to address LRRK2 variant dependent variance in progression.

In conclusion, we did not identify baseline predictors or longitudinal change in clinical, DAT, and biofluid biomarker characteristics in a large LRRK2 G2019S NMC cohort. These findings are of essential importance for the research community and highlight the challenges of designing clinical trials in LRRK2 G2019S NMCs both from the feasibility standpoint (number of participants to be screened versus the number who will qualify) and essential need for further enrichment biomarker discovery in addition to DAT deficit.

More longitudinal data and larger comprehensive biomarker panels will be necessary to identify LRRK2 G2019S NMCs at the highest risk of progression. It also will be essential to define multimodal progression trajectories that ultimately predict the development of PD in a time period amenable to future disease prevention therapeutic interventions.


Study design

Data used in the preparation of this manuscript were obtained from the PPMI database ( The aims and methodology of the study have been published elsewhere33,34. Study protocol and manuals are available at The data used for this paper were downloaded on June 30, 2020 and reflect LRRK2 NMCs enrolled between September 2013 and May 2019 from 33 participating sites worldwide.


The LRRK2 NMC cohort enrolled male or female participants age 45 years or older at baseline with a LRRK2 pathogenic variant confirmed by the Genetic Coordination Core. Participants were excluded if they had a clinical diagnosis of PD based on established diagnostic criteria35 or conditions that precluded safe performance of lumbar puncture. Recruitment of the NMC cohort was done via participating sites and a centralized recruitment initiative, described previously, specifically targeting first-degree relatives of PD patients of Ashkenazi Jewish descent36. The study was approved by the institutional review board at each site, and participants provided written informed consent. HCs were enrolled in PPMI based on previously published inclusion and exclusion criteria; notably, they were required to have a Montreal Cognitive Assessment (MoCA) score ≥27 at enrollment34. Due to the focus on short-term longitudinal change, we restricted analyses to NMCs and HCs who were assessed at baseline and at least one of the first two annual follow-up visits (i.e., at year 1 and/or year 2). As of the data freeze date used for this analysis, 36/176 LRRK2 G2019S NMCs (20%) had not completed a year 2 assessment, primarily because they were not yet due for this study visit (note: only four of these participants had withdrawn).

LRRK2 pathogenic variants testing for the NMC cohort

Genetic analysis was performed by the PPMI genetic core. Briefly, qualified individuals underwent screening for the LRRK2 G2019S pathogenic variants. Results were provided by phone by certified genetic counselors at Indiana University or by site-qualified personnel36. In addition, LRRK2 carrier data from genetic platforms including whole genome sequencing were downloaded from the PPMI database to identify pathogenic variant carriers in the broader PPMI cohort. Our analysis focused on LRRK2 G2019S NMCs (N = 176). As such, we excluded participants with other LRRK2 pathogenic variants including R1441G due to small sample size (N = 16). We also excluded 19 LRRK2 NMCs and 2 HCs who had pathogenic variants of GBA1. In addition, a polygenic risk score (PRS) was constructed from 89 loci associated with PD risk, which excluded the G2019S variant, using previously published methodology37.

Study outcomes

All participants enrolled in PPMI undergo a standard test battery of assessments described in detail previously7,8. The clinical battery relevant to this analysis included the MDS-UPDRS, MoCA, 15-item Geriatric Depression Scale, Scale for Outcomes in PD-Autonomic (SCOPA-AUT), State and Trait Anxiety Scale (STAI), Modified Schwab and England Activities of Daily Living Scale, Questionnaire for Impulsive-Compulsive Disorders in Parkinson’s Disease (QUIP), Epworth Sleepiness Scale (ESS), RBD Screening Questionnaire (RBDSQ) and UPSIT. UPSIT was analyzed based on age/sex-specific percentiles derived from normative data. All participants are expected to undergo DATscan at baseline and then every other year to assess dopamine transporter (DAT) binding analyzed according to the PPMI imaging technical operations manual ( We applied quantitative DATscan analysis using previously described methods to determine the minimum putamen specific binding ratio (SBR) and <65% age/sex-expected lowest putamen SBR was used as a cut-off for DAT deficit34. Of note, HCs were enrolled based on the visual assessment of DATscan (normal versus abnormal) as per regulatory approval of DATscan38. Visual and quantitative DAT have high but not 100% correlation39. Thus, some HCs had DAT deficit based on the quantitative analysis. Prodromal risk score was calculated based on previously published criteria14. PPMI collects an array of biofluid biomarkers ( Not all such measures are currently available since the analytes are measured in batches. The available biofluid biomarkers included CSF Abeta, total tau and phosphor(181)-tau; serum urate and NfL; and urine BMP. The analytical methods for monitoring these biomarkers have been published previously7,9,18.


Progression to a clinically defined diagnosis of PD or other degenerative disorder (phenoconversion) is determined by the site investigator based on established diagnostic criteria. The clinical and biomarkers dataset of every phenoconverted participant is reviewed by a panel of experts (concensus committee) who adjudicate on the final diagnosis.

Statistical methods

Statistical analyses were performed using SAS version 9.4 (SAS/STAT 15.1; SAS Institute Inc., Cary, NC) and R version 4.1.3 (R Foundation for Statistical Computing, Vienna, Austria). Baseline demographics were compared using chi-squared and t tests (Fisher’s exact and Wilcoxon rank-sum tests where appropriate) at α = 0.05. Cross-sectional comparisons of biofluid and imaging biomarkers, conducted using the SAS/STAT CAUSALTRT procedure, applied inverse probability weighting methods to control for age and sex. Within LRRK2 G2019S NMCs, associations between baseline biofluids and striatal DAT SBR were assessed using Spearman partial rank-order correlations controlling for age and sex, with 95% confidence intervals (CIs) obtained using Fisher’s z transformation.

All models were fit using generalized estimating equations (GEE) with an autoregressive correlation structure. For most clinical and biological outcomes, group mean trajectories did not support a linear trend assumption because the outcomes in years 1 and 2 were relatively comparable. Instead, we treated year 1 and year 2 as analogous “follow-up” visits (i.e., considered them equivalent with respect to time). To compare longitudinal changes between groups, GEE with inverse probability weighting to control for age and sex, conducted using the R geepack package (because an analagous SAS procedure was unavailable), modeled the change from baseline at follow-up (continuous outcomes) or binary response at follow-up (categorical outcomes) while adjusting for the baseline value. Within LRRK2 G2019S NMCs, the effects of baseline DAT deficit and hyposmia on longitudinal clinical changes were assessed using GEE, conducted using the SAS/STAT GENMOD procedure, modeling the change from baseline at follow-up (continuous outcomes) or binary response at follow-up (categorical outcomes) while adjusting for the baseline value. Additionally, change from baseline to follow-up in striatal DAT SBR (measured at baseline and year 2 only) and urine BMP (measured at baseline, year 1, and year 2) were assessed using one-way t tests and GEE, respectively.

For all applicable analyses, serum NfL and urine BMP values were log-transformed and CSF biomarkers were assigned to quintile categories or dichotomized using established cutoffs7. For ease of interpretation, the PRS was normalized by rescaling to mean=0 and standard deviation=1; also, analyses were adjusted for age, sex, and the first five principal components of the genetic data (representing population structure).

All statistical tests were two-sided. To account for multiple comparisons, we applied a family-wise error rate to each set of analyses. Specifically, a Bonferroni correction, computed as 0.05/number of family-wise hypotheses tested per table, was applied to Tables 25 and applicable supplementary tables.