*3.4. PASAT*

To see whether serum NfL is associated with cognitive decline in MS patients, we performed a correlation analysis of serum NfL and PASAT at all time points (Figure 4). After correcting for multiple testing, a significant correlation between PASAT at month 24 and serum NfL at time points 3 and 18 remained (Spearman *r* = 0.64 and 0.57 and adjusted *p*-value = 0.005 and 0.029, respectively).

**Figure 4.** Correlation matrix of Spearman correlation analysis for serum NfL and PASAT. Numbers in the cells are showing the respective Spearman correlation coefficient.

#### *3.5. EDSS in Patients with Active Disease within 24 Months*

To see whether serum NfL is associated with the disability in MS patients, we performed a correlation analysis of serum NfL and EDSS at all time points (Figure 5).

**Figure 5.** Correlation matrix of Spearman correlation analysis for serum NfL and EDSS. Numbers in the cells are showing the respective Spearman correlation coefficient.

We did not observe significant correlations for serum NfL and EDSS at any time point in the group with active disease (at least one relapse) within 24 months.

#### *3.6. Individual Serum NfL Courses in Patients Treated with Glatiramer Acetate*

The serum NfL courses of all patients treated with glatiramer acetate with available MRI scans (T2 lesions) are illustrated. Additionally, for every available time point, the EDSS and occurred relapses are shown (Figure 6).

**Figure 6.** *Cont*.

**Figure 6.** Illustration of individual serum NfL courses over 24 months in patients with initiation of disease-modifying treatment with glatiramer acetate after baseline. Green circles, red squares and blue triangles show serum NfL values (left y-axis), EDSS (right y-axis) and the number of T2 lesions (left y-axis), respectively. Vertical dashed lines show events of clinical activity in the form of a relapse at this time point.

### **4. Discussion**

Neurodegeneration and axonal loss are major hallmarks of MS [2]. NfL has been extensively investigated as a biomarker for those molecular processes [6,9,11]. Initially NfL was exclusively analyzed in CSF, but with improved analytical sensitivity, serum analyses became possible as well [17]. Serum NfL shows a good correlation with CSF level and thereby offers a window to monitor axonal loss in MS patients consecutively [18,22]. Therefore, numerous studies including serum NfL in MS are available and it is used frequently in clinical trials [9,11,23,30]. However, longitudinal serum NfL assessments are scarce [26,27], especially in individual MS patients before and after initiation of first-line therapies. For this purpose, we aimed at characterizing the influence of those therapies on serum NfL levels.

Our data suggested that serum NfL may be suitable as a marker for therapy responsiveness based on the following findings: (a) sNfL levels stayed at a consistent low level or even dropped significantly in relapse-free patients over time and (b) sNfL levels after 9 and 12 months were significantly lower in patients without relapse within 12 months compared with patients suffering from a relapse during this time period.

However, we want to point out that most MS patients in the early phase of the disease, which is the case for most of our patients, show serum NfL levels that are within a normal age-adjusted range [16].

Furthermore, our data showed that serum NfL levels were associated with relapses as they were higher in time points with a present relapse compared with non-relapse time points.

The individual serum NfL courses showed that effects that were seen on a group basis did not always hold for every individual. Although serum NfL levels increased during the event of a relapse and decreased after high-dose corticosteroid therapy in most patients, there were exceptions (e.g., patient 16). More consistently, in our cohort, we observed that serum NfL levels stayed at a constant low level in therapy-responsive patients, which might be helpful in therapy monitoring of patients treated with first-line therapies. Our data suggested that this effect can be seen after 9 months. Whereas a sampling interval of 3 or even 6 months seems appropriate in patients without disease activity, other studies of highly active and more severely affected patients sugges<sup>t</sup> a sampling interval and serum NfL testing every month [26]. We did not observe a positive correlation between serum NfL and EDSS for all time points, which is not surprising as this was also not seen in other studies [31] or only described in larger cohorts and with patients more severely affected by the disease and accordingly with higher EDSS [22,32]. The same was true for the correlation of serum NfL and PASAT as cognitive functions are only mildly affected in the early phase of the disease [33]. Even though, in a previous study, we observed an association of a more active disease course with higher levels of pro-inflammatory cytokines and lower levels of anti-inflammatory cytokines in a subpopulation of our study cohort [29], there was no correlation of serum NfL with any of the observed cytokines in the present study.

We also want to discuss the shortcomings of this study. As this was a retrospective analysis of serum NfL in a prospectively collected cohort, pre-analytical effects on serum NfL outcomes must be considered as samples were stored for more than 10 years. However, the observed values were in the same range as those of comparable patients [30,34,35] and of particular interest as no other therapies were available at this time and thereby we were able to monitor long-term outcomes of serum NfL in this specific study population. We can also not completely rule out spontaneous processes or regression that influences serum NfL (sNfL) levels, as we did not include untreated, stable MS patients. As this was an explorative study, these findings need to be confirmed in independent studies and it is desirable to have more detailed MRI data (e.g., number of gadolinium-enhanced lesions, atrophy, etc.) and complete data sets for every patient in those future studies because, for example, T1-hypointense lesions explain the severity of clinical disability better than T2-hyperintense white matter lesions and gadolinium-enhancing lesions correlate better with active disease status. Missing correlation with EDSS was similar to previous findings [31]. However, we were also unable to detect any correlations with the analyzed cytokines. This might be due to the small sample size or that inflammatory processes were either not present in patients or not displayed in the serum of those patients.

Monitoring of subclinical disease activity using MRI is an established procedure in the care of MS patients. Due to the method's invasiveness, this is not possible for CSF examination, although CSF parameters are appropriate to reflect intrathecal inflammatory processes. Serum NfL appears to be a promising marker for monitoring subclinical disease activity, as demonstrated in this cohort with longitudinal data collection under the same therapy over 24 months. However, this effect may not be seen in every patient as shown in our single-patient illustrations. In a heterogeneous disease like MS, a single biomarker is not sufficient to completely monitor and evaluate therapy efficacy. For this reason, all available information, clinical and paraclinical, should be gathered and taken into account for clinical decision making.

In summary, our study presents the first results on the effect of first-line therapies on serum NfL levels in mildly affected MS patients over 24 months. Here, serum NfL seems especially helpful in detecting therapy-responsive patients, but we also want to address the need for identifying factors that might influence serum NfL values. Among others, this includes processes involved in the transport of NfL from the CSF into serum as well as NfL clearance. The more we know about non-disease-related mechanisms that affect serum NfL, the better we can model serum NfL courses and identify real changes that are caused by pathological processes.

**Author Contributions:** Conceptualization, H.T.; data curation, A.H.; analysis, A.H. and B.M.; funding acquisition, H.T.; investigation, A.H.; methodology, J.K. and M.O.; project administration, A.C.L. and H.T.; resources, M.O.; supervision, A.C.L. and H.T.; validation, M.S. and A.A.; visualization, A.H.; writing—original draft, A.H.; writing—review and editing, M.S., A.A., J.K., A.C.L., M.O. and H.T. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was supported in part by the German Federal Ministry of Education and Research (KKNMS), the German Research Foundation/DFG (SFB1279) and the foundation of the state Baden-Württemberg.

**Acknowledgments:** We kindly thank all patients for their participation in this study and all members of our local biobank, CSF and research lab for their excellent work. We would like to thank Paula Klassen for the linguistic and grammatical revision of the manuscript.

**Conflicts of Interest:** A.H. has nothing to disclose. M.S. has received consulting and/or speaker honoraria from Alexion, Bayer, Biogen, Merck, Roche and Sanofi Genzyme. She has received travel support from Celgene and Teva. She has received research funding from the Hertha-Nathor ff-Program. A.A. received research funding from DMSG and travel grants from Biogen, all not related to the current study. B.M. has nothing to disclose. J.K. has nothing to disclose. A.C.L. has nothing to disclose. M.O. has nothing to disclose. H.T. reports funding for research projects, lectures and travel from Bayer, Biogen, Genzyme, Fresenius, Merck, Mylan, Novartis, Roche, Siemens Health Diagnostics and Teva, and received research support from Hertie-Stiftung, DMSG, BMBF and University of Ulm.
