Water transport and homeostasis as a major function of erythrocytes

Abstract

Erythrocytes have long been known to change volumes and shapes in response to different salt concentrations. Aquaporin-1 (AQP1) was discovered in their membranes more than 20 yr ago. The physiological roles of volume changes and AQP1 expression, however, have remained unclear. We propose that rapid water exchange through AQP1 coupled with large capacity for volume change may allow erythrocytes to play an important role in water regulation. In this study, we showed that erythrocytes in situ gradually reduced their volumes by 39% in response to the hyperosmotic corticomedullary gradient within mouse kidneys. AQP1 knockout (KO) erythrocytes, however, displayed only minimal reduction. Constructing a microfluidic device resembling capillary flow with an extracellular fluorescent reporter demonstrated that water exchanges between erythrocytes and their hypotonic or hypertonic surroundings in vitro reached steady state in ~60 ms. AQP1 KO erythrocytes, however, did not show significant change. To simulate the water transport in circulation, we built basic units consisting of three compartments (i.e., erythrocyte, plasma, and interstitial fluid) using Kedem-Katchalsky equations for membrane transport, and connected multiple units to account for the blood flow. These simulations agreed with experimental results. Importantly, volume-changing erythrocytes in capillaries always “increase” the osmotic gradient between plasma and interstitial fluid, making them function as “micropumps” to speed up the regulation of local osmolarity. Trillions of these micropumps, mobile throughout the body, may further contribute to water homeostasis. These insights suggest that the enhanced exchange of water, in addition to O₂ and CO₂, may well be the third major function of erythrocytes.

NEW & NOTEWORTHY Physiological roles of erythrocyte volume change and aquaporin-1 were proposed and investigated here. We conclude that fast water transport by aquaporin-1 coupled with large volume-change capacity allows erythrocytes to enhance water exchange with local tissues. Furthermore, their huge number and mobility allow them to contribute to body water homeostasis.

INTRODUCTION

A principal function of blood is to redistribute critical molecules in the body. Erythrocytes, which occupy ~45% of blood volume, function to enhance such redistribution. The efficient transport of O₂ and CO₂ as the two major functions of erythrocytes has long been documented; while O₂ diffuses passively across the lipid bilayer, the transport of CO₂ is achieved by the transmembrane anion exchanger (i.e., anion exchanger 1 or band 3). The erythrocyte membrane, including its molecular organization, geometry, and mechanical properties, has been extensively investigated in a number of laboratories, including ours (3, 38).

In 1992, a component of band 7, CHIP28, in the human erythrocyte membrane was cloned, characterized, and renamed aquaporin-1 (AQP1) after it turned out to be the long sought-after water channel (28). There are 200,000 copies of AQP1 per erythrocyte (15) with a molecular weight of 28 kDa for each monomer (37, 44). Crystallography revealed a highly selective filter for water (4a) and that transport depends on the osmotic gradient, requiring no ATP or other forms of energy (4a, 37). While water moves across an artificial lipid bilayer only in the range of 2.3–6.3 × 10⁻⁴ cm/s (39), the presence of AQP1 allows for a much higher water permeability of ~1.8 × 10⁻² cm/s in erythrocytes (46).

Despite continued, outstanding research on AQP1, its physiological role in erythrocytes has not been clearly demonstrated. Targeted disruption of the AQP1 gene at exon 1 globally resulted in complete lack of expression of AQP1 in the knockout (KO) mice from birth, which displayed severely impaired urinary concentrating ability and elevated plasma osmolality (22). However, AQP1 is also expressed in a few other cell and tissue types (4, 26, 27, 42), such as the endothelium of blood vessels and the epithelium of the lungs, kidneys, eyes, and lacteals. The observed phenotype could thus be attributed to the disruption of AQP1 expression in erythrocytes as well as in nonerythrocytes.

Although erythrocytes in mice also express an aquaglyceroporin, aquaporin-9, aquaporin-9-null mice display no change in membrane water permeability (21). Similarly, human erythrocytes express an aquaglyceroporin, aquaporin 3; however, AQP1 dominates water transport in the erythrocyte membrane (31).

Water constitutes ~70% of our body weight and is essential for life, serving as the universal medium for many biological reactions. As we gain water by drinking and metabolism and lose water by urine excretion and evaporation, variations of osmolarity in tissues (e.g., the intestine, lungs, skin, and kidneys) are created, making timely water redistribution important. Even though systemic circulation takes less than a minute to complete, microcirculation flows much slower at a rate of ~1 mm/s (13, 24, 40). We hypothesized that circulating erythrocytes equipped with AQP1 enhance the ability of blood to transport water and regulate tissue osmolarity.

Human and mouse erythrocytes are normally biconcave, giving them a very large surface-to-volume ratio. Although the membrane surface area in each is conserved, the erythrocyte volume can change dramatically due to its membrane deformability, lack of restraining transcellular cytoskeleton (except for the membrane skeleton), and extra space normally occupied by the nucleus and other organelles. Erythrocytes swell to become stomatocytes in hypoosmotic medium and shrink to become echinocytes in hyperosmotic medium. It is remarkable that a stomatocyte has a volume that is approximately three times that of an echinocyte. We propose that the large volume change capacity allows erythrocytes to carry and redistribute a large amount of water more effectively in the circulation compared with erythrocytes lacking this capacity.

This study took a multidisciplinary approach to investigate the role of erythrocytes in water transport and homeostasis. We quantified erythrocyte volume changes in kidneys in situ, constructed a microfluidic device to monitor the real-time response of erythrocytes to extracellular osmolarity, and simulated water transport by erythrocytes during capillary flow. In each investigation, we analyzed wild-type (WT) mouse erythrocytes in contrast with that from AQP1 KO mice (22). The results demonstrated that each erythrocyte, enabled by AQP1 and a large capacity to change volume, may function as a “micropump” to regulate local osmolarity and that water transport and homeostasis may be a major, previously undiscovered or underappreciated, function of erythrocytes.

MATERIALS AND METHODS

Animals used in in situ and in vitro experiments.

We used mice for experiments following an Institutional Animal Care and Use Committee-approved protocol issued by the University of California-San Diego, which covered all procedures, including breeding, tissue collection, and cervical dislocation. WT mice (with a background of C57/Black and 129/Svj) were from the laboratory of L. A. Sung (3). Mice were bred in an animal facility at the University of California-San Diego and were genotyped by Transnetyx (Cordova, TN). AQP1 KO mice (22) were kindly provided by Dr. Alan S. Verkman (University of California-San Francisco). They were shipped to the University of California-San Diego based on a Material Transfer Agreement in two separate batches of four mice (2 male mice and 2 female mice) each. Mice were not bred but euthanized for collections of kidneys (in situ experiments) or erythrocytes (in vitro experiments) within 24 h after arrival in the Molecular Bioengineering Laboratory.

Kidney and tissue preparation.

WT and AQP1 KO mice, aged 4–18 mo, were euthanized via cervical dislocation, and kidneys were excised within 15 s. Immediate immersion of kidneys in freshly made 4% paraformaldehyde at 4°C for 12–24 h was to prevent changes to erythrocytes. Kidneys were then transferred to a 30% sucrose solution at 4°C for 24–48 h for cryoprotection. Kidneys were embedded in OCT (Tissue-tek Optimum Cutting Temperature compound, Sakura Finetek USA, Torrence, CA), frozen by dry ice, and stored at −20°C until sectioned by cryostat (Leica, Nussloch, Germany). Other tissues such as the intestine, where water influx is likely to occur, were not studied in this investigation because the timing of water intake and/or osmotic gradient would be more difficult to control.

Tissue staining.

Tissue sections (20–30 μm) mounted on each slide were stained with 3,3′-diaminobenzidene (DAB kit, Innovex Biosciences, Richmond, CA) for 5 min. They were counterstained by a 1:100 diluted stock solution of ATTO 655 (Sigma-Aldrich, St. Louis, MO), a fluorescent dye, and imaged within 8 h.

Confocal microscopy and three-dimensional reconstruction.

Erythrocytes were imaged using an Olympus FV1000 spectral confocal microscope. Image stacks were obtained with a z-spacing of 0.2 μm and a resolution of 1,024 × 1,024 pixels, resulting in a 0.0308 μm/pixel ratio. The x/y coordinates of images were taken relative to each other and in the directionality of the loop of Henle. Image stacks were processed with IMOD (version 4.7.4, University of Colorado), an image processing, modeling, and display set of programs. Contours around erythrocytes were obtained in each image stack and then joined together using triangular meshing to create three-dimensional (3-D) polygons whose surfaces were composed of triangles. Using the center of mass for each erythrocyte and coordinates from each triangle, a series of triangular pyramids can be created, whose total volume would be the volume of the erythrocyte.

Erythrocyte purification.

Mouse blood was collected via heart puncture using acid citrate dextrose as an anticoagulant after cervical dislocation. This method was necessary to provide the amount of blood required for the microfluidic experiments. Erythrocytes were purified by washing and centrifugation, as previously described (20).

Chip manufacturing.

Polydimethylsiloxane (PDMS) chips were manufactured and prepared, as previously described by Voigt (43). A W-junction was used to change final salt concentrations by combining a high- and low-salt solution in two reservoirs at different adjustable heights. The main channel containing an erythrocyte suspension was 10 × 10 × 2,000 μm to mimic capillary flow. Channels were pretreated with 1% BSA in PBS for 30 min to reduce erythrocyte adhesion to the walls.

Microfluidic device setup.

Five 50-ml syringes per chip were assembled with blunt 26-gauge needles, each connected to 6 ft of Tygon tubing and to the chip with a bent 26-gauge needle. Two salt solutions (1,800 and 0 mosmol/l NaCl), an erythrocyte suspension with Sytox green (25 μM), and two waste solutions (deionized water) were loaded in their respective syringes. Before experimentation, calibration was performed by the addition of ATTO 655 to a single channel of the W-junction. By observing the laminar flow of the W-junction as well as the downstream Y-junction, the hydrostatic pressures could be adjusted to achieve consistent mixing and flow conditions within the main channel (1 mm/s).

Fluorescence and bright-field videos.

The microfluidic device was fixed to the stage of a fluorescence microscope and video was recorded (43). Bright-field video was taken with 10- or 20-ms intervals; fluorescence video, imaged at 530 nm, was taken with 10- or 100-ms intervals.

RESULTS

In situ volume measurement of erythrocytes in the kidney.

We chose mouse kidneys to study erythrocyte volume changes because tissues surrounding the loop of Henle provide a directional corticomedullary osmotic gradient (9, 34). We took advantage of a high level of naturally existing peroxidase activity inside erythrocytes to darkly stain them with 3,3′-diaminobenzidene. By confocal microscopy, darkly stained erythrocytes were revealed in both the cortex (Fig. 1A) and medulla (Fig. 1B), in sharp contrast against the fluorescent background (see Tissue staining). This novel approach allowed for high-resolution acquisition of image stacks without photobleaching and detailed contouring of complex cell shapes (Fig. 1C), leading to accurately meshed membranes (Fig. 1D) and quantified cell volumes of erythrocytes (Fig. 2).

Fig. 1.

Confocal imaging and three-dimensional reconstruction of erythrocytes in renal tissues. A: cortex region showing erythrocytes in a glomerulus (G). Erythrocytes darkly stained by 3,3′-diaminobenzidene were contrasted with extracellular fluorescent counterstain and tissue autofluorescence. B: medulla region showing several capillaries (near the center) inferred from the deformed erythrocytes inside. Representative images of erythrocytes are shown in C and D. C: erythrocytes from image stacks were contoured. D: contours from each image were joined by triangular meshing (see materials and methods).

Fig. 2.

Color map and histogram of erythrocyte volumes. A and B: three-dimensional distribution of erythrocyte [red blood cell (RBC)] volumes in a representative wild-type (WT; A) and aquaportin-1 (AQP1) knockout (KO) mouse kidney (B). Bright red to dark blue in the color bar above (A) and (B) indicates the color spectrum from larger to smaller erythrocyte volumes in femtoliters. The location in three-dimensional space (x, y, z) is the relative position of each erythrocyte within the kidney. The orientation of the kidney is indicated by labels for the cortex and medulla along the axes. C: histogram of erythrocyte volumes in WT kidneys (n = 5). Each bin corresponds to 100.0 µm. D: histogram of erythrocyte volumes in AQP1 KO kidneys (n = 4) analyzed similarly. Each bin corresponds to 63.6 µm. The boundary between the cortex and medulla is indicated by a vertical line. Scale bar = SD.

Erythrocyte volumes in situ and their 3-D arrangement in the kidney were recorded. A 3-D color map of a WT kidney (Fig. 2A) revealed reduced cell volumes (more blue) in the medulla relative to the cortex (more red). In contrast, erythrocytes in an AQP1 KO kidney showed no clear changes with location (Fig. 2B).

Indeed, the histogram of WT kidneys (Fig. 2C) showed a trend of decreasing erythrocyte volumes against a presumed gradual increase of interstitial osmolarity from the cortex to the inner medulla (9, 34). The maximal reduction of volumes was 39% in the inner medulla (31.55 ± 1.46 fl, mean ± SD) from that in the cortex (51.83 ± 3.26 fl). The histogram of AQP1 KO kidneys (Fig. 2D), on the other hand, showed no significant change of cell volumes throughout the kidney. The cell volume of AQP1 KO erythrocytes (49.53 ± 4.86 fl) agreed with the mean volume (45–50 fl) known for mouse erythrocytes from the literature (32) as well as the Mouse Phenome Database from The Jackson Laboratory, suggesting that our procedures did not cause significant changes in cell volumes during tissue preparations and that the trend of changes in erythrocyte volume observed in WT kidneys was not an artifact.

Water exchange by erythrocytes in microfluidic devices.

To quantify the response of erythrocytes in real time, we designed and manufactured a microfluidic device (Fig. 3A) in which the kinetics of water transport was examined in a two-compartment system with controlled conditions (without the influence of interstitial fluids). Erythrocytes purified from WT or AQP1 KO mice were mixed with a surrogate solution for plasma [consisting of NaCl, sodium phosphate buffer (pH 7.4), and 0.1% BSA] and loaded into the device. Capillary flow conditions were replicated in a main channel of 10 µm (in cross section) after an erythrocyte suspension (in 300 mosmol/l) was merged with various salt solutions through a Y-junction (Fig. 3A) to create a 20% hematocrit.

Fig. 3.

Microfluidic design with an extracellular fluorescent reporter. A: bright-field imaging of the Y-junction, showing an inlet of salt solution and that of a erythrocyte [red blood cell (RBC)] suspension. Positions 1 and 2 in the main channel are marked. An upstream W-junction (not shown; see materials and methods) mixed a high-salt solution and a low-salt solution to produce the desired osmolarity. B: fluorescence imaging of the same Y-junction, showing the fluorescent reporter (Sytox green, excitation wavelength: 488 nm) present in the erythrocyte suspension. C: changes of the fluorescence intensity in response to different osmolarities (150, 300, and 900 mosmol/l) seen between positions 3 and 5 (image processing was used to enhance contrast for this panel). Notice that at 150 mosmol/l, the fluorescence intensity at position 3 was lower than those of positions 4 and 5. On the other hand, at 900 mosmol/l, the fluorescence intensity at position 5 was lower than those of positions 3 and 4. At 300 mosmol/l, the fluorescence intensity remained the same across positions 3−5. Positions 1–5 were each separated by 15 µm.

A membrane-impermeable fluorescent indicator was premixed with the erythrocyte suspension to mark the extracellular (or plasma) compartment (Fig. 3B). Under a fluorescence microscope, changes in fluorescence intensity along the main channel (at positions 1−5 in Fig. 3) were recorded, which reflected the water exchange by erythrocytes in response to the changes of osmolarity. When erythrocytes were exposed to 300 mosmol/l (Fig. 3C, middle), no detectable changes were recorded over time along the main channel (from left to right, only positions 3, 4, and 5 are shown). Erythrocytes swelled under hypoosmotic conditions (data not shown), and the fluorescence intensity increased (Fig. 3C, top); under hyperosmotic conditions, erythrocytes shrank (data not shown), and the fluorescence intensity decreased (Fig. 3C, bottom).

Using an upstream W-junction (see Chip manufacturing), osmolarity in the main channel could be changed without interrupting the flow. We were able to reverse the fluorescence intensity of the erythrocyte suspension by reversing the upstream osmolarity (data not shown). Such pretests for each experiment ensured the reliability and reproducibility of the microfluidic setup.

A flow rate of 1 mm/s in the main channel was used in all experiments, resembling the physiological flow rate for mouse capillaries (13, 40). When WT erythrocytes were subjected to constant osmolarities (ranging from 150 to 900 mosmol/l), the fluorescence intensity changed along this channel, with the maximum changes occurring near position 5 (60 µm downstream from the Y-junction) over a period of 60 ms (Fig. 4A). In comparison, when KO erythrocytes were subjected to similar treatments, there was no significant change over the same distance and time period (Fig. 4B). The difference between WT and KO erythrocytes suggested that the observed change in fluorescence intensity was the result of water transport through AQP1 across the erythrocyte membrane. These dynamic measurements may reflect the redistribution of water between the cellular and extracellular compartments and corresponding volumetric and light-scattering changes of erythrocytes.

Fig. 4.

Kinetics of water transport by erythrocytes in microfluidics. A: measurements of fluorescence intensity were taken at positions 1–5 within the main channel containing a wild-type erythrocyte suspension (20% hematocrit) mixed with a solution of a different osmolarities. For each osmolarity, four independent experiments were carried out, and the fluorescence intensity of each position was measured a total of 40 times (n = 40). B: experiments were repeated using aquaporin-1 (AQP1) knockout (KO) erythrocytes. C: fluorescence intensities at position 5 (60 ms after mixing) from A were replotted against the calculated osmolarity after mixing. D: fluorescence measurements (n = 5) by spectrophotometry (arbitrary units) 20 min after mixing components identical to A.

To observe the steady state of water exchange, experiments were repeated by mixing WT erythrocytes in solutions of various osmolarities in a 96-well plate for 20 min with the fluorescent intensities quantified by spectrophotometry (Fig. 4D). Results were the same with longer incubation times (data not shown). Figure 4C shows replotted fluorescence intensities at position 5 or at 60 ms from Fig. 4A. The similar pattern between Fig. 4, C and D, indicates that the dynamic changes in fluorescence intensity in the microfluidic device had reached steady state near 60 ms. Spectrophotometry of KO erythrocytes was not performed because no significant changes in fluorescence intensity had been observed (Fig. 4B).

Computational modeling of erythrocyte water transport.

In addition to in situ and in vitro experiments, we conducted computational modeling to simulate and make predictions about water transport by erythrocytes in mice and humans. This was done using a single-unit approach, in which three compartments (an erythrocyte, plasma, and the surrounding interstitial fluid) composed a basic unit (Fig. 5A).

Fig. 5.

Computational modeling of erythrocyte [red blood cell (RBC)] water transport in capillary flow. A: schematic of the single unit consisting of an erythrocyte (1), plasma (2), and surrounding interstitial fluid (3). Linking single units longitudinally formed a capillary structure. B: erythrocyte volumes predicted as they passed through a hyperosmotic gradient similar to that in the kidney (300 to 1,200 mosmol/l, from the cortex to the medulla). The volumes for wild-type (solid line) and aquaporin-1 (AQP1) knockout (KO) erythrocytes (dotted line) are presented. The volumes predicted by simulations in different capillary blood flow velocities (C) and various hematocrits (D) are also shown. These erythrocytes passed through a capillary surrounded by an interstitial fluid with an osmotic gradient changing from 300 to 1,200 mosmol/l (●) or from 300 to 0 mosmol/l (○). E: illustration featuring the importance of the tight junction (t) between endothelial cells (E) that is permeable to both water (W) and salts (S) as well as that of the erythrocyte (R) membrane, which allows fast water transport through AQP1 (a), resulting in its volume changes. The change of erythrocyte volumes, regardless of the direction, always increases the osmotic gradient between plasma and interstitial fluid. Endothelial cells are shaded, indicating the presence of an intracellular cytoskeleton and organelles. The drawing is not to scale.

The general equations governing volumetric flow across the capillary wall and erythrocyte membranes were based on the Kedem-Katchalsky approach for describing transport across a membrane driven by osmotic and hydrostatic pressure (16).

The equation below describes the transport of water across the capillary wall (J_v) driven by osmotic (C) and hydrostatic pressure (P):

$$J_{\text{v}}\left(x,t\right)={\text{ρ}}_{\text{f}}A{\text{V}}_{\text{w}}\{{\text{C}}_{\text{i}}\left(x,t\right)-{\text{C}}_{\text{e}}\left(x,t\right)-\text{σ}\left[{\text{C}}_{\text{d,i}}\left(x,t\right)-{\text{C}}_{\text{d,e}}\left(x,t\right)\right]\}-\left[{\text{P}}_{\text{i}}\left(x,t\right)-{\text{P}}_{\text{e}}\left(x,t\right)\right]$$

where ρ_f is water permeability of the capillary wall, A is the surface area, V_w is the partial molar volume of water, σ is the reflection coefficient, subscript d denotes diffusible solutes, and subscripts i and e denote internal and external compartments. Together, ρ_f × V_w are equal to the hydraulic conductivity of the capillary wall (30).

The second equation below describes the corresponding flux of solutes (J_c) resulting from the water volumetric flux (J_v) across the capillary wall, whose tight junctions permit water and solute transport:

$$J_{\text{c}}\left(x,t\right)=\left(1-\text{σ}\right)\check{\text{C}}\left(x,t\right)J_{\text{v}}+{\text{ρ}}_{\text{d}}\left[{\text{C}}_{\text{d,e}}\left(x,t\right)-{\text{C}}_{\text{d,i}}\left(x,t\right)\right]$$

Where $\check{\text{C}}$ is the mean concentration difference of the solute between internal and external compartments and ρ_d is the permeability of the solute across the capillary wall. Together, these two equations adequately describe water and solute transport, respectively, between plasma and interstitial fluid (compartments 2 and 3, respectively, in Fig. 5A).

However, in the case of water transport across the erythrocyte membrane, we assumed there was no significant contribution of hydrostatic pressure because of the remarkable deformability of the cell membrane (anisotropic and bending) and the large capacity for erythrocytes to change volume. With the movement of water into the erythrocyte, there would be a minute, instantaneous increase in the hydrostatic pressure of the cell with a simultaneous, corresponding decrease in the hydrostatic pressure of the surrounding fluid. This minute imbalance, however, would be immediately corrected by the expansion of the erythrocyte volume through the movement of the membrane. In other words, given the high hydraulic conductivity and deformability of the membrane, the minute transient change would not be sufficient to contribute to the calculations in a meaningful way. Thus, the flux of water through the erythrocyte membrane, as described in the equation below, would depend only on the concentration difference between erythrocytes and plasma without the hydrostatic pressure difference, the last term in the first equation:

$$J_{\text{v}}\left(x,t\right)={\text{ρ}}_{\text{f}}A{\text{V}}_{\text{w}}\left\{{\text{C}}_{\text{i}}\left(x,t\right)-{\text{C}}_{\text{e}}\left(x,t\right)-\text{σ}\left[{\text{C}}_{\text{d,i}}\left(x,t\right)-{\text{C}}_{\text{d,e}}\left(x,t\right)\right]\right\}$$

In addition, there was no associated solute flux due to the specificity of the AQP1 channel for water. Although the movement of water would change the salt gradient across the erythrocyte membrane, the transport of salts is typically several orders of magnitude slower (12). Therefore, the above equation alone would be adequate to describe the erythrocyte water transport across its cell membrane (between compartments 1 and 2 in Fig. 5A). In these simulations, ρ_f of 1.8 × 10⁻² cm/s was used for WT erythrocytes and 3.0 × 10⁻³ cm/s for AQP1 KO erythrocytes (46).

Multiple basic units were then linked longitudinally to form a capillary-like structure to account for blood flow between units and molecular diffusion across the membranes (i.e., capillary wall and erythrocyte membrane). In the case of mice, the simulated diameter of biconcave erythrocytes is 5.5 µm (surface area of 90 µm²), capillary diameter is 6 µm, and the interstitial compartment is defined by a radius of 6 µm from the center of the capillary (Fig. 5A). In general, we used physiological conditions for simulations, e.g., hematocrit was 20% (8), flow velocity for the blood in the capillary was 1.0 mm/s (13, 24, 40), and the length of the capillary used was 1 mm. Hydrostatic pressure along the capillary (arterial to venous) ranged from 32 to 16 mmHg while the interstitial fluid was typically 24 mmHg.

The first simulations we performed resembled the vasa recta, the capillaries surrounding the loop of Henle, against an osmotic gradient in the interstitial fluid, ranging from 300 to 1,200 mosmol/l from the cortex to the inner medulla (9, 34). In this special case, the blood flow rate used was 0.75 mm/s. WT erythrocytes in these simulated capillaries were predicted to lose water, leading to a 40% volume reduction at 1,200 mosmol/l, while only a slight reduction (2%) was predicted for AQP1 KO erythrocytes (Fig. 5B). The erythrocyte volume changes predicted by simulations (Fig. 5B) were similar to those found experimentally in situ (Fig. 2, C and D).

The simulations also allowed us to explore other conditions difficult to replicate experimentally, such as variations in blood flow velocity (Fig. 5C) or hematocrit (Fig. 5D) against hyperosmatic and hypoosmatic gradients along a capillary length. For example, if the blood flowed through a capillary surrounded by a hypoosmotic interstitial fluid with a gradient that changed from 300 down to 0 mosmol/l over 1 mm, in the end the volume of erythrocytes would increase by 20% (50 fl × 1.2 = 60 fl) if the flow velocity (x = 1) and hematocrit (x = 20%) were in normal physiological condition.

It is interesting that our simulations with these blood flow velocities (Fig. 5C) or hematocrits (Fig. 5D) against a hypoosmotic gradient (from 300 down to 0 mosmol/l in the interstitial fluid) predicted no intravascular hemolysis over a wide range. Each individual erythrocyte, however, was predicted to have a greater volume change, reflecting a greater responsibility in regulating surrounding tissues (toward isoosmotic conditions), if the blood flow or the hematocrit dropped.

We also simulated water transport in human blood circulation. Here, we used 8 µm in diameter (2 µm in thickness and 140 µm² for the total surface area) for erythrocytes and 8 µm in diameter and 1 mm in length for the capillary. The interstitial space was set to be proportionally equivalent by volume to that in mice. The resulting water transport across the capillary wall was faster than that of mice (data not shown), mainly due to the larger surface areas of both erythrocytes and capillaries. However, the relative volume changes of erythrocytes along the capillary were not significantly different. This was due to the similar blood flow rate and the osmolarity gradient in the interstitial compartment.

It is reasonable that as erythrocytes travel through hyperosmotic tissues, they respond by releasing water and shrink (Fig. 5E), expanding the plasma volume, while keeping the total volume of the capillary constant. This would dilute the plasma concentration and increase the osmotic gradient between the plasma and interstitial fluid, which would further facilitate water movement out toward the hyperosmotic tissues.

On the other hand, as erythrocytes travel through “hypoosmotic” tissues, they would respond by expanding their own volume and reduce the space for plasma (Fig. 5E). This would effectively raise the plasma concentration and again also increase the osmotic gradient between the plasma and hypoosmotic interstitial fluid, which would facilitate water influx from the interstitial space into the plasma.

The responses described above would lead to an enhanced regulation of local osmolarity. As shown in Fig. 5E, there was a major difference between the flow of water together with salts through the tight junctions and the selective transport of water across the erythrocyte membrane. As a result, erythrocytes, through changing volumes, always “increase” the osmotic gradient between the plasma and interstitial fluid. Consequently, each erythrocyte may function as a “micropump” to facilitate water transport that brings the interstitial fluid toward isoosmolarity. Trillions of erythrocytes in our body may thus adjust tissue osmolarity locally, quickly, and simultaneously. At the same time, their remarkable mobility in microcirculation and systemic circulation further allows them to redistribute water effectively and participate in body water homeostasis.

DISCUSSION

It is well established that the efficient exchanges of O₂ and CO₂ in tissues are the two major functions of erythrocytes. We demonstrated here, analogous to O₂ and CO₂, why the efficient exchange of water should be another major function of erythrocytes. In the case of O₂, the exchange is via passive diffusion across the lipid bilayer and the ability to carry more O₂ is achieved by the presence of hemoglobin, which has a high affinity toward O₂. In the case of CO₂, the exchange is via the transmembrane anion exchanger, and the ability to carry more CO₂ is due to the presence of carbonic anhydrase, which converts CO₂ into highly soluble bicarbonate. Here, we showed that in the case of water, the exchange occurs rapidly through the water channel and the ability to carry more water is due to the large capacity of erythrocytes to change volume to accommodate the influx or efflux of water.

Large volume-change capacity.

Lack of a nucleus, organelles, and especially transcellular cytoskeleton allows erythrocytes to change volumes without much resistance. Their single layer of spectrin-actin membrane skeleton (20, 38) is able to accommodate bending and anisotropic deformation (2, 3, 5, 36), and thus, changes of cell volume. Other nucleated cells (e.g., endothelial cells) will need to overcome more hydrostatic pressure when water enters or leaves the cell, as their transcellular cytoskeleton would resist changes to cell volume.

The transport of water across the erythrocyte membrane is synonymous with cell volume changes (see Computational modeling of erythrocyte water transport). In the isosmotic condition, water occupies ~65% of the cell volume. Under hyperosmotic conditions, for example, in the inner medulla, we found a 39% reduction of erythrocyte volume in situ. This finding is in agreement with a 40% volume reduction by simulation at 1,200 mosmol/l. Interestingly, such volume reductions represent almost a complete loss of free water [~39% of cell volume, based on our calculations from the composition of an erythrocyte and a previous experimental study (19)]. This implies that in the vasa recta near the bottom of the medulla, all or nearly all of the free water in erythrocytes is transferred to the surrounding plasma and interstitial fluid. These results also suggest that the lack of hydrostatic pressure gradient between the erythrocyte and plasma assumed in simulations is not far from physiological conditions.

While water efflux is limited by the amount of free water inside the erythrocytes (from this study), water influx is ultimately restricted by the elastic stress and surface tension in the cell membrane (7). An earlier report has shown that, under hypoosmotic conditions, e.g., at 131 mosmol/l, erythrocytes in vitro increased their volume by 74%, became spherical, and were on the verge of hemolysis (7). Under our simulated conditions, there was no predicted intravascular hemolysis. The large capacity of the erythrocyte to swell would also help prevent intravascular hemolysis if prolonged hypoosmotic conditions occur.

As described above, there are limits of erythrocyte volume change: erythrocytes can shrink up to 40% after losing all free water under hyperosmotic conditions or swell 74% before they burst under hypoosmotic conditions. Therefore, erythrocytes have a greater capability to swell below 300 mosmol/l than to shrink above 300 mosmol/l. The asymmetric, bidirectional capacity of erythrocytes to take up or release water is of significance in water regulation. For example, an erythrocyte that has swollen 70% under hypoosmotic conditions can potentially release more water: 110% (70% + 40%) of its original biconcave volume when it later passes through hyperosmotic tissues.

Without such a large capacity to change volume, erythrocytes would be limited in their ability to regulate local osmolarity. A biconcave cell shape and high membrane deformability allow volume changes to accommodate the water influx and efflux, which, in turn, prevents significant alteration of the hydrostatic pressure both inside and outside of erythrocytes. Without a flexible membrane and deformable cell body, water transport across the erythrocyte membrane would be impeded by the opposing changes of the hydrostatic pressure.

It has been reported that some patients with sickle cell anemia are not able to concentrate their urine in early childhood, which may be reversible with blood transfusion (13a). There is a possibility that dehydration of a patient’s erythrocytes in the inner medulla may promote hemoglobin crystallization, resulting in reduced cell deformability. Such reduced volume-change capacity may prevent sickle cells from reabsorbing water when they ascend from the medulla to the cortex. This would result in water retention in the interstitial fluid, contributing to dilute urine in these patients. The reported reversibility by blood transfusion with normal erythrocytes supports this possibility.

Rapid water transport.

Our in vitro microfluidic study indicated that normal erythrocytes took ~60 ms or less to reach steady state in osmolarity with the surrounding fluid. This has important implications for erythrocyte water transport in physiology. Since the blood in capillaries flows at ~1.0 mm/s in vivo, water transport between erythrocytes and the plasma over 60 μm (only a fraction of a capillary) can be significant. Thus, not only do erythrocytes function to restore tissues toward isoomolarity, they do so rapidly and locally in response to hypoosmolarity or hyperosmolarity.

With much slower water transport, erythrocytes may not be able to regulate local osmolarity. Our findings that AQP1 KO erythrocytes lacked significant responses to osmotic changes, both in situ and in vitro, were in agreement with this prediction.

There have been three humans identified with mutations in AQP1 (lacking the Colton blood group) who appear phenotypically normal (29). However, two of these three patients with AQP1 deficiency were studied further and shown to have urinary concentrating defects under water deprivation (431 mosmol/kg in subject 1 and 460 mosmol/kg in subject 2 compared with 775−1,200 mosmol/kg in normal subjects) (14). At the time of these studies (2001), there were only six kindred confirmed to lack the Colton blood group, suggesting that the absence of AQP1 function and/or the Colton blood group may be associated with a low survival rate in the population. Although these followup studies only examined two patients, one of them had several complications with multiple pregnancies.

In contrast to humans, mice lacking AQP1 displayed a more drastic reaction to dehydration. After 36 h of water deprivation, KO mice had serum osmolality increases, greater body weight decreases, and severely impaired ability to concentrate urine (657 ± 59 mosmol/kg) compared with WT (>2,500 mosmol/kg) and heterozygous mice (22). It is known that monomeric urea transporter B, which also transports water, exists in the membrane of erythrocytes (45). However, on the basis of AQP1 KO experiments, AQP1 accounts for 90% of the total water permeability of the erythrocyte membrane (46).

Although the underlying mechanisms for these observed phenotypes are not completely understood, one possibility (suggested by our investigations) is that erythrocytes without normal AQP1 may not transport water fast enough, causing an asymmetry of the water transport in the vasa recta. If the speed of erythrocyte water transport relative to blood flow is fast enough to equilibrate with their surroundings, such as in the case of the WT mice, approximately equal amounts of water would be lost and regained (all other conditions being equal). As the speed of water transport slows down with KO or dysfunctional AQP1 mutations, a lag may develop between the erythrocyte’s intracellular osmolarity and the plasma. In such cases, erythrocytes in the ascending portion of the vasa recta may continue to lose water instead of reabsorbing water, leading to an asymmetry in the amount of water lost and reabsorbed. When the phase of water release is prolonged and the phase of water uptake is delayed, asymmetric water transport by erythrocytes between the descending and ascending vasa recta may occur. This net loss of water from the blood to the interstitial fluid may dilute the corticopapillary gradient, possibly leading to an inability to concentrate urine.

Since AQP1 KO mice have dilute urine, we propose to transfuse AQP1 KO mice with WT erythrocytes as well as transfuse WT mice with AQP1 KO erythrocytes to confirm whether erythrocytes play a significant role in concentrating urine and/or the formation of the medullary concentrating gradient. It would also be interesting to investigate the potential roles erythrocytes play in other physiological conditions, such as dehydration and antidiuresis, or in disorders such as renal failure, where blood urea nitrogen is elevated, a syndrome of inappropriate antidiuretic hormone secretion, and sickle cell and other anemic diseases. The procedures and designs established in this report would be helpful in aiding these future investigations.

Urea in the inner medulla.

It is understood that in the inner medulla, urea in addition to salts also contributes to the osmolarity (6, 45), especially during antidiuresis when more urea is released from the collecting ducts (6, 35, 47). Although the presence of urea increases the osmotic gradient in the inner medulla, urea is predicted to enter erythrocytes quickly through urea transporter B to equalize the urea concentration across the membrane (11, 33), minimizing the effect on erythrocyte volumes (23). Because of this, salt remains a major contributor to erythrocyte volume changes in the vasa recta.

Our in situ experiments used mouse kidneys under their general physiological conditions and detected a maximum reduction of volume by 39% in the vasa recta of the inner medulla. These results are comparable to our simulated erythrocyte volume reductions and that reported by Macey and Yousef (23) (~35% erythrocyte volume reduction), in which a combined gradient of 1,400 mosmol/kg (700 mosmol/kg from urea and 700 mosmol/kg from salt) was used to simulate the inner medulla. If intracellular urea increases from 0 to 700 mosmol/l, erythrocytes would increase their volumes by ~0.5%.

Participation of erythrocytes in water homeostasis.

It is well known that arginine vasopressin (AVP), also known as antidiuretic hormone, plays an important role in body water homeostasis by increasing water reabsorption (into the interstitial space) from the urine (10, 41). This is accomplished, in part, by inducing translocation of aquaporin 2, a different water channel expressed in renal principal cells of the collecting ducts, from the cytoplasm to the apical membrane (1, 17). AVP is synthesized and regulated by osmoreceptors in the hypothalamus, stored in the posterior pituitary gland, and released into the plasma (10). Previous studies have suggested that these osmoreceptors have a baseline threshold, above which plasma AVP increases linearly with osmolality (41). As a form of hormonal regulation, it takes time, not only for the microcirculation to feed into the systemic circulation, but also for the systemic circulation to bring AVP to the kidney to carry out its function.

In contrast, there are 20–30 trillion erythrocytes constantly circulating in a human body. They are mobile and are distributed throughout the body in close proximity to all tissues. With the combination of rapid water transport and large capacity for volume changes, they are able to function as micropumps to facilitate the restoration of plasma and tissues toward isoosmolarity, regardless of the location and in a very short time. In addition, unlike AVP and other mechanisms of hormonal control (or other cell types that utilize ion loading or unloading mechanisms to control their cell volumes) that require energy, regulation of water by erythrocytes is passively governed by osmotic pressure differences and does not require ATP or other chemical energy. In this context, erythrocytes act as a low-cost stabilizing tool, smoothing out transient and local fluctuations in plasma and interstitial osmolarity (minimizing osmolar shifts), and working in synergy with hormonal mechanisms to reach body water homeostasis.

In conclusion, our integrative in situ, in vitro, and in silico investigations, together with previous works, support the important role erythrocytes play in water transport and homeostasis. We demonstrated that the combination of several crucial properties: rapid water transport, large volume-change capacity, huge quantity, and great mobility make erythrocytes uniquely suited to regulate local osmolarity and participate in water homeostasis. Since water is the most abundant molecule in our body and is essential for survival, its timely and efficient regulation may be considered as the third major function of erythrocytes.

GRANTS

This work was supported by National Heart, Lung, and Blood Institute Cardiovascular Training Grant T32-HL-105373 (to J. Sugie) and Grant RO1-HL-092793 (to L. A. Sung).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

J.S., M.I., and L.A.S. conceived and designed research; J.S. performed experiments; J.S., M.I., and L.A.S. analyzed data; J.S., M.I., and L.A.S. interpreted results of experiments; J.S. and L.A.S. prepared figures; J.S. and L.A.S. drafted manuscript; J.S. and L.A.S. edited and revised manuscript; J.S., M.I., and L.A.S. approved final version of manuscript.