Severe cases of influenza infection are characterised by an excessive and detrimental inflammatory response, lung damage (Mauad et al., 2010) and cardiovascular complications (Dawood et al., 2012). Airway epithelial cells, macrophages and other inflammatory cells shape the inflammatory response to influenza A virus (IAV) infection, and this can induce significant lung damage. Endothelial cells undergo apoptosis in response to IAV, resulting in increased permeability of the monolayer, which could account for the pulmonary oedema often observed in seriously ill patients (Armstrong et al., 2012). Indeed, the targeting of endothelial cells using the anti-inflammatory agonist, 2-amino-4-(4-heptyloxyphenyl)-2-methylbutanol (AAL-R), can ameliorate the influenza-induced cytokine storm, implying that immune cell infiltration and cytokine production are both orchestrated by lung endothelial cells (Teijaro et al., 2011). If endothelial cells are critical for immune cell activation in the lung, then it was reasoned that the reactive oxygen species (ROS) status in these cells may control the amount of inflammation and pathogenesis during a viral infection.

ROS are also implicated in IAV pathogenesis (Vlahos et al., 2011; Vlahos et al., 2012; Vlahos and Selemidis, 2014). ROS are pleiotropic molecules, which serve both physiological and pathological functions in the context of IAV infections. The cellular source and subcellular compartmentalization of ROS production can impact on their ultimate biological function (Selemidis, 2019). For example, ROS can be generated by inflammatory cells such as neutrophils and macrophages via either an endosomal NOX2 NADPH oxidase or by alterations in cell metabolism resulting in mitochondrial ROS generation. Recent work by our group and others have shown that mice deficient in NOX2 have an ameliorated IAV-induced inflammatory response in the airways, which is associated with reduced lung oxidative stress and viral titres (Vlahos et al., 2011). Moreover, we showed that IAV of low to high pathogenicity activate endosomal NOX2, which generated ROS in macrophages (To et al., 2017). The resultant endosomal hydrogen peroxide (H₂O₂) modified critical cysteine residues on the TLR7 protein, which can alter signalling and lead to reduced antiviral cytokine production, promoting viral replication and disease pathogenesis (To et al., 2017). In contrast to NOX2, airway epithelial cells also generate significant levels of ROS, including H₂O₂ via the DUOX 1/2 enzymes, to exert anti-bacterial and antiviral responses, although the mechanisms are not well characterised. IAV infection leads to a DUOX2 up-regulation and DUOX-mediated ROS generation (Strengert et al., 2014). In vivo silencing of DUOX increased the viral load after intranasal infection of mice with 2009 pandemic H1N1 influenza virus (Strengert et al., 2014). This suggests that there is a dynamic balance in compartmentalised ROS production, which has the potential to regulate viral pathogenesis.

The endothelium is a major regulator of vascular permeability during viral infection. Therefore the role of enzymatic sources of endothelial ROS underpinning the pathogenesis of IAV infection is worthy of investigation. In this regard, the NOX4 isoform of the NADPH oxidase enzyme family is expressed in at least 100 fold higher amounts than NOX2 in endothelial cells, and is the major source of ROS by these cells (Van Buul et al., 2005). Importantly it has a unique mode of activation and pattern of physiological effects. Unlike NOX2, which requires the assembly of multiple subunits for activation, NOX4 is constitutively active once it forms a heterodimer with p22phox and is thought to be regulated primarily through its expression (Drummond et al., 2011). NOX4 oxidase also differs from the other NADPH oxidases in that, due to its extended E-loop, it directly produces H₂O₂ rather than superoxide. These differences in structure and function become important when considering the role of NOX2 and NOX4 in the vasculature. NOX2 and NOX4 have been shown to activate distinct kinase pathways in response to stimulation by agonists in HEK293 cells (Anilkumar et al., 2008). While NOX2 exacerbates oxidative stress, inflammation and infarct volume in mouse models of stroke (Walder et al., 1997; Chen et al., 2011; De Silva et al., 2011), NOX4 was found to be protective (Schroder et al., 2012). NOX4 accounted for approximately 75% of the H₂O₂ formed in the vasculature and promoted angiogenesis. NOX4 also limited angiotensin-induced vascular dysfunction, and promoted nitric oxide and heme oxygenase 1 (HO-1) production (Fredenburgh et al., 2007). Furthermore, NOX4 has been shown to have protective anti-inflammatory effects in atherosclerosis and is down regulated in patients with atherosclerosis and diabetes, and in mouse models of atherosclerosis (Gray et al., 2016). Interestingly, overexpressing NOX4 specifically in endothelial cells, resulted in enhanced vasodilation and reduced systolic blood pressure (Ray et al., 2011). The opposing effects of NOX4 and endosomal NOX2 on inflammation and cytokine signalling might therefore be of significance for the exacerbated inflammation and pathogenesis resulting from influenza and other viral infections.

While there have been several studies examining the role of NOX4 in inflammatory pathways in the setting of cardiovascular disease, there has been limited investigation of the role of NOX4 in influenza pathology. To the best of our knowledge, there has been only one study examining the potential role of NOX4 in influenza infection and pathogenesis (Amatore et al., 2015). Increased NOX4 expression was detected in human mucoepidermoid pulmonary carcinoma (NCI-H292) cells that had been infected with the highly pathogenic influenza strain A/Puerto Rico/8/34 H1N1 PR8 (Amatore et al., 2015). Treatment with NADPH oxidase inhibitor diphenyleneiodonium (DPI) and knocking down NOX4 expression with siRNA prevented viral replication in this in vitro model of infection. The complexity of cancer cell ROS biology could have a confounding effect on the conclusions from this initial viral study and there have been no in vivo studies examining the effect of endothelial NOX4 on influenza pathology in an animal model of pathogenesis. The aim of this study was therefore to determine if endothelial NOX4 expression could influence the amount of airway/lung inflammation, and morbidity in response to IAV infection in vivo. We have investigated viral pathogenesis in the endothelial NOX4 overexpressing mouse [2-3 fold increase in endothelial NOX4 expression (Ray et al., 2011)] and WT control mice to examine the differential effect of NOX4 expression on influenza pathogenesis.

Epithelial and inflammatory cells are critically involved in viral pathogenesis, but there is evidence that endothelial cells are also important orchestrators of the inflammatory response to IAV infection (Teijaro et al., 2011). ROS are also important mediators of IAV pathology, however, the role of ROS production by the endothelium in response to IAV infection has not previously been established. NOX4 is the most abundantly expressed isoform of the NOX enzyme family in lung endothelial cells and the expression of this, ROS producing enzyme was specifically suppressed in response to influenza infection. In NOX4 transgenic mice the expression of this enzyme and production of H₂O₂ is elevated by 2-3 fold (Schroder et al., 2012). Influenza A virus infection of these NOX4 TG mice resulted in a significantly attenuated airway and lung inflammatory response, reduced neutrophil infiltration, lower neutrophil chemotactic factor expression and a reduction in viral burden when compared to WT mice. Interestingly, this protective effect of NOX4 overexpression in the lung endothelium appeared to occur mainly during the early phases of the infection. This demonstrates that IAV infection suppresses lung endothelial NOX4 expression to promote an inflammatory response, which enables enhanced viral replication and exacerbated lung pathogenesis.

In some respects, our findings contrast with Amatore et al. (2015) who found that mucoepidermoid pulmonary carcinoma cells infected with the highly pathogenic PR8 influenza strain in vitro had increased NOX4 gene and protein expression. While the anti-NOX4 antibody used by Amatore et al. (2015) has been invalidated and raises questions about these findings (Altenhofer et al., 2012) it is important to consider that the infected cells were tumorigenic, which could have influenced NOX4 expression. In contrast, we examined lung NOX4 expression in an in vivo model, over multiple days and with a different strain of virus demonstrating significant reduction in NOX4 expression following IAV infection. However, it should be noted that the specific effects of IAV infection on NOX4 expression in cells other than the epithelium is largely unknown and that basal NOX4 expression in epithelial cells is low, particularly when compared to NOX2 and NOX1 (Kolarova et al., 2010). Given the low basal expression of NOX4 in the epithelial cells, the role of NOX4 in the endothelium may be more clinically relevant and a counter balance for the effects of the other NOX enzymes. Consequently, the decrease in NOX4 expression in the lungs of infected mice might reflect the pathogenic effects of NOX2 oxidase and the potential reciprocal regulation of these two NOX isoforms in vascular endothelial cells that requires further investigation. For instance, the silencing of NOX2 in lung endothelial cells has been shown to increase NOX4 expression and vice versa (Pendyala et al., 2009). A similar mechanism could be responsible for the decrease in NOX4 and it is also important to consider that this expression may be variable over the course of the infection. Moreover, nitric oxide has been shown to cause down regulation of NOX2 in human endothelial cells (Duerrschmidt et al., 2006), while H₂O₂ drives eNOS production in endothelial cells (Thomas et al., 2002). Since these NOX4 TG mice are producing ~2-3 times more hydrogen peroxide than WT mice, this could indicate an increase in NO production and, subsequently, NOX2 downregulation.

Weight loss was one of the indicators of morbidity assessed in this study and while all mice lost weight when infected with HK x-31, the NOX4 TG mice lost less weight than infected WT mice at the earlier time point after infection. This suggests that endothelial NOX4 has protective effects in the earlier stages of IAV infection. Additionally, there were similar findings with the other markers of morbidity, including lung weight, which was used as an indicator of pulmonary oedema and inflammation. WT mice infected with IAV had significantly increased lung weight, however, IAV did not significantly increase the lung weight of infected NOX4 TG mice. These findings suggest that NOX4 reduces pulmonary oedema and given its location at the endothelium this is a critical site to regulate this pathological process.

This study also examined the oxidative burst in lung inflammatory cells, which is solely due to NOX2 oxidase (Vlahos et al., 2011; Selemidis et al., 2013; Yatmaz et al., 2013; To et al., 2017). There was an increase in this extracellular NOX2 oxidase ROS production from the BALF cells of infected WT mice, which was significantly less in the NOX4 TG mice. NOX4 has been shown to drive the production of the antioxidant heme oxygenase 1 (HO1) (Schroder et al., 2012), which has protective effects in a mouse model of lung injury (Fredenburgh et al., 2007). This could account for the decrease in ROS production and the reduced lung weight in virus infected NOX4 TG mice. Alternatively, the reduction in ROS production in the NOX4 TG mice might be due to a reduction in the proportion of neutrophils in the BALF of those mice. Neutrophils are likely to significantly affect the overall ROS production of BALF inflammatory cells as they express high levels of NOX2 and drive an oxidative burst via this enzyme. For instance, the BALF from uninfected mice primarily consists of alveolar macrophages, whereas the BALF infected mice have a higher percentage of neutrophils, which could explain the differences in ROS production, owing to their greater capacity for NOX2 dependent ROS production. Compared to the infected WT mice, BALF from the infected NOX4 TG mice had a decreased percentage of neutrophils and this decrease in neutrophil infiltration could explain the decreased overall ROS levels seen in the BALF. Interestingly, ROS production at Day 7 was significantly lower than that observed at Day 3 for both WT and NOX4 TG mice, even though there were significant numbers of neutrophils still in the BALF. The marked reduction in ROS might be due to the relative lack of influenza virus detected at Day 7, and thus a reduction in stimulation of ROS generation. At Day 7 there is a significant adaptive immune response that ultimately clears the virus.

A hallmark feature of severe influenza infection is an excessive and detrimental inflammatory response (Dominguez-Cherit et al., 2009; Mauad et al., 2010). This study found an increase in airway inflammatory cell infiltration, alveolitis, peribronchiolar inflammation and overall lung inflammation in WT influenza-infected mice at both 3- and 7-days post infection. However, there was less airway inflammatory cell infiltration, alveolitis and peribronchiolar inflammation and overall inflammation in the lungs of infected NOX4 TG mice, 3-days post infection. This decrease in inflammation corresponded with reduced weight loss, lung weight and viral titres in infected NOX4 TG mice at day three. This demonstrated that enhanced ROS production from NOX4 oxidase has the capacity to attenuate the viral load and decrease the downstream pathogenesis caused by IAV infection at early phases of infection, however, additional mechanisms driving inflammation beyond Day 3 appear not to be influenced by endothelial NOX4.

Influenza infected NOX4 TG mice had decreased expression of CXCL10, CCL3, CXCL1 and CXCL2 three days post infection. The activation of CXCR3 by CXCL10 triggers influenza-induced neutrophil infiltration, and the deletion of CXCL10 and CXCR3 results in increased survival (Ichikawa et al., 2013). The decrease in CXCL10 expression in the influenza-infected NOX4 TG mice compared to infected WT mice is consistent with both the decrease in neutrophil infiltration and the decrease in lung weight seen in infected NOX4 TG mice. CXCL10 also contributes to the oxidative burst from neutrophils extracted from an acid-triggered acute lung injury (Ichikawa et al., 2013). This is consistent with the L-012 data in this study, which showed a decreased oxidative burst in infected NOX4 TG mice compared to the infected WT mice. CCL3 and CXCL2 have also been shown to be involved in the recruitment of neutrophils in models of bacterial infection (Zeng et al., 2003; De Filippo et al., 2013), and are both upregulated in influenza infection (Wareing et al., 2004). The downregulation of CCL3 in infected NOX4 TG explains the decrease in neutrophil infiltration. CCL3 and CXCL10 are known to trigger the infiltration of macrophages in influenza infection (Zeng et al., 2003; Ichikawa et al., 2013) yet, despite the decrease in these chemokines, there was no change in the number of macrophages in the infected NOX4 TG. Either these are redundant pathways, or other compensatory mechanisms are involved in monocyte/macrophage infiltration. It is interesting that the early effects on inflammation were more pronounced than at later time points suggesting that the viral infection eventually overrides the early effects of NOX4 ROS production presumably as the cytokine response gains momentum. Of importance, these cytokines aforementioned are involved in regulating the recruitment of both macrophages and neutrophils. Specifically, a reduction in virus-induced neutrophilic inflammation is likely explained by a reduction in the expression of CXCL1 (KC), a chemoattractant that mediates the recruitment and trafficking of neutrophils to the site of infection (De Filippo et al., 2013; Sawant et al., 2015). In corroboration with our observations, CXCR1/2 antagonism alleviated neutrophilia and pulmonary damage in mice challenged with streptococcus pneumoniae and influenza virus, highlighting the selective nature of the chemoattractant CXCL1 in modulating neutrophilic inflammation associated immunopathology (Tavares et al., 2017).

The seemingly opposing roles of NOX2 and NOX4 in IAV pathogenesis aided by the observational nature of this study, enables a series of new hypotheses and questions that warrant further experimentation to be proposed. For example, is there any relationship in the kinetics of NOX2 and NOX4 and their resulting effects on the immune response to IAV? What are the ratios of NOX2 and NOX4 expression at various stages of the course of IAV infection and is NOX4 activated following the inflammation by NOX2 to mediate control during an IAV infection? Also, given the different subcellular compartmentalisation of NOX2 and NOX4, i.e. endosomal Vs endoplasmic reticulum/mitochondria, is there any cross regulation such that one might provide a dominant response over the other in a temporal fashion to instil pro-inflammatory dominant vs anti-inflammatory response from the endothelium? It must be kept in mind though, that NOX2 expression is much higher than NOX4 in inflammatory cells such as macrophages and neutrophils, and is likely to dominate the inflammatory response in these cells. Whilst NOX4 expression could be under the influence of NOX2, there might be additional factors that regulate NOX4 expression including factors that influence p22phox. Downregulation of p22phox can lead to destabilisation of the NOX4-p22phox heterodimer and thereby suppression of NOX4 activity. A better understanding of the dynamics of NOX2 and NOX4 activity during IAV infection is important for future therapeutic development strategies that aim to target oxidative stress pathways. Indeed, these fundamental differences in roles of NOX4 and NOX2 would strongly suggest that broad based anti-oxidants are unlikely to be of benefit but rather a more targeted spatial-temporal approach is necessary to address IAV mediated ROS production.

In summary, NOX4 has been shown to be protective against inflammatory and ischemic stress. This study has shown for the first time that endothelial NOX4 also ameliorates some symptoms of IAV infection, including oxidative stress and neutrophil infiltration, lung damage, and cytokine expression. This study further highlights the role of the endothelium in IAV pathology and the opposing roles of NOX isoforms in this context. This exemplifies the current paradigm that ROS production is not always detrimental, and that the specific site of ROS production can have vastly different effects on pathogenesis. It appears that while endosomal ROS production drives pathogenesis, that of NOX4 derived ROS production has the capacity to attenuate it. Future studies are needed to decipher the subcellular roles of endothelial NOX4, in particular, its expression in mitochondria places it at a site that might regulate mitochondrial ROS levels and thereby inflammation. Consequently, the specific pharmacological induced increase of endothelial NOX4 activity could be a means of reducing IAV-induced inflammation and oxidative stress and improve patient outcomes by suppressing the viral replication and downstream pathogenesis.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal study was reviewed and approved by Monash University Animal Ethics Committee.

KH, ET, RL, FL, JE, and SL performed experiments. KH, ET and SS wrote the manuscript KH, ET, RL, FL, SL, JE, AS, JO’L, DB, RV, and SS provided intellectual input and edited the manuscript. SS, RV, DB, and JO’L were involved in conceiving and funding this research project. SS supervised and managed the overall study. All authors contributed to the article and approved the submitted version.

This work was supported by the Australian Research Council (ARC) Future Fellowship Scheme for SS (I.D. FT120100876) and The National Health and Medical Research Council of Australia (NHMRC Project I.D. 1122506, 1128276).

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.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.