Software: SwitchMi Designer

Toehold switches are the main part of ExoSwitch, as biosensors for cancerous microRNA. Toehold switches as de-novo-designed-riboregulators have been synthesized for the first time by Green et al. in 2014 ¹. These RNA-based regulatory systems enable the control of translation and transcription in response to the related RNA, taking advantage of the predictable Watson-Crick base pairing that limits crosstalks (Fig. 1).

The toehold switch is composed of two strands of RNA^1,2 :

• switch strand: contains a single stranded sequence (named unpaired region, see Fig.1) at the 5’ end which is followed by the hairpin module upstream the repressed gene. The hairpin module includes the ribosome binding site (RBS) in the loop of the hairpin and the start codon (AUG). The repressed gene encodes for a colorimetric, fluorescent, or electrochemical measurable output.
• trigger RNA (target): binds to the toehold sequence thanks to Watson-Crick base pairing, enabling the opening of the hairpin. Then, the exposition of the RBS site and start codon leads to the translation of the coding gene previously repressed.



In the context of the iGEM competition, our project ExoSwitch involves engineered toehold switches to detect and quantify miRNAs. Toehold switches are mainly used for the detection of viral and bacterial RNA ^3,4 but they have been used to detect mammalian miRNAs for the first time in 2019 by Wang S et al. ⁵.

Some tools already exist to design toehold switches such as the design software developed by team Hong Kong-CUHK in 2017 ⁶ or the ToeHolder, designed by the ULaval iGEM team in 2019. These previous tools do not work with microRNA as trigger RNA because micro-RNA are shorter than the recommended size.

Thus, we decided to build SwitchMi Designer, a new tool available on GitHub under the MIT license, that aims to generate libraries of toehold switch candidates for the detection of targetmicroRNAs. SwitchMi Designer is a modified version of the ToeHolder, previously developed by team ULaval for iGEM 2019. The Toeholder designs toehold switches targeting RNA sequences between 30 and 1000bp, but miRNA are short RNA sequences of 18-25 nucleotides. SwitchMi Designer aims to design toehold switch candidates with a sequence of minimum 7 complementary nucleotides with the trigger miRNA. This complementary sequence, also called as Toehold region (Fig.2), is composed of at least 4 free nucleotides at the 5’ end of the toehold switch and 3 nucleotides engaged in the hairpin root (starting with 2 weak (A-T) and 1 strong (G-C) pairing) ¹.

A general scheme of the workflow implemented in SwitchMi Designer is shown in Figure 3.

SwitchMi Designer takes the following data as input :

• path to the miRNA sequence in FASTA Format
• length of unpaired region
• length of paired region
• path to output folder
• minimum of unpaired residues in the secondary structure of the miRNA for a candidate trigger to be considered
• repressed gene (by default it is defined as EGFP)
• molecule type: DNA or RNA
• list of suitable amino acids.

As the length of the paired and unpaired sequence can be modified by the user, we decided to try several combinations to get our bank of candidates :

• 10 unpaired nucleotides and 4 paired nucleotides
• 12 unpaired nucleotides and 8 paired nucleotides
• 10 unpaired nucleotides and 9 paired nucleotides



To build SwitchMi Designer, we adapted the code written by the ULaval team with the ToeHolder to the new version of NUPACK (4.0). Then we modified different options regarding the toehold sequence like the length of the paired and unpaired sequence.

First, the miRNA sequence is parsed nucleotide by nucleotide with Biopython. Using NUPACK, a physical model is specified and the MFE proxy structure of the miRNA is calculated. A loop scans (Fig. 4) through the trigger RNA sequence as illustrated in figure 3 to find two weak (A-T) and one strong (C-G) bases that will form the hairpin base of the toehold. If a hairpin base is selected, a miRNA subsequence is created based on the length of unpaired and paired regions defined by the user. Unpaired residues in the secondary structure of miRNA are counted for the subsequence and the subsequence is discarded if there are less than 4 unpaired residues.



When a subsequence passes the requirement, MFE proxy structures and the free energy of the toehold of the toehold switch , of the miRNA and of the toehold switch bound to the miRNA are calculated with NUPACK.

The sequence following the start codon is translated into amino acids. Only toehold sequences containing the suitable amino acids in the first positions were selected. Indeed, once the toehold switch binds to its trigger miRNA, the hairpin opens and the repressed gene is translated into protein. As you can see on the figure 2, the start codon precedes the repressed genes with few nucleotides that vary with the candidate and the trigger miRNA. Some combinations of proteins directly translated after the start coding may not be favorable, such as MKN.

The toehold switch is selected as a candidate if all the following conditions are satisfied :

• there is no stop codon in the toehold switch
• the free energy of the toehold switch bound with the miRNA is smaller than the sum of the free energies of the miRNA and the toehold switch when they are not linked.
• the translated nucleic acid sequence is favorable for amino acid translation and stable.

Otherwise, the toehold is discarded and the next hairpin base is tested. Finally, the code returns two documents :

1. A first .csv file called "toehold_candidates" is created containing all toehold switch sequences which have the requirements (2 weak pairs / 1 strong pair at the hairpin base and minimum 4 unpaired residues), with the following information :
• Non paired residues = number of residues unpaired in the miRNA structure
• Structure = secondary structure of the toehold switch candidate in Dot-Parenthesis notation
• Sequence = the nucleic acid sequence of the subsequence candidate as toehold target
• Start = Position of the first nucleotides from miRNA that binds the toehold switch
• End = Position of the last nucleotides from miRNA that binds the toehold switch
• Length unpaired trigger = length of free nucleotides from the toehold switch that binds the trigger miRNA
• Length paired trigger = length of nucleotides engaged into the hairpin from the toehold switch that binds the trigger miRNA
2. A second .csv file called "selected_toeholds_results" is created and contains all toehold switches that passed the requirements (No stop codon in the sequence; No start codon except the one planned; The first amino acid is M and the next two amino acids are the predefined list of suitable amino acids). The file contains all previous data of these toehold switches (stored in toehold_candidates.csv) but also the following information :
• Index = Toehold sequence ID
• Free energy of MFE proxy structure of the toehold switch bound with the trigger miRNA
• Free energy of MFE proxy structure of the toehold switch
• Free energy of MFE proxy structure of the trigger miRNA
• GC content = percentage of G and C nucleic acid in the toehold switch sequence
• Protein sequence produced when the trigger miRNA binds to the toehold switch

Apart from those .csv files, the rest of the files are .txt files containing the sequence of each toehold selected, either in DNA or RNA.

After the bank of toehold switch candidates is generated, the user has to check that the MFE proxy structure of the toehold switch alone shows the good hairpin structure (Fig. 2), and if the 5’ end called as “unpaired” is free on RNAFold webserver. Further work on the SwithMi Designer would aim to implement these two functions.

What is microfluidics?

Microfluidics is a technology of manipulating fluids in micrometric channels. The systems work with very little amounts of fluids (only some nanoliters or microliters are enough). This field of research is inspired by nature, which already masters this technique. For example, trees distribute sap in millions of tiny capillaries in a controlled manner. Spiders possess a “microreactor” to produce the material to build their webs. It has emerged in recent years as a fundamental tool thanks to its many applications in multiple research fields, such as biology, chemistry, physics or medicine.⁷

Among the benefits offered by this technology, two of them stand out significantly. First, the small size of the device is very handy to handle tiny amounts of liquids to perform complex experiments, reducing the price of research. Second, such devices are ideal platforms for multi-step experiments due to their small size and ease of use. Thanks to microfluidics, human mistakes are greatly reduced, the experiments are highly efficient and easily reproducible. Automated protocols can be executed much easily on microfluidic devices, and a great range of experiments can be performed according to the design of the chip.⁸

Today, this technology has a huge potential, and is already used in various industrial fields : health (diagnostics, personalized therapy, pregnancy test), energy (oil recovery, marine power), cosmetics fragrances, antiperspirants), biotechnology (organ-on-a-chip, controlled release), environment (water purification, CO2 recycling, air treatment), agriculture (taste making, contamination analysis), … ⁹

The relevance of microfluidics in ExoSwitch

A microfluidic chip was an obvious option for us to implement our cancer miRNA detection tool. Microfluidic methods have been developed to isolate tumoral biomarkers in biological fluids. Circulating tumour cells, circulating tumoral nucleic acids, exosomes, … all of these biomarkers have been described as options to be integrated in microfluidic devices for cancer detection.¹⁰

As we are using toehold switches for detecting tumour-associated miRNAs present within exosomes, microfluidics offer us a unique opportunity to integrate all the steps of the detection, from the drop of blood to the miRNA quantification. In our proposed solution, a drop of blood would be the input of the microfluidic device. Then the exosomes would be separated inside the chip, then lysed to free the miRNAs present within them. These miRNAs would later be detected by specific toehold switches, with GFP being expressed when specific binding happens. At the end, fluorescence would be measured to know the concentration of tumour-associated miRNAs.

Integration of all these steps, and the ability of carrying all these separate experiments with small amounts of liquid and reagents, motivated us to consider the microfluidic chip as a solid option.

Design of the microfluidic chip

Independently of our wetlab results, we wanted to design the chips, so even if we couldn’t use them in our lab, the design would remain as a contribution for future iGEM teams wanting to work on cancer of miRNA detection using microfluidic devices.

The first step in designing a microfluidic chip is knowing how we want it to function. Therefore, the design of the microchannels is very important. In general, you want the microchannels to be the smallest possible while using low flow rates. Basic knowledge of fluid mechanics can be useful for this step. The size of the microchannels (their diameter), flow rate and the viscosity of the fluid are all key components to determine the Reynolds number (Re), a dimensionless number that helps predict flow patterns in different fluid flow situations. Inside the microchannels, for a better flow, you want to keep the flow laminar, which means that Re < 2000.¹¹

where: $$\rho$$ is the density of the fluid (kg/m3)
$$u$$ is the flow speed (m/s)
$$L$$ is a characteristic linear dimension (m)
$$\mu$$ is the dynamic viscosity of the fluid (Pa.s or N.s/m² or kg/m.s)
$$\nu$$ is the kinematic viscosity of the fluid (m²/s)

We designed our microchannels with the Clewin software, for drawing 2D and 3D models with great accuracy. Other softwares such as AutoCAD or Solidworks can also be used.

Some guidelines must be respected when designing the device, so that when you build the chip, you won’t be surprised by a difference in quality or efficiency¹²:

• Two adjacent channels should be separated by at least twice their width, and never less than 200 µm
• The minimum distance from the edges of the chip should be 2 mm
• Minimum feature depth should be 5 µm
• Minimum thickness of the device should be 500 µm
• Maximum feature width should be 4 mm

Our microchannels have a diameter of 200 µm. Therefore, the distance between adjacent channels is 400 µm. Our device includes a Y-shaped channel (3 mm length), with 2 ports (500 µm radius) for the inputs : one for the sample, and one for a lysis buffer. The two liquids merge together in a curved mixer channel (6 cm total length). The resulting liquid is then joined by a cell-free system solution containing our toehold switches from a port (500 µm), which are joining the main channel via a curve channel (total length is 5 mm). The solutions are then merged together in another curved mixer channel (5,5 cm total length). The final products are collected in a port (500 µm radius), serving as the output of our device. Below you can see our 2D and 3D models for our microfluidic device. The whole device (microchannels + extra space around it) measures approximately 4 x 2.5 cm.

How to build a microfluidic chip?

Microfluidic chips can be built in several materials : glass, silicon, or polymers such as PolyDiMethylSiloxane (PDMS). We only focus on this latest option for our microfluidic device. The microfluidic pattern must be printed beforehand on a photolithography mask.

The first step is to mix the PDMS. The base and a curing agent are mixed together in a falcon tube in a 10:1 ratio or 10:2 ratio, depending on the stiffness you want. Less than 10 mL of this PDMS solution should be enough to build some chips. The tube is then centrifuged for 30 seconds at 3000 rpm. The solution is then poured into a mold until it is covered halfway up to the top. All the excess solution can be stored in the freezer at -20°C to be used later for other chips.

The second step is degassing the PDMS, to get rid of the bubbles. It can be done in a vacuum chamber or in a fridge. All you have to do is put the chip in a vacuum chamber for 30 minutes or in the fridge overnight, and the bubbles will disappear.

The third step is the baking of the PDMS. Preheat an oven and prepare a Petri dish to insert the mould into the oven. You can use an aluminium foil if you bake at more than 100°C to avoid the Petri dish from melting. The baking should take between 2 and 3 hours, depending on the amount of curing agent you mixed in the first step.

The final step is assembly and cleaning of the chip. Once the chip is baked, cut the chip from the mould with a scalpel. Use a needle to make holes into the chip. Repeat this operation for each one of the inlets you need. Then you want to clean the PDMS. Cover the channels with scotch tape to take off eventual dust. Wash the chip in isopropanol. Finally, cover the sides of your chip with glass slides to protect the chip. Make sure compatible holes are made on the glass slides.

There you go! You should now have a PDMS microfluidic chip! ¹³

Going further

Unfortunately for our team, we were not able to actually use this microfluidic chip to test our detection system, so this remains only a model. We hope that future iGEM teams that are interested in cancer detection and / or microfluidics design or any applications of this device could use this design as a foundation to build on for their project.