INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

IN DIFFUSE LUNG INJURY such as acute respiratory distress syndrome, it has been shown that pulmonary vessels supplying injured areas are often occluded, possibly by hypoxic vasoconstriction, occlusion by perivascular cuffs of edema, or intraluminal thrombosis (7). In dogs, we found that anastomotic bronchial systemic-to-pulmonary blood flow increases acutely when an isolated left lower lobe is injured through either the vascular route or via the airways (9). Total bronchial blood flow also increases in sheep after smoke inhalation (13) and in goats after acid aspiration (1, 14). Prior studies have reported either anastomotic bronchial blood flow (i.e., in the part that drains into the pulmonary circulation) (9) or the total flow through the bronchial branch of the bronchoesophageal artery when using flow probes (13). We recently validated the use of fluorescent microspheres to measure regional lung perfusion by both pulmonary and bronchial circulations (3, 6). Perfusion to small areas of the lung (2-cm cubes) can now be measured with a degree of accuracy not previously possible. Spatial coordinates recorded from each small piece of lung allow regional changes to be determined in bronchial and pulmonary blood flow in response to segmental lung injury. In this study, we report the total and regional response of both pulmonary and bronchial circulations to a localized (segmental) injury to the lung.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study was approved by the University of Washington Animal Care Committee. All methods and treatment of animals were in accordance with the Helsinki Declaration.

Prone, anesthetized, mechanically ventilated sheep (an animal with a large bronchial blood supply, minimal collateral ventilation, and septated lung lobules) were studied. Six 20- to 40-kg sheep (Columbia Cross) were sedated with xylazine, anesthetized, and maintained with intravenous thiopental sodium (15 mg/kg). The animals were ventilated at a tidal volume of 7-10 ml/kg, at a rate sufficient to maintain normal arterial blood gases. Polyethylene catheters were placed 1) in the abdominal inferior vena cava via a femoral vein for injection of microspheres into the pulmonary circulation, 2) in the right pulmonary artery (balloon floatation, Swan-Ganz catheter) for measurement of cardiac output, and 3) in the femoral artery to monitor systemic arterial pressure and arterial blood gases. An additional catheter was placed in the left ventricle via the left carotid artery to inject bronchial blood flow markers into the systemic circulation. A left thoracotomy was performed in the left fourth intercostal space, and umbilical tape was placed loosely around the left pulmonary artery. After surgery, the animals were maintained at 5 cmH₂O positive end-expiratory pressure, the pleural space was evacuated, and the chest wall was closed. Systemic blood pressure, airway pressure, cardiac output (by thermodilution), body temperature, and arterial blood gases were monitored throughout the study.

When blood pressure, cardiac output, and arterial blood gases were stable (~15-20 min), two million 15-µm fluorescent microspheres were injected into the femoral vein over 1 min to measure pulmonary blood flow. A reference blood sample from the pulmonary artery was withdrawn at a rate of 5 ml/min by using a calibrated withdrawal pump to calculate pulmonary blood flow to the left lung in milliliters per minute. Bronchial blood flow to the left lung was then measured by injecting 80 million fluorescent microspheres of a different color into the left ventricle, via the left ventricle catheter. Pressure tracings from this catheter were recorded just before microsphere injection to confirm location of the catheter in the left ventricle. Immediately before injection of microspheres into the left ventricle, the preplaced umbilical tape was tied, and the left pulmonary artery was temporarily occluded for 3 min after completion of microsphere injection. This prevented microspheres that had passed through systemic arteriovenous communications from reaching the left lung via the pulmonary circulation. Reference blood samples were withdrawn from the femoral arteries to calculate bronchial blood flow to the left lung in milliliters per minute. Under fiber-optic bronchoscopic guidance, a dependent segment of the left lung was located, and the bronchoscope was wedged into the segment. Through a polyethylene catheter passed through the suction channel of the bronchoscope, 1 ml/kg of 0.1 N hydrochloric acid was instilled into the segment. One hour after hydrochloric acid administration, pulmonary and bronchial blood flows were remeasured by injecting microspheres of different fluorescent colors. The animals were then deeply anesthetized, heparinized, and exsanguinated; lungs were removed; and the pulmonary vasculature was flushed free of remaining blood with a solution of normal saline and 2% dextran. When the effluent from the lungs was clear, the lungs were removed from the chest and inflated to 30 cmH₂O, suspended from the tracheal cannula, and air-dried at 30 cmH₂O (total lung capacity) for 7 days. The lungs were then placed in a plywood box that was filled with polyurethane foam and sliced into 1.2-cm-thick sections in the transverse plane (from cranial to caudal) for a total of ~25-30 slices. The left lung slices were diced into 2-cm³ cubes in a miter-box, and the x-, y-, and z-coordinates for each piece were noted. This allowed for three-dimensional reconstruction of the left lung (x: right to left, y: dorsal to ventral, and z: cranial to caudal) and the spatial location of each cube of lung tissue. Each lung piece was weighed, coded for presence or absence of airways, and coded for the presence or absence of injury based on visual inspection. Pieces weighing <10 mg were excluded from analysis, since the amount of fluorescence in them is small, and there is larger error in both the weight and the fluorescence measurements. The pieces were then soaked in 2-ethoxyethyl acetate to extract the fluorescence from the microspheres, which was measured by using a fluorimeter (LS-50B, Perkin Elmer).

Statistical methods. Blood flow was calculated for each piece by using the standard reference flow formula (12)

(1)

where _i is blood flow to each piece i (ml/min), F_i is fluorescence from each piece, R is the rate of reference blood withdrawal (ml/min), and F_ref is the fluorescence from the reference blood sample. Blood flow _i was determined for each fluorescent color used to identify bronchial and pulmonary flows to each piece pre- and postinjury.

The percentage of total pulmonary and bronchial flows to injured or noninjured pieces was determined for pre- and postinjury conditions. We calculated relative pulmonary and bronchial blood flows to each piece separately for injured and noninjured pieces in pre- and postinjury conditions. Relative flow to a piece within a region is calculated by dividing the flow to each piece during one condition by the mean flow to all pieces in the region during that condition. The relative flow units allow a clearer analysis of redistribution of flow during the two conditions. Thus, in each piece from the injured region, four relative flow values were calculated for pulmonary and bronchial flows both before and after injury. Similarly, for each piece in the noninjured region, four values of relative flow were calculated.

Relative blood flow units remove any effect of changes in cardiac output, driving pressure, or total flow to a region and also allow distribution of blood flow within a region to be determined independently of total blood flow to the lung and to the region. This normalization is important to the analysis of redistribution of flow. For example, if flow increases to the injured region from pre- to postinjury, but postinjury flow is allocated to pieces in exactly the same proportion as before the injury (i.e., no redistribution is present within the region), then relative flow units for each piece remain unchanged, with no redistribution of flow within the region. Also, in this example, a regression slope calculated for relative flow vs. the distance from the center of injury would be unchanged from pre- to postinjury states. On the other hand, if absolute flow units were used, or if flow was normalized within the entire lung (but not within the region of interest), then the calculated flow per piece as well as any slopes would change from pre- to postinjury, even though the allocation of flow within the region did not change.

We also measured the trend in flow in relation to the distance from the center of injury, which was determined by visual inspection. The slope of relative flow vs. distance from the center of injury was calculated by using ordinary least squares regression. The slope is expressed as percent relative flow units per centimeter away from the center of injury. A slope value of 7.1%, for example, means that the change in flow per centimeter distance is 7.1% of the mean flow per piece to the region.

We used the coefficient of variation (CV) to characterize heterogeneity. Descriptive statistics include the mean and SD of physiological parameters, slopes, and CVs. Descriptive statistics pre- and postinjury are compared by using the paired t-test. This test is also used to compare descriptive statistics between injured and noninjured regions or between pulmonary and bronchial flow. We also calculated the impact of injury on flow distribution by partitioning the total variation of flow within a region into two components: 1) a "piece" effect, the unchanging relative flow that reaches each piece during both injured and noninjured states, and 2) the flow component that changes between injured and noninjured states, possibly because of a combination of injury, time variation, and measurement error. The calculations have been described previously (2).

The variance component caused by flow changes across states (injured and noninjured) (²_change) was estimated by

(2)

where Y_ij is the observed relative flow for piece i in the injured state j, _i is the mean relative flow for piece i across the states considered, n is the number of states, and m is the number of pieces.

The variance component due to piece effect ) was estimated by

(3)

The grand mean across all pieces and states () has a value of 1.0 when flow is normalized to a mean for each state. This formula for the piece effect differs slightly from that used in mixed-effects modeling. We did not subtract /n from the , a subtraction that was intended to correct for any purely random error. However, incorporates both systematic variation within a piece over states and purely random error. Subtraction would lead to an underestimate of (11).

With this method, measured flow variation across all pieces and conditions is partitioned into constant and changing components. These calculations determine the percent of flow variation caused by a fixed flow to each piece vs. the percent changes that arise from injury, time fluctuation, or random measurement error.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The total number of injured pieces from the left lung of six animals was 120 ± 65 (mean ± SD) (range 53-238). The mean number of noninjured pieces from the left lung was 518 ± 192 (range 331-842), resulting in a range from 6 to 36% of the left lung being injured. The injury caused a major redistribution of flow in the lung, within injured and noninjured regions alike.

Changes in pulmonary blood flow distribution were greater with injury than the bronchial flow. Mean pulmonary blood flow to noninjured pieces before injury was 727 ± 246 ml/min and postinjury it was 574 ± 347 ml/min (P = 0.22). In contrast, mean pulmonary blood flow to the injured pieces was 246 ± 143 ml/min, which decreased to 56 ± 46 ml/min after injury (P = 0.01).

Cardiac output fell in two animals after lung injury. Because cardiac output influences bronchial blood flow, we corrected for this by dividing bronchial blood flow by the cardiac output obtained just before we made the measurements of bronchial blood flow in the injured pieces. The bronchial blood flow to the injured region was 0.51 ± 0.17 ml · min¹ · l¹ cardiac output, and it increased to 1.43 ± 0.85 ml · min¹ · l¹ cardiac output (P = 0.04).

Cardiac output and allocation of flow. Pulmonary flow decreased by about one-third, and bronchial flow approximately doubled from pre- to postinjury (Table 1). The allocation of flow to injured and noninjured regions also changed substantially. Before injury, 26% of pulmonary flow was distributed to the region that was subsequently injured, whereas only 9% reached this region postinjury. This reallocation, combined with the overall decrease in total pulmonary flow to the lung, resulted effectively in a fivefold reduction in pulmonary flow to the injured region.

Bronchial blood flow had a very different redistribution pattern. Approximately one-fifth of the total bronchial flow was allocated to the injured region both before and after injury. This near-constant allocation combined with an increase in total bronchial flow to the lung tripled the mean bronchial flow to the injured region.

Redistribution of blood flow within injured and noninjured regions. The balance of results on flow is presented in relative flow units, calculated as described above. Before injury, bronchial and pulmonary flows showed a negative radial gradient; i.e., flow decreased in pieces farthest from the center of injury in both the injured and noninjured regions (Table 2). Whereas effects of injury varied across animals, the negative gradient for mean pulmonary flow either changed to a positive gradient in the injured region or became less steep in the noninjured region. This resulted in decreased pulmonary flow near the center of the injury, and this decrease was a statistically significant shift to the noninjured region. The bronchial blood flow showed a different redistribution pattern. In both injured and noninjured regions, the negative radial gradient of bronchial flow became steeper; i.e., flow to the center of the injury increased twofold after injury, causing a mean increase in the allocation of bronchial flow toward the center of the injury, although the change was not statistically significant.

Total lung heterogeneity as well as regional lung heterogeneity (injured and noninjured regions) increased substantially after injury for both bronchial and pulmonary blood flows. Mean pulmonary blood flow heterogeneity increased in the injured region from 59 to 134% (P < 0.05). The increase in the noninjured region was not as large, a 23% difference (P < 0.05). Mean bronchial blood flow heterogeneity also increased in the injured region from 70 to 86% (P = 0.07). Mean bronchial blood flow heterogeneity in the noninjured region increased 32% (P < 0.05). Total left lung heterogeneity increased 39% for the pulmonary circulation and 27% for the bronchial circulation (P < 0.05).

The large change in CVs from pre- to postinjury suggests substantial redistribution of both bronchial and pulmonary flows within regions, more clearly illustrated by partitioning the total flow variation. The total flow variation across all pieces and observation times can be partitioned into two components. One component is a "piece effect," which can be thought of as a fixed unchanging allocation of relative flow to each piece occurring both pre- and postinjury. The second component of variation results from flow changes caused by a combination of injury, random time variation, and methodological error. On average, 48% of relative pulmonary flow variation to the injured lung region was due to piece effect; the remaining 52% represents relative flow changes due to injury, time variation, and method error (Table 3). In the noninjured region, 67% of pulmonary flow variation was caused by piece effect, and 33% was due to changing allocation.

The bronchial flow also showed large components of variation due to change, although these were smaller than for pulmonary flow. The change components for the bronchial circulation were 30 and 23% in injured and noninjured regions, respectively (Table 3).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The present study shows that after acid-aspiration injury to a segment of the lung, pulmonary blood flow decreases while bronchial blood flow increases to the injured pieces. This reinforces previous observations that these two blood flows act differently and probably respond independently to injury (9, 13). This study also adds spatial information to this independence. We have shown that this decrease in pulmonary blood flow is greatest at the center of the injured area and that the fall becomes progressively less with distance from the injured area. In contrast, the increase in bronchial blood flow is greatest at the center of the injury and tends to fall off further away from the injury. The absolute increase in bronchial blood flow was considerably less than the absolute decrease in pulmonary blood flow in the injured pieces.

Stothert et al. (13) have shown that injury from smoke inhalation in sheep caused a 6-fold increase in systemic blood flow to the lower trachea and an 11- to 14-fold increase in the intrapulmonary central airways. They also observed an increased hyperemic response as airway diameter decreased from the trachea to 2-mm airways. In airways <2 mm, the hyperemic response diminished, suggesting regional differences in the reactivity in the bronchial circulation.

Stothert et al. (14) showed in a chronic sheep model that airway injury caused by aspiration of hydrochloric acid increased bronchial blood flow and that this increase could be attenuated by pretreatment with cyclooxygenase inhibitors. Inhibition of thromboxane synthetase did not have such a protective effect. The precise mechanism that causes pulmonary blood flow to decrease and bronchial blood flow to increase to an injured area is unknown. Mechanisms responsible for the fall in pulmonary blood flow after injury have been attributed to the release of vasoactive amines, intravascular thrombosis from white cell and platelet aggregations, and hypoxic pulmonary vasoconstriction (7). The bronchial circulation responds differently to hypoxia. Charan et al. (4) showed that anastomotic bronchial blood flow (that portion of the bronchial blood that drains into the pulmonary circulation) increased with systemic hypoxia. Warren and Powell (17) showed that acute alveolar hypoxia increased bronchopulmonary anastomotic flow in dogs. These investigators also observed cyclooxygenase inhibitors to block bronchial vasodilatation. Similarly, Wagner and Mitzner (16) showed in sheep that systemic hypoxemia (arterial PO₂ 30-45 Torr) decreased bronchovascular resistance. Although our animals were not hypoxemic (arterial PO₂ >200 Torr), it is possible that alveolar hypoxia existed in the injured area.

Deffebach et al. (5) have shown in dogs that pretreatment with cycloxygenase inhibitors attenuates the increased bronchial blood flow after both intravascular and airway injury. An interesting finding in our present study is that changes in pulmonary and bronchial blood flows diminished as one proceeded away from the central injury, even though several of the pieces away from the center of injury were visibly injured. This would suggest that vasoactive amines and other local stimuli in the center of the injured areas may affect the upstream vasculature and alter flow to the entire region. The precise underlying mechanism is unknown, but possibilities include 1) diffusion of vasoactive amines along bronchovascular bundles to the upstream flow regulating segments of the arteries, 2) neural reflexes, or 3) cell-to-cell communications.

Marshall et al. (10) reported that bronchial arterial hypoxemia attenuates hypoxic pulmonary vasoconstriction. Thus one role for an acute increase in bronchial blood flow to an injured area may be preservation or accentuation of the hypoxic pulmonary vasoconstriction that diverts pulmonary blood flow away from the injured area, thus minimizing low ventilation-to-perfusion ratio regions in the injured area of lung. In chronic inflammatory states, such as lung abscess or bronchiectasis, angiogenesis and new vessel formation develop in the bronchial circulation, but this regenerative power does not exist in the pulmonary circulation. The signal for angiogenesis is unknown, but an interesting hypothesis is that, during the acute phase of injury, dilation of the bronchial vasculature would provide oxygenated blood to the area of injury and, possibly, also stimulate new vessel formation. Although intravascular thrombosis occurs in the pulmonary circulation, this does not seem to occur in the bronchial circulation, as evidenced by an acute increase in bronchial blood flow. The bronchial circulation may be somewhat protected from acid aspiration by its submucosal position in the airway, whereas the pulmonary circulation may not because of its thin alveolar capillary barrier.

The absolute rise in bronchial blood flow was substantially less than the fall of pulmonary blood flow. In contrast to the pulmonary circulation, the bronchial circulation normally plays a minor role in gas-exchange function. Because the bronchial circulation transports arterialized oxygenated blood to the lung, it participates little in oxygen uptake, but does excrete carbon dioxide (15). The role of increased bronchial blood flow in response to inflammation in the injured area, as well as its beneficial effects in recovery, are unknown. A small amount of oxygen-rich bronchial blood perfusing an injured area may be sufficient to provide nutrients and cells to overcome the injury or to prevent hypoxia in cells of the injured area that may become necrotic in the face of decreased ventilation and pulmonary perfusion. Thus increased bronchial blood flow during injury may serve a nutrient rather than a gas-exchange function. Further studies are required to delineate the precise mechanisms that underlie the bronchial blood flow increases and the role they play in recovery from lung injury.

Bronchial blood flow heterogeneity (CV) has not been previously reported. Due to the branching nature of the airways and their blood supply from the bronchial circulation, one would expect bronchial blood flow to be heterogeneous in the lung. Before injury, bronchial blood flow heterogeneity was 81%. The increases in heterogeneity of both pulmonary and bronchial blood flow after segmental lung injury are a reflection of the marked changes in blood flow to the lung after injury. Bronchial blood flow increased in the injured region while pulmonary blood flow decreased. Both changes were not uniform in magnitude, since the CVs changed. Interestingly, heterogeneity increased more in the regions that lost flow (injured region for pulmonary flow and noninjured region for bronchial). This may be due to a heterogeneous lung injury or a nonhomogeneous humoral response to the injury.

The variation in pulmonary or bronchial blood flow between the many pieces in a lung, and between two points in time (without a perturbation), could be due to the structure of the vascular bed, to redistribution of blood flow, temporal changes in blood flow, or method artifact. Based on components of variation, our analysis permits partition or quantification of the relative contribution of the above-mentioned factors to the variation in blood flow within an experiment that includes a perturbation, in this case, aspiration injury due to hydrochloric acid. Previous studies from our laboratory (2, 8) on the effect of increased gravitational force or increased cardiac output due to exercise showed that, in normal noninjured lungs, structure accounts for ~70% of the blood-flow variation and thus is the major determinant of pulmonary blood flow. In this context, structure means a stable factor that remains stable regardless of time or physiological perturbation. Redistribution, temporal changes, and method error accounted for <30% of the observed variation in pulmonary blood flow. This is in contrast to what occurs in an area of injured lung. Our present study shows that 48% of total variation in pulmonary blood flow in an injured lung could be attributed to "structure," whereas the remaining 52% were due to a combination of redistribution of blood flow, time, and method error. In contrast, 70% of the total variation in bronchial blood flow was due to structure and 30% to redistribution, time, and method noise in the injured regions. This may be due to a greater structural integrity of the bronchial circulation then that of the pulmonary circulation. The causes underlying the behavior differences of the two circulations are unknown. Possibilities include the magnitude of decrease in pulmonary blood flow and increase in bronchial blood flow, the size (area) of the two circulations, or sites and mechanisms of vascular regulation for the two circulations. It is also not known what changes would be observed if the route of injury was intravascular or via the airway injury, and further studies are warranted.