GLUT2 Immunoreactivity in Gomori-positive Astrocytes of the Hypothalamus

Abstract

A specialized subtype of astrocyte, the Gomori-positive (GP) astrocyte, is unusually abundant and prominent in the arcuate nucleus of the hypothalamus. GP astro-cytes possess cytoplasmic granules derived from degenerating mitochondria. GP granules are highly stained by Gomori's chrome alum hematoxylin stain, by the Perl's reaction for iron, or by toluidine blue. The source of the oxidative stress causing mitochondrial damage in GP astrocytes is uncertain, but such damage could arise from the oxidative metabolism of glucose transported into astrocytes by high-capacity GLUT2 glucose transporters. In accord with this hypothesis, the reported anatomical distribution of astrocytes staining positively for GLUT2 glucose transporters closely matches that of GP astrocytes. To examine whether or not these two staining procedures detect the same population of astrocytes, immunocytochemistry was performed on semithin sections to detect GLUT2 protein and sections were then stained with toluidine blue to detect GP granules. It was determined that GP astrocytes are frequently immunoreactive for the GLUT2 transporter protein. These data support the possibility that GP astrocytes may have an important influence upon the reactivity of the hypothalamus to glucose and that a specialized glucose metabolism may in part underlie the development of mitochondrial abnormalities in hypothalamic GP astrocytes.

AN UNUSUAL subtype of astrocyte, the Gomori-positive (GP) astrocyte, is particularly abundant in the arcuate nucleus of the hypothalamus. These cells can be visualized with light microscopy as astrocytes with ovoid, euchromatic nuclei that are surrounded by a cloud of granules intensely stained by Gomori's chrome alum hematoxylin stain or by toluidine blue. The identity of these cells as astrocytes is confirmed by immunoreactivity for the astrocyte protein, glial fibrillary acidic protein (Young et al. 1990; Schipper and Mateescu-Cantuniari 1991). Electron microscopy has shown that cytoplasmic GP granules are derived from degenerating mitochondria engulfed within lysosomes (Brawer et al. 1994; Wang et al. 1995). A number of experiments have shown that the GP phenotype and mitochondrial degeneration appear to result from some type of oxidative stress (Schipper et al. 1993). The reasons why oxidative stress would be particularly intense or damaging to astrocyte mitochondria in the arcuate nucleus are uncertain.

Increased glucose metabolism may be one factor underlying an enhanced oxidative stress in the arcuate nucleus. Several studies have reported the presence of high-capacity GLUT2 glucose transporter proteins in arcuate astrocytes and in arcuate tanycytes (Leloup et al. 1994; Ngarmukos et al. 2001; García et al. 2003). High levels of the mRNA for GLUT2 transporters have also been detected in the arcuate nucleus (García et al. 2003; Li et al. 2003).

In the liver and islets of Langerhans, these high-capacity, low-affinity GLUT2 glucose transporter proteins represent a portion of the glucose-sensing mechanism of hepatocytes and β cells. In the retina, GLUT2 transporters allow glial cells to take up and metabolize glucose at a substantially greater rate than that of neurons (Poitry-Yamate and Tsacopoulos 1992; Watanabe et al. 1994). In the hypothalamus, GLUT2 transporters seem functionally significant, because intraventricular infusions of antisense oligonucleotides that bind GLUT2 mRNA alter the neuroendocrine response to hypoglycemia (Leloup et al. 1998; Wan et al. 1998).

The overall distribution of arcuate astrocytes immunoreactive for GLUT2 transporters reported in one study appears to match the anatomical distribution of GP astrocytes (Young et al. 1990; Leloup et al. 1994). Chronic exposure of neural tissue to hyperglycemia does provoke the type of oxidative stress and mitochondrial impairment that is thought to damage mitochondria in GP astrocytes (Yang et al. 1998; Schmeichel et al. 2003). Thus, an increased rate of glucose uptake and oxidation could conceivably be one factor underlying the development of the GP phenotype in arcuate astrocytes.

An initial step in examining this hypothesis is to determine if some or all GP astrocytes stain positively for the GLUT2 glucose transporter protein. That is the main goal of this study.

Materials and Methods

Experimental Animals

Animal protocols were reviewed and approved by the Institutional Animal Care and Use Committee of the Howard University College of Medicine. All rats used in this study were 6-8 months old, because the mitochondrial degeneration present in GP astrocytes is age dependent and becomes much more prominent and easily detectable in older animals (reviewed in Young et al. 1990). Six rats were overdosed with Nembutal (Sigma-Aldrich, St. Louis, MO) and perfused through the left ventricle with 10% formalin in 0.2 M phosphate buffer. Two brains were placed in phosphate buffer plus 20% sucrose for cryoprotection and a later preparation of 30-μm-thick frozen sections. The remaining brains were kept in phosphate buffer for preparation of 60-μm-thick vibratome sections. Vibratome sections of liver were also prepared for the demonstration of GLUT2 immunoreactivity at the surface of hepatocytes.

Immunocytochemistry for GLUT2 Transporters

Free-floating vibratome sections, prepared using an Oxford vibratome (Vibratome, St. Louis, MO), were stained to demonstrate GLUT2 immunoreactivity. Sections were first all washed in PBS-normal goat serum and then incubated overnight at 4C with either the antibody to GLUT2 or with nonimmune rabbit serum.

GLUT2 immunoreactivity was demonstrated by use of a polyclonal rabbit antibody generously provided by Dr. William Pardridge. This antibody identified GLUT2 immunoreactivity in arcuate astrocytes in a recent publication (Ngarmukos et al. 2001). The antibody was raised in rabbits against a synthetic peptide, corresponding to a 13-amino acid carboxy-terminal sequence of the GLUT2 protein, that was then conjugated to bovine thyroglobulin. To eliminate nonspecific reactivity to thyroglobulin-like epitopes, the antibody was diluted 1:100 in 1 ml of Tris buffer, 5 mg of bovine thyroglobulin was added (Sigma Chemical Co.; St. Louis, MO), and the mixture was shaken at 37C for 60 min and spun at 10,000 × g to produce a pellet, which was discarded (Cramer et al. 1992). Preliminary trials indicated that the optimal dilution of the antibody for staining was 1:900. As an additional control, an aliquot of antibody, diluted 1:900, was preabsorbed by shaking at 37C with the GLUT2 peptide (Biogenesis, Inc., Kingston, NH). Vibratome sections were incubated in this antibody or in nonimmune rabbit serum, as described above, to demonstrate GLUT2 immunoreactivity.

After this, sections were washed in PBS-0.1% Triton X-100, incubated for 30 min in biotinylated goat anti-rabbit IgG (Vectastain ABC kit; Vector Laboratories, Burlingame, CA), and washed once more in PBS-0.1% Triton X-100. To nullify endogenous sources of pseudoperoxidase such as the cytoplasmic granules of GP astrocytes, sections were then incubated for 30 min in PBS-0.3% hydrogen peroxide, followed by washing in PBS and a 60-min incubation in avidin-biotin-peroxidase complex. Immunoreactivity was then localized via the production of a brown precipitate of diaminobenzidine (DAB) at the sites of peroxidase-labeled antibody by exposing the sections to a 0.05% solution of DAB plus 0.03% hydrogen peroxide for 15 min (Young et al. 2000). Subsequently, vibratome sections were dehydrated in alcohol and flat embedded between two plastic cover slips in JB-4 methacrylate resin (Polysciences, Inc.; Warrington, PA). These sections were then glued to methacrylate blocks using cyanoacrylate glue and were cut at a thickness of 1 μm using dry glass knives and a JB-4 microtome. Semithin sections were dried down onto gelatinized slides and briefly (20 s) counterstained with 0.1% toluidine blue in 0.2 M acetate buffer (pH 4.5) (Young et al. 1990).

Staining of GP Astrocytes

Several methods were used to stain cytoplasmic GP granules. First, to demonstrate the overall distribution of GP granules, which contain heme-linked iron derived from mitochondria, serial frozen sections of the arcuate region were prepared and every other section was stained with an intensified version of the Perl's reaction for iron. Free-floating sections were immersed in 5% HCl and 5% potassium ferrocyanide for 40 min to generate precipitates of ferric ferrocyanide over GP granules. Sections were then rinsed in phosphate buffer and immersed in 0.1% DAB containing 0.005% hydrogen peroxide for an additional 15 min. This procedure, which uses ferric ferrocyanide to intensify the conversion of DAB into a dark brown precipitate by hydrogen peroxide, demonstrates the minute iron-rich GP granules scattered throughout the arcuate nucleus but leaves other cellular elements unstained (Young 1988). In a number of trials, attempts were made to combine this procedure with immunocytochemistry. However, the highly acid pH required for the Perl's reaction either degraded antigenicity or else dissolved the DAB precipitate, making this reaction unsuitable for staining GP granules either before or after immunocytochemistry.

In an attempt to examine another staining method for GP astrocytes that is more compatible with immunocytochemistry, an approach described by Schipper and Mateescu-Cantuniari (1991) was used. This approach takes advantage of the pseudoperoxidase activity inherent in the heme-linked iron present in GP granules to directly convert DAB into an insoluble precipitate without prior exposure to the reagents of the Perl's reaction. Every other hypothalamic section was incubated in Tris-buffered saline containing 0.05% DAB plus 0.002% hydrogen peroxide for 30 min, and then sections were washed in PBS, mounted on slides, dehydrated, and cover slipped. Numbers of brown-staining clusters of granules, corresponding to GP astrocytes, were compared with counts obtained using the Perl's stain on adjacent brain sections.

To demonstrate GP granules in semithin sections, immunostained 60-μm-thick vibratome sections were flat embedded, sectioned at a thickness of 1 μm, and counterstained with 0.1% toluidine blue in 0.2 M acetate buffer (pH 4.5) so that GP granules appeared bright blue and cellular immunoreactivity retained a brown color attributable to DAB (Young et al. 1990).

Results

Frozen sections of brain stained with the DAB-intensified Perl's reaction for iron displayed minute, iron-containing granules of GP astrocytes that were clustered around faintly stained astrocyte nuclei and were evenly distributed throughout the arcuate nucleus (Figure 1). Sections stained with the DAB procedure alone showed clusters of granules that were similar in general but less intensely stained and less easily detectable. Counts of GP granules in each section differed between the two staining procedures: an average of 45.7 ± 3.2 GP astrocytes were detectable via the Perl's method, whereas 11.3 ± 1.6 GP astrocytes were evident in each section of the arcuate nucleus after staining with DAB alone. It was concluded that the DAB stain, although more compatible with immunocyto-chemistry than the Perl's reaction, would not be suitable for a quantitative estimate of numbers of GP astrocytes in the arcuate nucleus.

Figure 1.

(A) Distribution of GP astrocytes within the arcuate nucleus of the hypothalamus. Perl's stain for iron was performed on a 30-μm-thick frozen section, showing iron-positive cytoplasmic granules (arrows). (B) Higher magnification view of a GP astrocyte, showing clusters of iron-containing cytoplasmic granules (arrows). (C) Semithin section of the arcuate nucleus, stained with toluidine blue to illustrate GP granules adjacent to an astrocyte nucleus (arrow). Bars: A = 100 μm; B, C = 50 μm.

Vibratome sections of liver showed GLUT2-immunoreactive (ir) at the perisinusoidal surfaces of hepatocytes (Figure 2A). Immunoreactivity was absent from endothelial cells or from surfaces between hepatocytes that contributed to bile canaliculi. These results are consistent with those of other studies (Cramer et al. 1992; Lane et al. 1999).

Figure 2.

(A) Immunoreactivity for GLUT2 transporters at the sinusoidal surfaces of hepatocytes. (B-D) Examples of GLUT2 immunoreactivity in astrocytes possessing cytoplasmic GP granules (arrows). Bar = 5 μm. (E) GP astrocyte immunoreactive for GLUT2 (arrows) adjacent to a long, thin tanycyte process immunoreactive for GLUT2 (arrowhead). All micrographs were photographed using a 100× oil-immersion lens.

Antibodies were able to penetrate into the vibratome sections for only a few micrometers, so that patterns of immunoreactivity were confined to only a few semithin sections taken off of the top and bottom surfaces of each 60-μm-thick vibratome section. Thus, out of ~240 semithin sections of hypothalamus, fewer than 40 sections manifested immunoreactivity in the arcuate region. In this brain region, GLUT2 immunoreactivity was visible as thin, brown streaks scattered throughout the arcuate nucleus. Highly elongated, thin immunoreactive processes that extended throughout the arcuate nucleus were readily identifiable as belonging to tanycytes (Figure 2E). In sections in which astrocyte cell nuclei happened to be in the same plane of section as the immunoreactivity, shorter GLUT2-ir cell processes could be seen adjacent to astrocyte nuclei that appeared to radiate away from the location of the cell nucleus. In most of these cases (44 out of 58 immunoreactive astrocytes), the GLUT2-ir was surrounded by a cloud of minute blue-stained, GP cytoplasmic granules (Figures 2B-2E). In one fortuitous section apparently tangential to the surface of the astrocyte, immunoreactivity clearly followed the boundaries of the astrocyte cytoplasm and was enclosed by a collection of GP granules (Figure 2B). Only 4 out of 53 GP astrocytes that were present near traces of immunoreactivity were themselves lacking in GLUT2 immunoreactivity.

No GLUT2 immunoreactivity was detected in brain regions dorsal to the hypothalamus or in sections exposed to nonimmune serum or to GLUT2 antibody preabsorbed with GLUT2 peptide (data not shown).

Discussion

This study detected immunoreactivity for GLUT2 glucose transporters in tanycytes and astrocytes of the arcuate nucleus of the hypothalamus. Two previous studies have also reported GLUT2 immunoreactivity in arcuate astrocytes, whereas a third study detected GLUT2 primarily in GFAP-positive arcuate tanycytes (Leloup et al. 1994; Ngarmukos et al. 2001; García et al. 2003). In our hands, GLUT2 immunoreactivity was detectable in both cell types. Variations in the properties of GLUT2 antibodies used in these four studies may explain these different results; all studies demonstrated GLUT2 immunoreactivity in glia or in glia-like cells but not in neurons. The majority of GLUT2 immunoreactive astrocytes in this study were GP astrocytes. This finding may have a number of functional implications.

An unusual glucose metabolism of GP astrocytes may be one factor provoking oxidative damage and an age-related mitochondrial degeneration in these cells. During episodes of high blood glucose, glucose uptake into cells bearing high-capacity GLUT2 transporters would be increased. An increased glucose metabolism can provoke the type of oxidative stress in both pancreatic β cells and in neural tissue that is thought to damage mitochondria in GP astrocytes (Yang et al. 1998; Robertson et al. 2003; Schmeichel et al. 2003). In contrast, during conditions of hypoglycemia, low-affinity GLUT2 transporters fail to maintain an adequate glucose uptake into cells (Olson and Pessin, 1996). Hypoglycemia would thus impose another type of metabolic stress upon GLUT2-immunoreactive cells. In this context, it appears significant that GP astrocytes exhibit increased levels of glucose-regulated protein-94, which is normally produced in response to a deficit in glucose uptake (Mydlarski and Schipper, 1993). To further examine the possible consequences of a specialized glucose metabolism, it would be desirable to determine if bouts of hyperglycemia or hypoglycemia alter the rate of development of mitochondrial degeneration in GP astrocytes in vivo.

GLUT2 immunoreactivity in GP astrocytes may also have implications for the function of adjacent neurons. Neurons in the basomedial hypothalamus show an unusual sensitivity to changes in circulating glucose levels (Tkacs et al. 2000; Young et al. 2000; Briski and Marshall 2001). The function of these neurons can in turn be modulated by infusions of lactate (Briski 1999; Borg et al. 2003). Because the main source of brain lactate is from the metabolism of astrocyte glycogen, carbohydrate metabolism by astrocytes may have an important modulatory effect upon glucose sensing by the hypothalamus. This proposition is supported by the ability of astrocyte toxins to diminish the effects of 2-deoxyglucose and goldthioglucose upon the hypothalamus (Young, 1988; Young et al. 2000; Pow, 2001).

Many of the glucose-sensing neurons of the hypothalamus appear to be dopaminergic (Briski 1998). We have previously established that GP astrocytes are in close contact with arcuate dopaminergic neurons at the ultrastructural level (Young et al. 1990). Thus, GP astrocytes have both anatomical and physiological characteristics that would enable them to modify the process of glucose sensing by the hypothalamus.

It would be desirable to determine if GP astrocytes in other brain regions also possess GLUT2 transporters. For example, GP astrocytes can also be detected in the hippocampus (Young et al., 1996; St. Jacques et al. 1999). It seems reasonable that GP astrocytes in other brain regions would share the phenotypic complement of proteins that they possess in the hypothalamus. An example of such a protein is brain fatty acid binding protein, which is unusually abundant in arcuate GP astrocytes and which also is expressed in GP astrocytes in the hippocampus (Young et al. 1996). It was not possible, however, to test this proposition in this study because immunoreactivity for GLUT2 transporters was not detected in extrahypothalamic brain regions. This was likely related to the much lower abundance of the mRNA for GLUT2 in cortical and subcortical structures relative to that in the hypothalamus (Li et al. 2003).

Finally, the age-related damage to mitochondria in GP astrocytes may compromise the function of adjacent neurons. It is known, for example, that a specific blockade of oxidative metabolism in glial mitochondria by fluoroacetate can have serious consequences for the function of nearby neurons (Hülsmann et al. 2000). Also, oxidative damage to astrocyte mitochondria now appears to contribute to neuronal dysfunction in Alzheimer's disease (St. Jacques et al. 1999; Abramov et al. 2004). One way to examine this question in the case of GP astrocytes would be to determine if the age-related increase in numbers of GP granules correlates with an altered response to glucose deprivation by neurons of the hypothalamus. If so, an age-related oxidative damage to these glucose-sensitive glia in the hypothalamus may have important implications for hypothalamic function and for the pathophysiology of diabetes mellitus. Clinical data suggest that an impairment of hypothalamic glucose sensing may be present in at least a portion of diabetic patients and may increase the likelihood and damaging effects of episodes of hypoglycemia (Cranston et al. 2001).