The TLR3 group is composed of TLR3, TLR7, TLR8 and TLR9 ( Figure 1D ). To date, 10 human TLRs (TLR1-10) have been identified. These receptors are localized into the cell membrane and endosomes and are expressed in a wide range of cell types, including immune cells and non-immune cells. Among the TLRs found in endosomes, TLR3 recognizes dsRNA, TLR7 and TLR8 recognize ssRNA, and TLR9 recognizes CpG DNA ( Figure 2 ). Structurally, TLRs consist of an N-terminal ectodomain with leucine-rich repeat (LRR) motifs responsible for recognizing PAMPs/DAMPs (damage-associated molecular patterns), a transmembrane domain, and a C-terminal cytoplasmic Toll/IL-1R (TIR) domain ( Figure 1A ).

TLR3 primarily senses dsRNA viruses (31–34). Studies have found that 40–50 bp dsRNA can bind TLR3 and induce its dimerization, leading to basal activation (35), whereas a robust immune response requires longer dsRNA of at least 90 bp (36–39). Upon binding to dsRNA, TLR3 dimerizes and signals through the adapter protein TRIF (TIR-domain-containing adapter-inducing IFN-β, also called TICAM1). This TLR3-TRIF-mediated signaling predominantly activates IRF3/7 to induce the production of type I IFN, particularly IFN-β; while also, to a lesser extent, activates NF-κB and AP-1 to promote the production of cytokines (e.g., IL-6, IL-12, IL-1α, and CSF-3) (40). TLR3 also recognizes (+) ssRNA viruses, such as West Nile virus (WNV) (41), DENV (42), PV (32), SARS-CoV-2 (34), and dsDNA viruses, including murine cytomegalovirus (MCMV) (43) and HSV-1 (31, 44). The precise mechanisms by which TLR3 recognizes ssRNA or dsDNA viruses remain unclear, but it is likely that dsRNA structures, including secondary stem-loop formations in ssRNA viral genomes and dsRNA intermediates generated during viral replication (45, 46) activate TLR3. Similarly, many DNA viruses produce dsRNA intermediates during transcription, possibly arising from the bidirectional transcription within their genomes (46). Notably, TLR3 has been shown to recognize the stem-loop structures of PV (32) and the non-coding RNA transcripts EBERs from EBV, which are transcribed by RNA polymerase III (24). TLR3 activation is influenced by the topology, secondary, and tertiary structures of dsRNA regions (32). Therefore, it is plausible that TLR3 recognizes dsRNA structures derived from ssRNA or dsDNA viruses, initiating immune responses.

The TLR7 subfamily consists of TLR7, TLR8, and TLR9, which share highly homologous structures and activate similar antiviral mechanisms upon detecting single-stranded nucleic acids. These receptors contain a Z-loop located between LRR14 and LRR15 in the extracellular domains, which must be cleaved for activation. All three sensors utilize the adaptor MyD88 to relay signaling, leading to the activation of NF-κB, AP-1, and IRF3/5/7 (47). TLR7, TLR8, and TLR9 possess two binding sites that contribute to subtle differences in ligand recognition. TLR7 has one site for guanosine and its derivatives (e.g., 2',3'-cGMP) and uridine-containing ssRNA at another (48, 49). TLR8 recognizes uridine at one site and short ssRNA at another (50). Recent evidence demonstrated that endonuclease RNase T2 is important for TLR7 and TLR8 ligands (51, 52). Specifically, RNase T2 generates guanosine 2′,3′-cyclic monophosphate-terminated RNA fragments, and PLD exonuclease further releases the terminal 2′,3′-cyclic guanosine monophosphate (2′,3′-cGMP), which becomes the ligand for TLR7 (52). RNase T2 preferentially cleaves ssRNA between purine and uridine residues, which generates catabolic uridine as a TLR8 ligand (51). TLR9 binds two types of DNA: one site recognizes CpG DNA, while another recognizes 5′-xCx DNA (cytosine at the second position from the 5′ end) (53, 54). TLR7 and TLR8 recognize a wide variety of ssRNA viruses, including Coxsackie virus (CV), Encephalomyocarditis virus (EMCV), influenza A virus (IAV), measles virus (MV), Sendai virus (SeV), vesicular stomatitis virus (VSV), and SARS-CoV-2 (55, 56), activating antiviral immune responses.

The discovery that TLR9 recognizes unmethylated CpG bacterial DNA and initiates immune responses was first reported by Dr. Shizuo Akira’s group (57). For viral infections, TLR9 is not only limited to recognizing CpG ssDNA but also detects dsDNA viruses with unmethylated CpG motifs in their genomes, such as MCMV, adenovirus, and HSV-1/2 (43, 58–60). Prior to viral infection, TLR9 is expressed in the endoplasmic reticulum, upon infection, it is translocated into endosomes, where it binds CpG-containing viral DNA and signals through MyD88 to activate antiviral inflammatory pathways (59, 61). Moreover, BAD-LAMP regulates the trafficking of TLR9 to LAMP1⁺ late endosomes in human plasmacytoid dendritic cells, promoting NF-κB activation and the production of pro-inflammatory cytokines. In contrast, TLR9 can also signal from a distinct VAMP3⁺/LAMP2⁺/LAMP1^- endolysosomal compartment, which primarily enhances type I interferon production (62). In addition, two independent studies demonstrated that during viral infections, such as DENV or SARS-CoV-2, TLR9 recognizes mitochondrial (mt) DNA that is released into the cytosol, leading to NF-κB activation and an antiviral immune response (63, 64). Notably, the study on TLR9-mediated responses in DENV infection confirmed the colocalization of TLR9 and DNA in DENV-infected dendritic cells (DCs) (63).

The RIG-I group comprises RIG-I, MDA5, ZAP and LGP2 ( Figure 1D ). The RLR subfamily of PRRs includes retinoic acid-inducible gene I (RIG-I, also known as DDX58), melanoma differentiation-associated protein 5 (MDA5), and laboratory of genetics and physiology 2 (LGP2, also known as DHX58). These cytoplasmic RNA sensors share a central DExD/H-box helicase core domain, which contains a DECH motif, and a carboxy-terminal domain (CTD) responsible for foreign RNA substrate recognition. Both RIG-I and MDA5 feature two amino-terminal caspase activation and recruitment domains (CARDs) that mediate signal transduction to downstream effectors. In contrast, LGP2 lacks these CARDs but still binds RNA with high affinity, with a preference for blunt-ended dsRNAs over internal dsRNA regions or RNA overhangs (65) and acts as a regulator of RLR signaling (66). Despite structural homology ( Figure 1 ), the RLRs exhibit slight differences in their RNA recognition preferences.

The CTD of RIG-I detects viral RNA bearing 5′ triphosphate or 5′ diphosphate specifically (66), playing a crucial role in the antiviral response against RNA viruses, such as paramyxoviruses, influenza virus, and Japanese encephalitis virus (67). Upon infection, viral RNA interacts with RIG-I through the CTD, triggering a conformational change that exposes the N-terminal CARD, allowing RIG-I oligomerization and interaction with the adaptor protein mitochondrial anti-viral-signaling protein (MAVS). CARD-dependent interaction with MAVS activates transcription factors IRF3 and NF-κB, promoting the expression of type I IFNs and ISGs. The spatiotemporal regulation of RIG-I involves coordinated control of its activation, localization, and signaling dynamics. In the absence of infection, RIG-I resides in the cytoplasm in an auto-repressed conformation. Upon activation by viral RNA, RIG-I translocates to mitochondrial-associated membranes, where it interacts with the adaptor protein MAVS. This signaling is amplified through K63-linked ubiquitination by E3 ligases such as TRIM25 or Riplet, which stabilizes RIG-I’s association with MAVS (68). To prevent excessive inflammation, RIG-I activity is negatively regulated by several mechanisms, including deubiquitinating enzymes such as CYLD and USP21 (69, 70), autophagic degradation (71), and suppression by SOCS proteins (72). RIG-I recognizes short dsRNA (10–300 bp) with 5′-triphosphate or 5′-diphosphate ends. In contrast, MDA5 preferentially binds dsRNA longer than 1 kb without end specificity. Like RIG-I, MDA5 binds viral dsRNA through its CTD, oligomerizes to form filaments along the viral dsRNA, and then activates MAVS through its CARDs (73).

Initially, LGP2 was thought to act as a negative regulator, interfering with RIG-I and MDA5 RNA recognition (74). However, recent studies indicate that LGP2 serves as a positive regulator, enhancing viral RNA recognition and MDA5 activation (75, 76), while negatively regulating RIG-I (77, 78).The exact mechanisms behind this remain under investigation, but LGP2’s ATPase activity facilitate binding to diverse dsRNA species and works synergistically with MDA5 to improve antiviral efficiency (65, 76, 79). Further studies show that LGP2 is recruited to MDA5-dsRNA filaments, promoting filament formation along the dsRNA and stabilizing MDA5’s interaction with dsRNA during (+) ssRNA EMCV and dsRNA Bell Pepper Endornavirus (BPEV) infections in HEK293T cells (76). Additionally, LGP2 induces a conformational change in MDA5, exposing its CARDs completely, which activates MAVS and leads to the activation of immune response genes, including type I IFNs. Like RIG-I, MDA5 is a cytoplasmic sensor that interacts with MAVS on the outer mitochondrial membrane upon recognition of viral dsRNA. However, unlike RIG-I, MDA5 typically responds during the later stages of infection and is negatively regulated by factors such as LGP2 (80).

Zinc-finger antiviral protein (ZAP), also known as zinc-finger CCCH-type antiviral protein 1 (ZC3HAV1) or poly (ADP-ribose) polymerase family member 13 (PARP13), functions as a sensor for a variety of RNA and DNA viruses. Its N-terminus contains an RNA-binding domain (RBD) with four zinc finger domains (ZnF1-4), while the middle of the protein contains the fifth zinc finger (ZnF5) and two tandem WWE modules (WWE1 and WWE2) at the C-terminus ( Figure 1A ). The RBD is responsible for binding ssRNA, and the WWE2 pocket specifically binds poly (ADP-ribose) (PAR). ZAP was originally identified as an antiviral factor restricting MLV via a cDNA expression library screening in 2002 (81). Recent CLIP-seq and ZAP’s structure with RNA showed that ZAP interacts with CpG dinucleotides in RNA (82, 83). Consistent with its structure, ZAP is a sensor for a variety of RNA viruses, such as hepatitis E virus (HEV) (84), Ebola virus, chikungunya virus, and hepatitis B virus (HBV) (85, 86). However, DNA viruses, such as HBV (85), HCMV (87), and vaccinia virus (VACV) (88), have been reported as targets of ZAP as well. Similar to other sensors, ZAP exist in four isoforms: S, M, L, and XL, which have distinct antiviral roles (85, 89).

The cGAS group is made up of cGAS, OAS1, PKR and IFI16 ( Figure 1D ). Compared with other groups, four members of this group have longer functional distances, indicating their divergent functions. Human cyclic GMP-AMP synthase (cGAS) was discovered by Dr. Zhijian Chen’s lab a decade ago (90). It consists of 522 amino acids, with a long unstructured, positively charged N-terminal and a catalytic C-terminal ( Figure 1A ). The C-terminus contains two key domains: the nucleotidyltransferase (NTase) core domain, which is essential or cGAS enzymatic function, catalyzing GTP and ATP into cGAMP, and the Mab21 domain, which features a Zn ribbon motif [H(X₅)CC(X₆)C] that mediates DNA binding and ensures specificity for B-form dsDNA (91–93). The C-terminus has three DNA binding sites (A, B, and C) composed of positively charged amino acid residues that interact with negatively charged DNA backbone (92–95). Collectively, both the positively charged C- and N-termini of cGAS form multivalent interactions with the negatively charged DNA in a sequence-independent manner, promoting the formation of liquid-like droplets containing activated cGAS (96). As a DNA sensor, cGAS detects viral dsDNA from HSV-1 (97, 98) and KSHV (99) and produces cGAMP. Then, cGAMP activates the STING-TBK1-IRF3- and -NF-κB-mediated inflammatory pathways to induce the expression of type I IFNs and proinflammatory cytokines (97–100).

Initial, studies reported that cGAS is localized in the cytoplasm, where it detects viral dsDNA (98, 99). More recent findings have identified additional localizations, including the plasma membrane (101) and nucleus (97, 102–104). At the plasma membrane, cGAS is anchored to the inner leaflet via phosphatidylinositol 4,5-bisphosphate, keeping it in an inactive state (101). In the nucleus, cGAS is tethered to chromatin, also maintained in an inactive state (97, 102–105). Notably, this sequestration prevents the indiscriminate activation of cGAS by endogenous DNA while positioning it to optimally detect viral DNA (97, 101). Upon viral infection, cGAS is released, allowing it to recognize viral DNA and trigger initiate immune signaling (97, 101, 104).

cGAS also plays a critical role in RNA virus infections. In SARS-CoV-2 infection, cGAS binds to genomic dsDNA released from the nucleus or mtDNA due to infection-induced cellular stress, activating the STING-mediated pathway to trigger an antiviral response (106). In HIV-1 infection, cGAS detects primary reverse transcribed cDNA, antisense ssDNA with stem-loop structures, during early infection. This activates STING-mediated signaling, triggering an innate immune response (107–110).

Oligoadenylate synthetases (OASs) play key roles in the innate immune defense against viral dsRNA in the cytosol of vertebrates. In humans, there are four members in this family: OAS1, OAS2, OAS3, and OASL. Each of them contains a polymerase beta (pol-β)-like nucleotidyl transferase domain (also known as OAS domain), varies in the number of this domain. OAS1 has one pol-β domain (OAS domain I, DI), OAS2 has two pol-β domains (DI and DII), and OAS3 has three pol-β domains (DI, DII, and DIII). While OAS1–3 have one enzymatical active OAS domain, additional OAS domains in OAS2–3 remain inactive (DI in OAS2, DI and DII in OAS3). In contrast, OASL contains a single pol-β domain at its N-terminus and two consecutive ubiquitin-like domains in its C-terminus but lacks enzymatic activity entirely. Interestingly, OAS proteins do not contain a canonical dsRNA-binding domain or motif; however, dsRNA binding is achieved through electrostatic interactions and hydrogen bonds between the dsRNA and positively charged residues in a groove located on the surface of OAS proteins (111–114). OAS1 has a bilobular structure composed of N-terminal and C-terminal lobes, with a positively charged groove between them responsible for binding two minor grooves of dsRNA ≥18 bp, leading to a conformational change and activation of the protein (111, 112, 115). In contrast, the presence of additional inactive domains in the N-terminal region of OAS2 and OAS3 requires longer dsRNA for activation, with OAS2 (two OAS domains) binding to four minor grooves of dsRNA and requiring > 35 bp (114), while OAS3 (three OAS domains) requires dsRNA > 50 bp for activation (113, 115). Higher activation of OAS proteins is observed with longer dsRNA, OAS3 showing greater activation than OAS1 in response to equal amounts of dsRNA > 60 bp, highlighting the importance of dsRNA length in distinguishing the self from non-self target (114, 115). During viral infection, OAS1 recognizes viral dsRNA and catalyzes the production of 2′-5′-oligoadenylate, a secondary messenger that activates RNase L, an enzyme responsible for ssRNA cleavage ( Figure 2 ). Studies have shown that OAS1 binds dsRNA structures (stem-loops) in the 5′-UTR of SARS-CoV-2, effectively blocking viral replication (116, 117). Whereas an isoform of OAS1, p46, confers strong antiviral activity against picornaviruses, flaviviruses, and SARS-CoV-2 (118).

Protein kinase R (PKR) is a cytosolic dsRNA sensor that contains an N-terminal dsRNA binding domain with two dsRNA-binding motifs (dsRBM1 and dsRBM2) and a C-terminal kinase domain ( Figure 1A ). PKR detects viral dsRNA, including those from IAV, newcastle disease virus (NDV), enteroviruses (PV, EV70, CVB3, CVB5), SeV and MV (119–122), as well as endogenous dsRNA (i.e., non-coding (nc) RNAs, mtRNA transcripts) released into the cytoplasm (122, 123). Upon binding to dsRNA, PKR undergoes dimerization and phosphorylation, which in turn phosphorylates eIF2α, blocking the translation of viral mRNA and inhibiting viral replication (124). Previous studies have demonstrated PKR’s critical role in activating RLRs, such as RIG-I and MDA5, within stress granules (SGs), thereby facilitating antiviral responses during infections with IAV, NDV, PV, and SeV (119–121, 125). Based on these findings, SGs were considered essential for PKR and RLR activation, as their impairment or absence significantly reduced the activation of these viral sensors and subsequent antiviral responses. However, recent studies suggest that PKR and RLR activation and localization are independent of SGs (122, 126, 127). One study revealed that rather than serving as sites for viral RNA sensing and PKR/RLR activation, SGs regulate antiviral responses by preventing excessive activation of RIG-I, PKR and OAS signaling, as well as apoptosis, while also intrinsically restrict viral replication and spread in a RLR/PKR-independent manner (127). Moreover, other studies have indicated that PKR does not localize to SGs or processing bodies (P-bodies) but forms distinct autonomous clusters in response to dsRNA stress and interacts with dsRNA-binding proteins (122, 126). Notably, dsRNA stress induces the formation of dsRNA-induced foci (dRIF), which contain dsRNA, PKR, and various dsRNA-binding proteins, including ADAR1, Stau1, DHX9, NLRP1, and protein activator of PKR, with the timing and localization of dRIF formation strongly correlating with PKR activation (126). To prevent aberrant PKR activation, a recent CRISPR screening revealed that PACT cooperates with ADAR1 to suppress PKR activation from self-dsRNAs in uninfected cells (128).

IFI16 (Interferon Gamma Inducible Protein 16) contains an N-terminal PYD domain and two DNA-binding HIN domains, known as HINa and HINb ( Figure 1A ). It is located in the nucleus and serves as a DNA sensor to detect viral DNA (129–133). Upon binding viral DNA through its HIN domains, IFI16 activates the STING-mediated inflammatory pathway, leading to the induction of type I IFNs and proinflammatory cytokines (133–135). In addition to inducing IFN signaling, IFI16 plays a role in epigenetic regulation. Studies highlight that IFI16 binds to HSV-1 gene promoter regions, promoting the formation of repressive heterochromatin (marked by H3K9me3), thereby suppressing HSV-1 gene transcription and replication (129–131, 135). Moreover, IFI16 binds to these promoters and inhibits the recruitment of RNA Pol II and transcription factors, such as TBP and Oct1, further suppressing HSV-1 gene transcription (130). Besides its role in sensing DNA viruses, IFI16 is implicated in recognizing RNA virus, such as IAV (136) and HIV-1 (137). A recent study demonstrated that IFI16 directly binds IAV (A Puerto Rico/8/1934 H1N1, PR8) via its HINa domain, leading to enhanced RIG-I transcription and interaction with RIG-I through its PYD domain, creating a synergistic antiviral response (136).

The NLRP1 group consists of NLRP1, DAI, NLRP6, and AIM2 ( Figure 1D ). NLRP1 and NLRP6 share similar structural organization, including a central nucleotide-binding and oligomerization (NACHT) domain flanked by an N-terminal pyrin (PYD) domain and C-terminal LRRs. Both proteins function as viral sensors, though they recognize different viral components. NLRP1 detects pathogen-derived enzymes, such as proteases, and senses dsRNA. In contrast, NLRP6 primarily detects viral dsRNA.

NLRP1 senses the viral RNA of Semliki Forest virus. Mechanistically, it binds dsRNA through its LRR domain, leading to a conformational rearrangement that facilitates immune activation (138). NLRP6 interacts with dsRNA directly and undergoes liquid liquid phase separation both in vitro and in cells. This phase separation is crucial for activation of the NLRP6 inflammasome, a process further supported by investigation in mouse models infected with mouse hepatitis virus or rotavirus (139). Notably, the human genome encodes 14 distinct NLRP family members (NLRP1–14), many of which play key roles in inflammasome activation.

DNA-dependent activator of interferon-regulatory factors (DAI), also known as ZBP1 or DLM-1, is a critical nucleic acid sensor in the innate immune system. Originally identified as a cytosolic DNA sensor, DAI detects foreign DNA and activates type I IFN production, along with other innate immune responses (140). Mechanistically, DAI binds dsDNA through its Z-DNA binding domains, facilitating the recruitment of IRF3 and TBK1 to enhance immune signaling. Interestingly, recent studies have highlighted DAI’s role as an RNA virus sensor. It was found to recognize IAV genomic RNA (Z-RNA), binding to RIPK3 and recruiting MLKL and RIPK1, ultimately activating cell death pathways (141). In addition, DAI can interact with RNA transcripts derived from DNA viruses like murine cytomegalovirus (MCMV) and VACV. This interaction triggers necroptosis through the RIPK3 pathway (142, 143).

AIM2 (Absent in melanoma 2) consists of a single PYD domain and a single HIN domain ( Figure 1A ). As a cytosolic innate immune receptor, it recognizes both microbial and self-dsDNA in the cytoplasm. AIM2 can recognize a variety of dsDNA viruses, such as VACV (144, 145), MCMV (144), EBV (146), and human cytomegalovirus (HCMV) (147). During viral infection, the positively charged HIN domain of AIM2 interacts with the negatively charged dsDNA (≥ 80 bp in length) in a sequence-independent manner, through hydrogen bonds at both the major and minor grooves, leading to the formation of AIM2-ASC-caspase-1 inflammasomes (148). This assembled inflammasome cleaves pro-caspase-1 into active caspase-1, which matures proinflammatory cytokines IL-1β and IL-18 and cleaves gasdermin D (GSDMD), forming pores in the cell membrane that induce pyroptosis (149, 150). As a result, pyroptotic cell death releases cytosolic contents, including IL-1β and IL18 into the extracellular space, amplifying the inflammatory response (149, 151, 152). Notably, AIM2 is also implicated in antiviral responses against RNA viruses, such as enterovirus (EV-A71) (153), IAV (PR8) (154) and SARS-CoV-2 (155).The precise mechanism of how AIM2 recognizes RNA viruses remain unclear, However, researchers propose that mtDNA release or the uptake of self-DNA from dead cells may contribute to AIM2 activation (154, 155).

The DHX36 group have DHX9, DHX36 and Dicer1 ( Figure 1D ), which all contain a helicase domain ( Figure 1A ). DHX9 (also known as RHA) and DHX36 (RHAU or G4R1) are DExD/H-box helicases characterized by the conserved DExD/H motif, specifically the DEIH motif for both. They are localized in both the nucleus and the cytoplasm (156, 157), where they regulate mRNA translation by unwinding G-quadruplex (G4) RNA structures in the 5' UTR of mRNA (158, 159). In addition to G4-RNA, both helicases unwind G4-DNA and other non-canonical nucleic acid structures, such as R-loops and D-loops (160–162). Structurally, DHX9 and DHX36 share a common core helicase domain of two RecA-like domains (RecA1 and RecA2) and an oligonucleotide/oligosaccharide-binding (OB)-like fold in the C-terminus ( Figure 1A ). However, there are slight differences in their N- and C-terminal regions and binding preferences. DHX9 has two dsRNA-binding domains (dsRBD1 and dsRBD2) in its N-terminus, while its C-terminus features a HA2 domain and glycine rich (G-patch) region, which enhances RNA duplex unwinding (163, 164). In contrast, DHX36 possesses a glycine-rich element followed by the DHX36-specific motif (DSM) in its N-terminus, which is critical for high affinity binding to G4 structures in DNA and RNA, as well as RNA duplexes (161, 162). Functionally, DHX9 preferentially binds and unwinds G4-RNA, R-loops, G4-DNA, D-loops, and RNA forks in descending efficiency, favoring RNA-containing duplexes with 3’ overhang tails (164); whereas DHX36 exhibits a broader preference for G4-RNA, G4-DNA, and RNA duplexes (161, 162).

Besides their roles in nucleic acid metabolism, DHX9 and DHX36 function as cytosolic viral sensors. They recognize a range of viral nucleic acids, including single-stranded CpG ODNs and dsDNA viruses such as HSV-1 and MHV (Murine Hepatitis Virus)-68 (165, 166). They also detect dsRNA viruses (e.g., reovirus) and ssRNA viruses (e.g., NDV, IAV) (120, 167), leading to the induction of type I IFNs and proinflammatory cytokines. The mechanisms underlying DHX9 and DHX36 viral recognition are well elucidated. The helicase core of DHX9 (RecA domains, HA domain, and OB-like fold) binds ssRNA (163, 164) and interacts with CpG ODNs via its OB-like fold (165). Its N-terminal RBDs recognize dsRNA structures in RNA viruses, further amplifying immune signaling (163, 164, 167). Similarly, DHX36 binds ssDNA/ssRNA through its helicase core and C-terminal domains (161), with CpG ODNs interacting with its DEAH motif (165). The conserved 5’-β-hairpin region within the helicase core and the OB-like fold in the C-terminus facilitate recognition of RNA duplexes and G4 structures (162). Additionally, the DSM motif in the N-terminus and the OB-like fold of DHX36 enhances binding to G4 structures, RNA duplexes (161, 162), and dsRNA forms of ssRNA viruses (120). Multiple studies have reported the presence of G4 structures in the viral genomes and mRNA transcripts (168–172). This raises the possibility that viral G4 structures may activate DHX9 and DHX36 during infection. Though further studies are needed to elucidate the mechanisms of G4-mediated immune responses by these helicases.

Human Dicer1 is a large protein (~219 kDa), mainly known as an endoribonuclease, not a helicase though containing a helicase domain. It is found in both the nucleus and cytoplasm. During viral infection, Dicer1 processes viral dsRNA or viral pre-microRNAs (pre-miRNAs) into small interfering RNAs (siRNAs) or miRNA (173, 174), which then bind to target viral mRNA or viral RNA genomes, promoting RNA interference (RNAi) (175). Dicer1 is crucial for processing dsRNA substrates into small RNAs to promote RISC-mediated RNAi and this process has been well established in numerous studies. Studies have shown that Dicer1 is responsible for the biogenesis of miRNAs and siRNAs derived from viral dsRNA, including those from HIV-1 (176), IAV/delNS1 (177), and EV-A71 with defective 3A protein (178), as well as dsRNA hybrid duplexes of HIV-1 genome and human tRNA (179). These small RNAs, incorporated into the RISC complex, bind viral mRNA or the sense strand of viral RNA genomes, to suppress viral gene expression and inhibit replication. Recently, the Reis e Sousa group identified a naturally occurring, alternatively spliced isoform of human Dicer1, named antiviral Dicer (aviD), which lacks the Hel2i domain due to the absence of exons 7 and 8 (180). Previous studies from other groups revealed that while the N-terminal helicase domain of Dicer negatively regulates its ability to process dsRNAs, its absence in mutant forms significantly enhances the processing of dsRNA substrates (181, 182). Consistent with these findings, the study demonstrated that aviD significantly enhances the processing of dsRNA substrates from SARS-CoV-2 and Zika viruses (ZIKV), generating higher levels of compared to canonical Dicer1 while retaining comparable miRNA production (180). This ultimately leads to improved RNAi in human and mouse stem cells. Not all small RNAs suppress viral replication through RNAi, but they can impede antiviral responses in humans. For instance, during SARS-CoV-2 infection, Dicer1 processes the stem‐loop structure of ORF7a to produce two miRNAs, CoV2-miR-07a.1 and CoV-2-miR-07a.2, in infected human cells (183). However, these miRNAs suppress host innate immune responses by binding to the 3′ UTR of ISGs and facilitating RISC-mediated degradation of these genes in infected cells.

The Ku70 group has Ku70 and MRE11 ( Figure 1D ), two well-known proteins involved in DNA double-stranded breaks (DSB) repair. Human Ku70 is a 609 amino acid nuclear protein consisting of three domains: an N-terminal von Willebrand factor domain, a β-barrel DNA-binding domain in the core, and a C-terminal SAP domain ( Figure 1A ). It exists as a heterodimer with Ku80, a complex that is essential for the stability of both proteins. Together, the Ku70/Ku80 heterodimer associates with DNA-dependent protein kinase catalytic subunit (DNA-PKcs) to form the holoenzyme DNA-PK. As a key component of DNA-PK, the Ku70/80 dimer plays a critical role in repairing DSB through non-homologous end joining (NHEJ). In addition to its role in DNA repair, Ku70 also functions as a cytosolic viral DNA sensor, activating innate immune responses against DNA viruses. Upon viral DNA infection, such as HSV-1 (184), HSV-2 (185), and HBV (186), Ku70 translocates from the nucleus to the cytoplasm, where it activates the STING-mediated signaling pathway. This leads to the activation of IRF1, IRF3, and IRF7, promoting the production of type I (IFN-α/β) and type III (IFN-λ1) IFNs (184–187). Further investigation revealed that acetylation in Ku70’s NLS domain promotes cytoplasmic translocation, enhancing the production of IFN-λ1 and inhibiting viral replication (184). Ku70 is also involved in the recognition of RNA viruses. One study revealed that Ku70 recognizes (+) ssRNA retrovirus human T-cell lymphotropic virus 1 (HTLV-1), leading to STING activation and the phosphorylation of NF-κB and IRF3, subsequently inducing an antiviral response (188).

MRE11 is best known for its role in DNA repair as part of the MRE11-RAD50-NBS1 (MRN) complex, which detects and initiates DSB repair. However, it also functions as a cytosolic exogenous dsDNA sensor by recognizing sequences such as HSV-1 and IFN stimulatory DNA, and stimulates type I interferon by regulating STING trafficking (189). More recently, it was reported that MRE11 is essential for cGAS-STING activation (190). Mechanically, cGAS is constitutively inhibited by high-affinity binding to the nuclear chromatin histone H2A-H2B acidic patch (AP) region on the nucleosome disk face that prevents its oligomerization and activation in response to dsDNA (91, 102, 105, 191–194), now MRN complex releases cGAS from nucleosomal AP surfaces and enables cGAS mobilization, cytoplasmic relocalization and activation by dsDNA, either from pathogens or host itself.

IFIX, another PYHIN family member, contains an N-terminal PYD domain and a single DNA-binding HIN domain. Though similar to IFI16 structurally, IFIX has different biological functions and is often associated with promyelocytic leukemia (PML) nuclear bodies. It is primarily localized in the nucleus but can translocate to the cytoplasm under specific conditions, such as during viral infection or in response to specific cellular signals (195, 196). This translocation is regulated by the acetylation of its NLS motifs, with acetylation at K138 in the NLS2 motif promoting shuffling from the nucleus to the cytosol during viral infection (196). Like other PYHIN/ALR family members, IFIX recognizes DNA viruses such as HSV-1 and VACV, binding dsDNA through its HIN domain (195), eliciting an antiviral response (195, 196). Proteomic studies using AP-MS revealed that IFIX interacts with components of PML bodies and DNA damage response effectors, displaying its role in antiviral defense (195). Additionally, nuclear and cytoplasmic proteomic analysis during HSV-1 and VACV infections suggest that IFIX upregulates proteins involved in immune signaling and responses (196).

Leucine-Rich Repeat Flightless-Interacting Protein 1 (LRRFIP1), also known as GC-binding factor 2 (GCF2), is a multifunctional protein involved in transcriptional repression, cytoskeletal organization, signal transduction, and immune responses. It consists of three domains: an N-terminal helix domain, a central coiled-coil domain and a C-terminal nucleic acid binding domain ( Figure 1A ). LRRFIP1 functions as a nucleic acid sensor, recognizing dsRNA from vesicular stomatitis virus and dsDNA from Listeria monocytogenes (197). All three domains are believed to contribute to its nucleic acid-binding ability (198). A recent study has shown that reduced LRRFIP1 protein level correlates with the severity of SARS-CoV-2 infection and LRRFIP1 suppresses SARS-CoV-2 infection in a type I IFN-independent manner (199). Another study found that patients with severe COVID-19 exhibit increased alternative splicing of the LRRFIP1 transcript, resulting in reduced levels of functional protein and increased levels of truncated isoforms, suggesting a correlation between perturbed LRRFIP1 splicing and the severity of SARS-COV-2 infection (200). However, the precise mechanisms by which LRRFIP1 suppresses SARS-CoV-2 infection remain unknown.

Human hnRNPA2B1 is a small yet significant protein. It consists of 353 amino acids and belongs to the A/B subfamily of ubiquitously expressed heterogeneous nuclear ribonucleoproteins (hnRNPs), a family well known for its roles in RNA processing. Structurally, hnRNPA2B1 contains two RNA recognition motifs (RRM1 and RRM2) in its N-terminus, along with an RGG box and a NLS at its C-terminus ( Figure 1A ). Dr. Xuetao Cao’s group used biotinylated genomic DNA of HSV-1, coupled with 2D SDS–PAGE and mass spectrometry, and identified hnRNPA2B1 as a sensor for HSV-1 DNA (201). Further analysis revealed that two RRM domains are responsible for binding HSV-1 DNA. This discovery suggests a broader role for hnRNPA2B1 in nucleic acid sensing and immune signaling.

DDX41 is a member of the DEAD-box family helicases within Superfamily 2 (SF2) and contains the conserved DEAD (Asp-Glu-Ala-Asp) sequence in motif II. Different from DEIH-box helicase DHX9 and DHX36, DDX41 exists as a separate cluster. Predominantly located in the nucleus, DDX41 plays versatile roles in RNA metabolism and the innate immune response. Notably, DDX41 serves as a PRR, detecting the dsDNA from HSV-1 (202, 203) and RNA: DNA hybrids derived from the retrovirus murine leukemia virus (MLV) (204).

The molecular mechanisms underlying DDX41’s recognition of viral DNA and its initiation of antiviral responses have been extensively studied. Research by Liu’s group demonstrated that motifs I (Walker A) and II (Walker B, containing the DEAD sequence) in DDX41 are essential for viral DNA recognition and for its interaction with STING, initiating type I IFN production (202). Following the binding of pathogenic dsDNA or RNA: DNA hybrids, DDX41 translocates from the nucleus to the cytosol, where it colocalizes with STING and activates the STING-TBK1-IRF3 and NF-κB pathways, leading to the induction of type I IFNs and proinflammatory cytokines (202–205). Recent findings from our group have confirmed that DDX41 acts as a viral sensor upstream of cGAS, triggering cGAS-STING-TBK1 to induce type I IFN production in response to dsDNA viruses, such as HSV-1 (203). Though these findings highlight the importance of DDX41 in antiviral signaling, the exact location(s) of viral DNA recognition by DDX41 remains unclear.

The 5’ cap is a modified guanosine, with a methyl group added to the 7-position (7-methylguanosine, m7G) at the 5’ end of RNA. The 5’ cap is a key feature found in eukaryotic mRNAs that enhances their stability and translation efficiency. Many viruses adopt this 5’ cap for their RNAs to mimic host mRNA, thereby evading detection by host PRRs like RIG-I. Notable examples include SARS-CoV-2 (213), influenza virus (214), and flaviviruses like DENV (215) and WNV (216).

Some viruses methylate their RNA to prevent recognition by host sensors. For example, SARS-CoV-2 uses its nsp16 protein to methylate the 2′-O position of the ribose in the first nucleotide of its RNA, thereby helping the virus evade the host immune response (217). Similarly, the N6-methyladenosine (m⁶A) modification in HBV and hepatitis C virus (HCV) transcripts prevents recognition by RIG-I (218).

Viruses can change their RNA sequences through processes like adenosine-to-inosine (A-to-I) editing. This editing can alter the viral RNA in a way that reduces its recognition by host sensors. Many viruses, including Hepatitis Delta virus (HDV) and Measles virus (219), employ this strategy.

The NSP15 protein of SARS-CoV-2 cleaves viral RNA at specific sites, particularly at uridine residues, thereby preventing dsRNAs accumulation, which otherwise be recognized by sensors like MDA5, OAS, and PKR (220).

ZIKV and DENV form RNase L-induced bodies that promote the degradation of viral RNA (221).

Notably, DNA viruses do not typically use DNA editing mechanisms in the same way that RNA viruses do.