William Check, PhD
To jump-start a debate on molecular classification of lymphoma at the
annual meeting of the Association for Molecular Pathology last November,
Howard Ratech, MD, who had organized the session, read the contents of
a letter which, he said, he "was able to retrieve from the trash of a
large urban medical center." The purported letter, addressed to the chairman
of pathology, began, "Please accept my resignation as director of hematopathology
in your department." It continued in part: "I base my decision on moral
grounds. I find it highly unethical to continue the practice of pathology,
recklessly using antiquated light microscopy, living under the constant
fear of misdiagnosing neoplastic lymphoid lesions, when gene chips have
made lymphoma subclassification highly accurate and reproducible."
After reading the letter, Dr. Ratech, who is professor of pathology,
director of hematopathology, and director of the molecular pathology laboratory
at Albert Einstein College of Medicine-Montefiore Medical Center, asked
the panel members: "When do you anticipate that you will write your own
letter of resignation as a classical hematopathologist?"
Clearly, Dr. Ratech intended to be provocative with this "letter," just
as he had been in formulating the session as a debate and in stating the
proposition that molecular classification should be adopted for the "primary
diagnosis" of lymphoma. He calls the letter "an educational tool to start
people talking and thinking about the issues of molecular diagnosis of
lymphomas." His goal in the session: "to raise the issue to a public debate
rather than private handwringing."
For the purpose of the debate, Dr. Ratech alphabetically assigned panelists
to pro or con. "Many of the panelists did not want to be on the pro side,"
he says. "Some had written contrary opinions in the literature." Even
though all the panelists were prominent molecular pathologists, Dr. Ratech
says, "I don’t believe any of them at this moment would leap to do lymphoma
diagnosis without recourse to the microscope or protein expression by
flow cytometry or immunohistochemistry."
Subsequent interviews with the panelists verify Dr. Ratechs assessment.
"In my career we are always going to be using morphology as our starting
point for classifying lymphomas," says Daniel A. Arber, MD, director of
clinical hematology and flow cytometry, associate director of molecular
pathology, and professor of pathology at Stanford University School of
Adam Bagg, MD, director of hematology and director of the minimal residual
disease resource laboratory in the Department of Pathology and Laboratory
Medicine at the University of Pennsylvania, acknowledges that there’s
been a lot of "hype that pathologists will become redundant and so-called
traditional technologies will fall by the wayside. But," he says, "even
those of us who practice molecular pathology on a daily basis and who
are very sensitive to the exciting data from microarrays and even apply
some of these data in our practice know that it is not going to supplant
what we now do in the workup of lymphomas." Dr. Bagg feels that all the
panelists came away in large part with a sense that gene expression analysis
has a role—"but just a role," he says. "It’s not the final answer.
Microscopy will certainly have a role in the future, maybe a lesser role
but still a central one."
Rita M. Braziel, MD, professor of pathology and director of hematopathology,
flow cytometry, and molecular pathology at Oregon Health Sciences University,
agrees. "We certainly aren’t ready to use molecular techniques alone as
a subclassification tool for lymphomas. But they are an important adjunct
to establishing an accurate diagnosis in some cases and in determining
In her opinion, microscopy will never be obsolete. "That is always the
starting point," she says. Nor does she think pathologists will become
obsolete. "Pathologists are very flexible creatures and adapt to new technologies
Microscopy is the "entry point" to diagnosis of lymphoma, in the words
of Kojo S.J. Elenitoba-Johnson, MD, associate professor of pathology,
director of molecular hematopathology and proteomics, and director of
the molecular genetic pathology program at the University of Utah. "Histology
informs the nature of subsequent tests that you order."
Dan Jones, MD, PhD, medical director of the molecular diagnostics laboratory
at the University of Texas MD Anderson Cancer Center, says, "People will
always do histology because it is cheap. And histology tells you the whole
outcome of the genetic program in the tumor. When you do morphologic evaluation,
you are looking at the end stage."
Moreover, Dr. Jones says, "At the moment there are a lot of publications
with microarray data. But in terms of day-to-day practice, we are still
in the phase where molecular techniques have not made an impact in lymphoma.
We are lagging a bit behind diagnostics in leukemias."
Daniel E. Sabath, MD, PhD, associate professor of laboratory medicine,
head of hematology, and director of the clinical hematology laboratory
at the University of Washington, says much progress is being made in characterizing
lymphomas with gene expression profiles. "But we are not yet at the point
where they can replace current diagnostic modalities—histology,
flow, and immunohistochemistry."
And Janina A. Longtine, MD, chief of molecular diagnostics in the Department
of Pathology at Brigham and Women’s Hospital and associate professor of
pathology at Harvard Medical School, adds this consideration: Therapeutic
options lag behind the ability at this point to classify lymphomas with
microarrays. "Right now we can say whether something is a good or bad
kind of chronic lymphocytic leukemia," she says, "but we lack drug options
to treat those two types."
Debating the role of molecular methods in diagnosing lymphomas is, to
some extent, moot, since molecular methods are already ensconced in the
hematopathology laboratory. "In some ways we have molecular classification
now," Dr. Arber says. "We think of some types of lymphomas in a more molecular
sense." Gene amplification methods and fluorescence in situ hybridization,
or FISH, are already used in diagnosis, such as in looking for t(14;18)
in follicular lymphoma and t(11;14) or its surrogate cyclin D1 gene expression
in mantle cell lymphoma. In Burkitt’s lymphoma, t(8;14) in the c-myc
gene is an important marker.
In addition, genetic testing for somatic hypermutation informs prognosis
in chronic lymphocytic leukemia and its cousin, small lymphocytic lymphoma.
By morphology and standard immunophenotyping CLL appears homogeneous,
but the presence or absence of heavy chain gene rearrangement divides
patients into good- and bad-prognosis categories.
"The nice thing about IgH somatic hypermutation analysis in CLL," Dr.
Jones says, "is that enough people have decided they want to do that test
that we have a marker we can evaluate across the country in large numbers
to ask how it is affecting treatment and outcome a few years down the
line." Hypermutation is a "nice story," he says, showing that a molecular
test can "make it through all the hurdles such as getting samples and
clinicians agreeing they want it." Dr. Jones is starting a parallel evaluation
of clinical hypermutation testing in small lymphocytic lymphoma.
While molecular methods such as gene amplification and FISH are already
part of the hematopathology laboratory’s toolbox, it is still controversial
whether microarrays should be added as well. Enthusiasm for doing so has
arisen largely from research by two consortia on diffuse large B-cell
lymphoma, which is curable in less than 50 percent of patients (Shipp
MA, et al. Nat
Med. 2002;8:68-74; Rosenwald A, et al. N
Engl J Med. 2002;346:1937-1947). Both groups used gene expression
profiles to identify prognostic subtypes of this malignancy. One group
segregated patients into two groups with divergent five-year survival
rates in response to standard chemotherapy 70 percent versus 12 percent.
The other group found three prognostically significant histological subtypes:
germinal center B-cell-like, activated B-cell-like, and type 3 diffuse
large B-cell lymphoma.
Whether and how to incorporate these analyses into clinical practice
remains the big question. "I am not aware of any groups routinely doing
this kind of analysis on large B-cell lymphomas," Dr. Jones says.
Dr. Braziel has worked on some studies done by one of the research consortia,
the Lymphoma/ Leukemia Molecular Profiling Project, or LLMPP. She took
part in two of the project’s studies trying to use the gene expression
data for diffuse large B-cell lymphoma to devise simpler tests to identify
the subgroups (Iqbal J, et al. Am
J Pathol. 2004;165:159-166; Hans CP, et al. Blood.
2004;103:275-282). In one study researchers set out to develop a small
chip called a tissue array that could rapidly identify lymphoma patients
with particular genetic profiles.
A standard expression microarray chip contains thousands to tens of thousands
of genes, to which cDNA or cRNA from a patient’s tissue is hybridized.
For a tissue array, on the other hand, two to three cores from each patient’s
biopsy are placed on a slide and stained with antibodies to proteins coded
by genes that discriminate the subgroups. Eighty cases can be analyzed
on a standard microscope slide. Using such a tissue array, which Dr. Braziel
calls "basically complex IHC," the group could identify the same clinically
relevant subsets as were identified on full microarrays. Some version
of this technique may well become feasible over the next five to 10 years,
in Dr. Braziel’s view. "It may actually change our algorithm somewhat,"
she says. "We wouldn’t have to test for so many specific genetic factors
like we do with targeted PCR. But we are certainly not there yet."
A more general lesson from this work, according to Dr. Braziel: "Pathologists
should not feel the least bit despondent. We can take information from
microarrays and narrow it down to manageable and clinically feasible profiles."
In mantle cell lymphoma, too, the Lymphoma/Leukemia Molecular Profiling
Project has reported that gene expression profiling may offer prognostic
information (Rosenwald A, et al. Cancer
Cell. 2003;3:185-197). In this microarray analysis, poor outcome
was predicted by factors related to proliferation, or aberrant cell cycle
regulation, such as cyclin D1, which is already tested for in classifying
MCL. However, a small number of patients did not have overexpression of
cyclin D1, but rather cyclin D2. "Those data are preliminary, so I don’t
know what to make of that yet," Dr. Braziel says. "I don’t think either
microarrays or other molecular testing has come up with any answers in
Dr. Elenitoba-Johnson says other investigators have to verify this work.
"We have one study that is making this point [about cyclin D1-negative
MCL]." Making a far-reaching conclusion based on one paper is difficult,
he notes. "We can’t change our diagnostic criteria every time someone
comes up with a new concept."
Dr. Braziel took part in an LLMPP study of follicular lymphoma (Dave
SS, et al. N
Engl J Med. 2004;351:2159-2169). "We have had no therapy that
impacted survival in follicular lymphoma," she says. In addition, prognosis
is heterogeneous: Some patients live 20 years and some die in one year,
with no apparent difference in morphological appearance. Molecular profiling,
on the other hand, revealed that profiling the expression of genes largely
related to immune response and immune status allows prediction and stratification
of follicular lymphoma patients, which Dr. Braziel calls "a noteworthy
achievement."As a member of the Southwest Oncology Group, Dr. Braziel
has seen many clinical trials of therapy for follicular lymphoma over
the years. "Nothing has had much of an impact or changed practice much,"
she says. "So making an observation that might play a role in patient
care is really exciting."
She says it hasn’t yet translated into anything that can be used as a
clinical test. "But many people are looking intensely at those data to
see if we can do that." One possible contribution to patient management
is that a poor prognostic index obtained at the time of initial diagnosis
could influence a decision to transplant earlier.
Currently, pathologists attempt to categorize follicular lymphoma as
grade 1, 2, or 3 or transformed into large cell lymphoma, Dr. Jones says.
But that is not a particularly reproducible or quantitative position.
"This is the next frontier for molecular methods in lymphoma," he says.
"Can molecular testing now step in and not only diagnose follicular lymphoma
but also show how far along the path from indolent to aggressive that
a particular tumor has progressed?"
While gene expression profiling by microarrays has brought new excitement
into the field of lymphoma diagnosis and prognosis, and provided a deeper
understanding of the genetic processes underlying development of lymphomas,
molecular pathologists are skeptical for many fundamental reasons that
gene expression microarrays in their current form will become routine
in the workup of lymphomas. Microarray technology is far more expensive
than FISH, for example. "With gene chips in particular, the investment
in equipment and personnel to bring them online is substantial," Dr. Ratech
says. He notes that this investment is disproportionate to the small number
of diseases they would be applied to—"no more than a few hundred
cases a year per institution in primary lymphoma diagnosis. Even at the
busiest medical center it would be hard to justify."
Another obstacle to clinical adoption of microarrays is that different
groups can get different results on the same disease entity, a situation
that has occurred in studies of diffuse large B-cell lymphoma. Dr. Sabath
says the main source of this discordance is that people use different
gene expression platforms. "In my laboratory we have been studying gene
expression profiles of lymphocytes with an eye toward developing diagnostic
tools ourselves. The hardest task for us has been finding a decent platform
to make a custom microarray."
Dr. Jones says nearly all are undecided now about what to do with expression
microarray classification of large B-cell lymphomas. Gene expression profiles
can segregate cases into prognostic groups, but, he says, "they are based
on platforms that generate large amounts of data and change gene targets
all the time, almost monthly." Not surprisingly, there is no consensus
yet on which markers to look at. He calls this "a huge issue."
"We are still waiting for a classification of lymphomas on a molecular
basis using one of these newer techniques," he says. Reaching consensus
will require agreement that a particular result was very important, and
it will have to be confirmed in several laboratories using different technologies.
"Expression microarrays are a bit lost in the wilderness right now," Dr.
Dr. Sabath says there are also many fundamental technical issues with
gene expression profiling—how to make RNA and label it, for instance.
"When we take specimens out of our freezers we see changes," he says.
"How do we prove RNA is adequate for microarray analysis?" He also points
out that clinical use of microarrays will require "a huge cultural change"
among pathologists, from routinely fixing and embedding tissue, which
is not optimal for microarray analysis of RNA, to rapid freezing. (Related
Dr. Jones says setting up an infrastructure to collect samples in the
right way to do microarray evaluation is also an obstacle. "Even in large
cancer studies it is difficult to get people to treat every biopsy like
it will potentially have microarray analysis." Tissue must be frozen quickly
and treated carefully to eliminate non-tumor elements, which Dr. Jones
calls "very laborious."
Dr. Arber, for one, doesn’t think microarrays are sufficiently useful
for applying to patients. "I see them as good for defining prognostic
groups. Then the key features found in the microarray would be tested
by other methods, such as PCR or an immunophenotypic method," he says,
noting that others have different opinions. "Perhaps mini-arrays [with
a few dozen or a hundred genes] will be made for specific diseases," he
allows. "But we are still some way from that."
Dr. Bagg agrees that microarrays "still stand in the experimental realm."
"I don’t know anyone applying those new findings directly to the diagnostic
scenario yet," he says. Still, he adds, "The data coming out of microarray
studies have turned dogma on its head in a wonderful way. This has been
a humbling but very exciting experience for us. We have all these rules
and guidelines, then along comes this new technology and turns things
Drs. Arber and Bagg cite the case of chronic lymphocytic leukemia to
illustrate their ideas. It was known that CLL is morphologically homogeneous
but clinically heterogeneous. Moreover, it was thought to be a malignancy
of naive B cells, lymphocytes that had not passed through germinal centers
of lymph nodes. Then molecular studies showed that in some cases of CLL
malignant cells had been through germinal centers and that they displayed
somatic hypermutation, providing what Dr. Bagg calls "one of the most
important prognostic variables." However, he says, "to do somatic hypermutation
testing on every CLL patient is not currently feasible. It is complex,
expensive, and labor-intensive."
A few years ago, when samples with somatic hypermutation were compared
with unmutated samples on gene expression microarrays, overexpression
of certain genes was observed in unmutated specimens. One gene most indicative
of this correlation was Zap-70, which Dr. Bagg considers "as good a discriminant
as somatic hypermutation." Now many laboratories are setting up Zap-70
measurement by flow cytometry or immunohistochemistry. Thus, from complicated
molecular analyses has emerged a useful assay in a more traditional form.
Microarrays are a "fantastic" discovery tool, Dr. Elenitoba-Johnson concurs.
"They have already shown us things we would not otherwise have seen. But
there might be other ways to look at the expression of discriminant or
informative genes that would probably not necessitate having to evaluate
an entire array." For the most part, he predicts that is the model that
will be pursued. He says quantitative PCR is particularly good, with a
dynamic range superior to that of microarrays. "For clinical purposes,
one might need a testing modality capable of detecting fairly small changes,"
For Dr. Longtine, envisioning when microarrays will be used in the lab
isn’t easy. She, too, thinks it more likely that subsets of markers discovered
on microarrays will be measured by traditional methods. Even the analysis
of data obtained from microarrays is more complicated than with other
methods, Dr. Longtine says, noting that bioinformatics is used to analyze
microarray data. "There is a discussion within the community of bioinformatics
people how best to analyze these data," she says. "So much data come out
of each of these microarray studies it is hard to distill which are the
Interpreting results of tens of thousands of genes requires expertise
that most pathologists don’t have, Dr. Ratech agrees. "There are software
programs for this, but if you are going to use data for diagnosis with
clinical implications, I think you would want to understand the principles
of that interpretation."
Although molecular hematopathologists are still not convinced of the
clinical value of microarrays, some are having to deal with pressure from
oncologists to start using them. Says Dr. Ratech: "Basically every time
there is a new technological advance presented at the American Society
of Hematology meeting or at ASCO, where most hematology/ oncology clinicians
go, they come back to us and say, ’Why aren’t we doing this technique?’
It can be embarrassing because a pathologist can recognize the value of
the technique, but not have the resources to evaluate it." There is also
an issue of "prudence," Dr. Ratech says, since the clinician is referring
to research papers and the technique has not been subjected to a real
trial in the clinical world to assess its reproducibility, accuracy, and
sensitivity and to determine whether it adds valuable information that
can’t be obtained any other way.
"More than 20 years ago, clinicians came to me and said, ’Why aren’t
we doing FISH to detect chromosomal translocations?’ At that time it was
a cutting-edge technique. I didn’t have—and most hematopathologists
didn’t have—training in FISH. Even classical cytogenetics was a
separate discipline." Yet today FISH is a standard part of making an accurate
diagnosis in hematopathology. On the other hand, many commercial tests
have appeared that Dr. Ratech calls "crazy" and that no pathologist would
ever endorse. For instance, at one time laboratories claimed to be able
to take a piece of fresh tumor, expose it to drugs in vitro, and calculate
the best combination drug profile. "I don’t think any of those panned
out," he says. "They were the negative expression of the desire to get
a new test on line."
While not feeling pressure to adopt microarrays for lymphoma at this
time, Dr. Longtine does acknowledge sometimes being pressured to set up
assays she considers not ready for prime time. "Handling these situations
requires close communication and discussion," she says. "Clinicians are
feeling pressure from patients to have the most current tools." For example,
clinicians are now being pressured to send breast cancer samples to for-profit
companies for microarray expression analysis. "It is unclear how that
information will be used," she says, "but patients sometimes demand these
tests." As a result, oncologists may pressure pathologists to perform
certain procedures before they are ready. "Part comes from patients and
some from oncologists who just came back from meetings and want those
tools available at their institution," Dr. Longtine says.
Dr. Arber finds it easy now to resist pressure to bring microarrays into
clinical use because "none are up to standard for clinical care. They
are still investigational," he says. "You can’t use gene arrays, in my
opinion, to guide treatment since they are not FDA approved or validated."
At the same time that they act prudently about introducing microarrays
into routine diagnosis, pathologists must become acquainted with this
technology and be prepared to work with it should the clinical demand
arise. Dr. Elenitoba-Johnson calls pathologists "the integrators of all
of these technologies."
"Historically, we acquire new tools and investigate their potential and
utility. We would be well advised to get comfortable with microarrays,"
He calls pathology itself "a game of serial acquisition of tools." To
his mind, progress in medicine occurs via a multiparameter approach. "The
more discriminating modalities you introduce, the easier it is to get
data relevant to finding distinct pathological, clinical, and molecular
entities," he says. "Now microarray-based data are being validated. That
will add to the complexity." Sometime in the next several years the clinical
utility of microarrays will be determined, with a timeline that differs
among pathological entities. For diffuse large B-cell lymphoma, for instance,
that determination may come as soon as three years.
"It would be a very radical change for pathologists to begin at the molecular
level," Dr. Arber says. "There is always the fear, too, that if you can
do that you don’t need a pathologist." That’s why he feels strongly that
pathologists should take the lead in investigating molecular methods.
If pathologists don’t get interested, other specialties will take it over.
"Then we will have a more limited role," he warns.
Dr. Bagg emphasizes the many variables associated with good performance
of microarrays and the complexity of obtaining usable clinical information
from them. "That is where pathologists come in," he says. "We are best
attuned to evaluate those parameters. We amongst all physicians are most
skilled at QC issues." Since many publications on microarrays have emanated
from nonpathology research groups, one concern has been that nonpathologists
would take over diagnosis of lymphomas. "But I don’t think they are as
skilled as pathologists at QC and standardization," Dr. Bagg says. "That
is where we as laboratorians are going to play a central role," in taking
clinically useful microarrays from the research setting to the clinical
Dr. Braziel says younger pathologists can learn about microarrays in
their training. For those already in practice, says Dr. Braziel, "CAP
puts on courses in molecular areas aimed to be practical enough to update
people." And every journal today, she notes, even the most anatomically
oriented, has papers on molecular applications.
Also, pathologists in practice can identify a person to whom they can
refer people who have questions. "That is one of the roles I serve," Dr.
Braziel says, "to be willing to talk to people—both clinicians and
patients. Some oncologists give my name to patients if they can’t explain
why the physician is ordering a particular test. I end up talking to a
fair number of patients."
One of Dr. Ratech’s mentors from 20 years ago attended the Association
for Molecular Pathology session on lymphoma. "He has nothing to do with
molecular diagnosis," Dr. Ratech says. "I asked him why he came." He replied,
"I want to see if I am out of a job, obsolete." Afterward Dr. Ratech asked
his mentor what he had decided. "He said he was still worried—but
he felt better."
William Check is a medical writer in Wilmette, Ill.