A battery of screening systems are under development to validate compounds
for further development.
By definition, the pharmaceutical industry is based on the ability to successfully
screen drug candidates. The approach of taking molecules from various sources
and testing them initially in some kind of nonhuman system that mimics a human
disease is a hallmark of modern drug discovery and the pharmaceutical industry's
growth.
With the advent of such technologies as combinatorial chemistry, however,
the traditional problems of lead generation no longer represent a bottleneck
in drug development. The real issue now is not how many new compounds can
one discover, but rather how many can one validate as potential therapeutic
candidates for further development—on a daily basis.
This so-called lead validation problem is so daunting that technologies
that accurately deliver a continue/discontinue prediction 70% of the time
are deemed "very successful" and highly sought after. Which technologies
will be able to break open this bottleneck in drug development?
Historical perspective The ability to rapidly screen drug candidates has developed dramatically
over the past 50 years. Screening targets for cancer is a good example of
this evolution.
The US National Cancer Institute (NCI; Bethesda, MD) began formal screening
efforts in 1955 to systematically find new anticancer compounds that had just
begun to appear in early experiments1. The NCI began this program
with mouse models of leukemia, sarcomas, and carcinomas. Yet after 20 years
of experimentation, NCI researchers found that they had developed agents that
cured cancer in mice but were largely ineffective in humans. From 1975 onward,
NCI researchers refocused their testing strategy on xenograft mouse models.
Human tumors were transplanted into immunosuppressed mice and compounds were
tested on their ability to limit or kill these tumors.
The problem with this approach was that many of the key disease events,
such as metastasis, did not occur in the xenograft models—limiting the
predictive value of the approach. To overcome this limitation, the next step
was to develop and test human cancer cell lines. The NCI currently has 60
human tumor lines, representing all major malignancies, which it uses to test
lead compounds. This approach is currently also employed by many drug-development
companies, but it must be remembered that these cell lines may not always
reflect human cancers, because they are chosen based on their ability to propagate
in vitro, outside the normal host environment. Nevertheless, the NCI's
screening based on human cell lines has produced about 5,000 agents with antitumor
activity1 from screens of over 60,000 potential anticancer compounds.
Of these 5,000 leads, an NCI database search that excludes those leads without
novel mechanisms of action has selected 1,200 of these compounds for further
testing.
Many high-throughput approaches today use automated one-to-one binding
assays. In this approach, a target of interest, such as an enzyme known to
be involved in a disease, is exposed to potential inhibitors in an enzyme-function
test. While this approach makes sense a priori—and forms the
cornerstone of nearly all current commercial validation programs—in
actual practice its accuracy is constrained by the fundamental assumption
that a good in vitro inhibitor of an enzyme will also work in the clinic.
To overcome this problem, the next generation of high-throughput lead-validation
platforms may be based on combinations of automated cellular assays, and cell-free
"molecular drug response cascades."
Current state The so-called clonogenic assay represents another aspect of the cellular
assay. Here, the patient's own tumor cells are cultured in vitro, and
potential drugs tested against them for ones that work specifically on that
cancer. Comparisons of a specific agent's effect on a xenograft mouse
model versus a clonogenic assay reveal significant differences. For example,
colon cancer cells in xenografts respond to radiation and disappear largely
because they are deficient in cell-cycle signaling checkpoint systems regulating
regrowth, whereas a clonogenic assay of similar cells is unaffected by the
checkpoint deficiency2.
Companies are now developing dedicated cell-based assays in order to automate
the process for more high-speed lead validation (see Table
1). These assays all attempt to more closely mimic a specific disease
state than is possible through xenograft or clongenic assays. For example,
Rational Therapeutics has developed an ex vivo apoptosis (EVA) assay, which
detects the differential morphology of tumor cells as they undergo cell death
(apoptosis) in response to anticancer agents3. Another company,
Synaptic Pharmaceutical inserts cloned human receptor genes into cell lines
and screens for compounds that bind to target receptor subtypes4.
Acacia Biosciences uses yeast cells to profile their response to lead compounds
at the level of specific gene mutations, and also phenotypic responses5. Exelixis Pharmaceuticals uses whole model organisms whose genome
is essentially known, such as the fruit fly Drosophila melanogaster
and the nematode Caenorhabditis elegans, to test mutations and phenotypic
responses as lead validation screens. Small Molecule Therapeutics (now MorphoChem)
is developing microbial screens and assays based on the transmembrane receptor
tyrosine kinase system.
Table 1. Selected screening systems for lead validation
In addition to these cell-line based assays, probing further refinement
of the one-to-one intermolecular interactions approach is proving to be a
key component of many ambitious screening programs. Most of these approaches
employ technologies that allow more sensitive detection and quantification
in a high-throughput format.
For example, confocal microscopy enables the detection of faint and/or
small fluorescent signals by eliminating fluorescent light that is outside
the plane of focus. Evotec BioSystems is incorporating a version of confocal
microscopy called fluorescence correlation spectroscopy (FCS) into an automated
ultrahigh-throughput screening system that the company calls EVOscreen. It
enables the quantification of parameters such as fluorescence half-life, energy
transfer, brightness, and spectral shift, which can be interpreted by customized
software as molecular binding constants, polymerization of molecules, conformational
changes in proteins, or cellular events such as transcription, signal transduction,
or endocytosis.
Industry challenges The key lead validation challenge is the development of high-throughput
screens that also provide clinically useful information. Inevitably, there
is a tradeoff between the ease and speed of a test, and value of the information
it gives. For screens based on intermolecular interactions where the target
is either free in solution or immobilized on a surface, miniaturization and
robotic liquid handling present a technical limitation. For these assays,
the synthesis of appropriately labeled molecules for fluorescence or radiometric
measurements is not always easy. Molecules derivatized in such ways often
lose the specific biological activities that made them useful tools in the
first place. Recombinant methods often help in such cases. For example, DNA
topoisomerase has been engineered to contain a biotin segment that enabled
the cloned fusion protein to be purified from crude cell extracts by streptavidin-coated
scintillation proximity beads. This assay allowed the enzyme to be purified
and also to be screened with inhibitor compounds6. Several dedicated
instruments have been developed to handle nanoliter volumes in multi-well
plate formats and also on chips. Companies that have developed such systems
include Evotec, Cartesian Engineering (Durham, NC), Packard Instrument Company
(Palo Alto, CA), Caliper Technologies, Aurora Biosciences, LJL BioSystems,
and others.
Another challenge to the industry is that lead validation assays are limited
by the intrinsic difficulties of the particular cell or organism models they
depend on. Such assays are essentially determined by the nature of the biological
event that the drug lead will play a role in. Sometimes the event is simulated
in a cell model, and in this case the assay is limited by the inherent difficulties
of the model. For example, cell death or apoptosis is a major biological event
that correlates with cancer, where the latter actually inhibits normally occurring
apoptosis. Here, the idea is to develop anti-cancer agents that induce apoptosis
in cancer cells. Lead validation assays, therefore, focus on apoptosis-specific
events, such as the expression of apoptosis-causing genes in cell models7. The limitation here is that these assays are transient transfection
ones, where the apoptosis genes are induced transiently and often produce
too much of the desired protein, thus invalidating the assay. There is, therefore,
a constant effort to improve these assays and make them as physiologically
relevant as possible.
The future Advances in our understanding of molecular pathways and cascades within
cells, such as signal transduction and apoptosis, which are at the heart of
many disease conditions, promise to provide highly specific assays for disease
states. Inappropriate signal transduction cascades, for example, are implicated
in cancer, autoimmune, inflammatory, neurologic, and cardiovascular disorders.
Reagents that enable the dissection of these cascades, such as highly specific
antibodies and small molecule modulators, are becoming available from companies
such as Upstate Biotechnology (Lake Placid, NY) and others. These reagents
can be used together with standardized cell lines in automated high-throughput
formats to validate potential leads in terms of specific effects on disease-specific
molecular cascades. Whether cell-based or cell-free, such molecular drug response
cascades are likely to offer information that is more clinically relevant
than current high-throughput validation systems, and offer, therefore, significant
promise toward solving or alleviating the lead validation issue of drug development.
In addition, lead validation assays will take more forms than at present
and benefit from novel approaches. For example, the use of quantitative polymerase
chain reaction (PCR) in animal models of leukemia can help predict the survival
of the animals when treated with different drugs8. Here, quantitative
PCR is used to measure tumor burden in the animals, which will decrease as
a result of any therapeutic benefit from anti-cancer leads.
Finally, the increasing integration of lead validation assays into systematic
lead development efforts will also become more commonplace. For example, DNA
gyrase catalyzes the condensation of the DNA structure, and is therefore a
significant target for novel drug discovery of anti-cancer agents, anti-microbials,
and anti-virals. In a recent report, random screening of lead libraries failed
to produce good lead inhibitors of the enzyme. However, the integration of
in silico screening for novel small inhibitors, lead validation assays based
on biased high-throughput DNA gyrase screens and biophysical methods, and
a systematic lead optimization process has produced an inhibitor 10-fold more
potent than the known inhibitor novobiocin9.
Conclusions With combinatorial chemistry libraries nearly ubiquitous in the drug-development
community, millions of potential drug leads are being screened (see Combinatorial
chemistry, pp. 50−52).
The methods described here are aimed at providing critical data that will
help decide which, if any, will be developed further as therapeutics. The
central issue these methods must address is whether they provide information
that is clinically relevant.
For high-throughput screens based on the interaction of a target molecule
with lead compounds in isolation, there is little clinical context to lend
confidence to their results, unless they are highly specific in mimicking
a disease state. At the other end of the scale, cell lines probed for genotypic
and phenotypic responses incorporate more of a clinical context than two molecules
in solution, but are not amenable at present to standardized high throughput.
The future is very exciting in this area, because of advances in instrumentation
that enable the handling of very small quantities of cells and reagents, and
also the development of various detection methods that report the occurrence
of specific molecular events. It is most likely that it will be a synthesis
of the information derived from as many of these screening techniques as possible
that will provide the most important insights into the behavior of a lead
in a clinical setting.
Reprinted from Nature Biotechnology 16, 100−101
(1998).
