In degenerative brain disorders characterized by neuronal cell loss, a decrease of N-acetyl aspartate (NAA) is frequently detected by noninvasive proton MR-spectroscopy (1H-MRS). Because NAA is almost exclusively found in neurons but not glial cells, this decrease is considered a marker for the neuronal loss (Simmons et al., 1991). However, recent MRS studies have failed to show a clear correlation between neuronal cell loss and NAA content in total hippocampus of patients with mesial temporal lobe epilepsy (MTLE) associated with Ammon's horn sclerosis (AHS) (Kuzniecky et al., 2001; Petroff et al., 2002). These were unexpected findings, since AHS entails a prominent neuronal cell loss and gliosis, predominantly in sectors CA1, CA3, and CA4 of the Ammon's horn (Cornu Ammonis, CA) (Margerison and Corsellis, 1966; Ben-Ari, 2001). An even more puzzling report demonstrated the reversibility of NAA reduction in MTLE patients after seizure control (Vermathen et al., 2002). Thus, it has remained unclear if the decrease of the “neuronal marker metabolite” NAA—which is consistently observed in sclerotic hippocampi—is really caused by the loss of pyramidal cells (Cendes et al., 1994; Woermann et al., 1999; Vermathen et al., 2002), or might be related to metabolic dysfunction (Bates et al., 1996, Clark, 1998; Vielhaber et al., 2003a). To address this problem, we used high-resolution 1H-MRS at 14.1 Tesla to study the distribution of NMR-visible metabolites in the individual hippocampal subfields of patients with AHS, taking in account the segmental hippocampal pathology. The data were compared to data from a smaller group of patients with lesion-caused MTLE, whose hippocampi did not exhibit the typical neuropathological features of AHS (Spencer, 1994; Clusmann et al., 2002).
Purpose: In patients with mesial temporal lobe epilepsy (MTLE) it remains an unresolved issue whether the interictal decrease in N-acetyl aspartate (NAA) detected by proton magnetic resonance spectroscopy (1H-MRS) reflects the epilepsy-associated loss of hippocampal pyramidal neurons or metabolic dysfunction.
Methods: To address this problem, we applied high-resolution 1H-MRS at 14.1 Tesla to measure metabolite concentrations in ex vivo tissue slices from three hippocampal subfields (CA1, CA3, dentate gyrus) as well as from the parahippocampal region of 12 patients with MTLE.
Results: In contrast to four patients with lesion-caused MTLE, we found a large variance of NAA concentrations in the individual hippocampal regions of patients with Ammon's horn sclerosis (AHS). Specifically, in subfield CA3 of AHS patients despite of a moderate preservation of neuronal cell densities the concentration of NAA was significantly lowered, while the concentrations of lactate, glucose, and succinate were elevated. We suggest that these subfield-specific alterations of metabolite concentrations in AHS are very likely caused by impairment of mitochondrial function and not related to neuronal cell loss.
Conclusions: A subfield-specific impairment of energy metabolism is the probable cause for lowered NAA concentrations in sclerotic hippocampi of MTLE patients.
Patients and Methods
Twelve patients (8 women, 4 men; age range: 25–66 years; median age: 41.5 years) with drug-resistant MTLE, who had been selected for epilepsy surgery according to the Bonn protocol (Elger et al., 1992), participated in this study (Table 1). The presurgical evaluation included (1) neurological and neuropsychological examinations, (2) video-EEG monitoring using scalp and sphenoidal electrodes or subdural electrodes, (3) cerebral high-resolution magnetic resonance imaging (MRI), and (4) an intracarotid amobarbital procedure (Wada test).
|Patient no/sex/age (y)||Age at onset of epilepsy (y)||Seizure type||EEG ictal onset||Drugs at surgery||MRI findings temporal lobe||MTLE pathology||Surgery; outcome*; interval (m)|
|1/F/57||27||1,2,3||Temp L||LEV||HS L||AHS||SAH L; IIIa; 24|
|2/F/41||1||2,3||Temp L||CBZ||HS L||AHS||SAH L, Ia; 12|
|3/F/66||18||2,3||Temp R||CBZ||HS R||AHS||SAH R, IIb; 16|
|4/F/45||5||2,3||Temp L||CBZ, LEV||HS L||AHS||SAH L; Ia; 6|
|5/F/54||15||2,3||Temp R||CBZ, LEV||HS R||AHS||SAH R, Ia; 12|
|6/M/37||16||2,3||Temp R||LTG||HS R||AHS||SAH R; Ib; 8|
|7/M/41||15||2,3||Temp R||CBZ||HS R||AHS||SAH R, Ia; 18|
|8/F/25||1||1,2,3||Temp R||CBZ, LEV||HS R||AHS||SAH R, Ia; 12|
|9/M/39||15||2,3||Temp L||CBZ||EH Cystic lesion L||Lesion||Ant 2/3 resect + HE L; IIa; 15|
|10/M/43||15||2,3||Temp R||CBZ, LTG||EH Tumor R||Lesion||SAH + LE R; Ia; 24|
|11/F/42||14||1,2,3||Temp R||CBZ, LEV||EH Tumor R||Lesion||SAH + LE R; Ib; 15|
|12/F/38||14||2,3||Temp L||LTG||EH Dysplastic lesion L||Lesion||SAH + LE L; Ia; 6|
The clinical neurological examination was unremarkable in all the patients studied. All were receiving either carbamazepine monotherapy or had add-on therapy with lamotrigine or levetiracetam (see Table 1). None had additional major medical complications (e.g., diabetes mellitus). The MTLE patients were divided into two groups according to their histopathological diagnosis (Table 1): lesion-caused MTLE (no histological signs of hippocampal sclerosis, n = 4) and AHS (severe hippocampal sclerosis, n = 8). Hippocampal sclerosis was diagnosed on Nissl-stained sections (Institute of Neuropathology, Bonn) and the cell densities were determined as described below. All patients gave informed consent before the surgery and for scientific studies of resected tissue. The research protocol was approved by the Human Medicine Ethical Committees of the University of Bonn and the University of Magdeburg.
Hippocampi were obtained with intact blood supply in the operating theatre and immediately placed in ice-cold oxygen-gassed medium (Kunz et al., 2000). Transverse sections of 400 μm thickness were prepared with a vibratome (Leica, Germany) from the middle of the hippocampus and adjacent regions for histological and neurochemical analysis. Slices were transferred to a holding chamber where they are equilibrated at 25°C in a carbogen-gassed Ringer solution, containing 125 mM NaCl, 3 mM KCl, 2 mM CaCl2, 2 mM MgCl2, 1.5 mM NaH2PO4, 26 mM NaHCO3 and 10 mM glucose. The living slices were microdissected within 15 min after the surgery into the hippocampal subfields CA1, CA3, the dentate gyrus (DG) and the parahippocampal gyrus (PH). Specimens of the hippocampal subfields consisting of 3–5 slices were snap frozen and stored in liquid nitrogen. In previous studies, this preparation technique has been used to obtain intact brain tissue specimens suitable for the quantitative measurement of mitochondrial respiration and for the determination of respiratory chain enzyme activities (Kudin et al., 1999; Kunz et al., 2000; Vielhaber et al., 2003b). Furthermore, it has been shown that the physiological intactness of the slices critically depends on the presence of bath glucose (Kudin et al., 1999). Specifically, the addition of 10 mM glucose is required to obtain the stability needed for quantitative physiological and biochemical measurements in living human brain slices.
In vivo MRS studies and data analysis
Presurgical in vivo single-voxel 1H-MRS data were available for the eight MTLE patients with AHS. A PRESS sequence (TR: 1,500 ms, TE: 135 ms, 256 averages) was chosen for volume selective spectroscopy. The rectangular voxel size was adapted to the entire hippocampal volume (2.0–6.0 cm3) on axial T1-weighted images to reduce partial volume effects due to surrounding cerebrospinal fluid. NAA, Cho, and Cr were quantified by the LCModel analysis (Provencher, 2001). MR spectra are characterized by three major peaks: N-acetyl aspartate (NAA) at 2.02 ppm, total creatine (creatine plus phosphocreatine) at 3.02 ppm, and choline at 3.22 ppm.
High-resolution MRS studies of excised tissue
For performing high-resolution in vitro 1H-MRS, individual hippocampal subfield specimens were weighed in the frozen state and homogenized at 0°C. Hydrophilic metabolites were extracted with 2.5 ml of cold 5% perchloric acid and centrifuged for 15 min at 4,000 g to remove precipitated proteins and membrane components. The supernatants were neutralized (to pH 7) with potassium hydroxide solutions. Precipitated potassium perchlorate was removed by centrifugation (10 min, 4,000 g) and the samples were lyophilized. The dried powder was dissolved in 600 μl deuterium oxide (D2O). The 1H-MR spectra were recorded on a Bruker DRX 600 spectrometer operating at 14.1 Tesla (600 MHz proton resonance frequency). The acquisition parameters for the proton spectra were: Flip-angle 90° (8.5 μs pulse length), 6.200 Hz spectral width, digital resolution 32 k data points, repetition time 10 s. The residual water signal was presaturated by a frequency selective pulse. Between 128 and 512 fee induction decays (FID) were accumulated for each spectrum depending on the wet weight of the investigated sample. The time-dependent signal was processed by applying a 0.5 Hz line broadening prior to Fourier transformation without zero filling. Metabolite concentrations were measured in units of mmol/kg(wet weight) with reference to an internal standard (sodium formiate), which was added to the D2O before dissolving the dry lyophylisate. All results are given as mean ± SEM.
Neuronal cell densities
All MTLE specimens were independently examined by two neuropathologists according to established criteria (Wolf et al., 1993), and classified with respect to the presence of AHS (n = 8; cf. Table 1) or a focal extrahippocampal lesion in the mesial temporal lobe without the histologic signs of AHS (n = 4; Table 1). The neuronal cell densities in the hippocampal subfields of areas CA1, CA3, CA4, and DG as well as in the PH (entorhinal region) were determined by cell counting in adjacent coronal sections of the paraffin-embedded hippocampus at mid or midposterior level (Kim et al., 1990; Foldvary et al., 1999; Vielhaber et al., 2003b). Estimation of neuronal density from Nissl-stained sections (4 μm) was made independently by two counters (A.P.K., S.V.) blind to their classification, using a CCD-camera coupled to a personal computer at a total magnification of 250-fold (pyramidal cells) and 500-fold (dentate granule cells). Pyramidal cells were counted in 200 × 400 μm areas at three evenly distributed sites within the subfields CA1, CA3, CA4, and PH. In the DG, granule cells were counted in 50 × 100 μm areas at six evenly distributed sites. The neuron numbers obtained were adjusted with Abercrombie's formula utilizing 12.2 μm for the nuclear diameter of pyramidal neurons and 9.2 μm for the dentate granule neurons (Abercrombie, 1946). The histological analysis of cell densities was validated by the determination of cell counts from haematoxylin and eosin-stained adjacent sections, which revealed no differences (data not shown). AHS was characterized by the extensive hippocampal cell loss and concomitant astrogliosis in CA1 and CA3. In lesion-caused MTLE patients (Table 1), focal lesions were found adjacent to the temporomesial neocortex but did not involve the hippocampus proper.
To assess directly possible alterations in mitochondrial function in hippocampi of MTLE patients with AHS or lesions, the activity of aconitase, a mitochondrial matrix enzyme that is sensitive to oxidative stress, was measured spectrophotometrically in the respective brain homogenates of CA1, CA3, DG, and PH by monitoring the formation of NADPH at 340 nm (Gardner and Fridovich, 1992). In addition, the activity of citrate synthase (CS), a mitochondrial marker enzyme, was determined by a standard method (Bergmeyer, 1970).
Statistical analyses were performed using SPSS 11. First a multivariate analysis of variance (ANOVA) was performed to compare alterations in the NMR-visible metabolites (NAA, succinate, lactate, glucose) within the hippocampal subfields (CA1, CA3, DG, PH) of the patients with AHS and lesion-caused MTLE. Prior to this analysis the data were tested and found to be normally distributed. The overall ANOVA with subsequent Bonferroni correction for multiple comparisons revealed significant main effects for the factors “disease group” (F1,160= 13.57, p < 0.0001) and “metabolite” (F3,160= 93.96, p < 0.0001) as well as interactions between “disease group” and “metabolite” (F3,160= 9.84, p < 0.0001) and between “hippocampal region” and “metabolite” (F9,160= 5.53, p < 0.0001). Further comparisons of the conditions defined by these factors were performed by means of t-tests to investigate the main effects and interactions in detail (see below).
Correlations between variables were obtained using Spearman rank-order correlation coefficients (rs) and their two-tailed significance levels were calculated. When the significant correlations were found with the Spearman method, parametric Pearson's correlation coefficients (rp) were chosen for the graphical and statistical presentation of results. The significance level was set at p < 0.05 for all statistical analyses. Results are presented as mean ± SEM.
Structural MRI and MRS findings
Presurgical clinical MRI of the MTLE patients revealed hippocampal sclerosis (AHS group) in eight patients and an extrahippocampal pathology in the mesial temporal lobe (lesion group) in four patients (Table 1). In agreement with previous studies in MTLE (Cendes et al., 1994; Woermann et al., 1999; Vermathen et al., 2002), presurgical single-voxel 1H-MRS of the entire human sclerotic hippocampus from the AHS group (n = 8) showed significantly lower NAA concentrations ipsilateral to pathology (7.65 ± 1.3 mmol/L) compared to the contralateral side (9.08 ± 1.06 mmol/L; p < 0.03), or compared to age-matched healthy controls (11.60 ± 2.13 mmol/L; p < 0.001). In vivo 1H-MRS data of the hippocampus in lesion-caused MTLE were not available for our four patients but NAA levels in the range of healthy controls have been reported for patients with seizures arising from extrahippocampal regions of the temporal lobes (Vermathen et al., 1997).
High-resolution MRS of tissue extracts in relation to histology and enzyme data
In comparison to lesion samples (Fig. 1A and C, filled bars) in the AHS specimens (Fig. 1B and C, open bars), the pyramidal neurons within the CA1 subfield were more severely depleted than in the CA3 subfield (p < 0.001). The pyramidal neurons of the CA3 subfield (including the CA2 subfield) are relatively preserved, similar to the dentate granule cells (Fig. 1B). In comparison to lesion-caused MTLE patients, we observed in our AHS patients the following pattern of neuronal cell loss: CA1 72%; CA3 36%; dentate granule cells 43%, and PH 33% (Fig. 1C).
Fig. 2 shows two representative high-resolution 1H-MR spectra from the hippocampal CA3 regions of an MTLE patient with AHS and a patient with lesion-caused MTLE. The statistical analysis of the full set of samples showed that the NAA concentration in specimens from the AHS group was significantly lower in the CA3 subfield relative to the lesion group (Fig. 3A). On the other hand, no significant differences were found in the adjacent areas CA1, DG, and PH. This decrease of NAA in the CA3 region of the AHS group compared to the lesion group cannot be explained by extreme values of individual measurements since all NAA concentrations in the CA3 region of the AHS group (1.26–4.26 mmol/kg(ww)) were consistently lower than the lesion group (4.93–9.79 mmol/kg(ww)).
The fact that the neuronal cell density was better preserved in the CA3 than the CA1 subfield (Fig. 1C) appears to contradict the MRS findings of a selective decline of the “neuronal marker” NAA in CA3 but not in CA1 (Figs. 2, 3A). Additionally, the loss of NAA in CA3 was accompanied by a substantial increase of lactate (Figs. 2 and 3A), which inversely correlates with the NAA concentration in this subfield (Fig. 4, left panel). However, in our study we observed elevated lactate levels in all the hippocampal sections compared to concentrations reported for in vivo conditions. This lactate increase could have been produced by our freezing protocol (Bachelard, 1989), or by the high glucose concentration employed in the tissue bath for increasing the stability of the living tissue slices (Kudin et al., 1999). To evaluate the latter possibility we studied the dependence of lactate concentration on bath glucose concentration in human hippocampal slices. As shown in Fig. 5, both lactate and (as expected) glucose concentrations in the tissue did depend on bath glucose levels. The lactate levels in the absence of high bath glucose concentrations (2.97 ± 0.29 mmol/kg (ww)) are in line with those typically reported in freeze-clamp studies of brain tissue (around 3.0 mM) (Bachelard, 1989; Jenkins et al., 1996). Therefore, the elevated lactate concentrations in our hippocampal slices can be attributed to the 10 mM bath glucose concentration. Additionally, it has to be underlined that the inverse correlation between NAA and lactate was restricted to the CA3 subfield, and that the increase of the lactate concentration in this subfield was significantly larger in the AHS than the lesion group (Fig. 3A). No dependence on bath glucose concentration was found with the other MR metabolites under investigation (Fig. 5). In addition to the significant increase in lactate also succinate concentrations were found to be elevated in subfield CA3 of AHS patients (Fig. 3B). Similar to lactate (see above), we detected an inverse correlation between the concentration of NAA and succinate (rP=−0.60; p = 0.013) in the CA3 subfield (Fig. 4, right panel), but not in the adjacent hippocampal areas (data not shown). These findings strongly suggest that the increased lactate concentrations in the CA3 subfield of AHS patients is not a result of increased glycolytic metabolism, but rather caused by a regional impairment of mitochondrial oxidative phosphorylation. The independent evidence of a regional compromised mitochondrial function in the AHS hippocampus comes from the analysis of activities of mitochondrial enzymes in the individual subfields. As shown in Fig. 6, the activity of the tricarboxylic acid cycle enzyme aconitase normalized to the mitochondrial marker enzyme CS revealed a CA3-specific decline, similarly as previously reported for respiratory chain complex I (Kunz et al., 2000).
NAA has been regarded as potential in vivo marker for neuronal cell loss and, therefore, has been used as a progression marker in various neurodegenerative disorders (Meyerhoff et al., 1994; Davie et al., 1995; Rooney et al., 1998; Sanchez-Pernaute et al., 1999). However, in several studies of MTLE patients MRS revealed a decline of the major neuronal marker NAA in epileptic hippocampi without pathological histology (Connelly et al., 1998; Woermann et al., 1999). In fact, in some cases an NAA recovery after seizure control has been reported (Vermathen et al., 2002). Additionally, even carefully conducted correlative studies between hippocampal NAA content and neuronal cell loss/gliosis in human MTLE have been inconclusive (Kuzniecky et al., 2001; Petroff, 2002). Here, we used high-resolution 1H-MRS to analyze the regional distribution of metabolites in the hippocampus of MTLE patients, taking into account the complex hippocampal pathology in AHS, and compared the distribution in AHS to the distribution found in cases of lesion-caused MTLE.
The most striking finding of our study is the unequal NAA distribution in different hippocampal subfields in AHS. In particular, we found an NAA loss of approximately 50% in the CA3 subfield of AHS patients compared to hippocampi from lesion-caused MTLE patients with extrahippocampal pathology. On the other hand, no significant changes of NAA concentrations were seen in the adjacent hippocampal subfields of CA1, DG, and PH. The subfield-specific loss of NAA in CA3 in the hippocampi of AHS patients is remarkable, because the extent of pyramidal cell loss (−36%) was comparable to that of dentate granule cells (−43%) and to the loss of neurons in PH (−33%), and in CA1 even larger (−72%). Thus the CA3 subfield-specific decrease of the neuronal marker NAA is in clear contrast to the pattern of neuronal cell loss that was observed by us and is commonly described in AHS (Margerison and Corsellis, 1966; Kim et al., 1990; Wolf et al., 1993; Ben-Ari, 2001).
On the other hand, the preserved concentrations of NAA in the CA1 subfield of AHS specimens raise the issue of whether NAA is also expressed in glial cells. Although astrocytes were reported to not contain NAA (Simmons et al., 1991; Urenjak et al., 1993), reactive astrocytes have been reported to express foetal characteristics of progenitor cells (Lin and Matesic, 1994), and previous 1H-MRS studies found significant amounts of NAA in oligodendrocyte-type 2 progenitors in culture (Urenjak et al., 1993). Therefore, the unexpected high NAA concentrations in the CA1 region of the AHS group despite more than 70% pyramidal cell loss might be explained by specific alterations of properties of glial cells in this hippocampal subfield, since these cells are known to express certain neuronal features in the hippocampus of MTLE patients (Hinterkeuser et al., 2000; Seifert et al., 2004). Interestingly, the specific activity of the enzyme from which NAA is synthesized— aspartate-N-acetyltransferase (Asp-NAT; EC 184.108.40.206)—has been shown to be similar in highly purified intact brain mitochondria prepared by gradient centrifugation and subfractions consisting of myelin (Madhavarao et al., 2003). In this context, it should be mentioned that recent studies indicate that NAA serves as an acetyl source for acetyl-CoA, which is required for lipid synthesis during myelination (Kirmani et al., 2003). Therefore, also densely packed myelinated axons might be the origin of the observed high NAA concentration in CA1 region of the hippocampi of AHS patients.
Another interesting point that merits some discussion is that NAA synthesis was reported to be mitochondrial (Patel and Clark, 1979) and, hence, a decrease of NAA is thought to reflect mitochondrial dysfunction (Clark, 1998). Elevated lactate concentrations in conjunction with increased succinate and glucose concentrations were observed specifically in the CA3 subfield of AHS hippocampus. This metabolite response cannot be explained by an increased rate of glycolytic metabolism, for example, by neuronal activation that would also result in increased lactate levels (Prichard et al., 1991). The concomitant increase of the citric acid cycle metabolite succinate and the precursor glucose directly indicates the decline of oxidative metabolism, which is in agreement with previous reports of segmental impairment of mitochondrial respiratory chain function in the CA3 subfield of the AHS hippocampus (Kunz et al., 2000; Vielhaber et al, 2003b), since succinate is a powerful substrate of respiratory complex II (succinate dehydrogenase complex) in brain mitochondria. Therefore, increased succinate levels in the CA3 subfield of AHS can be explained by an inhibition of succinate oxidation caused by an impairment of downstream respiratory chain.
In addition, this particular hippocampal subfield also showed a decreased CS—normalized activity of aconitase—a mitochondrial enzyme containing an oxygen radical-sensitive FeS cluster. This specific decrease in activity can be interpreted as an indication of an increased oxidative stress associated with complex I inhibition-related mitochondrial dysfunction (Gardner and Fridovich, 1992; Kunz, 2002; Kann et al., 2005). In this context it is noteworthy to mention that aconitase has also been found severely deficient in human tissues of varying origins deriving from patients with Friedreich's ataxia (Rötig et al., 1997)—a neurodegenerative disorder characterized by degeneration of the Purkinje neurons of the cerebellum. In this disease the deficiency of frataxin leads to an impairment of intracellular iron metabolism causing enhanced oxidative stress. Similar to this disorder and to our observations in human AHS, oxidative stress related to mitochondrial dysfunction has been also observed in experimental animal models of epilepsy (Liang and al, 2000; Patel and Li, 2003).
The critical role of the CA3 region in the epileptic circuit of the AHS hippocampus is corroborated by experimental studies in which induced rhythmic synchronous activity in hippocampal slices increased free radical production particularly in the CA3 region where neurons subsequently degenerated (Frantseva et al., 2000). Both free radical production and cell death correlated with a persistent and progressive increase in intracellular calcium, beyond the duration of the activity. In a further step it has been shown for CA3 pyramidal cells that epileptic activity results in Ca2+-dependent changes in mitochondrial function that might contribute to the neuronal injury during epilepsy (Kovacs et al., 2005). Thus, the complex relationship between neuronal activity, intramitochondrial calcium handling, metabolic rate, and mitochondrial depolarization might enhance seizure propensity in the AHS epileptic limbic system (Kann et al., 2005).
In conclusion, our data indicate a tight correlation between the subfield-specific alterations of metabolite levels detected by MRS and a regional impairment of mitochondrial oxidative phosphorylation in the AHS hippocampus. In particular, our results strongly suggest that the underlying mechanism causing the decline of hippocampal NAA concentrations in patients with MTLE and AHS is directly related to a subfield-specific impairment of energy metabolism but not to the neuronal cell loss.
The authors acknowledge the generous support of the Center of Advanced Imaging Magdeburg (CAI, BMBF-grants 01GO0202 & 01GO0504) and the German Research Council (SFB/TR3 TP A7) given to WSK, SV, and HJH.