| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
HIGHLIGHTED TOPIC
Physiological Imaging of the Lung
1School of Medicine, Division of Pulmonary and Critical Care Medicine; Departments of 2Physiology and Biophysics and 3Environmental and Occupational Health Sciences, University of Washington, Seattle; 4Mountain-Whisper-Light Statistical Consulting, Seattle, Washington
Submitted 7 June 2006 ; accepted in final form 6 November 2006
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
2-cm3 pieces. Flow on each injection day was determined for each piece. Despite the increase in the hydrostatic gradient in the lung with growth, there was a strong correlation between blood flow to the same lung piece when compared on days 3 and 71 (0.73 ± 0.12). Although a dorsal-ventral gradient of perfusion did not exist on day 3, blood flow increased more in the dorsal region by day 12 and then gradually became more uniform by day 71. Although most of the lung pieces did not show any discernable pattern of blood flow redistribution, there were spatial patterns of blood flow redistribution that were similar across animals. Our findings suggest that local mechanisms, shared across animals, guide regional changes in vascular resistance or vasoregulation during postnatal development. In the pig, these mechanisms act to produce more uniform flow in the normal posture for an ambulating quadruped. The stimuli for these changes have not yet been identified. blood flow distribution; remodeling; vertical gradient
While the numbers of pulmonary arteries and veins leading to terminal respiratory units are completely determined at birth (20, 24), new arteries and veins accompany the postnatal formation of new respiratory units (22). Vascular remodeling also occurs after birth with muscularization regressing in larger arteries and smooth muscle extending to more peripheral arterioles (22). Although these general observations are well recognized, regional differences in vascular growth and the effect on perfusion distributions have not been studied. This study measures the regional distribution of blood flow in neonatal pigs and at different time points as they grow. Pigs were selected as the animal model because of their rapid growth and similarities in vascular development compared with humans. These experiments test two hypotheses: 1) the caudal preference for regional pulmonary blood flow is established at birth and 2) regional pulmonary blood flow redistributes with growth, and stereotypical patterns of redistribution can be identified across different animals. We use spatial clustering methods to identify regional patterns of blood flow redistribution and use these observations to infer regional changes in postnatal vascular growth. A new statistical clustering method called "meta-clustering" is developed and used to characterize patterns of blood flow change across animals.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Following the final microsphere injection, each animal was deeply anesthetized, given intravenous heparin, exsanguinated, and given an overdose of intravenous pentobarbital sodium. A sternotomy was performed, large-bore catheters were placed in the pulmonary artery and left atrium, and the thoracic aorta was tied off. The lungs were perfused with 2% dextran (molecular weight 74,000) in normal saline until clear of blood, removed from the chest, and allowed to dry inflated at an airway pressure of 25 cmH2O.
When 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, 1.9-cm3 cubes. Foam adhering to lung pieces was removed and each lung piece was weighed and assigned a three-dimensional coordinate and lobe designation. The lungs of one animal developed large cavities while drying and could not be used. The lungs of an additional animal were fixed via the trachea with formalin and processed to retain fluorescent properties of injected microspheres in histological sections (16).
Blood flow measurements were determined in the remaining six lungs. The fluorescent signals for each color were determined by extracting the fluorescent dyes from each piece with an organic solvent and then measuring the concentration of fluorescence in each sample (6). A matrix inversion method (23) was used to correct for spillover between colors. A weight-normalized relative blood flow to each piece was calculated for each color of injection. This normalized value was calculated by first dividing the fluorescence of each piece by the weight of the piece to obtain a weight-normalized fluorescent signal. The weight-normalized value for each piece was then divided by the mean weight-normalized value across all pieces. The result is a weight-normalized relative flow that has a mean of 1.0 across all pieces for each color. A piece with blood flow twice the mean would have a value of 2.0. The data set for each animal consisted of an x, y, and z coordinate; weight; and relative flow for each lung piece in each posture.
To minimize observed flow heterogeneity caused by artifact or measurement noise, pieces weighing <50 mg were not included, eliminating uncertainty in flow and in weight. Also 16–22 pieces containing >25% airway by visual inspection were excluded prior to analyses in each of the six animals.
Statistical analysis. Relative weight-normalized flows are used for all analyses and are hereafter referred to as flow or perfusion. Mean values are presented along with the SD. The Pearson correlation coefficient (r), calculated between perfusion at two time points for the collection of pieces within a pig, is used to quantify the relationship between regional perfusion at different time points. Vertical gradients of perfusion are determined from simple linear regression with blood flow as a function of height up the lung. The height up the lung was determined from the air-dried lungs at the end of the study and therefore does not represent the vertical gradient at the time of microsphere injections. Although the slopes could be recalculated to reflect the true gradient as a function of estimated lung height, we chose to simplify the comparisons by keeping vertical gradients relative to a single vertical distance when the lungs were inflated and dried.
Hierarchical cluster analysis can identify lung pieces with common patterns of flow over time. Clustering is carried out separately in each animal using the residuals of flow from each piece's grand mean flow across all times. We previously refined the hierarchical method in this context to produce a small set of relatively homogeneous clusters (10). Once identified in the temporal domain, the clusters of pieces are characterized by their physical three-dimensional centroid in the lung and by their size, summarized as the mean distance of all pieces in the cluster from the cluster centroid. Under the null hypothesis, the pieces of any given cluster constitute a random sample of pieces in the lung, and we can compare the mean distance of a cluster's pieces to its centroid to the corresponding mean for the entire lung, using a simple t-test.
We also use a novel statistical tool we refer to as meta-clustering. The purpose of this approach is to identify stereotypical changes in flow that are common to all animals over time. The pieces from all animals are merged into one data set and the hierarchical clustering method is used to identify clusters that have the same interpretation across all animals. The clustering of pieces from individual animals show whether changes in flow over time occur in different parts of the lung, and the meta-clustering demonstrates whether the pattern of change is common across animals.
In the usual clustering method, pieces from a single animal are grouped into clusters that have a common pattern of change across the five times of perfusion assessment. In the meta-clustering analysis, data for all animals are merged with individual lung pieces represented in rows of the data set and the columns representing the five times of perfusion measurement. The data for each animal occupies from 1,107 to 1,395 rows (pieces). The values used in the clustering are the residuals of weight-normalized flow at each time for each piece. The residuals are calculated as the difference between the relative flow for a given time for a specific piece minus the mean relative flow for that given piece across all five times. The residuals are used because of the interest in underlying changes in flow rather than in the magnitude of flow.
To ensure that a single dominant animal does not influence our findings, the residuals for each pig are slightly adjusted prior to clustering. Each animal's set of residuals (for all of its pieces) is multiplied by a factor so that all pigs have the same mean within-piece variance across time. The methods used in the clustering have been used in previous publications (10). The perfusion clusters are created without reference to the spatial location of pieces within the lung. After the meta-clustering is completed, the focus turns to determining whether the temporal clustering of flow is reflected in spatial clustering within the lung.
To determine whether clusters occur in a common spatial location across pigs, we create a "meta-lung" that pools the individual lungs so that they can be compared. The meta-lung is created by simply rescaling the x, y, and z coordinates (lateral, dorsal-ventral, and caudal-cranial directions, respectively) to a common set of coordinates, so that the location of a piece can be expressed in uniform x, y, and z coordinates for all animals. In the six animals used in this study, the lungs were of a very similar size and shape, with a range of 13–15 cm in the x dimension (from the center of the leftmost piece to the center of the rightmost piece), 12–14 cm in the y dimension, and 20–24 cm in the z dimension. The range of x, y, and z for each animal was linearly adjusted so that each animal had a common minimum and maximum value for the x, y, and z dimensions.
The spatial location of the perfusion clusters are compared with the meta-lung using t-tests for the mean x, y, and z coordinates for each cluster compared with the corresponding mean for the entire meta-lung using the SD of each coordinate in the entire meta-lung. In addition, a rough test for the spatial size of clusters compared with the size of the lung is carried out using the t-test for the mean distance of each piece in a cluster to its cluster centroid. To compare the volume of pieces in a cluster to the volume of the lung, we also calculate principal components of the x, y, and z coordinates of the pieces in a cluster and compare that to the corresponding calculation for the entire lung.
To identify regions of the lung that included a cluster more frequently than other regions, we calculated the odds of finding a piece from a specific cluster at a specific node of a superimposed rectilinear grid with a pseudo-odds ratio (OR) >4.0. The OR for this analysis was calculated as 
, with p being the probability of encountering the cluster at the given node and P being the overall proportion of all pieces in the meta-lung falling into the cluster. (P = n/N, where n is the total number of pieces in a cluster and N is the total number of pieces in the meta-lung.) The choice of OR 
4.0 is arbitrary, but a large OR was chosen so that regions with a relatively high density of a cluster could be identified.
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
|
Despite the significant growth and increase in the hydrostatic gradient in the lung, there was a strong correlation between blood flow to the same lung piece when compared on days 3 and 71. On average, the correlation coefficient between days 3 and 71 was 0.73 ± 0.12 (Fig. 2).
|
|
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Pigs were studied because their postnatal lung development is similar to humans and because they mature rapidly (22). Porcine and human lungs are well developed at birth, yet important postnatal changes occur (11, 24). While the number of conducting airways does not change, airways lengthen and increase in diameter and alveolarization continues for years after birth (14, 24). Following its remarkable transformation from the fetal to postnatal circulation, the pulmonary vascular tree undergoes remolding characterized by thinning of the muscular segments, redistribution of smooth muscle, and extension into the newly formed air spaces (11). Although these general descriptors have been well documented since description by Burri, Weibel, and Thurlbeck in the mid-1970s (3, 24), little research has focused on regional difference in lung development. Difference in development across the lung may provide insights into the mechanisms directing postnatal development.
The vertical gradients of blood flow changed substantially over time. Because the vertical gradients are determined against lung height in the excised dried lung, changes in slopes are not due simply to the change in the vertical height as the animals grew. On average, vertical gradients of perfusion were negligible in the postnatal period, favored the dorsal direction while growing, and then returned to being more uniform by the last time point. The vertical gradient of perfusion can be thought of as representing a relative balance between vascular conductance and a hydrostatic pressure, An anatomical bias for vascular resistance to be less in dorsal vessels would produce higher blood flows posteriorly with positive slopes. As the animals grow, a larger hydrostatic pressure due to the greater vertical height would favor blood flow to the ventral lung regions. One potential explanation for the observed vertical gradients could be as follows: no significant dorsal conductance bias or hydrostatic gradient in the postnatal period, increased vascular conductance posteriorly with a relatively smaller increase in the hydrostatic pressure, and finally a balance between vascular conductance and the hydrostatic pressure at the last time point.
Some methodological caveats must be addressed. When microspheres are used to measure regional blood flow, the volume at which the lung is fixed must be chosen arbitrarily. We elected to dry our lungs at total lung capacity and present our data as blood flow normalized to piece weight. In lungs dried at total lung capacity, alveoli are of approximately uniform size and the number of alveoli per lung piece is proportional to the piece weight. This is appropriate for blood flow measurements obtained when animals are prone and there is little to no vertical pleural pressure gradient. All injections in this study were performed while the animals were in the prone posture. A prior study comparing radioactive and fluorescent microspheres to measure regional blood flow in a chronic animal model demonstrated that fluorescent microspheres are superior to radioactive microspheres because the fluorescent dyes remain stable over 2 mo (25). Histology from one of our animals did not show any evidence of local inflammation or thrombosis in the microvasculature. Another weaknesses of our study is that we are only able to determine regional blood flow with reference to the other lung pieces at the end of the experiment when the pigs have grown considerably after the first injection. The volume of lung that we observe at the end of the experiment clearly represents a much smaller volume of lung on day 3. We, however, believe that the spatial location of the pieces remain in the same location relative to the other pieces with growth. The animals were suspended in slings so that they would remain quiet during the injections. Pigs have very stiff sternums and thoracic cages. It is possible that the sling may have altered thoracic expansion slightly and this effect may have been variable with age. There may also be concern that injection of 1 million microspheres into a 2.5-kg piglet may have hemodynamic consequences. We previously showed that injection of 1 million microspheres into 250-g rats has little effect on pulmonary hemodynamics (7) and we have unpublished data in 2-wk-old piglets where injection of 7 million microspheres into the pulmonary circulation had no effect on pulmonary or systemic hemodynamics. We therefore believe that the microsphere injections in 3-day-old pigs would have little hemodynamic effects.
We studied blood flow distributions in individual animals at multiple time points to avoid the spatial registration problems that would occur if we had studied groups of animals at different time points. By injecting microspheres at different time points and then harvesting the lungs at the end of the experiment, we are assured that we are examining blood flow to the same lung regions over time. While both the size of any given region and its spatial location change with time, we are confident in our measures of flow to each individual piece. Because we do not have to be concerned about the registration of regions across different animals, we have much greater confidence in our spatial information. We arbitrarily chose to study 2-cm3 piece volumes of lung pieces in mature animals. Although the lung pieces represent much smaller regions at earlier time points, we are confident that we are measuring blood flow to the same lung region over time. We do not need to assume that all regions grow at the same rate and magnitude, only that the microspheres that lodge at earlier time points do not move from one region to another over time. This latter assumption has been validated in a prior study in rabbits over months (25).
A potential disadvantage of serial injections within an animal is that the distribution of microspheres in subsequent injections may be influenced by prior injections. This concern has been studied extensively in acute experiments in smaller animals. In a rat model, 150,000 microspheres of two different colors were serially injected (one color after the other) and the distribution of the two colors was determined with very high spatial resolution (7). The numbers of microspheres of each color were highly correlated (r2 = 1.0) demonstrating that the initial injection of microspheres did not influence the distribution of microspheres in the subsequent injection. The distribution of serial microsphere injections has also been directly visualized in a rat model using intravital microscopy of the lung circulation (15). These studies demonstrated that blood flow at the alveolar capillary level is not significantly altered by lodged microspheres. Although these studies have been performed in small animals, the numbers of microspheres injected per body weight and lung size are of similar scales between the rat and neonatal pig.
Vertical gradients of perfusion are crude measures of blood flow distribution because the mechanisms determining blood flow are assumed to be linear and the effects similar in all regions. Clearly, regional blood flow distributions are not fit perfectly by a linear model. We therefore characterized changes in blood flow during growth by clustering methods to look for spatial patterns of blood flow redistribution that may not be fit well by linear regression. We then applied meta-clustering to find patterns that were shared in similar spatial locations across all of the animals studied.
Our observations could be explained by either changes in vascular geometry (increasing numbers or diameters of vessels) or changes in the relative tone of vessels on a regional scale. There is no evidence in the literature to support regional differences in vessel numbers or diameters. All morphometric studies of the developing vascular tree have used a longitudinal branching order as their reference (4). None have preserved the dorsal-ventral spatial distinction to see if there are systematic regional differences. Rendas and colleagues (22) demonstrated that the diameters of arteries increase, arterial wall thickness decreases, the number of intra-acinar arterioles increase, and the distribution muscularization changes with age from newborn to 20 wk and then is stable until adulthood in pigs. Pulmonary vascular resistance decreased significantly from 19.2 to 2.9 mmHg·l–1·min between the first and fourth month of life. They saw little changes in the vascular tree. Fike and Kaplowitz (5) demonstrated changes in the longitudinal distribution of vascular resistance changes with age.
Changes in regional vascular anatomy or vasoregulation must be driven by some stimulus. Local vascular distending pressures and shear stress have been shown to be important regulators of endothelial proliferation and production of angiogenic factors such as vascular endothelial growth factor (VEGF) and platelet-derived growth factor (PDGF). Mechanisms therefore exist that sense regional blood flow and alter local growth of the vascular tree. Others have shown that shear stress can alter the production of vascular mediators such as prostaglandins (1) and nitric oxide production (19). Johnson and associates (13) showed that an increase in pulmonary blood flow induced by exercise enhances acetylcholine-stimulated vasorelaxation and the adaptation may be due to an increased expression of endothelial nitric oxide synthetase. It is possible that regional differences in vasoregulation may evolve as lungs mature, leading to redistribution of pulmonary blood flow with growth. There are regional differences in lung stretch with each inspiration, and mechanical stretch has been shown to influence regional neonatal lung growth (21). Lung stretch has also been shown to regulate VEGF, which in turn can induce local angiogenesis. Our study suggests that the most fruitful place to look for mechanisms regulating postnatal vascular development are the apical segments and right middle lobe between 2 and 10 wk. We did not follow our animals past 10 wk and cannot comment on how pulmonary blood flow redistributes after this time.
This is the first study to explore the regional changes in blood flow distribution in the growing lung. Our findings suggest that local mechanisms, shared across animals, must guide regional changes in vascular resistance or vasoregulation during postnatal development. In the pig, these mechanisms act to produce more uniform flow in the normal posture for an ambulating quadruped. The stimuli for these changes have yet to be identified. The observations from this work identify time points and spatial locations that would be most fruitful for further investigation.
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
|
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 |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  C. S. Rogers, W. M. Abraham, K. A. Brogden, J. F. Engelhardt, J. T. Fisher, P. B. McCray Jr., G. McLennan, D. K. Meyerholz, E. Namati, L. S. Ostedgaard, et al. The porcine lung as a potential model for cystic fibrosis Am J Physiol Lung Cell Mol Physiol, August 1, 2008; 295(2): L240 - L263. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. Glenny Counterpoint: Gravity is not the major factor determining the distribution of blood flow in the healthy human lung J Appl Physiol, May 1, 2008; 104(5): 1533 - 1535. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
