5-HT_1A Receptors Are Reduced in Temporal Lobe Epilepsy After Partial-Volume Correction

Subjects

We studied 22 patients (18 men; mean age ± SD, 37 ± 11 y) referred to the Clinical Epilepsy Section, National Institute of Neurologic Disorders and Stroke (NINDS), National Institutes of Health (NIH), for evaluation of medically refractory TLE (5). Ictal video-electroencephalography (EEG) confirmed each patient’s epileptogenic zone localization. Five had subdural recording after PET for language mapping. Epilepsy onset age was 15 ± 10 y (range, 1–34 y); disease duration was 23 ± 13 y (range, 5–46 y). Patients with structural lesions (except MTS), progressive neurologic disorders, or taking any medications aside from antiepileptic drugs (AEDs), were excluded. Eleven were included in our earlier study (5). None experienced seizures for at least 2 d, or a secondarily generalized tonic clonic seizure for at least 1 mo, before PET studies. Depressive symptoms were assessed with the Beck Depression Inventory (BDI) (17). The study was approved by the NINDS Investigational Review Board and Radiation Safety Committee. We studied 10 healthy volunteers (7 men; mean age ± SD, 35 ± 9 y), screened with general physical examination and routine laboratory tests.

MR Procedure

All subjects underwent MRI using a 1.5-T Horizon (GE Healthcare). In patients, T1-weighted spoiled gradient images (repetition time [TR], 27 ms; minimum echo time [TE]; flip angle, 20°) were acquired in coronal orientation. In healthy subjects, 3-dimensional (3D) IRprepped (variable TR; minimum TE; and inversion time, 300 ms) were acquired in the transverse plane. Coronal and transverse MR images had the same spatial resolution (voxel size, 0.94 × 0.94 × 1.5 mm; 256 × 256 × 124 slices). Transverse slices were resliced in coronal planes for analyses. Standard coronal T2 and FLAIR (Fluid Attenuated Inversion Recovery) images were obtained. All MR images were read by experienced neuroradiologists who were unaware of PET results.

Radiochemistry

¹⁸F-FCWAY was synthesized as described (3); specific activity at injection was >18.5 TBq/mmol (>500 Ci/mmol).

PET Procedure

We used an Advance Tomograph (GE Healthcare; full width at half maximum [FWHM], 6–7 mm), scanning 35 simultaneous slices with 4.25-mm slice separation. During scanning, in a quiet, dim room with eyes closed and ears unoccluded, a thermoplastic facemask held the subject’s head in place (5). Continuous EEG monitoring was performed to ensure interictal studies. A transmission scan was performed for attenuation correction. Subsequently, a 333 ± 74 MBq (9 ± 2 mCi) ¹⁸F-FCWAY bolus was injected over 60 s and dynamic frames (1–5 min) were acquired in 3D mode for 120 min.

¹⁸F-FCWAY Input Function

About 30 radial arterial samples were taken to quantify ¹⁸F-FCWAY concentration and selected samples used to measure the ¹⁸F-fluorocyclohexanecarboxylic acid metabolite (¹⁸F-FC) (18). Unbound ¹⁸F-FCWAY fraction plasma protein was measured with ultracentrifugation (5).

Data Analysis

For each subject, ¹⁸F-FCWAY radioactivity frames were registered to MR images using a mutual information cost function (16). Tissue activity frames were corrected for brain acid metabolite ¹⁸F-FC uptake, vascular radioactivity, and ¹⁸F-fluoride metabolite spillover (18). Pixel data were fitted to a 2-tissue compartment model with 3 parameters using a metabolite-corrected input function to provide parametric images of V (4). A BP measurement was derived using cerebellum as a region devoid of 5-HT_1A receptors (19) and calculated as: Eq. 1 where V_ROI and V_CEREB are distribution volumes in the target region of interest (ROI) and cerebellum, respectively, and f₁ is the ¹⁸F-FCWAY plasma free fraction. Asymmetry indices (AIs) were computed as 200·(I –C)/(I + C), where I and C represent PET values in regions ipsilateral and contralateral to the seizure focus, respectively. In control subjects, the right side was arbitrarily called ipsilateral.

Measurement of CMRglu

Patients underwent ¹⁸F-FDG PET for CMRglu measurement within 2 d of ¹⁸F-FCWAY scans, using a noninvasive approach (5). Generated CMRglu parametric images were registered to MR images. Dosimetric restrictions precluded ¹⁸F-FDG scans in control subjects.

PVC Algorithm

The algorithm (20) provides voxel-by-voxel binary masks of gray matter, white matter, and cerebrospinal fluid (CSF). Gray matter pixels are corrected for spill-out of gray matter activity as well as for spill-in of activity from white matter. CSF activity is set to zero. Brain white matter activity is assumed to be uniform (11). Visual inspection revealed poor segmentation in basal ganglia and thalamus, where a large fraction of gray pixels were misidentified as white matter, due probably to close white-gray intermixing (16). Since 5-HT_1A receptor concentration in striatum and thalamus is negligible, and a 10% error in the estimation of white matter value would introduce only an ∼1% error in gray matter PVC data (11,12,16), potential bias of ¹⁸F-FCWAY binding in cortical gray regions is expected to be negligible. Subsequently, the binary masks are smoothed to the PET resolution by convolving the binary mask with a 6-mm FWHM 3D gaussian kernel (based on the reconstructed image resolution); these masks are named as s_GM, s_WM, and s_CSF. The corrected activity gray matter values can be calculated as follows: Eq. 2 where C_PVC represents the corrected activity value in a gray matter pixel after PVC, C_ORIG is the original uncorrected pixel value, C_WM is the estimated white matter activity (16), and s_GM and s_WM are the pixel values from the smoothed masks for gray and white matter, respectively.

ROI Analysis

3D hippocampal ROIs were manually drawn on MR volumes using an anatomic atlas (21). Large regions ensured adequate gray matter sampling and increased signal-to-noise ratio (16). To minimize the risk of potential spillover of skull ¹⁸F-fluoride activity, cerebellar ROIs were initially drawn along the outer cortical edge and then uniformly shrunk by about 50%. Due to finer sampling of MR than PET data, ROIs were placed on 6 or 7 consecutive slices for amygdala, every second or third slice for hippocampus, fusiform, and parahippocampus gyrus, and every third or fourth slice for remaining cortical regions and cerebellum. A similar strategy for ROI sampling had been previously adopted in a PVC study by our group (16).

Raphe boundaries are best defined on PET transverse planes. Thus, PET images coregistered to (transverse and coronal) MR images were resampled to a common spatial resolution (0.94 × 0.94 × 0.94 mm) and subsequently resliced to transverse planes. Circular ROIs (about 6-mm diameter) were placed sequentially on the 13 coregistered PET slices that best visualized raphe activity. Each ROI was centered on the pixels with the highest activity. Small ROIs were chosen to minimize sensitivity to PVE.

ROI measurements were computed using only gray matter pixels, as defined by the gray matter segment (16). For cerebellum and raphe, where gray matter segment definition is not accurate, the mean ROI value was computed from unmasked PET images coregistered to MR volumes. Global gray matter values were calculated as cortical average values. PET measurements were not performed for thalamus and basal ganglia, where segmentation was poor. Statistical analysis was performed using t tests, and statistical significance set at P < 0.05.

PVC Algorithm

MR-based algorithms for PVC have been used successfully in PET studies of healthy aging and Alzheimer’s disease (10,11,16,26,27). In a previous study with ¹¹C-arachidonic acid, no increase in variability after PVC was found (16), whereas a significant increase in variability was detected for CMRglu in the current study. For PET data with large intersubject variability, such as FCWAY BP and arachidonic acid incorporation rate (16), the additional noise of PVC is undetectable. For CMRglu, with lower intersubject variability, a moderate increase in variability owing to PVC is detectable.

Choice of Binding Index

Uncorrected V_CEREB was significantly greater in patients than that in control subjects. Cerebellum is assumed to be devoid of 5-HT_1A receptors (19) and, thus, to provide a reliable estimate of free plus nonspecific binding (5–8). Since V_CEREB is proportional to f₁/f₂, where f₂ represents ¹⁸F-FCWAY brain free fraction (28), this finding might be related to a combination of changes in f₁ and f₂. Subtraction of higher V_CEREB for regional distribution volumes (V_ROI, Eq. 1) might, in principle, have overestimated group differences, most likely in low binding regions—that is, occipital and parietal lobes. However, no such differences were found before or after PVC. This issue might be more relevant if a reference tissue model were adopted (6,7) and a different BP derivation used. In this case, the BP (28), is equal to: Eq. 3 Thus, the denominator, not just the numerator (Eq. 1), of the ratio would be affected. When this model is used, the absence of significant group differences in cerebellum binding is assumed but cannot be verified (29,30).

The higher value of ¹⁸F-FCWAY free fraction (f₁) in patients than that in control subjects (0.121 ± 0.033 vs. 0.066 ± 0.024, respectively; P < 0.001) is most likely due to AED displacement of ¹⁸F-FCWAY from protein-binding sites (31). Almost all patients were taking at least one highly protein-bound AED, such as carbamazepine, phenytoin, or valproic acid, which could have led to this effect. Correction for f₁ eliminates the bias in distribution volume measurements that would be introduced by this difference.

Increases of ¹⁸F-FCWAY BP and CMRglu after PVC were lowest in amygdala, insula, and hippocampus and highest in occipital and parietal lobes. This finding appears to contrast with the difference between hippocampal and occipital involvement in TLE. However, regional differences in PVC effects are influenced by local geometry. Cortical gray matter pixels in thin gyri surrounded in 3 dimensions by several white matter pixels will have small s_GM and high s_WM values (Eq. 2). The opposite will occur in regions, such as hippocampus and amygdala, that are relatively homogeneous gray matter (16). This explains the strong linear relationship in Figure 2. The offset from the line of identity is due to a larger white matter spill-in for ¹⁸F-FDG than for ¹⁸F-FCWAY, since white matter BP is very small.

Biologic Interpretation

The BP used in this study (Eq. 1) is equivalent to B_max/K_d, where B_max (nmol/L/mL) is the maximum receptor concentration available for ligand binding, and K_d (nmol/L) is the ligand-receptor dissociation constant at equilibrium (28). Reduced BP may be due to decreased receptor or increased 5-HT concentration (decrease in B_max), decreased ligand receptor affinity (increased K_d), or a combination. Estimation of B_max and K_d would require at least 2 PET scans with different ¹⁸F-FCWAY specific activities. This is impractical in clinical studies. Moreover, ¹⁸F-FCWAY BP is not likely to be flow sensitive (32).

Previous studies in TLE patients reported greater decreases in ¹¹C-flumazenil binding than neuronal loss in mesial temporal regions ipsilateral to seizure foci (13). Autoradiography performed with ³H-flumazenil on surgical specimens showed reduced B_max and increased K_d (13). Preliminary results from an autoradiographic study on temporal lobectomy specimens from a subgroup of our patients indicated a decrease in B_max, with no changes in K_d in the hippocampus, whereas neuronal counts were diminished to a lesser extent (33).

Since some AEDs may increase 5-HT concentration (5), direct competition of the ligand with an increased 5-HT concentration cannot be excluded. However, for the serotonergic system, there is little evidence of sensitivity of high- affinity tracers, such as ¹⁸F-FCWAY, to changes in 5-HT concentration (34). Moreover, a preliminary analysis performed on PET data did not show significant AED effects on mesial and lateral temporal or raphe ¹⁸F-FCWAY BP values (35).

Our study, along with the preliminary autoradiographic results (33), suggests that receptor loss may be part of the initial phase of neuronal dysfunction in TLE, followed by hypometabolism and eventual structural atrophy. Since patients with new-onset epilepsy may have a low incidence of hypometabolism (36), ¹⁸F-FCWAY PET might be particularly useful for early detection of functional abnormalities in TLE patients. Early 5-HT_1A receptor loss might lower the seizure threshold (1,37). Results with both ¹¹C-flumazenil and ¹⁸F-FCWAY suggest that receptor-binding studies may be of clinical value for patients with uncontrolled epilepsy being considered for surgery, particularly when other imaging studies are equivocal (5,13). Studies with ligands such as 5-HT_2A receptors or the serotonin transporter, not yet performed in epilepsy patients, might be of interest as well.

The physiologic events underlying 5-HT_1A binding reduction in raphe remain elusive. In contrast to cortex, 5-HT_1A raphe receptors are somatodendritic. After rapid increases in extracellular 5-HT, these receptors inhibit serotonergic neuronal firing rate and, thus, 5-HT release in cortical areas (38). Desensitization of 5-HT_1A autoreceptors occurs after 3–4 wk of antidepressant therapy, and a similar mechanism cannot be excluded in TLE patients (38).

Epilepsy and Depression

Depression is found frequently in TLE patients. In a previous study, a significant inverse correlation between the Montgomery Åsberg Depression Rating scale and ipsilateral anterior cingulate 5-HT_1A binding was reported (8). We did not detect such a relationship, possibly owing to the lack of specific frontal lobe subregion sampling (39). However, we found a significant inverse relationship, both before and after PVC, between BDI and ¹⁸F-FCWAY BP in hippocampus ipsilateral to the seizure focus (Fig. 3). Interestingly, no significant relationship was detected between BDI and BP or CMRglu in any other region. Previous PET studies in patients with depression reported reduced 5-HT_1A receptor binding in regions relevant to epilepsy, such as raphe nuclei and the amygdala-hippocampus complex (40). Additional studies will be necessary to assess the relation of 5-HT_1A receptors to depression in TLE.