Neurofilament markers in serum and cerebrospinal fluid of patients with amyotrophic lateral sclerosis

Abstract This study aims to determine the serum and cerebrospinal fluid (CSF) levels of neurofilament light chain (NFL) and phosphorylated neurofilament heavy chain (pNFH) in amyotrophic lateral sclerosis (ALS) patients, and to explore their feasibility as valid biomarkers for quantifying disease progression and predicting individual prognosis. 52 patients with ALS and 30 controls with noninflammatory neurological diseases were included. NFL and pNFH levels in serum and CSF were measured by enzyme‐linked immunosorbent assay. Our findings showed that serum and CSF levels of NFL and pNFH in ALS patients were significantly increased. These values were negatively correlated with disease duration (except CSF NFL with disease duration) and ALSFRS‐r score, and positively correlated with disease progression rate (DPR) and upper motor neuron (UMN) score, but did not correlate with bilateral median and ulnar nerve compound muscle action potential (cMAP) amplitudes (except a weak correlation between CSF NFL and cMAP amplitudes). The optimal cut‐off values with high sensitivity and specificity were obtained in ROC curve analysis to discriminate ALS from controls. Kaplan‐Meier survival curves illustrated that survival was significantly shorter for patients with higher neurofilament levels at diagnosis. The Cox proportional hazards regressions confirmed that NFL and pNFH were significant predictors of survival. Overall, NFL and pNFH in serum and CSF can be used as reliable biomarkers in ALS.

Whitney U test was used to determine the significant differences between two groups for normally distributed data or non-normally distributed data, respectively. A Pearson's correlation or Spearman's correlation test was performed for the correlation analysis. The area under the curve (AUC) and optimal cut-off levels were calculated using receiver operating characteristic (ROC) curves. Two-sided tests were conducted for all statistical analysis, and the level of statistical significance was set at p < 0.05. Kaplan-Meier methods were adopted to estimate the survival of ALS patients and differences among groups were assessed with log-rank test (Mantel-Cox).
Univariate Cox proportional hazards regressions were performed to identify predictors of the mortality hazard ratio (HR) in ALS patients and parameters with a p-value below 0.1 were subsequently tested in a multivariate Cox model via the stepwise forward approach.

| RE SULTS
The demographic and clinical characteristics of the participants are displayed in Table 1 and Tables S1 and S2. Levels of serum NFL (S-NFL), CSF-NFL, serum pNFH (S-pNFH) and CSF-pNFH in the ALS group were significantly higher than in the control group ( Figure   S1). Among different subgroups based on median values, the serum and CSF NFs levels were higher in rapidly progressive group than in slowly progressive group (Table S3). In correlation analysis between NFs levels and ALS clinical parameters, there were no significant correlations between NFs levels and age, height, weight, gender or onset site (Table S4). The NFs concentrations were negatively correlated with ALSFRS-r score and disease duration (except CSF NFL with disease duration, rho = −0.204, p = 0.227), and positively correlated with DPR and UMN score, but did not correlate with cMAP amplitudes (except a weak correlation between CSF NFL and cMAP amplitudes, rho = −0.361, p = 0.039) ( Figures S2-S4). In addition, positive correlation was evident between S-NFL and CSF-NFL (rho = 0.517, p = 0.001) as well as between S-pNFH and CSF-pNFH (rho = 0.809, p < 0.0001) ( Figure S5). The optimal cut-off values to discriminate between ALS and controls were calculated at 500 pg/ ml S-NFL (a sensitivity of 88.5% and specificity of 83.3%), 408 pg/ ml S-pNFH (a sensitivity of 100% and specificity of 83.3%), 647 pg/ mL CSF-NFL (a sensitivity of 97.3% and specificity of 96.7%) and 474 pg/ml CSF-pNFH (a sensitivity of 97.8% and specificity of 97.2%) in ROC curves ( Figure S6). 43 cases of ALS had been successfully followed up by telephone calls or clinic visits until September 2021. Kaplan-Meier survival curves showed that survival was significantly shorter for patients with higher serum/CSF NFL or pNFH levels at diagnosis (log-rank test p < 0.0001, Figure 1). Univariate Cox proportional hazards regressions identified S-NFL, CSF-NFL, S-pNFH, CSF-pNFH, ALSFRS-r score, DPR, UMN score and cMAP amplitudes as predictors of the mortality hazard ratio in ALS patients among all the tested factors (Table S5). Subsequently, parameters including ALSFRS-r score, DPR, UMN score and cMAP amplitudes were tested as covariates when serum/CSF NFs were separately introduced in the multivariate Cox models. The results showed that S-NFL (HR = 3.626, 95%CI: 1.538-8.551, p = 0.003) and S-pNFH (HR = 3.694, 95% CI: 1.569-8.698, p = 0.003) were independent predictors of survival.

| DISCUSS ION
In this study, we measured levels of NFL and pNFH in the serum and CSF of ALS patients and controls, and explored their feasibility as potential biomarkers in ALS.
We confirmed that levels of NFL and pNFH in the serum and CSF of ALS patients were substantially increased compared with the control group, which was consistent with the results of most previous studies. 2-6,10 Though we found a negative correlation between baseline NFs levels and ALSFRS-r, the rates of changes of ALSFRS-r during follow-up were not sufficiently available in this study and whether the baseline NFs levels could predict the decline in neuronal functions of ALS patients still needs to be investigated. Similarly, the relationship between NF levels and disease duration observed in the cross-sectional measurement design is barely due to changes in NFs levels along the natural history of the disease, but rather reflects the fact that aggressive ALS patients with a fast disease progression are referred faster to the neuromuscular centre and investigated with a shorter delay from symptom onset as compared with the slow progressors. As demonstrated in our study and previous research, [11][12][13][14][15] levels of NFL and pNFH in serum and CSF were positively correlated with DPR, warranting further evidence for the use of pNFH and NFL at an early disease stage as biomarkers for predicting ALS disease progression. The fluctuations of NF levels throughout the entire disease course still need to be further verified with sufficient follow-up data. A longitudinal study indicated a stable S-NFL concentration profile after a robust increase S-NFL level in ALS patients. 16 This Note: Data are presented as n, mean ±SD, or median (interquartile range).
*p < 0.001, significant difference between the ALS and control groups.

F I G U R E 1 Kaplan-Meier survival
curves. Significant separation of survival of ALS patients was noted when stratified by median levels of NFL and pNFH in serum and CSF samples trend is illuminated to be associated with the kinetics of neurofilaments clearance from axons which in turns depends on the rate of MNs degeneration. 17,18 The positive correlations between CSF and serum NFs levels and UMN score found in this study agree with another study using diffusion tensor imaging that highlighted associations between higher NFL levels and lower DTI fractional anisotropy and increased radial diffusivity in the CSTs of ALS patients 15 In contrast, we found no Notably, data for several factors known to influence survival in ALS including nutritional status, respiratory function, gene mutations and frontotemporal dementia were not available, so firm conclusions on independent roles for NFs levels especially in serum samples should be drawn with caution. The main limitation of our work is the small sample size which may not represent the general population.
More robust data from larger prospective longitudinal cohorts using regimented protocols and standardized detection methods would be needed to support the application of serum and CSF NFs as reliable biomarkers in clinical practice.

ACK N OWLED G EM ENTS
This work was supported by grants from Shanxi Science and Technology Department (201704D13111584), People's Republic of China.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available on request from the corresponding author.