Between 2019 and 2021, 28 patients with PNP were prospectively recruited at the Department of Neurology, University of Würzburg, Germany, where they were seen for diagnostic work-up. In addition, 10 patients with acute headaches of unclear etiology were used as disease controls for the CSF analysis, and serum data from twelve healthy volunteers were included in the analysis to determine the inflammatory state of the headache group. Our study was approved by the Würzburg Medical Faculty Ethics Committee (# 15/19 and # 242/17) and all participants gave written informed consent prior to inclusion.

Diagnoses were based on history taking and neurological examination, laboratory studies, and nerve conduction examinations. All patients underwent laboratory tests including full blood count, electrolytes, kidney and liver function tests, C-reactive protein, thyroid stimulating hormone, vitamin B12, HbA1c, oral glucose tolerance test (OGTT), screening for autoimmune antibodies (i.e., ANA, ENA, ANCA, anti-ganglioside antibodies), and lumbar puncture. Electrophysiological assessment with nerve conduction studies of the affected nerves including tibial motor nerve and sural sensory nerve was performed in all PNP patients.

To rate the severity of the neuropathy, we used standardized scales, including the modified Toronto clinical neuropathy score (mTCNS), the overall disability sum score (ODSS) and the Medical Research Council-sumscore (MRC-sumscore). Depression was assessed with the “Allgemeine Depressionsskala” (ADS) (37). Pain was evaluated by a numerical rating scale (NRS) from 0 (no pain) to 10 (worst pain), the neuropathic pain symptom inventory (NPSI) and the graded chronic pain scale (GCPS). After the diagnostic work-up, the examining neurologists categorized the neuropathies into a mild (1), moderate (2) or severe (3) clinical phenotype.

From each patient, venous blood was drawn in the morning between 8 a.m. and 9 a.m. into S-Monovette^® Serum-Gel Tubes (Sarstedt, Nümbrecht, Germany) and Tempus™ Blood RNA Tubes (Thermo Fisher Scientific, Waltham, MA, USA). For the collection of sera, whole blood was left to clot for 30 min at room temperature and later centrifuged for 10 min at 1200 g. Supernatant was aliquoted and stored at -20°C until further analysis. Tempus™ Blood RNA Tubes were immediately shaken for 30 s to lyse the cells and stabilize the RNA and stored at -80°C until extraction.

CSF samples were obtained after a lumbar puncture at the L4/5 level. The amount of CSF taken varied between 10-15 ml. Samples of 2 ml were aliquoted and subsequently stored at -20 °C.

RNA was extracted from Tempus™ Blood RNA Tubes following the manufacturer’s protocol (38, 39) from the MagMAX™ for Stabilized Blood Tubes RNA Isolation Kit (Thermo Fisher Scientific, Waltham, MA, USA). RNA quality and quantity were assessed with a NanoDrop™ One (Thermo Fisher Scientific, Waltham, MA, USA), and RNA was stored at -80°C.

For cDNA synthesis from mRNA, TaqMan Reverse Transcription reagents (Thermo Fisher Scientific, Waltham, MA, USA) were used. 250 ng mRNA of each sample were pre-incubated with 5 µl random hexamer at 85°C for 3 min. Next, 10 μL 10× PCR buffer, 22 μL MgCl_2, 20 μL deoxyribonucleoside triphosphate, 6.25 μL multiscribe reverse transcriptase and 2 μL RNase inhibitor were added per sample. Lastly, reaction was performed under these conditions: annealing (25°C, 10 min), reverse transcription (48°C, 60 min), and enzyme inactivation (95°C, 5 min).

For miRNA, reverse transcription was carried out with the miRCURY LNA RT Kit (Qiagen, Hilden, Germany). 10 ng of RNA were mixed with 2 μL of 5x reaction buffer, 5 μL of nuclease free water and 1 μL of enzyme mix, per sample. Reaction was performed using the following program: reverse transcription (42°C, 60 min) and enzyme deactivation (95°C, 5 min).

Reactions were carried out on a PRISM 7700 Cycler (Applied Biosystems, Waltham, MA, USA) and transcribed cDNA was stored at -20°C until further analysis.

Real time qPCR of mRNA and miRNA targets was performed to analyze gene expression using the StepOnePlus Real-Time PCR System (Thermo Fisher Scientific, Waltham, MA, USA). For mRNA, RT-qPCR was carried out with TaqMan qRT-PCR reagents (all Thermo Fisher Scientific, Waltham, MA, USA) and pre-designed assays. For target normalization, different endogenous controls were used: ribosomal protein L13a (RPL13A), actin beta (ACTb), Hydroxymethylbilane Synthase (HMBS) and TATA-Box binding protein (TBP) were tested. TBP was the most stable across groups, and thus selected as suitable endogenous control. For each reaction, 3.5 µl cDNA (8.75 ng cDNA) was mixed with 0.5 µl nuclease free water, 5 µl Fast Advanced Mastermix and 0.5 µl TBP primer and 0.5 µl target primer (see list of primers in Table 1).

For miRNA, the miRCURY LNA SYBR Green PCR Kit (Qiagen, Hilden, Germany) and pre-designed miRCURY LNA miR PCR assays (Qiagen, Hilden, Germany) were used. Based on previous experience from our group (12, 13) and on the recommendations by TaqMan Advanced miRNA Assays (https://assets.thermofisher.com/TFS-Assets/GSD/Reference-Materials/identifying-mirna-normalizers-white-paper.pdf), the ribosomal RNA 5s and hsa-miR-16 were tested as endogenous controls. Due to differences found in the expression of 5s between groups, hsa-miR-16 was selected as suitable endogenous control, based on its comparability and standard deviation across groups and samples. Each miRNA was run adding 5 µl 2× miRCURY SYBR Green Master Mix with 1 µl ROX per 50 µl and 1 µl primer (see list of primers in Table 1) to 4 µl of 1:80 diluted cDNA.

Each mRNA and miRNA was amplified in triplicates and relative quantitation (RQ) values were obtained by the StepOnePlus™ Software v2.3 (Thermo Fisher Scientific, Waltham, MA, USA) using interplate calibrators through the 2-ΔΔCt method.

Cytokine levels were measured in serum and CSF using the Ella™ technology (ProteinSimple, San Jose, Cal, USA). Ella™ is a fully automated cartridge-based system that allows you to perform multiple sample, multi-analyte immunoassays with the specificity of a traditional single-plex ELISA (enzyme-linked immunosorbent assay). Samples were thawed on ice and more than two freeze-thaw cycles were avoided. Samples were diluted 1:1 in the appropriate sample diluent from each kit and 50 µl were added to each sample well of the Simple Plex™ cartridge, after 1 ml of washing buffer was added to their corresponding wells. Each sample was measured in triplicates, and the levels of each cytokine were displayed in pg/ml.

Serum and CSF samples were quantitatively analyzed for their levels of the endogenous metabolites carnitine (CAR), palmitoylcarnitine (PC) and oleoylcarnitine (OC) by LC-MS/MS (liquid chromatography coupled with tandem mass spectrometry).

All samples were thawed on ice for the first time for this analysis, in order to avoid freeze-thaw cycles, and processed within two years upon collection. Serum samples were analyzed at a neat dilution and after applying a 1:10 dilution step in 70% (v/v) ethanol. For sample preparation, 50 µl of serum or CSF samples were mixed with 150 µl internal standard solution (100 ng/ml acetylcarnitine-d3 in acetonitrile). Samples were centrifuged (13000 rpm, 1.5 min) and the supernatant was used for analysis.

Chromatographic separation was performed with a UHPLC System (1290 Infinity series, Agilent) using a Waters Acquity UHPL CSH™ Fluoro-Phenyl (75 × 2.1 mm, 1.7 µm) column. The injection volume was 2 µl and the flow rate was set to 0.3 mL/min. For the chromatographic separation a gradient was run using 0.1% formic acid in water (solvent A) and 0.1% formic acid in acetonitrile (solvent B). Detection was performed by means of a triple quadrupole mass spectrometer (API 4000^®, Sciex) in multiple reaction monitoring (MRM) mode. Measurements were carried out in the positive electrospray ionization (ESI) mode. The LC-MS/MS system was operated with the Software Analyst^®, Version 1.7.1 (Sciex).

Calibration samples were prepared in 30% acetonitrile and covered calibration ranges of 5 - 500 ng/ml for PC and OC and 5 - 1500 ng/ml for CAR. To verify method performance at medium analyte concentrations, QC samples with a concentration of 50.00 ng/ml CAR, OC and PC were prepared in 30% acetonitrile or in a surrogate matrix comparable to the analyzed biological samples (2% bovine serum albumin (BSA) or PBS buffer pH 7.4 diluted 1:10 in 30% acetonitrile were used to mimic serum or CSF, respectively).

Correlation coefficients of 0.9996 (CAR, OC) and 0.9992 (PC) were obtained. Accuracies of the calibration samples were found within +/-15% or within +/-20% for the lowest calibration level, respectively. Mean accuracies for all QC samples were found between 87.07% and 109.77% with CVs < 15%, indicating an acceptable method performance for all matrices investigated.

Further analytical information can be found in the Supplementary Material.

Statistical analysis was performed in SPSS 27 (IBM, Armonk, NY, USA), where the Shapiro-Wilk test was used to determine the normal distribution of the data. For parametric data, a T-test was used for comparison between two groups and a Pearson test was performed for correlations. In non-parametric data, the Mann–Whitney U Test was applied for comparison of two groups, and the Spearman test was used for correlations. Data results were plotted in GraphPad Prism 9 (GraphPad Software, Inc., La Jolla, CA, USA) for visualization. Graphical images were incorporated from Smart Servier Medical Art, https://smart.servier.com/, under the Creative common Attribution 3.0 Unported Licence.

Baseline characteristics of the study group and the diagnostic subgroups are summarized in Table 2. Patients were included if they presented for diagnostic work-up for their PNP, including lumbar puncture. To be included, they further needed to have either no pain (NRS = 0) or pain ≥ 4 at the time of admission. Patients with an NRS between one and three were excluded from the cohort. After applying the exclusion criteria, twenty-eight patients with PNP of different etiologies were included [median body max index (BMI) 28.3, range 19.2-35.1; median age 54.5 years, range 20–80]. The median disease duration was 2.5 years (range 0.02–29 years). Nine patients were diagnosed with an inflammatory neuropathy including non-systemic vasculitis (six patients), chronic inflammatory demyelinating polyneuropathy (CIDP) (two patients) and multifocal motor neuropathy (MMN) (one patient). Nineteen patients were classified as non-inflammatory, including idiopathic neuropathy (nine patients), hereditary neuropathy (five patients), a neuropathy caused by vitamin B deficiency (three patients) or diabetic neuropathy (two patients).

Patients were classified as painless when they presented an NRS = 0 (fourteen patients; 50%), and as painful with an NRS ≥ 4 (fourteen patients; 50%). From the full cohort of PNP patients, at the time of inclusion fifteen patients had been treated for their neuropathy with either an immunosuppressive/immunomodulatory drug (six patients) and/or pain treatment (twelve patients), while thirteen received no treatment.

Further laboratory and electrophysiological data are given in Table 3. The median CRP value in serum was 0.24 mg/dl (range 0-2.37 mg/dl). In CSF, patients presented a median of 2 leukocytes/µl (range 0-6 cells/µl) and a median total protein of 46.3 mg/dl (range 19.3-194.6 mg/dl). Nerve conduction studies showed a median sural nerve sensory nerve action potential (SNAP) of 4.9 µV (range 2.5-12.6 µV) with a median nerve conduction velocity (NCV) of 45.9 m/s (range 32.4-57.3 m/s), and in the tibial nerve a median compound motor action potential (CMAP) upon stimulation at the ankle of 5.4 mV (range 0.1-23.4 mV) and a median nerve conduction velocity (NCV) of 38.3 m/s (range 26.8-52.2 m/s).

Patients that presented with acute headaches of unclear etiology (AH) and had a lumbar puncture to exclude meningitis or a subarachnoid bleed were used as controls for the CSF data. Only patients that had less than four leukocytes/µl in their CSF and normal protein values were included. After full clinical work-up, five of these patients had some signs of inflammation (elevated CRP, n = 3; sinusitis, n = 1; or elevated ANA and ANCA titers, n = 1). Thus, the AH group was divided into two groups (unclear headache, UH, and inflammatory headache, IH), and only patients with UH were further used as controls for inflammatory markers (Table 4, Figure 1A). Twelve healthy volunteers (HC) were included as controls for NFL and inflammatory markers in serum. HC and AH did not differ in age or NFL levels in pg/ml in serum (Figures 1B, C), and in both groups a positive correlation between age and NFL was present (HC, p < 0.05; AH, p < 0.01) (Figure 1D). When separating AH into UH and IH, NFL levels did not differ between groups, but IL-6 was higher in IH in comparison to HC (p < 0.001) (Figure 1E). These results supported our decision to continue the study on inflammatory markers with the UH exclusively, while the whole cohort (AH) was used for the NFL analysis.

NFL was measured in serum and CSF in twenty-eight patients with PNP in comparison to ten age-matched AH (Figures 2A, B). PNP patients had higher serum levels of NFL than AH (p < 0.01) (Figure 2C). There was a positive correlation between age and NFL levels in AH (p < 0.01) but not in PNP patients (Figure 2D). NFL levels correlated negatively to sural nerve SNAP in PNP patients (Figure 2E), which translated into higher levels of NFL in patients with abnormal SNAP values (< 5 µV) in comparison to those with normal SNAP (≥ 5 µV) (p < 0.01) (Figure 2F).

The same analysis was performed with CSF samples (Figure 3A) and we observed higher levels of NFL in PNP patients in comparison to AH (p < 0.001) (Figure 3B). Again, a correlation between age and NFL was present in AH (p < 0.001) but not in PNP patients (Figure 3C). Furthermore, our results showed a negative correlation between the levels of NFL in CSF and the MRC sumscore (Figure 3D), therefore suggesting an involvement of NFL release in the severity of the neuropathy.

In twenty-seven out of twenty-eight PNP patients and five UH, we analyzed the gene expression of the receptors TRPV1 and TLR4; the deacetylase SIRT1; the pro-inflammatory cytokines TNFα, IL-1β, IL-6 and IL-8, and chemokine CCL2; the anti-inflammatory cytokine IL-10; and the microRNAs miR-146a-5p, miR-132-3p and miR-155-5p in whole blood (Figure 4A). The expression results of PNP patients, with painful and painless PNP, and inflammatory and non-inflammatory PNP are displayed as fold change in comparison to UH in Table 5. Comparisons between PNP patients and controls, as well as between painful and painless PNP subgroups, and inflammatory and non-inflammatory did not show any differences in the expression of these components.

Interestingly, PNP patients were next grouped into patients with and without systemic inflammation based on increased CRP levels in the serum (cut-off 0.5 mg/dl). Patients with systemic inflammation had a higher expression of IL-10 in comparison to those with normal CRP levels (≤0.5 mg/dl) (p < 0.001) (Figure 4B) and a positive correlation was found between CRP levels and IL-10 expression (p < 0.01) (Figure 4E). Furthermore, PNP patients with high CSF total protein (> 50 mg/dl) also presented lower expression of TNFα (Figure 4C) and higher expression of miR-155 (Figure 4D) than those with normal levels (≤ 50 mg/dl). Moreover, a multivariate correlation analysis (Figure 4E) showed that CCL2 correlated positively with the duration of the disease (p < 0.001) while negatively with several neuropathy scores such as sural nerve SNAP amplitudes (p < 0.01) or GCPS, indicating a relation between inflammation and axonal degeneration.

As indicators of systemic inflammation, we analyzed the levels of the pro-inflammatory cytokines TNFα, IL-1β, IL-6 and IL-8, the chemokine CCL2, and the anti-inflammatory cytokine IL-10 in sera from twenty-eight patients with PNP and five UH (Figure 5A). Furthermore, we measured the levels of nerve growth factor beta (NGF-β) in sera, since it is involved in nociceptive processing and a target of novel analgesics (40). The CSF from the same cohort was available and we measured the levels of the pro-inflammatory cytokines IL-6 and IL-8, and the chemokines CCL2 and fractalkine (Cx3CL1). TNFα, IL-1β and IL-10 were not detected in these samples (data not shown). All results are detailed in Table 6.

CCL2 was present in higher levels in serum in PNP in comparison to UH, while no differences were discovered for the other analyzed markers. CCL2 was specifically upregulated in PNP patients under analgesic treatment (p < 0.05, Figure 5B). No differences were discovered between the PNP subgroups inflammatory and non-inflammatory. On the other hand, patients with a painful PNP had higher levels of IL-10 in sera than those with no pain (p < 0.01, Figure 5C). A multivariate analysis showed that the levels of IL-10 positively correlated with the pain scores in NRS (p < 0.01), NPSI (p < 0.05) and GCPS (p < 0.05) (Figure 5D).

The proteins present in CSF, such as NFL or inflammatory components, may be intrathecally produced or have infiltrated from the blood vessels due to a permeabilization of the B-CSF barrier. In order to study the origin of NFL and cytokines in CSF, we compared the levels of albumin between CSF and serum (albumin ratio in CSF/in serum: Q_Alb) and divided the patients in subgroups according to the B-CSF barrier integrity: Normal B-CSF barrier integrity (Q_Alb < 0.007) and mild (0.007 < Q_Alb < 0.01), moderate (0.01 < Q_Alb < 0.02) and severe (Q_Alb > 0.02) B-CSF barrier dysfunction (Figure S1A) (41). Due to a low number of patients with severe B-CSF barrier dysfunction (2), moderate and severe groups were analyzed together. A moderate/severe B-CSF barrier dysfunction was present more often in inflammatory neuropathies than normal or mild dysfunction (Figure S1B). Furthermore, the albumin ratio (Q_Alb) correlated with the severity of the clinical phenotype (Figure S1C). As expected, the levels of total protein in CSF increased as the B-CSF barrier dysfunction became more severe (Figure S1D), thus indicating a higher permeabilization or a relation between an intrathecal production and a B-CSF barrier dysfunction.

NFL measurements in serum and CSF in the different subgroups of twenty-eight PNP patients and ten AH (Figure 6A) showed higher levels in those patients with moderate/severe B-CSF barrier dysfunction in comparison to those with a normal B-CSF barrier integrity and to AH (Figures 6B, C). This resulted in a constant Q_NFL (NFL ratio in CSF/serum) among subgroups and between PNP and AH (Figure 6D). Furthermore, NFL in serum and CSF correlated positively with the severity of the clinical phenotype (Figure 6E), suggesting that the release of NFL might be a consequence of the severe neurodegeneration.

Interestingly, we found a correlation between the levels of NFL and IL-8 in CSF (Figure 6F), thus indicating a relation between neurodegeneration and inflammation. With this discovery, we decided to study the levels of cytokines between CSF and serum, in order to determine their origin, in the different subgroups of twenty-eight PNP patients and five UH (Figure 7A). This analysis showed that PNP patients with a moderate/severe B-CSF barrier dysfunction present higher levels of IL-6 and IL-8 in CSF and a higher Q_IL-6 and Q_CCL2 (cytokine ratio in CSF/in serum), than those with a normal B-CSF barrier integrity (Figure 7B). A multivariate analysis showed that the levels of IL-6 in CSF correlated with the overall severity, thus indicating a relation between the production or release of IL-6 and severe PNP (Figure 7C). Interestingly, a second multivariate analysis showed that the levels of most cytokines analyzed in CSF correlated with each other (Figure 7D), therefore suggesting a common stimulus triggering their production, or their involvement in a common pathway

In order to complete our study with the analysis of lipid compounds involved in pro-inflammatory pathways, we performed LC-MS to measure the levels of carnitine, palmitoylcarnitine and oleoylcarnitine in serum and CSF from twenty-three PNP patients and five UH (Figure 8A). While carnitine and palmitoylcarnitine in serum did not differ between PNP and UH, oleoylcarnitine was found in higher levels in patients with PNP in comparison to UH (Figure 8B). Carnitine was also not different in CSF between PNP patients and UH (Figure 8C). Palmitoylcarnitine and oleoylcarnitine were not detected in CSF samples (data not shown).

Our results also showed that PNP patients with high CSF protein (>50mg/dl) had higher levels of carnitine in CSF and a lower serum/CSF ratio than those with normal protein levels (p < 0.001) (Figure 8D).

Furthermore, we observed that the levels of palmitoylcarnitine and oleoylcarnitine in serum positively correlated in patients with PNP (p < 0.0001) and in UH (p < 0.001) (Figure 8E). Interestingly, a multivariate analysis found a positive correlation between the levels of carnitine in serum and the mTCNS (p < 0.001) as well as between the levels of carnitine in serum and the overall severity (Figure 8F).