POC results – In-vivo Toehold Screening
As a major part of our project was aimed at developing a platform for generating effective toehold elements for specific eukaryotic translation, we had to provide evidence for our program-generated sequences’ efficiency. As not much formerly published work exists on toehold in eukaryotes, we had little to instruct us in choosing between different toehold characteristics. We thus utilized the Toehold tryouts as an opportunity to examine possible features of eukaryotic toehold design, with the hope of implementing the conclusions in the program and retesting the outputs in later cycles. Previous studies in the field have not focused on endogenously expressed triggers; thus, we chose not to focus on endogenously expressed triggers, as we figured this challenge should be addressed only after an effective toehold structure is attained for eukaryotes. We chose to start with Wang and colleagues’ original toehold as a point of comparison of our optimization results [1] as other works chose to use alternative structures to the classical toehold structures for specific translation in eukaryotes [2].
To effectively examine Wang et al. Toehold's performances compared to ours, we tried to test the sequences in a system as similar to the one used in their work as we could attain (Figure 1). Plasmids with different miR-155-reactive toeholds sequences positioned 5’ to GFP reporter genes were examined and were compared to the miR-155-GFP plasmids used by Wang and colleagues (Figure 2), which were sent to us by the courtesy of the author [1]. Our initial budget constraints led us to use miRNA expression plasmids as triggers instead of the miRNA mimics used in the article, which might have some impact on toeholds performances. All sequences were tested in HEK-293 cells, which were seeded in 24-well plates and transfected using Lipofectamine 2000 24 hours later. Each toehold-GFP plasmid was transfected twice into cells, either alone or with the miRNA expression plasmid. Its GFP expression and fold change between the two conditions were measured via Fluorescence Activated Flow Cytometry (FACS) and compared to Wang et al.’s plasmid and to a constitutive GFP expression plasmid (figure 1).
Figure 1: outline of toehold testing in mammalian cell lines
Sequences are cloned into plasmids and transfected to HEK-293 cells seeded in 24-well plates 24 hours prior. Toehold sequences are either transfected alone or together with a plasmid expressing the trigger 155-miRNA.
Results from Wang et al.’s sequence suggest a negative effect of miRNA on expression
Four versions of the miR-155-toehold-GFP were created using our program (Figure 2; Supplementary – cycles’ sequences): (1) A sequence designed such that translation would only start after the stem-loop, meaning that the Kozak sequence and start codon are found after the toehold; (2) a sequence conceptually similar to the design in Wang and colleagues, meaning that a start codon and a Kozak sequence exist both at the middle of the stem-loop and at the beginning of the GFP’s open reading frame (ORF) and that the GFP’s ORF is not aligned with the first start codon; (3) a sequence generated according to the ‘classical’ toehold design in bacteria [3], meaning that an AUG and the Kozak sequence exist only inside the stem-loop; and (4) a sequence similar to the former but with an additional step in the design pipeline that prevents the formation of pseudoknots surrounding the toehold structure (see Model page for a better description of sequences’ design pipeline).
Figure 2: Toehold sequences tried in the 1st cycle
Colors represent different sequence elements: Orange – trigger-binding area; Red – Kozak sequence and start codon; Green – beginning of the GFP’s ORF. a. The sequence taken from Wang and colleague [1], Kozak and AUG are present both inside the loop and adjacent to the ORF. b. Four sequences generated by our program, with the Kozak and start codon found after (1) or inside (3, 4) the toehold stem-loop, or both (2). Sequence 4 was added with a pseudoknot prevention step.
To minimize additional intervening factors in the comparison between the toeholds, we decided to use Wang and colleagues’ expression plasmid (pCDNA3(-)) as a backbone to clone our sequences into. We chose to construct our plasmids from 3 parts, with the 2nd half of the GFP ORF inserted independently from the rest of the sequence, allowing us to use a ‘universal’ insert in a manner that fitted our logistical constraints. We used the PCR reaction to open and amplify only the backbone from Wang et al.’s plasmid and then constructed the new plasmid via Gibson assembly (Figure 3a). As Gibson’s success rates proved low, and to minimize costly and time-consuming sequencings, we screened plasmids from colonies of bacteria transformed with DNA from the Gibson reaction through restriction of a sample with XhoI (which cuts only inside the original toehold but not in any of our inserts) and running it through gel electrophoresis, marking uncut wells as suspected successful reactions and sending only these colonies for sequencing (figure 3b).
Figure 3: Toehold’s cloning procedure
a. Scheme of our cloning procedure, the backbone of the plasmid from Wang et al.[1] was amplified using PCR. 155-Toehold-GFP sequences were constructed with Gibson assembly from a constant middle-end GFP part and a variable 155-Toehold part. b. Gel electrophoresis of DNA samples extracted from colonies transformed with Gibson reaction output. Samples were either loaded directly (lower panel) or after restriction with XhoI (upper panel, see Lab notebook). Unsuccessful reactions are cut and present one lighter bend in the upper panel. Cut samples were excluded from further analyses.
When examining Wang et al.’s toehold’s performances, we noticed that in contrast with the results reported when using miRNA mimetics [1], our run of the miR-155 Toehold provided consistently lower expressions when co-transfected with the trigger miRNA expression plasmid (Figure 4a). To examine if the reduction in expressing population is resulting from the excessive load of plasmid co-transfection or from a direct and specific effect of the miRNA interacting with the toeholds, we compared its activity alone and with the 155-miRNA to its activity when co-transfected with a miRNA-21 plasmid, which is not expected to interact with the toeholds as it doesn’t bare the relevant trigger sequence. We could not see a large difference in reduction caused when co-transfecting with miRNA-155 or miRNA-21, but were able to repeat this procedure only once, and therefore prefer to apply caution when interpreting this result (Figure 5b). We also compared the different conditions for a constitutive GFP plasmid instead of the toehold, which is not supposed to interact with either miRNA. When examining the effect of co-transfection with miRNA-155 on this plasmid, it is apparent that expression is in fact not reduced in the double-transfected population, suggesting a specific negative effect of the miRNA-toehold interactions on translation (Figure 4b).
Figure 4: FACS results for Wang et al.’s miR-155 toehold
Co-transfection with miRNA-155 specifically decreases GFP expression in miR-155-Toehold-GFP. a. GFP is expressed in cells transfected with the miR-155 toehold construct (two right panels) but is lower when miRNA-155 expression plasmid is present (right). b. Open (constitutively expressing) GFP plasmid is unaffected by the existence of miRNA-155 (low expression is present overall). Scatterplots of FACS measurements were taken from live, single cells. The threshold for GFP expression (X-axis) was set according to the untreated cells (see Methods).
1st cycle’s sequence 1 had the highest fold-change of all sequences examined
As in Wang and colleagues' toehold, two of our four sequences exhibited consistently lower expression rates when transfected with the trigger. Results were also worse in these sequences when compared to expression with the non-binding miRNA-21 sequence instead of the trigger miRNA-155, but only one repetition was performed with this miRNA (Figure 5b). Of the other sequences, sequence 1, in which the start codon and the Kozak sequence are found only after the stem-loop element, consistently exhibited similar or better expression rates when transfected with the trigger than when transfected alone (Figure 5). Sequence 1 thus presented a seemingly higher fold-change than all other toeholds, although insignificantly according to the student’s t-test (However we believe further repetition of the testing would solve this issue). Overall expression was higher in the sequences having only the upstream start codon (seq. 4, 3), however in a non-specific manner.
Figure 5: FACS results for our program-generated sequences
a.Summed FACS results (n = 3 repetitions) of the transfection experiments’ shown as fold-change 
. When co-transfected with the trigger, sequence 1 consistently presents results higher or similar to the ones measured when transfected alone. b. FACS scatterplots showing GFP fluorescence in cells transfected with Wang et al.’s[1] Toehold-GFP or with our 1sr cycle sequences: alone (upper row), together with the trigger miRNA-155 expression plasmid (middle row), or with a non-binding miRNA-21 expression plasmid (lower row). FACS measurements were taken from live, single cells (see Methods).
mRNA could be used as triggers for toeholds in eukaryotes
One of our project’s main goals was to use windows inside mRNAs as triggers for eukaryotic toeholds. Despite mRNAs being formerly used to trigger toeholds in prokaryotes3 and in IRES-based eukaryotic toehold-like structures [2], we could not find any work that uses the classical toehold structure with triggers in mRNA. The ability to use sequences in mRNAs could highly expand the ensemble of triggers to choose from and thus expand the applicability of these structures for targeting to different cell populations. Working with miRNA-triggered toeholds in the 1st cycle further strengthened our decision to move to mRNA triggers as we found it difficult to assess our success in expressing the trigger plasmid, because of the limited tools for miRNA transcripts identification and our limited resources and time. Specifically, with mRNA, we could use a fluorescent trigger to quickly overcome this issue. As results from the 1st cycle suggested a specific effect of the miRNA-toehold interaction, decreasing protein expression, we wanted to check if this effect stems from the miRNA’s nature and could be avoided when using mRNA instead.
Therefore, we decided to further examine our toehold sequences using mCherry-mRNA as a trigger. We did this cycle with S. cerevisiae instead of the mammalian cell line, which allowed us to screen sequences in higher throughput and also to verify our software’s applicability to eukaryotes other than mammalian cells. We examined sequences with the two best-targeting designs from the first cycle, sequences 1 (Kozak downstream to the stem-loop) and 4 (Kozak inside the loop), but this time we were able to use our computational assessment method to choose the few highest-ranked sequences for each design option (Supplementary – cycles’ sequences), Metrics used are specified in the Model section). Sequences in this cycle are therefore ranked according to the probability of a successful interaction to occur. We used Gap Repair to clone our different Toehold-GFP sequences into a constitutive expression vector (pRS425) and compared GFP levels to those in yeast also transfected with mCherry expression plasmids (cloned inside pRS426).
Figure 6: Fluorescence results for the second cycle in yeast
Fluorescence was measured in a plate reader and normalized to OD (600 nm). Results are shown as fold-change 
. Asterisks on each bar represent statistical significance of the difference between expression with or without trigger sequence. Sequences 1 and 6, predicted to be the optimally active sequences in each category (Supplementary – cycles’ sequences) show significantly higher fold-changes compared to other toeholds in each category. * p-value < 0.05, ** p-value < 0.01, *** p-value < 0.001, **** p-value < 0.0001.
Interestingly, the two highest-ranked sequences in both categories (Kozak up and inside) also exhibited the highest fold increase when administered with the trigger. These sequences, Y1 (Kozak after the stem-loop) and Y6 (Kozak inside) were the only ones constitutively higher expressed when transformed together with mCherry plasmid (Figure 6). These results strengthen our assessment method’s validity and suggest a successful Toehold triggering by mRNA, which had not yet been reported in eukaryotes. Of special interest is the relative success of toeholds with the Kozak and start codon situated after the stem loop, a structure different from the one commonly used in bacteria. Based on these results and other advancements in our software, we constructed a 3rd cycle of sequences, but were unable to complete its testing by the Wikis’ closure deadline. We also checked expression levels without the trigger normalized to the yeast’s autofluorescence (measured in WT yeast under the same conditions). And the fluorescence of the toehold-GFP sequences compared to a constitutively expressed GFP sequence. All toehold-GFP sequences’ fluorescence was shown to be insignificantly different from autofluorescence when administered without the trigger sequence (student’s t-test), meaning that all sequences exhibited low leakiness of expression in non-target cells (Figure 7a). When administered together with the mCherry trigger, both Y1 and Y6 exhibited similar expression rates to the constitutive GFP sequence (student’s t-test), which, together with the low expression without the trigger accounts for a highly effective toehold activity (Figure 7b).
Figure 7: mRNA-triggered toeholds in yeast exhibit low leakiness without the trigger and effective expression with it
a. Fluorescent from yeast transformed with toeholds only (off-state) divided by autofluorescence of non-treated wild-type yeast. All sequences exhibited insignificant differences between off-state expression autofluorescence, assessed by student’s t-test. b. Fluorescence in yeast applied both toehold-GFP sequences and the trigger mCherry sequence, divided by fluorescence in yeasts with a constitutively expressed GFP plasmid. * p-value < 0.05, ** p-value < 0.01, *** p-value < 0.001, **** p-value < 0.0001. Student’s t-test.
To conclude the results of our toeholds testing, we were able to use the successive cycles to further develop our software and adapt it according to the results (see Project Engineering for additional adaptations made). We were also able to show improved activity of one of our toeholds over the previously reported eukaryotic toehold. In the yeast experiments, we were able to exhibit - for the first time, to our knowledge - effective toehold activity of an mRNA-triggered toehold in eukaryotes. Interestingly, the structure which worked best both in mammals and in yeast was the one of which start codon and Kozak sequence was downstream to the stem-loop element, a design different from the ones previously reported. Based on these findings and on our toehold assessment method, we designed another cycle of toeholds to be tested in mammalian cells, for which results are still pending.
References
- Wang, S., Emery, N. J., & Liu, A. P. (2019). A novel synthetic toehold switch for microRNA detection in mammalian cells. ACS synthetic biology, 8(5), 1079-1088.
- Zhao, E. M., Mao, A. S., de Puig, H., Zhang, K., Tippens, N. D., Tan, X., ... & Collins, J. J. (2022). RNA-responsive elements for eukaryotic translational control. Nature Biotechnology, 40(4), 539-545.
- Green, A. A., Silver, P. A., Collins, J. J., & Yin, P. (2014). Toehold switches: de-novo-designed regulators of gene expression. Cell, 159(4), 925-939.
Parts’ Improvement
Genetic stability is a fundamental theme of synthetic biology. Genetic constructs, fulfilling a specific destination in a synthetic cycle, must preserve their functionality over long periods of time, avoiding mutations that lead to the loss of the gene’s functionality. However, such preservation requires energy production, resulting in a reduction in fitness. In order to form a stable-functioning construct with an intact fitness value, novel tools of synthetic biology need to be brought into action. The Evolutionary Stability Optimizer (ESO) software is a computational tool developed by the iGEM TAU 2020 team, which is directed at automatically generating genetic constructs with preserved functionality and a minimal loss of fitness [1].
For our project’s part-optimization component, we utilized the ESO program to improve the Evolutionary stability of an existing part stored on the iGEM registry, BBa_K079050. We optimized it using the software in two fashions, a milder one and a fully-optimized one (for more details about ESO, visit the ESO website). To experimentally examine the effect of this optimization, we compared the evolutionary stability of the standard iGEM part in comparison with the evolutionary stability of the same part after optimization, by looking at its fluorescent signal over time. We tried to show that the optimized part (in both versions) reserves its functionality for longer times than the non-optimized one, so we designed the experiment in a way that enabled us to follow the process of losing functionality.
Figure 1: Expression levels (fluorescence) of non-optimized and optimized parts
E. coli transformed with the different parts were grown in restrictive media, split to allow continuous growth, and their fluorescence levels were measured daily. The results are shown in Figure 1. Optimization by the ESO software stabilized expression levels of the construct: expression levels of the non-optimized part were significantly lower than those of the mild optimized one (p-value = 4*10-3), student’s t-test) and those of the full optimized one (p-value = 9.7*10-7); as mentioned, the evolutionary stability (figure 2) has been measured as the time it took the construct to lose its fluorescence (and thus, its functionality). It was evident that the original, non-optimized construct lost its fluorescent signal faster than both the optimized constructs (while the mild optimized construct lost its fluorescence faster than the fully optimized one).
Figure 2: GFP fluorescence over time
GFP fluorescence levels over time (excit. 484; emmit. 510), optimized and non-optimized parts. Green - original non-optimized GFP; Red - minor optimization and Blue - full optimization. Each one performed as triplicates, marked by dotted, dashed, and solid lines. The non-optimized part lost its functionality in a short time - a significant decrease in fluorescence just after 2 days and no fluorescence after 3 days. Both minor-optimization and fully optimized parts took longer to lose their functionality - a decrease in fluorescence after 3-4 days and a loss of functionality in 5 or more days.
We would like to report about another results of an optimization process taken out by the ESO tool that has been performed over an existing part of the iGEM parts registry - BBa_J04421 (an eCFP coding device, containing a promoter, an RBS, a coding sequence and a terminator). The same procedures were taken while dealing with that part, in order to detect its expression levels and evolutionary stability. The results are depicted in figure 3, both the evolutionary stability and the expression levels, represented by the GFP fluorescence values of the different versions of the part (non-optimized/merely optimized/fully optimized) in day 1, the initial measurements timepoint. It could be witnessed regarding that part as well that the ESO tool has optimized its evolutionary stability and its expression levels.
Figure 3: eCFP fluorescence over time
eCFP fluorescence levels over time (excit. 434; emmit. 477), optimized and non-optimized parts. Solid - original non-optimized eCFP; dashed - minor optimization and dotted - full optimization. Each one performed as triplicates, marked by different lines. The non-optimized part lost its functionality in a short time - a significant decrease in fluorescence just after 3 days and no fluorescence after 4 days. Both minor-optimization and fully optimized parts took longer to lose their functionality - a slight decrease in fluorescence after 4 days and a loss of functionality in 5 or more days.
To conclude, the successful application of the ESO software to optimize the BBa_K079050 part, our results support the possibility that the ESO tool bears the ability to improve evolutionary stability and expression levels of genetic constructs. This ability is an important one considering the need to preserve both the functionality and fitness of genetic constructs in various Syn. Bio. applications.
References
- (1) Menuhin-Gruman, I., Arbel, M., Amitay, N., Sionov, K., Naki, D., Katzir, I., ... & Tuller, T. (2021). Evolutionary Stability Optimizer (ESO): A Novel Approach to Identify and Avoid Mutational Hotspots in DNA Sequences While Maintaining High Expression Levels. ACS Synthetic Biology, 11(3), 1142-1151.