Identifying miRNAs in multiple sclerosis gray matter lesions that correlate with atrophy measures

Abstract Objective Multiple sclerosis (MS) is an inflammatory, demyelinating and neurodegenerative disease of the central nervous system (CNS). Though MS was initially considered to be a white matter demyelinating disease, myelin loss in cortical gray matter has been reported in all disease stages. We previously identified microRNAs (miRNAs) in white matter lesions (WMLs) that are detected in serum from MS patients. However, miRNA expression profiles in gray matter lesions (GMLs) from progressive MS brains are understudied. Methods We used a combination of global miRNAs and gene expression profiling of GMLs and independent validation using real‐time quantitative polymerase chain reaction (RT‐qPCR), immuno‐in situ hybridization, and immunohistochemistry. Results Compared to matched myelinated gray matter (GM) regions, we identified 82 miRNAs in GMLs, of which 10 were significantly upregulated and 17 were significantly downregulated. Among these 82 miRNAs, 13 were also detected in serum and importantly were associated with brain atrophy in MS patients. The predicted target mRNAs of these miRNAs belonged to pathways associated with axonal guidance, TGF‐β signaling, and FOXO signaling. Further, using state‐of‐the‐art human protein–protein interactome network analysis, we mapped the four key GM atrophy‐associated miRNAs (hsa‐miR‐149*, hsa‐miR‐20a, hsa‐miR‐29c, and hsa‐miR‐25) to their target mRNAs that were also changed in GMLs. Interpretation Our study identifies miRNAs altered in GMLs in progressive MS brains that correlate with atrophy measures. As these miRNAs were also detected in sera of MS patients, these could act as markers of GML demyelination in MS.


Introduction
The chronic inflammatory demyelinating and neurodegenerative disease MS is characterized by oligodendrocyte death, myelin loss, and remyelination failure. 1 Studies of demyelination in MS have mainly focused on myelin loss from white matter (WM) regions of central nervous system (CNS) tissue, which are easily detected by noninvasive magnetic resonance imaging (MRI) techniques. 2 Conversely, many clinical characteristics like cognitive dysfunction and neurological disability have indicated cortical involvement associated with demyelination or secondary neurodegeneration. 3 In addition, early cortical involvement has been found to be associated with more rapid disease progression. 4 Unfortunately, cortical brain regions are relatively difficult to identify with conventional MRI techniques. Hence, there is a pressing need to develop other markers to reliably capture cortical MS pathology, to understand disease pathogenesis, and to determine the potential impact of disease-modifying therapies.
MicroRNAs (miRNAs), which are small noncoding RNA molecules (19-23nt), regulate post-transcriptional gene expression by binding to the 3' untranslated region (UTR) of a target mRNA. 5 miRNA expression levels in different biological samples originating from MS patients have been found to be dysregulated. [6][7][8] We recently reported miRNA expression profiles of demyelinated white matter lesions (WMLs) from progressive MS brains. 9 Of these, 101 miRNAs were differentially expressed (48 upregulated and 53 downregulated, p < 0.05) between normal-appearing WM and WMLs in MS brains. In a separate study, serum miRNAs levels were correlated with brain atrophy and other MRI measurements. 10 We used our results and those from this study to identify a unique set of miRNAs in both serum and brain lesions. Though miRNA expression from cortical demyelinated gray matter lesions (GMLs) have not been widely reported, a recent study by Fritsche et al 11 identified a cohort of dysregulated miRNAs in type III cortical lesions. Of the seven significantly upregulated miRNAs in GMLs, five (miR-330-3p, miR-4286, miR-4488, let-7e-5p, and miR-432-5) had common a target, synaptotagmin7. However, comprehensive analyses of miRNAs in GMLs, including the pathways altered, target genes expressed, and their association with MRI measurements, are lacking.
In this current study, we therefore sought to (a) analyze the expression of miRNAs and mRNAs in GMLs, (b) identify the biological pathways altered in MS brains following cortical demyelination, and (c) identify the miR-NAs in MS GMLs that correlate with atrophy measures and are detected in sera from MS patients. Collectively, our findings identify miRNAs in GMLs in MS brains that correlate with atrophy measures, which opens up the possibility of using these as potential biomarkers of cortical lesions in MS.

Materials and methods
Human subjects and regulatory compliance Patient demographics for this study are listed in Table 1. All brains were collected as part of the tissue procurement program approved by the Cleveland Clinic Institutional Review Board. Brains were removed according to a rapid autopsy protocol and sliced (1 cm thick) using a guided box. Slices were either rapidly frozen for biochemical analysis or long fixed in 4% paraformaldehyde (PFA) for morphological studies. 12 Global miRNA and mRNA profiling and pathway analyses miRNAs expression profiling of GMLs (five MS normal appearing gray matter [NAGM] and five MS demyelinated cortical lesions) was performed similarly as previously reported. 9,13 For mRNA expression profiling, eight NAGM regions (n = 6 MS brains) and eight GMLs (n = 8 MS brains) were selected for microarray gene expression analysis. Details of RNA isolation, labeling, and microarray analysis have been described previously. 14,15 Significantly dysregulated miRNAs/genes have been presented as heatmaps created with the Broad Institute Morpheus online tool (https://software.broadinsti tute.org/morpheus). To identify miRNA-associated molecular pathways, we used DIANA-miRPath v3.0, which utilizes predicted or validated miRNA target interactions derived from DianaTarBase (http://www.microrna. gr/miRPathv3). 16 For this analysis, we applied default parameters (p value threshold-0.05 and MicroT threshold-0.08) and the human gene database to find associated KEGG pathways.
Further, to identify enriched pathways (activated and inhibited) categories associated with differentially expressed genes (DEGs), ingenuity pathway analysis (IPA) was used. We applied g:Profiler, a functional enrichment analysis tool, to classify gene ontology (GO) enriched pathways in GML DEGs. 17 Graphical presentations of KEGG pathways were created with GraphPad Prism8 with p value < 0.05.

Comparison of GML miRNAs to brain atrophy-associated miRNAs
Regev et al 10 found significant clinical associations (protective/pathogenic) between serum miRNA levels and cortical MRI measurements. We compared the reported miRNAs from the Regev et al 10 study with the current study to identify the miRNAs present in both brain tissue and sera of MS patients and correlated with cortical gray matter volume (cGMV). The study reported 43 miRNAs correlating with MRI measures in Cohort 1 and 45 in Cohort 2, with only one miRNA common between the two cohorts-hsa-miR-185-5p. 10 Quantitative RT-PCR Expression levels of miRNAs were determined in a separate validation cohort consisting of seven progressive MS brain samples (Table 1). To validate miRNA expression, small RNAs were isolated from GMLs and surrounding NAGM using a miRNA isolation kit (miRNeasy Mini Kit, #217004) as per manufacturer's instructions (Qiagen Inc., USA). Isolated miRNAs were reverse-transcribed to cDNA with a TaqMan miRNA RT Kit (Applied Biosystems, #4366596) as recommended by the supplier. The expression of selected miRNAs (Table S1) was validated using TaqMan miRNA assays (cat# 4427975). U6 snRNA (assay ID #001973) was used as endogenous controls in the reaction. TaqMan Gene Expression Assays had a polymerase chain reaction (PCR) efficiency of 100% (AE10%). Each sample was run in triplicate. Delta (D) Ct values were used to determine relative expression changes (fold change, 2 ÀDDCT ). All quantitative data were analyzed and are presented as mean AE standard error of the mean (mean AE sem) using Prism8 (GraphPad Software).

In situ hybridization
In situ hybridization (ISH) was performed using a modified in situ protocol with locked nucleic acid-modified oligonucleotide probes (Exiqon, Denmark) as previously described, 9 with a few modifications. Briefly, well-characterized formalin-fixed, paraffin-embedded (FFPE) 7-µMthick sections were de-paraffinized and rehydrated followed by treatment with proteinase K (40 ng) at 37°C/ 30 min and then fixed in 4% PFA. Next, phosphate-buffered saline (PBS) washed sections were incubated in imidazole buffer, followed by incubation in EDC-Imidazole solution for 90 minutes at room temperature. After washing the sections, a DIG-labeled probe was hybridized to each section overnight (54°C). The next morning, sections were washed in 0.1X saline-sodium citrate, followed by endogenous peroxidase activity blocking with 3% H 2 O 2 . Sections were then placed in blocking solution (Roche) for 1 hour and incubated in a-DIG-POD antibody (Roche) and 1°antibody (MBP, Abcam, cat#MAB386) overnight at room temperature. The next morning, sections were washed (PBS/Tris-HCl/Triton buffer) and incubated with fluorescent-tagged tyramide signal amplification (Perkin Elmer, cat# NEL741E001KT) to label the probe. After washing, sections were incubated with alexa-594-tagged secondary antibody (ThermoFisher Scientific) for 1 hour at room temperature. Slides were then washed in 1X PBS, fixed in filtered auto-fluorescent eliminator regent (Millipore, cat#2160), and subjected to a series of 70% ethanol washes (6 times), with a final wash in PBS. Sections were then mounted in prolong gold antifade reagent (Invitrogen, cat#P36930) and micrographed under a fluorescent microscope (Leica DM5500 B).

Immunohistochemistry
Immunohistochemistry was performed on fresh-frozen and paraffin-embedded human brain sections. 12 Freshfrozen human brain sections were fixed in 4% PFA for 10 minutes at room temperature, followed by peroxidase enzyme deactivation. Human MS brain paraffin sections were de-paraffinized and rehydrated followed by antigen

Ingenuity pathway analysis (IPA) of miRNAtarget gene signatures
To correlate miRNA regulation of gene expression in similar GMLs, datasets of miRNAs and mRNAs were queried for inverse correlations of miRNA expression and its predicted/confirmed target gene (upregulated miRNAs to downregulated target genes/downregulated miRNA expression to upregulated target genes) using IPA tools, with core analysis being performed with the default setting.
Searching through miRNA target gene-enriched molecular pathways (DIANA-miRPath v.3), 16 several diverse pathways were found to be targeted by the 82 miRNAs, such as extracellular matrix (ECM)-receptor interaction, axon guidance, adherens junction, and hippocampal signaling pathway (Table S2). In addition, KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway analysis through miRNA target gene-enriched molecular pathways showed lysine degradation, axonal guidance, GABAergic synapse, and morphine addiction as pathways targeted by the significantly upregulated miRNAs. Conversely, downregulated miRNAs targeted diverse pathways like ECM-receptor interaction, fatty acid biosynthesis, and the TGF-b signaling pathway (Fig. 1F,G, Table S3). These findings indicate possible involvement of these pathways in cortical demyelination of progressive MS brains.

Identifying brain atrophy-associated miRNAs in cortical demyelinated lesions of progressive MS brains
Regev et al 10 reported 87 miRNAs having significant relationships (Spearman coefficient correlation p < 0.05) between serum miRNAs and cortical gray matter volume (cGMV) ( Table S8). As there can be significant differences between the miRNAs detected in peripheral circulation versus cortical GMLs, we compared our current data (82 miRNAs) to the study by Regev et al 10 and found 13 miRNAs common between the two studies ( Table 3). Of these 13, 4 miRNAs were found to be significantly (p < 0.05) dysregulated in GMLs of progressive MS brain in our study. Interestingly, these miRNAs have also been found to be present in the peripheral circulation of MS patients, mainly expressed by immune cells. 25 DIANA-miRPath (v.3) analysis of the four atrophy associated miRNAs predicted to regulate genes associated with diverse pathways like ECM-receptor interaction, lysine degradation, and focal adhesion pathways (Table S9). Recent studies have highlighted importance of extra cellular matrix in CNS development and various neurodegenerative diseases including MS. 26,27 We further analyzed the relationships between the four atrophy-associated miRNAs (miR149*, miR-20a, miR-25a, and miR-29c) and significant DEGs in the GMLs in the context of the human protein interactome. Using a human interactome database that contains 351,444 PPIs (among 17,706 proteins) and a functional miRNA-target interactome of 8157 interactions (edges) among 735 miR-NAs and 2766 target genes, we extracted all of the PPIs of the selected miRNA targets and the DEGs. The results indicate extensive PPIs among the miRNA targets (blue nodes) as well as the DEGs (green nodes) of all four miR-NAs (dark blue nodes) (Fig. 3A-D). Interestingly, for hsa-miR-20a (À5p) and hsa-miR-29c (À3p), a few DEGs were direct targets of the respective miRNAs. For example, BCL2, RGS5, and PURA were directly targeted by hsa-miR-20a, whereas WNT4, TET2, COL5A2, and BCL2 were directly targeted by hsa-miR-29c. Indeed, many of these genes (BCL2, TET2, and COL5A2) have been indeed found to be dysregulated in biological samples from MS patients. 14,[28][29][30] We also performed validation using immunohistochemistry to detect the presence of BCL2positive cells within regions of GMLs (Fig. 3E,F). These observations establish that the four miRNAs in GMLs could potentially affect the expression of the DEGs through direct as well as indirect interactions.

Discussion
Recent reports suggest that the presence of cortical demyelination correlates with progressive neurological decline in MS patients. 31,32 As cortical lesions are difficult to detect by conventional MRI techniques, identifying markers to predict cortical involvement in MS cases would be helpful to predict the cortical component of MS. In the present study, we profiled miRNAs expression in GMLs from progressive MS brains and identified 82 miRNAs that belonged to biological pathways like ECMreceptor interaction, axon guidance, and adherens junction signaling. The target mRNAs of the identified miR-NAs showed enrichment of pathways like sphingosine-1phosphate signaling, chondroitin sulfate degradation, and axonal guidance. Finally, we identified miRNAs that are changed in GMLs and correlate with MRI measurements of cGMV. 10 Taken together, this study identified a list of miRNAs and biological pathways that are changed in GMLs from progressive MS brain tissues and correlate with the cGMV.
Several studies have been performed to analyze miRNA expression in different biological samples (serum/plasma,  cerebrospinal fluid, and immune cells) from MS patients using techniques like microarray, RT-qPCR, small RNA sequencing, and ISH. [33][34][35][36] However, there are limited studies that provide miRNA profiling data from brain tissues of MS patients. While the majority of these reports have focused on WMLs, 6,9 a recent study analyzed miRNA expression from MS brain cortical lesions and presented 20 upregulated and 11 downregulated miR-NAs. 11 Our study in comparison to this report had only two miRNAs that were common (miR-1180 and miR-219-2-3p). The differences between these two studies may be attributed to the platform used to profile miRNAs (Nanostring technology TM vs. microarray platform) and the use of only control tissue for comparison with GMLs from MS brains. Interestingly, several of the significantly dysregulated miRNAs (miR-1275, miR-129-5p, miR-26a, miR-23b, miR-16, and miR-338-5p) reported in this study were common with our previous report of miRNAs identified in MS brain WMLs. 9 The similarity of the previously published WML dataset to the GML miRNA profile suggests possible involvement of these miRNAs regulating common pathways in GMLs and WMLs from MS brains. In line with this possibility, miR-338 has been found to play a major role during CNS myelination and myelin repair, 37 while miR-23b has been reported to play a critical role in suppressing leucocyte migration and halting the progression of experimental autoimmune encephalomyelitis (EAE), an animal model of MS. 38 Similarly, miR-129 has been reported to control axonal regeneration via regulating insulin-like growth factor-1 in peripheral nerve injury. 39 Intriguingly, three miRNAs (miR-25, miR-30d, and miR-100) downregulated in GMLs were upregulated in WMLs. 9 Among the pathways targeted by these dysregulated miRNAs, axonal guidance was found to be significantly affected, probably due to loss of regulation of target gene expression (ADAM11, ITGA4, EPHA7, WNT4, and PAPPA). 40 Other pathways, such as ECM receptor interaction, fatty acid biosynthesis, and TGF-b-signaling, have been proposed to be critical targets for myelin regeneration therapy for demyelinating lesions. [41][42][43] One important aspect of our study was to investigate concurrent gene expression from similar lesion tissue types as used for miRNA expression analysis. This analysis helped us to identify the pathways that might be active in GMLs. Among these pathways, death receptor signaling associated molecules like death receptors (DR) and decoy receptors (DcR) have been reported in MS WMLs. 44 Likewise, genes relating to corticotrophin releasing hormone has been found to be associated with decreased risk for MS disease. 45 The results also identified several pathways that may be repressed in GMLs. For example, genes related to protein kinase A (cAMP-dependent protein kinase) signaling were found to be significantly decreased in GMLs. In the nervous system, this pathway has been found to be associated with dopamine signaling. 46 In addition to alteration of the genes and pathways, several miRNAs in GMLs were found to target pathways like fatty acid degradation/metabolism, as well as valine, leucine, and isoleucine degradation. Of particular interest was the identification of fatty acid metabolism as a top hit, which could be a potential therapeutic target for treating MS. 42 Cortical demyelination-associated neuronal and axonal loss are main factors connected with neurological dysfunction reported from MS patients. 47 Previously, miRNA expression was correlated with GM atrophy measures. 10 Based upon positive/negative correlations, the authors 10 grouped these miRNAs as protective or pathogenic. Based upon the expression levels of these miRNAs, we found that miR-149* (À3p) was significantly upregulated (identified as protective) in GMLs and found to be associated with genes involved in T-cell proliferation, inhibition of apoptosis by targeting PD-1, TIM-3, BTLA, Foxp1, IL-2, TNF-a, and IFN-c in breast cancer cells. 48 Similarly, miR-20a, miR-25, and miR-29c (pathogenic miRNAs) were significantly downregulated in GMLs. These three miR-NAs have been found to target genes involved with inflammation and apoptosis activity [49][50][51] and also have been found to be enriched in CNS microglia and tissue resident macrophages. 52 Moreover, the target gene analysis through human PPI interactome analysis identifies direct (BCL2, TET2, RGS5, and COL5A2) gene targets of these miRNAs that could form the basis of future research to investigate the role of these genes in GMLs. Interestingly, BCL2, an anti-apoptotic protein, is expressed by oligodendrocytes and has been found to be increased within demyelinating lesions compared to the peri-plaque WM, with the highest numbers in remyelinating lesions specifically during relapsing-remitting disease course, 30 thus highlighting oligodendrocyte preservation or loss in MS WMLs. Conversely, BCL2 also acts as downstream product of the protective ciliary neurotrophic factor (CNTF) signaling pathway reported to be active in MS cortex. 14 Serum miRNAs expression can also be influenced by the presence of WMLs. Although we did not find significant overlap between our findings and serum miRNAs that correlate with WMLs, 9,10 further studies are needed to confirm if cortical atrophy associated serum miRNAs could also be influenced by presence of WMLs. Nonetheless, identification of four miRNAs in cortical lesions provides preliminary evidence of participation of these miRNAs in affecting atrophy measures in MS cortical GM.
In summary, we report a new set of miRNAs associated with cortical demyelination that are detected in serum of MS patients and are associated with cortical atrophy. The target genes of these miRNAs were found to be associated with neuronal function, fatty acid synthesis, and inflammation. As we also correlated the expression of these miRNAs in serum, our results provide a method to study cortical changes in MS brains through peripheral monitoring. These results also lay the foundation for future studies to target these miRNAs and pathways in order to investigate mechanisms underlying lesion formation and neuroprotection in progressive MS.