Use of Elemental Analysis To Determine Comparative Performance of

Anal. Chem. 2006, 78, 4630-4633

Use of Elemental Analysis To Determine Comparative Performance of Established DNA Quantification Methods Claire A. English, Sheila Merson, and Jacquie T. Keer*

Bio-Molecular Innovation, LGC, Queens Road, Teddington, TW11 0LY, U.K.

Quantification of genomic DNA is critical for many analyses in molecular biology. Current methods include optical density (OD) measurements or fluorescent enhancement but both approaches have limitations on achievable accuracy. In this study we performed an elemental analysis to quantify genomic DNA to provide an independent value for comparing the performance of four quantification methods. Specifically ICP-OES (inductively coupled plasma-optical emission spectroscopy) was used to assign a concentration value to a DNA stock solution, based on the stoichiometry of phosphorus within the molecule. Two absorbance- and two fluorescence-based methods were then used to quantify the same DNA solution using replicate analyses. The precision of each method was assessed by measurement of replicate spread (coefficient of variation) and trueness by t-test. Results showed that performance of the methods was variable, both in terms of concordance with the independent ICP-OES value and repeatability of data. While need for expensive equipment and technical expertise may preclude widespread replacement of more traditional methods for DNA quantification, use of primary methods such as ICP-OES analysis for production of accurate calibrants may increase quantitative accuracy and give greater appreciation of the true performance of current methods. The requirement for absolute quantification of genomic DNA is fundamental to many molecular analyses, and demands for higher accuracy have increased with the development of quantitative diagnostic assays1,2 and genotyping.3,4 To have confidence in quantitative results, either the amount of starting material put into a detection assay or the concentration of genomic DNA used to construct calibration curves must be known. There is a range of methods available for quantification; however, in the absence of certified reference materials or an absolute measurement technique, determining the accuracy of each is currently a challenge. * Corresponding author. Tel.: +44 (0) 20 8943 7449. Fax: +44 (0) 20 8943 2767. E-mail: [email protected]. (1) Gal, S.; Fidler, C.; Lo, Y. M.; Taylor, M.; Han, C.; Moore, J.; Harris, A. L.; Wainscoat, J. S. Br. J. Cancer 2004, 90, 1211-15. (2) Zhong, X. Y.; Holzgreve, W.; Hahn, S. Hypertens. Pregnancy 2002, 21, 7783. (3) Duewer, D. L.; Kline, M. C.; Redman, J. W.; Butler, J. M. Anal. Chem. 2004, 76, 6928-34. (4) Kline, M. C.; Duewer, D. L.; Redman, J. W.; Butler, J. M. Anal. Chem. 2003, 75, 2463-69.

4630 Analytical Chemistry, Vol. 78, No. 13, July 1, 2006

In this study we used an elemental analysis of DNA based on phosphorus stoichiometry to assign a concentration value to a DNA solution. This determined concentration was used as the basis for comparison of four quantification methods in terms of their measurement accuracy. Comparative studies of DNA measurement techniques have been carried out in the past by researchers in different fields.5,6 However, without an “absolute” measurement technique for genomic DNA, the values are still benchmarked against one of the test methods, for example, an optical density (OD) reading, which is not a definitive measurement method. To develop a procedure for measuring genomic DNA concentration, we investigated the use of ICP-OES (inductively coupled plasma-optical emission spectroscopy) analysis to quantify a stock genomic DNA solution by calculating the phosphorus content of the DNA backbone. This method has previously been used to quantify short oligonucleotides7-9 but as yet not extended to complex high molecular weight species, where a practical requirement lies. The study presented here assessed two UV absorbance- and two fluorescence-based DNA measurement methods routinely used in molecular biology laboratories. With use of the ICP-OES quantification method to achieve an independent anchor point, the methods could be compared meaningfully against each other and assessed in terms of their overall performance. UV absorbance measurements should be used with some caution due to the inaccuracies introduced by the pH, buffer components, and other UV absorbing impurities.10 The development of fluorescent dyes has provided an alternative technique specific to double-stranded DNA.11 PicoGreen is a highly sensitive cyanine dye, with a wide dynamic range enabling selective detection of double-stranded DNA. It has benefits over other available dyes such as Hoechst 33258 because of its low background signals 12 and unbiased behavior toward GC-rich regions. In this study we have investigated the performance of (5) Haque, K. A.; Pfeiffer, R. M.; Beerman, M. B.; Struewing, J. P.; Chanock, S. J.; Bergen, A. W. BMC Biotechnol. 2003, 3, 20. (6) Rengarajan, K.; Cristol, S. M.; Mehta, M.; Nickerson, J. M. Mol. Vis. 2002, 8, 416-21. (7) Murphy, J. H.; Trapane, T. L. Anal. Biochem. 1996, 240, 273-82. (8) Yang, I.; Han, M. S.; Yim, Y. H.; Hwang, E.; Park, S. R. Anal. Biochem. 2004, 335, 150-61. (9) Donald, C. E.; Stokes, P.; O’Connor, G.; Woolford, A. J. J Chromatogr., B 2005, 817, 173-82. (10) Cavaluzzi, M. J.; Borer, P. N. Nucleic Acids Res. 2004, 32, e13. (11) Singer, V. L.; Jones, L. J.; Yue, S. T.; Haugland, R. P. Anal. Biochem. 1997, 249, 228-38. (12) Ahn, S. J.; Costa, J.; Emanuel, J. R. Nucleic Acids Res. 1996, 24, 2623-25. 10.1021/ac060174k CCC: $33.50

© 2006 American Chemical Society Published on Web 05/11/2006

Table 1. ICP-OES Conditions for Phosphorus Determination condition

Figure 1. Experimental protocol for quantification method comparison. Six individual samples from the stock DNA solution were diluted in two stages, first 1:2 and then a further 1:500, for measurement using two UV absorbance- and two fluorescence-based quantification methods.

(1) the U-2000 spectrophotometer (Hitachi, UK), (2) the ND-1000 spectrophotometer (Nanodrop, USA), (3) PicoGreen fluorescent dye (Molecular Probes, UK), and (4) Quant-iT fluorescent dye (Molecular Probes, UK) for quantification of purified genomic DNA, by running a series of replicate samples on each instrument. EXPERIMENTAL SECTION Experimental Design. The study protocol was designed to compare the performance of four DNA quantification methods in terms of their measurement accuracy (trueness and precision) relative to an anchor point. A single stock solution of DNA was prepared and quantified using an independent elemental analysis method (ICP-OES). To assess each approach, six 0.5 mL samples were taken from the DNA stock and diluted in two stages to create solutions at the appropriate concentration ranges for the methods being studied. Each sample was diluted 1:2 for the spectrophotometric measurements and then by a further 1:500 for the fluorescence readings. The six individual samples were each quantified in replicates of six to give a total of 36 measurements per method. A schematic diagram of the experimental design is shown in Figure 1. Materials. A homogeneous and purified stock of human placental DNA (Sigma, Poole, UK) was prepared semigravimetrically for the quantification, at an approximate concentration of 100 µg/mL. The solution was stirred overnight to ensure complete homogeneity and then dialyzed against water (Elga 18.2 MΩ) for 7 h at room temperature (10 K MWCO snake-skin dialysis tubing, Pierce, Rockford, IL) to remove any low molecular weight molecules and ions, particularly external phosphorus. ICP-OES Analysis. Previous work had already demonstrated the agreement of the method with independent Isotope Dilution Mass Spectrometry (IDMS) quantification.9 Three 10 mL aliquots of the DNA stock were taken from the solution under stirring conditions, together with three blank water samples (Elga 18.2

setting

RF power (W)

1300

gas (L/min) plasma, argon auxiliary nebulizer

15 0.5 0.8

nebulizer uptake rate (mL/min)

1.0

wavelength (nm) phosphorus gold (internal standard)

213.617 208.209

time (s) integration read

0.1 8.0

MΩ) for phosphorus concentration determination using ICP-OES analysis. The samples were not further diluted, digested, or otherwise pretreated prior to analysis, and an initial estimation gave a phosphorus value of 5 ppm. Samples were then either measured using single standard matching (averaging five individual sample replicates) or external calibration (with three replicate readings averaged) to give a final concentration value. Gold was used as the internal standard, at 0.7 ppm for external calibration, and at 2 ppm for the single matched standard. The phosphorus standard was prepared gravimetrically in water from 99.995% pure NH4H2PO4 (Alfa Aesar, Karlsruhe, Germany) after oven drying at 105 °C for 20 h. For external calibration the phosphorus standards were 0, 2.5, and 5.0 and 10.0 ppm, while for the single matched method the standard was 5.6 ppm. Analysis was performed on an Optima 3300RL radial view instrument (Perkin Elmer, Beaconsfield, UK) with manually set read times. A gem tip cross-flow nebulizer was used, with a Ryton double-pass Scott-type spray chamber. The machine operating conditions are detailed in Table 1. The phosphorus value was converted into the equivalent amount of DNA using the following calculation:

DNA mass )

av mol wt nucleotide × mass of P mol wt P

An average nucleotide mass value of 308.92 was calculated using molecular weights of each condensed nucleotide monophosphate, and a GC ratio of 41% was determined from the composition of the human genome.13 The density of the solution was determined as 0.9978 ((0.0033) g/mL, and this was used to convert mass measurements into concentration values. Measurement Methods. U-2000 Spectrophotometer. For each replicate measurement, 100 µL of the well-mixed sample was pipetted into a cuvette (Hellma 105.201-QS, Mullheim, Germany) and loaded into the U-2000 spectrophotometer (Hitachi, Berkshire, UK) with a 100 µL blank water sample (Elga 18.2 MΩ). Ultraviolet (UV) absorbance readings were taken at 280 and 260 nm. The (13) International Human Genome Sequencing Consortium. Nature 2001, 409, 860-921.

Analytical Chemistry, Vol. 78, No. 13, July 1, 2006

4631

cuvette was rinsed out between each sample reading with ultrapure water and blank readings were taken between each set of 6 sample readings. The DNA concentration was calculated from the A260 value for each replicate. ND-1000 Spectrophotometer. Each well-mixed sample (1.2 µL) was loaded onto the NanoDrop ND-1000 (NanoDrop, Delaware) instrument in turn and the UV absorbance reading taken at 230, 260, and 280 nm using the “DNA 50” settings. The software automatically calculated the DNA concentration in ng/µL that was corrected for dilution to determine the final concentration. Blank samples were run between each set of 6 replicate results using water (Elga 18.2 MΩ). PicoGreen Fluorogenic Dye. First, an 11-point standard curve was constructed directly into a 96-well plate using the “DNA standard solution for PicoGreen DNA Assays” (Cambio, Cambridge, UK) at 1 µg/mL diluted with 1× Tris-EDTA (TE) pH 7.5, to produce duplicate standards with final concentrations in the range of 0-175 ng/mL. Then 100 µL of each unknown sample was added to the 96-well plate in replicates of 6, together with the same volume of PicoGreen dye (Molecular Probes, Invitrogen, Paisley, UK) diluted 1:200 in TE buffer. After a 5 min incubation at room temperature the fluorescence was measured using a Denley Wellfluor microplate reader (Slough, UK) (excitation 485 nm, emission 530 nm). A standard curve was plotted from the data produced using Excel (Microsoft) and from this the unknown sample concentrations were calculated. Quant-iT Fluorescent Dye. Measurements were carried out according to the High Sensitivity Quant-iT kit protocol (Molecular Probes, Invitrogen, Paisley, UK) using a 96-well plate format. Two microliters of each λ DNA standard was added to the plate, to generate an 8-point standard curve with a final concentration range of 0-99 ng/mL (each reaction in duplicate). Six replicates of each unknown (20 µL) were pipetted into the appropriate wells of the plate. The fluorescence was measured under the same conditions as PicoGreen above. From the data produced, a standard curve was plotted in Excel and the concentration of unknowns calculated. Safety Considerations. The PicoGreen and Quant-iT reagents are potential mutagens and should be handled with care according to manufacturers recommendations and disposed of as hazardous waste. Statistical Data Analysis. Data from the four sets of measurements produced 36 replicate data points (6 per sample) for each approach. The precision of each method was assessed by plotting the data and calculating the coefficient of variation (CV). To establish the level of trueness and concordance with the assigned value, data from the four approaches were analyzed using a student’s t-test to compare the mean concentrations from each method to the mean ICP-OES value for the DNA stock. RESULTS AND DISCUSSION ICP-OES Analysis. ICP-OES analysis of three aliquots of the DNA stock solution gave a mean phosphorus value of 5.3 µg/g, with CVs of 1.52-1.64% for the externally calibrated measurements and 3.17% for the single standard matching, respectively. The CV for the single standard matching measurements was higher than expected, as this approach would usually give a more consistent value than external calibration. The mean phosphorus value was then used to establish the DNA concentration of the stock, 4632 Analytical Chemistry, Vol. 78, No. 13, July 1, 2006

Figure 2. Distribution of quantification results by method.

calculated to be 52.75 µg/mL. The mean DNA concentration results from each of the methods were compared to this assigned value. The ICP-OES method measures the number of phosphorus ions in the sample, which is related to the DNA concentration by the phosphorus to DNA nucleotide mass ratio. However, any environmental or extraneous phosphorus not removed before the analyses will contribute to the phosphorus concentration determined, leading to an overestimation of the DNA level. Dialysis was used to purify the DNA in this study and remove phosphorus not integral to the DNA molecule itself. The use of ICP-OES analysis for DNA quantification purposes has as yet been restricted to short length oligonucleotides, and the progression to high molecular weight DNA strands is a little explored area. The data presented here highlight the potential for accurate DNA quantification with this approach. Spectral Analysis. All the measurements collected from each of the approaches are displayed in Figure 2, with 36 data points for each method. Precision. The precision associated with each measurement method was assessed by calculation of the CV for each (Table 2). The precision analysis demonstrated varying degrees of dispersion in the measurements from the 4 methods. Results from method 4 (Quant-iT) exhibited the widest distribution of data ranging from 49.96 to 74.70 µg/mL and hence the lowest degree of precision. The PicoGreen results were the most precise, with high intra-assay repeatability as reflected by the CV results (3.5%). This value is relatively low in comparison with those from other researchers (8.3 and 13.6%,5,6 respectively), emphasizing the close attention to detail with which the measurements were performed together with consistent performance of the assay. Interestingly, the other fluorescence-based method, the Quant-iT kit, demonstrated the lowest precision (CV 9.2%) and the precision of the two UV absorbance methods was intermediate. The reason for the disparity in performance between the two fluorescence-based measurements is unclear as both use the same principles of quantification, so the specific protocol employed in performing each reading could account for the variation. Of the two absorbance-based methods the ND-1000 results were less precise. This may be attributable to the very small proportion of the total sample that is used for each measurement (1.2 µL) compared to that for the U-2000 (100 µL), with associated sampling issues. Trueness. Figure 3 displays the mean concentration value (n ) 36) obtained for the DNA solution with each method in relation to the assigned value derived from the ICP-OES analysis. The

Table 2. Statistical Summary of 4-Method Comparison

method

mean concentration (µg/mL)

standard deviation

% differencea

CV %

t-value

t-crit

P-value

(1) U-2000 Spec (Hitachi) (2) ND-1000 spec (NanoDrop) (3) PicoGreen (Molecular Probes) (4) Quant-iT (Molecular Probes)

72.84 67.92 48.44 58.59

2.89 4.71 1.67 5.39

+38.08 +28.77 -8.17 +11.07

4.0 6.9 3.5 9.2

-41.76 -19.32 +15.47 -6.51

2.03 2.03 2.03 2.03

1.89 × 10-31 3.00 × 10-20 3.20 × 10-17 1.66 × 10-7

a

From the assigned ICP-OES value.

assigned concentration of the DNA stock, although the uncertainty associated with the results is large. Confidence in the t-results is high for all methods as indicated by the P-values, but the low probability associated with the Quant-iT measurement indicates that a true difference from the assigned value is least assured.

Figure 3. Method trueness and precision. The horizontal line across the plot represents the assigned DNA concentration from the ICPOES analysis. The error bars represent the CV for each set of measurements.

closest agreement was produced by the PicoGreen method, with the Quant-iT method, NanoDrop, and Hitachi decreasing in concordance, respectively. The mean concentrations from each method are shown in Table 2 with the t-test results, which indicate the significance of the differences between each experimental mean and the assigned value from the ICP-OES analysis. The PicoGreen method gives a low DNA concentration relative to the assigned value, possibly because only double-stranded DNA is measured, so any single-stranded material present would not contribute to the measurement. By contrast, the phosphate present in both single- and double-stranded DNA would be detected by the ICP-OES when the molecule is broken down, potentially yielding a more accurate value. The Quant-iT, ND-1000, and U-2000 results all demonstrated a positive bias, measuring 1138% higher than the ICP-OES value. The absorbance readings are likely to overestimate the value due to measurement of contaminants absorbing at A260, contributing to an overall higher result. Here the spectrophotometrically determined values are between 28 and 38% higher than the ICP-OES measurement and are substantially greater than those calculated using UV determinations of oligonucleotide concentrations by other researchers.8 The higher purity of chemically synthesized, HPLC separated oligonucleotides may facilitate more accurate spectrophotometric measurement. In addition, analysis of the t-statistics showed all methods had t-values greater than the t-critical value, demonstrating significant differences (P < 0.05) to the assigned ICP-OES value. The QuantiT measurements exhibited the least significant difference, followed by the PicoGreen, ND-1000, and U-2000, respectively. This can be explained by consideration of the precision of the measurements (Figure 3). Due to the widespread results produced by the Quant-iT method, some of the data points fall near the

CONCLUSION The ICP-OES method described here enables the performance of different DNA quantification methods to be compared against an independent elemental analysis, which is based on a primary measurement approach. In this investigation of two UV absorbance- and two fluorescence-based quantification methods the fluorescence-based methods were the most accurate with closest concordance of concentration results to our assigned ICP-OES derived value, although precision was variable. Further definition of the uncertainty associated with the P-analysis is required to improve the application for measuring large biomolecules, but the technique provides potential as a route to standardizing DNA quantification for use in accurate routine analysis and for certification of quantitative nucleic acid reference standards. For routine laboratory use, the most appropriate quantification method will depend on the type of sample being analyzed, the volume of material available, and the approximate concentration range. The fluorescent methods, for example, are able to measure much lower concentrations of DNA than the spectrophotometric methods and so would be appropriate for very diluted or trace level analytes. If the amount of sample available is a limiting factor, then the ND-1000 might be most appropriate due to the very small volume required for each measurement. Provision of performance characteristics for each instrument, as reported here, will enable the most appropriate method to be chosen. In addition, the results highlight the variability in DNA quantification methods that are in current use, by comparison to an independent technique. An understanding of the potential variability in the methods used to calibrate many molecular analyses will increase appreciation of the level of measurement uncertainty associated with quantitative DNA measurements. ACKNOWLEDGMENT We would like to thank Malcolm Burns for statistical advice. This work was funded by the National Measurement System Directorate, Department of Trade and Industry, under the Measurement for Biotechnology (MfB) programme 2004-2007. Received for review January 26, 2006. Accepted April 10, 2006. AC060174K Analytical Chemistry, Vol. 78, No. 13, July 1, 2006

4633