Trastuzumab

EVOLUTION OF SURGICAL PATHOLOGY

Morphologic pathology has advanced considerably since the latter part of the 19th century when in the midst of significant initial skepticism in some quarters, surgical biopsy was introduced as a diagnostic tool in continental Europe. In the US, the formal institution of surgical pathology as a diagnostic discipline in clinical practice, can be traced to the appointment of Joseph Bloodgood as the first full-fledged surgical pathologist at Johns Hopkins Hospital in Baltimore, Maryland, by William Halsted the pioneer in breast cancer surgery.1 It is interesting, although not common knowledge, that the advent of surgical pathology was facilitated by the development of the freezing microtome in the early years and that frozen sections were an important and in some instances the sole means of obtaining tissue diagnoses. According to anecdotes, the ability of a pathologist to render a definitive intra-operative diagnosis in a matter of minutes (rather than days taken by the 'academic' pathologists) greatly enhanced the relevance and value of surgical pathology. 2 Over the last several decades advances in histological techniques have helped to establish morphologic pathology as the single most important tool in the hands of the surgical pathologist. In fact, the terms morphologic pathology and histopathology are used synonymously with surgical pathology.

The pathologic evaluation of cancer is based on assessment of a series of morphologic parameters of an individual tumor, including the tumor size, type, grade, involvement of lymphovascular channels and the status of regional lymph nodes. These assessments are made using the standard hematoxylin and eosin (H&E) stain on tissue sections. While this is the usual approach, in some instances additional studies are performed as diagnostic and prognostic parameters, including electron microscopy, in situ hybridization, polymerase chain reaction (PCR), cytogenetics and proteomic analyses. In breast cancer, for example, the expression of estrogen and progesterone receptor proteins (ER/PR) as well as that of Her2/neu gene product are tested using immunohistochemical methods. Although conventional H&E stainbased morphologic pathology is the single most important approach in surgical pathology, it has evolved and increased its strength by being able to "absorb" and incorporate numerous newer technical advances. In many ways these "advances" are imprints left by the technologies that were developing or had been developed at the time.

In the early decades of the 20th century, various histochemical tests that permitted the detection of microbial organisms or intra-cellular constituents such as mucin, glycogen or fat were added to surgical pathology. At the time, these newer tests were used with great enthusiasm, for example, in distinguishing an adenocarcinoma from non-adenocarcinomas or in discriminating Ewing's tumor from other similar appearing soft tissue/bone tumors. With advances in immunology in the second half of the last century, antigen-antibody reactions were adapted to tissue sections on the premise that the localization of a single antigen or groups of antigens would enable distinction of various cell types. This, in turn, would provide the pathologist with an objective diagnosis of the tumor type in contrast to the morphologic diagnosis which is based on subjective assessment of patterns. Similarly, when electron microscopes became available, they were quickly employed in diagnosing lesions based on their sub-cellular features. The same paradigm applies to the use of molecular techniques such as in situ hybridization and PCR.

During the evolution of surgical pathology its relevance as a diagnostic modality has been questioned on many occasions. The fact is, however, that surgical pathology has flourished over time and has either absorbed some of these techniques with the goal of increasing its own vitality or seen the demise of some of the very applications that threatened its relevance in clinical oncology. An example is the biochemical assays for ER and PR in breast cancers used in the 1960s and 1970s. These assays were performed by biochemists because morphologic predictors of ER/PR status were considered unreliable. However, with the immunohistochemical assays for these markers in the early 1980s and their rapid application to paraffin sections, the biochemical assays were rendered obsolete.

Over the years, our understanding of tumor biology has increased exponentially. We now know that various pathways involving many molecules play vital roles in the growth and development of tumors. In recent years, oncologists have targeted some of the key molecules involved in tumorigenesis in devising novel forms of therapy.4 An example in breast cancer is Her2/neu, a key tumor growth factor receptor. An anti-Her2/neu monoclonal antibody (commercially available as Trastuzumab or Herceptin) is used in the treatment of a subset of breast cancer patients with evidence of dramatic responses. Since purely morphological methods do not allow elucidation of the pathways or processes involved in tumorigenesis, these adjunctive approaches are used in tandem with conventional morphology to give us only snap-shots of molecules important in tumor growth. Following the sequencing of the human genome, the development of DNA microarrays on silicone chips have permitted gene profiling of cancers. It is likely that gene profiles will permit the identification of hundreds of genes involved in tumorigenesis and will also elucidate their interrelationships. It is likely that gene profiles will permit molecular classifications of tumors and that these approaches may supplant the current morphologic ones. Studies have also shown that gene profiling may help to identify genes that are responsible for organ specific metastatic potential of tumor cells.3 It is clear, therefore, that gene profiling will become a key approach in understanding the basic biology of tumors, in the clinical diagnosis of tumors, and in the development of novel therapeutic strategies. However, the results from c DNA microarray studies must be thoroughly validated since there is potential for the incorrect selection of candidate components within the tumor. The role of the surgical pathologist in this validation process will be crucial.

GENE EXPRESSION ANALYSIS: MORPHOLOGY VERSUS GENE (DNA) MICROARRAYS

By conventional light microscopy, the diagnosis of tumors is based on the recognition of specific patterns that are universally acknowledged as being characteristic of the particular entity. Considering the fact that a bewildering array of morphologic phenotypes are seen in tumors arising in a given organ, it is tempting to argue that the various patterns are, in fact, the ultimate expression of the underlying genotype. Figure 1 shows some of the less commonly encountered histologie types of breast carcinoma (A: Invasive lobular carcinoma, B: Invasive micropapillary carcinoma, C: Medullary carcinoma and D: Mucinous carcinoma). The recognition of these morphologic types has prognostic significance based on the retrospective analysis of thousands of such cases across numerous studies independently carried out by pathologists over a period of many years. Some of these tumors, such as the invasive micropapillary carcinoma, are known for their aggressive behavior, while others such as mucinous carcinoma and medullary carcinoma are associated with a favorable prognosis. Invasive lobular carcinoma, although not prognostically distinct, appears to be biologically distinct in that these tumors have a predilection to metastasize to unusual sites such as the stomach, endometrium or cervix. It would be interesting to determine if the gene profiles for each of the morphologic types is different and, if so, what the differences are.

Morphology also reveals the topographic relationship and the nature of the tumor stroma. For example, in Figure IA the stroma is fibrous and is significantly greater than that in Figure IB. On the other hand the stroma in figure 1C is composed of lymphocytic infiltrates whereas that in Figure 1D is composed of acellular mucin. The recognition and in depth knowledge of these morphologic attributes of tumor will be of great importance in analyzing the data that gene profiling studies deliver.

GENE MICROARRAYS: How THEY ARE PERFORMED

In contrast to morphology, DNA microarray studies directly demonstrate the over- or under-expression of thousands of genes or gene related sequences in comparison with another tumor or the baseline for that particular organ. Gene microarrays are performed using various types of expression arrays, the most common of which is the oligonucleotide array. Briefly, fresh or frozen tissue samples enriched for a representative area of the tumor are homogenized and subjected to a standardized protocol for extraction of messenger RNA (mRNA). The mRNA in the samples is then reverse transcribed to complementary DNA (cDNA) and biotinylated. The biotinylated sample is then incubated with a silicon or nylon membrane chip typically arrayed with millions of copies of oligonucleotide probe sets. The number of probes on a given platform (silicon chip) varies but usually several thousand probe sequences are arrayed on an individual platform. After the completion of incubation, a chromogenic or fluorescent reaction using conjugated streptavidin is carried out and the subsequent reaction is image analyzed. (Figure 2).

STATISTICAL ANALYSIS AND GENERATION OF DENDOGRAMS

A computer program uses algorithms to cluster differences in expression patterns between sample sets based on the level and similarities of expression patterns between samples. This approach is called the unsupervised clustering or classification. In the supervised methods of analysis, gene expression differences are studied in predetermined groups (typically based on available clinical information such as bad prognosis and good prognosis tumors). Once the clusters are determined, they are supplemented by visual display using tree-like dendograms. (Figure 3) The statistical methods are still evolving and no single statistical method is applicable to all situations. One of the major weakness of the statistical tests used in gene expression analysis is that whereas there are thousands of gene sequences, the total number of samples is relatively small.

GENE MICROARRAY: CURRENT STATUS

The number of studies on the subject of gene microarrays has increased exponentially in the last 3-4 years: over 400 studies focus on breast cancer alone. Several studies have demonstrated the considerable power of gene microarray analyses. For instance, in a study of morphologically similar appearing non Hodgkin's lymphomas of the diffuse large B cell type, two subtypes defined by differing patterns of gene expression were identified.5 One subset had gene patterns similar to those of germinal center cells (called as the germinal center diffuse large B-cell lymphoma) and the other had gene profiles similar to those seen in the peripherally activated B-cells (called the activated diffuse large B cell lymphoma). The activated cell type had a significantly worse prognosis based on the results of this study. Analyses of this type highlight the clinical significance of gene profiling studies and their superiority over conventional morphologic studies in predicting prognosis.

In another study, gene expression profiles of more than 175 cancers from 10 common sites including 12 metastatic tumors were studied.6 The authors were able to accurately identify the organ of origin in over 90% of cases, an astoundingly high accuracy rate. Using morphology and state of the art immunohistochemistry accurately predicting the site of origin of a metastatic lesion is often frustratingly difficult, especially in poorly differentiated tumors. In another study, a total of 295 early breast cancer cases (stage I/II) were analyzed. Based on the profiles using a set of signature genes, it was possible to divide the patients into poor and good prognostic groups. Eighty-five percent of the good prognosis patients survived disease free for over 5 years while only 50% of the patients in the bad prognosis group had more than 5 year survival. A study done by Sorlie et al8 proposed a molecular classification for breast cancer dividing the cases into two broad categories one with the 'luminaT type cytokeratin expression (generally ER positive) and the other with a 'basal' type cytokeratin expression (generally ER negative). It is clear that gene expression studies represent a major advance in the study of tumors. It appears that using this approach, it will be possible to predict prognosis, to classify tumors into biologically relevant groups, and, as described in another recent study, to identify genes that are responsible for organ specific metastatic potential.

VALIDATION OF GENE MICROARRAY DATA

It is critical to validate data on gene expression. A reference sample must always be assayed with the test sample as an essential quality assurance control. Ideally, all samples should be run in duplicate on different chips and the expression patterns should be compared to ensure that there are no "between run" differences. It is important to validate gene expression by actually demonstrating a gene product in tissue sections either by using in situ hybridization for the corresponding messenger RNA or by performing immunohistochemistry for the corresponding protein. An elegant approach is to prepare multitissue blocks, also referred to as tissue microarrays (Figure 4). Briefly 3mm cores of comparable areas of tumors that were subjected to gene expression analysis are obtained from paraffin embedded blocks and are re-embedded in a separate block. Potentially, several hundred tumor samples can be embedded in one block, and sections obtained from such blocks can be examined immunohistochemically or by in situ hybridization to verify differences in gene expression patterns (Figure 5).

PROBLEMS AND PITFALLS IN GENE MICROARRAY STUDIES

The published studies on gene micro arrays indicate that our understanding of tumor biology will be enhanced greatly using this technology. Morphologic methods may be supplemented or even supplanted by gene profiling studies; however, there is a need to understand several issues. First, only a few hundred cases have been examined for any given type of cancer and it would be incorrect to make generalizations based on these relatively few examples. Under the present circumstances it will probably not be feasible to study a large number of cases because of the need for fresh or frozen tissue samples. Tissue repositories with large numbers of tumors are relatively few; moreover, the study of banked specimens may not allow sufficiently long term follow-up studies.

Most tumors are morphologically heterogenous. Unless the sampling is performed by a trained surgical pathologist the results may be misleading or spurious. Even with good sampling, tissue contamination by stroma or other elements may be unavoidable. In such instances, it may be necessary to use laser capture microdissection to isolate pure tumor populations. Laser capture equipment is expensive and not available in most pathology laboratories. Finally, today's statistical methods are not uniform, and there is ongoing discussion on what may be the best approach to analyze the data obtained from these studies.

In conclusion, the routine use of gene profiling as a clinical tool may not quite be around the corner. Meticulous planning and well-supervised execution of these studies with a major input from a surgical pathologist is essential for the new technology to render the goods.

REFERENCES

1. Rosen G. Beginnings of surgical pathology. Am J Surg Pathol 1977;1361-4.

2. Wright JR. The development of frozen section techniques, the evolution of surgical biopsy. Bull Hist Med 1985;;9:295-326.

3. Minn AJ, Kang Y, et al. Distinct organ specific metastasis potential of individual breast cancer cells and primary tumors. J Clin Inv 2005;115:44-55.

4. Olson JA. Application of microarray profiling to clinical trials in cancer. Surgery 2004; 136:519-23.

5. Alizadeh AA. Distinct sub types of diffuse large B-cell lymphoma identified by gene expression profiling. Nature 2000;403:503-11.

6. Su AI, Welsh JB, etal. Molecular classification of human carcinomas by use of gene expression signatures. Cancer Res 2001; 61:7388-93.

7. Van de Vijer MJ, et al. A gene expression's signature as a predictor of survival in breast cancer. NEJM 2002;347:1999-2009.

8. Sorlie T, et al. Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications. Proceeding Nat Acad Science 2001;98:10869-74.

DILIP GIRI, MD

Dilip Giri, MD, is a Surgical Pathologist/Cytopathologist at Rhode Island Hospital and The Miriam Hospital, and Assisant Professor of Pathology and Laboratory Medicine, Brown Medical School.

CORRESPONDENCE:

Dilip Giri, MD

Rhode Island Hospital

593 Eddy Street

Providence, Rhode Island, 02903.

Phone: (401) 444-3122.

Fax: (401) 444-8514.

e-mail: dgiri@lifespan.org

Copyright Rhode Island Medical Society Jul 2005
Provided by ProQuest Information and Learning Company. All rights Reserved