Gene Expression Assays in Cancer

The technology and expertise required for analysis of gene expression were adopted at a very early stage in its development by the field of cancer biology. This was because it was recognized that, while there were available a number of useful clinical, morphological, and molecular parameters for diagnosis and/or prognosis of human malignancies [36], there remained a substantial margin of error. It can be the case that, even with these tools, patients receiving the same diagnosis can have markedly different treatment outcomes. Cancer diagnosis is often subject to subtyping of diagnostic categories as new diagnostic tools are developed. There was a pressing need for a diagnostic tool which would complement existing histopathological evaluation and enable the profiling of cancer cells in tissues [37]. In addition to providing accurate diagnosis, there was a further need to provide biomarkers which would be useful in (1) detecting cancerous cells early in the disease, (2) providing accurate pretreatment staging, (3) determining the biological aggressiveness of disease in order to tailor appropriate therapy, (4) predicting treatment outcome, and (5) monitoring disease progression [37, 38].

4.1. Gene Expression Assays in Breast Cancer

Some early studies demonstrated the potential of the new technology [39-41] to examine the differences between normal and malignant colonic tissue [39] and identifying potential subclassifications of breast cancer [41]. An interesting aside to these studies was the different analytical methods used by the authors to analyze and interpret the data. This variability of approach remains a feature of microarray studies and can hinder the comparison of data from different studies. However, there are now processes in place which attempt to standardize at least some of the analyses [42].

It is not possible to review all applications of gene expression arrays in the field of cancer and a number of excellent reviews are available to examine different aspects of the field, for example [43-48]. However, as an introduction to how the methodology has been applied, we will discuss breast cancer as an illustrative example.

4.1.1. Early Studies in Breast Cancer

Following the early pioneering work, a series of studies have been carried out to examine breast cancer malignancies more thoroughly. As stated previously, a number of tools are available for the classification of tumors. These include histopathology, immunocytochemistry, cytogenetics, fluorescent in situ hybridization, comparative genome hybridization, and whole genome allelotyping [49]. However, these can be limited in their diagnostic capability with subclasses of disease being indistinguishable and are of little prognostic value. The study of breast cancer by gene expression arrays, taken as an exemplar, can be used to demonstrate how the field has developed both technically and analytically. It is a feature of breast cancer, unlike pathologies such as leukemia, that there is no clear cytogenetic marker associated with the disease. As will be discussed later (Section 4.1.4) it has been clearly demonstrated that some breast cancers have a familial aspect [50, 51], but early studies of gene expression in breast cancer examined nonfamilial cases. A study by Perou et al. [52], using cDNA arrays featuring 8102 human genes, profiled 42 breast tissues from both tumor (39) and normal

(3) samples. Using a hierarchical clustering algorithm, the authors were able to distinguish a number of important differences between the samples. Twenty tumor samples were examined twice, once before treatment and again after treatment with doxorubicin for an average of 16 weeks. Pre- and posttreat-ment samples from the same individual were more similar to each other than samples from different individuals. The tumor samples could be separated according to whether they were estrogen receptor positive or negative. In addition, the estrogen receptor positive group were characterized by the expression of many genes expressed by breast luminal cells and were found, with one exception, to express low levels of the Erb-B2 gene. (Erb-B2, also known as HER-2 or neu, is a transmembrane tyrosine kinase receptor whose expression status has been further shown to be correlated with poor prognosis [49] in this and other cancers). The study identified four classifications of tissue: estrogen receptor positive/luminal cell-like, basal cell-like, Erb-B2 expressing, and normal breast tissue. An important feature of the data was the finding that the gene expression profile of a primary tumor is similar to that of a metastasis from the tumor. This would suggest a clonal development of the tumors and indicate a basis for individually defined treatment.

Bertucci et al. [49] carried out a similar study and reached a broadly similar result. Also, using a hierarchical clustering approach, 34 primary breast cancer samples were compared using cDNA arrays on a nylon membrane format. An important feature of this study was that the cDNAs spotted on the arrays were a select group predetermined to be important in breast cancer. Two distinct groups of tumor were identified, characterized by the outcome after therapy. As found in the earlier study [52], expression of Erb-B2 was associated with a poor prognosis, and this is suggested to reflect the behavior of the protein as an enhancer of cellular motility and hence possibly predisposing to a greater risk of metastasis.

4.1.2. Identification of Gene Signatures

Although early studies were designed to distinguish between normal and tumor material, a later study [53] set out to identify a gene signature that was associated with a short-time interval before metastasis and hence a poor prognosis. The current predictors of metastasis, lymph node status, and the histological grade of the tumor were not adequate predictive tools and it was recognized that many patients were receiving chemotherapy and/or hormone therapy when it was not needed. The study examined tumors from young patients, under the age of 55 years, who had received surgical treatment alone. An analysis which examined clinical outcome identified a poor prognosis signature comprising 70 genes that were associated with a short interval to metastasis from lymph node negative tumors. This offered a predictive model for patients who would benefit from adjuvant therapy and also would allow identification of those patients for whom adjuvant therapy was less likely to be needed. Further studies on a larger cohort of patients have indicated that the 70 gene signature is a powerful predictor of disease outcome [54]. A feature of these studies is that no attempt was made to examine different cell types from within the tumor, which limits the specificity of the analysis. The predictive power of the gene expression assay would be furthered if the precise cells that were likely to develop into tumors could be identified and analyzed.

4.1.3. Dissecting Out Tumor Types Using Gene Expression Analysis

As has been discussed (Section 3.2), the techniques of LCM enable the researcher to separate individual cells from within a heterogeneous tissue. This should increase the specificity of the gene expression assay and increase its predictive power. Sgroi et al. [55] were the first to apply LCM to the isolation of different cell types from breast cancer tumors. They examined normal, invasive, and metastatic cell populations and described gene expression differences between each of the groups of isolated cells, demonstrating the potential use of the combined approach. Ma et al. [56] used LCM and arrays to determine whether tumorigenesis was a multistep process with a defined series of stages occurring in response to the acquisition of advantageous gene changes. If this were the case, these changes might be similar across different individuals. In breast cancer, the designated stages of disease progression are: premalignant atypical ductal hyperplasia (ADH), followed by ductal carcinoma in situ (DCIS), and finally invasive ductal carcinoma (IDC). Unfortunately, these stages are poorly defined histologically in breast cancer, with an alternative grading system being used in which the tumors are described as well, moderately or poorly differentiated and given the numerical identifier I, II, or III, respectively. A number of different parameters were examined and a few features identified. It was not possible to define stage-specific gene expression profiles by hierarchical clustering but some distinction could be made between the tumor grades. In particular, grade I was quite different from grade III and grade II appeared to be a hybrid signature of the two extremes. A subset of genes were identified that could be associated with grade III (poorly differentiated) tumors and the DCIS stage. This subgroup was more likely to be associated with an occult invasive phenotype. A particular gene, RMM2, was identified which may play a part in disease progression. RMM2 is the rate-limiting component for the conversion of ribonucleotides to deoxyribonucleotides required for DNA synthesis. It is thus a point of control for the rate of cell division. Increased RMM2 production would support rapid cell division and proliferation. As has been noted before, the profile of expression of different stages of tumor from the same individual were quite similar, again indicating that the tumors are clonally derived and do not necessarily represent changes in gene expression that are common to the disease process.

A study [57] examined the gene expression profile of 286 patients with lymph node negative breast cancer. These authors were able to describe a 76 gene signature list: 60 for estrogen receptor positive and 16 for estrogen receptor negative, which could be used to recommend systemic adjuvant chemotherapy where it would be most appropriate. Sotiriou et al. [53] examined the gene expression profile of 99 breast cancer patients, both lymph node positive and negative. As has been seen previously, the major division was between the estrogen receptor positive, luminal cell-derived tumors, and estrogen receptor negative, basal cell-derived tumors, with some smaller subgroups being identified. It was suggested that a minimum classifying set of genes might become available when further studies had been carried out. A similar study [58] also set out to identify the genes that were associated with the transition from DCIS to IDC and used LCM to isolate the cells. The study also identified a series of genes differentially regulated in IDC compared to DCIS, with more upregulated than downregulated (445:101). However, the list of genes shared only four with the Ma study [56]. This reflects the differences in approach with different array platforms being used and, to some extent, the analytical methods applied. There were similarities in the outcome of both studies. In both instances, the progression stages from the same patient are more closely clustered than the corresponding stages from different patients. Both groups also consider that the approach provides a molecular insight into the transition from noninvasive to invasive tumor and begins to characterize a set of differentially expressed genes that might underpin the process.

4.1.4. Gene Expression Profiling of Familial Breast Cancer

It appears that the relative risk of breast cancer increases with the number of female relatives one has who suffer from the disease, particularly if the age of onset in the relatives is under 40 years [50]. The two major known inherited forms of breast cancer, which are caused by mutations in the genes BRCA1 or BRCA2, account for ~15% of the excess familial risk of the disorder, which suggests there are other potential inherited causes [50]. A few genes have been implicated: TP53, PTEN, CHK2, and ATM, but the numbers of affected individuals are very small. Several large-scale studies have examined the influence of the BRCA1 and BRCA2 gene mutations on familial breast cancer [51]. The study by Malone et al. reviewed some of these and demonstrates that, while the overall numbers of sufferers can differ between different populations, the population frequency of BRCA1 mutations in breast cancer patients was about 2.4% compared to only 0.04% of controls. The figure for BRCA2 mutations was 2.3% in patients and 0.4% in controls.

These data are similar to previous studies, although the frequency of BRCA1 mutations was slightly lower and BRCA2 slightly higher [51]. However, it is apparent that the level of gene mutation leading to breast cancer is low and other mechanisms leading to disease need to be investigated.

Hedenfalk et al. [8] examined the differential expression profiles of the inherited forms of breast cancer. They examined seven each of BRCA1, BRCA2, and sporadic breast cancer samples and demonstrated that there were different profiles for the groups. BRCA1-derived tumors differed from BRCA2 in the expression of 176 genes. In addition the BRCA1 mutations were more likely to be associated with estrogen receptor negative cells while BRCA2 tumors were more likely to be estrogen receptor positive. Therefore, the study demonstrated significant differences between the germ line mutations BRCA1 and BRCA2 at the level of their global gene expression profiles. The familial forms of breast cancer were also different from the sporadic form. The study also identified a sporadic case with altered expression of BRCA1 through a promoter defect, which gave confidence in relation to the power of the study.

4.1.5. Profiling the Effects of Drug Treatments

One area of research that has progressed in the field of breast cancer analysis has been the use of gene expression arrays to examine drug effects on tumor progression in order to improve treatment strategies [59-61]. The rationale behind such studies is that, while adjuvant systemic therapy after surgery for breast cancer is widely used and represents a crucial intervention in reducing mortality, it is difficult to predict which patients are likely to benefit from particular treatment regimes. Chang et al. [59] were interested in the prediction of response to taxane drug treatment using the drug docexatol. They were able to examine the effect of neoadjuvant therapy on tumor development and by association determine a gene expression profile that lent itself to a positive outcome to drug treatment. Using the Affymetrix HgU95-Av2 GeneChip, they focused on genes which gave a consistent response would enable the identification of gene expression patterns that could be used as a predictive test for the effect of the taxane treatment. The 92 genes that were identified consisted of 14 genes overexpressed in treatment-resistant tumors and 78 genes overexpressed in treatment-sensitive tumors. The genes associated with docetaxel treatment resistance fell into categories including protein translation, cell cycle, and RNA transcription suggesting an involvement in cell growth and division, while the categories represented in the treatment groups included stress or apoptosis, cell adhesion, cytoskeleton, protein transport, transduction, and RNA splicing which is in keeping with cell maintenance or cell death roles. These results were consistent with a role for docetaxel in apoptosis induction and suggested that a profile of genes associated with potential treatment strategies had been determined. The 92 genes identified in this experiment were tested in a leave-one out cross-validation study which showed them to be 92% successful for predicting a positive response to the drug and 83% successful in predicting a negative outcome. A further test of this profile in an independent cohort of samples was successful in describing six patients with drug-sensitive tumors. This level of predictive value compares favorably with other clinically validated markers.

A study by Ayers et al. [60] set out to describe a pathological complete response (pCR) to a combined neoadjuvant chemotherapy approach consisting of paclitaxel and fluorouracil + doxorubicin + cyclophosphamide. Forty-two patients were examined by gene expression array prior to drug treatment and the response monitored. Thirteen of the patients achieved complete tumor loss and these were split into two groups: six were examined to define a gene expression profile associated with sensitivity to the drug treatment and the remainder used in a validation set. A complex analytical methodology gave rise to a 74 gene predictive group which were tested on a validation set and were found to be 78% accurate in predicting the response of a tumor to the drug treatment (three patients predicted to demonstrate pCR showed a complete tumor loss). Further predictive models using smaller numbers of genes were less efficient. While the list of genes important in the predictive process included few previously considered important to the tumorigenic process, this experimental protocol indicates the value of the methodology and may be developed into a diagnostic tool with a larger number of patients.

It is not possible in a short review to examine fully how gene expression arrays have been used to examine breast cancer. However, the above discussion illustrates the principles of the approaches used and the potential clinical value of the data emerging. Brenton et al. [62] and Sorlie et al. [63] have attempted to mine the published data in order to synthesize an overview of the research in this field. It is apparent that breast cancer is a heterogeneous disorder with marked differences between familial disease related to expression of BRCA1 or BRCA2 and the sporadic form of the disease. The location of the initial tumor in a basal or luminal cell type is also significant in terms of prognosis. The gene expression studies have demonstrated that the profile of expression can be used as an adjunct to the histopathological and histochemical approaches, and provide additional information concerning the potential effects of adjuvant therapy. Hence, different subtypes can be identified using gene expression data: basal or luminal cell derived, Erb-B2, (HER-2), positive or negative, BRCA1 or BRCA2 positive, and estrogen receptor positive or negative. There remains a lack of consensus as to the most appropriate set of genes to examine, but it is noted that basal cell-derived tumors have the worst prognosis and this is exacerbated by the presence of Erb-B2 expression and estrogen receptor status. The potential of the approach has been demonstrated and what is now needed is a large-scale study, or reanalysis of the current datasets, to enhance the information that can be used in patient diagnosis.

4.2. Gene Expression Assays in Other Cancers

Although this chapter has focused on breast cancer, it is also interesting to briefly describe how the technology has impacted on the field of hematologi-cal cancers. This has been thoroughly discussed by Dunphy [64] and demonstrates how new stratifications of disease are being identified by their gene expression signature and hence new markers being made available for more cost-effective diagnostic methods such as IHC and flow cytometry. It is clear that gene expression profiling is useful as an adjunct approach with other methodologies.

Valk et al. [9] investigated genes differentially expressed in acute myeloid leukemia (AML). Current methodologies offer only about 50% of the information, predicting patient therapy and prognosis. In this study, 285 AML patients and 8 healthy controls had their gene expression profile of blood or bone marrow determined, and 16 clusters were identified. some of these clusters were associated with known karyotypic abnormalities or particular gene mutations, but several of the clusters had a normal karyotype. There was a unique cluster which was associated with AML cases with a poor prognostic outcome. With new clusters of AML having their own characteristic gene signatures, this demonstrated how the technology could aid diagnosis and provided a step toward the identification of a set of genes that could be ordered on a diagnostic array. Another study of AML examined 116 AML cases of which 45 had normal karyotypes [65]. Gene expression signatures were demonstrated for different subtypes. Progress of the disease was followed up in an attempt to construct a gene expression-based outcome predictor. A group of 133 genes could be used to predict outcome for karyotypically abnormal cases. In patients with a normal karyotype, there were two subgroups based on their gene expression profile. These studies demonstrate that there is a role for gene expression profiling in both improving classification and predicting the prognosis of AML.

4.3. Considerations When Comparing Array Studies

It is appropriate to consider some of the general themes that have already become apparent in the short time that microarray technology has been available. Cancer biology is complicated by the mechanisms that can lead to a tumor developing and progressing. As has been discussed, tumor development can be associated with chromosomal rearrangements or gene mutations which are subject to selection pressure during cancer progression leading to a cancer cell that has a selective advantage [66]. This explains the clonal development of tumors which has been demonstrated by gene expression profiling. Previously, many studies in cancer examined single markers and correlated them with a diagnosis and clinical outcome [67]. This process is a simplification of the etiology and progression of cancer.

Many pathways interact in the development of a tumor and the advantage that gene expression profiling brings is the potential, particularly through clustering analyses, to identify groups of genes that are differentially regulated together. This grouping process can further identify regulatory processes that are a part of the functional relationships modified in the tumor [67]. There are three potential approaches to tumor analysis: whole tumor examination, tumor-derived cell lines, and microdissection profiling [67]. Each approach has its benefits and drawbacks. The whole tumor will provide a gross picture of the expression profile but the cancerous cells may be a small fraction of the tumor mass. Cultured cells provide an idea of the expression profile of the cancerous cell but in an isolated environment. Microdissected cells are able to examine the gene expression changes in what may be a small population of cells within the tumor, but it may not be easy to identify the appropriate cell and the small numbers of cells will provide a limited quantity of RNA, which means additional processing steps are required to produce the gene expression profile. However, the field has progressed rapidly and microarray technologies have been used to profile the gene expression profile of tumors, leading to the discovery of particular disease susceptibility genes, therapeutic targets, and profiles related to disease outcome and drug sensitivity and resistance [66]. What has become apparent is that from the many studies that have been carried out in a variety of tumor types, a consensus has not yet been reached which can be applied in all studies. Different groups have used different array types and different analytical tools. This can lead to apparently discrepant results. There is now acute awareness of this problem in the field of microarray research and mechanisms are in place to try and address the difficulties. The adoption of the MIAME standards [42] for publication of material has focused attention on the parameters required to validate a study. The next step is a further examination of strategies for data analysis that will strengthen the data examination and allow studies to be more easily compared and interpreted. The challenge will be to correlate these findings with high-throughput proteomics analyses, and finally to the development of novel diagnostic, predictive, and prognostic biomarkers, as well as therapeutic targets.

10 Ways To Fight Off Cancer

10 Ways To Fight Off Cancer

Learning About 10 Ways Fight Off Cancer Can Have Amazing Benefits For Your Life The Best Tips On How To Keep This Killer At Bay Discovering that you or a loved one has cancer can be utterly terrifying. All the same, once you comprehend the causes of cancer and learn how to reverse those causes, you or your loved one may have more than a fighting chance of beating out cancer.

Get My Free Ebook

Post a comment