- Email: [email protected]
Functional reconstitution and characterization of AqpZ, the E. coli water channel protein1
Article No. jmbi.1999.3032 available online at http://www.idealibrary.com on
J. Mol. Biol. (1999) 291, 1169±1179
Functional Reconstitution and Characterization of AqpZ, the E. coli Water Channel Protein Mario J. Borgnia1, David Kozono1, Giuseppe Calamita4, Peter C. Maloney3 and Peter Agre1,2* 1
Department of Biological Chemistry 2
Department of Medicine
3
Department of Physiology Johns Hopkins University School of Medicine, 725 N Wolfe Street, Baltimore, MD 21205-2185, USA 4
Dipartimento di Fisiologia Generale ed Ambientale University of Bari, 165/A via Amendola, 70126, Bari, Italy
Understanding the selectivity of aquaporin water channels will require structural and functional studies of wild-type and modi®ed proteins; however, expression systems have not previously yielded aquaporins in the necessary milligram quantities. Here we report expression of a histidine-tagged form of Escherichia coli aquaporin-Z (AqpZ) in its homologous expression system. 10-His-AqpZ is solubilized and puri®ed to near homogeneity in a single step with a ®nal yield of 2.5 mg/l of culture. The histidine tag is removed by trypsin, yielding the native protein with the addition of three N-terminal residues, as con®rmed by microsequencing. Sucrose gradient sedimentation analysis showed that the native, solubilized AqpZ protein is a trypsin-resistant tetramer. Unlike other known aquaporins, AqpZ tetramers are not readily dissociated by 1 % SDS at neutral pH. Hydrophilic reducing agents have a limited effect on the stability of the tetramer in 1 % SDS, whereas incubations for more than 24 hours, pH values below 5.6, or exposure to the hydrophobic reducing agent ethanedithiol cause dissociation into monomers. Cys20, but not Cys9, is necessary for the stability of the AqpZ tetramer in SDS. Upon reconstitution into proteoliposomes, AqpZ displays very high osmotic water permeability (pf 5 10 10ÿ14 cm3 sÿ1 subunitÿ1) and low Arrhenius activation energy (Ea 3.7 kcal/mol), similar to mammalian aquaporin-1 (AQP1). No permeation by glycerol, urea or sorbitol was detected. Expression of native and modi®ed AqpZ in milligram quantities has permitted biophysical characterization of this remarkably stable aquaporin tetramer, which is being utilized for high-resolution structural studies. # 1999 Academic Press
*Corresponding author
Keywords: aquaporin; bacterial; water transport; channel structure; gene family
Introduction Water movement across biological membranes is not fully explained by simple diffusion through the lipid bilayer. Observations of highly water-permeable tissues led to the development of theoretiAbbreviations used: AQP1, aquaporin-1; AqpZ, aquaporin-Z; dodecylmaltoside, n-dodecyl-b-Dmaltoside; GlpF, glycerol facilitator protein; IPTG, isopropyl-b-D-thiogalactoside; Mops, 3-[nmorpholino]propanesulfonic acid; Ni-NTA, nickelnitrilotriacetic acid; NMDG, n-methyl-D-glucamine; octylglucoside, n-octyl-b-D-glucopyranoside; PMSF, phenylmethylsulfonyl ¯uoride; PVDF, polyvinylidene di¯uoride. E-mail address of the corresponding author: [email protected] 0022-2836/99/351169±11 $30.00/0
cal tools for the study of osmotic and diffusional processes across membranes and pores, and prompted the long pursuit of the molecular entity responsible for water transport (Finkelstein, 1987). The ®rst characterized water channel, aquaporin-1 (AQP1), was serendipitously identi®ed and puri®ed from human red blood cells (Denker et al., 1988; Smith & Agre, 1991). Expression of AQP1 mRNA in Xenopus oocytes (Preston et al., 1992) led to its recognition as the erythrocyte water channel, a conclusion con®rmed by reconstitution of highly puri®ed AQP1 into proteoliposomes (Zeidel et al., 1992, 1994; van Hoek & Verkman, 1992). Sequence analysis related AQP1 to the Escherichia coli glycerol facilitator protein (GlpF) and to the major intrinsic protein of mammalian lens (MIP, now designated AQP0). Screening of mRNA # 1999 Academic Press
1170 libraries prepared from different tissues permitted identi®cation of at least ten mammalian homologs. Multiple homologs have also been identi®ed in other vertebrates (Abrami et al., 1994; Ma et al., 1996), plants (Weig et al., 1997), yeast (Bonhivers et al., 1998) and bacteria (Calamita et al., 1995). Some of the identi®ed proteins are highly speci®c for water transport (``orthodox aquaporins'') whereas others are permeated by water, glycerol and other small, uncharged molecules (``aquaglyceroporins''). The amino acid sequences of aquaporins are approximately 30 % identical; however, the conserved residues are distributed in de®ned clusters, which themselves are absolutely conserved throughout the family (Park & Saier, 1996). All homologs are predicted to contain six transmembrane-spanning segments formed by the tandem duplication of three transmembrane domains joined in obverse symmetry. The N and C termini reside in the intracellular space (Smith & Agre, 1991), leaving three extracellular and two intracellular connecting loops (Figure 1). Loops B and E share the most highly conserved residues, including the Asn-Pro-Ala motifs, suggesting that they
Functional Reconstitution of AqpZ
may be directly involved in the selectivity ®lter of the channel. Because several residues in these loops are hydrophobic, it was proposed that loops B and E fold into the membrane-spanning region of each subunit and are surrounded by the six transmembrane segments forming a structure referred to as the ``hourglass'' (Jung et al., 1994). Although each subunit apparently contains a single aqueous pore, AQP1 and other aquaporins are believed to exist as tetramers in mild detergents (Smith & Agre, 1991) and in the native membrane (Jung et al., 1994; Verbavatz et al., 1993; Walz et al., 1994). Although certain well-de®ned residues distinguish aquaporins from aquaglyceroporins, the molecular determinants of the functional differences between the two groups are far from clear. Their elucidation awaits the structural and biophysical characterization of multiple homologs, chimeras and mutants. To date, most functional and comparative studies have been carried out with the oocyte expression system. Single channel water permeabilities are hard to determine in this system due to the lack of an accurate method for quanti®cation of plasma membrane protein
Figure 1. Predicted primary sequence and membrane topology of 10-His-AqpZ. To facilitate the puri®cation of the overexpressed protein, the sequence of AqpZ was modi®ed to include a 23-residue N-terminal tag containing ten consecutive histidine residues. The residue corresponding to the con®rmed trypsin cleavage site is indicated (arrowhead). Numbers indicate the residues that were subjected to mutagenesis. Highly conserved residues in the aquaporin family are marked as ®lled shapes. Basic residues are represented as diamonds, acidic residues appear as triangles and neutral residues are shown as circles. The putative transmembrane domains were determined by submitting an alignment including the predicted sequence of AqpZ to the Predict Protein e-mail server ([email protected]).
1171
Functional Reconstitution of AqpZ
expression, the possible contribution of endogenous proteins, and differences in membrane traf®cking. Moreover, the protein expression levels achieved in oocytes are several orders of magnitude below the levels needed for structural studies. Thus, red cell AQP1 is the only member of the aquaporin water channel family with a structural Ê (Walz et al., 1997; Li et al., model resolved at 6 A 1997; Cheng et al., 1997). Because AQP1 is puri®ed from human red blood cells, structure-function analysis is currently limited to the wild-type molecule. Thus, development of expression systems for the puri®cation of large quantities of wild-type and modi®ed aquaporins is essential for the understanding of the molecular basis of the biophysical properties of aquaporin. AqpZ, the aquaporin from E. coli (Calamita et al., 1995), was identi®ed as a good candidate for puri®cation following overexpression in bacteria (Calamita et al., 1998) and may be a good substrate for structure-function studies. Here, we report the functional reconstitution and characterization of highly puri®ed native and modi®ed AqpZ forms.
Results
Figure 2. Electrophoretic analysis of puri®ed 10-HisAqpZ. Aliquots of puri®ed material were incubated for one hour at room temperature in denaturing loading buffer containing 1 % SDS and 143 mM b-mercaptoethanol. (a) Silver staining of a sample resolved in 15 % SDS-PAGE. (b) Dependency of the apparent molecular mass on acrylamide concentration. Samples were resolved in gels of different polymer content and Coomassie blue stained. The apparent molecular mass was calculated by comparison with standards.
Expression and purification of AqpZ Puri®cation of membrane transport proteins has been greatly facilitated by construction of recombinants containing peptide domains such as poly(His), which are used for af®nity-chromatography. The AqpZ protein is believed to contain only short N-terminal and C-terminal cytoplasmic domains, so the aqpZ DNA was cloned into the pTrc10HisEcoRI plasmid (Tamai et al., 1997) yielding an AqpZ polypeptide with a 23-residue N-terminal extension containing ten consecutive His residues (Figure 1). E. coli cultures transformed with pTrc10HisAqpZ grew at the same rate as control bacteria transformed with the plasmid pTrc10HisEcoRI, encoding a truncated 25 residue polypeptide. In contrast, addition of 1 mM IPTG caused an almost complete arrest of growth of pTrc10HisAqpZ-transformed cells but was without appreciable effect on control bacteria, suggesting that the expression of 10-His-AqpZ is toxic to the cells. To generate reasonable levels of expression, cultures were propagated to higher cell densities (A 1.4 to 1.8) prior to induction. 10-His-AqpZ was puri®ed from the dodecylmaltoside-solubilized membrane fraction by Ni af®nity chromatography. Aliquots of the eluted material were diluted in sample buffer (12 % glycerol, 1 % SDS, 140 mM b-mercaptoethanol and 70 mM TrisHCl (pH 6.8)) for one hour at room temperature and were analyzed by SDS-PAGE with 15 % polyacrylamide slabs. A major molecular species of 80 kDa and a less-abundant species of 23 kDa were revealed by silver staining of the gels (Figure 2(a)). One liter of culture yielded up to 2.5 mg of puri®ed protein, depending on the density of the culture when induced.
Known aquaporins migrate as 25-30 kDa monomers during SDS-PAGE. To evaluate the molecular mass of the 80 kDa species, the electrophoretic behavior in SDS-PAGE was studied at different concentrations of acrylamide. The analysis revealed a linear relationship between the apparent molecular mass and the acrylamide concentration, with the higher molecular mass band exhibiting faster mobility in gels of lower polymer content (Figure 2(b)). This aberrant electrophoretic behavior is a known characteristic of certain hydrophobic proteins (Helenius & Simons, 1975) but precludes identi®cation of the apparent SDS-stable oligomeric state of 10-His-AqpZ. Macromolecular assembly of AqpZ To further characterize the eluted material, dodecylmaltoside-solubilized samples were analyzed by non-equilibrium centrifugation on a sucrose gradient, followed by SDS-PAGE analysis of fractions. Interestingly, both the 80 kDa and the 23 kDa species coincided in exactly the same fractions of the gradient. The apparent sedimentation coef®cient for the peak fraction was estimated at 5.7 by comparison of AqpZ mobility and the mobility of several standard proteins (Figure 3). This value is similar to that obtained for octylglucoside-solubilized AQP1 (Smith & Agre, 1991), which is known to be a tetramer (Verbavatz et al., 1993; Walz et al., 1994). Thus, our data suggest that AqpZ exists as tetramer when solubilized in mild detergents like dodecylmaltoside as well as in the
1172
Figure 3. Non-equilibrium sedimentation coef®cient of puri®ed AqpZ. Solubilized, puri®ed AqpZ was layered on top of a 5 % to 20 % continuous sucrose gradient containing 1.5 % dodecylmaltoside and centrifuged at 140,000 g for 18 hours. A total of 35 fractions were collected and resolved by SDS-PAGE. The distribution of the 80 kDa species was coincident with that of the 23 kDa species. The sedimentation coef®cient (sw,20) of AqpZ was determined by comparison with the following standards: cytochrome c (1.8), catalase (2.9), b-amylase (4.3), bovine serum albumin (8.9) and carbonic anhydrase (11.2).
harsh detergent SDS, which usually unfolds proteins and dissociates the subunits. Dissociation of the tetramer was attempted under several different conditions. Incubation of the samples in SDS-PAGE at 37 C or higher temperatures in the presence of chaotropic (8 M urea or guanidinium chloride) or hydrophilic reducing agents (140 mM b-mercaptoethanol or 100 mM dithiothreitol) caused aggregation of AqpZ, which failed to enter the separating SDS-PAGE gels (not shown). The effects of various reducing agents were studied (Figure 4). Increasing concentrations of the hydrophilic reducing agent b-mercaptoethanol caused a shift in the oligomerization of the protein towards the low molecular mass species but ultimately caused aggregation of the protein (Figure 4). No change in the distribution was observed with dithiothreitol in a range of concentrations up to 1 M (Figure 4). Incubation with the relatively water-insoluble reducing agent ethanedithiol at a nominal concentration of 60 mM led to almost complete dissociation of the high molecular mass species (Figure 4), con®rming its identity with the 23 kDa 10-His-AqpZ. Differential reactivity with these reducing agents suggests that the target residue may be located in a hydrophobic environment. Prolonged incubation (>24 hours) in 1 % SDS-containing sample buffer resulted in dissociation of the tetramer even without reducing agents (Figure 4). When sample-loading buffers covering a range of pH were used to incubate the
Functional Reconstitution of AqpZ
Figure 4. Dissociation of the SDS-resistant 10-HisAqpZ tetramer. Top: Effect of reducing agents. Puri®ed protein was incubated for one hour at room temperature in 100 ml aliquots of gel loading buffer (pH 6.8), to which appropriate volumes of dithiothreitol, b-mercaptoethanol or ethanedithiol were added to reach the indicated concentrations. Samples were then resolved in 14 % SDS-PAGE. Bottom left: Long-term dissociation. Puri®ed 10-His-AqpZ was incubated in gel loading buffer (pH 6.8) without reducing agents for the indicated times prior to electrophoresis in 14 % SDS-PAGE. Bottom right: Dissociation by pH. Aliquots of 10-His-AqpZ were incubated in gel loading buffer at the indicated pH for ten minutes, followed by electrophoresis in 14 % SDS-PAGE.
protein at room temperature prior to SDS-PAGE (Figure 4), a rapid transition from tetramer to monomer occurs below pH 5.6 (12 % glycerol, 1 % SDS and 100 mM sodium acetate). The dissociation induced by pH and SDS was not reversed by returning to neutral pH (data not shown). Moreover, the dissociation by pH was not due to the presence of the histidine tag, since experiments carried out on proteolytically cleaved 10-His-AqpZ exhibited the same behavior (see below). Digestion of octylglucoside-solubilized 10-HisAqpZ with trypsin was undertaken to remove the poly(His) tag. Controlled proteolysis with a tenfold molar excess of trypsin revealed a cleavage pattern comprising three intermediate species and a ®nal product (Figure 5(a)). Analysis of the 10-His-AqpZ open reading frame predicts the existence of 11 basic residues, including a potential trypsin cleavage site between the poly(His) tag and the initial Met of AqpZ (Figure 1). Stepwise removal of the N-terminal extension from each subunit of a 10-His-AqpZ tetramer may explain the stepwise degradation observed in the gels. The analysis of digested protein with an anti-His antibody revealed the presence of four distinct species (Figure 5(b)), presumed to correspond to tetramers
Functional Reconstitution of AqpZ
Figure 5. Tryptic removal of the 10-His tag. Puri®ed 10-His-AqpZ (1 mg) was incubated for two hours with trypsin. Lanes a-f: 0, 0.12, 0.37, 1.1, 3.3 and 10 mg of trypsin. The digested material was resolved by 10 % SDS-PAGE. Two different gels run in parallel were stained with (a) Coomassie Brilliant Blue and or (b) transferred to nitrocellulose and probed using a monoclonal antibody recognizing poly(His) (Qiagen).
containing four, three, two or one N-terminal 10-His tags. The ®fth, lowermost band visible with Coomassie Blue staining was not detected by antiHis immunoblot, indicating that it represents the tetramer composed by four monomers lacking the 10-His tag at the N terminus. N-terminal protein microsequencing of the lowermost band transferred to PVDF membranes revealed the sequence H-E-F-M-F-R, corresponding to residues ÿ3 to 3 of the native polypeptide. Thus, the band results from enzymic hydrolysis at a single Arg residue. In contrast, no cleavage was observed when the reconstituted protein was exposed to trypsin (see below), suggesting that the unique cleavage site is protected by the lipid bilayer.
1173 atus. Liposomes and proteoliposomes were rapidly subjected to an outwardly oriented osmotic gradient, and the time course of change in light scattering (lem 600 nm) was recorded. The changes in scattered light were proportional to the gradient imposed for both liposomes and proteoliposomes, indicating the regular osmotic behavior of the system. Owing to changes in ¯uidity of the lipid bilayer, basal water permeability of liposomes increases greatly with temperature. Therefore, to improve the signal to noise ratio, most measurements were performed at 6-8 C. When an osmotic gradient of 284 mosmol/l was imposed, the water permeability (Pf) for 10-His-AqpZ in membranes was proportional to the protein-to-lipid ratio up to 1:100 (Figure 6). Proteoliposomes formed with higher ratios of 10-His-AqpZ to lipid did not exhibit proportionally increased rates of equilibration. At a ratio of 1:200 (50 mg of protein in 10 mg of lipid) the observed Pf was markedly increased (Pf 0.0283 cm/second at 6.5 C) as compared to liposomes (Pf 0.0015 cm/second at 6.5 C).
Reconstitution of AqpZ into proteoliposomes Direct biophysical evaluation of transport functions may be performed after reconstitution of the puri®ed proteins into proteoliposomes. Two different methods, rapid dilution and dialysis, both permitted reconstitution of the puri®ed 10-His-AqpZ into proteoliposomes (see Materials and Methods); however, dialysis yielded higher and more reproducible recovery of the reconstituted material. The order in which the different components were added to the reconstitution mixture was found to be critical. To prevent aggregation and precipitation, the puri®ed AqpZ is placed in a solution of high detergent concentration before addition of the lipids. The diameter of the reconstituted proteoliposomes (liposomes) was 122(20) nm as measured by electron microscopy. Permeability of AqpZ proteoliposomes Water movement across the membrane of control liposomes and proteoliposomes reconstituted with 10-His-AqpZ at different protein-to-lipid ratios was analyzed using a stopped-¯ow appar-
Figure 6. Functional reconstitution of AqpZ. Proteoliposomes reconstituted at different protein concentrations (25, 50, 100, 200 or 400 mg protein in 10 mg lipids) or control liposomes were abruptly exposed to a twofold increase in osmolarity. Light scattering was monitored in a stopped-¯ow apparatus. The reduction in volume due to water ef¯ux caused an increase in light scattering. The collected data were normalized to ®t between zero and unity. The traces are the average of ten or more individual plots at each protein to lipid ratio, and are displaced on the y-axis for clarity. Inset: Pf values for each 10-His-AqpZ to lipid ratio. Rate constants were determined by curve ®tting of the plots to a single order exponential. Pf values were calculated as described in Materials and Methods.
1174 Assuming total incorporation of protein into proteoliposomes, a maximum of 125 monomers per vesicle of 10-His-AqpZ is estimated. Thus, the calculated osmotic water permeability (pf) for each AqpZ subunit at this protein to lipid ratio is 510 10ÿ14 cm3/second, close to the value obtained for AQP1 (11.7 10ÿ14 cm3/second) (Zeidel et al., 1992). Proteoliposomes reconstituted with 10-His-AqpZ after cleavage of the His tag by trypsin treatment were indistinguishable from those containing the intact tagged protein (not shown). Measurements of Pf at pH 5, 6 and 7 were indistinguishable. Water permeability measurements were performed at different temperatures on both proteoliposomes (1:100 protein-to-lipid ratio) and control liposomes (Figure 7). The Arrhenius activation energies calculated from the plot are consistent with water movement through a channel for AqpZ (Ea 3.7 kcal/mol), as opposed to diffusion across the lipid bilayer of liposomes (Ea 16.9 kcal/mol). The permeability of 10-His-AqpZ-containing proteoliposomes to small solutes was also measured (see Materials and Methods), but no difference was observed between proteoliposomes and liposomes for transmembrane movement rates for glucose, sorbitol, sucrose or mannitol (not shown) as well as for glycerol and urea (Table 1). The permeability of the solutes is determined by following the movement of water (see Materials and Methods). Unlike in the case proteoliposomes, where the permeability of water is much higher than that of formamide, the permeability of the latter will be underestimated in the liposomes where
Figure 7. Activation energy of water ¯ow across proteoliposomes. The kinetics of water transport were measured in both proteoliposomes containing reconstituted 10-His-AqpZ (circles) and control liposomes (®lled circles) at different temperatures in the range from 284.1 to 303.5 K. Rates were determined by curve ®tting from the average of ten measurements of water movement at each temperature. The natural logarithms of the rates are plotted against the reciprocal of the temperature. The Arrhenius activation energy (EA) was determined from the slope of linear regression of the plots.
Functional Reconstitution of AqpZ Table 1. Permeability of AqpZ to uncharged small solutes Rate constant
Glycerol Urea Formamide
Liposomes (X S.D.)
Proteoliposomes (X S.D.)
0.166 0.017 0.0962 0.0362 8.95 0.43
0.116 0.0142 0.0634 0.0255 11.45 0.32
S.D., standard deviation.
water movement is the limiting factor. Therefore, the small (1.33-fold) difference observed for formamide (Table 2) may be attributed to this phenomenon and is minor as compared to the 50-fold increase in permeability observed for water at similar protein:lipid ratio.
Expression of 10-His-AqpZ mutants Proteoliposomes reconstituted with puri®ed wild-type red cell AQP1 have been studied extensively (Zeidel et al., 1992, 1994; van Hoek & Verkman, 1992). Characterization of other wildtype or mutant aquaporins has been restricted to studies of proteins expressed in Xenopus laevis oocytes, a system with known limitations. Thus, a major goal of this study was to establish a system suitable for expression and puri®cation AqpZ variants. In the cysteine-less mutant, the two Cys residues in positions 9 and 20 of native AqpZ were replaced by Ser residues. Sedimentation analysis showed that the cysteine-less variant retains its tetrameric structure when solubilized in the mild detergent dodecylmaltoside at 1.5 % (Figure 3). When analyzed by SDS-PAGE, the monomeric form is predominant (Figure 8). Similar results were obtained for the Cys20-Ser variant, while the Cys9-Ser mutant largely remained an 80 kDa complex in SDS (Figure 8). In spite of this difference in SDS stability, the water permeability of AqpZ was not altered by modi®cation of both Cys residues (Figure 8). The recognized mercurial inhibition of AQP1 is due to the presence of the residue Cys189 preceding the second Asn-Pro-Ala motif (Preston et al., 1993). When inhibition of AqpZ proteoliposomes was attempted, no inhibition was observed in 1 mM HgCl2 (not shown). When the corresponding residue in AqpZ, Thr183, is replaced by a Cys, the stability in SDS is like that of wild-type AqpZ (Figure 8). The water permeability of the Thr183Cys mutant is equivalent to the wild-type AqpZ (Figure 8) and is not inhibited by 1 mM HgCl2 (not shown). In contrast, mutation of the highly conserved Arg in position 189 into Val or Ser resulted in inactivation of the channel (Figure 8). Despite differences in SDS stability or water permeability, sedimentation analysis in the mild detergent dodecylmaltoside at 1.5 % revealed that all mutants
Functional Reconstitution of AqpZ
Figure 8. Analysis of 10-His-AqpZ mutants. Top: Samples of puri®ed modi®ed 10-His-AqpZ were resolved in a 14 % SDS-PAGE. Bottom: Rate constants for water movement in proteoliposomes reconstituted with puri®ed wild-type or mutant 10-His-AqpZ variants.
exhibited mobility similar to that of wild-type AqpZ protein (not shown).
Discussion The key to understanding how aquaporins are selectively permeated by water resides in the threedimensional structure of the channel. So far, structural studies of aquaporins are limited by the lack of means for producing milligram quantities of new homologs or modi®ed variants of the protein. Attempts to express different wild-type and mutant aquaporins in heterologous systems have previously been undertaken. The Xenopus oocyte expression system is often limited by improper traf®cking of mutant aquaporins to the plasma membrane or insuf®cient levels of expression (Jung et al., 1994). Small amounts of AQP1 (Laize et al., 1995), AQP2 (Coury et al., 1998) and AQPcic (Lagree et al., 1998) have been expressed in yeast. Native and His-tagged AQP4 were puri®ed from recombinant-baculovirus-infected Sf9 insect cells (Yang & Verkman, 1997); however, the highest level of expression (0.11 mg/l of culture for AQP4) is insuf®cient for high-resolution structural studies, and the His-tagged form was reported to be inactive. Here we describe an expression system per-
1175 mitting puri®cation of wild-type and site-directed modi®ed AqpZ. The yield obtained is 20 times higher than the one obtained for AQP4 in the baculovirus system. We show that the puri®ed AqpZ is active, with unit permeability values that are similar or higher than those reported for AQP1. The AqpZ expression-reconstitution system was used to characterize the protein biochemically and biophysically. Several observations indicate that AqpZ is very hydrophobic and appears to be tightly packed. Extended tryptic digestion failed to achieve cleavage at any of the amino acids within the native structure of solubilized, His-tagged AqpZ, and no cleavage was observed in the reconstituted protein. Digestion of 10-His-AqpZ with protease V8 from Staphylococcus aureus followed by mass spectrometry analysis by MALDI-TOF con®rmed the resistance of the native core of the protein upon solubilization with mild detergents (data not shown). These results suggest that most of the structure of AqpZ is effectively shielded from protease activity by either the lipid bilayer or the solubilizing detergent. The resistance of the tetramer to dissociation and its abnormal migration pattern observed during SDS-PAGE implies that folding is at least partially preserved even in the presence of SDS. As stated above, sedimentation analysis con®rmed that octylglucoside and dodecylmaltosidesolubilized AqpZ is tetrameric, and this is corroborated by electron-microscopy (Ringler et al., 1999, accompanying paper). Thus, the SDS-resistant species detected by electrophoresis is most likely an undissociated tetramer rather than a result of de novo aggregation. The tight packing of the protein may explain the unusual stability of the reconstituted samples, which retained 100 % activity after six months of storage at 4 C (data not shown). This property makes AqpZ a particularly well-suited protein for structural studies. The two cysteine residues in AqpZ, Cys9 and Cys20 are predicted to be located at the two opposite ends of the ®rst transmembrane segment. Mutant forms of AqpZ including Cys20-Ser, Cys9Ser, and the double mutant, all exist as tetramers when analyzed by sedimentation in mild detergent (octylglucoside or dodecylmaltoside) and exhibit wild-type permeability when reconstituted into proteoliposomes. The resistance to dissociation by SDS-PAGE distinguishes between the different cysteine mutants even though both mutants remain functional. Cys20 is necessary and suf®cient to stabilize the tetramer, since it modi®cation causes the protein to migrate primarily as a monomer during SDS-PAGE, whereas modi®cation of Cys9 has no apparent effect (Figure 8). Dissociation caused by the reducing agent ethanedithiol of wild-type and Cys9-Ser tetramers argues that the oxidation state of Cys20 may be critical to the stability of the tetramer in SDS. Unlike wild-type AqpZ, octylglucoside or dodecylmaltoside-solubilized Cys20-Ser precipitates when stored for long periods at 4 C (not shown). The existence of disul®de bonds between subunits is not consistent with
1176 the observed rapid dissociation at low pH and the long-term dissociation at neutral pH in the presence of 1 % SDS. Dissociation of the tetramer in SDS may be explained independently for each condition as minor conformational changes induced by pH, steric interactions due to serine substitution of Cys20, the oxidation state of Cys20, or other unidenti®ed residue(s). Non-covalent tetrameric assemblies of other bacterial transport proteins have also been found to be stable in SDS (le Coutre et al., 1998). The coordination of divalent cations by four cysteine residues as a fold-stabilizing feature has been described recently (Huse et al., 1998; Bixby et al., 1999). This type of structure is sensitive to all three parameters affecting AqpZ tetramer stability in SDS, including pH, electrochemical reduction of the coordinated metal, and mutagenesis of Cys20. Thus, the idea that Cys20 in each subunit stabilizes the tetramer is appealing; however, attempts to reduce the tetramer by incubation in chelators have so far been unsuccessful (not shown). Existence of a cysteine-mediated intersubunit interaction will require involvement of the ®rst transmembrane segment of each monomer in protein-protein interaction close to the center of the tetramer rather than facing the lipid phase. The availability of large quantities of puri®ed wild-type and mutant forms of AqpZ may now permit highresolution structural studies that could clarify this. The role of the highly conserved Arg residue that follows the second NPA in most aquaporins was explored. The Arg189 in loop E was modi®ed into Val, the residue located in the homologous position relative to the NPA in loop B and in loop E in a subset of plant aquaporins. This mutation was initially undertaken with the aim of overcoming the negative selectivity of AqpZ to protons. This and other substitutions of Arg189 led to loss of activity, without change in the tetrameric state of the solubilized mutants, as judged by the sedimentation in mild detergents (not shown), whereas the stability of the protein in SDS was reduced (Figure 8). Thus, although Arg189 is presumed to reside in the hourglass close to the center of each monomer, the loss of activity is accompanied by perturbation of the structure at the interface between monomers. This result is consistent with the prediction that highly conserved residues in the hourglass pore-forming domain of the protein have a major function in holding the tetramer together (Mathai & Agre, 1999). The largest bene®t from the AqpZ expressionreconstitution system may be the availability of large amounts of wild-type or site-directed mutant forms of an aquaporin for 2D and 3D crystallography. The present information about aquaporin structure is largely derived from analysis of AQP1 from human and bovine red blood cells (Walz et al., 1997; Cheng et al., 1997; Li et al., 1997). The puri®ed AQP1 is somewhat heterogeneous due to the presence of a naturally occurring 22 kDa fragment and the existence of a complex N-linked glycan.
Functional Reconstitution of AqpZ
These may contribute to the lack of uniformity in the AQP1 membrane crystals; however, the AqpZ system appears to have neither problem. Moreover, the stability of the AqpZ tetramer in SDS suggests that the protein may be particularly rigid, a feature that may be favorable to preparation of highly uniform 2D and 3D crystals. The lack of rigidity is believed to be in part responsible for the lack of uniform 2D crystals prepared from abundant amounts of the E. coli Lac permease (Zhuang et al., 1999). The AqpZ prepared with our system Ê has already yielded structural information at 7 A resolution (Ringler et al., 1999, accompanying paper); however, it is recognized that eventual preparation of 3D crystals will be needed for X-ray diffraction. For all of these reasons, we believe that AqpZ will be an exceedingly interesting and useful member of the aquaporin family.
Materials and Methods Materials N-octyl-b-D-glucopyranoside (octylglucoside) and n-dodecyl-b-D-maltoside (dodecylmaltoside) were obtained from Calbiochem. Carboxy¯uorescein and anti¯uorescein antibody were from Molecular Probes. Escherichia coli total lipid extract acetone/ether preparation was from Avanti Polar-lipids. Ni-NTA-agarose was from Qiagen. Restriction enzymes were from New England Biolabs. Other reagents were from Sigma or Aldrich.
Expression plasmids and bacterial strains The plasmid pTrcHisUhpT (Tamai et al., 1997) was modi®ed to contain an EcoRI cleavage site downstream from the 10-His tag. Mutagenesis was performed using the Chameleon Double Stranded, Site-Directed Mutagenesis kit (Stratagene, CA); a primer obliterating the ScaI site on the vector was used for selection (Table 2). The product, pTrc10HisEcoRI, contained a stop codon in frame with the 10-His tag after the EcoRI site, leading to the expression of a truncated 25-residue peptide. A cDNA coding for AqpZ was digested with EcoRI and XhoI, and was ligated into pTrc10HisEcoRI previously digested with EcoRI and XbaI. The resulting construct, pTrc10HisAqpZ, contains the sequence of the commercially available pTrc99A plasmid (Pharmacia) without the insert NcoI-XbaI, which was replaced by an insertion containing the coding sequence for 10-His-AqpZ (Figure 1), followed by a fragment of aqpZ 30 UTR (up to the XhoI site). Upstream from this sequence, the vector contains a Trc promoter inducible by isopropylthiogalactoside (IPTG). The vector also encodes lacIq, the Trc repressor that ensures low transcriptional levels in the absence of inducer. The expression vector was transformed into the commercially available E. coli strain XL1B, and selected for by ampicillin resistance. The expression vector was modi®ed using the Chameleon system to yield the different mutant forms used in this study. Sequences of the primers used for mutagenesis are listed in Table 2.
1177
Functional Reconstitution of AqpZ Table 2. Primers utilized in site-directed mutagenesis Purpose T183C R189V R189S C9S C20S C9S EcoRI site generation Selection
Name T183C R189V R189S 2xC-S C9S 10HISECO ScaI-AgeI
Expression and purification of 10-His-AqpZ Luria broth cultures containing 50 mg/ml ampicillin were incubated for 13 to 16 hours at 37 C, diluted 100fold into fresh broth, and propagated to an A600 nm of about 1.5. Induction of 10-His-AqpZ was achieved by addition of 1 mM IPTG for two hours at 37 C before centrifugation (15 minutes at 9000 g). Harvested cells were resuspended in one 1/100 culture volume of ice cold lysis buffer containing: 100 mM K2HPO4, 1 mM MgSO4, 1 mM phenylmethylsulfonyl¯uoride (PMSF), 0.1 mg/ml deoxyribonuclease I (pH 7.0). Cells were subjected to three lysis cycles in a French press (125 106 Pa, at 4 C). Unbroken cells and debris were separated from the cell lysate by a 30 minute centrifugation at 10,000 g and discarded. Membrane fractions were recovered from the supernatant by a 60 minute centrifugation at 140,000 g. For protein extraction cell membranes were resuspended to the original volume in solubilization buffer (1.5 % dodecylmaltoside in a buffer containing 100 mM K2HPO4, 10 % glycerol, 5 mM b-mercaptoethanol, 200 mM NaCl, pH 8.0) and incubated on ice for one hour. Insoluble material was pelleted by 45 minute centrifugation at 140,000 g. The soluble fraction was mixed with 1/60 volume of pre-washed Ni-NTA-agarose beads (Qiagen) and incubated, with agitation, at 4 C for two hours. The resin was then packed in a column and washed with 100 volumes of a wash-buffer (1.5 % dodecylmaltoside, 100 mM K2HPO4, 10 % glycerol, 5 mM b-mercaptoethanol, 200 mM NaCl, 100 mM imidazole, pH 7.0) to remove non-speci®cally bound material. Residual washing buffer was removed by low speed centrifugation of the column (two minutes at 2,000 rpm in a bench-top microcentrifuge). Ni-NTA-agarose-bound material was eluted by incubation in one bed volume of elution buffer (1.5 % dodecylmaltoside, 100 mM K2HPO4, 10 % glycerol, 5 mM b-mercaptoethanol, 200 mM NaCl, 1 M imidazole, pH 7.0) for one hour at room temperature. For maximum recovery, repeated elution steps were performed. Typically, the ®rst elution step yielded pure protein at a concentration of 5-10 mg/ml, measured as described (Schaffner & Weissmann, 1973) using bovine serum albumin as a standard. Sedimentation analysis Solubilized, puri®ed 10-His-AqpZ (2-10 mg of protein in 100 ml sample volume) was layered on top of a 4 ml continuous sucrose gradient (5 % to 20 % sucrose, 1.5 % dodecylmaltoside, 5 mM EDTA, 20 mM Tris-HCl (pH 8)) and centrifuged at 140,000 g for 18 hours (34,000 rpm in a Beckman SW 50.1 Ti rotor). Fractions were collected and analyzed by SDS-PAGE. The sedimentation coef®cient (sw,20) was determined by comparison with the fol-
Sequence 0
5 -PCG GGT TGA CAG AAC AGT TAG TCA CC -30 50 -PCG CAG TGC TGA CCG CTG GGT TAA CAG-30 50 -PCG CAG TGC TGC TCG CTG GGT TAA CAG-30 50 -PGC ACT ACC AGA GCC ACC AAA AAC AAG CCA GAA GGT ACC AAA AGA TTC AGC TGC-30 50 -PCC AGA AGG TAC CAA AAG ATT CAG CTG C-30 50 -GCA TGC GAC CTT AGA ATT CGT GAC C-30 50 -PTT CTG TGA CCG GTG AAT ACT CAA CC-30
lowing internal standards: cytochrome c (1.8), catalase (2.9), b-amylase (4.3), bovine serum albumin (8.9) and carbonic anhydrase (11.2). Functional reconstitution of purified AqpZ Puri®ed 10-His-AqpZ was reconstituted into proteoliposomes. The low critical micellar concentration of dodecylmaltoside makes this detergent unsuitable for reconstitution by dialysis, so the detergent octylglucoside was used instead. Brie¯y, a reconstitution mixture was prepared at room temperature by sequentially adding: 100 mM Mops-Na (pH 7.5); 1.25 % (w/v) octylglucoside; puri®ed 10-His-AqpZ (®nal concentration 50-100 mg/ml); and 10 mg/ml lipids. E. coli total lipid extract (acetone/ ether preparation, Avanti Polar Lipids, Inc.) had previously been resuspended in 2 mM b-mercaptoethanol in water (®nal concentration 50 mg/ml) and incubated at room temperature for one hour, diluted to a ®nal concentration of 45 mg/ml in 100 mM Mops-Na (pH 7.5) and pulsed in a bath sonicator until a clear suspension was obtained. Lipid was always handled under a nitrogen atmosphere. The reconstitution mixture was loaded into SPECTRA/POR 2 dialysis tubing, molecular mass cut-off 12,000-14,000 (Spectrum Medical Industries, Inc.), and dialyzed against 100 volumes of assay buffer (50 mM MOPS, 150 mM N-methyl-D-glucamine, adjusted to pH 7.5 with HCl) for 24-72 hours at room temperature. Alternatively, the reconstitution mixture was rapidly injected into 25 volumes of assay buffer to dilute the detergent. Liposomes were harvested by centrifugation (45 minutes at 140,000 g) and resuspended into assay buffer. The diameter of the 10-His-AqpZ proteoliposomes obtained by dialysis was measured from micrographs of cryofractured material (122(20) nm) and by light scattering (130 nm). Membrane permeability measurements The osmotic behavior of reconstituted proteoliposomes and control liposomes was analyzed by following the light scattering of the preparation in a stopped-¯ow apparatus (SV.17MV, Applied Photophysics) with a measured dead time of 0.7 ms. The water permeability was measured by rapid mixture with hyperosmolar assay buffer containing sucrose, causing water to leave the vesicles. Light scattering was recorded at an emission wavelength of 600 nm. Under these conditions a reduction in vesicle volume leads to an increase in the signal. Data were ®tted to a exponential rise equation and initial rates (k) were calculated from the ®tted equation. The osmotic water permeability (Pf) was calculated using the following expression:
1178
Functional Reconstitution of AqpZ Pf k= S=V0 Vw osm
where S/V0 is the vesicle surface area to initial volume ratio, Vw is the partial molar volume of water (18 cm3), and osm is the difference in osmolarity between the intravesicular and extravesicular aqueous solutions (usually 300 mos mol). The radius of proteoliposomes was measured by electron microscopy. The permeability of single AqpZ molecules (pf) was estimated as the difference in permeability between proteoliposomes and liposomes, multiplied by the surface area of a proteoliposome, and divided by the number of AqpZ molecules incorporated to a proteoliposome. The number of AqpZ monomers was estimated from the amount of protein incorporated into the reconstitution mixture. The number of liposomes was estimated assuming an internal volume of 1 ml/mg of phospholipid (Ambudkar & Maloney, 1986). The permeability of the liposomes to small solutes was performed by rapidly mixing a volume of suspension vesicles at a given concentration of permeant solute with a volume of solution with the osmolarity equilibrated with a non-permeant solute. The effective concentration of the solute in the outer medium is then reduced to half, generating an outward gradient. The outward movement of permeant solute generates an outwardly oriented osmotic gradient, which drives water movement. The consequent change in volume is monitored by stopped-¯ow, as described above.
Acknowledgements This work was supported by a fellowship from the Human Frontier Science Program (M.J.B.), the National Institutes of Health and the Cystic Fibrosis Foundation (P.A.), and the National Science Foundation, no. MCB 9603997 (P.C.M.). We thank Dr Mon-Chou Fann and Dr John C. Mathai for discussion of this work, and assistance. We thank Gera D. Eytan for critical reading of the manuscript.
References Abrami, L., Simon, M., Rousselet, G., Berthonaud, V., Buhler, J. M. & Ripoche, P. (1994). Sequence and functional expression of an amphibian water channel, FA-CHIP: a new member of the MIP family. Biochim. Biophys. Acta, 1192, 147-151. Ambudkar, S. V. & Maloney, P. C. (1986). Bacterial anion exchange. Use of osmolytes during solubilization and reconstitution of phosphate-linked antiport from Streptococcus lactis. J. Biol. Chem. 261, 10079-10086. Bixby, K. A., Nanao, M. H., Shen, N. V., Kreusch, A., Bellamy, H., Pfaf®nger, P. J. & Choe, S. (1999). Zn2-binding and molecular determinants of tetramerization in voltage-gated K channels. Nature Struct. Biol. 6, 38-43. Bonhivers, M., Carbrey, J. M., Gould, S. J. & Agre, P. (1998). Aquaporins in Saccharomyces. Genetic and functional distinctions between laboratory and wild-type strains. J. Biol. Chem. 273, 27565-27572. Calamita, G., Bishai, W. R., Preston, G. M., Guggino, W. B. & Agre, P. (1995). Molecular cloning and characterization of AqpZ, a water channel from Escherichia coli. J. Biol. Chem. 270, 29063-29066.
Calamita, G., Kempf, B., Bonhivers, M., Bishai, W. R., Bremer, E. & Agre, P. (1998). Regulation of the Escherichia coli water channel gene aqpZ. Proc. Natl Acad. Sci. USA, 95, 3627-3631. Cheng, A., van Hoek, A. N., Yeager, M., Verkman, A. S. & Mitra, A. K. (1997). Three-dimensional organization of a human water channel. Nature, 387, 627630. Coury, L. A., Mathai, J. C., Prasad, G. V., Brodsky, J. L., Agre, P. & Zeidel, M. L. (1998). Reconstitution of water channel function of aquaporins 1 and 2 by expression in yeast secretory vesicles. Am. J. Physiol. 274, F34-F42. Denker, B. M., Smith, B. L., Kuhajda, F. P. & Agre, P. (1988). Identi®cation, puri®cation, and partial characterization of a novel Mr 28,000 integral membrane protein from erythrocytes and renal tubules. J. Biol. Chem. 263, 15634-15642. Finkelstein, A. (1987). Water Movement Through Lipid Bilayers, Pores, and Plasma Membranes: Theory and Reality, John Wiley & Sons, New York. Helenius, A. & Simons, K. (1975). Solubilization of membranes by detergents. Biochim. Biophys. Acta, 415, 2979. Huse, M., Eck, M. J. & Harrison, S. C. (1998). A Zn2 ion links the cytoplasmic tail of CD4 and the Nterminal region of Lck. J. Biol. Chem. 273, 1872918733. Jung, J. S., Preston, G. M., Smith, B. L., Guggino, W. B. & Agre, P. (1994). Molecular structure of the water channel through aquaporin CHIP. The hourglass model. J. Biol. Chem. 269, 14648-14654. Lagree, V., Pellerin, I., Hubert, J. F., Tacnet, F., Le, Caherec F., Roudier, N., Thomas, D., Gouranton, J. & Deschamps, S. (1998). A yeast recombinant aquaporin mutant that is not expressed or mistargeted in Xenopus oocyte can be functionally analyzed in reconstituted proteoliposomes. J. Biol. Chem. 273, 12422-12426. Laize, V., Rousselet, G., Verbavatz, J. M., Berthonaud, V., Gobin, R., Roudier, N., Abrami, L., Ripoche, P. & Tacnet, F. (1995). Functional expression of the human CHIP28 water channel in a yeast secretory mutant. FEBS Letters, 373, 269-274. le Coutre, J., Kaback, H. R., Patel, C. K., Heginbotham, L. & Miller, C. (1998). Fourier transform infrared spectroscopy reveals a rigid alpha-helical assembly for the tetrameric Streptomyces lividans K channel. Proc. Natl Acad. Sci. USA, 95, 6114-6117. Li, H., Lee, S. & Jap, B. K. (1997). Molecular design of aquaporin-1 water channel as revealed by electron crystallography. Nature Struct. Biol. 4, 263-265. Ma, T., Yang, B. & Verkman, A. S. (1996). cDNA cloning of a functional water channel from toad urinary bladder epithelium. Am. J. Physiol. 271, C1699C1704. Mathai, J. C. & Agre, P. (1999). Hourglass pore-forming domains restrict aquaporin-1 tetramer assembly. Biochemistry, 38, 923-928. Park, J. H. & Saier, M. H. (1996). Phylogenetic characterization of the MIP family of transmembrane channel proteins. J. Membr. Biol. 153, 171-180. Preston, G. M., Carroll, T. P., Guggino, W. B. & Agre, P. (1992). Appearance of water channels in Xenopus oocytes expressing red cell CHIP28 protein. Science, 256, 385-387. Preston, G. M., Jung, J. S., Guggino, W. B. & Agre, P. (1993). The mercury-sensitive residue at cysteine
Functional Reconstitution of AqpZ 189 in the CHIP28 water channel. J. Biol. Chem. 268, 17-20. Ringler, P., Borgnia, M. J., Stahlberg, H., Maloney, P., Agre, P. & Engel, A. (1999). Structure of water channel Aqpz from Escherichia coli revealed by electron crystallography. J. Mol. Biol. 291, 1181-1190. Schaffner, W. & Weissmann, C. (1973). A rapid, sensitive, and speci®c method for the determination of protein in dilute solution. Anal. Biochem. 56, 502514. Smith, B. L. & Agre, P. (1991). Erythrocyte Mr 28,000 transmembrane protein exists as a multisubunit oligomer similar to channel proteins. J. Biol. Chem. 266, 6407-6415. Tamai, E., Fann, M. C., Tsuchiya, T. & Maloney, P. C. (1997). Puri®cation of UhpT, the sugar phosphate transporter of Escherichia coli. Protein Exp. Purif. 10, 275-282. van Hoek, A. N. & Verkman, A. S. (1992). Functional reconstitution of the isolated erythrocyte water channel CHIP28. J. Biol. Chem. 267, 18267-18269. Verbavatz, J. M., Brown, D., Sabolic, I., Valenti, G., Ausiello, D. A., van Hoek, A. N., Ma, T. & Verkman, A. S. (1993). Tetrameric assembly of CHIP28 water channels in liposomes and cell membranes: a freeze-fracture study. J. Cell Biol. 123, 605618. Walz, T., Smith, B. L., Agre, P. & Engel, A. (1994). The three-dimensional structure of human erythrocyte aquaporin CHIP. EMBO J. 13, 2985-2993.
1179 Walz, T., Hirai, T., Murata, K., Heymann, J. B., Mitsuoka, K., Fujiyoshi, Y., Smith, B. L., Agre, P. & Engel, A. (1997). The three-dimensional structure of aquaporin-1. Nature, 387, 624-627. Weig, A., Deswarte, C. & Chrispeels, M. J. (1997). The major intrinsic protein family of Arabidopsis has 23 members that form three distinct groups with functional aquaporins in each group. Pl. Physiol. 114, 1347-1357. Yang, B. & Verkman, A. S. (1997). Water and glycerol permeabilities of aquaporins 1-5 and MIP determined quantitatively by expression of epitopetagged constructs in Xenopus oocytes. J. Biol. Chem. 272, 16140-16146. Zeidel, M. L., Ambudkar, S. V., Smith, B. L. & Agre, P. (1992). Reconstitution of functional water channels in liposomes containing puri®ed red cell CHIP28 protein. Biochemistry, 31, 7436-7440. Zeidel, M. L., Nielsen, S., Smith, B. L., Ambudkar, S. V., Maunsbach, A. B. & Agre, P. (1994). Ultrastructure, pharmacologic inhibition, and transport selectivity of aquaporin channel-forming integral protein in proteoliposomes. Biochemistry, 33, 1606-1615. Zhuang, J., Prive, G. G., Werner, G. E., Ringler, P., Kaback, H. R. & Engel, A. (1999). Two-dimensional crystallization of Escherichia coli lactose permease. J. Struct. Biol. 125, 63-75.
Edited by W. Baumeister (Received 9 April 1999; received in revised form 30 June 1999; accepted 16 July 1999)
J. Mol. Biol. (1999) 291, 1169±1179
Functional Reconstitution and Characterization of AqpZ, the E. coli Water Channel Protein Mario J. Borgnia1, David Kozono1, Giuseppe Calamita4, Peter C. Maloney3 and Peter Agre1,2* 1
Department of Biological Chemistry 2
Department of Medicine
3
Department of Physiology Johns Hopkins University School of Medicine, 725 N Wolfe Street, Baltimore, MD 21205-2185, USA 4
Dipartimento di Fisiologia Generale ed Ambientale University of Bari, 165/A via Amendola, 70126, Bari, Italy
Understanding the selectivity of aquaporin water channels will require structural and functional studies of wild-type and modi®ed proteins; however, expression systems have not previously yielded aquaporins in the necessary milligram quantities. Here we report expression of a histidine-tagged form of Escherichia coli aquaporin-Z (AqpZ) in its homologous expression system. 10-His-AqpZ is solubilized and puri®ed to near homogeneity in a single step with a ®nal yield of 2.5 mg/l of culture. The histidine tag is removed by trypsin, yielding the native protein with the addition of three N-terminal residues, as con®rmed by microsequencing. Sucrose gradient sedimentation analysis showed that the native, solubilized AqpZ protein is a trypsin-resistant tetramer. Unlike other known aquaporins, AqpZ tetramers are not readily dissociated by 1 % SDS at neutral pH. Hydrophilic reducing agents have a limited effect on the stability of the tetramer in 1 % SDS, whereas incubations for more than 24 hours, pH values below 5.6, or exposure to the hydrophobic reducing agent ethanedithiol cause dissociation into monomers. Cys20, but not Cys9, is necessary for the stability of the AqpZ tetramer in SDS. Upon reconstitution into proteoliposomes, AqpZ displays very high osmotic water permeability (pf 5 10 10ÿ14 cm3 sÿ1 subunitÿ1) and low Arrhenius activation energy (Ea 3.7 kcal/mol), similar to mammalian aquaporin-1 (AQP1). No permeation by glycerol, urea or sorbitol was detected. Expression of native and modi®ed AqpZ in milligram quantities has permitted biophysical characterization of this remarkably stable aquaporin tetramer, which is being utilized for high-resolution structural studies. # 1999 Academic Press
*Corresponding author
Keywords: aquaporin; bacterial; water transport; channel structure; gene family
Introduction Water movement across biological membranes is not fully explained by simple diffusion through the lipid bilayer. Observations of highly water-permeable tissues led to the development of theoretiAbbreviations used: AQP1, aquaporin-1; AqpZ, aquaporin-Z; dodecylmaltoside, n-dodecyl-b-Dmaltoside; GlpF, glycerol facilitator protein; IPTG, isopropyl-b-D-thiogalactoside; Mops, 3-[nmorpholino]propanesulfonic acid; Ni-NTA, nickelnitrilotriacetic acid; NMDG, n-methyl-D-glucamine; octylglucoside, n-octyl-b-D-glucopyranoside; PMSF, phenylmethylsulfonyl ¯uoride; PVDF, polyvinylidene di¯uoride. E-mail address of the corresponding author: [email protected] 0022-2836/99/351169±11 $30.00/0
cal tools for the study of osmotic and diffusional processes across membranes and pores, and prompted the long pursuit of the molecular entity responsible for water transport (Finkelstein, 1987). The ®rst characterized water channel, aquaporin-1 (AQP1), was serendipitously identi®ed and puri®ed from human red blood cells (Denker et al., 1988; Smith & Agre, 1991). Expression of AQP1 mRNA in Xenopus oocytes (Preston et al., 1992) led to its recognition as the erythrocyte water channel, a conclusion con®rmed by reconstitution of highly puri®ed AQP1 into proteoliposomes (Zeidel et al., 1992, 1994; van Hoek & Verkman, 1992). Sequence analysis related AQP1 to the Escherichia coli glycerol facilitator protein (GlpF) and to the major intrinsic protein of mammalian lens (MIP, now designated AQP0). Screening of mRNA # 1999 Academic Press
1170 libraries prepared from different tissues permitted identi®cation of at least ten mammalian homologs. Multiple homologs have also been identi®ed in other vertebrates (Abrami et al., 1994; Ma et al., 1996), plants (Weig et al., 1997), yeast (Bonhivers et al., 1998) and bacteria (Calamita et al., 1995). Some of the identi®ed proteins are highly speci®c for water transport (``orthodox aquaporins'') whereas others are permeated by water, glycerol and other small, uncharged molecules (``aquaglyceroporins''). The amino acid sequences of aquaporins are approximately 30 % identical; however, the conserved residues are distributed in de®ned clusters, which themselves are absolutely conserved throughout the family (Park & Saier, 1996). All homologs are predicted to contain six transmembrane-spanning segments formed by the tandem duplication of three transmembrane domains joined in obverse symmetry. The N and C termini reside in the intracellular space (Smith & Agre, 1991), leaving three extracellular and two intracellular connecting loops (Figure 1). Loops B and E share the most highly conserved residues, including the Asn-Pro-Ala motifs, suggesting that they
Functional Reconstitution of AqpZ
may be directly involved in the selectivity ®lter of the channel. Because several residues in these loops are hydrophobic, it was proposed that loops B and E fold into the membrane-spanning region of each subunit and are surrounded by the six transmembrane segments forming a structure referred to as the ``hourglass'' (Jung et al., 1994). Although each subunit apparently contains a single aqueous pore, AQP1 and other aquaporins are believed to exist as tetramers in mild detergents (Smith & Agre, 1991) and in the native membrane (Jung et al., 1994; Verbavatz et al., 1993; Walz et al., 1994). Although certain well-de®ned residues distinguish aquaporins from aquaglyceroporins, the molecular determinants of the functional differences between the two groups are far from clear. Their elucidation awaits the structural and biophysical characterization of multiple homologs, chimeras and mutants. To date, most functional and comparative studies have been carried out with the oocyte expression system. Single channel water permeabilities are hard to determine in this system due to the lack of an accurate method for quanti®cation of plasma membrane protein
Figure 1. Predicted primary sequence and membrane topology of 10-His-AqpZ. To facilitate the puri®cation of the overexpressed protein, the sequence of AqpZ was modi®ed to include a 23-residue N-terminal tag containing ten consecutive histidine residues. The residue corresponding to the con®rmed trypsin cleavage site is indicated (arrowhead). Numbers indicate the residues that were subjected to mutagenesis. Highly conserved residues in the aquaporin family are marked as ®lled shapes. Basic residues are represented as diamonds, acidic residues appear as triangles and neutral residues are shown as circles. The putative transmembrane domains were determined by submitting an alignment including the predicted sequence of AqpZ to the Predict Protein e-mail server ([email protected]).
1171
Functional Reconstitution of AqpZ
expression, the possible contribution of endogenous proteins, and differences in membrane traf®cking. Moreover, the protein expression levels achieved in oocytes are several orders of magnitude below the levels needed for structural studies. Thus, red cell AQP1 is the only member of the aquaporin water channel family with a structural Ê (Walz et al., 1997; Li et al., model resolved at 6 A 1997; Cheng et al., 1997). Because AQP1 is puri®ed from human red blood cells, structure-function analysis is currently limited to the wild-type molecule. Thus, development of expression systems for the puri®cation of large quantities of wild-type and modi®ed aquaporins is essential for the understanding of the molecular basis of the biophysical properties of aquaporin. AqpZ, the aquaporin from E. coli (Calamita et al., 1995), was identi®ed as a good candidate for puri®cation following overexpression in bacteria (Calamita et al., 1998) and may be a good substrate for structure-function studies. Here, we report the functional reconstitution and characterization of highly puri®ed native and modi®ed AqpZ forms.
Results
Figure 2. Electrophoretic analysis of puri®ed 10-HisAqpZ. Aliquots of puri®ed material were incubated for one hour at room temperature in denaturing loading buffer containing 1 % SDS and 143 mM b-mercaptoethanol. (a) Silver staining of a sample resolved in 15 % SDS-PAGE. (b) Dependency of the apparent molecular mass on acrylamide concentration. Samples were resolved in gels of different polymer content and Coomassie blue stained. The apparent molecular mass was calculated by comparison with standards.
Expression and purification of AqpZ Puri®cation of membrane transport proteins has been greatly facilitated by construction of recombinants containing peptide domains such as poly(His), which are used for af®nity-chromatography. The AqpZ protein is believed to contain only short N-terminal and C-terminal cytoplasmic domains, so the aqpZ DNA was cloned into the pTrc10HisEcoRI plasmid (Tamai et al., 1997) yielding an AqpZ polypeptide with a 23-residue N-terminal extension containing ten consecutive His residues (Figure 1). E. coli cultures transformed with pTrc10HisAqpZ grew at the same rate as control bacteria transformed with the plasmid pTrc10HisEcoRI, encoding a truncated 25 residue polypeptide. In contrast, addition of 1 mM IPTG caused an almost complete arrest of growth of pTrc10HisAqpZ-transformed cells but was without appreciable effect on control bacteria, suggesting that the expression of 10-His-AqpZ is toxic to the cells. To generate reasonable levels of expression, cultures were propagated to higher cell densities (A 1.4 to 1.8) prior to induction. 10-His-AqpZ was puri®ed from the dodecylmaltoside-solubilized membrane fraction by Ni af®nity chromatography. Aliquots of the eluted material were diluted in sample buffer (12 % glycerol, 1 % SDS, 140 mM b-mercaptoethanol and 70 mM TrisHCl (pH 6.8)) for one hour at room temperature and were analyzed by SDS-PAGE with 15 % polyacrylamide slabs. A major molecular species of 80 kDa and a less-abundant species of 23 kDa were revealed by silver staining of the gels (Figure 2(a)). One liter of culture yielded up to 2.5 mg of puri®ed protein, depending on the density of the culture when induced.
Known aquaporins migrate as 25-30 kDa monomers during SDS-PAGE. To evaluate the molecular mass of the 80 kDa species, the electrophoretic behavior in SDS-PAGE was studied at different concentrations of acrylamide. The analysis revealed a linear relationship between the apparent molecular mass and the acrylamide concentration, with the higher molecular mass band exhibiting faster mobility in gels of lower polymer content (Figure 2(b)). This aberrant electrophoretic behavior is a known characteristic of certain hydrophobic proteins (Helenius & Simons, 1975) but precludes identi®cation of the apparent SDS-stable oligomeric state of 10-His-AqpZ. Macromolecular assembly of AqpZ To further characterize the eluted material, dodecylmaltoside-solubilized samples were analyzed by non-equilibrium centrifugation on a sucrose gradient, followed by SDS-PAGE analysis of fractions. Interestingly, both the 80 kDa and the 23 kDa species coincided in exactly the same fractions of the gradient. The apparent sedimentation coef®cient for the peak fraction was estimated at 5.7 by comparison of AqpZ mobility and the mobility of several standard proteins (Figure 3). This value is similar to that obtained for octylglucoside-solubilized AQP1 (Smith & Agre, 1991), which is known to be a tetramer (Verbavatz et al., 1993; Walz et al., 1994). Thus, our data suggest that AqpZ exists as tetramer when solubilized in mild detergents like dodecylmaltoside as well as in the
1172
Figure 3. Non-equilibrium sedimentation coef®cient of puri®ed AqpZ. Solubilized, puri®ed AqpZ was layered on top of a 5 % to 20 % continuous sucrose gradient containing 1.5 % dodecylmaltoside and centrifuged at 140,000 g for 18 hours. A total of 35 fractions were collected and resolved by SDS-PAGE. The distribution of the 80 kDa species was coincident with that of the 23 kDa species. The sedimentation coef®cient (sw,20) of AqpZ was determined by comparison with the following standards: cytochrome c (1.8), catalase (2.9), b-amylase (4.3), bovine serum albumin (8.9) and carbonic anhydrase (11.2).
harsh detergent SDS, which usually unfolds proteins and dissociates the subunits. Dissociation of the tetramer was attempted under several different conditions. Incubation of the samples in SDS-PAGE at 37 C or higher temperatures in the presence of chaotropic (8 M urea or guanidinium chloride) or hydrophilic reducing agents (140 mM b-mercaptoethanol or 100 mM dithiothreitol) caused aggregation of AqpZ, which failed to enter the separating SDS-PAGE gels (not shown). The effects of various reducing agents were studied (Figure 4). Increasing concentrations of the hydrophilic reducing agent b-mercaptoethanol caused a shift in the oligomerization of the protein towards the low molecular mass species but ultimately caused aggregation of the protein (Figure 4). No change in the distribution was observed with dithiothreitol in a range of concentrations up to 1 M (Figure 4). Incubation with the relatively water-insoluble reducing agent ethanedithiol at a nominal concentration of 60 mM led to almost complete dissociation of the high molecular mass species (Figure 4), con®rming its identity with the 23 kDa 10-His-AqpZ. Differential reactivity with these reducing agents suggests that the target residue may be located in a hydrophobic environment. Prolonged incubation (>24 hours) in 1 % SDS-containing sample buffer resulted in dissociation of the tetramer even without reducing agents (Figure 4). When sample-loading buffers covering a range of pH were used to incubate the
Functional Reconstitution of AqpZ
Figure 4. Dissociation of the SDS-resistant 10-HisAqpZ tetramer. Top: Effect of reducing agents. Puri®ed protein was incubated for one hour at room temperature in 100 ml aliquots of gel loading buffer (pH 6.8), to which appropriate volumes of dithiothreitol, b-mercaptoethanol or ethanedithiol were added to reach the indicated concentrations. Samples were then resolved in 14 % SDS-PAGE. Bottom left: Long-term dissociation. Puri®ed 10-His-AqpZ was incubated in gel loading buffer (pH 6.8) without reducing agents for the indicated times prior to electrophoresis in 14 % SDS-PAGE. Bottom right: Dissociation by pH. Aliquots of 10-His-AqpZ were incubated in gel loading buffer at the indicated pH for ten minutes, followed by electrophoresis in 14 % SDS-PAGE.
protein at room temperature prior to SDS-PAGE (Figure 4), a rapid transition from tetramer to monomer occurs below pH 5.6 (12 % glycerol, 1 % SDS and 100 mM sodium acetate). The dissociation induced by pH and SDS was not reversed by returning to neutral pH (data not shown). Moreover, the dissociation by pH was not due to the presence of the histidine tag, since experiments carried out on proteolytically cleaved 10-His-AqpZ exhibited the same behavior (see below). Digestion of octylglucoside-solubilized 10-HisAqpZ with trypsin was undertaken to remove the poly(His) tag. Controlled proteolysis with a tenfold molar excess of trypsin revealed a cleavage pattern comprising three intermediate species and a ®nal product (Figure 5(a)). Analysis of the 10-His-AqpZ open reading frame predicts the existence of 11 basic residues, including a potential trypsin cleavage site between the poly(His) tag and the initial Met of AqpZ (Figure 1). Stepwise removal of the N-terminal extension from each subunit of a 10-His-AqpZ tetramer may explain the stepwise degradation observed in the gels. The analysis of digested protein with an anti-His antibody revealed the presence of four distinct species (Figure 5(b)), presumed to correspond to tetramers
Functional Reconstitution of AqpZ
Figure 5. Tryptic removal of the 10-His tag. Puri®ed 10-His-AqpZ (1 mg) was incubated for two hours with trypsin. Lanes a-f: 0, 0.12, 0.37, 1.1, 3.3 and 10 mg of trypsin. The digested material was resolved by 10 % SDS-PAGE. Two different gels run in parallel were stained with (a) Coomassie Brilliant Blue and or (b) transferred to nitrocellulose and probed using a monoclonal antibody recognizing poly(His) (Qiagen).
containing four, three, two or one N-terminal 10-His tags. The ®fth, lowermost band visible with Coomassie Blue staining was not detected by antiHis immunoblot, indicating that it represents the tetramer composed by four monomers lacking the 10-His tag at the N terminus. N-terminal protein microsequencing of the lowermost band transferred to PVDF membranes revealed the sequence H-E-F-M-F-R, corresponding to residues ÿ3 to 3 of the native polypeptide. Thus, the band results from enzymic hydrolysis at a single Arg residue. In contrast, no cleavage was observed when the reconstituted protein was exposed to trypsin (see below), suggesting that the unique cleavage site is protected by the lipid bilayer.
1173 atus. Liposomes and proteoliposomes were rapidly subjected to an outwardly oriented osmotic gradient, and the time course of change in light scattering (lem 600 nm) was recorded. The changes in scattered light were proportional to the gradient imposed for both liposomes and proteoliposomes, indicating the regular osmotic behavior of the system. Owing to changes in ¯uidity of the lipid bilayer, basal water permeability of liposomes increases greatly with temperature. Therefore, to improve the signal to noise ratio, most measurements were performed at 6-8 C. When an osmotic gradient of 284 mosmol/l was imposed, the water permeability (Pf) for 10-His-AqpZ in membranes was proportional to the protein-to-lipid ratio up to 1:100 (Figure 6). Proteoliposomes formed with higher ratios of 10-His-AqpZ to lipid did not exhibit proportionally increased rates of equilibration. At a ratio of 1:200 (50 mg of protein in 10 mg of lipid) the observed Pf was markedly increased (Pf 0.0283 cm/second at 6.5 C) as compared to liposomes (Pf 0.0015 cm/second at 6.5 C).
Reconstitution of AqpZ into proteoliposomes Direct biophysical evaluation of transport functions may be performed after reconstitution of the puri®ed proteins into proteoliposomes. Two different methods, rapid dilution and dialysis, both permitted reconstitution of the puri®ed 10-His-AqpZ into proteoliposomes (see Materials and Methods); however, dialysis yielded higher and more reproducible recovery of the reconstituted material. The order in which the different components were added to the reconstitution mixture was found to be critical. To prevent aggregation and precipitation, the puri®ed AqpZ is placed in a solution of high detergent concentration before addition of the lipids. The diameter of the reconstituted proteoliposomes (liposomes) was 122(20) nm as measured by electron microscopy. Permeability of AqpZ proteoliposomes Water movement across the membrane of control liposomes and proteoliposomes reconstituted with 10-His-AqpZ at different protein-to-lipid ratios was analyzed using a stopped-¯ow appar-
Figure 6. Functional reconstitution of AqpZ. Proteoliposomes reconstituted at different protein concentrations (25, 50, 100, 200 or 400 mg protein in 10 mg lipids) or control liposomes were abruptly exposed to a twofold increase in osmolarity. Light scattering was monitored in a stopped-¯ow apparatus. The reduction in volume due to water ef¯ux caused an increase in light scattering. The collected data were normalized to ®t between zero and unity. The traces are the average of ten or more individual plots at each protein to lipid ratio, and are displaced on the y-axis for clarity. Inset: Pf values for each 10-His-AqpZ to lipid ratio. Rate constants were determined by curve ®tting of the plots to a single order exponential. Pf values were calculated as described in Materials and Methods.
1174 Assuming total incorporation of protein into proteoliposomes, a maximum of 125 monomers per vesicle of 10-His-AqpZ is estimated. Thus, the calculated osmotic water permeability (pf) for each AqpZ subunit at this protein to lipid ratio is 510 10ÿ14 cm3/second, close to the value obtained for AQP1 (11.7 10ÿ14 cm3/second) (Zeidel et al., 1992). Proteoliposomes reconstituted with 10-His-AqpZ after cleavage of the His tag by trypsin treatment were indistinguishable from those containing the intact tagged protein (not shown). Measurements of Pf at pH 5, 6 and 7 were indistinguishable. Water permeability measurements were performed at different temperatures on both proteoliposomes (1:100 protein-to-lipid ratio) and control liposomes (Figure 7). The Arrhenius activation energies calculated from the plot are consistent with water movement through a channel for AqpZ (Ea 3.7 kcal/mol), as opposed to diffusion across the lipid bilayer of liposomes (Ea 16.9 kcal/mol). The permeability of 10-His-AqpZ-containing proteoliposomes to small solutes was also measured (see Materials and Methods), but no difference was observed between proteoliposomes and liposomes for transmembrane movement rates for glucose, sorbitol, sucrose or mannitol (not shown) as well as for glycerol and urea (Table 1). The permeability of the solutes is determined by following the movement of water (see Materials and Methods). Unlike in the case proteoliposomes, where the permeability of water is much higher than that of formamide, the permeability of the latter will be underestimated in the liposomes where
Figure 7. Activation energy of water ¯ow across proteoliposomes. The kinetics of water transport were measured in both proteoliposomes containing reconstituted 10-His-AqpZ (circles) and control liposomes (®lled circles) at different temperatures in the range from 284.1 to 303.5 K. Rates were determined by curve ®tting from the average of ten measurements of water movement at each temperature. The natural logarithms of the rates are plotted against the reciprocal of the temperature. The Arrhenius activation energy (EA) was determined from the slope of linear regression of the plots.
Functional Reconstitution of AqpZ Table 1. Permeability of AqpZ to uncharged small solutes Rate constant
Glycerol Urea Formamide
Liposomes (X S.D.)
Proteoliposomes (X S.D.)
0.166 0.017 0.0962 0.0362 8.95 0.43
0.116 0.0142 0.0634 0.0255 11.45 0.32
S.D., standard deviation.
water movement is the limiting factor. Therefore, the small (1.33-fold) difference observed for formamide (Table 2) may be attributed to this phenomenon and is minor as compared to the 50-fold increase in permeability observed for water at similar protein:lipid ratio.
Expression of 10-His-AqpZ mutants Proteoliposomes reconstituted with puri®ed wild-type red cell AQP1 have been studied extensively (Zeidel et al., 1992, 1994; van Hoek & Verkman, 1992). Characterization of other wildtype or mutant aquaporins has been restricted to studies of proteins expressed in Xenopus laevis oocytes, a system with known limitations. Thus, a major goal of this study was to establish a system suitable for expression and puri®cation AqpZ variants. In the cysteine-less mutant, the two Cys residues in positions 9 and 20 of native AqpZ were replaced by Ser residues. Sedimentation analysis showed that the cysteine-less variant retains its tetrameric structure when solubilized in the mild detergent dodecylmaltoside at 1.5 % (Figure 3). When analyzed by SDS-PAGE, the monomeric form is predominant (Figure 8). Similar results were obtained for the Cys20-Ser variant, while the Cys9-Ser mutant largely remained an 80 kDa complex in SDS (Figure 8). In spite of this difference in SDS stability, the water permeability of AqpZ was not altered by modi®cation of both Cys residues (Figure 8). The recognized mercurial inhibition of AQP1 is due to the presence of the residue Cys189 preceding the second Asn-Pro-Ala motif (Preston et al., 1993). When inhibition of AqpZ proteoliposomes was attempted, no inhibition was observed in 1 mM HgCl2 (not shown). When the corresponding residue in AqpZ, Thr183, is replaced by a Cys, the stability in SDS is like that of wild-type AqpZ (Figure 8). The water permeability of the Thr183Cys mutant is equivalent to the wild-type AqpZ (Figure 8) and is not inhibited by 1 mM HgCl2 (not shown). In contrast, mutation of the highly conserved Arg in position 189 into Val or Ser resulted in inactivation of the channel (Figure 8). Despite differences in SDS stability or water permeability, sedimentation analysis in the mild detergent dodecylmaltoside at 1.5 % revealed that all mutants
Functional Reconstitution of AqpZ
Figure 8. Analysis of 10-His-AqpZ mutants. Top: Samples of puri®ed modi®ed 10-His-AqpZ were resolved in a 14 % SDS-PAGE. Bottom: Rate constants for water movement in proteoliposomes reconstituted with puri®ed wild-type or mutant 10-His-AqpZ variants.
exhibited mobility similar to that of wild-type AqpZ protein (not shown).
Discussion The key to understanding how aquaporins are selectively permeated by water resides in the threedimensional structure of the channel. So far, structural studies of aquaporins are limited by the lack of means for producing milligram quantities of new homologs or modi®ed variants of the protein. Attempts to express different wild-type and mutant aquaporins in heterologous systems have previously been undertaken. The Xenopus oocyte expression system is often limited by improper traf®cking of mutant aquaporins to the plasma membrane or insuf®cient levels of expression (Jung et al., 1994). Small amounts of AQP1 (Laize et al., 1995), AQP2 (Coury et al., 1998) and AQPcic (Lagree et al., 1998) have been expressed in yeast. Native and His-tagged AQP4 were puri®ed from recombinant-baculovirus-infected Sf9 insect cells (Yang & Verkman, 1997); however, the highest level of expression (0.11 mg/l of culture for AQP4) is insuf®cient for high-resolution structural studies, and the His-tagged form was reported to be inactive. Here we describe an expression system per-
1175 mitting puri®cation of wild-type and site-directed modi®ed AqpZ. The yield obtained is 20 times higher than the one obtained for AQP4 in the baculovirus system. We show that the puri®ed AqpZ is active, with unit permeability values that are similar or higher than those reported for AQP1. The AqpZ expression-reconstitution system was used to characterize the protein biochemically and biophysically. Several observations indicate that AqpZ is very hydrophobic and appears to be tightly packed. Extended tryptic digestion failed to achieve cleavage at any of the amino acids within the native structure of solubilized, His-tagged AqpZ, and no cleavage was observed in the reconstituted protein. Digestion of 10-His-AqpZ with protease V8 from Staphylococcus aureus followed by mass spectrometry analysis by MALDI-TOF con®rmed the resistance of the native core of the protein upon solubilization with mild detergents (data not shown). These results suggest that most of the structure of AqpZ is effectively shielded from protease activity by either the lipid bilayer or the solubilizing detergent. The resistance of the tetramer to dissociation and its abnormal migration pattern observed during SDS-PAGE implies that folding is at least partially preserved even in the presence of SDS. As stated above, sedimentation analysis con®rmed that octylglucoside and dodecylmaltosidesolubilized AqpZ is tetrameric, and this is corroborated by electron-microscopy (Ringler et al., 1999, accompanying paper). Thus, the SDS-resistant species detected by electrophoresis is most likely an undissociated tetramer rather than a result of de novo aggregation. The tight packing of the protein may explain the unusual stability of the reconstituted samples, which retained 100 % activity after six months of storage at 4 C (data not shown). This property makes AqpZ a particularly well-suited protein for structural studies. The two cysteine residues in AqpZ, Cys9 and Cys20 are predicted to be located at the two opposite ends of the ®rst transmembrane segment. Mutant forms of AqpZ including Cys20-Ser, Cys9Ser, and the double mutant, all exist as tetramers when analyzed by sedimentation in mild detergent (octylglucoside or dodecylmaltoside) and exhibit wild-type permeability when reconstituted into proteoliposomes. The resistance to dissociation by SDS-PAGE distinguishes between the different cysteine mutants even though both mutants remain functional. Cys20 is necessary and suf®cient to stabilize the tetramer, since it modi®cation causes the protein to migrate primarily as a monomer during SDS-PAGE, whereas modi®cation of Cys9 has no apparent effect (Figure 8). Dissociation caused by the reducing agent ethanedithiol of wild-type and Cys9-Ser tetramers argues that the oxidation state of Cys20 may be critical to the stability of the tetramer in SDS. Unlike wild-type AqpZ, octylglucoside or dodecylmaltoside-solubilized Cys20-Ser precipitates when stored for long periods at 4 C (not shown). The existence of disul®de bonds between subunits is not consistent with
1176 the observed rapid dissociation at low pH and the long-term dissociation at neutral pH in the presence of 1 % SDS. Dissociation of the tetramer in SDS may be explained independently for each condition as minor conformational changes induced by pH, steric interactions due to serine substitution of Cys20, the oxidation state of Cys20, or other unidenti®ed residue(s). Non-covalent tetrameric assemblies of other bacterial transport proteins have also been found to be stable in SDS (le Coutre et al., 1998). The coordination of divalent cations by four cysteine residues as a fold-stabilizing feature has been described recently (Huse et al., 1998; Bixby et al., 1999). This type of structure is sensitive to all three parameters affecting AqpZ tetramer stability in SDS, including pH, electrochemical reduction of the coordinated metal, and mutagenesis of Cys20. Thus, the idea that Cys20 in each subunit stabilizes the tetramer is appealing; however, attempts to reduce the tetramer by incubation in chelators have so far been unsuccessful (not shown). Existence of a cysteine-mediated intersubunit interaction will require involvement of the ®rst transmembrane segment of each monomer in protein-protein interaction close to the center of the tetramer rather than facing the lipid phase. The availability of large quantities of puri®ed wild-type and mutant forms of AqpZ may now permit highresolution structural studies that could clarify this. The role of the highly conserved Arg residue that follows the second NPA in most aquaporins was explored. The Arg189 in loop E was modi®ed into Val, the residue located in the homologous position relative to the NPA in loop B and in loop E in a subset of plant aquaporins. This mutation was initially undertaken with the aim of overcoming the negative selectivity of AqpZ to protons. This and other substitutions of Arg189 led to loss of activity, without change in the tetrameric state of the solubilized mutants, as judged by the sedimentation in mild detergents (not shown), whereas the stability of the protein in SDS was reduced (Figure 8). Thus, although Arg189 is presumed to reside in the hourglass close to the center of each monomer, the loss of activity is accompanied by perturbation of the structure at the interface between monomers. This result is consistent with the prediction that highly conserved residues in the hourglass pore-forming domain of the protein have a major function in holding the tetramer together (Mathai & Agre, 1999). The largest bene®t from the AqpZ expressionreconstitution system may be the availability of large amounts of wild-type or site-directed mutant forms of an aquaporin for 2D and 3D crystallography. The present information about aquaporin structure is largely derived from analysis of AQP1 from human and bovine red blood cells (Walz et al., 1997; Cheng et al., 1997; Li et al., 1997). The puri®ed AQP1 is somewhat heterogeneous due to the presence of a naturally occurring 22 kDa fragment and the existence of a complex N-linked glycan.
Functional Reconstitution of AqpZ
These may contribute to the lack of uniformity in the AQP1 membrane crystals; however, the AqpZ system appears to have neither problem. Moreover, the stability of the AqpZ tetramer in SDS suggests that the protein may be particularly rigid, a feature that may be favorable to preparation of highly uniform 2D and 3D crystals. The lack of rigidity is believed to be in part responsible for the lack of uniform 2D crystals prepared from abundant amounts of the E. coli Lac permease (Zhuang et al., 1999). The AqpZ prepared with our system Ê has already yielded structural information at 7 A resolution (Ringler et al., 1999, accompanying paper); however, it is recognized that eventual preparation of 3D crystals will be needed for X-ray diffraction. For all of these reasons, we believe that AqpZ will be an exceedingly interesting and useful member of the aquaporin family.
Materials and Methods Materials N-octyl-b-D-glucopyranoside (octylglucoside) and n-dodecyl-b-D-maltoside (dodecylmaltoside) were obtained from Calbiochem. Carboxy¯uorescein and anti¯uorescein antibody were from Molecular Probes. Escherichia coli total lipid extract acetone/ether preparation was from Avanti Polar-lipids. Ni-NTA-agarose was from Qiagen. Restriction enzymes were from New England Biolabs. Other reagents were from Sigma or Aldrich.
Expression plasmids and bacterial strains The plasmid pTrcHisUhpT (Tamai et al., 1997) was modi®ed to contain an EcoRI cleavage site downstream from the 10-His tag. Mutagenesis was performed using the Chameleon Double Stranded, Site-Directed Mutagenesis kit (Stratagene, CA); a primer obliterating the ScaI site on the vector was used for selection (Table 2). The product, pTrc10HisEcoRI, contained a stop codon in frame with the 10-His tag after the EcoRI site, leading to the expression of a truncated 25-residue peptide. A cDNA coding for AqpZ was digested with EcoRI and XhoI, and was ligated into pTrc10HisEcoRI previously digested with EcoRI and XbaI. The resulting construct, pTrc10HisAqpZ, contains the sequence of the commercially available pTrc99A plasmid (Pharmacia) without the insert NcoI-XbaI, which was replaced by an insertion containing the coding sequence for 10-His-AqpZ (Figure 1), followed by a fragment of aqpZ 30 UTR (up to the XhoI site). Upstream from this sequence, the vector contains a Trc promoter inducible by isopropylthiogalactoside (IPTG). The vector also encodes lacIq, the Trc repressor that ensures low transcriptional levels in the absence of inducer. The expression vector was transformed into the commercially available E. coli strain XL1B, and selected for by ampicillin resistance. The expression vector was modi®ed using the Chameleon system to yield the different mutant forms used in this study. Sequences of the primers used for mutagenesis are listed in Table 2.
1177
Functional Reconstitution of AqpZ Table 2. Primers utilized in site-directed mutagenesis Purpose T183C R189V R189S C9S C20S C9S EcoRI site generation Selection
Name T183C R189V R189S 2xC-S C9S 10HISECO ScaI-AgeI
Expression and purification of 10-His-AqpZ Luria broth cultures containing 50 mg/ml ampicillin were incubated for 13 to 16 hours at 37 C, diluted 100fold into fresh broth, and propagated to an A600 nm of about 1.5. Induction of 10-His-AqpZ was achieved by addition of 1 mM IPTG for two hours at 37 C before centrifugation (15 minutes at 9000 g). Harvested cells were resuspended in one 1/100 culture volume of ice cold lysis buffer containing: 100 mM K2HPO4, 1 mM MgSO4, 1 mM phenylmethylsulfonyl¯uoride (PMSF), 0.1 mg/ml deoxyribonuclease I (pH 7.0). Cells were subjected to three lysis cycles in a French press (125 106 Pa, at 4 C). Unbroken cells and debris were separated from the cell lysate by a 30 minute centrifugation at 10,000 g and discarded. Membrane fractions were recovered from the supernatant by a 60 minute centrifugation at 140,000 g. For protein extraction cell membranes were resuspended to the original volume in solubilization buffer (1.5 % dodecylmaltoside in a buffer containing 100 mM K2HPO4, 10 % glycerol, 5 mM b-mercaptoethanol, 200 mM NaCl, pH 8.0) and incubated on ice for one hour. Insoluble material was pelleted by 45 minute centrifugation at 140,000 g. The soluble fraction was mixed with 1/60 volume of pre-washed Ni-NTA-agarose beads (Qiagen) and incubated, with agitation, at 4 C for two hours. The resin was then packed in a column and washed with 100 volumes of a wash-buffer (1.5 % dodecylmaltoside, 100 mM K2HPO4, 10 % glycerol, 5 mM b-mercaptoethanol, 200 mM NaCl, 100 mM imidazole, pH 7.0) to remove non-speci®cally bound material. Residual washing buffer was removed by low speed centrifugation of the column (two minutes at 2,000 rpm in a bench-top microcentrifuge). Ni-NTA-agarose-bound material was eluted by incubation in one bed volume of elution buffer (1.5 % dodecylmaltoside, 100 mM K2HPO4, 10 % glycerol, 5 mM b-mercaptoethanol, 200 mM NaCl, 1 M imidazole, pH 7.0) for one hour at room temperature. For maximum recovery, repeated elution steps were performed. Typically, the ®rst elution step yielded pure protein at a concentration of 5-10 mg/ml, measured as described (Schaffner & Weissmann, 1973) using bovine serum albumin as a standard. Sedimentation analysis Solubilized, puri®ed 10-His-AqpZ (2-10 mg of protein in 100 ml sample volume) was layered on top of a 4 ml continuous sucrose gradient (5 % to 20 % sucrose, 1.5 % dodecylmaltoside, 5 mM EDTA, 20 mM Tris-HCl (pH 8)) and centrifuged at 140,000 g for 18 hours (34,000 rpm in a Beckman SW 50.1 Ti rotor). Fractions were collected and analyzed by SDS-PAGE. The sedimentation coef®cient (sw,20) was determined by comparison with the fol-
Sequence 0
5 -PCG GGT TGA CAG AAC AGT TAG TCA CC -30 50 -PCG CAG TGC TGA CCG CTG GGT TAA CAG-30 50 -PCG CAG TGC TGC TCG CTG GGT TAA CAG-30 50 -PGC ACT ACC AGA GCC ACC AAA AAC AAG CCA GAA GGT ACC AAA AGA TTC AGC TGC-30 50 -PCC AGA AGG TAC CAA AAG ATT CAG CTG C-30 50 -GCA TGC GAC CTT AGA ATT CGT GAC C-30 50 -PTT CTG TGA CCG GTG AAT ACT CAA CC-30
lowing internal standards: cytochrome c (1.8), catalase (2.9), b-amylase (4.3), bovine serum albumin (8.9) and carbonic anhydrase (11.2). Functional reconstitution of purified AqpZ Puri®ed 10-His-AqpZ was reconstituted into proteoliposomes. The low critical micellar concentration of dodecylmaltoside makes this detergent unsuitable for reconstitution by dialysis, so the detergent octylglucoside was used instead. Brie¯y, a reconstitution mixture was prepared at room temperature by sequentially adding: 100 mM Mops-Na (pH 7.5); 1.25 % (w/v) octylglucoside; puri®ed 10-His-AqpZ (®nal concentration 50-100 mg/ml); and 10 mg/ml lipids. E. coli total lipid extract (acetone/ ether preparation, Avanti Polar Lipids, Inc.) had previously been resuspended in 2 mM b-mercaptoethanol in water (®nal concentration 50 mg/ml) and incubated at room temperature for one hour, diluted to a ®nal concentration of 45 mg/ml in 100 mM Mops-Na (pH 7.5) and pulsed in a bath sonicator until a clear suspension was obtained. Lipid was always handled under a nitrogen atmosphere. The reconstitution mixture was loaded into SPECTRA/POR 2 dialysis tubing, molecular mass cut-off 12,000-14,000 (Spectrum Medical Industries, Inc.), and dialyzed against 100 volumes of assay buffer (50 mM MOPS, 150 mM N-methyl-D-glucamine, adjusted to pH 7.5 with HCl) for 24-72 hours at room temperature. Alternatively, the reconstitution mixture was rapidly injected into 25 volumes of assay buffer to dilute the detergent. Liposomes were harvested by centrifugation (45 minutes at 140,000 g) and resuspended into assay buffer. The diameter of the 10-His-AqpZ proteoliposomes obtained by dialysis was measured from micrographs of cryofractured material (122(20) nm) and by light scattering (130 nm). Membrane permeability measurements The osmotic behavior of reconstituted proteoliposomes and control liposomes was analyzed by following the light scattering of the preparation in a stopped-¯ow apparatus (SV.17MV, Applied Photophysics) with a measured dead time of 0.7 ms. The water permeability was measured by rapid mixture with hyperosmolar assay buffer containing sucrose, causing water to leave the vesicles. Light scattering was recorded at an emission wavelength of 600 nm. Under these conditions a reduction in vesicle volume leads to an increase in the signal. Data were ®tted to a exponential rise equation and initial rates (k) were calculated from the ®tted equation. The osmotic water permeability (Pf) was calculated using the following expression:
1178
Functional Reconstitution of AqpZ Pf k= S=V0 Vw osm
where S/V0 is the vesicle surface area to initial volume ratio, Vw is the partial molar volume of water (18 cm3), and osm is the difference in osmolarity between the intravesicular and extravesicular aqueous solutions (usually 300 mos mol). The radius of proteoliposomes was measured by electron microscopy. The permeability of single AqpZ molecules (pf) was estimated as the difference in permeability between proteoliposomes and liposomes, multiplied by the surface area of a proteoliposome, and divided by the number of AqpZ molecules incorporated to a proteoliposome. The number of AqpZ monomers was estimated from the amount of protein incorporated into the reconstitution mixture. The number of liposomes was estimated assuming an internal volume of 1 ml/mg of phospholipid (Ambudkar & Maloney, 1986). The permeability of the liposomes to small solutes was performed by rapidly mixing a volume of suspension vesicles at a given concentration of permeant solute with a volume of solution with the osmolarity equilibrated with a non-permeant solute. The effective concentration of the solute in the outer medium is then reduced to half, generating an outward gradient. The outward movement of permeant solute generates an outwardly oriented osmotic gradient, which drives water movement. The consequent change in volume is monitored by stopped-¯ow, as described above.
Acknowledgements This work was supported by a fellowship from the Human Frontier Science Program (M.J.B.), the National Institutes of Health and the Cystic Fibrosis Foundation (P.A.), and the National Science Foundation, no. MCB 9603997 (P.C.M.). We thank Dr Mon-Chou Fann and Dr John C. Mathai for discussion of this work, and assistance. We thank Gera D. Eytan for critical reading of the manuscript.
References Abrami, L., Simon, M., Rousselet, G., Berthonaud, V., Buhler, J. M. & Ripoche, P. (1994). Sequence and functional expression of an amphibian water channel, FA-CHIP: a new member of the MIP family. Biochim. Biophys. Acta, 1192, 147-151. Ambudkar, S. V. & Maloney, P. C. (1986). Bacterial anion exchange. Use of osmolytes during solubilization and reconstitution of phosphate-linked antiport from Streptococcus lactis. J. Biol. Chem. 261, 10079-10086. Bixby, K. A., Nanao, M. H., Shen, N. V., Kreusch, A., Bellamy, H., Pfaf®nger, P. J. & Choe, S. (1999). Zn2-binding and molecular determinants of tetramerization in voltage-gated K channels. Nature Struct. Biol. 6, 38-43. Bonhivers, M., Carbrey, J. M., Gould, S. J. & Agre, P. (1998). Aquaporins in Saccharomyces. Genetic and functional distinctions between laboratory and wild-type strains. J. Biol. Chem. 273, 27565-27572. Calamita, G., Bishai, W. R., Preston, G. M., Guggino, W. B. & Agre, P. (1995). Molecular cloning and characterization of AqpZ, a water channel from Escherichia coli. J. Biol. Chem. 270, 29063-29066.
Calamita, G., Kempf, B., Bonhivers, M., Bishai, W. R., Bremer, E. & Agre, P. (1998). Regulation of the Escherichia coli water channel gene aqpZ. Proc. Natl Acad. Sci. USA, 95, 3627-3631. Cheng, A., van Hoek, A. N., Yeager, M., Verkman, A. S. & Mitra, A. K. (1997). Three-dimensional organization of a human water channel. Nature, 387, 627630. Coury, L. A., Mathai, J. C., Prasad, G. V., Brodsky, J. L., Agre, P. & Zeidel, M. L. (1998). Reconstitution of water channel function of aquaporins 1 and 2 by expression in yeast secretory vesicles. Am. J. Physiol. 274, F34-F42. Denker, B. M., Smith, B. L., Kuhajda, F. P. & Agre, P. (1988). Identi®cation, puri®cation, and partial characterization of a novel Mr 28,000 integral membrane protein from erythrocytes and renal tubules. J. Biol. Chem. 263, 15634-15642. Finkelstein, A. (1987). Water Movement Through Lipid Bilayers, Pores, and Plasma Membranes: Theory and Reality, John Wiley & Sons, New York. Helenius, A. & Simons, K. (1975). Solubilization of membranes by detergents. Biochim. Biophys. Acta, 415, 2979. Huse, M., Eck, M. J. & Harrison, S. C. (1998). A Zn2 ion links the cytoplasmic tail of CD4 and the Nterminal region of Lck. J. Biol. Chem. 273, 1872918733. Jung, J. S., Preston, G. M., Smith, B. L., Guggino, W. B. & Agre, P. (1994). Molecular structure of the water channel through aquaporin CHIP. The hourglass model. J. Biol. Chem. 269, 14648-14654. Lagree, V., Pellerin, I., Hubert, J. F., Tacnet, F., Le, Caherec F., Roudier, N., Thomas, D., Gouranton, J. & Deschamps, S. (1998). A yeast recombinant aquaporin mutant that is not expressed or mistargeted in Xenopus oocyte can be functionally analyzed in reconstituted proteoliposomes. J. Biol. Chem. 273, 12422-12426. Laize, V., Rousselet, G., Verbavatz, J. M., Berthonaud, V., Gobin, R., Roudier, N., Abrami, L., Ripoche, P. & Tacnet, F. (1995). Functional expression of the human CHIP28 water channel in a yeast secretory mutant. FEBS Letters, 373, 269-274. le Coutre, J., Kaback, H. R., Patel, C. K., Heginbotham, L. & Miller, C. (1998). Fourier transform infrared spectroscopy reveals a rigid alpha-helical assembly for the tetrameric Streptomyces lividans K channel. Proc. Natl Acad. Sci. USA, 95, 6114-6117. Li, H., Lee, S. & Jap, B. K. (1997). Molecular design of aquaporin-1 water channel as revealed by electron crystallography. Nature Struct. Biol. 4, 263-265. Ma, T., Yang, B. & Verkman, A. S. (1996). cDNA cloning of a functional water channel from toad urinary bladder epithelium. Am. J. Physiol. 271, C1699C1704. Mathai, J. C. & Agre, P. (1999). Hourglass pore-forming domains restrict aquaporin-1 tetramer assembly. Biochemistry, 38, 923-928. Park, J. H. & Saier, M. H. (1996). Phylogenetic characterization of the MIP family of transmembrane channel proteins. J. Membr. Biol. 153, 171-180. Preston, G. M., Carroll, T. P., Guggino, W. B. & Agre, P. (1992). Appearance of water channels in Xenopus oocytes expressing red cell CHIP28 protein. Science, 256, 385-387. Preston, G. M., Jung, J. S., Guggino, W. B. & Agre, P. (1993). The mercury-sensitive residue at cysteine
Functional Reconstitution of AqpZ 189 in the CHIP28 water channel. J. Biol. Chem. 268, 17-20. Ringler, P., Borgnia, M. J., Stahlberg, H., Maloney, P., Agre, P. & Engel, A. (1999). Structure of water channel Aqpz from Escherichia coli revealed by electron crystallography. J. Mol. Biol. 291, 1181-1190. Schaffner, W. & Weissmann, C. (1973). A rapid, sensitive, and speci®c method for the determination of protein in dilute solution. Anal. Biochem. 56, 502514. Smith, B. L. & Agre, P. (1991). Erythrocyte Mr 28,000 transmembrane protein exists as a multisubunit oligomer similar to channel proteins. J. Biol. Chem. 266, 6407-6415. Tamai, E., Fann, M. C., Tsuchiya, T. & Maloney, P. C. (1997). Puri®cation of UhpT, the sugar phosphate transporter of Escherichia coli. Protein Exp. Purif. 10, 275-282. van Hoek, A. N. & Verkman, A. S. (1992). Functional reconstitution of the isolated erythrocyte water channel CHIP28. J. Biol. Chem. 267, 18267-18269. Verbavatz, J. M., Brown, D., Sabolic, I., Valenti, G., Ausiello, D. A., van Hoek, A. N., Ma, T. & Verkman, A. S. (1993). Tetrameric assembly of CHIP28 water channels in liposomes and cell membranes: a freeze-fracture study. J. Cell Biol. 123, 605618. Walz, T., Smith, B. L., Agre, P. & Engel, A. (1994). The three-dimensional structure of human erythrocyte aquaporin CHIP. EMBO J. 13, 2985-2993.
1179 Walz, T., Hirai, T., Murata, K., Heymann, J. B., Mitsuoka, K., Fujiyoshi, Y., Smith, B. L., Agre, P. & Engel, A. (1997). The three-dimensional structure of aquaporin-1. Nature, 387, 624-627. Weig, A., Deswarte, C. & Chrispeels, M. J. (1997). The major intrinsic protein family of Arabidopsis has 23 members that form three distinct groups with functional aquaporins in each group. Pl. Physiol. 114, 1347-1357. Yang, B. & Verkman, A. S. (1997). Water and glycerol permeabilities of aquaporins 1-5 and MIP determined quantitatively by expression of epitopetagged constructs in Xenopus oocytes. J. Biol. Chem. 272, 16140-16146. Zeidel, M. L., Ambudkar, S. V., Smith, B. L. & Agre, P. (1992). Reconstitution of functional water channels in liposomes containing puri®ed red cell CHIP28 protein. Biochemistry, 31, 7436-7440. Zeidel, M. L., Nielsen, S., Smith, B. L., Ambudkar, S. V., Maunsbach, A. B. & Agre, P. (1994). Ultrastructure, pharmacologic inhibition, and transport selectivity of aquaporin channel-forming integral protein in proteoliposomes. Biochemistry, 33, 1606-1615. Zhuang, J., Prive, G. G., Werner, G. E., Ringler, P., Kaback, H. R. & Engel, A. (1999). Two-dimensional crystallization of Escherichia coli lactose permease. J. Struct. Biol. 125, 63-75.
Edited by W. Baumeister (Received 9 April 1999; received in revised form 30 June 1999; accepted 16 July 1999)