Near-cognate suppression of amber, opal and quadruplet codons competes with aminoacyl-tRNA^Pyl for genetic code expansion

Abstract

Over 300 amino acids are found in proteins in nature, yet typically only 20 are genetically encoded. Reassigning stop codons and use of quadruplet codons emerged as the main avenues for genetically encoding non-canonical amino acids (NCAAs). Canonical aminoacyl-tRNAs with near-cognate anticodons also read these codons to some extent. This background suppression leads to ‘statistical protein’ that contains some natural amino acid(s) at a site intended for NCAA. We characterize near-cognate suppression of amber, opal and a quadruplet codon in common Escherichia coli laboratory strains and find that the PylRS/tRNA^Pyl orthogonal pair cannot completely outcompete contamination by natural amino acids.

1. Introduction

Expanding protein synthesis to site-specifically introduce amino acids beyond the canonical 20 has impacted broad areas of molecular biology. Nearly 100 non-canonical amino acids (NCAAs) have been added to the genetic code [1–3]. Among these newly encoded NCAAs are useful biophysical probes [1,2] and post-translation modifications that play critical roles in basic processes ranging from cellular communication to oncogenesis [4–6]. Code expansion is made possible by employing an ‘extra’ aminoacyl-tRNA synthetase (AARS)/tRNA pair that does not cross react those of the host system and is, therefore, ‘orthogonal’ in that host cell. The AARS/tRNA pair is engineered to ligate a desired NCAA to its cognate tRNA which then directs co-translational site-specific insertion of the NCAA in protein by use of an ‘open’ codon. The vast majority of NCAAs are encoded by the UAG (amber) codon. To a lesser extent UAA (ochre), UGA (opal) and more rarely quadruplet codons (e.g., AGGA in [7]) have been used for code expansion.

Biosynthesis of biochemically homogenous proteins containing multiple non-canonical amino acids is a major goal of synthetic biology [8], yet it remains to be seen just how far the genetic code can be stretched. Current techniques enable simultaneous encoding of two NCAAs in the same polypeptide in vivo using stop [9,10] and quadruplet codons [7]. Quadruplet codons are of great interest, promising an astounding 256 ‘open’ codons. Since the initial discoveries of +1 frameshift suppressors [11,12], some progress has been made to enhance translation of quadruplet codons [7], but 4-base codons give rise to diverse translational products (e.g., [13]). Using supplementation incorporation, 3 analogs of canonical amino acids were encoded in a protein by the native E. coli aaRS/tRNA pairs using their cognate sense codons [14]. In vitro cell-free systems have incorporated up to 3 NCAAs [15].

Although in most organisms UAG, UGA and UAA signal translational stop, it is long known [16] that mutations resulting in premature stop codons can be suppressed by naturally evolved tRNAs that encode canonical amino acids with stop codons by virtue of anticodon mutation (e.g., glutamine amber suppressor supE) or by other mutations that stimulate near-cognate decoding of the stop codon (reviewed in [17]). Pyrrolysyl-tRNA synthetase and tRNA^Pyl are the only known system to have naturally evolved to promote amber codon re-assignment to a NCAA, pyrrolysine [18]. For this reason, ‘un-engineered’ PylRS and tRNA^Pyl were able to directly expand the genetic code of E. coli [19–21], and this pair has become a critical vehicle for genetic code expansion (reviewed in [3]). Biochemical [19,22] and structural [21] characterization of PylRS revealed that the enzyme does not recognize the anticodon bases, indicating that genetic code expansion beyond the amber codon is possible with PylRS and tRNA^Pyl. As a PylRS/tRNA^Pyl_UUA ochre codon decoding pair was recently characterized [10], we were prompted to evaluate the opal and quadruplet codon decoding capacity of PylRS/tRNA^Pyl in vivo in E. coli.

2. Materials and Methods

2.1 Plasmids, strains, chemicals, and protein purification

E. coli strains TOP10, BL21(DE3), and DH5α were from Invitrogen; MG1655 was from the American Type Culture Collection (ATCC). In vivo suppression experiments require two plasmids as detail previously [23]. pKTS, encodes the Methanosarcina mazei PylRS (pKTS-pylS). pTech, contains both the lacZ reporter and a tRNA expression cassette with the M. mazei pylT (pTech-pylT). The pTech plasmid with wild type lacZ reporter (lacZ(WT)) was subject to quick-change mutagenes is (Agilent) to produce the M 3 to TAG a nd M 3 to TGA variants and to convert the tRNA^Pyl_CUA to tRNA^Pyl_UCA in the opal reporter. pETtrio-pylT(CUA)-PylRS-sfGFP134TAG [24] (K.A.O. & W.R.L unpublished data) was quick-changed to pETtrio-pylT( UCCU)-sfGFP134AGGA, encoding tRNA^Pyl_UCCU and a C-terminal His₆-tagged superfolder green fluorescent protein (sfGFP) with AGGA at codon 134. tRNAs were cloned into pUC18 (XbaI/BamHI) for in vitro transcription, which was performed as described [25]. His₆-tagged PylRS (in pet15b) was over-expressed and purified as before [26]. E. coli ArgRS and TrpRS were purified from the ASKA collection strains JW1865 and JW3347 [27], respectively. His-tagged proteins were purified over Ni-NTA (Qiagen) according to manufacture’s instructions. Oligonucleotide primers (Integrated DNA Technologies) and Pyl analogs, N-ε-(tert-butyloxycarbonyl)- L-lysine (BocK) (ChemImpex) and N-ε-cyclopentyloxycarbonyl-L-lysine (Cyc) (Sigma), were purchased.

2.2 In vivo suppression assays

BL21(DE3) cells were co-transformed with pKTS (empty or bearing pylS) and pTECH vectors (empty or containing a tRNA^Pyl amber or opal variant and lacZ with a cognate amber or opal codon at M3). β-galactosidase activity was measured as before [28]. Cells were grown in LB (37°C) to early stationary phase (A₆₀₀ = 2.0); 20 μl of culture was used for each replicate and 6 independent cultures were measured in duplicate reactions. For each strain, intrinsic β-galactosidase activity (~3% of the activity of the LacZ(WT) reporter) in cells carrying empty pKTS and pTech plasmids is subtracted from all activity measurements.

BL21(DE3) cells were transformed with pETtrio-pylT(UCCU)-PylRS-sfGFP134AGGA, and sfGFP was expressed with 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) or 1 mM IPTG and 5 mM BocK.

2.3 Aminoacylation assay

Aminoacylation assays were performed for all AARS/tRNA pairs as described [25]. Enzyme, tRNA, amino acid, and ATP were included in excess concentrations to detect low activity reactions. For PylRS, ArgRS and TrpRS, aminoacylation of tRNAs was performed (37°C) in 100 mM Na-HEPES (pH 7.2), 30 mM KCl (or 60 mM KCl for PylRS only), 10 mM MgCl₂ (or 25 mM MgCl₂ for PylRS only), 10 μM AARS, 5 μM α-[³²P]-labeled tRNA, 10 mM amino acid, 2 mM ATP, and 4 mM dithiothreitol. The reaction was stopped with 1.5 volumes of 0.66 μg/μL nuclease P1 (Sigma) in 100 mM sodium citrate, pH 5.49. Because the tRNA is labeled only at A76, the radiolabeled reaction products after digestion are AMP (representing unaminoacylated tRNA) and aminoacyl-AMP (representing aminoacylated tRNA). The reaction products are separated on PEI-cellulose plates in 0.1 M ammonium acetate and 5% acetic acid then visualized, and quantified using phosphorimaging.

2.4 Mass spectrometry

Electrospray ionization mass spectrometry (ESI-MS) of sfGFP134AGGA was on an Applied Biosystems QSTAR Pulsar (Concord, ON, Canada) equipped with a nanoelectrospray ion source. The data were acquired in positive ion mode (500–2000 Da); spray voltage of +2000 V; flow rate of 700 nL/min. Spectra were analyzed with BioAnalyst software (Applied Biosystems). A mass range of m/z 500–2000 was used for spectral deconvolution and the output range was 20000 to 30000 Da (resolution of 0.1 Da and S/N threshold of 20).

3. Results

In evaluating PylRS/tRNA^Pyl pair for opal and 4-base decoding, we are interested in how and to what extent natural amino acids can be inserted at a protein locus intended for NCAA insertion. In the experiments below, we characterize background suppression of amber and opal codons, show aminoacylation activity of tRNA^Pyl variants, and attempt to install BocK with an AGGA decoding PylRS system.

3.1 Amber and opal background suppression in common E. coli strains

Given that directed evolution of AARS/tRNA orthogonal pairs requires a strain with high transformation efficiency to maintain library diversity (e.g., TOP10), while expression of NCAA-containing protein can be enhanced in a strain optimized for protein expression (e.g., BL21), we characterized amber and opal background suppression in common laboratory E. coli strains (Fig. 1). We also tested DH5α, a common cloning strain, and MG1655, which is the parent of an E. coli strain in which TAG was globally recoded to TAA for enhanced NCAA incorporation [29]. M3 of β-galactosidase was mutated to a TAG or TGA codon to generate amber and opal reporter constructs, which were separately transformed into different E. coli strains. Activity of the enzyme reports on translational read-through of amber or opal codons.

Fig. 1.

Relative amber (UAG) and opal (UGA) codon background suppression in different E. coli strains. β-galactosidase, containing either Methionine 3 (WT), M3UAG (Amber), or M3UGA (Opal), activities were calculated in Miller units. Since absolute activity varies between strains, WT activity for each strain was set to 100% to show relative suppression levels.

Amber suppression in DH5α is highly efficient, directed by the suppressor tRNA (supE44) encoded in its genome. The SupE gene product, tRNA^Gln_CUA, is aminoacylated by GlnRS and leads to Gln incorporation at UAG [30]. In MG1655, BL21 and TOP10 strains, no suppressor tRNA is present in their genomes. Despite this fact, background amber suppression is found in all three strains. Although low in Top 10 (1%, not significantly above background), greater suppression is seen in BL21 (3%) and MG1655 (17%). In both BL21 and MG1655, therefore, using the amber codon to encode NCAAs may lead to greater contaminating incorporation of canonical amino acids compared to TOP10. In BL21(DE3) and derived strains, near-cognate incorporation of Trp, Gln, and Tyr in response to UAG has also been reported [31], and code expansion systems for phosphoserine [5] and BocK (K.A.O. & W.R., unpublished data) displayed contamination by Gln and Trp at the UAG encoded locus, respectively.

We find that, in the absence of a suppressor tRNA, UGA displays generally higher background suppression than UAG (Fig. 1). Even when background amber suppression is low (e.g., 3% in BL21(DE3)), contamination of an NCAA encoded site with natural amino acids has been observed in BL21(DE3) and derived strains [5,31] (K.A.O. & W.R.L, unpublished data). We anticipate that UGA-based code expansion systems may encounter yet greater contamination by natural amino acids than amber codon systems.

3.2 Orthogonality of PylRS/tRNA^Pyl_UCA for opal codon decoding

It was recently shown that the PylRS/tRNA^Pyl can support incorporation of BocK in response to the opal codon in E. coli BL21(DE3) (K.A.O. & W.R.L, unpublished data). Mass spectroscopic analysis revealed that the sfGFP reporter contained both Trp and BocK at the position encoded by UGA. In the absence of BocK, only the Trp containing peptide was detected, but the source of Trp read-through was not confirmed. To rule out the possibility that Trp could be ligated to tRNA^Pyl_UCA, we assayed PylRS and TrpRS in vitro with α-[³²P]-labeled tRNA substrates (Fig. 2–3). Commercially available Pyl analogs (BocK and Cyc) are used for the assays. Plateau aminoacylation by PylRS with tRNA^Pyl_UCA and either BocK (52 ±11 %) or Cyc (46 ± 3%) is similar to that with wild type tRNA^Pyl_CUA and BocK (45 ± 5 %) or Cyc (43 ± 9 %) (Table 1, Fig. 2). The E. coli TrpRS, although active with tRNA^Trp, fails to ligate Trp to any tRNA^Pyl variant (Fig. 3).

Fig. 2.

In vitro aminoacylation of tRNA^Pyl variants by M. mazei PylRS with Pyl analogs N-ε-(tert-butyloxycarbonyl)- L-lysine (BocK) or N-ε-cyclopentyloxycarbonyl-L-lysine (Cyc). An additional spot (representing < 5% of total radiolabel) between the aa-AMP and AMP is an unidentified possibly degraded reaction product, often seen in aminoacylation assays (e.g., [53]).

Fig. 3.

In vitro aminoacylation of tRNA^Pyl variants and tRNA^Trp by E. coli TrpRS with Trp.

Table 1.

Plateau aminoacylation levels.

enzyme	amino acid	tRNAPyl	tRNAPyl opal	tRNAPyl opalA73	tRNAPyl opal C73	tRNAPyl opal U73	tRNAPyl UCCU	tRNATrp	tRNAArg
PylRS	BocK	45 ± 5%	52 ± 11%	13 ± 2%	25 ± 2%	40 ± 2%	40 ± 6%
	Cyc	43 ± 9%	46 ± 3%	6 ± 3%	13 ± 4%	29 ± 3%	40 ± 7%
TrpRS	Trp	n.d.	n.d.	n.d.	n.d.	n.d.	n.d.	77 ± 7%
ArgRS	Arg	n.d.	n.d.	n.d.	n.d.	n.d.	n.d.		89 ± 1%

Values are in % of total tRNA aminoacylated at plateau level (reached before the 10 min time point) of the reaction.

n.d. – no aminoacylation detected.

We conducted in vivo and in vitro assays to further characterize the PylRS/tRNA^Pyl sys tem for opal codon decoding. E. coli BL21 (DE3) cells were transformed with empty plasmids or with two vectors (pKTS and pTECH) for independent expression of PylRS or tRNA^Pyl_UCA, respectively (Fig. 4). These experiments lack the amino acid substrate for PylRS as here we evaluate background opal suppression, possible endogenous charging of tRNA^Pyl_UCA or canonical amino acid charging by PylRS. We observe an opal codon background suppression of 10% in BL21, and addition of tRNA^Pyl_UCA (11%) or the PylRS (13%) does not significantly enhance suppression. The tRNA^Pyl_UCA is, therefore, not a substrate for endogenous E. coli synthetases. Addition of both tRNA^Pyl_UCA and PylRS (19%) appears to marginally increase activity, but this is within the error of the background measurement for opal suppression.

Fig. 4.

Opal codon background suppression in E. coli BL21(DE3) is not significantly increased by the addition of tRNA^Pyl or PylRS.

Base 73 of the tRNA^Pyl_UCA, the so-called discriminator base, was mutated recently (K.A.O. & W.R.L, unpublished data) to find a tRNA^Pyl variant better able to compete with Trp-tRNA^Trp read-through of UGA. In vivo, these mutants did not significantly affect read-though of UGA, and we found in vitro that these substrates aminoacylated to a significantly lower level than tRNA^Pyl_UCA with the wild type G73 (Table 1, Fig. 2).

Taken together, the data show that background opal codon suppression in E. coli BL21 exists at a level of ~10% and this leads to Trp insertion (K.A.O. & W.R.L, unpublished data) in response to the UGA codon. It is well established [21,26,32], and demonstrated again in Fig. 4, that PylRS is not active with any of the canonical 20 amino acids, so there is no possibility of PylRS forming Trp-tRNA^Pyl or aminoacylating any tRNA^Pyl variant with natural amino acids. Since TrpRS has no activity towards tRNA^Pyl_UCA, Trp insertion likely results from near-cognate (one codon anti-codon mismatch) suppression of UGA by tRNA^Trp_CCA. PylRS/tRNA^Pyl_UCA can promote assignment of UGA to encode a NCAA, such as BocK, but BocK-tRNA^Pyl_UCA fails to completely outcompete background suppression by Trp-tRNA^Trp.

3.3 Near-cognate suppression outcompetes BocK-tRNA^Pyl_UCCU for quadruplet codon decoding

Since quadruplet codons could provide many new openings for genetic code expansion, we characterized PylRS/tRNA^Pyl_UCCU for its capacity to decode the 4-base codon AGGA. We used a sfGFP reporter with AGGA encoded at position N134. BL21(DE3) cells transformed with pETtrio-pylT(UCCU)-PylRS-sfGFP134AGGA showed significant and similar sfGFP expression levels both in the absence (122 mg/L culture) and in the presence of 5 mM BocK (138 mg/L culture). Although we suspected BocK should be incorporated into the sfGFP reporter to some extent, the ESI-MS analysis only detected Arg at position 134 (Table 2, Fig. 5). Addition of BocK to the medium did not change the ESI-MS patterns. Interestingly, we found about half as much background AGGA suppression in TOP10 cells transformed with this sfGFP reporter (70 mg/L compared to ~130 mg/L in BL21(DE3)).

Table 2.

ESI-MS characterization of sfGFP.

Protein	Calculated mass /Da	Detected mass /Da
sfGFPN134AGGAa	27867c	27868
	27735d	27737
sfGFPN134AGGAb	27867c	27867
	27735d	27736

^asfGFP expressed in cells with pETtrio-pylT(UCCU)-PylRS-sfGFP134AGGA, 1 mM IPTG.

^bsfGFP expressed in cells with pETtrio-pylT(UCCU)-PylRS-sfGFP134AGGA, 1 mM IPTG and 5 mM BocK.

^cFull length sfGFP proteins.

^dFull length sfGFP proteins without N-terminal Met1.

Fig. 5.

Quadruplet codon AGGA encodes arginine in vivo in E. coli. (A) Protein produced by BL21(DE3) cells transformed with pETtrio-pylT(UCCU)-sfGFP134AGGA, grown with or without BocK. (B) ESI-MS of sfGFP produced by cells grown with (B) or without (C) BocK.

We conducted in vitro aminoacylation assays to determine if tRNA^Pyl_UCCU is a competent substrate for PylRS (Fig. 2). With BocK, aminoacylation reached a plateau of 40 ± 6% for tRNA^Pyl_UCCU, compared to 45 ± 5% for tRNA^Pyl_CUA (the wild type substrate). The aminoacylation level for tRNA^Pyl_UCCU should be similar to amber and opal decoding tRNA^Pyl variants that support BocK incorporation in vivo [33] (K.A.O. & W.R.L unpublished data). To test if misacylation might contribute to the signal seen in Fig. 5, we assayed tRNA^Pyl variants for Arg accepting activity. Although, the E. coli ArgRS is highly active with its native substrate, no tRNA^Pyl variants could be aminoacylated with Arg (Fig. 6). In summary, the data indicate that the PylRS/tRNA^Pyl_UCCU system, although competent in BocK aminoacylation, cannot outcompete AGGA suppression by E. coli’s native Arg-tRNA^Arg.

Fig. 6.

In vitro aminoacylation of tRNA^Pyl variants and tRNA^Arg by E. coli ArgRS.

4. Discussion

In the experiments above, we showed that amber (5–10%) and opal codon (10–25%) background suppression vary in different commonly used laboratory E. coli strains, and this will likely lead to mis-incorporation of natural amino acid in many of the currently available orthogonal translation systems. Background amino acid incorporation at amber and opal codons has not been extensively described or quantified, but in certain cases mass spectrometry has detected the presence of TAG encoding both the desired NCAA and contaminant natural amino acid (e.g., [5]). Incorporation of BocK in response to UGA by PylRS/tRNA^Pyl is accompanied by significant Trp incorporation (K.A.O. & W.R.L unpublished data). It is well known that the ribosome is most permissive to codon anti-codon mismatch at the third nucleotide of the codon, so near-cognate suppression of UGA by Trp-tRNA^Trp_CCA is not unexpected [34–37].

Several engineered E. coli strains survive without release factor 1 (RF1), which is responsible for stimulating ribosome dissociation when the amber codon is encountered in the RNA message [38]. RF1 deletion leads to extension of natural E. coli proteins beyond their normal TAG stop codon, and some of these extensions are detrimental to cellular fitness [39,40]. These defective phenotypes can be alleviated either by compensating mutations in RF2 (possibly allowing RF2 to resolve ribosomes stalled at UAG) [31,41–43], or by reassignment of at least 7 critical TAG stop codons to TAA [39]. These strains are of great utility for recoding UAG to a sense codon because the RF1 deletion strains provide a higher level of UAG read-through [39,43], but background suppression has not been thoroughly characterized in these strains. In a ΔRF1 E. coli strain with a minimized genome (lacking ~500 genes), Johnson et al. [31] reported background suppression of UAG with Trp, Tyr, and Gln by single-base mismatch near-cognate suppression. In the presence of an evolved TyrRS/tRNA^Tyr orthogonal pair, incorporation of p-acetyl- L-phenylalanine was observed in the mass spectrum, but peptides resulting from background suppression with Trp, Tyr or Gln were not detected. They also observe a marked increase in background suppression upon RF1 deletion [31]. It is likely that for some NCAAs, expression of protein in a RF1 deletion strain will lead to higher rates of NCAA incorporation, but enhanced near-cognate suppression could yield greater natural amino acid contamination.

We found the situation is exacerbated when considering quadruplet codons. The background suppression of AGGA by Arg-tRNA^Arg is so efficient that it completely outcompetes AGGA reading by BocK-tRNA^Pyl_UCCU, even though we found tRNA^Pyl_UCCU is a competent substrate for PylRS. To address generally inefficient quadruplet decoding of the E. coli ribosome, Neumann et al. evolved an orthogonal ribosome for increased suppression of AGGA [7]. Using a TyrRS/tRNA^Tyr orthogonal pair, enhanced p-azido-L-phenylalanine (AzF) incorporation by the evolved ribosome was confirmed by mass spectrometry. The evolved ribosome allowed significantly higher background insertion of (presumably) natural amino acid when AzF was absent, but the level of natural amino acid contamination at the AGGA encoded locus in the presence of the orthogonal pair and AzF was not quantified.

Quadruplet decoding is still not completely understood (reviewed in [44,45]). Some data favor a 3-base decoding model followed by ribosomal slipping to the +1 frame, but almost all 4-base suppressors do have extended anticodon loops. In many 4-base decoding systems, establishing 4 W-C pairs is not essential for suppression activity, but it enhances suppression in certain cases. For AGGA, both Arg-tRNA^Arg_CCU and Bock-tRNA^Pyl_UCCU could make at least 3 W-C base pairs in the decoding center, and it may be that the ribosome is showing its natural tendency to favor Arg-tRNA^Arg_CCU over any other aa-tRNAs for AGG decoding and then ‘slipping’ to the +1 frame. Although quadruplet codons potentially allow incorporation of vastly more NCAAs simultaneously, competition with natural aminoacyl-tRNA will be a significant challenge. Designing efficient and accurate 4-base decoding systems is still an important direction for future work.

Current genetic code expansion technology is generally unable to completely eliminate contamination by natural amino acids. Fascinatingly, certain organisms can even derive a selective advantage from the production of such ‘statistical proteins’ [46]. For many structural, functional, and enzymological studies of NCAA-containing proteins, it is critical to produce homogenous product. Hyper-accurate ribosomes, with mutations in rpsL, may limit near-cognate suppression of amber and opal codons [47,48], providing a valuable genetic background for future studies.

The molecular mechanism of tRNA^Pyl decoding, in Methanosarcina or in the heterologous E. coli background, is not yet characterized. It is, therefore, not obvious why tRNA^Pyl does not compete more effectively against near-cognate suppression at UGA and AGGA. Since EF-Tu recognizes tRNA^Pyl as a ‘normal’ tRNA [49] and EF-Tu does not contact the anticodon, incompatibility with EF-Tu cannot explain the differential levels of near-cognate suppression permitted by amber, opal, and 4-base decoding tRNA^Pyl variants. Either there is an insufficient concentration of aminoacyl-tRNA^Pyl in the cell, or perhaps tRNA^Pyl is not optimally suited to decoding and translocation on the E. coli ribosome. In support of the latter view, we observed more efficient translation of Pyl-tRNA^Pyl in Methanosarcina acetivorans cells compared to E. coli, which shows significant levels of premature termination at the Pyl-encoding UAG [50,51]. Although mechanistic details have also not been established for other orthogonal pairs, it is already clear that for many orthogonal translation systems improved aminoacylation efficiency [52], or compatibility with EF-Tu [5] will be required to synthesize and deliver sufficient noncanonical aminoacyl-tRNA to the ribosome to outcompete near-cognate suppression by natural amino acids.

Highlights.

• Nonsense suppression by natural amino acids hinders genetic code expansion.

• Higher levels of near-cognate suppression found in opal versus amber codons.

• E. coli BL21 and MG1655 display higher background suppression than TOP10.

• ylRS/tRNA^Pyl_UCA cannot outcompete Trp incorporation at opal codons.

• PylRS/tRNA^Pyl_UCCU is outcompeted by Arg-tRNA^Arg in quadruplet decoding.