Main

Glucose is an essential substrate for adult brain oxidative metabolism(1, 2). Brain glucose consumption constitutes 90% of the total body insulin independent glucose utilization(3, 4). In contrast, the fetal and neonatal brain utilizes other non-glucose substrates, such as ketones(5) and lactate(6), as well as glucose(7). Still, the developing brain must rely upon a constant supply of glucose for ongoing cellular growth, differentiation, and functional development(814). An inadequate supply of glucose leads to adverse effects on the forming mammalian brain(15). As has been reported previously, an excessive supply of glucose results in enhanced fetal brain glucose and oxygen uptakes(16). In other tissues, increased circulating glucose leads to enhanced glycolysis with pyruvate and lactate accumulation(17).

Facilitative glucose transport constitutes the first step in brain glucose uptake and utilization. Currently, of the six glucose transporter isoforms cloned to date(1820), there are two main isoforms with a functional role in transporting glucose that are expressed in a predominant manner in adult brain. They are cell-specific, with Glut 1 mostly localized to the endothelial cells of the blood-brain barrier(2124), and Glut 3 to neuronal cells and processes(2526). As in the adult, Glut 1 and Glut 3 are also present in fetal brain, but at lower levels(27).

Previously, we examined the effect of maternal diabetes upon fetal brain Glut 1 and Glut 3 proteins in the nonobese pregnant diabetic mouse model(NOD). No substantive changes were found(27). This was in contrast to in vivo observations in adult rat brain, and in vitro, where in higher levels of extracellular glucose caused a decline in brain microvascular(28) and glial cell Glut 1 levels, with associated decreased glucose uptake(12). However, the question still remained that, perhaps, this absence of change in fetal brain glucose transporters encountered in vivo, was unique to the NOD model of genetically induced diabetes used by us and related to the chronicity of the diabetic disease process in this model.

To address this question further, we examined the effect of short-term maternal diabetes of relatively severe proportions on fetal brain Glut 1 and Glut 3 expression, by employing the more conventional, well described, and widely accepted streptozotocin induced model of diabetes in pregnancy.

METHODS

Antibodies and cDNA probes. The mouse Glut 3 peptide used to generate the anti-Glut 3 antibody, and the polyclonal anti-rat Glut 1 IgG were obtained from Dr. M. Mueckler (Washington University School of Medicine, St. Louis, MO). Mouse Glut 3 cDNA was a gift from Dr. G. I. Bell (Howard Hughes Institute, University of Chicago, Chicago, IL.) All other reagents were obtained from commercial vendors.

Animals. Gestationally timed pregnant Sprague-Dawley rats(Taconic Farm's Inc., Germantown, NY) were housed in individual cages, and were maintained with 12 h light-dark cycles. As approved by the St. Louis University School of Medicine's Animal Care Committee, the National Institutes of Health guidelines in the care and use of animals were followed.

The pregnant animals were allowed at least 1 d of acclimatization before experimental manipulation. All animals, on d 12-13 of gestation, received either 65 mg/kg of freshly prepared and chilled (on ice) streptozotocin (Sigma Chemical Co., St. Louis, MO) or an equal volume of vehicle intraperitoneally. Maternal tail vein glucose levels were monitored with a One Touch glucose analyzer (Lifescan, Inc., Milpitas, CA.) and the animals were divided into three groups based on the glucose values.

The vehicle injected animals demonstrated blood glucoses of less than 10 mM and were assigned to the control group [C (n = 24)]. The streptozotocin-injected animals were further divided into two groups. A portion of the injected animals failed to develop overt diabetes. They were assigned to the STZ-ND group (n = 12), and had glucose values of<10 mM. This group served as an additional control group that helped to distinguish the direct chemical effects of streptozotocin from the effects of maternal diabetes. The rest of the animals in this study that received streptozotocin and developed glucose levels greater than 19.5 mM were assigned to the severely diabetic group (SEVERE-D [n = 29]). There were no animals with glucose concentrations between 10 and 13.4 mM. Animals with glucose concentrations ranging between 13.5 and 19.4 mM were not included.

On d 20 of a 21-d gestation, pregnant rats were anesthetized with pentobarbital and underwent hysterotomy. Maternal blood was collected for plasma glucose and insulin assessments. Urine was checked for ketonuria. Fetal blood was obtained and pooled from single litters for the analysis of plasma glucose and insulin. Plasma glucose levels were analyzed by the glucose oxidase method(29) and insulin by the double antibody RIA(29). Fetal brains were collected, pooled from single or two litters, and stored at -70°C for further analyses. The exception was in the case of nuclei isolation, where pooled fetal brains were separately collected and immediately processed.

DNA quantification:. DNA content of whole tissue homogenates was determined microfluorometrically by employing the Hoechst 33258 dye as described previously(30).

Nuclear run-on experiments. Fetal brain nuclei were isolated as previously described(31). From pooled fetal brains of the three experimental groups, 2 × 107 nuclei were used per reaction. The nuclei were first washed with 20 mM Tris, pH 8.3, 5 mM MgCl2, 800 mM KCl, 500 μM dNTPs (ATP, CTP, GTP), and 40% glycerol. The nuclei were suspended in the same buffer containing 125 μCi of uridine[32P]triphosphate, and incubated at 30°C for 30 min. Reactions were terminated by adding 20 U of DNAse I and further incubated for 1 h at 37°C. The proteins were removed by treating the reaction mix with 100μg/mL proteinase K (in 250 mM NaCl, 10 mM Tris, pH 7.4, 25 mM EDTA, 1% SDS) for 45 min at 37°C. The reaction mix was then extracted twice with equal volumes of phenol, followed once with a phenol:chloroform:isoamyl alcohol mix(ratio 1:0.5:0.5), and then once with chloroform alone. Nascent transcripts were then precipitated with 2 volumes of ethanol and 0.3 mM sodium acetate with 20 μg of yeast tRNA, serving as a carrier. Precipitates were centrifuged 15 min at 4°C at 12,000 rpm, washed with 70% ethanol, and suspended in 100 μl of RNAse-free water. The 32P-labeled RNA was separated from unincorporated nucleotides by using the RNaid kit (Bio 101, Inc., La Jolla, CA.) following the kit's recommended protocol. After purification, 1 μl was counted from each sample to assess the incorporated radioactivity(31).

Slot blot analysis. Two micrograms of 2400-bp mouse Glut 1(32), 1600-bp mouse Glut 3(25), 1800-bp chicken β-actin, and pGEM vector (as negative control) each were immobilized onto a nylon membrane using a slot blot manifold. The DNA on the filters was hybridized with 7 × 106 cpm of 32P-labeled nascent transcripts obtained from the brain nuclei of control, STZ-ND, and SEVERE-D fetuses for 84 h at 42°C [50% deionized formamide, 0.1% polyvinylpyrrolidone, 0.1% BSA, 0.1% Ficoll, 5 × SSC (0.75 M sodium chloride, 0.075 M sodium citrate), 0.2% SDS, 200 μg/mL salmon sperm DNA, 0.05 M sodium pyrophosphate, and 12 μg/mL tRNA]. The filters were subsequently washed twice with 1 × SSC (0.15 M sodium chloride, 0.015 M sodium citrate; pH 7.0) and 0.5% SDS at room temperature, twice with 1 × SSC and 0.5% SDS at 37°C, and then exposed for 8 d to Kodak X-O-Mat film at -70°C(31). The autoradiographs were subjected to two-dimensional densitometry (Biomed Instruments, Fullerton, CA.).

Northern blot analysis. Whole tissue poly (A+) RNA was extracted using the Fast track kits followed by ethanol-3 M sodium acetate precipitation(27). The final RNA pellet was resuspended in TE buffer (10 mM Tris, pH 7.4, 0.1 mM EDTA, pH 8.0); RNA was quantitated spectrophotometrically at 260 nm; and the integrity of the sample determined by gel electrophoresis.

Ten to 20 μg of fetal brain poly(A+) RNA were separated by 1.2% agarose-2.2 M formaldehyde slab gel electrophoresis and transferred to Nytran filters(33). These filters were subjected to Northern blot analysis as described previously. Hybridization was performed at 55°C with 1 × 106 cpm/mL of a 32P-oligolabeled 2.4-kb mouse Glut 1 cDNA(32), or a 1.6-kb mouse Glut 3 cDNA fragment(25). The filters were subsequently washed twice in 2 × SSC (1 × SSC = 0.15 M sodium chloride, 0.015 M sodium citrate; pH 7.0) for 5 min each at room temperature, followed by two washes in 1 × SSC/0.1% SDS for 30 min each at 55°C. The filters were then exposed to autoradiography for varying periods of time. Specific mRNA abundance was quantitated by densitometry, and the same filters were stripped and rehybridized to a 1.8-kb chicken β-actin cDNA fragment(27). The β-actin mRNA levels closely reflected the ethidium bromide staining of the RNA samples, thereby serving as an internal control to standardize the amount of RNA loaded per individual lane. Specific mRNA/β-actin mRNA ratio for a given sample was then assessed.

Western blot analysis. From each of the three experimental groups, 100 μg of pooled fetal brain homogenates were solubilized in Laemmli's buffer(34). Protein content was determined by the Bradford's dye-binding assay(35). The samples were then separated by 10% discontinuous SDS-PAGE(34). The proteins were transferred to nitrocellulose filters by electroblotting(36), and the filters were subjected to Western blot analysis as described previously(27). The primary antibodies used were 2 μg/mL of a protein A -affinity-purified rabbit anti-rat Glut 1 C terminus peptide antibody(37, 38), or 20 μg/mL of a peptide-affinity-purified anti-mouse Glut 3 IgG, which was raised in a rabbit against the last 16 amino acids of the mouse Glut 3 C terminus peptide(25). The autoradiographs were subjected to two-dimensional densitometry (Biomed Instruments, Fullerton, Ca.).

Data analysis. All data presented are expressed as mean± SEM. All three groups were compared by the one-way ANOVA, and intergroup differences were validated with the Newman Kuel's test. Additionally nonparametric testing was also conducted using Kruskal-Wallis with tied ranks, followed by the Dunn's test, to ensure that the conclusions drawn were not due to either small numbers resulting in a β-error or due to a heteroscedastic distribution of observations. Using both parametric and nonparametric testing, statistically significant differences between groups were similar.

RESULTS

The characteristics of the three groups are shown in tabular form. There is no difference in final maternal weight, mean fetal weight, or number of pups per litter (Table 1). Additionally, either trace (<5 mg/dL) or no ketonuria was observed in the SEVERE-D animals. No significant differences were noted in fetal brain weight, protein content, or DNA content(Table 2). As per the study design, there were no differences in maternal or fetal glucose concentrations between the control and the STZ-ND groups (Table 3). Insulin values were like-wise no different in these two groups (Table 3). In contrast, the SEVERE-D group demonstrated a 4-fold increase in maternal glucose levels, and a 6-fold increase in fetal glucose concentrations(Table 3). A decline in maternal insulin levels was evident (Table 3). Unlike our previous study(39) but typical of the severe form of a streptozotocin-diabetic pregnancy, mild fetal hypoinsulinemia was also observed (Table 3).

Table 1 Animal demographics
Table 2 Fetal brain characteristics
Table 3 Plasma glucose and insulin concentrations

Figure 1 demonstrates the nuclear run-on experimental results. The rates of transcription/elongation for Glut 1 and Glut 3 in fetal brain were low compared with β-actin. Although densitometric evaluation of Glut 1 revealed absolute values higher than the pGEM control, Glut 3 values were minimally higher than the plasmid control. Given the limitation of accurately detecting differences at levels of minimal expression, densitometric values expressed as a percent of the vehicle control group revealed no substantive difference in fetal brain Glut 1 (C = 100 ± 0%; STZ-ND = 90 ± 13%; SEVERE-D = 96 ± 12%) or Glut 3 (C = 100± 0%; STZ-ND = 102 ± 5%; SEVERE-D = 105 ± 2%) transcription/elongation rates among the three groups. Both Glut 1 and Glut 3 results paralleled those seen with β-actin which served as the internal control (C = 100 ± 0%; STZ-ND = 92 ± 13%; SEVERE-D = 97 ± 14%).

Figure 1
figure 1

Nuclear run-on experiment. A representative autoradiograph of a nuclear run-on demonstrating pGEM (1st row), β-actin(2nd row), Glut 1 (3rd row), and Glut 3 (4th row) transcriptional/elongation rates in fetal brain nuclei obtained from vehicle control (C; 1st column), streptozotocin-treated nondiabetic (SC; 2nd column), and severely diabetic(SD; 3rd column) groups.

Figure 2 depicts the Northern (Fig. 2D) and Western blots (Fig. 2E). All mRNA data were standardized to the β-actin internal control (Fig. 2C). These standardized values paralleled the standardization to either the 28 S or 18 S ethidium bromide stained rRNA bands (Fig. 2C). No statistically significant differences are seen in the steady state Glut 1 or Glut 3 mRNA levels among the three groups (ANOVA with Newman-Kuel's test; Kruskal-Wallis with Dunn's test) (Fig. 2A).

Figure 2
figure 2

Quantitative densitometric analysis of Glut 1 and Glut 3 mRNA (panel A) and protein (panel B) levels in pooled fetal rat brains from the C, STZ-ND, and SEVERE-D groups. (Panel A)[Gult 1: (C) n = 5, (STZ-ND) n = 5, (SEVERE-D) n= 5; GLUT 3: (C) n = 6, (STZ-ND) n = 4, (SEVERE-D)n = 5; p (ANOVA) and Newman-Kuel's or Kruskal-Wallis and Dunn's test= 0.07 (Glut 1), = 0.06 (Glut 3). (Panel B) [Glut 1: (C) n= 4, (STZ-ND) n = 4, (SEVERE-D) n = 7; GLUT 3: (C)n = 8, (STZ-ND) n = 3, (SEVERE-D) n = 7#p (ANOVA) and Newman-Kuel's or Kruskal-Wallis and Dunn's test = 0.05 (Glut 1), = 0.0006 (Glut 3). (Panel C) An ethidium bromide stained 1% agarose gel demonstrates the 28 S and 18 S ribosomal RNA in C, STZ-ND, and severely diabetic (SD) groups (top). A representative Northern blot demonstrating the 1.8-kb β-actin mRNA in C, STZ-ND, and severely diabetic (SD) groups (bottom). (Panel D) Representative Northern blots demonstrating the 2.8-kb Glut 1 mRNA(top) and the 4.1-kb Glut 3 mRNA (bottom) from fetal brains of the C, STZ-ND, and severely diabetic (SD) groups. (Panel E) Representative Western blots demonstrating the ≈47-kD Glut 1(top) and the ≈50-kD Glut 3 (bottom) protein in fetal brains from the C, STZ-ND, and SEVERE-D groups.

A trend toward a decline in fetal brain Glut 1 protein concentrations is seen between the groups when all three groups are considered simultaneously via ANOVA or Kruskal-Wallis tests (p = 0.05), with no difference between the STZ-ND and SEVERE-D groups (Newman-Kuel's and Dunn's tests). A 35-50% decline in Glut 3 protein levels is seen in both the STZ-ND and the SEVERE-D groups when compared with the vehicle control group (ANOVA with Newman-Kuel's and Kruskal-Wallis with Dunn's tests) (Fig. 2B).

DISCUSSION

Prior attempts at elucidating the regulation of glucose's entry into neural tissue have taken many approaches. In vitro studies showing the effects of extracellular glucose upon fetal and neonatal glial cell cultures have shown reciprocal changes in Glut 1 mRNA, protein levels, and 2-deoxyglucose uptake with altered glucose concentrations(10, 12). However, in the whole animal, circulating glucose first has to cross the Glut 1-rich endothelial cells lining the blood-brain barrier before accessing the glial or neuronal cells. The fetus relies entirely on the maternal circulation for its glucose supply(40). Studies using fetal rat microvessels and microvessel-free cortical membranes have demonstrated the presence of Glut 1 in both these fractions attesting to the presence of Glut 1 in endothelial cells of the microvasculature and the neural cells of the cortical parenchyma(24, 41, 42). In vitro cultures of adult bovine brain capillary endothelial cells, in the face of hypoglycemia, have shown posttranscriptional enhancement of Glut 1 transporter gene expression via mRNA stabilization(43). However,in vitro studies do not reflect the in vivo situation. In any cultured cell type including embryonic postmitotic neurons in which Glut 1 is not detected in-situ(24), Glut 1 expression is observed(44).

Studies in vivo using isolated microvessels from the adult rat brain have shown that streptozotocin-induced diabetes causes an increase in Glut 1 mRNA with a decline in the corresponding protein levels(28, 45). Examination of adult brain parenchyma using whole brain homogenates showed that streptozotocin diabetes, although suppressing steady state Glut 1 mRNA levels, failed to demonstrate any comparable change in Glut 1 protein levels(46). These two studies support a diabetes-induced down-regulation of the adult brain microvascular Glut 1 protein alone. Investigations examining adult mouse and rat brain Glut 3 mRNA and protein levels demonstrated no change to a modest change in response to streptozotocin-induced diabetes(47, 48).

In the nonobese diabetic mouse model of genetically induced maternal diabetes, maternal and subsequent fetal hyperglycemia results from autoimmune destruction of maternal pancreatic β-islet cells(49). In this model, 21 d of fetal hyperglycemia with either fetal hyperinsulinemia or normoinsulinemia, although decreasing fetal brain Glut 1 mRNA levels, demonstrated only minimal (≈20%) changes in fetal brain Glut 1 protein levels. In these animals, fetal Glut 3 mRNA and protein levels were barely detectable(27). Unlike the mouse(27) and human(26) situations, detectable levels of Glut 3 mRNA and protein are observed in fetal rat brain, albeit at lower levels than in the adult. To differentiate the effects of long-term maternal diabetes as encountered in the NOD mouse and to make comparisons with previous in vivo adult studies of diabetes, a relatively short-term streptozotocin-induced severe diabetic pregnancy rat model was important to examine.

Administration of streptozotocin during pregnancy results in a heterogeneous presentation of maternal diabetes, resulting from variations of the dose of the chemical and the timing of its administration. Streptozotocin administered between d 0 and 8 of gestation causes symmetrical intrauterine growth restriction due to a teratogenic effect upon embryogenesis(5052). Streptozotocin administration later in gestation impairs growth asymmetrically from a diminution of uteroplacental blood flow leading to fetal ischemia(53). A lower dose of streptozotocin administered during early gestation on the other hand leads to fetal hyperinsulinemia with macrosomia(54). The dose and timing of administration of streptozotocin was carefully chosen in the present study to prevent the confounding variable of intrauterine growth aberration, which may exert its own independent effect on fetal glucose transporters(55, 56). Our resultant model demonstrated no changes in fetal body or brain weights.

In a separate study where severe maternal diabetes was induced early in gestation on d 5, in utero growth retardation was associated with fetal hyperglycemia and hyperinsulinemia. These in utero perturbations of 15-d duration enhanced fetal brain Glut 1 expression(56). Because streptozotocin treatment alone in the absence of maternal diabetes was not compared with the streptozotocin-treated diabetic group, it is difficult to determine whether this observed effect is independent of the chemical. The observations of this study contrast our present results, where no significant in utero growth retardation was associated with fetal hyperglycemia, modest fetal hypoinsulinemia, and no substantive changes in brain glucose transporter concentrations beyond that noted with streptozotocin treatment alone. These observations do not rule out the presence of a protein conformational change leading to an alteration in the fetal brain glucose transporter affinity.

Previous investigations suggested that some pregnant animals when treated with streptozotocin on d 5 of gestation, although not overtly diabetic, had abnormal glucose tolerance values. However, the presence of abnormal maternal glucose tolerance tests was not associated with any changes in fetal glucose and insulin values(57), and tissue (including the brain) glucose utilization rates(58). Conversely, changes seen more in Glut 3 than Glut 1 protein levels may have resulted from the direct chemical effects of streptozotocin. Evidence of neuronal specific toxicity is widely reported supporting a more pronounced effect upon the neuronal specific Glut 3(25) rather than Glut 1, which is detected in microvessels and cortical membranes(24, 41). The mechanism by which streptozotocin achieves tissue-specific targeting is unknown(62, 63), but this chemical has been noted to adversely affect the function of specific neuronal cells directly(5961). This cell-specific alteration may be related to the fact that streptozotocin causes DNA strand breaks(64, 65), and uncouples the process of oxidative phosphorylation by altering certain mitochondrial enzymes(66). Thus, cells with a higher oxidative metabolism, namely neurons, may be more vulnerable than others. Because no detectable changes in either Glut 1 or Glut 3 expression were noted pretranslationally at the mRNA level, the effect of streptozotocin appears to be on the translational and/or posttranslational processes regulating fetal brain glucose transporter expression.

The changes in fetal brain Glut 3 (and Glut 1) in the STZ-ND group being similar to the SEVERE-D rather than the vehicle control group supports the argument of an adverse effect of streptozotocin upon the fetal neural cells. This is unlike effects upon nonneural fetal tissues as previously described(39). Because embryonic rat neurons and pancreaticβ-islet cells share a common cellular lineage(67), it is feasible that these two cell types are more vulnerable to the chemical effects of streptozotocin. This is seen here as a more pronounced effect upon the neuron-specific Glut 3 rather than the microvascular and cortical Glut 1 along with fetal hypoinsulinemia. Our present study supports the need to further examine the direct effects of streptozotocin independent of glucose abnormalities upon neuronal glucose transporters.

In conclusion, the results demonstrated in this short-term chemically induced severe diabetic model parallel our previous observations in the NOD mouse model and support no substantive effect upon fetal brain glucose transporter expression. The fact that in vivo maternal diabetes, independent of streptozotocin fails to significantly affect, whereas streptozotocin independent of maternal diabetes alters fetal brain glucose transporter concentrations, advises caution in interpretation of studies employing streptozotocin to induce maternal diabetes and examining embryonic neural or pancreatic tissues without inclusion of the chemical control. In addition, our results dictate a need for examining in the future the intracerebroventricular (in vivo) and in vitro direct effects of streptozotocin independent of glucose upon microvascular and neuronal glucose transporter expression.