Serum neurofilament light levels correlate with severity measures and neurodegeneration markers in autosomal dominant Alzheimer’s disease

Autosomal dominant Alzheimer’s disease (ADAD), with its almost 100% penetrance and relatively predictable age of onset, allows the evaluation of disease-modifying treatments at early or pre-symptomatic stages of the disease [1]. Two trials, the Dominantly Inherited Alzheimer Network Trials Unit and the Alzheimer’s Prevention Initiative [2, 3], are already ongoing. Thus, there is considerable interest in finding non-invasive biomarkers that could track the disease progression or provide evidence of disease modification. Cerebrospinal fluid (CSF) biomarkers have shown strong correlations with clinical and cognitive measures in ADAD [4, 5]. However, repeated CSF sampling is not always feasible or well tolerated. Blood-based biomarkers are less invasive and allow more frequent determinations, although they are challenging due to the lower concentration of brain analytes. Recently, measurement of neurofilament light (NfL) concentration using an ultrasensitive immunoassay on the Single molecule array (Simoa) platform has been demonstrated to be feasible and reliable, both in sporadic Alzheimer’s disease (AD) [6] and ADAD [7], as well as in other neurodegenerative or neuroinflammatory diseases such as frontotemporal dementia [8], progressive supranuclear palsy [9], Huntington’s disease [10], and multiple sclerosis [11].

In this study, we analysed serum NfL levels in a Spanish cohort of ADAD individuals and investigated whether they were associated with clinical markers of disease severity and CSF biomarkers.

Twenty-two participants were symptomatic mutation carriers (SMC) and 38 were asymptomatic, with 20 of these being mutation carriers (AMC) and 18 being asymptomatic non-carriers (CTR). AMC were more than a decade younger than their parental age at onset (mean ± SD EYO, −14.26 ± 7.68 years; Table 1). As expected, SMC were significantly older and had lower scores on cognitive and clinical measures. SMC showed higher serum NfL levels compared with AMC and CTR (as well as higher total tau, phosphorylated tau, and NfL, and lower Aβ1–42 CSF levels). No significant differences were observed between AMC and CTR in demographic, clinical, or biochemical variables included in Table 1 except for EYO.

Serum NfL correlations with age, EYO, and clinical, cognitive, and CSF biochemical measures are shown in Table 2 and Fig. 1. In MC, serum NfL levels showed a negative correlation with Mini-Mental State Examination (MMSE; rho = −0.77, p < 0.001) and positive correlations with EYO and CDR sum of boxes (CDR-SOB) (rho = 0.75 and 0.70, respectively, both p < 0.001) (Fig. 1a-c and Table 2). When the analysis was restricted to SMC, serum NfL remained inversely correlated with MMSE score (rho = −0.48, p = 0.02); when the analysis was restricted to AMC only, we observed a weaker correlation with EYO (rho = 0.41, p = 0.073 (two-tailed), p = 0.037 (one-tailed)). Serum NfL levels showed significant correlations in MC with several cognitive tests besides MMSE, such as the Free and Cued Selective Reminding test sub-scores: total free recall (rho = −0.54, p = 0.02); total recall (rho = −0.68, p < 0.001); delayed free recall (rho = −0.57, p = 0,01); delayed total recall (rho = −0.57, p < 0.001); total Digits score (rho = −0.57, p < 0.01); Boston naming test (rho = −0.45, p = 0.08); trail making test part A (rho = 0.64, p < 0.001), and trail making test part B (rho = 0.60, p = 0.002). All these cognitive measures also showed strong significant correlations with MMSE.

Serum NfL correlated with CSF NfL (rho = 0.72, p < 0.001), total tau (rho = 0.71, p < 0.001), and phosphorylated tau (rho = 0.71, p < 0.001) (Fig. 1d-f) but not with Aβ1–42 levels (rho = −0.32, p = 0.12) in MC. In CTR, serum NfL correlated with CSF total tau levels, but not with other markers.

Six available longitudinal serum samples from MC obtained at least 1 year apart were also analysed. The mean ± SD annual longitudinal change was 1.67 ± 1.84 ng/L (Fig. 2). This value correlated with the baseline EYO (rho = 0.89, p = 0.019). No longitudinal samples from CTR were available for comparison.

We found that serum NfL concentrations correlated with severity measures in ADAD mutation carriers, as well as with CSF NfL, total tau, and phosphorylated tau levels. Serum NfL levels in MC correlated with the estimated years to symptom onset (EYO). In participants who were already symptomatic (SMC), NfL levels correlated with global cognition (MMSE). SMC showed increased NfL levels compared with both CTR and AMC but we did not find a significant difference between the CTR and the AMC groups who were, as a group, 14 years younger than the parental age of onset. In a previous study, Weston and colleagues studied serum NfL with the same methodology in 48 individuals from a different ADAD cohort [7]. In this cohort, like ours, serum NfL correlated with EYO and cognitive measures across MC. Different studies suggest that NfL is a non-specific marker of neurodegeneration [15]. Serum NfL levels are elevated and may reflect disease intensity, not only in sporadic AD [6] but also in amyotrophic lateral sclerosis [16], frontotemporal lobar degeneration [8], Huntington’s disease [10], and Creutzfeldt-Jakob disease [17]. However, more longitudinal data are needed to evaluate the reliability of serum NfL in monitoring the progression of neurodegeneration. Promisingly, serum NfL levels are normalized in response to disease-modifying therapies in multiple sclerosis [18].

NfL may also be useful for predicting symptom onset in genetic neurodegenerative conditions such as frontotemporal dementia [19], genetic Creutzfeldt-Jakob disease [17], or genetic amyotrophic lateral sclerosis [20], although it is still unclear when increased levels can first be detected in these different disorders and whether NfL rises in the asymptomatic phase of each disease or only with symptom onset.

Even if the mean concentrations were similar in the two studies (CTR 12.7 ± 7.2 pg/mL; AMC 16.7 ± 7.7 pg/mL, and SMC 46.0 ± 20.8 pg/mL), Weston and colleagues reported significantly increased serum NfL in AMC compared with CTR in ADAD; that differed from our results. Differences in the mean EYO in AMC could account for this discrepancy, as serum NfL correlated with EYO in both cohorts and the asymptomatic carriers were closer to symptom onset in the UCL cohort than in our study. Similarly, although they should be considered exploratory results due to the limited sample size and the absence of controls, in our study the serum NfL levels increased longitudinally with the rate of change in SMC higher than in the AMC, suggesting that the magnitude of annual NfL change might increase with disease progression. Thus, these results would support the idea of a progressive increase in serum NfL and, thus, neurodegeneration during the asymptomatic phase of the disease that might accelerate around the time of symptom onset. No individuals in the severe phases of the disease were included in any of the studies to evaluate if the increase in serum NfL is maintained in advanced phases of the disease.

Serum NfL levels significantly correlated with CSF NfL levels in the whole cohort and across MC as described in other studies in different neurodegenerative diseases [6, 15, 19]. Serum NfL also correlated with CSF total tau and phosphorylated tau levels, although these correlations were not statistically significant within diagnostic groups, suggesting that the pathological condition reflected by each biomarker may diverge in different stages of the AD continuum, as has been shown in sporadic AD [6]. However, we cannot rule out a type II error (false negative result) in our study due to the small sample size when AMC and SMC were analysed separately.

There are several limitations in this study. Although the sample size is relatively large for the rarity of ADAD, the sample size limited the interpretation of some of the analyses and it would be of interest to explore these results in larger cohorts. It would be of great interest to know if different mutations have a different effect on serum NfL levels, but unfortunately the low numbers of subjects from each mutation preclude this analysis. The lack of repeated individual samples in most of the participants also limits the interpretation of the reliability of the marker at the individual level in longitudinal studies.

In summary, serum NFL levels in ADAD MC are associated with expected time to symptom onset, with clinical and cognitive measures, and with CSF neurodegeneration markers. These findings suggest that serum NfL may be a non-invasive biomarker for the prediction of symptom onset and potentially for tracking disease severity in ADAD.