When a linear transform of a spherical normal has uncorrelated components

Let $\underline X\sim N_n(\underline\mu,\sigma^2I)$ and A be a $n\times n$ matrix. Then $\underline Y=A\underline X$ are uncorrelated normally distributed if and only if which of the following conditions hold?