Mercurial sensitivity of aquaporin 1 endofacial loop B residues KUNYAN KUANG,2 JORGE F. HALLER,2 GUANGPU SHI,2 FENGYING KANG,2,3 MIN CHEUNG,2,4 PAVEL ISEROVICH,2 AND JORGE FISCHBARG1,2 1 Department of Physiology and Cellular Biophysics, College of Physicians and Surgeons, Columbia University, New York, New York 10032, USA 2 Department of Opthalmology, College of Physicians and Surgeons, Columbia University, New York, New York 10032, USA
(RECEIVED February 7, 2001; FINAL REVISION April 16, 2001; ACCEPTED May 16, 2001)
Abstract The water channel protein aquaporin-1 (AQP1) has two asparagine-proline-alanine (NPA) repeats on loops B and E. From recent structural information, these loops are on opposite sides of the membrane and meet to form a pore. We replaced the mercury-sensitive residue cysteine 189 in AQP1 by serine to obtain a mercury-insensitive template (C189S). Subsequently, we substituted three consecutive cysteines for residues 71–73 near the first NPA repeat (76–78) in intracellular loop B, and investigated whether they were accessible to extracellular mercurials. AQP1 and its mutants were expressed in Xenopus laevis oocytes, and the osmotic permeability (Pf) of the oocytes was determined. C189S had wild-type Pf but was not sensitive to HgCl2. Expression of all three C189S cysteine mutants resulted in increased Pf, and all three mutants regained mercurial sensitivity. These results, especially the inhibitions by the large mercurial p-chloromercunbenzene-sulfonic acid (pCMBS) (∼6Å wide), suggest that residues 71–73 at the pore are accessible to extracellular mercurials. A 30-ps molecular dynamics simulation (at 300 K) starting with crystallographic coordinates of AQP1 showed that the width of the pore bottleneck (between Connolly surfaces) can vary (wavg ⳱ 3.9 Å, ⳱ 0.75; hydrated AQP1). Thus, although the pore width would be ⱖ 6 Å only for 0.0026 of the time, this might suffice for pCMBS to reach residues 71–73. Alternative explanations such as passage of pCMBS across the AQP1 tetramer center or other unspecified transmembrane pathways cannot be excluded. Keywords: Water channel; pore; hourglass model; osmotic permeability
Water channels proteins are currently the subject of great attention, as they appear important to cell function in a host of tissues. Analysis of the amino acid sequences of the members of the major intrinsic protein (MIP) family has delineated the presence of a twofold repeat in the primary structure of these proteins (Preston and Agre 1991; ChepeReprint requests to: Dr. Jorge Fischbarg, Department of Physiology and Cellular Biophysics, College of P. & S., Columbia University, 630 West 168th Street, New York, NY 10032, USA; e-mail:
[email protected]; fax: 212-305-2461. 3 Present address: Department of Ophthalmology, Weifang Medical College, Shandong, China. 4 Present address: Department of Radiation Oncology, 2 Donner, Hospital of the University of Pennsylvania, 3400 Spruce Street, Philadelphia, PA 19104, USA. Article and publication are at www.proteinscience.org/cgi/doi/10.1110/ ps.5901.
linsky 1994). Preston and Agre (1991) predicted that the two characteristic asparagine-proline-alanine (NPA) motifs in loops B and E of aquaporin-1 (AQP1) were on opposite sides of the membrane. Later on, Agre and colleagues (Jung et al. 1994) developed the hourglass model, in which the cytoplasmic loop B and the extracellular loop E overlap within the membrane to form the water pore. Recent electron crystallographic analysis of AQP1 structure (Cheng et al. 1997; Mitsuoka et al. 1999; Heymann and Engel 2000; Murata et al., 2000; Ren et al., 2000, 2001) and another MIP protein, the glycerol facilitator GlpF (Fu et al. 2000), are consistent with the NPA boxes delimiting part of the pore. The wild-type AQP1 water channel is characteristically sensitive to mercurials placed outside of cells. It has been established that one single cysteine residue in its chain (Cys 189) confers that property (Preston et al. 1993; Zhang et al.
Protein Science (2001), 10:1627–1634. Published by Cold Spring Harbor Laboratory Press. Copyright © 2001 The Protein Society
1627
Kuang et al.
1993), presumably by steric blockage of the aqueous pore through the protein. Replacing C189 by a non-sulfhydryl containing amino-acid-like serine (C189S) makes AQP1 mercurial insensitive (Preston et al. 1993; Zhang et al. 1993). Hence, by use of C189S as a template, further substitution of Cys residues at strategic locations can be used to test the accessibility to mercurials for the residues replaced. For this work, we substituted cysteine one at a time (Akabas et al. 1992; Xu and Akabas 1993) for three consecutive residues (71–73) near the first NPA repeat (76–78). The cysteine AQP1 mutants were expressed in Xenopus laevis oocytes and the osmotic permeability (Pf) of oocytes expressed was determined under control conditions and in the presence of the mercurial reagents HgCl2 and p-chloromercuribenzene-sulfonic acid (pCMBS). Unexpectedly, we found that all three mutants were sensitive to both mercurials. To study such behavior, we ran molecular dynamics (MD)simulations of AQP1 at 300 K, and found that its pore bottleneck can vary in width over a surprisingly wide range (avg. width ⳱ 3.9 Å, ⳱ 0.75; hydrated AQP1). This seems consistent with pCMBS (width ∼6 Å) possibly traversing the pore if the incubation time is long enough. We discuss how residues 71–73 could be accessible to external mercurials by this or alternative routes, and the possible implications for MIP protein selectivities.
Results Osmotic permeability of oocytes The Pf of the oocytes was determined 3 d after injection. Their volume change was monitored after osmotic challenge (going from 178 to 15 mOsm Barth’s medium at room temperature). As expected, oocytes injected with wild-type AQP1 cRNA had high Pf values (215.2 ± 12 m/s, Figs.1 and 2), whereas water-injected oocytes exhibited very small Pf (15.4 ± 1.4 m/s). The increased Pf by the oocytes expressing AQP1 was inhibited by 1-min incubation in 1 mM HgCl2 (25.7 ± 2.4 m/s, Figs. 1 and 2). Oocytes injected with cRNA encoding the mutant C189S exhibited wild-type Pf (205.1 ± 27.7 m/s, Figs. 1 and 2) but were no longer sensitive to external 1 mM HgCl2 (Pf ⳱ 200.7 ± 15.7 m/s). Such results are the same as those in prior studies (Preston et al. 1993; Zhang et al. 1993) that established that cysteine 189 is the Hg2+-sensitive residue in AQP1. Thereafter, by use of C189S as a template, cysteines were substituted to the individual residues 71–73 in loop B. The residue A73 (near the NPA motif in loop B) corresponds to the position C189 (near the NPA motif in loop E). Figure 1 (curve c) shows that the osmotically induced changes in oocytes expressing A73C/C189S are similar to those expressing AQP1 (curve a) or the mutant C189S
Fig. 1. Determination of osmotic permeability of oocytes. Oocytes (numbers shown in parenthesis) were injected with either 50 nL water or cRNA encoding wild-type AQP1 or its mutants 4 d prior to these experiments. In each experiment, an oocyte was subjected to a 92% hypotonic challenge (from 180 to 15 mOsm). Where noted, HgCl2 was added to the outside medium 1 min prior to hypotonic challenge. (䊐) AQP1 (5); (䊊) C189S (5); (䉭) A73C/C189S (6); (〫) H2O (5); (䊏) AQP1; HgCl2 (6); (䊉) C189S; HgCl2 (5); (䉱) A73C/C189S; HgCl2 (6); (⽧) H2O; HgCl2 (5). Data are AVG ± SEM.
1628
Protein Science, vol. 10
Mercurial sensitivity of AQP1 loop B residues
Fig. 2. Osmotic permeability of oocytes expressing wild-type AQP1 or its mutants, and the effect on it of 1 mM external HgCl2 and pCMBS. Pf values were determined from initial slopes of volume change after osmotic challenge for each oocyte. N is 5–6 oocytes per group. (Hatched bars) 1 mM pCMBS; (open bars) 1mM HgCl2; (solid bars) untreated. Data are AVG ± SEM.
(curve b). However, the mutant A73C/C189S was once again sensitive to HgCl2 (curve g). Oocytes expressing mutants S71C/C189S, G72C/C189S had comparatively high Pf values (Fig. 2: 72.3 ± 4.8, 166.5 ± 15.5, and 155.6 ± 14.1 m/s, respectively). Such values are similar to those observed with A73C/C189S, but somewhat lower than those seen with wild-type AQP1 or mutant C189S. The key result here is that all three mutants regained susceptibility to mercury inhibition; the Pf expressed by these mutants was inhibited by 1 mM HgCl2 (Fig. 2: 23.9 ± 2.4, 61.9 ± 15, and 20.3 ± 0.9 m/s, for S71C/C189S, G72C/C189S, and A73C/C189S, respectively). We also examined the effects of 1 mM pCMBS on oocytes incubated with it for 30 min. The Pf expressed by oocytes injected with wild-type AQP1, S71C/C189S, G72C/C189S, or A73C/C189S, was inhibited significantly by pCMBS (Fig. 2: 42.5 ± 3.3, 69.1 ± 18.1, and 51.9 ± 10 m/s, respectively).
sitions; this was quite visible when an animated sequence was made with the frames sampled every 50 fs (not shown). In keeping with this trend, the pore also changed its dimensions with time. Figure 5 shows the particularly narrow region of the pore, or vestibule, on which we have centered our attention. The figure exemplifies how the distance between the given opposing residues varied between the initial condition (Fig. 5A, 6.40 Å) to that 17.9-ps later (Fig. 5B, 7.66 Å) during the simulation by use of hydrated AQP1. Figure 6 shows the progression of pore widths during 20 ps (sampled at 100-fs intervals) for both simulations (AQP1 in vacuum, and hydrated). To define pore widths, three residues of interest (Phe 24, Leu 149, and Ile 191) were shown with Connolly (1983) surfaces (with InsightII; atom radius scale: 0.7 of VDW; probe radius 1.4 Å). The surfaces were at 0.8 Å of the H centers, from which pore widths were estimated accordingly from the interatomic distances. As can be seen, the estimated widths vary considerably during the short time interval examined. Figure 7 shows the distribution of pore widths for the two simulations with hydrated AQP1 given in Figure 6. As can be calculated from those data, although the average pore width was 3.9 ± 0.75 Å, there was a 0.003 probability that the pore would be ⱖ 6 Å. Discussion Because AQP1 contains four cysteines, but C189 is its only mercury-sensitive residue, it was accepted that residue C189 in extracellular loop E is in contact with the extracellular medium (Jung et al. 1994) before structural studies reaffirmed it. As for the other three AQP1 cysteines, C102 and C152 are in bilayer-spanning domains, presumably inaccessible locations. C87 is located in the short helix in loop B (A78-S86) (Mitsuoka et al. 1999). The short helices in
Molecular dynamics We utilized the AQP1 coordinates (at 3.8 Å resolution) communicated by Murata el al. (2000). Prior to the simulations, a recommended step of energy minimization was performed as described in Materials and Methods. Figures 5–7, below, show the results of the MD simulations. The simulations ran for 30 ps, during which time the positions of the atoms oscillate within a steady-state range throughout the simulation. The events during the first ∼20 ps depicted here are adequate to describe the process. The simulations revealed that the entire protein appeared quite mobile. Helices and loops changed their relative po-
Fig. 3. Dimensions of the two mercurial sulfhydryl reagents we have used. Three-dimensional models were obtained by use of the HyperChem program (Hypercube); dots represent space-filling spheres.
www.proteinscience.org
1629
Kuang et al.
Fig. 4. Stereoscopic view of AQP1 according to the coordinates of Murata et al. (2000); exofacial side at the top. Helices 1–6 are colored (blue, pink, green, red, orange, and yellow, respectively). Loops and short pore helices are white, and the residues we mutated (71–73) are in light blue.
loop B and loop E form a structural unit hydrophobic enough to span the membrane (Heymann and Engel 2000); given its apparent lack of reactivity, C87 is presumably also inaccessible to mercurials. It also seems relevant that, as Preston et al (1993) reported, oocytes expressing serine mutants of these four cysteines (C87S, C102S, C152S, and C189S) exhibited Pf values equivalent to those found with wild-type AQP1. From that report, apparently the introduction of a residue-like serine, smaller than the existing cysteine, did not disrupt the fold significantly. Moreover, HgCl2 inhibited the mutants C87S, C102S, or C152S to the same extent as wild-type AQP1, and did not inhibit C189S (Preston et al. 1993; Zhang et al. 1993). Therefore, we presume the other three cysteine residues are inaccessible to outside mercurials. The Pf values we determined for our mutants G72C/ C189S and A73C/C189S were relatively high, but somewhat lower than those obtained with wild-type AQP1. Hence, these mutations did not appear to affect the overall protein fold. The ∼30% decrease in Pf we observed for them compared with the wild type (Fig. 2, second and third bar groups) might be due to increased steric hindrance as C is bulkier than G or A. In support of this hypothesis, mutants A73F and A73M yielded much-reduced Pf in a prior study (Jung et al. 1994). However, different expression levels of the mutants cannot be excluded as an explanation. It seems also interesting that the mutation S71C resulted in a decrease in Pf to one-third of that in wild-type or C189S 1630
Protein Science, vol. 10
mutant-expressing oocytes (Fig. 2, fourth bar group ). S71 seems to be important to stabilize the molecular structure (Y. Fujiyoshi, pers. comm.), therefore, it is conceivable that replacement of this residue might lead to complex structural changes and lowered Pf. Rather surprisingly, our main observation is that the three C mutants (from 71to 73) recovered the sensitivity to both HgCl2 and pCMBS added to the outside (see Fig. 2). In support of our interpretations, we note that Jung et al. (1994) preceded us in showing that A73C/C189S exhibited mercurial sensitivity (∼60% inhibition for 1 mM HgCl2, compared with ∼90% inhibition in the present data). From this result, they concluded that residue 73 could be placed near the cytoplasmic face of the single aqueous pathway across the protein, which appears remarkably prescient in view of recent structural information. In our results, the membrane-impermeant sulfhydryl reagent pCMBS (see Fig. 2) blocks the Pf corresponding to mutant expression by some 70% (just as it does in the case of wild-type AQP1). How the mercurials, especially the larger one, pCMBS, would reach these residues is less clear. One possibility is that the mutations in residues 71–73 alter the stability of the molecule and produce slight conformation changes either enhancing the diameter of the pore or exposing cysteines that before were inaccessible to mercurials. For example, this might happen with Cys 102, which is close to the narrowest region of the pore (cf. Fig. 4). However, another explanation that deserves attention is that
Mercurial sensitivity of AQP1 loop B residues
Fig. 5. (A,B). These figures depict two states of hydrated AQP1 during a 30-ps molecular dynamics simulation (300 K). Both views represent transversal slices at an equatorial plane corresponding to the narrowest region of the pore. A shows the initial pore; B represents the frame at 17.9 ps depicting the widest pore observed. Distances shown are between the para H of Phe 24 and the nearest H in the opposing residue (either Ile 191 or Leu 149). Only the backbone is displayed, except for the residues around the bottleneck.
www.proteinscience.org
1631
Kuang et al.
Fig. 6. Time sequences of pore widths obtained from two molecular dynamics simulations of AQP1 (in vacuum and hydrated) sampled at 100-fs intervals for 20 ps. Pore widths were estimated by adding 1.6 Å to the distances between the atoms shown in Fig. 5. (䊐) In vacuum; (䉱) hydrated.
the pore dimensions might vary. In the MD simulations at 300 K we report here, the width of the pore went from the 3 Å described by Murata et al (2000) to, if briefly, as much as ∼6.1 Å (Fig. 6), a size large enough to allow pCMBS (∼6 Å wide, cf. Fig. 3) to traverse the pore. There is a range of values for the diameter of the AQP1 pore in different reports [∼3.0 Å (Murata et al. 2000); ∼4.5 Å (Mitsuoka et al. 1999);
Fig. 7. Histogram of the pore width values obtained in the simulation for hydrated AQP1 depicted in Fig. 6.
1632
Protein Science, vol. 10
∼6.5 Å (Ren et al. 2000); ∼4.0 ± 0.5 Å (Ren et al. 2001)]. This variation might be due mostly to the limited resolutions achieved this far, but perhaps might also reflect the existence of different conformations. In this connection, it seems interesting that there is evidence that in lactose permease, as temperature increases, additional regions are exposed to pCMBS inhibition (Venkatesan et al. 2000). As a note of caution, given that the starting point of our simulations is a relatively low-resolution model, there may be an inherent inaccuracy, especially for the positions of the side chains. On the other hand, we observed comparatively large mobility for the helical segments with both the esff and the cvff force fields (not shown), so this may correspond to the general behavior of the protein. In a related subject, the time sequence for the hydrated AQP1 in Figure 6 shows what are roughly two overall pore width levels. Interestingly, this might indicate that the pore could be gated. More work is required to refine these aspects. As stated in the Results, from the simulations with hydrated AQP1, the probability of the pore reaching a value ⱖ 6 Å appears to be only 0.003. Hence, for pCMBS to migrate all of the length of the pore to reach the intracellular end, comparatively long incubation times would be necessary. Yet, this is what happens in practice as oocytes are preincubated with pCMBS typically for 20–30 min. If pore widths are that variable, one would then need to contend with the possibility that substrates normally excluded by
Mercurial sensitivity of AQP1 loop B residues
AQP1 such as urea and glycerol might occasionally traverse the pore. On one hand, such events are rather rare, and, hence, almost undetectable. On the other hand, an increase in pore width might not necessarily increase the selectivity for such substrates. Such selectivity might rather stem crucially from favorable binding sites rather than steric factors. That this may be so can be gleaned from the cogent arguments given to explain glycerol selectivity by the glycerol facilitator GlpF (a glyceroaquaporin), for which crystallographic coordinates at 2.2 Å resolution have been recently communicated (Fu et al. 2000). In that case, several characteristic arrangements along the pore would add up to contribute to enhance glycerol selectivity. In addition, we note that another glyceroaquaporin, MIP26, has a water conductance of only ∼15% that of AQP1 (Zampighi et al. 1995), whereas its pore diameter has to be more than adequate for water passage. Therefore, conversely, it could be that in some glyceroaquaporins, the motifs that enhance water selectivity in aquaporins would be either absent or have less weight. Another question that this hypothesis poses is whether given substrates could interact with the protein and contribute to modify its dynamic behavior, enlarging the pore width as they move. Passage of pCMBS across AQP1 might be possible this way. Even if one contemplates the possibility of passage of pCMBS across another pathway such as the center of the AQP1 tetramer, from recent evidence, the corresponding width is only ∼ 3 Å (Ren et al. 2001), which is inadequate unless dynamic changes are once more postulated. Lastly, it is conceivable that pCMBS would traverse the membrane through pathways other than AQP1. In this view, a carrier permeable to pCMBS may be coexpressed with AQP1, or the expression of AQP1 might lead to unspecified local changes in the lipid allowing the mercurial(s) across. These possibilities appear somewhat unlikely to us, but have not been ruled out by the present work. Materials and methods Oligonucleotide-mediated mutagenesis and in vitro RNA synthesis AQP1 mutants were constructed with the Altered Sites II in vitro Mutagenesis System kit (Promega). Restriction enzymes were se-
lected so as to ensure proper insertion into the pAlter-1 vector (from the kit above); The cDNA in the vector pBluescript SK(−) (AQP-pBS) was thus digested with KpnI and BamHI (New England Biolabs). The resulting fragment of ∼2600 bases (which included the full-length AQP1 cDNA) was ligated into pAlter-1 cut with KpnI and BamHI. This vector had its tetracycline site active, which was used to select and amplify the ligation product (AQP-pAlter) by use of Escherichia coli cells. Mutation oligonucleotides (27 to 33 bases) were designed (by use of the Gene Runner program, Hastings Software) so as to generate the appropriate cysteine replacement or substitution while simultaneously introducing a silent restriction site; this usually requires a 3–4 base mismatch (Table 1). After generating single-strand AQP1-pAlter, two phosphorylated oligonucleotides were annealed to it as follows: (1) the one with the mutation, and (2) a second one that repairs the ampicillin-resistance site. After second-strand synthesis and transformation of temporary host-competent cells, colonies with the mutant were selected; the mutant DNA (AQPmutpAlter) was then extracted and used to transform XL1-blue E. coli cells, suited for long-term maintenance and segregation. Ampicillin-resistant colonies (20%–80% of total colonies) were selected; mutations were identified by restriction digests and confirmed by DNA sequencing.
Expression and assay of the mutants Mutants were excised from AQPmut-pAlter by use of KpnI and BamHI, and were ligated back into AQP-pBS cut with the same enzymes; such transfer was necessary to assay the mutant function by expressing it in Xenopus laevis (African toad) oocytes. For in vitro RNA transcription, AQP-pBS or AQPmut-pBS constructs were linearized with KpnI, and messenger RNA was synthesized by use of T3 RNA polymerase (mmessage mmachine, Ambion).
Preparation of oocytes and measurement of Pf Oocytes were removed from the ovaries of Xenopus laevis (NASCO), defolliculated, and separated (Echevarria et al. 1993). The largest undamaged oocytes (stages 5 and 6) were transferred to Barth’s medium. Oocytes were injected with 50 nL of either water or cRNA (25 ng/l) as described (Echevarria et al. 1993) and were incubated at 18°C for 72 h. Oocytes were then transferred to a glass-bottom chamber containing Barth’s medium at room temperature. The oocyte equatorial cross-section was viewed with a Nikon TMs inverted microscope equipped with a video camera (model NC-65, Dage-MTI) connected to a monitor screen. Oocytes were superfused at first in isotonic Barth’s solution (178 mOsm) for a period of 60 s, and then in hypotonic solution (by reducing NaCl, 15 mOsm) for another 100 s. A frame grabber recorded an image every 10 s, and a computer calculated oocyte
Table 1. Cysteine site-directed mutagenesis Wild-type AA Ser-71 Gly-72 Ala-73 Cys-189 a
Mutant
Codon
AA
Codon
Antisense oligonucleotidesa
AGT GGT GCT TGT
Cys-71 Cys-72 Cys-73 Ser-189
TGC TGT TGT TCC
5⬘287TTGAGGTGAGCACC[GCA]TATGTGACCCACA2583⬘ 5⬘287TTGAGGTGAGC[ACA]GCTGATGTGACCC2613⬘ 5⬘287TTGAGGTG[ACA]ACCGCTGATGTGAC2633⬘ 5⬘643GGGCAGGGTTGATTCC[GGA]GCCAGTGTAG6153⬘
The underlined bases are mismatched; bases in brackets represent codons of mutants.
www.proteinscience.org
1633
Kuang et al.
area and volume from such image. Pf values were calculated from the induced change in oocyte volume, Pf ⳱ (dV/dt) × [1/(A × Vw × ⌬C)], in which A was the area (assumed spherical) of the oocyte at zero time, dV/dt was the rate of volume change at zero time, and ⌬C was the osmolarity gradient at zero time.
Molecular dynamics MD were performed using as the starting position the AQP1 coordinates reported by Murata et al. (2000). We used the Discover 3 module of InsightII (Molecular Simulations Inc.; version 97.2); the force field was ESFF (Extensible Systemic Force Field). We ran simulations as follows: (1) with AQP1 in vacuum, and (2) with the exofacial and endofacial vestibules plus the pore of AQP1 hydrated (using the assembly/soak command of InsightII, which added 254 water molecules). Initially, before running dynamics, the energy was optimized with the steepest descent and conjugate algorithms (1000 steps each). Parameters for MD were as follows: equilibration 5000 steps, run 30,000 steps, temperature 300 K, time step 1 femtosecond, and no constraints. Otherwise, settings were default. History files for the 30-ps runs recorded the position of the atoms every 50 steps.
Acknowledgments This work was supported by National Institutes of Health Grant no. EY08918, and in part by Fight for Sight, Inc., and Research to Prevent Blindness, Inc. The publication costs of this article were defrayed in part by payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 USC section 1734 solely to indicate this fact.
References Akabas, M.H., Stauffer, D.A., Xu, M., and Karlin, A. 1992. Acetylcholine receptor channel structure probed in cysteine-substitution mutants. Science 258: 307–310. Cheng, A., van Hoekm A.N., Yeager, M., Verkman, A.S., and Mitra, A.K. 1997. Three-dimensional organization of a human water channel. Nature 387: 627–630.
1634
Protein Science, vol. 10
Chepelinsky, A.B. 1994. The MIP transmembrane channel family. In Handbook of membrane channels: Molecular and cellular physiology, (ed. C. Peracchia), pp. 413–432, Academic Press, San Diego, CA. Connolly, M.L. 1983. Solvent-accessible surfaces of proteins and nucleic acids. Science 221: 709–713. Echevarria, M., Kuangm, K., Iserovich, P., Li, J., Preston, G.M., Agre, P., and Fischbarg, J. 1993. Cultured bovine corneal endothelial cells express CHIP28 water channels. Am. J. Physiol. 265: C1349–C1355. Fu, D., Libson, A., Miercke, L.J., Weitzman, C., Nollert, P., Krucinski, J., and Stroud, R.M. 2000. Structure of a glycerol-conducting channel and the basis for its selectivity. Science 290: 481–486. Heymann, J.B. and Engel, A. 2000. Structural clues in the sequences of the aquaporins. J. Mol. Biol. 295: 1039–1053. Jung, J.S., Preston, G.M., Smith, B.L., Guggino, W.B., and Agre, P. 1994. Molecular structure of the water channel through aquaporin CHIP. The hourglass model. J. Biol. Chem. 269: 14648–14654. Mitsuoka, K., Murata, K., Walz, T., Hirai, T., Agre, P., Heymann, J.B., Engel, A., and Fujiyoshi, Y. 1999. The structure of aquaporin-1 at 4.5-A resolution reveals short alpha- helices in the center of the monomer. J. Struct. Biol. 128: 34–43. Murata, K., Mitsuoka, K., Hirai, T., Walz, T., Agre, P., Heymann, J.B., Engel, A., and Fujiyoshi, Y. 2000. Structural determinants of water permeation through aquaporin-1. Nature 407: 599–605. Preston, G.M. and Agre, P. 1991. Isolation of the cDNA for erythrocyte membrane protein of 28 kilodaltons: Member of an ancient channel familiy. Proc. Natl. Acad. Sci. 88: 11110–11114. Preston, G.M., Jung, J.S., Guggino, W.B., and Agre, P. 1993. The mercurysensitive residue at cysteine 189 in the CHIP28 water channel. J. Biol. Chem. 268: 17–20. Ren, G., Cheng, A., Reddy, V., Melnyk, P., and Mitra, A.K. 2000. Threedimensional fold of the human AQP1 water channel determined at 4 A resolution by electron crystallography of two-dimensional crystals embedded in ice. J. Mol. Biol. 301: 369–387. Ren, G., Reddy, S., Cheng, A., Melnyk, P., and Mitra, A.K. 2001. Visualization of a water-selective pore by electron crystallography in vitreous ice. Proc. Natl. Acad. Sci. 98: 1398–1403. Venkatesan, P., Hu, Y., and Kaback, H.R. 2000. Site-directed sulfhydryl labeling of the lactose permease of Escherichia coli: Helix X. Biochemistry 39: 10656–10661. Xu, M. and Akabas, M.H. 1993. Amino acids lining the channel of the gammaaminobutyric acid type A receptor identified by cysteine substitution. J. Biol. Chem. 268: 21505–21508. Zampighi, G.A., Kreman, M., Boorer, K.J., Loo, D.D.F., Bezanilla, F., Chandy, G., Hall, J.E., and Wright, E.M. 1995. A method for determining the unitary functional capacity of cloned channels and transporters expressed in Xenopus laevis oocytes. J. Membr. Biol. 148: 65–78. Zhang, R., van-Hoek, A.N., Biwersi, J., and Verkman, A.S. 1993. A point mutation at cysteine 189 blocks the water permeability of rat kidney water channel CHIP28k. Biochemistry 32: 2938–2941.
