INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

REMOVAL OF EDEMA LIQUID from the air space of the lung is a life-threatening problem that affects thousands of hospital patients annually. It is generally agreed that the difficulty in removing edema fluid is due to the relative impermeability of the air space epithelium to the plasma macromolecules that enter into the air space during the alveolar flooding stage of edema formation (6). These macromolecules exert an osmotic pressure that inhibits the removal of the edema fluid. This conclusion is based mainly on studies of rates of edema solute removal from the lungs of experimental animals. For example, Berthiaume and colleagues (3) measured albumin removal rates of 0.9-1.4%/h from the lungs of unanesthetized sheep that had had 100 ml of autologous serum instilled into their lungs.

Our understanding of the resolution of edema could be improved by more quantitative information about the permeability characteristics of the alveolar epithelium to macromolecules. To address this, Matsukawa and colleagues (13) measured rates of dextran flux across cultured lung epithelial cell monolayers. Their results suggested that the monolayers were perforated by pores with radii of 56 Å. Such pores would be relatively impermeable to albumin molecules, which have a radius of 35 Å and which are the main macromolecular component of edema liquid. The data of Matsukawa et al. additionally suggested that the epithelial cells also transferred dextran across the monolayers by endocytosis, a process that would allow the transport of molecules the size of albumin and larger (7).

Our approach was similar to that of Matsukawa and colleagues (13), except that we used hetastarch (Het) as a tracer instead of dextran. Furthermore, we measured Het flux across the alveolar epithelium of intact lungs that were partially inflated with Het, rather than across cell monolayers. The advantage of using Het is that it consists of a continuous spectrum of molecular masses that range from 10³ to 10⁷ Da. Inflating the lung with Het exposed the air space epithelium to the entire Het molecular spectrum. To evaluate alveolar epithelial permeability, we measured the masses of the Het molecules that entered the lung perfusate from the air space.

The Het molecules detected in the perfusate had a unique, bimodal mass distribution. We found a narrow peak, consisting of molecules with masses ranging from 5 to 46 kDa and a lower but much broader band consisting of molecules with masses up to 2,000 kDa. Our hypothesis was that each of these components represented a different route of molecular transfer across the epithelium. We hypothesized that the narrow peak represented Het flux through small pores in the epithelium, whereas the broad band represented Het flux through a large-diameter pore. We applied pore-modeling techniques to both the narrow peak and the broad band to determine the porosity and conductivity of the alveolar epithelium on the basis of a two-pore population. Our results suggest that macromolecules are cleared from the alveolar air space by diffusion through pores of two different radii.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

We anesthetized retired male breeder rats (442 ± 25 g; n = 9) with intraperitoneal ketamine (40 mg/kg), xylazine (6 mg/kg), and acepromazine (1 mg/kg), and tied them supine. We infused heparin (750 U/kg) through a cannula (PE-190) that was tied into a femoral vein. After 10 min, to allow the heparin to circulate, we cut the femoral artery and allowed the animals to die from exsanguination. We infused Ringer lactate into the venous cannula during exsanguination (1 ml/ml of shed blood; total 20 ml) to facilitate the removal of residual red blood cells from the pulmonary circulation. Once breathing ceased, we tied a cannula into the trachea, set the tracheal pressure to 5 cmH₂O with air, and opened the chest using a sternum-splitting incision. After tying cannulas (PE-190) into the pulmonary artery and left atrium, we removed the lungs from the chest and placed them into a styrofoam perfusion chamber.

The lungs were perfused with albumin in phosphate-buffered saline (see concentration below) and ventilated with air using a piston pump (25 breaths/min). Inflation and deflation pressures were set to 15 and 5 cmH₂O, respectively. We used a recirculating perfusion system. The perfusate that dripped from the left atrial cannula was pumped to a perfusion chamber set 10 cm above the base of the lung. These airway and vascular pressures were chosen to obtain an air space-vascular pressure gradient near zero in a ventilated lung. Thus the transport of Het from the air space to the vasculature was mainly due to concentration gradients. Perfusate flow was measured by using a drop counter placed beneath the left atrial cannula, and vascular and inflation pressures were measured via transducers (Statham) connected to an oscillograph (Grass). After 10 min, to allow the pulmonary circulation to clear of any remaining red blood cells, we stopped the perfusion and introduced 10 ml of Het (6%; oncotic pressure, 40.0 ± 0.6 mmHg) into the tracheal cannula, using a reservoir placed 10 cm above the lung.

We prepared two sets of lungs. In the first (n = 5), the perfusion solution contained 10% albumin in phosphate-buffered saline (oncotic pressure, 57.3 ± 1.7 mmHg) (oncotic gradient). The second (n = 4) contained 7.2% albumin (measured oncotic pressure, 39.7 ± 1.0 mmHg), to match the oncotic pressure of Het filling the air space (oncotically balanced). Oncotic pressures were measured by use of an oncometer (Wescor) equipped with a 10-kDa membrane (Fig. 1). Before reestablishing perfusion, we filled the perfusion system with fresh solution using a minimal volume (27 ml) to minimize dilution of Het entering the perfusate from the air space. Perfusion and ventilation with air were reestablished by using the baseline (pre-Het inflation) pressures (above).

View larger version (15K): [in this window] [in a new window]   Fig. 1.   Oncotic pressures (, in mmHg) measured in hetastarch (Het; solid symbols) and albumin (Alb; open symbols) solutions. The regression equation fit to the Het data has a value of  = 0.542 + 1.499 Het + 0.75 Het2 (R2 = 0.98). The regression equation fit to the Alb data has a value of  = 0.683 + 3.27 Alb + 0.27 Alb2 (R2 = 0.99).

Once each hour during perfusion, 2 ml of the perfusion solution were withdrawn for Het analysis and replaced with 2 ml of fresh solution. After the last perfusate sample was collected (hour 6), the lung was homogenized and a sample of the lung supernatant was collected after centrifugation.

Het concentrations and molecular distributions in the homogenate and in the hourly perfusate samples were measured by use of high-performance size-exclusion chromatography (HPSEC). Details of our HPSEC system have been published elsewhere (10). Briefly, the perfusate Het samples were pumped through three columns, arranged in series, each of which was packed with inert particles of specific size. The rate at which Het polymers flowed through the spaces between the particles was inversely proportional to the polymer molecular weight. Polymer emergence from the columns was measured by use of a refractance detector, the output of which was proportional to polymer concentration. Het consists of polymers ranging from 10³ to 10⁷ kDa, and refractance detector output for a Het stock solution produces a peak in which retention time is shown on the horizontal axis and concentration on the vertical axis. The retention times were converted to molecular mass on the basis of the results of a calibration plot consisting of the retention times of dextran samples of known molecular mass (2,000, 580, 71, and 11 kDa). The peak produced by a 6% Het sample consists of ~600 individual retention times (1 s apart), each of which represents a unique Het polymer weight. The area enclosed by such a peak is proportional to the Het sample concentration. Peak areas were converted to concentration by using a calibration plot composed of peak areas produced by Het samples of known concentration. Before lung perfusate samples were introduced into the HPSEC, trichloroacetic acid was added to each sample to precipitate all proteins and prevent protein contamination of the size-exclusion columns. After centrifugation to remove the precipitate, the pH of the supernatant was restored to neutral by using 5 N KOH before analysis of the supernatant by HPSEC (10).

Electron Microscopy

Statistics

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Baseline perfusate flows before Het inflation averaged 2.4 ± 0.9 ml/min in both groups. One hour after Het inflation, flows were 2.2 ± 1.1 ml/min in the oncotically balanced group and 2.9 ± 0.9 in the oncotic gradient group (not significant). After 6 h, flows in the two groups were, respectively, 1.7 ± 0.5 and 2.2 ± 0.9 ml/min (P = 0.03).

Examples of the molecular distribution in Het tracheal instillation solution and lung homogenate are shown in Fig. 2. Molecular masses in the instillation solutions for both groups ranged from 4.2 ± 2.9 to 14,095 ± 5,410 kDa with a peak at 228 ± 38 kDa that had an average height of 13,224 ± 230 (HPSEC refractance detector units) (Fig. 2). The lung homogenates, which did not differ significantly between the two groups, ranged from 5.0 ± 3.2 to 10,408 ± 2,120 kDa, with a peak at 304 ± 46 kDa that had an average height of 6,811 ± 3,301. 

View larger version (14K): [in this window] [in a new window]   Fig. 2.   Het molecular distributions in a 6% Het lung instillate solution and in lung homogenate after perfusion for 6 h (oncotic gradient study). The height of the homogenate peak is lower than that of the instillate because of dilution of the instillate concentration by the lung tissue. That is, the Het concentration in a mixture of Het + lung is lower than in the Het solution alone. There is some loss of the smallest Het fractions in the homogenate (<104 Da), perhaps due to tissue amylase activity.

An example of HPSEC analysis from a lung inflated with 6% Het and perfused with 10% albumin (oncotic gradient) is shown in Fig. 3. The most striking finding is the bimodal shape of the Het molecular distribution in the hourly perfused samples. Each sample is composed of a narrow, low-molecular-weight peak followed by a broader but lower high-molecular-weight band. This profile is distinctly different from that of the Het infusate or the Het remaining within the lung at the end of the perfusion period (homogenate), both of which have single peaks. These were the source of the Het entering the perfusate. Note too that the heights of the peaks in the perfusate samples increased with time, as a result of increasing Het concentration within the perfusate. The low peaks ranged from 6.3 ± 3.3 to 46.0 ± 6.9 kDa, with a maximum at 16.0 ± 1.5 kDa that had an average height of 448 ± 395 (Table 1). The high-molecular-weight band had a maximum mass of 2,120 ± 522 kDa with an average height of 63 ± 4.1 at 288 ± 85 kDa.

View larger version (33K): [in this window] [in a new window]   Fig. 3.   High-performance size-exclusion chromatography (HPSEC) Het peaks from a lung inflated with 6% Het ( = 35.7 mmHg) and perfused with 10% Alb ( = 59.6 mmHg) for 6 h. Het measured in each of the hourly perfusate samples had a bimodal distribution, consisting of a low-molecular-weight peak and a high-molecular-weight band. These shapes are distinctly different from those of the infusate and the homogenate (peak tops not shown), suggesting possible mechanisms by which Het crossed the alveolar epithelium to enter the perfusate (see DISCUSSION). Values used to characterize the perfusate Het molecular distributions are noted (Table 1). The vertical axis is expressed in units of HPSEC detector output, which is proportional to concentration. m.w., Molecular weight; max., maximum; min., minimum; Homog., homogenate.

HPSEC analysis from a lung perfused with 7.2% albumin (oncotically balanced) is shown in Fig. 4. The molecular weight ranges of the low-molecular-weight peaks and high-molecular-weight bands in these lungs were not significantly different from those of the 10% albumin group, with one exception. The low-molecular-weight peak occurred at 10.4 ± 1.0 kDa, which was significantly less than that for the oncotic gradient group (16.0 ± 1.5 kDa; P = 0.0001) (Table 1). The heights of the peaks for the 7.2% albumin group were also significantly lower than those of 10% albumin-perfused lungs (Table 1). The maximum height of the low-molecular-weight peak was only 15% of that in the oncotic gradient group, whereas the maximum height of the high-molecular-weight band was 38% of the gradient group (Table 1).

View larger version (25K): [in this window] [in a new window]   Fig. 4.   HPSEC peaks from a lung inflated with 6% Het ( = 35.7 mmHg) and perfused with an Alb solution of equal osmotic pressure (7.2%). The bimodal Het mass perfusate distributions seen in Fig. 3 are evident, although the peak heights are only ~[1/5] as great.

To estimate the fraction of the air space Het that entered the perfusate, we compared areas enclosed by the highest (hour 6) perfusate sample with those enclosed by the homogenate. These areas are proportional to the total amount of Het found in the perfusate and air space, respectively, after 6 h of perfusion. We summed these two areas, on the basis of the assumption that this represented the total amount of Het in the air space at the start of perfusion. We then divided the area enclosed by the hour 6 perfusate sample by this sum to obtain the percentage of Het initially in the air space that entered the perfusate after 6 h. The results (Table 2) show that 0.53 ± 0.57% of the air space Het entered the perfusate in lungs in which there was no oncotic gradient. The low-molecular-weight peak and high-molecular-weight band each accounted for about half of this total (Table 2). When the perfusate oncotic pressure exceeded that in air space by 17 mmHg, the percentage of air space Het that appeared in the perfusate after 6 h increased to 3.9 ± 2.9% (P = 0.04). The low-molecular-weight peak accounted for about three-fourths of the total.

Modeling the Small Pore System

View larger version (16K): [in this window] [in a new window]   Fig. 5.   Change in peak height vs. time for specific Het fractions in perfusate samples of oncotically balanced studies (means ± SE). The equation for the 15-kDa regression line is peak height = 23.3 + 9.0 h , n = 4, R2 = 0.88. The equation for the 20-kDa regression line is peak height = 19.2 + 7.8 h, n = 4, R2 = 0.86.

View larger version (12K): [in this window] [in a new window]   Fig. 6.   Diffusive Het mass flow () per unit of Het concentration difference (C) for specific Het fractions. The squares represent measured values (means ± SE). The curve, which represents a pore with a radius of 5 nm and a total diffusive area-to-diffusion distance ratio (Ad/L) of 71, provides a good fit to the data.

We attributed the monotonic decrease in /C with increasing molecular mass (M) to the diffusion of Het molecules of different size through cylindrical pores of constant radius (r). The diffusion of a spherical solute molecule through a cylindrical pore is given by (18)

(1)

Here D_free is the free diffusion coefficient of the solute in an unbounded solution;  is the solute distribution function, a measure of the steric exclusion of the solute from a region within one solute radius of the pore wall.  is equal to (1  a/r)² where a is the solute radius. The parameter k is the hydrodynamic drag coefficient of a sphere moving along the pore centerline, obtained as a function of a/r from tables (17). Application of the Starling equation to the excluded region near the wall and the remaining pore region results in the reflection coefficient () of the solute (1)

(2)

where a is the is the solute radius and r is the pore radius.

We obtained the pore radius and pore number required to fit the /C-vs.-M data (Fig. 6) as follows. We used Eq. 1 to compute D vs. M for a single pore using a chosen value of r. We assumed that the solute radius (a) of each molecular weight fraction was proportional to M^1/3, which was scaled to albumin, for which M is 66 kDa and the radius is 3.5 nm. Molecular radii for molecular weight fractions of 15, 20, 25, 30, 35, and 40 kDa were 2.14, 2.35, 2.53, 2.69, 2.83, and 2.96 nm, respectively. We assumed that D_free was proportional to M^1/2 and that D_free for albumin was 6 × 10⁷ cm²/s (21). We chose the value of r so that the computed values of D multiplied by a constant provided a good fit to the /C-vs.-M data. The constant (A_d/L) is given by Fick's equation for steady-state diffusion

(3)

Here A_d is the total diffusive area and L is the diffusion distance from the alveolar space to the capillary lumen. The constant A_d/L that converts D to /C provides an estimate of pore number if L is known. A pore radius of 5 nm with an A_d/L value of 71 provided a good fit to the data (curve, Fig. 6). Using an L value of 5.0 × 10⁵ cm (22) yielded a value for A_d of 3.6 × 10³ cm. This resulted in a total of 4.5 × 10⁹ pores per lung. Assuming 3 × 10⁷ alveoli per rat lung, this equates to 150 pores per alveolus (20) (Table 3).

The reflection coefficients () for each molecular weight fraction, calculated using Eq. 2, are shown in Fig. 7. The estimated  for albumin is 0.83. 

View larger version (14K): [in this window] [in a new window]   Fig. 7.   Reflection coefficients () calculated for specific Het fractions () and for Alb () using Eq. 2. Pore radius is assumed to be 5 nm.

Modeling Small-Pore Membrane Conductivity

(4)

A value for b can be obtained from the solute flux equation (19)

(5)

where s is the solute flux, Cm is the mean solute concentration in the membrane, and C is the concentration difference. The first term on the right-hand side of Eq. 5 is the solute flux due to osmotic flow that arises from the bulk flow due to the osmotic pressure difference in the Starling equation (4). Rewriting Eq. 5

(6)

We assumed that the last term of Eq. 6 was negligibly small relative to b. That is, the diffusive flux was small compared with the bulk flow. This assumption is verified after the calculation for b. With this assumption, b is equal to either the slope magnitude or the intercept of the linear equation fit to the s/Cm-vs.- data.

s/Cm was obtained from the oncotic gradient data (Fig. 3) as follows. First, we calculated the slope of the linear regression fit to the peak heights of each of the six Het fractions (15-40 kDa) over hours 1-6 (Fig. 8). s/Cm for each molecular fraction was equal to the slope divided by one-half of the height of the corresponding peak in the 6% Het sample (Cm), then multiplied by the perfusate volume (27 ml). s/Cm was plotted vs. its reflection coefficient (Fig. 7, Eq. 2), and a regression equation fit to the data (Fig. 9) was calculated: s/Cm = 0.99  + 0.73, n = 6, R² = 0.99, P < 0.0001. The slope magnitude (0.99 ml/h) is equal to b, the bulk flow (Eq. 6).

View larger version (17K): [in this window] [in a new window]   Fig. 8.   Change in peak height vs. time for specific low-molecular-weight Het fractions in perfusate samples of the oncotic gradient studies (means ± SE). The equation for the 15-kDa regression line is y = 108.9 + 23.6 h, n = 6, R2 = 0.89. The equation for the 40-kDa regression line is y = 1.22 + 8.06 h, n = 6, R2 = 0.83.

View larger version (16K): [in this window] [in a new window]   Fig. 9.   Ratio of s to mean solute concentration (Cm; s/Cm) for 6 low-molecular-weight Het fractions (15-40 kDa) (Eq. 6). The slope of the regression line through these data is equal to the bulk solute flow (b). This value was used to calculate the membrane conductivity (K; Eq. 4).

We applied this b value to Eq. 4 to calculate the membrane conductivity (K). We used the mean  value of 0.59 (Fig. 7), of 17 mmHg, and value for A_t of 5 × 10³ cm² on the basis of a mean alveolar radius for the rat of 35 µm (20). The resulting K value was 2.0 × 10⁵ ml · h ¹ · cm² · mmHg¹. This value for membrane conductivity is 10³-fold smaller than the value measured for systemic capillaries (16). Thus the resistance to flow from the air space to the vasculature is associated almost entirely with the epithelium.

Errors in the method. Only the solute flux due to osmotic flow was considered in the calculation of b; the diffusive flux was neglected. This was justified because the diffusive term in Eq. 6, DA_dC/(C_mL), was much smaller than b. The estimated value of the diffusive term, based on a D value of 3.6 × 10⁴ cm²/h (value for 15 kDa fraction), A_d/L of 71 cm, and a C/Cm value of 2, was 0.05 ml/h, which is 5% of the b value of 0.99 ml/h. From Eq. 6, b is the y-intercept of the linear regression fit to the s/Cm-vs.- data (Fig. 9) and was equal to 0.73 ml/h, which is 26% less than the estimate based on the slope and within the error on the basis of the scatter in the data.

Modeling the Large-Pore System

First, we calculated the slope of the linear regression fit to the peak heights of each of five Het fractions (100, 200, 400, 800, and 1,600 kDa) over hours 1-6 (Fig. 10). Molecular radii for these molecular weight fractions were 4.02, 5.06, 6.38, 8.04, and 10.1 nm, respectively. s/Cm for each molecular fraction was equal to the slope divided by one-half of the height of the corresponding peak in the 6% Het sample (Cm), then multiplied by the perfusate volume (27 ml). We used a trial-and-error procedure to fit the s/Cm- data to obtain the osmotic flow (b). We chose a value for the large pore (R_l), and from the a/R_l values for the five molecular weight fractions we computed the values for  (Eq. 2). The value R_l was adjusted until the y-intercept equaled the slope magnitude of the linear regression equation fit to the s/Cm- data. The resulting equation satisfied Eq. 6 for solute flux without the diffusion term: s/Cm = 0.080  0.082 b, R² = 0.98. The equivalent large-pore radius was 17 nm (Table 3). The osmotic flow through the large pores (b_l) was 0.082 ml/h, which is 8% of the total flow through both sets of pores (Fig. 11).

View larger version (18K): [in this window] [in a new window]   Fig. 10.   Change in peak height vs. time for specific high-molecular-weight Het fractions in perfusate samples of the oncotic gradient studies (means ± SE). The equation for the 200-kDa regression line is y = 1.56 + 13.1 h, n = 6, R2 = 0.83. The equation for the 1,600-kDa regression line is y = 1.57 + 2.74 h, n = 6, R2 = 0.64.

View larger version (16K): [in this window] [in a new window]   Fig. 11.   s/Cm for 6 high-molecular-weight Het fractions (100-1,600 kDa) (Eq. 6). The slope of the regression line through these data is equal to large-pore b.

In the absence of diffusive transport of the large-molecular-weight fractions in the oncotic balanced data, the diffusion area and number of the large pores were estimated as follows. The large-to-small pore number ratio (N_l/N_sm) was obtained by using Poiseuille's law for osmotic flow through cylindrical pores

(7)

Here the subscripts l and sm denote large-pore and small-pore variables, respectively. We used the molecular mass fractions of 16 and 288 kDa at the peak signals (oncotic gradient data, Table 1) to represent the transport of low- and high-molecular-weight fractions through the small (5 nm) and large (17 nm) pores. The large and small solute radius of 5.7 nm and 2.2 nm produced a value of _sm/_l of 0.46/0.31 (Eq. 2). With b_l/b_sm of 0.082 and R_sm/R_l of 5/17, N_l/N_sm was 1/1,100. For N_sm of 150 pores per alveoli, N_l was one pore for every seven alveoli (Table 3). The area ratio (A_l/A_sm), given by (N_l/N_sm)(R_l/R_sm)², was 1/95. On the basis of the estimated A_sm/L of 71 cm, A_l/L was 0.75 cm.

From the estimates of R_l and A_l/L, we justified omitting the diffusive term from Eq. 6 to calculate b_l, as follows. The magnitude of the diffusive term D_l(A_l/L)(C/Cm) was calculated by using the molecular mass fraction of 100 kDa. D_l was 5.4 × 10⁴ cm²/h, calculated from values for D_free, , and k (Eq. 1), appropriate for the molecular radius of 4 nm. The diffusive term was 0.0014 ml/h or 2% of the calculated Qb_l.

To determine the contribution of nondiffusive transport mechanisms to the transport of the large-molecular-weight band, we compared the mass flux of the large molecular fractions transported through the large pores by restricted diffusion to that measured in the oncotic balanced data (Table 2). The large-to-small pore mass flux ratio (s_l/s_sm) is related to the diffusion and area ratios as follows

(8)

We used the molecular weight fractions of 10 and 288 kDa at the peak signals to represent the low- and high-molecular-weight bands (oncotic balanced data, Table 1). Large- and small-pore diffusion coefficients D_l and D_sm were 1.6 × 10⁴ and 7.0 × 10⁴ cm²/h, calculated from values for D_free, , and k (Eq. 1) appropriate for Het molecular radii of 5.7 and 1.9 nm, respectively. s_l/s_sm was 1/530. This ratio for diffusive mass flux through the large pores was considerably smaller than that measured at 6 h in the oncotically balanced experiments (Table 2). Thus, in the absence of osmotic flow, the transport of the large molecular fractions must have occurred via a nondiffusive mechanism.

To determine the contribution of this nondiffusive mechanism, the mass flux of the large molecular fractions transported by osmotic flow through the large pores was compared with that from the oncotic gradient data (Table 2). From the solute flux equation (Eq. 6)

(9)

For high- and low-molecular-weight fractions of 288 and 16 kDa at the signal peaks,  was 0.31 and 0.46, respectively. Cm_l/Cm_sm was 2.5 ± 0.56 (Fig. 3). For b_l/b_sm of 0.082, s_l/s_sm was 0.26, which is half of the ratio of the perfusate large molecular mass (0.51 ± 0.37) measured at 6 h in the oncotic gradient experiments (Table 2). Thus half of the large molecular mass was transported by osmotic flow through large pores and half was transported by a nondiffusive mechanism. Of the half of the nondiffusive transport that occurred with the osmotic flow, ~10% would have occurred in the absence of osmotic flow (oncotically balanced data, Table 2), leaving ~40% that was associated with the oncotic gradient. Thus nondiffusive transport might have been increased by the presence of an oncotic gradient (Table 3).

The nature of this nondiffusive transport is unknown, but transport by cytoplasmic vesicles, which are known to be present in alveolar epithelial cells, is one possibility (5, 22, 23). To investigate the potential for vesicular transport, we used electron microscopy to examine alveolar epithelia in the lungs of rats into which colloidal gold had been instilled. We found gold particles to be present in epithelial cell vesicles of both type I and type II cells (Figs. 12 and 13). The number of particles present in each vesicle ranged from four to six to several dozen. We found no gold particles in intercellular junctions between alveolar epithelial cells. These results do not provide direct evidence for Het transport by epithelial cell vesicles, but they do demonstrate the capability of these vesicles to engulf particulates that are of approximately the same size as the largest Het molecules. We will return to this point in the DISCUSSION.

View larger version (164K): [in this window] [in a new window]   Fig. 12.   An alveolar septum in the lung of a rat into which colloidal gold (3-nm radius) had been instilled. Alveolar air spaces (A) are present on both sides of the septum, and the gold infusate forms a meniscus on the upper side. Boxed areas, enlarged at left, show vesicles that contain gold particles (arrows) within type I cells. Vesicles such as these may have been responsible for that portion of the large-molecular-weight Het transport that we could not explain in terms of pores (Table 3).

View larger version (143K): [in this window] [in a new window]   Fig. 13.   Alveolar type II cell from the lung of a rat into which 3-nm-radius colloidal gold solution had been instilled. Gold solution is present in the alveolar air space (A). Clumps of gold particles are also present within vesicles inside of the cell (insets, arrows). Insets are enlargements of the areas enclosed by rectangles. Vesicular uptake such as this might have accounted for some of the large-molecular Het transport (Table 3).

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Results of our calculations are summarized in Table 3. The table shows the radii of the small and large pores, the number of pores of each type, their fluid conductivity, and the percentage of instilled Het that passed through each. Also shown is our estimate of the contribution of nondiffusive mechanisms.

Our results can be compared with those of Matsukawa and colleagues (13), who measured dextran transport across rat alveolar epithelial type II cell monolayers generated in culture. Using fluorescent dextran fractions ranging from 4 to 150 kDa, these authors concluded that the smaller fractions crossed the monolayers via diffusion-limited pores, whereas the larger fractions (>70 kDa) crossed by means of nondiffusional transcellular mechanisms such as pinocytosis. They calculated the pore radius to be 56 Å, which is near to our estimate of 50 Å obtained using Het. On the basis of a pore area-to-length ratio of 0.05 for their monolayers, Matsukawa and colleagues estimated the number of equivalent pores to be 2.5 × 10⁶/cm². From our estimate of the number of pores per alveolus (150) and the surface area of a spherical alveolus with a radius 35 µm, the number of pores in the rat lung should be 1 × 10⁶/cm². This is 2.5-fold less than the monolayer estimate. However, this difference is within the expected margin of error, considering the differences between a type II cell monolayer and the intact lung epithelium, which consists primarily of type I cells. Furthermore, our estimate for epithelial fluid conductivity is near that estimated from in vivo sheep studies (see below).

Our results also support the conclusions of a previous study by Conhaim and colleagues (4), who inflated lungs with solutes of various diameters and measured their concentrations in perivascular interstitial liquid-filled cuffs. They concluded that the air space epithelium could be characterized as if it were perforated by pores with radii of 10, 400, and 4,000 Å, which accounted for, respectively, 68, 30, and 2% of total liquid flux across the epithelium. Our present results, obtained using the continuous molecular distribution of Het, refine this estimate.

The rates of Het disappearance from the air space we measured (Table 2) can be compared with those of Matthay, Berthiaume, and colleagues (3, 15), who measured rates of albumin disappearance from the lungs of unanesthetized sheep into which 100 ml of autologous, radiolabeled serum had been instilled. They measured a protein disappearance rate of 1.6%/h. By comparison, we measured a disappearance rate of 0.53% after 6 h (0.09%/h) in the absence of an oncotic gradient and 3.9% after 6 h (0.65%/h) at an oncotic gradient of 17 mmHg. Oncotic gradients in the sheep experiments were initially absent because the instillate osmotic pressure (29 mmHg) was similar to that in the blood plasma (15). Thus the initial rapid increase in liquid clearance in the first 4 h with an increasing instillate protein concentration was attributed to active transport mechanisms. The progressively reduced liquid clearance from 12 to 24 h was attributed to passive osmotic reabsorption (see below). This is consistent with the results of the oncotic gradient experiments that modeled macromolecular transport as osmotic flow through water-filled pores. The higher protein disappearance rate in the sheep studies might have been due to the larger epithelial surface area in contact with an instillate volume (100 ml) that was larger than that used in the present studies (10 ml).

The sheep studies showed that, by 12 h, liquid clearance progressively slowed as instillate protein became more concentrated with time. These data are used to estimate epithelial fluid conductivity as follows. Specifically, the rate of liquid absorption of the 100 ml of serum albumin instilled into the airways decreased from 3.2 ml/h at 12 h to 1.4 ml/h at 24 h, whereas the instillate protein concentration increased from 9 g/dl to 12 g/dl. At 12 h, the instillate volume was reduced to 30 ml (15). We assume that the reduction in absorption was due to a reduction in osmotic flow and estimated K from the Starling equation (Eq. 4). At 12 h, when b was 3.2 ml/h and instillate protein osmotic pressure (i) was 40 mmHg

(10)

Here c is capillary protein osmotic pressure, and instillate i was computed from Landis and Pappenheimer's (11) relation between protein concentration and osmotic pressure. At 24 h, when b was 1.4 ml/h and i was 60 mmHg

(11)

Assuming no change in P and c, subtraction of Eq. 10 from Eq. 11 results in a value of KA of 0.09. With a value for  (albumin) of 0.83 (Fig. 7), the calculated value for KA_t in the sheep lung was 0.11 ml · h¹ · mmHg¹, similar to that estimated for the rat lung. On the basis of mean alveolar diameters of 70 and 100 µm (20) and alveolar liquid volumes of 10 and 30 ml for the rat and sheep, respectively, the alveolar surface area in contact with alveolar liquid was twofold greater in the sheep than in the rat. Thus K for the sheep was half that for the rat. The effects of lymphatic drainage and a significant vascular-to-airway hydrostatic pressure gradient may also contribute to liquid and protein clearance in vivo.

Our electron microscopy studies showed that vesicles of both type I and type II cells are capable of internalizing colloidal gold particles present within the airspace. The radius of these particles (3 nm) corresponds to a protein molecular mass of 40 kDa. However, the particles were coated with albumin, which means that their effective radius was larger. The presence of several of these particles within each vesicle supports our speculation that Het molecules with masses up to 2,000 kDa (radius 11.5 nm) might have been transported out of the air space by this mechanism.

In a previous study, 5-nm-radius gold particles were not found to be present within either type I or type II cell vesicles in the lungs of rabbits (8). Possible explanations for these discrepant findings may be differences in the quantity of gold instilled, differences in the radius of the particles instilled, or differences between rabbit and rat lungs.

We did not find gold particles to be present in intercellular junctions between epithelial cells, which supports our estimate for the radius of the intercellular pores (5 nm radius). The reflection coefficient of a 40-kDa macromolecule by such a pore is 0.7 (Fig. 7). If one albumin molecule (3.5 nm) were attached to each gold particle (3 nm), its effective radius would be 6.5 nm and impermeable through a 5-nm radius pore.

Electron microscopic studies have shown that vesicular transport in rabbit lungs could be inhibited by using the pharmacological agents monensin and nocodazole (9). However, the authors of those studies also found that the clearance of albumin and IgG from rabbit lungs was not significantly decreased by these agents. They concluded that pathways other than endocytosis were responsible for the air space clearance of albumin and IgG measured. This is consistent with our conclusion that the large-molecular-weight Het fractions entered the perfusate through a combination of large-pore flux plus a nondiffusive mechanism that might have been vesicular transport.

Our results confirm the conclusions of others that the alveolar epithelium can be characterized as if it contains both small, diffusion-limited pores as well as larger transcellular pathways (4, 13). However, the continuous distribution of Het that we recorded allows the molecular weight limits of these pathways to be seen more readily and provides insight into the magnitude of these processes in intact lungs. The slow rate of Het removal from the air space that we measured confirms that the alveolar epithelium represents a significant impediment to the clearance of macromolecules in pulmonary edema. Our studies allowed us to quantitatively separate the transport of macromolecules by diffusion (and osmotic flow) through pores from that due to nondiffusive mechanisms. Vesicular transport may be such a mechanism.

The methods we have developed will facilitate the study of mechanisms that promote transcellular diffusion and vesicular transport and thereby enhance the resolution of this life-threatening condition.