INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

ENDOTOXIN CONTRIBUTES to gas exchange abnormalities in experimental acute lung injury by increasing ventilation-to-perfusion ratio (A/) heterogeneity and intrapulmonary shunt (8, 13, 18). Although frequently attributed to changes in the perfusion distribution (19), A/ heterogeneity is determined by changes in alveolar ventilation (A) heterogeneity, perfusion heterogeneity, and the correlation between regional ventilation and perfusion (29). Because the distributions of ventilation and perfusion have not been independently measured during endotoxemia, the precise changes in ventilation and perfusion that lead to A/ heterogeneity are unknown.

The purpose of this study was to determine how endotoxin changes the distributions of regional ventilation and perfusion and the contribution of these changes to A/ heterogeneity. We independently measured regional ventilation and perfusion using aerosolized and injected microspheres in endotoxemic pigs and quantified the relative importance of changes in ventilation heterogeneity, perfusion heterogeneity, and correlation between regional ventilation and perfusion in determining A/ heterogeneity during endotoxemia.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Animal Preparation

1

1

Study Protocol

1

1

Regional ventilation was measured at each time point by aerosolizing yellow, orange, or yellow-green 1-µm-diameter fluorescent microspheres (FluoSpheres, Molecular Probes, Eugene, OR) in the ventilator circuit (25), while regional perfusion was simultaneously measured by injection of crimson, blue-green, or green 15-µm fluorescent microspheres (FluoSpheres, Molecular Probes) through a femoral venous catheter. Microspheres were sonicated and vortexed immediately before administration. Fifteen-micrometer microspheres were suspended in normal saline to a total volume of 10 ml, manually injected in small intermittent boluses (in an attempt to achieve as constant an infusion as possible), and agitated frequently during injection to prevent settling. Microspheres were administered over a 10-min time period in the first five animals and, because of improvements in aerosol delivery, over 5 min in the last two animals. Different color microspheres were used to measure ventilation and perfusion at each time point, and color assignment varied across experiments. Arterial, mixed-venous, and exhaled gas samples were collected for determination of inert gas concentrations by use of a gas chromatograph (Varian 3300; Walnut Creek, CA) (15) immediately after each microsphere administration, and blood-gas measurements were made on arterial and venous samples (ABL 4, Radiometer, Copenhagen, Denmark).

Core body temperature, mean arterial pressure, pulmonary artery pressure, pulmonary artery occlusion pressure, peak and end-inspiratory pause airway pressures, cardiac output (in triplicate), respiratory exchange ratio, and arterial blood gases were measured immediately before and after each microsphere administration, and average values were used for comparisons between time points. Cardiac output was measured by thermodilution (Edward's COM 2, Baxter, Irvine, CA). Hematocrit was determined by centrifugation of duplicate samples drawn after each microsphere administration. The respiratory exchange ratio was calculated by using fractional O₂ and CO₂ contents in inspired and mixed expired gas that were measured with a mass spectrometer (MGA1100, Perkin-Elmer, Norwalk, CT).

Animals were given heparin (10,000 units) and papaverine (2 mg/kg) and then killed by exsanguination under deep anesthesia. Median sternotomy was performed, large-bore catheters were placed in the left atrium and main pulmonary artery, and the lungs were perfused with 2% dextran in normal saline until free of blood. The right kidney was removed from five animals and analyzed for fluorescence to determine whether right-to-left shunting was present.

Lungs were removed from the chest, inflated with 25 cmH₂O airway pressure, and air dried. They were encased in foam while suspended vertically in a squared box and then cut into 1.2-cm³-thick transverse slices, and each slice was cut into ~1.7-cm³ cubes in a miter box, yielding 851-1,221 lung pieces per animal. Pieces were visually scored for airway content and weighed, with pieces less than 8 mg excluded from the data set to minimize error due to uncertainty in flow or weight.

Fluorescent intensities for each color were determined by extracting fluorescent dye from each lung piece with the organic solvent 2-ethoxyethyl acetate (Aldrich Chemical, Milwaukee, WI). Dye concentration was read in a fluorimeter (LS50B, Perkin-Elmer) and corrected for background signal and spillover from adjacent colors (11). Based on the mean fluorescent intensity for each color in each animal and standard values for fluorescent intensity per microsphere, the mean number of microspheres per piece was ~1,300 for crimson, 1,200 for blue-green, 1,000 for green, 39,000 for yellow, 37,000 for yellow-green, and 56,000 for orange. Kidney fluorescence was determined by dissolving tissue with 4 M KOH, filtering the suspension with a 10-µm pore filter, extracting fluorescent dye from the filter with 2-ethoxyethyl acetate, and determining dye concentration in a fluorimeter (11).

Data Processing

Fluorescence. Fluorescent intensities were converted to flows in milliliters per minute by dividing the fluorescent intensity within each piece by the sum of intensities for that color in all pieces and then multiplying by total ventilation (for ventilation) or cardiac output (for perfusion). Ventilation was calculated by estimating anatomic dead space in three consecutive breaths from plots of exhaled CO₂ concentration vs. exhaled volume as described by Fowler (9).

1

1

A- and -weighted A/ distributions. Inert-gas data for an animal were excluded if the remaining sum of squares at any time point exceeded 10. Retentions and excretions of inert gases were used to compute 50-compartment A- and -weighted A/ distributions as described by Wagner et al. (27). Standard deviations of A- and -weighted A/ distributions (logSD and logSD) were calculated by using all A/ bins except shunt and dead space and again after exclusion of secondary A/ peaks (i.e., those smaller than the main peak) in high A/ regions. A- and -weighted A/ distributions were also calculated from microsphere data after pieces representing shunt and dead space were excluded.

Predicting gas exchange. Arterial PO₂, PCO₂, and alveolar-arterial O₂ differences were predicted from A/ distributions measured with microspheres, mixed venous blood gases, Hb concentration, and body temperature as described by Altemeier et al. (2). Briefly, end-capillary O₂ and CO₂ contents and regional alveolar PO₂ and PCO₂ were calculated in each lung piece by using its A/ and solving mass balance equations for O₂, CO₂, and N₂. End-capillary gas contents were perfusion weighted and summed to give arterial gas contents, and regional alveolar gas tensions were ventilation weighted and summed to give mixed alveolar gas tensions. Arterial O₂ and CO₂ contents were converted to gas tensions by using oxygen- and carbon dioxide-Hb dissociation curves for pigs.

Statistical Analysis

Comparisons between baselines 1 and 2 and between baseline 2 and endotoxemia were made by using two-tailed paired t-tests for physiological, gas exchange, and inert gas data and for coefficients of variation and standard deviations of the A, , and A/ distributions. Paired t-tests were also used to compare results of inert gas and microsphere techniques and predicted and measured gas exchange. Linear correlation coefficients were calculated for measurement of regional A and within the same piece at the same time point, and differences were evaluated by using paired t-tests after Fisher's z transformation.

Determinants of ventilation-perfusion heterogeneity. We partitioned changes in A/ heterogeneity during endotoxemia into those attributable only to changes in A heterogeneity, only to changes in heterogeneity, and only to changes in correlation between regional A and (i.e., A- correlation) by using the mathematical expression that relates these variables (29)

(1)

where F describes the variance of the A/ distribution in the natural log domain (),  is the correlation between ln A and ln within a piece, and and are the variances of the A and distributions in the natural log domain. Specifically, effects of , , and were quantified in each animal by calculating the hypothetical change in that would have resulted had endotoxin changed only one of the three variables. Each hypothetical change, H, was calculated by taking the difference between during baseline 2 (calculated by inserting values for , , and  from baseline 2 into Eq. 1), and that would have resulted had endotoxin changed only one of the three variables (calculated by inserting appropriate values for , , and  into Eq. 1)


where subscripts E and 2 denote distributions measured during endotoxemia and baseline 2. ANOVA using a randomized complete block design was used to assess differences between hypothetical changes in due to , , and , and post hoc comparisons between individual changes were evaluated with Fisher's protected least significant difference test.

Redistribution of ventilation and perfusion. Redistribution was defined as the decrease in correlation between pre- and postendotoxin measurements (e.g.,  [A_E, A₂]) compared with the correlation between measurements made during baselines 1 and 2 (e.g.,  [A₁, A₂]). Redistribution was considered significant if the confidence interval for the difference in correlation coefficients did not include zero. The magnitudes of ventilation and perfusion redistribution were considered significantly different if confidence intervals describing ventilation and perfusion redistribution did not overlap.

Effects of airway deposition. To evaluate the effect of airway content on ventilation and perfusion distributions, we recalculated coefficients of variation and redistribution for ventilation and perfusion after excluding lung pieces judged to contain more than 25% airways by volume.

Mechanisms of redistribution. Linear regression analysis was used to determine whether changes in predicted alveolar PO₂ (2) within a lung piece [(predicted alveolar PO₂ during endotoxemia)  (predicted alveolar PO₂ during baseline 2)] were predictive of changes in perfusion due to endotoxin in that piece [100 · (_E  2)/₂]. Pieces with flows during baseline 2 that were less than 0.05 times mean flow were excluded from this analysis. Slopes of regression lines were evaluated for difference from zero with unpaired t-tests.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Physiological Data

Endotoxemia decreased arterial and mixed venous PO₂ and increased mixed venous PCO₂ and the alveolar-arterial O₂ difference. Arterial PCO₂ did not increase significantly (Table 1). Endotoxin also increased peak airway and end-inspiratory hold pressure and resulted in hemoconcentration and acidemia. Physiological data did not change significantly between baseline 1 and baseline 2.

Microsphere Data: Ventilation, Perfusion, and Ventilation-Perfusion Distributions

View this table: [in this window] [in a new window]   Table 2.   Effects of endotoxin on A- correlation and A, , and A/ heterogeneity

View larger version (25K): [in this window] [in a new window]   Fig. 1.   Regional alveolar ventilation (A) vs. regional perfusion () within same lung piece measured during baseline 2 (A) and endotoxemia (B). Subscripts 2 and E denote time points baseline 2 and endotoxemia, respectively. Each point represents a single 1.7-cm3 lung region (n = 1,221). Correlation between regional A and is initially high but decreases during endotoxemia.

Inert-Gas Data: Ventilation-Perfusion Heterogeneity

Determinants of Ventilation-Perfusion Heterogeneity

View larger version (25K): [in this window] [in a new window]   Fig. 2.   Percent changes in ratio of A to (A/) heterogeneity attributable to change in A- correlation (change in ), A heterogeneity (change in A), heterogeneity (change in ), interaction between simultaneous changes in A and heterogeneity [interaction (A-)], and interaction between simultaneous changes in A- correlation and A and heterogeneity [interaction (-A,)]. On average, decrease in A- correlation had a greater impact on A/ heterogeneity than did changes in A or heterogeneity. Interaction (-A,) also significantly increased A/ heterogeneity because a decrease in A- correlation magnifies the effect that increases in A and heterogeneity have on A/ heterogeneity. Note that interaction (A-) decreased A/ heterogeneity because A/ heterogeneity achieves a local minimum when A and heterogeneity are similar, and A and heterogeneities are more alike when changes due to endotoxin are considered together. See text and Fig. 5 for further discussion.

Redistribution of Ventilation and Perfusion

View this table: [in this window] [in a new window]   Table 3.   Effect of endotoxin on redistribution of A and

View larger version (37K): [in this window] [in a new window]   Fig. 3.   A: regional during baseline 1 (1) vs. regional 2. B: regional during endotoxemia (E) vs. regional 2. C: regional at baseline 1 (A1) vs. regional A2. D: regional AE vs. regional A2. Each point represents a single 1.7-cm3 lung region (n = 913). Note linear correlation coefficient (r) and 95% confidence intervals (given in parentheses). Endotoxemia causes significant redistribution of perfusion, but redistribution of ventilation is modest.

Regional perfusion did not preferentially decrease in regions predicted to have the largest decrease in alveolar PO₂ (mean slope = 1.1 ± 0.4 Torr¹, P = 0.0005 for difference from zero) (Fig. 4A). Changes in regional ventilation and perfusion due to endotoxin were poorly correlated (mean r² = 0.09, 95% confidence interval 0.06  0.25) (Fig. 4B) even when animals with no or minimal ventilation redistribution were excluded from the analysis.

View larger version (16K): [in this window] [in a new window]   Fig. 4.   A: percent change in regional [100 · (E  2)/2] vs. change in predicted alveolar PO2 (PAO2) in the same lung piece [(predicted PAO2 during endotoxemia)  (predicted PAO2 during baseline 2)]. Note regression line. Perfusion does not preferentially decrease in regions with the greatest decrease in PAO2, suggesting that hypoxic pulmonary vasoconstriction does not determine perfusion redistribution during endotoxemia. B: percent change in regional A [100 · (E  2)/2] vs. percent change in regional [100 · (E  2)/2] in same lung piece. Note coefficient of determination R2. Changes in regional ventilation and perfusion are poorly correlated, suggesting that mechanisms that match ventilation and perfusion do not determine redistribution of ventilation or perfusion during endotoxemia. Each point represents a single 1.7-cm3 lung region (n = 913).

Comparison of Microsphere and Inert-Gas Data

View this table: [in this window] [in a new window]   Table 4.   Heterogeneity and shunt in A- and -weighted A/ distributions

Predicted vs. Measured Gas Exchange

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The main findings of this study are the following: 1) Endotoxemia decreases the correlation between regional ventilation and perfusion and increases perfusion heterogeneity. 2) Decrease in correlation between regional ventilation and perfusion during endotoxemia substantially increases A/ heterogeneity, whereas increase in perfusion heterogeneity has less effect. 3) Endotoxemia has only modest effects on the ventilation distribution. Therefore, the decrease in correlation between regional ventilation and perfusion during endotoxemia results primarily from changes in perfusion.

Despite a significant increase in perfusion heterogeneity, the increase in A/ heterogeneity during endotoxemia was determined principally by the decrease in A- correlation. The importance of the A- correlation follows directly from the mathematical relationship describing A/ heterogeneity as a function of ventilation and perfusion heterogeneity and A- correlation (Eq. 1) (29) that is illustrated graphically in Fig. 5.

View larger version (32K): [in this window] [in a new window]   Fig. 5.   Contour plots showing A/ heterogeneity for 2 different A- correlations: A- correlation before endotoxemia (A- correlation = 0.88; A), and A- correlation during endotoxemia (A- correlation = 0.56; B). Specifically, contour plots show A/ heterogeneity (variance A-) as a function of A heterogeneity (variance A), heterogeneity (variance ), and A- correlation as described in Eq. 1 (see Statistical Analysis). When the A- correlation is high (A), A/ heterogeneity achieves local minima in the region in which A and heterogeneity are similar (i.e., in the "valley" of the contour plot). In this region, changes in A and heterogeneity have little impact on A/ heterogeneity. When the A- correlation decreases (B), A/ heterogeneity becomes more sensitive to changes in A and heterogeneity, and A/ heterogeneity increases considerably even if A and heterogeneity are unchanged.

When the A- correlation is high (Fig. 5A), A/ heterogeneity is relatively low in the region in which ventilation and perfusion heterogeneity are similar (i.e., in the "valley" of the contour plot). In this region, even large increases in ventilation and perfusion heterogeneity have relatively little impact on A/ heterogeneity. In contrast, A/ heterogeneity increases dramatically when the A- correlation decreases (Fig. 5B), even if ventilation and perfusion heterogeneity do not change. When the A- correlation is low or ventilation and perfusion heterogeneity are very dissimilar, changes in ventilation and perfusion heterogeneity have a greater impact on A/ heterogeneity.

The importance of the A- correlation to A/ heterogeneity has not been previously emphasized. Historically, A/ heterogeneity has been attributed primarily to heterogeneity in the distributions of ventilation and perfusion because the A- correlation was thought to be weak (29). When regional ventilation and perfusion are directly and independently measured, however, the A- correlation is strong (3, 21, 25). Melsom et al. (21) found that this strong correlation was associated with a narrow A- distribution even though the individual distributions of ventilation and perfusion were broad. We extend their observations by showing that A/ heterogeneity increases during endotoxemia principally as a result of the decrease in the previously strong correlation between regional ventilation and perfusion. The importance of the A- correlation is unlikely to be unique to endotoxemia. Because we and others (3, 21, 25) have demonstrated that normal lung (i.e., before injury) is characterized by a strong A- correlation and has similar ventilation and perfusion heterogeneities, A/ heterogeneity should be relatively sensitive to changes in A- correlation and insensitive to changes in ventilation and perfusion heterogeneity regardless of their cause.

To what extent are changes in the ventilation and perfusion distributions independently responsible for the decrease in A- correlation during endotoxemia? The contribution of each to the decrease in A- correlation is roughly reflected by the magnitude of ventilation and perfusion redistribution. Because perfusion redistribution was significantly greater than ventilation redistribution, changes in perfusion are primarily responsible for the decrease in the A- correlation. Redistribution roughly reflects effects on the A- correlation because the A- correlation was strong before endotoxemia and changes in ventilation and perfusion within each piece during endotoxemia were weakly correlated. Therefore, redistribution of ventilation and perfusion is much more likely to decrease than to increase A- correlation.

Although mechanisms determining redistribution of perfusion during endotoxemia are unclear, redistribution is unlikely to be explained by regional hypoxic pulmonary vasoconstriction. Perfusion did not decrease in regions predicted to have the greatest decrease in alveolar PO₂ (Fig. 4A), and changes in regional ventilation and perfusion within each piece were not strongly correlated (Fig. 4B). This suggests that hypoxic pulmonary vasoconstriction or other mechanisms that preserve A/ matching do not determine redistribution of ventilation or perfusion during endotoxemia. Release of thromboxane A₂ has been shown to mediate endotoxin-induced pulmonary vasoconstriction (17, 26) and could potentially mediate heterogeneous changes in perfusion via regional differences in thromboxane A₂ release or responsiveness of the pulmonary vasculature. The presence of physiologically significant spatial heterogeneity in biochemical and cellular mediators of vasoregulation awaits direct confirmation.

Our conclusions rely on the ability of microspheres to accurately measure regional ventilation and perfusion. Fluorescent microspheres are a well-established method for measuring regional pulmonary blood flow. Microspheres of 15-µm diameter lodge in pulmonary capillaries in proportion to regional blood flow (4, 14, 20). In addition, fluorescent-labeled microspheres have been validated as markers of regional pulmonary blood flow by comparison with radiolabeled microspheres in injured (16) and uninjured (11) lungs. Although lung pieces with low relative blood flows may have contained fewer than 400 microspheres of a single color in this study, heterogeneity and correlation coefficients are well determined even when the "400-microsphere rule" is violated (24).

Aerosolized, 1-µm-diameter fluorescent microspheres have been validated as markers of regional ventilation in normal lung (25). They deposit nearly exclusively in gas-exchanging regions of the lung, provide reproducible measurements, and (combined with measurements of regional blood flow) can accurately predict respiratory and inert-gas exchange (2). Because deposition patterns of aerosolized microspheres have not been determined during lung injury, increased deposition in non-gas-exchanging regions (i.e., airways) during endotoxemia may have confounded estimates of regional ventilation. Although we cannot definitively exclude this possibility, changes in ventilation heterogeneity and redistribution were minimal when we excluded lung pieces with high airway content from our analysis.

A/ distributions generated from microsphere data most likely underestimate A/ heterogeneity during endotoxemia. Both overestimation of arterial PO₂ and underestimation of A/ heterogeneity compared with inert-gas data support this belief. Underestimation of A/ heterogeneity is most likely explained by development of A/ heterogeneity on a scale that is beneath the spatial resolution of our methods (i.e., within 1.7-cm³ lung pieces). This interpretation is consistent with that of Altemeier et al. (3), who showed that microsphere data accurately predicted arterial PO₂ in uninjured lungs but overestimated arterial PO₂ after vascular bead embolism.

Discrepancy between inert-gas and microsphere estimates of A/ heterogeneity (i.e., logSD and logSD) before endotoxemia is more difficult to explain. The limited spatial resolution of the microsphere method should result in lower estimates of A/ heterogeneity for microspheres than for MIGET. However, because microsphere data accurately predict gas exchange before endotoxemia, underestimation of A/ heterogeneity by the microsphere method before endotoxemia must be small. Although values for ventilation and perfusion heterogeneity and A- correlation before endotoxemia (from microsphere data) are consistent with previously published work (3, 5, 12, 21, 25), estimates of A/ heterogeneity made with MIGET are higher than those in the literature (8, 18). Therefore, discrepancy between MIGET and microsphere data before endotoxemia may be due to systematic error in our inert-gas data in addition to the limited spatial resolution inherent in the microsphere method.

Change in cardiac output may have affected the perfusion distribution during endotoxemia. The effect of cardiac output on regional perfusion has never been measured directly, but studies using the inert-gas technique show that A/ heterogeneity increases when cardiac output decreases (6, 22). Because increases in A/ heterogeneity in these studies were small compared with those that we and others (8, 18) have observed during endotoxemia, the decrease in cardiac output during endotoxemia is unlikely to have had a major impact on redistribution of blood flow.

Abnormalities in A- matching during porcine endotoxemia may resemble those in patients with acute respiratory distress syndrome and sepsis because endotoxin is frequently present in these conditions (23). However, the early time point that we chose to study, the degree of pulmonary hypertension that develops in pigs, and the presence of pulmonary intravascular macrophages in pigs but not humans (28) may limit the generalizability of our conclusions.

In conclusion, we have shown that decrease in correlation between regional ventilation and perfusion is the principal determinant of A/ heterogeneity during endotoxemia. This correlation decreases primarily as a result of changes in perfusion, because endotoxin's effects on the distribution of ventilation are modest. Although perfusion heterogeneity increases during endotoxemia, this increase has little effect on A/ heterogeneity.