The main goal of our iGEM project was to design an accurate and accessible diagnostic test to simplify the diagnosis of multiple sclerosis (MS). We divided our main goal into sub-parts to work efficiently. Our team investigated microRNA (miRNA) amplification, designed an RNA threshold system, and designed and tested in vitro miRNA-targeting toehold switches. Throughout our project, we used the Design-Build-Test-Learn (DBTL)-cycle to effectively work towards our goals. The two DBTL-cycles that we would like to highlight are the cycle of making accurate miRNA-targeting toehold switches and the cycle of developing an RNA threshold through toehold-mediated strand displacement.

RNA threshold detection system

During this engineering cycle, we focused on the design of an RNA threshold system for our diagnostic test. We designed the required RNA strands, determined the desired ratios informed by our model, and tested the behaviour of the system in vitro. The results were used to adapt the model for redesign of the system in a new round of the engineering cycle.

Introduction

miRADAR designed a diagnostic test for MS based on concentration-dependent miRNA detection. When a patient suffers from a disease, miRNA concentrations in the blood can change. A certain disease can therefore cause a specific pattern of upregulated or downregulated miRNA levels in blood. For our diagnostic test, it was essential to be able to distinguish between miRNA levels associated with healthy people and MS patients. We therefore worked on an RNA threshold system to ensure the test only gets activated once the miRNA levels meet the level that is associated with MS in literature.

Round 1

Design
The miRNA level that distinguishes between healthy people and MS patients, which we refer to as the ’threshold level’, can differ for each miRNA. Therefore, the threshold level of our test must be tunable. In addition, the threshold system must give an RNA output that only activates our toehold switch when the threshold is met (Figure 1a). We utilised the properties of a toehold-mediated strand displacement (TMSD) system, like the one developed by Zhang et al..¹ The TMSD system uses a combination of four RNA strands that (partially) anneal to each other: an input RNA strand (A) (BBa_K5106011), an inhibitor strand (B) (BBa_K5106012), a trigger strand (C) (BBa_K5106013), and an output RNA strand (D) (BBa_K5106014). Our threshold system contains an excess of strand B, equivalent to the desired threshold of the test. Strand B and C intially anneal (Figure 1b), and input strand A anneals to the free strand B molecules. If strand A exceeds the set threshold, there will be free strand A after binding to all free strand B molecules. Strand A will then displace strand C to anneal to strand B (Figure 1b), since strand B has a stronger affinity for strand A than for strand C. Free strand C molecules then activate output strand D, leading to an output signal above the set threshold. To study our threshold system independently from other parts of our test, we designed strands A, B, and C to be compatible with the inducible spinach aptamer output strand as designed by Wang et al..² Spinach is a folded RNA structure that can stabilise the side group of a fluorophore molecule (DFHBI), which allows it to fluoresce. The inducible spinach aptamer output strand contains a misfolded spinach aptamer, that upon annealing to a trigger (in this case strand C), can refold to stabilise DFHBI, leading to fluorescence. For our final test design, the input RNA strand A would be the MS-associated miRNA sequence and the output RNA strand D would be our toehold switch.

Generating an RNA threshold using the TMSD system. a) Our threshold system converts a linear signal output into a binary signal output, which means the test will only be activated above a certain threshold concentration level associated with MS. b) Schematic of toehold-mediated strand displacement with input strand A, inhibitor strand B, and trigger strand C, which activates an output signal.

Optimising the concentration of the different strands and their affinities is crucial to generate the desired threshold level. Both design aspects were supported by modelling data:
1. Our model predicted that the concentration of strand B is of great importance for the binary behaviour of the threshold. We found that the concentration of strand B should be higher than the concentration of strands C and D to give a successful threshold system. The optimal ratio was predicted to be 3:1 for both strand B:C and B:D.
2. Our model predicted that the annealing rate for strand A and strand B should be 1000x higher than the annealing rate for strand A and the duplex BC, to create the desired threshold system. We assumed that a higher affinity leads to a higher annealing rate. Therefore, we ensured that the affinity for strand A and strand B was larger than the affinity for strand B and strand C. The NUPACK software³ was used to confirm the desired differences in affinities of our designed RNA strands.

Build
Once the design of the strands was finalised, we converted the RNA sequences to DNA in silico, and added a T7 promoter (BBa_J64997) upstream to the sequences, to allow for in vitro transcription (IVT) of the RNA strands. In addition, we added a T7 terminator (BBa_K731721) downstream of the output RNA sequence D. The designed DNA sequences for the four strands were ordered as ssDNA oligo’s and annealed so they could be used as template for IVT to obtain high-yield RNA strands as a product.

Test
We performed multiple experiments to investigate whether the RNA strands had the desired functionality. Three experiments are highlighted in this cycle. First, we studied whether folding of the inducible spinach aptamer D led to a fluorescent signal, with and without trigger strand C. We expected to see a fluorescent signal only for the trigger strand C and strand D together, since strand C allows for correct folding of D, and no fluorescence when only strand D was added. We performed the IVT reactions to produce the RNA strands and measured fluorescence over time. The reaction containing strand D alone showed a low fluorescent signal while the co-transcription of strand C and strand D together resulted in a high fluorescent signal (Figure 2). Our positive control, the original spinach aptamer, did not lead to fluorescence, presumably due to misfolding of the RNA strand during IVT. The test confirmed that strand C enables aptamer folding and subsequent fluorophore stabilisation by strand D, although there was modest background fluorescence.

Trigger strand C enables folding of the inducible spinach aptamer strand D, leading to a fluorescent signal. Fluorescent signal over time of the IVT reaction (hours) for strand C, strand D and strand C and D together. Excitation = 470 nm, emission = 505 nm.

After confirming that strands C and D together lead to fluorescence, we studied the effect of the input strand A on the fluorescent signal. In the TMSD system, the input strand should not directly affect the fluorescent signal. We tested the effect of input strand A on the fluoresent signal to make sure that a signal is only obtained through activation of the output strand D by the trigger strand C. However, when strand A was added to strand C and D, we observed an increase in fluorescent signal (Figure 3), which suggested that strand A can also activate folding that leads to fluorescence.

Input strand A enables folding of the inducible spinach aptamer strand D, leading to fluorescence. Fluorescent signal for strands C and D together, compared to strands C, D and A together. Excitation = 470 nm, emission = 505 nm.

Finally, we added the inhibitor strand B to the mixture of strands A, C, and D. Strand B is expected to anneal to the input strand A first. If strand B is in excess of strand A, it also anneals to the trigger strand C, which is expected to cause a decrease in fluorescent signal. Varying amounts of inhibitor strand B were added to the mixture. We observed a decrease in fluorescent signal for all samples which meant strand B was in excess of strand A, allowing strand B to anneal to the trigger strand C (Figure 4). In addition, the decrease in fluorescent signal was higher when more inhibitor strand was added. This indicated that the inhibitor strand annealed to multiple trigger strands C, thereby inactivating a higher level of output strand D.

Inhibitor strand B decreases the fluorescent signal. Fluorescence was measured over time after the addition of strand B (minutes). Excitation = 470 nm, emmission = 505 nm.

Learn
The three experiments conducted provided valuable information about the behaviour of the system. We showed that the trigger strand C activates proper folding of output strand D, leading to fluorescence, although there is modest background fluorescence. In addition, it was confirmed that the inhibitor strand B captures strand C to decrease the fluorescent signal. However, the increase in fluorescence after the addition of the input strand A indicated that strand A can also activate fluorescence, presumably by annealing (partially) to the output strand D. This unexpected interaction indicated that strand D was in excess of strand C. Since the affinity of strands A and D is lower than strands C and D, strand A can only anneal to strand D if strand D is in excess of strand C. This suggested that the in vitro transcription of the output strand D was more efficient than the transcription of the trigger strand C during co-transcription, leading to free molecules of strand D in the mixture. Round one therefore reveals the importance of the ratio between the trigger strand C and output strand D in the reaction mixture. A higher concentration of strand C is required to prevent the input strand A from annealing to, and thus activating, the output strand D.

Round 2

Redesign
We learned that the input strand A can also activate the output strand D and therefore, the ratio of the trigger strand C and output strand D is of great importance for a working threshold. To predict the optimal ratio for new experiments in the lab, our model could be used. In the model we can incorporate the annealing possibility between strands A and D. Next, we can predict the optimal ratio again for both the annealing rates and input concentrations for the RNA strands in the TMSD system. To obtain this predicted optimal ratio of trigger strand C and output strand D in the lab, the co-transcription could be optimised. This could for instance be done by changing the input DNA template concentration in the IVT reaction during future experiments.

(1)

Zhang, D. Y.; Winfree, E. Control of DNA Strand Displacement Kinetics Using Toehold Exchange. Journal of the American Chemical Society 2009, 131 (47), 17303–17314. https://doi.org/10.1021/ja906987s.

(2)

Wang, T.; Simmel, F. C. Switchable Fluorescent Light-Up Aptamers Based on Riboswitch Architectures. Angewandte Chemie International Edition 2023, 62 (41), e202302858. https://doi.org/10.1002/anie.202302858.

(3)

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, 2022. https://doi.org/10.26434/chemrxiv-2022-xv98l.