The unfolded protein response (UPR) is an intracellular signaling pathway that provides homeostatic feedback regulation between the endoplasmic reticulum (ER) and the gene expression program in the nucleus. To this end, the UPR senses the conditions inside the ER, detecting an imbalance between newly made proteins and the protein folding capacity in the ER, and activates a corrective response. For signaling, the UPR uses a transmembrane sensor of ER-lumenal unfolded proteins, Ire1.
Ire1 is an ER membrane-resident receptor that serves as a primary signal transduction device in the UPR conserved from yeast to mammalian cells [1
]. Oligomerization of Ire1-lumenal domains is thought to be a key event in initiating signal propagation across the ER membrane that enables the cooperative assembly of Ire1's cytosolic kinase and RNase modules into an ordered oligomer with a defined three-dimensional structure [5
]. The oligomer is stabilized by phosphates resulting from autophosphorylation of the kinase domain and allows juxtaposition of Ire1's RNase domains, which presumably activate the RNase [5
]. Once activated, Ire1 initiates the non-conventional splicing of HAC1
mRNA (yeast) or XBP1
mRNA (metazoan) by cleaving the mRNA at two conserved sites to excise an intron [5
]. An RNA ligase (tRNA ligase in yeast and a still unknown enzyme in metazoan cells) rejoins the severed exons to complete the reaction. Intron removal allows for the production of the UPR transcription activators Hac1 and XBP1, respectively, which upregulate UPR target genes.
Ire1's RNase domain thereby serves the primary role in signal transmission. The mechanism underlying activation of Ire1 RNase by oligomerization and the mechanism of mRNA recognition and cleavage have remained elusive. To date, there are no known structural homologues of the Ire1 RNase domain that could help answering these questions. The best attempt at defining the mechanism of RNA cleavage has been made based on sequence conservation arguments and on a crystal structure of Ire1 dimer with a ligand-free RNase domain, which also lacked a fragment of the RNase active site [6
]. In this work, the RNase dimer was proposed to contain two independent catalytic centers, one per Ire1 monomer, which would simultaneously accommodate the two RNA stem-loops conserved in all known mRNA substrates of Ire1 [6
]. According to the proposed model, two stem-loops would form a kissing interaction for docking and recognition by Ire1. Subsequent studies proposed [5
] that such a kissing interaction is not likely because Ire1 exhibits no preference for RNA substrates with dual stem-loops over substrates with a single stem-loop, indicating that a single stem-loop structure is the cognate folded substrate of Ire1 RNase.
It has been also suggested [6
] that putative active site residues are positioned similarly in Ire1 RNase and in pre-tRNA splicing endonuclease (SEN), although these two endoribonucleases share no apparent sequence or structural homology and cleave dissimilar RNA substrates (stem-loop versus bulge-helix-bulge motif). The authors point out [6
] that the putative catalytic residues in Ire1 do not align completely with those in SEN and would require a 5-Å displacement of Y1043 (in yeast Ire1) to bring it into a position corresponding to that of Y249 in SEN. A 5-Å conformational change in Ire1 RNase upon substrate binding has been proposed based on these arguments [6
]. A new cocrystal structure of Ire1 RNase with an oligonucleotide bound, which we provide in this work, does not support considerable conformational changes in the position of Y1043. Therefore, Ire1 and SEN apparently cleave RNA using different arrangements of active site residues. Underscoring this difference, substitution of a single catalytic histidine residue H1061 produced orders-of-magnitude greater effect on the catalytic activity of Ire1 (this work) compared to only a 28-fold rate reduction in SEN [8
Conservation of residue H1061, the only invariant histidine in the RNase domain of Ire1, was used to propose that H1061 marks the catalytic center and serves for general acid-general base catalysis [6
]. In agreement with this model, mutation H1061A reduced RNA cleavage [6
]. However, this mutation was analyzed using a qualitative approach that had a narrow dynamic range and did not distinguish between effects on binding and catalysis. Furthermore, mutation H1061A not only removed the ability of histidine to transfer protons but also the ability to form hydrogen bonds, which alone could explain the observed effect on RNA cleavage, without invoking H1061 in general acid-general base catalysis. Considering that ribonucleases can cleave RNA stem-loops by using for proton transfer basic amino acids instead of conventional ionizable residues (histidine, glutamate and aspartate) [9
], the very involvement of H1061 in general acid-general base catalysis could not be deduced convincingly from any of the available data.
To understand the mechanism of RNA cleavage by Ire1, we combined quantitative analysis of rationally designed Ire1 mutants and X-ray crystallography. We show that proton transfer by histidine H1061 contributes greater than five orders of magnitude (≥ 7.6 kcal/mol) to catalysis of RNA cleavage, which experimentally defines a key catalytic functionality in the active center of Ire1 RNase. Using the non-disruptive Y1043F mutation we experimentally demonstrate a catalytic role of the OH group of Y1043 and also show an important role in RNA binding of a helix-loop element (HLE) residue, R1039. Our findings suggest parallels between catalytic centers of Ire1 and the well characterized RNases A and T1 and propose an unexpected mechanism of oligomerization-induced in trans RNA recognition by Ire1.