- Email: [email protected]
Anchoring of aquaporin-4 in brain: Molecular mechanisms and implications for the physiology and pathophysiology of water transport
Neuroscience 129 (2004) 999 –1010
ANCHORING OF AQUAPORIN-4 IN BRAIN: MOLECULAR MECHANISMS AND IMPLICATIONS FOR THE PHYSIOLOGY AND PATHOPHYSIOLOGY OF WATER TRANSPORT M. AMIRY-MOGHADDAM,* D. S. FRYDENLUND AND O. P. OTTERSEN
2004) and its precise colocalization with the inward rectifying potassium channel, Kir4.1 (Nagelhus et al., 1999) it has been proposed that AQP4 serves to facilitate K⫹ buffering, through an indirect mechanism that remains to be fully resolved (Nagelhus et al., 2004). An involvement of AQP4 in K⫹ buffering could underlie the extracellular volume changes that occur at sites of neuronal activation and would be in line with the view that aquaporins are needed to provide a rapid and selective transport of water through plasma membranes (Agre et al., 2002). An entirely different issue is whether AQP4 is engaged in constitutive water fluxes, which in the brain are likely to be small compared, e.g. to the situation in the kidney where aquaporins have well documented roles (Nielsen et al., 1999). Constitutive water fluxes occur across all internal and external brain surfaces. The net water flow from blood to brain has been estimated to 0.17 l/g/min (Oldendorf, 1970). At the pial surface of the brain the net water flux appears to be directed outwards but the magnitude of this flux remains to be determined. As pointed out elsewhere in this volume (Papadopoulos et al., 2004; Nagelhus et al., 2004), the most conspicuous feature of the AQP4 expression pattern in the CNS is the enrichment of this water channel at the interface between the brain and the associated liquid spaces (blood, subarachnoidal space, and ventricles; Frigeri et al., 1995; Nielsen et al., 1997). At the pial and vascular interfaces the AQP4 pool is concentrated in the subpial and perivascular endfeet, respectively, whereas the ventricle wall shows an accumulation of AQP4 in the basolateral membranes of ependymal cells and in subjacent glial processes (Nielsen et al., 1997). AQP4 is also expressed by brain endothelial cells albeit at a relatively low level (Amiry-Moghaddam et al., 2004). It is a reasonable assumption that the above pools of AQP4 take part in activity dependent and constitutive water fluxes and that AQP4 in this capacity confers an advantage that outweighs the increased vulnerability to edema that the expression of AQP4 entails. The first part of this review will discuss the molecular mechanisms that are thought to be responsible for the anchoring of AQP4 at brain– blood and brain–CSF interfaces. The second part will show how our insight in these anchoring mechanisms has provided a tool for analyses of the functional role of AQP4 in the brain. Notably, we will review data obtained in mice that are depleted of AQP4 at the brain– blood and brain–CSF interfaces following knockout of ␣-syntrophin, a major anchoring protein for brain AQP4 (Neely et al., 2001). Such mice show delayed onset of postischemic and hyponatremic edema, and a reduced
Centre for Molecular Biology and Neuroscience, Institute of Basic Medical Sciences, University of Oslo, POB 1105 Blindern, N-0317 Oslo, Norway
Abstract—Astrocytes show an enrichment of aquaporin-4 (AQP4) in those parts of the plasma membrane that are apposed to pial or perivascular basal laminae. This observation begged the following questions: 1, What are the molecular mechanisms that are responsible for the site specific anchoring of AQP4? 2, What are the physiological and pathophysiological roles of the AQP4 pools at these specialized membrane domains? Recent studies suggest that the site specific anchoring depends on the dystrophin complex. Further, ␣-syntrophin (a member of the dystrophin complex) is required to maintain a polarized expression of AQP4 in the perivascular membranes. Hence transgenic mice deficient in ␣-syntrophin provided a model where the perivascular pool of AQP4 could be removed for assessment of its functional roles. Data suggest that the perivascular pool of AQP4 plays a role in edema formation and that this pool (through its serial coupling with the AQP4 pools in other astrocyte membranes) is involved in Kⴙ siphoning. In the cerebral cortex, the astrocyte membrane domain contacting the pial basal lamina differs from the perivascular membrane domain in regard to the mechanisms for AQP anchoring. Thus deletion of ␣-syntrophin causes only a 50% loss of AQP4 from the former membrane (compared with a 90% loss in the latter), pointing to the existence of additional anchoring proteins. We will also discuss the subcellular distribution and anchoring of AQP4 in the other cell types that express this protein: endothelial cells, ependymal cells, and the specialized astrocytes of the osmosensitive organs. © 2004 IBRO. Published by Elsevier Ltd. All rights reserved. Key words: AQP4, aquaporin, water channels, syntrophin, brain edema, astrocyte.
There is now compelling evidence that AQP4, the predominant aquaporin in the CNS, is functionally active (AmiryMoghaddam and Ottersen, 2003; Nicchia et al., 2003; Solenov et al., 2004) and that it is implicated in the generation and resolution of different forms of experimentally induced brain edema (Manley et al., 2000; Vajda et al., 2002; AmiryMoghaddam et al., 2003a, 2004; Papadopoulos et al., 2004). Recent data have also provided some clues as to the physiological role of AQP4. Notably, judged by functional analyses (Amiry-Moghaddam et al., 2003b; Binder et al., *Corresponding author. Tel: ⫹47-22-85-1495; fax: ⫹47-22-85-1299. E-mail address: [email protected] (M. Amiry-Moghaddam). Abbreviations: AQP4, aquaporin-4; CSF, cerebrospinal fluid; dDAVP, [deamino-cysl, D-Arg8]-vasopressin; MCAO, medial cerebral artery occlusion.
0306-4522/04$30.00⫹0.00 © 2004 IBRO. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.neuroscience.2004.08.049
999
1000
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
Fig. 1. Distribution of proteins thought to be involved in the anchoring of AQP4 at astrocyte endfeet. Immunofluorescence analysis of neocortex (A, C) and cerebellar cortex (B, D) using antibodies to dystrophin (A, B) or ␣-syntrophin (C, D). The antibody to dystrophin recognizes the brain specific isoform Dp 71 (see Fig. 11). Arrows and arrowheads indicate subpial and perivascular labeling, respectively. Electron microscopy shows that the ␣-syntrophin labeling (Fig. 2) and dystrophin labeling (not illustrated) reside in perivascular astrocyte membranes. Scale bars⫽50 m.
capacity for K⫹ buffering (Amiry-Moghaddam et al., 2003a, b, 2004). These findings complement data obtained by use of other transgenic approaches (Manley et al., 2000; Vajda et al., 2002) and suggest that the AQP4 pools that are apposed to brain microvessels are of physiological and pathophysiological importance. The dystrophin molecular complex is expressed at the brain– blood and brain–CSF interfaces and is involved in AQP4 anchoring The dystrophin complex is known to play a crucial role in the molecular organization of the sarcolemma of striated muscle fibers (Zubrzycka-Gaarn et al., 1988). Mdx mice that lack dystrophin show a pronounced loss of AQP4 in striated muscle and were also found to exhibit a perturbed expression of AQP4 in brain (Frigeri et al., 2001). In agreement, biochemical studies have provided evidence of a molecular interaction between brain AQP4 and members of the dystrophin complex (Neely et al., 2001). These data called for an analysis of the expression pattern of dystrophin and associated molecules in brain. Immunofluorescence shows that the major brain dystrophin (Dp-71) is concentrated around brain microvessels
and near the pial surface (Fig. 1A, B). The pattern of labeling obtained with antibodies to Dp-71 is mimicked by antibodies to ␣-syntrophin (Fig. 1C, D), consistent with the idea that ␣-syntrophin is part of the brain dystrophin complexes (cf. Fig. 11). More detailed analysis in the electron microscope revealed that ␣-syntrophin (Fig. 2) and Dp-71 (not shown) are expressed in the perivascular and subpial endfeet of astrocytes, notably in those membrane domains that are directly apposed to the basal lamina. By immunocytochemistry it could be demonstrated that AQP4 is enriched at those sites that express Dp-71 and ␣-syntrophin (compare Fig. 3A, C with Fig. 1) and in the same membrane domains (compare Figs. 4 and 5 with Fig. 2). To test whether the precise colocalization between AQP4 and ␣-syntrophin reflects a coupling between the two molecules we analyzed mice in which the ␣-syntrophin gene had been deleted (referred to as ␣-syn⫺/⫺). These ␣-syn⫺/⫺ mice showed a dramatic reduction of AQP4 immunoreactivity in perivascular astrocyte membranes (Fig. 3–5; Amiry-Moghaddam et al., 2003a,b, 2004; Neely et al., 2001). However, some AQP4 labeling remained, most notably in the granule cell layer of the cerebellum and
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
Fig. 2. ␣-Syntrophin immunogold labeling in cerebellum (A) and neocortex (B) of wild type (control) animals. Immunogold particles signaling ␣-syntrophin are concentrated at the perivascular membrane (double arrowheads) of astrocyte endfeet, while the opposite membrane (i.e. the endfoot membrane domain facing neuropil; double-headed arrow) is devoid of labeling. End, endothelial cells; L, vessel lumen. Scale bars⫽500 nm.
in the zone just subjacent to the pial surface of the neocortex (Amiry-Moghaddam et al., 2004). The existence of such an ␣-syntrophin independent pool explains the mis-
1001
match between the patterns of ␣-syntrophin/Dp-71 labeling and AQP4 labeling in immunofluorescence preparations of wild type animals (compare Fig. 3 and Fig. 1). Immunogold analysis showed that the ␣-syntrophin independent AQP4 pools in the superficial parts of the neocortex and in the granule cell layer of the cerebellum reside in non-endfeet membranes of astrocytes (AmiryMoghaddam et al., 2004; Fig. 4). In fact, the labeling intensity of these membranes was enhanced following deletion of ␣-syntrophin, indicative of a subcellular redistribution of AQP4 from endfeet to non-endfeet membranes (Fig. 4C). That a deletion of ␣-syntrophin causes a mislocalization rather than a net loss of AQP4 is in line with data obtained from quantitative immunoblots (Neely et al., 2001). The mislocalization of AQP4 in ␣-syn⫺/⫺ animals can be viewed as a deficient polarization of AQP4 expression. Hence, in astrocytes that normally express low levels of AQP4 in non-endfeet membranes (the majority) the deletion of ␣-syntrophin will nearly abolish the endfeet labeling. This contrasts with the situation in the minority of astrocytes (located subjacent to the pial surface of the neocortex and in the granule cell layer of the cerebellum) that normally exhibit a sizable pool of AQP4 in non-endfeet membranes: here loss of polarization implies that endfeet membranes retain a significant density of AQP4, on a par with that in non-endfeet membranes. In other words, common for both types of astrocyte is that removal of ␣syntrophin leads to an even distribution of AQP4 across all
Fig. 3. Immunofluorescence labeling of AQP4 in the neocortex (A, B) and cerebellar cortex (C, D) of wild type (A, C) and ␣-syn⫺/⫺ animals (B, D). A comparison between the pattern of AQP4 labeling in A and C with the patterns of dystrophin and ␣-syntrophin labeling in Fig. 1 reveals a clear mismatch. Thus, non-endfeet pools of AQP4 are found in the superficial zone of the cerebral cortex (A) and in the granule cell layer of the cerebellum (C). (The expression in non-endfeet membranes was confirmed by electron microscopy). These pools remain after deletion of ␣-syntrophin (B, D). Electron microscopic analyses reveal that the ␣-syntrophin independent pool of AQP4 is distributed in a non-polarized fashion in astrocyte membranes (Fig. 4). Arrows, cortical surfaces; arrowheads and double arrows, labeled and unlabelled microvessels. Asterisks indicate Purkinje cell layer. Scale bars⫽50 m. From Amiry-Moghaddam et al., 2004.
1002
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
Fig. 4. Quantitative immunogold analysis of AQP4 distribution in wild type compared with ␣-syn⫺/⫺ animals. (A) Deletion of ␣-syntrophin is associated with a loss of the polarized AQP4 distribution in neocortical and cerebellar astrocytes. In wild type (WT) animals, perivascular membranes are enriched in AQP4 compared with endfeet membranes facing neuropil (see Fig. 2 for definition of membrane domains). In syn⫺/⫺ animals these two membrane domains show the same level of immunolabeling. (B) Effect of ␣-syntrophin deletion on AQP4 immunolabeling in subpial endfoot membranes. Same experimental design as in A. In the cerebellum of syn⫺/⫺ animals, the labeling of subpial membranes is similar to that of endfeet membranes facing neuropil. In the neocortex, the polarization of labeling is reduced but not abolished (see text). Labeling intensity is expressed as number of particles per um membrane. (C) Effect of ␣-syntrophin deletion on the AQP4 labeling intensity of non-endfeet membranes. In the superficial zone of the neocortex (i.e. the upper approximately 5 m of layer I) and in the deep part of the cerebellar cortex (i.e. granule cell layer) the labeling of non-endfoot membranes is increased after ␣-syntrophin deletion. No significant changes are seen deeper in the neocortex or more superficially in the cerebellar cortex. In this case, the labeling density was recorded as number of particles per m2, as many membranes could be traced only for short distances. Asterisks, significantly different from other member of the pair (P⬍0.01).
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
1003
Fig. 5. Immunogold analyses reveal two pools of AQP4 at the brain– blood interface. The major pool resides in the perivascular membranes of astrocytes (double arrows) and is depleted after ␣-syntrophin knockout. The minor pool resides in the abluminal (arrows) and adluminal membrane (arrowheads) of endothelial cells (with a predominance in the former) and is independent of ␣-syntrophin (B). (A) Wild type animals; (B) ␣-syn⫺/⫺ animals. From the molecular layer of the cerebellum. E, endothelial cell; L, capillary lumen; P, pericyte. Scale bars⫽500 nm. From Amiry-Moghaddam et al. (2004).
membrane domains (Fig. 4A). There is one exception to this rule: the subpial endfeet in the neocortex retain some polarized expression of AQP4 even after ␣-syntrophin deletion (Fig. 4B; Amiry-Moghaddam and Ottersen, unpublished observations), indicating that the enrichment of AQP4 in this domain does not depend on ␣-syntrophin alone but that another molecule (another syntrophin?) is also involved. It should be noted that deletion of ␣-syntrophin has no effect on the level or distribution of AQP4 in the specialized astrocytes in the supraoptic nucleus and other osmosensitive organs. In wild type animals (Nielsen et al., 1997) as in ␣-syn⫺/⫺ animals the latter astrocytes exhibit very high densities of AQP4 along their entire surfaces, except in those membrane domains that abut on neuronal elements. In conclusion, brain astrocytes can be divided in three categories, based on their AQP4 distribution pattern in wild type and alpha-syn⫺/⫺ animals. The most prevalent type has a modest density of AQP4 along non-endfeet membranes and a strong, ␣-syntrophin dependent enrichment of AQP4 in endfeet membranes. A second type, found superficially in the neocortex and deep in the cerebellar cortex, is similarly endowed with an ␣-syntrophin dependent AQP4 pool at the endfeet. However, it differs from the first type by showing relatively strong non-endfeet labeling that persists or even increases after ␣-syntrophin deletion. The third type is the specialized astrocytes in osmosensitive organs referred to above. The evidence discussed above suggests that ␣syntrophin is the key anchoring protein for the endfeet pools of AQP4. It remains to identify the protein (or proteins) that anchor AQP4 at non-endfeet membranes. The latter membranes are not labeled for Dp-71 (Fig. 1), indicating that dystrophin complexes are not involved (although one must open for the possibility that these mem-
branes express dystrophin complexes that are not recognized by the Dp-71 antibody). AQP4 in endothelial cells and ependymocytes AQP4 is expressed not only in astrocytes but also in endothelial cells (Nagelhus et al., 1998; Amiry-Moghaddam et al., 2004; Fig. 5) and ependymocytes (Nielsen et al., 1997; not shown). The former cells show AQP4 labeling in abluminal as well as in adluminal membranes. The linear gold particle density over in the abluminal membrane domain is about three times as high as that over the adluminal membrane. However, even the abluminal membrane is weakly AQP4 immunopositive compared with the perivascular astrocyte membrane (Fig. 5). Immunocytochemical analyses suggest that the AQP4 pool in ependymocytes is restricted to the basolateral membranes (not shown; Nielsen et al., 1997). Neither the endothelial nor the ependymal pool of AQP4 is significantly decreased after ␣-syntrophin knockout (Amiry-Moghaddam et al., 2004). ␣-Synⴚ/ⴚ animals: a model for assessing the function of the perivascular AQP4 pool The data reviewed above indicate that ␣-syn⫺/⫺ animals should provide a unique model for unraveling the functional role of the AQP4 pool that is enriched in the astrocyte membrane domains at the brain– blood interfaces. As explained above, depletion of ␣-syntrophin leads to a loss of AQP4 from these membrane domains. There is one exception: some AQP4 remains in the endfeet membranes of the minority of astrocytes that contain a sizeable nonendfoot pool of AQP4. Are other membrane molecules than AQP4 affected by deletion of ␣-syntrophin? If so, this could confound the interpretation of data generated by this model. Immunogold analyses did not reveal any significant changes in the
1004
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
Fig. 6. Removal of the perivascular pool of AQP4 by ␣-syntrophin deletion significantly reduces the extent of postischemic edema (22 h following 90 min of MCAO). The perivascular AQP4 immunolabeling (arrowheads), distinct in wild type animals (A), is nearly abolished after ␣-syntrophin knockout (B). ␣-Syn⫺/⫺ animals show preserved labeling of the endfeet membranes facing the neuropil and a sizable increase in endfoot volume (indicated by double arrows in A and B). E, endothelial cells; L, lumen. (C, D) The volume of postischemic edema (unstained areas in TTC stained frontal sections) is smaller in ␣-syn⫺/⫺ animals (D) than in wild type (C). Scale bars⫽500 nm. Modified from Amiry-Moghaddam et al. (2003a).
perivascular expression levels of Kir4.1 or of monocarboxylate, glutamate, or glucose transporters (AmiryMoghaddam et al., 2003b). However, it cannot be ruled out that the deletion of ␣-syntrophin affects one or several unknown endfoot molecules and that this modifies the water transport capacity of perivascular membranes. It also needs to be resolved whether the ␣-syntrophin deletion leads to alterations of the electrical properties of the endfoot membrane. It should be pointed out that the blood– brain barrier remains intact in the ␣-syn⫺/⫺ animals as they show no evidence of an increased uptake of gadolinium chloride from blood to brain (AmiryMoghaddam, unpublished observations). ␣-Synⴚ/ⴚ animals show a decreased vulnerability to postischemic and hyponatremic edema The ␣-syn⫺/⫺ animals showed a partial protection in models of postischemic and hyponatremic edema. The extent of postischemic edema was assessed 22 h after a 90 min period of medial cerebral artery occlusion (MCAO). Quantitative analyses revealed that the volume of the infarct was 50% smaller in the ␣-syn⫺/⫺ animals than in wild type controls (Fig. 6). The ultrastructure of the infarction core, assessed by electron microscopy, was better preserved in the ␣-syn⫺/⫺ than in the wild type animals (AmiryMoghaddam et al., 2003a; Fig. 7). Hyponatremic edema was induced by i.p. injections of distilled water, combined with 0.4 g/kg [deamino-
cysl, D-Arg8]-vasopressin (dDAVP; see legend to Fig. 8 for details). By use of diffusion weighted MRI it could be shown that the onset of fulminant edema (signaled by a precipitous decline in apparent diffusion coefficient) was delayed by 15–20 min in the ␣-syn⫺/⫺ animals compared with controls (Amiry-Moghaddam et al., 2004; Fig. 8). These results are in line with data from other transgenic approaches (Manley et al., 2000; Vajda et al., 2002) and indicate that the perivascular membrane determines the rate of water influx from blood in pathophysiological conditions. In other words, in such conditions the perivascular endfeet rather than the endothelial cells serve as the primary barrier for water uptake. It is not surprising that the removal of perivascular AQP4 provides only a partial protection in models of brain edema: water may pass in the narrow clefts between the perivascular endfeet (which are not joined by tight junctions) and by slow diffusion through the lipid bilayer of the cell membranes (cf. Fig. 10). In the infarction core of wild type animals there was an almost complete loss of perivascular AQP4 22 h after the ischemic insult (Fig. 7), in the face of a preserved immunolabeling for ␣-syntrophin (not shown). This suggests that the coupling between ␣-syntrophin and AQP4 is sensitive to ischemia. In a clinical setting, the occurrence of an ischemia-dependent loss of perivascular AQP4 would be expected to impede the resolution of the edema, which is
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
1005
Fig. 7. Loss of perivascular AQP4 (B) in the infarction core but not in penumbra (C) of wild type animals 22 h following a 90 min period of MCAO. The corresponding area in the contralateral hemisphere served as a control (A). The infarction core in the ␣-syn⫺/⫺ (D) animals reveals better preserved ultrastructure compared with the wild type (B). E, endothelial cells with nucleus (N); L, lumen; arrowheads, perivascular membrane.
likely to depend on an AQP4 mediated outward flux of water. Recent experimental data support the notion (Amiry-Moghaddam et al., 2003a) that AQP4 facilitates clearance of excess water from brain (Papadopoulos et al., 2004; Papadopoulos et al., 2004). This highlights an important issue that must be duly considered in attempts to develop novel therapeutical strategies: as AQP4 permits bidirectional transport of water, an inhibition of the activity of this protein— expected to confer protection in the buildup phase of edema—might prolong the resolution phase. The mechanisms underlying the postischemic loss of perivascular AQP4 are not known. However, the loss occurs following disintegration of the endothelial tight junctions and onset of vasogenic brain edema. It is conceivable that the molecular events in the endothelial lining affect those in the perivascular endfeet, through the bridging connections provided by dystroglycans and laminin/agrin. This would be in line with the data of Nico et al. (2003) who showed show that disruption of BBB by lipopolysaccharide leads to endfoot swelling and loss of perivascular AQP4 (Nicchia et al., 2004).
␣-Synⴚ/ⴚ animals shows a reduced capacity for Kⴙ clearance In our early studies on brain aquaporins (Nielsen et al., 1997; Nagelhus et al., 1998) we predicted that AQP4 could act in concert with those membrane molecules that are engaged in K⫹ buffering. The main arguments were 1, the density profile of AQP4 along the astrocyte membrane (high densities in endfeet compared with nonendfeet membranes) corresponds to the topological variation in K⫹ conductance, as recorded in retinal Müller cells (Newman, 1984); 2, In Müller cells, AQP4 is precisely coexpressed with Kir4.1, a K⫹ channel crucially involved in retinal K⫹ spatial buffering (Nagelhus et al., 1998); 3, The precise coexpression of Kir4.1 and AQP4 agrees with biochemical data indicating that the two molecules may be part of the same molecular complex (Connors et al., 2004; Guadagno and Moukhles, 2004; Fig. 11); 4, evoked activity induces changes in extracellular space volume (a reduction locally and an increase distally) that match the redistribution of excess K⫹ (Niermann et al., 2001). Our hypothesis implies that astrocytes, in terms of activity induced K⫹ and water transport, should be viewed
1006
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
Fig. 8. Removal of the perivascular pool of AQP4 by ␣-syntrophin deletion (syn⫺/⫺) delays the onset of fulminant edema (dashed lines), as recorded by diffusion weighted MRI. Ordinate shows apparent diffusion coefficients (ADC), abscissa shows time (in minutes) after an i.p. injection of distilled water (20% of body weight) and dDAVP (0.4 g/kg). Schematic drawings (left) indicate the effect of ␣-syntrophin knockout on the two pools of AQP4 (endothelial and perivascular) at the brain– blood interface (cf. Fig. 5): the endothelial pool is not affected. E, endothelial cells; L, lumen. Red dots indicate AQP4. From Amiry-Moghaddam et al., 2004.
as cells that have two distinct membrane domains that are serially coupled and that allow influx and efflux, respectively, of K⫹ and water. The larger domain, corresponding to the astrocyte surface that is apposed to the neuropil, mediates K⫹ and water uptake from active neurons. The smaller domain, corresponding to the perivascular and subpial membranes, represents the site where water and K⫹ exit the astrocytes under high neuronal activity. According to this scheme the role of astrocytes as “cellular
siphons” is not restricted to K⫹ (Newman et al., 1984) but includes water as well. The question remained whether the parallel fluxes of K⫹ and water reflect an obligatory coupling between the two. This question was addressed in the ␣-syn⫺/⫺ model (Fig. 9). Consistent with an obligatory coupling, the ␣-syn⫺/⫺ animals (lacking perivascular AQP4 but with a normal complement of Kir4.1) showed a protracted clearance of K⫹ from the termination areas of orthodromically
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
1007
cussed above) that the ␣-syn⫺/⫺ animals exhibit a protracted buildup as well as a protracted clearance of K⫹. A slower buildup implies increased threshold; a slower clearance implies an increased duration and severity of seizures once induced. In support of this idea, AQP4⫺/⫺ animals revealed stronger seizures than wild type animals in the pentylenetetrazole model (Manley et al., 2004). General discussion
Fig. 9. Removal of the perivascular pool of AQP4 by ␣-syntrophin deletion (Syn⫺/⫺) slows the recovery of extracellular K⫹ concentration following repetitive orthodromic stimulation of isolated hippocampal slices. Representative tracings of [K⫹]o signal in CA1 during and after a 10 Hz train delivered to the Schaffer collaterals from wild type (WT) and Syn⫺/⫺ animals shown separately or superimposed in overlay. Calibration bars show K⫹ concentration in mM and time in seconds. From Amiry-Moghaddam et al., 2003b.
activated fiber pathways (Schaffer collaterals and perforant path; Amiry-Moghaddam et al., 2003b). The molecular mechanisms responsible for this obligatory coupling are not resolved (see Nagelhus et al. (2004) this volume, for a tentative model). But our data indicate that the ability of astrocytes to flux water is a prerequisite for efficient K⫹ buffering. Obviously, volume changes of astrocytes are complementary to the extracellular volume changes. It is quite probable that the volume changes of the extracellular space are essential to provide a K⫹ concentration gradient that is steep enough to allow for a sufficiently rapid redistribution of this ion. This line of reasoning would explain why, in ␣-syn⫺/⫺ animals, there is not only a slower clearance of K⫹ but also a slower buildup of [K⫹]o during orthodromic activation (Fig. 9). The slow buildup could simply reflect an impaired shrinkage of the local extracellular space (due to loss of transglial, AQP4 mediated water transport). In the absence of extracellular volume reduction the released K⫹ will be distributed over a larger volume than in control animals. This concept would be compatible with our finding that the peak K⫹ concentration tends to be slightly lower in ␣-syn⫺/⫺ knockouts than in controls (Amiry-Moghaddam et al., 2003b). If AQP4 is involved in the clearance of K⫹ one would expect changes in its expression to affect cell excitability. Two recent studies have addressed this issue (AmiryMoghaddam et al., 2003b; Binder et al., 2004). AmiryMoghaddam et al. (2003b) showed that removal of perivascular AQP4 (by ␣-syntrophin deletion) enhanced the severity of hyperthermia-induced seizures. Binder et al. (2004) (based on pentylenetetrazol induced seizures) reported that seizure threshold was increased in AQP4⫺/⫺ animals. These two data sets, contradictory at first glance, can be reconciled by taking into account our findings (dis-
With the discovery of the aquaporins, water flow in the brain cannot any longer be discussed in terms of simple diffusion across lipid bilayers but must now be understood as a specific and facilitated flux of water through the hydrophilic interior of specialized membrane proteins. On this background the mechanisms that serve to anchor AQP4 at the plasma membrane are important, as they dictate the subcellular expression pattern of this molecule and hence direct the microscopic and macroscopic water fluxes in the brain. A discussion of the direction and magnitude of water fluxes in the brain must be based on three seminal observations: 1, water flux through aquaporins is bidirectional and is driven by osmotic and hydrostatic forces (Agre et al., 2002): 2, AQP4 —the predominant aquaporin in brain—is expressed in astrocytes and not in neurons; 3, most astrocytes show a remarkable polarization when it comes to AQP4 distribution, implying that some membrane domains are specialized for rapid transmembrane water exchange. However, the latter observation must not overshadow the fact that even in astrocytes with a polarized expression of AQP4 in endfeet membranes (such as the Müller cells), low but significant levels of AQP4 are found also in the nonendfeet membranes. In fact, in the case of the Müller cells the quantitative immunogold data suggest that the total amount of AQP4 in the nonendfeet membranes (which constitute approximately 90% of the total surface) corresponds to the total amount of AQP4 in endfeet membranes (which constitute the remaining approximately 10%; Nagelhus et al., 1999). This is consistent with the idea that astrocytes possess two separate and serially coupled membrane domains with regard to water transport (Nagelhus et al., 2004) as well as K⫹ conductance (Newman, 1984). In fact, as discussed above, evidence is accumulating to suggest that astrocytes mediate parallel fluxes of K⫹ and water in conditions which require buffering of extracellular K⫹. In other words, the spatial redistribution of K⫹ is associated with a similar redistribution of water. During high neuronal activity K⫹ and water may exit preferentially through the endfeet membranes, or (in the case of the astrocytes in the superficial part of the neocortex) through the nonendfeet membranes that form AQP4 enriched lamellae just subjacent to the pial surface (Niermann et al., 2001). In the eye (Nagelhus et al., 1998) and at the neocortical surface the AQP4 (Nielsen et al., 1997) enriched membrane domains are directly or indirectly apposed to liquid spaces (vitreous humor and subarachnoidal space) that act as efficient sinks for water and K⫹. In the brain neuropil the AQP4 enriched membrane domains face microvessels, as expected if anchoring depends on laminin and agrin (central constituents of subpial as well as perivascu-
1008
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
Fig. 10. Diagram showing the possible involvement of AQP4 (indicated in dark blue) in activity dependent and constitutive water flow in the neocortex of wild type and ␣-syn⫺/⫺ animals. Arrows show putative direction of water flux. (A) Activity dependent fluxes in wild type animals: Siphoning of K⫹ is associated with water uptake in astrocytes and subsequent water release through the perivascular and subpial membranes that are strongly enriched in AQP4. Constitutive water fluxes in wild type animals: There is a net secretion of water from the capillary bed and a net production of water from the energy metabolism. The enrichment of AQP4 in endfeet and non-endfeet membranes at the cortical surface serves to facilitate the transport of water into the subarachnoidal space, which acts as a sink. The activity dependent efflux of water through perivascular endfeet is likely to join the constitutive flux of water (asterisk) as osmotic gradients may not favor a net transendothelial water flux from brain to blood (but see text). (B) Possible scenario of changes in water transport after ␣-syntrophin knockout. The distinct swelling of perivascular and subpial endfeet (cf. Fig. 6) may reflect a reduced capacity for water efflux through the endfeet membranes.
lar basal laminae; Guadagno and Moukhles, 2004; Warth et al., 2004). Water that exits the perivascular membrane domains under high neuronal activity might be reflected to join the constitutive flux of water from blood to brain (see below and Fig. 10). It is still unclear if and to what extent brain AQP4 is involved in constitutive water fluxes. In general, aquaporins are essential primarily when there is a need for rapid movement of water across a plasma membrane (Agre et al., 2002). However, it should be recalled that AQP4 provides not only high transport capacity but also high selectivity (Amiry-Moghaddam and Ottersen, 2003). Constitutive water movement by way of cotransporters will inevitably cause redistribution of the cotransported compounds while AQP4 mediates transport of water alone. An involvement of AQP4 in constitutive water flux is consistent with the finding that absence of perivascular AQP4 (in mdx or syn⫺/⫺ mice) is associated with a distinct swelling of the endfeet (Amiry-Moghaddam et al., 2003a; Fig. 6). Control experiments in syn⫺/⫺mice indicate that
these changes are not secondary to a disruption of the blood– brain barrier or to changes in the expression level of known water cotransporters in perivascular membranes. An obvious source of the excess water is the energy metabolism, as water is formed when glucose is broken down. If one assumes that there is normally a constitutive efflux of water across the perivascular membranes, and that this manifests itself in swelling once the perivascular AQP4 pool is removed, then one is confronted with the problem that the direction of flux does not match the direction of net flux across the blood– brain interface. The latter flux is thought to be directed inwardly (Oldendorf, 1970, 1972). However, the concept of a net flux from blood to brain is based on an experimental design (recovery of systemically administered tritiated water; Oldendorf, 1970) that would fail to reveal any heterogeneities in the vascular bed. Thus the possibility remains that influx only occurs in one segment of the vascular bed and that this influx is balanced (or exceeded; Greitz, 2004) by an efflux in an-
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
1009
Fig. 11. Diagram showing the presumed molecular basis for the enrichment of AQP4 in perivascular and subpial membranes. The dystrophin complex is anchored to the basal lamina through laminin and agrin and binds AQP4 (and possibly other molecules with a C-terminal SXV sequence like Kir4.1) by way of ␣-syntrophin (␣-syn) or other syntrophins (syn). The coupling between AQP4 and syn may be indirect (indicated by double arrows). We assume that ␣-syn is crucially involved in the anchoring of the M23 isoform of AQP4 (as M23 is the isoform that forms orthogonal arrays of proteins, particularly enriched in endfeet membranes; Wolburg, 1995). Other syn may be primarily involved in the anchoring of the M1 isoform (which does not form OAPs; Furman et al., 2003). It should be understood that the identity and stoichiometry of the SXV-containing membrane molecules (SSV for AQP4; SNV for Kir4.1) may vary according to region, cell type, and membrane domain. ␣-DG and -DG, ␣ and  dystroglycan; Dp-71, major dystrophin isoform in brain. The PDZ binding domains of syn are indicated. H1 indicates the coiled-coil motif interaction between Dp71 (the major dystrophin isoform in brain) and ␣-dystrobrevin (Sadoulet-Puccio et al., 1997).
other segment. In any case, while transendothelial water flux and flux through perivascular membranes are serially coupled in pathophysiological states (Manley et al., 2000; Vajda et al., 2002; Amiry-Moghaddam et al., 2003a, 2004) this is not necessarily true in all vascular segments at the low flux rates that occur constitutively in physiological conditions. It is possible that the efflux through the perivascular membrane is balanced by a paracellular movement of water in the opposite direction (see above); alternatively, there may be a net movement of fluid along the basal lamina that separates the endfeet from the endothelial cells. This fluid could eventually be drained into the Virchow-Robin spaces which are known to sustain an “active” transport of water due to the piston-like effect of the pulsatile CSF waves (Rennels et al., 1985). The capacity for water transport along the paracapillary route is not known.
REFERENCES Agre P, King LS, Yasui M, Guggino WB, Ottersen OP, Fujiyoshi Y, Engel A, Nielsen S (2002) Aquaporin water channels-from atomic structure to clinical medicine. J Physiol 542:3–16. Amiry-Moghaddam M, Otsuka T, Hurn PD, Traystman RJ, Haug FM, Froehner SC, Adams ME, Neely JD, Agre P, Ottersen OP, Bhardwaj A (2003a) An alpha-syntrophin-dependent pool of AQP4 in astroglial end-feet confers bidirectional water flow between blood and brain. Proc Natl Acad Sci USA 100:2106 – 2111. Amiry-Moghaddam M, Ottersen OP (2003) The molecular basis of water transport in the brain. Nat Rev Neurosci 4:991–1001. Amiry-Moghaddam M, Williamson A, Palomba M, Eid T, de Lanerolle NC, Nagelhus EA, Adams ME, Froehner SC, Agre P, Ottersen OP (2003b) Delayed K⫹ clearance associated with aquaporin-4 mislocalization: phenotypic defects in brains of alpha-syntrophin-null mice. Proc Natl Acad Sci USA 100:13615–13620.
1010
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
Amiry-Moghaddam M, Xue R, Haug FM, Neely JD, Bhardwaj A, Agre P, Adams ME, Froehner SC, Mori S, Ottersen OP (2004) Alpha-syntrophin deletion removes the perivascular but not endothelial pool of aquaporin-4 at the blood-brain barrier and delays the development of brain edema in an experimental model of acute hyponatremia. FASEB J 18:542–544. Binder DK, Oshio K, Ma T, Verkman AS, Manley GT (2004) Increased seizure threshold in mice lacking aquaporin-4 water channels. Neuroreport 15:259 –262. Connors NC, Adams ME, Froehner SC, Kofuji P (2004) The potassium channel Kir4.1 associates with the dystrophin-glycoprotein complex via alpha-syntrophin in glia. J Biol Chem 279:28387–28392. Frigeri A, Gropper MA, Umenishi F, Kawashima M, Brown D, Verkman AS (1995) Localization of MIWC and GLIP water channel homologs in neuromuscular, epithelial and glandular tissues. J Cell Sci 108(Pt 9):2993–3002. Frigeri A, Nicchia GP, Nico B, Quondamatteo F, Herken R, Roncali L, Svelto M (2001) Aquaporin-4 deficiency in skeletal muscle and brain of dystrophic mdx mice. FASEB J 15:90 –98. Furman CS, Gorelick-Feldman DA, Davidson KG, Yasumura T, Neely JD, Agre P, Rash JE (2003) Aquaporin-4 square array assembly: opposing actions of M1 and M23 isoforms. Proc Natl Acad Sci USA 100:13609–13614. Greitz D (2004) The hydrodynamic hypothesis versus the bulk flow hypothesis. Neurosurg Rev 27:299 –300. Guadagno E, Moukhles H (2004) Laminin-induced aggregation of the inwardly rectifying potassium channel, Kir4.1, and the waterpermeable channel, AQP4, via a dystroglycan-containing complex in astrocytes. Glia 47:138 –149. Manley GT, Binder DK, C Marios, Papadopoulos MC, Verkman AS (2004) New insights into water transport and edema in the central nervous system from phenotype analysis of aquaporin-4 null mice. Neuroscience 129:983–991. Manley GT, Fujimura M, Ma T, Noshita N, Filiz F, Bollen AW, Chan P, Verkman AS (2000) Aquaporin-4 deletion in mice reduces brain edema after acute water intoxication and ischemic stroke. Nat Med 6:159 –163. Nagelhus EA, Mathiisen TM, and Ottersen OP (2004) Aquaporin-4 in the central nervous system: cellular and subcellular distribution and coexpression with Kir 4.1. Neuroscience 129:905–913. Nagelhus EA, Horio Y, Inanobe A, Fujita A, Haug FM, Nielsen S, Kurachi Y, Ottersen OP (1999) Immunogold evidence suggests that coupling of K⫹ siphoning and water transport in rat retinal Muller cells is mediated by a coenrichment of Kir4.1 and AQP4 in specific membrane domains. Glia 26:47–54. Nagelhus EA, Veruki ML, Torp R, Haug FM, Laake JH, Nielsen S, Agre P, Ottersen OP (1998) Aquaporin-4 water channel protein in the rat retina and optic nerve: polarized expression in Muller cells and fibrous astrocytes. J Neurosci 18:2506 –2519. Neely JD, Amiry-Moghaddam M, Ottersen OP, Froehner SC, Agre P, Adams ME (2001) Syntrophin-dependent expression and localization of aquaporin-4 water channel protein. Proc Natl Acad Sci USA 98:14108 –14113. Newman EA (1984) Regional specialization of retinal glial cell membrane. Nature 309:155–157. Newman EA, Frambach DA, Odette LL (1984) Control of extracellular potassium levels by retinal glial cell K⫹ siphoning. Science 225:1174 –1175. Nicchia GP, Nico B, Camassa LMA, Mola MG, Loh N, Dermietzel R, Spray DC, Svelto M, Frigeri A (2004) The role of aquaporin-4 in the
blood– brain barrier development and integrity: studies in animal and cell culture models. Neuroscience 129:935–945. Nicchia GP, Frigeri A, Liuzzi GM, Svelto M (2003) Inhibition of aquaporin-4 expression in astrocytes by RNAi determines alteration in cell morphology, growth, and water transport and induces changes in ischemia-related genes. FASEB J 17:1508 –1510. Nico B, Frigeri A, Nicchia GP, Corsi P, Ribatti D, Quondamatteo F, Herken R, Girolamo F, Marzullo A, Svelto M, Roncali L (2003) Severe alterations of endothelial and glial cells in the blood-brain barrier of dystrophic mdx mice. Glia 42:235–251. Nielsen S, Kwon TH, Christensen BM, Promeneur D, Frokiaer J, Marples D (1999) Physiology and pathophysiology of renal aquaporins. J Am Soc Nephrol 10:647– 663. Nielsen S, Nagelhus EA, Amiry-Moghaddam M, Bourque C, Agre P, Ottersen OP (1997) Specialized membrane domains for water transport in glial cells: high-resolution immunogold cytochemistry of aquaporin-4 in rat brain. J Neurosci 17:171–180. Niermann H, Amiry-Moghaddam M, Holthoff K, Witte OW, Ottersen OP (2001) A novel role of vasopressin in the brain: modulation of activitydependent water flux in the neocortex. J Neurosci 21:3045–3051. Oldendorf WH (1970) Measurement of brain uptake of radiolabeled substances using a tritiated water internal standard. Brain Res 24:372–376. Oldendorf WH (1972) Cerebrospinal fluid formation and circulation. Prog Nucl Med 1:336 –358. Papadopoulos MC, Saadoun S, Binder DK, Manley GT, Krishna S, Verkman AS (2004) Molecular mechanisms of brain tumor edema. Neuroscience 129:1011–1020. Papadopoulos MC, Manley GT, Krishna S, Verkman AS (2004) Aquaporin-4 facilitates reabsorption of excess fluid in vasogenic brain edema. FASEB J 18(11):1291–1293. Rennels ML, Gregory TF, Blaumanis OR, Fujimoto K, Grady PA (1985) Evidence for a paravascular fluid circulation in the mammalian central nervous-system, provided by the rapid distribution of tracer protein throughout the brain from the subarachnoid space. Brain Res 326:47– 63. Sadoulet-Puccio HM, Rajala M, Kunkel LM (1997) Dystrobrevin and dystrophin: an interaction through coiled-coil motifs. Proc Natl Acad Sci USA 94:12413–12418. Solenov E, Watanabe H, Manley GT, Verkman AS (2004) Sevenfoldreduced osmotic water permeability in primary astrocyte cultures from AQP-4-deficient mice, measured by a fluorescence quenching method. Am J Physiol Cell Physiol 286:C426 –C432. Vajda Z, Pedersen M, Fuchtbauer EM, Wertz K, Stodkilde-Jorgensen H, Sulyok E, Doczi T, Neely JD, Agre P, Frokiaer J, Nielsen S (2002) Delayed onset of brain edema and mislocalization of aquaporin-4 in dystrophin-null transgenic mice. Proc Natl Acad Sci USA 99:13131–13136. Warth A, Kroger S, Wolburg H (2004) Redistribution of aquaporin-4 in human glioblastoma correlates with loss of agrin immunoreactivity from brain capillary basal laminae. Acta Neuropathol (Berl) 107:311–318. Wolburg H (1995) Orthogonal arrays of intramembranous particles: a review with special reference to astrocytes. J Hirnforsch 36:239 –258. Zubrzycka-Gaarn EE, Bulman DE, Karpati G, Burghes AH, Belfall B, Klamut HJ, Talbot J, Hodges RS, Ray PN, Worton RG (1988) The Duchenne muscular dystrophy gene product is localized in sarcolemma of human skeletal muscle. Nature 333:466 – 469.
(Accepted 23 August 2004) (Available online 6 November 2004)
ANCHORING OF AQUAPORIN-4 IN BRAIN: MOLECULAR MECHANISMS AND IMPLICATIONS FOR THE PHYSIOLOGY AND PATHOPHYSIOLOGY OF WATER TRANSPORT M. AMIRY-MOGHADDAM,* D. S. FRYDENLUND AND O. P. OTTERSEN
2004) and its precise colocalization with the inward rectifying potassium channel, Kir4.1 (Nagelhus et al., 1999) it has been proposed that AQP4 serves to facilitate K⫹ buffering, through an indirect mechanism that remains to be fully resolved (Nagelhus et al., 2004). An involvement of AQP4 in K⫹ buffering could underlie the extracellular volume changes that occur at sites of neuronal activation and would be in line with the view that aquaporins are needed to provide a rapid and selective transport of water through plasma membranes (Agre et al., 2002). An entirely different issue is whether AQP4 is engaged in constitutive water fluxes, which in the brain are likely to be small compared, e.g. to the situation in the kidney where aquaporins have well documented roles (Nielsen et al., 1999). Constitutive water fluxes occur across all internal and external brain surfaces. The net water flow from blood to brain has been estimated to 0.17 l/g/min (Oldendorf, 1970). At the pial surface of the brain the net water flux appears to be directed outwards but the magnitude of this flux remains to be determined. As pointed out elsewhere in this volume (Papadopoulos et al., 2004; Nagelhus et al., 2004), the most conspicuous feature of the AQP4 expression pattern in the CNS is the enrichment of this water channel at the interface between the brain and the associated liquid spaces (blood, subarachnoidal space, and ventricles; Frigeri et al., 1995; Nielsen et al., 1997). At the pial and vascular interfaces the AQP4 pool is concentrated in the subpial and perivascular endfeet, respectively, whereas the ventricle wall shows an accumulation of AQP4 in the basolateral membranes of ependymal cells and in subjacent glial processes (Nielsen et al., 1997). AQP4 is also expressed by brain endothelial cells albeit at a relatively low level (Amiry-Moghaddam et al., 2004). It is a reasonable assumption that the above pools of AQP4 take part in activity dependent and constitutive water fluxes and that AQP4 in this capacity confers an advantage that outweighs the increased vulnerability to edema that the expression of AQP4 entails. The first part of this review will discuss the molecular mechanisms that are thought to be responsible for the anchoring of AQP4 at brain– blood and brain–CSF interfaces. The second part will show how our insight in these anchoring mechanisms has provided a tool for analyses of the functional role of AQP4 in the brain. Notably, we will review data obtained in mice that are depleted of AQP4 at the brain– blood and brain–CSF interfaces following knockout of ␣-syntrophin, a major anchoring protein for brain AQP4 (Neely et al., 2001). Such mice show delayed onset of postischemic and hyponatremic edema, and a reduced
Centre for Molecular Biology and Neuroscience, Institute of Basic Medical Sciences, University of Oslo, POB 1105 Blindern, N-0317 Oslo, Norway
Abstract—Astrocytes show an enrichment of aquaporin-4 (AQP4) in those parts of the plasma membrane that are apposed to pial or perivascular basal laminae. This observation begged the following questions: 1, What are the molecular mechanisms that are responsible for the site specific anchoring of AQP4? 2, What are the physiological and pathophysiological roles of the AQP4 pools at these specialized membrane domains? Recent studies suggest that the site specific anchoring depends on the dystrophin complex. Further, ␣-syntrophin (a member of the dystrophin complex) is required to maintain a polarized expression of AQP4 in the perivascular membranes. Hence transgenic mice deficient in ␣-syntrophin provided a model where the perivascular pool of AQP4 could be removed for assessment of its functional roles. Data suggest that the perivascular pool of AQP4 plays a role in edema formation and that this pool (through its serial coupling with the AQP4 pools in other astrocyte membranes) is involved in Kⴙ siphoning. In the cerebral cortex, the astrocyte membrane domain contacting the pial basal lamina differs from the perivascular membrane domain in regard to the mechanisms for AQP anchoring. Thus deletion of ␣-syntrophin causes only a 50% loss of AQP4 from the former membrane (compared with a 90% loss in the latter), pointing to the existence of additional anchoring proteins. We will also discuss the subcellular distribution and anchoring of AQP4 in the other cell types that express this protein: endothelial cells, ependymal cells, and the specialized astrocytes of the osmosensitive organs. © 2004 IBRO. Published by Elsevier Ltd. All rights reserved. Key words: AQP4, aquaporin, water channels, syntrophin, brain edema, astrocyte.
There is now compelling evidence that AQP4, the predominant aquaporin in the CNS, is functionally active (AmiryMoghaddam and Ottersen, 2003; Nicchia et al., 2003; Solenov et al., 2004) and that it is implicated in the generation and resolution of different forms of experimentally induced brain edema (Manley et al., 2000; Vajda et al., 2002; AmiryMoghaddam et al., 2003a, 2004; Papadopoulos et al., 2004). Recent data have also provided some clues as to the physiological role of AQP4. Notably, judged by functional analyses (Amiry-Moghaddam et al., 2003b; Binder et al., *Corresponding author. Tel: ⫹47-22-85-1495; fax: ⫹47-22-85-1299. E-mail address: [email protected] (M. Amiry-Moghaddam). Abbreviations: AQP4, aquaporin-4; CSF, cerebrospinal fluid; dDAVP, [deamino-cysl, D-Arg8]-vasopressin; MCAO, medial cerebral artery occlusion.
0306-4522/04$30.00⫹0.00 © 2004 IBRO. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.neuroscience.2004.08.049
999
1000
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
Fig. 1. Distribution of proteins thought to be involved in the anchoring of AQP4 at astrocyte endfeet. Immunofluorescence analysis of neocortex (A, C) and cerebellar cortex (B, D) using antibodies to dystrophin (A, B) or ␣-syntrophin (C, D). The antibody to dystrophin recognizes the brain specific isoform Dp 71 (see Fig. 11). Arrows and arrowheads indicate subpial and perivascular labeling, respectively. Electron microscopy shows that the ␣-syntrophin labeling (Fig. 2) and dystrophin labeling (not illustrated) reside in perivascular astrocyte membranes. Scale bars⫽50 m.
capacity for K⫹ buffering (Amiry-Moghaddam et al., 2003a, b, 2004). These findings complement data obtained by use of other transgenic approaches (Manley et al., 2000; Vajda et al., 2002) and suggest that the AQP4 pools that are apposed to brain microvessels are of physiological and pathophysiological importance. The dystrophin molecular complex is expressed at the brain– blood and brain–CSF interfaces and is involved in AQP4 anchoring The dystrophin complex is known to play a crucial role in the molecular organization of the sarcolemma of striated muscle fibers (Zubrzycka-Gaarn et al., 1988). Mdx mice that lack dystrophin show a pronounced loss of AQP4 in striated muscle and were also found to exhibit a perturbed expression of AQP4 in brain (Frigeri et al., 2001). In agreement, biochemical studies have provided evidence of a molecular interaction between brain AQP4 and members of the dystrophin complex (Neely et al., 2001). These data called for an analysis of the expression pattern of dystrophin and associated molecules in brain. Immunofluorescence shows that the major brain dystrophin (Dp-71) is concentrated around brain microvessels
and near the pial surface (Fig. 1A, B). The pattern of labeling obtained with antibodies to Dp-71 is mimicked by antibodies to ␣-syntrophin (Fig. 1C, D), consistent with the idea that ␣-syntrophin is part of the brain dystrophin complexes (cf. Fig. 11). More detailed analysis in the electron microscope revealed that ␣-syntrophin (Fig. 2) and Dp-71 (not shown) are expressed in the perivascular and subpial endfeet of astrocytes, notably in those membrane domains that are directly apposed to the basal lamina. By immunocytochemistry it could be demonstrated that AQP4 is enriched at those sites that express Dp-71 and ␣-syntrophin (compare Fig. 3A, C with Fig. 1) and in the same membrane domains (compare Figs. 4 and 5 with Fig. 2). To test whether the precise colocalization between AQP4 and ␣-syntrophin reflects a coupling between the two molecules we analyzed mice in which the ␣-syntrophin gene had been deleted (referred to as ␣-syn⫺/⫺). These ␣-syn⫺/⫺ mice showed a dramatic reduction of AQP4 immunoreactivity in perivascular astrocyte membranes (Fig. 3–5; Amiry-Moghaddam et al., 2003a,b, 2004; Neely et al., 2001). However, some AQP4 labeling remained, most notably in the granule cell layer of the cerebellum and
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
Fig. 2. ␣-Syntrophin immunogold labeling in cerebellum (A) and neocortex (B) of wild type (control) animals. Immunogold particles signaling ␣-syntrophin are concentrated at the perivascular membrane (double arrowheads) of astrocyte endfeet, while the opposite membrane (i.e. the endfoot membrane domain facing neuropil; double-headed arrow) is devoid of labeling. End, endothelial cells; L, vessel lumen. Scale bars⫽500 nm.
in the zone just subjacent to the pial surface of the neocortex (Amiry-Moghaddam et al., 2004). The existence of such an ␣-syntrophin independent pool explains the mis-
1001
match between the patterns of ␣-syntrophin/Dp-71 labeling and AQP4 labeling in immunofluorescence preparations of wild type animals (compare Fig. 3 and Fig. 1). Immunogold analysis showed that the ␣-syntrophin independent AQP4 pools in the superficial parts of the neocortex and in the granule cell layer of the cerebellum reside in non-endfeet membranes of astrocytes (AmiryMoghaddam et al., 2004; Fig. 4). In fact, the labeling intensity of these membranes was enhanced following deletion of ␣-syntrophin, indicative of a subcellular redistribution of AQP4 from endfeet to non-endfeet membranes (Fig. 4C). That a deletion of ␣-syntrophin causes a mislocalization rather than a net loss of AQP4 is in line with data obtained from quantitative immunoblots (Neely et al., 2001). The mislocalization of AQP4 in ␣-syn⫺/⫺ animals can be viewed as a deficient polarization of AQP4 expression. Hence, in astrocytes that normally express low levels of AQP4 in non-endfeet membranes (the majority) the deletion of ␣-syntrophin will nearly abolish the endfeet labeling. This contrasts with the situation in the minority of astrocytes (located subjacent to the pial surface of the neocortex and in the granule cell layer of the cerebellum) that normally exhibit a sizable pool of AQP4 in non-endfeet membranes: here loss of polarization implies that endfeet membranes retain a significant density of AQP4, on a par with that in non-endfeet membranes. In other words, common for both types of astrocyte is that removal of ␣syntrophin leads to an even distribution of AQP4 across all
Fig. 3. Immunofluorescence labeling of AQP4 in the neocortex (A, B) and cerebellar cortex (C, D) of wild type (A, C) and ␣-syn⫺/⫺ animals (B, D). A comparison between the pattern of AQP4 labeling in A and C with the patterns of dystrophin and ␣-syntrophin labeling in Fig. 1 reveals a clear mismatch. Thus, non-endfeet pools of AQP4 are found in the superficial zone of the cerebral cortex (A) and in the granule cell layer of the cerebellum (C). (The expression in non-endfeet membranes was confirmed by electron microscopy). These pools remain after deletion of ␣-syntrophin (B, D). Electron microscopic analyses reveal that the ␣-syntrophin independent pool of AQP4 is distributed in a non-polarized fashion in astrocyte membranes (Fig. 4). Arrows, cortical surfaces; arrowheads and double arrows, labeled and unlabelled microvessels. Asterisks indicate Purkinje cell layer. Scale bars⫽50 m. From Amiry-Moghaddam et al., 2004.
1002
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
Fig. 4. Quantitative immunogold analysis of AQP4 distribution in wild type compared with ␣-syn⫺/⫺ animals. (A) Deletion of ␣-syntrophin is associated with a loss of the polarized AQP4 distribution in neocortical and cerebellar astrocytes. In wild type (WT) animals, perivascular membranes are enriched in AQP4 compared with endfeet membranes facing neuropil (see Fig. 2 for definition of membrane domains). In syn⫺/⫺ animals these two membrane domains show the same level of immunolabeling. (B) Effect of ␣-syntrophin deletion on AQP4 immunolabeling in subpial endfoot membranes. Same experimental design as in A. In the cerebellum of syn⫺/⫺ animals, the labeling of subpial membranes is similar to that of endfeet membranes facing neuropil. In the neocortex, the polarization of labeling is reduced but not abolished (see text). Labeling intensity is expressed as number of particles per um membrane. (C) Effect of ␣-syntrophin deletion on the AQP4 labeling intensity of non-endfeet membranes. In the superficial zone of the neocortex (i.e. the upper approximately 5 m of layer I) and in the deep part of the cerebellar cortex (i.e. granule cell layer) the labeling of non-endfoot membranes is increased after ␣-syntrophin deletion. No significant changes are seen deeper in the neocortex or more superficially in the cerebellar cortex. In this case, the labeling density was recorded as number of particles per m2, as many membranes could be traced only for short distances. Asterisks, significantly different from other member of the pair (P⬍0.01).
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
1003
Fig. 5. Immunogold analyses reveal two pools of AQP4 at the brain– blood interface. The major pool resides in the perivascular membranes of astrocytes (double arrows) and is depleted after ␣-syntrophin knockout. The minor pool resides in the abluminal (arrows) and adluminal membrane (arrowheads) of endothelial cells (with a predominance in the former) and is independent of ␣-syntrophin (B). (A) Wild type animals; (B) ␣-syn⫺/⫺ animals. From the molecular layer of the cerebellum. E, endothelial cell; L, capillary lumen; P, pericyte. Scale bars⫽500 nm. From Amiry-Moghaddam et al. (2004).
membrane domains (Fig. 4A). There is one exception to this rule: the subpial endfeet in the neocortex retain some polarized expression of AQP4 even after ␣-syntrophin deletion (Fig. 4B; Amiry-Moghaddam and Ottersen, unpublished observations), indicating that the enrichment of AQP4 in this domain does not depend on ␣-syntrophin alone but that another molecule (another syntrophin?) is also involved. It should be noted that deletion of ␣-syntrophin has no effect on the level or distribution of AQP4 in the specialized astrocytes in the supraoptic nucleus and other osmosensitive organs. In wild type animals (Nielsen et al., 1997) as in ␣-syn⫺/⫺ animals the latter astrocytes exhibit very high densities of AQP4 along their entire surfaces, except in those membrane domains that abut on neuronal elements. In conclusion, brain astrocytes can be divided in three categories, based on their AQP4 distribution pattern in wild type and alpha-syn⫺/⫺ animals. The most prevalent type has a modest density of AQP4 along non-endfeet membranes and a strong, ␣-syntrophin dependent enrichment of AQP4 in endfeet membranes. A second type, found superficially in the neocortex and deep in the cerebellar cortex, is similarly endowed with an ␣-syntrophin dependent AQP4 pool at the endfeet. However, it differs from the first type by showing relatively strong non-endfeet labeling that persists or even increases after ␣-syntrophin deletion. The third type is the specialized astrocytes in osmosensitive organs referred to above. The evidence discussed above suggests that ␣syntrophin is the key anchoring protein for the endfeet pools of AQP4. It remains to identify the protein (or proteins) that anchor AQP4 at non-endfeet membranes. The latter membranes are not labeled for Dp-71 (Fig. 1), indicating that dystrophin complexes are not involved (although one must open for the possibility that these mem-
branes express dystrophin complexes that are not recognized by the Dp-71 antibody). AQP4 in endothelial cells and ependymocytes AQP4 is expressed not only in astrocytes but also in endothelial cells (Nagelhus et al., 1998; Amiry-Moghaddam et al., 2004; Fig. 5) and ependymocytes (Nielsen et al., 1997; not shown). The former cells show AQP4 labeling in abluminal as well as in adluminal membranes. The linear gold particle density over in the abluminal membrane domain is about three times as high as that over the adluminal membrane. However, even the abluminal membrane is weakly AQP4 immunopositive compared with the perivascular astrocyte membrane (Fig. 5). Immunocytochemical analyses suggest that the AQP4 pool in ependymocytes is restricted to the basolateral membranes (not shown; Nielsen et al., 1997). Neither the endothelial nor the ependymal pool of AQP4 is significantly decreased after ␣-syntrophin knockout (Amiry-Moghaddam et al., 2004). ␣-Synⴚ/ⴚ animals: a model for assessing the function of the perivascular AQP4 pool The data reviewed above indicate that ␣-syn⫺/⫺ animals should provide a unique model for unraveling the functional role of the AQP4 pool that is enriched in the astrocyte membrane domains at the brain– blood interfaces. As explained above, depletion of ␣-syntrophin leads to a loss of AQP4 from these membrane domains. There is one exception: some AQP4 remains in the endfeet membranes of the minority of astrocytes that contain a sizeable nonendfoot pool of AQP4. Are other membrane molecules than AQP4 affected by deletion of ␣-syntrophin? If so, this could confound the interpretation of data generated by this model. Immunogold analyses did not reveal any significant changes in the
1004
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
Fig. 6. Removal of the perivascular pool of AQP4 by ␣-syntrophin deletion significantly reduces the extent of postischemic edema (22 h following 90 min of MCAO). The perivascular AQP4 immunolabeling (arrowheads), distinct in wild type animals (A), is nearly abolished after ␣-syntrophin knockout (B). ␣-Syn⫺/⫺ animals show preserved labeling of the endfeet membranes facing the neuropil and a sizable increase in endfoot volume (indicated by double arrows in A and B). E, endothelial cells; L, lumen. (C, D) The volume of postischemic edema (unstained areas in TTC stained frontal sections) is smaller in ␣-syn⫺/⫺ animals (D) than in wild type (C). Scale bars⫽500 nm. Modified from Amiry-Moghaddam et al. (2003a).
perivascular expression levels of Kir4.1 or of monocarboxylate, glutamate, or glucose transporters (AmiryMoghaddam et al., 2003b). However, it cannot be ruled out that the deletion of ␣-syntrophin affects one or several unknown endfoot molecules and that this modifies the water transport capacity of perivascular membranes. It also needs to be resolved whether the ␣-syntrophin deletion leads to alterations of the electrical properties of the endfoot membrane. It should be pointed out that the blood– brain barrier remains intact in the ␣-syn⫺/⫺ animals as they show no evidence of an increased uptake of gadolinium chloride from blood to brain (AmiryMoghaddam, unpublished observations). ␣-Synⴚ/ⴚ animals show a decreased vulnerability to postischemic and hyponatremic edema The ␣-syn⫺/⫺ animals showed a partial protection in models of postischemic and hyponatremic edema. The extent of postischemic edema was assessed 22 h after a 90 min period of medial cerebral artery occlusion (MCAO). Quantitative analyses revealed that the volume of the infarct was 50% smaller in the ␣-syn⫺/⫺ animals than in wild type controls (Fig. 6). The ultrastructure of the infarction core, assessed by electron microscopy, was better preserved in the ␣-syn⫺/⫺ than in the wild type animals (AmiryMoghaddam et al., 2003a; Fig. 7). Hyponatremic edema was induced by i.p. injections of distilled water, combined with 0.4 g/kg [deamino-
cysl, D-Arg8]-vasopressin (dDAVP; see legend to Fig. 8 for details). By use of diffusion weighted MRI it could be shown that the onset of fulminant edema (signaled by a precipitous decline in apparent diffusion coefficient) was delayed by 15–20 min in the ␣-syn⫺/⫺ animals compared with controls (Amiry-Moghaddam et al., 2004; Fig. 8). These results are in line with data from other transgenic approaches (Manley et al., 2000; Vajda et al., 2002) and indicate that the perivascular membrane determines the rate of water influx from blood in pathophysiological conditions. In other words, in such conditions the perivascular endfeet rather than the endothelial cells serve as the primary barrier for water uptake. It is not surprising that the removal of perivascular AQP4 provides only a partial protection in models of brain edema: water may pass in the narrow clefts between the perivascular endfeet (which are not joined by tight junctions) and by slow diffusion through the lipid bilayer of the cell membranes (cf. Fig. 10). In the infarction core of wild type animals there was an almost complete loss of perivascular AQP4 22 h after the ischemic insult (Fig. 7), in the face of a preserved immunolabeling for ␣-syntrophin (not shown). This suggests that the coupling between ␣-syntrophin and AQP4 is sensitive to ischemia. In a clinical setting, the occurrence of an ischemia-dependent loss of perivascular AQP4 would be expected to impede the resolution of the edema, which is
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
1005
Fig. 7. Loss of perivascular AQP4 (B) in the infarction core but not in penumbra (C) of wild type animals 22 h following a 90 min period of MCAO. The corresponding area in the contralateral hemisphere served as a control (A). The infarction core in the ␣-syn⫺/⫺ (D) animals reveals better preserved ultrastructure compared with the wild type (B). E, endothelial cells with nucleus (N); L, lumen; arrowheads, perivascular membrane.
likely to depend on an AQP4 mediated outward flux of water. Recent experimental data support the notion (Amiry-Moghaddam et al., 2003a) that AQP4 facilitates clearance of excess water from brain (Papadopoulos et al., 2004; Papadopoulos et al., 2004). This highlights an important issue that must be duly considered in attempts to develop novel therapeutical strategies: as AQP4 permits bidirectional transport of water, an inhibition of the activity of this protein— expected to confer protection in the buildup phase of edema—might prolong the resolution phase. The mechanisms underlying the postischemic loss of perivascular AQP4 are not known. However, the loss occurs following disintegration of the endothelial tight junctions and onset of vasogenic brain edema. It is conceivable that the molecular events in the endothelial lining affect those in the perivascular endfeet, through the bridging connections provided by dystroglycans and laminin/agrin. This would be in line with the data of Nico et al. (2003) who showed show that disruption of BBB by lipopolysaccharide leads to endfoot swelling and loss of perivascular AQP4 (Nicchia et al., 2004).
␣-Synⴚ/ⴚ animals shows a reduced capacity for Kⴙ clearance In our early studies on brain aquaporins (Nielsen et al., 1997; Nagelhus et al., 1998) we predicted that AQP4 could act in concert with those membrane molecules that are engaged in K⫹ buffering. The main arguments were 1, the density profile of AQP4 along the astrocyte membrane (high densities in endfeet compared with nonendfeet membranes) corresponds to the topological variation in K⫹ conductance, as recorded in retinal Müller cells (Newman, 1984); 2, In Müller cells, AQP4 is precisely coexpressed with Kir4.1, a K⫹ channel crucially involved in retinal K⫹ spatial buffering (Nagelhus et al., 1998); 3, The precise coexpression of Kir4.1 and AQP4 agrees with biochemical data indicating that the two molecules may be part of the same molecular complex (Connors et al., 2004; Guadagno and Moukhles, 2004; Fig. 11); 4, evoked activity induces changes in extracellular space volume (a reduction locally and an increase distally) that match the redistribution of excess K⫹ (Niermann et al., 2001). Our hypothesis implies that astrocytes, in terms of activity induced K⫹ and water transport, should be viewed
1006
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
Fig. 8. Removal of the perivascular pool of AQP4 by ␣-syntrophin deletion (syn⫺/⫺) delays the onset of fulminant edema (dashed lines), as recorded by diffusion weighted MRI. Ordinate shows apparent diffusion coefficients (ADC), abscissa shows time (in minutes) after an i.p. injection of distilled water (20% of body weight) and dDAVP (0.4 g/kg). Schematic drawings (left) indicate the effect of ␣-syntrophin knockout on the two pools of AQP4 (endothelial and perivascular) at the brain– blood interface (cf. Fig. 5): the endothelial pool is not affected. E, endothelial cells; L, lumen. Red dots indicate AQP4. From Amiry-Moghaddam et al., 2004.
as cells that have two distinct membrane domains that are serially coupled and that allow influx and efflux, respectively, of K⫹ and water. The larger domain, corresponding to the astrocyte surface that is apposed to the neuropil, mediates K⫹ and water uptake from active neurons. The smaller domain, corresponding to the perivascular and subpial membranes, represents the site where water and K⫹ exit the astrocytes under high neuronal activity. According to this scheme the role of astrocytes as “cellular
siphons” is not restricted to K⫹ (Newman et al., 1984) but includes water as well. The question remained whether the parallel fluxes of K⫹ and water reflect an obligatory coupling between the two. This question was addressed in the ␣-syn⫺/⫺ model (Fig. 9). Consistent with an obligatory coupling, the ␣-syn⫺/⫺ animals (lacking perivascular AQP4 but with a normal complement of Kir4.1) showed a protracted clearance of K⫹ from the termination areas of orthodromically
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
1007
cussed above) that the ␣-syn⫺/⫺ animals exhibit a protracted buildup as well as a protracted clearance of K⫹. A slower buildup implies increased threshold; a slower clearance implies an increased duration and severity of seizures once induced. In support of this idea, AQP4⫺/⫺ animals revealed stronger seizures than wild type animals in the pentylenetetrazole model (Manley et al., 2004). General discussion
Fig. 9. Removal of the perivascular pool of AQP4 by ␣-syntrophin deletion (Syn⫺/⫺) slows the recovery of extracellular K⫹ concentration following repetitive orthodromic stimulation of isolated hippocampal slices. Representative tracings of [K⫹]o signal in CA1 during and after a 10 Hz train delivered to the Schaffer collaterals from wild type (WT) and Syn⫺/⫺ animals shown separately or superimposed in overlay. Calibration bars show K⫹ concentration in mM and time in seconds. From Amiry-Moghaddam et al., 2003b.
activated fiber pathways (Schaffer collaterals and perforant path; Amiry-Moghaddam et al., 2003b). The molecular mechanisms responsible for this obligatory coupling are not resolved (see Nagelhus et al. (2004) this volume, for a tentative model). But our data indicate that the ability of astrocytes to flux water is a prerequisite for efficient K⫹ buffering. Obviously, volume changes of astrocytes are complementary to the extracellular volume changes. It is quite probable that the volume changes of the extracellular space are essential to provide a K⫹ concentration gradient that is steep enough to allow for a sufficiently rapid redistribution of this ion. This line of reasoning would explain why, in ␣-syn⫺/⫺ animals, there is not only a slower clearance of K⫹ but also a slower buildup of [K⫹]o during orthodromic activation (Fig. 9). The slow buildup could simply reflect an impaired shrinkage of the local extracellular space (due to loss of transglial, AQP4 mediated water transport). In the absence of extracellular volume reduction the released K⫹ will be distributed over a larger volume than in control animals. This concept would be compatible with our finding that the peak K⫹ concentration tends to be slightly lower in ␣-syn⫺/⫺ knockouts than in controls (Amiry-Moghaddam et al., 2003b). If AQP4 is involved in the clearance of K⫹ one would expect changes in its expression to affect cell excitability. Two recent studies have addressed this issue (AmiryMoghaddam et al., 2003b; Binder et al., 2004). AmiryMoghaddam et al. (2003b) showed that removal of perivascular AQP4 (by ␣-syntrophin deletion) enhanced the severity of hyperthermia-induced seizures. Binder et al. (2004) (based on pentylenetetrazol induced seizures) reported that seizure threshold was increased in AQP4⫺/⫺ animals. These two data sets, contradictory at first glance, can be reconciled by taking into account our findings (dis-
With the discovery of the aquaporins, water flow in the brain cannot any longer be discussed in terms of simple diffusion across lipid bilayers but must now be understood as a specific and facilitated flux of water through the hydrophilic interior of specialized membrane proteins. On this background the mechanisms that serve to anchor AQP4 at the plasma membrane are important, as they dictate the subcellular expression pattern of this molecule and hence direct the microscopic and macroscopic water fluxes in the brain. A discussion of the direction and magnitude of water fluxes in the brain must be based on three seminal observations: 1, water flux through aquaporins is bidirectional and is driven by osmotic and hydrostatic forces (Agre et al., 2002): 2, AQP4 —the predominant aquaporin in brain—is expressed in astrocytes and not in neurons; 3, most astrocytes show a remarkable polarization when it comes to AQP4 distribution, implying that some membrane domains are specialized for rapid transmembrane water exchange. However, the latter observation must not overshadow the fact that even in astrocytes with a polarized expression of AQP4 in endfeet membranes (such as the Müller cells), low but significant levels of AQP4 are found also in the nonendfeet membranes. In fact, in the case of the Müller cells the quantitative immunogold data suggest that the total amount of AQP4 in the nonendfeet membranes (which constitute approximately 90% of the total surface) corresponds to the total amount of AQP4 in endfeet membranes (which constitute the remaining approximately 10%; Nagelhus et al., 1999). This is consistent with the idea that astrocytes possess two separate and serially coupled membrane domains with regard to water transport (Nagelhus et al., 2004) as well as K⫹ conductance (Newman, 1984). In fact, as discussed above, evidence is accumulating to suggest that astrocytes mediate parallel fluxes of K⫹ and water in conditions which require buffering of extracellular K⫹. In other words, the spatial redistribution of K⫹ is associated with a similar redistribution of water. During high neuronal activity K⫹ and water may exit preferentially through the endfeet membranes, or (in the case of the astrocytes in the superficial part of the neocortex) through the nonendfeet membranes that form AQP4 enriched lamellae just subjacent to the pial surface (Niermann et al., 2001). In the eye (Nagelhus et al., 1998) and at the neocortical surface the AQP4 (Nielsen et al., 1997) enriched membrane domains are directly or indirectly apposed to liquid spaces (vitreous humor and subarachnoidal space) that act as efficient sinks for water and K⫹. In the brain neuropil the AQP4 enriched membrane domains face microvessels, as expected if anchoring depends on laminin and agrin (central constituents of subpial as well as perivascu-
1008
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
Fig. 10. Diagram showing the possible involvement of AQP4 (indicated in dark blue) in activity dependent and constitutive water flow in the neocortex of wild type and ␣-syn⫺/⫺ animals. Arrows show putative direction of water flux. (A) Activity dependent fluxes in wild type animals: Siphoning of K⫹ is associated with water uptake in astrocytes and subsequent water release through the perivascular and subpial membranes that are strongly enriched in AQP4. Constitutive water fluxes in wild type animals: There is a net secretion of water from the capillary bed and a net production of water from the energy metabolism. The enrichment of AQP4 in endfeet and non-endfeet membranes at the cortical surface serves to facilitate the transport of water into the subarachnoidal space, which acts as a sink. The activity dependent efflux of water through perivascular endfeet is likely to join the constitutive flux of water (asterisk) as osmotic gradients may not favor a net transendothelial water flux from brain to blood (but see text). (B) Possible scenario of changes in water transport after ␣-syntrophin knockout. The distinct swelling of perivascular and subpial endfeet (cf. Fig. 6) may reflect a reduced capacity for water efflux through the endfeet membranes.
lar basal laminae; Guadagno and Moukhles, 2004; Warth et al., 2004). Water that exits the perivascular membrane domains under high neuronal activity might be reflected to join the constitutive flux of water from blood to brain (see below and Fig. 10). It is still unclear if and to what extent brain AQP4 is involved in constitutive water fluxes. In general, aquaporins are essential primarily when there is a need for rapid movement of water across a plasma membrane (Agre et al., 2002). However, it should be recalled that AQP4 provides not only high transport capacity but also high selectivity (Amiry-Moghaddam and Ottersen, 2003). Constitutive water movement by way of cotransporters will inevitably cause redistribution of the cotransported compounds while AQP4 mediates transport of water alone. An involvement of AQP4 in constitutive water flux is consistent with the finding that absence of perivascular AQP4 (in mdx or syn⫺/⫺ mice) is associated with a distinct swelling of the endfeet (Amiry-Moghaddam et al., 2003a; Fig. 6). Control experiments in syn⫺/⫺mice indicate that
these changes are not secondary to a disruption of the blood– brain barrier or to changes in the expression level of known water cotransporters in perivascular membranes. An obvious source of the excess water is the energy metabolism, as water is formed when glucose is broken down. If one assumes that there is normally a constitutive efflux of water across the perivascular membranes, and that this manifests itself in swelling once the perivascular AQP4 pool is removed, then one is confronted with the problem that the direction of flux does not match the direction of net flux across the blood– brain interface. The latter flux is thought to be directed inwardly (Oldendorf, 1970, 1972). However, the concept of a net flux from blood to brain is based on an experimental design (recovery of systemically administered tritiated water; Oldendorf, 1970) that would fail to reveal any heterogeneities in the vascular bed. Thus the possibility remains that influx only occurs in one segment of the vascular bed and that this influx is balanced (or exceeded; Greitz, 2004) by an efflux in an-
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
1009
Fig. 11. Diagram showing the presumed molecular basis for the enrichment of AQP4 in perivascular and subpial membranes. The dystrophin complex is anchored to the basal lamina through laminin and agrin and binds AQP4 (and possibly other molecules with a C-terminal SXV sequence like Kir4.1) by way of ␣-syntrophin (␣-syn) or other syntrophins (syn). The coupling between AQP4 and syn may be indirect (indicated by double arrows). We assume that ␣-syn is crucially involved in the anchoring of the M23 isoform of AQP4 (as M23 is the isoform that forms orthogonal arrays of proteins, particularly enriched in endfeet membranes; Wolburg, 1995). Other syn may be primarily involved in the anchoring of the M1 isoform (which does not form OAPs; Furman et al., 2003). It should be understood that the identity and stoichiometry of the SXV-containing membrane molecules (SSV for AQP4; SNV for Kir4.1) may vary according to region, cell type, and membrane domain. ␣-DG and -DG, ␣ and  dystroglycan; Dp-71, major dystrophin isoform in brain. The PDZ binding domains of syn are indicated. H1 indicates the coiled-coil motif interaction between Dp71 (the major dystrophin isoform in brain) and ␣-dystrobrevin (Sadoulet-Puccio et al., 1997).
other segment. In any case, while transendothelial water flux and flux through perivascular membranes are serially coupled in pathophysiological states (Manley et al., 2000; Vajda et al., 2002; Amiry-Moghaddam et al., 2003a, 2004) this is not necessarily true in all vascular segments at the low flux rates that occur constitutively in physiological conditions. It is possible that the efflux through the perivascular membrane is balanced by a paracellular movement of water in the opposite direction (see above); alternatively, there may be a net movement of fluid along the basal lamina that separates the endfeet from the endothelial cells. This fluid could eventually be drained into the Virchow-Robin spaces which are known to sustain an “active” transport of water due to the piston-like effect of the pulsatile CSF waves (Rennels et al., 1985). The capacity for water transport along the paracapillary route is not known.
REFERENCES Agre P, King LS, Yasui M, Guggino WB, Ottersen OP, Fujiyoshi Y, Engel A, Nielsen S (2002) Aquaporin water channels-from atomic structure to clinical medicine. J Physiol 542:3–16. Amiry-Moghaddam M, Otsuka T, Hurn PD, Traystman RJ, Haug FM, Froehner SC, Adams ME, Neely JD, Agre P, Ottersen OP, Bhardwaj A (2003a) An alpha-syntrophin-dependent pool of AQP4 in astroglial end-feet confers bidirectional water flow between blood and brain. Proc Natl Acad Sci USA 100:2106 – 2111. Amiry-Moghaddam M, Ottersen OP (2003) The molecular basis of water transport in the brain. Nat Rev Neurosci 4:991–1001. Amiry-Moghaddam M, Williamson A, Palomba M, Eid T, de Lanerolle NC, Nagelhus EA, Adams ME, Froehner SC, Agre P, Ottersen OP (2003b) Delayed K⫹ clearance associated with aquaporin-4 mislocalization: phenotypic defects in brains of alpha-syntrophin-null mice. Proc Natl Acad Sci USA 100:13615–13620.
1010
M. Amiry-Moghaddam et al. / Neuroscience 129 (2004) 999 –1010
Amiry-Moghaddam M, Xue R, Haug FM, Neely JD, Bhardwaj A, Agre P, Adams ME, Froehner SC, Mori S, Ottersen OP (2004) Alpha-syntrophin deletion removes the perivascular but not endothelial pool of aquaporin-4 at the blood-brain barrier and delays the development of brain edema in an experimental model of acute hyponatremia. FASEB J 18:542–544. Binder DK, Oshio K, Ma T, Verkman AS, Manley GT (2004) Increased seizure threshold in mice lacking aquaporin-4 water channels. Neuroreport 15:259 –262. Connors NC, Adams ME, Froehner SC, Kofuji P (2004) The potassium channel Kir4.1 associates with the dystrophin-glycoprotein complex via alpha-syntrophin in glia. J Biol Chem 279:28387–28392. Frigeri A, Gropper MA, Umenishi F, Kawashima M, Brown D, Verkman AS (1995) Localization of MIWC and GLIP water channel homologs in neuromuscular, epithelial and glandular tissues. J Cell Sci 108(Pt 9):2993–3002. Frigeri A, Nicchia GP, Nico B, Quondamatteo F, Herken R, Roncali L, Svelto M (2001) Aquaporin-4 deficiency in skeletal muscle and brain of dystrophic mdx mice. FASEB J 15:90 –98. Furman CS, Gorelick-Feldman DA, Davidson KG, Yasumura T, Neely JD, Agre P, Rash JE (2003) Aquaporin-4 square array assembly: opposing actions of M1 and M23 isoforms. Proc Natl Acad Sci USA 100:13609–13614. Greitz D (2004) The hydrodynamic hypothesis versus the bulk flow hypothesis. Neurosurg Rev 27:299 –300. Guadagno E, Moukhles H (2004) Laminin-induced aggregation of the inwardly rectifying potassium channel, Kir4.1, and the waterpermeable channel, AQP4, via a dystroglycan-containing complex in astrocytes. Glia 47:138 –149. Manley GT, Binder DK, C Marios, Papadopoulos MC, Verkman AS (2004) New insights into water transport and edema in the central nervous system from phenotype analysis of aquaporin-4 null mice. Neuroscience 129:983–991. Manley GT, Fujimura M, Ma T, Noshita N, Filiz F, Bollen AW, Chan P, Verkman AS (2000) Aquaporin-4 deletion in mice reduces brain edema after acute water intoxication and ischemic stroke. Nat Med 6:159 –163. Nagelhus EA, Mathiisen TM, and Ottersen OP (2004) Aquaporin-4 in the central nervous system: cellular and subcellular distribution and coexpression with Kir 4.1. Neuroscience 129:905–913. Nagelhus EA, Horio Y, Inanobe A, Fujita A, Haug FM, Nielsen S, Kurachi Y, Ottersen OP (1999) Immunogold evidence suggests that coupling of K⫹ siphoning and water transport in rat retinal Muller cells is mediated by a coenrichment of Kir4.1 and AQP4 in specific membrane domains. Glia 26:47–54. Nagelhus EA, Veruki ML, Torp R, Haug FM, Laake JH, Nielsen S, Agre P, Ottersen OP (1998) Aquaporin-4 water channel protein in the rat retina and optic nerve: polarized expression in Muller cells and fibrous astrocytes. J Neurosci 18:2506 –2519. Neely JD, Amiry-Moghaddam M, Ottersen OP, Froehner SC, Agre P, Adams ME (2001) Syntrophin-dependent expression and localization of aquaporin-4 water channel protein. Proc Natl Acad Sci USA 98:14108 –14113. Newman EA (1984) Regional specialization of retinal glial cell membrane. Nature 309:155–157. Newman EA, Frambach DA, Odette LL (1984) Control of extracellular potassium levels by retinal glial cell K⫹ siphoning. Science 225:1174 –1175. Nicchia GP, Nico B, Camassa LMA, Mola MG, Loh N, Dermietzel R, Spray DC, Svelto M, Frigeri A (2004) The role of aquaporin-4 in the
blood– brain barrier development and integrity: studies in animal and cell culture models. Neuroscience 129:935–945. Nicchia GP, Frigeri A, Liuzzi GM, Svelto M (2003) Inhibition of aquaporin-4 expression in astrocytes by RNAi determines alteration in cell morphology, growth, and water transport and induces changes in ischemia-related genes. FASEB J 17:1508 –1510. Nico B, Frigeri A, Nicchia GP, Corsi P, Ribatti D, Quondamatteo F, Herken R, Girolamo F, Marzullo A, Svelto M, Roncali L (2003) Severe alterations of endothelial and glial cells in the blood-brain barrier of dystrophic mdx mice. Glia 42:235–251. Nielsen S, Kwon TH, Christensen BM, Promeneur D, Frokiaer J, Marples D (1999) Physiology and pathophysiology of renal aquaporins. J Am Soc Nephrol 10:647– 663. Nielsen S, Nagelhus EA, Amiry-Moghaddam M, Bourque C, Agre P, Ottersen OP (1997) Specialized membrane domains for water transport in glial cells: high-resolution immunogold cytochemistry of aquaporin-4 in rat brain. J Neurosci 17:171–180. Niermann H, Amiry-Moghaddam M, Holthoff K, Witte OW, Ottersen OP (2001) A novel role of vasopressin in the brain: modulation of activitydependent water flux in the neocortex. J Neurosci 21:3045–3051. Oldendorf WH (1970) Measurement of brain uptake of radiolabeled substances using a tritiated water internal standard. Brain Res 24:372–376. Oldendorf WH (1972) Cerebrospinal fluid formation and circulation. Prog Nucl Med 1:336 –358. Papadopoulos MC, Saadoun S, Binder DK, Manley GT, Krishna S, Verkman AS (2004) Molecular mechanisms of brain tumor edema. Neuroscience 129:1011–1020. Papadopoulos MC, Manley GT, Krishna S, Verkman AS (2004) Aquaporin-4 facilitates reabsorption of excess fluid in vasogenic brain edema. FASEB J 18(11):1291–1293. Rennels ML, Gregory TF, Blaumanis OR, Fujimoto K, Grady PA (1985) Evidence for a paravascular fluid circulation in the mammalian central nervous-system, provided by the rapid distribution of tracer protein throughout the brain from the subarachnoid space. Brain Res 326:47– 63. Sadoulet-Puccio HM, Rajala M, Kunkel LM (1997) Dystrobrevin and dystrophin: an interaction through coiled-coil motifs. Proc Natl Acad Sci USA 94:12413–12418. Solenov E, Watanabe H, Manley GT, Verkman AS (2004) Sevenfoldreduced osmotic water permeability in primary astrocyte cultures from AQP-4-deficient mice, measured by a fluorescence quenching method. Am J Physiol Cell Physiol 286:C426 –C432. Vajda Z, Pedersen M, Fuchtbauer EM, Wertz K, Stodkilde-Jorgensen H, Sulyok E, Doczi T, Neely JD, Agre P, Frokiaer J, Nielsen S (2002) Delayed onset of brain edema and mislocalization of aquaporin-4 in dystrophin-null transgenic mice. Proc Natl Acad Sci USA 99:13131–13136. Warth A, Kroger S, Wolburg H (2004) Redistribution of aquaporin-4 in human glioblastoma correlates with loss of agrin immunoreactivity from brain capillary basal laminae. Acta Neuropathol (Berl) 107:311–318. Wolburg H (1995) Orthogonal arrays of intramembranous particles: a review with special reference to astrocytes. J Hirnforsch 36:239 –258. Zubrzycka-Gaarn EE, Bulman DE, Karpati G, Burghes AH, Belfall B, Klamut HJ, Talbot J, Hodges RS, Ray PN, Worton RG (1988) The Duchenne muscular dystrophy gene product is localized in sarcolemma of human skeletal muscle. Nature 333:466 – 469.
(Accepted 23 August 2004) (Available online 6 November 2004)