Created: 1st September 1999, last updated: 12th November 1999, © 1999 ABRF
Elaine R. Mardis
Technology Development, Genome Sequencing Center, Washington University School of Medicine, St. Louis, MO
Commercial availability of multicapillary DNA sequencing instruments can significantly impact the existing paradigm in high-throughput DNA sequencing facilities, shifting the rate-limiting step away from the electrophoresis and detection step. These instruments, although novel and poorly understood, represent a significant capital investment. This review informs the reader about the state-of-the-art instruments, although capillary sequencers are in a state of almost constant flux, with improvements emerging rapidly. (J Biomol Tech 1999;10:137-143)
Key words: multicapillary electrophoresis; automated DNA sequencing.
Address correspondence and reprint requests to Elaine R. Mardis, Director, Technology Development, Genome Sequencing Center, Washington University School of Medicine, 4444 Forest Park Boulevard, St. Louis, MO 63108 (email: emardis@watson.wustl.edu).
High-throughput DNA sequencing laboratories are poised on the brink of a paradigm shift in their workflow and operations. The commercial availability of multicapillary sequencing instruments signals that change is imminent and that a careful evaluation of these platforms, their capabilities, and their limitations is in order. This review considers two commercially available instruments, the MegaBACE 1000 DNA Sequencing System from Amersham Pharmacia/Molecular Dynamics (Sunnyvale, CA) and the 3700 DNA Analyzer from Perkin-Elmer Biosystems (Foster City, CA). Both offer simultaneous electrophoresis and detection of samples on 96 capillaries. The MegaBACE has been available for the past year and the 3700 since January of 1999. These instruments are relatively expensive, with the MegaBACE retailing at $200,000 and the 3700 at $300,000. As with any instrument, there are associated consumables, including capillary arrays, separation matrix, buffer, and sequencing reagents. The software for data collection and analysis on both instruments operates in a Microsoft Windows NT environment.
Ideally, capillary instruments offer several significant advantages over slab gel scanners. The most obvious advantage of multicapillary systems is the elimination of gel plate assembly and gel casting, because a flowable separation matrix is automatically injected into each capillary just before loading new samples. Some care must be taken to ensure that air bubbles are not introduced into the matrix, because their presence in a capillary causes noise from laser light diffraction and instability in the electrophoretic current. Loading samples onto capillaries also occurs in a straightforward manner that is much less error prone than manual application of samples into slab gel wells, especially at 96-lane density. Sample ions load onto a capillary by a process called electrokinetic injection. A high field voltage is applied between the sample and the anode end of the capillary, causing DNA ions to migrate into the capillary.
A major advantage of capillary instruments is the elimination of postrun lane retracking. This is a time-consuming and error-prone step in slab gel analysis, which scales up with increasing lane number. Capillaries do not require retracking because data collection occurs at previously determined spatial locations corresponding to each capillary in the array. A natural consequence of the elimination of retracking is that the time-consuming transfer of gel image files for off-line retracking and analysis is no longer necessary. Only 96 sample files, each around 250 kilobytes, are transferred to a host server or computer for analysis and archiving. This advantage immediately appeals to centers such as ours, where tight instrument loading schedules mean that gel image files from 100 or more sequencers must be transferred across busy local area networks at specific times. Comparatively speaking, capillary instruments require much less hands-on time from technicians than slab gel scanners, although to varying degrees. The PE 3700 allows users to place up to four 96-well microtiter plates on the workspace at one time and then performs the loading, electrophoresis, detection, and data analysis for each plate in succession without user intervention. The MegaBACE requires 10 to 15 minutes of user intervention for sample loading from each 96-well plate, after which electrophoresis and detection are performed automatically.
One commonly cited advantage of electrophoresis in capillaries over glass plates is the ability to run high field voltages because of the greater heat dissipation capability from capillaries. In practice, however, our electrophoresis duration is roughly equivalent when comparing the 3700 with the shortest run module on the ABI 377 sequencer (3.3 versus 3.5 hours) and is about 1 hour shorter on the MegaBACE (2.3 hours, a 34% improvement) than on the 377. These times primarily reflect changes that we have made to the manufacturer's recommendations to increase readlength, which is a factor of predominant importance in genomic sequencing projects. We hope future improvements to run matrices and run voltage profiles will allow significantly shorter run times.
One fundamental difference between these platforms is the detection schema used by each instrument. The MegaBACE is essentially a commercial version of a multicapillary sequencer devised by Richard Mathies1,2 and uses his concept of scanning the capillary array at a fixed point with a confocal microscope objective to deliver focused laser light and to collect the fluorescent emissions of laser-excited fragments. In this system, it is imperative that the laser beam is focused at the exact center of each capillary to provide maximum power to the sample bands and that the confocal objective collects emissions from these center pixel positions as well. This precise focusing optimizes sensitivity and signal to noise. In production, we routinely test each instrument for focus every 2 weeks, making adjustments if necessary. Photomultiplier tubes (PMTs) detect the bandpass-filtered fluorescent emissions from sequencing fragment bands.
Unlike the MegaBACE, the 3700 uses a sheath-flow detector system, including a charge-coupled device (CCD) camera. This instrument is a commercial adaptation of work from Kambara's laboratory in Japan.3 In sheath-flow detection, the sequencing fragment bands are excited and detected as they elute from the anode end of the capillary into the sheath-flow cuvette. The emerging fragment forms a plume that is swept off in a sheath flow of polymer through the laser beam path. The resulting emissions are collected by a spectrograph that uses a reflection grating to disperse the light into its component wavelengths and passes that spectrum onto the surface of the CCD camera. The CCD camera converts the resulting information into digital data that are collected for the run. Because the laser beam enters from the right side of the array, a person can easily detect by eye when the beam is out of alignment with the sheath flow, because the samples at the left of the array exhibit markedly weaker signal than those at the right. Correcting the laser beam entry path on the 3700 requires a trained service technician, whereas an experienced laboratory technician can refocus the MegaBACE optics.
Both capillary platforms offer sample loading by electrokinetic injection, although this is manifested differently in each instrument. In general, electrokinetic injection is a sensitive process that is influenced by the ionic strength of a sample, the amount of template DNA in the sample, and the geometry of the electrode and capillary tip within the sample. The process of electrokinetic injection involves the transfer of charged ions in an electrical field onto the capillary separation matrix. Because only ions transfer in this process, no liquid volume loss occurs from the sample. Ideally, this means that samples can be reloaded if a run is aborted or lost for other reasons.
The 3700 loading scheme allows the instrument to achieve unattended operation for analysis of up to four 96-well plates (or one 384-well plate) of reactions. The 3700 has a two-tip gantry-mounted autoloader that transfers 2.5-mL sample aliquots from microtiter plate wells into a multiwell loading bar. Each of the injection ends of the capillary array project into each of the wells on the loading bar, and electrokinetic injection takes place when a voltage is applied to the metal loading bar (acting as the cathode). After injection, the wells of the loading bar are washed out and filled with running buffer, and separation by electrophoresis and data collection begin. Because of the sample transfer process described, the sample aliquots transferred are lost after electrokinetic injection, which means that reloading is limited to the available sample volume remaining in the source plate. Another potential source of volume loss from plates queued on the 3700 workspace for loading is evaporation.
The amount of evaporative volume loss can be reduced in a variety of ways. First, we have found that turning off the overhead workspace light source in the 3700 reduces evaporation. Second, plates can be covered with a thin adhesive foil tape after resuspension of samples, and the worker can employ the foil-piercing hardware option available from Perkin-Elmer. With this option, the 3700 foil-piercing tips selectively pierce the foil tape over each well immediately before loading to allow pipette tip access for sample transfer to the load bar. This option appears to be robust for 96-well plates in limited testing, although it adds time to the overall workflow.
The MegaBACE requires attended operation to load each plate individually. After placing the sample plate onto the cathode stage and closing the stage door, the plate is raised up to the capillary tips. There is a thin platinum electrode positioned alongside each tip. The injection voltage is applied at these electrodes, and the samples migrate onto the capillary. The plate is replaced by a buffer-filled reservoir during electrophoresis and detection, so the sample plate can be immediately returned to a -20°C freezer and reloaded, if necessary, without sample degradation worries. The loading process requires about 10 minutes to complete.
In capillary systems, an injectable matrix functions to resolve sequencing fragments in the applied electrical field. Predictably, the matrix largely determines the potential resolution limit of the sequencing fragments. It also dictates instrument design and workflow. The currently used matrix for the MegaBACE is linear polyacrylamide (LPA) and for the 3700 is Performance Optimized Polymer 6 (POP6). Each has very different physical characteristics--LPA is extremely viscous, and POP6 has the consistency of jelly--that directly influence their resolution limit. These matrices also differ in their polymer pore size, a characteristic that may underlie the greater tendency of LPA to clog with template DNA during electrokinetic injection.
LPA is packaged into individual 2-mL tubes and is injected into capillaries from within a titanium pressure hull under nitrogen gas pressure. This injection requires that a high-pressure nitrogen tank be attached to the instrument. It also requires a "relaxation" period before sample injection and electrophoresis so that the polymer can retain its native configuration. Overall, the time required for matrix injection, relaxation, and sample loading is about 50 minutes (30 minutes for the hands-off process of matrix relaxation and prerun).
On the 3700, the POP6 matrix is supplied in a 200-mL bottle that nests inside the instrument door. A supply line inside the bottle of POP6 is used to supply a syringe pump that first fills the sheath flow cuvette and then injects POP6 into the capillaries under applied syringe pressure. No relaxation time follows this low-pressure injection, but subsequent flush of the cuvette and loading bar wells to clear excess matrix is required. The overall processing time between runs for these functions to complete is 51 minutes.
Predictably, Molecular Dynamics and Perkin-Elmer advocate the use of their proprietary labeling chemistries and enzymes for their capillary instruments. Amersham has developed specific sequencing reaction kits for the MegaBACE, with altered deoxynucleotide to dideoxynucleotide ratios to provide improved fragment length distributions from electrokinetic injection. Perkin-Elmer initially supported the use of their existing Big Dye terminator kit, although it appears that a similar deoxynucleotide-dideoxynucleotide optimization may be required.
In general, the Amersham kits are based on the Dyenamic Energy Transfer and ThermoSequenase DNA polymerase technologies. These include a Dyenamic ET primer kit and a Dyenamic ET terminator kit. Because this instrument has an operating temperature of 44°C, compressions in dye primer reactions occur. To address this issue, Amersham introduced a dITP-containing Dyenamic ET primer kit. Responsive to the urging of many high-throughput sequencing facilities, Molecular Dynamics now offers a set of bandpass filters that allow the user to detect and analyze the Big Dye labeling chemistry from Perkin-Elmer.
Perkin-Elmer has, for the near term, focused their development efforts on Big Dye terminator chemistry, although it is also possible to detect and analyze Big Dye primer reactions on the 3700. This limited capability primarily reflects the relatively short time that the instrument has been commercially available, and it is anticipated that additional functionality will be introduced over time.
Both instruments offer software that runs in a Windows NT environment. We have installed the version 1.0 software for each instrument. To improve user interaction in a production environment, we created custom interfaces to facilitate sample sheet import from and data export to our Unix databases. The following observations largely reflect our experiences with and opinions about both software packages.
Although NT-based software represents a radical shift from their tradition of Macintosh-based software, the transition seems smooth from a user standpoint. This is mainly because many of the user interface graphics and features of data collection and analysis from the ABI 377 sequencer have been maintained, including the concepts of editable Run Modules, run status monitoring windows, log files, and spreadsheet-based analysis queuing. This is not to say that the 3700 is without its differences; the main one is that the data collection and analysis software interact directly with an Oracle-based instrument database that is resident on the host computer. The Oracle database stores information about scheduled runs, including the run schedule and plate records, run logs and error logs, preference settings for the Data Collection software, and processed, unanalyzed fluorescence data that have been collected from the CCD. The database is integral to the instrument's operation and must be functional for the instrument to function. Other new features of the software include the Data Extractor program that assembles all relevant information to create analyzed sample files, communicating the data to the AutoAnalyzer program that performs the data analysis and basecalling. Presentation of trace data from individual sample files is virtually identical to the Mac-based software, with the ability to toggle between run information, textual sequence, raw data, and processed data extant.
The data collection and analysis software for the MegaBACE pose challenges, perhaps more so for the experienced PE software user than for the neophyte. Data collection occurs through the Instrument Control Manager (ICM), which guides the user through a predetermined workflow. Preparing the instrument for a run involves stepping through the workflow, during which there are instructions presented to the user at one of two separate liquid crystal diodes (LCDs) on the instrument. These LCDs flank the two stages---cathode and anode--that are accessed through doors at the front of the instrument as a means of supplying necessary items such as tubes of LPA for injection and samples for loading. The run setup requires about 10 to 15 minutes of user interaction time before electrophoresis and data collection are underway. Programs are available for monitoring the run, including a Run Image view that provides display information for each capillary at the top half and displays the electrophoretic profile of four color data from any selected individual capillary on the bottom half. Run status is also a part of this display, including time elapsed, PMT voltages, and run temperature. Another useful display is the Current Monitor window, which displays the current values for each capillary throughout the run. This display is useful because it can give the user a quick view of capillaries that are demonstrating significantly reduced current relative to others. These capillaries may turn out to be routinely problematic and require replacement. Alternatively, reduced current may reflect a clogging event where excessive template has interfered with loading of sequencing fragments onto the matrix.
Analysis software for MegaBACE reads uses a spreadsheet interface into which the run data from individual or multiple runs are entered and available for analysis. The user selects a "basecaller" (a data-processing algorithm), and then the input data from individual runs are analyzed. Electropherograms of the resulting data can be viewed by highlighting individual samples and requesting the electropherogram view from a pull-down menu or by selecting all samples in the spreadsheet and viewing the traces in groups of four.
In a production setting, the MegaBACE software has posed problems for technicians in our laboratory because the user interface does not allow straightforward import of the sample sheet or run parameter selection. The postrun data analysis should be automated and not require the manual procedures described. On a positive note, the manufacturer is working aggressively to address these issues and to provide a more tractable user interface. Molecular Dynamics has made this version 2.0 available to beta test sites, including our laboratory.
Two facts regarding our experiences with the MegaBACE and the PE 3700 are important. First, we have had access to MegaBACE sequencers for 2 years and to the 3700 for 5 months. As such, our recent experiences with the 3700 in some ways reflect struggles that we had with the initial MegaBACE units of 2 years past. Second, we address new instrumentation in light of our current workflow and protocols, trying to deviate as little as possible from proven processes and not always considering optional approaches that might be appropriate in another setting. In other words, what is right for our high-throughput sequencing laboratory will not necessarily be right for another and almost certainly will not be appropriate for a lower-throughput or core facility.
Both types of capillary instruments have been run at our facility in mock production mode, involving maximizing the throughput per machine per day. We are only now poised to introduce them into our genomic and EST production groups. Running mock production allows us to assess instrument performance at maximum throughput, thereby determining whether there are hardware stability problems. It allows us to gauge whether the preps and sequencing protocols and the run parameters we have optimized are sufficiently robust. Mock production generates feedback from a small user group on what additional hardware and software pieces of the puzzle are still required for straightforward implementation.
Table 1 provides a summary of the modifications we made to existing protocols and procedures associated with each instrument relative to their manufacturer's recommendations. In most cases, these modifications were meant to address shortcomings in the performance of each instrument. In the case of the 3700, we were primarily concerned with short readlengths caused by the low viscosity of the POP6 matrix and to the detection bias of shorter fragment lengths resulting from the mismatch of sheath flow velocity to electrophoretic velocity. For the MegaBACE, we were concerned with the relatively high percentage of overinjections experienced with the LPA matrix when excess DNA is injected, impeding the flow of current through the capillary and prohibiting bands from reaching the detector. We also developed the concept of "preinjection" of samples, for which the sample plate is mock injected onto matrix from the previous run to reduce the number of small ions that out-compete DNA sequencing fragments for injection. Subsequently, the samples are injected a second time onto freshly injected matrix, electrophoresed, and detected as usual. In general, we favor ethanol precipitation of reaction products over other cleanup methods for its ease of performance in a 96-well format and because it is cheap, technically simple, and easily performed in a high-throughput production setting. Specifically, we precipitate with 100% ethanol in the presence of ammonium or sodium acetate and use two 70% washes to help eliminate excess salt ions that can interfere with electrokinetic injection and to reduce the intensity of free dye terminator artifacts. Column cleanup methods have been investigated but have proven too expensive and irreproducible in our mock production studies. However, for projects in which unincorporated dye terminator signals will interfere with analysis, column cleanup is preferred. We also routinely resuspend precipitated samples in deionized water (as opposed to formamide-based solutions) before loading, because this provides the maximum yield of resuspended fragments from the pellet. On the 3700, samples resuspended only in water typically evaporate during the preceding runs, and we add one half of the final volume as deionized formamide (after the initial water resuspension) to prevent complete evaporation. On the MegaBACE, the addition of one half of the final volume as 0.2% deionized hydroxyethyl cellulose helps to prevent most of the overinjection failures that occur with LPA, presumably by acting as a molecular weight sieve.
TABLE 1
Modifications to Suggested Run Parameters Made for the PE 3700 or MD MegaBACE to Improve Performance
|
|
||||||
| Instrument | Suggested Parameter | Modified Parameter | Rationale for Modification | |||
|
|
||||||
| 3700 | 50°C oven and sheath-flow cuvette temperature |
37°C oven and sheath-flow cuvette temperature (dye terminator chemistry only) |
Lowering the run temperature decreases the matrix viscosity and allows better fragment resolution range |
|||
| 3700 | 10,000 sec data collection time | 12,000 sec data collection timea | Additional data collection time required for slower migration of fragments at 37°C |
|||
| MegaBACE | Not applicable | Preinject samples at 3 kV for 20 sec | Helps to rid samples of excess ionic content that can interfere with electrokinetic injection success |
|||
| MegaBACE | Loading from a standard formamide/EDTA solution |
Loading from an aqueous 0.1% hydroxyethyl cellulose solution |
Reduces the incidence of over- injected samples |
|||
| MegaBACE | Inject samples at 10 kV for 15 sec | Inject samples at 2 kV for 60 sec | Reduces the incidence of over- injected samples |
|||
| MegaBACE | Data collection for 90 min at 10 kV | Data collection for 140 min at 7 kV | Increases the resolution and readlength for most samples to >750 bases |
|||
|
|
||||||
aA consequence of longer data collection time on the 3700 is that the sheath-flow volume must be increased to 13,500 µL to provide sufficient amounts of polymer for the sheath flow elution of fragments.
There are two phases of instrument stability to be considered. The first phase is stability for a newly released instrument, which is typically poor, because a variety of design, manufacturing, installation, and shipping issues are being sorted out. Both the MegaBACE and the 3700 have lived up to expectations with regard to this phase of stability, with each exhibiting a unique set of challenges. The second phase is long-term stability, which is assessed after the manufacturer has overcome most of the aforementioned issues. Although not all instruments successfully transition to the second phase of stability, the MegaBACE has entered that phase. We are evaluating these units with respect to their percentage of "up" time over the ensuing months. So far, the stability has been excellent, approaching more than 95% up time over the past 8 months. The 3700, because of its recent introduction, is in the initial phase of stability assessment.
A major cost-related issue in capillary sequencers is the expected lifetime of the capillary arrays. In the 3700, the capillaries are all part of a single bundle of 110 uncoated, fused silica capillaries, consisting of 96 primary capillaries, 8 reserve capillaries, and 6 sheath-flow guide capillaries. The reserve capillaries can be designated for use in the event that between one and eight of the primary capillaries is exhibiting substandard performance. In the MegaBACE, the capillaries are bundled together as sets of 16 coated, fused silica capillaries each, such that 6 separate array bundles comprise the 96. Whether an internal capillary coating is required is a function of the separation matrix used and directly affects the potential capillary lifetime. Intuitively, uncoated capillaries should last significantly longer than coated capillaries because the coating will wear away with repeated usage. As such, the specifications for MegaBACE capillaries are a lifetime of up to 100 runs and for 3700 capillaries up to 300 runs. Perkin-Elmer reports that flushing capillaries with a nitric acid wash can regenerate them for up to 100 additional runs.
How does specification compare with reality in our experience? Our records for MegaBACE capillary arrays indicate that, on average, 170 runs can be obtained from a set of 16 capillaries before the resolution, judged by analysis of a standard control reaction, falls below the manufacturer's specification of 550 bases. Although our experience with the 3700 is much less extensive, we have not yet achieved 300 uses from an array of capillaries before a similar assessment process indicates that resolution of sequencing reaction fragments falls below 300 bases. Nitric acid treatment of these capillaries has not successfully restored them to function in a limited number of attempts on different arrays, but recent experience has taught us that several runs may be required before the acid-stripped capillaries again provide optimal performance.
The separation matrix is an integral part of the capillary sequencer and most of our development efforts have centered on overcoming the undesirable aspects of either matrix formulation. In the case of LPA, the matrix is extremely viscous and in our hands demonstrates the most potential for routinely obtaining sequences that span greater than 700 high-quality bases. Among its undesirable aspects, the pore size (among other physical characteristics) probably exacerbates the tendency of this matrix to overinject or "clog" with template molecules, proteins, or other species that out-compete the sequencing reaction fragments for loading. The incidence of clogging is undoubtedly a consequence of our choice to use lower purity template preparation methods, because our early experience with loading reactions obtained from polymerase chain reaction-generated templates did not indicate a high percentage of failure due to clogs. The addition of hydroxyethyl cellulose before loading reactions on the MegaBACE significantly reduces or eliminates this problem. The POP6 matrix, having different physical characteristics from LPA, does not tend to clog, but it exhibits lower resolution potential. The modifications that we have made to run parameters, shown in Table 1, have primarily been aimed at improving POP6 resolution, although we have only achieved an average of 500 high-quality bases for the Big Dye terminator chemistry with these in place.
Labs that are planning to have multiple capillary platforms need to think carefully about the necessary infrastructure to support these instruments, because it is significantly different from that required for slab gel sequencers. Depending on the laboratory size and throughput, infrastructure considerations may include a system to monitor capillary arrays for number of uses and performance, scheduled focusing assays, scheduled replacement of arrays on a rotating basis, and clearing of disk-archived data from hard drive storage. For the MegaBACE, a rotating supply of nitrogen tanks must be kept on hand. Some laboratories have reduced this need somewhat by using compressed air in place of a low-pressure nitrogen tank for driving the anode and cathode stages. For the 3700, routine nitric acid washes of capillaries may become necessary. In practice, these types of activities are best performed by a dedicated group that has the appropriate monitoring tools in place and is fully experienced in capillary replacement techniques and required postinstallation verification procedures.
This review was intended to be a comprehensive, unbiased report on our experience with two commercially available capillary sequencers, the MegaBACE 1000 and the Perkin-Elmer 3700. Potential buyers of capillary instrumentation should fully investigate the capabilities, specifications, and throughput of an instrument in light of the type of sequencing projects to be performed before a decision to purchase is made.
The author wishes to thank John Bashkin and Jackie Snider for critical reading of the manuscript. Thanks also are extended to two valued collaborators, John Bashkin and Paul McEwan. Bob Waterston and Rick Wilson have provided encouragement and enlightenment, as always.
1. Huang XC, Quesada MA, Mathies RA. Capillary array electrophoresis using laser-excited confocal fluorescence detection. Anal Chem 1992;64:967-972.
2. Marsh M, Tu O, Dolnik V, et al. High-throughput DNA sequencing on a capillary array electrophoresis system. J Capillary Electrophor 1997;4:83-89.
3. Kambara H, Takahashi S. Multiple-sheathflow capillary array DNA analyser. Nature 1993;361:565-566.