Bench-to-bedside review: Hypercapnic acidosis in lung injury - from 'permissive' to 'therapeutic'

Modern ventilation strategies for patients with acute lung injury and acute respiratory distress syndrome frequently result in hypercapnic acidosis (HCA), which is regarded as an acceptable side effect ('permissive hypercapnia'). Multiple experimental studies have demonstrated advantageous effects of HCA in several lung injury models. To date, however, human trials studying the effect of carbon dioxide per se on outcome in patients with lung injury have not been performed. While significant concerns regarding HCA remain, in particular the possible unfavorable effects on bacterial killing and the inhibition of pulmonary epithelial wound repair, the potential for HCA in attenuating lung injury is promising. The underlying mechanisms by which HCA exerts its protective effects are complex, but dampening of the inflammatory response seems to play a pivotal role. After briefly summarizing the physiological effects of HCA, a critical analysis of the available evidence on the potential beneficial effects of therapeutic HCA from in vitro, ex vivo and in vivo lung injury models and from human studies will be reviewed. In addition, the potential concerns in the clinical setting will be outlined.


Introduction
Worldwide, acute lung injury (ALI) and acute respiratory distress syndrome (ARDS) are associated with a high mortality rate (35 to 45%) [1]. Modern ventilation strategies include the use of low tidal volumes and/or limiting plateau pressure and have been shown to reduce morbidity and mortality in patients with ALI and ARDS [2][3][4]. Th e subsequent increase in arterial carbon dioxide tension (PaCO 2 ) is regarded as an acceptable side-eff ect ('permissive hypercapnia'). In current practice, mean maximum PaCO 2 and pH associated with permissive hypercapnia are around 8.9 kPa and 7.2, respectively [2], and are reported to be well tolerated as long as tissue perfusion and oxygenation are preserved and there are no contraindications [5][6][7]. Numerous studies have investi gated the eff ects of hypercapnic acidosis (HCA) in laboratory animals and humans; to date, however, it is unclear whether HCA should be considered as an acceptable adverse eff ect of lung-protective ventilation or as therapeutic by itself ('therapeutic hypercapnia'). Human trials studying the eff ect of carbon dioxide (CO 2 ) per se on outcome in patients with lung injury have not been carried out to date. After briefl y summarizing the physiological eff ects of HCA, we present a critical analysis of the available evidence on the potential benefi cial eff ects of therapeutic HCA from in vitro, ex vivo and in vivo lung injury models and from human studies.

Physiological eff ects of hypercapnic acidosis
HCA has a myriad of eff ects on many physiological processes. Th e recognition of these eff ects is important as it will aff ect the decision whether or not to allow the development of HCA in a specifi c patient. As outlined below, the fi nal eff ect of HCA on physiological functions depends on the level of hypercapnia, the context of the subject (healthy versus diseased) and many other factors. Th erefore, we will briefl y review the physiological eff ects of HCA.

Oxygenation
Th e benefi cial eff ects of HCA in increasing arterial and tissue oxygenation is evident from multiple in vivo studies [8][9][10][11][12][13][14][15][16] and has been demonstrated in healthy humans as well [17]. HCA can improve tissue oxygenation by several mechanisms. First, a rightward shift of the oxyhemoglobin dissociation curve during acute respiratory acidosis decreases the affi nity of hemoglobin for oxygen and facilitates oxygen release to the tissues (the Bohr eff ect) [18]. Second, HCA causes vasodilatation in microvessels, promoting oxygen delivery and tissue perfusion. However, high concentrations of PCO 2 (>13.3 kPa) will surpass the benefi cial vasodilatory eff ects of HCA and result in vasoconstriction [19]. Th ird, HCA

Abstract
Modern ventilation strategies for patients with acute lung injury and acute respiratory distress syndrome frequently result in hypercapnic acidosis (HCA), which is regarded as an acceptable side eff ect ('permissive hypercapnia'). Multiple experimental studies have demonstrated advantageous eff ects of HCA in several lung injury models. To date, however, human trials studying the eff ect of carbon dioxide per se on outcome in patients with lung injury have not been performed. While signifi cant concerns regarding HCA remain, in particular the possible unfavorable eff ects on bacterial killing and the inhibition of pulmonary epithelial wound repair, the potential for HCA in attenuating lung injury is promising. The underlying mechanisms by which HCA exerts its protective eff ects are complex, but dampening of the infl ammatory response seems to play a pivotal role. After briefl y summarizing the physiological eff ects of HCA, a critical analysis of the available evidence on the potential benefi cial eff ects of therapeutic HCA from in vitro, ex vivo and in vivo lung injury models and from human studies will be reviewed. In addition, the potential concerns in the clinical setting will be outlined.
improves ventilation-perfusion (V/Q) matching by poten tiat ing hypoxic pulmonary vasoconstriction [15,16]. In contrast, impaired V/Q matching has been demonstrated with HCA in patients with ARDS [20,21]. Th e diff erences in V/Q matching in these studies could be explained by the manner in which hypercapnia was achieved -inhaled CO 2 [15,16] versus low-volume (pressure-limited) ventilation-induced hypercapnia. In the latter case, atelectasis may develop, leading to increased intrapulmonary shunting [20,21]. As inhaled CO 2 theoretically results in a more uniform lung acidosis, it might be superior to low minute ventilation-induced hypercapnia in achieving improved V/Q matching and an anti-infl ammatory eff ect, as has been suggested by Sinclair and colleagues [22]. Fourth, as cardiac output is one of the major determinants of peripheral oxygen delivery, one can expect that a CO 2 -mediated increase in cardiac output augments peripheral oxygen delivery. However, an increase in cardiac output results in an increase in mixed venous oxygen tension which may lead to an increase in pulmonary shunting due to attenuation of hypoxic pulmonary vasoconstriction [21,23].

The lung Pulmonary compliance
As will be outlined below, it has been demonstrated in experimental studies that pulmonary compliance is improved by HCA. Th is may be explained by the pHmediated eff ect of HCA in improving surfactant secretion and its surface-tension-lowering properties [24,25].

Pulmonary vascular tone
Increases in pulmonary vascular tone may have particularly unfavorable consequences in patients with pulmonary hypertension. Experimental evidence is confl ict ing concerning the pulmonary vasodilatory or vasoconstrictive eff ect of HCA [9,[26][27][28][29][30]. Th ese apparent opposing eff ects may be attributable to the presence or absence of pH-buff er resulting in pulmonary vasodilatation or vasoconstriction, respectively [26,29,30].
However, clinical studies demonstrate that HCA causes an increase in mean pulmonary arterial pressure in ARDS [5,31]. Recently, Mekontso and colleagues [32] showed a lower right ventricular stroke index in patients with severe ARDS who were ventilated with higher positive end-expiratory pressure (PEEP; 10 to 11 mmHg) at a constant plateau pressure that subsequently led to HCA (pH 7.17 to 7.20, PaCO 2 9.44 to 9.98 kPa). An increase in pulmonary vascular resistance was postulated but no objective measurements were performed. Multivariate analysis demonstrated that pH, per se, and not CO 2 or PEEP, was responsible for the impaired right ventricular function [32]. Th erefore, caution is warranted with the use of 'permissive' or 'therapeutic' HCA in patients with pulmonary hypertension and depressed right ventricular function.

Cardiovascular system Cardiac output
HCA has a direct suppressive eff ect on cardiac contractility, but it can lead to a net increase in cardiac output by several mechanisms, as has been demonstrated in both animal and human studies [15,17,19,31,[33][34][35]. First, sympathetically mediated release of catecholamines due to neuroadrenal stimulation results in an increase in end-systolic volume and venous return [34,35]. In addition to an increase in heart rate, HCA induces ATPsensitive K + channel-mediated vasodilation, as has been demonstrated for the brain vasculature and coronary vessels [36,37], which could decreases left ventricular afterload. An increase of 1.33 to 1.60 kPa in PaCO 2 increases cardiac index by 14% in the critically ill and healthy mechanically ventilated patient [17,33]. In the clinical setting, however, care should be taken with patients exhibiting depressed myocardial function.

Myocardium
Acidosis has protective eff ects against myocardial ischemia-reperfusion injury [38,39]. Hydrogen ions inhibit Ca 2+ infl ux into the myocardial fi ber, which decreases myocardial contractility and oxygen demand, leading to less tissue injury during myocardial ischemia [39][40][41]. Furthermore, hypercapnia causes coronary vasodilatation, which may be of further benefi t during the period of reperfusion [40]. Th ese protective eff ects of hypercapnia can be of pivotal importance in the treatment of patients undergoing coronary artery bypass grafting with extracorporeal circulation and subsequently experiencing myocardial suppression.

Central nervous system Cerebral blood fl ow and tissue oxygenation
In the absence of intracranial hypertension, HCA may have benefi cial eff ects on the brain. Hypercapnia may improve cerebral blood fl ow by decreasing cerebrovascular resistance through dilatation of arterioles and improves tissue oxygenation, as has been demonstrated in both human and animal studies [13,[42][43][44]. Consequently, HCA has protective eff ects against cerebral hypoxic-ischemic injury, as has been demonstrated in rat models [45,46]. In a recent clinical study, it was shown that cerebral perfusion changed by 4.0 ml/100 g/minute for each 0.133 kPa change in the partial pressure of CO 2 (pCO 2 ) [44].
Hypercapnia results in cerebral vasodilatation and a subsequent rise in cerebral blood fl ow. In the presence of disturbed auto-regulation this can cause critical intra cranial pressure elevation and reduced cerebral perfusion (reviewed in [6]). Th erefore, HCA should be avoided in cases of intracranial pathology, in particular in the absence of intracranial pressure recording.

Eff ects of hypercapnic acidosis in experimental lung injury
Cell culture studies Alveolar macrophages Alveolar macrophages play a prominent role in the pathogenesis of ventilator-induced lung injury (VILI), possibly through the generation of cytokines, chemokines, nitric oxide (NO) and free radicals [47][48][49]. Upon stimulation, alveolar macrophages release proinfl ammatory cytokines and chemokines, resulting in the activation of polymorphonuclear leucocytes (reviewed in [50]). Activated polymorphonuclear leucocytes increase endothelial and epithelial permeability, resulting in tissue edema and accumulation of high-molecular-weight proteins in the airspaces (reviewed in [50]). HCA decreases cytokine release by alveolar macrophages and this eff ect appears to be primarily pH-mediated [51,52]. For instance, Lang and colleagues [53] demonstrated that CO 2 decreased lipopolysaccharide (LPS)-induced TNF-α secretion in rat alveolar macrophages in a dose-dependent fashion. Four hours after exposure to 20% CO 2 (pH 5.8), TNF-α secretion was only 50% compared to cells exposed to 2.5% CO 2 (pH 7.2). However, buff ering the culture medium pH to 7.2 completely abolished the eff ect of hyper capnia. A decrease in cell metabolic activity appeared to be responsible for the pH-induced decline in cytokine release in these cells as incubation with 20% CO 2 resulted in an approximately 40% reduction in metabolic activity and an equal reduction in LPSstimulated TNF-α secretion [54].

Alveolar type II epithelial cells
Hypercapnia has detrimental eff ects on alveolar type II epithelial cells. For example, in buff ered fetal rat type II alveolar epithelial cells, injury is caused by a hypercapniamediated increase in NO, NO synthase and 3-nitrotyrosine leading to protein tyrosine nitration [55]. Additionally, in a scratch injury model where primary type II rat alveolar epithelial cells were injured with a surgical blade, HCA caused more permanent plasma membrane defects and an impaired rate of plasma membrane repair [56]. Recently, O'Toole and colleagues [57] reported impaired pulmonary epithelial wound healing by HCA-induced diminished cellular migration through inhibition of NF-κB. NF-κB is a transcription factor responsible for the transcription of intercellular adhesion molecule (ICAM)-1 and pro-infl ammatory cytokines such as TNF-α, IL-1β, IL-6 and IL-8 (reviewed in [58,59]). NF-κB is present in the cytoplasm in an inactive form through its interaction with inhibitory proteins κB (IκB) and can be activated by multiple stimuli. Regarding the potential for HCA to delay pulmonary epithelial wound repair following mechanical injury, these studies raise concerns regarding epithelial repair in patients with ALI/ARDS undergoing 'permissive' hypercapnia.

Pulmonary artery endothelial cells
HCA has been demonstrated to inhibit NF-κB in pulmonary artery endothelial cells [60]. In isolated human pulmonary artery endothelial cells, Takeshita and colleagues [60] demonstrated that HCA suppressed IκB degradation, resulting in reduced NF-κB activity. Th is resulted in decreased expression of ICAM-1 and IL-8 with subsequent inhibition of neutrophil adherence ( Figure 1). As infl ammation plays an important role in the pathogenesis of VILI, this modulating eff ect of HCA on the infl ammatory response may further reduce lung injury during mechanical ventilation associated with hyper capnia.

Experiments in ex vivo lung preparations
In isolated perfused rabbit lungs it has been demonstrated that HCA (pH 6.84, PCO 2 15.96 kPa) prevents the development of microvascular permeability by warm ischemia-reperfusion and free-radical-mediated lung injury, possibly via inhibition of endogenous xanthine oxidase [61]. Also, less severe levels of HCA have been shown to attenuate ischemia-reperfusion injury in the isolated rabbit lung [26].
Xanthine oxidase is involved in the metabolism of purines and pyrimidines and generates superoxide and subsequently hydrogen peroxide when oxidizing hypoxanthine or xanthine to uric acid [62]. As various studies have demonstrated a possible role for reactive oxygen species (ROS) in the pathogenesis of ARDS, HCA may off er protection against ROS-mediated lung injury by inhibiting xanthine oxidase [63,64].
Broccard and colleagues [65] have demonstrated a protective role of HCA against reactive nitrogen species (RNS)-mediated lung injury by attenuating the rise in stable end-products of NO metabolism. However, the eff ects of HCA on RNS are complex. In addition to reducing RNS-mediated injury, HCA can enhance tissue nitration. Th is has been demonstrated by increased lung nitrotyrosine levels in animals treated with HCA following endotoxin injury and in animals subjected to VILI [10,66]. It appears that the net eff ect of HCA on nitrogen radicals may be benefi cial, perhaps because the oxidant pathway is more injurious. More studies are, however, necessary to clearly demonstrate this.
Both ROS and RNS are generated in response to various infl ammatory stimuli in lung endothelial, alveolar and airway epithelial cells as well as in activated alveolar macrophages and neutrophils [63]. Th is may result in oxidation, nitration and inactivation of important proteins, DNA and lipids. For example, peroxynitrite can oxidize and nitrate surfactant protein A, resulting in loss of its function [67,68]. Alterations in the function, production and composition of surfactant stimulates alveolar collapse with subsequent loss of compliance and deterioration in gas exchange. Impaired surfactant function has been reported in patients with ARDS and may aggravate respiratory failure (reviewed in [69]). As such, HCA may off er protection against lung injury by preventing surfactant nitration [70,71].
Furthermore, it has been demonstrated that HCA increases lamellar body volume density of type II pneumocytes in dog lungs. As lamellar body volume density of type II pneumocytes is known to be associated with intracellular storage and secretion of surfactant, HCA may have a stimulating role on surfactant release [24]. An indirect eff ect of HCA on surfactant function has also been envisioned. As HCA has been shown to diminish pulmonary microvascular permeability [26,56,61,70], it will prevent elevated bronchoalveolar lavage fl uid (BALF) protein concentration, which has been shown to decrease surfactant activity [72]. In the future, in vivo studies are mandatory to investigate the NF-κB can be activated by multiple stimuli, such as endotoxin (lipopolysaccharide), reactive oxygen species (ROS) and cytokines (IL-1β and TNF-α). Subsequently, phosphorylation of IκB (inhibitory proteins κB) occurs followed by its degradation, allowing NF-κB to be transported to the cell nucleus where it binds to specifi c promoter sites and activates transcription of target genes. Following activation of NF-κB, both intra-and extracellular feedback mechanism will subsequently regulate NF-κB activation, with IL-1β and TNF-α providing positive extracellular feedback. The potential mechanism by which hypercapnic acidosis (HCA) inhibits NF-κB activation appears to involve suppression of the degradation of IκB-α. Subsequently, this will result in suppressed production of IL-1β, IL-6, IL-8 and TNF-α. Suppression of intercellular adhesion molecule (ICAM)-1 and IL-8 will subsequently lead to inhibition of neutrophil adherence. HCA may also off er protection against ROS-mediated lung injury by inhibiting xanthine oxidase (XO).

Bacterial-and endotoxin-induced lung injury
In an established Escherichia coli-induced lung injury model, CO 2 was added 6 hours after E. coli instillation, allowing the development of severe pneu monia. Inspired CO 2 attenuated the fall in arterial oxygenation, the increase in peak airway pressure and the reduction in lung compliance. Moreover, histologic lung injury was reduced in the hypercapnic group compared to the normocapnic group in the presence of antibiotic therapy [73]. However, no diff erences in bacterial loads, BALF neutrophil counts, IL-6 or TNF-α levels were found. Th ese results were confi rmed by the same investi gators in the setting of evolving pneumonia-induced lung injury [12]. In this study, CO 2 was added immediately after E. coli instillation in both neutrophil-depleted and nondepleted rats, suggesting a neutrophil-independent mechanism for the eff ect of HCA [12].
HCA attenuated endotoxin-induced lung injury in rats as shown by improved arterial oxygenation, reduced alveolar infl ux of neutrophils and alveolar/tissue edema, reduced NO metabolite concentrations, improved lung compliance and improved histological indices of lung injury when given both prophylactically (before endotoxin instillation) as well as therapeutically (30 minutes after endotoxin instillation) [10]. As ALI is generally well established before the patient comes to the ICU, these data emphasize the potential clinical relevance of HCA. In contrast to these results, an increase in alveolarcapillary membrane permeability, lung wet-to-dry ratio, BALF cell counts, indices of oxidative infl ammatory reactions and lung histologic injury were observed with hypercapnia in an intravenous endotoxin-induced ALI rabbit model [74]. Possible explanations for the apparent discrepancies may be diff erences in plasma CO 2 tension (±9.7 kPa versus ±7.8 kPa), species used (rat versus rabbit), route of LPS administration (intratracheal versus intravenous) and the method of producing HCA (inhaled CO 2 versus low minute ventilation-induced hypercapnia) [10,74].

Ischemia-reperfusion lung injury
Pulmonary ischemia-reperfusion injury may occur in humans after, for instance, cardiopulmonary bypass, thrombolysis or embolectomy for pulmonary embolism and lung transplantation. In an open chest rabbit model of lung ischemia-reperfusion injury, therapeutic HCA has been demonstrated to exert benefi cial eff ects. Attenuation of the infl ammatory response was demonstrated by reduced TNF-α concentration and freeradical-mediated injury [70]. Th ese data are supported by the dose-dependent benefi cial eff ects of HCA on mesenteric ischemia-reperfusion-induced lung injury [9].

Stretch-induced lung injury
Mechanical ventilation with high tidal volumes or high peak pressure and/or low PEEP causes excessive lung stretch and shear forces, which are considered key determinants of VILI [78,79].
Increased lung compliance, higher arterial partial pressure of oxygen (PaO 2 ) and a decrease in lung injury have been demonstrated with HCA in a rabbit VILI model [11]. Stretch-induced lung injury was achieved by subjecting the rabbits to extremely high tidal volume ventilation (25 ml/kg) and zero PEEP. Although this study suggested that the eff ect of HCA was independent of tidal volume, it is of limited clinical relevance as both groups were ventilated with these high tidal volumes. However, HCA has demonstrated protective eff ects at more clinically relevant tidal volumes [8,75], though these eff ects were less impressive. Th is supports the theory that lung-protective ventilation strategies reduce VILI to a point that a protective eff ect of HCA is less detectable.
Recently, we studied the eff ect of HCA on the immune response by adding CO 2 to the inspiratory mixture in a mild VILI model in mice. Indeed, HCA decreased intrapulmonary cytokine levels, leucocyte infl ux and wet-dry ratio in a dose-dependent fashion [76,80] (Figure 2). Th ese results are in apparent contrast with the fi ndings of a previous study performed in a surfactant-depleted rabbit model [77]. Despite attenuation of BALF monocyte chemoattractant protein-1, no eff ect of HCA was found on other infl ammatory mediators, vascular permeability, lung mechanics or oxygenation. However, an inhaled CO 2 concentration of 12% was used, resulting in extremely high values of PaCO 2 (>18.6 kPa) and very low pH (<6.9), which may have infl uenced these results. Additionally, as the authors mentioned, the model is prone to atelectasis and HCA may, therefore, be less eff ective in attenuating lung injury. Furthermore, HCA may improve surfactant production and decrease surfactant surface tension. In this surfactant-depleted rabbit model these benefi cial eff ects of HCA will therefore be absent [24,25].

Human data
Th e ARDSnet trial showed that tidal volume ventilation using 6 ml/kg ideal body weight in patients suff ering from ALI/ARDS resulted in a signifi cant reduction in mortality of 8.8% compared to the use of tidal volumes of 12 ml/kg [3]. Although the PaCO 2 was slightly higher in the low tidal volume arm, the eff ects of HCA could not be separated from lung protective ventilation as an explanation for the decrease in mortality. However, a secondary analysis, using multivariate logistic regression and controlling for co-morbidities and severity of lung injury, demonstrated that HCA on day 1 (pH <7.35, PaCO 2 >6 kPa) was associated with reduced mortality in patients ventilated with 12 ml/kg but not in patients ventilated with 6 ml/kg [81]. A dose-response relationship between HCA and mortality in this group was demonstrated.
Recently, Terragni and colleagues [82] showed a reduction of tidal hyperinfl ation and an attenuation of pulmonary infl ammation during ventilation with low tidal volumes of 4.2 ± 0.3 ml/kg compared to tidal volumes of 6.3 ± 0.3 ml/kg. Th e development of HCA was eff ectively and safely managed by extracorporeal CO 2 removal. Th e question remains if morpho logical and infl ammatory parameters would even be further improved without the use of extracorporeal CO 2 removal. To date, however, human trials studying the eff ect of CO 2 per se on outcome in patients with lung injury have not been performed.
In conclusion, based on the experimental data, HCA may have direct protective eff ects against lung injury as described above. Potential mechanisms responsible for this benefi cial eff ect are complex and probably multi factorial and may be diff erent depending on the cause of lung injury. At least a reduction in infl ammatory mediators via the NF-ĸB pathway, a reduction in RNS and ROS via inhibiting xanthine oxidase and an improvement in surfactant function appear to play an important role in the prevention of lung injury by HCA.

Hypercapnic acidocis versus hypercapnia
Th e protective eff ects of HCA in lung injury models may be a function of pH or the CO 2 per se. Th is issue is of signifi cant relevance when considering the need for buff ering HCA in the clinical context.
Th e protective eff ect of HCA in experimental studies (cell culture studies [55], ex vivo studies [26] as well as in in vivo studies [83]) demonstrate that pH buff ering of HCA attenuates its lung protective eff ect. Th is suggests that the protective eff ect of HCA appears to be a function of pH, rather than elevated CO 2 per se. However, a synergistic eff ect between CO 2 and pH may exist, as HCA seems to be more protective than metabolic acidosis in the setting of ALI [26].
Besides the relevance of the type of acidosis (hypercapnic versus metabolic), the type of buff er (sodium bicarbonate versus tris-hydroxymethyl aminomethane (THAM)) seems to be of particular importance. Despite correcting the arterial pH, administration of sodium bicarbonate has the disadvantage that it may worsen intracellular acidosis. Th e CO 2 generated diff uses rapidly across cell membranes to equilibrate between intracellular and extracellular compartments, leading to intracellular acidosis [84].
Since no human trials have been performed to investigate the eff ect of buff ering during the use of 'therapeutic hypercapnia' , no advise about buff ering at the bedside can be given.

Other evidence of harm with hypercapnic acidosis
Host response to infections HCA may attenuate lung injury by reducing neutrophil activity, concentrations of key cytokines, such as TNF-α, IL-1β, IL-6 and IL-8, and the expression of ICAM-1 [10,12,51,52,54,60,70,76,85]. However, the phagocytic activity and bactericidal capacity of neutrophils and macrophages are essential for an eff ective host response to invading bacteria. As one of the most common causes of ALI/ARDS remains sepsis [86], concerns have been expressed about the possible deleterious eff ects of HCA on bacterial killing [87,88]. Recently, Costello and colleagues [89] reported a benefi cial eff ect of HCA in reducing the severity of lung injury in early and prolonged systemic sepsis. Additionally, no eff ect of HCA on bacterial load was demonstrated, providing some reassurance regarding the safety of HCA in the clinical setting of sepsis.

Inhibition of phagocytosis
Various studies have demonstrated an acidosis-mediated suppressant eff ect on the phagocytic activity of neutrophils and macrophages [90] (reviewed in [91]). Recently, it has been demonstrated that sustained HCA in the presence of prolonged pulmonary infection without antibiotic therapy increases bacterial load and worsens lung injury, mainly through inhibition of neutrophil phagocytosis. However, no diff erence in lung damage between the normocapnic and HCA group was found with co-administration of antibiotics [92]. Th is unfavorable eff ect of HCA is in contrast with other in vivo studies reporting no diff erence [93] or a modest protective eff ect of HCA in evolving and established pneumonia-induced lung injury [12,73]. It is also in contrast with the benefi cial eff ects of HCA in the setting of early and prolonged systemic sepsis. Th is suggests that the eff ects of HCA appear to depend on the site of infection as well as the stage of infection.

Reduced neutrophil respiratory burst
An acidosis-mediated suppressant eff ect on intracellular killing by reduced production of ROS by macrophages and neutrophils has also been demonstrated [90,94] (reviewed in [91]). In human neutrophils, HCA was associated with a pH-dependent decrease in intracellular oxidant production and IL-8 secretion [85].

Neuromuscular system
Prolonged hypercapnia may have negative eff ects on the neuromuscular function of the diaphragm. Degenerative changes of the diaphragm were observed after keeping rats in hypercapnic chambers for 6 weeks or more [95,96]. Th ese data are of particular importance in the clinical context, where neuromuscular function of the diaphragm plays a pivotal role in the success of weaning from the ventilator.

Milieu interne
HCA decreases hyperlactatemia in the context of global hypoxemia as well as during normoxia [15,70,97]. Prevention of hyperlactatemia is probably due to a pHmediated suppressive eff ect of HCA on lactic acid generation by decreasing cell metabolism through inhibition of glycolysis (reviewed in [98]). Subsequently, this leads to diminished cellular fuel utilization in times of ischemia. It is reasonable to think that a cellular metabolic shutdown can be benefi cial for, for example, the kidney, but it may have unfavorable eff ects for the brain

Clinical perspective and recommendations
Th e protective eff ect of 'therapeutic' HCA has been demonstrated in various lung injury models [9,10,12,70,73,74]. However, caution should be taken when extrapolating these results to the clinical setting. Importantly, the studies performed use diff erent models (that is, levels of hypercapnia, duration and timing of hypercapnia, healthy or injured lungs), with diff erent and sometimes confl icting outcomes that make comparison diffi cult. Accordingly, the optimal CO 2 level for mechanically ventilated patients with ALI is unknown. Th e concept of an optimal CO 2 concentration is essential as most physiological systems are saturable and it is therefore reasonable that an eff ective upper limit of CO 2 , a point beyond which advantages shift towards harmful eff ects, exists.
Despite these uncertainties, the potential for therapeutic hypercapnia in attenuating lung injury is promising. Th is review supports the need for studying thera peutic HCA at the bedside in patients without contraindications in a pilot setting. Diff erent methods of eliciting hypercapnia (inhaled CO 2 versus low minute ventilation-induced hypercapnia) need further investigation, especially in human studies

Conclusion
Modern ventilation strategies have demonstrated a reduction in mortality in patients with ALI and ARDS. Th e subsequent HCA is regarded as an acceptable side eff ect and is generally well tolerated. Experimental studies have reported a myriad of eff ects of HCA on many physiological processes. Despite the fact that concerns remain regarding HCA, in particular impaired bacterial killing and the inhibition of pulmonary epithelial wound repair, the potential for therapeutic HCA in attenuating lung injury is promising.