Kenneth C. Saunders^,

Automation Team, Department of Pharmacokinetics Dynamics and Metabolism, Pfizer Global Research and Development, Ramsgate Road, Sandwich, Kent, UK CT13 9NJ

The use of automated sample processing, analytics and screening technology for profiling absorption, distribution, metabolism and excretion (ADME) and physicochemical properties, early in the drug discovery process, is becoming more widespread. The use and application of these technologies is both diverse and innovative. High-throughput screening (HTS) technologies have been utilised enabling the profiling of an increased number of compounds emerging from the drug discovery process. Although the drivers for using these technologies are common, different approaches can be taken.
 

Section Editors:

Han van de Waterbeemd, Christopher Kohl – Pfizer Global Research & Development, Sandwich Laboratories, PDM (Pharmacokinetics, Dynamics and Metabolism), Sandwich, Kent, UK CT13 9NJ

The increasing demand for high-quality ADME and physicochemical data in early drug discovery have resulted in semi- or fully automated robotic systems for the measurement of the key properties in medium to high-throughput. Based on many years experience in the field of automation and the bioanalysis of ADME samples, Ken Saunders presents here the state-of-the-art in ADME screening technology. Modern systems need to be balanced between good data quality, high-throughput, and flexibility to allow easy addition or removal of particular assays. The move to automated high-throughput ADME data generation system creates an increased demand for improved data analysis tools and in silico predictive models.

Introduction
There is an increasing demand to profile more in vitro absorption, distribution, metabolism and excretion (ADME; see Glossary) and physicochemical properties of newly synthesised compounds early during a discovery program 1 and 2. Chemistry departments in pharmaceutical companies have the capacity to generate hundreds of compounds per day, thus creating a requirement for innovative automated high-throughput solutions to provide these data in a rapid manner. Many different automated platforms have been developed and the assays have been adapted to operate at high capacities [3]. High-throughput liquid chromatography–mass spectrometry (LC–MS; see Glossary) has accelerated the development of ADME assays in recent years and they tend to be configured in micro-titre plate formats [4]. The sensitivity, selectivity and speed of this technique have enabled samples from CYTOCHROME P450 (see Glossary) and permeability assays, such as CACO-2 (see Glossary) and parallel artificial membrane permeation assay (PAMPA; see Glossary), to be analysed.

Although work in this field has been focused on assay miniaturisation and the analytical end-points, automating the whole process has advantages including throughput, robustness and release of resources. Automated systems have been developed for single assays or a multiple assay process [5]. All components of this type of ADME process can be automated (sample submission through to result reporting).

Automation approaches
One of the emerging areas of robotics and automation is in the profiling of in vitro ADME and physicochemical properties. Automation for use in drug metabolism and pharmacokinetic (DMPK; see Glossary) studies can be exploited to varying degrees [6]. Complex integrated systems are available that are specialised for high-capacity and simple protocols. This type of system is widely used in biological throughput screening (HTS; see Glossary). The high sample capacity of these systems is optimised for the vast compound numbers encountered in biological HTS (i.e. multiple compounds per screen) [7]. ADME screens are configured for multiple screens per compound and some of the assays cannot be miniaturised to use this type of instrumentation. The most common approach used in DMPK studies is semi-automated; this approach uses individual workstations, such as 96-well instruments, and has been established in many DMPK departments [8]. It has the advantage of being flexible, enabling manual intervention at the main stages of the assay for tasks that include loading reagents and centrifugation. Unfortunately, this is also a disadvantage because unattended operation is often preferred, particularly with respect to loading and preparing samples for analysis.

An alternative approach reported is “industrial automation”, which is used by companies such as Cyprotex ( http://www.cyprotex.com / ), EvotecOAI ( http://www.evotecoai.com  / ), The Automation Partnership ( http://www.automationpartnership.com  / ) and other pharmaceutical companies. This approach uses a combination of HTS technologies that have been integrated into the relevant stages of the ADME screens, such as plate replication, reagent addition and analysis. These companies use leading-edge screening technologies to run traditional in vitro ADME tests via a “HTS-like” process. This approach can provide a big increase in throughput and can also reduce running costs (after the initial capital outlay).

A fully automated approach can be taken, using a centralised robotic platform (or arm) integrated with ADME workstations, which can carry out single or multiple assays [9]. These systems are designed for fully automated sample preparation, data analysis and management of results generated from in vitro ADME assays. Fully automated robots have also been integrated with sample submission and the LC–MS analytical systems [Green, C.E. et al. (2004) ALIAS: an automated laboratory in vitro assay system for candidate selection. Abstracts of Papers ELRIG: Next Meeting on ADME Advances]. Systems with a centralised robotic arm combine a series of modular assay work stations to provide a robotic platform, which is flexible, upgradeable and easily reconfigured when assays change or are newly developed (Fig. 1).

(59K)

The scheduling and programming of these robots is key to their success and unattended, parallel and overnight operation is routine. Facile sample submission and data reporting “on the fly” can be done remotely via a web interface. Although the capital outlay for these systems is initially high, the benefits of using less manpower, more time to visualise the data and more efficient use of a scientist's time are the primary advantages. These systems give a fully integrated approach to high-throughput ADME evaluation in support of drug discovery.

Parallel approaches to both sample preparation and analysis are widely employed to improve throughput and efficiency of in vitro ADME processes. Automated systems can be programmed and scheduled to prioritise multiple tasks in a procedure. These systems are also programmed and integrated with analytical instruments, such as plate readers or mass spectrometers, allowing unattended loading and analyses of samples.

Parallel approaches are most commonly employed with LC–MS systems and analysis of samples generated from in vitro ADME screens. Multiplexing both chromatography systems and ion sources are commonly used in this field. Multiple samples (usually four or eight samples) are injected onto a parallel LC system simultaneously and then analysed using a parallel ion source [10]. Increased throughput can be obtained by incorporating cassette experiments into the assay protocol (whereby more than one compound is added to the experiment at once) or sample pooling (cassette analysis strategies, see below). Mass spectrometers can monitor multiple analytes simultaneously in a sample mixture without significantly compromising sensitivity. This feature enables the dosing of multiple compounds in certain experiments and monitoring for all these analytes, which provides significant gains in efficiency.

An alternative strategy is to combine the samples produced for analysis at the end of the ADME experiment (sample pooling strategy) [11]. Cassette and pooling strategies have disadvantages, such as compounds interfering with each other in either the experiment or the detection system. Nonetheless, they can be applied successfully provided that safeguards are put in place during the sample submission, preparation and data acquisition processes.

Sample management and submission systems
The systems used for sample tracking and submission of liquid or solid compounds in automated ADME is pivotal. Most pharmaceutical companies now have liquid or solid compound sample banks that allow compounds to be provided in a flexible micro-plate format with an identifying bar code. The compounds must be submitted to the DMPK laboratory in solid or liquid form and all the appropriate information tracked and reported. This can be done in a variety of ways using databases, sample banks and web sites, which provide facile methods of sample submission. The compound sample stock solutions are normally provided in micro-titre plates in the form of low volume, concentrated solutions in a solvent such as dimethylsulphoxide (DMSO; see Glossary). The sample plates are then entered into the appropriate systems and processed accordingly.

LC–MS
High-performance liquid chromatography (HPLC; see Glossary) coupled to mass spectrometry is the analytical technique that is most widely used within this expanding area of science and technology. LC–MS is ideally suited for pharmaceutical compounds owing to the typical compound physicochemical properties and molecular weight. LC–MS can accept small volume samples in a micro-titre format and monitor multiple analytes in the same sample in complex mixtures such as biomatrices. Furthermore, LC–MS does not rely on the property of a compound for detection, such as a UV chromophore or fluorescence. Mass spectrometers are relatively large, expensive to purchase and costly to operate. This issue has driven innovative research into the more efficient use of LC–MS or replacement of this detection technique with cheaper alternatives.

In summary, this technique is employed for in vitro ADME analyses owing to its all-round capability to detect most compounds, enhanced sensitivity, selectivity and ease of automation when compared with traditional analytical techniques. LC–MS in a high-throughput mode is the end-point most commonly used to quantify the vast number of samples generated from these screens. Innovative ways to both analyse and process these samples by LC–MS have been reported 12 and 13. Significant effort continues in this area to reduce analysis times. This can be done in a variety of ways by multiplexing, mixing samples, using fast chromatography and HPLC column switching.

In this regard, analytic throughput has improved significantly by shortening HPLC run time. Analysis times have now been reduced to around 15 s per sample to assess ADME properties, such as metabolic clearance in liver micosomes. This is achieved by using fast gradient column switched HPLC systems, which elute the analytes into the MS detector in a very short time.

In accordance with quicker run times, fast, automated data processing of the analytical MS data is a key component in providing rapid turn around of data to discovery project teams. The first part of the process is the rapid determination of mass spectrometric parameters of each single compound for quantification purposes [14]. Once the MS conditions have been determined the samples are analysed and acquired as efficiently as possible. The data can then be processed and ADME results reported. This can be done in a number of ways but customised automated software packages have been developed to allow the processing and extraction of the relevant information as quickly as possible [15]. The speed of analysis can therefore be the key to the throughput of these assays and new ionisation techniques are being researched to allow samples to be analysed in less than a second using LC–MS coupled to a matrix-assisted laser desorption ionisation (MALDI; see Glossary) type ionisation source [Cole, M.J. et al. (2004) Characterization and performance of MALDI on a triple quadrupole mass spectrometer for analysis and quantitation of small molecules. Abstracts of Papers, 227th ACS National Meeting ANYL-014]. This type of system has the potential to eliminate the LC–MS analysis time as a bottleneck.

Physicochemical methods
Solubility
Compound solubility is one of the most important properties measured at an early stage. Low-solubility compounds are more difficult to develop and produce variable data from screens, such as Caco-2 and lipophilicity. Therefore, a rapid low-cost method for determining solubility before running the more costly ADME screens is a useful tool.

Turbidimetry is an established method often used as a quick and simple assessment of solubility [16]. Compounds dissolved in DMSO are diluted with aqueous buffer and the degree precipitation is measured. This method is robust but consumes more material than laser nephelometry, which has been shown to be a reliable and sensitive technique for the measurement of kinetic solubility in 96-well or 384-well plate format [17]. This process can easily be automated on a liquid sample processor, such as a Tecan Genesis (Tecan, http://www.tecan.com/ ).

Thermodynamic or equilibrium solubility is considered the gold standard procedure for measuring solubility. This method, although accurate, consumes far more compound and requires over 24 h shaking or equilibration and therefore has relatively poor throughput for screening purposes. As such, this method is generally used on development compounds where solubility and dissolution of the dose is more important and there is more crystalline material available.

Other dedicated hybrid automated systems have been developed by Sirius (http://www.sirius-analytical.com/) to measure solubility, such the new CheqSol instrument. These specialised instruments are optimised to measure and profile solubility in high- to medium-throughput mode. Depending on the scientists’ requirements, they also provide the flexibility to use different approaches to assess solubility.

pKa (ionisation constant)
Measuring the dissociation constant(s) or the PKA (see Glossary) of new discovery compounds can be an important parameter when considering the charged state of compound in the body, as this can affect the distribution, solubility and permeability of the compound.

Potentiometric titration methods are well established 18 and 19, but the amount of compound required and capacity restrictions limit their use during early discovery. Hence, alternative strategies have been used to allow more rapid and automated screening using capillary electrophoresis [20] or high-throughput pKa instruments [21] available from CombiSep (Presearch, http://www.presearch.co.uk/) and Sirius, respectively.

Lipophilicity
Distribution coefficient (LOGD; see Glossary) and partition coefficient (LOGP; see Glossary) are the most common lipophilicity parameters determined in drug discovery. The data generated are used to rank novel compounds in terms of their lipophilicity and help predict parameters such as in vivo permeability and metabolic stability. The assay involves measuring the distribution of a compound between a lipophilic phase, such as octanol, and an aqueous phase, such as water or buffer, at physiological pH. The established approach is to use the shake-flask methodology in which a compound is partitioned and then quantified in an octanol-aqueous system [22]. This process can be easily automated on a liquid sample processor and quantification can be carried out using HPLC–MS or HPLC–UV end-points. log D or log P has been estimated with success using HPLC retention factors [23]. This determination is completely automated and involves injecting compound solutions onto a reversed-phase HPLC system, they are then separated using a lipophilic HPLC column and detected using UV detection. Sirius provides instruments that operate automated procedures measuring both log P and pKa utilising titrimetric methodology [24]. This is considered a gold standard determination for pKa. Also available are more sensitive and high-throughput instruments that use a UV end-point. log P and pKa can both be determined using these instruments and log D can then be calculated from these measured values.

Permeability and drug absorption using trans-well assays
The most common high-throughput in vitro permeability assays use either artificial membranes or cell-based trans-well systems. PAMPA offers a fast robust and cost-effective method for assessing the intrinsic permeability of novel discovery compounds 25 and 26. Phospholipids are added to an organic solvent, typically dodecane, on a trans-well filter plate to mimic an intestinal membrane. Compounds are dosed to one side of the trans-well plate at physiological pH and the apparent permeability is measured as flux across the lipid layer over a period of over time. This assay is configured in a 96-well format and has been automated on a single workstation (robotic liquid handler). The assay is easily automated and the analysis performed using either a UV plate reader or mass spectrometer. In pharmaceutical companies this type of assay tends to be used during early discovery, in a high-throughput configuration (rather than other cell-based assays), as it has the ability to rank compounds in terms of intrinsic permeability in a cost-effective manner.

In vitro ADME methods
Caco-2 (or MDCK) trans-well assay
The most established cell-based assay for estimating intestinal permeability uses the Caco-2 [27] or Madin Darby Canine Kidney cells (MDCK; see Glossary) cell line. Caco-2 cells have the advantage that they express a variety of transport systems present in the human intestine. The cells must be cultured and then grown on the trans-well filter plates before carrying out the assay. The recommended culture time is 18–21 days to ensure that all the necessary transporters are present. MDCK cells can be used in a similar manner to Caco-2 cells and have a shorter culture time (approximately five days), but have the disadvantage in that they are “non-human” in nature. The trans-well plates are used in either a 24-well or more recently in a 96-well format. The permeability experiment is carried out in a bi-directional mode to estimate the contribution of any active transport processes. Compound is dosed to either side of the trans-well plate chambers and the rate of transport (Papp) is estimated after 1–2 h. Samples from the assay are usually quantified using LC–MS, as each compound can produce up to 20 samples for analysis. This assay has been automated on a number of platforms including Beckman Biomek liquid handlers (Beckman-coulter, http://www.beckmancoulter.com/). Automating the assay procedure and cell culture process and optimising the analytics have enhanced the throughput. However, the disadvantages of this method are the fairly complex process of automation and high-throughput and the costs of producing the cultured plates.

Metabolic clearance
Prediction of metabolic clearance from in vitro methods is one of the main screening tools used by most ADME groups 15 and 28. Measurement of a discovery compound's metabolic stability in liver microsomes or hepatocytes can be a very useful tool in predicting desirable drug-like properties [29]. Compounds are incubated with liver microsomes or hepatocytes and their disappearance measured over time, from which the clearance can be derived. This type of assay has been automated on liquid sample processors fitted with suitable incubators that are programmed to sample at set time points and the metabolism quenched with an organic solvent, such as acetonitrile. Quantification is usually performed using LC–MS.

Cytochrome P450 inhibition and drug–drug interactions (DDI) assays
Compounds that are potent inhibitors of the major drug metabolising enzymes (cytochrome P450s) can lead to drugs that yield undesirable drug–drug interactions in the clinic when co-administered with other medication. Therefore, compounds are screened early in discovery to identify the major metabolic pathways and any potential cytochrome P450 inhibition. Assays are designed to monitor these potential DDIS (see Glossary) with human microsomes or recombinant cytochrome P450 using specific probe substrates with well-characterised metabolic pathways. New discovery compounds are co-incubated with these test substrates and the formation of metabolites monitored, allowing the assessment of the any potential DDI. Traditionally this has been carried out with quantification of the substrates and metabolites by LC–MS [30] and the assay procedure can be easily automated. A more recent development uses an “inhibition cocktail” approach for high-throughput inhibition screening of the major human cytochrome P450 enzymes [31]. This assay is configured using ascending substrate standard concentrations (as a cocktail or mixture) for quantification, alternatively a single substrate cocktail concentration may be used, thus allowing even higher throughput. Fluorescence-based approaches are now widely used [32] because these have the advantage of high capacity with no requirement for the more costly LC–MS analysis. This screen can be easily automated in a 384-well plate format, as this set up is similar to many high-throughput biological screens.

Conclusions
The increase in new chemical entities emerging from discovery is a result of high-throughput medicinal chemistry, such as parallel synthesis and the associated biological screening of the compounds for potency. The recent strategy to define ADME and physicochemical properties early on has driven the requirement for high-capacity ADME assays [33].

Improvements in the capacity, speed and efficiency of the drug discovery process are constantly being sought after in the pharmaceutical industry.

The philosophy is that profiling these “drug-like” properties early on will identify potential development liabilities, thereby increasing efficiency and reduce compound attrition later on in a compound's development (where the cost is exponentially more) [34]. The modern paradigm of placing in vitro ADME and physicochemical screens during early discovery processes has led to increased pressures on ADME groups to develop automated solutions that address the bottlenecks in sample handling, analysis and data reporting. These automated solutions and technologies have advantages and disadvantages (Table 1). Although automated solutions for in vitro ADME assays should in reality yield an increase in capacity, accuracy, throughput, reliability and lower costs, this might not necessarily be the case [35]. Laboratory automation will continue to shape the future of in vitro ADME techniques used in early discovery. It will continue to help optimise efficiencies using increased throughput and miniaturisation. Miniaturisation and HTS is continually pushing the boundaries of screening science 7 and 36 and this will continue to impact this field in the future. Implementation of automation gives benefits of throughput and reliability. Automated systems can track samples throughout the assay process (using bar-coding and database technologies). They can also be much quicker, for example, using micro titre plate technology. This is particularly true of processes that can be done in parallel. Other advantages include unattended (overnight) operation, better use of analytical resources, reduced inter-operator variability, and releasing resources for other tasks. The development of sound automated ADME and physicochemical procedures requires a diverse skill set from computer programming to the hands on science of DMPK. In vitro ADME procedures are often lengthy and relatively complex because they require extensive optimisation to automate them. The trend nowadays is therefore to use cheaper and simpler procures wherever possible, for example, using a PAMPA screen rather than a Caco-2 assay for high-throughput. ADME assays are relatively expensive in terms of resources, reagents and detection techniques, and hence it is too costly to screen every compound synthesised. As a result, big pharmaceutical companies are using in silico approaches in combination with (or in place of) these more expensive automated in vitro systems.
 

Links

• http://labautomation.org /: LabAutomation, conference and exhibition on emerging laboratory technologies

• http://www.elrig.org /: European Laboratory Robotics Interest Group
 

Glossary :
ADME : absorption, distribution, metabolism and excretion.
Caco-2 cells :human colon adenocarcinoma cells.
Cytochrome P450 :major sub-family of isoenzymes present in the liver responsible for the metabolism of drug-like compounds.
DDI :drug–drug interaction with respect to co-administered medication and drug-metabolising enzymes such as cytochrome P450.
DMPK :drug metabolism and pharmacokinetics.
DMSO :dimethylsulphoxide.
HPLC : high-performance liquid chromatography.
HTS :high-throughput screening.
log D :the logarithm of the distribution coefficient at a certain pH.
log P : the logarithm of the partition coefficient at a neutral pH.
MALDI :matrix-assisted laser desorption ionisation.
MDCK :Madin Darby Canine Kidney cells.
LC–MS :liquid chromatography–mass spectrometry.
PAMPA :parallel artificial membrane permeation assay.
pKa :the ionisation constant for a particular compound

Source : sciencedirect