1. Introductory remarks
Reflecting on the past 15 years of the history of large-scale DNA sequencing, it is incredible to realize just how far this discipline has progressed in a relatively short period of time. The efforts, hopes, fears, and failures of many molecular biologists, engineers, and others have been played out in the arena of academic and industrial labs – all pursuing the advancement of the field. Nowhere have their efforts been more critical to these achievements than in the incorporation of robotics and automation to render tasks once entrusted to skilled technicians into the routine, programmed movements of robotic systems. This overview aims to trace the early origins of those efforts and their metamorphosis over time, as well as to present a state-of-the-art picture of high-throughput DNA sequence production for the interested reader.
2. Overview: critical components that rendered DNA sequence an “automatable” process
Before one can begin to chart the history of robotics and automation in large-scale DNA sequencing efforts, however, it is important to point out many of the factors that played a role in rendering DNA sequencing an “automatable” process. First and foremost, sequencing developed, over time, into a routine process that utilized the same series of steps and the same components at each iteration. One enormous contributor to the routine nature of sequencing reactions was encompassed by the development of cycled sequencing – an offshoot of PCR, in which a single primer is extended by a thermostable DNA polymerase on a template in the presence of dNTPs and ddNTPs, producing a great excess of dideoxy terminated fragments (in comparison to the input template amount) and eliminating stepwise addition of DNA sequencing reagents as in the past (McBride et al., 1989; Craxton, 1993). Later, a series of substantial improvements to DNA sequencing methods (enzymology and DNA fragment labeling in particular) and to associated methods such as DNA template preparation further enhanced the ability to automate reaction assembly by robotic means. Several of these key developments are highlighted below. Another important contribution to the routine nature of DNA sequencing was that the detection hardware, DNA sequencing instruments, also improved significantly over time. A third development was the adaptation, borrowed from the clinical laboratory, of the multiwell microtiter plate format for sample handling, which provided a reproducible footprint and well-to-well spacing for access and plate handling by robotic devices. Along the way, industrial practices such as bulk solution manufacturing and quality control were developed and implemented by many high-throughput DNA sequencing facilities, contributing a reliability factor to the reagents used in these processes. The combination of these factors, along with the ever-increasing scale of DNA sequencing throughput became such that only robotic solutions were appropriate, especially in their elimination of manual errors, their ease of sample tracking (via bar code reading/entry), and the reproducible nature of their product.
3. Early days – the “do-it-yourself” era
When we and others first started thinking about automating DNA sequencing and its associated processes in mid-1980, suffice it to say that there were very few instrument manufacturers thinking about the same things. This situation improved slowly over time, but many first efforts at “automation” were fairly rudimentary, not truly “automated”, and nearly always enlisted devices that were borrowed from other disciplines or were devised by genome center-associated engineering teams responding to demands for increased throughput. Predominantly, early efforts were aimed at applying robotics to individual processes in the DNA sequencing workflow rather than at the development of larger, integrated systems.
One of the first areas to be pinpointed for robotics application was that of liquid transfer, since reliable, reproducible manual liquid transfer of microliter volumes is a function of the quality of the pipettor and the skill of the person using it (not to mention adjunct factors of fatigue and interruption!). Most commercially available liquid-handling robots were designed for clinical applications that utilized the multiwell microtiter plate format. Hence, the adaptation of 96 well microtiter plates into DNA sequencing methods likely was a happenstance of the need for robotic liquid handling. Although many of the robots were standalone devices, they brought a reliability and reproducibility aspect to the sequencing process and enhanced throughput at very early stages. For example, our Center used the Robbins Hydra96™ pipettors to both prepare DNA templates (Mardis, 1994) and to pipette sequencing reactions for many years, and very early reports of sequencing reaction and DNA isolation methods used the Beckman Biomek 1000 robot (Wilson et al., 1988; Mardis and Roe, 1989; Koop etal., 1990).
Upstream of the DNA isolation and sequencing processes, harvesting recombinant clones from agar plates was another process that was targeted early on for automation, for quite obvious reasons – the process initially was done by technicians whose tools consisted of lightboxes (for imaging the plaques or colonies) and beakers of sterile wooden toothpicks (for harvesting the plaques or colonies). Several robots were designed and built, both in companies and academic labs (Panussis et al., 1996), for the purpose of picking M13 plaques and/or plasmid-containing colonies from agar plates. These robots typically combined a vision system for imaging plates, an algorithmic selection of “pickable” plaques or colonies, a picking mechanism that contacted each plaque/colony to harvest it, and an axis/gantry system that precisely located the picking device over the plaque/colony to be harvested and over the microtiter plate well to be inoculated. The implementation of these devices into high-throughput sequencing labs provided an automated solution to a very tedious process and enabled a greatly enhanced scale of operations, not to mention making a significant dent in wooden toothpick sales worldwide.
The advent of large-scale cycle sequencing necessitated improvements to thermal cycler design – faster temperature transitions, smaller footprints, and multiple blocks were all developed to address the needs of high-throughput labs. Again, this was an area where both academic and commercial entities contributed their own designs and concepts, most of which again centered around the 96 well (and later 384) microtiter plate format. Hand in hand with thermal cycler development was the genesis of injection-molded microtiter plates that included the use of plastics with low thermal deformation and improved heat transfer characteristics, as well as design aspects that enabled both manual and robotic handling.
4. Impact of industrial/academic collaborations
As mentioned previously, there were a number of key discoveries or applications of existing technology that provided further ease of automation for DNA sequencing and related processes. Many of these resulted from efforts in academic labs funded to provide technology development for the Human Genome Project (HGP) by federal agencies such as NIH and DOE. These discoveries, in turn, were further developed into commercial products by companies that either became interested in, or were created to supply reagents, instrumentation, and develop technology for DNA sequencing.
Tabor and Richardson (1995) published a report of a mutated Taq polymerase with a single amino acid change in its nucleotide-binding site that mirrored the amino acid present at this position in the viral T7 polymerase. This single amino acid change effectively eliminated the polymerase’s incorporation bias of dNTPs over ddNTPs, allowing a significant reduction in the ddNTP concentrations in sequencing reactions. Also in 1995, Ju and Mathies reported the use of fluorescence resonance energy transfer (FRET) technology in the design of base-specific dye primers (Ju et al., 1995a,b), effectively transforming a technique long used in protein structure determination into a fluorescent labeling strategy for modern day DNA sequencing applications. Energy Transfer (“ET”) dye primers, and later “Big Dye” dye terminators (Lee et al., 1997), significantly enhanced DNA sequencing by dramatically reducing the amount of input template required to produce high quality sequence patterns, based on the enhanced quantum yield obtained from energy transfer from donor to acceptor dyes. Ultimately, the combination of FRET-based (“Big Dye™”) dye terminators and the active site-modified thermostable polymerase yielded the most readily automated combination of technologies, and presently experiences widespread use in automated DNA sequencing. Namely, dye terminators enable one reaction per template (instead of four reactions when using dye primers), the modified enzyme allows faster thermal cycling (fewer cycles, shorter extension times), and easier reaction cleanup to remove unincorporated terminators (since the ratio of fluorescent labeled ddNTP terminators to dNTPs is significantly reduced).
In addition to the commercialization of these key sequencing chemistry improvements, during the time period 1997-2001, many large-scale sequencing centers partnered with commercial robotics vendors to devise custom robotics solutions for specific high-throughput processes (Oefner et al., 1996; Marziali et al., 1999). Several of these robots, or scaled-down versions, subsequently transitioned into products for the sequencing market. In general, these robots either used a centrally positioned, articulated robotic arm or linear conveyor belts with gantry-mounted grippers to position microtiter plates onto various stations (liquid handling, plate sealing, mixing, lidding/delidding, etc.) on the workspace, and utilized scheduling software to keep all stations as occupied as possible in order to maximize throughput.
It is important to point out the critical contribution of sample tracking via barcodes and databases to these efforts. Without these tools, large numbers of samples would require manual logging as they progress through the DNA workflow, ultimately limiting the throughput.
Ultimately, DNA sequencing technology implementation is only as good as the instrument on which separation and excitation/detection occurs. As such, much of the ability to automate DNA sequencing was enabled by the evolution of DNA sequencing instrumentation, which has been considerable since commercial introduction of the Applied Biosystems 370 A in mid-1980 (Mardis and Roe, 1989). Early sequencing instruments, while automated in terms of laser scanning, fluorescence detection, and data analysis, required much user interaction including gel pouring, gel loading, and manual assignment of sample lanes (“tracking”). Developments that enhanced the automation of these instruments included replacing slab gels with capillaries (eliminating gel pouring and tracking), capillary illumination with a nonscanning laser beam (eliminating the time required for side-to-side scanning), and providing onboard robotics that scan barcodes and automate loading directly from a microtiter plate stacker (eliminating nearly all manual intervention) (Mardis, 1999).
5. Advent of high-throughput integrated DNA sequencing automation systems
The recent commercial introduction of DNA sequencing instruments with very high sensitivity has enabled the latest round of integrated robotics that approach full automation – that elusive ideal often referred to as DNA in, sequence out. In some of these efforts, the conventional, microtiter plate-based approach has been abandoned in favor of capillary tubes (an ironic twist of fate, since originally glass capillaries were used for DNA sequencing reaction vessels prior to the introduction of polypropylene microcentrifuge tubes). In other efforts, such as our own, 384 well microtiter plates are being used to prepare a sufficient amount of DNA for one sequencing reaction in each well and then directly sequencing that DNA in the same plate. Regardless of the approach, the combination of enzymology, fluorescent labeling technology, submicroliter pipetting capability, rapid thermal cycling (due to enhanced thermal transfer and fewer cycles for small volumes), and detection sensitivity are making more fully integrated robotics approaches possible.
These efforts fall short of a fully automated approach since some preliminary work is required to grow the subclones in culture and since the sequencing products ultimately are loaded and detected on a separate instrument. As yet, the elusive goal of full automation continues to be the focus of many academic and commercial efforts to miniaturize the DNA preparation, sequencing, separation, and detection processes, by a variety of approaches. The most successful of these efforts should provide us with the next generation of automated DNA sequencing instrumentation systems, hopefully in the next few years.
