| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Departments of 1Physiology and Biophysics and 2Medicine, University of Washington, Seattle 98195-6522; and 3The Mountain-Whisper-Light Statistical Consulting, Seattle, Washington 98112-2913
Submitted 28 February 2003 ; accepted in final form 18 November 2003
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONREFERENCES |
|---|
) and alveolar ventilation (
A) were measured by using intravenous infusion of 15 µm and inhalation of 1-µm fluorescent microspheres, respectively. PrO2 was calculated for each piece at each FIO2. Lung pieces differed in their response to hypoxia in a manner related to their initial A/ with FIO2 = 0.21. Reducing FIO2 < 0.15 decreased to the initially high A/ (higher PrO2) regions and forced into the low A/ (dorsal-caudal) regions. Resistance increased in most lung pieces as PrO2 decreased, reaching a maximum resistance when PrO2 is between 40 and 50 Torr. Local resistance decreased at PrO2 < 40 Torr. Pieces were statistically clustered with respect to their relative response pattern to each FIO2. Some clusters were shown to be spatially organized. We conclude that HPV is spatially heterogeneous. The heterogeneity of response may be related, in part, to the heterogeneity of baseline A/. hypoxia; pulmonary circulation; high-altitude pulmonary edema; heterogeneity; ventilation-perfusion; fluorescent microspheres
) away from the hypoxic area (21). HPV is thought to be a primary adaptive mechanism of the pulmonary circulation necessary to preserve arterial blood oxygenation in the face of regional lung disease. It works by increasing pulmonary vascular resistance to hypoxic alveoli, thereby diverting pulmonary away from these regions. The reduction in from poorly ventilated (hypoxic) alveoli to better ventilated alveoli preserves the matching of alveolar ventilation (A)-perfusion (A/) and arterial oxygenation (6).
Whereas heterogeneity of HPV has been thought to be present in mammalian lung for some time, the importance of HPV heterogeneity was first hypothesized by Hultgren (16, 17). It was thought that heterogeneity of HPV was the primary explanation for high-altitude pulmonary edema in that vasoconstriction in response to hypoxia occurred only in some regions. It was reasoned that a shift in from the constricted regions to the nonconstricted regions resulted in overperfusion, elevated capillary pressure, and increased fluid flux into the interstitium and eventually to the alveolar spaces. Heterogeneity of HPV response is a prerequisite for such a mechanism to work. The present study demonstrates regional heterogeneity of HPV in the mammalian lung by using direct measurements of regional A and from which regional PO2 (PrO2) is calculated.
This study was undertaken to determine the heterogeneity of HPV by quantifying the HPV response within 
2.0 cm3 pieces. We chose the supine pig (without positive end-expiratory pressure) in order to maximize A/ heterogeneity. We hypothesized that reduction of inspired O2 fraction (FIO2) would initially cause vasoconstriction and a decrease in in the lower A/ regions with lower PrO2. As a direct consequence, would shift toward high A/ regions, increasing in these pieces with high PrO2. Eventually, with further reduction of FIO2, initially higher A/ (and PO2) regions would begin to vasoconstrict, shifting toward the originally low A/ regions. Because we expected regional A distribution to change only slightly, the anticipated effect would be an increased A/ heterogeneity with reduced FIO2.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
The experimental protocol was approved by the University of Washington Animal Care Committee. Pigs of either sex (n = 5, weighing 21.5 ± 0.4 kg) were premedicated with 2 mg/kg xylazine and ketamine 20 mg/kg im hydrochloride. After placement of an ear vein catheter, anesthesia was maintained with continuous infusion of thiopental sodium titrated to suppress hemodynamic and motor responses to noxious stimuli. A tracheotomy was performed, and a cuffed 7.0 mm endotracheal tube was inserted and tied in. Animals were mechanically ventilated with a constant-volume piston pump (Harvard Apparatus, South Natick, MA), at a tidal volume (VT) of 14 ml/kg body wt. The respiratory rate (RR; 14-20 breaths/min) was adjusted to maintain arterial PCO2 between 35 and 40 Torr and held constant throughout the experiment. The pigs were ventilated with air during surgery. FIO2 was varied during the experiment, as described below. The animals were placed on a heating pad to keep body temperature between 37.2 and 39.6°C.
Preparation and Measurements
A catheter was placed in a femoral artery to monitor systemic arterial pressure and to permit blood-gas sampling. An introducer was placed in the right external jugular vein, and a 7-French Swan-Ganz thermodilution catheter (Baxter, Irvine, CA) was advanced into the pulmonary artery to measure mean Ppa, pulmonary capillary wedge pressure (Ppcw), and temperature, and for blood sampling. Cardiac outputs (thermodilution technique) and blood temperature were measured with a cardiac output computer (Baxter Edwards Sat-2, Irvine, CA). Femoral venous catheters were inserted for infusion of anesthetic, fluid maintenance, and microsphere administration. Respiratory pressure, VT, RR, and minute ventilation were measured continuously with a digital spirometer (KORR, Medical Technologies Research Spirometry System, Salt Lake City, UT). End-tidal (expiratory) CO2, RR, and systemic arterial PO2 (PaO2) were continuously monitored with a CO2SMO (Nova Metrix Medical System, Wallingford, CT). Blood pressures, heart rate, as well as airway pressure (Paw) were measured with a Mark 12 Data Managed System, DMS 1000 (Graphtec, Irvine, CA). Vascular pressures, Paw, end-tidal CO2, and VT were digitally recorded with PowerLab, ADInstruments (Grand Junction, CO) on a PowerPC. Arterial and mixed-venous blood gases and hemoglobin were analyzed with ABL 5 and OSM 3 hemoximeter machines (Radiometer, Copenhagen, Denmark). Inspired O2 concentrations were measured spectrometrically with a MGA-1100 (Perkin-Elmer Medical Instruments, Norwalk, CT).
Fowler Dead Space
Lungs were briefly inflated and held at total lung capacity for a few seconds. After 1-2 min of normal ventilation, expired CO2 concentration and expired lung volume were sampled 100 times per second by PowerLab, and dead space was estimated by Fowler's method (10).
Protocol and Microsphere Administration
After dead space determination, animals were ventilated with a series of FIO2: 0.50, 0.21; then (in random order) 0.15, 0.12, 0.09; and a repeat 0.50 at the end. After 10 min at each FIO2, lungs were briefly inflated to total lung capacity in an attempt to expand any atelectasis. Five minutes later, during which physiological measurements were obtained, intravenous and inhaled fluorescent microspheres (FMS) were simultaneously administered for 5 min (1, 2). FMS of 11 different colors [intravenous (15 µm): blue, blue-green, green, crimson, carmine, scarlet; and aerosolized (1 µm): yellow-green, yellow, orange, orange-red, red], obtained from Molecular Probes (Eugene, OR), were used for these experiments. The order of colors given, both by aerosol and intravenously, was predetermined in a random fashion; one exception was carmine, which was always used to mark for the 0.50 FIO2 repeat, during which no aerosolized FMS were given. After FMS administration, arterial and mixed-venous blood samples were taken.
Shortly after the last microsphere injection at 0.50 FIO2, the animals were given heparin (5,000 units) and papaverine (30 mg) intravenously before being exsanguinated under deep anesthesia. A sternotomy was performed, and the lungs were perfused with a 2% dextran solution through the pulmonary artery until clear of blood. The lungs were removed from the chest, inflated to a Paw of 25 cmH2O, and dried for 4 days.
Lung Preparation and Data Normalization
Once dry, the lungs were coated with Kwik Foam (DAP, Dayton OH), suspended vertically in a plastic-lined squared box, and embedded in rapidly setting urethane foam (2-lb. Polyol and Isocyanate, International Sales, Seattle, WA) to create a rigid form to which a three-dimensional coordinate system was applied. The foam block was sliced and cut into uniformly sized 2.0-cm3 cubes. Foam adhering to lung pieces was removed, and each lung piece was weighed and assigned its three-dimensional coordinate, lobe designation, and percent airway code.
The fluorescent signal for each color was determined by extracting the fluorescent dyes from each piece with an organic solvent (Cello-solve, Sigma-Aldrich) and by measuring the fluorescence concentration in each sample (11). Spillover from adjacent colors was corrected by using a matrix inversion method (34).
Every piece in each animal's data set consists of x-, y-, and z-coordinates, lobe designation, weight, percent airway code, and fluorescent signal for and A at each FIO2. Each piece's fluorescent signal is linearly proportional to A or to that piece. The signal is converted to milliliters per minute by multiplying its fraction of the total fluorescence by the total A or cardiac output, respectively. All pieces weighing <8 mg were excluded (to minimize uncertainty in fluorescence and in weight). The A is calculated by subtracting the Fowler dead space from the VT and multiplying by the RR. Alveolar and arterial partial pressure and end-capillary contents for O2 and CO2 for each lung piece are determined by solving mass balance equations for each gas, given that piece's A/ (1). This approach assumes that all alveoli in the piece are in equilibrium. We employed the method of Altemeier et al. (1) to simultaneously measure regional A and in 2.0-cm3 cubes of lung with microsphere techniques. This allows the calculation of PrO2 within each piece of the lung. A comparison of the -weighted PO2 sum to measured PaO2 is show below.
To compare A and among pieces, some being incomplete tissue cubes, each piece was weight normalized (WN) by dividing piece fluorescence by piece weight. To minimize the effect of nonalveolar tissue on weight normalization, pieces consisting of >20% airway tissue (as determined by eye) were not included (averaging 56 pieces out of an average total of 1,008 per animal). Mean lung piece weight was 30.7 ± 13.7 (SD) mg. Furthermore, we assigned relative (R) and relative A (RA) values to each piece by dividing the piece WN value at a given FIO2 by the average WN value of all pieces in the lung at that FIO2 setting. After this step, the mean WN relative value at the given FIO2 setting had a mean of unity for both WNR (WNR) and WN RA (WNRA). This WN, -normalized adjustment allowed comparison among animals with different cardiac outputs and, within each animal, comparison of at different FIO2 settings. Piece resistance was calculated as (Ppa - Ppcw)/WN to that piece.
Statistical Analysis
All data are presented as means ± SD, except where noted. The goodness of the linear fit, R2, between two variables is used to quantify the strength of the relationship. Statistical significance is assumed when P < 0.05. The coefficient of variation (SD/mean) is used to characterize the heterogeneity of and A at each FIO2. For comparison at various FIO2, we use statistical significance based on repeated-measures ANOVA and Fisher's paired least significant difference post hoc test.
Cluster analysis. The data are analyzed to find clusters of lung pieces that trace a similar pattern of changes in (therefore, resistance) vs. FIO2. This procedure is completed without consideration of spatial location or the PrO2. Once a cluster of pieces is defined, the spatial location of these pieces is investigated as the potential locus for the associated regional HPV (HPVR) response.
Cluster analysis is a statistical method for grouping items, such as pieces of lung, into "clusters" that share similar characteristics (8). A description of the clustering algorithms can be found in Glenny et al. (13). The observations used in the cluster analysis consist of each piece having its own unique pattern of values across the five settings of FIO2. Clustering was also carried out simultaneously on all pieces from all animals merged together (see Metaclustering, below).
Hierarchical clustering was implemented in S-Plus (Insightful, Seattle, WA) to define an initial set of clusters defined by HPVR response. Then a method of "sharpening" was used that reallocates pieces in very small clusters into the large clusters that are similar to them (13). Once the clusters are defined, a process that is carried out without reference to the x-, y-, and z-coordinates of the pieces, the spatial location of the pieces in a cluster is then of particular interest. Descriptive statistics for location and size of the cluster (based on x-, y-, and z-coordinates) consist of the cluster centroid, the mean distance of pieces to the cluster centroid, and shape and orientation parameters defined by principal components of the spatial coordinates.
Metaclustering. Pieces are grouped together that have a common hypoxic-response pattern across animals. This simple technique is designated as "metaclustering." In metaclustering, the data sets for all animals are merged, with individual lung pieces represented in rows of the data set while the columns contain the response to experimental conditions, such as the WNR at specified hypoxia levels. Because the intent of the clustering is to find pieces with a similar pattern of changes in across the FIO2 settings, the clustering was carried out, not on the five values, but, instead, on the deviations of around the mean . Specifically, for each piece, the residuals of , after subtracting the mean, were used in the clustering. Thus, for example, a piece with WNR of (0.4, 0.6, 0.5, 0.6, 0.4) for the five FIO2 settings, respectively, would fall in the same cluster as a piece with the same changes, but with values of (1.4, 1.6, 1.5, 1.6, 1.4). The two pieces have quite different means, but both have residuals of (-0.1, +0.1, 0.0, +0.1, -0.1) and would fall in the same cluster.
A small adjustment to the residuals was also carried out for each animal before clustering. The motivation for the adjustment was that each animal might have its own unique level of within-piece variability of , with one pig, say, being generally more variable across the FIO2 settings than another pig. This could have an undesirable effect on the metaclusters. As an extreme example, which is unrealistic but illustrative, suppose that all of the pieces of one animal had residuals of (-0.1, +0.1, 0.0, +0.1, -0.1), and all of the pieces of the second animal had residuals with twice the variability, namely, residuals of (-0.2, +0.2, 0.0, +0.2, -0.2). In this case, the two pigs would fall into two completely different clusters (hardly "meta" clusters), even though their patterns are proportional and similar. In order to avoid any tendency for individual pigs to dominate any of the derived clusters, as in this example, we multiplied all of the residuals for each pig by a factor such that the mean of the within-piece variances was the same for all animals. This multiplicative factor was close to unity for all pigs, indicating that there was only a minor difference among the five pigs in their variability of across the FIO2 settings. After adjusting the residuals, the clusters of pieces were then derived for the full data set, including pieces (and their adjusted residuals) from all animals.
We compared the five metaclusters to determine whether, in addition to the significantly different spatial distribution of the clusters within the lung, the PrO2 values also differed. We used Kruskal-Wallis nonparametric ANOVA to test the null hypothesis that, when FIO2 = 0.21 (room air), the PrO2 values for pieces were not associated with the five metaclusters. Specifically, the null hypothesis is that the five metaclusters have the same mean PrO2. In addition, we compared the mean PO2 values between each pair of metaclusters (e.g., cluster 2 vs. 5) using the same procedure. For the 10 pairwise comparisons of metaclusters, we used the Bonferroni correction by multiplying each P value by 10 (with a maximum P value of 1.0).
Spatial location of HPV clusters and the metalung. Our most effective method of exploring the spatial location of HPV clusters is with rotating three-dimensional plots of the lung and its pieces, with each piece color-coded to its particular HPVR cluster. However it is not possible to present such a display in a paper. Figure 4 shows an example of a color-coded cluster display that shows obvious spatial clustering of HPVR response in two views (lateral and cranial) in one of the animals. The clusters for this display were determined based only on the data from this animal. To display and test for spatial clustering using data from the metaclusters, a "metalung" was created by linear transformations (stretching and compressing) along the x-, y-, and z-axes of the coordinates for each animal. Specifically, the range of the x spatial dimension was determined for each animal's lung and averaged across the five animals to give the range of x (Rx), for an ideal metalung. Similar ideal ranges were determined for the y-and z-dimensions (Ry and Rz). The x-coordinates of every piece for every animal were then linearly transformed to "ideal" coordinates so that all animals had ranges Rx, Ry, and Rz for the three dimensions. The animal lungs were quite similar in size and shape, and the stretching and compressing along the various dimensions were very modest. All subsequent spatial analysis was carried out by using the ideal coordinates. Figures 1, 2, 3, 4, 5, 6, 7, 8 are shown using the original animal lung coordinates.
|
|
|
|
|
|
|
|
The distribution of the HPV clusters within the three-dimensional lung can be examined to determine whether there is spatial clustering of the pieces. Under the null hypothesis, there is only a random relationship between a -based cluster and its spatial location within the lung. Such a cluster would not differ spatially from a randomly selected sample of pieces from the lung. Thus, the expected mean of the x-, y-, and z-coordinates from such a random cluster would be the same as the mean coordinates for the entire lung. Similarly, the expected mean distance of a piece to the centroid of its random cluster would be the same as the mean of the distances of all pieces in the lung to the lung centroid. In this study, the x-, y-, and z-coordinates of the centroid of each cluster and the mean distance of each piece to the cluster centroid were compared to the corresponding values for the entire lung by using a single-sample t-test. Principal component analysis was used to provide a descriptive statistic on the relative volume of the clusters compared with the whole lung.
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Physiological data are shown in Table 1 for each run. With progressive hypoxia, vasoconstriction results in an increase in total pulmonary resistance and an increase in Ppa. Cardiac output remains unchanged except at the lowest FIO2. Thus the overall pulmonary vascular resistance increases with hypoxia (as FIO2 is reduced from 0.50 to 0.12) but reduces slightly at the lowest FIO2 of 0.09. Systemic arterial pressure increases with the increase in cardiac output. PaO2 reduces with reduced FIO2. Arterial PCO2 remains constant, except for a slight hypercapnia at the most hypoxic inspired FIO2. For comparison, the maximum resistance (Table 1) occurs at an alveolar PO2 of 40 Torr [estimated by the method of Altemeier et al. (1)].
|
Microsphere Data: A, , and A/ Distribution
The overall matching of A and can be determined by evaluation of A and within all of the pieces within each animal. An average for all five animals is presented in Table 2. Progressive decreases in hypoxia lead to a progressive increase in A/ heterogeneity as measured by the -weighted SD of the natural log of A/, which increased from 0.32 to 0.44, 0.61, and 0.79 for FIO2 = 0.21, 0.15, 0.12, and 0.09, respectively. Figure 1 illustrates WNR as a function of FIO2 (Fig. 1A) or PrO2 (Fig. 1B) among pieces of differing A/ and thus differing PrO2 for one animal (greater A/ corresponds to greater PrO2). As FIO2 is reduced, PrO2 decreases, inducing local changes in that alter the local A/ and further influencing PrO2. As FIO2 decreases from 0.50 to 0.21, little change is seen in regional in the lowest A/ regions. Elevation of resistance and decrease in regional begin to occur within the low- to mid-A/ regions (blues in Fig. 1). The decrease in to these A/ regions results in a shift of toward the high A/ regions. This serves to decrease A/ in the high-A/ regions, while increasing A/ in the low-A/ regions, thus making the overall A/ more homogeneous. As FIO2 is lowered further to 0.12, continues to shift, along with increasing Ppa. At this point, the high-A/ regions reach a PO2 low enough to cause some vasoconstriction in these previously nonconstricted regions. This causes a shift in from the high-A/ regions back toward the low-A/ regions due to higher Ppa and possible vasodilation at extremely low PO2. As the FIO2 progressively decreases from 0.21 to 0.12, the heterogeneity (as defined by the SD) of A/ increases (see below). The relative heterogeneity of WNR and WNRA averaged over all five animals is shown in Table 3. There was a slight but significant decrease in heterogeneity of WNR, which returned to the normal degree of heterogeneity when FIO2 = 0.50 was repeated at the end of the experiment. Because becomes more uniform with hypoxia, yet A remains heterogeneous, the initial good matching of A/ is lost.
|
|
Individual Pig HPV Response Patterns
The WNR is shown for each piece in one animal in Figure 2. To take advantage of all the data available, rather than just grouping lung pieces by their A/ at FIO2 of 0.21, we clustered lung pieces by their HPV response using the residuals of their WNR response pattern to FIO2. The residual for a piece at a specified FIO2 setting is the difference between the R at that setting minus the piece mean of R for all FIO2 settings. WNRA residuals are defined in a similar way. There are five main clusters with pieces responding similarly in this one animal. Pieces in cluster 1 respond to hypoxia by increasing R. Pieces in cluster 5 respond to hypoxia by decreasing R in response to hypoxia. Other responses are found in clusters 2, 3, and 4.
The mean changes of R and RA are shown for the five main clusters within this animal (Figure 3). Figure 3A shows the relative changes of vs. FIO2, and Fig. 3B shows the relative changes in A vs. FIO2 within the five clustered regions in response to hypoxia. Cluster 1 represents regions with very low A/. to regions in cluster 1 remains low until FIO2 <0.15, when the increase in Ppa results in an increase in regional . The R for the different clusters varies up to 200% and in different directions with hypoxia. Relative A changes very little, <10%, with hypoxia for each cluster. All five pigs showed similar HPV response cluster patterns with little changes in relative A. The mean A (WNRA) differed significantly among the clusters (P = 0.004, repeated-measures ANOVA), and the pattern of mean WNRA across the FIO2 settings also differed significantly (P = 0.05, cluster-by-FIO2 interaction).
The spatial distribution of clusters represented in Figure 3 is shown in Figure 4. The lungs for this animal are shown in the supine posture from the left side (slightly skewed to aid visualization) and the caudal-cranial view from the ventral surface. It is obvious from Fig. 4 that there is a spatial correlation between pieces with similar response patterns. In this animal, the cluster with the lowest A (cluster 1, blue) is located primarily in the dorsal-caudal regions, whereas the cluster with the highest A (cluster 4, orange) is located in the ventral regions. The number of clusters analyzed per individual animal ranged from five to nine. Of the 37 clusters obtained from the individual analysis of all animals, 90% where found to be significantly more spatially compact than a random sample of pieces from the corresponding lung (null hypothesis test).
The heterogeneity of A/ distribution is shown in Figure 5 for different values of FIO2 in one animal. In this animal, SD ln A/ decreased from 0.83 at FIO2 = 0.50 to 0.50 at FIO2 = 0.21. With decreased FIO2, SD ln A/ increased to 0.68, 0.71, and 0.93 for FIO2 = 0.15, 0.12, and 0.09, respectively. This animal is closest to representing the mean response of all five animals. In the normal lung, some HPV occurs during room-air breathing. Increased levels of hypoxia worsen A/ distribution because of a significant redistribution of pulmonary with little redistribution of A.
HPV Response Patterns Across Pigs
The similarity in response to hypoxia across all pigs is examined by using metaclustering, which assigned all pieces of all animals to one of five main response patterns (see METHODS). In Figure 6, WNR (A) and WN resistance (B) vs. PrO2 for each cluster are shown. For lung pieces that had no , a resistance value of 20 (
100 times normal piece resistance) is assigned. The pieces are grouped with other pieces having a similar profile of across the FIO2 settings (see Metaclustering above). Table 4 shows that each cluster is represented across all animals. Clusters 2, 4, and 5 represent a response qualitatively similar to the whole isolated lung (36). The resistance in the isolated (whole) lung is a sum of the resistance of all of the individual pieces and, therefore, is not identical to any individual piece. However, the majority of pieces demonstrate the same increase in resistance at low PrO2 (<70 Torr) and a reduction in resistance at very low PrO2 (<40 Torr). Cluster 3 (orange) represents pieces with the highest A/ (as indicated by the initial high PrO2 and from Fig 5). This cluster is just reaching its maximum resistance at an FIO2 of 0.09 and a PrO2 near 40 Torr. Cluster 1 (blue) includes pieces with the lowest A/ and with a very high initial resistance, which dramatically decreases as PrO2 falls <60 Torr. The increase in and reduction in resistance seen at extremely low PrO2 may be due to the increase in Ppa, but there may also be a component of active or passive vasodilation in those regions at extreme levels of hypoxia.
|
Spatial Metacluster
Figure 7 shows the color-coded spatial locations of regions (see Fig. 6) within each of the five animals. Figure 7A is the same animal as in Fig. 4, shown in the same views but color-coded to reflect the metaclustering. The other four animals are shown in the caudal-cranial view from the ventral surface only. Pieces in clusters 1 and 2 (blue and green) are located primarily in the dorsal-caudal regions in four of the animals, representing regions with the lowest A/. Pieces in cluster 4 (red) form a small cluster located primarily in the dorsal-caudal region in all animals. Pieces in clusters 3 and 5 (orange and purple) represent over one-half (2,536/4,750 = 0.53) of pieces in the five animals and are spatially located throughout all but the most dorsal-caudal regions in all five animals. Whereas each of the animals is unique, there is a general similarity in the spatial distribution of pieces with similar HPV response. A more statistical exploration of this distribution follows.
All clusters obtained for this metacluster analysis were found to have centroids that differed significantly from that of the entire lung and to be significantly more spatially compact than the metalung. The location of the centroids and the mean distances of pieces to centroids are shown in Table 5. Table 5 shows that the five clusters differ significantly from the entire lung in their centroid locations in the y (dorsal-ventral) and z (caudal-cranial) dimensions and considerably less so in the more symmetric x (right-left) dimension. All of the clusters are smaller in size than the lung, as reflected in the smaller mean distances to the centroid (all statistically significant). The differences in location of centroids, and the sizes, although significantly different than the entire lung, are only modestly different in magnitude. The clusters are fairly extensive and are not concentrated exclusively in a single region of the metalung. However, as shown by the three-dimensional display in Fig. 7, each cluster is strikingly more prominent in certain localized regions. The volume of the region spanned by each cluster can be compared among the clusters and with the whole lung. If the volume of the lung is taken as an index value of 1.0, then, relative to this volume, the volume of clusters 1-5 is, respectively, 0.84, 0.81, 0.76, 0.33, and 0.73. (The value noted is the square root of the product of the first three eigen-values from singular value decomposition of the x, y, and z locations of all pieces in the cluster, divided by the corresponding value for the entire lung. For a three-dimensional multivariate normal distribution, this relative value would correspond to the volume inside an ellipsoid containing a specified proportion of the distribution.) Thus, whereas the displays show greater concentration of the clusters in specific lung regions, each cluster spans a large proportion of the volume of the entire lung (about four-fifths), except cluster 4, which spans only about one-third of the lung.
|
To determine whether the response is dependent on A/, we evaluated whether each metacluster was different from each other metacluster in their mean PrO2 at FIO2 = 0.21. The test for differences among the five clusters in mean PrO2 was highly significant (P < 0.0001), indicating that pieces that cluster with similar patterns of across FIO2 settings also cluster nonrandomly on their PrO2, when FIO2 is the same as room air. Furthermore, the 10 pairwise comparisons of mean PrO2 for the metaclusters were also statistically very significant (P < 0.001), except for the comparison of clusters 4 and 5 (uncorrected P = 0.8, Bonferroni-corrected P = 1.0). R2 equals 0.27 for cluster designation as a predictor of PrO2. The mean values of PrO2 for the five metaclusters are listed in Table 6, with similar means only for clusters 4 and 5.
|
When clusters of pieces were created for the animals individually (using each animal's data separately), nonrandom clustering of PrO2 was also evident. There were from five to nine clusters for individual animals, and, for every animal, the null hypothesis of only random distribution of PrO2 values across clusters was rejected (P < 0.0001), using the same procedures as for the metaclusters. Furthermore, the results of Bonferroni-corrected tests between pairs of clusters within each animal yielded only 22 out of 150 hypothesis tests, where two clusters compared did not have a significantly different mean PrO2 (P < 0.05).
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
A/. Methods Evaluation
The methods used in this study have been evaluated before (12, 22, 23, 25, 33). Microspheres with a 15-µm diameter are almost completely entrapped in the small pulmonary arterioles (32) and adequately reflect the local pulmonary (4, 22). Aerosolized FMS were used to measure regional A. This method has been validated by Robertson et al. (33) and Melsom et al. (23).
With the present state of the art, it is not possible to directly measure PrO2 within individual pieces. The best that can be done is to calculate the PrO2 based on the and A to a region based on the FMS measurements. This approach was developed by Altemeier et al. (1) based on the mathematical model of Olszowka and Farhi (27). A test of the reliability of calculated PaO2 values with experimentally measured PaO2 is shown in Figure 8 for each animal and each FIO2. The calculated and measured PO2 match very closely for all FIO2 levels, except FIO2 = 0.50. The difference at high PO2 may be due to the presence of A/ heterogeneity within each piece that is ignored in our analysis (1) and/or errors in measurement of PO2 at high PO2 levels (15).
Other Evidence for Anatomic Distribution of Hypoxic Responsiveness
Earlier studies have identified a spatial variation in the response to hypoxia by observing a shift in from lower lung regions toward upper lung regions in sheep (26), isolated rabbit lungs (29), exercising rats (19), and human subjects (3, 14). Each of the above authors concluded that the shift in toward upper regions of the lungs with hypoxia was consistent with the presence of a heterogeneous HPV response. Whereas the above studies used methods with relatively poor resolution (compared with those used in our study), all are consistent with our findings. Pelletier et al. (31) found regional differences in endothelium-mediated relaxation caused by differences in the magnitude of the endothelial release of nitric oxide (NO) in horses, which points to the possibility of heterogeneity in vascular responses.
Dependence of HPV on A/
In the pig, we found that HPV varies among spatial regions in a manner dependent on A/. Lower A/ regions tend to vasoconstrict at higher FIO2 than higher A/ regions. At lower FIO2 values, vasoconstriction in the higher A/ regions tends to force to the lower A/ regions, which may also vasodilate at their extremely low PrO2.
Effect of Hypoxia on Overall A/ Heterogeneity
With FIO2 = 0.21, A and are highly correlated with r = 0.89 (Table 2). Progressive reduction in FIO2 shifts away from the more hypoxic regions due to HPV. Because there is little change in regional A, the correlation between A and decreases to r = 0.87, 0.79, and 0.65 for FIO2 = 0.15, 0.12, and 0.09, respectively. The worsening correlation is further reflected by an increase in measures of A/ heterogeneity as hypoxia progresses. The decrease in correlation between A and as FIO2 decreases is primarily due to the deteriorating correlation of at lower FIO2 levels with its piece-specific at FIO2 = 0.21. The mean correlation of at FIO2 = 0.21 with at lower FIO2 = 0.15, 0.12, and 0.09 is r = 0.93, 0.85, and 0.77, respectively.
For the same set of correlations of A at FIO2 = 0.21 with A at lower FIO2 = 0.15, 0.12, and 0.09, r = 0.96, 0.97, and 0.96, respectively, there was an insignificant decrease in correlation. The greater decrease in correlation of than of A as FIO2 decreases is statistically significant (P = 0.009, t-test). The deterioration of A/ matching with hypoxia is, therefore, primarily due to changes in distribution.
Similarities Across Animals
Sylvester et al. (36) and Brimioulle et al. (5) showed a reduction in HPV response at very low PO2 values (PO2 < 50 Torr). Sylvester speculated that the decrease of resistance at very low PO2 may be due to release of a vasodilator, such as NO, or the loss of ATP for active smooth muscle contraction. Our data in supine intact pigs are similar in pattern, yet different in magnitude from the data of Sylvester et al. in isolated pig lungs. With the isolated lung preparation, PO2 is controlled and identical in all regions. The measurement of total vascular resistance does not allow identification of the response of individual regions. We measured vascular resistance within regions. The overall response of the whole lung is a -weighted sum of all of the regions. On average, our intact pigs increased their mean Ppa from 25 to 46 cmH2O as FIO2 decreased from 0.21 to 0.09 (see Table 1). In this intact preparation, we were not able to decrease the FIO2 to dramatically low values. However, we did see a reduction in resistance in some regions at very low PrO2 values (see clusters 1, 2, and 4 in Figure 6B).
Potential Mechanisms and Significance of Heterogeneity of HPV
This study is the first to evaluate HPV within small regions (2.0 cm3) and to show that variation in HPV is distributed in an anatomically defined pattern. The reasons for this spatial heterogeneity are not clear. However, the present study shows that the spatial heterogeneity is, in part, due to variation in A/. Variation in strength of HPV among species has been described. HPV is strongest in cows, pigs, and young humans, compared with dogs and sheep (18, 30). Medial thickness of the small pulmonary arteries is highly correlated with the development of pulmonary hypertension and right ventricular hypertrophy in hypoxic animals (37). It is known that endothelial cell function is heterogeneous among cells (24, 35). However, the study of endothelial function is just beginning.
A combination of mechanisms contributes to HPV. Hypoxia affects voltage-dependent potassium or other K+ channels in membranes of pulmonary arterial myocytes, resulting in depolarization and Ca2+ influx through voltage-dependent Ca2+ channels, leading to myocyte contraction (38). Hypoxia also stimulates production and release of endothelin (ET)-1 from the endothelium (20, 28). ET is a potent vasoconstrictor and acts by stimulating ETA and ETB receptors on smooth muscle and endothelium. Another modulator counteracting the vasoconstrictor role of ET is NO, the principal modulator of endothelium-dependent vasodilation in the pulmonary circulation (9). It is possible that any or all of these factors may be distributed in a heterogeneous manner.
Whereas the importance of this heterogeneous HPV is not yet known, it is possible that heterogeneous HPV may be a partial explanation for the heterogeneous pattern of pulmonary edema observed in high-altitude pulmonary edema (16). Regions with stronger HPV can divert toward regions with weaker HPV. The resulting overperfusion of the weak HPV regions may lead to elevated vascular pressures and capillary fluid leak.
|
FOOTNOTES |
|---|
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  A. T. Lovering, L. M. Romer, H. C. Haverkamp, D. F. Pegelow, J. S. Hokanson, and M. W. Eldridge Intrapulmonary shunting and pulmonary gas exchange during normoxic and hypoxic exercise in healthy humans J Appl Physiol, May 1, 2008; 104(5): 1418 - 1425. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
