Abstract
The many successes of synthetic biology have come in a manner largely different from those in other engineering disciplines; in particular, without well-characterized and simplified prototyping environments to play a role analogous to wind-tunnels in aerodynamics and breadboards in electrical engineering. However, as the complexity of synthetic circuits increases, the benefits—in cost savings and design cycle time—of a more traditional engineering approach can be significant. We have recently developed an in vitro ‘breadboard’ prototyping platform based on E. coli cell extract that allows biocircuits to operate in an environment considerably simpler than but functionally similar to in vivo. The simplicity of the cell-free transcription-translation breadboard makes it a promising tool for rapid biocircuit design and testing, as well as for probing the fundamentals of gene circuit functions that are normally masked by cellular complexity. In this work we characterize the cell-free breadboard using real-time and simultaneous measurements of transcriptional and translational activities of a small set of reporter genes and a transcriptional activation cascade. We determine the effects of promoter strength, gene, and nucleoside triphosphate concentrations on biocircuit properties, and we isolate contributions of the essential components—core RNA polymerase, housekeeping sigma factor, and ribosomes—to overall performance. Importantly, we show how limits on essential resources, particularly those involved in translation steps, are manifested as reduced expression in the presence of orthogonal genes functioning as load processes.
Abbreviations
- TX
- transcription
- TL
- translation
- FP
- fluorescent protein
- MGapt
- malachite green RNA aptamer
- UTR
- untranslated region
- RNAP
- RNA polymerase
- NTP
- nucleoside triphosphate
- RBS
- ribosome binding site
Introduction
The field of synthetic biology has matured to the point where biological parts are regularly assembled into modestly complex circuits with wide-ranging applications [1]. Unfortunately, the development of new biological circuits typically involves long and costly ad hoc design cycles characterized by trial-and-error and lacking the prototyping stage essential to other engineering disciplines. More often than not, designed circuits fail to operate as expected. The reason for these failures is in many cases related to context: the poorly characterized environment in which the system is operating [2–4]. This includes the finite and variable (from cell to cell, condition to condition, and time to time) pools of biomolecular resources such as transcription/translation machinery and nucleoside triphosphates (NTPs), weak control over the component DNA concentrations, unpredicted interactions between components and circuits and their cellular hosts (see, e.g., [5,6]), and any number of other system properties with unknown or unknowable effects.
We have recently developed an in vitro biomolecular ‘breadboard’ based on E. coli cell extract that provides a functional environment similar to in vivo but with significantly reduced complexity [7,8]. DNA and mRNA endogenous to the cells is eliminated during extract preparation, so that transcription-translation circuits of interest may be operated in isolation without interference by a cellular host. The cell-free breadboard also allows for a degree of control over reaction conditions and component concentrations that cannot be achieved in vivo. As a prototyping platform, the cell-free breadboard provides for a considerable reduction in circuit design cycle time, not only because of its relative simplicity when compared with in vivo, but also because it eliminates much of the lengthy cloning and cell transformation steps typically required in biocircuit development (see, e.g., [9]). Indeed, cell-free applications for synthetic biology are quickly expanding [10, 11]. But beyond its potential as an improved circuit development platform, our cell-free breadboard has another significant advantage: its simplicity reveals important details of biocircuit operation normally masked by cellular complexity.
In this work we show a detailed and quantitative characterization of the cell-free breadboard—an essential precursor to any biocircuit development and testing application—and explore a number of fundamental aspects of biocircuit operation not easily studied in vivo. Central to our work is the use of a novel reporter that combines the malachite green RNA aptamer and a fluorescent protein for a real-time and simultaneous read-out of the system’s transcription and translation activity. We establish the functional implications of intrinsic biocircuit properties such as component concentration and promoter strength, as well as those of the extrinsic biomolecular resource pool that includes nucleoside triphosphates (NTPs), sigma factors, and other transcription/translation machinery. Importantly, we show how limits on essential resources, particularly those involved at the translational level, manifest themselves in the form of reduced expression and ‘crosstalk’ between orthogonal genes. Implications for biocircuit prototyping are discussed.
Results
To best characterize transcription-and translation-level performance in the cell-free breadboard platform, we use a reporter construct encoding an optimized green or cyan fluorescent protein (deGFP/deCFP) along with the malachite green RNA aptamer (MGapt) in the 3′ UTR (Figure 1A). The 35-base MGapt sequence contains a binding pocket for the malachite green dye [12] and allows for real-time fluorescence monitoring of RNA dynamics with a temporal resolution significantly higher than what has been previously achieved in cell extract using radio-labeling and gel analysis (e.g., in [13]). The use of MGapt as a measure of RNA was validated using real-time PCR and by comparing deGFP levels with and without the aptamer in the 3′ UTR (Figures S1 and S2).
Constitutive gene expression under standard conditions
The deGFP-MGapt reporter placed under the control of a strong constitutive promoter, Pr (the lambda repressor Cro promoter), and strong untranslated region (UTR) serves as our baseline construct. MGapt expression curves demonstrate complex RNA dynamics that include production via transcription machinery and degradation by RNases (Figure 1B). Saturation of the DNA by the transcription machinery can be seen in the slopes of the MGapt curves during the early stages of the experiment; these early production rates are well described by a Michaelis-Menten–type function with Michaelis constant of ∼24 nM (Figure S3). The absence of a steady-state level of MGapt indicates that RNA production does not continue indefinitely. We note a distinct qualitative change in the MGapt profiles between template DNA concentrations of 2 and 5 nM, from curves characterized by relatively broad peaks and slow decays to ones that are more sharply-peaked.
Protein expression curves (Figure 1C) show simpler translation dynamics—protein degradation machinery is absent from the standard cell-free breadboard—but again we see qualitative differences above and below a 2–5 nM DNA threshold. Above, deGFP production slows continuously before stopping at ∼ 500 minutes. Below the threshold the profiles may be described as piecewise linear functions with roughly constant positive slope for times t < tend,TL ∼ 330 minutes and zero slope for t > tend,TL. Given that tend,TL appears fixed for all concentrations in this regime, it is unlikely that the cessation of protein production is due to complete consumption of necessary resources by the translation machinery. Similar results have been noted previously [14] with the suggestion that a number of other processes, including NTP hydrolysis and enzyme denaturation, may lead to early termination of protein synthesis reactions [15, 16].
We use two simple measures to better compare transcription-and translation-level performance under different conditions: MGapt integrated over the course of the experiment and the concentration of deGFP at the end of the experiment (deGFP(tend) = [deGFP]end). The choice of these performance metrics is motivated by the relationship between integrated MGapt and deGFP concentrations under ‘ideal’ conditions: in the absence of protein degradation, and under the naive assumptions of unlimited resources and conditions unchanging with time, a simple model for deGFP production may be written as for constants kTL and tmat, and thus deGFP at any time t∗ may be expected to be proportional to ∫t∗MGapt (with a short delay for deGFP maturation). This model was previously validated for up to one hour of expression [13]. Below the 2–5 nM DNA threshold described above and for times t<tend,TL we find that this proportionality continues to hold true (Figure 2A), although the relationship between deGFP(t∗) and ∫t∗MGapt becomes much less straightforward as resources are consumed and system conditions change with time (see Supporting Information). Plotting ∫MGapt and [deGFP]end as a function of plasmid concentration (Figure 2B), we see a ‘linear’ regime in which [deGFP]end is proportional to DNA concentration and a ‘saturation’ regime in which [deGFP]end versus DNA concentration is sublinear. These regimes correspond to the qualitative differences in the MGapt and deGFP expression curves described above. Surprisingly, we see no significant change in ∫MGapt at the transition between regimes.
Transcription, translation, and promoter strength
To determine how promoter strength affects transcription and translation in our cell-free breadboard, we tested the reporter construct under the control of two additional constitutive promoters Pr1 and Pr2 made weaker than Pr by single base mutations in the -35 and -10 region, respectively (see Materials and Methods). We find that the concentration at which the system transitions from the linear regime to the saturation regime is increased for these weaker promoters, up to ∼10 nM for Pr1 and ∼20 nM for Pr2 (Figure S5). Thus, we see something of a performance trade-off between DNA concentration and promoter strength: a weaker promoter allows for ‘linear’ performance with higher template concentrations.
The overall performance under different promoters can be summarized and compared in a plot of [deGFP]end versus ∫MGapt, effectively a measure of protein produced per transcript (Figure 3). Worth noting is the dramatic increase in the Pr curve at the regime transition point, an increase not seen for the weaker promoters. This may be explained by the differential transcriptional dynamics in the ‘linear’ and ‘saturation’ regimes—distributed versus peaked—coupled with decreasing activity of the translation machinery. Such a time-dependent reduction in translational efficiency, reported in other cell-free systems [14], would mean that although transcription may take place later in the experiment, the resulting transcripts are less translatable.
What is clear from Figure 3 is that the relationship between DNA template concentration, promoter strength, integrated RNA, and final protein concentration is not a simple one; for example, with 2 nM DNA, ∫MGapt produced using Pr1 and Pr2 is 40% and 15% of the Pr value, respectively, and [deGFP]end is 11% and 0.4% of that produced by Pr. With 20 nM DNA, the percentages are different: ∫MGapt produced using Pr1 and Pr2 is 70% and 32% of Pr, respectively, and [deGFP]end is 30% and 2% of Pr. We note that at this higher concentration, the strong promoter is operating in the ‘saturation’ regime, while the weaker promoters are not.
The role of NTPs
The standard platform contains the natural NTPs essential for biocircuit operation, in concentrations of 1.5 mM ATP and GTP and 0.9 mM CTP and UTP [17]. Among their many cellular functions, ATP and GTP play a crucial role in translation, and all four NTPs are used in transcription as substrates in the synthesis of RNA. NTPs thus serve to couple a biocircuit’s transcription and translation layers together, with an impact that is not intuitively obvious but that can be significant (see, e.g., [18]). As a result, understanding precisely how changes in NTP concentration affect performance is of paramount importance.
We supplemented the system with an additional 1.25 mM of each NTP, an increase of ∼83% ATP/GTP and ∼138% CTP/UTP. In the linear regime, the extra NTPs have little effect on the shapes of the MGapt and deGFP profiles, save for an increase in tend,TL to ∼450–500 minutes (Figure S6). In the saturation regime however, the MGapt curves are broadened dramatically while the deGFP curves are more compressed at late times. The overall effect may be more easily seen on a plot of [deGFP]end versus ∫MGapt (Figure 3). We find that the additional NTPs support a ∼30% increase in [deGFP]end in the linear regime, a result that we primarily attribute to the increase in tend,TL. That is, the rate of production is relatively fixed but the productive period is extended. ∫MGapt also increases at low DNA concentrations. A more surprising result is seen at high DNA concentration, where ∫MGapt increases considerably but [deGFP]end is actually reduced by up to 20%. This suggests that NTPs do in fact help at the transcription level but that those excess transcripts are not translatable, and that perhaps the resources used to produce those transcripts may have been taken at the expense of reporter protein production.
Additional ‘housekeeping’ sigma factors
The association of a sigma factor (σ) with the catalytic core RNA polymerase (RNAP) is necessary for promoter recognition and transcription initiation in bacteria [19]. Thus, both σ and core RNAP are potential bottlenecks on transcription. In the cell-free breadboard, only the E. coli ‘housekeeping’ σ70 is present at appreciable concentrations [8]. To address the possibility that σ70 is in short supply and that it introduces an additional, NTP-independent limit on transcription capacity as a result, we supplemented the system with a plasmid carrying the σ70 gene under the control of the Pr promoter and assessed the system performance. Looking at MGapt kinetics at early times (Figure 4), when resources that might otherwise be consumed in the production of σ70 are still plentiful, we see that 0.1–0.5 nM Pr-σ70 increases the level of MGapt for all reporter DNA concentrations tested relative to standard conditions. 1 nM Pr-σ70 only has a positive effect with 10 nM reporter. Taken together these results suggest that additional σ70 does in fact help initiate more transcription events, although the effect is a mild one. We note that the Pr-σ70 kinetic traces deviate from the nominal curves at ∼30-40 minutes, a time that allows for the accumulation of the additional σ70. (Further discussion of the effect of additional Pr-σ70 can be found in the Supporting Information.)
Performance of a simple transcription-translation cascade
We also investigated the effect of adding an intermediate layer of transcription and translation on our reporters. The ‘cascade’ consists of constitutively-expressed T7 RNAP under the control of Pr, Pr1, or Pr2 and the deGFP-MGapt construct downstream of a T7-specific promoter (Figure 5). T7 RNAP is convenient to use for this purpose since, unlike the native E. coli RNAP, it is a single-subunit RNAP that is easy to incorporate onto a single plasmid, and it does not compete with the core RNAP for sigma factors. This cascade circuit allows us to further determine if NTP consumption by transcription/translation processes is a limiting factor, since the additional layer of transcription and translation would lead to a reduction in output relative to constitutive expression as it consumes more of these resources more quickly. Alternatively, if expression was limited (at least in part) by a reduction in the activity of the native RNAP, the introduction of T7 RNAP may extend the lifetime of the system.
There are substantial qualitative differences in T7 cascade expression as compared with a single-stage constitutive promoter. RNA increases rapidly and exhibits a long, slow decay (see, e.g., Figure S7), and there is a ∼30-60 minute delay in protein expression (see, e.g., Figure S8). A higher reporter DNA concentration leads to a shorter delay and faster rise in expression, but the final deGFP concentration is often below the level achieved with a lower reporter concentration. This suggests a trade-off in cascadedriven protein production that may be the result of fuel consumption: if deGFP is produced more quickly, then the production appears to arrest sooner.
We find that the T7 cascade protein output is dictated by the concentration of the first-stage T7 RNAP plasmid and the identity of the promoter that drives T7 RNAP expression (Figure 5). Weaker promoters (Pr1 and Pr2) controlling T7 RNAP expression lead to a wide range of deGFP levels with only small variations in the T7 RNAP concentration, and for any fixed concentration of the T7 RNAP plasmid, adjusting the reporter concentration over an order of magnitude does not affect deGFP output appreciably. When the strong Pr promoter is used, deGFP levels saturate at a level independent of the Pr-T7 RNAP concentration while MGapt levels vary substantially. The T7 cascade thus provides for independent tuning of RNA and protein outputs. Interestingly, we find regions of overlap where cascades with high concentrations of weaker first-stage promoters behave identically to low concentrations of stronger first-stage promoters. This equivalence was not present with the one-stage simple expression and may be due to the fact that in all versions of the cascade the promoters driving deGFP are identical.
In the simple expression case we found that adding NTPs to the system led to a considerable increase in transcription in the ‘saturation’ regime, but that the excess transcripts were not translated, and moreover that the resources used to produce those transcripts may have been taken at the expense of reporter protein production. We thus set out to see how the same addition of NTPs affects output of the T7 cascade with a strong first-stage promoter. As before, we supplemented the system with an additional 1.25 mM of each NTP. The resulting kinetics can be seen in Figures S9 and S10, and compared with Figures S7 and S8. Again we see a significant increase in the transcriptional activity; peaks are taller and broadened and the differences between different PT7-deGFP-MGapt concentrations are more pronounced. And while we do not see a decrease in deGFP as we did with the Pr-deGFP-MGapt construct, there is little benefit to excess NTPs at the translational level.
Resource sharing and crosstalk between orthogonal genes
Recent work has highlighted indirect coupling between genes, or ‘crosstalk’, as a prominent side-effects of resource sharing in biocircuits [20–26]. Resources include those which are consumed during circuit operation (e.g., NTPs) as well as transcription/translation machinery that is present in limited amounts and that different parts of the circuit are forced to share. To clearly distinguish between crosstalk at the transcription stage (that may arise due to competition for RNAP or σ70) and at the translation stage (arising from, e.g., a limited ribosome pool), we assayed system performance with the reporter Pr-deCFP-MGapt and two different orthogonal loads present: (1) deGFP, driven by the same Pr promoter as the reporter, and (2) the same Pr-deGFP construct but with the ribosome binding site (RBS) deleted (Pr-ΔRBS-deGFP). Crosstalk at the transcription and translation levels appears as changes in MGapt and deCFP fluorescence, respectively; with the Pr-ΔRBS-deGFP, any crosstalk must be strictly transcriptional as there is no RBS that can sequester ribosomes away from the deCFP reporter.
We find that an increase in loading generally leads to a decrease in reporter expression, although it affects the transcription and translation levels differently (Figure 6). For example, there is a similar ∼250 nM·hr variation in ∫MGapt as load increases for all reporter concentrations. At the translation level, however, the effect is highly dependent on load and reporter concentrations: the influence of the load on [deCFP]end is small at 1 nM reporter DNA but significant at 10 nM reporter. Thus, as the reporter DNA concentration increases, and the resources needed to produce reporter protein are more in demand, the translational crosstalk becomes much more pronounced. This also highlights the maximum translation capacity of the system that limits the total amount of protein that can be produced; as [deGFP]end goes up, [deCFP]end necessarily goes down. The top right corner of the [deGFP]end–[deCFP]end plot in Figure 6 represents a translation performance regime that appears to be inaccessible.
The loading effects seen in Figure 6 suggest that translation resources may be more limiting to system performance. To confirm this result, we compare the ∫ MGapt–[deCFP]end relationships for the Pr-deGFP and Pr-ΔRBS-deGFP constructs (Figure 7). When the RBS is present, in general we see that an increase in load leads to a decrease in ∫MGapt and [deCFP]end (Figure 6, left). When the RBS is absent (Figure 6, right), for the most part the load has no effect on performance, except for at high concentrations of both load and reporter, at which point we note a decrease in [deCFP]end and increase in ∫MGapt.
Discussion
Interest in simplified in vitro environments that approximate in vivo conditions is rooted in the desire to build and better understand biological circuits without the confounding factors that exist in live cells. Key to the success of these development and testing platforms is detailed and quantitative characterization. In this work we characterized a recently-developed, cell extract–based ‘breadboard’ which we then applied to the study of how biomolecular resources are used and shared in simple biocircuits. A novel reporter construct consisting of the malachite green RNA aptamer and a fluorescent protein allowed us to monitor transcription and translation simultaneously and in real-time. The use of the malachite green RNA aptamer may offer advantage over molecular beacons and binary probes [27, 28] in terms of wider dynamic range and faster response times. An analogous strategy connecting an RNA aptamer that mimics GFP [29] with fluorescent proteins spurred interest in concurrent measurement of transcription-translation activities in vivo [30, 31]. Recent studies have indicated that MGapt is also compatible with in vivo characterization of synthetic circuits [], rendering this study relevant to in vivo studies as well as expanding the toolsets for real-time RNA monitoring with different spectral properties. Our results confirmed that transcriptional activity is a good predictor of translation-level behavior within a linear regime of DNA template concentrations for up to 6 hrs or more, beyond typical results for batch-mode cell-free reactions [14, 32]. On the other hand, the transcriptional and translational capacity of the system shows saturation dependent on different factors, adding to the consensus that there is significant value in the development and use of reliable transcription-translation reporters for concurrent mRNA and protein measurement.
In what follows we further discuss the connections between cell-free and in vivo results, and the implications of this work for more effective biocircuit prototyping.
Relevance for in vivo
While it can be expected that specific circuit behaviors will manifest themselves to different degrees in cell-free systems versus in vivo, cell-free work has significant potential for contributing to our understanding of how circuits function in living systems. One example may be found in our ‘resource competition’ assays, through which we were able to quickly and clearly observe the translation machinery serving as a significant limiting resource. While a translational bottleneck has been noted in cell-free systems previously [33], a systematic characterization of loads on the system at transcription and translation levels has not been reported. Similar ribosome loading effects have been suggested by recent theoretical work [24] and several other studies on ribosome utilization [25,34–36], despite the fact that live cells are able to produce additional translation machinery.
Other consequences of a limited ribosome pool may also be found in vivo. For example, it has been established that ribosomes protect mRNAs from the action of endonucleases [37] and that ribosome spacing is a determinant of degradation rate [38]. Thus, if demands on a system are such that the available ribosome pool is insufficient to densely cover the number of transcripts, increased endonucleolytic activity would lead to an increase in the RNA degradation rate constant. Interestingly, a sharp increase in degradation rate is precisely what is seen in the MGapt expression profiles when the system transitions from the ‘linear’ to the ‘saturation’ regime (Figure 1B).
Of course, ribosomes are not the only molecular resource that can find themselves in short supply, in our cell-free system or in vivo. It was recently shown that the E. coli ClpXP protein degradation machinery can be overloaded, leading to significant crosstalk between unrelated networks [21]. More recent theoretical work has suggested that competition for a relatively small number of RNases by a large number of RNA molecules can also lead to crosstalk [18]. Evidence for this form of crosstalk may be found in our results: in the increase in ∫MGapt that occurs when the untranslated ΔRBS-deGFP load is added in amounts higher than or comparable to the reporter (Figure 7). In this case, the load presents a large number of new targets for degradation enzymes, drawing them away from the RNA reporter and thus indirectly leading to the increase in ∫MGapt. We are currently unaware of in vivo results demonstrating crosstalk-via-RNases; however, given the ribosome loading effects seen in both cell-free and in vivo systems, it is an intriguing possibility worthy of exploration, especially for RNA-based synthetic regulatory circuits [39–41].
And there are still other examples of how the cell-free breadboard may be used to predict or confirm various effects that arise in cells due to resource limits. In a recent modeling study it was suggested that different combinations of promoter and RBS strengths can result in comparable protein output with different loads on the cellular expression machinery, and that codon usage can introduce a bottleneck that impacts the expression of other genes [42]. The degree of precise control that exists in the cell-free breadboard—for example, control over DNA concentration and known induction levels without an intervening cell membrane—makes it an ideal platform for investigating this and other related questions.
On biocircuit prototyping
Despite recent developments in standardized part libraries and rapid assembly tools (e.g., [43, 44]), synthetic biology still lacks the accepted prototyping platforms and protocols common to other engineering disciplines. For the purposes of prototyping, one particular advantage of our cell-free breadboard is the rapid testing cycle it permits: save for the initial cloning, transformation, and plasmid preparation, none of the individual assays performed in this work required the many hours of cell treatment typically needed for in vivo studies. With plasmids in hand, the time from cell-free experiment setup to first results is a matter of minutes.
Problems associated with limits on the cell-free breadboard system capacity may be mitigated when operating in regimes that yield predictable response; the ‘linear’ regime, for example, where the protein production rate is constant until a well-defined end time and the amount of protein is proportional to template DNA concentration. The ‘linear’ regime boundaries can change with promoter strength and NTP concentration, as shown in Figure 8. But even when the strongest promoter Pr is used, we find that we can achieve ∼6 hours of predictable performance, with measurable FP signal over 3 orders of magnitude of template DNA concentrations. We advocate using the ‘linear’ regime for cell-free circuit testing or other applications that require linear response. The ‘saturation’ regime may be used when maximum yield is desired but the linearity of the DNA–protein relationship is not essential.
And there are a number of ways in which limits on the capacity of the cell-free breadboard may be raised. The functional lifetime of the system, which in bulk operation is limited by unidentified mechanisms reducing the activity of the transcription/translation machinery, may be increased using dialysis membranes and vesicles, up to 16 and 100 hours, respectively [8, 45]. Reaction times may be further extended with the use of microfluidics or other continuous-flow devices, as demonstrated with other cell-free environments [32, 46]. Also, the addition of purified proteins such as T7 RNAP or σ70 could potentially support an increase in capacity at no additional cost to the system; in related work it has been shown that purified GamS protein can be added to prevent degradation of linear DNA [9]. Ideally, a combination of strategies should be employed to take maximum advantage of the cell-free breadboard. The ease with which these strategies can be employed, along with the relative simplicity of the system and the control that it offers, makes it a promising platform for synthetic biocircuit prototyping.
Materials and Methods
Cell-free system and reactions
The breadboard environment consists of a crude cytoplasmic extract from E. coli containing soluble proteins, including the entire endogenous transcription-translation machinery and mRNA and protein degradation enzymes [7,8]. Detailed instructions for extract preparation can be found in [17]. To avoid variation between different extract preparations we used the same batch of extract for all experiments. Reactions took place in 10 µl volumes at 29°C. Measurements were made in a Biotek plate reader at 3 minute intervals using excitation/emission wavelengths set at 610/650 nm (MGapt), 485/525 nm (deGFP), and 433/475 nm (deCFP). No significant toxicity was observed for typical deGFP expression experiments when up to 20 µM malachite green (MG) dye was included in the reaction; the MG dye concentration was thus fixed at 10 µM for all experiments.
Fluorescence measurements and reporters
Real-time fluorescence monitoring of mRNA dynamics was performed using the malachite green aptamer (MGapt) [12] incorporated in the 3′ UTR of the fluorescent protein reporter genes, 15 bases downstream of the stop codons. This location of MGapt insertion was chosen after a number of other possible locations were tested and found to give less accurate measures of RNA dynamics. For example, incorporation of MGapt within the 5′ UTR upstream of the RBS led to decreased RNA stability, a result that may be due to the preference of 5′ end degradation by the dominant endonuclease in E. coli, RNase E [47]. This is consistent with a recent study on the Spinach fluorescent RNA aptamer [30] in which it was reported that incorporation of the aptamer in the 3′ UTR region led to stronger fluorescence than in the 5′ UTR. It was also found that a 6-base spacing between the stop codon of deGFP and MGapt affected the protein expression level to some extent, but a 10-base and 15-base spacing showed equivalent MGapt fluorescence signal levels without affecting protein expression. The fluorescent reporter variants deGFP and deCFP were previously designed to be more translatable in the cell-free system [7]. The UTR controlling translation of deGFP (eGFP-Δ6-229) and deCFP contained the T7 gene 10 leader sequence for highly efficient translation initiation [7]. All the ORFs were terminated by T500 except for the PT7-deGFP-MGapt construct which contained T7 terminator.
Plasmids and Bacterial strains
Plasmids was created using standard cloning methods. The plasmid pBEST-Luc (Promega) was used as a template for all constructs except for the PT7-deGFP-MGapt construct which was derived from the plasmid pIVEX2.3d (Roche). The same antibiotic resistance gene was used with each plasmid to ensure that any burden on the system due to the expression of these ‘background’ proteins was the same for each construct. All plasmids used are listed in Table S1. Plasmid DNAs used in cell-free experiments were prepared using Qiagen Plasmid Midi prep kits. E. coli strains KL740 (which contains lambda repressor to control for Pr promoter) or JM109 were used. LB media with 100 µg/mL carbenicillin was used to culture cells.
Promoters Pr1 and Pr2 were each modified from Pr with a single base mutation in the -35 and -10 region, respectively. The sequences, with mutations highlighted by □, are
Preparation of pure mRNA and qRT-PCR
RNA was transcribed using a linear template PCR-amplified from pIVEX2.3d PT7-deGFP-MGapt including T7 promoter and T7 terminator region. The transcription reaction was prepared as a total volume of 100 µL with 0.1 µM linear DNA template, 20% (v/v) T7 RNA polymerase (Cellscript), 7.5 mM each NTP (Epicentre), 24 mM MgCl2 (Sigma), 10% (v/v) 10× transcription buffer, and 1% (v/v) thermostable inorganic pyrophosphatase (New England Biolabs). After an overnight incubation at 37°C, the reaction mixture was run on 1% agarose gel, RNA bands that correpond to full-length transcript were excised and eluted from gel by the Freeze-N-Squeeze column (Biorad) and resuspended in water. Concentrations of purified RNA were determined spectrophotometrically using Nanodrop.
For qRT-PCR, 1 µL samples were taken at different time points from a tube containing reaction mixture at 29°C and diluted 50-fold in water. These samples were stored at -80°C until used. Afterward the samples were further diluted to a final dilution of 1:5000. Two µL of samples were analyzed in 50 µL reactions of the Power SYBR Green RNA-to-CT 1-Step kit (Life Technologies) in the MX3005 real-time PCR machine (Agilent Technologies). Primers amplified a region of the deGFP gene closer to its 3′ end (424-597 nt) and were used at 0.3 µM concentrations. Concentrations of deGFP-MGapt RNA in the sample were determined from a standard curve of dilutions of purified mRNA in a range from 0.6 to 60 pM mRNA per PCR reaction.
Acknowledgments
The authors would like to thank Eduardo Sontag and members of the Murray Group for helpful discussions. This research is funded in part by the Gordon and Betty Moore Foundation through Grant GBMF2809 to the Caltech Programmable Molecular Technology Initiative, and by the Defense Advanced Research Projects Agency (DARPA/MTO) Living Foundries program, contract number HR0011-12-C-0065. Z.A.T. was partially supported by grants TAMOP-4.2.1-B-11/2/KMR-2011-0002, TAMOP-4.2.2./B-10/1-2010-0014 and OTKA NF 104706.
The views and conclusions contained in this document are those of the authors and should not be interpreted as representing officially policies, either expressly or implied, of the Defense Advanced Research Projects Agency or the U.S. Government.