Human endogenous retrovirus W in multiple sclerosis: transcriptional activity is associated with decline in oligodendrocyte proportions in the white matter of the brain

The myelin sheath plays a crucial role in facilitating fast axonal signal transmission and preventing signal loss or interference, thus enabling efficient neural communication (Frankenhaeuser and Schneider 1951). Moreover, myelin sheaths actively contribute to supporting axonal energy needs by producing adenosine triphosphate (ATP) (Ravera and Panfoli 2015). Oligodendrocytes, the cells responsible for producing the myelin sheaths in the central nervous system (CNS), play a crucial role in the continuous maintenance of myelin integrity (Aber et al. 2022) but they also actively support axonal function and survival (Nave and Trapp 2008).

Multiple sclerosis (MS) is an autoimmune disease characterized by inflammatory processes that gradually degrade myelin sheaths surrounding CNS axons (Wu and Alvarez 2011). Consequently, the loss of myelin sheath exposes axons, rendering them vulnerable to degeneration (Simkins et al. 2021). While white matter is predominantly affected by demyelination (Trapp et al. 2018), spinal cord involvement is also observed (Marrodan et al. 2020). These demyelinating events occur within focal lesions with variable regional distribution, contributing to the diverse spectrum of symptoms (Ghasemi et al. 2017).

The precise etiology and underlying mechanisms driving the development of MS remain incompletely elucidated. However, a well-established phenomenon in the pathogenesis of MS is the infiltration of peripheral lymphocytes across the blood–brain barrier (BBB) into the CNS. This infiltration sets off a cascade of inflammatory responses within the CNS, leading to demyelination and axonal degeneration (Rodríguez Murúa et al. 2022). Autoreactive lymphocytes targeting CNS autoantigens are thought to be involved in this process (Liu et al. 2022). It has also been proposed that demyelination in MS may involve multiple mechanisms. These mechanisms include T cell–mediated demyelination, antibody–mediated demyelination and oligodendrocyte death (Cudrici et al. 2006; Staugaitis and Trapp 2012). Additionally, CNS-resident cells, such as astrocytes and microglia, play significant roles in MS pathogenesis (Absinta et al. 2021).

The onset of MS is thought to be determined by the interplay between genetic predisposition and environmental factors (Waubant et al. 2019). Among the various environmental triggers, viral infections, such as Epstein-Barr virus (EBV), human herpesvirus 6 (HHV-6), and varicella-zoster virus (VZV) have been associated with the onset of MS due to their potential mechanism involving molecular mimicry between viral antigens and CNS autoantigens (Libbey et al. 2007). Human endogenous retroviruses (HERVs), remnants of ancient retroviral infections integrated to human genome millions of years ago, have also been linked to MS development and progression. The human genome harbors an estimated 100,000 individual HERV fragments that constitute about 8% of our genetic material (Belshaw et al. 2004). The prototypic HERV sequence consists of four main retroviral genes (gag, pro, pol, env) that code for structural proteins, viral enzymes, and envelope glycoproteins (Grandi and Tramontano 2018a). It is worth noting that HERV open reading frames (ORFs), that contain the viral genes, have undergone extensive mutations and deletions over time, rendering them incapable of producing viral particles (Grandi and Tramontano 2018b). However, numerous HERV elements are transcriptionally active, producing RNA, across tissues (She et al. 2022). A small subset of these elements can produce full-length or truncated proteins that may have physiological functions or contribute to disease mechanisms (Grandi and Tramontano 2018a).

Especially the HERV-W subfamily of HERVs has been of particular interest in the context of MS. Transcriptional activity of numerous HERV-W loci in the brain has been reported in both normal conditions and disease states (Schmitt et al. 2013; Elkjaer et al. 2021). Research has also indicated an increased expression of the HERV-W envelope protein (env) within lesions found in individuals with MS (Grandi and Tramontano 2018a). This protein is believed to originate from one or a few specific well-preserved HERV-W loci and it has been shown to have pro-inflammatory effects, contributing to the disruption of the blood–brain barrier and myelin destruction (Kremer et al. 2019). There is also evidence that HERV-W env protein is cytotoxic to oligodendrocytes through the induction of the release of redox reactants (Antony et al. 2004). While 7q21.2 (syncytin-1) and Xq22.3 have been identified as primary candidate loci due to their structural integrity, there is still uncertainty about the origin of the env protein. Human genome harbors at least 13 HERV-W loci that have full-length env, but it has been speculated that HERV-W loci with truncated and mutated env genes might produce shorter proteins that could also have a role in disease (Grandi et al. 2016). While the pathogenicity of the HERV-W in MS is commonly attributed to the env protein, research has also shown that HERV-W RNA can induce neuroinflammation by activating NF-κB and triggering the production of type 1 interferons (Morris et al. 2019; Rangel et al. 2022; Russ and Iordanskiy 2023). Consequently, in the context of MS, a broader investigation including HERV-W loci without protein-coding potential may hold significance.

In this study, we utilized RNA sequencing data to analyze the transcriptional activity of 215 distinct HERV-W loci in the white matter of the brain comparing MS and control samples. In addition, we employed a transcriptional deconvolution technique to estimate the proportions of neuronal, glial, and endothelial cells and analyzed whether there are associations between transcriptional activity of HERV-W loci and cell proportions. This approach could provide valuable insights into pinpointing the HERV-W loci associated with MS pathogenesis, while also shedding light on associated mechanisms.

The proportions of glial cells, neuronal cells and endothelial cells were estimated from the RNA sequencing data obtained from white matter brain samples. For all samples average proportions for oligodendrocytes, astrocytes, excitatory neurons, endothelial cells, microglia, and inhibitory neurons were 0.621, 0.213, 0.130, 0.03, 0.004, and 0.004, respectively. Thus, oligodendrocytes, astrocytes and excitatory neurons were the most numerous cell types, making up almost 97% of all estimated cells and oligodendrocytes were distinctly the most common cell type. To assess the differences in the cell type distributions between the MS and control samples, average cell type proportions were obtained for both groups. Cell type distributions were similar in both groups as shown in Fig. 1. However, the proportions of astrocytes and endothelial cells were significantly higher (p < 0.05) in MS individuals and proportion of oligodendrocytes was significantly lower in MS individuals as shown in Table 1.

Locus-specific HERV-W RNA expression was obtained by summing up the aligned reads at each genomic location specified by Grandi et al. (2016). There was a total of 215 assessed HERV-W loci of which 65 were considered to have reliable expression (average number of aligned reads > 15). HERV-W loci 15q21.3 and 7q21.2 had overall highest expression with average of 353, and 317 reads, respectively. Rest of the loci had lower expression intensities ranging between average of 15 and 160 reads. Xq22.3 had mean read count of 31. Read counts for all studied HERV-W are provided in supplementary materials.

To assess the differences of HERV-W loci expression between the MS and control samples the differential expression analysis was performed. As shown in Table 2, a total of 14 HERV-W loci were found to be differentially expressed between MS and control samples. Of these, the majority (12) of HERV-W loci were overexpressed in MS samples (2q24.3, 2q13, 1q25.2, 2q22.2, 2p16.2, Xq13.3, 3p11.1, 4q23, 15q21.3, 6q21a, 11p14.2, and 6q21c), whereas only two loci (12q13.3 and 1q22) were expressed in lesser amounts in MS samples. The highest fold change in the difference of means was observed in case of HERV-W locus 2q24.3, where the expression was four-fold higher in the MS samples (16.74 vs 67.45 reads). One differentially expressed HERV-W locus was located in the X chromosome (Xq13.3). HERV-W locus 7q21.2, that codes for syncytin-1 was not amongst the differentially expressed loci. HERV-W locus Xq22.3, one of the putative loci for MSRV, was not observed to be differentially expressed either.

The association between differentially expressed HERV-W loci RNA expression intensities and estimated cell type proportions was studied with correlation analysis. As shown in Fig. 2, significant correlations were observed in case of oligodendrocytes, astrocytes, and excitatory neurons. In case of 7 HERV-W loci (6q21c, 11p14.2, 3p11.1, Xq13.3, 1q25.2, 2q13, and 2q24.3), the HERV-W expression intensity was significantly, inversely related with the oligodendrocyte proportion. Expression intensity of only one HERV-W locus (1q22) was significantly, positively correlated to oligodendrocyte proportion. Two HERV-W loci (2p16.2 and 2q22.2) had positive, significant correlation with astrocyte proportion and one locus (1q22) had negative, significant correlation. Four loci (6q21c, 11p14.2, 3p11.1, and Xq13.3) had positive, significant correlation with excitatory neuron proportions.

Finally, GSEA was utilized to shed light into biological processes to which HERV-W expression is associated. In case of three HERV-W loci, oligodendrocyte-related functions were identified to be suppressed as shown in Fig. 3. 2p16.2 was associated with suppression of oligodendrocyte development, suppression of ensheathment of neurons, and suppression of myelination. 2q13 was associated with suppression of oligodendrocyte development and differentiation. Xq13.3 was associated with suppression of ensheathment of neurons and suppression of myelination. In addition, HERV-W loci 2p16.2 and 2q13 were associated with immune system related processes, such as T cell activation, positive regulation of cytokine production, and antigen receptor-mediated signaling pathway. GSEA results for all differentially expressed HERV-W, that had mean expression > 15 counts, are available in supplementary material.

Viral infections have been suggested as triggering environmental factor for MS. Especially HERV-W, a subfamily of HERVs, has been associated with MS in several studies. Although the precise pathogenic mechanisms and the specific HERV-W locus involved in MS remain unclear, it is suggested that the pathological effects are attributed to the env protein, that is translated from one or more HERV-W loci. In this study, we utilized high-throughput RNA sequencing data to investigate the transcriptional activity of 215 distinct HERV-W loci in individuals with MS and with healthy controls. Our analysis indicated that 65 HERV-W loci were transcriptionally active in the white matter of the brain. Among the expressed transcriptionally active HERV-W, 14 exhibited differential expression between MS brain samples and control samples. In addition, we found significant correlation between differentially expressed HERV-W loci and oligodendrocyte proportions in the white matter of the brain. Our findings demonstrate a substantial overlap with a previous study by Schmitt et al. (2013), where they utilized RT-PCR to quantify HERV-W locus expression and reported that 100 loci are transcriptionally active in the brain. Despite the difference in the quantification technique, the HERV-W loci that were transcriptionally most active, were remarkably similar to what we observed. In addition, both Schmitt et al. (2013) and our study identified the HERV-W locus at locus 7q21.2, which codes for syncytin-1, as having exceptionally high expression levels in the brain. These results suggest that the expression profiles of HERV-W loci in the brain are consistent across different cohorts. The differences in the number of transcriptionally active HERV-W loci can be attributed to the variation in sample size and differences in the quantification technique.

In the context of MS, neither Schmitt et al. (2013) nor our results observed upregulation of the locus at 7q21.2. The study conducted by Elkjaer et al. (2021) has also identified the HERV-W locus at 7q21.2 as the most transcriptionally active HERV-W locus and reported that it is not upregulated in MS. In contrast to these findings, earlier studies by Antony et al. (2006, 2007) have reported transcriptional upregulation of HERV-W locus at 7q21.2. Our results further challenge the initial suspicion of 7q21.2 being a source of increased env protein observed in MS. Another prime candidate for the source of env protein observed in MS is the HERV-W sequence at locus Xq22.3, which contains a nearly complete env ORF, producing truncated env protein resembling the structure of syncytin-1 (Roebke et al. 2010). Earlier studies have associated this locus as the source of env protein observed in the MS brain lesions (Rolland et al. 2006; Laufer et al. 2009; García-Montojo et al. 2014). Interestingly, our results indicated only moderate transcriptional activity (mean read count of 31) at Xq22.3 locus without significant upregulation in the MS patients. These findings suggest that Xq22.3 may not be the pathogenic HERV-W locus associated with MS. Our results are consistent with Schmitt et al. (2013), where they also did not observe MS-associated upregulation at this locus. Consequently, our study suggests, that transcriptional activities of HERV-W loci 7q21.2 and Xq22.3, previously considered prime candidates in MS pathogenicity, remain unaltered in MS.

However, we identified 12 distinct upregulated and 2 down-regulated HERV-W loci in MS patients. Three of the upregulated HERV-W loci (6q21a, 15q21.3, and 4q32.3) are considered as full-length but they have developed internal stop codons (Grandi et al. 2016). In addition, 15q21.3 was the HERV-W loci with highest transcriptional activity (mean of 353 read counts) in the white matter of brain. Interestingly, previous study by Laufer et al. (2009) has pointed out the loci 15q21.3 and 6q21c as a potential sources of env protein observed in MS. Seven of the upregulated HERV-W loci were insertions within the human coding genes: 1q25.2 (RASAL2), 2p16.2 (ASB3), 2q22.2 (KYNU), 2q24.3 (COBLL1), 6q21a (ATG5), 6q21c (SLC16A10), and 11p14.2 (ANO3). As for these loci, their upregulation could reflect the changes of the expression of these genes as a part of the transcriptional changes in MS. In fact, upregulation of ATG5 has been previously reported in MS (Alirezaei et al. 2009). Three of the upregulated HERV-W loci were insertions within non-coding genes: 2q24.3 (TCONS_00004484), 4q23 (LOC100507053), and Xq13.3 (TCONS_00016997). In this study, we utilized RNA-seq data obtained from Elkjaer et al. (2019), who also investigated the transcriptional activity of HERV-W loci using the same dataset (Elkjaer et al. 2021). Elkjaer et al. (2021) did not observe upregulation of the HERV-W loci in MS when total reads where accounted, but found some upregulated and downregulated HERV-W loci on chromosome 7 in MS patients. In contrast, our analysis did not identify any differential HERV-W transcriptional activity in chromosome 7. This discrepancy may be attributed to the use of a different HERV-W references. Elkjaer et al. (2021) utilized more comprehensive set of references (HERVd database), whereas we used a more compact set of references by Grandi et al. (2016). It is also worth noting that Elkjaer et al. (2021) did not analyze differential expression of individual HERV-W loci outside of chromosome 7.

To assess the biological implications of transcriptional activity of HERV-W RNA in MS, we estimated the abundance of the neuronal, glial, and endothelial cells in the white matter of the brain. Plenty of studies have utilized CIBERSORTx for the deconvolution of peripheral immune cells, for which it is well validated. However, less studies have utilized CIBERSORTx for the deconvolution of central nervous system cells due to lack of well validated signature matrices. Astrocytes are often considered as the predominant glial cell in the CNS (Miller 2018). However, in the context of the white matter, the oligodendrocytes are the most abundant glial cell (Walhovd et al. 2014; Sigaard et al. 2016; Seeker et al. 2022). This is due to the role of oligodendrocytes as maintainers of myelin sheaths around neuronal axons, which reside in the white matter of the brain. Overall, the cell proportions obtained in this study are well in line with what is known about the cell abundances in the white matter. Specifically, our results highlight the predominance of oligodendrocytes in the white matter of the brain. Neurons, on the other hand, are sparsely distributed in the white matter as they primarily reside in the grey matter (Bae et al. 2020). Thus, our results further validate the performance of CIBERSORTx in estimating cell type proportions in the CNS. In the context of MS, we identified three statistically significant alterations in the cell type proportions between MS and control samples. The most prominent change was the decrease in oligodendrocyte proportions in the MS samples. This was expected as it is well known that oligodendrocyte loss is part of the MS pathophysiology (Cudrici et al. 2006). However, the specific mechanism leading to this decrease remains unclear. Both, oligodendrocyte apoptosis (Staugaitis and Trapp 2021) and the impairment of oligodendrocyte precursor cells (OPCs) (Cui et al. 2013) have been suggested to be relevant in MS. Unfortunately, we were not able to differentiate these possibilities, as the used signature matrix cannot distinguish the OPCs.

The significant increase of the astrocyte proportions in MS is consistent with findings from previous literature where proliferation and activation of astrocytes, astrogliosis, have been reported in MS (Correale and Farez 2015). Observed increase in the endothelial cell proportions could be result of peripheral immune cell infiltration to the brain due to inflammatory response. Alternatively, altered transcriptional profile might also affect how the cell type proportions are determined. While microglia also play an essential role in the MS (Luo et al. 2017), we did not observe any significant differences in microglia proportions. It is important to note that activation of microglia rather than their proportion have been reported previously in MS (O’Loughlin et al. 2018).

To assess the pathogenic potential of the 14 differentially expressed HERV-W loci, their expression intensities were correlated with the obtained cell proportions. The most prominent finding was that expression intensities of 8 HERV-W loci (6q21c, 11p14.2, 3p11.1, Xq13.3, 2p16.2, 1q25.2, 2q13, and 2q24.3) were strongly, inversely correlated to oligodendrocyte proportions. This finding suggests a potential link between the dysregulated expression of these HERV-W loci and oligodendrocyte function or survival, possibly implicating them in the pathogenesis of MS. This result was also supported by GSEA as three of the HERV-W loci were associated with suppression of oligodendrocyte-related processes, such as oligodendrocyte development and myelination. However, none of these HERV-W loci contain particularly well-preserved env genes suggesting that these loci do not contribute to the production of env protein observed in MS. Therefore, we found no evidence linking HERV-W loci capable of encoding env protein to the observed reduction in oligodendrocyte abundance. The study by Antony et al. (2004) presents a contrasting view showing that Syncytin-1, env protein of HERV-W at 7q21.2, is cytotoxic to oligodendrocytes. On the other hand, HERV-W mRNA has been shown to activate NF-κB and increase the production of type 1 interferons (Morris et al. 2019; Rangel et al. 2022; Russ and Iordanskiy 2023), and thus has pathogenic potential. Consequently, in theory, any HERV-W loci could be implicated in the MS pathogenesis as long as it is recognized as foreign RNA by the innate immune system. The exact mechanisms underlying the pro-inflammatory effects of HERV-W mRNA remain to be elucidated and further studies are needed to understand the molecular interactions between innate immune system and HERV-W mRNA. The underlying cause of the upregulation of the observed HERV-W is unclear. One explanation could be changes in transcriptional and/or epigenetic landscapes resulting from the disease. Alternatively, the upregulation might be attributed to a proportional increase of specific cells, that exhibit more relaxed HERV-W expression in these loci. For instance, infiltrating immune cells recruited to the CNS because of the disease might contribute to this.

There are some inherent limitations in HERV studies due to the multicopy nature and high sequence similarity between elements. One notable concern is the lack of standardized nomenclature and classification of HERVs. This can lead to inconsistencies in naming across studies and make comparisons difficult. Additionally, variety of methodologies are used for HERV detection and analysis. These include PCR techniques, RNA sequencing and bioinformatics approaches, and they can yield incomparable results. Lastly, deconvolution methods that are based on the transcriptome are still far from well-established. Namely, the lack of suitable reference profiles is a major limitation. Here we have used healthy brain transcriptome as a reference, but disease can induce gene expression changes, which could affect the cell type abundance estimates, potentially leading to inaccurate results.