(Issue 7 ? posted May 2, 1997; archived May 30, 1997)
Abstract
A successful drug must not only be active against its target, but also
nontoxic. Toxicity testing in
animals is expensive and morally questionable, so any method for picking
up toxicities early in drug
development would be valuable. Xenometrix, Inc. is using the responses of
single genes to stress as a
marker for toxicity in an attempt to use single cells as the test
organism.
Stress is on the minds of pharmaceutical companies these
days. The stress they are
worried about is not, however, that suffered by their chemists, working
long hours
to find the elusive next drug. Rather it is the stress responses of single
cells,
which Xenometrix, Inc. (Boulder,
Colorado) is using
in a reductionist approach to toxicology.
As pharmaceutical chemists amass huge collections of
chemicals, they cannot hope to
test them all in animals. Using rows upon rows of animals would not be
acceptable,
either ethically or financially.
To fish out the few active compounds, companies have
resorted to assays that test the
responses of cultured cells or purified proteins. But even when a compound
appears to
be active by this measure, it may still be unacceptably toxic. Xenometrix
hopes to reveal
these toxicities by examining which of the candidate drugs alters the
expression of genes
that are induced by stress.
From Ames to Ames II
The first person to substitute single cells for animals in
toxicology was Bruce Ames
of
the University of California at Berkeley. In the 1960s, Ames developed a
strain
of Salmonella typhimurium to test for the mutagenicity (and so
carcinogenicity) of
compounds. The strain carried a mutation in one of its histidine synthesis
genes that could
be effectively repaired by any one of several missense mutations. Ames
exposed these
bacteria to various chemicals and found that the number of bacteria that
were then able
to make histidine was proportional to the ability of the chemicals to
induce carcinogenic
mutations in animal models.
The Ames II assay, developed by Pauline Gee in Ames's
laboratory, is now part of the
Xenometrix product line. Gee constructed six Salmonella strains,
each of which
can regain the ability to synthesize histidine only by one of six possible
missense
mutations. Two other strains, developed as part of the original Ames test,
detect a
one- or two-base pair frameshift. Mutagenic compounds vary in the types of
mutations that
they introduce, so the pattern of histidine synthesis in the test strains
acts as a
fingerprint for each compound.
Stress alerts
Detecting the mutagenic potential of compounds is a
routine procedure in drug
companies,
but there is more to toxicity than mutagenesis. Chemicals can damage cells
and organisms in
many ways, and cells have, in turn, developed mechanisms to detect and
counteract that
damage. This is where the Xenometrix leap of faith comes in. "We're
looking at [the induction
of] genes - the first response that cells have - and we think that is
indicative of the
metabolic response further on," says Gee, who is now the vice
president for research and
development at Xenometrix.
Each Xenometrix assay uses a bank of inducible promoters,
some of which are synthetic
promoters with the active elements from more complex promoters. Each
promoter and its
associated reporter gene is inserted into an individual bacterial or human
cell line.
The readout is the activity or presence of reporter gene products.
A simplified list of the toxic effects detected by two of
the Xenometrix assays is
given
in Table 1. The real results, however, are more
complicated. Any
given chemical may cause several types of damage, and genes typically
respond to a
complicated combination of signals, some of which have not been
characterized. "The surprises
are probably less than 10%," says Gee, "but it's those that you
notice, and you spend 90%
of your time on them." One way of making sense of certain responses
is to look at the kinetics
of induction in an attempt to determine which responses are primary and
which follow the
induction of other genes.
But does the system work?
Just as the Ames assay had to struggle for recognition as
a measure of genotoxicity,
stress gene induction must be validated as a good indicator of other
toxicities. Most
toxicologists are enthusiastic about the general approach, but have
reservations about applying
the methods. "I don't think we know enough about what the readout
means,"
says Richard
Morimoto of
Northwestern University. The threshold of damage that cells respond to is
poorly
characterized, and so the significance of the signal is hard to interpret.
"These monitors
may be a little too sensitive," says Morimoto.
Jeffrey Theiss, the director of molecular toxicology at
Parke-Davis, believes that
more work is needed. "Is the signal reflective of a particular
toxicity, or of adaptive
changes that protect the cell from toxicity down the road?" he asks.
"We don't really
know that yet."
Gee agrees with these caveats, but feels that Xenometrix
is responding adequately by
building
a database - a survey of the responses to drugs whose toxicities are
known. "We have to
decide at what point [the drug has] overwhelmed the cell," she
explains, "and we are
working to decide what that cutoff might be."
To help interpret the mountain of data, Xenometrix is
developing pattern recognition
software. Based on the assay results of compounds with known toxicities,
the computer
designs rules for predicting likely toxicities of novel compounds. This
eliminates the bias
that could come from specific scientific hypotheses.
Bacteria or human
Xenometrix produces the stress gene assays in both human
cultured cells and bacteria.
The
bacterial assays have several advantages over the human assays, including
promoters that
are sensitive to osmotic stress, an oxidative stress response that is
better understood, and
simplicity. To mimic some of the metabolic reactions that happen in
humans, a liver homogenate
is added to these assays, and a cell-wall mutation is used to increase
permeability to
drugs.
The human assays use more complicated culture conditions,
but they have the obvious
advantage
of being closer to the clinical situation, an advantage they also have over various animal
assays.
Applications
Gee estimates that, by next year, approximately 60% of
Xenometrix's business will
come from pharmaceutical and biotechnology companies, with the remainder
split between cosmetic,
chemical, and environmental companies. For pharmaceutical companies,
explains Theiss,
the assays can be used in one of two settings. "If you have a drug
pretty far along in
development and identify a toxicity," he says, "with these
systems you can understand why a
toxicity is there." Once the mechanism of toxicity is identified,
researchers can try to
rationally modify the drug to remove the toxicity while retaining the
therapeutic
properties.
The second application comes earlier in drug development,
when the number of lead
compounds
is being whittled down. The stakes here are higher, and the drug companies
are correspondingly
more cautious about implementing the new assays. "If we develop a
high level of confidence,
then we could test compounds early on," says Theiss. But a false
positive in a toxicity
assay could result in a valuable lead being discarded. (This may explain
the low level of
competition in this field; see sidebar.)
Theoretically, these assays could be used to replace
either animal or human testing,
but neither Xenometrix nor its customers feel that this is imminent. The
assays can, however,
reduce the number of compounds that need to be tested in animals by first
eliminating those
that are obviously toxic. And by using more stress-related genes (a gene
discovery effort is
now beginning) and more powerful interpretive software, the detection of
stress seems set to
become a central feature in drug discovery.
William Wells, Ph.D., is a scientific journalist with Biotext, Ltd. in San Francisco, California.
Previous Profile Articles:
What's your view? Use our Feedback to let us know
privately, or to post a Letter to HMS Beagle.
Virtual Toxicology
The Ames II assay, developed by Pauline Gee in Ames's
laboratory, is now part of the
Xenometrix product line. Gee constructed six Salmonella strains,
each of which
can regain the ability to synthesize histidine only by one of six possible
missense
mutations. Two other strains, developed as part of the original Ames test,
detect a
one- or two-base pair frameshift. Mutagenic compounds vary in the types of
mutations that
they introduce, so the pattern of histidine synthesis in the test strains
acts as a
fingerprint for each compound.
A simplified list of the toxic effects detected by two of
the Xenometrix assays is
given
in 
Gee agrees with these caveats, but feels that Xenometrix
is responding adequately by
building
a database - a survey of the responses to drugs whose toxicities are
known. "We have to
decide at what point [the drug has] overwhelmed the cell," she
explains, "and we are
working to decide what that cutoff might be."
The second application comes earlier in drug development,
when the number of lead
compounds
is being whittled down. The stakes here are higher, and the drug companies
are correspondingly
more cautious about implementing the new assays. "If we develop a
high level of confidence,
then we could test compounds early on," says Theiss. But a false
positive in a toxicity
assay could result in a valuable lead being discarded. (This may explain
the low level of
competition in this field; see 
Previous Profile Articles: