Because of potential physiological importance, we investigated transport of water, glycerol, urea, and β-hydroxybutyrate by AQP9 (Fig. 1). Osmotic swelling assays in oocytes showed that rat AQP9 transports water much less rapidly than a water-selective channel like AQP4. In contrast, isoosmotic swelling assays confirmed that AQP9 oocytes, but not AQP4 oocytes, transport glycerol. Rates of β-hydroxybutyrate transport were too low to measure accurately. These results were confirmed by measuring uptake of 14C-urea and 14C-glycerol. At pH 7.5, permeability of AQP9 oocytes to 14C-β-hydroxybutyrate was negligible. At pH 5.5 (still above the pKa of 4.7), a modest increase in permeability was detected.
Studies of purified AQP9 protein have not previously been reported. Proteoliposomes reconstituted with purified AQP9 were highly permeable to glycerol and urea, whereas β-hydroxybutyrate permeability was essentially the same as background, and water permeability was only slightly above background (Fig. 2). Thus, AQP9 is much more permeable to glycerol and urea than to water and is essentially impermeable to β-hydroxybutyrate at physiological pH levels.
In vivo studies of AQP9 protein have not previously been reported. Fasting is known to increase blood glycerol levels, and up-regulation of AQP9 was observed in liver as expected. Moreover, AQP9 protein levels continued to rise in liver after fasting up to 96 h, and refeeding caused a gradual return to baseline levels (Fig. 3C). During starvation, the liver releases β-hydroxybutyrate and other ketone bodies to substitute for glucose. Kidney and skeletal muscle use ketone bodies but do not express AQP9 (2). Although AQP9 is expressed in brain, fasting-induced changes in protein levels were not observed (not shown).
We investigated other metabolic states known to cause elevations of serum β-hydroxybutyrate or urea; however, levels of AQP9 were not increased in liver. Other urea transporters (18) and β-hydroxybutyrate transporters (19) have been identified in liver; the latter are up-regulated in brain of rats on ketogenic diets (14) and in liver of fasted rats (not shown). The only known mammalian glycerol channels or transporters are aquaporins 3, 7, and 9. AQP7 is not expressed in rat liver (20), and it is not known whether AQP3 is up-regulated in situations where the liver requires increased glycerol import. AQP7 is expressed in adipose tissue (6), where the protein is believed to facilitate the release of glycerol during fasting (21–23). Coordinated regulation of genes encoding mouse AQP7 and AQP9 was recently reported (9). This regulation is thought to maximize glycerol release from adipocytes and glycerol uptake by the liver.
As during fasting, blood insulin levels are low in type 1 diabetes mellitus. Injection of streptozotocin produces a well established animal model of insulin-dependent type 1 diabetes (24). Compared with control rats, streptozotocin-injected rats had elevated blood glucose levels and significantly increased AQP9 protein levels in liver. Treatment of streptozotocin-injected rats with long-acting NPH insulin restored AQP9 to baseline (Fig. 4B). One day after administering rapid-acting regular insulin to normal rats, AQP9 levels in the liver decreased 50%, even though blood glucose levels had returned to normal (Fig. 4A). These results suggest that insulin represses AQP9 expression in vivo.
It is not certain why AQP9 levels are reduced more slowly after refeeding fasted rats than after insulin treatment (compare Figs. 3C and 4A). In the refed state, insulin levels presumably returned to normal soon after initial refeeding, so it is likely that additional metabolic signals, such as glucocorticoids, may complicate the response. Glucocorticoid levels in the blood increase during fasting and are required for the increased gluconeogenesis in starvation and diabetes (25, 26). In cultured hepatoma cells, dexamethasone increased expression of AQP9, whereas insulin and dexamethasone added together had little effect on AQP9 protein levels (not shown). The human AQP9 promoter contains a glucocorticoid receptor binding motif (27), but it is not known whether elevated glucocorticoid levels persist proportionally longer after longer fasts.
Glycerol is not normally an important gluconeogenic precursor for humans after overnight fasting, but after fasting human subjects for up to 3.5 days, as much as 21% of circulating glucose is derived from glycerol (28), and fasted rodents have even higher values of circulating glycerol (29, 30). Taken together, recently published studies of AQP9 mRNA (9) and our studies of AQP9 protein strongly support the role for AQP9 expression in liver as a molecular mechanism for maximizing glycerol influx and urea efflux during states requiring increased gluconeogenesis.
We thank M. Daniel Lane and Nicholas F. LaRusso for critical readings of the manuscript, Visvanathan Chandramouli for helpful discussions, and Ida Maria Jalk, Birgit Bonefeld, Weidong Wang, and Lynn Wachtman for assistance. This work was supported in part by grants from the U.S. Public Health Service, National Institutes of Health, the Human Frontier Science Program, the National Danish Research Foundation, the Karen Elise Jensen Foundation, and the European Commission (QLK3-CT-2000-0078).