Abstract
The human unr gene encodes an 85 kDa protein which contains five cold shock domains (CSD). The capacity of Unr to interact in vitro with RNA and its intracellular localization suggest that Unr could be involved in some aspect of cytoplasmic mRNA metabolism. As a step towards identification of Unr mRNA targets, we investigated the RNA-binding specificity of Unr by an in vitro selection approach (SELEX). Purine-rich sequences were selected by Unr, leading to the identification of two related consensus sequences characterized by a conserved core motif AAGUA/G or AACG downstream of a purine stretch. These consensus sequences are 11-14 nt long and appear unstructured. RNAs containing a consensus sequence were bound specifically by Unr with an apparent dissociation constant of 1 x 10(-8) M and both elements, the 5' purine stretch and the core motif, were shown to contribute to the high affinity. When the N-terminal and C-terminal CSD were analyzed individually, they exhibited a lower affinity than Unr for winner sequences (5- and 100-fold, respectively) but with similar binding specificity. Two combinations of CSDs, CSD1-2-3 and CSD1*2-3-4-5 were sufficient to achieve the high affinity of Unr, indicating some redundancy between the CSDs of Unr for RNA recognition. The SELEX-generated consensus motifs for Unr differ from the AACAUC motif selected by the Xenopus Y-box factor FRGY2, indicating that a diversity of RNA sequences could be recognized by CSD-containing proteins.