A Model for Efficient Assay Development and Screening at a Small Research Site

Building an automated assay group 

In the 1990s, when high throughput screening (HTS) became a central part of the drug discovery armory for all major pharmaceutical companies, the Terlings Park research site of Merck in the United Kingdom, in common with many similar research sites, set up an automation group for rapid screening of the Merck compound collection against new targets for distributed hit generation and confirmation. Automated linear track systems were installed at each site, capable of screening collections of singlicate compounds or mixtures. Each automation group performed single point screening primarily by radioligand binding for identifying hits for the projects at its site. Terlings Park purchased and set up two Beckman–Sagian (Fullerton, CA) linear track robotic systems to support this process. Identified hits were then followed up by medicinal chemistry. Lead identification and lead optimization assays (LI & LO) were run within the project teams, in a largely manual fashion. A handful of Biomek 2000s were purchased and spread throughout the Terlings Park lab groups to make the follow-up process somewhat more efficient.

Toward the end of the 1990s, it became possible to run cell-based assays at high throughput with the development of new imaging platforms, led by the first-generation fluorescent imaging plate readers (FLIPR, Molecular Devices Corporation, Sunnyvale, CA) and voltage-ion probe readers (VIPR, originally from Aurora Biosciences now Aurora Discovery, San Diego, CA). In 1998, Terlings Park set up a nascent cell imaging assay group, which took on responsibility for implementing semiautomated cellular assays for ion channel and G-protein coupled receptor (GPCR) targets, in particular developing innovative medium and high throughput functional ion channel assays.1, 2, 3 A major paradigm shift in screening occurred at the end of the 1990s—the increased sophistication and increased cost of automation technology meant that it became seen to be more efficient to set up a single centralized facility. This would be responsible for screening for hits from a greatly expanded singlicate compound collection to support the global research effort.4, 5 When this core facility was set up, some of the Merck & Company research sites dismantled their automation groups, and key staff left to pursue automation within other companies. Centralized ultra high throughput screening (uHTS) generated confirmed active hits; all responsibilities for lead generation and optimization fell to the distributed sites. The responsibility at Terlings Park for providing primary medicinal chemistry compound screening data shifted from the project team to the local automation group. Terlings Park chose to merge their automation group with the cell imaging assay group, and gave the combined group an expanded portfolio of responsibilities. Occasional HTS screens were still performed where appropriate due to, for example, patent situations in the United States, but the primary work began to involve many types of assay development, development of uHTS screens, and automation and miniaturization of all frontline medicinal chemistry supported LI & LO assays. In this article, the development of a centralized assay team that supports all automated assay development and multiple types of screening will be described.

Strategy for centralizing assay support 

The automated assay team over time has developed and expanded its capability for performing a broad range of assays, to future-proof the screening options for new chemistry-supported projects. The current responsibility covers all LI & LO biochemical assays, principally in vitro enzyme assays. Also, fluorescence-based cellular signaling assays are performed including calcium, membrane potential, and gene reporter assays, and related flash luminescence assays (aequorin). Some radiometric assays are still used, notably uptake assays for transporters. An increasing number of cell-based enzyme assays are performed, using a variety of formats—peptide secretion by Meso Scale Discovery reader or substrate conversion and phosphorylation assays using fluorescence intensity, fluorescence polarization, or time-resolved fluorescence. In addition, endpoint and kinetic high-content screening (HCS) assays, and in vitro electrophysiology assays, are run regularly. The strategy at Terlings Park has not been to adopt a one-size-fits-all strategy for all project teams, but to evaluate what benefit centralization and automation offer on a project-by-project basis. For example, when setting up this automated assay team, we did not centralize screening for every project at once. Instead, we implemented a gradual roll-out of the new model, led by projects that would most immediately benefit. FLIPR and radioligand binding projects were taken on first, because a FLIPR (Fig. 1C) and a filtration binding system were already present on the Beckman–Sagian robots at that time from the previous HTS function. Buy-in from project teams for this new business model was consequently based on demonstration of increased productivity, reliability, and added value, rather than by top down enforcement, which is always likely to lead to resentment and division.

View Large Image Download to PowerPoint

Figure 1. Key items of automated equipment. (A, B) Thermo CRS Robolab system, supporting FLIPR ion detection (A), and after moving into an expanded lab was rebuilt with additional deck space to incorporate additional equipment including Tecan Genios Pro plate reader (B) for assays such as homogenous time resolved fluorescence. C is a Beckman–Sagian ORCA track supporting FLIPR assays. D shows a safety cabinet exhausted to stack to remove any hazardous vapors. Each of our robots is contained within a separate room, with glass panels enabling viewing between rooms, located off the main plate preparation work area. E, F two views of an RTS dual Stäubli arm, rather than track, robotic system. The scheduling software allows each arm to perform individual assays or to work cooperatively on a single assay, with dynamic scheduling of plates and assays to be read on Tecan Genios Pro or Ultra readers.

Developing the team 

The automated assay team personnel was assembled by a mixture of happenstance (the availability of suitable scientists from other parts of the site due to reorganization), along with strategic planning for key recruitment. There was a particular emphasis on bringing on board scientists who had previous Pharma assay development expertise and understanding. In developing this group, it was viewed that the optimum team for automated assays would consist of a balanced mixture of ex-high throughput screening group members, with their well-developed automation and troubleshooting skills, and conventional lab scientists with prior expertise in assay development and optimization with strong links to the project teams. Some key staff were in some cases transferred from within the organization and others were recruited from academia (wherever possible with some industrial experience) or directly from other pharmaceutical organizations to develop a team with the required broad skillset and mixture of expertise.

From their base skillset, all individuals have broadened their automated assay skillset by formal and informal training from internal and external experts, by ‘over-the-shoulder’ learning from others within the team with expertise in and responsibility for particular technologies, and by direct exposure to having the opportunity to develop and implement innovative technologies and assays. This means that each experienced team member is able to develop screening assays, optimize and miniaturize assays, or deliver screening results on a regular basis to support long-term medicinal chemistry projects, depending on site needs at a given time. All team members are identified as ‘power users’ of at least one technology, with time set aside in their annual objectives to become expert and then for them to develop others' skills and expertise with that system.

Whereas within some organizations, assay development, LI & LO are separate functions, in this operational model they are performed on the same instruments by a single team. Among the benefits of this approach have been freeing up individual scientists in the project teams to focus on target validation, proof of concept experiments, and identification of new project targets within their given therapeutic franchise. Also, it allows automated assay scientists to focus on maximizing productivity and expertise in a single area, such as miniaturization, automated cell culture, or robotic programming. Screening is not divided into silos depending on the project stage, such a single assay scientist can be responsible for project support from initiation through to acceptance of preclinical drug candidates. Individuals frequently develop assays, and run the consequent screen. As well as providing a more interesting job, this approach broadens scientific skills enabling flexible deployment of staff to meet project needs for assay development or screening support to address shifting bottlenecks, and provides a greater incentive to develop stable robust assays, which are not just “thrown over the wall” to another group.

Further expansion of responsibilities for this centralized assay group occurred in 2003, when the compound management team joined the group, and again in 2005 when the existing in vitro electrophysiology team was merged into the group. These mergers enabled cross-skilling of individual scientists from these focussed areas, and has led to a more efficient, flexible deployment of staffing resources. A single, representative, automated assay scientist with initial skills in patch clamping has learnt fully automated biochemical kinase screening and workstation-based cellular HCS assays to a high standard. Most screening assays are fully developed within the automated assay group. However, flexibility, not forced rigid rules on such matters, has worked well. Where a particular expertise and knowledge of the therapeutic area are required for assay design, some assays are initially developed in the project team's biochemistry group and then transferred to the automated assay group for optimization and miniaturization. Some novel assay types have been outsourced for the initial assay test case, typically to the vendor, which has developed the new technology. This has provided the optimum opportunity for testing the potential utility of their new tool. If the vendor cannot demonstrate the assay generates a robust screening window in our test case, there is no need for us to spend the time to bring a technology in-house, learn its idiosyncrasies and develop a screen. In some cases, an unusual assay type is required, which we do not anticipate requiring regularly. In such cases, outsourcing to the experienced scientists within the vendor company can again be the most efficient path forward.

Key technologies 1–HCS 

New technologies have been brought into the group over time. The automated assay group developed significant expertise in HCS, initially using Cellomics ArrayScanII, (Cellomics, Pittsburgh, PA) for lead optimization and novel target identification.6, 7 After the successful implementation of Cellomics ArrayScanII, the expanding demands for HCS assays led to further HCS systems being purchased (BD Pathway HT (BD Biosciences, Rockville, MD), GE IN Cell 1000 (GE Healthcare, Chalfont St. Giles, U.K.)). These expanded the team's support work into earlier stage work in target validation, library screening, and stem cell projects. Using such platforms, we have pioneered the implementation of assays that involve multiple high-content readouts, or both conventional and high-content readouts, from the same wells. These ‘multiplexed’ screens provide directly comparable data sets for target activity and for compound toxicity or off-target effects (Figs. 2D and E).8, 9, 10 The ArrayScanII has an integral Twister to supply plates. With an automation vendor (Process Analysis & Automation, Farnborough, U.K.), we ensured that InCell 1000 was integrated with a TwisterII plate handler (Caliper Life Sciences, Hopkinton, MA).

View Large Image Download to PowerPoint

Figure 2. Plate readers deployed supporting multiple assay types. A is a Meso Scale Discovery SectorHTS which has been integrated onto a Beckman–Sagian linear track system, alongside a PE Fusion reader, so that certain assay types can have multiplexed readouts using both technologies. B shows a Aurora Discovery topology compensating plate reader and flying reagent dispenser, which are used for development of 1536 and 3456 well uHTS assays. C is a CyBio Lumax SD flash luminescence reader with stacker, for economic and rapid screening of aequorin detecting intracellular calcium. D is a BD PathwayHT kinetic cell imaging system driven by a Thermo CRS CataLyst-5 robot arm. E is a GE IN Cell 1000 for HCS with Caliper Twister II integration. F shows a semiautomated workstation for supporting specialized assay applications—in this case ion channel membrane potential assays on Aurora Biosciences VIPR supported by plate preparation on a PE PlateTrak.

We have fully automated for the first time a BD Pathway HT kinetic HCS cell imaging system (Fig. 2D) for single cell oscillation assays that enable signaling studies in mixed primary cells and stem cell subpopulations.8, 9, 10 Because Terlings Park was an early adopter of this technology, no commercial driver was available. We worked closely with Thermo-CRS (Burlington, Canada) and BD to design and develop an automated system, containing kinetic and endpoint screen management that met our flexible needs.9 This automation integration was not developed for HTS, because the reader images one well at a time often for several minutes in kinetic mode.9 However, these long-term assays can require cells to be removed from the reader to a cell incubator and back over the course of many hours, or to be automatically moved from the reader and immediately fixed and immunocytochemically labeled, then returned to the same position on the reader. Particularly as such assays can last potentially for 24h or more, this type of imaging-based screening requires fully automated control and management of imaging, liquid handling and cell labeling, and cell incubation. Automation of assays can, thereby, be not just to increase throughput but also for reasons of precision and practicality.

Key technologies 2–robots 

Second-generation flexible and robust linear track screening systems were designed and purchased, which can be used for multiple assay types (Figs. 1A, B, E and F, 2D; Table 1). To minimize downtime, screening assays are scheduled on both automated FLIPR track systems (Figs. 1B and C) or on duplicate Tecan reader-enabled robotic systems (Figs. 1B and E). A key aspect of our technology purchase and automation design has been to ensure availability of a fallback position should one robot be temporarily unavailable to users (e.g., due to servicing, upgrades, or system failure). High-priority readers are duplicated on at least two systems, or if necessary backed up with a manually accessible reader of the same class. We ensure that assay performance is validated and extremely comparable on the two sister systems, so that if a single screening system is unavailable priority assays can continue unaffected on the other system (Fig. 3).

Table 1. Assay equipment as laid out within the Terlings Park automated assay team

	System	Automation vendor	Integrated readers	Assay formats
	Robot 1	Beckman–Sagian	FLIPR	FLINT
	Robot 2	Beckman–Sagian replaced by RTS	TriLux replaced with Ultra & GeniosPro	Radioligand binding replaced with HTRF/FLINT/FP
	Robot 3	Beckman–Sagian	SectorHTS & Fusion	Cytotox/ECL/glow luminescence/HTRF/FLINT
	Robot 4	ThermoCRS Robolab	FLIPR & GeniosPro	HTRF/FLINT
	Robot 5	ThermoCRS Polara	PathwayHT	Kinetic FLINT/FRET
	Semi-automated	PlateTrak, Biomek FX, Matrix 2+2, Flexdrop, Flying reagent dispenser	VIPRs	FLINT/FRET
			ArrayScan, INCell 1000	HCS
			Topology compensating plate reader	1536/3456 FLINT
			Lumax	Flash luminescence
			TopCount	Glow luminescence

FLINT=fluorescence intensity; FRET=Forster resonance energy transfer; HTRF=homogenous time resolved fluorescence; ECL=electrochemiluminescence; FP=fluorescence polarization.

View Large Image Download to PowerPoint

Figure 3. Workflow for assay development. A standard operating protocol has been written up that describes how we will work together collaboratively on this side to develop cell lines and screening assays. The protocol is summarized in these two drawings, and detailed at greater length in the document shared with all participating parties.

Some technologies not widely deployed in the industry for automated screening purposes were automated by our team for particular, high priority, projects. These include a Meso Scale Discovery SectorHTS reader (MSD, Gaithersburg, MD) (Fig. 2A) for sensitive antibody-based assays. No robotic driver was commercially available for this reader, even though it has become increasingly popular within the industry for lower throughput determinations later in the screening cascade. At our site, it was being deployed for a frontline screen in a high-profile lead optimization project, as well as for other lower throughput project support, justifying automation onto one of our systems. We worked closely with Meso Scale Discovery and Beckman to ensure that a competent driver was developed and validated on one of our Beckman–Sagian track system. As this was a custom project, there was an initial need for us to convince the automation vendor that this product would be of interest to others; further optimization of the initial release might be desirable to meet a broader range of other customer needs in the future. As described above, we have also automated for the first time a BD Pathway HT kinetic cell imaging system.9 Other companies have subsequently expressed interest in a similar integration of this technology. Radioligand binding assay demands have declined, as functional cell-based assays became the primary assay of choice—for space reasons a Beckman–Sagian binding assay robot was dismantled in 2004 and replaced with a custom-designed dual Stäubli arm robot (RTS, Manchester, U.K.) suitable for multiple functional assay types (Figs. 1E and F).14 This RTS system was designed with inbuilt expansion space to meet future reconfiguration needs. Also to cope with future reconfiguration needs, our Thermo-CRS track was originally built with only one side on the system, to fit in a small lab space (Fig. 1A). It was subsequently rebuilt in an expanded laboratory space with equipment on both sides of the track (Fig. 1B), enabling us to add additional liquid handling and to integrate a second reader type, Tecan Genios Pro as well as FLIPR.

A variety of plate readers have been deployed for toxicity, GPCR gene reporter assays, and enzymatic readouts (Figs. 2A–C; Table 1) (drug metabolism, pharmacokinetic, and cellular toxicology assays are run within a separate department). For ion channel work, generally our two FLIPRs,1, 211 and for particular assays the lower throughput VIPR-I and VIPR-II,312 are used. Full assay automation is very powerful, but can be cumbersome for some assays where the throughput does not justify this. Again, flexible deployment and allocation of assays to particular automated or workstation-supported readers ensure that the technology and scientists are operating efficiently. Many assays with modest chemistry support have been run not fully automated, but using appropriate components of the automation laboratory and liquid handling technologies that suit these medium throughput assays. Stand-alone workstations, of varying complexity and abilities, were purchased and installed. We have chosen to always buy from the best vendor for the particular equipment type, not become tied to a partnership with a preferred vendor. Both FLIPRs were automated onto linear track systems (Figs. 1A–C) whereas VIPRs were supported by workstation plate preparation systems, such as PlateTrak (PerkinElmer, Boston, MA) (Fig. 2F).

Miniaturizing assays 

The Terlings Park automated assay team works very closely with colleagues in the centralized uHTS screening facility, in developing many uHTS-compatible assays. For this, we use a range of technologies, including 1536 and 3456 well-format liquid handling and plate readers (Aurora Discovery, San Diego, CA; Fig. 2B) for fluorescence assays such as enzymatic readouts, or gene reporter assays typically for GPCRs.4513 For other assays such as ion channels, we typically develop 384-format assays that we ensure will be compatible with miniaturized technologies at the uHTS site.15 Benefits of this partnership approach with our uHTS colleagues include ongoing communication between regular points of contact so all parties know what is required for successful uHTS development. Also, because the assay is developed within the Terlings Park automated assay team, this expertise becomes immediately transferable to hit titration at Terlings Park after uHTS is complete, and then to automated LI & LO assays. It should be emphasized that we do not currently have a need to run routine lead generation/optimization assays in formats beyond 384, given the numbers of compounds being generated by internal and outsourced chemistry for each project. We have identified and agreed a preferred 384 well-plate format, which is recommended for use by team scientists, to simplify liquid handler programming, and to assay quality control (QC) and data analysis design. It has been interesting, however, to note that when we have miniaturized assays from 384 format, often using preplated cells, into 3456 format using suspension cells for transfer to uHTS, screening windows and Z' values have remained robust in virtually all cases.

Technology project management 

Expensive equipment is regularly used within a centralized assay group for multiple projects and franchises, thereby increasing return on investment. As a relatively small research site, it has been crucial that we have bought equipment carefully at relatively modest expense. The strategic plan for equipment purchase has been enhanced with the identification and deployment of a senior team member into a technology project management role. This site-wide role enables the individual to focus on identifying new technologies of importance to the site, followed up by setting up demonstrations and evaluations on-site. When new technologies are deemed valuable and affordable, the technology project manager then helps the assay scientists through the internal purchase approval process, then takes responsibility for ensuring value by cost negotiations, factory visits, and close ongoing liaison with the vendor(s). This ‘ownership’ of the technology identification and implementation process by a single individual has ensured that key technologies have been delivered on time, that site acceptance testing is rigorous and consistent, and that subsequent equipment service provision is closely monitored. After installation, daily documentation of technology performance is fed into quarterly formal meetings with our robotic suppliers under the technology project manager, to ensure close monitoring of equipment reliability and service support levels.

Project support 

It is important to demonstrate to project team leaders the benefits of this dedicated assay team model practically to ensure their confidence and support. In one example project, the LO assay was developed and run within a project biochemistry group before the centralization of assay screening occurred. Originally, the difficult-to-grow cells were cultured by the project team, and the 3-day-long assay was performed in the lab group as the responsibility of three senior scientists. Along with the transfer of the assay to the assay automation group, cell provision was transferred to our centralized cell culture group and automated on SelecT (The Automation Partnership, Royston U.K.). They implemented various modifications to growth conditions by the cell culture group that have significantly improved cell reliability. The assay has been automated on one of our robot systems (Fig. 4), with protocol modifications and quality control checks that have again improved week-to-week data consistency. Rather than being a responsibility of multiple senior postdoctoral scientists, the assay is now fully (and well) performed by a single junior team member.

View Large Image Download to PowerPoint

Figure 4. Comparative data obtained from a manual assay with a modified automated version of the same assay that became the frontline LO assay for that project. Very strong correlation over many months was demonstrated, with both assays running in parallel until the project team was happy to make a full transition.

Whereas in the early days, project team members were occasionally initially reluctant to let go of their assay, support for our model has grown greatly. This is assisted by changes to cell culture and screen design, by people whose full-time job was cell culture and screening and therefore have a broad perspective on assay optimization, resulting in high-quality data. Again, our model has been to demonstrate the benefits rather than to argue them. We did not miniaturize the assay format in the example discussed above beyond 96-well, as throughput did not demand it and the screening window was already relatively modest—this is unusual for us and we miniaturize to 384 in most other cases. Screening data analysis itself has gone through several iterations in recent years. When HTS assays were routinely performed here, the site relied upon ActivityBase (IDBS, Guildford, U.K.) for data management. This was replaced by in-house custom Excel-based macros during the first phase of automating LI & LO assays. More recently, a widespread consultation process within the organization resulted in agreement on an internal data analysis solution that can now be shared across each of the distributed sites.

The Terlings Park assay development model is to set up a small team involving scientists from the project team, cell culture group, and automated assay team (Fig. 3). An assay development coordinator is identified, who could come from any of these teams depending on which individual is most appropriate, to liaise with project and departmental managements and ensure that sufficient FTE resources are available, and assay development milestones are agreed. This small interdisciplinary team makes the decisions as to who does what during the assay development process—close collaboration and agreement between these individuals should enhance the satisfaction of the project team and maximize the quality of the final assay. This has become a fast, streamlined approach in which everybody understands their roles, with a clear standard operating procedure to follow to minimize potential for disagreements or misunderstandings (Fig. 3).

Automated assay group organization 

There are a number of ways of organizing an automated assay group into subteams to meet the needs of multiple projects. Organization can be by technical expertise on particular screening systems or assay types, or can use project franchise alignment with skills across a range of techniques within each subteam. The Terlings Park automated assay group follows the latter approach, with screening subteams each being headed by a senior team member and closely integrated with particular therapeutic franchises. The franchise-based focus helps project leaders to work with us in prioritizing work on the targets and screens within that franchise, as it generates clarity for them as to the options for deployment of the finite resources available. On occasions when demand is greater on one franchise in another, scientists can be deployed flexibly outside their usual franchise. These decisions are made easier by regular communication between the automated assay group leadership and project leaders. This occurs primarily by three mechanisms—by regular 1:1 meetings between project and automated assay group leadership at least every 8 weeks, project core team meetings, and by the issuing of a monthly report to all project leaders on the work of the automated assay team (Fig. 5). Monthly reports have engendered confidence in our collaborators and stakeholders, and result in a broader understanding of the competing demands on assay support scientists' time.

View Large Image Download to PowerPoint

Figure 5. Monthly report examples. Different types of summary reports are written for exploratory assay development, and screening assay work. (A) Typical screening project update; and (B) robot performance tracking table.

It is important for the automated assay group to interact with all the projects and franchises to ensure we are closely aligned at any given time. Projects have changed very rapidly at this site so a carefully developed, robust screen may very rapidly no longer be needed, with an immediate expectation that an entirely new approach is to be automated and ready to go in a matter of days or weeks. We have also had to carefully manage workload—having achieved full buy-in of the site we were in danger of becoming victims of our success, making it necessary on several occasions to argue successfully for additional headcount and technology purchases to ensure we could continue to meet the site's expanding in vitro automated assay needs.

Conclusions 

At Terlings Park a specialist function screening model has been established, with a single group having responsibility for developing and automating biochemical and cellular assays for lead generation, lead optimization, and HTS assay design and transfer to centralized uHTS. Close collaboration occurs with the local automated cell culture group, and automated compound management is fully integrated with the assay team.16 A combination of workstation and fully automated robots has been implemented; duplication of readers on sister robot systems has ensured key priority assays experience minimal downtime. Development of a new role of technology project manager for the site has ensured that resources are available to fully evaluate potential new systems, thereby maximizing benefits from the available budget. Successful collaboration with project teams and franchises has been aided by clarity in the division of responsibility, partly by using the communication and SOP processes outlined above, and also partly by the strong support and encouragement of departmental management.

Ensuring reliability of our automated systems has also been key to ensuring success—buying intelligently to upgrade our capabilities and reliability, and developing backup systems to ensure that individual system downtime does not excessively delay priority screening, have been central to this. Overall, the paradigm established at the U.K. Terlings Park site and similar approaches at two sister Merck sites, have come to be viewed as a model of efficiency, interdisciplinary cooperation and development of successful automated endeavors within the organization. The model can now be seen as suitable for expanding to large sites, broadening the benefits to be derived from the efficient assay centralization model, which we have established. This tutorial illustrates a paradigm in which an automated assay group can be built and can become expertly skilled, and appropriate automation resources can be installed, at relatively modest expense and recruitment levels and having widespread benefits to efficiency and productivity.