Type I Interferon (IFN)-Regulated Activation of Canonical and Non-Canonical Signaling Pathways

Canonical and Non-canonical IFN Signaling in Malignancies

Type I IFNs have been studied in a wide range of cancers in the last few years, as illustrated in Table 1 . These studies have focused exclusively on the IFNα and IFNβ subtypes, demonstrating their clinical relevance over other type I IFN subtypes. The signaling analyses in the last few years have still focused more on the JAK/STAT cascades, specifically STAT1 effects in type I IFN signaling. However, some reports explored the impact of STAT3 versus STAT1, as well as the non-canonical involvement of MAPKs, SIRT2, and SLFN5.

STAT1 phosphorylation and the induced expression of various ISGs such as OASL and ISG15 have commonly been used as indicators of a type I IFN response (19, 27, 28, 34, 35). In a study on cervical cancer, the importance of IFN-inducible activation of STAT1 and STAT2 was demonstrated through the use of STAT1 and STAT2 knockout human HeLa cells, yet the STAT3 knockout did not have any effect on ISGs (24). By contrast, in colorectal cancer, inhibition of p-STAT3 but not p-STAT1 decreased IFNα and IFNβ induced granzyme B expression in cytotoxic T lymphocytes (25). These differences highlight how different effectors activated by type I IFNs are dependent on cell type and disease specificity. Bazhin et al. also explored IFN-activated STAT3 effects, identifying a non-canonical interaction with p38 MAPK on STAT3 phosphorylation in mature DCs (31). ULK1 has been identified as a regulator of p38 MAPK and ISGs, downstream of mTOR, in type I IFN signaling in myeloproliferative neoplasms (20). This demonstrates a connection between both major IFN activated non-canonical signaling pathways. Another MAPK, extracellular signal-regulated kinase (ERK), is involved in non-canonical type I IFN signaling in malignancy, where mitogen-activated protein kinase kinase kinase 8 (MAP3K8) and ERK phosphorylation were decreased upon IFNα treatment in bladder cancer cells (23). Further elucidation is needed on the STAT-dependent and -independent non-canonical functions of the many MAPK pathway proteins.

Additional effects of non-canonical type I IFN-induced signaling in various malignancies have been examined. A glioblastoma study identified SLFN5 as a regulator of STAT1 induction by type I IFNs (26). In leukemia and lymphoma cells, type I IFN induced phosphorylation of STAT1 on serine 727 is mediated by cyclin dependent kinase 9 (CDK9), and this activation is dependent on the deacetylation of CDK9 by SIRT2 (19). Additionally, quercetin, a natural compound, decreases Src Homology Phosphatase 2 (SHP2), a negative regulator of STAT1 (27).

An important issue related to the clinical use of IFNs is toxicity and adverse events. Although approved in 1986 by the FDA for the treatment of malignancies and viral disorders, with demonstrated positive disease outcomes, IFNα is currently not commonly used in cancer treatment due to adverse effects (36, 37). A pilot study looked at the potential of decreasing the dose of IFN-α2b for the treatment of melanoma over the course of an 11-month treatment period. Despite the dose reduction, p-STAT1 levels were induced at comparable levels throughout the 11 months, and the IFN was well-tolerated (28). An alternative strategy has been to stimulate the endogenous type I IFN response in immune cells. Tsuchiya et al. genetically engineered induced pluripotent stem cell (iPSC)-derived proliferation myeloid cells (iPSC-pMCs) to produce IFNα. When injected into mice, these IFN-producing iPSC-pMCs exerted immunomodulatory effects analogous to direct type I IFN administration, yet without adverse effects or hematopoietic stem cell exhaustion (37). Brown et al. studied recombinant poliovirus/rhinovirus chimera PVSRIPO effects in cancer immunosuppression and found PVSRIPO infection of DCs increased IFNβ production and a sustained type I IFN response, as indicated by p-STAT1 and ISG induction (IFIT1, ISG15) (34). In a separate study, the use of photodynamic therapy (PDT) lead to the upregulation of type I IFNs in melanoma cells and DCs co-cultured with the PDT treated cells; the authors proposed this ex vivo strategy of stimulating DCs with the use of PDT as a possible immunotherapy (29).

Distinct from the positive outcomes of type I IFN treatment for malignancies, a number of studies have addressed the potential link of IFN treatment with chemotherapy resistance, immunosuppression, and driving of cancer stemness. Qadir et al. found chronic CD95 activation leading to cancer stemness was driven by IFNα/β-STAT1 canonical signaling (32). They also provided evidence that radio-resistant squamous cancer cells had increased p-STAT1 and ISG expression and that type I IFN treatment of breast and squamous cancer cells increased stemness and sphere formation, which was blocked by JAK inhibition, indicative of the involvement of canonical signaling.

Several studies have evaluated the effects of type I IFN administration in combination with immunotherapy. One group showed that IFNα increased programmed death-ligand 1 (PD-L1) expression on various immune cells through non-canonical p38/STAT3 signaling (31). The inference is that combining immunotherapy with IFNα treatment would limit the immunosuppressive effects of IFN treatment and permit effective growth inhibition. Similarly, another study provided evidence that IFNα-iPSC-pMC treatment increased PD-L1 mRNA, and when combined with a PD-L1 inhibitor, synergistic anti-tumor effects were reported (37). The poliovirus/rhinovirus type I IFN induced response likewise increases PD-L1 expression (34). Additionally, a bioinformatics examination of IFN gene deletions revealed that homozygous deletion of IFN was significantly associated with non-response to anti-CTLA4 treatment among melanoma patients (38). Overall, these studies suggest type I IFNs may have a critical role in immunotherapy strategies, possibly via a combination of type I IFN treatment with PD-L1 inhibition. Moreover, the data suggest that PD-L1 expression may be affected by IFN-induced non-canonical signaling.

Canonical and Non-canonical Signaling in Autoimmune Diseases

Accumulating evidence implicates chronic and persistent type I IFN signaling in systemic inflammation that promotes the pathogenesis of some autoimmune diseases, including systemic lupus erythematosus (SLE), rheumatoid arthritis, multiple sclerosis (MS), type I diabetes (T1D) and Sjögren’s syndrome, among others (13, 39). These conditions are associated with different clinical symptoms and management strategies, yet there are common features related to the underlying inflammatory signaling pathways involved and the dysregulated immune response. Figure 2 summarizes the cell type-specific type I IFN-induced canonical and non-canonical signaling pathways recently implicated in autoimmune diseases.

IFNα has been shown to impact the onset and progression of T1D, which involves the autoimmune attack of pancreatic β cells (40). One study demonstrated that IFNα activated STAT1, STAT2, and STAT3 in pancreatic β cells through TYK2, and that STAT2 was more critical than STAT1 in mediating the inflammatory and endoplasmic reticulum (ER) stress response (41). Another study likewise reported on IFNα induction of ER stress in pancreatic β cells, leading to the downregulation of insulin production and influence on T1D onset (42). A mouse model study revealed that inhibition of IFNα, but not IFNβ, in the pre-diabetes stage prevented the onset of T1D and blocked autoreactive T cells from entering and killing β cells in the pancreatic islets (43). Notably, patients with neutralizing autoantibodies to type I IFNs, specifically IFNαs, are less likely to develop T1D (44). These studies identify the negative impact of IFNα on the development of T1D.

Sjögren’s syndrome is an autoimmune disease with glandular lymphocyte infiltration leading to symptoms of dry mouth and eyes, where approximately 50% of patients have a type I IFN signature (45, 46). Given that this IFN signature is not present in all patients, one study analyzed the effects of IFN-α2b treatment of peripheral blood mononuclear cells (PBMCs) from patients with Sjögren’s compared with PBMCs from healthy donors, including type I IFN signature-positive and negative patients (45). Baseline effector protein phosphorylation levels differed predominantly in T cells in Sjögren’s patients compared with healthy individuals, with higher p-p38 and p-STAT1 (Y701, S727). Sjögren’s patients also exhibited increased IFNα-inducible JAK phosphorylation of STAT1 (Y701). Further, IFNα-2b treatment of PBMCs upregulated p-STAT1 (Y701) in B cells and downregulated p-STAT3 on S727 in T cells in type I IFN signature-positive patients.

SLE manifestations include organ damage and skin rash (47, 48). There is an IFNα signature in sera of SLE patients. A recent study using inducible IFNα transgenic mice found that upregulation of IFNα alone was capable of inducing an SLE phenotype (47). SLE pathogenesis is characterized by inflammasome overactivation; one study demonstrated that prolonged IFNα treatment increased inflammasome activity, which was eliminated with knockdown of IRF1 in SLE monocytes (49). IFNα treatment increased p-STAT1 and p-STAT2 at tyrosine residues, indicative of a classical JAK/STAT driven response. Another group that analyzed B cells from SLE patients, found increased baseline p-STAT3 (Y705), not p-STAT1, compared to B cells from healthy individuals (50). Additionally, these investigators found that IFNα treatment polarized naïve B cell differentiation towards a lupus-like phenotype, which was reversed by a STAT3 inhibitor and was absent in STAT3-deficient donor naïve B cells. In SLE monocytes, Gkirtzimanaki et al. identified IFNα induced mTOR activity, which promoted oxidative stress, revealing non-canonical IFNα signaling in SLE (51).

Cognizant of the persistent IFNα signature in SLE patients, a phase IIb clinical trial evaluated the effects of vaccination with IFNα kinoid, which produces anti-IFNα antibodies (52). Although the trial did not see a benefit in Based Composite Lupus Assessment (BICLA), the drug did provoke anti-IFN-α2b serum antibodies and decreased the IFN gene signature in 91% of patients. An anti-IFNAR1 monoclonal antibody, anifrolumab, has been evaluated in 11 clinical trials for SLE (9), Sjögren’s (1), and rheumatoid arthritis (1), with encouraging results (53). A recent phase III trial in SLE did not meet its primary endpoint of response, as per the SLE Responder Index; however, the same group conducted another phase III trial using the of British Isles Lupus Assessment Group (BILAG)-BICLA response as the primary endpoint and reported a statistically significant higher percentage of patients having a response as well as seeing a decrease in secondary endpoints, suggesting that a chronic IFNα response in SLE patients may contribute to disease pathogenesis (48).

Interestingly, while IFNα has been implicated in the pathogenesis of various autoimmune diseases, IFNβ has been used to successfully treat MS (54). Employing a mouse model of MS, studies with mice that lack the IFNβ gene revealed that in the absence of IFNβ the mice had a more severe disease with earlier onset and that the lack of IFNβ predisposed the mice to a pro-inflammatory Th17 immunophenotype (55, 56). Given the heterogeneity of the disease, and differing patient responses to IFNβ treatment, the identification of potential biomarkers of response to IFNβ therapy is receiving considerable attention. One study suggested predictors of response could be based on cell type-specific responses to type I IFN signaling, such as higher activation of STAT1, STAT3, and p38, leading to higher TRAIL expression in monocytes of IFN responders (57). Hurtado-Guerrero et al. analyzed monocytes from MS patients ex vivo, either left untreated (baseline) or after short-term IFNβ treatment (58). At baseline, there were no detectable differences in the levels of IFNAR1, IFNAR2, p-STAT1, and p-STAT2 among responders and non-responders, yet following IFNβ treatment, differences were observed. They found a pattern of decreased IFNAR1 and increased IFNAR2, p-STAT1, and p-STAT2 levels representing 68.4% of responder IFNβ-stimulated monocytes. Other groups have employed bioinformatics to uncover gene signatures that determine a response to IFNβ. One study used a feature selection computational method on a longitudinal microarray dataset of relapse-remitting MS (RRMS) patients treated with IFNβ-1b, and found a predictive seven gene signature (CXCL9, IL2RA, CXCR3, AKT1, CSF2, IL2RB, GCA) with 65.08% predictive accuracy (59). Using an alternative method of Elastic net modeling, Fukushima et al. analyzed time-course microarray datasets from PBMCs of MS patients and identified eleven (ZBTB16, ZFP37, HPS5, HOPX, ARFGAP3, CALML5, VPS26A, SLC5A4, MBL2, DLGAP4, CACNA1C) and eight (SMA4, MIR7114_NSMF, LSM8, FLAD1, RRN3P1, RASL10A, IER3IP1, CDH2) genes predictive of an IFNβ response, with 81% and 78% accuracy, respectively, for each dataset (60). A different study employed the GeneRank method to identify monotonically expressed genes (MEGs) that determine a good response (AFTPH, ALOX5, ATG7, MYD88, LILRB1, PRKAB1, PSEN1, VAMP3) and a bad response (AGFG1, CHM, IGLL1, PELI1, PTEN) for responders, and two bad response MEGs for non-responders (NAP1L4, MMS19) in IFNβ treated RRMS patients (61). As an alternative strategy to gene analysis, a logistic regression modeling method was used to examine metabolites from the sera of a cohort of MS patients to predict the production of anti-drug antibodies (ADA) to IFNβ treatment (62). Differences in 29 metabolites were shown to be indicative of ADA production, and the top ten most significant metabolites were lipid related. Another study using a systems immunology approach evaluated ADA production differences in three IFNβ treated cohorts and showed reduced baseline NOTCH2 expression and that a pro-inflammatory phenotype in monocytes was predictive of ADA development (63). Given the preceding, there is a need for further identification and characterization of biomarkers that are reproducibly predictive of an IFNβ response in RRMS patients.

The differences between IFNα and IFNβ in the generation of effects in autoimmune diseases requires additional analysis. Although both type I IFNs bind to and initiate signaling cascades through the dimeric IFNAR, they do differ in primary amino acid sequences and in binding affinity to the receptor which may account for varying impacts of the response on cells (54, 64). Binding affinity for IFNAR1 and IFNAR2 varies among IFNα subunits, with overall higher affinity for IFNAR2 over IFNAR1, and IFNβ has tighter binding to each receptor subunit than any of the IFNα subunits (64, 65). How the induced signaling can differ after the type I IFN ligand binds is not well understood but studies have shown differences further downstream in genes and transcription factor binding sites of IFNα versus IFNβ signaling, such as enrichment of IRF8 binding sites in IFNβ response (54). As previously mentioned, cell-type and disease state lead to variance in type I IFN signaling which is further complicated by the differences invoked by IFNα and IFNβ and requires further studies, especially to understand the protein signaling cascades after binding of type I IFNs to the IFNAR.

Canonical and Non-Canonical IFN Signaling in Antiviral Responses

IFNs are critical effectors of an antiviral response in mammalian cells. Following viral infection, type I IFNs are produced by immune and non-immune cells, bind to and activate IFNAR, and signal through canonical and non-canonical pathways (66–68). An area of interest has been the involvement of the IFN system in the pathophysiology of Coronavirus Disease 19 (COVID-19).

Since the emergence of severe acute respiratory syndrome coronavirus (SARS-CoV) in 2003 and Middle East respiratory syndrome coronavirus (MERS-CoV) in 2012, therapeutic options for treatment have been limited (69). Type I IFNs are attractive therapeutic candidates because of their ability to clear virus through direct inhibition of viral replication of both DNA and RNA viruses and their effects on the activation of specific immune cell subsets to assist with viral clearance (70). Many viruses, including coronaviruses, evade an IFN antiviral response by inhibiting the production of type I and III IFNs (71–73). Scrutiny of the SARS-CoV genome identified the genes NSP1, NSP3, ORF3b, and ORF6 that are antagonists for type I IFNs, as well as the N protein (74). ORF6 not only inhibits the production of IFN but can also inhibit the expression of ISGs by inhibiting STAT1 nuclear translocation, through disruption of karyopherin-mediated transport. IRF3 is an important transcription factor necessary for IFNβ expression. The papain-like protease (PLpro), conserved in both SARS-CoV and SARS-CoV-2, inhibits the phosphorylation required for IRF3 homodimerization and nuclear translocation leading to its association with CBP/p300 and NF-κB for IFNβ expression (75–77). Comparing the gene sequences between SARS-CoV and SARS-CoV-2 for NSP3, ORF3b, and ORF6, revealed sequence differences that may contribute to the greater sensitivity of SARS-CoV-2 to type I IFNs (77). Konno et al. made the observation that ORF3b inhibits type I IFN induction more so in SARS-CoV-2 than in SARS-CoV, and a naturally arising SARS-CoV-2 variant exerts even greater antagonism of type I IFN induction by ORF3b (78). Accumulating data continue to provide further evidence of a blunted IFN response in COVID-19 cases (79–83).

Recently, data have emerged that indicate that SARS-CoV-2 is sensitive to the antiviral effects of both IFNα and IFNβ in cell culture assays, similar to the sensitivity of SARS-CoV in vitro (84–86). A pilot clinical study during the SARS outbreak of 2003 demonstrated that treatment with an IFNα resulted in reduced disease-associated impaired oxygen saturation and rapid resolution of lung abnormalities (87). The evidence of SARS-CoV-2 sensitivity to IFN treatment and accumulating clinical studies suggest that IFN treatment may have therapeutic benefits for COVID-19 (88). Early on in the pandemic, Zhou et al. provided evidence that treating COVID-19 patients with nebulized IFN-α2b with or without the antiviral drug, arbidol, accelerated viral clearance from the airways of infected patients and also reduced the circulating levels of the inflammatory cytokines, IL-6 and CRP (89). Following up from this exploratory study, there have been several clinical studies evaluating the therapeutic benefit of IFNα and IFNβ treatment for COVID-19 (see Table 2 ). In vitro studies suggested greater antiviral effectiveness of IFNβ over IFNα against SARS CoV (97). This prompted the WHO SOLIDARITY randomized controlled trial of a combination of lopinavir/ritonavir, ribavirin, and IFNβ-1b versus lopinavir/ritonavir in SARS-CoV-2 (90). The findings suggest that the triple combination treatment was more effective than lopinavir/ritonavir alone, reducing symptom severity and time to viral clearance. Given the emerging evidence that lopinavir/ritonavir treatment may be ineffective against SARS-CoV-2, the ongoing trial had been amended to compare the therapeutic effectiveness of IFNβ with remdesivir, a viral polymerase inhibitor that has demonstrated limited therapeutic efficacy in COVID-19 cases. A prospective observational study was conducted to assess the therapeutic efficacy of IFN-α2b in SARS-CoV-2 patients during the first month after the COVID-19 outbreak began in Cuba. Intramuscular administration of IFN-α2b improved both the rate of recovery and case fatalities (91). However, a retrospective cohort study demonstrated that there is great importance on the timing of administration of IFN-α2b with reduction of in-hospital mortality when administered the first five days of admission but increased mortality and delayed recovery was seen if given later (92). Additionally, inborn errors of type I IFNs and presence of autoantibodies against type I IFNs can be determinants of severity of disease and effectiveness of type I IFN treatment (95, 98). Roughly 10% of COVID-19 patients with severe pneumonia in a cohort of 987 patients had neutralizing autoantibodies against IFNα, IFNω, or both, where patients with no or mild symptoms had no detectable autoantibodies (95). These findings demonstrate that administration of IFNα may not be effective in patients with severe condition and autoantibodies, but since IFNβ autoantibodies were uncommon in the same patients, IFNβ may provide a more beneficial treatment. The same group analyzed a separate cohort of patients with life-threatening pneumonia and found 3.5% had inborn errors in type I IFN related genes, specifically in loci pertaining to TLR3- and IRF7-dependent type I IFN induction (98). This showed a commonality with influenza since similar type I IFN related gene defects have been demonstrated in life-threatening influenza pneumonitis (99).

Similar to SARS-CoV, SARS-CoV-2 interacts with the angiotensin-converting enzyme 2 (ACE2) for cell entry, while MERS-CoV exploits the dipeptidyl peptidase 4 (DPP4) receptor for entry into human cells (100–102). Ziegler et al. demonstrated that nasal secretory cells (goblet cells), type II pneumocytes, and absorptive enterocytes of the ileum are positive for the two critical receptors for SARS-CoV-2 cell entry, ACE2 and the type II transmembrane serine protease, TMPRSS2 (103). Their observation that ACE2 expression is induced by type I IFNs in primary upper airway basal cells and lung tissue is hard to reconcile with IFNs inhibiting infection by SARS-CoV-2, yet recent emerging data suggesting a role for the renin-angiotensin pathway in protection from specific clinical features of COVID-19 would support a role for ACE2 in limiting COVID-19 severity. The inability of mice to uptake SARS-CoV-2 infection through the mouse ortholog of entry receptor ACE2 prompted Israelow et al. to create an adeno associated virus-mediated human ACE2 mouse model that can be utilized to analyze SARS-CoV-2 in mice, and they found increased type I IFN signaling ISGs in the lungs and limited control of SARS-CoV-2 replication by type I IFNs (104). The involvement of canonical versus non-canonical pathways in the induction of IFN-responses against SARS-CoV2 remains to be elucidated.

SARS-CoV-2 and influenza viruses are respiratory infections where disease severity results in lung hyper-inflammation and acute respiratory distress. Findings from clinical studies suggest that the early viral phase of both infections is associated with a blunted IFN response, yet progression to severe disease shows no such failed IFN response, specifically elevated levels of ISGs in PBMCs are observed (105–107). The implications are that the therapeutic benefits of IFN treatment are applicable in the early viral phases of COVID-19 and influenza, but that once the pulmonary phases of both infections progress to hyper-inflammation, IFN treatment is likely to be contra-indicated. Non-canonical effects in type I IFN signaling in influenza have been demonstrated as with p38 MAPK signaling, shown to be important in affecting type I IFN production and signaling in highly pathogenic avian influenza virus infected endothelial cells (108). Additionally, IFN-κ treatment inhibits influenza replication in lung cells, dependent on IFNAR, p38, CHD6, and Fos activation, but not STAT1 (109). Notably, IFNα induced STAT3 activation is crucial for inhibition of influenza viral replication and ISG transcription in mouse embryonic fibroblasts (110).

Although antiretroviral therapy (ART) for Human Immunodeficiency Virus (HIV) infection has transformed this infection from a fatal one to a chronic disease, viral reservoirs complicate efforts for HIV elimination, and a recent review paralleled HIV reservoir persistence to immuno-editing and immune evasion in cancer (111). The roles of type I IFNs in the pathophysiology of HIV infection are not fully understood, but IFNα has been implicated as an adverse factor in the persistence of HIV-1. When circulating levels of IFNα were measured for healthy donors, primary-infected, and chronically-infected patients, higher IFNα levels were associated with higher viral loads and higher expression of the ISG, USP18, which negatively regulates IFNα signaling by displacing JAK2 bound to IFNAR2 (112). Humanized mouse models have provided evidence that whereas type I IFNs suppress early HIV infection, type I IFN signaling induces T cell depletion and impaired functionality during persistent infection. When IFN signaling is blocked in HIV-infected mice or in monkeys receiving ART, this reduces the HIV reservoir, rescues anti-HIV T cells, and reduces HIV-induced inflammation (113–115). Notably, HIV-1 proteins, Vpu and Nef, inhibit ISG expression through canonical IFNα mediated JAK/STAT1 signaling, blocking any antiviral benefits from IFNα (116). Knockout of IFNAR1 in an HIV-induced brain injury mouse model provided memory benefits and neuronal injury protection while suppressing p38 activation, indicating involvement of type I IFN non-canonical signaling in HIV-1–related neurotoxicity (117). Indeed, there is accumulating evidence that sustained type I IFN signaling, surprisingly, can promote viral replication for a number of viruses, mediated by induction of certain ISGs and inhibition of IRFs (14). IFN induced 2′5′-oligoadenylate synthetase-like (OASL) limits RNA virus replication through enhancing RIG-I signaling yet inhibits cGAS and promotes viral replication for DNA viruses such as HSV.

Of late, there are emergent data that SLFN proteins, non-canonical effectors of type I IFN signaling, have a role as antivirals. IFN induced SLFN11 expression controls protein synthesis by regulating tRNA abundance, limiting West Nile virus, dengue virus, and Zika virus replication, all (+) ssRNA viruses, but having little effect on (–) ssRNA viruses (118). Interestingly, SLFN 11 control of HIV-1 infection is independent of type I IFN signaling (119). IFNβ induced SLFN14 exhibits antiviral activity in mouse macrophages, limiting infection with influenza virus or the DNA virus, varicella-zoster virus (120).

Besides the duration of type I IFN signaling influencing whether there is inhibition or enhancement of viral replication (105, 112), cell environmental factors also contribute to a type I IFN response. In a mouse model of vesicular stomatitis virus infection, high salt levels augment type I IFN signaling through the non-canonical p38 pathway (121). In neurons, viral infection may cause pain hypersensitivity; type I IFNs elicit pain sensitization in neurons, by promoting MAPK interacting kinase phosphorylation of eukaryotic initiation translation factor (122).