Subjects
The study protocol was reviewed and approved by the Ethical Committee of St-Joseph University – Hôtel-Dieu de France hospital, and has therefore been performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments. All subjects gave written informed consent in order to participate in the study.
30 volunteers were included in this prospective study. A medical history was recorded for every volunteer. Handedness of the subjects was determined by the Edinburgh Handedness Inventory. Inclusion criteria was trilinguism with a different age acquisition of each language. The 3 spoken languages were English, French and Arabic.
Exclusion criteria included: ambidexterity, as well as neurologic, psychiatric, or any other significant medical disease.
The volunteers were divided in two groups of 15 subjects each, a left-handed and a right-handed group. The two groups were matched for age and gender. Each group included 8 women and 7 men. The mean age was 27.2 years of age with a 2.8 SD in the right-handed group and 25.5 years of age with a 3.6 SD in the left-handed group.
We labelled L1, L2 and L3 the native language, the second and the last acquired language respectively. The mean age of acquisition was birth for L1, 3 years of age for L2 and 9 years of age for L3. The Arabic language represented L1 in 70% of the cases, the other 30% had the French language as a native language. L2 was the French language in 70% of our subjects, Arabic in 26.7% and English in 3.3% of our population.
The Common European Framework of Reference for Languages CEFRL was used as a questionnaire to objectively determine the language proficiency of the subjects. Their language habits including how many hours per day and in which context (social, professional) they use each language were also reported. Subjective proficiency was also assessed with the subjects ranking L1, L2 and L3 from the language they feel the most comfortable using (score of 1) to the language they feel the least comfortable using (score of 3).
Image acquisition
Scanning was performed on a 3T MRI scanner (GE Healthcare, Milwaukee, Wisconsin) using an eight-channel head coil. An axial FLAIR sequence (repetition time/echo time (TR/TE) = 10000/140.2 ms, field of view (FOV) = 220 × 220 mm, flip angle = 90°, thickness = 4 mm) was performed to rule out parenchymal abnormalities. Three-dimensional axial T1-weighted images were acquired using a spoiled gradient-recalled echo sequence (TR/TE = 8.4/2.6 ms, FOV = 260 × 260 mm, flip angle = 15°, matrix = 256 × 256 voxels, thickness = 1.2 mm). fMRI data were then acquired using a single shot gradient echo echo-planar imaging (EPI) sequence: TR/TE = 3000/30 ms, FOV = 260 × 260 mm, flip angle = 90°, matrix = 128 × 128 voxels, thickness = 4.5 mm, no skip, EPI voxel size = 1.875 × 1.875 × 4.5 mm. A visual responsive naming paradigm 28 was repeated 3 times, once for each language. The order of administration of L1, L2 and L3 paradigms was random and independent of the age of acquisition and language proficiency of the subject and it was counterbalanced across participants. A block design was used with a 30 s of activation (five questions) followed by a 30 s rest period where the subject was asked to fixate on a cross-hair. The block design was repeated five times (5 min in total). The blocks of activation (Epoch) were constructed similarly for each language and consisted of the same questions for each language. The order of the questions was however random within each block. The order of the blocks within each paradigm was also different to account for the effect of habituation.
DTI scans were acquired using a whole-brain 34 direction diffusion-weighted images (DWI) with the following parameters: TR/TE = 8000/86 ms, FOV = 260 mm, matrix = 128 × 128 voxels, slice thickness = 2 mm, b values of 0 and 1000 s/mm2, voxel size = 2 x 2 x 2.6 mm.
fMRI data analysis
Post processing and analysis of the fMRI images were performed using SPM12 available on the web: (https://www.fil.ion.ucl.ac.uk/spm/software/spm12/). The images were normalized into the International Consortium for Brain Mapping (ICBM) space template for European brains. Realignment of the functional, blood oxygenation level dependent (BOLD) images and 3D axial T1 weighted images was performed using the mean image as reference. In order to improve signal to noise ratio, realignment was then followed by smoothing using a Gaussian kernel of 8mm full-width at half maximum.
We obtained statistical analyses of fMRI at the level of a single subject and then we performed another statistical analyses at a group level for both the left and the right-handed groups.
A General Linear Model technique was used to generate the statistical parametric maps. The task and rest conditions were compared using a t test at p < 0.001 (uncorrected).
Language evaluation
The number of active clusters in the language areas and the number of active voxels (voxel size of 2 x 2 x 2 mm) in a cluster were reported by a neuroradiologist, blinded to the handedness of the subject, the age of acquisition and the language proficiency. The language areas that were studied included the activated clusters in the inferior frontal gyrus, the middle frontal gyrus, the superior frontal gyrus, the superior temporal gyrus, the middle temporal gyrus, the supramarginal gyrus and the angular gyrus.
Language Lateralization by fMRI
The language laterality index (LI) was calculated for every subject for L1, L2 and L3 using the standard formula 29 for fMRI: fMRI-LI=(L − R)/(L + R), where L and R are the numbers of voxels in the clusters of language in the left and right hemispheres, respectively. The fMRI-LI ranged from − 1 (complete right dominance) to + 1 (complete left dominance). Right hemisphere language laterality was defined as − 1 ≤ fMRI-LI < − 0.2, bilaterality as − 0.2 ≤ fMRI-LI ≤ 0.2, and left hemisphere language laterality as 0.2 < fMRI-LI ≤ 1.
DTI data analysis
DTI images were reconstructed using the dcm2niigui application in MRIcron, a cross-platform NIfTI format image viewer (https://www.nitrc.org/projects/mricron). Alignment of the anatomical and DTI data for each subject was performed in AFNI software (https://afni.nimh.nih.gov) after gradient distortion correction and head motion correction. Images were then normalized into the International Consortium for Brain Mapping (ICBM) space template. The aligned and motion-corrected DTI data and diffusion gradients DTI were processed with the Diffusion Toolkit software developed by the Martinos Center for Biomedical Imaging, Massachusetts General Hospital (http://www.trackvis.org) 30. For each voxel, a tensor matrix was derived. After diagonalization of the matrix, eigenvalues were obtained and maps of diffusivity (MD) and fractional anisotropy (FA) were created. Trackvis (http://www.trackvis.org/) was then used to reconstruct the AF in the left and in the right hemispheres using a region of interest (ROI) approach (Fig. 1). Deterministic tract reconstruction using a fiber association by continuous tracking algorithm (35-degree angular threshold) was performed for fiber-tracking analysis
The images were processed by a neuroradiologist blinded to the patient’s handedness and to the results of the fMRI. Regions of interest for tract segmentation were placed manually on the color FA maps cross referenced to the b0 DWI images in the frontal and temporal lobes on the left and on the right, in the expected location of the AF. The AF was derived by a two-ROI approach, one at the level of the rolandic operculum in coronal view and the other laterally to the ventricular trigone on an axial view. Fibers that pass through both regions defined the AF 26. Any aberrant fibers in reference to the known anatomy were removed. The AF in each subject was categorized as present on the left only, on the right only, or bilaterally. Mean Diffusivity (MD), mean FA, Number of Fibers, Fiber Length and Fiber Volume were then calculated for each identifiable AF.
Language Lateralization by DTI
The language laterality index determined by DTI (DTI-LI) was calculated for every subject using the standard DTI-LI formula: DTI-LI = (Left AFV − Right AFV) / (Left AFV + Right AFV), where AFV is the AF Volume. DTI-LI ≥ + 0.1 value indicates left hemispheric lateralization, ≤ − 0.1 indicates right hemispheric lateralization and the values between + 0.1 and − 0.1 represent bilateral lateralization 17,12,14.
Comparison of DTI-LI and fMRI-LI
The first language acquired being the native language of the subject, language lateralization on fMRI for the first language acquired L1 (LI-L1) was adopted as the language lateralization of the subject on fMRI. A prior fMRI study conducted on these patients demonstrated that irrespective of language proficiency and age of acquisition, the lateralization of language does not change for right-handed subjects. It does however change for some left-handed subjects, with the language of the least proficiency (usually L2 or L3) presenting a bilateral activation in few cases31.
Statistical analysis was performed using the Statistical Package for Social Sciences (SPSSv23). Correlation between FA Index (Left FA – Right FA) / (Left FA + Right FA), DTI-LI and fMRI Laterality Index (LI-L1) was determined using Pearson correlation. Comparison between the left AF and the right AF regarding MD, FA, Number of Fibers, Fiber Length, Fiber Volume was evaluated using Student’s t-test. The Independent Sample t test was used to determine if there were differences between the AF of the right-handed subjects and the AF of the left-handed subjects in the right hemisphere and in the left hemisphere. The statistical significance threshold was set to p < 0.01, adjusted for multiple comparisons with Bonferroni correction.