Interferons (IFNs) are a group of secreted proteins that play key roles in host antiviral immunity. IFNs are typically induced by the activation of host pattern-recognition receptors (PRRs), mainly RIG-I-like receptors (RLR) and toll-like receptors (TLR), during viral infection (1, 2). To date, three types of IFNs (I, II, and III) have been identified. Type I IFNs (referred to as IFNs in this review) compose the largest IFN family (3). Type II IFNs comprise only IFN-γ, which is unrelated to type I IFNs because it uses different receptors and is encoded by a different chromosomal locus (3, 4). Type III IFNs were recently discovered and comprise IFN-λ1, IFN-λ2, IFN-λ3, and IFN-λ4 (4, 5). IFN-λ signals through a unique receptor but activates the same pathway as that of type I IFNs (4, 6, 7). The classification of different IFN types along with their corresponding receptors is summarized in Table 1.

All types of IFNs are capable to activate the Janus kinase/signal transducer and activator of transcription (JAK/STAT) pathway. In this review, recent progress of canonical or non-canonical activation JAK/STAT pathway, viral antagonism of the JAK/STAT pathway, removing of the JAK/STAT antagonist from viral genome for virus attenuation, and the potential pathogenesis roles of tyrosine phosphorylation-independent non-canonical STATs activation during virus infection are discussed in detail to provide new insight to understand the interplay between JAK/STAT signaling and viral antagonism.

Like other cytokines, IFNs bind their receptors and lead to the activation of certain signaling pathways, mainly the JAK/STAT pathway (8). Based on analysis of the flanking 5′-regions of genes locus in response to IFN, an IFN-stimulated response element (ISRE) was identified (9, 10). Probing the lysate of IFN-treated cells using ISREs, a cellular factor (named ISGF3) consisting of four preexisting proteins with sizes of 48, 84, 91, and 113 kDa was identified (10). Three of the four proteins with sizes of 91, 84, and 113 kDa were thought to belong to the same protein family and are currently known as STAT1α, STAT1β, and STAT2, respectively. The 48-kDa protein was later renamed interferon regulatory factor 9 (IRF9) (11). The size differences of STAT1α and STAT1β are due to alternative splicing of the same gene product (12). Moreover, the involvement of a kinase in IFN-α-induced IFN-stimulated gene (ISG) expression led to the discovery of the JAK family and STAT phosphorylation (13).

There are four members of the JAK family: JAK1, JAK2, JAK3, and tyrosine kinase 2 (TYK2). They are all characterized as having a C-terminal catalytic domain and a related, but enzymatically inactive, pseudo-kinase or kinase-like domain (14). They also share sequence similarity in five additional domains in the N-terminal region (15). The seven domains are now called Janus homology domains (JHD) 7 to 1 from the N- to C-terminal region of the JAKs (15). The four JHDs in N-terminal regions of JAKs (JHD7 to JHD4) are also called band 4.1 domains (C-terminus of JAKs) due to their homology to the band 4.1/ezrin–radixin–moesin protein family (16). The band 4.1 domains and Src homology 2 (SH2) domain (JHD5) are responsible for receptor binding (17). The pseudo-kinase domain (JHD2) is thought to regulate the kinase activity of JAKs via an interaction with the kinase domain (JHD1) (18).

The STAT family in mammalian hosts has seven members: STAT1, 2, 3, 4, 5A, 5B, and 6 (19). Sequence analysis shows remarkable similarity among the STAT genes, with the exception of STAT2 (19). All STATs share a very similar structure: an N-terminal domain, a coiled-coil domain, a DNA-binding domain (DBD), a linker domain, an SH2 domain, and a transactivation domain (TAD) (20). Schematic illustration of JAKs and STAT protein structure is listed as Figure 1. However, isoforms resulting from similar patterns of RNA splicing or protein proteolytic processing were reported for all STATs except STAT2 (21–24). These STAT isoforms lack the C-terminal TAD domain, which implies a regulatory role for STAT activation (22, 25, 26).

Janus kinases are generally non-covalently associated with the cytoplasmic tail of specific receptors. Upon cytokine binding, receptor dimerization or oligomerization leads to JAK apposition and autophosphorylation on tyrosine residues, releasing their intrinsic catalytic activity. Tyrosine phosphorylation of cytokine-receptor cytoplasmic domains by activated JAKs then provides binding sites for the SH2 domains of the STAT proteins. The STATs are then recruited to the JAKs, whereupon they are phosphorylated at a tyrosine residue (near residue 700 of their 750–850 aa-long sequence) (15). Upon activation, STAT/STAT interactions occur immediately through reciprocal SH2 interactions (27). All STATs are able to form homodimers, and the formation of STAT heterodimers is different depending on cytokines activating upstream signaling (15, 28).

STAT1 and STAT2 are the major players in type I IFN-mediated signaling (28). Other complexes induced by type I IFNs include STAT1–STAT1, STAT3–STAT3, STAT4–STAT4, STAT5–STAT5, and STAT6–STAT6 homodimers as well as STAT1–STAT3, STAT1–STAT4, STAT1–STAT5, STAT2–STAT3, and STAT5–STAT6 heterodimers (28). Karyopherin α1 (KPNA1) is the essential importin for the nuclear transport of phosphorylated STAT1 (29). STAT1 possesses a non-classical NLS, and KPNA1 binds between two STAT1 monomers, with two major binding determinants in the SH2 and DBDs (30). In addition to mediating IFN signaling, STATs are also responsible for transducing signals for several families of cytokines (Table 2).

The phosphorylation of tyrosine residues (near residue 700) in STATs is generally considered an essential step in the canonical activation of the JAK/STAT pathway induced by IFN or other cytokines (31). However, STATs can also be phosphorylated on serine residues in the C-terminal TAD (31). Initially, phosphorylation of serine residue in the TAD domain was considered to contribute to the maximal transcriptional activity of STAT in addition to tyrosine-dependent STAT activation (32). However, in recent years, STATs without tyrosine phosphorylation have been found to undergo continuous nuclear import/export and to contribute to alternative gene expression as non-canonical STAT activation (33, 34). It was shown that EBV the can specifically promote the expression of several ISGs including STAT1 without stimulating IFN induction or JAK/STAT activation but depends its early lytic nuclear protein SM protein (35). This observation challenges the canonical model of STATs activation, which generally views expression of ISGs as a consequence of STATs activation requires tyrosine phosphorylation.

Moreover, it has been demonstrated that unphosphorylated STAT1 and STAT2 with IRF9 can form unphosphorylated ISGF3 (U-ISGF3) (36). U-ISGF3 formation requires high levels of IRF9, STAT1, and STAT2 without tyrosine phosphorylation, and U-ISGF3 could also be induced by low level IFN-β. It was proposed that phosphorylated ISGF3 drives the first rapid-response phase, while U-ISGF3 drives the second prolonged response by binding to distinct ISREs, which are different from the ISREs in the rapid phase (36). Moreover, recent reports also demonstrated that U-ISGF3 drives the constitutive expression of ISGs to protect against viral infection under homeostatic conditions (37, 38). In addition to U-STAT1 and U-ISGF3, tyrosine unphosphorylated form of STAT as transcription activator has been reported for other STATs as well and is proposed to play roles in cytokine signaling, cell proliferation, hematopoietic differentiation, and cancer prognosis (34, 39–43). Moreover, it appears that unphosphorylated STATs other than STAT1/2 can form homodimers or heterodimers (41).

In addition to unphosphorylated STATs, mono-phosphorylation of the serine residues of different STATs has been reported (as serine phosphorylation of the TAD without tyrosine phosphorylation) in recent years (31, 44), which represents a novel non-canonical pathway of STAT activation (45). Unlike JAKs inducing tyrosine phosphorylation, the kinase involved in the serine phosphorylation of STATs is still unclear and might be involve the p38 MAPK pathway or occur via ERK or cyclin-dependent kinase 8 (CDK8) (31, 46–48). Serine mono-phosphorylation of STAT1 has been investigated more often than that of other STATs (49). In a mouse model of bacterial infection, a modest gain-of-function in antibacterial immunity was found in a STAT1Y701F mutant compared with that in Stat1^−/− mice (33, 49), suggested that serine mono-phosphorylated STAT1 at S727 site might contribute the partial restoration of antibacterial immunity in STAT1Y701F mutated mice. Notably, a STAT1Y701F mutant partially retained NK cell cytotoxicity, in contrast to a complete loss in Stat1^−/− mice. However, the NK maturation defect in the STAT1Y701F mutant was similar to that found in Stat1^−/− mice. A single mutation of serine phosphorylation (STAT1-S727A) enhances NK cell cytotoxicity against a range of tumor cells (50, 51). In acute myeloid leukemia (AML), it appears that serine mono-phosphorylation and nuclear translocation of STAT1 were promoted by ERK, with certain chemokines and ISGs upregulated by STAT1-S727 (48). Moreover, a higher level of serine mono-phosphorylation of STAT5 was found in AML and appears to be CDK8 dependent (52).

The involvement of STAT2 in non-canonical STAT activation is also interesting. Phosphorylation of STAT2 at serine 734 appears to negatively regulate the IFN-α-induced antiviral response (53). However, when IFN stimulation is lacking, STAT2 (unphosphorylated form) constitutively binds to activated STAT1, thus specifically precluding the nuclear translocation of STAT1 in response to IFN-γ, IL-6, and IL-27 (54). Moreover, STAT2 can form an ISGF3-like complex with IRF9 in the absence of STAT1 to evoke a prolonged ISGF3-like transcriptional response and antiviral activity (55).

The current understanding of tyrosine phosphorylation-independent non-canonical STAT activation is still limited, and more investigation is needed. The existing literature presents investigations of U-STAT that mainly focus on tyrosine phosphorylation and rarely on serine phosphorylation (38, 56, 57). However, it is not known whether the function of the serine mono-phosphorylation of STATs correlates with U-STATs or the transcription complex such as U-ISGF3 (44). It is also unclear whether U-STATs and serine mono-phosphorylated STATs are functionally equal or whether they actually have unique functions. Therefore, the functions of U-STATs and serine mono-phosphorylated STATs should be delineated in future studies. Moreover, except for serine phosphorylation in the TAD domain of STATs, more phosphorylation sites in STATs have been reported, such as serine 287, threonine (T) 800, and T597 in STAT2 (58–60). The functions of these novel phosphorylation sites and their correlation with non-canonical STAT activation will be further explored in the next decade.

Since the discovery of virally encoded IFN antagonists, it has been proposed that JAK/STAT pathway antagonism is a virulence factor that might offer a novel route of virus attenuation during vaccine development using a modified live virus. Mice lacking intact JAK/STAT signaling, such as IFN-receptor or STAT1 knockout mice, are more susceptible to virus infection than wild-type mice (214–217). In addition, in vivo data suggest that a fast type I IFN response protects astrocytes from flavivirus (tick-borne encephalitis virus, JEV, WNV, and ZIKV) infection (218). Notably, as reverse genetics technology facilitates the manipulation of virus genomes, point mutations or deletions of JAK/STAT antagonists have been explored to reduce viral virulence.

As described earlier, the measles virus P gene (encodes three proteins P, V, and C) is the major antagonist that interferes with IFN-mediated JAK/STAT signaling (219). Tyrosine 110, valine 112, and histidine 115 in the shared domain of the P and V proteins determine the STAT1-antagonizing function of these two proteins (219). A recombinant measles virus with a mutation at tyrosine 110 of the P protein fails to antagonize STAT1. Compared with the wild-type measles virus, the mutant virus leads to short-lived viremia, without a skin rash and other clinical signs in rhesus monkeys, which suggests attenuation.

Similar to these observations from the measles virus, another neurotropic virus, Sindbis virus, is capable of suppressing both type I and type II IFN-mediated responses by disrupting JAK/STAT signaling (220). However, two avirulent strains that are unable to cause detectable disease in adult mice were shown to be relatively inefficient inhibitors of STAT1/2 activation (220). Further analysis demonstrated that a single amino acid determinant, the Thr at aa 538 of nsP1 of Sindbis virus, restores the STAT1 inhibition of nsP1 when it is introduced into avirulent strains and is required for Sindbis virus virulence in vivo (220). Moreover, as another well-defined JAK/STAT antagonist, NS5 from the flavivirus member WNV has been linked to virulence. The NS5 protein from a naturally attenuated WNV strain was shown to be a poor suppressor of pY-STAT1. Restoration of a single residue in NS5 of attenuated WNV to the analogous residue in virulent WNV demonstrated efficient inhibition of STAT1 activation and conferred the virulence phenotype (221). Furthermore, as observed by crystallization, the STAT1 antagonist VP24 from a virulent Ebola Sudan strain has a novel, pyramidal fold structure, which contains a site on a particular face of the pyramid exhibits reduced solvent exchange when in complex with STAT1 (153). Compared with VP24 from the non-pathogenic Reston strain, this site is above two highly conserved pockets in VP24 that contain key residues previously implicated in Ebola virus virulence (153).

One of the most promising examples of virus attenuation promisingly based on removing the JAK/STAT antagonist is the influenza viruses. As described earlier, NS1 encoded by influenza A and B viruses antagonizes IFN-activated JAK/STAT signaling at multiple steps. Investigations into generating an attenuated influenza virus based on deleting NS1 from the influenza virus genome have been ongoing for decades. The complete deletion of NS1 from influenza A generates a viable virus, but the virus replicates at a much lower level (multiple log reduction of the viral titer) than wild type in normal MDCK cells; however, replication can be partially restored in IFN-deficient Vero cells (222). Meanwhile, the NS1-deleted influenza A virus maintains pathogenicity in STAT1 knockout mice but is no longer pathogenic in wild-type mice, suggesting that the attenuation of the influenza A virus by NS1 deletion is JAK/STAT dependent (222). Since then, more studies have been conducted. As a naturally truncated NS1 variant was identified and highly attenuated in the host (222, 223), to avoid over attenuation, partial deletions of NS1 (removing the C-terminal effector domain but maintaining the N-terminal RNA binding domain of NS1) were conducted to generate mutant viruses that maintain an avirulent phenotype and evoke a protective immune response in mice (224, 225). Similar results were also observed for the influenza viruses in other hosts, including pigs, birds, and macaques (226–229). In humans, when NS1 was completely deleted from the H1N1 influenza A virus, the virus was tested in clinical trials and was demonstrated to induce higher levels of strain-specific and cross-neutralizing antibodies in a dose-dependent manner after one dose of immunization, despite the highly attenuated replication-deficient phenotype (230). Therefore, the deletion of a JAK/STAT antagonist appears to be a promising approach for the rapid attenuation of the virulence phenotype for influenza viruses.

While the deletion of a JAK/STAT antagonist for virus attenuation appears to be promising, several issues remain unclear. First, most studies that screen for JAK/STAT antagonists from viral proteins have relied on the transient expression of a viral protein in mammalian cells. The correlation between the genotype of a virus-encoded JAK/STAT antagonist and the IFN-antagonizing phenotype of an entire virus requires further investigation. As a typical example, the comparison of the neurovirulent and attenuated variant of JEV in Stat-1-deficient mice demonstrated that the attenuated phenotype of JEV is completely lost (231), suggesting an important role for the IFN-activated JAK/STAT pathway in controlling JEV infection. However, it is notable that the NS5 protein (acting as an IFN antagonist among all flavivirus) of the attenuated JEV strain maintains its potential for antagonizing IFN similar to the neurovirulent strain. Conversely, a single Glu to Lys mutation at aa 138 in the JEV envelope protein demonstrated both IFN sensitivity and the attenuated phenotype in inoculated animals (231). Research from our lab PRRSV showed a similar result. The IFN and JAK/STAT antagonist of PRRSV was mapped to the first 4 kb of the PRRSV genome, which includes the coding region for NSP1α, NSP1β, and NSP2 (232). However, one novel PRRSV isolate, A2MC2, which maintains an IFN-inducing phenotype and does not block JAK/STAT signaling in cell culture, contains an identical sequence for the first 4 kb when compared with the PRRSV strain inhibiting IFN induction and signaling (232). Therefore, when elucidating the mechanism of the JAK/STAT IFN antagonist, data gained from artificial overexpression of putative viral JAK/STAT antagonists should be carefully reviewed, and its putative role should be further verified in virus-infected cells.

Moreover, single amino acid mutation-mediated relief of JAK/STAT antagonism and virus attenuation has been reported (219, 220). Considering the natural mutation rate of virus replication, especially for RNA viruses, the restoration of virulence from an attenuated phenotype is a significant concern, even if multiple amino acid substitutions are introduced. As a typical example, after alanine-scanning mutagenesis, PRRSV-NSP1β, a well-defined IFN-JAK/STAT antagonist encoded by PRRSV, was substituted with alanines to aa 16–20 of nsp1β in mutant PRRSV and generated a viable virus with attenuated phenotype in vitro. However, after infecting pigs, the recombinant virus exhibited reduced growth at early infection times but quickly regained wild-type growth properties as a result of substitutions within the mutated sequence (233), suggesting high selection pressure toward maintaining the IFN-JAK/STAT inhibitory property of the virus in vivo. Conversely, partial deletion of the JAK/STAT antagonist, such as a truncated influenza NS1, may be a preferred approach to single or multiple amino acid mutations. However, since a viral JAK/STAT antagonist may be indispensable for viral replication and deletion of JAK/STAT antagonist from viral genome may be lethal, introducing a non-lethal but stable deletion for a viable recombinant virus requires careful investigation and a deep understanding of viral protein function (234).

As discussed in Section “Viral Interference of IFN-Activated JAK/STAT,” the phosphorylation of tyrosine resides (near residue 700) in STATs is generally considered the activation of the canonical JAK/STAT pathway (31). However, mono-phosphorylation of serine residues of different STATs has been frequently reported as non-canonical TAD serine phosphorylation without tyrosine phosphorylation (31), which may imply a novel function for STATs during virus infection and pathogenesis.

Although less investigated, virus-induced serine mono-phosphorylation of STATs with different functions than tyrosine-phosphorylated STATs has been reported for both DNA and RNA viruses, such as EBV, HIV, and PRRSV (44, 235, 236). Based on our literature research, EBV was first reported for its ability to induce serine mono-phosphorylation of STAT1 (235). EBV-induced serine mono-phosphorylated STAT1 is able to bind DNA in EBV-infected cells (235). However, researchers in this study postulated that EBV uses serine mono-phosphorylation of STAT1 to restrict IFN-stimulated STAT1-DNA binding, therefore preventing IFN-activated JAK/STAT signaling. In a later study conducted on HIV-1, serine mono-phosphorylation of STAT1 and STAT3 was observed in HIV-1-infected human brain microvascular endothelial cells and correlated with HIV-1-induced inflammatory responses and neuropathogenesis (236).

Porcine reproductive and respiratory syndrome virus is known for its capability to inhibit both IFN induction and IFN-activated JAK/STAT signaling, and several PRRSV antagonists for JAK/STATs have been identified (172, 198, 237, 238). However, it is notable that PRRSV infection promotes the IFN-independent serine mono-phosphorylation of STAT1 (S727) via nsp12 and is linked to higher expression of proinflammatory cytokines in vitro (44). Moreover, mono-phosphorylation of STAT1 (S727) is correlated with viral virulence, as a vaccine strain demonstrated a minimal effect on pSTAT1-S727 (44). This observation is interesting because the PRRSV genome encodes several nsps to block both PRR signaling (TLR or RLR) and JAK/STAT signaling (238, 239). However, aberrant sustained expression of proinflammatory cytokines and chemokines is considered to contribute to the virulence of high-pathogenesis PRRSV (240). Therefore, it appears that the expression of proinflammatory cytokines and chemokines promoted by IFN-independent mono-phosphorylation of STAT1 offers an alternative explanation for the cytokine storm that occurs during PRRSV infection. However, this speculation requires further investigation.

Research on KSHV also shows that the latent protein kaposin B of KSHV promotes the mono-phosphorylation of STAT3 at S727 in the absence of the phosphorylation of Y705 (241). It appears that mono-phosphorylation of STAT3 at S727 is activated by the host kinase mitogen-activated protein kinase-activated protein kinase 2 (MK2) and leads to elevation of STAT3-dependent genes, including CCL5 (241). Moreover, kaposin B of KSHV upregulates inflammatory cytokine levels, which correlates to KS pathogenesis (242). This finding is consistent with the putative function of serine mono-phosphorylated STAT3 in HIV-1 (236).

To date, serine mono-phosphorylation induced by virus infection has only been reported in STAT1 and STAT3. The available data imply a correlation between serine mono-phosphorylation of STATs (STAT1 and STAT3) and the proinflammatory response caused by virus infection. It is still unclear whether virus-induced serine mono-phosphorylation is common among all STATs or restricted to STAT1 and STAT3, since canonical activation of STAT1 and STAT3 also induces a proinflammatory response. Moreover, it is interesting to determine whether U-ISGF3 carries mono-phosphorylated STAT1, as previous reports of U-ISGF3 only focus on tyrosine phosphorylation, without testing serine phosphorylation (36, 38). Together, the correlation of non-canonical STAT activation, serine mono-phosphorylated STATs, and unphosphorylated STATs during viral infection requires further study and may yield insights regarding viral pathogenesis, such as virally induced cytokine storms.

The induction and signaling of type I IFNs are well-defined, and the antagonism of IFN-JAK/STAT pathway by many viruses is known. However, many questions about the type I IFN-activated JAK/STAT pathway remain unanswered. Although type I IFN subtypes appear to be functionally redundant because all type I IFNs bind to the same receptors, the differences among type I IFN subtypes are still unclear. Although activating the same pathway as type I IFNs, the function of type III IFNs requires further study. A recent report on Yellow fever virus (YFV) showed that type III IFN-mediated signaling is critical for controlling the infection of live attenuated YFV in vivo (243). Meanwhile, in HepG2 cells with persistent HEV infection, persistent activation of JAK/STAT signaling by type III IFNs renders the infected cells refractory to exogenous type I IFN treatment, and depletion of the receptors for type III IFNs restores IFN responsiveness (244). It would be interesting to examine the cross talk between type I and type III IFN-mediated signaling.

Although JAK/STAT antagonists have been identified for many viruses, for viruses that cause chronic infection, the role of type I IFN-activated JAK/STAT signaling in viral pathogenesis and virulence is more complicated. It appears that type I IFN-induced negative regulatory pathways are emerging as key drivers of chronic inflammation during chronic virus infection (245). During chronic HCV infection, the activation of endogenous type I IFN signaling and the elevation of hepatic ISGs contribute to HCV persistence (246). Therefore, the role of JAK/STAT in chronic viral infection should be reconsidered carefully, since the antiviral effects of type I IFNs are primarily manifested in an acute infection (247).

Virally encoded antagonists of type I IFN signaling are generally considered virulence factors that can be explored for virus attenuation. Current attenuation methods based on the mutation of type I IFN antagonists are still premature. Although reverse genetics provide a useful tool to manipulate viral genomic sequences, the restoration of a type I IFN-JAK/STAT-antagonizing phenotype in a recombinant virus during infection is a concern. As a result, a deep understanding of the structure and function of virally encoded JAK/STAT antagonists is more important than simple identification of JAK/STAT antagonists from viral genome. Furthermore, the potential link between virus-induced serine mono-phosphorylation of STATs and viral pathogenesis suggests that an interplay between viruses and the JAK/STAT pathway is more complicated than simply counteraction of each other. In the coming decade, we expect that more attention will be paid to these aspects to increase our understanding of the type I IFN-activated JAK/STAT pathway, the mechanism of viral-coded type I IFN antagonists and the role of non-canonical STAT activation in viral pathogenesis.

All authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.