Characterization of the Concurrent Metabolic Changes in Brain and Plasma During Insulin-Induced Moderate Hypoglycemia Using ¹H NMR Spectroscopy in Juvenile Rats

Abstract

Treatment of hypoglycemia in children is currently based on plasma glucose measurements. This approach may not ensure neuroprotection since plasma glucose does not reflect the dynamic state of cerebral energy metabolism. To determine whether cerebral metabolic changes during hypoglycemia could be better characterized using plasma metabolomic analysis, insulin-induced acute hypoglycemia was induced in 4-week-old rats. Brain tissue and concurrent plasma samples were collected from hypoglycemic (N=7) and control (N=7) rats after focused microwave fixation to prevent post-mortem metabolic changes. The concentration of 29 metabolites in brain and 34 metabolites in plasma were determined using ¹H NMR spectroscopy at 700 MHz and examined using partial least squares-discriminant analysis. The sensitivity of plasma glucose for detecting cerebral energy failure was assessed by determining its relationship to brain phosphocreatine. The brain and plasma metabolite profiles of the hypoglycemia group were distinct from the control group (brain: R²=0.92, Q²=0.31; plasma: R²=0.95, Q²=0.74). Concentration differences in glucose, ketone bodies and amino acids were responsible for the intergroup separation. There was 45% concordance between the brain and plasma metabolite profiles. Brain phosphocreatine correlated with brain glucose (control group: R²=0.86; hypoglycemia group: R²=0.59; p<0.05), but not with plasma glucose. The results confirm that plasma glucose is an insensitive biomarker of cerebral energy changes during hypoglycemia and suggest that a plasma metabolite profile is superior for monitoring cerebral metabolism.

1. Introduction

Hypoglycemia is common in infants and children. Inadequately treated hypoglycemia leads to cerebral energy failure and neuronal injury [1]. Therefore, early detection and correction of cerebral energy failure is the primary goal of treatment. At present, hypoglycemia is managed based on plasma glucose levels [2, 3]. While this approach is easy to implement in clinical practice, it may not ensure neuroprotection. Plasma glucose levels do not reflect the dynamic cerebral energy metabolism during hypoglycemia [4, 5]. Although glucose is its primary energy substrate, the developing brain is capable of energy production using alternative substrates such as ketone bodies and lactate [4–8]. Once these substrates have been depleted, brain glutamine and glutamate are used to support energy production via the tricarboxylic acid (TCA) cycle [5, 9]. Cerebral energy failure occurs only after all alternative energy sources have been depleted [5]. A comprehensive knowledge of the cerebral metabolic changes is necessary to ensure energy sufficiency during hypoglycemia and prevent brain injury.

In vivo ¹H NMR spectroscopy (MRS) is a robust method for monitoring brain metabolism during hypoglycemia [5, 10]. However, in vivo MRS is not practical in clinical settings. Methods that combine the sensitivity of in vivo MRS with the practicality of easily accessible biofluids are necessary. ¹H MRS-based plasma metabolomic analysis provides a comprehensive insight into the dynamic state of metabolism and has been used for monitoring altered cerebral metabolism in adult humans and animal models [11–16]. Limited data demonstrate a 30–70% concordance between the plasma and brain metabolomic profiles [17]. The objective of the present study was to determine the concurrent metabolic changes in brain and plasma during hypoglycemia using ¹H MRS in 4-week-old rats. Rats of this age are used to model the effects of hypoglycemia in young children due to similarities in the stage of brain development, substrate preference and vulnerability to hypoglycemia-induced injury [18–20].

2. Material and Methods

2.1. Animal Preparation

Four-week-old male and female Sprague Dawley (average body weight: 150 gm) rats were used. Animals were housed under standard laboratory conditions and allowed ad libitum food and water. The University of Minnesota Institutional Animal Care and Use Committee approved the study.

2.2. Induction of Acute Hypoglycemia

Littermates were randomly assigned to control (Control group; n=7) and hypoglycemia (HG group; n=7) groups. Rats in the HG group were subjected to acute hypoglycemia using human regular insulin, 10 U/kg i.p. after overnight fasting as previously described [19, 21]. The target blood glucose concentration was <2.5 mmol/l (<40 mg/dl) based on previous studies [19, 21]. Rats in the Control group were similarly fasted and injected with an equivalent volume of normal saline. Fasting was continued and the ambient temperature was maintained at 34°C. Blood glucose concentration was measured every 15 min in tail vein samples using a glucometer (Accu-Chek Compact Plus, Roche, Indianapolis, IN). Animals were continuously monitored for seizures and coma.

2.3. Plasma and Brain Tissue Collection

Animals were killed 240 min after insulin or normal saline injection using focused microwave fixation (4 kW for 1.1 sec; Gerling Applied Electronics, Modesto, CA) under brief (<30 sec) isoflurane anesthesia. This method rapidly inactivates enzymes and prevents post-mortem changes in metabolite concentrations [22–24]. The brain was collected. A blood sample was obtained and the plasma was separated. Brain and plasma samples were stored at −80°C.

2.4. ¹H MRS-Based Metabolomic Analysis

Metabolite profiles were determined using published methods with minor modifications [11, 12, 14, 15, 25]. Brain tissue was weighed and digested in 3 ml ice-cold perchloric acid containing 1M trimethylsilylpropionic acid (TSP; Sigma-Aldrich, St. Louis, MO). After centrifugation (3000g for 15 min at 4°C × 2), the supernatant was neutralized with 200 μl of 9M KOH and pH was adjusted to 7.0±0.1. A chelating ion exchange resin (Chelex^® 100 Resin; Bio-Rad, Hercules, CA) was added to remove paramagnetic compounds. Aliquots of 250μl were frozen, lyophilized overnight and resuspended in distilled H₂O and 10% D₂O for a final volume of 60 μl. One-dimensional ¹H NMR spectra (128 scans) were acquired using a 700 MHz NMR spectrometer (Bruker Avance, Billerica, MA) with a 1.7-mm TCI ¹H-enhanced cryoprobe and 1D-nuclear Overhauser effect pulse sequence [26]. Metabolites were identified by their chemical shift in relation to the TSP peak (δ = 0.0 ppm) and the concentrations were determined (Chenomx NMR Suite 8.0; Chenomx Inc., Edmonton, Alberta, Canada).

For plasma analysis, D₂O and 1mM of 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS; Sigma-Aldrich, St. Louis, MO) were added [27] to achieve a final volume of 550 μl. One-dimensional ¹H NMR spectra (64 scans) were acquired using the 700 MHz NMR spectrometer with a 5-mm TCI ¹H-enhanced cryoprobe and Carr-Purcell-Meiboom-Gill presaturation pulse sequence with a spectral width of 10 kHz [26]. A line broadening of 0.3 Hz was applied, followed by fast Fourier transformation, autophasing and autobaseline corrections using a software program (TopSpin; Bruker, Billerica, MA). Metabolites were identified by their chemical shift in relation to the DSS peak (δ=0.0 ppm) and the concentrations were determined.

2.5. Statistical Analysis

A web server designed for comprehensive metabolomic data analysis was used (MetaboAnalyst 3.0; www.metaboanalyst.ca) [28]. Intergroup differences in brain and plasma metabolites were determined using unpaired t-tests. Multivariate analysis was performed using partial least squares - discriminant analysis (PLS-DA) to quantify the contribution of each metabolite to the classification of the groups. Metabolite concentrations were log-transformed and scaled by subtracting the mean and dividing by the standard deviation before PLS-DA using the R software (https://www.R-project.org). Data are presented as mean±SEM. Statistical significance was set at p<0.05.

3. Results

3.1. Blood Glucose Concentration

The target blood glucose concentration (<2.5 mmol/l) was reached 30 min after the insulin injection and was maintained within the target range until tissue collection at 240 min (210 min of hypoglycemia). The mean blood glucose concentration from the period of insulin administration until tissue collection was lower in the HG group (Control group, 4.4±0.1 mmol/l [79.9±2.0 mg/dl]; HG group, 2.0±0.2 mmol/l [36.9±4.1 mg/dl]; p<0.001). Hypoglycemic rats were less active, but conscious and seizure-free. There was no mortality.

3.2. Metabolomic Analysis

High-resolution ¹H NMR spectra of the brain extract and plasma were obtained from all 14 animals (Supplemental Figure 1, online). The concentration of 29 brain metabolites and 34 plasma metabolites were quantified (Supplemental Table 1, online).

3.2.1. Univariate Analysis

Relative to the Control group, brain alanine, glucose, β-hydroxybutyrate, lactate, threonine and valine concentrations were lower in the HG group (p<0.05; Table 1). In the plasma, acetoacetate, glucose, glutamine, β-hydroxybutyrate, 3-hydroxy-3-methylglutarate and isoleucine concentrations were lower in the HG group (p<0.05; Table 1).

Table 1:

Brain and Plasma Metabolites Altered in the Hypoglycemia Group on Univariate Analysis

	Metabolite	Control Group	Hypoglycemia Group*
Brain	Alanine	0.58 ± 0.08	0.25 ± 0.03
	Glucose	0.59 ± 0.20	0.10 ± 0.02
	β-hydroxybutyrate	0.36 ± 0.04	0.23 ± 0.04
	Lactate	3.94 ± 0.74	1.82 ± 0.41
	Threonine	0.63 ± 0.13	0.17 ± 0.05
	Valine	0.10 ± 0.01	0.04 ± 0.01
Plasma	Acetoacetate	0.53 ± 0.08	0.16 ± 0.04
	Glucose	4.44 ± 0.13	2.06 ± 0.24
	Glutamine	0.88 ± 0.06	0.57 ± 0.09
	β-hydroxybutyrate	2.88 ± 0.34	0.34 ± 0.13
	3-hydroxy-3-methylglutarate	0.16 ± 0.03	0.04 ± 0.02
	Isoleucine	0.22 ± 0.02	0.12 ± 0.03

Values are mean ± SEM metabolite concentrations in μmol/ml; n=7 per group.

^*p<0.05 vs. Control group for all metabolites in the brain and plasma (unpaired t tests).

3.3.2. Multivariate Analysis – Brain

The brain metabolomic profile was constructed using the 29 quantified metabolites. PLS-DA demonstrated clear separation of the Control and HG groups (R²=0.92; Q²=0.31; Figure 1A). Loading plots (Figure 1B) demonstrated that alanine, valine, threonine, glucose, β-hydroxybutyrate, lactate and glutamine (lower in HG group), and phenylalanine and τ-methylhistidine (higher in HG group) were responsible for the intergroup separation (variable importance in projection [VIP] score ≥ 1.0 for each; Supplemental Figure 2, online).

Figure 1.

Partial least squares-discriminant analysis (PLS-DA) score and loading plots of the brain (A, B) and plasma (C, D) metabolomic profiles in the control and hypoglycemia groups. A and C: PLS-DA scores plot of principal component (PC) 1 and PC2 showing clustering of the metabolomic profiles separating the control and hypoglycemia groups in brain (R²=0.92; Q²=0.31; A) and plasma (R²=0.92; Q²=0.31; C). Control group, filled circles (n=7); hypoglycemia group, open triangles (n=7). B and D: Metabolite loadings plot corresponding to the scores plot shows metabolites responsible for the intergroup separation in brain (B) and plasma (D).

3.3.3. Multivariate Analysis – Plasma

PLS-DA demonstrated clear separation of the plasma metabolite profiles of the Control and HG groups (R²=0.95; Q²=0.74; Figure 1C). Loading plots (Figure 1D) revealed that β-hydroxybutyrate, glucose, acetoacetate, glutamine, isoleucine, proline, 3-hydroxy-3-methylglutarate and α- hydroxybutyrate (lower in HG group), and phenylalanine (higher in HG group) were responsible for the intergroup separation (VIP score ≥ 1.0 for each; Supplemental Figure 2, online).

3.4. Relationship between Glucose and Phosphocreatine

Management of hypoglycemia is currently based on plasma glucose measurements. The sensitivity of plasma glucose for detecting cerebral energy failure was assessed by determining the relationship between plasma glucose and brain phosphocreatine, which indexes energy reserves in the brain [29]. Brain phosphocreatine positively correlated with brain glucose (Control: R²=0.86; HG: R²=0.59; p<0.05), but not with plasma glucose in either group.

4. Discussion

MRS-based metabolomic analysis demonstrates that acute insulin-induced moderate hypoglycemia induces extensive and parallel metabolic changes in the brain and plasma of juvenile rats. The data confirm the well-known changes in glucose and ketone bodies during hypoglycemia, and demonstrate significant alterations in the amino acid concentrations in brain and plasma. The results also demonstrate that plasma glucose is a poor biomarker of the cerebral energy changes in hypoglycemia.

The metabolites primarily responsible for the intergroup stratification (as determined by VIP score ≥1.0) in the brain and plasma were glucose, ketone bodies and amino acids. Four of the nine metabolites (glucose, β-hydroxybutyrate, glutamine and phenylalanine) were common to the two compartments. The 45% plasma vs. brain concordance is within the 30–70% concordance reported in a previous study [17]. Most of the differentiating metabolites had lower concentrations in the hypoglycemia group, suggesting their utilization for energy production. It is well known that the developing brain can use ketone bodies and lactate during neuroglycopenia [6–8]. The lower concentration of acetoacetate, α- and β-hydroxybutyrate and 3-hydroxy-3-methylglutarate in the plasma, and β-hydroxybutyrate and lactate in the brain is consistent with this possibility. The liver mitochondria in rats are capable of transforming 3-hydroxy-3-methylglutarate to acetoacetate [30], which can be transported across the blood-brain barrier and used for energy production. The capacity for ketone body uptake and utilization is maximal in the pre-weaning period [4]. However, it is still substantial in the post-weaning period. β-hydroxybutyrate administration attenuates hypoglycemia-induced brain injury in 4-week-old rats [21].

The results demonstrate that multiple amino acids are altered during hypoglycemia. Amino acid concentrations are decreased in other causes of hypoglycemia (e.g., due to starvation) [31, 32]. Lower plasma and brain glutamine concentration is consistent with the utilization of this amino acid in the TCA cycle once glucose and ketone bodies have been exhausted [5, 9, 10]. Likewise, the lower concentration of isoleucine and proline in plasma, and alanine, threonine and valine in brain, may imply their utilization for energy production [33]. Plasma branched amino acid concentrations decrease initially during insulin-induced moderate hypoglycemia, but increase in the later stages despite stable hyperinsulinemia. This response, likely mediated by the counterregulatory hormones, counters the antiproteolytic action of insulin and promotes gluconeogenesis, net hepatic glucose output and amino acid oxidation [34–36]. Of note, the branched chain amino acids, leucine, isoleucine and valine supply the amide group for glutamate and glutamine synthesis. Depletion of these amino acids delays the post-hypoglycemic recovery of glutamine and glutamate concentrations in the brain [5, 9]. Phenylalanine also could be used for energy production after conversion to tyrosine [33]. Thus, phenylalanine concentration should have decreased in the hypoglycemia group. Instead, increased levels were found in plasma and brain. A parsimonious explanation is that phenylalanine to tyrosine conversion was inhibited by the antagonistic action of insulin on phenylalanine hydroxylation [37].

Brain phosphocreatine, an index of tissue energy reserves, had a robust relationship with brain glucose. This is not surprising given that glucose is the primary energy substrate to the rat brain at this age [4]. The lack of correlation between plasma glucose and brain phosphocreatine in either group suggests that plasma glucose is an insensitive biomarker of the cerebral energy changes and challenges the validity of current plasma glucose-based treatment strategies in children. Unlike our previous study [5], there was no relationship between brain glutamate and phosphocreatine, likely because the brain tissue was harvested prior to the onset of glutamate decrease. A pronounced and parallel decrease in brain glutamate and phosphocreatine commences only after the complete depletion of brain glutamine [5]. Brain glutamine concentration was 34% of the control group in the hypoglycemia group (0.48±0.07 μmol/g vs. 1.42±0.45 μmol/g; P=NS) in the present study.

Our study has several limitations. The sample size may be considered small; but it is comparable to previous ¹H MRS-based metabolomic studies in brain injury models [38, 39]. The number of metabolites identified (29 in brain and 32 in plasma) is higher than the numbers (6–15) reported in those studies and supports our approach. Our results are specific to insulin-induced moderate hypoglycemia. A more severe hypoglycemia may cuase additional metabolic changes, particularly in markers of energy failure and purine metabolism [9]. Similarly, the metabolic changes in other causes of hypoglycemia, such as failure of glucose production (e.g., starvation, panhypopituitarism and glycogen storage disorders) or increased glucose utilization due to lack of alternative substrates (e.g., fatty acid oxidation disorders) may be different. Recurrent hypoglycemia also has adverse metabolic and functional outcomes [20], but was not part of the study. Other limitations include lack of dose-response and time course assessments, and non-inclusion of neuronal injury and neurodevelopment assessments, all of which were not possible with our study design. Finally, LC-MS/MS-based analysis might have uncovered additional metabolites. However, ¹H MRS offers many advantages over the LC-MS/MS method, including not requiring elaborate sample preparation [40] and a lower cost.

5. Conclusions:

Our study confirms that plasma glucose by itself is a poor biomarker of the cerebral energy changes during hypoglycemia and suggests that a plasma metabolite profile may be superior for this purpose. The interspecies differences in rats and humans in brain energy metabolism and substrate utilization preclude direct extrapolation of the results to humans. Validation of this approach in human infants and children, preferably through longitudinal assessment during hypoglycemia and after its treatment, may improve clinical care of those at risk for hypoglycemia.

Abbreviations

DSS, 4

4-dimethyl-4-silapentane-1-sulfonic acid

HG

hypoglycemia

MRS

NMR spectroscopy

PLS – DA

partial least squares - discriminant analysis

TCA

tricarboxylic acid

TSP

trimethylsilylpropionic acid

VIP

variable importance in projection