A Sample Purification Method for Rugged and High-Performance DNA

to achieve high performance and rugged DNA sequencing analysis. DNA sequencing fragments are introduced into a capillary electrophoresis (CE) column ...
0 downloads 0 Views 114KB Size
Anal. Chem. 1998, 70, 1528-1535

A Sample Purification Method for Rugged and High-Performance DNA Sequencing by Capillary Electrophoresis Using Replaceable Polymer Solutions. B. Quantitative Determination of the Role of Sample Matrix Components on Sequencing Analysis Oscar Salas-Solano, Marie C. Ruiz-Martinez,† Emanuel Carrilho,‡ Lev Kotler, and Barry L. Karger*

Barnett Institute and Department of Chemistry, Northeastern University, Boston, Massachusetts 02115

In the previous paper, a sample cleanup procedure for DNA sequencing reaction products was developed, in which template DNA was removed by ultrafiltration and the total concentration of salts (chloride and di- and deoxynucleotides) was decreased below 10 µM using gel filtration. In this paper, a quantitative study of the effects of these sample solution components on the injected amount and separation efficiency of the sequencing fragments in capillary electrophoresis is presented. The presence of chloride and deoxynucleotides in a total concentration above 10 µM in the sample solution significantly decreased the amount of DNA sequencing fragments injected into the capillary column. However, the separation efficiency was not affected upon increasing the amount of salt. On the other hand, in the presence of only 0.1 µg of template in the sample (one-third of the lowest quantity recommended in cycle sequencing) and at very low chloride concentration (∼5 µM), the separation efficiency decreased by 70%, and the injected amount of DNA sequencing fragments was 40% lower compared to the sample cleaned by the new purification method. The deleterious effect of template DNA on the separation of sequencing fragments was suppressed in the presence of salt in a concentration above 100 µM in the sample solution. Separately, it was found that both the electric field strength and duration of injection affected the resolution of DNA sequencing fragments when the cleaned up sample solution was used. Separation efficiencies of 15 × 106 theoretical plates/m were achieved when the sample was loaded at low electric field, e.g., 25 V/cm for 80 s or less. The results demonstrate that the sample solution components (chloride, deoxynucleotides, template DNA) and injection conditions must be controlled to achieve high performance and rugged DNA sequencing analysis.

nents of the sample solution can dramatically influence what and how much is injected by this method.2-7 However, the effect of the sample solution purity on the separation performance of the DNA sequencing fragments is an issue that has not been sufficiently addressed in CE. The small ionic species present in a sequencing reaction mixture, such as di- and deoxynucleotides, and buffer anionic components (e.g., chloride), all affect the amount of DNA sequencing fragments loaded during an electrokinetic injection.8-10 An increased amount of sample injected is necessary to scale down the sequencing reaction preparations and thus reduce the overall cost of DNA sequencing. Besides low-molecular-weight ionic species, the presence of template DNA in the sample is also a factor influencing the DNA sequencing results. Injection of template into the column has been shown to produce instability and reduce resolution in capillary cross-linked gels,11 as well as impair signal in slab gels.12-14 The presence of template DNA could further decrease the amount of

DNA sequencing fragments are introduced into a capillary electrophoresis (CE) column filled with a separation matrix by means of electrokinetic injection.1 It is well-known that compo-

* To whom correspondence should be addressed: (e-mail) bakarger@ lynx.dac.neu.edu. † Department of Energy Human Genome Project Distinguished Postdoctoral Fellow. Current address: Curagen Corp., 322 E. Main St. Brandford, CT 06504. ‡ Current address: University of Sa ˜o Paulo, IQSC/DQFM 13560-970, Sa˜o Carlos-S.P., Brazil. (1) Cohen, A. S.; Najarian, D. R.; Paulus, A.; Guttman, A.; Smith, J. A.; Karger, B. L. Proc. Natl. Acad. Sci. U.S.A. 1988, 85, 9660-9663. (2) Chien, R.-L.; Burgi, D. S. Anal. Chem. 1992, 64, 489A-496A. (3) Williams, P. E.; Marino, M. A.; Delrio, S. A.; Turni, L. A.; Devaney, J. M. J. Chromatogr., A 1994, 680, 525-540. (4) Kleparnik, K.; Garner, M.; Bocek, P. J. Chromatogr., A 1996, 698, 375383. (5) Devaney, J.; Marino, M.; Williams, P.; Weaver, K.; S., D. R.; Turner, K.; Belgrader, P. Appl. Theor. Electrophor. 1996, 6, 11-14. (6) Belgrader, P.; Devaney, J. M.; Delrio, S. A.; Turner, K. A.; Weaver, K. R.; Marino, M. A. J. Chromatogr., B 1996, 683, 109-114. (7) Figeys, D.; Ahmadzedeh, H.; Arriaga, E.; Dovichi, N. J. J. Chromatogr., A 1996, 744, 325-331. (8) Swerdlow, H.; Jones, B. J.; Witter, C. T. Anal. Chem. 1997, 69, 848-855. (9) Schwartz, H. E.; Ulfelder, K.; Sunzeri, F. J.; Busch, M. P.; Brownlee, R. G. J. Chromatogr. 1991, 559, 267-283. (10) Ruiz-Martinez, M. C.; Salas-Solano, O.; Carrilho, E.; Karger, B. L. Anal. Chem. 1998, 70, 1516-1527. (11) Swerdlow, H.; Dew-Jager, K. E.; Brady, K.; Grey, R.; Dovichi, N. J.; Gesteland, R. Electrophoresis 1992, 13, 475-483. (12) Tong, X. C.; Smith, L. M. Anal. Chem. 1992, 64, 2672-2677.

1528 Analytical Chemistry, Vol. 70, No. 8, April 15, 1998

S0003-2700(97)01144-X CCC: $15.00

© 1998 American Chemical Society Published on Web 03/18/1998

DNA sequencing fragments loaded into a capillary column during electrokinetic injection. The preceding paper has described a sample cleanup procedure for DNA sequencing by CE, which results in removal of template DNA and leaves negligible amounts of low-molecularweight ionic components in the sequencing sample solution.10 In this paper, the maximum levels of charged components other than DNA sequencing fragments (e.g., template DNA, chloride, and deoxynucleotides) in a sample have been determined by a systematic study of the effects of these species on the injected amount and separation efficiency. It has also been shown that the deleterious effect of template DNA depends on the salt concentration in the sample. In addition, it was found that the overall quality of sequencing data depends on the injection field strength and time. Optimum injection conditions for purified DNA sample solutions were thus established. EXPERIMENTAL SECTION Experimental conditions are described in the preceding paper (10). RESULTS AND DISCUSSION In electrokinetic injection, DNA sequencing fragments are introduced into the capillary column along with other negatively charged ions present in the sample solution. The measure of the fraction of the total electric current carried by each ionic species in solution during the injection process can be expressed by the transference number T.4,19 Assuming a sequencing sample containing n solutes in free solution, the transference number of any given DNA fragment during the electrokinetic injection from free solution into the capillary filled with polymer solution may be written as

TDNA )

CDNAzDNAµDNA n

(1)

∑C z µ

i i i

i)1

where Ci is the concentration of ion i, zi is the charge on the ith ion, and µi is the free solution mobility of the ith ion. Equation 1 describes the situation just before the ions enter the capillary. Since the capillary contains a sieving matrix, the mobilities of most ions (especially of larger DNA fragments) will change, resulting in a change of the transference number.19 It is also to be noted that chloride ions migrate faster than DNA molecules in free solution, and in the polymer solution, these mobility differences should be even greater. Therefore, the injection process should be biased toward faster migrating small ions and shorter DNA fragments.4 According to eq 1, the amount of DNA sequencing fragments injected could be increased by reducing the concentra(13) Swerdlow, H.; Dewjager, K.; Gesteland, R. F. Biotechniques 1994, 16, 684693. (14) Tong, X. C.; Smith, L. M. DNA Sequence 1993, 4, 151-162. (15) Figeys, D. R., A.; Dovichi, N. J. Electrophoresis 1994, 15, 1512-1517. (16) Ruiz-Martinez, M. C.; Berka, J.; Belenkii, A.; Foret, F.; Miller, A. W.; Karger, B. L. Anal. Chem. 1993, 65, 2851-2858. (17) Carrilho, E.; Ruiz-Martinez, M. C.; Berka, J.; Smirnov, I.; Goetzinger, W.; Miller, A. W.; Brady, D.; Karger, B. L. Anal. Chem. 1996, 68, 3305-3313. (18) Goetzinger, W.; Karger, B. L. Int. Patent WO 9623220, 1996. (19) Spencer, M.; Kirk, J. M. Electrophoresis 1983, 4, 46-52.

tion of other ionic components of the samples. These components include chloride, di- and deoxynucleotides, phosphates, template DNA, and excess primer, which are all present in the sample solution. In the current study, the effects of salt concentration (e.g., chloride and deoxynucleotides) and template DNA on the injected amount and separation efficiency of sequencing fragments were quantitatively determined. Furthermore, the role of injection field strength and time on the separation efficiency of the purified sequencing samples in a solution of low ionic strength have been examined. Effect of Chloride Concentration. As already noted, the presence of ionic impurities affects the amount of DNA loaded into the capillary column.5-10 To quantify the role of the salt concentration in the samples on the electrokinetic injection of DNA sequencing fragments, known amounts of sodium chloride were added back to dye-primer cycle sequencing samples purified with the cleanup protocol. Seven sequencing reactions were prepared with AmpliTaq-FS DNA polymerase using labeled primers on a ssM13mp18 template. After the reactions were complete, all 28 single-color mixtures were pooled together and then split into seven identical aliquots of 50 µL each. The template was removed using the poly(ether sulfone) membranes deactivated with LPA (700-1000 kDa). Next, the samples were desalted using two prewashed Centri-Sep columns per sample. After the purification, the samples were again mixed together, and the final chloride concentration in the mixture was ∼5 µM, as determined using the CE-indirect UV method.10 The mixed sample was again split in seven aliquots of 50 µL each to maintain the amount of DNA fragments constant. Six aliquots were dried under vacuum and then dissolved in 50 µL of various NaCl solutions to obtain final chloride concentrations of 10, 25, 50, 95, 200, and 300 µM. These levels of salt content covered the typical range of chloride concentration found in DNA samples purified by conventional methods.10 The seventh aliquot, which was not dried, represented a typical sample purified with the cleanup protocol (∼5 µM chloride and no template) and was used as a control. This latter solution resulted in the highest CDNA/Ci ratio and hence gave the highest DNA transference number of the DNA sequencing fragments (eq 1). For the CE analysis, an aliquot of 10 µL from each sample was taken and diluted to 25 µL of H2O. As discussed in the previous paper, the purpose of dilution was to facilitate the injection to the column, but no change in the DNA transference number resulted. For the analysis, the peak area and number of theoretical plates of the DNA fragments 78, 289, and 842 bases long were measured and normalized to the corresponding value obtained with the control sample. Figure 1 shows the effect of increasing the chloride concentration in the sample solution on the injected amount and separation efficiency on the fragment 289 bases long. Similar trends were observed for other DNA fragments, as measured for those of 78 and 842 bases long. At 10 µM chloride in the sample solution, the injected amount of DNA fragments was 10% lower than the control. By increasing the chloride concentration in the sample solution from ∼5 to 25 µM, which is only in the lowest range of chloride concentration found after ethanol precipitation of sequencing samples, the injected amount of the DNA fragments Analytical Chemistry, Vol. 70, No. 8, April 15, 1998

1529

Figure 1. Normalized peak area (9) and number of theoretical plates (b) vs sample chloride concentration of dye-primer cycle sequencing sample made with M13mp18 template. All peak areas and number of theoretical plates of the base 289 were normalized to the values obtained with the cleanup sample: chloride concentration of 5.0 µM and no template. Error bars are shown at the 95% confidence interval (n ) 3). The electropherograms of two samples corresponding to the C-terminated sequence of M13mp18 at 5 and 200 µM, respectively, are shown in the inset. Electrophoretic conditions: capillary effective length 15 cm, total length 25 cm; 75-µm-i.d., 365-µm-o.d. coated capillary (poly(vinyl alcohol)); polymer solution, 2% (w/w) LPA (MW >5.5 × 106 MDa); running buffer Tris-TAPS-EDTA. The samples were injected at a constant voltage of 100 V/cm for 20 s, and electrophoresis was performed at 200 V/cm at room temperature.

decreased by 30%. This result makes clear that the chloride concentration in the sample solution should be below 10 µM for maximum amount injected. The DNA transference number depends not only on the ionic strength but also on the concentration of DNA in the sample solution (eq 1). The observed level of the sequencing fragments injected at higher chloride concentration could be even lower if less DNA were present in the sample solution. It is known that in the case of ethanol precipitation there is a possibility of low DNA recovery.20,21 Indeed, the high variability in chloride concentration after the ethanol precipitation step, typically 20230 µM, could lead to a lack of signal for the DNA fragments 1530 Analytical Chemistry, Vol. 70, No. 8, April 15, 1998

under a fixed set of injection conditions. Moreover, the yield of the fragments from a Sanger reaction may be sequence dependent. Thus, less than maximum injection could decrease the ruggedness of automated DNA sequencing in a multiple capillary array or microchip instrument. Upon further addition of chloride, a gradual decrease in injected amount of sequencing fragments was observed (Figure 1). At a level of 200 µM of chloride, the peak area decreased by more than 50% and at a concentration of 300 µM by 80%, compared to the control sample with ∼5 µM chloride. These results were (20) Talbot, D. Amicon Appl. Note 1996, 387, 11-15. (21) Hilderman, D.; Muller, D. Biotechniques 1997, 22, 878-879.

predictable, since upon increasing the salt concentration, DNA sequencing fragments become a smaller and smaller fraction of the total ionic strength of the sample solution. Interestingly, in the presence of substantial amount of chloride in the sample solution, the separation efficiency of the fragments did not change compared to the control sample (see Figure 1). As an example, the inset in this figure shows that the separation of the C-terminated fragments 101-125 bases long was not affected by the almost 40-fold increase of the amount of chloride in the sample solution. A similar study to determine the influence of the deoxynucleotides on DNA sequencing data was performed. Different amounts of 2′-deoxyadenosine 5′-triphosphate disodium salt (dATP) were added to the purified sequencing samples. Upon increase of the dATP concentration in the sample solution, a sharp drop in signal of the sequencing fragments with little or no change in separation efficiency was observed. The 10% decrease in signal was already reached at 5 µM dATP. Higher concentrations of dATP caused substantial losses in signal. For example, at 40 µM, the signal was only 45% of the control, and it was less than 20% at 100 µM dATP in the sample solution. Therefore, the deoxynucleotides in the sample solution appeared to cause higher loss in signal, i.e., decrease in the amount of DNA injected, compared to chloride. This may be a result of 4-fold higher negative charge carried by deoxynucleotide ions. However, the effect on signal intensity caused by deoxynucleotides was less than the factor of 4. Effect of Template DNA. At present there is no universal agreement on the role of template DNA on the electrophoretic separation of sequencing reaction products by CE.10 Some have suggested that the template reduces gel stability and resolution of the sequencing fragments.7,11,22 However, traditional schemes of sample purification such as ethanol precipitation or gel filtration which do not include the removal of template from the sample solution are still commonly used.22-26 To determine the influence of the template in the sample solution, different amounts (up to 3.0 µg) of circular ssM13mp18 DNA were added back to a sample purified by the cleanup protocol. An aliquot of 10 µL from each sample was taken and diluted to 25 µL with distilled water. As shown in Figure 2, a dramatic decrease in both the peak area and the separation efficiency of the sequencing fragments was observed when the amount of template in the low ionic strength sample solution was increased. In the presence of only 0.1 µg of template, which is one-third of the lowest recommended quantity for cycle sequencing, the number of theoretical plates decreased by ∼70%, and the amount of DNA sequencing fragments injected into the column was 40% lower compared to the control. The effect of adding 0.1 µg of M13mp18 to a templatefree sample is further shown in the inset. The fluorescence signal and separation efficiency of the C-terminated fragments in the range of 101-125 bases were significantly decreased. It was also observed that the migration times increased by more than 5 min, (22) Swerdlow, H.; Jones, B. J.; Wittwer, C. T. Anal. Chem. 1997, 69, 848-855. (23) Fang, Y.; Zhang, J. Z.; Hou, J. Y.; Lu, H.; Dovichi, N. J. Electrophoresis 1996, 17, 1436-1442. (24) Bashkin, J.; Marsh, M.; Barker, D.; Johnston, R. Appl. Theor. Electrophor. 1996, 6, 23-28. (25) Wu, C. H.; Quesada, M. A.; Schneider, D. K.; Farinato, R.; Studier, F. W.; Chu, B. Electrophoresis 1996, 17, 1103-1109. (26) Tan, H.; Yeung, E. S. Anal. Chem. 1997, 69, 664-674.

which was likely related to the sudden drop of current from 10 to 2 µA during the separation process. These deleterious effects were more pronounced when the amount of the template DNA in the sample solution was further increased, as seen in Figure 2. Indeed, when 1.2 µg of template was added (the highest quantity used in cycle sequencing per sample), the signal from the sequencing fragments was only 10% of the control sample. These results demonstrated that there was a clearly harmful effect of template DNA on the injected amount and separation of the sequencing fragments in sample solutions with a low salt content. The effect of the template is probably due to the fact that, after minimizing the presence of small ions, more DNA template is injected. It can be assumed that this template, by virtue of its low electrophoretic mobility in polyacrylamide solutions,27 creates a zone of high electrical resistance at the head of the column. This region could cause the current to decrease in the electrophoretic system, increasing local heating11 and leading to a disturbance in the electric field. Furthermore, this local heating could alter the polymer solution and decrease the resolution of the sequencing fragments. The amount of fragments injected into the head of the column could also decrease, as shown in Figure 2. Additionally, concentrated template DNA could precipitate in the head of the capillary column, clogging the polyacrylamide network, and therefore contributing to the problems cited above. It is important to note that bubble formation and declining current, which caused instability of capillary gels11,28 and also in polymer solutions,7 have also been prevented with the new sample cleanup protocol. This is a significant factor in the design of a robust, reproducible, and automatable DNA sequencing methodology. Effect of Chloride Concentration in the Presence of Template. In previous studies, a read length of more than 1000 bases using replaceable LPA solutions in CE was achieved with cycle sequencing samples purified by ethanol precipitation and then dissolved in water.17 In that work, the template was not removed, and yet the deleterious effects on the separation of the DNA reaction products were not observed, in contrast to the results in Figure 2. The chloride concentration in a sequencing sample after ethanol precipitation could be as high as 230 µM.10 It would thus appear that there is a relationship between the salt concentration in the sample solution and the influence of template DNA on the separation. To study this relationship, six four-color dye-primer cycle sequencing samples with 0.3 µg of M13mp18/sample were prepared, and then all 24 reactions were mixed and separated into six aliquots of 50 µL each to maintain the amount of DNA constant. One sample out of the six was purified with the cleanup protocol, and the other five samples were simply desalted without template removal. The chloride concentration of each of the five samples was determined to be ∼5 µM. Four of these five aliquots were vacuum-dried and then dissolved in NaCl solutions to obtain final chloride concentrations of 25, 50, 95, and 200 µM. The final volume of each of these sample solutions was also 50 µL. A fifth aliquot that was not dried was used as a control. As indicated above, 10 µL of each sample was taken for the electrophoresis analysis. (27) Garner, M. M.; Chrambach, A. Electrophoresis 1992, 13, 176-178. (28) Karger, A. E. Electrophoresis 1996, 17, 144-151.

Analytical Chemistry, Vol. 70, No. 8, April 15, 1998

1531

Figure 2. Normalized peak area (9) and number of theoretical plates (b) vs the amount of template (M13mp18) added to a purified sample. The peak areas and number of theoretical plates for base 289 were normalized to the values obtained with the cleaned-up sample: chloride concentration of 5 µM and no template. Error bars are shown at the 95% confidence interval (n ) 3). The electropherograms of two samples corresponding to the C-terminated sequence of M13mp18 are shown in the inset. See Figure 1 for experimental conditions.

Figure 3 shows the effect of the template on DNA sequencing analysis as a function of the salt concentration in the sample. In agreement with the results in Figure 2, the separation efficiency was only ∼15% of the control. Interestingly, the efficiency increased with the increase in salt content. At 50 µM chloride in the sample solution, the number of theoretical plates increased to 70% of the control. Further increases in the salt content to higher than 100 µM resulted in the number of theoretical plates increasing to become similar to the template-free sequencing sample. The electropherograms in the inset clearly show the gain in separation at higher salt content when the template is present. From the above results, it is clear that the content of chloride in the sequencing samples at the level of ∼100 µM or above suppresses the deleterious effects of the template DNA on separation. This result is probably due to chloride ions that are 1532 Analytical Chemistry, Vol. 70, No. 8, April 15, 1998

preferentially injected into the capillary column, thus reducing the amount of template DNA injected into the column. Figure 3 may provide an explanation as to why some approaches do not include template DNA removal from the sample solution.8,22-26 In these methods, the samples were purified using ethanol precipitation or gel filtration columns without prewashing of the packing. Such methods leave higher chloride content (25230 µM) in the sample solution.10 Moreover, in some cases, the samples were then dissolved in a formamide-0.5 M EDTA (49: 1) solution, which further increases the ionic strength of the sample solution. Therefore, harmful effects of the template on the separation efficiency in such cases were likely minimized. However, an increase in ionic strength of the sample solution caused a significant decrease in the amount of DNA fragments loaded (Figure 3). Comparison of the plot of normalized peak

Figure 3. Normalized peak area (9) and number of theoretical plates (b) vs chloride concentration of dye-primer cycle sequencing sample containing 0.3 µg of M13mp18 template. The peak areas and number of theoretical plates for the base 289 were normalized to the values obtained with the purified sample: chloride concentration of ∼5 µM and no template. Error bars are shown at the 95% confidence interval (n ) 3). The electropherograms of two samples corresponding to the C-terminated sequence of M13mp18 are shown in the inset. See Figure 1 for experimental conditions.

area in Figure 1 to that in Figure 3 reveals that the influence of increasing chloride concentration in the sample solution is more pronounced when the template is present in the sample. These results indicate that removal of template DNA and the maintenance of low chloride concentration (∼5 µM) in the sample is important, as a significantly higher amount of the DNA sequencing fragments can be injected without compromising separation efficiency. Optimization of Electric Field Strength and Time of Injection. It has been reported previously that the CE separation of oligonucleotides dissolved in water or other low ionic strength solvents in cross-linked gels is affected by the injection field strength and duration.29,30 It was therefore decided to examine in more detail the effect of the electric field strength and injection time on the separation of sequencing fragments purified with the

developed purification protocol using LPA polymer solutions. Table 1 presents the results of experiments in which the number of injection coulombs was held constant (∼60 µA‚s) while the electric field strength was varied. Upon decreasing the electric field from 400 to 25 V/cm. the separation efficiency increased almost 10-fold. Below 25 V/cm, changes in the efficiency were negligible. It is important to note that the separation efficiency reaches as high as 15 × 106 plates/m for the DNA fragments in sequencing samples, among the highest values reported to date. The effect of injection field strength on the separation of the sequencing fragments is shown in Figure 4. After decreasing the field from 200 to 25 V/cm, the multiple peaks corresponding to the fragments 101-125 bases long were almost baseline resolved. (29) Paulus, A.; Ohms, J. L. J. Chromatogr. 1990, 507, 113-123. (30) Demorest, D.; Dubrow, R. J. Chromatogr. 1991, 559, 43-56.

Analytical Chemistry, Vol. 70, No. 8, April 15, 1998

1533

Figure 4. Effect of injection field strength on the separation of DNA sequencing fragments using replaceable linear polyacrylamide solutions. The electropherograms correspond to the C-terminated sequence of M13mp18. (A) Electrokinetic injection for 10 s at 200 V/cm, (B) 20 s at 100 V/cm, and (C) 80 s at 25 V/cm. See Figure 1 for other experimental conditions.

In addition, injecting the sample at 200 V/cm caused severe distortion (tailing) of the sequencing fragment peaks. The doublets in the region of 101-102 and 168-169 bases were clearly observed with the sample being injected at 25 V/cm, given that the separation efficiency was 5 times higher. The improvement in the separation efficiency is likely related to the fact that the electrokinetic injection of a low ionic strength sample into a column with a high ionic strength buffer results in a much higher electric field at the injection point than in the column.2 When the samples were injected under strong electric fields (e.g., 200-400 V/cm), the velocity of the DNA sequencing fragments, and therefore the electrokinetic force at which these molecules enter the polymer solution, may be higher than the resisting force of the polymer. The interaction of DNA fragments with the polymer strands could deform the linear polyacrylamide structure, leading to a loss of resolution.31 Moreover, Joule 1534 Analytical Chemistry, Vol. 70, No. 8, April 15, 1998

heating would be most severe at the capillary tip under high injection fields.32 This heating could also increase the risk of column electrical breakdown by the formation of bubbles in the capillary tip and failure of the sequencing run. To maximize the amount of DNA fragments injected into the column without compromising the high separation performance, the effect of injection time on the separation efficiency was next investigated, using a fixed low injection field. Several aliquots of the same purified sample were injected at 25 V/cm for different time intervals, and the results are presented in Table 2. The separation efficiency of the sequencing fragments was almost identical when the duration of the electrokinetic injection was increased from 40 to 80 s. After 80 s, a decrease in the separation (31) Kozulic, B. Anal. Biochem. 1995, 231, 1-12. (32) Best, N.; Arriaga, E.; Chen, D. Y.; Dovichi, N. J. Anal. Chem. 1994, 66, 4063-4067.

Table 1. Separation Efficiency of Sequencing Fragments as a Function of the Injection Field Strengtha injection time (s)

electric field (V/cm)

injection current (µA)

peak area (mV‚s)

efficiencyb (×106 plates/m)

5 10 20 40 80 160

400 200 100 50 25 12

12 6 3 1.5 0.7 0.3

61 57 55 58 56 59

2 3 8 11 15 14

a The peak area and the number of theoretical plates are for the fragment 289 bases long. Constant injection coulombs, 60 s‚µA. See Figure 1 for other experimental conditions. b Analysis was performed in triplicate for each experimental condition, and the resulting RSDs for peak area and theoretical plates were 9 and 5%, respectively.

Table 2. Separation Efficiency of the Sequencing Fragments as a Function of the Injection Time of the Sequencing Samplesa injection time (s)

coulombs (s‚µA)

peak areab (mV‚s)

efficiencyb (×106 plates/m)

40 50 60 80 100 120 160

28 35 42 56 70 84 112

30 42 55 60 78 83 102

16 15 15 16 11 6 3

a Injection at constant electric field of 25 V/cm (0.7 µA). The peak areas, theoretical plates numbers, and peak widths are for the fragment 289 bases long. Experimental conditions were as in Figure 1. b Analysis was performed in triplicate for each experimental condition, and the resulting RSDs for peak area and theoretical plates were 9 and 6%, respectively.

efficiency was observed, leading to a ∼5-fold loss when injection time was 160 s. This increase in the peak width with loading time was expected, because more DNA fragments were injected into the capillary column and driven through the separation medium before all the DNA was loaded. For optimum separation efficiency and good signal under these column conditions, injection of the sample for 80 s or less at 25 V/cm would appear optimum. CONCLUSIONS The development of capillary electrophoresis for highthroughput DNA sequencing as a fast and economical alternative to slab gel electrophoresis is an important goal. This process requires rugged operation to provide for the highest throughput. The new DNA sequencing sample purification protocol for CE developed in this work should contribute to these goals. The quantitative studies in this paper have demonstrated that reducing salt concentration below 10 µM in total, along with DNA

template removal from the sequencing sample, is an essential step to significantly increase the amount of DNA sequencing fragments injected into the capillary column. In addition to the improvement of sequencing results, maintaining the low level of salts in the sample solution could lead to a reduction of the amounts and costs of reagents used in the sequencing reactions. The effects of the template on DNA sequencing by CE have been clarified, and it has been demonstrated that template removal from the sample solution also significantly increases the separation efficiency of the sequencing reaction products. Furthermore, template removal decreases column failures such as spikes due to bubble formation and declining current. Conditions for electrokinetic injection of the purified sample have been established, further optimizing the sequencing methodology. DNA sequencing samples purified with the new cleanup protocol feature constant chemical composition, which is a very important factor to consider in the reproducible and automatic operation of multiple capillary arrays instrumentation. Further optimization of the separation of the polymer solution has resulted in the electrophoretic separation of more than 1000 DNA sequencing fragments on M13mp18 in less than 1 h, with a base-calling accuracy of over 99%.33 In addition, improvements have been made in the preparation of the polymer solution matrix in which LPA (MW >9 × 1012) has been produced as a white powder using emulsion polymerization methods.34 As noted in the preceding paper, the principles developed here should also prove useful to microchip array electrophoresis.35 Work to adapt the cleanup method to a robotics station for effective automation is currently underway. ACKNOWLEDGMENT The authors acknowledge the DOE under the Human Genome Project Grant DE-FGO2-90ER 60985 and NIH grant under Project 5R 01 HG01413-02 for support of this work. Support by DOE does not constitute an endorsement of the views expressed in this article. M.C.R.M. thanks the Department of Energy under the Human Genome Project Distinguished Postdoctoral Fellowship Program for their financial support. The authors also thank Dr. Arthur W. Miller for the development of the base-calling software and Marek Minarik for development of the buffer system for the CE indirect-UV method. This is contribution 701 from the Barnett Institute. Received for review October 16, 1997. Accepted January 29, 1998. AC9711448 (33) Salas-Solano, O.; Carrilho, E.; Kotler, L.; Sosic, Z.; Miller, A.; Goetzinger, W.; Karger, B. L., to be published. (34) Goetzinger, W.; Kotler, L.; Carrilho, E.; Ruiz-Martinez, M. C.; Salas-Solano, O.; Karger, B. L. Electrophoresis, in press. (35) Woolley, A. T.; Mathies, R. A. Anal. Chem. 1995, 67, 3676-3680.

Analytical Chemistry, Vol. 70, No. 8, April 15, 1998

1535