The aim of ExoSwitch is to create an early cancer diagnosis tool using toehold switches to detect and quantify exosomal miRNAs from blood samples. We modeled a microfluidic device which allows exosome separation from the blood, followed by exosomes lysis and exosomal miRNAs detection using toehold switches in a cell-free system.

Specific miRNAs signatures are correlated with cancer initiation and spreading. The aberrant expression profile of certain miRNAs indicates cancer setup and progression. This means that these expression profiles can hold a diagnostic and prognostic value¹.
For our proof-of-concept, we first modeled a bank of toehold switches targeting specific miRNAs using SwitchMi Designer (a software we have designed). We then decided to focus on the detection of two miRNAs, miR-21-5p and miR-141-5p. The miR-141-5p is overexpressed in cancer patients such as in the case of breast or prostate cancer². The miR-21-5p is commonly found in abnormal quantities in patients suffering from cancer and it is overexpressed in cervix and breast cancer cell lines’ exosomes, such as HeLa and MCF-7^3-5. We tried to produce the banks of toehold switches targeting these two miRNAs in order to characterize them (phase 1 of our proof-of-concept).
At the same time, we cultured HeLa and MCF-7 cell lines to extract their exosomes and isolate the exosomal miRNAs to fit phase 2 of our proof-of-concept.

Phase 1

A plasmid to produce any toehold switch candidates

Pierre Nioche kindly provided us with a GFP tagged protein and pET24d(+) plasmid. The first step was to build a pET24d(+)_GFP plasmid by assembling GFP and pET-24d(+) using molecular cloning with restriction enzymes. (Figure 1)

Figure 1: Assembly of the pET-24d(+)_GFP plasmid with molecular cloning method using restriction sites. Created with BioRender.

We first amplified the GFP fragment from the GFP tagged protein via polymerase chain reaction (PCR) to get a higher quantity of DNA for molecular cloning. At each step, samples were analyzed by gel electrophoresis to validate the step. Figure 2 shows that after PCR we have obtained a unique fragment (one band in the red box) that is slightly below the 800bp band from the NEB 100bp ladder. This fragment corresponds to the expected EGFP fragment with the two restriction sites EcorI and XhoI at its extreme sides, which is 738bp long.

Figure 2: GFP fragment amplified via PCR controlled on an agarose gel. Samples and NEB 100bp ladder were run on a 1.2% agarose gel for 30min. The NEB 100bp ladder was loaded in the first well. The DNA fragment encoding for GFP is labeled in the red box, with a size between 800 and 700 bp (expected size= 738bp).

We next aimed to integrate this PCR amplified GFP fragment into the plasmid pET-24d(+). GFP fragments and pET-24d(+) are thus digested using the restriction enzymes EcorI and XhoI overnight. The efficiency of digestion was checked by electrophoresis. Figure 3 shows that pET-24d(+) was successfully digested: the well loaded with the digest samples exhibit a unique band corresponding to linearized DNA (open and digested plasmid), compared to the well loaded with an undigested plasmid that shows two bands for the two forms of a plasmid.

Figure 3: Digested pET-24d(+) controlled on a gel. Samples were run on a 1.2% agarose gel for 30min. The two first wells are loaded with pET24d(+) plasmid which was previously digested using EcorI and XhoI restriction enzymes and show the band of linearized DNA. The third well has been loaded with undigested pET24d(+) plasmid, showing the two bands of circular and supercoiled forms.

The GFP fragment was then integrated into the linearized pET-24d(+) plasmid using 10 minute ligation protocol with the NEB T4 ligase. To amplify the plasmid, Bl21(DE3) E. coli were transformed with the ligated product using heat shock protocol. Since pET-24d(+) plasmid contains the kanamycin resistance gene, bacteria were cultured overnight on kanamycin LB (Luria Broth) to select the E.coli who have integrated the plasmid. From this transformation, we selected 8 clones that we're able to grow on our kanamycin media and form colonies. After DNA extraction from these 8 clones, we have made a PCR to amplify the GFP fragment to verify the integration of the GFP fragment into the plasmid. Figure 4 shows that PCR has amplified a fragment of more than 500pb from DNA of the colonies 2, 4, 5,6 and not from other colonies. This fragment presented the expected size of GFP, suggesting the success of the ligation. To confirm the good construction of our plasmid, we sent it to sequencing (Eurofins Genomics) and clones 2,4,6 were positive with our new construction (Figure 4).

Figure 4: pET-24d(+) and EGFP Ligation results. 1.2% Agarose Gel with NEB Ladder 100Bp. This figure presents the results of PCR performed on the different clones obtained through the transformation.

Figure 5: Vector map of pET-24d(+)_GFP created by assembly of GFP insert and pET-24d(+) (BBa_K3878000). Created with SnapGene with the sequencing results.

We named this newly engineered plasmid pET-24d(+)_GFP and it contains the part BBa_K3878000 (EGFP followed by a T7terminator); its vector map is detailed on Figure 5. This plasmid pET-24d(+)_GFP is the receiving vector to insert the variable part of the toehold switch. The plasmid keeps the T7 promoter and most of the restriction site. One point that can be addressed is the disparition of the EcorI site.

Production of complete toehold switches

We tried several methods (detailed in /wetlab) to get the toehold switch directly fully synthesized or the variable part of the toehold switch (Figure 6) ready to be assembled into a plasmid that already encodes the repressed gene, GFP.

Figure 6: Scheme of a toehold switch. In our project ExoSwitch, we define the variable part of the toehold switch the part that differs depending on the target RNA. Indeed, the Toehold part is complementary to the target miRNA. Thus, the complementary nucleotide at the root of the hairpin differs with the target miRNA.

DNA synthesis by SOE PCR

SOE PCR is the first technique we tried to synthesize one of the full Toehold Switch sequences with. Three sets of backward primers were designed and ordered to enable the step-by-step assembly. The size of the primers depends on the set but is on average from 40 to 70 bps.

SOE with the first set of primers: The PCR1 performed with the first backward primer leads to a successful elongation of the starting sequence (EGFP) with the right expected size (Figure 7A). Indeed, electrophoresis gel migration reveals the bands corresponding to the PCR1 product being placed slightly higher than the band corresponding to the GFP migration (Figure 7A). The same is true for the PCR2 performed using the product from the PCR1 with the second backward primer together (Figure 7B). The PCR2 product still migrates less compared to the PCR1 and is then of higher molecular weight (Figure 7B). However, the PCR3 performed with the PCR2 product together with the third backward primer did not complete the Toehold Switch synthesis. The PCR3 product is of very low quantity, and moreover, it seems to be placed lower than the PCR2 product, which means that the sequence has not been elongated… To solve the problem we reproduced the three PCRs several times, but the PCR3 always leads to confusing results. We then decide to redesign the primers to solve the problem.

SOE with the second and third sets of primers: Second and third sets of primers were ordered and delivered. However, only the first two PCRs (PCR1 and PCR2) have been ever successful. The new primers used have not changed the results - after each PCR we were obtaining again something similar to the data represented in Figure 7C. We decided to abandon the strategy after several weeks of attempts.

Figure 7: Electrophoretic analysis of the SOE PCR products obtained with the first set of backward primers. EGFP is used as the starting sequence for the Toehold Switch synthesis. NEB Ladder 1KBp is used as a reference for all three gels (A,B and C). Three consecutive SOE PCRs were performed: PCR1 (taking GFP as the starting sequence, Xho1 as a forward primer, and the first backward primer), PCR2 (taking PCR1 product as the starting sequence, Xho1 as a forward primer, and the second backward primer), PCR3 (taking PCR2 product as the starting sequence, Xho1 as a forward primer, and the third backward primer). A: Post-PCR1 gel electrophoresis. The starting sequence (GFP) is successfully elongated. B: Post-PCR2 gel electrophoresis. The PCR1 product is successfully elongated. C: Post-PCR3 gel electrophoresis. The PCR2 seems to be of higher molecular weight compared to the PCR3 product, no elongation of the PCR2 product was observed.

Hypotheses on what went wrong: We found out that the first set of primers is indeed inaccurate. We hypothesize that the last (third) primer is not able to fix correctly to the amplified DNA due to the specificity of its secondary structure. But at the same time, second and third sets of primers are designed correctly and do not lead to the right PCR3 product formation either. That’s why we still suppose that the secondary structure somehow makes the SOE PCR synthesis challenging.

Moreover, we noticed that a long particular region of the Toehold sequence is rich in GC bases. This makes the pairing of the strands very strong. Thus, we suppose that despite the opening of the DNA during a PCR cycle, the third primer does not have enough time to fixate to the corresponding place. This is what we have decided to call a competition between primers and the toehold sequence, and what can be one of the reasons for challenges we met on this step.

Molecular cloning using restriction enzymes

The synthesized fragments, the variable part of the toehold switch targeting hsa-miR-141-5p, are received into plasmids from IDT and amplified with 10-beta E. coli bacteria transformation. Plasmids with the variable part of the toehold switch are then successfully digested overnight using BamHI and BglII restriction enzymes. Figure 8 shows a unique band slightly smaller than 300pb, corresponding to the purified and digested fragment (between 260 to 290bp depending on the candidate).

Figure 8: Purification of the digested variable part of toehold switch targeting hsa-miR-141-5p confirmed on a gel. 1.2% agarose gel loaded with the NEB 100bp ladder on the first well and the 6 candidates.

The assembly of the variable part of the toehold switch into the plasmid pET-24d(+)_GFP was performed as previously but did not work as shown after sequencing of the plasmids with Eurofins Genomics (Figure 9).
Hypotheses on what went wrong: We obtained a very low amount (~40 ng/μL) of plasmid and insert after the purification step, too low a concentration of both components significantly decreased our chance to get the good assembly product. In addition, BglII and BamHI generate the same sticky ends, meaning that the plasmid can close itself easily. We should have used a higher concentration of plasmids and inserts and better optimized our strategy to use other restriction enzymes that do not generate the same overhangs.

Figure 9: Sequencing of the pET-24d(+) plasmid after the ligation step does not show the correct insertion of the variable part of the toehold switch candidates targeting miRNA_141_5p. Created with Snapgene.

Molecular cloning using golden gate assembly

In parallel with the restriction enzyme method, we worked on another method using the NEB® Golden Gate Assembly Kit (BsaI-HF®v2). With the help of Manish Kushwaha, we designed the fragments coding for the variable part of the toehold switches targeting miR-21-5p preceded by T7 promoter. iGEM Evry kindly gave us the BBa_K3453101 as a plasmid for the golden gate assembly. The variable parts of the toehold switch are synthesized by IDT and received into a plasmid. We successfully amplified the fragment via PCR to get a high quantity of the fragments (between 100 and 200 ng/μl).
The amplified fragment was inserted into the BBa_K3453101 plasmid using NEB® Golden Gate Assembly Kit (BsaI-HF®v2). E.coli was transformed with the golden gate product. Since receiving plasmid contains the Tetracycline resistance gene, bacteria were cultured overnight on kanamycin LB (Luria Broth) to select the E.coli who have integrated the plasmid. We obtained 200 colonies per candidate. To confirm the success of the ligation, we performed a PCR amplification on the variable part of the toehold switch fragments. Figure 10 shows a similar band for each candidate, even for the negative control. We concluded that our construction is not a success and that none of the plasmids contains the variable part of the toehold switch. In addition, we sequenced some plasmids to the sequencing, and the results confirmed the non-assembly of the construct. None of them had the fragment cloned into the plasmid, as indicated in the gel run with PCR amplified fragment front.

Figure 10: The ligation of the variable part of the toehold switch targeting hsa-miR-21-5p into the receiving plasmid did not worked. 1.2% Agarose Gel run on the PCR product performed on clones that grow after the transformation with Golden Gate Protocol. Well 1 and 16 were loaded with the negative control (no fragment) and wells 2 to 15 were run with the samples.

Hypotheses on what went wrong: We tried the golden assembly only once without optimizing the protocol. We hypothesize that we may not have well designed our fragments or that the ratio fragment/plasmid must be improved.

Phase2

MCF-7 and HeLa cell lines Exhibit Exosomal miRNAs

In this part of the project, we aimed to verify that MCF-7 and HeLa cell lines are expressing abnormal quantities of exosomal miRNAs. These cell lines have been selected because they are well-characterized cancer cells, for human breast cancer cells and human cervical cancer, respectively (see our wetlab section for more details)⁵. MCF-7 and HeLa cells were grown separately at a 70% of confluence as we can see in Figure 11.

Figure 11: HeLa cells at 70% confluency.

Exosome extraction from HeLa cells and MCF-7 cells:

Extraction of small vesicles was performed on the two cell lines using a protocol based on differential centrifugation in order to obtain pure exosomes free from all types of residues and contamination. Due to the transparency of the vesicles, a BiCinchoninic acid Assay (BCA) test has been performed to evaluate the amount of total proteins in our samples. Vesicle membrane contains proteins, thus we can correlate the concentration of proteins with the concentration of small vesicles.

Table 1: protein concentrations in the samples of MCF-7, HeLa cells culture respectively after exosome extraction.

n°	Sample	Protein concentration (µg/mL)
1	MCF-7 Extract A	555,90
2	MCF-7 Extract B	660,07
3	Exosomes HeLa	350

BCA results show protein concentration between 350 and 660 µg/mL, as described in Table 1. These concentrations are decent, leading to the assumption that we had vesicles.

A western blot has been conducted to confirm the presence of exosomes in the vesicle extract, using CD9+ and CD63+ as surface markers and Calnexin as a negative control (non-exosomal protein). As shown on Figure 12, we observed a band for the CD63+ markers for the extracts from HeLa cells but not from MCF-7 cells. The negative control was The negative control Calnexin was expressed. Calnexin is an abundant 90kDA chaperone protein that resides in the membrane of the endoplasmic reticulum, and is not supposed to be expressed in exosomes. Calnexin was detected in samples from MCF-7 cells (A and B), meaning that it may not be pure exosomes. None of the vesicle extracts exhibit CD9+. This result is not that surprising because CD9 expression on exosomes vary with the cell type and the physiological conditions^6-7. CD63+ was detected but not Calnexin for the sample from HeLa cells, we concluded that the vesicle extract from HeLa cells contained exosomes.

Figure 12: Vesicles extracted from HeLa and MCF-7 cell media. Western blot analysis of CD9, CD63 and calnexin protein respectively in the MCF-7 and HeLa exosomes

Thus, we performed miRNAs extraction and quantification from HeLa exosomes. We first isolate miRNAs from HeLa exosome sample.

To quantify the targeted miRNA, we performed reverse transcription on the RNA sample we have and obtained the complementary DNA (cDNA). PolyA-tag was added to the miRNAs and we performed a qRT-PCR on these polyA-tagged miRNAs to quantify the hsa-miR-141-5p and hsa-miR-21-5p.

Comparison of the variation of Cq using U6 primers, specific miRNA primers (targeting hsa-miR-34a, hsa-miR-141-5p and hsa-miR-21-5p), and between water and cDNA sample was done. H2O was added in place of cDNA sample as control value. We analyzed the difference between H2O and sample Cq, which is the number of cycles necessary to extract fluorescence from the noise of the qRT-PCR. Low Cq means high RNA quantity. U6 was the positive control of the well functionning of the qRT-PCR reaction. H2O Cq was higher than our cDNA Cq meaning that no errors have been done following the protocol. The Cq measured for miR-34a for the cDNA sample is equal to the Cq for the water. This indicates that miRNA-34a was not present in the exosomes extracted from HeLa cells, which was expected as this miRNA selected as a negative control for assessing hsa-miR-21-5p and hsa-miR-141-5p presence in cancer derived exosomes. The Cq measured for the hsa-miR-141-5p and miR-21-5p are lower in the cDNA sample than in the water (table 2), meaning that these miRNAs were contained within exosomes from the HeLa cells.

Table 2: Final results of qRT-PCR with mean Cq from each miRNA We calculate the mean Cq of the triplicate of each miRNA with their primers in order to compare them with their respective blank.

Sample	Cq(hsa-miR-141-5p)	Cq(hsa-miR21-5p)	Cq(hsa-miR34a)	Cq(miU6)
H2O	29,37	17,65	25,18	37,17
cDNA sample from Hela cells' Exosomes	19,85	14,85	24,71	21,98

We were able to demonstrate that exosomal extracts from HeLa cell lines contain miRNAs such as hsa-miR-141-5p and hsa-miR-21-5p. However, we were not able to quantify these miRNAs. In order to find the precise concentration of our initial sample, we will need to compare our Cq value with a range of dilution of synthetic miRNA.

Perspectives

Toehold switches exhibit a very specific secondary structure that makes their synthesis particularly difficult. Nevertheless, we are convinced that the construction of the plasmid encoding for the complete toehold switch can be successfully done with more time and protocol optimisation.
Once the plasmid encoding for the complete toehold switch is produced we will double transform BL21(DE3) E. coli with a miRNA (either the trigger one or a negative control with another miRNA) to demonstrate the specificity and sensibility of the toehold switch candidates.
After this step, we hope that at least one candidate per toehold switch bank targeting a specific miRNA will give satisfactory results.
We will characterize the toehold switch with a cell-free system. We will measure the fluorescence of :

• the toehold switch without trigger miRNA to confirm the toehold switch does not produce any fluorescence,
• the toehold switch with another miRNA that is not its target to confirm its specificity,
• the miRNA alone to confirm that signal was not from autofluorescent proteins produced from the miRNA
• the toehold switch with a predefined quantity of trigger miRNAs to confirm if the quantity of fluorescence is correlated with the quantity of miRNAs and how.

We hope these results will tell us if our SwitchMi Designer is well able to model toehold switches to detect and quantify specific miRNAs and that we may be able to improve it. The toehold switch that gave the best results on the characterization experiment will be tested on a cell-free system with the miRNAs extracted from our MCF-7 and HeLa cell lines previously isolated. We will compare the results given by the toehold switch with the one we obtained using the classical qRT-PCR method.