To be relevant for the molecular pathogenesis of MS, candidate miRNAs would be expected either to be pivotal players in activation/function of immune responses, or to be expressed in tissues affected by the disease process (e.g., brain), or to map within regions that have been linked/associated with MS. Accordingly, we selected 22 immunity-related miRNAs (Table 1) based on the following criteria:

MiRNA expression profiling was performed on RNA extracted from PBMCs of 10 MS cases and an equal number of age- and sex-matched controls, using locked nucleic acid (LNA)-modified probes in a liquid-phase bead-based array. Although other tissues and cell types, such as MS brain lesions or specific immunologic cell populations, might be more directly connected to MS pathogenic changes, we chose to investigate miRNA expression patterns in PBMCs, which are readily accessible and could potentially represent an optimal clinical sample for biomarker detection.

After normalization, four RNA samples (all control subjects) were excluded from further analyses because of abnormally high/low background bead signals. Moreover, mir-18a and mir-105 did not show appreciable expression signals in all of the PBMC samples, and were not further analyzed. Monitoring for differences in expression levels between cases and controls showed a significant up-regulation of four miRNAs: mir-155, mir-92a, let-7f, and mir-19a (P < 0.05) (Figure 1(A)). The fold change was comprised between 3.3 (top hit: mir-155, P = 0.013) and 1.2 (let-7f, P = 0.0026). No down-regulated miRNAs were observed.

To validate the most striking result obtained from miRNA profiling, the mature mir-155 and its long non-coding RNA precursor (i.e., the mir-155 host gene MIR155HG, also known as BIC, B-cell integration cluster) were quantitated by semi-quantitative reverse transcription-PCR (qRT-PCR). Real-time experiments were performed on RNA extracted from all of the 20 available subjects (including the four RNAs that failed the microbead-based screening).

As expected, both transcripts resulted up-regulated in MS cases, and the difference in expression levels between the two analyzed groups was significant (P = 0.012) or close to significance (P = 0.053) in the case of BIC and mir-155, respectively (Figure 1(B)). However, fold changes were not as marked as those observed in microbead-based experiments, being on average mir-155 1.43-fold and BIC 2.12-fold more expressed in cases compared to controls.

Even though differences in expression levels in cases vs. controls were not highly significant, in our opinion the identified miRNA-signature warrants further investigation. In fact, this represents the first case of a specific miRNA dysregulation in MS that has been replicated in two independent studies (this work and [21]) and in different clinical samples: actually, previous studies showed that mir-155 is strongly up-regulated (about 12 fold) also in active MS lesions compared to normal brain white matter [21]. The aberrant expression of mir-155 in PBMCs of our MS patients could, to some extent, reflect the corresponding alterations in the brain. Interestingly, an up-regulation of mir-155 was reported also in CD4+ cells from the spleen, lymph nodes, and CNS of EAE mice [45], strengthening the hypothesis of mir-155 overexpression as an MS signature, and suggesting a possible correlation with disease severity and CNS infiltration of autoimmune cells. It should be noted that a very recent study by Fenoglio and colleagues [29] did not evidence a significant dysregulation of mir-155 in MS cases compared to controls. However, this difference might be due to the study design, as only RR patients in the relapse phase were analyzed [29], whereas we chose to focus only on the remission phase, to avoid the possible confounding effect of the ongoing inflammation, whose amount is different during the diverse phases of the disease [3]. Therefore, it would be important in the future to evaluate mir-155 expression in larger MS population samples, to assess its reliability as peripheral biomarker for the disease and/or its progression.

Apart from MS, mir-155 was repeatedly reported as up-regulated in other autoimmune disorders, such as RA, SLE, and ulcerative colitis (UC), suggesting shared pathogenic mechanisms with MS. In particular, mir-155 was shown to be over-expressed in: (i) synovial tissue and fluids, as well as synovial fibroblasts and PBMCs of RA patients [46,47]; (ii) inflamed colonic mucosa of UC patients [48]; and (iii) splenic B and T cells, as well as T-reg cells in the standard induced mouse model of SLE (i.e., collagen-induced arthritis mice, CIA) [49,50]. Notably, knock-out mice for mir-155 (mir-155^−/−) not only did not develop CIA, but also resulted highly resistant to EAE [51].

Unquestionably, both adaptive and innate immune responses are highly regulated, and mir-155 represents a key modulator in the development, maturation, maintenance, and function of different immune-competent cells, like Th1, Th2, B, dendritic, and T-reg cells (Figure 2) [43]. In addition, several pieces of evidence indicate that mir-155 act as a positive regulator of autoimmune inflammation in EAE, by favoring the emergence of Th1 and Th17 cells and the consequent production of proinflammatory cytokines [45,52]. Hence, mir-155 might also represent a promising therapeutic target for MS, as suggested by the reduced clinical severity of EAE after anti-mir-155 treatment [45].

As mir-155 levels are significantly altered in MS patients, we next asked if genetic variations within the BIC/mir-155 gene might predispose to MS. Indeed, expression studies are often strengthened by the support of genetic analyses, even those showing genome-wide significant association signals, as in the case of the ILR2A gene [9,14]. Concerning the mir-155 locus, no association data were available so far. Indeed, surfing the dbGaP database, we found four single nucleotide polymorphisms (SNPs), located in close proximity of the BIC/mir-155 gene, and covering a genomic region of 19 kb (Figure 3), which were nominally associated (p < 0.05) with MS in previously performed GWAS. We therefore attempted to replicate these associations in our case-control Italian MS cohort of 360 cases and 662 controls.

Polymorphisms were genotyped by high-resolution melting (HRM) analysis; the mean genotyping success rate was 95.2% and the accuracy was >99%, according to random duplicated genotyping of 5% of samples. Moreover, 10% of samples were analyzed using also a second technique (i.e., Sanger sequencing), a step introduced for genotyping quality control. None of the polymorphic loci showed a significant deviation from Hardy-Weinberg equilibrium (HWE).

All of the four SNPs were analyzed using the PLINK program: allele and genotype frequencies between MS patients and controls were compared, but no statistically significant associations were found (Tables 2 and 3). However, SNP rs2829806 showed a weak trend toward significance, both in the allelic association (P = 0.076) and under a dominant mode of inheritance (P = 0.055; data not shown).

To facilitate the analysis of allelic haplotypes, the level of linkage disequilibrium (LD) among the four SNPs was evaluated using the Haploview program. We observed a unique haplotype block, with the three downstream SNPs showing a lower degree of LD among each other (Figure 3). Hence, haplotype analysis was accomplished using a “sliding window” option, considering three adjacent SNPs across the region at a time (Table 4).

The haplotype GTT, composed of SNPs rs2829803, rs2282471, and rs2829806, was over-represented in MS cases (13.5%) compared to controls (10.3%), thus resulting associated with the disease status (P = 0.035). This haplotype confers a 1.36-fold increased genetic risk of developing MS [95% confidence interval (CI) = 1.05–1.77] (Table 4 and Figure 3), again supporting the role of mir-155 in the pathogenesis of MS. Notably, other two haplotypes, determined by the same polymorphisms and both showing a protective effect, also resulted significantly associated with the disease (GTG, P = 0.016; GCG, P = 0.039). However, these haplotypes are quite rare in the analyzed population (Table 4).

Taken together, these results suggest that the mir-155 locus can be genetically associated with MS. Although the precise mir-155/BIC variants responsible for the association and their functional effect are still unknown, one might speculate that polymorphisms affecting the expression of this miRNA might directly contribute to MS susceptibility.

PBMCs were collected from 10 RR MS patients, all in the remission phase, and 10 healthy subjects. Of the 10 patients (mean age 36.9 years; range 24–50; SD ± 10.1), six were females and four males. The choice to focus on the RR subtype of the disease comes from the notion that in RR-MS patients inflammatory events seem to have a major role, whereas, in PP-MS cases, neurodegeneration is likely prominent [53,54]. Controls were matched with cases in terms of gender and age (mean age 33.9 years; range 25–48; SD ± 7.3).

All patients had not received any immunomodulatory therapy within three months prior to blood withdrawal. Moreover, to avoid possible confounder effects due to diurnal variation in immune function, all samples were collected between 8–11 am.

PBMCs were isolated from heparinized blood by centrifugation on a Lympholyte Cell separation media (Cederlane Laboratories Limited, Hornby, Ontario, Canada) gradient. All blood samples were processed immediately after phlebotomy. Total RNA was isolated using the Eurozol kit (Euroclone, Wetherby, UK).

RNA concentration was determined using the NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA). RNA quality was assessed on an Agilent Bioanalyzer 2100 using the Agilent RNA 6000 Nano Assay kit (Agilent Technologies, Santa Clara, CA, USA).

Expression profiling of 22 candidate miRNAs (Table 1) was carried out by using the liquid-phase Luminex Microbead miRNA Profiling system with a custom FlexmiR Select miRNA Assay (Luminex, Austin, TX, USA). Briefly, 1 μg of total RNA was biotin labeled with the FlexmiR miRNA Labeling kit (Luminex), which labels all RNA molecules, including small RNAs, by first removing the 5′-phosphates from the terminal end of the miRNAs by a Calf Intestinal Phosphatase (CIP) treatment, and then enzymatically attaching a biotin label to the 3′-end of the miRNAs in the total RNA sample. Treated RNA samples were then hybridized to beads containing different fluorophores and coated with LNA-modified capture probes (a different fluorophore and a single probe for each bead) complementary to the selected miRNAs. After washing away the unbound RNA, streptavidin-phycoerythrin (SAPE) reporter molecules were added to the reaction and then the expression of miRNAs was analyzed on a Luminex analyzer (Luminex).

Normalization of samples and calculation of median fluorescence intensity was performed according to the Luminex protocol.

The expression patterns of unfiltered data were determined using supervised hierarchical clustering of samples, using the DNA-Chip Analyzer software (dChip) [55]. To further define miRNAs differentially expressed between groups (patients vs controls), the data were filtered on significance of differences using the t test (P < 0.05).

Stem-loop qRT-PCR for mature mir-155 (TaqMan microRNA Assay ID 002623; Applied Biosystems, Foster City, CA, USA) was performed according to the manufacturer-recommended protocols: 300 ng of total RNA were reverse transcribed in a 20-μL reaction volume, using the ImProm-II Reverse Transcriptase (Promega, Madison, WI, USA); 0.8 μL of the RT reaction were used for subsequent real-time PCRs. Data were normalized to mir-146a (TaqMan microRNA Assay ID 000468) levels, a miRNA that was shown, by our microbead-based profiling experiments, to be readily detectable and not affected by the analyzed conditions.

Real-time qRT-PCRs for the quantitation of the mir-155 host gene were carried out using the FastStart SYBR Green Master Mix (Roche Applied Science, Indianapolis, IN, USA). In this case, random nonamers and the SuperScript-III reverse transcriptase (Invitrogen Life Technologies, Carlsbad, CA, USA) were used to perform first-strand complementary DNA (cDNA) synthesis, starting from 500 ng of total RNA, according to the manufacturer’s instructions. Of a total of 20 μL of RT reaction, 1 μL was used as template for PCR amplifications with gene-specific primers. Expression levels were normalized using the hydroxymethylbilane synthase (HMBS) and beta-actin (ACTB) housekeeping genes. Primer sequences and PCR conditions can be provided on request.

In all cases, real-time qRT-PCR assays were performed in triplicate on a LightCycler 480 (Roche Applied Science), and expression levels were analyzed by the GeNorm software [56]. An unpaired, one-tailed t-test was performed to test for significant differences between the MS cohort and healthy controls.

A total of 360 unrelated patients affected by clinically definite MS according to the revised McDonald’s criteria [57] and classified according to Lublin’s criteria [58] as having RR, SP, and PP courses, entered the study: they were consecutively selected from the patients referring to three MS Centers of Northern Italy. To avoid population stratification due to different genetic background, only patients of Caucasian origin coming from mainland Italy (Northern part of the peninsula) were included in the study.

The female-to-male ratio was approximately 2:1 (236 females, 124 males); the mean age was 46.4 ± 12 years (range 22–79 years); the mean age of onset was 30.4 ± 9.1 years (range 11–61 years); the mean disease duration was 15.6 ± 9.6 years (ranging from 0.5 to 55 years); the mean expanded disability status scale (EDSS) was 2.86 ± 2.43 (ranging from 0.5 to 9). As expected, the disease duration and the EDSS were significantly higher in the SP (83 patients) and PP (21 patients) groups compared to the RR (256 patients) group.

As controls, 662 healthy volunteers were recruited. Among them, 360 were matched for age, gender, and geographic origin with the MS patients; the remaining 302 had the same female/male distribution of the patients, and an age above 52 years (mean age 57.9 years; range 52–79; SD ± 4.5). All control subjects had no sign of or familial history for neurological diseases.

Genomic DNA extraction was performed from peripheral blood using either the salting-out procedure or an automated DNA extractor (Maxwell 16 System; Promega). DNA samples were quantified by using a PicoGreen assay (Invitrogen Life Technologies) and the microplate reader Wallac 1420 VICTOR³ V (Perkin Elmer, Waltham, MA, USA).

SNP genotyping was performed by HRM analysis. For each reaction, 7.5 ng of genomic DNA were amplified with a touch-down thermal profile in a final volume of 10 μL. Reactions were performed in 384-well plates using the LightCycler 480 HRM Master Mix on a LightCycler 480 (Roche Applied Science). Amplicons were analyzed with the Gene Scanning Software (Roche Applied Science).

Primer sequences, as well as the specific PCR conditions for each primer couple are available on request.

SNP case-control analyses on allele and genotype frequency data were performed with χ² statistics (Fisher exact test) by using the PLINK software v.1.07 [59]. All analyzed SNPs had a minimal overall call rate of 95% and were tested for HWE in controls before inclusion in the analyses (P_HWE > 0.05).

The LD structure of the genomic region was determined from our SNP data using the Haploview v4.0 program [60]. Haplotype analysis (sliding-window option) and haplotype phasing were performed using the PLINK software. Only phased haplotypes with posterior probability of 1 were included for determining OR and 95% CI (risk haplotype vs all of the other alleles). All reported p values were not corrected for the number of comparisons.

Power estimates indicated that, if each analyzed polymorphism (disease allele frequency of 15%) were to directly confer a 1.5-fold increase in the relative risk of MS, the case/control cohort used in this research would be of sufficient size to have 92% power to detect a significant association at the 0.05 level.

The URLs for data presented in this work are as follows: