RPA Overview

Recombinase Polymerase Amplification (RPA) is an isothermal DNA amplification method which combines a recombinase enzyme with a strand-displacing polymerase to rapidly generate new DNA copies (see Diagnostic Overview). Like polymerase chain reaction (PCR), RPA utilizes specific primers to selectively amplify target sequences of interest (see Fig. 1).

Figure 1. Animation of the recombinase polymerase amplification process.

RPA Methodology

LANCET’s proximity-dependent ligation (PDL) assay generates a limited number of ligated DNA products, requiring amplification for reliable diagnosis from a small blood sample. To ensure the output from PDL can be easily detected, we incorporated RPA, allowing us to produce a sufficient number of copies for successful detection (see Fig. 2).

Figure 2. Depiction of exponential DNA amplification using RPA.

RPA vs. PCR

Compared to PCR, the traditional method of DNA amplification, RPA offers faster isothermal replication with greater sensitivity, making LANCET well suited for rapid point-of-care testing (see Table 1).

RPA Design

RPA Target Construct Design

To ensure maximum efficacy of our RPA assay, we designed our target construct to contain all RPA amplicons. Amplicons were designed according to the TwistDx® DNA Amplification Kit Assay Design Manual for RPA with key parameters as follows (Lambert iGEM, 2025; TwistDx, 2025a):

• Length: 100–200 nucleotides
• GC content: 40–60%
• Minimize number of repetitive, palindromic sequences (see Engineering Success - Diagnostic).
  • Secondary condition to be reduced after two initial conditions for length and GC content are met

Because longer amplicons (>200–300 base pairs) amplify less efficiently due to interfering secondary structures (TwistDx, 2025b), we restricted our target regions to 100-200 base pairs. GC content was maintained in the optimal range and repetitive sequences were minimized to improve amplification speed by reducing the product-to-noise ratio (see Engineering Success - Diagnostic).

In addition, all amplicons were designed to include the same Cas12a sgRNA binding sites, enabling us to maintain strong compatibility with the downstream assay (see Fig. 2). This strategy allows us to significantly reduce experimentation workload and keep the conditions of the Cas12a-DETECTR assay more consistent.

RPA Primer Design

The template DNA sequence provided for our RPA assay is the 232 base pair product resulting from our PDL assay, which we then used to design our RPA primer pairs (see Fig. 3).

Figure 3. Full length sequence map denoting all RPA primer binding sites on PDL product (Lambert iGEM, 2025).

RPA Primer Set 1.1

For the forward and reverse RPA primer set 1.1 (BBa_250TF2EO and BBa_25X1JFV4), we independently designed the sequences using the TwistDx® Design Manual (TwistDx, 2025a). The RPA primer design restrictions as set out by TwistDx® include these guidelines:

• Length: 30–36 nucleotides
• GC content: 20–70%
• T_m (Melting temperature): 50–100°C
• Maximum length of a mononucleotide repeat: 5 nucleotides

We intentionally positioned the primers near the ends of the PDL product to facilitate amplification of both aptamer sequences. This guarantees that only fully ligated products, rather than single aptamers, would be amplified (see Engineering Success - Diagnostic).

The primer sequences were also cross-referenced against the full target region to ensure that each primer had only one complementary binding site within the desired amplicon, reducing off target effects and ensuring proper alignment. Although shorter sequences such as RPA primers (~30 nucleotides) generally do not exhibit significant secondary structures, we further screened our primer sequences to minimize potential hairpins or primer-dimers that could create additional primer noise (Daher et al., 2015) (see Engineering Success - Diagnostic).

RPA Primer Sets 1.2–1.4

After finding limited success with our manually designed primers, we generated additional RPA primer pairs 1.2-1.4 (BBa_25VUQNNR, BBa_25RMXLR4, BBa_25NVR9HK, BBa_25PUGKNO, and BBa_257J0EPT) using the EZassay software (see Engineering Success - Diagnostic). From the software’s set of possible primer combinations, we filtered the results to meet specific requirements of our assay which EZassay’s algorithm did not account for. We excluded amplicons that were too short for our assay (<100 base pairs) or those that did not encompass both aptamer constructs, which could lead to a false positive result (see RPA Primer Set 1.1).

RPA Primer Set 1.5

To further validate our software, we generated an additional RPA primer pair, Set 1.5 (BBa_25F80ARP and BBa_25N9JU4X), using CASPER. Unlike previous sets designed with external tools such as EZassay, this primer pair was produced entirely within our own platform, which takes into account T_m balance, GC content, and dimerization risk (see Software Results).).

Among all candidate designs, primer set 1.5 achieved the highest composite score from CASPER, reflecting its strong predicted performance. After CASPER generated the primer set, we also confirmed that primers amplified both aptamer sequences, reducing the risk of false positives in our assay’s process. By experimentally testing primer set 1.5, we were able to evaluate the effectiveness of CASPER’s scoring system and confirm its ability to generate functional primers tailored to our diagnostic assay (see Validating Predictions from CASPER; see Software Results).).

RPA Experimentation

We performed multiple separate experimental workflows to guide our testing with RPA. First, we validated the functionality of our TwistAmp® kit by using a commercial positive control. Then, for experimentation unique to LANCET, we began by amplifying a construct that simulated the PDL product, and finally directly used the double stranded DNA (dsDNA) output from PDL to ensure that RPA could be effectively integrated within our diagnostic pipeline.

All final reactions, including controls, were run in triplicates to ensure that results were reproducible and statistically significant.

RPA Positive Control Testing

To validate the efficacy of our commercial RPA kits, we decided to use the TwistAmp® Exo positive control and TwistAmp® Exo positive control oligo mix to begin our experiments. We tested the TwistAmp® Basic kit and verified the output using gel electrophoresis and utilized the TwistAmp® Exo kit to get a quantitative fluorescent readout using a plate reader. Our initial results showed a band for RPA at ~150 bp as expected and a significant increase in fluorescence, confirming that the kit functioned as expected (TwistDx, 2025a; see Figs. 4-5).

Figure 4. Gel electrophoresis results of RPA positive control reaction, showing band for positive control at the expected length of ~150 bp.

Figure 5. Fluorescent output of TwistAmp® Exo kit, quantified using plate reader; results indicate a significant difference in RFU between positive and negative controls.

RPA Target Construct Testing

After verifying the performance of the kit, we purchased a synthetic dsDNA RPA target construct to serve as a template that contained binding sites for all of our primers. To simulate our assay’s output, we modeled the expected concentration of the PDL product, and diluted the target construct to imitate experimental conditions (see Fig. 6); (see PDL Modeling Results); (see Experiments).

Figure 6. Predicted generation of dsDNA PDL product over time, informing experimental protocol for RPA target construct reactions.

We began our experimentation with RPA primer set 1.1, which proved to be ineffective (see Fig. 7).

Figure 7. Gel electrophoresis results of RPA on target construct with primer set 1.1, showing no bands present for experimental and control samples.

Subsequently, we utilized CASPER, our software tool for RPA-Cas12a assay design, which predicted that primer sets 1.2-1.4 would perform more effectively (see RPA Primer Sets 1.2-1.4) (see Software Results).

Experimentally, RPA primer sets 1.2 and 1.4 demonstrated greater efficacy, producing gel bands within the expected base pair length around ~100-150 base pairs for each amplicon and showing significantly greater expression than the negative control (see Fig. 8).

Figure 8. Gel electrophoresis of RPA with sets 1.2–1.4, demonstrating bands at 153 (a), 133 (b), and 154 (c) bp for primers 1.2-1.4, respectively.

By using the ImageJ software to analyze mean intensity and area covered by the bands, we were also able to quantify the success of our RPA assay (see Fig. 9) (see Engineering Success - Diagnostic).

Figure 9. Comparative ImageJ analysis of RPA target construct reaction on agarose gel electrophoresis with primer sets 1.2-1.4, demonstrating relative efficiency of RPA primers.

Results from ImageJ analysis quantitatively confirmed that primer set 1.3 resulted in the greatest amplification, followed by 1.4, then 1.2, which was consistent with predictions from CASPER (see Software Results). Moving forward, we decided to only pursue experimentation with primer set 1.3.

Conjugating RPA with PDL

The final step of our experimentation was to ensure that RPA could be integrated with PDL within our overall diagnostic workflow. We followed the optimized experimental protocol for PDL, including the same negative controls and upscaled reaction conditions for RPA as we did with PCR (see Experiments ; see Optimizing Aptamer Concentration).

Figure 10. Gel electrophoresis of post-PDL RPA products, demonstrating bands at expected lengths for experimental samples and empty lanes for control samples, validating the success of RPA.

From conjugating PDL and RPA together, we successfully found bands at the expected length of 133 base pairs, indicating successful assay activity. Our results show that we were able to successfully amplify the PDL product using RPA, demonstrating the compatibility of the two processes within our diagnostic workflow (see Fig. 10).

Testing Protein Concentration Over Time

To ensure that LANCET remained effective across multiple stages of Lyme, we decided to test the amplification efficacy of RPA by varying the protein concentration of CspZ. By using approximated blood-borne protein concentrations at 2, 25, 50, 75, 100, and 150 days post infection (pi), we were able to confirm that the combined PDL-RPA process could function over an extended period of time (see CspZ Protein Concentration).

Our gel electrophoresis results visualized fluorescent bands at the expected 133 base pairs for PDL-RPA samples. Bands were qualitatively observed in samples with protein concentration up to 100 days pi, providing a preliminary basis for the functionality of LANCET (see Fig. 11).

Previous slideNext slide

Slide 0 of 0

Figure 11. Gel electrophoresis results of post-PDL RPA reaction with protein concentrations at 2 (a), 25 (a), 50 (a), 75 (b), 100 (b), 150 (b), 200 (c), 250 (d), and 300 (d) days pi, qualitatively validating that assay remains viable until at least 100 days pi.

By using the ImageJ software to analyze mean intensity and area covered by the bands, we were also able to quantify the success of our RPA amplification (see Fig. 12; see Engineering Success - Diagnostic).

Figure 12. ImageJ analysis of PDL-RPA on agarose gel electrophoresis with protein concentrations from 2-300 days pi, quantitatively validating assay viability for at least 150 days pi.

Results from ImageJ analysis illustrate that RPA amplification of PDL remains functional until at least 150 days pi, which surpasses the capability of PCR amplification (see Fig. 12; see Testing Protein Concentration Over Time). Similarly to PCR, the signal intensity of amplification was the highest at 2 days pi and assay activity gradually decreased over the 200 day period, with a significant loss of amplification potential by day 200.

All negative controls (no aptamers, no ligase, no protein) produced significantly lower intensities, which further validated our assay’s specificity when using RPA amplification and limited the possibility of false positive results.

These findings demonstrate that, as expected, our approach with RPA is more sensitive and robust than traditional PCR amplification. Additionally, we illustrate here that our RPA system can successfully amplify dsDNA products generated by PDL throughout the progression of Lyme disease. By amplifying the limited PDL products, we ensure that enough DNA copies are available for detection by the downstream Cas12a system and that LANCET functions as a reliable and rapid point-of-care diagnostic.

In Silico Validation

To further characterize our RPA system, we modeled the expected concentration of DNA after amplification using a system of deterministic ordinary differential equations (ODEs) (see Fig. 13; see RPA Modeling). Our model predicted an exponential increase in DNA products from RPA by correlating the initial concentration of PDL product to expected concentration of RPA product over time.

Figure 13. ODE model of exponentially increasing dsDNA RPA product created by correlating the expected concentration of PDL product to RPA product over time.

The integrated densities achieved from our ImageJ software for RPA amplification were normalized by the negative controls and standardized by DNA ladder. By using bands on the ladder of known nanogram weight, the standardized integrated densities of samples can be converted to copies of DNA produced (see Fig. 14).

Figure 14. Comparison of experimental and theoretical RPA amplification on PDL products over time; higher values from the ODE model as compared with observed values shows that experimental samples are lacking in amplification efficiency.

We then compared the copies produced in the RPA target construct reaction, 2 day pi post-PDL RPA reaction with primer set 1.3, and the RPA ODE model (see Fig. 14). The data from our experimentation parallels our predictive ODE model; however, the model predicts a higher DNA concentration produced by RPA as compared to the target construct reaction and post-PDL reaction. Thus, our in silico predictions demonstrate that our experimental data still falls short of predicted values when examining the relationship between the copies of DNA produced from RPA and the initial concentration of reagents in our PDL assay.

Statistical Analysis

ANOVA

In addition to our ODE model, we conducted a significance test using RStudio to validate the statistical significance of our results with RPA. We performed an Analysis of Variance (ANOVA) to compare the mean signal intensity of standardized integrated densities across PDL protein concentrations corresponding to 2, 25, 50, 75, 100, and 150 days pi after amplification with PCR and RPA.

The critical statistical values for our ANOVA analysis include the following (Bevans, 2020):

• F value: the test statistic from an F test; the greater the F value, the greater the likelihood that the variation between groups is caused by the independent variable(s) and not from random chance
• Pr(>F): the p value of the F statistic; a measure of the significance of the F value and how likely it is that the resultant F value could occur due to random chance under the null hypothesis
• Residual variance: the sum of the squares of the residuals of the data set; a smaller residual variance indicates that the independent variables explain more of the variation

We ran a two-way ANOVA analysis that tested whether the independent variables have an interactive effect on each other, where time elapsed may also affect the activity of the amplification methods. Because the ‘Method:Time’ variable has a relatively high F value, low p value of variance (p < 0.001), and small residual variance, much of the variation can be explained by the interaction between the independent variables (see Table 2).

If the differences were not statistically significant, the data would follow the null hypothesis and there would be less than a 2 x 10^-16% probability for the results to occur under random chance (see Table 2). Therefore, the interaction between our amplification methods and time elapsed past infection experimentally show a significant difference that is likely not due to random chance.

Tukey HSD Post-Hoc Test

ANOVA analyses are prone to confounding experimental errors, especially as the number of groups compared increases. For a statistical significance level of 5%, the overall error rate for our ANOVA can reach up to 1 - [(1-0.05)^a], which for 66 comparisons between all of our groups is 1 - [(1-0.05)⁶⁶] = 0.97, or 97% (Greenwood & Banner, n.d.). To control this variability, we ran Tukey’s Honest Significant Difference (HSD) post-hoc test to adjust the p-values at a 95% confidence interval.

The faceted plot with letter displays revealed that all PCR groups (a) remain statistically indistinguishable and cluster at low intensities (~0.2-0.3), indicating no significant time effect for PCR (see Fig. 15). RPA groups, however, are separated into distinct subsets with RPA:2 being the highest (b), RPA:25 & RPA:50 intermediate (c), RPA:75 & RPA:100 lower (d), and RPA:150 as the lowest (e) (see Fig. 15).

Figure 15. Faceted plot of signal intensity by method and time with Tukey HSD group comparisons; all PCR groups (a), RPA:2 (b), RPA:25 and RPA:50 (c), RPA:75 and RPA:100 (d), and RPA:150 (e).

In addition to the Tukey post-hoc test, we graphed the standardized integrated densities of post-PDL PCR and post-PDL RPA at 2 days pi. The plot clearly showed that the signal intensity for RPA was greater by a factor of ten compared to PCR (see Fig. 16).

Figure 16. ImageJ analysis comparing post-PDL PCR reaction and post-PDL RPA reaction, quantitatively showing a significantly greater amplification activity by RPA.

Together, the adjusted p-values from the post-hoc Tukey HSD test and graph of post-PDL PCR compared to post-PDL RPA validates our choice of RPA within LANCET’s diagnostic workflow. Our statistical analysis demonstrates that RPA groups are always significantly greater than PCR groups, even at the latest time points. These findings support RPA’s greater sensitivity and stronger amplification efficiency as compared to PCR, experimentally confirming that LANCET is better suited for point-of-care testing as compared to traditional methods of DNA amplification.

Validating Predictions from CASPER

To characterize the predictive efficacy of our software tool, we also used primer sequences specifically generated by CASPER (see RPA Primer Set 1.5; see Software Results). When utilizing primer set 1.5, we found that we were able to visualize fluorescent bands at the expected amplicon base pair length of 193 base pairs using gel electrophoresis (see Fig. 17). Bands were observed in samples with protein concentration up to 250 days pi, providing a basis for the functionality of our software’s primer sequences (see Fig. 17).

Previous slideNext slide

Slide 0 of 0

Figure 17. Gel electrophoresis results of post-PDL RPA reaction with protein concentrations at 2 (a), 25 (a), 50 (a), 75 (b), 100 (b), 150 (b), 200 (c), 250 (c), and 300 (c) days pi, qualitatively validating that PDL-RPA with primer set 1.5 remains viable until at least 250 days pi.

Once again, we used ImageJ to quantify the standardized integrated density covered by our gel bands and compared the success of our in-house primers with those designed using the commercial software (see Fig. 18).

Figure 18. Comparative ImageJ analysis of PDL-RPA reaction on gel electrophoresis with primer sets 1.3 and 1.5, quantitatively validating viability of primer set 1.5 for at least 250 days pi and demonstrating greater amplification efficiency using CASPER-generated sequences.

Results from ImageJ analysis illustrate that RPA primer set 1.5 was able to outperform primer set 1.3 at some of the time intervals tested (see Fig. 18). Additionally, the sequences generated by CASPER were viable for a longer period than primers generated by commercial software, as signal intensity was significantly greater than those of the negative controls for 250 days pi.

Our experimental findings for primer set 1.5 illustrated that CASPER’s scoring system is in fact able to generate RPA primer sequences that have a greater amplification efficiency. Not only do we verify CASPER’s predictive capability by empirically validating the higher ranking for primer set 1.5 as compared to other primer sets, CASPER is indeed able to outperform current RPA-Cas12a design tools in creating functional primers.

In Silico Validation

After testing with our software’s generated primers, we compared the 2 day pi post-PDL RPA reaction using primer set 1.5 to our deterministic RPA ODE model (see Fig. 13; see RPA Modeling). The integrated densities from ImageJ analysis were again converted to copies of DNA produced (see Fig. 19).

Figure 19. Comparison of experimental and theoretical RPA amplification on PDL products over time; alignment between predicted values from the ODE model with observed values validates results from our experimentation.

The data from our experimentation closely parallels our predictive ODE model, with there being no significant difference between the experimental sample using primer set 1.5 and the modeled RPA reaction (see Fig. 19). Thus, our model confirms that our data is an accurate relationship of DNA copies produced from RPA and indicates that primers designed by CASPER better fit the expected behavior of our system.

ANOVA

To determine the statistical significance of the variation between data from RPA primer sets 1.3 and 1.5, we developed an ANOVA model to compare the signal intensity of PDL-RPA samples with protein concentrations at 2, 25, 50, 75, 100, 150, 200, 250, and 300 days pi.

We ran a two-way ANOVA analysis that tested if primer sequences had an interactive effect with time elapsed to explain the variation between data sets. Because the ‘Primer:Time’ variable has a relatively high F value, low p value of variance (p < 0.001), and small residual variance, much of the variation can be explained by the interaction between the independent variables (see Table 3).

If the differences were not statistically significant, the data would follow the null hypothesis and at the greatest, there would be less than a 1.12 x 10^-5% probability for the results to occur under random chance (see Table 3). Therefore, the interaction between our primer sequences and time elapsed past infection experimentally show a significant difference that is likely not due to random chance.

Tukey HSD Post-Hoc Test

ANOVA analyses are prone to confounding experimental errors, especially as the number of groups compared increases. For a statistical significance level of 5%, the overall error rate for our ANOVA can reach up to 1 - [(1-0.05)^a], which for 153 comparisons between all of our groups is 1 - [(1-0.05)¹⁵³] = 0.9996, or 99.96% (Greenwood & Banner, n.d.). To control this variability, we ran Tukey’s Honest Significant Difference (HSD) post-hoc test to adjust the p-values at a 95% confidence interval (see Fig. 20).

Figure 20. Faceted plot of signal intensity by primer and time with Tukey HSD group comparisons [(see Engineering Success - Diagnostic).](/lambert-ga/engineering/)

The adjusted p-values from the post-hoc Tukey HSD demonstrates that primer 1.5 exhibits higher amplification than 1.3, with letter labels for 1.5 at later time points overlapping with those of earlier time points for 1.3 (see Fig. 20). Primer 1.5 also shows a greater degree of amplification as time progresses, reflecting its greater sensitivity for the dsDNA PDL product. These results validate the selection of primer 1.5 by CASPER for our diagnostic workflow, as it maximizes detectable DNA output.

Blood Serum Testing

Using primer set 1.5, which we validated to have a greater amplification efficiency than previous primer sets, we conducted post-PDL RPA amplification on spiked samples of simulated blood serum. The simulated blood replicated the pH and buffer conditions of serum without containing any actual blood products, making it safe for use.

We introduced CspZ concentrations corresponding to predicted values over time at 2, 25, 50, 75, 100, 150, 200, 250, and 300 days and performed gel electrophoresis to visualize our results (see Fig. 21). Distinct bands were observed for samples up to 200 days post-infection, confirming that the primer sequences generated by our software remained functional across time even in simulated blood serum (see Fig. 21).

Previous slideNext slide

Slide 0 of 0

Figure 21. Gel electrophoresis results of post-PDL RPA reaction with simulated blood serum with protein concentrations at 2 (a), 25 (a), 50 (a), 75 (b), 100 (b), 150 (b), 200 (c), 250 (c), and 300 (c) days pi, qualitatively validating that PDL-RPA with primer set 1.5 in simulated blood remains viable until at least 200 days pi.

Using ImageJ, we quantified the standardized integrated densities of the gel bands to compare the amplification efficiency of our RPA primers under simulated blood conditions with standard conditions in vitro (see Fig. 22).

Figure 22. ImageJ analysis of PDL-RPA on gel electrophoresis with primer set 1.5 using spiked simulated blood serum samples, quantitatively validating assay viability for at least 250 days pi.

Results from ImageJ show that running the assay under simulated blood serum reduced overall assay activity compared to standard in vitro conditions for certain protein concentrations, specifically 100-200 days pi (see Fig. 22). However, while the simulated blood did interfere with the assay, no significant difference in amplification efficiency was observed across other time points, and the assay under blood with RPA primer set 1.5 still remains viable 250 days pi. These findings confirm that our RPA assay maintains functional viability under physiologically relevant conditions.

ANOVA

To determine the statistical significance of the variation between data from amplification in vitro and in simulated blood, we developed an ANOVA model to compare the signal intensity of PDL-RPA samples with protein concentrations at 2, 25, 50, 75, 100, 150, 200, 250, and 300 days pi.

We ran a one-way ANOVA analysis that tested if the assay conditions had a statistically significant effect variation between data sets. Because the ‘Condition’ variable has a relatively high F value, low p value of variance (p < 0.001), and small residual variance, much of the variation can be explained by the interaction between the independent variables (see Table 4).

Because the differences between conditions are not statistically significant (p = 0.432>0.05), the data follows the null hypothesis, meaning that the variation in signal intensity between conditions is likely due to random chance (see Table 4). These findings suggest that the PDL-RPA assay maintains its amplification efficiency even in simulated blood components, demonstrating its robustness and applicability in physiologically relevant samples without significant loss of signal.

ANOVA analyses between numerous comparisons between groups can cause confounding experimental errors, especially as the number of groups compared increases. However, for a statistical significance level of 5%, the overall error rate for this one-way ANOVA reaches up to 1 - [(1-0.05)¹] = 0.05, or 5%, since there is only 1 comparison made (Greenwood & Banner, n.d.). Therefore, there was no need for a post-hoc adjustment for this ANOVA analysis, as it inherently maintains a 95% confidence interval for the variance between conditions.

References