Toehold switches for the detection of miRNA

miRADAR focuses on microRNAs (miRNA) as potential biomarkers to diagnose MS patients. In our final test platform, miRNAs from the blood will be amplified after which a threshold system will be implemented. This only results in the production of a colour output when miRNA levels indicating MS patients are present. Toehold switches are riboregulators where a gene is repressed by a toehold structure which contains an unavailable ribosome binding site (RBS) in the loop and a start codon in the bulge. In response to an RNA trigger generated by our threshold system, this toehold structure unfolds, which allows the ribosome to bind leading to a colour change.¹ The output from the threshold system will be longer in nucleotides compared to miRNAs. Although the toehold switches in our diagnostic test platform will detect the longer output of the threshold system, we were interested in detection of miRNAs with toehold switches as a proof of concept.

This part of our dry-lab aims to design these miRNA-targeting toehold switches for an MS- specific miRNA trigger. We specifically focused on detecting hsa-miR-484 (BBa_K5106000) since this miRNA is shown to be upregulated in relapsing-remitting MS (RRMS) patients by Regev et al. (2018).² Eventually, the designed toehold switches will be tested in the lab.

Software tools for the design of miRNA-detecting toehold switches

SwitchMi Designer

For the design of the toehold switches we used the software SwitchMi Designer developed in Python by the iGEM team of Uparis_BME in 2021. This software program is a modified version of a software tool built by the iGEM team of Ulaval in 2019, which was able to design toehold switches for longer RNA sequences.
SwitchMi Designer uses the distinction of the paired and unpaired region of toehold switches (Figure 1). The number of nucleotides of the miRNA that anneals to the unpaired and paired region can be given as a users’ input. The software loops through the miRNA and searches for two weak (A or T) and one strong (G or C) base in the trigger miRNA sequence, of which the reverse complement was found to be the proper base for the bottom stem in the toehold switch by Green et al. (2014).¹ These 3 bases of the miRNA will anneal to the first three nucleotides in the paired region of the toehold structure (Figure 1). The software checks if a toehold switch annealed to an miRNA has a lower minimum free energy (MFE) compared to the toehold switch and miRNA alone in the solution. For an miRNA to hybridize to a toehold switch properly, it should be energetically more favourable for the toehold switch to anneal an miRNA compared to the toehold switch and miRNA not hybridized to each other.³ The translated protein will also be checked for stop codons, for the first amino acid to be methionine and for all amino acids in the part between start codon and protein sequence to be from the users’ input. In addition, the linker should not introduce a frameshift in the translation product. The top stem, bulge and bottom stem should also be in total 19 base pairs long. When the designed toehold switches do not meet these requirements, they are discarded.

Structure of a toehold switch and miRNA. The structural names of the toehold hairpin structure are indicated with labels. The toehold hairpin structure contains a bottom stem, bulge, top stem and loop. The sequence names are indicated with an arrow. The sequences are unpaired and paired region, and the linker. In green an miRNA is indicated. The orange square in the miRNA indicates the three nucleotides the program searches, after which an orange dotted line indicates the cut between the region in the miRNA hybridizing to the unpaired and paired region of the toehold switch. The reverse transcribed parts of these three nucleotides are indicated in the same colour in the toehold switch.

Adjusting the software

For the design of our toehold switches we used an adjusted version of SwitchMi Designer. This process is described in one of our engineering cycles. The software was adjusted in Python (version 3.10.12) such that a toehold switch was designed to hybridize with an miRNA at as many nucleotides as possible without removing the switch’s key feature of finding the proper base of the bottom stem. Due to these adjustments, we obtained a tool where the input variables stating the number of miRNA nucleotides annealing to the unpaired region and paired region could be discarded.

Input variables SwitchMi Designer:
Path to miRNA input sequence (FASTA format)
Path to output folder
Length of the paired region of the toehold
Length of the unpaired region of the toehold
Sequence reporter gene
Molecule type of input sequence (RNA/DNA)
Minimum number of unpaired bases in the secondary structure of the miRNA
List of suitable amino acids

Input variables adjusted software:
Path to miRNA input sequence (FASTA format)
Path to output folder
Sequence reporter Gene
Molecule type of input sequence (RNA/DNA)
Minimum number of unpaired bases in the secondary structure of the miRNA
List of suitable amino acids

The sequence located in the top stem and loop is conserved among all toe- hold switches designed by SwitchMi Designer and will contain the RBS and the start codon. Therefore, the sequences of these regions are fixed in the code of the software. The toehold structures obtained by SwitchMi Designer are closely related to toehold switches designed by Green et al. (2014).¹ For the design of our toehold switches we chose to include the bulge in the conserved region and use the B series designed by Pardee et al. (2016) as this toehold switch structure is optimized for shorter (mi)RNA molecules.⁴ The conserved region of the B series toehold switch includes the loop, top stem, and bulge of the toehold structure. Due to the longer conserved region, the restriction of 19 nucleotides in the top stem, bulge and bottom stem was not suitable for this design anymore. Therefore, we enforced that the bottom stem should consist of 11 base pairs matching the B series toehold switch. If the miRNA sequence of the bottom stem is shorter than eleven nucleotides, the program will add A-U base pairs to obtain a bottom stem of eleven base pairs. If it is longer than eleven, the program will indicate the number of nucleotides not annealed to the toehold switch, such that the user can decide on which toehold switch to use. We also adjusted the linker to match the linker of the B series toehold switch.

SwitchMi Designer:
Conserved region: AUACAGAAACAGAGGAGAUAU
Linker: ACCUGGCGGCAG
Adjusted software:
Conserved region: GGACUUUAGAACAGAGGAGAUAAAGAUG
Linker: AAACCTGGCGGCAGCGCAAAAG

Instead of NUPACK, we also used ViennaRNA (version 2.6.4) for the prediction of the MFE and the secondary structure.^5,6 Wayment-Steele et al. (2022) found that the predicted structures by ViennaRNA were more correlated with experimental data compared to the structures predicted by NUPACK.⁷ The secondary structures of the toehold switches predicted by ViennaRNA were also visualised in figures, using the Python library Forgi (version 2.2.3).⁸

Running the software to design toehold switches

As stated previously, we design toehold switches to detect hsa-miR-484 as the miRNA was found to be upregulated in RRMS.² The RNA sequence of the miRNA was obtained through the miRBase database (v22) and was formatted to a FASTA file.⁹ Additionally to the miRNA sequence FASTA file as input, we also specified that the toehold switch sequence from start codon until the protein sequence should encode amino acids with low molecular weight. We did not specify the reporter as we were only interested in the toehold switch structure with the linker. According to Brennecke et al. (2015), the minimal target site of an miRNA should at least contain seven bases for the proper annealing of the miRNA.¹⁰ Therefore, the parameter indicating the minimum unpaired residues was set to seven.

Sequence: ‘/path/gene_name.fasta’
Output_folder: ‘/path/output_directory’
Reporter: ‘’
mol_type:
RNA min_unpaired: 7
suitable_aa: [’G’, ’A’, ’S’, ’P’, ’V’, ’T’, ’C’]

Toehold switches for the detection of hsa-miR-484

With the previously mentioned input variables, the software predicted three possible toehold switch sequences with index 8, 10 and 16 (Table 1). The index is given to a toehold switch when the software tries find a proper base for the bottom stem. It represents the first nucleotide of the paired region in the toehold switch (Figure 1). If the requirements for a toehold switch are not met, the predicted toehold switch will be discarded. The main difference between the toehold switches is the number of added A-U base pairs in the bottom stem to obtain eleven base pairs in this region.

Predicted toehold switch sequences. Sequences of toehold switches with index 8, 10, and 16 predicted by our model meeting all requirements. Conserved regions are indicated in bold.
Index	Toehold Sequence
8	GGAUCGGGAGGGGACUGAGCCUGAGGACUUUAGAACAGAGGAGAUAAAGAUGUCAGGCUCAGUAAACCUGGCGGCAGCGCAAAAG
10	GGAUCGGGAGGGGACUGAGCCUGAUUGGACUUUAGAACAGAGGAGAUAAAGAUGAAUCAGGCUCAAAACCUGGCGGCAGCGCAAAAG
16	GGAUCGGGAGGGGACUGAGCCUGAUUUUUUUUGGACUUUAGAACAGAGGAGAUAAAGAUGAAAAAAAAUCAAAACCUGGCGGCAGCGCAAAAG

The software also predicted the MFE of the miRNA, toehold switch and the miRNA annealed to the toehold switch (Table 2). As the MFE of the toehold+miRNA is lower compared to the MFE of the toehold switch added to the MFE of the miRNA, our predicted toehold switches energetically favour being annealed to the miRNA confirming this requirement.

Results of the MFE. The MFE (kcal/mol) of the miRNA, toehold switches (index 8, 10, and 16), and miRNA annealed toehold switches
Index	MFE miRNA (kcal/mol)	MFE toehold (kcal/mol)	MFE toehold+miRNA (kcal/mol)
8	-0.70	-31.4	-58.1
10	-0.70	-27.9	-58.1
16	-0.70	-20.4	-59.2

The software program also generates a figure of the secondary structure predicted by ViennaRNA (Figure 2a-c). Through these figures we were able to identify the loop with the RBS, a top stem of 5 base pairs long, a bulge with the start codon, and a bottom stem of eleven base pairs long. The bulge shows incorrect base pairing in the middle. We assume this will not be a problem for the linearization of toehold switch, as the sequence was found to operate properly by Pardee et al. (2016).⁴ From these figures, we can conclude that the toehold hairpin structures predicted by ViennaRNA match the requirements. However, in all predicted toeholds switches the linker is annealed to the unpaired region, which could interfere with the hybridization of the miRNA.

To validate the annealing of the miRNA to the toehold switch, the web browser of NUPACK was used (Figure 2d-f).⁵ We generated secondary structures of hybridization between the miRNA and toehold switches and noticed that the miRNA was able to fully bind to all three toehold switches. Additionally, we performed a one tube analysis of all toehold switches including the trigger miRNA (Figure 3). For this analysis, all three toehold switches were set on a concentration of 1.0 \muM and the trigger miRNA was set on a concentration of 3.0 \muM. This showed that the toehold switches anneal all miRNA molecules present in the solution.

As the predicted structure of the toehold switch with index 16 shows more annealing between the unpaired region and the linker compared to the other two toehold switches (Figure 2c), we chose the toehold switches with index 8 and 10 (BBa_K5106001 and BBa_K5106002 ) to be tested in the lab (Figure 2a-b).

Secondary structures of the (annealed) predicted toehold switches. The secondary structure of the toehold switches predicted by our software program to meet all criteria. The complementary region, RBS, start codon, linker, and trigger miRNA are indicated with a red, blue, yellow, green and purple shade behind the secondary structure respectively. a, b, c: The secondary structure predicted by ViennaRNA as output from the software where a loop is indicated in blue, a bulge is indicated in yellow, free nucleotides at the end of the sequence are indicated in orange, and free nucleotides in between the sequence are indicated in red. The parts of toehold hairpin structures (Bottom Stem, Bulge, Top Stem, Loop) are indicated with a label. d, e, f: The secondary structure of the toehold switch annealed to the trigger miRNA hsa-miR-484 predicted by NUPACK where red/orange/yellow nucleotides indicate a high probability of the possible structure and green/blue nucleotides indicate a low probability of the structure. a: toehold switch with index 8; b: toehold switch with index 10; c: toehold switch with index 16; d: toehold switch with index 8 annealed to hsa-miR-484; e: toehold switch with index 10 annealed to hsa-miR-484; f: toehold switch with index 16 annealed to hsa-miR-484.

NUPACK hybridisation analysis of the trigger miRNA hsa-miR-484 (Trigger) with three predicted toehold switch sequences (Toehold 1-3), provided by the improved software. Toehold 1 indicates the toehold switch with index 8, Toehold 2 indicates the toehold switch with index 10 and Toehold 3 indicates the toehold switch with index 16. The concentration (nM) of each complex present in the system is shown.

Future perspective

By adjusting the software tool SwitchMi Designer, we obtained a tool more suitable for the design of miRNA-detecting toehold switches. By providing the miRNA sequence of hsa-miR-484 as an input, the software tool predicted three possible toehold switches for the detection of hsa-miR-484. Due to the predicted structure of these switches, we chose two to be tested in the lab.

Despite our ability to predict miRNA-detecting toehold switches, there are some points where we could improve the software. The hybridization of miRNA to the toehold switch is now validated in the NUPACK web browser. To make the analysis of the predicted toehold switch structures more consistent, the software tool could generate a figure of the hybridization of the toehold switch and miRNA predicted by ViennaRNA. By adding this to the software, it would also take less time to generate these figures and validate the hybridization of the toehold switch to the miRNA. In addition, we could add the predictive model built by the iGEM team of SASTRA in 2019. They built a model to predict the ON/OFF ratio of miRNA detection toehold switches which helps predict the efficacy of toehold switches. By combining the code of SASTRA with our software, we could verify the predicted toehold switches further. By further improving the software tool, future iGEM teams can use the tool to design and validate toehold switches for their own input

(1)

Green, A. A.; Silver, P. A.; Collins, J. J.; Yin, P. Toehold Switches: De-Novo-Designed Regulators of Gene Expression. Cell 2014, 159 (4), 925. https://doi.org/10.1016/j.cell.2014.10.002.

(2)

Regev, K.; Healy, B. C.; Paul, A.; Diaz-Cruz, C.; Mazzola, M. A.; Raheja, R.; Glanz, B. I.; Kivisäkk, P.; Chitnis, T.; Jagodic, M.; Piehl, F.; Olsson, T.; Khademi, M.; Hauser, S.; Oksenberg, J.; Khoury, S. J.; Weiner, H. L.; Gandhi, R. Identification of MS-Specific Serum miRNAs in an International Multicenter Study. Neurology Neuroimmunology & Neuroinflammation 2018, 5 (5), e491. https://doi.org/10.1212/NXI.0000000000000491.

(3)

Lane, A. N.; Jenkins, T. C. Thermodynamics of Nucleic Acids and Their Interactions with Ligands. Quarterly Reviews of Biophysics 2000, 33 (3), 255–306. https://doi.org/10.1017/S0033583500003632.

(4)

Pardee, K.; Green, A. A.; Takahashi, M. K.; Braff, D.; Lambert, G.; Lee, J. W.; Ferrante, T.; Ma, D.; Donghia, N.; Fan, M.; Daringer, N. M.; Bosch, I.; Dudley, D. M.; O’Connor, D. H.; Gehrke, L.; Collins, J. J. Rapid, Low-Cost Detection of Zika Virus Using Programmable Biomolecular Components. Cell 2016, 165 (5), 1255–1266. https://doi.org/10.1016/j.cell.2016.04.059.

(5)

Fornace, M. E.; Huang, J.; Newman, C. T.; Porubsky, N. J.; Pierce, M. B.; Pierce, N. A. NUPACK: Analysis and Design of Nucleic Acid Structures, Devices, and Systems; preprint; Chemistry, 2022. https://doi.org/10.26434/chemrxiv-2022-xv98l.

(6)

Lorenz, R.; Bernhart, S. H.; Höner zu Siederdissen, C.; Tafer, H.; Flamm, C.; Stadler, P. F.; Hofacker, I. L. ViennaRNA Package 2.0. Algorithms for Molecular Biology 2011, 6 (1), 26. https://doi.org/10.1186/1748-7188-6-26.

(7)

Wayment-Steele, H. K.; Kladwang, W.; Strom, A. I.; Lee, J.; Treuille, A.; Becka, A.; Das, R. RNA Secondary Structure Packages Evaluated and Improved by High-Throughput Experiments. Nature Methods 2022, 19 (10), 1234–1242. https://doi.org/10.1038/s41592-022-01605-0.

(8)

Thiel, B. C.; Beckmann, I. K.; Kerpedjiev, P.; Hofacker, I. L. 3D Based on 2D: Calculating Helix Angles and Stacking Patterns Using Forgi 2.0, an RNA Python Library Centered on Secondary Structure Elements., 2019. https://doi.org/10.12688/f1000research.18458.2.

(9)

Kozomara, A.; Birgaoanu, M.; Griffiths-Jones, S. miRBase: From microRNA Sequences to Function. Nucleic Acids Research 2019, 47 (D1), D155–D162. https://doi.org/10.1093/nar/gky1141.

(10)

Brennecke, J.; Stark, A.; Russell, R. B.; Cohen, S. M. Principles of MicroRNA–Target Recognition. PLOS Biology 2005, 3 (3), e85. https://doi.org/10.1371/journal.pbio.0030085.