The Role of the Cognitive Control System in Recovery from Bilingual Aphasia: A Multiple Single-Case fMRI Study

Aphasia in bilingual patients is a therapeutic challenge since both languages can be impacted by the same lesion. Language control has been suggested to play an important role in the recovery of first (L1) and second (L2) language in bilingual aphasia following stroke. To test this hypothesis, we collected behavioral measures of language production (general aphasia evaluation and picture naming) in each language and language control (linguistic and nonlinguistic switching tasks), as well as fMRI during a naming task at one and four months following stroke in five bilingual patients suffering from poststroke aphasia. We further applied dynamic causal modelling (DCM) analyses to the connections between language and control brain areas. Three patients showed parallel recovery in language production, one patient improved in L1, and one improved in L2 only. Language-control functions improved in two patients. Consistent with the dynamic view of language recovery, DCM analyses showed a higher connectedness between language and control areas in the language with the better recovery. Moreover, similar degrees of connectedness between language and control areas were found in the patients who recovered in both languages. Our data suggest that engagement of the interconnected language-control network is crucial in the recovery of languages.


Patient 2
General aphasia evaluation at T1, showed slight impairment in object naming, verbal fluency, yes/no questions and following oral semi-complex and complex commands in L1 and slight impairment in verbal fluency, word/phrase repetition and following oral and written semi-complex and complex orders L2. At T2, the scores remained stable for both languages in all subtests except for an improvement in object naming in L1 and word and phrase repetition in L2. The total GAE score showed improvement in the total GAE score of L1 (Χ 2 : 5.14, P: 0.023). In L2 no improvement was found in the total GAE score.
Picture naming during fMRI acquisition showed low accuracy score in both L1 and L2 at T1 (26 of 40 images correctly named in L1 and 28 of 40 picture correctly named in L2). At T2, there was an improvement in picture naming in L1 (Χ 2 : 5.14, p: 0.02), while no improvement was found in picture naming in L2.

Patient 3
General aphasia evaluation at T1, showed a generally good performance in all the subtests except a slight impairment in object naming and moderate impairment in verbal fluency and following written semi-complex and complex commands in both L1 and L2. At T2, her performance improved generally in both L1 and L2 (except for the absence of improvement in verbal fluency in L2). Mcnemar chi-squared test over the total GAE score showed a significant improvement in the total GAE score of both L1 (Χ 2 : 7.11, P: 0.007) and L2 (Χ 2 : 8.1, P: 0.004).
Picture naming during fMRI acquisition showed a significant improvement at T2 in picture naming in L1 (Χ 2 : 5.3, p: 0.02), however, no improvement was found in picture naming in L2 across time.

Patient 4
General aphasia evaluation at T1 in L2 showed impairment in object naming, verbal fluency and description, while L1 performance in all subtests were spared except for one error in object naming and two errors in yes/no questions. His performance in the total GAE score was stable in both L1 and L2 between T1 and T2; L1 (Χ 2 : 3.2, P: 0.07) and L2 (Χ 2 : 3.2, P: 0.07).
Picture naming during fMRI acquisition showed an impaired performance in L2 at T1 (mainly due to no responses), while picture naming in L1 was performed quite correctly. At T2, no improvement was found in picture naming in L2 (Χ 2 : 2.25, P: 0.13).

Patient 5
General aphasia evaluation at T1, showed a slight impairment in object naming (only one language switching error), impaired verbal fluency, series and impairment in following auditory and written commands in L1. However, in L2, he showed impairment in object naming, verbal fluency and following auditory and written commands. At T2, his performance improved in written command in both L1 and L2. He also showed improvement in object naming in L2, however he performed more errors in yes/no questions in L2. Mcnemar chi-squared test over the total GAE score showed no significant improvement in the total GAE score of both L1 (Χ 2 : 2.25, P: 0.13) and L2 (Χ 2 : 3.2, P: 0.07).

Supplementary Data 2:
Control subjects: Control subjects recruitment: We recruited six healthy subjects matched by age (65.6±8.5), education and AoA (18.6±4.3) however one subject was removed from the analyses as he did not meet our criteria for DCM analyses (i.e. lack of activation in one of the ROIs). The control subjects all speak French as their second language; two subjects speak Italian as L1, two speak German and one speaks Spanish as L1. In all control subjects, their main language usage and exposure at work and with family and friends is L2.
Control subjects were evaluated in a single session using behavioral language control tasks (linguistic and non-linguistic switching) and fMRI picture naming tasks. The MRI recording was performed in the scanner 2 for all the control subjects.

FMRI data analyses:
For the control group, brain activation during picture naming in L1 and L2 was shown using Fixed Effect Analyses (FFX) of the activation with a threshold of p<0.05 FWE corrected (Supplementary figure 1 a). compared to L2; i.e., higher connectedness within the language-control network for L1 compared to L2.
Pattern of activation in control subjects: Fixed Effect analyses of fMRI data shows the pattern of brain activation in our control group while picture naming in L1 and L2. A significance threshold of p<0.05 (FWE corrected) was applied.