Brain X chromosome inactivation is not random and can protect from paternally inherited neurodevelopmental disease.

Non-random (skewed) X chromosome inactivation (XCI) in the female brain can ameliorate X-linked phenotypes, though clinical studies typically consider 80-90% skewing favoring the healthy allele as necessary for this effect1-10. Here we quantify for the first time whole-brain XCI at single-cell resolution and discover a preferential inactivation of paternal to maternal X at [~]60:40 ratio, which surprisingly impacts disease penetrance. In Fragile-X-syndrome mouse model, Fmr1-KO allele transmitted maternally in [~]60% brain cells causes phenotypes, but paternal transmission in [~]40% cells is unexpectedly tolerated. In the affected maternal Fmr1-KO(m)/+ mice, local XCI variability within distinct brain networks further determines sensory versus social manifestations, revealing a stochastic source of X-linked phenotypic diversity. Taken together, our data show that a modest [~]60% bias favoring the healthy allele is sufficient to ameliorate X-linked phenotypic penetrance, suggesting that conclusions of many clinical XCI studies using the 80-90% threshold should be re-evaluated. Furthermore, the paternal origin of the XCI bias points to a novel evolutionary mechanism acting to counter the higher rate of de novo mutations in male germiline11-16. Finally, the brain capacity to tolerate a major genetic lesion in [~]40% cells is also relevant for interpreting other neurodevelopmental genetic conditions, such as brain somatic mosaicism.