Aquaporin Nt-TIPa can account for the high permeability of tobacco cell vacuolar membrane to small neutral solutes


*For correspondence (fax +33 1 69 82 37 68;


Members of the major intrinsic protein (MIP) family, described in plants as water-selective channels (aquaporins), can also transport small neutral solutes in other organisms. In the present work, we characterize the permeability of plant vacuolar membrane (tonoplast; TP) and plasma membrane (PM) to non-electrolytes and evaluate the contribution of MIP homologues to such transport. PM and TP vesicles were purified from tobacco suspension cells by free-flow electrophoresis, and membrane permeabilities for a wide range of neutral solutes including urea, polyols of different molecular size, and amino acids were investigated by stopped-flow spectrofluorimetry. For all solutes tested, TP vesicles were found to be more permeable than their PM counterparts, with for instance urea permeabilities from influx experiments of 74.9 ± 9.6 × 10–6 and 1.0 ± 0.3 × 10–6 cm sec–1, respectively. Glycerol and urea transport in TP vesicles exhibited features of a facilitated diffusion process. This and the high channel-mediated permeability of the same TP vesicles to water suggested a common role for MIP proteins in water and solute transport. A cDNA encoding a novel tonoplast intrinsic protein (TIP) homologue named Nicotiana tabacum TIPa (Nt-TIPa) was isolated from tobacco cells. Immunodetection of Nt-TIPa in purified membrane fractions confirmed that the protein is localized in the TP. Functional expression of Nt-TIPa in Xenopus oocytes showed this protein to be permeable to water and solutes such as urea and glycerol. These features could account for the transport selectivity profile determined in purified TP vesicles. These results support the idea that plant aquaporins have a dual function in water and solute transport. Because Nt-TIPa diverges in sequence from solute permeable aquaporins characterized in other organisms, its identification also provides a novel tool for investigating the molecular determinants of aquaporin transport selectivity.


It is now well established that the central vacuole of plant cells is a pivotal organelle that serves multiple cellular functions such as metabolite storage and degradation, cell signalling, and pH and turgor regulation (Taiz 1992). Consistent with this versatility, numerous and specific transport activities have been demonstrated in the tonoplast (TP) (Barkla & Pantoja 1996; Martinoia & Ratajczak 1997). Recent studies have revealed in particular a high water permeability for the TP of tobacco suspension or wheat root cells (Maurel et al. 1997; Niemietz & Tyerman 1997). This property suggested a role for the vacuole in cytoplasm osmoregulation and in the control of transcellular water flow.

The high water permeability of some plant cellular membranes has recently been assigned to the activity of water channels (Kaldenhoff et al. 1998; Maurel et al. 1997; Niemietz & Tyerman 1997; Rivers et al. 1997). A class of water channel proteins named aquaporins is indeed present in plants (for reviews, see Maurel 1997; Schäffner 1998; Tyerman et al. 1999). These proteins belong to the major intrinsic protein (MIP) family of membrane channels with cognates in a wide variety of organisms, and exhibit a typical structure with an intramolecular repeat resulting from an ancient gene duplication, six putative membrane-spanning domains and highly conserved residues including two Asn-Pro-Ala (NPA) motifs (Agre et al. 1998; Park & Saier 1996).

Plant MIP homologues can be classified in three sequence subclasses (Maurel 1997; Schäffner 1998). Two of these correspond to the so-called tonoplast intrinsic proteins (TIPs) and plasma membrane intrinsic proteins (PIPs) which have been consistently localized in the plant TP and plasma membrane (PM), respectively. The third subclass comprises close homologues of nodulin-26 (NOD26), an MIP homologue from the peribacteroid membrane of symbiotic root nodules (Fortin et al. 1987). Some of these homologues have been identified in non-legume species and their cell localization remains to be determined (Weig et al. 1997).

Heterologous expression in Xenopus oocytes has provided a powerful tool for the functional characterization of MIP channels. Several plant homologues have been shown to function as water channels (Maurel 1997; Schäffner 1998; Tyerman et al. 1999), and a high selectivity for water was established for some of them (Maurel et al. 1993). NOD26 was the first MIP homologue identified in plants (Fortin et al. 1987) but its function is as yet not firmly established. Earlier reconstitution experiments of the purified protein in artificial membranes revealed an ion channel activity (Weaver et al. 1994) which could mediate the exchange of organic acids across the peribacteroid membrane (Ouyang et al. 1991). More recently, evidence for a role in the transport of water and small non-electrolytes has been presented (Rivers et al. 1997).

In contrast to plant MIPs, some homologues in other organisms have been unambiguously identified as solute transporters. The archetype of such proteins is GlpF, a glycerol facilitator of the inner membrane of Escherichia coli, encoded in an operon determining glycerol uptake and assimilation (Heller et al. 1980; Maurel et al. 1994). Fps1, an MIP homologue of yeast, provides a path for glycerol release in response to hypotonicity (Luyten et al. 1995). Several animal aquaporins have also been shown to transport small non-electrolytes such as urea or glycerol in addition to water, but the physiological significance of this mixed selectivity remains poorly understood (Ishibashi et al. 1997, 1998; Kuriyama et al. 1997; Ma et al. 1997; Sasaki et al. 1998).

In recent years, simultaneous advances in electrophysiological and molecular approaches have brought about a large body of information on the transport of ions across plant cell membranes (Barkla & Pantoja 1996; Logan et al. 1997). The mechanisms of non-electrolyte transport have been investigated to a much lesser extent and concern mostly the PM level. In particular, transporters specific for amino acids and sugars have been molecularly identified in this membrane (Logan et al. 1997; Tanner & Caspari 1996). Yet the transport mechanisms in the plant plasma and internal membranes of other important uncharged metabolites, such as compatible solutes, remain largely unknown (Bohnert et al. 1995; McCue & Hanson 1990; Martinoia & Ratajczak 1997).

In the present work, we characterized by stopped-flow spectrophotometry the permeability to a series of non-electrolytes of PM and TP vesicles purified from tobacco suspension cells. Our measurements revealed a very high permeability of TP vesicles to small molecules such as urea and glycerol, with typical features of a facilitated diffusion mechanism. Such a selectivity profile, reminiscent of the activity of MIP solute channels, prompted us to isolate and characterize Nt-TIPa, a novel TIP expressed in the TP of tobacco cells. When expressed in Xenopus oocytes, Nt-TIPa transports urea and glycerol in addition to water. We propose that Nt-TIPa can account for the high permeability of the tobacco TP to small non-electrolytes, and more generally, that plant aquaporins have a dual function in water and solute transport.


TP vesicles of tobacco cells show a high permeability to small non-electrolytes

Membrane vesicles enriched in PM or TP were purified by free-flow electrophoresis (FFE) from tobacco suspension cells (Maurel et al. 1997) and characterized for their capacity to transport small non-electrolytes. Urea and glycerol were used in our study as reference solutes. Polyols of different molecular size, comprising up to six carbons, were tested to evaluate the size dependence of the solute permeation. Proline, glycine and myo-inositol were also investigated because these molecules or their derivatives serve as compatible solutes in plant cells (Bohnert et al. 1995; McCue & Hanson 1990). To investigate solute transport, vesicles were loaded with 10 mm 6-carboxyfluorescein (CF) and submitted in a stopped-flow apparatus to an inwardly directed osmotic gradient generated by a high extra-vesicular concentration of the solute of interest. For all solutes tested, a rapid vesicle shrinking due to osmotic water efflux is observed, which results in an increase in the entrapped fluorophore concentration and a decrease in the fluorescence signal (Fig. 1). Over a longer period of time, the external solute may diffuse into the vesicles, driving an osmotic water influx, and the vesicles may recover their initial volume (Fig. 1). This second phase depends primarily on the membrane permeability to the solute and a large range of solutes can be easily tested using this approach.

Figure 1.

Time course of solute influx in tobacco membrane vesicles.

Tonoplast (a) and plasma membrane (b) vesicles containing 10 mm CF, 100 mm sucrose, 20 mm Tris–HEPES, pH 7.5, were abruptly mixed in a stopped-flow apparatus with an equal volume of a solution made hyperosmotic with a solute of interest. This resulted in a extra-vesicular solution containing 100 mm sucrose, 20 mm Tris–HEPES, pH 7.5, and 325 mm of the indicated solute. Change in vesicle volume can be followed by a proportional change in CF fluorescence indicated in arbitrary units. During the first phase, vesicles shrink to about 25% of their initial volume because of osmotic water efflux. In the second phase, solute influx predominates and is accompanied by water influx. The rate of vesicle swelling is primarily determined by the permeability of the membrane to the tested solute. Representative traces corresponding to an average of 7–10 individual time-course. The inset in (a) shows the same recordings on a reduced time scale.

Figure 1(a) illustrates typical biphasic kinetics of tobacco TP vesicle volume adjustment and shows that urea permeation across the TP was faster than for any other solute. When challenged with a high urea concentration, the TP-enriched vesicles recovered their initial volume in about 1 sec while they barely swelled in the presence of proline (Fig. 1a). The rate constant of vesicle swelling allowed estimation of a urea permeability value (Purea) of 74.9 ± 9.6 × 10–6 cm sec–1 (Table 1). Glycerol was the second most permeable solute (Pglycerol = 8.3 ± 1.1 ×10–6 cm sec–1). Tonoplast vesicles also exhibited a significant permeability to erythritol, glycine and proline, in the range of 10–6 cm sec–1 (Table 1). Permeation of molecules of higher molecular size could not be resolved with our technique (Psol ≤ 0.2 × 10–6 cm sec–1).

Table 1.  Stopped-flow measurements of permeability of membrane vesicles to solutesa
 UreaGlycerol (C3)bErythritol (C4)bXylitol (C5)bMannitol (C6)bMyo-inositol (C6)bGlycineProline
  1. a Measurements were performed in solute uptake experiments as exemplified in Fig. 1 and explained in the text. b Number of carbon atoms in the molecule. c Exponential rate constant (mean ± SE) fitted to the second phase (solute phase) of vesicle volume relaxation. Kinetics with a rate constant less than 0.015 sec–1 could not be fitted (–). dPermeability to solute (mean ± SE). A rate constant greater than 0.015 sec–1 yields resolution limits for Psol of 0.2 × 10–6 and 0.05 × 10–6 cm sec–1 for TP and and PM vesicles, respectively. The symbols ≤ and < indicate whether Psol was close or far below this limit. eNumber of independent membrane preparations tested.

k (sec–1)c6.84 ± 0.870.70 ± 0.090.11 ± 0.030.11 ± 0.010.08 ± 0.01
Psold (× 10–6 cm sec–1)74.9 ± 9.68.3 ± 1.11.3 ± 0.3 < 0.2 < 0.2 < 0.21.0 ± 0.10.9 ± 0.1
k (sec–1)c0.20 ± 0.060.22 ± 0.040.04 ± 0.03
Psold (× 10–6 cm sec–1)1.0 ± 0.31.1 ± 0.20.2 ± 0.2< 0.05< 0.05< 0.05≤ 0.05≤ 0.05

A similar approach using purified PM vesicles showed glycerol and urea to be the most permeant tested solutes in these membranes (Fig. 1b), with respective permeabilities Pglycerol = 1.1 ± 0.2 × 10–6 and Purea = 1.0 ± 0.3 ×10–6 cm sec–1. Except for erythritol (Peryth = 0.2 ± 0.2 × 10–6 cm sec–1), the permeability to all other tested solutes was too low to be measured (Psol ≤ 0.05 × 10–6 cm sec–1) (Table 1).

To calculate permeability values in the experiments described above, assumptions are required about the intra-vesicular solute concentration at the beginning of the second (solute) phase of the relaxation. For a given membrane or water permeability, the more permeable the solute is, the less accurate the approximation is. Such assumptions are not required in experiments where abrupt transfer of solute-loaded vesicles in a iso-osmotic solution but with reduced solute concentration results in solute efflux and concomitant vesicle shrinking (Priver et al. 1993). Figure 2 shows examples of monophasic change in CF fluorescence associated with the efflux from TP and PM vesicles of urea and glycerol, the most permeable solutes in these membranes. The permeability values derived from these measurements were Purea = 38.0 ± 9.1 × 10–6 cm sec–1 (n = 4) (TP) and Pglycerol = 1.2 ± 0.6 × 10–6 cm sec–1 (n = 3) (PM). The range of these two values is similar to the range of respective permeability values derived from biphasic relaxations (Table 1). This validates to some extent the series of solute permeability measurements performed with the latter method.

Figure 2.

Time course of solute efflux in tobacco membrane vesicles.

Tonoplast (TP) and plasma membrane (PM) vesicles containing 10 mm CF, 100 mm sucrose, 20 mm Tris–HEPES, pH 7.5, and 280 mm of urea (TP) or glycerol (PM) were abruptly mixed in a stopped-flow apparatus with an equal volume of a isoosmotic buffer deprived of urea or glycerol. This resulted in an extra-vesicular solution containing 240 mm sucrose, 20 mm Tris–HEPES, pH 7.5, and 140 mm of the indicated solute. Solute efflux down its concentration gradient and the accompanying water efflux result in vesicle shrinking as monitored by a decrease in CF fluorescence. Representative traces corresponding to an average of 7–10 individual time-courses.

Altogether, data obtained with tobacco TP and PM vesicles reveal that each of the purified membrane fractions exhibited a specific selectivity profile. In addition, TP vesicles exhibited a significantly higher permeability to non-electrolytes than PM vesicles. Urea and glycerol permeabilities were 35–70- and 7–8-fold higher in TP than in PM vesicles. Permeation of larger molecules such as glycine and proline was moderate in the TP but was much lower in the PM, in which it could not be resolved with our approach.

Pore-mediated solute transport in tobacco TP vesicles

The temperature dependence of solute flow in purified membrane vesicles was determined between 6 and 32°C. The deduced Arrhenius activation energy (Ea) values of urea and glycerol transport in TP vesicles were 4.0 ± 2.4 (n = 2) and 5.4 ± 0.2 kcal mol–1 (n = 2), respectively. In contrast, an Ea of 14.9 ± 2.9 kcal mol–1 (n = 3) was measured for urea permeation in PM vesicles. Such a high temperature dependence may reflect a solubility-diffusion mechanism across the PM lipid bilayer, whereas the low Ea of urea and glycerol transport across the TP suggests a facilitated diffusion mechanism.

Mercury is a reversible blocker of most of aquaporins. Pre-treatment of TP vesicles with mercuric chloride decreased their urea permeability by up to 88 ± 3% (n = 2) after 5 min in the presence of 2 mm HgCl2 (Fig. 3). Pre-treatment of 15 min with 0.5 mm CuSO4 inhibited by 57 ± 10% (n = 3) the rate of urea transport across the TP (not shown). These results suggest the presence of solute-transporting proteinaceous pores in the TP vesicles.

Figure 3.

Mercury inhibition of urea transport in TP vesicles.

TP vesicles were pre-incubated for 5 min in the presence of the indicated HgCl2 concentration and assayed for urea transport as described in Fig. 1 Inhibition of urea influx relative to untreated membranes was 73% (0.2 mm HgCl2) and 90% (2 mm HgCl2).

Molecular cloning of Nt-TIPa, an MIP homologue of tobacco cells

It was previously shown that water channels confer a high, mercury-sensitive water permeability on TP vesicles from tobacco suspension cells (Maurel et al. 1997). The present data extend these observations, showing that the same vesicles exhibit a high, mercury-sensitive, permeability to small non-electrolytes. These findings led us to the idea that the same channel proteins, possibly MIP homologues, may mediate both water and solute transport in the tobacco TP. MIP genes expressed in tobacco cells were investigated by RT–PCR using degenerate oligonucleotides corresponding to the two NPA motifs typically conserved in this family (Park & Saier 1996). Sequence analysis of 15 independently cloned PCR fragments showed them to define seven distinct MIP-like sequences, three of them belonging to the PIP subfamily while the four others fall into the TIP subfamily. One of the TIP sequences, referred to as TIPa, is slightly divergent from the three others (Fig. 4a), which encode close homologues of δ-TIP, an aquaporin of Aradadopsis thaliana previously characterized as a selective water channel (Daniels et al. 1996). Because TIPa mRNA is strongly expressed in suspension cells (not shown), it may represent a good candidate to account for the selectivity profile of TP vesicles. The TIPa PCR fragment was used as a probe to screen a cDNA library from tobacco suspension cells. A cDNA clone of 924 bp, whose sequence perfectly matched that of the TIPa PCR fragment, was isolated and found to encode a putative protein of 247 amino acids (25.9 kDa) (Fig. 4b), subsequently named Nicotiana tabacum TIPa (Nt-TIPa) (GenBank accession no. AJ237751). Nt-TIPa exhibits the typical features of MIP channels, with in particular two NPA motifs and six predicted membrane-spanning domains (Fig. 4b,c).

Figure 4.

Molecular characterization of Nt-TIPa.

(a) Sequence relationship of TIP-like cDNA fragments amplified by RT–PCR from tobacco suspension cells. Amplification was performed using degenerate oligonucleotides corresponding to the two highly conserved NPA motifs. The dendrogram obtained with the PILEUP procedure of the GCG software package (Devereux et al. 1984) shows the relationship between four distinct TIP sequences identified in 8 of the 15 independent PCR fragments analysed. The number of PCR fragments corresponding to each sequence and the percentage of amino acid identity between two adjacent sequences are presented.

(b) Nucleotide and deduced amino acid sequence of Nt-TIPa cDNA (GenBank accession no. AJ237751). The sequences corresponding to the oligonucleotides used for PCR amplification are underlined. Two potential binding sites for copper at residues 76–78 and 90–92 are double-underlined. Putative phosphorylation sites are indicated by an asterisk.

(c) Kyte and Doolittle hydropathy analysis of the deduced amino acid sequence of Nt-TIPa with a window of 11 residues (Kyte & Doolittle 1982). Putative membrane spanning domains are numbered from I to VI.

Nt-TIPa is highly expressed in a TP-enriched fraction

Antibodies raised against a synthetic peptide corresponding to the last 13 amino acids of the Nt-TIPa C-terminal domain were used to probe oocyte proteins in immunoblot experiments. A 24 kDa protein was revealed in oocytes injected with sense Nt-TIPa cRNA but not in oocytes injected with the corresponding antisense cRNA (Fig. 5a). This result confirms the specificity of the anti-Nt-TIPa antibody. To investigate the localization of Nt-TIPa in the plant cell, the antibodies were used to probe tobacco membrane fractions. An immunoreactive peptide was detected at 24 kDa in a crude microsomal membrane preparation. Similar to the previously characterized TP enzymatic markers, pyrophosphatase and nitrate-sensitive ATPase (Maurel et al. 1997), this peptide was enriched in the most electronegative FFE fractions, in particular in the TP-enriched fraction typically used in stopped-flow experiments (Fig. 5b). Previous work showed that, in contrast, this fraction is strongly depleted (10–25% of the peak level) in the endoplasmic reticulum, Golgi and PM markers, cyt c reductase, latent UDPase and vanadate-sensitive ATPase, respectively (Maurel et al. 1997). These results point to the TP as the major localization of Nt-TIPa in tobacco suspension cells.

Figure 5.

Immunodetection of Nt-TIPa.

(a) Immunoblot analysis of total protein extracts from oocytes injected with antisense or sense Nt-TIPa cRNA. Equal amounts of protein corresponding to 0.1 oocyte were loaded in each lane.

(b) Immunoblot analysis of microsomal membranes and pooled membrane fractions (A–E) purified from tobacco cells by FFE. The corresponding FFE separation profile determined by absorbance at 280 nm is shown below. For immunoblot analysis, equal amounts of protein (10 μg) were loaded in each lane. The immunoreactive signal at approximately 42 kDa probably corresponds to an aggregate of Nt-TIPa which migrates as a dimeric form (Daniels et al. 1996). Molecular mass markers are indicated on the left and on the right.

Nt-TIPa transports glycerol and urea in addition to water

The transport properties of Nt-TIPa were investigated after expression in Xenopus oocytes and compared with those of plant aquaporin γ-TIP (Maurel et al. 1993) and bacterial glycerol facilitator GlpF (Maurel et al. 1994). Oocytes were injected with cRNAs transcribed in vitro from antisense Nt-TIPa, sense Nt-TIPa, GlpF or γ-TIP cDNA. After 3 days, water transport was assayed by swelling experiments in hypotonic conditions. Oocytes injected with Nt-TIPa sense cRNA showed a two- fold increase in osmotic water permeability (Pf = 17.2 ± 1.9 × 10–4 cm sec–1) compared to control oocytes injected with antisense Nt-TIPa cRNA (Pf = 8.0 ± 0.9 × 10–4 cm sec–1) (Fig. 6a). This stimulation was lower than that observed after γ-TIP expression which raised eightfold the oocyte water permeability (Pf = 55.9 ± 12.9 × 10–4 cm sec–1) (Fig. 6a). As previously described (Maurel et al. 1994), GlpF cRNA did not alter water transport in oocytes.

Figure 6.

Functional analysis of Nt-TIPa in Xenopus oocytes.

Water (a) and solute (b) transport were assayed at 20°C 3 days after cRNA injection. Results (mean ± SE) obtained in oocytes expressing tobacco Nt-TIPa, Arabidopsisγ-TIP and bacterial GlpF are presented as a percentage of the reference value measured in control oocytes injected with antisense Nt-TIPa cRNA. Reference values (100%) are: (a) Pf = 8.0 × 10–4 cm sec–1; (b) glycerol: 153 pmol min–1 oocyte–1; urea, 28 pmol min–1 oocyte–1. Results were pooled from several independent experiments and the total number of tested oocytes is indicated.

Uptake experiments with radiolabelled urea and glycerol were performed to investigate the solute transport activity of Nt-TIPa. When compared to control antisense Nt-TIPa oocytes, oocytes expressing sense Nt-TIPa cRNA exhibited an increased capacity for both urea and glycerol uptake, by factors of 3.9 and 1.5, respectively (significantly different from controls at P≥ 0.99) (Fig. 6b). We calculated that the stimulation in solute permeability specifically induced by Nt-TIPa corresponded to a net increase in permeability of 1.5 ± 0.3 × 10–6 and 0.2 ± 0.1 × 10–6 cm sec–1 for urea and glycerol, respectively. In the same experiments, GlpF enhanced glycerol uptake fivefold but had no effect on urea uptake (Fig. 6b).


Measurements of membrane permeability using isolated vesicles allow resolution of the transport features of well-defined subcellular membranes and are devoid of most artefacts generated at the cell and tissue levels by unstirred layers (Niemietz & Tyerman 1997; Verkman 1995). We previously used FFE in conjunction with stopped-flow spectrophotometry to characterize the water transport properties of PM and TP vesicles purified from tobacco suspension cells (Maurel et al. 1997). In the present work, we extend these studies and investigate the permeability of these purified membranes to small non-electrolytes. The permeability profiles determined here show that, as for water, tobacco TP and PM vesicles have strikingly different solute transport properties. When compared to PM vesicles, TP vesicles exhibited a 35–70-fold higher permeability to urea and were significantly more permeable than their PM counterparts to larger molecules such as glycine and proline. Whether such differential properties occur in living tobacco cells or in other plant membrane preparations remains to be determined. Nevertheless, the present report extends to small non-electrolytes the idea that the plant TP, similar to the PM, exhibits specific and highly specialized transport functions (Barkla & Pantoja 1996).

Aquaporin-mediated transport of small non-electrolytes in the plant TP

The solute permeability values determined on tobacco TP vesicles, in the case of urea in particular, surpass values reported with artificial membranes (Lande et al. 1995; Walter & Gutknecht 1986). In addition, the relative permeabilities of TP vesicles to glycerol and urea clearly deviate from values expected from the respective lipid–water partition coefficients of the two solutes (Priver et al. 1993; Walter & Gutknecht 1986). These observations, along with the weak temperature dependence of glycerol and urea transport, suggested that permeation of the two molecules may occur by facilitated diffusion across pores rather than by a solubility-diffusion mechanism across the lipid bilayer. The blockade of solute transport by mercury and copper further suggested the involvement of channel protein(s) in this path.

The transport of small non-electrolytes across plant membranes has been discussed in terms of permeation through pores since the early 1960s (Dainty & Ginzburg 1964a,b; Stadelmann 1969). Recently, a thorough analysis of reflection coefficients for small organic solutes in giant internodal cells of Chara has provided evidence for slippage of solutes across water channels (Schütz & Tyerman 1997; Steudle & Henzler 1995). However, the molecular structure of these pores remains unknown. The determination of a selectivity profile reminiscent of MIP solute facilitators in the tobacco TP, a membrane that, in addition, exhibits a high, channel-mediated water permeability, led us to suppose that aquaporin-like proteins could account for both water and solute transport in this membrane. More specifically, we focused our attention on Nt-TIPa, a novel tobacco MIP which slightly diverges in sequence from other TIPs previously characterized as selective water channels in other species (Daniels et al. 1996; Maurel et al. 1993, 1995). Functional expression of Nt-TIPa in Xenopus oocytes showed this protein to transport water, and also small solutes such as glycerol and urea, with a greater efficiency for the latter. These features correctly match the solute selectivity profile determined at the vesicle level. Consistent with this, MIP homologues may be major constituents of native plant membranes and the intense signal obtained after immunodetection of Nt-TIPa suggests that this protein is efficiently expressed in the tobacco cell TP. The pharmacology of water and solute transport also suggests parallels between TP vesicles and oocyte membranes expressing Nt-TIPa. Water transport in TP vesicles can be inhibited by mercury, and solute transport was blocked by both mercury and copper. Mercury is a common blocker of most aquaporin channels and inhibited Nt-TIPa-mediated water transport in the oocyte membrane. Also, the protein contains two Gly-Gly-His motifs at positions 76–78 and 90–92 in the first predicted cytoplasmic loop (B loop) which may serve as binding sites for copper. However, non-specific effects of both mercury and copper on the solute permeability of native oocytes precluded any further pharmacological characterization of Nt-TIPa-mediated solute transport in this system. Nevertheless, our data strongly suggest that Nt-TIPa significantly contributes to solute transport across the tobacco cell TP.

We found, however, that the permeability to solutes conferred on oocyte membranes by Nt-TIPa expression is much lower than that found in TP vesicles. This may simply reflect the fact that Nt-TIPa was much less efficiently expressed in the former than in the latter membranes. However, the water transport capacity of Nt-TIPa in oocytes, relative to its urea transport capacity, was moderate, suggesting that Nt-TIPa probably cannot account for the extremely high water permeability of the tobacco TP (Pf≈ 7 × 10–2 cm sec–1) (Maurel et al. 1997). We believe that other aquaporin isoforms may also contribute to water transport across the tobacco TP. A large number of MIP genes are indeed present in plants. Eleven TIP-like cDNAs have been reported in Arabidopsis (Weig et al. 1997), and four TIP homologues, at least, are expressed in the tobacco cell suspension. Plant MIPs whose transport activity was previously characterized in detail all functioned as selective water channels. Thus, it is likely that tobacco cells and probably other plant cells, similar to mammalian erythrocytes or renal collecting duct cells (Roudier et al. 1998; Sasaki et al. 1998), can express aquaporins with distinct selectivity profiles in the same membrane compartment. More generally, the identification of Nt-TIPa as a solute-transporting aquaporin of plants points to transport selectivity as another feature, along with specific cell localization and regulation properties, to rationalize the diversity of MIP proteins in these organisms.

Physiological relevance of non-electrolyte transport by aquaporins

In bacteria, water and glycerol transports are mediated by AqpZ and GlpF, respectively, two MIP homologues with distinct transport properties (Calamita et al. 1995; Maurel et al. 1994). In multi-cellular organisms, the interplay between numerous aquaporin isoforms with complex but specific selectivity profiles may also provide a basis for independent regulation of solute and water transport. The physiological significance of solute transport by mammalian aquaporins remains poorly understood. An exception to this may be human AQP9/AQP7L, which is specifically expressed in adipose tissues and could provide a path for glycerol export after lipolysis (Kuriyama et al. 1997).

Evidence is emerging from the present and other works that plant water channels can also transport solutes. For experimental reasons, none of the small molecules used so far for demonstrating this property has a clear physiological relevance (Henzler & Steudle 1995; Rivers et al. 1997) but these studies will surely foster future work on the physiology of non-electrolyte transport in plant membranes. For instance, the peribacteroid membrane of soybean symbiotic nodules exhibits a high permeability to formamide and glycerol which is probably due to the activity of NOD26 (Rivers et al. 1997). Carbon dioxide transport by erythrocyte AQP1 has recently been demonstrated (Nakhoul et al. 1998; Prasad et al. 1998), and a role for NOD26 in the exchange between the two symbiotic partners of gaseous substances such as CO2 and NH3 requires investigation (Tyerman et al. 1999). This function is however unlikely for TP aquaporins. In contrast, a transport of glycine was detected in the tobacco TP, corresponding possibly to a channel-like activity reported in barley mesophyll vacuoles (Goerlach & Willms-Hoff 1992). This would avoid a cytoplasmic accumulation of glycine and subsequent metabolic inhibition at high rates of photorespiration. Compatible solutes might also be relevant substrates for TP aquaporins. Numerous studies have described in detail the metabolism of compatible solutes in the response of plants to water deficit, but very little is known about the transport and compartmentation of these molecules in the plant cell (Bohnert et al. 1995; Leigh et al. 1981; McCue & Hanson 1990; Martinoia & Ratajczak 1997). We and colleagues have proposed that the differential permeability of the plant PM and TP to water provides a basis for efficient cytosol osmoregulation in the case of a rapid change in external water potential (Maurel et al. 1997; Niemietz & Tyerman 1997; Tyerman et al. 1999). Here we found that tobacco TP vesicles were much more permeable to proline than PM vesicles. A differential transport of solutes across the PM and TP may provide another mechanism for longer-term regulation of the respective volumes of the cytosol and vacuole. This idea was raised in an early work by Leigh et al. (1981) who suggested that in red beet the exchange of proline and glycinebetaine between cytoplasm and vacuole might be regulated and may play a role in the response of plants to moderate water stress.

In conclusion, our results support the idea that plant aquaporins, similar to their homologues in other organisms, have a dual function in water and solute transport. The physiology of small non-electrolyte transport in plant cells is poorly understood and the present study will surely stimulate future work in this research area. The identification of a novel solute-transporting aquaporin which clearly diverges from those previously characterized in other organisms also provides a valuable tool for investigating the molecular biophysics of water and solute transport by aquaporins.

Experimental procedures

Vacuolar and plasma membrane purification

A microsomal membrane fraction was isolated from exponentially growing tobacco (Nicotiana tabacum cv. Xanthi) suspension cells and separated by free-flow electrophoresis (FFE) as previously described (Maurel et al. 1997). Pooled membrane fractions enriched in PM or TP were centrifuged at 70 000 g for 36 min, resuspended in a minimal volume of FFE chamber buffer (Maurel et al. 1997), and stored at –80°C. The purity of the PM- and the TP-enriched fractions was assessed in a previous work (Maurel et al. 1997)

Fluorescence measurements of membrane vesicle volume

Stopped-flow light scattering was previously used to characterize water transport in membranes purified from tobacco suspension cells (Maurel et al. 1997). We found, however, that the light scattering signal is not adequate to follow solute transport in tobacco membrane vesicles. Changes in intra-vesicular solute concentration probably alter the refractive index of the vesicles (Verkman 1995) and this results in a complex light scattering response. To avoid this, a fluorimetric method was used in which osmotically induced variations of vesicle volume were followed by means of a vesicle-entrapped fluorophore with concentration-dependent self-quenching properties (Chen et al. 1988). Purified membrane vesicles were washed in 70 mm sucrose, 20 mm Tris–HEPES, pH 7.5, and centrifuged at 70 000 g for 20 min at 4°C. Membrane vesicles were incubated overnight at 4°C in the same medium supplemented with 10 mm 6-carboxyfluorescein (CF) (Sigma). Extra-vesicular CF was removed by one passage of the vesicle suspension through a Sephadex® G-25 M column (Pharmacia) equilibrated in 100 mm sucrose, 20 mm Tris–HEPES, pH 7.5. To minimize CF leakage, CF-loaded vesicles were then conserved on ice prior to solute permeability measurements. We checked that, under these conditions, the vesicle-specific fluorescence signal (see below) was stable for at least 2 h (data not shown).

The fluorescence signal of CF-loaded vesicles was characterized in a previously described SFM3 stopped-flow spectrophotometer (Biologic, Claix, France) (Maurel et al. 1997). Excitation was at 477 nm and emitted light was filtered with a 520 nm cut-off filter. When vesicles were submitted to hyperosmotic shocks of varying amplitude, it was found that the amplitude of the time-dependent decrease in fluorescence signal was linearly correlated to the expected relative change in vesicle volume as calculated from the imposed external osmolalities (data not shown). These results show that the vesicles behaved as osmometers and that CF fluorescence can be used as a direct reporter of the relative vesicle volume.

Determination of the solute permeability by efflux experiments

All operations were performed at 20°C. CF-loaded vesicles (50 μg protein ml–1) were equilibrated for 10 min in a medium containing 280 mm of the appropriate permeant solute, 100 mm sucrose, 20 mm Tris–HEPES, pH 7.5. Aliquots of vesicles (75 μl) were rapidly mixed in the stopped-flow apparatus (dead time < 6 ms), with an equal volume of an isotonic medium but devoid of permeant solute (380 mm sucrose, 20 mm Tris–HEPES, pH 7.5). Diffusion of the permeant solute through the membrane down its concentration gradient is accompanied by an osmotic efflux of water and results in vesicle shrinking. The time course of vesicle volume decrease is monophasic and reflected by a parallel decrease in fluorescence signal. Data from at least seven time courses were averaged and fitted to a single exponential function using the Simplex procedure of the Biokine software (Biologic). The membrane permeability to solute (Psol) was calculated using the following equation for solute diffusion (Priver et al. 1993; Verkman 1995):


where Jn is the net flow of permeant solute, n(t) is the number of molecules of permeant solute inside the vesicles at time t, and S is the average surface area of the vesicles. The concentrations in permeant solute inside and outside the vesicles are denoted by C(t) and Cout, respectively.

The quasi-osmotic equilibrium of the vesicles is described by:


where Iout and I(t) represent the concentration in impermeant solute, outside and inside the vesicle, respectively. Since the vesicle volume is negligible compared with the extra-vesicular volume, it was assumed that Iout and Cout remained constant throughout the experiment.

The conservation of the impermeant solute inside the vesicle can be written as:


where V(t) is the mean vesicle volume at time t. In addition




Volume variations were simulated using eqn 5 and the EXCEL software (Microsoft) for various arbitrary values of Psol between 10–4 and 10–7 cm sec–1, with the other parameters as in our experimental conditions. S and V(0) were calculated assuming that, at t = 0, PM and TP vesicles were spheres with average diameters of 154 and 320 nm, respectively (Maurel et al. 1997). The theoretical curves were fitted to a single exponential (AXOGRAPH, Axon Instruments, Foster City, California, USA), and the deduced rate constant was plotted against the imposed permeability. The calibration curve obtained was used to determine the solute permeability from the exponential rate constant experimentally determined under the same initial conditions. We verified that simulated kinetics mostly depended on Psol and were minimally changed when a reflection coefficient less than unity and a solvent drag term were introduced in eqn 1 (Verkman 1995).

Determination of the solute permeability by uptake experiments

The stopped-flow procedure was essentially the same as for efflux experiments except that vesicles containing 10 mm CF, 100 mm sucrose, 20 mm Tris–HEPES, pH 7.5, were mixed with an equal volume of a hypertonic medium containing 650 mm of the tested solute, 100 mm sucrose, 20 mm Tris–HEPES, pH 7.5. This resulted first in a rapid osmotic water efflux and shrinking of the vesicles. After this first process has been completed, solute diffusion into the vesicles predominates. Solute influx is accompanied by an osmotic water influx and vesicles re-swell to their initial volume. The overall process results in a biphasic time course of CF fluorescence changes. The second phase of the relaxation essentially depends on the membrane permeability for the solute and can be fitted by a single-exponential function using the Biokine software. For the solute permeability determination, the second (solute) phase was simulated according to eqn 5 and using the following assumption. Since the membrane vesicles are more permeable to water than to any solute by at least two orders of magnitude, solute influx is negligible during the water efflux phase. Mixture of the vesicle suspension with an equal volume of hypertonic medium (see above) leads to an initial 3.1-fold increase in external osmolality. This yields the following initial conditions for the solute phase: 310 mm sucrose, 62 mm Tris–HEPES, 31 mm CF inside the vesicles and 325 mm tested solute, 100 mm sucrose, 20 mm Tris–HEPES in the external medium.

Isolation and cloning of Nt-TIPa cDNA

The degenerate oligonucleotides 5′-GGIGGICA(C/T)ITIAA(C/T)CCIGCIGTNAC-3′ and 5′-GCIGGICC(A/G)AAI(C/G)(A/C/T)IC(G/T)ICGIGG(A/G)TT-3′ where N = A, T, C or G, were designed according to the two following amino acid motifs, GGH(I/V/L)NPAVT and NPAXX(F/L)GPR, which have been shown to be highly conserved among plant MIPs (Park & Saier 1996; Schäffner 1998). These oligonucleotides were used for PCR amplification of tobacco suspension cell cDNA (one cycle of 4 min at 94°C, 45 sec at 50°C and 1 min at 72°C, followed by 30 cycles of 45 sec at 94°C, 45 sec at 50°C and 1 min at 72°C). PCR products were cloned, and sequence analysis of 15 clones showed all of them to encode MIP homologues. Four clones corresponded to a TIP homologue, subsequently named Nt-TIPa. The corresponding PCR fragment (tipa) was labelled with [α32P]dCTP (Megaprime DNA Labelling system, Amersham, UK) and used as a probe to screen 700 000 phages from a λZAPII cDNA library from tobacco suspension cells (provided by Dr Claire Lurin, CNRS, Gif-sur-Yvette, France) according to the manual instructions (Stratagene). Hybridizations were performed in 5× SSC, 7% SDS, 10× Denhardt's solution and 20 mm phosphate buffer, pH 7 at 65°C (Sambrook et al. 1989). Membranes were washed three times at 65°C, first in 2× SSC and 0.5% SDS, then in 1× SSC and 0.5% SDS, and finally in 0.5× SSC and 0.5% SDS. Excised pBluescript plasmids from several positive phage clones were characterized by restriction enzyme analysis using EcoRI and ApaI. The longest cDNA fragment corresponding to tipa was sequenced (dye Terminator, Applied Biosystems 373A, Perkin Elmer). Sequence data were analysed using the Genetics Computer Group (Madison, Wisconsin, USA) version 8.0 software package (Devereux et al. 1984). For further expression in Xenopus oocytes, the Nt-TIPa cDNA insert was cloned as an EcoRV–BamHI fragment in pT7TS between the 5′ and 3′ untranslated regions of a β-globin gene of Xenopus (Abrami et al. 1994). The Nt-TIPa cDNA was cloned similarly in pT7TS in an antisense orientation as an EcoRV–SpeI fragment. The constructs were used for in vitro cRNA transcription as previously described (Maurel et al. 1993).

Cell localization of Nt-TIPa

A synthetic peptide corresponding to the C-terminus of Nt-TIPa (residues 235–247; VRTHVPLPSDESF) was coupled to keyhole limpet haemocyanin (Neosystem, Strasbourg, France). Rabbits were boosted three times at 4–6 week intervals by subcutaneous injection of 300 μg of this conjugate in Freund's adjuvant. For immunoblot analysis, proteins from Xenopus oocytes or from plant membrane fractions purified by FFE were separated by SDS–PAGE and transferred to Immobilon PVDF membranes (Millipore, Bedford, Massachusetts, USA). Proteins were probed with the crude anti-Nt-TIPa antiserum and visualized using goat anti-rabbit IgG coupled to alkaline phosphatase (Biosys, Compiègne, France) with 5-bromo-4-chloro-3 indolyl phosphate and nitrobluetetrazolium reagents (BioRad, Hercules, California, USA).

Water and solute transport assays in Xenopus oocytes

Stage V and VI Xenopus oocytes were isolated as described previously (Maurel et al. 1993), injected with 50 ng of in vitro transcribed capped cRNA (Maurel et al. 1993), and incubated at 19°C in Barth's solution (88 mm NaCl, 1 mm KCl, 0.41 mm CaCl2, 2.4 mm NaHCO3, 0.33 mm Ca(NO3)2, 0.82 mm MgSO4, 1 mg l–1 gentamycin, 10 mm HEPES, pH 7.4). After 72 h, the oocytes were transferred into Barth's solution diluted fivefold with distilled water and subsequent changes in cell volume were followed using a video camera. The osmotic water permeability coefficient, Pf, was calculated from:


where V is the oocyte volume at time t, and with initial oocyte volume Vo = 9 × 10–4 cm3, oocyte surface area S = 0.045 cm2 and the molar volume of water Vw = 18 cm3 mol–1. Osmin and Osmout represent the osmolality inside and outside the oocyte, respectively.

For solute transport assays, oocytes were incubated in a solution containing radiolabelled solute (150 mm[14C]glycerol or 20 mm[14C]urea; 10 μCi ml–1) and complemented with a partially diluted Barth's solution to reach iso-osmolar conditions (200 mosmol kg–1). After 10–30 min incubation at room temperature, oocytes were rinsed four times in 100–200 ml of ice-cold Barth's solution and lysed overnight in 2% SDS before scintillation counting. We checked that solute uptake was linear with time over the incubation period.


We thank Frédérique Tacnet and Gabriela Amodeo for helpful discussions and critical reading of the manuscript. We also thank Claire Lurin for providing the tobacco cDNA library, and Marck Cock for the gift of MIP degenerate oligonucleotides. This work was supported by the Centre National de la Recherche Scientifique (UPR0040), the Commissariat à l’Energie Atomique (Direction des Sciences du Vivant, segment 28), and the Commission of the European Union (Contract BIO4-CT98-0024).


  1. GenBank accession number AJ237751 (Nt-TIPa).

  2. Note added in proofThe water and glycerol properties of Nodulin 26 were recently demonstrated by Dean et al. (1999) Biochemistry, 38, 347–353.