LRRK2 and GBA1 variant carriers have higher urinary bis(monacylglycerol) phosphate concentrations in PPMI cohorts

We quantified concentrations of three isoforms of the endolysosomal lipid, bis(monoacylglycerol) phosphate (BMP) in the urine of deeply phenotyped cohorts in the Parkinson’s Progression Markers Initiative: LRRK2 G2019S PD (N = 134) and non-manifesting carriers (NMC) (G2019S+ NMC; N = 182), LRRK2 R1441G PD (N = 15) and R1441G+ NMC (N = 15), GBA1 N409S PD (N = 76) and N409S+ NMC (N = 178), sporadic PD (sPD, N = 379) and healthy controls (HC) (N = 190). The effects of each mutation and disease status were analyzed using nonparametric methods. Longitudinal changes in BMP levels were analyzed using linear mixed models. At baseline, all LRRK2 carriers had 3–7× higher BMP levels compared to HC, irrespective of the disease status. GBA1 N409S carriers also showed significant, albeit smaller, elevation (~30–40%) in BMP levels compared to HC. In LRRK2 G2019S PD, urinary BMP levels remained stable over two years. Furthermore, baseline BMP levels did not predict disease progression as measured by striatal DaT imaging, MDS-UPDRS III Off, or MoCA in any of the cohorts. These data support the utility of BMP as a target modulation biomarker in therapeutic trials of genetic and sPD but not as a prognostic or disease progression biomarker.


INTRODUCTION
Rare missense mutations in the gene encoding a multidomain protein, leucine-rich repeat kinase 2 (LRRK2), are a major cause of autosomal dominant Parkinson's disease (PD) 1,2 , whereas common variants in the LRRK2 gene are associated with sporadic PD 3,4 . Variants in GBA1, the gene encoding the lysosomal hydrolase, glucocerebrosidase (GCase), represent the most common genetic risk factors for PD and related synucleinopathies 5 . The precise mechanisms by which LRRK2 or GBA1 mutations cause PD is not entirely clear. However, recent cell biological studies have begun to implicate endolysosomal trafficking and lysosomal dysfunction as key pathogenic mechanisms associated with LRRK2-and GBA1associated PD [6][7][8] . In aggregate, these studies have spurred a robust pipeline of therapeutics targeting LRRK2 and GCase 9,10 . We have recently shown that LRRK2 kinase negatively regulates the activity of lysosomal GCase in human induced pluripotent cells 11 , further demonstrating a convergence between LRRK2 and GCase pathways 12 . It has been suggested that LRRK2-and GCase-targeted therapies may benefit sporadic PD (sPD) also on the basis of an increase in LRRK2 kinase activity 13 and a deficiency in GCase 14 in brains of sPD cases, though the former data have not been replicated broadly 6,15 . The heterogeneity of sPD and low penetrance of LRRK2 16 and GBA1 5 variants for PD necessitate a patient enrichment strategy to support clinical development of therapies. Thus, biomarkers that could detect endolysosomal dysfunction may inform patient enrichment strategies for LRRK2and GCase-targeted therapeutics. Additionally, such biomarkers could also constitute pharmacodynamic markers to demonstrate relevant target modulation at clinically used doses.
Several lines of evidence indicate that bis(monoacylglycerol) phosphate (BMP), previously called LBPA (lysophosphosphatidic acid), is an atypical phospholipid that regulates and is affected by endolysosomal functions [17][18][19] . We recently examined 11 isoforms of BMP in the urine from two independent LRRK2 cohorts and observed an elevation in all BMP isoforms assayed in LRRK2 mutation carriers compared to non-carriers 20 . A correlational analysis indicated that total di-18:1-BMP and total di-22:6-BMP most strongly discriminated the LRRK2 mutation carriers from noncarriers. Furthermore, levels of di-22:6-BMP and its 2,2′ isoform negatively correlated with MoCA scores in the carriers. However, this study had two major shortcomings. One, it was performed in smaller cross-sectional cohorts that lacked deep phenotyping data which precluded our ability to assess correlations between BMP levels and clinical progression. Second, some participants in the LRRK2 cohort also carried GBA1 pathogenic variants, which may have confounded outcomes and data interpretation. Thus, further investigation of urinary BMP as a biomarker that could facilitate PD therapeutic development is needed. On the basis of our previous study, we focused the current study on the assessment of total di-18:1-BMP, total di-22:6-BMP, and 2,2′-di-22:6-BMP.
The objectives of the present study were to utilize the deeply phenotyped and large longitudinal observational cohorts from the Parkinson's Progression Markers Initiative (PPMI) to extend our previous studies 20 and assess: (a) effects of LRRK2 and GBA1 pathogenic variants on baseline urinary BMP levels in PD manifesting and non-manifesting carriers (NMC), (b) longitudinal changes in BMP levels in LRRK2 carriers, and (c) whether baseline BMP levels predict disease progression in LRRK2, GBA1 or sPD cohorts.

RESULTS
Baseline BMP concentrations are higher in LRRK2 cohorts compared to sPD and HC Demographics, baseline clinical characteristics and mean striatal specific binding ratios (SBR) computed from the dopamine transporter (DaT) SPECT imaging data (DaTscan) for PD manifesting and NMC of LRRK2 G2019S+ and R1441G+ were compared to sPD and HC groups (Table 1a, b, respectively). Neither G2019S+ nor R1441G+ carriers with PD or NMC differed significantly in age from sPD or HC, respectively. Consistent with PPMI's enrollment strategy, disease duration at baseline was significantly longer for both G2019S+ and R1441G+ PD manifesting groups compared to the sPD group. Consequently, the LRRK2 G2019S+ PD group showed greater deficits in mean striatal DaT binding and MoCA than sPD. Similarly, the R1441G+ PD group showed greater deficits in MoCA and UPDRS III Off than the sPD cohort. The sex distribution was also different between sPD and LRRK2 genetic cohorts. Baseline BMP concentrations for each LRRK2 genotype were compared across the four groups after adjusting for age and sex. Consistent with our previous report 20 , concentrations of all three BMP isoforms were significantly higher (3-7-fold) in LRRK2 G2019S+ carriers compared to HC and sPD, but the levels did not differ between the PD manifesting and NMC groups (Table 1a and Fig. 1). This effect of LRRK2 genotype on BMP elevation, irrespective of disease status, was also seen in R1441G+ carriers (Table 1b and Fig. 1). A comparison of baseline BMP levels in G2019S+ and R1441G+ sub-cohorts showed statistically significant overall effect and higher levels in R1441 G+ NMC sub-cohort compared to G2019S+ NMC (Supplemental Table 1). Unlike the genetic cohort, the sPD group did not show an elevation in any BMP isoform when compared to HC (Table 1a).
Longitudinal Changes in BMP levels in LRRK2 mutation carriers Given the robust increases in baseline BMP levels associated with LRRK2 genotypes, we examined whether BMP isoforms change over time in LRRK2 G2019S+ or R1441G+ individuals manifesting PD and LRRK2 NMCs. Table 2a, b show annualized change in BMP levels and an assessment of whether disease status affects the longitudinal change in LRRK2 G2019S+ and R1441G+ carriers (note the log of the BMP level was modeled due to the skewed distribution of the data). In the G2019S+ PD group, di-18:1-BMP increased longitudinally whereas the R1441G+ individuals showed quantitatively greater changes in all three BMP isoforms in the NMC group. However, there was no statistically significant difference in longitudinal BMP changes in the PD manifesting versus NMC groups for either LRRK2 genotype. Supplemental Table 2 shows the raw concentrations of each BMP isoform at baseline, year 1 and year 2 by disease status for the two LRRK2 genotypes. These data can be useful to facilitate power calculations for future studies.
Baseline BMP concentrations are higher in the GBA1 N370S+ cohort, but not in other GBA1 genotypes compared to sPD and HC Given the previously reported convergence in cellular processes affected by LRRK2 and GCase, we assessed whether risk variants of GBA1 affect urinary BMP levels. Table 3 shows demographics and baseline characteristics for GBA1 N409S+ PD (N = 76) and NMC (N = 178) compared with the sPD (N = 379) and HC (N = 190) groups. There was no overall difference in age between the GBA1 N409S+ groups, HC, or sPD. As with the LRRK2 G2019S+ /R1441G + PD groups, the GBA1 N409S+ PD had significantly longer disease duration and associated higher MDS-UPDRS III Off and lower MoCA scores when compared to sPD. However, unlike the lower mean striatal DaT SBR in the G2019S+ PD group when compared to sPD (Table 1a), the GBA1 N409S+ PD group did not differ in striatal SBR from sPD. Interestingly, total di-22:6-BMP and 2,2′ di-22:6-BMP isoforms were higher in GBA1 N409S+ individuals (PD and NMC) but not total di-18:1-BMP. However, the levels of di-22:6-BMP isoforms did not differ between those with and without PD (Table 3 and Fig. 2), once again indicating that BMP is a trait marker even in the GBA1 carriers. It is noteworthy that the magnitude of the increase in BMP concentration in N409S+ was 40%, much smaller than the 3-7X elevation seen with the LRRK2 genotypes.
We also examined baseline BMP levels in three additional GBA1 sub-cohorts, though each had a small number of individuals: those associated with severe Gaucher disease, L483P (N = 8 with PD and N = 3 NMC), and a group of rare variants combined (IVS2+ 1 G > A, L29Afs18, T408M+ R159W, R502C; N = 6 with PD and N = 4 NMC) as well as the PD-specific common risk variant, E365K (N = 19 with PD and N = 5 NMC). Surprisingly, there was no overall difference in baseline BMP levels in these GBA1 sub-cohorts when compared to sPD, and HC (Supplemental Table 3a-c).
A small number of LRRK2 G2019S carriers also carried the GBA1 N409S variant (N = 5 with PD and N = 15 NMC). Although the small numbers do not allow definitive conclusions, BMP levels in these individuals had a similar pattern to those seen in LRRK2 G2019S+ subjects, i.e., BMP levels were elevated in the compound mutation carriers, but did not differ between those with and without PD (Supplemental Table 4). When comparing BMP levels in the LRRK2 G2019S+ with those with both LRRK2 G2019S+ and GBA1 N409S+, there was no overall statistical difference in concentrations of any BMP isoform indicating that there is no interaction between these genotypes in the sub-cohorts studied.
Baseline BMP levels do not predict a decline in striatal DaT and clinical outcomes over 5 years Our final objective was to examine whether baseline BMP isoforms could predict a longitudinal decline in striatal DaT, MDS-UPDRS III Off, or MoCA in the two largest genetic groups, LRRK2 G2019S+, and GBA1 N409S+ as well as the sPD and HC group (note that the HC group is excluded from the modeling of the DaT data since this group did not undergo DaT SPECT following baseline visit). Overall, we did not observe a relationship between any isoform of BMP at baseline and the change in striatal DaT SBR, MDS-UPDRS III Off, or MoCA in either genetic cohort or sPD and HC group (Table 4a-c). Since LRRK2 R1441G is thought to be more pathogenic than LRRK2 G2019S, we explored whether baseline BMP would predict clinical or radiographical decline and failed to see an association (Table 4a-c). The summary statistics for each outcome are shown in Supplementary Table 5a-c.

DISCUSSION
Here we present the largest dataset of analysis of BMP isoforms in individuals with the most common pathogenic/risk variants in LRRK2 (G2019S and R1441G) and GBA1 (N409S) as well as other PD-associated GBA1 variants and compared them to sPD and HC. In the LRRK2 carriers, our results indicate that: (i) baseline BMP levels are elevated in both G2019S+ and R1441G+ individuals, (ii) the elevation in baseline BMP levels did not differ between variant carriers+ PD and NMC, (iii) baseline total di-18:1-BMP levels are higher in R1441G+ NMC when compared to G2019S+ NMC subcohort but not in PD manifesting groups for the two LRRK2 BMP, first identified in 1967, is a negatively charged glycerophospholipid with an unusual structure and is localized almost exclusively to late endosomal and lysosomal membranes 19 . BMPs contain two fatty acid acyl chains, the composition of which is celltype specific and contributes to the biochemical functions of the various BMP species. Here we examined levels of total di-18:1-BMP and di-22:6-BMP on the basis of findings from our previous study 20 showing that these two analogs are the most abundant in the urine and also most significantly associated with the LRRK2 G2019S genotype. In addition, we examined the 2,2′ isomer of di-22:6-BMP since it is the major isoform in the late endosomes that is more active at forming multivesicular bodies than other BMP isoforms 21 and was associated with worse cognition as measured by MoCA 20 . Key roles of BMPs include lysosomal stability, hydrolase activity, vesicle formation, and endosomal trafficking 19 . Furthermore, BMP accumulates in several lysosomal storage disorders, including Niemann-Pick disease 22,23 , a lysosomal disorder caused by diminished acid sphingomyelinase levels. Interestingly, variants in the SMPD1 gene encoding acid sphingomyelinase are also risk factors for PD 24 . BMP accumulation has also been reported in Gaucher's disease 19 , as well as iatrogenic phospholipidosis 25,26 . Thus, BMP appears to be an effective marker of endolysosomal homeostatic functions that could be interrogated in PD cohorts and controls.
Here we observed 4-7× elevation in baseline BMP isoform levels in the LRRK2 G2019S+ group compared to HC, which is consistent with our previous report 20 . Additionally, we observed a similar magnitude of increase in those with the LRRK2 R1441G variant despite a smaller number of individuals in this sub-cohort of the PPMI study. However, in the NMC sub-cohorts for the two LRRK2 variants, urinary total di-18:1-BMP levels were significantly higher in the R1441G+ than in G2019S+ individuals. It is noteworthy that disease status did not affect BMP levels in either G2019S+ or R1441G+ individuals, indicating that urine BMP is a trait, but not a state, marker. In the G2019S+ carriers, urinary BMP isoforms remained stable over two years in both PD-manifesting and non-manifesting individuals. On the other hand, R1441G+ carriers showed a significant and somewhat greater (~15%) annualized increase in BMP concentration in the NMC group, although there was no statistical difference in the longitudinal change between those with and without PD. R1441G is a rarer but more penetrant variant of LRRK2 than G2019S. Whether the higher baseline levels of total di-18:1-BMP or the apparent higher rate of increase in BMP levels in R1441G is due to its greater pathogenicity would require further studies with a larger sample size. Another novel finding of the present study is that GBA1 N409S was also associated with significantly higher levels of total di-22:6 and 2, 2′-di-22:6-BMP at baseline, albeit the magnitude of increase was only~30-40% when compared to those without the variant. Although the levels of total di-18:1-BMP were numerically higher in the GBA1 N409S carriers, they did not reach statistical significance. Whether the apparent isoform-specific changes in LRRK2 vs. GBA1 variant carriers reflect cell-type specific effects of these variants or a reflection of the relatively smaller magnitude of change associated with the N409S variant remains to be seen. Future cell biological studies of mechanisms associated with BMP regulation by LRRK2 and GCase activity could shed light on this issue. As with LRRK2, there was no difference between baseline BMP concentrations for N409S+ with PD as compared to NMCs, indicating again that urine BMP is a trait marker of the N409S genotype. GBA1 N409S significantly reduces GCase activity and causes generally mild and non-neuronopathic Gaucher disease and a mild clinical phenotype in PD 27,28 . Hence, we also examined BMP concentrations in the urine from Gaucher disease-associated severe GBA1 variant carriers. Unexpectedly, there was no elevation in baseline BMP levels in these PPMI participants. It is possible that the lack of association of the severe Gaucher disease GBA1 variants to urinary BMP levels was due to the small numbers of individuals with L483P (N = 10) and the group of rare point/frameshift mutations (N = 11). Another GBA1 variant in the PPMI study is E365K (N = 24), a PD risk factor that does not cause Gaucher disease in homozygote carriers. We did not observe an elevation in BMP levels in E365K+ carriers either. Given the high prevalence of this variant in the United States population, it would be important to examine BMP levels in a larger cohort of E365K+ to ensure that the limited data available in PPMI was not a contributing factor. In fact, assessing BMP levels across GBA1 variants in larger cohorts is critical to assess whether a correlation exists between GCase activity and BMP concentrations 28 .
In our previous paper 20 , we proposed that LRRK2-mediated phosphorylation of Rab substrates 8 may increase BMP concentrations in biofluids by affecting the biogenesis, motility, or extracellular release of endolysosomal vesicles. Indeed, a recent paper indicates that inhibition of LRRK2 kinase activity reduces the release of BMP-containing vesicles from the kidney 29 . Thus the observed increase in urinary BMP levels in LRRK2 G2019S+ and R1441G+ individuals is likely due to the higher kinase activity of these pathogenic variants. On the other hand, the GBA1 N409S variant reduces GCase activity and thereby induces lysosomal stress, which may, in turn, affect lysosomal vesicular release via a mechanism distinct from LRRK2-mediated BMP release. Ongoing cell biological studies investigating the regulation of biosynthesis, metabolism, and secretion of BMP by LRRK2 and GCase would provide insights into the mechanisms associated the observed increases in urinary BMP levels reported here. In our previous unbiased proteomics study also, we observed minimal overlap in lysosomal proteins affected by LRRK2 G2019S and GBA1 N409S variants, suggesting that the two variants affect distinct cell biological pathways 30 . Regardless of the mechanism, our observation that PD-associated variants in LRRK2 and GBA1 increase urinary BMP levels further indicates that the LRRK2 and GCase pathways converge on endolysosomal dysfunction. Our previous cell biology studies in iPS-derived dopaminergic neurons have also shown a convergence between LRRK2 and GCase 11 . There were only twenty individuals with both G2019S+ and N409S+ mutations (5 with PD), precluding our ability to reach firm conclusions around the interaction between these genotypes in the regulation of urinary BMP. However, numerically, these double mutation carriers showed a similar pattern of BMP elevation as seen in LRRK2 variant carriers without an overall statistical difference when compared to G2019S+ sub-cohort. We examined BMP levels in the sPD cohorts since recent studies have indicated higher LRRK2 kinase activity 13 and reduced GCase activity in sporadic PD 14 . Unlike the genetic cohorts studied here, the sPD group did not differ in baseline BMP levels from HCs. Thus, unfortunately, urinary BMP assessment would not inform the enrichment of sPD patients with endolysosomal deficits for therapeutic development. Although disappointing, this observation is consistent with our unbiased proteomic studies on urine 30 and cerebrospinal fluid 31 where we observed an enrichment in alterations in lysosome-associated proteins in sPD compared to HC in the cerebrospinal fluid but not the urine, despite seeing significant changes in both biometrics in LRRK2 G2019S+ individuals 30,31 . We also evaluated whether baseline BMP would prognosticate PD progression assessed by MDS-UPDRS III Off, MoCA, or striatal DaT SBR in either the genetic cohorts or the sPD cohort. We did not see an association between any BMP isoform and PD progression, but it is noteworthy that the LRRK2 and GBA1 genetic cohorts in PPMI show the minimal progression on these clinical and radiometric outcomes over 5 years since enrollment.
The source(s) of urinary BMP in the urine is an area of active research although recently direct secretion from kidneys has been suggested 30 . However, urinary BMP has been studied as a target modulation biomarker of LRRK2 kinase inhibitors in preclinical 32,33 and clinical studies 29 . Here we demonstrate that in the LRRK2 G2019S carriers, BMP isoforms remain stable over two years, strengthening its utility as a pharmacodynamic biomarker. To our knowledge, alterations in BMP isoforms have not been examined as a biomarker of GCase-targeted therapies, either preclinically or clinically. It would be interesting to study whether GCase activation can restore BMP levels in GBA1 mutation carriers or in cellular and animal models of GCase deficiency. Clinical development of such therapeutics would be facilitated by biomarkers that enable the assessment of pharmacodynamic effects and patient enrichment.
In conclusion, our study shows that urinary BMP is a trait marker of LRRK2 G2019S, LRRK2 R1441G, and GBA1 N409S variants with more robust elevations in BMP concentrations in LRRK2 than GBA1 carriers. The relative stability of BMP over two years in G2019S+ individuals strengthens its reported utility as a target modulation biomarker. The data provided here could facilitate power calculations for the assessment of BMP as a pharmacodynamic biomarker for LRRK2 or GCase-targeted therapies. However, we were not able to see the prognostic utility of baseline BMP on PD progression monitored by MDS-UPDRS III Off, MoCA, or striatal DaT imaging in either genetic or sPD cohorts of PPMI, precluding its use for patient enrichment or monitoring disease progression. The overall test was non-significant thus pairwise comparisons were not computed.
K.M. Merchant et al.

Participants and biospecimens included in the analyses
Urine aliquots from four cohorts classified by the presence/ absence of pathogenic gene variants were analyzed: healthy controls (HC) and sPD without pathogenic variants, LRRK2 mutation carriers with G2019S (G2019S+) or R1441G (R1441G+) manifesting PD and NMC and GBA1 N409S carriers (N409S+) with PD and NMC. We focused the majority of statistical modeling on LRRK2 (G2019S+, R1441G+) and GBA1 N409S+ carriers. Note that for the purpose of these analyses, GBA1 E365K and T408M were not considered pathogenic and a few individuals with one of these variants were present in G2019S+, N409S+, sPD, or HC cohorts analyzed here (see Supplementary Table 6). For GBA1 variants other than N409S, we generated Descriptive Statistics only because of relatively small sample sizes. The additional GBA1 variants analyzed are the pathogenic variant, L483P, a group of rare pathogenic variants combined together due to a small sample size (IVS2 + 1 G > A, L29Afs*18, T408M + R159W, and R502C) and the common risk variant, E365K. For the Descriptive Statistics, we have provided the specific variant information by disease status in the respective tables. Only those subjects who had baseline BMP data and non-missing genetic information were included in the analyses. In addition, the LRRK2+ cohort was analyzed for longitudinal changes (baseline, year 1 and year 2) in BMP levels.

Study outcomes
All participants enrolled in PPMI undergo a standard test battery of clinical assessments described in detail previously 7,8 . In the current analyses, we focused on motor and cognitive outcomes using the Movement Disorders Society-Unified Parkinson's Disease Rating Scale Part III OFF (MDS-UPDRS III Off) and Montreal Cognitive Assessment (MoCA) data up through year 5. In addition, we used dopamine transporter (DaT) SPECT imaging data (DaTscan) collected at baseline and then every other year. Specifically, mean striatal specific binding ratios (SBR) computed as described previously 33 were incorporated into the analyses.
Quantitative assessment of BMP isoforms in the urine BMPs can exist in three geometrical isoforms (2, 2′-, 2, 3′-, and 3, 3′-BMP), which may influence their functional properties 36,37 . Based on our previous observation that among all the isoforms of BMP tested, di-18:1-BMP and total di-22:6-BMP most strongly Quantitation was performed using authentic di-22:6-BMP and di-18:1-BMP reference standards. Internal standards were used for each analyte reported. The intensities of the analytes and internal standards were determined by the integration of extracted ion peak areas using AnalystTF 1.7 and MultiQuant 3.0 software. Calibration curves were prepared by plotting the peak area ratios for each analyte to internal standard versus concentration. The model for the calibration curve was linear with (1/x 2 ) weighting. Each urine aliquot was assayed also for creatinine concentration by a colorimetric assay (method of Jaffé) with Parameter Creatinine Assay test reagents (R&D Systems, Minneapolis, MN)  20,26 . Concentrations of urine BMPs (ng/mL) were normalized to the concentration of urine creatinine and reported as ng/mg creatinine.

Statistical analyses
Descriptive statistics of demographic and baseline characteristics were computed for each genetic variant group by disease status. For quantitative Gaussian characteristics, the means for each variant and disease status were modeled using Welch's one-way analysis of variance (ANOVA). The overall F-test was examined for significance. If a significant difference was present, pairwise differences of interest were examined using Student's t-test, assuming unequal variances. For quantitative non-Gaussian characteristics (except BMP), the Kruskal-Wallis test was used to examine an overall difference among groups. If a difference was present, pairwise differences of interest were examined using a Mann-Whitney U-test. Baseline BMP levels were compared using a rank-based linear model with adjustment for age and sex using a similar protected approach. For qualitative variables, a chi-square test was used to analyze proportional differences among the cohorts. A Bonferroni correction was applied to all pairwise comparisons for a specific characteristic as indicated in the result tables.
Longitudinal urine BMP levels between PD and NMC mutation carriers among the LRRK2+ cohort were compared using linear mixed models (LMMs). The log of the BMP level was modeled due to the skewed distribution. We included fixed effects of disease status (PD vs NMC), time (months), and interaction of disease status and time with adjustment for age and sex. We report the estimated annualized change in log BMP for PD and NMC as well as the difference between PD and NMC groups.
To assess the effect of baseline BMP measures on the longitudinal change in MoCA and MDS-UPDRS III Off within a specific variant group (G2019S+, R1441G+, or N409S+) or the sPD and HC combined group, we used longitudinal Tobit analysis. A separate model was fit for each group and included participants from both disease states. The models were fit using the SAS procedure Proc NLMixed with Adaptive Gaussian Quadrature (15 quadrature points). The parameter values from an LMM with a random intercept were used as starting values. For examining the effect of baseline BMP levels on mean striatal DaT SBR over time, we used LMMs. Fixed effects for time and baseline BMP, as well as the interaction of time and baseline BMP, were included in the longitudinal models for MoCA, MDS-UPDRS III Off, and mean striatal DaT SBR. All models adjusted for disease status, age, and sex. The MoCA models also adjusted for education. Akaike's Information Criterion (AIC) was used to determine the inclusion of an interaction of disease status and time in each model. The main effect of interest was the interaction of time and baseline BMP.
Analyses use a significance level of 0.05 unless otherwise stated and all statistical tests were two-sided. Analyses were performed using SAS software Version 9.4 (SAS Institute, Cary, NC).

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.