Skip to main content
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2018 Jul 6.
Published in final edited form as: Angew Chem Int Ed Engl. 2018 May 9;57(24):7215–7219. doi: 10.1002/anie.201713215

A Facile Method for Producing Selenocysteine-Containing Proteins

Takahito Mukai 1, Anastasia Sevostyanova 2, Tateki Suzuki 3, Xian Fu 4, Dieter Söll 5,
PMCID: PMC6035045  NIHMSID: NIHMS978983  PMID: 29631320

Abstract

Selenocysteine (Sec, U) confers new chemical properties on proteins. Improved tools are thus required that enable Sec insertion into any desired position of a protein. We report a facile method for synthesizing selenoproteins with multiple Sec residues by expanding the genetic code of Escherichia coli. We recently discovered allo-tRNAs, tRNA species with unusual structure, that are as efficient serine acceptors as E. coli tRNASer. Ser-allo-tRNA was converted into Sec-allo-tRNA by Aeromonas salmonicida selenocysteine synthase (SelA). Sec-allo-tRNA variants were able to read through five UAG codons in the fdhF mRNA coding for E. coli formate dehydrogenase H, and produced active FDHH with five Sec residues in E. coli. Engineering of the E. coli selenium metabolism along with mutational changes in allo-tRNA and SelA improved the yield and purity of recombinant human glutathione peroxidase 1 (to over 80%). Thus, our allo-tRNAUTu system offers a new selenoprotein engineering platform.

Keywords: genetic code expansion, protein engineering, selenocysteine, selenoproteins, synthetic biology


Selenocysteine (Sec, U) is a fascinating building block for recombinant proteins:[1] it is more active and more resistant to irreversible overoxidation than cysteine (Cys),[2] is chemically modifiable,[3] and a diselenide bond is more stable than a disulfide bond in proteins.[4] Sec residues in proteins can be chemically converted into different amino acid side chains via a dehydroalanine intermediate.[3a,5] Recent advances in genetic code expansion have established in vivo methods in E. coli for site-specific, UAG-dependent Sec insertion into recombinant proteins mediated by the elongation factor Tu (EF-Tu), without assistance from the Sec-dedicated elongation factor SelB and Sec-insertion sequence (SECIS) element.[6] Although wildtype tRNASec species have antideterminants against EF-Tu,[7] mutations in the acceptor and the T-stem of tRNASec remove the antideterminants.[6c,7] Such EF-Tu-compatible variants of E. coli tRNASec have enabled the production of bacterial and human selenoproteins in E. coli cells.[6,8]

Sec-tRNASec is synthesized in two steps in bacteria: seryl-tRNA synthetase (SerRS) attaches serine (Ser) to tRNASec, and SelA converts the Ser moiety to Sec[9] by using selenophosphate provided by selenophosphate synthase (SelD).[10] One of the essential identity elements for E. coli SelA is the non-canonical 13-base pair (13-bp) branch (acceptor plus T-stems).[11] Therefore, E. coli tRNASec and its EF-Tu-compatible variants (UTu, UTuX, UTu6, and SecUx)[6a,c,d,8] have a 13-bp branch. While the 13-bp branch structure may be optimized for SelB-mediated recoding,[12] canonical tRNAs with a 12-bp branch may be more suitable for EF-Tu-mediated translation.[13]

To overcome this drawback, we chose SelA species that can recognize tRNASec with a 12-bp branch. A close relative of E. coli, Aeromonas salmonicida (As) subsp. pectinolytica 34mel, has one such SelA and tRNASec pair.[14] Initial attempt to develop EF-Tu-compatible variants of As tRNASec was not successful (data not shown), thus implying that even the 12-bp tRNASec structure is not suitable for SerRS-mediated amino-acylation and EF-Tu-mediated translation. Recently, we discovered a group of tRNA with unusual cloverleaf structures, allo-tRNAs, while searching for tRNASec.[15] Some allo-tRNA species were revealed to function as tRNASer in E. coli[15] and are expected to have identity elements for SelA (Figure 1A). The amber suppressor variant of an allo-tRNA species of metagenomic origin (named allo-tRNAUTu1) exhibited similar suppression efficiency as E. coli supD tRNASer (Figure 1A and Figure S1), which is one of the most active suppressor tRNAs in E. coli.[16] Thus, we coupled this allo-tRNA with As SelA.

Figure 1.

Figure 1

Translating the UAG stop codon as serine and selenocysteine. A) Cloverleaf structures of E. coli supD tRNASer and allo-tRNAUTu1. B) FDHH expression in E. coli ΔselABC ΔfdhF (ME6) cells at 25 °C with allo-tRNAUTu1, with or without Aeromonas salmonicida (As) SelA, and with reporter fdhF variant genes. Am =amber UAG codon.

We used the E. coli fdhF gene, which encodes formate dehydrogenase H (FDHH; Figure 1B), to examine EF-Tu-mediated Sec incorporation.[6a,c,d,17] The UGA codon 140 specifying the catalytic Sec residue was changed to UAG (Figure 1B). As expected, co-expression of allo-tRNAUTu1 and As SelA led to Sec insertion and the expression of wildtype FDHH, which reduced benzyl viologen to a purple dye (Figure 1B). For further estimation, we changed four Cys codons at positions 8, 11, 15 and 42 to UAG in fdhF. Each of the four Cys residues can be separately replaced with Sec without impairing the FDHH activity (Miller and Reynolds et al., unpublished) but cannot be replaced with Ser (Figure 1B). Our fdhF gene variants have one to five UAG codons that must be translated as Sec; premature translation stop by release factor 1 (RF-1) and Ser incorporation by Ser-allo-tRNAUTu1 make inactive FDHH. Surprisingly, the allo-tRNAUTu1 and As SelA pair enabled the translation of up to five UAG codons (Figure 1B). Thus, the FDHH variants may have up to five Sec residues. However, premature stop or Ser incorporation also occurred, since increasing the number of UAG codons in the fdhF reading frame decreased the FDHH activity of the cell spots (Figure 1B). Although allo-tRNAUTu1 was apparently superior to the earlier tRNASec variants in terms of the yield of selenoproteins (Figure S2), it may be inferior in terms of the purity of the selenoproteins produced.[6c] Therefore, we decided to optimize the selenium metabolism, allo-tRNA and SelA structures, and the total system, in order to improve the Ser-to-Sec conversion rate on allo-tRNA.

To increase the amount of selenium donor for As SelA, we added As SelD and Treponema denticola (Td) Sec-containing thioredoxin (Trx1; Figure 2A).[18] Expression of As SelD greatly improved the yield of the FDHH variant with 5 Sec residues (Figure 2B). We then cloned a mutant Td Trx1 gene with a Sec residue that is UAG encoded (Figure 2A). To determine the efficiency of Sec incorporation, we used human glutathione peroxidase 1(GPx1) that contains an active-site Sec residue at position 49. GPx1 mRNA with a UAG codon 49[6a] was overexpressed from a pET vector in comparable yield to a GPx1(Cys49) variant, and then analyzed by intact protein mass spectrometry. While in the absence of Td Trx1, more than 50% of the GPx1 contained Ser49 (Figure S3), while expression with Td Trx1 raised the level of Sec incorporation to more than 50% (Figure 2C and Figure S4), as judged by a comparison of mass peak intensities. Thus, As SelD and Td Trx1 may support rapid selenium supply during high demand. Since this pSecUAG-ADT system (Figure 2A) produced red-colored elemental selenium from 10 μM selenite in the growth media, the cells may have reduced enough amounts of selenite. Actually, addition of other proteins and enzymes (e.g., thioredoxin reductase, selenocysteine lyase, putative S-transporters possibly involved in Se-transfer; Figure S5A) produced no improvement (see Figures S4A and S5B).

Figure 2.

Figure 2

Se metabolism engineering. A) Putative pathways of selenium transfer to As SelA in engineered E. coli carrying pSecUAG-ADT. B) FDHH expression in ME6 cells at 25 °C. C) Intact protein mass spectrometry of the human GPx1(Ser49 and Sec49) mixture obtained from ME6 cells carrying pSecUAG-ADT. The calculated masses are 23133 Da for GPx1(Ser49) and 23193 Da for GPx1(Sec49). In many cases, putative peaks for mono-oxidized (+16 Da), di-oxidized (+32 Da), and a slightly larger (+42–47 Da) GPx1 were observed. Sec incorporation is estimated to be 55%.

We then engineered the local and overall structures of allo-tRNA (Figure 3A). An allo-tRNAUTu1 variant (named allo-tRNAUTu1D) carrying a part of the A. salmonicida tRNASec D-stem (Figure 3A) inserted Sec more efficiently than allo-tRNAUTu1 (Figures S6, S7). The U14·G21 wobble base pair in allo-tRNAUTu1D might contribute to the enhanced binding. We designed allo-tRNAUTu2D (Figure 3A and Figure S6) from an allo-tRNAAla species that carries a different cloverleaf structure to allo-tRNAUTu1.[15] Surprisingly, GPx1 proteins produced with allo-tRNAUTu1D and allo-tRNAUTu2D exhibited similar glutathione peroxidase (GPx) activities (Figure 3B), which is also supported by mass spectrometry data (Figure S7). Because allo-tRNAUTu2D expressed from a weak tRNA promoter was more easily sequestered by excess SelA than allo-tRNAUTu1D (Figure 3C), the former tRNA required a smaller amount of As SelA for Sec-tRNA formation. We chose allo-tRNAUTu2D, since SelA overexpression is a potential burden to cells.

Figure 3.

Figure 3

Engineering allo-tRNA structures. A) Cloverleaf structures of two allo-tRNA variants. Introduced mutations are indicated with red letters. B) Glutathione peroxidase (GPx) activities of GPx1 produced with allo-tRNAUTu1D and allo-tRNAUTu2D in ME6 cells at 25°C. Each bar represents the average of three independent experiments using different E. coli colonies. C) FDHH expression in ME68z cells at 25 °C. As SelA was expressed with a strong promoter (++++) or a weak promoter (+). Allo-tRNAsUTu were expressed with the indicated promoters (PargW >PselC).

Then we engineered the As SelA structure for a better positioning of allo-tRNA (Figure 4A). The SelA N-terminal domain binds the elbow region of Ser-tRNASec to bring the Ser moiety into the catalytic site of the SelA core domain, and is fixed to the core domain by hydrophobic interactions with a few residues (Figure 4A).[9] We expected that engineering of this inter-domain interaction would fine-tune the orientation of the N-terminal domain and consequently improve Ser-to-Sec conversion on engineered allo-tRNA. Based on structural information from a SelA-tRNASec complex,[9] we modified residues Pro68, Leu69, Gln72, and Cys173 of As SelA, which corresponds to Ile25, Tyr26, Lys29, and Glu129 of Aquifex aeolicus SelA (Figure 4A and Figure S8). For screening a mutant library of As SelA, we used a NMC-A β-lactamase variant[6c] in which the essential disulfide bond must be replaced by a diselenide bond in E. coli C321.ΔA.opt ΔselAB (Figure 4B). Several good variants were obtained (Table S1, Figure 4B, and Figure S9). Variants #1.9 and #2.1 carry multiple mutations (A68-I69-E72-R173 and P68-F69-S72-V173), while variant #1.9 also gained a spontaneous Pro2-to-Thr2 mutation. These As SelA variants drastically enhanced the expression of FDHH (with 5 Sec residues) at 30 °C (Figure 4C and Figure S10) and 37°C (Figure S11). Furthermore, they remained orthogonal to E. coli tRNASer (Figure S12). Finally, we combined #2.1 with Thr2 to develop SelAEvol.

Figure 4.

Figure 4

Engineering of A. salmonicida SelA. A) The amino acid residues involved in the fixation of the SelA N-terminal domain in the crystal structure of Aquifex aeolicus SelA and Thermoanaerobacter tengcongensis tRNASec (PDB ID: 3w1k). B) Screening for highly active As SelA variants by an NMC-A β-lactamase reporter assay in E. coli C321.ΔA.opt ΔselAB. Serial dilutions of cells expressing wildtype or mutant SelA were spotted on ampicillin-containing agar plates and incubated at 30°C. C) FDHH expression at 30 °C in ME6 fdhF(5 UAG codons) cells with pSecUAG-D-allo-tRNAUTu1D and pMWcat-AsSelA-(GUG) expressing wildtype or mutant SelA. Sodium selenite was added to a final concentration of 5 μM. D) FDHH expression level in ME68z fdhF (5 UAG codons and ΔSECIS) cells at 37°C was highest when both allo-tRNAUTu2D and SelAEvol were expressed at a moderate level (arabinose (ara) 0.01% and “++”, respectively).

SelAEvol with allo-tRNAUTu2D were subcloned to develop a series of pSecUAG-Evol (No. 1–4) variants, since the relative expression levels of tRNA and SelA appear to be critical. The SelAEvol expression level is controlled with four different transcription/translation initiation signals (No. 1 > 2 >3 >4). In this way, the allo-tRNA expression level can be regulated by arabinose concentration. Since unwanted suppression of amber codons at the ends of essential genes could impose burdens on the cells, the E. coli ME6 strain was transformed with a plasmid encoding essential genes with a UAG-to-UAA replacement,[19] to establish ME68z (see Materials and Methods in the Supporting Information). For FDHH yields at 37 °C, pSecUAG-Evol3 with 0.01% arabinose was the best (Figure 4D). By using pSecUAG-Evol2 with 0.001% arabinose at 20°C, we produced selenoglutaredoxin,[6a,d,20] an E. coli glutaredoxin Grx1(Cys11Sec/Cys14Ser) variant, in a yield of 0.9 mg from 2 L cell culture in a pure form using thiolsepharose chromatography (Figures S13,S14).[6a]

GPx1 was produced using pSecUAG-Evol2 with 0.001% arabinose at 25°C (Figure 5A). The protein yield and suppression efficiency were about 2–3 mgL−1 and 70%, respectively. An intact mass analysis suggested predominant (>80%) Sec incorporation (Figure 5B). We realized that the increase in the mass of GPx1(49Sec) by 44–46 Da (Figure 5B) may correspond to a Cys-to-Sec substitution at any of the five Cys positions.[21] Obviously, high levels of SelAEvol expression may lead to Sec-allo-tRNA levels in excess of the amount needed for UAG translation; hydrolysis of Sec-allo-tRNA may generate free Sec, which would produce cysteinyl tRNA synthetase mediated Sec misincorporation at Cys codons. Therefore, expression levels of SelAEvol and allo-tRNAUTu2D should be kept moderate. Curiously, in SDS-PAGE analysis, the GPx1 band disappeared, while another main band emerged around 55 kDa when the highest SelAEvol expression level (pSecUAG-Evol1) was employed (Figure 5A). Excess SelAEvol may have sequestered allo-tRNAUTu2D. Furthermore, SelAEvol was revealed to have affinity to nickel resin through the N-terminal His-rich region. Therefore, its AHSHS sequence was changed to PYR (from another Aeromonas SelA) or ASSAS (Figure S15). GPx1 proteins produced with SelAEvol and the ASSAS variant exhibited similar GPx activities, while the PYR variant was slightly less productive (Figure 5C).

Figure 5.

Figure 5

Optimizing the Sec-encoding system in ME68z cells. A) SDS-PAGE of purified GPx1 proteins produced with pSecUAG-Evol 1 and 2. Arabinose concentrations were 0.01% (++) or 0.001% (+). The supernatant and flow-through fractions were also applied to the gel. Unexpectedly, SelAEvol was also purified by nickel chromatography. B) Intact protein mass spectrometry of the rightmost sample of panel (A) produced by using pSecUAG-Evol2. Sec incorporation is estimated to be 84%. The peak of 23237.601 might derive from GPx1(Sec49) with a Cys-to-Sec substitution at any of the five Cys positions. C) Glutathione peroxidase (GPx) activities of GPx1 produced with SelAEvol variants. Each bar represents the average of three independent experiments using different E. coli colonies. D) The cleavage patterns of the MXB variants produced with As SelA variants plus SufS(C364A). In vitro intein cleavage was achieved with 100 mM dithiothreitol at room temperature over 16 h.

Finally, pSecUAG-Evol2 was modified to encode the ASSAS variant and selenocysteine lyase SufS(C364A)[22] to develop pSecUAG2. Predominant Sec incorporation was confirmed by an in vitro intein cleavage assay of an Mxe GyrA intein (MXB) variant with the substitution of a Sec for the catalytic Cys384.[6b] The yield of the full-length protein (70 kDa) at 25°C was about 10 mgL−1. Because less than 10% of the full-length protein remained intact after reaction (Figure 5D and Figure S16), and because the Ser384 variant is inactive,[6b] the Sec incorporation rate is estimated to be more than 90% by assuming complete intein reaction.

Compared to earlier tRNAUTu variants, which carry a 13-bp branch, the allo-tRNA system is intrinsically decoupled from the SelB/SECIS-mediated machinery and fully compatible with the canonical translation apparatus. Future optimization studies should improve the yield and specificity of the desired recombinant selenoproteins. The simplicity of the system should facilitate further engineering and its application to other organisms.

Supplementary Material

pdf

Acknowledgments

We thank Anna Merkuryev for help with SelA engineering and Navin Rauniyar, Ana Crnković, and Omer Ad for dedicated efforts on the MS analyses. T.M. is a Japan Society for the Promotion of Science postdoctoral fellow for research abroad. We acknowledge support of this work from the US National Institutes of Health grants (R01GM22854 and R35GM122560 to D.S.); the Department of Energy DE-FG02-98ER20311 grant (to D.S.) supported the genetic experiments.

Footnotes

Conflict of interest

The authors declare no conflict of interest.

Supporting information and the ORCID identification number(s) for the author(s) of this article can be found under: https://doi.org/10.1002/anie.201713215.

Contributor Information

Dr. Takahito Mukai, Department of Molecular Biophysics and Biochemistry Yale University, New Haven, CT 06520 (USA)

Dr. Anastasia Sevostyanova, Department of Molecular Biophysics and Biochemistry Yale University, New Haven, CT 06520 (USA)

Dr. Tateki Suzuki, Department of Molecular Biophysics and Biochemistry Yale University, New Haven, CT 06520 (USA)

Dr. Xian Fu, Department of Molecular Biophysics and Biochemistry Yale University, New Haven, CT 06520 (USA)

Prof. Dieter Söll, Department of Molecular Biophysics and Biochemistry Yale University, New Haven, CT 06520 (USA). Department of Chemistry, Yale University New Haven, CT 06520 (USA)

References

Associated Data

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

Supplementary Materials

pdf

RESOURCES