Meeting Gold Medal Criteria Introduction 1. Validation of Sandwich Hybridisation Steps 1.1. Demonstrating Successful APTMS Addition 1.2. Demonstrating Successful DNA Probe Addition 1.3. Demonstrating Successful Sandwich Hybridisation 1.4. Demonstrating Successful Storage and Use 1.5. Issues experienced and how to deal with them 2. Calibrations for other iGEM teams Footnotes

Introduction

The main measurements that we have taken and analysed have been those associated with the detection step of our diagnostic device. More specifically, the validation of the success of numerous different stages in the process of the development of our two working prototypes heavily relied on fluorescent readouts. We thoroughly documented our protocols and different calibration curves through this process. One of our major goals going into this project was to develop a standard set of measurements and protocols that other iGEM teams, looking to adapt nucleic acid based diagnostic methods centered around paper based technology, can reference as the basis of their own device.

1. Validation of Sandwich Hybridisation Steps

It was important to us to make our diagnostic tool both useful and repeatable for others, to allow future iGEM teams to utilise the work we have carried out in their own projects. In terms of the real-life application of our test by farmers, it was also important for us to maximise the outputted levels of fluorescence, to help increase the resilience of the test’s outputs when subjected to imperfect following of the instructions. In order to facilitate both of these aims, categorising and describing each of the individual steps of the hybridisation process well, by carrying out well designed experiments with sufficient repeats and enough controls to confidently determine the cause of any differences in measurements, was an absolutely central part of our project.

1.1. Demonstrating Successful APTMS Addition

The first step towards a Sandwich hybridisation disc prototype was the addition of APTMS to the hydroxyl functional groups of the cellulose discs. Two different methods for binding APTMS to paper were tested: the one step and two step method (procedures described in protocols). We additionally varied the incubation times and the concentrations of various solutions in an attempt to optimise a binding method for the use of future teams.

Figure 1. Summary diagram of the one step method for APTMS addition. The two step method is similar to this

To validate the binding of APTMS to the paper, the paper was incubated in a FITC solution overnight (described in protocols) and then the fluorescence was subsequently measured using a plate reader. FITC is a fluorescent molecule that binds to APTMS, so the fluorescence intensity of the discs should be proportional to the concentration of APTMS bound to them.

Different controls were done to validate the functionality of the FITC quantification method. In one set of controls, APTMS was not added to the discs (though all other preparation steps were kept the same). This gave a baseline fluorescence intensity that was due to FITC binding non-specifically to the paper as opposed to the functional groups of the APTMS molecules. It additionally showed that the increased fluorescence in the APTMS discs with FITC added was not solely due to improper washing of the FITC.

Figure 2. Showing the fluorescence of the discs after different treatment protocols with APTMS and with no APTMS. Error bars are given with +/- SD

In order to properly compare the differences in the fluorescence of APTMS and non-APTMS treated discs with many different disc conditions, we grouped the discs by their method of preparation, independently of APTMS addition. The values were then normalised by dividing the fluorescence by the mean fluorescence of the group - this allowed for direct comparison of the relative fluorescences of APTMS and non-APTMS treated data between many different conditions. Finally, we took the log of this normalised fluorescence value, which was to make the data more normally distributed and increase the heterogeneity of the variance - this is often the norm with fluorescence that spans orders of magnitude, and log-transforming in this way is functionally equivalent to comparing geometric means of the untransformed data. An ANOVA was performed to test the significance of the differences in fluorescence between APTMS treated paper, and paper without APTMS treatment, and the difference was found to be significant (ANOVA, F_1,199 = 49.509, p=3.11x10^-11). The results indicated that the addition of APTMS does significantly increase fluorescence intensity obtained subsequent to FITC addition validating this method for quantification of APTMS concentration.

Figure 3. Showing the log normalised fluorescence of the discs not treated with APTMS, compared to the discs treated with APTMS. The log of the normalised raw data was taken to improve normality for data analysis. Error bars with +/- SD, and the pairwise significance are shown.

Additionally, controls were done where discs were incubated without FITC to validate that any fluorescence increase due to the addition of APTMS was not due to the APTMS molecules themselves fluoresceing, but was instead due to the APTMS molecules binding more FITC. Further controls were performed to ensure that any fluorescence changes obtained were not an artefact of a different stage of the disc preparation. Results achieved from these controls were significant, giving supporting evidence of the successful binding of APTMS to paper (and the subsequent coupling of FITC to APTMS). For example, pooling and normalising data of the fluorescence of APTMS discs, based on whether they were treated with FITC or without FITC, produced significant results (Kruskal-Wallis, χ²(1)=57.759, p=2.962x10^-14); a Kruskal-Wallis test was used due to the extreme heteroskedasticity and deviation from normality that was not fixed by transformation. This shows that a significant proportion of the increased fluorescence of APTMS discs treated with FITC compared to non-APTMS discs, comes from the binding of FITC and not from APTMS’s own fluorescence.

Figure 4. Example of one of the sets of controls carried out to confirm that fluorescence was due to bound APTMS immobilising FITC to paper. Normalised fluorescence of APTMS discs treated with or without FITC. Error bars with +/- SD, and the pairwise significance are shown.

1.2. Demonstrating Successful DNA Probe Addition

The second step towards the prototype was the binding of the immobilised DNA probe to the paper. This was done using two different methods that were adapted from existing literature; the ionic method and the covalent method (procedure described in protocols). The ionic addition method was a relatively novel DNA addition method that had been tested few times in prior literature and not documented in detail. We therefore wanted to validate the success of this method and develop a more thorough protocol for future iGEM teams also working with paper based nucleic acid centered diagnostic technology. The covalent method was relatively well documented in literature- though lacking consistent incubation times and concentrations- and was therefore was a good standard against the ionic addition method. In addition to comparisons being made between the two methods, different conditions were also varied within each method in an attempt to optimise the protocol.

Figure 5. Summary diagram of the ionic addition of the DNA probe to the paper discs.

Figure 6. Summary diagram of the covalent addition of the DNA probe to the paper discs.

In both methods, the successful binding of the DNA probes to the paper was validated by using Cy3 bound DNA probes and using fluorescence measurements to quantify the concentration of the probe actually bound to paper. The Cy3 dye was chosen due to its high fluorescence intensity (making differences in fluorescence more detectable) and small size (making it less likely to interfere with the process of the probe binding to the paper as for the ultimate sandwich hybridisation, the immobilised probe is not bound to a fluorophore).

The ionic method of addition was measured first, with controls being carried out for the same reasons as those with FITC - to ensure that the increased fluorescence was due to actual binding of DNA probes to the APTMS functional groups. Specifically, these controls were not prior treated with APTMS meaning that the fluorescence intensity obtained from them was due to the non specific binding of the DNA probe and ineffective washing steps. Additionally, controls were done where discs were incubated without the DNA probe to validate that any fluorescence increase shown due to the addition of the DNA probe was not due to the APTMS molecules themselves. As well as 3 different control conditions, the DNA immobilisation was also carried out with the DNA suspended in 3 different solutions - TE, PB, or water. This buffer data was important in helping direct the DBTLs for optimising each of the steps to sandwich hybridisation, by informing and directing future protocols.

Figure 7. A bar chart depicting the fluorescence of paper discs following treatment with different combinations of DNA and APTMS, and under different buffers

In order to confirm whether the DNA probes were being successfully added, the data was grouped by the reagents present in the addition (DNA and APTMS, DNA without APTMS, APTMS without DNA, and no DNA or APTMS), and normalised by dividing the data points by each group’s mean - this allows for comparison between data of different buffers. The data did not meet assumptions of normality and homogeneity of variance, and common transformations were not able to solve for this. We therefore tested significance of the differences in the median fluorescence of the four conditions with a Kruskal-Wallis test. The difference in the was significant (Kruskal-Wallis, χ²(3)=36.028, p=7.389x10^-8), and subsequent post-hoc testing was carried out with a Dunn test to compare pairwise differences. The median fluorescence of discs treated with DNA and APTMS was significantly different to discs treated with DNA and no APTMS, discs treated with no DNA, and discs treated with no DNA or APTMS (p=0.000023, p=0.0000025, and p=0.0000032 respectively). The differences in medians between discs treated with DNA and no APTMS, discs treated with no DNA, and discs treated with no DNA or APTMS, were not significant (all with p>0.3).

Figure 8. Dot plot depicting the normalised fluorescence of ionic discs prepared with different conditions - DNA and APTMS, DNA without APTMS, and no DNA or APTMS. Error bars of +/- SD, and all significant pairwise differences are given.

Farmer not washing being okay

1.3. Demonstrating Successful Sandwich Hybridisation

The third step of the protocol came with the proof of concept for the Sandwich hybridisation method. For the development of the paper disc in this experiment, a non-Cy3 bound DNA probe was bound to the cellulose discs. A pre-hybridisation step was then performed between the fluorophore bound mobile probe and the miR399f molecules in solution before the addition of this to the pre-prepared paper discs bound to immobilised DNA probe (procedure described in protocols).

Figure 9. Summary diagram of the final sandwich hybridisation protocol.

The pre-hybridisation step used a mobile probe bound to Cy3. In addition to reasons stated earlier about why Cy3 was chosen, it was also well suited to being the fluorophore in this scenario as it is able to stabilise nucleic acid duplex formation. The selection of the correct fluorophore in this step was especially important, as depending on the limit of detection of the assay, and the concentration of miRNA that is able to be obtained from the extraction procedure, the final product may possibly not need any signal amplification and therefore use a fluorophore bound mobile DNA probe. This means that the fluorophore for the mobile DNA probe should be chosen carefully to maximise the efficiency of the hybridisation step.

Multiple controls and experiments were used to validate different aspects of this experiment. Negative controls added only the diluted fluorophore bound probe solution to the paper discs to obtain the background fluorescence intensity caused by non specific binding of the fluorophore bound mobile probe to the paper. Another experiment was carried out in parallel to this using the human miR16 to test the specificity of the test.

Figure 10. Figure showing the differences in fluorescence intensity between discs with different binding methods (covalent referring to the covalent method used for binding the immobilised DNA probe to the cellulose paper and ionic for the ionic method) and whether the paper was blocked prior to testing or not. Controls were done with miR16 to test if non-specific binding would contribute to a higher fluorescence signal. Controls were also done without miRNA to get a background reading of fluorescence.

The data was pooled into the 4 different disc preparation methods that we carried out, and then normalised by dividing the data points points by their respective group mean fluorescent value. This gave normalised fluorecent values for miR399, miR16, and no miRNA, which we took the log of to increase normality and homogeneity of variance. Once logged, the assumptions of ANOVA were sufficiently met, and the differences in normalised fluorescence were shown to be significant (ANOVA, F_2,44=29.308, p=8.113x10^-9). Subsequent post-hoc testing was carried out with a Tukey’s HSD test, and the differences in the mean of miR399 and miR16, and the differences in the mean of miR399 and no miRNA, were both found to be significant (p=2.4x10^-8, and p=1.1x10^-6 respectively).

Figure 11. Dot plot of the log normalised fluorescence data for the sandwich hybridisations when no miRNA, miR16, or miR399 were used. The log of the normalised data was taken in order to improve normality and homogeneity of the variance for data analysis. Error bars with +/- SD, and all significant pairwise differences are given

Additionally, for this final sandwich hybridisation step, the effect that blocking the paper had on the level of background noise was tested (procedure described in protocols). Though we obtained the blocking buffer from existing literature, it had only been used prior in conjunction with the use of gold nanoparticles ¹ for detection of nucleic acids with this method not having used APTMS. Our test therefore shows that this blocking buffer is compatible with the rest of our protocol for binding DNA to paper. This is important due to the lack of a singular well defined and easily accessible protocol for binding DNA to paper, with lots of existing protocols- for example using ready made blocking buffer solutions ² - increasing the cost and decreasing the accessibility of their methods. The final protocol that we developed for blocking and determination of a cheap, easily made blocking buffer therefore makes the use of paper discs in this way more financially accessible to future iGEM teams as we are aware of the budget issues that can be associated with the project.

1.4. Demonstrating Successful Storage and Use

The final validation of the protocol was demonstrating that the hybridisation assay functioned on paper discs that had been stored for periods of time, just as they would be if shipped across the globe. Paper discs were prepared as before, both with the ionic and covalent binding protocols, before being stored for a period of 4 days; this was carried out either by simple storage in a fridge, or by lyophilising the paper discs and then being kept in a dry environment. The sandwich hybridisation was performed using the same protocol as before, and once again, controls in the form of no miRNA and the human miR16 were used.

Figure 12. The differences in fluorescent intensity between stored discs with different binding methods and different storage techniques. The fluorescence of discs treated with miR399, and the two controls, miR16 and no miRNA, are shown. Error bars with +/- SD are given.

Once again, in order to test the significance of our measurements we normalised the data from the four different groups, and pooled the results. The data was sufficiently normally distributed without a log-transformation, but heterogeneity of variance meant that a Welch’s ANOVA was performed to test the significance of the differences in normalised fluorescence between discs treated with miR399, miR16, and no miRNA. The difference in normalised fluorescence was significant (Welch’s ANOVA, F_2,35.561 = 37.296, p = 1.86x10^-9). There was a significant difference in fluorescence between miR399 and miR16 (Games-Howell post-hoc test, 36 df, p=1.1x10^-7), and between miR399 and no miRNA (Games-Howell post-hoc test, 36 df, p=4.1x10^-9). The difference in fluorescence between miR16 and no miRNA was not significant (Games-Howell post-hoc test, 36 df, p=0.4). These measurements indicate that our protocols managed to produce a final disc that is capable of being stored for long periods of time, and that produces a detectable increase in fluorescent intensity only in the presence of the specific target miRNA.

Figure 13. The normalised fluorescence produced by the treatment of discs with miR399, miR16, and no miRNA. Error bars with +/- SD, and all significant pairwise differences are given.

1.5. Issues experienced and how to deal with them

Despite care during the various steps of the sandwich hybridisation experiment to ensure reproducible results, there were some key sources of error. One of these is the fact that different people have different disc washing techniques, in part due to the lack of thorough description in literature. Through the various tests, a standardised washing technique has been developed and outlined to ensure reproducibility in future team’s results. Another source of error from the APTMS addition test was the fact that we only had half-area 96 well plates available to us, with our discs not fitting flat in these wells. To get around this, we inserted them in a way such that they folded against the well walls, however, this introduced a large source of error between triplicate repeats due to the discs being inserted differently. To overcome this, we used only the central four pixels from the matrix disc scan in our fluorescence measurements. Though we did obtain full area 96-well plates around the time of the sandwich hybridisation testing, ideally given time, a next step in validating the APTMS condition results would have been to repeat the different conditions using these new full area plates. Additionally, using black paper discs as opposed to white paper discs in attempt to minimise background fluorescence readouts due to the paper would be ideal given the time and funding.

Most critically, an important future step would be to carry out further experimental repeats for the efficacy differences of different DNA binding techniques on the final sandwich hybridisation readouts.Only ionic method experimental repeats were able to be carried out given the time, however, further covalent repeats should also be carried out. This would allow more reliable comparisons to be made between the efficacy of the ionic and covalent binding methods to validate if the ionic method is a cheaper alternative to the covalent method that could be used for the final product.

Testing of the different binding methods with lower concentrations of miRNA solutions is something else that would ideally be tested experimentally to find the limit of detection of the device. This could furthermore be used to determine if the optimisation results of various steps throughout the entire paper preparation and final sandwich hybridisation protocol decrease the limit of detection significantly.

Note on the use of miR16 as a control: miR16 is a human miRNA that is upregulated in breast cancer. A diagnostic device for human breast cancer using miR16 as a biomarker and the sandwich hybridisation method for detection was found. miR16 was therefore thought to be a safe choice to use when considering which miRNA to use as a control as it had preexisting primer sequences available in literature and the efficiency of these primers had been validated. Additionally, its sequence was sufficiently different from miR399f such that it was unlikely to bind to the miR399f DNA probes. As it is a human miRNA, it will not be found in plants and therefore can be used for spiking in for positive controls ³.

2. Calibrations for other iGEM teams

One of the major ways in which we validated different steps in our process of binding nucleic acids to paper was to measure the fluorescence of relevant fluorophores (mNeonGreen, Thioflavin T, FITC and Cy3). To measure the fluorescence intensities, we used the VANTAstar microplate reader.

To ensure that other iGEM teams can compare their results directly to ours despite not having the same plate reader, we ran dilutions of our fluorophores (ThT, FITC and Cy3) which allows comparison of their results to ours.

Figure 14. Standard curve for FITC in DMSO

Unfortunately, due to our tight budget, we were not able to order Cy3 dye on its own for our dilution sets and instead used our Cy3 bound to DNA. This means that other iGEM teams referencing our results will need to order the Cy3 bound to our specific DNA sequence to do their dilutions (documented below) which can be ordered using the iGEM IDT sponsorship. Fluorescence measurements were taken for dilution series of concentrations 0 uM, 2.5 uM, 5 uM, 7.5 uM, and 10 uM of Cy3 in water, for each of the different Cy3 bound nucleic acid probes. Triplicate repeats were performed for each concentration, with each well having 200 uL of the relevant fluorophore bound probe. This is necessary as being bound to DNA affects the fluorescence intensity of Cy3 with the base sequence of the DNA bound also affecting this. As seen in Figure 15, all standard curves for the three probes fit well to a logarithmic model.

Figure 15. How the fluorescence intensity changes for different Cy3 bound nucleic acids and the differences between them

To ensure that results on different plates were comparable, we utilised the ‘enhanced dynamic range’ option on the plate reader so that the amplification of the signal was the same except for Thioflavin T (ThT) as its free state and bound state emit at different wavelengths, thus requiring an emission spectral scan. Another reason is that ThT in solution phase and in paper phase might also exhibit different emission wavelengths. For this case, we experimented with the gain percentage and arbitrarily set it to 20% for subsequent setups as, at 20%, it does not saturate the sensor of the microplate reader.

For ThT, a serial dilution was performed including 100, 10, 5, 1, 0.5 and 0.1 µM ThT and measured for fluorescence, each as triplicates. Spectral scans were performed in case of any emission shift as a consequence of concentration difference. The spectral scan (average from triplicates) obtained in Figure 16 saw the shift of emission peak from 445 nm, according to literature ⁴, to around 470-480 nm. This may be attributed to the ThT contamination of Milli-Q water used as a blank as the majority of the signals surrounding the 440 nm mark were negative i.e. they were lower than the blank’s signals. A replicate of 0.5 µM ThT was an anomalous signal and biased the average spectral scan towards more positive values. The data point, therefore, was removed during data cleaning, and a more polished version of the spectral scan was obtained as seen in Figure 17.

Figure 16 Average spectral scans including an anomalous result for 0.5 µM Thioflavin T (ThT) (Wavelength in nm)

Figure 17. Average spectral scan for each triplicate and how the fluorescence intensity varies at different wavelengths and concentrations of Thioflavin T (ThT) (Wavelength in nm)

A ThT standard curve was derived from the fluorescence intensity data at 445 nm. Standard deviations for samples were calculated and used as error bars. Due to the contamination issue, at lower concentrations, the readouts after blank correction were negative. An interesting point is that if 10 and 100 µM ThT included, the standard curve would fit better to a logarithmic trend as in Figure 18, but if 10 and 100 µM ThT was removed, a linear model would be more appropriate for the calibration curve as in Figure 19. This may suggest that at higher concentrations of ThT, ThT behaves in a certain manner that influences the relationship between fluorescence and concentration, resulting in a non-linear relationship. Xue et al. (2017) ⁵ reported the self-fluorescence behaviour of ThT likely due to the micelle formation of ThT at concentrations above 5 µM, but an elevated fluorescence readout was described, which would lead to an exponential relationship. Another contributing factor for substantial variations might be the degradation of ThT. As evidenced in an article ⁶, both ThT powder stored at 4℃, and exposed to irradiation for 60 mins decomposed into other derivatives, but these derivatives were shown to have a blue-shifted absorption maximum, which does not align with our finding that the emission peaks were red-shifted.

Figure 18. How the fluorescence changes as Thioflavin T (ThT) increases in concentration, including 10 and 100 µM and the plot fitted with a logarithmic curve (Horizontal axis is in logarithmic scale)

Figure 19. How the fluorescence changes as Thioflavin T (ThT) increases in concentration, excluding 10 and 100 µM and the plot fitted with a linear model

Spectral scans were also conducted for dry paper discs and paper discs impregnated with ThT at varying concentrations, including 100, 10, 5, 1, 0.5 and 0.1 µM ThT, and Milli-Q water was used as a negative control. Figure 20 illustrates that the dry paper disc has an emission peak at around 447 nm but was only marginal compared to other spectral scans. Other emission peaks sat around 470-480 nm, and 476 nm was arbitrarily chosen as the data points for constructing a standard curve since 476 nm lies between the other emission peaks. Despite coming from the same source, the negative control did not display any sign of significant contamination unlike that of the solution-phase spectral scans. This may be due to cross-talks between adjacent wells for solution phase or contamination during aliquoting ThT dilutions into each well.

Figure 20. Spectral scans for dry paper disc and different concentrations of Thioflavin T on paper discs (Wavelength in nm)

Likewise, a standard curve was constructed for ThT on paper discs. The fluorescence readouts at 476 nm were corrected with a blank (water), and similar to the solution phase, if 10 and 100 µM are excluded, the linear regression fits better with the data points as demonstrated in Figure 21 & 22 with the latter possessing a higher R² value.

Figure 21. How fluorescence intensity at 476 nm varies with different Thioflavin T concentrations on paper discs, including 10 and 100 µM and the plot fitted with a linear model (Horizontal axis is in logarithmic scale)

Figure 22. How fluorescence intensity at 476 nm varies with different Thioflavin T concentrations on paper discs, excluding 10 and 100 µM and the plot fitted with a linear model

Footnotes

1. Zhou, F., Noor, M. and Krull, U. (2015). A Paper-Based Sandwich Format Hybridization Assay for Unlabeled Nucleic Acid Detection Using Upconversion Nanoparticles as Energy Donors in Luminescence Resonance Energy Transfer. Nanomaterials, 5(4), pp.1556–1570. doi:https://doi.org/10.3390/nano5041556. ↩

2. Yaichi Kawakatsu, Okada, R., Hara, M., Tsutsui, H., Yanagisawa, N., Tetsuya Higashiyama, Arima, A., Baba, Y., Ken-ichi Kurotani and Michitaka Notaguchi (2024). Microfluidic device for simple diagnosis of plant growth condition by detecting miRNAs from filtered plant extracts. Plant phenomics. doi:https://doi.org/10.34133/plantphenomics.0162. ↩

3. Clancy, E., Burke, M., Vahid Arabkari, Barry, T., Kelly, H., Dwyer, R.M., Kerin, M.J. and Smith, T.J. (2017). Amplification-free detection of microRNAs via a rapid microarray-based sandwich assay. Analytical and Bioanalytical Chemistry, 409(14), pp.3497–3505. doi:https://doi.org/10.1007/s00216-017-0298-6. ↩

4. Levine, H. (1993). Thioflavine T interaction with synthetic Alzheimer’s disease β-amyloid peptides: Detection of amyloid aggregation in solution. Protein Science, 2(3), pp.404–410. doi:https://doi.org/10.1002/pro.5560020312. ↩

5. Xue, C., Lin, T.Y., Chang, D. and Guo, Z. (2017). Thioflavin T as an amyloid dye: fibril quantification, optimal concentration and effect on aggregation. Royal Society Open Science, [online] 4(1), p.160696. doi:https://doi.org/10.1098/rsos.160696. ↩

6. Hsu, J.C.-C., Chen, E.H.-L., Snoeberger, R.C., Luh, F.Y., Lim, T.-S., Hsu, C.-P. and Chen, R.P.-Y. (2013). Thioflavin T and Its Photoirradiative Derivatives: Exploring Their Spectroscopic Properties in the Absence and Presence of Amyloid Fibrils. The Journal of Physical Chemistry B, 117(13), pp.3459–3468. doi:https://doi.org/10.1021/jp309331u. ↩