Identification of miRNA targets

Once a disease develops, a specific combination of miRNAs is dysregulated due to their role in gene regulation.¹ Therefore, it is crucial to select a specific combination of these miRNAs to target when producing a diagnostic test. In our project, this combination was identified using a machine learning algorithm, based on miRNA expression data of relapsing-remitting MS patients and a healthy control group. The algorithm identified a combination of miRNAs which can distinguish between these groups. The levels of all of these miRNAs need to be dysregulated to accurately distinguish between MS patients and healthy control. When applying this to logic gates, the fact that all need to be present before a diagnostic test can give a positive results means that they can be detected by AND gates. We also compared MS to a disease that shares symptoms with MS, further referred to as the mimic disease, that therefore hinders diagnosis. To distinguish MS from this mimic, we identified mimic-specific miRNAs that are dysregulated for patients with the mimic disease, but not MS patients. These miRNAs will be included in the circuit as a NOT gate, since the presence of this miRNA indicates the mimic disease and not MS. Everything together results in an MS-specific combination of dysregulated miRNAs which can distinguish between MS, the mimic disease, and the healthy control (Figure 2).

Amplification and Threshold Detection of miRNAs

Before the selected miRNA combination can be detected with our diagnostic test, concentration or amplification of the miRNA sample is necessary. This is because the miRNAs are present in the blood in the fM - pM range,² whereas the concentration limit for our detection method is in the nM range.³ To achieve this amplification of the miRNAs, we make use of nucleic-acid sequence-based amplification (NASBA). This test can be performed as a one-pot isothermal reaction, which means that the reaction can take place in an incubator with constant temperature, allowing the test to take place at most medical facilities in the Netherlands.⁴ In addition, NASBA allows for exponential amplification with a high fold-change, and is used in other cell-free tests for viral RNA detection.^5,6

NASBA consists of two main parts: DNA template production (Figure 3a), and the RNA amplification cycle (Figure 3b). First, a double-stranded DNA template (dsDNA) is produced. A stem-loop primer anneals to a part of the selected miRNA, which forms the starting point for reverse transcription. The reverse transcriptase enzyme then synthesises a complementary DNA (cDNA) strand using the miRNA as template (Figure 3a (I)). The enzyme RNase H degrades the miRNA in the DNA/RNA duplex (Figure 3a (II)). With the help of a forward primer, reverse transcriptase synthesises the second DNA strand to obtain double-stranded DNA (Figure 3a (III)). The forward primer integrates a T7 promoter sequence upstream of the miRNA sequence, which is required for the RNA amplification cycle. During this second part of NASBA, the enzyme T7 RNA polymerase transcribes RNA products from the dsDNA template (Figure 3b (I)). A universal reverse primer allows for reverse transcription of the RNA product (Figure 3b (II)). After degradation of the RNA product in the DNA/RNA duplex (Figure 3b (III)), the universal forward primer enables synthesis of a new dsDNA template (Figure 3b (IV)). This cycle generates more DNA templates which can be transcribed again, making NASBA an exponential amplification method.⁷

In patients with autoimmune illnesses, there are signs of inflammation that can be detected in the blood, such as the up- or downregulation of miRNA (Figure 4a). These dysregulated miRNA patterns can be used to diagnose patients. However, some of these miRNA dysregulations are subtle, as little as a 5% marginal increase in concentration have been cited as significant in correlation with diseases.⁸ This is a problem because toehold switches increase their output linearly (Figure 4b), meaning a small increase in input will result in a small change in output. This causes an intermediate output rather than a binary one (Figure 4b). A binary output (being ON or OFF) is preferred because this gives a clear signal for MS diagnosis. To distinguish these minor changes in miRNA concentration in patients, a threshold detection system is required. A threshold would mean that only if the miRNA is above a certain concentration, an output signal can be produced (Figure 4b). Since different miRNAs can have different concentrations in the blood, the level of the threshold must be tuneable. This means that for any miRNA, regardless of sequence or concentration, a concentration threshold can be constructed.

One method to construct such a tuneable threshold, is by utilising the properties of toehold mediated strand displacement (TMSD). TMSD is the process of replacing an annealed RNA strand, with one that has a higher binding affinity. This preferential affinity of RNA strands with better annealing, can create a sharp transition in output signal (Figure 4b).

In TMSD there are at least three strands of RNA present: A, B and C (Figure 5). In the initial state, strand B and C are annealed, however strand A has a higher affinity for B. Because there is a free region of RNA on strand B, strand A can partially bind. This is known as the toehold, and is essential to stabilise the kinetic bottleneck in the reaction. When the toehold is at least six bases long, the forward reaction kinetics in the exchange reaction improve, with equilibrium constant K increasing up to the power six.⁹ Once partially bound, strand A can compete with strand C, ending in strand C being displaced from strand B. To create a TMSD threshold you need three RNA strands where all binding affinities are controlled so that the binding energy of RNA pairs is ranked: AB>BC>CD, with D being an external output receptor for C. This is because A must anneal stronger to B, otherwise a very large ratio of A:C is required to displace C.

This alone does not create a threshold, for this we need a surplus of strand B. By creating a pool of free strand B, low levels of strand A will bind to these, not releasing any of strand C. By controlling the concentration of free B, a threshold can be created. In the situation where there is less A present than B, indicated as the OFF state (Figure 6a), A binds to B, but due to the excess of B, no C is released. When A is present in higher concentrations (Figure 6b), the unbound B is sequestered and C is displaced from B. This means output signal C is only released when A is present in concentrations above the set threshold.

The slope of this threshold is dependent on the ratio between B and C. If the ratio of B to C is 1:1 when any A is present, C will be released, causing a linear relation between A and C (Figure 7a). However, when the ratio of B is increased, a larger fraction of A is required to bind the free B than to release the C (Figure 7b). The smaller the fraction of A that releases C becomes, the more the output signal starts to resemble a full binary signal. Computer modelling has shown that a B:C ratio larger than 3 can create a near binary output response (Figure 7c).

This system allows for an easy and modular system for thresholding in cell-free systems. The RNA sequences can be easily designed based on a desired input and/or output sequence, and the threshold can be altered by simply changing the concentration RNA strand B. The output signal C can be modified for any downstream detector (D), which can then detect the output signal of the threshold system C. This downstream detector D can be a toehold switch or an inducible fluorescent aptamer, thereby giving a visible signal in a cell-free test.

Detection of target miRNAs

The detection of released miRNAs after establishing the threshold is executed using toehold switches and logic circuits made from toehold switches. The circuits combine Boolean gates, such as AND, OR or NOT, with toehold switches to add specific selection criteria to the identified miRNA combination we test for. We specifically use toehold switches to detect our miRNA targets, and ribocomputing to detect a combination of these targets.

Toehold switches

miRADAR uses toehold switches to detect miRNAs. A toehold switch is an RNA structure present on the 5’-UTR of a messenger RNA that prevents translation of a gene until a specific trigger RNA is present. This is due to the secondary structure of the switch RNA, which consists of a hairpin structure. This hairpin is formed by the ribosome binding site (RBS) and start codon (AUG), forming a loop and bulge respectively (Figure 8a). After this hairpin, a 21-nucleotide linker sequence and the mRNA of an output gene are present. The toehold domain of the switch RNA is complementary to the trigger RNA, allowing for linear RNA-RNA interactions between the two.¹⁰

When the trigger RNA is not present, the RBS is in the hairpin structure, disabling the possibility for the ribosome to bind, and thereby preventing translation of the protein. However, when the trigger RNA is present, it can bind the toehold domain of the switch, thereby opening the hairpin structure by a strand displacement reaction. This exposes the RBS and start codon, allowing ribosome binding and subsequent mRNA translation (Figure 8b).

Toehold switch circuits

Since the presence of diseases leads to a specific pattern of up- or downregulated miRNAs, we aim to detect a specific combination in one test. The same principle of toehold switches can be extended into a circuit of YES/NO tests, as well as logic gates- e.g. AND, OR, and NOT gates. This results in a Boolean logic circuit that only produces an output if all necessary triggers are present.¹¹

A toehold switch capable of performing Boolean logic functions works with a similar mechanism as the conventional toehold switch, but has a slightly modified structure. We can make a toehold switch capable of AND logic for two input triggers. This means that both triggers need to be present before an output signal could be observed. To modify the structure of a regular toehold switch for this, we add additional hairpins structures, where each hairpin contains a complementary region for a trigger RNA (Figure 9a). After the first trigger binds the switch, the top left hairpin structure is linearized. This results in a conventional toehold switch structure, allowing the second trigger to bind. After binding, complete linearisation of the switch RNA, allowing the ribosome to bind, and translation of repressed protein to start (Figure 9b). This same principle can be used to generate larger AND gates, where more trigger RNAs need to present before an output can be observed. Similarly, this can be used to generate even more complex circuits, including other Boolean gates, such as OR and NOT. In our test, the use of (larger) AND gates allow us to specifically detect multiple miRNAs at once, while generating a single output. The use of more complex logic gates makes it possible to place additional conditions on the system, making a toehold switch circuit completely customisable.¹¹

Paper test platform and output signal

After the detection of the disease-specific combination of miRNAs, the desired easy-to-read output is generated on a simple test platform. To make sure that the test fits our accessible requirements, we develop a paper-based blood test. miRADAR uses the commercial kit PURExpress^{\tiny{\textregistered}}, which is based on the PURE System Technology,¹³ to which we add the components necessary for establishing the threshold and the detection of the trigger miRNAs. After embedding all these parts into paper discs, these can be freeze-dried. This process results in a high stability of the components, thereby allowing for easy long-term storage. To make use of the dried test, it can be easily reactivated by adding a solution of the NASBA-amplified miRNAs. If this solution contains the specific set of miRNAs at a concentration above the threshold, an colour change of the paper disc can be observed.

The output can be generated in two different ways. The first option is by including a reporter gene, such as LacZ, after the toehold switch. Translation of the enzyme encoded by lacZ results in the substrate Chlorophenol Red-\beta-D-Galactopyranoside (CPRG) being converted into Chlorophenol Red (CPR). This results in a clear colour change from yellow to purple on the paper disc in our test. A big advantage of the use of a paper-based system is that the reactions can occur at room temperature and that the test can be performed with standard equipment.