Peripheral Immune Cell Numbers and C-Reactive Protein in Parkinson’s Disease: Results from a Population-Based Study

INTRODUCTION

Parkinson’s disease (PD) is a complex neurodegenerative disease that is characterized by a combination of motor and non-motor symptoms [1]. These symptoms are predominantly caused by a lack of dopamine due to neuronal death in the substantia nigra [1–3]. However, the pathology of PD is known to be more widespread and may even initiate outside of the brain [4–6]. Recent studies have suggested an essential role of the immune system in the pathophysiology of PD [1–3, 7–9].

Evidence for both central and peripheral immune responses in the pathogenesis of PD is extensive [8, 9]. In clinical PD, alterations in the gut microbiome have been shown to induce systemic inflammatory responses [8, 11] and impaired humoral as well as cellular immune responses have been found [2, 12]. At an earlier stage of pathogenesis, infectious agents have been suggested to act as a trigger of auto-immune responses against dopaminergic neurons [13]. Indeed, alterations in differential leukocyte counts and cytokine levels have already been described before the onset of clinical PD [14–17].

Despite the emerging evidence of the involvement of the peripheral immune system in PD pathophysiology, it remains largely unclear how peripheral immune or inflammation markers change during the prediagnostic phase of the disease as well as after the diagnosis [8]. The objective of this study was to investigate the temporal relation of peripheral immune cell numbers and C-reactive protein (CRP) with PD. Hereto, we studied the association of these markers with prevalent and incident PD.

METHODS

Assessment of peripheral immune cell numbers and CRP

Participants underwent a maximum number of three venous blood samples at the research center after overnight fasting. Differential blood counts included leukocytes, monocytes, lymphocytes, granulocytes, and platelets in 10⁹ per liter, which were assessed directly following the blood draw using the COULTER AcT diff2 Hematology Analyzer (Beckman Coulter, San Diego, CA, USA) or the Sysmex XS-800 Hematology analyzer (Sysmex, Norderstedt, Germany) for the latest measurement. These peripheral immune cell numbers were measured between 2002 and 2016: At three subsequent visits from the fourth visit of RS-I, at three subsequent visits from the second visit of RS-II and at two subsequent visits from the first visit of RS-III. High-sensitivity CRP was determined in serum which was stored at –20°C until performance of the measurements using Rate Near Infrared Particle Immunoassay (Immage Immunochemistry System, Beckman Coulter, Fullerton, CA). CRP was determined in several study rounds between 1990 and 2008: At the first, third and fourth visit of RS-I, at the first and second visit of RS-II and at the first visit of RS-III. The measurements and number of participants in each study round are shown in Fig. 1.

OPEN IN VIEWER

Fig. 1

Study measurements and number of participants. CRP, C-reactive protein. Numbers in the boxes represent the number of individuals who participated in the shown study round. Numbers in orange are participants who undertook CRP measurements and numbers in green participants who undertook measurements of peripheral immune cell numbers. In RS-I-4 and RS-II-2, CRP was only measured in a random subset, by taking into account sex and age representation. The last measurement of each of the markers (shown in shaded orange or green) was used for the analyses of prevalent PD, if the last measurement was missing, the previous measurement was used (n prevalent PD for analyses CRP = 105, n prevalent PD for analyses peripheral immune cell numbers = 34). For the peripheral immune cell numbers the total number of individuals with any of the markers available is shown, the numbers in the analyses might differ because the specific measurement is missing.

Because neutrophil counts were not available, we used granulocyte counts to calculate the granulocyte-to-lymphocyte ratio (GLR) as a proxy for the neutrophil-to-lymphocyte ratio (NLR). The GLR was calculated as the granulocyte count divided by the lymphocyte count. The platelet-to-lymphocyte ratio (PLR) was calculated as the platelet count divided by the lymphocyte count [19]. The systemic immune-inflammation index (SII) is usually calculated as the NLR times the platelet count [20]. Because of the absence of neutrophil counts in our study, we calculated an adapted SII as the GLR times the platelet count [19].

RESULTS

Table 1 shows the baseline characteristics of the study population, including participants with at least one measurement of the peripheral immune cell numbers or CRP. The mean (standard deviation, SD) age of our sample was 65.1 (9.8) years and 57.5%were women. Current smokers included 21.0%of the total population and the average BMI was 26.9 (4.1) kg/m². Immune system influencing drugs were used by 11.4%of the population. The median [interquartile range, IQR] number of CRP measurements was 1 [1–2] and the median [IQR] number of measurements for the other markers was 2 [1–2].

We included 34 participants with prevalent PD (mean (SD) age 76.6 (7.8) years, 52.9%women) in the logistic regression models of the association between peripheral immune cell numbers and prevalent PD. Figure 2 shows the results of the analyses for prevalent PD per doubling of the markers, model 1 and model 2 showed comparable results. Participants with a higher lymphocyte count were less likely to have prevalent PD (odds ratio (OR) model 2 : 0.34, 95%CI 0.17–0.68). Participants with a higher GLR, PLR, and adapted SII were more likely to have prevalent PD; however, these effects disappeared after subsequent adjustment for the lymphocyte count (OR GLR model 3 : 1.40, 95%CI 0.68–2.87, OR PLR model 3 : 0.98, 95%CI 0.41–2.35, OR adapted SII model 3 : 1.16, 95%CI 0.71–1.91). At the last measurement of CRP, 105 participants (mean (SD) age 78.4 (8.5) years, 61%women) had prevalent PD. CRP was not associated with prevalent PD. We subsequently tested the interaction effects of sex with the peripheral immune cell numbers and CRP, which were non-significant in all analyses of prevalent PD.

OPEN IN VIEWER

Fig. 2

Association between peripheral immune cell numbers, CRP, and prevalent Parkinson’s disease. GRL, granulocyte-to-lymphocyte ratio; PLR, platelet-to-lymphocyte ratio; Adapted SII, adapted systemic immune-inflammation index; CRP, C-reactive protein. Odds ratios are shown per doubling of the marker. Model 1 was adjusted for age and sex. Model 2 was adjusted for age, sex, BMI, smoking, and immune system influencing drugs. Model 3 was adjusted for age, sex, BMI, smoking, immune system influencing drugs, and lymphocytes. A total of 8,553 participants were included in the analyses with leukocytes, 8,550 in analyses with monocytes and lymphocytes, 8,551 in analyses with granulocytes, 8,552 in analyses with platelets, 8,550 in analyses with GLR, 8,548 in analyses with PLR and adapted SII, and 12,426 in analyses with CRP. Prevalent Parkinson’s disease was found in 34 participants for all analyses except for analyses with CRP, which included 105 participants with prevalent Parkinson’s disease. p-values model 1: Leukocytes: 0.22, Monocytes: 0.28, Lymphocytes: 0.001, Granulocytes: 0.74, Platelets: 0.51, GLR: 0.003, PLR: 0.01, Adapted SII: 0.04, CRP: 0.70. p-values model 2: Leukocytes: 0.46, Monocytes: 0.32, Lymphocytes: 0.002, Granulocytes: 0.42, Platelets: 0.55, GLR: 0.003, PLR: 0.03, Adapted SII: 0.04, CRP: 0.62. p-values model 3: GLR: 0.36, PLR: 0.97, Adapted SII: 0.04.

In the analyses of the peripheral immune cell numbers, during a median [IQR] follow-up of 14.4 years [10.0–19.0], 73 participants developed incident PD and 2,354 participants died. The mean (SD) age at diagnosis of these incident PD events was 76.6 (8.9) years and 47.1%were women. The hazard ratios of the joint models for higher leukocytes, lymphocytes, platelets, PLR, and adapted SII were suggestive of a decreased risk of incident PD and higher monocytes of an increased risk of incident PD (Table 2). However, confidence intervals were wide and none of these results were statistically significant. In the analyses of CRP, during a median [IQR] follow-up of 11.8 years [9.1–17.7], 160 participants developed incident PD and 5,735 participants died. The mean (SD) age at diagnosis was 77.4 (7.9) years and 47.5%were women. There was no evidence for an association of CRP with the risk of PD. Subsequent analyses of the interaction between sex and the peripheral immune cell numbers and CRP in the joint models revealed no significant results.

Nested case-control analyses

For the nested case-control analyses, two cases could not be matched to the minimum number of three controls and were excluded. Figure 3 visualizes the estimated temporal evolution of the lymphocyte count for participants with incident PD (n = 85) and matched controls (n = 255) using linear mixed models. This figure suggests that the lymphocyte count starts deviating in PD cases compared to controls at the time of disease diagnosis.

OPEN IN VIEWER

Fig. 3

Visualization of the association between lymphocyte count and Parkinson’s disease. The lines show the predicted lymphocyte count for participants with an average BMI (26.4) who do not use immune system influencing drugs. The shaded areas represent the corresponding 95%credible intervals. Of the 87 incident Parkinson’s disease events, 85 events could be matched with three controls and were included in the analysis. Matching was performed on birth year, sex, cohort, and the model was adjusted for BMI and immune system influencing drugs. Adjustment for smoking status was not possible because of the small number of smoking participants.

DISCUSSION

In this population-based study, we found a lower lymphocyte count in people with clinical PD. Although the GLR, PLR, and adapted SII were also associated with prevalent PD, these effects were driven by the lymphocyte count. Contrary to previous studies, we found no association between peripheral immune cell numbers, CRP, and incident PD.

Both the innate and adaptive immune system have been shown to be involved in the pathogenesis of PD [2, 9]. However, whether immune system alterations trigger α-synuclein aggregation or act merely as a response thereof continues to be a source of debate [3, 8]. Studies have shown that α-synuclein can activate immune responses [3, 28], but also that peripheral lymphocytes may infiltrate in the central nervous system, initiate α-synuclein aggregation and stimulate uptake of α-synuclein in the brain [3, 29]. Our results support the hypothesis that peripheral immune system alterations act as a response to already ongoing pathogenesis and neurodegeneration.

Our finding that participants with a higher lymphocyte count were less likely to have prevalent PD corroborates previous findings of lower levels of circulating lymphocytes in patients with prevalent PD [30–36]. This decrease has been shown to occur predominantly because of reduced numbers of CD4 + and CD3 + lymphocytes [37, 38]. Interestingly, our observed association between lymphocytes and prevalent PD was attenuated after adjustment for anti-parkinsonian drugs. Emerging evidence indicates that dopamine importantly connects the nervous system with the immune system [39–41]. Although lymphocytes express dopaminergic receptors and dopaminergic agents might thus act directly on peripheral immune cells [40], recent studies found no or only small effects of anti-parkinsonian drugs on peripheral immune cells [30, 41]. The influence of anti-parkinsonian drugs on the association between lymphocytes and PD is thus of interest for future research.

In addition to the effect of lymphocytes, we found that participants with a higher GLR, PLR, and adapted SII [19, 42] were more likely to have prevalent PD. However, this effect was driven by the lymphocyte count. A few cross-sectional studies have previously studied the relation between NLR and prevalent PD. Two studies found statistically significant higher mean NLR in patients with PD compared to controls (adjusted OR Liu et al.: 1.7, 95%CI 1.0–2.9) [43, 44], whereas three other studies reported largely similar NLR between cases and controls [36, 46]. None of these previous studies adjusted their analysis for the lymphocyte count.

We also studied the relation between CRP and prevalent PD and, in contrast to a 2019 meta-analysis which found a standardized mean difference of 1.1 of patients with PD compared to controls [12], we did not find an association between this acute-phase protein and prevalent PD. The difference with the recent meta-analysis [12] might be partially explained by publication bias, as the meta-analysis included multiple small studies with large effect sizes.

Only a few previous studies have determined the association between leukocyte count and the risk of incident PD [14, 15]. Ton et al. studied the association between leukocyte count and the risk of PD in 5,828 participants of the Cardiovascular Health Study [47]. They reported a higher risk of incident PD with an increased leukocyte count, although this effect was not statistically significant [47]. In addition, Yazdani et al. repeatedly measured blood immune and inflammation markers in a longitudinal cohort study of 812,073 participants and followed the participants for the occurrence of PD [15]. In contrast to Ton et al., they found that an increased baseline leukocyte count (HR per SD increase: 0.86, 95%CI 0.80–0.93), driven by lymphocytes (HR per SD increase: 0.74, 95%CI 0.59–0.94), was associated with a decreased risk of PD. Finally, Jensen et al. studied differential leukocyte counts and the risk of incident PD in 312,590 participants of the UK biobank [14]. They showed that a lower lymphocyte count was associated with an increased risk of incident PD (OR per SD reduction: 1.18, 95%CI 1.07–1.32), which was also confirmed by their subsequent Mendelian randomization analyses. Jensen et al. also described a higher risk of PD with decreased eosinophil count, monocyte count and CRP, although these findings were not robust to sensitivity analyses.

Although our effect estimates of the association between repeatedly measured lymphocyte counts and PD risk might suggest a protective effect of a higher lymphocyte count, these findings were not statistically significant, were not reproduced in the subset without comorbidities and are in contrast with our nested case-control analyses suggesting that the lymphocyte count starts to deviate from the expected average after PD diagnosis. For the monocyte count, a previous study in people with REM-sleep behavior (RBD), which is considered a prodromal state of PD, found no difference between the percentage of total monocytes, but an increase in percentage of classical monocytes compared to controls [48]. In addition, Jensen et al. found some evidence of an increased risk of PD with reduced monocyte count and the study of Yazdani et al. found no association between monocytes and the risk of PD [14, 15]. Because of these conflicting previous findings and the wide confidence intervals around our estimates, the association between monocyte count and PD warrants further investigation.

Differences between our study and the studies by Yazdani et al. and Jensen et al. [14, 15] which reported a lower lymphocyte count to be associated with an increased risk of incident PD could be explained by several factors. First, these two previous studies [14, 15] used hospital discharge international classification of diseases (ICD) codes or self-reported PD to define incident events. Both studies reported a significant delay of multiple years between the first symptom onset and the date of incident PD recorded in the study [14, 49]. We defined the date of diagnosis as the first moment the criteria for PD were met based on continuous follow-up of clinical records of general practitioners, nursing homes and hospitals, which has probably led to an earlier and more accurate event date of PD. However, exclusion of follow-up time several years before diagnosis did not substantially alter the results of the previous studies, suggesting that this discrepancy does not fully explain the different results. Second, selection bias may have played a role in the previous studies as the study of Yazdani et al. included blood samples from general health check-ups or referrals from outpatient visits and the study of Jensen et al. analyzed a subset of the already selected UK biobank cohort. Finally, and perhaps most importantly, our study only included 73 events of incident PD, thus giving rise to uncertainty about the estimated effects.

Strengths of this study include the representative population-based design with continuous follow-up for the occurrence of PD, which ensures timely and standardized diagnosis of PD events. Furthermore, we performed structural and repeated measurements of peripheral immune cell numbers and CRP, not relying on measurements performed as part of the clinical assessment of a patient.

The main limitation of the current study includes the small number of PD events. Although the incidence of PD in our study was similar to other population-based studies [50], the number of events remained small for the analyses of interest. The small number of events led to wide confidence intervals around the estimated effects and prohibited us from performing further stratifications to take into account the heterogeneity of PD. In addition, given the previously described association between peripheral immune cell numbers, CRP, and mortality [42, 52], death could be considered a competing event in our analyses which could have biased the results. Furthermore, the joint and linear mixed models we used depend on the correct specification of fixed and random effects to estimate the underlying value of the marker, misspecification of these models could have impacted our findings. Finally, we did not measure lymphocyte subtypes and did not assess whether cytokines, which could have pro- or anti-inflammatory properties, were associated with prevalent or incident PD. A 2016 meta-analysis and several subsequent studies showed altered concentrations of peripheral pro- and anti-inflammatory cytokine levels in people with PD [53–56] and in people with REM-sleep behavior disorder [57] compared to controls, but the association with incident PD remains largely unknown [16, 17].

In conclusion, we found that participants with a higher lymphocyte count were less likely to have prevalent PD. In contrast to previous studies, however, we did not find an association between peripheral immune cell numbers nor CRP and the risk of incident PD. The longitudinal relation between the number of circulating immune cells, biomarkers of inflammation and the risk of PD warrants further investigation by large population-based studies.