Skip to main content
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2014 Apr 5.
Published in final edited form as: Science. 2013 Apr 5;340(6128):75–78. doi: 10.1126/science.1229521

Decameric SelA•tRNASec Ring Structure Reveals Mechanism of Bacterial Selenocysteine Formation

Yuzuru Itoh 1,2,3, Markus J Bröcker 4, Shun-ichi Sekine 1,2, Gifty Hammond 4, Shiro Suetsugu 3, Dieter Söll 4,5,*, Shigeyuki Yokoyama 1,2,*
PMCID: PMC3976565  NIHMSID: NIHMS567086  PMID: 23559248

Abstract

The 21st amino acid, selenocysteine (Sec), is synthesized on its cognate transfer RNA (tRNASec). In bacteria, SelA synthesizes Sec from Ser-tRNASec, whereas in archaea and eukaryotes SepSecS forms Sec from phosphoserine (Sep) acylated to tRNASec. We determined the crystal structures of Aquifex aeolicus SelA complexes, which revealed a ring-shaped homodecamer that binds 10 tRNASec molecules, each interacting with four SelA subunits. The SelA N-terminal domain binds the tRNASec-specific D-arm structure, thereby discriminating Ser-tRNASec from Ser-tRNASer. A large cleft is created between two subunits and accommodates the 3′-terminal region of Ser-tRNASec. The SelA structures together with in vivo and in vitro enzyme assays show decamerization to be essential for SelA function. SelA catalyzes pyridoxal 5′-phosphate–dependent Sec formation involving Arg residues nonhomologous to those in SepSecS. Different protein architecture and substrate coordination of the bacterial enzyme provide structural evidence for independent evolution of the two Sec synthesis systems present in nature.


The micronutrient selenium is required for human and animal health (1). Selenium is present in proteins in the form of the 21st amino acid, selenocysteine (Sec), in which the thiol moiety of cysteine is replaced by a selenol group (2). Sec is located in the active sites of many redox enzymes and is encoded by a UGA stop codon in all three domains of life (3). Sec lacks its own aminoacyl-tRNA synthetase and is synthesized by the tRNA-dependent conversion of Ser (3). The first step in Sec synthesis is the formation of Ser-tRNASec by seryl-tRNA synthetase (SerRS) (3). In bacteria, the selenocysteine synthase SelA then converts the Ser-tRNASec to Sec-tRNASec. Archaea and eukaryotes use an intermediate step in which the hydroxyl group of Ser-tRNASec is phosphorylated by O-phosphoseryl-tRNA kinase (PSTK) (4) to give Sep-tRNASec, the substrate for the homo-tetrameric enzyme SepSecS, the final synthetic enzyme (5, 6). Both SelA and SepSecS are fold-type-I pyridoxal 5′-phosphate (PLP)–dependent enzymes (7) that use selenophosphate as the selenium donor (3, 8). All Sec-synthesis systems must strictly discriminate Ser-tRNASec from Ser-tRNASer. tRNASec is the longest tRNA (9), and its tertiary structure is quite different from that of a canonical tRNA (8, 10). In archaea and eukaryotes, PSTK discriminates Ser-tRNASec from Ser-tRNASer for phosphorylation (11), and then SepSecS recognizes Sep-tRNASec in a phosphate-dependent manner (8). However, because of the lack of crystallographic studies on SelA, the discrimination mechanism in bacteria has remained elusive. Cryogenic electron microscopy suggested SelA to be a homodecameric enzyme of >500 kD (12, 13), in contrast to the ~220-kD SepSecS homotetramer (8).

Here, we present the crystal structures of full-length decameric Aquifex aeolicus SelA (SelA-FL, residues 1 to 452), alone (3.9 Å resolution) and in complex with Thermoanaerobacter tengcongensis tRNASec (7.5 Å), and of a SelA mutant lacking the N-terminal domain (SelA-ΔN, residues 62 to 452) with and without thiosulfate (3.25 and 3.20 Å, respectively). Biochemical and genetic experiments (table S1) were performed with Escherichia coli SelA (A. aeolicus numbering is used in this Report, fig. S1).

SelA is a homodecamer in which the 10 subunits form a pentamer of dimers (Fig. 1). Each subunit consists of the N-terminal domain (residues 1 to 66), the N-linker (residues 67 to 89), the core domain (residues 90 to 338), and the C-terminal domain (residues 339 to 452) (fig. S2A). The intimate dimer in SelA contains two catalytic sites formed at the subunit-subunit interface, where the cofactor PLP is covalently linked to a conserved Lys285 (fig. S2, B to E). The N-terminal domain protrudes from the central pentagon, is intrinsically mobile (fig. S3), and only contacts the other core domain of the intimate dimer. The orientation of the C-terminal domain (relative to the core domain) differs substantially from that of the C-terminal domains in other fold-type-I PLP-dependent enzymes (7) (fig. S4). Its orientation creates a large space between the core and C-terminal domains (fig. S2A), which allows the interaction with the neighboring intimate dimer for decamerization (fig. S4B) and the formation of a large cleft between two intimate dimers (Fig. 1).

Fig. 1. Structure of SelA.

Fig. 1

Overall structure of A. aeolicus SelA. PLPs are represented as sphere models. The 10 subunits are shown in different colors. The disordered regions in the N-terminal domains are colored gray.

The structure (Fig. 2A) of A. aeolicus SelA complexed to T. tengcongensis tRNASec at 7.5-Å resolution (cocrystals with the homologous tRNA only diffracted to ~20 Å) revealed that the SelA decamer binds up to 10 tRNASec molecules. Despite the low resolution of the ~0.8-MD ribonucleo-protein, the positions of the tRNASec molecules were unambiguously detected in the electron density map, except for the CCA terminus (Fig. 2B). Four SelA subunits interact with one Ser-tRNASec, with one dimer holding tRNASec and the other providing the catalytic site (Fig. 2B). The large cleft (Fig. 1) provides the space for tRNASec to approach the catalytic site on the neighboring dimer. Therefore, the proper relative positioning of the two dimers is required to perform the overall reaction and is fixed by the ring closure to form the decameric structure (Fig. 2C). If SelA assumed the quaternary structure of any other fold-type-I PLP enzyme [e.g., the tetrameric SepSecS (8)], the tRNA-binding and catalytic sites could not work together. In fact, a quadruple mutation introduced at the dimer-dimer interface (Thr191-Thr192-Asp199-Tyr220→Tyr191-Tyr192-Arg199-Pro220) caused a dimeric quaternary structure (fig. S5A) and abolished SelA activity in vivo (table S1) and in vitro (fig. S5B).

Fig. 2. Structure of the SelA•tRNASec complex.

Fig. 2

(A) Overall structure of A. aeolicus SelA complexed with T. tengcongensis tRNASec (colored brown). (B) Interaction of tRNASec with intimate dimers I•J and A•B (close-up view). The omit Fobs-Fcalc map (3.0 sigma level) of the tRNASec is shown in a blue mesh. The putative position of the CCA region is shown as a dashed line. (C) A schematic diagram of the interaction of tRNASec with dimers I•J and A•B. (D and E) Discrimination of tRNASec from tRNASer by SelA. T. thermophilus tRNASer [PDB ID 1SER (14)] was docked to the SelA•tRNASec complex by superimposing the backbone phosphorus atoms of the T arm. The tRNASec surface is complementary to that of the SelA N-terminal domain (D), whereas that of tRNASer is not complementary (E).

SelA interacts with the D-arm side of the L-shaped tRNASec and does not contact either the extra arm or the anticodon arm (Fig. 2B). The N-terminal domain of SelA binds the D arm and the T loop of tRNASec (fig. S6, A and B). The SelA-ΔN mutant, lacking the N-terminal domain, did not bind Ser-tRNASec or tRNASec, whereas the dimeric mutant dSelA, which retains the N-terminal domain, can still bind tRNASec (fig. S5, C and D). The dissociation constant (KD) for dSelA is 624 ± 84 (SD) nM, whereas that for the wild type is 75 ± 13 nM, indicating that more than 90% of the standard change in Gibbs free energy (ΔG°) for tRNASec binding is ascribed to the tRNASec interactions within one dimer, particularly with the N-terminal domain. The specific interaction between the SelA N-terminal domain and the tRNASec D arm accomplishes the discrimination of the substrate Ser-tRNASec from the structurally divergent nonsubstrate Ser-tRNASer [the maximal six-base-pair D stem and minimal four-nucleotide D loop in tRNASec compared with the three-base-pair D stem and 11-nucleotide D loop in tRNASer (fig. S7)]. The D stem of T. tengcongensis tRNASec contains the fifth and sixth pairs G14:U21 [G14:C21 in A. aeolicus tRNASec (fig. S7)] and C15:G20a, whereas D loop nucleotide U16 forms the unique tertiary base pair U16:U59. The SelA N-terminal domain contacts C15, U16, G20a, and U21 because the shape of the unique D-arm structure of tRNASec is complementary to the protein structure (Fig. 2D and fig. S6D). Because of the incompatible shape of tRNASer, the corresponding nucleotides in the canonical augmented D helix of tRNASer cannot interact with the SelA N-terminal domain (Fig. 2E and fig. S6D). In contrast, the difference in the number of base pairs in the acceptor stem between tRNASec and tRNASer (eight and seven, respectively) barely contributes to the discrimination, probably because of the inherent flexibility of the N-terminal domain (fig. S6D).

The large cleft described above enables SelA to hold the end of the acceptor stem: G1 from the G1:C72 pair and the G73 discriminator nucleotide interact with the C-terminal domain of the neighboring subunit (Fig. 2B and fig. S6C). The deletion of residues 423 and 424 within the β18-loop region, which contacts G73 (fig. S6C), yielded completely inactive enzymes (table S1), indicating that the binding of G1 and G73 is required for properly directing the acceptor arm toward the catalytic site. Although the CCA end is not visible, probably because tRNASec is not ligated with serine, the length of CCA-Ser is consistent with the distance between G73 in tRNASec and PLP in the catalytic site (fig. S6C). Therefore, the large cleft can accommodate the terminal region, from the first base pair of the acceptor stem to the Ser-adenosine moiety of Ser-tRNASec (fig. S8A), whereas the N-terminal domain recognizes the tRNASec-specific D-arm structure. In contrast, SepSecS binds Sep-tRNASec on the surface of the protein globule in a completely different manner from that of SelA and is therefore unable to recognize the D arm (fig. S8, B and C).

We also determined the crystal structure of SelA-ΔN in complex with thiosulfate, which revealed thiosulfate ions (TS1 to TS3) bound to one SelA subunit (Fig. 3A, fig. S9A). TS1 binds in the putative selenophosphate-binding pocket formed by Arg86A (subunit A), Arg312B, and Arg315B (subunit B). Mutations of these residues to Ala resulted in markedly reduced activities. Furthermore, Arg119 and Asp284, which interact with Arg312B and Asp86A, respectively, are required for catalytic activity (table S1). TS2 and TS3 are thought to mimic the phosphate group of the terminal A76 of tRNASec. Because the Ala mutations of Asn218 and Phe224 drastically reduced the activity in vivo (table S1), Asn218A and Phe224J might form part of the A76-binding pocket at the interface between the subunits A and J (Fig. 3, B and C). Notably, the interaction of A76 with both Asn218A and Phe224J can only occur in the decameric SelA (fig. S9B). The mechanism of PLP-dependent Ser-to-Sec conversion by SelA (depicted in Fig. 4 and explained in the legend) is similar to that of SepSecS (8) and, unlike other fold-type-I PLP enzymes, involves a series of crucial Arg residues. However, SelA and SepSecS use respective sets of nonhomologous Arg residues.

Fig. 3. Catalytic site of SelA.

Fig. 3

(A) The catalytic site of SelA-ΔN in complex with thiosulfate. PLP and thiosulfate ions (TS1 to TS3) are shown. TS1 mimics the substrate selenophosphate (Se-P). The subunits of the intimate dimer A•B are colored as in fig. S2A, whereas subunit J is colored gray. (B) Docking model of the Ser-ligated terminal tRNASec nucleotide (A76) and Se-P. (C) SelA surface, showing that A76 binds to the boundary of subunits J and A. The docked A76-Ser and G73 of tRNASec are connected through C74-C75.

Fig. 4. The mechanism of Sep-to-Sec conversion.

Fig. 4

In the waiting state, the PLP forms an internal aldimine with Lys285 (A). The amino group of Ser-tRNASec replaces Lys285 to form an external aldimine (B), followed by deprotonation of the α position of the Ser moiety (C). Lys285 protonates and eliminates the β hydroxyl group to form a 2-aminoacrylyl moiety (D). The nucleophilic addition of selenophosphate generates a Se-phosphorylated Sec moiety (E), which is hydrolyzed to a Sec moiety and a phosphate (F). Lastly, the α position of the Sec moiety is protonated (G), and Sec-tRNASec is released (H).

In summary, the SelA pentamer-of-dimers forms a ring-shaped structure and binds 10 tRNASec molecules. Each tRNASec interacts with all subunits from the two neighboring dimers, and this interaction critically depends on the pentagonal ring architecture. The SelA N-terminal domain recognizes the tRNASec-specific D-arm structure, thereby discriminating Ser-tRNASec from Ser-tRNASer. Moreover, the large cleft created between two SelA dimers accommodates the Ser-tRNASec 3′-terminal region. Thus, the functional form of SelA differs from that of the tetrameric SepSecS, which recognizes the phosphoseryl moiety of Sep-tRNASec without tRNASec/tRNASer discrimination. Moreover, SelA catalyzes Sec formation by using a different set of active-site Arg residues from that of SepSecS. The entirely different structures and divergent catalytic residues of the bacterial and archaeal/eukaryotic Sec synthases are evidence for convergent evolution of two Sec synthesis systems present in nature.

Supplementary Material

SOM

Acknowledgments

Y.I. was supported by research fellowships from the Japan Society for the Promotion of Science (JSPS). M.J.B. holds a Feodor Lynen Postdoctoral Fellowship of the Alexander von Humboldt Foundation (Bonn, Germany). We thank the staff members of SPring-8 BL41XU and the Photon Factory beam lines for assistance with our data collection, A. Ishii and T. Nakayama for assistance in the manuscript preparation, and M. Simonovic (University of Illinois, Chicago) for critical discussions. We are grateful to the National BioResource Project in Japan for providing the ASKA E. coli K12 ORF library. This work was supported in part by JSPS Grants-in-Aid for Scientific Research (A) to S.Y. and (C) to S.S. and the Targeted Proteins Research Program of the Ministry of Education, Culture, Sports, Science and Technology (to S.Y.). D.S. acknowledges support from the Division of Chemical Sciences, Geosciences, and Biosciences, Office of Basic Energy Sciences of the U.S. Department of Energy (DE-FG02-98ER20311) for funding the genetic experiments; the National Institute of General Medical Sciences (GM22854); and Defense Advanced Research Projects Agency (contracts N66001-12-C-4020 and N66001-12-C-4211). The plasmids for protein expression and tRNA transcription are available from S.Y. under a material transfer agreement with RIKEN. The atomic coordinates and structure factors have been deposited in the Protein Data Bank (PDB IDs 3W1H, 3W1I, 3W1J, and 3W1K).

Footnotes

References and Notes

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

SOM

RESOURCES