Team:Evry Paris-Saclay/Engineering
**Figure 1.** Toehold switches principle (schematic inspired from iGEM Athens 2018).
The Rosewood biosensor project was inspired by the engineering framework of Synthetic Biology: Design, Build, Test and Learn [DBTL]. We can clearly identify two complete DBTL cycles during the project duration that were invaluable for our team’s progress.
## Rosewood DBTL Cycle #1
### Design
The first iteration of the DBTL cycle used the toehold switches predicted by the CUHK tool [[6]](#ref6). This was done with two purposes in mind. First, we wanted to get used to the experimental protocols relating toehold switches construction and testing while our own model was under construction. More importantly, we wanted to build these toehold switches with the goal of comparing our model against the previously published software [[6]](#ref6) (Check our ["Model"](https://2020.igem.org/Team:Evry_Paris-Saclay/Model) page for a more detailed description).
### Build
The toehold switch sensors (Table 1) were assembled by Golden Gate in the low copy plasmid pSB3T5. For this, fragments containing the T7 promoter followed by different toehold switches and RBS were synthesized with flanking type IIS restriction sites, and then interested into one of the acceptor vectors BBa_K3453101, BBa_K3453102 or BBa_K3453103 in which Golden Gates adapters are present upstream of sfGFP-LVAtag with either BsaI, BsmBI or BbsI sites, respectively.
The trigger sequences (Table 2) were placed under the control of the T7 promoter and followed by the strong SBa_000587 synthetic terminator and the resulting transcriptional units were synthesized and assembled in the high copy plasmid pSB1C3.
**Table 1.** Rosewood toehold switches sensors designed by the CUHK tool [[6]](#ref6).
| Label | Toehold Switch Part Number | sfGFP Expression Cassettes’ Part Number |
|---|---|---|
| Rosewood DmMatK Toehold Switch 1.1 | BBa_K3453011 | BBa_K3453111 |
| Rosewood DmMatK Toehold Switch 1.3 | BBa_K3453012 | BBa_K3453112 |
| Rosewood DmMatK Toehold Switch 1.4 | BBa_K3453013 | BBa_K3453113 |
| Rosewood DmRbcL Toehold Switch 1.1 | BBa_K3453031 | BBa_K3453131 |
| Rosewood DmRbcL Toehold Switch 1.2 | BBa_K3453032 | BBa_K3453132 |
| Rosewood DmRbcL Toehold Switch 1.3 | BBa_K3453033 | BBa_K3453133 |
| Rosewood DmTrnL-UAA Toehold Switch 1.1 | BBa_K3453051 | BBa_K3453151 |
| Rosewood DmTrnL-UAA Toehold Switch 1.2 | BBa_K3453052 | BBa_K3453152 |
| Rosewood DmTrnL-UAA Toehold Switch 1.3 | BBa_K3453053 | BBa_K3453153 |
| Label | Trigger Part Number | Trigger Transcriptional Unit Part Number |
|---|---|---|
| Rosewood DmMatK Trigger 1.1 | BBa_K3453021 | BBa_K3453121 |
| Rosewood DmMatK Trigger 1.3 | BBa_K3453022 | BBa_K3453122 |
| Rosewood DmMatK Trigger 1.4 | BBa_K3453023 | BBa_K3453123 |
| Rosewood DmRbcL Trigger 1.1 | BBa_K3453041 | BBa_K3453141 |
| Rosewood DmRbcL Trigger 1.2 | BBa_K3453042 | BBa_K3453142 |
| Rosewood DmRbcL Trigger 1.3 | BBa_K3453043 | BBa_K3453143 |
| Rosewood DmTrnL-UAA Trigger 1.1 | BBa_K3453061 | BBa_K3453161 |
| Rosewood DmTrnL-UAA Trigger 1.2 | BBa_K3453062 | BBa_K3453162 |
| Rosewood DmTrnL-UAA Trigger 1.3 | BBa_K3453063 | BBa_K3453163 |
**Figure 2.** *In vivo* characterization of sfGFP expression by *E. coli* BL21 Star™(DE3) cells harbouring the positive controls (BBa_K3453104 and BBa_K3453105) and the negative controls BBa_K3453103 (no promoter, no RBS) or an empty pSB3T5 vector. Cells were co-transformed with an empty pSB1C3 backbone in order to rule out any antibiotic influence on the *E. coli* growth and metabolism. The data and error bars are the mean and standard deviation of at least three measurements on independent biological replicates.
#### Fluorescence results
When tested against their cognate triggers (Figure 3), 7 of the 9 first-generation switches showed fluorescence with a mean comparable to that of the positive controls, except for two switches (DmMatK 1.3 and DmRbcL 1.3). When tested in the no trigger condition, three of the first-generation toehold switches (DmRbcL 1.2, DmTrnL-UAA 1.1 and DmTrnL 1.2) exhibited fluorescence higher than 50% of the average of positive controls which we define as leaky behaviour.
**Figure 3.** *In vivo* characterization of sfGFP expression by *E. coli* BL21 Star™(DE3) cells harbouring the 1st generation of rosewood toehold switches (Table 1) and triggers (Table 2). The negative controls have been performed with an empty pSB3T5, pSB1C3 (no trigger) and BBa_K3453103 (no promoter, no RBS) and the positive controls with BBa_K3453104 and BBa_K3453105. The data and error bars are the mean and standard deviation of at least three measurements on independent biological replicates.
#### ON/OFF Ratio
The positive controls showed on average 320 times the fluorescence of the negative controls confirming the validity of the experimental setup. Of the 9 first-generation toehold switches tested, DmMatK 1.1 exhibits an ON/OFF ratio greater than 100, DmMatK 1.4 shows an ON/OFF level close to 30 and DmMatK 1.3, DmRbcL 1.1 and DmTrnL 1.3 display at least a level greater than 5 (Figure 4).
**Figure 4.** MEFL / Particle fold changes of the 1st generation of rosewood toehold switches (Table 1) in the presence of the rosewood triggers (Table 2) compared to the MEFL / Particle value in the absence of the trigger.
#### Crosstalk
To measure orthogonality of the toehold switches, we performed additional crosstalk experiments. In these experiments, we tested the expression of the reporter gene when the trigger of a different toehold switch for the same gene was added inside the cell. Only three of the 9 first-generation toehold switches showed a higher fluorescence measurement in the presence of their non-cognate trigger or even in the absence of any trigger: DmRbcL 1.2, DmTrnL-UAA 1.1 and 1.2. The other 6 toehold switches generated a fluorescent response only in the presence of a trigger containing the corresponding binding site. It should be noted that the triggers designed with the CUHK tool [[6]](#ref6) are 120 nt long and some contain binding sites for several switches.
#### Full trigger
In addition, we tested the toehold switches in the presence of the respective full known fragments of *Dalbergia maritima* genes, which we refer to as the full triggers. The MEFL / Particle values were generally lower which indicates lower levels of switch - trigger binding most probably due to steric constrains.
#### Toehold switches showed a dynamic range that leverages the full expression capability of the cassette
The high dynamic range of fluorescence outputs without and with their cognate triggers showed us that we had enough expression change to easily distinguish between the negative and the positive results. We expected that the measurements obtained from the toehold switches will show a reduced dynamic range between the ranges shown by the control. However, as shown in Figure 3, we obtained positive results that surpassed the fluorescence of the positive controls suggesting that we leveraged the full capacity of the expression cassette.
#### Low leaky expression is key for obtaining high-performance toehold switches
We obtained a range of ON/OFF ratios expanding over two orders of magnitude (Figure 4). The highest performer switch, DmMatK 1.1, showed an ON/OFF ratio of 130, on the same order of performance as some of the best switches reported in literature. The positive signal of the switch is approximately 40% higher to the signal obtained for the positive controls, but not the highest positive signal that we obtained on the dataset. In contrast, the fluorescence measured in the absence of the trigger was the lowest among all the first-generation switches. For comparison, the switch DmMatK 1.4 had a positive signal also comparable to that of positive controls but its negative output was almost 5 times higher than that of DmMatK 1.1. Consequently, the ON/OFF ratio of DmMatK 1.4 is markedly reduced. Thus, large ON/OFF ratios are based primarily on very low levels of leakage expression found in the absence of the trigger.
#### Some genes targets seemed more easily detectable
On this very small sample size, the three toehold switches for DmMatK showed favorable behavior albeit with very different ON/OFF ratios. On the contrary, we obtained only one favorable switch for DmTrnL-UAA. It is tempting to suggest that DmMatK is an easier target in terms of the secondary structures that can be achieved to detect its sequence. However, given our small sample size and several other parameters that may have reduced DmTrnL-UAA toehold switches performance we would have to more rigorously test this hypothesis.
#### Crosstalk is detectable but at desirable levels
In contrast to other sensing purposes where a set of toehold switches that are highly orthogonal is desirable, the ability to detect triggers that are slightly off from the cognate trigger desired means that we will be able to detect triggers that have some mutations in their sequences.
The extent of this crosstalk needs to be measured as we clearly want to avoid false positives in our sensing. A possible approach for this would be to create a spectrum of triggers from negative controls to full triggers that measures the activation curve of the switches in relationship to the similarity to the cognate trigger sequence.
#### Efficacy score of the CUHK model was a poor predictor of toehold switch performance
As can be seen in Table 3, the efficacy score weakly predicted fluorescent measurements of the toehold switches. In particular, the two designs with highest expected performance *in silico* did not produce functional toehold switches. In addition, the poorest design in terms of efficacy turned out to be the toehold switch with the second highest fold change ratio. These results suggest that we could have obtained better results using toehold switches with lower predicted ranks and confirm the challenging nature of predicting RNA secondary structure dynamics *in silico*.
**Table 3.** Efficacy scores of the Rosewood toehold switches sensors designed by the CUHK tool [[6]](#ref6) vs their ranking according to fold change ratio measured *in vivo* (Figure 4). It should be noted that the last digit of the toehold name is the rank predicted by the CUHK tool.
| Label | Efficacy score in silico | Rank in vivo |
|---|---|---|
| Rosewood DmMatK Toehold Switch 1.1 | 98 | 1 |
| Rosewood DmMatK Toehold Switch 1.3 | 55 | 5 |
| Rosewood DmMatK Toehold Switch 1.4 | 44 | 3 |
| Rosewood DmRbcL Toehold Switch 1.1 | 35 | 6 |
| Rosewood DmRbcL Toehold Switch 1.2 | 32 | - |
| Rosewood DmRbcL Toehold Switch 1.3 | 15 | 2 |
| Rosewood DmTrnL-UAA Toehold Switch 1.1 | 115 | - |
| Rosewood DmTrnL-UAA Toehold Switch 1.1 | 110 | - |
| Rosewood DmTrnL-UAA Toehold Switch 1.3 | 110 | 4 |
| Label | Toehold Switch Part Number | sfGFP Expression Cassettes’ Part Number |
|---|---|---|
| Rosewood DmMatK Toehold Switch 2.1 | BBa_K3453014 | BBa_K3453114 |
| Rosewood DmMatK Toehold Switch 2.2 | BBa_K3453015 | BBa_K3453115 |
| Rosewood DmMatK Toehold Switch 2.3 | BBa_K3453016 | BBa_K3453116 |
| Rosewood DmRbcL Toehold Switch 2.1 | BBa_K3453034 | BBa_K3453134 |
| Rosewood DmRbcL Toehold Switch 2.2 | BBa_K3453035 | BBa_K3453135 |
| Rosewood DmRbcL Toehold Switch 2.3 | BBa_K3453036 | BBa_K3453136 |
| Rosewood DmTrnL-UAA Toehold Switch 2.1 | BBa_K3453054 | BBa_K3453154 |
| Rosewood DmTrnL-UAA Toehold Switch 2.2 | BBa_K3453055 | BBa_K3453155 |
| Rosewood DmTrnL-UAA Toehold Switch 2.3 | BBa_K3453056 | BBa_K3453156 |
| Label | Trigger Part Number | Trigger Transcriptional Unit Part Number |
|---|---|---|
| Rosewood DmMatK Trigger 2.1 | BBa_K3453024 | BBa_K3453124 |
| Rosewood DmMatK Trigger 2.2 | BBa_K3453025 | BBa_K3453125 |
| Rosewood DmMatK Trigger 2.3 | BBa_K3453026 | BBa_K3453126 |
| Rosewood DmRbcL Trigger 2.1 | BBa_K3453044 | BBa_K3453144 |
| Rosewood DmRbcL Trigger 2.2 | BBa_K3453045 | BBa_K3453145 |
| Rosewood DmRbcL Trigger 2.3 | BBa_K3453046 | BBa_K3453146 |
| Rosewood DmTrnL-UAA Trigger 2.1 | BBa_K3453064 | BBa_K3453164 |
| Rosewood DmTrnL-UAA Trigger 2.2 | BBa_K3453065 | BBa_K3453165 |
| Rosewood DmTrnL-UAA Trigger 2.3 | BBa_K3453066 | BBa_K3453166 |
**Figure 5.** *In vivo* characterization of sfGFP expression by *E. coli* BL21 Star™(DE3) cells harbouring the 2nd generation of rosewood toehold switches (Table 4) and triggers (Tables 2 and 5). The negative controls have been performed with an empty pSB3T5, pSB1C3 (no trigger) and BBa_K3453103 (no promoter, no RBS) and the positive controls with BBa_K3453104 and BBa_K3453105. The data and error bars are the mean and standard deviation of at least three measurements on independent biological replicates.
#### ON/OFF Ratio
ON/OFF ratios of the nine second-generation toehold switches were calculated by dividing MEFL / Particle values of switch with trigger and switch without trigger (Figure 6). DmRbcL 2.2 and DmTrnL-UAA 2.2 show very high fold-change with values exceeding 40 and 20 respectively. DmMatK 2.1 ranks third with a ratio of almost 10, while the other six switches don’t exceed a fold change of five.
**Figure 6.** MEFL / Particle fold changes of the 2nd generation of rosewood toehold switches (Table 4) in the presence of the rosewood triggers (Table 5) compared to the MEFL / Particle value in the absence of the trigger.
#### Crosstalk
Broad crosstalk tests were not performed for the second-generation switches, though five of the nine switches were also tested with triggers of other switches, if sequence overlap occurred.
#### The sensing performance of the second-generation switches is impaired by high values in negative scenarios
Even though the toehold switches showed high expression in the presence of their cognate trigger, we report high fluorescence values in their negative scenarios. This effect caused a reduction in most of the ON/OFF ratios and greatly impacted the sensing capabilities. Several reasons could explain this leakiness, among them:
1. Switch sequences cannot fold into a hairpin structure stable enough to resist ribosome binding and stem unzipping.
2. Endogenous RNAs from *E. coli* bind and unwind stem structure
A criterion we use to prevent choosing leaky switches, is TIR fold-change and OFF-TIR values predicted by the RBS calculator [[8]](#ref8). TIR itself is put together by different variables including Shine-Dalgarno similarity, ribosome binding free energy penalty, potential for ribosome sliding and others that rely on RNA secondary structure and RNA biophysics predictions.
To improve leakiness, we aim to both adjust the cutoffs for TIR and the other metrics and, to generate a more diverse library and have a broader range of possibilities, predict switches of multiple architectures, that vary in domain length (toehold, stem, loop) and predefined sequence and structural constraints.
#### Unlike that suggested by the DBTL Cycle #1 data, our gene targets do not seem to influence toehold switch performance
In contrast to behaviour seen in DBTL Cycle #1, DmTrnL-UAA in Cycle #2 proved to be a gene easier to detect with the second-generation toehold switches, and, on the contrary, DmMatK showed leaky behaviour. This suggests that the difficulties encountered in the two cycles are not particular to the sequence of the gene tested but rather more likely due to the toehold design methods.
## Conclusions
We have successfully designed, built, tested and analysed 18 rosewood toehold switches, half of which were are able to act as toehold switches and (i) efficiently repress the downstream reporter gene expression in the absence of the cognate trigger and (ii) release the translation inhibition in the presence of the cognate trigger. Moreover, we obtained several functional toehold switches for each of the three *Dalbergia maritima* genes tested: MatK, RbcL and TrnL-UAA, and 6 of them were further tested in a cell-free experimental setup as a Proof of Concept experiment for the implementation of our Rosewood detection tool (Check our ["Proof of Concept"](https://2020.igem.org/Team:Evry_Paris-Saclay/Proof_Of_Concept) page for a more detailed description).
## References
[1] Green AA, Silver PA, Collins JJ, Yin P. Toehold switches: de-novo-designed regulators of gene expression. Cell (2014) 159: 925–939.
[2] Hong F, Ma D, Wu K, Mina LA, Luiten RC, Liu Y, Yan H, Green AA. Precise and programmable detection of mutations using ultraspecific riboregulators. Cell (2020) 180: 1018-1032.e16.
[3] Pardee K, Green AA, Takahashi MK, Braff D, Lambert G, Lee JW, Ferrante T, Ma D, Donghia N, Fan M, Daringer NM, Bosch I, Dudley DM, O’Connor DH, Gehrke L, Collins JJ. Rapid, low-cost detection of Zika virus using programmable biomolecular components. Cell (2016) 165: 1255–1266.
[4] Boettner B. Harvard Wyss Institute’s nasal swab and toehold switch technologies licensed to Agile Biodetection to facilitate SARS-CoV-2 diagnostic efforts. Wyss Institute2020 Oct 5.
[5] Koksaldi IC, Ahan RE, Kose S, Haciosmanoglu N, Kehribar ES, Gungen MA, Ozkul A, Seker UOS. SARS-CoV-2 detection with de novo designed synthetic riboregulators. medRxiv (2020): 2020.07.28.20164004.
[6] To AC-Y, Chu DH-T, Wang AR, Li FC-Y, Chiu AW-O, Gao DY, Choi CHJ, Kong S-K, Chan T-F, Chan K-M, Yip KY. A comprehensive web tool for toehold switch design. Bioinformatics (2018) 34: 2862–2864.
[7] Zadeh JN, Steenberg CD, Bois JS, Wolfe BR, Pierce MB, Khan AR, Dirks RM, Pierce NA. NUPACK: Analysis and design of nucleic acid systems. Journal of Computational Chemistry (2011) 32: 170–173.
[8] Salis HM. The ribosome binding site calculator. Methods in Enzymology (2011) 498: 19–42.