1. DNA and RNA G-Quadruplexes

As introduced earlier, G-quadruplexes are relevant both at the DNA and RNA levels, with RNA G4s being often more stable than DNA G-quadruplexes. When a DNA sequence can fold into a G4 structure, its corresponding RNA motif may also adopt a G4 topology (Krafčíková et al., 2017b). Consequently, a sequence involved in the regulation of transcription may also be involved in the regulation of translation or other mechanisms related to mRNA functions if the sequence is transcribed. Overall, the formation of G4 structures by both DNA and RNA sequences makes G-quadruplexes relevant for most viruses as long as their genome contains G-rich regions. Although this feature is attractive for antiviral agents—G4 motifs being targetable both as DNA and RNA—it may complicate the deconvolution of biologic effects of G4-based antiviral agents.

Over 200 G4 structures are currently available in the Protein Database (PDB) structure database. Unfortunately, only a few of them correspond to viral G-quadruplexes. Two solution NMR structures of DNA G4 motifs found in the HIV-1 long terminal repeat (LTR) are known, and one example is presented (Fig. 4) (De Nicola et al., 2016; Butovskaya et al., 2018). This HIV G-quadruplex offers specific epitopes for recognition and can be considered as “druggable.” The other solved structure by NMR belongs to the DNA motif of human papillomavirus type 52 (Marušič and Plavec, 2019).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

The structure of HIV-1 LTR G-quadruplex (6H1K), elucidated by the Richter and Phan teams (Butovskaya et al., 2018). This motif involves both a G4 and a duplex part. (A) The surface of G4 displays various cavities, unlike the duplex DNA. (B) A ball-and-stick representation of the same structure.

G-quadruplex stability depends on a number of factors. Intracellular ionic conditions are favorable for G4 formation, and molecular crowding further stabilizes G4s (Matsumoto et al., 2020). Long G-runs allow the formation of more quartets, and this is often correlated with high stability.

Stability is not the only key parameter affected by the primary sequence. G-quadruplexes are inherently polymorphic, and different topologies can be formed (Phan et al., 2006), such as parallel, antiparallel, or hybrid conformations. Even left-handed G-quadruplexes may be observed (Bakalar et al., 2019). Additional levels of variability may be provided by loop-loop interactions, capping base triplets or base pairs, bulges, and noncanonical quartets [for example, a CGCG quartet (Lim et al., 2009)]. As a consequence, although most G-quadruplexes mostly or exclusively rely on the formation of the same elementary bricks, the G-quartets, they can each adopt a unique fold. This druggability feature is due to the intricate assemblage of guanine residues, which provide a patchwork of aromatic surfaces, clefts, and valleys as well as unique electrostatics. These properties enable the G-quadruplex to interact specifically with proteins or even smaller ligands, which can compete with proteins or mimic some of their roles. Obviously, nucleic acids, including G4s, do not offer equivalent targets as protein binding sites. Still, the observation that cellular and viral proteins as well as artificial and natural compounds selectively interact with these structures suggests that these G-quadruplexes are attractive targets for drug design.

2. Searching for G-Quadruplexes

Initial attempts to identify potential G4 motifs in a relatively small genome involved the manual search of GG-, GGG-, and/or GGGG-islands on the same strand. A variety of bioinformatics tools are now available to predict motifs susceptible to adopt a G4 structure. Putative G-quadruplexes were initially searched in the human genome by using computational approaches employing the sequence motif G₃₊N_1-7G₃₊N_1-7G₃₊N_1-7G₃₊, in which G represents the guanine residues in each guanine tract, usually directly involved in G-tetrad formation, and N indicates any combination of loop residues (Huppert and Balasubramanian, 2005; Todd et al., 2005). The so-called Quadparser algorithm was first applied to the human genome with the abovementioned sequence motif as a query, and it was allowed to pick up over 300,000 candidate sequences. Since then, alternative search tools, such as G-Quadruplex forming G-Rich Sequences (QGRS) Mapper (Kikin et al., 2006) and G4-Hunter (Bedrat et al., 2016), have been designed [for a recent review, see Lombardi and Londoño-Vallejo (2020)]. Computational bioinformatics studies indicate that the density of G4s in viral genomes is moderately but not exclusively related to the G/C content of the viruses. A global analysis of human versus viral PQSs suggests that short-looped PQSs are more frequent and have a similar composition across viral taxonomic groups. Besides, there is a higher number of pyrimidine loops in viruses infecting animals irrespective of the viral genome type. Computational and statistical studies by Puig Lombardi et al., (2019) advocate the idea that the genome of viruses is rich in C-looped G-quadruplexes, possibly acting as a transcriptional binding site for the host transcription factors. The PQS density is 2- to 3-fold greater in viruses infecting vertebrate hosts than any other host (Puig Lombardi et al., 2019), implying a possible coevolution of these structures, as vertebrates also tend to be G4-rich.

Table 1 presents typical examples of G4 motifs in the virus genomes. We have chosen to rank these sequences according to the G4-Hunter score, which correlates with G4 propensity and stability, as discussed before. Although this ranking does not reflect the relative strength of G4 sequences in each virus, it illustrates that it is much easier to find a stable G4 motif in the KSHV genome than in the one of SARS-CoV-2. This table illustrates the differences in G-richness between motifs studied: Not all G4s are equal, and although some of them are probably so stable that dedicated helicases are required to unfold them, others are far more labile and may have different functions when folded or unfolded.

3. Viral G-Quadruplexes

Given that both DNA and RNA G-rich sequences are prone to G4 formation, it is not surprising that G4-motifs have been identified in nearly all viruses, including double-stranded DNA (ds-DNA), single-stranded DNA, and single-stranded RNA (ss-RNA) (e.g., Retroviridae) (Lavezzo et al., 2018). Among the viruses that contain definitive G4 structures in their genome, characterization of these structures in cancer-related viruses like KSHV, EBV, HBV, HCV, HIV, and human papillomavirus (HPV) deserves paramount attention (Saranathan and Vivekanandan, 2019). For a full review of the importance of G-quadruplexes in viral pathogenesis, one can refer to a recent review by Ruggiero and Richter (2020).

ds-DNA viruses, such as members of the Herpesviridae family, are especially enriched in G4 motifs, and most of the computationally identified G4-prone motifs were proven to be exceedingly stable under physiologic conditions. In DNA viruses, such as adeno-associated viruses and human herpesviruses, G4s modulate DNA replication, whereas G4s in the promoter region of HBV and the mRNA of EBV modulate transcription and translation, respectively (Ravichandran et al., 2018). Human herpesviruses are categorized into three subcategories: alphaherpesviruses, betaherpesviruses, and gammaherpesviruses. PQS frequencies (predicted by Quadparser) are higher in modulatory regions of immediate early genes compared with early and late genes in most herpesviruses. HSV-2 has the highest number of PQSs (n = 318) among human herpesviruses, with PQS densities as high as 1.037/kb (i.e., 7-fold greater than the PQS density in the human DNA genome). The PQS densities of HSV-1 and HSV-2 are currently the highest reported for any genome ever sequenced, with the exception of some Archaea (Brázda et al., 2020). The repeat regions in herpesvirus genomes are important for the maintenance of the episomal form of the genome. PQS densities are significantly enriched within the repeat regions of herpesviruses. The telomeric-like motif “GGG​TTA​GGG​TTA​GGG​TTA​GGG” is repeated 81 times within a 3-kb region of the HHV-7 genome and a total of 204 times in the whole genome of this virus (Biswas et al., 2016).

An important feature of G-quadruplexes in some viruses, such as HSV-1, is that regions of the virus genome with higher G4 densities tend to be close to recombination breakpoints. Approximately 11% of breakpoints are located within a G4 motif, proffering these DNA secondary structures as hot spots for recombination regulation in the HSV-1 genome (Saranathan et al., 2019). Intriguing findings of the versatile functions of HSV-1 G-quadruplexes do not end here. Biswas et al. (2018) divulged that the packaging signal (pac-1 signal) in the termini of HSV-1 genome is crucially involved in the recognition of protein machinery required for the cleavage of the viral genome, its encapsidation, and virion assembly. Interestingly, the mouse monoclonal antibody 1H6 was shown to visualize G4 structures in host cells infected with HSV-1 by the aid of immunofluorescence and immune-electron microscopy (Artusi et al., 2016). G-quadruplex formation in the viral genome was found to be cell cycle–dependent and peaked at the time of viral DNA replication, traveled to the nuclear membrane at the time of virus nuclear egress (the export of viral capsids from the nucleus to the cytoplasmic area), and was later tracked in HSV-1 immature virions released from the host nucleus. However, this antibody is not truly G4-specific and has been reported to bind to other structures; thus its use for the detection of viral G4 can lead to ambiguous findings (Kazemier et al., 2017). A more specific G4 antibody, such as the now commercially available ScFv version of the BG4 antibody, would be recommended (Biffi et al., 2013). However, a recent study indicates that BG4 can also bind some cytosine-rich sequences on single-stranded DNA (Ray et al., 2020).

In RNA viruses, such as retroviruses, flaviviruses, and filoviruses, high variability may be observed, as evidenced by Coronaviridae, in which an astonishing intraspecies variance in G4s density was recently found (Lavezzo et al., 2018). In spite of their high genetic variability, all retroviruses besides ε-retroviruses contain highly conserved putative G-quadruplex–forming sequences in their promoter regions (Glazko and Kosovsky, 2013; Ruggiero et al., 2019). In fact, the G-rich clusters in LTR regions are strikingly conserved in all primate lentiviruses. Also, the majority of PQSs (∼70%) are located in the U3 region just upstream of the transcription start site.

For RNA viruses, it is of quintessential importance to elude the host RNA decay machinery for survival. One of the characterized ways to accomplish this task is to target the host 5'-3' exoribonuclease 1 (XRN1). The 3ʹ untranslated regions (3' UTRs) of several phleboviruses and arenaviruses contain RNA structures that block the ribonuclease activity of XRN1. RVFV, a member of the phlebovirus of the Bunyaviridae family, exploits this mechanism of action to evade RNA destruction. The 3ʹ-terminal segment of the nucleocapsid (N) mRNA of RVFV can repress XRN1, and it is likely that this evasion is carried out by employing a G-quadruplex structure (Charley et al., 2018). Among viruses, RVFV is unusual in that it is composed of two negative-sense and one ambisense RNA during infection, generating transcripts from both strands. Zika virus, a member of the Flaviviridae family, displays seven PQSs that are markedly conserved within the genomes of >58 flaviviruses. Notwithstanding that these sequences are in general conserved between different species, their exact biologic role has remained nebulous (Göertz et al., 2018).

1. Viral G-Quadruplex–Binding Proteins

The role of G4s in the viral life cycle is becoming increasingly clear as viral proteins are found to have a G4-binding or unwinding activity. The first example was found in SV40, which bears a large multifunctional protein called T-antigen, which is capable of unwinding G4s (Plyler et al., 2009).

More recently, Rajendran et al. (2013) and Butovskaya et al. (2019) reported that HIV-1 nucleocapsid protein 7 (NCp7) binds to stable G4 structures in viral RNA and may promote its unfolding to assist in viral reverse transcription. The discovery of G4 chaperones or helicases in viral genomes is very promising and also implies that the regulation of viral processes, such as reverse transcription or replication, is so vital that viruses have implemented several regulatory elements like G4 and protein chaperones to supervise these steps. NCp7 has an impressive affinity (K_d around 10 nM) for the central DNA flap region [central polypurine tract (cPPT) region] of HIV-1, which may fold into an intermolecular parallel DNA G4 (Lyonnais et al., 2003). Potassium ions and a dibenzophenanthroline derivative coined as MMQ3 were able to stabilize these G-quadruplexes (Lyonnais et al., 2002), hinting that capsid assembly and packing of the viral genome may also be affected by G4 formation.

2. Cellular G-Quadruplex–Binding Proteins

In addition to viral proteins, several host G4-binding proteins have been shown to regulate virus life cycles. More cellular G4-interactive proteins are characterized mainly because of their relevance in various diseases that are linked to G4s, especially cancer.

Nucleolin has been implicated in several pathologic processes, such as tumorigenesis (in which lots of G4 structures in the promoters of the oncogenes are defined) and viral pathogenicity. Nucleolin binds to G4 structures found in the HIV-1 LTR promoter as well as in host promoters, such as c-myc. In addition, nucleolin may also be recruited to G4 present in aborted RNA transcripts generated from the hexanucleotide repeat motif (GGGGCC)n as well as to the Epstein-Barr virus–encoded nuclear antigen (EBNA) 1 mRNA. Indeed, nucleolin acts as a vital host protein capable of binding to the viral RNA G4 structure; this protein can stabilize these structures and suppress viral replication (Bian et al., 2019).

Another case of cellular G4 binding protein is nucleophosmin (NPM1) that augments the infectivity potential of adeno-associated viruses (Satkunanathan et al., 2017). NPM1 stabilizes the G4 structures in the c-myc oncogene and is also implicated in the repression of cellular as well as viral replication. It has been suggested that this protein directly interacts with the G4 motifs to mobilize the packaging of viral genome and encapsidation. Our current understanding of the interaction between NPM1 and viral G4 is far from complete, but knockdown of this protein results in the increased replication of viral DNA and localization of packaged adeno-associated virus particles in the cytoplasm. It also seems that NPM1 stabilizes G4 structures in viral genome and thus creates a blockage in the replication process, which is contradictory to its cooperative role in viral packaging and pathogenesis (Satkunanathan et al., 2017).

Heterogeneous nuclear ribonucleoprotein families have also been revealed as unfolding agents of viral G-quadruplex structures. Among this family, heterogeneous nuclear ribonucleoprotein A2/B1 untwists the G4 structure in the LTR of HIV-1 and boosts the level of transcription (Scalabrin et al., 2017). Stabilization of these structures by stabilizing compounds can halt viral transcription without seriously affecting telomere G4 stability (Perrone et al., 2015).

Finally, the SUD domain of SARS Nsp3 protein can interact with G4 structures, and this interaction has been proposed to regulate SARS transcription/replication (Kusov et al., 2015). This regulation and its potential use as a therapeutic target will be discussed in the chapter dedicated to coronaviruses.

1. Immune Modulation

To address the role of host G-quadruplexes in viral pathogenesis, we first need to clarify the function of these G4s for the host cell. Oncogenic promoters, telomeres, introns, and both 5′ and 3′ UTRs of mRNAs are the most well characterized locations where G-quadruplexes have been reported in the human genome and transcriptome (Carvalho et al., 2020). Among RNA, G4 can be found in noncoding RNAs as well. Studies have revealed that a type of long noncoding RNAs transcribed at telomeres called “TElomeric Repeat-containing RNA” is actively engaged in the mechanisms orchestrating telomere sustenance and chromosome end sheltering (Bettin et al., 2019). Telomeric RNA/G4-forming sequences suppress the expression [STAT1 (signal transducer and activator of transcription 1), ISG15, and 2′,5′-oligoadenylate synthetase (OAS3)] of the innate immune system in three-dimensional cultures. This suppression was similarly triggered by the nontelomeric G-rich DNA aptamer AS1411. Both TElomeric Repeat-containing RNA and AS1411 fold into G4 structures, which inhibit the induction of specific innate immune genes in cancer cells (Hirashima and Seimiya, 2015). We first need to elaborate the underlying cellular pathways involved in these genes—STAT1, IFN-stimulated gene (ISG) 15, and OAS3—to assess the druggability of these host sequences:

• STAT1 is a key target of IFN-γ, acting as a transcription factor of various genes associated with antiviral proteins and enzymes, microbicidal molecules, phagocytosis-related receptors, cytokines, chemokines, inflammatory pathways, and also antigen-presenting molecules (Hu and Ivashkiv, 2009).

• ISG15, encodes a ubiquitin-like protein induced mainly by type I interferons and viral infections. This protein is conjugated to numerous cellular proteins, a process styled as ISGylation. A plethora of proteins involved in antiviral signaling, including RIG-I (for retinoic acid-inducible gene I), MDA-5 (for melanoma differentiation-associated protein 5), Mx1 (for Myxovirus resistance protein 1), PKR (protein kinase R), STAT1, and JAK1 (Janus kinase I), have been characterized as target proteins for ISGylation. Some viruses stimulate the production of viral-specific proteins that can deconjugate ISG15 from its target proteins or inhibit the ISGylation of the mentioned proteins, therefore abolishing the antiviral response of the immune system (Jeon et al., 2010).

• OAS proteins are a group of proteins with antiviral activity stimulated by the IFN-α and IFN-β (Lee et al., 2019). They are implicated in the degradation of viral RNA with the aid of ribonuclease L (RNase L) (Choi et al., 2015).

Overall, these telomeric G4 structures inhibit the transcription of antiviral proteins. Therefore, targeting host-cell telomeric G4s by ligands may aid in the body’s response to the viral pathogens. This mechanism may explain why some ligands display spurious results in vitro but contradictory or disappointing results in vivo. The selectivity and specificity toward viral G4s should be high enough to override the stabilization caused in the immune-related genes. Apart from this point, this stabilization can also be helpful in viral conditions in which a massive cytokine storm is noted, as found in severe cases of COVID-19 (Ye et al., 2020). In this case, the immune activation by the alluded genes and pathways of the interferons is only aggravating the condition, and a broad-spectrum stabilizer might be more efficient in subjugating the viral disease.

There are some other aspects of the immune system that are interestingly regulated by G4 structures. G-quartet nuclease 1 is a human nuclease capable of cutting G4 DNA. This protein is linked to heavy-chain class-switch recombination in immunoglobulin genes (Sun et al., 2001). Typically, this switch entails a shift between expressing IgM and IgD to the expression of IgG, IgA, or IgE, inducing an upregulation in the humoral response of the immune system that is essential in viral elimination (Stavnezer and Schrader, 2014). Whether the activity of this nuclease can be exploited in antiviral drug development remains an open question.

2. Oncogenic Viruses

A diverse set of viruses known as oncoviruses is known to cause cancer in humans. Some viruses, such as EBV, KSHV, HPV, HBV, and HCV, which are all known to have G4 sequences with functional importance, are classified as type 1 carcinogens by the International Agency for Research on Cancer (Kumar et al., 2020). G4 ligands that stabilize cellular oncogenes suppress the expression of proliferative genes, and they can also prevent the expression of viral genes. Thus, application of such ligands should be explored for preclinical and clinical applications.

Oncogene promoters are among the best-studied forms of G-quadruplexes. Oncogenes are known to have regulatory effects on the immune-related genes. Here we summarize a brief perspective about the immune significance of these oncogenes to assess whether these oncogenes can have positive or negative effects on the immune system and viral resistance. Depending on the type of oncogene, this effect can be upregulatory or downregulatory. For instance, the Myc oncogene is well recognized as one of the proteins involved in the mitigation of immune elicitation by the highly proliferative cells (Casey et al., 2018). VEGF (vascular endothelial growth factor) can cause a defect in the functional maturation of dendritic cells from their progenitors. VEGF also prevents the differentiation and function of different immune cells during hematopoiesis (Li et al., 2016). There is an increasing amount of evidence that advocates the idea that KRAS (Karsten-RAS) is closely tied with the immune evasion of tumor cells and the production of Th1-2–suppressive cytokines (van Maldegem and Downward, 2020). HIF-1α (hypoxia-inducible factor 1 alpha) activity is stimulated in response to viral pathogens, but increasing evidence indicates that the result of this activation can favor the virus and not the host. Some viruses have developed mechanisms to stabilize HIF-1α to produce an antiapoptotic effect that sustains the survival of the infected cells (Palazon et al., 2014). On the other hand, c-kit (tyrosine-protein kinase KIT, CD117) is known to have very complicated signaling pathways in the immune system, and c-kit mutation with a gain of function is reportedly related to mast cell proliferation and allergic reactions. c-kit has also been implicated in skewing the differentiation of T cells to Th2 and Th17 cells and away from Th1 (Ray et al., 2010). Overexpression of BCL-2 (B-cell lymphoma 2) in the mice is linked with higher immune response and prolonged survival of B cells (Renault and Chipuk, 2013). RET (rearranged during transfection), another proto-oncogene, is engaged in the proinflammatory pathways and also homeostasis of the immune system (Rusmini et al., 2013). To conclude, the crosstalk between oncogenes and the immune system is complex and needs to be better understood if one wants to exploit it against viral infections.

3. Cellular Pathways

Recently, researchers discovered that a G-quadruplex is present in the promoter region of the human TMPRSS2 (transmembrane protease, serine 2) gene, which encodes a type II transmembrane serine protease that can sever hemagglutinin of many subtypes of influenza viruses and spike glycoprotein of coronaviruses (Shen et al., 2020). Benzoselenoxanthene analogs are capable of stabilizing this G-quadruplex, downregulating this gene and manifesting demonstrable antiviral activity commensurate with the inhibitory activity of oseltamivir, a neuraminidase inhibitor of influenza viruses. It remains to be determined whether this antiviral effect was mediated by TMPRSS2 inhibition or via the stabilization of other viral or host-cell G-quadruplexes. In fact, benzothioxanthene derivatives, which are structurally related to benzoselenoxanthenes, also stabilize telomeric G4s (Mergny et al., 1998). In any case, these studies sprout the idea that stabilization of human G4 structures may be beneficial for the treatment of viral infections.

4. Telomeric Integration

HHV-6A and HHV-6B are two distinct types of ds-DNA viruses that belong to the subfamily Betaherpesvirinae. HHV-6B is a ubiquitous virus that infects nearly 100% of the human population. HHV-6A infection causes sixth disease (exanthema subitem or roseola infantum) in children, and the presence of HHV-6 in normal brains suggests a latent phase in the central nervous system, which, in some rare cases, can be later reactivated and lead to encephalitis (Limeres Posse et al., 2017; Fida et al., 2019). In the latent phase, human herpesviruses typically maintain their genomes as extrachromosomal nuclear episomes. How HHV-6A/B achieves latency is still enigmatic. The HHV-6A/B genomes consist of a unique sequence that is flanked by G-rich direct repeat regions that harbor the packaging sequences (pac-1 and pac-2) and two arrays of either perfect or imperfect telomeric repeats (TMRs) at the genome termini. The fluidity of telomeres is important for efficient chromosomal integration of HHV-6A, and that interference with telomerase activity negatively affects the generation of cellular clones containing integrated HHV-6A. Gilbert-Girard and colleagues (2017) have examined the effects of a G-quadruplex binding and stabilizing agent, BRACO-19, on HHV-6A chromosomal integration. BRACO-19 reduced the number of clones harboring integrated HHV-6A and negatively affected HHV-6A integration in telomerase-expressing cells (Gilbert-Girard et al., 2017). Although the effects of BRACO-19 on the viral TMRs and pac-1 G4s remain partially unexplained, their observation (stabilization of host G4s in telomeres) suggests that telomeric G4 ligands can serve as a prevention therapy for those at risk of HHV-6A virus or even a treatment for related viruses that integrate their genome at the telomeric segments of the host like Marek disease virus [MDV, an oncogenic alphaherpesvirus (Previdelli et al., 2019)].

A. Specific Studies on G-Quadruplexes in Coronaviruses: Response to Coronavirus Disease 2019 Pandemic

The recent COVID-19 outbreak stimulated an unprecedented research effort to find new targets against the SARS-CoV-2 virus. Knowing that G-quadruplex–prone motifs were predicted in the genome of the Coronaviridae family, including the MERS and SARS, it was interesting to determine whether putative G-quadruplex sequences are also present in the SARS-CoV-2 genome. Screening with QGRS Mapper, a web-based tool for identifying quadruplex-prone sequences, suggested that 25 such sequences are present but only with G2 motifs, meaning that they correspond to relatively unstable G4s (Ji et al., 2020; Panera et al., 2020). Two of these short RNA sequences were experimentally confirmed to form G4 structures in vitro but with low thermal stability (melting temperature, or Tm <37°C; meaning these structures are predominantly unfolded at physiologic temperature) (Ji et al., 2020). The same authors suggested that PQSs may be present in the open reading frames of several SARS-CoV-2 genes, such as spike glycoprotein, membrane, and nucleocapsid genes.

Bartas et al. (2020) analyzed the occurrence of putative G4 motifs within the 109 genomes of all known Nidovirales, such as coronaviruses, including SARS-CoV, MERS-CoV, and SARS-CoV-2. Rather than QGRS Mapper, they used G4Hunter to predict G4-prone motifs, keeping a minimal threshold of 1.2 (not a single motif with a G4-Hunter score of 1.6 or above was found in all Nidovirales). This threshold value, which has previously been shown to maximize accuracy (Bedrat et al., 2016), will miss relatively unstable G4 with lower scores – in other words, it is more stringent than the QGRS studies discussed above, explaining why fewer hits would be found. The lowest G4 density was found in Nidovirales infecting humans (0.06), including SARS-CoV-2, for which a single motif is present on the negative strand and none are present on the positive strand, with a G4-Hunter score above 1.2. The G4 density in this virus is lower than expected by chance, arguing for a counter-selection evolution against G-prone motifs in SARS-CoV-2. The 25 motifs reported by Ji et al. (2020) or Panera et al. (2020) all have lower G4-Hunter scores.

Interestingly, some coronaviruses code for G4-binding proteins. For example, the Nsp3 of SARS-CoV possesses a SUD, indispensable for the replication/transcription procedure of viral pathogenesis (Kusov et al., 2015). SUD is a 338-amino-acid domain that is absent in less-deadly coronaviruses and is therefore regarded as one of the domains that enhances the pathogenicity of SARS with respect to other members of the Coronaviridae family. SUD is composed of three subdomains dubbed as SUD-N [macrodomain 2], SUD-M (macrodomain 3), and SUD-C (domain preceding Ubl2 (ubiquitin-like protein 2) and PL2^pro (papain-like protease 2). The SARS-CoV SUD domain binds to G-quadruplexes (Tan et al., 2009; Kusov et al., 2015; Lei et al., 2018). SUD-NM interacts with RNA G-quadruplexes since Nsp3 has not been found in the nucleus of the cells. Mutations of some lysine residues in SUD-M abolish G4 binding and inhibit viral replication. SUD-C has also been announced as a regulator of the specificity of the RNA-binding activity of SUD-M (Johnson et al., 2010). By altering the pattern of gene expression in infected cells through the interaction of SUD to G4 structures present in mRNAs, SARS-CoV may wield the cellular processes in its desired states. As this SUD domain is conserved between SARS-CoV and SARS-CoV-2 (M. Lavigne et al., manuscript in preparation), this mechanism may contribute to the high pathogenicity of these two viruses.

The fact that the SARS virus codes for a G4-binding protein while being itself relatively G4-poor may seem surprising. However, besides viral G4, one should not forget that the host cell also contains G4-prone nucleic acids that may well be relevant for viral infections. In addition, other host-cell proteins important for infection, such as TMPRSS2 may possibly be regulated by G4 ligands. Both aspects will be discussed in the next section.

B. Assets and Drawbacks in the Development of Antiviral Therapies Targeting G-Quadruplexes

Nearly all currently used drugs are known to target protein structures, with a few exceptions, such as antibiotics, aimed at ribosomal RNA. Our understanding of how proteins behave far exceeds that for nucleic acids. There are countless ligands with versatile structures that are capable of modulating the behavior of different sets of proteins. Protein cavities are adroitly designed to interact with a panoply of other proteins, cellular metabolites, and xenobiotics. Hence, they are more druggable than their nucleic acid cousins, which are often considered to be less intricate than proteins and large peptides. The unique folding of proteins—with the notable exception of intrinsically disordered ones—makes them appropriate targets for medicinal chemists, especially when binding or catalytic sites are known. This can be exploited by a variety of ligands, including small molecules, peptides, aptamers, or nucleotides. In contrast, this subtlety is not the same for nucleic acids, as they offer relatively limited diversity in terms of “residue” properties (all nucleotides have identical charges and relatively similar hydrophilicities). In addition, RNA tends to be inherently flexible, and apart from noncanonical structures, it often lacks a specific binding pocket that can be exploited for specific recognition.

These drawbacks do not mean that nucleic acid targeting is impossible: for example, a number of antibiotics work by binding to the prokaryotic ribosome. In addition, G-quadruplexes constitute original targets among nucleic acids with unique advantages: 1) They are often well defined structures; 2) their electrostatics are fundamentally different from other DNA/RNA fold because of the presence of four sugar-phosphate backbones and a central spine of positively charged ions; 3) terminal quartets offer a stacking platform for planar aromatic ligands; and 4) loops, grooves, bulges, and flanking sequences offer additional epitopes for specific recognition. Overall, intramolecular G-quadruplexes should be considered as globular shapes rather than linear DNA or RNA polymers. Their folding landscape and timescale are also reminiscent of proteins.

These properties were exploited by a number of teams who identified hundreds of compounds binding to G-quadruplexes. Unfortunately, many of these molecules are not drug-like and violate some of Lipinski’s rule of five. Even though Lipinski’s rule of five was initially applied to oral drugs and has been challenged these years, at least for some parameters, this rule provides useful guidelines for optimizing pharmacokinetic tolerance and efficacy (Doak et al., 2014). For example, most G4 ligands tend to have a relatively large size; Quarfloxin (CX-3543), the first G4-targeting drug to reach the clinic [it went in phase II clinical trials for the treatment of cancer (Buket et al., 2014)] violates this rule in terms of molecular weight (Buket et al., 2014). In addition, while G4 offer unique advantages among nucleic acid targets, they also pose specific problems: 1) Some, but not all, G4s tend to be polymorphic; this structural variability is illustrated with the human telomeric motif (or, to a lesser extent, to viral sequences, such as the one present in the HIV LTR region) for which multiple folds are known. Identifying which ones are relevant in a physiologic environment is not straightforward; 2) although distinguishing G4 from non-G4 structures is relatively simple, making the distinction between various G4s is harder, especially when planar aromatic compounds are considered and most will bind to all G4 offering an accessible terminal tetrad for stacking. Given the high number of potential G4 sequences in the human genome, it may be well advised to limit off-target effects by selectively target a subset of G4 motifs; 3) the physiologic and biochemical pathways altered by G4 ligands are still largely elusive, in part because many biologic studies reported their effects on a few genes only and not at the transcriptome or genome-wide level; and 4) related to that, and in contrast with a widely accepted belief, G-quadruplex structures are not always acting as transcriptional repressors, and predicting the impact of a G4 ligand on the expression profile of a given gene is not straightforward.

These opportunities and drawbacks apply to all G4-based strategies, aimed at fighting cancer or transmittable diseases. There are, however, additional hurdles in antiviral research, as the biologic roles of viral G4s only start to be uncovered. It took decades to unambiguously prove the formation of G-quadruplexes in human cells (Lipps and Rhodes, 2009). Not all G-rich motifs found in viruses may adopt a G4 fold, and even then, not all G4 may be appropriate targets (i.e., they may not be essential for viral survival or pathogenesis).

Importantly, whenever viral G4 targets are considered, the genome and transcriptome of the host cell offer thousands of potential off-sites that can divert G4 ligands from their intended viral target. These cellular sites may be present in large molar excess, even when the virus is actively replicating. Searching for highly efficient binders with low interference with the host-cell functions may therefore be the way to go to design potent antiviral agents capable of fighting the current and future viral outbreaks. Some G-quadruplex ligands may have a higher affinity for viral G4 structures than for human structures. A c-exNDI induced significant virus inhibition with no cytotoxicity by inhibiting viral DNA replication, with consequent impairment of viral genes transcription. c-exNDI preferentially targeted HSV-1 G4s over cellular telomeric G4s, one of the most well established G4s within host cells, whereas other less abundant cellular G4s were also recognized in the host (Callegaro et al., 2017). Naphthalene diimide derivatives also displayed significant antiviral activity at nanomolar concentrations by targeting HIV-1 LTR promoter G4s (Perrone et al., 2015). The regulation of transcription banks on the binding of host-cell transcription factors and associated regulatory proteins to the LTRs. Treatment of HSV-1–infected cells with BRACO-19 caused a pronounced halt of virus production and viral DNA synthesis (Artusi et al., 2015; Frasson et al., 2019). On the other hand, one can propose that the broad stabilization of both host and viral G4s may have synergistic effects and constitute an alternative strategy for antiviral drug discovery.

To make progress, the physiologic role of host and pathogen G-quadruplexes ought to be uncovered, and to this end, we should first establish which sequences of viral genomes fold into these ordered structures. It may be possible that, as it is increasingly accepted for human G4s, only a small fraction of potential G4-forming motifs adopts a G4 fold under physiologic conditions. The same can be true for viral candidate sequences. The current number of putative G4 viral structures massively outnumbers the number of well characterized ones. Many of these sequences are annotated as putative G4-forming sequences (PQSs), pending exact in vitro and in vivo characterization. Identifying and validating a new G4 structure in the genome of a virus is a laborious task. These PQSs have been proposed in viral genomes that have not been fully investigated, such as the Zika virus. Moreover, HSV-1 genome indicates multiple clusters of repeated sequences creating highly stable G-quadruplexes involved in the replication of viral DNA (Puig Lombardi et al., 2019). Thus, our identification techniques should be further improved for a better understanding of G-quadruplexes.

We summarize the major drawbacks in utilizing G4 targetability in the antiviral context (Fig. 8). The first step is to identify appropriate targets. A complete in silico mapping of the known viral genomes has already been performed, and the currently available prediction tools allow testing every new virus genome in a few minutes. The recent example of the SARS-CoV-2 situation demonstrates that even this analysis is not straightforward, as teams may disagree on what a candidate sequence should look like and which algorithm should be used. As a consequence, the number of candidate sequences oscillates between 0 and 25. The next step will be to validate these candidate motifs, first by showing their formation on short DNA/RNA oligonucleotides in vitro (which may be harder than it seems) and then on longer fragments and ideally on the full transcript or the whole viral genome. The use of electronic circular dichroism allows an easy and straightforward technique to infer at least in vitro not only the presence of G4s but also their specific topology (parallel, antiparallel, hybrid, etc.) thanks to the emergence of specific spectral signatures (Cheng et al., 2018; del Villar-Guerra et al., 2018).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 8.

The general scheme of the obstacles of G-quadruplexes, as a mean to direct drug-like molecules toward them. For the development of a new ligand molecule in drug design and discovery, first we need to identify, validate, and determine the structure of a relatively stable G-quadruplex in the viral (or host) cells. This G-quadruplex should have pertinent role to viral infectivity. With the aid of medicinal chemistry, the affinity of the ligand should be optimized to reach the desired in vitro and in vivo efficacy. Certain precautions are needed to avoid genomic instability and toxicological damage to the normal cells.

Once validated, structural determination, although not indispensable, would definitely facilitate drug design but seems hard to achieve, as we do not have a single X-ray crystallographic structure of a viral G-quadruplex structure. Although one can argue that G4 structures can manifest themselves into a variety of topologies, their “targetability” or “ligandability” may not vary as much as for protein structures.

A number of screening approaches may then be proposed to identify G4 ligands binding to the target(s). Initial screening, either in silico with virtual libraries or directly with chemical libraries, is used to identify hit molecules. Using structure-activity relationship (SAR) between the previously characterized compounds allows establishing a correlation between activity and structure. This process can be also quantitatively accomplished using quantitative SAR. The next step will be to convert hits into leads and then drug-like compounds. This part should be familiar to any medicinal chemist, as it is in large part identical to any other drug development program. It may prove harder than for other targets though, as most G4 ligands display mediocre drug-likeness potential when compared. A series of drug-likeness evaluations should be implemented to modify and improve structure toward molecules with a balanced lipophilicity/hydrophilicity, appropriate size and molecular weight, and fair permeability to enter the cells. The penultimate step should be to minimize off-target effects and toxicity toward the host, which is crucial given the presence of multiple potential targets in normal cells. Long-term genomic instability studies have to be considered in this stage, since these compounds may interact with DNA (even if preferentially aimed at RNA). G-quadruplex ligand PDS, for example, was thought to cause double-strand breaks in some studies (Moruno-Manchon et al., 2017), whereas in some others it was found to even mildly mitigate the formation of such DNA damage (Kumari et al., 2019). The evaluation of the in vivo effect of the compound, along with its bioavailability, biodistribution, and pharmacokinetic profile is necessary, and the antiviral effect of the compound can first be determined in cellular models and then in vivo.

C. Hit-to-Lead Optimization for G-Quadruplex Ligands

In the rational drug design, hit-to-lead optimization can be accomplished by comparing the effects of compounds belonging to the same family but differing by chemical groups or atoms at specific locations. This allows to determine SARs, and this is often coupled with structural studies, wherein the impact of such modification on the binding to its target is determined thanks to high-resolution structural methods coupled to molecular dynamics simulations. Indeed, recently Hognon et al. (2020) assessed the exact binding interactions that favors the dimerization of SUD in cooperation with G-quadruplexes via molecular dynamics simulations and free-energy profiles. In their inspiring review about the application of molecular modeling and simulations in antiviral drug discovery, Francés-Monerris and colleagues (2020) state that the “open inactive” structure of SUD was shown to be disfavored by RNA G-quadruplexes. Their work suggests that specific ligands that perturb this mutual interaction could be employed in therapeutic approaches. Combined quantum mechanics/molecular mechanics studies are gaining attention despite their computationally intensive nature. The works by Batista group are a suitable example of application of these hybrid methods in high-quality prediction of PQS structures (Ho et al., 2014), which should be extended to study viral G-quadruplexes. The coupling of computational molecular simulation with experimental spectroscopy (Gattuso et al., 2016) provides a useful tool to unravel the detailed folding of G-quadruplexes and their interactions with other proteins and ligands, which are crucial in understanding viral infections, and can also reveal the drug mechanism of actions.

Alternative attractive strategies are based on fragment-based or scaffold-based hit optimization. Amiodoxime, cationic bis(acylhydrazones), benzoselenoxanthene, naphtalene diimides, dibenzophenanthroline, bisquinolines/quinolinium, acridine, and porphyrin are the scaffolds that have been tested for stabilization of viral G-quadruplexes. The relative paucity of medicinal chemistry studies on G4 ligands against viruses makes it difficult to provide a general core or scaffold for G-quadruplex ligands [for a review on G-quadruplex ligand structure, please refer to Sun et al. (2019)]. Hit optimization also uses in silico toxicological profile software and databases to ensure the relative safety of the compounds and their lack of mutagenic (and possibly carcinogenic) potential.

Unfortunately, in the G-quadruplex field, 1) the target is not always known, 2) some ligands may bind to multiple targets, and 3) either the three-dimensional structure of the target is poorly known, especially in a physiologic environment (topology may be affected by (macro)molecular crowding (Amjadi Oskouie and Abiri, 2021), for example), or no data are available on ligand-binding mode: although there are over 250 G4 structures in the Protein Database (PDB), only a few high-resolution structures of drug-G4 complexes are known.

There is, therefore, considerable work to be done to progress from hit to lead, and from lead to drug, and a G4 ligand has yet to reach clinical trial against a viral infection. The knowledge accumulated on G4 agents active against cancer should be precious, as some of the issues anticipated for a G4-against-virus strategy are shared for anticancer agents.

D. What Remains Untouched?

There are some untouched areas that may boost the significance of viral G4s in future with further experiments. For instance:

1. It remains to be investigated whether virulence and pathogenicity are related to the density of G4s in viral genome or not. Owing to the ability of G4s to shift viral cycle to lytic/latent phases, there could be a relationship with acuteness or chronicity of viral infections (Bohálová et al., 2021). According to results found for EBNA1 of EBV (Lista et al., 2017; Reznichenko et al., 2019), replication compartments of HSV-1 (Artusi et al., 2016), LANA of KSHV (Madireddy et al., 2016), and Sp1 binding-site region in HIV-1 LTR (Piekna-Przybylska et al., 2020), we can assume that G4s are more likely to switch the active phase to the latent phase, but the correlation of this latency with the duration of diseases (acuteness or chronicity) requires further analysis. Viral latency is involved in some poorly diagnosed or quiescent forms of cancers like cervical cancer caused by HPV. G4s are one of the pivotal players in viral latency, and their targeting can serve as both therapeutic and diagnostic goals.

2. Autophagy is a cellular mechanism that is mainly used to emancipate the cells from the junk products as well as external pathogens like viruses. There is some evidence that indicates the role of G4s in modulating autophagy (Beauvarlet et al., 2019a,b, 2020; Lejault et al., 2020). This modulation could impede viral elimination by possible drug molecules but has not yet been studied.

3. The role of G4s in chromatin packaging for viruses that integrate their genome in the host DNA requires further investigation. Also, the interaction between the G4s and epigenome is scantly investigated.

4. The safety of G4 ligands in animal models with viral infections is not assessed and demands scrutiny.

5. Should we selectively target the viral G4s or not? This is probably the hottest unexplored conundrum regarding the use of drugs that target G4s. The evidence seems to suggest that this kind of selectivity may not be necessary.

6. Some viruses seem to be devoid of stable G4 structures. Among the most notable examples of these viruses are measles virus, mumps virus, poliovirus, and hantavirus. Whether this lack of G4 occurrence is due to insufficient studies or is due to the real absence of these elements requires further explorations.

7. Higher-order multimerization of G4 structures is being investigated for genomic sequences. Some mutations have also been characterized to facilitate the formation of such higher-order structures (Kolesnikova et al., 2017). It would be interesting to study whether such dimeric or tetrameric assembly (multimerization) really exists for viral G4 motifs and to study whether they have some functional role in viral pathogenesis, which may be further used for therapeutic strategies.