This study aims at assessing the burden of rare (minor allele frequency < 1%) predicted damaging variants in the whole exome of 92 bipolar I disorder (BD) patients and 1051 controls of French ancestry. Patients exhibiting an extreme phenotype (earlier onset and family history of mood disorder) were preferentially included to increase the power to detect an association. A collapsing strategy was used to test the overall burden of rare variants in cases versus controls at the gene level. Only protein-truncating and predicted damaging missense variants were included in the analysis. Thirteen genes exhibited p values exceeding 10-3 and could be considered as potential risk factors for BD. Furthermore, the validity of the association was supported when the Exome Aggregation Consortium database non-Finnish European population was used as controls for eight of them. Their gene products are involved in various cerebral processes, some of which were previously implicated in BD and belong to pathways implicated in the therapeutic effect of lithium, the main mood stabilizer. However, exome-wide threshold for association study was not reached, emphasizing that larger samples are needed.