Two populations were included in the study. Blood samples for enzyme-linked immunosorbent assay (ELISA) analyses were collected from 34 patients with PD and 35 controls (Table 1). PD patients had a mean age of 74.4 ± 8.6 years and controls 75.3 ± 8.2 years. Mean duration of PD was 6.4 ± 6.8 years. Whole blood samples for total RNA isolation and quantitative real time polymerase chain reaction (qRT-PCR) analysis were collected from 60 PD patients and 30 controls (Table 1). The mean age of the PD patients and healthy controls were 73.2 ± 8.0 and 73.0 ± 7.9 years, respectively. The mean duration of PD was 7.1 ± 6.8 years. The study was approved by the Research Ethics Committee of the University of Tartu, and all subjects provided written informed consent.

Only patients who fulfilled the Queen Square Brain Bank Criteria for PD (Gibb and Lees, 1988; Poewe et al., 2017) were recruited. Patients with PD underwent medical assessment, which included clinical history, comorbid diseases, toxins in history, and medication use. The disease severity was staged according to the Hoehn and Yahr (H&Y) scale (Hoehn and Yahr, 1967), and the disability was assessed with the Schwab and England Activities of Daily Living (SE-ADL) scale (Schwab et al., 1969). The severity of the PD symptoms was identified by using the Movement Disorders Society-Unified Parkinson’s Disease Rating Scale (MDS-UPDRS), as a rule rated in the ON-stage of the disease (Goetz et al., 2008). Cognitive function was assessed using the Mini Mental State Examination (MMSE) (Folstein et al., 1975). Depression was rated with the Beck Depression Inventory scoring (Beck et al., 1961) based on 21 questions with ratings from 0 to 3 to measure severity of depression. A score of 18 points signifies depression in PD patients (Silberman et al., 2006; Tumas et al., 2008). Olfactory performance was assessed with the Sniffin’ Sticks Olfactory test (SS-12) (Hummel et al., 1997). Current PD medications of the patients involved in the ELISA measurements are provided in Supplementary Table S1. Comorbidities and other physiological or medical conditions of the PD patients are listed in Supplementary Table S2.

The control group was formed from healthy volunteers without any current neurological condition, and individuals who visited outpatient or inpatient clinics of the Department of Neurology and Neurosurgery of Tartu University Hospital without any current neurodegenerative disease but because of other neurological conditions.

Blood samples were collected to BD Vacutainer^® tubes and allowed to clot at room temperature. Serum was separated by centrifugation at 1300 × g for 10 min at room temperature (RT), and stored at −80°C until analysis.

The development and validation of human MANF ELISA for the measurement of MANF in serum has been described earlier (Galli et al., 2016). The dynamic range of MANF ELISA is 62.5–2000 pg/ml, and its sensitivity is 45 pg/ml. Human MANF ELISA does not recognize recombinant human CDNF. For MANF quantitation, serum samples were diluted at 1:20 in blocking buffer [1% casein in phosphate buffered saline, 0.05% Tween-20 (PBST)]. To block interference caused by heterophilic antibodies (HAs), Immunoglobulin Inhibiting Reagent (IIR; BioIVT) was added to the serum samples at concentration of 500 mg/l. Diluted samples and standards were applied to 96-well plates coated with goat anti-human MANF antibody (AF3748, R&D Systems), and incubated overnight at +4°C in agitation. After washing with PBST, bound MANF on the plates was detected using horseradish peroxidase (HRP)-conjugated mouse anti-human MANF antibody (4E12, Icosagen) and DuoSet ELISA Development System (R&D Systems). Absorbance was read at 450 nm and 540 nm using VICTOR³ plate reader.

The development of human pAb CDNF ELISA has been described (Back et al., 2013). Sensitivity of the CDNF ELISA is 6.2 pg/ml (determined as the mean + 3 standard deviations (SDs) of 10 blank samples) and the quantitation range 7.8–500 pg/ml (Supplementary Table S3). The antibodies used in CDNF ELISA were goat anti-hCDNF (1 μg/ml, R&D Systems) for coating, and rabbit anti-CDNF (0.1 μg/ml, Lindholm et al., 2007) followed by HRP-conjugated donkey anti-rabbit (1:2000, GE Healthcare) for detection. The mean intra- and inter-assay variations (analyzed at three different concentration levels) are 9.3 ± 2.5 and 9.7 ± 1.5% CV, respectively. For testing the blockage of HA-interference, we added IIR to serum samples at concentration of 100 and 500 mg/l, and monitored the background with a control ELISA (cELISA) as described (Galli et al., 2016). The cELISA was run exactly as the CDNF ELISA; the only difference was a coating antibody, which was goat anti-hMANF (1 μg/ml, R&D Systems), opposed to goat anti-hCDNF used in the CDNF ELISA. Cross-reactivity of cELISA for neither recombinant CDNF nor MANF was verified at a concentration of 2 ng/ml. Linearity of dilution and recovery of spiked CDNF in serum was tested in the presence of IIR.

For fractionation, we pooled 10 subjects’ blood obtained from the Finnish Red Cross (license 43/2017). A total of 50 ml citrate-blood was centrifuged at 150 × g for 20 min at room temperature with no brake applied. The platelet-rich plasma was transferred into a new tube without disturbing the buffy coat, and HEP buffer, pH7.4, (140 mM NaCl, 5 mM EDTA, 3.8 mM HEPES, 2.7 mM KCl) was added at a 1:1 ratio. The sample was mixed gently to prevent platelet activation and centrifuged at 100 × g for 15–20 min at RT to pellet contaminating red and white blood cells (RBCs and WBCs). The remaining supernatant was centrifuged at 800 × g for 20 min at RT to pellet platelets. Isolated platelets were rinsed twice with 1 mM EDTA in PBS (pH 7.4) and homogenized in 400 μl of lysis buffer (20 mM Tris–HCl, pH 8.2, 137 mM NaCl, 1% NP40, 10% glycerol, 0.5 mM Na₃VO₄, Complete Mini protease inhibitor cocktail (Roche)). After 20 min incubation on ice, the sample was centrifuged at 12,000 × g for 20 min at + 4°C, and the supernatant was stored at −80°C until analysis.

WBCs were collected by isolating the buffy coat of the initial centrifugation. Contaminating RBCs were lysed by the addition of 0.5 ml of eBioscience^TM RBC Lysis Buffer (Invitrogen) and incubated for 15 min at RT. WBCs were pelleted by centrifugation at 500 × g for 5 min at RT. The supernatant was decanted, and the pellet was carefully resuspended in cold PBS and centrifuged again. The washed pellet was homogenized in 400 μl of lysis buffer.

For RBC isolation, the top layer contaminated with buffy coats was removed and 400 μl of the RBC layer was collected. 8 ml of PBS was added to the RBC layer and centrifuged at 500 × g for 10 min at RT. The supernatant was removed, and the washing was repeated two more times. The pellet of RBCs was homogenized in 800 μl of lysis buffer.

Mesencephalic astrocyte-derived neurotrophic factor in the homogenates was quantitated by hMANF ELISA and the result normalized to the total protein concentration measured by DC Protein Assay (Bio-Rad Laboratories).

Protein samples derived from fractionated blood cells were reduced by boiling with β-mercaptoethanol, separated on 4–15% SDS-PAGE (MiniProtean^® TGX^TM, Bio-Rad; 50 μg/lane) and blotted on a nitrocellulose membrane. Proteins were stained with Ponceau S (0.1% in 5% acetic acid) after which MANF was detected with rabbit anti-MANF (1 μg/ml, Icosagen) followed by HRP-conjugated secondary antibody against rabbit IgG and enhanced chemiluminescence (ECL) Western Blotting Substrate kit (Pierce).

Whole blood samples were collected into Tempus Blood RNA tubes (Applied Biosystems) and frozen at −80°C. Total RNA was isolated using Tempus Spin RNAS isolation kit (Applied Biosystems). 200 ng of RNA was used for cDNA synthesis using High Capacity cDNA Reverse Transcription Kit with RNase Inhibitor (Applied Biosystems). cDNA was used as a template for TaqMan^® qRT-PCR analysis in the ABI Prism 7900HT Sequence Detection System (Applied Biosystems, Foster City, CA, United States). TaqMan^® Gene Expression Master Mix and FAM labeled TaqMan (Applied Biosystems) gene assays were used to detect the mRNA expression level of the gene of interest and of actin as a reference gene. The TaqMan probes used were: Hs00180640_m1 (MANF), Hs00418490_m1 (CDNF), and Hs01060665_g1 (ActB). Reactions were carried out in four replicates. Data was analyzed using the 2(-Delta Delta C(T)) method, where the gene expression levels were normalized to the level of the actin beta housekeeping gene.

The distributions of MANF and CDNF concentrations were skewed (Shapiro–Wilk test, P < 0.05 for both), thus non-parametric statistical tests were carried out to analyze serum data. Bivariate correlations were analyzed with Spearman’s correlation test. Kruskal–Wallis, along with Tukey HSD post hoc test, and Mann–Whitney U-tests were used to compare MANF and CDNF serum levels in different groups. qRT-PCR data was analyzed with the parametric unpaired t-test. P ≤ 0.05 was considered statistically significant for all tests. Statistical analyses were performed using Software Package for Social Science (SPSS) v. 21.0. All results are expressed as mean ± SD.

The average serum MANF concentration was twice as high in PD patients compared to their age-matched controls (Figure 1A). Serum MANF concentration was 10.2 ± 8.6 ng/ml (median: 6.7 ng/ml, n = 34) in PD patients and 5.0 ± 4.0 ng/ml (median: 4.3 ng/ml, n = 35) in the control group (P < 0.001). The difference in MANF serum concentration between the groups was 5.2 ng/ml, which represent a 105% increase.

Serum MANF level was not dependent on duration or severity of PD (Table 2), or used medication (Supplementary Table S1). However, a positive correlation was found between serum MANF concentration with BDI scoring (r_s = 0.42; P = 0.02; Figure 1B). In PD patients who scored 18 or more points the average serum MANF level was 14.7 ± 12.8 ng/ml (n = 10), while in patients scoring 2–17 points the average MANF concentration was 7.1 ± 4.2 ng/ml (n = 21, P = 0.031; Figure 1C).

It has been reported that MANF mRNA is expressed in the peripheral WBCs (Wang J. et al., 2014). Thus, it is possible that circulating MANF is derived from blood cells. We used ELISA to quantitate MANF levels in different types of blood cells isolated from healthy human donors. MANF levels in RBCs, WBCs, and platelets in a pooled blood sample were 0.1, 79.5, and 358.8 ng/mg of total protein, respectively (Supplementary Figure S1A). In agreement, MANF was detected in platelets but not in RBCs by Western blot (Supplementary Figure S1B).

Next, we used qRT-PCR analysis to measure MANF mRNA levels in whole blood derived from the PD patients and controls. However, MANF transcripts were not elevated in the blood of PD patients compared to controls (P = 0.44; Figure 1D).

We have previously found a significant interference in our in-lab-designed hMANF ELISA by HAs present in human blood circulation (Galli et al., 2016). Similar interference was observed in the hCDNF ELISA as analyzed by a cELISA. The cELISA was constructed on an antibody pair derived from the same host species as the antibodies in the original ELISA, but without reactivity for either recombinant CDNF or MANF. Addition of IIR to the serum samples decreased the background detected by the cELISA. The best blockage of background caused by HA was obtained with 1:4 serum dilution and 500 mg/l IIR when analyzed on the hCDNF ELISA (Supplementary Table S4). The recovery of recombinant CDNF added at concentrations of 50, 100, or 250 pg/ml to sera diluted at 1:4 and supplemented with 500 mg/l IIR was 76.6 ± 3.3% (range 72.0–82.8%, n = 7). Linearity of dilution of endogenous CDNF in human sera was 95.5 ± 12.1% (n = 3 sera in 4 sequential dilutions, Supplementary Table S5).

In general, CDNF levels in serum were much lower than those of MANF. The range of calculated serum CDNF concentrations in the study population was 4.3–134.4 pg/ml. As the lower limit of detection (LLOD) of our hCDNF ELISA is 6.2 pg/ml, and the sera need to be diluted 1:4 for optimal background control, the minimal CDNF concentration that can reliably be distinguished from the blank is 24.8 pg/ml.

Cerebral dopamine neurotrophic factor (CDNF) concentrations in 7 (21%) samples form the PD and 6 (17%) samples from the control group fell below the LLOD of our hCDNF ELISA. We approached the data analysis by three simple methods used for dealing with samples under LLOD (Helsel, 1990): (1) including all sample values; (2) replacing the sample values below LLOD by 0; and (3) replacing the sample values below LLOD by half of the LLOD (i.e., 12.4 pg/ml), which can be considered as a rough mean value of the samples that fall below the detection limit. All the three methods gave similar results indicating that serum CDNF levels in PD patients and controls do not significantly differ in the current study population. P-values for each of the analysis methods were following: (1) 0.25, (2) 0.28, and (3) 0.28. In the Figure 2A we show the values also below the LLOD for not causing an artificial skew in the distribution. However, all values below LLOD should be taken with a level of uncertainty. When all sample values were included in the calculations, the average serum CDNF concentration in the PD patients was 40.3 ± 23.5 pg/ml (median: 38.1 pg/ml, n = 34), and in the controls it was 33.3 ± 12.9 pg/ml (median: 32.5 pg/ml, n = 35; Figure 2A).

MANF and CDNF concentrations correlated positively in the sera from the PD patients (r_s = 0.41, P = 0.016, n = 34, Figure 2B) but not in those from the controls (r_s = 0.18, P = 0.31, n = 35).

Using qRT-PCR we observed that CDNF gene expression was upregulated in whole blood samples of PD patients by 26% as compared to controls (P = 0.015; Figure 2C).