Primary bile is generated by an osmotic process generated by the active secretion of biliary constituents from the portal blood or the hepatocyte cell interior into the bile canalicular lumen. Luminal accumulation of osmotically active solutes is followed by the passive movement of water into the bile canaliculus [14,15]. Canalicular bile is then modified by cholangiocytes, the epithelial cells lining the intrahepatic bile ducts, via secretory and absorptive processes, stored and concentrated in the gallbladder (with few exceptions), and released into the intestine. At a protein level, at least seven distinct aquaporins (AQP0, 1, 3, 4, 8, 9, and 11) have been identified in the hepatobiliary tract and appear to play a key role in liver, biliary tree and gallbladder physiology (Tables 1 and 2) .
Rodent hepatocytes express four AQPs: AQP0, 8, 9 and 11 [17–20], while a fifth one (AQP3) is found in human liver . Although redundant, liver AQPs have distinctive subcellular localizations.
AQP8, the most abundant aquaporin in mouse and rat hepatocyte, features multiple subcellular localizations  being found in the inner membrane of mitochondria and in membranes of the smooth endoplasmic reticulum (SER) adjacent to glycogen granules , as well as in subapical vesicles and in the canalicular membrane [17,19,22–24]. At least two alternatively spliced isoforms of AQP8 with distinctions in their N-termini have been identified by our laboratory both in mouse and rat hepatocytes [25,26]. We recently suggested that rodent hepatocytes contain two pools of AQP8: an intracellular pool residing permanently within the hepatocyte cell interior (SER and mitochondria), and a choleretic pool shuttling between cytoplasm (subapical vesicles) and canalicular membrane which is controlled by choleretic stimuli . Regarding the intracellular pool, while SER AQP8 may have a role in preserving cytoplasmic osmolality during glycogen synthesis and degradation in the liver , the physiological relevance of mitochondrial AQP8 remains to be fully assessed . Potential roles for mitochondrial AQP8 in the uptake of NH4+ to supply the urea cycle [29,30] and in releasing H2O2 from the mitochondrial matrix during generation of reactive oxygen species , have been timidly hypothesized. The choleretic pool of AQP8 has been suggested to be relevant to canalicular secretion of water. Indeed, following stimulation by choleretic agonists, such as dibutyryl cyclic adenosine monophosphate (cAMP) or glucagon, subapical AQP8 translocates to the canalicular plasma membrane. This redistribution increases the apical cell surface permeability leading to facilitation of the osmotic water transport into the canalicular lumen [23,31,32]. This hypothesis was extended by studies showing that, like other canalicular transporters mediating primary bile secretion, hepatocyte AQP8 is significantly and specifically increased in apical lipid ‘raft’ microdomains involved in canalicular secretion, following exposure to the choleretic agonist glucagon [33,34]. As suggested by Gradilone et al. , lack of canonical protein kinase A consensi for phosphorylation in AQP8, indicate that the cAMP/glucagon-induced redistribution of such AQP to the canalicular plasma membrane likely occurs via phosphatidylinositol-3-kinase–dependent microtubule-associated trafficking of vesicles. This is the case in the isoform 2 of the Cl−/HCO3− exchanger (AE2) and the multidrug resistance-associated protein 2 (MRP2), two carriers mediating canalicular bile secretion. However, additional work is required to explain the apparent absence of bile flow phenotypes in AQP8 knockout mice . Potential answers may relate to the redundancy characterizing AQPs in hepatocyte, functional modification of other genes in response to the elimination of the target gene as well as the fact that a 60% reduction of AQP8 protein in the rat hepatocyte apical membrane corresponds to a 15% decrease in the overall canalicular osmotic permeability .
AQP9 is an aquaglyceroporin being permeable to a wide variety of neutral solutes including polyols, carbamides, purines, pyrimidines and toxic metalloids (i.e. antimonite and arsenite) in addition to water [37,38]. AQP9 is localized in the hepatocyte basolateral (sinusoidal) membrane [18,39] (Fig. 1d) where it appears unaffected by the choleretic stimulus [31,40,41]. The expression of hepatic AQP9 is sex-linked as male rat hepatocytes express higher levels of AQP9 than the female counterpart . AQP9 is the entry pathway for plasma glycerol deriving from adipose lipolysis [42–46], a major substrate for hepatic gluconeogenesis. This function is strongly supported both by the fact that AQP9 is up-regulated by fasting in wild type mice but not in mice lacking the γ isoform of the peroxisome proliferator-activated receptor (PPAR γ) , and the 2-fold increase to which hepatic AQP9 undergoes in vivo in rats fasted for 96 h . Metabolic relevance for AQP9 is also indicated by its negative regulation by insulin . AQP9 was also suggested to be responsible for the osmotic uptake of water from the sinusoidal blood into the rat hepatocyte cell interior , and thus, together with AQP8, contributes to primary bile secretion. AQP9 might also be important for the rapid shifts of water across, into or out of the hepatocyte underlying the so called hepatocellular hydration state, an efficient mechanism of short-term control of canalicular secretion and hepatocyte volume [47,48]. Moreover, a role for AQP9 as an exit channel for urea produced within the hepatocyte or solutes, such as purines and pyrimidines derived from nucleotide synthesis de novo, lactate and ketone bodies has been hypothesized . Based on its proven capacity to transport certain heavy metals AQP9 has been suggested to represent the entry route of arsenic in hepatocyte  whose consequent poisoning is known to lead to hepatocellular damage and hepatocellular carcinoma.
AQP11, an AQP marked by an unusual pore-forming NPA motif , was found to be expressed in mouse liver membranes . While the function of AQP11 as a water channel remains controversial [50,51], of note is the phenotype expressed by the knockout mice lacking AQP11 as such mutant mice die before weaning as a consequence of the uraemia from polycystic kidney disease .
The functional role of AQP0 remains unclear so far. AQP0 is an AQP characterized by weak permeability to water  and is weakly expressed in intracellular vesicles of rat pericentral hepatocytes . Clues in understanding liver AQP0 may be provided by recent work showing that lens AQP0 forms membrane junctions in vivo .
AQP3 mRNA is found in liver parenchima  where it results up-regulated during hepatic steatosis , an observation suggesting potential involvement in lipid metabolism. However, additional work is needed to fully assess both the subcellular localization and functional meaning of AQP3 in liver.
Intrahepatic bile duct AQPs
Although cholangiocytes, the epithelial cells lining the intrahepatic bile ducts, represent only 3–5% of the liver cell population, they play a considerable role in bile formation by contributing up to 40% of the daily output of bile fluid to form the so called ductal bile . At a protein level, cholangiocytes express at least two AQPs, AQP1 and AQP4, whose regulation and functional involvement has been the object of intense investigation (see Masyuk and LaRusso for a review ). In basal conditions, cholangiocyte AQP1 has been reported to reside in the membrane of subapical vesicles whereas under active choleresis secretion would stimulate the exocytotic insertion of AQP1 in the apical plasma membrane with consequent facilitation of water transport into the biliary lumen [56–58]. On its side, AQP4, the basolateral membrane AQP of cholangiocytes, has been suggested as the main channel in mediating the uptake of water from the peribiliary vascular plexus surrounding the bile duct to the cell interior of rat cholangiocytes [14,59,60].
AQP1 and AQP4 would also be implicated in the intrahepatic bile duct absorption of water as suggested by Masyuk and co-workers . Most of the driving force underlying the ductal absorption of water would be generated by the uptake of conjugated bile salts (cholehepatic circulation of bile salts) and glucose by the apical Na+-dependent bile acid transporter ASBT and Na+-coupled glucose transporter SGLT1, respectively, expressed by cholangiocytes [62,63]. Reabsorption of glucose and/or bile salts by cholangiocytes would be stimulated by somatostatin [64,65], an observation consistent with the known cholestatic action of this hormone. However, although the working model involving AQPs as transcellular membrane pathways in bile duct secretion and absorption appears as a plausible one, its ultimate validation requires additional studies, especially in the light of reported observations that deletion of AQP1 does not affect bile formation in mice . Work is also needed to evaluate the potential presence of the AQP11 protein in cholangiocytes.
The epithelial cells of human and mouse gallbladder express AQP1 and AQP8 [67,68].
In the human gallbladder, AQP1 is localized on both the apical and basolateral plasma membranes of epithelial cells lining the neck portion of the organ . A different profile is found in the mouse gallbladder, where AQP1 is localized in the apical and basolateral plasma membranes as well as in subapical vesicles of the epithelial cells of the neck and corpus portions  (Fig. 1e). Mouse gallbladder AQP1 has been found to be moderately regulated at transcriptional level by the satiety hormone leptin . In mouse gallbladder, AQP8 is localized primarily apically and intracellularly to a lesser extent .
The pattern of subcellular distribution of AQP8 and AQP1 strongly corroborates the hypothesis of a transcellular route for the movement of water across the gallbladder epithelium. Osmotic water would cross the apical membrane through AQP8 and AQP1, while AQP1 would be the facilitated pathway for the movement of water across the basolateral membrane. The presence of two distinct AQPs in the apical membrane is an unusual finding and may relate to the membrane's ability both to absorb and secrete fluid. A tempting idea is that AQP1 is hormonally translocated to the gallbladder apical membrane to secrete water as in the bile duct epithelium, a functional homologue of the gallbladder epithelium, whereas apical AQP8 may account for the absorption of water from gallbladder bile. Additional experimental work is needed to prove this hypothesis.
A third aquaporin, AQP4, has been recently detected at a transcript level in mouse gallbladder by a study using leptin-deficient (Lepob) animals . Leptin replacement moderately enhanced AQP1, whereas down-regulated AQP4 leptin was suggested to alter gallbladder volume by mediating gallbladder absorption/secretion of water by acting on the expression of aquaporins. Nevertheless, the presence and localization of the AQP4 protein in the gallbladder remain to be demonstrated.