Regulation of plant aquaporin activity


  • François Chaumont,

    Corresponding author
    1. Unité de Biochimie Physiologique, Institut des Science de la Vie, Université catholique de Louvain, Croix du Sud 2-20, B-1348 Louvain-la-Neuve, Belgium
    Search for more papers by this author
  • Menachem Moshelion,

    1. Unité de Biochimie Physiologique, Institut des Science de la Vie, Université catholique de Louvain, Croix du Sud 2-20, B-1348 Louvain-la-Neuve, Belgium
    Search for more papers by this author
    • The Robert H. Smith Institute for Plant Sciences and Genetics in Agriculture, Faculty of Agricultural, Food and Environmental Quality Sciences, The Hebrew University of Jerusalem, Rehovot 76100, Israel.

  • Mark J. Daniels

    1. Department of Cell Biology, The Scripps Research Institute, La Jolla, CA 92037, U.S.A.
    Search for more papers by this author



Accumulating evidence indicates that aquaporins play a key role in plant water relations. Plant aquaporins are part of a large and highly divergent protein family that can be divided into four subfamilies according to amino acid sequence similarity. As in other organisms, plant aquaporins facilitate the transcellular movement of water, but, in some cases, also the flux of small neutral solutes across a cellular membrane. Plant cell membranes are characterized by a large range of osmotic water permeabilities, and recent data indicate that plant aquaporin activity might be regulated by gating mechanisms. The factors affecting the gating behaviour possibly involve phosphorylation, heteromerization, pH, Ca2+, pressure, solute gradients and temperature. Regulation of aquaporin trafficking may also represent a way to modulate membrane water permeability. The aim of this review is to integrate recent molecular and biophysical data on the mechanisms regulating aquaporin activity in plant membranes and to relate them to putative changes in protein structure.

Abbreviations used:

abscissic acid


green fluorescent protein


hemi-helix E


nodulin26-like intrinsic protein


plasma membrane intrinsic protein


small and basic intrinsic protein


tonoplast intrinsic protein. The two-letter prefix denotes the species, e.g. At, Arabidopsis thaliana


Plant growth and development are dependent on the tight regulation of water uptake and transport across cellular membranes and tissues. Since the discovery of the first aquaporin or water channel in plants more than 10 years ago (Maurel et al., 1993), many studies have indicated that aquaporins represent an important selective pathway for water to move across cellular membranes, and have definitively changed our view about how plants might regulate water flow in different physiological conditions (reviewed in Tyerman et al., 1999, 2002; Johansson et al., 2000; Javot and Maurel, 2002; Maurel et al., 2002).

Since the first report of a plant water channel, an increasing number of aquaporin genes have been reported. Unlike in animals, plants appear to express a surprisingly high number of aquaporin homologues. For instance, 35 aquaporin genes have been identified in the Arabidopsis genome (Johanson et al., 2001; Quigley et al., 2002) and 36 expressed aquaporin genes have been detected in maize (Zea mays) (Chaumont et al., 2001). On the basis of sequence similarity, plant aquaporins have been classified into four subfamilies, the plasma membrane intrinsic proteins (PIPs), the tonoplast intrinsic proteins (TIPs), the nodulin26-like intrinsic proteins (NIPs) and a small group named the small and basic intrinsic proteins (SIPs) (Chaumont et al., 2001; Johanson and Gustavsson, 2002; Johanson et al., 2001). Although PIPs and TIPs are largely targeted to the plasma membrane and the vacuolar membrane or tonoplast respectively, complexities in localization and trafficking exist (see below). The subcellular location of members from the NIP and SIP subfamilies is still uncertain. The divergence of plant aquaporins into four subfamilies had been already established when land plants emerged, as indicated by the presence of aquaporins in the moss Physcomitrella patens falling into the same four subfamilies (Borstlap, 2002). More recent aquaporin differentiation has occurred, as indicated by the number of TIP subgroups present in the tracheophytes and absent in bryophytes (Borstlap, 2002). Even more recent gene duplications, arising after the monocot—dicot separation, have been hypothesized from the presence of several proteins on a single branch of the aquaporin phylogenetic tree in maize that do not have concomitant homologues in Arabidopsis (Chaumont et al., 2001). The increase of aquaporin isoforms in plants suggests that gene duplications offer adaptive advantages for growth in different environmental conditions, possibly as a result of divergent transport selectivities or regulatory mechanisms.

Monitoring aquaporin gene expression patterns in many plant species in specific tissues, cell types or in response to phytohormones or environmental factors has highlighted the putative roles of water channels. Aquaporins have been shown to control the extensive water transport from the roots to the leaves during the transpiration stream, and also to regulate other processes, such as the transport of assimilates through sieve elements in the phloem, the closure or aperture of stomata in leaves, the movement of leaves and the control of cytoplasmic homoeostasis. These data have been reported in detail by Maurel et al. (2002) and Tyerman et al. (2002), and will not be discussed further in the present review.

The aim of the present review is to integrate recent molecular and biophysical data on the mechanisms regulating aquaporin activity in plants. After describing the structural features of plant aquaporins, we will discuss both conclusive and putative roles of phosphorylation, heteromerization, pH, Ca2+, high solute concentration and pressure on aquaporin activity, and relate these mechanisms to possible structural changes or channel gating. The term gating is used here to describe the conformational changes of a channel leading to pore opening or closure. Gating of ion channels is known to be regulated by several mechanisms, some depending on factors such as the intracellular metabolic state or second messengers, and others depending on the electrical potential of the cell membrane, as observed for the voltage-gated ion channels (Aggarwal and MacKinnon, 1996; Seoh et al., 1996). The immense amount of information available on ion-channel gating comes mainly from patch-clamp experiments that allow the measurement of ion-channel activity with high accuracy and at the single-channel level. Unfortunately, such accurate techniques do not exist in measuring the water permeability of membranes, due to the electro-neutral character of water and the technical difficulties inherent in such measurements. Evidence for aquaporin gating has been reported for several mammalian and plant isoforms (Yasui et al., 1999; Nemeth-Cahalan and Hall 2000; Cho et al., 2002; Tournaire-Roux et al., 2003; Wan et al., 2004), but the molecular mechanism(s) causing the conformational changes and the physiological relevance of gating remain poorly understood.

Plant membranes have a high range of osmotic water permeability

The involvement of aquaporins in the regulation of membrane water permeability has been recently questioned in Hill et al. (2004), which argues that lipid membranes are very permeable to water and that there is no need for aquaporin-facilitated water movement in most physiological situations. However, contrary to this long-standing notion, there is evidence of very low water permeability of biological membranes, including the plant plasma membrane. Osmotic water permeability coefficient (Pf) values below 10 μm·s−1 have been recorded for plasma membrane from tobacco (Nicotiana tabacum), onion (Allium cepa), wheat (Triticum aestivum), Mimosacea tree (Samanea saman) and maize cells using the plasmometric method, which does not involve any plasma membrane injury (Table 1) (Ramahaleo et al., 1999; Morillon et al., 2001a; Moshelion et al., 2002, 2004; Siefritz et al., 2002; Suga et al., 2003). Interestingly, higher Pf values (up to 1280 μm·s−1) have been measured using similar procedures for cells from the same species, but from different cell types or developmental stages (Table 1) (Ramahaleo et al., 1999; Morillon et al., 2001b; Suga et al., 2003). The observation that Pf was reduced by the addition of mercury chloride [a commonly used aquaporin inhibitor, see Tyerman et al. (1999) and Javot and Maurel (2002) for its action and limitations] strongly implicates the involvement of aquaporins in the regulation of plasma membrane Pf. The permeability of internal membranes, such as the tonoplast and the peribacteroid membrane, measured using either stopped-flow spectrophotometry or the plasmometric method, is generally very high (Pf>200–500 μm·s−1) and can be inhibited by mercury, demonstrating aquaporin-mediated water movement through these membranes (reviewed in Maurel et al., 2002). In addition, when determined, the transmembrane water flux exhibits a low dependence on the temperature (low Arrhenius activation energy) and the ratio of osmotic to diffusional water permeability is higher than 1 (Maurel et al., 2002), both indicating the presence of facilitated transmembrane water flow. Finally, many plant aquaporins have been demonstrated to increase the transmembrane flux when expressed in the Xenopus oocyte expression system (see below). Altogether, these observations suggest that plant cell membrane permeability can be greatly enhanced by active aquaporins.

Table 1.  Osmotic water permeability coefficients of plant protoplastsThe information collected in this Table represents only methods of direct measurements of the cell water permeability coefficient (Pf). The measurements were performed on the whole protoplast and involve no injury to the cell plasma membrane. The number of days refers to the time elapsed since germination. TC, transference chamber; SEC, solution-exchange chamber.
Cell sourceMeasuring methodPf range (μm·s−1)Reference
Wheat root   
 3-day oldTC0.9–5Ramahaleo et al. (1999)
 5-day oldTC1.2–640Ramahaleo et al. (1999)
Maize root   
 5-day old capSEC0–3.2M. Moshelion and F. Chaumont, unpublished data
 5-day old meristematic cellSEC0–4.6M. Moshelion and F. Chaumont, unpublished data
 5-day old elongation zoneSEC0–7.6M. Moshelion and F. Chaumont, unpublished data
 5-day old root hair zoneSEC0–32.5M. Moshelion and F. Chaumont, unpublished data
 2-day oldTC1.25–10Ramahaleo et al. (1999)
 5-day oldTC2.5–1280Ramahaleo et al. (1999)
Tobacco roots tipTC4–128Siefritz et al. (2002)
Rape (Brassica napus) root   
 3–5-day oldTC80–640Ramahaleo et al. (1999)
 2-day oldTC0.9–180Ramahaleo et al. (1999)
Flax (Linum usitatissimum) rootTC150–263Ramahaleo et al. (1999)
 3–5-day old   
Radish root   
 4-day old endodermisCapillary transference120–550Suga et al. (2003)
 4-day old cortexCapillary transference260–580Suga et al. (2003)
Melon (Cucumis melo) root 6-day oldSolution—dilution15Carvajal et al. (2000)
Rape hypocotyl 5-day oldTC220–550Ramahaleo et al. (1999)
Arabidopsis hypocotyl 15-day oldTC2.5–1280Morillon et al. (2001b)
Radish hypocotylSEC0.1–550M. Moshelion and F. Chaumont, unpublished data
Onion leafTC3–17Ramahaleo et al. (1999)
Petunia (Petunia hybrida) leafTC0.63–640Ramahaleo et al. (1999)
Arabidopsis leaf mesophylTC1.25–640Morillon et al. (2001b)
Maize BMS (Black Mexican Sweet) cell lineSEC0.1–24Moshelion et al. (2004)
Samanea motor organSEC3–5Moshelion et al. (2002)

Other evidence for a direct role of aquaporins in plant water relations comes from the manipulation of aquaporin gene expression by over-expression techniques (Aharon et al., 2003; Uehlein et al., 2003; Hanba et al., 2004), and from gene silencing by antisense suppression (Kaldenhoff et al., 1998; Martre et al., 2002; Siefritz et al., 2002, 2004) or T-DNA insertion (Javot et al., 2003). Over-expression of AtPIP1; 2 in tobacco significantly increased plant growth rate, transpiration rate, stomatal density and photosynthetic efficiency under favourable growth conditions, but caused faster wilting in drought conditions (Aharon et al., 2003). Arabidopsis plants expressing antisense AtPIP1; 2 or AtPIP2; 3 gene(s) alone or together (double antisense) showed a reduction in transcript or protein levels for several PIP1 and/or PIP2 homologues (Kaldenhoff et al., 1998; Martre et al., 2002). The altered PIP expression resulted in a reduction of Pf in isolated protoplasts and a decrease in the total root hydraulic conductivity. Interestingly, an increase in the root-to-leaf dry-mass ratio was observed, which may be a response to compensate for the reduced cell water permeability. Because the leaf hydraulic conductance per leaf area was similar to the control plants, the overall hydraulic conductance of the plant was unchanged. In other experiments, lowered expression of several PIP1 homologues in tobacco NtAQP1 antisense plants also resulted in a significant decrease of protoplast membrane water permeability, but contrary to Arabidopsis PIP antisense lines, NtAQP1 antisense tobacco did not show any increase in root mass (Siefritz et al., 2002). Instead, tobacco root hydraulic conductivity was reduced by approx. 60$, plant water potential was decreased and plant respiration was diminished, probably in order to limit water loss. In addition, the diurnal epinastic leaf movement was reduced, pointing to the importance of aquaporin-mediated water transport in this process (Moshelion et al., 2002; Siefritz et al., 2004). Finally, in Arabidopsis plants bearing T-DNA insertions in AtPIP2; 2, the hydraulic conductivity of root cortex cells and the osmotic hydraulic conductivity were reduced by approx. 30$ and 14$ respectively when compared with wild-type plants (Javot et al., 2003). These Arabidopsis knock-out lines demonstrated the involvement of a single aquaporin in root water uptake. In summary, these experiments are evidence for the important contribution of aquaporins in plant water relations via the regulation of water movement in plant cells, tissues and organs.

Aquaporin structure and specificity

Aquaporin structure has been defined by the atomic-resolution structures solved for mammalian AQP0 and AQP1, and the homologous bacterial glycerol facilitator GlpF (Fu et al., 2000; Murata et al., 2000; Sui et al., 2001; Gonen et al., 2004). These proteins illustrate the characteristic aquaporin structure, with six membrane-spanning α-helices and both C- and N-termini facing the cytosol (Figure 1). The cytosolic loop between the second and third transmembrane domain (loop B) and the extra-cytosolic loop between the fifth and sixth transmembrane domain (loop E) both form short hydrophobic helices that dip halfway into the membrane from opposite sides. These two loops contain generally conserved Asn-Pro-Ala (NPA) motifs, in which the two asparagine residues are key to the formation of the pore water-selectivity filter. However, exceptions to this NPA aquaporin signature were observed in some Arabidopsis and maize NIP and SIP isoforms where the alanine residue is replaced by leucine, valine, serine, threonine or cysteine residues (Chaumont et al., 2001; Johanson et al., 2001). Aquaporins usually exist as tetramers in which each monomer forms an independent water channel. Molecular structure analysis of two plant aquaporins, kidney bean (Phaseolus vulgaris) PvTIP3;1 (α-TIP) and spinach (Spinacia oleracea) leaf SoPIP2;1 (PM28A), confirmed the typical tetrameric conformation found in animal and bacterial aquaporins (Daniels et al., 1999; Fotiadis et al., 2001; Karlsson et al., 2003). Tetramer assembly is likely to be important for protein folding, stability and/or transport to the target membranes.

Figure 1.

A topological model of maize ZmPIP1;1 and ZmPIP1;2

The figure is based on the mammalian AQP0 and AQP1 and bacterial GlpF structures (Fu et al., 2000; Murata et al., 2000; Sui et al., 2001; Gonen et al., 2004), and shows the six transmembrane helices (TM1–TM6) and the two short helices in the structural loops B and E [hemi-helix B (HB) and HE]. Residues in yellow are conserved in all maize PIPs (Chaumont et al., 2001). Residues in green are important for the pore formation and selectivity according to the structure of AQP0, AQP1 and GlpF. Residues in red are different amino acid residues found in ZmPIP1;1. The asterisk indicates a putative phosphorylated serine conserved in all plant PIPs. This figure is adapted from (Chaumont et al., 2001) ©American Society of Plant Biologists and is reprinted with permission.

Many more aquaporins are found in plants than in animals. This greater diversity of plant aquaporins could indicate a wider range of channel solute specificity. Some plant aquaporins, such as AtTIP1;1 (Maurel et al., 1993), are characterized as being highly specific for water, whereas other isoforms were demonstrated to facilitate the passage of glycerol, urea or CO2 in addition to water (Rivers et al., 1997; Biela et al., 1999; Dean et al., 1999; Gerbeau et al., 1999; Guenther and Roberts, 2000; Weig and Jakob, 2000; Liu et al., 2003; Uehlein et al., 2003). Interestingly, the PIP, TIP and NIP subfamilies include at least one isoform transporting another solute in addition to water (Biela et al., 1999; Dean et al., 1999; Gerbeau et al., 1999; Weig and Jakob, 2000). Evidence suggests that some plant aquaporins can be permeable to ammonia and ammonium (Niemietz and Tyerman, 2000; Jahn et al., 2004), boron (Dordas et al., 2000), hydrogen peroxide (Henzler and Steudle, 2000) and small alcohols (Hertel and Steudle, 1997; Schutz and Tyerman, 1997). As a cautionary note, it has been found that a mammalian aquaporin expressed in the Xenopus oocyte system (commonly used to characterize channel activity and specificity) can demonstrate solute or ion fluxes that are not observed in the native cell membrane (Fang et al., 2002; Tsunoda et al., 2004).

As an example of the physiological relevance of channel solute diversity, the aquaporin-facilitated CO2 movement through plant cellular membranes might be important for reducing the resistance to internal CO2 diffusion and for optimizing carboxylation of ribulose-1,5-biphosphate carboxylase (reviewed in Tyerman et al., 2002). A higher CO2 membrane permeability was observed in tobacco plants over-expressing the aquaporin NtAQP1, and a lower CO2 membrane permeability was found in antisense plants (Uehlein et al., 2003). An increase of internal CO2 conductance and CO2 assimilation was also measured in rice (Oryza sativa) over-expressing barley (Hordeum vulgare) HvPIP2;1 (Hanba et al., 2004).

The molecular basis of AQP1 and GlpF selectivity is due to two filters within the pore; the first one formed by the conserved dual NPA motif which appears to act as a size-exclusion selectivity filter and an electrostatic filter preventing proton conduction, and the second one formed by a constriction region [the aromatic/arginine (ar/R) filter] recently proposed to be involved in hydrogen-bond interruption (de Groot and Grubmuller, 2001; Sui et al., 2001; de Groot et al., 2003). Structural models of the putative pore regions of Arabidopsis aquaporins, on the basis of homology modelling with mammalian AQP1 and GlpF structures, has indicated that plant aquaporins have a common fold, but show differences in pore apertures, hydrogen-bond formation and pore amphiphilicity that could result in different selectivities for channel substrates (Wallace and Roberts, 2004).


Phosphorylation of plant aquaporins was reported even before their function and specificities were known, and was the first modification known to directly affect channel activity (Weaver and Roberts, 1991, 1992; Johnson and Chrispeels, 1992; Miao et al., 1992). In vivo and in vitro phosphorylation of serine residues within the N- or C-termini of bean seed PvTIP3;1 (at Ser7), spinach leaf SoPIP2;1 (at Ser274 and Ser277) and soya-bean (Glycine max) root nodule GmNodulin26 (at Ser262) has been reported (Weaver and Roberts, 1991, 1992; Johnson and Chrispeels, 1992; Miao et al., 1992). Evidence suggests that this phosphorylation is due to a membrane-associated calcium-dependent protein kinase. Phosphorylation of Arabidopsis and maize plasma membrane aquaporins has also been detected in vivo by mass spectrometry or radioactive labelling (Chaumont et al., 2000a; Santoni et al., 2003; V. Van Wilder, H. Degand, R. Derua, E. Waelkens and F. Chaumont, unpublished data).

Heterologous expression of aquaporins in Xenopus oocytes is a very convenient system to address the function and the regulation of aquaporins (Preston et al., 1992; Maurel et al., 1993). Compared with control oocytes, oocytes expressing an aquaporin in their plasma membrane swell much more rapidly following a shift to hypo-osmotic medium. A useful feature of this system is that agonists or antagonists of kinases and phosphatases can be added, allowing the role of phosphorylation on aquaporin activity to be investigated. For instance, the water-channel activity of PvTIP3;1 in Xenopus was enhanced by adding a cAMP agonist that stimulated oocyte protein kinase A (Maurel et al., 1995). Similarly, SoPIP2;1 and GmNodulin26 activity increased when oocytes were incubated in the presence of the phosphatase inhibitor okadaic acid (Johansson et al., 1998; Guenther et al., 2003). These studies also investigated mutants highlighting the importance of specific putative phosphorylated serine residues in this regulatory mechanism. Altogether, these experiments demonstrated that plant aquaporin phosphorylation can modulate the water-channel activity; however, the mechanism is still unknown. It has been proposed that plant aquaporin phosphorylation is directly involved in rapid and reversible in situ channel gating (Johansson et al., 2000). Because the in vivo-identified phosphorylation sites were located in the cytosolic N- or C-termini of the proteins, and since structural data for these terminal domains could not be obtained, computational modelling of a negatively charged phosphate group at these positions could not be performed. Additional data will be required to characterize the gating mechanism(s) driven by N- or C-terminal phosphorylation.

It cannot yet be excluded that plant aquaporin phosphorylation is a way to regulate protein trafficking, as observed for mammalian AQP2 (reviewed in Brown, 2003). In collecting-duct principal cells, the diuretic hormone vasopressin raises the levels of intracellular cAMP with consequent activation of protein kinase A, which can then phosphorylate Ser256 of AQP2. Phosphorylation of AQP2, which is normally present in intracellular vesicles, triggers AQP2 vesicle fusion to the apical plasma membrane and leads to an increase in membrane Pf. The characterization of the plasma membrane and endomembrane localization of plant aquaporins expressed in oocytes (and plant cells) in response to the application of pharmacological compounds affecting their phosphorylation status could be performed in order to determine whether the observed water-channel activity increase is due to an increased amount of protein in the plasma membrane or a gating mechanism.

In addition to the phosphorylation of N- or C-terminal serines, there is indirect experimental data suggesting that a serine located in the cytoplasmic loop close to the first NPA motif can also be phosphorylated (Figure 1). This serine is conserved in all plant PIPs and several TIPs, and lies within protein kinase consensus phosphorylation sequence (Arg/Xaa-Lys-Xaa-Ser-Xaa-Xaa-Arg) recognized by several protein kinases, including calcium-dependent protein kinases (Johansson et al., 2000). Replacement of this serine residue with an alanine in SoPIP2;1 or PvTIP3;1 resulted in reduced water-channel activity when either protein was expressed in oocytes (Maurel et al., 1995; Johansson et al., 1998). The high degree of conservation of this serine suggests that it has an important structural and functional role. Phosphorylation of this residue has not yet been observed in in vivo or in vitro phosphorylation experiments. However, an antibody raised against a conserved peptide containing the phosphorylated serine of wheat PIP2 recognized a band with the expected PIP2 size in maize extract that disappeared after phosphatase treatment (Aroca et al., 2005).

Regulation of aquaporin phosphorylation in plants may be controlled according to developmental stages or environmental factors. For example, phosphorylation of PvTIP3;1 detected in vitro peaked in developing seeds and declined during seed imbibition (Johnson and Chrispeels, 1992), possibly reflecting an enhanced phosphorylation of PsTIP3;1 in vivo during the same period and suggesting a dynamic control of vacuolar swelling and fusion (reviewed in Maurel et al., 1997). GmNodulin26 phosphorylation in symbiosomes reached a peak when nodules matured and were fully developed, and was also enhanced during salt and water stress, suggesting an important role in osmoregulation of the infected cell cytosol (Guenther et al., 2003). SoPIP2;1 phosphorylation at Ser274 was enhanced by increasing leaf water potential and was reduced during water deficit, perhaps to prevent water efflux from leaf cells (Johansson et al., 1996, 1998). Recently, phosphorylation and dephosphorylation of a tulip (Tulipa gesneriana) plasma membrane aquaporin was reported to regulate petal opening at 20°C and closing at 5°C respectively (Azad et al., 2004). These observations were coupled with different rates of water transport through the stem to the petals and were dependent on Ca2+. These examples clearly demonstrate the crucial physiological role of aquaporin phosphorylation and dephosphorylation in plants, even if the molecular consequence of this modification remains unknown.


Aquaporin activity can be affected by the oligomeric state of the protein. Aquaporins are generally found to be tetrameric, both in vitro and in vivo. Tetramerization appears to involve interactions between neighbouring monomers via the membrane-spanning α-helices and the extramembrane loops, which may contribute to tetramer stability (Murata et al., 2000). Many experiments have demonstrated that various aquaporins form stable homotetramers, culminating in the high-resolution structural analysis of AQP1 and GlpF crystals that showed these proteins as homotetramers, and a number of other aquaporins have also been shown to form homotetramers (Fu et al., 2000; Murata et al., 2000; Fotiadis et al., 2001).

However, several plant isoforms have been found to form heterotetramers. In one instance, heteromers of two tonoplast aquaporins (25 and 26 kDa in size) from lentil (Lens culinaris) seed were detected in crosslinking experiments. Incubation of tonoplast-enriched membranes in the presence of a cross-linking reagent induced the formation of dimers, trimers and tetramers, made of both 25 and 26 kDa proteins (Harvengt et al., 2000). Heteromerization of plasma membrane aquaporins has been inferred through studies of ZmPIPs co-expressed in Xenopus oocytes (see below). Plant PIPs can be divided into two major groups, PIP1s and PIP2s, on the basis of their sequence and their water-channel activity. All PIP2 proteins exhibit high water-channel activity in Xenopus oocytes or in yeast vesicles, whereas PIP1 proteins are often inactive or have low activity (Daniels et al., 1994; Yamada et al., 1995; Johansson et al., 1998; Biela et al., 1999; Chaumont et al., 2000b; Marin-Olivier et al., 2000; Dixit et al., 2001; Moshelion et al., 2002; Gaspar et al., 2003; Suga and Maeshima, 2004).

Co-expression of ZmPIP1; 2 and different ZmPIP2 genes in Xenopus oocytes demonstrated a positive co-operative effect in aquaporin activity which was probably the result of enhanced plasma-membrane targeting (Fetter et al., 2004). The physical interaction of ZmPIP1;2 and ZmPIP2s was demonstrated by affinity chromatography, and this heteromerization resulted in higher levels of ZmPIP1;2 in the plasma membrane (Fetter et al., 2004). Although ZmPIP1;1 and ZmPIP1;2 isoforms showed poor aquaporinactivity when expressed individually, they showed increased aquaporin activity when co-expressed together. Also, co-expression of ZmPIP1;2 and ZmPIP2;5 resulted in increased water-channel activity, but no co-operative effect was observed between ZmPIP1;1 and ZmPIP2;5. Curiously, the ZmPIP1;1LE mutant, in which the loop E sequence of ZmPIP1;1 is changed to that of ZmPIP1;2 by the substitution of four amino acids, behaved identically to ZmPIP1;2 in these co-expression experiments. Such results indicate that ZmPIP1 heteromerization is required for these isoforms to act as functional water channels.

A possible mechanism for this observation is suggested by comparing the structural models of PIP1 and PIP2 isoforms (Figure 2). The mutant ZmPIP1;1LE was created from ZmPIP1;1 by the substitutions V246I, Q250R, H251D and A255N, which produces a loop E sequence identical with that found in ZmPIP1;2 (Fetter et al., 2004). Putative structures of ZmPIP1;1, ZmPIP1;1LE, ZmPIP1;2 and ZmPIP2;5 were generated by homology modelling (Sanchez and Sali, 2000), followed by a constant temperature thermal annealing step performed as a molecular dynamics simulation.

Figure 2.

Structural models of hemi-helix B bearing the second NPA motif, loop E and transmembrane helix 6, shown for the ZmPIP1;1 aquaporin (orange ribbon), the mutant ZmPIP1;1LE (red ribbon), ZmPIP1;2 (grey ribbon), and ZmPIP2;5 (green ribbon)

Four amino acids in loop E of ZmPIP1;1 were changed to create the mutant ZmPIP1;1LE. These changes produce a loop E sequence in ZmPIP1;1LE identical with that found in the ZmPIP1;2 aquaporin. This change causes ZmPIP1;1LE to have the same positive co-operative effect as ZmPIP1;2 when co-expressed in Xenopus oocytes with other ZmPIPs (Fetter et al., 2004). The four mutated residues of ZmPIP1;1LE (I246, R250, D251 and N255) are shown in spherical atom form. Putative structures of ZmPIP1;1, ZmPIP1;1LE, ZmPIP1;2, and ZmPIP2;5 were generated by homology modelling (Sanchez and Sali, 2000) using bovine and human AQP1 and sheep AQP0 as templates. Each tetrameric model was initially equilibrated using conjugate gradient energy minimization, then thermally annealed in vacuo for 400 ps in 1 fs steps at a temperature of 290 K. During the thermal annealing step, a harmonic restraint constant of 12 kcal/mol per A2 was applied to all α-helices. The CNS software package (Brünger et al., 1998) was used for molecular dynamics simulations. Scale bar, 5 Å.

Our models indicate that V246I is located at the end of hemi-helix E (HE) and faces the lipid bilayer. The introduction of a larger hydrophobic residue may force HE into the pore vestibule. The substitutions Q250R and H251D appear in the middle of loop E and introduce a pair of oppositely charged residues on the external face of the protein. These residues can form ion pairs with each other or neighbouring polar residues, or form strong solvent interactions, thus constraining the loop structure. A255N is located at the N-terminal end of helix 6 and substitutes a non-polar residue with a larger polar residue, which appears to pull the side chain into a position more exposed to solvent. The overall effect is to push loop E of ZmPIP1;1LE, ZmPIP1;2, and ZmPIP2;5 towards the pore vestibule and to move HE so that it is more perpendicular to the plane of the membrane. These observations hint that loop E residues are important in propagating structural changes (i.e. multimeric state and structure of adjacent monomer) from helix 6 to NPA HE. Alternatively, changes in loop E can either affect the structure of HE containing the second NPA motif and thus affect channel activity, or affect the structure of transmembrane helix 6 and thus alter protein oligomerization.

The crucial role of loop E in aquaporin activity was reported previously in studies of chimeras of bovine AQP0 and human AQP2 (Kuwahara et al., 1999). Another study of protein chimeras showed that exchanging loop E of the glycerol channel GlpF with loop E of the aquaporin AQPcic (aquaporin of Cicadella) altered either oligomer assembly or tetramer stability, which could have affected targeting to the plasma membrane (Duchesne et al., 2002). More recently, Suga and Maeshima (2004) further demonstrated the importance of the loop E in their studies of radish (Raphanus sativus) PIPs expressed in yeast. Substituting Ile244 of RsPIP1;3 with valine (present at the same position in RsPIP2;2 and corresponding to Val246 of ZmPIP1;1, see Figures 1 and 2) increased the aquaporin activity of RsPIP1;3 more than 2-fold. A similar single mutation (V246A) in ZmPIP1;1 has yet to be tested.

Further experiments will be necessary to reveal the interactions and oligomerization of aquaporins in plant cells and to elucidate the physiological importance of such regulatory mechanisms.

Regulation by pH and Ca2+

The regulation of aquaporin permeability by Ca2+ and/or protons concentrations has been reported for mammalian AQP0, AQP3 and AQP6, and for plant PIPs (Yasui et al., 1999; Zeuthen and Klaerke, 1999; Nemeth-Cahalan and Hall, 2000; Gerbeau et al., 2002; Tournaire-Roux et al., 2003; Nemeth-Cahalan et al., 2004). The water permeability of plasma membrane from Arabidopsis suspension cells or root cells was reduced in the presence of free Ca2+ and/or low pH (Gerbeau et al., 2002; Tournaire-Roux et al., 2003). Cytosol acidosis is observed during oxygen deprivation or anoxia resulting from soil flooding and is linked to a reduction of root cell water permeability (Zhang and Tyerman, 1991; Tournaire-Roux et al., 2003). Interestingly, it was shown that in Xenopus oocytes the Arabidopsis water channels AtPIP2;1, AtPIP2;2, AtPIP2;3 and AtPIP1;2 close upon a shift of the cytosolic pH from 7 to 6. The residue primarily responsible for pH-mediated gating of AtPIP2;2 was shown to be His197 which is localized in intracellular loop D. Substitution of His197 by an alanine residue reduces the effect of cytosol acidification (Tournaire-Roux et al., 2003).

A structural model of AtPIP2;2 suggests a possible mechanism for the pH-dependent gating of this aquaporin. We performed homology modelling to generate a structure for AtPIP2;2 using the atomic-resolution structures of bovine and human AQP1 (de Groot et al., 2001; Sui et al., 2001) and sheep AQP0 (Gonen et al., 2004) as templates. This model shows that the cytoplasmic face of AtPIP2;2, excluding the N-terminus, has a large number of basic amino acids and presents a negatively charged surface (Figure 3).

Figure 3.

Models for the open state (A) and closed state (B) of the Arabidopsis aquaporin AtPIP2;2

Molecular dynamics simulations suggest that with cytosol acidosis loop D can fold to occlude the water channel, a state that could be stabilized by the interaction of the acidic N-terminus (pI 3.8) with the positively charged His197 and other basic residues on the cytoplasmic face of the protein. The availability of several atomic-resolution structures for eukaryotic aquaporins (bovine and human AQP1 and sheep AQP0) allowed us to generate a probable structure of AtPIP2;2 by homology modelling (Sanchez and Sali, 2000). The first 25 residues of the N-terminus could not be modelled as there is no corresponding template structure. To represent the open channel at pH 7 all histidine residues were set to an uncharged state, and to represent the closed channel at pH 6 all cytoplasmic histidines were set to their protonated and positively charged state. The model structures were initially equilibrated using conjugate gradient energy minimization, then thermally annealed in vacuo for 400 ps in 1 fs time steps at 290 K. During the thermal annealing step, a harmonic restraint constant of 20 kcal/mol per A2 was applied to all atoms apart from those within 15 Å of H197 which were left unrestrained. The CNS software package (Brünger et al., 1998) was used for molecular dynamics simulations. Monomers within the homotetramer are shown as differently coloured ribbon models. Basic residues on the cytoplasmic face of a monomer and one neighbour are shown in spherical atom form and labelled with both their single-letter amino-acid code and residue number. Residues of the neighbouring monomer are labelled with an asterisk. Nt; N-terminus, Ct; C-terminus. Scale bar, 10 Å.

Under conditions of cytosol acidosis, His197 and other cytosol-exposed histidine residues would exist in a protonated positively charged state. In particular, His103 forms a key part of the pore channel (Sui et al., 2001; Tajkhorshid et al., 2002) and is highly conserved in membrane intrinsic protein family proteins, whereas His197 and His264 are highly conserved in plant PIPs and may therefore play an important role in these isoforms. To examine the effects of pH on AtPIP2;2 structure, tetrameric models were generated in which His103, His197 and His264 were set to either an unprotonated or a protonated state. Molecular dynamic simulations were then performed in which residues in spatial proximity to His197 were allowed to equilibrate. At the conclusion of the simulation, the unprotonated histidine model was found in an open state (Figure 3A) and the protonated histidine model was found in a closed state (Figure 3B), in which loop D was folded over the cytoplasmic vestibule of the channel. We suggest that when loop D is folded over the pore, the acidic N-terminus of AtPIP2;2 can interact with the basic residues Lys190, Arg191, Arg194 and protonated His197 of loop D, which would stabilize the loop in a structure that occludes the water channel (Figure 3B).

This model explains the observed effects of mutations in AtPIP2;2 (Tournaire-Roux et al., 2003). H197D creates a constitutively open channel, which is likely to be due to the presence of the negatively charged aspartate that would interfere with ionic interactions between loop D and the negatively charged N-terminus. The mutant H197K results in a pH-insensitive water channel with low activity, which would be caused by the positively charged lysine enabling a pH-independent interaction between loop D and the N-terminus, leading to a generally closed channel.

pH sensivity in the mutant H197A was not eliminated, but significantly diminished (Tournaire-Roux et al., 2003). This observation could imply a shift in the pH—response curve and the participation of acidic amino acids, such as aspartate or glutamate, in the pH sensor. pH sensitivity was also reduced in the mutants R194A and D195A, but was eliminated in the double mutants R194A/H197A and D195A/H197A. These observations indicate that His197 is not the sole pH-sensing moeity, but is the major pH-sensing site under physiological conditions. Our models suggest that His197 forms a crucial part of a charge network, which includes the residues Arg194 and Asp195, that stabilizes the closure of the water channel by ionic interactions with the AtPIP2;2 N-terminus.

An alternative mechanism of water-channel gating was recently proposed to explain the pH-dependent regulation of mammalian AQP0 (Nemeth-Cahalan et al., 2004). The pH sensitivity of AQP0 is dependent on the position of histidine in loop A, and the authors suggest that a charged histidine in loop A could reorganize water molecules around the pore in a manner that restricts the flow of water through the channel.

The mechanism leading to membrane water permeability reduction in the presence of Ca2+, as observed in cultured Arabidopsis cells, is still unclear. In the case of mammalian AQP0, lowering either the Ca2+ concentration or the pH increased channel water permeability, whereas lowering both Ca2+ and pH further decreased the channel water permeability (Nemeth-Cahalan and Hall, 2000). Increasing or lowering intracellular Ca2+ concentration abolished sensitivity to external Ca2+, indicating that this regulation occurs at a cytoplasmic site. Recent data show that both Ca2+ and pH modulations are separable (Nemeth-Cahalan et al., 2004). Whereas the pH sensitivity of AQP0 is dependent on the position of histidine in loop A, Ca2+ regulation appears to involve the binding of calmodulin to the C-terminus. Similar to the effect of a charged histidine residue in loop A, a Ca2+ ion bound to the C-terminus via calmodulin could reorganize water molecules around the pore in a manner that restricts the flow of water through the channel. The effect of Ca2+ on the water-channel activity of wild-type and mutant Arabidopsis AtPIP2;2 would be useful to elucidate the molecular mechanism(s) of Ca2+-dependent channel closure.

Gating by high solute concentration and pressure pulses

Another mechanism controlling the gating of aquaporins might be the cohesion/tension forces in the presence of high concentrations of osmotic solutes (Ye et al., 2004). According to this model, osmolytes excluded from the channels may cause tensions (negative pressures) inside the pore and induce a deformation of the protein and eventually its closure. This mechanism might explain the observed inhibition of Chara cell membrane hydraulic conductivity by a high concentration of osmotic solutes (Steudle and Tyerman, 1983). In more recent experiments, membrane permeability parameters (hydraulic conductivity, permeability and reflection coefficient) of Chara cells were measured using a cell pressure probe as the concentration of variously sized osmolytes was increased (Ye et al., 2004). Water-channel activity, inferred from the membrane permeability parameters, was seen to decrease with increasing osmotic pressure and with increasing osmolyte size. A cohesion/tension mechanism was proposed to account for the gating of Chara water channels, which suggests that the osmotic tension generated within the channel by a high osmotic potential leads to the structural collapse and closure of the protein channel. Similar effects of high salinity on membrane water permeability as reported in several species could result, at least partly, from the same mechano-sensitive mechanism (Azaizeh et al., 1992; Carvajal et al., 1996, 1999; Martinez-Ballesta et al., 2000).

A different mechanism of mechanical inhibition was recently reported in young maize roots (Wan et al., 2004). Hydraulic conductivity of cortex cell membrane was measured using a cell pressure probe after variously sized pulses of turgor pressure. Whereas small- or medium-sized pressure pulses (<0.2 MPa) caused a reversible inhibition of the hydraulic conductivity, larger pressure pulses induced changes that were not reversible, except in the presence of the stress hormone abscissic acid (ABA). The authors suggested that water flow through the aquaporin could cause conformational changes in the channel, proportional to the water flux, as a result of the input of kinetic energy to the channel constriction (Wan et al., 2004). The effect of ABA in rapid positive regulation of aquaporin gating might suggest a direct binding to the channel.

Both energy-input and tension-gating mechanisms might be used by the plant to sense changes in turgor pressure and surrounding water availability, and to adapt the membrane water permeability in a ABA-dependent manner (Wan et al., 2004). It should be noted that computational simulations suggest that aquaporins can tolerate extraordinary hydrostatic and osmotic pressures, so further experiments will be required to verify these conclusions. Molecular dynamics simulations carried out with human AQP1 showed that the protein structure was stable up to applied pressures of 200 MPa, and evidence of protein unfolding was not observed until a pressure of 400 MPa was applied (Zhu et al., 2002, 2004).

Finally, other internal or external factors, such as heavy metals, nutrient, temperature and reactive oxygen species, have been shown to modify aquaporin activity (Preston et al., 1993; Daniels et al., 1996; Zhang and Tyerman, 1999; Clarkson et al., 2000; Henzler and Steudle, 2000; Niemietz and Tyerman, 2002; Lee et al., 2004). Except for the Hg2+ ion effect, which requires the presence of specific cysteine(s) located proximal to the pore, the precise molecular mechanisms by which these factors affect aquaporin activity are not yet characterized.

Regulation by membrane trafficking

Plant aquaporins are specifically targeted to certain membranes. Immunocytochemistry, immunodetection and expression of green fluorescent protein (GFP)-fusion proteins have found most aquaporins to be localized either in the plasma membrane, the vacuolar membrane or in the peribacteroid membrane of nitrogen-fixing symbiotic root nodules (Fortin et al., 1987; Morrison et al., 1988; Daniels et al., 1994; Kammerloher et al., 1994; Robinson et al., 1996; Fleurat-Lessard et al., 1997; Chaumont et al., 1998, 2000b; Barkla et al., 1999; Barrieu et al., 1999; Cutler et al., 2000; Kirch et al., 2000; Reisen et al., 2003). To reach their destination, aquaporins have to move through the secretory pathway, including the endoplasmic reticulum, through the Golgi, and then into different types of vesicles according to their membrane target(s). Regulation of plant aquaporin membrane trafficking may therefore represent a basic mechanism for modulating membrane water permeability in response to different environmental conditions and water availability.

The cellular distribution of some aquaporins appears to be more complex than simple plasma membrane or tonoplast localization. For instance, Arabidopsis plasma membrane PIP1 homologues were found in convoluted plasma membrane invaginations called plasmalemmasomes and may facilitate the water exchange between the apoplast and vacuole (Robinson et al., 1996). Maize ZmPIP1;2 and ZmPIP2;5, fused to GFP, were detected not only in the plasma membrane, but also in intracellular membranes and the perinuclear compartment, locations that probably correspond to stages of PIP transport in the secretory pathway (Chaumont et al., 2000b). Tests on maize protoplasts showed that the membrane osmotic water permeability changed during a hypo-osmotic challenge. This dynamic behaviour of Pf was interpreted as reflecting a modification of aquaporin trafficking and/or activity (Moshelion et al., 2004). The localization of Mesembryanthemum crystallinum aquaporins also revealed a complex distribution pattern (Kirch et al., 2000; Vera-Estrella et al., 2004). McPIP1;4 was not detected in the plasma membrane fractions, but was found in the tonoplast and other fractions of intermediate density, and McPIP2;1 was present, but not exclusive to the plasma membrane. Finally, in a significant development, careful analysis of tonoplast aquaporin localization has shown that there exist three different types of vacuole in plants, which can be distinguished by the type of TIP(s) in the membrane (Paris et al., 1996; Jauh et al., 1999).

Glycosylation is important for the proper routing and membrane insertion of a number of proteins, and glycosylated forms of mammalian AQP1 and AQP2 have been reported (Smith et al., 1994; Baumgarten et al., 1998). Glycosylation of AQP2 is important for exit from the Golgi complex and sorting to the plasma membrane (Hendriks et al., 2004). Recently, a study of the effects of water stress on McTIP1;2 localization revealed a new mechanism of plant aquaporin regulation possibly related to protein glycosylation (Vera-Estrella et al., 2004). Mannitol-induced osmotic stress induced a shift in McTIP1;2 localization from tonoplast fractions to higher-density fractions, which was correlated to the detection of spherical vesicular compartments decorated with McTIP1;2 antibodies. Interestingly, the osmotic stress also resulted in the appearance of a larger McTIP1;2 isoform corresponding to a glycosylated state that may be required for membrane relocalization (Vera-Estrella et al., 2004). The distribution of McTIP1;2 was inhibited by different compounds affecting the secretory pathway, such as brefeldin A, wortmannin and cytochalasin D, and appears to involve signalling through phosphorylation events, in a manner resembling the regulation of AQP2 localization (see above). This mechanism of McTIP1;2 redistribution to endosomal compartments under osmotic stress may be important in regulating the water balance of different tissues.


The regulation of plant aquaporin activity in cellular membranes appears to involve many different mechanisms. Beyond the initial regulatory step of altering gene expression according to cell type, plant developmental stage and environmental conditions, the subsequently translated aquaporins have to be sent to their target membrane, and then, when required, facilitate the transmembrane flux of water and/or small non-electrolytes. Gating of aquaporins through the different mechanisms that we have discussed could represent a rapid pathway of response to environmental constraints, such as anoxia, salt and water stress, and also to any modifications of cell water homoeostasis occurring during the daily life cycle of the plant. The tight regulation of aquaporin activity that has been observed may be essential to adjust the overall hydraulic conductivity of plant tissues. The molecular mechanisms leading to aquaporin opening or closure are not well understood, although a number of essential amino acid residues have been identified and several structural motifs have been predicted. Structural analysis of plant aquaporins (both wild-type and mutant isoforms) will be useful to define conformational changes occurring during gating. Studying the effects of different stress conditions or compounds known to modulate membrane permeability in plant cells expressing wild-type or mutant aquaporins will tell us about the mechanisms linking effectors and aquaporins. The elucidation of these mechanisms will contribute to our further understanding of water relations in the whole plant.


This work was supported by grants from the Belgian Fund for Scientific Research and the Interuniversity Attraction Poles Programme – Belgian Science Policy (to F.C.) and the U.S. National Institutes of Health (to M.J.D.), and a Marie Curie European individual fellowship (to M.M.).


  1. Osmotic water permeability coefficient (Pf): Each membrane can be characterized by a hydraulic permeability (Lp) or the related osmotic water permeability coefficient (Pf) that describe overall water movement in response to hydrostatic or osmotic pressure gradient. This is described in the equation Pf=LpRT/Vw, where R is the gas constant, T is the absolute temperature and Vw is the partial molar volume of water.

  2. Plasmometric method: A technique that records protoplast volume change in response to osmotic challenge.

  3. Root hydraulic conductivity: The water flow measured at the level of excised roots.

  4. Epinastic leaf movement: This is due to differential growth of cells at the upper and lower leaf surface.

  5. Reflection coefficient: This quantifies the membrane selectivity for a solute.