Evidence of aquaporin involvement in human central pontine myelinolysis

All lesions identified in CNS tissues examined from the 6 CPM patients exhibited active demyelination (Figure 1a, e, i; Figure 2g, h; Figure 3c). Lesions affected the base of the central pons symmetrically (Figure 1a, e, i; Figure 2a; Figure 3a). There was relative preservation of axons and neurons (Figure 2l, m) and, to some extent, of myelin (Figure 1e, i; Figure 2c, g, h; Figure 3c). All lesions were heavily infiltrated by activated macrophages (Figure 2a, b; Figure 3a, b) containing abundant myelin degradation products, a hallmark of active demyelination (Figure 2g, h; Figure 3c). There was a mild degree of T lymphocyte infiltration around vessels and in the parenchyma (Figure 2n; Figure 3o), and evidence of axonal swelling (Figure 2l) and oligodendrocyte apoptosis (Figure 2o). No deposits of complement activation products were found.

Nonlesional pontine white matter tracts were frequently vacuolated (Figure 4d-i), compatible with intramyelinic edema (Figure 4e, inset), and were infiltrated by microglia (Figure 4i). However, myelin was preserved (Figure 4b), and AQP4 (Figure 4f), AQP1 (Figure 4g) and GFAP (Figure 4h) immunoreactivities were of normal intensity. In normal appearing gray and white matter there was evidence of prominent microglial activation (Figure 4j, k), rare clusters of small macrophages (Figure 4j), and ballooned and red ischemic neurons (Figure 4l, m).

Astrocytes in the normal pons expressed both AQP1 and AQP4 (Figure 5). No differences were observed between the female and male controls.

Astrocytes in the pontine white matter tracts expressed preferentially, but not exclusively, AQP1 (Figure 5a-c). AQP1 in the white matter was highly concentrated on the membrane of the astrocytic cell bodies and in all astrocytic foot processes, including those abutting blood vessels (Figure 5d). AQP4 immunoreactivity in the white matter was mainly observed in astrocytic foot processes abutting blood vessels (Figure 5f).

Astrocytes in the pontine grey matter nuclei preferentially, but not exclusively, expressed AQP4 (Figure 5a-c). AQP4 positive astrocytic processes enveloped the blood vessels as well as the neuronal cell bodies (Figure 5g). AQP1 immunoreactivity in the pontine nuclei was preferentially observed around blood vessels and around some, but not all neurons (Figure 5e).

Loss of AQP1 and AQP4 was evident within demyelinating lesions in four of the six cases (Table 1, Figure 1b, f, c, g; Figure 2d, e). Areas with loss of AQPs (Figure 1b, c, f, g; Figure 2d, e) corresponded to areas of demyelination (Figure 1a, e; Figure 2d, e) and macrophage infiltration (Figure 2a, b), and AQP immunoreactivity was enhanced at the lesion borders (Figure 4a, b). All four cases demonstrated a loss of the characteristic perivascular distribution of AQP4 (Figure 2i) and AQP1 (Figure 2j) within the area of active demyelination. Beyond the demyelinating lesions, AQP1 and AQP4 immunoreactivities were preserved (Figure 1b, c, f, g). Lesions uniformly retained GFAP-immunoreactive astrocytes, despite AQP1 and AQP4 loss (Figure 2f, k), but astrocyte numbers were reduced. Lesional astrocytes were smaller and had fewer and shorter processes (Figure 2k) than perilesional astrocytes (Figure 4c), giving the impression of GFAP loss within lesions when sections were observed under low magnification (Figure 1d, h). Gemistocytes were rare within lesions, but were abundant at lesion borders (Figure 4a-c). The nuclear condensation and fragmentation typical of apoptosis was not observed in either lesional or non-lesional astrocytes. Occasional macrophages contained GFAP-positive material (Figure 2k, lower inset), and Rosenthal fibers were present (Figure 2p).

In two of the six cases astrocytes within demyelinating lesions exhibited enhanced AQP1 and AQP4 immunoreactivity (Table 1; Figure 1j, k; Figure 3d-g, j-l). In contrast to lesions with loss of AQPs, these lesions exhibited astroglial activation, numerous gemistocytes, and abundant mitoses (Figure 3d-l), as well as pronounced GFAP immunoreactivity (Figure 3m, n). Occasional macrophages contained in their cytoplasm degradation products of AQP4 (Figure 3d), AQP1 (Figure 3j) and GFAP (Figure 3m).

Blinded review of medical records revealed that all 4 CPM cases with lesional loss of AQP1 and AQP 4 immunoreactivities were male, whereas the 2 cases with increased lesional AQP1 and AQP4 immunoreactivities were female. Median age at hospitalization was 49 years (range 24–68). Median disease duration (interval to death from arrival at hospital or from clinical onset of CPM) was 9 days (range 1–60).

All cases had clinical evidence of an osmotic disturbance or had an underlying condition which predisposed to the development of CPM (Table 1). Patient 1 had severe hyponatremia in the setting of alcoholism. In correcting the hyponatremia, serum Na⁺ increased by 2 mEq/L in the first 24 hours and by another 12 mEq/L in the next 24 hours (Figure 1a-d) [3, 6, 11‐16]. Patient 2 had malnutrition, severe dehydration, and hypernatremia (172 mEq/L) (Figure 1e-h; Figure 2) [11, 13‐16]. Patient 3 had hyponatremia following liver transplantation [11, 13‐16, 29‐31]. Patient 4 had small-cell lung carcinoma with brain and liver metastases, abnormal liver function tests, hyperkalemia, hyperphosphatemia and SIAD (presumed to be paraneoplastic) as the precipitating factor for CPM [11, 14]. Patient 5 had hepatic failure and hypokalemia (Figure 1i-l; Figure 3) [1, 11, 13‐16, 32]. Patient 6 had diabetes mellitus type II [11, 13‐16, 33, 34].

Apoptosis and loss of oligodendrocytes are traditionally considered the pathological hallmarks of CPM. However we found in a subset of CPM cases that astrocytes were variably reduced in numbers, were small with fewer and shorter processes, and AQP1 and AQP4 immunoreactivities were lost within actively demyelinating regions with relative preservation of myelin. These findings suggest that prominent degenerative changes in astrocytes may precede demyelination. Astrocyte damage has been described in a single previous report [8], however this is the first study characterizing in detail the predominant astrocytic injury in human CPM. Although autopsy studies do not permit definite determination of whether this astrocytic damage is primary or secondary, a rat model of osmotic demyelination suggests that rapid correction of hyponatremia triggers astrocytic damage in male rats, and that ensuing loss of trophic communication between astrocytes and oligodendrocytes results in inflammation, microglial activation, oligodendrocyte injury, and subsequent demyelination [35]. The constellation of neuropathological findings we observed, particularly among four of the CPM patients, implies a similar sequence of pathogenic events, with two notable differences. First, GFAP immunoreactivity is lost in rat lesional astrocytes, but is retained in lesional astrocytes of patients with CPM, despite a loss of both AQP1 and AQP4. This argues against the loss of AQPs being strictly due to a loss of astrocytes. Second, in the rat CPM model, lesional astrocytes show signs of extensive apoptosis, whereas apoptotic astrocytes were not evident in human CPM. Interspecies and interregional astrocytic heterogeneity may explain these discrepancies. Most human CPM lesions reside in the pons, while rodent CPM lesions typically reside in the external capsule, claustrum, corpus striatum, neocortex, hippocampus and anterior commissure [3, 4]. Furthermore, unlike rodent astrocytes which only express AQP4, astrocytes in the human brain express both AQP1 and AQP4 [27, 28]. Species and CNS differences with respect to astrocyte morphology, properties and functions therefore exist and have been described previously in both the normal and diseased brain [36‐38].

A unique finding in our study is that immunoreactivity for AQP proteins is lost in some actively demyelinating lesions in human CPM. This finding was not universal. Two of the six cases had an increase of AQP1 and AQP4 immunoreactivities within actively demyelinating CPM lesions. Complex dynamic processes that maintain water homeostasis likely contribute to the discordance of these findings. As the principal water channels in human CNS astrocytes, both AQP1 and AQP4 are involved indirectly in osmolyte movement through functional and molecular interactions with numerous ion channels and osmolyte transporters [18‐21].

The observed loss of AQP1 and AQP4 in four of six cases may reflect a compensatory astrocytic response to a hypotonic milieu. The astrocyte response anticipated in a hypotonic environment is swelling due to osmotically-driven water influx through AQP channels. The astrocyte’s physiological regulatory volume decrease preserves cell volume homeostasis by releasing inorganic and organic osmolytes, which in turn are followed by water [19, 20, 39, 40]. The therapeutic use of a hyperosmolar solution to rapidly correct chronic hypotonicity causes endothelial cell shrinkage and blood–brain barrier disruption, vasogenic edema and increased osmolarity of the CNS extracellular space [41, 42]. Water efflux from astrocytes to compensate for intracellular and extracellular osmolarities, further accentuates the shrinkage of astrocytes. Loss of AQP4 and AQP1 could represent a protective mechanism whereby astrocytes restrict water loss and prevent the triggering of apoptosis precipitated by a loss of cell volume [40, 43‐49]. The reduced number and small size of GFAP-positive (non-apoptotic) astrocytes that we observed in regions of AQP1 and AQP4 loss within CPM lesions are compatible with this hypothesis.

In contrast, two of the six CPM cases were characterized by astrogliosis associated with increased AQP1 and AQP4 immunoreactivities. In the setting of vasogenic edema accompanying CPM, an increase in AQP1 and AQP4 should facilitate water removal [21, 50]. Administration of urea or reinduction of hyponatremia during inadvertent rapid correction of hyponatremia are both known to decrease the number and severity of CPM lesions and to increase AQP4 expression [51‐56]. However, AQPs are only one of several factors that may be involved in the genesis and resolution of CPM. Although this was a small series, we did not observe differences between the severity of lesions, nor did the clinical outcomes differ in the patients regardless of observed tissue AQP status.

AQP4 co-localizes and acts in concert with the inwardly-rectifying Kir4.1 K⁺ channel [57‐65], the K⁺ clearing electro neutral co-transporter NKCC [66], volume regulated anion channels [67], Ca²⁺ activated K⁺ channels [68], the Kv1.5 voltage-gated K⁺ channel [69], two-pore domain K⁺ channel [70, 71], transient receptor potential vanniloid related channel 4 [72] and excitatory amino acid transporter 2 (EAAT2) [73]. Although the co-localization of AQP1 with ion channels on CNS astrocytes has not been studied so far, AQP1 is known to co-localize with the Na-K-2Cl cotransporter in meningiomas [74], and with the α-epithelial Na(+) channel in the lung [75]. Therefore, loss of AQP1 and AQP4 in CPM lesions would be expected to not only impair bidirectional water fluxes in astrocytes, but could also impair the transport of ions and organic osmolytes, and possibly aggravate osmotic stress [76] and glutamate excitotoxicity [73].

An unexpected finding in our study was that all CPM cases exhibiting lesional loss of AQP1 and AQP4 immunoreactivity were male patients and that female cases exhibited a lesional increase in astrocytic AQP immunoreactivity. The small number of patients precludes definitive conclusions concerning sexual dimorphism in astrocytic AQP regulation. However, female and male sex hormones have been reported to have opposite effects on ion transporters, and therefore could exert similar effects on the expression of water channels. For instance, it is known that estrogens impair brain adaptation to hyponatremia whereas androgens enhance it through respective inhibition and stimulation of Na⁺-K⁺-ATPase [77‐84]. Furthermore, it is known that estrogen induces AQP1 expression by activating the estrogen-response element in the promoter of the Aqp1 gene during angiogenesis in human breast and endometrial carcinomas [85]. Thus, it is conceivable that in the setting of osmotic stress, estrogens and androgens could have opposite effects on AQP1 and AQP4 expression in female and male patients. If downregulation of AQP1 and AQP4, and/or failure of their upregulation were an androgen-dependent mechanism, this would, on the one hand protect the astrocyte from apoptosis, but also be expected to worsen the osmotic disturbance. This outcome could plausibly explain the higher incidence of CPM observed in the male population [5, 16].CPM is also known to occur in association with a multitude of several underlying conditions, with alcoholism, liver transplantation, malnutrition, hepatic cirrhosis, burns and SIAD being among the most common. Although most of these conditions cause changes in serum Na⁺ levels, changes in other serum osmolytes have also been associated with CPM including hypokalemia, hyperglycemia, hypophosphatemia, and correction of hyperammonemia [11, 13‐16]. Furthermore, hypoxia and arginine vasopressin, in addition to estrogen, affects the brain’s ability to adapt to cellular edema [39]. Therefore the differences observed in the expression of AQP1 and AQP4 among the human CPM lesions may alternatively be explained by different underlying disease processes that contributed to the development of CPM.

Our findings, corroborated by prior studies, justify considering CPM a primary astrocytopathy with secondary demyelination [86‐88]. Furthermore, we demonstrate for the first time that AQP1 and AQP4 contribute to human CPM pathology and pathogenesis. A primary role for AQP4 has been associated with neuromyelitis optica (NMO), an autoimmune inflammatory astrocytopathy caused by complement-activating IgG autoantibodies directed against AQP4 [88‐90]. Early active demyelinating lesions in both NMO and a subgroup of CPM patients are characterized by a spectrum of astrocytic damage, AQP4 loss, retention of small GFAP immunoreactive astrocytes with fewer and shorter processes, intramyelinic edema and apoptosis of oligodendrocytes with secondary demyelination. Loss of AQP4 in NMO is caused in part by the internalization of AQP4 by some astrocytes following its interaction with NMO-IgG (antigenic modulation). Complement, when available, is activated by NMO-IgG remaining on internalization-resistant AQP4 [88‐91] expressing astrocytes. In contrast to previously published studies [92], we did not find evidence for either complement activation or AQP4 astrocyte internalization in CPM lesions, suggesting that loss of AQP4 can occur in the absence of complement-activating antibodies. The CPM lesions characterized by AQP4 and AQP1 loss, but preserved GFAP staining of astrocytes in the absence of antibodies, complement activation or AQP4 astrocyte internalization, somewhat resemble the type 6 NMO lesions recently described which show a more pronounced loss of AQP4 and AQP1 than GFAP, and also occur in the absence of immunoglobulin deposition or complement activation [27].

We propose that loss of AQP4 and AQP1 in the setting of CPM may be due to compensatory changes in response to the local osmotic environment, resulting in astrocytic injury as a consequence of the osmotic stress. Type 6 NMO lesions may similarly mirror astrocytic injury caused by osmotic stress due to disturbances of the local osmotic environment either due to a possible direct effect of NMO-IgG on water transport [90] or triggered by astrocytes in nearby regions that are damaged via antibody- and complement-mediated mechanisms [88‐91].

To characterize AQP1 and AQP4 immunoreactivity and neuropathological characteristics of CPM, we analyzed microscopic transverse sections of the pons in archival autopsied tissue from two control cases (17 year old female and 15 year old male), and six pathologically confirmed CPM cases. Clinical information was obtained from medical records. The study was approved by the Institutional Review Board of the Mayo Clinic, Rochester.

Specimens were fixed in 10–15% formalin and embedded in paraffin. Sections were stained with haematoxylin and eosin (HE) for morphological evaluation and Luxol-fast blue-periodic acid-Schiff (LFB/PAS) to demonstrate myelin and its degradation products. Lesions were classified immunohistochemically with respect to demyelinating activity, as previously described [93]. Immunohistochemical analyses used avidin–biotin or alkaline phosphatase/anti-alkaline phosphatase technique without modification [94]. Tissues were exposed (16 hrs, at 4°C), to IgGs specific for: AQP1 (rabbit polyclonal 1:500; Santa Cruz, USA), AQP4 (affinity-purified rabbit polyclonal 1:250; Sigma-Aldrich, USA), myelin proteolipid protein (PLP, rabbit polyclonal 1:500; Serotec, Oxford, USA), 2′3′-cyclic-nucleotide 3′-phosphodiesterase (CNPase, mouse monoclonal 1:2000; Sternberger, USA), myelin oligodendrocyte protein (MOG, mouse monoclonal 1:200; gift of Dr Sarah Piddlesden, Cardiff, UK), myelin-associated glycoprotein (MAG, mouse monoclonal 1:10; Chemicon, USA), glial fibrillary acidic protein (GFAP, mouse monoclonal 1:4000; Dako, Denmark), neurofilament protein (NF, mouse monoclonal 1:800; Dako, Denmark), T lymphocytes (CD3, rat monoclonal 1:400; Serotec, USA), macrophages/microglial cells (KiM1P, mouse monoclonal 1:1000; gift from Dr. Radzun, University of Göttingen, Germany) and activated complement antigen (C9neo, mouse monoclonal 1:200; gift from Dr. Paul Morgan, Department of Biochemistry, Cardiff, UK). Primary antibodies were omitted in control staining. Antigen retrieval was performed as previously described [88].