The role of human endogenous retroviruses in gliomas: from etiological perspectives and therapeutic implications

Abstract

Accounting for approximately 8% of the human genome, human endogenous retroviruses (HERVs) have been implicated in a variety of cancers including gliomas. In normal cells, tight epigenetic regulation of HERVs prevent aberrant expression; however, in cancer cells, HERVs expression remains pervasive, suggesting a role of HERVs in oncogenic transformation. HERVs may contribute to oncogenesis in several ways including insertional mutagenesis, chromosomal rearrangements, proto-oncogene formation, and maintenance of stemness. On the other hand, recent data has suggested that reversing epigenetic silencing of HERVs may induce robust anti-tumor immune responses, suggesting HERVs’ potential therapeutic utility in gliomas. By reversing epigenetic modifications that silence HERVs, DNA methyltransferase, and histone deacetylase inhibitors may stimulate a viral-mimicry cascade via HERV-derived dsRNA formation that induces interferon-mediated apoptosis. Leveraging this anti-tumor autoimmune response may be a unique avenue to target certain subsets of epigenetically-dysregulated gliomas. Nevertheless, the role of HERVs in gliomas as either arbitrators of oncogenesis or forerunners of the innate anti-tumor immune response remains unclear. Here, we review the role of HERVs in gliomas, their potential dichotomous function in propagating oncogenesis and stimulating the anti-tumor immune response, and identify future directions for research.

First described in 1981, human endogenous retroviruses (HERV) account for approximately 8% of the human genome and serve as relics of ancestral germline retroviral infection.^1–3 HERVs can be identified by the presence of canonical retroviral genes: gag (core structural proteins), pol (viral replication enzymes including reverse transcriptase) env (envelope protein), and LTR (long terminal repeats containing promoters, enhancers, and other regulatory proteins; Figure 1). Currently, there are 31 HERV families (classified according to their tRNA primer-binding site) that are divided into three main classes based on their homology to exogenous retroviruses (see Table 1). Due to their accumulation of coding deletions and postinsertional mutations, many HERVs have inherently lost their ability to replicate and form functional viral particles. However, a few HERVs (mainly HERV-K, subtype HML-2) have retained near full-length sequences with some degree of transcriptional activity.^4–6 Polymorphisms have also been identified in HERVs however, their clinical significance remains unclear.⁷ The transcription of these “semi-active” HERVs remains tightly regulated by epigenetic modification of retroviral LTRs and nuclear transcriptional factors. As such, many HERV elements are expressed in early embryogenesis and are critical in organ development but remain silent in normal adults.⁸ However, HERVs may be re-expressed in pathologic conditions with epigenetic dysregulation such as amyotrophic lateral sclerosis (ALS), multiple sclerosis (MS), and cancer. Over the last several years, an association between human cancers (melanoma, breast, prostate, and lymphoma) and HERV transcriptional activation has been reported, but a clear etiologic role of HERVs in cancer has yet to be established.^9–12 A thorough examination of HERVs in brain neoplasms, specifically glioma, has yet to be performed. Here, we review the potential role of HERVs in glioma pathophysiology and highlight therapeutic options to leverage HERVs for immunotherapy.

Fig. 1.

Schema of original HERV infection and integration into human genome. HERVs maintain similar retroviral phylogeny with conserved RNA sequences gag (group antigens), pol (polymerases), and env (envelope proteins). Retroviral integration into the genome requires reverse transcriptase and integrase. Once integrated, HERVs are typically silenced by epigenetic machinery, but occasionally, HERV env splice products (HERV-K hml) may be transcribed into functional proteins, rec, and np9.

Table 1.

Human Endogenous Retrovirus Classification

HERV Class	Homology	Type
HERV Class I	Mammalian Type C retroviruses (murine leukemia virus)	HERV-H, HERV-F, HERV-RW, HERV-W, HERV-P, HERV-E,HERV-ERI, HERV-R, HERV-T, HERV-I, HERV-FRD
HERV Class II	Type D retroviruses Mammalian Type B (Mouse mammary tumor virus)	HERV-K (HML1-10)
HERV Class III	Spumavirus (foamy virus)	HERV-L

HERV Expression in Gliomas

Currently, the presence or significance of HERV expression in glioma samples remains uncertain. Most of the extensive bioinformatic HERV analyses do not include gliomas due to the heterogeneity of these cancers, limitations caused by the RNA preparation method, lack of deep sequencing, and inadequate control samples. However, a few bioinformatics studies have suggested that HERV protein expression is increased in gliomas. A large study of over 300 glioma samples from the Chinese Glioma Genome Atlas (CGGA) demonstrated that HERVK-113 was detected in the genome of over 90% of samples, although no association between virus expression and survival was noted.¹³ Additionally, it has also been suggested that HERV LTRs are differentially upregulated in glioma tissue compared to normal brain samples, albeit these results are preliminary.¹⁴ Despite these findings, quantitative mRNA qPCR analysis from established glioma cell lines and patient tissue samples failed to demonstrate a robust expression of either full-length HERV-K envelope protein or its splice products. Since HERV-K envelope splice products have been associated with tumor transformation, their absence suggests that HERV-K splice products may not be associated with human astrocytic tumors. Aside from HERV-K, the association of other families of HERVs and gliomas has not been investigated thoroughly. Although limited in nature, the aforementioned studies support the need for a more thorough investigation of the association of HERVs and gliomas.

HERVs as Oncogenic Drivers in Glioma

Research over the last four decades has implicated endogenous retroviruses in cellular de-differentiation and malignant transformations.¹⁵ HERVs may contribute to tumor formation/pathogenesis in two main ways: indirect transcriptional regulation of oncogenes/tumor suppressors and expression of oncogenic HERV proteins. However, these oncogenic properties have not been methodically investigated in gliomas. Therefore, we aim to highlight potential mechanisms by which HERVs may influence glioma oncogenesis (Figure 2).

Fig. 2.

Theoretical oncogenic role of HERVs in cancer.¹⁶ HERV LTR can be rearranged to act as cis-promoters of downstream oncogenes. Nonhomologous recombination of HERVs can lead to chromosomal rearrangements, leading to disruption of functional cell regulator genes. Insertional mutagenesis of HERV elements can also inactivate tumor suppressor genes. HERV-K envelop splice products (np9 and rec) have also been associated with oncogenic activity. HERV-FRD and HERV-W have been implicated in cell-cell fusion via syncitin-1, and have been implicated in several gynecological cancers.¹⁷

Oncogenic HERV Proteins

Although none of the HERVs possess established oncogenes, protein products from HERV transcripts have been implicated in inducing tumorigenesis by directly altering cellular signaling pathways, promoting tumor immune evasion, and inducing cellular fusion.⁵ Splice products from the envelope protein of HERV-K (HML-2), rec, and Np9 have been linked to tumorigenesis in a variety of cancers due to their activation of the c-MYC proto-oncogene.¹⁸ By binding to the transcriptional repressor promyelocytic leukemia zinc finger (PLZF), rec, and np9 may dysregulate the cell cycle and modulate cellular proliferation in specific cancers. Aside from mediating the expression of the protoconcogene C-MYC, PLZF also regulates tumor suppressor genes p53 and p21.¹⁹ Of note, PLZF was noted to be increased more than three-fold compared to normal controls in glioblastoma samples, suggesting that this is a potential target for HERV-K proteins in human high-grade gliomas.^19,20 Nevertheless, the role of PLZF in human gliomas as a tumor initiator has not yet been established. The other splice product np9 has also been associated with disinhibiting Notch signaling by binding with the Ligand of Numb Protein X (LNX) and initiating proteasome degradation.²¹ Since the LNX protein has been shown to be downregulated in glioma cells, an investigation of the role of np9-LNX complex may help understand aberrant Notch signaling pathways in gliomas.²²

Aside from cellular proliferation, HERVs have also been implicated in promoting tumor invasion and metastasis for specific cancer types. Most notably, HERV-H has been associated with expansion of a immune-evasive CD271+ cell populations in metastatic colon cancer cells undergoing epithelial-mesenchymal transitions. In contrast, depletion of the HERV-H-mediated CD271 populations resulted in tumor regression.²³ CD271 (p75 neurotrophin receptor) is a well-known mediator of cellular migration and invasion in gliomas and medulloblastomas; furthermore, CD271+ cells also form a stem-cell niche of treatment-refractory glioma cells that resist conventional chemoradiation and evade immunosurveillance.^24–27 As such, investigations into the role of HERV-H in regulating CD271 expression in gliomas may be warranted.

Indirect Gene Dysregulation

HERVs have also been implicated in indirectly dysregulating proto-oncogenes due to nonhomologous recombination in various cancers. In glioma, HERV-K derived sequences were discovered interspersed within tumor suppressor and DNA repair genes, BRCA2, and XRCC1, respectively.²⁸ These viral-derived interspersed rearrangements share homology to HERV-K and may be implicated in gene inactivation of regulatory genetic machinery. Additionally, chromosomal translocation of HERV LTRs has also been associated with enhanced expression of downstream oncogenes in several cancers.²⁹ In prostate cancer, the enhanced expression of the ETV1 oncogene has been associated with upstream HERV-K 22q11.23 5′-UTR LTR translocation; the ETV1 oncogene has also been noted to be overexpressed in a variety of capicua (CIC) mutated oligodendrogliomas and has been associated with an aggressive phenotype.^30–32 Similar, oncogenic chromosomal rearrangements of HERVs have been described in childhood glioma forming syndromes such as constitutional mismatch repair deficiency (CMMRD) due to biallelic inactivation of the mismatch repair gene, PMS2.^33,34

The promoter function of HERV LTRs also serves as a well-established mechanism of HERV-induced cellular dysregulation. Through cis activation of a downstream oncogene, an HERV-derived LTR can induce tumorigenesis.^35,36 In such a manner, HERV LTRs can promote overexpression of proteins such as CSF1R and ERBB4 in certain lymphomas. These proteins have also been linked to an aggressive phenotype in glioblastoma and medulloblastoma, respectively, but a causal link between HERVs and their upregulation has not been explored.^36–40 The transcriptional regulation of cancer-promoting genes by LTRs has been well described in a variety of cancers. Still, cis LTR regulation of oncogenes in gliomas has yet to be definitively established (see Table 2).

Table 2.

Potential Linkages of HERVs as Cancer-Promoting Agents in Glioma

Study	HERV-element	Mechanism	Cancer Cellular Gene	Gene Role in Glioma
18	Rec/Np9	Activation of C-MYC oncogene	PLZF	Upregulated three-fold in glioma samples
21	Np9	Amplifying Notch signaling via LNX degradation	LNX	Downregulated in gliomas
28	HERV-K	Gene inactivation by insertional mutagenesis	BRCA2	Germline inactivation mutations noted in some samples
28	HERV-K	Gene inactivation by insertional mutagenesis	XRCC1	Polymorphisms associated with increased risk of gliomas
32	HERV-K LTR	Chromosomal rearrangement activating ETV1 proto-oncogene	ETV1	ETV1 overexpression associated with aggressive phenotype in oligodendrogliomas
33	HERV LTR	Chromosomal rearrangement	PMS2	Association with CMMRD
37	HERV LTR	cis activation of promoter	CSF1R	Implicated in Glioma invasion41
40	HERV LTR	cis activation of promoter	ERBB4	Associated with aggressive glioma phenotype42
43	HERV-K	Correlation to CD133+ stem-cell markers in melanoma	CD133	Glioma stem-cell marker, associated with aggressive phenotype
8	HERV-K (Hml-2)	Maintenance of stemness through interactions with CD98HC/mTOR	CD98HC	CD98 upregulated in astrocytic tumors, facilitates tumor proliferation, amino acid transport44

Abbreviations: CSF1R, colony-stimulating factor 1 receptor; ERBB4, receptor tyrosine-protein kinase; ETV1, ETS translocation variant 1; LNX, E3 ubiquitin-protein ligase; PLZF, promyelocytic leukemia zinc finger protein; PMS2, Mismatch repair endonuclease; XRCC1, X-ray repair cross-complementing protein 1.

HERVs May Be Responsible for Stem-Cell Phenotype in Cancer Cells

One of the earliest examples of HERV function was discovered in embryonic stem cells, which activate HERV-K expression to maintain pluripotency and an undifferentiated state. Through epigenetic modifications (histone/DNA hypomethylation), HERV-K expression is induced during embryogenesis/development to protect the host by inducing antiviral defense mechanisms through the envelope splice product, rec.⁴⁵ Additionally, envelope proteins from HERV-W and FRD, syncitin-1 and 2, both facilitate cell-cell fusion during embryogenesis and trophectoderm development.⁴⁶ In cancer, associations between HERVs and cancer stem cells have been described. CD133 (a common glioma stem-cell marker) was found to be tightly correlated to HERV expression in melanoma cell lines, and treatment with reverse transcriptase inhibitors restricted both HERV-K expression and CD133+ cancer cell populations.⁴³ The HERK-K envelope protein (HML-2) is found to be overexpressed in human pluripotent stem cells (PSC), and consequently downregulated during neuronal differentiation. The HERV-K (HML-2) interacts with CD98HC and activates the mTOR pathway, while concurrently inducing epigenetic changes through lysophosphatidylcholine acyltransferase (LPCAT1).⁸ CD98 is widely expressed in astrocytic tumors, where it may impact tumor biology by facilitating amino acid transport and promoting oncogenic transformation.⁴⁴ Accordingly, epigenetic silencing or downregulation of HML-2 reverses stemness and promotes neuronal differentiation, suggesting a potential role of HERV-K envelope protein in neurogenerative conditions and neuro-oncology.⁸

HERVs as a Biomarker of Cancer

The association of HERV expression and cancer suggests that HERVs could be utilized as a prognostic biomarker for certain cancers. The most notable prognostic biomarker HERV-K (HML-2) has been studied in breast and ovarian cancer patients; HERV-K (HML-2) expression in serum and tissues was quantitatively correlated to both breast and ovarian cancer histopathology (low-grade lesions with demonstrably lower expression of HERV-K (HML-2) and antibodies compared to higher-grade lesions).^12,47,48 Additionally, HERV expression was also noted to increase after induction of isocitrate dehydrogenase (IDH) R132 mutation in glioma cells. In this model, immortalized astrocytes with inducible IDH1 mutant enzyme were noted to have a global increase in HERV expression with concomitant epigenetic changes.⁴⁹ Similarly, HERV upregulation (HERV-K, T, W, F) was also noted to increase in glioma-initiating cells after exposure to cytomegalovirus. Given the potential etiologic role of CMV in glioma, there is a suggestion that HERVs may have a bystander role in glioma tumorigenesis.⁵⁰ Analogous findings were noted in non-HIV permissive glioma cell lines (U87) after exposure to HIV-1.⁵¹ However, it is unclear if HERV expression is a consequence of oncogenesis or an intrinsic cellular immune defense mechanism. Nevertheless, these findings suggest a potential role for HERVs as a biomarker of disease in gliomas.

Therapeutic Implications of HERVs in Gliomas

In addition to the potential tumorigenic role of HERVs, recent studies suggest that HERV expression may help augment immunotherapy for gliomas.

Adaptive Immune Response Against HERVs

The adaptive immune response to HERVs is inherently suppressed because of two factors: the presence of immune tolerance to intrinsic HERV protein and epigenetic silencing of HERV-specific lymphocytes. However, in cancer, upregulation of HERV expression activates the immune response to a certain extent producing an antibody response against HERVs proteins.⁵² Anti-HERV-K antibodies were discovered in several cancers, including prostate, breast, germ-cell tumors, and melanoma.^{36,48,53–55} To amplify the efficacy of this intrinsic antibody response, direct antibody therapy, and vaccination strategies have been suggested for a variety of cancers. Indeed, exogenous monoclonal antibodies against HERV env protein have demonstrated considerable efficacy in a variety of experimental models of cancers and in a human study of multiple sclerosis.^52,56 Preclinical vaccination strategies against HERV envelope or gag proteins have also demonstrated considerable safety and efficacy in reducing the metastatic tumor burden in renal carcinoma and other cancers.^5,57,58 Of note, vaccination strategies against HERV proteins in gliomas have not yet been investigated.

HERV-Mediated Viral-Mimicry

As mentioned previously, HERV expression in normal cells is tightly mediated by epigenetic repressive changes that silence retroviral transcription. Epigenetic modifications, including histone/DNA hypermethylation and histone deacetylation, prevent HERV transcription, thereby silencing innate immune responses. Repressive histone markers such as H3K9me3 have been associated with HERV silencing.^59,60 These repressive histone markers are mainly mediated by histone methyltransferases (SUV39h1, SUV39h2, SETDB1), and knock-down of histone methyltransferase SETDB1 has been associated with increased HERV expression in embryonic stem cells.^61–63 As such, a retroviral silencing complex (RSC) comprising SETDB1 (histone methyltransferases), HUSH complex, tripartite motif-containing protein 28 (TRIM28), and others has been established to regulate the expression of transposable elements and HERVs.⁶⁴ Of interest, H3K9me3 and SETDB1 have both been demonstrated to play a critical role in gliomagenesis, especially in gliomas with a hypermethylator CIMP (CpG island methylator) phenotype, and TRIM28 has also been shown to increase the proliferation of glioma cells.^65–67

Given these findings, there has been considerable interest in capitalizing on HERV expression by reversing epigenetic changes with the fundamental premise that transient HERV expression may induce an innate antiviral immune response (i.e. viral mimicry). Since most of the epigenetic silencing hinges on DNA/histone hypermethylation or deacetylation, targeted pharmaceutical blockade of these enzymes has been proposed. For example, recent evidence has demonstrated that DNA methyltransferase inhibitors (DNMTi) could induce pro-apoptotic interferon responses that are mediated by increased HERV expression.^68,69 Specifically, DNMTi therapy elicited dsRNA from HERVs that are sensed by TLR3 and MDA5, which precipitate a robust interferon cytokine response that ultimately suppresses proliferation and induces apoptosis (Figure 3). Similarly, it was demonstrated that DNMTi increased transposable element expression and HERV-derived peptides in established glioblastoma cell lines, suggesting a potential synergistic role of DNMTi-induced overexpression of HERVs and immunotherapy.⁷⁰ Early phase 1 studies using DNMTi (5-Azacitidine) monotherapy in recurrent high-grade IDHm gliomas achieved disease stabilization in approximately 40% of patients but failed to achieve a durable radiographic response (PFS and OS of 4.7 and 25 months, respectively).⁷¹ Of note, the study was relatively limited by the small sample size (n = 12), histological heterogeneity and concomitant use of bevacizumab. Similar to DNMTi, activation of a viral-mimicry cascade can also be elicited through histone deacetylase inhibitors (HDACi) which relieve repressive histone marks. Currently, early-phase clinical trials using HDACi (vorinostat) in high-grade gliomas have demonstrated safety/tolerability in treating recurrent glioblastoma, albeit with some dose-limiting toxicity.^72–74 However, HDACi and immunotherapy have not been combined in a clinical setting to treat high-grade gliomas, and the clinical efficacy of HDACi in improving overall survival/progression-free survival has yet to be demonstrated. The clinical role of epigenetic therapy in actuating HERVs in gliomas requires further exploration, which optimally would include a “window of opportunity” study to confirm that HDACi therapy alters the epigenetic profile of the tumor and increase HERV expression.

Fig. 3.

Viral mimicry cascade. Epigenetic therapies may de-repress HERV expression, leading to formation and release of nuclear dsRNA. dsRNA binds MDA5 which signals MAVS (mitochondrial antiviral signaling protein) to activate IRF and NF- kB. The end-result of this cascade is the induction of a pro-inflammatory immune response mediated by interferon (IFN). Anti-HERV antibodies are also produced by activation of TLR by dsRNA.

The Dichotomous Role of HERVs in Oncogenesis and Immunotherapy

The divergent roles of HERVs as either oncogenic drivers or arbitrators of the immune system have yet to be defined in gliomas. Most data suggest that specific subsets of HERVs may contribute to oncogenesis through envelope proteins (most commonly HERV-K, hml-2), insertional mutagenesis, and nonhomologous recombination.^5,29,35 In gliomas particularly, increases in HERV expression were concomitantly found after induction of IDH mutations, a precursor of gliomagenesis; however, these associations have not established causality.⁴⁹ Therefore, HERV’s direct contribution to oncogenic transformation in gliomas has yet to be established. On the other hand, the reactivation of HERVs through viral-mimicry pathways in gliomas has been demonstrated in several studies through epigenetic reprogramming.⁷⁰ Exposure to demethylating agents or histone modifiers induces total HERV expression and secondary endogenous antiviral immune responses in gliomas through secondary dsRNA-mediated activation interferon pathways. However, selectively inducing viral-mimicry responses in gliomas with epigenetic drugs may be complicated since aberrant reactivation of HERVs in normal cells may prompt de novo mutations/transformations and unregulated autoimmune responses. Therefore, there must be a delicate balance when considering activating HERVs for immunomodulation to circumvent systemic toxicity. Ultimately, reactivation of HERVs through a viral-mimicry state in glioma remains an appealing concept for the future of glioma immunotherapy.

Limitations of Classifying HERVs in Gliomas

As mentioned previously, identifying and quantifying HERV expression in human brain tumors remains challenging. Primarily, restricting analysis to the tumor component remains a principal issue given the intrinsic heterogeneity of glioblastoma and a tumor microenvironment containing intervening normal brain tissue. Any computational platform for transposable elements must account for this heterogeneity and adjust for tumor sample purity. Analysis of tumors with poor tumor purity and intervening normal parenchyma/non-tumor proteins must be amended to accurately reflect the parent tumor phenotype. Additionally, epigenetic changes within glioma specimens vary widely; IDH-mutant gliomas carry a distinctive epigenetic phenotype that may significantly alter HERV expression, so separate analyses of each tumor subset may be necessary. Single-cell analysis may be necessary to identify heterogeneity within the tumor and to identify any somatic mutations and novel insertions. As mentioned previously, HERV expression is tightly regulated by epigenetic silencing mechanisms including TRIM28, HMTs, HDAC, and HUSH complexes; hence, analysis of tumor tissues should consider the tumor’s distinct epigenetic phenotype that may regulate HERV expression. Tumors with pervasive histone/DNA modifications could theoretically repress HERV expression and should be accounted for when comparing results among multiple patients. Lastly, typical RNA-seq platforms and datasets are not optimized to accurately detect HERVs and their associated polymorphisms. Large-scale bioinformatic studies on HERV expression in known databases such as TCGA or CCGA have insufficient reading depth (<100 bp, <50 million reads) or inadequate preparation (non-rRNA depleted) to accurately detect HERV sequences; therefore, false-negative rates may be relatively high. To date, no comprehensive database of HERV expression exists for gliomas due to the aforementioned reasons. Only a handful of established total RNA-seq datasets are appropriately prepared to quantify HERV expression, such as the encyclopedia of DNA elements (ENCODE) for certain tumor cell lines; however, the applicability of cell-line specific analyses has yet to be demonstrated for primary tumor tissues.^75,76 Additionally, users must carefully select computational tools to adequately estimate the retrotranscriptome of these RNA-seq data sets; several software algorithms have been proposed to adequately quantify retrotransposable elements including TE-transcripts, Telescope, Salmon-TE, etc.^77–79 As such, focused bioinformatic analyses of glioma specimens may be required to better understand the role of HERVs in gliomas.

Conclusion

Research data suggest several mechanisms by which HERVs could play a role in glioma pathogenesis. HERVs may facilitate tumor formation through direct protein formation, its splice products or indirectly dysregulating proto-oncogene/tumor suppressor expression. In addition to maintaining stemness of pluripotent cells, HERVs may also act as biomarkers of glioma progression or acquisition of mutations (IDH1 mutation). Although there is decent data to implicate the role of HERVs in glioma pathogenesis, in-depth -omics on the genome, epigenome, and transcriptome may help further elucidate their function. Aside from their role in oncogenic transformation, HERVs can also elicit robust anti-tumor immune responses in gliomas through a viral-mimicry cascade which may synergize with immunotherapy. New strategies to capitalize on HERV-induced anti-tumor immune responses in glioma should be considered therapeutic options.

Funding

This work was supported by National Institute of Neurological Disorders and Stroke (R25NS108937-01).

Conflict of interest statement. The authors have no financial or personal conflicts of interest.

Authorship statement. A.H.S. contributed to design and preparation of manuscript. M.G., A.N., and J.H. critically reviewed paper and provided senior mentorship.