In vitro Study of Starling’s Hypothesis in a Cultured Monolayer of Bovine Aortic Endothelial CellsPang Z. · Tarbell J.M.
Biomolecular Transport Dynamics Laboratory, Departments of Chemical Engineering and Bioengineering, The Pennsylvania State University, University Park, Pa., USA Corresponding Author
Starling’s hypothesis that fluid movement across the microvascular wall is determined by the transmural differences in hydrostatic and osmotic pressures was tested using an in vitro model comprised of bovine aortic endothelial cells grown on a porous support. In all experiments, a 1% bovine serum albumin (BSA) solution was maintained in the abluminal reservoir and the luminal reservoir contained either a 1 or a 5.5% BSA solution. The global osmotic pressure difference across the endothelial layers was thus either 0 or 20.3 cm H2O. When the luminal concentration of BSA was changed from 1 to 5.5% at a hydrostatic pressure differential of 5, 10 or 20 cm H2O, no reverse flow (in the reabsorption direction) was observed even though the hydrostatic pressure differential was far below the global osmotic pressure differential. In another case, the hydrostatic pressure differential was dropped quickly from 20 to 5 cm H2O, while a constant osmotic pressure differential was maintained by 5.5% BSA in the luminal reservoir. A strong transient reabsorption flow was observed over a 30-second period which diminished to undetectable levels within 2.5 min; then a sustained steady-state filtration flow was observed after 20 min. These in vitro experiments support other studies in capillaries showing transient reabsorption that decays to steady-state filtration at longer times.
Copyright © 2003 S. Karger AG, Basel
Starling’s pioneering hypothesis that fluid movement across microvascular walls is determined by the transmural difference in hydrostatic and oncotic pressures  has become a general principle of cardiovascular and renal physiology . Starling noted that higher protein concentration in the plasma acts as an opposing force to the hydrostatic pressure and that hydrostatic and oncotic pressures reach a balance which determines the flux. The Starling relationship between the transmural fluid flow rate per unit area of capillary wall, Jv/A, and the differentials between capillary and interstitial values of hydrostatic pressure, Pc – Pi, and osmotic pressure, πc – πi, is usually expressed as
Jv/A = LP · [Pc – Pi – σ · (πc – πi)] (1)
where LP, the hydraulic conductivity, and σ, the reflection coefficient, are properties of the capillary wall. An inference of the classical Starling hypothesis is that there is a net filtration in the capillaries on the arterial side and reabsorption (flow against the hydrostatic pressure gradient) on the venous side where the osmotic pressure differential is higher than the hydrostatic pressure differential.
Studies by Intaglietta and Zweifach  challenged the notion that significant reabsorption occurs at the venous end of the microcirculation where the osmotic force exceeds the hydrostatic force. Their investigations indicated that fluid reabsorption by way of the Starling mechanism was demonstrable in short-term experiments (where protein does not attain steady-state distribution in the extravascular space), but not on a long-term, steady-state basis. This was clearly confirmed by Michel and Phillips  who performed experiments on isolated, perfused, frog mesenteric microvessels using the Landis technique and observed transient reabsorption over a period of 15–30 s after the hydrostatic pressure was dropped below the osmotic pressure, as expected from Starling’s law. When the measurements were extended 2–5 min after the pressure drop, however, there was no evidence of reabsorption even though the osmotic pressure was significantly greater than the hydrostatic pressure.
To explain this paradox in the steady state, Michel and Phillips  presented a novel assumption that the pericapillary concentration, Ci, that determines πi in equation 1, was given by a mixing condition such that
Ci = Js/Jv (2)
where Js is the transmural flow rate of the osmotic solute. In general, Ci differs from the global tissue concentration, which in experiments is typically fixed by bathing solutions or external reservoirs. This assumption, when combined with the classical convective coupling equation for solute transport , leads to the following modified Starling law .
Pe = Jv(1 – σ)/Pd (4)
is the Péclet number and Pd is the diffusive permeability of albumin. This theory predicts that when the solute permeability is finite, steady-state reabsorption of fluid from tissue to capillary cannot be maintained by the Starling mechanism alone.
A much more detailed microstructural model of the endothelial transport barrier accounting for the surface glycocalyx layer and the fine structure of the interendothelial tight junction has provided further insights into the Starling mechanism [6, 7]. This model predicts that the Starling forces are determined by the local differences in hydrostatic and osmotic pressures across the surface glycocalyx layer. Additional experiments in frog mesenteric microvessels which support this new view of Starling’s hypothesis have been reported recently .
The controlled experiments which support a modified Starling hypothesis have been limited to frog mesenteric capillaries. Our group, and many others, have studied endothelial transport properties using in vitro models comprised of cultured endothelial cells grown on porous supports [9, 10, 11]. To further test the Starling hypothesis, in the present study, we used an in vitro model which was well established in our laboratory, bovine aortic endothelial cells (BAECs ) on polycarbonate filters, and conducted experiments paralleling those described by Michel and Phillips  and Hu et al. . We also observed transient reabsorption after lowering hydrostatic pressure which reverted to filtration in the steady state.
The following chemicals were obtained from Sigma Chemical (St. Louis, Mo., USA): bovine serum albumin (BSA; 30% solution, fraction V), trypsin, minimal essential media (MEM), glutamic acid, sodium bicarbonate, fetal bovine serum (FBS), fibronectin (cell culture grade from bovine plasma, lyophilized), acetic acid (glacial) and penicillin-streptomycin solution. Polycarbonate filters (Transwell Chamber, 0.4 μm pore size, 24.5 mm diameter) were obtained from Costar (Cambridge, Mass., USA).
Endothelial cells were either harvested from bovine thoracic aortas or purchased from VEC Technologies, Inc. (Rensselaer, N.Y., USA), and cultured in MEM supplemented with 10% FBS as described previously . Cells of passage 8–11 were seeded at a density of 2.5 × 105 cells/cm2 on polycarbonate filters pretreated with gelatin (5 mg/l) and fibronectin (30 μg/ml). All experiments were performed on monolayers 7–9 days following seeding.
A detailed description of the experimental apparatus used to measure volume flux was presented in Sill et al. . Briefly, the entire apparatus was housed in a Plexiglas box and kept at an ambient air temperature of 37°C. The polycarbonate membrane Transwell filter which contained the BAEC monolayer was sealed within a two-piece polycarbonate assembly that separated the luminal (monolayer side) and abluminal compartments. A 5% CO2-95% air gas port provided continual positive pressure outgassing and maintained pH at 7.4. The abluminal chamber was attached by Tygon tubing to a borosilicate glass tube followed by additional Tygon tubing to a reservoir. The difference in level of the fluid in the luminal compartment and in the reservoir provided an adjustable hydrostatic pressure head (ΔP) which acted as a driving force for fluid movement. The experimental media was MEM with different concentrations of BSA and no FBS.
For the measurement of water flux, an air bubble was introduced into the medium in the borosilicate glass tube and bubble movement, indicative of volume flux, was measured using a custom-designed tracking device interfaced with a computer. The tracking unit consisted of a traveling spectrophotometer that recorded the bubble’s displacement versus time, allowing calculation of the volumetric flow rate as follows:
Jv = (Δd/Δt) · F (5)
where Δd/Δt is bubble displacement per unit time and F is a tube calibration factor (i.e. fluid volume contained in a known length of tubing). Because a bubble tracker could follow bubble motion in one direction only, two bubble trackers were used to measure bubble motion in opposite directions simultaneously so that a change in flow direction (reabsorption) could be detected when luminal pressure was suddenly dropped below osmotic pressure.
At the beginning of an experiment, MEM supplemented with 1% BSA (MEM/1% BSA), was added to the luminal and abluminal chambers to minimize the osmotic pressure differential. Only 1.5 ml was added to the luminal side. A hydrostatic pressure differential was not applied at this stage. After a 1-hour equilibration period, a 5, 10 or 20 cm H2O pressure differential was gradually imposed across the monolayer over a 1-min time interval and volumetric flow rate was recorded by the bubble tracker for 2.5 h. The first hour was required to obtain a stable baseline volume of Jv. At 60 min, 1.5 ml of MEM/10% BSA was added to the luminar chamber, providing a mixed luminal concentration of 5.5% BSA to initiate the osmotic pressure difference. The experiments were then conducted for another 1.5 h.
In another set of experiments, after the 1-hour equilibration period, 1.5 ml MEM/10% BSA was added to the luminal chamber, while a 20 cm H2O hydraulic pressure differential was imposed almost simultaneously. But after 60 min, the abluminal reservoir was smoothly raised up 15 cm over a 1-min time interval, reducing the hydrostatic pressure differential from 20 to 5 cm H2O (still a positive pressure drop). The movement of the bubble was monitored for another 1.5 h. The same protocol with 1% BSA in both the luminal and abluminal chambers was used as a control.
In order to drive steady-state transmural flow from the abluminal to the luminal compartment (negative flow or reabsorption), MEM/1% BSA was placed in both the luminal and abluminal compartments, and after a 1-hour equilibration period, a 10 cm H2O hydrostatic pressure differential was applied to the monolayer for 1 h. Then the abluminal reservoir was raised over a period of 1 min to obtain –2, –4, and –10 cm H2O pressure differentials, each level being held for 20 min. Finally, the +10 cm H2O condition was restored and the final value of Lp was compared with the initial value at +10 cm H2O.
Selected monolayers were visualized for damage after exposure to the various experimental treatments for 150 min. A fluorescent indicator, calcein AM (Molecular Probes), was used to check cell viability. Briefly, monolayers were rinsed with PBS, followed by incubation in 0.005 mg/ml calcein AM for 15 min at 37°C. An Olympus IMT-2 inverted microscope was used to visualize the cell monolayer at excitation of 480 nm and emission of 535 nm. Viable cells were stained in green and viewed at a magnification of ×100.
Figure 1 shows the time course of changing the luminal BSA concentration from 1 to 5.5%. Upon application of a hydrostatic pressure gradient of 10 cm H2O at time 0, a characteristic time-dependent decrease in Jv occurred that reached a steady state within an hour. This response, termed the ‘sealing effect’, was previously reported by Sill et al. . The value of Jv at 55 min was chosen as the baseline, and other Jv data were normalized by this value to facilitate comparisons among different monolayers. The abluminal concentration was kept at 1% BSA throughout the experiment. For nearly a 10-min period after the change of luminal concentration at 60 min, transmural flow was not detectable, buth there was no reversal of flow either. Then the bubble tracker began to record positive flow again at about 70 min.
Fig. 1. Effect of osmotic pressure on volumetric flow rate at 10 cm H2O positive pressure differential. The luminal concentration of albumin was changed from 1 to 5.5% at 60 min. The abluminal concentration of albumin was maintained at 1%. The baseline Jv at 55 min was 2.38 ± 0.33 (SEM) × 10–6 cm/s · cm H2O (n = 5).
The osmotic pressure of BSA solutions (π in cm H2O) can be determined from the BSA concentration (C in g/dl), using π = 3.305 C + 0.136 C2 + 0.0084 C3 . If we use the BSA concentrations in the luminal chamber and abluminal compartment for determining the osmotic pressure differential in Starling’s law (equation 1), we estimate that π (5.5%) – π (1%) = 20.3 cm H2O. For a reasonable value of the reflection coefficient, say σ >0.7, the osmotic driving force is expected to be >14.2 cm H2O and reabsorption should occur when the hydrostatic pressure differential is 10 cm H2O. The absence of adsorption in the steady state suggests that the global BSA concentrations do not determine the osmotic pressure differential in Starling’s law.
Similar experiments were conducted at 5 cm H2O and 20 cm H2O and similar patterns were observed. Steady Jv values before and after changing the osmotic pressure at 5, 10, and 20 cm H2O are shown in figure 2. A decrease in steady-state volume flux associated with the increased osmotic pressure in the luminal chamber was clearly observed. But once again, there was no evidence of reabsorption (negative flow) at any time during these experiments even in the case of 5 cm H2O hydrostatic pressure differential.
Fig. 2. The steady-state reduction of volume flux by osmotic pressure at different positive pressure differentials. At 5 cm H2O, volume flux was reduced from 1.98 ± 0.27 × 10–6 cm/s (n = 9) at 1%/1% to 0.46 ± 0.10 × 10–7 cm/s (n = 9) at 5.5%/1%. At 10 cm H2O, volume flux was reduced from 2.96 ± 0.29 × 10–6 cm/s (n = 10) at 1%/1% to 1.35 ± 0.3 × 10–7 cm/s (n = 10) at 5.5%/1%. At 20 cm H2O, volume flux was reduced from 9.9 ± 0.5 × 10–6 cm/s (n = 9) at 1%/1% to 4.9 ± 0.80 × 10–6 cm/s (n = 9) at 5.5%/1%.
Because it is difficult to introduce a sudden change in osmotic pressure across the endothelial layer, as we attempted in the experiments summarized in figures 1 and 2, we introduced a rapid change in hydrostatic pressure in the next series of experiments. With 5.5% BSA in the luminal compartment and 1% BSA in the abluminal compartment, the hydrostatic pressure differential was quickly changed from 20 to 5 cm H2O. In this case, a transient reversal of low was observed (fig. 3). The magnitude of the peak reverse flow was much higher than expected for the overall Starling driving force, but it decayed quickly as indicated in figure 4. There was a period of nearly 20 min after decay of the reversal flow during which transmural flow could not be detected in either direction (fig. 3). After that period, positive flow resumed and approached a steady-state level.
Fig. 3. Transient reversal of volumetric flow rate after a sudden reduction of hydrostatic pressure differential from 20 cm H2O to 5 cm H2O. The luminal and abluminal albumin concentrations (5.5%/1%) were not changed. The transient reversal phase is enlarged in figure 4. The baseline Jv was 3.11 ± 0.07 × 10–6 cm/s · cm H2O (n = 4).
Fig. 4. Transient reversal of volumetric flow rate: enlargement of figure 3.
In the control experiment with 1% BSA on both sides of the endothelial layer, where the global osmotic pressure difference was zero throughout the experiment, the sudden change of hydrostatic pressure from 20 to 5 cm H2O did not cause reabsorption (data not shown), indicating that the transient reabsorption was due to the osmotic pressure differential, not an artifact such as instantaneous volume change due to the deformation of the filter membrane after the pressure change or significant fluid accumulation under the monolayer when the pressure differential was reversed.
The high reverse flow recorded in the experiments of figure 3 suggested that the hydraulic conductivity of the in vitro monolayer might be asymmetric, with higher values for flow in the direction of reabsorption. To test this, we placed 1% BSA in both the luminal and abluminal compartments to minimize the osmotic pressure effect so that the hydraulic conductivity could be determined from the classical Starling law (equation 1) as follows:
In view of the more general modified Starling law (equation 3), equation 6 must be considered an approximation. However, using reasonable values of the parameters σ and Pd (discussed in a subsequent section), the approximation underestimates Lp by only 16% at 10 cm H2O. The Lpvalues calculated using equation 6 are listed in table 1.
Table 1. Asymmetric nature of the endothelial transport barrier
The hydraulic conductivity for reverse flow was much higher (about 100-fold) than for forward flow, indicating a highly asymmetric transport barrier. Figure 5 shows a typical time course of Jv when pressure differential was alternated. When pressure differential was changed from 10 to –2 cm H2O, a high reverse flow was observed that kept increasing during the first 5 min and reached a plateau at 10 min. Return to a positive pressure differential (10 cm H2O) was followed by a ‘sealing’ period similar to the one following the initial exposure to 10 cm H2O pressure differential. Recovery (LpfinalLpinitial) from –2 cm H2O differential pressure was 120% after return to +10 cm H2O differential pressure, suggesting that the cell monolayer was not altered very much by the reverse flow period. However, at higher reverse pressures of –4 and –10 cm H2O, recovery data indicated somewhat greater alteration of the monolayer.
Fig. 5. A typical time course of Jv at alternately positive and negative pressure differentials. Pressure differential was 10 cm H2O for the first hour and a negative pressure differential (–2 cm H2O) was then applied for another 20 min before it was returned to 10 cm H2O.
Selected monolayers stained with calcein AM were visualized under a microscope at a magnification of ×100. There were no apparent areas of endothelial denudation or cell death after experimental treatments including reverse flow experiments.
This study was the first to test the proposal of a modified Starling law [6, 15] in an in vitro endothelial cell monolayer. We observed that the effective osmotic driving force under steady-state conditions was substantially lower than predicted by Starling’s law (equation 1) using global concentrations, and that steady-state reabsorption (negative flow) could not be observed even when the global osmotic gradient was substantially higher than the hydrostatic gradient. These observations are consistent with experiments in the microvessels of the frog mesentery [4, 8] and more general considerations of microvascular transport . The results support the hypothesis that local concentrations, not global concentrations, determine the osmotic pressure differential .
To assess the steady-state data, the model of Michel and Phillips  was used. The parameters Lp, Pd, and σ in equations 3 and 4 were assumed constant (independent of pressure and protein concentration) and were adjusted to minimize the square of the error between the predictions and the data. The resulting values of the parameters are: Lp = 5.1 × 10–7 cm/s/cm H2O; Pd = 1.0 × 10–6 cm/s; σ = 0.71 (dashed lines on figure 2). The Lp and Pd values are consistent with previous measurements of these parameters for BAEC monolayers in vitro [10, 17] and for porcine coronary arterioles in vitro  and the σ value is consistent with many in vivo measurements in capillaries [4, 18]. The predicted Péclet numbers vary from 0.013 for the 5.5%/1% case at 5 cm H2O up to 2.87 for the 1%/1% case at 20 cm H2O. The more sophisticated 3-dimensional model of Hu and Weinbaum , which focuses on the endothelial surface glycocalyx layer as the primary molecular sieve for plasma proteins, was not investigated.
We were unable to observe transient flow reversal when the luminal osmotic pressure was increased by adding a concentrated albumin solution to the luminal compartment. This would appear to be the result of slow equilibration of the osmotic solute with the endothelial layer due to transport limitations associated with the unstirred layer on the luminal surface of the endothelium.
In order to obtain a more rapid change in the overall Starling driving force, we equilibrated the system at 20 cm H2O differential pressure with 5.5% BSA in the luminal compartment and then quickly dropped the pressure differential to 5 cm H2O. We were able to observe a transient reversal which decayed rapidly during the first 30 s after the pressure change and was no longer observable after about 2 min (fig. 4). A zero flow state then persisted for about 20 min before a positive steady-state flow was finally established (fig. 3). The dynamics of this transient reversal are similar to observations reported by Michel and Phillips  and Hu et al.  in frog mesenteric capillaries. They reported transient flow reversals during 5- to 20-second vessel occlusions and steady-state positive flows after 2–5 min of occlusion. The time required to establish steady flow was somewhat longer in our in vitro experiments and may have been associated with the transport barrier provided by the supporting filter and the size of the abluminal chamber (20 cm3).
A surprising finding of the present study was the asymmetry of the endothelial transport barrier in vitro. When hydrostatic pressure was used to reverse the direction of volumetric flux from the abluminal compartment to the luminal compartment, there was about a 100-fold increase in Lp (table 1). The in vitro endothelium, thus, has the characteristics of a check valve: high resistance to flow in one direction and low resistance in the opposite direction. Michel and Phillips  and Hu et al.  did not conduct experiments with the hydrostatic gradient reversed, but their transient osmotic reversal experiments suggested that the hydraulic conductivity was essentially the same in both flow directions, with no evidence of asymmetry. This difference in symmetry of the transport barrier is a distinct difference between the in vitro BAEC model and the frog mesenteric capillary model.
There was a previous report that albumin was transported in an asymmetric fashion across porcine pulmonary artery endothelium in vitro . The transport of albumin from the abluminal to the luminal side was approximately 10-fold greater than the transport from the luminal to the abluminal side. But Siflinger-Birnboim et al.  were unable to confirm this finding in a similar system. Huxley and Curry  reported asymmetry in barrier hydraulic conductivity with alterations in the albumin content of the luminal and abluminal sides of the blood vessel.
The elevated hydraulic conductivity that we observed for volume flux from the abluminal to the luminal compartment may have derived from the nature of the attachment of the endothelium to its basement matrix which must withstand fluid flow forces tending to cause separation when flow is the direction of reabsorption. When volume flux is in the normal filtration direction, fluid flow forces would tend to stabilize this interface. It should also be noted that there was no ‘sealing effect’ either when the flow was in the direction of reabsorption (fig. 5). In fact, the hydraulic conductivity actually increased over the time period of reverse flow. This further supports the notion that the change in direction of the pressure and flow forces is somehow responsible for the asymmetry of the transport barrier in vitro.
In spite of a significant difference in the symmetry of the endothelial transport barrier with respect to volume flux, the in vitro BAEC model and frog mesenteric capillaries display similar transport characteristics with respect to osmotic gradients. The fundamental observation in frog mesentery that an overall osmotic gradient favoring classical Starling reabsorption in fact produces steady-state filtration is captured by the in vitro model. In addition, the dynamics of transient reabsorption are similar for both systems. The in vitro model, therefore, should continue to be useful for studies of the endothelial transport barrier when the hydrostatic pressure is highest on the luminal side of the barrier – the most common physiological situation.
This work was supported by NHLBI/NIH Grant R01 HL57093.
Dr. John M. Tarbell
Steinman Hall, Department of Biomedical Engineering
City University of New York
New York, NY 10031 (USA)
Tel. +1 212 650 6841, Fax +1 212 650 5768, E-Mail email@example.com
Received: December 16, 2002
Accepted after revision: March 6, 2003
Published online: July 29, 2003
Number of Print Pages : 8
Number of Figures : 5, Number of Tables : 1, Number of References : 21
Journal of Vascular Research (Incorporating International Journal of Microcirculation)
Founded 1964 as Angiologica by M. Comèl and L. Laszt (1964–1973) continued as Blood Vessels by J.A. Bevan (1974–1991)
Official Journal of the European Society for Microcirculation
Vol. 40, No. 4, Year 2003 (Cover Date: July-August 2003)
Journal Editor: U. Pohl, Munich
ISSN: 1018–1172 (print), 1423–0135 (Online)
For additional information: http://www.karger.com/jvr