Purification and functional characterization of aquaporin-8


  • Kun Liu,

    1. Department of Biological Chemistry, School of Medicine, Johns Hopkins University, Baltimore, MD 21205, U.S.A.
    Search for more papers by this author
  • Hiroaki Nagase,

    1. Department of Biological Chemistry, School of Medicine, Johns Hopkins University, Baltimore, MD 21205, U.S.A.
    Search for more papers by this author
  • Chunyi George Huang,

    1. Department of Biological Chemistry, School of Medicine, Johns Hopkins University, Baltimore, MD 21205, U.S.A.
    Search for more papers by this author
  • Giuseppe Calamita,

    1. Department of General and Environmental Physiology, University of Bari, via Amendola 165/A, I-70126 Bari, Italy
    Search for more papers by this author
  • Peter Agre

    Corresponding author
    1. Department of Biological Chemistry, School of Medicine, Johns Hopkins University, Baltimore, MD 21205, U.S.A.
    2. Department of Medicine, School of Medicine, Johns Hopkins University, Baltimore, MD 21205, U.S.A.
      Department of Biological Chemistry, School of Medicine, Johns Hopkins University (email pagre@jhmi.edu).
    Search for more papers by this author

Department of Biological Chemistry, School of Medicine, Johns Hopkins University (email pagre@jhmi.edu).


Background information. Aquaporins (AQPs) are a family of channels permeable to water and some small solutes. In mammals, 13 members (AQP0–AQP12) have been found. AQP8 is widely distributed in many tissues and organs. Previous studies in frog oocytes suggested that AQP8 was permeable to water, urea and ammonium, but no direct characterization had yet been reported.

Results. We expressed recombinant rAQP8, hAQP8 and mAQP8 (rat, human and mouse AQP8 respectively) in yeast, purified the proteins to homogeneity and reconstituted them into proteoliposomes. Although showing high sequence similarity, AQP8 proteins from the three species had to be purified with different detergents prior to reconstitution. In stopped-flow studies, all three AQP8 proteoliposomes showed water permeability, which was inhibited by mercuric chloride and rescued by 2-mercaptoethanol. rAQP8 and hAQP8 proteoliposomes did not transport glycerol or urea but were permeable to formamide, which was also inhibited by mercuric chloride. In the oocyte transport assay, hAQP8-injected oocytes showed significantly higher [14C]methylammonium uptake than water-injected oocytes.

Conclusions. In the present study, we successfully purified rAQP8, hAQP8 and mAQP8 proteins and characterized their biochemical and biophysical properties. All three AQP8 proteins transport water. rAQP8 and hAQP8 are not permeable to urea or glycerol. Moreover, hAQP8 is permeable to ammonium analogues (formamide and methylammonium). Our results suggest that AQP8 may transport ammonium in vivo and physiologically contribute to the acid—base equilibrium.

Abbreviations used:



mAQP, rAQP, human, mouse and rat AQP respectively


n-dodecyl β-D-maltoside


n-decyl β-D-maltoside


deoxycholic acid


horseradish peroxidase


modified Barth's solution






N-lauroylsarcosine sodium salt


Nonidet P40


n-octyl β-D-glucoside


Triton X-100


Triton X-114


Aquaporins (AQPs) are a family of membrane channel proteins that are selectively permeated by water (classical AQPs) or water plus small uncharged solutes such as glycerol and urea (aquaglyceroporins). AQPs facilitate membrane transport either constitutively (e.g. AQP1) or under regulation (e.g. AQP2 by vasopressin). In either case, the movement of water is driven by an osmotic gradient and leads to rapid equilibrium across the membrane. The functional distinction between water-transporting members of the family (such as AQP0, AQP1, AQP2, AQP4 and AQP5) and glycerol-transporting members (AQP3, AQP7, AQP9 etc.) is correlated with specific differences in the primary amino acid sequences supporting the existence of two subfamilies.

AQP8 cDNAs have been cloned from rat (Ishibashi et al., 1997; Koyama et al., 1997), mouse (Ma et al., 1997) and human (Koyama et al., 1998). AQP8 protein sequences from rat and mouse have a similarity of 90%, but the similarity between hAQP8 (human AQP8) and rAQP8 (rat AQP8) or between hAQP8 and mAQP8 (mouse AQP8) is slightly lower (75%). Sequence alignment analyses indicate AQP8 as a classical AQP (Ishibashi et al., 1997; Koyama et al., 1997; Ma et al., 1997). However, genomic and evolutionary constraints make AQP8 significantly divergent when compared with the classical mammalian AQPs (Calamita et al., 1999; Zardoya, 2005). Figure 1(A) shows the phylogenetic tree of mammalian AQPs where AQP8 appears close to the most divergent group of paralogues, AQP11 and AQP12.

Figure 1.

Sequence alignment, phylogenetic analysis and modelled structure of AQP8

(A) Phylogenetic tree of mammalian AQPs. The three AQP8 proteins are in red. The scale bar represents genetic distance between homologues, representing 0.1 amino acid substitutions per site. (B) Protein sequence alignment between hAQP8 and TaTIP2;1. Residues in red are those identical between the two proteins. Asterisks mark key residues lining the putative pore of hAQP8, which are also shown in (C). Dashes represent gaps between the two sequences. (C) Modelled pore region of hAQP8 based on bovine AQP0 atomic structure. Key residues in the pore region (His72, Ile198, Gly207, Cys208 and Arg213) are shown in ball-and-stick representation, and the other residues are shown as cartooned helices.

AQP8 has been detected in many tissues and organs, such as salivary gland, liver, pancreas, small intestine, colon, testis, heart and kidney (Ishibashi et al., 1997; Koyama et al., 1997, 1998; Ma et al., 1997). rAQP8 has been detected at the apical membrane of a salivary epithelial cell line and contributes to the water movement in response to an osmotic gradient (Hoque et al., 2002). In cultured rat hepatocytes, cAMP caused AQP8 to traffic from intracellular vesicles to the plasma membrane; this redistribution was accompanied by an increase in water permeability of the membrane (Garcia et al., 2001). Recently, AQP8 has been found on the mitochondrial inner membrane of rat liver and may facilitate rapid mitochondrial expansions during oxidative phosphorylation and apoptosis (Calamita et al., 2005). In mouse liver, AQP8 may preserve cytoplasmic osmolality during glycogen metabolism to maintain the mitochondrial volume (Ferri et al., 2003). Mice lacking AQP8 have an increase in testicular weight and a mild hypertriglyceridaemia after 21 days on a 50% fat diet (Yang et al., 2005).

Functional evaluations of AQP8 are inconsistent. When ectopically expressed in Xenopus oocytes, rAQP8, hAQP8 and mAQP8 were shown to transport water but not glycerol (Ishibashi et al., 1997; Koyama et al., 1997, 1998; Ma et al., 1997). rAQP8- and hAQP8-injected oocytes did not transport [14C]-urea in uptake assays (Koyama et al., 1997, 1998), whereas mAQP8-injected oocytes showed significantly higher [14C]urea uptake than water-injected oocytes (Ma et al., 1997). A yeast growth complementation study suggested that hAQP8 might transport ammonium (Jahn et al., 2004). A similar function for hAQP8 has been indicated by functional studies with Xenopus oocytes (Holm et al., 2005). The sequence similarity between hAQP8 and the wheat AQP TaTIP2;1, which is able to transport ammonium, is 32% and, their putative pores share most of the key residues (Figures 1B and 1C).

In the present study, AQP8 proteins from rat, human and mouse were expressed in yeast, purified to homogeneity and functionally reconstituted into proteoliposomes. Optical scattering studies using a stopped-flow apparatus showed that all three have water permeability. rAQP8 and hAQP8 did not transport urea or glycerol. Intriguingly, rAQP8 and hAQP8 proteoliposomes were permeable to formamide and hAQP8-injected oocytes had a significantly higher [14C]methylammonium uptake than water-injected oocytes. Formamide and methylammonium are ammonium analogues (Jahn et al., 2004), so our results suggest that AQP8 may physiologically transport ammonium.

Results and discussion

Expression and purification of rAQP8, hAQP8 and mAQP8 from yeast

His10-tagged rAQP8, hAQP8 or mAQP8 was expressed in pep4Δ Saccharomyces cerevisiae. After induction with 2% (w/v) galactose and fractionation by ultracentrifugation, the yeast lysate was separated, and AQP8 protein was shown by immunoblotting to be enriched in the membrane fraction (results not shown).

Protein solubility was screened with several detergents: OG (n-octyl β-D-glucoside), DM (n-decyl β-D-maltoside), DDM (n-dodecyl β-D-maltoside), TX100 (Triton X-100), TX114 (Triton X-114), NL (N-lauroylsarcosine sodium salt), NP40 (Nonidet P40), DOC (deoxycholic acid) or CHAPS. Although AQP8 protein sequences from the three species have high similarity, their detergent solubilities were very different. Comparing the band intensity of immunoblotting demonstrated that rAQP8 had the best solubility with TX100. In terms of hAQP8, OG, DDM, TX114 and NL had a similar effect. For mAQP8 solubilization, NL showed the best ability and TX114 was the second best (Figure 2A). According to the sequence alignment, different amino acid residues of three AQP8 homologues are spread throughout the whole proteins. Therefore it seems that all the domains of AQP8 are responsible for the distinct solubilities.

Figure 2.

Purification of AQP8 from the yeast expression system

(A) Solubilities of rAQP8, hAQP8 and mAQP8 (top, middle and bottom respectively). Detergents are (1) OG, (2) DM, (3) DDM, (4) TX100, (5) TX114, (6) NL, (7) NP40, (8) DOC, (9) CHAPS and (10) SDS. Solubility profiles employed for purifications are boxed: for rAQP8, TX100; for hAQP8, OG; and for mAQP8, NL. (B) Coomassie Blue-stained SDS gels and immunoblots of purified rAQP8, hAQP8 and mAQP8 (top, middle and bottom panels respectively). Arrows indicate the monomer bands of AQP8. The bands with higher molecular mass shown on both the Coomassie Blue gels and immunoblots are thought to be AQP8 oligomers.

OG can be easily dialysed for future functional analysis, so it is a favoured detergent in solubilization of many overexpressed AQPs. In the present study, the purification of hAQP8 with OG was successful; however, rAQP8 and mAQP8 were solubilized poorly by OG. TX100 was used to extract rAQP8 from the yeast membrane and was replaced by OG during the affinity chromatography. In mAQP8 purification, the non-ionic detergent TX114 was tried first, but the His-tagged protein was absorbed poorly by the nickel column (results not shown). Therefore NL was finally employed for mAQP8 purification. The purities of AQP8 proteins were checked on Coomassie Blue-stained SDS gels and the bands were confirmed by immunoblotting (Figure 2B). More than approx. 90% of the purified protein was AQP8. There were two faint bands bigger than 55.4 kDa on the hAQP8 Coomassie Blue gel; they were assumed to be impurities since they were not probed in immunoblotting (Figure 2B, middle panel).

Functional reconstitution of AQP8 proteins

Purified AQP8 proteins were then reconstituted into proteoliopsomes. The presence of AQP8 was confirmed by silver staining (Figure 3, insets) and immunoblotting (results not shown). Light-scattering stopped-flow experiments showed that all three AQP8s were permeable to water (Figure 3). The rate constants of AQP8 proteoliposomes were at least 10 times higher than that of control liposomes prepared identically but lacking AQP8 proteins. The water permeability was inhibited by 1 mM HgCl2 and partially restored with 5 mM 2-ME (2-mercaptoethanol), indicating that the water transport was due to AQP8 (Table 1). These properties of AQP8 are similar to those of other AQPs. The HgCl2-sensitive residue of AQP1 is Cys189 (Preston et al., 1993). By sequence alignment, the putative mercury-sensitive residue of AQP8 is Cys209 (in rAQP8), Cys208 (in hAQP8) and Cys207 (in mAQP8). Consistent with our results, it is reasonable that AQP8 has been found in kidney, liver, pancreas and salivary gland where there is active fluid transport (Yang et al., 2005). The Arrhenius activation energy (Ea) of rAQP8 proteoliposomes (4.33 kcal/mol; 1 cal∼4.184 J) is well below that of the control liposomes (13.05 kcal/mol), consistent with AQP8 acting as an open channel.

Figure 3.

Water permeability of AQP8

Purified rAQP8, hAQP8 and mAQP8 were reconstituted into proteoliposomes and they were all permeable to water in the stopped-flow assay. Black tracings represent light scatter of AQP8 proteoliposomes. Grey tracings represent light scatter of control liposomes. Silver-stained SDS gels of rAQP8, hAQP8 and mAQP8 proteoliposomes are shown in the insets, indicating the presence of AQP8 proteins. Arrows indicate the monomer bands of AQP8.

Table 1.  HgCl2 inhibition and 2-ME rescue of AQP8 water permeability
Rate constant (k) is expressed as mean±S.D.
 k (s−1)
   Control liposomes3.91±0.03
   rAQP8+1 mM HgCl24.85±0.032
   rAQP8+1 mM HgCl2+5 mM 2-ME75.09±1.38
   Control liposomes4.90±0.0056
   hAQP8+1 mM HgCl24.85±0.037
   hAQP8+1 mM HgCl2+5 mM 2-ME23.97±1.12
   Control liposomes4.83±0.0084
   mAQP8+1 mM HgCl25.32±0.12
   mAQP8+1 mM HgCl2+5 mM 2-ME91.21±12.85

In the stopped-flow assay on AQP8 water permeability, sorbitol was first used as the impermeant solute. Later, it turned out that sucrose and mannitol were also impermeant solutes (results not shown). Purification of mAQP8 required anionic NL. Sometimes ionic detergents may disturb the correct protein folding, but our study revealed that NL could still be applied for protein purification under certain conditions. The diameters of control liposomes and AQP proteoliposomes were similar under electron microscopy, and different detergents (OG or NL) did not affect the sizes of liposomes. As a result, the rate constants were reliable to compare the permeabilities between controls and AQP proteoliposomes. AQP8 has been widely studied in many systems, including oocytes, cultured cells and knockout mice. Our study using proteoliposomes supplied clear evidence on AQP8 permeabilities, directly confirming that AQP8 is a water-selective channel. This work has provided new technical insights into permeability measurements of AQP8 and, maybe in the future, other AQPs.

Solute permeabilities of AQP8

Glycerol and urea are small non-electrical solutes transported by aquaglyceroporins. AQP8 is expressed in liver and kidney where glycerol and urea fluxes are known to be essential for metabolism; however, it is unresolved whether AQP8 contributes to glycerol or urea transport. Light-scattering stopped-flow assays failed to detect glycerol or urea permeability of rAQP8 and hAQP8 proteoliposomes. The rate constants of rAQP8 and hAQP8 proteoliposomes were quite similar to those of the controls (Table 2).

Table 2.  Solute permeabilities of rAQP8 and hAQP8
 Control liposomes

We found that control liposomes prepared with OG or NL showed different solute permeabilities. This finding introduced some complexity into interpreting the results. OG was used in purification of rAQP8 and hAQP8, whereas NL was necessary for mAQP8 purification. Thus, OG was used for the control liposomes of rAQP8 and hAQP8 proteoliposomes, while NL was used for the control liposomes of mAQP8 proteoliposomes. In the transport assays for glycerol and urea, the control liposomes prepared with NL had a lower rate constant than those prepared with OG, although the lipid was from the same batch and the liposome diameters were similar. Furthermore, it took five times longer for the control liposomes prepared with NL to reach the plateau (results not shown). We suspected that NL altered lipid packing, making the liposomes less permeable to solutes. Thus measurements on glycerol and urea permeability of mAQP8 proteoliposomes were not considered reliable. mAQP8 has been reported to transport isotope-labelled urea when ectopically expressed in oocytes (Ma et al., 1997) — unlike rAQP8 and hAQP8 (Koyama et al., 1997, 1998). In our light-scattering assays, rAQP8 and hAQP8 were confirmed to lack urea and glycerol permeability. However, due to the detergent issue mentioned above, urea and glycerol permeabilities of mAQP8 were not conclusive.

Formamide permeability of rAQP8 and hAQP8 proteoliposomes

hAQP8 is able to rescue the growth of yeast defective in ammonium uptake (Jahn et al., 2004). Since multiple transport proteins reside on the yeast membrane, growth complementation could be either a direct or an indirect effect of AQP8 expression. Formamide (HCONH2), a small uncharged solute, has been used as an ammonium analogue (Jahn et al., 2004) and is suitable for our light-scattering system. In the present study, both rAQP8 and hAQP8 proteoliposomes were permeable to formamide. rAQP8 proteoliposomes had a rate constant 4.4 times that of control liposomes and hAQP8 proteoliposomes had a rate constant 2.6 times that of control liposomes (Figure 4). After the treatment with 1 mM HgCl2, the rate constants of rAQP8 and hAQP8 proteoliposomes were 3.2±0.02 and 3.5±0.03 s−1 respectively, slightly below that of control liposomes (4.0±0.1 s−1). The fact that HgCl2 entirely blocks the formamide transport indicates that the permeation occurred through AQP8 instead of the lipid bilayer. To check whether the formamide permeability was specific for AQP8, rAQP4 was also tested in the assay. The rate constant of rAQP4 proteoliposomes was minimal (2.5±0.03 s−1), only 1.2-fold that of control liposomes (2.1±0.04 s−1). Therefore formamide transport seems specific to AQP8 and is not a common activity of AQPs.

Figure 4.

Formamide permeability of AQP8

Both rAQP8 (A) and hAQP8 (B) proteoliposomes were permeable to formamide in the stopped-flow assay. Black tracings represent light scatter of AQP8 proteoliposomes and grey tracings represent light scatter of control liposomes. Rate constants (k) are shown besides the curves as means±S.D.

Acetamide (CH3CONH2) is another uncharged solute, structurally similar to formamide, but with one more methyl group. In our system, minimal acetamide permeation was detected. The rate constant of hAQP8 proteoliposomes (4.2±0.05 s−1) was only 1.2-fold that of the control liposomes (3.4±0.08 s−1), suggesting that acetamide is not a good ammonium analogue for the stopped-flow assay.

hAQP8-mediated [14C]methylammonium uptake in oocytes

To verify further the ammonium permeability of AQP8, radioactively labelled [14C]methylammonium (C14H3NH2), another ammonium analogue (Jahn et al., 2004), was used in the oocyte uptake assay. Oocytes injected with hAQP8 cRNA exhibited a relatively small but significant increase in [14C]methylammonium accumulation (5.7±0.8 pmol/oocyte), 1.85-fold the accumulation by water-injected oocytes (3.1±0.3 pmol/oocyte). The uptake was entirely inhibited by co-incubation with 5 mM unlabelled ammonium chloride, showing that the uptake was ammonium-specific (Figure 5).

Figure 5.

[14C]Methylammonium uptake by hAQP8-injected oocytes

Xenopus oocytes injected with 50 nl of water or 15 ng of hAQP8 cRNA were cultured for 3 days and used in [14C]-methylammonium uptake assay. After a 15 min incubation, hAQP8-injected oocytes showed significant methylammonium uptake, which was 1.85-fold that by water-injected oocytes. The uptake was entirely blocked by co-incubation with 5 mM ammonium chloride. ***P<0.001.

Physiological significance of formamide and methylammonium transport by AQP8

The formamide and methylammonium permeabilities of AQP8 suggest that AQP8 may transport ammonium in vivo, which is consistent with recent Xenopus oocyte studies (Holm et al., 2005) as well as the observation that expression of hAQP8 is able to enhance the yeast growth under ammonium-limiting conditions (Jahn et al., 2004). In kidney, ammonium efflux is essential for the generation of bicarbonate, and thus is critical for the urine pH adjustment and the acid—base equilibrium of body fluid (Nakhoul and Hamm, 2004). RhBG and RhCG, the non-erythyroid members of the Rh family, also transport ammonium in the collecting duct (Quentin et al., 2003), so it seems that redundant pathways exist. The transport mechanism is not yet clear. Further efforts are needed to determine whether the uncharged gas form (NH3) or the cation form (NH4+) is transported by AQP8.

Materials and methods


Microbial growth medium components were obtained from Q-biogene (Carlsbad, CA, U.S.A.) and Difco (Detroit, MI, U.S.A.). Restriction enzymes were purchased from New England Biolabs (Beverly, MA, U.S.A.). Penta-His—HRP-conjugate antibody (where HRP stands for horseradish peroxidase) and Ni-NTA—agarose (where Ni-NTA stands for Ni2+-nitrilotriacetate) were obtained from Qiagen (Valencia, CA, U.S.A.). Rabbit anti-rAQP8 antibody was from Alpha Diagnostic (San Antonio, TX, U.S.A.). OG was from Anatrace (Maumee, OH, U.S.A.). DM and DDM were from Calbiochem (La Jolla, CA, U.S.A.). Complete EDTA-free protease inhibitor cocktail was purchased from Roche (Indianapolis, IN, U.S.A.). Escherichia coli polar lipids were purchased from Avanti Polar Lipids (Alabaster, AL, U.S.A.). [14C]Methylammonium chloride was from American Radiolabeled Chemicals (St. Louis, MO, U.S.A.). Bio-Safe II scintillation-counting buffer was from Research Products (Mount Prospect, IL, U.S.A.). Other reagents were from Sigma, Fisher Scientific (Newark, DE, U.S.A.) or J.T. Baker (Phillipsburg, NJ, U.S.A.). HRP-conjugated donkey anti-rabbit IgG and ECL® (enhanced chemiluminescence) Western blotting detection system were purchased from Amersham Biosciences.

Sequence alignment and structural modelling

AQP protein sequences were obtained from NCBI (http:www.ncbi.nlm.nih.gov) as reference sequences. Multiple sequence alignment was performed by ClustalW 1.82 with the matrix Gonnet 250. The alignment was then put into TreeView 1.6.6 to generate the phylogenetic tree. The modelled structure of the AQP8 pore was based on the structure of bovine AQP0 (Harries et al., 2004), which has the highest sequence similarity to AQP8 among the AQPs with resolved structures. Modelling was performed by the program JACKAL (http:trantor.bioc.columbia.eduprogramsjackal). Figure 1(C) was prepared by PyMol (http:www.pymol.org).

Strains, plasmids and media

The yeast strain was the protease-deficient S. cerevisiae (pep4Δ), which had lowered proteolytic activity for heterologous proteins. Expression plasmids pYES2-10His-rAQP8, pYES2-10His-hAQP8 and pYES2-10His-mAQP8 contained full-length cDNA of rAQP8, hAQP8 and mAQP8, whose expressions were controlled by GAL4 promoter. A His10 tag was fused at the N-terminus of AQP8. AQP8 cDNAs were amplified by PCR primers with EcoRI and XbaI sites at the 5′- and 3′-ends respectively from template plasmids pTA8, pHA8 and pMA8, which were constructed by inserting rAQP8, hAQP8 and mAQP8 in pCR2.1-topo vector (Invitrogen, San Diego, CA, U.S.A.). AQP8 fragments and the high-copy-number vector pYES2-10His-hAQP1 (Saparow et al., 2001) were digested with EcoRI and XbaI and then hAQP1 was replaced by AQP8 fragments. AQP4 expression plasmid pYES2-10His-rAQP4 was constructed in a previous study (Kozono et al., 2003). The selecting medium was Ura (27 g of dropout base and 0.77 g of complete supplement mixture minus uracil/litre) and the inducing medium was YP-Gal (20 g of peptone and 10 g of yeast extract/litre, with 2% galactose) (Kozono et al., 2003). The Xenopus expression plasmid pXβG-ev1-hAQP8 was constructed by ligating hAQP8 cDNA with pXβG-ev1 vector (Preston et al., 1992).

Expression of AQP8 in yeast, and membrane fraction extraction

Yeasts carrying His-tagged AQP8 were first propagated in Ura medium at 30 °C to an A600 of approx. 1.0 and then induced in YP-Gal medium at 30 °C for approx. 20 h until A600 was approx. 5. Yeasts (4 litres) were harvested by centrifugation at 6000 g and resuspended in 150 ml of lysis buffer (100 mM K2HPO4 and 1× protease inhibitor cocktail). Three French press cycles (1.4×108 Pa) at 4 °C were used to break the yeast cell wall. Unbroken cells and debris were separated from the cell lysate by a 10 min centrifugation at 6000 g and discarded. The membrane fraction was recovered from the supernatant by a 60 min centrifugation at 200000 g.

Detergent screening

The membrane fraction was divided into aliquots, mixed with different detergents at a final concentration of 2% (w/v) on ice for 30 min, and then centrifuged at 200000 g. The solubilized fraction (supernatant) was used for SDS/PAGE. Another aliquot was solubilized by 2% (w/v) SDS as the positive control for the screening. The solubilities of detergents were compared by immunoblotting with anti-AQP8 or anti-His antibody.

Purification of His-tagged AQP proteins

The membrane fraction was resuspended with 2–3% detergent (TX-100 for rAQP8, OG for hAQP8 and NL for mAQP8) in buffer A [100 mM K2HPO4, 10% (v/v) glycerol, 5 mM 2-ME, 200 mM NaCl and 1× protease inhibitor cocktail, pH 7.5] and incubated on ice for 1 h. Insoluble material was removed by a 45 min centrifugation at 200000 g. The soluble fraction was mixed with 0.6–1 ml of pre-equilibrated Ni-NTA beads and incubated with gentle agitation at 4 °C overnight. The beads were then packed in a glass/plastic Econo column (Bio-Rad Laboratories, Hercules, CA, U.S.A.) and washed with 100 bed volumes of less detergent [1.2% (w/v) OG for rAQP8 and hAQP8, and 2% NL for mAQP8] and 20 mM imidazole in buffer A to remove non-specifically bound materials. Ni-NTA—agarose-bound material was eluted with 0.2 ml of 750 mM imidazole in buffer A. Purification of rAQP4 was described previously (Kozono et al., 2003).

In vitro reconstitution of AQP8 proteoliposomes

The method was described previously (Kozono et al., 2003). In brief, the protein/lipid ratio (w/w) was 1:50 (AQP8 proteoliposomes) or 1:100 (AQP4 proteoliposomes). During the reconstitution, 1.2% OG was used for all the AQP samples. Dialysis was performed at least five times to remove the detergents. Diameters of liposomes (as means±S.D.) were determined by electron microscopy (Zeidel et al., 1992). Diameters of rAQP8, hAQP8 proteoliposomes and the control liposomes were 118.6±17.8, 116.1±15.7 and 115.1±15.6 nm respectively. Diameters of mAQP8 proteoliposomes and the control liposomes were 103.4±13.0 and 106.0±14.3 nm respectively. AQP4 proteoliposomes and the control liposomes were prepared with another batch of lipid, and their diameters were 165.2±33.2 and 163.3±24.3 nm respectively.

Permeability measurements and the determination of Arrhenius activation energy (Ea)

Water permeabilities of liposomes were measured by detecting the light scattering of the preparations in a stopped-flow apparatus with a dead time of ≤1 ms (SF-2003; KinTek Instruments, University Park, PA, U.S.A.). Then, 20 μl of AQP proteoliposomes or control liposomes was rapidly mixed with the same volume of hyperosmolar solution (570 mosmol of sorbitol, sucrose or mannitol) at 4 °C for 1 s. Because sorbitol, sucrose and mannitol are impermeant for proteoliposomes, the osmotic gradient (285 mosmol) drives the water efflux, and the consequent reduction in liposome volume is measured as an increase in the intensity of scattered light at 600 nm. The rate constant of the normalized light intensity increase indicates the rate constant of water efflux, which is proportional to the water permeability coefficient (Kozono et al., 2003; Calamita et al., 2005). The following equation describes the light intensity increase as a function of k with time:


where Y is the normalized light intensity (0–1) and t is time (s); k is the rate constant (s−1), calculated by fitting the Yt curve by non-linear regression. One-exponential function is used for control liposomes and two-exponential function is used for AQP8 proteoliposomes. A1 and C are constants. Glycerol, urea, formamide and acetamide permeabilities were measured using a previous method (Kozono et al., 2003). The curves were first fit with the one-exponential model and rate constants were compared between AQP proteoliposomes and control liposomes. If the fitting was poor, the two-exponential model was then used and the secondary rate constants were compared. Control experiments showed that there was no change in light intensity in the absence of osmotic gradient.

Ea of rAQP8 proteoliposomes or control liposomes was determined by measuring the water transport at various temperatures in the range 4–25 °C. The method was described previously (Borgnia et al., 1999). Inhibition by HgCl2 or rescue by 2-ME was performed by incubating the liposomes with 1 mM HgCl2 or 5 mM 2-ME on ice for at least 15 min before the stopped-flow assay. The osmolalities of samples were measured by a vapour pressure osmometer (Vapro osmometer; Wescor, Logan, UT, U.S.A.). Each stopped-flow experiment was independently performed at least twice, and at least six individual traces were used to calculate the mean±S.D. of the rate constant (k).

hAQP8 expression in oocytes and [14C]methylammonium uptake

Capped cRNA of full-length hAQP8 was synthesized in vitro from XbaI-linearized expression plasmid pXβG-ev1-hAQP8 by using T3 RNA polymerase and purified with the RNeasy Mini kit from Qiagen. Defolliculated Xenopus laevis oocytes (stages V and VI) were injected with 50 nl of water or 15 ng of cRNA, and cultured in MBS (modified Barth's solution) at 16 °C. [14C]-Methylammonium uptake was measured 3 days post-injection. Experiments were performed at room temperature (23 °C) by placing groups of five to ten oocytes in 2 ml of MBS containing 1 μCi/ml [14C]methylammonium chloride and 20 μM unlabelled methylammonium chloride with or without 5 mM ammonium chloride as the competitor. Oocytes were incubated for 15 min, washed four times in 3 ml of ice-cold MBS with 20 μM unlabelled methylammonium chloride and solubilized in 200 μl of SDS (10%). Bio-Safe II buffer (5 ml) was added to the lysate and the intracellularly accumulated radioactivity was analysed by a liquid-scintillation counter.


We thank Shahram Khademi for discussion and Sally Craig and Trish Ward for editorial assistance. This work was funded by NIH (National Institutes of Health) grants (HL33991, HL48268 and EY11239).