A PET study on regional coexpression of 5-HT1A receptors and 5-HTT in the human brain
Several lines of evidence suggest inter-dependency between the serotonin transporter (5-HTT) and the 5HT1A receptor, two recognised targets for the treatment of anxiety and depression.
to examine the correlation of regional expression levels for these two serotonergic markers in the human brain in vivo.
Twelve male control subjects were examined with PET twice on the same day, using the radioligands [11C]WAY 100635 and [11C]MADAM for quantification of the 5-HT1A receptor and the 5-HTT, respectively. The binding potential (BP) was calculated for raphe nuclei, hippocampus and frontal cortex.
In all regions, the BP for both [11C]WAY 100635 (raphe nuclei 1.85–4.71, hippocampus 2.52–6.17, frontal cortex 2.03–3.79) and [11C]MADAM (2.70–7.65, 0.47–1.76, 0.18–0.51) varied several fold between subjects. In the raphe nuclei, where the two markers are situated on the same neurons, the ratio of [11C]WAY 100635 binding to [11C]MADAM BP binding varied considerably (0.43–1.05). There was a positive correlation between the two markers in the raphe nuclei (rxy = 0.68, p < 0.05) and in the hippocampus (rxy = 0.97, p < 0.001) but not in the frontal cortex (rxy = −0.25, p = 0.44).
The results support a correlation between density levels of the 5-HT1A-receptor and the 5-HTT in the raphe nuclei and hippocampus but not in the frontal cortex. A suggested clinical implication is that the inter-individual variability in 5-HT1A-receptor and 5-HTT densities, as well as the ratio of these, is of particular interest in relation to individual responses to selective serotonin reuptake inhibitor treatment.
KeywordsPET5-HT1A-receptor5-HTT[11C]MADAM[11C]WAY 100635HumanIn vivo
Of the 14 5-HT receptors described so far (Hoyer and Martin 1997), the 5-HT1A subtype is one of the best characterised. This G-protein coupled receptor has a wide anatomical distribution in the human brain with highest concentration in the raphe nuclei, limbic structures and neocortex (Hoyer et al. 1986; Verge et al. 1986). A line of evidence from animal studies support its role in a variety of brain functions involving both cognition and emotion (for reviews, see e.g. Barnes and Sharp 1999; Buhot 1997; Pucadyil et al. 2005). These findings have not been confirmed in humans in vivo. Interestingly, in a recent PET study on 24 male control subjects, no significant correlation was found between inter-individual variability in 5-HT1A receptor density and cognitive performance (Borg et al. 2006).
While the 5-HT1A subtype is a postsynaptic receptor in serotonergic projection areas, it is situated presynaptically on cell bodies in the dorsal raphe nuclei (DRN), where it is the only 5-HT-receptor subtype hitherto identified (Barnes and Sharp 1999; Miquel et al. 1991; Sotelo et al. 1990). As these autoreceptors mediate inhibition of cell firing and, thereby, serotonin release in projection areas, it has been suggested that the 5-HT1A receptor in the raphe may have a role in the general regulation of serotonergic activity (Artigas et al. 1996; Borg et al. 2003).
The serotonin transporter (5-HTT) is another functionally important protein in serotonergic transmission. The 5-HTT protein is only expressed on 5-HT neurons and controls the 5-HT concentration in the synaptic cleft (Bengel et al. 1997; Fujita et al. 1993; Qian et al. 1995). With its high concentration on cell bodies and nerve terminals of the raphe nuclei, it may be in an equally good position as the 5-HT1A receptor to have a central role in regulation of serotonergic neurotransmission (Cortes et al. 1988; Hoffman et al. 1998; Plenge et al. 1990; Zhou et al. 1996).
The two proteins are of central interest in psychopharmacology. The 5-HTT is the target for selective serotonin reuptake inhibitors (SSRI), which have a well documented effect in the treatment of depression and anxiety disorders (for recent reviews, see Cryan et al. 2005; Vaswani et al. 2003). The 5-HT1A receptor is the target for the partial agonist buspirone, which is prescribed in anxiety disorders, and, interestingly, the 5-HT1A receptor partial agonist pindolol has been suggested to accelerate the onset of the antidepressant effect of SSRIs. The presynaptic 5-HT1A receptor inhibits 5-HT neurotransmission when stimulated. An acute effect of SSRI treatment is increased 5-HT concentration in the raphe nuclei and a 5-HT1A receptor-mediated decrease in 5-HT neurotransmission. This decrease is normalised in chronic SSRI administration because of a 5-HT1A receptor downregulation (Artigas et al. 1996; Ballesteros and Callado 2004; Hales et al. 1997).
The short (S) allele of the frequent 5-HTT gene (SLC6A4) polymorphism, 5-HTT gene-linked polymorphic region has been associated with restricted transcriptional activity in vitro (Collier et al. 1996). Interestingly, carriers of the S allele have also been shown to have an increased frequency of anxiety and mood disorders (Caspi et al. 2003; Heils et al. 1995; Lesch et al. 1996; Melke et al. 2001). In addition, it has been reported that carriers of the S allele have a lower 5-HT1A receptor density (David et al. 2005). This finding suggests a dependency between the expression levels of these two markers for serotonergic neurotransmission.
The possible relationship between 5-HT1A and 5-HTT gene expression levels has, to some extent, been approached experimentally. In 5-HTT knock-out (KO) mice, both 5-HT1A receptor proteins and mRNA have been shown to be decreased in the DRN, increased in the hippocampus and unchanged in other forebrain areas (Fabre et al. 2000). A pharmacological examination of the 5-HT1A receptor mediated cell activity in 5-HTT KO mice showed, correspondingly, a desensitisation in the DRN but no alteration in the hippocampus (Mannoury la Cour et al. 2001). In an autoradiography study on prefrontal cortex of suicide victims and control subjects, a negative correlation was found between the two markers, suggesting a common regulatory factor (Arango et al. 1995).
These experimental observations in animal models and humans together with the molecular genetic findings may support a relationship between expression levels of the two 5-HT markers. A direct examination of this hypothesis in humans in vivo has recently been allowed by the availability of [11C]WAY 100635 and [11C]MADAM, two suitable PET radioligands for regional quantification of 5-HT1A receptor and 5-HTT binding also in raphe (Farde et al. 1998; Gunn et al. 2000; Lundberg et al. 2005). PET studies of control subjects have shown an at least twofold inter-individual variability in protein expression levels of both 5-HTT and the 5-HT1A receptor (Borg et al. 2003; Lundberg et al. 2005). This high variability provides an advantageous biological condition for correlative studies on these markers.
The aim of the present study was to examine the regional correlation between expression levels of the 5-HT1A receptor and the 5-HTT in the human brain. Twelve control subjects were examined with [11C]WAY 100635 and [11C]MADAM in the same day. The selection of regions included DRN, where both markers are expressed on the 5-HT neurons and the projection areas frontal cortex and the hippocampal complex, where 5-HTT is expressed presynaptically and the 5-HT1A receptor is expressed postsynaptically.
Materials and methods
Subjects and design
The study was approved by the Ethics and the Radiation Safety committees of the Karolinska Hospital. Twelve male subjects, aged 22–55, participated after giving informed consent. All subjects were healthy according to history, psychiatric interview, physical examination, blood and urine analysis and magnetic resonance imaging (MRI) of the brain. They did not use any medication and they were all non-smokers. All subjects were examined with PET twice in the same day: at 11 am immediately after IV injection of [11C]MADAM and at 2 pm immediately after IV injection of [11C]WAY-100635 (eight subjects) or in the opposite order (four subjects).
MRI and the head fixation system
The MRI system used was Signa, 1.5 T (GE Medical Systems, Milwaukee, WI, USA). T2- and T1-weighted images (matrix 256 × 256 × 156; pixel size 1.0156 × 1.0156 × 1.0 mm) were obtained. The T2 sequence is a 2-D fast spin echo protocol with the following settings: relaxation time (TR) 5,000 ms, echo time (TE) 68 ms, axial field of view (FOV) 26 cm, 260 × 260 matrix, 44 × 3.0 mm slices, 1 number of excitations (NEX) 4 min. The slice gap is set to 0.125 mm to make the centre-to-centre distance 3.125 mm, which is that of the ECAT PET images. The T1 sequence is a 3-D spoiled gradient recalled protocol with the following settings: TR 23 ms, TE 4 ms, flip angle 50°, FOV 260 × 180 × 156, matrix 256 × 192 × 156, 156 × 1.0 mm slices, and 1 NEX 8 min 45 s. To allow the same head positioning in the two imaging modalities, a head fixation system with an individual plaster helmet was used in both the PET and MRI measurements (Bergstrom et al. 1981).
[11C]MADAM was obtained by methylation of ADAM using [11C]methyl iodide, as described previously (Hall et al. 1997; Tarkiainen et al. 2001). Between 286 and 318 MBq was injected intravenously. The specific radioactivity of the radioligand injected varied between 196 and >100,000 Ci/mmol, corresponding to a mass injected of <0.01 to 11.9 μg.
[11C]WAY-100635 was prepared from 11C-acylation of WAY100634 with carbonyl-11C-cyclohexanecarbonyl chloride, as described previously (Hall et al. 1997). Between 146 and 323 MBq was injected intravenously. The specific radioactivity of the radioligand injected varied between 199 and 1,983 Ci/mmol, corresponding to a mass injected of 1.5 to 15.2 μg. The variation of specific activity and injected mass was due to technical problems with the cyclotron unit.
PET experimental procedure
The PET system used was ECAT EXACT HR 47 (Siemens, Berlin and Munich, Germany), which was run in the 3-D mode (Wienhard et al. 1994). The in-plane and axial resolution are about 3.8 and 4.0 mm, respectively, full width at half maximum. The reconstructed volume was displayed as 47 sections with a centre-to-centre distance of 3.125 mm.
In each PET measurement, the subject was placed recumbent with his head in the PET system. A sterile physiological phosphate buffer (pH = 7.4) solution containing the radioligand was injected as a bolus for 2 s into a cannula inserted into the right antecubital vein. The cannula was then immediately flushed with 10 ml saline.
Brain radioactivity was measured in a series of consecutive time frames. After injection with [11C]MADAM, the examination lasted for 93 min and consisted of 20 frames (3 × 1′; 4 × 3′; 13 × 6′), except in one subject where the examination lasted for 87 min (3 × 1′; 4 × 3′; 12 × 6′). After injection with [11C]WAY100635, the examination lasted for 69 min and consisted of 16 frames (3 × 1′; 4 × 3′; 9 × 6′).
In addition to an elaborated head fixation system allowing for repositioning within less than 3 mm (Bergstrom et al. 1981), a coregistration procedure was applied. For each subject, the MR image was adjusted to position the anterior–posterior commissural (AC–PC) line in the horizontal plane, and the inter-hemispheric plane orthogonal to the AC–PC plane. It was resampled and cropped to generate a 256 × 256 × 141 matrix with 1 mm2 pixels before it was used for manual definition of regions of interest (ROIs). The PET images were resampled to a 2-mm2 pixel size and coregistered to a corresponding MRI half-resolution dummy. Coregistration was done using SPM2 (Maes et al. 1997).
Rationale for selection of ROIs
The aim of the study was to examine regional correlation between the two markers and to contrast regions with the two markers on the same neuron (the raphe nuclei) to regions where they are situated pre- (5-HTT) and postsynaptically (the 5-HT1A-receptor; hippocampus, frontal cortex). The latter two were chosen as they, together with the raphe nuclei, represent three different developmental levels of the central nervous system (brainstem, allocortex, isocortex).
Definition of ROIs
ROIs were defined according to anatomical boundaries for neocortex, hippocampal complex, raphe nuclei and cerebellum. All ROIs but those for raphe nuclei were delineated in ten consecutive sections on the MR images and transferred to the corresponding reconstructed PET images.
On MR images the raphe nuclei cannot be differentiated from surrounding tissue. Therefore, these ROIs were delineated directly on the summated PET images (frame 6-onwards) of [11C]MADAM and [11C]WAY 100635 in three to five sections, according to a method previously applied in PET measurements of the 5-HTT and 5-HT1A proteins in the raphe nuclei (Andree et al. 2002; Lundberg et al. 2005).
Time activity curves and quantitative analysis
To obtain the average radioactivity concentration for the whole volume of interest, data for each ROI were pooled. Regional radioactivity was calculated for each frame, corrected for decay and plotted vs time, thus providing time–activity curves for each region. Right and left regions for bilateral ROIs were analysed both separately and averaged.
Binding potentials (BPs; Bmax/Kd) (Mintun et al. 1984) were calculated by means of the simplified reference tissue model (Lammertsma and Hume 1996). The applicability of this method using the cerebellum as the reference region has previously been shown for both [11C]WAY-100635 and [11C]MADAM (Gunn et al. 1998; Lundberg et al. 2005).
For each region, the association between [11C]MADAM and [11C]WAY 100635 binding was examined by calculation of the Pearson’s correlation coefficient using SPSS 12.0.1 for Windows. Two-tailed tests for significance were performed and the results were corrected for multiple comparisons.
[11C]WAY 100635 and [11C]MADAM binding potentials (BPs) in the three examined regions
[11C]WAY 100635 BP
Mean ± SD
Mean ± SD
3.26 ± 1.13
4.64 ± 1.67
4.69 ± 1.39
0.81 ± 0.33
2.78 ± 0.56
0.29 ± 0.11
Experimental observations together with molecular genetic findings support a postulated relationship between expression levels of the 5-HT1A receptor and the 5-HT transporter. The large inter-individual variability in these expression levels detected in previous PET-studies provides a possibility to examine this relationship directly in humans in vivo.
In the present study in adult male subjects, there was a positive correlation between 5-HTT protein and 5-HT1A receptor protein binding in the raphe nuclei and in the hippocampal complex. No correlation was found in the frontal cortex. In the following, the implication of these findings will be discussed in relation to the organisation of the serotonin system and the pharmacology of mood and anxiety disorders.
In the raphe nuclei, the proteins are most likely situated on the same neurons (Burnet et al. 1995; Hoffman et al. 1998; Miquel et al. 1991; Sur et al. 1996; Zhou et al. 1996). This colocalisation of the proteins provides a neuroanatomical underpinning for the assumed linear correlation between expression levels should the density be an index of the number of neuronal cell bodies alone. Our data set can thus verify such a correlation.
In clinical SSRI treatment of depression, the response rate has been reported to be about 2/3 (Barbui and Hotopf 2001). A suggested interpretation of the relatively low response rate is that clinical patient samples might be heterogeneous with regards to aetiology. An individual variability in the molecular response to SSRI treatment is an alternative reading. Along this line, it has been suggested that a large group of non-responders share an unfavourable biochemical phenotype (Kampf-Sherf et al. 2004). Indeed, using single photon emission tomography and [123I]β-CIT high 5-HTT binding in diencephalon prior to SSRI treatment has been shown to predict better treatment response in subjects with major depression (Kugaya et al. 2004). In the present study, using PET and [11C]MADAM, we confirm a several-fold inter-individual variability in 5-HTT binding in three different brain regions. Thus, we conclude that individual 5-HTT binding, i.e. the target for SSRIs, might serve as a marker to predict SSRI response.
The 5-HT1A receptor has also been discussed extensively in relation to SSRI treatment of depression. The response rate has been shown to increase by co-administration with the 5-HT1A/β-adrenoreceptor partial agonist pindolol (Ballesteros and Callado 2004). Moreover, a substantial body of preclinical evidence suggests that this effect of pindolol is due to inhibition of somatodendritic 5-HT1A receptors that otherwise mediate an acute SSRI-induced decrease in 5-HT transmission (for a review, see Artigas et al. 1996). In our data set, there was a large inter-individual variability in the raphe nuclei 5-HT1A-receptor/5-HTT BP ratio (Fig. 2 range 0.43–1.10, mean 0.73, SD 0.21;). It would be of interest to examine if this variability also exists in depressed patients and if it also may be associated with SSRI treatment response.
The 5-HT1A-receptor and the 5-HTT have hitherto been used separately in research of the relation between mood disorders and the serotonin system in a large number of PET studies (Bhagwagar et al. 2003, 2004; Drevets et al. 1999; Meltzer et al. 2004; Meyer et al. 2004; Parsey et al. 2006a, b, c, d; Sargent et al. 2000). Importantly, the large inter-individual HT1A-receptor/5-HTT BP ratio variability in our data set suggests the two markers not to be mutually exchangeable but rather examined in parallel when used as markers for the 5-HT system.
In neocortex and hippocampal complex, the 5-HT1A-receptor and 5-HTT are expressed on separate neurons. The 5-HTT is most likely expressed presynaptically on 5-HT neurons projecting from the raphe nuclei and the 5-HT1A-receptor postsynaptically on glutamatergic pyramidal cells (Barnes and Sharp 1999; Burnet et al. 1995; Hoffman et al. 1998). A functional relation between the two molecules is still suggested from a study on prenatally stressed rats used as a model for depression. Here, 5-HTT blockade with imipramine was shown to normalise an initial 5-HT1A receptor mRNA level increase (Morley-Fletcher et al. 2004). In the present study, at the protein expression level, this could be confirmed in the hippocampal complex where a significant correlation between 5-HT1A-receptor and 5-HTT binding was found (Fig. 3b). Although this correlation to a large degree was driven by one subject, it also remained significant after exclusion of this subject (rxy = 0.63, p < 0.05).
It cannot be ruled out that the lack of correlation between 5-HT1A-receptor and 5-HTT binding in the frontal cortex is due to the fact that 5-HTT expression is driven not primarily by neurodevelopment but rather by ongoing regulatory mechanisms. One or more of the three other 5-HT receptors hitherto detected in cortex (the 1B, 1E, 1F, 2A and 2C subtypes) may have a role in this regard (Barnes and Sharp 1999; Moret and Briley 2000). Another possible reason for the lack of a significant correlation may be the less favourable [11C]MADAM signal-to-noise ratio in the neocortex (Lundberg et al. 2005).
Additionally, the finding in the raphe nuclei has to be viewed in relation to the reliability of the methodology. The DRN is a thin elongated structure of only about 70 mm3 (Baker et al. 1990). Its small size is not ideal for reliable PET measurements. This is true for both [11C]WAY 100635 and [11C]MADAM, and the findings thus have to be interpreted with some caution.
A strength of the present study is that the measurements of [11C]MADAM and [11C]WAY 100635 binding was performed with only about 1 h interval, giving a minimal possibility for changes in 5-HT1A receptor and 5-HTT densities to take place during the course of the study. The BP is related to the protein density (Bmax) and the apparent in vivo affinity for the radioligand binding to the protein examined (Kd). inter-individual differences in BP may thus be related to variation in Bmax or Kd (Mintun et al. 1984). In healthy volunteers, inter-individual variance in BP has been suggested to be due mainly to variability in Bmax rather than Kd (Farde et al. 1995). Accordingly, a basic assumption in this study on protein densities is that BP differences between individuals are mainly based on variation in Bmax. Another limitation is the small number of subjects examined. This is, however, compensated for by the considerable inter-individual variability in both [11C]MADAM and [11C]WAY 100635 binding previously reported (Borg et al. 2003; Lundberg et al. 2006). ROIs for the raphe nuclei were defined directly on the summated [11C]MADAM and [11C]WAY 100635 PET images. This allows for differences in partial volume effect on the calculation of BP for the two ligands. However, as the ROI volumes for the raphe nuclei did not differ statistically [ROI volume[11C]MADAM = 107 ± 45, ROI volume[11C]WAY 100635 = 98 ± 46 (mm3, mean ± SD, p = 0.33, paired t test, two-tailed)], this would not influence the result to a significant degree.
Our result provides support for a hypothesised correlation in expression of the 5-HT1A-receptor and the 5-HTT in the raphe nuclei. The large inter-individual variability in BP quotient in this region may represent a potential biological phenotype of SSRI non-responders. This hypothesis should be addressed in future studies. In addition, the variability, which is also present in the hippocampal complex and neocortex, suggests the two markers not to be mutually exchangeable but rather examined in parallel when used as markers for the 5-HT system.
All members of the PET group at Karolinska Institutet are greatly acknowledged. This work was supported by the Swedish Research Council (grant 09114). The experiments comply with Swedish law.
Statement of interest