Introduction

Parkinson’s disease (PD) is a common neurodegenerative disorder that is characterized primarily by motor dysfunction, such as tremors, muscle rigidity, bradykinesia, and postural instability. Its incidence increases with age, typically manifesting at approximately 60 years of age, with a significant increase in prevalence among individuals over 80 years of age1,2. Patients with PD also present with a series of nonmotor symptoms, such as abnormal constipation, sleep‒wake cycle disorders, orthostatic hypotension, urinary disorders, emotional disorders, a loss of the sense of smell and cognitive impairment3. Among these factors, cognitive impairment may have a greater impact on the quality of life of patients than motor symptoms, and it is also a risk factor for early death4,5. The pathogenesis of PD is complex. Studies have shown that various mechanisms and pathways, including oxidative stress, mitochondrial dysfunction, cellular calcium imbalance, neuroinflammation and other neurotransmitter system defects, play roles in the pathogenesis of PD5,6.

Glycoprotein nonmetastatic melanoma protein B (GPNMB) is a glycosylated type 1 transmembrane protein located in the cell membrane. Mature GPNMB located on the cell membrane is easily cleaved by ADAM10 to form soluble GPNMB (sGPNMB)7. GPNMB is widely expressed in various tissues and participates in cell differentiation, migration, inflammation/anti-inflammation, and tissue regeneration8. A recent study showed that GPNMB increased the risk of PD by interacting with α-synuclein; the plasma GPNMB levels in patients with PD increased with increasing disease severity9. These results indicate that GPNMB is involved in the occurrence and development of PD. Recently, another study showed that plasma GPNMB levels are closely related to cognitive dysfunction10.

The diagnosis of PD currently relies primarily on postmortem confirmation, making research into antemortem biomarkers crucial. Neuroimaging biomarkers, such as PET and MRI, can assess neurodegenerative processes at different stages of PD, but they are complex, costly, and have limited sensitivity to early-stage lesions11,12. Retinal imaging, while noninvasive, is susceptible to ocular diseases, which reduces its specificity and reliability13. The α-synuclein seed amplification assay shows high sensitivity and specificity for diagnosing prodromal α-synucleinopathies, but it is complex, expensive, and lacks standardized protocols14,15. Autonomic function tests have low specificity and are difficult to use as standalone diagnostic tools16,17. Phosphorylated α-synuclein detection in skin biopsies has potential, but issues with sensitivity, specificity, technical standardization, and sample heterogeneity remain unresolved18,19. The isolation and purification of CNS-derived extracellular vesicles face challenges due to their technical complexity, insufficient biomarker specificity and sensitivity, and difficulties in sample collection and processing20,21. Current research on the relationship between these biomarkers and cognitive function is limited, although preliminary evidence suggests that certain markers (e.g., α-synuclein and imaging changes) may be associated with cognitive dysfunction5,22. Among the existing methods, voxel-based morphometry (VBM), a mature structural MRI technique, is cost-effective and suitable as a routine screening tool.

The purpose of this study was to determine the relationship between plasma GPNMB levels and cognitive dysfunction in patients with PD and to explore the possible brain regions affected by GPNMB through VBM to provide a basis for targeted drug treatment of patients with PD and cognitive dysfunction in the future.

Methods

Participants

From January 2023 to September 2023 115 consecutive patients with PD were enrolled from the outpatient and inpatient departments of the Department of Neurology of Henan Provincial People’s Hospital. MRI data from 67 patients were obtained. The inclusion criteria for patients with PD were as follows: all patients with PD were diagnosed by experienced neurologists based on the Movement Disorder Society’s Parkinson’s Disease Clinical Diagnostic Criteria23. The exclusion criteria were as follows: (1) parkinsonian syndrome, such as multiple system atrophy, progressive supranuclear paralysis or secondary (drug-induced, toxic or vascular) parkinsonian syndrome; (2) patients without PD or related syndromes who may present Parkinsonian-like clinical manifestations accompanied by cognitive impairment; (3) a history of tumor or deep brain stimulation surgery; and (4) other central or peripheral inflammatory diseases. Simultaneously, 51 healthy individuals who visited the hospital for routine physical examinations were recruited as the control group. Their ages ranged from 45 to 80 years, and they had no neurological diseases or conditions that might affect GPNMB levels.This case‒control study followed the principles of the Declaration of Helsinki and was approved by the Medical Ethics Committee of Henan Provincial People’s Hospital (2024‒46). All the experiments were conducted in accordance with the relevant guidelines and regulations. All the subjects agreed to participate in this study and signed the informed consent form.

Clinical features

The following general demographic data of all the subjects were recorded: sex, age, and years of education. The course of PD (the time between the first onset of motor symptoms and participation in this study) and the levodopa equivalent daily dose (LEDD) were recorded. The LEDD was evaluated according to the levodopa conversion formula24. Assessments of motor and nonmotor symptoms were performed. The Hoehn‒Yahr stage (H‒Y stage) to assess the severity of the disease. Motor symptoms were evaluated by the Unified PD Rating Scale (UPDRS) Part III Motor Score25.The UPDRS-III consists of 18 items, each scored on a scale of 0 to 4, resulting in a total score ranging from 0 to 72. Higher scores indicate more severe motor symptoms. Importantly, the UPDRS-III exclusively assesses motor symptoms and does not include nonmotor symptoms such as cognitive impairment or mood disorders. Neuropsychiatric symptoms were assessed with the following scales: the 14-point Hamilton Anxiety Scale (HAMA)26, 17-point Hamilton Depression Scale (HAMD)27, REM sleep behavior disorder questionnaire-HongKong (RBDQ-HK)28, Argentine hyposmia rating scale (AHRS)29, Parkinson’s Disease Sleep Scale (PDSS)30, Epworth Sleepiness Scale (ESS)31, Non-Motor Symptom Scale (NMSS) and Parkinson’s Disease Questionnaire-39 (PDQ-39)32. Cognitive impairment was defined as described below. Cognitive function was measured using the Montreal Assessment of Cognitive Ability (MoCA). The PD and cognitive impairment (PD-CI) group consisted of patients clinically diagnosed with PD who presented with cognitive impairment defined by Montreal Cognitive Assessment (MoCA) score cutoffs as follows: ≤13 for illiterate individuals, ≤ 19 for those with 1–6 years of education, and ≤ 24 for those with 7 or more years of education33,34. All motor and nonmotor symptoms were evaluated in the “on” state by the same experienced neurologist.

Detection of plasma GPNMB levels

We collected 5 ml blood samples from each participant at 7:00 am (overnight fast for 12 h), placed the samples in tubes with an anticoagulant (EDTAK2), and centrifuged them at 25 °C (2,500 × g for 15 min) within 1 h of collection. The supernatant was collected, aliquoted into 1.5 ml centrifuge tubes, and stored at -80 °C until analysis. GPNMB protein levels in human plasma samples were measured using an enzyme-linked immunoassay kit (R&D Systems, Catalog Number: DY008B) according to the manufacturer’s instructions. The plasma samples were diluted at 1:30 to measure the optical density within the standard range(R&D Systems, Catalog Number: DY995). All samples were analyzed in duplicate, and the absorbance at 450 nm was determined. The intra-assay coefficient of variation (CV) was 5.07%, and the interassay coefficient of variation (CV) was 11.67%.

Neuroimaging data acquisition and preprocessing

All MR images were acquired using a Siemens MAGNETOM Prisma 3 T MRI scanner with a 64-channel head coil. Structural images were obtained using a 3D MRI magnetization-prepared rapid gradient-echo (MPRAGE) T1-weighted sequence [matrix = 256 × 256; flip angle = 4°; field of view (FOV) = 256 × 256 mm; slice thickness = 1 mm; TR = 5,000 ms; echo time (TE) = 3.43 ms; voxel size = 1 × 1 × 1 mm].

The quality of the acquired MR images was visually assessed to exclude scans with severe vascular lesions, space-occupying lesions, or motion artifacts. VBM was performed with SPM81 running on MATLAB 2013b. T1 image preprocessing was implemented using VBM82 as follows35: (1) format conversion—DICOM images were converted to the parsable NIFTI format; (2) standardization—after image registration, all the anatomical images were registered to the Montreal Neurological Institute (MNI) coordinate space with a voxel size of 1 × 1 × 1 mm3; (3) segmentation—the brain tissue was divided into gray matter (GM), white matter and cerebrospinal fluid; and (4) smoothing using an 8 mm half-peak full width (FWHM) Gaussian kernel36.

Whole-brain and region of interest (ROI) analyses

Two sets of between-group analyses were performed on the differences in GM volume (after modulation and smoothing) using a two-sample t test in SPM8. Age and H–Y stage were included as covariates. The significance threshold was defined as an uncorrected p = 0.001 (p-unc = 0.001) at the voxel level and a familywise error (FWE)-corrected p = 0.05 (p-FWE = 0.05) at the cluster level. False discovery rate (FDR)-corrected peak level results (p-FDRs) are also reported. The brain regions with definite differences between groups were used as ROIs, and we also analyzed the ROI of the left putamen. The XJVIEWER module was used to save the ROI as a mask, and DPABI was used to extract the GM volume of the ROI. The results of the SPM8 analysis were visualized using MRICRON3236,37.

Statistical analysis

The normality of the distribution was evaluated by the Shapiro‒Wilk normality test or Kolmogorov‒Smirnov test. Continuous data with a normal distribution are presented as the means and standard errors (SEs). Data that did not conform to a normal distribution are reported as medians (first quartiles, third quartiles). A t test was used to compare two sets of normally distributed variables. One-way ANOVA was used to compare variables between three or more groups. Variables that did not conform to a normal distribution were evaluated by the Mann‒Whitney U test (two groups) or the Kruskal‒Wallis (KW) rank sum test (three or more groups). The Spearman correlation coefficient was calculated to determine whether the GPNMB level was associated with cognitive impairment, and the Pearson correlation coefficient was calculated to analyze the correlation between the plasma GPNMB level and ROI volume. Receiver operating characteristic (ROC) curves were constructed to assess the accuracy of plasma GPNMB levels in diagnosing patients with PD. Points on the ROC curve with the largest Youden index are considered the best GPNMB cutoff points for diagnosing the disease.

Statistical analyses were performed using SPSS software version 27 (IBM, USA) and the R language (version 4.4.2). Graphs were generated using GraphPad Prism version 10.1.2 (GraphPad Prism Software, Inc., San Diego, CA, USA). A p value < 0.05 was considered to indicate a statistically significant difference.

Results

Demographics and clinical characteristics

Table 1 summarizes the demographic and clinical characteristics of the PD-CI patients, PD-NC patients and healthy controls. Significant differences in age, sex and GPNMB levels were observed among the three groups (p < 0.05)(Table 1)(Fig. 1A). However, no significant difference in the distribution of GPNMB levels was observed among participants of different sexes (p = 0.351). Subjects aged ≥ 60 years and < 60 years were defined as elderly and young, respectively. The level of GPNMB in the elderly group was higher than that in the young group (p = 0.022). The levels of GPNMB in the HC group, PD-CN group and PD-CI group were analyzed in layers to eliminate the influence of age, and the level of GPNMB in the PD-CI group was higher than that in the other two groups. Statistically significant differences were not observed between the PD-CI group and the PD-NC group in years of education, total LEDD score or UPDRS-III score, but statistically significant differences in disease duration, H–Y stage or MoCA score were observed between the two groups (p < 0.05) (Table 1).

Table 1 Demographic data of healthy controls and PD patients.
Full size table
Fig. 1
figure 1

(A) The Kruskal‒Wallis test was used for intergroup comparisons, followed by pairwise comparisons with Bonferroni’s correction to further analyze the differences between groups. ****p < 0.0001 indicates a significant difference, and ns indicates no significant difference. (B) Among the H‒Y stages, 0 represents the control group. The plasma GPNMB level in patients with PD was higher than that in healthy controls, and as the disease progressed, the plasma GPNMB level gradually increased (p = 0.001).

Full size image

We conducted a stratified analysis comparing GPNMB levels between the PD-NC and PD-CI groups at equivalent disease stages (early vs. advanced) to investigate whether the associations between cognitive impairment and biomarkers are influenced by the disease stage. The analysis revealed that GPNMB levels were consistently and significantly increased in the PD-CI group compared with those in the PD-NC group, regardless of the disease stage (Fig. S1A, B). Interestingly, when comparing GPNMB levels across different disease stages within the same cognitive status groups, we observed a nonsignificant trend toward higher levels in patients with advanced-stage PD than in patients with early-stage PD (PD-NC: 24.21 ± 0.82 vs. 25.77 ± 1.48, p = 0.32; PD-CI: 30.22 ± 0.93 vs. 31.16 ± 1.42, p = 0.56). These findings collectively suggest that while disease progression may contribute to increased GPNMB levels, the observed increase in GPNMB levels appears to be more strongly associated with the cognitive status than with the disease stage. The patients were divided into two groups based on the median disease duration of 3 years to investigate the effect of the disease duration on the GPNMB levels. The GPNMB levels were then compared between the PD-CI and PD-NC groups within each stratum. The results revealed significant differences between the two groups in both the short-duration stratum (30.30 ± 0.91 vs. 25.33 ± 1.05, p = 0.001) (Fig. S1 C) and the long-duration stratum [30.82 ± 1.24 vs. 22.19 (20.50, 26.28), p < 0.001] (Fig. S1 D).

GPNMB levels in relation to the H–Y stage and nonmotor symptoms

After controlling for confounding factors such as sex and age, GPNMB concentrations were significantly correlated with MoCA scores (p = 0.019) and the H–Y stage (p = 0.037) (Table 2). GPNMB levels were negatively correlated with MoCA scores, suggesting that when patients’ MoCA scores were lower, their levels of GPNMB were higher. GPNMB levels were positively correlated with the H–Y stage, as shown in Fig. 1B. With an increase in the H‒Y stage, the GPNMB level gradually increased (p = 0.001). GPNMB levels were also positively correlated with HAMD-17 scores; namely, the higher the level of GPNMB, the higher the HAMD-17 score. GPNMB levels were not significantly correlated with HAMA scores; constipation; sleep, smell, or nonmotor symptom total scores; or PDQ-39 scores in patients with PD.

Correlations between plasma GPNMB levels, H-Y stages and MoCA subitem scores

Plasma GPNMB levels were negatively correlated with scores for the following subitems of the MoCA: attention, language and delayed recall (p < 0.05) (Table 3). The plasma GPNMB level parallels the progression of the disease, and thus a further analysis of the correlation between the H–Y stage and scores for MoCA subitems in patients with PD is necessary. Among the two functional domains, there was an overlap with the previous results, but a significant correlation was not identified between the H–Y stage and delayed memory (p = 0.194) (Table 3). We further verified whether the relationships between GPNMB levels and MoCA subitem scores were independent of the H–Y stage, age, and disease duration by calculating partial correlation coefficients. After controlling for the H–Y stage, age, and disease duration, the partial correlation between GPNMB levels and the score for the delayed memory subitem remained significant (specific results are shown in Supplementary Table 1). These findings suggest that the association between GPNMB levels and cognitive impairment, especially delayed memory, may be independent of disease severity.

Table 2 Partial correlations between plasma GPNMB levels and nonmotor symptoms.
Full size table
Table 3 Correlations between the GPNMB levels, H–Y stages and MoCA subitem scores.
Full size table

Construction and analysis of the ROC curves

We constructed ROC curves (Fig. 2) to differentiate PD-CI patients from those with PD alone and to evaluate the utility of GPNMB levels as a potential biomarker for PD-CI. The optimal cutoff value for the GPNMB level was determined to be 28.16 ng/mL (sensitivity = 70.2%, specificity = 75.9%, AUC = 0.775, p < 0.001) (Fig. 2A). For the GMV/ICV ratio, the cutoff value was 41.93% (sensitivity = 65.2%, specificity = 67.2%, AUC = 0.671, p = 0.022) (Fig. 2B). The combined predictive model incorporating both plasma GPNMB levels and the GMV/ICV ratio demonstrated improved performance, as shown in Fig. 2C (sensitivity = 75.4%, specificity = 75.9%, AUC = 0.788, p < 0.001). The weights for the combined model were adjusted based on AUC values using R, resulting in an optimized AUC of 0.803. The Akaike information criterion (AIC) values for the three models were 136.98, 84.87, and 71.07, respectively. Compared with the individual ROC models, the combined prediction model presented the lowest AIC value, indicating a superior balance between goodness-of-fit and model complexity.

Fig. 2
figure 2

(A) ROC curve showing the discriminative ability of GPNMB levels between the PD-CI and PD-NC patients. The AUC value was 0.775. (B) ROC curve showing the discriminative ability of the GMV/ICV ratio between the PD-CI and PD-NC patients. The AUC value was 0.671. (C) ROC curve showing the discriminative ability of the combined GPNMB levels and the GMV/ICV ratio between the PD-CI group and the PD-NC group. The AUC value was 0.803.

Full size image

VBM analysis of differences in brain regions between the high GPNMB group and low GPNMB group

Patients with PD (with MRI data collected) were divided into a high GPNMB group (> 24.67 ng/ml) and a low GPNMB group (≤ 24.67 ng/ml) according to the median GPNMB level. The differences in brain regions between the high GPNMB group and low GPNMB group were compared by a VBM analysis. The results (Fig. 3) revealed that the high GPNMB group had significant differences in the left putamen (cluster size 418, p = 0.015; Table 4) compared with the low GPNMB group. Each volumetric image section in the high GPNMB group presented a reduced area of the left putamen compared with that in the low GPNMB group (Fig. 3A). The brain regions with different volumes between the low and high GPNMB groups are summarized in Table 2. As shown in Fig. 3B, the volume of the ROI was significantly larger in the low GPNMB group than in the high GPNMB group after the extraction of volumes in different brain areas (p < 0.001). Additionally, the correlation analysis indicated a moderate negative correlation between the ROI volumes and plasma GPNMB levels in the PD group (r = − 0.572, p < 0.001; Fig. 3B).

Table 4 Overview of the VBM results.
Full size table
Fig. 3
figure 3

VBM analysis of differences in brain areas between the groups with low and high levels of GPNMB. (A) T1-weighted imaging-based VBM analysis of the whole brain in a series of images of sections. A statistically significant reduction in clusters of the left putamen concentration was observed in the high GPNMB PD group compared with the control group. The data for patients with low GPNMB PD is shown. Only clusters comprising at least 10 suprathreshold voxels were included in the analysis and are shown in the figure. For visualization, shaded red sections were added to highlight all the clusters. The color bar indicates the level of concentration according to the depth of the color. (B) Correlation between the GPNMB level and left putamen volume.

Full size image

Differences in brain regions between the low- and high-GPNMB groups. ∗ indicates that no areas were significantly different. MNI, Montreal Neurological Institute; VBM, voxel-based morphometry; the T value is statistically significant according to a two-sample t test.

Discussion

Recently, the use of quantitative trait loci (eQTLs) or protein quantitative trait loci (pQTLs) in blood, cerebrospinal fluid (CSF) and the brain as functional intermediates to investigate the potential biological mechanism of GPNMB in PD has become very popular38,39,40. These findings revealed that blood GPNMB levels are associated with an increased risk of PD, laying a theoretical foundation for our research. Although genetic data from brain samples are optimal for research into neurodegenerative diseases, peripheral blood is deemed as a reasonable surrogate that avoids several limitations related to postmortem brain tissue40. Our findings are consistent with the results of previous studies by Diaz-Ortiz ME and others9. Plasma GPNMB levels are significantly correlated with the severity of motor symptoms in Parkinson’s disease patients (such as the H‒Y stage). This result may indicate that GPNMB is closely related to the overall progression of PD. Although GPNMB is associated with motor severity, after controlling for influencing factors such as the H‒Y stage, disease duration, and age, the results still revealed a significant independent association between GPNMB levels and cognitive dysfunction (especially delayed memory in the MoCA subitems). These findings suggest that the relationship between GPNMB levels and cognitive decline may extend beyond the severity of motor symptoms. This independent association further supports the potential of GPNMB as a biomarker for cognitive dysfunction.

Neuroinflammation is one of the key factors involved in the progression of PD can lead to neuronal damage and cognitive decline41; the increased levels of GPNMB, a protein associated with inflammation, may reflect the intensification of the inflammatory response. Microglial activation may play a significant role in this process. Semra Smajic and colleagues found that Parkinson’s disease cells are differentially distributed along the microglial activation trajectory and are enriched in the activated state42. Furthermore, a specific group of central nervous system macrophages, known as boundary-associated macrophages (BAMs), highly expresses GPNMB. The genes expressed by the clusters activated by BAMs are involved in inflammation (Il1b), lymphocyte chemotaxis (Cxcl5 and Cxcl10), and extracellular matrix remodeling (Mmp14)43. In animal experiments, Huang et al. reported that plasma GPNMB levels were closely related to inflammatory activity in the brain. Eight hours after the intraperitoneal injection of LPS into the abdominal cavity of the mice, an increase in the number of GPNMB and OX42 double-positive cells was detected in the last region. This finding provides evidence that blood-derived macrophages expressing GPNMB penetrate the systemic circulation and participate in immune/inflammatory responses44. LPS induces the high expression of GPNMB in microglia in vitro, and GPNMB participates in the proinflammatory effect of microglia by increasing the expression level of MMP-345. MMP3 levels are elevated in the postmortem brain tissue of patients with PD. MMP3 released by stressed DA neurons is involved in disease progression46. Moloney and coworkers reported that GPNMB plays a role in lipid accumulation and lysosomal dysfunction. Additionally, GPNMB levels are selectively increased in the substantia nigra of patients with PD and increases after lysosomal stress44. A recent study suggested that GPNMB is functionally related to the cellular uptake of α-synuclein, enabling GPNMB to promote the spread of α-synuclein in the brains of patients with PD9.

However, the role of GPNMB in PD is still widely debated. Studies have also shown that GPNMB plays a protective role in PD. Neal ML et al. reported that GPNMB may play a neuroprotective role by reducing astrocyte-mediated neuroinflammation in a CD44-dependent manner47. The study by Smajic et al. involved GPNMB and CD44. Although the trends of the changes in these two proteins were consistent, the author did not emphasize the dependence of GPNMB on CD4442. Subsequently, a mouse model of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine-induced PD was generated. The transgenic overexpression of GPNMB prevents dopaminergic neurodegeneration48. Treatment with the anti-inflammatory cytokine IL-4 results in increased GPNMB levels in astrocytes and mesenchymal cells9. High expression of GPNMB can reduce the secretion of inflammatory factors and act as a negative regulator of the inflammatory response of macrophages49. Therefore, the role of GPNMB in PD still needs to be explored.

We explored differences in the anatomical structures of brain regions in patients with different GPNMB levels. Interestingly, our VBM analysis revealed a decrease in the volume of the left putamen in the PD group with high GPNMB levels compared with the group with low GPNMB levels. The left putamen was extracted and analyzed as a ROI, and significant differences were still observed between the two groups. The ROI was moderately negatively correlated with the plasma GPNMB level. Putamen atrophy has previously been investigated as a potential biomarker for mild cognitive impairment in patients with PD50. According to Murthy et al., the amount of the GPNMB mRNA distributed in the putamen region is considerable51. Hunerli et al.‘s small cohort study mentioned that the putamen volume in PD-MCI patients was lower than that in the control group52. Interestingly, Hunerli-Gunduz et al. reported that PD-MCI patients exhibited putamen atrophy but not in the right putamen53. Therefore, we speculate that GPNMB may be involved in activating microglia and promotes the release of inflammatory factors during the onset of cognitive impairment in patients with PD, which may indirectly lead to apoptosis or directly regulate apoptosis, thus providing imaging evidence of relative atrophy of the putamen at the voxel level.

In this study, certain baseline differences were observed between the PD-NC and PD-CI groups, particularly in terms of age and sex. The PD-CI group had an older average age, which may have contributed to the elevated levels of GPNMB. Age is a well-established risk factor for the development of cognitive impairment in patients with PD54. In this study, plasma GPNMB levels exhibited a weak yet statistically significant positive correlation with age across all participants, which is consistent with the findings reported by Gillett et al.55. In addition, Lv et al. revealed that MALAT1 upregulated GPNMB protein levels by inhibiting miR-135b-5p, thereby inhibiting cell proliferation and promoting cell apoptosis in MPP+-stimulated PD cell models56. Suda et al. identified glycoprotein nonmetastatic melanoma protein B (GPNMB) as a molecular target for senolytic therapy. The antiaging vaccine also improved the normal and pathological phenotypes associated with aging and extended the lifespan of male progeroid mice57. Collectively, these studies highlight the close associations of GPNMB with aging and apoptosis. Moreover, sex may also have an effect on GPNMB levels. In our data analysis, we statistically controlled for age and sex to minimize the influences of these confounding factors on the results. However, these baseline differences may still have some impact on the interpretation of the results. Future studies need to better control these confounding factors to assess the relationship between GPNMB levels and PD symptoms more accurately.

GPNMB is not very specific, limiting its ability as a biomarker of PD. An increase in GPNMB levels has also been reported in several other neurodegenerative diseases in addition to PD. Examples include AD58, ALS59,60, Gaucher’s disease and Niemann-Pick type C disease61,62. The plasma GPNMB level in patients with Gaucher’s disease was 10 times higher than that in the control group, which is different from findings in patients with other neurodegenerative diseases.

The main strength of this study is that it is one of very few studies to measure peripheral plasma GPNMB levels in persons with PD and control subjects and to analyze the correlation between their levels and clinical symptoms in combination with VBM. A significant limitation of this study was the lack of matching in terms of age and sex between the healthy controls and patients with PD. This mismatch may serve as a potential source of confounding factors, especially for age-related biomarkers such as GPNMB. In addition, the sample size was relatively small, and the study was cross-sectional, which limited our ability to infer causality. Future studies should consider stricter matching strategies and validate the potential of GPNMB as a biomarker in larger longitudinal cohorts, confirming whether changes in GPNMB levels over time are closely related to cognitive decline and exploring the role of confounding factors. Further investigations into the role of GPNMB in neuroinflammation and the specific pathways through which it affects cognitive function are needed.

Conclusions

In summary, the plasma GPNMB level is related to the severity of the disease and PD-related cognitive impairment. It is manifested in the scores of Montreal Cognitive Assessment subitems, such as delayed memory. The left putamen may be one of the brain regions where GPNMB acts in patients with Parkinson’s disease. Nevertheless, these factors provide potential biological markers for the screening and monitoring of PD-related cognitive dysfunction.