Improvement of an Unusual Twin-Arginine Transporter Leader Peptide by a Codon-Based Randomization Approach

Abstract

Secretion of Escherichia coli penicillin acylase was improved by codon-based random mutagenesis of its signal peptide. The mutagenesis technology was applied to the gene region coding for positions Lys2 to Thr13 (N half) and Ala14 to Leu25 (C half) of the signal peptide. Protein secretion was higher in several signal peptide variants (up to fourfold with respect to the wild-type value).

Penicillin acylase (PAC) from Escherichia coli is an important enzyme for the production of semisynthetic antibiotics. PAC is synthesized in the bacterial cytoplasm as a precursor containing an N-terminal signal peptide, an α-subunit, a connector peptide and a β-subunit. Once the cytoplasmic precursor is translocated to the periplasm, the signal peptide is removed and the periplasmic precursor is processed by various autoproteolytic reactions (9) into the mature heterodimeric, secreted protein. Much work has been done to improve transcriptional (4), translational (4), and posttranslational (7, 13) steps to enhance the production of penicillin acylase. Ignatova and coworkers (8) recently demonstrated that the PAC signal peptide, which does not contain a typical twin-arginine motif, could still target the precursor protein to the twin-arginine transporter (Tat) system (3, 12). We previously reported a selection system based on penicillin G (PenG) resistance which requires a functional PAC in the bacterial periplasm (6). Exploiting this selection system, we aimed to optimize the leader peptide of PAC with the purpose of improving translocation. Optimization was performed by a codon-based mutational approach using the whole signal peptide. The results could be helpful for identifying functional residues and, most importantly, finding signal peptide variants that improve the secretion rate of this industrial enzyme.

The pac gene from E. coli ATCC 11105 was obtained by PCR amplification with oligonucleotides designed to add NdeI and XhoI restriction sites at the start and stop codons, respectively, and cloned into vector pT4Bla (11) (Km^r) to produce pT4BlaPAC (see Fig. S3 in the supplemental material). A new SacI restriction site at the positions coding for Ser3 and Ser4 of the mature enzyme was introduced by site-directed mutagenesis to produce plasmid pT4BlaPACSac (Fig. S3). For randomization, the pac gene region coding for the signal peptide was divided into two segments: the region coding for positions Lys2 to Thr13 was mutagenized under nonsaturating conditions by spiking each of the wild-type codons with a mixture of 20 codons (Table 1; see also Fig. S1), to produce the N-half library. The pac region coding for positions Ala14 to Leu25 was mutagenized similarly, but in this case a mixture of 20 anticodons was used (Table 1; see also Fig. S1) to produce the C-half library. Oligonucleotide synthesis conditions are briefly described below. Duplexes of the randomized regions were generated by extension of 500 pmol of complementary primer over 500 pmol of mutagenic oligonucleotide using the Klenow fragment of DNA polymerase I (Fig. S2).

TABLE 1.

List of trinucleotides prepared for codon and anticodon mutagenesis^a

^aList of Fmoc-protected trinucleotide phosphoramidites (FTPs) used to assemble the two mutagenic oligonucleotides reported in this study. The sense mutagenic oligonucleotide (AGG AGG CAT ATG AAA AAT CGC AAT CGT ATG ATC GTG AAC TGT GTT ACT GCT TCC CTG ATG) was assembled using an equimolar pool of the trinucleotides with entries in the “codon” column; codons in bold were subjected to partial mutagenesis. The antisense mutagenic oligonucleotide (AT CTC ACT TGA GCT CTG CTC AGC CAG TGC AGG TAA GCT CCA ATA ATA CAT CAG GGA AGC AGT AAC ACA GTT C) was assembled using an equimolar pool of the trinucleotides with entries in the “anticodon” column; codons in bold were subjected to partial mutagenesis. The basic structure of FTPs is shown above the table.

For all experiments, around 40 pmol of the double-stranded and double-digested mutagenic DNA fragment was ligated to 2 pmol of the pT4BlaPACSac cloning vector using T4 DNA ligase (Fig. S4). The recombinant plasmids were electroporated into E. coli XL1-Blue cells already transformed with the pACYC184 vector containing a 6-amino penicillanic acid (6-APA)-specific β-lactamase (6). The transformants were spread on plates containing kanamycin, chloramphenicol, and increasing amounts of PenG (concentrations ranging from 60 μg to 120 μg per ml of culture media). Plasmid DNA from bacterial colonies was purified, and the pac gene coding for the signal peptide was sequenced.

In additional experiments, designed to assess the variability of the libraries, the mutant duplexes were cloned into an empty pT4Bla vector at the NdeI and SacI restriction sites (Fig. S4). The resultant recombinant plasmids were used to transform E. coli XL1-Blue cells, and some clones were randomly selected to isolate plasmids for nucleotide sequence analysis.

For protein detection (Western) experiments, a decapeptide tag (NH-Tyr Pro Tyr Asp Val Pro Asp Tyr Ala Ser-COOH) was added to the end of the β domain of all resulting functional clones (22 clones). The decapeptide tag was added to each clone by replacing a BglII/XhoI fragment containing the final 913 nucleotides of the pac gene with a similar DNA fragment also encoding the tag (Fig. S5). The decapeptide tag is specifically recognized by the monoclonal antibody 12Ca5 (I. A. Wilson, TSRI, San Diego, CA). All final constructs were verified by DNA sequencing.

E. coli XL1-Blue was transformed by electroporation with the different expression plasmids and grown overnight with agitation at 30°C in 5 ml of LB medium containing kanamycin. Before the cells were harvested by centrifugation, the optical density at 600 nm (OD₆₀₀) of the cultures was measured and the cell pellets were resuspended in 50 mM phosphate-buffered saline (PBS) with the appropriate volume (around 500 μl) in order to normalize all samples to the same OD₆₀₀. The resuspended cell pellets were disrupted by sonication, and total and soluble (collected by centrifugation) protein fractions were taken.

Western blotting was performed according to published procedures (14). Bovine serum albumin (3%) in PBS was used to block nonspecific sites in the nitrocellulose paper. The PAC band was revealed using an anti-mouse immunoglobulin alkaline phosphatase-conjugated secondary antibody and ready-to-use alkaline phosphatase liquid substrate system for membranes from Sigma (St. Louis, MO). Finally, PAC specific activity assays using total soluble protein extracts were determined under substrate saturation conditions (PenG final concentration, 2%) by the paradimethyl amino-benzaldehyde method (1).

Codon-based random amino acid substitutions in the PAC signal peptide.

Six Fmoc-protected trinucleotide phosphoramidites with the sequences ATA, CCA, CGG, GCA, GCT, and TTC were chemically synthesized as described by Yáñez and coworkers (15). These trinucleotides and the 20 compounds previously described (15) are enough to prepare two mutagenic pools corresponding to 20 codons and 20 anticodons, as seen in Table 1. Each mixture can be used to modify DNA segments of either the coding or the noncoding strand, respectively, and still encode all natural amino acids. Fmoc-protected trinucleotide phosphoramidites are valuable, unique reagents that can be substoichiometrically incorporated during the ordinary assembly of oligonucleotides to produce libraries of mutagenic oligonucleotides containing wild-type and mutant codons interspersed. The Fmoc-trinucleotide approach eliminates codon redundancy as well as stop codons and allows the production of libraries of mutant proteins containing few amino acid replacements in a target region comprising several residues. The method has been thoroughly described elsewhere (6a).

For the present study, two mutagenic oligonucleotides were assembled to explore mutations along the complete pac signal peptide. The oligonucleotide with the sequence 5′-AGG AGG CAT ATG AAA AAT CGC AAT CGT ATG ATC GTG AAC TGT GTT ACT GCT TCC CTG ATG-3′, encoding amino acids 2 to 13, was assembled using the pool of trinucleotides corresponding to codons. The codons subjected to replacement are those in bold. The total concentration of the mutagenic pool was 4 mM, and all their components were equimolar, yielding an average mutagenesis rate of 0.15 per substituted codon. The oligonucleotide with the sequence 5′-AT CTC ACT TGA GCT CTG CTC AGC CAG TGC AGG TAA GCT CCA ATA ATA CAT CAG GGA AGC AGT AAC ACA GTT C-3′, encoding amino acids 14 to 25, was assembled using the pool of trinucleotides corresponding to anticodons. For this oligonucleotide, the codons in bold were subjected to replacement using a higher mutagenic pool concentration (8 mM) to yield on average a mutagenesis rate of 0.3 per codon. Thus, different mutagenesis rates were designed for each oligonucleotide to induce fewer amino acid changes in the N-half region and several more changes in the C-half region of the signal peptide. Finally, the wild-type pac gene coding for the signal peptide was replaced with a cassette coding for a library of mutagenized signal peptides (see Fig. S2 to S4 for details about the procedures). Libraries containing approximately 10⁵ different members were obtained.

Library diversity.

Data in Table 2 show the variability that resulted from the two different mutagenic strategies. The N-half region contained mostly nonmutagenized signal peptides (34 of 58 randomly chosen variants; data not shown). The observed mutagenesis rate for this experiment (0.047) was lower than the theoretical (0.15). In contrast, we did not find a wild-type signal peptide within the randomly chosen set of C-half variants (17 clones). The actual mutagenesis rate for this experiment (0.33) was close to the theoretical (0.3). For these experiments, the mutagenic duplexes were cloned in a vector not containing the penicillin acylase gene, such that the variability obtained is free from the known bacterial detrimental effects caused by penicillin acylase gene expression (13). It is important to note the high mutation rate that resulted in the C-half region. For instance, serine at position 15 was replaced in 11 of 17 randomly selected variants, resulting in 10 different amino acids at this position. This sequence space exploration is possible only by codon-based mutagenic methods such as the one we describe here.

TABLE 2.

Library diversity

Variant(s)	Amino acid at positiona:
	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21	22	23	24	25	26
WT	M	K	N	R	N	R	M	I	V	N	C	V	T	A	S	L	M	Y	Y	W	S	L	P	A	L	A
1, 25		C										C		G		P				I		G	S		V
2, 26			M									C		H	Q			D		H	L	A
3, 27				M	H	C	*	D			M			N	R								M	T
4, 28					H									E	*		P		N					I
5, 29						M								G	M			*			T	Y
6, 30						F									A			N	K
7, 31							R								D		A		C		W	N		I
8, 32							F			G					N		F	R							S
9, 33								S							E			A						V
10, 34								K		A			V		E				C		W			T	W
11, 35									S						P	C	*		V					E	P
12, 36									M						*	D			T			A
13, 37									Q						I	A		*					S	C	Y
14, 38										I						G				F	D		E		G
15, 39										S							Q	N
16, 40										G			F								T
17, 41											*		I									I		*
18, 42											V
19												I
20												R
21												N
22													S
23													H
24													D

^aUndesigned modifications (mostly single nucleotide deletions, not shown) occurred at positions marked with an asterisk; the segments of sequence that do correspond to the design are shown properly aligned. The mutagenized regions are the regions from position 2 to 13 and from position 14 to 25. The amino acids are designated by the single-letter nomenclature.

PAC selection system.

To identify functional signal peptides present in the libraries, we used our previously reported PAC selection system (6). Briefly, the selection system works as follows. Any functional signal peptide able to translocate PAC to the periplasm will confer resistance to PenG because the 6-APA formed as a result of hydrolysis of PenG by PAC is then degraded by a 6-APA-specific β-lactamase expressed in the bacterial cell using a compatible plasmid. Colonies are then selected on plates containing PenG. To select for improved signal peptides, we used plates containing PenG at concentrations above the level conferred by the wild-type signal peptide (around 60 μg/ml of PenG in the expression system described here).

PAC leader peptide functional variants.

From the data in Table 2 and Table 3, we show that some amino acid positions have a very low replacement frequency among the functional signal peptides. For instance, according to Table 2, leucine at position 25 has a mutation frequency of around 0.35 (6 clones out of 17 showed a replacement at position 25). The 13 functional variants described in Table 3 for the C-half region resulted from plating approximately 50,000 colonies. If the amino acid at position 25 is irrelevant for the signal peptide function, the expectation would be that 4 out of 13 functional clones will show a replacement at this position. This was clearly not the case. Similar analyses for every position suggest that positions 14, 15, and 17 to 22 are freely replaceable residues. The mutation frequency observed for the C-half library (0.33; see above) and the small number of variants rescued during selection (13 clones [Table 3]) strongly suggest that we get only a fraction of the potential functional clones. A library size of several million different variants will be required to sample all the possible single-, double-, and triple-mutant combinations in our library. In contrast, in the N-half library, the low mutation frequency (0.047; see above) and the small number of colonies that were subjected to the selection system (around 50,000 colonies) make it impossible to reach similar conclusions. However, it is interesting that the N-half functional clones accepted a higher mutational load (8 out of 9 selected clones are at least double mutants) compared with the variability observed within the nonselected group (around 60% of the clones are single mutants [Table 2]). This result indicates that this part of the signal peptide could tolerate a high substitution rate. However, more than 50% of the replacements in the N-half functional variants contained aromatic or glycine residues. Unexpectedly, one of the best signal peptides from this library contains a phenylalanine in the N region (Table 3, clone 9).

TABLE 3.

Functional variants

Variant(s)	Amino acid at positiona:
	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21	22	23	24	25	26
WT	M	K	N	R	N	R	M	I	V	N	C	V	T	A	S	L	M	Y	Y	W	S	L	P	A	L	A
1, 10											F			V							V	P		H
2, 11									R				W	I	K			F	G
3, 12							V	W		K			W	G	M		W		W		G
4, 13						V	N			G			Y		A						N		F
5, 14							L		G						I			G	A
6, 15						G							C							S
7, 16							N			G											T
8, 17				S					G	I					M		L	N	W			V
9, 18			F							G					C			N
19															V				L		V
20															A		F			Q
21														G	L	F		L		I	Q			V
22														G				T

^aThe mutagenized regions are the regions from position 2 to 13 and from position 14 to 25. The amino acids are designated by the single-letter nomenclature.

Recently, a method (called TatP) to predict twin-arginine signal peptides was described (2). TatP was designed to potentially predict variant Tat signal peptides not containing the consensus twin-arginine motif (Ser/Thr Arg Arg X Phe Leu Lys). Unfortunately, TatP was unsuccessful in identifying the wild-type signal peptide or any of the functional variants described herein as a twin-arginine signal peptide. However, we found that by replacing a pair of tyrosine residues at positions 18 and 19 with several different residues, the resultant signal peptide was considered a “potential Tat signal peptide” without a Tat motif. TatP positive prediction of these modified PAC signal peptides could be due to the fact that Tat signal peptides present a less hydrophobic central region than classical signal peptides (5). Further experimental work is needed to find if the identified “potential Tat signal peptides” are able to secrete the enzyme through the Tat system.

PAC expression and activity levels.

Due to the stringency of the selection step, we were able to isolate mainly gain-of-function signal peptide variants (Fig. 1 and 2). The three- to fourfold enhancement in the specific PAC activity shown by almost all of the C-half variants (Fig. 1) indicates that a number of substitutions in this region make a larger improvement in the PAC secretion level than mutations in the N-half region (Fig. 2). As expected, the measured enzyme activity levels compared very well with the quantified expression levels of secreted protein (as estimated by densitometric analysis of the Western experiments).

FIG. 1.

Relative PAC activities of the N-half variants. The expression level of the respective variant is shown above each bar. PAC wild-type specific activity of 30 μg of 6-APA produced per 100 μg of total soluble protein was taken as 1. Numbers inside the bars identify the variant as described in Table 3.

FIG. 2.

Relative PAC activities of the C-half variants. The expression level of the respective variant is shown above each bar. PAC wild-type specific activity of 30 μg of 6-APA produced per 100 μg of total soluble protein was taken as 1. Numbers inside the bars identify the variant as described in Table 3.

In conclusion, a codon-based mutagenesis method allowed us to make a leader peptide engineering effort to improve PAC secretion to the bacterial periplasm. The best gain-of-function leader peptides obtained included replacements in residues from the C-half region of the signal peptide.