January 2010
Feature Story

William Check, PhD

In its first two editions, the textbook of which David J. Dabbs, MD, is editor and one of the authors was titled Diagnostic Immunohistochemistry. Dr. Dabbs told CAP TODAY that when the third edition comes out, early this year, it will have a new title: Diagnostic Immunohistochemistry with Theranostic and Genomic Applications. “It is not easy to have a publisher change a title,” says Dr. Dabbs, chief and professor of pathology at Magee-Womens Hospital, University of Pittsburgh Medical Center. “But I really wanted to emphasize that immunohistochemistry [IHC] is not just useful in diagnosing carcinoma, melanoma, germ cell tumors, lymphoma, and sarcoma. There is a whole battery of new applications that is available now in solid tumors. We are using molecular morphology of IHC and tissue to set the stage for treatment and further genetic testing for which the patient may be eligible.” IHC is far more powerful than it was two or three decades ago, he notes, and adds, “We pathologists need to think about how to incorporate new markers and knowledge into our daily clinical pathology practice.”

To extend the reach of that message, Dr. Dabbs organized a symposium at the 2009 American Society for Clinical Pathology meeting embodying the same theme as the textbook. He invited five pathologists who contributed chapters to speak on their areas of expertise—breast, lung, gastrointestinal, head and neck, and genitourinary cancers. “My main reason for putting this together was to make pathologists aware of the applications that exist for diagnostic modalities related to IHC and molecular diagnostic pathology,” Dr. Dabbs says. He views the disciplines of IHC and molecular pathology like tectonic plates pressed against each other and constantly grinding. “Once in a while one slips over the other,” he says. “I wanted to make participants aware of the current state of the art of diagnostic IHC and where molecular pathology picks up. Right now the lines between them are blurred; there is considerable overlap.” At the same time, he adds, he wanted to show how IHC can be used in more than a diagnostic sense, in genomic and theranostic applications as well.

Diagnostic IHC is well established. Dr. Dabbs’ aim was to highlight the rise of “genomic IHC”—the ability of IHC to detect clinically important genomic products without necessarily having to do molecular testing. “This is an aspect [of IHC] that is not terribly well recognized,” he says. “Theranostic IHC” is a second aspect of IHC that Dr. Dabbs wanted to highlight. He notes that IHC was the method of choice for one of the first theranostic tests—estrogen and progesterone receptor detection, which got its start in the 1980s as a way to guide breast cancer chemotherapy. “Now,” he says, “there are new tests, such as HER2, EGFR, and KIT, that not only represent genomic IHC for diagnostic purposes but also represent theranostic, prognostic, and predictive markers.”

In the symposium, George J. Netto, MD, chair-elect of the solid tumor subdivision of the Association for Molecular Pathology and associate professor of pathology, oncology, and urology at Johns Hopkins Medical Institutions, discussed the promising theranostic role of novel biomarkers in genitourinary tumors. Speaking more generally, he told CAP TODAY, “The symposium addressed how molecular diagnostic assays are finally gaining utility in determining prognosis and predicting response to therapy in the solid tumor domain following their widespread use in hematologic neoplasms and genetic disorders.” The strong attendance, primarily by general anatomic and surgical pathologists, he says, is indicative of the increased interest in gaining knowledge in novel applications of molecular diagnostic tests as part of anatomic pathology case workup. As surgical pathologists become more involved in ordering such tests, Dr. Netto believes, they will have to become familiar not only with new IHC markers that resulted from recent genetic discoveries but also with the DNA-based molecular tests that oncologist colleagues will increasingly request.

As two examples of this trend, Dr. Netto cites the advances in managing colon and lung cancer patients. “Recent clinical trials not only supported the role of targeted tyrosine kinase inhibitors therapy for these prevalent tumors, but also demonstrated the great utility of molecular testing in predicting response to anti-EGFR monoclonal antibodies and EGFR inhibitors. Mutational analyses of somatic EGFR and KRAS genes in pulmonary adenocarcinoma and KRAS oncogene mutational analysis in advanced colorectal adenocarcinoma are now the standard of care.” These anti-EGFR Rx companion molecular tests predict response to expensive targeted therapy. “Our hope as molecular diagnosticians is to have similar theranostic strategies in most, if not all, solid organ tumors in the not-so-distant future,” Dr. Netto says. “We will have tests to help clinicians determine who is likely to respond and who is not, in addition to determining how aggressive a given tumor is.”

Also talked about in the ASCP session is the need to adopt uniform, standardized algorithms for performing companion molecular tests for new targeted therapy agents. “We also discussed how some of these upcoming genetic markers could be translated into easier-to-perform IHC tests,” Dr. Netto says, “which will help address issues of turnaround time, cost, and widespread availability.”

Discussing breast cancer, Rohit Bhargava, MD, assistant professor of pathology and co-director of surgical pathology at Magee-Womens Hospital, noted that it is heterogeneous at the morphologic, IHC, and molecular levels. Histologically, breast cancer has been classified into ductal and lobular. However, Dr. Bhargava says, “Knowing about the biomarker status of a tumor, especially its ER and HER2 status, may be more important than thinking in terms of ductal and lobular.” He finds that oncologists increasingly want to know biomarker information: “Morphologic classification is still important, but from a treatment perspective, biomarker status is more important.”

During about the past 10 years, researchers have used gene microarrays in an attempt to devise classifications of breast cancer with improved prognostic significance relative to traditional morphology-based classification. Based on genomic clustering, this work has recognized at least four subclasses of breast cancer: luminal A, luminal B, basal-like, and ERBB2. The luminal tumors form the ER-positive subgroup, and the basal-like and ERBB2 tumors constitute the ER-negative subgroup. Retrospective comparison showed differences in survival among the groups (Sorlie T, et al. Proc Natl Acad Sci USA. 2001;98:10869–10874).

Several studies have attempted to extrapolate or compare molecular findings to immunohistochemical criteria, initially focusing on basal-like breast carcinomas. Dr. Bhargava said that, by IHC, basal-like breast tumors lack ER and PR expression and are also negative for HER2 (“triple-negative”). They are called basal-like carcinomas because they express basal-phenotype markers, such as high-molecular-weight cytokeratins (CK5, CK5/6, CK14, CK17) and EGFR.

Basal-like tumors were compared with triple-negative tumors to see whether the additional elements in the gene expression profile improved prognosis relative to the three commonly measured IHC biomarkers. Outcomes were mixed. One group found that basal-like tumors were not more strongly associated with patient survival (Jumppanen M, et al. Breast Cancer Res. 2007;9:R16). A second study showed that basal-like breast cancer defined by five biomarkers (ER, PR, HER2, EGFR, CK5/6) had superior prognostic value versus triple-negative tumors (Cheang MC, et al. Clin Cancer Res. 2008;14:1368–1376). However, the difference was modest: 10-year cancer-specific survival of 67 percent versus 62 percent.

The first attempt to use IHC to define all molecular subtypes was the Carolina Breast Cancer Study (Carey LA, et al. JAMA. 2006;295:2492–2502). In a more limited way, attempts have been made with IHC to reproduce the luminal A and B division, since these two classes have different prognoses. “Luminal A tumors show the best prognosis,” Dr. Bhargava says. They are well differentiated and strongly ER- and PR-positive. Luminal B tumors have a worse prognosis than luminal A and are poorly differentiated and supposedly have low to moderate expression for hormone receptors. Dr. Bhargava and his colleagues used semiquantitative IHC, taking into account both percentage and intensity of staining to segregate ER-positive/HER2-negative tumors into two groups, attempting to replicate luminal A and B. They concluded: “Immunohistochemistry is a reliable surrogate tool to classify breast carcinoma according to the gene expression profile classification” (Bhargava R, et al. Int J Clin Exp Pathol. 2009;2:444–455).

Two other approaches have been taken to divide luminal tumors into A and B. One used IHC for two additional genes, FOXA1 and GATA3, in ER-positive/HER2-negative tumors (Badve S, Nakshatri H. J Clin Pathol. 2009;62:6–12) and the other used quantitative Ki-67 staining (Cheang MC, et al. J Natl Cancer Inst. 2009;101:736– 750). These latter investigators proposed that ER-positive/HER2-negative tumors with a Ki-67 labeling index of 14 percent or higher be classified as luminal B and the remainder as luminal A. Dr. Bhargava says, “It remains to be established in further studies” whether these additional markers can neatly separate the two luminal subtypes. Because the two classes were devised on the basis of gene expression profiling, it may not be possible to reproduce them exactly with IHC biomarkers. Certainly no single additional biomarker can accomplish this. “There will always be tumors that will bridge luminal A and luminal B,” Dr. Bhargava says.

Moving to theranostics, Dr. Bhargava notes that pathologic complete response to neoadjuvant therapy is seen mostly in basal-like and ERBB2 molecular classes, and not often in luminal tumors. He and his colleagues showed that semiquantitative IHC for ER, PR, and HER2 can provide similar prediction (similar to gene expression profiling): Pathologic complete response was identified in 33 percent of ERBB2 tumors, 30 percent of triple-negative tumors, and eight percent of weak ER-positive/HER2-positive tumors. This work is now in press. “We can apply semiquantitative IHC as a surrogate for gene expression profiling,” Dr. Bhargava concludes.

The evidence shows, in his view, that gene expression profiles can be translated directly into IHC and that “IHC is here to stay” and will continue to be useful. To Dr. Bhargava, this is not surprising. “Many profiling studies are heavily influenced by the ER status of tumors,” he says. “Since we measure that and other biomarkers, we get a great amount of information. Even just ER, PR, and HER2 values provided an extraordinary amount of information as to which tumors were going to have pathologic complete response.” In his practice, Dr. Bhargava currently measures these three biomarkers, as well as the recently added Ki-67, which is especially helpful in ER-positive tumors. As an indicator of how proliferative a tumor is, Ki-67 has a much broader dynamic range than counting mitotic figures (number of mitoses per high-power field), Dr. Bhargava points out. He does not report molecular classes. “We want to see how acceptable this classification becomes,” he says. “All oncologists and surgeons at our institutions are well aware which tumors have a high likelihood of response based on ER, PR, and HER2 status. They take into account clinical parameters plus markers to make patient decisions.”

At the same time, Dr. Bhargava says, “We should study and validate new­er biomarkers, not just stick with the ones we currently have. We know that, among triple-negative breast cancers (as identified by immunohistochemistry), 30 percent to 40 percent show complete pathological response, which is similar to using gene expression profiling. Basal-like tumors identified by expression profiling have that degree of response.” However, it is impossible to determine which individual tumors will respond, and additional biomarkers might help in this regard.

Gastrointestinal cancers are second in seniority to breast cancer in terms of the contribution of biomarkers determined by IHC and, more recently, by molecular methods. Alyssa M. Krasinskas, MD, director of the GI Pathology Center of Excellence and associate professor of pathology at UPMC, discussed two types of GI tumors in which pathologists can provide diagnostic, prognostic, and predictive functions: colorectal cancer (CRC) and gastrointestinal stromal tumors (GIST). “CRC and GIST are two examples where we, as pathologists, can use our skills to help clinicians provide personalized medicine to their patients,” she told symposium participants.

With regard to CRC, Dr. Krasinskas told CAP TODAY that pathologists should not view this condition as only one disease. She has written, “Wasn’t life easy for the surgical pathologist when CRC was believed to be one disease? All we needed to do was make a diagnosis of ‘adenocarcinoma’ on a colonoscopic biopsy or properly stage a resected tumor.” There are currently at least two major divisions of CRC. Most CRCs arise by the chromosomal instability pathway and most are sporadic. (About one percent are inherited.) About 15 percent of CRCs arise via the microsatellite instability (MSI) pathway. These cancers, too, are largely sporadic, and most result from hypermethylation of the promoter region of the mismatch repair gene MLH1.

However, two percent to five percent of tumors that arise via the MSI pathway are inherited, resulting from germline mutations in one of the mismatch repair genes MLH1, MSH2, MSH6, or PMS2. Whether due to sporadic or inherited mutations, a deficiency in DNA repair results in various sizes of microsatellites—short repeating DNA segments present throughout the genome—or microsatellite instability.

It is this small fraction of CRC cases, those due to inherited mutations in one of the mismatch repair genes, that makes it important to determine the pathway of colorectal carcinogenesis. Patients with these mutations have Lynch syndrome (also known as hereditary nonpolyposis colorectal cancer, or HNPCC). Patients with Lynch syndrome are at high risk for additional primary cancers; their relatives are also at risk for developing certain cancers and should be offered genetic testing. Another reason for identifying CRC cases due to a high level of microsatellite instability (MSI-H) is that these individuals may have a better prognosis and may respond less well to 5-FU. However, according to Dr. Krasinskas, “Clinicians would like to see more data before they adopt MSI testing for this purpose.”

Several schemes have been developed to guide screening for Lynch syndrome, such as the Amsterdam criteria or the Bethesda guidelines. Dr. Krasinskas pointed out one weakness of all the schemes: They use 50 years of age as a cutoff, “which is a rather arbitrary age,” she says. A recent study found that about half of Lynch syndrome patients were over age 50 (Hampel H, et al. J Clin Oncol. 2008;26:5783–5788). Based on this study and other similar studies, one working group has suggested offering genetic testing for Lynch syndrome to all newly diagnosed CRC patients (Berg AO. Genet Med. 2009;11:35–41).

Lynch syndrome can be detected by using IHC to look for loss of one of the mismatch repair proteins or PCR to detect MSI using a panel of probes. Sequencing of the suspected gene confirms the diagnosis. When using IHC, it is possible to test for only two proteins, MSH6 and PMS2, because of interactions among the four mismatch repair proteins. If a laboratory decides to use two stains, Dr. Krasinskas emphasizes the need to confirm the results by either adding the other antibodies or by applying molecular testing.

In the algorithm used at the University of Pittsburgh Medical Center, all resected CRCs are tested for microsatellite instability by PCR, which was chosen because its reproducibility approaches 100 percent and it avoids the pitfalls of IHC, such as variable staining, Dr. Krasinskas explains. On the other hand, PCR doesn’t reveal which protein is missing. “Since not all MSI-H tumors are inherited, we need to do additional testing,” Dr. Krasinskas says. The next step in the algorithm is to test for BRAF mutations; CRCs with BRAF mutations are sporadic. For MSI-H tumors that are BRAF-negative, the clinician is contacted and IHC is done, if requested. “Usually the clinicians want IHC,” Dr. Krasinskas says. If loss of MLH1 is seen, one could look for hypermethylation of the MLH1 promoter. When Lynch syndrome is suspected, the clinicians often seek genetic counseling for their patients.

A second molecular aspect of CRC is what Dr. Krasinskas calls “the EGFR-KRAS story.” And, she adds, “It really is a story.” When the anti-EGFR monoclonal antibody cetuximab (Erbitux) was approved for treating CRC in 2004, it was limited to tumors that were positive for EGFR by IHC. However, further work failed to support the correlation between EGFR status and response to cetuximab, and IHC for this receptor was “quickly abandoned,” Dr. Krasinskas says. Replacing this is a molecular test for a mutation in the KRAS gene. When KRAS carries an activating mutation, the Ras protein is altered and the EGFR pathway is constitutively activated. Currently, an ASCO Provisional Clinical Opinion recommends that all candidates for anti-EGFR antibody therapy have their tumor tested for KRAS mutations in a CLIA-accredited laboratory. About 40 percent of CRC patients have a KRAS mutation. “We perform KRAS testing by PCR on all metastatic [M1] CRC disease and at the request of clinicians,” Dr. Krasinskas says. “We test for BRAF [in KRAS-negative disease] at the clinician’s request. Since they are requesting it more often, we may add BRAF testing to our algorithm.” The BRAF V600E mutation is mutually exclusive with KRAS mutations. All of these mutations confer resistance to anti-EGFR therapy.

“Colorectal cancer is one gastrointestinal neoplasm that has brought IHC and molecular diagnostics together into the daily practice of surgical pathology,” Dr. Krasinskas says.

In GIST, which is the most common mesenchymal tumor of the GI tract, IHC is an important part of diagnosis: Staining for KIT (CD117) is positive in 95 percent of GIST. A newer IHC marker, DOG1, is as sensitive and appears to be more specific. Molecular analysis can also make the diagnosis by showing an activating mutation in the genes for KIT or platelet-derived growth factor receptor A (PDGFRA). However, mutational analysis is negative in 10 percent to 15 percent of GIST. Dr. Krasinskas uses IHC to look for KIT and DOG1. “DOG1 is the new kid on the block,” she says. If KIT is strongly positive, she omits DOG1. But if KIT staining is equivalent, she may do DOG1, depending on morphology. “You can often suspect it’s a GIST on the H&E stain,” she says, “and IHC is just confirming your suspicion. In other cases we depend more on staining.” If a suspicious tumor is negative for both proteins, it may be sent for mutational analysis. Depending on the clinical situation, a patient with a KIT- or DOG1-positive GIST on IHC may be either followed or treated with the tyrosine kinase inhibitor imatinib (Gleevec).

Over the past few years, studies have shown that KIT status in GIST has both prognostic and predictive value. Patients with KIT-negative untreated GISTs by IHC have inferior overall survival compared with those with KIT-positive untreated disease (Lee HE, et al. J Clin Pathol. 2008;61:722–729). The same pattern of survival is seen in treated GISTs (Heinrich MC, et al. J Clin Oncol. 2008;26: 5360–5367).

In the theranostic context, Dr. Krasinskas says, “The genotype of the tumor can predict treatment outcome.” Patients with GISTs bearing the most common KIT exon 11 mutations have the highest response rates (72 percent versus 44 percent for exon 9-carrying GIST or wild type) to the standard 400-mg dose of imatinib, while patients with GISTs that harbor exon 9 mutations benefit from higher doses of imatinib—response rates of 67 percent versus 17 percent with 800 mg/day versus 400 mg/day, respectively (Heinrich MC, et al. J Clin Oncol. 2008;26:5360–5367). This same study also showed that patients treated with imatinib have increased time to progression and overall survival if the tumors harbor an exon 11 mutation compared with those that harbor exon 9 or no (wild-type) mutations. Of note, GISTs with PDGFRA mutations do not respond to TKI. As a result of these and other data, National Comprehensive Cancer Network guidelines support “routine mutational analysis for all newly diagnosed intermediate- and high-risk patients, as well as for overtly malignant GISTs.”

In no cancer are advances more urgently needed than in lung cancer, whose five-year survival has remained stuck at 13 to 15 percent for three decades. Fortunately, therapy, diagnosis, and theranostics are proceeding hand-in-hand-in-hand. Sanja Dacic, MD, PhD, associate professor and director of the FISH and aCGH laboratory in the Department of Pathology at UPMC, says there has been an “explosion” of directed therapies for lung cancer in the past few years, most importantly EGFR-TKI. “EGFR-TKI may produce a short-term survival advantage, with a remarkable radiographic response and improvements in symptoms,” she says. “The main issue with anti-EGFR targeted therapies, unfortunately, is that almost all patients develop resistance within six months.” So there are still questions about their long-term value. Perhaps they will make a bigger impact in combination with agents that attack different targets.

At the symposium, Dr. Dacic first discussed platinum-based therapies, which are both toxic and variable in response. Expression of a gene called ERCC1 (excision repair cross-complementation group 1) has been suggested to predict sensitivity to platinum-based therapy (Olaussen KA, et al. N Engl J Med. 2006;355:983– 991). “This is a controversial area,” Dr. Dacic says. “Published studies used different assessment methods of ERCC1, and it is currently unclear what should be the method of choice in a clinical practice.” Dr. Dacic’s laboratory has implemented the IHC method used in the International Adjuvant Lung Cancer Trial and does it by clinician request. Commercial laboratories offer IHC and PCR-based tests as well.

Monoclonal antibodies like cetuximab and bevacizumab (Avastin) that target EGFR and vascular endothelial growth factor (VEGF), respectively, as well as EGFR-TKI such as gefitinib and erlotinib, offer hope for lung cancer patients. Use of these therapies affects pathology practice in several ways. First, Dr. Dacic says, “The diagnostic approach to lung cancer has changed in the last five years. Now we must subclassify non-small cell lung carcinomas.” Traditionally, lung cancer has been divided into small cell versus non-small cell (NSCLC). It is certain that histologic type of lung carcinomas is predictive of the safety or efficacy, or both, of these novel therapies. There is, therefore, a rationale to distinguish NSCLC subtypes.

Making this distinction can require additional tools. “We pathologists do a good job subclassifying well- or moderately differentiated lung carcinoma by H&E,” Dr. Dacic says. “With poorly differentiated tumors, about 30 percent will require use of additional stains.” Histochemical stains such as mucicarmine and PASD can help in some cases. A small IHC panel, which can include TTF-1, surfactant apoprotein, p63, and CK 5/6, may be extremely helpful in such cases. Immunoexpression of TTF-1 and surfactant apoproteins would favor adenocarcinoma, whereas expression of p63 and CK 5/6 favor squamous cell carcinoma.

Several predictors of response to targeted therapies are available, of which the most important are KRAS and EGFR mutations. KRAS mutations occur most often in smokers and are negative predictors of response to EGFR-TKI therapy; about 30 percent of lung adenocarcinomas have a KRAS mutation. Mutations in the EGFR gene are found in about 10 percent of lung adeno­carcinomas, and these mutations occur mostly in women and never smokers. Tumors bearing the newly described EML4-ALK fusion mutation are more frequent in adenocarcinomas, never smokers, and wild-type EGFR and KRAS, and are highly sensitive to novel agents that target this mutation (Shaw AT, et al. J Clin Oncol. 2009;27:4247–4253). Laboratories that intend to implement testing for these predictors of tumor response should be aware there is no standardization of methods or specimen type.

“We routinely test all non-squamous cell lung cancers for KRAS and EGFR mutations, and for ALK rearrangements by FISH,” Dr. Dacic says. “Oncologists and pathologists practicing in community hospitals are more frequently asking for this type of testing as well.”

The tumor is evaluated for mutation in the EGFR gene, which strongly predicts response to EGFR-TKI such as erlotinib and gefitinib. Based on the results of mutational analysis, patients can be triaged for adequate therapies. Recent development of EGFR mutation-specific antibodies may offer an exciting new avenue of screening for EGFR mutations in clinical practice, Dr. Dacic says.

Methods to qualify patients for anti-EGFR antibodies are still uncertain, but initial clinical trials suggest possible use of EGFR-FISH and possibly IHC. Dr. Dacic notes that EGFR IHC results depend highly on the antibody type, procedure protocols, and interpretation criteria—none of which are standardized.

Pathologists should use molecular and immunohistochemical methods in assessing lung cancer specimens to improve and properly direct the management of lung cancer patients, Dr. Dacic concluded, advice that applies to breast cancer and colorectal cancer as well. It is a simple-sounding piece of advice, but putting it into practice while biomarkers, detection methods, and targeted therapies proliferate will require that pathologists keep up with the latest advances, make difficult decisions, and adopt new technologies as needed.