Permeability of AQP9 in Oocytes. X. laevis oocytes injected with cRNAs encoding rat AQP9, AQP4, or both were compared with oocytes injected with 50 nl of water. Coefficients of osmotic water permeability (Pf) were determined from rates of swelling after transfer from 200 to 70 mosM Barth's solution. Water-injected oocytes exhibited low water permeability, whereas the Pf of AQP9 oocytes was increased ≈4-fold, AQP4 oocytes ≈20-fold, and AQP9 + AQP4 oocytes ≈12-fold (Fig. 1A Left).
Coefficients of Ps were determined by measuring oocyte swelling after transfer to isoosmotic Barth's solution containing 100 mosM solute (glycerol, D- or L-β-hydroxybutyrate). Water-injected oocytes failed to swell after transfer to glycerol, and AQP4 oocytes exhibited minimal permeability (Fig. 1A Right). AQP9 oocytes exhibited increased glycerol permeability, but permeability of the AQP9 + AQP4 oocytes was 2-fold higher. This synergistic response is presumed to represent rapid glycerol influx through AQP9 followed by rapid water influx through AQP4. D- and L-β-hydroxybutyrate permeabilities were too low to measure accurately.
Solute permeabilities were also determined by measuring 14C-labeled solute uptake. At neutral pH, AQP9 oocytes exhibited much higher permeability to 14C-urea and 14C-glycerol than AQP4 oocytes (Fig. 1B). At pH 7.5, the permeability of AQP9 oocytes to the racemic mixture DL-14C-β-hydroxybutyrate was too low to measure accurately. At pH 5.5, the permeability of AQP9 oocytes for DL-14C-β-hydroxybutyrate was 10 times higher than the AQP4 oocytes (Fig. 1B).
Permeability of AQP9 in Proteoliposomes. Oocytes express endogenous membrane proteins, so a defined system was used to measure permeability of purified AQP9 protein. Rat AQP9 with a polyhistidine tag at the N terminus was expressed in S. cerevisiae and purified. Coomassie staining and immunoblotting of SDS/PAGE showed a major band at ≈32 kDa and a ladder of oligomers (Fig. 2A). Purified AQP9 and pure E. coli lipids were reconstituted into resealed proteoliposomes for comparison with control liposomes reconstituted from lipid without protein.
AQP9 proteoliposomes loaded with glycerol, urea, or DL-β-hydroxybutyrate were rapidly transferred by stopped-flow to isoosmotic fluid containing sucrose as a nonpermeant osmolyte. Permeable membranes release solute, causing water efflux measurable by light scattering. Glycerol permeability of AQP9 proteoliposomes was 63 times higher than liposomes and was almost entirely blocked with 1 mM HgCl2 (Fig. 2B); urea permeability of AQP9 proteoliposomes was 90 times higher than liposomes. When exposed to a simple osmotic gradient, AQP9 proteoliposomes were only 1.4 times as permeable as liposomes. At neutral pH, DL-β-hydroxybutyrate permeability of AQP9 proteoliposomes and liposomes was equivalent (Fig. 2B), even if the proteoliposomes were reconstituted in the presence of DL-β-hydroxybutyrate. Precipitations appeared when DL-β-hydroxybutyrate permeability measurements were attempted at pH 5.5 (not shown).
AQP9 Expression in Liver. AQP9 protein levels were increased in liver of rats within 12 h of fasting (not shown). Continued fasting for up to 96 h with free access to water caused incremental increases in AQP9 levels (Fig. 3A). During this time, the average glucose and urea concentrations decreased by approximately half, whereas β-hydroxybutyrate increased 4-fold (Fig. 3B). Liver AQP9 levels increased in all five of our experiments, rising as high as 20-fold above baseline. When rats fasted 96 h were refed, levels of AQP9 in liver gradually declined after 3 days and returned to baseline levels after 1 week (Fig. 3C).
Ketogenic and high protein diets did not alter liver AQP9 levels. Ketogenic diets increase serum concentrations of β-hydroxybutyrate (14), but after a week rats did not exhibit changes in liver AQP9 despite a 20-fold increase in serum β-hydroxybutyrate (not shown). Rats fed a high protein diet for a week did not exhibit changes in liver AQP9, despite a 4-fold increase in serum urea (not shown).
Effects of Insulin on AQP9 Expression in Liver. A single injection of 5 units/kg regular insulin (rapid-acting) caused a reduction in AQP9 in liver after 24 h (Fig. 4A). To evaluate the chronic effects of insulin, rats were injected with 100 mg/kg streptozotocin to destroy pancreatic β-cells; control rats were injected with buffer. After 5 days, the streptozotocin-treated rats were divided into two groups: one group was injected twice daily with 13 units/kg NPH insulin (long-acting) for 4 days; the other group received no injections. By 5 days, the streptozotocin-injected rats were all frankly diabetic, and by 9 days serum glucose levels were ≈8-fold higher than control rats, whereas serum glucose levels were normal in streptozotocin-injected rats treated with insulin (Fig. 4B). AQP9 expression in liver was increased >2-fold in the livers of streptozotocin-injected rats when compared with livers from control rats or streptozotocin-injected rats treated with insulin (Fig. 4B). Changes in liver AQP9 expression were not observed in rats injected s.c. with 3 mg/kg dexamethasone or 450 μg/kg glucagon when compared with control buffer-injected rats (data not shown).
Confocal Immunofluorescence Microscopy of AQP9 in Liver. Consistent with previous studies (2–4), distinct AQP9 immunolabeling was observed exclusively in plasma membranes at the sinusoidal surfaces of hepatocyte plates of livers from control rats (Fig. 5A). The overall distribution of AQP9 was similar in livers of fasted animals, although the level of immunolabeling was dramatically increased throughout the lobules (Fig. 5B). Furthermore, weak staining of nonsinusoidal surfaces was observed, and hepatocytes were of markedly smaller size. Similar changes were observed after streptozotocin treatment (Fig. 5C).