Abstract :
[en] Secretion of brain-derived neurotrophic factor (BDNF) is essential for synaptic plasticity in the central nervous system during neurodevelopment [Huang]. A common human non-synonymous SNIP in the BDNF gene (Val66Met, rs6265) decreases activity-dependent BDNF release in neurons transfected with the human A allele (Met-BDNF). We reasoned that the persistent differential activity-dependent BDNF release implied by this polymorphism should also be associated with differences in adult brain structure.
The study population comprised 36 healthy subjects (aged 18-25): 15 (9 male) were identified as carrying the Met allele (“Met carrier” group) and 21 (9 male) were homozygotes for the Val allele (“Val/Val” group). The groups did not vary significantly in IQ, age nor scores for a battery of psychological tests. A high-resolution T1-weighted image (sMRI), 7 unweighted (b=0) and a set of diffusion-weighted (b=1000) images using 61 non-collinear directional gradients were acquired for each subject.
The processing workflow relied on several pieces of software and was developed in Python and Nipype. The sMRIs were segmented using the automated labeling of Freesurfer [Desikan] and further parcellated using the Lausanne2008 atlas into 1015 regions of interest (ROIs) [Cammoun]. DWIs were corrected for image distortions (due to eddy currents) using linear coregistration functions from FSL [Smith]. Fractional anisotropy maps were generated, and a few single-fiber (high FA) voxels were used to estimate the spherical harmonic coefficients (order 8) of the response function from the DWIs [Tournier]. Then orientation distribution functions were obtained at each voxel. Probabilistic tractography was performed throughout the whole brain using seeds from subject-specific white-matter masks and a predefined number of tracts (300,000), see Fig. 1. The tracks were affine-transformed into the subject's structural space with Dipy [Garyfallidis]. Connectome mapping was performed by considering every contact point between each tract and the outlined ROIs (unlike in [Hagmann]): the connectivity matrix was incremented every time a single fiber traversed between any two ROIs.
We trained a Gaussian Process Classifier [Rasmussen] (interfaced by PRoNTo [Schrouff]) on these connectivity matrices. The accuracy and generalization ability of the classification were assessed with a leave-one-subject-out cross-validation procedure. With this linear kernel method weights were also obtained indicating the contribution to the classification output (in favor of either genotypic group) of each edge in the network. The same method was employed to discriminate features related to the subjects' gender and genotype for the ADA gene.
The classifier was able to discriminate between Val/Val and Met carriers with 86.1% balanced accuracy. The predictive value for the Val/Val and Met carrier groups were 94.4% (p=0.001) and 77.8% (p=0.003), respectively. In Fig. 2 the weights obtained by the classifier are visualized as edges in the brain network. For the classifier trained to identify gender or the subjects' ADA genotype, the global accuracy reached 63.9% (n.s.) and 58.3% (n.s.) respectively.
Using high-resolution connectome mapping from normal young healthy human volunteers grouped based on the Met allele of the BNDF gene, we show that the BDNF genotype of an individual can be significantly identified from his structural brain wiring. These differences appear specific to this allele; no such difference could be found for the polymorphism in the ADA gene, or even for gender. We propose that the decreased availability of BDNF leads to deficits in axonal maintenance in Met carriers, and that this produces mesoscale changes in white matter architecture.
Acknowledgements: the FNRS, the ULg, the Queen Elisabeth Medical Foundation, the Léon Fredericq Foundation, the Belgian Inter-University Attraction Program, the Welbio program, and the MCITN in Neurophysics (PITN-GA-2009-238593).
Cammoun L. et al. (2011), ‘Mapping the human connectome at multiple scales with diffusion spectrum MRI’, J Neuroscience Methods, 203:386–397.
Desikan R.S. et al. (2006), ‘An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest’, Neuroimage, 31:968-980.
Hagmann P. et al. (2008), ‘Mapping the structural core of human cerebral cortex’, PLoS Biology, 6:e159
Huang E.J., Reichardt L.F. (2001), ‘Neurotrophins: roles in neuronal development and function’, Annual Review of Neuroscience, 24:677-736.
Garyfallidis E. et al. (2011), ‘Dipy - a novel software library for diffusion MR and tractography’, 17th Annual Meeting of the Organization for Human Brain Mapping. http://nipy.sourceforge.net/dipy/
Rasmussen C.E. (2006), Gaussian processes for machine learning.
Schrouff J. et al. (2012), ‘PRoNTo: Pattern Recognition for Neuroimaging Toolbox’, 18th Annual Meeting of the Organization for Human Brain Mapping. http://www.mlnl.cs.ucl.ac.uk/pronto
Smith S.M. et al. (2004), ‘Advances in functional and structural MR image analysis and implementation as FSL’, Neuroimage, 23 Suppl 1:S208-S219.
Tournier J.D., et al. (2007), ‘Robust determination of the fibre orientation distribution in diffusion MRI: non-negativity constrained super-resolved spherical deconvolution’, Neuroimage, 35:1459-1472.
