Bruce R. McCord¹, Bruce Budowle¹, Alice R. Isenberg², Ralph O. Allen^2
1FBI Laboratory, Quantico, VA 22135
²University of Virginia, Charlottesville, VA 22903

Length polymorphisms generated using PCR can be easily and efficiently analyzed using capillary electrophoresis (1-5). In CE, DNA molecules are separated by size as they migrate through a sieving medium contained within a narrow 50-75 um diameter capillary. The use of a narrow diameter capillary enables efficient heat dissipation during electrophoresis and allows much higher separation voltages to be used. The separation medium inside the capillary need not be a rigid physical gel; instead, a water-soluble entangled polymer solution is pumped into the capillary and replaced at the conclusion of each sample analysis. The result is a high speed, high resolution separation that can be repeated many times before the capillary needs to be replaced.

Systems allowing complete automation of DNA analysis using robotic sampling and multiwavelength detection are becoming commercially available. We have been evaluating one such instrument, the ABI PRISM 310 Genetic Analyzer (Perkin Elmer) to determine its suitability for forensic DNA analysis. This particular CE system allows four channel multiwavelength detection with software capable of rapid size estimation based on in-lane internal standards. The use of a power-driven syringe permits viscous polymers to be pumped into the capillary, however capillaries may only be flushed with the buffer solution. Automated washing of the capillary with cleaning solutions such as hydroxide or methanol is not possible when viscous polymers are used. For this reason, capillary lifetime is generally limited to under 100 runs.

Since this instrument has the capability to perform separations of ssDNA using highly viscous polymers, a study was undertaken to assess the effects of polymer, field strength, and temperature on the separation. Principally, these experiments were performed using the polymer hydroxyethyl cellulose (HEC). Polymers supplied by Perkin Elmer were also evaluated. Two different PCR-amplified short tandem repeat multiplex systems [AmpFISTR Blue (Perkin Elmer) and GenePrint™ Fluorescent STR System/CCTv (Promega Corp.)] to assess the performance of the CE and polymer matrices. The goals of these experiments were: single-base resolution to 350 bp, analytical precision better than 0.2 bp and long capillary lifetimes.

Fluorescently-labeled allelic ladders were obtained from Promega Corp. (GenePrint™ Fluorescent STR System (CTTv), Madison, WI) and Perkin Elmer (AmpFISTR Blue, Foster City, CA). Amplified samples were prepared using standard protocols and primers from their respective companies. Trizma base and boric acid were obtained from Sigma Chemical Co. (St. Louis, MO), EDTA was purchased from Fisher Scientific (Fair Lawn, NJ), and the hydroxyethyl cellulose (cat # 30,863-3) was obtained from Aldrich Chemical Co. (Milwaukee, WI). GS350 Rox standard and GS 500 Tamra standard were obtained from Perkin-Elmer. Deionized formamide was prepared by adding approximately 2g of Amberlite MB-1 (Sigma) to 50 ml of high purity formamide (Sigma) [6]. For CE analysis, 2 µl of PCR product and 2 µl of internal standard were added to 24 µl of deionized formamide. Some allelic ladder samples were subjected to a float dialysis procedure utilizing a Millipore (Milford, MA) VCWP 013 00 (0.1 um) filter to remove excess buffer salts and to ensure reproducible injections [7,8]. This dialysis procedure is unnecessary for most amplified samples [4].

A Perkin Elmer/ABI 310 genetic analyzer was used for all experiments and was calibrated using matrix standards labeled with ROX, JOE, TAMRA, and FAM fluorescent dyes [6]. Capillary columns used included uncoated 50 um capillaries from Perkin-Elmer and Supelco (Bellefonte, PA) as well as 50m m DB-17 phenyl methyl coated capillaries from J & W Scientific (Folsom, CA) One sieving medium contained 100mM tris-borate, 2mM EDTA, (TBE) pH 8.6 adjusted with Cs)H with 2.0% hydroxyethyl cellulose and 7.4 M urea. The HEC polymer solution was prepared by mixing a 3% solution of the polymer in TBE and allowing the solution to stir overnight. 12.8g of urea were added to 20 ml of this solution to produce a solution containing 7.4 M urea and 2% HEC. The mixture was periodically shaken and allowed to dissolve over a period of several hours. It was then filtered through a 5 um syringe filter (Gelman Sciences pn/ 4199). To avoid problems with crystallization of saturated urea solutions, buffer chambers were filled with 3% cellulose solution containing no urea. Additionally, a proprietary solution from Perkin-Elmer was examined (Genetic Analyzer Buffer with High Resolution Polymer 4). Proprietary buffer solutions were prepared as described by the manufacturer.

Samples were analyzed using 40cm capillaries at field strengths of 250-375 V/cm. Coated capillaries were prepared by rinsing with methanol for 10 minutes prior to installation. Uncoated capillaries were used as received for the High Resolution Polymer 4. Injections were made at 2.4 kV for 5-30s. Data were analyzed using the Perkin-Elmer GeneScan genetic analysis software and Microsoft Excel.

The goal of this study was to optimize ssDNA separations of multiplexed STR loci using the PE/ABI 310 CE instrument. A number of different parameters were evaluated, including temperature, field strength, and polymer matrix. Experiments were performed to determine how various changes in analysis parameters affect precision, resolution, and capillary lifetime.

A double polymer sieving medium developed for use with double-stranded DNA [4] was adapted to separate single-stranded DNA samples. The polymer used, HEC, has been well characterized [11]; however with the exception of a recent study on its use as a polymer for sequencing via CE [12], little work has been published on the capability of HEC to resolve ssDNA. To adapt the sieving medium for the separation of ssDNA fragments, the HEC polymer concentration was increased to 2%, and urea was added at a concentration of 7.4 M. The procedure used in this paper is similar to that utilized by Bashkin et al. [12], although a shorter polymer was used to keep viscosity to a minimum, and formamide was not added to the buffer.

During experiments to determine the optimum polymer concentration, a concomitant increase in resolution was observed as the polymer concentration rose. However, to allow fill times of four minutes or less and to minimize viscosity, the polymer concentration was maintained at 2%. The urea concentration was adjusted to 7.4 M to reduce problems with peak splitting and to improve resolution. As a result of these changes, single base resolution was possible at fragment sizes ranging to 300 bases in length. For example, figure 1 illustrates the analysis of two alleles of approximately 200 bases long from the TH01 STR locus which differ in length by a single base. The resolution between the 9.3 and 10 alleles is 1.1 or 0.93 bases.

Upon development of the above procedure, a series of tests was performed to evaluate the precision of the instrument with the above polymer system. The first experiment was a measurement of the effect of injection time on resolution and peak height. The sample in Figure 1 was analyzed using injection times from 5 to 30 seconds at 2.4 kV. These results were anticipated as the quantity of material determined on a CE system using an electric injection as a complex function of injection time, salt content and voltage [13]. Multiple injections from the same sample vary widely as ionic strength and sample concentration change. To eliminate the effect of multiple injections, 18 individual samples at the same concentration were sequentially injected. In these results, the sample area still varied with a relative standard deviation of 14%. Hydrodynamic injections which produce lower standard deviations are not possible with the PE/ABI 310 instrument [10].

Additional experiments were carried out to assess the effect of temperature and voltage on sample resolution. Table 1 shows these results:

The resolution between any two individual peaks was calculated using the standard chromatographic definition, modified to permit measurement using the width at ½ the peak height:

where T is the peak retention time, W is the peak width and W (½) is the peak width at half height [9]. Dividing the size difference of the two peaks by the resolution produces an estimate of the peak resolution in units of bases.

The results in Table 1A show an improvement in precision of size estimates at higher temperature. This may be due to partial reannealing of the ssDNA fragments at lower temperatures. Concomitantly, there is a slight loss of resolution at higher temperatures due to increased dispersion of the DNA fragments as they migrate through the capillary column. Balance of these two separation parameters is required to develop a procedure to perform STR analyses. Currently, separations are carried out at 60°C, as this appears to be an effective temperature for analysis. The effect of field strength on the separation is shown in Table 1B. While higher field strengths will reduce analysis times, they can affect resolution due to an increase in sample dispersion caused by the higher temperatures generated at these voltages.

A study using the AmpFISTR blue STR multiplex (containing the loci FGA, vWA, and D3S1358) was performed using both HEC and High Resolution Polymer 4 (Perkin Elmer). Approximately 45 separate analyses, consisting of allelic ladders and amplified samples, were performed with both media. In general, both media provided good precision, with a standard deviation of approximately 0.2 bases produced by pooling the results from all fragments in the allelic ladder. Resolution was measured to be between 1 and 1.6 bases at DNA fragment lengths of approximately 200 bases. Figure 2 illustrates the analysis of the AmpFISTR blue allelic ladder using the HEC matrix. The HEC analyses, performed at 30°C, showed lower precision for larger-sized DNA fragments. As mentioned above, precision can be improved by performing the analyses at temperatures of 45°C and above. The polymer system from Perkin-Elmer, used at 60°C, produced better precision, even though the resolution was not as high. When available, the results obtained were compared with those acquired using slab gel analysis. In spite of the variations in temperature and polymer, in all cases the allele types matched those from the gels.

One problem observed with both analytical protocols was the decrease in column performance over time. It was determined that column performance for the HEC system using coated capillaries could be restored by flushing the system with methanol for 30 minutes prior to each set of analyses. Presumably, this step removes chemisorbed materials from the column walls. A similar step could be performed on the uncoated capillaries of the proprietary polymer system by rinsing with 1M NaOH, although this procedure was not attempted.

An additional observation was that size estimates varied by 7 - 12 bp between the two polymer systems. While such differences might be expected given the variations in analysis conditions, changes in size estimates underscore the importance of the use of standard allelic ladders as references when performing any STR analysis in which results from different methodologies are compared.

A series of experiments was carried out to evaluate the utility of capillary electrophoresis with multiwavelength fluorescence and a soluble polymer sieving medium for the analysis of multiplexed STR loci. The results show a precision of 0.2 bases or better using either of two different polymer systems, and the analyzed data are consistent with those typed using gel systems. In tests of a novel polymer system consisting of 2% HEC in a TBE buffer, resolution of a single-base difference between TH01 alleles was demonstrated. Evaluation of temperature effects on the separation revealed that higher temperatures produced increased precision at the cost of some loss in resolution. The results indicate the potential of capillary electrophoresis as a useful tool for automated fluorescent analysis of DNA.

The authors would like to thank Jim Robertson, Kathleen M. Keys, and Jill Smerick for their technical expertise in support of this work. Major funding for A.R.I. was provided by the National Institute of Justice under grant #93-IJ-CX-0030. This is publication #97-01 of the Laboratory Division of the Federal Bureau of Investigation. Names of commercial manufacturers are provided for identification purposes only, and do not imply endorsement by the Federal Bureau of Investigation.