IT’S ALMOST ALIVE. The swollen heart on Denis Noble’s screen throbs and
quivers as he hammers at the keyboard. He hits one key and it pounds furiously;
another and it writhes as if in agony, kicking up a hurricane of dazzling
colours. Watching the display, you would think that Noble, professor of
physiology at the University of Oxford, is hooked on the latest computer
game.
But this is no mere game. This heart runs on Europe’s largest supercomputer,
and might just save your life.
If you could look inside that computer, you would find it struggling with 30
million intricate equations. And if you could put the heart under a microscope,
you’d find more than a million virtual cells—each with its own complex
internal biochemistry—feeding on sugar molecules and burning oxygen. Each
cell would have its own ion pumps and channels, and be joined to its neighbours
by connective tissue that comes complete with a blood supply. Noble and his
colleagues have created a model of such staggering detail that it behaves almost
exactly like a living human heart.
Advertisement
Already, this virtual heart can develop the bouts of irregular beating that
precede heart attacks—or be saved from them by timely treatment. It is so
lifelike that for the first time researchers can see how drugs affect the
performance of the heart without ever dosing up a real human being. The idea? To
bring the logic of life into the lab while leaving flesh and bones and the
capacity for suffering far behind. Today, it’s happening for the heart. And in a
few years, Noble insists, researchers will be stitching heart, lungs and other
organs together into an electronic Frankenstein—a functioning and
realistic virtual corpus.
That’s not exactly what Noble had in mind when he started, 35 years ago. His
aim then was simply to model the workings of a single cell. That in itself took
more than two decades to get right. He had to become a proficient mathematician
first. In the meantime, computers got faster. Noble soon found himself putting
his cells together, and building the framework of a model that would act just
like a living heart. Get the molecular details right, he argues, and the cell
should work. Get the cell right, and the organ should follow suit. At every
step, he and his colleagues do exacting experiments to be sure that the model is
lifelike.
This painstaking approach is beginning to pay dividends. Two years ago, for
example, the multinational pharmaceuticals giant Hoffman-LaRoche approached
Noble and his colleagues to help with a problem. During the approval process for
the company’s high blood pressure treatment mibefradil, the US Food and Drug
Administration had noticed a glitch in the ECGs of people taking the drug.
“Clinicians told them that the drug was dangerous,” Noble recalls. And
understandably so. The glitch appeared at the end of the heart’s contraction
cycle at a point associated with a deadly cardiac malfunction. Ordinarily,
Hoffman-LaRoche would have scrapped the drug or started more, costly clinical
trials.
Applying the drug to their virtual heart, the researchers found that it too
developed the same glitch. And with the model, they could “get inside”
individual cells to pinpoint the cause of the trouble. They found that the
glitch was not a sign of major malfunction. “We were able to satisfy the FDA
that it wouldn’t cause any trouble,” says Noble, and consequently, the drug was
approved.
The FDA was impressed. “The results of the model were presented at a meeting
of the FDA advisory committee, and people listened,” recalls Ray Lipicky,
chairman of the FDA’s committee charged with approving new cardiac drugs. “That
alone is testimony to how far this approach has come. Ten years ago, no one
would have listened.” As more doctors and researchers begin to pay attention,
the virtual heart should increasingly replace real hearts in research.
But if the real power of the model lies in gauging the performance of the
whole organ, the secret to its success lies at a lower level. Heart muscle cells
are wonders of complexity, and capturing the essence of that complexity in
mathematics is Noble’s forte.
Long and thin, each cell contains fibre-like proteins capable of contracting,
and a membrane to separate the cell from the outside. Proteins embedded in the
membrane carry crucial signals or materials from one side to the other. Tiny
pumps, for example, push positive and negative ions across the membrane and
maintain different concentrations of ions inside and outside the cell. This
“polarises” the cell, creating a voltage difference between the inside and
outside, known as the membrane potential. Other proteins in the membrane, called
ion channels, act as floodgates. When they open, ions rush through
(see Diagram).
Heart cells contract by virtue of the delicate interplay of the ion pumps and
channels and the chemical and electrical differences across the membrane. Take,
for example, a cell in one of the heart’s ventricles, the larger chambers in the
lower half of the heart. At rest, the membrane potential in each cell is
negative—the inside is more negative than the outside.
Swarming calcium
But things change fast if an electrical stimulus from a neighbouring cell
momentarily decreases this difference, depolarising the cell. When that happens,
sodium channels suddenly open and sodium ions (Na+) pour into the cell. This,
in turn, triggers calcium channels to open. And when these calcium ions (Ca2+)
swarm round the cell’s fibrous proteins, it contracts.
At this point, the sodium and calcium channels close, and ion pumps begin to
push ions back out of the cell to restore the cell to its original state. In a
healthy cell, the cycle happens in a fraction of a second.
That’s the basic picture. The devil, as ever, is in the detail. For there are
several kinds of channels for both sodium and calcium ions. There are other
important channels, too, for potassium and chloride ions, each of which opens
and closes at its own pace, and in response to characteristic signals. “When I
first started doing this back in the 1960s,” recalls Noble, “I started with only
four channels.” Now he is up to thirty or forty.
In addition to pumps and channels, the membrane is home to proteins, called
exchangers, that ferry several ions across the membrane at once. And there are
still others—for hormones such as angiotensin or adrenaline—which,
when triggered, alter the cell’s internal chemistry. The model also captures the
creation and consumption of oxygen in the cell, generation of carbon dioxide,
changes of acidity and the cell’s production and consumption of ATP
molecules—the fuel that powers its pumps.
But things get more complicated yet, for the heart has many kinds of muscle
cells, each with its own unique internal chemistry. As well as ventricular
cells, there are the cells that surround the atria, the heart’s upper chambers.
The sinoatrial cells are the heart’s pacemakers and occupy a region called the sinoatrial node
(see Diagram). These need no stimulus to contract. Each
ticks on its own like a clock, and together they generate the electrical
impulses that drive the heart.
Then there are cells of the “Purkinje fibres”—the heart’s telephone
lines—which carry the regular impulses from the sinoatrial cells quickly
downwards, first through the atria and on to the ventricles. All of which means
that Noble really has had to build a variety of cellular models. “The proteins
are more or less the same in each cell,” he says, but some have more of one kind
of ion channel than another.
Unpredictable
While the behaviour of these cells is undoubtedly complex, their precise
arrangement in the real heart gives the whole organ an extra layer of complexity
that simply cannot be predicted by studying individual cells. Working together,
the cells create a wave of electrical excitation that makes the atria contract
first, filling the ventricles with blood, then the ventricles, which pump it to
the rest of the body. Modelling this sort of behaviour meant that Noble needed
to assemble his cells in just the right geometry.
To tackle this three-dimensional puzzle, Noble joined forces early this
decade with biomedical engineer Raimond Winslow of Johns Hopkins University in
Baltimore. They needed to know not only which cell type goes where, but also the
orientation of all the cells. The spindly cells contract along their long axis,
and in a real heart the orientation of those axes differs for cells in different
regions.
Fortunately for Noble and Winslow, Peter Hunter and his colleagues at the
University of Auckland in New Zealand were making the most detailed map ever of
the layout of heart cell orientations in the canine heart. Researchers in
Hunter’s group measured the orientations of cells in half-millimetre squares all
over the dog heart. Last year, they met with Winslow and Noble’s teams to put it
all together.
So far, their virtual creation has only ventricles (the atria are being added
now), but already it is giving heart researchers opportunities they’ve never had
before, such as studying the root cause of disease. During a heart attack, for
example, the interruption of blood flow deprives some of the heart’s cells of
oxygen. This can cause a heart to develop a second pacemaker (see “Intelligent
cells”). Doctors suspect, says Winslow, “that this is what triggers the
arrhythmias that can lead to sudden cardiac death”. The computer simulations
show how and also suggest ways to prevent a fatal attack.
In the computer, Noble can simulate a second pacemaker and the wave of
excitation it produces. The simulations show that this wave can run out of
control to create “re-entrant tachycardia”, in which an uncontrolled wave races
around the heart in a loop, re-exciting the same tissue over and over. At the
surface of the heart, this contraction pattern often looks like a rotating
spiral with two long “arms” rotating about the centre. In this state, a heart
beats at several hundred times a minute compared with about 70 beats per minute
for a healthy heart. But this is only the precursor to something worse.
In the final few minutes of life, the spiral breaks apart and tiny “wavelets”
of excitation zip haphazardly about the tissue. This is what doctors call
fibrillation. “If you held the fibrillating heart in your hand,” says Arun
Holden, an expert in the area from the University of Leeds, “it would feel like
a ball of writhing worms, quivering chaotically.” And the trouble is it doesn’t
pump any blood.
To save a person whose heart is fibrillating, doctors jolt the heart back to
its senses with a huge electrical shock that resets the membrane potentials of
all the cells by brute force. “This makes people jump,” says Noble, “and because
of the terrible pain, patients won’t tolerate built-in defibrillators very
well.” But using the model, several groups are looking at crafty ways to deliver
tiny shocks to the heart to drive out any spirals that form, or, if a spiral has
broken up, to force it back together, and then drive it out. “We know that the
energy required to defibrillate is very low—if only it could be applied
intelligently,” says Noble.
Virtual disease
Another opportunity open to researchers is to generate a health
problem—such as congestive heart failure, in which the heart cannot pump
enough blood around the lungs or body—and then administer different drugs
to find out which works best. In the computer, drugs act on the model by
changing the performance of one or more of the cells’ ion pumps or channels, or
some other aspect of their metabolism. Digitalis, for example, which makes the
heart beat more forcefully, inhibits a sodium-potassium exchanger.
A drug’s effect on a single living cell is not difficult to measure in the
lab and add to virtual cells in the model. Once programmed in, the model shows
how a drug’s molecular interactions affect the performance of the whole heart,
without ever putting a human at risk. And here is perhaps the model’s biggest
potential benefit—in searching for new drugs. On the one hand,
pharmaceuticals companies can screen potentially useful drugs simply by “giving”
them to the model. On the other, they can study which cellular pumps, channels
or other mechanisms are worth targeting to produce a particular physiological
response.
Unlike cystic fibrosis and other diseases that are caused by a single
defective gene, most heart disorders arise from the complex interaction of the
proteins made by many genes—and the interactions of millions of cells
built from those proteins. With the model, researchers can reach into the
heart’s cells and “tweak” one variable or another and see the result. They can
boost the action of an exchanger, perhaps, or limit the effectiveness of a pump,
and then watch as this triggers a cascade of other changes that alter the
heart’s beating pattern. By running through the possible targets one after
another, drugs companies can search for those that might be beneficial.
“This area is going to explode,” says Winslow. Banking on that belief, he,
Noble and a handful of other scientists have formed a company called Physiome
Sciences to market their model to medical researchers and drugs companies. The
model, Noble hopes, will help to prevent tragedies like the one that happened a
decade ago with the drugs encainide and flecainide.
These drugs were designed to inhibit an ion channel that was known to trigger
episodes of arrhythmia. Doctors expected them to be a real boon for people who
had suffered a heart attack. But, in a trial involving more than 2000 such
patients, those taking encainide and flecainide turned out to be at a higher
risk of fatal heart attacks than people taking a placebo. It was a disaster, and
the panel in charge of the trial brought it to an abrupt halt.
The harsh lesson is that what works for a single heart cell may not work for
the whole organ, and that traditional strategies for drug design don’t
adequately respect this fact. “An ion channel or exchanger may have many roles
in the cell’s economy,” says Noble, “so inhibiting its activity can have
unpredictable effects”. As he sees it, biology has reached a stage where further
understanding has to take this complexity into account. And that means taking a
computational approach to biology.
He sees as prophetic the words used in May 1997 by geneticist Sidney Brenner
of the Molecular Sciences Research Institute in La Jolla, California. “Genes can
only specify the properties of the proteins they code for,” Brenner said, “and
any integrative properties of the system must be `computed’ by their
interactions”. The word “computed” may seem odd in a biological context, but
Brenner’s point was that computation is precisely the effect of the myriad
chemical reactions taking place in our bodies. Genes do not determine how the
heart beats—only proteins and their interactions do that.
And that has implications. “Brenner meant not only that biological systems
themselves `compute’ these interactions,” says Noble, “but also that in order to
understand them, we need to compute them too.” Which is precisely what he and
his colleagues are doing.
Of course, the job is only part done. The model still needs atria, and if it
is really to simulate how a blocked artery triggers a heart attack, the
interface between its blood supply and muscle cells must be improved. Within
three years, Noble expects to have these problems ironed out and a virtual
heart, lungs and circulation working together. What about other organs? The
brain is just too complex to contemplate, he says. Modelling the remainder of
the body is conceivable, but it would take a major international effort. The
saving grace, Noble says, is that once you’ve modelled one organ, others become
that much easier to do.
IT DOESN’T take a whole heart to learn real lessons from virtual cells. Last
year, Denis Noble of Oxford University and his student Frederick Ch’en used one
of their mathematical cells to take a closer look at the chemistry behind heart
attacks.
Most heart attacks happen when one of the arteries feeding the heart with
blood becomes blocked. The muscle cells fed by the artery can be starved of
oxygen and other essentials. This turns them acidic, and if the trauma doesn’t
kill them, they begin beating furiously on their own. “That produces an
additional `pacemaker’,” says Noble. The pulses from this extra transmitter can
disrupt the orderly progress of the wave of contraction round the heart.
For years. researchers have known that the main trigger for this unruly
beating is an ion pump called the sodium-calcium exchanger. After calcium rushes
into a cell and triggers a contraction, this pump drives calcium ions back out
of the cell in exchange for sodium ions. For every calcium ion (Ca2+) going
out, it sends three sodium ions (Na+) in, bringing a positive charge into the
cell. But under acidic conditions, this pump works too well. It brings in too
much charge and “depolarises” the cell again, initiating another beat cycle
before the rest of the heart is ready.
How can that be prevented? Pharmaceuticals companies have already developed
drugs that inhibit the sodium-calcium exchanger. But in simulations, Noble and
Ch’en discovered that inhibiting the sodium-calcium exchanger actually makes the
runaway beating worse. “Paradoxically,” says Noble, “to suppress that beating we
need to increase the activity of the exchanger.” He and Ch’en are now working on
their model to find out why.
For Noble, this is a golden example of how even the single-cell model is
already more “intelligent” than any researcher. The interactions in a cell are
so complex that no one can foresee the consequences of making a change. “Beyond
a certain level of complexity,” says Noble, “armchair theorising becomes
inadequate to the task.”
Intelligent cells
-
Further reading:
The Limits of Reductionism in Biology
Proceedings of the Novartis Foundation Symposium 213,
edited by Gregory Bock and Jamie Goode (John Wiley and Sons, 1998) -
Computational Biology of the Heart
edited by A. V. Paniflov and A. V. Holden (John Wiley and Sons, 1997) -
For links to researchers and their work see
www.physiome.com/psinc_text/science.html
