NfL and pNfH are increased in Friedreich’s ataxia

A total of 99 patients (median age 38 years, range 16–68) with genetically confirmed Friedreich’s ataxia were recruited through the European Friedreich’s Ataxia Consortium for Translational Studies (EFACTS). All patients carried GAA repeat expansions on both alleles. Repeat lengths of the shorter allele varied between 67 and 1167 GAA repeats and on the longer allele between 200 and 1500 repeats. We assessed clinical disease severity with the Scale for the Assessment and Rating of Ataxia (SARA), a validated score that allows to quantify an individuals’ degree of ataxia and ranges from 0 (no ataxia) to 40 points (most severe ataxia) [16]. Age at onset was defined by first reported symptoms, and disease duration as the period between onset and time of sampling. In addition, 30 individuals (median age 48 years, range 18–68) were enrolled at the Department of Neurodegenerative Disorders, Hertie Institute for Clinical Brain Research, University Hospital Tübingen, as healthy controls. All controls were assessed by neurologists with special expertise in neurodegenerative diseases, ascertaining that none of them had a history or clinical signs of neurodegenerative disease or of any other major neurological disorder. The study has been approved by the institutional review board, and all subjects gave written informed consent prior to participation.

The biomaterial from Friedreich’s ataxia patients used in this study was provided by the centralized biomaterial bank of the medical faculty of the RWTH Aachen University (RWTH cBMB) and used in accordance with the biomaterial bank’s regulations and vote 206/09 of the ethics committee of the medical faculty of the RWTH Aachen University. Serum samples from healthy controls were provided by the biobank of the Hertie Institute for Clinical Brain Research (HIH), University of Tübingen, and used in accordance with the biomaterial bank’s regulations and vote 199/2011BO1 of the ethics committee of the medical faculty of the University of Tübingen. Samples were frozen at − 80 °C within 90 min after collection, and analysed without any previous thaw–freeze cycle.

NfL concentrations were analysed by single molecule array (Simoa) assay as previously described [11]. Inter-assay coefficients of variation (CV) for three native serum samples were 14.6%, 7.5%, and 2.1% for control samples with mean concentrations of 7.7 pg/ml, 22.6 pg/ml, and 77.4 pg/ml, respectively. The mean intra-assay CV of duplicate determinations for concentration was 3.7%. pNfH was quantified by a commercially available Kit (Quanterix) on the Simoa platform on a single run. The mean intra-assay CV of duplicate determinations for concentration was 9.5%.

Statistics were performed using nonparametric tests in bivariate analysis. Multiple regression models were used to assess dependency of NfL on age. Data in the results section report median and range, data in graphs represent median and 95% confidence interval (CI). Demographic data were compared using two-tailed Mann–Whitney test. For pNfH, the Mann–Whitney U test was used to compare values between patients and controls. A two-tailed Wilcoxon test was applied to compare paired data from longitudinal measurements in identical subjects. Correlation was assessed by computing Spearman r. One outlier was removed in the pNfH control group (pNfH = 346.1 pg/ml), by the robust regression and outlier removal (ROUT) method (Q = 0.1%) in GraphPad Prism 7. To assess a possible use of NfL measurements for diagnosis, an age-corrected ROC curve was constructed. This was done by first calculating a quadratic regression model for NfL dependent on age in the control group only. Second, the equation of this model was used to calculate predicted values of NfL for both, the control and the patient group. Third, the differences between observed and predicted values of NfL were calculated and used as continuous classificator in a ROC analysis. This led to the “age-corrected ROC curve”. Regression and ROC analyses were performed using SPSS 25 for Windows. All other analyses were performed with GraphPad Prism 7 for Mac.

To investigate whether serum NfL and pNfH might serve as biomarkers in Friedreich’s ataxia, we compared NfL serum levels between 30 healthy controls and 99 patients with Friedreich’s ataxia as well as pNfH serum levels in a subgroup of 9 controls and 20 patients. The median NfL concentration in controls was 21.15 pg/ml (range 3.6–49.3), while the concentration in Friedreich’s ataxia was significantly higher with 26.1 pg/ml (range 0–78.1; p = 0.002) (Fig. 1a). Similarly, pNfH levels were significantly elevated in patients compared to controls (controls 23.5 pg/ml, range 13.3–43.2; Friedreich’s ataxia 92 pg/ml, range 3.1–303; p = 0.0004) (Fig. 1b).

Controls (45.27 ± 14.11 years) were older than patients with Friedreich’s ataxia (38.37 ± 13.05 years; p = 0.02), but covered a comparable age range (controls 18–68 years, patients 16–68 years). Thus, age dependency of NfL was assessed in detail for both groups. In healthy individuals, there was a clear quadratic dependency of NfL on age (Fig. 2a, dashed line and Supplementary Figure S1a): in controls aged < 30 years, there was no increase of NfL with age. Between 30 and 50 years of age, there was a moderate increase and in the age range of 50–65, there was a steep increase (r-square linear model = 0.55, quadratic model 0.64, both p < 0.00001). In diseased subjects, no significant age dependency of NfL levels could be detected with a non-significant trend to a quadratic model (r-square linear model = 0.00, p = 0.90, quadratic model 0.05, p = 0.095). The area under the age-corrected ROC curve (cf. “Methods” section) was 0.78 (95% CI 0.69–0.86). Separate analysis within three equally sized strata of age (16–31 years, 32–47 years, 48–68 years) revealed areas under the ROC curve of 0.99 (CI 0.95–1.00, p < 0.001), 0.81 (0.70–0.92, p = 0.002) and 0.49 (0.28–0.70, p = 0.94). Thus, based on NfL measurements, there was an excellent classification in healthy or affected for younger individuals, a moderately good classification for middle aged individuals and no classification better than chance for older individuals.

We assessed serum NfL levels in correlation with disease severity as defined by the Scale for the Assessment and Rating of Ataxia (SARA) score. We found that serum NfL did not correlate with the SARA score (r = − 0.13, 95% CI − 0.32 to 0.08, n = 99; p = 0.22) (Fig. 3a). Similarly, there was no correlation between NfL concentration and age at onset (r = 0.05, 95% CI − 0.15 to 0.25, n = 99; p = 0.62) (Fig. 3b) or disease duration (r = − 0.06, 95% CI − 0.26 to 0.14, n = 99; p = 0.53) (Fig. 3c). In patients, there was a small, but significant inverse correlation between levels of NfL and the length of the smaller GAA repeat (allele 1) (r = − 0.24; 95% CI − 0.42 to − 0.03; n = 99, p = 0.02) (Fig. 3d), but not with the GAA repeat length in the larger allele (allele 2) (r = − 0.01; 95% CI − 0.21 to 0.20, n = 99; p = 0.95) (Fig. 3e).

Serum pNfH levels were measured in a subgroup of 20 patients that were selected a priori to represent moderately (SARA score 10–20) or severely affected (SARA score 30–40) individuals, to increase the visibility of potential differences from disease severity despite the small group size. Analogous to NfL, there was no correlation between pNfH level and SARA score (r = − 0.20, 95% CI − 0.60 to 0.28; n = 20, p = 0.41) (Supplementary Figure S2). Interestingly, there was even a tendency towards lower pNfH levels in more severely affected patients, albeit this tendency did not reach significance (SARA 10–20: 139.5 pg/ml, range 3.1–303, n = 10; SARA 30–40: 87.75 pg/ml, range 23.1–258.6, n = 10; p = 0.17) (Fig. 3f). pNfH also did not correlate with age, neither in controls nor in Friedreich’s ataxia patients (controls r = 0.10, p = 0.80, n = 9; patients r = − 0.11, 95% CI − 0.54 to 0.36, n = 20; p = 0.64) (Supplementary Figure S1c, d) or age at onset (r = 0.18, 95% CI − 0.30 to 0.59, n = 20; p = 0.44) (Fig. 3g). There were small, but non-significant correlations between pNfH concentration and disease duration (r = − 0.39, 95% CI − 0.72 to 0.08, n = 20; p = 0.09) (Fig. 3h) and also with GAA repeat length of allele 1 (r = − 0.32, 95% CI − 0.68 to 0.16, n = 20; p = 0.17), but not with repeat length of allele 2 (r = 0.06, 95% CI − 0.41 to 0.50, n = 20; p = 0.81) (Supplementary Figure S3).

To assess progression dynamics of NfL in Friedreich’s ataxia, we used a longitudinal approach in a group of 14 patients by measuring serum NfL at baseline (BL) and 2 years later (2FU). On individual level, we observed an increase of serum NfL in 9 of 14 patients (64.3%) while concentrations decreased in 4 patients (28.6%) and stayed the same in one patient (Fig. 4a). Overall, there was no significant change during the 2-year period (BL 27.5 pg/ml, range 5.3–53.1; 2FU 34.1 pg/ml, range 11–80.8; n = 14, p = 0.06) (Fig. 4a). While there was a significant increase in the SARA score over time (BL 21.7 points, range 6–32.5; 2FU 23.5 points, range 13.5–32.5, n = 14; p = 0.007) (Supplementary Figure S4), the individual dynamics of NfL (increase/decrease) did not match the dynamics of the SARA score (Fig. 4b), congruent to the lack of correlation between NfL levels and disease severity measured by SARA (see Fig. 3c).