Engineering Cycles

Figure 1a. Engineering overview - From Concept to Design.

Figure 1e. Summary of the Engineering development.

Design

Inspiration for the Project

Colorectal cancer (CRC) is among the most prevalent types of cancer in Hong Kong, accounting for about 20% of all cancer cases in recent years [1]. Regrettably, the incidence rate is increasing, with approximately 5000 new cases diagnosed annually [2]. One notable feature of CRC is that it takes years to develop, allowing ample time for early detection and treatment. Research indicates early detection can improve survival rates to over 90% [3]. However, mainstream screening methods, such as colonoscopy, FOBT, and FIT, are severely limited. This inspired our desire to create a non-invasive microRNA-based colorectal cancer detection kit.

After consulting professionals from the biomedical and healthcare fields, along with reviewing the literature, a possible solution emerged: utilizing the upregulation of certain microRNAs during the early stages of CRC in faecal samples. We aim to develop a device that can detect the presence of those microRNAs using toehold switches. This would result in an accessible, accurate, affordable alternative and complement to the current screening methods.

Design of Constructs

1. Choice of microRNA Targets

The targeted microRNA for this project was chosen to meet four critical criteria: significant overexpression in faecal samples from the early stages of CRC. We first conducted a literature review to choose several potential candidates based on published articles.

From the literature review, we discovered several frequently reported microRNAs within our criteria, showing levels within patients significantly elevated compared to healthy controls in the faecal samples of colorectal cancer patients. As such, miR-21 [1], miR-135b [2], miR-92a [3], and miR-29a [4] were preliminarily selected based on the prevalence of their reporting in the literature. While a comprehensive panel of microRNAs can refine diagnostic accuracy, considering our scope and practical limitations, we decided to refine our selection to two microRNAs based on the rationale listed below.

By evaluating previous iGEM teams, namely the UGM-Indonesia 2023 iGEM team (https://2023.igem.wiki/ugm-indonesia/engineering/), the complementary sequence of miR-21 was reported to include a stop codon, which halts protein expression. The team performed a workaround by using a second-generation toehold switch containing multiple components. We decided to deprioritize this candidate since we preferred to work with fewer components (first-generation toehold switch) in the diagnostic kit. The remaining candidates did not contain stop codons within their complementary sequence, which we prioritized in our design.

We then further analyzed miR-135b-5p, miR-92a-3p, and miR-29a based on their specificity and sensitivity in faecal samples. While the three candidates all exhibit good diagnostic capabilities in published studies, their specificity and sensitivity vary. miR-92a-3p, miR-135b-5p, and miR-29a exhibit sensitivity in faecal sample testing for CRC of 87.3% [5], 78% [2], and 59% [6] respectively, and a specificity of 95% [5], 68% [2], and 61% [6] respectively. Based on previous diagnostic performance, we finally decided to base our project on miR-92a-3p and miR-135b-5p due to their superior diagnostic performance among the three toehold candidates.

2. Design of Toehold Switch Constructs

Programmes used for Toehold Switches Design

We employed two programmes — miRADAR (developed by WageningenUR iGEM team 2024), based on Biopython, and NUPACK.

These programs were used to create a panel of toehold switch candidates with varying ability to detect the same microRNA. Each generated toehold switch has a unique sequence, which we use to predict its binding efficiency to the target microRNA based on a series of criteria listed below.

Criteria for evaluating and selecting optimal toehold switches

Several considerations were made during the process of designing a robust and functional toehold switch, and four main criteria were chosen to decide the optimal toehold switch listed as follows:

CG content
A range of optimum CG nucleotide content of 20%-60% was agreed upon after evaluation of external references, as CG bonds have a higher strength than AU bonds, it allows for more sturdy toehold switches, but overflowing of content may cause difficulty in expressing positive results.

Bulge loop size and symmetry
A suitable internal loop size also contributes to how vigorous the structure is, with 1X1, 2X1, and 2X2 being the most ideal configurations.

Unbound or unpaired nucleotides
Unbound or unpaired nucleotides may contribute to incomplete and unspecific reactions (false positives); thus, a lower number of unbound or unpaired nucleotides would yield a more accurate result.

HMFE and TMFE comparison
HMFE (hybrid minimum free energy) and TMFE (toehold minimum free energy) could be detrimental to the function of the toehold switch if the HMFE is higher than the TMFE. In such case, the retaining of the hairpin structure of the toehold switch will be favoured over being linearized by targets, causing switches to be overly stable and maintained at the OFF state. The desirable state would be when the HMFE is significantly lower than the TMFE, then the trigger microRNA can efficiently bind to the toehold switch and readily disrupt the OFF state, activating the toehold switch, thereby achieving effective ON/OFF switch behaviour.

Design of toehold switches 1,2,3 for targeting miR-92a-3p + design of toehold switch 9,8,7 for targeting miR-135b-5p

Figure 2. Toehold 1

Switch 4 has one nucleotide bound to the toehold switch. Thus, it is deemed unfit as it may cause nonspecific and incomplete interactions. Switch 4 is eliminated.

For the remaining switches, by comparison of HMFE (Hybrid minimum free energy) and TMFE (Toehold minimum free energy), the switch having the largest difference between them is the prime comparison. The values will be compared against each other below.

Toehold 2: 21.70000076 kcal/mol
Toehold 3: 22.10000038 kcal/mol
Toehold 5: 22.09999847 kcal/mol

Despite toehold 3 and toehold 5 having an approximately similar minimum free energy difference, toehold 3 was selected due to its slightly higher difference between HMFE and TMFE. Moreover, toehold 3 had a more desirable inner loop structure compared to toehold 5’s, having a 3X3 configuration compared to 5’s 1X3.

After applying the aforementioned criteria, Fig. 2 (Toehold 1) was deemed to be the exemplary candidate for the design of the toehold switch.

Figure 3. Toehold 3

Figure 4. Toehold 9

As seen in Fig. 4, the hybrid’s bonds are relatively weak, allowing a comparatively effortless separation when the reaction occurs. It has 0.065uM over 1uM bonded into a secondary structure (over 96%), with a low minimal free energy of -48.12 kcal/mol, indicating a firm structure. The internal bonds of the toehold switch 9 are robust, resulting in an effective bonding with miR-135b-5p.

Figure 5. Toehold 8

Toehold switch 8 followed the rules of having strong internal bonds, simultaneously having a low minimal free energy of -48.09 kcal/mol, being proportionately stable. Additionally, its hairpin structure is resolved, with all the remaining bonds not being as sturdy. Toehold switch 8 is also able to bind well with our desired miR-135b-5p.

Figure 6. Toehold 7

The above toehold switch 7 has a significant equilibrium stability of bonds when binding to miR-135b-5p; while its hybrid bond is weak, its internal bonds are strong.

Furthermore, nearly all miR-135b-5p binds to the toehold switch 7. Bottoming at a minimum free energy of -46.56 kcal/mol, which matches our criteria when choosing an optimal toehold switch for future binding.

- Hybrid’s bonds are weak
- 0.965um over 1um bonded into a hybrid/secondary structure (over 96%)
- Low minimal free energy, -48 kcal/mol
- The toehold switch 9’s bond is relatively strong
- All toehold switches bonded with miR-135b-5p
- The toehold switch’s bonds are strong (>0.5)

Alterations to the program

Error
While generating a batch of new toehold switch configurations for a different microRNA, an indexing error occurred during the first run of the program caused by invalid codons, where the total number of nucleotides of the microRNA was indivisible by 3. This caused the program to crash due to the input of an invalid key into the dictionary for searching, as the dictionary in the program only had keys for triplets of nucleotides.

Patches
Assisted by AI, an additional if statement was added to the program, which validates if the codon is divisible by 3, halting the program instead of crashing it (via breaking). Moreover, a check was added to handle cases where no start codons were found, passing if none were found, and also checked whether the codons were valid by searching through the codon tables, stopping the program if it was not found, rather than allowing the program to crash.

Build

Cloning of Toehold Switches

1. Plasmid selection

We chose the pET-14b-EGFP as the plasmid for inserting the toehold switches. It is because the plasmid already contains a T7 promoter (enabling strong, inducible transcription using T7 RNA polymerase—ideal for cell-free protein synthesis systems), an EGFP reporter (providing a sensitive, quantifiable output for switch activation), and a T7 terminator. It also has an AmpR gene, which lets the bacteria gain Ampicillin resistance for selection.

Figure 7. Structure of pET-14b-EGFP

2. Cloning of Toehold Switches 1,2,3,7,8,9 (BBa_25ZJK2J7, BBa_2528ZLEG,BBa_25I8AOF6, BBa_25LB2PQH, BBa_252PS1IQ, and BBa_25I8RMIB)

Unsuccessful attempts

For our first attempt, we transformed the original plasmid into bacterial cells and cultured them for cloning of the purchased original plasmid. The bacterial cell chosen is DH5α, as toehold switches often involve precise RNA sequences and regulatory elements. Moreover, DH5α’s recombination-deficient background ensures these elements remain intact during replication. Last but not least, DH5α is a common laboratory bacterial strain used for routine cloning procedures.

After carrying out miniprep of the plasmid, we tested the concentration of the plasmid we collected, and the final results were lower than the normal range.

We then carried out digestion, NcoI and NdeI are chosen as the restriction enzyme sites. The restriction enzyme NcoI recognizes the DNA sequence C/CATGG and cleaves between the C and the first A. The restriction enzyme NdeI recognizes and cleaves the sequence 5'-CATATG-3'. The reason for choosing the two sites is to allow us to add the toehold switch between the T7 promoter and the EGFP. This allows the T7 promoter to synthesize our toehold switch, EGFP (reporter) to be expressed, and the T7 terminator to stop synthesizing our RNA product. We will add the restriction sites to our oligo (toehold switch), allowing us to cut both the vector and insert.

Figure 8. Position of relevant restriction enzyme sites

We utilize restriction enzymes NcoI and NdeI to ligate the plasmid, the working principle of which is demonstrated in the diagram below.

Figure 9. Working principle of restriction enzyme using EcoRI as an example

The samples underwent gel electrophoresis, and the suitable part of the gel was cut out. We then attempt to ligate the toehold switches into the plasmid.

Using the above principle, multiple attempts in ligation of the plasmid and the toehold switches 1 and 2 were made. However, after transforming the plasmids into the bacterial cell, and upon sequencing, it was discovered that the toehold switches were not ligated to the plasmid in all attempts.

We suspected that the result is caused by the following reasons:
- Some procedural failings and not using enough bacteria for the minipreps.
- Some contamination issues during gel cutting and excessive agar gel cutting.
- Some issues during ligation. The first batch of oligos ordered did not include a 5’ phosphate group modification for the formation of phosphodiester bonds, which prevents the toehold switch from being inserted in the plasmid.
- The ratio of concentration of the plasmid to the toehold switch is not optimal.

We have made the following modifications to plasmid extraction and ligation upon some additional research and consultation with our advisor:

Initially, to enhance the yield from miniprep, we doubled the volume of bacteria used to 4 mL and used vortex instead of just pipetting up and down to resuspend the cells more thoroughly, as vortex should be able to ensure complete resuspension and can increase the final yield after all the procedures.

After multiple attempts, the repeated minipreps are found to be inefficient, and we would like to obtain a higher concentration of extracted plasmid from bacteria at once for future use as well. We decided to use maxiprep instead of miniprep. More consistent data is also believed to be found since we would be using the plasmids extracted from the same amount. We were pleased to see that a much higher concentration of uncut plasmid was extracted after the maxiprep.

To enhance ligation, we ensured that the oligos ordered from the company include a 5’ phosphate group modification for the formation of phosphodiester bonds, allowing the toehold switch to be inserted in the plasmid.

Successful Cloning

Insights gained

Drawing on insights from previous trials and implementing the modifications outlined above, we opted to increase the concentration of the toehold switch insert in the ligation process. We experimented on plasmid-to-toehold switch concentration ratio 1:10 in an attempt to optimize the ligation efficiency. Additionally, we decided that an extension of the ligation time was also a reasonable approach to increase the likelihood of successful ligation. Ultimately, by using another 1:20 plasmid-to-toehold switch concentration and conducting overnight ligation, we successfully ligated one toehold switch (toehold 8) into the pET-14b-EGFP plasmid and transformed the recombinant plasmids into E. coli, as confirmed by colony growth and positive DNA sequencing results.

Positive cloning of toehold switch 8

Positive control
The positive control contains E. coli that has only the plasmid vector transformed (without the toehold switch insert), as the plasmid already contains an antibiotic resistance gene, which ensures the E. coli colonies can be grown.

Recombinant plasmid
The E. coli was transformed with the recombinant plasmid (containing the toehold switch 8 insert and the antibiotic resistance gene) and grew into cultures.

Negative control
The negative control contains E. coli transformed with cut and linearized plasmids, thus E. coli were killed by antibiotics with no colonies found.

Figure 10. E.coli colonies observed on plates for the positive control (left) and toehold switch 8 recombinant plasmids (middle), but absent on the negative control plate (right).

Figure 11. Sequencing result of toehold switch 8 recombinant plasmids extracted from E.coli colonies

The transformed plasmids were sent to DNA sequencing and we confirmed the toehold switch 8 insert was successfully ligated into our plasmid.

Due to time constraints, we decided to order synthesized plasmids for the remaining toehold switch candidates.

Test

Test of Reporter Signals upon binding of toehold switches to microRNA targets

1. Binding of Toehold Switches 1,2,3 (BBa_252B5JD8, BBa_25I9UFIG, BBa_25FRRH5T) to miR-92a-3p

To test the binding of toehold switches, we use an IVE kit, which allows for transcription and translation of the plasmid gene in a cell-free system. We mix 5µM of microRNA samples with the plasmid with or without the toehold switches and observe the samples under blue light. Green fluorescence would be observed if the microRNA successfully binds with the toehold switches and opens the hairpin structure of the toehold and allows for the expression of EGFP.

After initial testing, a minimal difference in terms of fluorescence was found between the positive and negative control, suggesting that optimization of the positive control is required.

Using a plate reader, we tested the fluorescence under different concentrations of pET plasmid and concluded that a plasmid concentration of 7.5 ng in each tube (11µL).

Upon our initial testing using a plate reader, we found that for toeholds 1 and 2, the fluorescence increases, indicating them as promising toehold switches. However, the magnitude of the fold change is less than optimal, as the fold change ranges from 1.595 to 2.251. It is concluded from the data that toeholds 1 and 2 seem to be the most promising for the detection of miR-92a-3p. However, it is believed that the current fold change is not enough to constitute a qualitative test. For toehold 3 and the plasmid alone, a decrease in fluorescence is found. It is speculated that for the plasmid, the microRNA sample contains buffer or other biomolecules that scatter the light. For toehold 3, unintended intramolecular base-pairing occludes the designed single-stranded toehold segment, stabilizing the “off” state of the toehold, which may lead to a decrease in gene expression. This suggests that toehold 3 is not a suitable candidate, and the target microRNA sample may decrease the fluorescence, suggesting that the method of fluorescence currently chosen may not be the most optimal.

With the result in mind, we believed other means of protein expression, such as colored protein and LacZ reporter, should be tested.

2. Binding of Toehold Switches 7,8,9 (BBa_250ZV6M0, BBa_250JCZ3C, BBa_25SFF6VD) to miR-135b-5p

In our experiments testing the IVE kit with toehold 7, 8, and 9, valid results could not be obtained due to experimental errors occurring where the background control level of the setups was higher than the negative control group, which rendered the experimental readouts of toehold 7, 8, and 9 groups to be invalid.

Therefore, due to time constraints, there are no presentable data on the binding of toehold 7, 8, and 9 as of the competition. We foresee further testing to be carried out in the future.

Learn

As we aim to develop a kit with convenience as one of the main goals, the requirement of a blue light to expose the EGFP to the naked eye has been taken into consideration as an inconvenience. Moreover, it was established that the presence of microRNA would affect the fluorescence of samples, making EGFP not the most optimal option for a reporter gene.

Suggestions from students and professors noted that the EGFP could be replaced with a colored protein, which bypasses the requirement of blue light for results, although this has yet to be tested as a viable option. Another alternative raised by professionals is the use of the LacZ reporter gene, which can also be explored.

To optimize the kit as a qualitative test, it is believed that the use of threshold gates, which would allow the toehold to open if only a certain level of microRNAs is detected, can be explored. We could do so by using Toehold-Mediated Strand Displacement (TMSD). We can design a threshold strand that competes with the target microRNA for binding to the toehold switch or a strand that is complementary to the microRNA sequence. At low microRNA concentrations, the threshold strand sequesters the switch, preventing activation. Once microRNA levels surpass the threshold, it outcompetes the threshold strand, triggering the switch.

Another optimization would be setting an appropriate read time. Experiments can be carried out using the level of CRC patients and a normal sample to determine the time taken for observable changes to occur.

References