The transport of water across epithelia has remained an enigma ever since it was discovered over 100 years ago that water was transported across the isolated small intestine in the absence of osmotic and hydrostatic pressure gradients. While it is accepted that water transport is linked to solute transport, the actual mechanisms are not well understood. Current dogma holds that active ion transport sets up local osmotic gradients in the spaces between epithelial cells, the lateral intercellular spaces, and this in turn drives water transport by local osmosis. In the case of the small intestine, which in humans absorbs about 8 l of water a day, there is no direct evidence for either local osmosis or aquaporin gene expression in enterocytes. Intestinal water absorption is greatly enhanced by glucose, and this is the basis for oral rehydration therapy in patients with secretory diarrhoea. In our studies of the intestinal brush border Na+-glucose cotransporter we have obtained evidence that there is a direct link between the transport of Na+, glucose and water transport, i.e. there is cotransport of water along with Na+ and sugar, that will account for about 50 % of the total water transport across the human intestinal brush border membrane. In this short review we summarize the evidence for water cotransport and propose how this occurs during the enzymatic turnover of the transporter. This is a general property of cotransporters and so we expect that this may have wider implications in the transport of water and other small polar molecules across cell membranes in animals and plants.
The question of how water is transported across epithelial cells in the absence of external driving forces has intrigued physiologists for over a century, since Reid (1892) demonstrated that water is transported across the intestine in the absence of transepithelial osmotic and hydrostatic pressure differences. Much of the stimulus for investigations of the past 40 years originated with the hypothesis that active Na+ transport was responsible for the generation of local osmotic gradients within epithelial tissues, and that water is coupled to Na+ movement by osmosis. Peter Curran proposed a three-compartment model for intestinal fluid absorption (Curran & McIntosh, 1962), and Jared Diamond proposed a local osmosis model for the gall bladder (Diamond & Bossert, 1967). The local osmotic compartment was not defined in the Curran model, but Diamond proposed that the local osmotic gradient was generated in the lateral intercellular spaces. The latter hypothesis is consistent with the location of the Na+-K+-pumps in the basolateral membrane of absorptive epithelia. Although considerable energy has been devoted to determine the osmolarity of the lateral spaces, especially in the gall bladder, there is little evidence for any significant hypertonicity (see Spring, 1998). One possible explanation is that the hypertonicity is small (< 3 %) and the water permeability of the basolateral membrane is high.
With Peter Agre's discovery of water channels (Preston et al. 1992), the aquaporins, there has been resurgence of interest in epithelial water transport, especially in the kidney. In the intestine there is little, if any, expression of aquaporins in enterocytes and this raises the possibility that the high water permeability of the brush border and basolateral membranes is either due to a high intrinsic water permeability of the lipid bilayer or the presence of other water channels. Yet another possibility is that local osmosis does not account for water transport across the intestine.
There is ample evidence that intestinal water absorption, 8 l a day in man, is intimately linked to ‘active’ solute absorption and that glucose in the gut lumen greatly enhances salt and water absorption. Glucose stimulates sodium transport across the brush border membrane of the enterocytes by Na+-glucose cotransport, and the Na+ that enters the cell is pumped out into the lateral intercellular spaces by the basolateral Na+-K+ pump. It is commonly assumed that water follows sugar and salt transport by local osmosis. These links between glucose, salt and water transport form the basis of the therapy used to treat secretory diarrhoeas such as cholera, oral rehydration therapy (Hirschorn & Greenough, 1991).
Our interest in tackling the enigma of the mechanism of intestinal water transport was rekindled by two independent observations. The first followed our successful cloning and expression of the brush border Na+-glucose cotransporter, SGLT1 (Hediger et al. 1987), when it was discovered that SGLT1 also behaves as a water channel (Zampighi et al. 1995). The second was that water transport across the apical membrane of the choroid plexus was found to be closely linked to the activity of the K+-Cl− cotransporter (Zeuthen 1991, 1994). We then initiated a collaborative study in Los Angeles and Copenhagen to examine more closely the relationship between Na+, glucose and water transport using the cloned transporter (SGLT1) expressed in Xenopus laevis oocytes (Loo et al. 1996, 1999; Zeuthen et al. 1997, 2001; Meinild et al. 1998; Leung et al. 2000). The focus of this brief review is to summarize the evidence that SGLT1 behaves as a water channel and a Na+-glucose-water cotransporter, and to provide an alternative hypothesis for the mechanism of water transport across the intestine in the absence of external osmotic or hydrostatic forces (Reid, 1892).
Our strategy has been to overexpress rabbit or human SGLT1s in Xenopus laevis oocytes, and to determine whether the cotransporter changes the water transport of the oocytes. This offers several unique advantages. (1) Oocytes are large cells, 1 mm in diameter. (2) Native oocytes have very low water permeabilities. (3) While native oocytes do not express SGLTs to any significant level, they do efficiently translate SGLT1 cRNAs injected into the cell and efficiently insert fully functional SGLTs into the plasma membrane, up to 1 × 1011 copies of SGLT1 are found in the plasma membrane (Zampighi et al. 1995). (4) Na+-glucose is electrogenic and so it is simple to measure the rate of Na+-glucose cotransport continuously using electrophysiological methods (Parent et al. 1992a; Loo et al. 1996). There is tight coupling between Na+ and glucose transport, two Na+ ions are transported along with one glucose molecule, and the glucose-activated inward current is a Na+ current (Mackenzie et al. 1998). The cotransporter currents are as high as 2-3 μA, whereas the currents in control oocytes are less than 0.001 μA. (5) It is possible to continuously measure the volume of voltage-clamped oocytes with great precision (0.03 %) and time resolution (1 s) using optical methods (Loo et al. 1996; Zeuthen et al. 1997).
Thus it is straightforward to continuously measure Na+-glucose cotransport and water transport in the same cell with high precision and time resolution. Water transport is measured by monitoring the changes in cross-sectional area of the oocyte, and Na+-glucose cotransport is measured using the two-electrode voltage clamp technique. Na+- glucose cotransport is recorded as a glucose-stimulated inward Na+ current, and cotransport is activated, or inactivated, by (1) the addition or withdrawal of sugar, or (2) changing the membrane potential (Na+-glucose cotransport is highly voltage dependent between 0 and -100 mV; Parent et al. 1992a). In our systems, we can accurately record both water transport and Na+-glucose cotransport in a single cell over a time scale of seconds to minutes.
SGLT1 is a water channel
Expression of SGLT1 in oocytes revealed that the cotransporter behaves as a low conductance water channel (Zampighi et al. 1995; Loo et al. 1996, 1999; Loike et al. 1996; Meinild et al. 1998). The osmotic water permeability (Lp) of SGLT1-expressing oocytes is directly proportional to the number of cotransporters expressed in the membrane with a slope of 1.4 × 10−14 cm s−1, which is about 5 % of that for a single aquaporin (AQP1) molecule. The water permeability of SGLT1 is independent of the size and direction of the osmotic gradient (-100 to +300 mosmol), and for at least 50 s is independent of time of exposure to the osmotic gradient (Zeuthen et al. 2002). The SGLT1 Lp is also independent of the presence or absence of ligands (Na+ and sugar), but phlorizin inhibited water flow through SGLT1 with a Ki of ∼5 μM in the presence of Na+. Phlorizin (100 μM) reduces the Lp of oocytes expressing SGLT1 to that of control oocytes. The activation energy for passive water flow through SGLT1 is 5 kcal mol−1, similar to that for AQP1 but much lower than that for water flow through the native oocyte membrane (14 kcal mol−1). We conclude that SGLT1 behaves as a water channel.
Thus the high density of SGLT1 proteins in the brush border of enterocytes, 250 000 copies per cell, are expected to contribute to the passive fluxes of water and other small polar molecules in and out of the cell. The channel properties of SGLT1 are shared by other cotransporters such as the thyroid Na+-iodide, the brain Na+-Cl−-GABA, the renal Na+-dicarboxylate and plant H+-amino acid cotransporters (Loo et al. 1996; Meinild et al. 2000; Leung et al. 2000).
Cotransport of water
When oocytes expressing SGLT1 are incubated in a sugar-free isotonic solution there is no change in the volume of the oocyte with time. However, when the cotransporter is activated by sugar there is an immediate increase in cell volume. Figure 1 shows one experiment where the solution superfusing an oocyte expressing hSGLT1 is abruptly changed to one containing sugar. Note that the osmolarity of the superfusate was maintained constant by replacing 10 mm mannitol with 10 mmα-methyl-d-glucopyranoside (αMDG, a model non-metabolized SGLT1 substrate). There was an immediate increase in the inward Na+ current through SGLT1 that reached a steady value of 1540 nA within 30 s, and returned towards the baseline when the specific blocker phlorizin was added (Fig. 1A). The volume of the oocyte also increased immediately on activation of the Na+-glucose cotransporter, and phlorizin blocked this volume increase (Fig. 1B). The initial increase in volume of the oocyte can be accounted for if we assume that the transport of two Na+ ions and one glucose molecule is coupled to the transport of 249 water molecules. In this series of experiments 203 ± 20 water molecules were transported for each turnover of the transporter (16 oocytes). In view of the rapid response, we hypothesized that the initial component of the sugar-coupled water flow is due to water cotransport.
What are the properties of sugar-coupled water flow?
Our studies have revealed the following characteristics of the initial rate of sugar coupled-water transport.
(1) There is a stoichiometric relationship between Na+- glucose cotransport and the rate of water transport. The coupling is independent of the level of SGLT1 expression and the rate of Na+-glucose cotransport. In these studies Na+-glucose cotransport was varied by changing the membrane potential, the external Na+ and sugar concentrations, and temperature (Loo et al. 1996; Meinild et al. 1998).
(4) The coupling coefficient is the same when cotransport is driven by either Na+ or H+ gradients (D. D. F. Loo & A.-K. Meinild, unpublished observations).
(5) The coupling coefficient is different for different cotransporters. For example, in one series of experiments the amount of water transport per turnover of the cotransporter varied from 264 for human SGLT1 to 424 for rabbit SGLT1 (Zeuthen et al. 2001), and in another series the water coupling ranged from 50 for a plant H+-amino acid cotransporter (AAP5) to 200 for the rat thyroid Na+-iodide cotransporter (NIS) (Loo et al. 1996).
Osmosis and Na+-glucose cotransport
The studies outlined above clearly demonstrate that there is a close link between Na+-glucose cotransport and the initial rate of water transport. Furthermore, the stoichiometry between Na+, glucose and water transport immediately after activation of hSGLT1 indicates that the transporter fluid is hypertonic to the saline bathing the oocyte, i.e. ∼100 water molecules per transported solute through hSGLT1 compared with 258 water molecules per solute in the bathing solution. This means that the intracellular fluid will become hypertonic as transport proceeds, and this is expected to draw more water into the cell by osmosis. Figure 2 shows an experiment where Na+-glucose cotransport was maintained for over 5 min. When transport was switched on by the addition of sugar the volume of the cell increased within 2 s at an initial rate of 13 pl s−1 (Δ1, Fig. 2), equivalent to a coupling ratio of 269 ± 12 (n= 9) water molecule per turnover of the protein. As cotransport continued there was a slow increase in the rate of volume increase until a steady state is reached some 3-5 min after Na+-glucose cotransport was initiated. The steady-state rate was on average 3 times the initial rate and this remained constant for at least another 10 min. Note that in these experiments the sugar concentration was reduced from 10 to 1 mm and the lower transport rate, ∼600 nA, declined by less than 15 % over the duration of the experiment. As expected, in the steady state, fluid transport was isotonic (775 water molecules per turnover).
However, if cotransport is rapidly switched off, by removing sugar, or by depolarizing the membrane potential, there is an immediate (within 1 s) decrease in the rate of volume increase of the oocyte (Δ2, Fig. 2). In five hSGLT1 oocytes the initial increase in water transport with turning cotransport on by the addition of 1 mm sugar was 15 ± 2 pl s−1 (inward current 550 ± 74 nA), and after reaching a steady state (41 ± 6 pl s−1) the initial reduction of water transport on removing sugar was 14.5 ± 1 pl s−1, i.e. the initial rates of volume change of the oocyte were identical on switching the cotransport on and off by the addition and removal of sugar (Zeuthen et al. 2001). We estimate that in these five oocytes cotransport accounts for ∼35 % of water transport in the steady state and that osmosis accounts for the remainder (≈35 % by osmosis through SGLT1 and ∼30 % by osmosis through the plasma membrane).
Phlorizin cannot be used to specifically inhibit the water cotransport component of the steady-state water flow as it also blocks osmosis through SGLT1 (see above). In this series of experiments phlorizin initially reduced the steady-state flow by 70 %, i.e. twice as much as removing glucose.
Differentiation between water cotransport and osmosis
We attempted to dissociate the glucose-stimulated water transport from osmosis by measuring the initial rate of water transport before Na+-glucose cotransport produces any significant change in intracellular osmolarity. The simplest way to do so is to change the rate of Na+-glucose cotransport by stepping the membrane potential within a few milliseconds using the two-electrode voltage clamp technique. Figure 3A illustrates one experiment with rbSGLT1 where the oocyte was superfused with saline containing 5 mmαMDG and the membrane potential was initially held at 0 mV. After stepping the membrane potential to -100 mV the transporter current increased immediately from 280 to 950 nA and the water transport rate increased within seconds from 9 to 40 pl s−1. Figure 3B illustrates another experiment where the membrane potential was initially held at -80 mV and then depolarized to 0 mV. The cotransporter current immediately decreased from 1100 to ∼350 nA and the water flow decreased from 48 to 13 pl s−1 within 1.1 ± 0.7 s (n= 18). In these experiments the number of water molecules transported per turnover of the cotransporter was identical to that obtained by adding or withdrawing sugar from the perfusate (Zeuthen et al. 2001).
Our interpretation of these results is that there is cotransport of water through SGLT1, and that osmosis only becomes a contributing factor after sufficient time has lapsed to build up an intracellular osmotic gradient. Additional evidence in support of this hypothesis is as follows.
(1) The rate of water cotransport is independent of the osmotic gradient across the membrane, and can even occur uphill against an osmotic gradient (Loo et al. 1996; Meinild et al. 1998).
(3) The stoichiometry for SGLT1 is constant when the rate of transport varies by over an order of magnitude but varies considerably from cotransporter to cotransporter. The coupling coefficients range from 50 for the plant H+-amino acid cotransporter, to 176 for the renal Na+-dicarboxylate cotransporter, to 210 for the human Na+-glucose cotransporter, and to as high as 424 for the rabbit Na+-glucose cotransporter. This is not expected if the coupling between water transport and solute cotransport is purely osmotic.
(4) Finally, urea is transported by SGLT1 and other cotransporters such as the low affinity Na+-glucose, the Na+-iodide and the Na+-Cl−-GABA cotransporters (Leung et al. 2000). Although urea is transported through the cotransporter water channel (see above), the rate of urea transport increases when the cotransporter is activated by the addition of substrates. To a first approximation, the ratio of urea to water transport is in the same range as their concentrations in the saline bathing the oocytes. It is unlikely that the sugar-activated urea transport through SGLT1 is due to solvent drag as the reflection coefficient for urea is close to 1 (see Zeuthen et al. 2001).
How is water transport coupled to cotransport?
We propose that water cotransport is simply a consequence of how sodium and glucose transport through SGLT1 are coupled. Our kinetic model for SGLT1 incorporating water cotransport is shown in Fig. 4. This six-state alternating access model accounts for all the known characteristics of SGLT1. Na+-coupled sugar transport is envisaged as a series of conformational changes induced by the binding of ligands (see Parent et al. 1992b; Loo et al. 1998; Meinild et al. 2002).
SGLT1 is an integral protein with an apparent valence of -2. At a membrane potential of -50 mV and in the absence of Na+, all of the transporters are in the C6 form. Addition of saturating Na+ to the external solution results in two Na+ ions binding to the protein and a change in formation to the C2 form (80 %). In the absence of external sugar, Na+ is transported at a low rate (turnover 5 s−1, uniport mode). In the presence of external sugar the open C2 conformation permits sugar to bind to form C3, and the transport cycle proceeds to C4, where sugar dissociates from SGLT1 at the internal membrane surface, followed by Na+ dissociation due to the low intracellular Na+ concentration. The protein then returns to the original conformation C1 to complete the transport cycle. In the presence of saturating external Na+ and sugar, the turnover rate of human SGLT1 transport cycle is ∼60 s−1 at 22 °C. This ordered reaction mechanism accounts for the strict coupling between Na+ and sugar transport and the kinetics of Na+ uniport in the absence of sugar.
According to this scheme water and urea are cotransported along with Na+ and glucose simply as a consequence of the asymmetrical conformational changes that occur during the transport cycle. When external Na+ binds and produces the conformational change that permits high affinity sugar binding (C2 to C3), water and urea ‘bind’ to the open sugar binding pocket and pass to the cytosolic surface along with Na+ and glucose translocation (C3 to C4). Sugar and Na+ then dissociate from SGLT1, and the protein returns to the closed conformation (C6), extruding water and urea to the cytoplasm. Finally, SGLT1 in the closed conformation (C6) returns to the original state (C1) without any significant water movement. The ‘binding’ of more than 200 water molecules to the sugar-binding pocket is quite reasonable given that SGLT1 can transport glucosides with a molecular volume of 1700 Å3 (20 × 12 × 75 Å, Lostao et al. 1994; Hirayama et al. 2001). The volume occupied by 200-400 water molecules (6000-12 000 Å3) only amounts to 7 % of the volume of the cotransporter (100 000 Å3). The driving force for coupled transport of Na+, glucose and water across the membrane through SGLT1 is simply the sum of the electrochemical potential gradients across the membrane for these solutes. Normally it is the Na+ gradients that provide the energy for uphill sugar and water cotransport.
Physiological implications of water pumps
These studies cast some light on the mechanism of water transport across the brush border membrane of enterocytes, and across the epithelium. The human intestine absorbs about 1 mol of d-glucose and 8 l of water per day and given that the stoichiometry of hSGLT1 is 2 Na+:1 glucose:264 water, this indicates that cotransport alone can account for ∼4 l of water transport. Since the cotransported fluid is hypertonic this will result in an additional component due to osmosis, but the exact amount is difficult to estimate due to uncertainties about the magnitude of the osmotic gradient and the water permeability of the membrane. These links between sodium, glucose and water transport across the brush border membrane rationalize the use of oral salt solutions containing glucose to treat patients with secretory diarrhoea, i.e. oral rehydration therapy. How is water transported across the basolateral membrane of the enterocyte in the absence of aquaporins and a hypertonic fluid in the lateral intercellular spaces? One possibility is that basolateral membrane transport proteins act as water pumps and channels. We have demonstrated that cotransporters in general behave as water channels and water cotransporters, which supports the notion that cotransporters such as the K+-Cl− cotransporter play an important role in water exit from the cell (see also Zeuthen, 2000, 2002). The exact role will depend on both the type and density of cotransporters present in the basolateral membrane. Finally, it is possible that other transporters such as the exchangers and ATPases also transport water as a direct consequence of conformational changes.
The Na+-glucose cotransporter behaves as a low conductance water channel, and there is a sugar-coupled water flow that couples Na+, sugar and water stoichiometrically. In the steady state, direct sugar-coupled water flow constitutes one-third of the total water transport in isotonic water flow, the remaining being osmotic water flow. The sugar-coupled water flow or water cotransport, is distinguished from osmotic water flow by the instantaneous response to step jumps in the rate of Na+-glucose cotransport. Recently, Duquette et al. (2001) repeated our experiments on hSGLT1 and many of their results agree with ours. Unfortunately, these authors did not attempt to resolve the initial rates of water flow owing to technical difficulties. Under the assumption of extremely low intracellular mobilities (Zeuthen et al. 2002), and on the basis of steady-state flows alone, they concluded that water flow was linked to Na+-glucose cotransport by simple osmosis. As we have noted above, we concluded that 65 % of the steady-state water flow was osmotic. More definitive experiments are needed to confirm or reject our hypothesis that water cotransport accounts for 35 % of the steady-state flow. One obvious approach is to measure the thermodynamics of Na+, sugar and water transport. However this is difficult since the osmotic gradients required to influence Na+-sugar cotransport are large. Use of the Gibbs equation predicts that osmotic gradients of 100 mosmol l−1 will shift the reversal potential by only about 5 mV, a prediction supported by preliminary experiments (Zeuthen & MacAulay, 2002). Even with these experiments the interpretation will be clouded by the same ‘unstirred layer’ questions that bedeviled earlier studies of water transport. Ultimately the question will be resolved when the 3-D structure of SGLT1 is obtained. This may not be in the too distant future given current progress in isolating sufficient quantities of recombinant SGLTs for structural studies (Turk et al. 2000; Quick & Wright, 2002).
This work was supported by funds from the NIH (DK19567), USDA (99-35304-7975), the Danish Research Council, the Lundbeck Foundation, and the Novo-Nordisk Foundation. We gratefully acknowledge the contributions of colleagues to these studies on SGLT1: Drs G. Chandy, B. A. Hirayama, D. Klaerke, D. Leung, A.-K. Meinild, M. Panayotova-Heiermann and G. Zampighi.
Presented at The Journal of Physiology Synthesium on Water Transport Controversies, Christchurch, New Zealand, 30 August, 2001.