Regulator of oligodendrocyte maturation, miR-219, a potential biomarker for MS

Lumbar CSF samples were obtained from the Radboud UMC, Nijmegen, the Netherlands (cohort 1), the Erasmus University Medical Center, Rotterdam, the Netherlands (cohort 2), and the VU University Medical Center, Amsterdam, the Netherlands (cohort 3). CSF samples were collected, prepared, and stored in agreement with international consensus guidelines [19]. Table 1 shows the demographic and clinical characteristics of the included participants in all three cohorts. All CSF donors signed written informed consent for use of material and clinical information for research purposes.

MS and CIS were diagnosed according to state-of-the-art diagnostic criteria at time of CSF collection [20, 21], and further classified as having relapsing remitting (RRMS), secondary progressive (SPMS), or primary progressive (PPMS) according to Lublin and Reingold [22]. Control groups consisted of individuals with other non-inflammatory neurological diseases (NIND), inflammatory neurological diseases other than MS (IND), Alzheimer’s disease (AD), and idiopathic intracranial hypertension (IIH). Neurologically healthy individuals (NHI) had been assessed for a neurological disorder but were diagnosed with either a systemic disease without neurological manifestations, or, for example, (tension-type) headache.

CSF samples (0.5 ml) of cohorts 1 and 2 were spiked with 1 μg MS2 carrier RNA (Roche Applied Science) to increase the yield of RNA isolation [23]. Liquid was removed by freeze-drying on an Alpha 1–2 LDplys freeze dryer (Christ, Osterode am Harz, Germany) and samples were resuspended with 200 μl RNase-free water. RNA was isolated using the miRCURY RNA Isolation kit for biofluids (Exiqon, Vedbaek, Denmark). Equivalent CSF volumes were used as input for the RNA isolation. The expression profiles of miR-24 and miR-16 have been reported to be stable in several bodily fluids and tissues [24‐27], and were therefore used for normalization purposes.

RNA isolation of cohort 3 samples was performed on 300 μl CSF again using the miRCURY RNA isolation kit for biofluids (Exiqon, Vedbaek, Denmark) according to the manufacturer’s protocol and including the on-column DNase treatment. To optimize RNA yield per sample, 2 μg glycogen carrier was added to the lysis solution. Secondly, to monitor RNA isolation and proper reference miRNA normalization, per sample, 150 pmol synthetic UniSP6 RNA spike-in (Exiqon) was added to the lysis solution. Eluted RNA was directly stored at − 80 °C.

The RT reaction in RNA samples of cohorts 1 and 2 was performed as previously described [23] using the Taqman MicroRNA RT Kit (Life Technologies, Landsmeer, the Netherlands) and individual miRNA RT primers in a specific stem-loop conformation. Individual reverse transcription (RT) and quantitative polymerase chain reaction (qPCR) primers for miRNAs (hsa-miR-16, hsa-miR-24, hsa-miR-219) were obtained from Life Technologies. RT products were directly subjected to a pre-amplifaction step using Taqman PreAmp Master Mix (Life Technologies) and individual Taqman miRNA primers for qPCR. The procedure was carried out according to the manufacturer’s protocol for multiplexed pre-amplification. The pre-amplification product was then diluted eight times in RNAse-free 0.1× TE (1 mM Tris-HCl pH 8.0/100 μM EDTA) buffer. First strand cDNA synthesis on the RNA samples of cohort 3 was performed using the Universal cDNA synthesis kit II (Exiqon) according to the manufacturer’s protocol. Per sample 1 μl RNA was used in a 10 μl reaction mixture. An inter-plate control (IPC) was created by pooling RNA of 10 samples and used for cDNA synthesis. To control for potential DNA contamination the IPC reactions were also performed omitting the enzyme mix (−RT). cDNA samples were aliquoted and stored at − 20 °C until PCR.

For samples of cohorts 1 and 2, 5 μl of diluted pre-amplification products of CSF was used according to the manufacturer’s procedure with Taqman Universal Master Mix II (Life Technologies) and individual Taqman miRNA primers for qPCR. Samples were measured in duplicate. Amplification was performed using a qPCR Thermal Cycler (Applied Biosystems, Nieuwerkerk aan den IJssel, The Netherlands) with a denaturation step at 95 °C for 10 min, followed by 50 cycles of 95 °C for 15 s and 60 °C for 60 s. Data of these qPCR experiments were normalized using the geometric mean [26, 28] of the two uniformly expressed miRNAs, i.e., miR-16 and miR-24. Relative expression levels (REL) were calculated using the formula REL = 2 ^− ∆Ct, where Ct is cycle threshold, and ∆Ct = Ct (miRNA) – Ct (geometric mean of miR-16 and miR-24).

Quantitative PCR (qPCR) on the cDNA samples of cohort 3 was performed using the ExiLENT SYBR Green master mix (Exiqon) in combination with primer-mixes for UniSP6, miR-150, miR-219, and miR-204 (Exiqon). cDNA samples were diluted 10× of which 1 μl was added to each 10-μl reaction. qPCRs were done according to the manufacturer’s protocol with the addition of 0.2 μl ROX reference dye (Invitrogen, Waltham, MA, USA) per reaction. All samples were run in triplicate on an ABI7300. On each 96-well plate, the IPC, the IPC-RT, and non-template controls were included in triplicate. Raw Ct (cycle threshold) values were extracted; mean and SDs were calculated for each sample and adjusted using the mean IPC Ct values. Sample values were only included for further analysis when at least two measurements were detectable with less than two Ct difference. Based on the study performed by Bergman and coworkers [29], miR-204 Ct values were used for normalization after confirming high correlation with UniSP6 spike-in Ct levels (Spearman’s rho p < 0.0001, r = 0.7093), qualifying miR-204 as a reference miRNA in CSF. Relative miRNA expression levels were calculated using REL = 2^− ∆Ct, where ∆Ct = Ct target miRNA – Ct miR-204. To identify the causes for samples to be undetectable by qPCR for the miRNAs under investigation, the levels of UniSP6 and miR-204 were used. The minimally required RNA isolation yield was determined by identifying the highest detectable Ct value of UniSP6 providing detectable Ct values for each miRNA separately. The minimally required total miRNA level was determined by identifying the highest detectable Ct value of miR-204 providing detectable Ct values.

Statistical differences were determined using the ANOVA procedure in SPSS 20.0 for Windows Software (SPSS Inc., Chicago, IL, USA). Using cross tabulation, the incidence of MS and/or CIS patients with undetectable miR-219 expression was compared with that in control patients by calculating odds ratios (OR) with 95% confidence intervals (CIs) and its p value (two-tailed Fisher Exact test). OR above 1 denotes a positive association between the non-measurability of miR-219 expression and MS. As data were not normally distributed as tested with the Shapiro-Wilk test, differences between multiple groups or two groups were analyzed using the Kruskall-Wallis test and the Mann-Whitney U test, respectively. Correlations were tested using Spearman’s non-parametric correlation coefficient. For estimation of a biasing effect of age, analysis of variance (ANOVA) on the ranks was applied with age as covariate. A test was considered significant when p < 0.05.

We investigated the potential of miR-219 as a CSF biomarker discriminating MS from NHI in a first exploratory cohort. miR-24 and miR-16, surrogate markers for total miRNA expression, were detectable in all samples. In contrast, miR-219 could not be detected in all samples (Fig. 1a). Mean relative expression of miR-219 in individuals with detectable miR-219 was significantly lower in CSF of SPMS and PPMS patients compared to NHI (Fig. 1b). The odds ratio (OR) for a diagnosis of progressive MS (SPMS and PPMS) versus NHI was 20.8 (p = 0.0294) times higher if CSF miR-219 was undetectable (Table 2).

We next sought to validate these findings in a larger, independent second cohort of MS patients, NHI and NIND, such as AD and IIH and IND. Again, miR-24 and miR-16 were detected in all samples, while the proportion of individuals with undetectable miR-219 was higher in the MS patients’ groups (Fig. 1c). However, mean relative expression of miR-219 in individuals with detectable miR-219 did not significantly differ in MS patients compared to controls in this cohort (Fig. 1d). There was no correlation between CSF miR-219 levels and age, disease duration, CSF erythrocyte count, or CSF leukocyte count. The OR for a diagnosis of MS (RRMS, SPMS, PPMS, CIS) versus controls was 2.6 (p = 0.0262) times higher if CSF miR-219 was undetectable, and the OR for a diagnosis of progressive MS versus non-MS was 3.6 (p = 0.0165) times higher if CSF miR-219 was undetectable (Table 2). The OR for a diagnosis of CIS versus controls (NHI, AD, IIH, IND) was not significant.

In a third cohort consisting of 90 MS patients and 22 controls, by omitting the pre-amplification step, we used a slightly different method to measure miRNA levels. Pre-amplification has a higher chance of resulting in false positive signals, especially in miRNAs with very low expression levels such as miR-219. The omission did result in less measurable total miRNA providing 38 MS and 8 controls with sufficient total miRNA to be included in the analysis of miR-219 expression. No differences in mean expression levels were observed between groups (data not shown), but again, individuals with undetectable miR-219 were more frequent in the MS patients’ groups (Fig. 1e). In this cohort, the OR for a diagnosis of MS (RRMS, SPMS, and PPMS) versus controls (IND and NIND) was 39.7 (p = 0.0002) times higher if CSF miR-219 was undetectable (Table 2). Significance was also observed when selecting specific MS groups (RRMS, SPMS, or PPMS) versus controls. miR-219 absence of detection was not correlated with gender or EDSS.

In addition to miR-219, we measured miR-150 levels in this cohort, as two groups independently showed discriminative power to differentiate MS patients from controls using this miRNA [29, 30]. This miRNA was thus used to validate the reliability of the miR-219 measurements in our study. miR-150 was detectable in 55 of the 112 CSF samples, while the frequency of undetectable miR-150 was similar between groups in terms of insufficient RNA yield, insufficient total miRNA level, or insufficient miR-150 level. Pair-wise comparisons showed significant differences in mean levels of miRNA-150 between RRMS and PPMS (p = 0.007), and between progressive-onset (PPMS) and relapse-onset (RRMS + SPMS) disease (p = 0.023), but there were no differences compared to controls (Fig. 1f). MiR-150 level was negatively correlated to age (r = − 0.298, p = 0.027). To estimate the biasing effect of age, analysis of variance (ANOVA) on the ranks was applied with age as covariate. This led to a loss of significance between RRMS and PPMS (p = 0.092), and between progressive-onset (PPMS) and relapse-onset (RRMS + SPMS) disease (p = 0.089). No significant differences were found between miR-150 levels in relapse-onset patients who did or did not use interferon-beta. miR-150 expression levels were not correlated with gender or EDSS.