The use of gene expression profiling in tumour classification and management

Authors:

Details:

Murdoch Children’s Research Institute
Women’s and Children’s Health
University of Melbourne
Melbourne, VIC

Murdoch Children’s Research Institute
Melbourne, VIC

Women’s and Children’s Health
University of Melbourne
Melbourne, VIC


Introduction

Microarray-based transcript profiling is a technique used to measure the steady-state levels of gene expression on a genome-wide scale in a cell or tissue. The technology essentially involves making a collection of labelled probes from all the genes expressed in a cell, and then defining the gene and its level of expression – by hybridising them to a target of sequences representing thousands of genes, which have been previously immobilised on a solid substrate. This snapshot of the gene expression reflects the degree to which genes are switched on or off in the cell, a process that is tightly controlled for each individual cell and tissue type. Given that the total cellular composition of the mRNA species encoded by the active genes largely will determine which proteins are expressed in that cell, and thus much of its biology, the ability to profile gene expression at the genomic level can provide major insights into the molecular mechanisms driving the function of normal tissues and of cancers. The uniqueness of a gene expression profile seen in individual tissues can also extend to different subgroups of cancer.

This technology is therefore useful in the field of cancer management in the context of improving our ability for accurate diagnosis, prognosis, choice of therapy and monitoring of therapeutic response. Further, global gene expression profiling can give us valuable insights into the molecular pathogenesis of individual cancer types, which can in turn, provide a basis for more rational management.

This review will highlight several advances in the application of microarray profiling to the clinical management of various cancer types. In addition, it will touch upon potential novel applications that could result from ongoing research.

Haematolymphoid malignancies

It is now apparent that microarray expression analysis will substantially augment the established diagnostic protocols of morphology, immunophenotype and cytogenetic analysis in the diagnosis of leukaemia and lymphoma. The dysregulated genes involved in the translocation breakpoints defined by cytogenetic analysis appear to be integrally involved in the genesis of these tumours. However, the diversity of clinical response observed within the translocation-defined categories of leukaemia and lymphoma indicates that other, accompanying molecular events can be equally important in determining the outcome of these malignancies. Global gene expression signatures have been identified that correlate extremely strongly with several of these chromosomal translocations, and may thus in future replace the need to perform cytogenetic and immunophenotypic analysis. Moreover, gene expression profiling has now begun to identify many of the additional key-molecular abnormalities related to outcome, and these insights into molecular pathogenesis have begun to facilitate novel therapeutic options for the management of these tumour types.

Leukaemias

Paediatric acute myeloid leukaemias (AML) are a group of malignancies displaying heterogeneity at the clinical and molecular pathogenic levels. It is difficult to accurately predict the likely therapeutic outcome, and thus to select the optimal, risk-adjusted treatment protocols for AML using the current morphological, cytogenetic and immunophenotypic diagnostic approaches. The implementation of risk-adjusted therapies for AML is based partly on the likelihood of failure to induce remission, or on the chances of relapse. In the case of paediatric AML, several genes associated with prognosis have now been identified by gene expression profiling, thus facilitating an improved classification based on the likely risk of relapse, which is not dependent on cytogenetics or morphology. In addition these microrarray-based analyses have identified molecular pathways which are not associated with conventional classification systems, and which may constitute novel targets for therapy, eg dysregulation of the NF-kappa B signalling pathway1.

There is a similar need to assign a patient to an individual risk of recurrence group for tailoring therapy in the paediatric acute lymphoblastic leukaemias (ALL). Gene expression profiling has now been used in several studies to classify the known prognostic subtypes of ALL (such as T-ALL, E2A-PBX1, TEL-AML1, MLL rearrangements, BCR-ABL and hyperdiploid >50 chromosomes). In one study, a small number of genes differentially expressed between these subtypes were able to accurately distinguish between these different classes of ALL with an overall 97% success rate. This achievement makes the evaluation of a diagnostic array containing relatively few genes feasible2.

A new insight into T-cell ALL leukaemogenesis resulted from the demonstration of activation of five T-cell oncogenes encoding the developmentally-important transcription factors HOX11, TAL1, LYL1, LMO1 and LMO2, in the absence of chromosomal translocations3. Over-expression of LYL1, HOX11 and TAL1 were each in turn associated with the abnormal expression of other groups of genes, which together indicated developmental arrest of the leukaemic cells at the pro-T (LYL1 signature), early cortical thymocyte (HOX11 signature) and late cortical thymocyte (TAL1 signature) stages. This study was therefore able to classify the leukaemias according to these various newly-identified oncogenic pathways, and was also able to highlight cases with a more favourable prognosis (eg HOX11 signature) versus those with a worse prognosis (eg TAL1 and LYL1 signatures).

Chronic lymphocytic leukaemia (CLL) is the most common leukaemia encountered in humans. Microarray-based profiling has supported the view that this disease has a distinct gene expression profile, when compared with similar analyses of other lymphoid malignancies. It now appears that the distinct subgroups of CLL, namely the slowly-progressive form in which the cells have somatic mutations of the immunoglobulin genes, and the more aggressive form in which the cells lack immunoglobulin gene mutations (and for which patients require immediate therapy), each have a distinct pattern of altered gene expression. Interestingly, one gene (ZAP-70) was found to be strongly differentially expressed in the aggressive form of the disease, and this gene alone was able to identify this aggressive form of CLL with over 90% accuracy. If this finding is confirmed in further studies, it would be an example of microarray analysis advancing knowledge of the molecular basis of a disease at a genome level, followed by the identification of a single gene as a discriminator between disease subtypes in the clinical setting4.

Non-Hodgkin’s lymphomas

The molecular alterations driving lymphoma tumourigenesis are frequently linked to fundamental events occurring during the ontogeny of lymphoid cells. It is now accepted that the cells giving rise to the majority of non-Hodgkin’s lymphomas (NHL) are derived from germinal centre cell (GCC). The propensity for these GCCs to develop malignancy is presumed to result partly from the genetic instability, which exists during the B-cell developmental phases of immunoglobulin gene recombination, somatic hypermutation and class switching. This transient genetic instability predisposes towards the development of oncogenic chromosomal translocations, resulting in dysregulation of genes controlling cellular proliferation, apoptotic and differentiation pathways.

Diffuse large B-cell lymphoma (DLBCL) is a common adult lymphoma showing a variable response to conventional chemotherapy, and thus may constitute a collection of distinct disease types, all with a similar histological appearance. Hierarchical clustering of global gene expression data has supported this hypothesis, by identifying at least three distinct clinical subgroups of DLBCL. One of these, the germinal centre B-cell-like subgroup (in which all the commonly-encountered oncogenic changes of bcl-2 translocation and c-rel amplification were encountered) had a far greater likelihood of response to anthracycline-based chemotherapy than the other groups – an activated B-cell like group, and a group designated as DLBCL type 3. When these three tumour categories were further analysed to identify the individual genes that influenced outcome, four functional groups of genes were highlighted. The worst outcome was associated with the increased expression of genes associated with cellular proliferation, and the best outcome was seen in the tumours that had high expression of genes, which might reflect a competent immune response to the tumour cell, including MHC class II genes. The gene expression-based prognosis was superior to the currently used international prognostic index in predicting treatment response, and is likely to be of great utility in stratifying patients for future clinical trials4.

Mantle cell lymphoma (MCL) is a form of NHL which is presently incurable, and which displays a diverse clinical course, with survival ranging from one to 10 years after diagnosis. Information resulting from genome scale expression profiling has now begun to elucidate the molecular drivers of these tumours, including the determinants of the observed clinical heterogeneity. Insights into the microanatomical location of tumour formation include the normal expression level of the CC motif chemokine receptor, CCR7 (involved in homing of B-cell subsets to primary lymphoid follicles) by tumour cells, and the abnormal expression of other B-cell trafficking receptors such as CCR5 and CCR65. In a different study, two subtypes of MCL were identified, one of which carried the commonly-encountered cyclin D1 over-expression, (usually resulting from the t(11;14)), and a cyclin D1-negative cohort. Within the cyclin D1 positive cohort, a signature of over-expression of several genes associated with cellular proliferation was an extremely powerful prognostic indicator, and was able to define cohorts of tumour cases with widely variable outcomes. The importance of this prognostic signature in predicting outcome argues for stratification of future clinical trials based on this predictor, and for the development of novel therapies directed at the altered G1/S checkpoint function believed to result from abnormal cyclin D1 activity4.

Brain tumours

A number of studies have employed global gene expression analysis in an attempt to improve on the limitations of the histological classification systems as a basis for therapeutic decision making for malignant gliomas. A gene expression-based prediction model proved far superior than histological assessment in distinguishing between tumours with atypical histology which were more likely to behave as glioblastomas (and which were chemoresistant) than those more likely to behave as anaplastic oligodendrogliomas (which were chemosensitive)6. Further evidence regarding the possible utility of gene expression-based diagnostic algorithms was revealed in a study on the heterogeneous group of malignancies of the CNS known as embryonal tumours. This analysis was able to distinguish between medulloblastomas and other tumour types such as primitive neuroectodermal tumours and rhabdoid tumours, and indicated that medulloblastomas might arise due to dysregulation of the Sonic Hedgehog (SHH) pathway in cerebellar granular cells. The gene expression profile also proved much more successful at predicting clinical outcome than the widely-used morphological criteria7.

Breast carcinoma

Breast cancer patients with localised disease that are deemed by conventional clinicopathological criteria to have a high risk of distant metastases are frequently treated with hormonal or chemotherapy. However, only approximately one-third of these cases would have gone on to develop distant metastasis, and therefore the remainder of patients would have received this therapy unnecessarily. A gene expression-based signature utilising 70 genes has now been defined that acts independently of other prognostic systems to accurately predict the likelihood of distant metastasis and the prognosis of breast carcinomas. It is hoped that, if reproducible, such signatures might assist in the optimal choice of therapy for this disease. In the context of the management of individuals with a potential inherited mutation in the breast and ovarian cancer predisposition gene BRCA1, a gene expression signature has now been defined that appears to identify tumours arising on the basis of BRCA1 protein dysfunction. The ability to accurately identify BRCA1 associated tumours is useful in guiding decision relating to the diagnosis of the underlying germline mutation8.

Future outlook

The tumour types highlighted above constitute a snapshot of the possible applications of gene expression signatures to enhance our comprehension of the molecular basis of cancer, and to improve our ability to make decisions in the clinical context. Numerous additional studies have been performed, on tumours ranging from paediatric sarcomas to gastric carcinomas. Together with this revolution in our understanding of cancer pathobiology, there is a concomitant technological revolution represented by efforts to produce a ‘lab-on-a-chip’. This concept involves the coordinated endeavours of experts in fields as diverse as nanotechnology (eg narrow-bore capillaries), microfluidics and microimaging, to make the tools of gene expression profiling available to the deliverers of primary medical care in a time frame which will be useful to these practitioners.

References

1. T Yagi, A Morimoto, M Eguchi, S Hibi, et al. “Identification of a gene expression signature associated with prognosis of pediatric AML.” Blood, 8 (2003): 8.

2. ME Ross, X Zhou, G Song, SA Shurtleff, et al. “Classification of pediatric acute lymphoblastic leukemia by gene expression profiling.” Blood, 1 (2003): 1.

3. AA Ferrando, DS Neuberg, J Staunton, ML Loh, et al. “Gene expression signatures define novel oncogenic pathways in T cell acute lymphoblastic leukemia”. Cancer Cell, 1 (2002): 75-87.

4. LM Staudt. “Molecular diagnosis of the hematologic cancers.” N Engl J Med, 348 (2003): 1777-85.

5. S Ek, CM Hogerkorp, M Dictor, M Ehinger, CA Borrebaeck. “Mantle cell lymphomas express a distinct genetic signature affecting lymphocyte trafficking and growth regulation as compared with subpopulations of normal human B cells”. Cancer Res, 62 (2002): 4398-405.

6. CL Nutt, DR Mani, RA Betensky, P Tamayo, et al. “Gene expression-based classification of malignant gliomas correlates better with survival than histological classification”. Cancer Res, 63 (2003): 1602-7.

7. SL Pomeroy, P Tamayo, M Gaasenbeek, LM Sturla, et al. “Prediction of central nervous system embryonal tumour outcome based on gene expression.” Nature, 415 (2002): 436-42.

8. LJ van’t Veer, J Dai, MJ van de Vijver, YD He, et al. “Gene expression profiling predicts clinical outcome of breast cancer.” Nature, 415 (2002): 530-6.

Be the first to know when a new issue is online. Subscribe today.