The aim of this project is to develop Sci-Phi 29: a versatile platform to expand the repertoire of bacterial species currently used in synthetic biology. Sci-Phi 29 enables host independent replication of genetic vectors using the concept of orthogonality and ensures controllable expression of a gene of interest across different bacterial species with the use of an incoherent feed-foward loop. To achieve this goal, we investigated the bacteriophage Φ29 replication system (orthogonality module) and engineered an incoherent feed-forward loop (controllability module).
To circumvent the need to discover bacterial host specific parts and re-characterize parts for every host, we created a collection of parts. The collection provides essential tools for controllable expression of genes across different bacterial hosts. All parts in our collection have been assembled by golden gate assembly based on a hierarchical cloning system called Modular Cloning (MoClo). We extended the traditional golden gate assembly method for modular cloning of circuits in host specific vectors. The success of our design has been demonstrated by transfer of a genetic circuit (expressing GFP) between E. coli and P. putida backbones in a one-pot reaction.
Furthermore, we have successfully cloned ribozyme insulators through modular cloning (Figure 1) as described on our design page.
In addition, to establish optimal conditions for in vivo orthogonal replication using phi29 components, a series of 16 composite parts were constructed. Each of these parts were used to determine optimal concentrations of the phi29 components necessary to potentially demonstrate orthogonal replication.
In vitro characterization of the orthogonal replication parts
To verify that our constructs for orthogonal replication were functionally expressed, we characterized the four proteins needed for orthogonal replication, DNAP, TP, SSB (p5) and DSB (p6) in vitro. For expressing these constructs, we used PUREfrex® (GeneFrontier Corporation, Japan). The PURE (Protein Synthesis Using Recombinant Elements) system contains all the purified components found to be sufficient to sustain the reactions involved in protein synthesis. Specifically, we employed PUREfrex®2.0 for its high protein yield and its prolonged-expression lifetime in bulk. After cell-free gene expression of each construct (3 hours, 37°C) we verified for the presence of our protein of interest by SDS-PAGE and/or mass spectrometry. For fluorescent labeling, BODIPY-Lys-tRNALys (FluoroTectTM GreenLys, Promega) was added, this binds to the translation products at the lysine residues sites. Since there are a multitude of proteins present in the PUREfrex® system, for some of our constructs, especially DNAP and TP, it was difficult to properly identify the presence of our expressed protein via SDS-PAGE. Luckily, using mass spectrometry, we were able to determine whether these proteins were present. To perform mass spectrometry analysis, PURE system samples were first trypsin-digested and then analyzed by liquid chromatography-coupled mass spectrometry (LC-MS). All mass spectrometry data were normalized to the presence of EF-TU, an elongation factor that can be found in the same concentration in all PURE system reactions. The mass spectrometer looks for the mass of unique peptide sequences, and their elution time. The raw data and optimized parameters for the mass spectrometry method can be found here and here, respectively.
DNA Polymerase ﹀
The expression of DNAP from our construct was confirmed by mass spectrometry, as reported by Figure 2. For DNAP, the mass spectrometer screened for the following peptide sequences: ENGALGFR and LVEGSPDDYTDIK, both were detected by mass spectrometry, so DNAP expression from our construct was successful.
Figure 2: Identification by mass spectrometry of ENGALGFR peptide (left) and of LVEGSPDDYTDIK peptide (right) in the PUREfrex sample expressing DNAP.
Terminal Protein ﹀
The expression of TP from our construct was confirmed by mass spectrometry, as reported by Figure 3. For TP, the mass spectrometer screened for the following peptide sequences: IAEIER, LVDEK, ILSYLEQYR. All three peptides were detected by mass spectrometry, so TP expression from our construct was also successful.
Figure 3: Identification by mass spectrometry of IAEIER peptide (left), of LVDEK peptide (middle) and of ILSYLEQYR peptide (right) in the PUREfrex sample expressing TP.
Single Stranded Binding Protein (p5) ﹀
The p5 expression starting from our three DNA constructs was confirmed by mass spectrometry, as reported by Figure 4. For p5, the mass spectrometer screened for the following peptide sequences: IFNAQTGGGQSFK and TVAEAASDLIDLVTR. The difference in height can be attributed to the strength of the promoters, fewer peptides were measured with decreasing strength. From this data, we can conclude that p5 expression was successful.
Moreover, the expression of p5 (SSB) was identified by an 18% SDS-PAGE gel. From the 10-μL of expression reaction, 8 μL was loaded on the SDS-PAGE, while the other 2 μL was analyzed by the mass spectrometer for further confirmation. Click here for the protocols. An SDS-PAGE was carried out for the p5 protein with three different promoter strengths (Wild-Type, medium, and weak). As a negative control, one expression reaction was run in the absence of DNA. In Figure 5, the three samples with different promoter strengths and the control can be seen on the protein gel. The presence of a band (depicted with a red star) at the expected molecular weight (13,3 kDa) confirmed the presence of the expressed p5. In the control line with no gene added, a smear background of GreenLys is visible that is distinct from gene-specific bands. The other upper band could be due to contamination in the expression reaction as it is also present in the negative control.
Figure 4: Identification by mass spectrometry of FNAQTGGGQSFK peptide (left) and of TVAEAASDLIDLVTR peptide (middle) in the p5 expressed sample. Figure 5: SDS-PAGE gels of SSB (p5) after cell-free expression. Translation prodfucts were analyzed by fluorescence imaging at 488 nm of an 18% gel. The bands depicted with an upper red asterisk correspond to the protein of interest with expected molecular weight. In the negative control lane where no DNA was added (most right) no specific band can be observed.
Double Stranded Binding Protein (p6) ﹀
As for p5, the expression of p6 (DSB) was confirmed by mass spectrometry, as reported by Figure 6. For p6, the mass spectrometer screened for the following peptide sequences: GEPVQVVSVEPNTEVYELPVEK and FLEVATVR. The difference in height can be attributed to the strength of the promoters, fewer peptides were measured with decreasing promoter strength.
Furthermore, the expression was confirmed by an 18% SDS-PAGE gel and mass spectrometry. Click here for the protocols. An SDS-PAGE was carried out for the three p6 encoding constructs with the three different promoter strengths (Wild-Type, medium, and weak). As a negative control, one expression reaction was run in the absence of DNA. In Figure 7, the three samples and the control can be seen on the protein gel. The presence of a band (depicted with a red star) at the expected molecular weight (12 kDa) confirmed the presence of the expressed p6. In the control line with no gene added, a smear background of GreenLys is visible that is distinct from gene-specific bands. The other upper band could be due to contamination in the expression reaction as it is also present in the negative control. This data confirms that we were able to express p6 successfully.
Figure 6: Identification by mass spectrometry of FLEVATVR peptide (left) and of GEPVQVVSVEPNTEVYELPVEK peptide (right) in the p6 expressed sample. Figure 7: SDS-PAGE gels of p6 (DSB) after cell-free expression. Translation products were analyzed by fluorescence imaging at 488 nm of an 18% agarose gel. The bands depicted with an upper red asterisk correspond to the protein of interest with the expected molecular weight. In the control lane where no DNA was added (most right) no specific band was observed.
We were able to successfully identify DNAP, TP, p5 and p6 after in vitro expression of our constructs. The results indicate that our constructs are correct and can lead to the expression of the protein of interest. Furthermore, we were able to observe a difference in expression level when using different promoter strengths, for example, a lower amount of protein was produced when using the weak promoter.
For this module we also wanted to demonstrate in vitro replication of our linear construct (OriL-GFP-Kan-OriR), to show that our construct can be propagated by the phi29 replication machinery.
Results -- In Vitro Replication ﹀
For this experiment, we used an in vitro transcription-translation system (PUREfrex®2.0) supplemented with dNTPs and purified auxiliary proteins: single-stranded binding protein (SSB) and double-stranded binding protein (DSB), needed for an efficient DNA replication. We used 1 nM as the starting DNA concentration of our OriL-GFP-Kan-OriR with a twofold excess of plasmid DNA encoding for both DNAP and TP. After a 4 hour incubation at 30°C, we purified the DNA from the reaction as described in the complete protocol. The size of our construct was determined by DNA gel electrophoresis. As a control, we used samples with and without dNTPs. On our gel, we observed that our linear construct (~2,3 kb) was successfully amplified, as shown in Figure 8. These results show that the phi29 replication machinery can be used for orthogonal DNA replication in vitro. We demonstrated that our origin-flanked linear construct (OriL-GFP-Kan-OriR) could be replicated in vitro by the phi29 minimal replication machinery.
Figure 8: Purified OriL-GFP-Kan-OriR from a coupled DNA transcription-translation and replication reaction with (+) and without (-) dNTPs, amplified in vitro using the PUREfrex 2.0 system An intense band in the expected size (~2.3 kb) is observed from the +dNTPs reaction, demonstrating the OriL-GFP-Kan-OriR in vitro replication.
Successful replication of our linear construct has been shown on the agarose gel (Figure 8). A very intense band is visible at a size of 2.3 kb, corresponding to the size of the OriL-GFP-Kan-OriR construct. This indicates that there has been successful in vitro amplification of our construct.
When trying to express the phi29 replication machinery in E. coli we came into contact with, Dr. Chang Liu and Dr. Julian Willis, who informed us that the expression of these four proteins needs to be tightly controlled to avoid toxic events. To find conditions where DNAP, TP, p5, and p6 are expressed in cells that can show exponential growth, we conducted a toxicity assay. We individually expressed in E. coli BL21 (DE3) pLysS cells all the four components of the phi29 machinery under three different T7 promoter variants with varying strengths. The medium T7 promoter has a relative strength of 0.64 compared to the Wild-Type T7 promoter. The weak T7 promoter has a relative expression of around 0.1. The used pLysS strain used exhibits some leaky expression, allowing limited gene expression without IPTG induction (Rosano et al., 2014). Furthermore, to better find a window of non-toxicity, we induced these constructs with different IPTG concentrations. E. coli BL21(DE3) pLysS cells have their T7 RNA polymerase (RNAP) under the control of a lac promoter. IPTG binds to the lac repressor allowing the expression of the genes which are under the control of the lac operator. Therefore, the higher the IPTG concentrations, the higher the expression of T7 RNAP and the genes under the T7 promoter.
Results -- Toxicity Assay﹀
We transformed all 15 constructs from the orthogonal module separately in E. coli BL21 (DE3) pLysS cells. All cells transformed with the constructs were plated on LB agar supplemented with the corresponding antibiotics; chloramphenicol and ampicillin. Single colonies for all constructs were grown overnight in LB with the corresponding antibiotics, and diluted the next day to have the same Optical Density (OD) at 600nm. The diluted cells were directly pipetted into a 96 well plate with different IPTG concentrations. The OD of the cells was measured over 20 hours in a plate reader. From the raw data the OD over time was plotted. The maximum slope of the exponential phase was calculated from the collected data. This maximum was normalized against the maximum slope of a strain expressing GFP with the corresponding T7 promoter. These results are plotted in bar graphs, as reported in Figure 9.
Figure 9: The graphs reported in this figure show the maximum growth rate of cells containing the 5 different constructs (DNAP, TP, p5, p6 and GFP as a control). Each construct has been studied in the presence of three different strength promoters, here indicated as weak, medium, and Wild-Type promoter. For all these conditions three different IPTG inductions were applied: Top left) no IPTG induction and Top right) 1 mM IPTG induction and Bottom left) 10 mM IPTG induction
The graphs reported in Figure 9 show that the maximum growth rate of the cells containing the constructs for the four replication proteins decreases whenever the promoter strength increases. The growth rates of cells expressing GFP were normalized so the relative difference could be compared to each other. When comparing the construct containing cells and the GFP one, we can conclude that expressing our constructs of interest using a higher promoter strength results in a higher display of toxicity by the cells. In some conditions, such as cells containing constructs for DNAP and p5, the cells show higher maximum growth rates than the cells containing the GFP construct. This is especially observed when strains are grown in the presence of high IPTG, which increases the concentration of expressed proteins. This can be due to the fact that high levels of GFP expression can lead to some toxicity due to H2O2 formation (Ganini et al., 2017). Cells containing TP and p6 show a significantly lower slope than cells containing GFP expressed under a Wild-Type and medium strength promoter, indicating that these proteins are more toxic to cells than GFP. When higher IPTG concentrations are supplemented to the sample, there is a higher expression of genes under the T7 promoter. This explains why higher IPTG concentrations seem to show higher toxicity for cells containing TP and p6 when normalized to cells containing GFP. In contrast, cells containing DNAP and p5 show lower toxicity at higher IPTG contractions compared to cells containing GFP. As mentioned above, this is probably due to the fact that GFP produces H2O2, and this production is also induced with IPTG.
Cells containing DNAP, TP, p5, and p6 under the expression of a strong promoter show higher toxicity compared to cells where these proteins are less strongly expressed, when normalized to GFP. Cells containing DNAP and p5 do not show extra toxicity compared to cells containing GFP, while cells containing TP and p6 do show toxicity. Overall, our data (Figure 9) suggests that the ideal condition to express this protein in vivo is by using medium or low strength promoter in low IPTG induction, especially for TP and p5. Until now, we have studied some conditions for toxicity for every single construct. However, if we want to express all the components together, more fine tuning will be required.
Control of gene expression levels with iFFL
The behavior of genetic parts and circuits in different bacterial species is unpredictable as it is influenced by host-dependent variations (Yang et al., 2018). We engineered a unique implementation of the iFFL incoherent feed-forward loop motif to achieve gene expression independent of these variables.
We have incorporated T7 promoter variants into our iFFL systems, which have been used to show independence of transcriptional rates as well as being compatible with portable T7 expression systems. Furthermore, a broad host range version of the system was developed and demonstrated in E. coli and P. putida.
Transcriptional variations may be caused by 3 host-dependent variables:
- Copy number
- Promoter strength
- T7 RNAP concentration
Results -- Copy Number Independence﹀
The expression levels in a genetic circuit are strongly correlated to the plasmid copy number of the DNA template (Segall-Shapiro et al., 2018). The amount of plasmid copy number can change when the plasmid is transferred between organisms. Therefore, there is a need for expression levels independent of plasmid copy number if the same genetic circuit is used in different organisms. The steady-state solution of our model tells us that when our repressor binding is fully non-cooperative, we have complete independence of plasmid copy number.
Figure 11: Steady-state GOI production for gene plasmid copy number 1 to 600 (genome integration to high plasmid copy number plasmid).
Figure 11 shows the steady-state level of the gene of interest as predicted by the model.
The model without assumptions has the same expression level independent of plasmid copy number (Figure 11). We therefore can transfer our circuit between organisms and expect the expression of the gene of interest to be independent of the changes in plasmid copy number of our orthogonal plasmid.
Results -- Independence to Promoter Strengths﹀
To validate our model prediction, we designed T7 promoters based iFFL systems with varying promoter strengths. We compared our wild-type T7 promoter based iFFL system to a iFFL system based on a T7 promoter variant with 50% strength compared to the wild-type (medium T7 based iFFL system) (Ryo Komura et al., 2018). As a control, we express GFP without TALE regulation. We expect our iFFL systems to give the same expression level independent of the promoter strength.
The output GFP fluorescence in E. coli was measured using flow cytometry during logarithmic growth phase after induction with 1mM IPTG. As a reference for background fluorescence, we used E. coli BL21(DE3) cells without any plasmid. The most dense region (determined by eye) in the scatter plot (forward and side-scatter) was selected for gating in order to only compare cells of similar morphology. Furthermore, the fluorescence histogram was gated to discern between fluorescent and non-fluorescent cells, by gating the E. coli BL21(DE3) cells without any plasmid. The median of the background was subtracted from the median of the samples, and the resulting values were plotted (Figure 12).
Figure 12: Steady-state GFP fluorescence measurement of promoter variation using flow cytometry. The graph depicts T7 and medium T7 iFFL systems, expected to give the same fluorescence according to the model. As a control, GFP under control of an unrepressed T7 promoter was used.
In Figure 12, similar GFP fluorescence can be observed for the T7 and medium T7 iFFL systems while the unrepressed control system shows higher fluorescence. This suggests successful insulation of gene expression from change in promoter strengths. This validates our model, which predicts that as long as the ratio of promoter strengths of both genes are kept constant we have stabilized expression.
Raw data from flow cytometry experiments can be found here.
Results -- Tunability﹀
The predictions made by modeling not only tell us that we can maintain the same level of gene expression but also that we can easily tune the expression levels by changing one of the promoters. By changing one of the promoters to another variant of T7, we can expect a different level of expression, while at the same time expect it to behave similarly when transferred between organisms. Through the use of T7 variants, we could achieve wide ranges of expression levels, which can be used to establish complex genetic circuits, while also expecting it to work similarly in different biological contexts.
We tested the prediction by changing the promoter controlling the TALE protein to a weak variant of T7, which has been shown to be about 10% in strength compared to the wild-type (Ryo Komura et al., 2018). According to the model, the expression should increase.
We measure GFP fluorescence using flow cytometry. As a reference for background fluorescence, we use E. coli BL21(DE3) cells without any plasmid. The most dense region (determined by eye) in the scatter plot (forward and side-scatter) is selected for gating in order to only compare cells of similar morphology. Furthermore, the fluorescence histogram is gated to discern between cells that are "off" or "on" (expressing GFP or not), by gating the E. coli BL21D(E3) cells without any plasmid.
The median of the background is subtracted from the samples and are compared.
Figure 13 indicates that there is higher fluorescence when a weaker T7 promoter variant is used to express the TALE protein in comparison to the original system.
Figure 13: Steady-state GFP fluorescence measurement of E. coli BL21(DE3) cells expressing our iFFL systems. The graph depicts a different T7 iFFL systems, one with both promoters T7, one with both medium strength and one where the promoter controlling of the TALE is weak. As a control GFP under control of an unrepressed T7 promoter was used.
The data suggests that we can increase expression of the gene of interest by decreasing the promoter strength of the TALE gene, through the use of a weaker T7 promoter variant. As orthogonal promoter varients will keep their ratio in strengths independent of biological context this would allow for both tunability and predictability. Therefore, we expect the use of T7 promoter variants to allow for high-precision tuning of expression levels in genetic circuits in a host-independent manner.
Results -- Independence to T7RNAP Concentration﹀
Transcriptional variations may also be caused due to changes in RNAP concentrations in expression strains of different species. This is also the case for portable T7 RNAP expression systems where variations are caused by translational differences between organisms. The iFFL loop was tested for different concentrations of IPTG, and thus on the effect of changes in E. coli BL21(DE3) T7 RNAP concentrations. In unrepressed systems, the expression of the GOI is a function of IPTG concentrations. However, in our iFFL system, since the transcriptional rates of TALE and GFP are under the control of T7 promoters, both genes have a constant ratio of transcription rates, thus similar GOI expression is expected. As a control, we expressed GFP under the control of the T7sp1 promoter.
The output GFP fluorescence in E. coli was measured using flow cytometry during logarithmic growth phase after induction with different concentrations of IPTG. As a reference for background fluorescence, we used E. coli BL21(DE3) cells without any plasmid. The densest region (determined by eye) in the scatter plot (forward and side-scatter) was selected for gating in order to only compare cells of similar morphology. Furthermore, the fluorescence histogram was gated to discern between cells that are "off" or "on" (expressing GFP or not), by gating wild-type E. coli BL21(DE3). The median of the background was subtracted and the samples were compared to each other.
In Figure 14, the GFP fluorescence of the unrepressed control changes with changing concentrations of IPTG while the iFFL system shows the same GFP expression across different IPTG concentrations. Thus, the iFFL system has been shown to insulate gene expression against changes in transcription rates (achieved by varying IPTG concentrations). There is a high difference in levels of expression between the controlled and uncontrolled samples. The large difference could be due to overexpression of TALE protein which significantly repressed the expression of GFP.
Figure 14: Steady-state GFP fluorescence measurement of IPTG titration using flow cytometry. The graph depicts a T7 iFFL system, induced using different levels of IPTG, which, according to the model, should give the same result. As a control, GFP under control of an unrepressed T7 promoter was used.
We can infer from the data that with our orthogonal iFFL, we can maintain constant levels of expression in varying T7 RNAP concentrations, which further corroborates with the potential of developing a portable system integrated to our loop. Unfortunately, a great decrease in fluorescence was observed. This is likely due to still fairly high rates of expression of TALEsp1, as its expressed by the same RBS. Therefore, a next step would be the design of RBS sequences that show lower translation rates of TALE or the use of lower strength T7 promoter variants.
Results above indicate successful implementation of the iFFL system to insulate from transcriptional variations. Varying promoter strengths and different IPTG concentrations do not affect expression levels as long as a constant ratio of transcription rate of TALE and GFP is maintained. Furthermore, we demonstrated the tunability of the iFFL system to achieve different levels of gene expression. As predicted by our model, we achieved gene expression independent of changes in transcriptional rates by maintaining constant ratios of transcriptional rates of TALE and GFP genes.
Portable T7 Expression System
Combining our iFFL circuit with portable T7 expression systems we see the potential of establishing universal systems for controllable expression.
Results -- Cotransformation with Portable T7 Expression System ﹀
To test the compatibility of our system with the portable T7 RNAP expression system, we first cloned our T7 promoter based optimized iFFL into low and medium copy number backbones (pICH82113, and pICH82094 respectively) from the MoClo toolkit. These plasmid backbones contain origins of replication compatible with ColE1. The clones were then co-transformed with the plasmid containing the portable T7 expression system into E. coli TOP10. Wild-type E. coli TOP10 does not endogenously express T7 RNAP.
Figure 15: Fluorescence histogram of cotransformations obtained by flow cytometry. Black is regular E. coli TOP10 cells, green is cotransformation of our T7 promoter based optimized iFFL, pink is cotransformation of an iFFL without repression.
Figure 15 shows the fluorescence histogram obtained by flow cytometry. A higher level of fluorescence is observed when our plasmid is cotransformed with the UBER plasmid.
We demonstrated the compatibility of our iFFL circuit with orthogonal T7 RNAP expression circuits. Setting the basis for universal systems that establish stable levels of expression across bacteria.
Strengths of ribosome binding sites (RBSs) across organisms change due to differing anti-Shine Dalgarno sequences (Salis, Mirsky, & Voigt, 2009) therefore, circuits need to be retuned. The iFFL system can be used to achieve controllable gene expression despite differences in RBS strengths. In order to see the effect of translational variation on the expression levels of the gene of interest (GOI) we modeled our system for a range of translational rates for both genes (TALE and GOI).
Figure 16: Steady-state GOI production while translation rates of both TALE and GOI are changed. The lines indicate the constant ratio of the translation rates.
As can be seen in Figure 16, the kinetic model maintains the same level of GFP expression when the translation rates for both genes remain in a constant ratio. To keep the same ratio in translation rates across organisms we used the same ribosome binding site (RBS) for both genes. Using the same RBS ensures that translation initiation for both genes change in a similar manner (Salis et al., 2009), more on the design choices can be found here.
According to our model solution we can maintain the same level of GOI expression when both translation rates (for TALE and for the GOI) remain constant. We therefore designed our system to contain the same RBS in front of both TALE and the GOI.
Expression across Different Organisms
In order to test our system in different organisms we made a broad host range version of our iFFL. This allows us to test the functioning of our iFFL circuits in E. coli and P. putida. Thereby demonstrating gene expression insulated from variations associated with microbial hosts.
Results -- Expression Across Organisms﹀
Broad host range promoter (PBHR) was designed by combining the conserved -10 and -35 regions from E.coli and B.subtilis (Yang S et al., 2018) and the promoter was engineered to contain a binding site for TALE repressor (PBHRsp1). Using the PBHR and PBHRsp1, we constructed an iFFL genetic circuit driving GFP expression. The circuit was transformed in E. coli and P. putida and, output fluorescence was measured by flow cytometry during logarithmic growth phase. To correct for background fluorescence, E. coli and P. putida without plasmids were used as blanks. GFP under the control of PBHRsp1 was used as a positive (unrepressed) control. As the cell morphologies of E. coli and P. putida are different they cannot be compared directly, gating was based on the most dense regions in the scatter plot for each organism. In order to compare the GFP expression levels between each organism, the background fluorescence for each organism was subtracted by its respective blank.
In Figure 17, the median fluorescence of the gated populations is plotted. A large difference in expression levels is observed between the unrepressed controls and the broad host range promoter based iFFL systems in E. coli and P. putida. However, similar levels of expression were observed from iFFL systems in E. coli and P. putida. The difference in expression levels between the unrepressed circuit is significantly higher than the difference in expression levels between the iFFL system (578530 and 2351.2, respectively).
Figure 17: Comparison of expression across E. coli and P. putida. Blue shows constitutive expression and red shows the expression when our iFFL system is implemented.
The data suggests that our system succesfully decreased the differences in expression levels between E. coli and P. putida. More tests would need to be done to truly demonstrate that our system can achieve similar expression across bacterial species. However, we strongly believe that with our wet lab results in combination with modeling predictions we have laid the foundation stabilization of genetic circuits across species barriers.
- Ganini, D., Leinisch, F., Kumar, A., Jiang, J., Tokar, E. J., Malone, C. Mason, R. P. (2017). Fluorescent proteins such as eGFP lead to catalytic oxidative stress in cells. Redox biology, 12, 462–468.
- Rosano, G. L., & Ceccarelli, E. A. (2014). Recombinant protein expression in Escherichia coli: advances and challenges. Frontiers in microbiology, 5, 172.
- Ryo Komura, W. A., Keisuke Motone, Atsushi Satomura, Mitsuyoshi Ueda (2018). High-throughput evaluation of T7 promoter variants using biased randomization and DNA barcoding." PLOS ONE.
- Salis, H. M., et al. (2009). Automated design of synthetic ribosome binding sites to control protein expression. Nature Biotechnology 27(10): 946-950.
- Segall-Shapiro, T. H., et al. (2018). Engineered promoters enable constant gene expression at any copy number in bacteria. Nature Biotechnology 36: 352.
- Yang, S., Liu, Q., Zhang, Y., Du, G., Chen, J., & Kang, Z. (2018). Construction and Characterization of Broad-Spectrum Promoters for Synthetic Biology.