As occurs in pluripotent embryos, many cancers are known to reactivate transposable elements (TE) through chromatin remodeling, DNA hypomethylation, and histone modifications [
12]. TEs are fundamentally heterogeneous, comprising two main types: DNA transposons and retrotransposons, which include Long Interspersed Nuclear Elements (LINEs), Short Interspersed Nuclear Elements (SINEs), and Long Terminal Repeat (LTR) Retrotransposons. Our focus in this review will be on LTR retrotransposons or human ERVs (hERVs).
The genome of an intact hERV provirus comprises at least 5’ and 3’ long terminal repeats (LTRs) flanking an internal
Gag (group-specific antigen)
-Pro (protease)
-Pol (polymerase) polyprotein-coding sequence.
Gag is cleaved by
Pro to generate a virus-like particle that contains the fusion protein and ERV mRNA. hERVs may also contain a remnant envelope (
Env) and other accessory genes but are generally not infectious [
12]. HERV-K (HML-2), first reported in 1986, is the most recent hERV integrated into genomic DNA. Unlike other hERVs, HERV-K contains a near-full-length transcript in the human genome that includes open reading frames (ORFs) of HERV-K
Gag,
Pol, and
Env, which can be read and translated into functional retroviral proteins [
61].
ERV regulatory networks fundamentally consist of pre-transcriptional and post-transcriptional regulatory networks [
63]. Pre-transcriptional regulators include zinc-finger proteins (ZFPs), TRIM28-SETDB1, human silencing hub (HUSH) complex, SWItch-sucrose non-fermentable (SWI-SNF) complex, MORC proteins, Lymphoid-specific helicase (LSH), and P-element-induced wimpy testis (PIWI)-interacting RNAs (piRNAs). Alternatively, piRNAs, nuclear exosome targeting (NEXT) complex, and RNA methylation comprise the post-transcriptional modulators. Table
2 summarizes the proteins and function of the ERV expression regulators.
Table 2
Known Regulators of hERVs
ZFPs & TRIM28-SETDB1 | Recruits HIRA complex, heterodimeric protein complex of DAXX, and ATRX | Binds to and methylates proximal DNA and trimethylates proximal H3K9 residues to silence the transcription of ZFP-bound ERV loci Replace canonical H3 proteins with the transcriptionally repressive non-canonical H3.3 variant | |
HUSH Complex | Composed of three protein subunits, MPP8, PPHLN1, and TASOR Recruits MORC2 | Universally represses TE transposition and expression by performing targeted repression of loci corresponding to intron-less RNA | |
SWI-SNF Complex | Composed of PBRM1 | Rearrange location of histone proteins in chromatin, which impacts loci accessibility for epigenetic modification | |
MORC | Facilitates HUSH, DNA methylators, and DAXX-ATRX | Compacts chromatin through multimeric assemblies that trap DNA loops to restrict access, which influences gene expression | |
LSH | Facilitates DNMT enzymes and H3K9 trimethylating enzymes | Permits accessibility of DNA transcription factors and chromatin-modifying enzymes to regulate ERV expression | |
PIWI-interacting RNAs | Utilizes argonaute proteins Guides H3K9 tri-methylating and DNA methylating enzymes | Guides DNA and histone silencing activity proximal to loci complementary to the piRNA guide | |
Post-transcriptional |
PIWI-interacting RNAs | Utilizes argonaute proteins | Recognize and cleave complementary ERV mRNAs | |
NEXT Complex | Composed of RBM7, ZCCHC8, and MTR4 Recruited by HUSH | Degrades intron-less RNA | |
m6A RNA Methylation | Deposited by METTL3-METTL14 Recognized by YTHDFs | m6A on 5'UTR signals for ERV mRNA degradation | |
hERVs may drive oncogenesis in two main ways: (1) indirect transcriptional regulation of oncogenes/tumor suppressors and (2) expression of oncogenic HERV proteins [
11].
There is accumulating support for hERVs as an oncogenic driver and emerging target for treatment via epigenetic silencing. On the other hand, the role of post-transcriptional silencing and RNA modifications of hERV transcripts in glioma remains virtually unexplored. This subsection briefly recounts major hERV-mediated oncogenic findings and provides context for the onco-exaptation discussion later in the review.
Akin to its role in development, hERVs may serve as alternative promoters for proximal genes in malignant cells and cryptic transcription start sites to produce aberrant CDS mRNA [
12]. Similarly, onco-exaptation may proceed by hERVs acting as enhancers [
64]. Moreover, HERV-K-derived sequences appear to interrupt and inactivate tumor suppressor BRCA2 and DNA repair gene XRCC1 in glioma cells [
65].
Through c-MYC proto-oncogene activation, HERV-K
Env and accessory proteins Rec and Np9 have been linked to tumorigenesis in various cancers [
66]. Rec and Np9 have been proposed to bind to the transcriptional repressor promyelocytic leukemia zinc finger (PLZF), which mediates the expression of the protooncogene c-MYC and suppressor genes p53 and p21 [
67]. Np9 has also been associated with amplifying Notch signaling via binding to and initiating Ligand of Numb Protein X (LNX) degradation [
68]. Interestingly, in glioma cells, LNX protein has been found to be decreased [
69]. Therefore, the HERV-K Rec and Np9 represent putative oncogenes and require further investigation as therapeutic targets in gliomas displaying signs of hERV activation.
Finally, hERVs have been proposed to be responsible for the stem-cell phenotype in cancer stem cells (CSCs) [
11,
61]. CD133, a common glioma stem-cell marker, strongly correlates with hERV expression in melanoma cell lines. Treatment with reverse transcriptase inhibitors lowers HERV-K expression and CD133 + melanoma cell populations [
70]. Moreover, the HERV-K env is overexpressed in pluripotent stem cells (PSC) but downregulated during neuronal differentiation. This same protein interacts with CD98HC, activates the mTOR pathway, and induces epigenetic changes through lysophosphatidylcholine acyltransferase (LPCAT1) [
71]. CD98 is widely expressed in astrocytic tumors, where it has been suggested to promote oncogenic transformation by facilitating amino acid transport [
72].
Epigenetic and m6A epitranscriptomic crosstalk
Chemical modifications on DNA, RNA, and proteins (e.g., histones) impact gene regulation. As discussed above, installing an RNA m6A modification alters mRNA stability and translation. Furthermore, emerging data suggests that m6A methylation influences physiological regulation beyond post-transcriptional mechanisms [
73]. While loss of m6A writers and nuclear m6A readers are known to be developmentally lethal, knockout of the YTHDF family, cytoplasmic m6A readers involved in transcript decay, has not been shown to recapitulate lethality [
74,
75]. Although alternative explanations, such as compensation from other readers, are undoubtedly reasonable, evidence that suggests m6A methylation feeds back onto epigenetic circuits has rapidly accumulated.
A vital function of heterochromatin is restraining the activity of embedded satellite repeats and transposable elements [
76]. Endogenous retroviruses (ERVs) are a prominent class of retrotransposons that necessitate constitutive silencing by regulation machineries, traditionally understood to comprise epigenetic processes. Remarkably, three separate research groups have recently reported a role for m6A in regulating ERVs through an element known as intracisternal A particle (IAP) in mouse embryonic stem cells (mESCs) [
77‐
79]. m6A on the 5′UTR of the IAP mRNA recruits YTHDF readers for mRNA degradation. Thus, m6A levels are inversely correlated with mRNA and protein levels of IAP. Knockout of Mettl3 writers and rescue by a catalytically inactive form failed to restore H3K9me3 levels at IAP elements, while knockout of Alkbh5 erasers significantly increased H3K9me3 levels at these sites. Thus, this data suggests a positive link between m6A deposition and levels of H3K9me3, a significant molecular feature of heterochromatin.
The current explanation for these findings is that m6A RNA modifications catalyzed by Mettl3 are known to interact with Ythdc1 readers. This interaction, confirmed via chromatin immunoprecipitation followed by sequencing (ChIP-seq) with enrichment at H3K9me3-rich transposable elements, has a role in mediating retrotransposon silencing and maintaining mESC identity [
79]. In addition, Ythdc1 appears to guide Mettl3 and facilitate its interaction with chromatin, tripartite motif containing 28 (TRIM28), and SET-domain-bifurcated histone lysine methyltransferase 1 (SETDB1). This aptly named m6A methyltransferase complex regulates H3K9me3 deposition at IAPs.
Despite some inconsistencies in the proposed mechanism, these studies [
77‐
79] provide convincing evidence that m6A directly impacts heterochromatin formation. Furthermore, expanding evidence has revealed the role of m6A methylation on chromosome-associated regulatory RNAs (carRNAs) or mRNA-encoding histone-modifying enzymes and accessible chromatin. There is evidence of reverse feedback of histones on m6A modifications and of the histone elongation mark H3K36me3 guiding m6A deposition [
80]. Kan et. Al more extensively discusses these concepts [
73].
It is currently unknown if and how RNA modifications other than m6A feedback on epigenetic checkpoints.
Viral mimicry
Viral mimicry describes an active anti-viral cellular state triggered by an endogenous stimulus. It may evoke innate and adaptive immune responses and can be triggered by cytosolic RNA or DNA [
81].
Two landmark papers in 2015 describe “viral mimicry” as a process involving inhibition of epigenetic silencing, retrotransposon transcription, and IFN activation. The host cells interpret aberrant repetitive element RNA expression as a viral infection and activate an IFN response [
82,
83]. Recognition of retrotransposon-derived duplex RNAs (dsRNA) by cytosolic RNA sensors like MDA5 or endosomal RNA sensors such as Toll-like receptor 3 (TLR3) initiates antiviral signaling. Stimulation of RNA sensors is propagated by mitochondrial antiviral-signaling protein (MAVS) aggregation on the mitochondrial surface, which induces a TBK1-mediated phosphorylation cascade that results in the phosphorylation, dimerization, and nuclear localization of IRF3/7 to activate either type I or III IFN signaling. This pro-inflammatory immune response ultimately suppresses proliferation and induces apoptosis in the affected cell. In addition, viral mimicry enhances adaptive immune responses as hERV-derived peptides form tumor-associated antigens (TAAs) that may elicit CD8 + T-cell responses. Glioblastoma is characteristically considered to be an “immunologically cold” tumor. Using adjunctive agents to enhance intratumoral viral mimicry and its innate and adaptive immune responses could bolster the effects of present immunotherapies.
In glioblastoma cell lines, DNA methyltransferase inhibition (DNMTi) increased the expression of TE and HERV-derived peptides [
84]. However, clinical trials utilizing adjuvant DNMTi have been unsuccessful to date. A phase 1 trial using 5-Azacitidine monotherapy in recurrent high-grade IDHm gliomas reached disease stabilization in approximately 40% of patients but failed to accomplish a durable radiographic response [
85]. Histone deacetylase inhibitors (HDACi) such as Vorinostat, which has a narrow therapeutic index, have shown safety/tolerability for the treatment of recurrent glioblastoma [
86‐
88] but have not yet been combined with immunotherapies for clinical treatment of high-grade gliomas. In addition, the clinical efficacy of HDACi in improving progression-free or overall survival has yet to be demonstrated.
Several recurrent cancer-driving mutations are known to activate TEs that prime tumors for viral mimicry induction. One of the most studied examples is the H3.3
K27M mutation in high-grade gliomas [
89]. This mutation impairs the recruitment of the Polycomb complex, reduces facultative heterochromatinization, and thus activates DNA transposons, LINEs, and SINEs [
81]. However, despite elevated retrotransposon expression, H3.3
K27M gliomas do not have elevated IFN signaling, likely secondary to a cancer-specific compensatory mechanism. Treatment of high-grade glioma with either DNMTi or HDACi has been shown to more strongly promote MAVS-dependent induction of IFN and IFN-stimulated genes in H3.3K27M cancers than H3.3 wild-type cancers. The resulting enhanced dsRNA responses promote PKR-mediated cell death [
89]. Therefore, although H3.3K27M gliomas lack IFN induction at baseline, H3K27me3 loss results in elevated retrotransposon expression and primes for viral mimicry responses to DNMTi and HDACi treatment.
Aside from ERVs, numerous studies have linked LINEs and SINEs to a viral mimicry response [
12]. In fact, the cytosolic dsRNA sensor MDA5, a vital player in viral mimicry, was found to preferentially bind to the stem-loop structure formed by inverted-repeat Alus (IR-Alus) relative to the bidirectionally transcribed dsRNA structures of ERVs and LINEs [
90]. IR-Alus is also the primary substrate for the A-to-I mRNA editor ADAR1 [
91], which produces a modification that disrupts the RNA duplex and prevents MDA5-mediated dsRNA sensing [
92]. Therefore, ADAR1 depletion increases cytosolic dsRNA levels and sensitizes cancer cells to treatment by viral mimicry inducers such as decitabine or CADK4/6 [
90]. Furthermore, tumors with intrinsically high IFN signaling have been found to be ADAR1 dependent [
93,
94], and thus, ADAR1 inhibition may be exploited for these viral mimicry-primed cells. On the other hand, immune-checkpoint blockade-resistant tumors in mice were found to be sensitive to both IFN and anti-PD1 blockade in ADAR1 deletion [
95]. As immune checkpoint blockade resistance is common in high-grade gliomas, combined ADAR and epigenetic therapies may represent a novel path for a synthetic vulnerability to immunotherapies.
The mechanisms that dictate whether elevated dsRNAs induce sublethal or lethal IFN responses remain unclear. Survival of the subset of cancer cells with both high dsRNA levels and ADAR1 deficiency suggests that ADAR1-mediated editing is not the only mechanism cancer cells utilize to evade immunogenic cell death. As we strive to develop more robust and clinically relevant drugs, other RNA regulators and editors providing compensatory immunosuppression must also be explored. One recently found example is the RNA helicase DHX9 in breast cancer cells [
96].
Long read methylome data indicates that hERV CpG methylation is lower at baseline than other TEs and the remaining genome in normal tissues [
97]. Thus, it is probable that in many tumor cells, DNA methylation may not be the principal mechanism limiting ERV activation [
98]. Though HDAC inhibitors and lysine methyltransferase inhibitors have been shown to synergize with DNMTis to activate ERVs [
99,
100], other repressive and less characterized pathways likely provide compensatory effects. Further studies and improvements in epitranscriptomic mapping of gliomas are necessary to identify specific cancer-driving mutations in RNA modification regulators that prime tumors for viral mimicry. With crosstalk to epigenetic silencing already established, it is probable such mutations exist and may represent promising targets for novel high-grade glioma treatment in combination with immunotherapies and pre-transcriptional ERV regulation inhibitors (i.e., DNMTi and HDACi). Though several small molecule inhibitors of RNA modification regulators already exist, many must be redesigned to improve their BBB penetrance and levels in GBM. CRISPR-Cas systems developed for programmable RNA modification editing [
101] and nucleotide-specific editing [
102] expand the scope of RNA engineering and facilitate mechanistic understanding of the epitranscriptome. Like their DNA-targeting counterparts, RNA-targeting CRISPR-Cas systems may begin to be translated into the clinic in the coming years.
Post-transcriptional methods of viral mimicry induction are undergoing further study for recurrent cancers like glioblastoma. Despite its promise, the potential benefits of leveraging inducible hERV activity against solid tumors must be closely balanced with possible unintended consequences such as ERV onco-exaptation.