Mild Endotoxemia during Mechanical Ventilation Produces Spatially Heterogeneous Pulmonary Neutrophilic Inflammation in Sheep

• ❖ Endotoxemia causes pulmonary inflammation, but whether this is homogeneous throughout the lungs during large tidal volume ventilation is not known

• ❖ In sheep with mild endotoxemia and large tidal volume ventilation within ranges used clinically, neutrophil inflammation occurred heterogeneously, related to local aeration and perfusion

• ❖ In this situation, regional pulmonary injury might occur by increased mechanical stress and localized inflammation

ENDOTOXEMIA and mechanical ventilation are frequently associated in clinical practice.1–3Acute lung injury (ALI) due to endotoxemia has been characterized as a generalized process of lung inflammation.4–6In contrast, a key element in ALI is the heterogeneous spatial distribution of lung aeration.7–10Studies to date indicate that endotoxin exposure either preceding or accompanying injurious mechanical ventilation augments lung inflammation and injury.11–13Indeed, isolated cell, ex vivo  and in vivo  small animal studies showed that exposure to high tidal volumes and endotoxin leads to increased production of neutrophil-attracting cytokines,11–16increased neutrophil counts in the bronchoalveolar lavage,13,16and worsening of gas exchange and respiratory mechanics.12,13,17Although these investigations illustrate basic inflammatory processes induced by the combination of excessive lung strain and endotoxin, they do not allow for a straightforward translation of those findings to specific regions of the heterogeneously aerated and perfused large animal lung. Specifically, it is not known whether, in the setting of endotoxemia occurring in a large animal whose lungs present mechanical heterogeneity similar to that in humans, inflammation during mechanical ventilation would develop homogeneously or heterogeneously either in nondependent areas potentially subjected to overdistension18,19or in dependent regions subjected to cyclic recruitment and/or atelectrauma.9,20Also unknown is whether there is a dependence of inflammation on regional lung aeration or perfusion. Furthermore, it is unclear whether any inflammatory changes can be detected at less injurious levels of endotoxemia and mechanical ventilation than the exaggerated conditions used in most studies.

A study of these regional effects is relevant because aeration heterogeneity is regarded as a key factor to regionally amplify mechanical forces during mechanical ventilation21and produce lung injury.7–9Understanding such heterogeneities could also assist clinical decisions on the optimal use of strategies aimed at minimizing lung injury by reducing heterogeneity of aeration, such as the open lung approach22and high-frequency ventilation,23and redistributing perfusion, such as inhaled nitric oxide24and noisy ventilation,25,26according to the degree of lung injury severity, which is itself highly correlated with the percentage of recruitable lung.27 

Positron emission tomography (PET) imaging after intravenous injection of ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) has been used to quantify pulmonary inflammation in vivo  and noninvasively in the nontumoral lung.6,28–30Neutrophils are an essential component of ALI due to either endotoxin or mechanical ventilation. The increases in both neutrophil numbers and activity contribute to increased ¹⁸F-FDG uptake during lung inflammation.31,32Because ¹⁸F-FDG-PET imaging can detect changes in lung neutrophil kinetics before their migration into the alveolar space,6it has been proposed as a potentially powerful tool to study the early phases of neutrophil trafficking and state of activation during ALI. In line with such arguments, we showed that whole-lung ¹⁸F-FDG uptake is increased after 90 min of injurious mechanical ventilation, correlates with neutrophil infiltration, and potentially precedes lung dysfunction.29 

In this study, we used a sheep model of large tidal volume mechanical ventilation bounded by clinically accepted pressure limits, designed to promote lung derecruitment (positive end-expiratory pressure [PEEP]= 0) and maximal inflation within accepted plateau pressures (P^plat= 30-32 cm H²O) for 2 h. This strategy was chosen to deliberately promote, in a healthy lung, regional heterogeneity of aeration within clinically observed alveolar pressures ranges and not to test specific ventilatory settings applicable to a clinical condition. By using this model and methods of regional pulmonary ¹⁸F-FDG kinetics modeling, we sought to test whether combining mechanical ventilation of a heterogeneously expanded lung with mild endotoxemia could yield measurable and anatomically localized levels of neutrophilic inflammation.

The experimental protocols were approved by the Massachusetts General Hospital Subcommittee on Research Animal Care (Boston, Massachusetts). Twelve sheep (22.3 ± 5.9 kg, approximately 3 months old) were fasted overnight and premedicated with intramuscular ketamine (4 mg/kg) and midazolam (2 mg/kg). After intravenous induction of anesthesia with ketamine (4 mg/kg), an endotracheal tube was inserted. General anesthesia was maintained with a continuous infusion of propofol and fentanyl titrated to heart rate and blood pressure. Pancuronium 0.1 mg/kg at induction and repeated every 90 min (0.02-0.04 mg/kg) was used for muscle paralysis. Each sheep was placed supine in the PET scanner (Scanditronix PC4096; General Electric, Milwaukee, WI) with the caudal end of the field of view just superior to the dome of the diaphragm. After a recruitment maneuver, they were mechanically ventilated with P^plat= 30-32 cm H²O, PEEP = 0, inspired O²fraction = 0.3 (adjusted to an arterial O²saturation >0.88), inspiratory-to-expiratory time ratio = 1:2, respiratory rate = 18 breaths/min or higher to maintain the arterial carbon dioxide pressure (Paco²) between 32 and 45 mmHg. If Paco²were less than 32 mmHg when respiratory rate = 18 breaths/min, a variable dead space was added to the breathing circuit aiming at that Paco²range. Physiologic data were collected, and PET scans were acquired both at the start of the protocol and after 2 h of mechanical ventilation, except for the FDG scan performed at the end of the study. After the initial set of scans, six sheep (endotoxin group) received a continuous 10 ng · kg⁻¹· min⁻¹intravenous infusion of endotoxin (Escherichia coli  O55:B5, List Biologic Laboratories Inc., Campbell, CA), whereas six did not (controls). Studies were performed sequentially in each group.

The experimental system and methods of analysis have been described in detail.29,30,33–35Scans consisted of 15 cross-sectional slices of 6.5-mm thickness over a 9.7-cm-long axis field, providing three-dimensional data for an estimated 70% of the total lung volume.35For each slice, resulting images, consisting of 128 × 128 voxels of 6 × 6 × 6.5-mm size, were low-pass filtered to 12 × 12 mm to a final volumetric resolution of approximately 0.9 cm³.

1. Transmission scans: used to correct for attenuation and to calculate regional gas fraction (F^gas). We categorized pulmonary parenchyma as nonaerated (F^gas<0.1), poorly aerated (0.1 ≤ F^gas<0.5), normally aerated (0.5 ≤ F^gas< 0.85), and hyperinflated (F^gas≥ 0.85).36,37 

2. Emission scans

   • Intravenous ¹³N-nitrogen (¹³NN)-saline: used to measure regional pulmonary perfusion and shunt from the lung tracer kinetics after a bolus injection of ¹³NN-saline during a 60-s apnea at mean lung volume.35,38Because of the low solubility of nitrogen in blood and tissues (partition coefficient water-to-air is 0.015 at 37°C), in perfused and aerated regions, virtually all ¹³NN diffuses into the alveolar airspace at first pass, accumulating in proportion to regional perfusion. In regions with shunting alveolar units, there is a peak of tracer concentration in the early PET frames, corresponding to arrival of the bolus of tracer with pulmonary blood flow, followed by a decrease toward a plateau due to lack of retention of ¹³NN in nonaerated units. The magnitude of this decrease is related to regional shunt. Perfusion and shunt fraction were calculated with a tracer kinetics model.39 

   • Intravenous ¹⁸F-FDG: used for quantification of regional neutrophilic inflammation.29,30After ¹³NN clearance, ¹⁸F-FDG (5-10 mCi) was infused at a constant rate through the jugular catheter over 60 s, and, starting at the beginning of ¹⁸F-FDG infusion, sequential PET frames (9 × 10 s, 4 × 15 s, 1 × 30 s, 7 × 60 s, 15 × 120 s, 1 × 300 s, and 3 × 600 s) were acquired while pulmonary arterial blood was sampled at 5.5, 9.5, 25, 37, and 42.5 min. Blood samples (1 ml) were spun down, and the activity of plasma was measured in a gamma counter cross-calibrated with the PET scanner. The plasmatic activities of those samples were used to calibrate the blood-pool region of interest (ROI) (see Selection of Voxels for Analysis section) and obtain an image-derived input function taking into account partial-volume and spillover effects.33 ¹⁸F-FDG PET scans were acquired only after injury because of the 110-min half-life of ¹⁸F-FDG.

Inside cells, ¹⁸F-FDG is phosphorylated by hexokinase to ¹⁸F-FDG-6-phosphate, which accumulates in proportion to cellular metabolic rate. ¹⁸F-FDG net uptake rate (Kⁱ), a measure of cellular metabolic activity, was calculated at the ROI level by fitting the ¹⁸F-FDG kinetics with a three-compartment model of Sokoloff et al .30,40To account for potential effects of lung inflation and blood volume on regional Kⁱ, we standardized Kⁱby lung tissue, thus computing a specific Kⁱas K^is= Kⁱ/F^tissue, where F^tissue= (1-F^gas-F^blood) and F^bloodis the fractional volume of the blood compartment obtained from the model of Sokoloff et al . K^isis proportional to ¹⁸F-FDG uptake per gram of lung tissue.

Patlak analysis41was used to compute Kⁱat the voxel level and construct parametric images (fig. 1) and illustrations of regional kinetics (fig. 2).

Identification of the aerated lung fields was done by thresholding the transmission scans. Early frames of the ¹³NN-saline infusion emissions scans were used to identify nonaerated perfused lung fields. The combination of those fields resulted in the final lung mask for each animal. We manually excluded areas corresponding to main bronchi and large pulmonary vessels.

Three ROIs of same vertical height (ventral, middle, and dorsal) were defined by dividing the three-dimensional lung mask with two horizontal planes and used for quantification of regional F^gas, and ¹³NN and ¹⁸F-FDG kinetics.

A blood-pool ROI was defined by thresholding the regional activity of ¹³NN during the first 5 s after ¹³NN-saline injection. During this time, ¹³NN is confined mostly to the right heart cavities and pulmonary arteries, and only a minor amount has diffused into the alveolar gas volume.33 

Lungs from four animals of the control group and five animals of the endotoxin group were excised at the end of the experiment and fixed with Trump's fixative (4% formaldehyde and 1% glutaraldehyde in phosphate-buffered saline) at a pressure of 25 cm H²O. A block of lung tissue was sampled from ventral and dorsal regions and embedded in paraffin. Sections of 5-μm thickness were cut, mounted, and stained with hematoxylin-eosin for light microscopy. Lung neutrophils were counted in 40 randomly selected high-power (400×) fields per animal (10 per region) by two investiga-tors, who were blinded to the group assignment. In addition, perivascular and alveolar edema, alveolar hemorrhage, septal thickening, and capillary congestion were evaluated semiquantitatively with a four-grade scale (absent = 0, mild = 1, moderate = 2, and marked = 3).

Variables were tested for normality with the Shapiro-Wilk test. We expressed values as means and standard deviations for normally distributed variables and median and interquartile ranges (25-75%) otherwise. For normally distributed variables, we used the independent samples Student t  test for comparisons between groups and paired t  tests for comparisons between time points in the same group. For not normally distributed variables, we used the Wilcoxon rank sum test for comparisons between groups and the Wilcoxon signed rank test for comparisons between time points in the same group. Regional F^gas, perfusion, and shunt in ventral, middle, and dorsal regions within and between groups and before and after 2 h of mechanical ventilation were compared with a linear mixed-effect model.42This model was chosen because the analysis involved observations in the same individual at different time points and topographic lung regions. The categorical variables group, region, and time were modeled as fixed effects, and the variation among individuals was modeled by assuming random coefficients for the intercept (lme4 package, R statistical environment, R 2.6.2, Vienna, Austria). To study the dependence of K^ison F^gasand perfusion, plots of K^isversus  perfusion and F^gaswere built as follows: in each animal, F^gasand perfusion were computed voxel-by-voxel; functional compartments of aeration and perfusion were created by grouping together voxels belonging to tertiles of low, intermediate, and high F^gas, or of low, intermediate, and high perfusion; and K^iscomputed for the ROIs defined as the set of voxels in each tertile of F^gasor perfusion. Furthermore, we sought to identify the best predictors of K^isin the endotoxin group. For this, perfusion, F^gas, squared F^gas, and lung volume in each one of the aeration categories were computed in the ventral, middle, and dorsal ROIs and univariately regressed against K^is. Variables significantly associated with K^isin univariate analyses (P < 0.1) were included in a backward stepwise multivariate mixed-effect model using similar considerations as those described earlier. The multivariate regression analysis was applied. All statistical tests were two-tailed, and the significance was set at P < 0.05.

By experimental design, PEEP and P^platwere kept at 0 and 30 cm H²O, respectively. Along 2 h of mechanical ventilation, mean tidal volume decreased in both groups, significantly in the endotoxin group (table 1). In this group, respiratory rate was increased to maintain the PaCO²within the predefined acceptable range. The shift in PaCO²toward the upper limit of the accepted range was associated with a nonsignificant difference in pH (table 1). Oxygenation was reduced in the endotoxin group during the 2 h of the study (P = 0.02). Hemodynamics was stable during the experiment and comparable between groups. Circulating neutrophil counts decreased significantly in the endotoxin group (P = 0.02).

Mean F^gasof the imaged lung was similar in both groups at baseline (controls = 0.57 ± 0.08 and endotoxin group = 0.58 ± 0.04, NS) and after 2 h of mechanical ventilation (controls = 0.56 ± 0.08, endotoxin group = 0.53 ± 0.06, NS; figs. 1 and 3). The regional distribution of F^gasat baseline and after 2 h of mechanical ventilation was not statistically different between the two groups (fig. 3). F^gasdecreased significantly after 2 h in the dorsal regions in both groups and in the ventral regions of the control group (fig. 3).

There was no statistical difference in aeration between groups at baseline and after 2 h of mechanical ventilation (fig. 4). Most voxels in the ventral and middle regions were normally aerated, whereas voxels in the dorsal regions were either poorly aerated or nonaerated (fig. 4). The proportion of hyperinflated voxels was small in all regions. After 2 h of mechanical ventilation, the size of the nonaerated compartment increased significantly only in the dorsal ROIs of the endotoxin group (P = 0.02).

The regional distribution of perfusion before and at the end of the 2-h period was comparable between groups (figs. 1 and 3). The distribution of perfusion was inversely related with that of aeration along the vertical ROIs. In the ventral and middle regions of both groups, shunt fraction was small and remained unchanged after 2 h of mechanical ventilation (fig. 3). Shunt increased significantly in the dorsal regions of the endotoxin group.

PET assessment of ¹⁸F-FDG kinetics yielded regional plots that were suitable for analysis in all cases of both groups (figs. 2A and B). These plots were characterized by an early peak followed by either a continuously decreasing curve in controls or a plateau or slightly ascending slope in the endotoxin group. Quantification of the kinetics with Patlak plots (figs. 2C and D) or computations derived from the model of Sokoloff et al.  (table 2) evidenced the differences in regional ¹⁸F-FDG uptake, particularly high in dorsal ROIs.

In the control group, regional Kⁱwas low, showed a small interanimal variability in all ROIs, and was not significantly different among the ROIs of the same animal (table 2; fig. 5A). Likewise, specific Kⁱ(K^is), representing the standardization of Kⁱby the amount of lung tissue in the ROI, was not significantly different for the different ROIs (table 2; fig. 5B), despite the differences in regional perfusion and aeration. In contrast, global and regional Kⁱwere larger in the endotoxin group, with mean values of global Kⁱand K^ismore than twofold greater than those measured in the control group (table 2; fig. 5C). Furthermore, there were large differences in Kⁱand K^isamong ROIs in the endotoxin group, in addition to larger interanimal variability in these parameters. On average, Kⁱand K^isincreased progressively from ventral to dorsal ROIs (figs. 5C and D). As a result, Kⁱin the endotoxin group was 168% and K^iswas 46%, larger in dorsal than in ventral ROIs.

K^isincreased monotonically with perfusion in the endotoxin group, but not in the controls (figs. 6A and B). A different relationship was observed between K^isand F^gasin the endotoxin group (figs. 6C and D). K^isvalues in the two extremes of aeration were larger than those at the intermediate F^gastertile. No changes in K^iswere observed in the control group for the different F^gastertiles. We approximated the relationship between K^isand F^gaswith a quadratic function. Univariate analyses showed that the amount of lung in the nonaerated (P = 0.08), poorly aerated (P = 0.03), and normally aerated compartments (P = 0.03), and squared-F^gas(P = 0.05), F^blood(P = 0.02), and perfusion (P = 0.01), were significantly associated with K^is. Multivariate regression showed that only perfusion (slope = 0.5 × 10⁻²· min⁻¹, n = 18 observations, P < 0.01) and squared-F^gas(slope = 4.7 × 10⁻²· min⁻¹, n = 18 observations, P < 0.01) remained significantly associated with K^isamong the ROIs in the endotoxin group.

Neutrophil counts in the dorsal regions of the endotoxin group were more than double the counts in the same regions of the control group (P < 0.05), whereas no significant difference was present in the ventral regions (table 3). Moderate or severe alveolar hemorrhage was focal and present in only 11% of the high-power fields studied. On average, hemorrhage was mild and more intense in the dorsal regions of the control group (table 3). Perivascular and alveolar edema, septal thickening, and capillary congestion were very mild in both groups, without statistically significant differences.

The most important findings of the current study were as follows: (1) mild short-term endotoxemia in the presence of heterogeneous lung aeration and perfusion and large tidal volume mechanical ventilation bounded by clinically acceptable pressure limits produced spatially heterogeneous increases in pulmonary neutrophilic inflammation; (2) these regional increases of neutrophilic inflammation were significantly related to both regional aeration and perfusion and were detectable with PET within 2 h of injury; and (3) in the absence of endotoxemia, pulmonary ¹⁸F-FDG uptake was low and uniformly distributed at those settings of mechanical ventilation.

Pulmonary ¹⁸F-FDG uptake has been shown to be a marker of the concentration and degree of activation of neutrophils in the nontumoral lung.6,28We used the model of Sokoloff et al.  40to derive regional parameters from ¹⁸F-FDG kinetics. This choice was based on the observation of “Sokoloff-type” tracer kinetics in both control and endotoxin groups (fig. 2), according to a previously described strategy for model selection.30Based on that analysis, Sokoloff-type kinetics is suggestive of a minimal degree of lung edema.30 

The ventilator settings chosen for the study produced in both groups the intended heterogeneity in lung expansion with significant fraction of nonaerated and poorly aerated units in dependent regions (PEEP = 0) and a predominance of normal aeration in nondependent regions (limitation of P^platto 30-32 cm H²O). In fact, nondependent areas of the control group presented values of Kⁱand K^isthat matched those observed in uninjured lungs of prone sheep in two previous studies29,30and did not show any detectable histopathologic injury. PEEP = 0 was the likely cause for the decrease in F^gasin both groups after 2 h of mechanical ventilation. As intended, the used dose of endotoxin did not lead to changes in perfusion distribution to ventral, middle, and dorsal regions (fig. 3B). This result emphasizes the small dose used in our study, which was either equivalent to or lower than doses considered mild and devoid of lung injurious effects in previous investigations.6,43The nonaerated lung compartment increased in the most dependent regions with endotoxemia. At least three factors could account for this observation: surfactant dysfunction, increase in regional blood volume, and regional edema. All three are compatible with the increased regional shunt found in dependent regions. Given the very mild degrees of edema apparent on histologic samples, the two first factors are the most likely.

Endotoxin has been shown to induce systemic inflammation and to increase sequestration of neutrophils in the lungs yielding increased whole-lung ¹⁸F-FDG uptake.4–6Previous ex vivo  and in vivo  small animal models described an increase in the release of inflammatory mediators and a modification in inflammatory cell infiltration due to endotoxin, which depended on tidal volume.4,11,12,14However, no previous study has investigated whether and how those results can be extrapolated to the heterogeneously aerated and perfused large animal lung. Such extrapolation is complex, given that mechanical strain is heterogeneously distributed in the mechanically ventilated heterogeneously aerated lung and that there is still controversy over the contribution of overdistension and low volume ventilation to inflammation during ventilator-induced lung injury. Thus, it is difficult to predict how mild endotoxemia in the presence of heterogeneous pulmonary perfusion would interact with that heterogeneous distribution of lung inflation to affect lung inflammation, a condition related to clinical situations such as intraoperative mechanical ventilation during surgical interventions involving subclinical endotoxemia. A previous large animal study showed that endotoxemia produced increased whole lung neutrophil activation.6However, no data were provided on regional inflammation or on the combined effect of endotoxemia and mechanical ventilation. Understanding regional heterogeneities is essential to optimize the use of strategies to reduce lung injury by reducing heterogeneity of aeration22,23and perfusion,26,44according to the degree of lung injury severity.

Our results indicate that the combination of mild endotoxin doses with ventilatory settings, which did not by themselves cause injury, resulted in a substantial increase in pulmonary neutrophilic inflammation. This increase showed a heterogeneous spatial distribution, characterized by a vertical dependence of neutrophilic inflammation (Kⁱand K^is), more intense in dependent regions. The fact that the observed increase in ¹⁸F-FDG uptake was still significant after correction for regional lung tissue (K^is), including correction for regional lung collapse and blood volume, supports the conclusion that it was not a mere consequence of the increase in the amount of pulmonary tissue per unit volume of lung or in regional blood volume in derecruited regions because it was still significant after correction for regional lung tissue (K^is). The increase in neutrophil counts in dependent regions of the endotoxin group reinforces the imaging findings.

Remarkably, changes in global and regional ¹⁸F-FDG uptake in the endotoxemia group were markedly larger than those in regional aeration and perfusion. The average Kⁱand K^isin the endotoxin group were more than twofold compared with those in the control group, whereas average F^gasand pulmonary perfusion distributions were similar in the two groups. Such similarity supports the inference that perfusion distribution per se  could not explain the changes in global and regional ¹⁸F-FDG uptake during endotoxemia. Furthermore, although deterioration of regional aeration and shunt in the endotoxin group was limited to the dorsal regions, and distributions of F^gasand perfusion were similar in the two groups, ¹⁸F-FDG uptake (Kⁱ) in the endotoxin group was larger than that in the control group in all lung regions: 47% larger in ventral regions, 103% in middle regions, and 150% in dorsal regions. These results suggest that ¹⁸F-FDG may be an early and sensitive marker of regional pulmonary inflammation induced by the combination of endotoxemia and ventilator-induced lung injury.

Endotoxemia could contribute to the development of regional inflammation during mechanical ventilation by activation of neutrophils, resulting in their augmented response to the inflammatory stimuli produced by localized mechanical strain, which could be excessive in heterogeneous lungs. Our findings are, therefore, in line with the two-hit theory,11although we used both injurious stimuli simultaneously and not sequentially as in typical two-hit studies. The changes in circulating neutrophil counts suggest another form of interaction between mechanical ventilation and endotoxemia. The nonsignificant increase of circulating neutrophils in the control group, a trend noted in previous studies,6,10might be explained by release from the marginated pool due to mechanical ventilation. The lack of migration of these cells to the lungs or the lack of activation of these cells in the lungs implies that an additional hit would have a greater than normal supply of neutrophils to call on. The observed fall in circulating neutrophils in the endotoxin group shows that this occurred and that neutrophils have either marginated in the capillaries or migrated into the lungs. The increase in Kⁱand K^isin the endotoxin group indicates that these neutrophils became activated.

The multivariate regression analysis showed that both perfusion and regional aeration were important to explain the observed regional values of K^is. The linear relationship between K^isand regional perfusion (fig. 6) could represent the increase in regional inflammation with the increase in regional load of endotoxin, inflammatory cells, including neutrophils, and mediators of inflammation to the more perfused regions predominant in the dependent lung. This association between regional lung perfusion and K^isis important because therapeutic interventions can modify perfusion distribution, for example, as recently shown with noisy ventilation.26,44In contrast to the direct relationship between K^isand regional perfusion, the relationship between K^isand F^gaswas biphasic with large K^isin the low and high extremes of aeration and low K^isfor F^gasvalues in the normal range (fig. 6). The large K^isfor low F^gasmay represent the contribution of low volume lung injury associated with nonaerated and poorly aerated predominantly in dependent regions,9,20whereas the high K^isfor larger F^gasvalues suggest the contribution of lung overdistension. Although we did not find a significant fraction of hyperinflated areas, our measurements performed at mean lung volume underestimate hyperinflation, as discussed below. Thus, we speculate that overdistension could have occurred at end inspiration in regions of high F^gas.

Despite the substantial heterogeneity in lung aeration, we found a low and uniform distribution of neutrophilic inflammation in the control group. Such regional findings contrast with commonly invoked interdependence mechanisms,21which predict that the used P^platwould generate high local forces in regions lying in the interface between aerated and derecruited lung. These forces would be expected to result in inflammation and increased ¹⁸F-FDG uptake. At least two factors could account for our finding. First, ¹⁸F-FDG uptake could be an insensitive marker of regional lung mechanical injury. However, we showed that peak pressures of 50 cm H²O applied for 90 min increased lung ¹⁸F-FDG uptake in homogeneously expanded sheep lungs, supporting the sensitivity of the technique.29Given that interdependence mechanisms would predict local pressures higher than 130 cm H²O for P^plat= 30-32 cm H²O,21if those local values were present, they should be detectable with our technique. Indeed, our histologic findings support the absence of parenchymal inflammation in the control group. Consequently, the second possibility is that the assumptions of that previously theorized interdependence model21may not accurately represent the expansion of the dorsal regions of a normal lung.

Advantages and limitations of the used imaging techniques have been discussed in detail previously.29,30,33,38,39Specifically to this research, the regional aeration of the lung was computed from PET transmission scans. Because these scans are collected for 10 min during uninterrupted mechanical ventilation, the calculated F^gasof a region represents its average aeration over the breathing cycle. F^gasis not only affected by motion but also by filtering during image reconstruction and processing, and partial volume effects.45The consequent limited spatial resolution (13 mm) could result in underestimation of the degree of regional hyperinflation compared with that measured from computed tomography images acquired during end-inspiratory breath holds at much higher spatial resolution. Additional limitations include (1) species differences: young sheep as studied in our work are known to have reduced collateral ventilation, which might make them more prone than adult humans to reabsorption atelectasis and lung collapse.46Also, presence of pulmonary intravascular macrophages in sheep may make them differently sensitive to endotoxemia-induced lung injury when compared with humans47,48; (2) ventilatory settings: we chose settings aimed at maximizing aeration heterogeneity. Settings used clinically might be associated with regional inflammation of different intensity and distribution from those found in the current work. For example, lung inflammation may be reduced by PEEP and lower tidal volumes,49despite the presence of a proinflammatory response even during mild mechanical ventilation.50Also, in contrast to controlled ventilation, noisy pressure support ventilation could modify inflammation by changing lung mechanics and gas exchange25,26; (3) effects of mechanical ventilation in normal and injured lungs: mechanical ventilation with low and high tidal volumes affect lungs differently, depending on their previous degree of injury.51,52Consequently, regional inflammation magnitude and distribution may differ in the case of previously injured lungs or lungs affected by other injurious mechanisms51; (4) interference with anesthesia: because both inhaled53and intravenous54,55anesthetic agents can modulate inflammation, use of different anesthetic regimens could modify the observed inflammatory pattern.

In summary, marked spatially heterogeneous increases in pulmonary neutrophilic inflammation result from mild short-term endotoxemia in the presence of heterogeneous lung aeration and perfusion, and large tidal volume mechanical ventilation bounded by clinically acceptable pressure limits. ¹⁸F-FDG uptake may be a sensitive early marker of pulmonary neutrophilic inflammation in the studied conditions. The increase in spatial heterogeneity of inflammation was dependent both on regional perfusion and aeration, suggesting that factors beyond gas distribution can contribute to regional neutrophilic inflammation during ALI due to endotoxemia and mechanical ventilation.

The authors thank Wellington V. Cardoso, M.D., Ph.D. (Professor of Medicine and Pathology, Boston University School of Medicine, Boston, Massachusetts), and Mauro Tucci, M.D., Ph.D. (Research Fellow, Department of Anesthesia and Critical Care, Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts, and Respiratory Intensive Care Unit, University of Sao Paulo School of Medicine, Sao Paulo, Brazil), for support with histopathologic analysis; Hui Zheng, Ph.D. (Assistant Professor, Biostatistics Center, Massachusetts General Hospital), for assistance with statistical analysis; Steven B. Weise, B.S. (Senior Research Technician, Division of Nuclear Medicine, Massachusetts General Hospital), for image acquisition and processing; Peter A. Rice, B.S., R.Ph., and Stephen C. Dragotakes, R.Ph. (Certified Nuclear Pharmacists, Department of Radiology, Massachusetts General Hospital), John A. Correia, Ph.D. (Associate Professor of Radiology, Harvard Medical School), and William M. Buceliewicz, B.S., and David F. Lee, B.S. (Senior Cyclotron Engineers, Department of Radiology, Massachusetts General Hospital), for preparation of the radioisotopes; and Ivany Schettino, M.D., Ph.D., Tommaso Mauri, M.D., and Maria Avila, M.D. (Research Fellows, Department of Anesthesia and Critical Care, Massachusetts General Hospital and Harvard Medical School), for assistance with animal preparation.