**3. Results**

### *3.1. microRNA Expression Profile*

Microarray analysis was able to detect 2827 probesets in at least one sample out of the total 6659 human probesets (42.45%) (Figure 2A). Among detected probesets, 33 showed a differential expression between positive and negative patients for LS-OCMBs, and more importantly, the top 10 probesets that show the highest expression are able to group patients according to their LS-OCMBs status (Figure 2B). Among these 10 probesets, we can find five mature miRNA, two pre-miRNA and two small nucleolar RNAs. For validation purposes, we selected the four mature miRNA for which TaqMan Advanced miRNA Assays were available (hsa-miR-6800-5p, hsa-miR-6821-5p, hsa-miR-4485 and hsa-miR-4741). However, RT-qPCR validation results did not confirm the alteration of these four miRNAs (Figure 2C).

### *3.2. circRNA Expression Profile*

RNA-seq analysis pipeline of circular transcriptome detected 27,630 bona fide circRNAs and 5431 circRNAs (19.6%) with sum of reads ≥ 10 (Figure 3A). Di fferential expression analysis revealed that 124 circRNAs are altered (*p* < 0.05; FC > |2|) between the two groups, 72 of them being upregulated, while 52 are downregulated.

We selected ten candidate circRNAs to confirm their di fferential expression by RT-qPCR in a larger cohort. First, the technical test for primer amplification confirmed a single and specific circRNA amplification for nine out of the 10 circRNA candidates. circMETRNL was the one that did not show a good amplification and a single band in agarose gel, so we discarded it in subsequent validation experiments. Additionally, Sanger sequencing of the PCR product confirmed that the amplicons span the BSJ (Figure S1). As it can be observed in Figure 3C, the lower expression of hsa\_circ\_0000478 and hsa\_circ\_0116639 was confirmed by qPCR in PBMCs from patients with positive LS-OCMBs (FC = −1.5 and FC = −1.65, respectively; *p* < 0.01).

**Figure 2.** Results of miRNA expression profile by microarrays. ( **A**) Heatmap showing the expression (microarray intensity signal) of miRNA that are expressed in at least one sample. (**B**) Heatmap of the expression of the ten DEmiRNAs showing the highest expression. Hierarchical clustering shows that the expression pattern of these ten miRNAs is able to group patients according to their LS-OCMBs status. ( **C**) RT-qPCR validation results of selected candidate miRNAs. P—positive LS-OCMB status. N—negative LS-OCMB status. DE—di fferentially expressed.

**Figure 3.** *Cont.*

**Figure 3.** Results of circRNAs and linear RNAs expression profile by RNA-seq. (**A**) Volcano plot showing the expression difference (log2 FC) between positive and negative group of circRNA that show a sum of reads across all samples higher than 10. DEcircRNA are highlighted in red and labels point at those DEcircRNA with a base mean higher than 5. (**B**) Volcano plot showing the expression difference (log2 FC) of linear RNAs between positive and negative groups. Differentially expressed linear RNAs are highlighted in red and labels point at those DE linear RNAs with an adjusted *p*-value < 0.01. (**C**) RT-qPCR validation results of selected candidate circRNAs. Asterisk indicates a statistically significant difference (*p* < 0.01). circ\_0000478 and circ\_0116639 validation includes more samples than the rest of the circRNAs because of limitations in sample amount (Table S1). (**D**) RT-qPCR validation results of selected candidate linear RNAs. Asterisk indicates a statistically significant difference (*p* < 0.05). P: positive LS-OCMB status. N: negative LS-OCMB status.

### *3.3. Linear Transcripts Expression Profile*

In the present study, we also analyzed the linear transcriptome by RNA-seq, identifying 115,869 transcripts with ≥10 reads in total. We applied an additional filter using a detection criterion, which permitted to identify 84,863 linear RNAs with a consistent expression pattern, from which 92.8% were detected in both groups. Among them, 2441 transcripts were di fferentially expressed (*p* < 0.05 and FC >| 2|), from which 1421 were detected in both groups (Figure 3B). The rest of the transcripts are specific to one of the groups, 382 being expressed only in PBMCs from patients with negative LS-OCMBs status while 638 transcripts are only expressed in patients with positive LS-OCMBs.

We selected ten candidate linear RNAs to confirm their di fferential expression by RT-qPCR in a larger cohort. As it can be observed in Figure 3D, the lower expression of IRF5 and MTRNR2L8 was confirmed in PBMCs from patients with positive LS-OCMBs (FC = −1.33 in both transcripts; *p* < 0.05).

Di fferentially expressed linear RNAs are enriched in biological processes regarding mainly the immune system. The most significant and enriched terms (FDR < 0.01 and fold-enrichment > 2) are shown in Figure 4. Complement activation, humoral immune response and type I interferon signaling pathway appear among the top enriched terms.

**Figure 4.** Results from gene overrepresentation test of 2441 DEmRNA. The most significant (fold-enrichment > 2) and enriched (FDR < 0.01) GO biological processes are shown. Bars are colored according to their FDR value. DE—di fferentially expressed. FDR—false discovery rate.

#### *3.4. Evaluation of circRNA and Linear RNAs as Biomarkers of a Highly Active Disease*

In order to assess the potential of the four validated RNAs as blood biomarkers of the LS-OCMB status we performed the ROC curve analysis. As shown in Table 2, di fferent combinations of RNAs were tested to find the best performance in discriminating both groups. We found that di fferent combinations of both circRNAs and linear RNAs improves the performance, reaching AUC values of around 70%.


**Table 2.** ROC analysis results of the four candidate transcripts and different combinations. ROC—receiver operating characteristic.

### **4. Discussion**

In this work, we characterized the whole-transcriptome of PBMCs from MS patients with distinct LS-OCMB status. This profiling and the validation experiments revealed that the global transcriptome of PBMCs from patients with positive LS-OCMBs is different from those patients with negative LS-OCMB status.

RNA-seq results reveal that there are 124 circRNAs and 2441 linear RNAs differentially expressed between the two groups. Interestingly, 58% of linear RNAs are detected only in one of the groups, highlighting that there is a specific expression pattern in PBMCs from patients with different LS-OCMB status. It is true; however, that our RNA-seq sample size is limited and these observations should be taken with care.

To the best of our knowledge, none of the circRNAs have been previously related with MS. Circ\_0000478 is located in chromosome 13 and its host gene is von Willebrand factor A domain containing 8 (*VWA8*). Interestingly, the linear transcript of this gene is also among the differentially expressed linear RNAs in our dataset, having a fold-change of −3.23 (*p* = 0.047). This transcript has been recently described and codes a mitochondrial protein with still uncertain function [32]. Regarding circ\_0116639, it is located in chromosome 22, overlapping the *EP300* gene. EP300 is a histone acetyltransferase, which is detected in our study but is not differentially expressed. These differences in expression patterns of the circRNA with their host genes reveals that the biogenesis of circRNAs is independently regulated from that of their host genes, as it has been previously described [33].

Of note, interferon regulatory factor 5 (*IRF5*) is one of the confirmed downregulated transcripts, and polymorphisms in this gene has been associated to the risk of developing MS in last genome-wide association analysis, performed by the International Multiple Sclerosis Genetics Consortium, as well as in replication studies in Spanish cohorts [34,35]. Additionally, SNPs in this gene have been related to increased levels of CXCL13 in CSF, a chemokine related to highly active disease [36]. On top of that, it has been shown in animal models that IRF5 is related with the microglial polarization towards

a pro-inflammatory state (M1) in response to stroke and in Alzheimer's disease [37,38]. This could sugges<sup>t</sup> microglial implication in a more aggressive disease course, as it has been suggested by other authors in a exome sequencing project [39]. Nonetheless, our findings are made in peripheral PBMCs, so the *IRF5* source might be peripheral monocytes or macrophages. In any case, further research is needed to uncover *IRF5* implication in a highly active MS disease course.

On the other hand, MT-RNR2 like 8 transcript (*MTRNR2L8*), the other validated linear transcript showing lower expression in the LS-OCMB+ group, is coded in chromosome 11, and, interestingly, it spans the location of the miR-4485 stem-loop location, one of the four miRNAs selected for validation in the present study. According to microarray data, miR-4485 is downregulated in LS-OCMB+ groups, as it is *MTRNR2L8*. The fact that *MTRNR2L8* is a host gene for mir-4485 may explain that both transcripts appear downregulated in this group of patients, even if the miRNA could not have been validated. These results point at a possible interaction between these two RNAs that might be related to a more active disease, but further research is needed to confirm this hypothesis. According to UniProt, MTRNR2L8 has a role as a neuroprotective and antiapoptotic factor and it has been found to be increased in a specific area of the brain from patients with major depressive disorder [40]. Although the function of this gene is still unclear, the possible role as a neuroprotective factor is very interesting in the case of MS, since we find this gene downregulated in PBMCs. But again, further studies should be conducted to uncover the possible link between *MTRNR2L8* expression in peripheral blood and disease activity, as well as the mechanisms underlying this relationship.

Gene overrepresentation test shows that biological processes related to complement activation, humoral immune response and type I interferon response are among the most enriched term. Of note, intrathecal synthesis of IgM antibodies has previously been correlated with intrathecal complement activation [41], and its role in demyelination and axonal damage has been clearly demonstrated [42]. Interestingly, di fferent components of the complement system have been related to MS and plasma levels of some of the components have been proposed as MS disease state biomarkers [43]. Of note, these processes are enriched among altered genes in PBMC from patients with a marker of a high activity disease. This analysis may reveal that those patients may have an altered immune response that may explain their higher disease activity, but further and other kind of studies will be needed to explore this hypothesis.

The main aim of this study was to identify in blood a biomarker or a group of them that could di fferentiate between positive and negative LS-OCMB patients, thus could be used as a more accessible marker of highly active disease. ROC curve analysis revealed that these markers in combination have around 70% of accuracy, which is considered, in general, an acceptable performance [44]. Taking into account that these are measured in blood, they may help in the clinics to perform repeated measurements, while taking serial samples of CSF has serious drawbacks. E fforts are being made to define patients with a highly active disease course, given that their e ffective therapeutic window might be narrower than a more benign or less active disease form [45]. In line with this, several biomarkers have been proposed to identify individuals with aggressive disease course, such as CSF levels of neurofilament light chain, CXCL13 or CHI3L1 [46–48]. Other authors have also reported di fferences in non-coding transcriptome between patients with di fferent disease activity. Quintana and colleagues studied the expression in CSF of a panel of miRNAs patients with MS with LS-OCMB characterization and patients with other neurological diseases (ONDs) [49]. They found di fferences in the expression of miR-30a-5p, miR-150, miR-645 and miR-191 between the OND group and the LS-OCMB+ group, but they could not find any di fference between patients with positive and negative LS-OCMBs. Due to di fferences in the profiling platforms between their study and the present one, we are unable to check the expression of our candidate miRNA in data from Quintana and colleagues. Another study reported that serum levels of miR-24-3p correlate with disease progression index, calculated dividing EDSS value by disease duration, but the status of LS-OCMB is not measured [50]. More recently, the expression in blood of long non-coding RNAs has also been described to show the

capacity of discriminating highly active MS patients from others with less active disease course, based on a age-related MS severity score [51].

All these studies highlight the need of biomarkers that are able to identify patients with highly active disease and the effort that the scientific community is doing in this direction. In addition, it is also important to have a consensus on the definition of what is an aggressive disease course, given that each study uses different parameters to measure and classify patients in each of the groups, which makes the comparison between them really challenging [52,53]. An early recognition of a patient in this group could benefit from different therapeutic advice and; therefore, the possibility to improve the long-term outcome of the disease.

Our miRNA profiling study discovered some miRNA differentially expressed between the two groups, although they have not been validated by RT-qPCR experiments. In this regard, other techniques such as droplet digital PCR (ddPCR) may help in reducing variability and discover small differences between groups. Although it has been reported that the sensitivity is not increased with ddPCR in gene expression studies, and that the performance is similar provided that the amount of starting material is enough and there are no contaminants in the reaction [54,55]. Nonetheless, it is a technique that could be considered in future validation projects aiming at definitely discarding or including those miRNAs as biomarkers for a more aggressive MS course.
