Decreased hippocampal volume on MRI is associated with increased extracellular glutamate in epilepsy patients

Authors


  • Walid Abi-Saab is currently in Abbott Laboratories.

Address correspondence to Idil Cavus, M.D., Ph.D., Yale University, Department of Psychiatry, 300 George St., 9th Floor, Room 27, New Haven, CT 06511, U.S.A. E-mail: idil.cavus@yale.edu

Summary

Purpose: Temporal lobe epilepsy (TLE) is associated with smaller hippocampal volume and with elevated extracellular (EC) glutamate levels. We investigated the relationship between the hippocampal volume and glutamate in refractory TLE patients.

Methods: We used quantitative MRI volumetrics to measure the hippocampal volume and zero-flow microdialysis to measure the interictal glutamate, glutamine, and GABA levels in the epileptogenic hippocampus of 17 patients with medication-resistant epilepsy undergoing intracranial EEG evaluation. The relationships between hippocampal volume, neurochemical levels, and relevant clinical factors were examined.

Results: Increased EC glutamate in the epileptogenic hippocampus was significantly related to smaller ipsilateral (R2= 0.75, p < 0.0001), but not contralateral hippocampal volume when controlled for glutamine and GABA levels, and for clinical factors known to influence hippocampal volume. Glutamate in the atrophic hippocampus was significantly higher (p = 0.008, n = 9), with the threshold for hippocampal atrophy estimated as 5 μM. GABA and glutamine levels in the atrophic and nonatrophic hippocampus were comparable. Decreased hippocampal volume was related to higher seizure frequency (p = 0.008), but not to disease duration or febrile seizure history. None of these clinical factors were related to the neurochemical levels.

Conclusions: We provide evidence for a significant association between increased EC glutamate and decreased ipsilateral epileptogenic hippocampal volume in TLE. Future work will be needed to determine whether the increase in glutamate has a causal relationship with hippocampal atrophy, or whether another, yet unknown factor results in both. This work has implications for the understanding and treatment of epilepsy as well as other neurodegenerative disorders associated with hippocampal atrophy.

Hippocampal atrophy on MRI is seen in 60%–75% of patients with medication-resistant temporal lobe epilepsy (TLE), as well as in other neurodegenerative conditions including Alzheimer's, Huntington's and Parkinson's disorders and posttraumatic stress disorder (Spencer et al., 1993; Geuze et al., 2005). In epilepsy, hippocampal atrophy is tightly related to the disorder, as resection of the epileptogenic atrophic hippocampus is one of the few factors predicting better clinical outcome (Spencer et al., 2005). Although the mechanism underlying this atrophy remains unknown, several associated clinical factors have been identified. These include genetic background, trauma, infections, ischemia, elevated glucocorticoids, and prolonged childhood febrile seizures (Mathern et al., 2002; Baulac et al., 2004; Geuze et al., 2005; Lewis, 2005). In addition, glutamatergic excitotoxicity associated with seizures has long been proposed as a potential mechanism for the neuronal loss (Olney et al., 1986) underlying hippocampal atrophy (Luby et al., 1995; Briellmann et al., 2002). Studies in animal models have indicated that the initial status epilepticus alone (Gorter et al., 2003) or recurrent seizures can result in significant hippocampal damage (Cavazos & Sutula, 1990; Sutula et al., 2003), which is reduced by N-methyl-d-aspartate (NMDA) receptor antagonists (Brandt et al., 2003).

Using in vivo microdialysis in patients with medication-resistant epilepsy, we have previously reported that the extracellular (EC) glutamate levels in the epileptogenic hippocampus can increase dramatically during spontaneous seizures (During & Spencer, 1993). Furthermore, our subsequent work demonstrates that in the epileptogenic hippocampus of these patients, the basal glutamate levels are elevated chronically during the interictal period (Cavus et al., 2005). Since chronic exposure to high glutamate has been related to neurotoxicity and cell loss (Olney et al., 1986; Tanaka et al., 1997; Cid et al., 2003), we investigated the relationship between basal EC glutamate levels and the hippocampal volume in patients with TLE. We used quantitative MRI to measure the hippocampal volume and zero-flow microdialysis to measure the basal EC levels of glutamate and its metabolites glutamine and gamma-aminobutyric acid (GABA) in the epileptogenic hippocampus of patients with medication-resistant TLE. The effects of clinical factors known to affect hippocampal volume, such as seizure frequency, duration of epilepsy, and history of febrile seizures (Spencer et al., 1993; Theodore et al., 1999; Kalviainen & Salmenpera, 2002; Bernasconi et al., 2005) were also examined.

Methods

Subjects

Patients with medication-resistant complex partial seizures (CPS) evaluated at Yale University Epilepsy Surgery Program were invited to participate in the brain MRI and microdialysis research protocols approved by the Yale University School of Medicine Human Investigation Committee and in compliance with the Declaration of Helsinki. The in vivo microdialysis procedure in epilepsy patients was first developed in our Center (During, 1991), and has proven to be a safe and reliable method for measurement of the EC neurochemicals in the human brain. None of the patients studied over several years have developed adverse effects attributable to microdialysis. We studied 26 consecutive patients using quantitative hippocampal volumetric MRI, followed by the hippocampal microdialysis study within 1 month during the intracranial EEG monitoring phase. Both studies were performed interictally, when patients were on their antiepileptic drugs (AEDs). Clinical data was obtained from patients, their families and hospital records during the phased evaluation of the patient. Seizure frequency was calculated as average total seizures (complex partial and generalized clonic–tonic seizures) per month over the 12 months prior to the intracranial study, as reported by the patient and their family during structured interviews every 3–4 months. Duration of illness was calculated as years from the age of onset of the first habitual seizure.

Procedures

Microdialysis probes incorporated into depth electrodes (Spencer probe, Ad-Tech Instrument Co., Racine, WI, U.S.A.) were implanted using a stereotactic neuronavigation device (BrainLAB, Westchester, IL, U.S.A.) in the anterior region of the suspected epileptic hippocampus (Fig. 1) as well as in other relevant brain regions. In few cases, additional depth electrodes were placed in the posterior ipsilateral hippocampus or in the contralateral hippocampus. Two types of microdialysis probes were used. The Yale probe (diameter 0.3 mm, membrane 10 mm with 5 kDa cutoff) is attached to a flexible depth electrode (Ad-Tech Instrument Co.) and has been described before (During, 1991; During & Spencer, 1993; Cavus et al., 2005). The second, more recent design is a custom-modified CMA/20 probe (0.67 mm diameter, 70 mm length, 10 mm 20 kDa cutoff membrane, CMA, North Chelmsford, MA, U.S.A.) 0.67 mm diameter, 70 mm length, and 10 mm 20 kDa cutoff membrane, which is inserted into a depth electrode with perforations between contacts 1 and 2 to allow for fluid exchange. All probes are sterilized by gamma radiation. Following surgery, 3D coregistered CT and MRI was used to verify the location of the probes (Fig. 1) and after 1–2 days of recovery, the AEDs were tapered in order to allow spontaneous seizures. Seizure onset site was interpreted by Yale epileptologists experienced in intracranial electrophysiology (see Table 1). The intracranial EEG data obtained from the depth electrode contacts that flank the microdialysis probe was used to determine if the probe was within the epileptogenic (site of habitual seizure origin) or nonepileptogenic hippocampus. Only data obtained from the epileptogenic hippocampus were included in the analysis.

Figure 1.


Magnetic resonance image of a patient's brain implanted with a modified Spencer depth probe in the right hippocampus. The microdialysis membrane (10 mm long), which is not visible on MRI, lies between depth electrode contact 1 and 2 in the anterior hippocampus. The image is taken 1 day after implantation and shows a coronal section of the brain. Contacts 7 and 8 are outside of the brain and appear curved, as the depth electrode is flexible.

Table 1.  Subject characteristics
PtAgeSexIllness duration (yrs)Seizure/monthFebrile seizuresClinical MRIProbe siteSeizure focusHipp pathologyAEDsa
  1. aAEDs during the microdialysis study.

  2. b No surgery due to functional hippocampus.

  3. Pt, patient; B, bilateral; L, left; R, right; Hipp, hippocampus; EC, entorhinal cortex; TC, temporal cortex; NA, not applicable; AED, antiepileptic drug; CBZ, carbamazepine; CLZ, clonazepam; DPH, phenytoin; GBP, gabapentin; LTG, lamotrigine; LEV, levetiracetam; OXC, oxcarbazepine; VA, valproic acid; ZNS, zonisamide.

 125F15 5NoNormalL hippL hippNAbCBZ, ZNS
 220F 811NoNormalL hippL hipp & TCNAbCBZ, ZNS
 346F 9 4NoNormalL hippL hippMinimal cell lossLEV, VA
 424M23  2.3YesNormalR hippR hippMinimal cell lossDPH, GBP
 538F30  4.5NoNormalR hippR hippMinimal cell lossTPM
 637F 8 8NoNormalR hippR hipp & TCMinimal cell lossDPH, ZNS
 739M13 4NoL hipp atrophyL hippL hipp & TCNAbLEV, LTG
 854M3518YesL hipp atrophyL hippL hipp & R TCNAbCBZ
 941M40 2YesR hipp atrophyL hippL hipp & R TCNAbCBZ, VA
1052F 8 6NoB hipp atrophyL hippL hippNAbLEV, VA
1130F23 5NoL hipp atrophyL hippL hipp & TCSclerosisLTG, OXC, ZNS
1226M18 8NoL hipp atrophyL hippL hippSclerosisCBZ, DPH
1340M2312YesL hipp atrophyL hippL hippSclerosisCLZ, LTG
1417M1213YesR hipp atrophyR hippR hippSclerosisLTG
1537M 816NoB hipp atrophyL hippL hipp & TCSclerosisLTG, OXC
1652F17 6NoB hipp atrophyR hippR hipp & TCSclerosisDPH
1732F28 4YesB hipp atrophyR hippR hippSclerosisCBZ

Zero-flow microdialysis study

The zero-flow quantitative microdialysis method allows for estimation of the absolute basal concentration of neurochemicals in the EC fluid under zero-flow steady-state conditions (Hutchinson et al., 2002). This method yields results comparable to those obtained using no-flow microdialysis, the accepted gold standard method for quantitative measurement of EC substrates. However the no-flow method is not suitable for human research, since it requires perfusion of the measured neurochemicals. The method used here was described before (Cavus et al., 2005). Briefly, to avoid any possible confounds related to disturbed blood–brain barrier (Benveniste et al., 1987), circadian rhythm or stress, the study was conducted 2–5 days following probe implantation, in the afternoon, when patients were quietly resting, at least 6 h from any seizure activity. For clinical care reasons, most patients were on some combination of AEDs (Table 1). Sterile artificial cerebrospinal fluid (135 mM NaCl, 3 mM KCl, 1 mM MgSO4× 7H2O, 1.2 mM CaCl2× 2H2O in 1 mM sodium phosphate buffer, pH 7.4) was infused using portable CMA107 syringe pumps (CMA) for 1–2 h at 2.0 μl/min flow rate for equilibration, two 20 μl dialysate samples were collected and then the flow rate was decreased to 0.5 and 0.2 μl/min, allowing for 60–90 min equilibration and additional sample collection at each step. The experiment was completed in ∼6 h, samples were stored over dry ice, and then in –80°C freezer for HPLC analysis. Basal levels were determined using regression analysis with fit to 2nd polynomial order to a flow of zero, i.e., steady state.

HPLC analysis

Glutamate and glutamine were analyzed as described before (Cavus et al., 2005) using O-phthaldialdehyde (OPA) derivatization and fluorescence detector (Shimadzu Scientific Instruments, Inc., Columbia, MD, U.S.A.). Sensitivity limits were 0.1 μM for glutamate and 10 μM for glutamine, with signal to noise ratio of 10:1. GABA was analyzed on a dedicated system (Bioanalytical Systems, Inc., Indianapolis, IN, U.S.A.). Briefly, 2 μl dialysate was added to 4 μl internal standard, derivatized with 13 μl OPA, and injected onto column (BAS, ODS column, Phase II 100 × 3.2 mm 3μm). Mobile phase is 0.1 M acetic acid (pH 4.85) in 37% acetonitrile solution, 0.8 ml/min flow rate. Electrochemical detection (ESA, Chelmsford, MA, U.S.A.) is achieved at 430 mV and chromatograms are complete in 22 min. Sensitivity limit was 8 nM. Peak areas of amino acids are compared to external standards to determine concentration in the samples using EZCHROME elite software (Scientific Software, Inc., Pleasanton, OH, U.S.A.).

Quantitative hippocampal MRI

Quantitative MR imaging data were acquired using a 4T Varian Inova whole body MR system with cavity resonator head coil. T1 gradient echo-weighted imaging using an inversion recovery preparation was used to acquire high-contrast images at 1.5 mm isotropic resolution. Off-midline sagital imaging was used to acquire triply oblique images through the bilateral hippocampi to visualize the planum temporale. Manual volumetry using coronal reformatted images perpendicular through planum temporale was performed according to the published methods (Jack et al., 1992; Watson et al., 1992) so as to include the tail, body and pes hippocampus, subiculum, and dentate. The volumetry was performed such that those pixels defined as being within the tracing are included in the final measurements. Whole brain volumetry was performed using 3 mm thick slice whole brain quantitative T1 imaging (Hetherington et al., 2001). These images were processed with manual tracing of the cerebellum, which was not included in the determination of whole brain volume. Using these methods, our healthy control hippocampal measurements were 2.69 ± 0.32 (left hippocampus) and 2.84 ± 0.33 (right hippocampus), which supports the generally observed view that the right is larger than the left hippocampus and is in relatively good agreement with published literature (Jack et al., 1989).

Intracranial spike analysis

Intracranial EEG spikes were counted with a commercially available program (Reveal, Persyst Development Corp., Prescott, AZ, U.S.A.). We have previously evaluated this program against three epileptologists experienced in intracranial EEG. In the evaluation, the algorithm's performance agreed with the consensus of experts with an acceptable sensitivity (0.714) and specificity (0.819).

Statistical analysis

Data was analyzed using JMP statistical software (version 5.0.1.2, SAS Institute Inc, Cary, NC, U.S.A.). Since raw data was not normally distributed, all data was log transformed to achieve normalization. We used linear regression and multiple regression models to examine the relationships between hippocampal volume, glutamate, glutamine, GABA, and clinical factors (disease duration, history of febrile seizures, and frequency of seizures). Group differences were examined using t-tests. Significance was set as p < 0.05.

Results

Out of 26 consecutive patients studied with hippocampal MRI volumetry and microdialysis, 17 (eight female, age 37.0 ± 11.5 and nine male, age 34.7 ± 11.7, mean ± SD; two left-handed) had habitual seizures originating from the sampled hippocampus (epileptogenic hippocampus) and in some cases from both the hippocampus and the temporal cortex (Table 1). Data from patients where hippocampus was not (n = 2) or secondarily involved (n = 5), or where seizures were not localized (n = 2), were excluded from analysis. Only data from anterior hippocampus were used, since few patients had additional probes in the posterior hippocampus. Hippocampal volumetry was measured at least 48 h following any seizures. Microdialysis studies were 79 ± 28 h (mean ± SD) from any clinical or subclinical seizure activity. The relative in vivo recovery rates for the CMA custom-designed microdialysis probes at 0.2 μl/min flow rate were: glutamate 82 ± 5% (mean ± SD, n = 10), glutamine 83 ± 6% and GABA 83 ± 5%. Since these rates were comparable to the recovery rates of the Yale probe at the same flow rate (80 ± 13% glutamate, 81 ± 6% glutamine and 82 ± 6% GABA, n = 10), results obtained from the custom CMA (n = 7) and Yale probes (n = 12) were combined.

Glutamate, glutamine, and GABA levels in the epileptogenic hippocampus

Zero-flow glutamate levels ranged from 1.0 to 25.4 μM (7.1 ± 1.6 μM, mean ± SE, n = 17) and glutamine levels were 112–2543 μM (688 ± 153 μM, n = 16), consistent with values reported earlier (Cavus et al., 2005). GABA levels were 0.03–2.8 μM (0.7 ± 0.2 μM, n = 16).

Hippocampal volume versus EC glutamate

We have hypothesized that since exposure to chronically elevated glutamate can be excitotoxic (Olney et al., 1986; Tanaka et al., 1997; Cid et al., 2003), higher basal glutamate levels are related to increased neuronal cell loss and smaller hippocampal volume. Indeed, increased glutamate was significantly related to decreased absolute hippocampal volume, with 75% of the variability in the volume being attributable to the glutamate levels (F1,15= 45.8, p < 0.0001, log-transformed data, Fig. 2). Similar results were obtained using a nonparametric test (r =−0.90, p < 0.0001, Spearman's rank correlation test, raw data) and normalized hippocampal volume (R2= 0.76, F1,11= 34.5, p < 0.0001, total cerebral volume for four subjects was not available). Therefore our data indicate that there is a highly significant negative relationship between the volume and the EC glutamate levels of the epileptogenic hippocampus. In addition, whether the glutamate metabolites glutamine and GABA have effect on the hippocampal volume was examined in a multiple regression model. Glutamate (F1,11= 26.2, p = 0.0003), but not glutamine (F1,11= 0.5, p = 0.48) or GABA (F1,11= 0.75, p = 0.40) had significant effect, indicating that when controlled for the levels of glutamine and GABA, only glutamate is significantly related to the hippocampal volume. There were no significant relationships between any of the measured neurochemicals and the volume of the contralateral hippocampus (all tests p > 0.05, regression analysis).

Figure 2.


In patients with medication-resistant TLE, the volume of the epileptogenic hippocampus measured by quantitative MRI is significantly related to the basal EC glutamate measured by zero-flow microdialysis method during the interictal period (R2= 0.75, n = 17, p < 0.0001, logarithmic fit on raw data). The threshold for hippocampal atrophy (dashed line) corresponds to 5 μM glutamate.

Epilepsy duration, seizure frequency, and history of febrile seizures

Longer disease duration, higher seizure frequency, and history of febrile seizures have been associated with smaller hippocampus in patients with epilepsy (Spencer et al., 1993; Theodore et al., 1999; Kalviainen & Salmenpera, 2002; Bernasconi et al., 2005). In our sample, smaller hippocampal volume was related to higher seizure frequency (R2= 0.39, n = 17, p < 0.008 for absolute hippocampal volume and R2= 0.35, n = 13, p < 0.03 for normalized hippocampal volume; Fig. 3), but not to disease duration or history of febrile seizures (p > 0.05 for all tests). The relative contribution of these clinical factors and the hippocampal volume on glutamate levels was examined in a multiple regression model. Only hippocampal volume (F1,12= 34.4, p < 0.0001) and none of the examined clinical factors was significantly related to glutamate (p > 0.05 for all other effects). Thus, smaller hippocampal volume is associated with higher EC glutamate even when controlled for disease duration, seizure frequency, and history of febrile seizures, and none of these clinical factors have significant effect on interictal glutamate. Analysis using normalized hippocampal volume yielded similar results (F1,8= 23.6, p = 0.001 for the effect of glutamate, p > 0.05 for all other effects). Interestingly, subjects with history of febrile seizures (n = 6) had higher glutamate (9.6 ± 2.7 vs. 5.8 ± 1.9 μM) and lower GABA levels (0.4 ± 0.3 vs. 0.8 ± 0.3 μM), but these differences were not statistically significant (p > 0.05, t-test).

Figure 3.


Decreased volume of the epileptogenic hippocampus in patients with medication-resistant TLE is related to increased seizure frequency (average complex partial and tonic–clonic seizures per month over the last 12 months; R2= 0.39, n = 17, p = 0.008, logarithmic fit on raw data).

Atrophic versus nonatrophic epileptogenic hippocampus

Nine out of 17 subjects had atrophic hippocampus, which was defined as being 2 SDs smaller then the average hippocampal volume in control subjects. The atrophic hippocampus had significantly higher glutamate (10.9 ± 2.4 μM vs. 3.2 ± 0.7 μM, p = 0.008, Fig. 4A), while glutamine (685 ± 160 μM vs. 752 ± 309 μM, p = 0.8, Fig. 4B) and GABA levels (0.9 ± 0.3 μM vs. 0.4 ± 0.2 μM, p = 0.4, Fig. 4C) were comparable. Using linear regression model formula [Log (glutamate μM) = 23.15 − 2.8 × Log (hippocampal volume cc)], the glutamate level corresponding to the threshold for hippocampal atrophy (2.18 cc in this sample) was calculated as 5.1 μM (Fig. 2).

Figure 4.


Zero-flow microdialysis levels of interictal EC glutamate (A), glutamine (B), and GABA (C) in the atrophic (n = 9) and nonatrophic (n = 8) epileptogenic hippocampus of patients with medication-resistant TLE. Basal glutamate levels are significantly elevated in the atrophic hippocampus (p = 0.008, t-test on log-transformed data). The glutamine and GABA levels are comparable. Data are plotted as mean ± SEM.

Relationship between hippocampal spikes and EC neurochemical levels

We have previously reported that basal glutamate levels in the human epileptogenic hippocampus are not related to the interictal spiking rate (Cavus et al., 2005). In this sample, in addition to glutamate, we explored if interictal spiking is related to glutamine and GABA. Spikes recorded at depth electrode contacts flanking the dialysate membrane were counted for 10 probe sites during the sample collection at each equilibrated flow rate and expressed as total spike number (total time 170 min). None of the measured substrates were related to the intracranial spiking (p > 0.05 for all tests).

Discussion

We report that in patients with medication-resistant TLE, smaller volume of the epileptogenic hippocampus on MRI is related to higher basal glutamate measured with zero-flow microdialysis. This relationship was significant even when controlled for levels of glutamate metabolites, glutamine and GABA, and for clinical factors reported to influence the hippocampal volume, such as seizure frequency, duration of epilepsy, and history of febrile seizures (Spencer et al., 1993; Theodore et al., 1999; Kalviainen & Salmenpera, 2002; Bernasconi et al., 2005). While none of these clinical factors had significant effect on glutamate levels, higher seizure frequency was associated with smaller epileptogenic hippocampus. In addition, glutamate levels were significantly higher in the atrophic versus nonatrophic epileptogenic hippocampus.

Basal EC glutamate in the epileptogenic human hippocampus ranged from 1 to 25 μM, with 75% of this variability being attributable to the degree of hippocampal volume loss. Significant hippocampal atrophy on MRI was observed at concentrations above 5 μM. In contrast, in our previous study, glutamate levels in the nonepileptogenic human hippocampus were found to be much lower, estimated as 0.5–4 μM (Cavus et al., 2005) and similar to the ones in control rats (Lerma et al., 1986). Although the results of the present study indicate that there is a significant association between higher basal glutamate levels and decreased volume in the epileptogenic hippocampus, they do not establish a causal link. Several possible mechanisms may account for this association. One interpretation is that the chronic glutamate elevation in the epileptogenic hippocampus can result in neurotoxic injury, cell loss (Olney et al., 1986; Tanaka et al., 1997; Cid et al., 2003), and hippocampal atrophy. Indeed, in animal models, progressive epileptogenesis has been associated with increasing basal glutamate levels (Lothman et al., 1987), with glutamate levels above 5 μM being neurotoxic in cell culture (Cid et al., 2003). Accordingly, in our patient population, the higher EC glutamate levels in the epileptogenic hippocampus were also related to lower neuronal count (Cavus et al., 2006). Alternatively, the initial sclerotic insult could result in both smaller hippocampus and in increased glutamate through an unidentified mechanism. Early seizures, trauma, stroke, or infections are all associated with hippocampal atrophy (Geuze et al., 2005) as well as with glutamate elevation (Doble, 1999). A third possibility is that the EC glutamate could accumulate in the already atrophied hippocampus through multiple mechanisms. These include impaired uptake of the glutamate released with seizures (Ueda et al., 2001), enhanced tonic or stimulated glial release (Del Arco et al., 2003; Cavelier et al., 2005), or enhanced neuronal release. Whether glia or neurons, or both are the source of the higher glutamate in human epileptogenic hippocampus is currently unknown. There is some evidence that glial glutamate reuptake in TLE patients is impaired, since despite adequate expression of the transporters (Bjornsen et al., 2007), there is a decrease in the transporter protein (Proper et al., 2002; Bjornsen et al., 2007), in synaptosomal reuptake (Hoogland et al., 2004) and in the clearance of glutamate release with seizures (During & Spencer, 1993). Impaired glial glutamate—glutamine metabolism is another candidate mechanism for ongoing glial glutamate release (Eid et al., 2004). We did not find any relationship between the glutamate levels in the epileptogenic hippocampus and the contralateral nonepileptogenic hippocampal volume, suggesting that the related pathogenic process is local to the affected site. Interictal neuronal activity most likely does not contribute significantly to the elevated glutamate in the atrophic hippocampus, as there are relatively fewer neurons, and the contribution of the action potential—dependent release to microdialysate glutamate and GABA is known to be negligible (Del Arco et al., 2003). This is also supported by our present and previous data (Cavus et al., 2005) showing no relation between intracranial EEG spiking and basal glutamate and GABA levels.

Regardless of the source of glutamate, its elevation alone in the intact brain is not sufficient to cause excitotoxic injury or seizures (Obrenovitch et al., 2000), due to several protective mechanisms such as avid glial glutamate uptake (Danbolt, 2001) and rapid receptor desensitization (Zorumski et al., 1996). However, in the epileptic hippocampus, failure in these mechanisms and other pathologic adaptations, such as altered NMDA receptor channel dynamics (Isokawa et al., 1997; Lieberman & Mody, 1999), may facilitate the glutamatergic injury. Impaired cellular energetics in epilepsy and in other neurodegenerative disorders (Doble, 1999; Kunz, 2002) can also enhance excitotoxicity (Jabaudon et al., 2000). In animal models, exposure to chronically elevated EC glutamate is reported to promote epileptogenicity and excitotoxicity (Tanaka et al., 1997; Sierra-Paredes et al., 2001; Vazquez-Lopez et al., 2005), possibly through cellular reorganization and increased expression of the extrasynaptic NMDA receptors (Vazquez-Lopez et al., 2005). The extrasynapic NMDA receptors are exposed to the EC milieu sampled by microdialysis (Vizi & Mike, 2006), and their over activation can lead to enhanced tonic excitability (Dalby & Mody, 2003; Le Meur et al., 2007) and neuronal death (Hardingham & Bading, 2003). Higher extrasynaptic glutamate can also activate the presynaptic metabotropic glutamate receptors and suppress further glutamate release, controlling excitability. However, in the sclerotic epileptogenic human hippocampus, some aspects of this critical regulation seem to be impaired (Dietrich et al., 1999, 2002).

EC glutamate was significantly related only to the ipsilateral hippocampal volume, but not to disease duration, frequency of seizures, or history of febrile seizures, suggesting that the relationship between glutamate and hippocampal volume is independent from these clinical factors. However, interpretation of these results is probably limited by a relatively small sample size. Such limitation could also explain our negative findings for any significant associations between hippocampal volume, disease duration, and history of febrile seizures, which have been reported in larger samples (Spencer et al., 1993; Theodore et al., 1999; Kalviainen & Salmenpera, 2002; Fuerst et al., 2003; Bernasconi et al., 2005), although not by all (Spanaki et al., 2000; Liu et al., 2005). On the other hand, we also found that smaller hippocampal volume was related to higher seizure frequency suggesting that the effect of seizure frequency may be more robust, despite concerns about the accuracy of self-reports. Whether the atrophic hippocampus is a cause (Spencer et al., 2005) or a result of recurrent seizures (Sutula et al., 2003), or both (Mathern et al., 2002), has remained a disputed issue. However, that smaller hippocampal volume was related independently to higher glutamate levels and to higher seizure frequency may be suggestive of independent mechanisms mediating the two conditions, where glutamate levels are not predictors or consequence of recurrent seizures.

Neither GABA nor glutamine levels were related to the hippocampal volume. GABA levels in the epileptogenic hippocampus were 0.03–2.8 μM, and although they were higher in the atrophic hippocampus, the difference was not significant. For comparison, EC GABA in the nonepileptic rat hippocampus is 0.8 ± 0.15 μM (Lerma et al., 1986). The wide range in GABA levels in our patients may reflect various factors, including the effects of AEDs, the degree of GABAergic cell preservation and GABA clearance (During et al., 1995; Patrylo et al., 2001), and the menstrual cycle phase in women (Epperson et al., 2002). Nevertheless, higher GABA levels do not seem to protect against glutamate-induced neurotoxicity (Yeh et al., 2005), as they also did not have a significant effect on the hippocampal volume in our study. Surprisingly, glutamine levels in the atrophic hippocampus were not decreased, despite the reported suppression of glutamine synthase (Eid et al., 2004), the glial enzyme converting glutamate to glutamine. In fact, glutamine levels were correlated with increased gliosis in the excised hippocampus from patients with TLE (Cavus et al., 2006), suggesting that the equivalent glutamine levels in the atrophic and nonatrophic hippocampus may result from a combination of suppressed glutamine synthase and increased gliosis.

In summary, we report that higher EC glutamate in the epileptogenic hippocampus of patients with TLE is related to lower hippocampal volume. Future work will be needed to elucidate any causal mechanisms of this association. In addition, if the glutamate levels in poorly controlled patients increase as the hippocampal volume progressively decreases (Fuerst et al., 2003) is also unknown. Other less invasive methods, such as magnetic resonance spectroscopy (MRS) may be more suitable for repeated measures of brain glutamate (Pan et al., 2006), although the relationship between tissue and EC glutamate will need to be elucidated. If the elevation in EC glutamate indeed has a role in the pathophysiology of epilepsy, then treatments that suppress EC glutamate or inhibit the extrasynaptic NMDA receptors (Chazot, 2004) would be of great interest. In animal models, such drugs are reported to be neuroprotective even when administered long after epileptic activity (Brandt et al., 2003; Ayala & Tapia, 2005). Our results may also be relevant for the understanding and treatment of other neurodegenerative and psychiatric conditions associated with hippocampal atrophy and glutamate dysregulation, such as Alzheimer's disease, Parkinson's and Huntington's disorders, amyotrophic lateral sclerosis, posttraumatic stress disorder, and depression (Doble, 1999; Geuze et al., 2005).

Acknowledgments

This work was supported by NIH-National Institute of Neurological Disorders and Stroke (P01NS39092, D.D.S), NIH-National Institute of Drug Abuse (BIRWCH 1K12DA14038-01, I. C), NIH-National Institute of Biomedical Imaging and Bioengineering (000473, H.P.H. and J.W.P., and EB-005438 for H.P.H.) and was approved by the IRBs at Yale University School of Medicine and Albert Einstein College of Medicine. We thank David Ocame, Sarah Forselius, Michael Cassaday, and Willard Kasoff, M.D., for technical assistance with microdialysis and HPLC, and Ralitza Gueorgieva, Ph.D., for advice on statistical methods. We also thank Anne Williamson, Ph. D., for review and suggestions on the manuscript.

Conflict of interest: We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines. J. H. Krystal provided paid scientific consultation to the following companies during the prior 3 years: Janssen, Takeda Industries, Lilly, Merz, GlaxoSmithKline, Bristol-Myers Squibb, Organon, Shire, Sumitomo, Cypress Pharmeuticals. W.Abi-Saab is currently employed by Abbott Pharmaceuticals. None of the other authors have any conflicts of interest.

Ancillary