Genetic variation on the BDNF gene is not associated with differences in white matter tracts in healthy humans measured by tract-based spatial statistics


Dr C. Montag, Department of Psychology, University of Bonn, Kaiser-Karl-Ring 9, 53111 Bonn, Germany. E-mail:


The brain-derived neurotrophic factor (BDNF) is a member of the neurotrophin family and involved in nerve growth and survival. It has also become a major research focus in the investigation of both cognitive and affective processes in the human brain in the last years. Especially, a single nucleotide polymorphism on the BDNF gene called BDNF Val66Met gained a lot of attention, because of its effect on activity-dependent BDNF secretion and its link to negative emotionality and impaired memory processes. A well-replicated finding from genetic structural imaging showed that carriers of the less frequent 66Met allele show diminished gray matter volume in several areas of the temporal lobe. New imaging techniques like diffusion tensor imaging now allow investigating the influence of BDNF Val66Met on white matter integrity. We applied tract-based spatial statistics in a brain image dataset including n = 99 healthy participants. No significant differences between the 66Met and homozygous 66Val carriers were observed when correcting for multiple comparisons. In summary, the BDNF Val66Met polymorphism seems not to play a substantial role with respect to the modulation of the white matter integrity in healthy subjects. Although not in the focus of this study, we also investigated the influence of Eysenck's Personality Questionnaire on the white matter tracts. No significant results could be observed.

The brain-derived neurotrophic factor (BDNF) is a protein from the neurotrophin family crucially influencing synaptic plasticity of neurons (Martinowich et al. 2007). It is coded by the gene MIM113505. A prominent genetic variant on the BDNF gene – the BDNF Val66Met polymorphism (rs6265) – is located on human chromosome 11p14.1 and leads to an amino acid exchange from valine to methionine. The 66Met allele has been associated with a diminished activity-dependent BDNF secretion and impaired memory processes (Egan et al. 2003). Recently, a functional magnetic resonance imaging (fMRI) study showed that carriers of the 66Met allele also respond with stronger amygdala activity to emotional stimuli (Montag et al. 2008). Several structural magnetic resonance imaging (MRI) studies showed an influence of BDNF Val66Met, especially on the gray matter volume of the temporal lobe (Montag et al. 2009; Pezawas et al. 2004). To date, no study has investigated the influence of the BDNF Val66Met polymorphism on white matter architecture using diffusion tensor imaging (DTI) in combination with a new powerful statistical tool in genetic imaging: the tract-based spatial statistics (TBSS).

Diffusion tensor imaging is widely used to characterize the white matter architecture of the human brain in vivo, providing information about integrity and organization of fiber tracts. In particular, DTI has been applied as one promising technique for parameterization of the brain's white matter. Several scalar indices have been proposed to correlate with the underlying structure, fractional anisotropy (FA) being the one most commonly used to quantify the directionality of diffusion and thereby structural integrity of the tissue. It is highest in white matter tracts, significantly lower in gray matter and theoretically zero in cerebrospinal fluid, therefore allowing to differentiate between different types of brain matter (Le Bihan et al. 2001; Pierpaoli & Basser 1996). It is preferable to analyze changes of FA on a whole brain basis, to avoid a restriction to a priori defined brain region. Different voxel-based techniques for whole brain analysis of FA differences have been used extensively, voxel-based morphometry (VBM) being the most commonly used approach (Ashburner & Friston 2000; Jones et al. 2005). This method usually requires smoothing and spatial normalization of images (Focke et al. 2008; Smith et al. 2006), which is inappropriate for scalar values as FA. Fractional anisotropy values are of special interest in the context of genetic imaging, because it has been shown by Kochunov et al. (in press) that whole brain average FA values show heritability estimates of 52%. Tract-based spatial statistics is a rather novel and increasingly used VBM-like approach to register and analyze white matter changes in FA datasets, using exclusively non-linear registration techniques. The use of smoothing methods is not required, as FA values are projected on a skeleton of white matter tracts based on the individual subject studied. It has been shown that this method increases the sensitivity and interpretability of white matter tract changes in neurological diseases (Focke et al. 2008; Smith et al. 2006).

Given the impact of BDNF on the structure and functionality of the human brain, the aim of the current study was to test for the effects of BDNF Val66Met on white matter integrity by using TBSS in a large genetic DTI study.



The sample (n = 99) consisted of 45 males (mean age: 30.45, SD = 10.24) and 54 females (mean age: 28.15, SD = 11.60). The participants were recruited in Cologne and Bonn via telephone advertisement and in psychology classes at the Department of Psychology at the University of Bonn, Germany. All of the participants provided buccal cells for genotyping of the BDNF Val66Met polymorphism. The participants filled in self-report questionnaires assessing life-time history of any neurological/psychopathological disorder and were excluded in case of any present or past occurring disorder. Furthermore, we administered the Beck Depression Inventory (BDI). The mean score of the total sample yielded unobtrusive BDI scores (M = 6.19, SD = 5.72) representing a non-clinical range. The BDI score of five participants was missing. In general, clinical structural interviews would have been a better approach to check on psychopathological/neurological disorders. Unfortunately, we were not able to conduct these interviews, because of lacking resources. As an addition to the BDI, the participants filled in Eysenck's Personality Questionnaire-Revised (EPQ-R; Eysenck & Eysenck 1991). The EPQ-R consists of 102 items with a dichotomous response format. It measures the personality dimensions: neuroticism, extraversion and psychoticism. Furthermore, it is possible to build a scale measuring social desirability. The data of the EPQ-R were gathered for other research purposes, but here we were also able to investigate the influence of personality dimensions on white matter tracts. As personality variables had no influence on the FA values and are not in the focus of this study, they are not discussed later on. Nevertheless, we report the non-significant results in the Results section for future studies. The study was approved by the local medical ethics committee at the University of Bonn. Participants gave written consent to participate in the study.


DNA was extracted from buccal cells. Automated purification of genomic DNA was conducted by means of the MagNA Pure® LC system using a commercial extraction kit (MagNA Pure LC DNA isolation kit; Roche Diagnostics, Mannheim, Germany). Genotyping of the BDNF Val66Met polymorphism was performed by real-time polymerase chain reaction (PCR) using fluorescence melting curve detection analysis by means of the Light Cycler System 1.5 (Roche Diagnostics).

The primers and hybridization probes (TIB MOLBIOL, Berlin, Germany) for BDNF Val66Met are as follows: forward primer, 5′-ACTCTGGAGAGCGTGAATGG-3′; reverse primer, 5′-CCAAAGGCACT TGACTACTGA-3′; anchor hybridization probe, 5′-LC640-CGAACACAT GATAGAAGAGCTGTT-phosphate-3′; sensor hybridization probe, 5′-AAGAGGCTTGACATCATTGGCTGACACT-fluorescein-3′.

Statistical analyses

Owing to the skew distribution of the BDNF Val66Met polymorphism in the Caucasian population, a 66Met+ group consisting of Val66Met and Met66Met carriers was built to be tested vs. the homozygous Val66 variant called 66Met− group.

Imaging protocol

Magnetic resonance imaging was performed at the Life&Brain Center in Bonn on a 3-Tesla scanner (Magnetom Trio; Siemens, Erlangen, Germany). A neurovascular eight-channel head coil was used for signal reception. All subjects underwent the same imaging protocol consisting of whole brain T1-weighted, T2-weighted and diffusion-weighted structural imaging. The total study time was approximately 110 min per subject and all images were obtained in one session.

Diffusion-weighted data were obtained using an in-house DTI sequence. Images were acquired using single-shot, dual-echo, spin-echo echo planar imaging (EPI; repetition time TR = 12seconds, echo time TE = 100milliseconds, 72 axial slices, resolution 1.72 × 1.72 × 1.7 mm, no cardiac gating). As parallel imaging scheme, a Generalized Autocalibrating Partially Parallel Acquisition (GRAPPA) technique (acceleration factor of 2.0) was chosen. Diffusion weighting was isotropically distributed along 60 directions (b-value = 1000 s/mm2). In addition, seven datasets with no diffusion weighting (b0, b-value = 0 s/mm2) were acquired initially and interleaved after each block of 10 diffusion-weighted images as anatomical reference for motion correction. The high angular resolution of the diffusion weighting directions improves the robustness of probability density estimation by increasing the signal-to-noise ratio (SNR) and reducing directional bias. To further increase SNR, scanning was repeated three times for averaging.

Directly after the DTI, a three-dimensional (3D)-T2-weighted dataset with 192 slices was obtained (RARE; TR = 2seconds, TE = 355 milliseconds, resolution 1.0 × 1.0 × 1.0 mm, flip angle 180°). T1-weighted images were obtained using an MP-RAGE sequence with 160 slices (TR = 1300milliseconds, TI = 650milliseconds, TE = 3.97 milliseconds, resolution 1.0 × 1.0 × 1.0 mm, flip angle 10°).

Data analysis

All imaging data were transferred to a cluster of Linux workstations for processing. The structural images were visually inspected for any structural abnormalities by a board-certified neurologist. Preprocessing and analysis of diffusion data were performed with an in-house protocol using FSL 4.1 tools (available at First, the DICOM images were converted into a 4D Nifti-file. Motion correction was then applied on all images using 7-parameter global rescale registration with a mutual information cost function and trilinear interpolation as implemented in FLIRT (FMRIB Linear Image Registration Tool; Jenkinson & Smith 2001). All baseline b0 images were aligned to a reference b0 image and the resulting linear transformation matrixes were then applied to the diffusion-weighted images following each baseline b0 image. After correction for eddy currents, the three repetitions were averaged to improve the SNR. A binary mask differentiating between brain and skull structures was calculated for brain extraction using Brain Extraction Tool (BET) (Smith 2002) and applied to all images. Then FA, mean diffusivity (MD), as well as the eigenvalues of the diffusion tensor were generated using the DTI fit algorithm (Smith et al. 2004).

For voxel-wise analysis of FA, we applied the TBSS tool , i.e. also included in FSL (Smith et al. 2006, 2007). Tract-based spatial statistics is a novel registration approach that has advantages over conventional VBM-like analysis of FA data with regard to partial volume effects and multiple comparison problems (Smith et al. 2006), leading to a higher sensitivity for identifying white matter changes (Focke et al. 2008). All FA maps were aligned to the 1 × 1 × 1 mm FMRIB58 standard space FA template using non-linear registration (Rueckert et al. 1999). By averaging of the individual FA maps, a mean FA image was generated. A skeleton representing the major tracts was then derived from the FA maps and visually inspected, to determine a suitable threshold (a threshold of 0.2 has been used). The final thresholded FA skeleton for each subject (contained in a 4D Nifti-image) was calculated and used to carry out voxel-based statistics. FSLmath was used to calculate parallel diffusion values. Registration and skeletonizing of MD, axial and parallel diffusion maps was performed accordingly using the TBSS non-FA tool. Group analysis was carried out using FSL Randomise with 5000 permutations (Nichols & Holmes 2002). It included group comparisons using two-sample t-tests between the respective genotype groups. The group analysis was complemented by regression analysis with regard to different personality and depression scores using results of the EPQ-R and BDI (see above). Analysis was run on all tracts of the white matter skeleton first and then complemented by region of interest (ROI) analysis focusing on the frontal and temporal lobes, respectively. Identical parameters were used for all analyses. Owing to the lower number of voxels in the ROI analysis, the statistical power is higher in the ROI analyses than in the whole brain analysis. To avoid artifacts caused by inappropriate thresholds, the threshold-free cluster enhancement (TFCE) method (Smith & Nichols 2009) was used for all statistical analysis. Threshold-free cluster enhancement aims to keep the sensitivity benefits of cluster-based thresholding while at the same time minimizing errors due to arbitrarily chosen thresholds for the formation of clusters and for spatial smoothing. It uses a raw statistic image and produces an output image in which the voxel-wise values represent the amount of cluster-like local spatial support. Family-wise error (FWE) correction on the cluster level with a P-value of <0.05 was applied to all statistical images. The resulting statistical maps (both corrected and uncorrected P-values of the whole brain and ROI approaches) were superimposed on the mean FA image and the group skeleton as well as the MNI152 template supplied by FSL. FSLView and its atlas tools (ICBM-DTI-81 white-matter labels atlas; JHU white-matter tractography atlas) as well as general neuroanatomical handbooks were used to allocate FA changes detected by TBSS to the different anatomical structures in the MNI152 space (Hua et al. 2008; Wakana et al. 2004).



Genotyping for the BDNF Val66Met polymorphism yielded the following genotype frequencies: Val66Val, 57; Val66Met, 38; Met66Met, 4. The genotype distribution is in Hardy–Weinberg equilibrium (dfχ2 = 0.57, n.s.). In accordance to the structural and functional genetic BDNF imaging studies (Montag et al. 2009), we tested carriers of the 66Met+ group (n = 42) vs. carriers of the 66Met− group (n = 57).

Tract-based spatial statistics and the BDNF Val66Met polymorphism

The whole brain analysis showed a number of clusters where the FA in the Met66− is significantly higher than in the Met66+ group before FWE correction. The more extensive changes were noted and matched to the respective structures (Table 1). Applying an ROI analysis in regions known to be affected by BDNF, i.e. the temporal and frontal lobes show additional effects (Table 2; Figs 1 and 2). The analysis of MD data, the first eigenvalues (axial diffusion) and parallel diffusion using TBSS showed no significant differences between the Met66+ and Met66− groups.

Table 1.  Detected FA changes Met+ < Met− (whole brain analysis)
StructureSideUncorrected P-valuesP-value (FWE-corr)
Superior longitudinal fasciculus (sup frontal gyrus)Rt97 = 1.47, P < 0.050.237
Superior longitudinal fasciculus (sup frontal gyrus)Lt97 = 1.28, P < 0.050.215
CingulumLt97 = 1.34, P < 0.050.215
Inferior fronto-occipital fasciculusRt97 = 1.40, P < 0.050.206
Uncinate fasciculusRt97 = 1.52, P < 0.050.254
Inferior fronto-occipital fasciculusLt97 = 1.49, P < 0.050.218
Forceps majorLt97 = 1.46, P < 0.050.256
Inferior longitudinal fasciculusLt97 = 1.49, P < 0.050.234
Table 2.  Detected FA changes Met+ < Met− (temporal and frontal lobe ROI analyses)
 SideUncorrected P-valuesP-value (FWE-corr)
Structure (temporal lobe ROI analysis)   
Superior longitudinal fasciculus (temporal part)Lt97 = 1.64, P < 0.050,248
Inferior longitudinal fasciculusLt97 = 1.53, P < 0.050.322
Parahippocampal gyrusLt97 = 1.38, P < 0.050.340
Structure (frontal lobe ROI analysis)   
Superior longitudinal fasciculus (sup frontal gyrus)Lt97 = 1.36, P < 0.050.284
Superior longitudinal fasciculus (sup frontal gyrus)Rt97 = 1.59, P < 0.050.308
Uncinate fasciculusRt97 = 1.49, P < 0.050.246
Uncinate fasciculusLt97 = 1.20, P < 0.050.274
Inferior fronto-occipital fasciculusRt97 = 1.50, P < 0.050.266
Inferior fronto-occipital fasciculusLt97 = 1.25, P < 0.050.224
Figure 1.

Both (a) and (b) show the region of interest analysis of the temporal lobes in different axial slices with respect to FA values contrasting carriers of the Met66> Met66+ variant (P < 0.05, not FWE corrected, left temporal lobe only). (a) The affected tracts shown are the left inferior longitudinal fasciculus (ILF) and the left inferior fronto-occipital fasciculus (IOF). (b) Shows affections of the ILF and superior longitudinal fasciculus (SLF).

Figure 2.

Both (a) and (b) show the region of interest analysis of the frontal lobes in different axial slices with respect to FA values contrasting carriers of the Met66> Met66+ variant (P < 0.05, not FWE corrected). (a) Shows bilateral affections of the ILF and the uncinate fasciculus (UF). (b) Bilateral affections of the SLF are illustrated.

None of the reported significant FA differences between the 66Met+ and 66Met− groups does survive FWE correction for multiple comparisons.

A regression model with regard to personality measures collected by the EPQ-R and BDI scores as a measurement of mood showed no significant results.


The aim of the present study was to investigate the influence of the BDNF Val66Met polymorphism on the white matter integrity of the human brain. Mounting evidence from functional and structural imaging showed that this single nucleotide polymorphism (SNP) influences cognitive (Egan et al. 2003) and affective neuronal circuits (Montag et al. 2008, 2009). A new way of analyzing DTI datasets in genetic imaging – the TBSS – has never been used before to investigate the influence of BDNF Val66Met on brain connectivity, although it is a more sensitive technique compared with other VBM-based DTI procedures (Smith et al. 2006). Besides the importance to examine the effects of BDNF Val66Met on white matter integrity in the whole brain, especially the temporal lobe region is in the focus of BDNF research. Therefore, the temporal lobe is also of special interest for our BDNF Val66Met TBSS-DTI study: a well-replicated finding in structural imaging is that carriers of the less frequent 66Met+ variant show smaller gray matter volume of structures such as the hippocampus (Bueller et al. 2006; Pezawas et al. 2004), the parahippocampal gyrus and the right amygdala in healthy participants (Montag et al. 2009). This might be of high importance for a better understanding of mood disorders, because smaller gray matter volumes of temporal lobe structures have been associated, e.g. with depression (Frodl et al. 2002; Sheline et al. 2003). The upregulation of BDNF in this particular area of the brain is one of the prominent explanations for antidepressive effects of BDNF through its key role in modulating synaptic plasticity and axonal length (Chen et al. 2006). Therefore, the BDNF Val66Met SNP – influencing activity-dependent BDNF secretion (Egan et al. 2003) – could have an influence not only on gray matter volume but also on white matter integrity.

White et al. (2008) recently reviewed DTI studies in psychiatry research and came to the conclusion that white matter integrity seems to be impaired in a lot of psychiatric disorders, such as schizophrenia or depression, but that the findings are very heterogeneous. In part this might be because of the more insensitive VBM-based DTI techniques used in the last years. Owing to its potential impact of BDNF Val66Met on trait anxiety (Montag et al. accepted) and also depression (Groves 2007), it is of interest to give an account of a TBSS study investigating the influence of depression on the white matter tract architecture. Kieseppäet al. (2009) found that patients with major depression disorder showed a trend toward lower FA values in the left sagittal striatum. Taylor et al. (2004) reported that patients with late-onset depression showed, compared to healthy controls, smaller values in the right superior frontal gyrus white matter but without using TBSS.

In our large sample, no significant effects of the BDNF Val66Met polymorphism on white matter integrity could be detected after FWE correction, although our power is with an n = 99, high enough to detect those effects, compared with other MRI and VBM studies (De Geus et al. 2008). Therefore, in healthy humans, the effect of BDNF Val66Met on white matter integrity seems to be negligible compared to its well-replicated effect on gray matter volume. Stimulated by the growing body of literature showing the influence of BDNF Val66Met on negative emotionality and cognitive processes, we also conducted ROI analyses of the frontal and temporal lobe. Still we could not find an influence of the BDNF Val66Met polymorphism on the FA values or other diffusion properties in both mentioned regions. As we investigated only healthy participants, it is possible that the effects of the BDNF Val66Met polymorphism on white matter integrity become more apparent if the 66Met+ variant covaries with other environmental or genetic factors (e.g. critical life events or psychiatric disorders such as depression). In line with this, Taylor et al. (2008) reported that patients with late-onset depression showed significantly greater white matter hyperintensity volumes, but only when being a 66Met+ carrier.

The next interesting – but very time-consuming – step would be to manually outline the hippocampus as a well-known target of BDNF, in order to search for an influence of BDNF Val66Met on afferent and efferent white matter tracts projecting to and from this part of the brain. Kennedy et al. (2009) also planted such a ‘seed’ for their analysis in the corpus callosum (CC) in order to be able to detect age-related effects of 66Met+ on the FA splenium of the CC. This very narrow approach with a focus on a distinct neuroanatomical target might help to further clarify the role of BDNF Val66Met for white matter integrity in several parts of the brain. Furthermore, the investigation of epistasis effects on white matter tracts will be of high importance. Recently, we reported an epistasis effect of the prominent DRD2 Taq Ia polymorphism and the BDNF Val66Met polymorphism on the anterior cingulate cortex (ACC) in healthy humans (Montag et al. 2010). Carriers of the configuration 66Met+/A1+ (A1+ is associated with a 30–40% reduced D2 receptor density in the striatum) showed the lowest gray matter volume of the ACC. Both afferent and efferent white matter tracts could also be modulated by this epistasis effect.

In conclusion, the fact that we could not detect a significant influence of BDNF Val66Met on white matter integrity in the present large sample of healthy participants makes it very unlikely that this SNP has a strong effect in healthy human subjects. Despite this negative finding, the here described technique of analyzing genetic effects on white matter brain architecture by means of TBSS is a promising field in the neurosciences and basic research in psychiatry.


J.-C.S.-B. was supported by a grant of the ‘Verein fuer Epilepsieforschung Bonn’ and Deutsche Forschungsgemeinschaft (DFG) as part of the German trans-regional research cluster on temporal lobe epilepsy (SFB TR3, project A8).

We thank Peter Trautner for help with image processing and Beate Newport and Silke Schiller for their help with MRI scans and study organization.