In the introduction to his book,The Order of Things, the Frenchstructuralist philosopher MichelFoucault posed the question: "Whenwe establish a considered classification,when we say that a cat and adog resemble each other less thantwo greyhounds do, on what groundsare we able to establish the validityof this classification with completecertainty?" He went on: "On whattable, according to what grid of identities,similarities, analogies do wesort out so many different and similarthings."
In the introduction to his book,The Order of Things, the Frenchstructuralist philosopher MichelFoucault posed the question: "Whenwe establish a considered classification,when we say that a cat and adog resemble each other less thantwo greyhounds do, on what groundsare we able to establish the validityof this classification with completecertainty?" He went on: "On whattable, according to what grid of identities,similarities, analogies do wesort out so many different and similarthings."Individualized Diagnosis
As Rosenwald summarizes in hisarticle, the grid that has been recentlyused with success in studying lymphomatypes and subtypes and holdspromise for future clinical applicationis the DNA microarray. The reasonthat DNA microarrays-by virtue ofanalyzing the expression levels ofthousands of genes-are such powerfultools is simply the following basicnotion (as expressed by the 18th centuryFrench naturalist Buffon): "Themore we increase the number of divisionsin the production of nature, thecloser we shall approach to the truth,since nothing really exists in natureexcept the individual." In his article,Rosenwald illustrates how lymphomadiagnosis can be "individualized" andwhat the clinical potential of doing socould be. But in the words of Foucault,"we have to critically evaluate thegrounds on which we are able to establishthe validity of a more individualizedlymphoma classification."Prior to the 1970s, the diagnosisand classification of lymphoma wasbased purely on histologic examination. Later, with the advent ofmonoclonal antibodies specific for amultitude of lymphoid antigens andthe application of these antibodies inimmunophenotyping, lymphoma diagnosiscould be more refined. Duringthe early days of molecular diagnosticsof lymphoid malignancies, theability to establish a diagnosis of neoplasticdisease by demonstrating theclonal nature of the lymphoid populationrepresented another quantumleap. In subsequent years, a vastamount of information has accumulatedon the molecular pathology ofvarious lymphoid malignancies.These developments ultimately led tothe recognition of biological entities,reflected in the Revised European-American Classification of LymphoidNeoplasms and updated in the recentWorld Health Organization lymphomaclassification.Further biologic variables of theseestablished lymphoma entities havealready been recognized, but as discussedby Rosenwald, microarrayanalysis can help to identify thesevariables more exhaustively, objectively,and reproducibly. This bringsus one step closer to an "individualized"diagnosis. In addition, andimportantly, aberrant molecularpathways can be revealed in lymphomas,opening the door to morespecific treatment. Indeed, establishinga direct link between precise diagnosisand treatment should be theultimate goal.Gene Expression Profiling
DNA microarray technology ispossible largely due to the structuralgenomic foundation established byseveral large-scale human genomeprojects. It allows the study of genome-wide gene expression profiles inphysiologic and disease processes. Thetechnology has been employed tostudy a wide variety of human tumorswith the expectation that it will engendera better understanding of the molecularmechanisms underlying thebehavior of a tumor. Various microarrayplatforms have been used forstudying gene expression profiles. Thestudies conducted by Alizadeh,Rosenwald, and coworkers employeda cDNA microarray, the Lymphochip.Major concerns about the use of thistechnology include considerations ofhow readily we can compare studiesacross different platforms and how suchcomparisons can be facilitated. In thisregard, it is important that sufficient informationon experimental design anddata processing be available so that theexperimental data can be independentlyanalyzed by other investigators. A crossplatformcomparison of data is a challengingendeavor, and one example isthe recently published comparison oftwo sets of microarray data on diffuselarge B-cell lymphoma (DLBCL)-onebased on the Affymetrix platform andthe other, on the Lymphochip platform. It is interesting that the germinalcenter B-cell and activatedB-cell subsets of DLBCL could be definedbased on Affymetrix data, andfurthermore, that the subsets had differentoverall survival rates.Molecular Prognosticators
While the molecular prognosticatorsdefined by the studies of Rosenwaldet al can be used in addition to or alternativelyto the International PrognosticIndex and have been validatedin the studies by these investigators,several unresolved issues remain. Thenumber of array elements on theLymphochip, although substantial,represents only a fraction of thetranscriptome. Will the predictorschange substantially if the study isrepeated using an array with a morecomplete representation of thetranscriptome?As mentioned above, a molecularprognosticator will be of value onlyas long as it reflects the variables ofthe biologic response to specific treatments.The molecular prognosticatorfor DLBCL established by Rosenwaldet al was based on a series of patientstreated with anthracycline-basedmultiagent chemotherapy. Currently,patients with DLBCL are oftentreated with regimens containingrituximab (Rituxan), which may significantlyalter their response and survival.Will the molecular prognosticatorbased on samples from patientstreated with anthracycline-based chemotherapybe valid for patients treatedwith current modalities?Although gene expression profilingis a powerful technique, additional informationis useful in supervising thediscovery process. Clinical informationwas used to supervise the discoveryof prognosticators, and immunoglobulinvariable region gene mutationinformation was essential to findingdifferentially expressed genes in mutatedand nonmutated chronic lymphocyticleukemia cases.[6,7] It is likelythat certain unique genetic abnormalitieswithin a tumor subset may also beused to guide similar discovery processes.The integration of global geneticand gene expression data willcertainly be highly synergistic in futureinvestigations.Conclusions
Nonetheless, Rosenwald's conclusionis well founded. Gene expressionprofiling will have a dramatic impacton the diagnosis of lymphoma, addingmany variables to the grid bywhich diagnoses are made. In the foreseeablefuture, the initial characterizationof a lymphoma case may beperformed on a single platform, whichcould be either a diagnostic DNAminichip,an antibody panel, or a realtimereverse transciptase-polymerasechain reaction plate, and, no doubt,these instruments will be refined further.The ultimate goal is to treat patientswith tailored therapy, based ontargeting aberrant biologic pathwaysdetected in the individual tumor.
The author(s) have no significant financial interest or other relationship with the manufacturers of any products or providers of any service mentioned in this article.
Rappaport H: Tumors of the hematopoieticsystem, fascicle 8, first series. Atlas of TumorPathology. Washington, DC, Armed ForcesInstitute of Pathology, 1966.
Harris NL, Jaffe ES, Stein H, et al: A revisedEuropean-American classification oflymphoid neoplasms: A proposal from the InternationalLymphoma Study Group. Blood84:1361-1392, 1994.
Jaffe ES, Harris NL, Stein H, et al (eds):WHO Classification of Tumours; Pathology andGenetics of Tumours of Hematopoietic and LymphoidTissues. Lyon, France, IARC Press, 2001.
Wright G, Tan B, Rosenwald A, et al: Agene expression-based method to diagnoseclinically distinct subgroups of diffuse large Bcell lymphoma. Proc Natl Acad Sci U S A100:9991-9996, 2003.
Rosenwald A, Wright G, Chan WC, et al:The use of molecular profiling to predict survivalafter chemotherapy for diffuse large-B-celllymphoma. N Engl J Med 346:1937-1947, 2002.
Rosenwald A, Alizadeh AA, Widhopf G,et al: Relation of gene expression phenotypeto immunoglobulin mutation genotype in B cellchronic lymphocytic leukemia. J Exp Med194:1639-1647, 2001.
Klein U, Tu Y, Stolovitzky GA, et al: Geneexpression profiling of B cell chronic lymphocyticleukemia reveals a homogeneous phenotyperelated to memory B cells. J Exp Med194:1625-1638, 2001.