Heart and Circulatory Physiology

Fibroblast alignment under interstitial fluid flow using a novel 3-D tissue culture model

Chee Ping Ng, Melody A. Swartz

This article has a correction. Please see:


Interstitial flow is an important component of the microcirculation and interstitial environment, yet its effects on cell organization and tissue architecture are poorly understood, in part due to the lack of in vitro models. To examine the effects of interstitial flow on cell morphology and matrix remodeling, we developed a tissue culture model that physically supports soft tissue cultures and allows microscopic visualization of cells within the three-dimensional matrix. In addition, pressure-flow relationships can be continuously monitored to evaluate the bulk hydraulic resistance as an indicator of changes in the overall matrix integrity. We observed that cells such as human dermal fibroblasts aligned perpendicular to the direction of interstitial flow. In contrast, fibroblasts in static three-dimensional controls remained randomly oriented, whereas cells subjected to fluid shear as a two-dimensional monolayer regressed. Also, the dynamic measurements of hydraulic conductivity suggest reorganization toward a steady state. These primary findings help establish the importance of interstitial flow on the biology of tissue organization and interstitial fluid balance.

  • cell culture
  • shear stress
  • mechanical stress
  • hydraulic conductivity

physiological interstitial flow is the movement of fluid through the extracellular matrix of a tissue, often between blood vessels and lymphatic capillaries. It provides convection necessary for the transport of large proteins through the interstitial space and constitutes an important component of the microcirculation (Fig. 1). Aside from its role in transport, interstitial flow also provides a specific mechanical environment to cells in the interstitium that could play an important role in determining interstitial organization and architecture, based on abundant evidence that mechanical forces help govern the architecture of tissues such as the lung (16,33), bone (1, 11, 13), articular cartilage (8, 20), and vascular tissues (3, 4, 17, 23,25). Importantly, many cell types including fibroblasts, smooth muscle cells, osteocytes, and chondrocytes reside within a three-dimensional (3-D) environment and are exposed to interstitial fluid forces; this is in contrast to cells such as endothelial and epithelial cells that form a monolayer to create a lumen or surface and may be exposed to shear stresses across the surface.

Fig. 1.

Schematic of a microcirculatory fluid transport in a microvascular bed. Blood flow in the capillaries (BC) brings fluid, which is then transported to cells in the interstitium (I) via interstitial flow (arrows). Fluid is either then returned to the blood microvessel or drained into the lymphatic microvessel (LC).Inset: besides its role in transport, interstitial flow also provides a mechanical environment directly on the interstitial cells via the induction of shear stress (τ) and hydrostatic pressure (P), or indirectly via cell-extracellular matrix (ECM) signaling by imposing a strain (ε) and elastic stress (ς) on the matrix fibers that the cells are attached to via integrin receptors.

Despite its importance, the biological regulation of interstitial fluid balance is poorly understood, largely because of the lack of experimental models. Although many in vivo and in vitro studies have been performed to characterize the mechanics of interstitial fluid balance and estimate the Darcy permeability of a variety of tissues, there have been very few studies to examine how interstitial flow affects cell response and how interstitial fluid is regulated in a soft tissue environment. In their seminal study, Wang and Tarbell (30) investigated the effects of transvascular flow on smooth muscle cells seeded in a collagen gel model and found that the production levels of prostaglandins were 10-fold lower in cells under interstitial fluid shear than in a two-dimensional (2-D) monolayer shear model using a rotating disk, demonstrating that cell response to fluid flow in 2-D configurations poorly mimics the response of fluid flow on interstitial cells in their natural 3-D environment. Furthermore, it has been demonstrated that medium perfusion improves the architecture of engineered tissues such as cardiac muscles (2), downregulates extracellular protein accumulation in engineered cartilage (19), and enhances the viability and functionality of bone cells seeded in collagen constructs (7,21).

In this study, we develop an in vitro flow chamber to directly examine in situ the effects of interstitial flow on cell organization in soft tissue cultures over relatively long periods of time (days to weeks) where we can 1) specifically control the flow environment,2) directly observe cell organization, and 3) measure the mechanical properties (e.g., hydraulic resistance) “online.” In doing so, we observe that fibroblasts align perpendicular to interstitial flow in 3-D culture. These novel findings, along with the model we developed, help to form an experimental basis for understanding the biological regulation of interstitial fluid balance in soft tissues.


Cell culture.

Human dermal fibroblasts (CCD1079sk, ATCC; Manassas, VA) were cultured in α-modified minimum essential medium Eagle (α-MEM; Sigma-Aldrich; St. Louis, MO) with 10% fetal bovine serum (GIBCO; Carlsbad, CA).

Flow model setup and preparation.

The interstitial flow chamber consisted of a cell-suspended collagen gel sandwiched between two glass coverslips, 1.6 mm apart, through which a radial flow of culture medium was directed (Fig.2). The gel was affixed between the flow inlet and outlet by acid-treated porous polyethylene (PE), which anchored the collagen and prevented contraction (14). To enhance binding of the collagen gel to the top and bottom glass surfaces, they were cleaned (26), silanized with 2% 3-aminopropyltriethoxysilane (Pierce; Rockford, IL) for 15 min, and functionalized with 0.1% glutaraldehyde for 30 min before being rinsed thoroughly with deionized water. The components of the flow chamber were assembled under sterile conditions with sterile silicone glue.

Fig. 2.

Model and flow chamber setup. A: materials used for the chamber consist of silastic tubing (a), functionalized glass slide (b), porous polyethylene rod (c), porous polyethylene ring (e), and functionalized glass coverslip (f). Interstitial flow is delivered through the cell-populated collagen matrix (d). B: photo of assembled flow chamber. C: to induce flow, the flow chamber (FC) is connected to a sterile media reservoir (R) via a peristaltic pump (P) and a manometer (M).

After chamber assembly, the chamber was filled with a cell-populated 0.35% collagen solution. The collagen solution was prepared from rat tail tendons according to Pasternak and Miller (22). Human dermal fibroblasts were added to the collagen at a density of 105 cells/ml. The cell-gel solution (of ∼150 μl) was injected into the chamber and incubated at 37°C for 30 min to allow the gel to polymerize and react with activated glass surfaces. The entire setup was then placed in a 35-mm plate and immersed in media overnight for cell attachment in a 37°C, 5% CO2incubator.

To induce flow, the chamber was connected to a sterile media reservoir via a peristaltic pump and a pressure manometer. The flow was delivered at 10 μl/min (leading to an average velocity of 13 μm/s at the inlet and 3.6 μm/s at the outlet, and an average residence time of ∼14.5 min), which induced an average pressure drop of 3.5 cmH2O. A static chamber used as a control was set up exactly as the experimental flow chamber except that it was not connected to the flow delivery apparatus. In both groups, medium surrounded the chamber and could diffuse through the outer and inner PE rings; this medium was replaced every 2 days. All cultures were maintained in a humidified 37°C, 5% CO2 incubator for the duration of the experiment. The manometer readings were monitored every 3 h, and the culture system was examined under light microscopy (Nikon TE 200) daily.

2-D monolayer shear studies.

To compare the effects of interstitial flow with flow across a 2-D monolayer, we used the same flow chamber setup as the interstitial flow studies, but instead of filling the chamber with cell-suspended gel, we seeded the cells at a density of 104 cells/cm2atop a thin coating of 0.35% collagen that was covalently bound to the bottom functionalized surface of the chamber. After 24–48 h of static culture to allow the cells to attach to the surface, at which time the cells reached a confluence of 30–40%, flow was induced over the cell monolayer for 2 days under the same volumetric flow rate as in the 3-D cultures (10 μl/min).

Immunofluorescent staining.

Fluorescein phalloidin (Molecular Probes; Eugene, OR) was used to examine the alignment of the cells when subjected to interstitial flow by staining its actin fibers. The entire chamber was fixed by immersion in 4% paraformaldehyde in PBS for 30 min, rinsed with PBS, and permeabilized in 0.5% Triton for 30 min. The chamber was then immersed overnight in a solution containing 200 nM fluorescein phalloidin. The gels were rinsed in PBS, and images were taken with epifluorescence microscopy (Nikon TE 200 with SPOT RT Slider digital camera) and laser scanning confocal microscopy (Leica LCS Laser Microscope System). For the flow studies on fibroblast monolayers (used for comparison of cell behavior in 3-D gels), the fibroblasts were live-labeled with 5 μM calcein-AM (Molecular Probes) for 10 min before the induction of flow. This calcein-AM live stain was unsuitable for long-term use in the 3-D studies because the dye was not retained in viable cells after several cell divisions (10).

Computation of bulk Darcy permeability.

Continuous online measurements of overall Darcy permeabilityK′, or the inverse hydraulic resistance, were used as a bulk indicator of changes in matrix integrity. K′ (in cm2) was evaluated according to Darcy's law in a one-dimensional radial configuration where K′ = [Qμ ln(r o/r i)]/[2πhΔP(t)], Q is the constant volumetric flow rate delivered by the peristaltic pump, μ is the viscosity of the perfusing fluid (equals 1.2 cp),h is the gel thickness (equals 1.6 mm), ΔP(t) is the pressure gradient driving the flow (i.e., the difference between the inlet pressure as measured from the manometer and the outlet pressure of 1.082 cmH2O equal to the height of medium in the dish), and r o and r iare the outer and inner radii of the gel, respectively (equal 5.6 mm and 1.6 mm). [Note that the hydraulic conductivity K(cm4 · dyn−1 · s−1) is also commonly used in the literature, where K =K′/μ.] The collagen gel dominated the overall hydraulic resistance because the PE had a value of K′ that was two orders of magnitude greater than that of the gel (10−4 vs. ∼10−6 cm2, respectively). Also, the Reynold's number was estimated at 10−5–10−6 for this system, validating the use of Darcy's law in modeling interstitial flow through the gel culture system.

Measurement of cell angle.

Cell orientation in the composite radial section micrographs was quantified for representative flow and static conditions using Adobe Photoshop 6.0 (San Jose, CA) by measuring the angle α of each cell from the horizontal axis or the axis parallel to the direction of radial flow. Because of the very high aspect ratio of these cells in 3-D (typically >10), the major axis of each cell was readily identified and the angle calculated digitally from a manual trace. The cell angles, averaged in increments of 0.4 mm, were then plotted as a function of radial distance. Because the average angle is 90° regardless of the degree of alignment (as reflected in the scatter plots in Fig. 4, fourth row), the variance of the average cell angle was used as an indicator of the randomness of the cell orientation at a particular radial coordinate: the larger the variance, the more random the cell orientation.

Computation of shear stress on cells under flow.

The shear stress on cells subjected to interstitial fluid flow was estimated by using a model by Wang and Tarbell for estimating shear stresses resulting from transvascular flow on smooth muscle cells (29). In this model, stress τ is estimated byτ = [Qμ/(2πrh)](K′) ½(30). τ was calculated at the inner and outer radii (r i and r o) to obtain estimates of the maximum and minimum, respectively, representing the range of shear stresses experienced by the cells undergoing interstitial flow.

For flow on cells plated as 2-D monolayers, the radial profile of the wall shear stress was derived from the velocity profile for steady radial flow between two parallel disks with entrance effects ignored:τw = (3Qμ)/(πra 2), where Q is the volumetric flow rate and a is the distance between the top and bottom disks.


Model features.

Our in vitro interstitial flow apparatus meets the following criteria:1) the culture system is supported against contraction and allows interstitial flow without ECM compaction, 2) the system is easily maintained over several days to allow sufficient time to evaluate cell response, 3) cell morphology and migration can be monitored and visualized under microscopy in situ over a period of days, and 4) the hydraulic resistance of the system can be continuously monitored to determine changes in the most relevant mechanical property (i.e., flow resistance) and indirectly indicate matrix integrity. In addition, the superficial flow velocityv and interstitial pressure P vary across the radial distance of the gel, allowing a range of conditions to be examined and compared in one experiment. All of these model features are demonstrated through the following results.

Cells align perpendicular to interstitial flow.

Flow was induced through fibroblast-populated collagen gels for 2–3 days at a volumetric flow rate of 10 μl/min. Surprisingly, the fibroblasts displayed perpendicular alignment to the radial direction of flow (Fig. 3 A) and had a more noticeable spindle-shaped morphology than those in static conditions. In contrast, the cells in static cultures retained a random orientation (Fig. 3 B) and appeared more branched in shape. In addition, we compared the effect of flow on cells seeded in 3-D cultures with that on cells plated as a monolayer. The cells plated in a monolayer were found to undergo regression under the same flow rate (Fig. 3 C) or shear stress (data not shown).

Fig. 3.

Comparison of fibroblast orientation in three-dimensional (3-D) gels subjected to interstitial flow (A) and in static controls (B) after 2 days with fluid-sheared fibroblasts plated as a subconfluent monolayer on a thin collagen layer after 3 h (C). Arrows under micrographs indicate radial direction of the chamber. Taken at ×100 magnification, the images for the 3-D experiments (A and B) are inverted confocal micrographs of maximum projections of fluorescein phalloidin-stained cells in 150-μm-thick sections, whereas the image for the monolayer shear stress experiment (C) is an inverted epifluorescence micrograph of calcein-AM-labeled cells. Scale bar = 200 μm.

The radial design of our model allowed us to examine the dependence of cell orientation on velocity due to the radial variations in velocity. Figure 4 compares the profiles of cells undergoing interstitial flow with those in static conditions. In the cultures subjected to flow, velocity and pressure decreased radially (Fig. 4, first and second row, respectively), whereas the hydrostatic pressure in the static gel (v = 0 cm/s) was maintained at 1.082 cmH2O. On inspection of cell morphology (Fig. 4, third row), cells subjected to interstitial flow were perpendicularly aligned at higher velocities but became more randomly oriented with increasing radial distance and thus decreasing bulk velocity. This was confirmed by the cell angle distributions (Fig. 4, fourth row) and the increasing variance of the average cell angle with radial distance (Fig. 4, fifth row). The estimated shear stress (30) on the cells in the collagen matrix varied from 1.135 to 0.324 dyn/cm2 (at r i tor o, respectively). Under static conditions, cells were randomly oriented everywhere, as evidenced by the high and consistent variance in angle across the radial distance.

Fig. 4.

Comparison of typical radial sections from gels under interstitial flow (A) and static conditions (B). Velocity (v; 1st row) and pressure profiles (P; 2nd row) with cell orientation α (4th row) and variance Var( Embedded Image ) (5th row) were plotted against the radial distance (arrows). Inverted confocal images (3rd row) are maximum projections of 150-μm-thick sections of gels with fluorescein phalloidin-stained cells taken at ×100 magnification.

Fibroblasts regress under 2-D fluid shear stress within 3 h.

The effects of flow across a 2-D monolayer of fibroblasts were very different from those of interstitial flow on 3-D cultured fibroblasts. Within the same chamber and under the same volumetric flow rate of 10 μl/min, cells regressed (Fig. 3 C) within 3 h. This flow rate yielded a range of shear stresses on the surface of 0.048–0.014 dyn/cm2 from r i tor o. We took time-lapsed images to determine the kinetics of regression and to ascertain whether there was alignment of cells before regression (data not shown), but alignment was not observed at any time before cell rounding. We also observed that with higher flow rates, cells regressed even faster; thus a comparison of cell morphology with similar shear stress as that estimated for our interstitial flow studies (over an order of magnitude larger) was not feasible. This was in contrast to static 2-D controls, which retained a normal morphology throughout the experiment.

Darcy permeability reflects cell and matrix reorganization.

K′ was measured by monitoring the pressure drop across the chamber and comparing it with the imposed flow rate according to Darcy's law. When the daily average of K′, K¯ , was plotted against t, we noticed that K¯ for the fibroblast-populated cultures showed an increase in the first day (Fig.5). However, K¯ eventually decreased before it leveled off at a steady value of ∼0.25 × 10−6 cm2 by day 5. On the other hand, the K¯ for the acellular gel under interstitial flow initially increased and leveled off at ∼4.5 × 10−6 cm2.

Fig. 5.

One of the unique features of the flow model is that we can monitor Darcy permeability (K′) changes over time. Variability in the average bulk Darcy permeability data for the fibroblasts (⧫) and acellular gels (▴), averaged daily and presented in the graph as Embedded Image ′, suggests that interstitial flow was inducing matrix reorganization toward some optimal state.


The effects of interstitial flow on cells are poorly understood despite the importance of interstitial flow in tissue function. Here, we present a unique in vitro interstitial flow model for cell-populated biological matrices along with novel observations of cell alignment under interstitial flow within a 3-D matrix. Our interstitial flow model maintains the following design criteria: 1) it can be visualized microscopically in situ, 2) it allows tracking of matrix remodeling via changes in hydraulic conductivity, 3) it does not contract macroscopically, and 4) it can be maintained over long periods of time because some cell and matrix organizational changes are typically observed over a period of days to weeks. Although the data we present here utilize dermal fibroblasts, which is a relevant system for investigating interstitial flow in skin, the model can be easily extended to other cell systems such as vascular cells, chondrocytes, or smooth muscle cells as well as with other matrices.

The main challenge to the creation of the model was to maintain the mechanical integrity of the soft and highly compliant gel cultures undergoing interstitial flow without compaction or fluid channeling around, rather than through, the gel. The problem was complicated by cell contraction, which posed a serious challenge to our ability to monitor the hydraulic conductivity of our cultures because gel contraction caused the sample to detach, shrink, and reroute fluid flow. To address these problems, we used acid-treated porous PE material to anchor the collagen gels on the lateral edges and surface modification of the glass surfaces on the top and bottom. Therefore, the collagen gel was anchored in the radial direction by the PE rings and covalently bound to the top and bottom glass surfaces to inhibit detachment while allowing visual observations in situ. In addition, any gel contraction and fluid rerouting around the gel would significantly decrease the hydraulic resistance and thus be reflected in the pressure manometer readings. It should be noted that, whereas compaction might be a relevant and desirable feature for modeling interstitial flow through tissues such as cartilage (where compaction drives the flow) or the arterial wall (where the pressure gradient that drives the flow also expands the vessel lumen and compacts the wall), the boundaries of the interstitial space between the capillary and lymphatics are fixed and thus the pressure gradients driving this flow would tend to swell rather than compact the matrix.

We observed that interstitial cells (e.g., fibroblasts) respond to interstitial fluid flow in a 3-D environment very differently from that to pure shear flow when plated in a 2-D monolayer, as in parallel plate flow chamber experiments. The fibroblasts under interstitial flow align normal to the direction of flow, whereas cells in static controls retained a random orientation. In contrast, fibroblasts regressed into rounded structures when they were plated as a 2-D monolayer and subjected to similar flow rates. Under higher flow rates, such as to approach the estimated shear stress range in the 3-D cultures, the cells regressed even more quickly, making a morphological comparison between 2-D and 3-D cultures impractical. Furthermore, a direct comparison of shear stress between the two systems may not be relevant because the mechanical environment is more complex in the 3-D culture undergoing interstitial fluid flow due to the stresses on cell-ECM attachments (e.g., shear stress on ECM fibers leads to ECM strain and thus cytoskeletal strain via integrins). Nonetheless, these results clearly demonstrate that 2-D flow studies and 3-D static studies do not accurately reflect morphological behaviors of interstitial cells that are mechanically stimulated in their natural 3-D configuration. It also suggests that interstitial flow may affect how cells organize their environment.

The perpendicular alignment of the fibroblasts is consistent with observations in other soft tissues where interstitial flow is significant, such as smooth muscle cells in the arterial wall (12); however, it is important to note that circumferential stretch of the artery wall is likely to be responsible for this alignment (18). One potential rationale for our observations is that the cells realign to align the matrix to alter its mechanical properties, as seen in matrix remodeling in other tissues in response to mechanical stress (2, 12, 24). Interestingly, this perpendicular alignment under interstitial flow is different from other kinds of mechanical loading of fibroblast-populated collagen matrices where parallel alignment of the cells to the direction of the applied load is typically observed (6, 28).

The circumferential alignment of the fibroblasts appeared to occur in a critical range of bulk velocities (Fig. 4). The general trend we observed was that at high superficial velocities, fibroblasts showed a rounded and regressed morphology; as bulk velocity decreased, fibroblast morphology became more perpendicularly aligned and then displayed random orientation at near-static conditions. We hypothesized the cells regressed at higher shear stress. This was indirectly supported by morphological observations of rounded fibroblasts at higher interstitial flow rates (e.g., 100 μl/min, data not shown). In addition, the regressed cells were similar in morphology to those in the 2-D fluid shear studies under much lower fluid shear rates. Furthermore, the anchoring of a fibroblast in a 3-D network of extracellular fibers provides more support against a shearing force than if it were attached only on one side as in the 2-D monolayer; this may account for the observations that fibroblasts could withstand higher shear stresses in 3-D than in 2-D cultures.

One of the unique features about our model is that we can monitor changes in the mechanical properties of the tissue culture under interstitial flow as an indirect online monitor of matrix integrity and fluid channeling, because matrix composition and architecture is the primary determinant of K′ (15). In our results, we noticed that K′ for the cell-gel culture increased initially but later decreased and leveled off at a steady value, whereas K′ increased for the acellular gel but remained high. Although we have yet to elucidate the mechanisms behind these observations, they may be indirect indications of the fibroblasts' ability to regulate their mechanical environment. The initial increase in K′ may be due to fluid channeling within the gel, leading to a decrease in hydraulic resistance. In cell-populated gels, the subsequent return of K′ to its lower initial value may be the result of matrix synthesis and remodeling by the fibroblasts as well as perhaps their realignment. Moreover, the measurements ofK′ were consistently found to be in the narrow range of 0.25–2.5 × 10−6 cm2, suggesting that whereas cells could affect the hydraulic conductivity of the gel, they did not elicit drastic changes over the time period of 5–7 days. This average K′ value is consistent with that obtained previously for type I collagen (5).

In vivo measurements of K′ vary greatly due to location, method of measurement, and state of hydration of the tissue but are generally estimated to be in the range of 10−8 to 10−12 cm2 (15), with mouse tail skin recently estimated at 10−8 cm2(27). Reconstituted collagen gels have a much higher value of K′ because collagen concentrations in cell culture systems are by necessity very small (the maximum is ∼3.5 mg/ml), and also because reconstituted collagen gels do not have any specialized architecture nor are they incorporated with other ECM molecules such as proteoglycans, which strongly decrease K′ (15). Given this limitation, we could either mimic in vivo interstitial flow rates or interstitial fluid pressure gradients in our system, but not both. Interstitial flow rates are extremely difficult to measure in vivo, and estimates vary greatly due to measurement technique and tissue type (31). On the other hand, interstitial fluid pressure, although also difficult to measure and questionable, has been consistently found within a narrow range, with average values in the skin roughly within ±2 cmH2O (32). Thus our pressure gradient driving the flow is physiologically relevant, even if the flow rates that result are high due to the high K′.However, it is important to note that in vivo, K′ varies widely according to hydration and thus level of swelling in a tissue. For example, Guyton et al. (9) showed that changes in interstitial fluid pressure from −6.8 to +0.8 cmH2O led to changes in K′ of over five orders of magnitude; this infers that interstitial fluid velocity would also increase by five orders of magnitude for a given pressure gradient in a slightly edematous tissue. Thus, because of the large variability of fluid balance parameters found in vivo and particularly in edematous states, the range of interstitial fluid pressure and average fluid velocity as well as the hydraulic resistance in our system has clear physiological relevance.

In summary, the biological regulation of the interstitial fluid environment is critical to understanding interstitial architecture in health and disease, including fluid balance changes seen in edema. Our model can serve as a useful tool to study both short- and long-term effects of interstitial flow on cells and can easily be extended to examine other cell systems and processes, including those involved in angiogenesis, arterial wall remodeling, and bone remodeling, to name a few. Other interests include the study of fluid redistribution in the body due to gravitational effects such as deep sea diving and weightlessness experienced in space. Using our model, we observed fibroblast alignment perpendicular to interstitial flow. Furthermore, the variability in the bulk Darcy permeability K′ of the cultures suggested that interstitial flow was inducing matrix reorganization toward some steady state, although the mechanisms responsible have yet to be determined. These findings establish interstitial flow as having distinct effects on cells different from those of 2-D shear flow, and our newly developed model is a useful tool in understanding the biological regulation of interstitial fluid flow through soft tissues.


The authors thank Daniel Sedehi, Rick Boardman, Dr. Ranee Stile, Dr. Matt Glucksberg, and Dr. William Russin for invaluable technical assistance and advice.


  • This work was supported by the Whitaker Foundation and the National Science Foundation.

  • Address for reprint requests and other correspondence: M. A. Swartz, Dept. of Biomedical Engineering, Northwestern Univ., 2145 Sheridan Rd., Evanston, IL 60208-3107 (E-mail: m-swartz2{at}northwestern.edu).

  • 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.

  • First published January 16, 2003;10.1152/ajpheart.01008.2002


View Abstract