Motivated by analysis of gene expression data measured in multiple tissues, we introduce a matrix-normal graphical model for studying the conditional independence structures among genes using data across multiple tissues.We develop a regularized estimation method for model estimation and graph selection.We present theoretical estimation bounds and sparsistency results of our estimates.We then apply the methods for a gene expression data set of mice measured over multiple issue.The resulting graph is biologically more interpretable than methods based on single issue or methods based on lnmping all tissues together.