Crystal structures of the human elongation factor eEFSec suggest a non-canonical mechanism for selenocysteine incorporation

Abstract

Selenocysteine is the only proteinogenic amino acid encoded by a recoded in-frame UGA codon that does not operate as the canonical opal stop codon. A specialized translation elongation factor, eEFSec in eukaryotes and SelB in prokaryotes, promotes selenocysteine incorporation into selenoproteins by a still poorly understood mechanism. Our structural and biochemical results reveal that four domains of human eEFSec fold into a chalice-like structure that has similar binding affinities for GDP, GTP and other guanine nucleotides. Surprisingly, unlike in eEF1A and EF-Tu, the guanine nucleotide exchange does not cause a major conformational change in domain 1 of eEFSec, but instead induces a swing of domain 4. We propose that eEFSec employs a non-canonical mechanism involving the distinct C-terminal domain 4 for the release of the selenocysteinyl-tRNA during decoding on the ribosome.

Figure 1. The overall structure and domain organization of human eEFSec.

(a) Cartoon and (b) surface representation diagrams of the chalice-like structure of human eEFSec shown in two views rotated ∼90° clockwise around vertical axis. Individual domains, linker and extreme C-terminus regions are labelled and coloured according to the scheme in (a). GDPCP is shown as sticks and Mg²⁺ as a grey sphere. (c) Schematic diagram showing domain organization in human eEFSec. White bars denote regions connecting individual domains.

Figure 2. The GTP-to-GDP exchange on human eEFSec induces a conformational change in D4 and not in D1.

(a) The global superimpositioning of eEFSec:GDPCP (blue) and eEFSec:GDP (light red) reveals a lack of the canonical conformational change in the EF-Tu-like domain. Instead, the C-terminal D4 swings ∼26° towards the dorsal face of the molecule and away from the tRNA-binding site. Two views related by ∼90° clockwise rotation around a vertical axis are shown. The view on the left is oriented so that the tRNA-binding site is in the paper plane. The view on the right is oriented so that the tRNA-binding site (or ventral face) is on the left and perpendicular to the paper plane. (b) Modelling of the canonical conformational change in eEFSec reveals that the movement of D1 might be prevented by a steric clash between the enlarged loop β17–β18 (blue) in D3 and helix α4 (red) in D1. The steric clash is highlighted with black box and shown in a close-up view.

Figure 3. The structure of the GTPase site in human eEFSec.

(a) The GTPase site is located in D1 and is composed of the structural elements conserved in small GTPases: the P loop, switch 1, switch 2, the guanine-binding sequence and Mg²⁺ ion. The binding of the GTP analog, GDPCP, causes partial ordering of switch 1, which orients Thr48 and Asp92 to interact with Mg²⁺ and waters 1 and 2 (W1, W2). (b) The GTP-to-GDP transition induces structural rearrangements mainly restricted to switch 1 and switch 2 regions. In the GDP-bound state, switch 1 is almost completely disordered and switch 2 adopts a different orientation. The conserved residues in the GTPase site are shown as sticks, water molecules are shown as red spheres, Mg²⁺ is a grey sphere and H-bonds are dashed lines. (c) Mutations of conserved residues in GTPase site (Thr48, Asp92 and His96) severely impair read-through of the Sec UGA codon and selenoprotein synthesis in vitro. Error bars represent standard deviation (s.d.) calculated from three replicates.

Figure 4. The structure of the putative Sec-binding pocket in human eEFSec.

(a) The site is located at the interface of D1 (Phe53; blue) and D2 (Asp229, His230, Arg285; red). Asp229 and Arg285 are conserved across the kingdoms, whereas His230 and Phe53 are present in eukaryotic and archaeal orthologs, but are replaced by Arg and Tyr in bacterial SelB. (b) The superimpositioning of the amino acid-binding sites reveals differences between eEFSec (blue) and EF-Tu (beige). The overall positive charge of the site in eEFSec is suggested to complement for negatively charged Sec moiety of Sec-tRNA^Sec. (c) The replacement with alanine of any of the residues in the Sec-binding pocket (Asp229, His230, and Arg285) completely abolishes the read-through of Sec UGA codon and selenoprotein synthesis in vitro. Error bars represent s.d. calculated from three replicates.

Figure 5. GTP analogues lock eEFSec in the GTP-bound state.

The overlay of human eEFSec:GDPCP (blue) onto EF-Tu:GDPNP (a, beige) and EF-Tu:GDP (b, green), reveals that the arrangement of D1-3 in eEFSec:GDPCP is similar to that in the GTP-bound state of EF-Tu. Thus, GDPCP and GDPNP are optimal GTP mimics when bound to human eEFSec. The orientation of D1 in EF-Tu:GDP can only be seen when the view in (a) and (b, left) is rotated ∼90° clockwise around the vertical axis.

Figure 6. A model of decoding of the Sec UGA codon by human eEFSec.

The 80S ribosome pauses when encountering the UGA codon. The ternary eEFSec:GTP:Sec-tRNA^Sec complex and SBP2 (grey sphere) bind to the ribosome through interactions with the SECIS element, which is in the 3′UTR of the selenoprotein mRNA. After codon recognition, eEFSec hydrolyses GTP and D1 (shades of blue) and D2 (red) move in a ratchet-like motion towards and away from the tRNA-binding (or ventral) side, respectively. Also, the domain D4 (orange) swings ∼26° away from the ventral side. These movements are emphasized in the boxed inset; the GTP- and GDP-bound states of eEFSec are shown in solid and dashed lines, respectively. After eEFSec:GDP dissociation, Sec-tRNA^Sec accommodates and formation of the peptide bond occurs. Note: the variable arm of tRNA^Sec (pointing to the right), D4 of eEFSec, SBP2 and SECIS are most likely oriented perpendicularly to the plane of the paper and towards the reader.