Amplification and Threshold Selection of miRNAs

miRNA amplification using NASBA

Toehold switches have a detection limit that is 10^3 to 10^6 higher than the concentrations of miRNAs in blood. This means we needed to engineer an amplification step upstream of our detection module that comprises the toehold switch circuit. Our chosen amplification technique is NASBA, a one-pot, isothermal amplification technique for RNA which can give exponential amplification up to a fold change of 10^9 relative to original levels.¹ The technique consists of two sub-reactions: reverse transcription (RT) and in vitro transcription (IVT). Initially, we aimed to reproduce the two sub-reactions separately. First, we performed RT with miRNA hsa-miR-484, which was found to be upregulated specifically with Relapse-Remitting MS.² PCR was used on the RT products to confirm the production of DNA during RT. Bands with DNA of the expected size, 144 bp, were obtained, which confirmed the RT reaction was successful (Figure 1).

Since RT is the first essential step in the NASBA amplification reaction, we investigated the detection limit of this sub-reaction. We performed RT for miRNA concentrations ranging from 70 nM to 7 fM, and the RT products were used as templates for a PCR. We found that RT is convincing for an miRNA concentration as low as 700 pM. This can be seen by clear bands at 144 bp, the same height as the positive control, as well as an increase in faintness of the bands below 100, which are indicative of the primers. Notably, the bands in the negative controls indicate a possible contamination in the PCR mastermix. However, the experiment could not be repeated due to time constraints. Further optimisation of RT Is required to amplify miRNA from a blood sample directly, due to miRNA concentration in the fM-pM range. Alternatively, a concentration step of isolated RNA from blood is required beforehand.

The second sub-reaction in the NASBA amplification is IVT. During this step, the DNA template produced during RT is transcribed to obtain RNA products containing the selected miRNA sequence. We prepared the dsDNA template by doing PCR on the RT product. During PCR, the T7 promoter was incorporated in the forward primer to add the T7 promoter upstream of the miRNA sequence. The IVT was successful as shown by new bands at 122 nt underneath the template DNA band at 144 bp (Figure 3, left).

After IVT, the effect of DNase I treatment and RNA purification on RNA stability, yield and purity was investigated. DNase I treatment was not fully achieved since the DNA template band is still visible at 144 bp after treatment (Figure 3, centre). However, the DNA template has almost fully degraded after RNA purification, since there is no clear DNA template band left. This indicates our sample contained pure RNA product. In addition, the intensity of the bands after RNA purification shows that the RNA product is well-concentrated (Figure 3, right). It is important to note that during the RNA purification step, the RNA was concentrated due to the elution of the product with a lower amount of buffer. During the DNase I treatment, the RNA was slightly diluted. After these steps, the samples were not normalised to RNA amount before loading on the gel, since this step was mainly performed to show that the chosen methods qualitatively work.

To conclude, we showed that miRNA hsa-miR-484 can be reverse transcribed when present at a concentration of 700 pM or higher. The obtained DNA template can produce a high yield of RNA product during IVT. Through downstream processing, this RNA product can be purified and concentrated for the next step in our test.

Establishing miRNA threshold with TMSD

For establishing the threshold of miRNAs in our test, we used a Toehold-Mediated Strand Displacement (TMSD) threshold system, further referred to as TTS. The TTS system consists of four composite parts: BBa_K5106015, BBa_K5106016, BBa_K5106017, and BBa_K5106018. After we designed and confirmed the sequences of the TTS in NUPACK, we converted them into DNA and combined them with a T7 promotor. This allowed us to order the sequences as DNA oligos and transcribe them using in vitro transcription (IVT) back into RNA. The T7 promotor allows RNA polymerase to bind and amplify the RNA sequence at 37 ^{\circ}C. This was done in a thermocycler to ensure optimal temperature and minimal evaporation. RNA from IVT was directly loaded in a 96-black clear bottom plate for fluorescence detection. No purification was performed as this caused irredeemable misfolding of the fluorescent aptamer. Spinach-DFHBI-1t fluorescence was measured at 470/505 nm at 25 ^{\circ}C, 30 minutes after adding the RNA fragments to allow the mixture to reach equilibrium.

Alternatively, RNA was transcribed in a plate reader at 37 ^{\circ}C to directly measure the fluorescence and thereby the concentration of RNA in the system. DNA with IVT buffer and enzymes was directly loaded in a 96-well black clear bottom plate for fluorescence detection. To obtain a fluorescent signal, C and D had to be co-transcribed to result in the correct folding of the fluorescent aptamer. Spinach-DFHBI-1t fluorescence was measured at 470/505 nm at 37 ^{\circ}C in 30-minute intervals, to reduce photobleaching.

We observed that fluorescence was produced in the presence of TTS part D alone (Figure 4), this background signal was to be expected. Since it was less than 20% of the signal of CD, this was deemed acceptable. Notably more fluorescence was observed when part C and D were co-transcribed in the same system, indicating that co-transcription is necessary to ensure correct folding. When fragment C is produced separately, no fluorescent output signal was observed as expected. This was also the case for the negative control. The positive control sample consisting of the spinach-2 fluorescent aptamer also showed no fluorescence. This was hypothesised to be caused by misfolding of the RNA during IVT. This was shown by heat denaturation of the spinach-2 aptamer, after which the RNA was able to refold, and fluorescence was restored (not shown).

After the first experiment in which we showed that TTS part C and D can produce fluorescence, RNA strand A was added to the reaction mix. This was to produce a threshold, opposite of that originally designed, where the subsequent addition of B would inactivate the fluorescence. Since A should only have affinity for strand B, no change was expected. However, an almost 3-fold higher fluorescent signal was observed after the addition of A (Figure 5). Since A itself cannot anneal and stabilise DFHBI-1T, it was speculated that it might have affinity for D.

This was confirmed by a NUPACK analysis (Figure 6), which indicated that A could anneal with D. This was missed since in previous analyses C and D were expected, and therefore modelled, to be at equal concentration. If the concentration of D is greater than that of C, A can anneal to D. Since NUPACK showed these criteria are essential for this reaction to occur, we must assume this is what caused the increase in fluorescence due to the addition of A (Figure 6). This is in line with protocols about IVT, which state that longer RNA sequences transcribe more efficiently than short RNA sequences. Therefore we hypothesise that; 1: Due to the longer DNA sequence, fragment D has been transcribed more; 2: Due to the anti-sense compatibility A can anneal to D (when C is present in smaller concentrations than D), and induce fluorescence.

To test the full functionality of the TTS system, we investigated how well part B could inhibit the fluorescent signal caused by part C,D and A. This is a simplified (and reverse) version of the TTS system, as it was unsure whether C could again anneal to D and produce fluorescence after release by B. This system was predicted to give a similar threshold reaction, however the threshold would be based on the concentration A, and the signal would be inhibited once the concentration of B exceeded that of A.

Since the annealing affinity between part B and C (and especially A and B) is higher than between part C and D, addition of part B should lead to a decrease in the observed fluorescence. The fluorescence of CD was measured at 470/505 nm with varying concentrations strand B (Figure 7). It was observed that fluorescence indeed went down over time with the addition of part B. In addition, fluorescence went down to a plateau that was also determined by the concentration of inhibitor B, whereby adding more B resulted in lower fluorescence. This indicated that part B can indeed inhibit fluorescence by annealing to part C, and can thus be used to create the TTS system together with part C and D.

However, since the fluorescence of all wells dropped immediately without any threshold function being observed, therefore we hypothesise that C/A has not saturated all of D, therefore when any B is present, CD/AD pair is split up and fluorescence decreases.

Altogether, we showed that fluorescence caused by the secondary structure formed by part C and D together works. In addition, inhibition of part B is also successful. The next step would be proving that the inhibition of B, only occurs once the threshold of A is exceeded. Secondly it must be shown that A can release part B again as required for the complete TTS system.

Detection of target miRNAs

We designed two toehold switches for the detection of the miRNA hsa-miR-484. These toehold switches, referred to as A and B (BBa_K5106001 and BBa_K5106002), were constructed from our improved software for toehold switch design. A third toehold switch, referred to as C (BBa_K5106003), was designed using just the NUPACK python software tool. In addition, two other toehold switches were tested, referred to as 1 and 2. These were selected from prior research, the first from iGEM team CLSB-UK 2017, the latter from Pardee et al. (2014).³ However, we were not able to test them with their own trigger (mi)RNA so they will not be discussed further. Finally, a previously reported toehold AND gate (BBa_K5106007) was produced and tested.⁴

Detection of miRNAs using toehold switches

We performed in vitro experiments of the MS-specific toehold switches A, B, and C. The respective toehold switches were expressed on a plasmid under the control of the T7 promoter and terminator, followed by the lacZ gene which encodes \beta-galactosidase (composite parts BBa_K5106004, BBa_K5106005 and BBa_K5106006). This enzyme hydrolyses the substrate Chlorophenol Red-\beta-D-Galactopyranoside (CPRG) to Chlorophenol Red (CPR), resulting in a colour change from yellow to purple.

To check whether the constructs were activated by miRNA, we performed a test in 2 \muL PURExpress, with \sim 25 ng plasmid containing the respective toehold switch. A negative control, containing no DNA template, and a positive control, containing only lacZ under the control of the T7 promoter and terminator, were included as well. In addition, 0.5 \muL miRNA hsa-miR-484 (5 \muM) was added to part of the samples. The tubes were incubated at 37 ^{\circ}C for two hours.

After two hours, \beta-galactosidase (LacZ) performed as expected, since no colour change from yellow to purple was observed in the negative control, whereas the colour change was observed in the positive control (Figure 8). For the three different MS-specific toehold switches, we observed that the visible colour change was strongest for toehold switch B, followed by A, and that toehold switch C does not show a clear difference between presence or absence of the miRNA trigger. In addition, the three toehold switches do not appear to be leaky, since we did not observe the colour change when no trigger was added (Figure 8). Expression of toehold switch B did cause slight discolouration, but the difference between the absence and presence of the trigger miRNA was still distinguishable by eye.

In addition to this qualitative test, we performed a quantitative measurement to evaluate the performance of each toehold switch. For this, 1 \muL trigger miRNA (5 \muM) was added to 5 \muL PURExpress containing \sim 30 ng of DNA template. The tubes were incubated at 37 ^{\circ}C for two hours. Every 20 minutes, a 0.5 \muL sample was taken and diluted in 75 \muL nuclease-free water. We measured the absorption at 570 nm of the diluted samples using a plate reader.

Similar to the qualitative test, toehold switch B produced the biggest absolute increase in output signal over time, indicating that this toehold switch is more efficient than our other constructs (Figure 9). However, we saw that this switch also showed the highest signal without trigger RNA present, and is thus most leaky. Toehold switch A was also activated by the trigger miRNA, but has considerably less output signal compared to toehold switch B. Similar to B, toehold switch A produces an increase in output signal over time, even when no trigger is present, and is therefore also leaky. Toehold switch C did not show any increase in absorption over time. We performed a one-sided t-test (df=4) to compare the absorption in the samples with and without trigger miRNA. Both toehold switch A and toehold switch B gave significantly different signal output in the presence and absence of trigger miRNA after three hours (p = 0.0093 and p = 0.0048).

Toehold switches A and B were designed using an improved version of the SwitchMi Designer software from iGEM team UParis BME 2021, whereas toehold switch C was designed with a simpler approach, using only NUPACK. Since toehold switches A and B were activated by miRNA and toehold switch C was not, we can conclude that a dedicated tool for toehold switch design improves the chance of designing a working toehold switch. It is therefore advised to use our toehold switch design software for future design of toehold switches that can regulate expression based on the presence of a known miRNA sequence.

Furthermore, it is important to note that we used different methods of analysis for the qualitative (Figure 8) and quantitative tests (Figure 9). In both cases, the colour change or absorption was clear in cases where the trigger miRNA has-miR-484 was added. In the more extensive quantitative measurements, we saw an increase in absorbance for the samples without the addition of trigger miRNA as well. In addition, the difference in colour we observed between the presence and absence of trigger miRNA was expected to be larger, but the exact reason for this apparent leakiness remains to be investigated. The data obtained from this experiment was fitted to our continuous ODE model to look for sources of leakiness and the overall dynamics of the toehold switches.

From the combined results, we can conclude that toehold switch B works the best of the three different toehold switches made for the detection of hsa-miR-484, upregulated with Relapse-Remitting MS.² Therefore, this construct was used in further testing of our cell-free paper-based test system.

Toehold switch AND gate

We performed in vitro experiments of the AND gate designed by Green et al. (2017).⁴ The AND gate was expressed on a plasmid under the control of the T7 promoter and terminator, followed by the lacZ gene. This AND gate is activated by RNA. In the same experiment as the toehold switches, we performed a test in 2 \muL PURExpress with \sim 25 ng plasmid containing the AND gate. The same positive and negative controls were included. To activate the AND gate, 0.5 \muL of both trigger RNAs (5 \muM) was added to one sample. The tubes were incubated at 37 ^{\circ}C for two hours.

For the first qualitative test (Figure 10), the AND gate showed a clear colour change when both trigger RNAs were added. We also observed leakiness, since the tube without trigger RNA shows an orange colour, instead of the yellow we see in the negative control. However, to determine how the AND gate functions over time, we performed a quantitative in vitro measurement on samples with no trigger RNA, with either one of the two trigger RNAs and with both triggers present (Figure 11). For this, 1 \muL trigger RNA (5 \muM) was added to 5 \muL PURExpress containing \sim 30 ng of DNA template. The tubes were incubated at 37 ^{\circ}C for two hours. Every 20 minutes, a 0.5 \muL sample was taken and diluted in 75 \muL nuclease-free water. We measured the absorption at 570 nm of the diluted samples using a plate reader.

In this measurement, we observe that the toehold switch AND gate shows an increase in absorption over time in the presence of both triggers, indicating that the AND gate was indeed activated by RNA. However, we also observed notable leakiness in the system with all the samples producing the same level of absorption after 180 minutes. However, the difference between the presence and absence of trigger RNAs between 80 and 120 minutes was still considerable. This indicates that after two hours, the toehold switch AND gate is less applicable to a diagnostic test due to increased chances of false positives. We also performed a one-sided t-test to compare the absorption in the samples with and without trigger RNA, as well as compare between the presence of both trigger RNAs and only one of the trigger RNAs. After 120 minutes, the presence of both trigger RNAs caused significantly different absorption values when compared to the samples where no trigger RNAs were added (p < 0.001). After 180 minutes this difference was still significant. Additionally, the difference in absorption between the presence of both trigger RNAs and the presence of either one of the RNAs was significant (p < 0.001 for T1 and p = 0.0048 for T2). However, after 180 minutes, these differences were no longer significant, suggesting that the AND gate was partially activated by single RNAs.

The source of leakiness requires further research. This data was also fitted to our continuous ODE model to look at the dynamics of the system. Furthermore, the differences between the absence of trigger RNA and the presence of either trigger RNA is significant after 180 minutes (p = 0.0015 for T1, p < 0.001 for T2), whereas it is not significant at 120 minutes. This result, coupled to the quantitative measurement of the MS-specific toehold switches, also suggests that it may be necessary to read the test output directly after incubation to ensure an accurate test result. In the final test, the output is based on the colour change so no imagers are necessary. To fully characterise the toehold switch and AND gate output over time, we could investigate alternatives to absorption in the future, such as UV-Vis or colourimetry, to visualise the differences between the presence and absence of trigger miRNA more clearly.

Paper test platform and output signal

Cell-free paper-based test system

We tested the expression of toehold switch B and the AND gate in paper. Cell-free expression of toehold switches in paper was first reported by Pardee et al. (2014), which is the method we followed as well.³ In brief, filter paper was pre-treated with Bovine Serum Albumin (BSA) to block nonspecific interactions between the cell-free components and the cellulose matrix. After drying, 2 mm discs were cut out and put into a 384-well plate. Cell-free reactions (PURExpress) were assembled in 2 \muL volumes and applied to the discs, which were subsequently freeze-dried. After 24h, the discs were rehydrated with 2 \muL nuclease-free water with or without trigger miRNA. The discs were incubated at 37 ^{\circ}C for two hours.

We showed that this method of producing freeze-dried tests and rehydrating them activates the cell-free transcription and translation machinery (Figure 12). The negative control discs remained yellow, while the positive control discs turned purple. We have also shown that the discs containing the toehold switches can directly be rehydrated with trigger miRNA, simulating the final testing environment. The discs with toehold switch B were rehydrated with hsa-miR-484, which activated the toehold switch leading to a purple colour. The discs containing the AND gate were reactivated with both trigger RNAs, leading to a purple colour. Freeze-drying the reactions onto paper ensures that the biochemical components are stable during long-term storage at room temperature, and by working in small volumes we require fewer materials per disc. These properties contribute to the accessibility of the test.

Alternative signal output

We have performed several experiments to design a proof-of-principle DNAzyme and tested whether DNAzymes are directly compatible for detection of the miRADAR toehold switches or that reaction conditions should be optimised. We theorised that hemin should be added after the IVT in order to prevent unwanted interference by oxidation and briefly incubated at room temperature. IGEM Leiden 2020 already described negative effects of dithiothreitol (DTT) on the DNAzyme activity. Therefore, reaction buffers containing DTT should be avoided or otherwise adjusted accordingly. In addition, we found that the presence of antioxidants (\beta-mercaptoethanol (\betaME) and ethylenediaminetetraacetic acid (EDTA)) in the T7 storage buffer inhibits the final reaction from taking place, however by increasing the hemin concentration (500 nM to \sim 2.0 – 2.5 \muM in final reaction mix) this could be overcome. Finally, freshly dissolved ABTS and H_2O_2 were added and a colour change could be observed within 3 to 20 minutes (Figure 13).

During trouble shooting we tried a new buffer composition designed by Chen et al. (2020) which replaces the conventional potassium and/or sodium salts with NH4Cl, which was found to improve the colour formation and retention (Figure 14).⁵

Combining G4/hemin probes with our designed toehold switches has not been successful as of yet. Some probes were able to produce an increased signal when a DNA variant of the target sequence was present, though hybridisation with the RNA switch was not achieved (Figure 15). Additionally, some probes produced a background signal when no target was present, meaning there is a slight level of auto-hybridisation of the DNAzymes (Figure 15).