Hyperoxia and modulation of pulmonary vascular and immune responses in COVID-19.

Oxygen is the most commonly used therapy in hospitalized patients with COVID-19. In those patients who develop worsening pneumonia and acute respiratory distress syndrome (ARDS), high concentrations of oxygen may need to be administered for prolonged time periods, often together with mechanical ventilation. Hyperoxia, although lifesaving and essential for maintaining adequate oxygenation in the short term, may have adverse long-term consequences upon lung parenchymal structure and function. How hyperoxia per se impacts lung disease in COVID-19 has remained largely unexplored. Numbers of experimental studies have previously established that hyperoxia is associated with deleterious outcomes inclusive of perturbations in immunologic responses, abnormal metabolic function, and alterations in hemodynamics and alveolar barrier function. Such changes may ultimately progress into clinically evident lung injury and adverse remodeling and result in parenchymal fibrosis when exposure is prolonged. Given that significant exposure to hyperoxia in patients with severe COVID-19 may be unavoidable to preserve life, these sequelae of hyperoxia, superimposed on the cytopathic effects of SARS-CoV-2 virus, may well impact pathogenesis of COVID-19-induced ARDS.

INTRODUCTION

Hypoxemic respiratory failure is the most common reason for hospitalization of patients with COVID-19. High fractions of inspired oxygen (), either administered noninvasively by high-flow nasal cannula or invasively by mechanical ventilation, are commonly needed to maintain adequate oxygenation in patients with severely impaired lung function, as in those with COVID-19-induced acute respiratory distress syndrome (ARDS) (21). In a large clinical cohort from Italy, 89% of the mechanically ventilated patients with COVID-19 received of at least 50%, and 12% of the patients received of 100% (23).

The exposure of the injured lungs to hyperoxia might also be notably prolonged, as cohort studies of COVID-19 ARDS reported mean durations of mechanical ventilation to exceed 2 wk (47). Additional periods of noninvasive oxygen administration are typically required following extubation due to residual respiratory insufficiency. Despite the key role of oxygen in the treatment of COVID-19 and the associated pneumonia, the effects of prolonged exposure to hyperoxia in these patients are unexplored and still remain unknown.

HYPEROXIA AND CLINICAL OUTCOMES IN ARDS

In patients with moderate and severe ARDS, maintaining normoxemia [arterial partial pressure of oxygen () = 80–100 mmHg] would typically require levels of of 50% and higher, and these may pose additional injury to the lung with prolonged exposure (3). To limit the exposure to high levels of , it has been recommended by ARDS Network that arterial levels between 55 and 80 mmHg are targeted in mechanically ventilated patients. However, in a recent multicenter randomized trial that included patients with various etiologies of ARDS, targeting lower limits of this range (55–70) with conservative oxygen therapy did not improve patient survival when compared with liberal oxygen therapy with targets of 90–105 mmHg (10). Moreover, cases of mesenteric ischemia were reported in the conservative oxygen therapy group. We therefore posit that comparable lower levels of could also be deleterious in critically ill patients with COVID-19, given that intestinal ischemia is a commonly reported complication (36). On the other hand, liberal use of oxygen resulting in hyperoxemia ( of >100 mmHg) should also be avoided.

It has been documented that levels of are not always adjusted by clinicians in the ICU when hyperoxemia is found on arterial blood gas of mechanically ventilated patients (16). A proactive approach to limit even milder levels of hyperoxemia would be supported by a retrospective analysis of 10 randomized trials of ARDS which found that oxygen levels above the goal ( of >80 mmHg) were associated with increased mortality (4). Based on these data, a strategy that avoids both hypoxemia and hyperoxemia should be pursued in the setting of COVID-19 ARDS, and measurements of should guide a close titration of to limit unnecessary exposure of the lungs to hyperoxia. Where arterial sampling is not available, avoidance of peripheral oxygen saturation () levels below 90% (7) and above 96% (41) may be advised.

EVOLUTION OF HYPEROXIC LUNG INJURY

Hyperoxia is known to induce complex endothelial, epithelial, platelet, and immune cell responses in the lung that over a period of days to weeks of continuous exposure result in diffuse alveolar damage, respiratory failure, and high rates of lethality in experimental animal models (33). Exposures that are prolonged but sublethal typically result in lung fibrosis (31).

Clinically, pulmonary hyperoxia in spontaneously breathing or mechanically ventilated patients does not produce obvious respiratory decompensation and may therefore not be perceived by clinicians as an immediate threat to patient’s outcome.

The relatively slow development of various features of hyperoxia-induced lung injury contrasts to other more acute forms of injury where noxious stimuli (e.g., bacterial endotoxin, gastric acid aspiration) can rapidly trigger pulmonary edema, immune cell recruitment, respiratory failure, and shock.

In human volunteers, breathing 95% oxygen for 17 h is associated with detectable increases in alveolocapillary permeability (15). Histologically, hyperoxia drives progressive destruction of alveolocapillary membranes leading to enlargement of air spaces, alveolar hemorrhage, and vascular network remodeling (28, 32). Regression of lung capillary network (capillary rarefaction) has been demonstrated angiographically and histologically and is a result of the destruction of alveolocapillary membranes, obstruction of capillaries by microthrombi, and vascular pruning (27, 40, 46). The loss of the proportion of functional microcapillaries is accompanied by marked dilation and congestion of remaining vessels and diversion of blood flows (and whole cardiac output) through these remaining larger vessels, leading to pulmonary hypertension and intrapulmonary shunting (28, 32).

Intra-alveolar immune infiltrates are not pronounced after short periods of exposure to hyperoxia and are primarily restricted to interstitial space (interstitial pneumonitis), unlike other types of more acute lung injury. Exposure to hyperoxia, however, leads to changes in the composition of lymphoid and myeloid immune populations of the lung. Depletion of immunoregulatory immune cell populations in the lung (e.g., myeloid regulatory cells, regulatory B-cells) (24) and subsequent recruitment of proinflammatory leukocyte populations inclusive of natural killer T (NKT) cells (38) were described by our group. Alveolar macrophage phenotypes are also altered by hyperoxia (24). Dense perivascular immune cell infiltrates and a collapse of normal alveolar architecture are observed in the final stages of experimental hyperoxic injury. Given the key role of oxygen in the care of patients with COVID-19, these potential superimposed immunomodulatory and vascular effects of hyperoxia may be additive to the virus-induced lung injury.

SHARED FEATURES OF HYPEROXIC TOXICITY AND COVID-19-INDUCED LUNG INJURY

We note that many of the well-characterized and specific pathological features of hyperoxic lung injury are also present in patients with COVID-19 ARDS (Fig. 1). Recent reports in COVID-19, examining the lungs of patients with associated ARDS at autopsies, demonstrate significant enlargement of alveolar spaces due to alveolar septal destruction, significant engorgement of capillaries, capillary thrombosis, alveolar edema, alveolar hemorrhage, perivascular infiltrates, and scattered fibrosis (1, 34). The observed vascular remodeling has been attributed, at least in part, to endothelialitis, pericyte alterations, and pathological forms of intussusceptive angiogenesis (1, 11). The marked vascular dilations, seen in both hyperoxia and COVID-19 microscopically (32, 34) and even radiologically (39), are uncommon in other types of lung injury. It is also reasonable to posit that these pathognomonic vascular changes require extended time periods (e.g., days to weeks) to fully develop. This would be consistent with a typically prolonged clinical course of both severe COVID-19 pneumonia and hyperoxic lung injury. Increased recruitment and activation of unconventional T-cells in the lung—mucosal-associated invariant T cells (MAIT) cells in patients with COVID-19 (29) and NKT cells in mouse models (38)—represents another shared feature of hyperoxia-induced and COVID-19-related lung injury.

Figure 1.

Overlapping aspects of SARS-CoV-2 infection and exposure to hyperoxia in lung injury.

When characterizing pathogenesis of novel appearing pulmonary diseases, such as with COVID-19 pneumonia, it is important to consider all iatrogenic effects (hyperoxia and prolonged mechanical ventilation, in particular), as these insights clearly will dictate ongoing development of therapeutics. Due to spatial heterogeneity of pulmonary involvement in COVID-19, it is likely that segments of the lung that are severely damaged by SARS-CoV-2 and largely consolidated experience lesser exposure to hyperoxia, whereas segments that are not consolidated and well ventilated are more prone to hyperoxia- and ventilator-induced lung injury. The reported vascular and immunological features of COVID-19 ARDS may reflect, at least in part, changes ascribed to prolonged hyperoxia that are superimposed on the original COVID-19-related disease, or develop in parallel in the well-ventilated segments of the lung.

FUTURE DIRECTIONS AND POTENTIAL THERAPEUTIC TARGETS

We propose that animal models of severe SARS-CoV-2 infection should also incorporate hyperoxic exposure to better replicate molecular events occurring during the treatment of severe COVID-19 in humans and to delineate whether hyperoxia influences mucosal and systemic immune responses to coronaviruses.

Hyperoxia was previously found to modulate acute lung injury in several experimental models of pneumonia, having deleterious effects. Exposure of mice to 60% significantly augments acute lung injury induced by low-dose intratracheal lipopolysaccharide, by increasing alveolocapillary barrier permeability, neutrophil recruitment, and macrophage activation and reducing the numbers of regulatory T-cells (5). In a mouse model of Legionella pneumonia, prolonged exposure to >90% led to increased mortality, and hyperoxia was found to increase alveolocapillary permeability and promote epithelial cell and macrophage apoptosis (44). Adverse effects of hyperoxia ( 95%) upon alveolar macrophage bactericidal function were documented in a mouse model of Klebsiella pneumoniae infection (9), and the observed macrophage dysfunction was linked to a decreased granulocyte-macrophage colony-stimulating factor (GM-CSF) production by epithelial cells during hyperoxia (8). Secondary bacterial pneumonias are relatively common in patients with COVID-19 ARDS (17). Although the factors predisposing to pneumonia in these patients could be numerous (e.g., sedation, prolonged mechanical ventilation), a prolonged exposure to hyperoxia and its effects upon innate immunity may play a contributory role.

Angiotensin-converting enzyme-2 (ACE2), expressed by several cell subsets in the lung including alveolar epithelial cells, endothelial cells, and pericytes, is a key component of COVID-19 pathogenesis. SARS-CoV-2 uses ACE2 as a receptor for cellular entry, and binding of the virus leads to ACE2 internalization, resulting in decreased cell surface expression (13, 26, 45). ACE2 and the product of its enzymatic activity, angiotensin 1–7, were found to be protective in an animal model of hyperoxic lung injury via mechanisms involving inhibition of the proinflammatory NF-κB pathway and activation of the antioxidant Nrf2/HO-1/NQO1 pathway (22). Administration of ACE2 agonist diminazene aceturate ameliorated lung injury, whereas ACE2 inhibitor MLN-4760 potentiated lung injury. By decreasing ACE2 expression and activities, SARS-CoV-2 may further exacerbate lung susceptibility to hyperoxia-induced injury, further highlighting the importance of limiting hyperoxia exposure.

Other common respiratory viruses (e.g., RSV, hMPV) have been also found to decrease Nrf2 (nuclear factor erythroid 2-related factor 2) activation and thereby increase lung susceptibility to oxidative stress. Nrf2 inducers and antioxidant enzyme mimetics are therefore proposed as potential therapies in viral pneumonia (30a), and such an approach could be applicable in COVID-19 ARDS.

Other methods to protect the lung from deleterious effects of prolonged hyperoxia may involve targeting immune cells and inflammatory mediators. We have found that deletion of NKT cells (38), and targeting certain lipid inflammatory mediators (e.g., lysophosphatidic acid) (37), ameliorates hyperoxia-induced lung injury in mice, and we have established a pathogenic role for the NKT cell-autotaxin-lysophosphatidic acid axis in hyperoxia (37). These same therapeutic strategies could be applicable in COVID-19 ARDS, an approach strongly supported by significantly elevated levels of lysophosphatidic acid in the plasma of these patients and lysophosphatidic acid being identified as a disease marker of COVID-19 in a recent multiomic screen (42).

Recent data also demonstrate that prolonged exposure to hyperoxia increases epithelial expression of a putative SARS-CoV-2 coreceptor, namely, TMPRSS11D, in lungs of neonatal mice and human lung tissue. These findings also suggest that chronic and/or intermittent oxygen administration in individuals with chronic lung disease might also facilitate infection by SARS-CoV-2 (35).

Lung parenchymal stress, injury, or cell death that is imposed by viral infection, hyperoxia, and mechanical ventilation leads to release of intracellular purines including adenosine 5′-triphosphate (ATP) and adenosine 5′-diphosphate (ADP) into the microenvironment and circulation where these molecules act as “danger signals” (25). Binding of ATP and ADP to P2 purinergic receptors on immune cells and platelets has largely proinflammatory (e.g., P2X7-mediated) and prothrombotic (P2Y12-mediated) effects. Enzymatic breakdown of ATP to adenosine 5′-monophosphate (AMP) and adenosine by concerted actions of ectonucleotidases CD39 (ENTPD1) and CD73 (NT5E) terminates P2 receptor signaling and leads to an opposing, P1 receptor-mediated adenosinergic signaling that is largely anti-inflammatory and tissue protective (20).

Hence, the availability of the ATP-hydrolyzing enzymes CD39 and CD73 and adenosine receptors in the lung vasculature and on immune cells is crucial for return to homeostasis and limiting further injury. Adenosinergic signaling via adenosine receptor 2B (ADORA2B) has been shown to mitigate both ventilator- (18) and oxygen-induced lung injury (14) in experimental models, and such an approach to bolster adenosinergic responses may therefore be protective in COVID-19.

The expression patterns of certain protective purinergic elements (CD39, CD73, ADORA2A, ADORA2B) are known to be enhanced by hypoxia. As an example, hypoxia triggers Sp-1 (specificity protein 1)-dependent induction of CD39 (19), hypoxia-inducible factor 1-α (HIF1α)-dependent induction of ADORA2B (18) and CD73 (43), and HIF2α-dependent induction of ADORA2A (6). It remains to be determined whether prolonged exposure to hyperoxia as seen in COVID-19 adversely impacts the expression of these tissue-protective purinergic systems linked to endothelial, epithelial, or immune cell homeostasis. Given this mechanism is quite likely, it is of interest to note that HIF1 activators have been proposed as another potential intervention to mitigate against SARS-CoV-2-induced lung injury, possibly promoting local generation of adenosine and expression of adenosine receptors and also diminishing the adverse effects of hyperoxia (2, 30).

CONCLUSIONS

There appear to be overlapping pathological effects upon the lung of SARS-CoV-2 infection and exposure to high concentrations of inhaled oxygen. Hyperoxia may exacerbate the local cytopathic effects of the virus. Therefore, judiciously limiting the exposure to hyperoxia to the lowest levels that are absolutely required may be prudent. We propose that the extensive studies of hyperoxia and lung injury conducted over several decades, and resulting in the identification of pathogenic, mechanistic pathways, will inform future therapeutic efforts in COVID-19 and ARDS.

GRANTS

This study is supported by a Clinical Investigator Award from National Heart, Lung and Blood Institute (K08HL141694) (to D. Hanidziar), by a grant from the Department of Defense (W81XWH-16-1-0464, DAMP-Mediated Innate Immune Failure and Pneumonia after Trauma – PR1511953) (to S. C. Robson), and by the Jane Siegel BIDMC Research Fund.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

D.H. drafted manuscript; S.C.R. edited and revised manuscript; S.C.R. approved final version of manuscript.