Kinase Gcn2 is activated by
amino acid starvation and downregulates translation initiation by phosphorylating the alpha subunit of translation
initiation factor 2 (eIF2alpha). The Gcn2
kinase domain (KD) is inert and must be activated by
tRNA binding to the adjacent regulatory domain. Previous work indicated that Saccharomyces cerevisiae Gcn2 latency results from inflexibility of the hinge connecting the N and C lobes and a partially obstructed
ATP-binding site in the KD. Here, we provide strong evidence that a network of hydrophobic interactions centered on Leu-856 also promotes latency by constraining helix alphaC rotation in the KD in a manner relieved during
amino acid starvation by
tRNA binding and autophosphorylation of Thr-882 in the activation loop. Thus, we show that mutationally disrupting the hydrophobic network in various ways constitutively activates eIF2alpha phosphorylation in vivo and bypasses the requirement for a key
tRNA binding motif (m2) and Thr-882 in Gcn2. In particular, replacing Leu-856 with any nonhydrophobic residue activates Gcn2, while substitutions with various hydrophobic residues maintain
kinase latency. We further provide strong evidence that parallel, back-to-back dimerization of the KD is a step on the Gcn2 activation pathway promoted by
tRNA binding and autophosphorylation. Remarkably, mutations that disrupt the L856 hydrophobic network or enhance hinge flexibility eliminate the need for the conserved
salt bridge at the parallel dimer interface, implying that KD dimerization facilitates the reorientation of alphaC and remodeling of the active site for enhanced
ATP binding and catalysis. We propose that hinge remodeling, parallel dimerization, and reorientation of alphaC are mutually reinforcing conformational transitions stimulated by
tRNA binding and secured by the ensuing autophosphorylation of T882 for stable
kinase activation.