Correlation of Hippocampal Glucose Oxidation Capacity and Interictal FDG-PET in Temporal Lobe Epilepsy


Address correspondence and reprint requests to Dr. S. Vielhaber at Department of Neurology II, University of Magdeburg Medical Center, Leipziger Str. 44, D-39120 Magdeburg, Germany. E-mail:


Summary:  Purpose: Interictal [18F]fluorodeoxyglucose (FDG) positron emission tomography (PET) demonstrates temporal hypometabolism in the epileptogenic zone of 60–90% of patients with temporal lobe epilepsy. The pathophysiology of this finding is still unknown. Several studies failed to show a correlation between hippocampal FDG-PET hypometabolism and neuronal cell loss. Because FDG is metabolized by hexokinase bound to the outer mitochondrial membrane, we correlated the glucose-oxidation capacity of hippocampal subfields obtained after surgical resection with the corresponding hippocampal presurgical FDG-PET activity.

Methods: In 16 patients with electrophysiologically confirmed temporal lobe epilepsy, we used high-resolution respirometry to determine the basal and maximal glucose-oxidation rates in 400-μm-thick hippocampal subfields obtained after dissection of human hippocampal slices into the CA1 and CA3 pyramidal subfields and the dentate gyrus.

Results: We observed a correlation of the FDG-PET activity with the maximal glucose-oxidation rate of the CA3 pyramidal subfields (rp = 0.7, p = 0.003) but not for the regions CA1 and dentate gyrus. In accordance with previous studies, no correlation of the FDG-PET to the neuronal cell density of CA1, CA3, and dentate gyrus was found.

Conclusions: The interictal hippocampal FDG-PET hypometabolism in patients with temporal lobe epilepsy is correlated to the glucose-oxidation capacity of the CA3 hippocampal subfield as result of impaired oxidative metabolism.

With [18F]fluorodeoxyglucose positron emission tomography (FDG-PET), interictal hypometabolism is found among 60–90% of patients with the clinical syndrome of temporal lobe epilepsy (TLE) (1–3). FDG-PET has become an accepted tool for localization of temporal lobe foci and for prediction of surgical outcome (4–7). FDG-PET hypometabolism is proposed to be associated with the presence of underlying mesial temporal sclerosis (MTS) and seems to correlate with the degree of hippocampal volume loss measured by volumetric magnetic resonance imaging (MRI) (8,9). Conversely, several studies failed to show a clear relation between hippocampal neuronal cell loss, associated atrophy, and interictal hypometabolism (10–12). The pathophysiologic mechanism accounting for the FDG-PET hypometabolism is therefore still under discussion. The purpose of our study was to examine the existence of a relation between regional FDG metabolism in PET scans and intrinsic glucose metabolism in surgical specimens in TLE patients. Because FDG is metabolized in the brain by hexokinase bound to the outer mitochondrial membrane (13), we tried to correlate the glucose-oxidation capacity quantified in resected hippocampal subfield specimens with the corresponding hippocampal FDG-PET in patients with TLE.


Patient population

Patients corresponding to the following criteria were included:

  • 1Drug-resistant TLE with confirmed single epileptogenic zone by presurgical evaluation according to the Bonn protocol (14), which included, in addition to FDG-PET, (a) neurologic and neuropsychological examination, (b) video-EEG monitoring with scalp and sphenoidal electrodes or subdural electrodes (see later), (c) cerebral high-resolution MRI, and (d) intracarotid amobarbital procedure (Wada test).
  • 2No other neurologic condition than epilepsy.
  • 3No reported partial or generalized seizures within 48 h before PET or during FDG uptake.
  • 4Excised temporal lobe tissue suitable in quantity and quality for neurochemical and histologic workup.

Sixteen patients (seven female patients; age range, 13–63 years; median age, 40 years) meeting these criteria were included. The clinical and laboratory data are summarized in Table 1. Clinical neurologic examination was unremarkable in all investigated subjects, and the patients were receiving carbamazepine (CBZ) monotherapy or had add-on therapy with lamotrigine (LTG). MR scans were carefully evaluated, especially for the presence of putative abnormalities of left and right hippocampal formations (asymmetry of size, signal alterations) (15). Seizures of unilateral temporal origin were demonstrated by continuous interictal and ictal video-EEG monitoring with scalp and sphenoidal electrodes in 10 patients and additionally by intracranial recordings in six patients (no. 5, 6, 9, 13, 15, and 16). All patients underwent operations on the side of the seizure origin (16). The study was approved by the University of Bonn Medical Ethical Committee.

Table 1.  Clinical and laboratory data
no/sex/age (yr)
Age at
onset of
epilepsy (yr)
ictal onset
MRIPathologySurgery; outcomea; interval (mo)
  • Temp, temporal; l, left; r, right; HS, hippocampal sclerosis; AHS, Ammon's horn sclerosis; SAH, selective amygdala-hippocampectomy; Ant resect, anterior resection; HE, hippocampectomy; 1, simple partial seizures; 2, complex partial seizures; 3, secondarily generalized tonic–clonic seizures; mo, month; yr, year.

  • a

     Classification of postoperative outcome (cit>31).

  • b  Temporal lobe specimen showed a pseudocystic lesion or cmultiple sclerotic zones but nearly normal hippocampal formation.

 1/F/6311,2,3Temp LHS RAHSSAH R; IId; 22
 2/M/41131,2Temp LHS LAHSSAH L; Ia; 24
 3/M/34162,3Temp LHS LAHSSAH L; Ia; 24
 4/M/30191,2Temp RHS RAHSSAH R; Ia; 6
 5/F/39111,2,3Temp LHS LAHSAnt 2/3 resect. + HE L; IIb; 24
 6/M/4231,2,3Temp RHS RAHSSAH R; IIb; 24
 7/M/130.61,2,3Temp RHS RAHSSAH R; Ia; 6
 8/F/34111,2,3Temp RHS RAHSSAH R; Ia; 24
 9/M/2381,2,3Temp LHS LAHSTemp. pole resect. + HE L; Ia; 13
10/F/41372Temp LHS LAHSSAH L; IIa; 24
11/M/56431,2,3Temp LHS LAHSSAH L; IId; 24
12/M/47151,2,3Temp RHS RAHSSAH R; Ia; 4
13/M/43181,2,3Temp RNormalNormalAnt 2/3 resect. + HE R; IIIa; 18
14/F/34131,2,3Temp RCystic defect temp RExtrahipp. lesionsbAnt 2/3 resect. + HE R; IVb; 24
15/F/28101,2,3Temp RNormalNormalSAH R; Ia; 24
16/F/4281,2Temp RResidual of meningitisExtrahipp. lesionscAnt. 2/3 resect. + HE R; Ia; 12

PET procedures

Data acquisition

PET scans were performed on a high-resolution head-dedicated PET camera (ECAT EXACT 47; Siemens, Erlangen, Germany) with an axial full field of view of ∼16.2 cm (5.8-mm in-plane and 5-mm axial resolution), which provides 31 slices at 3.37-mm intervals. PET was done 30 min after intravenous injection of 185 MBq of [18F]fluorodeoxyglucose (FDG)/70 kg body weight. Before the injection of the tracer, the fasting blood glucose level was assured to be within physiologic limits, and the patients were placed comfortably on a couch in a quiet room with the lights dimmed to a minimum. The patients were asked to keep their eyes closed. During the period of image acquisition, the patient was carefully monitored for head movements and immediately repositioned if necessary. Data acquisition consisted of a 30-min emission scan of the brain and a 10-min transmission scan for attenuation correction with a germanium 68 source. Axial, coronal, and sagittal slices as well as slices angulated perpendicular or parallel to the longitudinal axis of the hippocampal body were reconstructed.

Data analysis

For interindividual comparison PET scans were spatially normalized by SPM 99 software (Wellcome Department of Cognitive Neurology, London, U.K., Extraction of the signal from background noise was performed by using spatially normalized images with a thresholding technique (AnalyzePC 3.0 software; Biomedical Imaging Resource, Mayo Foundation, Rochester, MN, U.S.A.). Mean cerebral activity was normalized according to a previously described technique (17). To evaluate regional metabolism differences, we defined volumes of interest (VOIs) by manually segmenting the hippocampal region of each hemisphere on a SPM MNI-template, which was used for normalization in the previous steps. Because asymmetry indices have been found to be more sensitive to metabolic alterations in unilateral TLE than the absolute values and regional-to-bihemispheric ratio averages (3,18–20), we calculated the difference between the activity of hippocampal VOIs of the epileptogenic (ipsilateral) and nonepileptogenic (contralateral) side in our patients according to the formula: [(VOI ipsilateral – VOI contralateral)/(VOI ipsilateral + VOI contralateral)]× 200 (20). These differences were then used in further analyses. Furthermore, the biochemical data and the normalized ipsilateral activity of the hippocampal VOIs were compared.

The investigator (J.v.O.) segmenting the hippocampus was blinded to the results of the oxygen consumption analysis (see later) and the clinical and presurgical laboratory data of the patients.

Neuropathologic evaluation of hippocampal specimens

All TLE specimens were independently examined by two neuropathologists and classified with respect to the presence of Ammon's horn sclerosis (AHS, n =12; Table 1) or focal extrahippocampal lesions in the temporal lobe (n = 2; Table 1) or no pathology (n = 2; Table 1) according to established criteria (21). The neuronal cell loss in the hippocampal CA1, CA3 pyramidal subfields, and the dentate gyrus (in comparison to control hippocampal tissue from autopsy cases) was studied by semiquantitative cell count determinations in Nissl-stained sections of the paraffin-embedded hippocampus (22).

Biochemical studies

Slice preparation

After surgery the hippocampus was immediately placed in ice-cold medium [90 mM NaCl, 3 mM KCl, 2 mM MgSO4, 2 mM CaCl2, 1 mM sodium pyruvate, 10 mM glucose, 105 mM sucrose, and 10 mM Hepes buffer, pH 7.4 (4-2-hydorxyethyl-1-piperazine-2-ethanesulfonic acid)]. The 400-μm transverse sections were prepared with a vibratome (Leica, Germany) according to standard methods. The slices were then dissected into the hippocampal subfields [CA1 pyramidal subfield, the CA3 pyramidal subfield, and the dentate gyrus (DG)] and were transferred to a holding chamber where they were stored at 25°C in a standard carbogen-gassed Ringer's solution, containing 10 mM glucose, until further use. The individual subfields of one or two slices were evaluated for each subject. Before the experiments, the slices were preequilibrated for ≥60 min in the holding chamber.

Slice Respiration

To assess possible alterations in mitochondrial function of hippocampal samples from patients with therapy-resistant TLE, oxygen-uptake measurements were performed by applying high-resolution respirometry. This technique has been shown to allow a quantitative and hippocampal subfield–specific detection of mitochondrial respiratory chain dysfunction in patients with TLE (23). The maximal oxygen consumption in human brain microslices consisting of 400-μm-thick hippocampal subfields was determined at 30°C in a medium consisting of 125 mM NaCl, 3 mM KCl, 2 mM CaCl2, 1.25 mM sodium phosphate, 2 mM MgCl2, and 20 mM HEPES-NaOH (pH 7.4) with a PC-supported Oroboros high-resolution oxygraph (24). For determination of the rates of oxygen consumption of single hippocampal slices, the first derivate of the digitally acquired oxygen-concentration trace was used. The consumption rate was calculated as the average of datapoints of the first derivate trace after steady state until the next addition. The oxygen consumption was determined in each hippocampal subfield in the presence of 10 mM glucose (resting rate of respiration), and in the uncoupled state (maximal rate of glucose utilization), by addition of TTFB (4,5,6,7-tetrachloro-2-trifluoromethylbenzimidazole). The protein content was determined by using a protein assay kit based on Peterson's modification of the micro-Lowry method according to the instructions of the manufacturer (Sigma, Deisenhofen, Germany).

Statistical analysis

Multivariate analysis of variance (ANOVA) with the factors cell density in the three hippocampal subfields (CA1, CA3, and DG) was performed on multiple variables (neuronal cell density, glucose-oxidation rate, FDG-PET). Significant relations were compared in separate ANOVAs. Correlations between metric variables (FDG-PET and glucose-oxidation rates) were obtained by calculating of the nonparametric Spearman correlation coefficients (rs) and the corresponding two-tailed significance levels. When the observed correlation, as measured by the Spearman method, was significant, parametric Pearson's coefficients (rp) were chosen for graphic and statistical presentation of results. For all statistical analyses, the significance level was set at p < 0.05.


In this study we compared the presurgical hippocampal FDG-PET scan of 16 TLE patients with the ipsilateral oxidative glucose-consumption rates in individual hippocampal subfields (CA1, CA3, and DG regions) of the resected temporal specimens. A representative oxygraphic trace of a single 400-μm slice of the CA3 subfield from human hippocampus is shown in Fig. 1. The addition of 10 mM glucose (GLUC) increased the endogenous respiration of the slice as visible from the increasing slope of the oxygen content trace (upper trace) and the increasing level of the first derivative trace (lower trace), which directly corresponds to the rate respiration. This rate is proportional to the basal adenosine triphosphate (ATP) turnover of the corresponding hippocampal subfield. The addition of the inhibitor of the mitochondrial adenine nucleotide translocase, atractyloside (ATR), did not lead to a marked inhibition of respiration. We added 10 μM TTFB to the oxygraph chamber. This led to a marked increase in the slice respiration because of the protonophoretic action of this compound, causing a collapse of the mitochondrial membrane potential (uncoupling of mitochondrial oxidative phosphorylation). The further improvement of substrate supply by 10 mM pyruvate (PYR) only slightly increased the rate of respiration. For the quantitative evaluation of the oxidation rates, we determined the amount of oxygen consumed per minute referred to the protein content of the sample (see later).

Figure 1.

Representative oxygen-consumption trace of a single 400-μm slice of the CA3 pyramidal subfield from human epileptic hippocampus surgically removed from a patient with Ammon's horn sclerosis (patient 10 in Table 1). Sequential additions were made to detect rate changes of oxygen consumption. The addition of glucose (GLUC, 10 mM) caused an increase in slice respiration (higher slope of the oxygen concentration curve and higher level of the first derivative, lower curve). By subsequent addition of atractyloside (ATR, 200 μM), an inhibitor of the mitochondrial adenine nucleotide translocase, no inhibition of slice respiration was achieved. The maximal glucose-supported slice respiration was induced by uncoupling with TTFB (4,5,6,7-tetrachloro-2-trifluoromethylbenzimidazole, 2 μM). Pyruvate (PYR, 10 mM) further stimulated the uncoupled respiration by improved mitochondrial supply of reducing equivalents. The respiration chamber contained 255 μg protein in 1.5 ml medium for oxygraphic determinations.

The abnormality of mitochondria in the CA3 region of hippocampus of TLE patients becomes apparent in succinate dehydrogenase (SDH)/cytochrome c oxidase (COX) histochemistry (Fig. 2). In contrast to nonepileptic controls (A), six of 12 patients with TLE due to AHS exhibited an intense blue staining of some neuronal cell somata within the CA3 region (B). Dense accumulation of the blue SDH reaction product at a week COX reactivity (brown) within cells is the major feature of mitochondrial dysfunction. These COX-deficient SDH-positive neurons were observed predominantly in the CA3 region, showing, in comparison to CA1, only a moderate neuronal cell loss [CA3, 74%± 12% vs. CA1, 90 ± 10% (p = 0.007)].

Figure 2.

Succinate dehydrogenase (SDH) staining in a patient (no. 10 in Table 1) with temporal lobe epilepsy (B) with increased cellular staining in CA3 subfield, indicating mitochondrial proliferation. Cytochrome c oxidase–deficient, SDH-positive pyramidal neurons in CA3 also are observed. Nonepileptic control subject (A). Bar = 50 μm.

To investigate the putative relation between neuronal cell density glucose-oxidation rates and FDG-PET, we performed a multivariate ANOVA with the factors neuronal cell density in the three hippocampal subfields (CA1, CA3, and DG). This analysis revealed neither significant main effects (CA1: F = 0.149; p = 0.708, CA3: F = 1.886; p = 0.203, DG: F = 0.084; p = 0.92) nor interactions, indicating that neither the PET nor the glucose-oxidation rates are dependent on the neuronal cell counts in the three investigated hippocampal subfields.

The relations between FDG-PET and basal/maximal glucose-oxidation rates were investigated in separate ANOVAs with the factors oxidation rate in the three hippocampal subfields (CA1, CA3, and DG). Neither a significant main effect (CA1: F = 0.285; p = 0.763, CA3: F = 0.265; p = 0.777, DG: F = 2.568; p = 0.168) nor interactions were found for the basal glucose-oxidation rates. However, a significant main effect was found for the maximal glucose-oxidation rates in the CA3 subfield (F = 6.778; p = 0.031), revealing a strong dependence of the PET activity on the maximal oxidation rate in this subfield.

Further to characterize this effect, we correlated maximal glucose-oxidation rates of the individual subfields with the PET findings. As shown in Fig. 3B, a positive correlation of FDG-PET (asymmetry indices, AIs) with the maximal glucose-oxidation rates of the hippocampal region, was found to be significant for the CA3 region (rp = 0.7; p = 0.003). Conversely, neither ANOVA revealed any significant effect or interaction, nor was any correlation observed for the CA1 subfield (rp = 0.4; Fig. 3A) or for the DG (rp = –0.045; Fig. 3C). A comparable positive correlation with the maximal glucose-oxidation rates of the hippocampal region also was found for the normalized FDG-PET activity of the ipsilateral hippocampus in the CA3 subfield (rp = 0.61; p = 0.012) but again not for the CA1 subfield or the DG (data not shown).

Figure 3.

Figure 3.

Relation of maximal glucose-oxidation rates of hippocampal subfields to hippocampal FDG-PET activity expressed as asymmetry indices (AJs) {[volume of interest (VOI) ipsilateral – VOI contralateral]/[VOI ipsilateral + VOI contralateral]}× 200 (20). The term ipsilateral is related to the epileptogenic hippocampus. There is a positive correlation for the region CA3 (rp = 0.7; p = 0.003; B) but not for the subfields CA 1 (rp = 0.4; A) and dentate gyrus (DG; rp = –0.045; C). The numbers in the scatterplot correspond to the patients in Table 1.

Figure 3.

Figure 3.

Relation of maximal glucose-oxidation rates of hippocampal subfields to hippocampal FDG-PET activity expressed as asymmetry indices (AJs) {[volume of interest (VOI) ipsilateral – VOI contralateral]/[VOI ipsilateral + VOI contralateral]}× 200 (20). The term ipsilateral is related to the epileptogenic hippocampus. There is a positive correlation for the region CA3 (rp = 0.7; p = 0.003; B) but not for the subfields CA 1 (rp = 0.4; A) and dentate gyrus (DG; rp = –0.045; C). The numbers in the scatterplot correspond to the patients in Table 1.

Figure 3.

Figure 3.

Relation of maximal glucose-oxidation rates of hippocampal subfields to hippocampal FDG-PET activity expressed as asymmetry indices (AJs) {[volume of interest (VOI) ipsilateral – VOI contralateral]/[VOI ipsilateral + VOI contralateral]}× 200 (20). The term ipsilateral is related to the epileptogenic hippocampus. There is a positive correlation for the region CA3 (rp = 0.7; p = 0.003; B) but not for the subfields CA 1 (rp = 0.4; A) and dentate gyrus (DG; rp = –0.045; C). The numbers in the scatterplot correspond to the patients in Table 1.


In ∼60–90% of patients with TLE, glucose hypometabolism is observed with FDG-PET, and this method has become an accepted tool for localization of temporal lobe foci (1–3,25). It has been proposed that FDG-PET hypometabolism is closely associated with the presence of underlying MTS and correlates with the degree of hippocampal volume loss measured by means of volumetric MRI (8,9,26). Conversely, several studies failed to show a clear relation between hippocampal neuronal cell loss, which is correlated with hippocampal atrophy, and interictal FDG-PET hypometabolism (10–12). We examined the hypothesis that the hypometabolic pattern in patients with TLE is related to the oxidative glucose-consumption rate. Confirming the absence of any correlation of the hippocampal FDG-PET activity to hippocampal neuronal cell density, we found a significant correlation of FDG-PET activity to the glucose-oxidation rates of CA3 pyramidal subfield. This cell layer showed a moderate neuronal cell loss in MTS, which was much more pronounced in area CA1, whereas the granular cell layer of the dentate gyrus remained relatively preserved (27). Accordingly, the maximal glucose-oxidation rates were lower in the region CA1 than in CA3 or the DG. Nevertheless, the neuronal cell density is obviously not the determinant for the mesial temporal lobe epilepsy–associated alteration of glucose metabolism observed with FDG-PET. The uncoupled glucose-oxidation rate of hippocampal subfield, which corresponds roughly to the total mitochondrial capacity, shows a pronounced correlation to the FDG-PET, whereas neuronal cell density failed to do so. To explain this discrepancy, an intrinsic metabolic impairment leading to a decreased glucose metabolism in the vulnerable pyramidal cell layers of the sclerotic hippocampus must be proposed. Possible mechanisms could include a direct impairment of mitochondrial oxidative phosphorylation, as recently shown in the CA3 hippocampal subfield of patients with TLE (23). This putative mechanism is further supported by our histologic observations in patients with AHS showing mitochondrial proliferation in some CA3 pyramidal neurons (indicated by strong SDH reactivity) at low cytochrome c oxidase staining intensity. The presence of COX-negative neurons points to a direct impairment of enzymes of mitochondrial respiratory chain and is a characteristic feature of mitochondrial encephalopathies (28). However, COX-deficient SDH-positive neurons have been also reported in neurodegenerative diseases, like Alzheimer disease, which have been not associated with specific mtDNA mutations, suggesting that specific mutations are not necessarily required to observe this COX enzyme defect in distinct neurons (29).

Our data suggest that the intrinsic alteration of neuronal energy metabolism seems to be a stronger contributing factor than the neuronal cell count. Although a relation between hippocampal volume and glucose metabolism in temporal lobes has been observed (8,9) recent data suggest that this dependency seems to break down in the epileptic focus (30), which is consistent with our findings. In addition, the correlation of PET and CA3 oxygen consumption even in patients showing normal MRI and hippocampal histology supports the intrinsic metabolic concept and is in accordance with previous observations of abnormal PET with normal MRI in TLE (9).