Glial and neuronal markers in cerebrospinal fluid predict progression in multiple sclerosis

Study participants and clinical assessments

We analyzed a total of 301 samples matched with clinical data from patients recruited and prospectively followed at the MS Unit, Department of Neurology, Bellvitge University Hospital. All clinical data were entered into the European Database for Multiple Sclerosis (EDMUS). ¹⁷ Data were sequentially updated at follow-up visits, by a trained neurologist. We included patients at the time point when their CIS or MS diagnosis was made. CIS was defined as the first neurological event suggestive of MS. Definite MS was diagnosed according to the Poser ¹⁸ and McDonald ¹⁹ criteria, as appropriate.^18–20

All patients underwent magnetic resonance imaging (MRI) studies. Patients were classified as having clinically-definite RRMS (n = 192) and CIS (n = 109), according to the disease course at the time of lumbar puncture (LP). Relapse was defined as having signs of neurological dysfunction lasting more than 24 hours. ¹⁹ For inclusion into the relapsing group, the first signs of relapse had to have started within 1 month of sampling. Clinical manifestations at MS onset or during relapse were classified as long-tract disorders, brainstem affection, myelitis or optic neuritis, according to neurological signs and MRI studies. The neurological deficits were scored using the Expanded Disability Status Scale (EDSS). ²¹ Disability endpoints were considered as the EDSS score that was stable for at least 1 year. We estimated disease duration from the onset of the first neurological symptoms. At the last clinical follow-up, patients were classified into four groups, according to their disease course: CIS, CIS converters to RRMS, RRMS and RRMS converters to secondary progressive multiple sclerosis (SPMS).

CSF sampling

We obtained all CSF samples from the Bellvitge Biomedical Research Institute (IDIBELL), CSF Biobank MS Unit collection in Bellvitge University Hospital. This biobank contains coded CSF samples from patients having undergone diagnostic lumbar punctures at the Department of Neurology, in the course of routine diagnostic measures. We obtained CSF samples from around the time of the first neurological examination at our MS Unit. CSF was collected in glass tubes and immediately analyzed for their cell count, protein concentration and IgG oligoclonal bands (OCB). For banking, we stored 1-ml aliquot samples in polypropylene tubes at – 80ºC, until analysis. Our samples were collected between 1997 and 2008, and their clinical assessment ranged from May 1997 to November 2011. All the samples analyzed were thawed only once. Our study samples were coded and analyzed by board-certified laboratory technicians, in a blinded fashion.

CSF analyses

We analyzed the concentrations of NFL in CSF using the UmanDiagnostics NFlight^® assay (UmanDiagnostics, Umea, Sweden), as was previously described. ²² The assay is a solid-phase sandwich enzyme-linked immunoassay (ELISA), with no known cross-reactivity with other brain antigens. The detection limit was 31 ng/L. The respective intra- and inter-assay coefficients of variation (CV) were < 6% and < 9%, respectively. GFAP was measured using a previously-described in-house ELISA procedure. ²³ The detection limit of the GFAP assay was 32 ng/L. The respective intra- and inter-assay CV were 4% and 8%. CSF levels of S100B were measured using the Modular system^® and the S-100B reagent kit (Roche Diagnostics, Basel, Switzerland). Its intra- and inter-assay CV were below 3%.

We analyzed t-tau and p-tau with INNOTEST^® (Innogenetics, Gent, Belgium). The respective detection limits were 60 pg/mL and 15.6 pg/mL. Intra- and inter-assay CV were below 10%.

We analyzed YKL-40 with a Quantikine^® ELISA (R&D Systems, Minneapolis, MN, USA). We diluted CSF samples 1:100 in this analysis. The detection limit was 8.15 pg/mL. The respective intra- and inter-assay CV were 4.7% and 6.9%. We analyzed MCP-1, α-sAPP, β-sAPP, Aβ38, Aβ40, and Aβ42 assays with the MSD Multiarray^TM 96-well immunoassay (Meso Scale Discovery, Gaithersburg, MD, USA). The detection limits were 4.2 pg/mL for MCP-1, 120 pg/mL for α-sAPP, 52 pg/mL for β-sAPP, 5 pg/mL for Aβ38, 5 pg/mL for Aβ40, and 4 pg/mL for Aβ42. Intra- and inter-assay CV were below 10%. The samples with levels below the detection limit were assigned the detection limit value.

Statistics

We described continuous variables by their mean and standard deviation (m(SD)), or median and interquartile range (IQR), depending on their distribution, and their categorical variables by numbers and percentages. Frequency distributions were examined using binominal and χ² contingency tests. Differences between groups were analyzed with the Kruskal-Wallis test, followed by pairwise post-hoc comparisons, using the Mann-Whitney U test. A p value ≤ 0.05 was considered significant. The p values of post-hoc comparisons were adjusted using the Bonferroni (Holms) correction.

We evaluated the time to CDMS in the CIS group, and the time to disability endpoints in relapsing-remitting forms of MS groups (CIS and RRMS). We used the median biomarker level as a cut-off to classify patients into two groups: Those with high levels and those with low levels. We categorized by age at first symptoms, age at LP, disease duration and EDSS, using as cut-offs the median of each variable. For univariate analysis, we used the Kaplan-Meier estimator to assess our patients’ survival: Univariate analysis included gender, age at first symptoms, age at LP, disease duration at LP, EDSS at LP, the OCB, storage time and the 12 CSF biomarkers. We considered p values ≤ 0.1 for inclusion in the multivariate analysis. For the multivariate analysis, we used a Cox Hazard regression model to find the independent risk variables for time conversion to CDMS, and the period of time to reach EDSS 3 and EDSS 6. We prepared univariate and multivariate statistical analyses using Statistical Package for the Social Sciences 20.0 package (SPSS, Chicago, IL, USA).

CSF biomarker levels in MS

There were no significant differences in any biomarker levels between the RRMS and CIS groups (Table 2).

CSF biomarker levels and conversion from CIS to CDMS

We performed survival analyses to evaluate the association of CSF biomarker levels and demographic variables, with the time to conversion from CIS to CDMS. Univariate analysis showed that disease duration at LP, NFL levels and β-sAPP levels were associated with the time from CIS to CDMS (Table 4). A Cox hazard regression model showed that only CIS patients with increased CSF levels of NFL were associated with earlier conversion to CDMS (HR (95% CI): 2.69 (1.75 – 4.15); p < 0.0001) (Figure 1).

OPEN IN VIEWER

Table 4.

Univariate and multivariate analysis.

	Univariate			Multivariate
	CIS to CDMS	Time to EDSS 3	Time to EDSS 6	CIS to CDMS	Time to EDSS 3	Time to EDSS 6
Gender (female)	p = 0.112	p = 0.127	p = 0.016	–	–	HR 0.29 (0.09–0.99)
						p < 0.047
Age at first symptom	p = 0.584	p = 0.522	p = 0.833	–	–	–
Age at LP	p = 0.427	p = 0.479	p = 0.743	–	–	–
Disease duration a	p = 0.015	p = 0.158	p = 0.825	–	–	–
Freezer time a	p = 0.0001	p = 0.219	p = 0.916	–	–	–
EDSS at LP ≥2	p = 0.494	p = 0.0001	p = 0.1	–	HR 3.42 (2.03–5.78)	p = 0.265
					p < 0.0001
OCB	p = 0.235	p = 0.593	p = 0.099	–	–	p = 0.113
NFL levels b	p < 0.0001	p = 0.15	p = 0.228	HR 2.69 (1.75–4.15)	–	–
				p < 0.0001
GFAP levels b	p = 0.558	p = 0.0001	p = 0.072	–	HR 1.83 (1.01–3.35)	p = 0.4
					p < 0.048
YKL-40 levels b	p = 0.343	p = 0.0001	p = 0.006	–	HR 2.78 (1.48–5.23)	HR 4.57 (1.01–20.83)
					p < 0.001	p < 0.05
MCP-1 levels b	p = 0.162	p = 0.005	p = 0.025	–	p = 0.135	p = 0.18
t-tau levels b	p = 0.759	p = 0.902	p = 0.317	–	–	–
p-tau levels b	p = 0.196	p = 0.493	p = 0.719	–	–	–
α-sAPP levels b	p = 0.156	p = 0.485	p = 0.249	–	–	–
β-sAPP levels b	p = 0.076	p = 0.901	p = 0.657	p = 0.458	–	–
S100B levels b	p = 0.327	p = 0.915	p = 0.654	–	–	–
Aβ38 levels b	p = 0.962	p = 0.852	p = 0.9	–	–	–
Aβ40 levels b	p = 0.486	p = 0.975	p = 0.591	–	–	–
Aβ42 levels b	p = 0.463	p = 0.988	p = 0.616	–	–	–

For our univariate analysis, we used the Kaplan-Meier estimator to assess survival analyses, p–values are shown. p-values ≤0.1 were considered for inclusion in the multivariate analysis. For multivariate analysis, a Cox Hazard regression model was used to find independent risk variables for time conversion to CDMS, and time to reach EDSS 3 and EDSS 6. We display the HR, 95% CI and p-values data.

a

This variable was not considered for multivariate analysis due to its correlation to the final endpoint (time-dependent bias).

b

Comparison between patients with CSF biomarker levels above the median value and patients with CSF biomarker levels below the median value.

Aβ38: 38 amino acid long fragment of amyloid β; Aβ40: 40 amino acid long fragment of amyloid β; Aβ42: 42 amino acid long fragment of amyloid β; α-sAPP: α-cleaved soluble amyloid-precursor protein; β-sAPP: β-cleaved soluble amyloid-precursor protein; CIS: clinically isolated syndrome; CDMS: clinically definite multiple sclerosis; CI: confidence interval; CSF: cerebrospinal fluid; EDSS: Expanded Disability Status Scale; GFAP: glial fibrillary acidic protein; HR: hazard ratio; LP: lumbar puncture; MCP1: monocyte chemoattractant protein 1; NFL: neurofilament light protein; OCB: oligoclonal bands; p-tau: tau phosphorylated at threonine 181; S100B: S-100B protein; t-tau: total-tau; t-tau: total-tau; YKL-40: human chitinase 3-like 1 protein.

OPEN IN VIEWER

Figure 1.

NFL levels and conversion from CIS to CDMS.

NFL and YKL-40 concentrations were significantly higher in the CIS patients whom had converted to CDMS in less than 1 year (NFL: 1770 ng/L (1202 – 5152); YKL-40: 138 ng/mL (84 – 174)), than in the non-converters (NFL: 850 ng/L (470 – 1800); YKL-40: 83 ng/mL (59 – 111)) (NFL: p = 0.001 and YKL-40: p = 0.03).

CSF biomarker levels and disability progression

We also investigated CSF biomarker levels in CIS and RRMS patients, in relation to disability progression. Survival analyses were performed to evaluate the association of CSF biomarker levels and demographic variables with the period of time to reach EDSS 3 and EDSS 6. Univariate analysis showed that EDSS at LP > 2.0 and levels of GFAP, YKL-40 and MCP-1 were associated with the time to reach EDSS 3 (Table 4). Independently of basal EDSS > 2.0 (HR (95% CI): 3.42 (2.03 – 5.78); p < 0.0001), the Cox Hazard regression model showed that high levels of YKL-40 (2.78 (1.48 – 5.23); p = 0.001) and GFAP (1.83 (1.01 – 3.35); p = 0.04) were independently associated with earlier progression to EDSS 3 (Figure 2(a) and Figure 2(b)). Univariate analysis showed that gender, OCB, EDSS at LP > 2.0, and levels of GFAP, YKL-40 and MCP-1 were also associated with the time to reach EDSS 6 (Table 4). The Cox Hazard regression model showed that male gender (female gender HR (95% CI): 0.29 (0.09 – 0.99); p = 0.04) and high levels of YKL-40 (4.57 (1.01 – 20.83); p = 0.05) were independent risk factors to reach EDSS 6 (Figure 3).

OPEN IN VIEWER

Figure 2.

YKL-40 and GFAP levels and disability progression in the relapsing forms of MS.

OPEN IN VIEWER

Figure 3.

YKL-40 levels and disability progression in the relapsing forms of MS.

CSF biomarker levels and relapses

We used the RRMS and CIS samples to search for differences between the relapsing and remitting phases of MS. Patients with acute relapses had significantly lower GFAP, MCP-1, t-tau, p-tau, α-sAPP and β-sAPP levels, as compared with patients in remission (Table 2). Patients with acute relapses had significantly higher NFL levels, compared with patients in remission, only when the relapse period was extended to 3 months prior to sampling (relapse: 1455 ng/L (677 – 2670); stable phase/remission: 830 ng/L (520 – 1835); p = 0.009). In contrast, there were no differences in the levels of YKL-40 and S100B, between relapse and remission (Table 2); hence, the difference in CSF biomarker levels between the relapse and remission phase was large for GFAP, MCP-1 and NFL, while small for YKL-40.