Overview
We developed Directed Chassis-agnostic Evolution, or DiCE, a novel, easy-to-implement selection-based directed evolution platform built off Qbeta replicase, an RNA-based RNA polymerase. We worked towards demonstrating DiCE’s ability to evolve proteins in both E. coli and cell-free environments. For DiCE In Vivo, we 1) Demonstrated that Qbeta could replicate RNA in our in vivo system and thereby create the “anti-sense RNA” strand 2) Showed that Qbeta could “flip” an antisense chloramphenicol target gene and produce antibiotic resistance in E. coli, and 3) Modeled theoretical error frequency and distribution for our DiCE In Vivo system. For DiCE in vitro we 1) Showed Qbeta-dependent RNA self-replication with a spinach aptamer 2) Showed Qbeta-dependent expression of a GFP target gene between MDVs and 3) Successfully implemented a full proof-of-concept DiCE In Vitro to evolve a T7 polymerase to have 400% increased transcriptional activity for a T3 promoter. Furthermore, we generated standard selection schema compatible with PREDCEL (Heidelberg 2017) to expand the range of synthetic biological parts that can be created by any SSS. We were able to make first steps towards monitoring evolutionary progress with the infection reporter that was developed and created a new framework for conducting M13-based evolutions. Finally, we designed a selection schema to evolve a novel anti-CRISPR protein using our DiCE in vivo system. Taken together, our work on SSS presents a foundational advance towards a future where part creation is easier, faster, and more accessible.
DiCE in vivo
Currently, results have largely been impeded with the optimized constructs due to time and difficulty successfully establishing a working procedure of gene delivery into our electrocompetent E. coli. To the best of our knowledge, a single paper has been published within the last 40 years on the transfection of E. coli with mRNA by Akira Taketo, which has greatly limited the foundation from which we could work upon.8 This procedure is still under experimentation and being modified in order to move on to additionally planned experiments and proofs-of-concept in order to validate the hypothesized mechanism above.
Despite these current impediments to the in vivo portion of the DiCE project, data was gathered with the initial construct design. This data shows promise that Qβ replicase functions within bacteria and qualitatively suggests that mutant proteins were created for increased resistance to chloramphenicol. The latter, however, can’t be validated like the former due to a lack of sequencing data and inherent flaws of the initial construct design.
Experiment 1: Results
Experiment 2: Results:
Experiment 3: Results:
We are still gathering data and troubleshooting. We will edit this page when publishable data is available. :)
DiCE in vitro
If Qbeta was able to self replicate with the MDV regions as expected, then fluorescence should increase as a result of the spinach aptamer tag added to the Qbeta subunit RNA. Our results showed significant differences in fluorescent output between MDV-Qbeta-Spinach and mRFP-Spinach, with nearly a 500% increase in the fluorescence of the MDV-Qbeta-Spinach construct compared to the mRFP-Spinach background. This suggests that MDV-Qbeta-Spinach was operational in vitro in the used cell free lysate.
Testing MDV Region Impact on Replicability of Qbeta System
(IMAGE)We hypothesized that flanking a given gene of interest, especially Qbeta, with the midivariants should make the whole sequence more conducive to replication and amplification by Qbeta. This was tested by comparing the fluorescent output of trials of MDV-Qbeta-Spinach construct in vitro with trials of a Qbeta-Spinach only construct, which lacked the MDV flanking regions. Run in triplicate, the results suggested a clear improvement of the MDV flanked Qbeta-Spinach over the control Qbeta-Spinach only, with no trial of the MDV-Qbeta-Spinach dipping below the highest performing trial of the Qbeta-Spinach. The differences in amplification of each construct suggested a ~30% improvement in the amplification rate with the MDV flanked construct versus the non-MDV flanked construct. MDV regions clearly play a critical role in Qbeta replicatibility, and can greatly enhance replication and amplification.
MDV Regions with Genes Beyond Qbeta
(IMAGE)After demonstrating that the MDV regions allowed Qbeta self replication and were an important factor to the replicability of the Qbeta construct, the MDV regions were then applied to other target genes. The replicability of GFP RNA with and without MDV regions when exposed to Qbeta was determined in vitro, with a Qbeta construct featuring just the subunit as well as the full linked fusion protein. As shown in this data, GFP flanked by the MDV regions, in both the Qbeta full linker construct as well as with the subunit alone, were both shown to undergo replication with Qbeta in vitro, all present roughly ~2x above the background signal.
GFP Expression Inside Droplets:
(Image)Droplets were clearly visible and their fluorescence was detectable under the microscope, showing that the cell free system and Qbeta replication/amplification worked successfully within the emulsification process. This means that the emulsification process is viable for creating the link between phenotype and genotype necessary for self selecting directed evolution, allowing us to proceed to the next step of developing an in vitro system to achieve this.
DiCE - Evolving a T7 RNA Polymerase to Bind T3 RNA Polymerase’s promoter In Vitro:
(Image)Finally, once the feasibility of making non-Qbeta genes replicable and amplifiable by the MDV regions was demonstrated, the MDV regions were applied to the T7 Polymerase gene as an application of the Qbeta based DiCE method of directed evolution in vitro. By adding MDV regions to the T7 Polymerase, the T7 Polymerase was made replicable by Qbeta, which is a highly error-prone RNA transcriptase. This introduces mutations into each round of replication, which when put under selective pressure and linked to the replicability or availability of Qbeta in a system, should result in the accumulation of mutations that favor more well adapted candidates to the environment. In this system, the goal is the evolution of a T7 polymerase capable of binding to a T3 promoter region, and the selective pressure is introduced by linking Qbeta availability, which is responsible for replication, to binding and transcribing a T3 promoter. After 5 rounds of evolution, which involve passaging and expression inside cell free solution in increments, a ~400% increase was demonstrated in the rates at which T7 polymerase resulted in the expression of fluorescent reporter protein underneath T3 promoter regions with round 5 RNA compared to control non-evolved stock RNA for the polymerase. This suggests the power, speed, and accessibility of DiCE in vitro evolution as well as the usage of MDV regions for creating sequences replicable by Qbeta.
Conclusion
With these experiments, we first initially were able to demonstrate the ability of Qbeta to replicate itself and undergo RNA amplification, with an increase in fluorescence of 500% compared to the baseline of the non-amplifying control. We were then able to demonstrate that the MDV regions increased the replicability of RNA sequences by Qbeta, showing that there was a ~30% increase in the fluorescence of the MDV-Qbeta-Spinach compared to the Qbeta-Spinach only construct. Next, we were able to show the ability of any arbitrary gene to be amplified and replicated by Qbeta through the addition of the MDV regions, with the MDV-GFP constructs which showed ~2x the activity of non-MDV GFP controls. Then we were able to demonstrate that we could create a link between genotype and phenotype that could be used for self selection in vitro via an emulsification process, which succeeded and was clearly visible with GFP constructs. After establishing all of these elements, we were able to conduct a T7 polymerase to T3 promoter binding affinity evolution and increase the activity of T7 polymerase with T3 promoter regions by ~400% compared to a non-evolved control. These results demonstrate that the DiCE In Vitro technique is a powerful and viable method of directed evolution. Featuring rapid prototyping, deep tunability, and a wide range of modifiable environmental variables, DiCE In Vitro demonstrates the ability to evolve constructs rapidly, effectively and without the barriers of traditional in vivo directed evolution techniques. More work needs to be done to test the power and limits of this new technique, but it is clear that the success of DiCE is beyond random.
Novel Selection Schema
Infection Reporter Characterization
Due to the similarity between each of the infection reporters, we expected both of them to give high expression when infected and low expression if not. However, when both were expressed in a plasmid with a CloDF13 origin of replication (copy number 20-4014), they both exhibited a high variability in expression, with some infected cells giving no fluorescence levels above baseline, others giving slight increases in fluorescence, and still others giving large increases in fluorescence.
A single colony of either TG-1 and S2060 cells with infection reporters (parts BBa_K3258045 and BBa_K3258006) driven by the respective biobrick parts in a BBa_K3258044 backbone was picked and immediately infected with 2•107 pfu. Cells were then grown in LB with 35 µg/mL chloramphenicol and 25 mM glucose for 20 hours shaking at 37ºC.
Based on the results in the figure above, we decided to focus on characterizing Ppsp in S2060 cells. To potentially elucidate the reasons for high expression variability, we hypothesized that the promoter was differentially active at different phases of growth. Thus, we decided to infect the cells at various stages in the E. coli growth cycle as shown below:
An overnight culture of S2060 cells with infection reporters (parts BBa_K3258045 and BBa_K3258006) driven by the respective biobrick parts in a BBa_K3258044 backbone was picked and grown to saturation overnight. Cells were diluted 1:100 into fresh LB with 35 µg/mL chloramphenicol and 25 mM glucose, grown at 37ºC, and infected at the indicated time points. After 8 hours, the cells were grown for an additional 24 hours and their fluorescence and OD600 values measured.
Based on the results above, we decided to stick with using part BBa_K2832003 as the promoter for our infection reporter as opposed to the new BBa_K2832000.
Despite our best efforts, we were unable to conduct a successful evolution of allowing the T7 polymerase to bind to the T3 promoter using PREDCEL+ due to ambiguous results as a result of our passaging experiments. We will continue to troubleshoot these issues after iGEM. However, we had initial early signs of success when we saw plaques on plates transformed with all plasmids necessary for the evolution (SP, HP, and APMP).
Uninfected cells were diluted 1:5000 into LB with 50 µg/mL kanamycin, 35 µg/mL chloramphenicol, and 25 mM arabinose with variable amounts of rhamnose (10 mM to no rhamnose based on plaques assays) along with 0.5 µL of infected cells from a previous round and cultured in 50 mL conical tubes shaking at 37ºC overnight. Normalized fluorescence (ex. 488 nm, em. 530 nm) measurements were taken to monitor infection, but proved unreliable when compared to plaque assays or rolling circle amplification followed by PCR amplification (RCA-PCR).
After 6 rounds of evolution with the final round growing without rhamnose supplementation, an RCA-PCR was performed and a band detected corresponding to a 500 bp region of T7 where mutations were expected. However, sequencing results returned the consensus T7 polymerase sequence corresponding to that sequence.
We will continue to troubleshoot this evolution before and after the Jamboree and, once validated, move to test our biosensor selection schema.
Novel DE Application
Results
Because cell free is expensive, we started with testing just the dCas9 groups first to demonstrate that all of our individual components were functioning as expected. Following a few suboptimal trials where we struggled with bubbles in the plate reader, we obtained the following data.
Our positive and negative controls gave the expected results: our RFP fragment fluoresced, and our H2O negative control had minimal background. We saw marked lower expression for one our RFP positive controls, perhaps due to pipetting error.
Our RFP + scrambled sgRNA + dCas9 group fluoresced as expected (because the sgRNA variable region is scrambled and therefore unable to guide the dCas9 to the RFP gene for knockdown). What’s more, our trial with RR1 and dCas9 led to significant decrease. However, these initial results did not demonstrate AcrIIA4-dependent return of RFP expression: both our RFP+RR1+dCas9 + either dead AcrIIA4 or normal AcrIIA4 groups behaved the same and did not fluoresce.
After rerunning these AcrIIA4 groups specifically and seeing similar results, we sequence verified to confirm that we had successfully added the T7 promoter, RBS, and T7 terminator to our linear fragment. Because this sequencing data checked out, we were left with inconclusive results for AcrIIA4. We had two hypotheses as to why this might have been:
- In Vitro
- Occasionally proteins that are functional in vivo are not able to fold properly in cell free extract. Though this is uncommon it is possible that AcrIIA4 might have been one of these proteins.
- Perhaps more likely, the AcrIIA4 was struggling to express because it’s DNA sequence was mammalian codon optimized instead of bacterial. Though it is often not a problem to express mammalian optimized vectors in bacteria or bacteria extract, it can sometimes lead to far lower expression levels that could be responsible for the observed lack of effectiveness.
At this point in time, we unfortunately realized we were running low on our PURExpress kit. The cost to purchase more reactions was significantly high and we were still quite far from answering our initial question, so we decided to pivot our energy to other subprojects.
Instead, we shifted our focus to implementing a cost-effective protocol to develop our own cell-free extract, which we called ourTXTL. Check out this process here! Our hope was to return to cell free testing once we finalized this protocol, but unfortunately we only got successful results with ourTXTL later in the summer at which point it was too late to pick up this project again!
We did, however, carefully design a three plasmid system that could effectively evolve AcrIIA4 to be more effective against dxCas9 using the DiCE in vivo evolutionary system. See below.
Conclusion
While we were not able to conclude whether AcrIIA4 is effective against dxCas9 this summer, we do indeed think this project is worth pursuing in the future. Especially given that cell free expression is far more accessible given the ourTXTL process we worked on this summer, we hope to return to additional testing following the Jamboree.
Furthermore, we went ahead and envisioned the following DiCE in vivo selection schema to evolve AcrIIA4 to be better at inhibiting dxCas9. Once we get the answer to our initial question of AcrIIA4’s current effectiveness against dxCas9, we can pretty easily clone the following system by simply swapping in AcrIIA4 to be between the Qbeta binding MDVs with restriction enzymes, and changing the 20 base pair variable region of the sgRNA to block CmR expression. See our video for a more detailed explanation of our AcrIIA4 DiCE selection system!