Aquaporin-4 in Astroglial Cells in the CNS and Supporting Cells of Sensory Organs—A Comparative Perspective

Aquaporin-4 (AQP4) is a classical, water-specific aquaporin. Although first discovered in the brain [1] AQP4 was subsequently localized in many other organs [2,3]. Playing a role in fast water transport through epithelia seems to be its prevalent function in the body. This includes epithelia in the lung, salivary glands, intestinal tract, to name just a few [4,5,6] (see also other articles in this issue). So far, the reports on its cellular distribution showed its presence exclusively in the basolateral membrane domain whereas other aquaporins have also been localized to the apical membrane domain. Besides epithelial cell expression, AQP4 occurs also on skeletal muscle.

In the vertebrate central nervous system (CNS), AQP4 is expressed by astrocytes or astroglial cells, and ependymal cells. Astrocytes are one of the two main types of macroglial cells in the vertebrate CNS, ependymal cells are lining the brain ventricles and are sometimes considered a separate class of glial cells However, they share some features with astrocytes, one of them is the expression of AQP4. In sensory epithelia, AQP4 has been found in supporting cells.

Up to about 30 years ago, glial cells were considered glue or structural support cells (“Nervenkitt”) which had little to do with information processing [7]. When astrocytes were recognized as being more than structural support, their enzymatic activity and ion channels were studied and it was established that they are necessary for transmitter recycling, nutritional support, and potassium removal from the extracellular space (trophic, metabolic, and synaptic functions) [8,9].

The history of the discovery of aquaporins in the mammalian brain is curiously remarkable in the respect that the main water channel in the CNS, aquaporin-4 (AQP4), was investigated on ultra-morphological grounds before the first reports on the existence of specific water channels were published. So-called orthogonal arrays of particles (OAPs) were found in membranes of astrocytes by the freeze fracture method employing electron microscopy [10,11]. These OAPs occurred especially in high densities at places where astrocytes faced the basal lamina and therefore extracellular fluid spaces as is the case around blood vessels (Figure 1a,b) and under the pia mater, at the superficial glia limitans [12]. Before it became known that OAPs consist of AQP4, there were many speculations on their functional involvement ranging from ion channels [13] to regeneration, and even a role for water transport, reviewed in Wolburg [12]. The suggestion for a role in regeneration resulted from comparisons of various brain structures across different species, reviewed in [14]. Within the brain, OAPs were also found in membranes of ependymal cells, the cells that line the ventricular surface [15], and of course in many cells outside the CNS [16].

At this time, the involvement of OAPs in water transport was speculated on just by few researchers [17] and not accepted in the research community, nor was there good evidence for it. In 1986, Benga and co-workers published evidence for the existence of such a channel [18], and in 1992, the first water channel was cloned by the Agre group [19], and soon after water channels were found in the CNS [1], and as the forth water channel sequenced it was named aquaporin-4 (AQP4). For the history of the discovery of water channels, see Benga [20]. The first indication that OAPs were formed by water channels originated from studies of AQP4 knock-out mice that did not form any OAPs [21]. Direct evidence in support of the idea that the particle arrays consisted of water channel proteins was then provided by freeze fracture immunolabeling [22]. This notion is now generally accepted.

Since then, aquaporins in the vertebrate brain have been suggested to be involved in many physiological processes such as waste removal and fine-tuning of potassium homeostasis, reviewed in [23] some of which we will discuss below.

In this review, we focus on AQP4 as the main water channel occurring in the vertebrate brain, and its expression in the brain and in neuro-sensory structures, more specifically the retina, inner ear, olfactory mucosa, and taste buds. In addition, we will include data on non-mammalian AQP4 expression to arrive at an evolutionary perspective.

Like other aquaporins, AQP4 occurs as tetramers [25] with two major isoforms, M1 and M23 differing in their initiation sites at Met1 or Met23, respectively at the N terminus [1], reviewed in Papadopoulos [26]. The combined expression of both isoforms determines the formation of arrays as multiples of heterotetramers [27,28,29]. When CHO-K1 cells were transfected with M1 isoform of AQP4, no arrays formed in the membranes, whereas transfecting with the M23 isoforms led to large raft-like lattices. The expression of both isoforms after transfection resulted in OAPs that resembled those of astrocytes in vivo [30], suggesting that both isoforms coexist in one array [31]. With the development of superresolution techniques in light microscopy, some attempts have been undertaken to study array formation by localization microscopy: Transfecting cells with fluorescent-tagged AQP4 M1 and M23 isoforms showed that the longer M1 isoform was more mobile in the cell membrane than the shorter M23, suggesting a stronger interaction at the N-terminus of the latter [32]. Two-color dSTORM indicated that M1 and M23 co-assemble in OAPs with a M1-enriched periphery surrounding a M23-enriched core [33]. Further studies suggested that M1-M23 aggregation state has effects on the AQP4 mobility in the plasma membrane [29]. Although many reports have scrutinized the expression patterns and conditions of square array and higher order structures of AQP4, the precise function of array formation remains elusive [16]. Suggestions range from ideal high density packaging to adhesion function, gas transport, and anchoring at the basement membrane, summarized in [23].

AQP4 has been described to be expressed in epithelial cells of many organs, including kidney, intestine, salivary glands, sensory organs, and skeletal muscle [3]. In all cases of epithelial cell expression, AQP4 was found to be localized in the basolateral membrane domain [14,34]. This general principle, however, is not followed in the specialized epithelium of the cochlea, nor in the neuroepithelium derived CNS. As will be outlined and discussed below, two of inner ear supporting cell types express AQP4 exclusively in the basal membrane domains, and CNS astrocytes show high AQP4 presence at their endfeet membranes.

In the mammalian brain, antibody studies as well as freeze fracture EM revealed a high density of AQP4 expression on astrocytic endfeet membranes at the perivascular and the superficial glia limitans (Figure 1c,d). This expression pattern of AQP4, and other channel expression such as the potassium channel Kir4.1, indicates that astrocytes are highly polarized cells. This is not surprising in the sense that epithelial cells are polarized cells, and astrocytes are derived from neuroepithelial cells during development [35]. Yet, due to their star-shaped morphology, there are no defined apical and basolateral membrane domains in astrocytes, nor is there a sub-apical tight junction. The perivascular endfeet are part of the neurovascular unit and directly behind the actual blood-brain barrier (BBB) formed by the endothelial cell tight junctions [36].

Several aspects may shed light on the function of AQP4 distribution in the mammalian brain. Firstly, AQP4 expression does not start in a polarized pattern of cellular distribution, but is spread over cell membrane of radial glial cells and early astrocytes during cortical development [37]. The beginning of polarized expression coincides roughly with the appearance of OAPs. Several studies have established that the microenvironment and molecules of the extracellular matrix such as the dystrophin associated protein complex and agrin play a decisive role in AQP4 polarization and OAP formation [35,38,39,40,41,42]. Secondly, in ependymal cells AQP4 is expressed in basolateral membrane domains (Figure 1c) but not in the specialized ependymal cells covering the stroma and blood vessels of the choroid plexus. These latter cells form subapical tight junctions (the basis for the blood-cerebrospinal fluid barrier) and express aquaporin-1 in the apical membrane but no AQP4 [14,16]. This is similar to circumventricular organs where the blood-brain barrier is shifted from the endothelial to the glial barrier. Thirdly, in response to pathological conditions, such as traumatic injury, stroke, or tumor, astrocytes are reactive and often upregulate intermediate filaments, and also AQP4. This increase correlates with a loss in polarized AQP4 expression and a reduction in OAPs [43].

The two major macroglial cells in the CNS, astroglial and oligodendroglial cells, are present in most vertebrates, including teleost fish. Unlike in mammals however, the astroglial cells in fish do not show the typical star-shaped morphology; they rather have long processes and often contact the ventricle and the pial surface of the brain [73,74]. They are referred to as radial glial cells, present for example in the optic tectum and can at the same time be considered ependymal cells since they contact the ventricle, and no other specialized cuboidal cells lining the ventricle exist like in the mammalian brain [75]. These adult radial glial cells form junctional complexes at the ventricular surface [74] and thus retain epithelial-like polarity. According to the infolding of the neural tube during development, the side facing the ventricle represents the apical side, the pial side, the basal or basolateral surface.

When stained for AQP4 by immunohistochemistry, radial glial cells in the zebrafish brain were positive but did not show the polarized distribution as it is known from mammalian astrocytes [74]. OAPs occurring in high density in mammalian perivascular astrocytic endfeet have never been found in the brain of fish. Thus, the expression pattern of AQP4 on radial glial cells and the lack of OAPs in fish brain resemble the situation of radial glial cells during cortical development. However, the radial glial (Müller) cells of the goldfish (Carassius auratus) retina have been reported to form OAPs at their endfeet facing the vitreous humor [76], a finding we could recently confirm in the retina of cichlid fish Astatotilapia burtoni (Figure 3). Thus, the formation of OAPs by AQP4 on glial cells is a trait that evolved prior to tetrapod evolution but is not prevalent in the fish brain.

In amphibians and reptiles, radial glial types predominate, although smaller astrocyte-like cells with multiple processes might have evolved several times during vertebrate evolution. Aquaporins were present at the beginning of deuterostome and vertebrate evolution. This includes AQP4 as one of the classical aquaporins [77]. Besides in the teleost CNS, the expression of AQP4 in the brain of sharks has been documented [78]. However, information on the distribution of AQP4 in the brain of other fish groups, reptiles, and amphibians is scarce. Freeze-fracture data from the 1980–1990s (summarized in [16]) revealed OAPs on retinal Müller cells in all major vertebrate groups. In amphibians, Müller cells of urodeles formed OAPs, whereas those of anurans did not. In the lizard thoracic spinal cord OAPs were present, but the caudal spinal cord was OAP-negative. Generally, the brains of elasmobranchs, hagfish, or lamprey are completely devoid of OAPs.

In birds, astrocytes are formed and show more or less the mammalian pattern of AQP4 distribution in brain [79,80] and retina [81].

All major groups of eukaryotic organisms show expression of water channels [82]. Within vertebrates, aquaporin 4 has been reported to occur in the gills of the jawless hagfish [83] confirming much earlier reports on OAPs [84], and in many tissues of sharks including kidney, gill, and brain [85]. In the gills of hagfish, there was clearly basolateral expression of AQP4, but a polarized expression on astroglial processes as seen in mammals has not been demonstrated. It is noteworthy that a glial-based BBB was common in early vertebrate brain evolution [86].

Although many functions have been suggested for AQP4 besides water transport such as facilitating cell migration and cell adhesion, the control of water balance and homeostasis is likely its main role. In the evolutionary context, it is interesting that in the sarcopterygean lineage leading to tetrapods, an ancient aquaporin gene cluster evolved and diverged into paralogous forms of AQP2, -5, or -6 [77]. This presumably enabled increased water conservation necessary for survival in terrestrial habitats. In the actinopterygian lineage, a genome duplication presumably occurred during early teleost evolution. Thus, 18 members of the aquaporin gene family were identified in the zebrafish [87], more than in mammals where 13 AQPs are usually found (numbered AQP0-12) [82]. Regarding AQP4, two AQP4 gene sequences have been predicted for the cichlid fish Astatotilapia burtoni, and we have recently confirmed the expression of both genes in brain and retina of this fish (unpublished observations). For further aspects on the evolution of aquaporin genes and their occurrence in vertebrate phylogeny, see [77].

The high density of AQP4 around blood vessels in the mammalian brain is thought to facilitate water flow when potassium is released from the endfeet to extracellular space and blood vessels (potassium siphoning) thus serving ionic homeostasis [88,89]. In addition, a convective water flow from perivascular arterial spaces through nervous system parenchyma has been suggested to play a major role in waste removal [90,91]. This is, in part, a paracellular fluid movement but is thought to be facilitated by AQP4 channels and their polarized distribution. Impairment of AQP4 leads to accumulation of waste products in the brain, which might be one of causes for neurodegenerative diseases. Re-distribution of AQP4 in brain diseases such as tumors can lead to cytotoxic edema. Thus, the polarized distribution in the mammalian brain is essential for water balance and ionic equilibration. As pointed out above, in fish brain, the shape and AQP4 distribution of astroglial cells differs from the situation in mammalian astrocytes. Since there is a lack or low degree of polarized distribution of water channels, the fluid movements in the fish brain are likely different from the mammalian brain. This is corroborated by the maintained radial morphology of astroglial cells in the fish brain suggesting an apical-basal (i.e., surface-ventricular) dominated polarity rather than a perivascular-parenchymal polarity. Thus, a bulk convective water flow as suggested for the mammalian brain seems unlikely in the fish brain.

A strict polarity with a basolateral AQP4 expression is maintained in the supporting cells of the sensory epithelia of the olfactory mucosa. A more differentiated expression is seen in the inner ear where inner sulcus cells and Claudius cells show a restricted basal expression. This AQP4 expression is likely providing a water balance to control together with ion channels the ionic concentration surrounding the sensory cells, thereby enabling them to transduce sensory stimuli into voltage changes. Interestingly, in the mammalian retina, the AQP4 polarization on radial glial processes around blood vessels and expression of scaffolding proteins is less pronounced than in astrocytes. In mammalian astrocytes, there is a strong polarity reflected in AQP4 expression but a basolateral domain cannot be defined. An overview of the AQP4 expression patterns is presented in Table 1.

Derouiche et al. [92] proposed three membrane domains for Müller cells and astrocytes, ventricular, neuronal, and basal lamina facing, based on distinct molecule localizations including AQP4. Given the heterogeneity of astrocytes in the mammalian brain [93], even more differentiated domain subdivisions might be anticipated.

In summary, a polarized expression of AQP4 is part of a control mechanism for a stable ionic environment. This is beneficial for the neighboring sensory or nerve cells and neuronal signaling. This is easily achieved in epithelial-like structures of sensory organs, i.e., the basolateral membrane domain of supporting cells. In the mammalian brain, this polarity is shifted to astrocytic endfeet at the endothelial and superficial glia limitans and serves as a part of the neurovascular unit to efficiently maintain homeostasis.