Introduction

Permissive hypercapnia is the acceptance of elevated CO2 that results from the limitation of inspiratory tidal volume (or pressure) in patients receiving mechanical ventilation [1]. In this context, the most important principle of supportive therapy is to minimize lung injury induced by mechanical stretch, i.e., ventilator-associated lung injury. While lessening tidal volume (V T) is associated with improved patient survival [2], hypercapnia may have additional roles [3, 4]. First, the ‘permissive’ hypercapnia resulting from V T reduction may coincidentally improve survival, an effect that may be additive to the protective effects of reducing V T [5]. Second, the deliberate elevation of CO2, therapeutic hypercapnia [6], has been hypothesized to directly protect against organ injury, and this has been demonstrated experimentally [7–15]. However, while salutatory effects of CO2 have been identified, there are potentially serious adverse effects. These include pulmonary or intracranial hypertension [16, 17], as well as more recently described molecular and cellular alterations [18–24], such as injury from nitrogen-derived oxidants [22–24], and adverse protein modification of important enzyme systems (e.g., Na/K-ATPase) [25]. Such adverse events may explain recent reports of worsening progression of sepsis associated with hypercapnia [14, 20, 26].

Balancing adverse effects, a series of cellular and molecular processes that are considered beneficial in the context of injury have been observed with hypercapnia. Such processes include induction of Axl (an anti-apoptotic protein) in cultured umbilical vein endothelial cells [27], attenuation of NF-κB signaling in cultured pulmonary artery endothelial cells [28], inhibition of interleukin-8 release from isolated neutrophils [29], inhibition of xanthine oxidase ex vivo [7], activation of K-ATPase channels in vascular smooth muscle [30] and the in vivo inhibition of cytokine release [8, 12].

Whether any specific mechanism is pathogenic or protective may depend on the nature and timing of the injury and on the inflammatory milieu. For example, in experimental models of sepsis, hypercapnia can exert net harm [26]—or net benefit [14]—depending on trajectory of the infection. Indeed, in short-term sepsis models, nitrotyrosine formation (conventionally considered harmful) was increased with hypercapnia [19, 20, 31]. However, in ventilator-induced lung injury without sepsis, the reported effects of hypercapnia have so far been beneficial [8, 9, 11]—or neutral [32]—but not adverse.

In this study we sought: (1) to confirm that hypercapnic acidosis attenuates in vivo ventilator-induced lung injury, (2) to determine whether additional features of biological plausibility supported such protection (e.g., dose-response, time series, accompanying inflammatory profile) and (3) if hypercapnia would protect lung tissue biochemistry from changes associated with injurious ventilation.

Materials and methods

All experiments were performed according to the protocols approved by the Animal Care and Use Committee of the Research Institute at the Hospital for Sick Children, Toronto, Canada.

Animal experiments

C57BL/6 female mice (20–25 g, Charles River, St. Constant, Canada) were anesthetized (ketamine/xylazine 150/15 mg kg−1, IP), a tracheostomy was inserted, and they were ventilated using a computer-controlled small animal ventilator (SCIREQ, Flexivent, Montreal, Canada) [33] with the following settings, which do not cause vivo ventilator-induced lung injury (VILI) in the in vivo mouse lung over the time interval studied, in the absence of a pre-ventilation injury: V T 10 ml kg−1, PEEP 1.0 cmH2O, frequency 135 min−1, FiO2 0.75 and FiN2 0.25. Core temperature was maintained (36.5–38.0°C) and supplemental fluids and anesthesia administered as required throughout the experiment. Lung elastance was measured at baseline and hourly thereafter. Evans Blue (Sigma-Aldrich, Oakville, ON, Canada; 30 ml kg−1, IV) was given, and animals were randomized to receive injurious ventilation (V T 45 ml kg−1, PEEP 0 cmH2O, frequency 30–35 min−1) or control ventilation (V T 12 ml kg−1, PEEP 1 cmH2O, frequency 135 min−1), in the setting of either normocapnia (FiO2 0.75, FiCO2 0, balance N2) or hypercapnia (FiO2 0.75, FiCO2 0.12, balance N2). All of these ventilator settings are known, from pilot data, to result in normocapnia where FiCO2 is zero. Arterial blood gas measurements were performed at the end of each experiment and the animals killed by exsanguination. Bronchoalveolar lavage was followed by removal of the lungs en bloc, and the lung tissue retained (−80°C). Additional animals were studied to examine the effects of increasing concentrations of FiCO2 (0.05, 0.12, 0.25) with FiO2 0.75 and balance N2 in all cases. Pulmonary microvascular permeability was determined using a modification of the Evans’ Blue technique as previously described [34], where BAL concentration of Evans blue corrected for the presence of heme pigments [35].

Mediator assays

Bronchoalveolar lavage samples were analyzed for cytokines (interleukin-6, IL-6; keratinocyte chemoattractant factor, KC; macrophage chemoattractant protein-1, MCP-1), the soluble cell adhesion molecules (soluble vascular cell adhesion molecule, sV-CAM; soluble E-selectin) and matrix metallo-proteinase-9 (MMP-9), were measured with commercial ELISA kits (LINCOplex Kits, LINCO Research, St. Charles, MI), according to manufacturer’s instructions. In addition, TNFα bioactivity in BAL was measured with a murine fibroblast cytotoxicity assay [36], where cell death was assessed using the methylthiazoletetrazolium method [37]. Stable hydrolysis products of prostaglandins, leucotrienes and lipoxins were measured in total lung tissue by liquid chromatography–tandem mass spectrometry (LC–MS/MS) using an API4000 triple-quadruple mass spectrometer (MDS SCIEX, Concord, ON, Canada) [38].

Histopathology and immunohistochemistry

Right lungs were pressure fixed (10% buffered formalin, 20 cmH2O), serially sectioned from apex to base, randomly selected, stained with hematoxylin-eosin and examined by an observer blinded to experimental group. Paraffin tissue sections were deparaffinized, rehydrated, microwaved for antigen retrieval and incubated (3% H2O2) to quench endogenous peroxidase. Nonspecific adsorption was minimized (incubation with 10% goat serum) and the sections then incubated with primary anti-MPO polyclonal antibody, polyclonal anti-3-NT or with control solutions (buffer alone; non-specific purified rabbit IgG or blocking peptides). Specific labeling was detected with biotinylated goat anti-rabbit IgG, followed by incubation with avidin DH and biotinylated horseradish peroxidase and diaminobenzidine tetrahydrochloride as chromogen. Counterstaining was with hematoxylin.

Cyclooxygenase and nitrotyrosine protein

COX-2 and 3-nitrotyrosine (3-NT) protein expression was examined using Western blot analysis of lung tissue using polyclonal rabbit anti-mouse antibodies (Cayman Chemicals, Ann Arbor, MI), incubated with horseradish peroxidase-conjugated anti-rabbit IgG (BioRad, Mississauga, ON, Canada) and visualized by enhanced chemiluminescence (ECL, Amersham Biosciences, Piscataway, NJ); the data were normalized against the housekeeping protein GAPDH.

Cyclooxygenase mRNA

COX-2 mRNA expression was assessed on Northern blots of total mRNA isolated from lung tissue using TRIzol (Invitrogen, Burlington, ON, Canada), hybridized with a [32P]dCTP-labeled cDNA encoding mouse COX-2. Following stringent washing, COX-2 mRNA was visualized by autoradiography, and blots were stripped and reprobed with GAPDH for normalization.

Statistics

Data were analyzed using SigmaStat (version 2.0; Jandel Corporation, San Rafael, CA). Group comparisons were performed with ANOVA followed by Student-Newman-Keuls or t tests for parametric data, and with ANOVA on ranks followed by Dunn’s tests for non-parametric data. Significance was set at P < 0.05. Results are expressed as mean ± standard deviation (SD).

Results

Lung pathophysiology

Baseline animal weight and respiratory system compliance were comparable among the groups (Table 1). The blood gas and acid-base data at the end of the experiment confirmed hypercapnic acidosis in both hypercapnia groups, mild metabolic acidosis in the injurious ventilation/normocapnia group and close to normal acid-base status with protective ventilation/normocapnia (Table 1).

Table 1 Blood gases and acid-base variables
Full size table

Elastance was similar among all groups at baseline and became significantly increased in the two injurious ventilation groups over time, but to a significantly lesser extent in the injurious ventilation/hypercapnia group (Fig. 1a). The order for final elevation in elastance was: injurious ventilation/normocapnia > injurious ventilation/hypercapnia > protective ventilation/normocapnia ≈ protective ventilation/hypercapnia. The same order of injury was reflected in lung microvascular leakage (BAL Evans’ Blue; Fig. 1b), as well as in the histologic appearance (Fig. 2a) and intensity of myeloperoxidase staining (Fig. 2b).

Fig. 1
figure 1

Serial elastance and microvascular leakage: injurious ventilation resulted in a progressive worsening in respiratory system elastance (vs. non-injurious ventilation; a). Addition of 12% CO2 to the inspired gas resulted in a smaller increase in elastance versus normocapnia. Injurious ventilation resulted in greater microvascular leak (BAL Evans’ Blue concentration; b), which was attenuated by hypercapnia. Data are means ± SD n = 13 injurious ventilation, n = 8 protective ventilation. *Injurious ventilation normocapnia versus all other groups (two-way repeated measures ANOVA on ranks, P < 0.001; Dunn’s test, P < 0.05)

Full size image
Fig. 2
figure 2

Histology and myeloperoxidase: representative histologic sections (a hematoxylin and eosin (H&E) ×400; b myeloperoxidase immunoreactivity counterstained with H&E, ×400) indicates greatest injury in the injurious ventilation/normocapnia sections, less injury with injurious ventilation/hypercapnia and normal appearances with non-injurious ventilation (whether normocapnia or hypercapnia). (HM hyaline membrane, IC infiltrating cells, E edema.) MPO-immunoreactive (brown staining) cells include neutrophils and pulmonary alveolar macrophages (b)

Full size image

In an additional group of experiments, a dose-response effect was also apparent wherein higher concentrations of added CO2 (0 vs. 5, 12 or 25% inspired CO2) were associated with incrementally better protection against injury in terms of elastance (Fig. 3a) and microvascular leakage (Fig. 3b).

Fig. 3
figure 3

Dose-response effects of CO2 on injury: injurious ventilation resulted in progressive worsening of respiratory system elastance during 3 h of ventilation (a). There was a dose-response effect whereby higher concentrations of inspired CO2 resulted in progressively more effective attenuation of injury over time (a). Similarly, higher doses of inspired CO2 resulted in less microvascular leakage at the end of 3 h of injurious ventilation (b). Data are means ± SD, 4–6 per group. *P < 0.05 (ANOVA on ranks)

Full size image

Inflammatory mediators

The BAL fluid concentrations of several cytokines (interleukin-6, IL-6; keratinocyte chemoattractant factor, KC; macrophage chemoattractant protein-1, MCP-1) (Figure 1S), the soluble cell adhesion molecules (soluble vascular cell adhesion molecule, sV-CAM; soluble E-selectin) (Figure 2S) and matrix metallo-proteinase-9 (MMP-9) (Figure 2S) all mirrored the above patterns. Because TNFα is known to exhibit phasic responses to injurious ventilation [39], time-course experiments were performed to assess the activity over time, and hypercapnia attenuated both peaks (45 and 180 min) of TNFα activity (Figure 3S).

Cyclooxygenase and eicosanoids

The above cytokines mirrored the physiologic impact of injurious ventilation and the patterns of protection afforded by hypercapnic acidosis. A number of these mediators regulate the expression of cyclooxygenase (COX) enzymes, which are important proximal regulators of prostanoid synthesis. Prostanoids have been progressively implicated in the pathogenesis of acute lung injury [38], and because the acutely responsive (i.e., inducible) eicosanoid-producing enzyme in the lung is COX-2, the COX-2 enzyme protein level was measured. COX-2 protein was increased by injurious ventilation, and this increase was less in the presence of hypercapnia; as well, hypercapnia appeared to reduce the baseline COX-2 levels (Fig. 4a). Overall the order of lung tissue COX-2 protein levels was: injurious ventilation/normocapnia > injurious ventilation/hypercapnia > protective ventilation/normocapnia > protective ventilation/hypercapnia. In order to determine whether the increase in COX-2 protein represented new COX-2 synthesis, the expression of COX-2 mRNA was measured and found to be overwhelmingly induced by injurious ventilation; this effect was strongly attenuated by hypercapnia (Fig. 4b).

Fig. 4
figure 4

Tissue cyclooxygenase protein and mRNA: tissue COX-2 protein levels (Western blot) were increased by injurious ventilation, and the increase was attenuated by hypercapnia (a). A similar pattern was observed for tissue COX-2 mRNA levels (Northern blot, b). †Injurious ventilation normocapnia versus all other groups (ANOVA, P < 0.001; Student-Neuman-Keuls test P < 0.05); *hypercapnia versus all other groups (ANOVA, P < 0.001; Student-Neuman Keuls test P < 0.05)

Full size image

However, despite the attenuation of COX-2 protein levels by hypercapnia, the levels of COX-2 products were only moderately suppressed. In addition, it was noted that arachidonic acid, the substrate for eicosanoid production, was also significantly lower in the presence of hypercapnia (Figure 4S).

Lung tissue nitrotyrosine

Sepsis models have frequently demonstrated increases in tissue nitrotyrosine levels. Injurious ventilation was associated with a modest increase in nitrotyrosine immunostaining (Fig. 5a) and increased detection of 3-nitrotyrosine on Western blot (Fig. 5b). However, despite otherwise overall protective effects, hypercapnia significantly increased the lung tissue nitrotyrosine load.

Fig. 5
figure 5

Lung tissue nitrotyrosine: injurious ventilation was associated with a modest increase in nitrotyrosine immunostaining (a) and increased detection of 3-nitrotyrosine on Western blot (b). However, despite otherwise overall protective effects, hypercapnia significantly increased the lung tissue nitrotyrosine load. *P < 0.05 hypercapnia versus normocapnia (unpaired t test)

Full size image

Discussion

The current data confirm the protective effects of therapeutic hypercapnia initiated concurrently with injurious ventilation in an in vivo model of ventilator-induced lung injury, and characterize its effects on multiple mediators, including cytokines, eicosanoids and tissue nitration. The enhancing effect of hypercapnia on oxygenation may be due to amelioration of injury or to a direct enhancement of PaO2 (previously reviewed) [6].

Cytokine responses

The inflammatory cytokines assayed in the current study were chosen because of their strong association with ventilator-induced lung injury [40, 41] and the reliability of murine assays. The cytokine response mirrored the injury responses, and the TNFα responses were studied at several time points because of the known biphasic time course during high tidal volume ventilation [42]. In the current study, both the early and the late phases of TNFα elevation were inhibited by hypercapnia, such that the profile was similar to the control (un-injured) protocols in the current and previous [39] reports. The remaining cytokine responses were closely aligned with each other, and with the overall injury pattern, consistent with previous reports [40, 41]. The impact of hypercapnia was uniform among all assayed mediators.

Eicosanoids, injury and hypercapnia

Several of the cytokine mediators are important regulators of COX expression, which in turn regulates prostanoid synthesis. We measured tissue eicosanoid levels and found that hypercapnia attenuated the elevation of ‘core’ eicosanoid molecules (i.e., arachidonic acid, PGF2α, thromboxane B2). The BAL samples—assayed for a broader range of eicosanoids—revealed that although almost all classes were elevated by injurious ventilation (i.e., paralleled the degree of injury), the COX-2-dependent eicosanoids poorly paralleled the protection observed with hypercapnia.

In general, inflammatory prostanoids appear to be important in acute lung injury [43]. COX-2 is inducible and its expression increases following exposure to inflammatory mediators such as growth factors, cytokines and endotoxin [44], as well as stretch [38]. IL-1β, a proximal pro-inflammatory cytokine that is ubiquitously increased in ventilator-induced lung injury [40, 41], directly induces COX-2 via NF-κB [45]. Indeed, increased expression of COX-2 is mediated by several signaling pathways including cAMP, MAP-kinase and NF-κB [46], the latter being activated in ventilator-induced lung injury [47]. However, in the current study, despite hypercapnia-mediated suppression of COX-2 induction, substantial quantities of distal prostanoids were measured, indicating that the constitutive COX-1 may have a substitutive role in prostanoid generation in this model. Indeed, since the decreases in PGF2α and TXB2 are similar to the decrease in the COX substrate arachidonic acid, we cannot differentiate whether the decreases in prostanoid production under hypercapnia were due to the suppression of COX-2 or to suppression of arachidonic acid levels.

Mechanisms of hypercapnic protection

A small number of studies point to potential mechanisms of protection from injury associated with hypercapnic acidosis. The release of cytokines is an ubiquitous association of the development of almost all forms of acute lung injury [40, 41, 47]. In several studies demonstrating hypercapnic protection against lung injury, attenuation of cytokine release has been a consistent feature [8, 10, 12]. However, in such cases, it is not clear whether the inhibition of inflammatory mediators is a cause—or a consequence—of the lesser injury that occurred in the context of hypercapnia. One study has examined the release of pro-inflammatory cytokines in cultured leukocytes, in the absence of additional injury [29]. That study demonstrates that hypercapnia inhibits the release of interleukin-8 from stimulated leukocytes; such inhibition appears to be on the basis of increased intracellular H+, is dependent on intracellular carbonic anhydrase and is the converse of the effects observed with hypocapnic alkalosis [29]. Despite the consistency of the association between lung injury and elevated cytokines, there are significant doubts about the pathogenicity of cytokines in the development of ventilator-associated lung injury [48]. Thus, exploration of the effects of hypercapnia must extend beyond individual pro-inflammatory cytokines.

The IκB/NF-κB signaling pathway is a central pathway linking signal transduction with transcription regulation in sepsis [49] and mechanical stretch [47]. In the context of endotoxin-induced lung injury, hypercapnic acidosis has been shown to inhibit the translocation of NF-κB by stabilization of IκB [28]. Several additional pathways mediate ventilator-induced lung injury, including apoptosis and oxidant-induced injury; limited studies suggest hypercapnic acidosis may also affect such pathways [27].

Mechanisms of harm

In parallel with explorations of potential mechanisms of protection, several groups have reported on the molecular basis for how hypercapnia may cause harm. In the absence of sepsis—or any injury—increased levels of CO2 result in increased peroxynitration of surfactant protein-A, and the increased level of nitration is directly correlated with functional impairment (i.e. ability to aggregate lipids) [18]. Lange et al. [19] reported that increased CO2 resulted in higher levels of nitration of cultured epithelial cells—a phenomenon that was accentuated in the presence of pro-inflammatory cytokines—and suggested that the source of the increased nitration was linked to upregulation of nitric oxide synthase. In addition, CO2 promotes nitrotyrosine at the chemical level in that binding to peroxynitrite (ONOO) it forms an adduct (ONOOCO2 −), which promotes tyrosine nitration even more than peroxynitrite alone. Therefore, even if the levels of NOS, NO and superoxide are not elevated, the addition of CO2 will increase the propensity for nitrotyrosine formation. The same pattern of concern was reproduced in vivo in a follow-up study by Lange et al. [20], where hypoventilation resulted in worsening of LPS-induced lung injury, with an increased intensity of tissue nitration. Tissue nitration is usually an adverse development, as examples from the literature indicate that it is usually associated with a significant lack of protein function whether in the setting of hypercapnia [18] or otherwise. However, it is impossible to determine net benefit or harm from our overall measures of tissue nitration, since one could speculate that nitration and hence inhibition of pro-inflammatory or pro-apoptotic proteins might confer net benefit on the lung. In addition, peroxynitrate and nitrotyrosine have been reported to influence src family kinase activity, thus modifying intracellular signaling [50, 51], potentially influence lung injury. Finally, while tyrosine nitration is a marker of peroxynitrite formation, peroxynitrite itself has many effects beyond nitration of tyrosine. CO2 by forming the ONOOCO2 − adduct enhances nitration of tyrosine. While on balance nitrotyrosine would be expected to have negative effects on protein functions, elevated levels of peroxynitrite may be far more adverse in terms of, for example, lipid peroxidation, depletion of glutathione, inhibition of mitochondrial respiration and direct DNA damage in the absence of ONOOCO2 − formation [52].

Limitations and future studies

The in vivo mouse is naturally limited in terms of immediate applicability. Murine models, while invaluable for assay availability, knowledge of the genome and the ability to develop genetically modified models, possess fundamentally different physiological characteristics. In our mouse model, it should be noted that the PaO2 was higher than would be anticipated clinically; while it is not possible to rule out an additive or synergistic effect of hyperoxia with injurious ventilation, hyperoxia appears innocuous in the adult mouse (in contrast to neonatal rats). Nonetheless, they provide insights that are sometimes not otherwise possible. Subsequent investigation in this area could involve longer term survival or in vivo imaging [53] that could identify time course, trajectory and tissue-specificity of both inflammation and evolution of lung injury, as well as its inhibition by hypercapnia. For maximal relevance to clinical practice, application of hypercapnia after injury has become established will be important. The effects of peroxynitrate and tissue nitration are complex, and the net effect of increased nitrotyrosine formation in the hypercapnic setting is unknown. Considerable future work will be needed to determine the key target proteins, whether these proteins are activated or inhibited, and the net effect of these activity changes on lung injury.

Conclusion

The current report confirms that hypercapnia provides in vivo protection in a model of ventilator-induced lung injury, illustrates the pattern of responses of important regulatory mediators, identifies inconsistency regarding the eicosanoid profile and identifies the potentially harmful generation of nitrotyrosine formation. Thus, clinicians and those designing clinical trials need to be aware of the potential for detrimental effects when using hypercapnia, even in the absence of sepsis, in order to balance benefits versus harm with this approach.