Aquaporins: Phylogeny, Structure, and Physiology of Water Channels

Aquaporins: Phylogeny, Structure, and Physiology of Water Channels J. Bernard Heymann and Andreas Engel How water permeates cellular membranes and what this means for cell functioning and several diseases are now emerging from the study of the aquaporins (AQPs), the water channel family. A combination of sequence analysis, three-dimensional structure determination, and physiology of the AQP family proteins provides a glimpse into the workings of water channels. E very living cell must deal with the osmotic and hydrostatic pressure changes in its environment. However, the mechanisms cells employ to transport water and maintain turgor were largely unknown until the discovery of particular membrane proteins serving as channels specific for water and other small nonionic molecules. The first channel cloned and expressed in Xenopus oocytes to show water permeation was aquaporin (AQP)1 from human red blood cells (10). The existence of protein channels for nonionic compounds and, in particular, water has important implications for water management in living organisms and presents interesting requirements of their molecular design. The structure of such proteins must enforce high specificity while allowing a high flux through the channel. Some variance in specificity among the different members of the water channel family appears to be associated with subtle changes in the sequences. How these features are encoded is the subject of ongoing structural studies of various AQPs. The need for such channels is clearly demonstrated in diseases such as nephrogenic diabetes insipidus associated with defects in functional AQP2 expression in the kidney collecting duct and cataract formation as a consequence of mutant AQP0 (= MIP, major integral protein) in the eye lens. Although water may pass through other channels such as those for sodium and calcium, the rates are insufficient to account for the large fluxes observed (in the range of 109–1010 molecules·s-1·channel-1). The questions posed vary from how the channel works to how this ties in with water management and diseases. The family of water channels Long before anything was known of its function, the MIP (=AQP0) of the eye lens was J. B. Heymann and A. Engel are in the M. E. Müller Institute for Microscopic Structural Biology, Biozentrum, University of Basel, CH-4056 Basel, Switzerland. 0886-1714/99 5.00 © 1999 Int. Union Physiol. Sci./Am.Physiol. Soc. sequenced (for a review of the history, see Ref. 5). AQP1 was then cloned and classified with some related sequences of suspected channel proteins, followed by actual demonstration that it is a water channel (10). With the subsequent dramatic increase in similar sequences found in many organisms, these proteins became known as the MIP family. However, using the MIP designation is problematic, because a search in any of the popular sequence and structure databases for the keyword “MIP” also yields proteins not related to the aquaporins, such as the mitochondrial intermediate peptidase and the macrophage infectivity potentiator. A renaming of the family to the “aquaporin family” or “AQP family” is suggested and is used in this review. The subsequent explosion in the number of sequences obtained for organisms of most major taxa suggests that the AQP family consists of old proteins that are important for life. From the scope and variety of AQP family sequences, the impression is that homologs occur in most organisms and most cell types. In some wellstudied organisms such as rats, humans, and Arabidopsis thaliana, several AQP isoforms with somewhat different specificities and distributions have been found. What is the significance of these similarities and differences for structure and function? The rapid increase in the number of sequences is not complemented by sufficient physiological and functional studies, resulting in a confusion of naming schemes and classifications. The names given to the sequences are laboratory specific and strongly influenced by historical issues. Therefore, there exists a need to develop a consistent nomenclature based on function, cellular location, and sequence similarity. Agre (1) initiated this effort by establishing a consistent naming scheme for mammalian AQPs. In this section, the available sequences are analyzed to establish relationships and to examine features possibly associated with function, leading to classification and nomenclature proposals. News Physiol. Sci. • Volume 14 • October 1999 “. . . the AQP family consists of old proteins that are important for life.” 187 Downloaded from www.physiology.org/journal/physiologyonline by ${individualUser.givenNames} ${individualUser.surname} (168.151.137.021) on November 6, 2018. Copyright © 1999 American Physiological Society. All rights reserved. FIGURE 1. Phylogenetic analysis of the aquaporin (AQP) family suggests a classification into two clusters [AQP and glycerol facilitator-like protein (GLP)], 16 subfamilies, and 46 types. The types are considered to be representative of the whole family of 160 sequences obtained from Genbank, SWISS-PROT, EMBL, and the genome databases. “. . .some members of the AQP family have a very high specificity for water. . . .” 188 Phylogenetic analysis showed subdivisions of the aquaporin family agreeing mostly with those given by Park and Saier (8) (Fig. 1, Table 1). Some subfamilies and groupings within subfamilies are overpopulated with highly similar members [such as the tonoplast intrinsic protein (TIP) and plasma membrane intrinsic protein (PIP) subfamilies], whereas others feature only single sequences (such as the single sequence for the archaea, AQParc). The complete absence of sequences from many other taxonomic groups further supports the anticipation of even more complexity in the family. Among the 160 sequences available, many appear to be minor variants from different species (such as AQP2) or different tissues (such as those in Arabidopsis thaliana), leading to overrepresentation of some sequence types. To compare the sequences without overrepresentation, only one sequence of a type was further used in alignment and analysis. A type is defined as a set of sequences with a phylogenetic divergence of <13%. This definition was designed to maintain the various mammalian isoforms in the AQP0 subfamily as distinct types, as given in Agre (1). These type sequences for the AQP family were classified into clusters and subfamilies (following Ref. 8). Subfamilies were assigned by visual inspection of the most consistent phylogenetic trees (the consensus shown in Fig. 1) and can be defined based on major taxonomic groupings (e.g., animal, plant, yeast, bacteria, archaea). The subfamilies show sequence divergences between 13 and 35%. The phylogenetic analysis revealed a suggestive dichotomy, with two clear clusters of subfamilies [Fig. 1; also evident in the work of Park and Saier (8)]. In addition, the archaea aquaporin (AQParc) is positioned as an intermediate sequence between these clusters, suggesting a very ancient divergence. The phylogenetic divergence for the two clusters is ∼43%, not much higher than for the subfamilies. We suggest that these clusters be called the AQP and glycerol facilitator-like protein (GLP) clusters (Fig. 1). It is well known from physiological data that some members of the AQP family have a very high specificity for water, whereas others also allow the passage of larger nonionic compounds such as glycerol and urea. It is therefore tempting to associate this apparent functional distinction with the phylogenetic clusters. That this picture is too simple is emphasized by the absence of plant sequences in the GLP cluster, whereas some proteins such as AQP0 also transport glycerol (7). The two clusters are therefore unlikely to represent strict functional distinctions. Also, whereas the fundamental design apparently remained unaltered through evolution (6 helices around 2 central loops, see below), functional diversity News Physiol. Sci. • Volume 14 • October 1999 Downloaded from www.physiology.org/journal/physiologyonline by ${individualUser.givenNames} ${individualUser.surname} (168.151.137.021) on November 6, 2018. Copyright © 1999 American Physiological Society. All rights reserved. Table 1. Classification of aquaporin family sequences based on phylogeny into 2 clusters, 16 subfamilies, and 46 types Subfamily AQP0 (Animal) AQP8 (Animal) TIP (Plant tonoplast intrinsic proteins) PIP (Plant plasma membrane intrinsic proteins) NIP (Plant nodulin 26-like intrinsic proteins) AQPD (Slime mold) AQPY (Yeast) AQPZ (Bacteria) AQPS (Synechococcus sp.) AQPA (Archaeoglobus fulgidus) GLPA (Animal) GLPY1 (Yeast) GLPY2 (Yeast) GLPB1 (Gram-negative bacteria) GLPB2 (Gram-positive bacteria) GLPB3 (Mycoplasma, gram positive) Synonyms† Type Type Sequence* AQP0 AQP1 AQP2 AQP4 AQP5 AQP6 AQPcic AQPbib AQPhin AQP8 AQPcel aTIP gTIP dTIP TIPpic PIP1 AQP cluster MIP_HUMAN AQP1_HUMAN AQP2_HUMAN AQP4_HUMAN AQP5_HUMAN AQP6_HUMAN g1279358 BIB_DROME g1262285 AQP8_RAT g1359543 TIPA_ARATH TIPG_ARATH g1145697 g3021538 WC1A_ARATH PIP2 NIPgly NIPnic NIPara NIPory AQPdic WC2A_ARATH NO26_SOYBN PMIP_NICAL E1249637 Q40746 g1562532 C32C4.2,F40F9.9 bTIP,MP23,MP28 TIP1,RB7,VM23,TIPR,TIPS,NOD26§ DIP,RB7,TIP7,TIP18,RT-TIP MIPr WC1,TIPW,PIPB,PMIP,RAMP, WSI-TIP, PTOM75, MIPA, MIPB, MIPD, PAMIP1, EMIP, PM28B WC2,AQUA,MIPC,MIPE,PM28A,PIP3 Nodulin-26 PMIP MIP,NLM1 MIP WacA AQY AQPL_YEAST YPR192W,AQY1,AQY2 AQPZ AQPscy AQPsco AQPZ_ECOLI P73809 Q55998 AQParc O28846 AQP3 AQP7 AQP9 GLPcel1 GLPcel2 GLPcel3 GLPcel4 GLPY1 GLP cluster AQP3_HUMAN AQP7_RAT g2887407 g669031 g1065485 Q21473 g521003 YFF4_YEAST F32A5.5 K02G10.7 M02F4.8 C01G6.1 YFF4,YFL054C GLPY2 FPS1_YEAST FPS1,YLL043W GLPFeco GLPFhae GLPFpse GLPFsal GLPFbac GLPFstr GLPFlac GLPFmga GLPFmge GLPFmpn GLPF_ECOLI GLPF_HAEIN GLPF_PSEAE PDUF_SALTY GLPF_BACSU GLPF_STRPN YDP1_LACLC GLPF_MYCGA GLPF_MYCGE GLPF_MYCPN MIP26 CHIP28, DER2, AQP-CHIP,‡ FA-CHIP WCH2, WCH-CD, AQP-CD‡ WCH4,MIWC‡ hKID,WCH3‡ Big brain protein GLIP,BLIP‡ AQPAP,AQP7L PDUF *Names for type sequences from SWISS-PROT and accession numbers from Genbank (g) or TREMBL (O, P and Q); synonyms include all sequences classified within the same type; ‡see Ref. 1; §three sequences can be obtained in the databases for nodulin-26. Two of these are highly similar to gTIP (Q39882 and Q39883 in TREMBL), while the other (NO26_SOYBN) is classified in the NIP subfamily. † News Physiol. Sci. • Volume 14 • October 1999 189 Downloaded from www.physiology.org/journal/physiologyonline by ${individualUser.givenNames} ${individualUser.surname} (168.151.137.021) on November 6, 2018. Copyright © 1999 American Physiological Society. All rights reserved. based on subtle changes may have developed several times and in different taxonomic groups. The AQP1 three-dimensional structure “A . . . path for water flow . . . is difficult to distinguish. . . .” 190 Hydropathy analysis of the majority of sequences in the AQP family suggests six transmembrane segments. Although a few sequences yield somewhat ambiguous hydropathy plots, the average hydropathy of the whole family, based on the same alignment used in the phylogeny analysis, definitively indicates six transmembrane segments (Fig. 2A) consistent with topology and mutagenic analysis, leading to the hourglass model (Fig. 2B; Ref. 6). Determination of the structure of AQP1 at 6Å resolution by electron crystallography combined with metal-shadowing electron microscopy and atomic force microscopy provided a clear picture of the architecture of the water channel (3, 11). The hourglass model (6) was confirmed, and examination of the threedimensional (3-D) map allowed an assignment of the six predicted helices, with implications for water permeation (5). The three 3-D maps of AQP1 all clearly demarcate the monomers in the tetrameter as sixhelix bundles surrounding density assigned to the functional loops B and E (3, 11). Also in agreement is the likely position of the water channel through the protein (Fig. 3), adjacent to the central density (loop E) and toward the fourfold axis of the tetramer (3, 5). The putative water channel starts on the extracellular side with a cavity between the central density and the helices around the fourfold axis with a width of ∼10-12 Å. Below this opening, within the cytoplasmic half of the monomer, the presence of a large amount of density suggests that this is the narrow part of the channel. A particular path for water flow through this region is difficult to distinguish, and multiple paths may be possible. It is interesting to compare the conserved residues within the helices and loops B and E with residues known to line water pockets in membrane proteins. Such an analysis on the photosynthetic reaction center, cytochrome-c oxidase, bacteriorhodopsin, and cytochrome f led to a speculative model of the channel-lining residues (5). The salient feature of this model is a string of highly conserved polar residues that can be placed adjacent to the functional loops and along the center of the six-helix bundle. The inherent sequence symmetry of the molecule is also reflected in the symmetrical arrangement of these residues with respect to the membrane plane. This assignment imposes constraints on the angular orientation of the helices and also locates these conserved residues away from the lipid environment. As this is in agreement with available mutagenesis data, it may help in producing a reasonable atomic model of a water channel protein. The physiology of water permeation Because the AQP family has only recently been discovered, physiological and functional analyses are quite limited. Of particular interest is the high flux combined with high selectivity for water found for the best-studied member, AQP1. Water channels are a physiological necessity to deal with osmotic pressure variations and cell volume changes. However, it is unclear how the passage of ions is prevented, especially protons, which are able to tunnel through the matrix of hydrogen bonds in water. Flux of water through AQPs Water channel proteins lower the activation energy of a membrane for water permeation from 10-20 kcal/mol to <5 kcal/mol. This allows a high flux of water through the channel when the membrane is subjected to hydrostatic or osmotic pressure. To measure the water flux through membrane channels, AQPs are commonly expressed in Xenopus oocytes, giving a typical permeability coefficient Pf = 0.02 cm/s compared to Pf = 0.001 cm/s without channels (6). This allows semiquantitative measurements, but the uncertainty in the level of expression of particular proteins makes interpretation difficult. Two-dimensional crystallization of AQP1 in the presence of lipids into closed vesicles offered a system in which the protein density in the membrane is known with high accuracy (12). In these crystalline vesicles of ∼3-µm diameter, the osmotic permeation coefficient for water flowing through the AQP1 channel was found to be Pf = 0.472 cm/s (12). For a crystalline unit cell of 9.6 nm x 9.6 nm containing eight AQP1 monomers (11.5 nm2/channel), this translates into a unit osmotic permeability coefficient of pf = 5.43 x 10-14 cm3·s-1·channel-1 = 1.8 x 109 water molecules·s-1·channel-1. An attractive and simple model for the selectivity of the aquaporins is a size-exclusion mechanism, originally proposed by Heller et al. (4) for GlpF from Escherichia coli. It was concluded that the AQP1 channel specificity is also based on size exclusion, with the channel showing decreasing permeability for formamide compared with water (13). This indicated that the channel has a narrow part with a diameter about the size of a water News Physiol. Sci. • Volume 14 • October 1999 Downloaded from www.physiology.org/journal/physiologyonline by ${individualUser.givenNames} ${individualUser.surname} (168.151.137.021) on November 6, 2018. Copyright © 1999 American Physiological Society. All rights reserved. FIGURE 2. A: average hydrophobicity of the 46 AQP family sequence types, showing only those residues represented in AQP1 (smoothed over a 19-residue window). B: the hourglass model: topology and functional loops in aquaporin [major integral protein (MIP)] family proteins derived from sequence analysis, mutagenesis, labeling, and functional analysis (6). molecule, ∼4 Å [cross-sectional area (A) ≈ 13 Å2]. However, the existence of glycerol-specific channels with a lower water permeability than the water channels (7) suggests a different selectivity mechanism, with the width of the channel constriction being only one consideration. The high selectivity of the AQP1 channel suggests that it functions according to the single-file model, in which a single chain of water molecules has to pass through the narrowest part of the pore. This allows the determination of the length (L) of this narrow part of the channel from the ratio of osmotic to diffusive flow of water (the file number). Analysis of the low permeability of formamide indicated that this constriction might be quite short, L ≈ 9–13 Å (13). The 6-Å 3-D structure of AQP1 shows a wide funnel on the extracellular side of ∼10- to 12-Å diameter, narrowing towards the middle of the membrane plane. The narrow part of the channel thus appears to lie on the cytoplasmic side and may vary from 10 to 20 Å in length. Given these measurements for the AQP1 channel, the apparent diffusion coefficient of water in the channel can be calculated as Dw ≈ 0.4 x 10-5 to 0.8 x 10-5 cm2/s (Dw = (L/A)pf). Compared to bulk water diffusion, Dw = 2.1 x 10-5 cm2/s, diffusion in the channel is therefore about three to five times lower. Of course, the implicit assumption is that there is only one course of water or solute flow. If there are several paths for water passage, the apparent diffusion coefficient within the channel must be lower. Other measurements have been done for AQP1, giving the unit permeability coefficient pf = 1–16 x 10-14 cm3·s-1·channel-1. The high variability in the determined permeabilities is likely caused by uncertainty about the number of channels per unit membrane area. The determination for the crystalline vesicles is taken as definitive, with several other measurements yielding comparable permeabilities. Some aquaporins have low water permeability, pf = 0.03–0.3 x 10-14 cm3·s-1·channel-1, such as that for AQP0, whereas AQP4 has been reported to have a pf = 15–24 x 10-14 cm3·s-1·channel-1 (14). Because AQP4 remains selective for water (14), the higher flux through AQP4 may be associated with multiple paths for water permeation to yield a diffusion coefficient significantly lower than in bulk water. Other aquaporins have permeabilities comparable to that of AQP1 (2–5 x 10-14 cm3·s-1·channel-1). “. . . it functions according to the single-file model. . . .” Selectivity for and flux of nonionic compounds The weight of the literature data on the AQP family is in favor of channels permeated by small nonionic molecules in the size range of water to glycerol. Although xylitol is the largest compound reported to pass through the E. coli GlpF, it was impermeant to pentoses (4). The linear xylitol molecule has a width similar to that of glycerol, whereas the sugars are ring structures with a larger cross section. The functional difference between the two clusters of the family shown in Fig. 1 is not clearcut. For instance, the AQP cluster proteins AQP0 and nodulin 26 (NIP)gly both transport glycerol, whereas AQP9 of the GLP cluster is impermeant to glycerol. The characterization of the AQP family physiology is therefore still shrouded in a cloud of questions. The permeability of glycerol and urea of GLP cluster proteins is three to four orders of magnitude lower than the water permeability, as shown, for example, for AQP0 (7). This suggests that in these channels a large flux of water accompanies solute permeation. News Physiol. Sci. • Volume 14 • October 1999 191 Downloaded from www.physiology.org/journal/physiologyonline by ${individualUser.givenNames} ${individualUser.surname} (168.151.137.021) on November 6, 2018. Copyright © 1999 American Physiological Society. All rights reserved. FIGURE 3. A proposed course of the channel through AQP1. The 6-Å 3-D map of the AQP1 monomer (11) is shown cut through the middle to expose the functional loops B and E (cytoplasmic side at bottom). The channel starts on the extracellular side as a wide opening, narrowing down toward the cytoplasmic side (~4 Å wide). Also shown are the assigned helices 1–4, with helices 5 and 6 completely cut away. Selectivity against ions “. . . the AQP1 channel has no intrinsic ion conductance. . . .” 192 As demonstrated in several laboratories, the AQP1 channel has no intrinsic ion conductance ability (2, 15). If the flow of water through the channel is accompanied by ions without selection, the conductivity in a 100 mM salt is expected to be (0.1 M/55 M) 3.3 x 109 water molecules·s-1·channel-1 = 6 x 106 ions·s-1·channel-1, which is comparable to the single-channel current observed for ion channels. AQP1 and many of the other AQP family members show a strong concentration of positive charge on the cytoplasmic side (especially loop D) and negative charge on the extracellular side. Perhaps these strongly charged entrances to the channel form highly efficient electrostatic filters preventing the passage of ions. Alternatively, the inability of the AQP1 channel to conduct ions may be explained by the size of hydrated ions. However, protons should be easily conducted along a water chain, as is the case for the gramicidin A channel (9). In addition, at a rate of water passage through AQP1 of ∼109 molecules·s-1·channel-1, only the protons swept along with the water should lead to a fast drop in the pH gradient across the membrane. However, a pH shift experiment did not show any proton conductance through AQP1, whereas the control, gramicidin A, dissipated the proton gradient efficiently (15). An examination of the mechanisms of proton conductance through the gramicidin A channel suggested a fast forward reaction associated with the propagation of an ionic defect and a slower backward reaction associated with a reorientation of the water chain through propagation of a bonding defect (9). The passage of protons is thus similar to a tunneling effect, with proton propagation much faster than the movement of the actual water molecules. This study also suggested that clusters of water bound to the protein slow down proton transfer, whereas a single-file hydrogen-bonded chain of water molecules conducts a proton more efficiently (especially when electrostatic interactions with the protein wall of the channel are switched off; Ref. 9). Hence the structure of the water at the charged channel entrances may prevent proton flow. Furthermore, the orientations of the water molecules required for the transmission of the defect along the water News Physiol. Sci. • Volume 14 • October 1999 Downloaded from www.physiology.org/journal/physiologyonline by ${individualUser.givenNames} ${individualUser.surname} (168.151.137.021) on November 6, 2018. Copyright © 1999 American Physiological Society. All rights reserved. chain may be restricted or disallowed within the channel. Thus the specific interactions and arrangement of the water in the pore may provide a barrier to proton conduction. Another mechanism might rely on the requirement of the transmission of a defect along the water chain to reorient the water molecules correctly for proton translocation. A gap in the water chain too large to bridge by significant tunneling of a proton might prevent proton conductance. Conclusion That the AQP family is an important group of proteins is emphasized by its abundance and its involvement in diseases. The emergence of a large amount of sequence data on this family, together with progress on the 3-D structure of some of its members and physiological studies, provides a fascinating picture of how such a channel is constructed and how it functions. The diversity of isoforms and functional variation in many organisms indicates multiple roles in cellular and tissue functioning. Here an attempt is made to classify the available sequences into a consistent scheme, taking into account the similarities on different levels, namely, clusters, subfamilies, and types. Sequence analysis and the available 3-D structures suggest one architecture for the family, that of a tetramer with each monomer a six-helix bundle surrounding two inward-pointing loops. The channel is thought to be located within this bundle next to the loops, with a narrow region towards the cytoplasmic side. The available evidence indicates that all these channels allow permeation of the very small water molecule. Some of the channels transport glycerol as well, but three to four orders of magnitude more slowly than water. Although some suggestions regarding the exclusion of ions have been made, the real explanation for ion nonconductance must still be found. Determination of an atomic model and further functional studies are essential to understand how these channels work. We wish to thank all the people in the aquaporin field contributing valuable suggestions, in particular Peter Agre, Christophe Maurel, Peter Deen, Frédérique Tacnet, and Soren Nielsen. This work was supported by the Swiss National Foundation for Scientific Research (grant 31-42435.94 to A. Engel) and the Maurice E. Müller Foundation of Switzerland. References 1. Agre, P. Molecular physiology of water transport: aquaporin nomenclature workshop. Mammalian aquaporins. Biol. Cell 89: 255–257, 1997. 2. Agre, P., M. D. Lee, S. Devidas, and W. B. Guggino. Aquaporins and ion conductance. Science 275: 1490, 1997. 3. Cheng, A., A. N. van Hoek, M. Yeager, A. S. Verkman, and A. K. Mitra. Three-dimensional organization of a human water channel. Nature 387: 627–630, 1997. 4. Heller, K. B., E. C. C. Lin, and T. H. Wilson. Substrate specificity and transport properties of the glycerol facilitator of Escherichia coli. J. Bacteriol. 144: 274–278, 1980. 5. Heymann, J. B., P. Agre, and A. Engel. Progress on the structure and function of aquaporin 1. J. Struct. Biol. 121: 191–206, 1998. 6. Jung, J., G. Preston, B. Smith, W. Guggino, and P. Agre. Molecular structure of the water channel through aquaporin CHIP. The hourglass model. J. Biol. Chem. 269: 14648–14654, 1994. 7. Kushmerick, C., K. Varadaraj, and R. Mathias. Effects of lens major intrinsic protein on glycerol permeability and metabolism. J. Membr. Biol. 161: 9–19, 1998. 8. Park, J. H., and M. H. Saier. Phylogenetic characterization of the MIP family of transmembrane channel proteins. J. Membr. Biol. 153: 171–180, 1996. 9. Pomès, R., and B. Roux. Structure and dynamics of a proton wire: a theoretical study of H+ translocation along the single-file water chain in the gramicidin A channel. Biophys. J. 71: 19–39, 1996. 10. Preston, G. M., T. P. Carroll, W. B. Guggino, and P. Agre. Appearance of water channels in Xenopus oocytes expressing red cell CHIP28 protein. Science 256: 385–387, 1992. 11. Walz, T., T. Hirai, K. Murata, J. B. Heymann, K. Mitsuoka, Y. Fujiyoshi, B. L. Smith, P. Agre, and A. Engel. The 6Å three-dimensional structure of aquaporin-1. Nature 387: 624–627, 1997. 12. Walz, T., B. Smith, M. Zeidel, A. Engel, and P. Agre. Biologically active two-dimensional crystals of aquaporin CHIP. J. Biol. Chem. 269: 1583–1586, 1994. 13. Whittembury, G., E. Gonzalez, A. Gutierrez, M. Echevarria, and C. Hernandez. Length of the selectivity filter of aquaporin-1. Biol. Cell 89: 299–306, 1997. 14. Yang, B., and A. S. Verkman. Water and glycerol permeabilities of aquaporins 1-5 and MIP determined quantitatively by expression of epitope-tagged constructs in Xenopus oocytes. J. Biol. Chem. 272: 16140–16146, 1997. 15. Zeidel, M., S. Nielsen, B. Smith, S. Ambudkar, A. Maunsbach, and P. Agre. Ultrastructure, pharmacologic inhibition, and transport selectivity of aquaporin channel-forming integral protein in proteoliposomes. Biochemistry 33: 1606–1615, 1994. News Physiol. Sci. • Volume 14 • October 1999 “Some of the channels transport glycerol as well. . . .” 193 Downloaded from www.physiology.org/journal/physiologyonline by ${individualUser.givenNames} ${individualUser.surname} (168.151.137.021) on November 6, 2018. Copyright © 1999 American Physiological Society. All rights reserved.