Abstract

   Immune disease-associated non-coding SNPs, which often locate in
   tissue-specific regulatory elements, are emerging as key factors in
   gene regulation. Among these elements, long non-coding RNAs (lncRNAs)
   participate in many cellular processes, and their characteristics make
   these molecules appealing therapeutic targets. In this study, we have
   studied lncRNA LOC339803 in the context of neuronal cells, which is
   located in autoimmunity-associated region 2p15 and recently described
   to have a proinflammatory role in intestinal disorders. Using human
   brain samples and a wide variety of in vitro techniques, we have showed
   a differential function of this lncRNA in neuronal cells. We have
   further demonstrated the role of LOC339803 in maintaining hexokinase 2
   (HK2) levels and thus mitochondrial integrity, partially explaining the
   implication of the lncRNA in multiple sclerosis (MS) pathogenesis. Our
   results show the importance of cell-type-specific studies in the case
   of regulatory lncRNAs. We present LOC339803 as a candidate for further
   studies as a mitochondrial dysfunction marker or possible therapeutic
   target in neurodegeneration.

   Keywords: MT: non-coding RNAs, lncRNA, SNP, cell-specificity, RNA
   modification, mitochondria, metabolism, neuroinflammation, multiple
   sclerosis, brain

Graphical abstract

   graphic file with name fx1.jpg
   [55]Open in a new tab
     __________________________________________________________________

   This study reveals that the autoimmunity-associated lncRNA LOC339803,
   downregulated in the cortex of multiple sclerosis patients, helps
   maintain mitochondrial integrity by regulating HK2 levels in neurons.
   Its link to mitochondrial dysfunction highlights a potential
   therapeutic role in neurodegeneration.

Introduction

   Over the last few decades, Genome-Wide Association Studies (GWAS) and
   Immunochip mapping have identified numerous autoimmune disease (AID)
   loci, providing valuable insights into complex diseases.[56]^1^,[57]^2
   The identification of gene variants implicated in immune disorders has
   aided not only in risk prediction but also in illuminating some
   underlying biological mechanisms, thereby opening venues for novel
   strategies in prevention, diagnosis, and treatment.[58]^3 Nevertheless,
   the use of genetic evidence in drug discovery remains limited,[59]^4
   most probably due to their predominant localization in non-coding
   genomic regions and their individual small effect size, making it
   difficult to understand the underlying biological function or the key
   players associated with these variants.[60]^5^,[61]^6 However, a recent
   study has highlighted the importance of the use of genetic evidence in
   drug discovery and showed that a small genetic effect size is not
   negatively correlated with clinical success rates. Furthermore, this
   report highlights the value of studies based on genetic evidence for
   increased success in drug development.[62]^4

   In the past decade, it has been noted that disease-associated single
   nucleotide polymorphisms (SNPs) are enriched in long non-coding RNAs
   (lncRNAs),[63]^7^,[64]^8 suggesting that they are interesting
   disease-susceptibility candidate genes. LncRNAs are a varied category
   of RNA molecules spanning over 200 nucleotides in length that have low
   or no coding potential and their expression is rather tissue and cell
   type specific in comparison to that of protein-coding genes.[65]^9
   While they have been described to localize in all cell compartments,
   they predominantly reside within the nucleus. Thus far, lncRNAs have
   been linked to virtually every stage of the gene life cycle, including
   transcription, mRNA splicing, RNA decay, and translation. Indeed,
   lncRNAs can respond to external stimuli in a quick and
   cell-type-specific manner, acting both transcriptionally or
   post-transcriptionally.[66]^7^,[67]^8^,[68]^10^,[69]^11 Moreover, a
   lncRNA can participate in the pathogenesis of more than one disease,
   activating specific pathways, which lead to diverse clinical
   phenotypes.[70]^7^,[71]^12^,[72]^13^,[73]^14 Hence, deciphering the
   role of immunity-associated SNPs in lncRNA function may help to
   understand the pathogenesis of these complex disorders.

   AIDs are a clinically heterogeneous group of diseases that affect more
   than 7% of the population. Genetic and environmental factors have been
   identified to play a major role in the development of these disorders.
   Aberrant cytokine responses to environmental triggers produced by
   immune and non-immune cells, such as epithelial cells, keratinocytes,
   or neurons, lead to chronic inflammation and destruction of healthy
   tissues causing a wide range of autoimmune diseases, such as rheumatoid
   arthritis (RA), lupus, multiple sclerosis (MS), and type 1 diabetes,
   among others.[74]^3 Although symptomatic therapies may help in some
   cases, more effective and specific treatments for these diseases are
   still needed.[75]^15^,[76]^16 In this line, studying AID-associated
   SNPs that reside on lncRNAs is a poorly explored but promising approach
   toward disease-specific successful therapies, as these molecules have
   shown to be valuable disease-specific biomarkers as well as potential
   targets for future personalized therapeutic
   approaches.[77]^17^,[78]^18^,[79]^19

   Starting from an inflammatory-associated SNP, we have recently
   performed a functional study describing the implication of LOC339803
   lncRNA in the development of intestinal inflammatory disorders. We
   showed that allele-specific RNA methylation levels affect LOC339803
   function in intestinal cells, facilitating the formation of a chromatin
   repressor complex in the presence of the more methylated risk
   allele.[80]^20 Our study demonstrated that increased levels of this
   lncRNA exacerbate inflammation in the intestine, predisposing to
   develop intestinal inflammatory disorders and emerging as a potential
   therapeutic target. The region in which this lncRNA is located is also
   associated with other immune disorders,[81]^21 such as MS,[82]^22 a
   chronic AID affecting the central nervous system (CNS), including the
   brain and spinal cord.[83]^23 MS pathogenesis is still unclear and
   there is a need of new disease markers and targeted
   therapies.[84]^24^,[85]^25 Given the increasing evidence that cell- and
   tissue-specific expression of lncRNAs provide these molecules the
   capacity to exert critical functions in the brain,[86]^26^,[87]^27 we
   aimed to better understand the implication of LOC339803 in disease
   development as well as to elucidate its possible use as a therapeutic
   target in MS.

   Using human brain samples and an in vitro neuronal cell line model, we
   have characterized the involvement of LOC339803 in mitochondrial
   integrity in neuronal cells. Our results point to a possible
   implication of this lncRNA in MS, emerging as a good candidate for
   future studies finding biomarkers or targeted drugs in MS and stressing
   the importance of tissue-specific studies of lncRNAs.

Results

Inflammation-associated LOC339803 in human cortical samples and MS

   As mentioned previously, the genomic region where LOC339803 is located
   (2q15) is associated with several immune disorders including
   MS[88]^20^,[89]^28 ([90]Figure 1A). The MS-associated SNP rs1177228 is
   located within this region[91]^22 ([92]Figures S1A and S1B), and its
   reference allele A is correlated with the reference allele A of the SNP
   rs11498, associated with intestinal inflammation, and located within
   LOC339803 lncRNA ([93]Figure S1B). Interestingly, while G genotype
   increases the risk in intestinal disorders, it is the protective allele
   for MS ([94]Figure S1C).

Figure 1.

   [95]Figure 1
   [96]Open in a new tab

   Inflammation-associated LOC339803 in human cortical samples and MS

   (A) Association of 2q15 genomic region in multiple sclerosis, Crohn’s
   disease, and rheumatoid arthritis using LocusZoom. LOC339803 is
   highlighted in red. (B) Heatmap graph for LOC339803 expression in
   different tissues according to online data (GTEX and NCBI) and RT-qPCR
   quantification of a commercial RNA pool. (C) scRNAseq data of LOC339803
   in cerebral cortex data according to Allen Brain Atlas. (D) Left,
   representative images and right quantification of LOC339803 by RNAscope
   in control and MS patients (n = 4). Scale bars, 20 μm. Data represent
   the mean and standard error. p values determined by one-way ANOVA test.
   ∗p < 0.05, ∗∗p < 0.01, ns = not significant. GM = gray matter; NAGM =
   normal appearing gray matter; DMGM = demyelinated gray matter.

   We found that LOC339803 is highly expressed in brain tissue according
   to different databases[97]^29^,[98]^30 and RT-qPCR quantification of a
   commercial RNA pool from diverse human tissues ([99]Figure 1B).
   Analysis of single cell RNA sequencing (scRNAseq) data from Allen Brain
   Atlas revealed that the expression of LOC339803 varies across different
   brain regions and it is more widely expressed in neuronal than
   non-neuronal cells from cerebral cortex ([100]Figure 1C). We also
   quantified LOC339803 expression by RT-qPCR in human cortical samples
   and intestinal samples and confirmed the abundant expression of this
   lncRNA in cortical samples compared to intestinal samples
   ([101]Figure S1D). Of note, LOC339803 expression did not vary when sex
   of individuals was considered ([102]Figure S1E) and showed no
   correlation with the age of individuals (Spearman r = 0.2355, p =
   0.1732) in cortical samples ([103]Figure S1F).

   Interestingly, when analyzing demyelinated gray matter from MS
   patients, we observed reduced LOC339803 levels, both by RT-qPCR
   ([104]Figure S1G) and RNAscope ([105]Figure 1D). Considering LOC339803
   is more abundant in neuronal cells than in non-neuronal
   ([106]Figure 1C), our results provide evidence of decreased levels of
   LOC339803 in cortical neurons in MS patients. Therefore, we aimed to
   elucidate the function of LOC339803 in the neuronal cells.

LOC339803 shows cell-type-specific characteristics

   Increased LOC339803 levels result in a m^6A methylation-dependent
   reduction of COMMD1 gene transcription in intestinal cells.[107]^20 We,
   therefore, wanted to study whether LOC339803 has the same function in
   neuronal cells. We first quantified LOC339803 expression in the human
   SH-SY5Y neuroblastoma cell line as well as other intestinal and immune
   cell lines. As observed in human tissues, LOC339803 was highly
   expressed in SH-SY5Y cells compared to the rest of cell lines
   ([108]Figure S2A), suggesting that SH-SY5Y cells are a good model for
   our studies.

   Although COMMD1 expression is lower in neuronal SH-SY5Y than in
   intestinal HCT-15 cells (data not shown) the overexpression or
   silencing of LOC339803 in SH-SY5Y cells did not affect COMMD1
   transcription ([109]Figures S2B and S2C), indicating that LOC339803 has
   a different activity in brain and intestinal tissues. Additionally,
   when analyzing scRNAseq data from Allen Brain Atlas, we could observe
   that unlike LOC339803, COMMD1 is expressed both in neuronal and
   non-neuronal cell types at high levels ([110]Figure S2D), showing no
   apparent correlation with lncRNA expression. Moreover, when m^6A
   methylation status of LOC339803 was determined, differences between
   cell lines were also detected. We performed m^6A RNA
   immunoprecipitation (meRIP) in both cell lines and found high levels of
   methylation of LOC339803 in HCT-15 intestinal cells, mainly in the LOC
   G form ([111]Figure 2A, right). However, m^6A methylation was barely
   detectable in SH-SY5Y neuronal cells ([112]Figure 2A, left), suggesting
   that m^6A methylation may not be crucial for its function in this cell
   type.

Figure 2.

   [113]Figure 2
   [114]Open in a new tab

   LOC339803 shows cell-type-specific characteristics

   (A) m^6A RNA immunoprecipitation (meRIP) and allele-specific LOC339803
   levels (LOC A and LOC G) assessed by RT-qPCR in SH-SY5Y neuronal (n =
   3) and HCT-15 intestinal (n = 4) cells. (B) Venn diagram of proteins
   identified by mass spectrometry in cell-type-specific LOC339803 IVT
   pull down. (C) Subcellular localization of LOC339803 in neuronal
   SH-SY5Y cell line by fluorescence in situ hybridization (FISH). Two
   different probes with Digoxigenin (DIG) tags (red) were used (LOC1
   located in exon1, and LOC3 located in exon2). α-Tubulin (green) was
   used as cytoplasmic control and Hoechst (blue) as nuclear marker. Scale
   bars, 10 μm. (D) Subcellular localization of LOC339803 in SH-SY5Y cell
   line by biochemical fractionation using RPLP0 and ARGI lncRNA as
   cytoplasmic and nuclear controls respectively (n = 4). (E) Subcellular
   localization of LOC339803 using RPLP0 (cytoplasmic) and MALAT1
   (nuclear) as controls upon SRSF3 silencing (n = 3) in SH-SY5Y cells.
   All p values were determined by two-way ANOVA test. Data represent the
   mean and standard error ∗p < 0.05, ns = not significant. Enrichment
   relative to control IgG ##p < 0.01.

   Secondary structure of lncRNAs is usually essential for their mechanism
   of action, so we next addressed whether m^6A methylation and the SNP
   genotype could be affecting the secondary structure of LOC339803, what
   in turn, could be influencing its cell-type-specific function. With
   this aim, we analyzed the predicted secondary structure of the lncRNA
   depending on rs11498 SNP genotype (LOC A or LOC G forms) and m^6A
   motif. We took the advantage of RNA structure ensemble approach because
   structure ensemble can provide more effective information than single
   folded structure.[115]^31 Boltzmann sampling of RNA structures was used
   to sample 1000 structures for both SNP-U (RNA from LOC A form) and
   SNP-C (RNA from LOC G form) LOC339803 sequences containing m^6A
   site.[116]^32 Five major groups and representative structures of all
   the groups were identified after K-means clustering of combined 2000
   structures. Cluster 2, 4, and 5 contain SNP-U structures (RNA from LOC
   A form) as major populations, while 96.4% of structures in cluster 3
   belong to SNP-C ensemble (RNA from LOC G form) ([117]Figure S2E). All
   three SNP-U representative structures show double-strandedness at m^6A
   site and single-strandedness at SNP-U site. In SNP-C representative
   structure, SNP site is folded with m^6A site in single strand region,
   pointing that this form will be more accessible to be methylated, as we
   could observe in HCT-15 cells ([118]Figure 2A). These results suggest
   that both the SNP genotype and m^6A methylation could modulate lncRNA
   structure, and thus, lncRNA function.

   Both subcellular localization and protein binding are key
   characteristics to understand the cellular function of lncRNAs.[119]^33
   To further confirm the differential function of LOC339803 in each cell
   type, we performed an in vitro transcribed (IVT) lncRNA
   immunoprecipitation followed by mass spectrometry in the SH-SY5Y and
   HCT-15 cell lines. Although some binding proteins were common in both
   cell lines, LOC339803 showed cell-type-specific interacting partners
   ([120]Figure 2B and [121]Table S1). We also confirmed that, while it is
   primary nuclear in intestinal cells[122]^20 ([123]Figure S2F),
   LOC339803 is widely present in the cytoplasm of neurons, both in the
   SH-SY5Y cell line ([124]Figures 2C and 2D) and in cortical sections
   from control individuals ([125]Figure S2G), reinforcing that this
   lncRNA functions differently in each cell type.

   We next aimed to study how LOC339803 is exported to the cytoplasm in
   neuronal cells and whether m^6A methylation is involved in this
   process. It has been described that different nuclear export mechanisms
   are involved in cytoplasmic lncRNA export.[126]^34 The YTHDC1 m^6A
   reader, which interacts with LOC339803 and regulates the nuclear
   localization of the lncRNA in intestinal cells,[127]^20 has also been
   linked with the export mechanism of SRSF3-NXF1.[128]^35 LOC339803 is a
   transcript with few and long exons ([129]Figure S1A) and based on
   previous studies[130]^34 NXF1 silencing influenced its nuclear export
   ([131]Figure S2H). We analyzed the protein expression and localization
   of both NXF1 and SRSF3 export proteins in SH-SY5Y and HCT-15 cell lines
   and showed that the expression levels and subcellular localization are
   different in each cell line ([132]Figure S2I). Indeed, while SRSF3 is
   nuclear in HCT-15 cells, it is also present in the cytoplasm of SH-SY5Y
   cells ([133]Figures S2I and 2J). We next wanted to confirm that these
   proteins interact with LOC399803 for its nuclear export in SH-SY5Y
   cells. Although we could not detect NXF1 enrichment in our IVT RIP in
   SH-SY5Y cell fraction, we could observe SRSF3 and LOC339803 interaction
   by mass spectrometry ([134]Table S1) which was further confirmed by RNA
   immunoprecipitation ([135]Figure S2K). In addition, silencing of SRSF3
   protein resulted in increased nuclear retention of the lncRNA,
   indicating that the export of LOC339803 to the neuron cytoplasm is
   mediated by an NXF1/SRSF3 export complex, being the interaction between
   LOC339803 and SRSF3 key for this export ([136]Figure 2E). In accordance
   with the low m^6A methylation levels ([137]Figure 2A) and low YTHDC1
   protein expression ([138]Figure S2L) in SH-SY5Y cells; LOC339803 was
   shown to contain an m^6A independent cytoplasmic localization in
   neuronal cells, as despite interacting with YTHDC1 ([139]Figure S2M),
   silencing of the m^6A reader YTHDC1 did not influence its cytoplasmic
   localization ([140]Figure S2N).

   Overall, these results suggested that LOC339803 has a tissue-specific
   function and while it regulates transcription of COMMD1 in the nucleus
   of intestinal cells by an m^6A-mediated mechanism, it is exported to
   the cytoplasm by an m^6A independent mechanism in neuronal cells, where
   it likely participates in other cellular processes.

Cytoplasmic LOC339803 interacts with metabolic proteins

   To further study LOC339803 cellular function in neuronal cells, we
   performed an enrichment analysis of the cytoplasmic proteins bound to
   LOC339803 in SH-SY5Y cells using ShinyGO. We observed an enrichment of
   proteins related to ribosome and metabolic pathways ([141]Figures 3A
   and [142]S3A). We could not confirm interaction between LOC339803 and
   ribosomal protein RLP10 by RIP ([143]Figure S3B). Moreover, an in
   silico analysis of coding capacity using Coding Potential
   Calculator[144]^36 points to a weak coding capacity of LOC339803
   (score: 0.78). Among metabolic proteins, hexokinase 2 (HK2) was of
   particular interest as it participates in the first step of glycolysis
   but has also been related to mitochondrial membrane protection,[145]^37
   both key processes in neuronal cell metabolism and MS
   development.[146]^38^,[147]^39^,[148]^40^,[149]^41 We confirmed that
   LOC339803 binds HK2 protein using RIP (Sup [150]Figure 3C). Moreover,
   our data showed that this interaction most likely occurs in the
   cytoplasm, as it is where the majority of the lncRNA is located
   ([151]Figure 3B).

Figure 3.

   [152]Figure 3
   [153]Open in a new tab

   Cytoplasmic LOC339803 interacts with metabolic proteins

   (A) Gene ontology enrichment of cytoplasmic proteins that bind
   LOC339803 in SH-SY5Y neuronal cells represented as dotblot chart by
   ShinyGO, number of genes represented as dot size, and fold enrichment
   in colors. (B) Subcellular localization of LOC339803 (LOC) in
   cytoplasmic and mitochondrial fractions from SH-SY5Y cell line. RPLP0
   (P0) and mitochondrial encoded CO2 (mt-CO2) were used as cytoplasmic
   and mitochondrial controls respectively. Data represent the mean and
   standard error (n = 4). (C) HK2 protein-level quantification and
   representative immunoblot by western blot in SH-SY5Y cells transfected
   with control (sictr) or two LOC339803 targeting siRNA (siLOC1 and
   siLOC2), GAPDH was used as loading control. Data represent the mean and
   standard error (n = 4). p values determined by one-way ANOVA test. (D)
   Colocalization between mitochondria (Tom20, red) and cytochrome c
   (green) in SH-SY5Y cells transfected with control (control) or two
   LOC339803 targeting siRNA (siLOC-1 and siLOC-2), nuclei were visualized
   using DAPI (blue). Scale bars, 20 μm. (E) HK2 immunoprecipitation and
   quantification of bound allele-specific LOC339803 (LOC A and LOC G)
   levels assessed by RT-qPCR in SH-SY5Y neuronal cells. Right,
   representative immunoblot of the RIP experiment with GAPDH as negative
   control for the IP. Data represent the mean and standard error (n = 3).
   p values were determined by two-way ANOVA test. (F) HK2 protein level
   quantification and representative immunoblot by western blot in SH-SY5Y
   cells with a mutation of the SNP region in LOC339803 using CRISPR-Cas9,
   GAPDH was used as loading control. Data represent the mean and standard
   error (n = 4). p value determined by unpaired Student’s t
   test. +p < 0.1, ∗∗p < 0.01, ∗∗∗∗p < 0.0001, ns = not significant.
   Enrichment relative to control IgG #p < 0.01, ###p < 0.001.

   To better understand the biological implication of the interaction
   between LOC339803 and HK2 in neuronal function and MS pathogenesis, we
   silenced the lncRNA using two different siRNAs ([154]Figure S2C)
   mimicking what we observed in DMGM human samples ([155]Figure 1D).
   LOC339803-silenced cells had reduced HK2 protein levels
   ([156]Figure 3C) together with an increase of cytochrome c (Cyt C)
   release ([157]Figure 3D), indicating that its expression is critical to
   avoid mitochondrial damage.

   We then tested the influence of LOC339803-associated SNP in the
   predisposition to MS development. No differences in the subcellular
   localization of LOC339803 were detected when comparing the risk (LOC A)
   and protective (LOC G) SNP forms ([158]Figure S3D). Similarly, the
   stability of the lncRNA was not affected by the SNP genotype of
   LOC339803 (Sup [159]Figure 3E). However, the interaction with HK2 was
   enhanced by the presence of the protective LOC G form ([160]Figure 3E),
   indicating the interaction is allele specific. We, therefore,
   hypothesized that the interaction between the RNA and HK2 occurs within
   this region. To test this possibility, we performed a CRISPR-Cas9
   deletion of the SNP region that disturbed the predicted secondary
   structure of the RNA ([161]Figure S3F). In these mutant cells, we
   observed significant reduction of HK2 protein levels ([162]Figure 3F)
   despite levels of LOC339803 were unchanged ([163]Figure S3G), showing
   that the structure of LOC339803 may be important to maintain proper HK2
   protein levels.

   Altogether, these results support that LOC339803 levels, and their
   secondary structures may be key to maintain HK2 protein amounts and
   suggest their involvement in mitochondrial protection.

LOC339803 targeting to reverse altered mitochondrial function in MS

   Our results show that reduced levels of LOC339803 lead to lower HK2
   protein levels and increased Cyt C release in neuronal cells (sign of
   mitochondrial membrane disruption) ([164]Figure 3). These results
   suggest a possible implication of LOC339803 in mitochondrial damage and
   altered metabolism characteristic of MS.[165]^39^,[166]^40^,[167]^41 To
   confirm this, we generated stable LOC339803 knockdown cells using short
   hairpin RNAs (Sup [168]Figure 4A). As before ([169]Figure 3D), these
   cells had augmented Cyt C protein levels ([170]Figures 4A and
   [171]S4B). Moreover, higher CXCL10 expression together with increased
   mitochondrial encoded RNA levels were detected in LOC339803 silenced
   cells ([172]Figures 4A and [173]S4C), pointing to a possible activation
   of the STING pathway and alterations of the mitochondrial activity when
   LOC339803 is silenced. The number of mitochondria were not
   significantly altered in these cells (Sup [174]Figure 1D);
   nevertheless, we observed increased mitochondria roundness in
   shLOC339803 cells ([175]Figure 4B), a sign of mitochondrial fission.
   Moreover, LOC339803 mutant cells ([176]Figures 3F, [177]S3F, and S3G)
   also showed increased CXCL10 proinflammatory cytokine and mitochondrial
   encoded ND1 (mt-ND1) expression ([178]Figure S4E), again suggesting
   that the structure of LOC339803 seems important to maintain
   mitochondrial integrity.

Figure 4.

   [179]Figure 4
   [180]Open in a new tab

   LOC339803 targeting to reverse altered mitochondrial function in MS

   (A) Left, quantification of cytochrome c protein levels by western blot
   in control plK0 and LOC339803 silenced cells using short hairpin RNA
   silencing (shLOC) in SH-SY5Y neuronal cells, HSP90 was used as loading
   control. Right, relative RNA quantification of CXCL10 and mt-ND1 by
   RT-qPCR in plK0 and shLOC. RPLP0 was used as endogenous control. Data
   represent the mean and standard error (n = 3). (B) Mitochondrial
   morphology in control cells (plK0) and LOC339803 silenced cells (shLOC)
   using mitotracker and concofal microscopy, nuclei were visualized using
   Hoechst. Scale bars, 20 μm. (C) Seahorse assay results for OCRs of plK0
   control and shLOC cells. Data represent the mean and standard error
   (n = 5). All p values were determined by two-way ANOVA test. ∗p < 0.05,
   ∗∗∗∗p < 0.0001. (D) Gene ontology enrichment of differentially
   expressed genes from RNAseq data of relapsing-remitting multiple
   sclerosis (RRMS), control samples represented as dotblot chart by
   ShinyGO, and number of genes represented as dot size and fold
   enrichment in colors. (E) Volcano plot of RNAseq data of RRMS and
   control samples, and genes involved in metabolic pathways are labeled.
   (F) LOC339803 expression level upon different drug treatments according
   to DToxS dataset. (G) Gene signature of RRMS plotted against L1000
   firework display data. Yellow dots correspond to cyclosporine
   signatures.

   As HK2 reduction could be influencing both glycolysis and mitochondrial
   membrane integrity, we performed Seahorse assays to quantify both
   glycolysis and mitochondrial respiration. No significant changes were
   detected between plK0 and shLOC cells in their glycolytic capacity
   ([181]Figure S4F). However, when mitochondrial respiration was
   assessed, increased non-mitochondrial oxygen consumption rate (OCR),
   maximal OCR, and proton leak were detected ([182]Figures 4C and
   [183]S4G), which indicates altered mitochondrial activity and
   mitochondrial membrane disruption.

   These results show that reduced LOC339803 levels not only affect HK2
   protein levels but can also increase mitochondrial membrane disruption.
   Accumulation of unhealthy mitochondria in cells leads to increased
   superoxide formation and mitochondrial DNA damage, which would increase
   risk to MS development. Indeed, pathway enrichment of differentially
   expressed genes from online available RNAseq data ([184]GSE16461) of
   relapsing remitting MS (RRMS) patients and control individuals showed a
   high number of genes participating in metabolic pathways
   ([185]Figure 4D), being most of them upregulated ([186]Figure 4E). In
   accordance with our results in MS samples ([187]Figure 1), LOC339803
   expression is also downregulated in RRMS patients compared to controls.
   Hence, we next overexpressed both LOC3398903 forms in SH-SY5Y cells to
   assess if this could reverse disease risk signature. An increase of HK2
   protein levels, more significantly with the LOC G form
   ([188]Figure S4H), which binds HK2 more efficiently ([189]Figure 3E),
   was observed. Moreover, Cyt C protein levels were also reduced in these
   cells ([190]Figure S4I). While no specific LOC339803 targeting small
   molecules have been described, LOC339803 signature in different cell
   lines was analyzed in the DToxS dataset which includes data of FDA
   approved drugs using iLINCS. The average signature of LOC339803 in
   response to each drug was plotted, Olmertasan being the most effective
   drug on increasing LOC339803 expression ([191]Figure 4F). Then, the
   gene signature of RRMS was plotted against L1000 firework display data
   to identify putative drugs capable to reverse MS signature.
   Cyclosporine, which we found could induce LOC33903 expression, showed a
   significant (p = 3,10-e5) opposite signature to that observed in the MS
   patients ([192]Figure 4G), suggesting it could reverse the disease
   phenotype. Additional in vitro experiments confirmed that cyclosporine
   induces LOC339803 expression in the SH-SY5Y cell line
   ([193]Figure S4J). What is more, in vitro experiments in other cell
   types showing different degrees of LOC338903 increase after
   cyclosporine treatment ([194]Figure S4K) support the ability of this
   drug to induce LOC339803 expression.

   Collectively, we confirmed that reduced LOC339803 levels not only
   affect HK2 protein levels but also lead to mitochondrial damage as
   represented by higher mitostress values, increased mitochondrial
   electron transport chain gene expression (as mt-ND1) and mitochondrial
   roundness. As a result, these cells express more Cyt C and CXCL10,
   which could enhance caspase activation and apoptosis in neuronal cells.
   Interestingly, cyclosporine, capable of increasing LOC339803 levels,
   seems to reverse MS signature, pointing to its possible therapeutic
   use.

Discussion

   This study of lncRNA LOC339803 in neuronal cells highlights the
   intricate relationship between RNA molecules, interaction with
   proteins, and cell type specificity. We found that the function of
   LOC339803 varies drastically between neuronal and intestinal cells,
   indicating a context-dependent role of this lncRNA. While it regulates
   transcription in intestinal cells, it binds to HK2 protein in the
   cytoplasm in neuronal cells, modulating its levels and affecting
   mitochondrial integrity. These results highlight the importance of
   studying the cell-type-specific lncRNA function using appropriate cell
   models.

   We have observed differences between neuronal and intestinal cells in
   terms of protein expression and m^6A methylation. Differences in each
   cell type environment could in turn influence the lncRNA function. For
   instance, m^6A methylation levels were described to influence the
   secondary structure of RNA molecules, which is crucial for the ability
   of lncRNAs to bind their target proteins.[195]^42 We have seen that
   methylation levels in LOC339803 are barely detectable in SH-SY5Y cells,
   while they are abundant in intestinal cells. Therefore, it is plausible
   to think that the differential methylation levels in each cell type
   could influence the secondary structure of LOC339803 and thus affect
   the cell-type-specific protein binding affinity. Indeed, LOC339803
   binds cell-type-specific proteins, leading to its cytoplasmic
   localization in neuronal cells while it is located in the nucleus in
   intestinal cells. Moreover, we could also confirm that the genotype of
   the SNP could alter the secondary structure of the lncRNA
   ([196]Figure S2D) what may affect the allele-specific binding of
   LOC339803 to HK2 ([197]Figure 3E). All these results highlight the
   importance of studying context-specific secondary structures of lncRNAs
   as it could be key to understand their mechanism of
   action.[198]^42^,[199]^43

   The interaction between LOC339803 and HK2 unveils significant
   implications for mitochondrial function and cellular homeostasis. Apart
   from its role in glucose phosphorylation, HK2 exhibits the ability to
   bind to the voltage-dependent anion channel (VDAC) situated on the
   mitochondrial outer membrane.[200]^37^,[201]^44 This specific binding
   facilitates the effective use of ATP generated within the mitochondria
   as glycolytic fuel, while also serves to stabilize the mitochondrial
   membrane and to hinder the release of pro-apoptotic factors, including
   Cyt C.[202]^44 Silencing of LOC339803 in neuronal cells resulted in
   reduced HK2 and increased Cyt C release, most probably due to reduced
   binding of HK2 to mitochondrial membrane. Indeed, disruption of the
   interaction between HK2 and VDAC could lead to reduction of ATP
   provision and destabilization of mitochondrial membranes, which could
   in turn enable mt-DNA release to the cytoplasm and activate STING
   pathway too.[203]^45 However, it is still unclear what is the mechanism
   by which LOC339803 regulates HK2 protein levels.

   Mitochondrial dysfunction emerges as a pivotal factor in the
   pathogenesis of MS, providing new insights into understanding and
   potentially treating this complex neurological
   disorder.[204]^39^,[205]^41^,[206]^46 Neurons are particularly affected
   by mitochondrial dysfunction, where in pathological conditions
   dysregulated mitochondrial dynamics increase negative signals for the
   cell.[207]^47 We have observed increased expression of mitochondrial
   electron transport chain mt-ND1 as well as increased mitochondrial
   respiration in LOC339803 silenced cells. Mitochondrial dysfunction has
   been proposed to be prior to neurodegeneration in an experimental
   autoimmune encephalitis mouse model[208]^48; hence, mitochondrial
   alterations caused by downregulation of LOC339803 could be enhancing
   neuroinflammation that could later lead to neurodegeneration. Moreover,
   as a characteristic of mitochondrial membrane disruption due to the
   reduced binding of HK2 to mitochondria, we also observed increased
   proton leak levels in LOC339803 silenced cells. Altogether, these cells
   presented altered mitochondrial activity, leading to Cyt C release,
   which can lead to mitochondrial-dependent apoptosis causing neuronal
   death.[209]^47^,[210]^49 Interestingly, CXCL10 proinflammatory cytokine
   induction was also observed in LOC339803 silenced cells. Elevated
   CXCL10 has been correlated with more aggressive disease course[211]^50
   while neutralization of this chemokine showed improved neurological
   function in an animal model.[212]^51 These findings underscore the
   importance of investigating LOC339803 involvement in mitochondrial
   dysfunction in MS and related neurodegenerative diseases. Indeed,
   overexpression of both LOC3398903 forms in SH-SY5Y cells resulted in an
   increase of HK2 protein amounts, and reduced Cyt C protein levels,
   suggesting that LOC339803 could be an interesting therapeutic target.
   What is more, we were able to identify a variety of drugs capable to
   induce LOC339803 expression levels. In the case of cyclosporine, apart
   from inducing LOC339803 expression in SH-SHY5 cells and other cell
   types, MS signature was found to be reversed, pointing that targeting
   LOC339803 regulated pathways could be a good approach for MS therapy.

   In summary, the tissue-specific function of LOC339803, coupled with its
   involvement in mitochondrial dysfunction and neuronal damage, positions
   it as a promising biomarker and therapeutic target for complex
   neurological disorders like MS. Moreover, the differential
   cell-type-specific m^6A methylation of LOC339803 emphasizes the
   importance of epigenetic regulation in shaping cellular function.
   Understanding the role of LOC339803 in modulating mitochondrial
   integrity could provide valuable insights into disease mechanisms.
   Hence, future research aimed at elucidating the precise mechanisms
   underlying LOC339803 function and exploring its therapeutic potential
   holds promise for advancing precision medicine approaches in
   neurological disorders.

Materials and methods

Experimental model and study participant details

Study participants

   For human brain studies, postmortem frontal cortex samples from
   individuals without neuropsychiatric disorders were obtained at autopsy
   performed in the Basque Institute of Legal Medicine. Samples were
   obtained according to Spanish regulations for postmortem studies and
   were stored in the brain repository of the UPV/EHU (Spain) until assay.
   Myelinated and demyelinated gray matter samples from progressive MS
   patients were obtained from Cleveland Clinic, Lerner Research Institute
   (USA), and from the UK Multiple Sclerosis Tissue Bank at Imperial
   College, London via prospective donor scheme following ethical approval
   by the National Research Ethics Committee in the UK (08/MRE09/31). The
   use of human samples was approved by the Institutional Review Board of
   the UPV/EHU (CEIC, E16/46).

Cell lines and treatments

   Neuronal SH-SY5Y cell line was purchased from ATCC (#CRL-2266) and
   cultured in DMEM (Lonza, Basel, Switzerland, #12-604F) supplemented
   with 10% FBS (Millipore, Burlington, MA, USA, #S0115), 100 units/mL
   penicillin and 100 μg/mL streptomycin (Lonza, #17-602E). Intestinal
   HCT-15 (#91030712) cell line was purchased from Sigma-Aldrich (Poole,
   UK) and cultured in RPMI (Lonza, #12-115F) with 10% FBS, 100 units/mL
   penicillin and 100 μg/mL streptomycin. U937 (#CRL-3253), THP-1
   (#TIB-202) and JURKAT (#TIB-152) cell lines were purchased from ATCC
   and cultured as suggested by manufacturer.

   SH-SY5Y cells were treated with actinomycin D (Sigma-Aldrich, #A9415)
   at a final concentration of 5 μg/mL for 2 h, 4 h, and 8 h for lncRNA
   stability assay.

   SH-SY5Y cells were treated with cyclosporine (Sigma-Aldrich, #30024) at
   final concentration of 1 μM for 2 h, 6 h, and 24 h.

Method details

Gene expression analyses

   We used 500–1000 ng of RNA for the retrotranscription reaction using
   iScript cDNA Synthesis Kit (BioRad, CA, USA, #1708890). Expression
   values were determined by RT-qPCR using Sybr Green (iTaq SYBR Green
   Supermix, Bio-Rad, #1725124) and specific primers. RPLP0 gene was used
   as endogenous control in both human samples and cell lines. Reactions
   were run in a BioRad CFX384 and melting curves were analyzed to ensure
   the amplification of a single product. All qPCR measurements were
   performed in duplicates and expression levels were analyzed using the
   2–ΔΔCt method and normalization to the highest value was used for
   relative RNA expression calculation. For LOC339803 expression analysis,
   a TaqMan assay with a specific fluorophore for each allele was used
   when evaluating both alleles at the same time (this could be done
   because both used cell lines, SH-SY5Y and HCT-15, are heterozygous for
   rs11498 SNP). For the rest of the experiments, specific primers for the
   lncRNA and SYBR Green approach were used. All primers are listed in
   [213]Table S2.

RNAscope

   For single-molecule fluorescence in situ hybridization (smFISH)
   validation, we used frozen human cryosections. Staining was performed
   on representative selection of samples using ACD RNAscope 2.5 HD Red
   assay. Slides were directly transferred from −80°C into 4%
   paraformaldehide for fixation: first 15 min at 4°C, and then 2 h at
   room temperature (RT). After fixation, slides were incubated with H2O2
   for 10 min and then boiled in target retrieval solution (ACD) for
   5 min. Following a washing step in dH2O slides were dehydrated in 100%
   EtOH before protease treatment (Protease IV, ACD, 30 min at RT). The
   target probe (C1) was then incubated on slides for 2 h at 40°C in a
   specific RNAScope hybridization oven (ACD). The following human
   RNAScope assay probe was used: C2orf74-DT (C1). Slides were then washed
   and kept in 5X saline sodium citrate (SSC) overnight. Next day,
   amplification and probe channel detection steps were performed
   following manufacturer’s recommendation. Slides were then
   counterstained with 30% hematoxylin for 15 sec and then mounted with
   Fluoromount reagent (Thermo Fisher Scientific). Brightfield images were
   taken using a Leica DMi 8 microscope with a Leica DFC7000 GT camera.

FISH

   For SH-SY5Y cell line, cells were seeded in 12 mm coverslips and
   incubated at 37°C. When cells were confluent, the coverslips were
   transferred to a new 24-well plate with cold 1X PBS and carefully
   washed by aspirating. Cells were fixed with 4% paraformaldehyde (PFA)
   and 4% sucrose in 1X PBS for 20 min at 37°C, washed twice with 1X PBS
   and cold 90% methanol and incubated for 5 min at RT. After washing
   twice with 1X PBS, cells were blocked with 10% BSA, 0.5% Triton X-100
   in 1X PBS buffer, and incubated for 1 h at RT. Cells were then washed
   three times with 0.5% Triton X-100 in 1X PBS buffer for 10 min each and
   once with pre-warmed solution 2 (0.1X SSC) for 5 min. 50 μl of the
   denatured DIG-Probe Mix were spotted on top a Parafilm, coverslips were
   transferred over the drop (with cells facing the solution) and
   incubated at 57°C for 60 min in dry oven. Cells were washed with
   pre-warmed solution 1 (2X SSC, 0.1% Tween 20) and three times with
   pre-warmed wash solution 2 for 5 min each. The coverslips were
   incubated with blocking solution (3% BSA, 4xSSC, and 0.1% Tween 20) for
   45 min at 37°C in a humid chamber. The primary antibody in a buffer of
   0.5% BSA, 0.2% Triton X-100 (TX) in 1X PBS was incubated overnight at
   4°C in a humidified chamber. Next day, cells were washed three times
   with 0.5% Triton X-100 in 1X PBS buffer for 10 min each and incubated
   in the secondary antibody in a buffer of 0.5% BSA, 0.2% Triton X-100 in
   1X PBS for 2 h at RT in the dark (in a humidified chamber). Finally,
   cells were washed three times with 0.5% Triton X-100 in PBS buffer and
   three times with 1X PBS for 10 min each. Coverslips were mounted with
   vectashield, images were taken in an Axio Observer-Zeiss microscope and
   images were analyzed using the open-software FIJI tool.

   For HCT-15 cell line, cells were seeded in 12 mm coverslips and
   incubated at 37°C. When cells were confluent, they were washed once
   with 1X PBS at RT and fixed with 4% paraformaldehyde in PBS for 45 min
   at 4°C. Cells were washed twice with 1X PBS at RT for 5 min and
   permeabilized with 0.5% Triton X-100 in 1X PBS for 5 min at RT. Cells
   were washed three times in 1X PBS and once with pre-warmed solution 2
   for 5 min each. 50 μl of the denatured DIG-Probe Mix were spotted on
   top a Parafilm, coverslips were transferred over the drop (with cells
   facing the solution) and incubated at 57°C for 60 min in dry oven.
   Cells were washed with pre-warmed solution 1 (2X SSC, 0.1% Tween 20)
   and three times with pre-warmed wash solution 2 for 5 min each. The
   coverslips were incubated with blocking solution (3% BSA, 4xSSC, and
   0.1% Tween 20) for 45 min at 37°C in a humid chamber. The primary
   antibody in a buffer of 3% BSA in 1X PBS was incubated overnight at 4°C
   in a humidified chamber. Next day, cells were washed three times with
   1X PBS buffer for 5 min each and incubated in the secondary antibody in
   1X PBS for 1 h at RT in the dark (in a humidified chamber). Finally,
   cells were washed three times with 1X PBS for 5 min each. Coverslips
   were mounted with vectashield, images were taken in an Axio
   Observer-Zeiss microscope, and images were analyzed using the
   open-software FIJI tool.

m^6A RNA immunoprecipitation

   We fragmented 4 μg of precleared RNA per sample with RNA fragmentation
   buffer (100 mM Tris, 2 mM MgCl2) for 3 min at 95°C and placed on ice
   immediately after heating. 10% of RNA was kept as input; 1 μg of m6A
   antibody (Abcam, #ab151230) and control antibody (IgG, Santa Cruz
   Biotechnologies, Dallas, USA, #sc-2025) were coupled to agarose A beads
   (GE Healthcare, Chicago, USA) in a rotation wheel for 1 h at 4°C. After
   incubation, beads were washed twice in reaction buffer (150 mM NaCl,
   10 mM Tris-HCl, 0.1% NP-40). RNA was added to the antibody-coupled
   beads and incubated for 3 h at 4°C in a rotating wheel. Subsequently,
   beads were washed 3X in reaction buffer, 3X in low salt buffer (50 mM
   NaCl, 10 mM TrisHCl, and 0.1% NP-40) and 3X in high salt buffer (500 mM
   NaCl, 10 mM TrisHCl, and 0.1% NP-40). After the last wash, beads were
   resuspended in lysis buffer and RNA was extracted using the PureLink
   RNA extraction kit (Invitrogen, Carlsbad, USA, #12183016).

RNA immunoprecipitation assay

   For SRSF3 RIP experiments, SH-SY5Y cells were lysed in RIP buffer
   (150 mM KCl, 25 mM Tris, 0.5 mM DTT, 0.5% NP-40, PI), kept on ice for
   15 min and homogenized using a syringe. Lysates were pre-cleared with
   protein A-agarose beads (GE Healthcare, Chicago, USA) for 1 h in a
   wheel shaker at 4°C. Beads were blocked with 20% BSA and mixed with
   pre-cleared lysates and 2 μg of anti-IgG antibody (negative control;
   Santa Cruz Biotechnologies, #sc-2025) or SRSF3 (Santa Cruz
   Biotechnologies, #sc-13510). After overnight incubation in a wheel
   shaker at 4°C, beads were washed 3X with RIP buffer, 3X with low salt
   buffer (50 mM NaCl, 10 mM Tris-HCl, and 0.1% NP-40) and 3X with high
   salt buffer (500 mM NaCl, 10 mM Tris-HCl, and 0.1% NP-40). After the
   washes, 70% of beads were resuspended in RNA extraction buffer and 30%
   was used for western blot (WB).

   For HK2, RPL10, and YTHDC1 RIP experiments, SHSY5Y cells were lysed in
   RIP buffer (150 mM KCl, 25 mM Tris, 0.5 mM DTT, 0.5% NP-40, PI), kept
   on ice for 15 min and homogenized using a syringe. Lysates were
   pre-cleared with proteinG dynabeads (Thermo Fisher, Waltman, MA, USA)
   for 1 h in a wheel shaker at 4°C. Pre-cleared lysates were incubated
   with 2 μg of anti-IgG antibody or antibody of interest (HK2-
   Proteintech, #22029-1-AP; RLP10- Proteintech, #17013-1-AP; YTHDC1- Cell
   Signaling, #E4I9E) for 2 h at RT. After incubation 20 μL dynabeads were
   added and further incubated for 30 min at RT. The immunoprecipitation
   was washed three times with RIP buffer, three times with low salt
   buffer (50 mM NaCl, 10 mM Tris-HCl, 0.1% NP-40) and three times with
   high salt buffer (500 mM NaCl, 10 mM Tris-HCl, and 0.1% NP-40). After
   the washes, 70% of beads were resuspended in RNA extraction buffer and
   30% was used for WB.

Western blot

   Laemmli buffer (62 mM Tris-HCl, 100 mM dithiothreitol (DTT), 10%
   glycerol, 2% SDS, 0.2 mg/mL bromophenol blue, 5% 2-mercaptoethanol) was
   added to the protein extracts in radioimmunoprecipitation assay buffer
   (RIPA) (50 mM Tris HCl, 150 mM NaCl, 1.0% IGEPAL, 0.5% Sodium
   Deoxycholate, 1.0 mM EDTA, 0.1% SDS, 0.01% sodium azide) and were
   denatured by heat. Proteins were migrated on 10% SDS-PAGE gels.
   Following electrophoresis, proteins were transferred onto
   nitrocellulose membranes using a Transblot-Turbo Transfer System
   (Biorad) and blocked in 5% non-fatty milk diluted in TBST (20 mM Tris,
   150 mM NaCl, and 0.1% Tween 20) at RT for 1 h. The membranes were
   incubated overnight at 4°C with primary antibodies diluted 1:1000 in
   TBST. Immunoreactive bands were revealed using the Clarity Max ECL
   Substrate (BioRad, #1705062) after incubation with a horseradish
   peroxidase-conjugated anti-mouse or anti-rabbit (1:10000 dilution in
   2.5% non-fatty milk) secondary antibody for 1 h at RT. The
   immunoreactive bands were detected using a Bio-Rad Molecular Imager
   ChemiDoc XRS and quantified using the ImageJ software. The following
   antibodies were used for western blotting: YTHDC1 (Abcam, #264375, Cell
   Signaling, #E4I9E), HK2 (Santa Cruz Biotechnologies, #sc-374091), HSP90
   (Cell Signaling; #4874), CYTC (Santa Cruz Biotechnologies, #sc-13156),
   GAPDH (Santa Cruz Biotechnologies, #sc-47724), SRSF3 (Santa Cruz
   Biotechnologies, #sc-13510), and ACTIN (Santa Cruz Biotechnologies,
   #sc47778).

Plasmid construction and overexpression

   LOC339803 was amplified from human cDNAs containing A or G allele for
   rs11498 SNP and cloned into a pCMV6 vector (Origene, #PS100001) using
   KpnI and FseI restriction sites. The primers used for cloning are
   listed in [214]Table S2.

   For LOC339803 overexpression, 200 ng of plasmids per 100.000 cells were
   used. Cells were seeded and transfection was performed with X-TremeGENE
   HP DNA transfection reagent (Sigma-Aldrich, #6366546001) for 48 h.

Silencing experiments

   For LOC339803, SRSF3 or YTHDC1 silencing, 30 nM of 2 different siRNAs
   against LOC339803 (IDT, #hs.Ri.LOC339803.13.1 and
   hs.Ri.LOC339803.13.2), SRSF3 (IDT, #hs.Ri.SRSF3.13.2 and
   hs.Ri.SRSF3.13.3), YTHDC1 (IDT, #hs.Ri.YTHDC1.13.1 and
   hs.Ri.YTHDC1.13.3) or negative control siRNA (IDT #51-01-14-01) were
   transfected into cells using Lipofectamine RNAimax reagent (Invitrogen)
   for 48 h.

   pLKO.1-TRC Cloning vector (Addgene, #10878) was used for the
   construction of plKO.1-shLOC339803.1 plasmid. Following Addgene’s
   protocol shRNAs were designed to knock down human LOC339803. The
   sequences for the oligos are shown in [215]Table S2. For LOC339803
   knockdown in SH-SY5Y cell line, viral particles were produced in
   HEK293FT cells transfected with 1 μg pLKO.1 shRNA plasmid, 750 ng
   psPAX2 packaging plasmid (Addgene, #12260) and 250 ng pMD2.G envelope
   plasmid (Addgene, #12259) using X-TremeGENE HP DNA transfection reagent
   (Sigma-Aldrich, #6366244001) in DMEM without antibiotics and cells were
   incubated o/n at 37°C. Transfection media was replaced with fresh
   complete DMEM and viral particle containing media was harvested after
   24 h and 48 h. Collected media was centrifuged and concentrated 10
   times using Lenti-X concentrator (Takara, #631232). Viral particles
   were stored in aliquots at −80°C until used. SH-SY5Y cells were
   infected with sh-LOC339803 or plKO.1 as negative control following
   Addgene’s protocol.

Cellular fractionation

   For the quantification of RNA amounts in nuclear and cytoplasmic
   compartments, nuclei were isolated using C1 lysis buffer (1.28 M
   sucrose, 40 mM Tris-HCl pH 7.5, 20 mM MgCl2, and 4% Triton X-100). The
   amounts of LOC339803-A/G, MALAT1 (nuclear control) and RPLP0
   (cytoplasmic control) were measured by RT-qPCR and compared to the
   total amount of those RNAs in the whole cell lysate.

   For the quantification of protein amounts in nuclear and cytoplasmic
   compartments, cells were resuspended in NARA buffer (10 mM HEPES pH
   7.9, 10 mM KCl, and 0.1 mM EDTA) with proteinase inhibitors (PI) and
   incubated in ice for 10 min. After adding NP-40 to final concentration
   0.05%, lysates were incubated 5 min in ice and centrifuged at 400 g for
   2 min. The supernatant was the cytosolic fraction. Pellet was washed 3X
   with NARA buffer and resuspended in NARC buffer (20 mM HEPES, 400 mM
   NaCl, and 1 mM EDTA) + PI, shaken at 4°C for 30 min and centrifuged at
   16.000 g for 10 min. The supernatant was the nuclear extract.

   For the quantification of RNA and protein amounts in mitochondrial and
   cytoplasmic compartments, cells were resuspended in MIB buffer (210 mM
   Mannitol, 70 mM sucrose, 10 mM HEPES pH 7.5, and 1 mM EDTA) + PI and
   homogenized in ice using a 25-gauge needle fitted on a syringe. After 3
   centrifugations at 1500 g for 5 min, all supernatants were collected in
   ice for further centrifugation at 10000 g for 10 min. Supernatant was
   cytosolic fraction and pellet the mitochondrial fraction. Mitochondrial
   fraction was resuspended in 50 μL RIPA+PI and 70% of supernatants were
   used for RNA extraction and 30% was used for WB.

RNA immunoprecipitation assay followed with mass spectrometry

   For RIP-MS experiments, sense and antisense LOC339803 were amplified
   from cDNA using a T7 promoter primer. The PCR product was purified and
   used for in vitro transcribing biotinylated RNA using the T7 polymerase
   (Takara) and RNA biotin labeling kit (Roche). 1 mg of purified
   LOC339803 RNA was mixed and incubated with whole cell extracts from
   HCT-15 and SH-SY5Y cell. Streptavidin beads were added to the reaction
   and further incubated. After incubation, beads were washed 5 times.
   Samples were sent for mass spectrometry and subjected to in-solution
   digestion followed by nano LC-MS/MS analysis. The list of bound
   proteins was retrieved based on the fold change (sense/antisense). We
   considered proteins with a fold change higher than 2 as potential
   interactors of the sense RNA. List of RIP-MS results are shown in
   [216]Table S1.

Immunofluorescence

   For immunostaining of siRNA LOC339803, 1,00,000 SH-SY5Y cells were
   cultured in μ-slide 8-well high plates (Ibidi, #80806) and washed with
   PBS 1x after 48 h. Cells were fixed in 4% paraformaldehyde for 10 min
   and blocked in 0.25% Triton and 3% NGS in PBS1X for 1 h at RT. Samples
   were incubated with primary antibodies Cyt C (Santa Cruz, #sc-13156)
   and TOM20 (Abcam, ab186735) (dilution 1:50) overnight at 4°C in a wet
   chamber. After two washes, the cells were incubated with secondary
   antibodies at final concentration 1:500 anti-mouse Alexa Fluor 488
   (Thermo Fisher Scientific, A-10680) and anti-rabbit Alexa Fluor 594
   (Thermo Fisher Scientific, A-11008) respectively for 1 h at RT.
   Finally, sample were covered with DAPI-mounting medium (Ibidi, #50011)
   and images were acquired with a confocal microscope (Zeiss LSM800).

   To study mitochondrial integrity, 100000 SH-SY5Y cells infected with
   sh-LOC339803 or plKO.1 were plated in IBIDI plates (#80826) overnight.
   MitoTracker Orange CM-H2TMRos (Invitrogen, #M7511) was added at 1 mM
   final concentration for 30 min at 37°C and nuclei were visualized using
   Hoechst (Sigma-Aldrich, #B2261) at 1.7 μg/mL. Mitochondria were
   visualized using confocal microscopy (Zeiss LSM880 Airyscan).

Seahorse

   Glycolysis and mitochondrial respiration were measured employing an
   XF24 extracellular flux analyzer (Agilent Technologies), and the
   measurements were normalized to cellular protein amount.

   A mix of 3 different SH-SY5Y cells infected with sh-LOC339803 or plKO.1
   were used. 80,000 cells were plated directly in Seahorse plates and
   incubated in their growth media overnight.

   For both OCR and glycolysis determination, the medium of the cells was
   changed to XF Seahorse medium (Agilent Technologies) supplemented with
   4 mM glutamine, 10 mM pyruvate, and 25 mM glucose and were incubated
   for 1 h at 37°C without CO2.

   Then, Mitostress test was performed according to manufacturer’s
   instructions by taking 3 oxigen consumption rate (OCR) and
   extracellular acidification rate (ECAR) measurements at baseline and
   after sequentially adding oligomycin (1 μM), FCCP (1 μM), and
   antimycin/rotenone (0.5 μM)

   For glycolytic rate assay 3 measurements are taken at baseline and
   after sequentially adding antimycin/rotenone (0.5 μM) and 2-DG (50 mM).

Mutant cells generation using CRISPR-Cas9

   For mutant cell line generation, two sgRNAs close to the associated SNP
   were designed and cloned in px458 GFP and px330 mCherry vectors.
   SH-SY5Y cells were transfected with 250 ng of each plasmid. Mixed
   population of the SH-SY5Y cells presenting the expected deletion was
   used. The sequences for the sgRNAs are shown in [217]Table S1.

LOC339803 gene expression analysis in online databases

GTEX and NCBI

   Bulk tissue gene expression for [218]AC016747.3 in thyroid, brain,
   kidney, spleen, lung, heart, muscle, liver, colon, small intestine, and
   stomach from GTEX database
   ([219]https://www.gtexportal.org/home/gene/AC016747.3) was retrieved,
   and values were normalized to the highest value.

   LOC339803 expression from RNA sequencing of total RNA from 20 human
   tissues and Human Protein Atlas (HPA) RNA-seq normal tissues projects
   was retrieved from NCBI database
   ([220]https://www.ncbi.nlm.nih.gov/gene/?term=LOC339803) and values
   were normalized to the highest value.

Allen brain map

   Gene expression of the Human Brain Atlas
   ([221]https://human.brain-map.org) was used to assess the expression of
   LOC339803 in the different brain structures. Allen Brain Cell Atlas
   ([222]https://knowledge.brain-map.org/abcatlas) was used to visualize
   the expression of LOC339803 in neuronal and non-neuronal cell.

DToxS and L1000FWD

   For the selection of drugs that could induce LOC339803, we used DToxS
   database ([223]https://iyengarlab.org/dtoxs/). We plotted the mean
   expression of LOC339803 in different human cell lines after exposure to
   FDA approved drugs. The drugs that presented induction of the lncRNA
   were checked in L1000FWD ([224]https://maayanlab.cloud/l1000fwd/) to
   evaluate their ability to revert MS related gene signature.

Pathway enrichment and protein association analysis

   First, we performed a cellular component analysis using gene ontology
   resource[225]^52^,[226]^53^,[227]^54 selecting the proteins that were
   enriched only in SH-SY5Y cells according to our IVT-RIP MS results
   ([228]Table S1). Then, only the cytoplasmic proteins were selected to
   perform a pathway enrichment using ShinyGo 0.80 KEGG pathway enrichment
   analysis (FDR cutoff: 0.1).[229]^55

   The STRING database ([230]https://string-db.org/) was used to analyze
   the protein association networks (both physical and functional
   interactions) of the cytoplasmic proteins that interact with LOC339803
   and were enriched according to ShinyGo KEGG pathway enrichment.

   For the analysis of gene ontology in RNAseq data from RRMS patients and
   controls ([231]GSE16461
   ), pathway enrichment was performed using ShinyGo 0.80 KEGG pathway
   enrichment analysis (FDR cutoff: 0.05).

Quantification and statistical analyses

   All the statistical analyses were performed using GraphPad Prism 8
   (GraphPad Software). Significance was calculated using Student’s t
   test, Mann-Whitney test or ANOVA test as specified in figure legends.
   The statistically significance level was set at p < 0.05. p values
   lower than 0.1 are marked with a + sign.

Limitations of the study

   One limitation of this work is the small amount of human brain samples
   studied. Opposite to intestinal biopsies that can be acquired with
   routine endoscopy or colonoscopy techniques, brain samples are
   collected postmortem, reducing the access to the samples. Therefore,
   the use of available databases has been helpful to confirm our results
   in bigger sample sets. Additionally, one of the main drawbacks of our
   work, that also applies to many others studying lncRNAs,[232]^56 is the
   lack of a murine homolog for LOC339803, making it complicated to
   perform preclinical studies in murine disease models.

Data availability

     * •
       Plasmids generated in this study are available upon request.
     * •
       Data and code used in this study are available upon request.

Acknowledgments