Introduction

The brain-derived neurotrophic factor (BDNF) is the most prominent member in the neurotrophin family and involved in development and activity-dependent regulation of neuronal structures [1]. Cumulating evidence demonstrated a functional interplay between BDNF and the neurotransmitter serotonin (5-HT), constituting common intracellular signaling pathways and transcription factors, BDNF control over the development and function of serotonergic neurons as well as serotonergic regulation of BDNF gene expression and signaling [2].

Briefly, BDNF is linked with at least three major intracellular signaling cascades: the phosphoinositide-3 kinase pathway enabling cell survival, the phospholipase-gamma pathway effecting synaptic plasticity and the mitogen-activated protein kinase pathway associated with neuronal differentiation and neurite outgrowth [3]. Beside the p75 neurotrophin receptor, which is activated by proBDNF and all other neurotrophins, BDNF releases it’s effects by binding to tropomyosin-kinase related receptor B (TrkB) [4]–[6]. Thereby, BDNF is a major factor in the proper development and plastic regulation of the central nervous system and highly active in limbic structures such as the hippocampus and the amygdala, where long-term potentiation, learning and memory are facilitated [7]. However, it should be stated here that most of the evidence of BDNF in this context is based on rodent data.

The BDNF gene is located at chromosome 11p13-14, including many splice sites and promoters. All BDNF mRNAs are initially translated into proBDNF and are then cleaved into mature BDNF [8]. The most investigated polymorphism of the BDNF gene exists in the codon 66 of proBDNF (Val66Met, rs6265) and consists of a valine to methionine substitution, which is associated with reduced intracellular proBDNF trafficking, synaptic secretion of BDNF, and thus a lower extracellular BDNF concentration in met-allele carriers [9]. Thought to trigger deficits in neuronal development and plasticity, the Val66Met polymorphism is of major interest in neuropsychiatric research [2], [7].

Interestingly, in humans the molecular connections between 5-HT and BDNF, and how alterations in one system affect the other are hardly known. Due to the lack of current methods to measure BDNF, TrkB or p75 in the living human brain, in vivo research in humans mainly focuses on the investigation of alterations of serotonergic structures thought to be mediated via changes in BDNF. In imaging genetics studies, serotonergic markers are labeled by radioligands and their binding is measured using PET. As yet, there exist three studies investigating alterations of BDNF, as represented by the Val66Met polymorphism, and it’s association with binding of 5-HT_1A, 5-HT_2A receptors as well as the 5-HTT in the human brain [10]–[12]. Two previous studies failed to detect links between Val66Met and binding of 5-HT_1A and 5-HT_2A receptors. On the other side, a recently published study reports lower 5-HT_1A binding in healthy subjects carrying the met-allele compared to val-homozygotes, a difference which was not observed in depressed subjects [12]. As far as 5-HTT is concerned, in one study, applying the serotonin transporter (5-HTT) specific radioligand [¹¹C]-MADAM (N = 25) with PET and [¹²³I]-ß-CIT (N = 18) with single photon emission tomography (SPECT) in two independent samples, the authors found increased 5-HTT binding in val-homozygote male subjects and compared to met-allele carriers [10]. On the other hand applying the radioligand [¹¹C]DASB (N = 49), the second study failed to detect any effect of Val66Met genotype status on 5-HTT binding [11].

To resolve contradictory results we conducted an imaging genetics study investigating the association between 5-HTT binding using PET with the radioligand [¹¹C]DASB and the Val66Met genotype status in healthy subjects as well as in depressive patients. We also measured 5-HT_1A receptor binding in healthy subjects genotyped for Val66Met, in order to resolve two equivocal findings. We hypothesized, that Val66Met impacts on 5-HTT binding in patients with major depression and healthy subjects. Furthermore, we hypothesized that significant differences are detected between BDNF genotype status and 5-HT_1A binding in healthy subjects.

Methods

Subjects

In a neuroimaging genetics study with a cross-sectional design in total 92 subjects, aged 18–65 years were included. The study was divided into two groups, in the first one 51 healthy adult volunteers (37 female) were included and measured with [carbonyl-¹¹C]WAY-10063. In the second group 25 healthy subjects (HS) and 16 currently depressed patients with an Hamilton Depression Rating Scale ≥16 (HAMD: 19.7±3.5, mean ± SD) were included (for further details see table 1) and measured with [¹¹C]DASB. None of the subjects received both radioligands. The study population originates from a pooled sample, which is part of previously published studies [13]–[16]. Genotyping data of BNDF were previously not published. All subjects underwent a psychiatric screening by the help of the complete Structured Clinical Interview for DSM-IV type disorders (SCID I+II), physical and neurological examination, clinical history, ECG, routine laboratory analysis, urinary drug and pregnancy screening. All subjects were at least three months free of any psychotropic medication. Every study subject was enrolled in study participation after detailed oral information about all study procedures and subsequent signing of a written informed consent form. The study and all study related procedures were approved by the Ethics Committee of the Medical University of Vienna.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Demographic variables of the entire study sample.

https://doi.org/10.1371/journal.pone.0106810.t001

Statistical Analysis

For normally distributed demographic variables and clinical measures student’s t-tests, for nominal variables chi-squared tests were performed. Significance was determined as p<0.05 and all tests were two-sided.

Differences of 5-HT_1A and 5-HTT BP_ND between BDNF Val66Met genotype groups were calculated using a voxel-wise and a ROI-based approach. For the voxel-wise analysis both in the 5-HTT and the 5-HT_1A – groups an ANOVA was performed as implemented in SPM8. Grouped genotype status (val/val, vs. met-carrier  =  GG vs. A-carrier) served as factor and radioligand specific activity, sex and age served as covariates. In the 5-HTT-collective diagnosis was added as additional factor in a second step analysis. F-tests and group-wise post-hoc t-tests between genotype groups were calculated and contrasted in SPM8. Additionally, in the 5-HTT-group an interaction between diagnosis and genotype status was contrasted by weighting contrast vectors in SPM according to group size. An absolute image threshold was set at 0.1 BP_ND to remove voxels with low signal-to-noise ratio and a cluster threshold was set at 50 voxels. A statistical level of p<0.05 corrected for multiple comparisons by the family-wise error rate (FWE) at voxel-level was considered significant, for subsequent explorative analysis an uncorrected threshold of p<0.001 was accepted.

In the ROI-based analyses differences between genotypes groups (val/val vs. met carrier) were calculated with a linear mixed model in SPSS 19 (IBM Corp. Released 2010. IBM SPSS Statistics for Windows, Version 19.0. Armonk, NY: IBM Corp.). Thereby, subject served as the random effect and BDNF genotype status, region, sex and age served as fixed effects. Ten representative regions were chosen due to their a priori known high distribution of 5-HT_1A receptors and 5-HTT and implications in psychiatric disorders (see tables 1,2 and figures 1,2). Diagnosis was taken as additional factor in the 5-HTT-study collective. Significance was determined as p<0.05. Post-hoc t-tests were conducted two-sided in 10 AAL ROIs (see tables 1,2 and figures 1,2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Bar chart plotting serotonin-1A binding potential (5-HT_1A BP_ND) according to BNDF Val66Met genotype status.

Values at the y-axis represent 5-HT_1A BP_ND separated for val/val and met-carrier, respectively, x-axis shows regions of interest. Regions and values correspond to table 2. ACC: anterior cingulate cortex, AMY: amygdala, MCC: medial cingulate cortex, HIPP: hippocampus, INS: insula, paraHIPP: parahippocampus, PCC: posterior cingulate cortex, TempPole: temporal pole, DRN: dorsal raphe nucleus.

https://doi.org/10.1371/journal.pone.0106810.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Bar chart plotting serotonin transporter binding potential (5-HTT BP_ND) according to BNDF Val66Met genotype status.

Values at the y-axis represent 5-HTT BP_ND in pooled healthy subjects and depressive patients. Binding potential is separated for val/val and met-carriers, respectively, x-axis shows regions of interest. Because healthy subjects and depressive patients were pooled here, regions do, but values do not correspond to table 3. ACC: anterior cingulate cortex, AMY: amygdala, MCC: medial cingulate cortex, HIPP: hippocampus, CAUD: caudatum, PUT: putamen, THAL: thalamus, STRIA: striatum, MID: Midbrain, NACC: nucleus accumbens.

https://doi.org/10.1371/journal.pone.0106810.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Post-hoc t-tests comparing serotonin-1A receptor (5-HT_1A) binding potential (BP_ND) according to BDNF Val66Met genotype status in 51 healthy subjects.

https://doi.org/10.1371/journal.pone.0106810.t002

Results

Out of the 51 HS in the 5-HT_1A-group 30 carried GG, 18 carried AG and 3 AA. The 5-HTT-group had 25 HS with 19 carrying GG, 5 carried AG and 1 AA, whereas in the MDD group with 16 depressed patients 13 carried GG, 3 carried AG and 0 the AA allele (table 1). Allele frequencies of the BDNF gene in all study groups were distributed in accordance with the Hardy-Weinberg equilibrium [5-HT_1A-group: X² = 0.02, p = 0.891, 5-HTT-group HS: X² = 0.72 p = 0.4, MDD patients X² = 0.17 p = 0.68). The AA and AG+GG study groups did not differ in demographical, clinical measures or radiopharmaceutical measures (table 1). The allelic distribution was not associated with diagnosis in the 5-HTT-group (X² = 0.157, p = 0.692).

In the voxel-wise analysis there was no significant association of BDNF genotype (GG vs. A-carrier) status with 5-HT_1A BP_ND (F-test: all p>0.05 FWE corr. and all p>0.001 uncorr.). Furthermore, there was no significant association of BDNF genotype (GG vs. A-carrier) with 5-HTT BP_ND (F-test: all p>0.05 FWE corr. and all p>0.001 uncorr.). There was no interaction between BDNF genotype status, diagnosis or sex and 5-HTT BP_ND (t-test: all p>0.05 FWE corr. and all p>0.001 uncorr).

The mixed model analyses of ROIs in the 5-HT_1A-group, controlling for potential effects of sex, age and specific radioligand activity, yielded no significant difference of 5-HT_1A BP_ND in selected ROIs between GG homozygotes and A-allele carriers (F = 0.342, df = 1,45, p = 0.562). In the 5-HTT-group, the mixed model revealed no significant difference between 5-HTT BP_ND in the selected ROIs between GG homozygotes and A-allele carriers (F = 0,41, df = 1,33, p = 0.526). There was no interaction between diagnosis and allele in the statistical model (p = 0.989). Post-hoc t-tests and average BP_ND values for both study groups are shown in table 2 and table 3, BP_ND-values of allele groups are displayed in figure 1 and figure 2. Here, in the 5-HTT-group, a significant difference between GG and A-carriers was observed in HS in the midbrain (p = 0.040, uncorr., table 3) as well as between GG in HS and GG in MDD patients (p = 0.034, uncorr.), with BP_ND increases in GG-carriers, respectively. All other post-hoc tests (5-HT_1A: GG vs. A-carrier; 5-HTT HS: GG vs. A-carrier, MDD GG vs. A-carrier, HS vs. MDD GG, HS males GG vs. HS males A-carrier) did not yield significant results (all p>0.05 uncorr.).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Post-hoc t-tests comparing serotonin transporter (5-HTT) binding potential (BP_ND) according to BDNF Val66Met genotype status in 25 healthy subjects and 16 depressed patients.

https://doi.org/10.1371/journal.pone.0106810.t003

Discussion

In a voxel-wise analysis as well as in a ROI-based approach, we did not observe significant differences of 5-HT_1A-receptor BP_ND nor of 5-HTT BP_ND according to BDNF genotype status. There was no interaction between MDD diagnosis or sex and 5-HTT BP_ND. In the midbrain, weak increases of 5-HTT-BP_ND in healthy subjects between val-homozygotes and met-carriers were found. Furthermore, weak increases of 5-HTT BP_ND were observed in the midbrain in val-homozygote healthy subjects compared to val-homozygote MDD patients. There was no association between allelic distribution and major depression. To sum up, all voxel-wise and ROI-based testing yielded negative results and none of the post-hoc tests survived correction.

Our results are in concordance with a previous PET study applying [¹¹C]DASB in 49 healthy subjects, where the authors neither detected differences in 5-HTT binding in relation to BDNF genotype nor a correlation between blood BDNF levels and central 5-HTT binding [11]. Additionally, no effect on 5-HT_2A binding was shown in this work. Here, the authors calculated the radiotracer BP_ND similar to our study by applying a fully automated reference region model (MRTM2) [22] and an automated ROI-delineation. The only other currently published human PET-study investigating the impact of BDNF polymorphisms on 5-HTT binding reports differences in men and shows no effect of genotype status on 5-HT_1A binding [10]. Men homozygous for the val-allele exhibited significantly higher 5-HTT binding in regions such as the hippocampus, insula or dorsal raphe compared to met-carrier, while this effect was absent in women. Furthermore, reductions of 5-HTT binding in met-carrier (n = 3) compared to val-homozygotes (n = 6) in an independent [¹²³I]-ß-CIT-study with male suicide attempters were demonstrated, but this reduction was absent when pooled with healthy controls. The authors also used a reference region model with [¹¹C]-MADAM, a tracer exhibiting a comparable 5-HTT affinity to [¹¹C]DASB [27], the ROIs were manually delineated on individual magnetic resonance images (MRI). Notably, our group previously reported strong correlations of BP_ND values between automatically and manually delineated ROIs [23]. The radioligand and the method of ROI generation are on these grounds an unlikely source of variance leading to alternative results. Importantly, in search of arguments for this difference, one must mention that the number of male met-carriers in that collective was low (n = 4), which makes the analysis vulnerable to outliers and hence may increase type-I errors. Likewise, our study exhibits a subgroup with a low subject number and indeed we saw an outlier in the MDD met-carrier group (n = 3) when we plotted the individual BP_ND values (data not shown). Hence, our results in depressed patients have to be interpreted with caution. But the fact that both the study by Klein et al., which exhibits a large sample size of healthy volunteers, as well as our study did not reproduce higher 5-HTT binding in val-homozygote healthy subjects, rather speaks for an absent effect of BDNF Val66Met on 5-HTT binding.

Apart from this, our study agrees with the data by Henningsson et al., on an absent effect of Val66Met on 5-HT_1A receptor binding in healthy subjects [10]. Both studies apply the same radioligand, i.e. [carbonyl-¹¹C]WAY-100635, exhibit an almost identical number of subjects (n = 53 in Henningsson et al.), and modeled 5-HT_1A binding by a reference region model (BP_ND). These results are in contradiction to a recent finding reporting 5-HT_1A reductions in healthy met-allele carriers [12], which is not present in MDD patients. In this study 50 healthy subjects and 50 MDD patients were measured with the radioligand [carbonyl-¹¹C]WAY-100635, yet 5-HT_1A binding was calculated by an arterial input function (BP_F). Most interestingly, when the authors repeated their analysis with BP_ND values, the reduction of 5-HT_1A binding in healthy met-carriers was not detectable, suggesting that this finding was associated with the method of radioligand modeling. Following the discussion of the authors, one cannot rule out that Val66Met causes differences of radioligand binding in the blood leading to a bias in the arterial input function. Although, our results are in agreement with all previous studies on 5-HT_1A binding using reference tissue models [10], [12], validation by a different tracer not susceptible to modeling methodology is further needed. Taken together, while there are currently contradicting findings on the in vivo effect of BDNF Val66Met genotypes on 5-HTT binding [10], [11], this study adds data emphasizing the absence of such an effect. Moreover, this work corroborates previous results by reference tissue models demonstrating no association between BDNF Val66Met genotype status and 5-HT_1A receptor binding [10], [12] and is in contradiction with a study reporting binding values modeled with arterial blood sampling [12].

Preclinical data report that BDNF promotes development and function of serotonergic neurons by enhancing survival and differentiation [28], increasing local 5-HT [29] modifying the firing pattern of serotonergic raphe neurons [28], [30] and altering the function of serotonergic receptors such as the 5-HT_1A and 5-HT_2A receptors and the 5-HTT [2], [29], [31]. Vice-versa, raised extracellular 5-HT levels occurring upon administration of SSRIs are thought to increase local BDNF levels by enhanced phosphorylation of serotonergic receptor coupled cAMP response element-binding (CREB) protein [32]–[34], a common target of BDNF and G protein-coupled serotonergic receptors [2]. Confronted with this evidence, one is puzzled upon the lack of strong evidence for an association between BDNF and serotonergic structures in humans in vivo. However, preclinical studies are not consistent and negative results regarding the expression of 5-HT receptors and transporter are reported [31], [35]. Although the interaction between the BNDF and 5-HT provides a promising bridge between structural and functional neuronal activity, and serves as explanatory hypothesis for neuronal plasticity deficits in neuropsychiatric disorders, exact mechanisms underlying the regulation of the cross connection between BDNF and 5-HT in humans remain unresolved [36]. Our data in concert with above referred work speak for a similar expression of 5-HTT and 5-HT_1A receptors upon life-time BDNF reduction, but unfortunately do not illuminate the mechanisms leading to this observation. Theoretically, counter-regulatory or compensatory effects may have altered 5-HTT and 5-HT_1A expression. Furthermore, it is possible that not absolute numbers but functional activity of serotonergic structures is altered by BDNF.

The evidence on connections between depression and BNDF genotype status is inconsistent as well. Meta-analytical research suggested an association of Val66Met with major depressive disorder antidepressant treatment response or hippocampal volume and a role of gender and ethnicity [37]–[39]. However, recent meta-analyses refuted these associations and detected power deficits in many trials [40]–[42]. Low serum levels of BDNF were suggested as potential peripheral marker of depression and increase of serum BDNF as response to the appropriate first-line treatment with selective 5-HT reuptake inhibitors (SSRIs). Likewise, this association is weaker than initially thought and there is no relationship between symptom severity and BDNF serum concentration [43]. Our results suggest no association between allelic distribution and diagnosis. Our small number of MDD subjects remain a limiting factor in that regard.

Limitations

Unfortunately a common problem of human PET studies is weak power resulting from low subject numbers, owed to the large effort of conducting PET-imaging. This is even more intrinsic to genetic PET studies reporting results based on genotype subgroups [44] and in SNP neuroimaging studies where pooling of rare genotype groups is common practice. The low subject number in the MDD met-carrier group could therefore be a limitation of our study. One elegant way to circumvent this issue in future studies would be pooling data between PET centers, which is already common in MRI studies. Second, mean age of genotype groups is heterogeneous, yet controlled for in all statistical analyses. Finally, we did not model PET data with an arterial input function [45], because arterial blood data were not collected. This would have been useful to confirm reported differences according to the methodology for calculating 5-HT_1A binding with [carbonyl-¹¹C]WAY-100635, an issue we are trying to resolve in future studies [46].

Conclusion

Although others have investigated the effects of the BDNF gene on 5-HTT and 5-HT_1A binding with PET, this study adds data to the ongoing discussion about the cross connection between 5-HT and BDNF. While previous work in humans demonstrated contradicting results, due to this work the conclusion of an absent influence of Val66Met on 5-HTT and 5-HT_1A has gained substantial support.

Acknowledgments

The authors are grateful to U. Moser, E. Akimova, P. Stein, M. Fink, C. Spindelegger, A. Höflich, I. Hofer-Irmler, S. Zgud, S. Pichler, A. Kautzky and D. Winkler for medical and administrative support, and M. Savli for technical support. We thank the PET team, especially G. Karanikas, T. Traub-Weidinger, L.-K. Mien, J. Ungersboeck, K. Kletter, L. Nics, and C. Philippe for technical support. Further, we thank the genetics team of D. Rujescu, especially M. Friedl, A. Hartmann, I. Giegling.

The study is part of C. Kraus’ thesis “Serotonin and Neuroplasticity” supervised by R. Lanzenberger in the Clinical Neurosciences PhD program at the Medical University of Vienna, Austria. Parts of this study have been presented by P. Baldinger at the 19^th at the 11^th World Congress of Biological Psychiatry (WFSBP), June 23^rd–27^th, 2013, Kyoto, Japan.