Introduction

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease resulting from a progressive degeneration of motoneurons (MNs) in the motor cortex, brainstem and spinal cord [1], leading to a worsening muscular weakness and respiratory failure, with a median survival of 3 years from symptom onset [2]. Most of the cases are sporadic (sALS), while a family history was reported in 10% of cases (fALS) [3]. The disease may occur in several clinical phenotypes characterized by a wide variability of course and survival [4, 5].

The diagnostic work-up in ALS is mainly guided by the exclusion of ALS-mimicking diseases. Several studies have tried to identify biological markers of the disease burden [6,7,8]; however, confounding and conflicting results have been reported. To date, only the neurofilaments have been proven to be markers of disease severity, indicating their potential prognostic role [9,10,11,12].

One of the most debated biomarkers proposed for diagnosis of ALS is the Tau protein [13,14,15,16], a microtubule-associated protein that plays the role of microtubule stabilizer and assembler promoter [17]; it can be phosphorylated by specific kinases in the phospho-Tau protein isoform (p-Tau), leading to its separation from tubuline, with microtubule instability and disintegration. Increased CSF total Tau (t-Tau) levels have been found in several CNS diseases, including Alzheimer’s disease, encephalomyelitis and ischemic stroke [18,19,20]. Recent studies raised new interest in the diagnostic power of t-Tau and p-Tau in ALS, highlighting that reduced levels of p-Tau and p-Tau/t-Tau ratio in CSF could be suggestive of ALS [21]. On the other hand, other studies refuted this hypothesis, reporting an increase in t-Tau and questioning on the proposed role of p-Tau as a diagnostic biomarker of ALS [22].

The aim of this study is to test the diagnostic and prognostic role of CSF t-Tau, p-Tau protein, p-Tau/t-Tau ratio compared to that of neurofilament light (NFL) in patients with ALS, amyotrophic lateral sclerosis mimicking (AM) and other non-neurodegenerative diseases (ONND).

Patients and methods

Patients

Eighty-five incident ALS patients, 33 definite and 52 probable ALS according to the revised El Escorial criteria [23], were recruited. Only sporadic ALS, genotyped as negative for mutations in major ALS genes (SOD1, TDP43, FUS, C9orf72), and not responding to FTD diagnostic criteria [24], have been included in the study. Each patient underwent clinical, neurophysiological, neuropsychological and MR evaluation. Exposure to environmental toxins (Y/N) was reported by patient history. Disease duration from symptom onset; site of onset (bulbar or spinal); prevalence of UMN or LMN at onset; time interval from symptom onset to diagnosis (ODI) were evaluated at time of inclusion in the study. ALSFRS-r [25], manual muscle testing (MMT) [26]; time to generalization (time interval between disease spreading from spinal to bulbar district or vice versa—TTG) [27], progression rate [(48–ALSFRS-r)/disease duration in months] [28] and Sniff Nasal Inspiratory Pressure (SNIP) [29] were evaluated at the time of inclusion in the study and after 1-year follow-up. After the evaluation of general cognitive status with the MMSE, patients underwent a standardized frontal assessment battery, including Cognitive Behavioural Screen (ALSCBS) [30] and Edinburgh Cognitive and Behavioural ALS Screen (ECAS) [31]. No patient assumed riluzole at the time of inclusion in the study.

Control groups included 30 patients with AM (n.5 hereditary spastic paraparesis, n.4 spinal muscular atrophy, n.6 multifocal motor neuropathies, n.15 spondylotic myelopathy) and 51 patients with ONND (including migraine, CIDP, diabetic neuropathies, Bell’s palsies). Routine blood tests and brain MRI were normal in all patients.

All cases (ALS and Control patients) were consecutively recruited at the diagnosis time amongst patients attending the Neurology Unit of our Department, from March 2009 to May 2017. The three diagnostic groups were age- and sex-matched. The ALS cohort was followed until May 2018. Population’s demographic and clinical characteristics are listed in Table 1.

Table 1 Demographic and Clinical features—at inclusion in the study—of ALS, AM and ONND diagnostic groups
Full size table

This study was approved by the local ethics committee. Patient or family members’ written consent was obtained from each participant.

CSF t-Tau, p-Tau and NFL assay

All patients underwent spinal tap for CSF analysis included in the diagnostic work-up of the disease. CSF samples were collected according established procedures, in polypropylene tubes. Lumbar puncture was performed in the morning (from 9.00 a.m. to 11.00 a.m.), then the specimen was centrifuged as soon as possible from the withdrawal (within 60 min), aliquoted in smaller polypropylene tubes. Considering the high susceptibility of NFL to degradation by proteases [32], a protease inhibitor (AEBSF-4-2-aminoethyl benzenesulfonyl fluoride hydrochloride Sigma-Aldrich; St. Louis, MO, USA) was added to the aliquot of CSF used for NFL assay. All aliquots were stored at − 80 °C until analysis.

Commercially available ELISA kits were used for CSF t-Tau and p-Tau assay (Innogenetics; inter-assay variability coefficient < 10%), as previously reported [14, 16, 22, 33], and for NFL assay (ELISA test Uman Diagnostic AB; Umea, Sweden) (inter-assay variability coefficient < 9%) [9, 11, 34].

Statistical analysis

Demographic and clinical variables were expressed as mean (± SD) or frequencies for continuous and categorical variables. Non-parametric statistics were used because of the non-normal distribution of most of the variables. The Kruskal–Wallis test for multiple independent groups was used for data comparison, followed by the pairwise Dunn’s post hoc test. Receiver operating curve (ROC) was used to calculate an optimal cut-off concentration of CSF t-Tau, p-Tau, p-Tau/t-Tau ratio and NFL to discriminate patients with ALS from AM and ONND. Correlations between parameters were calculated using Spearman’s Rho.

ALS survival was evaluated with tracheostomy or death as events, recording time in months from the symptom onset of the disease. Patients who were alive and had not undergone tracheostomy by the last follow-up were censored. The prognostic ability of CSF t-Tau, p-Tau and NFL to predict the risk of the hard end point (tracheostomy or death) was assessed using univariate stepwise forward multivariate Cox model and reported as hazard ratios (HRs) with their 95% confidence intervals (CIs), after stratifying patients according to the median of CSF t-Tau (184 pg/mL), p-Tau (32 pg/mL) and NFL (6103 ng/L). Time-to-event analysis was performed using the Kaplan–Meier method and the survival curves were compared using the log-rank (Mantel–Cox) test in ALS stratified according to the median of CSF t-Tau, p-Tau and NFL concentration.

Statistical analysis was performed with IBM SPSS statistics data editor version 22.

Results

CSF t-Tau, p-Tau levels and p-Tau/t-Tau ratio did not correlate with CSF NFL levels (rs = 0.073, p = 0.59; rs = − 0.093, p = 0.49; rs = − 0.237; p = 0.076, respectively).

CSF t-Tau, p-Tau/t-Tau ratio and NFL significantly differed between groups (Kruskal–Wallis test, p < 0.001). In particular ALS patients had significantly higher t-Tau levels and lower p-Tau/t-Tau ratio than AM and ONND (Dunn’s post hoc test). On the contrary, CSF p-Tau levels did not differ between ALS, AM and ONND. The concentration of CSF NFL was significantly higher in ALS than in all other diagnostic groups (Table 2). A boxplot of CSF t-Tau, p-Tau, p-Tau/t-Tau ratio and NFL in the diagnostic groups is showed in Fig. 1a–d.

Table 2 CSF t-Tau, p-Tau levels, p-Tau/t-Tau ratio and NFL in ALS, AM and ONND cohorts
Full size table
Fig. 1
figure 1

CSF t-Tau levels (a), p-Tau levels (b), p-Tau/t-Tau ratio (c) and NFL levels (d) in ALS, AM and ONND cohorts (p values refer to post hoc Dunn’s test). a ALS vs AM (p = 0.005); ALS vs ONND (p < 0.001); b ALS vs AM (ns); ALS vs ONND (ns); c ALS vs AM (p = 0.006); ALS vs ONND (p < 0.001); d ALS vs AM (p < 0.001); ALS vs ONND (p < 0.001)

Full size image

Using ROC analysis, the optimal diagnostic power of CSF t-Tau in discriminating ALS from AM was defined by a cut-off of 143 pg/mL that resulted in a sensitivity of 69% and specificity of 60%. For the discrimination of ALS from ONND the best diagnostic performance was obtained with a CSF t-Tau cut-off value of 106 pg/mL, with a sensitivity of 88% and specificity of 51%. p-Tau protein had a much weaker discriminative power in discriminating ALS from AM (best cut-off 32.5 pg/mL; sensitivity 51%, specificity 54%) and ONND (best cut-off 35.5 pg/mL; sensitivity 45%, specificity 63%). Similarly, p-Tau/t-Tau ratio was not efficient in discriminating ALS from AM (best cut-off 0.19, sensitivity 41%, specificity 34%) and ONND (best cut-off 0.20, sensitivity 38%, specificity 16%) (Fig. 2a, c).

Fig. 2
figure 2

Receiver Operating Curve (ROC) analysis. Area under curve (AUC) of CSF t-Tau protein: 0.70 (95% confidence, limits of area 0.6–0.8) (a, solid line), p-Tau protein: 0.54 (95% confidence, limits of area 0.42–0.66) (a, dotted line) and NFL: 0.83 (95% confidence, limits of area 0.72–0.94) (b, dashed line) for discrimination between ALS and AM. Area under curve (AUC) of CSF t-Tau protein: 0.74 (AUC: 0.74; 95% confidence interval, limits of area 0.64–0.83) (c, solid line), p-Tau protein: 0.54 (95% confidence limits of area 0.44–0.64) (c, dotted line) and NFL: 0.82 (95% confidence, limits of area 0.70–0.93) (d, dashed line) for discrimination between ALS and ONND

Full size image

A CSF NFL concentration above 2180 ng/L yielded the optimal discrimination between ALS and AM at sensitivity of 86% and specificity of 87%, with a positive predictive value (PPV) of 92% and a negative predictive value (NPV) of 51%. CSF NFL showed good diagnostic performance also in discriminating ALS from ONND with an optimal cut-off of 2720 ng/L that resulted in 83% sensitivity and 75% specificity (Fig. 2b, d).

Relations with demographic and clinical characteristics

In pooled diagnostic groups, CSF t-Tau, p-Tau and NFL correlated with age (rs = 0.377 p = 0.001, rs = 0.283 p = 0.01 and rs = 0.427 p = 0.001, respectively). However, when considering the groups individually, this correlation was not confirmed in ALS. Gender did not influence t-Tau, p-Tau, p-Tau/t-Tau ratio and NFL levels.

CSF t-Tau, p-Tau, p-Tau/t-Tau ratio and NFL did not differ in ALS stratified for occupational exposure to environmental toxins, site of onset (bulbar/spinal and UMN/LMN), cognitive impairment. Furthermore, t-Tau did not correlate with ALSFRS-r, MMT and TTG, all calculated both at the time of inclusion of the study and after 1-year follow-up. On the other hand, CSF t-Tau protein positively correlated with progression rate at the time of spinal tap, and inversely with ODI (robust trend towards significance) and SNIP measured at 1-year follow-up. No correlations were found between CSF p-Tau and all clinical features (Table 3).

Table 3 Correlation between CSF t-Tau, p-Tau, p-Tau/t-Tau ratio and ALS clinical features
Full size table

CSF NFL inversely correlated with ODI, ALSFRS-r score at spinal tap and after 1-year follow-up, with TTG, MMT at spinal tap and after 1 year follow-up, FVC measured after 1-year follow-up and directly with progression rate at spinal tap and—even stronger—after 1-year follow-up (Table 4).

Table 4 Correlation between CSF NFL levels and ALS clinical features
Full size table

The Kaplan–Meier analysis showed that ALS patients with high levels of CSF t-Tau at the time of diagnosis (higher than 184 pg/mL) had a short survival (log-rank test χ2 = 5.219; p = 0.022) (Fig. 3a). The median survival was 31.20 months in the group with high CSF t-Tau levels and 52.96 months for low CSF t-Tau (p = 0.001). Cox univariate hazard ratio of tracheostomy/death was 2.08 (p = 0.021, 95% CI 1.09–3.94), meaning that patients with higher CSF levels of t-Tau had a two-fold increased risk of tracheostomy/death at any time during the follow-up compared to patients with lower levels of CSF t-Tau. After adjusting for gender, age, site of onset, ODI, TTG, cognitive impairment and SNIP measured after 1-year follow-up, the significance was not modified. On the other hand, the inclusion of progression rate as covariate in the multivariate model led to a loss of significance for t-Tau as predictor of death/tracheotomy risk (Table 5).

Fig. 3
figure 3

Kaplan–Meier survival curve in ALS stratified according to the median CSF t-Tau levels (184 pg/mL). Survival is reduced in patients with CSF t-Tau above (a, dashed line) compared with those below (a, solid line) median value (log-rank test χ2 = 5.124; p = 0.02). Kaplan–Meier survival curve in ALS stratified according to the median CSF NFL levels (6103 ng/L). Survival is reduced in patients with CSF NFL above (b, dashed line) compared with those below (b, solid line) median value (log-rank test χ2 = 7.365; p = 0.007)

Full size image
Table 5 Summary of univariate and multivariate Cox analysis for survival to death/tracheostomy in the cohort of ALS patients, referred to t-Tau levels
Full size table

No difference in overall survival was found between patients with high/low p-Tau CSF levels neither for high/low p-Tau/t-Tau ratio.

CSF NFL levels were predictive of survival in ALS cohort, with poor survival in patients with NFL levels higher than 6103 ng/L (log-rank test χ2 = 7.365; p = 0.007) (Fig. 3b). Cox multivariate model revealed that higher levels of CSF NFL were related to a 4.22 hazard ratio of death/tracheotomy in each time of the follow-up (p = 0.007, 95% CI 1.48–12.03), a result confirmed after adjusting for gender, age, site of onset, ODI, TTG, progression rate, cognitive impairment and SNIP measured after 1-year follow-up. However, in the multivariate model, also the progression rate and ODI represented significant prognostic factors for survival even after adjusting for all the other clinical and demographic covariates (Table 6).

Table 6 Summary of univariate and multivariate Cox analysis for survival to death/tracheostomy in the cohort of ALS patients, referred to NFL levels
Full size table

Discussion

Our cross-sectional study confirmed high CSF t-Tau levels in ALS patients, reflecting the burden of neurodegeneration that characterizes the disease [22]. Furthermore, the reduction of p-Tau/t-Tau ratio is due to the increased t-Tau and not to decreased p-Tau, differently from what reported in a previous study [21]. On the other hand, the process of phosphorylation of Tau, that is one of the major pathological finding in AD, occurs in a lesser extent in other neurodegenerative disorders, such as ALS [33].

The accuracy of CSF t-Tau in distinguishing ALS from ONND showed a sensitivity of 88% and specificity of 51%; a little increase of the diagnostic accuracy is found in distinguishing from the diseases with clinical features similar to ALS (AM), where the specificity rises up to 60%. Nevertheless, the significance of these values is suboptimal in comparison to that of CSF NFL which better discriminated ALS from ONND or AM (specificity of 75 and 87%, respectively), also according to other authors [10, 11, 33]. These data suggest a lower diagnostic power of t-Tau protein for ALS [33, 35], making it an unreliable diagnostic biomarker and confirming NFL as a reference diagnostic biomarker for ALS, as already previously stated [9]. For these reasons, the diagnostic potential of CSF t-Tau and p-Tau levels has been extensively investigated in ALS, with conflicting results. Some authors found higher levels of t-Tau protein in ALS patients than in control cohorts [15, 22], differently from other studies where the authors concluded that Tau protein is not a useful diagnostic biomarker for ALS [11, 14, 16, 36, 37]. In other studies, attention was focused on the CSF p-Tau with conflicting results too [21, 22].

A relationship between high CSF t-Tau levels, severity of disease (expressed by progression rate [38, 39]) and respiratory impairment (as expressed by SNIP measured at 1-year follow-up) has been observed. These results partially conflict with previous studies in which the prognostic role of this CSF biomarker was resized [14, 15, 33]. The potential prognostic value of t-Tau is confirmed after stratifying the cohort according to the median CSF t-Tau levels: we found in fact that the overall survival was shorter in patients with higher levels of CSF t-Tau protein. Anyway, when we used the Cox multivariate model to quantify the predictive power of high CSF t-Tau levels on the risk of death/tracheostomy we found that this biomarker was predictive also when the model was adjusted for sex, age and SNIP but it lost its predictive power with the insertion of the progression rate in the model, a parameter with a well-established and evidently stronger predictive power [38, 39]. CSF p-Tau levels showed no relation with the survival. Given that, when we compare the diagnostic and prognostic potential of CSF t-Tau in comparison to that of NFL, our data confirm the absolute superiority of CSF NFL, considered as the most reliable and efficient tool for diagnosis and prediction of clinical progression and survival in ALS patients. Nevertheless, the significant relation with survival observed also for CSF t-Tau suggests a prognostic potential of this biomarker.

This study was approved by the local ethics committee. Patient or family members’ written consent was obtained from each participant. The study has been performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments.