INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES APPENDIX

AS RED BLOOD CELLS MOVE from pulmonary arterioles to venules, they cross several alveolar walls, each of which contains an intricate network of capillaries. The total number of capillary segments crossed from arteriole to venule varies among species; in our canine preparation, the average is ~60 (5). Blood perfuses the capillaries in a complex manner: some segments are nearly always perfused and form interconnecting pathways across the alveolar wall (8, 9), while in other regions of the same alveolar wall, the blood frequently switches among segments, turning them on and off (7, 8). In the present study, we concentrated our analysis on the pattern of switching among segments within individual alveolar walls.

In previous work (12), we observed alveolar-capillary networks for a period of 1 min; if a single red blood cell moved through a capillary segment during that 1-min time period, we considered that segment to be perfused. That approach provided a reasonable measure of the functional gas exchange surface area. Because the blood switched around among the segments comprising the network many times in <1 min, the 1-min time periods underestimated the switching frequency. Therefore, in this study, we shortened the observed time period to 4 s. From continuous observations made over a 16-min-long period, we determined the state of perfusion (on or off) of each capillary segment in the network during each 4-s interval. From these measurements, we obtained a data stream from each alveolar-capillary network of sufficient length to apply fractal analysis and thereby determine whether the capillary perfusion pattern was random or nonrandom. In addition, we compared the patterns among neighboring alveoli. The results were unexpected and counterintuitive.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES APPENDIX

Animal preparation. In accordance with our institutional guidelines, healthy adult male mongrel dogs (18-27 kg, n = 5) were anesthetized with pentobarbital sodium (30-40 mg/kg iv), intubated, and mechanically ventilated. After heparinization (1,000 U/kg), the animals were rapidly exsanguinated via the left common carotid artery. After a left thoracotomy, the left lower lobar pulmonary artery was cannulated, and the left lower lobe was excised, along with a cuff of left atrium, and placed on a microscope stand. The left atrial cuff was secured around another cannula, and the lobe was perfused with autologous heparinized whole blood (hematocrit = 30-41%). The time interval to reperfusion was <30 min. Blood was pumped by a Masterflex 7522-10 pump drive and 7024-20 pump head that was controlled by a feedback system that kept flow constant (<1% variation). After it left the pump, the blood flowed through a windkessel to dampen high-frequency pressure oscillations from the pump as well as to trap bubbles, through a filter (20-µm pore size, Fenwal 4C7700) to remove microaggregates, and through a heat exchanger (Bentley HE-30) to warm the blood to 37-38°C before entering the lobe (Fig. 1). Venous blood drained passively from the lobe into a reservoir. The lobe was ventilated (model 607D, Harvard Apparatus) with 6% CO₂-17% O₂-77% N₂ at a tidal volume of 100 ml. This produced blood gas tensions in the normal range for arterial blood. End-expiratory pressure was set at 5 mmHg. Pulmonary arterial and venous pressures and airway pressures were monitored continuously with transducers (Statham P23 XL) that were zeroed at the site at which the microcirculatory observations were being made. Pulmonary venous pressure was set at 1 mmHg by adjusting the height of the reservoir, and pump flow rate was set at a value [337 ± 75 (SD)] which resulted in a pulmonary arterial pressure of 10-15 mmHg. These zone 2 conditions caused about one-half of the capillaries to be perfused in the observed field. The lobe was suspended by two small spring-backed paper clips attached to opposite edges of the lobe and raised until the uppermost pleural surface came into contact with a transparent window (Fig. 1). A 1.3-cm² area on the surface of the lobe was observed through the window, which was surrounded by a vacuum ring to prevent lateral tissue movement (10).

View larger version (29K): [in this window] [in a new window]   Fig. 1.   Schematic of apparatus used to perfuse left lower lung lobe under conditions of steady blood flow, microvascular pressure, and alveolar inflation pressure. CCD, charge-coupled device. PA, alveolar pressure; Ppa, pulmonary arterial pressure; Ppv, pulmonary venous pressure.

Microcirculatory observations. The subpleural microcirculation under the window was observed with a Leitz Ultropak surface-illuminating microscope (×11 objective) coupled to a 200-W mercury arc lamp. The light was heavily filtered to prevent tissue damage by infrared and ultraviolet light. A narrow band-pass interference filter was used to illuminate the field by using only the mercury green line (546 nm). This wavelength was absorbed by hemoglobin, thereby increasing the contrast between the erythrocytes and surrounding tissue (11).

View larger version (27K): [in this window] [in a new window]   Fig. 2.   Tracings of 3 alveolar walls (A-C) on surface of lung. Each capillary segment that was perfused by at least 1 red blood cell during the 16-min observation period was traced and numbered as shown on the figure. All 3 alveolar neighbors were fed by same arteriole and drained by same venule. In this isolated pump-perfused lobe, pressure and flow were held constant throughout the observation period. Average alveolar diameter was ~100 µm.

Data analysis. To analyze the switching pattern, we progressively divided the set of 240 4-s data points from the 16-min video recording into subsets by using all integer factors of 240, i.e., 2, 3, 4, 5, 6, 8, ..., 80, 120. At each step, the number of perfused segments within subsets was pooled. For example, when we divided the set of 240 by 2, we obtained two 8-min-long subsets, each containing 120 measurements of the number of perfused segments in 4-s intervals. After pooling the 120 measurements in each subset, we obtained two measurements of the number of perfused segments, one for each of the two 8-min observation periods. The relative dispersion (SD/mean) of the number of perfused segments was calculated for each integer division of the 240 data points. Plots of the log of the relative dispersion of the number of perfused segments vs. the log of the size of the time interval were made. The fractal dimension (D_t) was calculated from the slope of these plots (4).

Statistics. To test for stability of the experimental preparation, we used the paired two-tailed t-test to compare blood-gas and pressure measurements made at the beginning of the study with those made at the end of the study. Pressure measurements made every 5 min were tested for changes over time by using a two-way analysis of variance.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES APPENDIX

The path of the flowing blood frequently switched among capillary segments, producing a fluctuation in the number of perfused segments (Fig. 3, left). The log-log plots of the relative dispersion vs. the size of the time interval (Fig. 3, right) were linear [r² = 0.94 ± 0.04 (SD)] up to a time interval of ~100 s, thus supporting a fractal relationship. Fractal analysis showed the fluctuation to be nonrandom. The average D_t of three adjoining alveoli in each of five animals was 1.12 ± 0.04 (Table 1). This demonstrates that the fluctuation in the number of perfused segments was ordered and self-similar in the time domain (a D_t of 1.0 indicates a perfectly uniform, temporally correlated process, whereas a D_t of 1.5 indicates a random process without temporal correlation).

View larger version (33K): [in this window] [in a new window]   Fig. 3.   Left: patterns of fluctuation in number of perfused segments from 3 alveolar walls shown in Fig. 2, A-C. Each 16-min recording of the alveoli was divided into consecutive 4-s observation periods. A segment was considered perfused during a 4-s observation period if 1 or more red blood cells flowed through the segment during that period. No. of perfused capillary segments during each 4-s period is shown for each alveolus. Right: plots of log of relative dispersion (SD/mean) of no. of perfused capillary segments vs. log of the size of measurement time interval (alveoli A, B, and C in Fig. 2). Fractal dimension (Dt) was calculated from slope of plots (4). For all alveoli, average Dt was 1.12 ± 0.04 (SD) with an average r2 of 0.94 ± 0.04 for fit to the plot. A Dt of 1.0 indicates a perfectly uniform, temporally correlated process, whereas a Dt of 1.5 indicates a random process without temporal correlation. Dt were sufficiently close to 1.0 to indicate that capillary perfusion within each alveolus was a temporally correlated process that was self-similar.

When we increased pressure and flow, we were able to estimate the maximum number of capillary segments in each network. Of the 15 alveoli studied (3 each in 5 animals), 9 alveoli had every segment perfused at least once during the 16-min period, 4 alveoli had all but 1 segment perfused, and 1 alveolus had all but 3 segments perfused. Of the total of 287 segments perfused when flow and pressure were raised, 97% were perfused at some point during the 16-min period of observation.

Using linear regression, we found virtually no relation of perfusion patterns between any of the alveoli (see APPENDIX); the mean r² for all pairs was 0.06 ± 0.06 (SD) (Table 2). The perfusion relationships among the three adjoining alveoli shown in Fig. 2 are plotted in Fig. 4.

View larger version (18K): [in this window] [in a new window]   Fig. 4.   Cross correlations between neighboring alveoli in Fig. 2. Each no. on axes gives no. of perfused segments at same time in a given alveolus vs. its immediate neighbor. Mean r2 for all pairs was 0.06 ± 0.06 (SD). Very low r2 values show that there was virtually no correlation in capillary perfusion patterns in adjacent neighbors (see APPENDIX), i.e., the perfusion pattern of each alveolar wall was independent of neighboring alveoli. In each graph, 240 points were plotted, but many overlie each other.

The preparations were stable over the time course of the experiment. Perfusion pressures and blood-gas values were in the normal range (Table 3), with no difference between values at the beginning of the experiment and those at the end of the experiment. Furthermore, there was no variation among the pulmonary arterial or pulmonary venous pressure measurements that were made every 5 min (P  0.3).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES APPENDIX

For decades, we have considered the noisy-looking pattern of pulmonary capillary perfusion to be random, the result of highly flexible red blood cells coursing through complex capillary networks. Another possible cause of blood-flow switching between segments could be subtle alterations in pressure and flow into the capillary network. To test these ideas, we made a considerable effort in these experiments to maintain stable conditions. Flows were held steady by the pump-perfusion system. Pulmonary arterial and venous pressures were constant, thus indicating that pulmonary vascular resistance was stable (Table 3). Blood-gas values, blood temperatures, and airway pressures were stable as well (Table 3). We expected that stabilizing these criteria would optimize the conditions for stable capillary perfusion. Despite these efforts, the path of the flowing blood switched frequently among capillary segments, causing the number of perfused segments to vary remarkably over time (Fig. 3, left). The application of fractal analysis unexpectedly showed that the fluctuation in the number of perfused segments was not random (Fig. 3, right), as each alveolar-capillary network had a perfusion pattern that was self-similar over time (Table 1).

One design consideration in these experiments was selection of the observation period to accurately assess the switching frequency. Previously, we had observed alveolar-capillary networks for a period of 1 min (12). If a single red blood cell moved through a capillary segment during that period, we considered it to be perfused. Although this was practical, this time period underestimated the switching frequency, because numerous switches could occur in <1 minute. As we reduced the time of observation to very short periods, however, we found that another error occurred in the opposite direction. To illustrate the new error, consider three capillary segments connected in series. If a red blood cell traveling through the feeding segment was separated by a large enough plasma gap from a second red blood cell traveling through the draining segment, the middle segment would be considered unperfused, because no red blood cell was perfusing it. When the first red blood cell entered the middle segment, that segment was considered perfused. If both red blood cells separated by the plasma gap could be observed to be moving along a single path at the same velocity, the observer would readily deduce that the middle segment was perfused, albeit with plasma alone. Unfortunately, it was not always possible to make this kind of deduction accurately. Through trial and error, we determined that if a 4-s observation period were used for each segment when perfused by blood of normal hematocrit, the switching could be reasonably assessed while avoiding overestimation of switching by counting plasma gaps as unperfused segments.

Because the observed capillary networks lay between a single feeding arteriole and a single draining venule and had perfusion patterns that repeated over time, it seemed reasonable to expect that the perfusion of neighboring alveoli would be correlated. One likely possibility was that perfusion of neighboring alveoli would be positively correlated. Although overall pulmonary hemodynamics were stable, regional alterations in the incoming arteriolar flow might cause the three neighbors to increase or decrease their level of perfusion in concert, thereby correlating positively with each other. Another possibility was a negative correlation; e.g., if flow increased in alveolus A, it might do so at the expense of alveolus B or C. Stealing of this kind would lead to a negative correlation. The possibility no one anticipated was that there would be virtually no correlation of the number of perfused segments in adjoining alveolar walls. That surprising and counterintuitive result, however, is what we found (Fig. 4, Table 2).

The lobar perfusion conditions were set in zone 2 (pulmonary arterial pressure > pulmonary venous pressure > alveolar pressure, all pressures being referenced to the site of microcirculatory observations). There was, however, considerable variation in perfusion of individual alveolar walls (Fig. 3). There were instances in which a given alveolar wall was not perfused at all (zone 1), while its immediate neighbor was partially perfused (zone 2) or even fully recruited (zone 3). Examples of those conditions are shown by the points lying on either the abscissa or ordinate in Fig. 4. These observations suggest that multiple perfusion zones can exist in the ultimate isogravitational plane, a perfusion plane that is one alveolar wall thick. These data provide support for the idea, recently advanced by Glenny and Robertson (3), of multiple perfusion zones present in isogravitational planes.

The lack of correlation between neighboring alveoli (Fig. 4) is remarkable for two reasons. First, the alveolar networks in each preparation were fed and drained by common vessels, suggesting that inlet and outlet conditions were similar for each alveolus. Second, when larger volumes of lung are compared, blood flow in neighboring regions is similar. The smallest regions compared in this way are 1.2-cm cubes. Even in regions that small, Glenny (1) showed that flow distribution is similar among neighboring cubes, i.e., the r² for adjoining 1.2-cm cubes is ~0.51, P < 0.001. Thus, if a given small region had high flow, the adjacent small regions were likely to have high flow, and, conversely, low-flow small regions were likely to have low-flow neighbors. Furthermore, Glenny et al. (2) found these flow relationships were stable over time: high-flow regions had high flow on each of 5 consecutive days, and low-flow regions consistently had low flow (r² > 0.85). These data indicate that, as flow progresses from generation to generation of the pulmonary arterial tree, regional blood flow is similar to that of its neighbors and is determined by stable structural characteristics of the pulmonary arterial branches. Given the stable flow and pressure conditions of our preparation and the high coefficients of determination for adjacent small regions of lung, it is striking that the similarity of flow relationships throughout the pulmonary arterial tree disappears so completely once the capillary bed is reached. That is, although neighboring flow patterns do correlate on a macro scale, there is virtually no relation between flow patterns in neighboring alveolar walls, with r² values approaching zero.

These data support the idea that the perfusion pattern within each alveolar wall is autonomous and independent. Therein, we speculate, lie two important design features of the lung that provide robustness. First, fractal patterns are themselves robust (13), i.e., fractal patterns that repeat over time will, if perturbed, return to their original repeating pattern on their own. Thus the fractal character of perfusion of an individual alveolar-capillary network is inherently robust. The second characteristic that is robust is the independence of neighbors. If there were dependence of neighboring alveolar perfusion patterns, especially if they were tightly correlated, then, if the perfusion pattern of one alveolar wall were disrupted, the neighbors would likely be affected in a potentially disruptive way. The independent perfusion pattern displayed by alveolar capillaries makes them appear to be impervious to the behavior of their neighbors.

Ideally, a robust design must contain both a flexible component to adapt to new conditions and a more rigid component to provide long-term continuity. The functional independence of individual alveolar-capillary networks, as shown by the very low r² values, is the flexible component that can adapt to new conditions and is undisturbed by the perfusion state of its immediate neighbors. The fractal nature of perfusion, the automatically repeating component of perfusion within an individual capillary network, may serve as the more rigid component.

Whether the cause of the switching is active or passive is not clear. The work of Kiani et al. (6) on the cheek pouch of the hamster has shown that, in the absence of vasomotion, fluctuations in flow among arteriolar branches occur as a result of the particulate nature of blood and the nonlinear relationship between the partitioning of cell flow and total blood flow at bifurcations. Alternatively, active mechanisms may also be present. Recently, Ying et al. (15) demonstrated oscillations in the intracellular calcium concentration of rat pulmonary capillary endothelial cells. Such oscillations could be associated with changes in capillary tone, thereby having an effect on the distribution of flow within the capillary network. Because data from these and other experiments can be used to generate arguments either supporting the idea that the fractal perfusion pattern and the switching pattern are caused by an active switching system, or that the switching is the passive result of a particulate fluid moving through a complex network, more experiments will be required to settle the issue.

In either case, the perfusion pattern of each alveolar wall is created by the unique characteristics of its own network. These patterns result in function that is independent of neighboring walls. If this reasoning is correct, then the capillary bed has, in addition to its delicate yet remarkably strong walls (14), an unsuspected elegance of design that imparts a robust character to the perfusion of pulmonary gas exchange vessels.