Mass-terpiece in the works for imaging protein networks?
Imaging mass spectrometry
By all appearances, Richard M. Caprioli, PhD, is a medical
His titles at Vanderbilt University bear that out: professor of biochemistry
and director of the Mass Spectrometry and Proteomics Center. He counts
pathologists among his closest colleagues.
With a little imagination, however, it’s easy to see Dr. Caprioli
as an artist. Like all artists, he’s trying to help us see the world
a little differently—in his case, to see proteins from a new
Dr. Caprioli and his colleagues are using MALDI (matrix-assisted
laser desorption/ionization) mass spectrometry as an imaging device.
The Vanderbilt team does this by placing tissue samples directly
into the instrument. By comparing the spectra from targeted cell
populations, they hope to single out diagnostic and therapeutic
"It’s one of the more interesting technologies being developed
in proteomics today," says Emanuel Petricoin, PhD, co-director of
the FDA/ NCI Clinical Proteomics Program and senior investigator
at FDA’s Center for Biologics Evaluation and Research.
Of course, proteomics, like any semifuturistic field, has more
than its share of rosy technologies. Followup studies, clinical
applications, cost, and the like have sent many such methods packing.
But more than one company has begun to produce mass spectrometers
capable of using Dr. Caprioli’s method. Dr. Caprioli is running
courses at his laboratory to teach others how to do this work. And,
he says, "The people who are most excited about this work, and whom
I’m working most actively with, are pathologists."
To fully appreciate the potential of this approach,
it helps to remember a few protein basics. Common wisdom
once held that proteins act alone. "That’s probably still a mindset
in many who aren’t involved in this field directly," says Dr. Caprioli.
However, the vast majority of proteins act within complex networks—for
example, in the signaling cascades. "We don’t think that’s just
an oddball system. We think that almost all proteins are arranged
in complex networks," he says.
These networks are the ideal target for MALDI MS. By using it
as an imaging device, Dr. Caprioli and his colleagues are able to
extract the spacial orientation of a protein.
Immunohistochemistry does this as well, he acknowledges, as do
several other methods. But they have their limitations—they’re
less successful at handling the complexity of cross-reacting proteins
that make up a network, for instance. "You have to know what protein
you’re interested in," Dr. Caprioli says. "It’s not that these aren’t
wonderful techniques. But they’re not very good for this purpose."
MALDI MS takes another tack. With Dr. Caprioli’s method, users
raster a piece of tissue with a laser beam to produce a large number
of pixels. Then they travel from one pixel to another, producing
a protein pattern for each spot. "You keep doing this, so that in
a high-resolution mode, you might have several tens of thousands
of pixels, just as you would in a digital image on a computer screen,"
Dr. Caprioli says. "But now, you have in that one pixel hundreds
of signals. So you can go back and reproduce the image in any one
of those signals. You then get a molecular-weight-specific photograph
of the piece of tissue."
This is no academic exercise. Clinical applications are well underway
at Vanderbilt. In one endeavor, the researchers are conducting comparative
proteomics studies to obtain molecular markers for the staging of
glioblastomas and lung tumors. By looking only at protein expression,
they can distinguish between a squamous cell carcinoma, an adenocarcinoma,
or a large cell carcinoma. While it’s true that many of the proteins
are the same, Dr. Caprioli reports they are finding suites of proteins
that are unique to each type of tumor.
In another study, the researchers are rastering breast biopsies.
"We’ll take the needle biopsy taken with maybe a 22-gauge needle,
so we’ve got a very small piece of tissue, about two centimeters
long. And our laser beam is very small—about 50 microns. So
we’re looking at a specimen that is 2,000 by 500 microns—we
can get lots of pixels across this tissue," Dr. Caprioli says.
In scanning this tissue, the researchers are discerning differences
between invasive tumor cells, growing tumor cells, and normal cells.
"Pathologists don’t need me to do this," Dr. Caprioli says. "They
can raster the piece of tissue on their own and determine the proportion
of invading cells to normal cells and determine the patient’s risk."
"We have a long way to go with this, though we’ve done the proof
of principle. What we’re really excited about is bringing together
drug therapy and proteomics," Dr. Caprioli says.
"If a drug arrives in a tumor in a certain spot, or near a tumor,
what is the change in the local protein concentration?" he asks.
If a drug shrinks a tumor, this should be accompanied by the disappearance
of tumor-specific proteins; likewise, the proteins that disappeared
from normal tissue when the tumor was growing should start to reappear.
"And we’re beginning to see that," Dr. Caprioli says.
He envisions predictive therapy, in which clinicians and pathologists
use imaging mass spectrometry to identify which patient is likely
to respond to therapy—and, just as important, who won’t. "We
haven’t proven that we can predict this through the proteome," he
cautions. "I want to be clear on that. But we have one or two experiments
that we’ve done that show for a few proteins it does work. So my
belief is that it will work.
"It’s just common sense," he continues. "One of the biggest changes
we see in tumor genesis is the disappearance of the differentiated
proteins in the tissue. So if the tumor begins to shrink, we should
be able to see these differentiated proteins begin to come back
as the tissue goes back toward normal. And so far—and again,
it’s very early—but so far that’s exactly what we’re seeing."
Also falling into the a-little-too-early-but-don’t-discount-it
realm is the suggestion from one of Dr. Caprioli’s clinical colleagues,
a neurosurgeon who would like to use the technology to look at his
surgical margins. "He says, ’I don’t have the luxury of taking out
another centimeter of tissue—every little bit I take out compromises
the patient. So I want to know, if I cut here, how many migrating
or tumor cells are left?"
"He wants us to run before we can walk," says Dr. Caprioli. "But
from a research point of view, it’s certainly doable.
"I’m not suggesting that my laboratory will be doing all these
things," he adds. "But it opens the door to molecular profiling
of disease and therapy for pathologists."
That would include Roy Jensen, MD, associate professor of
pathology, cancer biology, and cell and developmental biology at Vanderbilt.
Dr. Jensen and his pathologist colleagues are working closely
with Dr. Caprioli and his team, and Drs. Jensen and Caprioli are
authors of a recently published paper on the technique (Xu BJ, et
al. Direct analysis of laser capture microdissected cells by MALDI
mass spectrometry. J Am Soc Mass Spectrom. 2002;11:1292-1297).
In particular, Dr. Jensen is interested in looking at preneoplastic
breast disease as well as the transitions from in situ carcinoma
to invasive carcinoma, in both epithelial and stromal components.
Dr. Jensen enjoins a certain amount of caution when discussing
imaging mass spectrometry. "This technology generates a huge amount
of data. I wouldn’t call it ’information’ just yet," he says.
Imagine looking at Georges Seurat’s pointillist masterpiece, "A
Sunday on La Grande Jatte-1884." Early viewers may have wondered,
"What’s with all the dots, Georges?" But those who step back from
the thousands upon thousands of paint dots, placed on the canvas
with scientific precision, see one of painting’s optical marvels.
"It’s sort of like we’re trying to figure out the legend to the
map that Richard’s creating," Dr. Jensen says. "We can scan an entire
slide with his instrument and get mass spectra from literally thousands
of different points. We can visualize specific peaks and the location
of protein expressions across the whole slide. But we don’t know
what these specific peaks correspond to as far as specific proteins.
So we want to identify peaks that are associated with, say, ductal
carcinoma in situ. And then what we can do, we hope, is image that
particular peak over the whole slide."
To take an example, once it’s clear that a particular peak corresponds
to, say, collagen type IV, says Dr. Jensen, "we can look for the
expression of type IV collagen across the entire slide." Again,
this may hardly seem remarkable in and of itself, he says, since
immunohistochemistry can do this too. "The big deal is we can pick
any peak in that spectra and show where the distribution of that
type of protein is across the entire slide. So it is really like
doing a thousand immunohistochemistries at once, by a different
Dr. Jensen suggests roughly a five-year time frame for the trek
toward clinical applications.
"If we identify a select number of proteins whose expression patterns
turn out to be crucial...if we can develop standard antibodies against
them, then we can apply the technology more quickly to the clinic,"
he says. "Ideally, what would happen later on is that the technology
Richard is developing would be sufficiently advanced so that we
can do this imaging on more of a real-time basis." Now, scanning
an entire slide is an overnight enterprise, which will hardly make
it the workhorse of frozen section analysis.
In five to 10 years, "I can see a situation where you cut a frozen
section, you analyze it in this scanning mass spec device, and you
could literally generate a tremendous amount of information before
the patient’s even off the table," Dr. Jensen says. "Diagnosis and
prognosis and perhaps even ways to select chemotherapy."
The technology is alluring in its relative simplicity, he says.
"Essentially, all you really need to do is cut a good-quality frozen
section, spray on a matrix material, and mount it in the machine.
Basically, everything else is just pushing a button, and it generates
Certainly the technology will need to advance. "But it’s not like
there’s a huge technical hurdle," says Dr. Jensen "It will just
be incremental advances in computing power and optics, those sorts
It can certainly be done, says Marvin Vestal, PhD, scientific fellow
at Applied Biosystems, one of the companies that is collaborating with Dr. Caprioli.
"Some aspects of the mass spectrometer will undoubtedly have to
be modified and refined, to make it into something people can use
more practically," he says. "We haven’t accomplished that, but we’re
seeing what would need to be done if it turns out that this really
is an application a lot of people would want to use."
The most obvious challenge is to obtain data at a higher rate.
Says Dr. Vestal: "If it takes a year to map a few centimeters of
tissue, that’s not acceptable. If it takes a few seconds, then it’s
very acceptable. The truth is probably somewhere in between." The
most advanced MALDI mass spectrometers run the laser at 200 Hz,
which translates into 200 pulses per second. "We can get a good
spectrum every second or every few seconds," says Dr. Vestal. "But
you may want to do 10,000 or even a million spectra to get a complete
image, so these rates are clearly not satisfactory for everyday
As the technology evolves, its appeal will grow. "Mass spectrometry
is a totally foreign and alien technique as far as most people are
concerned," he continues. "But if all you have to do is prepare
a sample on a slide and insert it into the instrument—which
is what Dr. Caprioli is doing—and then let the instrument
do its work, without having to intervene, then it will be quite
Indeed, mass spectrometry is heading in that direction anyhow,
he says. "So if the incentive is there to use it the way Richard
is using it, then we will certainly find a way to overcome whatever
hurdles we need to."
Though Dr. Vestal is careful to note that his company has not
decided flat-out to commercialize this method, no one is backing
away from it, either. "It could have a great future."
Karen Titus is CAP TODAY contributing editor and co-managing