Research Article pubs.acs.org/synthbio
Evolutionary Design of Choline-Inducible and -Repressible T7-Based Induction Systems Kohei Ike,† Yusuke Arasawa,† Satoshi Koizumi,‡ Satoshi Mihashi,‡ Shigeko Kawai-Noma,† Kyoichi Saito,† and Daisuke Umeno*,†,§ †
Department of Applied Chemistry and Biotechnology, Faculty of Engineering, Chiba University, 1-33 Yayoi-Cyo, Inage-ku, Chiba 263-8522, Japan ‡ Technology Development & Research Department, Kyowa Hakko Bio Co., Ltd., 1-6-1, Ohtemachi, Chiyoda-ku, Tokyo 100-8185, Japan § Precursory Research for Embryonic Science and Technology (PRESTO), Japan Science and Technology Agency (JST), 4-1-8 Honcho, Kawaguchi, Saitama 332-0012, Japan S Supporting Information *
ABSTRACT: By assembly and evolutionary engineering of T7-phagebased transcriptional switches made from endogenous components of the bet operon on the Escherichia coli chromosome, genetic switches inducible by choline, a safe and inexpensive compound, were constructed. The functional plasticity of the BetI repressor was revealed by rapid and highfrequency identification of functional variants with various properties, including those with high stringency, high maximum expression level, and reversed phenotypes, from a pool of BetI mutants. The plasmid expression of BetI mutants resulted in the choline-inducible (Bet-ON) or choline-repressible (Bet-OFF) switching of genes under the pT7/betO sequence at unprecedentedly high levels, while keeping the minimal leaky expression in uninduced conditions. KEYWORDS: betI repressor, synthetic biology, self-cloning, genetic switch, directed evolution
I
proteins and valuable chemicals. However, these regulatory systems are deeply embedded in central metabolic networks and are constrained to be under the strong influence of cellular physiology. Their inducers (sugars) may be relatively cheap but are quickly metabolized. Synthetic analogues, such as isopropyl β-D-thiogalactoside (IPTG), are resistant to decomposition and therefore allow long-lasting and robust control over the genes of interest; however, their unit prices are unrealistically high for commercial bioprocesses.14,15 In a quest for genetic switches induced by inexpensive and safe compounds, cumate-16 and NaCl-17 responsive systems have been explored. In the present work, we set the goal of developing a new class of genetic switches that are inducible with other safe and cheap compounds that can be widely applied in biomanufacturing processes in Escherichia coli (E. coli). In search of an ideal motif, we limited our search to endogenous components (those encoded on the chromosome of E. coli). In many countries, genetically engineered microorganisms created using genetic materials derived from organisms between which natural gene transfer or exchange is possible are categorized as “self-cloning” and exempted from the requirements of the Cartagena Protocol on Biosafety as long as they are used under contained conditions (Council Directive 2001/18/EC).
nduction systems are indispensable components for the bacterial production of recombinant proteins and as tools for metabolic engineering and synthetic biology. Because whichever products are ultimately toxic to the cell upon their hyperproduction, gene expression is preferred to be tightly regulated.1 Despite awareness of the utility of induction systems,2−4 most existing systems are not suitable for practical industrial bioprocesses, particularly for the large-scale production of biocommodities. First, they must be induced in safe and inexpensive ways. Researchers have developed chemical-free induction systems, such as light- and temperature-inducible systems;5−9 however, they require specially made bioreactor setups. If chemical inducers are to be used, they must be safe, low in price, and with excellent disposition in cell. They must be efficiently transported into the cell, exhibiting reasonable half-lives to ensure induction states of hours or days. Second, the induction systems must meet strict specification requirements such as high stringency, high maximum expression level, and proper inducer sensitivity and selectivity, depending on the working context. Because all natural systems have switching thresholds that are adapted to their natural contexts, a few natural genetic switches behave appropriately in highly synthetic contexts without engineering efforts. Sugar-based induction systems, such as lactose-,10,11 arabinose-,12 and rhamnose-inducible13 systems, have been widely used for the large-scale production of pharmaceutical © XXXX American Chemical Society
Received: June 7, 2015
A
DOI: 10.1021/acssynbio.5b00107 ACS Synth. Biol. XXXX, XXX, XXX−XXX
Research Article
ACS Synthetic Biology
inserted betO downstream of the +2 position relative to the pT7 transcription initiation site, resulting in pT7/betO (Figure 1A). It should be noted that this pT7/betO was placed in the plasmid in multiple copies to overload the repression capacity of the chromosomally encoded BetI.
Among the 336 entries described as transcriptional control systems in EcoCyc,18 the bet operon drew our particular attention. The bet operon is induced by choline, which is an essential nutrient for all animals and thus highly friendly to the production of food additives.19 On the basis of cost-effective synthetic routes already established,20 or possibly by the extraction from abundant sources such as egg yolk, this safe inducer can be obtained at a very competitive unit price. Considering the above advantages, we speculated that bet-based transcriptional switches could be promising for cost-efficient and scalable bioprocesses, if appropriately engineered. In E. coli, the bet operon comprises four structural genes: betT (choline importer), betA (choline dehydrogenase), betB (betaine aldehyde dehydrogenase), and a regulatory gene: betI (transcriptional repressor). The expression of bet genes is negatively regulated by BetI through binding at or near the promoter region,21,22 and its repression is canceled upon complex formation with choline,23 thereby inducing the bet genes. Other than that, however, little is known about how bet operon is operated or about the physiological dynamics of choline in E. coli. The capacity or titer of choline uptake, halflife of choline in the cell, and even the intercellular choline concentration have not been clearly described in the literature. In addition, the bet operon appears to be under the influence of oxygen availability and osmotic conditions21−25 and probably of other unknown factors that participate in the complex osmoresponsive network in E. coli. Another concern was crosstalk. All the bet components (betI and betO) we wanted to use are virtually identical to (and thus indistinguishable from) the original copies on the chromosome. In the present study, we present the fast-track construction of functional choline-inducible switches made of endogenous parts that are constrained to crosstalk with endogenous transcriptional networks per se. Given this unique situation, we prioritized the stoichiometric dominance of the components of our induction system, rather than searching for the proper balance of the expression levels of the components. Not surprisingly, a prototype switch was nonfunctional. After a round of directed evolution, however, we quickly obtained highly efficient, highly stringent choline-inducible and -repressible variants. The resulting Bet-ON/Bet-OFF switches exhibited equal/higher stringency and higher induction level than a pET15b-based induction system. Because all these systems were fully functional in the presence of an intact bet operon on the chromosome, they should be of great use in various strains and in various contexts.
Figure 1. Design and characterization of a prototype BetI-pT7/betO system. (A) Construct design. Repressor BetI is constantly expressed by a pL promoter and represses T7RNAP-dependent transcription by binding to the betO locus of pT7/betO. T7RNAP is expressed by the chromosome of E. coli BL21-AI. BetI derepression is induced by choline. Note that BetI is also expressed by the chromosomal bet operon, which is embedded in the complex regulatory network influenced by choline, osmotic stress, oxygen level, temperature,21−25 level of global regulator Cra,37 and probably other unknown factors. (B) Expression profile of BetI-pT7/betO system as a function of choline chloride concentration with (filled circles) or without (open circles) the plasmid expression of wild-type BetI. The mean values of relative fluorescence intensity per cell, calculated from three individual experiments, were measured with various concentrations of choline chloride added to the medium. Error bars represent standard deviations (SD).
■
RESULTS Construction of pT7/betO Hybrid Promoter. A bacteriophage T7 RNA polymerase/promoter system (T7 system) is commonly used for achieving the high expression of recombinant proteins in E. coli.26,27 Expression strains such as BL21(DE3)28 and BL21-AI carrying the gene encoding T7RNAP on the chromosome have been developed. Given that E. coli is the natural host for bacteriophage T7, T7RNAPencoding strains still fall within the category of self-cloning. A chemically inducible T7 promoter (pT7) can be constructed simply by the insertion of an operator sequence downstream of the transcription initiation site.29−32 The operator sequence for BetI (betO) is well-defined.18,21 We synthesized a 20-nucleotide sequence (5′-TTAA TTGA ACGT TCAA TTAA-3′) annotated as betO in EcoCyc,18 as the component of our system. Given the above information, we
To measure the expression level, the reporter gene (of either superfolder gf p (sfgf p)33 or gf puv34) was placed under pT7/ betO. To prevent unnecessary limiting the transcription initiation capacity by the toxic effect of overexpression of the reporter gene (GFP), the RBS sequence for sfgf p was adjusted to be relatively low (with RBS score 685 in an RBS calculator35) (Figure S1). Upon transformation of the plasmid pT7/betO-685sfgfp, BL21-AI cells exhibited bright fluorescence irrespective of the choline concentration in the medium (Figure 1B, open circles). From this “always-ON” phenotype, we concluded that the level of endogenously (chromosomally) expressed BetI was well short of covering all plasmid-borne betO loci on the plasmids (∼20 copies). This phenomenon is B
DOI: 10.1021/acssynbio.5b00107 ACS Synth. Biol. XXXX, XXX, XXX−XXX
Research Article
ACS Synthetic Biology frequently seen, for example in lac repressor-mediated regulation.11,36 At very high choline concentrations (>20 mM), we observed a slight increase in relative fluorescence per cell (Figure 1B, open circles). This observation seemed unique to the gene under betO: fluorescence per cell of BL21-AI harboring pT7/ lacO-685sfgf p stayed constant even at high choline concentration (Figure S2C, gray circles). We do not know whether this increase was due to the choline-induced derepression of endogenous BetI. At and above this range of concentration, choline exerted increasingly negative effects on cell growth (Figures S2A and S3). Effect of Plasmid (Over-)Expression of BetI on Behavior of pT7/betO. We constructed a p15A-based vector for the constitutive expression of BetI. To ensure high repression capacity, the reading frame of BetI was placed under the control of a strong promoter, lambda pL, and with a highly efficient translation initiation site (RBS score 72,659, with an RBS calculator35). This plasmid (pAC-pL-betI) was cotransformed into BL21-AI with pT7/betO-685sfgf p. The resulting transformant cells were subjected to fluorescence analysis in various concentrations of choline. This time, no virtual fluorescence was observed from the transformant in all choline concentration (Figure 1B, filled circles). This clearly indicates the level of BetI is high enough to tightly suppress all pT7/betO with negligible (undetectable) leakage (Table S1). However, the repression of pT7/betO could not be canceled even at the maximum feedable concentration of choline. To establish choline-inducible pT7/ betO systems that function appropriately, we elected to reduce the repression without ruining its stringency (with minimal leaky expression under uninduced conditions). Screening for Choline-Responsive Mutants. In theory, too-tight repression could be relaxed by lowering either the effective concentration of BetI or the affinity between BetI and betO. To keep plasmid-borne BetIs dominant in the cell (over chromosome-encoded ones), we decided not to modify the expression level of BetI. We accordingly constructed a gene library of BetI using error-prone PCR mutagenesis.38 The resulting library (pAC-pL-[betI]LIB, library size, 2 × 105) was transformed into BL21-AI harboring pT7/betO-gf puv and incubated on LB agar plates containing 0−100 mM choline chloride and 0.2% (w/v) L-arabinose (for expression of T7RNAP) and screened for GFPuv fluorescence (Figure 2). Approximately 40% of the clones showed fluorescence on LB agar without choline, representing variants inactivated by deleterious mutations. The fraction of fluorescent clones was slightly higher (ca. 45%) in the presence of choline, suggesting that an appreciable fraction of the variants (ca. 5%) could be choline-induced. From agar plates containing 0.1 and 1 mM choline, 135 fluorescent clones were picked and screened for switching properties by stamping onto LB agar plates containing 0−100 mM choline using a 96-pin replicator. The majority of the clones exhibited GFP fluorescence also in the absence of choline, indicating that they were nonfunctional. We found 9 clones that were nonfluorescent in the absence but fluorescent in the presence of choline (choline-inducible clones). Interestingly, we found six clones with “reversed” phenotypes: they were fluorescent in the absence but nonfluorescent in the presence of >1 mM choline. Choline-Inducible BetI Mutants: Bet-ON System. Among the nine choline-inducible variants (named Bet-ON
Figure 2. Directed evolution of choline-responsive genetic switches. The reading frame of BetI was randomized by error-prone PCR and inserted back into the regulator plasmid. The resulting BetI library plasmids were introduced into BL21-AI harboring a screening plasmid and subjected to fluorescence-based screening in varying concentrations of choline, from which choline-induced (Bet-ON) and choline-repressed (Bet-OFF) switches were identified. For details of the construction of library and screening plasmids, see Figure 1A.
mutants after the Tet-ON system39), we selected five representative mutants and analyzed their transfer function to choline (Figure 3A and Table 1). All variants exhibited increases in cellular fluorescence with increasing choline concentration. We found variation in the maximum expression level (from high to low, BetImut1 > BetImut2 > BetImut3 > BetImut4) and in stringency (leaky expression without choline induction: from high to low, BetImut2 > BetImut1 > BetImut3 > BetImut4) among the mutants. Notably, the fluorescence level of the cell harboring BetImut1 was undistinguishable from those not harboring plasmid-borne BetI (open circle) on addition of 10 mM choline. Thus, BetImut1 could completely adopt the unbound state upon choline addition, restoring maximum promoter activity. BetI is known to function as a homodimer,23 but the variants did not exhibit the typical transfer curves with sigmoidal shape. This observation was partially because of the cytotoxic effect above high (>10 mM) concentrations of choline. By fitting the fluorescence values normalized to the cell to the sigmoidal curve, we could obtain the half-maximal effective concentration (EC50) of choline for BetImut1 (4.1 mM: see Table 1). We could not obtain either EC50 values or Hill constants for other mutants. Sequence analysis revealed that all these Bet-ON mutants carried mutations in N-terminal domains (Table 1). Mapping onto the predicted BetI structure (Figure 3B), which was built based on the putative transcriptional regulator from Pseudomonas aeruginosa PAO1 crystal structures (PDB ID, 3E7Q) using the SWISS-MODEL server,40 showed that these mutations are spread throughout the N-terminal, DNA-binding domain. Together with the relatively high frequency of occurrence of Bet-ON phenotypes in the library pool, and with the slight decrease in stringency, mutations found in N-terminal domains are likely to have decreased the affinity between betO and choline-bound BetI, thereby restoring the choline-induced depression of BetI expressed from highly active promoters on multicopy plasmids. C
DOI: 10.1021/acssynbio.5b00107 ACS Synth. Biol. XXXX, XXX, XXX−XXX
Research Article
ACS Synthetic Biology
Choline-Suppressed BetI Mutants (BetIrev): Bet-OFF System. Among the six mutants with reversed phenotypes (named BetIrev or Bet-OFF mutants), five were further analyzed with respect to their transfer function to choline (Figure 4A). All BetIrev mutants showed gradual decreases in GFP fluorescence with increasing choline concentration. All BetIrev mutants showed high stringency in the presence of >10 mM choline; GFP fluorescence was under background level. In contrast, we observed marked variation in fluorescence level under the maximum induced condition (without choline; from high to low, BetIrev1 > BetIrev2 > BetIrev3 > BetIrev4 > BetIrev5). In particular, BetIrev1 proved to be fully induced (with fluorescence reaching the level of pT7/betO-685sfgfp in the cell without plasmid-borne BetIs) while preserving perfect stringency under the uninduced condition (open circles). Sequence analysis (Table 1) showed that BetIrev3, BetIrev4, and BetIrev5 carry only one nonsynonymous mutation each (L88F, A84E, and L192P, respectively). Thus, each of these mutations alone could reverse the switching behaviors of BetI. We found multiple amino acid substitutions in BetIrev1 and BetIrev2. To identify the mutations responsible for the reversed phenotype of BetIrev1 and BetIrev2, six BetI mutants, BetIL76F, BetIT99K, BetIL183P, BetIS8P, BetIF112L, and BetIG156D, were constructed by site-directed mutagenesis42 and their transfer functions to choline chloride were analyzed (Figure S4 and Table S1). BetIL183P showed a reversed phenotype (Figure S4A), whereas BetIL76F and BetIT99K were not distinguishable from wild-type BetI. Thus, we concluded that L183P is responsible for the reversed phenotype of BetIrev1. Interestingly, the transfer function of BetIL183P was not identical to that of BetIrev1 but showed ten times lower apparent EC50 (0.7 mM) and 30% lower maximum expression level. This difference is despite the functional neutrality of both L76F and T99K alone. None of the mutations found in BetIrev2, S8P, F112L, and G156D conferred reversed phenotype to BetI alone (Figure S4B). The four mutations confirmed to reverse the choline response of BetI were distant from one another in the primary structure; however, when mapped to clusters on the predicted BetI structure, they were found to be packed in a region of the
Figure 3. Choline-inducible (Bet-ON) mutants of BetI. (A) Doseresponsive expression of pT7/betO in cells expressing BetI mutants. Relative fluorescence per cell is plotted as a function of choline chloride concentration. Each plot represents the average of three independent experiments and error bars represent standard deviations. (B) Structural mapping of mutations found in the choline-inducible variants. BetI structure was modeled using Swiss-Model server40 based on the crystal structure of putative transcriptional regulator from Pseudomonas aeruginosa PAO1 (PDB: 3E7Q). The modeled structure is displayed as a homodimer, after structure alignment with the crystal structure of the DNA/tetracycline repressor TetR complex (1QPI).41 DNA structure is taken from the crystal structure of 1QPI.
Table 1. Nucleotide and Amino Acid Substitutions in BetI Variants and GFP Expression Profile by pT7/betO in Cells Expressing BetI Mutants with or without Choline Chloride BetI
amino acid (nucleotide) change
BetIrev3 BetIrev4 BetIrev5 autofluorescence (no GFP) no BetI
− I22T (T65C) T31A (A91G) A30G (C89G) Y49C (A146G), A71V (C212T) T31A (A91G) L76F (A228C), T99K (C296A), L183P (T548C) S8P (T22C), A78A (A234G), F112L (T334C), G156D (G467A) L88F (A264C) A84E (C251A), S102S (C306T) L192P (T575C) − −
BetIwild-type BetImut1 BetImut2 BetImut3 BetImut4 BetImut5 BetIrev1 BetIrev2
fluorescence/ODa choline (−) (0 mM)
fluorescence/ODa choline (+) (100 mM)
EC50b [mM]
0.624 (0.028) 9.78 (1.7) 16.3 (3.5) 7.19 (1.1) 5.57 (0.65) − 140 (2.7) 54.7 (2.6)
0.763 (0.097) 176 (4.0) 124 (9.5) 115 (6.3) 82.3 (4.9) − 1.10 (0.20) 1.05 (0.15)
N.D. 4.1 N.D. N.D. N.D. − 9.6 3.1
23.4 (1.2) 19.9 (8.9) 2.97 (2.1) 1.11 136 (1.8)
0.989 (0.22) 1.33 (0.27) 1.60 (0.20) 1.92 188 (4.2)
0.91 0.43 0.25 N.D. N.D.
a Data were calculated from three different experiments. Each value represents the average of three measurements. The numbers in parentheses represent standard deviations. bEC50 was defined as the half-maximal effective concentration of choline chloride, at which relative fluorescence value reaches half of the maximum.
D
DOI: 10.1021/acssynbio.5b00107 ACS Synth. Biol. XXXX, XXX, XXX−XXX
Research Article
ACS Synthetic Biology
to the T7 promoter by placing lacO downstream the promoter. Probably owing to having the same design as pT7/betO with pET15b (Figure S7A), our pT7/betO also behaved as a highly stringent switch. In terms of the maximum expression level, pT7/betO was markedly different from pT7/lacO (in pET15b): the fluorescent signal of pT7/betO-685sfgf p was about 3 (for BetON) and 2-fold (Bet-OFF) of those by pET15b-685sfgf p (Figure S7B and S7C and Tables S1 and S2). It emerged that pT7/lacO of pET15b vector was partially repressed by plasmidcoded lacIQ even at a high IPTG concentration: removal of plasmid-coded from pET15b resulted in an approximately 3fold elevation in the expression level of pT7/lacO, whereas the pT7/lacO could not be turned off even in the absence of IPTG. In contrast, pT7/betO showed both high expression level and stringency with plasmid-encoded BetI mutants. Without expressing plasmid-coded regulators, pT7/betO exhibited always-ON phenotypes as pT7/lacO; however, the maximum expression level of pT7/betO was ca. 50% higher than that of pT7/lacO even in the absence of plasmid-coded lacIQ. This difference was observed despite the placement of betO operator sequences downstream of pT7 exactly like pT7/ lacO. It is known that the modification of several downstream bases (+1 to +6) near pT7 can influence promoter strength by affecting the ability of the T7RNAP to transition from the initiation complex to the elongation complex.52 Interestingly, pT7/lacO was higher in sequence similarity to the original T7 promoter than was pT7/betO. The difference in their expression level could be ascribed to the tendency of the transcript of pT7/lacO to form a more stable secondary structure than that of pT7/betO (Figure S8). Overexpression of Recombinant Proteins Using BetON/-OFF Systems. The T7 system is designed to overexpress target proteins, and some proteins could be expressed to the level of 50% of total protein upon induction.28 We tested the performance of our systems for the overproduction of sfGFP. BL21-AI harboring pT7/betO-35,174sfgf p and pAC-pL-betImut1 (Bet-ON cells) accumulated sfGFP (26.8 kDa) up to 40% of total protein 4 h after the addition of choline to the medium. BL21-AI carrying pT7/betO-35,174sfgf p and pAC-pL-betIrev1 (BetOFF cells) produced the same amount after the removal of choline chloride from the medium (Figure S9). Measurements of the amount of sfGFP in the cell pellet showed that the apparent fraction of sfGFP peaked 4 h after induction (Figure S10). After this time point, the majority of the sfGFP was released into the medium either by cell lysis or by excretion after a certain time (Figure 5A). SDS-PAGE analysis revealed that as much as 82 and 86% of the total proteins in the entire cultures (media and cell) was sfGFP for Bet-ON cells and Bet-OFF cells, respectively (Figure 5B). After removing the cell debris by centrifugation, we obtained medium containing 0.2 mg/mL of sfGFP with minimal impurity
