GPNMB biomarker levels in GBA1 carriers with Lewy body disorders

Abstract

Background:

The GPNMB SNP rs199347 and GBA1 variants both associate with Lewy body disorder (LBD) risk. GPNMB encodes glycoprotein nonmetastatic melanoma protein B (GPNMB), a biomarker for GBA1-associated Gaucher’s disease (GD).

Objectives:

Determine whether GPNMB levels (1) differ in LBD with and without GBA1 variants and (2) associate with rs199347 genotype.

Methods:

We quantified GPNMB levels in plasma and CSF from 124 LBD individuals with one GBA1 variant (121 plasma, 14 CSF), 631 LBD individuals without GBA1 variants (626 plasma, 41 CSF), 9 neurologically normal individuals with one GBA1 variant (plasma), and 2 individuals with two GBA1 variants (plasma). We tested for associations between GPNMB levels and rs199347 or GBA1 status.

Results:

GPNMB levels associate with rs199347 genotype in plasma (p=0.022) and CSF (p=0.007), but not with GBA1 status.

Conclusions:

rs199347 is a protein quantitative trait locus (pQTL) for GPNMB. GPNMB levels are unaltered in individuals carrying one GBA1 variant.

INTRODUCTION

Parkinson’s disease (PD) is a neurodegenerative disorder characterized by progressive motor symptoms that frequently manifest alongside cognitive decline. Both common genetic variants with small effects and rarer variants with stronger effects have been associated with risk for PD (1). For example, genome-wide association studies (GWAS) have linked 90 loci to PD risk (2). From these GWAS-nominated common variant risk loci, the sentinel single-nucleotide polymorphism (SNP) rs199347 near GPNMB associates with PD risk and with RNA expression levels of GPNMB (3,4). GPNMB encodes glycoprotein nonmetastatic melanoma protein B (GPNMB), a multifaceted protein that is widely expressed throughout the body, in both a transmembrane and secreted form (3,4). Colocalization analyses support GPNMB as the target gene for the rs199347 locus, with higher expression linked to PD risk. Mechanistic studies also implicate GPNMB in PD pathogenesis (3–7). Specifically, GPNMB interacts with alpha-synuclein and may be necessary and sufficient for the uptake of fibrillar alpha-synuclein into cells (3). Moreover, GPNMB levels are elevated in PD brain and plasma (3,5).

GPNMB is also a biomarker for the lysosomal storage disorder Gaucher’s disease (GD) (8,9). GD is an autosomal recessive disorder caused by mutations in GBA1, which encodes glucocerebrosidase (GCase), an enzyme that breaks down glucosylceramide and glucosylsphingosine (10). Plasma and CSF levels of GPNMB protein can reliably distinguish between those with GD and healthy controls, and they correlate with disease severity (8,9). In GD, GBA1 mutations are classified into severe variants (which cause neuronopathic forms of GD) and mild variants (which cause non-neuronopathic GD) (10,11). Common GBA1 variants that do not result in GD also exist (10).

In addition to causing GD, GBA1 variants have been linked to PD, as well as other LBD such as dementia with Lewy bodies (DLB) (12). Variants in GBA1 are, in fact, the most common genetic risk factor for PD, found in 5-15% of PD patients, with up to 30% of carriers developing PD by age 80 (13–15). Intriguingly, mild GBA1 mutations, severe GBA1 mutations, and non-GD-causing GBA1 variants all confer risk for PD when inherited on one or both alleles (10,11). Because GBA1 and GPNMB variants are both implicated in PD risk, with GBA1 variants implicated in DLB as well, and because GPNMB levels are elevated in GD, we sought to characterize GPNMB in GBA1-associated LBD. Specifically, we evaluated GPNMB protein levels in the plasma and CSF from GBA1 variant carriers with LBD, comparing them to LBD individuals without GBA1 variants (GBA-neg LBD).

MATERIALS AND METHODS

Samples

Plasma and cerebrospinal fluid (CSF) samples were collected from individuals with LBD and neurologically normal controls (NC). Specifically, samples were obtained from 743 PD, 3 DLB, 11 individuals with parkinsonism not otherwise specified (PNOS), and 9 NC (Table 1). CSF and plasma samples were collected from individuals enrolled in ongoing research studies at the University of Pennsylvania (Penn), as previously described (15–17). PD individuals met UK Brain Bank criteria for diagnosis (18), while DLB individuals met McKeith criteria for diagnosis (19). Informed consent was obtained for all participants under approval by the University of Pennsylvania Institutional Review Board.

Table 1.

Participant demographics, GBA1 status, and clinical diagnosis

	Plasma Cohort 1				Plasma Cohort 2			CSF Cohort
GBA1 Variant	None	Risk	Mild	Severe	Risk	Mild	Severe	None	Risk	Mild	Severe
n	626	47	20	8	22	15	9	41	9	2	3
sex (% male)	66.3%	61.7%	75.0%	37.5%	68.2%	80.0%	55.6%	51.2%	55.6%	50.0%	33.3%
age	67.7 ±9.1	67.6 ±7.8	67.4 ±10.0	65.0 ±12.2	66.0 ±8.3	65.7 ±11.1	61.4 ±8.1	65.9 ±6.4	69.8 ±5.0	64.0 ±1.4	59.0 ±7.2
disease duration	6.9 ±6.0	6.5 ±6.6	7.0 ±6.0	3.8 ±4.5	6.8 ±4.9	9.0 ±5.3	6.9 ±4.9	7.2 ±4.2	8.0 ±4.8	5.5 ±2.12	7.3 ±6.7
Diagnosis
PD	616	47	19	8	21	15	8	41	9	2	2
DLB	1	0	0	0	1	0	1	0	0	0	1
PNOS	9	0	1	0	0	0	0	0	0	0	0
rs199347
AA	36.3%	38.3%	40.0%	25.0%	22.7%	33.3%	66.7%	24.4%	22.2%	50.0%	33.3%
GA	44.4%	48.9%	55.0%	37.5%	50.0%	46.7%	22.2%	58.5%	33.3%	50.0%	66.7%
GG	19.3%	12.8%	5.0%	37.5%	27.3%	20.0%	11.1%	17.1%	44.4%	0.0%	0.0%

Age and disease duration are presented as mean ± standard deviation. Table does not include 2 GBA1 homozygous mutation carriers (1 PD and 1 PNOS) or 9 neurologically normal controls.

Abbreviations: PD, Parkinson’s disease; DLB, Dementia with Lewy Bodies; PNOS, Parkinsonism not otherwise specified.

Genotyping

Individuals were screened for 14 GBA1 variants (Supplementary Table 1) and genotyped at rs199347 using SNPtype assays designed using the D3 Assay Design Tool (Standard BioTools, San Francisco, CA), as previously described (15). The identified GBA1 variants were confirmed by Sanger sequencing or Taqman assay.

GPNMB quantification

Human plasma and CSF samples were tested for GPNMB protein levels using the Human Osteoactivin/GPNMB DuoSet ELISA (DY2550, R&D Systems) according to the manufacturer’s instructions. Plasma and CSF samples were diluted 1:30 and 1:1, respectively, and run in duplicate. Optical density was measured at absorbance of 450nm on a microplate reader (Cohort 1 – Tristar LB 941, Berthold Technologies; Cohort 2 – FLUOstar Omega, BMG technologies). Samples with a coefficient of variation (CV) greater than 20% or optical density above or below the limits of detection set by the standard curve were excluded. The log₁₀GPNMB concentration for each sample was determined on a 4-parameter logistic standard curve. Homozygous GBA1 variant carriers were used as a positive control on each ELISA plate in Cohort 1, and GPNMB measures for these individuals were averaged across all runs. In evaluating performance characteristics of this ELISA with replicate controls, we found it reliable for relative, but not absolute, quantification in our hands. Thus, only samples assayed in the same run were compared to one another in our analyses.

Statistical Analysis:

Pairwise comparisons were performed using Mann-Whitney U tests. We evaluated impact of GBA1 variant and rs199347 genotype on GPNMB levels using two-way analysis of variance (ANOVA).

RESULTS

Cohort characteristics

Plasma samples from 749 LBD individuals were analyzed, in two cohorts. Cohort 1 comprised 626 GBA-neg LBD individuals, 47 LBD carriers of variants associated with LBD risk but not with GD (GBA-risk LBD), 20 LBD carriers of GBA1 mild mutations (GBA-mild LBD), 8 LBD carriers of GBA1 severe mutations (GBA-severe LBD), and 2 LBD individuals homozygous for GBA1 mild variants, along with 9 neurologically normal GBA1 mild mutation carriers. Cohort 2 consisted of 22 GBA-risk LBD, 15 GBA-mild LBD, and 9 GBA-severe LBD individuals.

CSF samples from 55 individuals were analyzed. The CSF cohort comprised 41 GBA-neg LBD, 9 GBA-risk LBD, 2 GBA-mild LBD, and 3 GBA-severe LBD individuals.

Demographic information is summarized in Table 1, and GPNMB measures for all individuals are shown in Supplementary Table 2.

GPNMB plasma and CSF levels do not associate with GBA1 status

We first compared plasma GPNMB levels in GBA1 variant carriers (n=75) versus non-carriers (n=626) with LBD in Cohort 1. We found no difference in plasma GPNMB levels comparing LBD individuals with versus without variants in GBA1 (Figure 1A). Because GBA1 variants have differential impact on GCase activity, we evaluated GBA-risk LBD, GBA-mild LBD, and GBA-severe LBD groups separately, again finding no difference in plasma GPNMB levels among the different GBA1 variant classes (Supplemental Figure 1A).

Figure 1. GPNMB levels are not associated with GBA1 status.

Plasma and/or CSF from 766 individuals were tested for GPNMB levels. (A) Plasma or (B) CSF GPNMB from GBA1 carriers and non-carriers with LBD are compared. Mann-Whitney U-test p-value is shown. (C) Plasma or (D) CSF GPNMB levels from LBD individuals are stratified by rs199347 genotype and GBA1 variant class. P-values for individual effects by two-way ANOVA are shown. (E) Among carriers of one GBA1 N409S mutations, plasma levels for GPNMB do not differ comparing neurologically normal (NC) individuals to LBD individuals. However, LBD individuals with two GBA1 N409S mutations show previously reported increases in GPNMB levels. Mann-Whitney U-test p-value comparing NC and LBD is shown. (A-E) DLB and PNOS individuals (bolder color) admix with PD individuals (pastel shade). Lines and error bars display median and 95% confidence interval (CI), respectively.

To confirm our findings, we tested GPNMB plasma levels in a cohort of LBD individuals with GBA1 variants of different severities, recruited primarily after our initial analyses. In Cohort 2, consisting of 46 additional LBD individuals with GBA1 variants, we again found no association between GBA1 variants and plasma GPNMB levels (Supplemental Figure 1B).

Next, we assessed the relationship between GBA1 status and GPNMB levels in the CSF. As for plasma, we found no differences in CSF GPNMB levels between LBD individuals with versus without GBA1 variants (Figure 1B).

GPNMB plasma and CSF levels associate with rs199347 genotype

We have previously reported that rs199347 genotype associates with GPNMB levels in the CSF (3). Here, we tested for effects of rs199347 genotype and GBA1 status on GPNMB levels in the CSF and in the plasma of LBD individuals by two-way ANOVA. As shown in Figure 1C (plasma) and 1D (CSF), rs199347 genotype, but not GBA1 status, significantly associated with GPNMB protein levels in both plasma and CSF.

GPNMB plasma levels do not differ in LBD versus neurologically normal controls with GBA1 N409S variants

We have previously reported that, in the absence of GBA1 variants, GPNMB levels are higher in the plasma of LBD individuals compared with NC (3). Thus, we tested for differences in plasma GPNMB among carriers of one GBA1 N409S mutation, comparing individuals with (n=19) versus without (n=9) PD. We did not detect any differences in PD compared with NC for this single mutation group, but we did confirm prior reports that carriers of two GBA1 N409S mutations have elevated plasma GPNMB (Figure 1E).

DISCUSSION

Here, we assessed GPNMB protein levels in the CSF and plasma of 55 and 749 individuals with LBD, respectively, for evidence of association with GBA1 status and with GPNMB PD risk genotypes. We found no differences in GPNMB levels by GBA1 status. We confirmed our previously reported associations between rs199347 PD risk SNP genotypes and GPNMB CSF levels and extended this protein quantitative trait locus (pQTL) finding to the plasma for the first time.

The major strength of our study is its sample size, incorporating plasma samples from 121 LBD individuals carrying a single GBA1 variant, and 2 LBD individuals carrying two GBA1 variants, comparing them with samples from 626 LBD individuals without GBA1 variants. With a few notable exceptions (20–23), most GBA1-asssociated LBD biomarker studies are limited to fewer than 50 individuals carrying variants in GBA1 (24–28). Moreover, because the majority of plasma samples were collected from our MIND Initiative study (15), which used a whole-clinic-enrollment strategy, our results may be more reflective of a general movement disorders clinic population.

Limitations of our study should also be acknowledged. First, because LBD individuals were recruited from one clinical site, it is possible that in other populations, stratification of GPNMB levels by GBA1 status might be observed. We note, however, that a prior report of several other established GD biomarkers (ferritin, CD162, CCL18, and chitotriosidase) also failed to find differences in PD individuals with versus without GBA1 variants (20). Moreover, a recent PD CSF proteomic study investigating differentially expressed CSF biomarkers in PD individuals with versus without GBA1 variants also did not find differences in GPNMB levels, despite identifying GPNMB as the top candidate target gene associated with PD by GWAS in Mendelian randomization analyses (29). It is possible that the level of compromise in glucocerebrosidase activity or lysosomal function found in carriers of one GBA1 variant is inadequate to impact GPNMB, whereas the more severe compromise seen in GD leads to “spillover” of GPNMB into biofluids. Second, the ELISA used to measure GPNMB levels, while more robust than “-omic”-scale platforms, is reliable for relative, but not absolute, quantification in our hands. Indeed, this limitation led to our decision to analyze Cohorts 1 and 2 separately, as GPNMB measures were performed in separate batches for each cohort. Finally, while we did not detect differences in plasma GPNMB levels comparing GBA1 N409S carriers with versus without LBD, this aspect of our study lacked adequate power to be conclusive.

Taken together, our data do not support a model whereby GBA1 is upstream of GPNMB in PD pathogenesis. However, they do not rule out the possibility that GPNMB genotypes and expression impact glucocerebrosidase or GBA1-pathway-related activity, which was not assessed here. Moreover, they confirm and extend prior work demonstrating that PD risk genotypes at the GPNMB locus may act by elevating GPNMB expression. As such, they add to a growing body of evidence suggesting that decreasing GPNMB expression may be beneficial as a therapeutic strategy in PD.