Introduction
A central role for 5-hydroxytryptamine1A (5-HT1A) receptors in the pathophysiology of depression and mechanism of action of antidepressants is proposed (Deakin & Graeff, Reference Deakin and Graeff1991; Blier & de Montigny, Reference Blier and de Montigny1994; Savitz et al. Reference Savitz, Lucki and Drevets2009; Albert et al. Reference Albert, Le Francois and Millar2011). 5-HT1A receptors are expressed centrally in two functionally distinct locations – on the soma and dendrites of the serotonergic neurones of the raphé nuclei (‘presynaptic’ or ‘somatodendritic’ receptors) and postsynaptically (Albert et al. Reference Albert, Zhou, Van Tol, Bunzow and Civelli1990; Aznavour et al. Reference Aznavour, Benkelfat, Gravel, Aliaga, Rosa-Neto, Bedell, Zimmer and Descarries2009). Activation of somatodendritic 5-HT1A receptors is inhibitory (Kelly et al. Reference Kelly, Larkman, Penington, Rainnie, McAllister-Williams and Hodgkiss1991; McAllister-Williams & Kelly, Reference McAllister-Williams and Kelly1995), reducing the firing rate of raphé neurones and thus the release of 5-HT onto postsynaptic receptors in projection areas (Hjorth & Sharp, Reference Hjorth and Sharp1991).
While there are consistent data suggesting a decrease in the number and function of postsynaptic 5-HT1A receptors in depression (Savitz et al. Reference Savitz, Lucki and Drevets2009), the situation is less clear with regards to somatodendritic autoreceptors. Early positron emission tomography (PET) studies suggested a decrease in the number of somatodendritic 5-HT1A binding sites in currently depressed (Drevets et al. Reference Drevets, Frank, Price, Kupfer, Holt, Greer, Huang, Gautier and Mathis1999, Reference Drevets, Thase, Moses-Kolko, Price, Frank, Kupfer and Mathis2007; Sargent et al. Reference Sargent, Kjaer, Bench, Rabiner, Messa, Meyer, Gunn, Grasby and Cowen2000; Meltzer et al. Reference Meltzer, Price, Mathis, Butters, Ziolko, Moses-Kolko, Mazumdar, Mulsant, Houck, Lopresti, Weissfeld and Reynolds2004) and remitted patients (Bhagwagar et al. Reference Bhagwagar, Rabiner, Sargent, Grasby and Cowen2004). However a more recent study, while finding a significant reduction in postsynaptic 5-HT1A receptors, found no change in binding in the raphé (Hirvonen et al. Reference Hirvonen, Karlsson, Kajander, Lepola, Markkula, Rasi-Hakala, Nagren, Salminen and Hietala2008), while another group has found an increase in raphé binding (Parsey et al. Reference Parsey, Oquendo, Ogden, Olvet, Simpson, Huang, Van Heertum, Arango and Mann2006) which they have replicated (Parsey et al. Reference Parsey, Ogden, Miller, Tin, Hesselgrave, Goldstein, Mikhno, Milak, Zanderigo, Sullivan, Oquendo and Mann2010). It is possible that these discrepancies resulted from whether or not antidepressant-naive subjects were studied, whether the binding was estimated using an arterial blood input and/or the reference region used in the analysis (Hirvonen et al. Reference Hirvonen, Karlsson, Kajander, Lepola, Markkula, Rasi-Hakala, Nagren, Salminen and Hietala2008; Savitz et al. Reference Savitz, Lucki and Drevets2009; Parsey et al. Reference Parsey, Ogden, Miller, Tin, Hesselgrave, Goldstein, Mikhno, Milak, Zanderigo, Sullivan, Oquendo and Mann2010). Post-mortem studies have also been inconsistent with decreases (Arango et al. Reference Arango, Underwood, Boldrini, Tamir, Kassir, Hsiung, Chen and Mann2001) and increases (Stockmeier et al. Reference Stockmeier, Shapiro, Dilley, Kolli, Friedman and Rajkowska1998) in somatodendritic 5-HT1A receptors reported. The data described by Arango et al. (Reference Arango, Underwood, Boldrini, Tamir, Kassir, Hsiung, Chen and Mann2001) have been reanalysed, suggesting one possible explanation of such inconsistencies. This analysis revealed that in 10 suicide victims there was a higher level of 5-HT1A binding caudally but lower binding rostrally compared with 10 controls (Boldrini et al. Reference Boldrini, Underwood, Mann and Arango2008). What the functional implications might be of regionally divergent differences in depressed patients is unclear.
The possibility that somatodendritic 5-HT1A receptors might be increased in depression is supported by genetic studies. A human C(−1019)G 5-HT1A polymorphism has been identified in the repressor/enhancer region of the receptor's gene. The G allele has been found to be associated with depression and suicide (Albert et al. Reference Albert, Le Francois and Millar2011). The effect of this allele is dependent on cell type. In postsynaptic, non-serotonergic, cells it is associated with enhancement of the repression of 5-HT1A receptor expression, but an inhibition of repression, and hence overexpression of receptors, in the raphé is seen (Czesak et al. Reference Czesak, Lemonde, Peterson, Rogaeva and Albert2006). It has been reported that an increase in raphé 5-HT1A receptor binding found in PET studies of depressed patients correlates with the G/G genotype (Parsey et al. Reference Parsey, Oquendo, Ogden, Olvet, Simpson, Huang, Van Heertum, Arango and Mann2006, Reference Parsey, Ogden, Miller, Tin, Hesselgrave, Goldstein, Mikhno, Milak, Zanderigo, Sullivan, Oquendo and Mann2010).
More important than the simple number of binding sites is the functional status of 5-HT1A receptors. Administration of 5-HT1A agonists leads to the release of cortisol, prolactin and growth hormone (GH), though this is a postsynaptic receptor response (Cowen, Reference Cowen2000; Blier et al. Reference Blier, Seletti, Gilbert, Young and Benkelfat2002). The hypothermic response to 5-HT1A agonists has been argued to result from somatodendritic 5-HT1A activation (Lesch et al. Reference Lesch, Mayer, Disselkamp-Tietze, Hoh, Schoellnhammer and Schulte1990), with this effect being blunted in depression (Lesch et al. Reference Lesch, Mayer, Disselkamp-Tietze, Hoh, Schoellnhammer and Schulte1990; Meltzer & Maes, Reference Meltzer and Maes1995). However, while a somatodendritic origin for this hypothermia is supported in mice due to the effect being blocked by the 5-HT neurotoxin 5,7-dihydroxytryptamine (Martin et al. Reference Martin, Phillips, Hearson, Prow and Heal1992), in man tryptophan depletion has no significant effect on the degree of hypothermia induced by buspirone, suggesting that it predominantly results from postsynaptic 5-HT1A activation (Blier et al. Reference Blier, Seletti, Gilbert, Young and Benkelfat2002). An alternative method of assessing somatodendritic 5-HT1A receptors is the effect of the 5-HT1A agonist buspirone on the EEG frequency spectrum (Alhaj et al. Reference Alhaj, Wisniewski and McAllister-Williams2011). Buspirone causes a negative shift in the theta and slow α frequency ranges (Murasaki et al. Reference Murasaki, Miura, Ishigooka, Ishii, Takahashi and Fukuyama1989; Barbanoj et al. Reference Barbanoj, Anderer, Antonijoan, Torrent, Saletu and Jané1994; Holland et al. Reference Holland, Wesnes and Dietrich1994; Anderer et al. Reference Anderer, Saletu and Pascual-Marqui2000; McAllister-Williams & Massey, Reference McAllister-Williams and Massey2003). Although buspirone is a relatively non-selective drug (it is a dopamine D2 antagonist) and systemic administration makes localization of effects difficult, several pieces of evidence indicate that this negative shift in cortical EEG originates from activation of somatodendritic 5-HT1A receptors. First, the EEG frequency effects are mimicked by other more selective 5-HT1A agonists in man (Seifritz et al. Reference Seifritz, Moore, Trachsel, Bhatti, Stahl and Gillin1996; Anderer et al. Reference Anderer, Saletu and Pascual-Marqui2000) and are seen in animals following administration of buspirone as well as the selective full 5-HT1A receptor agonist 8-OH-DPAT [8-hydroxy-2-(di-n-propylamino)tetralin], but not with the D2 antagonist haloperidol (Bogdanov & Bogdanov, Reference Bogdanov and Bogdanov1994). Second, source localization of the effect of buspirone demonstrates a significant increase in theta EEG activity in the hippocampus as well as neighbouring cortical areas (Anderer et al. Reference Anderer, Saletu and Pascual-Marqui2000). Hippocampal theta is well known to be under ascending serotonergic control. Local application of 8-OH-DPAT into the raphé activating somatodendritic 5-HT1A receptors causes an increase in hippocampal theta (Vertes et al. Reference Vertes, Kinney, Kocsis and Fortin1994; Nitz & McNaughton, Reference Nitz and McNaughton1999), an effect blocked by 5-HT1A antagonists (Marrosu et al. Reference Marrosu, Fornal, Metzler and Jacobs1996). Third, acute administration of selective serotonin reuptake inhibitors (SSRIs), which increase 5-HT in the raphé leading to somatodendritic 5-HT1A activation (Bel & Artigas, Reference Bel and Artigas1992), causes a similar shift in the EEG (Saletu et al. Reference Saletu, Grunberger and Linzmayer1986). Fourth, pindolol, which acts as a 5-HT1A antagonist at postsynaptic receptors but a partial agonist at somatodendritic receptors (Clifford et al. Reference Clifford, Gartside, Umbers, Cowen, Hajos and Sharp1998), mimics rather than blocks the effect of buspirone (McAllister-Williams & Massey, Reference McAllister-Williams and Massey2003).
Here we utilize the effect of buspirone on the EEG frequency spectrum as an index of somatodendritic 5-HT1A function to investigate the status of these receptors in a cohort of drug-free, currently depressed, patients versus healthy controls. We hypothesize that depressive illness is associated with an increase in receptor function.
Method
Subjects
A total of 15 patients (nine males) with Semi-Structured Interview Schedule (SCID; First et al. Reference First, Spitzer, Gibbon and Williams1997) confirmed Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) diagnoses of major depressive disorder and 15 age-, intelligence quotient (IQ)- and gender-matched healthy controls took part in this study. Depressed patients and healthy controls were recruited by advertisements and from general practitioner referrals. The inclusion criteria required a verbal IQ of 90 or more as assessed by the National Adult Reading Test and right-handedness as assessed using Briggs' handedness inventory (Briggs & Nebes, Reference Briggs and Nebes1975). Patients and controls were excluded if they had used any psychotropic medication within the previous 6 weeks. Patients with psychotic symptoms or meeting the diagnostic criteria for additional psychiatric disorders (including personality disorder and substance misuse) were excluded. Controls were not included if they had a history of any DSM-IV Axis I or II diagnosis or had a first-degree relative with a history of any DSM-IV Axis I diagnosis.
The severity of depression was assessed at baseline using the 17-item version of the Hamilton Depression Rating Scale (HAMD; Hamilton, Reference Hamilton1967), Montgomery–Asberg Depression Rating Scale (MADRS; Montgomery & Asberg, Reference Montgomery and Asberg1979) and the Beck Depression Inventory (BDI; Beck et al. Reference Beck, Ward, Mendelson, Mock and Erbaugh1961) ratings of mood. Patients were required to have a score of ⩾17 points on the HAMD, while controls had to score < 8 points for inclusion.
Written informed consent was obtained from all participants. Ethical approval was granted by the local research ethics committee. Control subjects were paid an honorarium of £40 as compensation for time spent in taking part in the study and patients and controls were reimbursed for any expenses (such as travel) incurred, in line with the ethical approval obtained for the study.
Experimental design and procedure
On the day of investigation, subjects reported to the Department of Psychiatry, Royal Victoria Infirmary, Newcastle upon Tyne, at 09:00 hours. Mood was rated using the HAMD, MADRS and BDI. Subjects were seated in the EEG recording facility, fitted with an electrode cap and given free access to drinking water but were otherwise fasted during the experiment. They completed a neurocognitive task reported elsewhere (Alhaj et al. Reference Alhaj, Massey and McAllister-Williams2007) followed by the 5-HT function task reported here. At baseline (approximately 12.00 hours) five visual analogue scales (VASs) were completed, rating symptoms of ‘depression’, ‘drowsiness’, ‘restlessness’, ‘nausea’ and ‘light-headedness’. EEG was recorded, a blood sample (via cannula inserted for the duration of the study) was drawn for estimations of GH, prolactin and cortisol concentrations, and sub-lingual temperature measured using a standard clinical digital thermometer. Following baseline measurements, buspirone (30 mg) was administered orally. Further sets of measurements (VASs, blood samples, temperature and EEG recording) were conducted 25–40, 55–70 and 85–100 min (referred to as t = 30, 60 and 90 min) after buspirone administration.
EEG recordings and analysis
EEG recordings and analysis were as previously described (McAllister-Williams & Massey, Reference McAllister-Williams and Massey2003). EEG was recorded from 29 silver/silver chloride electrodes positioned using an elasticated cap (Easy Caps, Germany) and sited in accordance with the international 10–20 system. Further electrodes were placed on the right and left mastoid processes. All channels were recorded relative to the left mastoid. Vertical electro-occulograms (EOGs) was recorded between electrodes placed on the nazion and electrodes below the centre of each eye. A horizontal EOG was recorded between electrodes placed on the outer canthus of the eyes. EEG and EOG were filtered with a bandpass of 0.1–100 Hz and sampled at a rate of 400 Hz. At baseline and t = 30, 60 and 90 min, subjects were instructed to keep their eyes open and to maintain their gaze on the fixation point (a red cross) displayed on a computer monitor. They were asked to remain still and relaxed for three 3-min periods interspersed with two 30-s rest intervals between while continuous EEG was acquired during this 10-min period. In between EEG recordings, subjects rested and watched emotionally neutral videos.
A standardized EEG analysis procedure was followed using Neuroscan Scan 4.3 software (Neurosoft Inc., USA). This included blink-correction (Semlitsch et al. Reference Semlitsch, Anderer, Schuster and Presslich1986) and a principle component analysis to remove ECG artefact. Data were algebraically re-referenced to represent recordings from an average mastoid reference and epoched into segments 10.24 s long (to give a power of two number of points). Any epoch in which any channel, except vertical EOG, had a voltage deflection greater than ± 75 μV was excluded to remove residual artefacts. Remaining epochs underwent fast Fourier transformation (FFT) and were averaged. The precision of the FFT was 0.098 Hz. Potential shifts in the EEG frequency spectrum were assessed by calculating the centroid frequency between 6 and 10.5 Hz spanning theta and slow α frequency bands as previously described (Anderer et al. Reference Anderer, Barbanoj, Saletu and Semlitsch1993; McAllister-Williams & Massey, Reference McAllister-Williams and Massey2003; McAllister-Williams et al. Reference McAllister-Williams, Massey and Fairchild2007).
Plasma endocrine assays
Blood samples were taken into EDTA (ethylenediaminetetra-acetic acid) tubes, and centrifuged at 3000 rpm for 10 min. Plasma was removed and stored at −20°C. Samples were analysed for GH, prolactin and cortisol using radioimmunoassay (RIA) kits (Immuno Diagnostic Systems Ltd, UK). Intra- and inter-assay coefficients of variation were 10.0% and 6.6% for GH, 3.2% and 3.4% for prolactin and 8.1% and 7.6% for cortisol, respectively. Of the participants, one patient and one healthy control had some samples that were spoilt or not obtainable. The data on the endocrine effects of buspirone administration therefore represent those from 14 patients and 14 controls.
Hypothalamic–pituitary–adrenal (HPA) axis functional assessment
HPA axis function was assessed by measuring morning and evening salivary cortisol concentrations (Goodyer et al. Reference Goodyer, Herbert, Altham, Pearson, Secher and Shiers1996) on two consecutive days prior to the test, collected using standard saliva collection devices (‘Salivettes’, Sarstedt, Germany) at 08:00 and 20:00 hours. Tubes were stored in each subject's fridge until the time of their visit. Cortisol assays were performed by means of RIA using Gamma-B 125I corticosteroid RIA kits obtained from Immunodiagnostic Systems Ltd (UK). Intra- and inter-assay coefficients of variation were 6.8 ± 0.46% and 4.7 ± 0.22%.
Statistical analysis
SPSS (version 19.0; IBM, USA) was utilized throughout. For the analysis of the shift in EEG frequency spectrum, the change in centroid frequency, calculated between 6 and 10.5 Hz, following buspirone administration was quantified as an area under the curve (AUC) at t = 30, 60 and 90 min relative to that at baseline (t = 0) using the trapezoid method. As such, this measure is ‘AUC-ground’ rather than ‘AUC-increase’ as previously defined (Fekedulegn et al. Reference Fekedulegn, Andrew, Burchfiel, Violanti, Hartley, Charles and Miller2007). The AUC data from all 29 scalp electrodes were analysed using a repeated-measures analysis of variance (ANOVA) employing a between-subject factor of ‘group’ (patient versus control) and a within-subject factor of electrode ‘site’. To examine the time course of the effect, data from a single electrode (at the position Cz), chosen on the basis of previously observed effects of buspirone at this site (McAllister-Williams & Massey, Reference McAllister-Williams and Massey2003), were analysed in a similar way with ‘time’ (baseline and t = 30, 60 and 90 min) as an additional within-subject factor. The plasma concentrations of GH, prolactin and cortisol were each analysed using ANOVA with a between-subject factor of ‘group’ and a within-subject factor of ‘time’, as well as calculating AUCs as for the centroid data. VAS data for each of the five symptoms were analysed separately using ANOVAs in a similar way to the endocrine data. Temperature data were analysed using the same AUC method as for the centroid frequency. Salivary cortisol levels were averaged across the 2 days they were obtained. The two resulting measures, and other demographic variables, were compared between groups using an independent-samples t test or the Mann–Whitney U test dependent on whether the data were normally distributed or not (assessed using the Kolmogorov–Smirnov test). Correlational analysis was conducted using Spearman's correlation to avoid assumptions about normality of distribution of data. All ANOVA analyses incorporated the Greenhouse–Geisser correction for inhomogeneity of covariance, with F ratios reported with uncorrected degrees of freedom for clarity. All values are reported as mean values with their standard errors (s.e.m.).
Results
Participants
Table 1 lists the demographic features and depression rating scale scores of patients and controls. The two groups of subjects were matched for age and IQ (p > 0.5). In addition to there being equal numbers of men and women in each group, there were also equal numbers of pre-menopausal women (two in each group) who were matched for stage of menstrual cycle. There were more smokers amongst patients (five compared with one in the control group) but equal numbers who reported a history of recreational drug use (two in each group). Alcohol usage was matched (p > 0.25). Patients scored significantly higher on mood rating scales than the healthy controls (see Table 1). The average duration of current episode of depression for patients was 24 months (range 1 month to 10 years) and the average number of previous depressive episodes was 3.9 (range 1–10 episodes). In the patient group, four were psychotropic drug naive and the other 11 had had antidepressants in the past, with the mean period since discontinuation being 13 months (range 2–60 months).
Patients and controls demography a
s.e.m., Standard error of the mean; IQ, intelligence quotient; NART, National Adult Reading Test; HAMD, Hamilton Depression Rating Scale; MADRS, Montgomery–Asberg Depression Rating Scale; BDI, Beck Depressive Inventory.
a Demographic characteristics and depression rating scale scores in the depressed patient and control cohorts. The right-hand column reports the p value from two-tailed independent-samples t tests between the two groups.
Effect of buspirone on EEG centroid frequency
As expected, buspirone administration caused a decrease in the centroid frequency. This effect was more evident in patients with depression. Fig. 1 shows the effect of buspirone on the centroid frequency (analysed as AUC) across the 29 scalp electrodes. The mean AUC across all electrodes was −12.97 (s.e.m. = 2.82) and −3.98 (s.e.m. = 2.07) Hz.min for patients and controls, respectively. ANOVA of data from all 29 scalp electrodes revealed a significant difference between patients and controls (F 1,28 = 6.55, p = 0.016). Subsidiary two-tailed independent-samples t tests revealed significant (p < 0.05) group differences at 21 out of 29 electrodes (see Fig. 1), most consistently at fronto-central electrode sites. No group × electrode site interaction was found on ANOVA. The time course of the effect of buspirone on the EEG is illustrated in Fig. 2 which shows data from the Cz electrode. ANOVA of these data revealed a group × time interaction (F 3,84 = 3.38, p = 0.038).
Topography of the effect of buspirone (30 mg) on the electroencephalographic (EEG) frequency spectrum. Area under the curve analysis of the shift in EEG centroid frequency compared with baseline was computed for both patients (■) and controls (
) for all 29 scalp electrode sites. Results are shown as if looking down on the scalp from above. Data are means, with standard errors of the mean represented by vertical bars. * Electrode sites where there was a significant difference between patients and controls (p < 0.05, two-tailed, non-paired t test).
Time course of the shift in electroencephalographic centroid frequency with buspirone (30 mg). Data were recorded at the electrode at the Cz position from patients and controls before (time 0) and after administration of buspirone. The data plotted are the estimated marginal means and standard errors of the mean calculated in SPSS from an analysis of variance of the centroid data with baseline as a covariate as recommended for the analysis of drug effects (Vickers & Altman, Reference Vickers and Altman2001).
Given that AUC values were calculated relative to baseline centroid values, this was compared across the two groups. The mean baseline centroid frequency across all electrode sites was higher for patients [8.64 (s.e.m. = 0.11) Hz] than for the controls [8.38 (s.e.m. = 0.08) Hz], though an ANOVA for all 29 electrode sites did not reveal a significant group difference (F 1,28 = 3.57, p = 0.069). Post-hoc two-tailed independent-samples t tests did, however, reveal significant (p < 0.05) group differences at five out of 29 electrodes (positions Pz, P4, CP4, O1 and O2) with a posterior scalp distribution. As a result of these marginal baseline differences, the ANOVA of the centroid AUC values from all 29 electrode sites was re-run with baseline centroid value as a covariate. A significant patient versus control group effect remained (F 1,28 = 4.86, p = 0.036). Consistent with the group effect remaining significant when covarying for baseline centroid values, there were no significant correlations between individual electrode baseline and AUC values either for the full sample or the patients and controls analysed separately.
Neuroendocrine and temperature response to buspirone
Fig. 3 shows the GH, prolactin and cortisol responses to buspirone in the patients and controls. In contrast to the effects of buspirone on the EEG centroid frequency, there was little difference in the endocrine responses between patients and controls, with, if anything, the response of patients being numerically less than controls. However, ANOVA revealed no group or group × time interaction for GH, prolactin or cortisol (all p's > 0.3). There were no significant correlations between the effect of buspirone on the EEG centroid frequency and its effect on neuroendocrine levels for any of the three hormones, consistent with our hypothesis that the effect of buspirone on the EEG centroid frequency is mediated by activation of somatodendritic 5-HT1A receptors, while the neuroendocrine responses are due to postsynaptic 5-HT1A receptor activation. This was the case both with the raw neuroendocrine AUC data and with more normally distributed reciprocal transformed data (all p's > 0.2).
Neuroendocrine effects of buspirone administration (30 mg). (a) Growth hormone (GH), (b) prolactin and (c) cortisol plasma levels in patients and controls before (time 0) and following administration of buspirone. Data are means, with standard errors of the mean represented by vertical bars.
There were technical problems with a number of temperature measures made, resulting in temperature AUC data being available for 13 patients and eight controls. Mean AUC showed a numerically greater hypothermia in the controls [−12.6 (s.e.m. = 40.9) v. 0.4 (s.e.m. = 9.1) °C/min], but there was no significant difference when analysed by t test (p = 0.4).
Effect of potential confounds on the EEG response to buspirone
To compare the effect of previous psychotropic drug exposure, drug-naive patients were compared with those with previous exposure [four v. 11 in each group, respectively; centroid AUC values: −10.3 (s.e.m. = 7.8) v. −13.7 (s.e.m. = 4.1) Hz.min]. Also, those with exposure more than 12 months previously (including those who were drug naive) were compared with those with more recent exposure [seven v. eight in each group, respectively; centroid AUC values: −11.1 (s.e.m. = 5.0) v. −14.3 (s.e.m. = 2.8) Hz.min]. In both cases there was no significant difference between groups (p > 0.5 in both cases).
With regard to illness history and mood symptoms, there was no significant correlation between the AUC for the effect of buspirone at the Cz electrode for the patients and the number of previous episodes or duration of current episode, nor within either the patient-only data, or both patients and controls, with HAMD, MADRS or BDI.
There was no significant difference between patients and controls in salivary cortisol at 08.00 hours [16.2 (s.e.m. = 3.8) v. 14.0 (s.e.m. = 2.1) nmol/l, Mann–Whitney U test, p > 0.5]. However, a trend was seen for the 20.00 hours cortisol level to be higher in patients compared with controls [2.6 (s.e.m. = 2.1) v. 1.8 (s.e.m. = 0.7) nmol/l, Mann–Whitney U test, p = 0.093]. While no effect was seen in the patient-only sample, there was a significant correlation in the combined population between the AUC for the effect of buspirone on the EEG with the 20.00 hours salivary cortisol level (r = 0.40, p = 0.031). The 20.00 hours salivary cortisol level also significantly correlated with all three measures of mood across the combined population (HAMD: r = 0.38, p = 0.043; MADRS: r = 0.38, p = 0.042; BDI: r = 0.51, p = 0.005).
VAS data were available for 14 subjects in each group. Data for patients at baseline were significantly higher for patients compared with controls for ‘depression’, ‘drowsiness’, ‘restlessness’ and ‘light-headedness’ and were just short of significant for ‘nausea’ (see Table 2). ANOVA of the data from the five VAS measures revealed a significant ‘patient versus control’ effect for all scales, with the exception of ‘nausea’, due to the patients rating all symptoms higher than controls at all time points (see Table 2). There was a significant effect of ‘time’ for the VASs relating to ‘drowsiness’ (F 1,26 = 4.3, p = 0.015), ‘nausea’ (F 1,26 = 10.3, p = 0.001) and ‘light-headedness’ (F 1,26 = 11.8, p < 0.001), consistent with buspirone administration being associated with an increase in all of these symptoms. There was a significant ‘patient versus control’ × time interaction for ‘light-headedness’, with this interaction being just short of significance for ‘drowsiness’ and ‘nausea’ due to the effects of buspirone being greater in the patients. However, there were no significant correlations between the AUC for the effect of buspirone on the EEG and any of the VAS measures in either the combined population, or patient-only group.
Visual analogue scores for patients and controls a
Data are given as mean (standard error of the mean).
t = 0, Baseline, prior to oral buspirone (30 mg) administration; t = 30, 30 min after buspirone administration; t = 60, 60 min after buspirone administration; t = 90, 90 min after buspirone administration; df, degrees of freedom; ANOVA, analysis of variance.
a The Table shows the visual analogue scores in mm at all time points for the patient and control groups. The Table also shows the results of t tests on the baseline data comparing the two groups and the ANOVA of all time points, firstly a main effect of group and then a group × time interaction.
Discussion
This is the first study to utilize the described EEG method for the examination of putative somatodendritic 5-HT1A receptor function in a population of medication-free depressed patients. The findings suggest an increase in receptor function in depressed patients compared with a cohort of well-matched controls. This is consistent with some PET (Parsey et al. Reference Parsey, Oquendo, Ogden, Olvet, Simpson, Huang, Van Heertum, Arango and Mann2006, Reference Parsey, Ogden, Miller, Tin, Hesselgrave, Goldstein, Mikhno, Milak, Zanderigo, Sullivan, Oquendo and Mann2010) and post-mortem (Stockmeier et al. Reference Stockmeier, Shapiro, Dilley, Kolli, Friedman and Rajkowska1998) data and an hypothesis of a transcriptional dysregulation of 5-HT1A receptors in depression with increased expression in raphé serotonergic neurones (Albert et al. Reference Albert, Le Francois and Millar2011), leading to an overall impairment of 5-HT neurotransmission in depression. The hypothesized effects of antidepressants, particularly SSRIs, in down-regulating somatodendritic 5-HT1A receptors (Blier & de Montigny, Reference Blier and de Montigny1994) would directly counter the difference between patients and controls that we have seen here.
It has been suggested that discrepancies in the PET studies of 5-HT1A receptor binding might arise due to differences in the medication status of patients (Hirvonen et al. Reference Hirvonen, Karlsson, Kajander, Lepola, Markkula, Rasi-Hakala, Nagren, Salminen and Hietala2008; Savitz et al. Reference Savitz, Lucki and Drevets2009; Parsey et al. Reference Parsey, Ogden, Miller, Tin, Hesselgrave, Goldstein, Mikhno, Milak, Zanderigo, Sullivan, Oquendo and Mann2010). The present study has a strength in that all patients had been free of psychotropic medication for at least 2 months. A comparison between those who were drug naive versus those with previous exposure, and between those with exposure more than 12 months previously versus more recent exposure, both found no difference in somatodendritic 5-HT1A receptor function. However, the sample sizes were small and this issue requires further exploration.
The current study examined the HPA axis status of the included patients and controls. There was a trend for a higher 20:00 hours salivary cortisol level in patients compared with controls, which is in the expected direction given the frequently described hypercortisolaemia, particularly in regards to trough levels, seen in depression (McAllister-Williams et al. Reference McAllister-Williams, Ferrier and Young1998). The finding of a positive correlation between somatodendritic 5-HT1A receptor function and 20.00 hours salivary cortisol levels across the combined patient/control population is of interest. It is at odds with our previous demonstration that repeated administration of cortisol to healthy subjects attenuates the effect of buspirone on the EEG (McAllister-Williams et al. Reference McAllister-Williams, Massey and Fairchild2007). However, the 20:00 hours salivary cortisol level correlated with all three measures of mood used and these differed in a dichotomous way between the patients and controls. So while there was significant overlap in both the EEG measure and the salivary cortisol levels between patients and controls, the relationship between these two could have simply been driven by the group difference in somatodendritic 5-HT1A receptor function and higher cortisol being related to low mood in the depressed patients.
There was no significant difference between the depressed patients and controls in either the neuroendocrine or hypothermic response to buspirone, albeit with a rather small dataset for the latter. This is consistent with previous findings examining the corticotrophin (ACTH), cortisol and prolactin responses to buspirone in depression (Meltzer & Maes, Reference Meltzer and Maes1994; Navinés et al. Reference Navinés, Gómez-Gil, Martín-Santos, de Osaba, Escolar and Gastó2007).
While there is no definitive means of confirming that the effect of buspirone on the EEG centroid frequency results from somatodendritic 5-HT1A receptor activation, previous pharmacological characterization of the effect supports this contention (McAllister-Williams & Massey, Reference McAllister-Williams and Massey2003). The current study also provides circumstantial support for this notion in that not only was there no significant correlation between the effects of buspirone on the EEG centroid frequency and its neuroendocrine and hypothermic effects, if anything, the relative effect of buspirone on these measures was in the opposite direction in the patients versus controls. Further, the time course of the effect on the centroid frequency appeared somewhat different compared with that on the neuroendocrine responses. Given that the GH, prolactin and cortisol responses to buspirone are accepted to result from postsynaptic 5-HT1A receptor activation, this would be consistent with the effect on the centroid frequency being mediated via a different population of 5-HT1A receptors.
There are a number of potential limitations of the current study. The lack of a placebo treatment arm for both groups means that it is not possible to completely rule out a difference in the change in EEG centroid frequency simply over time between patients and controls. However, this seems an unlikely explanation for the differences seen. Overall there was no baseline centroid difference between groups, though a difference was seen at five electrode sites. However, these had a difference topography (being posteriorly distributed) compared with the scalp regions where the effect of buspirone was most prominent (fronto-centrally), in both the current study and in previous reports (McAllister-Williams & Massey, Reference McAllister-Williams and Massey2003). Further, the patient versus control difference in the change in centroid frequency as assessed by the AUC measure remained significant when covarying for baseline centroid values. With regard to changes in centroid values over time, we have previously reported that following placebo treatment there is no significant change over the course of 90 min, albeit in healthy subjects (McAllister-Williams & Massey, Reference McAllister-Williams and Massey2003). We cannot rule out the possibility that the situation would be different in depressed patients. However, observation of the data in Fig. 2 shows that the main separation between groups occurs within the first 30 min of administration of buspirone. It seems unlikely that this sharp separation between the groups could occur over such a short period of time simply as a result of differences in diurnal variation in centroid frequency between patients and controls.
Blood levels of buspirone were not obtained and so it is not possible to exclude the possibility that there is a pharmacokinetic difference between patients and controls. We are not aware of any such difference being reported in any previous studies.
In conclusion, this study provided the first in vivo evidence that somatodendritic 5-HT1A autoreceptors may be hyperfunctional in depressed patients. This finding may represent a central abnormality that contributes to serotonergic hypofunction in at least some patients with depressive illnesses.
Acknowledgements
We thank Professor Nicol Ferrier and Dr Sasha Gartside for helpful comments on the manuscript and all of the patients and controls who contributed data to this study. The study was funded via a UK Medical Research Council Clinician Scientist Fellowship award to R.H.McA.-W. (G108/335).
Declaration of Interest
None.
