Applying a Linear Amplification Strategy to Recombinase Polymerase Amplification for Uniform DNA Library Amplification.

Recombinase polymerase amplification (RPA) is an isothermal DNA amplification method with broad applications as a point-of-care test and in molecular biology techniques. Currently, most of the applications are focused on target-specific amplification. Because RPA has the advantage of amplifying DNA under isothermal conditions, we utilized RPA as a DNA library amplification tool. In this study, we used a sheared genomic DNA library and an oligonucleotide (oligo) library for the comparison of polymerase chain reaction and RPA. For the sheared DNA library, we observed biased amplification after RPA was conducted. Thus, to amplify the size-variable DNA library uniformly, we introduced a linear amplification strategy with RPA and successfully improved the uniformity. On the other hand, using the same-sized oligo library, we confirmed that RPA amplified this library uniformly without modification of the protocol. These results demonstrate that RPA can be applied not only to amplify a specific target as previously demonstrated but also to amplify a complex DNA library composed of a large number of different DNA molecules.

Introduction

Recombinase polymerase amplification (RPA) was first reported in 2006 and has rapidly become an alternative method to polymerase chain reaction (PCR) for various applications.[1] Typically, RPA requires a recombinase, single-stranded DNA-binding protein and a polymerase[1] to amplify nucleic acids. The mechanism of RPA is similar to that of PCR. First, the recombinase and primer form a complex and attach to the target site. Then, the polymerase recognizes the primer and extends through the template DNA. During extension, single-stranded DNA-binding proteins attach to the opposite strand to stabilize the RPA reaction intermediate.

RPA has two advantages over PCR—reaction temperature and reaction time. RPA usually proceeds between 37 and 42 °C under isothermal conditions, which is advantageous for instrument-free field diagnosis and prevention of heat-induced DNA mutations. The reaction time for RPA is typically 15[2] or 20–40 min based on the manufacturer’s protocol, which is shorter than most PCR reactions. Because of these advantages, RPA is used for detection and diagnosis of specific bacteria[2−4] and viruses.[5−8] Moreover, RPA can be conducted on solid-phase materials,[9−11] including lateral flow dipstick,[10,12−14] which permits the technique to be used in point-of-care tests.

Studies have examined replacing existing PCR-based molecular biology techniques with RPA-based techniques. For example, allele-specific RPA is performed for genotyping SNPs using 19-base pair (bp) to 21 bp primers designed in silico.[15] Blocked RPA uses wild-type binding blocked oligo and mutation-specific amplification,[16] and miRPA detection is performed by amplifying two ligated probes using the target miRNA as a splint oligo.[17] In addition, digital droplet RPA reduces the reaction time to 30 min by replacing the amplification method from PCR to RPA.[18] Another example is multiplex RPA, which detects multiple pathogens simultaneously using multiple primer pairs.[19,20]

With the development of next-generation sequencing (NGS) technology, the importance of error-free and unbiased DNA library amplification is becoming increasingly important. To achieve a reduced error rate and unbiased amplification, the researchers proposed several different methods. For example, researchers compared the fidelity of DNA polymerase[21] and used high-fidelity DNA polymerase to reduce the error which was generated during DNA amplification. Other researchers used molecular DNA barcode to remove the DNA duplicate[22] and amplification bias.[23,24] However, most of the methods used PCR as a DNA amplification method, which has a chance to induce heat-related damages such as deamination of cytosine to uracil.[25]

We hypothesized that replacing PCR with RPA can reduce these heat-related damages. Thus, in this study, we applied RPA to amplify the DNA library and designed several experiments to properly implement it. We amplified sheared DNA and an oligonucleotide DNA library using RPA and identified the appropriate RPA method for each DNA library. Through this study, we show that RPA is suitable for NGS sample amplification, further broadening the potential applications of the RPA method.

Results and Discussion

Random-Sheared DNA Library Amplification Using RPA

To evaluate amplification by RPA, we prepared sheared DNA libraries using genomic DNA from human NA12878 or E. coli EcNR2.[26] DNA was sheared into two different sizes, each averaging between 180 bp and 300 bp (Figure S1), which we termed human small, human large, E. coli small, and E. coli large library, respectively. Sheared DNA libraries were treated with several enzymes for end repair, dA-tailing, adaptor ligation, and uracil cleavage, as described in the Materials and Methods (“NGS sample preparation”). After uracil cleavage, the sequencing adaptor had flanking sequences longer than 30 bp. Besides, the RPA reaction has a suggested primer length of ≥30 bp. Thus, we hypothesized that RPA could be employed directly for the index attachment step instead of PCR.

To confirm our hypothesis, we conducted index attachment steps with PCR and two-primer RPA (n = 3) with the small and large E. coli and human DNA libraries, which were treated through serial NGS sample preparation steps. Although RPA products were smaller than PCR products (Figure S2), the samples were successfully sequenced with the Illumina NextSeq platform (Tables S1, S2). Sequencing data were analyzed using AdapterRemoval,[27] Burrows–Wheeler alignment (BWA),[28] Samtools,[29] and in-house python codes. The most abundant fragment size of RPA-amplified E. coli small and large libraries were 59 bp and 50 bp, respectively, which were smaller than their corresponding PCR products of 92 bp and 147 bp. The human small and large libraries showed similar trends as the E. coli libraries (Figure S3). Moreover, the RPA amplicons of the human and E. coli large DNA libraries showed similar size distributions as those of their small-sized library amplicons.

The genome sequence of EcNR2 is already known; therefore, we were able to calculate the uniformity values of every EcNR2 genome base. Using these uniformity values, we calculated the percentage of bases in the uniform range (the uniformity values in a range of 0.5–1.5) to compare the uniformity difference between samples. Thus, if the percentage was higher, the sample was amplified more uniformly. Using the above calculation, we calculated the percentage of uniform range of small and large DNA libraries which showed 76.6 and 84.0% of bases in the uniform range, respectively. By contrast, the RPA amplicons from small and large DNA libraries showed 56.3 and 49.2%, respectively (Figure S5). We assumed that the size shift might be due to faster amplification of small-sized DNA by RPA versus PCR. Thus, the difference in the size-dependent amplification rate caused biased amplification with reduced uniformity.

Linear Amplification RPA Improves the Uniformity and Size Variation of DNA Libraries

To solve the size shift and lower uniformity problem, we examined single-primer linear RPA as a potential solution to the size shift problem of DNA libraries amplified with two-primer RPA. We hypothesized that a single primer would prevent reamplification of the RPA amplicon by the opposite primer, blocking acceleration of small-sized DNA amplification. Moreover, because of the strand displacement activity of RPA, large DNA should be amplified simultaneously in multiple strands. Taken together, to improve the size shift and uniformity, we adopted a single-primer linear amplification strategy.[30] To test our hypothesis, we amplified E. coli small (n = 3) and large (n = 3) libraries using single-primer linear RPA, which we refer to as linear RPA (Figure ). After adaptor ligation and USER treatment of E. coli libraries, 20 min of linear RPA was performed with the P7 index primer (Table S3). A short cycle of PCR was performed with the P5 index and P7 end primers to attach the P5 index for NGS. We confirmed the size distribution of products by agarose gel electrophoresis prior to performing NGS and found broader size distributions (Figure S4) than that obtained with two-primer RPA (Figure S2). Amplified libraries were sequenced with the Illumina NextSeq platform and analyzed as above. From the analysis result, the size distribution and uniformity of linear RPA products were improved compared with products from two-primer RPA (Figure ). The average size distributions of the triplet experiment with the small and large libraries were 56 bp and 45 bp, respectively, by RPA and 90 bp and 109 bp, respectively, by linear RPA (Figure a). Also, the average percentages of bases in the uniform range (uniformity value between 0.5 and 1.5) about the triplet experiment was improved from 56.3 and 49.2% by two-primer RPA to 73.6 and 75.7% by linear RPA for the small and large DNA libraries, respectively (Figures b and S5).

Figure 1

Schematic diagram of linear RPA. The schematic diagram of single-primer linear RPA and linear RPA product-specific PCR. Black and gray designate the P5 and P7 sides of the NGS adaptor, respectively. Purple designates the P7 index sequence, and blue designates the P5 index sequence.

Figure 2

Size distribution, uniformity, and substitution error ratio of linear RPA. Comparison of (a) size distribution, (b) uniformity, and (c) substitution error ratio of PCR, RPA, and linear RPA amplicons from small and large E. coli sheared DNA libraries. (a) Dot plots show average % ratio of each aligned length for the size distributions of DNA libraries from NGS data. The aligned DNA size data were obtained from Sam files after BWA alignment. (b) Violin plots show uniformity values for all replicates of #1 E. coli EcNR2 samples from Figure S5, including PCR, RPA, and linear RPA. Y-axis is the uniformity value, which was calculated by the following equation (sequencing depth of given position/average depth of whole EcNR2 genome). The numbers above the plot represent the percentage of data with a uniformity value between 0.5 and 1.5. (c) Bar plot shows the substitution error ratio of samples. The bar plots of experimental samples were generated using the average value for each sample; the error bar represents standard deviation values.

Next, we analyzed the substitution error of linear RPA products. We calculated the error rate and substitution type and then compared the substitution error rate with those obtained with KAPA HiFi HotStart Ready Mix[31] and NGS.[32] From the result, no significant substitution error rate differences were noted among PCR and RPA samples (Figure c). The substitution error ratio of samples was similar to the NGS error ratio but not the amplification error ratio. Thus, linear RPA improved the library amplification with respect to size distribution and uniformity without significant substitution error rate change.

Microarray Oligonucleotide Library Amplification Using RPA

Linear RPA was used to overcome the bias which occurs when amplifying the size-variable DNA library with two primers by RPA. We wanted to investigate whether the uniformity is low when amplifying the same-sized DNA library by RPA. To test this, we designed and synthesized a microarray oligonucleotide library (oligo library) with 150 bp long sequences (Supporting Information 1). The terminal 20 bp and 30 bp at both ends were designed to contain common flanking sequences as primer binding sites for PCR and RPA, respectively (Figure S6, Supporting Information 1). One thousand insert sequences with 90 bp were extracted from the genomic DNA sequence of Synechocystis sp PCC 6803. We conducted RPA and PCR for 20 min and 25 cycles, respectively, with 20 ng oligo library in each reaction as a template (n = 3). Agarose gel electrophoresis showed the correct amplicon size with both methods (Figure S7). After NGS sample preparation with eight cycles of index attachment PCR, the samples were sequenced using the Illumina NextSeq platform. Sequencing data for both the PCR and RPA oligo library amplicons were analyzed using BWA, Samtools, and in-house python codes. We assessed the uniformity of the PCR and RPA amplicons by calculating the uniformity value (sequencing depth of the single oligo/average depth of all 1000 oligos) of each oligo in the library. All replicate experiments are visualized on a dot plot showing the uniformity values (Figure a). Results indicated that 77.0% of oligos were in the uniform range (uniformity value 0.5–1.5) after PCR amplification, and 63.2% of oligos were in the uniform range by RPA.

Figure 3

Uniformity and correlation of amplification order between PCR and RPA amplicons for the microarray oligonucleotide library. (a) Comparison of uniformity values for PCR and RPA amplicons shown in a violin plot. Y-axis represents the uniformity value, which was calculated by the following equation (sequencing depth of given position/average depth of whole 1000 oligoes). The numbers above the plot represent the percentage of data with a uniformity value between 0.5 and 1.5. (b) Uniformity value of each oligo in the descending order. The left and the right plots were rearranged in the descending order based on PCR replicate #1 and RPA replicate #1, respectively. (c) Correlation coefficient value between samples using uniformity values of each oligo.

The uniformity value can also represent the order of amplification for each of the content in the library. Values for PCR and RPA replicates were arranged in descending order based on the rank of PCR and RPA replicate 1 to produce dot plots (Figure b). We calculated the correlation coefficient of the uniformity values for the PCR and RPA replicates (Figure c). Interestingly, the uniformity value of each of oligo content showed a different ranking order for PCR and RPA amplicons. We did not identify the reason for the difference in amplification preferences but believe it may be due to the differences in polymerases used for PCR and RPA. Taken together, the data show that RPA uniformly amplifies DNA libraries of the same size and has different amplification preferences than PCR.

Discussion

In this study, we introduced a novel DNA library amplification method based on RPA. At first, we hypothesized that RPA can be a replacement of PCR for DNA library amplification. However, during the experiment, an accelerated small-sized DNA amplification by two-primer RPA reaction caused lower uniformity compared to PCR. To solve this problem, we adapted a linear amplification strategy. We hypothesized that the single-primer linear RPA would result in unbiased, single-stranded DNA amplicons by the strand displacement activity of RPA. As a result, we were able to obtain the more uniform products compared to two-primer RPA products. Nevertheless, the size distribution was not superior compared to PCR products. We believe that further optimization is necessary for linear RPA to be competitive over PCR.

We also examined the uniformity of RPA when amplifying the same-sized oligonucleotide library. We designed and synthesized a 150 bp oligo DNA library from the genome sequence of Synechocystis sp PCC 6803. The RPA product was slightly lower than the PCR amplicons but showed similar uniformity. It was noted that during the analysis, different amplification preferences were found between the PCR and RPA amplified oligo library sequencing data. The correlation coefficient of PCR–PCR or RPA–RPA replicates was close to 1, but the correlation coefficient of PCR–RPA showed 0.6 (Figure b,c). We were unable to elucidate the reason for these different preferences, but acknowledge that differences in RPA versus PCR polymerases may contribute.

As described above, one of the reasons why we used RPA is to reduce the heat deamination of DNA during PCR. To prove it, we analyzed the substitution error rate of our samples. However, the substitution error ratio showed a similar trend as that of the NGS substitution error rate ratio. Thus, we were not able to find any evidence that heat-induced C to T deamination was reduced by RPA. We assumed that the deamination error rate might be lower than the sequencing error rate in our study.

Taken together, we show that single-primer linear RPA can be one of the alternative methods to PCR for DNA library amplification. Still further optimization is needed, and we believe that the linear RPA can be used for simultaneous amplification and barcoding with a single oligo, resulting in shorter reaction times and milder reaction conditions than PCR. Thus, we envision that our study will extend the scope of RPA in the field of genomics.

Materials and Methods

PCR Conditions

All primers were purchased from Integrated DNA Technologies (USA) or Macrogen (Korea) (Table S3). We used KAPA HiFi HotStart Ready Mix (Kapa Biosystems, USA) for PCR experiments. The 50 μL PCR reaction consisted of 25 μL of 2× KAPA HiFi HotStart Ready Mix, 2.4 μL of 10 μM forward primer, 2.4 μL of 10 μM reverse primer, and template DNA or ddH2O (negative control). The amplicon was purified with the MinElute PCR Purification Kit (Qiagen, Germany) according to the manufacturer’s protocol.

RPA Conditions

Two-primer RPA was conducted at 37 °C with the TwistAmp Basic kit (TwistDx Limited, UK) following the manufacturer’s protocol. Briefly, 25 μL of 2× Reaction Buffer, 3.68 μL of 25 mM dNTP mix, 5 μL of 10× Basic E-mix, 2.4 μL of 10 μM P5 primer, 2.4 μL of 10 μM P7 primer, and 44.5 μL of ddH2O were included. Samples were vortexed and spun down, and 2.5 μL of 20× Core Reaction Mix was added to the reaction mixture. The solution was mixed vigorously, centrifuged, and transferred to a new dry tube. Next, 2.5 μL of 200 mM MgOAc and 0.5 μL of template DNA were added separately to the cap. Finally, the contents were centrifuged to start the 20 min reaction and incubated at 37 °C. Amplified DNA was purified with the MinElute PCR Purification Kit (Qiagen, Germany) following the manufacturer’s protocol.

For all single-primer linear RPA reactions, everything was the same until the DNA purification step, except replacing 2.4 μL of P5 primer with 2.4 μL of ddH2O. After purification, linearly amplified RPA products were subjected to eight cycles of PCR with P5 primer and end sequence of P7 primer (Table S3).

NGS Sample Preparation

All NGS samples were prepared using the 5× ER/A-tailing Enzyme Mix (Enzymatics, USA) and WGS ligase (Enzymatics, USA) without deviating from the manufacturer’s protocol. DNA was first mixed with the ER/A-tailing enzyme mix and incubated at 20 °C for 30 min and 65 °C for 30 min. Sequencing adaptor (10 μL) from NEBNext Multiplex Oligos for Illumina (NEB, USA) and WGS ligase were added to the A-tailed product. After ligation for 30 min at 23 °C, 3 μL of USER enzyme mix (New England BioLabs, USA) was directly mixed into the ligation tube and incubated for 30 min at 37 °C. The USER cleaved product was then purified using Agencourt AMPure XP beads with a 1.2× excess volume of USER cleaved product (Beckman-Coulter, USA). The products were amplified with either PCR or RPA using the index primer pairs provided in the NEBNext Multiplex Oligos for Illumina kit. The product was purified using AMPure XP beads with a 1.2× excess volume of AMPure XP beads and sequenced using the Illumina NextSeq platform.

Sequencing Data Analysis

We used the AdapterRemoval,[27] BWA,[28] SAMtools,[29] and in-house python scripts to analyze read counts, sequencing depths, and uniformities of each library.

Generation of Human and E. coli Genomic DNA Library

In this study, we used five different DNA libraries. Four were generated by random shearing of human NA12878 and E. coli EcNR2[26] genomic DNA. Human NA12878 genomic DNA was purchased from the Coriell Institute (USA). E. coli EcNR2 strain was incubated in Luria–Bertani media (BD Biosciences, USA) at 30 °C in a shaking incubator. After harvesting EcNR2 cells by centrifugation, we precipitated the genomic DNA using the GeneAll Exgene Cell SV Kit (GeneAll, Korea) according to the manufacturer’s protocol. Both genomic DNAs were sheared into 180 bp and 300 bp fragments using a M220 focused ultrasonicator (Covaris, USA). The size of sheared DNA was measured by agarose gel electrophoresis and an Agilent 4200 TapeStation system (Agilent, USA).

Design and Synthesis of the Microarray Oligonucleotide Library

The microarray oligonucleotide library was designed to have a thousand unique 150 bp sequences and synthesized with the DNA microarray synthesizer (CustomArray, USA). Each sequence possessed a 30 bp flanking sequence at both sides, which is complementary to the RPA primer. The terminal 20 bp at each end sequence possessed a Tm value near 60 °C, enabling their use as a primer pair for PCR.