Plant aquaporins: multifunctional water and solute channels with expanding roles


  • S. D. Tyerman,

    Search for more papers by this author
    • *

      Present address: Department of Horticulture Viticulture and Oenology, Plant Research Centre, Adelaide University, Waite Campus, PMB#1 Glen Osmond SA 5064, Australia.

  • C. M. Niemietz,

    Search for more papers by this author
    • *

      Present address: Department of Horticulture Viticulture and Oenology, Plant Research Centre, Adelaide University, Waite Campus, PMB#1 Glen Osmond SA 5064, Australia.

  • H. Bramley

    Search for more papers by this author
    • *

      Present address: Department of Horticulture Viticulture and Oenology, Plant Research Centre, Adelaide University, Waite Campus, PMB#1 Glen Osmond SA 5064, Australia.

Correspondence: Stephen D. Tyerman. E-mail:


There is strong evidence that aquaporins are central components in plant water relations. Plant species possess more aquaporin genes than species from other kingdoms. According to sequence similarities, four major groups have been identified, which can be further divided into subgroups that may correspond to localization and transport selectivity. They may be involved in compatible solute distribution, gas-transfer (CO2, NH3) as well as in micronutrient uptake (boric acid). Recent advances in determining the structure of some aquaporins gives further details on the mechanism of selectivity. Gating behaviour of aquaporins is poorly understood but evidence is mounting that phosphorylation, pH, pCa and osmotic gradients can affect water channel activity. Aquaporins are enriched in zones of fast cell division and expansion, or in areas where water flow or solute flux density would be expected to be high. This includes biotrophic interfaces between plants and parasites, between plants and symbiotic bacteria or fungi, and between germinating pollen and stigma. On a cellular level aquaporin clusters have been identified in some membranes. There is also a possibility that aquaporins in the endoplasmic reticulum may function in symplasmic transport if water can flow from cell to cell via the desmotubules in plasmodesmata. Functional characterization of aquaporins in the native membrane has raised doubt about the conclusiveness of expression patterns alone and need to be conducted in parallel. The challenge will be to elucidate gating on a molecular level and cellular level and to tie those findings into plant water relations on a macroscopic scale where various flow pathways need to be considered.


Water uptake and flow across membranes is a fundamental requirement for plant growth, both at the cellular level and at the whole plant level, as a consequence of gas exchange. Hence it is highly significant that in the last decade we have learnt that water could cross plant membranes through proteinaceous channels formed by members of the aquaporin superfamily. To fix one kilogram of carbon, terrestrial plants typically transport several hundred kilograms of water. If we assume that at least 50% of transpired water in the biosphere passes through plant cell membranes in an overall composite ‘plant epithelium’, then it could be argued that plant aquaporins transport the highest total mass of a substance compared to any other class of membrane transport. This large-scale significance may be reflected in inhibition of transpiration and needle growth after spraying pine tree stands with the aquaporin antagonist phenylmercuric acetate (Turner & Waggoner 1968), although unspecific inhibition of other enzymes may have contributed to the observed effect (e.g. Loos & Lüttge 1984 and discussed in Tyerman et al. 1999).

The aquaporin superfamily consists of hydrophobic proteins with six membrane-spanning alpha helixes that have molecular weights in the range of 26–34 kDa (see Structure below). The family is also referred to as the major intrinsic protein (MIP) superfamily based on the first sequenced member, the Major Intrinsic Protein of bovine lens cells (Gorin et al. 1984), now called AQP0. Aquaporin function was first identified in an homologous protein from human red cells, CHIP28 (now called AQP1), based upon its ability to dramatically increase water permeability of Xenopus oocytes expressing the mRNA of the CHIP28 gene (Preston et al. 1992). The discovery of a tonoplast integral protein (γTIP) that functions as an aquaporin (Maurel et al. 1993) triggered the discovery of further plant aquaporin genes, functional studies of aquaporins, and a less mechanical view of plant water relations that has been extensively reviewed in the last 3 years (Schäffner 1998; Tyerman et al. 1999; Kaldenhoff & Eckert 1999; Kjellbom et al. 1999; Johansson et al. 2000; Santoni et al. 2000; Maurel & Chrispeels 2001; Chrispeels et al. 2001; Maurel et al. 2001). Aquaporins are common to most organisms and may function as water-selective channels (aquaporins, proper), or relatively non-selective channels for water and other small non-electrolytes (aquaglyceroporins) (Agre, Bonhivers & Borgnia 1998; Borgnia & Agre 2001). Under certain circumstances some MIPs may also allow ion permeation (Weaver et al. 1994; Lee et al. 1995; Yasui et al. 1999).

The discovery of aquaporins in plants has resulted in a paradigm shift in the understanding of plant water relations (Maurel & Chrispeels 2001). This is because water flow across plant membranes in the past was considered to be relatively invariant depending mainly on lipid properties. The connection with membrane proteins was not made until 1990 (Wayne & Tazawa 1990). Before that time the hypothetical pathway for water flow across plant membranes included the possible existence of pores (Dainty 1963), as evidenced by low activation energies (Dainty & Ginzburg 1964a), high frictional interactions (e.g. Dainty & Ginzburg 1964b; Steudle & Tyerman 1983) and from osmotic effects on water permeability, but not ion permeability (Kiyosawa & Tazawa 1987). A specific test of flow through pores is a high ratio of osmotic to diffusional water permeability (Pos/Pd) (Dainty 1963) reflecting either the effect of pore geometry on viscous versus diffusional flow, or the effect of filing of molecules in a narrow pore (Hill 1995). This was tested in several cases ‘preaquaporins’ but a question mark always stood over high values of Pos/Pd because of the greater effect of unstirred layers on Pd compared to Pos (Dainty 1963). Notwithstanding this difficulty, water movement through a pore-less lipid bilayer could not account for the high water permeabilities observed for the Characeae.

The most important contribution of the new paradigm is that permeation of water through a proteinaceous pore extends the potential for control over water flow by living cells. This has been incorporated into the composite transport model of water flow across roots (Steudle 1994), where the role of aquaporins has been suggested to regulate the ‘cell-to-cell’ pathway, which dominates when water flow is governed primarily by osmotic gradients or when the apoplastic pathway is blocked by suberization or lignification (Steudle 2000; Hose, Steudle & Hartung 2000; Barrowclough, Peterson & Steudle 2000). These conditions are more likely to occur under water stress.

Diversity of plant aquaporins

The wide diversity of MIPs in plants was first signalled when Weig, Deswarte & Chrispeels (1997) identified 23 different MIP genes in Arabidopsis that clustered into three groups. This has been extended to over 35 MIPs within four groups in Arabidopsis (Johanson et al. 2001) and at least 31 MIPs in maize that also occur within four distinct groups (Chaumont et al. 2001). One group corresponds to location in the tonoplast (Tonoplast Integral Proteins, TIPs; Johnson, Höfte & Chrispeels 1990; Höfte et al. 1992; Daniels et al. 1996; Karlsson et al. 2000) and another to the plasma membrane (PIPs) (Daniels, Mirkov & Chrispeels 1994; Kammerloher et al. 1994; Robinson et al. 1996; Johansson et al. 1996; Barone, Shih & Wasserman 1997; Chaumont et al. 2000a; Frangne et al. 2001; Suga, Imagawa & Maeshima 2001). The PIPs separate out into two subgroups (Weig et al. 1997; Tyerman et al. 1999; Chaumont et al. 2001). The two subgroups correlate with high (PIP2) and low (PIP1) water channel activity (Chaumont et  al. 2000a). However, there are some exceptions where water channel activity has been observed for members of the PIP1 subgroup; Nt-AQP1 (Biela et al. 1999) and Bo-PIP1b1 and Bo-PIP1b2 from Brassica oleracea (Marin-Olivier et al. 2000).

The assignment of plant MIPs to an exclusive membrane site based on sequence data should be taken as putative because some PIP antibodies reacted with purified plasma membrane and tonoplast (Barkla et al. 1999). When membranes from Mesembryanthemum crystallinum were separated on a sucrose gradient, antibodies to three PIPs (MIPA, MIPB and MIPC) reacted with a range of membrane fractions including that identified as tonoplast (Kirch et al. 2000).

The nodulin-like MIPs (NIPs) form a third distinct group based on sequence for which the membrane location is not known in maize and Arabidopsis. The archetype NIP, NOD 26 is located in the peribacteroid membrane (PBM) of soybean nodule cells (Fortin, Morrison & Verma 1987; Weaver et al. 1991), and LIMP2, the NOD 26 orthologue in Lotus japonicus, is also probably located in the PBM (Guenther & Roberts 2000). The fourth major group in both Arabidopsis and maize consists of small basic MIPs (SIPs) (Urban Johanson, MIP 2000 conference, July 1–5, 2000 Göteborg, Sweden; Chaumont et al. 2001). The SIPs are quite divergent from the other plant MIPs and show significant sequence differences in loop B where the Asn-Pro-Ala (NPA) box consists of Asn-Pro-Thr or Asn-Pro-Leu (Chaumont et al. 2001). Given the importance of residues in loop B for transport characteristics (Murata et al. 2000; Fu et al. 2000) it is likely that the SIPs could display quite different transport selectivity.

The diversity of MIPs in plants contrasts with only 10 different MIPs in the mouse and in humans that cluster into two groups (Ishibashi, Imai & Sasaki 2000). The diversity in plants may indicate a wider range of function that may include a greater range in selectivity (Santoni et al. 2000).

Selectivity of transport

MIPs have been classified on the basis of their water and solute permeability and with respect to the later glycerol permeability has been taken as a key discriminator. This stems in part from the fact that some MIPs were originally identified as selective glycerol facilitators (e.g. GlpF, Maurel et al. 1994). However, even GlpF is permeable to water (Borgnia & Agre 2001). Other MIPs are permeable to both glycerol and water (aquaglyceroporins). This includes NOD 26 (Rivers et al. 1997; Dean et al. 1999), and its ortholog LIMP2 from Lotus japonicus nodules (Guenther & Roberts 2000). Two NIPs (NLM1 and NLM2) from Arabidopsis are also glycerol facilitators (Weig & Jakob 2000a) and NLM1 was already known for its water channel activity (Weig et al. 1997). In terms of the major MIP groups in plants, there is at least one aquaglyceroporin in all plant MIP groups so far examined (excluding SIPs): TIP (Nt-TIPa; Gerbeau et al. 1999), PIP (Nt-AQP1 Biela et al. 1999); NIP (Dean et al. 1999). There are as yet no exclusive glycerol facilitators discovered in plants. Glycerol is not known to be an osmoticum in the infected cells of nodules where NOD26 is situated in the peribacteroid membranes. However, glycerol can be produced in plants and transported to leaves when plants are flooded or under anoxic stress (Gerber et al. 1988). It has been suggested that the role of NLMs, which are predominantly expressed in roots, be assessed with respect to anoxic or hypoxic stress (Weig & Jakob 2000b). It will be important to test the permeation of solutes in context with possible functional roles related to the site of MIP expression. This may include testing permeation of some gases that can be functionally important in the particular system where the MIP occurs.

Permeability to gases

Carbon dioxide

The possibility that CO2 may permeate through AQP1 has been examined because AQP1 is expressed in mammalian cells where CO2 transport is important (red cells and lung epithelium). Studies on oocytes expressing AQP1 (Nakhoul et al. 1998; Cooper & Boron 1998) and on phospholipid vesicles with reincorporated AQP1 (Prasad et al. 1998) indicated that CO2 could permeate via AQP1. However, two recent studies have cast doubt over this interpretation. Erythrocytes of AQP-1 null mice showed the same CO2 permeability as wild type (Yang, Brown & Verkman 2000). These authors also demonstrated that carbonic anhydrase, which is used to accelerate the pH changes associated with CO2 fluxes in some of the previous studies was inhibited by HgCl2. However, this does not explain the results of Cooper & Boron (1998) who used p-chloromercuribenzenesulfonic acid (PCMBS), which is membrane impermeant, did not use carbonic anhydrase, and demonstrated that a mercurial insensitive mutant of AQP1 enhanced CO2 induced acidification independent of the presence of PCMBS. Another study has examined the effect on CO2 permeability of different levels of expression of AQP1 in corneal endothelial cells (Sun et al. 2001). Although the water permeability changed, CO2 permeability did not and no effect of PCMBS on CO2 transport was observed. The degree to which aquaporins may facilitate CO2 permeation will depend on the permeability coefficient for CO2 in the lipid pathway. Since CO2 has a high oil-water partition coefficient, then for most membranes the permeability will be governed more by diffusion through the unstirred layers around the membrane than by the membrane itself (e.g. Gutknecht, Bisson & Tosteson 1977; Yang et al. 2000).

The possibility that CO2 may permeate AQPs in plants has not been examined yet. It is an important issue since the resistance to CO2 diffusion within C3 leaves to the sites of fixation in the stroma of chloroplasts is important in determining the concentration of CO2 that the RIBISCO enzyme is exposed to. The greater the diffusive resistances the lower the concentration of CO2 within the stroma will be to sustain the required flux. Since Rubisco has a relatively low affinity for CO2 it would be preferable for there to be an overall low resistance to internal diffusion of CO2. The internal resistance to CO2 diffusion will ultimately affect the degree of photorespiration, the nitrogen use efficiency and water use efficiency of the leaf. It has been noted that the exchange rate of CO2 across the interfaces in mammalian lung and leaves are similar, but the rapid rate in leaves is achieved with a much lower concentration gradient; thus the leaf interfaces have an overall high conductance to CO2 (Evans & von Caemmerer 1996). The high conductance could result from high CO2 permeabilities of the plasma membrane and chloroplast outer membranes, but the actual CO2 permeabilities of higher plant membranes or the basis of apparent high permeabilities is unknown.

Chlamydomonas reinhardtii is a unicellular alga that has a CO2 concentrating mechanism. The plasma membrane permeability coefficient for CO2 is rather low, being between 12 and 18 µm s−1, and is not significantly affected by the degree of inorganic carbon accumulation (Sültemeyer & Rinast 1996). This PCO2 is some 100 times less than the values estimated for lipid bilayer membranes (Gutknecht et al. 1977). Despite the low PCO2Sültemeyer & Rinast (1996) conclude that no substantial gradient of CO2 concentration exists across the plasma membrane based on the energy demands to maintain the gradient and other observations. Chlamydomonas also lacks turgor and has a low Pos for the plasma membrane, which would be a requirement for the energy efficiency of active water transport via the contractile vacuole (Raven 1982, 1995). This led to the suggestion that Chlamydomonas may lack aquaporins (Raven 1995).

In this respect it is interesting that in Chara there was shown to be a positive correlation between CO2 permeability and water permeability resulting from growing cells in either high CO2 or low CO2 (Wayne, Mimura & Shimmen 1994). Water permeability was not associated with HCO3 transport. Wayne et al. (1994) suggest that either the lipid composition was changed affecting both water and CO2 permeability or both molecules move through the same transport protein. Perhaps water flow into cells via arrays of aquaporins sweeps CO2 to and from the surface of the membrane reducing the unstirred layer limitation. However, in higher plant leaves, water and CO2 may well move in opposite directions across mesophyll cell membranes, in which case if they both moved through the same aquaporin there could be a negative effect on CO2 diffusion. On the other hand could there be some sort of coupling like an antiporter? It has been suggested from the angle of the pore in AQP1 monomers that the tetramer rotates about its central axis when water flow occurs (Verkman & Mitra 2000). Could this rotation and the disturbance created in adjacent lipid molecules influence gas permeation?

One of the most abundant proteins in the spinach leaf plasma membrane is the plasma membrane located MIP (PM28A) (Johansson et al. 1996). Low water potential reduces phosphorylation of PM28A in spinach leaf plasma membrane. Phosphorylation is carried out by a Ca2+-dependent membrane-bound protein kinase (Johansson et al. 1996). When expressed in oocytes, PM28A can also be phosphorylated, and decreased phosphorylation reduces the water permeability of oocytes expressing PM28A (Johansson et al. 1998). Taken together, this seminal work suggests that PM28A is important in regulating water flow through leaf tissue, but the possibility also exists that like AQP1, PM28A may also be capable of enhancing CO2 permeation, thus contributing to the overall high apparent CO2 permeability of leaf cells. Low leaf water potential is known to reduce the fixation of CO2 but the mechanism is unknown.


Ammonia (NH3) and ammonium (NH4+) fluxes across plant membranes occur during processes of nitrogen assimilation in roots and leaves, and as a result of photorespiration in C3 leaves. Nitrogen fixation in legumes requires that NH3/NH4+ is transferred across the peribacteroid membrane between symbiotic bacteria and the cytoplasm of host cells (Day et al. 2001). Nodulin 26 densely studs this membrane and has been suggested to have various roles in the symbiotic partnership including malate transport (Weaver et al. 1994) and osmoregulation (Dean et al. 1999). Ammonia gas (NH3) permeation across the symbiosome membrane was examined and found to be protein mediated, possibly via NOD 26 (Niemietz & Tyerman 2000). The permeation of NH3 was reduced by about 40% with HgCl2 and the activation energy was more than doubled. The activation energy however, was higher than would be expected for permeation being dominated by an open pore (52 kJ mol−1 or 12 kcal mol−1), but is consistent with a combination of permeation via the lipid and via channels. Furthermore, the dose–response to HgCl2 showed greater sensitivity for NH3 permeation compared to H2O. These differences do not necessarily mean that NH3 does not permeate via NOD26. Such differences between solute and water transport through the same MIP (AQP3) have been observed before (Echevarria, Windhager & Frindt 1996) and it was proposed that perhaps different pathways through the MIP exist, but other evidence suggests that this may not be the case (Verkman & Mitra 2000). Alternatively, NH3 may permeate the same pore as water but may not have such an easy passage, i.e. there may be higher energy-barriers to overcome. For this reason also slight conformational changes induced by low concentrations of HgCl2 (see Barone et al. 1997) may have a bigger effect on NH3 permeation.

Despite the evidence for channel mediated permeation of NH3 through the PBM, the overall NH3 permeability of the peribacteroid membrane (80 µm s−1) was rather low compared to other cell membranes, e.g. the erythrocyte membrane (2100 µm s−1) where it has been shown that NH3 does not permeate via CHIP28 (AQP1) (Zeidel et al. 1994). Membrane lipid composition has a large influence on NH3 permeation (Priver, Rabon & Zeidel 1993; Lande, Donovan & Zeidel 1995; Negrete et al. 1996) and one could expect that for channel-mediated NH3 permeation to have most effect in terms of control, the basal lipid permeability should be low. Control is the key feature of our new paradigm. In this respect NOD26 is the major phospho-protein in the PBM and that phosphorylation is carried out by a membrane associated Ca2+ -dependent protein kinase (Weaver et al. 1994). Phosphorylation of NOD26 can induce changes in ion channel activity that it displays in artificial bilayers (Lee et al. 1995). Therefore Niemietz & Tyerman (2000) tested the effect of ATP pre-incubation on NH3 and H2O transport. It was found that ATP inhibited NH3 permeation but stimulated H2O permeation. The inhibition of NH3 permeation makes sense in the context of NH3 release from the symbiosome where under energized conditions (high proton pump activity acidifying the inside of the symbiosome) the NH4+ cation channel can account for most of the nitrogen release (Tyerman, Whitehead & Day 1995). Under these circumstances there is the possibility that NH3 could leak back into the symbiosome because of the pH gradients across the membrane (acid inside and slightly alkaline in the cytoplasm), so a low NH3 permeability (channels gated closed to NH3) would be advantageous. When the membrane is not energized as a result of reduced ATP availability the inhibition may be released from the channel pathway for NH3 and this could then account for the majority of released nitrogen.

To conclusively test that NH3 can permeate through NOD 26, purification of the protein and reincorporation into phospholipid vesicles is required (e.g. Dean et al. 1999). These experiments are currently underway. Another way would be to use oocytes to express NOD26 and then test NH3/NH4+ permeability by examining changes in cytosolic pH when the external NH3/NH4+ concentration is altered. A preliminary report by Cooper et al. (2000) indicates that NOD26 may be permeable to both NH3 and NH4+. The increased NH4+ permeability is very interesting as patch clamp experiments on the PBM have demonstrated in two legume species (soybean and Lotus japonicus) that there is a very low conductance (sub-picoSiemens) channel that allows NH4+ permeation in this membrane (Tyerman et al. 1995; Whitehead, Day & Tyerman 1998; Roberts & Tyerman 2001). A feature of the channel is that it appears to occur at very high density estimated at about 1000 channels per µm2 (Tyerman et al. 1995). This density is typical of MIP densities in membranes (Verkman & Mitra 2000) and is actually less than that estimated by Rivers et al. (1997) for NOD26 (approximately 150 000 per µm2). NOD 26 has also displayed ion channel activity when incorporated in artificial lipid bilayers but in this case they displayed slight anion selectivity and had very large conductances (Weaver et al. 1994; Lee et al. 1995). These characteristics are not typically observed for patch-clamped PBM membranes.

Ammonia gas is also produced in leaves through a variety of mechanisms that include photorespiration, NH4+ uptake by roots (Mattsson & Schjoerring 1996), NO3 reduction, deamination and protein degradation (Temple, Vance & Gantt 1998). NH3 can be lost from leaves when the atmospheric concentration is less than the NH3 compensation point (Farquhar et al. 1980). The compensation point reflects a variety of processes including the degree of photorespiratory production of NH3, NH3 assimilation mainly by GOGAT (Temple et al. 1998), the permeation of NH3 across the plasma membrane, the pH and NH4+ concentration in the leaf apoplast (Mattsson, Husted & Schjoerring 1998), and NH4+ uptake from the apoplast (Raven & Farquhar 1981), which can have high affinity for NH4+ (Ninnemann, Jauniaux & Frommer 1994). A low NH3 permeability of leaf mesophyll plasma membranes would result in less energy used in NH4+ scavenger transport. Therefore, MIPs situated in leaf mesophyll plasma membrane might be expected to have high selectivity and not to transport NH3.

Other important non-ectrolytes that may be transported by aquaporins

Other molecules relevant to plant function, for which there is evidence for permeation via MIPs in plants and other organisms, include boron (Dordas, Chrispeels & Brown 2000), antimonite (Sanders et al. 1997), H2O2 (Henzler & Steudle 2000), small alcohols (Hertel & Steudle 1997; Schütz & Tyerman 1997) and compatible solutes (Gerbeau et al. 1999). It is also possible that un-dissociated organic acids may permeate via MIPs and this could be important for release of organic acids to the cytosol from acidic compartments or transport of organic acids across biotrophic interfaces. Recently Chen et al. (2001) demonstrated that inhibiting expression of tomato ripening-associated membrane protein (TRAMP) increased the organic acid content and decreased sugar content of fruit. They did not observe any phenotype that could be associated with altered water relations, which is in contrast to the effects of reduced expression of PIPs in Arabidopsis (Kaldenhoff et al. 1998). Chen et al. (2001) proposed that the TRAMP protein may be located in the tonoplast and be involved in organic acid and sugar transport between the cytosol and vacuole.


Boron is an essential nutrient and is taken up by plant roots as boric acid (H3BO3), which is uncharged at normal pHs (Loomis & Durst 1992). It can also be in excess in some soils and cause toxicity. There are significant differences in genotypes in terms of boron uptake and also in tolerance to high soil concentrations. Dordas & Brown (2000) first established the effect of lipid bilayer composition on boron permeability. Using Arabidopsis mutants differing in sterol composition and length of fatty acid tails, they found up to two-fold differences in permeability. Dordas et al. (2000) then investigated the possible permeation via aquaporins. Plasma membrane vesicles isolated from squash roots displayed relatively high boron permeability compared with microsomal vesicles, although the value (3 × 10−3 µm s−1) was relatively low compared with other small solutes, and this was reversibly inhibited by HgCl2 and phloretin. The activation energy for boron transport was rather high and similar to that for NH3 permeation in the peribacteroid membrane (Tyerman & Niemietz 2000). This could be for the reasons already explained above. A similar boric acid permeability was estimated by Nuttall (2000) for wheat root cells and plasma membrane vesicles (less than 6 × 10−3 µm s−1). Recently Dordas & Brown (2001) have demonstrated a reversible HgCl2 inhibition of boric acid uptake into intact  squash roots. They also showed that various non-electrolytes of similar size to boric acid inhibited boric acid uptake. To establish that some known MIPs may be permeable to boron Dordas et al. (2000) also demonstrated that Zm-PIP1 could slightly (but significantly) stimulate boron uptake into oocytes, but that At-NLM1 Zm-PIP3 and GlpF did not. Experiments on oocytes expressing At-PIP1b, At-PIP2a and At-PIP2b by Nuttall (2000) also suggest that boric acid could permeate some PIP aquaporins. These are very interesting results since Zm-PIP1 expressed on its own has not been demonstrated to increase the permeability to any other solute (at least not by itself), whereas NLM1 is non-selective and GlpF is selective for glycerol. Apart from the potential for passive uptake in roots through PIPs, boric acid could also be taken up by a high affinity transporter as has been shown to occur in Chara (Stangoulis et al. 2001).

Hydrogen peroxide

Hydrogen peroxide is an important by-product of photorespiration and is also used in plant defence reactions. Henzler & Steudle (2000) investigated the H2O2 permeability of Chara plasma membranes using the pressure probe and analysis of the breakdown reaction in the cell. They were able to demonstrate that H2O2 permeability was inhibited by HgCl2 and suggested that perhaps some aquaporins functioned as ‘peroxoporins’. Using the pressure probe Henzler & Steudle (2000) needed to use fairly high and unphysiological concentrations of H2O2. It would be worthwhile in the future to assess H2O2 permeation in a vesicle system (notwithstanding the associated problems, see below) incorporating a dye that is sensitive to H2O2 at lower concentrations. Subsequently Henzler has reported that water channels seem to be blocked by inducing the Fenton reaction to produce OH radicles (T. Henzler, MIP 2000 conference). This is a potentially powerful tool as it offers another way to block water channels. However the contribution from lipid peroxidation needs to be assessed.


The general structure of aquaporins has been reviewed recently (Verkman & Mitra 2000; Engel, Fijiyoshi & Agre 2000) with most information coming from studies on AQP1 that can be easily purified from red cells. Aquaporins are typically predicted to contain 6 transmembrane α-helices (helix 1–6), with the N- and C- termini located on the cytoplasmic side. There are five loops (A–E) joining the transmembrane helices. Two of the loops (B and E) are hydrophobic each consisting of a small helix. At the end of each small helix is the highly conserved asparagine–proline–alanine (NPA) sequence. Each half of the polypeptide (3·5 helixes) shows significant sequence similarity. The two halves of the polypeptide show obverse symmetry, with the hydrophobic loops containing the NPA overlapping in the middle of the lipid bilayer to form two hemipores that together create a narrow channel proposed to be similar in shape to an hour-glass (Jung et al. 1994).

Aquaporin polypeptides generally form homotetramers in the membrane (Engel et al. 2000) but with each monomer forming a single water pore (van Hoek et al. 1991; Jung et al. 1994; Shi, Skach & Verkman 1994). However, heterotetramers are formed between two aquaporin isoforms (25 and 26 kDa) in the tonoplast of protein-storage vacuoles from lentil seeds (Harvengt et al. 2000). In addition the closely related MIPs AqpZ and GlpF show different degrees of aggregation where GlpF can exist as monomers under low ionic strength whereas AqpZ forms tetramers (Borgnia & Agre 2001). Recently it has been demonstrated that two aquaporin isoforms (PM28A and PM28C) from spinach leaf plasma membrane, probably form separate homotetramers but with very different structural characteristics (Fotiadis et al. 2001).

The structure of AQP1 has been determined from electron cystallographic studies to increasingly high resolutions (e.g. Walz et al. 1995; Jap & Li 1995; Cheng et al. 1997; Walz et al. 1997; Murata et al. 2000) enabling an atomic model to be resolved at about 0·38 nm (Murata et al. 2000; Ren et al. 2001). Resolution to 0·22 nm has been obtained for an aquaglyceroporin, GlpF (Fu et al. 2000). Despite some differences in the matrix used for the two-dimensional protein crystals, these studies confirm the basic hour-glass structure and the presence of six transmembrane helices tilted from the perpendicular axis to form a cylindrical barrel. The projection structure of the plant α-TIP indicates a similar structure to animal MIPs (Daniels, Chrispeels & Yeager 1999). Likewise specific residues in AQP1 that are thought to be involved in helix interactions and stabilization of the pore are conserved in all maize AQPs (Chaumont et al. 2001).

The resolution to the size of a water molecule has enabled models to be constructed for water and solute permeation through the putative pore (Murata et al. 2000; Fu et al. 2000) and to determine why, for instance, water can rapidly permeate but the hydronium ion (H3O+) does not (Murata et al. 2000). The four-fold axis of symmetry in the centre of the tetramer that generally displays low electron density, has been speculated to be a potential ion channel pore because it posses some similarities to K+ channels (Fu et al. 2000). Structural aspects of MIPs will also be considered below with respect to selectivity and control of water channel activity.

Molecular basis of selectivity

The molecular basis of aquaporin selectivity is currently not well understood. Substrate exclusion by size in a narrow aqueous pore has long been considered to be the basis of water channel selectivity (Macey 1984). However, this model is unlikely to be sufficient to account for the various selectivities among the different members of the MIP superfamily. The proposed pore size in aquaporins (0·2–0·25 nm) (Cheng et al. 1997; Hill 1994) compared to the size of a water molecule is consistent with water moving as a single file such that the flux of an individual molecule depends on the flux of other water molecules. This accounts for the high ratio of osmotic permeability to diffusive permeability (Pos/Pd) that is used as a functional assay of water channels/pores (Dainty 1963; Niemietz & Tyerman 1997; Rivers et al. 1997). CO2 has a cylindrical shape with its radius (0·14 nm) being small enough so that it may traverse the pore along its axis. Therefore it would be expected that water flow through AQP1 would affect the permeation of CO2. Ammonia is a pyramid-shaped molecule that could permeate the selectivity barrier if it was oriented with the nitrogen atom facing the side of the pore. A recent atomic model of water permeation through AQP1 requires that the oxygen atom in a water molecule hydrogen bonds with polar groups on one side of the pore at the narrowest point where the ends of the two half helixes meet (Murata et al. 2000). This breaks the hydrogen-bonded chain of water molecules through the pore and may prevent H+ moving via a ‘proton wire’ or via H3O+ permeation. Thus the cross-sectional area at the selectivity filter needs to be wide enough to permit H2O to move sideways. If this is the case then NH3 can fit through this cross-sectional area.

It is generally considered that hydrated ions are excluded from the water pore because of their size and the hydrophobic residues lining the pore make it energetically unfavourable for the ions to be dehydrated. The anion channel activity observed with AQP6 is related to a positively charged residue in the water pore (Yasui et al. 1999) and ion channel activity observed in other MIPs may be the result of permeation via the central axis of the tetramer (Fu et al. 2000).

Boric acid permeation may be similar to the mechanism of glycerol transport. Boric acid interacts strongly with OH groups and has three OH groups, as does glycerol. However, despite the range of polyols that GlpF is permeable to (Fu et al. 2000), this does not include boric acid (Dordas et al. 2000). For glycerol the mechanism of permeation through a MIP pore has been described as being analogous to sliding on a ‘greasy pole’. Based on a crystal structure of GlpF at 0·22 nm resolution, Fu et al. (2000) proposed that the alkyl backbone of glycerol moves against a hydrophobic wedge of the selectivity filter and the OH groups form hydrogen-bonds with residues on the other side. Polyol sugars that have the OH groups lined up in the same stereo-specific relationship to the alkyl backbone were able to permeate Glpf, e.g. ribitol. Even sorbitol was able to permeate but mannitol does not, illustrating the stereo-specificity (Fu et al. 2000). Perhaps this is the basis of the lack of boric acid permeation through GlpF since boric acid is a planar molecule with the OH-groups separated by bond angles of 120°. The exclusion of water by GlpF and perhaps by Zm-PIP1 may be related to the more hydrophobic wall of the channel. Fu et al. (2000) point out that in water-selective aquaporins the hydrophobic corner in the narrow constriction is smaller.

Animal aquaglyceroporins can be distinguished from water-selective aquaporins on the basis of five amino acid residues, including two that follow the NPA sequence in loop E (Froger et al. 1998) or a pair of residues in helix 1 and 4 (Heyman & Engel 2000); however, this does not apply to the plant MIPs (Santoni et al. 2000). For the insect AQPcic changing two amino acid residues near to loop E changed the selectivity from water to glycerol, and this was also related to the protein-forming monomers rather than tetramers (Lagree et al. 1999). More of the plant MIPs need to be tested for non-electrolyte selectivity before specific residues are correlated with selectivity.

Control of water permeability

If we assume that aquaporins act independently of each other (but see Chaumont et al. 2000b) and that their water channel activity is not influenced by their density in the membrane, the total ensemble water flow (F) through aquaporins, under a constant gradient across an area of cell membrane (Am), should be equal to the density of aquaporin pores in the membrane (D), times the flow through a single open pore (fo), times the open probability of the pore (Popen), times the membrane area (Am):

inline image

Thus, the flow can be varied by changing the density (D), changing the flow through a single open pore (fo, change its conductance), or changing the fraction of the time that the pore is in a conducting state (Popen). With the current methods for measuring water channel activity it is not possible to distinguish between a change in fo and a change in Popen, however, it should be possible to separate changes in fo × Popen from changes in D. We will refer to fo × Popen as gating of AQPs.

In animal kidney, vasopressin induces large increases in the water permeability of the apical membrane of principal cells in the collecting duct. The signalling includes a cAMP dependent pathway and protein kinase A. The water permeability in the apical membrane is increased by incorporation of vesicles containing AQP2 (i.e. increase in D, reviewed in Nielsen & Agre (1995). A similar situation is now proposed to occur for rat hepatocytes where the location of AQP8 is mostly intracellular and insertion to the plasma membrane is stimulated by cAMP (Garcia et al. 2001). Control of plant water channel activity via changes in D through an analogous vesicle shuttle mechanism has also been suggested based on the distribution of PIP proteins within plant cells (Kirch et al. 2000). As yet however, we do not have an equivalent model-cell system like the kidney collecting-duct to test this in. The observation that abscisic acid (ABA) increases water permeability in cortex cells of roots (Hose et al. 2000) would be useful particularly in conjunction with the technique of Ramahaleo et al. (1999) for measuring isolated protoplast Pos. If this could be combined with simultaneous whole-cell patch clamping so that membrane capacitance is monitored to determine exocytosis (Homann & Tester 1998), it would be possible to determine if such a mechanism occurs also in plants.

Some plant aquaporins when phosphorylated can display higher water channel activity in Xenopus (Maurel et al. 1995; Johansson et al. 1998). In the case of PM28a the protein is present constitutively at high levels in the plasma membrane and is phosphorylated in situ. Thus it would appear that gating of the channel via phosphorylation occurs for this aquaporin.

The mammalian aquaporins AQP0, AQP3 and AQP6 show sensitivity to reduced pH (Zeuthen & Klaerke 1999; Yasui et al. 1999; Nemeth-Cahalan & Hall 2000) in a manner that suggests a gating effect. Chloride channel gating was observed for AQP6 at low pH (Yasui et al. 1999) and gating is observed for NOD 26 in lipid bilayers (Weaver et al. 1994; Lee et al. 1995) illustrating that MIPs can intrinsically gate and conduct osmotically important ions. The pH effects are very relevant for plants as acidification of the cytoplasm often occur with stress responses that cause dramatic changes in cell water permeability, e.g. during anoxia (Zhang & Tyerman 1991) cytosolic pH declines (Kurkdjian & Guern 1989; Gout et al. 2001). pH gradients across membranes may also determine the directionality of weak base and weak acid transport such as NH3 transport across the PBM discussed above. The effect of pH on water permeability has seldom been examined for plants. The effect of external pH on water permeability for Chara and Nitella has been tested and found to not have a significant effect (Ord, Cameron & Fensom 1977; Tyerman & Steudle 1984; Wayne et al. 1994), except in Nitella at pH 4·4 where a slight drop was recorded (Ord et et alal. 1977). A decrease in water permeability of outer cortex cells exposed to pH 4·5 was also observed for a maize variety that was acid sensitive, but was not observed in a variety that was acid tolerant (Gunsé, Poschenrieder & Barceló 1997). The effect of permeable weak acids as a means to reduce internal pH would be worthwhile investigating.

Response to turgor and osmotic pressure

In the yeast Saccharomyces cerevisiae, turgor is regulated by varying the concentration of glycerol. Under hypoosmotic shock glycerol is released via the Fpslp MIP (Tamas et al. 1999). This also occurs under anaerobic conditions where glycerol accumulates in the cells as a metabolic end product. There is no evidence for mechanosensitivity of Fps1p but the hydrophilic C- and N-terminus of the protein is much longer than for other MIPs and a 20 amino-acid region of the N-terminus is required for turgor/osmosensitivity (reviewed by Hohman et al. 2000). This illustrates the potential for MIPs to be involved also in volume or turgor homeostasis in plants. Cell pressure probe studies occasionally reveal sudden changes in hydraulic conductivity (Cosgrove & Steudle 1981) and low turgor pressure has also been known to increase the hydraulic conductivity of some plant cells (Steudle, Zimmermann & Zillikens 1982) suggesting that mechanosensitivity of MIPs should be investigated. Mechanosensitivity of ion channels is well established and in some cases the channel can respond to membrane tension without linker proteins for transmission of stress. The best studied channel in this class is the Escherichia coli MscL channel, the structure of which is known (Rees, Chang & Spencer 2000). Using amphipathic cations and anions that will distort one or other leaflet of the lipid bilayer, Martinac, Adler & Kung (1990) showed that the force for channel gating comes directly from the lipid membrane. We have tested these compounds on water permeation across the PBM and found that the cationic amphipath (chlopromazine, CPZ) increases water permeability but not glycerol permeability (Niemietz and Tyerman unpublished).

Increased tension in vesicles derived from kidney brush borders seems to decrease water flow through water channels (Soveral, Macey & Moura 1997a, b). As tension could change during the kinetics of volume changes induced in vesicles to measure water permeation this could lead to unusual kinetics, for example an apparent delay in the onset of the volume change (Soveral et al. 1997a). We have observed these delays in PBM vesicles only in the presence of HgCl2. However, HgCl2 also causes rectification of glycerol permeation across PBM vesicles (Tyerman & Niemietz 2000) suggesting that a conformational change occurs in the channel as proposed for PMIP31 from Beta vulgaris (Barone et al. 1997).

Ramahaleo et al. (1999) found no difference in Pos between swelling and shrinking of protoplasts despite large increases in surface area being observed for swelling. The large area increase observed for root protoplasts (34% increase) means that invaginations of plasma membrane unfold or new membrane is incorporated, because the elastic expansion of the membrane can only account for about a 3% increase (Wolfe & Steponkus 1981). Since Pos appeared to be constant this indicates that water channels were already present in the membrane invaginations or were already present in the internal membrane stores.

Osmotic gradient sensitivity?

Gating in ion channels often results in the phenomenon of rectification, or unequal fluxes in each direction. This is due to the closure of the channel when the gradient causes ion flow in a particular direction. Rectification can also occur for a constantly open pore with equal ion concentrations on each side of the pore when there are asymmetrical energy barriers about the mid-point along the pore. The symmetrical structure of water channels suggested that flow in opposite directions would be equal and this was confirmed for AQP 0–5 (Meinild, Klaerke & Zeuthen 1998). However, GlpF does not have a symmetrical pore (Fu et al. 2000) and some plant MIPs may also share this feature. Rectification can occur from such asymmetry depending on the size of the solutes used to create the osmotic gradient, i.e. how far they penetrate into the pore (Hill 1995). Rectification of flow has been observed in both algal (Chara, Tazawa, Asai & Iwasaki 1996) and animal (Farmer & Macey 1970) membranes. In the case of Chara the rectification disappeared when water channels were inhibited (Tazawa et al. 1996). We have found that rectification did not occur for PBM vesicles when the external solute concentration was the same for opposite osmotic gradients across the membrane. However, because water permeability increases with decreasing solute concentration, rectification appears to occur if water flow in different directions is compared with different final external solute concentrations (Niemietz and Tyerman, unpublished).

Rectification of flow may occur as a result of phosphorylation/dephosphorylation of aquaporins in response to external osmotic pressure or turgor pressure. PM28A may be gated closed in response to decreased water potential (see above, and model described in Johansson et al. 2000). If the signalling leading to dephosphorylation is very rapid in response to a sudden increase in external osmotic pressure, then Pos for exosmosis would be lower than for endosmosis. In Chara such rectification is only observed for osmotic gradients, but not for hydrostatic pressure gradients (Steudle & Tyerman 1983). Could the rectification be due to sensing of osmotic pressure and a resulting signal cascade resulting in changed phosphorylation status of aquaporins?

It is well known that water permeability declines in Chara with increasing osmotic pressure and this has been suggested to be due to dehydration of pores in the membrane (Dainty & Ginzburg 1964c; Kiyosawa & Tazawa 1972; Steudle & Tyerman 1983). Niemietz & Tyerman (1997) found that the Pos of endomembrane vesicles from wheat roots declined with increased external osmotic pressure. In this case it was not known if the effect was caused by the magnitude of the gradient or the absolute value of the osmotic pressure as occurs in Chara (Steudle & Tyerman 1983). Another example of an osmotic effect on water channel activity occurs with NOD26. Measurements of the Pos of PBM vesicles by Niemietz & Tyerman (2000) yielded values that were lower than those measured by Rivers et al. (1997). The difference is related to the much higher osmotic pressure used to bath the vesicles by Niemietz & Tyerman (2000 and unpublished results).

Are there structural clues for gating mechanisms in aquaporins? It is not clear what structural elements of aquaporins could lead to the gating that was clearly observed by Yasui et al. (1999). In terms of osmotic pressure sensitivity, dehydration of the protein may restrict the pore thereby reducing single channel conductance (Zimmerberg & Parsegian 1986). The relatively large vestibule in the central axis of the AQP1 tetramer could potentially play a role here since if solutes are excluded from this space but water is not, then a hydrostatic pressure (tension) will occur in the vestibule that is dependent on the osmotic pressure in the solution. This could alter the conformation of the surrounding water pores in the monomers to change single channel conductance.

Improvements in the measurement of water channel activity

A problem with functional characterization of aquaporins is that no technique has yet been fully developed to assess water flow across single membranes from specific cells. This is in contrast to ion channels where specific membranes can be isolated and single channels measured using the patch-clamp technique. A technique has been developed for measuring the water permeability of single protoplasts (Ramahaleo et al. 1999), but these measurements will reflect the composite water transport across the tonoplast, cytoplasm and plasma membrane in series. However, this technique can also be applied to vacuoles (Morillon & Lassalles 1999). Being able to identify the type of cell being measured would greatly enhance this technique and it should be possible to use enhancer trap lines of Arabidopsis with expression of green fluorescent protein to identify specific cell types (e.g. Maathuis et al. 1998; Kiegle et al. 2000)

Alternatively specific membrane types can be isolated and kinetics of vesicle shrinking/swelling can be measured using stopped-flow fluorescence spectroscopy. This is compromised by the need for relatively large amounts of tissue and potentially several cell types being used. This can be overcome by using cell cultures (Maurel et al. 1997; Gerbeau et al. 1999) or isolation of specific organelles such as the symbiosome from legume nodules (Rivers et al. 1997; Niemietz & Tyerman 2000). However, there is a need for a microtechnique equivalent to patch clamping that can be used to assess aquaporin activity on a small region of a single membrane. Such a technique will enable rapid evaluation of types of post-translational modifications to aquaporins and would also greatly enhance the screening of more specific inhibitors. The method developed by Pohl, Saparov & Antonenko (1997) for artificial membrane bilayers shows some promise in this respect (Saparov et al. 2000). In this technique ion concentration profiles adjacent to the membrane with aquaporins incorporated are measured with ion sensitive microelectrodes. The water flow across the small area of membrane will sweep ions toward (endosmotic side) or away from (exosmotic side) the membrane and the steepness of the concentration profile can be used to measure the magnitude of the water flow.

An important tool in measurement of aquaporin activity is blockade by mercurials. However, these can have non-specific effects on metabolism in living cell systems and has been discussed elsewhere (Tyerman et al. 1999). Zinc has been shown to reduce water permeability in Chara reversibly and to be more effective at high pH (Rygol, Arnold & Zimmermann 1992; Tazawa et al. 1996). Some alternatives to metals and organo-metals, are emerging. Some K+ channel blockers have been reported to inhibit water channel activity. This includes tetraethyl ammonium (TEA) (Brooks, Regan & Yools 2000) and nonyltriethylammonium (C9) (Wayne & Tazawa 1990; Tazawa, Sutou & Shibasaka 2001). For AQP1 expressed in oocytes the inhibition by TEA was as strong as with HgCl2 at the same concentration (100 µm) and mutating a tyrosine residue to a phenylalanine in loop E of the water-pore abolished sensitivity to TEA (Brooks et al. 2000). Zhang & Tyerman (1999) have tested TEA on wheat root cortex cells but did not observe an effect. Another blocker that has been used is dimethyl sulphoxide (e.g. Garcia et al. 2001) but this has to be used at high concentrations (500 mm).

Patterns of mip location

Clues to function of MIPs may be obtained from the patterns of mRNA and protein expression in plant organs, cell types and membrane types. High expression levels of particular PIP and TIP aquaporins have been associated with: (i) high cell expansion rates (Ludevid et al. 1992; Kaldenhoff et al. 1995; Chaumont et al. 1998; Weig & Eisenbarth 2000) (ii) cell differentiation (Kaldenhoff et al. 1995); (iii) cells that exhibit large changes in volume, e.g. guard cells and pulvinar motor cells (Kaldenhoff et al. 1995; Sarda et al. 1997; Fleuratlessard et al. 1997) (iv) high rates of water flow through cells (Barrieu, Chaumont & Chrispeels 1998; Kirch et al. 2000; Otto & Kaldenhoff 2000; Frangne et al. 2001) (v) the formation of a large central vacuole (Karlsson et al. 2000). For three TIP isoforms (γαδ) there is an association with different types of vacuole (Jauh et al. 1998; Jauh, Phillips & Rogers 1999). However, the expression pattern of a spinach δTIP (So-δTIP) was completely different to δTIPs in other species leading Karlsson et al. (2000) to conclude that physiological roles based on expression patterns could not be extrapolated across species.

With regard to conclusions drawn from sites of mRNA accumulation (e.g. in situ hybridizations) there are limitations because potentially there are three levels of regulation that could govern water channel activity subsequent to transcription of the aquaporin gene; regulation of the amount of protein translated (post-transcriptional regulation), regulation of the amount of protein incorporated into the target membrane (post-translational regulation) and regulation of the aquaporin activity (opening and closing or gating). Suga et al. (2001) showed that the amounts of MIP protein might not reflect the mRNA levels. They compared mRNA expression of six PIPs and two TIPs in radish organs, and also examined protein levels for two of the PIPs (PAQ1 and PAQ2) and TIPs. They concluded that transcription of MIP genes are independently regulated in different organs and at different stages of development confirming previous studies (e.g. Kirch et al. 2000). In some organs at specific stages all the MIP mRNAs were expressed. However, the amount of protein for PAQ2 was virtually undetectable despite the mRNA being abundant. Suga et al. (2001) suggest that the mRNAs for PAQ2 may be ready for rapid translation in response to some stimuli that could include water stress. Kirch et al. (2000) also concluded for Mesembryanthemum crystallinum that MIP protein expression was cell specific and depended on developmental stage. MIP-A (a PIP) was associated with phloem cells whereas MIP-B (also a PIP) with xylem parenchyma. MIP-B was also located between xylem vessels. MIP-F, which was clearly confined to the tonoplast, was most highly expressed in cells that should have high water fluxes through them but was also ubiquitously expressed throughout the plant (Kirch et al. 2000). Otto & Kaldenhoff (2000) showed that a PIP from tobacco (NtAQP1) that is water and urea permeable has its mRNA mainly expressed in roots. The protein was shown by immuno-localization to be in cells closely associated with xylem vessels and phloem cells. This was also found to be the case in petioles.

Plant biotrophic interfaces

Specific MIPs are also expressed in plant biotrophic interfaces (Fortin et al. 1987; Weaver et al. 1991; Opperman, Taylor & Conkling 1994; Serraj et al. 1998; Krajinski et al. 2000; van Dongen et al. 2000) where water and solute flows can be concentrated, and are potentially a critical component in the interaction between organisms (Ikeda et al. 1997). The development of arbuscular mycorrhiza (AM) in Medicago truncatula increases the expression of a TIP mRNA (Mtaqp1) that when expressed in Xenopus oocytes gives increased water permeability (Krajinski et al. 2000). This was suggested to allow for more rapid water equilibration between vacuole and cytoplasm in root cells containing arbuscules. Another consideration, with respect to the location of up-regulated aquaporins in AM, is their role in facilitating water transport across the interfaces and the possibility of an advection effect in facilitating solute fluxes, in particular of phosphate. Arbuscules are likely to be the sites of phosphate release to the plant whereas intercellular hyphae could be the sites of sugar release to the fungus (Smith, Dickson & Smith 2001). Since water flow from fungal hyphae in the soil would be expected to be directed toward the roots, preferential flow at the arbuscular interface due to a greater density of aquaporins in either plant or fungal membranes would also facilitate the movement of phosphate. High water flows across the intercellular hyphal interface would impede the movement of sugar and could be a disadvantage.

Plant parasitic interfaces are also likely to involve aquaporins either in the process of infection by the parasite or in the function of the interface. When tomato (Lycopersicon esculentum) is challenged by the higher plant parasite, Cuscuta reflexa, an incompatibility reaction ensues. In this reaction tomato increases expression of LeAqp2, a PIP1 family member, but TRAMP expression was not increased (Werner et al. 2001). Auxin has been found to be elevated in tomato during the interaction, but when auxin (IAA) was applied independently it increased the expression of both LeAqp2 and TRAMP (Werner et al. 2001). The TIP family member TobRB7 has also been reported to be up-regulated in tobacco (Nicotiana tabacum) roots during infestation with root knot nematodes (Opperman et al. 1994). In addition, a homologue of TobRB7 has been found to be up-regulated in root knot nematode feeding sites in tomato (Bird 1996). The cis-acting sequence responsible for root-specific or nematode-inducible expression of TobRB7 has been isolated and characterized (Opperman et al. 1994), and used in gene silencing strategies to produce transgenic plants with enhanced nematode resistance. Interference with TobRB7 expression during giant cell induction by root knot nematodes possibly interferes with development of functional feeding sites and suggests the importance of MIPs in this biotrophic interface.

Membrane clusters

Clustering of MIPs within a particular region of a membrane may also provide functional clues. In Arabidopsis and Brassica napus PIP1 isoforms have been imunolocated to plasmalemmasomes (Robinson et al. 1996; Frangne et al. 2001). Plasmalemmasomes are highly folded regions of the plasma membrane and are often associated with cells that have high transport activity, e.g. transfer cells. Robinson et al. (1996) suggest that in regions where the cytoplasm is thin, the high water flux through the PIP1-dense plasmalemmasomes would allow more direct access of water flow to the tonoplast. This is an interesting concept considering what might happen when a water flow is stimulated, as flow across the plasma membrane is driven by a combination of hydrostatic and osmotic gradients, whereas that across the tonoplast can only be driven by osmotic gradients. As a result of different flow densities across the plasma membrane concentration gradients may well develop within the cytoplasm as suggested in the composite transport model for water channels in Chara membranes (Henzler & Steudle 1995). If the water flux of adjacent tonoplast and plasma membrane regions do not match there is the possibility that local changes in cytoplasmic thickness will occur causing the tonoplast to be appressed or withdrawn away from the plasma membrane. TIPs have also been noted to occur in aggregates (Jauh et al. 1998). The animal MIP, AQP4 that has the highest measured single channel water permeability can form orthogonal arrays of particles. These were proposed to amplify water transport through a bulk-flow siphoning mechanism (Yang et al. 1996).

Effect of water stress on the expression of mip genes

Over the last decade evidence has accumulated showing that water stress can change expression patterns of MIP genes (Table 1). In Pisum sativum Guerrero and coworkers (Guerrero, Jones & Mullet 1990; Guerrero & Crossland 1993) found that a number of PIP-genes were up-regulated in wilted shoots and leaves. A similar increase in PIP-expression was reported for the rd28 gene from Arabidopsis thaliana (Yamaguchi-Shinozaki et al. 1992) and for the TRAMP protein in tomato (Fray et al. 1994). Other PIP-genes increasingly expressed are mipA in Brassica oleracea (Ruiter et al. 1997), NeMIP1 to NeMIP3 in Nicotiana excelsior (Yamada et al. 1997), CpPIPa6 and CpPIPa2 in the resurrection plant Craterostigma plantagineum (Mariaux et al. 1998), and two rice clones OsPIP1a and OsPIP2a (Malz & Sauter 1999).

Table 1.  Changed MIP expression in plants under water stress
Observed expressionGenePlantTissue/organ testedMembrane
Aquaporin activityReference
PIPs, up-regulatedclone 7aPisum sativumshootnot determinednot determinedGuerrero et al. 1990
rd28Arabidopsis thalianawhole plant extractnot determinednot determinedYamaguchi-Shinozaki
et al. 1992
rd28Arabidopsis thalianaall organs, highest in seeds,
flower bolts
antibosies bind to
increases Pos of Xenopus
oocytes; no Hg
Daniels et al. 1994
TRAMPLycopersicumwilted stemsnot determinednot determinedFray et al. 1994
trg31Pisum sativumleaf, midribnot determinednot determinedGuerrero & Crossland
mip ABrassica oleracesanthers; probably pollen
not determinednot determinedRuiter et al. 1997
leaves in response to root
antibodies bind to
not determinedMariaux et al. 1998
Nicotiana excelsiorleavesnot determinednot determinedYamada et al. 1997
Oryza sp.internodal growing zonenot determinednot determinedMalz & Sauter 1999
rootsantibodies bind to
not determinedKirch et al. 2000
PIPs, down-regulatedMC-MIPA,
roots, vascular bundles,
not determinedincreases Pos of Xenopus
Yamada et al. 1995
rwc1Oryza sp.leavesnot determinedincreases Pos of Xenopus
Li et al. 2000
TIPs, up-regulatedrTIPOryza sp.shoots, rootsnot determinednot determinedLiu et al. 1994
SunTIP7Helianthus annuusguard cellsnot determinedincreases Pos of Xenopus
oocytes, Hg inhibition
Sarda et al. 1997
Bob-TIP26-1Brassica oleraceameristematic regions
of florets
not determinedincreases Pos of Xenopus
oocytes, Hg inhibition
Barrieu et al. 1999
TIPs, down-regulatedCp-TIPCraterostigma
leaves, callusnot determinednot determinedMariaux et al. 1998
SunTIP18Helianthus annuusrootsnot determinedincreases Pos of Xenopus
oocytes, Hg inhibition
Sarda et al. 1999
leavesantibodies bind to
TP fraction
not determinedKirch et al. 2000

Some PIP-genes, however, are down-regulated under water stress. Yamada et al. (1995) found a family of PIP transcripts in the common iceplant, Mesembryanthemum crystallinum that were down-regulated, when the plants were exposed to drought stress. Rice water channel 1 rwc1 (Li et al. 2000) showed a similar decrease in expression.

Some members of the TIP-family also showed changed expression patterns when plants were exposed to water stress: rTIP in rice (Liu, Umeda & Uchimiya 1994), SunTIP7 in sunflower (Sarda et al. 1997), and Bob-TIP26 in cauliflower (Barrieu et al. 1999) were all expressed more. Conversely SunTIP18 in sunflower (Sarda et al. 1999) and Cp-TIP in the resurrection plant (Mariaux et al. 1998) and Mc-MIP F in Mesembryanthemum crystallinum (Kirch et al. 2000) were down-regulated, when the plants lacked water.

Another potential layer of regulation consists in protein modification. Johansson et al. (1996, 1998) demonstrated that in spinach leaves PM28A was dephosphorylated and thus inactivated under drought conditions. This could not be stimulated by ABA treatment. Johansson et al. (1998) suggest that the plasma membrane-associated Ca2+-dependent protein kinase that is responsible for phosphylating PM28a may be triggered by Ca2+ influx through stretch activated channels.

As is the case for other drought-induced proteins (Ingram & Bartels 1996) ABA can be involved in triggering expression of aquaporin-like genes. In rice OsPIP1a (Malz & Sauter 1999) and rTIP1 (Liu et al. 1994) were increasingly expressed in ABA-treated plants. Mariaux et al. (1998) showed that although Cp-PIPa2 responded to both desiccation and ABA-treatment, the closely related, drought-responsive CpPIPa6 was unaffected by ABA. rd28 in Arabidopsis thaliana (Yamaguch-Shinozaki et al. 1992) and the tomato protein TRAMP (Fray et al. 1994) were both insensitive to ABA.

It is quite obvious that aquaporin-like genes are involved in drought response, but a causal interpretation of the data is made difficult by several issues: cDNA libraries used to identify genes differentially expressed under drought conditions are often obtained from whole plant extracts, where it is impossible to localize wilting-responsive gene expression to a specific tissue. Drought response, however, can occur as localized as, for example in guard cells (Sarda et al. 1997) or in anthers (Ruiter et al. 1997). In Mesembryanthemum crystallinum (Yamada et al. 1995) aquaporin-like genes are specifically expressed in root vascular bundles and in the endodermis. Sometimes individual cells in a tissue can be up-regulated in aquaporin expression (Kirch et al. 2000).

In addition to this, subcellular localization of the described transcripts is often unknown. Genes are allocated to PIP and TIP families solely based on sequence homology, a practice that has been shown to be rather ambiguous (Barkla et al. 1999). In only two of the studies reported above for water-stress effects (Daniels et al. 1994; Mariaux et al. 1998) antibodies were used to show that the encoded PIPs were localized in the plasma membrane.

Another problem arises because water channel activity of the encoded proteins is often unknown. However, the number of studies, where the genes in question were expressed in Xenopus oocytes, and aquaporin activity was demonstrated, is rising (Daniels et al. 1994; Yamada et al. 1995; Sarda et al. 1997; 1999; Barrieu et al. 1999; Li et al. 2000).A clear picture of the function of MIP genes in drought-response will only emerge after the above uncertainties have been resolved.

Functional characterization of mip-mediated transport in native membranes

Plasma membranes

In plants the discovery of new MIP genes has outpaced the functional characterization of the underlying transport mediated by MIPs particularly in their native membranes. We have obtained many important clues from the gene sequences, heterologous function and expression patterns but ultimately we need to determine the role of MIPs in the native membrane. This is not an easy task and is complicated by alterations to cells and membranes caused by the measurement process. It is also complicated by another important route for water flow in intact cells, through plasmodesmata.

A surprising observation that has been made independently by three research groups is that water channel activity was low or absent in isolated plasma membrane vesicles (Maurel et al. 1997; Niemietz & Tyerman 1997; Dordas et al. 2000) despite the probable presence of aquaporins in the membrane. Niemietz & Tyerman (1997) and Maurel et al. (1997) both found low Pos, high activation energies, and lack of HgCl2 inhibition. However, Niemietz & Tyerman (1997) did observe a Pos/Pd ratio for plasma membrane that was significantly greater than one, indicating that there may have been some water channel activity. Dordas et al. (2000) found a low Pos value (24 µm/s) for squash root plasma membrane that was not inhibited by HgCl2. Interestingly the boric acid permeability of squash root plasma membrane was inhibited by HgCl2 (see selectivity discussion above). In maize one subgroup of the PIP family (ZmPIP1a and b) displays low aquaporin activity in Xenopus, but ZmPIP2a does show activity as do other members of the PIP2 group from other plants (Chaumont et al. 2000a). There seems also to be some co-operative effect between ZmPIP1b and ZmPIP2a in oocytes, but it is not known if this is via increased incorporation of PIP2a in the membrane, resulting from the presence of PIP1b, or a co-operative effect between the two proteins within the membrane (Chaumont et al. 2000b). So PIPs can account for water channel activity, at least in Xenopus oocytes.

A low water permeability of the plasma membrane compared with the tonoplast makes sense in theory because it reduces the magnitude of volume disturbances in the cytoplasm during osmotic shocks (Maurel et al. 1997; Niemietz & Tyerman 1997; modelled in Tyerman et al. 1999). However, modelling of the kinetics of pressure-relaxations obtained with the cell pressure probe on intact wheat root cells, and substituting the values obtained by Niemietz & Tyerman (1997) showed that water channels could be active in the plasma membranes of intact cells (Zhang & Tyerman 1999). However, the contribution from plasmodesmata to water flow induced from a single intact cell did not allow a firm conclusion to be made. Based on measurements made using the pressure clamp technique where cell osmotic volume was estimated, Zhang & Tyerman (1991) concluded that plasmodesmata may account for decrease in hydraulic conductivity caused by anoxia. However, experiments by Cleland, Fujiwara & Lucas (1994) showed that the size exclusion limit of plasmodesmata actually increased with anoxia suggesting that the plasmodesmatal anulus could have a higher hydraulic conductivity. Recently it has been shown that the transport of solutes through plasmodesmata can also occur through the desmotubule (Cantrill, Overall & Goodwin 1999), but it is not known if this pathway is affected by anoxia. It is an exciting possibility that water may also take the desmotubule pathway via the endoplasmic reticulum lumen. It would therefore be interesting to examine the occurrence of aquaporins in the endoplasmic reticulum membrane providing a link between aquaporins and symplasmic water transport. Organelles that act as calcium stores have been proposed to have water channel activity in Nitella (Kikuyama & Tazawa 1998).

The complicating effect of plasmodesmata can be avoided by using isolated protoplasts. Measurements on protoplasts from wheat roots showed that the stage of plant development correlated with different patterns of water permeability (Ramahaleo et al. 1999). Here the majority of protoplasts from wheat roots had Pos values similar to that obtained for the bulked plasma membranes from wheat roots (Niemietz & Tyerman 1997), except for protoplasts from 5-d-old roots where some had values that were over 10-fold higher and similar to those obtained for all intact cells measured with the cell pressure probe (Zhang & Tyerman 1999). Thus control elements of aquaporins may have been altered or conditions during measurement were not optimal for water channel activity to be observed in vesicles or the majority of protoplasts. The absence of turgor is one obvious difference between intact cells and protoplasts or vesicles.

An alternative explanation for the lack of water channel activity in wheat root plasma membrane vesicles measured by Niemietz & Tyerman (1997) relate to the nutrient status of the plants since it has been shown that nutrient stress can reduce water permeability of roots, root cells and root membranes (Radin & Mathews 1989; Carvajal, Cooke & Clarkson 1996; Clarkson et al. 2000). Although no evidence of nutrient deficiency was evident in the seedlings used by Niemietz & Tyerman (1997), Clarkson et al. (2000) found low values of water permeability in isolated plasma membrane vesicles and low hydraulic conductivity of exuding roots in the early stages of deprivation of the nutrient anions NO3 H2PO42– or SO42–. These changes may be the result of variation in expression of aquaporin mRNAs which also occurs diurnally (Henzler et al. 1999; Clarkson et al. 2000), but this does not exclude the possibility of post-translational control of aquaporins as well. Henzler et al. (1999) found that despite diurnal oscillations in whole root hydraulic conductivity this could not be attributed to changes in cortex cells. Thus they suggested that a specific cell type such as the endodermal cells in the root may have exclusively displayed the alterations in activity of aquaporins. Maggio & Joly (2000) have also observed a strong circadian rhythm in water flux through excised roots, which, based on HgCl2 inhibition, was suggested to be the result of oscillations in water channel activity.

Ramahaleo et al. (1999) found that there was a bimodal distribution of Pos for protoplasts isolated from roots, which changed with root development. For young roots (less than 3 d old) Pos values were less than 10 µm s−1, values that do not indicate aquaporin activity and for wheat root protoplasts a high activation energy was obtained. For 3- to 5-day-old roots, a higher proportion of protoplasts had high Pos values, well above 100 µm s−1. This was particularly evident for rape roots where a very low activation energy was also measured. A wide range in Pos was also observed for protoplasts from Arabidopsis leaves (wild type; Pos 5–500 µm s−1) (Barrieu, Morillon & Chrispeels 2000). Reduced expression of RD28 (PIP2 family) clearly resulted in a reduction in Pos to median values of 1·2–2·5 µm s−1. This compares with a median value for wild-type and over-expressing plants of about 10–20 µm s−1 (Barrieu et al. 2000). Thus although a Pos of 10–20 µm s−1 is rather low it still seems to reflect some water channel activity as was also indicated for wheat root plasma membrane based on a Pos/Pd ratio of 3 (Niemietz & Tyerman 1997).

A link between growth and water channel activity has been indicated from a correlation between Pos of hypocotyl protoplasts and hypocotyl growth induced by brassinolide in an Arabidopsis mutant defective in brassinolide synthesis (Morillon et al. 2001a). Again over a 100-fold difference in Pos between individual protoplasts was measured for wild-type, but brassinolide-treated mutant plants appeared to have a significant subpopulation of cells with higher Pos than untreated controls. It was suggested that the higher Pos may be required to reduce the hydraulic resistance in the cell-to-cell pathway under increased water potential gradients generated by rapid growth.

It is not clear why such a wide range in Pos is measured for individual protoplasts in the studies discussed above, but it may be related to post-translational control or gating of aquaporins in response to as yet uncharacterized conditions or original position of cells within tissues. Earlier work with the pressure probe also showed wide ranges in Lp[= (Pos × Vw)/RT) for the same type of plant cells (e.g. Cosgrove & Steudle 1981; Brinckmann et al. 1984).

There are clearly discrepancies between different ways of measurement of water channel activity in plasma membranes that may be related to several factors including: (1) nutritional status of plants; (2) changes in water channel activity during membrane or protoplast isolation procedures; (3) differences in aquaporin activity between cell types and stages of development; and (4) contribution to water permeation measured in intact cells from plasmodesmata. With the possible exception of the last point, all indicate tight control of plasma membrane aquaporins. With respect to plasmodesmata it is also possible that aquaporins in the endoplasmic reticulum could control transcellular water flow if flow is via the desmotubule (Cantrill et al. 1999). It will be possible to study plasma membrane aquaporin regulation in more detail using new techniques available by variation to isolation procedures including greater attention to phosphorylation status, free Ca2+ concentration and pH. Promotor-trapped lines using GFP to identify specific cell types would be very useful to use with the technique of Ramahaleo et al. (1999).


Studies with tonoplast vesicles (Maurel et al. 1997; Gerbeau et al. 1999) and isolated vacuoles (Morillon & Lassalles 1999) have clearly demonstrated that the tonoplast has a very high water permeability as the result of aquaporin activity. Three criteria for water channel activity were demonstrated for tonoplast vesicles from tobacco suspension cells by Maurel et al. (1997): High Pos, reversible inhibition with HgCl2, and low activation energy. Niemietz & Tyerman (1997) used the lower phase of two phase-partitioned vesicles as an endo-membrane fraction that included vacuole vesicles and concluded that water channel activity was high, based on the same criteria used by Maurel et al. (1997), but in addition, a value of Pos/Pd of 7 was observed. Morillon & Lasalles (1999) found that even with substantial inhibition by HgCl2 there was still evidence for water flow through water channels based upon the activation energy remaining low.

The evidence to date suggests that tonoplast has constitutive water channel activity. However, based upon TIP expression studies we might expect differences between tonoplast membranes from different cell types and during different stages of cell expansion (see above). It is possible that water channel activity is constitutive but the differences between cells and developmental stages may be more subtle, such as differences in selectivity or differences in response to control factors such as pH. Gerbeau et al. (1999) found that the tonoplast from cultured tobacco cells had high permeability to a range of small non-electrolytes in addition to water and that HgCl2 inhibited this. A similar selectivity profile was observed for a TIP (Nt-TIPa) from tobacco cells when expressed in oocytes. This TIP was immuno-located to the tonoplast of tobacco cells.

Evidence for the participation of aquaporins in whole plant response to water stress

Water potential gradients and the conductance of the flow pathways govern water movement through a plant. The hydraulic conductance of components in the pathway may be variable, particularly in the root system (Passioura & Tanner 1985; Passioura 1988; Steudle 2001). The dynamics of water transport in roots are multi-faceted due to the complex anatomical structure of the root system and modifications to the endodermis and exodermis (if present), caused by the deposition of suberin and lignin (Perumalla & Peterson 1986; Enstone & Peterson 1997). There are three different pathways for radial water flow, which are around protoplasts (apoplast), through the cytoplasm of cells via plasmodesmata (symplast) and from cell to cell across the membranes. Depending on the nature of the driving force for water transport, one pathway may dominate the route of radial water flow or transport may incorporate a combination of the different pathways (Steudle & Peterson 1998). Suberized layers restrict water flow in the apoplast and so water transport must occur via the cell-to-cell pathway in some root tissues (Barrowclough et al. 2000; Zimmerman et al. 2000).

Abiotic stress affects the hydraulic conductivity of roots and a number of studies have shown that nutrient deprivation, water deficit, salinity and waterlogging reduces hydraulic conductivity (Everard & Drew 1987; Joly 1989; Azaizeh, Gunse & Steudle 1992; Birner & Steudle 1993; Carvajal et al. 1996; Lu & Neumann 1999; Else et al. 2001). During stress, anatomical changes that decrease hydraulic conductance, occur in the roots of some plant species (Steudle 2000). However, anatomical changes provide only slowly reversible (growth dependent) or no reversible adjustment to root hydraulic conductivity and so there must also be physiological regulation controlling water transport across the root.

A tentative indicator of aquaporin involvement is the observation that HgCl2 can reversibly reduce hydraulic conductivity of roots (Maggio & Joly 1995; Carvajal et al. 1996; Wan & Zwiazek 1999; Quintero, Fournier & Benlloch 1999; Barrowclough et al. 2000; Carvajal, Cerdá & Martinez 2000; Maggio & Joly 2000; Martinez-Ballesta, Martinez & Carvajal 2000; Martre, North & Nobel 2001). Based upon the extent of inhibition caused by HgCl2, the cell-to-cell pathway with aquaporins in the serial array of membranes could account for between 21 and 85% of root hydraulic conductance in different species under non-water-stressed conditions (discussed in Martre et al. 2001; Maurel et al. 2001). The correlation between diurnal oscillations in root hydraulic conductivity and aquaporin expression also indicates that aquaporins are involved (Henzler et al. 1999).

North & Nobel (2000) reported that in Agave deserti, a desert succulent, drought-stressed root segments have a lower hydraulic conductivity than segments of the same root in moist soil. Only moist roots showed a reduction of hydraulic conductivity by HgCl2, suggesting that under drought conditions, some root water channels are switched off. This was also suggested for Opuntia acanthocarpa where aquaporins (defined as the HgCl2 inhibited component) accounted for 38% of the reduction in root hydraulic conductance in response to drought. Reduction in hydraulic conductance by water stress was also reported by Lu & Neumann (1999), although they found that mercury treatment of roots only inhibited leaf growth in drought-stressed seedlings, but not in control plants. They suggested that in their system (rice plants with a plentiful water supply) aquaporins were normally inactive or not expressed and mercury inhibition in drought-stressed plants was a consequence of up-regulated aquaporin genes. Water stress appears to diminish the amplitude of oscillations in water flux through excised roots under constant pressure (Maggio & Joly 2000) and it was proposed that perhaps water flux through roots is co-ordinated, via aquaporin activity, with stomatal conductance in the shoot.

The effects of ABA on the hydraulic conductivity of roots and root cells (Glinka 1973; Quintero et al. 1999; Hose et al. 2000), and also of giant algal cells (Ord et al. 1977), has been investigated. Variable effects related to technique were observed in the earlier literature (discussed by Hose et al. 2000). In maize root cells the increase in water permeability, after a lag of about 2000 s, was transient (over a period of 2000 s) and specific for undissociated (+)-cis-trans-ABA isomer (Hose et al. 2000). Generally an increase in hydraulic conductivity is observed that correlates broadly with increased aquaporin gene expression observed by others (see above), but it is not known if the changes observed in the physiological studies are related to increased protein synthesis.

Salt stress has been shown to lower hydraulic conductivity of whole maize roots (Azaizeh & Steudle 1991), and of excised paprika pepper plants to a level similar to that obtained by mercury inhibition (Carvajal, Martinez & Alcaraz 1999). This has been matched to similar effects on cell hydraulic conductivity (Azaizeh et al. 1992) or protoplast Pos (Martinez-Ballesta et al. 2000; Carvajal et al. 2000). An ameliorative effect of Ca2+ has also been reported (Azaizeh et al. 1992; Martinez-Ballestra et al. 2000; Carvajal et al. 2000). Based on a partial rescue of the salt inhibition of Pos by the phosphatase inhibitor okadaic acid, Carvajal et al. (2000) suggest that the salt inhibition may be via reduced phosphorylation of aquaporins along similar lines to the proposed regulation of PM28A in spinach leaves (Johansson et al. 1996, 1998). Not all studies have observed a reduction in root cell Lp in response to salinity. For tobacco root, no effect of salinity on Lp was observed on epidermal and cortex cells, but interestingly the Lp values were rather low compared with other plant roots (Tyerman et al. 1989).

It is difficult to predict what would be the most advantageous strategy for alterations to aquaporin activity in response to water stress or salinity. On one hand a greater aquaporin activity in roots would lead to smaller water potential gradients between xylem and soil. This could have the effect of relieving xylem tensions, provided the apoplastic pathway does not dominate. However, according to the composite transport model of the root, increasing xylem tensions should divert more flow via the apoplast since hydrostatic gradients generate proportionately more flow through this pathway than via the cell-to-cell pathway, for which osmotic and hydrostatic gradients generate flow (reviewed by Steudle 2001). If transpiration is stopped by closure of stomata it may be a better strategy to reduce water channel activity in order to reduce water loss to the soil (Martre et al. 2001). In the case of salinity, advection of NaCl to the root surface may be minimized by reduced aquaporin activity.

The reduction in plasma membrane water permeability, that is assumed to occur as a result of dephosphorylation in spinach leaves, was proposed to reduce the rate of water loss by the plant, providing a delay that could be utilized for synthesis of osmolytes (Johansson et al. 1998). This is an attractive hypothesis, but within the possible two orders of magnitude change that is likely to occur in Pos this would take the time for a cell to equilibrate with the apoplast from a few seconds to a few hundred seconds and may still be too rapid for accumulation of osmolytes. Alternatively if flow initially occurred predominantly through the cell-to-cell pathway a decrease in Pos of the cells may divert water flow more to the apoplast pathway, or through different regions of the tissue, perhaps containing cells with lower volumetric elastic modulus that act as water stores. As an additional factor to be considered for the cell-to-cell pathway, elevation of cytosolic Ca2+ can transiently close plasmodesmata (Holdaway-Clarke et al. 2000).

With respect to the xylem it has been suggested that embolism repair may involve aquaporins located in the xylem parenchyma cells adjacent to vessels (Holbrook & Zwieniecki 1999). However, unless gradients are favourable for water flow into embolized vessels aquaporins are of no use. This has been discussed in more detail by Steudle (2001). Recently Morillon et al. (2001b) proposed an interesting mechanism involving active water transport and aquaporin gating for rapid volume changes in motor cells. Based on their model, if water could be transported actively into xylem parenchyma cells, under conditions when aquaporins are closed, perhaps a sufficient gradient may be generated to allow a rapid volume flow into embolized vessels, by opening of aquaporins. This possibility requires more scrutiny.

Conclusions and future prospects

There is now an array of methods with solid theoretical backgrounds for measuring water channel activities in plants at different levels. These methods can be used for both water permeation and solute permeation (see reviews by Verkman 1995; Tomos & Leigh 1999; Agre et al. 1999). Some require consideration of alternative pathways, e.g. apoplast in whole root pressure probe measurements (Barrowclough et al. 2000) or plasmodesmatal transport in single cell pressure probe measurements (Murphy & Smith 1998). It is possible that water transport through plasmodesmata may be governed by aquaporins in the endoplasmic reticulum given that solute transport through the desmotubule has been demonstrated (Cantrill et al. 1999).

At the membrane level where vesicles are isolated, activity may be lost by not adequately controlling potential phosphorylation status of aquaporins but this also provides the opportunity to assess gating behaviour. Aquaporin activity can now be measured in a small area of artificial lipid bilayer by measurement of the unstirred layer profiles with ion sensitive microelectrodes (Pohl et al. 1997; Saparov et al. 2000) and this offers promise for further smaller scale measurements.

Plant MIPs may be permeable to an array of physiologically relevant molecules including boric acid, hydrogen peroxide, ammonia and carbon dioxide. Other molecules such as nitrous oxide, ethylene and undissociated organic acids would be worth investigation. Ultimately, it will be difficult to judge the main function of some MIPs simply based on selectivity profiles, and gene-silencing techniques will be required to test specific hypotheses on function (Kaldenhoff et al. 1998; Barrieu et al. 2000; Chen et al. 2001).

Water channels may be gated by pH and could be osmosensitive and mechanosensitive, either directly as discussed above or indirectly via Ca2+-dependent phosphorylation (Johansson et al. 1998). The effect of pH may be very important for rapid responses to stresses such as salinity or anoxia in order to divert water and nutrient uptake to regions of the soil that are more favourable to the plant.

Osmosensitivity and mechanosensitivity, if these can be demonstrated conclusively, may link volume regulation of endomembrane compartments and turgor regulation of the cell with water channel activity. In terms of the whole root or leaves these characteristics could also adjust the overall hydraulic resistance of the cell-to-cell pathway and therefore alter hydraulic pathways within whole tissues. This will affect the water potential gradients required for flow, perhaps relieving xylem tension and ultimately the degree of stomatal opening required.


We thank the Australian Research Council for funding to S.D.T. and a reviewer for comments that enhanced this manuscript significantly.

Received 14 May 2001; received in revised form 13 August 2001; accepted for publication 13 August 2001