Fish TLR5 develops a taste for viral RNA.

These days not only humans but also freshwater fish battle with infections by RNA viruses (Zou & Nie, 2017). This observation prompted Liao et al to turn their attention to viral recognition in Grass carp (Ctenopharyngodon idella), the most important cultivated freshwater fish with 5.7 million tons and 13 billion USD in fishery exports per year (FAO, 2021). Grass carp and other freshwater fish, such as the model organism Danio rerio (zebrafish), have a sophisticated innate immune system that helps them to detect microbial and viral pathogens by employing a variety of pattern recognition receptors (PRRs; Zou & Nie, 2017). Toll-like receptors (TLRs) are one class of PRRs that detect microbe-associated molecular patterns (MAMPs), such as flagellin or viral double-stranded (ds)RNA. In mammals, TLR3 is specialized in sensing viral dsRNA (Liu et al, 2008), while TLR5 recognizes the MAMP flagellin (Yoon et al, 2012; Fig 1). The well-established notion of TLR5 as a purely "bacterial" flagellin TLR has now been challenged by Liao et al in this issue of EMBO Reports (Liao et al, 2022). The authors' intriguing and unexpected results indicate that fish TLR5 is involved in viral recognition, a function lost in mammals, and shed light on hitherto inexplicable links of mammalian TLR5 to antiviral immune signaling.

Teleost fish have undergone whole genome duplication and grass carp encodes two tandem TLR5 genes, CiTLR5a and CiTLR5b. Liao et al (2022) started by comparing the sensitivity of the two CiTLR5 paralogs to different MAMPs. Various flagellins or flagellated bacteria were sensed by CiTLR5a and CiTLR5b heterodimers for activation of NF‐κB, a transcription factor regulating the expression of pro‐inflammatory genes in both fish and mammals. Very surprisingly, they found that CiTLR5b can also function as a homodimer to signal to Interferon‐responsive elements (ISRE) in response to a control MAMP, the dsRNA analog poly(I:C), or to viral infection of Ctenopharyngodon idella cells (Fig 1). Notably, the CiTLR5 ectodomains directly bound poly(I:C). Through a comprehensive series of domain swaps and mutagenesis, the authors showed that flagellin recognition requires canonical TLR5 recognition sites in both CiTLR5a and CiTLR5b. However, poly(I:C) binding sites differed between the TLR5 paralogs. These differences in ligand‐receptor interactions correlated well with their surprising finding that in vitro the response of CiTLR5b to poly(I:C) was blocked by co‐expression of CiTLR5a. Interestingly, in fish tissues CiTLR5a and 5b are co‐expressed; however, 5b usually exceeds the mRNA expression of 5a by a factor of ~2–10 in most analyzed C. idella tissues, suggesting that, at steady‐state levels, viral dsRNA sensing is operational. Indeed, viral challenge in live grass carp induced an antiviral transcriptional program that was dampened by RNA interference with CiTLR5b expression, confirming that CiTLR5b is involved in viral sensing in vivo. Although the structural framework for dsRNA vs. flagellin binding and NF‐κB vs. ISRE signaling requires further exploration, Laio et al’s meticulous and comprehensive study does show that fish TLR5 is a bona fide viral sensor.

Figure 1

Viral sensing is a unique function of fish TLR5

In teleosts, TLR5 proteins can be involved in both flagellin and dsRNA sensing. Specifically, TLR5a‐TLR5b heterodimers recognize flagellin and induce NF‐κB‐dependent cytokine genes. However, TLR5b homodimers are able to respond to dsRNA to trigger antiviral transcriptional programs. Conversely, in mammals bacterial flagellin and viral dsRNA sensing are sensed separately by TLR5 and TLR3, respectively, and trigger largely distinct transcriptional profiles. A possible reminiscence of an antiviral role for mammalian TLR5 is the regulation of IL‐22 upon flagellin engagement. The structural basis for these similarities and differences both at the ligand engagement and intracellular TIR domain remains elusive.

The study immediately extends its scope to another teleost, zebrafish, and also explores the intriguing possibility that mammalian TLR5 may respond to dsRNA. In fact, responsiveness to dsRNA appears a unique capacity of teleost TLR5: Whereas teleost tandem TLR5s can sense flagellin via TLR5a/b heterodimers and dsRNA via a TLR5b homodimer, mammalian TLR5 seems confined to sense only flagellin in a homodimeric manner (Fig 1). Nevertheless, the study of Liao et al has important implications for our understanding of mammalian TLR5 function.

Firstly, when it comes to the structural connotations of TLR5 sensing, most of what we know about flagellin‐TLR5 recognition in mammals is based on the structure of a truncated zebrafish DrTLR5b homodimer in complex with Salmonella flagellin (Yoon et al, 2012). As insightful as this structure has been, this present study and another (Voogdt et al, 2018) remind us that in fish TLR5 homodimers cannot form an active flagellin‐sensing complex, suggesting that the 2012 structure by Yoon et al is most likely that of an inactive flagellin‐TLR5 complex. Therefore, conclusions drawn over the years about the human sensing of flagellin have to be treated with caution and require experimental confirmation by a human TLR5‐flagellin complex. En route, the precise details of the supposedly active structures of a Ci (or Dr)TLR5a/b heterodimer in complex with flagellin and a Ci (or Dr)TLR5b homodimer in complex with poly(I:C) would be very informative and exciting to compare. In the meantime, it may prove insightful to compare the modeling studies offered by Liao et al with the first reported TLR structure of human TLR3 binding the same ligand (Liu et al, 2008). Furthermore, Liao's work raises the possibility that Ci/DrTLR5b may exist in two different “ON” states, one signaling toward NF‐κB together with Ci/DrTLR5a upon flagellin engagement and another signaling toward ISRE in response to poly(I:C). Further comparative studies of intracellular Ci/Dr vs. human TLR5(a/b) and TLR3 Toll/Interleukin‐1 receptor(TIR) signal transduction domains may thus prove insightful.

Secondly, the study begs the intriguing question whether mammalian TLR5 still has a connection to antiviral immune responses, that is, is there a “reminiscence of antiviral function” in human TLR5? CiTLR5b stimulation with dsRNA or viral infection triggered the transcription of interferon (IFN) 1, IFN3 and/or interleukin 22 (IL22) in vitro as well as in vivo (Fig 1). The antiviral properties of IFNs in humans are textbook knowledge, but human IL‐22 was also shown to have antiviral properties (Das et al, 2020) and to display similarity to fish IL‐22 (Siupka et al, 2014). Liao et al now help us to make more sense of reports linking TLR5 activity with antiviral properties in humans: for example, hTLR5‐induced IL‐22 restricted rotavirus infection in the intestine (Zhang et al, 2020), and flagellin reduced influenza A virus replication independently of type I IFN and IL‐2, thereby improving the efficacy of the antiviral drug, oseltamivir (Georgel et al, 2019).

Collectively, this study uncovers fascinating and completely unexpected facets of TLR5 sensing that have relevance beyond fish. Importantly, it also re‐opens many questions that appeared settled and warrants exciting structural and biochemical work that could provide interesting new insights into the human immune system.

Disclosure and competing interests statement

The author declare that they have no conflict of interest.