DNA assembly with error correction on a droplet digital microfluidics platform

DNA assembly with error correction on a droplet digital microfluidics platform Background: Custom synthesized DNA is in high demand for synthetic biology applications. However, current technologies to produce these sequences using assembly from DNA oligonucleotides are costly and labor-intensive. The automation and reduced sample volumes afforded by microfluidic technologies could significantly decrease materials and labor costs associated with DNA synthesis. The purpose of this study was to develop a gene assembly protocol utilizing a digital microfluidic device. Toward this goal, we adapted bench-scale oligonucleotide assembly methods followed by enzymatic error correction to the Mondrian™ digital microfluidic platform. Results: We optimized Gibson assembly, polymerase chain reaction (PCR), and enzymatic error correction reactions in a single protocol to assemble 12 oligonucleotides into a 339-bp double- stranded DNA sequence encoding part of the human influenza virus hemagglutinin (HA) gene. The reactions were scaled down to 0.6-1.2 μL. Initial microfluidic assembly methods were successful and had an error frequency of approximately 4 errors/kb with errors originating from the original oligonucleotide synthesis. Relative to conventional benchtop procedures, PCR optimization required additional amounts of MgCl , Phusion polymerase, and PEG 8000 to achieve amplification of the assembly and error correction products. After one round of error correction, error frequency was reduced to an − 1 average of 1.8 errors kb . Conclusion: We demonstrated that DNA assembly from oligonucleotides and error correction could be completely automated on a digital microfluidic (DMF) platform. The results demonstrate that enzymatic reactions in droplets show a strong dependence on surface interactions, and successful on-chip implementation required supplementation with surfactants, molecular crowding agents, and an excess of enzyme. Enzymatic error correction of assembled fragments improved sequence fidelity by 2-fold, which was a significant improvement but somewhat lower than expected compared to bench-top assays, suggesting an additional capacity for optimization. Keywords: DNA, Gibson assembly, Digital microfluidics, Error correction Background Gene synthesis is a costly and labor-intensive process. Over the last decade, major research advances in gen- The cost of synthetic DNA is directly related to the cost ome sequencing (i.e. “DNA reading”) are slowly being of oligonucleotides, and a large amount of hands-on matched by advances in synthetic biology (i.e. “DNA labor required for conventional oligonucleotide-based writing”)[1, 2]. Rapid advances in synthetic biology are gene assembly is also a significant cost [6–8]. The fueling a demand for synthetic DNA that will only in- cheapest oligonucleotides that can be purchased from crease in the future. However, the ability to synthesize commercial suppliers are usually unpurified and contain long DNA molecules in a short period of time without errors. Thus, the genes assembled from the unpurified significant expense remains one of the main challenges oligonucleotides must be sequence verified to find a cor- in synthetic biology [3–5]. rect assembly. Implementation of an enzymatic error correction step greatly improves the sequence fidelity of the assemblies, which reduces the number of clones that * Correspondence: griffin@stanford.edu have to be individually cloned and sequence verified [7, Stanford Genome Technology Center, Stanford University, 3165 Porter Drive, 9, 10]. Unfortunately, this extra error correction step Palo Alto, CA 94304, USA also significantly increases the hands-on time required Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Khilko et al. BMC Biotechnology (2018) 18:37 Page 2 of 14 to complete assembly. Integration of digital microfluidics can be replaced and performed by programmable droplet into DNA assembly coupled with error correction can generation and routing over heater bars underneath the potentially ease this labor burden by allowing for a “set microfluidic cartridge. The sequential reactions can be up and walk away” approach to managing the entire performed on a single microfluidic cartridge without any process. human intervention [18]. Software automation programs Digital microfluidics (DMF) is a technology based on can be designed to conduct multiple experiments in paral- the electrowetting phenomenon. The phenomenon lel. Since the devices are fully automated, the sources of describes a change of surface tension at a solid/liquid/ human errors and labor costs can be greatly reduced. In gas interface by application of an electric field [11, 12]. addition, DNA assembly programs can be distributed be- The voltage applied to the electrodes lowers the surface tween laboratories, so that scientists can share robust tension, which leads to a reduction of the contact angle protocols. and increases the wettability of the surface. Conse- Because DNA assembly and error correction reactions quently, the liquid spreads over the surface where the require the use of expensive enzymes, scaling down to voltage is applied. Thus, a hydrophobic surface becomes smaller reaction volumes reduces reagent expenses. Due hydrophilic. By the application of a voltage on a dielec- to the large surface-to-volume ratio, micro-droplet reac- tric surface, the liquids can be transported over the tors have high heat and mass transfer rates. This makes surface of a microfluidic cartridge. it possible to increase kinetics and speed up reactions In electrowetting on dielectric (EWOD) devices, a [13, 17, 19]. Integration of effective error correction pro- droplet is sandwiched between two hydrophobic plates cedures could permit DNA assembly on a single micro- and the remaining volume is filled with immiscible li- fluidic cartridge without the need for expensive and quid, for example, a silicone oil (Fig. 1). The oil prevents lengthy sequence verification. evaporation of the aqueous droplets and facilitates trans- A number of DNA assembly protocols have been de- port. The bottom plate is the array of electrodes, which veloped to date (Table 1). For the scope of this study, can locally control the surface tension by the application only assembly methods from oligonucleotides will be of a voltage. Digital microfluidic devices are completely discussed. The most popular gene construction methods programmable and do not require any pumps or valves for microfluidic applications are polymerase-based and to move the liquids. A cartridge can be inserted into a endonuclease-based assembly. Both approaches utilize micro-controller that is operated by a software program oligonucleotides as DNA building blocks [20]. The [13–15]. The program turns the voltage on and off at polymerase-based assembly method utilizes the same ap- certain electrodes, so the droplets can be directed any- proach as PCR [21–24]), but instead of using forward where on a chip. They can also be dispensed, trans- and reverse primers, oligonucleotides overlap and serve ported, split, fused, mixed and held in certain regions. as templates for a complementary strand. The oligonu- Digital microfluidic devices are applicable for gene as- cleotides are designed to be either a part of the top or sembly because DNA is typically handled in microliter the bottom DNA strand. In the first assembly cycle, oli- amounts. The microfluidic devices are capable of gen- gonucleotides overlap partially, and the polymerase ex- erating droplets in the microliter to picoliter range tends the complementary strand in a 5′ to 3′ direction. [16, 17]. The microliter droplets act like reaction and In the next cycle, the double-stranded DNA pieces are transport vessels. The ability to program liquid hand- separated and hybridized to other oligonucleotides or as- ling operations such as dispense, transport, merge, sembled single-stranded fragments. The process of de- mix, and split allows a researcher to automate and naturation, annealing, and extension is repeated until simplify the gene assembly process. Time-consuming the desired sequence has been built [25]. steps like pipetting, transferring reagents, tube labeling, The studies shown in Table 1 utilized different types incubation at certain temperatures, and thermocycling of microfluidics to perform DNA assembly. Among Fig. 1 A cross-section of an EWOD cartridge Khilko et al. BMC Biotechnology (2018) 18:37 Page 3 of 14 Table 1 Summary of microfluidic assembly methods Authors Type of microfluidics Assembly method Assembly size Error rates Kong et al. [22] Droplet Polymerase-based (PCA) 500 –1000 bp 1.78 errors/kb Huang et al. [21] Microchannel Polymerase-based (PCA) 760 bp 4.1 errors/kb Quan et al. [23] Microarray Polymerase-based (PCA) 500 – 1000 bp 1.9 errors/kb Tian et al. [24] Microarray Polymerase-based (PAM) 14,500 bp 2.2 errors/kb Linshiz et al. [48] Chanel Exonuclease-based (Gibson) and 754 bp N/A polymerase-based (IHDC) Shih et al. [49] Combined digital and droplet microfluidics Exonuclease-based assembly of 2100 bp N/A dsDNA (Gibson) Tangen et al. [50] Droplets Polymerase-based (PCA) and 525 bp N/A Exonuclease-based (Gibson) Ben-Yehezkel et al. [18] Digital microfluidics Polymerase-based (POP) 800 bp 2.22 errors/kb those studies, only the work by Ben-Yehezkel et al. was multiple rounds of error correction after assembly and conducted using digital microfluidics, the same type of PCR to decrease the incidence of errors originating from microfluidics utilized in the present study. The group the oligonucleotides [28]. Gibson assembly has been developed an innovation on the polymerase-based assem- used successfully to assemble entire genes (1.5-1.7 kb) in bly method called programmable order polymerization a single step, and this method is arguably the most effi- (POP). The method was first successfully automated on cient to assemble genes from multiple oligonucleotides the Mondrian™ microfluidic device. The assembly reaction [28]. Using Gibson assembly, higher numbers of oligo- proceeded to assemble the sequence from the inside out. nucleotides can be assembled in a single reaction than In each of four phases, a double-stranded DNA (dsDNA) by PCR assembly. For this reason, we decided to imple- fragment was extended by a pair of oligonucleotides such ment Gibson assembly on a DMF device. that one oligonucleotide bound to each end of the se- To design a DNA assembly protocol for a programmable quence. Multiple rounds of thermocycling for each phase digital microfluidic device, we developed a process consist- with an oligonucleotide pair ensured that most of the ing of three major parts (Fig. 2). First, DNA oligonucleo- product was extended in each step. The group reported tides were assembled into a double-stranded DNA − 1 an error rate of 1 in 450 bp (2.22 errors kb ) for their fragment. Second, the assembly was amplified by PCR, and assembly method, and the errors were identified as third, errors from the original oligonucleotides were substitutions [18]. removed. We used Sanger DNA sequencing of the recov- Whereas Ben-Yehezkel used four separate rounds of ered, error-corrected products to verify the efficiency of the DNA polymerization to sequentially extend and assem- error correction process and develop an efficient DNA as- ble two DNA fragments in each round, we used a single sembly and error correction protocol. The ultimate goal is “one-pot” assembly of 12 fragments in a single droplet to design a reliable and cost-effective DNA assembly proto- and successfully assembled the complete sequence. In col that will be widely applicable in biological research. addition, we performed a round of error correction on the digital microfluidic device. At some point, a large Methods number of DNA fragments in a one-pot reaction would DNA sequences and oligonucleotides lead to miss-hybridization, so it is interesting to contem- Our 339-bp test assembly sequence comprised a partial plate a process that combines both these methods in a sequence from the human influenza virus H9N2 DNA assembly reaction to reduce the assembly time and hemagglutinin (HA) gene (283-bp, nt 211-493 of the HA errors even further. coding region) flanked on each side by 8-bp NotI restric- In summary, the one-pot assembly method used in tion endonuclease sites and 20-bp homology regions to this manuscript is a one-step isothermal Gibson assem- the pUC19 cloning vector. The 339-bp fragment was di- bly developed at the J. Craig Venter Institute ([26]) and vided into 12 overlapping oligonucleotides (See Table 2). is distinctly different in process from the work described The final assembled test sequence is listed in Fig. 3. by Ben-Yehezkel et al. With this technique, double- or single-stranded DNA pieces are joined into longer frag- Mondrian™ digital microfluidics (DMF) device ments by three enzymes: T5 exonuclease, DNA polymer- The main goal of this study was to develop a robust and ase, and Taq DNA ligase. The reagents are incubated at reliable DNA assembly and error correction protocol for 50 °C for 0.5-1 h after which the assembled product is the Mondrian™ DMF device (Illumina, Inc.). The whole typically amplified by PCR [27]. Published protocols use protocol involved four consecutive enzymatic reactions. Khilko et al. BMC Biotechnology (2018) 18:37 Page 4 of 14 Fig. 2 A schematic of influenza HA gene assembly on DMF. This diagram shows the steps of the process which were all performed sequentially on a microfluidic cartridge. The steps are Gibson assembly of 12 oligonucleotides, PCR amplification of a double-stranded DNA piece, error removal using an error correction enzyme, and PCR amplification of corrected sequences Prior to incorporating the four enzymatic steps of the device. To observe the behavior of the droplets, a digital gene assembly in a complete protocol, each enzymatic camera was mounted above the cartridge to produce a step was optimized individually. All liquid handling op- magnified image of the DMF cartridge onto a computer erations were programmed using the Application Devel- screen. The Mondrian™ cartridge that was used in the opment Environment software (Illumina, Inc.). experiments (Fig. 4a), consisted of two plates, a plastic The Mondrian™ microfluidic system included a micro- top plate, and a printed circuit board (PCB) substrate. controller that was connected to a computer and digital The area between the plates was filled with a 2 cSt sili- microfluidic cartridges that were inserted into the cone oil. As seen in Fig. 4b, the configuration of the Table 2 Primers used in this study Primer Name Primer Sequence (5′->3′) Puc-049cloning-R CCGGGTACCGAGCTCGAATTCACTG Puc-049cloning-F GATCCTCTAGAGTCGACCTGCAGGC Puc19-5’F TCCCAGTCACGACGTTGTAAAACGAC HA049-1 CAGGTCGACTCTAGAGGATCGCGGCCGCGACACATGCACTATTGAAGGACTT HA049-2 AGATCACAAGAAGGGTTACCATAGACAAGTCCTTCAATAGTGCATGTGTCGC HA049-3 GTCTATGGTAACCCTTCTTGTGATCTGTTGTTGGGGGGAAGAGAATGGTCCT HA049-4 TACAGCTGATGGTCTTTCAACGATGTAGGACCATTCTCTTCCCCCCAACAAC HA049-5 ACATCGTTGAAAGACCATCAGCTGTAAATGGAACGTGTTACCCTGGGAATGT HA049-6 GTGTTCTGAGTTCCTCTAAGTTTTCCACATTCCCAGGGTAACACGTTCCATT HA049-7 GGAAAACTTAGAGGAACTCAGAACACTCTTTAGTTCCTCTAGTTCCTACCAA HA049-8 ATTGTGTCTGGGAATATTTGGATTCTTTGGTAGGAACTAGAGGAACTAAAGA HA049-9 AGAATCCAAATATTCCCAGACACAATCTGGAATGTGACTTACACTGGAACAA HA049-10 GTAGAATGAATCTGAACATGATTTGCTTGTTCCAGTGTAAGTCACATTCCAG HA049-11 GCAAATCATGTTCAGATTCATTCTACAGGAACATGAGATGGCTGACTCAAAG HA049-12 GAATTCGAGCTCGGTACCCGGCGGCCGCTTTGAGTCAGCCATCTCATGTTCCT Khilko et al. BMC Biotechnology (2018) 18:37 Page 5 of 14 >HA049_sequence GAATTCGAGCTCGGTACCCGGCGGCCGCTTTGAGTCAGCCATCTCATGTTCCTGTAGAATGAATCT GAACATGATTTGCTTGTTCCAGTGTAAGTCACATTCCAGATTGTGTCTGGGAATATTTGGATTCTTT GGTAGGAACTAGAGGAACTAAAGAGTGTTCTGAGTTCCTCTAAGTTTTCCACATTCCCAGGGTAAC ACGTTCCATTTACAGCTGATGGTCTTTCAACGATGTAGGACCATTCTCTTCCCCCCAACAACAGAT CACAAGAAGGGTTACCATAGACAAGTCCTTCAATAGTGCATGTGTCGCGGCCGCGATCCTCTAGA GTCGACCTG Fig. 3 Oligonucleotide alignment for the 339-bp test assembly sequence. a Arrangement of DNA oligonucleotides used for assembly of the HA049 sequence. b FASTA formatted HA049 sequence DMF cartridge allowed eight processes to be performed used to set temperatures for the enzymatic reactions. in parallel. The reagents were loaded through 50 μLor Additionally, an area of the cartridge could be cooled 10 μL ports on the cartridge top plate, and the samples down with a Peltier device. Fig. 4c shows a close-up view were withdrawn through other ports. There were also of one lane with three different temperature zones, seven reservoirs dedicated for the collection of waste which were maintained during reactions using the droplets. The microfluidic cartridge had three heater heaters and the cooler. The device was operated by the bars that contacted the back of the PCB, which were Application Development Environment (ADE) software. Fig. 4 The Mondrian™ microfluidic cartridge. a Image of the cartridge. b Diagram of the cartridge electrode paths. This diagram of the chip comes from a screenshot of an ADE software. There are 50 μL reservoirs highlighted in blue, 10 μL reservoirs highlighted in red. Orange reservoirs were used to hold waste. Green reservoirs were used to collect final products. The configuration of the chip allowed to perform 8 reactions in parallel. c Magnified image of one lane of the microfluidic cartridge where the reactions were performed. The area highlighted in yellow was used for error correction reaction. The area highlighted in purple was used for Gibson assembly and PCR annealing/extension. The area highlighted in red was used for DNA denaturation during PCR and error correction pretreatment Khilko et al. BMC Biotechnology (2018) 18:37 Page 6 of 14 Prior to each experiment, a program was designed to An automation program for the DMF was developed direct droplets through the liquid handling operations. to perform microfluidic PCR amplification experiments. The device was operated at voltages between 90 V and The reactions were carried out in 1.2 μL droplets. The 300 V and at a frequency of 30 Hz. droplets were brought to the PCR area of the chip, The liquid volumes of 0.3, 0.6, and 1.2 μL were gener- which consisted of two temperature zones. The denatur- ated and manipulated on the microfluidic cartridge. To ation zone was set to 98 °C, and the annealing/extension dispense a 0.3 μL or 0.6 μL droplet, three electrodes zone was set to 72 °C. adjacent to the reagent input port were activated, which The thermocycling was performed at a reduced voltage caused the liquid to spread over three electrodes (Fig. (90 V), which eliminated nonspecific adsorption of poly- 5a). The electrode #2 was switched off to obtain a 0.3 μL merase on the microfluidic surface and reduced spurious droplet (Fig. 5b). The double 0.6 μL droplet was dis- bubble formation at high temperatures [18, 27, 29–32]. pensed by turning off the electrode #3 (Fig. 5c). To cre- An initial denaturation was performed by moving the ate a 1.2 μL droplet, two 0.6 μL droplets are brought droplets to the 98 °C zone where they were held for next to each other and separated by one inactive elec- 30 s. Then thirty PCR cycles were performed by cycling trode as shown in Fig. 5d. Then the electrode between the droplets from 98 °C to 72 °C at 1.5 s/electrode, and the two 0.6 μL droplets was turned on, merging both from 72 °C to 98 °C at 1 s/electrode. Annealing/exten- into one 1.2 μLdroplet (Fig. 5e). Refer to Additional sion of the droplets was done by switching on/off volt- file 1: Video 1 to see the liquid handling operation de- age of the area of three electrodes every half of a second scribed here. Materials to see all liquid handling operations for 20 s, and denaturation was performed by holding the used in present work. droplets at 98 °C for 10 s. After 30 cycles of PCR, DNA was held for 10 min at 72 °C to allow a final extension. Then, the voltage was switched back to 300 V, so that Optimization of microfluidic PCR the samples could be transported to the collection Each on-chip PCR reaction contained 1X HF Phusion reservoirs. buffer, detergent-free (Thermo Fisher Scientific), 0.8 μM forward and reverse primers, 1.75 ng/μL HA-049 DNA template (plasmid-cloned HA-049 sequence), 0.1 U μL Optimization of microfluidic Gibson assembly Phusion polymerase (Thermo Fisher Scientific). The In each experiment, 50 μL master mixes were made reactions were set up to allow the addition of 1 mM from fresh reagents and prepared according to previ- MgCl , 1.25 mM PEG 8000, 0.2 mM NAD, 2 mM DTT ously published Gibson Assembly protocols [33] with to the reaction mixture or the combination of 1.25 mM additional modifications described below. Assembly, PEG 8000 and 1 mM MgCl , 0.2 mM NAD, and 2 mM oligonucleotide, and PCR master mixes were prepared at DTT. Final concentrations in the reaction droplets are 2X concentrations such that once two equal size given for all reagents. droplets were merged, the final enzyme mixtures would Fig. 5 Generation of 0.3, 0.6, and 1.2 μL droplets on DMF. a Stretching the liquid over three electrodes. b Generation of 0.3 μLdroplet. c Generation of 0.6 μLdroplet. d Two 0.6 μL droplets separated by 1 electrode. e 1.2 μL droplet resulted from merging two 0.6 μL droplets Khilko et al. BMC Biotechnology (2018) 18:37 Page 7 of 14 be at the correct 1X concentration. The oligo master original size (339-bp) and the cleaved size (approxi- mix containing a mixture of all oligonucleotides was pre- mately 170-bp). In case of a failure, only one 339-bp pared by diluting a 1 μM stock solution in DI water con- band was visible. Preliminary experiments demonstrated taining 0.01% Tween 20. The surfactant was a necessary that a microfluidic error correction reaction with stand- component to reduce the surface tension, which facili- ard benchtop reagents was not successful due microflui- tated droplet dispensing and movement. The amount of dic surface interactions and nonspecific adsorption of surfactant required was determined for each master mix. CorrectASE™. To test the hypothesis of CorrectASE™ ad- The enzymes suspended in storage buffers contained sorption on the droplet oil/water interface, the mix of stabilizers. It was observed that the droplets with en- DNA was treated with additional reagents. The reactions zyme solutions were easily dispensed and manipulated were performed with extra CorrectASE™, 0.01% Tween on a cartridge without any additional surfactant. Thus, 20, 1.25 mM PEG 8000, and 2.5 mM MgCl to deter- the assembly master mix and PCR master mix did not mine which could improve reaction performance. contain Tween 20. The final (1X) concentrations of re- agents in the assembly reaction were 1X isothermal (iso) Protocol for DNA assembly with error correction − 1 − 1 buffer, 0.05 U μL of Phusion polymerase, 4 U μL The protocol consisted of four consecutive enzymatic re- − 1 DNA ligase, 0.08 U μL T5 exonuclease, and 250 nM actions. The process started with Gibson assembly that oligonucleotides. was carried out for 60 min. Then, the assembly products To perform DNA assembly experiments, an automa- were amplified in 30 PCR cycles. Next, DNA was treated tion program was created. The temperature in the as- with CorrectASE™ for 60 min. The error correction sembly area was set to 50 °C. Next, 0.3 μL droplets products were amplified in a second PCR. According to containing oligonucleotides were dispensed. The drop- this protocol, the final concentrations of reagents in Gib- lets were transported to a waiting area where they were son assembly reaction were 1X isothermal (iso) buffer, 0. − 1 held while another dispenser generated 2X Gibson mas- 05 U μL of Phusion polymerase (Thermo Fisher − 1 − 1 ter mix droplets. The oligonucleotide and Gibson master Scientific), 4 U μL DNA ligase (NEB), 0.08 U μL T5 mix droplets were merged to obtain a double 0.6 μL exonuclease (NEB), and 50 nM oligonucleotides (IDT droplet and brought to the assembly area where they DNA). After assembly, the product was diluted with 0. were incubated for 15-60 min at 50 °C. To ensure ad- 01% Tween 20 (Sigma Aldrich) by 8-fold. Diluted assem- equate mixing, the droplets were moved up and down blies were merged with equal size droplets of PCR mas- − 1 over 4 electrodes. When the reaction was finished, the ter mixes to achieve 0.1 U μL Phusion polymerase assembly droplets were merged with 0.6 μL PCR drop- (Thermo Fisher Scientific), 1X HF detergent-free buffer lets, so the total volume of each droplet became 1.2 μL. (Thermo Fisher Scientific), 0.25 mM of each dNTP The polymerase chain reaction was performed as de- (Thermo Fisher Scientific), 0.8 μM of forward and re- scribed above. After PCR, the products were diluted. To verse primers (IDT DNA, 0.625 mM PEG 8000 (Sigma), perform dilutions, a dispenser containing DI water and 0.5 mM MgCl (Thermo Fisher Scientific) in the reac- 0.05% Tween 20 generated 0.6 μL droplets. Then, the tions. After amplification, two of the eight droplets were droplets were merged with the assembly droplets, mixed, recovered from the chip, and the rest of the droplets and split into two equal size droplets. This step was iter- were diluted by 2-fold with 0.01% Tween 20 solution to ated to achieve the desired dilutions. When assembly continue to the error correction step. time was variable, 0.6 μL droplets containing both the The EC denature/anneal step of the protocol was imple- oligonucleotides and Gibson assembly reagents were mented to expose the errors in DNA sequence for further held in a waiting area, and two droplets were moved to CorrectASE™ treatment. During the denaturation step, − 1 the assembly incubation zone in 15 min increments. DNA was diluted to 20-25 ng μL in 1X CorrectASE™ This way, each condition was tested twice in two differ- buffer and incubated at 98 °C for 2 min, 25 °C for 5 min, ent experimental droplets. and 37 °C for 5 min. Then, the droplets were merged with CorrectASE™ master mix to a final concentration of 2X Optimization of enzymatic error correction CorrectASE™ (Invitrogen), 0.01% Tween 20 (Sigma The optimization of enzymatic error correction step was Aldrich), and 2.5 mM PEG 8000 (Sigma). The master performed using a mix of two equal molar amounts of mixes contained double amounts of reagents to obtain 1X 339-bp PCR products. Sequences were amplified from concentration after the equal size droplets were merged. two DNA templates. The first template had a completely The reagents were loaded on a DMF cartridge into correct sequence and the other had a mutation approxi- dedicated dispensers as prescribed by the automation mately in the middle of the 339-bp sequence. If the error program. All master mixes except CorrectASE™ were correction reaction was successful, two DNA bands were loaded on the cartridge at the beginning of the process. visualized on an agarose gel corresponding to the To ensure that the enzyme stayed active, CorrectASE™ Khilko et al. BMC Biotechnology (2018) 18:37 Page 8 of 14 was loaded into the dispenser three minutes before it was x  1000 f ¼ ð1Þ to be used by the program. At the end of the process, all n  l droplets were collected in 20 μL water containing 0.05% Tween 20 and retrieved from the device manually. where x is the number of errors in a single clone, n is the number of sequenced clones not including clones Cloning and sequencing of assembled and amplified with mis-assemblies, and l is the length of a sequence in products bases. Recovered products were brought up to 50 μLin water and an equal volume of Agencourt AMPure XP Results beads (Beckman Coulter) was added and mixed. After Optimization of microfluidic PCR 5 min incubation to bind DNA to the beads, the tube Optimization of PCR on DMF demonstrated that the was placed on a magnet and allowed to settle for additives improved amplification efficiency. Control 5 min. The supernatant was removed and the beads − 1 samples, which contained 0.1 U μL of Phusion were washed two times with 80% ethanol. After a polymerase, did not show any bands on the agarose gel final 5 min incubation with the caps open to allow (data not shown). On the other hand, PCR that the beads to dry, the DNA was eluted in 15 μL contained the iso buffer used for DNA assembly resulted 10 mM TRIS buffer (pH 8.5). in the desired 339-bp bands. To determine the compo- The purified products were assembled into a pUC19 nents of the iso buffer that contributed to the successful vector that was amplified by primers Puc-049cloning-R PCR reaction, we tested each component individually + Puc-049cloning-L (Table 2) using OneTaq polymerase and in combination. When PEG 8000, DTT, NAD, and (NEB). Assembly of the product into the pUC19 vector MgCl were added separately to the reaction, only PEG was by Gibson assembly [7, 8, 33], assembly reactions 8000 demonstrated some amplification of DNA tem- were electroporated into E. coli strain Epi300 (Epicentre) plate, but the result was not as good as PCR supple- , and resulting clones were selected on LB plates con- mented with the iso buffer (data not shown). Based on − 1 taining ampicillin at 100 μgmL . Colonies were these results PEG 8000 was combined with either NAD, screened using 20 μL colony PCR reactions with primers DTT, or MgCl to find out if DNA would be amplified pUC19-5’F and pUC19-3’R (OneTaq, NEB). Colonies at the same level as with the iso buffer. As seen in Fig. 6, containing a plasmid with an insert sequence of 339- the combination of 1.25 mM PEG 8000 and 1 mM bp were grown overnight in 5 mL LB broth and puri- MgCl showed comparable band intensity as the iso buf- fied using the QIAprep miniprep kit (Qiagen). The fer. This result demonstrated that microfluidic PCR insert sequences of the resulting plasmids were ana- must be performed with an excess of Phusion enzyme lyzed by Sanger DNA sequencing. For each treatment, and supplemented with additional MgCl and PEG 8000. 10-20 independent clones were sequenced using the Multiple methods to reduce biofouling in microfluidics pUC19-5’Fprimer. have been tested in this work. The only method that im- proved PCR yield and transport of droplets was the re- Data analysis duction of the electrowetting voltage from 300 V to Samples recovered from Lanes 1 and 2, 3-8, which cor- 90 V during high-temperature PCR [18]. For more responded to assembly-only and EC treatments, respect- ively, were pooled and analyzed by DNA gel electrophoresis on a 2% agarose gel, and using the 1 Kb plus DNA ladder (Invitrogen) as a size standard. The indication of a successful experimental run was the presence of a 339-bp band. For the in-depth error ana- lysis, the samples were cloned into pUC19 vectors and Sanger sequenced. Sequencing data was analyzed by aligning sequencing output files with the template DNA (Additional file 2). Each sequence alignment was inspected for errors in the newly assembled sequence. The errors were categorized in three groups: deletions, insertions, and substitutions. The sequences that had Fig. 6 On-chip polymerase chain reaction performed with two components of the iso buffer as described in the text. DNA from misincorporated oligonucleotides were treated as “mis- the reactions was separated by agarose gel electrophoresis on a assemblies”. The error frequency per 1 kb (f) was calcu- 2% agarose gel lated using Eq. 1 [34]. Khilko et al. BMC Biotechnology (2018) 18:37 Page 9 of 14 information refer to the Additional file 3: Video 2 in- improve the performance of CorrectASE™ enzyme. The cluded in Additional files. reactions were supplemented with either PEG 8000, Tween 20, and excess CorrectASE™ or combination of Optimization of microfluidic Gibson assembly the additives. Our test for CorrectASE™ activity was to The first group of experiments tested optimum reaction add a mixture of two PCR products with one containing time for oligonucleotide assembly. When we tested reac- a nucleotide mismatch in the sequence relative to the tion times spanning 15-60 min, the bands for all tested other at approximately the midpoint of the sequence. times were of similar brightness (Fig. 7), suggesting the Thus, successful error correction led to cleavage of the oligonucleotides were assembled in the 15-60 min time full-length product and resulted in two bands (339-bp period. and 170-bp) on the agarose gel with comparable inten- Figure 8 demonstrates the results of microfluidic as- sities. As seen in Fig. 10, the presence of 0.01% Tween sembly experiments with a dilution of assembly con- 20, 2X CorrectASE™ and 1.25 mM of PEG 8000 in the structs prior to PCR. Dilution of the assembly constructs reaction droplet gave the most even brightness in both from 2-fold to 16-fold resulted in comparable amounts bands on the agarose gel. of the PCR product. However, 16-fold was the maximum dilution rate that could be achieved before the PCR tem- Validation of DMF protocol for assembly with error plate was too dilute to amplify. A dilution rate greater correction than 32-fold did not result in amplification of the assem- The error analysis of DNA assembly of 12 oligonucleo- bly product. tides followed by CorrectASE™ treatment is shown in To investigate if the concentration of oligonucleotides Fig. 11 and Additional file 2. The average error fre- in the assembly reaction could influence the fidelity of as- quency of assembly samples from three separate runs − 1 sembly constructs, two sets of samples obtained by the as- was found to be about 4 errors kb , which corresponds sembly of 50 nM or 250 nM oligonucleotides were to what is widely reported for phosphoramidite DNA sequenced. The average error rate from five separate runs synthesis chemistry, where error rates of approximately for each oligonucleotides concentration is shown in Fig. 9. 1:200 are typical. The average error frequency of the − 1 It was determined that the average error rate for 250 nM samples after error correction was about 2 errors kb , and 50 nM oligonucleotides was similar, at 3.15 errors which corresponds to an average error reduction by 2- − 1 − 1 kb and 2.94 errors kb , respectively. fold. The average error reduction using a conventional Single-base deletions comprised the bulk of errors. benchtop protocol with the same sequence was found to There was no preference for errors to occur between A/ be about 10-fold (data not shown). T or C/G bases. Both sets of samples had comparable As seen in Table 3, the enzyme was effective in remov- percentages of deletions and the same amount of ing deletion and insertion errors but failed to remove multiple-base deletions. The assembly of 50 nM oligonu- substitutions. Overall, error correction was successful, cleotides resulted in 80% deletion errors, whereas and in all three experiments about half of the sequenced 250 nM assembly had 83% deletions, with the remainder clones were found to be error-free. being multiple base deletions, insertions or substitutions. Discussion Optimization of enzymatic error correction Optimization of microfluidic PCR An optimization of the error correction reaction was The results of the microfluidic PCR experiments demon- conducted to determine the reagents that reduce adsorp- strated that due to the high surface-to-volume ratio, re- tion of CorrectASE™ on the oil/water interface and actions performed on the microfluidic device show a Fig. 7 On-chip oligonucleotide assembly and PCR amplifications testing different assembly reaction durations. DNA from the reactions was separated by agarose gel electrophoresis on a 2% agarose gel Khilko et al. BMC Biotechnology (2018) 18:37 Page 10 of 14 Fig. 8 Dilution of the oligonucleotide assembly product (assembled with 50 nM each oligo) prior to PCR. a 2 to 16-fold and b 32 to 128-fold. After dilution and PCR, products were separated by agarose gel electrophoresis on 2% agarose gels strong dependence on surface interactions. Protein must be increased up to 10-fold in microfluidic droplets molecules can adsorb at the oil/water interface, which [36, 37]. The results of PCR experiments presented here reduces the surface tension over time [27, 35]. Addition- demonstrated that sufficient and repeatable PCR amplifi- ally, adsorption of a protein at the droplet aqueous/oil cation could be achieved with a 5-fold increase of interface could facilitate segregation of hydrophobic Phusion polymerase. groups that may lead to change in protein conformation The efficiency and specificity of PCR are affected by 2+ and inactivation. At high temperatures, the exposed the Mg concentration. Magnesium ions help the hydrophobic groups of the protein could lead to protein polymerase to fold in the active conformation [38]. Also, 2+ denaturation. The combined effect of protein adsorption Mg stabilizes dsDNA and increases the melting and denaturation could reduce the amount of available temperature (T ) of primers. Thus, it is crucial to have enzyme and decrease reaction efficiency. It has been re- the correct amount of free magnesium, and this ported previously that to achieve amplification efficiency concentration must often be optimized for each primer similar to a benchtop PCR, the amount of polymerase pair. It has been observed that the concentration of free Fig. 9 Average error frequency for sequences assembled from 250 nM and 50 nM oligonucleotides. Average error frequency from five independent experiments is plotted with error bars indicating one standard deviation from the mean Khilko et al. BMC Biotechnology (2018) 18:37 Page 11 of 14 Fig. 10 CorrectASE™ optimization on the DMF platform. DNA from the reactions was separated by agarose gel electrophoresis on a 2% agarose gel 2+ Mg can be decreased because of precipitation on Also, the polymers preserve native protein conformation microfluidic surfaces, capture by chelating agents and facilitate binding to a substrate. It has been shown present in reagents and storage buffers, and by binding that PEG 8000 stabilized Taq polymerase at high tem- to dNTPs [37]. According to the Phusion polymerase peratures [39, 41]. Since Phusion is a polymerase, it is product literature (Thermo), the optimum concentration possible that PEG 8000 formed weak bonds with the en- of MgCl is between 0.5-1 mM. The experimental zyme and reduced hydrophobic interactions with the results demonstrated that the addition of 0.5-1 mM of Teflon coating. As a result, the activity of the enzyme magnesium to the 1.5 mM MgCl present in Phusion was increased and amplification yield was improved HF buffer, improved polymerase activity, but this effect [42]. Consequently, microfluidic PCR is affected by ad- was inconsistent from lane to lane. However, it was sorption as well as by interactions of reaction compo- shown that the synergistic effect of magnesium and nents with interfaces. In order to achieve amplification PEG 8000 created favorable conditions for PCR on the DMF, the reaction must be carried out with the − 1 amplification (Fig. 6). final concentration of 0.1 U μL of Phusion (a 5-fold Polyethylene glycol (PEG) is recognized as a molecular increase relative to standard benchtop conditions), 0.5- crowding agent and frequently used as a PCR enhancer 1 mM of MgCl , and 0.625-1.25 mM of PEG 8000. and an enzyme immobilization agent [39, 40]. Molecular Improved PCR yield at 90 V showed that at the lower crowding creates the conditions similar to a natural cell voltage, the oil film between the aqueous droplet and environment in which the enzyme was evolved. It was the Teflon coated surface stayed intact and eliminated reported that macromolecular crowding affects the en- hydrophobic interactions between the polymerase and zyme reaction kinetics by increasing the viscosity of a the surface. According to Kleinert et al., the actuation medium that in turn influences the diffusion of reagents. voltage has a significant influence on the oil film [31]. At Fig. 11 Average error frequency of assembly samples followed by CorrectASE™ treatment. The average error rate of three independent experiments is plotted with error bars indicating one standard deviation from the mean Khilko et al. BMC Biotechnology (2018) 18:37 Page 12 of 14 Table 3 Error analysis of DNA obtained using DMF protocol Run 1 Run 2 Run 3 Assembly EC Assembly EC Assembly EC Deletions 10 4 9 4 10 4 Insertions 0 0 2 0 1 0 Substitutions 3 3 1 0 0 1 Total clones sequenced 8 14 10 10 9 7 Total of clones with correct sequences 0 7 1 5 0 2 − 1 high actuation voltage when the droplet moves, the film Kosuri et al. reported 4 errors kb , and Yehezkel et al. − 1 becomes unstable, breaks down, and tiny oil droplets get reported 2.2 errors kb [18, 44–46]. The analysis of trapped under the aqueous phase. error types demonstrated that the majority of errors In addition, excess surfactant destabilizes the oil film. belonged to single-base deletions with a small percent- Mohajeri and colleagues demonstrated that the critical age of insertions and substitutions. These results are micellar concentration of nonionic surfactants such as comparable to 75.6% deletions, 2.2% insertions, and 22. Tween 20 decreases at higher temperatures [32]. Thus, 2% substitutions, obtained by Sequeira et al. [46]. How- in the denaturation zone, less surfactant is necessary to ever, several clones in both data sets had mis- reduce the surface tension. If there is an excessive incorporated oligonucleotides. This issue could be solved amount of surfactant, the oil film becomes unstable, and by improving the design of the overlapping oligonucleo- the adsorption of the protein occurs, which is further tide sequences. Since the 50 nM oligonucleotide dataset enhanced at high temperatures. It is important to use had 1.5 times more clones with mis-assemblies, degrad- lower voltage and minimize the amount of Tween 20 to ation of some oligonucleotides by T5 exonuclease could avoid loss of Phusion polymerase and subsequent drop- be the cause of mis-incorporation. The results demon- let transport failure. strated that the Gibson assembly method performed on the DMF device is efficient. The error frequencies for Optimization of microfluidic Gibson assembly microfluidic synthesized sequences are in line with those Microfluidic DNA assembly protocols developed in this found for benchtop DNA synthesis in the published work produce results similar to the results published in literature. the literature. Our results show that even 15 min was an acceptable length of time for an efficient microfluidic Optimization of enzymatic error correction DNA assembly. In bench-top reactions, DNA assembly Since the best CorrectASE™ activity was obtained with 0. reactions proceed in reaction times between 15 and 01% Tween, 2X CorrectASE™, and 1.25 mM PEG addi- 60 min [26, 33, 43]. tives, the adsorption of the enzyme on the oil/water Dilution of the assembly product prior to the PCR interface of the aqueous droplet is the most likely ex- amplification is an additional step that should be in- planation of error correction in previous runs. Accord- cluded in a microfluidic Gibson assembly protocol. Since ing to Baldursdottir et al., protein molecules tend to the goal was to assemble the product that had the mini- aggregate on the oil/water interface in a multilayer. The mum number of errors, it was important to remove adsorption rate is affected by the molecular weight and a unreacted oligonucleotides, oligonucleotide fragments, saturation concentration. Large protein molecules tend and mis-assemblies that were present at a low level be- to adsorb faster than small ones due to the large surface fore amplification. Based on these results, we kept the area available for contact with the interface. Also, hydro- dilution of the assembly product no greater than 16-fold. phobic proteins tend to adsorb more due to interactions If the dilution step before PCR is employed, the amplifi- with the hydrophobic coated surface [47]. If some of the − 1 cation mix must contain 0.1 U μL of Phusion, 0. protein molecules adsorb on the interface, hydrophobic 625 mM PEG 8000, and 0.5 mM MgCl . and hydrophilic groups will rearrange, and it will cause The results of the error analysis suggest that the con- the protein to change conformation, leading to a loss of centration of oligonucleotides during assembly did not activity, and the reaction will not proceed with the max- affect the fidelity of the resulting sequence. Both DNA imum yield. assembly methods demonstrated an error frequency in We previously showed for PCR reactions that the − 1 the 1-10 errors kb range, which was similar to the presence of a molecular crowding agent such as PEG values reported in the literature for microfluidic DNA significantly increased the activity of Phusion polymer- assembly [9]. For instance, Saem et al. reported 1.9 ase. According to Sasaki et al., the activity of DNase I to − 1 − 1 errors kb , Sequeira et al. reported 3.45 errors kb , degrade supercoiled DNA and linear DNA was improved Khilko et al. BMC Biotechnology (2018) 18:37 Page 13 of 14 in the presence of 20% w/v PEG [40]. A kinetic analysis Additional files demonstrated that the rate of the DNA cleavage reaction Additional file 1: Video demonstration of droplet liquid handling increased with the increasing of concentration of PEG. operations used in the study. (MP4 9178 kb) However, molecular crowding did not improve the Additional file 2: Sequencing reads aligned to a template. (PDF 1973 kb) activity of Exonuclease III and inhibited the activity Additional file 3: Video demonstration of droplet movement during of Exonuclease I [40]. Consequently, macromolecular 300 V and 90 V PCR cycle. (MP4 19348 kb) crowding could be the reason why the error correc- tion reaction was improved on the DMF platform Abbreviations with the addition of PEG. DMF: Digital microfluidics; DTT: Dithiothreitol; EC: Error correction; NAD: Nicotinamide adenine dinucleotide; PCR: Polymerase chain reaction; Surfactants in digital microfluidic electrowetting on PEG: Polyethylene glycol dielectric (EWOD) are very important. The excess of surfactant can lead to a destruction of the oil film under Acknowledgements the droplet, which could cause the adsorption of hydro- We gratefully acknowledge the supply of electrowetting instruments and cartridges from Illumina, Inc. under a joint project agreement with Stanford phobic molecules on the microfluidic surface. Insuffi- University. cient surfactant could also cause interface instability that in turn could cause the adsorption of enzymes on the Funding The study was funded by NIH grant 5R21GM104694-03. oil/water interface. Usually, to be able to generate and manipulate the droplets on DMF the concentration of Availability of data and materials Tween 20 has to be 0.01-0.05% [18, 31]. However, en- Small-scale Sanger sequencing data is archived with the researchers and will zymatic reactions contain multiple components which be made available upon request. can potentially affect surface tension. Thus, the amount Authors’ contributions of Tween 20 has to be optimized for individual reac- YK, PBG, PDW, JIG, MDA designed experiments. YK, PBG, PDW conducted tions. It has been demonstrated in this study that even research and performed experiments. YK,PBG,PDW,JIG, MDA,MAM analyzed data and wrote the manuscript. All authors read and approved the presence of 0.001% of Tween 20 in reaction droplets the final manuscript. in conjunction with the excess of CorrectASE™ and PEG 8000 gives reproducible error correction results. Ethics approval and consent to participate Not applicable. Validation of DMF protocol for assembly with error Competing interests correction The authors declare that they have no competing interests. The results of our microfluidic protocol demonstrated that some inhibition of CorrectASE™ was still occurring Publisher’sNote during error correction on the DMF platform. Lower Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. error reduction could also be related to over dilution of error correction products prior to PCR or errors in Author details amplification. This suggests that further optimization is Stanford Genome Technology Center, Stanford University, 3165 Porter Drive, Palo Alto, CA 94304, USA. Department of Biomedical, Chemical and possible on the DMF device. Materials Engineering, San Jose State University, 1 Washington Sq, San Jose, CA 95192, USA. J. Craig Venter Institute, 4120 Capricorn Lane, La Jolla, CA Conclusion 92037, USA. Oligonucleotide assembly and error correction protocols Received: 31 August 2017 Accepted: 24 April 2018 for the Mondrian™ digital microfluidic device were developed. The process involved automation of the poly- merase chain reaction, Gibson assembly of 12 oligonu- References 1. Hutchison CA 3rd, Chuang RY, Noskov VN, Assad-Garcia N, Deerinck TJ, cleotides, and enzymatic error correction reaction with Ellisman MH, Gill J, Kannan K, Karas BJ, Ma L, Pelletier JF, Qi ZQ, Richter RA, CorrectASE™. The final protocol consisted of the assem- Strychalski EA, Sun L, Suzuki Y, Tsvetanova B, Wise KS, Smith HO, Glass JI, bly of oligonucleotides, two PCR steps, and an error cor- Merryman C, Gibson DG, Venter JC. Design and synthesis of a minimal bacterial genome. Science. 2016;351:6280. rection reaction. To achieve PCR amplification on the 2. Shendure J, Ji H. Next-generation DNA sequencing. Nat Biotech. 2008; DMF platform, the reactions were supplemented with 26(10):1135–45. PEG, MgCl , and 5-fold increased amount of polymerase 3. Hughes AR, Miklos, EA, Ellington AD. Gene Synthesis: Methods and Applications, Methods in Enzymology. 2011;498:277–309 (relative to benchtop conditions). The error correction 4. Chris Voigt. Synthetic Biology, Part B; Computer Aided Design and DNA reaction was supplemented with PEG, Tween 20, and an Assembly. Elsevier Press. 2011: p. 550. excess of CorrectASE™ (2-fold increase relative to bench- 5. Kosuri S, Church GM. Large-scale de novo DNA synthesis: technologies and applications. Nat Meth. 2014;11(5):499–507. top conditions). The final protocol assembled DNA se- 6. Gibson DG: Gene and genome construction in yeast. In Curr Protoc Mol − 1 quences with an average of 4 errors kb and reduced Biol. Volume Chapter 3. Edited by Ausube FM. 2011. DOI: https://doi.org/10. errors after error correction by 2-fold. 1002/0471142727.mb0322s94. Khilko et al. BMC Biotechnology (2018) 18:37 Page 14 of 14 7. Gibson DG, Lei Y, Chuang R-Y, Venter JC, Hutchison CA, Smith HO, 32. Mohajeri E, Noudeh GD. Effect of temperature on the critical micelle Enzymatic assembly of DNA molecules to several hundred kilobases. Nat concentration and Micellization thermodynamic of nonionic surfactants: Meth. 2009;6:343–5. Polyoxyethylene Sorbitan fatty acid esters. E-Journal of Chem. 2012;9(4): 8. Gibson DG, Glass JI, Lartigue C, Noskov VN, Chuang R-Y, Algire MA, Benders 2268–74. GA, Montague MG, Ma L, Moodie MM, Merryman C, Vashee S, Krishnakumar 33. Gibson DG, Caiazza N, Richardson TH. Materials and methods for the synthesis R, Assad-Garcia N, Andrews-Pfannkoch C, Denisova EA, Young L, Qi Z-Q, of error-minimized nucleic acid molecules US 20130225451 A1. 2013. https:// Segall-Shapiro T, Calvey CH, Parmar PP, Hutchison CA, Smith HO, Venter JC. www.google.com/patents/US20130225451. Accessed 25 Jun 2017. Creation of a bacterial cell controlled by a chemically synthesized genome. 34. Fuhrmann M, Oertel W, Berthold P, Hegemann P. Removal of mismatched Science. 2010;329:52. bases from synthetic genes by enzymatic mismatch cleavage. Nucleic Acids Res. 2005;33(6):e58. 9. Ma S, Saaem I, Tian J. Error correction in gene synthesis technology. Trends 35. Beverung CJ, Radke CJ, Blanch HW. Protein adsorption at the oil/water Biotechnol. 2011;30(3):147–54. interface: characterization of adsorption kinetics by dynamic interfacial 10. Ma S, Tang N, Tian J. DNA synthesis, assembly and applications in synthetic tension measurements. Biophys Chem. 1999;81(1):59–80. biology. Curr Opin Chem Biol. 2012;16(3):260–7. 36. Krishnan M, Burke DT, Burns MA. Polymerase chain reaction in high surface- 11. Lee J, Moon H, Fowler J, Schoellhammer T, Kim C. Electrowetting and to-volume ratio SiO2 microstructures. Anal Chem. 2004;76(22):6588–93. electrowetting-on-dielectric for microscale liquid handling. Sensors 37. Wang F, Burns MA. Performance of nanoliter-sized droplet-based Actuators A Phys. 2002;95(2):259–68. microfluidic PCR. Biomed Microdevices. 2009;11(5):1071–80. 12. Pollack MG, Fair RB, Shenderov AD. Electrowetting-based actuation of liquid 38. Patel PH, Loeb LA. Getting a grip on how DNA polymerases function. Nat droplets for microfluidic applications. Appl Phys Lett. 2000;77(11):1725–6. Struct Mol Biol. 2001;8(8):656–9. 13. Teh S, Lin R, Hung L, Lee AP. Droplet microfluidics. Lab Chip. 2008;8(2):198–220. 39. Zimmerman SB, Harrison B. Macromolecular crowding increases binding of 14. Baroud CN, Gallaire F, Dangla R. Dynamics of microfluidic droplets. Lab DNA polymerase to DNA: an adaptive effect. Proc Natl Acad Sci U S A. 1987; Chip. 2010;10(16):2032–45. 84(7):1871–5. 15. Song JH, Evans R, Lin Y, Hsu B, Fair RB. A scaling model for electrowetting- 40. Sasaki Y, Miyoshi D, Sugimoto N. Effect of molecular crowding on DNA on-dielectric microfluidic actuators. Microfluid Nanofluid. 2009;7(1):75–89. polymerase activity. Biotechnol J. 2006;1(4):440–6. 16. Vergauwe N, Witters D, Ceyssens F, Vermeir S, Verbruggen B, Puers R, 41. Saiki RK, Scharf S, Faloona F, Mullis KB, Horn GT, Erlich HA, Arnheim N. Lammertyn J. A versatile electrowetting-based digital microfluidic platform Enzymatic amplification of beta-globin genomic sequences and restriction for quantitative homogeneous and heterogeneous bio-assays. J Micromech site analysis for diagnosis of sickle cell anemia. Science. 1985;230(4732): Microengineering. 2011;21(5):054026. 1350–4. 17. Jebrail MJ, Fau BM, Patel KD. Digital microfluidics: a versatile tool for 42. Xia YM, Hua ZS, Gular E, Srivannavit O, Ozel AB. Minimizing the surface applications in chemistry, biology and medicine. Lab Chip. 2012;12:2452–63. effect on PCR in PDMS-glass chips by dynamic passivation. J Chem Technol 18. BenYehezkelT,Rival A,Raz O,CohenR,MarxZ,CamaraM,DubernJF,Koch B, and Biotechnol. 2007;82:33–8. Heeb S, Krasnogor N, Delattre C, Shapiro E. Synthesis and cell-free cloning of DNA 43. Akama-Garren EH, Joshi NS, Tammela T, Chang GP, Wagner BL, Lee DY, libraries using programmable microfluidics. Nucleic Acids Res. 2016;44(4):e35. Rideout Iii WM, Papagiannakopoulos T, Xue W, Jacks T. A modular assembly 19. Song H, Chen DL, Ismagilov RF. Reactions in droplets in microfluidic platform for rapid generation of DNA constructs. Sci Rep. 2015; https://doi. channels. Angew Chem Int Ed. 2006;45(44):7336–56. org/10.1038/srep16836 3. 20. Czar MJ, Anderson JC, Bader JS, Peccoud J. Gene synthesis demystified. 44. Kosuri S, Eroshenko N, LeProust EM, Super M, Way J, Li JB, Church GM. Trends Biotechnol. 2009;27(2):63–72. Scalable gene synthesis by selective amplification of DNA pools from high- 21. Huang MC, Ye H, Kuan YK, Li M, Ying JY. Integrated two-step gene synthesis fidelity microchips. Nat Biotechnol. 2010;28(12):1295–9. in a microfluidic device. Lab Chip. 2009;9(2):276–85. 45. Saaem I, Ma S, Quan J, Tian J. Error correction of microchip synthesized 22. Kong DS, Carr PA, Chen L, Zhang S, Jacobson JM. Parallel gene synthesis in genes using surveyor nuclease. Nucleic Acids Res. 2012;40(3):e23. a microfluidic device. Nucleic Acids Res. 2007;35(8):e61. 46. Sequeira AF, Guerreiro CI, Vincentelli R, Fontes CM. T7 endonuclease I 23. Quan J, Saaem I, Tang N, Ma S, Negre N, Gong H, White KP, Tian J. Parallel mediates error correction in artificial gene synthesis. Mol Biotechnol. 2016; on-chip gene synthesis and application to optimization of protein 58(8-9):573–84. expression. Nat Biotechnol. 2011;29(5):449–52. 47. Baldursdottir SG, Fullerton MS, Nielsen SH, Jorgensen L. Adsorption of 24. Tian J, Gong H, Sheng N, Zhou X, Gulari E, Gao X, Church G. Accurate proteins at the oil/water interface-observation of protein adsorption by multiplex gene synthesis from programmable DNA microchips. Nature. interfacial shear stress measurements. Colloids Surf B Biointerfaces. 2010; 2004;432(7020):1050–4. 79(1):41–6. 25. Stemmer WPC, Crameri A, Ha KD, Brennan TM, Heyneker HL. Single-step 48. Linshiz G, Jensen E, Stawski N, Bi C, Elsbree N, Jiao H, Kim J, Mathies R, assembly of a gene and entire plasmid from large numbers of Keasling JD, Hillson NJ. End-to-end automated microfluidic platform for oligodeoxyribonucleotides. Gene. 1995;164(1):49–53. synthetic biology: from design to functional analysis. J Biol Eng. 2016;10(1):1. 26. Gibson DG, Smith HO, Hutchison CA III, Venter JC, Merryman C. Chemical 49. Shih SC, Goyal G, Kim PW, Koutsoubelis N, Keasling JD, Adams PD, Hillson synthesis of the mouse mitochondrial genome. Nat Meth. 2010;7(11):901–3. NJ, Singh AK. A versatile microfluidic device for automating synthetic 27. Yoon J, Garrell RL. Preventing biomolecular adsorption in electrowetting- biology. ACS Synth Biol. 2015;4(10):1151–64. based biofluidic chips. Anal Chem. 2003;75(19):5097–102. 50. Tangen U, GAS M, Sharma A, Wagler PF, Cohen R, Raz O, Marx T, Ben- 28. Dormitzer PR, Suphaphiphat P, Gibson DG, Wentworth DE, Stockwell TB, Algire Yehezkel T, JS MC. DNA-library assembly programmed by on-demand MA, Alperovich N, Barro M, Brown DM, Craig S, Dattilo BM, Denisova EA, De nano-liter droplets from a custom microfluidic chip. Biomicrofluidics. Souza I, Eickmann M, Dugan VG, Ferrari A, Gomila RC, Han L, Judge C, Mane S, 2015;9(4):044103. Matrosovich M, Merryman C, Palladino G, Palmer GA, Spencer T, Strecker T, Trusheim H, Uhlendorff J, Wen Y, Yee AC, Zaveri J, Zhou B, Becker S, Donabedian A, Mason PW, Glass JI, Rappuoli R, Venter JC. Synthetic generation of influenza vaccine viruses for rapid response to pandemics. Sci Transl Med. 2013; https://doi.org/10.1126/scitranslmed.3006368. 29. Prakash AR, Amrein M, Kaler KV. Characteristics and impact of Taq enzyme adsorption on surfaces in microfluidic devices. Microfluid Nanofluid. 2008; 4(4):295–305. 30. Erill I, Campoy S, Erill N, Barbé J, Aguiló J. Biochemical analysis and optimization of inhibition and adsorption phenomena in glass–silicon PCR- chips. Sensors Actuators B Chem. 2003;96(3):685–92. 31. Kleinert J, Srinivasan V, Rival A, Delattre C, Velev OD, Pamula VK. The dynamics and stability of lubricating oil films during droplet transport by electrowetting in microfluidic devices. Biomicrofluidics. 2015;9(3):034104. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png BMC Biotechnology Springer Journals

DNA assembly with error correction on a droplet digital microfluidics platform

Free
14 pages
Loading next page...
 
/lp/springer_journal/dna-assembly-with-error-correction-on-a-droplet-digital-microfluidics-Xj06LkD4Qs
Publisher
BioMed Central
Copyright
Copyright © 2018 by The Author(s).
Subject
Chemistry; Applied Microbiology; Biotechnology; Biochemical Engineering; Genetic Engineering; Plant Breeding/Biotechnology; Biomedical Engineering/Biotechnology
eISSN
1472-6750
D.O.I.
10.1186/s12896-018-0439-9
Publisher site
See Article on Publisher Site

Abstract

Background: Custom synthesized DNA is in high demand for synthetic biology applications. However, current technologies to produce these sequences using assembly from DNA oligonucleotides are costly and labor-intensive. The automation and reduced sample volumes afforded by microfluidic technologies could significantly decrease materials and labor costs associated with DNA synthesis. The purpose of this study was to develop a gene assembly protocol utilizing a digital microfluidic device. Toward this goal, we adapted bench-scale oligonucleotide assembly methods followed by enzymatic error correction to the Mondrian™ digital microfluidic platform. Results: We optimized Gibson assembly, polymerase chain reaction (PCR), and enzymatic error correction reactions in a single protocol to assemble 12 oligonucleotides into a 339-bp double- stranded DNA sequence encoding part of the human influenza virus hemagglutinin (HA) gene. The reactions were scaled down to 0.6-1.2 μL. Initial microfluidic assembly methods were successful and had an error frequency of approximately 4 errors/kb with errors originating from the original oligonucleotide synthesis. Relative to conventional benchtop procedures, PCR optimization required additional amounts of MgCl , Phusion polymerase, and PEG 8000 to achieve amplification of the assembly and error correction products. After one round of error correction, error frequency was reduced to an − 1 average of 1.8 errors kb . Conclusion: We demonstrated that DNA assembly from oligonucleotides and error correction could be completely automated on a digital microfluidic (DMF) platform. The results demonstrate that enzymatic reactions in droplets show a strong dependence on surface interactions, and successful on-chip implementation required supplementation with surfactants, molecular crowding agents, and an excess of enzyme. Enzymatic error correction of assembled fragments improved sequence fidelity by 2-fold, which was a significant improvement but somewhat lower than expected compared to bench-top assays, suggesting an additional capacity for optimization. Keywords: DNA, Gibson assembly, Digital microfluidics, Error correction Background Gene synthesis is a costly and labor-intensive process. Over the last decade, major research advances in gen- The cost of synthetic DNA is directly related to the cost ome sequencing (i.e. “DNA reading”) are slowly being of oligonucleotides, and a large amount of hands-on matched by advances in synthetic biology (i.e. “DNA labor required for conventional oligonucleotide-based writing”)[1, 2]. Rapid advances in synthetic biology are gene assembly is also a significant cost [6–8]. The fueling a demand for synthetic DNA that will only in- cheapest oligonucleotides that can be purchased from crease in the future. However, the ability to synthesize commercial suppliers are usually unpurified and contain long DNA molecules in a short period of time without errors. Thus, the genes assembled from the unpurified significant expense remains one of the main challenges oligonucleotides must be sequence verified to find a cor- in synthetic biology [3–5]. rect assembly. Implementation of an enzymatic error correction step greatly improves the sequence fidelity of the assemblies, which reduces the number of clones that * Correspondence: griffin@stanford.edu have to be individually cloned and sequence verified [7, Stanford Genome Technology Center, Stanford University, 3165 Porter Drive, 9, 10]. Unfortunately, this extra error correction step Palo Alto, CA 94304, USA also significantly increases the hands-on time required Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Khilko et al. BMC Biotechnology (2018) 18:37 Page 2 of 14 to complete assembly. Integration of digital microfluidics can be replaced and performed by programmable droplet into DNA assembly coupled with error correction can generation and routing over heater bars underneath the potentially ease this labor burden by allowing for a “set microfluidic cartridge. The sequential reactions can be up and walk away” approach to managing the entire performed on a single microfluidic cartridge without any process. human intervention [18]. Software automation programs Digital microfluidics (DMF) is a technology based on can be designed to conduct multiple experiments in paral- the electrowetting phenomenon. The phenomenon lel. Since the devices are fully automated, the sources of describes a change of surface tension at a solid/liquid/ human errors and labor costs can be greatly reduced. In gas interface by application of an electric field [11, 12]. addition, DNA assembly programs can be distributed be- The voltage applied to the electrodes lowers the surface tween laboratories, so that scientists can share robust tension, which leads to a reduction of the contact angle protocols. and increases the wettability of the surface. Conse- Because DNA assembly and error correction reactions quently, the liquid spreads over the surface where the require the use of expensive enzymes, scaling down to voltage is applied. Thus, a hydrophobic surface becomes smaller reaction volumes reduces reagent expenses. Due hydrophilic. By the application of a voltage on a dielec- to the large surface-to-volume ratio, micro-droplet reac- tric surface, the liquids can be transported over the tors have high heat and mass transfer rates. This makes surface of a microfluidic cartridge. it possible to increase kinetics and speed up reactions In electrowetting on dielectric (EWOD) devices, a [13, 17, 19]. Integration of effective error correction pro- droplet is sandwiched between two hydrophobic plates cedures could permit DNA assembly on a single micro- and the remaining volume is filled with immiscible li- fluidic cartridge without the need for expensive and quid, for example, a silicone oil (Fig. 1). The oil prevents lengthy sequence verification. evaporation of the aqueous droplets and facilitates trans- A number of DNA assembly protocols have been de- port. The bottom plate is the array of electrodes, which veloped to date (Table 1). For the scope of this study, can locally control the surface tension by the application only assembly methods from oligonucleotides will be of a voltage. Digital microfluidic devices are completely discussed. The most popular gene construction methods programmable and do not require any pumps or valves for microfluidic applications are polymerase-based and to move the liquids. A cartridge can be inserted into a endonuclease-based assembly. Both approaches utilize micro-controller that is operated by a software program oligonucleotides as DNA building blocks [20]. The [13–15]. The program turns the voltage on and off at polymerase-based assembly method utilizes the same ap- certain electrodes, so the droplets can be directed any- proach as PCR [21–24]), but instead of using forward where on a chip. They can also be dispensed, trans- and reverse primers, oligonucleotides overlap and serve ported, split, fused, mixed and held in certain regions. as templates for a complementary strand. The oligonu- Digital microfluidic devices are applicable for gene as- cleotides are designed to be either a part of the top or sembly because DNA is typically handled in microliter the bottom DNA strand. In the first assembly cycle, oli- amounts. The microfluidic devices are capable of gen- gonucleotides overlap partially, and the polymerase ex- erating droplets in the microliter to picoliter range tends the complementary strand in a 5′ to 3′ direction. [16, 17]. The microliter droplets act like reaction and In the next cycle, the double-stranded DNA pieces are transport vessels. The ability to program liquid hand- separated and hybridized to other oligonucleotides or as- ling operations such as dispense, transport, merge, sembled single-stranded fragments. The process of de- mix, and split allows a researcher to automate and naturation, annealing, and extension is repeated until simplify the gene assembly process. Time-consuming the desired sequence has been built [25]. steps like pipetting, transferring reagents, tube labeling, The studies shown in Table 1 utilized different types incubation at certain temperatures, and thermocycling of microfluidics to perform DNA assembly. Among Fig. 1 A cross-section of an EWOD cartridge Khilko et al. BMC Biotechnology (2018) 18:37 Page 3 of 14 Table 1 Summary of microfluidic assembly methods Authors Type of microfluidics Assembly method Assembly size Error rates Kong et al. [22] Droplet Polymerase-based (PCA) 500 –1000 bp 1.78 errors/kb Huang et al. [21] Microchannel Polymerase-based (PCA) 760 bp 4.1 errors/kb Quan et al. [23] Microarray Polymerase-based (PCA) 500 – 1000 bp 1.9 errors/kb Tian et al. [24] Microarray Polymerase-based (PAM) 14,500 bp 2.2 errors/kb Linshiz et al. [48] Chanel Exonuclease-based (Gibson) and 754 bp N/A polymerase-based (IHDC) Shih et al. [49] Combined digital and droplet microfluidics Exonuclease-based assembly of 2100 bp N/A dsDNA (Gibson) Tangen et al. [50] Droplets Polymerase-based (PCA) and 525 bp N/A Exonuclease-based (Gibson) Ben-Yehezkel et al. [18] Digital microfluidics Polymerase-based (POP) 800 bp 2.22 errors/kb those studies, only the work by Ben-Yehezkel et al. was multiple rounds of error correction after assembly and conducted using digital microfluidics, the same type of PCR to decrease the incidence of errors originating from microfluidics utilized in the present study. The group the oligonucleotides [28]. Gibson assembly has been developed an innovation on the polymerase-based assem- used successfully to assemble entire genes (1.5-1.7 kb) in bly method called programmable order polymerization a single step, and this method is arguably the most effi- (POP). The method was first successfully automated on cient to assemble genes from multiple oligonucleotides the Mondrian™ microfluidic device. The assembly reaction [28]. Using Gibson assembly, higher numbers of oligo- proceeded to assemble the sequence from the inside out. nucleotides can be assembled in a single reaction than In each of four phases, a double-stranded DNA (dsDNA) by PCR assembly. For this reason, we decided to imple- fragment was extended by a pair of oligonucleotides such ment Gibson assembly on a DMF device. that one oligonucleotide bound to each end of the se- To design a DNA assembly protocol for a programmable quence. Multiple rounds of thermocycling for each phase digital microfluidic device, we developed a process consist- with an oligonucleotide pair ensured that most of the ing of three major parts (Fig. 2). First, DNA oligonucleo- product was extended in each step. The group reported tides were assembled into a double-stranded DNA − 1 an error rate of 1 in 450 bp (2.22 errors kb ) for their fragment. Second, the assembly was amplified by PCR, and assembly method, and the errors were identified as third, errors from the original oligonucleotides were substitutions [18]. removed. We used Sanger DNA sequencing of the recov- Whereas Ben-Yehezkel used four separate rounds of ered, error-corrected products to verify the efficiency of the DNA polymerization to sequentially extend and assem- error correction process and develop an efficient DNA as- ble two DNA fragments in each round, we used a single sembly and error correction protocol. The ultimate goal is “one-pot” assembly of 12 fragments in a single droplet to design a reliable and cost-effective DNA assembly proto- and successfully assembled the complete sequence. In col that will be widely applicable in biological research. addition, we performed a round of error correction on the digital microfluidic device. At some point, a large Methods number of DNA fragments in a one-pot reaction would DNA sequences and oligonucleotides lead to miss-hybridization, so it is interesting to contem- Our 339-bp test assembly sequence comprised a partial plate a process that combines both these methods in a sequence from the human influenza virus H9N2 DNA assembly reaction to reduce the assembly time and hemagglutinin (HA) gene (283-bp, nt 211-493 of the HA errors even further. coding region) flanked on each side by 8-bp NotI restric- In summary, the one-pot assembly method used in tion endonuclease sites and 20-bp homology regions to this manuscript is a one-step isothermal Gibson assem- the pUC19 cloning vector. The 339-bp fragment was di- bly developed at the J. Craig Venter Institute ([26]) and vided into 12 overlapping oligonucleotides (See Table 2). is distinctly different in process from the work described The final assembled test sequence is listed in Fig. 3. by Ben-Yehezkel et al. With this technique, double- or single-stranded DNA pieces are joined into longer frag- Mondrian™ digital microfluidics (DMF) device ments by three enzymes: T5 exonuclease, DNA polymer- The main goal of this study was to develop a robust and ase, and Taq DNA ligase. The reagents are incubated at reliable DNA assembly and error correction protocol for 50 °C for 0.5-1 h after which the assembled product is the Mondrian™ DMF device (Illumina, Inc.). The whole typically amplified by PCR [27]. Published protocols use protocol involved four consecutive enzymatic reactions. Khilko et al. BMC Biotechnology (2018) 18:37 Page 4 of 14 Fig. 2 A schematic of influenza HA gene assembly on DMF. This diagram shows the steps of the process which were all performed sequentially on a microfluidic cartridge. The steps are Gibson assembly of 12 oligonucleotides, PCR amplification of a double-stranded DNA piece, error removal using an error correction enzyme, and PCR amplification of corrected sequences Prior to incorporating the four enzymatic steps of the device. To observe the behavior of the droplets, a digital gene assembly in a complete protocol, each enzymatic camera was mounted above the cartridge to produce a step was optimized individually. All liquid handling op- magnified image of the DMF cartridge onto a computer erations were programmed using the Application Devel- screen. The Mondrian™ cartridge that was used in the opment Environment software (Illumina, Inc.). experiments (Fig. 4a), consisted of two plates, a plastic The Mondrian™ microfluidic system included a micro- top plate, and a printed circuit board (PCB) substrate. controller that was connected to a computer and digital The area between the plates was filled with a 2 cSt sili- microfluidic cartridges that were inserted into the cone oil. As seen in Fig. 4b, the configuration of the Table 2 Primers used in this study Primer Name Primer Sequence (5′->3′) Puc-049cloning-R CCGGGTACCGAGCTCGAATTCACTG Puc-049cloning-F GATCCTCTAGAGTCGACCTGCAGGC Puc19-5’F TCCCAGTCACGACGTTGTAAAACGAC HA049-1 CAGGTCGACTCTAGAGGATCGCGGCCGCGACACATGCACTATTGAAGGACTT HA049-2 AGATCACAAGAAGGGTTACCATAGACAAGTCCTTCAATAGTGCATGTGTCGC HA049-3 GTCTATGGTAACCCTTCTTGTGATCTGTTGTTGGGGGGAAGAGAATGGTCCT HA049-4 TACAGCTGATGGTCTTTCAACGATGTAGGACCATTCTCTTCCCCCCAACAAC HA049-5 ACATCGTTGAAAGACCATCAGCTGTAAATGGAACGTGTTACCCTGGGAATGT HA049-6 GTGTTCTGAGTTCCTCTAAGTTTTCCACATTCCCAGGGTAACACGTTCCATT HA049-7 GGAAAACTTAGAGGAACTCAGAACACTCTTTAGTTCCTCTAGTTCCTACCAA HA049-8 ATTGTGTCTGGGAATATTTGGATTCTTTGGTAGGAACTAGAGGAACTAAAGA HA049-9 AGAATCCAAATATTCCCAGACACAATCTGGAATGTGACTTACACTGGAACAA HA049-10 GTAGAATGAATCTGAACATGATTTGCTTGTTCCAGTGTAAGTCACATTCCAG HA049-11 GCAAATCATGTTCAGATTCATTCTACAGGAACATGAGATGGCTGACTCAAAG HA049-12 GAATTCGAGCTCGGTACCCGGCGGCCGCTTTGAGTCAGCCATCTCATGTTCCT Khilko et al. BMC Biotechnology (2018) 18:37 Page 5 of 14 >HA049_sequence GAATTCGAGCTCGGTACCCGGCGGCCGCTTTGAGTCAGCCATCTCATGTTCCTGTAGAATGAATCT GAACATGATTTGCTTGTTCCAGTGTAAGTCACATTCCAGATTGTGTCTGGGAATATTTGGATTCTTT GGTAGGAACTAGAGGAACTAAAGAGTGTTCTGAGTTCCTCTAAGTTTTCCACATTCCCAGGGTAAC ACGTTCCATTTACAGCTGATGGTCTTTCAACGATGTAGGACCATTCTCTTCCCCCCAACAACAGAT CACAAGAAGGGTTACCATAGACAAGTCCTTCAATAGTGCATGTGTCGCGGCCGCGATCCTCTAGA GTCGACCTG Fig. 3 Oligonucleotide alignment for the 339-bp test assembly sequence. a Arrangement of DNA oligonucleotides used for assembly of the HA049 sequence. b FASTA formatted HA049 sequence DMF cartridge allowed eight processes to be performed used to set temperatures for the enzymatic reactions. in parallel. The reagents were loaded through 50 μLor Additionally, an area of the cartridge could be cooled 10 μL ports on the cartridge top plate, and the samples down with a Peltier device. Fig. 4c shows a close-up view were withdrawn through other ports. There were also of one lane with three different temperature zones, seven reservoirs dedicated for the collection of waste which were maintained during reactions using the droplets. The microfluidic cartridge had three heater heaters and the cooler. The device was operated by the bars that contacted the back of the PCB, which were Application Development Environment (ADE) software. Fig. 4 The Mondrian™ microfluidic cartridge. a Image of the cartridge. b Diagram of the cartridge electrode paths. This diagram of the chip comes from a screenshot of an ADE software. There are 50 μL reservoirs highlighted in blue, 10 μL reservoirs highlighted in red. Orange reservoirs were used to hold waste. Green reservoirs were used to collect final products. The configuration of the chip allowed to perform 8 reactions in parallel. c Magnified image of one lane of the microfluidic cartridge where the reactions were performed. The area highlighted in yellow was used for error correction reaction. The area highlighted in purple was used for Gibson assembly and PCR annealing/extension. The area highlighted in red was used for DNA denaturation during PCR and error correction pretreatment Khilko et al. BMC Biotechnology (2018) 18:37 Page 6 of 14 Prior to each experiment, a program was designed to An automation program for the DMF was developed direct droplets through the liquid handling operations. to perform microfluidic PCR amplification experiments. The device was operated at voltages between 90 V and The reactions were carried out in 1.2 μL droplets. The 300 V and at a frequency of 30 Hz. droplets were brought to the PCR area of the chip, The liquid volumes of 0.3, 0.6, and 1.2 μL were gener- which consisted of two temperature zones. The denatur- ated and manipulated on the microfluidic cartridge. To ation zone was set to 98 °C, and the annealing/extension dispense a 0.3 μL or 0.6 μL droplet, three electrodes zone was set to 72 °C. adjacent to the reagent input port were activated, which The thermocycling was performed at a reduced voltage caused the liquid to spread over three electrodes (Fig. (90 V), which eliminated nonspecific adsorption of poly- 5a). The electrode #2 was switched off to obtain a 0.3 μL merase on the microfluidic surface and reduced spurious droplet (Fig. 5b). The double 0.6 μL droplet was dis- bubble formation at high temperatures [18, 27, 29–32]. pensed by turning off the electrode #3 (Fig. 5c). To cre- An initial denaturation was performed by moving the ate a 1.2 μL droplet, two 0.6 μL droplets are brought droplets to the 98 °C zone where they were held for next to each other and separated by one inactive elec- 30 s. Then thirty PCR cycles were performed by cycling trode as shown in Fig. 5d. Then the electrode between the droplets from 98 °C to 72 °C at 1.5 s/electrode, and the two 0.6 μL droplets was turned on, merging both from 72 °C to 98 °C at 1 s/electrode. Annealing/exten- into one 1.2 μLdroplet (Fig. 5e). Refer to Additional sion of the droplets was done by switching on/off volt- file 1: Video 1 to see the liquid handling operation de- age of the area of three electrodes every half of a second scribed here. Materials to see all liquid handling operations for 20 s, and denaturation was performed by holding the used in present work. droplets at 98 °C for 10 s. After 30 cycles of PCR, DNA was held for 10 min at 72 °C to allow a final extension. Then, the voltage was switched back to 300 V, so that Optimization of microfluidic PCR the samples could be transported to the collection Each on-chip PCR reaction contained 1X HF Phusion reservoirs. buffer, detergent-free (Thermo Fisher Scientific), 0.8 μM forward and reverse primers, 1.75 ng/μL HA-049 DNA template (plasmid-cloned HA-049 sequence), 0.1 U μL Optimization of microfluidic Gibson assembly Phusion polymerase (Thermo Fisher Scientific). The In each experiment, 50 μL master mixes were made reactions were set up to allow the addition of 1 mM from fresh reagents and prepared according to previ- MgCl , 1.25 mM PEG 8000, 0.2 mM NAD, 2 mM DTT ously published Gibson Assembly protocols [33] with to the reaction mixture or the combination of 1.25 mM additional modifications described below. Assembly, PEG 8000 and 1 mM MgCl , 0.2 mM NAD, and 2 mM oligonucleotide, and PCR master mixes were prepared at DTT. Final concentrations in the reaction droplets are 2X concentrations such that once two equal size given for all reagents. droplets were merged, the final enzyme mixtures would Fig. 5 Generation of 0.3, 0.6, and 1.2 μL droplets on DMF. a Stretching the liquid over three electrodes. b Generation of 0.3 μLdroplet. c Generation of 0.6 μLdroplet. d Two 0.6 μL droplets separated by 1 electrode. e 1.2 μL droplet resulted from merging two 0.6 μL droplets Khilko et al. BMC Biotechnology (2018) 18:37 Page 7 of 14 be at the correct 1X concentration. The oligo master original size (339-bp) and the cleaved size (approxi- mix containing a mixture of all oligonucleotides was pre- mately 170-bp). In case of a failure, only one 339-bp pared by diluting a 1 μM stock solution in DI water con- band was visible. Preliminary experiments demonstrated taining 0.01% Tween 20. The surfactant was a necessary that a microfluidic error correction reaction with stand- component to reduce the surface tension, which facili- ard benchtop reagents was not successful due microflui- tated droplet dispensing and movement. The amount of dic surface interactions and nonspecific adsorption of surfactant required was determined for each master mix. CorrectASE™. To test the hypothesis of CorrectASE™ ad- The enzymes suspended in storage buffers contained sorption on the droplet oil/water interface, the mix of stabilizers. It was observed that the droplets with en- DNA was treated with additional reagents. The reactions zyme solutions were easily dispensed and manipulated were performed with extra CorrectASE™, 0.01% Tween on a cartridge without any additional surfactant. Thus, 20, 1.25 mM PEG 8000, and 2.5 mM MgCl to deter- the assembly master mix and PCR master mix did not mine which could improve reaction performance. contain Tween 20. The final (1X) concentrations of re- agents in the assembly reaction were 1X isothermal (iso) Protocol for DNA assembly with error correction − 1 − 1 buffer, 0.05 U μL of Phusion polymerase, 4 U μL The protocol consisted of four consecutive enzymatic re- − 1 DNA ligase, 0.08 U μL T5 exonuclease, and 250 nM actions. The process started with Gibson assembly that oligonucleotides. was carried out for 60 min. Then, the assembly products To perform DNA assembly experiments, an automa- were amplified in 30 PCR cycles. Next, DNA was treated tion program was created. The temperature in the as- with CorrectASE™ for 60 min. The error correction sembly area was set to 50 °C. Next, 0.3 μL droplets products were amplified in a second PCR. According to containing oligonucleotides were dispensed. The drop- this protocol, the final concentrations of reagents in Gib- lets were transported to a waiting area where they were son assembly reaction were 1X isothermal (iso) buffer, 0. − 1 held while another dispenser generated 2X Gibson mas- 05 U μL of Phusion polymerase (Thermo Fisher − 1 − 1 ter mix droplets. The oligonucleotide and Gibson master Scientific), 4 U μL DNA ligase (NEB), 0.08 U μL T5 mix droplets were merged to obtain a double 0.6 μL exonuclease (NEB), and 50 nM oligonucleotides (IDT droplet and brought to the assembly area where they DNA). After assembly, the product was diluted with 0. were incubated for 15-60 min at 50 °C. To ensure ad- 01% Tween 20 (Sigma Aldrich) by 8-fold. Diluted assem- equate mixing, the droplets were moved up and down blies were merged with equal size droplets of PCR mas- − 1 over 4 electrodes. When the reaction was finished, the ter mixes to achieve 0.1 U μL Phusion polymerase assembly droplets were merged with 0.6 μL PCR drop- (Thermo Fisher Scientific), 1X HF detergent-free buffer lets, so the total volume of each droplet became 1.2 μL. (Thermo Fisher Scientific), 0.25 mM of each dNTP The polymerase chain reaction was performed as de- (Thermo Fisher Scientific), 0.8 μM of forward and re- scribed above. After PCR, the products were diluted. To verse primers (IDT DNA, 0.625 mM PEG 8000 (Sigma), perform dilutions, a dispenser containing DI water and 0.5 mM MgCl (Thermo Fisher Scientific) in the reac- 0.05% Tween 20 generated 0.6 μL droplets. Then, the tions. After amplification, two of the eight droplets were droplets were merged with the assembly droplets, mixed, recovered from the chip, and the rest of the droplets and split into two equal size droplets. This step was iter- were diluted by 2-fold with 0.01% Tween 20 solution to ated to achieve the desired dilutions. When assembly continue to the error correction step. time was variable, 0.6 μL droplets containing both the The EC denature/anneal step of the protocol was imple- oligonucleotides and Gibson assembly reagents were mented to expose the errors in DNA sequence for further held in a waiting area, and two droplets were moved to CorrectASE™ treatment. During the denaturation step, − 1 the assembly incubation zone in 15 min increments. DNA was diluted to 20-25 ng μL in 1X CorrectASE™ This way, each condition was tested twice in two differ- buffer and incubated at 98 °C for 2 min, 25 °C for 5 min, ent experimental droplets. and 37 °C for 5 min. Then, the droplets were merged with CorrectASE™ master mix to a final concentration of 2X Optimization of enzymatic error correction CorrectASE™ (Invitrogen), 0.01% Tween 20 (Sigma The optimization of enzymatic error correction step was Aldrich), and 2.5 mM PEG 8000 (Sigma). The master performed using a mix of two equal molar amounts of mixes contained double amounts of reagents to obtain 1X 339-bp PCR products. Sequences were amplified from concentration after the equal size droplets were merged. two DNA templates. The first template had a completely The reagents were loaded on a DMF cartridge into correct sequence and the other had a mutation approxi- dedicated dispensers as prescribed by the automation mately in the middle of the 339-bp sequence. If the error program. All master mixes except CorrectASE™ were correction reaction was successful, two DNA bands were loaded on the cartridge at the beginning of the process. visualized on an agarose gel corresponding to the To ensure that the enzyme stayed active, CorrectASE™ Khilko et al. BMC Biotechnology (2018) 18:37 Page 8 of 14 was loaded into the dispenser three minutes before it was x  1000 f ¼ ð1Þ to be used by the program. At the end of the process, all n  l droplets were collected in 20 μL water containing 0.05% Tween 20 and retrieved from the device manually. where x is the number of errors in a single clone, n is the number of sequenced clones not including clones Cloning and sequencing of assembled and amplified with mis-assemblies, and l is the length of a sequence in products bases. Recovered products were brought up to 50 μLin water and an equal volume of Agencourt AMPure XP Results beads (Beckman Coulter) was added and mixed. After Optimization of microfluidic PCR 5 min incubation to bind DNA to the beads, the tube Optimization of PCR on DMF demonstrated that the was placed on a magnet and allowed to settle for additives improved amplification efficiency. Control 5 min. The supernatant was removed and the beads − 1 samples, which contained 0.1 U μL of Phusion were washed two times with 80% ethanol. After a polymerase, did not show any bands on the agarose gel final 5 min incubation with the caps open to allow (data not shown). On the other hand, PCR that the beads to dry, the DNA was eluted in 15 μL contained the iso buffer used for DNA assembly resulted 10 mM TRIS buffer (pH 8.5). in the desired 339-bp bands. To determine the compo- The purified products were assembled into a pUC19 nents of the iso buffer that contributed to the successful vector that was amplified by primers Puc-049cloning-R PCR reaction, we tested each component individually + Puc-049cloning-L (Table 2) using OneTaq polymerase and in combination. When PEG 8000, DTT, NAD, and (NEB). Assembly of the product into the pUC19 vector MgCl were added separately to the reaction, only PEG was by Gibson assembly [7, 8, 33], assembly reactions 8000 demonstrated some amplification of DNA tem- were electroporated into E. coli strain Epi300 (Epicentre) plate, but the result was not as good as PCR supple- , and resulting clones were selected on LB plates con- mented with the iso buffer (data not shown). Based on − 1 taining ampicillin at 100 μgmL . Colonies were these results PEG 8000 was combined with either NAD, screened using 20 μL colony PCR reactions with primers DTT, or MgCl to find out if DNA would be amplified pUC19-5’F and pUC19-3’R (OneTaq, NEB). Colonies at the same level as with the iso buffer. As seen in Fig. 6, containing a plasmid with an insert sequence of 339- the combination of 1.25 mM PEG 8000 and 1 mM bp were grown overnight in 5 mL LB broth and puri- MgCl showed comparable band intensity as the iso buf- fied using the QIAprep miniprep kit (Qiagen). The fer. This result demonstrated that microfluidic PCR insert sequences of the resulting plasmids were ana- must be performed with an excess of Phusion enzyme lyzed by Sanger DNA sequencing. For each treatment, and supplemented with additional MgCl and PEG 8000. 10-20 independent clones were sequenced using the Multiple methods to reduce biofouling in microfluidics pUC19-5’Fprimer. have been tested in this work. The only method that im- proved PCR yield and transport of droplets was the re- Data analysis duction of the electrowetting voltage from 300 V to Samples recovered from Lanes 1 and 2, 3-8, which cor- 90 V during high-temperature PCR [18]. For more responded to assembly-only and EC treatments, respect- ively, were pooled and analyzed by DNA gel electrophoresis on a 2% agarose gel, and using the 1 Kb plus DNA ladder (Invitrogen) as a size standard. The indication of a successful experimental run was the presence of a 339-bp band. For the in-depth error ana- lysis, the samples were cloned into pUC19 vectors and Sanger sequenced. Sequencing data was analyzed by aligning sequencing output files with the template DNA (Additional file 2). Each sequence alignment was inspected for errors in the newly assembled sequence. The errors were categorized in three groups: deletions, insertions, and substitutions. The sequences that had Fig. 6 On-chip polymerase chain reaction performed with two components of the iso buffer as described in the text. DNA from misincorporated oligonucleotides were treated as “mis- the reactions was separated by agarose gel electrophoresis on a assemblies”. The error frequency per 1 kb (f) was calcu- 2% agarose gel lated using Eq. 1 [34]. Khilko et al. BMC Biotechnology (2018) 18:37 Page 9 of 14 information refer to the Additional file 3: Video 2 in- improve the performance of CorrectASE™ enzyme. The cluded in Additional files. reactions were supplemented with either PEG 8000, Tween 20, and excess CorrectASE™ or combination of Optimization of microfluidic Gibson assembly the additives. Our test for CorrectASE™ activity was to The first group of experiments tested optimum reaction add a mixture of two PCR products with one containing time for oligonucleotide assembly. When we tested reac- a nucleotide mismatch in the sequence relative to the tion times spanning 15-60 min, the bands for all tested other at approximately the midpoint of the sequence. times were of similar brightness (Fig. 7), suggesting the Thus, successful error correction led to cleavage of the oligonucleotides were assembled in the 15-60 min time full-length product and resulted in two bands (339-bp period. and 170-bp) on the agarose gel with comparable inten- Figure 8 demonstrates the results of microfluidic as- sities. As seen in Fig. 10, the presence of 0.01% Tween sembly experiments with a dilution of assembly con- 20, 2X CorrectASE™ and 1.25 mM of PEG 8000 in the structs prior to PCR. Dilution of the assembly constructs reaction droplet gave the most even brightness in both from 2-fold to 16-fold resulted in comparable amounts bands on the agarose gel. of the PCR product. However, 16-fold was the maximum dilution rate that could be achieved before the PCR tem- Validation of DMF protocol for assembly with error plate was too dilute to amplify. A dilution rate greater correction than 32-fold did not result in amplification of the assem- The error analysis of DNA assembly of 12 oligonucleo- bly product. tides followed by CorrectASE™ treatment is shown in To investigate if the concentration of oligonucleotides Fig. 11 and Additional file 2. The average error fre- in the assembly reaction could influence the fidelity of as- quency of assembly samples from three separate runs − 1 sembly constructs, two sets of samples obtained by the as- was found to be about 4 errors kb , which corresponds sembly of 50 nM or 250 nM oligonucleotides were to what is widely reported for phosphoramidite DNA sequenced. The average error rate from five separate runs synthesis chemistry, where error rates of approximately for each oligonucleotides concentration is shown in Fig. 9. 1:200 are typical. The average error frequency of the − 1 It was determined that the average error rate for 250 nM samples after error correction was about 2 errors kb , and 50 nM oligonucleotides was similar, at 3.15 errors which corresponds to an average error reduction by 2- − 1 − 1 kb and 2.94 errors kb , respectively. fold. The average error reduction using a conventional Single-base deletions comprised the bulk of errors. benchtop protocol with the same sequence was found to There was no preference for errors to occur between A/ be about 10-fold (data not shown). T or C/G bases. Both sets of samples had comparable As seen in Table 3, the enzyme was effective in remov- percentages of deletions and the same amount of ing deletion and insertion errors but failed to remove multiple-base deletions. The assembly of 50 nM oligonu- substitutions. Overall, error correction was successful, cleotides resulted in 80% deletion errors, whereas and in all three experiments about half of the sequenced 250 nM assembly had 83% deletions, with the remainder clones were found to be error-free. being multiple base deletions, insertions or substitutions. Discussion Optimization of enzymatic error correction Optimization of microfluidic PCR An optimization of the error correction reaction was The results of the microfluidic PCR experiments demon- conducted to determine the reagents that reduce adsorp- strated that due to the high surface-to-volume ratio, re- tion of CorrectASE™ on the oil/water interface and actions performed on the microfluidic device show a Fig. 7 On-chip oligonucleotide assembly and PCR amplifications testing different assembly reaction durations. DNA from the reactions was separated by agarose gel electrophoresis on a 2% agarose gel Khilko et al. BMC Biotechnology (2018) 18:37 Page 10 of 14 Fig. 8 Dilution of the oligonucleotide assembly product (assembled with 50 nM each oligo) prior to PCR. a 2 to 16-fold and b 32 to 128-fold. After dilution and PCR, products were separated by agarose gel electrophoresis on 2% agarose gels strong dependence on surface interactions. Protein must be increased up to 10-fold in microfluidic droplets molecules can adsorb at the oil/water interface, which [36, 37]. The results of PCR experiments presented here reduces the surface tension over time [27, 35]. Addition- demonstrated that sufficient and repeatable PCR amplifi- ally, adsorption of a protein at the droplet aqueous/oil cation could be achieved with a 5-fold increase of interface could facilitate segregation of hydrophobic Phusion polymerase. groups that may lead to change in protein conformation The efficiency and specificity of PCR are affected by 2+ and inactivation. At high temperatures, the exposed the Mg concentration. Magnesium ions help the hydrophobic groups of the protein could lead to protein polymerase to fold in the active conformation [38]. Also, 2+ denaturation. The combined effect of protein adsorption Mg stabilizes dsDNA and increases the melting and denaturation could reduce the amount of available temperature (T ) of primers. Thus, it is crucial to have enzyme and decrease reaction efficiency. It has been re- the correct amount of free magnesium, and this ported previously that to achieve amplification efficiency concentration must often be optimized for each primer similar to a benchtop PCR, the amount of polymerase pair. It has been observed that the concentration of free Fig. 9 Average error frequency for sequences assembled from 250 nM and 50 nM oligonucleotides. Average error frequency from five independent experiments is plotted with error bars indicating one standard deviation from the mean Khilko et al. BMC Biotechnology (2018) 18:37 Page 11 of 14 Fig. 10 CorrectASE™ optimization on the DMF platform. DNA from the reactions was separated by agarose gel electrophoresis on a 2% agarose gel 2+ Mg can be decreased because of precipitation on Also, the polymers preserve native protein conformation microfluidic surfaces, capture by chelating agents and facilitate binding to a substrate. It has been shown present in reagents and storage buffers, and by binding that PEG 8000 stabilized Taq polymerase at high tem- to dNTPs [37]. According to the Phusion polymerase peratures [39, 41]. Since Phusion is a polymerase, it is product literature (Thermo), the optimum concentration possible that PEG 8000 formed weak bonds with the en- of MgCl is between 0.5-1 mM. The experimental zyme and reduced hydrophobic interactions with the results demonstrated that the addition of 0.5-1 mM of Teflon coating. As a result, the activity of the enzyme magnesium to the 1.5 mM MgCl present in Phusion was increased and amplification yield was improved HF buffer, improved polymerase activity, but this effect [42]. Consequently, microfluidic PCR is affected by ad- was inconsistent from lane to lane. However, it was sorption as well as by interactions of reaction compo- shown that the synergistic effect of magnesium and nents with interfaces. In order to achieve amplification PEG 8000 created favorable conditions for PCR on the DMF, the reaction must be carried out with the − 1 amplification (Fig. 6). final concentration of 0.1 U μL of Phusion (a 5-fold Polyethylene glycol (PEG) is recognized as a molecular increase relative to standard benchtop conditions), 0.5- crowding agent and frequently used as a PCR enhancer 1 mM of MgCl , and 0.625-1.25 mM of PEG 8000. and an enzyme immobilization agent [39, 40]. Molecular Improved PCR yield at 90 V showed that at the lower crowding creates the conditions similar to a natural cell voltage, the oil film between the aqueous droplet and environment in which the enzyme was evolved. It was the Teflon coated surface stayed intact and eliminated reported that macromolecular crowding affects the en- hydrophobic interactions between the polymerase and zyme reaction kinetics by increasing the viscosity of a the surface. According to Kleinert et al., the actuation medium that in turn influences the diffusion of reagents. voltage has a significant influence on the oil film [31]. At Fig. 11 Average error frequency of assembly samples followed by CorrectASE™ treatment. The average error rate of three independent experiments is plotted with error bars indicating one standard deviation from the mean Khilko et al. BMC Biotechnology (2018) 18:37 Page 12 of 14 Table 3 Error analysis of DNA obtained using DMF protocol Run 1 Run 2 Run 3 Assembly EC Assembly EC Assembly EC Deletions 10 4 9 4 10 4 Insertions 0 0 2 0 1 0 Substitutions 3 3 1 0 0 1 Total clones sequenced 8 14 10 10 9 7 Total of clones with correct sequences 0 7 1 5 0 2 − 1 high actuation voltage when the droplet moves, the film Kosuri et al. reported 4 errors kb , and Yehezkel et al. − 1 becomes unstable, breaks down, and tiny oil droplets get reported 2.2 errors kb [18, 44–46]. The analysis of trapped under the aqueous phase. error types demonstrated that the majority of errors In addition, excess surfactant destabilizes the oil film. belonged to single-base deletions with a small percent- Mohajeri and colleagues demonstrated that the critical age of insertions and substitutions. These results are micellar concentration of nonionic surfactants such as comparable to 75.6% deletions, 2.2% insertions, and 22. Tween 20 decreases at higher temperatures [32]. Thus, 2% substitutions, obtained by Sequeira et al. [46]. How- in the denaturation zone, less surfactant is necessary to ever, several clones in both data sets had mis- reduce the surface tension. If there is an excessive incorporated oligonucleotides. This issue could be solved amount of surfactant, the oil film becomes unstable, and by improving the design of the overlapping oligonucleo- the adsorption of the protein occurs, which is further tide sequences. Since the 50 nM oligonucleotide dataset enhanced at high temperatures. It is important to use had 1.5 times more clones with mis-assemblies, degrad- lower voltage and minimize the amount of Tween 20 to ation of some oligonucleotides by T5 exonuclease could avoid loss of Phusion polymerase and subsequent drop- be the cause of mis-incorporation. The results demon- let transport failure. strated that the Gibson assembly method performed on the DMF device is efficient. The error frequencies for Optimization of microfluidic Gibson assembly microfluidic synthesized sequences are in line with those Microfluidic DNA assembly protocols developed in this found for benchtop DNA synthesis in the published work produce results similar to the results published in literature. the literature. Our results show that even 15 min was an acceptable length of time for an efficient microfluidic Optimization of enzymatic error correction DNA assembly. In bench-top reactions, DNA assembly Since the best CorrectASE™ activity was obtained with 0. reactions proceed in reaction times between 15 and 01% Tween, 2X CorrectASE™, and 1.25 mM PEG addi- 60 min [26, 33, 43]. tives, the adsorption of the enzyme on the oil/water Dilution of the assembly product prior to the PCR interface of the aqueous droplet is the most likely ex- amplification is an additional step that should be in- planation of error correction in previous runs. Accord- cluded in a microfluidic Gibson assembly protocol. Since ing to Baldursdottir et al., protein molecules tend to the goal was to assemble the product that had the mini- aggregate on the oil/water interface in a multilayer. The mum number of errors, it was important to remove adsorption rate is affected by the molecular weight and a unreacted oligonucleotides, oligonucleotide fragments, saturation concentration. Large protein molecules tend and mis-assemblies that were present at a low level be- to adsorb faster than small ones due to the large surface fore amplification. Based on these results, we kept the area available for contact with the interface. Also, hydro- dilution of the assembly product no greater than 16-fold. phobic proteins tend to adsorb more due to interactions If the dilution step before PCR is employed, the amplifi- with the hydrophobic coated surface [47]. If some of the − 1 cation mix must contain 0.1 U μL of Phusion, 0. protein molecules adsorb on the interface, hydrophobic 625 mM PEG 8000, and 0.5 mM MgCl . and hydrophilic groups will rearrange, and it will cause The results of the error analysis suggest that the con- the protein to change conformation, leading to a loss of centration of oligonucleotides during assembly did not activity, and the reaction will not proceed with the max- affect the fidelity of the resulting sequence. Both DNA imum yield. assembly methods demonstrated an error frequency in We previously showed for PCR reactions that the − 1 the 1-10 errors kb range, which was similar to the presence of a molecular crowding agent such as PEG values reported in the literature for microfluidic DNA significantly increased the activity of Phusion polymer- assembly [9]. For instance, Saem et al. reported 1.9 ase. According to Sasaki et al., the activity of DNase I to − 1 − 1 errors kb , Sequeira et al. reported 3.45 errors kb , degrade supercoiled DNA and linear DNA was improved Khilko et al. BMC Biotechnology (2018) 18:37 Page 13 of 14 in the presence of 20% w/v PEG [40]. A kinetic analysis Additional files demonstrated that the rate of the DNA cleavage reaction Additional file 1: Video demonstration of droplet liquid handling increased with the increasing of concentration of PEG. operations used in the study. (MP4 9178 kb) However, molecular crowding did not improve the Additional file 2: Sequencing reads aligned to a template. (PDF 1973 kb) activity of Exonuclease III and inhibited the activity Additional file 3: Video demonstration of droplet movement during of Exonuclease I [40]. Consequently, macromolecular 300 V and 90 V PCR cycle. (MP4 19348 kb) crowding could be the reason why the error correc- tion reaction was improved on the DMF platform Abbreviations with the addition of PEG. DMF: Digital microfluidics; DTT: Dithiothreitol; EC: Error correction; NAD: Nicotinamide adenine dinucleotide; PCR: Polymerase chain reaction; Surfactants in digital microfluidic electrowetting on PEG: Polyethylene glycol dielectric (EWOD) are very important. The excess of surfactant can lead to a destruction of the oil film under Acknowledgements the droplet, which could cause the adsorption of hydro- We gratefully acknowledge the supply of electrowetting instruments and cartridges from Illumina, Inc. under a joint project agreement with Stanford phobic molecules on the microfluidic surface. Insuffi- University. cient surfactant could also cause interface instability that in turn could cause the adsorption of enzymes on the Funding The study was funded by NIH grant 5R21GM104694-03. oil/water interface. Usually, to be able to generate and manipulate the droplets on DMF the concentration of Availability of data and materials Tween 20 has to be 0.01-0.05% [18, 31]. However, en- Small-scale Sanger sequencing data is archived with the researchers and will zymatic reactions contain multiple components which be made available upon request. can potentially affect surface tension. Thus, the amount Authors’ contributions of Tween 20 has to be optimized for individual reac- YK, PBG, PDW, JIG, MDA designed experiments. YK, PBG, PDW conducted tions. It has been demonstrated in this study that even research and performed experiments. YK,PBG,PDW,JIG, MDA,MAM analyzed data and wrote the manuscript. All authors read and approved the presence of 0.001% of Tween 20 in reaction droplets the final manuscript. in conjunction with the excess of CorrectASE™ and PEG 8000 gives reproducible error correction results. Ethics approval and consent to participate Not applicable. Validation of DMF protocol for assembly with error Competing interests correction The authors declare that they have no competing interests. The results of our microfluidic protocol demonstrated that some inhibition of CorrectASE™ was still occurring Publisher’sNote during error correction on the DMF platform. Lower Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. error reduction could also be related to over dilution of error correction products prior to PCR or errors in Author details amplification. This suggests that further optimization is Stanford Genome Technology Center, Stanford University, 3165 Porter Drive, Palo Alto, CA 94304, USA. Department of Biomedical, Chemical and possible on the DMF device. Materials Engineering, San Jose State University, 1 Washington Sq, San Jose, CA 95192, USA. J. Craig Venter Institute, 4120 Capricorn Lane, La Jolla, CA Conclusion 92037, USA. Oligonucleotide assembly and error correction protocols Received: 31 August 2017 Accepted: 24 April 2018 for the Mondrian™ digital microfluidic device were developed. The process involved automation of the poly- merase chain reaction, Gibson assembly of 12 oligonu- References 1. Hutchison CA 3rd, Chuang RY, Noskov VN, Assad-Garcia N, Deerinck TJ, cleotides, and enzymatic error correction reaction with Ellisman MH, Gill J, Kannan K, Karas BJ, Ma L, Pelletier JF, Qi ZQ, Richter RA, CorrectASE™. The final protocol consisted of the assem- Strychalski EA, Sun L, Suzuki Y, Tsvetanova B, Wise KS, Smith HO, Glass JI, bly of oligonucleotides, two PCR steps, and an error cor- Merryman C, Gibson DG, Venter JC. Design and synthesis of a minimal bacterial genome. Science. 2016;351:6280. rection reaction. To achieve PCR amplification on the 2. Shendure J, Ji H. Next-generation DNA sequencing. Nat Biotech. 2008; DMF platform, the reactions were supplemented with 26(10):1135–45. PEG, MgCl , and 5-fold increased amount of polymerase 3. Hughes AR, Miklos, EA, Ellington AD. Gene Synthesis: Methods and Applications, Methods in Enzymology. 2011;498:277–309 (relative to benchtop conditions). The error correction 4. Chris Voigt. Synthetic Biology, Part B; Computer Aided Design and DNA reaction was supplemented with PEG, Tween 20, and an Assembly. Elsevier Press. 2011: p. 550. excess of CorrectASE™ (2-fold increase relative to bench- 5. Kosuri S, Church GM. Large-scale de novo DNA synthesis: technologies and applications. Nat Meth. 2014;11(5):499–507. top conditions). The final protocol assembled DNA se- 6. Gibson DG: Gene and genome construction in yeast. In Curr Protoc Mol − 1 quences with an average of 4 errors kb and reduced Biol. Volume Chapter 3. Edited by Ausube FM. 2011. DOI: https://doi.org/10. errors after error correction by 2-fold. 1002/0471142727.mb0322s94. Khilko et al. BMC Biotechnology (2018) 18:37 Page 14 of 14 7. Gibson DG, Lei Y, Chuang R-Y, Venter JC, Hutchison CA, Smith HO, 32. Mohajeri E, Noudeh GD. Effect of temperature on the critical micelle Enzymatic assembly of DNA molecules to several hundred kilobases. Nat concentration and Micellization thermodynamic of nonionic surfactants: Meth. 2009;6:343–5. Polyoxyethylene Sorbitan fatty acid esters. E-Journal of Chem. 2012;9(4): 8. Gibson DG, Glass JI, Lartigue C, Noskov VN, Chuang R-Y, Algire MA, Benders 2268–74. GA, Montague MG, Ma L, Moodie MM, Merryman C, Vashee S, Krishnakumar 33. Gibson DG, Caiazza N, Richardson TH. Materials and methods for the synthesis R, Assad-Garcia N, Andrews-Pfannkoch C, Denisova EA, Young L, Qi Z-Q, of error-minimized nucleic acid molecules US 20130225451 A1. 2013. https:// Segall-Shapiro T, Calvey CH, Parmar PP, Hutchison CA, Smith HO, Venter JC. www.google.com/patents/US20130225451. Accessed 25 Jun 2017. Creation of a bacterial cell controlled by a chemically synthesized genome. 34. Fuhrmann M, Oertel W, Berthold P, Hegemann P. Removal of mismatched Science. 2010;329:52. bases from synthetic genes by enzymatic mismatch cleavage. Nucleic Acids Res. 2005;33(6):e58. 9. Ma S, Saaem I, Tian J. Error correction in gene synthesis technology. Trends 35. Beverung CJ, Radke CJ, Blanch HW. Protein adsorption at the oil/water Biotechnol. 2011;30(3):147–54. interface: characterization of adsorption kinetics by dynamic interfacial 10. Ma S, Tang N, Tian J. DNA synthesis, assembly and applications in synthetic tension measurements. Biophys Chem. 1999;81(1):59–80. biology. Curr Opin Chem Biol. 2012;16(3):260–7. 36. Krishnan M, Burke DT, Burns MA. Polymerase chain reaction in high surface- 11. Lee J, Moon H, Fowler J, Schoellhammer T, Kim C. Electrowetting and to-volume ratio SiO2 microstructures. Anal Chem. 2004;76(22):6588–93. electrowetting-on-dielectric for microscale liquid handling. Sensors 37. Wang F, Burns MA. Performance of nanoliter-sized droplet-based Actuators A Phys. 2002;95(2):259–68. microfluidic PCR. Biomed Microdevices. 2009;11(5):1071–80. 12. Pollack MG, Fair RB, Shenderov AD. Electrowetting-based actuation of liquid 38. Patel PH, Loeb LA. Getting a grip on how DNA polymerases function. Nat droplets for microfluidic applications. Appl Phys Lett. 2000;77(11):1725–6. Struct Mol Biol. 2001;8(8):656–9. 13. Teh S, Lin R, Hung L, Lee AP. Droplet microfluidics. Lab Chip. 2008;8(2):198–220. 39. Zimmerman SB, Harrison B. Macromolecular crowding increases binding of 14. Baroud CN, Gallaire F, Dangla R. Dynamics of microfluidic droplets. Lab DNA polymerase to DNA: an adaptive effect. Proc Natl Acad Sci U S A. 1987; Chip. 2010;10(16):2032–45. 84(7):1871–5. 15. Song JH, Evans R, Lin Y, Hsu B, Fair RB. A scaling model for electrowetting- 40. Sasaki Y, Miyoshi D, Sugimoto N. Effect of molecular crowding on DNA on-dielectric microfluidic actuators. Microfluid Nanofluid. 2009;7(1):75–89. polymerase activity. Biotechnol J. 2006;1(4):440–6. 16. Vergauwe N, Witters D, Ceyssens F, Vermeir S, Verbruggen B, Puers R, 41. Saiki RK, Scharf S, Faloona F, Mullis KB, Horn GT, Erlich HA, Arnheim N. Lammertyn J. A versatile electrowetting-based digital microfluidic platform Enzymatic amplification of beta-globin genomic sequences and restriction for quantitative homogeneous and heterogeneous bio-assays. J Micromech site analysis for diagnosis of sickle cell anemia. Science. 1985;230(4732): Microengineering. 2011;21(5):054026. 1350–4. 17. Jebrail MJ, Fau BM, Patel KD. Digital microfluidics: a versatile tool for 42. Xia YM, Hua ZS, Gular E, Srivannavit O, Ozel AB. Minimizing the surface applications in chemistry, biology and medicine. Lab Chip. 2012;12:2452–63. effect on PCR in PDMS-glass chips by dynamic passivation. J Chem Technol 18. BenYehezkelT,Rival A,Raz O,CohenR,MarxZ,CamaraM,DubernJF,Koch B, and Biotechnol. 2007;82:33–8. Heeb S, Krasnogor N, Delattre C, Shapiro E. Synthesis and cell-free cloning of DNA 43. Akama-Garren EH, Joshi NS, Tammela T, Chang GP, Wagner BL, Lee DY, libraries using programmable microfluidics. Nucleic Acids Res. 2016;44(4):e35. Rideout Iii WM, Papagiannakopoulos T, Xue W, Jacks T. A modular assembly 19. Song H, Chen DL, Ismagilov RF. Reactions in droplets in microfluidic platform for rapid generation of DNA constructs. Sci Rep. 2015; https://doi. channels. Angew Chem Int Ed. 2006;45(44):7336–56. org/10.1038/srep16836 3. 20. Czar MJ, Anderson JC, Bader JS, Peccoud J. Gene synthesis demystified. 44. Kosuri S, Eroshenko N, LeProust EM, Super M, Way J, Li JB, Church GM. Trends Biotechnol. 2009;27(2):63–72. Scalable gene synthesis by selective amplification of DNA pools from high- 21. Huang MC, Ye H, Kuan YK, Li M, Ying JY. Integrated two-step gene synthesis fidelity microchips. Nat Biotechnol. 2010;28(12):1295–9. in a microfluidic device. Lab Chip. 2009;9(2):276–85. 45. Saaem I, Ma S, Quan J, Tian J. Error correction of microchip synthesized 22. Kong DS, Carr PA, Chen L, Zhang S, Jacobson JM. Parallel gene synthesis in genes using surveyor nuclease. Nucleic Acids Res. 2012;40(3):e23. a microfluidic device. Nucleic Acids Res. 2007;35(8):e61. 46. Sequeira AF, Guerreiro CI, Vincentelli R, Fontes CM. T7 endonuclease I 23. Quan J, Saaem I, Tang N, Ma S, Negre N, Gong H, White KP, Tian J. Parallel mediates error correction in artificial gene synthesis. Mol Biotechnol. 2016; on-chip gene synthesis and application to optimization of protein 58(8-9):573–84. expression. Nat Biotechnol. 2011;29(5):449–52. 47. Baldursdottir SG, Fullerton MS, Nielsen SH, Jorgensen L. Adsorption of 24. Tian J, Gong H, Sheng N, Zhou X, Gulari E, Gao X, Church G. Accurate proteins at the oil/water interface-observation of protein adsorption by multiplex gene synthesis from programmable DNA microchips. Nature. interfacial shear stress measurements. Colloids Surf B Biointerfaces. 2010; 2004;432(7020):1050–4. 79(1):41–6. 25. Stemmer WPC, Crameri A, Ha KD, Brennan TM, Heyneker HL. Single-step 48. Linshiz G, Jensen E, Stawski N, Bi C, Elsbree N, Jiao H, Kim J, Mathies R, assembly of a gene and entire plasmid from large numbers of Keasling JD, Hillson NJ. End-to-end automated microfluidic platform for oligodeoxyribonucleotides. Gene. 1995;164(1):49–53. synthetic biology: from design to functional analysis. J Biol Eng. 2016;10(1):1. 26. Gibson DG, Smith HO, Hutchison CA III, Venter JC, Merryman C. Chemical 49. Shih SC, Goyal G, Kim PW, Koutsoubelis N, Keasling JD, Adams PD, Hillson synthesis of the mouse mitochondrial genome. Nat Meth. 2010;7(11):901–3. NJ, Singh AK. A versatile microfluidic device for automating synthetic 27. Yoon J, Garrell RL. Preventing biomolecular adsorption in electrowetting- biology. ACS Synth Biol. 2015;4(10):1151–64. based biofluidic chips. Anal Chem. 2003;75(19):5097–102. 50. Tangen U, GAS M, Sharma A, Wagler PF, Cohen R, Raz O, Marx T, Ben- 28. Dormitzer PR, Suphaphiphat P, Gibson DG, Wentworth DE, Stockwell TB, Algire Yehezkel T, JS MC. DNA-library assembly programmed by on-demand MA, Alperovich N, Barro M, Brown DM, Craig S, Dattilo BM, Denisova EA, De nano-liter droplets from a custom microfluidic chip. Biomicrofluidics. Souza I, Eickmann M, Dugan VG, Ferrari A, Gomila RC, Han L, Judge C, Mane S, 2015;9(4):044103. Matrosovich M, Merryman C, Palladino G, Palmer GA, Spencer T, Strecker T, Trusheim H, Uhlendorff J, Wen Y, Yee AC, Zaveri J, Zhou B, Becker S, Donabedian A, Mason PW, Glass JI, Rappuoli R, Venter JC. Synthetic generation of influenza vaccine viruses for rapid response to pandemics. Sci Transl Med. 2013; https://doi.org/10.1126/scitranslmed.3006368. 29. Prakash AR, Amrein M, Kaler KV. Characteristics and impact of Taq enzyme adsorption on surfaces in microfluidic devices. Microfluid Nanofluid. 2008; 4(4):295–305. 30. Erill I, Campoy S, Erill N, Barbé J, Aguiló J. Biochemical analysis and optimization of inhibition and adsorption phenomena in glass–silicon PCR- chips. Sensors Actuators B Chem. 2003;96(3):685–92. 31. Kleinert J, Srinivasan V, Rival A, Delattre C, Velev OD, Pamula VK. The dynamics and stability of lubricating oil films during droplet transport by electrowetting in microfluidic devices. Biomicrofluidics. 2015;9(3):034104.

Journal

BMC BiotechnologySpringer Journals

Published: Jun 1, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off