|Home | About | Journals | Submit | Contact Us | Français|
The quantification of plasmid DNA by the PicoGreen dye binding assay has been automated, and the effect of quantification of user-submitted templates on DNA sequence quality in a core laboratory has been assessed. The protocol pipets, mixes and reads standards, blanks and up to 88 unknowns, generates a standard curve, and calculates template concentrations. For pUC19 replicates at five concentrations, coefficients of variance were 0.1, and percent errors were from 1% to 7% (n = 198). Standard curves with pUC19 DNA were nonlinear over the 1 to 1733 ng/μL concentration range required to assay the majority (98.7%) of user-submitted templates. Over 35,000 templates have been quantified using the protocol. For 1350 user-submitted plasmids, 87% deviated by ≥ 20% from the requested concentration (500 ng/μL). Based on data from 418 sequencing reactions, quantification of user-submitted templates was shown to significantly improve DNA sequence quality. The protocol is applicable to all types of double-stranded DNA, is unaffected by primer (1 pmol/μL), and is user modifiable. The protocol takes 30 min, saves 1 h of technical time, and costs approximately $0.20 per unknown.
Templates submitted to biotechnology core facilities are typically of diverse vector types, and have been purified and quantified by diverse methods. Although this facility requests templates at 500 ng/μL, in preliminary experiments, template concentrations ranged from 0 to 6000 ng/μL (J. O’Shaughnessy, Hoechst 33258 assay, unpublished data). There is anecdotal evidence that template concentration can affect automated DNA sequencing success rates1,2 and that high template concentrations can impact capillary viability and contaminate adjacent lanes on slab gel sequencers.
A number of methods are available to quantify double-stranded (ds) DNA. The absorbance at λ = 260 nm is relatively insensitive (an A260nm of 0.1 corresponds to 5 μg/mL dsDNA)3 and nonspecific since numerous organic molecules and biopolymers including phenol, nucleotides, single-stranded (ss) DNA, protein, and RNA absorb at 260 nm. Fluorimetric dye binding assays for dsDNA have increased sensitivity and specificity. The ethidium bromide assay has higher sensitivity for dsDNA (1 μg/mL) than A260nm, but the fluorescence yield is constant over a limited dye/DNA ratio, and the reagent fluoresces with RNA and ssDNA.4–6 The Hoechst 33258 assay exhibits improved sensitivity (ca. 10 ng/mL) and specificity for dsDNA, but assays at two salt concentrations are required for high specificity.7,8
Compared with the aforementioned methods, the PicoGreen dye binding assay for dsDNA has greater sensitivity, specificity, and linear range.9,10 The Pico-Green assay is sensitive to 25 pg dsDNA/mL, which is 400-fold more sensitive than Hoechst 33258 dye. The PicoGreen assay has high specificity for dsDNA versus RNA or ssDNA, and is unaffected by protein, phenol, and numerous other reagents. For calf thymus DNA, the PicoGreen assay has a linear range of four orders of magnitude with a single dye concentration.9
There have been few reports on automation of DNA quantification with application to large numbers of user-submitted samples in a core or high throughput facility. Haque et al. compared PicoGreen, A260nm, and quantitative PCR (for a specific gene target) for quantification of a narrow range of dsDNA concentrations (up to 20 ng/μL) for a high-throughput genomics setting; they concluded that the PicoGreen assay had advantages of specificity and sensitivity.11 Tecan Robotics and Beckman Coulter offer automated DNA quantification protocols for integrated absorbance or fluorimetry plate readers.12–15 However, there are no published reports of application of these protocols to large numbers of samples.
An automated method to quantify plasmid DNA based on the PicoGreen assay has been developed, and over 35,000 DNA templates have been quantified. The protocol provides accurate determination of a broad range of DNA template concentrations, i.e., 30 to 1733 ng/μL (initial concentration). Based on 418 DNA sequencing runs, the quantification of user-submitted templates by the reported protocol significantly improved DNA sequence quality.
The PicoGreen dsDNA Quantitation Kit [PicoGreen (320 μM in DMSO), 20× TE buffer (200 mM Tris-HCl/20 mM EDTA buffer, pH 7.5), and lambda DNA (0.1 μg/μL TE)] was from Molecular Probes (Eugene, OR). pUC19 (1 μg/μL) was from New England Bio-labs (Beverly, MA). Corning-Costar clear v-bottom and black flat bottom 96-well plates were purchased from PGC Scientific (Frederick, MD) or Greiner (Pleasanton, CA). Pipet tips were from Robbins Scientific (Sunnyvale, CA) or Molecular BioProducts (San Diego, CA). Liquinox detergent was from Alconox, Inc. (New Hyde Park, NY).
The Beckman Coulter (Palo Alto, CA) Biomek 2000 robotic workstation included left- and right-side modules, and the pipette and gripper tools and holders shown in Figure 11.. The BMG FLUOStar 97 fluorescence plate reader (Durham, NC) was attached to the left side of the Biomek deck. The FLUOStar 97 was operated at a fixed gain of 20 with BMG 485-12 excitation and 520–35 emission filters. An IBM 300GL computer with Microsoft Windows NT operating system, Microsoft Excel 97, Bioworks v.3.1 and SILAS software (Beckman Coulter), and FLUOStar Galaxy software v.4.30-0 (BMG), control the Biomek and FLUOStar 97.
The setup of the deck for the Biomek protocol is shown in Figure 11.. The following solutions are loaded onto the Biomek: pUC19 (100 μL, 0.1 μg/μL TE) in a 1.5-mL microfuge tube and an empty 1.5-mL microfuge tube are placed in positions A1 and B1, respectively, of the microfuge 24 holder. For some experiments, lambda DNA (0.1 μg/μL TE) replaced pUC19 as DNA standard. TE and PicoGreen (1.6 μM in TE) are added to the quarter reservoirs. Approximately 10 μL of the templates to be quantified are placed in columns 2 through 12 of the PE 9600 plate. The read plate wells contain, in a total volume of 200 μL, 0.8 μM PicoGreen, TE, and up to 0.43 ng/μL (87 ng total) pUC19 DNA. The wells containing blanks, unknowns, and standards, and the concentrations of the latter are entered into the FLUOStar Galaxy software. FloPatterns 1 and 2 of the Biomek Bioworks software are modified for the number of templates to be quantified.
The Plasmid Quant protocol generates tables of fluorescence intensities (uncorrected and corrected for blanks; not shown), a standard curve (Fig. 22),), and a table of the calculated concentrations of unknowns. DNA concentrations are given for the original solutions. The pUC standards correspond to 1733, 1200, 700, 400, and 100 ng/μL of DNA, and a 1-ng/μL point is included for curve-fitting purposes. DNA unknowns of higher concentrations were flagged by the protocol, manually diluted five-fold, and reassayed.
An overview of the protocol is given in Table 11.. The protocol is in 96-well format and performs all pipetting, mixing, plate reading, and data handling. The reproducibility and accuracy of the Biomek MP20 tool and the manual Rainin Pipet-Lite L8–10 multichannel pipet (Oakland, CA) in pipeting 1-μL aliquots of pUC19 into 99 μL of TE were compared. For the MP20 robotic pipeting, the coefficient of variance (CV) ranged from 30% to 42%, and percent error from theoretical was from 5% to 32% (mean 13.6%). The manual multichannel pipettor (CV ca. 10%) was found to be three-fold more reproducible than the Biomek robot for pipeting 1 μL, if the remainder of the protocol was pipetted robotically. In all further experiments, 1-μL volumes were pipetted manually with the multichannel pipettor.
The protocol can be modified by the user at numerous levels. The number of columns to be pipetted can be modified based on the number of templates to be assayed. The deck setup and pipeting parameters can be modified, as can be the range of the standard curve and the dilution of standards and unknowns. In addition, the protocol can be truncated at the end of the pipeting and mixing to produce read plates suitable for a nonintegrated fluorimetry plate reader.
Of the 35,000 templates quantified with the protocol, approximately 18,000 were sequenced with BigDye chemistry version 2.0, 4000 with BigDye v.3.0, and 13,500 with BigDye v.3.1. Data are presented only for DNA sequenced with BigDye Terminator chemistry v.3.1 on an Applied Biosystems (ABI) PRISM 3700 capillary sequencer with POP6 polymer (ABI, Foster City, CA), as described previously.16 Reaction mixtures contained 185–500 ng template DNA (unless otherwise indicated), primer (3.5 pmol), BigDye Terminator Master Mix (1.2 μL), 5× Sequencing buffer (1.8 μL), and water in a total volume of 10 μL. One M13F primer/pGEM sequencing reaction was run per 96 well plate. Average and 1st Phred <20 scores were generated automatically by dnaLIMS software (dna-Tools Inc., Ft. Collins, CO). The Average Phred score is calculated over the entire sequencing run and the 1st Phred <20 score is the first base that the Phred score drops below 20. Unsuccessful runs were defined by Average Phred scores below 25 or 1st Phred <20 scores below 300.
Statistical analysis was performed using the Student’s t-test for unpaired data, unless otherwise indicated. A significant difference between means was taken as p <0.05.
The choice of the DNA standard for the PicoGreen assay was not trivial since different types of dsDNA had variable fluorescent yields with PicoGreen. In addition, although standard curves for several types of DNA including bacteriophage lambda and calf thymus dsDNA were linear, those for pUC19 and øX174 were nonlinear.9 Since most templates quantified in this facility are plasmids, pUC19 was chosen as the DNA standard. Over the concentration range (up to 1733 ng/μL) needed to assay a high percentage of user-submitted templates (Fig. 33),), the standard curve for pUC19 remained nonlinear in over 25 interim protocols, with varying concentrations of standards and PicoGreen (data not shown) and in the final protocol (Fig. 22).
Replicate determinations (n = 98) at two concentrations of pUC19 DNA (294 and 571 ng/μL) had CVs and % errors of approximately 10% (Table 22).). Quantification of replicates of two concentrations of pUC19 (n = 32) using a lambda DNA standard curve resulted in CVs (ca. 10%) similar to those found with a pUC19 standard curve, but % errors of approximately 30%, i.e., three-fold higher for the lambda DNA standard curve. In order to minimize systematic error in the PicoGreen assay, it is essential to choose the most appropriate DNA standard for the type of DNA assayed, namely pUC19 for plasmid quantification.
Data were compiled for replicates for the calculated concentrations and fluorescence intensities for each of the five concentrations of pUC19 used to generate the standard curve. The theoretical concentrations and the experimentally determined mean, standard deviation (SD), minimum, maximum, CV, % error, and n values are given in Table 33.. The calculated concentrations of the standards agreed closely with the theoretical values for concentrations from 100 ng/μL to 1733 ng/μL (% error of <1% to 7%). The CVs increased from 2% to 15% as the concentration of the standard decreased from 1733 to 100 ng/μL. In contrast, the fluorescence intensity for each pUC19 concentration varied strikingly from run to run. For the 1733 ng/μL standard, the fluorescence intensity had a mean of 35,400, but ranged from 18,000 to 64,500 (Table 33).). Similar broad ranges of fluorescence intensity were observed at each concentration of pUC19, with CVs ranging from 34% to 47%. The CVs for fluorescence intensity were 2.5-fold to 20-fold greater than the CVs for pUC19 concentration, indicating that a standard curve should be generated for each run.
The assay was not developed for accuracy for DNA concentrations below 30 ng/μL since the facility has anecdotal evidence from dRhodamine and older Dye Terminator chemistries run on slab gel and capillary sequencers that DNA sequencing reactions on templates submitted below 60 ng/μL have a high probability of failure (data not shown).
Over 35,000 templates have been quantified with the Plasmid Quant protocol. Representative data for 1350 plasmids are given in Figure 33.. All templates were ostensibly quantified by users and submitted at 500 ng/μL. Only 13% of the templates were found to have concentrations within 20% of the target concentration, i.e., between 401 and 600 ng/μL (cross-hatched). Approximately 28% of the submitted templates were below 100 ng/μL, i.e. more than five-fold below the target concentration. Fifteen percent of the submitted templates were in each of the following ranges: 101–200, 201–300, and 301–400 ng/μL, and 2% to 5% of the submitted templates were in the 100 ng/μL concentration ranges between 600 and 1000 ng/μL. Above 1000 ng/μL, each 100 ng/μL range accounted for less than approximately 1% of the total.
With a standard curve up to 1733 ng/μL, 98.7% of the 1350 templates assayed fell within the standard curve, and only 1.3% or 17 templates required dilution and re-assay. If the concentration range of the standard curve were decreased, a greater percentage of samples would require dilution and re-reading. For example, with a standard curve up to 500 ng/μL, 24%, or 318 templates, would require dilution and re-assay. M13 Universal primer (1 pmol/μL), which is at a three-fold higher concentration than that used in sequencing reactions, did not affect plasmid quantification (Table 22).
The effect of template quantification on the quality of DNA sequencing data for user-submitted templates with inserts was assessed in three experiments. In the first experiment, 99 templates were sequenced without quantification. Seventy four percent of the templates had Average Phred scores over 25 (mean scores of 44 ± 9) and were classified as successful, while 26% had Average Phred scores below 25 (mean scores of 12 ± 5) and were considered unsuccessful. In contrast, 81% of 183 randomly selected templates quantified by the protocol with the template concentration in the sequencing reaction mixture optimized, had Average Phred scores greater than 25 (mean scores of 49 ± 7), i.e., a 10% increase in DNA sequencing success rate for quantified templates.
Two subsequent experiments focused on templates that quantified below 60 ng/μL or over 800 ng/μL. Each template was sequenced at two concentrations, with all other reaction conditions held constant. This generated results equivalent to (i) no in-house template quantification and (ii) adjusting template concentration for quantification as much as possible.
For 40 templates that quantified below 60 ng/μL, each template was run (i) as if it was at the requested concentration of 500 ng/μL (0.5 μL added to reaction mixtures), and (ii) optimizing the amount of DNA in the reaction mixture based on quantification (6 μL added to reaction mixtures). Option (i) resulted in 15 ± 11 ng DNA in reaction mixtures, and only 12.5% or 5 of 40 successful sequencing runs (Average Phred scores of 11 ± 9). Option (ii) resulted in 173 ± 132 ng DNA in reaction mixtures, and an 87.5% success rate or 35/40 successful runs (Average Phred scores of 36 ± 11). Quantification of this set of templates increased Average Phred scores four-fold (4.4 ± 1.8) (p < 0.001). Analogous results were obtained with 1st Phred <20 scores or average signal intensities (data not shown).
For 28 templates that quantified between 814 and 1,952 ng/μL, each template was run (i) as if it was at the requested concentration of 500 ng/μL (0.5 μL added to reaction mixtures), and (ii) optimizing the amount of DNA added to reaction mixtures (1 μL of a 1:3 dilution). For options (i) and (ii), reaction mixtures contained 723 ± 145 ng versus 482 ± 97 ng DNA, respectively. Option (i) had a success rate of 35.7% (10/28 templates sequenced successfully) with Average Phred scores of 18 ± 12, while option (ii) had a success rate of 78.6% (22/28) with Average Phred scores of 38 ± 14. Quantification of high concentration templates (800 ng/μL) improved Average Phred scores three-fold (3.0 ± 2.1) (p < 0.001). Analogous results were obtained with 1st Phred <20 scores (data not shown).
Since 26% of the user-submitted templates quantified by the Plasmid Quant protocol were at concentrations below 60 ng/μL or above 800 ng/μL, with analysis of 35,000 samples, quantification would identify 9250 templates with a low probability of successful runs and save users and/or the facility the cost of these runs and troubleshooting.
The applicability of the protocol to other types of dsDNA was validated by substitution of lambda DNA (100 ng/μL) for the pUC19 standard in several runs. The resulting standard curves were linear (r = 0.998). For replicate assays at two concentrations of lambda DNA, CVs and % errors (ca. 10%) were comparable to those for pUC19 (Table 22).
Automation of template quantification saves approximately 1 h of technician’s time per 88 templates quantified. The manual duties eliminated by the protocol include numerous pipetting and mixing steps, construction of the standard curve, and calculation of template concentrations. The savings in personnel time is equal to an annual savings of $4500, assuming one full quantification run per day and salary and benefits of $35,000 annually. The automated protocol is twice as fast as the manual equivalent, requiring only 30 min per full run. The cost to run the protocol for the standard curve plus 88 samples is $18.74 for consumables (including $3.20 for the use of pUC19), which is equal to $0.21 per sample. The total cost can be reduced by washing the 96 well plates in Liquinox and reusing them.
In conclusion, the application of the Plasmid Quant protocol to user-submitted templates in a core facility setting improves the quality of DNA sequencing data and facilitates troubleshooting. The protocol has significant advantages over commercially available alternatives. Firstly, the choice of the appropriate type of DNA for standard, and the nonlinearity of standard curves with plasmid DNA has been extensively addressed in this protocol, but neither has been addressed in the commercially available protocols. Secondly, none of the commercially available protocols have standard curves beyond 500 ng/μL, which would necessitate dilution and re-assay of approximately 25% of submitted templates in a core facility setting. Finally, the Plasmid Quant protocol is available upon request at no charge, versus the cost of approximately $3000 for commercially available protocols.
This work ws supported by NIH grants SIG S10RR15938 and S10RR13949 to KI for the Applied Biosystems 3700 DNA sequencer and Beckman Biomek 2000.