Efficient modification and preparation of circular DNA for expression in cell culture

Conceptual introduction

With this method, which consists of a single tube two-step reaction followed by a typical spin column-based DNA cleanup, a researcher can produce a batch of Circular Vectors in 3 h. The Circular Vector method simplifies making small variable targeting components that work in conjunction with an invariable large plasmid coding for Cas9 and/or other proteins and genes. This large plasmid can be cloned in large quantity in bacterial culture, extracted with one of the commercially available maxi-preps, and used for subsequent months or even years of experiments. Notably, the variable component for each editing target—the Circular Vector—is now a breeze to make.

Circularized DNA is resistant to exonuclease degradation in the cytoplasm¹⁴. Linear DNA lacks such stability, which is a significant barrier to utilizing it for gene delivery. Moreover, our method takes advantage of this difference in properties by using T5 exonuclease to digest unreacted or misreacted linear DNA fragments, thereby purifying the Circular Vector DNA.

Our circularization method for the CRISPR guide expression vectors requires an initial DNA sequence design, which can be created using the researcher’s preferred tools (we used SnapGene 6.0.7), followed by ordering dsDNA fragments from a commercial manufacturer.

The prefabricated dsDNA should—at a minimum—contain a promoter (U6, in our example), followed by a DNA segment coding for epegRNA with a terminator³⁸ (see Fig. 1a and the detailed map in Supplementary Fig. 1). The epegRNA can be designed using a published tool suited to this purpose^15,16,39. Two matching type IIS restriction enzyme recognition sites create complementary 4 nt overhangs near both ends of the dsDNA fragment. The Circular Vector can be conveniently validated with Sanger sequencing (see Fig. 1b). Placing two primers, Sanger-1 and Sanger-2, near the opposite locations of the circular structure has proven to be efficient and reliable, covering most of the circular structure with the overlap that includes the opposing primer near the middle of the FASTA sequence. If the whole plasmid sequencing is preferred, companies performing such sequencing—for example, Plasmidsaurus (www.plasmidsaurus.com)—require a minimum of 2000 bp circle, however techniques based on rolling circle amplification, like one developed by Octant⁴⁰ may be able to sequence short circular dsDNA.

The matching overhangs created by the restriction enzyme cuts are designed to provide a high-fidelity ligation⁴¹, which is also aided by the absence of any other overhangs that would be present in a typical Golden Gate assembly (see Fig. 1c). The ’nnnnnn’ sequences on Fig. 1c represent the minimum of six base pairs required on the back end of the restriction enzyme. Some manufacturers (e.g., Twist Bioscience) prefer to deliver their DNA fragments with end adapters that serve as well-designed PCR amplification primers and simultaneously eliminate the need to incorporate end sequences.

Step-by-step protocol

We aimed to make the method of Circular Vector production user-friendly and rapid. Thus, we primarily used reaction enzymes and buffers based on commercial kits, with only a few additional reagents. If desired, the kits may be substituted with individual reagents. If a sufficient amount of dsDNA is received from a commercial manufacturer, it may be used directly in the reaction; otherwise, it may need to be amplified by PCR to obtain the desired quantity. The protocol for the preparation of the circularized DNA vectors consists of three steps:

Step 1: Circularization of the source DNA (Fig. 1d). Mix the reaction reagents listed in Table 2 by pipetting, preferably on ice, into a 200 μl PCR tube or another suitable tube with at least double the Table 2 reaction volume. The extra volume will be required for the addition of the digestion reagents in Step 2. Place the tube in a thermal cycler and run the reaction as described in Table 2 cycler schedule.

Step 2: Digesting all remaining linear DNA (Fig. 1e). Prepare the T5 exonuclease reaction mix as listed in Table 3 and mix it in 1:1 proportion into the tube containing the reaction mix from the previous step. For example, if a 50 μl reaction was performed in Step 1, 50 μl will need to be added. Program and run the thermal cycler as listed in Table 3 cycler schedule.

Step 3: Cleanup of the circularized DNA (Fig. 1f). Use a spin column kit such as QIAquick PCR Purification Kit, or your preferred method.

The above protocol steps are sufficient for a practical application of the method. For in depth explanation of the steps and reasoning behind them read the following section.

Protocol Implementation Details

1. 1.

   The restriction enzyme digestion and ligation reaction is followed by 15 min of heat inactivation to end the ligation reaction (see Fig. 1D). The constant 37 ^oC reaction temperature was chosen to minimize mismatch ligation^42,43. T4 DNA ligase inactivates at 65 ^oC. Since the restriction enzyme was chosen with a higher inactivation temperature, the restriction enzyme will remain active and slightly assist in the following digestion step. The concentrations of ligase and restriction enzyme were experimentally determined to work better with a slightly higher concentration than is typical in Golden Gate assembly. This higher concentration serves two purposes: it facilitates rapid processing, and, more importantly, the restriction enzyme rapidly cuts to free overhangs on both ends of a DNA fragment, thereby allowing the self-ligation of these overhangs on the same DNA fragment and limiting the proportion of multiple DNA fragments ligating on each other (see further discussion in the Results section). The concentrations of input linear dsDNA were tested in a range from 7.5 to 120 ng/μl and produced indistinguishable circularized DNA yield and composition. A concentration of 120 ng/μl results in the input of 6.0 μg DNA per 50 μl reaction, which was considered optimal from handling and reagent cost perspectives. The addition of ATP may be optional for a 1 h ligation duration, with increasing ATP concentrations from 1 mM provided by the NEB T4 DNA ligase buffer to 2 mM showing a slight improvement in yield, particularly at longer ligation durations. This may be due to the exhaustion of ATP with a high concentration of ligated DNA, as well as ATP inactivation at longer reaction durations. The 2 mM ATP concentration is well within the optimum range for T4 DNA ligase mixes⁴⁴, and the ATP reagent is inexpensive. The restriction enzyme substitution to a different type IIS enzyme may be required in cases where the sequence coding for epegRNA includes the BsaI recognition site (for an example of using BbsI in a slightly different scenario, see Supplementary Fig. 2).

2. 2.

   Digestion of non-circularized DNA (Fig. 1e). This step consists of adding an equal amount of the T5 DNA exonuclease reagents to the ligation reaction. For example, 50 μl of exonuclease reaction mix is added to 50 μl of ligation reaction output, resulting in a comfortable level of 100 μl of reaction mix in a typical 200 μl Eppendorf PCR tube thermocycler (larger tubes and dry baths, etc. may be used if desired). The concentration of T5 DNA exonuclease was validated as sufficient to completely digest all linear DNA in under 1 h for all input DNA lengths tested in this research. We found that adding only T5 exonuclease to the ligation reaction resulted in a very slow digestion rate due to the absence of potassium ions. The role of NEBuffer 4 is to provide potassium anions that accelerate exonuclease reactions. While the concentration of potassium in the resulting mix is half that of the exclusively NEBuffer 4 reaction mix, it remains close to the optimum concentration for cleaving reactions⁴⁵. The optimal balance between exonuclease and endonuclease activity for T5 exonuclease is at the reaction pH 7.9–8.0⁴⁶. This is the pH of NEBuffer 4, while the ligation buffer pH is 7.5. Thus, a suitable amount of Tris base was added to increase pH to this range, as presented in Table 2 (main article). Notably, it is not recommended to excessively increase pH for the T5 exonuclease reaction mix. When we tested pH increases above 8.5, all DNA—including the circularized DNA—was completely and rapidly digested.

3. 3.

   Cleanup of the circularized DNA. This step involves using a quality DNA cleanup kit to extract Circular Vector DNA of sufficient concentration and purity. We used a spin column kit, which was selected as described below. With a large number of such products available, the choice of kit is up to the researcher’s preference; however, the yield may vary. We aimed to select a simple spin column-based extraction kit that would provide a consistently high yield, a low level of impurities, and a sufficiently high DNA concentration for accurate and consistent measurements. This purity was important for the direct use of the Circular Vectors in cell culture transfection and for the accuracy of concentration and yield measurements reported by this study. A number of kits were reviewed, and three kits were then tested and evaluated for their efficiency in DNA extraction and low chaotropic salt contamination: the Takara Bio NucleoSpin Gel and PCR Cleanup 740609.50⁴⁷, QIAGEN QIAquick PCR Purification Kit 28106⁴⁸, and NEB Monarch PCR & DNA Cleanup Kit T1030S⁴⁹.

   Based on a validation test with the same set of PCR products, the QIAquick kit was chosen based on it having the best recovery of input to output DNA when compared to the other two kits (20% better than NucleoSpin and 30% better than NEB) and exhibiting a low amount of captured chaotropic salts when testing a blank PCR product cleanup without any input DNA. In addition, even though the QIAquick kit rates 30 μl as a minimal elution volume, it performs with a good compromise of yield and higher DNA concentrations and shows excellent purity when eluting with volumes below 20 μl, which was preferable for our prime editing validation experiments. The QIAquick PCR Purification Kit has proven very reliable and consistent, with only one outlier that had approximately half the yield observed in the whole set of experiments. This outlier result may have been due to a defect in the spin column; thus, the sample was discarded.

   The QIAquick kit includes an optional pH indicator that is recommended to add to the binding buffer due to its importance in maximizing DNA yield. The indicator reagent allows us to verify whether the mix pH is optimal for binding. Notably, we determined that pH adjustment was required via the addition of one or more 10 μl of 3M pH = 5.2 sodium acetate, as recommended by QIAGEN⁴⁸.

Special attention was given to minimizing pipetting losses. The amount of input DNA was adjusted to result in output product DNA in the range of 2–4 μg, which provided a sufficient amount of the final DNA product for measurement and testing. In extreme cases, the amount of input dsDNA for the 1782 bp Circular Vector had to be scaled up to 27 μg, which still resulted in only a 1.3–1.7 μg yield, while an input DNA amount of 6–9 μg was sufficient in 450–950 bp range. Such low yield, which would require high DNA input and the use of large quantities of reagents to produce sufficient product, indicates that long Circular Vectors are impractical to synthesize. We preferably aimed to achieve a concentration of output DNA well over 80 ng/μl, which would result in high-purity DNA based on 260/230 and 260/280 ratios. This was additionally aided by performing an extra wash step when using the purification kits, which resulted in consistently pure DNA output with a minimal potential sacrifice in yield. The DNA concentration measurements were primarily performed on a NanoDrop Eight Spectrophotometer (ThermoFisher Scientific) and repeated on a DS-11 Series Spectrophotometer/Fluorometer (DeNovix). Notably, the measured concentrations were closely matched between these two devices.

Extraction of Band 1 for testing gene editing performance of the pure single Circular Vector with no duplicates was performed by DNA separation by electrophoresis on 1% E-Gel EX Agarose Gels (Invitrogen). The Circular Vector DNA was cleaned using Monarch DNA Gel Extraction Kit (T1020S) with one extra wash step.

Equations used in data analysis

1. 1.

   The maximum possible yield ratio of the circularization reaction is calculated by dividing the circularized DNA length, as determined by the restriction enzyme cut sites based on input dsDNA length. Typically, a type IIS restriction enzyme requires six base pairs or more between it and the end of the dsDNA strand to ensure efficient cutting. The restriction enzyme recognition site itself, an offset between the recognition site and the cut location, and one length of the overhang are also discarded. For our handling, it was convenient to retain the Twist Bioscience end adapters, which are 22 bp in length and can be used with Twist Bioscience primers for the PCR amplification of these DNA fragments. In this case, the maximum possible yield ratio denominator is the length of the dsDNA ordered from Twist Bioscience plus 44 bp. All of the elements can be formally accounted for by the following equation:

where R is the maximum yield ratio, L_c is the length of the circularized DNA, L_{enz_site} is the recognition enzyme site, L_{enz_offset} is the offset between the enzyme recognition site and the beginning of the cut, L_{cut_length} is the length of the cut (typically 4 bp), and L_{end_pad} is either padding or end adapters (as in our case), assuming that they are of equal length on either end. As an example, for dsDNA delivered by Twist Bioscience with adapters for making a 452 bp Circular Vector, the maximum possible yield ratio is 0.88. Thus, if 6 μg of DNA is used as an input, the maximum possible yield will be 6 * 0.88 = 5.28 μg.

1. 2.

   We define the molar multiplier of the circularization reaction product as the ratio of the combined relative weight produced by circularization to the molar relative fraction of a single circular guide. We assess the relative band brightnesses from the gel electrophoresis image, which was processed using the NIH’s ImageJ (see Supplementary Fig. 3 for the complete set of unmodified gel elecrtophoresis images). Since we know the relative brightness, we must sum all band relative brightnesses, which, by definition, will be = 1 divided by the sum of each band’s relative brightness (I_i) divided by that band’s number i. Thus, the molar multiplier M is calculated as:

Transfection, handling, and sequencing of edited HEK293T cells

HEK293T cells were seeded on 24-well plates (Corning) at 140 ⋅ 10³ cells per well in DMEM supplemented with 10% FBS. Approximately 24 h after seeding, the cells were transfected at 60% confluency with 2.0 mL of Lipofectamine 2000 in Optimem (Thermo Fisher Scientific) according to the manufacturer’s protocol and 800 ng of Cas9 prime editor plasmid, with molar ratios of 4:1 (110 ng for purified single band 452 bp Circular Vectors, 175 ng for a 452 bp Circular Vectors, and 370 ng for an 952 bp Circular Vectors) for each of the location guides (HBB, CDKL5 and PRNP single base substitution based on¹⁵).

Then, 14–16 h after transfection, media was replaced with fresh DMEM plus 10% FBS and 6 ng/μl of blasticidin (Thermo Fisher Scientific) to select for cells containing prime editor. This was followed 24 h later by 6 ng/μl. Then, a media change was performed after 24 h (based on a recommendation by Xiong et al.⁵⁰). Genomic DNA was extracted 72 h later using a Zymo Research Quick-DNA Microprep Kit (D3020).

Genomic areas of interest were amplified from 4 ng of whole genome DNA using Q5 High-Fidelity 2× Master Mix (New England BioLabs) at 24–26 cycles using the master mix protocol (see Data Availability data files for a list of the primers). This was followed by DNA separation by electrophoresis on 1% E-Gel EX Agarose Gels (Invitrogen) and the use of a Monarch DNA Gel Extraction Kit (T1020S) with one extra wash step. Sanger sequencing was performed by Azenta Genewiz using our PCR primers.

Reagents, materials and costs

The protocol low reagent cost is exemplified here by the breakdown of a typical 50 μl reaction. The reagents are used in quantities listed in Table 2 and Table 3. The manufacturer sources and identifiers are provided in Table 4. The bulk of the expense inherent in Circular Vectors is the cost of linear dsDNA ordered from a commercial supplier. We will present the costs for the example of our 452 bp Circular Vector. The prices we paid for the reagents were typical for retail university buyers, without any extra discounts. With a circularized length of 452 bp, a 470 bp dsDNA fragment was ordered from Twist Bioscience at the list price of $0.07 per base pair, resulting in an input DNA cost of $32.90.

Usually, 1.5–2.0 μg of dsDNA fragment was delivered by Twist Bioscience. The costs of the reagents for a 50 μl protocol are presented in Supplementary Table 1, in quantities listed in Step-by-Step Protocol section. The planned output is 2.0–2.5 μg of Circular Vector, which requires an input of 6.0 μg of dsDNA. Such yield is sufficient for more than 10 edits on a 24-well plate and ~40 edits on a 96-well plate. The cost of this Circular Vector protocol is $9.30. If your supplier can deliver a sufficient quantity of DNA fragments, this is all your expense.

In our study, we had to run a PCR amplification step (Optional, priced in Supplementary Table 1), which added $4.76 to the cost. If only a small amount of Circular Vectors is needed, the reaction can be scaled down to a smaller quantity of available DNA, with a lower use of reagents and lower resulting cost.

Statistics and reproducibility

All protocol reactions were performed in independent reaction duplicates (n = 2) or triplicates (n = 3), as indicated, and yield percentage values were averaged. Gel electrophoresis of the samples above was analyzed using National Institutes of Health’s ImageJ software¹⁷ in independent sample duplicates (n = 2), and resulting concatemer percentage values were averaged.

The editing efficiency analysis performed from Sanger sequencing using EditR^22,23 from independent samples in triplicates (n = 3) or quadruplicates (n = 4). The P values for editing efficiency calculated by EditR were P < 5 ⋅ 10⁻⁸ for all samples. The editing efficiency values were averaged, and the corresponding standard deviation values presented in Table 1.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.