In vivo incorporation of multiple unnatural amino acids through nonsense and frameshift suppression

Site-speciﬁc incorporation of unnatural amino acids (UAAs) into proteins is a valuable tool for studying structure–function rela-tionships, incorporating biophysical probes, and elucidating pro-tein–protein interactions. In higher eukaryotic cells, the methodology is currently limited to incorporation of a single UAA in response to a stop codon, which is known as nonsense suppression. Frameshift suppression is a unique methodology for incorporating UAAs in response to quadruplet codons, but currently, it is mostly limited to in vitro protein translation systems. Here, we evaluate the viability of frameshift suppression in Xenopus oocytes. We demonstrate UAA incorporation by using yeast phenylalanine frameshift suppressor (YFFS) tRNAs that recognize two different quadruplet codons (CGGG and GGGU) in vivo . Suppression efﬁciency of the YFFS tRNAs increases nonlinearly with the amount of injected tRNA, suggesting a signiﬁcant competition with endogenous, triplet-recognizing tRNA. Both frameshift suppressor tRNAs are less efﬁcient than the amber suppressor tRNA THG73 ( Tetrahymena thermophila G73), which has been used extensively for UAA incorporation in Xenopus oocytes. However, the two YFFS tRNAs are more ‘‘orthogonal’’ to the Xenopus system thanTHG73,andtheyofferaviablereplacementwhensuppressing at promiscuous sites. To illustrate the potential of combining nonsense and frameshift suppression, we have site-speciﬁcally incorporated two and three UAAs simultaneously into a neuroreceptor

T he site-specific incorporation of unnatural amino acids (UAAs) into proteins biosynthetically is a powerful methodology that is seeing increasing use. The primary approach has been stop codon (nonsense) suppression using a specially designed tRNA with an anticodon that recognizes the stop codon. A wide range of in vitro translation systems has been used, along with expression in Escherichia coli and, to a lesser extent, yeast. Nonsense suppression in higher eukaryotes has, for the most part, been limited to the Xenopus oocyte, where microinjection of the required mRNA and aminoacyl tRNA is straightforward and electrophysiology provides a sensitive probe of UAA incorporation (1,2). Other experiments in higher eukaryotes have relied on the evolution of a unique tRNA and a complementary aminoacyl-tRNA synthetase (aaRS) to insert a UAA in response to the UAG or UGA stop codon, but currently, only 3-iodo-tyrosine (3), p-benzoyl-phenylalanine (4), and 5hydroxy-tryptophan (5) have been incorporated.
A remarkable variant of this approach is the use of quadruplet codons, a process that is termed frameshift suppression and was pioneered by Sisido and coworkers (6,7). The success of this approach opens up the possibility of developing multiple additional codons, thus incorporating several different UAAs into a protein.
This multiple incorporation, in turn, would enable the use of innovative biophysical approaches such as incorporating FRET pairs, structural probes such as unique cross-linking approaches, and more detailed structure-function studies.
To date, frameshift suppression in vivo has been performed only in E. coli by using a unique tRNA͞aaRS pair, and homoglutamine is the only UAA incorporated by this method. Frameshift suppres-sion was used simultaneously with nonsense suppression to incorporate two UAAs in E. coli (8). It has yet to be established whether frameshift suppression by chemically aminoacylated tRNA can be effective in vivo in general and in eukaryotic cells such as the Xenopus oocyte in particular. In fact, a previous attempt to perform frameshift suppression in Xenopus oocytes showed very poor suppression efficiency (9). Here, we show that with appropriately designed frameshift suppressor (FS) tRNAs, frameshift suppression is a viable approach to UAA incorporation in eukaryotic cells. Also, the efficiency of frameshift suppression can be substantially improved by ''masking'' the mRNA of all in-frame quadruplet sequences that match the frameshift suppression site. In particular, we describe two tRNAs with four-base anticodons that can deliver UAAs in response to the quadruplet codons CGGG and GGGU. When directly compared with an amber suppressor (AS) tRNA (THG73, Tetrahymena thermophila G73) that has been used extensively in Xenopus oocytes, the FS tRNAs are less efficient at delivering UAAs. However, both FS tRNAs are more ''orthogonal'' than THG73, producing much less incorporation of undesired natural amino acids at promiscuous sites. We also show that suppression by FS tRNAs increases nonlinearly with the amount of injected tRNA. To illustrate the potential of this methodology, we have successfully incorporated two and three different UAAs simultaneously into a neuroreceptor expressed in a Xenopus oocyte.

Testing Frameshift Suppression Viability in Vivo.
To determine whether frameshift suppression is viable in Xenopus oocytes, we chose to use a tRNA that can be aminoacylated in vivo. We selected the human serine AS (HSAS), because it is aminoacylated (with serine) in eukaryotic cells and the seryl-tRNA synthetase does not recognize the anticodon (10)(11)(12). The CUA anticodon of HSAS was replaced with CCCG and ACCC to create the human serine FSs (HSFS CCCG and HSFS ACCC ) (cloverleaf structures shown in Fig. 1A), which recognize the quadruplet codons CGGG and GGGU. Prior research showed that these four-base codons are efficient in vitro (7). Injection of wild-type muscle nicotinic acetylcholine receptor (nAChR) mRNA and either HSFS CCCG or HSFS ACCC (2.5 or 10 ng per oocyte; no amino acid ligated to the tRNA) into Xenopus oocytes resulted in no detectable channel expression. The addition of the original AS HSAS with wild-type nAChR mRNA did show channel expression with 2.5 ng of tRNA per oocyte but not with 10 ng. These results suggested that the HSFS tRNAs were causing ϩ1 frameshifts, resulting in undesirable truncation of wild-type protein and thus a lack of detectable current. Analysis of the four nAChR subunits revealed four CGGG and one GGGU in-frame quadruplet codons, which were mutated to degenerate codons (see Materials and Methods) to avoid sup-pression. We refer to the resulting mRNAs as the ''masked'' constructs. Other groups have similarly removed undesired inframe quadruplets (7,9,13). Injection of 2.5 ng per oocyte of either unligated HSFS plus the masked nAChR mRNAs resulted in functional channels with the same EC 50 as channels expressed without tRNA (data not shown). Unless otherwise noted, all subsequent experiments used such masked constructs.
To test whether a naturally occurring amino acid (serine) could be incorporated in response to a quadruplet codon, we probed a highly conserved leucine of the nAChR M2 domain, a site designated Leu-9Ј. This site is a promiscuous site in the nicotinic receptor, and replacement of the native leucine with essentially any natural amino acid produces a functional receptor, usually with a quite noticeable shift in EC 50 . In particular, prior research showed that a leucine-to-serine mutation in the ␤-subunit (␤9Ј) resulted in an Ϸ33-fold increased sensitivity to acetylcholine (ACh) (14). This site was mutated to UAG, CGGG, or GGGU. When mutant mRNA was injected into Xenopus oocytes along with 2.5 ng of unligated HSAS or HSFS tRNA, which should be aminoacylated with serine by the endogenous seryl-tRNA synthetase, significant channel expression was seen. However, the EC 50 values varied depending on the incubation time (Table 1). This finding suggested that natural amino acids other than serine were being placed at the ␤9Ј site with 2-day incubations, because the conventional mutant, ␤9ЈSer, shows no change in EC 50 ( Table 1). The variability in EC 50 between 1-and 2-day incubations suggests that the tRNAs are being modified to accept other amino acids. Modification of yeast phe-nylalanine tRNA in Xenopus oocytes has been shown to increase greatly from 1-to 2-day incubation times (15). Thus, we avoid this complication by incubation for 1 day. Amber suppression is highly efficient when the average maximal peak current (I max ) is measured at 1.25 ng of tRNA per oocyte and decreases slightly when 2.5 ng is added (Table 2). CGGG shows lower suppression than GGGU, in agreement with previous in vitro studies (7,16). CGGG suppression is highly nonlinear, with a 330% increase in current when twice as much tRNA is injected (Table 2). GGGU, however, shows an almost linear relationship, with an increase of 86% in response to doubling ( Table 2). These data suggest that HSFS ACCC is a more efficient tRNA at recognizing its cognate quadruplet codon and͞or has less competition with endogenous triplet tRNA in Xenopus oocytes than HSFS CCCG . These experiments establish that frameshift suppression is viable in Xenopus oocytes and that UAA incorporation should be feasible when using the appropriate FS tRNA.

UAA Incorporation by Frameshift Suppression.
THG73 is an AS tRNA (cloverleaf structure shown in Fig. 1 A) (17) that is used extensively for incorporating UAAs into various ion channels expressed in Xenopus oocytes (2). Initially, a FS derived from THG73 that recognizes the quadruplet codon CGGG (THG73FS CCCG ) was tested for UAA incorporation. Attempts to suppress ␤9ЈCGGG with THG73FS CCCG -L, where L was chemically aminoacylated onto the tRNA, showed no current in vivo. This result is consistent with data from Voss and coworkers (9), who saw  We then chose to screen yeast phenylalanine FS (YFFS) tRNA, which was used successfully by Sisido and colleagues (7,16) in vitro. We studied both YFFS CCCG and YFaFS ACCC (yeast phenyalanine containing acceptor stem mutations FS); Fig. 1 A shows cloverleaf structures. The latter contains acceptor stem mutations (denoted by the ''a'') incorporated to reduce glycyl-tRNA synthetase recognition (7). We first evaluated a nonpromiscuous position of the nAChR, ␣149W, an agonist-binding site tryptophan that makes a cation-interaction with ACh (18). Wild-type recovery (i.e., suppressing the ␣149 quadruplet codons with YFFS CCCG -W or YFaFS ACCC -W) resulted in functional, wild-type channels ( Table  3). To demonstrate UAA incorporation, we relied on previous work using the AS THG73 that established that 5-fluoro-tryptophan (WF1) (structure shown in Fig. 1B) incorporated at ␣149 decreased the cation-interaction and caused an Ϸ4-fold increase in EC 50 (18). YFFS CCCG -WF1 suppression at ␣149CGGG resulted in a comparable increase in EC 50 (Table 3), establishing the successful incorporation of the UAA WF1.
We next considered the previously mentioned Leu-9Ј residue. Suppression at ␤9ЈGGGU and ␦9ЈGGGU with YFaFS ACCC -Aba (where Aba is ␣-aminobutyric acid) and YFaFS ACCC -Nval (where Nval is norvaline) (UAA structures shown in Fig. 1B), respectively, resulted in reductions in EC 50 ( Table 3) that were consistent with previous studies that used the same UAAs and the AS THG73 (14). All frameshift suppression experiments had an I max between Ϫ1.6 and Ϫ4.4 A, which is more than adequate for UAA studies in vivo and should allow for the incorporation of multiple UAAs. In all cases, injection of full-length tRNA that had no amino acid attached to the 3Ј end resulted in no detectable currents in response to added ACh, directly showing a lack of aminoacylation by endogenous Xenopus aaRSs.
Masking Effects on Frameshift Suppression. Experiments with HSFS required the masking of the nAChR subunits to avoid protein truncation caused by ϩ1 frameshifts. To demonstrate the effect on UAA incorporation, suppression experiments were performed with wild-type and masked constructs. The quadruplet codon GGGU was chosen because there was only one in-frame quadruplet in the signaling sequence of the nAChR ␤-subunit and none in the ␣-, ␥-, or ␦-subunits. Wild-type recovery was performed by suppressing at ␣149GGGU with YFaFS ACCC -W and adding either wild-type or masked ␤ mRNA to the injection mixture. Table 6, which is published as supporting information on the PNAS web site, shows the dramatic effect of masking one position on frameshift suppression. With a 1:1:1:1 ratio of ␣:␤:␥:␦, the masked construct gives a 2.7-fold increase in I max relative to wild type. As the amount of ␣-subunit (which contains the suppression site) is increased, the masking effect decreases to 1.5-fold and 1.2-fold with subunit ratios of 5:1:1:1 and 10:1:1:1, respectively. Calculations that assume two equally efficient quadruplet codons reproduce this trend ( Table 6), suggesting that the ␣149GGGU and the GGGU present in the ␤-subunit have similar suppression efficiencies.

Comparison of Frameshift and Nonsense Suppression Efficiencies.
To compare frameshift and nonsense suppression, the ␣149 and ␤9Ј sites were studied in more detail. Suppression of ␣149CGGG or GGGU with 10 ng of YFFS CCCG -W or YFaFS ACCC -W resulted in 38% and 48%, respectively, of the current from 10 ng of THG73-W suppression at ␣149UAG ( Table 4). Suppression of ␤9ЈUAG with 2 ng of THG73-L resulted in the largest I max (Table 4). Suppression at ␤9ЈCGGG or GGGU with 2 ng of YFFS CCCG -L or YFaF-S ACCC -L resulted in 14% and 36%, respectively, of the current from THG73-L (Table 4). We conclude that amber suppression is more efficient than frameshift suppression, in agreement with a trend previously seen in a eukaryotic cell-free translation system (16). In particular, the suppression efficiency observed here follows the order: THG73 Ͼ YFaFS ACCC Ͼ YFFS CCCG .
Interestingly, the yield of receptors from frameshift suppression at the ␤9Ј site was substantially improved by increasing the amount of tRNA injected. Suppression with 6 ng of YFFS CCCG -L or   YFaFS ACCC -L gave dramatic increases in I max , with a percentage change of 950% and 630%, respectively (Table 4). This large change in I max in response to a modest increase in tRNA concentration implicates a competition with endogenous triplet tRNA that responds nonlinearly to the amount of injected FS tRNA. A comparable increase in the amount of injected THG73-L led to complications due to reacylation of the tRNA by endogenous aaRSs (undesired) and incorporation of natural amino acids other than leucine, an issue that is addressed in detail in the following section and in Discussion.

Comparison of Aminoacylation of Suppressor tRNA and Read-Through
of Suppression Sites. To evaluate aminoacylation in vivo, which is undesirable for any tRNA used to incorporate UAAs, the ␤9Ј site was again studied, because most amino acids produce functional receptors when substituted at this position (14). In all experiments, tRNAs that had been ligated to dinucleotide deoxyCA (dCA) but did not contain an amino acid at the 3Ј end were injected to more closely mimic the biologically active, full-length tRNA. To maximize the potential for aminoacylation by endogenous aaRSs, 2-day incubations and relatively large mRNA injections (16.5 ng) were used. Surprisingly, THG73-dCA, which has been used extensively for UAA incorporation in Xenopus oocytes, showed significant aminoacylation in vivo, with an I max of Ϫ4.8 and Ϫ8.2 A for 2 and 6 ng of tRNA, respectively ( Table 4). Note that under other conditions (less mRNA and shorter incubations), previous work has found no complications from aminoacylation using THG73-dCA in Xenopus oocytes (9,14,17). Still, the present results establish that THG73 is susceptible to aminoacylation by aaRSs, which is undesired. No aminoacylation was seen with 2 ng of THG73-L, suggesting that aminoacylation by endogenous aaRSs is more likely when nonaminoacylated THG73 is injected, as noted previously (17). Both FS tRNAs show much lower amounts of aminoacylation by aaRSs, as evidenced by the decrease in I max (Table 4). YFFS CCCG -dCA shows only 8.8% and 15% of the I max of THG73-dCA at 2 and 6 ng, respectively. The most orthogonal suppressor is YFaFS ACCC -dCA, with 1.9% and 3.3% of the I max of THG73-dCA at 2 and 6 ng, respectively. The orthogonality trend thus follows the order: YFaFS ACCC -dCA Ͼ YFFS CCCG -dCA Ͼ THG73-dCA. YFaFS ACCC is the most orthogonal and efficient FS tRNA, and it therefore offers a viable replacement for THG73, especially when aminoacylation by aaRSs poses a problem in vivo.
Read-through at the ␤9Ј site was also assessed by injection of mRNA only (Table 4). ␤9ЈUAG showed the most read-through, presumably because there is only one in-frame stop codon before desired termination. ␤9ЈCGGG and ␤9ЈGGGU show 23% and 21% read-through, respectively, relative to the UAG stop codon. This finding is consistent with the idea that an endogenous triplet tRNA recognizing the first three bases of a quadruplet codon causes a Ϫ1 frameshift, which then presents multiple stop codons. Again, we designed this experiment to enhance signals from read-through by injecting large amounts of mRNA (50 ng). No current was detectable after injection of mRNA containing UAG, CGGG, or GGGU at position ␣149, confirming that this site is much less promiscuous than ␤9Ј.
Incorporation of Two UAAs. To investigate the simultaneous incorporation of two UAAs, we again built on previous work that used THG73 to incorporate UAAs into the nAChR at positions ␣149, ␤9Ј, and ␦9Ј. Importantly, EC 50 changes associated with mutations at these sites are independent of one another (18,19), allowing one to qualitatively anticipate the consequences of multiple mutations. In particular, both ␤9ЈAba and ␦9ЈNval produce predictable reductions in EC 50 that should be reproduced when combined with mutations at ␣149 (14). That is, the previously noted 4-fold increase in EC 50 that is seen when the native tryptophan at ␣149 is replaced by WF1 should persist when in combination with ␤9ЈAba or ␦9ЈNval.
Successful incorporation of two UAAs to produce large AChinduced currents could be seen when a 5-fold excess of mutant to wild-type mRNA was used. Suppression with ␣149UAG͞ THG73-W and ␤9ЈCGGG͞YFFS CCCG -L is a wild-type recovery experiment that gave the expected EC 50 for ACh of 50 M (Table  5). Maintaining ␤9ЈCGGG͞YFFS CCCG -L but substituting ␣149UAG͞THG73-WF1 resulted in the anticipated 4-fold increase in EC 50 (Table 5) (18). For incorporation of two UAAs, ␣149UAG͞THG73-W or WF1 was combined with either ␤9ЈCGGG͞YFFS CCCG -Aba or ␦9ЈGGGU͞YFaFS ACCC -Nval (Table 5 and Fig. 2; representative traces are shown in Fig. 4, which is published as supporting information on the PNAS web site). The ␣149 WF1:W EC 50 ratios are 4.4 for the ␤9Ј and ␦9Ј mutants. These experiments establish that frameshift suppression can be combined with nonsense suppression to incorporate two UAAs in a eukaryotic system.

Incorporation of Three UAAs.
To demonstrate the incorporation of three UAAs, we combined the two-UAA incorporation experiments described above, taking advantage of the knowledge that EC 50 is lowered monotonically by appropriate 9Ј mutations at multiple subunits (20). Thus, one expects a lower EC 50 when ␤9ЈAba and ␦9ЈNval are incorporated simultaneously. Suppression of ␣149UAG:␤9ЈCGGG:␥:␦9ЈGGGU by using an mRNA ratio of 5:5:1:5 with THG73-W, YFFS CCCG -Aba, and YFaFS ACCC -Nval resulted in functional channel expression with an EC 50 of 4.5 M ACh (Fig. 3), which is lower than either of the two UAAs (Aba or Nval) incorporated separately. However, the same conditions with THG73-WF1 yielded only small currents. To obtain more expression, ␣149UAG mRNA and THG73-WF1 were initially injected, and, 24 h later, ␤9ЈCGGG:␥:␦9ЈGGGU (5:1:5) was injected with YFFS CCCG -Aba and YFaFS ACCC -Nval (final mRNA ratio of 5:5:1:5). This strategy resulted in adequate expression and an EC 50 of 19 M ACh (Fig. 3). The ratio of the EC 50 s (␣149 WF1:W) is 4.2, confirming that three UAAs were simultaneously incorporated in vivo.

Discussion
The present results establish that frameshift suppression is viable in a eukaryotic, vertebrate cell and that it can be used to incorporate multiple UAAs in a single experiment. Previous work in Xenopus oocytes found that UAA incorporation using THG73FS ACCC was inefficient, and it was proposed that either the Xenopus translational machinery was not compatible with frameshift suppression or that THG73FS ACCC was a poor template for quadruplet recognition (9). Our results support the second rationalization, and a second FS derived from THG73, THG73FS CCCG , is also not viable. It thus appears that THG73derived FS tRNAs are misfolded, not recognized by the elongation factor EF-Tu, or not accepted by other components of the translational machinery. However, frameshift suppression is viable in the Xenopus oocyte by using either HSFS or YFFS tRNAs. We find that, in Xenopus oocytes, the quadruplet GGGU is suppressed more efficiently by both HSFS ACCC and YFaFS ACCC than the corresponding CGGG͞ tRNA pairs. This difference is seen despite the fact that in Xenopus, the GGG triplet is used twice as frequently (12.9 per 1,000) as the CGG triplet (21). Frameshift suppression must compete with endogenous triplet-recognizing tRNAs. Codon usage is apparently not a perfect predictor of frameshift suppression efficiency.
We have evaluated three different tRNAs: the AS THG73 and the FSs YFFS CCCG and YFaFS ACCC . For UAA incorporation in the Xenopus oocyte, both YFFS tRNAs are less efficient than the AS THG73. This finding parallels results from earlier in vitro studies (16). Apparently, the competition between release factors and the AS tRNA is less detrimental than the competition between FS tRNAs and endogenous, triplet-recognizing tRNA. This view is supported by the rapid, nonlinear rise in suppression efficiency when the amount of YFFS tRNA is increased (Table 4). CGGGrecognizing tRNAs are more sensitive to the amount injected than GGGU-recognizing tRNAs. Increasing the amount of FS tRNA for UAA incorporation is essential to maximize suppression efficiency.
The incorporation of UAAs site-specifically into proteins requires the suppressor tRNA to be orthogonal to the endogenous aaRSs. Read-through of the suppression site or aminoacylation of the suppressor tRNA (once the chemically ligated UAA has been removed) can result in the undesired incorporation of natural amino acids at the suppression site. The two YFFS tRNAs studied here exhibit much more orthogonality than THG73 under the extreme conditions (extended incubation time and increased mRNA) used in Table 4. However, THG73 is an orthogonal suppressor tRNA to the Xenopus oocyte when used properly; THG73 has been used to successfully incorporate Ͼ100 residues at scores of sites in 20 different proteins (1,2). Even promiscuous sites, such as the ␤9ЈUAG, can be efficiently suppressed by THG73-UAA when using less tRNA, mRNA, and incubation time (14). ␤9ЈUAG injected with THG73-dCA shows no greater current than mRNA alone with similar conditions. The small current is Ͻ1% of typical UAA incorporation experiments and is caused by readthrough of the UAG codon (17). Voss and coworkers (9) found that THG73 incorporated three UAAs and Phe with efficiencies of 93.5-99.5% (determined by THG73-UAA incorporation relative to natural amino acids placed by read-through or aminoacylation of THG73-dCA) using luciferase expressed in Xenopus oocytes. The current results show that the YFFS tRNAs are even more orthogonal, and so the efficiency of UAA incorporation (relative to natural amino acids) should be greater than with THG73.
An important contributor to our ability to efficiently incorporate two and three UAAs is the masking of undesired quadruplets to prevent loss of UAA. In general, the requirement for masking of mRNA to remove undesirable quadruplet codons does complicate the frameshift suppression approach. The only previous examples of UAA incorporation in higher eukaryotes were performed by nonsense suppression (1)(2)(3)(4)(5)10). Frameshift suppression may be limited in vivo to cells that are dormant (such as Xenopus oocytes), express large quantities of the target mRNA, or come from genetically engineered organisms. Also, suppressor tRNAs may be limited to rare codons because of possible toxicity arising from undesired suppression in other proteins (22).
The combination of nonsense and frameshift suppression allows one to incorporate multiple UAAs site-specifically into proteins expressed in Xenopus oocytes. These methods are compatible with our entire library of UAAs (2, 23) and will allow for multiple UAAs to be incorporated into other ion channels for structure-function studies, cross-linking, and FRET experiments. In principle, further quadruplet codons could be used to simultaneously incorporate more than three UAAs.

Materials and Methods
Materials. All oligonucleotides were synthesized by the California Institute of Technology Biopolymer Synthesis facility or Integrated DNA Technologies (Coralville, IA) (sequences are listed in Table  7, which is published as supporting information on the PNAS web site). NotI was purchased from Roche Applied Science (Indianapolis). BamHI, EcoRI, FokI, T4 DNA ligase, and T4 RNA ligase were purchased from NEB (Beverly, MA). Kinase Max, T7 MEGAshortscript, and T7 mMessage mMachine kits were from Ambion (Austin, TX). dCA and 6-nitroveratryloxycarbonylprotected dCA-UAA were prepared as reported in refs. 14, 18, and 24. ACh chloride was purchased from Sigma-Aldrich. Gene Construction and RNA Preparation. The ␣-, ␤-, ␥-, and ␦subunits of nAChR were previously subcloned in the pAMV vector (25). All four in-frame CGGGs were mutated (shown in italics) to degenerate codons (␣182CGC, ␤23AGG, ␤402 AGG, and ␦195AGG), and one GGGT was mutated at the fourth position (␤1 AGC); these constructs are termed "masked." ␣149TAG, CGGG, GGGT; ␤9ЈTAG, CGGG, GGGT; and ␦9ЈGGGT mutations were placed on masked constructs by QuikChange sitedirected mutagenesis (Stratagene). Mutations were verified by DNA sequencing (at the California Institute of Technology Se-quencing͞Structure Analysis Facility). Template DNA was linearized with NotI and mRNA prepared with the T7 mMessage mMachine kit. mRNA was purified by using the RNeasy Mini kit (Qiagen, Valencia, CA) and quantified by absorption at 260 nm.
THG73 and HSAS in pUC19 vector were previously made (10,17). Genes for HSFS CCCG , THG73FS CCCG , and YFFS CCCG (sequence from ref. 6) with flanking EcoRI and BamHI overhangs were phosphorylated by using the Kinase Max kit, annealed, and ligated with T4 DNA ligase into EcoRI and BamHI linearized pUC19 vector as described in ref. 24 dCA and dCA-UAA Ligation to Suppressor tRNA. dCA and 6-nitroveratryloxycarbonyl-protected dCA-UAA were coupled to suppressor tRNA by using T4 RNA ligase for 30 min as described in refs. 24 and 26, desalted by using CHROMA SPIN-30 DEPC-H 2 O columns, and quantified by absorption at 260 nm. tRNA ligation efficiency was determined by MALDI mass spectrometry (26), and all tRNA dCA or dCA-UAA ligations were Ͼ75%.
Electrophysiology. Recordings employed two-electrode voltage clamp on the OpusXpress 6000A (Axon Instruments, Union City, CA). ACh was stored at Ϫ20°C as a 1 M stock, diluted in Ca 2ϩ -free ND96, and delivered to oocytes by computer-controlled perfusion system. For HSAS and HSFS experiments, the holding potential was Ϫ60 mV, and all UAA experiments were performed at either Ϫ60 or Ϫ80 mV. Dose-response data were obtained from at least nine ACh concentrations, and comparisons were tested at one drug concentration, except ␤9Ј(UAG, CGGG, or GGGU) with tRNA-L, for which two concentrations, 10 M and 1 mM, were used to check for aminoacylation (Table 4). Dose-response relations were fit to the Hill equation to determine EC 50 and the Hill coefficient (n H ). All reported values are represented as a mean Ϯ SE of the tested oocytes [number (n) is listed with each table].
Supporting Information. Masking experiments, representative traces for two UAA experiments (␣149W or WF1 and ␤9ЈAba), and oligonucleotides used in this study are detailed in Fig. 4 and Tables 6 and 7. E.A.R. is a National Science Foundation Predoctoral Fellow. This work was supported by National Institutes of Health Grants NS-34407 and NS-11756.